1 /*******************************************************************************
3 Intel 10 Gigabit PCI Express Linux driver
4 Copyright(c) 1999 - 2015 Intel Corporation.
6 This program is free software; you can redistribute it and/or modify it
7 under the terms and conditions of the GNU General Public License,
8 version 2, as published by the Free Software Foundation.
10 This program is distributed in the hope it will be useful, but WITHOUT
11 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 You should have received a copy of the GNU General Public License along with
16 this program; if not, write to the Free Software Foundation, Inc.,
17 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
19 The full GNU General Public License is included in this distribution in
20 the file called "COPYING".
23 Linux NICS <linux.nics@intel.com>
24 e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25 Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
27 *******************************************************************************/
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
36 #include <linux/interrupt.h>
38 #include <linux/tcp.h>
39 #include <linux/sctp.h>
40 #include <linux/pkt_sched.h>
41 #include <linux/ipv6.h>
42 #include <linux/slab.h>
43 #include <net/checksum.h>
44 #include <net/ip6_checksum.h>
45 #include <linux/etherdevice.h>
46 #include <linux/ethtool.h>
48 #include <linux/if_vlan.h>
49 #include <linux/if_macvlan.h>
50 #include <linux/if_bridge.h>
51 #include <linux/prefetch.h>
52 #include <scsi/fc/fc_fcoe.h>
53 #include <net/vxlan.h>
56 #include <linux/of_net.h>
60 #include <asm/idprom.h>
65 #include "ixgbe_common.h"
66 #include "ixgbe_dcb_82599.h"
67 #include "ixgbe_sriov.h"
69 char ixgbe_driver_name
[] = "ixgbe";
70 static const char ixgbe_driver_string
[] =
71 "Intel(R) 10 Gigabit PCI Express Network Driver";
73 char ixgbe_default_device_descr
[] =
74 "Intel(R) 10 Gigabit Network Connection";
76 static char ixgbe_default_device_descr
[] =
77 "Intel(R) 10 Gigabit Network Connection";
79 #define DRV_VERSION "4.2.1-k"
80 const char ixgbe_driver_version
[] = DRV_VERSION
;
81 static const char ixgbe_copyright
[] =
82 "Copyright (c) 1999-2015 Intel Corporation.";
84 static const char ixgbe_overheat_msg
[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
86 static const struct ixgbe_info
*ixgbe_info_tbl
[] = {
87 [board_82598
] = &ixgbe_82598_info
,
88 [board_82599
] = &ixgbe_82599_info
,
89 [board_X540
] = &ixgbe_X540_info
,
90 [board_X550
] = &ixgbe_X550_info
,
91 [board_X550EM_x
] = &ixgbe_X550EM_x_info
,
94 /* ixgbe_pci_tbl - PCI Device ID Table
96 * Wildcard entries (PCI_ANY_ID) should come last
97 * Last entry must be all 0s
99 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
100 * Class, Class Mask, private data (not used) }
102 static const struct pci_device_id ixgbe_pci_tbl
[] = {
103 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598
), board_82598
},
104 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_DUAL_PORT
), board_82598
},
105 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AF_SINGLE_PORT
), board_82598
},
106 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT
), board_82598
},
107 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598AT2
), board_82598
},
108 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_CX4
), board_82598
},
109 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_CX4_DUAL_PORT
), board_82598
},
110 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_DA_DUAL_PORT
), board_82598
},
111 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM
), board_82598
},
112 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_XF_LR
), board_82598
},
113 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598EB_SFP_LOM
), board_82598
},
114 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82598_BX
), board_82598
},
115 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4
), board_82599
},
116 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_XAUI_LOM
), board_82599
},
117 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KR
), board_82599
},
118 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP
), board_82599
},
119 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_EM
), board_82599
},
120 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_KX4_MEZZ
), board_82599
},
121 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_CX4
), board_82599
},
122 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_BACKPLANE_FCOE
), board_82599
},
123 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_FCOE
), board_82599
},
124 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_T3_LOM
), board_82599
},
125 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_COMBO_BACKPLANE
), board_82599
},
126 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T
), board_X540
},
127 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF2
), board_82599
},
128 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_LS
), board_82599
},
129 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_QSFP_SF_QP
), board_82599
},
130 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599EN_SFP
), board_82599
},
131 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_82599_SFP_SF_QP
), board_82599
},
132 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X540T1
), board_X540
},
133 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550T
), board_X550
},
134 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550EM_X_KX4
), board_X550EM_x
},
135 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550EM_X_KR
), board_X550EM_x
},
136 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550EM_X_10G_T
), board_X550EM_x
},
137 {PCI_VDEVICE(INTEL
, IXGBE_DEV_ID_X550EM_X_SFP
), board_X550EM_x
},
138 /* required last entry */
141 MODULE_DEVICE_TABLE(pci
, ixgbe_pci_tbl
);
143 #ifdef CONFIG_IXGBE_DCA
144 static int ixgbe_notify_dca(struct notifier_block
*, unsigned long event
,
146 static struct notifier_block dca_notifier
= {
147 .notifier_call
= ixgbe_notify_dca
,
153 #ifdef CONFIG_PCI_IOV
154 static unsigned int max_vfs
;
155 module_param(max_vfs
, uint
, 0);
156 MODULE_PARM_DESC(max_vfs
,
157 "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
158 #endif /* CONFIG_PCI_IOV */
160 static unsigned int allow_unsupported_sfp
;
161 module_param(allow_unsupported_sfp
, uint
, 0);
162 MODULE_PARM_DESC(allow_unsupported_sfp
,
163 "Allow unsupported and untested SFP+ modules on 82599-based adapters");
165 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
166 static int debug
= -1;
167 module_param(debug
, int, 0);
168 MODULE_PARM_DESC(debug
, "Debug level (0=none,...,16=all)");
170 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
171 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
172 MODULE_LICENSE("GPL");
173 MODULE_VERSION(DRV_VERSION
);
175 static bool ixgbe_check_cfg_remove(struct ixgbe_hw
*hw
, struct pci_dev
*pdev
);
177 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter
*adapter
,
180 struct pci_dev
*parent_dev
;
181 struct pci_bus
*parent_bus
;
183 parent_bus
= adapter
->pdev
->bus
->parent
;
187 parent_dev
= parent_bus
->self
;
191 if (!pci_is_pcie(parent_dev
))
194 pcie_capability_read_word(parent_dev
, reg
, value
);
195 if (*value
== IXGBE_FAILED_READ_CFG_WORD
&&
196 ixgbe_check_cfg_remove(&adapter
->hw
, parent_dev
))
201 static s32
ixgbe_get_parent_bus_info(struct ixgbe_adapter
*adapter
)
203 struct ixgbe_hw
*hw
= &adapter
->hw
;
207 hw
->bus
.type
= ixgbe_bus_type_pci_express
;
209 /* Get the negotiated link width and speed from PCI config space of the
210 * parent, as this device is behind a switch
212 err
= ixgbe_read_pci_cfg_word_parent(adapter
, 18, &link_status
);
214 /* assume caller will handle error case */
218 hw
->bus
.width
= ixgbe_convert_bus_width(link_status
);
219 hw
->bus
.speed
= ixgbe_convert_bus_speed(link_status
);
225 * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
226 * @hw: hw specific details
228 * This function is used by probe to determine whether a device's PCI-Express
229 * bandwidth details should be gathered from the parent bus instead of from the
230 * device. Used to ensure that various locations all have the correct device ID
233 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw
*hw
)
235 switch (hw
->device_id
) {
236 case IXGBE_DEV_ID_82599_SFP_SF_QP
:
237 case IXGBE_DEV_ID_82599_QSFP_SF_QP
:
244 static void ixgbe_check_minimum_link(struct ixgbe_adapter
*adapter
,
247 struct ixgbe_hw
*hw
= &adapter
->hw
;
249 enum pci_bus_speed speed
= PCI_SPEED_UNKNOWN
;
250 enum pcie_link_width width
= PCIE_LNK_WIDTH_UNKNOWN
;
251 struct pci_dev
*pdev
;
253 /* Some devices are not connected over PCIe and thus do not negotiate
254 * speed. These devices do not have valid bus info, and thus any report
255 * we generate may not be correct.
257 if (hw
->bus
.type
== ixgbe_bus_type_internal
)
260 /* determine whether to use the parent device */
261 if (ixgbe_pcie_from_parent(&adapter
->hw
))
262 pdev
= adapter
->pdev
->bus
->parent
->self
;
264 pdev
= adapter
->pdev
;
266 if (pcie_get_minimum_link(pdev
, &speed
, &width
) ||
267 speed
== PCI_SPEED_UNKNOWN
|| width
== PCIE_LNK_WIDTH_UNKNOWN
) {
268 e_dev_warn("Unable to determine PCI Express bandwidth.\n");
273 case PCIE_SPEED_2_5GT
:
274 /* 8b/10b encoding reduces max throughput by 20% */
277 case PCIE_SPEED_5_0GT
:
278 /* 8b/10b encoding reduces max throughput by 20% */
281 case PCIE_SPEED_8_0GT
:
282 /* 128b/130b encoding reduces throughput by less than 2% */
286 e_dev_warn("Unable to determine PCI Express bandwidth.\n");
290 e_dev_info("PCI Express bandwidth of %dGT/s available\n",
292 e_dev_info("(Speed:%s, Width: x%d, Encoding Loss:%s)\n",
293 (speed
== PCIE_SPEED_8_0GT
? "8.0GT/s" :
294 speed
== PCIE_SPEED_5_0GT
? "5.0GT/s" :
295 speed
== PCIE_SPEED_2_5GT
? "2.5GT/s" :
298 (speed
== PCIE_SPEED_2_5GT
? "20%" :
299 speed
== PCIE_SPEED_5_0GT
? "20%" :
300 speed
== PCIE_SPEED_8_0GT
? "<2%" :
303 if (max_gts
< expected_gts
) {
304 e_dev_warn("This is not sufficient for optimal performance of this card.\n");
305 e_dev_warn("For optimal performance, at least %dGT/s of bandwidth is required.\n",
307 e_dev_warn("A slot with more lanes and/or higher speed is suggested.\n");
311 static void ixgbe_service_event_schedule(struct ixgbe_adapter
*adapter
)
313 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
) &&
314 !test_bit(__IXGBE_REMOVING
, &adapter
->state
) &&
315 !test_and_set_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
))
316 schedule_work(&adapter
->service_task
);
319 static void ixgbe_remove_adapter(struct ixgbe_hw
*hw
)
321 struct ixgbe_adapter
*adapter
= hw
->back
;
326 e_dev_err("Adapter removed\n");
327 if (test_bit(__IXGBE_SERVICE_INITED
, &adapter
->state
))
328 ixgbe_service_event_schedule(adapter
);
331 static void ixgbe_check_remove(struct ixgbe_hw
*hw
, u32 reg
)
335 /* The following check not only optimizes a bit by not
336 * performing a read on the status register when the
337 * register just read was a status register read that
338 * returned IXGBE_FAILED_READ_REG. It also blocks any
339 * potential recursion.
341 if (reg
== IXGBE_STATUS
) {
342 ixgbe_remove_adapter(hw
);
345 value
= ixgbe_read_reg(hw
, IXGBE_STATUS
);
346 if (value
== IXGBE_FAILED_READ_REG
)
347 ixgbe_remove_adapter(hw
);
351 * ixgbe_read_reg - Read from device register
352 * @hw: hw specific details
353 * @reg: offset of register to read
355 * Returns : value read or IXGBE_FAILED_READ_REG if removed
357 * This function is used to read device registers. It checks for device
358 * removal by confirming any read that returns all ones by checking the
359 * status register value for all ones. This function avoids reading from
360 * the hardware if a removal was previously detected in which case it
361 * returns IXGBE_FAILED_READ_REG (all ones).
363 u32
ixgbe_read_reg(struct ixgbe_hw
*hw
, u32 reg
)
365 u8 __iomem
*reg_addr
= ACCESS_ONCE(hw
->hw_addr
);
368 if (ixgbe_removed(reg_addr
))
369 return IXGBE_FAILED_READ_REG
;
370 value
= readl(reg_addr
+ reg
);
371 if (unlikely(value
== IXGBE_FAILED_READ_REG
))
372 ixgbe_check_remove(hw
, reg
);
376 static bool ixgbe_check_cfg_remove(struct ixgbe_hw
*hw
, struct pci_dev
*pdev
)
380 pci_read_config_word(pdev
, PCI_VENDOR_ID
, &value
);
381 if (value
== IXGBE_FAILED_READ_CFG_WORD
) {
382 ixgbe_remove_adapter(hw
);
388 u16
ixgbe_read_pci_cfg_word(struct ixgbe_hw
*hw
, u32 reg
)
390 struct ixgbe_adapter
*adapter
= hw
->back
;
393 if (ixgbe_removed(hw
->hw_addr
))
394 return IXGBE_FAILED_READ_CFG_WORD
;
395 pci_read_config_word(adapter
->pdev
, reg
, &value
);
396 if (value
== IXGBE_FAILED_READ_CFG_WORD
&&
397 ixgbe_check_cfg_remove(hw
, adapter
->pdev
))
398 return IXGBE_FAILED_READ_CFG_WORD
;
402 #ifdef CONFIG_PCI_IOV
403 static u32
ixgbe_read_pci_cfg_dword(struct ixgbe_hw
*hw
, u32 reg
)
405 struct ixgbe_adapter
*adapter
= hw
->back
;
408 if (ixgbe_removed(hw
->hw_addr
))
409 return IXGBE_FAILED_READ_CFG_DWORD
;
410 pci_read_config_dword(adapter
->pdev
, reg
, &value
);
411 if (value
== IXGBE_FAILED_READ_CFG_DWORD
&&
412 ixgbe_check_cfg_remove(hw
, adapter
->pdev
))
413 return IXGBE_FAILED_READ_CFG_DWORD
;
416 #endif /* CONFIG_PCI_IOV */
418 void ixgbe_write_pci_cfg_word(struct ixgbe_hw
*hw
, u32 reg
, u16 value
)
420 struct ixgbe_adapter
*adapter
= hw
->back
;
422 if (ixgbe_removed(hw
->hw_addr
))
424 pci_write_config_word(adapter
->pdev
, reg
, value
);
427 static void ixgbe_service_event_complete(struct ixgbe_adapter
*adapter
)
429 BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
));
431 /* flush memory to make sure state is correct before next watchdog */
432 smp_mb__before_atomic();
433 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
436 struct ixgbe_reg_info
{
441 static const struct ixgbe_reg_info ixgbe_reg_info_tbl
[] = {
443 /* General Registers */
444 {IXGBE_CTRL
, "CTRL"},
445 {IXGBE_STATUS
, "STATUS"},
446 {IXGBE_CTRL_EXT
, "CTRL_EXT"},
448 /* Interrupt Registers */
449 {IXGBE_EICR
, "EICR"},
452 {IXGBE_SRRCTL(0), "SRRCTL"},
453 {IXGBE_DCA_RXCTRL(0), "DRXCTL"},
454 {IXGBE_RDLEN(0), "RDLEN"},
455 {IXGBE_RDH(0), "RDH"},
456 {IXGBE_RDT(0), "RDT"},
457 {IXGBE_RXDCTL(0), "RXDCTL"},
458 {IXGBE_RDBAL(0), "RDBAL"},
459 {IXGBE_RDBAH(0), "RDBAH"},
462 {IXGBE_TDBAL(0), "TDBAL"},
463 {IXGBE_TDBAH(0), "TDBAH"},
464 {IXGBE_TDLEN(0), "TDLEN"},
465 {IXGBE_TDH(0), "TDH"},
466 {IXGBE_TDT(0), "TDT"},
467 {IXGBE_TXDCTL(0), "TXDCTL"},
469 /* List Terminator */
475 * ixgbe_regdump - register printout routine
477 static void ixgbe_regdump(struct ixgbe_hw
*hw
, struct ixgbe_reg_info
*reginfo
)
483 switch (reginfo
->ofs
) {
484 case IXGBE_SRRCTL(0):
485 for (i
= 0; i
< 64; i
++)
486 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_SRRCTL(i
));
488 case IXGBE_DCA_RXCTRL(0):
489 for (i
= 0; i
< 64; i
++)
490 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_DCA_RXCTRL(i
));
493 for (i
= 0; i
< 64; i
++)
494 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDLEN(i
));
497 for (i
= 0; i
< 64; i
++)
498 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDH(i
));
501 for (i
= 0; i
< 64; i
++)
502 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDT(i
));
504 case IXGBE_RXDCTL(0):
505 for (i
= 0; i
< 64; i
++)
506 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RXDCTL(i
));
509 for (i
= 0; i
< 64; i
++)
510 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAL(i
));
513 for (i
= 0; i
< 64; i
++)
514 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_RDBAH(i
));
517 for (i
= 0; i
< 64; i
++)
518 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAL(i
));
521 for (i
= 0; i
< 64; i
++)
522 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDBAH(i
));
525 for (i
= 0; i
< 64; i
++)
526 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDLEN(i
));
529 for (i
= 0; i
< 64; i
++)
530 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDH(i
));
533 for (i
= 0; i
< 64; i
++)
534 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TDT(i
));
536 case IXGBE_TXDCTL(0):
537 for (i
= 0; i
< 64; i
++)
538 regs
[i
] = IXGBE_READ_REG(hw
, IXGBE_TXDCTL(i
));
541 pr_info("%-15s %08x\n", reginfo
->name
,
542 IXGBE_READ_REG(hw
, reginfo
->ofs
));
546 for (i
= 0; i
< 8; i
++) {
547 snprintf(rname
, 16, "%s[%d-%d]", reginfo
->name
, i
*8, i
*8+7);
548 pr_err("%-15s", rname
);
549 for (j
= 0; j
< 8; j
++)
550 pr_cont(" %08x", regs
[i
*8+j
]);
557 * ixgbe_dump - Print registers, tx-rings and rx-rings
559 static void ixgbe_dump(struct ixgbe_adapter
*adapter
)
561 struct net_device
*netdev
= adapter
->netdev
;
562 struct ixgbe_hw
*hw
= &adapter
->hw
;
563 struct ixgbe_reg_info
*reginfo
;
565 struct ixgbe_ring
*tx_ring
;
566 struct ixgbe_tx_buffer
*tx_buffer
;
567 union ixgbe_adv_tx_desc
*tx_desc
;
568 struct my_u0
{ u64 a
; u64 b
; } *u0
;
569 struct ixgbe_ring
*rx_ring
;
570 union ixgbe_adv_rx_desc
*rx_desc
;
571 struct ixgbe_rx_buffer
*rx_buffer_info
;
575 if (!netif_msg_hw(adapter
))
578 /* Print netdevice Info */
580 dev_info(&adapter
->pdev
->dev
, "Net device Info\n");
581 pr_info("Device Name state "
582 "trans_start last_rx\n");
583 pr_info("%-15s %016lX %016lX %016lX\n",
590 /* Print Registers */
591 dev_info(&adapter
->pdev
->dev
, "Register Dump\n");
592 pr_info(" Register Name Value\n");
593 for (reginfo
= (struct ixgbe_reg_info
*)ixgbe_reg_info_tbl
;
594 reginfo
->name
; reginfo
++) {
595 ixgbe_regdump(hw
, reginfo
);
598 /* Print TX Ring Summary */
599 if (!netdev
|| !netif_running(netdev
))
602 dev_info(&adapter
->pdev
->dev
, "TX Rings Summary\n");
603 pr_info(" %s %s %s %s\n",
604 "Queue [NTU] [NTC] [bi(ntc)->dma ]",
605 "leng", "ntw", "timestamp");
606 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
607 tx_ring
= adapter
->tx_ring
[n
];
608 tx_buffer
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_clean
];
609 pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
610 n
, tx_ring
->next_to_use
, tx_ring
->next_to_clean
,
611 (u64
)dma_unmap_addr(tx_buffer
, dma
),
612 dma_unmap_len(tx_buffer
, len
),
613 tx_buffer
->next_to_watch
,
614 (u64
)tx_buffer
->time_stamp
);
618 if (!netif_msg_tx_done(adapter
))
619 goto rx_ring_summary
;
621 dev_info(&adapter
->pdev
->dev
, "TX Rings Dump\n");
623 /* Transmit Descriptor Formats
625 * 82598 Advanced Transmit Descriptor
626 * +--------------------------------------------------------------+
627 * 0 | Buffer Address [63:0] |
628 * +--------------------------------------------------------------+
629 * 8 | PAYLEN | POPTS | IDX | STA | DCMD |DTYP | RSV | DTALEN |
630 * +--------------------------------------------------------------+
631 * 63 46 45 40 39 36 35 32 31 24 23 20 19 0
633 * 82598 Advanced Transmit Descriptor (Write-Back Format)
634 * +--------------------------------------------------------------+
636 * +--------------------------------------------------------------+
637 * 8 | RSV | STA | NXTSEQ |
638 * +--------------------------------------------------------------+
641 * 82599+ Advanced Transmit Descriptor
642 * +--------------------------------------------------------------+
643 * 0 | Buffer Address [63:0] |
644 * +--------------------------------------------------------------+
645 * 8 |PAYLEN |POPTS|CC|IDX |STA |DCMD |DTYP |MAC |RSV |DTALEN |
646 * +--------------------------------------------------------------+
647 * 63 46 45 40 39 38 36 35 32 31 24 23 20 19 18 17 16 15 0
649 * 82599+ Advanced Transmit Descriptor (Write-Back Format)
650 * +--------------------------------------------------------------+
652 * +--------------------------------------------------------------+
653 * 8 | RSV | STA | RSV |
654 * +--------------------------------------------------------------+
658 for (n
= 0; n
< adapter
->num_tx_queues
; n
++) {
659 tx_ring
= adapter
->tx_ring
[n
];
660 pr_info("------------------------------------\n");
661 pr_info("TX QUEUE INDEX = %d\n", tx_ring
->queue_index
);
662 pr_info("------------------------------------\n");
663 pr_info("%s%s %s %s %s %s\n",
664 "T [desc] [address 63:0 ] ",
665 "[PlPOIdStDDt Ln] [bi->dma ] ",
666 "leng", "ntw", "timestamp", "bi->skb");
668 for (i
= 0; tx_ring
->desc
&& (i
< tx_ring
->count
); i
++) {
669 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
670 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
671 u0
= (struct my_u0
*)tx_desc
;
672 if (dma_unmap_len(tx_buffer
, len
) > 0) {
673 pr_info("T [0x%03X] %016llX %016llX %016llX %08X %p %016llX %p",
677 (u64
)dma_unmap_addr(tx_buffer
, dma
),
678 dma_unmap_len(tx_buffer
, len
),
679 tx_buffer
->next_to_watch
,
680 (u64
)tx_buffer
->time_stamp
,
682 if (i
== tx_ring
->next_to_use
&&
683 i
== tx_ring
->next_to_clean
)
685 else if (i
== tx_ring
->next_to_use
)
687 else if (i
== tx_ring
->next_to_clean
)
692 if (netif_msg_pktdata(adapter
) &&
694 print_hex_dump(KERN_INFO
, "",
695 DUMP_PREFIX_ADDRESS
, 16, 1,
696 tx_buffer
->skb
->data
,
697 dma_unmap_len(tx_buffer
, len
),
703 /* Print RX Rings Summary */
705 dev_info(&adapter
->pdev
->dev
, "RX Rings Summary\n");
706 pr_info("Queue [NTU] [NTC]\n");
707 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
708 rx_ring
= adapter
->rx_ring
[n
];
709 pr_info("%5d %5X %5X\n",
710 n
, rx_ring
->next_to_use
, rx_ring
->next_to_clean
);
714 if (!netif_msg_rx_status(adapter
))
717 dev_info(&adapter
->pdev
->dev
, "RX Rings Dump\n");
719 /* Receive Descriptor Formats
721 * 82598 Advanced Receive Descriptor (Read) Format
723 * +-----------------------------------------------------+
724 * 0 | Packet Buffer Address [63:1] |A0/NSE|
725 * +----------------------------------------------+------+
726 * 8 | Header Buffer Address [63:1] | DD |
727 * +-----------------------------------------------------+
730 * 82598 Advanced Receive Descriptor (Write-Back) Format
732 * 63 48 47 32 31 30 21 20 16 15 4 3 0
733 * +------------------------------------------------------+
734 * 0 | RSS Hash / |SPH| HDR_LEN | RSV |Packet| RSS |
735 * | Packet | IP | | | | Type | Type |
736 * | Checksum | Ident | | | | | |
737 * +------------------------------------------------------+
738 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
739 * +------------------------------------------------------+
740 * 63 48 47 32 31 20 19 0
742 * 82599+ Advanced Receive Descriptor (Read) Format
744 * +-----------------------------------------------------+
745 * 0 | Packet Buffer Address [63:1] |A0/NSE|
746 * +----------------------------------------------+------+
747 * 8 | Header Buffer Address [63:1] | DD |
748 * +-----------------------------------------------------+
751 * 82599+ Advanced Receive Descriptor (Write-Back) Format
753 * 63 48 47 32 31 30 21 20 17 16 4 3 0
754 * +------------------------------------------------------+
755 * 0 |RSS / Frag Checksum|SPH| HDR_LEN |RSC- |Packet| RSS |
756 * |/ RTT / PCoE_PARAM | | | CNT | Type | Type |
757 * |/ Flow Dir Flt ID | | | | | |
758 * +------------------------------------------------------+
759 * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
760 * +------------------------------------------------------+
761 * 63 48 47 32 31 20 19 0
764 for (n
= 0; n
< adapter
->num_rx_queues
; n
++) {
765 rx_ring
= adapter
->rx_ring
[n
];
766 pr_info("------------------------------------\n");
767 pr_info("RX QUEUE INDEX = %d\n", rx_ring
->queue_index
);
768 pr_info("------------------------------------\n");
770 "R [desc] [ PktBuf A0] ",
771 "[ HeadBuf DD] [bi->dma ] [bi->skb ] ",
772 "<-- Adv Rx Read format\n");
774 "RWB[desc] [PcsmIpSHl PtRs] ",
775 "[vl er S cks ln] ---------------- [bi->skb ] ",
776 "<-- Adv Rx Write-Back format\n");
778 for (i
= 0; i
< rx_ring
->count
; i
++) {
779 rx_buffer_info
= &rx_ring
->rx_buffer_info
[i
];
780 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
781 u0
= (struct my_u0
*)rx_desc
;
782 staterr
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
783 if (staterr
& IXGBE_RXD_STAT_DD
) {
784 /* Descriptor Done */
785 pr_info("RWB[0x%03X] %016llX "
786 "%016llX ---------------- %p", i
,
789 rx_buffer_info
->skb
);
791 pr_info("R [0x%03X] %016llX "
792 "%016llX %016llX %p", i
,
795 (u64
)rx_buffer_info
->dma
,
796 rx_buffer_info
->skb
);
798 if (netif_msg_pktdata(adapter
) &&
799 rx_buffer_info
->dma
) {
800 print_hex_dump(KERN_INFO
, "",
801 DUMP_PREFIX_ADDRESS
, 16, 1,
802 page_address(rx_buffer_info
->page
) +
803 rx_buffer_info
->page_offset
,
804 ixgbe_rx_bufsz(rx_ring
), true);
808 if (i
== rx_ring
->next_to_use
)
810 else if (i
== rx_ring
->next_to_clean
)
819 static void ixgbe_release_hw_control(struct ixgbe_adapter
*adapter
)
823 /* Let firmware take over control of h/w */
824 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
825 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
826 ctrl_ext
& ~IXGBE_CTRL_EXT_DRV_LOAD
);
829 static void ixgbe_get_hw_control(struct ixgbe_adapter
*adapter
)
833 /* Let firmware know the driver has taken over */
834 ctrl_ext
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_CTRL_EXT
);
835 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_CTRL_EXT
,
836 ctrl_ext
| IXGBE_CTRL_EXT_DRV_LOAD
);
840 * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
841 * @adapter: pointer to adapter struct
842 * @direction: 0 for Rx, 1 for Tx, -1 for other causes
843 * @queue: queue to map the corresponding interrupt to
844 * @msix_vector: the vector to map to the corresponding queue
847 static void ixgbe_set_ivar(struct ixgbe_adapter
*adapter
, s8 direction
,
848 u8 queue
, u8 msix_vector
)
851 struct ixgbe_hw
*hw
= &adapter
->hw
;
852 switch (hw
->mac
.type
) {
853 case ixgbe_mac_82598EB
:
854 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
857 index
= (((direction
* 64) + queue
) >> 2) & 0x1F;
858 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(index
));
859 ivar
&= ~(0xFF << (8 * (queue
& 0x3)));
860 ivar
|= (msix_vector
<< (8 * (queue
& 0x3)));
861 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(index
), ivar
);
863 case ixgbe_mac_82599EB
:
866 case ixgbe_mac_X550EM_x
:
867 if (direction
== -1) {
869 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
870 index
= ((queue
& 1) * 8);
871 ivar
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_IVAR_MISC
);
872 ivar
&= ~(0xFF << index
);
873 ivar
|= (msix_vector
<< index
);
874 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_IVAR_MISC
, ivar
);
877 /* tx or rx causes */
878 msix_vector
|= IXGBE_IVAR_ALLOC_VAL
;
879 index
= ((16 * (queue
& 1)) + (8 * direction
));
880 ivar
= IXGBE_READ_REG(hw
, IXGBE_IVAR(queue
>> 1));
881 ivar
&= ~(0xFF << index
);
882 ivar
|= (msix_vector
<< index
);
883 IXGBE_WRITE_REG(hw
, IXGBE_IVAR(queue
>> 1), ivar
);
891 static inline void ixgbe_irq_rearm_queues(struct ixgbe_adapter
*adapter
,
896 switch (adapter
->hw
.mac
.type
) {
897 case ixgbe_mac_82598EB
:
898 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
899 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS
, mask
);
901 case ixgbe_mac_82599EB
:
904 case ixgbe_mac_X550EM_x
:
905 mask
= (qmask
& 0xFFFFFFFF);
906 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(0), mask
);
907 mask
= (qmask
>> 32);
908 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EICS_EX(1), mask
);
915 void ixgbe_unmap_and_free_tx_resource(struct ixgbe_ring
*ring
,
916 struct ixgbe_tx_buffer
*tx_buffer
)
918 if (tx_buffer
->skb
) {
919 dev_kfree_skb_any(tx_buffer
->skb
);
920 if (dma_unmap_len(tx_buffer
, len
))
921 dma_unmap_single(ring
->dev
,
922 dma_unmap_addr(tx_buffer
, dma
),
923 dma_unmap_len(tx_buffer
, len
),
925 } else if (dma_unmap_len(tx_buffer
, len
)) {
926 dma_unmap_page(ring
->dev
,
927 dma_unmap_addr(tx_buffer
, dma
),
928 dma_unmap_len(tx_buffer
, len
),
931 tx_buffer
->next_to_watch
= NULL
;
932 tx_buffer
->skb
= NULL
;
933 dma_unmap_len_set(tx_buffer
, len
, 0);
934 /* tx_buffer must be completely set up in the transmit path */
937 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter
*adapter
)
939 struct ixgbe_hw
*hw
= &adapter
->hw
;
940 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
944 if ((hw
->fc
.current_mode
!= ixgbe_fc_full
) &&
945 (hw
->fc
.current_mode
!= ixgbe_fc_rx_pause
))
948 switch (hw
->mac
.type
) {
949 case ixgbe_mac_82598EB
:
950 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXC
);
953 data
= IXGBE_READ_REG(hw
, IXGBE_LXOFFRXCNT
);
955 hwstats
->lxoffrxc
+= data
;
957 /* refill credits (no tx hang) if we received xoff */
961 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
962 clear_bit(__IXGBE_HANG_CHECK_ARMED
,
963 &adapter
->tx_ring
[i
]->state
);
966 static void ixgbe_update_xoff_received(struct ixgbe_adapter
*adapter
)
968 struct ixgbe_hw
*hw
= &adapter
->hw
;
969 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
973 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
975 if (adapter
->ixgbe_ieee_pfc
)
976 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
978 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) || !pfc_en
) {
979 ixgbe_update_xoff_rx_lfc(adapter
);
983 /* update stats for each tc, only valid with PFC enabled */
984 for (i
= 0; i
< MAX_TX_PACKET_BUFFERS
; i
++) {
987 switch (hw
->mac
.type
) {
988 case ixgbe_mac_82598EB
:
989 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXC(i
));
992 pxoffrxc
= IXGBE_READ_REG(hw
, IXGBE_PXOFFRXCNT(i
));
994 hwstats
->pxoffrxc
[i
] += pxoffrxc
;
995 /* Get the TC for given UP */
996 tc
= netdev_get_prio_tc_map(adapter
->netdev
, i
);
997 xoff
[tc
] += pxoffrxc
;
1000 /* disarm tx queues that have received xoff frames */
1001 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
1002 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
1004 tc
= tx_ring
->dcb_tc
;
1006 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
1010 static u64
ixgbe_get_tx_completed(struct ixgbe_ring
*ring
)
1012 return ring
->stats
.packets
;
1015 static u64
ixgbe_get_tx_pending(struct ixgbe_ring
*ring
)
1017 struct ixgbe_adapter
*adapter
;
1018 struct ixgbe_hw
*hw
;
1021 if (ring
->l2_accel_priv
)
1022 adapter
= ring
->l2_accel_priv
->real_adapter
;
1024 adapter
= netdev_priv(ring
->netdev
);
1027 head
= IXGBE_READ_REG(hw
, IXGBE_TDH(ring
->reg_idx
));
1028 tail
= IXGBE_READ_REG(hw
, IXGBE_TDT(ring
->reg_idx
));
1031 return (head
< tail
) ?
1032 tail
- head
: (tail
+ ring
->count
- head
);
1037 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring
*tx_ring
)
1039 u32 tx_done
= ixgbe_get_tx_completed(tx_ring
);
1040 u32 tx_done_old
= tx_ring
->tx_stats
.tx_done_old
;
1041 u32 tx_pending
= ixgbe_get_tx_pending(tx_ring
);
1043 clear_check_for_tx_hang(tx_ring
);
1046 * Check for a hung queue, but be thorough. This verifies
1047 * that a transmit has been completed since the previous
1048 * check AND there is at least one packet pending. The
1049 * ARMED bit is set to indicate a potential hang. The
1050 * bit is cleared if a pause frame is received to remove
1051 * false hang detection due to PFC or 802.3x frames. By
1052 * requiring this to fail twice we avoid races with
1053 * pfc clearing the ARMED bit and conditions where we
1054 * run the check_tx_hang logic with a transmit completion
1055 * pending but without time to complete it yet.
1057 if (tx_done_old
== tx_done
&& tx_pending
)
1058 /* make sure it is true for two checks in a row */
1059 return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED
,
1061 /* update completed stats and continue */
1062 tx_ring
->tx_stats
.tx_done_old
= tx_done
;
1063 /* reset the countdown */
1064 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &tx_ring
->state
);
1070 * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1071 * @adapter: driver private struct
1073 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter
*adapter
)
1076 /* Do the reset outside of interrupt context */
1077 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
1078 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
1079 e_warn(drv
, "initiating reset due to tx timeout\n");
1080 ixgbe_service_event_schedule(adapter
);
1085 * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1086 * @q_vector: structure containing interrupt and ring information
1087 * @tx_ring: tx ring to clean
1089 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector
*q_vector
,
1090 struct ixgbe_ring
*tx_ring
)
1092 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1093 struct ixgbe_tx_buffer
*tx_buffer
;
1094 union ixgbe_adv_tx_desc
*tx_desc
;
1095 unsigned int total_bytes
= 0, total_packets
= 0;
1096 unsigned int budget
= q_vector
->tx
.work_limit
;
1097 unsigned int i
= tx_ring
->next_to_clean
;
1099 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
1102 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
1103 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
1104 i
-= tx_ring
->count
;
1107 union ixgbe_adv_tx_desc
*eop_desc
= tx_buffer
->next_to_watch
;
1109 /* if next_to_watch is not set then there is no work pending */
1113 /* prevent any other reads prior to eop_desc */
1114 read_barrier_depends();
1116 /* if DD is not set pending work has not been completed */
1117 if (!(eop_desc
->wb
.status
& cpu_to_le32(IXGBE_TXD_STAT_DD
)))
1120 /* clear next_to_watch to prevent false hangs */
1121 tx_buffer
->next_to_watch
= NULL
;
1123 /* update the statistics for this packet */
1124 total_bytes
+= tx_buffer
->bytecount
;
1125 total_packets
+= tx_buffer
->gso_segs
;
1128 dev_consume_skb_any(tx_buffer
->skb
);
1130 /* unmap skb header data */
1131 dma_unmap_single(tx_ring
->dev
,
1132 dma_unmap_addr(tx_buffer
, dma
),
1133 dma_unmap_len(tx_buffer
, len
),
1136 /* clear tx_buffer data */
1137 tx_buffer
->skb
= NULL
;
1138 dma_unmap_len_set(tx_buffer
, len
, 0);
1140 /* unmap remaining buffers */
1141 while (tx_desc
!= eop_desc
) {
1146 i
-= tx_ring
->count
;
1147 tx_buffer
= tx_ring
->tx_buffer_info
;
1148 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
1151 /* unmap any remaining paged data */
1152 if (dma_unmap_len(tx_buffer
, len
)) {
1153 dma_unmap_page(tx_ring
->dev
,
1154 dma_unmap_addr(tx_buffer
, dma
),
1155 dma_unmap_len(tx_buffer
, len
),
1157 dma_unmap_len_set(tx_buffer
, len
, 0);
1161 /* move us one more past the eop_desc for start of next pkt */
1166 i
-= tx_ring
->count
;
1167 tx_buffer
= tx_ring
->tx_buffer_info
;
1168 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
1171 /* issue prefetch for next Tx descriptor */
1174 /* update budget accounting */
1176 } while (likely(budget
));
1178 i
+= tx_ring
->count
;
1179 tx_ring
->next_to_clean
= i
;
1180 u64_stats_update_begin(&tx_ring
->syncp
);
1181 tx_ring
->stats
.bytes
+= total_bytes
;
1182 tx_ring
->stats
.packets
+= total_packets
;
1183 u64_stats_update_end(&tx_ring
->syncp
);
1184 q_vector
->tx
.total_bytes
+= total_bytes
;
1185 q_vector
->tx
.total_packets
+= total_packets
;
1187 if (check_for_tx_hang(tx_ring
) && ixgbe_check_tx_hang(tx_ring
)) {
1188 /* schedule immediate reset if we believe we hung */
1189 struct ixgbe_hw
*hw
= &adapter
->hw
;
1190 e_err(drv
, "Detected Tx Unit Hang\n"
1192 " TDH, TDT <%x>, <%x>\n"
1193 " next_to_use <%x>\n"
1194 " next_to_clean <%x>\n"
1195 "tx_buffer_info[next_to_clean]\n"
1196 " time_stamp <%lx>\n"
1198 tx_ring
->queue_index
,
1199 IXGBE_READ_REG(hw
, IXGBE_TDH(tx_ring
->reg_idx
)),
1200 IXGBE_READ_REG(hw
, IXGBE_TDT(tx_ring
->reg_idx
)),
1201 tx_ring
->next_to_use
, i
,
1202 tx_ring
->tx_buffer_info
[i
].time_stamp
, jiffies
);
1204 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
1207 "tx hang %d detected on queue %d, resetting adapter\n",
1208 adapter
->tx_timeout_count
+ 1, tx_ring
->queue_index
);
1210 /* schedule immediate reset if we believe we hung */
1211 ixgbe_tx_timeout_reset(adapter
);
1213 /* the adapter is about to reset, no point in enabling stuff */
1217 netdev_tx_completed_queue(txring_txq(tx_ring
),
1218 total_packets
, total_bytes
);
1220 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1221 if (unlikely(total_packets
&& netif_carrier_ok(tx_ring
->netdev
) &&
1222 (ixgbe_desc_unused(tx_ring
) >= TX_WAKE_THRESHOLD
))) {
1223 /* Make sure that anybody stopping the queue after this
1224 * sees the new next_to_clean.
1227 if (__netif_subqueue_stopped(tx_ring
->netdev
,
1228 tx_ring
->queue_index
)
1229 && !test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
1230 netif_wake_subqueue(tx_ring
->netdev
,
1231 tx_ring
->queue_index
);
1232 ++tx_ring
->tx_stats
.restart_queue
;
1239 #ifdef CONFIG_IXGBE_DCA
1240 static void ixgbe_update_tx_dca(struct ixgbe_adapter
*adapter
,
1241 struct ixgbe_ring
*tx_ring
,
1244 struct ixgbe_hw
*hw
= &adapter
->hw
;
1248 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
1249 txctrl
= dca3_get_tag(tx_ring
->dev
, cpu
);
1251 switch (hw
->mac
.type
) {
1252 case ixgbe_mac_82598EB
:
1253 reg_offset
= IXGBE_DCA_TXCTRL(tx_ring
->reg_idx
);
1255 case ixgbe_mac_82599EB
:
1256 case ixgbe_mac_X540
:
1257 reg_offset
= IXGBE_DCA_TXCTRL_82599(tx_ring
->reg_idx
);
1258 txctrl
<<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599
;
1261 /* for unknown hardware do not write register */
1266 * We can enable relaxed ordering for reads, but not writes when
1267 * DCA is enabled. This is due to a known issue in some chipsets
1268 * which will cause the DCA tag to be cleared.
1270 txctrl
|= IXGBE_DCA_TXCTRL_DESC_RRO_EN
|
1271 IXGBE_DCA_TXCTRL_DATA_RRO_EN
|
1272 IXGBE_DCA_TXCTRL_DESC_DCA_EN
;
1274 IXGBE_WRITE_REG(hw
, reg_offset
, txctrl
);
1277 static void ixgbe_update_rx_dca(struct ixgbe_adapter
*adapter
,
1278 struct ixgbe_ring
*rx_ring
,
1281 struct ixgbe_hw
*hw
= &adapter
->hw
;
1283 u8 reg_idx
= rx_ring
->reg_idx
;
1285 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
1286 rxctrl
= dca3_get_tag(rx_ring
->dev
, cpu
);
1288 switch (hw
->mac
.type
) {
1289 case ixgbe_mac_82599EB
:
1290 case ixgbe_mac_X540
:
1291 rxctrl
<<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599
;
1298 * We can enable relaxed ordering for reads, but not writes when
1299 * DCA is enabled. This is due to a known issue in some chipsets
1300 * which will cause the DCA tag to be cleared.
1302 rxctrl
|= IXGBE_DCA_RXCTRL_DESC_RRO_EN
|
1303 IXGBE_DCA_RXCTRL_DATA_DCA_EN
|
1304 IXGBE_DCA_RXCTRL_DESC_DCA_EN
;
1306 IXGBE_WRITE_REG(hw
, IXGBE_DCA_RXCTRL(reg_idx
), rxctrl
);
1309 static void ixgbe_update_dca(struct ixgbe_q_vector
*q_vector
)
1311 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
1312 struct ixgbe_ring
*ring
;
1313 int cpu
= get_cpu();
1315 if (q_vector
->cpu
== cpu
)
1318 ixgbe_for_each_ring(ring
, q_vector
->tx
)
1319 ixgbe_update_tx_dca(adapter
, ring
, cpu
);
1321 ixgbe_for_each_ring(ring
, q_vector
->rx
)
1322 ixgbe_update_rx_dca(adapter
, ring
, cpu
);
1324 q_vector
->cpu
= cpu
;
1329 static void ixgbe_setup_dca(struct ixgbe_adapter
*adapter
)
1333 /* always use CB2 mode, difference is masked in the CB driver */
1334 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
1335 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
,
1336 IXGBE_DCA_CTRL_DCA_MODE_CB2
);
1338 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
,
1339 IXGBE_DCA_CTRL_DCA_DISABLE
);
1341 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
1342 adapter
->q_vector
[i
]->cpu
= -1;
1343 ixgbe_update_dca(adapter
->q_vector
[i
]);
1347 static int __ixgbe_notify_dca(struct device
*dev
, void *data
)
1349 struct ixgbe_adapter
*adapter
= dev_get_drvdata(dev
);
1350 unsigned long event
= *(unsigned long *)data
;
1352 if (!(adapter
->flags
& IXGBE_FLAG_DCA_CAPABLE
))
1356 case DCA_PROVIDER_ADD
:
1357 /* if we're already enabled, don't do it again */
1358 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
1360 if (dca_add_requester(dev
) == 0) {
1361 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
1362 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
,
1363 IXGBE_DCA_CTRL_DCA_MODE_CB2
);
1366 /* Fall Through since DCA is disabled. */
1367 case DCA_PROVIDER_REMOVE
:
1368 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
1369 dca_remove_requester(dev
);
1370 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
1371 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
,
1372 IXGBE_DCA_CTRL_DCA_DISABLE
);
1380 #endif /* CONFIG_IXGBE_DCA */
1382 #define IXGBE_RSS_L4_TYPES_MASK \
1383 ((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1384 (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1385 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1386 (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1388 static inline void ixgbe_rx_hash(struct ixgbe_ring
*ring
,
1389 union ixgbe_adv_rx_desc
*rx_desc
,
1390 struct sk_buff
*skb
)
1394 if (!(ring
->netdev
->features
& NETIF_F_RXHASH
))
1397 rss_type
= le16_to_cpu(rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
) &
1398 IXGBE_RXDADV_RSSTYPE_MASK
;
1403 skb_set_hash(skb
, le32_to_cpu(rx_desc
->wb
.lower
.hi_dword
.rss
),
1404 (IXGBE_RSS_L4_TYPES_MASK
& (1ul << rss_type
)) ?
1405 PKT_HASH_TYPE_L4
: PKT_HASH_TYPE_L3
);
1410 * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1411 * @ring: structure containing ring specific data
1412 * @rx_desc: advanced rx descriptor
1414 * Returns : true if it is FCoE pkt
1416 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring
*ring
,
1417 union ixgbe_adv_rx_desc
*rx_desc
)
1419 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1421 return test_bit(__IXGBE_RX_FCOE
, &ring
->state
) &&
1422 ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK
)) ==
1423 (cpu_to_le16(IXGBE_ETQF_FILTER_FCOE
<<
1424 IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT
)));
1427 #endif /* IXGBE_FCOE */
1429 * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1430 * @ring: structure containing ring specific data
1431 * @rx_desc: current Rx descriptor being processed
1432 * @skb: skb currently being received and modified
1434 static inline void ixgbe_rx_checksum(struct ixgbe_ring
*ring
,
1435 union ixgbe_adv_rx_desc
*rx_desc
,
1436 struct sk_buff
*skb
)
1438 __le16 pkt_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.pkt_info
;
1439 __le16 hdr_info
= rx_desc
->wb
.lower
.lo_dword
.hs_rss
.hdr_info
;
1440 bool encap_pkt
= false;
1442 skb_checksum_none_assert(skb
);
1444 /* Rx csum disabled */
1445 if (!(ring
->netdev
->features
& NETIF_F_RXCSUM
))
1448 if ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN
)) &&
1449 (hdr_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_TUNNEL
>> 16))) {
1451 skb
->encapsulation
= 1;
1454 /* if IP and error */
1455 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_IPCS
) &&
1456 ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_IPE
)) {
1457 ring
->rx_stats
.csum_err
++;
1461 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_L4CS
))
1464 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_TCPE
)) {
1466 * 82599 errata, UDP frames with a 0 checksum can be marked as
1469 if ((pkt_info
& cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP
)) &&
1470 test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR
, &ring
->state
))
1473 ring
->rx_stats
.csum_err
++;
1477 /* It must be a TCP or UDP packet with a valid checksum */
1478 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1480 if (!ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_OUTERIPCS
))
1483 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_ERR_OUTERIPER
)) {
1484 ring
->rx_stats
.csum_err
++;
1487 /* If we checked the outer header let the stack know */
1488 skb
->csum_level
= 1;
1492 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring
*rx_ring
,
1493 struct ixgbe_rx_buffer
*bi
)
1495 struct page
*page
= bi
->page
;
1498 /* since we are recycling buffers we should seldom need to alloc */
1502 /* alloc new page for storage */
1503 page
= dev_alloc_pages(ixgbe_rx_pg_order(rx_ring
));
1504 if (unlikely(!page
)) {
1505 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1509 /* map page for use */
1510 dma
= dma_map_page(rx_ring
->dev
, page
, 0,
1511 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1514 * if mapping failed free memory back to system since
1515 * there isn't much point in holding memory we can't use
1517 if (dma_mapping_error(rx_ring
->dev
, dma
)) {
1518 __free_pages(page
, ixgbe_rx_pg_order(rx_ring
));
1520 rx_ring
->rx_stats
.alloc_rx_page_failed
++;
1526 bi
->page_offset
= 0;
1532 * ixgbe_alloc_rx_buffers - Replace used receive buffers
1533 * @rx_ring: ring to place buffers on
1534 * @cleaned_count: number of buffers to replace
1536 void ixgbe_alloc_rx_buffers(struct ixgbe_ring
*rx_ring
, u16 cleaned_count
)
1538 union ixgbe_adv_rx_desc
*rx_desc
;
1539 struct ixgbe_rx_buffer
*bi
;
1540 u16 i
= rx_ring
->next_to_use
;
1546 rx_desc
= IXGBE_RX_DESC(rx_ring
, i
);
1547 bi
= &rx_ring
->rx_buffer_info
[i
];
1548 i
-= rx_ring
->count
;
1551 if (!ixgbe_alloc_mapped_page(rx_ring
, bi
))
1555 * Refresh the desc even if buffer_addrs didn't change
1556 * because each write-back erases this info.
1558 rx_desc
->read
.pkt_addr
= cpu_to_le64(bi
->dma
+ bi
->page_offset
);
1564 rx_desc
= IXGBE_RX_DESC(rx_ring
, 0);
1565 bi
= rx_ring
->rx_buffer_info
;
1566 i
-= rx_ring
->count
;
1569 /* clear the status bits for the next_to_use descriptor */
1570 rx_desc
->wb
.upper
.status_error
= 0;
1573 } while (cleaned_count
);
1575 i
+= rx_ring
->count
;
1577 if (rx_ring
->next_to_use
!= i
) {
1578 rx_ring
->next_to_use
= i
;
1580 /* update next to alloc since we have filled the ring */
1581 rx_ring
->next_to_alloc
= i
;
1583 /* Force memory writes to complete before letting h/w
1584 * know there are new descriptors to fetch. (Only
1585 * applicable for weak-ordered memory model archs,
1589 writel(i
, rx_ring
->tail
);
1593 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring
*ring
,
1594 struct sk_buff
*skb
)
1596 u16 hdr_len
= skb_headlen(skb
);
1598 /* set gso_size to avoid messing up TCP MSS */
1599 skb_shinfo(skb
)->gso_size
= DIV_ROUND_UP((skb
->len
- hdr_len
),
1600 IXGBE_CB(skb
)->append_cnt
);
1601 skb_shinfo(skb
)->gso_type
= SKB_GSO_TCPV4
;
1604 static void ixgbe_update_rsc_stats(struct ixgbe_ring
*rx_ring
,
1605 struct sk_buff
*skb
)
1607 /* if append_cnt is 0 then frame is not RSC */
1608 if (!IXGBE_CB(skb
)->append_cnt
)
1611 rx_ring
->rx_stats
.rsc_count
+= IXGBE_CB(skb
)->append_cnt
;
1612 rx_ring
->rx_stats
.rsc_flush
++;
1614 ixgbe_set_rsc_gso_size(rx_ring
, skb
);
1616 /* gso_size is computed using append_cnt so always clear it last */
1617 IXGBE_CB(skb
)->append_cnt
= 0;
1621 * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1622 * @rx_ring: rx descriptor ring packet is being transacted on
1623 * @rx_desc: pointer to the EOP Rx descriptor
1624 * @skb: pointer to current skb being populated
1626 * This function checks the ring, descriptor, and packet information in
1627 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1628 * other fields within the skb.
1630 static void ixgbe_process_skb_fields(struct ixgbe_ring
*rx_ring
,
1631 union ixgbe_adv_rx_desc
*rx_desc
,
1632 struct sk_buff
*skb
)
1634 struct net_device
*dev
= rx_ring
->netdev
;
1636 ixgbe_update_rsc_stats(rx_ring
, skb
);
1638 ixgbe_rx_hash(rx_ring
, rx_desc
, skb
);
1640 ixgbe_rx_checksum(rx_ring
, rx_desc
, skb
);
1642 if (unlikely(ixgbe_test_staterr(rx_desc
, IXGBE_RXDADV_STAT_TS
)))
1643 ixgbe_ptp_rx_hwtstamp(rx_ring
->q_vector
->adapter
, skb
);
1645 if ((dev
->features
& NETIF_F_HW_VLAN_CTAG_RX
) &&
1646 ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_VP
)) {
1647 u16 vid
= le16_to_cpu(rx_desc
->wb
.upper
.vlan
);
1648 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), vid
);
1651 skb_record_rx_queue(skb
, rx_ring
->queue_index
);
1653 skb
->protocol
= eth_type_trans(skb
, dev
);
1656 static void ixgbe_rx_skb(struct ixgbe_q_vector
*q_vector
,
1657 struct sk_buff
*skb
)
1659 skb_mark_napi_id(skb
, &q_vector
->napi
);
1660 if (ixgbe_qv_busy_polling(q_vector
))
1661 netif_receive_skb(skb
);
1663 napi_gro_receive(&q_vector
->napi
, skb
);
1667 * ixgbe_is_non_eop - process handling of non-EOP buffers
1668 * @rx_ring: Rx ring being processed
1669 * @rx_desc: Rx descriptor for current buffer
1670 * @skb: Current socket buffer containing buffer in progress
1672 * This function updates next to clean. If the buffer is an EOP buffer
1673 * this function exits returning false, otherwise it will place the
1674 * sk_buff in the next buffer to be chained and return true indicating
1675 * that this is in fact a non-EOP buffer.
1677 static bool ixgbe_is_non_eop(struct ixgbe_ring
*rx_ring
,
1678 union ixgbe_adv_rx_desc
*rx_desc
,
1679 struct sk_buff
*skb
)
1681 u32 ntc
= rx_ring
->next_to_clean
+ 1;
1683 /* fetch, update, and store next to clean */
1684 ntc
= (ntc
< rx_ring
->count
) ? ntc
: 0;
1685 rx_ring
->next_to_clean
= ntc
;
1687 prefetch(IXGBE_RX_DESC(rx_ring
, ntc
));
1689 /* update RSC append count if present */
1690 if (ring_is_rsc_enabled(rx_ring
)) {
1691 __le32 rsc_enabled
= rx_desc
->wb
.lower
.lo_dword
.data
&
1692 cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK
);
1694 if (unlikely(rsc_enabled
)) {
1695 u32 rsc_cnt
= le32_to_cpu(rsc_enabled
);
1697 rsc_cnt
>>= IXGBE_RXDADV_RSCCNT_SHIFT
;
1698 IXGBE_CB(skb
)->append_cnt
+= rsc_cnt
- 1;
1700 /* update ntc based on RSC value */
1701 ntc
= le32_to_cpu(rx_desc
->wb
.upper
.status_error
);
1702 ntc
&= IXGBE_RXDADV_NEXTP_MASK
;
1703 ntc
>>= IXGBE_RXDADV_NEXTP_SHIFT
;
1707 /* if we are the last buffer then there is nothing else to do */
1708 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1711 /* place skb in next buffer to be received */
1712 rx_ring
->rx_buffer_info
[ntc
].skb
= skb
;
1713 rx_ring
->rx_stats
.non_eop_descs
++;
1719 * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1720 * @rx_ring: rx descriptor ring packet is being transacted on
1721 * @skb: pointer to current skb being adjusted
1723 * This function is an ixgbe specific version of __pskb_pull_tail. The
1724 * main difference between this version and the original function is that
1725 * this function can make several assumptions about the state of things
1726 * that allow for significant optimizations versus the standard function.
1727 * As a result we can do things like drop a frag and maintain an accurate
1728 * truesize for the skb.
1730 static void ixgbe_pull_tail(struct ixgbe_ring
*rx_ring
,
1731 struct sk_buff
*skb
)
1733 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1735 unsigned int pull_len
;
1738 * it is valid to use page_address instead of kmap since we are
1739 * working with pages allocated out of the lomem pool per
1740 * alloc_page(GFP_ATOMIC)
1742 va
= skb_frag_address(frag
);
1745 * we need the header to contain the greater of either ETH_HLEN or
1746 * 60 bytes if the skb->len is less than 60 for skb_pad.
1748 pull_len
= eth_get_headlen(va
, IXGBE_RX_HDR_SIZE
);
1750 /* align pull length to size of long to optimize memcpy performance */
1751 skb_copy_to_linear_data(skb
, va
, ALIGN(pull_len
, sizeof(long)));
1753 /* update all of the pointers */
1754 skb_frag_size_sub(frag
, pull_len
);
1755 frag
->page_offset
+= pull_len
;
1756 skb
->data_len
-= pull_len
;
1757 skb
->tail
+= pull_len
;
1761 * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1762 * @rx_ring: rx descriptor ring packet is being transacted on
1763 * @skb: pointer to current skb being updated
1765 * This function provides a basic DMA sync up for the first fragment of an
1766 * skb. The reason for doing this is that the first fragment cannot be
1767 * unmapped until we have reached the end of packet descriptor for a buffer
1770 static void ixgbe_dma_sync_frag(struct ixgbe_ring
*rx_ring
,
1771 struct sk_buff
*skb
)
1773 /* if the page was released unmap it, else just sync our portion */
1774 if (unlikely(IXGBE_CB(skb
)->page_released
)) {
1775 dma_unmap_page(rx_ring
->dev
, IXGBE_CB(skb
)->dma
,
1776 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
1777 IXGBE_CB(skb
)->page_released
= false;
1779 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[0];
1781 dma_sync_single_range_for_cpu(rx_ring
->dev
,
1784 ixgbe_rx_bufsz(rx_ring
),
1787 IXGBE_CB(skb
)->dma
= 0;
1791 * ixgbe_cleanup_headers - Correct corrupted or empty headers
1792 * @rx_ring: rx descriptor ring packet is being transacted on
1793 * @rx_desc: pointer to the EOP Rx descriptor
1794 * @skb: pointer to current skb being fixed
1796 * Check for corrupted packet headers caused by senders on the local L2
1797 * embedded NIC switch not setting up their Tx Descriptors right. These
1798 * should be very rare.
1800 * Also address the case where we are pulling data in on pages only
1801 * and as such no data is present in the skb header.
1803 * In addition if skb is not at least 60 bytes we need to pad it so that
1804 * it is large enough to qualify as a valid Ethernet frame.
1806 * Returns true if an error was encountered and skb was freed.
1808 static bool ixgbe_cleanup_headers(struct ixgbe_ring
*rx_ring
,
1809 union ixgbe_adv_rx_desc
*rx_desc
,
1810 struct sk_buff
*skb
)
1812 struct net_device
*netdev
= rx_ring
->netdev
;
1814 /* verify that the packet does not have any known errors */
1815 if (unlikely(ixgbe_test_staterr(rx_desc
,
1816 IXGBE_RXDADV_ERR_FRAME_ERR_MASK
) &&
1817 !(netdev
->features
& NETIF_F_RXALL
))) {
1818 dev_kfree_skb_any(skb
);
1822 /* place header in linear portion of buffer */
1823 if (skb_is_nonlinear(skb
))
1824 ixgbe_pull_tail(rx_ring
, skb
);
1827 /* do not attempt to pad FCoE Frames as this will disrupt DDP */
1828 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
))
1832 /* if eth_skb_pad returns an error the skb was freed */
1833 if (eth_skb_pad(skb
))
1840 * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1841 * @rx_ring: rx descriptor ring to store buffers on
1842 * @old_buff: donor buffer to have page reused
1844 * Synchronizes page for reuse by the adapter
1846 static void ixgbe_reuse_rx_page(struct ixgbe_ring
*rx_ring
,
1847 struct ixgbe_rx_buffer
*old_buff
)
1849 struct ixgbe_rx_buffer
*new_buff
;
1850 u16 nta
= rx_ring
->next_to_alloc
;
1852 new_buff
= &rx_ring
->rx_buffer_info
[nta
];
1854 /* update, and store next to alloc */
1856 rx_ring
->next_to_alloc
= (nta
< rx_ring
->count
) ? nta
: 0;
1858 /* transfer page from old buffer to new buffer */
1859 *new_buff
= *old_buff
;
1861 /* sync the buffer for use by the device */
1862 dma_sync_single_range_for_device(rx_ring
->dev
, new_buff
->dma
,
1863 new_buff
->page_offset
,
1864 ixgbe_rx_bufsz(rx_ring
),
1868 static inline bool ixgbe_page_is_reserved(struct page
*page
)
1870 return (page_to_nid(page
) != numa_mem_id()) || page_is_pfmemalloc(page
);
1874 * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1875 * @rx_ring: rx descriptor ring to transact packets on
1876 * @rx_buffer: buffer containing page to add
1877 * @rx_desc: descriptor containing length of buffer written by hardware
1878 * @skb: sk_buff to place the data into
1880 * This function will add the data contained in rx_buffer->page to the skb.
1881 * This is done either through a direct copy if the data in the buffer is
1882 * less than the skb header size, otherwise it will just attach the page as
1883 * a frag to the skb.
1885 * The function will then update the page offset if necessary and return
1886 * true if the buffer can be reused by the adapter.
1888 static bool ixgbe_add_rx_frag(struct ixgbe_ring
*rx_ring
,
1889 struct ixgbe_rx_buffer
*rx_buffer
,
1890 union ixgbe_adv_rx_desc
*rx_desc
,
1891 struct sk_buff
*skb
)
1893 struct page
*page
= rx_buffer
->page
;
1894 unsigned int size
= le16_to_cpu(rx_desc
->wb
.upper
.length
);
1895 #if (PAGE_SIZE < 8192)
1896 unsigned int truesize
= ixgbe_rx_bufsz(rx_ring
);
1898 unsigned int truesize
= ALIGN(size
, L1_CACHE_BYTES
);
1899 unsigned int last_offset
= ixgbe_rx_pg_size(rx_ring
) -
1900 ixgbe_rx_bufsz(rx_ring
);
1903 if ((size
<= IXGBE_RX_HDR_SIZE
) && !skb_is_nonlinear(skb
)) {
1904 unsigned char *va
= page_address(page
) + rx_buffer
->page_offset
;
1906 memcpy(__skb_put(skb
, size
), va
, ALIGN(size
, sizeof(long)));
1908 /* page is not reserved, we can reuse buffer as-is */
1909 if (likely(!ixgbe_page_is_reserved(page
)))
1912 /* this page cannot be reused so discard it */
1913 __free_pages(page
, ixgbe_rx_pg_order(rx_ring
));
1917 skb_add_rx_frag(skb
, skb_shinfo(skb
)->nr_frags
, page
,
1918 rx_buffer
->page_offset
, size
, truesize
);
1920 /* avoid re-using remote pages */
1921 if (unlikely(ixgbe_page_is_reserved(page
)))
1924 #if (PAGE_SIZE < 8192)
1925 /* if we are only owner of page we can reuse it */
1926 if (unlikely(page_count(page
) != 1))
1929 /* flip page offset to other buffer */
1930 rx_buffer
->page_offset
^= truesize
;
1932 /* move offset up to the next cache line */
1933 rx_buffer
->page_offset
+= truesize
;
1935 if (rx_buffer
->page_offset
> last_offset
)
1939 /* Even if we own the page, we are not allowed to use atomic_set()
1940 * This would break get_page_unless_zero() users.
1942 atomic_inc(&page
->_count
);
1947 static struct sk_buff
*ixgbe_fetch_rx_buffer(struct ixgbe_ring
*rx_ring
,
1948 union ixgbe_adv_rx_desc
*rx_desc
)
1950 struct ixgbe_rx_buffer
*rx_buffer
;
1951 struct sk_buff
*skb
;
1954 rx_buffer
= &rx_ring
->rx_buffer_info
[rx_ring
->next_to_clean
];
1955 page
= rx_buffer
->page
;
1958 skb
= rx_buffer
->skb
;
1961 void *page_addr
= page_address(page
) +
1962 rx_buffer
->page_offset
;
1964 /* prefetch first cache line of first page */
1965 prefetch(page_addr
);
1966 #if L1_CACHE_BYTES < 128
1967 prefetch(page_addr
+ L1_CACHE_BYTES
);
1970 /* allocate a skb to store the frags */
1971 skb
= napi_alloc_skb(&rx_ring
->q_vector
->napi
,
1973 if (unlikely(!skb
)) {
1974 rx_ring
->rx_stats
.alloc_rx_buff_failed
++;
1979 * we will be copying header into skb->data in
1980 * pskb_may_pull so it is in our interest to prefetch
1981 * it now to avoid a possible cache miss
1983 prefetchw(skb
->data
);
1986 * Delay unmapping of the first packet. It carries the
1987 * header information, HW may still access the header
1988 * after the writeback. Only unmap it when EOP is
1991 if (likely(ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
)))
1994 IXGBE_CB(skb
)->dma
= rx_buffer
->dma
;
1996 if (ixgbe_test_staterr(rx_desc
, IXGBE_RXD_STAT_EOP
))
1997 ixgbe_dma_sync_frag(rx_ring
, skb
);
2000 /* we are reusing so sync this buffer for CPU use */
2001 dma_sync_single_range_for_cpu(rx_ring
->dev
,
2003 rx_buffer
->page_offset
,
2004 ixgbe_rx_bufsz(rx_ring
),
2007 rx_buffer
->skb
= NULL
;
2010 /* pull page into skb */
2011 if (ixgbe_add_rx_frag(rx_ring
, rx_buffer
, rx_desc
, skb
)) {
2012 /* hand second half of page back to the ring */
2013 ixgbe_reuse_rx_page(rx_ring
, rx_buffer
);
2014 } else if (IXGBE_CB(skb
)->dma
== rx_buffer
->dma
) {
2015 /* the page has been released from the ring */
2016 IXGBE_CB(skb
)->page_released
= true;
2018 /* we are not reusing the buffer so unmap it */
2019 dma_unmap_page(rx_ring
->dev
, rx_buffer
->dma
,
2020 ixgbe_rx_pg_size(rx_ring
),
2024 /* clear contents of buffer_info */
2025 rx_buffer
->page
= NULL
;
2031 * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2032 * @q_vector: structure containing interrupt and ring information
2033 * @rx_ring: rx descriptor ring to transact packets on
2034 * @budget: Total limit on number of packets to process
2036 * This function provides a "bounce buffer" approach to Rx interrupt
2037 * processing. The advantage to this is that on systems that have
2038 * expensive overhead for IOMMU access this provides a means of avoiding
2039 * it by maintaining the mapping of the page to the syste.
2041 * Returns amount of work completed
2043 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector
*q_vector
,
2044 struct ixgbe_ring
*rx_ring
,
2047 unsigned int total_rx_bytes
= 0, total_rx_packets
= 0;
2049 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2051 unsigned int mss
= 0;
2052 #endif /* IXGBE_FCOE */
2053 u16 cleaned_count
= ixgbe_desc_unused(rx_ring
);
2055 while (likely(total_rx_packets
< budget
)) {
2056 union ixgbe_adv_rx_desc
*rx_desc
;
2057 struct sk_buff
*skb
;
2059 /* return some buffers to hardware, one at a time is too slow */
2060 if (cleaned_count
>= IXGBE_RX_BUFFER_WRITE
) {
2061 ixgbe_alloc_rx_buffers(rx_ring
, cleaned_count
);
2065 rx_desc
= IXGBE_RX_DESC(rx_ring
, rx_ring
->next_to_clean
);
2067 if (!rx_desc
->wb
.upper
.status_error
)
2070 /* This memory barrier is needed to keep us from reading
2071 * any other fields out of the rx_desc until we know the
2072 * descriptor has been written back
2076 /* retrieve a buffer from the ring */
2077 skb
= ixgbe_fetch_rx_buffer(rx_ring
, rx_desc
);
2079 /* exit if we failed to retrieve a buffer */
2085 /* place incomplete frames back on ring for completion */
2086 if (ixgbe_is_non_eop(rx_ring
, rx_desc
, skb
))
2089 /* verify the packet layout is correct */
2090 if (ixgbe_cleanup_headers(rx_ring
, rx_desc
, skb
))
2093 /* probably a little skewed due to removing CRC */
2094 total_rx_bytes
+= skb
->len
;
2096 /* populate checksum, timestamp, VLAN, and protocol */
2097 ixgbe_process_skb_fields(rx_ring
, rx_desc
, skb
);
2100 /* if ddp, not passing to ULD unless for FCP_RSP or error */
2101 if (ixgbe_rx_is_fcoe(rx_ring
, rx_desc
)) {
2102 ddp_bytes
= ixgbe_fcoe_ddp(adapter
, rx_desc
, skb
);
2103 /* include DDPed FCoE data */
2104 if (ddp_bytes
> 0) {
2106 mss
= rx_ring
->netdev
->mtu
-
2107 sizeof(struct fcoe_hdr
) -
2108 sizeof(struct fc_frame_header
) -
2109 sizeof(struct fcoe_crc_eof
);
2113 total_rx_bytes
+= ddp_bytes
;
2114 total_rx_packets
+= DIV_ROUND_UP(ddp_bytes
,
2118 dev_kfree_skb_any(skb
);
2123 #endif /* IXGBE_FCOE */
2124 ixgbe_rx_skb(q_vector
, skb
);
2126 /* update budget accounting */
2130 u64_stats_update_begin(&rx_ring
->syncp
);
2131 rx_ring
->stats
.packets
+= total_rx_packets
;
2132 rx_ring
->stats
.bytes
+= total_rx_bytes
;
2133 u64_stats_update_end(&rx_ring
->syncp
);
2134 q_vector
->rx
.total_packets
+= total_rx_packets
;
2135 q_vector
->rx
.total_bytes
+= total_rx_bytes
;
2137 return total_rx_packets
;
2140 #ifdef CONFIG_NET_RX_BUSY_POLL
2141 /* must be called with local_bh_disable()d */
2142 static int ixgbe_low_latency_recv(struct napi_struct
*napi
)
2144 struct ixgbe_q_vector
*q_vector
=
2145 container_of(napi
, struct ixgbe_q_vector
, napi
);
2146 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2147 struct ixgbe_ring
*ring
;
2150 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
2151 return LL_FLUSH_FAILED
;
2153 if (!ixgbe_qv_lock_poll(q_vector
))
2154 return LL_FLUSH_BUSY
;
2156 ixgbe_for_each_ring(ring
, q_vector
->rx
) {
2157 found
= ixgbe_clean_rx_irq(q_vector
, ring
, 4);
2158 #ifdef BP_EXTENDED_STATS
2160 ring
->stats
.cleaned
+= found
;
2162 ring
->stats
.misses
++;
2168 ixgbe_qv_unlock_poll(q_vector
);
2172 #endif /* CONFIG_NET_RX_BUSY_POLL */
2175 * ixgbe_configure_msix - Configure MSI-X hardware
2176 * @adapter: board private structure
2178 * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2181 static void ixgbe_configure_msix(struct ixgbe_adapter
*adapter
)
2183 struct ixgbe_q_vector
*q_vector
;
2187 /* Populate MSIX to EITR Select */
2188 if (adapter
->num_vfs
> 32) {
2189 u32 eitrsel
= (1 << (adapter
->num_vfs
- 32)) - 1;
2190 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, eitrsel
);
2194 * Populate the IVAR table and set the ITR values to the
2195 * corresponding register.
2197 for (v_idx
= 0; v_idx
< adapter
->num_q_vectors
; v_idx
++) {
2198 struct ixgbe_ring
*ring
;
2199 q_vector
= adapter
->q_vector
[v_idx
];
2201 ixgbe_for_each_ring(ring
, q_vector
->rx
)
2202 ixgbe_set_ivar(adapter
, 0, ring
->reg_idx
, v_idx
);
2204 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2205 ixgbe_set_ivar(adapter
, 1, ring
->reg_idx
, v_idx
);
2207 ixgbe_write_eitr(q_vector
);
2210 switch (adapter
->hw
.mac
.type
) {
2211 case ixgbe_mac_82598EB
:
2212 ixgbe_set_ivar(adapter
, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX
,
2215 case ixgbe_mac_82599EB
:
2216 case ixgbe_mac_X540
:
2217 case ixgbe_mac_X550
:
2218 case ixgbe_mac_X550EM_x
:
2219 ixgbe_set_ivar(adapter
, -1, 1, v_idx
);
2224 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITR(v_idx
), 1950);
2226 /* set up to autoclear timer, and the vectors */
2227 mask
= IXGBE_EIMS_ENABLE_MASK
;
2228 mask
&= ~(IXGBE_EIMS_OTHER
|
2229 IXGBE_EIMS_MAILBOX
|
2232 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIAC
, mask
);
2235 enum latency_range
{
2239 latency_invalid
= 255
2243 * ixgbe_update_itr - update the dynamic ITR value based on statistics
2244 * @q_vector: structure containing interrupt and ring information
2245 * @ring_container: structure containing ring performance data
2247 * Stores a new ITR value based on packets and byte
2248 * counts during the last interrupt. The advantage of per interrupt
2249 * computation is faster updates and more accurate ITR for the current
2250 * traffic pattern. Constants in this function were computed
2251 * based on theoretical maximum wire speed and thresholds were set based
2252 * on testing data as well as attempting to minimize response time
2253 * while increasing bulk throughput.
2254 * this functionality is controlled by the InterruptThrottleRate module
2255 * parameter (see ixgbe_param.c)
2257 static void ixgbe_update_itr(struct ixgbe_q_vector
*q_vector
,
2258 struct ixgbe_ring_container
*ring_container
)
2260 int bytes
= ring_container
->total_bytes
;
2261 int packets
= ring_container
->total_packets
;
2264 u8 itr_setting
= ring_container
->itr
;
2269 /* simple throttlerate management
2270 * 0-10MB/s lowest (100000 ints/s)
2271 * 10-20MB/s low (20000 ints/s)
2272 * 20-1249MB/s bulk (12000 ints/s)
2274 /* what was last interrupt timeslice? */
2275 timepassed_us
= q_vector
->itr
>> 2;
2276 if (timepassed_us
== 0)
2279 bytes_perint
= bytes
/ timepassed_us
; /* bytes/usec */
2281 switch (itr_setting
) {
2282 case lowest_latency
:
2283 if (bytes_perint
> 10)
2284 itr_setting
= low_latency
;
2287 if (bytes_perint
> 20)
2288 itr_setting
= bulk_latency
;
2289 else if (bytes_perint
<= 10)
2290 itr_setting
= lowest_latency
;
2293 if (bytes_perint
<= 20)
2294 itr_setting
= low_latency
;
2298 /* clear work counters since we have the values we need */
2299 ring_container
->total_bytes
= 0;
2300 ring_container
->total_packets
= 0;
2302 /* write updated itr to ring container */
2303 ring_container
->itr
= itr_setting
;
2307 * ixgbe_write_eitr - write EITR register in hardware specific way
2308 * @q_vector: structure containing interrupt and ring information
2310 * This function is made to be called by ethtool and by the driver
2311 * when it needs to update EITR registers at runtime. Hardware
2312 * specific quirks/differences are taken care of here.
2314 void ixgbe_write_eitr(struct ixgbe_q_vector
*q_vector
)
2316 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2317 struct ixgbe_hw
*hw
= &adapter
->hw
;
2318 int v_idx
= q_vector
->v_idx
;
2319 u32 itr_reg
= q_vector
->itr
& IXGBE_MAX_EITR
;
2321 switch (adapter
->hw
.mac
.type
) {
2322 case ixgbe_mac_82598EB
:
2323 /* must write high and low 16 bits to reset counter */
2324 itr_reg
|= (itr_reg
<< 16);
2326 case ixgbe_mac_82599EB
:
2327 case ixgbe_mac_X540
:
2328 case ixgbe_mac_X550
:
2329 case ixgbe_mac_X550EM_x
:
2331 * set the WDIS bit to not clear the timer bits and cause an
2332 * immediate assertion of the interrupt
2334 itr_reg
|= IXGBE_EITR_CNT_WDIS
;
2339 IXGBE_WRITE_REG(hw
, IXGBE_EITR(v_idx
), itr_reg
);
2342 static void ixgbe_set_itr(struct ixgbe_q_vector
*q_vector
)
2344 u32 new_itr
= q_vector
->itr
;
2347 ixgbe_update_itr(q_vector
, &q_vector
->tx
);
2348 ixgbe_update_itr(q_vector
, &q_vector
->rx
);
2350 current_itr
= max(q_vector
->rx
.itr
, q_vector
->tx
.itr
);
2352 switch (current_itr
) {
2353 /* counts and packets in update_itr are dependent on these numbers */
2354 case lowest_latency
:
2355 new_itr
= IXGBE_100K_ITR
;
2358 new_itr
= IXGBE_20K_ITR
;
2361 new_itr
= IXGBE_12K_ITR
;
2367 if (new_itr
!= q_vector
->itr
) {
2368 /* do an exponential smoothing */
2369 new_itr
= (10 * new_itr
* q_vector
->itr
) /
2370 ((9 * new_itr
) + q_vector
->itr
);
2372 /* save the algorithm value here */
2373 q_vector
->itr
= new_itr
;
2375 ixgbe_write_eitr(q_vector
);
2380 * ixgbe_check_overtemp_subtask - check for over temperature
2381 * @adapter: pointer to adapter
2383 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter
*adapter
)
2385 struct ixgbe_hw
*hw
= &adapter
->hw
;
2386 u32 eicr
= adapter
->interrupt_event
;
2388 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
2391 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) &&
2392 !(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_EVENT
))
2395 adapter
->flags2
&= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2397 switch (hw
->device_id
) {
2398 case IXGBE_DEV_ID_82599_T3_LOM
:
2400 * Since the warning interrupt is for both ports
2401 * we don't have to check if:
2402 * - This interrupt wasn't for our port.
2403 * - We may have missed the interrupt so always have to
2404 * check if we got a LSC
2406 if (!(eicr
& IXGBE_EICR_GPI_SDP0_8259X
) &&
2407 !(eicr
& IXGBE_EICR_LSC
))
2410 if (!(eicr
& IXGBE_EICR_LSC
) && hw
->mac
.ops
.check_link
) {
2412 bool link_up
= false;
2414 hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
2420 /* Check if this is not due to overtemp */
2421 if (hw
->phy
.ops
.check_overtemp(hw
) != IXGBE_ERR_OVERTEMP
)
2426 if (adapter
->hw
.mac
.type
>= ixgbe_mac_X540
)
2428 if (!(eicr
& IXGBE_EICR_GPI_SDP0(hw
)))
2432 e_crit(drv
, "%s\n", ixgbe_overheat_msg
);
2434 adapter
->interrupt_event
= 0;
2437 static void ixgbe_check_fan_failure(struct ixgbe_adapter
*adapter
, u32 eicr
)
2439 struct ixgbe_hw
*hw
= &adapter
->hw
;
2441 if ((adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) &&
2442 (eicr
& IXGBE_EICR_GPI_SDP1(hw
))) {
2443 e_crit(probe
, "Fan has stopped, replace the adapter\n");
2444 /* write to clear the interrupt */
2445 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1(hw
));
2449 static void ixgbe_check_overtemp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2451 struct ixgbe_hw
*hw
= &adapter
->hw
;
2453 if (!(adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
))
2456 switch (adapter
->hw
.mac
.type
) {
2457 case ixgbe_mac_82599EB
:
2459 * Need to check link state so complete overtemp check
2462 if (((eicr
& IXGBE_EICR_GPI_SDP0(hw
)) ||
2463 (eicr
& IXGBE_EICR_LSC
)) &&
2464 (!test_bit(__IXGBE_DOWN
, &adapter
->state
))) {
2465 adapter
->interrupt_event
= eicr
;
2466 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_EVENT
;
2467 ixgbe_service_event_schedule(adapter
);
2471 case ixgbe_mac_X540
:
2472 if (!(eicr
& IXGBE_EICR_TS
))
2479 e_crit(drv
, "%s\n", ixgbe_overheat_msg
);
2482 static inline bool ixgbe_is_sfp(struct ixgbe_hw
*hw
)
2484 switch (hw
->mac
.type
) {
2485 case ixgbe_mac_82598EB
:
2486 if (hw
->phy
.type
== ixgbe_phy_nl
)
2489 case ixgbe_mac_82599EB
:
2490 case ixgbe_mac_X550EM_x
:
2491 switch (hw
->mac
.ops
.get_media_type(hw
)) {
2492 case ixgbe_media_type_fiber
:
2493 case ixgbe_media_type_fiber_qsfp
:
2503 static void ixgbe_check_sfp_event(struct ixgbe_adapter
*adapter
, u32 eicr
)
2505 struct ixgbe_hw
*hw
= &adapter
->hw
;
2506 u32 eicr_mask
= IXGBE_EICR_GPI_SDP2(hw
);
2508 if (!ixgbe_is_sfp(hw
))
2511 /* Later MAC's use different SDP */
2512 if (hw
->mac
.type
>= ixgbe_mac_X540
)
2513 eicr_mask
= IXGBE_EICR_GPI_SDP0_X540
;
2515 if (eicr
& eicr_mask
) {
2516 /* Clear the interrupt */
2517 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, eicr_mask
);
2518 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2519 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
2520 adapter
->sfp_poll_time
= 0;
2521 ixgbe_service_event_schedule(adapter
);
2525 if (adapter
->hw
.mac
.type
== ixgbe_mac_82599EB
&&
2526 (eicr
& IXGBE_EICR_GPI_SDP1(hw
))) {
2527 /* Clear the interrupt */
2528 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_GPI_SDP1(hw
));
2529 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2530 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
2531 ixgbe_service_event_schedule(adapter
);
2536 static void ixgbe_check_lsc(struct ixgbe_adapter
*adapter
)
2538 struct ixgbe_hw
*hw
= &adapter
->hw
;
2541 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
2542 adapter
->link_check_timeout
= jiffies
;
2543 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
2544 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_LSC
);
2545 IXGBE_WRITE_FLUSH(hw
);
2546 ixgbe_service_event_schedule(adapter
);
2550 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter
*adapter
,
2554 struct ixgbe_hw
*hw
= &adapter
->hw
;
2556 switch (hw
->mac
.type
) {
2557 case ixgbe_mac_82598EB
:
2558 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2559 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, mask
);
2561 case ixgbe_mac_82599EB
:
2562 case ixgbe_mac_X540
:
2563 case ixgbe_mac_X550
:
2564 case ixgbe_mac_X550EM_x
:
2565 mask
= (qmask
& 0xFFFFFFFF);
2567 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(0), mask
);
2568 mask
= (qmask
>> 32);
2570 IXGBE_WRITE_REG(hw
, IXGBE_EIMS_EX(1), mask
);
2575 /* skip the flush */
2578 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter
*adapter
,
2582 struct ixgbe_hw
*hw
= &adapter
->hw
;
2584 switch (hw
->mac
.type
) {
2585 case ixgbe_mac_82598EB
:
2586 mask
= (IXGBE_EIMS_RTX_QUEUE
& qmask
);
2587 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, mask
);
2589 case ixgbe_mac_82599EB
:
2590 case ixgbe_mac_X540
:
2591 case ixgbe_mac_X550
:
2592 case ixgbe_mac_X550EM_x
:
2593 mask
= (qmask
& 0xFFFFFFFF);
2595 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(0), mask
);
2596 mask
= (qmask
>> 32);
2598 IXGBE_WRITE_REG(hw
, IXGBE_EIMC_EX(1), mask
);
2603 /* skip the flush */
2607 * ixgbe_irq_enable - Enable default interrupt generation settings
2608 * @adapter: board private structure
2610 static inline void ixgbe_irq_enable(struct ixgbe_adapter
*adapter
, bool queues
,
2613 struct ixgbe_hw
*hw
= &adapter
->hw
;
2614 u32 mask
= (IXGBE_EIMS_ENABLE_MASK
& ~IXGBE_EIMS_RTX_QUEUE
);
2616 /* don't reenable LSC while waiting for link */
2617 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
2618 mask
&= ~IXGBE_EIMS_LSC
;
2620 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
)
2621 switch (adapter
->hw
.mac
.type
) {
2622 case ixgbe_mac_82599EB
:
2623 mask
|= IXGBE_EIMS_GPI_SDP0(hw
);
2625 case ixgbe_mac_X540
:
2626 case ixgbe_mac_X550
:
2627 case ixgbe_mac_X550EM_x
:
2628 mask
|= IXGBE_EIMS_TS
;
2633 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
2634 mask
|= IXGBE_EIMS_GPI_SDP1(hw
);
2635 switch (adapter
->hw
.mac
.type
) {
2636 case ixgbe_mac_82599EB
:
2637 mask
|= IXGBE_EIMS_GPI_SDP1(hw
);
2638 mask
|= IXGBE_EIMS_GPI_SDP2(hw
);
2640 case ixgbe_mac_X540
:
2641 case ixgbe_mac_X550
:
2642 case ixgbe_mac_X550EM_x
:
2643 if (adapter
->hw
.device_id
== IXGBE_DEV_ID_X550EM_X_SFP
)
2644 mask
|= IXGBE_EIMS_GPI_SDP0(&adapter
->hw
);
2645 if (adapter
->hw
.phy
.type
== ixgbe_phy_x550em_ext_t
)
2646 mask
|= IXGBE_EICR_GPI_SDP0_X540
;
2647 mask
|= IXGBE_EIMS_ECC
;
2648 mask
|= IXGBE_EIMS_MAILBOX
;
2654 if ((adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) &&
2655 !(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
2656 mask
|= IXGBE_EIMS_FLOW_DIR
;
2658 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMS
, mask
);
2660 ixgbe_irq_enable_queues(adapter
, ~0);
2662 IXGBE_WRITE_FLUSH(&adapter
->hw
);
2665 static irqreturn_t
ixgbe_msix_other(int irq
, void *data
)
2667 struct ixgbe_adapter
*adapter
= data
;
2668 struct ixgbe_hw
*hw
= &adapter
->hw
;
2672 * Workaround for Silicon errata. Use clear-by-write instead
2673 * of clear-by-read. Reading with EICS will return the
2674 * interrupt causes without clearing, which later be done
2675 * with the write to EICR.
2677 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICS
);
2679 /* The lower 16bits of the EICR register are for the queue interrupts
2680 * which should be masked here in order to not accidentally clear them if
2681 * the bits are high when ixgbe_msix_other is called. There is a race
2682 * condition otherwise which results in possible performance loss
2683 * especially if the ixgbe_msix_other interrupt is triggering
2684 * consistently (as it would when PPS is turned on for the X540 device)
2688 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, eicr
);
2690 if (eicr
& IXGBE_EICR_LSC
)
2691 ixgbe_check_lsc(adapter
);
2693 if (eicr
& IXGBE_EICR_MAILBOX
)
2694 ixgbe_msg_task(adapter
);
2696 switch (hw
->mac
.type
) {
2697 case ixgbe_mac_82599EB
:
2698 case ixgbe_mac_X540
:
2699 case ixgbe_mac_X550
:
2700 case ixgbe_mac_X550EM_x
:
2701 if (hw
->phy
.type
== ixgbe_phy_x550em_ext_t
&&
2702 (eicr
& IXGBE_EICR_GPI_SDP0_X540
)) {
2703 adapter
->flags2
|= IXGBE_FLAG2_PHY_INTERRUPT
;
2704 ixgbe_service_event_schedule(adapter
);
2705 IXGBE_WRITE_REG(hw
, IXGBE_EICR
,
2706 IXGBE_EICR_GPI_SDP0_X540
);
2708 if (eicr
& IXGBE_EICR_ECC
) {
2709 e_info(link
, "Received ECC Err, initiating reset\n");
2710 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
2711 ixgbe_service_event_schedule(adapter
);
2712 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_ECC
);
2714 /* Handle Flow Director Full threshold interrupt */
2715 if (eicr
& IXGBE_EICR_FLOW_DIR
) {
2716 int reinit_count
= 0;
2718 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
2719 struct ixgbe_ring
*ring
= adapter
->tx_ring
[i
];
2720 if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE
,
2725 /* no more flow director interrupts until after init */
2726 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_EIMC_FLOW_DIR
);
2727 adapter
->flags2
|= IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
2728 ixgbe_service_event_schedule(adapter
);
2731 ixgbe_check_sfp_event(adapter
, eicr
);
2732 ixgbe_check_overtemp_event(adapter
, eicr
);
2738 ixgbe_check_fan_failure(adapter
, eicr
);
2740 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2741 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2743 /* re-enable the original interrupt state, no lsc, no queues */
2744 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2745 ixgbe_irq_enable(adapter
, false, false);
2750 static irqreturn_t
ixgbe_msix_clean_rings(int irq
, void *data
)
2752 struct ixgbe_q_vector
*q_vector
= data
;
2754 /* EIAM disabled interrupts (on this vector) for us */
2756 if (q_vector
->rx
.ring
|| q_vector
->tx
.ring
)
2757 napi_schedule(&q_vector
->napi
);
2763 * ixgbe_poll - NAPI Rx polling callback
2764 * @napi: structure for representing this polling device
2765 * @budget: how many packets driver is allowed to clean
2767 * This function is used for legacy and MSI, NAPI mode
2769 int ixgbe_poll(struct napi_struct
*napi
, int budget
)
2771 struct ixgbe_q_vector
*q_vector
=
2772 container_of(napi
, struct ixgbe_q_vector
, napi
);
2773 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
2774 struct ixgbe_ring
*ring
;
2775 int per_ring_budget
, work_done
= 0;
2776 bool clean_complete
= true;
2778 #ifdef CONFIG_IXGBE_DCA
2779 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
)
2780 ixgbe_update_dca(q_vector
);
2783 ixgbe_for_each_ring(ring
, q_vector
->tx
)
2784 clean_complete
&= !!ixgbe_clean_tx_irq(q_vector
, ring
);
2786 if (!ixgbe_qv_lock_napi(q_vector
))
2789 /* attempt to distribute budget to each queue fairly, but don't allow
2790 * the budget to go below 1 because we'll exit polling */
2791 if (q_vector
->rx
.count
> 1)
2792 per_ring_budget
= max(budget
/q_vector
->rx
.count
, 1);
2794 per_ring_budget
= budget
;
2796 ixgbe_for_each_ring(ring
, q_vector
->rx
) {
2797 int cleaned
= ixgbe_clean_rx_irq(q_vector
, ring
,
2800 work_done
+= cleaned
;
2801 clean_complete
&= (cleaned
< per_ring_budget
);
2804 ixgbe_qv_unlock_napi(q_vector
);
2805 /* If all work not completed, return budget and keep polling */
2806 if (!clean_complete
)
2809 /* all work done, exit the polling mode */
2810 napi_complete_done(napi
, work_done
);
2811 if (adapter
->rx_itr_setting
& 1)
2812 ixgbe_set_itr(q_vector
);
2813 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2814 ixgbe_irq_enable_queues(adapter
, ((u64
)1 << q_vector
->v_idx
));
2820 * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
2821 * @adapter: board private structure
2823 * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
2824 * interrupts from the kernel.
2826 static int ixgbe_request_msix_irqs(struct ixgbe_adapter
*adapter
)
2828 struct net_device
*netdev
= adapter
->netdev
;
2832 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2833 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
2834 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
2836 if (q_vector
->tx
.ring
&& q_vector
->rx
.ring
) {
2837 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2838 "%s-%s-%d", netdev
->name
, "TxRx", ri
++);
2840 } else if (q_vector
->rx
.ring
) {
2841 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2842 "%s-%s-%d", netdev
->name
, "rx", ri
++);
2843 } else if (q_vector
->tx
.ring
) {
2844 snprintf(q_vector
->name
, sizeof(q_vector
->name
) - 1,
2845 "%s-%s-%d", netdev
->name
, "tx", ti
++);
2847 /* skip this unused q_vector */
2850 err
= request_irq(entry
->vector
, &ixgbe_msix_clean_rings
, 0,
2851 q_vector
->name
, q_vector
);
2853 e_err(probe
, "request_irq failed for MSIX interrupt "
2854 "Error: %d\n", err
);
2855 goto free_queue_irqs
;
2857 /* If Flow Director is enabled, set interrupt affinity */
2858 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
2859 /* assign the mask for this irq */
2860 irq_set_affinity_hint(entry
->vector
,
2861 &q_vector
->affinity_mask
);
2865 err
= request_irq(adapter
->msix_entries
[vector
].vector
,
2866 ixgbe_msix_other
, 0, netdev
->name
, adapter
);
2868 e_err(probe
, "request_irq for msix_other failed: %d\n", err
);
2869 goto free_queue_irqs
;
2877 irq_set_affinity_hint(adapter
->msix_entries
[vector
].vector
,
2879 free_irq(adapter
->msix_entries
[vector
].vector
,
2880 adapter
->q_vector
[vector
]);
2882 adapter
->flags
&= ~IXGBE_FLAG_MSIX_ENABLED
;
2883 pci_disable_msix(adapter
->pdev
);
2884 kfree(adapter
->msix_entries
);
2885 adapter
->msix_entries
= NULL
;
2890 * ixgbe_intr - legacy mode Interrupt Handler
2891 * @irq: interrupt number
2892 * @data: pointer to a network interface device structure
2894 static irqreturn_t
ixgbe_intr(int irq
, void *data
)
2896 struct ixgbe_adapter
*adapter
= data
;
2897 struct ixgbe_hw
*hw
= &adapter
->hw
;
2898 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
2902 * Workaround for silicon errata #26 on 82598. Mask the interrupt
2903 * before the read of EICR.
2905 IXGBE_WRITE_REG(hw
, IXGBE_EIMC
, IXGBE_IRQ_CLEAR_MASK
);
2907 /* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
2908 * therefore no explicit interrupt disable is necessary */
2909 eicr
= IXGBE_READ_REG(hw
, IXGBE_EICR
);
2912 * shared interrupt alert!
2913 * make sure interrupts are enabled because the read will
2914 * have disabled interrupts due to EIAM
2915 * finish the workaround of silicon errata on 82598. Unmask
2916 * the interrupt that we masked before the EICR read.
2918 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2919 ixgbe_irq_enable(adapter
, true, true);
2920 return IRQ_NONE
; /* Not our interrupt */
2923 if (eicr
& IXGBE_EICR_LSC
)
2924 ixgbe_check_lsc(adapter
);
2926 switch (hw
->mac
.type
) {
2927 case ixgbe_mac_82599EB
:
2928 ixgbe_check_sfp_event(adapter
, eicr
);
2930 case ixgbe_mac_X540
:
2931 case ixgbe_mac_X550
:
2932 case ixgbe_mac_X550EM_x
:
2933 if (eicr
& IXGBE_EICR_ECC
) {
2934 e_info(link
, "Received ECC Err, initiating reset\n");
2935 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
2936 ixgbe_service_event_schedule(adapter
);
2937 IXGBE_WRITE_REG(hw
, IXGBE_EICR
, IXGBE_EICR_ECC
);
2939 ixgbe_check_overtemp_event(adapter
, eicr
);
2945 ixgbe_check_fan_failure(adapter
, eicr
);
2946 if (unlikely(eicr
& IXGBE_EICR_TIMESYNC
))
2947 ixgbe_ptp_check_pps_event(adapter
, eicr
);
2949 /* would disable interrupts here but EIAM disabled it */
2950 napi_schedule(&q_vector
->napi
);
2953 * re-enable link(maybe) and non-queue interrupts, no flush.
2954 * ixgbe_poll will re-enable the queue interrupts
2956 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
))
2957 ixgbe_irq_enable(adapter
, false, false);
2963 * ixgbe_request_irq - initialize interrupts
2964 * @adapter: board private structure
2966 * Attempts to configure interrupts using the best available
2967 * capabilities of the hardware and kernel.
2969 static int ixgbe_request_irq(struct ixgbe_adapter
*adapter
)
2971 struct net_device
*netdev
= adapter
->netdev
;
2974 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
2975 err
= ixgbe_request_msix_irqs(adapter
);
2976 else if (adapter
->flags
& IXGBE_FLAG_MSI_ENABLED
)
2977 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, 0,
2978 netdev
->name
, adapter
);
2980 err
= request_irq(adapter
->pdev
->irq
, ixgbe_intr
, IRQF_SHARED
,
2981 netdev
->name
, adapter
);
2984 e_err(probe
, "request_irq failed, Error %d\n", err
);
2989 static void ixgbe_free_irq(struct ixgbe_adapter
*adapter
)
2993 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
2994 free_irq(adapter
->pdev
->irq
, adapter
);
2998 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++) {
2999 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[vector
];
3000 struct msix_entry
*entry
= &adapter
->msix_entries
[vector
];
3002 /* free only the irqs that were actually requested */
3003 if (!q_vector
->rx
.ring
&& !q_vector
->tx
.ring
)
3006 /* clear the affinity_mask in the IRQ descriptor */
3007 irq_set_affinity_hint(entry
->vector
, NULL
);
3009 free_irq(entry
->vector
, q_vector
);
3012 free_irq(adapter
->msix_entries
[vector
++].vector
, adapter
);
3016 * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3017 * @adapter: board private structure
3019 static inline void ixgbe_irq_disable(struct ixgbe_adapter
*adapter
)
3021 switch (adapter
->hw
.mac
.type
) {
3022 case ixgbe_mac_82598EB
:
3023 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, ~0);
3025 case ixgbe_mac_82599EB
:
3026 case ixgbe_mac_X540
:
3027 case ixgbe_mac_X550
:
3028 case ixgbe_mac_X550EM_x
:
3029 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC
, 0xFFFF0000);
3030 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(0), ~0);
3031 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EIMC_EX(1), ~0);
3036 IXGBE_WRITE_FLUSH(&adapter
->hw
);
3037 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
3040 for (vector
= 0; vector
< adapter
->num_q_vectors
; vector
++)
3041 synchronize_irq(adapter
->msix_entries
[vector
].vector
);
3043 synchronize_irq(adapter
->msix_entries
[vector
++].vector
);
3045 synchronize_irq(adapter
->pdev
->irq
);
3050 * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3053 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter
*adapter
)
3055 struct ixgbe_q_vector
*q_vector
= adapter
->q_vector
[0];
3057 ixgbe_write_eitr(q_vector
);
3059 ixgbe_set_ivar(adapter
, 0, 0, 0);
3060 ixgbe_set_ivar(adapter
, 1, 0, 0);
3062 e_info(hw
, "Legacy interrupt IVAR setup done\n");
3066 * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3067 * @adapter: board private structure
3068 * @ring: structure containing ring specific data
3070 * Configure the Tx descriptor ring after a reset.
3072 void ixgbe_configure_tx_ring(struct ixgbe_adapter
*adapter
,
3073 struct ixgbe_ring
*ring
)
3075 struct ixgbe_hw
*hw
= &adapter
->hw
;
3076 u64 tdba
= ring
->dma
;
3078 u32 txdctl
= IXGBE_TXDCTL_ENABLE
;
3079 u8 reg_idx
= ring
->reg_idx
;
3081 /* disable queue to avoid issues while updating state */
3082 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), 0);
3083 IXGBE_WRITE_FLUSH(hw
);
3085 IXGBE_WRITE_REG(hw
, IXGBE_TDBAL(reg_idx
),
3086 (tdba
& DMA_BIT_MASK(32)));
3087 IXGBE_WRITE_REG(hw
, IXGBE_TDBAH(reg_idx
), (tdba
>> 32));
3088 IXGBE_WRITE_REG(hw
, IXGBE_TDLEN(reg_idx
),
3089 ring
->count
* sizeof(union ixgbe_adv_tx_desc
));
3090 IXGBE_WRITE_REG(hw
, IXGBE_TDH(reg_idx
), 0);
3091 IXGBE_WRITE_REG(hw
, IXGBE_TDT(reg_idx
), 0);
3092 ring
->tail
= adapter
->io_addr
+ IXGBE_TDT(reg_idx
);
3095 * set WTHRESH to encourage burst writeback, it should not be set
3096 * higher than 1 when:
3097 * - ITR is 0 as it could cause false TX hangs
3098 * - ITR is set to > 100k int/sec and BQL is enabled
3100 * In order to avoid issues WTHRESH + PTHRESH should always be equal
3101 * to or less than the number of on chip descriptors, which is
3104 if (!ring
->q_vector
|| (ring
->q_vector
->itr
< IXGBE_100K_ITR
))
3105 txdctl
|= (1 << 16); /* WTHRESH = 1 */
3107 txdctl
|= (8 << 16); /* WTHRESH = 8 */
3110 * Setting PTHRESH to 32 both improves performance
3111 * and avoids a TX hang with DFP enabled
3113 txdctl
|= (1 << 8) | /* HTHRESH = 1 */
3114 32; /* PTHRESH = 32 */
3116 /* reinitialize flowdirector state */
3117 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
3118 ring
->atr_sample_rate
= adapter
->atr_sample_rate
;
3119 ring
->atr_count
= 0;
3120 set_bit(__IXGBE_TX_FDIR_INIT_DONE
, &ring
->state
);
3122 ring
->atr_sample_rate
= 0;
3125 /* initialize XPS */
3126 if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE
, &ring
->state
)) {
3127 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
3130 netif_set_xps_queue(ring
->netdev
,
3131 &q_vector
->affinity_mask
,
3135 clear_bit(__IXGBE_HANG_CHECK_ARMED
, &ring
->state
);
3138 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), txdctl
);
3140 /* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3141 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3142 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3145 /* poll to verify queue is enabled */
3147 usleep_range(1000, 2000);
3148 txdctl
= IXGBE_READ_REG(hw
, IXGBE_TXDCTL(reg_idx
));
3149 } while (--wait_loop
&& !(txdctl
& IXGBE_TXDCTL_ENABLE
));
3151 e_err(drv
, "Could not enable Tx Queue %d\n", reg_idx
);
3154 static void ixgbe_setup_mtqc(struct ixgbe_adapter
*adapter
)
3156 struct ixgbe_hw
*hw
= &adapter
->hw
;
3158 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
3160 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3163 /* disable the arbiter while setting MTQC */
3164 rttdcs
= IXGBE_READ_REG(hw
, IXGBE_RTTDCS
);
3165 rttdcs
|= IXGBE_RTTDCS_ARBDIS
;
3166 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
3168 /* set transmit pool layout */
3169 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
3170 mtqc
= IXGBE_MTQC_VT_ENA
;
3172 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
3174 mtqc
|= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
3175 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
3176 mtqc
|= IXGBE_MTQC_32VF
;
3178 mtqc
|= IXGBE_MTQC_64VF
;
3181 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_8TC_8TQ
;
3183 mtqc
= IXGBE_MTQC_RT_ENA
| IXGBE_MTQC_4TC_4TQ
;
3185 mtqc
= IXGBE_MTQC_64Q_1PB
;
3188 IXGBE_WRITE_REG(hw
, IXGBE_MTQC
, mtqc
);
3190 /* Enable Security TX Buffer IFG for multiple pb */
3192 u32 sectx
= IXGBE_READ_REG(hw
, IXGBE_SECTXMINIFG
);
3193 sectx
|= IXGBE_SECTX_DCB
;
3194 IXGBE_WRITE_REG(hw
, IXGBE_SECTXMINIFG
, sectx
);
3197 /* re-enable the arbiter */
3198 rttdcs
&= ~IXGBE_RTTDCS_ARBDIS
;
3199 IXGBE_WRITE_REG(hw
, IXGBE_RTTDCS
, rttdcs
);
3203 * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3204 * @adapter: board private structure
3206 * Configure the Tx unit of the MAC after a reset.
3208 static void ixgbe_configure_tx(struct ixgbe_adapter
*adapter
)
3210 struct ixgbe_hw
*hw
= &adapter
->hw
;
3214 ixgbe_setup_mtqc(adapter
);
3216 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
3217 /* DMATXCTL.EN must be before Tx queues are enabled */
3218 dmatxctl
= IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
);
3219 dmatxctl
|= IXGBE_DMATXCTL_TE
;
3220 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
, dmatxctl
);
3223 /* Setup the HW Tx Head and Tail descriptor pointers */
3224 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
3225 ixgbe_configure_tx_ring(adapter
, adapter
->tx_ring
[i
]);
3228 static void ixgbe_enable_rx_drop(struct ixgbe_adapter
*adapter
,
3229 struct ixgbe_ring
*ring
)
3231 struct ixgbe_hw
*hw
= &adapter
->hw
;
3232 u8 reg_idx
= ring
->reg_idx
;
3233 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
3235 srrctl
|= IXGBE_SRRCTL_DROP_EN
;
3237 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3240 static void ixgbe_disable_rx_drop(struct ixgbe_adapter
*adapter
,
3241 struct ixgbe_ring
*ring
)
3243 struct ixgbe_hw
*hw
= &adapter
->hw
;
3244 u8 reg_idx
= ring
->reg_idx
;
3245 u32 srrctl
= IXGBE_READ_REG(hw
, IXGBE_SRRCTL(reg_idx
));
3247 srrctl
&= ~IXGBE_SRRCTL_DROP_EN
;
3249 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3252 #ifdef CONFIG_IXGBE_DCB
3253 void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
3255 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter
*adapter
)
3259 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
3261 if (adapter
->ixgbe_ieee_pfc
)
3262 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
3265 * We should set the drop enable bit if:
3268 * Number of Rx queues > 1 and flow control is disabled
3270 * This allows us to avoid head of line blocking for security
3271 * and performance reasons.
3273 if (adapter
->num_vfs
|| (adapter
->num_rx_queues
> 1 &&
3274 !(adapter
->hw
.fc
.current_mode
& ixgbe_fc_tx_pause
) && !pfc_en
)) {
3275 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3276 ixgbe_enable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3278 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3279 ixgbe_disable_rx_drop(adapter
, adapter
->rx_ring
[i
]);
3283 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3285 static void ixgbe_configure_srrctl(struct ixgbe_adapter
*adapter
,
3286 struct ixgbe_ring
*rx_ring
)
3288 struct ixgbe_hw
*hw
= &adapter
->hw
;
3290 u8 reg_idx
= rx_ring
->reg_idx
;
3292 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3293 u16 mask
= adapter
->ring_feature
[RING_F_RSS
].mask
;
3296 * if VMDq is not active we must program one srrctl register
3297 * per RSS queue since we have enabled RDRXCTL.MVMEN
3302 /* configure header buffer length, needed for RSC */
3303 srrctl
= IXGBE_RX_HDR_SIZE
<< IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT
;
3305 /* configure the packet buffer length */
3306 srrctl
|= ixgbe_rx_bufsz(rx_ring
) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT
;
3308 /* configure descriptor type */
3309 srrctl
|= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF
;
3311 IXGBE_WRITE_REG(hw
, IXGBE_SRRCTL(reg_idx
), srrctl
);
3315 * Return a number of entries in the RSS indirection table
3317 * @adapter: device handle
3319 * - 82598/82599/X540: 128
3320 * - X550(non-SRIOV mode): 512
3321 * - X550(SRIOV mode): 64
3323 u32
ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter
*adapter
)
3325 if (adapter
->hw
.mac
.type
< ixgbe_mac_X550
)
3327 else if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
3334 * Write the RETA table to HW
3336 * @adapter: device handle
3338 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3340 void ixgbe_store_reta(struct ixgbe_adapter
*adapter
)
3342 u32 i
, reta_entries
= ixgbe_rss_indir_tbl_entries(adapter
);
3343 struct ixgbe_hw
*hw
= &adapter
->hw
;
3346 u8
*indir_tbl
= adapter
->rss_indir_tbl
;
3348 /* Fill out the redirection table as follows:
3349 * - 82598: 8 bit wide entries containing pair of 4 bit RSS
3351 * - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3352 * - X550: 8 bit wide entries containing 6 bit RSS index
3354 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
3355 indices_multi
= 0x11;
3357 indices_multi
= 0x1;
3359 /* Write redirection table to HW */
3360 for (i
= 0; i
< reta_entries
; i
++) {
3361 reta
|= indices_multi
* indir_tbl
[i
] << (i
& 0x3) * 8;
3364 IXGBE_WRITE_REG(hw
, IXGBE_RETA(i
>> 2), reta
);
3366 IXGBE_WRITE_REG(hw
, IXGBE_ERETA((i
>> 2) - 32),
3374 * Write the RETA table to HW (for x550 devices in SRIOV mode)
3376 * @adapter: device handle
3378 * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3380 static void ixgbe_store_vfreta(struct ixgbe_adapter
*adapter
)
3382 u32 i
, reta_entries
= ixgbe_rss_indir_tbl_entries(adapter
);
3383 struct ixgbe_hw
*hw
= &adapter
->hw
;
3385 unsigned int pf_pool
= adapter
->num_vfs
;
3387 /* Write redirection table to HW */
3388 for (i
= 0; i
< reta_entries
; i
++) {
3389 vfreta
|= (u32
)adapter
->rss_indir_tbl
[i
] << (i
& 0x3) * 8;
3391 IXGBE_WRITE_REG(hw
, IXGBE_PFVFRETA(i
>> 2, pf_pool
),
3398 static void ixgbe_setup_reta(struct ixgbe_adapter
*adapter
)
3400 struct ixgbe_hw
*hw
= &adapter
->hw
;
3402 u32 reta_entries
= ixgbe_rss_indir_tbl_entries(adapter
);
3403 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3405 /* Program table for at least 2 queues w/ SR-IOV so that VFs can
3406 * make full use of any rings they may have. We will use the
3407 * PSRTYPE register to control how many rings we use within the PF.
3409 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) && (rss_i
< 2))
3412 /* Fill out hash function seeds */
3413 for (i
= 0; i
< 10; i
++)
3414 IXGBE_WRITE_REG(hw
, IXGBE_RSSRK(i
), adapter
->rss_key
[i
]);
3416 /* Fill out redirection table */
3417 memset(adapter
->rss_indir_tbl
, 0, sizeof(adapter
->rss_indir_tbl
));
3419 for (i
= 0, j
= 0; i
< reta_entries
; i
++, j
++) {
3423 adapter
->rss_indir_tbl
[i
] = j
;
3426 ixgbe_store_reta(adapter
);
3429 static void ixgbe_setup_vfreta(struct ixgbe_adapter
*adapter
)
3431 struct ixgbe_hw
*hw
= &adapter
->hw
;
3432 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3433 unsigned int pf_pool
= adapter
->num_vfs
;
3436 /* Fill out hash function seeds */
3437 for (i
= 0; i
< 10; i
++)
3438 IXGBE_WRITE_REG(hw
, IXGBE_PFVFRSSRK(i
, pf_pool
),
3439 adapter
->rss_key
[i
]);
3441 /* Fill out the redirection table */
3442 for (i
= 0, j
= 0; i
< 64; i
++, j
++) {
3446 adapter
->rss_indir_tbl
[i
] = j
;
3449 ixgbe_store_vfreta(adapter
);
3452 static void ixgbe_setup_mrqc(struct ixgbe_adapter
*adapter
)
3454 struct ixgbe_hw
*hw
= &adapter
->hw
;
3455 u32 mrqc
= 0, rss_field
= 0, vfmrqc
= 0;
3458 /* Disable indicating checksum in descriptor, enables RSS hash */
3459 rxcsum
= IXGBE_READ_REG(hw
, IXGBE_RXCSUM
);
3460 rxcsum
|= IXGBE_RXCSUM_PCSD
;
3461 IXGBE_WRITE_REG(hw
, IXGBE_RXCSUM
, rxcsum
);
3463 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
3464 if (adapter
->ring_feature
[RING_F_RSS
].mask
)
3465 mrqc
= IXGBE_MRQC_RSSEN
;
3467 u8 tcs
= netdev_get_num_tc(adapter
->netdev
);
3469 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
3471 mrqc
= IXGBE_MRQC_VMDQRT8TCEN
; /* 8 TCs */
3473 mrqc
= IXGBE_MRQC_VMDQRT4TCEN
; /* 4 TCs */
3474 else if (adapter
->ring_feature
[RING_F_RSS
].indices
== 4)
3475 mrqc
= IXGBE_MRQC_VMDQRSS32EN
;
3477 mrqc
= IXGBE_MRQC_VMDQRSS64EN
;
3480 mrqc
= IXGBE_MRQC_RTRSS8TCEN
;
3482 mrqc
= IXGBE_MRQC_RTRSS4TCEN
;
3484 mrqc
= IXGBE_MRQC_RSSEN
;
3488 /* Perform hash on these packet types */
3489 rss_field
|= IXGBE_MRQC_RSS_FIELD_IPV4
|
3490 IXGBE_MRQC_RSS_FIELD_IPV4_TCP
|
3491 IXGBE_MRQC_RSS_FIELD_IPV6
|
3492 IXGBE_MRQC_RSS_FIELD_IPV6_TCP
;
3494 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV4_UDP
)
3495 rss_field
|= IXGBE_MRQC_RSS_FIELD_IPV4_UDP
;
3496 if (adapter
->flags2
& IXGBE_FLAG2_RSS_FIELD_IPV6_UDP
)
3497 rss_field
|= IXGBE_MRQC_RSS_FIELD_IPV6_UDP
;
3499 netdev_rss_key_fill(adapter
->rss_key
, sizeof(adapter
->rss_key
));
3500 if ((hw
->mac
.type
>= ixgbe_mac_X550
) &&
3501 (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)) {
3502 unsigned int pf_pool
= adapter
->num_vfs
;
3504 /* Enable VF RSS mode */
3505 mrqc
|= IXGBE_MRQC_MULTIPLE_RSS
;
3506 IXGBE_WRITE_REG(hw
, IXGBE_MRQC
, mrqc
);
3508 /* Setup RSS through the VF registers */
3509 ixgbe_setup_vfreta(adapter
);
3510 vfmrqc
= IXGBE_MRQC_RSSEN
;
3511 vfmrqc
|= rss_field
;
3512 IXGBE_WRITE_REG(hw
, IXGBE_PFVFMRQC(pf_pool
), vfmrqc
);
3514 ixgbe_setup_reta(adapter
);
3516 IXGBE_WRITE_REG(hw
, IXGBE_MRQC
, mrqc
);
3521 * ixgbe_configure_rscctl - enable RSC for the indicated ring
3522 * @adapter: address of board private structure
3523 * @index: index of ring to set
3525 static void ixgbe_configure_rscctl(struct ixgbe_adapter
*adapter
,
3526 struct ixgbe_ring
*ring
)
3528 struct ixgbe_hw
*hw
= &adapter
->hw
;
3530 u8 reg_idx
= ring
->reg_idx
;
3532 if (!ring_is_rsc_enabled(ring
))
3535 rscctrl
= IXGBE_READ_REG(hw
, IXGBE_RSCCTL(reg_idx
));
3536 rscctrl
|= IXGBE_RSCCTL_RSCEN
;
3538 * we must limit the number of descriptors so that the
3539 * total size of max desc * buf_len is not greater
3542 rscctrl
|= IXGBE_RSCCTL_MAXDESC_16
;
3543 IXGBE_WRITE_REG(hw
, IXGBE_RSCCTL(reg_idx
), rscctrl
);
3546 #define IXGBE_MAX_RX_DESC_POLL 10
3547 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter
*adapter
,
3548 struct ixgbe_ring
*ring
)
3550 struct ixgbe_hw
*hw
= &adapter
->hw
;
3551 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3553 u8 reg_idx
= ring
->reg_idx
;
3555 if (ixgbe_removed(hw
->hw_addr
))
3557 /* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3558 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3559 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3563 usleep_range(1000, 2000);
3564 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3565 } while (--wait_loop
&& !(rxdctl
& IXGBE_RXDCTL_ENABLE
));
3568 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not set within "
3569 "the polling period\n", reg_idx
);
3573 void ixgbe_disable_rx_queue(struct ixgbe_adapter
*adapter
,
3574 struct ixgbe_ring
*ring
)
3576 struct ixgbe_hw
*hw
= &adapter
->hw
;
3577 int wait_loop
= IXGBE_MAX_RX_DESC_POLL
;
3579 u8 reg_idx
= ring
->reg_idx
;
3581 if (ixgbe_removed(hw
->hw_addr
))
3583 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3584 rxdctl
&= ~IXGBE_RXDCTL_ENABLE
;
3586 /* write value back with RXDCTL.ENABLE bit cleared */
3587 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3589 if (hw
->mac
.type
== ixgbe_mac_82598EB
&&
3590 !(IXGBE_READ_REG(hw
, IXGBE_LINKS
) & IXGBE_LINKS_UP
))
3593 /* the hardware may take up to 100us to really disable the rx queue */
3596 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3597 } while (--wait_loop
&& (rxdctl
& IXGBE_RXDCTL_ENABLE
));
3600 e_err(drv
, "RXDCTL.ENABLE on Rx queue %d not cleared within "
3601 "the polling period\n", reg_idx
);
3605 void ixgbe_configure_rx_ring(struct ixgbe_adapter
*adapter
,
3606 struct ixgbe_ring
*ring
)
3608 struct ixgbe_hw
*hw
= &adapter
->hw
;
3609 u64 rdba
= ring
->dma
;
3611 u8 reg_idx
= ring
->reg_idx
;
3613 /* disable queue to avoid issues while updating state */
3614 rxdctl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(reg_idx
));
3615 ixgbe_disable_rx_queue(adapter
, ring
);
3617 IXGBE_WRITE_REG(hw
, IXGBE_RDBAL(reg_idx
), (rdba
& DMA_BIT_MASK(32)));
3618 IXGBE_WRITE_REG(hw
, IXGBE_RDBAH(reg_idx
), (rdba
>> 32));
3619 IXGBE_WRITE_REG(hw
, IXGBE_RDLEN(reg_idx
),
3620 ring
->count
* sizeof(union ixgbe_adv_rx_desc
));
3621 IXGBE_WRITE_REG(hw
, IXGBE_RDH(reg_idx
), 0);
3622 IXGBE_WRITE_REG(hw
, IXGBE_RDT(reg_idx
), 0);
3623 ring
->tail
= adapter
->io_addr
+ IXGBE_RDT(reg_idx
);
3625 ixgbe_configure_srrctl(adapter
, ring
);
3626 ixgbe_configure_rscctl(adapter
, ring
);
3628 if (hw
->mac
.type
== ixgbe_mac_82598EB
) {
3630 * enable cache line friendly hardware writes:
3631 * PTHRESH=32 descriptors (half the internal cache),
3632 * this also removes ugly rx_no_buffer_count increment
3633 * HTHRESH=4 descriptors (to minimize latency on fetch)
3634 * WTHRESH=8 burst writeback up to two cache lines
3636 rxdctl
&= ~0x3FFFFF;
3640 /* enable receive descriptor ring */
3641 rxdctl
|= IXGBE_RXDCTL_ENABLE
;
3642 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(reg_idx
), rxdctl
);
3644 ixgbe_rx_desc_queue_enable(adapter
, ring
);
3645 ixgbe_alloc_rx_buffers(ring
, ixgbe_desc_unused(ring
));
3648 static void ixgbe_setup_psrtype(struct ixgbe_adapter
*adapter
)
3650 struct ixgbe_hw
*hw
= &adapter
->hw
;
3651 int rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
;
3654 /* PSRTYPE must be initialized in non 82598 adapters */
3655 u32 psrtype
= IXGBE_PSRTYPE_TCPHDR
|
3656 IXGBE_PSRTYPE_UDPHDR
|
3657 IXGBE_PSRTYPE_IPV4HDR
|
3658 IXGBE_PSRTYPE_L2HDR
|
3659 IXGBE_PSRTYPE_IPV6HDR
;
3661 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3669 for_each_set_bit(pool
, &adapter
->fwd_bitmask
, 32)
3670 IXGBE_WRITE_REG(hw
, IXGBE_PSRTYPE(VMDQ_P(pool
)), psrtype
);
3673 static void ixgbe_configure_virtualization(struct ixgbe_adapter
*adapter
)
3675 struct ixgbe_hw
*hw
= &adapter
->hw
;
3676 u32 reg_offset
, vf_shift
;
3677 u32 gcr_ext
, vmdctl
;
3680 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
3683 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VT_CTL
);
3684 vmdctl
|= IXGBE_VMD_CTL_VMDQ_EN
;
3685 vmdctl
&= ~IXGBE_VT_CTL_POOL_MASK
;
3686 vmdctl
|= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT
;
3687 vmdctl
|= IXGBE_VT_CTL_REPLEN
;
3688 IXGBE_WRITE_REG(hw
, IXGBE_VT_CTL
, vmdctl
);
3690 vf_shift
= VMDQ_P(0) % 32;
3691 reg_offset
= (VMDQ_P(0) >= 32) ? 1 : 0;
3693 /* Enable only the PF's pool for Tx/Rx */
3694 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
), (~0) << vf_shift
);
3695 IXGBE_WRITE_REG(hw
, IXGBE_VFRE(reg_offset
^ 1), reg_offset
- 1);
3696 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
), (~0) << vf_shift
);
3697 IXGBE_WRITE_REG(hw
, IXGBE_VFTE(reg_offset
^ 1), reg_offset
- 1);
3698 if (adapter
->bridge_mode
== BRIDGE_MODE_VEB
)
3699 IXGBE_WRITE_REG(hw
, IXGBE_PFDTXGSWC
, IXGBE_PFDTXGSWC_VT_LBEN
);
3701 /* Map PF MAC address in RAR Entry 0 to first pool following VFs */
3702 hw
->mac
.ops
.set_vmdq(hw
, 0, VMDQ_P(0));
3705 * Set up VF register offsets for selected VT Mode,
3706 * i.e. 32 or 64 VFs for SR-IOV
3708 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
3709 case IXGBE_82599_VMDQ_8Q_MASK
:
3710 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_16
;
3712 case IXGBE_82599_VMDQ_4Q_MASK
:
3713 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_32
;
3716 gcr_ext
= IXGBE_GCR_EXT_VT_MODE_64
;
3720 IXGBE_WRITE_REG(hw
, IXGBE_GCR_EXT
, gcr_ext
);
3723 /* Enable MAC Anti-Spoofing */
3724 hw
->mac
.ops
.set_mac_anti_spoofing(hw
, (adapter
->num_vfs
!= 0),
3727 /* Ensure LLDP and FC is set for Ethertype Antispoofing if we will be
3728 * calling set_ethertype_anti_spoofing for each VF in loop below
3730 if (hw
->mac
.ops
.set_ethertype_anti_spoofing
) {
3731 IXGBE_WRITE_REG(hw
, IXGBE_ETQF(IXGBE_ETQF_FILTER_LLDP
),
3732 (IXGBE_ETQF_FILTER_EN
|
3733 IXGBE_ETQF_TX_ANTISPOOF
|
3736 IXGBE_WRITE_REG(hw
, IXGBE_ETQF(IXGBE_ETQF_FILTER_FC
),
3737 (IXGBE_ETQF_FILTER_EN
|
3738 IXGBE_ETQF_TX_ANTISPOOF
|
3742 /* For VFs that have spoof checking turned off */
3743 for (i
= 0; i
< adapter
->num_vfs
; i
++) {
3744 if (!adapter
->vfinfo
[i
].spoofchk_enabled
)
3745 ixgbe_ndo_set_vf_spoofchk(adapter
->netdev
, i
, false);
3747 /* enable ethertype anti spoofing if hw supports it */
3748 if (hw
->mac
.ops
.set_ethertype_anti_spoofing
)
3749 hw
->mac
.ops
.set_ethertype_anti_spoofing(hw
, true, i
);
3751 /* Enable/Disable RSS query feature */
3752 ixgbe_ndo_set_vf_rss_query_en(adapter
->netdev
, i
,
3753 adapter
->vfinfo
[i
].rss_query_enabled
);
3757 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter
*adapter
)
3759 struct ixgbe_hw
*hw
= &adapter
->hw
;
3760 struct net_device
*netdev
= adapter
->netdev
;
3761 int max_frame
= netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
3762 struct ixgbe_ring
*rx_ring
;
3767 /* adjust max frame to be able to do baby jumbo for FCoE */
3768 if ((adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
) &&
3769 (max_frame
< IXGBE_FCOE_JUMBO_FRAME_SIZE
))
3770 max_frame
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
3772 #endif /* IXGBE_FCOE */
3774 /* adjust max frame to be at least the size of a standard frame */
3775 if (max_frame
< (ETH_FRAME_LEN
+ ETH_FCS_LEN
))
3776 max_frame
= (ETH_FRAME_LEN
+ ETH_FCS_LEN
);
3778 mhadd
= IXGBE_READ_REG(hw
, IXGBE_MHADD
);
3779 if (max_frame
!= (mhadd
>> IXGBE_MHADD_MFS_SHIFT
)) {
3780 mhadd
&= ~IXGBE_MHADD_MFS_MASK
;
3781 mhadd
|= max_frame
<< IXGBE_MHADD_MFS_SHIFT
;
3783 IXGBE_WRITE_REG(hw
, IXGBE_MHADD
, mhadd
);
3786 hlreg0
= IXGBE_READ_REG(hw
, IXGBE_HLREG0
);
3787 /* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
3788 hlreg0
|= IXGBE_HLREG0_JUMBOEN
;
3789 IXGBE_WRITE_REG(hw
, IXGBE_HLREG0
, hlreg0
);
3792 * Setup the HW Rx Head and Tail Descriptor Pointers and
3793 * the Base and Length of the Rx Descriptor Ring
3795 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3796 rx_ring
= adapter
->rx_ring
[i
];
3797 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
3798 set_ring_rsc_enabled(rx_ring
);
3800 clear_ring_rsc_enabled(rx_ring
);
3804 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter
*adapter
)
3806 struct ixgbe_hw
*hw
= &adapter
->hw
;
3807 u32 rdrxctl
= IXGBE_READ_REG(hw
, IXGBE_RDRXCTL
);
3809 switch (hw
->mac
.type
) {
3810 case ixgbe_mac_82598EB
:
3812 * For VMDq support of different descriptor types or
3813 * buffer sizes through the use of multiple SRRCTL
3814 * registers, RDRXCTL.MVMEN must be set to 1
3816 * also, the manual doesn't mention it clearly but DCA hints
3817 * will only use queue 0's tags unless this bit is set. Side
3818 * effects of setting this bit are only that SRRCTL must be
3819 * fully programmed [0..15]
3821 rdrxctl
|= IXGBE_RDRXCTL_MVMEN
;
3823 case ixgbe_mac_X550
:
3824 case ixgbe_mac_X550EM_x
:
3825 if (adapter
->num_vfs
)
3826 rdrxctl
|= IXGBE_RDRXCTL_PSP
;
3827 /* fall through for older HW */
3828 case ixgbe_mac_82599EB
:
3829 case ixgbe_mac_X540
:
3830 /* Disable RSC for ACK packets */
3831 IXGBE_WRITE_REG(hw
, IXGBE_RSCDBU
,
3832 (IXGBE_RSCDBU_RSCACKDIS
| IXGBE_READ_REG(hw
, IXGBE_RSCDBU
)));
3833 rdrxctl
&= ~IXGBE_RDRXCTL_RSCFRSTSIZE
;
3834 /* hardware requires some bits to be set by default */
3835 rdrxctl
|= (IXGBE_RDRXCTL_RSCACKC
| IXGBE_RDRXCTL_FCOE_WRFIX
);
3836 rdrxctl
|= IXGBE_RDRXCTL_CRCSTRIP
;
3839 /* We should do nothing since we don't know this hardware */
3843 IXGBE_WRITE_REG(hw
, IXGBE_RDRXCTL
, rdrxctl
);
3847 * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
3848 * @adapter: board private structure
3850 * Configure the Rx unit of the MAC after a reset.
3852 static void ixgbe_configure_rx(struct ixgbe_adapter
*adapter
)
3854 struct ixgbe_hw
*hw
= &adapter
->hw
;
3858 /* disable receives while setting up the descriptors */
3859 hw
->mac
.ops
.disable_rx(hw
);
3861 ixgbe_setup_psrtype(adapter
);
3862 ixgbe_setup_rdrxctl(adapter
);
3865 rfctl
= IXGBE_READ_REG(hw
, IXGBE_RFCTL
);
3866 rfctl
&= ~IXGBE_RFCTL_RSC_DIS
;
3867 if (!(adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
))
3868 rfctl
|= IXGBE_RFCTL_RSC_DIS
;
3869 IXGBE_WRITE_REG(hw
, IXGBE_RFCTL
, rfctl
);
3871 /* Program registers for the distribution of queues */
3872 ixgbe_setup_mrqc(adapter
);
3874 /* set_rx_buffer_len must be called before ring initialization */
3875 ixgbe_set_rx_buffer_len(adapter
);
3878 * Setup the HW Rx Head and Tail Descriptor Pointers and
3879 * the Base and Length of the Rx Descriptor Ring
3881 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
3882 ixgbe_configure_rx_ring(adapter
, adapter
->rx_ring
[i
]);
3884 rxctrl
= IXGBE_READ_REG(hw
, IXGBE_RXCTRL
);
3885 /* disable drop enable for 82598 parts */
3886 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
3887 rxctrl
|= IXGBE_RXCTRL_DMBYPS
;
3889 /* enable all receives */
3890 rxctrl
|= IXGBE_RXCTRL_RXEN
;
3891 hw
->mac
.ops
.enable_rx_dma(hw
, rxctrl
);
3894 static int ixgbe_vlan_rx_add_vid(struct net_device
*netdev
,
3895 __be16 proto
, u16 vid
)
3897 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3898 struct ixgbe_hw
*hw
= &adapter
->hw
;
3900 /* add VID to filter table */
3901 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), true);
3902 set_bit(vid
, adapter
->active_vlans
);
3907 static int ixgbe_vlan_rx_kill_vid(struct net_device
*netdev
,
3908 __be16 proto
, u16 vid
)
3910 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
3911 struct ixgbe_hw
*hw
= &adapter
->hw
;
3913 /* remove VID from filter table */
3914 hw
->mac
.ops
.set_vfta(&adapter
->hw
, vid
, VMDQ_P(0), false);
3915 clear_bit(vid
, adapter
->active_vlans
);
3921 * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
3922 * @adapter: driver data
3924 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter
*adapter
)
3926 struct ixgbe_hw
*hw
= &adapter
->hw
;
3930 switch (hw
->mac
.type
) {
3931 case ixgbe_mac_82598EB
:
3932 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3933 vlnctrl
&= ~IXGBE_VLNCTRL_VME
;
3934 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3936 case ixgbe_mac_82599EB
:
3937 case ixgbe_mac_X540
:
3938 case ixgbe_mac_X550
:
3939 case ixgbe_mac_X550EM_x
:
3940 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3941 struct ixgbe_ring
*ring
= adapter
->rx_ring
[i
];
3943 if (ring
->l2_accel_priv
)
3946 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3947 vlnctrl
&= ~IXGBE_RXDCTL_VME
;
3948 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3957 * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
3958 * @adapter: driver data
3960 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter
*adapter
)
3962 struct ixgbe_hw
*hw
= &adapter
->hw
;
3966 switch (hw
->mac
.type
) {
3967 case ixgbe_mac_82598EB
:
3968 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
3969 vlnctrl
|= IXGBE_VLNCTRL_VME
;
3970 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
3972 case ixgbe_mac_82599EB
:
3973 case ixgbe_mac_X540
:
3974 case ixgbe_mac_X550
:
3975 case ixgbe_mac_X550EM_x
:
3976 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
3977 struct ixgbe_ring
*ring
= adapter
->rx_ring
[i
];
3979 if (ring
->l2_accel_priv
)
3982 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_RXDCTL(j
));
3983 vlnctrl
|= IXGBE_RXDCTL_VME
;
3984 IXGBE_WRITE_REG(hw
, IXGBE_RXDCTL(j
), vlnctrl
);
3992 static void ixgbe_restore_vlan(struct ixgbe_adapter
*adapter
)
3996 ixgbe_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), 0);
3998 for_each_set_bit(vid
, adapter
->active_vlans
, VLAN_N_VID
)
3999 ixgbe_vlan_rx_add_vid(adapter
->netdev
, htons(ETH_P_8021Q
), vid
);
4003 * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4004 * @netdev: network interface device structure
4006 * Writes multicast address list to the MTA hash table.
4007 * Returns: -ENOMEM on failure
4008 * 0 on no addresses written
4009 * X on writing X addresses to MTA
4011 static int ixgbe_write_mc_addr_list(struct net_device
*netdev
)
4013 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4014 struct ixgbe_hw
*hw
= &adapter
->hw
;
4016 if (!netif_running(netdev
))
4019 if (hw
->mac
.ops
.update_mc_addr_list
)
4020 hw
->mac
.ops
.update_mc_addr_list(hw
, netdev
);
4024 #ifdef CONFIG_PCI_IOV
4025 ixgbe_restore_vf_multicasts(adapter
);
4028 return netdev_mc_count(netdev
);
4031 #ifdef CONFIG_PCI_IOV
4032 void ixgbe_full_sync_mac_table(struct ixgbe_adapter
*adapter
)
4034 struct ixgbe_hw
*hw
= &adapter
->hw
;
4036 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4037 if (adapter
->mac_table
[i
].state
& IXGBE_MAC_STATE_IN_USE
)
4038 hw
->mac
.ops
.set_rar(hw
, i
, adapter
->mac_table
[i
].addr
,
4039 adapter
->mac_table
[i
].queue
,
4042 hw
->mac
.ops
.clear_rar(hw
, i
);
4044 adapter
->mac_table
[i
].state
&= ~(IXGBE_MAC_STATE_MODIFIED
);
4049 static void ixgbe_sync_mac_table(struct ixgbe_adapter
*adapter
)
4051 struct ixgbe_hw
*hw
= &adapter
->hw
;
4053 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4054 if (adapter
->mac_table
[i
].state
& IXGBE_MAC_STATE_MODIFIED
) {
4055 if (adapter
->mac_table
[i
].state
&
4056 IXGBE_MAC_STATE_IN_USE
)
4057 hw
->mac
.ops
.set_rar(hw
, i
,
4058 adapter
->mac_table
[i
].addr
,
4059 adapter
->mac_table
[i
].queue
,
4062 hw
->mac
.ops
.clear_rar(hw
, i
);
4064 adapter
->mac_table
[i
].state
&=
4065 ~(IXGBE_MAC_STATE_MODIFIED
);
4070 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter
*adapter
)
4073 struct ixgbe_hw
*hw
= &adapter
->hw
;
4075 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4076 adapter
->mac_table
[i
].state
|= IXGBE_MAC_STATE_MODIFIED
;
4077 adapter
->mac_table
[i
].state
&= ~IXGBE_MAC_STATE_IN_USE
;
4078 eth_zero_addr(adapter
->mac_table
[i
].addr
);
4079 adapter
->mac_table
[i
].queue
= 0;
4081 ixgbe_sync_mac_table(adapter
);
4084 static int ixgbe_available_rars(struct ixgbe_adapter
*adapter
)
4086 struct ixgbe_hw
*hw
= &adapter
->hw
;
4089 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4090 if (adapter
->mac_table
[i
].state
== 0)
4096 /* this function destroys the first RAR entry */
4097 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter
*adapter
,
4100 struct ixgbe_hw
*hw
= &adapter
->hw
;
4102 memcpy(&adapter
->mac_table
[0].addr
, addr
, ETH_ALEN
);
4103 adapter
->mac_table
[0].queue
= VMDQ_P(0);
4104 adapter
->mac_table
[0].state
= (IXGBE_MAC_STATE_DEFAULT
|
4105 IXGBE_MAC_STATE_IN_USE
);
4106 hw
->mac
.ops
.set_rar(hw
, 0, adapter
->mac_table
[0].addr
,
4107 adapter
->mac_table
[0].queue
,
4111 int ixgbe_add_mac_filter(struct ixgbe_adapter
*adapter
, u8
*addr
, u16 queue
)
4113 struct ixgbe_hw
*hw
= &adapter
->hw
;
4116 if (is_zero_ether_addr(addr
))
4119 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4120 if (adapter
->mac_table
[i
].state
& IXGBE_MAC_STATE_IN_USE
)
4122 adapter
->mac_table
[i
].state
|= (IXGBE_MAC_STATE_MODIFIED
|
4123 IXGBE_MAC_STATE_IN_USE
);
4124 ether_addr_copy(adapter
->mac_table
[i
].addr
, addr
);
4125 adapter
->mac_table
[i
].queue
= queue
;
4126 ixgbe_sync_mac_table(adapter
);
4132 int ixgbe_del_mac_filter(struct ixgbe_adapter
*adapter
, u8
*addr
, u16 queue
)
4134 /* search table for addr, if found, set to 0 and sync */
4136 struct ixgbe_hw
*hw
= &adapter
->hw
;
4138 if (is_zero_ether_addr(addr
))
4141 for (i
= 0; i
< hw
->mac
.num_rar_entries
; i
++) {
4142 if (ether_addr_equal(addr
, adapter
->mac_table
[i
].addr
) &&
4143 adapter
->mac_table
[i
].queue
== queue
) {
4144 adapter
->mac_table
[i
].state
|= IXGBE_MAC_STATE_MODIFIED
;
4145 adapter
->mac_table
[i
].state
&= ~IXGBE_MAC_STATE_IN_USE
;
4146 eth_zero_addr(adapter
->mac_table
[i
].addr
);
4147 adapter
->mac_table
[i
].queue
= 0;
4148 ixgbe_sync_mac_table(adapter
);
4155 * ixgbe_write_uc_addr_list - write unicast addresses to RAR table
4156 * @netdev: network interface device structure
4158 * Writes unicast address list to the RAR table.
4159 * Returns: -ENOMEM on failure/insufficient address space
4160 * 0 on no addresses written
4161 * X on writing X addresses to the RAR table
4163 static int ixgbe_write_uc_addr_list(struct net_device
*netdev
, int vfn
)
4165 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4168 /* return ENOMEM indicating insufficient memory for addresses */
4169 if (netdev_uc_count(netdev
) > ixgbe_available_rars(adapter
))
4172 if (!netdev_uc_empty(netdev
)) {
4173 struct netdev_hw_addr
*ha
;
4174 netdev_for_each_uc_addr(ha
, netdev
) {
4175 ixgbe_del_mac_filter(adapter
, ha
->addr
, vfn
);
4176 ixgbe_add_mac_filter(adapter
, ha
->addr
, vfn
);
4184 * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4185 * @netdev: network interface device structure
4187 * The set_rx_method entry point is called whenever the unicast/multicast
4188 * address list or the network interface flags are updated. This routine is
4189 * responsible for configuring the hardware for proper unicast, multicast and
4192 void ixgbe_set_rx_mode(struct net_device
*netdev
)
4194 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
4195 struct ixgbe_hw
*hw
= &adapter
->hw
;
4196 u32 fctrl
, vmolr
= IXGBE_VMOLR_BAM
| IXGBE_VMOLR_AUPE
;
4200 /* Check for Promiscuous and All Multicast modes */
4201 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
4202 vlnctrl
= IXGBE_READ_REG(hw
, IXGBE_VLNCTRL
);
4204 /* set all bits that we expect to always be set */
4205 fctrl
&= ~IXGBE_FCTRL_SBP
; /* disable store-bad-packets */
4206 fctrl
|= IXGBE_FCTRL_BAM
;
4207 fctrl
|= IXGBE_FCTRL_DPF
; /* discard pause frames when FC enabled */
4208 fctrl
|= IXGBE_FCTRL_PMCF
;
4210 /* clear the bits we are changing the status of */
4211 fctrl
&= ~(IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
4212 vlnctrl
&= ~(IXGBE_VLNCTRL_VFE
| IXGBE_VLNCTRL_CFIEN
);
4213 if (netdev
->flags
& IFF_PROMISC
) {
4214 hw
->addr_ctrl
.user_set_promisc
= true;
4215 fctrl
|= (IXGBE_FCTRL_UPE
| IXGBE_FCTRL_MPE
);
4216 vmolr
|= IXGBE_VMOLR_MPE
;
4217 /* Only disable hardware filter vlans in promiscuous mode
4218 * if SR-IOV and VMDQ are disabled - otherwise ensure
4219 * that hardware VLAN filters remain enabled.
4221 if (adapter
->flags
& (IXGBE_FLAG_VMDQ_ENABLED
|
4222 IXGBE_FLAG_SRIOV_ENABLED
))
4223 vlnctrl
|= (IXGBE_VLNCTRL_VFE
| IXGBE_VLNCTRL_CFIEN
);
4225 if (netdev
->flags
& IFF_ALLMULTI
) {
4226 fctrl
|= IXGBE_FCTRL_MPE
;
4227 vmolr
|= IXGBE_VMOLR_MPE
;
4229 vlnctrl
|= IXGBE_VLNCTRL_VFE
;
4230 hw
->addr_ctrl
.user_set_promisc
= false;
4234 * Write addresses to available RAR registers, if there is not
4235 * sufficient space to store all the addresses then enable
4236 * unicast promiscuous mode
4238 count
= ixgbe_write_uc_addr_list(netdev
, VMDQ_P(0));
4240 fctrl
|= IXGBE_FCTRL_UPE
;
4241 vmolr
|= IXGBE_VMOLR_ROPE
;
4244 /* Write addresses to the MTA, if the attempt fails
4245 * then we should just turn on promiscuous mode so
4246 * that we can at least receive multicast traffic
4248 count
= ixgbe_write_mc_addr_list(netdev
);
4250 fctrl
|= IXGBE_FCTRL_MPE
;
4251 vmolr
|= IXGBE_VMOLR_MPE
;
4253 vmolr
|= IXGBE_VMOLR_ROMPE
;
4256 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
4257 vmolr
|= IXGBE_READ_REG(hw
, IXGBE_VMOLR(VMDQ_P(0))) &
4258 ~(IXGBE_VMOLR_MPE
| IXGBE_VMOLR_ROMPE
|
4260 IXGBE_WRITE_REG(hw
, IXGBE_VMOLR(VMDQ_P(0)), vmolr
);
4263 /* This is useful for sniffing bad packets. */
4264 if (adapter
->netdev
->features
& NETIF_F_RXALL
) {
4265 /* UPE and MPE will be handled by normal PROMISC logic
4266 * in e1000e_set_rx_mode */
4267 fctrl
|= (IXGBE_FCTRL_SBP
| /* Receive bad packets */
4268 IXGBE_FCTRL_BAM
| /* RX All Bcast Pkts */
4269 IXGBE_FCTRL_PMCF
); /* RX All MAC Ctrl Pkts */
4271 fctrl
&= ~(IXGBE_FCTRL_DPF
);
4272 /* NOTE: VLAN filtering is disabled by setting PROMISC */
4275 IXGBE_WRITE_REG(hw
, IXGBE_VLNCTRL
, vlnctrl
);
4276 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
4278 if (netdev
->features
& NETIF_F_HW_VLAN_CTAG_RX
)
4279 ixgbe_vlan_strip_enable(adapter
);
4281 ixgbe_vlan_strip_disable(adapter
);
4284 static void ixgbe_napi_enable_all(struct ixgbe_adapter
*adapter
)
4288 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++) {
4289 ixgbe_qv_init_lock(adapter
->q_vector
[q_idx
]);
4290 napi_enable(&adapter
->q_vector
[q_idx
]->napi
);
4294 static void ixgbe_napi_disable_all(struct ixgbe_adapter
*adapter
)
4298 for (q_idx
= 0; q_idx
< adapter
->num_q_vectors
; q_idx
++) {
4299 napi_disable(&adapter
->q_vector
[q_idx
]->napi
);
4300 while (!ixgbe_qv_disable(adapter
->q_vector
[q_idx
])) {
4301 pr_info("QV %d locked\n", q_idx
);
4302 usleep_range(1000, 20000);
4307 static void ixgbe_clear_vxlan_port(struct ixgbe_adapter
*adapter
)
4309 switch (adapter
->hw
.mac
.type
) {
4310 case ixgbe_mac_X550
:
4311 case ixgbe_mac_X550EM_x
:
4312 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_VXLANCTRL
, 0);
4313 #ifdef CONFIG_IXGBE_VXLAN
4314 adapter
->vxlan_port
= 0;
4322 #ifdef CONFIG_IXGBE_DCB
4324 * ixgbe_configure_dcb - Configure DCB hardware
4325 * @adapter: ixgbe adapter struct
4327 * This is called by the driver on open to configure the DCB hardware.
4328 * This is also called by the gennetlink interface when reconfiguring
4331 static void ixgbe_configure_dcb(struct ixgbe_adapter
*adapter
)
4333 struct ixgbe_hw
*hw
= &adapter
->hw
;
4334 int max_frame
= adapter
->netdev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
4336 if (!(adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
)) {
4337 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4338 netif_set_gso_max_size(adapter
->netdev
, 65536);
4342 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4343 netif_set_gso_max_size(adapter
->netdev
, 32768);
4346 if (adapter
->netdev
->features
& NETIF_F_FCOE_MTU
)
4347 max_frame
= max(max_frame
, IXGBE_FCOE_JUMBO_FRAME_SIZE
);
4350 /* reconfigure the hardware */
4351 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
) {
4352 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
4354 ixgbe_dcb_calculate_tc_credits(hw
, &adapter
->dcb_cfg
, max_frame
,
4356 ixgbe_dcb_hw_config(hw
, &adapter
->dcb_cfg
);
4357 } else if (adapter
->ixgbe_ieee_ets
&& adapter
->ixgbe_ieee_pfc
) {
4358 ixgbe_dcb_hw_ets(&adapter
->hw
,
4359 adapter
->ixgbe_ieee_ets
,
4361 ixgbe_dcb_hw_pfc_config(&adapter
->hw
,
4362 adapter
->ixgbe_ieee_pfc
->pfc_en
,
4363 adapter
->ixgbe_ieee_ets
->prio_tc
);
4366 /* Enable RSS Hash per TC */
4367 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
4369 u16 rss_i
= adapter
->ring_feature
[RING_F_RSS
].indices
- 1;
4376 /* write msb to all 8 TCs in one write */
4377 IXGBE_WRITE_REG(hw
, IXGBE_RQTC
, msb
* 0x11111111);
4382 /* Additional bittime to account for IXGBE framing */
4383 #define IXGBE_ETH_FRAMING 20
4386 * ixgbe_hpbthresh - calculate high water mark for flow control
4388 * @adapter: board private structure to calculate for
4389 * @pb: packet buffer to calculate
4391 static int ixgbe_hpbthresh(struct ixgbe_adapter
*adapter
, int pb
)
4393 struct ixgbe_hw
*hw
= &adapter
->hw
;
4394 struct net_device
*dev
= adapter
->netdev
;
4395 int link
, tc
, kb
, marker
;
4398 /* Calculate max LAN frame size */
4399 tc
= link
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
+ IXGBE_ETH_FRAMING
;
4402 /* FCoE traffic class uses FCOE jumbo frames */
4403 if ((dev
->features
& NETIF_F_FCOE_MTU
) &&
4404 (tc
< IXGBE_FCOE_JUMBO_FRAME_SIZE
) &&
4405 (pb
== ixgbe_fcoe_get_tc(adapter
)))
4406 tc
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
4409 /* Calculate delay value for device */
4410 switch (hw
->mac
.type
) {
4411 case ixgbe_mac_X540
:
4412 case ixgbe_mac_X550
:
4413 case ixgbe_mac_X550EM_x
:
4414 dv_id
= IXGBE_DV_X540(link
, tc
);
4417 dv_id
= IXGBE_DV(link
, tc
);
4421 /* Loopback switch introduces additional latency */
4422 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
4423 dv_id
+= IXGBE_B2BT(tc
);
4425 /* Delay value is calculated in bit times convert to KB */
4426 kb
= IXGBE_BT2KB(dv_id
);
4427 rx_pba
= IXGBE_READ_REG(hw
, IXGBE_RXPBSIZE(pb
)) >> 10;
4429 marker
= rx_pba
- kb
;
4431 /* It is possible that the packet buffer is not large enough
4432 * to provide required headroom. In this case throw an error
4433 * to user and a do the best we can.
4436 e_warn(drv
, "Packet Buffer(%i) can not provide enough"
4437 "headroom to support flow control."
4438 "Decrease MTU or number of traffic classes\n", pb
);
4446 * ixgbe_lpbthresh - calculate low water mark for for flow control
4448 * @adapter: board private structure to calculate for
4449 * @pb: packet buffer to calculate
4451 static int ixgbe_lpbthresh(struct ixgbe_adapter
*adapter
, int pb
)
4453 struct ixgbe_hw
*hw
= &adapter
->hw
;
4454 struct net_device
*dev
= adapter
->netdev
;
4458 /* Calculate max LAN frame size */
4459 tc
= dev
->mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
4462 /* FCoE traffic class uses FCOE jumbo frames */
4463 if ((dev
->features
& NETIF_F_FCOE_MTU
) &&
4464 (tc
< IXGBE_FCOE_JUMBO_FRAME_SIZE
) &&
4465 (pb
== netdev_get_prio_tc_map(dev
, adapter
->fcoe
.up
)))
4466 tc
= IXGBE_FCOE_JUMBO_FRAME_SIZE
;
4469 /* Calculate delay value for device */
4470 switch (hw
->mac
.type
) {
4471 case ixgbe_mac_X540
:
4472 case ixgbe_mac_X550
:
4473 case ixgbe_mac_X550EM_x
:
4474 dv_id
= IXGBE_LOW_DV_X540(tc
);
4477 dv_id
= IXGBE_LOW_DV(tc
);
4481 /* Delay value is calculated in bit times convert to KB */
4482 return IXGBE_BT2KB(dv_id
);
4486 * ixgbe_pbthresh_setup - calculate and setup high low water marks
4488 static void ixgbe_pbthresh_setup(struct ixgbe_adapter
*adapter
)
4490 struct ixgbe_hw
*hw
= &adapter
->hw
;
4491 int num_tc
= netdev_get_num_tc(adapter
->netdev
);
4497 for (i
= 0; i
< num_tc
; i
++) {
4498 hw
->fc
.high_water
[i
] = ixgbe_hpbthresh(adapter
, i
);
4499 hw
->fc
.low_water
[i
] = ixgbe_lpbthresh(adapter
, i
);
4501 /* Low water marks must not be larger than high water marks */
4502 if (hw
->fc
.low_water
[i
] > hw
->fc
.high_water
[i
])
4503 hw
->fc
.low_water
[i
] = 0;
4506 for (; i
< MAX_TRAFFIC_CLASS
; i
++)
4507 hw
->fc
.high_water
[i
] = 0;
4510 static void ixgbe_configure_pb(struct ixgbe_adapter
*adapter
)
4512 struct ixgbe_hw
*hw
= &adapter
->hw
;
4514 u8 tc
= netdev_get_num_tc(adapter
->netdev
);
4516 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
||
4517 adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
4518 hdrm
= 32 << adapter
->fdir_pballoc
;
4522 hw
->mac
.ops
.set_rxpba(hw
, tc
, hdrm
, PBA_STRATEGY_EQUAL
);
4523 ixgbe_pbthresh_setup(adapter
);
4526 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter
*adapter
)
4528 struct ixgbe_hw
*hw
= &adapter
->hw
;
4529 struct hlist_node
*node2
;
4530 struct ixgbe_fdir_filter
*filter
;
4532 spin_lock(&adapter
->fdir_perfect_lock
);
4534 if (!hlist_empty(&adapter
->fdir_filter_list
))
4535 ixgbe_fdir_set_input_mask_82599(hw
, &adapter
->fdir_mask
);
4537 hlist_for_each_entry_safe(filter
, node2
,
4538 &adapter
->fdir_filter_list
, fdir_node
) {
4539 ixgbe_fdir_write_perfect_filter_82599(hw
,
4542 (filter
->action
== IXGBE_FDIR_DROP_QUEUE
) ?
4543 IXGBE_FDIR_DROP_QUEUE
:
4544 adapter
->rx_ring
[filter
->action
]->reg_idx
);
4547 spin_unlock(&adapter
->fdir_perfect_lock
);
4550 static void ixgbe_macvlan_set_rx_mode(struct net_device
*dev
, unsigned int pool
,
4551 struct ixgbe_adapter
*adapter
)
4553 struct ixgbe_hw
*hw
= &adapter
->hw
;
4556 /* No unicast promiscuous support for VMDQ devices. */
4557 vmolr
= IXGBE_READ_REG(hw
, IXGBE_VMOLR(pool
));
4558 vmolr
|= (IXGBE_VMOLR_ROMPE
| IXGBE_VMOLR_BAM
| IXGBE_VMOLR_AUPE
);
4560 /* clear the affected bit */
4561 vmolr
&= ~IXGBE_VMOLR_MPE
;
4563 if (dev
->flags
& IFF_ALLMULTI
) {
4564 vmolr
|= IXGBE_VMOLR_MPE
;
4566 vmolr
|= IXGBE_VMOLR_ROMPE
;
4567 hw
->mac
.ops
.update_mc_addr_list(hw
, dev
);
4569 ixgbe_write_uc_addr_list(adapter
->netdev
, pool
);
4570 IXGBE_WRITE_REG(hw
, IXGBE_VMOLR(pool
), vmolr
);
4573 static void ixgbe_fwd_psrtype(struct ixgbe_fwd_adapter
*vadapter
)
4575 struct ixgbe_adapter
*adapter
= vadapter
->real_adapter
;
4576 int rss_i
= adapter
->num_rx_queues_per_pool
;
4577 struct ixgbe_hw
*hw
= &adapter
->hw
;
4578 u16 pool
= vadapter
->pool
;
4579 u32 psrtype
= IXGBE_PSRTYPE_TCPHDR
|
4580 IXGBE_PSRTYPE_UDPHDR
|
4581 IXGBE_PSRTYPE_IPV4HDR
|
4582 IXGBE_PSRTYPE_L2HDR
|
4583 IXGBE_PSRTYPE_IPV6HDR
;
4585 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
4593 IXGBE_WRITE_REG(hw
, IXGBE_PSRTYPE(VMDQ_P(pool
)), psrtype
);
4597 * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
4598 * @rx_ring: ring to free buffers from
4600 static void ixgbe_clean_rx_ring(struct ixgbe_ring
*rx_ring
)
4602 struct device
*dev
= rx_ring
->dev
;
4606 /* ring already cleared, nothing to do */
4607 if (!rx_ring
->rx_buffer_info
)
4610 /* Free all the Rx ring sk_buffs */
4611 for (i
= 0; i
< rx_ring
->count
; i
++) {
4612 struct ixgbe_rx_buffer
*rx_buffer
= &rx_ring
->rx_buffer_info
[i
];
4614 if (rx_buffer
->skb
) {
4615 struct sk_buff
*skb
= rx_buffer
->skb
;
4616 if (IXGBE_CB(skb
)->page_released
)
4619 ixgbe_rx_bufsz(rx_ring
),
4622 rx_buffer
->skb
= NULL
;
4625 if (!rx_buffer
->page
)
4628 dma_unmap_page(dev
, rx_buffer
->dma
,
4629 ixgbe_rx_pg_size(rx_ring
), DMA_FROM_DEVICE
);
4630 __free_pages(rx_buffer
->page
, ixgbe_rx_pg_order(rx_ring
));
4632 rx_buffer
->page
= NULL
;
4635 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
4636 memset(rx_ring
->rx_buffer_info
, 0, size
);
4638 /* Zero out the descriptor ring */
4639 memset(rx_ring
->desc
, 0, rx_ring
->size
);
4641 rx_ring
->next_to_alloc
= 0;
4642 rx_ring
->next_to_clean
= 0;
4643 rx_ring
->next_to_use
= 0;
4646 static void ixgbe_disable_fwd_ring(struct ixgbe_fwd_adapter
*vadapter
,
4647 struct ixgbe_ring
*rx_ring
)
4649 struct ixgbe_adapter
*adapter
= vadapter
->real_adapter
;
4650 int index
= rx_ring
->queue_index
+ vadapter
->rx_base_queue
;
4652 /* shutdown specific queue receive and wait for dma to settle */
4653 ixgbe_disable_rx_queue(adapter
, rx_ring
);
4654 usleep_range(10000, 20000);
4655 ixgbe_irq_disable_queues(adapter
, ((u64
)1 << index
));
4656 ixgbe_clean_rx_ring(rx_ring
);
4657 rx_ring
->l2_accel_priv
= NULL
;
4660 static int ixgbe_fwd_ring_down(struct net_device
*vdev
,
4661 struct ixgbe_fwd_adapter
*accel
)
4663 struct ixgbe_adapter
*adapter
= accel
->real_adapter
;
4664 unsigned int rxbase
= accel
->rx_base_queue
;
4665 unsigned int txbase
= accel
->tx_base_queue
;
4668 netif_tx_stop_all_queues(vdev
);
4670 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4671 ixgbe_disable_fwd_ring(accel
, adapter
->rx_ring
[rxbase
+ i
]);
4672 adapter
->rx_ring
[rxbase
+ i
]->netdev
= adapter
->netdev
;
4675 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4676 adapter
->tx_ring
[txbase
+ i
]->l2_accel_priv
= NULL
;
4677 adapter
->tx_ring
[txbase
+ i
]->netdev
= adapter
->netdev
;
4684 static int ixgbe_fwd_ring_up(struct net_device
*vdev
,
4685 struct ixgbe_fwd_adapter
*accel
)
4687 struct ixgbe_adapter
*adapter
= accel
->real_adapter
;
4688 unsigned int rxbase
, txbase
, queues
;
4689 int i
, baseq
, err
= 0;
4691 if (!test_bit(accel
->pool
, &adapter
->fwd_bitmask
))
4694 baseq
= accel
->pool
* adapter
->num_rx_queues_per_pool
;
4695 netdev_dbg(vdev
, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
4696 accel
->pool
, adapter
->num_rx_pools
,
4697 baseq
, baseq
+ adapter
->num_rx_queues_per_pool
,
4698 adapter
->fwd_bitmask
);
4700 accel
->netdev
= vdev
;
4701 accel
->rx_base_queue
= rxbase
= baseq
;
4702 accel
->tx_base_queue
= txbase
= baseq
;
4704 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++)
4705 ixgbe_disable_fwd_ring(accel
, adapter
->rx_ring
[rxbase
+ i
]);
4707 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4708 adapter
->rx_ring
[rxbase
+ i
]->netdev
= vdev
;
4709 adapter
->rx_ring
[rxbase
+ i
]->l2_accel_priv
= accel
;
4710 ixgbe_configure_rx_ring(adapter
, adapter
->rx_ring
[rxbase
+ i
]);
4713 for (i
= 0; i
< adapter
->num_rx_queues_per_pool
; i
++) {
4714 adapter
->tx_ring
[txbase
+ i
]->netdev
= vdev
;
4715 adapter
->tx_ring
[txbase
+ i
]->l2_accel_priv
= accel
;
4718 queues
= min_t(unsigned int,
4719 adapter
->num_rx_queues_per_pool
, vdev
->num_tx_queues
);
4720 err
= netif_set_real_num_tx_queues(vdev
, queues
);
4724 err
= netif_set_real_num_rx_queues(vdev
, queues
);
4728 if (is_valid_ether_addr(vdev
->dev_addr
))
4729 ixgbe_add_mac_filter(adapter
, vdev
->dev_addr
, accel
->pool
);
4731 ixgbe_fwd_psrtype(accel
);
4732 ixgbe_macvlan_set_rx_mode(vdev
, accel
->pool
, adapter
);
4735 ixgbe_fwd_ring_down(vdev
, accel
);
4739 static void ixgbe_configure_dfwd(struct ixgbe_adapter
*adapter
)
4741 struct net_device
*upper
;
4742 struct list_head
*iter
;
4745 netdev_for_each_all_upper_dev_rcu(adapter
->netdev
, upper
, iter
) {
4746 if (netif_is_macvlan(upper
)) {
4747 struct macvlan_dev
*dfwd
= netdev_priv(upper
);
4748 struct ixgbe_fwd_adapter
*vadapter
= dfwd
->fwd_priv
;
4750 if (dfwd
->fwd_priv
) {
4751 err
= ixgbe_fwd_ring_up(upper
, vadapter
);
4759 static void ixgbe_configure(struct ixgbe_adapter
*adapter
)
4761 struct ixgbe_hw
*hw
= &adapter
->hw
;
4763 ixgbe_configure_pb(adapter
);
4764 #ifdef CONFIG_IXGBE_DCB
4765 ixgbe_configure_dcb(adapter
);
4768 * We must restore virtualization before VLANs or else
4769 * the VLVF registers will not be populated
4771 ixgbe_configure_virtualization(adapter
);
4773 ixgbe_set_rx_mode(adapter
->netdev
);
4774 ixgbe_restore_vlan(adapter
);
4776 switch (hw
->mac
.type
) {
4777 case ixgbe_mac_82599EB
:
4778 case ixgbe_mac_X540
:
4779 hw
->mac
.ops
.disable_rx_buff(hw
);
4785 if (adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
) {
4786 ixgbe_init_fdir_signature_82599(&adapter
->hw
,
4787 adapter
->fdir_pballoc
);
4788 } else if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
) {
4789 ixgbe_init_fdir_perfect_82599(&adapter
->hw
,
4790 adapter
->fdir_pballoc
);
4791 ixgbe_fdir_filter_restore(adapter
);
4794 switch (hw
->mac
.type
) {
4795 case ixgbe_mac_82599EB
:
4796 case ixgbe_mac_X540
:
4797 hw
->mac
.ops
.enable_rx_buff(hw
);
4803 #ifdef CONFIG_IXGBE_DCA
4805 if (adapter
->flags
& IXGBE_FLAG_DCA_CAPABLE
)
4806 ixgbe_setup_dca(adapter
);
4807 #endif /* CONFIG_IXGBE_DCA */
4810 /* configure FCoE L2 filters, redirection table, and Rx control */
4811 ixgbe_configure_fcoe(adapter
);
4813 #endif /* IXGBE_FCOE */
4814 ixgbe_configure_tx(adapter
);
4815 ixgbe_configure_rx(adapter
);
4816 ixgbe_configure_dfwd(adapter
);
4820 * ixgbe_sfp_link_config - set up SFP+ link
4821 * @adapter: pointer to private adapter struct
4823 static void ixgbe_sfp_link_config(struct ixgbe_adapter
*adapter
)
4826 * We are assuming the worst case scenario here, and that
4827 * is that an SFP was inserted/removed after the reset
4828 * but before SFP detection was enabled. As such the best
4829 * solution is to just start searching as soon as we start
4831 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
4832 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
4834 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
4835 adapter
->sfp_poll_time
= 0;
4839 * ixgbe_non_sfp_link_config - set up non-SFP+ link
4840 * @hw: pointer to private hardware struct
4842 * Returns 0 on success, negative on failure
4844 static int ixgbe_non_sfp_link_config(struct ixgbe_hw
*hw
)
4847 bool autoneg
, link_up
= false;
4848 int ret
= IXGBE_ERR_LINK_SETUP
;
4850 if (hw
->mac
.ops
.check_link
)
4851 ret
= hw
->mac
.ops
.check_link(hw
, &speed
, &link_up
, false);
4856 speed
= hw
->phy
.autoneg_advertised
;
4857 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
))
4858 ret
= hw
->mac
.ops
.get_link_capabilities(hw
, &speed
,
4863 if (hw
->mac
.ops
.setup_link
)
4864 ret
= hw
->mac
.ops
.setup_link(hw
, speed
, link_up
);
4869 static void ixgbe_setup_gpie(struct ixgbe_adapter
*adapter
)
4871 struct ixgbe_hw
*hw
= &adapter
->hw
;
4874 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
) {
4875 gpie
= IXGBE_GPIE_MSIX_MODE
| IXGBE_GPIE_PBA_SUPPORT
|
4877 gpie
|= IXGBE_GPIE_EIAME
;
4879 * use EIAM to auto-mask when MSI-X interrupt is asserted
4880 * this saves a register write for every interrupt
4882 switch (hw
->mac
.type
) {
4883 case ixgbe_mac_82598EB
:
4884 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4886 case ixgbe_mac_82599EB
:
4887 case ixgbe_mac_X540
:
4888 case ixgbe_mac_X550
:
4889 case ixgbe_mac_X550EM_x
:
4891 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4892 IXGBE_WRITE_REG(hw
, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4896 /* legacy interrupts, use EIAM to auto-mask when reading EICR,
4897 * specifically only auto mask tx and rx interrupts */
4898 IXGBE_WRITE_REG(hw
, IXGBE_EIAM
, IXGBE_EICS_RTX_QUEUE
);
4901 /* XXX: to interrupt immediately for EICS writes, enable this */
4902 /* gpie |= IXGBE_GPIE_EIMEN; */
4904 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
4905 gpie
&= ~IXGBE_GPIE_VTMODE_MASK
;
4907 switch (adapter
->ring_feature
[RING_F_VMDQ
].mask
) {
4908 case IXGBE_82599_VMDQ_8Q_MASK
:
4909 gpie
|= IXGBE_GPIE_VTMODE_16
;
4911 case IXGBE_82599_VMDQ_4Q_MASK
:
4912 gpie
|= IXGBE_GPIE_VTMODE_32
;
4915 gpie
|= IXGBE_GPIE_VTMODE_64
;
4920 /* Enable Thermal over heat sensor interrupt */
4921 if (adapter
->flags2
& IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
) {
4922 switch (adapter
->hw
.mac
.type
) {
4923 case ixgbe_mac_82599EB
:
4924 gpie
|= IXGBE_SDP0_GPIEN_8259X
;
4931 /* Enable fan failure interrupt */
4932 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
)
4933 gpie
|= IXGBE_SDP1_GPIEN(hw
);
4935 switch (hw
->mac
.type
) {
4936 case ixgbe_mac_82599EB
:
4937 gpie
|= IXGBE_SDP1_GPIEN_8259X
| IXGBE_SDP2_GPIEN_8259X
;
4939 case ixgbe_mac_X550EM_x
:
4940 gpie
|= IXGBE_SDP0_GPIEN_X540
;
4946 IXGBE_WRITE_REG(hw
, IXGBE_GPIE
, gpie
);
4949 static void ixgbe_up_complete(struct ixgbe_adapter
*adapter
)
4951 struct ixgbe_hw
*hw
= &adapter
->hw
;
4955 ixgbe_get_hw_control(adapter
);
4956 ixgbe_setup_gpie(adapter
);
4958 if (adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)
4959 ixgbe_configure_msix(adapter
);
4961 ixgbe_configure_msi_and_legacy(adapter
);
4963 /* enable the optics for 82599 SFP+ fiber */
4964 if (hw
->mac
.ops
.enable_tx_laser
)
4965 hw
->mac
.ops
.enable_tx_laser(hw
);
4967 if (hw
->phy
.ops
.set_phy_power
)
4968 hw
->phy
.ops
.set_phy_power(hw
, true);
4970 smp_mb__before_atomic();
4971 clear_bit(__IXGBE_DOWN
, &adapter
->state
);
4972 ixgbe_napi_enable_all(adapter
);
4974 if (ixgbe_is_sfp(hw
)) {
4975 ixgbe_sfp_link_config(adapter
);
4977 err
= ixgbe_non_sfp_link_config(hw
);
4979 e_err(probe
, "link_config FAILED %d\n", err
);
4982 /* clear any pending interrupts, may auto mask */
4983 IXGBE_READ_REG(hw
, IXGBE_EICR
);
4984 ixgbe_irq_enable(adapter
, true, true);
4987 * If this adapter has a fan, check to see if we had a failure
4988 * before we enabled the interrupt.
4990 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
4991 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
4992 if (esdp
& IXGBE_ESDP_SDP1
)
4993 e_crit(drv
, "Fan has stopped, replace the adapter\n");
4996 /* bring the link up in the watchdog, this could race with our first
4997 * link up interrupt but shouldn't be a problem */
4998 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
4999 adapter
->link_check_timeout
= jiffies
;
5000 mod_timer(&adapter
->service_timer
, jiffies
);
5002 /* Set PF Reset Done bit so PF/VF Mail Ops can work */
5003 ctrl_ext
= IXGBE_READ_REG(hw
, IXGBE_CTRL_EXT
);
5004 ctrl_ext
|= IXGBE_CTRL_EXT_PFRSTD
;
5005 IXGBE_WRITE_REG(hw
, IXGBE_CTRL_EXT
, ctrl_ext
);
5008 void ixgbe_reinit_locked(struct ixgbe_adapter
*adapter
)
5010 WARN_ON(in_interrupt());
5011 /* put off any impending NetWatchDogTimeout */
5012 adapter
->netdev
->trans_start
= jiffies
;
5014 while (test_and_set_bit(__IXGBE_RESETTING
, &adapter
->state
))
5015 usleep_range(1000, 2000);
5016 ixgbe_down(adapter
);
5018 * If SR-IOV enabled then wait a bit before bringing the adapter
5019 * back up to give the VFs time to respond to the reset. The
5020 * two second wait is based upon the watchdog timer cycle in
5023 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
5026 clear_bit(__IXGBE_RESETTING
, &adapter
->state
);
5029 void ixgbe_up(struct ixgbe_adapter
*adapter
)
5031 /* hardware has been reset, we need to reload some things */
5032 ixgbe_configure(adapter
);
5034 ixgbe_up_complete(adapter
);
5037 void ixgbe_reset(struct ixgbe_adapter
*adapter
)
5039 struct ixgbe_hw
*hw
= &adapter
->hw
;
5040 struct net_device
*netdev
= adapter
->netdev
;
5042 u8 old_addr
[ETH_ALEN
];
5044 if (ixgbe_removed(hw
->hw_addr
))
5046 /* lock SFP init bit to prevent race conditions with the watchdog */
5047 while (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
5048 usleep_range(1000, 2000);
5050 /* clear all SFP and link config related flags while holding SFP_INIT */
5051 adapter
->flags2
&= ~(IXGBE_FLAG2_SEARCH_FOR_SFP
|
5052 IXGBE_FLAG2_SFP_NEEDS_RESET
);
5053 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
5055 err
= hw
->mac
.ops
.init_hw(hw
);
5058 case IXGBE_ERR_SFP_NOT_PRESENT
:
5059 case IXGBE_ERR_SFP_NOT_SUPPORTED
:
5061 case IXGBE_ERR_MASTER_REQUESTS_PENDING
:
5062 e_dev_err("master disable timed out\n");
5064 case IXGBE_ERR_EEPROM_VERSION
:
5065 /* We are running on a pre-production device, log a warning */
5066 e_dev_warn("This device is a pre-production adapter/LOM. "
5067 "Please be aware there may be issues associated with "
5068 "your hardware. If you are experiencing problems "
5069 "please contact your Intel or hardware "
5070 "representative who provided you with this "
5074 e_dev_err("Hardware Error: %d\n", err
);
5077 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
5078 /* do not flush user set addresses */
5079 memcpy(old_addr
, &adapter
->mac_table
[0].addr
, netdev
->addr_len
);
5080 ixgbe_flush_sw_mac_table(adapter
);
5081 ixgbe_mac_set_default_filter(adapter
, old_addr
);
5083 /* update SAN MAC vmdq pool selection */
5084 if (hw
->mac
.san_mac_rar_index
)
5085 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
5087 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
))
5088 ixgbe_ptp_reset(adapter
);
5090 if (hw
->phy
.ops
.set_phy_power
) {
5091 if (!netif_running(adapter
->netdev
) && !adapter
->wol
)
5092 hw
->phy
.ops
.set_phy_power(hw
, false);
5094 hw
->phy
.ops
.set_phy_power(hw
, true);
5099 * ixgbe_clean_tx_ring - Free Tx Buffers
5100 * @tx_ring: ring to be cleaned
5102 static void ixgbe_clean_tx_ring(struct ixgbe_ring
*tx_ring
)
5104 struct ixgbe_tx_buffer
*tx_buffer_info
;
5108 /* ring already cleared, nothing to do */
5109 if (!tx_ring
->tx_buffer_info
)
5112 /* Free all the Tx ring sk_buffs */
5113 for (i
= 0; i
< tx_ring
->count
; i
++) {
5114 tx_buffer_info
= &tx_ring
->tx_buffer_info
[i
];
5115 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer_info
);
5118 netdev_tx_reset_queue(txring_txq(tx_ring
));
5120 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
5121 memset(tx_ring
->tx_buffer_info
, 0, size
);
5123 /* Zero out the descriptor ring */
5124 memset(tx_ring
->desc
, 0, tx_ring
->size
);
5126 tx_ring
->next_to_use
= 0;
5127 tx_ring
->next_to_clean
= 0;
5131 * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
5132 * @adapter: board private structure
5134 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter
*adapter
)
5138 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
5139 ixgbe_clean_rx_ring(adapter
->rx_ring
[i
]);
5143 * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
5144 * @adapter: board private structure
5146 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter
*adapter
)
5150 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5151 ixgbe_clean_tx_ring(adapter
->tx_ring
[i
]);
5154 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter
*adapter
)
5156 struct hlist_node
*node2
;
5157 struct ixgbe_fdir_filter
*filter
;
5159 spin_lock(&adapter
->fdir_perfect_lock
);
5161 hlist_for_each_entry_safe(filter
, node2
,
5162 &adapter
->fdir_filter_list
, fdir_node
) {
5163 hlist_del(&filter
->fdir_node
);
5166 adapter
->fdir_filter_count
= 0;
5168 spin_unlock(&adapter
->fdir_perfect_lock
);
5171 void ixgbe_down(struct ixgbe_adapter
*adapter
)
5173 struct net_device
*netdev
= adapter
->netdev
;
5174 struct ixgbe_hw
*hw
= &adapter
->hw
;
5175 struct net_device
*upper
;
5176 struct list_head
*iter
;
5179 /* signal that we are down to the interrupt handler */
5180 if (test_and_set_bit(__IXGBE_DOWN
, &adapter
->state
))
5181 return; /* do nothing if already down */
5183 /* disable receives */
5184 hw
->mac
.ops
.disable_rx(hw
);
5186 /* disable all enabled rx queues */
5187 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
5188 /* this call also flushes the previous write */
5189 ixgbe_disable_rx_queue(adapter
, adapter
->rx_ring
[i
]);
5191 usleep_range(10000, 20000);
5193 netif_tx_stop_all_queues(netdev
);
5195 /* call carrier off first to avoid false dev_watchdog timeouts */
5196 netif_carrier_off(netdev
);
5197 netif_tx_disable(netdev
);
5199 /* disable any upper devices */
5200 netdev_for_each_all_upper_dev_rcu(adapter
->netdev
, upper
, iter
) {
5201 if (netif_is_macvlan(upper
)) {
5202 struct macvlan_dev
*vlan
= netdev_priv(upper
);
5204 if (vlan
->fwd_priv
) {
5205 netif_tx_stop_all_queues(upper
);
5206 netif_carrier_off(upper
);
5207 netif_tx_disable(upper
);
5212 ixgbe_irq_disable(adapter
);
5214 ixgbe_napi_disable_all(adapter
);
5216 adapter
->flags2
&= ~(IXGBE_FLAG2_FDIR_REQUIRES_REINIT
|
5217 IXGBE_FLAG2_RESET_REQUESTED
);
5218 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
5220 del_timer_sync(&adapter
->service_timer
);
5222 if (adapter
->num_vfs
) {
5223 /* Clear EITR Select mapping */
5224 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_EITRSEL
, 0);
5226 /* Mark all the VFs as inactive */
5227 for (i
= 0 ; i
< adapter
->num_vfs
; i
++)
5228 adapter
->vfinfo
[i
].clear_to_send
= false;
5230 /* ping all the active vfs to let them know we are going down */
5231 ixgbe_ping_all_vfs(adapter
);
5233 /* Disable all VFTE/VFRE TX/RX */
5234 ixgbe_disable_tx_rx(adapter
);
5237 /* disable transmits in the hardware now that interrupts are off */
5238 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5239 u8 reg_idx
= adapter
->tx_ring
[i
]->reg_idx
;
5240 IXGBE_WRITE_REG(hw
, IXGBE_TXDCTL(reg_idx
), IXGBE_TXDCTL_SWFLSH
);
5243 /* Disable the Tx DMA engine on 82599 and later MAC */
5244 switch (hw
->mac
.type
) {
5245 case ixgbe_mac_82599EB
:
5246 case ixgbe_mac_X540
:
5247 case ixgbe_mac_X550
:
5248 case ixgbe_mac_X550EM_x
:
5249 IXGBE_WRITE_REG(hw
, IXGBE_DMATXCTL
,
5250 (IXGBE_READ_REG(hw
, IXGBE_DMATXCTL
) &
5251 ~IXGBE_DMATXCTL_TE
));
5257 if (!pci_channel_offline(adapter
->pdev
))
5258 ixgbe_reset(adapter
);
5260 /* power down the optics for 82599 SFP+ fiber */
5261 if (hw
->mac
.ops
.disable_tx_laser
)
5262 hw
->mac
.ops
.disable_tx_laser(hw
);
5264 ixgbe_clean_all_tx_rings(adapter
);
5265 ixgbe_clean_all_rx_rings(adapter
);
5269 * ixgbe_tx_timeout - Respond to a Tx Hang
5270 * @netdev: network interface device structure
5272 static void ixgbe_tx_timeout(struct net_device
*netdev
)
5274 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5276 /* Do the reset outside of interrupt context */
5277 ixgbe_tx_timeout_reset(adapter
);
5281 * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
5282 * @adapter: board private structure to initialize
5284 * ixgbe_sw_init initializes the Adapter private data structure.
5285 * Fields are initialized based on PCI device information and
5286 * OS network device settings (MTU size).
5288 static int ixgbe_sw_init(struct ixgbe_adapter
*adapter
)
5290 struct ixgbe_hw
*hw
= &adapter
->hw
;
5291 struct pci_dev
*pdev
= adapter
->pdev
;
5292 unsigned int rss
, fdir
;
5294 #ifdef CONFIG_IXGBE_DCB
5296 struct tc_configuration
*tc
;
5299 /* PCI config space info */
5301 hw
->vendor_id
= pdev
->vendor
;
5302 hw
->device_id
= pdev
->device
;
5303 hw
->revision_id
= pdev
->revision
;
5304 hw
->subsystem_vendor_id
= pdev
->subsystem_vendor
;
5305 hw
->subsystem_device_id
= pdev
->subsystem_device
;
5307 /* Set common capability flags and settings */
5308 rss
= min_t(int, ixgbe_max_rss_indices(adapter
), num_online_cpus());
5309 adapter
->ring_feature
[RING_F_RSS
].limit
= rss
;
5310 adapter
->flags2
|= IXGBE_FLAG2_RSC_CAPABLE
;
5311 adapter
->max_q_vectors
= MAX_Q_VECTORS_82599
;
5312 adapter
->atr_sample_rate
= 20;
5313 fdir
= min_t(int, IXGBE_MAX_FDIR_INDICES
, num_online_cpus());
5314 adapter
->ring_feature
[RING_F_FDIR
].limit
= fdir
;
5315 adapter
->fdir_pballoc
= IXGBE_FDIR_PBALLOC_64K
;
5316 #ifdef CONFIG_IXGBE_DCA
5317 adapter
->flags
|= IXGBE_FLAG_DCA_CAPABLE
;
5320 adapter
->flags
|= IXGBE_FLAG_FCOE_CAPABLE
;
5321 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
5322 #ifdef CONFIG_IXGBE_DCB
5323 /* Default traffic class to use for FCoE */
5324 adapter
->fcoe
.up
= IXGBE_FCOE_DEFTC
;
5325 #endif /* CONFIG_IXGBE_DCB */
5326 #endif /* IXGBE_FCOE */
5328 adapter
->mac_table
= kzalloc(sizeof(struct ixgbe_mac_addr
) *
5329 hw
->mac
.num_rar_entries
,
5332 /* Set MAC specific capability flags and exceptions */
5333 switch (hw
->mac
.type
) {
5334 case ixgbe_mac_82598EB
:
5335 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_CAPABLE
;
5337 if (hw
->device_id
== IXGBE_DEV_ID_82598AT
)
5338 adapter
->flags
|= IXGBE_FLAG_FAN_FAIL_CAPABLE
;
5340 adapter
->max_q_vectors
= MAX_Q_VECTORS_82598
;
5341 adapter
->ring_feature
[RING_F_FDIR
].limit
= 0;
5342 adapter
->atr_sample_rate
= 0;
5343 adapter
->fdir_pballoc
= 0;
5345 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
5346 adapter
->flags
&= ~IXGBE_FLAG_FCOE_ENABLED
;
5347 #ifdef CONFIG_IXGBE_DCB
5348 adapter
->fcoe
.up
= 0;
5349 #endif /* IXGBE_DCB */
5350 #endif /* IXGBE_FCOE */
5352 case ixgbe_mac_82599EB
:
5353 if (hw
->device_id
== IXGBE_DEV_ID_82599_T3_LOM
)
5354 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
5356 case ixgbe_mac_X540
:
5357 fwsm
= IXGBE_READ_REG(hw
, IXGBE_FWSM(hw
));
5358 if (fwsm
& IXGBE_FWSM_TS_ENABLED
)
5359 adapter
->flags2
|= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE
;
5361 case ixgbe_mac_X550EM_x
:
5362 case ixgbe_mac_X550
:
5363 #ifdef CONFIG_IXGBE_DCA
5364 adapter
->flags
&= ~IXGBE_FLAG_DCA_CAPABLE
;
5366 #ifdef CONFIG_IXGBE_VXLAN
5367 adapter
->flags
|= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
;
5375 /* FCoE support exists, always init the FCoE lock */
5376 spin_lock_init(&adapter
->fcoe
.lock
);
5379 /* n-tuple support exists, always init our spinlock */
5380 spin_lock_init(&adapter
->fdir_perfect_lock
);
5382 #ifdef CONFIG_IXGBE_DCB
5383 switch (hw
->mac
.type
) {
5384 case ixgbe_mac_X540
:
5385 case ixgbe_mac_X550
:
5386 case ixgbe_mac_X550EM_x
:
5387 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= X540_TRAFFIC_CLASS
;
5388 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= X540_TRAFFIC_CLASS
;
5391 adapter
->dcb_cfg
.num_tcs
.pg_tcs
= MAX_TRAFFIC_CLASS
;
5392 adapter
->dcb_cfg
.num_tcs
.pfc_tcs
= MAX_TRAFFIC_CLASS
;
5396 /* Configure DCB traffic classes */
5397 for (j
= 0; j
< MAX_TRAFFIC_CLASS
; j
++) {
5398 tc
= &adapter
->dcb_cfg
.tc_config
[j
];
5399 tc
->path
[DCB_TX_CONFIG
].bwg_id
= 0;
5400 tc
->path
[DCB_TX_CONFIG
].bwg_percent
= 12 + (j
& 1);
5401 tc
->path
[DCB_RX_CONFIG
].bwg_id
= 0;
5402 tc
->path
[DCB_RX_CONFIG
].bwg_percent
= 12 + (j
& 1);
5403 tc
->dcb_pfc
= pfc_disabled
;
5406 /* Initialize default user to priority mapping, UPx->TC0 */
5407 tc
= &adapter
->dcb_cfg
.tc_config
[0];
5408 tc
->path
[DCB_TX_CONFIG
].up_to_tc_bitmap
= 0xFF;
5409 tc
->path
[DCB_RX_CONFIG
].up_to_tc_bitmap
= 0xFF;
5411 adapter
->dcb_cfg
.bw_percentage
[DCB_TX_CONFIG
][0] = 100;
5412 adapter
->dcb_cfg
.bw_percentage
[DCB_RX_CONFIG
][0] = 100;
5413 adapter
->dcb_cfg
.pfc_mode_enable
= false;
5414 adapter
->dcb_set_bitmap
= 0x00;
5415 adapter
->dcbx_cap
= DCB_CAP_DCBX_HOST
| DCB_CAP_DCBX_VER_CEE
;
5416 memcpy(&adapter
->temp_dcb_cfg
, &adapter
->dcb_cfg
,
5417 sizeof(adapter
->temp_dcb_cfg
));
5421 /* default flow control settings */
5422 hw
->fc
.requested_mode
= ixgbe_fc_full
;
5423 hw
->fc
.current_mode
= ixgbe_fc_full
; /* init for ethtool output */
5424 ixgbe_pbthresh_setup(adapter
);
5425 hw
->fc
.pause_time
= IXGBE_DEFAULT_FCPAUSE
;
5426 hw
->fc
.send_xon
= true;
5427 hw
->fc
.disable_fc_autoneg
= ixgbe_device_supports_autoneg_fc(hw
);
5429 #ifdef CONFIG_PCI_IOV
5431 e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
5433 /* assign number of SR-IOV VFs */
5434 if (hw
->mac
.type
!= ixgbe_mac_82598EB
) {
5435 if (max_vfs
> IXGBE_MAX_VFS_DRV_LIMIT
) {
5436 adapter
->num_vfs
= 0;
5437 e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
5439 adapter
->num_vfs
= max_vfs
;
5442 #endif /* CONFIG_PCI_IOV */
5444 /* enable itr by default in dynamic mode */
5445 adapter
->rx_itr_setting
= 1;
5446 adapter
->tx_itr_setting
= 1;
5448 /* set default ring sizes */
5449 adapter
->tx_ring_count
= IXGBE_DEFAULT_TXD
;
5450 adapter
->rx_ring_count
= IXGBE_DEFAULT_RXD
;
5452 /* set default work limits */
5453 adapter
->tx_work_limit
= IXGBE_DEFAULT_TX_WORK
;
5455 /* initialize eeprom parameters */
5456 if (ixgbe_init_eeprom_params_generic(hw
)) {
5457 e_dev_err("EEPROM initialization failed\n");
5461 /* PF holds first pool slot */
5462 set_bit(0, &adapter
->fwd_bitmask
);
5463 set_bit(__IXGBE_DOWN
, &adapter
->state
);
5469 * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
5470 * @tx_ring: tx descriptor ring (for a specific queue) to setup
5472 * Return 0 on success, negative on failure
5474 int ixgbe_setup_tx_resources(struct ixgbe_ring
*tx_ring
)
5476 struct device
*dev
= tx_ring
->dev
;
5477 int orig_node
= dev_to_node(dev
);
5481 size
= sizeof(struct ixgbe_tx_buffer
) * tx_ring
->count
;
5483 if (tx_ring
->q_vector
)
5484 ring_node
= tx_ring
->q_vector
->numa_node
;
5486 tx_ring
->tx_buffer_info
= vzalloc_node(size
, ring_node
);
5487 if (!tx_ring
->tx_buffer_info
)
5488 tx_ring
->tx_buffer_info
= vzalloc(size
);
5489 if (!tx_ring
->tx_buffer_info
)
5492 u64_stats_init(&tx_ring
->syncp
);
5494 /* round up to nearest 4K */
5495 tx_ring
->size
= tx_ring
->count
* sizeof(union ixgbe_adv_tx_desc
);
5496 tx_ring
->size
= ALIGN(tx_ring
->size
, 4096);
5498 set_dev_node(dev
, ring_node
);
5499 tx_ring
->desc
= dma_alloc_coherent(dev
,
5503 set_dev_node(dev
, orig_node
);
5505 tx_ring
->desc
= dma_alloc_coherent(dev
, tx_ring
->size
,
5506 &tx_ring
->dma
, GFP_KERNEL
);
5510 tx_ring
->next_to_use
= 0;
5511 tx_ring
->next_to_clean
= 0;
5515 vfree(tx_ring
->tx_buffer_info
);
5516 tx_ring
->tx_buffer_info
= NULL
;
5517 dev_err(dev
, "Unable to allocate memory for the Tx descriptor ring\n");
5522 * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
5523 * @adapter: board private structure
5525 * If this function returns with an error, then it's possible one or
5526 * more of the rings is populated (while the rest are not). It is the
5527 * callers duty to clean those orphaned rings.
5529 * Return 0 on success, negative on failure
5531 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter
*adapter
)
5535 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
5536 err
= ixgbe_setup_tx_resources(adapter
->tx_ring
[i
]);
5540 e_err(probe
, "Allocation for Tx Queue %u failed\n", i
);
5546 /* rewind the index freeing the rings as we go */
5548 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
5553 * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
5554 * @rx_ring: rx descriptor ring (for a specific queue) to setup
5556 * Returns 0 on success, negative on failure
5558 int ixgbe_setup_rx_resources(struct ixgbe_ring
*rx_ring
)
5560 struct device
*dev
= rx_ring
->dev
;
5561 int orig_node
= dev_to_node(dev
);
5565 size
= sizeof(struct ixgbe_rx_buffer
) * rx_ring
->count
;
5567 if (rx_ring
->q_vector
)
5568 ring_node
= rx_ring
->q_vector
->numa_node
;
5570 rx_ring
->rx_buffer_info
= vzalloc_node(size
, ring_node
);
5571 if (!rx_ring
->rx_buffer_info
)
5572 rx_ring
->rx_buffer_info
= vzalloc(size
);
5573 if (!rx_ring
->rx_buffer_info
)
5576 u64_stats_init(&rx_ring
->syncp
);
5578 /* Round up to nearest 4K */
5579 rx_ring
->size
= rx_ring
->count
* sizeof(union ixgbe_adv_rx_desc
);
5580 rx_ring
->size
= ALIGN(rx_ring
->size
, 4096);
5582 set_dev_node(dev
, ring_node
);
5583 rx_ring
->desc
= dma_alloc_coherent(dev
,
5587 set_dev_node(dev
, orig_node
);
5589 rx_ring
->desc
= dma_alloc_coherent(dev
, rx_ring
->size
,
5590 &rx_ring
->dma
, GFP_KERNEL
);
5594 rx_ring
->next_to_clean
= 0;
5595 rx_ring
->next_to_use
= 0;
5599 vfree(rx_ring
->rx_buffer_info
);
5600 rx_ring
->rx_buffer_info
= NULL
;
5601 dev_err(dev
, "Unable to allocate memory for the Rx descriptor ring\n");
5606 * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
5607 * @adapter: board private structure
5609 * If this function returns with an error, then it's possible one or
5610 * more of the rings is populated (while the rest are not). It is the
5611 * callers duty to clean those orphaned rings.
5613 * Return 0 on success, negative on failure
5615 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter
*adapter
)
5619 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
5620 err
= ixgbe_setup_rx_resources(adapter
->rx_ring
[i
]);
5624 e_err(probe
, "Allocation for Rx Queue %u failed\n", i
);
5629 err
= ixgbe_setup_fcoe_ddp_resources(adapter
);
5634 /* rewind the index freeing the rings as we go */
5636 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
5641 * ixgbe_free_tx_resources - Free Tx Resources per Queue
5642 * @tx_ring: Tx descriptor ring for a specific queue
5644 * Free all transmit software resources
5646 void ixgbe_free_tx_resources(struct ixgbe_ring
*tx_ring
)
5648 ixgbe_clean_tx_ring(tx_ring
);
5650 vfree(tx_ring
->tx_buffer_info
);
5651 tx_ring
->tx_buffer_info
= NULL
;
5653 /* if not set, then don't free */
5657 dma_free_coherent(tx_ring
->dev
, tx_ring
->size
,
5658 tx_ring
->desc
, tx_ring
->dma
);
5660 tx_ring
->desc
= NULL
;
5664 * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
5665 * @adapter: board private structure
5667 * Free all transmit software resources
5669 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter
*adapter
)
5673 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
5674 if (adapter
->tx_ring
[i
]->desc
)
5675 ixgbe_free_tx_resources(adapter
->tx_ring
[i
]);
5679 * ixgbe_free_rx_resources - Free Rx Resources
5680 * @rx_ring: ring to clean the resources from
5682 * Free all receive software resources
5684 void ixgbe_free_rx_resources(struct ixgbe_ring
*rx_ring
)
5686 ixgbe_clean_rx_ring(rx_ring
);
5688 vfree(rx_ring
->rx_buffer_info
);
5689 rx_ring
->rx_buffer_info
= NULL
;
5691 /* if not set, then don't free */
5695 dma_free_coherent(rx_ring
->dev
, rx_ring
->size
,
5696 rx_ring
->desc
, rx_ring
->dma
);
5698 rx_ring
->desc
= NULL
;
5702 * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
5703 * @adapter: board private structure
5705 * Free all receive software resources
5707 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter
*adapter
)
5712 ixgbe_free_fcoe_ddp_resources(adapter
);
5715 for (i
= 0; i
< adapter
->num_rx_queues
; i
++)
5716 if (adapter
->rx_ring
[i
]->desc
)
5717 ixgbe_free_rx_resources(adapter
->rx_ring
[i
]);
5721 * ixgbe_change_mtu - Change the Maximum Transfer Unit
5722 * @netdev: network interface device structure
5723 * @new_mtu: new value for maximum frame size
5725 * Returns 0 on success, negative on failure
5727 static int ixgbe_change_mtu(struct net_device
*netdev
, int new_mtu
)
5729 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5730 int max_frame
= new_mtu
+ ETH_HLEN
+ ETH_FCS_LEN
;
5732 /* MTU < 68 is an error and causes problems on some kernels */
5733 if ((new_mtu
< 68) || (max_frame
> IXGBE_MAX_JUMBO_FRAME_SIZE
))
5737 * For 82599EB we cannot allow legacy VFs to enable their receive
5738 * paths when MTU greater than 1500 is configured. So display a
5739 * warning that legacy VFs will be disabled.
5741 if ((adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) &&
5742 (adapter
->hw
.mac
.type
== ixgbe_mac_82599EB
) &&
5743 (max_frame
> (ETH_FRAME_LEN
+ ETH_FCS_LEN
)))
5744 e_warn(probe
, "Setting MTU > 1500 will disable legacy VFs\n");
5746 e_info(probe
, "changing MTU from %d to %d\n", netdev
->mtu
, new_mtu
);
5748 /* must set new MTU before calling down or up */
5749 netdev
->mtu
= new_mtu
;
5751 if (netif_running(netdev
))
5752 ixgbe_reinit_locked(adapter
);
5758 * ixgbe_open - Called when a network interface is made active
5759 * @netdev: network interface device structure
5761 * Returns 0 on success, negative value on failure
5763 * The open entry point is called when a network interface is made
5764 * active by the system (IFF_UP). At this point all resources needed
5765 * for transmit and receive operations are allocated, the interrupt
5766 * handler is registered with the OS, the watchdog timer is started,
5767 * and the stack is notified that the interface is ready.
5769 static int ixgbe_open(struct net_device
*netdev
)
5771 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5772 struct ixgbe_hw
*hw
= &adapter
->hw
;
5775 /* disallow open during test */
5776 if (test_bit(__IXGBE_TESTING
, &adapter
->state
))
5779 netif_carrier_off(netdev
);
5781 /* allocate transmit descriptors */
5782 err
= ixgbe_setup_all_tx_resources(adapter
);
5786 /* allocate receive descriptors */
5787 err
= ixgbe_setup_all_rx_resources(adapter
);
5791 ixgbe_configure(adapter
);
5793 err
= ixgbe_request_irq(adapter
);
5797 /* Notify the stack of the actual queue counts. */
5798 if (adapter
->num_rx_pools
> 1)
5799 queues
= adapter
->num_rx_queues_per_pool
;
5801 queues
= adapter
->num_tx_queues
;
5803 err
= netif_set_real_num_tx_queues(netdev
, queues
);
5805 goto err_set_queues
;
5807 if (adapter
->num_rx_pools
> 1 &&
5808 adapter
->num_rx_queues
> IXGBE_MAX_L2A_QUEUES
)
5809 queues
= IXGBE_MAX_L2A_QUEUES
;
5811 queues
= adapter
->num_rx_queues
;
5812 err
= netif_set_real_num_rx_queues(netdev
, queues
);
5814 goto err_set_queues
;
5816 ixgbe_ptp_init(adapter
);
5818 ixgbe_up_complete(adapter
);
5820 ixgbe_clear_vxlan_port(adapter
);
5821 #ifdef CONFIG_IXGBE_VXLAN
5822 vxlan_get_rx_port(netdev
);
5828 ixgbe_free_irq(adapter
);
5830 ixgbe_free_all_rx_resources(adapter
);
5831 if (hw
->phy
.ops
.set_phy_power
&& !adapter
->wol
)
5832 hw
->phy
.ops
.set_phy_power(&adapter
->hw
, false);
5834 ixgbe_free_all_tx_resources(adapter
);
5836 ixgbe_reset(adapter
);
5841 static void ixgbe_close_suspend(struct ixgbe_adapter
*adapter
)
5843 ixgbe_ptp_suspend(adapter
);
5845 if (adapter
->hw
.phy
.ops
.enter_lplu
) {
5846 adapter
->hw
.phy
.reset_disable
= true;
5847 ixgbe_down(adapter
);
5848 adapter
->hw
.phy
.ops
.enter_lplu(&adapter
->hw
);
5849 adapter
->hw
.phy
.reset_disable
= false;
5851 ixgbe_down(adapter
);
5854 ixgbe_free_irq(adapter
);
5856 ixgbe_free_all_tx_resources(adapter
);
5857 ixgbe_free_all_rx_resources(adapter
);
5861 * ixgbe_close - Disables a network interface
5862 * @netdev: network interface device structure
5864 * Returns 0, this is not allowed to fail
5866 * The close entry point is called when an interface is de-activated
5867 * by the OS. The hardware is still under the drivers control, but
5868 * needs to be disabled. A global MAC reset is issued to stop the
5869 * hardware, and all transmit and receive resources are freed.
5871 static int ixgbe_close(struct net_device
*netdev
)
5873 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
5875 ixgbe_ptp_stop(adapter
);
5877 ixgbe_close_suspend(adapter
);
5879 ixgbe_fdir_filter_exit(adapter
);
5881 ixgbe_release_hw_control(adapter
);
5887 static int ixgbe_resume(struct pci_dev
*pdev
)
5889 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5890 struct net_device
*netdev
= adapter
->netdev
;
5893 adapter
->hw
.hw_addr
= adapter
->io_addr
;
5894 pci_set_power_state(pdev
, PCI_D0
);
5895 pci_restore_state(pdev
);
5897 * pci_restore_state clears dev->state_saved so call
5898 * pci_save_state to restore it.
5900 pci_save_state(pdev
);
5902 err
= pci_enable_device_mem(pdev
);
5904 e_dev_err("Cannot enable PCI device from suspend\n");
5907 smp_mb__before_atomic();
5908 clear_bit(__IXGBE_DISABLED
, &adapter
->state
);
5909 pci_set_master(pdev
);
5911 pci_wake_from_d3(pdev
, false);
5913 ixgbe_reset(adapter
);
5915 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
5918 err
= ixgbe_init_interrupt_scheme(adapter
);
5919 if (!err
&& netif_running(netdev
))
5920 err
= ixgbe_open(netdev
);
5927 netif_device_attach(netdev
);
5931 #endif /* CONFIG_PM */
5933 static int __ixgbe_shutdown(struct pci_dev
*pdev
, bool *enable_wake
)
5935 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
5936 struct net_device
*netdev
= adapter
->netdev
;
5937 struct ixgbe_hw
*hw
= &adapter
->hw
;
5939 u32 wufc
= adapter
->wol
;
5944 netif_device_detach(netdev
);
5947 if (netif_running(netdev
))
5948 ixgbe_close_suspend(adapter
);
5951 ixgbe_clear_interrupt_scheme(adapter
);
5954 retval
= pci_save_state(pdev
);
5959 if (hw
->mac
.ops
.stop_link_on_d3
)
5960 hw
->mac
.ops
.stop_link_on_d3(hw
);
5963 ixgbe_set_rx_mode(netdev
);
5965 /* enable the optics for 82599 SFP+ fiber as we can WoL */
5966 if (hw
->mac
.ops
.enable_tx_laser
)
5967 hw
->mac
.ops
.enable_tx_laser(hw
);
5969 /* turn on all-multi mode if wake on multicast is enabled */
5970 if (wufc
& IXGBE_WUFC_MC
) {
5971 fctrl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
5972 fctrl
|= IXGBE_FCTRL_MPE
;
5973 IXGBE_WRITE_REG(hw
, IXGBE_FCTRL
, fctrl
);
5976 ctrl
= IXGBE_READ_REG(hw
, IXGBE_CTRL
);
5977 ctrl
|= IXGBE_CTRL_GIO_DIS
;
5978 IXGBE_WRITE_REG(hw
, IXGBE_CTRL
, ctrl
);
5980 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, wufc
);
5982 IXGBE_WRITE_REG(hw
, IXGBE_WUC
, 0);
5983 IXGBE_WRITE_REG(hw
, IXGBE_WUFC
, 0);
5986 switch (hw
->mac
.type
) {
5987 case ixgbe_mac_82598EB
:
5988 pci_wake_from_d3(pdev
, false);
5990 case ixgbe_mac_82599EB
:
5991 case ixgbe_mac_X540
:
5992 case ixgbe_mac_X550
:
5993 case ixgbe_mac_X550EM_x
:
5994 pci_wake_from_d3(pdev
, !!wufc
);
6000 *enable_wake
= !!wufc
;
6001 if (hw
->phy
.ops
.set_phy_power
&& !*enable_wake
)
6002 hw
->phy
.ops
.set_phy_power(hw
, false);
6004 ixgbe_release_hw_control(adapter
);
6006 if (!test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
))
6007 pci_disable_device(pdev
);
6013 static int ixgbe_suspend(struct pci_dev
*pdev
, pm_message_t state
)
6018 retval
= __ixgbe_shutdown(pdev
, &wake
);
6023 pci_prepare_to_sleep(pdev
);
6025 pci_wake_from_d3(pdev
, false);
6026 pci_set_power_state(pdev
, PCI_D3hot
);
6031 #endif /* CONFIG_PM */
6033 static void ixgbe_shutdown(struct pci_dev
*pdev
)
6037 __ixgbe_shutdown(pdev
, &wake
);
6039 if (system_state
== SYSTEM_POWER_OFF
) {
6040 pci_wake_from_d3(pdev
, wake
);
6041 pci_set_power_state(pdev
, PCI_D3hot
);
6046 * ixgbe_update_stats - Update the board statistics counters.
6047 * @adapter: board private structure
6049 void ixgbe_update_stats(struct ixgbe_adapter
*adapter
)
6051 struct net_device
*netdev
= adapter
->netdev
;
6052 struct ixgbe_hw
*hw
= &adapter
->hw
;
6053 struct ixgbe_hw_stats
*hwstats
= &adapter
->stats
;
6055 u32 i
, missed_rx
= 0, mpc
, bprc
, lxon
, lxoff
, xon_off_tot
;
6056 u64 non_eop_descs
= 0, restart_queue
= 0, tx_busy
= 0;
6057 u64 alloc_rx_page_failed
= 0, alloc_rx_buff_failed
= 0;
6058 u64 bytes
= 0, packets
= 0, hw_csum_rx_error
= 0;
6060 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6061 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6064 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
) {
6067 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
6068 rsc_count
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_count
;
6069 rsc_flush
+= adapter
->rx_ring
[i
]->rx_stats
.rsc_flush
;
6071 adapter
->rsc_total_count
= rsc_count
;
6072 adapter
->rsc_total_flush
= rsc_flush
;
6075 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
6076 struct ixgbe_ring
*rx_ring
= adapter
->rx_ring
[i
];
6077 non_eop_descs
+= rx_ring
->rx_stats
.non_eop_descs
;
6078 alloc_rx_page_failed
+= rx_ring
->rx_stats
.alloc_rx_page_failed
;
6079 alloc_rx_buff_failed
+= rx_ring
->rx_stats
.alloc_rx_buff_failed
;
6080 hw_csum_rx_error
+= rx_ring
->rx_stats
.csum_err
;
6081 bytes
+= rx_ring
->stats
.bytes
;
6082 packets
+= rx_ring
->stats
.packets
;
6084 adapter
->non_eop_descs
= non_eop_descs
;
6085 adapter
->alloc_rx_page_failed
= alloc_rx_page_failed
;
6086 adapter
->alloc_rx_buff_failed
= alloc_rx_buff_failed
;
6087 adapter
->hw_csum_rx_error
= hw_csum_rx_error
;
6088 netdev
->stats
.rx_bytes
= bytes
;
6089 netdev
->stats
.rx_packets
= packets
;
6093 /* gather some stats to the adapter struct that are per queue */
6094 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
6095 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
6096 restart_queue
+= tx_ring
->tx_stats
.restart_queue
;
6097 tx_busy
+= tx_ring
->tx_stats
.tx_busy
;
6098 bytes
+= tx_ring
->stats
.bytes
;
6099 packets
+= tx_ring
->stats
.packets
;
6101 adapter
->restart_queue
= restart_queue
;
6102 adapter
->tx_busy
= tx_busy
;
6103 netdev
->stats
.tx_bytes
= bytes
;
6104 netdev
->stats
.tx_packets
= packets
;
6106 hwstats
->crcerrs
+= IXGBE_READ_REG(hw
, IXGBE_CRCERRS
);
6108 /* 8 register reads */
6109 for (i
= 0; i
< 8; i
++) {
6110 /* for packet buffers not used, the register should read 0 */
6111 mpc
= IXGBE_READ_REG(hw
, IXGBE_MPC(i
));
6113 hwstats
->mpc
[i
] += mpc
;
6114 total_mpc
+= hwstats
->mpc
[i
];
6115 hwstats
->pxontxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXONTXC(i
));
6116 hwstats
->pxofftxc
[i
] += IXGBE_READ_REG(hw
, IXGBE_PXOFFTXC(i
));
6117 switch (hw
->mac
.type
) {
6118 case ixgbe_mac_82598EB
:
6119 hwstats
->rnbc
[i
] += IXGBE_READ_REG(hw
, IXGBE_RNBC(i
));
6120 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC(i
));
6121 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC(i
));
6122 hwstats
->pxonrxc
[i
] +=
6123 IXGBE_READ_REG(hw
, IXGBE_PXONRXC(i
));
6125 case ixgbe_mac_82599EB
:
6126 case ixgbe_mac_X540
:
6127 case ixgbe_mac_X550
:
6128 case ixgbe_mac_X550EM_x
:
6129 hwstats
->pxonrxc
[i
] +=
6130 IXGBE_READ_REG(hw
, IXGBE_PXONRXCNT(i
));
6137 /*16 register reads */
6138 for (i
= 0; i
< 16; i
++) {
6139 hwstats
->qptc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPTC(i
));
6140 hwstats
->qprc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QPRC(i
));
6141 if ((hw
->mac
.type
== ixgbe_mac_82599EB
) ||
6142 (hw
->mac
.type
== ixgbe_mac_X540
) ||
6143 (hw
->mac
.type
== ixgbe_mac_X550
) ||
6144 (hw
->mac
.type
== ixgbe_mac_X550EM_x
)) {
6145 hwstats
->qbtc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBTC_L(i
));
6146 IXGBE_READ_REG(hw
, IXGBE_QBTC_H(i
)); /* to clear */
6147 hwstats
->qbrc
[i
] += IXGBE_READ_REG(hw
, IXGBE_QBRC_L(i
));
6148 IXGBE_READ_REG(hw
, IXGBE_QBRC_H(i
)); /* to clear */
6152 hwstats
->gprc
+= IXGBE_READ_REG(hw
, IXGBE_GPRC
);
6153 /* work around hardware counting issue */
6154 hwstats
->gprc
-= missed_rx
;
6156 ixgbe_update_xoff_received(adapter
);
6158 /* 82598 hardware only has a 32 bit counter in the high register */
6159 switch (hw
->mac
.type
) {
6160 case ixgbe_mac_82598EB
:
6161 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXC
);
6162 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCH
);
6163 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCH
);
6164 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORH
);
6166 case ixgbe_mac_X540
:
6167 case ixgbe_mac_X550
:
6168 case ixgbe_mac_X550EM_x
:
6169 /* OS2BMC stats are X540 and later */
6170 hwstats
->o2bgptc
+= IXGBE_READ_REG(hw
, IXGBE_O2BGPTC
);
6171 hwstats
->o2bspc
+= IXGBE_READ_REG(hw
, IXGBE_O2BSPC
);
6172 hwstats
->b2ospc
+= IXGBE_READ_REG(hw
, IXGBE_B2OSPC
);
6173 hwstats
->b2ogprc
+= IXGBE_READ_REG(hw
, IXGBE_B2OGPRC
);
6174 case ixgbe_mac_82599EB
:
6175 for (i
= 0; i
< 16; i
++)
6176 adapter
->hw_rx_no_dma_resources
+=
6177 IXGBE_READ_REG(hw
, IXGBE_QPRDC(i
));
6178 hwstats
->gorc
+= IXGBE_READ_REG(hw
, IXGBE_GORCL
);
6179 IXGBE_READ_REG(hw
, IXGBE_GORCH
); /* to clear */
6180 hwstats
->gotc
+= IXGBE_READ_REG(hw
, IXGBE_GOTCL
);
6181 IXGBE_READ_REG(hw
, IXGBE_GOTCH
); /* to clear */
6182 hwstats
->tor
+= IXGBE_READ_REG(hw
, IXGBE_TORL
);
6183 IXGBE_READ_REG(hw
, IXGBE_TORH
); /* to clear */
6184 hwstats
->lxonrxc
+= IXGBE_READ_REG(hw
, IXGBE_LXONRXCNT
);
6185 hwstats
->fdirmatch
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMATCH
);
6186 hwstats
->fdirmiss
+= IXGBE_READ_REG(hw
, IXGBE_FDIRMISS
);
6188 hwstats
->fccrc
+= IXGBE_READ_REG(hw
, IXGBE_FCCRC
);
6189 hwstats
->fcoerpdc
+= IXGBE_READ_REG(hw
, IXGBE_FCOERPDC
);
6190 hwstats
->fcoeprc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPRC
);
6191 hwstats
->fcoeptc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEPTC
);
6192 hwstats
->fcoedwrc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWRC
);
6193 hwstats
->fcoedwtc
+= IXGBE_READ_REG(hw
, IXGBE_FCOEDWTC
);
6194 /* Add up per cpu counters for total ddp aloc fail */
6195 if (adapter
->fcoe
.ddp_pool
) {
6196 struct ixgbe_fcoe
*fcoe
= &adapter
->fcoe
;
6197 struct ixgbe_fcoe_ddp_pool
*ddp_pool
;
6199 u64 noddp
= 0, noddp_ext_buff
= 0;
6200 for_each_possible_cpu(cpu
) {
6201 ddp_pool
= per_cpu_ptr(fcoe
->ddp_pool
, cpu
);
6202 noddp
+= ddp_pool
->noddp
;
6203 noddp_ext_buff
+= ddp_pool
->noddp_ext_buff
;
6205 hwstats
->fcoe_noddp
= noddp
;
6206 hwstats
->fcoe_noddp_ext_buff
= noddp_ext_buff
;
6208 #endif /* IXGBE_FCOE */
6213 bprc
= IXGBE_READ_REG(hw
, IXGBE_BPRC
);
6214 hwstats
->bprc
+= bprc
;
6215 hwstats
->mprc
+= IXGBE_READ_REG(hw
, IXGBE_MPRC
);
6216 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
6217 hwstats
->mprc
-= bprc
;
6218 hwstats
->roc
+= IXGBE_READ_REG(hw
, IXGBE_ROC
);
6219 hwstats
->prc64
+= IXGBE_READ_REG(hw
, IXGBE_PRC64
);
6220 hwstats
->prc127
+= IXGBE_READ_REG(hw
, IXGBE_PRC127
);
6221 hwstats
->prc255
+= IXGBE_READ_REG(hw
, IXGBE_PRC255
);
6222 hwstats
->prc511
+= IXGBE_READ_REG(hw
, IXGBE_PRC511
);
6223 hwstats
->prc1023
+= IXGBE_READ_REG(hw
, IXGBE_PRC1023
);
6224 hwstats
->prc1522
+= IXGBE_READ_REG(hw
, IXGBE_PRC1522
);
6225 hwstats
->rlec
+= IXGBE_READ_REG(hw
, IXGBE_RLEC
);
6226 lxon
= IXGBE_READ_REG(hw
, IXGBE_LXONTXC
);
6227 hwstats
->lxontxc
+= lxon
;
6228 lxoff
= IXGBE_READ_REG(hw
, IXGBE_LXOFFTXC
);
6229 hwstats
->lxofftxc
+= lxoff
;
6230 hwstats
->gptc
+= IXGBE_READ_REG(hw
, IXGBE_GPTC
);
6231 hwstats
->mptc
+= IXGBE_READ_REG(hw
, IXGBE_MPTC
);
6233 * 82598 errata - tx of flow control packets is included in tx counters
6235 xon_off_tot
= lxon
+ lxoff
;
6236 hwstats
->gptc
-= xon_off_tot
;
6237 hwstats
->mptc
-= xon_off_tot
;
6238 hwstats
->gotc
-= (xon_off_tot
* (ETH_ZLEN
+ ETH_FCS_LEN
));
6239 hwstats
->ruc
+= IXGBE_READ_REG(hw
, IXGBE_RUC
);
6240 hwstats
->rfc
+= IXGBE_READ_REG(hw
, IXGBE_RFC
);
6241 hwstats
->rjc
+= IXGBE_READ_REG(hw
, IXGBE_RJC
);
6242 hwstats
->tpr
+= IXGBE_READ_REG(hw
, IXGBE_TPR
);
6243 hwstats
->ptc64
+= IXGBE_READ_REG(hw
, IXGBE_PTC64
);
6244 hwstats
->ptc64
-= xon_off_tot
;
6245 hwstats
->ptc127
+= IXGBE_READ_REG(hw
, IXGBE_PTC127
);
6246 hwstats
->ptc255
+= IXGBE_READ_REG(hw
, IXGBE_PTC255
);
6247 hwstats
->ptc511
+= IXGBE_READ_REG(hw
, IXGBE_PTC511
);
6248 hwstats
->ptc1023
+= IXGBE_READ_REG(hw
, IXGBE_PTC1023
);
6249 hwstats
->ptc1522
+= IXGBE_READ_REG(hw
, IXGBE_PTC1522
);
6250 hwstats
->bptc
+= IXGBE_READ_REG(hw
, IXGBE_BPTC
);
6252 /* Fill out the OS statistics structure */
6253 netdev
->stats
.multicast
= hwstats
->mprc
;
6256 netdev
->stats
.rx_errors
= hwstats
->crcerrs
+ hwstats
->rlec
;
6257 netdev
->stats
.rx_dropped
= 0;
6258 netdev
->stats
.rx_length_errors
= hwstats
->rlec
;
6259 netdev
->stats
.rx_crc_errors
= hwstats
->crcerrs
;
6260 netdev
->stats
.rx_missed_errors
= total_mpc
;
6264 * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
6265 * @adapter: pointer to the device adapter structure
6267 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter
*adapter
)
6269 struct ixgbe_hw
*hw
= &adapter
->hw
;
6272 if (!(adapter
->flags2
& IXGBE_FLAG2_FDIR_REQUIRES_REINIT
))
6275 adapter
->flags2
&= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT
;
6277 /* if interface is down do nothing */
6278 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
6281 /* do nothing if we are not using signature filters */
6282 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_HASH_CAPABLE
))
6285 adapter
->fdir_overflow
++;
6287 if (ixgbe_reinit_fdir_tables_82599(hw
) == 0) {
6288 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
6289 set_bit(__IXGBE_TX_FDIR_INIT_DONE
,
6290 &(adapter
->tx_ring
[i
]->state
));
6291 /* re-enable flow director interrupts */
6292 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMS_FLOW_DIR
);
6294 e_err(probe
, "failed to finish FDIR re-initialization, "
6295 "ignored adding FDIR ATR filters\n");
6300 * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
6301 * @adapter: pointer to the device adapter structure
6303 * This function serves two purposes. First it strobes the interrupt lines
6304 * in order to make certain interrupts are occurring. Secondly it sets the
6305 * bits needed to check for TX hangs. As a result we should immediately
6306 * determine if a hang has occurred.
6308 static void ixgbe_check_hang_subtask(struct ixgbe_adapter
*adapter
)
6310 struct ixgbe_hw
*hw
= &adapter
->hw
;
6314 /* If we're down, removing or resetting, just bail */
6315 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6316 test_bit(__IXGBE_REMOVING
, &adapter
->state
) ||
6317 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6320 /* Force detection of hung controller */
6321 if (netif_carrier_ok(adapter
->netdev
)) {
6322 for (i
= 0; i
< adapter
->num_tx_queues
; i
++)
6323 set_check_for_tx_hang(adapter
->tx_ring
[i
]);
6326 if (!(adapter
->flags
& IXGBE_FLAG_MSIX_ENABLED
)) {
6328 * for legacy and MSI interrupts don't set any bits
6329 * that are enabled for EIAM, because this operation
6330 * would set *both* EIMS and EICS for any bit in EIAM
6332 IXGBE_WRITE_REG(hw
, IXGBE_EICS
,
6333 (IXGBE_EICS_TCP_TIMER
| IXGBE_EICS_OTHER
));
6335 /* get one bit for every active tx/rx interrupt vector */
6336 for (i
= 0; i
< adapter
->num_q_vectors
; i
++) {
6337 struct ixgbe_q_vector
*qv
= adapter
->q_vector
[i
];
6338 if (qv
->rx
.ring
|| qv
->tx
.ring
)
6339 eics
|= ((u64
)1 << i
);
6343 /* Cause software interrupt to ensure rings are cleaned */
6344 ixgbe_irq_rearm_queues(adapter
, eics
);
6348 * ixgbe_watchdog_update_link - update the link status
6349 * @adapter: pointer to the device adapter structure
6350 * @link_speed: pointer to a u32 to store the link_speed
6352 static void ixgbe_watchdog_update_link(struct ixgbe_adapter
*adapter
)
6354 struct ixgbe_hw
*hw
= &adapter
->hw
;
6355 u32 link_speed
= adapter
->link_speed
;
6356 bool link_up
= adapter
->link_up
;
6357 bool pfc_en
= adapter
->dcb_cfg
.pfc_mode_enable
;
6359 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
))
6362 if (hw
->mac
.ops
.check_link
) {
6363 hw
->mac
.ops
.check_link(hw
, &link_speed
, &link_up
, false);
6365 /* always assume link is up, if no check link function */
6366 link_speed
= IXGBE_LINK_SPEED_10GB_FULL
;
6370 if (adapter
->ixgbe_ieee_pfc
)
6371 pfc_en
|= !!(adapter
->ixgbe_ieee_pfc
->pfc_en
);
6373 if (link_up
&& !((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) && pfc_en
)) {
6374 hw
->mac
.ops
.fc_enable(hw
);
6375 ixgbe_set_rx_drop_en(adapter
);
6379 time_after(jiffies
, (adapter
->link_check_timeout
+
6380 IXGBE_TRY_LINK_TIMEOUT
))) {
6381 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_UPDATE
;
6382 IXGBE_WRITE_REG(hw
, IXGBE_EIMS
, IXGBE_EIMC_LSC
);
6383 IXGBE_WRITE_FLUSH(hw
);
6386 adapter
->link_up
= link_up
;
6387 adapter
->link_speed
= link_speed
;
6390 static void ixgbe_update_default_up(struct ixgbe_adapter
*adapter
)
6392 #ifdef CONFIG_IXGBE_DCB
6393 struct net_device
*netdev
= adapter
->netdev
;
6394 struct dcb_app app
= {
6395 .selector
= IEEE_8021QAZ_APP_SEL_ETHERTYPE
,
6400 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_IEEE
)
6401 up
= dcb_ieee_getapp_mask(netdev
, &app
);
6403 adapter
->default_up
= (up
> 1) ? (ffs(up
) - 1) : 0;
6408 * ixgbe_watchdog_link_is_up - update netif_carrier status and
6409 * print link up message
6410 * @adapter: pointer to the device adapter structure
6412 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter
*adapter
)
6414 struct net_device
*netdev
= adapter
->netdev
;
6415 struct ixgbe_hw
*hw
= &adapter
->hw
;
6416 struct net_device
*upper
;
6417 struct list_head
*iter
;
6418 u32 link_speed
= adapter
->link_speed
;
6419 const char *speed_str
;
6420 bool flow_rx
, flow_tx
;
6422 /* only continue if link was previously down */
6423 if (netif_carrier_ok(netdev
))
6426 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
6428 switch (hw
->mac
.type
) {
6429 case ixgbe_mac_82598EB
: {
6430 u32 frctl
= IXGBE_READ_REG(hw
, IXGBE_FCTRL
);
6431 u32 rmcs
= IXGBE_READ_REG(hw
, IXGBE_RMCS
);
6432 flow_rx
= !!(frctl
& IXGBE_FCTRL_RFCE
);
6433 flow_tx
= !!(rmcs
& IXGBE_RMCS_TFCE_802_3X
);
6436 case ixgbe_mac_X540
:
6437 case ixgbe_mac_X550
:
6438 case ixgbe_mac_X550EM_x
:
6439 case ixgbe_mac_82599EB
: {
6440 u32 mflcn
= IXGBE_READ_REG(hw
, IXGBE_MFLCN
);
6441 u32 fccfg
= IXGBE_READ_REG(hw
, IXGBE_FCCFG
);
6442 flow_rx
= !!(mflcn
& IXGBE_MFLCN_RFCE
);
6443 flow_tx
= !!(fccfg
& IXGBE_FCCFG_TFCE_802_3X
);
6452 adapter
->last_rx_ptp_check
= jiffies
;
6454 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
))
6455 ixgbe_ptp_start_cyclecounter(adapter
);
6457 switch (link_speed
) {
6458 case IXGBE_LINK_SPEED_10GB_FULL
:
6459 speed_str
= "10 Gbps";
6461 case IXGBE_LINK_SPEED_2_5GB_FULL
:
6462 speed_str
= "2.5 Gbps";
6464 case IXGBE_LINK_SPEED_1GB_FULL
:
6465 speed_str
= "1 Gbps";
6467 case IXGBE_LINK_SPEED_100_FULL
:
6468 speed_str
= "100 Mbps";
6471 speed_str
= "unknown speed";
6474 e_info(drv
, "NIC Link is Up %s, Flow Control: %s\n", speed_str
,
6475 ((flow_rx
&& flow_tx
) ? "RX/TX" :
6477 (flow_tx
? "TX" : "None"))));
6479 netif_carrier_on(netdev
);
6480 ixgbe_check_vf_rate_limit(adapter
);
6482 /* enable transmits */
6483 netif_tx_wake_all_queues(adapter
->netdev
);
6485 /* enable any upper devices */
6487 netdev_for_each_all_upper_dev_rcu(adapter
->netdev
, upper
, iter
) {
6488 if (netif_is_macvlan(upper
)) {
6489 struct macvlan_dev
*vlan
= netdev_priv(upper
);
6492 netif_tx_wake_all_queues(upper
);
6497 /* update the default user priority for VFs */
6498 ixgbe_update_default_up(adapter
);
6500 /* ping all the active vfs to let them know link has changed */
6501 ixgbe_ping_all_vfs(adapter
);
6505 * ixgbe_watchdog_link_is_down - update netif_carrier status and
6506 * print link down message
6507 * @adapter: pointer to the adapter structure
6509 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter
*adapter
)
6511 struct net_device
*netdev
= adapter
->netdev
;
6512 struct ixgbe_hw
*hw
= &adapter
->hw
;
6514 adapter
->link_up
= false;
6515 adapter
->link_speed
= 0;
6517 /* only continue if link was up previously */
6518 if (!netif_carrier_ok(netdev
))
6521 /* poll for SFP+ cable when link is down */
6522 if (ixgbe_is_sfp(hw
) && hw
->mac
.type
== ixgbe_mac_82598EB
)
6523 adapter
->flags2
|= IXGBE_FLAG2_SEARCH_FOR_SFP
;
6525 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
))
6526 ixgbe_ptp_start_cyclecounter(adapter
);
6528 e_info(drv
, "NIC Link is Down\n");
6529 netif_carrier_off(netdev
);
6531 /* ping all the active vfs to let them know link has changed */
6532 ixgbe_ping_all_vfs(adapter
);
6535 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter
*adapter
)
6539 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
6540 struct ixgbe_ring
*tx_ring
= adapter
->tx_ring
[i
];
6542 if (tx_ring
->next_to_use
!= tx_ring
->next_to_clean
)
6549 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter
*adapter
)
6551 struct ixgbe_hw
*hw
= &adapter
->hw
;
6552 struct ixgbe_ring_feature
*vmdq
= &adapter
->ring_feature
[RING_F_VMDQ
];
6553 u32 q_per_pool
= __ALIGN_MASK(1, ~vmdq
->mask
);
6557 if (!adapter
->num_vfs
)
6560 /* resetting the PF is only needed for MAC before X550 */
6561 if (hw
->mac
.type
>= ixgbe_mac_X550
)
6564 for (i
= 0; i
< adapter
->num_vfs
; i
++) {
6565 for (j
= 0; j
< q_per_pool
; j
++) {
6568 h
= IXGBE_READ_REG(hw
, IXGBE_PVFTDHN(q_per_pool
, i
, j
));
6569 t
= IXGBE_READ_REG(hw
, IXGBE_PVFTDTN(q_per_pool
, i
, j
));
6580 * ixgbe_watchdog_flush_tx - flush queues on link down
6581 * @adapter: pointer to the device adapter structure
6583 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter
*adapter
)
6585 if (!netif_carrier_ok(adapter
->netdev
)) {
6586 if (ixgbe_ring_tx_pending(adapter
) ||
6587 ixgbe_vf_tx_pending(adapter
)) {
6588 /* We've lost link, so the controller stops DMA,
6589 * but we've got queued Tx work that's never going
6590 * to get done, so reset controller to flush Tx.
6591 * (Do the reset outside of interrupt context).
6593 e_warn(drv
, "initiating reset to clear Tx work after link loss\n");
6594 adapter
->flags2
|= IXGBE_FLAG2_RESET_REQUESTED
;
6599 #ifdef CONFIG_PCI_IOV
6600 static inline void ixgbe_issue_vf_flr(struct ixgbe_adapter
*adapter
,
6601 struct pci_dev
*vfdev
)
6603 if (!pci_wait_for_pending_transaction(vfdev
))
6604 e_dev_warn("Issuing VFLR with pending transactions\n");
6606 e_dev_err("Issuing VFLR for VF %s\n", pci_name(vfdev
));
6607 pcie_capability_set_word(vfdev
, PCI_EXP_DEVCTL
, PCI_EXP_DEVCTL_BCR_FLR
);
6612 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter
*adapter
)
6614 struct ixgbe_hw
*hw
= &adapter
->hw
;
6615 struct pci_dev
*pdev
= adapter
->pdev
;
6616 struct pci_dev
*vfdev
;
6619 unsigned short vf_id
;
6621 if (!(netif_carrier_ok(adapter
->netdev
)))
6624 gpc
= IXGBE_READ_REG(hw
, IXGBE_TXDGPC
);
6625 if (gpc
) /* If incrementing then no need for the check below */
6627 /* Check to see if a bad DMA write target from an errant or
6628 * malicious VF has caused a PCIe error. If so then we can
6629 * issue a VFLR to the offending VF(s) and then resume without
6630 * requesting a full slot reset.
6636 pos
= pci_find_ext_capability(pdev
, PCI_EXT_CAP_ID_SRIOV
);
6640 /* get the device ID for the VF */
6641 pci_read_config_word(pdev
, pos
+ PCI_SRIOV_VF_DID
, &vf_id
);
6643 /* check status reg for all VFs owned by this PF */
6644 vfdev
= pci_get_device(pdev
->vendor
, vf_id
, NULL
);
6646 if (vfdev
->is_virtfn
&& (vfdev
->physfn
== pdev
)) {
6649 pci_read_config_word(vfdev
, PCI_STATUS
, &status_reg
);
6650 if (status_reg
& PCI_STATUS_REC_MASTER_ABORT
)
6652 ixgbe_issue_vf_flr(adapter
, vfdev
);
6655 vfdev
= pci_get_device(pdev
->vendor
, vf_id
, vfdev
);
6659 static void ixgbe_spoof_check(struct ixgbe_adapter
*adapter
)
6663 /* Do not perform spoof check for 82598 or if not in IOV mode */
6664 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
6665 adapter
->num_vfs
== 0)
6668 ssvpc
= IXGBE_READ_REG(&adapter
->hw
, IXGBE_SSVPC
);
6671 * ssvpc register is cleared on read, if zero then no
6672 * spoofed packets in the last interval.
6677 e_warn(drv
, "%u Spoofed packets detected\n", ssvpc
);
6680 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused
*adapter
)
6685 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused
*adapter
)
6688 #endif /* CONFIG_PCI_IOV */
6692 * ixgbe_watchdog_subtask - check and bring link up
6693 * @adapter: pointer to the device adapter structure
6695 static void ixgbe_watchdog_subtask(struct ixgbe_adapter
*adapter
)
6697 /* if interface is down, removing or resetting, do nothing */
6698 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6699 test_bit(__IXGBE_REMOVING
, &adapter
->state
) ||
6700 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6703 ixgbe_watchdog_update_link(adapter
);
6705 if (adapter
->link_up
)
6706 ixgbe_watchdog_link_is_up(adapter
);
6708 ixgbe_watchdog_link_is_down(adapter
);
6710 ixgbe_check_for_bad_vf(adapter
);
6711 ixgbe_spoof_check(adapter
);
6712 ixgbe_update_stats(adapter
);
6714 ixgbe_watchdog_flush_tx(adapter
);
6718 * ixgbe_sfp_detection_subtask - poll for SFP+ cable
6719 * @adapter: the ixgbe adapter structure
6721 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter
*adapter
)
6723 struct ixgbe_hw
*hw
= &adapter
->hw
;
6726 /* not searching for SFP so there is nothing to do here */
6727 if (!(adapter
->flags2
& IXGBE_FLAG2_SEARCH_FOR_SFP
) &&
6728 !(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
6731 if (adapter
->sfp_poll_time
&&
6732 time_after(adapter
->sfp_poll_time
, jiffies
))
6733 return; /* If not yet time to poll for SFP */
6735 /* someone else is in init, wait until next service event */
6736 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
6739 adapter
->sfp_poll_time
= jiffies
+ IXGBE_SFP_POLL_JIFFIES
- 1;
6741 err
= hw
->phy
.ops
.identify_sfp(hw
);
6742 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
6745 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
) {
6746 /* If no cable is present, then we need to reset
6747 * the next time we find a good cable. */
6748 adapter
->flags2
|= IXGBE_FLAG2_SFP_NEEDS_RESET
;
6755 /* exit if reset not needed */
6756 if (!(adapter
->flags2
& IXGBE_FLAG2_SFP_NEEDS_RESET
))
6759 adapter
->flags2
&= ~IXGBE_FLAG2_SFP_NEEDS_RESET
;
6762 * A module may be identified correctly, but the EEPROM may not have
6763 * support for that module. setup_sfp() will fail in that case, so
6764 * we should not allow that module to load.
6766 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
6767 err
= hw
->phy
.ops
.reset(hw
);
6769 err
= hw
->mac
.ops
.setup_sfp(hw
);
6771 if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
)
6774 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_CONFIG
;
6775 e_info(probe
, "detected SFP+: %d\n", hw
->phy
.sfp_type
);
6778 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
6780 if ((err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) &&
6781 (adapter
->netdev
->reg_state
== NETREG_REGISTERED
)) {
6782 e_dev_err("failed to initialize because an unsupported "
6783 "SFP+ module type was detected.\n");
6784 e_dev_err("Reload the driver after installing a "
6785 "supported module.\n");
6786 unregister_netdev(adapter
->netdev
);
6791 * ixgbe_sfp_link_config_subtask - set up link SFP after module install
6792 * @adapter: the ixgbe adapter structure
6794 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter
*adapter
)
6796 struct ixgbe_hw
*hw
= &adapter
->hw
;
6798 bool autoneg
= false;
6800 if (!(adapter
->flags
& IXGBE_FLAG_NEED_LINK_CONFIG
))
6803 /* someone else is in init, wait until next service event */
6804 if (test_and_set_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
))
6807 adapter
->flags
&= ~IXGBE_FLAG_NEED_LINK_CONFIG
;
6809 speed
= hw
->phy
.autoneg_advertised
;
6810 if ((!speed
) && (hw
->mac
.ops
.get_link_capabilities
)) {
6811 hw
->mac
.ops
.get_link_capabilities(hw
, &speed
, &autoneg
);
6813 /* setup the highest link when no autoneg */
6815 if (speed
& IXGBE_LINK_SPEED_10GB_FULL
)
6816 speed
= IXGBE_LINK_SPEED_10GB_FULL
;
6820 if (hw
->mac
.ops
.setup_link
)
6821 hw
->mac
.ops
.setup_link(hw
, speed
, true);
6823 adapter
->flags
|= IXGBE_FLAG_NEED_LINK_UPDATE
;
6824 adapter
->link_check_timeout
= jiffies
;
6825 clear_bit(__IXGBE_IN_SFP_INIT
, &adapter
->state
);
6829 * ixgbe_service_timer - Timer Call-back
6830 * @data: pointer to adapter cast into an unsigned long
6832 static void ixgbe_service_timer(unsigned long data
)
6834 struct ixgbe_adapter
*adapter
= (struct ixgbe_adapter
*)data
;
6835 unsigned long next_event_offset
;
6837 /* poll faster when waiting for link */
6838 if (adapter
->flags
& IXGBE_FLAG_NEED_LINK_UPDATE
)
6839 next_event_offset
= HZ
/ 10;
6841 next_event_offset
= HZ
* 2;
6843 /* Reset the timer */
6844 mod_timer(&adapter
->service_timer
, next_event_offset
+ jiffies
);
6846 ixgbe_service_event_schedule(adapter
);
6849 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter
*adapter
)
6851 struct ixgbe_hw
*hw
= &adapter
->hw
;
6854 if (!(adapter
->flags2
& IXGBE_FLAG2_PHY_INTERRUPT
))
6857 adapter
->flags2
&= ~IXGBE_FLAG2_PHY_INTERRUPT
;
6859 if (!hw
->phy
.ops
.handle_lasi
)
6862 status
= hw
->phy
.ops
.handle_lasi(&adapter
->hw
);
6863 if (status
!= IXGBE_ERR_OVERTEMP
)
6866 e_crit(drv
, "%s\n", ixgbe_overheat_msg
);
6869 static void ixgbe_reset_subtask(struct ixgbe_adapter
*adapter
)
6871 if (!(adapter
->flags2
& IXGBE_FLAG2_RESET_REQUESTED
))
6874 adapter
->flags2
&= ~IXGBE_FLAG2_RESET_REQUESTED
;
6876 /* If we're already down, removing or resetting, just bail */
6877 if (test_bit(__IXGBE_DOWN
, &adapter
->state
) ||
6878 test_bit(__IXGBE_REMOVING
, &adapter
->state
) ||
6879 test_bit(__IXGBE_RESETTING
, &adapter
->state
))
6882 ixgbe_dump(adapter
);
6883 netdev_err(adapter
->netdev
, "Reset adapter\n");
6884 adapter
->tx_timeout_count
++;
6887 ixgbe_reinit_locked(adapter
);
6892 * ixgbe_service_task - manages and runs subtasks
6893 * @work: pointer to work_struct containing our data
6895 static void ixgbe_service_task(struct work_struct
*work
)
6897 struct ixgbe_adapter
*adapter
= container_of(work
,
6898 struct ixgbe_adapter
,
6900 if (ixgbe_removed(adapter
->hw
.hw_addr
)) {
6901 if (!test_bit(__IXGBE_DOWN
, &adapter
->state
)) {
6903 ixgbe_down(adapter
);
6906 ixgbe_service_event_complete(adapter
);
6909 #ifdef CONFIG_IXGBE_VXLAN
6910 if (adapter
->flags2
& IXGBE_FLAG2_VXLAN_REREG_NEEDED
) {
6911 adapter
->flags2
&= ~IXGBE_FLAG2_VXLAN_REREG_NEEDED
;
6912 vxlan_get_rx_port(adapter
->netdev
);
6914 #endif /* CONFIG_IXGBE_VXLAN */
6915 ixgbe_reset_subtask(adapter
);
6916 ixgbe_phy_interrupt_subtask(adapter
);
6917 ixgbe_sfp_detection_subtask(adapter
);
6918 ixgbe_sfp_link_config_subtask(adapter
);
6919 ixgbe_check_overtemp_subtask(adapter
);
6920 ixgbe_watchdog_subtask(adapter
);
6921 ixgbe_fdir_reinit_subtask(adapter
);
6922 ixgbe_check_hang_subtask(adapter
);
6924 if (test_bit(__IXGBE_PTP_RUNNING
, &adapter
->state
)) {
6925 ixgbe_ptp_overflow_check(adapter
);
6926 ixgbe_ptp_rx_hang(adapter
);
6929 ixgbe_service_event_complete(adapter
);
6932 static int ixgbe_tso(struct ixgbe_ring
*tx_ring
,
6933 struct ixgbe_tx_buffer
*first
,
6936 struct sk_buff
*skb
= first
->skb
;
6937 u32 vlan_macip_lens
, type_tucmd
;
6938 u32 mss_l4len_idx
, l4len
;
6941 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
6944 if (!skb_is_gso(skb
))
6947 err
= skb_cow_head(skb
, 0);
6951 /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
6952 type_tucmd
= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
6954 if (first
->protocol
== htons(ETH_P_IP
)) {
6955 struct iphdr
*iph
= ip_hdr(skb
);
6958 tcp_hdr(skb
)->check
= ~csum_tcpudp_magic(iph
->saddr
,
6962 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
6963 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6964 IXGBE_TX_FLAGS_CSUM
|
6965 IXGBE_TX_FLAGS_IPV4
;
6966 } else if (skb_is_gso_v6(skb
)) {
6967 ipv6_hdr(skb
)->payload_len
= 0;
6968 tcp_hdr(skb
)->check
=
6969 ~csum_ipv6_magic(&ipv6_hdr(skb
)->saddr
,
6970 &ipv6_hdr(skb
)->daddr
,
6972 first
->tx_flags
|= IXGBE_TX_FLAGS_TSO
|
6973 IXGBE_TX_FLAGS_CSUM
;
6976 /* compute header lengths */
6977 l4len
= tcp_hdrlen(skb
);
6978 *hdr_len
= skb_transport_offset(skb
) + l4len
;
6980 /* update gso size and bytecount with header size */
6981 first
->gso_segs
= skb_shinfo(skb
)->gso_segs
;
6982 first
->bytecount
+= (first
->gso_segs
- 1) * *hdr_len
;
6984 /* mss_l4len_id: use 0 as index for TSO */
6985 mss_l4len_idx
= l4len
<< IXGBE_ADVTXD_L4LEN_SHIFT
;
6986 mss_l4len_idx
|= skb_shinfo(skb
)->gso_size
<< IXGBE_ADVTXD_MSS_SHIFT
;
6988 /* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
6989 vlan_macip_lens
= skb_network_header_len(skb
);
6990 vlan_macip_lens
|= skb_network_offset(skb
) << IXGBE_ADVTXD_MACLEN_SHIFT
;
6991 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
6993 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0, type_tucmd
,
6999 static void ixgbe_tx_csum(struct ixgbe_ring
*tx_ring
,
7000 struct ixgbe_tx_buffer
*first
)
7002 struct sk_buff
*skb
= first
->skb
;
7003 u32 vlan_macip_lens
= 0;
7004 u32 mss_l4len_idx
= 0;
7007 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
7008 if (!(first
->tx_flags
& IXGBE_TX_FLAGS_HW_VLAN
) &&
7009 !(first
->tx_flags
& IXGBE_TX_FLAGS_CC
))
7011 vlan_macip_lens
= skb_network_offset(skb
) <<
7012 IXGBE_ADVTXD_MACLEN_SHIFT
;
7017 struct ipv6hdr
*ipv6
;
7021 struct tcphdr
*tcphdr
;
7025 if (skb
->encapsulation
) {
7026 network_hdr
.raw
= skb_inner_network_header(skb
);
7027 transport_hdr
.raw
= skb_inner_transport_header(skb
);
7028 vlan_macip_lens
= skb_inner_network_offset(skb
) <<
7029 IXGBE_ADVTXD_MACLEN_SHIFT
;
7031 network_hdr
.raw
= skb_network_header(skb
);
7032 transport_hdr
.raw
= skb_transport_header(skb
);
7033 vlan_macip_lens
= skb_network_offset(skb
) <<
7034 IXGBE_ADVTXD_MACLEN_SHIFT
;
7037 /* use first 4 bits to determine IP version */
7038 switch (network_hdr
.ipv4
->version
) {
7040 vlan_macip_lens
|= transport_hdr
.raw
- network_hdr
.raw
;
7041 type_tucmd
|= IXGBE_ADVTXD_TUCMD_IPV4
;
7042 l4_hdr
= network_hdr
.ipv4
->protocol
;
7045 vlan_macip_lens
|= transport_hdr
.raw
- network_hdr
.raw
;
7046 l4_hdr
= network_hdr
.ipv6
->nexthdr
;
7049 if (unlikely(net_ratelimit())) {
7050 dev_warn(tx_ring
->dev
,
7051 "partial checksum but version=%d\n",
7052 network_hdr
.ipv4
->version
);
7058 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_TCP
;
7059 mss_l4len_idx
= (transport_hdr
.tcphdr
->doff
* 4) <<
7060 IXGBE_ADVTXD_L4LEN_SHIFT
;
7063 type_tucmd
|= IXGBE_ADVTXD_TUCMD_L4T_SCTP
;
7064 mss_l4len_idx
= sizeof(struct sctphdr
) <<
7065 IXGBE_ADVTXD_L4LEN_SHIFT
;
7068 mss_l4len_idx
= sizeof(struct udphdr
) <<
7069 IXGBE_ADVTXD_L4LEN_SHIFT
;
7072 if (unlikely(net_ratelimit())) {
7073 dev_warn(tx_ring
->dev
,
7074 "partial checksum but l4 proto=%x!\n",
7080 /* update TX checksum flag */
7081 first
->tx_flags
|= IXGBE_TX_FLAGS_CSUM
;
7084 /* vlan_macip_lens: MACLEN, VLAN tag */
7085 vlan_macip_lens
|= first
->tx_flags
& IXGBE_TX_FLAGS_VLAN_MASK
;
7087 ixgbe_tx_ctxtdesc(tx_ring
, vlan_macip_lens
, 0,
7088 type_tucmd
, mss_l4len_idx
);
7091 #define IXGBE_SET_FLAG(_input, _flag, _result) \
7092 ((_flag <= _result) ? \
7093 ((u32)(_input & _flag) * (_result / _flag)) : \
7094 ((u32)(_input & _flag) / (_flag / _result)))
7096 static u32
ixgbe_tx_cmd_type(struct sk_buff
*skb
, u32 tx_flags
)
7098 /* set type for advanced descriptor with frame checksum insertion */
7099 u32 cmd_type
= IXGBE_ADVTXD_DTYP_DATA
|
7100 IXGBE_ADVTXD_DCMD_DEXT
|
7101 IXGBE_ADVTXD_DCMD_IFCS
;
7103 /* set HW vlan bit if vlan is present */
7104 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_HW_VLAN
,
7105 IXGBE_ADVTXD_DCMD_VLE
);
7107 /* set segmentation enable bits for TSO/FSO */
7108 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSO
,
7109 IXGBE_ADVTXD_DCMD_TSE
);
7111 /* set timestamp bit if present */
7112 cmd_type
|= IXGBE_SET_FLAG(tx_flags
, IXGBE_TX_FLAGS_TSTAMP
,
7113 IXGBE_ADVTXD_MAC_TSTAMP
);
7115 /* insert frame checksum */
7116 cmd_type
^= IXGBE_SET_FLAG(skb
->no_fcs
, 1, IXGBE_ADVTXD_DCMD_IFCS
);
7121 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc
*tx_desc
,
7122 u32 tx_flags
, unsigned int paylen
)
7124 u32 olinfo_status
= paylen
<< IXGBE_ADVTXD_PAYLEN_SHIFT
;
7126 /* enable L4 checksum for TSO and TX checksum offload */
7127 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
7128 IXGBE_TX_FLAGS_CSUM
,
7129 IXGBE_ADVTXD_POPTS_TXSM
);
7131 /* enble IPv4 checksum for TSO */
7132 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
7133 IXGBE_TX_FLAGS_IPV4
,
7134 IXGBE_ADVTXD_POPTS_IXSM
);
7137 * Check Context must be set if Tx switch is enabled, which it
7138 * always is for case where virtual functions are running
7140 olinfo_status
|= IXGBE_SET_FLAG(tx_flags
,
7144 tx_desc
->read
.olinfo_status
= cpu_to_le32(olinfo_status
);
7147 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
7149 netif_stop_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
7151 /* Herbert's original patch had:
7152 * smp_mb__after_netif_stop_queue();
7153 * but since that doesn't exist yet, just open code it.
7157 /* We need to check again in a case another CPU has just
7158 * made room available.
7160 if (likely(ixgbe_desc_unused(tx_ring
) < size
))
7163 /* A reprieve! - use start_queue because it doesn't call schedule */
7164 netif_start_subqueue(tx_ring
->netdev
, tx_ring
->queue_index
);
7165 ++tx_ring
->tx_stats
.restart_queue
;
7169 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring
*tx_ring
, u16 size
)
7171 if (likely(ixgbe_desc_unused(tx_ring
) >= size
))
7174 return __ixgbe_maybe_stop_tx(tx_ring
, size
);
7177 #define IXGBE_TXD_CMD (IXGBE_TXD_CMD_EOP | \
7180 static void ixgbe_tx_map(struct ixgbe_ring
*tx_ring
,
7181 struct ixgbe_tx_buffer
*first
,
7184 struct sk_buff
*skb
= first
->skb
;
7185 struct ixgbe_tx_buffer
*tx_buffer
;
7186 union ixgbe_adv_tx_desc
*tx_desc
;
7187 struct skb_frag_struct
*frag
;
7189 unsigned int data_len
, size
;
7190 u32 tx_flags
= first
->tx_flags
;
7191 u32 cmd_type
= ixgbe_tx_cmd_type(skb
, tx_flags
);
7192 u16 i
= tx_ring
->next_to_use
;
7194 tx_desc
= IXGBE_TX_DESC(tx_ring
, i
);
7196 ixgbe_tx_olinfo_status(tx_desc
, tx_flags
, skb
->len
- hdr_len
);
7198 size
= skb_headlen(skb
);
7199 data_len
= skb
->data_len
;
7202 if (tx_flags
& IXGBE_TX_FLAGS_FCOE
) {
7203 if (data_len
< sizeof(struct fcoe_crc_eof
)) {
7204 size
-= sizeof(struct fcoe_crc_eof
) - data_len
;
7207 data_len
-= sizeof(struct fcoe_crc_eof
);
7212 dma
= dma_map_single(tx_ring
->dev
, skb
->data
, size
, DMA_TO_DEVICE
);
7216 for (frag
= &skb_shinfo(skb
)->frags
[0];; frag
++) {
7217 if (dma_mapping_error(tx_ring
->dev
, dma
))
7220 /* record length, and DMA address */
7221 dma_unmap_len_set(tx_buffer
, len
, size
);
7222 dma_unmap_addr_set(tx_buffer
, dma
, dma
);
7224 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
7226 while (unlikely(size
> IXGBE_MAX_DATA_PER_TXD
)) {
7227 tx_desc
->read
.cmd_type_len
=
7228 cpu_to_le32(cmd_type
^ IXGBE_MAX_DATA_PER_TXD
);
7232 if (i
== tx_ring
->count
) {
7233 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
7236 tx_desc
->read
.olinfo_status
= 0;
7238 dma
+= IXGBE_MAX_DATA_PER_TXD
;
7239 size
-= IXGBE_MAX_DATA_PER_TXD
;
7241 tx_desc
->read
.buffer_addr
= cpu_to_le64(dma
);
7244 if (likely(!data_len
))
7247 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
^ size
);
7251 if (i
== tx_ring
->count
) {
7252 tx_desc
= IXGBE_TX_DESC(tx_ring
, 0);
7255 tx_desc
->read
.olinfo_status
= 0;
7258 size
= min_t(unsigned int, data_len
, skb_frag_size(frag
));
7260 size
= skb_frag_size(frag
);
7264 dma
= skb_frag_dma_map(tx_ring
->dev
, frag
, 0, size
,
7267 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
7270 /* write last descriptor with RS and EOP bits */
7271 cmd_type
|= size
| IXGBE_TXD_CMD
;
7272 tx_desc
->read
.cmd_type_len
= cpu_to_le32(cmd_type
);
7274 netdev_tx_sent_queue(txring_txq(tx_ring
), first
->bytecount
);
7276 /* set the timestamp */
7277 first
->time_stamp
= jiffies
;
7280 * Force memory writes to complete before letting h/w know there
7281 * are new descriptors to fetch. (Only applicable for weak-ordered
7282 * memory model archs, such as IA-64).
7284 * We also need this memory barrier to make certain all of the
7285 * status bits have been updated before next_to_watch is written.
7289 /* set next_to_watch value indicating a packet is present */
7290 first
->next_to_watch
= tx_desc
;
7293 if (i
== tx_ring
->count
)
7296 tx_ring
->next_to_use
= i
;
7298 ixgbe_maybe_stop_tx(tx_ring
, DESC_NEEDED
);
7300 if (netif_xmit_stopped(txring_txq(tx_ring
)) || !skb
->xmit_more
) {
7301 writel(i
, tx_ring
->tail
);
7303 /* we need this if more than one processor can write to our tail
7304 * at a time, it synchronizes IO on IA64/Altix systems
7311 dev_err(tx_ring
->dev
, "TX DMA map failed\n");
7313 /* clear dma mappings for failed tx_buffer_info map */
7315 tx_buffer
= &tx_ring
->tx_buffer_info
[i
];
7316 ixgbe_unmap_and_free_tx_resource(tx_ring
, tx_buffer
);
7317 if (tx_buffer
== first
)
7324 tx_ring
->next_to_use
= i
;
7327 static void ixgbe_atr(struct ixgbe_ring
*ring
,
7328 struct ixgbe_tx_buffer
*first
)
7330 struct ixgbe_q_vector
*q_vector
= ring
->q_vector
;
7331 union ixgbe_atr_hash_dword input
= { .dword
= 0 };
7332 union ixgbe_atr_hash_dword common
= { .dword
= 0 };
7334 unsigned char *network
;
7336 struct ipv6hdr
*ipv6
;
7339 struct sk_buff
*skb
;
7340 #ifdef CONFIG_IXGBE_VXLAN
7342 #endif /* CONFIG_IXGBE_VXLAN */
7345 /* if ring doesn't have a interrupt vector, cannot perform ATR */
7349 /* do nothing if sampling is disabled */
7350 if (!ring
->atr_sample_rate
)
7355 /* snag network header to get L4 type and address */
7357 hdr
.network
= skb_network_header(skb
);
7358 if (skb
->encapsulation
) {
7359 #ifdef CONFIG_IXGBE_VXLAN
7360 struct ixgbe_adapter
*adapter
= q_vector
->adapter
;
7362 if (!adapter
->vxlan_port
)
7364 if (first
->protocol
!= htons(ETH_P_IP
) ||
7365 hdr
.ipv4
->version
!= IPVERSION
||
7366 hdr
.ipv4
->protocol
!= IPPROTO_UDP
) {
7369 if (ntohs(udp_hdr(skb
)->dest
) != adapter
->vxlan_port
)
7372 hdr
.network
= skb_inner_network_header(skb
);
7373 th
= inner_tcp_hdr(skb
);
7376 #endif /* CONFIG_IXGBE_VXLAN */
7378 /* Currently only IPv4/IPv6 with TCP is supported */
7379 if ((first
->protocol
!= htons(ETH_P_IPV6
) ||
7380 hdr
.ipv6
->nexthdr
!= IPPROTO_TCP
) &&
7381 (first
->protocol
!= htons(ETH_P_IP
) ||
7382 hdr
.ipv4
->protocol
!= IPPROTO_TCP
))
7387 /* skip this packet since it is invalid or the socket is closing */
7391 /* sample on all syn packets or once every atr sample count */
7392 if (!th
->syn
&& (ring
->atr_count
< ring
->atr_sample_rate
))
7395 /* reset sample count */
7396 ring
->atr_count
= 0;
7398 vlan_id
= htons(first
->tx_flags
>> IXGBE_TX_FLAGS_VLAN_SHIFT
);
7401 * src and dst are inverted, think how the receiver sees them
7403 * The input is broken into two sections, a non-compressed section
7404 * containing vm_pool, vlan_id, and flow_type. The rest of the data
7405 * is XORed together and stored in the compressed dword.
7407 input
.formatted
.vlan_id
= vlan_id
;
7410 * since src port and flex bytes occupy the same word XOR them together
7411 * and write the value to source port portion of compressed dword
7413 if (first
->tx_flags
& (IXGBE_TX_FLAGS_SW_VLAN
| IXGBE_TX_FLAGS_HW_VLAN
))
7414 common
.port
.src
^= th
->dest
^ htons(ETH_P_8021Q
);
7416 common
.port
.src
^= th
->dest
^ first
->protocol
;
7417 common
.port
.dst
^= th
->source
;
7419 if (first
->protocol
== htons(ETH_P_IP
)) {
7420 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV4
;
7421 common
.ip
^= hdr
.ipv4
->saddr
^ hdr
.ipv4
->daddr
;
7423 input
.formatted
.flow_type
= IXGBE_ATR_FLOW_TYPE_TCPV6
;
7424 common
.ip
^= hdr
.ipv6
->saddr
.s6_addr32
[0] ^
7425 hdr
.ipv6
->saddr
.s6_addr32
[1] ^
7426 hdr
.ipv6
->saddr
.s6_addr32
[2] ^
7427 hdr
.ipv6
->saddr
.s6_addr32
[3] ^
7428 hdr
.ipv6
->daddr
.s6_addr32
[0] ^
7429 hdr
.ipv6
->daddr
.s6_addr32
[1] ^
7430 hdr
.ipv6
->daddr
.s6_addr32
[2] ^
7431 hdr
.ipv6
->daddr
.s6_addr32
[3];
7434 #ifdef CONFIG_IXGBE_VXLAN
7436 input
.formatted
.flow_type
|= IXGBE_ATR_L4TYPE_TUNNEL_MASK
;
7437 #endif /* CONFIG_IXGBE_VXLAN */
7439 /* This assumes the Rx queue and Tx queue are bound to the same CPU */
7440 ixgbe_fdir_add_signature_filter_82599(&q_vector
->adapter
->hw
,
7441 input
, common
, ring
->queue_index
);
7444 static u16
ixgbe_select_queue(struct net_device
*dev
, struct sk_buff
*skb
,
7445 void *accel_priv
, select_queue_fallback_t fallback
)
7447 struct ixgbe_fwd_adapter
*fwd_adapter
= accel_priv
;
7449 struct ixgbe_adapter
*adapter
;
7450 struct ixgbe_ring_feature
*f
;
7455 return skb
->queue_mapping
+ fwd_adapter
->tx_base_queue
;
7460 * only execute the code below if protocol is FCoE
7461 * or FIP and we have FCoE enabled on the adapter
7463 switch (vlan_get_protocol(skb
)) {
7464 case htons(ETH_P_FCOE
):
7465 case htons(ETH_P_FIP
):
7466 adapter
= netdev_priv(dev
);
7468 if (adapter
->flags
& IXGBE_FLAG_FCOE_ENABLED
)
7471 return fallback(dev
, skb
);
7474 f
= &adapter
->ring_feature
[RING_F_FCOE
];
7476 txq
= skb_rx_queue_recorded(skb
) ? skb_get_rx_queue(skb
) :
7479 while (txq
>= f
->indices
)
7482 return txq
+ f
->offset
;
7484 return fallback(dev
, skb
);
7488 netdev_tx_t
ixgbe_xmit_frame_ring(struct sk_buff
*skb
,
7489 struct ixgbe_adapter
*adapter
,
7490 struct ixgbe_ring
*tx_ring
)
7492 struct ixgbe_tx_buffer
*first
;
7496 u16 count
= TXD_USE_COUNT(skb_headlen(skb
));
7497 __be16 protocol
= skb
->protocol
;
7501 * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
7502 * + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
7503 * + 2 desc gap to keep tail from touching head,
7504 * + 1 desc for context descriptor,
7505 * otherwise try next time
7507 for (f
= 0; f
< skb_shinfo(skb
)->nr_frags
; f
++)
7508 count
+= TXD_USE_COUNT(skb_shinfo(skb
)->frags
[f
].size
);
7510 if (ixgbe_maybe_stop_tx(tx_ring
, count
+ 3)) {
7511 tx_ring
->tx_stats
.tx_busy
++;
7512 return NETDEV_TX_BUSY
;
7515 /* record the location of the first descriptor for this packet */
7516 first
= &tx_ring
->tx_buffer_info
[tx_ring
->next_to_use
];
7518 first
->bytecount
= skb
->len
;
7519 first
->gso_segs
= 1;
7521 /* if we have a HW VLAN tag being added default to the HW one */
7522 if (skb_vlan_tag_present(skb
)) {
7523 tx_flags
|= skb_vlan_tag_get(skb
) << IXGBE_TX_FLAGS_VLAN_SHIFT
;
7524 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
7525 /* else if it is a SW VLAN check the next protocol and store the tag */
7526 } else if (protocol
== htons(ETH_P_8021Q
)) {
7527 struct vlan_hdr
*vhdr
, _vhdr
;
7528 vhdr
= skb_header_pointer(skb
, ETH_HLEN
, sizeof(_vhdr
), &_vhdr
);
7532 tx_flags
|= ntohs(vhdr
->h_vlan_TCI
) <<
7533 IXGBE_TX_FLAGS_VLAN_SHIFT
;
7534 tx_flags
|= IXGBE_TX_FLAGS_SW_VLAN
;
7536 protocol
= vlan_get_protocol(skb
);
7538 if (unlikely(skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP
) &&
7539 adapter
->ptp_clock
&&
7540 !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS
,
7542 skb_shinfo(skb
)->tx_flags
|= SKBTX_IN_PROGRESS
;
7543 tx_flags
|= IXGBE_TX_FLAGS_TSTAMP
;
7545 /* schedule check for Tx timestamp */
7546 adapter
->ptp_tx_skb
= skb_get(skb
);
7547 adapter
->ptp_tx_start
= jiffies
;
7548 schedule_work(&adapter
->ptp_tx_work
);
7551 skb_tx_timestamp(skb
);
7553 #ifdef CONFIG_PCI_IOV
7555 * Use the l2switch_enable flag - would be false if the DMA
7556 * Tx switch had been disabled.
7558 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
7559 tx_flags
|= IXGBE_TX_FLAGS_CC
;
7562 /* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
7563 if ((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) &&
7564 ((tx_flags
& (IXGBE_TX_FLAGS_HW_VLAN
| IXGBE_TX_FLAGS_SW_VLAN
)) ||
7565 (skb
->priority
!= TC_PRIO_CONTROL
))) {
7566 tx_flags
&= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK
;
7567 tx_flags
|= (skb
->priority
& 0x7) <<
7568 IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT
;
7569 if (tx_flags
& IXGBE_TX_FLAGS_SW_VLAN
) {
7570 struct vlan_ethhdr
*vhdr
;
7572 if (skb_cow_head(skb
, 0))
7574 vhdr
= (struct vlan_ethhdr
*)skb
->data
;
7575 vhdr
->h_vlan_TCI
= htons(tx_flags
>>
7576 IXGBE_TX_FLAGS_VLAN_SHIFT
);
7578 tx_flags
|= IXGBE_TX_FLAGS_HW_VLAN
;
7582 /* record initial flags and protocol */
7583 first
->tx_flags
= tx_flags
;
7584 first
->protocol
= protocol
;
7587 /* setup tx offload for FCoE */
7588 if ((protocol
== htons(ETH_P_FCOE
)) &&
7589 (tx_ring
->netdev
->features
& (NETIF_F_FSO
| NETIF_F_FCOE_CRC
))) {
7590 tso
= ixgbe_fso(tx_ring
, first
, &hdr_len
);
7597 #endif /* IXGBE_FCOE */
7598 tso
= ixgbe_tso(tx_ring
, first
, &hdr_len
);
7602 ixgbe_tx_csum(tx_ring
, first
);
7604 /* add the ATR filter if ATR is on */
7605 if (test_bit(__IXGBE_TX_FDIR_INIT_DONE
, &tx_ring
->state
))
7606 ixgbe_atr(tx_ring
, first
);
7610 #endif /* IXGBE_FCOE */
7611 ixgbe_tx_map(tx_ring
, first
, hdr_len
);
7613 return NETDEV_TX_OK
;
7616 dev_kfree_skb_any(first
->skb
);
7619 return NETDEV_TX_OK
;
7622 static netdev_tx_t
__ixgbe_xmit_frame(struct sk_buff
*skb
,
7623 struct net_device
*netdev
,
7624 struct ixgbe_ring
*ring
)
7626 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7627 struct ixgbe_ring
*tx_ring
;
7630 * The minimum packet size for olinfo paylen is 17 so pad the skb
7631 * in order to meet this minimum size requirement.
7633 if (skb_put_padto(skb
, 17))
7634 return NETDEV_TX_OK
;
7636 tx_ring
= ring
? ring
: adapter
->tx_ring
[skb
->queue_mapping
];
7638 return ixgbe_xmit_frame_ring(skb
, adapter
, tx_ring
);
7641 static netdev_tx_t
ixgbe_xmit_frame(struct sk_buff
*skb
,
7642 struct net_device
*netdev
)
7644 return __ixgbe_xmit_frame(skb
, netdev
, NULL
);
7648 * ixgbe_set_mac - Change the Ethernet Address of the NIC
7649 * @netdev: network interface device structure
7650 * @p: pointer to an address structure
7652 * Returns 0 on success, negative on failure
7654 static int ixgbe_set_mac(struct net_device
*netdev
, void *p
)
7656 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7657 struct ixgbe_hw
*hw
= &adapter
->hw
;
7658 struct sockaddr
*addr
= p
;
7661 if (!is_valid_ether_addr(addr
->sa_data
))
7662 return -EADDRNOTAVAIL
;
7664 ixgbe_del_mac_filter(adapter
, hw
->mac
.addr
, VMDQ_P(0));
7665 memcpy(netdev
->dev_addr
, addr
->sa_data
, netdev
->addr_len
);
7666 memcpy(hw
->mac
.addr
, addr
->sa_data
, netdev
->addr_len
);
7668 ret
= ixgbe_add_mac_filter(adapter
, hw
->mac
.addr
, VMDQ_P(0));
7669 return ret
> 0 ? 0 : ret
;
7673 ixgbe_mdio_read(struct net_device
*netdev
, int prtad
, int devad
, u16 addr
)
7675 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7676 struct ixgbe_hw
*hw
= &adapter
->hw
;
7680 if (prtad
!= hw
->phy
.mdio
.prtad
)
7682 rc
= hw
->phy
.ops
.read_reg(hw
, addr
, devad
, &value
);
7688 static int ixgbe_mdio_write(struct net_device
*netdev
, int prtad
, int devad
,
7689 u16 addr
, u16 value
)
7691 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7692 struct ixgbe_hw
*hw
= &adapter
->hw
;
7694 if (prtad
!= hw
->phy
.mdio
.prtad
)
7696 return hw
->phy
.ops
.write_reg(hw
, addr
, devad
, value
);
7699 static int ixgbe_ioctl(struct net_device
*netdev
, struct ifreq
*req
, int cmd
)
7701 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7705 return ixgbe_ptp_set_ts_config(adapter
, req
);
7707 return ixgbe_ptp_get_ts_config(adapter
, req
);
7709 return mdio_mii_ioctl(&adapter
->hw
.phy
.mdio
, if_mii(req
), cmd
);
7714 * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
7716 * @netdev: network interface device structure
7718 * Returns non-zero on failure
7720 static int ixgbe_add_sanmac_netdev(struct net_device
*dev
)
7723 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7724 struct ixgbe_hw
*hw
= &adapter
->hw
;
7726 if (is_valid_ether_addr(hw
->mac
.san_addr
)) {
7728 err
= dev_addr_add(dev
, hw
->mac
.san_addr
, NETDEV_HW_ADDR_T_SAN
);
7731 /* update SAN MAC vmdq pool selection */
7732 hw
->mac
.ops
.set_vmdq_san_mac(hw
, VMDQ_P(0));
7738 * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
7740 * @netdev: network interface device structure
7742 * Returns non-zero on failure
7744 static int ixgbe_del_sanmac_netdev(struct net_device
*dev
)
7747 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7748 struct ixgbe_mac_info
*mac
= &adapter
->hw
.mac
;
7750 if (is_valid_ether_addr(mac
->san_addr
)) {
7752 err
= dev_addr_del(dev
, mac
->san_addr
, NETDEV_HW_ADDR_T_SAN
);
7758 #ifdef CONFIG_NET_POLL_CONTROLLER
7760 * Polling 'interrupt' - used by things like netconsole to send skbs
7761 * without having to re-enable interrupts. It's not called while
7762 * the interrupt routine is executing.
7764 static void ixgbe_netpoll(struct net_device
*netdev
)
7766 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7769 /* if interface is down do nothing */
7770 if (test_bit(__IXGBE_DOWN
, &adapter
->state
))
7773 /* loop through and schedule all active queues */
7774 for (i
= 0; i
< adapter
->num_q_vectors
; i
++)
7775 ixgbe_msix_clean_rings(0, adapter
->q_vector
[i
]);
7779 static struct rtnl_link_stats64
*ixgbe_get_stats64(struct net_device
*netdev
,
7780 struct rtnl_link_stats64
*stats
)
7782 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7786 for (i
= 0; i
< adapter
->num_rx_queues
; i
++) {
7787 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->rx_ring
[i
]);
7793 start
= u64_stats_fetch_begin_irq(&ring
->syncp
);
7794 packets
= ring
->stats
.packets
;
7795 bytes
= ring
->stats
.bytes
;
7796 } while (u64_stats_fetch_retry_irq(&ring
->syncp
, start
));
7797 stats
->rx_packets
+= packets
;
7798 stats
->rx_bytes
+= bytes
;
7802 for (i
= 0; i
< adapter
->num_tx_queues
; i
++) {
7803 struct ixgbe_ring
*ring
= ACCESS_ONCE(adapter
->tx_ring
[i
]);
7809 start
= u64_stats_fetch_begin_irq(&ring
->syncp
);
7810 packets
= ring
->stats
.packets
;
7811 bytes
= ring
->stats
.bytes
;
7812 } while (u64_stats_fetch_retry_irq(&ring
->syncp
, start
));
7813 stats
->tx_packets
+= packets
;
7814 stats
->tx_bytes
+= bytes
;
7818 /* following stats updated by ixgbe_watchdog_task() */
7819 stats
->multicast
= netdev
->stats
.multicast
;
7820 stats
->rx_errors
= netdev
->stats
.rx_errors
;
7821 stats
->rx_length_errors
= netdev
->stats
.rx_length_errors
;
7822 stats
->rx_crc_errors
= netdev
->stats
.rx_crc_errors
;
7823 stats
->rx_missed_errors
= netdev
->stats
.rx_missed_errors
;
7827 #ifdef CONFIG_IXGBE_DCB
7829 * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
7830 * @adapter: pointer to ixgbe_adapter
7831 * @tc: number of traffic classes currently enabled
7833 * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
7834 * 802.1Q priority maps to a packet buffer that exists.
7836 static void ixgbe_validate_rtr(struct ixgbe_adapter
*adapter
, u8 tc
)
7838 struct ixgbe_hw
*hw
= &adapter
->hw
;
7842 /* 82598 have a static priority to TC mapping that can not
7843 * be changed so no validation is needed.
7845 if (hw
->mac
.type
== ixgbe_mac_82598EB
)
7848 reg
= IXGBE_READ_REG(hw
, IXGBE_RTRUP2TC
);
7851 for (i
= 0; i
< MAX_TRAFFIC_CLASS
; i
++) {
7852 u8 up2tc
= reg
>> (i
* IXGBE_RTRUP2TC_UP_SHIFT
);
7854 /* If up2tc is out of bounds default to zero */
7856 reg
&= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT
);
7860 IXGBE_WRITE_REG(hw
, IXGBE_RTRUP2TC
, reg
);
7866 * ixgbe_set_prio_tc_map - Configure netdev prio tc map
7867 * @adapter: Pointer to adapter struct
7869 * Populate the netdev user priority to tc map
7871 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter
*adapter
)
7873 struct net_device
*dev
= adapter
->netdev
;
7874 struct ixgbe_dcb_config
*dcb_cfg
= &adapter
->dcb_cfg
;
7875 struct ieee_ets
*ets
= adapter
->ixgbe_ieee_ets
;
7878 for (prio
= 0; prio
< MAX_USER_PRIORITY
; prio
++) {
7881 if (adapter
->dcbx_cap
& DCB_CAP_DCBX_VER_CEE
)
7882 tc
= ixgbe_dcb_get_tc_from_up(dcb_cfg
, 0, prio
);
7884 tc
= ets
->prio_tc
[prio
];
7886 netdev_set_prio_tc_map(dev
, prio
, tc
);
7890 #endif /* CONFIG_IXGBE_DCB */
7892 * ixgbe_setup_tc - configure net_device for multiple traffic classes
7894 * @netdev: net device to configure
7895 * @tc: number of traffic classes to enable
7897 int ixgbe_setup_tc(struct net_device
*dev
, u8 tc
)
7899 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
7900 struct ixgbe_hw
*hw
= &adapter
->hw
;
7903 /* Hardware supports up to 8 traffic classes */
7904 if (tc
> adapter
->dcb_cfg
.num_tcs
.pg_tcs
)
7907 if (hw
->mac
.type
== ixgbe_mac_82598EB
&& tc
&& tc
< MAX_TRAFFIC_CLASS
)
7910 pools
= (find_first_zero_bit(&adapter
->fwd_bitmask
, 32) > 1);
7911 if (tc
&& pools
&& adapter
->num_rx_pools
> IXGBE_MAX_DCBMACVLANS
)
7914 /* Hardware has to reinitialize queues and interrupts to
7915 * match packet buffer alignment. Unfortunately, the
7916 * hardware is not flexible enough to do this dynamically.
7918 if (netif_running(dev
))
7920 ixgbe_clear_interrupt_scheme(adapter
);
7922 #ifdef CONFIG_IXGBE_DCB
7924 netdev_set_num_tc(dev
, tc
);
7925 ixgbe_set_prio_tc_map(adapter
);
7927 adapter
->flags
|= IXGBE_FLAG_DCB_ENABLED
;
7929 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
) {
7930 adapter
->last_lfc_mode
= adapter
->hw
.fc
.requested_mode
;
7931 adapter
->hw
.fc
.requested_mode
= ixgbe_fc_none
;
7934 netdev_reset_tc(dev
);
7936 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
7937 adapter
->hw
.fc
.requested_mode
= adapter
->last_lfc_mode
;
7939 adapter
->flags
&= ~IXGBE_FLAG_DCB_ENABLED
;
7941 adapter
->temp_dcb_cfg
.pfc_mode_enable
= false;
7942 adapter
->dcb_cfg
.pfc_mode_enable
= false;
7945 ixgbe_validate_rtr(adapter
, tc
);
7947 #endif /* CONFIG_IXGBE_DCB */
7948 ixgbe_init_interrupt_scheme(adapter
);
7950 if (netif_running(dev
))
7951 return ixgbe_open(dev
);
7956 #ifdef CONFIG_PCI_IOV
7957 void ixgbe_sriov_reinit(struct ixgbe_adapter
*adapter
)
7959 struct net_device
*netdev
= adapter
->netdev
;
7962 ixgbe_setup_tc(netdev
, netdev_get_num_tc(netdev
));
7967 void ixgbe_do_reset(struct net_device
*netdev
)
7969 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7971 if (netif_running(netdev
))
7972 ixgbe_reinit_locked(adapter
);
7974 ixgbe_reset(adapter
);
7977 static netdev_features_t
ixgbe_fix_features(struct net_device
*netdev
,
7978 netdev_features_t features
)
7980 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7982 /* If Rx checksum is disabled, then RSC/LRO should also be disabled */
7983 if (!(features
& NETIF_F_RXCSUM
))
7984 features
&= ~NETIF_F_LRO
;
7986 /* Turn off LRO if not RSC capable */
7987 if (!(adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
))
7988 features
&= ~NETIF_F_LRO
;
7993 static int ixgbe_set_features(struct net_device
*netdev
,
7994 netdev_features_t features
)
7996 struct ixgbe_adapter
*adapter
= netdev_priv(netdev
);
7997 netdev_features_t changed
= netdev
->features
^ features
;
7998 bool need_reset
= false;
8000 /* Make sure RSC matches LRO, reset if change */
8001 if (!(features
& NETIF_F_LRO
)) {
8002 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
8004 adapter
->flags2
&= ~IXGBE_FLAG2_RSC_ENABLED
;
8005 } else if ((adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
) &&
8006 !(adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)) {
8007 if (adapter
->rx_itr_setting
== 1 ||
8008 adapter
->rx_itr_setting
> IXGBE_MIN_RSC_ITR
) {
8009 adapter
->flags2
|= IXGBE_FLAG2_RSC_ENABLED
;
8011 } else if ((changed
^ features
) & NETIF_F_LRO
) {
8012 e_info(probe
, "rx-usecs set too low, "
8018 * Check if Flow Director n-tuple support was enabled or disabled. If
8019 * the state changed, we need to reset.
8021 switch (features
& NETIF_F_NTUPLE
) {
8022 case NETIF_F_NTUPLE
:
8023 /* turn off ATR, enable perfect filters and reset */
8024 if (!(adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
))
8027 adapter
->flags
&= ~IXGBE_FLAG_FDIR_HASH_CAPABLE
;
8028 adapter
->flags
|= IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
8031 /* turn off perfect filters, enable ATR and reset */
8032 if (adapter
->flags
& IXGBE_FLAG_FDIR_PERFECT_CAPABLE
)
8035 adapter
->flags
&= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE
;
8037 /* We cannot enable ATR if SR-IOV is enabled */
8038 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
)
8041 /* We cannot enable ATR if we have 2 or more traffic classes */
8042 if (netdev_get_num_tc(netdev
) > 1)
8045 /* We cannot enable ATR if RSS is disabled */
8046 if (adapter
->ring_feature
[RING_F_RSS
].limit
<= 1)
8049 /* A sample rate of 0 indicates ATR disabled */
8050 if (!adapter
->atr_sample_rate
)
8053 adapter
->flags
|= IXGBE_FLAG_FDIR_HASH_CAPABLE
;
8057 if (features
& NETIF_F_HW_VLAN_CTAG_RX
)
8058 ixgbe_vlan_strip_enable(adapter
);
8060 ixgbe_vlan_strip_disable(adapter
);
8062 if (changed
& NETIF_F_RXALL
)
8065 netdev
->features
= features
;
8067 #ifdef CONFIG_IXGBE_VXLAN
8068 if ((adapter
->flags
& IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
)) {
8069 if (features
& NETIF_F_RXCSUM
)
8070 adapter
->flags2
|= IXGBE_FLAG2_VXLAN_REREG_NEEDED
;
8072 ixgbe_clear_vxlan_port(adapter
);
8074 #endif /* CONFIG_IXGBE_VXLAN */
8077 ixgbe_do_reset(netdev
);
8082 #ifdef CONFIG_IXGBE_VXLAN
8084 * ixgbe_add_vxlan_port - Get notifications about VXLAN ports that come up
8085 * @dev: The port's netdev
8086 * @sa_family: Socket Family that VXLAN is notifiying us about
8087 * @port: New UDP port number that VXLAN started listening to
8089 static void ixgbe_add_vxlan_port(struct net_device
*dev
, sa_family_t sa_family
,
8092 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8093 struct ixgbe_hw
*hw
= &adapter
->hw
;
8094 u16 new_port
= ntohs(port
);
8096 if (!(adapter
->flags
& IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
))
8099 if (sa_family
== AF_INET6
)
8102 if (adapter
->vxlan_port
== new_port
)
8105 if (adapter
->vxlan_port
) {
8107 "Hit Max num of VXLAN ports, not adding port %d\n",
8112 adapter
->vxlan_port
= new_port
;
8113 IXGBE_WRITE_REG(hw
, IXGBE_VXLANCTRL
, new_port
);
8117 * ixgbe_del_vxlan_port - Get notifications about VXLAN ports that go away
8118 * @dev: The port's netdev
8119 * @sa_family: Socket Family that VXLAN is notifying us about
8120 * @port: UDP port number that VXLAN stopped listening to
8122 static void ixgbe_del_vxlan_port(struct net_device
*dev
, sa_family_t sa_family
,
8125 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8126 u16 new_port
= ntohs(port
);
8128 if (!(adapter
->flags
& IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE
))
8131 if (sa_family
== AF_INET6
)
8134 if (adapter
->vxlan_port
!= new_port
) {
8135 netdev_info(dev
, "Port %d was not found, not deleting\n",
8140 ixgbe_clear_vxlan_port(adapter
);
8141 adapter
->flags2
|= IXGBE_FLAG2_VXLAN_REREG_NEEDED
;
8143 #endif /* CONFIG_IXGBE_VXLAN */
8145 static int ixgbe_ndo_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
8146 struct net_device
*dev
,
8147 const unsigned char *addr
, u16 vid
,
8150 /* guarantee we can provide a unique filter for the unicast address */
8151 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
)) {
8152 if (IXGBE_MAX_PF_MACVLANS
<= netdev_uc_count(dev
))
8156 return ndo_dflt_fdb_add(ndm
, tb
, dev
, addr
, vid
, flags
);
8160 * ixgbe_configure_bridge_mode - set various bridge modes
8161 * @adapter - the private structure
8162 * @mode - requested bridge mode
8164 * Configure some settings require for various bridge modes.
8166 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter
*adapter
,
8169 struct ixgbe_hw
*hw
= &adapter
->hw
;
8170 unsigned int p
, num_pools
;
8174 case BRIDGE_MODE_VEPA
:
8175 /* disable Tx loopback, rely on switch hairpin mode */
8176 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_PFDTXGSWC
, 0);
8178 /* must enable Rx switching replication to allow multicast
8179 * packet reception on all VFs, and to enable source address
8182 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VMD_CTL
);
8183 vmdctl
|= IXGBE_VT_CTL_REPLEN
;
8184 IXGBE_WRITE_REG(hw
, IXGBE_VMD_CTL
, vmdctl
);
8186 /* enable Rx source address pruning. Note, this requires
8187 * replication to be enabled or else it does nothing.
8189 num_pools
= adapter
->num_vfs
+ adapter
->num_rx_pools
;
8190 for (p
= 0; p
< num_pools
; p
++) {
8191 if (hw
->mac
.ops
.set_source_address_pruning
)
8192 hw
->mac
.ops
.set_source_address_pruning(hw
,
8197 case BRIDGE_MODE_VEB
:
8198 /* enable Tx loopback for internal VF/PF communication */
8199 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_PFDTXGSWC
,
8200 IXGBE_PFDTXGSWC_VT_LBEN
);
8202 /* disable Rx switching replication unless we have SR-IOV
8205 vmdctl
= IXGBE_READ_REG(hw
, IXGBE_VMD_CTL
);
8206 if (!adapter
->num_vfs
)
8207 vmdctl
&= ~IXGBE_VT_CTL_REPLEN
;
8208 IXGBE_WRITE_REG(hw
, IXGBE_VMD_CTL
, vmdctl
);
8210 /* disable Rx source address pruning, since we don't expect to
8211 * be receiving external loopback of our transmitted frames.
8213 num_pools
= adapter
->num_vfs
+ adapter
->num_rx_pools
;
8214 for (p
= 0; p
< num_pools
; p
++) {
8215 if (hw
->mac
.ops
.set_source_address_pruning
)
8216 hw
->mac
.ops
.set_source_address_pruning(hw
,
8225 adapter
->bridge_mode
= mode
;
8227 e_info(drv
, "enabling bridge mode: %s\n",
8228 mode
== BRIDGE_MODE_VEPA
? "VEPA" : "VEB");
8233 static int ixgbe_ndo_bridge_setlink(struct net_device
*dev
,
8234 struct nlmsghdr
*nlh
, u16 flags
)
8236 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8237 struct nlattr
*attr
, *br_spec
;
8240 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
8243 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
8247 nla_for_each_nested(attr
, br_spec
, rem
) {
8251 if (nla_type(attr
) != IFLA_BRIDGE_MODE
)
8254 if (nla_len(attr
) < sizeof(mode
))
8257 mode
= nla_get_u16(attr
);
8258 status
= ixgbe_configure_bridge_mode(adapter
, mode
);
8268 static int ixgbe_ndo_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
8269 struct net_device
*dev
,
8270 u32 filter_mask
, int nlflags
)
8272 struct ixgbe_adapter
*adapter
= netdev_priv(dev
);
8274 if (!(adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
))
8277 return ndo_dflt_bridge_getlink(skb
, pid
, seq
, dev
,
8278 adapter
->bridge_mode
, 0, 0, nlflags
,
8282 static void *ixgbe_fwd_add(struct net_device
*pdev
, struct net_device
*vdev
)
8284 struct ixgbe_fwd_adapter
*fwd_adapter
= NULL
;
8285 struct ixgbe_adapter
*adapter
= netdev_priv(pdev
);
8286 int used_pools
= adapter
->num_vfs
+ adapter
->num_rx_pools
;
8290 /* Hardware has a limited number of available pools. Each VF, and the
8291 * PF require a pool. Check to ensure we don't attempt to use more
8292 * then the available number of pools.
8294 if (used_pools
>= IXGBE_MAX_VF_FUNCTIONS
)
8295 return ERR_PTR(-EINVAL
);
8298 if (vdev
->num_rx_queues
!= vdev
->num_tx_queues
) {
8299 netdev_info(pdev
, "%s: Only supports a single queue count for TX and RX\n",
8301 return ERR_PTR(-EINVAL
);
8304 /* Check for hardware restriction on number of rx/tx queues */
8305 if (vdev
->num_tx_queues
> IXGBE_MAX_L2A_QUEUES
||
8306 vdev
->num_tx_queues
== IXGBE_BAD_L2A_QUEUE
) {
8308 "%s: Supports RX/TX Queue counts 1,2, and 4\n",
8310 return ERR_PTR(-EINVAL
);
8313 if (((adapter
->flags
& IXGBE_FLAG_DCB_ENABLED
) &&
8314 adapter
->num_rx_pools
> IXGBE_MAX_DCBMACVLANS
- 1) ||
8315 (adapter
->num_rx_pools
> IXGBE_MAX_MACVLANS
))
8316 return ERR_PTR(-EBUSY
);
8318 fwd_adapter
= kzalloc(sizeof(*fwd_adapter
), GFP_KERNEL
);
8320 return ERR_PTR(-ENOMEM
);
8322 pool
= find_first_zero_bit(&adapter
->fwd_bitmask
, 32);
8323 adapter
->num_rx_pools
++;
8324 set_bit(pool
, &adapter
->fwd_bitmask
);
8325 limit
= find_last_bit(&adapter
->fwd_bitmask
, 32);
8327 /* Enable VMDq flag so device will be set in VM mode */
8328 adapter
->flags
|= IXGBE_FLAG_VMDQ_ENABLED
| IXGBE_FLAG_SRIOV_ENABLED
;
8329 adapter
->ring_feature
[RING_F_VMDQ
].limit
= limit
+ 1;
8330 adapter
->ring_feature
[RING_F_RSS
].limit
= vdev
->num_tx_queues
;
8332 /* Force reinit of ring allocation with VMDQ enabled */
8333 err
= ixgbe_setup_tc(pdev
, netdev_get_num_tc(pdev
));
8336 fwd_adapter
->pool
= pool
;
8337 fwd_adapter
->real_adapter
= adapter
;
8338 err
= ixgbe_fwd_ring_up(vdev
, fwd_adapter
);
8341 netif_tx_start_all_queues(vdev
);
8344 /* unwind counter and free adapter struct */
8346 "%s: dfwd hardware acceleration failed\n", vdev
->name
);
8347 clear_bit(pool
, &adapter
->fwd_bitmask
);
8348 adapter
->num_rx_pools
--;
8350 return ERR_PTR(err
);
8353 static void ixgbe_fwd_del(struct net_device
*pdev
, void *priv
)
8355 struct ixgbe_fwd_adapter
*fwd_adapter
= priv
;
8356 struct ixgbe_adapter
*adapter
= fwd_adapter
->real_adapter
;
8359 clear_bit(fwd_adapter
->pool
, &adapter
->fwd_bitmask
);
8360 adapter
->num_rx_pools
--;
8362 limit
= find_last_bit(&adapter
->fwd_bitmask
, 32);
8363 adapter
->ring_feature
[RING_F_VMDQ
].limit
= limit
+ 1;
8364 ixgbe_fwd_ring_down(fwd_adapter
->netdev
, fwd_adapter
);
8365 ixgbe_setup_tc(pdev
, netdev_get_num_tc(pdev
));
8366 netdev_dbg(pdev
, "pool %i:%i queues %i:%i VSI bitmask %lx\n",
8367 fwd_adapter
->pool
, adapter
->num_rx_pools
,
8368 fwd_adapter
->rx_base_queue
,
8369 fwd_adapter
->rx_base_queue
+ adapter
->num_rx_queues_per_pool
,
8370 adapter
->fwd_bitmask
);
8374 #define IXGBE_MAX_TUNNEL_HDR_LEN 80
8375 static netdev_features_t
8376 ixgbe_features_check(struct sk_buff
*skb
, struct net_device
*dev
,
8377 netdev_features_t features
)
8379 if (!skb
->encapsulation
)
8382 if (unlikely(skb_inner_mac_header(skb
) - skb_transport_header(skb
) >
8383 IXGBE_MAX_TUNNEL_HDR_LEN
))
8384 return features
& ~NETIF_F_ALL_CSUM
;
8389 static const struct net_device_ops ixgbe_netdev_ops
= {
8390 .ndo_open
= ixgbe_open
,
8391 .ndo_stop
= ixgbe_close
,
8392 .ndo_start_xmit
= ixgbe_xmit_frame
,
8393 .ndo_select_queue
= ixgbe_select_queue
,
8394 .ndo_set_rx_mode
= ixgbe_set_rx_mode
,
8395 .ndo_validate_addr
= eth_validate_addr
,
8396 .ndo_set_mac_address
= ixgbe_set_mac
,
8397 .ndo_change_mtu
= ixgbe_change_mtu
,
8398 .ndo_tx_timeout
= ixgbe_tx_timeout
,
8399 .ndo_vlan_rx_add_vid
= ixgbe_vlan_rx_add_vid
,
8400 .ndo_vlan_rx_kill_vid
= ixgbe_vlan_rx_kill_vid
,
8401 .ndo_do_ioctl
= ixgbe_ioctl
,
8402 .ndo_set_vf_mac
= ixgbe_ndo_set_vf_mac
,
8403 .ndo_set_vf_vlan
= ixgbe_ndo_set_vf_vlan
,
8404 .ndo_set_vf_rate
= ixgbe_ndo_set_vf_bw
,
8405 .ndo_set_vf_spoofchk
= ixgbe_ndo_set_vf_spoofchk
,
8406 .ndo_set_vf_rss_query_en
= ixgbe_ndo_set_vf_rss_query_en
,
8407 .ndo_set_vf_trust
= ixgbe_ndo_set_vf_trust
,
8408 .ndo_get_vf_config
= ixgbe_ndo_get_vf_config
,
8409 .ndo_get_stats64
= ixgbe_get_stats64
,
8410 #ifdef CONFIG_IXGBE_DCB
8411 .ndo_setup_tc
= ixgbe_setup_tc
,
8413 #ifdef CONFIG_NET_POLL_CONTROLLER
8414 .ndo_poll_controller
= ixgbe_netpoll
,
8416 #ifdef CONFIG_NET_RX_BUSY_POLL
8417 .ndo_busy_poll
= ixgbe_low_latency_recv
,
8420 .ndo_fcoe_ddp_setup
= ixgbe_fcoe_ddp_get
,
8421 .ndo_fcoe_ddp_target
= ixgbe_fcoe_ddp_target
,
8422 .ndo_fcoe_ddp_done
= ixgbe_fcoe_ddp_put
,
8423 .ndo_fcoe_enable
= ixgbe_fcoe_enable
,
8424 .ndo_fcoe_disable
= ixgbe_fcoe_disable
,
8425 .ndo_fcoe_get_wwn
= ixgbe_fcoe_get_wwn
,
8426 .ndo_fcoe_get_hbainfo
= ixgbe_fcoe_get_hbainfo
,
8427 #endif /* IXGBE_FCOE */
8428 .ndo_set_features
= ixgbe_set_features
,
8429 .ndo_fix_features
= ixgbe_fix_features
,
8430 .ndo_fdb_add
= ixgbe_ndo_fdb_add
,
8431 .ndo_bridge_setlink
= ixgbe_ndo_bridge_setlink
,
8432 .ndo_bridge_getlink
= ixgbe_ndo_bridge_getlink
,
8433 .ndo_dfwd_add_station
= ixgbe_fwd_add
,
8434 .ndo_dfwd_del_station
= ixgbe_fwd_del
,
8435 #ifdef CONFIG_IXGBE_VXLAN
8436 .ndo_add_vxlan_port
= ixgbe_add_vxlan_port
,
8437 .ndo_del_vxlan_port
= ixgbe_del_vxlan_port
,
8438 #endif /* CONFIG_IXGBE_VXLAN */
8439 .ndo_features_check
= ixgbe_features_check
,
8443 * ixgbe_enumerate_functions - Get the number of ports this device has
8444 * @adapter: adapter structure
8446 * This function enumerates the phsyical functions co-located on a single slot,
8447 * in order to determine how many ports a device has. This is most useful in
8448 * determining the required GT/s of PCIe bandwidth necessary for optimal
8451 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter
*adapter
)
8453 struct pci_dev
*entry
, *pdev
= adapter
->pdev
;
8456 /* Some cards can not use the generic count PCIe functions method,
8457 * because they are behind a parent switch, so we hardcode these with
8458 * the correct number of functions.
8460 if (ixgbe_pcie_from_parent(&adapter
->hw
))
8463 list_for_each_entry(entry
, &adapter
->pdev
->bus
->devices
, bus_list
) {
8464 /* don't count virtual functions */
8465 if (entry
->is_virtfn
)
8468 /* When the devices on the bus don't all match our device ID,
8469 * we can't reliably determine the correct number of
8470 * functions. This can occur if a function has been direct
8471 * attached to a virtual machine using VT-d, for example. In
8472 * this case, simply return -1 to indicate this.
8474 if ((entry
->vendor
!= pdev
->vendor
) ||
8475 (entry
->device
!= pdev
->device
))
8485 * ixgbe_wol_supported - Check whether device supports WoL
8486 * @hw: hw specific details
8487 * @device_id: the device ID
8488 * @subdev_id: the subsystem device ID
8490 * This function is used by probe and ethtool to determine
8491 * which devices have WoL support
8494 int ixgbe_wol_supported(struct ixgbe_adapter
*adapter
, u16 device_id
,
8497 struct ixgbe_hw
*hw
= &adapter
->hw
;
8498 u16 wol_cap
= adapter
->eeprom_cap
& IXGBE_DEVICE_CAPS_WOL_MASK
;
8499 int is_wol_supported
= 0;
8501 switch (device_id
) {
8502 case IXGBE_DEV_ID_82599_SFP
:
8503 /* Only these subdevices could supports WOL */
8504 switch (subdevice_id
) {
8505 case IXGBE_SUBDEV_ID_82599_SFP_WOL0
:
8506 case IXGBE_SUBDEV_ID_82599_560FLR
:
8507 /* only support first port */
8508 if (hw
->bus
.func
!= 0)
8510 case IXGBE_SUBDEV_ID_82599_SP_560FLR
:
8511 case IXGBE_SUBDEV_ID_82599_SFP
:
8512 case IXGBE_SUBDEV_ID_82599_RNDC
:
8513 case IXGBE_SUBDEV_ID_82599_ECNA_DP
:
8514 case IXGBE_SUBDEV_ID_82599_LOM_SFP
:
8515 is_wol_supported
= 1;
8519 case IXGBE_DEV_ID_82599EN_SFP
:
8520 /* Only this subdevice supports WOL */
8521 switch (subdevice_id
) {
8522 case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1
:
8523 is_wol_supported
= 1;
8527 case IXGBE_DEV_ID_82599_COMBO_BACKPLANE
:
8528 /* All except this subdevice support WOL */
8529 if (subdevice_id
!= IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ
)
8530 is_wol_supported
= 1;
8532 case IXGBE_DEV_ID_82599_KX4
:
8533 is_wol_supported
= 1;
8535 case IXGBE_DEV_ID_X540T
:
8536 case IXGBE_DEV_ID_X540T1
:
8537 case IXGBE_DEV_ID_X550T
:
8538 case IXGBE_DEV_ID_X550EM_X_KX4
:
8539 case IXGBE_DEV_ID_X550EM_X_KR
:
8540 case IXGBE_DEV_ID_X550EM_X_10G_T
:
8541 /* check eeprom to see if enabled wol */
8542 if ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0_1
) ||
8543 ((wol_cap
== IXGBE_DEVICE_CAPS_WOL_PORT0
) &&
8544 (hw
->bus
.func
== 0))) {
8545 is_wol_supported
= 1;
8550 return is_wol_supported
;
8554 * ixgbe_get_platform_mac_addr - Look up MAC address in Open Firmware / IDPROM
8555 * @adapter: Pointer to adapter struct
8557 static void ixgbe_get_platform_mac_addr(struct ixgbe_adapter
*adapter
)
8560 struct device_node
*dp
= pci_device_to_OF_node(adapter
->pdev
);
8561 struct ixgbe_hw
*hw
= &adapter
->hw
;
8562 const unsigned char *addr
;
8564 addr
= of_get_mac_address(dp
);
8566 ether_addr_copy(hw
->mac
.perm_addr
, addr
);
8569 #endif /* CONFIG_OF */
8572 ether_addr_copy(hw
->mac
.perm_addr
, idprom
->id_ethaddr
);
8573 #endif /* CONFIG_SPARC */
8577 * ixgbe_probe - Device Initialization Routine
8578 * @pdev: PCI device information struct
8579 * @ent: entry in ixgbe_pci_tbl
8581 * Returns 0 on success, negative on failure
8583 * ixgbe_probe initializes an adapter identified by a pci_dev structure.
8584 * The OS initialization, configuring of the adapter private structure,
8585 * and a hardware reset occur.
8587 static int ixgbe_probe(struct pci_dev
*pdev
, const struct pci_device_id
*ent
)
8589 struct net_device
*netdev
;
8590 struct ixgbe_adapter
*adapter
= NULL
;
8591 struct ixgbe_hw
*hw
;
8592 const struct ixgbe_info
*ii
= ixgbe_info_tbl
[ent
->driver_data
];
8593 int i
, err
, pci_using_dac
, expected_gts
;
8594 unsigned int indices
= MAX_TX_QUEUES
;
8595 u8 part_str
[IXGBE_PBANUM_LENGTH
];
8596 bool disable_dev
= false;
8602 /* Catch broken hardware that put the wrong VF device ID in
8603 * the PCIe SR-IOV capability.
8605 if (pdev
->is_virtfn
) {
8606 WARN(1, KERN_ERR
"%s (%hx:%hx) should not be a VF!\n",
8607 pci_name(pdev
), pdev
->vendor
, pdev
->device
);
8611 err
= pci_enable_device_mem(pdev
);
8615 if (!dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(64))) {
8618 err
= dma_set_mask_and_coherent(&pdev
->dev
, DMA_BIT_MASK(32));
8621 "No usable DMA configuration, aborting\n");
8627 err
= pci_request_selected_regions(pdev
, pci_select_bars(pdev
,
8628 IORESOURCE_MEM
), ixgbe_driver_name
);
8631 "pci_request_selected_regions failed 0x%x\n", err
);
8635 pci_enable_pcie_error_reporting(pdev
);
8637 pci_set_master(pdev
);
8638 pci_save_state(pdev
);
8640 if (ii
->mac
== ixgbe_mac_82598EB
) {
8641 #ifdef CONFIG_IXGBE_DCB
8642 /* 8 TC w/ 4 queues per TC */
8643 indices
= 4 * MAX_TRAFFIC_CLASS
;
8645 indices
= IXGBE_MAX_RSS_INDICES
;
8649 netdev
= alloc_etherdev_mq(sizeof(struct ixgbe_adapter
), indices
);
8652 goto err_alloc_etherdev
;
8655 SET_NETDEV_DEV(netdev
, &pdev
->dev
);
8657 adapter
= netdev_priv(netdev
);
8659 adapter
->netdev
= netdev
;
8660 adapter
->pdev
= pdev
;
8663 adapter
->msg_enable
= netif_msg_init(debug
, DEFAULT_MSG_ENABLE
);
8665 hw
->hw_addr
= ioremap(pci_resource_start(pdev
, 0),
8666 pci_resource_len(pdev
, 0));
8667 adapter
->io_addr
= hw
->hw_addr
;
8673 netdev
->netdev_ops
= &ixgbe_netdev_ops
;
8674 ixgbe_set_ethtool_ops(netdev
);
8675 netdev
->watchdog_timeo
= 5 * HZ
;
8676 strlcpy(netdev
->name
, pci_name(pdev
), sizeof(netdev
->name
));
8679 memcpy(&hw
->mac
.ops
, ii
->mac_ops
, sizeof(hw
->mac
.ops
));
8680 hw
->mac
.type
= ii
->mac
;
8681 hw
->mvals
= ii
->mvals
;
8684 memcpy(&hw
->eeprom
.ops
, ii
->eeprom_ops
, sizeof(hw
->eeprom
.ops
));
8685 eec
= IXGBE_READ_REG(hw
, IXGBE_EEC(hw
));
8686 if (ixgbe_removed(hw
->hw_addr
)) {
8690 /* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
8691 if (!(eec
& (1 << 8)))
8692 hw
->eeprom
.ops
.read
= &ixgbe_read_eeprom_bit_bang_generic
;
8695 memcpy(&hw
->phy
.ops
, ii
->phy_ops
, sizeof(hw
->phy
.ops
));
8696 hw
->phy
.sfp_type
= ixgbe_sfp_type_unknown
;
8697 /* ixgbe_identify_phy_generic will set prtad and mmds properly */
8698 hw
->phy
.mdio
.prtad
= MDIO_PRTAD_NONE
;
8699 hw
->phy
.mdio
.mmds
= 0;
8700 hw
->phy
.mdio
.mode_support
= MDIO_SUPPORTS_C45
| MDIO_EMULATE_C22
;
8701 hw
->phy
.mdio
.dev
= netdev
;
8702 hw
->phy
.mdio
.mdio_read
= ixgbe_mdio_read
;
8703 hw
->phy
.mdio
.mdio_write
= ixgbe_mdio_write
;
8705 ii
->get_invariants(hw
);
8707 /* setup the private structure */
8708 err
= ixgbe_sw_init(adapter
);
8712 /* Make it possible the adapter to be woken up via WOL */
8713 switch (adapter
->hw
.mac
.type
) {
8714 case ixgbe_mac_82599EB
:
8715 case ixgbe_mac_X540
:
8716 case ixgbe_mac_X550
:
8717 case ixgbe_mac_X550EM_x
:
8718 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
8725 * If there is a fan on this device and it has failed log the
8728 if (adapter
->flags
& IXGBE_FLAG_FAN_FAIL_CAPABLE
) {
8729 u32 esdp
= IXGBE_READ_REG(hw
, IXGBE_ESDP
);
8730 if (esdp
& IXGBE_ESDP_SDP1
)
8731 e_crit(probe
, "Fan has stopped, replace the adapter\n");
8734 if (allow_unsupported_sfp
)
8735 hw
->allow_unsupported_sfp
= allow_unsupported_sfp
;
8737 /* reset_hw fills in the perm_addr as well */
8738 hw
->phy
.reset_if_overtemp
= true;
8739 err
= hw
->mac
.ops
.reset_hw(hw
);
8740 hw
->phy
.reset_if_overtemp
= false;
8741 if (err
== IXGBE_ERR_SFP_NOT_PRESENT
) {
8743 } else if (err
== IXGBE_ERR_SFP_NOT_SUPPORTED
) {
8744 e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
8745 e_dev_err("Reload the driver after installing a supported module.\n");
8748 e_dev_err("HW Init failed: %d\n", err
);
8752 #ifdef CONFIG_PCI_IOV
8753 /* SR-IOV not supported on the 82598 */
8754 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
)
8757 ixgbe_init_mbx_params_pf(hw
);
8758 memcpy(&hw
->mbx
.ops
, ii
->mbx_ops
, sizeof(hw
->mbx
.ops
));
8759 pci_sriov_set_totalvfs(pdev
, IXGBE_MAX_VFS_DRV_LIMIT
);
8760 ixgbe_enable_sriov(adapter
);
8764 netdev
->features
= NETIF_F_SG
|
8767 NETIF_F_HW_VLAN_CTAG_TX
|
8768 NETIF_F_HW_VLAN_CTAG_RX
|
8774 netdev
->hw_features
= netdev
->features
| NETIF_F_HW_L2FW_DOFFLOAD
;
8776 switch (adapter
->hw
.mac
.type
) {
8777 case ixgbe_mac_82599EB
:
8778 case ixgbe_mac_X540
:
8779 case ixgbe_mac_X550
:
8780 case ixgbe_mac_X550EM_x
:
8781 netdev
->features
|= NETIF_F_SCTP_CSUM
;
8782 netdev
->hw_features
|= NETIF_F_SCTP_CSUM
|
8789 netdev
->hw_features
|= NETIF_F_RXALL
;
8790 netdev
->features
|= NETIF_F_HW_VLAN_CTAG_FILTER
;
8792 netdev
->vlan_features
|= NETIF_F_TSO
;
8793 netdev
->vlan_features
|= NETIF_F_TSO6
;
8794 netdev
->vlan_features
|= NETIF_F_IP_CSUM
;
8795 netdev
->vlan_features
|= NETIF_F_IPV6_CSUM
;
8796 netdev
->vlan_features
|= NETIF_F_SG
;
8798 netdev
->hw_enc_features
|= NETIF_F_SG
| NETIF_F_IP_CSUM
|
8801 netdev
->priv_flags
|= IFF_UNICAST_FLT
;
8802 netdev
->priv_flags
|= IFF_SUPP_NOFCS
;
8804 #ifdef CONFIG_IXGBE_VXLAN
8805 switch (adapter
->hw
.mac
.type
) {
8806 case ixgbe_mac_X550
:
8807 case ixgbe_mac_X550EM_x
:
8808 netdev
->hw_enc_features
|= NETIF_F_RXCSUM
|
8815 #endif /* CONFIG_IXGBE_VXLAN */
8817 #ifdef CONFIG_IXGBE_DCB
8818 netdev
->dcbnl_ops
= &dcbnl_ops
;
8822 if (adapter
->flags
& IXGBE_FLAG_FCOE_CAPABLE
) {
8823 unsigned int fcoe_l
;
8825 if (hw
->mac
.ops
.get_device_caps
) {
8826 hw
->mac
.ops
.get_device_caps(hw
, &device_caps
);
8827 if (device_caps
& IXGBE_DEVICE_CAPS_FCOE_OFFLOADS
)
8828 adapter
->flags
&= ~IXGBE_FLAG_FCOE_CAPABLE
;
8832 fcoe_l
= min_t(int, IXGBE_FCRETA_SIZE
, num_online_cpus());
8833 adapter
->ring_feature
[RING_F_FCOE
].limit
= fcoe_l
;
8835 netdev
->features
|= NETIF_F_FSO
|
8838 netdev
->vlan_features
|= NETIF_F_FSO
|
8842 #endif /* IXGBE_FCOE */
8843 if (pci_using_dac
) {
8844 netdev
->features
|= NETIF_F_HIGHDMA
;
8845 netdev
->vlan_features
|= NETIF_F_HIGHDMA
;
8848 if (adapter
->flags2
& IXGBE_FLAG2_RSC_CAPABLE
)
8849 netdev
->hw_features
|= NETIF_F_LRO
;
8850 if (adapter
->flags2
& IXGBE_FLAG2_RSC_ENABLED
)
8851 netdev
->features
|= NETIF_F_LRO
;
8853 /* make sure the EEPROM is good */
8854 if (hw
->eeprom
.ops
.validate_checksum(hw
, NULL
) < 0) {
8855 e_dev_err("The EEPROM Checksum Is Not Valid\n");
8860 ixgbe_get_platform_mac_addr(adapter
);
8862 memcpy(netdev
->dev_addr
, hw
->mac
.perm_addr
, netdev
->addr_len
);
8864 if (!is_valid_ether_addr(netdev
->dev_addr
)) {
8865 e_dev_err("invalid MAC address\n");
8870 ixgbe_mac_set_default_filter(adapter
, hw
->mac
.perm_addr
);
8872 setup_timer(&adapter
->service_timer
, &ixgbe_service_timer
,
8873 (unsigned long) adapter
);
8875 if (ixgbe_removed(hw
->hw_addr
)) {
8879 INIT_WORK(&adapter
->service_task
, ixgbe_service_task
);
8880 set_bit(__IXGBE_SERVICE_INITED
, &adapter
->state
);
8881 clear_bit(__IXGBE_SERVICE_SCHED
, &adapter
->state
);
8883 err
= ixgbe_init_interrupt_scheme(adapter
);
8887 /* WOL not supported for all devices */
8889 hw
->eeprom
.ops
.read(hw
, 0x2c, &adapter
->eeprom_cap
);
8890 hw
->wol_enabled
= ixgbe_wol_supported(adapter
, pdev
->device
,
8891 pdev
->subsystem_device
);
8892 if (hw
->wol_enabled
)
8893 adapter
->wol
= IXGBE_WUFC_MAG
;
8895 device_set_wakeup_enable(&adapter
->pdev
->dev
, adapter
->wol
);
8897 /* save off EEPROM version number */
8898 hw
->eeprom
.ops
.read(hw
, 0x2e, &adapter
->eeprom_verh
);
8899 hw
->eeprom
.ops
.read(hw
, 0x2d, &adapter
->eeprom_verl
);
8901 /* pick up the PCI bus settings for reporting later */
8902 if (ixgbe_pcie_from_parent(hw
))
8903 ixgbe_get_parent_bus_info(adapter
);
8905 hw
->mac
.ops
.get_bus_info(hw
);
8907 /* calculate the expected PCIe bandwidth required for optimal
8908 * performance. Note that some older parts will never have enough
8909 * bandwidth due to being older generation PCIe parts. We clamp these
8910 * parts to ensure no warning is displayed if it can't be fixed.
8912 switch (hw
->mac
.type
) {
8913 case ixgbe_mac_82598EB
:
8914 expected_gts
= min(ixgbe_enumerate_functions(adapter
) * 10, 16);
8917 expected_gts
= ixgbe_enumerate_functions(adapter
) * 10;
8921 /* don't check link if we failed to enumerate functions */
8922 if (expected_gts
> 0)
8923 ixgbe_check_minimum_link(adapter
, expected_gts
);
8925 err
= ixgbe_read_pba_string_generic(hw
, part_str
, sizeof(part_str
));
8927 strlcpy(part_str
, "Unknown", sizeof(part_str
));
8928 if (ixgbe_is_sfp(hw
) && hw
->phy
.sfp_type
!= ixgbe_sfp_type_not_present
)
8929 e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
8930 hw
->mac
.type
, hw
->phy
.type
, hw
->phy
.sfp_type
,
8933 e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
8934 hw
->mac
.type
, hw
->phy
.type
, part_str
);
8936 e_dev_info("%pM\n", netdev
->dev_addr
);
8938 /* reset the hardware with the new settings */
8939 err
= hw
->mac
.ops
.start_hw(hw
);
8940 if (err
== IXGBE_ERR_EEPROM_VERSION
) {
8941 /* We are running on a pre-production device, log a warning */
8942 e_dev_warn("This device is a pre-production adapter/LOM. "
8943 "Please be aware there may be issues associated "
8944 "with your hardware. If you are experiencing "
8945 "problems please contact your Intel or hardware "
8946 "representative who provided you with this "
8949 strcpy(netdev
->name
, "eth%d");
8950 err
= register_netdev(netdev
);
8954 pci_set_drvdata(pdev
, adapter
);
8956 /* power down the optics for 82599 SFP+ fiber */
8957 if (hw
->mac
.ops
.disable_tx_laser
)
8958 hw
->mac
.ops
.disable_tx_laser(hw
);
8960 /* carrier off reporting is important to ethtool even BEFORE open */
8961 netif_carrier_off(netdev
);
8963 #ifdef CONFIG_IXGBE_DCA
8964 if (dca_add_requester(&pdev
->dev
) == 0) {
8965 adapter
->flags
|= IXGBE_FLAG_DCA_ENABLED
;
8966 ixgbe_setup_dca(adapter
);
8969 if (adapter
->flags
& IXGBE_FLAG_SRIOV_ENABLED
) {
8970 e_info(probe
, "IOV is enabled with %d VFs\n", adapter
->num_vfs
);
8971 for (i
= 0; i
< adapter
->num_vfs
; i
++)
8972 ixgbe_vf_configuration(pdev
, (i
| 0x10000000));
8975 /* firmware requires driver version to be 0xFFFFFFFF
8976 * since os does not support feature
8978 if (hw
->mac
.ops
.set_fw_drv_ver
)
8979 hw
->mac
.ops
.set_fw_drv_ver(hw
, 0xFF, 0xFF, 0xFF,
8982 /* add san mac addr to netdev */
8983 ixgbe_add_sanmac_netdev(netdev
);
8985 e_dev_info("%s\n", ixgbe_default_device_descr
);
8987 #ifdef CONFIG_IXGBE_HWMON
8988 if (ixgbe_sysfs_init(adapter
))
8989 e_err(probe
, "failed to allocate sysfs resources\n");
8990 #endif /* CONFIG_IXGBE_HWMON */
8992 ixgbe_dbg_adapter_init(adapter
);
8994 /* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
8995 if (ixgbe_mng_enabled(hw
) && ixgbe_is_sfp(hw
) && hw
->mac
.ops
.setup_link
)
8996 hw
->mac
.ops
.setup_link(hw
,
8997 IXGBE_LINK_SPEED_10GB_FULL
| IXGBE_LINK_SPEED_1GB_FULL
,
9003 ixgbe_release_hw_control(adapter
);
9004 ixgbe_clear_interrupt_scheme(adapter
);
9006 ixgbe_disable_sriov(adapter
);
9007 adapter
->flags2
&= ~IXGBE_FLAG2_SEARCH_FOR_SFP
;
9008 iounmap(adapter
->io_addr
);
9009 kfree(adapter
->mac_table
);
9011 disable_dev
= !test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
);
9012 free_netdev(netdev
);
9014 pci_release_selected_regions(pdev
,
9015 pci_select_bars(pdev
, IORESOURCE_MEM
));
9018 if (!adapter
|| disable_dev
)
9019 pci_disable_device(pdev
);
9024 * ixgbe_remove - Device Removal Routine
9025 * @pdev: PCI device information struct
9027 * ixgbe_remove is called by the PCI subsystem to alert the driver
9028 * that it should release a PCI device. The could be caused by a
9029 * Hot-Plug event, or because the driver is going to be removed from
9032 static void ixgbe_remove(struct pci_dev
*pdev
)
9034 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9035 struct net_device
*netdev
;
9038 /* if !adapter then we already cleaned up in probe */
9042 netdev
= adapter
->netdev
;
9043 ixgbe_dbg_adapter_exit(adapter
);
9045 set_bit(__IXGBE_REMOVING
, &adapter
->state
);
9046 cancel_work_sync(&adapter
->service_task
);
9049 #ifdef CONFIG_IXGBE_DCA
9050 if (adapter
->flags
& IXGBE_FLAG_DCA_ENABLED
) {
9051 adapter
->flags
&= ~IXGBE_FLAG_DCA_ENABLED
;
9052 dca_remove_requester(&pdev
->dev
);
9053 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_DCA_CTRL
,
9054 IXGBE_DCA_CTRL_DCA_DISABLE
);
9058 #ifdef CONFIG_IXGBE_HWMON
9059 ixgbe_sysfs_exit(adapter
);
9060 #endif /* CONFIG_IXGBE_HWMON */
9062 /* remove the added san mac */
9063 ixgbe_del_sanmac_netdev(netdev
);
9065 #ifdef CONFIG_PCI_IOV
9066 ixgbe_disable_sriov(adapter
);
9068 if (netdev
->reg_state
== NETREG_REGISTERED
)
9069 unregister_netdev(netdev
);
9071 ixgbe_clear_interrupt_scheme(adapter
);
9073 ixgbe_release_hw_control(adapter
);
9076 kfree(adapter
->ixgbe_ieee_pfc
);
9077 kfree(adapter
->ixgbe_ieee_ets
);
9080 iounmap(adapter
->io_addr
);
9081 pci_release_selected_regions(pdev
, pci_select_bars(pdev
,
9084 e_dev_info("complete\n");
9086 kfree(adapter
->mac_table
);
9087 disable_dev
= !test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
);
9088 free_netdev(netdev
);
9090 pci_disable_pcie_error_reporting(pdev
);
9093 pci_disable_device(pdev
);
9097 * ixgbe_io_error_detected - called when PCI error is detected
9098 * @pdev: Pointer to PCI device
9099 * @state: The current pci connection state
9101 * This function is called after a PCI bus error affecting
9102 * this device has been detected.
9104 static pci_ers_result_t
ixgbe_io_error_detected(struct pci_dev
*pdev
,
9105 pci_channel_state_t state
)
9107 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9108 struct net_device
*netdev
= adapter
->netdev
;
9110 #ifdef CONFIG_PCI_IOV
9111 struct ixgbe_hw
*hw
= &adapter
->hw
;
9112 struct pci_dev
*bdev
, *vfdev
;
9113 u32 dw0
, dw1
, dw2
, dw3
;
9115 u16 req_id
, pf_func
;
9117 if (adapter
->hw
.mac
.type
== ixgbe_mac_82598EB
||
9118 adapter
->num_vfs
== 0)
9119 goto skip_bad_vf_detection
;
9121 bdev
= pdev
->bus
->self
;
9122 while (bdev
&& (pci_pcie_type(bdev
) != PCI_EXP_TYPE_ROOT_PORT
))
9123 bdev
= bdev
->bus
->self
;
9126 goto skip_bad_vf_detection
;
9128 pos
= pci_find_ext_capability(bdev
, PCI_EXT_CAP_ID_ERR
);
9130 goto skip_bad_vf_detection
;
9132 dw0
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
);
9133 dw1
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
+ 4);
9134 dw2
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
+ 8);
9135 dw3
= ixgbe_read_pci_cfg_dword(hw
, pos
+ PCI_ERR_HEADER_LOG
+ 12);
9136 if (ixgbe_removed(hw
->hw_addr
))
9137 goto skip_bad_vf_detection
;
9140 /* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
9141 if (!(req_id
& 0x0080))
9142 goto skip_bad_vf_detection
;
9144 pf_func
= req_id
& 0x01;
9145 if ((pf_func
& 1) == (pdev
->devfn
& 1)) {
9146 unsigned int device_id
;
9148 vf
= (req_id
& 0x7F) >> 1;
9149 e_dev_err("VF %d has caused a PCIe error\n", vf
);
9150 e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
9151 "%8.8x\tdw3: %8.8x\n",
9152 dw0
, dw1
, dw2
, dw3
);
9153 switch (adapter
->hw
.mac
.type
) {
9154 case ixgbe_mac_82599EB
:
9155 device_id
= IXGBE_82599_VF_DEVICE_ID
;
9157 case ixgbe_mac_X540
:
9158 device_id
= IXGBE_X540_VF_DEVICE_ID
;
9160 case ixgbe_mac_X550
:
9161 device_id
= IXGBE_DEV_ID_X550_VF
;
9163 case ixgbe_mac_X550EM_x
:
9164 device_id
= IXGBE_DEV_ID_X550EM_X_VF
;
9171 /* Find the pci device of the offending VF */
9172 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
, device_id
, NULL
);
9174 if (vfdev
->devfn
== (req_id
& 0xFF))
9176 vfdev
= pci_get_device(PCI_VENDOR_ID_INTEL
,
9180 * There's a slim chance the VF could have been hot plugged,
9181 * so if it is no longer present we don't need to issue the
9182 * VFLR. Just clean up the AER in that case.
9185 ixgbe_issue_vf_flr(adapter
, vfdev
);
9186 /* Free device reference count */
9190 pci_cleanup_aer_uncorrect_error_status(pdev
);
9194 * Even though the error may have occurred on the other port
9195 * we still need to increment the vf error reference count for
9196 * both ports because the I/O resume function will be called
9199 adapter
->vferr_refcount
++;
9201 return PCI_ERS_RESULT_RECOVERED
;
9203 skip_bad_vf_detection
:
9204 #endif /* CONFIG_PCI_IOV */
9205 if (!test_bit(__IXGBE_SERVICE_INITED
, &adapter
->state
))
9206 return PCI_ERS_RESULT_DISCONNECT
;
9209 netif_device_detach(netdev
);
9211 if (state
== pci_channel_io_perm_failure
) {
9213 return PCI_ERS_RESULT_DISCONNECT
;
9216 if (netif_running(netdev
))
9217 ixgbe_down(adapter
);
9219 if (!test_and_set_bit(__IXGBE_DISABLED
, &adapter
->state
))
9220 pci_disable_device(pdev
);
9223 /* Request a slot reset. */
9224 return PCI_ERS_RESULT_NEED_RESET
;
9228 * ixgbe_io_slot_reset - called after the pci bus has been reset.
9229 * @pdev: Pointer to PCI device
9231 * Restart the card from scratch, as if from a cold-boot.
9233 static pci_ers_result_t
ixgbe_io_slot_reset(struct pci_dev
*pdev
)
9235 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9236 pci_ers_result_t result
;
9239 if (pci_enable_device_mem(pdev
)) {
9240 e_err(probe
, "Cannot re-enable PCI device after reset.\n");
9241 result
= PCI_ERS_RESULT_DISCONNECT
;
9243 smp_mb__before_atomic();
9244 clear_bit(__IXGBE_DISABLED
, &adapter
->state
);
9245 adapter
->hw
.hw_addr
= adapter
->io_addr
;
9246 pci_set_master(pdev
);
9247 pci_restore_state(pdev
);
9248 pci_save_state(pdev
);
9250 pci_wake_from_d3(pdev
, false);
9252 ixgbe_reset(adapter
);
9253 IXGBE_WRITE_REG(&adapter
->hw
, IXGBE_WUS
, ~0);
9254 result
= PCI_ERS_RESULT_RECOVERED
;
9257 err
= pci_cleanup_aer_uncorrect_error_status(pdev
);
9259 e_dev_err("pci_cleanup_aer_uncorrect_error_status "
9260 "failed 0x%0x\n", err
);
9261 /* non-fatal, continue */
9268 * ixgbe_io_resume - called when traffic can start flowing again.
9269 * @pdev: Pointer to PCI device
9271 * This callback is called when the error recovery driver tells us that
9272 * its OK to resume normal operation.
9274 static void ixgbe_io_resume(struct pci_dev
*pdev
)
9276 struct ixgbe_adapter
*adapter
= pci_get_drvdata(pdev
);
9277 struct net_device
*netdev
= adapter
->netdev
;
9279 #ifdef CONFIG_PCI_IOV
9280 if (adapter
->vferr_refcount
) {
9281 e_info(drv
, "Resuming after VF err\n");
9282 adapter
->vferr_refcount
--;
9287 if (netif_running(netdev
))
9290 netif_device_attach(netdev
);
9293 static const struct pci_error_handlers ixgbe_err_handler
= {
9294 .error_detected
= ixgbe_io_error_detected
,
9295 .slot_reset
= ixgbe_io_slot_reset
,
9296 .resume
= ixgbe_io_resume
,
9299 static struct pci_driver ixgbe_driver
= {
9300 .name
= ixgbe_driver_name
,
9301 .id_table
= ixgbe_pci_tbl
,
9302 .probe
= ixgbe_probe
,
9303 .remove
= ixgbe_remove
,
9305 .suspend
= ixgbe_suspend
,
9306 .resume
= ixgbe_resume
,
9308 .shutdown
= ixgbe_shutdown
,
9309 .sriov_configure
= ixgbe_pci_sriov_configure
,
9310 .err_handler
= &ixgbe_err_handler
9314 * ixgbe_init_module - Driver Registration Routine
9316 * ixgbe_init_module is the first routine called when the driver is
9317 * loaded. All it does is register with the PCI subsystem.
9319 static int __init
ixgbe_init_module(void)
9322 pr_info("%s - version %s\n", ixgbe_driver_string
, ixgbe_driver_version
);
9323 pr_info("%s\n", ixgbe_copyright
);
9327 ret
= pci_register_driver(&ixgbe_driver
);
9333 #ifdef CONFIG_IXGBE_DCA
9334 dca_register_notify(&dca_notifier
);
9340 module_init(ixgbe_init_module
);
9343 * ixgbe_exit_module - Driver Exit Cleanup Routine
9345 * ixgbe_exit_module is called just before the driver is removed
9348 static void __exit
ixgbe_exit_module(void)
9350 #ifdef CONFIG_IXGBE_DCA
9351 dca_unregister_notify(&dca_notifier
);
9353 pci_unregister_driver(&ixgbe_driver
);
9358 #ifdef CONFIG_IXGBE_DCA
9359 static int ixgbe_notify_dca(struct notifier_block
*nb
, unsigned long event
,
9364 ret_val
= driver_for_each_device(&ixgbe_driver
.driver
, NULL
, &event
,
9365 __ixgbe_notify_dca
);
9367 return ret_val
? NOTIFY_BAD
: NOTIFY_DONE
;
9370 #endif /* CONFIG_IXGBE_DCA */
9372 module_exit(ixgbe_exit_module
);