| Message ID | e722c29f7291b9c31533fd8b2e84d7f469c3ac88.1526452889.git.osandov@fb.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, May 15, 2018 at 11:45:38PM -0700, Omar Sandoval wrote: > From: Omar Sandoval <osandov@fb.com> > > generic_swapfile_activate() doesn't allow holes, so we should be > consistent here. This is also a bit safer: if the user creates a > swapfile with, say, truncate -s $SIZE followed by mkswap, they should > really get an error and not much less swap space than they expected. > swapon(8) will error out before calling swapon(2) if the file has holes, > anyways. > > Fixes: 9d93388b0afe ("iomap: add a swapfile activation function") > Signed-off-by: Omar Sandoval <osandov@fb.com> > --- > Hey, Darrick, I noticed this while writing up a generic xfstest to test > that the Btrfs swap support patches don't allow a swapfile with holes. > It'd be nice if we were all consistent :) This is based on > xfs-linux/for-next. Feel free to fold it in to your patch or apply it > separately as you see fit. Thanks! I sent a testcase of my own ("generic: test swapfile creation, activation, and deactivation") a while back; would you mind sending out yours so we can combine them into a single testcase? > fs/iomap.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/fs/iomap.c b/fs/iomap.c > index d193390a1c20..ba559adaa327 100644 > --- a/fs/iomap.c > +++ b/fs/iomap.c > @@ -1214,9 +1214,9 @@ static loff_t iomap_swapfile_activate_actor(struct inode *inode, loff_t pos, > struct iomap_swapfile_info *isi = data; > int error; > > - /* Skip holes. */ > + /* No holes. */ > if (iomap->type == IOMAP_HOLE) > - goto out; > + goto err; Ok. I agree that it looks weird to mount a swap file with holes, so I guess the least-surprise principle applies here and we should emulate the old behavior completely. Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> --D > > /* Only one bdev per swap file. */ > if (iomap->bdev != isi->sis->bdev) > -- > 2.17.0 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, May 16, 2018 at 08:50:00AM -0700, Darrick J. Wong wrote: > > Hey, Darrick, I noticed this while writing up a generic xfstest to test > > that the Btrfs swap support patches don't allow a swapfile with holes. > > It'd be nice if we were all consistent :) This is based on > > xfs-linux/for-next. Feel free to fold it in to your patch or apply it > > separately as you see fit. Thanks! > > I sent a testcase of my own ("generic: test swapfile creation, > activation, and deactivation") a while back; would you mind sending out > yours so we can combine them into a single testcase? Wasn't the desire to support holes the rationale for the Aleksei version of the iomap swapfile patch?
On Wed, May 16, 2018 at 08:51:46AM -0700, Christoph Hellwig wrote: > On Wed, May 16, 2018 at 08:50:00AM -0700, Darrick J. Wong wrote: > > > Hey, Darrick, I noticed this while writing up a generic xfstest to test > > > that the Btrfs swap support patches don't allow a swapfile with holes. > > > It'd be nice if we were all consistent :) This is based on > > > xfs-linux/for-next. Feel free to fold it in to your patch or apply it > > > separately as you see fit. Thanks! > > > > I sent a testcase of my own ("generic: test swapfile creation, > > activation, and deactivation") a while back; would you mind sending out > > yours so we can combine them into a single testcase? > > Wasn't the desire to support holes the rationale for the Aleksei > version of the iomap swapfile patch? Ah, so it was. FWIW I'm not sure why you'd /want/ a holey swapfile? TBH my motivation for iomap swapfile is purely to remove bmap callers. :) --D > -- > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, May 16, 2018 at 08:56:38AM -0700, Darrick J. Wong wrote: > On Wed, May 16, 2018 at 08:51:46AM -0700, Christoph Hellwig wrote: > > On Wed, May 16, 2018 at 08:50:00AM -0700, Darrick J. Wong wrote: > > > > Hey, Darrick, I noticed this while writing up a generic xfstest to test > > > > that the Btrfs swap support patches don't allow a swapfile with holes. > > > > It'd be nice if we were all consistent :) This is based on > > > > xfs-linux/for-next. Feel free to fold it in to your patch or apply it > > > > separately as you see fit. Thanks! > > > > > > I sent a testcase of my own ("generic: test swapfile creation, > > > activation, and deactivation") a while back; would you mind sending out > > > yours so we can combine them into a single testcase? Sure thing, I have a small pile of tests. I'm still working on some Btrfs-specific ones, but I can send out the generic ones and we can figure out how to merge them. > > Wasn't the desire to support holes the rationale for the Aleksei > > version of the iomap swapfile patch? > > Ah, so it was. FWIW I'm not sure why you'd /want/ a holey swapfile? From reading the old thread, it looks like Aleksei just wanted fallocated swap files to work: "I've traced the problem to bmap(), used in generic_swapfile_activate call, which returns 0 for blocks inside holes created by fallocate". Are holes in that sense are different from actual holes in the iomap sense?
> > - /* Skip holes. */ > + /* No holes. */ > if (iomap->type == IOMAP_HOLE) > - goto out; > + goto err; If we end up not allowing holes this check can just be removed entirely. We already check for the allow types below. I have to admit I would move that type check above the bdev check, though.
On Wed, May 16, 2018 at 09:19:55AM -0700, Omar Sandoval wrote: > > > Wasn't the desire to support holes the rationale for the Aleksei > > > version of the iomap swapfile patch? > > > > Ah, so it was. FWIW I'm not sure why you'd /want/ a holey swapfile? > > >From reading the old thread, it looks like Aleksei just wanted > fallocated swap files to work: "I've traced the problem to bmap(), used > in generic_swapfile_activate call, which returns 0 for blocks inside > holes created by fallocate". Oh, that makes more sense. > Are holes in that sense are different from > actual holes in the iomap sense? Unwritten extents aren't actually holes in any sense, so they are very different and should work with the iomap swapfile code.
On Wed, May 16, 2018 at 09:23:29AM -0700, Christoph Hellwig wrote: > > > > - /* Skip holes. */ > > + /* No holes. */ > > if (iomap->type == IOMAP_HOLE) > > - goto out; > > + goto err; > > If we end up not allowing holes this check can just be removed > entirely. We already check for the allow types below. I have to > admit I would move that type check above the bdev check, though. True, although I wonder if we should be logging more specific error messages instead of a generic "not a valid swap file" message. At least for my Btrfs patches, that's what I do. I'll send a v2 with a second patch doing that.
diff --git a/fs/iomap.c b/fs/iomap.c index d193390a1c20..ba559adaa327 100644 --- a/fs/iomap.c +++ b/fs/iomap.c @@ -1214,9 +1214,9 @@ static loff_t iomap_swapfile_activate_actor(struct inode *inode, loff_t pos, struct iomap_swapfile_info *isi = data; int error; - /* Skip holes. */ + /* No holes. */ if (iomap->type == IOMAP_HOLE) - goto out; + goto err; /* Only one bdev per swap file. */ if (iomap->bdev != isi->sis->bdev)