nfsd: Return EPERM, not EACCES, in some SETATTR cases
authorzhengbin <zhengbin13@huawei.com>
Fri, 30 Nov 2018 08:04:25 +0000 (16:04 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 21 Dec 2019 09:41:24 +0000 (10:41 +0100)
[ Upstream commit 255fbca65137e25b12bced18ec9a014dc77ecda0 ]

As the man(2) page for utime/utimes states, EPERM is returned when the
second parameter of utime or utimes is not NULL, the caller's effective UID
does not match the owner of the file, and the caller is not privileged.

However, in a NFS directory mounted from knfsd, it will return EACCES
(from nfsd_setattr-> fh_verify->nfsd_permission).  This patch fixes
that.

Signed-off-by: zhengbin <zhengbin13@huawei.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
fs/nfsd/vfs.c

index 8f0b19a3ca810993d02eb20fb160b2da017dcd59..b8cd100cfcd6dd3d79766671d938d7894a45ba41 100644 (file)
@@ -395,10 +395,23 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
        bool            get_write_count;
        bool            size_change = (iap->ia_valid & ATTR_SIZE);
 
-       if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
+       if (iap->ia_valid & ATTR_SIZE) {
                accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
-       if (iap->ia_valid & ATTR_SIZE)
                ftype = S_IFREG;
+       }
+
+       /*
+        * If utimes(2) and friends are called with times not NULL, we should
+        * not set NFSD_MAY_WRITE bit. Otherwise fh_verify->nfsd_permission
+        * will return EACCESS, when the caller's effective UID does not match
+        * the owner of the file, and the caller is not privileged. In this
+        * situation, we should return EPERM(notify_change will return this).
+        */
+       if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME)) {
+               accmode |= NFSD_MAY_OWNER_OVERRIDE;
+               if (!(iap->ia_valid & (ATTR_ATIME_SET | ATTR_MTIME_SET)))
+                       accmode |= NFSD_MAY_WRITE;
+       }
 
        /* Callers that do fh_verify should do the fh_want_write: */
        get_write_count = !fhp->fh_dentry;