On Wed, Dec 02, 2020 at 12:29:06PM -0500, Vivek Goyal wrote:
> On Wed, Dec 02, 2020 at 12:02:43PM -0500, Jeff Layton wrote:
>
> [..]
> > > > diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
> > > > index 290983bcfbb3..82a096a05bce 100644
> > > > --- a/fs/overlayfs/super.c
> > > > +++ b/fs/overlayfs/super.c
> > > > @@ -261,11 +261,18 @@ static int ovl_sync_fs(struct super_block *sb, int wait)
> > > > struct super_block *upper_sb;
> > > > int ret;
> > > >
> > > >
> > > >
> > > >
> > > > - if (!ovl_upper_mnt(ofs))
> > > > - return 0;
> > > > + ret = ovl_check_sync(ofs);
> > > > + /*
> > > > + * We have to always set the err, because the return value isn't
> > > > + * checked, and instead VFS looks at the writeback errseq after
> > > > + * this call.
> > > > + */
> > > > + if (ret < 0)
> > > > + errseq_set(&sb->s_wb_err, ret);
> > >
> > > I was wondering that why errseq_set() will result in returning error
> > > all the time. Then realized that last syncfs() call must have set
> > > ERRSEQ_SEEN flag and that will mean errseq_set() will increment
> > > counter and that means this syncfs() will will return error too. Cool.
> > >
> > > > +
> > > > + if (!ret)
> > > > + return ret;
> > > >
> > > >
> > > >
> > > >
> > > > - if (!ovl_should_sync(ofs))
> > > > - return 0;
> > > > /*
> > > > * Not called for sync(2) call or an emergency sync (SB_I_SKIP_SYNC).
> > > > * All the super blocks will be iterated, including upper_sb.
> > > > @@ -1927,6 +1934,8 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
> > > > sb->s_op = &ovl_super_operations;
> > > >
> > > >
> > > >
> > > >
> > > > if (ofs->config.upperdir) {
> > > > + struct super_block *upper_mnt_sb;
> > > > +
> > > > if (!ofs->config.workdir) {
> > > > pr_err("missing 'workdir'\n");
> > > > goto out_err;
> > > > @@ -1943,9 +1952,10 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
> > > > if (!ofs->workdir)
> > > > sb->s_flags |= SB_RDONLY;
> > > >
> > > >
> > > >
> > > >
> > > > - sb->s_stack_depth = ovl_upper_mnt(ofs)->mnt_sb->s_stack_depth;
> > > > - sb->s_time_gran = ovl_upper_mnt(ofs)->mnt_sb->s_time_gran;
> > > > -
> > > > + upper_mnt_sb = ovl_upper_mnt(ofs)->mnt_sb;
> > > > + sb->s_stack_depth = upper_mnt_sb->s_stack_depth;
> > > > + sb->s_time_gran = upper_mnt_sb->s_time_gran;
> > > > + ofs->upper_errseq = errseq_sample(&upper_mnt_sb->s_wb_err);
> > >
> > > I asked this question in last email as well. errseq_sample() will return
> > > 0 if current error has not been seen yet. That means next time a sync
> > > call comes for volatile mount, it will return an error. But that's
> > > not what we want. When we mounted a volatile overlay, if there is an
> > > existing error (seen/unseen), we don't care. We only care if there
> > > is a new error after the volatile mount, right?
> > >
> > > I guess we will need another helper similar to errseq_smaple() which
> > > just returns existing value of errseq. And then we will have to
> > > do something about errseq_check() to not return an error if "since"
> > > and "eseq" differ only by "seen" bit.
> > >
> > > Otherwise in current form, volatile mount will always return error
> > > if upperdir has error and it has not been seen by anybody.
> > >
> > > How did you finally end up testing the error case. Want to simualate
> > > error aritificially and test it.
> > >
I used the blockdev error injection layer. It only works with ext2, because
ext4 (and other filesystems) will error and go into readonly.
dd if=/dev/zero of=/tmp/loop bs=1M count=100
losetup /dev/loop8 /tmp/loop
mkfs.ext2 /dev/loop8
mount -o errors=continue /dev/loop8 /mnt/loop/
mkdir -p /mnt/loop/{upperdir,workdir}
mount -t overlay -o volatile,index=off,lowerdir=/root/lowerdir,upperdir=/mnt/loop/upperdir,workdir=/mnt/loop/workdir none /mnt/foo/
echo 1 > /sys/block/loop8/make-it-fail
echo 100 > /sys/kernel/debug/fail_make_request/probability
echo 1 > /sys/kernel/debug/fail_make_request/times
dd if=/dev/zero of=/mnt/foo/zero bs=1M count=1
sync
I tried to get XFS tests working, but I was unable to get a simpler repro than
above. This is also easy enough to do with a simple kernel module. Maybe it'd be
neat to be able to inject in errseq increments via the fault injection API one
day? I have no idea what the VFS's approach here is.
> >
> > If you don't want to see errors that occurred before you did the mount,
> > then you probably can just resurrect and rename the original version of
> > errseq_sample. Something like this, but with a different name:
> >
> > +errseq_t errseq_sample(errseq_t *eseq)
> > +{
> > + errseq_t old = READ_ONCE(*eseq);
> > + errseq_t new = old;
> > +
> > + /*
> > + * For the common case of no errors ever having been set, we can skip
> > + * marking the SEEN bit. Once an error has been set, the value will
> > + * never go back to zero.
> > + */
> > + if (old != 0) {
> > + new |= ERRSEQ_SEEN;
> > + if (old != new)
> > + cmpxchg(eseq, old, new);
> > + }
> > + return new;
> > +}
>
> Yes, a helper like this should solve the issue at hand. We are not
> interested in previous errors. This also sets the ERRSEQ_SEEN on
> sample and it will also solve the other issue when after sampling
> if error gets seen, we don't want errseq_check() to return error.
>
> Thinking of some possible names for new function.
>
> errseq_sample_seen()
> errseq_sample_set_seen()
> errseq_sample_consume_unseen()
> errseq_sample_current()
>
> Thanks
> Vivek
>
I think we can just replace the code in super.c with:
ofs->upper_errseq = READ_ONCE(&upper_mnt_sb->s_wb_err);
And then add an errseq helper which checks:
int errseq_check_ignore_seen(errseq_t *eseq, errseq_t since)
{
errseq_t cur = READ_ONCE(*eseq);
if ((cur == since) || (cur == since | ERRSEQ_SEEN))
return 0;
return -(cur & MAX_ERRNO);
}
---
This extra (cur == since | ERRSEQ_SEEN) ignores the situation where cur has
"been seen". We do not want to do the cmpxchg I think because that would hide
the situation from the user where if they do a syncfs we hide the error from
the user.
If the since had seen already set, but cur does not have seen set, it means
we've wrapped.
