On Sat, Jul 17, 2021 at 10:17:13AM -0700, Darrick J. Wong wrote:
> I experience the same problem that Murphy does, and I tracked it down
> to this chunk of inode_do_switch_wbs:
>
> /*
> * Count and transfer stats. Note that PAGECACHE_TAG_DIRTY points
> * to possibly dirty pages while PAGECACHE_TAG_WRITEBACK points to
> * pages actually under writeback.
> */
> xas_for_each_marked(&xas, page, ULONG_MAX, PAGECACHE_TAG_DIRTY) {
> here >>>>>>>>>> if (PageDirty(page)) {
> dec_wb_stat(old_wb, WB_RECLAIMABLE);
> inc_wb_stat(new_wb, WB_RECLAIMABLE);
> }
> }
>
> I suspect that "page" is really a pfn to a pmem mapping and not a real
> struct page.
I think you're right.
Running scripts/decodecode on the original report, that's:
0: 48 8b 50 08 mov 0x8(%rax),%rdx
RAX: 0000000005b0f661
so rax is not a struct page, it's a PFN (/DAX) entry.
We shouldn't even be calling inode_do_switch_wbs() for DAX inodes because
we don't do writeback for DAX inodes (at least as far as the kernel's
writeback infrastructure is concerned; obviously the CPU does writeback
from its caches to PMEM).
Maybe some check at a higher level would be appropriate? I don't know
much about this part of the kernel.
