On Tue, 11 Feb 2025 10:32:55 +0100 "Arnd Bergmann" arnd@arndb.de wrote:
On Tue, Feb 11, 2025, at 10:23, Naresh Kamboju wrote:
Regressions on arm64 Juno-r2 device while running LTP syscalls tests madvise01 warnings on the Linux next-20250210 tag.
First seen on next-20250210 Good: next-20250207 Bad: next-20250210
Cc SeongJae Park sj@kernel.org
[...]
This is mmap_write_unlock() triggering rwsem_assert_held_write(&mm->mmap_lock);
That code was changed in these commits:
415553982a1e mm-madvise-remove-redundant-mmap_lock-operations-from-process_madvise-fix ec68fbd9e99f mm/madvise: remove redundant mmap_lock operations from process_madvise() f19c9d7b57cf mm/madvise: split out madvise() behavior execution a4d5a73cdc14 mm/madvise: split out madvise input validity check 948a0a9ea070 mm/madvise: split out mmap locking operations for madvise()
You could try reverting them to see if that fixes it again.
Thank you for kindly Cc-ing me and further providing this nice information, Arnd! I posted a fix and continuing discussion there: https://lore.kernel.org/20250211063201.5106-1-sj@kernel.org
Thanks, SJ
[...]