| blob | 2a4ee6302122f8942ac08f6d26dbdf2f369f4e6e |
2 PAT (Page Attribute Table)
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.
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, Write-through and Uncached Minus.
18 PAT APIs
19 --------
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.
29 -------------------------------------------------------------------
30 API | RAM | ACPI,... | Reserved/Holes |
31 -----------------------|----------|------------|------------------|
32 | | | |
33 ioremap | -- | UC- | UC- |
34 | | | |
35 ioremap_cache | -- | WB | WB |
36 | | | |
37 ioremap_uc | -- | UC | UC |
38 | | | |
39 ioremap_nocache | -- | UC- | UC- |
40 | | | |
41 ioremap_wc | -- | -- | WC |
42 | | | |
43 ioremap_wt | -- | -- | WT |
44 | | | |
45 set_memory_uc | UC- | -- | -- |
46 set_memory_wb | | | |
47 | | | |
48 set_memory_wc | WC | -- | -- |
49 set_memory_wb | | | |
50 | | | |
51 set_memory_wt | WT | -- | -- |
52 set_memory_wb | | | |
53 | | | |
54 pci sysfs resource | -- | -- | UC- |
55 | | | |
56 pci sysfs resource_wc | -- | -- | WC |
57 is IORESOURCE_PREFETCH| | | |
58 | | | |
59 pci proc | -- | -- | UC- |
60 !PCIIOC_WRITE_COMBINE | | | |
61 | | | |
62 pci proc | -- | -- | WC |
63 PCIIOC_WRITE_COMBINE | | | |
64 | | | |
65 /dev/mem | -- | WB/WC/UC- | WB/WC/UC- |
66 read-write | | | |
67 | | | |
68 /dev/mem | -- | UC- | UC- |
69 mmap SYNC flag | | | |
70 | | | |
71 /dev/mem | -- | WB/WC/UC- | WB/WC/UC- |
72 mmap !SYNC flag | |(from exist-| (from exist- |
73 and | | ing alias)| ing alias) |
74 any alias to this area| | | |
75 | | | |
76 /dev/mem | -- | WB | WB |
77 mmap !SYNC flag | | | |
78 no alias to this area | | | |
79 and | | | |
80 MTRR says WB | | | |
81 | | | |
82 /dev/mem | -- | -- | UC- |
83 mmap !SYNC flag | | | |
84 no alias to this area | | | |
85 and | | | |
86 MTRR says !WB | | | |
87 | | | |
88 -------------------------------------------------------------------
90 Advanced APIs for drivers
91 -------------------------
92 A. Exporting pages to users with remap_pfn_range, io_remap_pfn_range,
93 vm_insert_pfn
95 Drivers wanting to export some pages to userspace do it by using mmap
96 interface and a combination of
97 1) pgprot_noncached()
98 2) io_remap_pfn_range() or remap_pfn_range() or vm_insert_pfn()
100 With PAT support, a new API pgprot_writecombine is being added. So, drivers can
101 continue to use the above sequence, with either pgprot_noncached() or
102 pgprot_writecombine() in step 1, followed by step 2.
104 In addition, step 2 internally tracks the region as UC or WC in memtype
105 list in order to ensure no conflicting mapping.
107 Note that this set of APIs only works with IO (non RAM) regions. If driver
108 wants to export a RAM region, it has to do set_memory_uc() or set_memory_wc()
109 as step 0 above and also track the usage of those pages and use set_memory_wb()
110 before the page is freed to free pool.
112 MTRR effects on PAT / non-PAT systems
113 -------------------------------------
115 The following table provides the effects of using write-combining MTRRs when
116 using ioremap*() calls on x86 for both non-PAT and PAT systems. Ideally
117 mtrr_add() usage will be phased out in favor of arch_phys_wc_add() which will
118 be a no-op on PAT enabled systems. The region over which a arch_phys_wc_add()
119 is made, should already have been ioremapped with WC attributes or PAT entries,
120 this can be done by using ioremap_wc() / set_memory_wc(). Devices which
121 combine areas of IO memory desired to remain uncacheable with areas where
122 write-combining is desirable should consider use of ioremap_uc() followed by
123 set_memory_wc() to white-list effective write-combined areas. Such use is
124 nevertheless discouraged as the effective memory type is considered
125 implementation defined, yet this strategy can be used as last resort on devices
126 with size-constrained regions where otherwise MTRR write-combining would
127 otherwise not be effective.
129 ----------------------------------------------------------------------
130 MTRR Non-PAT PAT Linux ioremap value Effective memory type
131 ----------------------------------------------------------------------
132 Non-PAT | PAT
133 PAT
134 |PCD
135 ||PWT
136 |||
137 WC 000 WB _PAGE_CACHE_MODE_WB WC | WC
138 WC 001 WC _PAGE_CACHE_MODE_WC WC* | WC
139 WC 010 UC- _PAGE_CACHE_MODE_UC_MINUS WC* | UC
140 WC 011 UC _PAGE_CACHE_MODE_UC UC | UC
141 ----------------------------------------------------------------------
143 (*) denotes implementation defined and is discouraged
145 Notes:
147 -- in the above table mean "Not suggested usage for the API". Some of the --'s
148 are strictly enforced by the kernel. Some others are not really enforced
149 today, but may be enforced in future.
151 For ioremap and pci access through /sys or /proc - The actual type returned
152 can be more restrictive, in case of any existing aliasing for that address.
153 For example: If there is an existing uncached mapping, a new ioremap_wc can
154 return uncached mapping in place of write-combine requested.
156 set_memory_[uc|wc|wt] and set_memory_wb should be used in pairs, where driver
157 will first make a region uc, wc or wt and switch it back to wb after use.
159 Over time writes to /proc/mtrr will be deprecated in favor of using PAT based
160 interfaces. Users writing to /proc/mtrr are suggested to use above interfaces.
162 Drivers should use ioremap_[uc|wc] to access PCI BARs with [uc|wc] access
163 types.
165 Drivers should use set_memory_[uc|wc|wt] to set access type for RAM ranges.
168 PAT debugging
169 -------------
171 With CONFIG_DEBUG_FS enabled, PAT memtype list can be examined by
173 # mount -t debugfs debugfs /sys/kernel/debug
174 # cat /sys/kernel/debug/x86/pat_memtype_list
175 PAT memtype list:
176 uncached-minus @ 0x7fadf000-0x7fae0000
177 uncached-minus @ 0x7fb19000-0x7fb1a000
178 uncached-minus @ 0x7fb1a000-0x7fb1b000
179 uncached-minus @ 0x7fb1b000-0x7fb1c000
180 uncached-minus @ 0x7fb1c000-0x7fb1d000
181 uncached-minus @ 0x7fb1d000-0x7fb1e000
182 uncached-minus @ 0x7fb1e000-0x7fb25000
183 uncached-minus @ 0x7fb25000-0x7fb26000
184 uncached-minus @ 0x7fb26000-0x7fb27000
185 uncached-minus @ 0x7fb27000-0x7fb28000
186 uncached-minus @ 0x7fb28000-0x7fb2e000
187 uncached-minus @ 0x7fb2e000-0x7fb2f000
188 uncached-minus @ 0x7fb2f000-0x7fb30000
189 uncached-minus @ 0x7fb31000-0x7fb32000
190 uncached-minus @ 0x80000000-0x90000000
192 This list shows physical address ranges and various PAT settings used to
193 access those physical address ranges.
195 Another, more verbose way of getting PAT related debug messages is with
196 "debugpat" boot parameter. With this parameter, various debug messages are
197 printed to dmesg log.
199 PAT Initialization
200 ------------------
202 The following table describes how PAT is initialized under various
203 configurations. The PAT MSR must be updated by Linux in order to support WC
204 and WT attributes. Otherwise, the PAT MSR has the value programmed in it
205 by the firmware. Note, Xen enables WC attribute in the PAT MSR for guests.
207 MTRR PAT Call Sequence PAT State PAT MSR
208 =========================================================
209 E E MTRR -> PAT init Enabled OS
210 E D MTRR -> PAT init Disabled -
211 D E MTRR -> PAT disable Disabled BIOS
212 D D MTRR -> PAT disable Disabled -
213 - np/E PAT -> PAT disable Disabled BIOS
214 - np/D PAT -> PAT disable Disabled -
215 E !P/E MTRR -> PAT init Disabled BIOS
216 D !P/E MTRR -> PAT disable Disabled BIOS
217 !M !P/E MTRR stub -> PAT disable Disabled BIOS
219 Legend
220 ------------------------------------------------
221 E Feature enabled in CPU
222 D Feature disabled/unsupported in CPU
223 np "nopat" boot option specified
224 !P CONFIG_X86_PAT option unset
225 !M CONFIG_MTRR option unset
226 Enabled PAT state set to enabled
227 Disabled PAT state set to disabled
228 OS PAT initializes PAT MSR with OS setting
229 BIOS PAT keeps PAT MSR with BIOS setting