On Tue, 21 Feb 2023, David Howells wrote: > David Howells <dhowells@xxxxxxxxxx> wrote: > > > > > + /* At this point we hold neither the i_pages lock nor the > > > > + * page lock: the page may be truncated or invalidated > > > > + * (changing page->mapping to NULL), or even swizzled > > > > + * back from swapper_space to tmpfs file mapping > > > > > > Where does this comment come from? This is cifs, not tmpfs. You'll > > > never be asked to writeback a page from the swap cache. Dirty pages > > > can be truncated, so the first half of the comment is still accurate. > > > I'd rather it moved down to below the folio lock, and was rephrased > > > so it described why we're checking everything again. > > > > Actually, it's in v6.2 cifs and I just move it in the patch where I copy the > > afs writepages implementation into cifs. afs got it in 2007 when I added > > write support[1] and I suspect I copied it from cifs. cifs got it in 2005 > > when Steve added writepages support[2]. I think he must've got it from > > fs/mpage.c and the comment there is prehistoric. > > The ultimate source is Hugh Dickins, it would seem: > > commit 820ef9df32856bb54fe5bc995153feb276420e15 > Author: Andrew Morton <akpm@xxxxxxxxx> > Date: Fri Nov 15 18:52:38 2002 -0800 > > [PATCH] handle pages which alter their ->mapping > > Patch from Hugh Dickins <hugh@xxxxxxxxxxx> > > tmpfs failed fsx+swapout tests after many hours, a page found zeroed. > Not a truncate problem, but mirror image of earlier truncate problems: > swap goes through mpage_writepages, which must therefore allow for a > sudden swizzle back to file identity. > > Second time this caught us, so I've audited the tree for other places > which might be surprised by such swizzling. The only others I found > were (perhaps) in the parisc and sparc64 flush_dcache_page called > from do_generic_mapping_read on a looped tmpfs file which is also > mmapped; but that's a very marginal case, I wanted to understand it > better before making any edit, and now realize that hch's sendfile > in loop eliminates it (now go through do_shmem_file_read instead: > similar but crucially this locks the page when raising its count, > which is enough to keep vmscan from interfering). > > Maybe we should delete or amend the comment now? Yes, that comment does not belong in afs or btrfs or cifs - though it does explain why we have sometimes chosen to compare folio_mapping(folio) with expected mapping, rather than against NULL. But "now" is not the moment to amend it: it looks like these sources are in flux at present. And truncate_cleanup_folio() has a "swizzles" comment without even a mapping to compare with. Hugh
