On Tue, Feb 25, 2025 at 01:32:00PM -0800, Nhat Pham wrote: > Currently, we crash the kernel when a decompression failure occurs in > zswap (either because of memory corruption, or a bug in the compression > algorithm). This is overkill. We should only SIGBUS the unfortunate > process asking for the zswap entry on zswap load, and skip the corrupted > entry in zswap writeback. Some relevant observations/questions, but not really actionable for this patch, perhaps some future work, or more likely some incoherent illogical thoughts : (1) It seems like not making the folio uptodate will cause shmem faults to mark the swap entry as hwpoisoned, but I don't see similar handling for do_swap_page(). So it seems like even if we SIGBUS the process, other processes mapping the same page could follow in the same footsteps. (2) A hwpoisoned swap entry results in VM_FAULT_SIGBUS in some cases (e.g. shmem_fault() -> shmem_get_folio_gfp() -> shmem_swapin_folio()), even though we have VM_FAULT_HWPOISON. This patch falls under this bucket, but unfortunately we cannot tell for sure if it's a hwpoision or a decompression bug. (3) If we run into a decompression failure, should we free the underlying memory from zsmalloc? I don't know. On one hand if we free it zsmalloc may start using it for more compressed objects. OTOH, I don't think proper hwpoison handling will kick in until the page is freed. Maybe we should tell zsmalloc to drop this page entirely and mark objects within it as invalid? Probably not worth the hassle but something to think about.
