d27554d8 |
1 | |
2 | PAT (Page Attribute Table) |
3 | |
4 | x86 Page Attribute Table (PAT) allows for setting the memory attribute at the |
5 | page level granularity. PAT is complementary to the MTRR settings which allows |
6 | for setting of memory types over physical address ranges. However, PAT is |
7 | more flexible than MTRR due to its capability to set attributes at page level |
8 | and also due to the fact that there are no hardware limitations on number of |
9 | such attribute settings allowed. Added flexibility comes with guidelines for |
10 | not having memory type aliasing for the same physical memory with multiple |
11 | virtual addresses. |
12 | |
13 | PAT allows for different types of memory attributes. The most commonly used |
14 | ones that will be supported at this time are Write-back, Uncached, |
15 | Write-combined and Uncached Minus. |
16 | |
59dfc3f8 |
17 | |
18 | PAT APIs |
19 | -------- |
20 | |
d27554d8 |
21 | There are many different APIs in the kernel that allows setting of memory |
22 | attributes at the page level. In order to avoid aliasing, these interfaces |
23 | should be used thoughtfully. Below is a table of interfaces available, |
24 | their intended usage and their memory attribute relationships. Internally, |
25 | these APIs use a reserve_memtype()/free_memtype() interface on the physical |
26 | address range to avoid any aliasing. |
27 | |
28 | |
29 | ------------------------------------------------------------------- |
30 | API | RAM | ACPI,... | Reserved/Holes | |
31 | -----------------------|----------|------------|------------------| |
32 | | | | | |
59dfc3f8 |
33 | ioremap | -- | UC- | UC- | |
d27554d8 |
34 | | | | | |
35 | ioremap_cache | -- | WB | WB | |
36 | | | | | |
59dfc3f8 |
37 | ioremap_nocache | -- | UC- | UC- | |
d27554d8 |
38 | | | | | |
39 | ioremap_wc | -- | -- | WC | |
40 | | | | | |
59dfc3f8 |
41 | set_memory_uc | UC- | -- | -- | |
d27554d8 |
42 | set_memory_wb | | | | |
43 | | | | | |
44 | set_memory_wc | WC | -- | -- | |
45 | set_memory_wb | | | | |
46 | | | | | |
59dfc3f8 |
47 | pci sysfs resource | -- | -- | UC- | |
d27554d8 |
48 | | | | | |
49 | pci sysfs resource_wc | -- | -- | WC | |
50 | is IORESOURCE_PREFETCH| | | | |
51 | | | | | |
59dfc3f8 |
52 | pci proc | -- | -- | UC- | |
d27554d8 |
53 | !PCIIOC_WRITE_COMBINE | | | | |
54 | | | | | |
55 | pci proc | -- | -- | WC | |
56 | PCIIOC_WRITE_COMBINE | | | | |
57 | | | | | |
59dfc3f8 |
58 | /dev/mem | -- | WB/WC/UC- | WB/WC/UC- | |
d27554d8 |
59 | read-write | | | | |
60 | | | | | |
59dfc3f8 |
61 | /dev/mem | -- | UC- | UC- | |
d27554d8 |
62 | mmap SYNC flag | | | | |
63 | | | | | |
59dfc3f8 |
64 | /dev/mem | -- | WB/WC/UC- | WB/WC/UC- | |
d27554d8 |
65 | mmap !SYNC flag | |(from exist-| (from exist- | |
66 | and | | ing alias)| ing alias) | |
67 | any alias to this area| | | | |
68 | | | | | |
69 | /dev/mem | -- | WB | WB | |
70 | mmap !SYNC flag | | | | |
71 | no alias to this area | | | | |
72 | and | | | | |
73 | MTRR says WB | | | | |
74 | | | | | |
59dfc3f8 |
75 | /dev/mem | -- | -- | UC- | |
d27554d8 |
76 | mmap !SYNC flag | | | | |
77 | no alias to this area | | | | |
78 | and | | | | |
79 | MTRR says !WB | | | | |
80 | | | | | |
81 | ------------------------------------------------------------------- |
82 | |
a2ced6e1 |
83 | Advanced APIs for drivers |
84 | ------------------------- |
67bac792 |
85 | A. Exporting pages to users with remap_pfn_range, io_remap_pfn_range, |
a2ced6e1 |
86 | vm_insert_pfn |
87 | |
67bac792 |
88 | Drivers wanting to export some pages to userspace do it by using mmap |
a2ced6e1 |
89 | interface and a combination of |
90 | 1) pgprot_noncached() |
91 | 2) io_remap_pfn_range() or remap_pfn_range() or vm_insert_pfn() |
92 | |
67bac792 |
93 | With PAT support, a new API pgprot_writecombine is being added. So, drivers can |
a2ced6e1 |
94 | continue to use the above sequence, with either pgprot_noncached() or |
95 | pgprot_writecombine() in step 1, followed by step 2. |
96 | |
97 | In addition, step 2 internally tracks the region as UC or WC in memtype |
98 | list in order to ensure no conflicting mapping. |
99 | |
67bac792 |
100 | Note that this set of APIs only works with IO (non RAM) regions. If driver |
101 | wants to export a RAM region, it has to do set_memory_uc() or set_memory_wc() |
a2ced6e1 |
102 | as step 0 above and also track the usage of those pages and use set_memory_wb() |
103 | before the page is freed to free pool. |
104 | |
105 | |
106 | |
d27554d8 |
107 | Notes: |
108 | |
109 | -- in the above table mean "Not suggested usage for the API". Some of the --'s |
110 | are strictly enforced by the kernel. Some others are not really enforced |
111 | today, but may be enforced in future. |
112 | |
113 | For ioremap and pci access through /sys or /proc - The actual type returned |
114 | can be more restrictive, in case of any existing aliasing for that address. |
115 | For example: If there is an existing uncached mapping, a new ioremap_wc can |
116 | return uncached mapping in place of write-combine requested. |
117 | |
118 | set_memory_[uc|wc] and set_memory_wb should be used in pairs, where driver will |
119 | first make a region uc or wc and switch it back to wb after use. |
120 | |
121 | Over time writes to /proc/mtrr will be deprecated in favor of using PAT based |
122 | interfaces. Users writing to /proc/mtrr are suggested to use above interfaces. |
123 | |
124 | Drivers should use ioremap_[uc|wc] to access PCI BARs with [uc|wc] access |
125 | types. |
126 | |
127 | Drivers should use set_memory_[uc|wc] to set access type for RAM ranges. |
128 | |
59dfc3f8 |
129 | |
130 | PAT debugging |
131 | ------------- |
132 | |
133 | With CONFIG_DEBUG_FS enabled, PAT memtype list can be examined by |
134 | |
135 | # mount -t debugfs debugfs /sys/kernel/debug |
136 | # cat /sys/kernel/debug/x86/pat_memtype_list |
137 | PAT memtype list: |
138 | uncached-minus @ 0x7fadf000-0x7fae0000 |
139 | uncached-minus @ 0x7fb19000-0x7fb1a000 |
140 | uncached-minus @ 0x7fb1a000-0x7fb1b000 |
141 | uncached-minus @ 0x7fb1b000-0x7fb1c000 |
142 | uncached-minus @ 0x7fb1c000-0x7fb1d000 |
143 | uncached-minus @ 0x7fb1d000-0x7fb1e000 |
144 | uncached-minus @ 0x7fb1e000-0x7fb25000 |
145 | uncached-minus @ 0x7fb25000-0x7fb26000 |
146 | uncached-minus @ 0x7fb26000-0x7fb27000 |
147 | uncached-minus @ 0x7fb27000-0x7fb28000 |
148 | uncached-minus @ 0x7fb28000-0x7fb2e000 |
149 | uncached-minus @ 0x7fb2e000-0x7fb2f000 |
150 | uncached-minus @ 0x7fb2f000-0x7fb30000 |
151 | uncached-minus @ 0x7fb31000-0x7fb32000 |
152 | uncached-minus @ 0x80000000-0x90000000 |
153 | |
154 | This list shows physical address ranges and various PAT settings used to |
155 | access those physical address ranges. |
156 | |
157 | Another, more verbose way of getting PAT related debug messages is with |
158 | "debugpat" boot parameter. With this parameter, various debug messages are |
159 | printed to dmesg log. |
160 | |