idle_page_tracking.rst (5481B)
1.. _idle_page_tracking: 2 3================== 4Idle Page Tracking 5================== 6 7Motivation 8========== 9 10The idle page tracking feature allows to track which memory pages are being 11accessed by a workload and which are idle. This information can be useful for 12estimating the workload's working set size, which, in turn, can be taken into 13account when configuring the workload parameters, setting memory cgroup limits, 14or deciding where to place the workload within a compute cluster. 15 16It is enabled by CONFIG_IDLE_PAGE_TRACKING=y. 17 18.. _user_api: 19 20User API 21======== 22 23The idle page tracking API is located at ``/sys/kernel/mm/page_idle``. 24Currently, it consists of the only read-write file, 25``/sys/kernel/mm/page_idle/bitmap``. 26 27The file implements a bitmap where each bit corresponds to a memory page. The 28bitmap is represented by an array of 8-byte integers, and the page at PFN #i is 29mapped to bit #i%64 of array element #i/64, byte order is native. When a bit is 30set, the corresponding page is idle. 31 32A page is considered idle if it has not been accessed since it was marked idle 33(for more details on what "accessed" actually means see the :ref:`Implementation 34Details <impl_details>` section). 35To mark a page idle one has to set the bit corresponding to 36the page by writing to the file. A value written to the file is OR-ed with the 37current bitmap value. 38 39Only accesses to user memory pages are tracked. These are pages mapped to a 40process address space, page cache and buffer pages, swap cache pages. For other 41page types (e.g. SLAB pages) an attempt to mark a page idle is silently ignored, 42and hence such pages are never reported idle. 43 44For huge pages the idle flag is set only on the head page, so one has to read 45``/proc/kpageflags`` in order to correctly count idle huge pages. 46 47Reading from or writing to ``/sys/kernel/mm/page_idle/bitmap`` will return 48-EINVAL if you are not starting the read/write on an 8-byte boundary, or 49if the size of the read/write is not a multiple of 8 bytes. Writing to 50this file beyond max PFN will return -ENXIO. 51 52That said, in order to estimate the amount of pages that are not used by a 53workload one should: 54 55 1. Mark all the workload's pages as idle by setting corresponding bits in 56 ``/sys/kernel/mm/page_idle/bitmap``. The pages can be found by reading 57 ``/proc/pid/pagemap`` if the workload is represented by a process, or by 58 filtering out alien pages using ``/proc/kpagecgroup`` in case the workload 59 is placed in a memory cgroup. 60 61 2. Wait until the workload accesses its working set. 62 63 3. Read ``/sys/kernel/mm/page_idle/bitmap`` and count the number of bits set. 64 If one wants to ignore certain types of pages, e.g. mlocked pages since they 65 are not reclaimable, he or she can filter them out using 66 ``/proc/kpageflags``. 67 68The page-types tool in the tools/vm directory can be used to assist in this. 69If the tool is run initially with the appropriate option, it will mark all the 70queried pages as idle. Subsequent runs of the tool can then show which pages have 71their idle flag cleared in the interim. 72 73See :ref:`Documentation/admin-guide/mm/pagemap.rst <pagemap>` for more 74information about ``/proc/pid/pagemap``, ``/proc/kpageflags``, and 75``/proc/kpagecgroup``. 76 77.. _impl_details: 78 79Implementation Details 80====================== 81 82The kernel internally keeps track of accesses to user memory pages in order to 83reclaim unreferenced pages first on memory shortage conditions. A page is 84considered referenced if it has been recently accessed via a process address 85space, in which case one or more PTEs it is mapped to will have the Accessed bit 86set, or marked accessed explicitly by the kernel (see mark_page_accessed()). The 87latter happens when: 88 89 - a userspace process reads or writes a page using a system call (e.g. read(2) 90 or write(2)) 91 92 - a page that is used for storing filesystem buffers is read or written, 93 because a process needs filesystem metadata stored in it (e.g. lists a 94 directory tree) 95 96 - a page is accessed by a device driver using get_user_pages() 97 98When a dirty page is written to swap or disk as a result of memory reclaim or 99exceeding the dirty memory limit, it is not marked referenced. 100 101The idle memory tracking feature adds a new page flag, the Idle flag. This flag 102is set manually, by writing to ``/sys/kernel/mm/page_idle/bitmap`` (see the 103:ref:`User API <user_api>` 104section), and cleared automatically whenever a page is referenced as defined 105above. 106 107When a page is marked idle, the Accessed bit must be cleared in all PTEs it is 108mapped to, otherwise we will not be able to detect accesses to the page coming 109from a process address space. To avoid interference with the reclaimer, which, 110as noted above, uses the Accessed bit to promote actively referenced pages, one 111more page flag is introduced, the Young flag. When the PTE Accessed bit is 112cleared as a result of setting or updating a page's Idle flag, the Young flag 113is set on the page. The reclaimer treats the Young flag as an extra PTE 114Accessed bit and therefore will consider such a page as referenced. 115 116Since the idle memory tracking feature is based on the memory reclaimer logic, 117it only works with pages that are on an LRU list, other pages are silently 118ignored. That means it will ignore a user memory page if it is isolated, but 119since there are usually not many of them, it should not affect the overall 120result noticeably. In order not to stall scanning of the idle page bitmap, 121locked pages may be skipped too.