1 2PAT (Page Attribute Table) 3 4x86 Page Attribute Table (PAT) allows for setting the memory attribute at the 5page level granularity. PAT is complementary to the MTRR settings which allows 6for setting of memory types over physical address ranges. However, PAT is 7more flexible than MTRR due to its capability to set attributes at page level 8and also due to the fact that there are no hardware limitations on number of 9such attribute settings allowed. Added flexibility comes with guidelines for 10not having memory type aliasing for the same physical memory with multiple 11virtual addresses. 12 13PAT allows for different types of memory attributes. The most commonly used 14ones that will be supported at this time are Write-back, Uncached, 15Write-combined and Uncached Minus. 16 17 18PAT APIs 19-------- 20 21There are many different APIs in the kernel that allows setting of memory 22attributes at the page level. In order to avoid aliasing, these interfaces 23should be used thoughtfully. Below is a table of interfaces available, 24their intended usage and their memory attribute relationships. Internally, 25these APIs use a reserve_memtype()/free_memtype() interface on the physical 26address range to avoid any aliasing. 27 28 29------------------------------------------------------------------- 30API | RAM | ACPI,... | Reserved/Holes | 31-----------------------|----------|------------|------------------| 32 | | | | 33ioremap | -- | UC- | UC- | 34 | | | | 35ioremap_cache | -- | WB | WB | 36 | | | | 37ioremap_nocache | -- | UC- | UC- | 38 | | | | 39ioremap_wc | -- | -- | WC | 40 | | | | 41set_memory_uc | UC- | -- | -- | 42 set_memory_wb | | | | 43 | | | | 44set_memory_wc | WC | -- | -- | 45 set_memory_wb | | | | 46 | | | | 47pci sysfs resource | -- | -- | UC- | 48 | | | | 49pci sysfs resource_wc | -- | -- | WC | 50 is IORESOURCE_PREFETCH| | | | 51 | | | | 52pci proc | -- | -- | UC- | 53 !PCIIOC_WRITE_COMBINE | | | | 54 | | | | 55pci proc | -- | -- | WC | 56 PCIIOC_WRITE_COMBINE | | | | 57 | | | | 58/dev/mem | -- | WB/WC/UC- | WB/WC/UC- | 59 read-write | | | | 60 | | | | 61/dev/mem | -- | UC- | UC- | 62 mmap SYNC flag | | | | 63 | | | | 64/dev/mem | -- | WB/WC/UC- | WB/WC/UC- | 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 | | | | 75/dev/mem | -- | -- | UC- | 76 mmap !SYNC flag | | | | 77 no alias to this area | | | | 78 and | | | | 79 MTRR says !WB | | | | 80 | | | | 81------------------------------------------------------------------- 82 83Advanced APIs for drivers 84------------------------- 85A. Exporting pages to users with remap_pfn_range, io_remap_pfn_range, 86vm_insert_pfn 87 88Drivers wanting to export some pages to userspace do it by using mmap 89interface and a combination of 901) pgprot_noncached() 912) io_remap_pfn_range() or remap_pfn_range() or vm_insert_pfn() 92 93With PAT support, a new API pgprot_writecombine is being added. So, drivers can 94continue to use the above sequence, with either pgprot_noncached() or 95pgprot_writecombine() in step 1, followed by step 2. 96 97In addition, step 2 internally tracks the region as UC or WC in memtype 98list in order to ensure no conflicting mapping. 99 100Note that this set of APIs only works with IO (non RAM) regions. If driver 101wants to export a RAM region, it has to do set_memory_uc() or set_memory_wc() 102as step 0 above and also track the usage of those pages and use set_memory_wb() 103before the page is freed to free pool. 104 105 106 107Notes: 108 109-- in the above table mean "Not suggested usage for the API". Some of the --'s 110are strictly enforced by the kernel. Some others are not really enforced 111today, but may be enforced in future. 112 113For ioremap and pci access through /sys or /proc - The actual type returned 114can be more restrictive, in case of any existing aliasing for that address. 115For example: If there is an existing uncached mapping, a new ioremap_wc can 116return uncached mapping in place of write-combine requested. 117 118set_memory_[uc|wc] and set_memory_wb should be used in pairs, where driver will 119first make a region uc or wc and switch it back to wb after use. 120 121Over time writes to /proc/mtrr will be deprecated in favor of using PAT based 122interfaces. Users writing to /proc/mtrr are suggested to use above interfaces. 123 124Drivers should use ioremap_[uc|wc] to access PCI BARs with [uc|wc] access 125types. 126 127Drivers should use set_memory_[uc|wc] to set access type for RAM ranges. 128 129 130PAT debugging 131------------- 132 133With 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 137PAT memtype list: 138uncached-minus @ 0x7fadf000-0x7fae0000 139uncached-minus @ 0x7fb19000-0x7fb1a000 140uncached-minus @ 0x7fb1a000-0x7fb1b000 141uncached-minus @ 0x7fb1b000-0x7fb1c000 142uncached-minus @ 0x7fb1c000-0x7fb1d000 143uncached-minus @ 0x7fb1d000-0x7fb1e000 144uncached-minus @ 0x7fb1e000-0x7fb25000 145uncached-minus @ 0x7fb25000-0x7fb26000 146uncached-minus @ 0x7fb26000-0x7fb27000 147uncached-minus @ 0x7fb27000-0x7fb28000 148uncached-minus @ 0x7fb28000-0x7fb2e000 149uncached-minus @ 0x7fb2e000-0x7fb2f000 150uncached-minus @ 0x7fb2f000-0x7fb30000 151uncached-minus @ 0x7fb31000-0x7fb32000 152uncached-minus @ 0x80000000-0x90000000 153 154This list shows physical address ranges and various PAT settings used to 155access those physical address ranges. 156 157Another, more verbose way of getting PAT related debug messages is with 158"debugpat" boot parameter. With this parameter, various debug messages are 159printed to dmesg log. 160 161