2014-03-10 09:42:05

by Yan, Zheng

Subject: [PATCH v3] nfsd4: fix memory leak in nfsd4_encode_fattr()

The temporary file handle should be freed when nfsd4_encode_fattr()
finishes its job. Christoph Hellwig suggests to move the code that
generates the temporary file handle into nfsd4_encode_dirent_fattr()

Signed-off-by: Yan, Zheng <[email protected]>
---
fs/nfsd/nfs4xdr.c | 34 ++++++++++++++++++----------------
1 file changed, 18 insertions(+), 16 deletions(-)

diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index 63f2395..73698f7 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -2058,7 +2058,6 @@ nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
u32 bmval1 = bmval[1];
u32 bmval2 = bmval[2];
struct kstat stat;
- struct svc_fh *tempfh = NULL;
struct kstatfs statfs;
int buflen = count << 2;
__be32 *attrlenp;
@@ -2104,17 +2103,6 @@ nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
if (err)
goto out_nfserr;
}
- if ((bmval0 & (FATTR4_WORD0_FILEHANDLE | FATTR4_WORD0_FSID)) && !fhp) {
- tempfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
- status = nfserr_jukebox;
- if (!tempfh)
- goto out;
- fh_init(tempfh, NFS4_FHSIZE);
- status = fh_compose(tempfh, exp, dentry, NULL);
- if (status)
- goto out;
- fhp = tempfh;
- }
if (bmval0 & (FATTR4_WORD0_ACL | FATTR4_WORD0_ACLSUPPORT
| FATTR4_WORD0_SUPPORTED_ATTRS)) {
err = nfsd4_get_nfs4_acl(rqstp, dentry, &acl);
@@ -2499,8 +2487,6 @@ out:
security_release_secctx(context, contextlen);
#endif /* CONFIG_NFSD_V4_SECURITY_LABEL */
kfree(acl);
- if (tempfh)
- fh_put(tempfh);
return status;
out_nfserr:
status = nfserrno(err);
@@ -2524,6 +2510,7 @@ nfsd4_encode_dirent_fattr(struct nfsd4_readdir *cd,
const char *name, int namlen, __be32 **p, int buflen)
{
struct svc_export *exp = cd->rd_fhp->fh_export;
+ struct svc_fh *tempfh = NULL;
struct dentry *dentry;
__be32 nfserr;
int ignore_crossmnt = 0;
@@ -2573,9 +2560,24 @@ nfsd4_encode_dirent_fattr(struct nfsd4_readdir *cd,

}
out_encode:
- nfserr = nfsd4_encode_fattr(NULL, exp, dentry, p, buflen, cd->rd_bmval,
- cd->rd_rqstp, ignore_crossmnt);
+ if ((cd->rd_bmval[0] & (FATTR4_WORD0_FILEHANDLE | FATTR4_WORD0_FSID))) {
+ tempfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
+ if (!tempfh) {
+ nfserr = nfserr_jukebox;
+ goto out_put;
+ }
+ fh_init(tempfh, NFS4_FHSIZE);
+ nfserr = fh_compose(tempfh, exp, dentry, NULL);
+ if (nfserr)
+ goto out_put;
+ }
+ nfserr = nfsd4_encode_fattr(tempfh, exp, dentry, p, buflen,
+ cd->rd_bmval, cd->rd_rqstp, ignore_crossmnt);
out_put:
+ if (tempfh) {
+ fh_put(tempfh);
+ kfree(tempfh);
+ }
dput(dentry);
exp_put(exp);
return nfserr;
--
1.8.5.3

2014-03-31 21:11:14

by J. Bruce Fields

Subject: Re: [PATCH v3] nfsd4: fix memory leak in nfsd4_encode_fattr()

On Mon, Mar 10, 2014 at 03:21:01AM -0700, Christoph Hellwig wrote:
> On Mon, Mar 10, 2014 at 05:41:41PM +0800, Yan, Zheng wrote:
> > The temporary file handle should be freed when nfsd4_encode_fattr()
> > finishes its job. Christoph Hellwig suggests to move the code that
> > generates the temporary file handle into nfsd4_encode_dirent_fattr()
> >
> > Signed-off-by: Yan, Zheng <[email protected]>
>
> I'd also add a BUG_ON to assert we have the file handle if needed in
> nfsd4_encode_fattr.
>
> But maybe I was a bit too fast in my initial reply and we should put in
> your simple fix first for the current -rc and -stable and have the
> cleanup on top of it. There's a lot more refactoring that should be
> done to nfsd4_encode_fattr anyway.

Makes sense; applying the bare fix for 3.15 with a stable cc, assuming
somebody will follow up with any cleanup. (Moving this case to
nfsd4_encode_dirent_fattr() as in this patch certianly looks fine to
me.)

--b.

2014-03-10 10:21:02

by Christoph Hellwig

Subject: Re: [PATCH v3] nfsd4: fix memory leak in nfsd4_encode_fattr()

On Mon, Mar 10, 2014 at 05:41:41PM +0800, Yan, Zheng wrote:
> The temporary file handle should be freed when nfsd4_encode_fattr()
> finishes its job. Christoph Hellwig suggests to move the code that
> generates the temporary file handle into nfsd4_encode_dirent_fattr()
>
> Signed-off-by: Yan, Zheng <[email protected]>

I'd also add a BUG_ON to assert we have the file handle if needed in
nfsd4_encode_fattr.

But maybe I was a bit too fast in my initial reply and we should put in
your simple fix first for the current -rc and -stable and have the
cleanup on top of it. There's a lot more refactoring that should be
done to nfsd4_encode_fattr anyway.