nfsd shoudn't call mnt_want_write twice
Subsystems (2)
cover nfsd shoudn't call mnt_want_write twice
From: "J. Bruce Fields" <[email protected]>
A fuzzer recently triggered lockdep warnings about potential sb_writers deadlocks caused by fh_want_write, showing that we haven't been careful to pair each fh_want_write() with an fh_drop_write().
This isn't normally a problem since fh_put() will call fh_drop_write() for us. And that's OK for NFSv3. But an NFSv4 protocol fuzzer can do weird things like call unlink twice in a compound.
So we can either make it safe to call fh_want_write() twice, or we can fix all the callers to call fh_drop_write().
For now I think we have to do the former just to get the bug fixed.
Long term I don't know whether it's best to stay with that or to fix up the callers. I fixed nfsd_unlink just because it's easy, but maybe that's pointless, it's others (like setattr) that are the complicated ones.
J. Bruce Fields (2):
nfsd: allow fh_want_write to be called twice
nfsd: fh_drop_write in nfsd_unlink
fs/nfsd/vfs.c | 8 +++++--- fs/nfsd/vfs.h | 5 ++++- 2 files changed, 9 insertions(+), 4 deletions(-)
Signature · 1 line
2.21.0
PATCH 1/2 nfsd: allow fh_want_write to be called twice
From: "J. Bruce Fields" <[email protected]>
A fuzzer recently triggered lockdep warnings about potential sb_writers deadlocks caused by fh_want_write().
Looks like we aren't careful to pair each fh_want_write() with an fh_drop_write().
It's not normally a problem since fh_put() will call fh_drop_write() for us. And was OK for NFSv3 where we'd do one operation that might call fh_want_write(), and then put the filehandle.
But an NFSv4 protocol fuzzer can do weird things like call unlink twice in a compound, and then we get into trouble.
I'm a little worried about this approach of just leaving everything to fh_put(). But I think there are probably a lot of fh_want_write()/fh_drop_write() imbalances so for now I think we need it to be more forgiving.
--- fs/nfsd/vfs.h | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
fs/nfsd/vfs.h +4 -1
@@ -120,8 +120,11 @@ void nfsd_put_raparams(struct file *file, struct raparms *ra);static inline int fh_want_write (struct svc_fh *fh) { -int ret =mnt_want_write (fh->fh_export->ex_path.mnt); +int ret; +if (fh->fh_want_write) +return 0 ; + ret =mnt_want_write (fh->fh_export->ex_path.mnt);if (!ret) fh->fh_want_write = true;return ret;
Signature · 1 line
2.21.0
PATCH 2/2 nfsd: fh_drop_write in nfsd_unlink
In reply to J. Bruce FieldsFrom: "J. Bruce Fields" <[email protected]>
fh_want_write() can now be called twice, but I'm also fixing up the callers not to do that.
Other cases include setattr and create.
--- fs/nfsd/vfs.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
fs/nfsd/vfs.c +5 -3
@@ -1786,12 +1786,12 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, rdentry =lookup_one_len (fname, dentry, flen); host_err =PTR_ERR (rdentry);if (IS_ERR (rdentry)) - goto out_nfserr; + goto out_drop_write;if (d_really_is_negative (rdentry)) {dput (rdentry); - err = nfserr_noent; - goto out; + host_err = -ENOENT ; + goto out_drop_write; }if (!type)
@@ -1805,6 +1805,8 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, host_err =commit_metadata (fhp);dput (rdentry); +out_drop_write: +fh_drop_write (fhp); out_nfserr: err =nfserrno (host_err); out:
Signature · 1 line
2.21.0