| Message ID | 20191107025927.GA6219@magnolia (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | iomap: iomap_bmap should check iomap_apply return value | expand |
On Wed, Nov 06, 2019 at 06:59:27PM -0800, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Check the return value of iomap_apply and return 0 (i.e. error) if it > didn't succeed. And how could we set the bno value if we didn't succeed?
On Thu, Nov 07, 2019 at 09:30:50AM +0100, Christoph Hellwig wrote: > On Wed, Nov 06, 2019 at 06:59:27PM -0800, Darrick J. Wong wrote: > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > Check the return value of iomap_apply and return 0 (i.e. error) if it > > didn't succeed. > > And how could we set the bno value if we didn't succeed? The iomap_bmap caller supplies an ->iomap_end that returns an error. Granted there's only one caller and it doesn't, so we could dump this patch and just tell Coverity to shut up, but it's odd that this is the one place where we ignore the return value. OTOH it's bmap which has been broken for ages; the more insane behavior seen in the wild, the better to scare away users. :P --D
On Thu, Nov 07, 2019 at 07:36:17AM -0800, Darrick J. Wong wrote: > On Thu, Nov 07, 2019 at 09:30:50AM +0100, Christoph Hellwig wrote: > > On Wed, Nov 06, 2019 at 06:59:27PM -0800, Darrick J. Wong wrote: > > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > > > Check the return value of iomap_apply and return 0 (i.e. error) if it > > > didn't succeed. > > > > And how could we set the bno value if we didn't succeed? > > The iomap_bmap caller supplies an ->iomap_end that returns an error. > > Granted there's only one caller and it doesn't, so we could dump this > patch and just tell Coverity to shut up, but it's odd that this is the > one place where we ignore the return value. > > OTOH it's bmap which has been broken for ages; the more insane behavior > seen in the wild, the better to scare away users. :P Oh well. I guess the patch is fine, it just isn't really needed as-is. Reviewed-by: Christoph Hellwig <hch@lst.de>
On Fri, Nov 08, 2019 at 06:51:51AM +0100, Christoph Hellwig wrote: > On Thu, Nov 07, 2019 at 07:36:17AM -0800, Darrick J. Wong wrote: > > On Thu, Nov 07, 2019 at 09:30:50AM +0100, Christoph Hellwig wrote: > > > On Wed, Nov 06, 2019 at 06:59:27PM -0800, Darrick J. Wong wrote: > > > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > > > > > Check the return value of iomap_apply and return 0 (i.e. error) if it > > > > didn't succeed. > > > > > > And how could we set the bno value if we didn't succeed? > > > > The iomap_bmap caller supplies an ->iomap_end that returns an error. > > > > Granted there's only one caller and it doesn't, so we could dump this > > patch and just tell Coverity to shut up, but it's odd that this is the > > one place where we ignore the return value. > > > > OTOH it's bmap which has been broken for ages; the more insane behavior > > seen in the wild, the better to scare away users. :P > > Oh well. I guess the patch is fine, it just isn't really needed as-is. Thanks for the review. :) --D > Reviewed-by: Christoph Hellwig <hch@lst.de>
diff --git a/fs/iomap/fiemap.c b/fs/iomap/fiemap.c index 690ef2d7c6c8..bccf305ea9ce 100644 --- a/fs/iomap/fiemap.c +++ b/fs/iomap/fiemap.c @@ -133,12 +133,16 @@ iomap_bmap(struct address_space *mapping, sector_t bno, struct inode *inode = mapping->host; loff_t pos = bno << inode->i_blkbits; unsigned blocksize = i_blocksize(inode); + int ret; if (filemap_write_and_wait(mapping)) return 0; bno = 0; - iomap_apply(inode, pos, blocksize, 0, ops, &bno, iomap_bmap_actor); + ret = iomap_apply(inode, pos, blocksize, 0, ops, &bno, + iomap_bmap_actor); + if (ret) + return 0; return bno; } EXPORT_SYMBOL_GPL(iomap_bmap);