[deliverable/linux.git] / drivers / edac / e7xxx_edac.c

/*
 * Intel e7xxx Memory Controller kernel module
 * (C) 2003 Linux Networx (http://lnxi.com)
 * This file may be distributed under the terms of the
 * GNU General Public License.
 *
 * See "enum e7xxx_chips" below for supported chipsets
 *
 * Written by Thayne Harbaugh
 * Based on work by Dan Hollis <goemon at anime dot net> and others.
 *	http://www.anime.net/~goemon/linux-ecc/
 *
 * Contributors:
 *	Eric Biederman (Linux Networx)
 *	Tom Zimmerman (Linux Networx)
 *	Jim Garlick (Lawrence Livermore National Labs)
 *	Dave Peterson (Lawrence Livermore National Labs)
 *	That One Guy (Some other place)
 *	Wang Zhenyu (intel.com)
 *
 * $Id: edac_e7xxx.c,v 1.5.2.9 2005/10/05 00:43:44 dsp_llnl Exp $
 *
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/pci.h>
#include <linux/pci_ids.h>
#include <linux/slab.h>
#include <linux/edac.h>
#include "edac_core.h"

#define	E7XXX_REVISION " Ver: 2.0.2 " __DATE__
#define	EDAC_MOD_STR	"e7xxx_edac"

#define e7xxx_printk(level, fmt, arg...) \
	edac_printk(level, "e7xxx", fmt, ##arg)

#define e7xxx_mc_printk(mci, level, fmt, arg...) \
	edac_mc_chipset_printk(mci, level, "e7xxx", fmt, ##arg)

#ifndef PCI_DEVICE_ID_INTEL_7205_0
#define PCI_DEVICE_ID_INTEL_7205_0	0x255d
#endif				/* PCI_DEVICE_ID_INTEL_7205_0 */

#ifndef PCI_DEVICE_ID_INTEL_7205_1_ERR
#define PCI_DEVICE_ID_INTEL_7205_1_ERR	0x2551
#endif				/* PCI_DEVICE_ID_INTEL_7205_1_ERR */

#ifndef PCI_DEVICE_ID_INTEL_7500_0
#define PCI_DEVICE_ID_INTEL_7500_0	0x2540
#endif				/* PCI_DEVICE_ID_INTEL_7500_0 */

#ifndef PCI_DEVICE_ID_INTEL_7500_1_ERR
#define PCI_DEVICE_ID_INTEL_7500_1_ERR	0x2541
#endif				/* PCI_DEVICE_ID_INTEL_7500_1_ERR */

#ifndef PCI_DEVICE_ID_INTEL_7501_0
#define PCI_DEVICE_ID_INTEL_7501_0	0x254c
#endif				/* PCI_DEVICE_ID_INTEL_7501_0 */

#ifndef PCI_DEVICE_ID_INTEL_7501_1_ERR
#define PCI_DEVICE_ID_INTEL_7501_1_ERR	0x2541
#endif				/* PCI_DEVICE_ID_INTEL_7501_1_ERR */

#ifndef PCI_DEVICE_ID_INTEL_7505_0
#define PCI_DEVICE_ID_INTEL_7505_0	0x2550
#endif				/* PCI_DEVICE_ID_INTEL_7505_0 */

#ifndef PCI_DEVICE_ID_INTEL_7505_1_ERR
#define PCI_DEVICE_ID_INTEL_7505_1_ERR	0x2551
#endif				/* PCI_DEVICE_ID_INTEL_7505_1_ERR */

#define E7XXX_NR_CSROWS		8	/* number of csrows */
#define E7XXX_NR_DIMMS		8	/* FIXME - is this correct? */

/* E7XXX register addresses - device 0 function 0 */
#define E7XXX_DRB		0x60	/* DRAM row boundary register (8b) */
#define E7XXX_DRA		0x70	/* DRAM row attribute register (8b) */
					/*
					 * 31   Device width row 7 0=x8 1=x4
					 * 27   Device width row 6
					 * 23   Device width row 5
					 * 19   Device width row 4
					 * 15   Device width row 3
					 * 11   Device width row 2
					 *  7   Device width row 1
					 *  3   Device width row 0
					 */
#define E7XXX_DRC		0x7C	/* DRAM controller mode reg (32b) */
					/*
					 * 22    Number channels 0=1,1=2
					 * 19:18 DRB Granularity 32/64MB
					 */
#define E7XXX_TOLM		0xC4	/* DRAM top of low memory reg (16b) */
#define E7XXX_REMAPBASE		0xC6	/* DRAM remap base address reg (16b) */
#define E7XXX_REMAPLIMIT	0xC8	/* DRAM remap limit address reg (16b) */

/* E7XXX register addresses - device 0 function 1 */
#define E7XXX_DRAM_FERR		0x80	/* DRAM first error register (8b) */
#define E7XXX_DRAM_NERR		0x82	/* DRAM next error register (8b) */
#define E7XXX_DRAM_CELOG_ADD	0xA0	/* DRAM first correctable memory */
					/*     error address register (32b) */
					/*
					 * 31:28 Reserved
					 * 27:6  CE address (4k block 33:12)
					 *  5:0  Reserved
					 */
#define E7XXX_DRAM_UELOG_ADD	0xB0	/* DRAM first uncorrectable memory */
					/*     error address register (32b) */
					/*
					 * 31:28 Reserved
					 * 27:6  CE address (4k block 33:12)
					 *  5:0  Reserved
					 */
#define E7XXX_DRAM_CELOG_SYNDROME 0xD0	/* DRAM first correctable memory */
					/*     error syndrome register (16b) */

enum e7xxx_chips {
	E7500 = 0,
	E7501,
	E7505,
	E7205,
};

struct e7xxx_pvt {
	struct pci_dev *bridge_ck;
	u32 tolm;
	u32 remapbase;
	u32 remaplimit;
	const struct e7xxx_dev_info *dev_info;
};

struct e7xxx_dev_info {
	u16 err_dev;
	const char *ctl_name;
};

struct e7xxx_error_info {
	u8 dram_ferr;
	u8 dram_nerr;
	u32 dram_celog_add;
	u16 dram_celog_syndrome;
	u32 dram_uelog_add;
};

static const struct e7xxx_dev_info e7xxx_devs[] = {
	[E7500] = {
		.err_dev = PCI_DEVICE_ID_INTEL_7500_1_ERR,
		.ctl_name = "E7500"
	},
	[E7501] = {
		.err_dev = PCI_DEVICE_ID_INTEL_7501_1_ERR,
		.ctl_name = "E7501"
	},
	[E7505] = {
		.err_dev = PCI_DEVICE_ID_INTEL_7505_1_ERR,
		.ctl_name = "E7505"
	},
	[E7205] = {
		.err_dev = PCI_DEVICE_ID_INTEL_7205_1_ERR,
		.ctl_name = "E7205"
	},
};

/* FIXME - is this valid for both SECDED and S4ECD4ED? */
static inline int e7xxx_find_channel(u16 syndrome)
{
	debugf3("%s()\n", __func__);

	if ((syndrome & 0xff00) == 0)
		return 0;

	if ((syndrome & 0x00ff) == 0)
		return 1;

	if ((syndrome & 0xf000) == 0 || (syndrome & 0x0f00) == 0)
		return 0;

	return 1;
}

static unsigned long ctl_page_to_phys(struct mem_ctl_info *mci,
		unsigned long page)
{
	u32 remap;
	struct e7xxx_pvt *pvt = (struct e7xxx_pvt *) mci->pvt_info;

	debugf3("%s()\n", __func__);

	if ((page < pvt->tolm) ||
			((page >= 0x100000) && (page < pvt->remapbase)))
		return page;

	remap = (page - pvt->tolm) + pvt->remapbase;

	if (remap < pvt->remaplimit)
		return remap;

	e7xxx_printk(KERN_ERR, "Invalid page %lx - out of range\n", page);
	return pvt->tolm - 1;
}

static void process_ce(struct mem_ctl_info *mci,
		struct e7xxx_error_info *info)
{
	u32 error_1b, page;
	u16 syndrome;
	int row;
	int channel;

	debugf3("%s()\n", __func__);
	/* read the error address */
	error_1b = info->dram_celog_add;
	/* FIXME - should use PAGE_SHIFT */
	page = error_1b >> 6;  /* convert the address to 4k page */
	/* read the syndrome */
	syndrome = info->dram_celog_syndrome;
	/* FIXME - check for -1 */
	row = edac_mc_find_csrow_by_page(mci, page);
	/* convert syndrome to channel */
	channel = e7xxx_find_channel(syndrome);
	edac_mc_handle_ce(mci, page, 0, syndrome, row, channel, "e7xxx CE");
}

static void process_ce_no_info(struct mem_ctl_info *mci)
{
	debugf3("%s()\n", __func__);
	edac_mc_handle_ce_no_info(mci, "e7xxx CE log register overflow");
}

static void process_ue(struct mem_ctl_info *mci,
		struct e7xxx_error_info *info)
{
	u32 error_2b, block_page;
	int row;

	debugf3("%s()\n", __func__);
	/* read the error address */
	error_2b = info->dram_uelog_add;
	/* FIXME - should use PAGE_SHIFT */
	block_page = error_2b >> 6;  /* convert to 4k address */
	row = edac_mc_find_csrow_by_page(mci, block_page);
	edac_mc_handle_ue(mci, block_page, 0, row, "e7xxx UE");
}

static void process_ue_no_info(struct mem_ctl_info *mci)
{
	debugf3("%s()\n", __func__);
	edac_mc_handle_ue_no_info(mci, "e7xxx UE log register overflow");
}

static void e7xxx_get_error_info (struct mem_ctl_info *mci,
		struct e7xxx_error_info *info)
{
	struct e7xxx_pvt *pvt;

	pvt = (struct e7xxx_pvt *) mci->pvt_info;
	pci_read_config_byte(pvt->bridge_ck, E7XXX_DRAM_FERR,
			&info->dram_ferr);
	pci_read_config_byte(pvt->bridge_ck, E7XXX_DRAM_NERR,
			&info->dram_nerr);

	if ((info->dram_ferr & 1) || (info->dram_nerr & 1)) {
		pci_read_config_dword(pvt->bridge_ck, E7XXX_DRAM_CELOG_ADD,
				&info->dram_celog_add);
		pci_read_config_word(pvt->bridge_ck,
				E7XXX_DRAM_CELOG_SYNDROME,
				&info->dram_celog_syndrome);
	}

	if ((info->dram_ferr & 2) || (info->dram_nerr & 2))
		pci_read_config_dword(pvt->bridge_ck, E7XXX_DRAM_UELOG_ADD,
				&info->dram_uelog_add);

	if (info->dram_ferr & 3)
		pci_write_bits8(pvt->bridge_ck, E7XXX_DRAM_FERR, 0x03, 0x03);

	if (info->dram_nerr & 3)
		pci_write_bits8(pvt->bridge_ck, E7XXX_DRAM_NERR, 0x03, 0x03);
}

static int e7xxx_process_error_info (struct mem_ctl_info *mci,
		struct e7xxx_error_info *info, int handle_errors)
{
	int error_found;

	error_found = 0;

	/* decode and report errors */
	if (info->dram_ferr & 1) {	/* check first error correctable */
		error_found = 1;

		if (handle_errors)
			process_ce(mci, info);
	}

	if (info->dram_ferr & 2) {	/* check first error uncorrectable */
		error_found = 1;

		if (handle_errors)
			process_ue(mci, info);
	}

	if (info->dram_nerr & 1) {	/* check next error correctable */
		error_found = 1;

		if (handle_errors) {
			if (info->dram_ferr & 1)
				process_ce_no_info(mci);
			else
				process_ce(mci, info);
		}
	}

	if (info->dram_nerr & 2) {	/* check next error uncorrectable */
		error_found = 1;

		if (handle_errors) {
			if (info->dram_ferr & 2)
				process_ue_no_info(mci);
			else
				process_ue(mci, info);
		}
	}

	return error_found;
}

static void e7xxx_check(struct mem_ctl_info *mci)
{
	struct e7xxx_error_info info;

	debugf3("%s()\n", __func__);
	e7xxx_get_error_info(mci, &info);
	e7xxx_process_error_info(mci, &info, 1);
}

/* Return 1 if dual channel mode is active.  Else return 0. */
static inline int dual_channel_active(u32 drc, int dev_idx)
{
	return (dev_idx == E7501) ? ((drc >> 22) & 0x1) : 1;
}


/* Return DRB granularity (0=32mb, 1=64mb). */
static inline int drb_granularity(u32 drc, int dev_idx)
{
	/* only e7501 can be single channel */
	return (dev_idx == E7501) ? ((drc >> 18) & 0x3) : 1;
}


static void e7xxx_init_csrows(struct mem_ctl_info *mci, struct pci_dev *pdev,
		int dev_idx, u32 drc)
{
	unsigned long last_cumul_size;
	int index;
	u8 value;
	u32 dra, cumul_size;
	int drc_chan, drc_drbg, drc_ddim, mem_dev;
	struct csrow_info *csrow;

	pci_read_config_dword(pdev, E7XXX_DRA, &dra);
	drc_chan = dual_channel_active(drc, dev_idx);
	drc_drbg = drb_granularity(drc, dev_idx);
	drc_ddim = (drc >> 20) & 0x3;
	last_cumul_size = 0;

	/* The dram row boundary (DRB) reg values are boundary address
	 * for each DRAM row with a granularity of 32 or 64MB (single/dual
	 * channel operation).  DRB regs are cumulative; therefore DRB7 will
	 * contain the total memory contained in all eight rows.
	 */
	for (index = 0; index < mci->nr_csrows; index++) {
		/* mem_dev 0=x8, 1=x4 */
		mem_dev = (dra >> (index * 4 + 3)) & 0x1;
		csrow = &mci->csrows[index];

		pci_read_config_byte(pdev, E7XXX_DRB + index, &value);
		/* convert a 64 or 32 MiB DRB to a page size. */
		cumul_size = value << (25 + drc_drbg - PAGE_SHIFT);
		debugf3("%s(): (%d) cumul_size 0x%x\n", __func__, index,
			cumul_size);
		if (cumul_size == last_cumul_size)
			continue;	/* not populated */

		csrow->first_page = last_cumul_size;
		csrow->last_page = cumul_size - 1;
		csrow->nr_pages = cumul_size - last_cumul_size;
		last_cumul_size = cumul_size;
		csrow->grain = 1 << 12;	/* 4KiB - resolution of CELOG */
		csrow->mtype = MEM_RDDR;	/* only one type supported */
		csrow->dtype = mem_dev ? DEV_X4 : DEV_X8;

		/*
		 * if single channel or x8 devices then SECDED
		 * if dual channel and x4 then S4ECD4ED
		 */
		if (drc_ddim) {
			if (drc_chan && mem_dev) {
				csrow->edac_mode = EDAC_S4ECD4ED;
				mci->edac_cap |= EDAC_FLAG_S4ECD4ED;
			} else {
				csrow->edac_mode = EDAC_SECDED;
				mci->edac_cap |= EDAC_FLAG_SECDED;
			}
		} else
			csrow->edac_mode = EDAC_NONE;
	}
}

static int e7xxx_probe1(struct pci_dev *pdev, int dev_idx)
{
	u16 pci_data;
	struct mem_ctl_info *mci = NULL;
	struct e7xxx_pvt *pvt = NULL;
	u32 drc;
	int drc_chan;
	struct e7xxx_error_info discard;

	debugf0("%s(): mci\n", __func__);

	/* make sure error reporting method is sane */
	switch(edac_op_state) {
		case EDAC_OPSTATE_POLL:
		case EDAC_OPSTATE_NMI:
			break;
		default:
			edac_op_state = EDAC_OPSTATE_POLL;
			break;
	}

	pci_read_config_dword(pdev, E7XXX_DRC, &drc);

	drc_chan = dual_channel_active(drc, dev_idx);
	mci = edac_mc_alloc(sizeof(*pvt), E7XXX_NR_CSROWS, drc_chan + 1);

	if (mci == NULL)
		return -ENOMEM;

	debugf3("%s(): init mci\n", __func__);
	mci->mtype_cap = MEM_FLAG_RDDR;
	mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED |
			EDAC_FLAG_S4ECD4ED;
	/* FIXME - what if different memory types are in different csrows? */
	mci->mod_name = EDAC_MOD_STR;
	mci->mod_ver = E7XXX_REVISION;
	mci->dev = &pdev->dev;
	debugf3("%s(): init pvt\n", __func__);
	pvt = (struct e7xxx_pvt *) mci->pvt_info;
	pvt->dev_info = &e7xxx_devs[dev_idx];
	pvt->bridge_ck = pci_get_device(PCI_VENDOR_ID_INTEL,
					pvt->dev_info->err_dev,
					pvt->bridge_ck);

	if (!pvt->bridge_ck) {
		e7xxx_printk(KERN_ERR, "error reporting device not found:"
			"vendor %x device 0x%x (broken BIOS?)\n",
			PCI_VENDOR_ID_INTEL, e7xxx_devs[dev_idx].err_dev);
		goto fail0;
	}

	debugf3("%s(): more mci init\n", __func__);
	mci->ctl_name = pvt->dev_info->ctl_name;
	mci->dev_name = pci_name(pdev);
	mci->edac_check = e7xxx_check;
	mci->ctl_page_to_phys = ctl_page_to_phys;
	e7xxx_init_csrows(mci, pdev, dev_idx, drc);
	mci->edac_cap |= EDAC_FLAG_NONE;
	debugf3("%s(): tolm, remapbase, remaplimit\n", __func__);
	/* load the top of low memory, remap base, and remap limit vars */
	pci_read_config_word(pdev, E7XXX_TOLM, &pci_data);
	pvt->tolm = ((u32) pci_data) << 4;
	pci_read_config_word(pdev, E7XXX_REMAPBASE, &pci_data);
	pvt->remapbase = ((u32) pci_data) << 14;
	pci_read_config_word(pdev, E7XXX_REMAPLIMIT, &pci_data);
	pvt->remaplimit = ((u32) pci_data) << 14;
	e7xxx_printk(KERN_INFO,
		"tolm = %x, remapbase = %x, remaplimit = %x\n", pvt->tolm,
		pvt->remapbase, pvt->remaplimit);

	/* clear any pending errors, or initial state bits */
	e7xxx_get_error_info(mci, &discard);

	/* Here we assume that we will never see multiple instances of this
	 * type of memory controller.  The ID is therefore hardcoded to 0.
	 */
	if (edac_mc_add_mc(mci,0)) {
		debugf3("%s(): failed edac_mc_add_mc()\n", __func__);
		goto fail1;
	}

	/* get this far and it's successful */
	debugf3("%s(): success\n", __func__);
	return 0;

fail1:
	pci_dev_put(pvt->bridge_ck);

fail0:
	edac_mc_free(mci);

	return -ENODEV;
}

/* returns count (>= 0), or negative on error */
static int __devinit e7xxx_init_one(struct pci_dev *pdev,
		const struct pci_device_id *ent)
{
	debugf0("%s()\n", __func__);

	/* wake up and enable device */
	return pci_enable_device(pdev) ?
		-EIO : e7xxx_probe1(pdev, ent->driver_data);
}

static void __devexit e7xxx_remove_one(struct pci_dev *pdev)
{
	struct mem_ctl_info *mci;
	struct e7xxx_pvt *pvt;

	debugf0("%s()\n", __func__);

	if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
		return;

	pvt = (struct e7xxx_pvt *) mci->pvt_info;
	pci_dev_put(pvt->bridge_ck);
	edac_mc_free(mci);
}

static const struct pci_device_id e7xxx_pci_tbl[] __devinitdata = {
	{
		PCI_VEND_DEV(INTEL, 7205_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
		E7205
	},
	{
		PCI_VEND_DEV(INTEL, 7500_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
		E7500
	},
	{
		PCI_VEND_DEV(INTEL, 7501_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
		E7501
	},
	{
		PCI_VEND_DEV(INTEL, 7505_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
		E7505
	},
	{
		0,
	}	/* 0 terminated list. */
};

MODULE_DEVICE_TABLE(pci, e7xxx_pci_tbl);

static struct pci_driver e7xxx_driver = {
	.name = EDAC_MOD_STR,
	.probe = e7xxx_init_one,
	.remove = __devexit_p(e7xxx_remove_one),
	.id_table = e7xxx_pci_tbl,
};

static int __init e7xxx_init(void)
{
	return pci_register_driver(&e7xxx_driver);
}

static void __exit e7xxx_exit(void)
{
	pci_unregister_driver(&e7xxx_driver);
}

module_init(e7xxx_init);
module_exit(e7xxx_exit);

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh et al\n"
	"Based on.work by Dan Hollis et al");
MODULE_DESCRIPTION("MC support for Intel e7xxx memory controllers");
module_param(edac_op_state, int, 0444);
MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
Commit	Line	Data
806c35f5 AC	1	/*
	2	* Intel e7xxx Memory Controller kernel module
	3	* (C) 2003 Linux Networx (http://lnxi.com)
	4	* This file may be distributed under the terms of the
	5	* GNU General Public License.
	6	*
	7	* See "enum e7xxx_chips" below for supported chipsets
	8	*
	9	* Written by Thayne Harbaugh
	10	* Based on work by Dan Hollis <goemon at anime dot net> and others.
	11	* http://www.anime.net/~goemon/linux-ecc/
	12	*
	13	* Contributors:
e7ecd891 DP	14	* Eric Biederman (Linux Networx)
	15	* Tom Zimmerman (Linux Networx)
	16	* Jim Garlick (Lawrence Livermore National Labs)
806c35f5 AC	17	* Dave Peterson (Lawrence Livermore National Labs)
	18	* That One Guy (Some other place)
	19	* Wang Zhenyu (intel.com)
	20	*
	21	* $Id: edac_e7xxx.c,v 1.5.2.9 2005/10/05 00:43:44 dsp_llnl Exp $
	22	*
	23	*/
	24
806c35f5 AC	25	#include <linux/module.h>
	26	#include <linux/init.h>
	27	#include <linux/pci.h>
	28	#include <linux/pci_ids.h>
	29	#include <linux/slab.h>
c0d12172	30	#include <linux/edac.h>
20bcb7a8	31	#include "edac_core.h"
806c35f5	32
20bcb7a8	33	#define E7XXX_REVISION " Ver: 2.0.2 " __DATE__
929a40ec	34	#define EDAC_MOD_STR "e7xxx_edac"
37f04581	35
537fba28	36	#define e7xxx_printk(level, fmt, arg...) \
e7ecd891	37	edac_printk(level, "e7xxx", fmt, ##arg)
537fba28 DP	38
537fba28 DP	39	#define e7xxx_mc_printk(mci, level, fmt, arg...) \
e7ecd891	40	edac_mc_chipset_printk(mci, level, "e7xxx", fmt, ##arg)
537fba28	41
806c35f5 AC	42	#ifndef PCI_DEVICE_ID_INTEL_7205_0
	43	#define PCI_DEVICE_ID_INTEL_7205_0 0x255d
	44	#endif /* PCI_DEVICE_ID_INTEL_7205_0 */
	45
	46	#ifndef PCI_DEVICE_ID_INTEL_7205_1_ERR
	47	#define PCI_DEVICE_ID_INTEL_7205_1_ERR 0x2551
	48	#endif /* PCI_DEVICE_ID_INTEL_7205_1_ERR */
	49
	50	#ifndef PCI_DEVICE_ID_INTEL_7500_0
	51	#define PCI_DEVICE_ID_INTEL_7500_0 0x2540
	52	#endif /* PCI_DEVICE_ID_INTEL_7500_0 */
	53
	54	#ifndef PCI_DEVICE_ID_INTEL_7500_1_ERR
	55	#define PCI_DEVICE_ID_INTEL_7500_1_ERR 0x2541
	56	#endif /* PCI_DEVICE_ID_INTEL_7500_1_ERR */
	57
	58	#ifndef PCI_DEVICE_ID_INTEL_7501_0
	59	#define PCI_DEVICE_ID_INTEL_7501_0 0x254c
	60	#endif /* PCI_DEVICE_ID_INTEL_7501_0 */
	61
	62	#ifndef PCI_DEVICE_ID_INTEL_7501_1_ERR
	63	#define PCI_DEVICE_ID_INTEL_7501_1_ERR 0x2541
	64	#endif /* PCI_DEVICE_ID_INTEL_7501_1_ERR */
	65
	66	#ifndef PCI_DEVICE_ID_INTEL_7505_0
	67	#define PCI_DEVICE_ID_INTEL_7505_0 0x2550
	68	#endif /* PCI_DEVICE_ID_INTEL_7505_0 */
	69
	70	#ifndef PCI_DEVICE_ID_INTEL_7505_1_ERR
	71	#define PCI_DEVICE_ID_INTEL_7505_1_ERR 0x2551
	72	#endif /* PCI_DEVICE_ID_INTEL_7505_1_ERR */
	73
806c35f5 AC	74	#define E7XXX_NR_CSROWS 8 /* number of csrows */
	75	#define E7XXX_NR_DIMMS 8 /* FIXME - is this correct? */
	76
806c35f5 AC	77	/* E7XXX register addresses - device 0 function 0 */
	78	#define E7XXX_DRB 0x60 /* DRAM row boundary register (8b) */
	79	#define E7XXX_DRA 0x70 /* DRAM row attribute register (8b) */
	80	/*
	81	* 31 Device width row 7 0=x8 1=x4
	82	* 27 Device width row 6
	83	* 23 Device width row 5
	84	* 19 Device width row 4
	85	* 15 Device width row 3
	86	* 11 Device width row 2
	87	* 7 Device width row 1
	88	* 3 Device width row 0
	89	*/
	90	#define E7XXX_DRC 0x7C /* DRAM controller mode reg (32b) */
	91	/*
	92	* 22 Number channels 0=1,1=2
	93	* 19:18 DRB Granularity 32/64MB
	94	*/
	95	#define E7XXX_TOLM 0xC4 /* DRAM top of low memory reg (16b) */
	96	#define E7XXX_REMAPBASE 0xC6 /* DRAM remap base address reg (16b) */
	97	#define E7XXX_REMAPLIMIT 0xC8 /* DRAM remap limit address reg (16b) */
	98
	99	/* E7XXX register addresses - device 0 function 1 */
	100	#define E7XXX_DRAM_FERR 0x80 /* DRAM first error register (8b) */
	101	#define E7XXX_DRAM_NERR 0x82 /* DRAM next error register (8b) */
	102	#define E7XXX_DRAM_CELOG_ADD 0xA0 /* DRAM first correctable memory */
	103	/* error address register (32b) */
	104	/*
	105	* 31:28 Reserved
	106	* 27:6 CE address (4k block 33:12)
	107	* 5:0 Reserved
	108	*/
	109	#define E7XXX_DRAM_UELOG_ADD 0xB0 /* DRAM first uncorrectable memory */
	110	/* error address register (32b) */
	111	/*
	112	* 31:28 Reserved
	113	* 27:6 CE address (4k block 33:12)
	114	* 5:0 Reserved
	115	*/
	116	#define E7XXX_DRAM_CELOG_SYNDROME 0xD0 /* DRAM first correctable memory */
	117	/* error syndrome register (16b) */
	118
	119	enum e7xxx_chips {
	120	E7500 = 0,
	121	E7501,
	122	E7505,
	123	E7205,
	124	};
	125
806c35f5 AC	126	struct e7xxx_pvt {
	127	struct pci_dev *bridge_ck;
	128	u32 tolm;
	129	u32 remapbase;
	130	u32 remaplimit;
	131	const struct e7xxx_dev_info *dev_info;
	132	};
	133
806c35f5 AC	134	struct e7xxx_dev_info {
	135	u16 err_dev;
	136	const char *ctl_name;
	137	};
	138
806c35f5 AC	139	struct e7xxx_error_info {
	140	u8 dram_ferr;
	141	u8 dram_nerr;
	142	u32 dram_celog_add;
	143	u16 dram_celog_syndrome;
	144	u32 dram_uelog_add;
	145	};
	146
	147	static const struct e7xxx_dev_info e7xxx_devs[] = {
	148	[E7500] = {
e7ecd891 DP	149	.err_dev = PCI_DEVICE_ID_INTEL_7500_1_ERR,
	150	.ctl_name = "E7500"
	151	},
806c35f5	152	[E7501] = {
e7ecd891 DP	153	.err_dev = PCI_DEVICE_ID_INTEL_7501_1_ERR,
	154	.ctl_name = "E7501"
	155	},
806c35f5	156	[E7505] = {
e7ecd891 DP	157	.err_dev = PCI_DEVICE_ID_INTEL_7505_1_ERR,
	158	.ctl_name = "E7505"
	159	},
806c35f5	160	[E7205] = {
e7ecd891 DP	161	.err_dev = PCI_DEVICE_ID_INTEL_7205_1_ERR,
	162	.ctl_name = "E7205"
	163	},
806c35f5 AC	164	};
806c35f5 AC	165
806c35f5 AC	166	/* FIXME - is this valid for both SECDED and S4ECD4ED? */
	167	static inline int e7xxx_find_channel(u16 syndrome)
	168	{
537fba28	169	debugf3("%s()\n", __func__);
806c35f5 AC	170
	171	if ((syndrome & 0xff00) == 0)
	172	return 0;
e7ecd891	173
806c35f5 AC	174	if ((syndrome & 0x00ff) == 0)
806c35f5 AC	175	return 1;
e7ecd891	176
806c35f5 AC	177	if ((syndrome & 0xf000) == 0 \|\| (syndrome & 0x0f00) == 0)
806c35f5 AC	178	return 0;
e7ecd891	179
806c35f5 AC	180	return 1;
	181	}
	182
e7ecd891 DP	183	static unsigned long ctl_page_to_phys(struct mem_ctl_info *mci,
e7ecd891 DP	184	unsigned long page)
806c35f5 AC	185	{
	186	u32 remap;
	187	struct e7xxx_pvt pvt = (struct e7xxx_pvt ) mci->pvt_info;
	188
537fba28	189	debugf3("%s()\n", __func__);
806c35f5 AC	190
806c35f5 AC	191	if ((page < pvt->tolm) \|\|
e7ecd891	192	((page >= 0x100000) && (page < pvt->remapbase)))
806c35f5	193	return page;
e7ecd891	194
806c35f5	195	remap = (page - pvt->tolm) + pvt->remapbase;
e7ecd891	196
806c35f5 AC	197	if (remap < pvt->remaplimit)
806c35f5 AC	198	return remap;
e7ecd891	199
537fba28	200	e7xxx_printk(KERN_ERR, "Invalid page %lx - out of range\n", page);
806c35f5 AC	201	return pvt->tolm - 1;
	202	}
	203
e7ecd891 DP	204	static void process_ce(struct mem_ctl_info *mci,
e7ecd891 DP	205	struct e7xxx_error_info *info)
806c35f5 AC	206	{
	207	u32 error_1b, page;
	208	u16 syndrome;
	209	int row;
	210	int channel;
	211
537fba28	212	debugf3("%s()\n", __func__);
806c35f5 AC	213	/* read the error address */
	214	error_1b = info->dram_celog_add;
	215	/* FIXME - should use PAGE_SHIFT */
e7ecd891	216	page = error_1b >> 6; /* convert the address to 4k page */
806c35f5 AC	217	/* read the syndrome */
	218	syndrome = info->dram_celog_syndrome;
	219	/* FIXME - check for -1 */
	220	row = edac_mc_find_csrow_by_page(mci, page);
	221	/* convert syndrome to channel */
	222	channel = e7xxx_find_channel(syndrome);
e7ecd891	223	edac_mc_handle_ce(mci, page, 0, syndrome, row, channel, "e7xxx CE");
806c35f5 AC	224	}
806c35f5 AC	225
806c35f5 AC	226	static void process_ce_no_info(struct mem_ctl_info *mci)
806c35f5 AC	227	{
537fba28	228	debugf3("%s()\n", __func__);
806c35f5 AC	229	edac_mc_handle_ce_no_info(mci, "e7xxx CE log register overflow");
	230	}
	231
e7ecd891 DP	232	static void process_ue(struct mem_ctl_info *mci,
e7ecd891 DP	233	struct e7xxx_error_info *info)
806c35f5 AC	234	{
	235	u32 error_2b, block_page;
	236	int row;
	237
537fba28	238	debugf3("%s()\n", __func__);
806c35f5 AC	239	/* read the error address */
	240	error_2b = info->dram_uelog_add;
	241	/* FIXME - should use PAGE_SHIFT */
e7ecd891	242	block_page = error_2b >> 6; /* convert to 4k address */
806c35f5 AC	243	row = edac_mc_find_csrow_by_page(mci, block_page);
	244	edac_mc_handle_ue(mci, block_page, 0, row, "e7xxx UE");
	245	}
	246
806c35f5 AC	247	static void process_ue_no_info(struct mem_ctl_info *mci)
806c35f5 AC	248	{
537fba28	249	debugf3("%s()\n", __func__);
806c35f5 AC	250	edac_mc_handle_ue_no_info(mci, "e7xxx UE log register overflow");
	251	}
	252
806c35f5 AC	253	static void e7xxx_get_error_info (struct mem_ctl_info *mci,
	254	struct e7xxx_error_info *info)
	255	{
	256	struct e7xxx_pvt *pvt;
	257
	258	pvt = (struct e7xxx_pvt *) mci->pvt_info;
	259	pci_read_config_byte(pvt->bridge_ck, E7XXX_DRAM_FERR,
e7ecd891	260	&info->dram_ferr);
806c35f5	261	pci_read_config_byte(pvt->bridge_ck, E7XXX_DRAM_NERR,
e7ecd891	262	&info->dram_nerr);
806c35f5 AC	263
	264	if ((info->dram_ferr & 1) \|\| (info->dram_nerr & 1)) {
	265	pci_read_config_dword(pvt->bridge_ck, E7XXX_DRAM_CELOG_ADD,
e7ecd891	266	&info->dram_celog_add);
806c35f5	267	pci_read_config_word(pvt->bridge_ck,
e7ecd891 DP	268	E7XXX_DRAM_CELOG_SYNDROME,
e7ecd891 DP	269	&info->dram_celog_syndrome);
806c35f5 AC	270	}
	271
	272	if ((info->dram_ferr & 2) \|\| (info->dram_nerr & 2))
	273	pci_read_config_dword(pvt->bridge_ck, E7XXX_DRAM_UELOG_ADD,
e7ecd891	274	&info->dram_uelog_add);
806c35f5 AC	275
806c35f5 AC	276	if (info->dram_ferr & 3)
e7ecd891	277	pci_write_bits8(pvt->bridge_ck, E7XXX_DRAM_FERR, 0x03, 0x03);
806c35f5 AC	278
806c35f5 AC	279	if (info->dram_nerr & 3)
e7ecd891	280	pci_write_bits8(pvt->bridge_ck, E7XXX_DRAM_NERR, 0x03, 0x03);
806c35f5 AC	281	}
806c35f5 AC	282
806c35f5 AC	283	static int e7xxx_process_error_info (struct mem_ctl_info *mci,
	284	struct e7xxx_error_info *info, int handle_errors)
	285	{
	286	int error_found;
	287
	288	error_found = 0;
	289
	290	/* decode and report errors */
	291	if (info->dram_ferr & 1) { /* check first error correctable */
	292	error_found = 1;
	293
	294	if (handle_errors)
	295	process_ce(mci, info);
	296	}
	297
	298	if (info->dram_ferr & 2) { /* check first error uncorrectable */
	299	error_found = 1;
	300
	301	if (handle_errors)
	302	process_ue(mci, info);
	303	}
	304
	305	if (info->dram_nerr & 1) { /* check next error correctable */
	306	error_found = 1;
	307
	308	if (handle_errors) {
	309	if (info->dram_ferr & 1)
	310	process_ce_no_info(mci);
	311	else
	312	process_ce(mci, info);
	313	}
	314	}
	315
	316	if (info->dram_nerr & 2) { /* check next error uncorrectable */
	317	error_found = 1;
	318
	319	if (handle_errors) {
	320	if (info->dram_ferr & 2)
	321	process_ue_no_info(mci);
	322	else
	323	process_ue(mci, info);
	324	}
	325	}
	326
	327	return error_found;
	328	}
	329
806c35f5 AC	330	static void e7xxx_check(struct mem_ctl_info *mci)
	331	{
	332	struct e7xxx_error_info info;
	333
537fba28	334	debugf3("%s()\n", __func__);
806c35f5 AC	335	e7xxx_get_error_info(mci, &info);
	336	e7xxx_process_error_info(mci, &info, 1);
	337	}
	338
13189525 DT	339	/* Return 1 if dual channel mode is active. Else return 0. */
13189525 DT	340	static inline int dual_channel_active(u32 drc, int dev_idx)
806c35f5	341	{
13189525 DT	342	return (dev_idx == E7501) ? ((drc >> 22) & 0x1) : 1;
13189525 DT	343	}
806c35f5	344
e7ecd891	345
13189525 DT	346	/* Return DRB granularity (0=32mb, 1=64mb). */
	347	static inline int drb_granularity(u32 drc, int dev_idx)
	348	{
806c35f5	349	/* only e7501 can be single channel */
13189525 DT	350	return (dev_idx == E7501) ? ((drc >> 18) & 0x3) : 1;
13189525 DT	351	}
e7ecd891	352
806c35f5	353
13189525 DT	354	static void e7xxx_init_csrows(struct mem_ctl_info mci, struct pci_dev pdev,
	355	int dev_idx, u32 drc)
	356	{
	357	unsigned long last_cumul_size;
	358	int index;
	359	u8 value;
	360	u32 dra, cumul_size;
	361	int drc_chan, drc_drbg, drc_ddim, mem_dev;
	362	struct csrow_info *csrow;
806c35f5	363
806c35f5	364	pci_read_config_dword(pdev, E7XXX_DRA, &dra);
13189525 DT	365	drc_chan = dual_channel_active(drc, dev_idx);
	366	drc_drbg = drb_granularity(drc, dev_idx);
	367	drc_ddim = (drc >> 20) & 0x3;
	368	last_cumul_size = 0;
806c35f5	369
13189525	370	/* The dram row boundary (DRB) reg values are boundary address
806c35f5 AC	371	* for each DRAM row with a granularity of 32 or 64MB (single/dual
	372	* channel operation). DRB regs are cumulative; therefore DRB7 will
	373	* contain the total memory contained in all eight rows.
	374	*/
13189525	375	for (index = 0; index < mci->nr_csrows; index++) {
806c35f5	376	/* mem_dev 0=x8, 1=x4 */
13189525 DT	377	mem_dev = (dra >> (index * 4 + 3)) & 0x1;
13189525 DT	378	csrow = &mci->csrows[index];
806c35f5	379
37f04581	380	pci_read_config_byte(pdev, E7XXX_DRB + index, &value);
806c35f5 AC	381	/* convert a 64 or 32 MiB DRB to a page size. */
806c35f5 AC	382	cumul_size = value << (25 + drc_drbg - PAGE_SHIFT);
537fba28 DP	383	debugf3("%s(): (%d) cumul_size 0x%x\n", __func__, index,
537fba28 DP	384	cumul_size);
806c35f5	385	if (cumul_size == last_cumul_size)
13189525	386	continue; /* not populated */
806c35f5 AC	387
	388	csrow->first_page = last_cumul_size;
	389	csrow->last_page = cumul_size - 1;
	390	csrow->nr_pages = cumul_size - last_cumul_size;
	391	last_cumul_size = cumul_size;
13189525 DT	392	csrow->grain = 1 << 12; /* 4KiB - resolution of CELOG */
13189525 DT	393	csrow->mtype = MEM_RDDR; /* only one type supported */
806c35f5 AC	394	csrow->dtype = mem_dev ? DEV_X4 : DEV_X8;
	395
	396	/*
	397	* if single channel or x8 devices then SECDED
	398	* if dual channel and x4 then S4ECD4ED
	399	*/
	400	if (drc_ddim) {
	401	if (drc_chan && mem_dev) {
	402	csrow->edac_mode = EDAC_S4ECD4ED;
	403	mci->edac_cap \|= EDAC_FLAG_S4ECD4ED;
	404	} else {
	405	csrow->edac_mode = EDAC_SECDED;
	406	mci->edac_cap \|= EDAC_FLAG_SECDED;
	407	}
	408	} else
	409	csrow->edac_mode = EDAC_NONE;
	410	}
13189525	411	}
806c35f5	412
13189525 DT	413	static int e7xxx_probe1(struct pci_dev *pdev, int dev_idx)
	414	{
	415	u16 pci_data;
	416	struct mem_ctl_info *mci = NULL;
	417	struct e7xxx_pvt *pvt = NULL;
	418	u32 drc;
	419	int drc_chan;
	420	struct e7xxx_error_info discard;
	421
	422	debugf0("%s(): mci\n", __func__);
c0d12172 DJ	423
	424	/* make sure error reporting method is sane */
	425	switch(edac_op_state) {
	426	case EDAC_OPSTATE_POLL:
	427	case EDAC_OPSTATE_NMI:
	428	break;
	429	default:
	430	edac_op_state = EDAC_OPSTATE_POLL;
	431	break;
	432	}
	433
13189525 DT	434	pci_read_config_dword(pdev, E7XXX_DRC, &drc);
	435
	436	drc_chan = dual_channel_active(drc, dev_idx);
	437	mci = edac_mc_alloc(sizeof(*pvt), E7XXX_NR_CSROWS, drc_chan + 1);
	438
	439	if (mci == NULL)
	440	return -ENOMEM;
806c35f5	441
13189525 DT	442	debugf3("%s(): init mci\n", __func__);
	443	mci->mtype_cap = MEM_FLAG_RDDR;
	444	mci->edac_ctl_cap = EDAC_FLAG_NONE \| EDAC_FLAG_SECDED \|
	445	EDAC_FLAG_S4ECD4ED;
	446	/* FIXME - what if different memory types are in different csrows? */
	447	mci->mod_name = EDAC_MOD_STR;
	448	mci->mod_ver = E7XXX_REVISION;
	449	mci->dev = &pdev->dev;
	450	debugf3("%s(): init pvt\n", __func__);
	451	pvt = (struct e7xxx_pvt *) mci->pvt_info;
	452	pvt->dev_info = &e7xxx_devs[dev_idx];
	453	pvt->bridge_ck = pci_get_device(PCI_VENDOR_ID_INTEL,
	454	pvt->dev_info->err_dev,
	455	pvt->bridge_ck);
	456
	457	if (!pvt->bridge_ck) {
	458	e7xxx_printk(KERN_ERR, "error reporting device not found:"
	459	"vendor %x device 0x%x (broken BIOS?)\n",
	460	PCI_VENDOR_ID_INTEL, e7xxx_devs[dev_idx].err_dev);
	461	goto fail0;
	462	}
	463
	464	debugf3("%s(): more mci init\n", __func__);
	465	mci->ctl_name = pvt->dev_info->ctl_name;
c4192705	466	mci->dev_name = pci_name(pdev);
13189525 DT	467	mci->edac_check = e7xxx_check;
	468	mci->ctl_page_to_phys = ctl_page_to_phys;
	469	e7xxx_init_csrows(mci, pdev, dev_idx, drc);
	470	mci->edac_cap \|= EDAC_FLAG_NONE;
537fba28	471	debugf3("%s(): tolm, remapbase, remaplimit\n", __func__);
806c35f5	472	/* load the top of low memory, remap base, and remap limit vars */
37f04581	473	pci_read_config_word(pdev, E7XXX_TOLM, &pci_data);
806c35f5	474	pvt->tolm = ((u32) pci_data) << 4;
37f04581	475	pci_read_config_word(pdev, E7XXX_REMAPBASE, &pci_data);
806c35f5	476	pvt->remapbase = ((u32) pci_data) << 14;
37f04581	477	pci_read_config_word(pdev, E7XXX_REMAPLIMIT, &pci_data);
806c35f5	478	pvt->remaplimit = ((u32) pci_data) << 14;
537fba28	479	e7xxx_printk(KERN_INFO,
e7ecd891 DP	480	"tolm = %x, remapbase = %x, remaplimit = %x\n", pvt->tolm,
e7ecd891 DP	481	pvt->remapbase, pvt->remaplimit);
806c35f5 AC	482
806c35f5 AC	483	/* clear any pending errors, or initial state bits */
749ede57	484	e7xxx_get_error_info(mci, &discard);
806c35f5	485
2d7bbb91 DT	486	/* Here we assume that we will never see multiple instances of this
	487	* type of memory controller. The ID is therefore hardcoded to 0.
	488	*/
	489	if (edac_mc_add_mc(mci,0)) {
537fba28	490	debugf3("%s(): failed edac_mc_add_mc()\n", __func__);
13189525	491	goto fail1;
806c35f5 AC	492	}
	493
	494	/* get this far and it's successful */
537fba28	495	debugf3("%s(): success\n", __func__);
806c35f5 AC	496	return 0;
806c35f5 AC	497
13189525 DT	498	fail1:
	499	pci_dev_put(pvt->bridge_ck);
	500
	501	fail0:
	502	edac_mc_free(mci);
806c35f5	503
13189525	504	return -ENODEV;
806c35f5 AC	505	}
	506
	507	/* returns count (>= 0), or negative on error */
e7ecd891 DP	508	static int __devinit e7xxx_init_one(struct pci_dev *pdev,
e7ecd891 DP	509	const struct pci_device_id *ent)
806c35f5	510	{
537fba28	511	debugf0("%s()\n", __func__);
806c35f5 AC	512
	513	/* wake up and enable device */
	514	return pci_enable_device(pdev) ?
e7ecd891	515	-EIO : e7xxx_probe1(pdev, ent->driver_data);
806c35f5 AC	516	}
806c35f5 AC	517
806c35f5 AC	518	static void __devexit e7xxx_remove_one(struct pci_dev *pdev)
	519	{
	520	struct mem_ctl_info *mci;
	521	struct e7xxx_pvt *pvt;
	522
537fba28	523	debugf0("%s()\n", __func__);
806c35f5	524
37f04581	525	if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
18dbc337 DP	526	return;
	527
	528	pvt = (struct e7xxx_pvt *) mci->pvt_info;
	529	pci_dev_put(pvt->bridge_ck);
	530	edac_mc_free(mci);
806c35f5 AC	531	}
806c35f5 AC	532
806c35f5	533	static const struct pci_device_id e7xxx_pci_tbl[] __devinitdata = {
e7ecd891 DP	534	{
	535	PCI_VEND_DEV(INTEL, 7205_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
	536	E7205
	537	},
	538	{
	539	PCI_VEND_DEV(INTEL, 7500_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
	540	E7500
	541	},
	542	{
	543	PCI_VEND_DEV(INTEL, 7501_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
	544	E7501
	545	},
	546	{
	547	PCI_VEND_DEV(INTEL, 7505_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
	548	E7505
	549	},
	550	{
	551	0,
	552	} /* 0 terminated list. */
806c35f5 AC	553	};
	554
	555	MODULE_DEVICE_TABLE(pci, e7xxx_pci_tbl);
	556
806c35f5	557	static struct pci_driver e7xxx_driver = {
680cbbbb	558	.name = EDAC_MOD_STR,
806c35f5 AC	559	.probe = e7xxx_init_one,
	560	.remove = __devexit_p(e7xxx_remove_one),
	561	.id_table = e7xxx_pci_tbl,
	562	};
	563
da9bb1d2	564	static int __init e7xxx_init(void)
806c35f5 AC	565	{
	566	return pci_register_driver(&e7xxx_driver);
	567	}
	568
806c35f5 AC	569	static void __exit e7xxx_exit(void)
	570	{
	571	pci_unregister_driver(&e7xxx_driver);
	572	}
	573
	574	module_init(e7xxx_init);
	575	module_exit(e7xxx_exit);
	576
806c35f5 AC	577	MODULE_LICENSE("GPL");
806c35f5 AC	578	MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh et al\n"
e7ecd891	579	"Based on.work by Dan Hollis et al");
806c35f5	580	MODULE_DESCRIPTION("MC support for Intel e7xxx memory controllers");
c0d12172 DJ	581	module_param(edac_op_state, int, 0444);
c0d12172 DJ	582	MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");