[Linux-cachefs] [PATCH 5/5] ceph: fold ceph_update_writeable_page into ceph_write_begin
Jeff Layton
jlayton at kernel.org
Fri Jun 11 14:52:55 UTC 2021
On Fri, 2021-06-11 at 10:14 -0400, Andrew W Elble wrote:
> We're seeing file corruption while running 5.10, bisected to 1cc1699070bd:
>
> > > +static int ceph_write_begin(struct file *file, struct address_space *mapping,
> > > + loff_t pos, unsigned len, unsigned flags,
> > > + struct page **pagep, void **fsdata)
>
> <snip>
>
> > > + /*
> > > + * In some cases we don't need to read at all:
> > > + * - full page write
> > > + * - write that lies completely beyond EOF
> > > + * - write that covers the the page from start to EOF or beyond it
> > > + */
> > > + if ((pos_in_page == 0 && len == PAGE_SIZE) ||
> > > + (pos >= i_size_read(inode)) ||
>
> Shouldn't this be '((pos & PAGE_MASK) >= i_size_read(inode)) ||' ?
>
> Seems like fs/netfs/read_helper.c currently has the same issue?
>
Yeah...I think you may be right. Have you tried a patch that does that
and does it fix the issue?
Also, do you happen to have a testcase that we could stick in xfstests
for this? If not, we can probably write one, but if you already have one
that'd be great.
> > > + (pos_in_page == 0 && (pos + len) >= i_size_read(inode))) {
> > > + zero_user_segments(page, 0, pos_in_page,
> > > + pos_in_page + len, PAGE_SIZE);
> > > + break;
> > > + }
>
--
Jeff Layton <jlayton at kernel.org>
More information about the Linux-cachefs
mailing list