zsmalloc: fix races between modifications of fullness and isolated
authorAndrew Yang <andrew.yang@mediatek.com>
Fri, 21 Jul 2023 06:37:01 +0000 (14:37 +0800)
committerAndrew Morton <akpm@linux-foundation.org>
Fri, 4 Aug 2023 20:03:40 +0000 (13:03 -0700)
commit4b5d1e47b69426c0f7491d97d73ad0152d02d437
tree0390b7e00167153964fac16bf16d713894ee5214
parent5d0c230f1de8c7515b6567d9afba1f196fb4e2f4
zsmalloc: fix races between modifications of fullness and isolated

We encountered many kernel exceptions of VM_BUG_ON(zspage->isolated ==
0) in dec_zspage_isolation() and BUG_ON(!pages[1]) in zs_unmap_object()
lately.  This issue only occurs when migration and reclamation occur at
the same time.

With our memory stress test, we can reproduce this issue several times
a day.  We have no idea why no one else encountered this issue.  BTW,
we switched to the new kernel version with this defect a few months
ago.

Since fullness and isolated share the same unsigned int, modifications of
them should be protected by the same lock.

[andrew.yang@mediatek.com: move comment]
Link: https://lkml.kernel.org/r/20230727062910.6337-1-andrew.yang@mediatek.com
Link: https://lkml.kernel.org/r/20230721063705.11455-1-andrew.yang@mediatek.com
Fixes: c4549b871102 ("zsmalloc: remove zspage isolation for migration")
Signed-off-by: Andrew Yang <andrew.yang@mediatek.com>
Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Cc: Matthias Brugger <matthias.bgg@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/zsmalloc.c