- 09 Nov, 2022 39 commits
-
-
Liu Shixin authored
The priority of hotplug memory callback is defined in a different file. And there are some callers using numbers directly. Collect them together into include/linux/memory.h for easy reading. This allows us to sort their priorities more intuitively without additional comments. Link: https://lkml.kernel.org/r/20220923033347.3935160-9-liushixin2@huawei.comSigned-off-by:
Liu Shixin <liushixin2@huawei.com> Cc: Christoph Lameter <cl@linux.com> Cc: David Hildenbrand <david@redhat.com> Cc: Kefeng Wang <wangkefeng.wang@huawei.com> Cc: Waiman Long <longman@redhat.com> Cc: zefan li <lizefan.x@bytedance.com> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org>
-
Liu Shixin authored
Remove unused register_hotmemory_notifier(). Link: https://lkml.kernel.org/r/20220923033347.3935160-8-liushixin2@huawei.comSigned-off-by:
David Hildenbrand <david@redhat.com> Cc: Christoph Lameter <cl@linux.com> Cc: Kefeng Wang <wangkefeng.wang@huawei.com> Cc: Waiman Long <longman@redhat.com> Cc: zefan li <lizefan.x@bytedance.com> Signed-off-by:
-
Liu Shixin authored
Commit 76ae8474 ("Documentation: raise minimum supported version of GCC to 5.1") updated the minimum gcc version to 5.1. So the problem mentioned in f02c6968 ("include/linux/memory.h: implement register_hotmemory_notifier()") no longer exist. So we can now switch to use hotplug_memory_notifier() directly rather than register_hotmemory_notifier(). Link: https://lkml.kernel.org/r/20220923033347.3935160-7-liushixin2@huawei.comSigned-off-by:
-
Liu Shixin authored
Commit 76ae8474 ("Documentation: raise minimum supported version of GCC to 5.1") updated the minimum gcc version to 5.1. So the problem mentioned in f02c6968 ("include/linux/memory.h: implement register_hotmemory_notifier()") no longer exist. So we can now switch to use hotplug_memory_notifier() directly rather than register_hotmemory_notifier(). Link: https://lkml.kernel.org/r/20220923033347.3935160-6-liushixin2@huawei.comSigned-off-by:
-
Liu Shixin authored
Commit 76ae8474 ("Documentation: raise minimum supported version of GCC to 5.1") updated the minimum gcc version to 5.1. So the problem mentioned in f02c6968 ("include/linux/memory.h: implement register_hotmemory_notifier()") no longer exist. So we can now switch to use hotplug_memory_notifier() directly rather than register_hotmemory_notifier(). Link: https://lkml.kernel.org/r/20220923033347.3935160-5-liushixin2@huawei.comSigned-off-by:
-
Liu Shixin authored
Commit 76ae8474 ("Documentation: raise minimum supported version of GCC to 5.1") updated the minimum gcc version to 5.1. So the problem mentioned in f02c6968 ("include/linux/memory.h: implement register_hotmemory_notifier()") no longer exist. So we can now switch to use hotplug_memory_notifier() directly rather than register_hotmemory_notifier(). Link: https://lkml.kernel.org/r/20220923033347.3935160-4-liushixin2@huawei.comSigned-off-by:
-
Liu Shixin authored
Commit 76ae8474 ("Documentation: raise minimum supported version of GCC to 5.1") updated the minimum gcc version to 5.1. So the problem mentioned in f02c6968 ("include/linux/memory.h: implement register_hotmemory_notifier()") no longer exist. So we can now switch to use hotplug_memory_notifier() directly rather than register_hotmemory_notifier(). Link: https://lkml.kernel.org/r/20220923033347.3935160-3-liushixin2@huawei.comSigned-off-by:
-
Liu Shixin authored
Patch series "mm: Use hotplug_memory_notifier() instead of register_hotmemory_notifier()", v4. Commit f02c6968 ("include/linux/memory.h: implement register_hotmemory_notifier()") introduced register_hotmemory_notifier() to avoid a compile problem with gcc-4.4.4: When CONFIG_MEMORY_HOTPLUG=n, we don't want the memory-hotplug notifier handlers to be included in the .o files, for space reasons. The existing hotplug_memory_notifier() tries to handle this but testing with gcc-4.4.4 shows that it doesn't work - the hotplug functions are still present in the .o files. Since commit 76ae8474 ("Documentation: raise minimum supported version of GCC to 5.1") has already updated the minimum gcc version to 5.1. The previous problem mentioned in f02c6968 does not exist. So we can now revert to use hotplug_memory_notifier() directly rather than register_hotmemory_notifier(). In the last patch, we move all hotplug memory notifier priority to same file for easy sorting. This patch (of 8): Commit 76ae8474 ("Documentation: raise minimum supported version of GCC to 5.1") updated the minimum gcc version to 5.1. So the problem mentioned in f02c6968 ("include/linux/memory.h: implement register_hotmemory_notifier()") no longer exist. So we can now switch to use hotplug_memory_notifier() directly rather than register_hotmemory_notifier(). Link: https://lkml.kernel.org/r/20220923033347.3935160-1-liushixin2@huawei.com Link: https://lkml.kernel.org/r/20220923033347.3935160-2-liushixin2@huawei.comSigned-off-by:
-
Kefeng Wang authored
Since commit 2f031c6f ("mm/rmap: Convert rmap_walk() to take a folio"), page_not_mapped() takes folio as parameter, rename it to be consistent. Link: https://lkml.kernel.org/r/20220927063826.159590-1-wangkefeng.wang@huawei.comSigned-off-by:
Kefeng Wang <wangkefeng.wang@huawei.com> Reviewed-by:
-
David Hildenbrand authored
Let's trigger a R/O longterm pin on three cases of R/O mapped anonymous pages: * exclusive (never shared) * shared (child still alive) * previously shared (child no longer alive) ... and make sure that the pin is reliable: whatever we write via the page tables has to be observable via the pin. Link: https://lkml.kernel.org/r/20220927110120.106906-8-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
We want an easy way to take a R/O or R/W longterm pin on a range and be able to observe the content of the pinned pages, so we can properly test how longterm puns interact with our COW logic. [david@redhat.com: silence a warning on 32-bit] Link: https://lkml.kernel.org/r/74adbb51-6e33-f636-8a9c-2ad87bd9007e@redhat.com [yang.lee@linux.alibaba.com: ./mm/gup_test.c:281:2-3: Unneeded semicolon] Link: https://bugzilla.openanolis.cn/show_bug.cgi?id=2455 Link: https://lkml.kernel.org/r/20221020024035.113619-1-yang.lee@linux.alibaba.com Link: https://lkml.kernel.org/r/20220927110120.106906-7-david@redhat.comSigned-off-by:
Yang Li <yang.lee@linux.alibaba.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Christoph von Recklinghausen <crecklin@redhat.com> Cc: Don Dutile <ddutile@redhat.com> Cc: Jason Gunthorpe <jgg@nvidia.com> Cc: John Hubbard <jhubbard@nvidia.com> Cc: Mike Rapoport <rppt@kernel.org> Cc: Nadav Amit <namit@vmware.com> Cc: Peter Xu <peterx@redhat.com> Cc: Shuah Khan <shuah@kernel.org> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by:
-
David Hildenbrand authored
io_uring provides a simple mechanism to test long-term, R/W GUP pins -- via fixed buffers -- and can be used to verify that GUP pins stay in sync with the pages in the page table even if a page would temporarily get mapped R/O or concurrent fork() could accidentially end up sharing pinned pages with the child. Note that this essentially re-introduces local_config support that was removed recently in commit 6f83d6c7 ("Kselftests: remove support of libhugetlbfs from kselftests"). [david@redhat.com: s/size_t/ssize_t/ on `cur', `total'.] Link: https://lkml.kernel.org/r/445fe1ae-9e22-0d1d-4d09-272231d2f84a@redhat.com Link: https://lkml.kernel.org/r/20220927110120.106906-6-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
Let's run all existing test cases with all hugetlb sizes we're able to detect. Note that some tests cases still fail. This will, for example, be fixed once vmsplice properly uses FOLL_PIN instead of FOLL_GET for pinning. With 2 MiB and 1 GiB hugetlb on x86_64, the expected failures are: # [RUN] vmsplice() + unmap in child ... with hugetlb (2048 kB) not ok 23 No leak from parent into child # [RUN] vmsplice() + unmap in child ... with hugetlb (1048576 kB) not ok 24 No leak from parent into child # [RUN] vmsplice() before fork(), unmap in parent after fork() ... with hugetlb (2048 kB) not ok 35 No leak from child into parent # [RUN] vmsplice() before fork(), unmap in parent after fork() ... with hugetlb (1048576 kB) not ok 36 No leak from child into parent # [RUN] vmsplice() + unmap in parent after fork() ... with hugetlb (2048 kB) not ok 47 No leak from child into parent # [RUN] vmsplice() + unmap in parent after fork() ... with hugetlb (1048576 kB) not ok 48 No leak from child into parent Link: https://lkml.kernel.org/r/20220927110120.106906-5-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
Let's add various THP variants that we'll run with our existing test cases. Link: https://lkml.kernel.org/r/20220927110120.106906-4-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
We'll reuse it in the anon_cow test next. Link: https://lkml.kernel.org/r/20220927110120.106906-3-david@redhat.comSigned-off-by:
-
David Hildenbrand authored
Patch series "selftests/vm: test COW handling of anonymous memory". This is my current set of tests for testing COW handling of anonymous memory, especially when interacting with GUP. I developed these tests while working on PageAnonExclusive and managed to clean them up just now. On current upstream Linux, all tests pass except the hugetlb tests that rely on vmsplice -- these tests should pass as soon as vmsplice properly uses FOLL_PIN instead of FOLL_GET. I'm working on additional tests for COW handling in private mappings, focusing on long-term R/O pinning e.g., of the shared zeropage, pagecache pages and KSM pages. These tests, however, will go into a different file. So this is everything I have regarding tests for anonymous memory. This patch (of 7): Let's start adding tests for our COW handling of anonymous memory. We'll focus on basic tests that we can achieve without additional libraries or gup_test extensions. We'll add THP and hugetlb tests separately. [david@redhat.com: s/size_t/ssize_t/ on `cur', `total', `transferred';] Link: https://lkml.kernel.org/r/51302b9e-dc69-d709-3214-f23868028555@redhat.com Link: https://lkml.kernel.org/r/20220927110120.106906-1-david@redhat.com Link: https://lkml.kernel.org/r/20220927110120.106906-2-david@redhat.comSigned-off-by:
-
Andrey Konovalov authored
Migrate the workqueue_uaf test to the KUnit framework. Initially, this test was intended to check that Generic KASAN prints auxiliary stack traces for workqueues. Nevertheless, the test is enabled for all modes to make that KASAN reports bad accesses in the tested scenario. The presence of auxiliary stack traces for the Generic mode needs to be inspected manually. Link: https://lkml.kernel.org/r/1d81b6cc2a58985126283d1e0de8e663716dd930.1664298455.git.andreyknvl@google.comSigned-off-by:
Andrey Konovalov <andreyknvl@google.com> Reviewed-by:
Marco Elver <elver@google.com> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com> Cc: Dmitry Vyukov <dvyukov@google.com> Signed-off-by:
-
Andrey Konovalov authored
Migrate the kasan_rcu_uaf test to the KUnit framework. Changes to the implementation of the test: - Call rcu_barrier() after call_rcu() to make that the RCU callbacks get triggered before the test is over. - Cast pointer passed to rcu_dereference_protected as __rcu to get rid of the Sparse warning. - Check that KASAN prints a report via KUNIT_EXPECT_KASAN_FAIL. Initially, this test was intended to check that Generic KASAN prints auxiliary stack traces for RCU objects. Nevertheless, the test is enabled for all modes to make that KASAN reports bad accesses in RCU callbacks. The presence of auxiliary stack traces for the Generic mode needs to be inspected manually. Link: https://lkml.kernel.org/r/897ee08d6cd0ba7e8a4fbfd9d8502823a2f922e6.1664298455.git.andreyknvl@google.comSigned-off-by:
-
Andrey Konovalov authored
Switch KUnit-compatible KASAN tests from using per-task KUnit resources to console tracepoints. This allows for two things: 1. Migrating tests that trigger a KASAN report in the context of a task other than current to KUnit framework. This is implemented in the patches that follow. 2. Parsing and matching the contents of KASAN reports. This is not yet implemented. Link: https://lkml.kernel.org/r/9345acdd11e953b207b0ed4724ff780e63afeb36.1664298455.git.andreyknvl@google.comSigned-off-by:
-
Thomas Weißschuh authored
Without this check open() will open large files on tmpfs although O_LARGEFILE was not specified. This is inconsistent with other filesystems. Also it will later result in EOVERFLOW on stat() or EFBIG on write(). Link: https://lore.kernel.org/lkml/76bedae6-22ea-4abc-8c06-b424ceb39217@t-8ch.de/ Link: https://lkml.kernel.org/r/20220928104535.61186-1-linux@weissschuh.netSigned-off-by:
Thomas Weißschuh <thomas.weissschuh@amadeus.com> Acked-by:
Hugh Dickins <hughd@google.com> Signed-off-by:
-
Kamalesh Babulal authored
Replace the checks for memcg is root memcg, with mem_cgroup_is_root() helper. Link: https://lkml.kernel.org/r/20220930134433.338103-1-kamalesh.babulal@oracle.comSigned-off-by:
Kamalesh Babulal <kamalesh.babulal@oracle.com> Reviewed-by:
Muchun Song <songmuchun@bytedance.com> Acked-by:
Michal Hocko <mhocko@suse.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Kamalesh Babulal <kamalesh.babulal@oracle.com> Cc: Roman Gushchin <roman.gushchin@linux.dev> Cc: Shakeel Butt <shakeelb@google.com> Cc: Tom Hromatka <tom.hromatka@oracle.com> Signed-off-by:
-
Deming Wang authored
Using vma_lookup() verifies the start address is contained in the found vma. This results in easier to read the code. Link: https://lkml.kernel.org/r/20221007030345.5029-1-wangdeming@inspur.comSigned-off-by:
Deming Wang <wangdeming@inspur.com> Signed-off-by:
-
Lukas Bulwahn authored
After the rework of shmem_get_folio_gfp() to use a folio, the local variable hindex is only needed to be set once before passing it to shmem_add_to_page_cache(). Remove the unneeded initialization and assignments of the variable hindex before the actual effective assignment and first use. No functional change. No change in object code. Link: https://lkml.kernel.org/r/20221007085027.6309-1-lukas.bulwahn@gmail.comSigned-off-by:
Lukas Bulwahn <lukas.bulwahn@gmail.com> Cc: Hugh Dickins <hughd@google.com> Cc: Matthew Wilcox <willy@infradead.org> Signed-off-by:
-
Rolf Eike Beer authored
There is no eprotect(), so I assume this is about mprotect(). Link: https://lkml.kernel.org/r/2385684.8vm7BOzihM@mobilepool36.emlix.comSigned-off-by:
Rolf Eike Beer <eb@emlix.com> Signed-off-by:
-
Uros Bizjak authored
Use try_cmpxchg instead of cmpxchg (*ptr, old, new) == old in update_used_max. x86 CMPXCHG instruction returns success in ZF flag, so this change saves a compare after cmpxchg (and related move instruction in front of cmpxchg). Also, reorder code a bit to remove additional compare and conditional jump from the assembly code. Together, hese two changes save 15 bytes from the function when compiled for x86_64. No functional change intended. Link: https://lkml.kernel.org/r/20221018145154.3699-1-ubizjak@gmail.comSigned-off-by:
Uros Bizjak <ubizjak@gmail.com> Reviewed-by:
Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Signed-off-by:
-
Vishal Moola (Oracle) authored
Initially, find_get_entries() was being passed in the start offset as a value. That left the calculation of the offset to the callers. This led to complexity in the callers trying to keep track of the index. Now find_get_entries() takes in a pointer to the start offset and updates the value to be directly after the last entry found. If no entry is found, the offset is not changed. This gets rid of multiple hacky calculations that kept track of the start offset. Link: https://lkml.kernel.org/r/20221017161800.2003-3-vishal.moola@gmail.comSigned-off-by:
Vishal Moola (Oracle) <vishal.moola@gmail.com> Cc: Hugh Dickins <hughd@google.com> Cc: Matthew Wilcox <willy@infradead.org> Signed-off-by:
-
Vishal Moola (Oracle) authored
Patch series "Rework find_get_entries() and find_lock_entries()", v3. Originally the callers of find_get_entries() and find_lock_entries() were keeping track of the start index themselves as they traverse the search range. This resulted in hacky code such as in shmem_undo_range(): index = folio->index + folio_nr_pages(folio) - 1; where the - 1 is only present to stay in the right spot after incrementing index later. This sort of calculation was also being done on every folio despite not even using index later within that function. These patches change find_get_entries() and find_lock_entries() to calculate the new index instead of leaving it to the callers so we can avoid all these complications. This patch (of 2): Initially, find_lock_entries() was being passed in the start offset as a value. That left the calculation of the offset to the callers. This led to complexity in the callers trying to keep track of the index. Now find_lock_entries() takes in a pointer to the start offset and updates the value to be directly after the last entry found. If no entry is found, the offset is not changed. This gets rid of multiple hacky calculations that kept track of the start offset. Link: https://lkml.kernel.org/r/20221017161800.2003-1-vishal.moola@gmail.com Link: https://lkml.kernel.org/r/20221017161800.2003-2-vishal.moola@gmail.comSigned-off-by:
-
Ma Wupeng authored
Commit 2555283e ("mm/rmap: Fix anon_vma->degree ambiguity leading to double-reuse") use num_children and num_active_vmas to replace the origin degree to fix anon_vma UAF problem. Update the comment in anon_vma_clone to fit this change. Link: https://lkml.kernel.org/r/20221014013931.1565969-1-mawupeng1@huawei.comSigned-off-by:
Ma Wupeng <mawupeng1@huawei.com> Signed-off-by:
-
Sidhartha Kumar authored
Helper function to retrieve hstate information from a hugetlb folio. Link: https://lkml.kernel.org/r/20220922154207.1575343-6-sidhartha.kumar@oracle.comSigned-off-by:
Sidhartha Kumar <sidhartha.kumar@oracle.com> Reported-by:
kernel test robot <lkp@intel.com> Reviewed-by:
Mike Kravetz <mike.kravetz@oracle.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Colin Cross <ccross@google.com> Cc: David Howells <dhowells@redhat.com> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Hugh Dickins <hughd@google.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Muchun Song <songmuchun@bytedance.com> Cc: Peter Xu <peterx@redhat.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: William Kucharski <william.kucharski@oracle.com> Signed-off-by:
-
Sidhartha Kumar authored
Remove the last caller of delete_from_page_cache() by converting the code to its folio equivalent. Link: https://lkml.kernel.org/r/20220922154207.1575343-5-sidhartha.kumar@oracle.comSigned-off-by:
-
Sidhartha Kumar authored
Allow hugetlbfs_migrate_folio to check and read subpool information by passing in a folio. Link: https://lkml.kernel.org/r/20220922154207.1575343-4-sidhartha.kumar@oracle.comSigned-off-by:
-
Sidhartha Kumar authored
Allow struct folio to store hugetlb metadata that is contained in the private field of the first tail page. On 32-bit, _private_1 aligns with page[1].private. Link: https://lkml.kernel.org/r/20220922154207.1575343-3-sidhartha.kumar@oracle.comSigned-off-by:
-
Sidhartha Kumar authored
Patch series "begin converting hugetlb code to folios", v4. This patch series starts the conversion of the hugetlb code to operate on struct folios rather than struct pages. This removes the ambiguitiy of whether functions are operating on head pages, tail pages of compound pages, or base pages. This series passes the linux test project hugetlb test cases. Patch 1 adds hugeltb specific page macros that can operate on folios. Patch 2 adds the private field of the first tail page to struct page. For 32-bit, _private_1 alinging with page[1].private was confirmed by using pahole. Patch 3 introduces hugetlb subpool helper functions which operate on struct folios. These patches were tested using the hugepage-mmap.c selftest along with the migratepages command. Patch 4 converts hugetlb_delete_from_page_cache() to use folios. Patch 5 adds a folio_hstate() function to get hstate information from a folio and adds a user of folio_hstate(). Bpftrace was used to track time spent in the free_huge_pages function during the ltp test cases as it is a caller of the hugetlb subpool functions. From the histogram, the performance is similar before and after the patch series. Time spent in 'free_huge_page' 6.0.0-rc2.master.20220823 @nsecs: [256, 512) 14770 |@@@@@@@@@@@@@@@@@@@@@@@@@@@ |@@@@@@@@@@@@@@@@@@@@@@@@@ | [512, 1K) 155 | | [1K, 2K) 169 | | [2K, 4K) 50 | | [4K, 8K) 14 | | [8K, 16K) 3 | | [16K, 32K) 3 | | 6.0.0-rc2.master.20220823 + patch series @nsecs: [256, 512) 13678 |@@@@@@@@@@@@@@@@@@@@@@@@@@@ | |@@@@@@@@@@@@@@@@@@@@@@@@@ | [512, 1K) 142 | | [1K, 2K) 199 | | [2K, 4K) 44 | | [4K, 8K) 13 | | [8K, 16K) 4 | | [16K, 32K) 1 | | This patch (of 5): Allow the macros which test, set, and clear hugetlb specific page flags to take a hugetlb folio as an input. The macrros are generated as folio_{test, set, clear}_hugetlb_{restore_reserve, migratable, temporary, freed, vmemmap_optimized, raw_hwp_unreliable}. Link: https://lkml.kernel.org/r/20220922154207.1575343-1-sidhartha.kumar@oracle.com Link: https://lkml.kernel.org/r/20220922154207.1575343-2-sidhartha.kumar@oracle.comSigned-off-by: -
Peter Xu authored
After converting all the three relevant testcases (uffd, madvise, mremap) to use memfd, no test will need the hugetlb mount point anymore. Drop the code. Link: https://lkml.kernel.org/r/20221014144015.94039-1-peterx@redhat.comSigned-off-by:
Peter Xu <peterx@redhat.com> Reviewed-by:
Axel Rasmussen <axelrasmussen@google.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Signed-off-by:
-
Peter Xu authored
For dropping the hugetlb mountpoint in run_vmtests.sh. Cleaned it up a little bit around the changed codes. Link: https://lkml.kernel.org/r/20221014144013.94027-1-peterx@redhat.comSigned-off-by:
-
Peter Xu authored
For dropping the hugetlb mountpoint in run_vmtests.sh. Since no parameter is needed, drop USAGE too. Link: https://lkml.kernel.org/r/20221014143921.93887-3-peterx@redhat.comSigned-off-by:
-
Peter Xu authored
Patch series "selftests/vm: Drop hugetlb mntpoint in run_vmtests.sh", v2. Clean the code up so we can use the same memfd for both hugetlb and shmem which is cleaner. This patch (of 4): We already used memfd for shmem test, move it forward with hugetlb too so that we don't need user to specify the hugetlb file path explicitly when running hugetlb shared tests. Link: https://lkml.kernel.org/r/20221014143921.93887-1-peterx@redhat.com Link: https://lkml.kernel.org/r/20221014143921.93887-2-peterx@redhat.comSigned-off-by:
-
Johannes Weiner authored
We noticed a 2% webserver throughput regression after upgrading from 5.6. This could be tracked down to a shift in the anon/file reclaim balance (confirmed with swappiness) that resulted in worse reclaim efficiency and thus more kswapd activity for the same outcome. The change that exposed the problem is aae466b0 ("mm/swap: implement workingset detection for anonymous LRU"). By qualifying swapins based on their refault distance, it lowered the cost of anon reclaim in this workload, in turn causing (much) more anon scanning than before. Scanning the anon list is more expensive due to the higher ratio of mmapped pages that may rotate during reclaim, and so the result was an increase in %sys time. Right now, rotations aren't considered a cost when balancing scan pressure between LRUs. We can end up with very few file refaults putting all the scan pressure on hot anon pages that are rotated en masse, don't get reclaimed, and never push back on the file LRU again. We still only reclaim file cache in that case, but we burn a lot CPU rotating anon pages. It's "fair" from an LRU age POV, but doesn't reflect the real cost it imposes on the system. Consider rotations as a secondary factor in balancing the LRUs. This doesn't attempt to make a precise comparison between IO cost and CPU cost, it just says: if reloads are about comparable between the lists, or rotations are overwhelmingly different, adjust for CPU work. This fixed the regression on our webservers. It has since been deployed to the entire Meta fleet and hasn't caused any problems. Link: https://lkml.kernel.org/r/20221013193113.726425-1-hannes@cmpxchg.orgSigned-off-by:
Johannes Weiner <hannes@cmpxchg.org> Cc: Rik van Riel <riel@surriel.com> Signed-off-by:
-
Mike Kravetz authored
During discussions of this series [1], it was suggested that hugetlb handling code in follow_page_mask could be simplified. At the beginning of follow_page_mask, there currently is a call to follow_huge_addr which 'may' handle hugetlb pages. ia64 is the only architecture which provides a follow_huge_addr routine that does not return error. Instead, at each level of the page table a check is made for a hugetlb entry. If a hugetlb entry is found, a call to a routine associated with that entry is made. Currently, there are two checks for hugetlb entries at each page table level. The first check is of the form: if (p?d_huge()) page = follow_huge_p?d(); the second check is of the form: if (is_hugepd()) page = follow_huge_pd(). We can replace these checks, as well as the special handling routines such as follow_huge_p?d() and follow_huge_pd() with a single routine to handle hugetlb vmas. A new routine hugetlb_follow_page_mask is called for hugetlb vmas at the beginning of follow_page_mask. hugetlb_follow_page_mask will use the existing routine huge_pte_offset to walk page tables looking for hugetlb entries. huge_pte_offset can be overwritten by architectures, and already handles special cases such as hugepd entries. [1] https://lore.kernel.org/linux-mm/cover.1661240170.git.baolin.wang@linux.alibaba.com/ [mike.kravetz@oracle.com: remove vma (pmd sharing) per Peter] Link: https://lkml.kernel.org/r/20221028181108.119432-1-mike.kravetz@oracle.com [mike.kravetz@oracle.com: remove left over hugetlb_vma_unlock_read()] Link: https://lkml.kernel.org/r/20221030225825.40872-1-mike.kravetz@oracle.com Link: https://lkml.kernel.org/r/20220919021348.22151-1-mike.kravetz@oracle.comSigned-off-by:Baolin Wang <baolin.wang@linux.alibaba.com> Tested-by:
-
- 06 Nov, 2022 1 commit
-
-
Linus Torvalds authored
-
