lkml.org 
[lkml]   [2020]   [Sep]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: More filesystem need this fix (xfs: use MMAPLOCK around filemap_map_pages())
On Sat 12-09-20 09:19:11, Amir Goldstein wrote:
> On Tue, Jun 23, 2020 at 8:21 AM Dave Chinner <david@fromorbit.com> wrote:
> >
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > The page faultround path ->map_pages is implemented in XFS via
> > filemap_map_pages(). This function checks that pages found in page
> > cache lookups have not raced with truncate based invalidation by
> > checking page->mapping is correct and page->index is within EOF.
> >
> > However, we've known for a long time that this is not sufficient to
> > protect against races with invalidations done by operations that do
> > not change EOF. e.g. hole punching and other fallocate() based
> > direct extent manipulations. The way we protect against these
> > races is we wrap the page fault operations in a XFS_MMAPLOCK_SHARED
> > lock so they serialise against fallocate and truncate before calling
> > into the filemap function that processes the fault.
> >
> > Do the same for XFS's ->map_pages implementation to close this
> > potential data corruption issue.
> >
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > ---
> > fs/xfs/xfs_file.c | 15 ++++++++++++++-
> > 1 file changed, 14 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
> > index 7b05f8fd7b3d..4b185a907432 100644
> > --- a/fs/xfs/xfs_file.c
> > +++ b/fs/xfs/xfs_file.c
> > @@ -1266,10 +1266,23 @@ xfs_filemap_pfn_mkwrite(
> > return __xfs_filemap_fault(vmf, PE_SIZE_PTE, true);
> > }
> >
> > +static void
> > +xfs_filemap_map_pages(
> > + struct vm_fault *vmf,
> > + pgoff_t start_pgoff,
> > + pgoff_t end_pgoff)
> > +{
> > + struct inode *inode = file_inode(vmf->vma->vm_file);
> > +
> > + xfs_ilock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
> > + filemap_map_pages(vmf, start_pgoff, end_pgoff);
> > + xfs_iunlock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
> > +}
> > +
> > static const struct vm_operations_struct xfs_file_vm_ops = {
> > .fault = xfs_filemap_fault,
> > .huge_fault = xfs_filemap_huge_fault,
> > - .map_pages = filemap_map_pages,
> > + .map_pages = xfs_filemap_map_pages,
> > .page_mkwrite = xfs_filemap_page_mkwrite,
> > .pfn_mkwrite = xfs_filemap_pfn_mkwrite,
> > };
> > --
> > 2.26.2.761.g0e0b3e54be
> >
>
> It appears that ext4, f2fs, gfs2, orangefs, zonefs also need this fix
>
> zonefs does not support hole punching, so it may not need to use
> mmap_sem at all.

So I've written an ext4 fix for this but before actually posting it Nikolay
working on btrfs fix asked why exactly is filemap_map_pages() actually a
problem and I think he's right it actually isn't a problem. The thing is:
filemap_map_pages() never does any page mapping or IO. It only creates PTEs
for uptodate pages that are already in page cache. As such it is a rather
different beast compared to the fault handler from fs POV and does not need
protection from hole punching (current serialization on page lock and
checking of page->mapping is enough).

That being said I agree this is subtle and the moment someone adds e.g. a
readahead call into filemap_map_pages() we have a real problem. I'm not
sure how to prevent this risk...

Honza
--
Jan Kara <jack@suse.com>
SUSE Labs, CR

\
 
 \ /
  Last update: 2020-09-16 23:04    [W:0.097 / U:12.792 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site