kernel/linux.git/tools/testing/selftests/mm, branch v6.18.21

selftests/mm/charge_reserved_hugetlb: drop mount size for hugetlbfs

2026-03-04T12:21:10+00:00

[ Upstream commit 1aa1dd9cc595917882fb6db67725442956f79607 ] charge_reserved_hugetlb.sh mounts a hugetlbfs instance at /mnt/huge with a fixed size of 256M. On systems with large base hugepages (e.g. 512MB), this is smaller than a single hugepage, so the hugetlbfs mount ends up with zero capacity (often visible as size=0 in mount output). As a result, write_to_hugetlbfs fails with ENOMEM and the test can hang waiting for progress. === Error log === # uname -r 6.12.0-xxx.el10.aarch64+64k #./charge_reserved_hugetlb.sh -cgroup-v2 # ----------------------------------------- ... # nr hugepages = 10 # writing cgroup limit: 5368709120 # writing reseravation limit: 5368709120 ... # write_to_hugetlbfs: Error mapping the file: Cannot allocate memory # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 # Waiting for hugetlb memory reservation to reach size 2684354560. # 0 ... # mount |grep /mnt/huge none on /mnt/huge type hugetlbfs (rw,relatime,seclabel,pagesize=512M,size=0) # grep -i huge /proc/meminfo ... HugePages_Total: 10 HugePages_Free: 10 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 524288 kB Hugetlb: 5242880 kB Drop the mount args with 'size=256M', so the filesystem capacity is sufficient regardless of HugeTLB page size. Link: https://lkml.kernel.org/r/20251221122639.3168038-3-liwang@redhat.com Fixes: 29750f71a9b4 ("hugetlb_cgroup: add hugetlb_cgroup reservation tests") Signed-off-by: Li Wang Acked-by: David Hildenbrand (Red Hat) Acked-by: Waiman Long Cc: Mark Brown Cc: Shuah Khan Cc: Signed-off-by: Andrew Morton Signed-off-by: Sasha Levin

selftests/mm: fix usage of FORCE_READ() in cow tests

2026-02-26T22:59:19+00:00

[ Upstream commit bce1dabd310e87fefe0645fec9ba98b84d37e418 ] Commit 5bbc2b785e63 ("selftests/mm: fix FORCE_READ to read input value correctly") modified FORCE_READ() to take a value instead of a pointer. It also changed most of the call sites accordingly, but missed many of them in cow.c. In those cases, we ended up with the pointer itself being read, not the memory it points to. No failure occurred as a result, so it looks like the tests work just fine without faulting in. However, the huge_zeropage tests explicitly check that pages are populated, so those became skipped. Convert all the remaining FORCE_READ() to fault in the mapped page, as was originally intended. This allows the huge_zeropage tests to run again (3 tests in total). Link: https://lkml.kernel.org/r/20260122170224.4056513-5-kevin.brodsky@arm.com Fixes: 5bbc2b785e63 ("selftests/mm: fix FORCE_READ to read input value correctly") Signed-off-by: Kevin Brodsky Acked-by: SeongJae Park Reviewed-by: wang lian Acked-by: David Hildenbrand (Red Hat) Reviewed-by: Dev Jain Cc: Jason Gunthorpe Cc: John Hubbard Cc: Lorenzo Stoakes Cc: Mark Brown Cc: Paolo Abeni Cc: Ryan Roberts Cc: Shuah Khan Cc: Usama Anjum Cc: Yunsheng Lin Signed-off-by: Andrew Morton Signed-off-by: Sasha Levin

tools/testing/selftests: fix gup_longterm for unknown fs

2026-01-23T10:21:25+00:00

commit 21c68ad1d9771d331198cc73cbf6e908d7915f35 upstream. Commit 66bce7afbaca ("selftests/mm: fix test result reporting in gup_longterm") introduced a small bug causing unknown filesystems to always result in a test failure. This is because do_test() was updated to use a common reporting path, but this case appears to have been missed. This is problematic for e.g. virtme-ng which uses an overlayfs file system, causing gup_longterm to appear to fail each time due to a test count mismatch: # Planned tests != run tests (50 != 46) # Totals: pass:24 fail:0 xfail:0 xpass:0 skip:22 error:0 The fix is to simply change the return into a break. Link: https://lkml.kernel.org/r/20260106154547.214907-1-lorenzo.stoakes@oracle.com Fixes: 66bce7afbaca ("selftests/mm: fix test result reporting in gup_longterm") Signed-off-by: Lorenzo Stoakes Reviewed-by: David Hildenbrand (Red Hat) Cc: Jason Gunthorpe Cc: John Hubbard Cc: Liam Howlett Cc: "Liam R. Howlett" Cc: Mark Brown Cc: Michal Hocko Cc: Mike Rapoport Cc: Peter Xu Cc: Shuah Khan Cc: Suren Baghdasaryan Cc: Vlastimil Babka Cc: Signed-off-by: Andrew Morton Signed-off-by: Greg Kroah-Hartman

tools/testing/selftests: add forked (un)/faulted VMA merge tests

2026-01-23T10:21:24+00:00

commit fb39444732f02c32a8312c168d97e33d872c14d3 upstream. Now we correctly handle forked faulted/unfaulted merge on mremap(), exhaustively assert that we handle this correctly. Do this in the less duplicative way by adding a new merge_with_fork fixture and forked/unforked variants, and abstract the forking logic as necessary to avoid code duplication with this also. Link: https://lkml.kernel.org/r/1daf76d89fdb9d96f38a6a0152d8f3c2e9e30ac7.1767638272.git.lorenzo.stoakes@oracle.com Fixes: 879bca0a2c4f ("mm/vma: fix incorrectly disallowed anonymous VMA merges") Signed-off-by: Lorenzo Stoakes Cc: David Hildenbrand (Red Hat) Cc: Jann Horn Cc: Jeongjun Park Cc: Liam Howlett Cc: Pedro Falcato Cc: Rik van Riel Cc: Vlastimil Babka Cc: Yeoreum Yun Cc: Harry Yoo Cc: Signed-off-by: Andrew Morton Signed-off-by: Greg Kroah-Hartman

tools/testing/selftests: add tests for !tgt, src mremap() merges

2026-01-23T10:21:24+00:00

commit 0ace8f2db6b3b4b0677e559d1a7ab7fd625d61ec upstream. Test that mremap()'ing a VMA into a position such that the target VMA on merge is unfaulted and the source faulted is correctly performed. We cover 4 cases: 1. Previous VMA unfaulted: copied -----| v |-----------|.............| | unfaulted |(faulted VMA)| |-----------|.............| prev target = prev, expand prev to cover. 2. Next VMA unfaulted: copied -----| v |.............|-----------| |(faulted VMA)| unfaulted | |.............|-----------| next target = next, expand next to cover. 3. Both adjacent VMAs unfaulted: copied -----| v |-----------|.............|-----------| | unfaulted |(faulted VMA)| unfaulted | |-----------|.............|-----------| prev next target = prev, expand prev to cover. 4. prev unfaulted, next faulted: copied -----| v |-----------|.............|-----------| | unfaulted |(faulted VMA)| faulted | |-----------|.............|-----------| prev next target = prev, expand prev to cover. Essentially equivalent to 3, but with additional requirement that next's anon_vma is the same as the copied VMA's. Each of these are performed with MREMAP_DONTUNMAP set, which will cause a KASAN assert for UAF or an assert on zero refcount anon_vma if a bug exists with correctly propagating anon_vma state in each scenario. Link: https://lkml.kernel.org/r/f903af2930c7c2c6e0948c886b58d0f42d8e8ba3.1767638272.git.lorenzo.stoakes@oracle.com Fixes: 879bca0a2c4f ("mm/vma: fix incorrectly disallowed anonymous VMA merges") Signed-off-by: Lorenzo Stoakes Cc: David Hildenbrand (Red Hat) Cc: Jann Horn Cc: Jeongjun Park Cc: Liam Howlett Cc: Pedro Falcato Cc: Rik van Riel Cc: Vlastimil Babka Cc: Yeoreum Yun Cc: Harry Yoo Cc: Signed-off-by: Andrew Morton Signed-off-by: Greg Kroah-Hartman

selftests/mm: fix thread state check in uffd-unit-tests

2026-01-08T09:17:11+00:00

commit 632b874d59a36caf829ab5790dafb90f9b350fd6 upstream. In the thread_state_get() function, the logic to find the thread's state character was using `sizeof(header) - 1` to calculate the offset from the "State:\t" string. The `header` variable is a `const char *` pointer. `sizeof()` on a pointer returns the size of the pointer itself, not the length of the string literal it points to. This makes the code's behavior dependent on the architecture's pointer size. This bug was identified on a 32-bit ARM build (`gsi_tv_arm`) for Android, running on an ARMv8-based device, compiled with Clang 19.0.1. On this 32-bit architecture, `sizeof(char *)` is 4. The expression `sizeof(header) - 1` resulted in an incorrect offset of 3, causing the test to read the wrong character from `/proc/[tid]/status` and fail. On 64-bit architectures, `sizeof(char *)` is 8, so the expression coincidentally evaluates to 7, which matches the length of "State:\t". This is why the bug likely remained hidden on 64-bit builds. To fix this and make the code portable and correct across all architectures, this patch replaces `sizeof(header) - 1` with `strlen(header)`. The `strlen()` function correctly calculates the string's length, ensuring the correct offset is always used. Link: https://lkml.kernel.org/r/20251210091408.3781445-1-wakel@google.com Fixes: f60b6634cd88 ("mm/selftests: add a test to verify mmap_changing race with -EAGAIN") Signed-off-by: Wake Liu Acked-by: Peter Xu Reviewed-by: Mike Rapoport (Microsoft) Cc: Bill Wendling Cc: Justin Stitt Cc: Liam Howlett Cc: Lorenzo Stoakes Cc: Michal Hocko Cc: Nathan Chancellor Cc: Shuah Khan Cc: Suren Baghdasaryan Cc: Vlastimil Babka Cc: Signed-off-by: Andrew Morton Signed-off-by: Greg Kroah-Hartman

selftests/mm: fix division-by-zero in uffd-unit-tests

2025-11-24T22:25:17+00:00

Commit 4dfd4bba8578 ("selftests/mm/uffd: refactor non-composite global vars into struct") moved some of the operations previously implemented in uffd_setup_environment() earlier in the main test loop. The calculation of nr_pages, which involves a division by page_size, now occurs before checking that default_huge_page_size() returns a non-zero This leads to a division-by-zero error on systems with !CONFIG_HUGETLB. Fix this by relocating the non-zero page_size check before the nr_pages calculation, as it was originally implemented. Link: https://lkml.kernel.org/r/20251113034623.3127012-1-cmllamas@google.com Fixes: 4dfd4bba8578 ("selftests/mm/uffd: refactor non-composite global vars into struct") Signed-off-by: Carlos Llamas Acked-by: David Hildenbrand (Red Hat) Reviewed-by: Lorenzo Stoakes Reviewed-by: Mike Rapoport (Microsoft) Reviewed-by: Ujwal Kundur Cc: Brendan Jackman Cc: Liam Howlett Cc: Michal Hocko Cc: Peter Xu Cc: Shuah Khan Cc: Suren Baghdasaryan Cc: Vlastimil Babka Cc: Signed-off-by: Andrew Morton

selftests/mm: add fork inheritance test for ksm_merging_pages counter

2025-09-28T18:51:32+00:00

Add a new selftest to verify whether the `ksm_merging_pages` counter in `mm_struct` is not inherited by a child process after fork. This helps ensure correctness of KSM accounting across process creation. Link: https://lkml.kernel.org/r/e7bb17d374133bd31a3e423aa9e46e1122e74971.1758648700.git.donettom@linux.ibm.com Signed-off-by: Donet Tom Acked-by: David Hildenbrand Cc: Aboorva Devarajan Cc: Chengming Zhou Cc: "Ritesh Harjani (IBM)" Cc: Wei Yang Cc: xu xin Signed-off-by: Andrew Morton

selftests/mm: skip soft-dirty tests when CONFIG_MEM_SOFT_DIRTY is disabled

2025-09-23T21:14:16+00:00

The madv_populate and soft-dirty kselftests currently fail on systems where CONFIG_MEM_SOFT_DIRTY is disabled. Introduce a new helper softdirty_supported() into vm_util.c/h to ensure tests are properly skipped when the feature is not enabled. Link: https://lkml.kernel.org/r/20250917133137.62802-1-lance.yang@linux.dev Fixes: 9f3265db6ae8 ("selftests: vm: add test for Soft-Dirty PTE bit") Signed-off-by: Lance Yang Acked-by: David Hildenbrand Suggested-by: David Hildenbrand Cc: Lorenzo Stoakes Cc: Shuah Khan Cc: Gabriel Krisman Bertazi Cc: Signed-off-by: Andrew Morton

selftests/mm: protection_keys: fix dead code

2025-09-21T21:22:34+00:00

The while loop doesn't execute and following warning gets generated: protection_keys.c:561:15: warning: code will never be executed [-Wunreachable-code] int rpkey = alloc_random_pkey(); Let's enable the while loop such that it gets executed nr_iterations times. Simplify the code a bit as well. Link: https://lkml.kernel.org/r/20250912123025.1271051-3-usama.anjum@collabora.com Signed-off-by: Muhammad Usama Anjum Reviewed-by: Sidhartha Kumar Reviewed-by: Zi Yan Cc: Baolin Wang Cc: Barry Song Cc: David Hildenbrand Cc: Dev Jain Cc: Jason Gunthorpe Cc: Kevin Brodsky Cc: Lance Yang Cc: Leon Romanovsky Cc: Liam Howlett Cc: Lorenzo Stoakes Cc: Mariano Pache Cc: Michal Hocko Cc: Mike Rapoport Cc: Ryan Roberts Cc: Shuah Khan Cc: Suren Baghdasaryan Cc: Vlastimil Babka Signed-off-by: Andrew Morton