On Sat, Oct 13, 2018 at 7:10 AM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
>
> On Thu, Oct 11, 2018 at 05:38:14PM +0300, Amir Goldstein wrote:
> > Some anon_bdev filesystems (e.g. overlayfs, ceph) don't have s_blocksize
> > set. Returning zero from FIGETBSZ ioctl results in a Floating point
> > exception from the e2fsprogs utility filefrag, which divides the size of
> > the file with the value returned by FIGETBSZ.
> >
> > Fix the interface by returning -EINVAL for these filesystems.
> >
> > Fixes: d1d04ef8572b ("ovl: stack file ops")
> > Signed-off-by: Amir Goldstein <amir73il@xxxxxxxxx>
> > ---
> > fs/ioctl.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/fs/ioctl.c b/fs/ioctl.c
> > index 2005529af560..0400297c8d72 100644
> > --- a/fs/ioctl.c
> > +++ b/fs/ioctl.c
> > @@ -669,6 +669,9 @@ int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
> > return ioctl_fiemap(filp, arg);
> >
> > case FIGETBSZ:
> > + /* anon_bdev filesystems may not have a block size */
> > + if (!inode->i_sb->s_blocksize)
> > + return -EINVAL;
> > return put_user(inode->i_sb->s_blocksize, argp);
>
> Probably makes sense... Out of curiosity - what does the same utility do
> when faced with -EINVAL here?
filefrag was actually converted from FIBMAP/FIGETBSZ to FIEMAP/statfs(2)
but apparently that didn't work well for some fs, so FIGETBSZ was
brought back from retirement to override b_size.
When ioctl(FIGETBSZ) fails filefrag continues with f_bsize without any
complains.
FYI, another utility in e2fsprogs that uses FIGETBSZ, copy_sparse, will
simply exit when ioctl(FIGETBSZ) fails. But copy_sparse only has
FIBMAP implementation so it is going to fail on overlay file anyway.
Cheers,
Amir.
commit 01824c9bbcd2c1037085213f6cd7d591f78db9c5
Author: Rakesh Pandit <rakesh@xxxxxxxxxx>
Date: Sun Jul 27 20:04:48 2014 -0400
filefrag: fix block size value
ioctl(FIGETBSZ) was used to get block size earlier but 2508eaa7
(filefrag: improvements to filefrag FIEMAP handling) moved to fstatfs
f_bsize which doesn't work well for many files systems.
Block size returned using fstatfs isn't block size but "optimal
transfer block size" as per man page. Even stat st_blksize is
"preferred I/O block size" and in may file systems it may even vary
from file to file (POSIX). This patch changes filefrag to use
FIGETBSZ preferentially over f_bsize.
[ Modified by tytso to add the fallback to f_bsize if FIGETBSZ fails
for some reason ]
Signed-off-by: Rakesh Pandit <rakesh@xxxxxxxxxx>
Signed-off-by: Theodore Ts'o <tytso@xxxxxxx>
