On Wed, Nov 19, 2025 at 01:27:08PM +0000, Pedro Falcato wrote:
On Wed, Nov 19, 2025 at 01:16:36PM +0000, Mark Brown wrote:
On Fri, Nov 07, 2025 at 04:11:48PM +0000, Lorenzo Stoakes wrote:
It is useful to be able to designate that certain flags are 'sticky', that is, if two VMAs are merged one with a flag of this nature and one without, the merged VMA sets this flag.
I'm seeing regressions on multiple arm64 platforms in at least the LTP clone302 and madvise10 selftests, both of which have bisected to one of the fixups to this patch. Especially given the other tests that also bisected to the same place I've not investigated further. There's a number of other LTP tests that started failing today including relevant seeming ones munlockall01, mprotect04, madvise10, mprotect03 and futex_cmp_requeue01 but I don't have bisects to confirm they're the same thing.
Thanks for the reports!
clone302:
tst_buffers.c:57: TINFO: Test is using guarded buffers tst_tmpdir.c:316: TINFO: Using /tmp/LTP_clorMwMMw as tmpdir (nfs filesystem) tst_test.c:1953: TINFO: LTP version: 20250530 tst_test.c:1956: TINFO: Tested kernel: 6.18.0-rc6-next-20251119 #1 SMP PREEMPT @1763523415 aarch64
next-20251119 still has the v3 version of the patchset, which is Known Bad(tm) after a couple of buggy fixups. v4 should hopefully work properly.
-- Pedro
Thanks for the reports Mark and also for following up Pedro, and yeah this should now be fully resolved in v4, see [0].
The fixups were just broken but luckily Pedro noticed the mistake I'd made and so was able to get a fix out quickly! :)
[0]:https://lore.kernel.org/all/cover.1763460113.git.lorenzo.stoakes@oracle.com/
Cheers, Lorenzo