It's theoretically possible for a single SETATTR call to come in that
sets the mode and the uid/gid. In that case, assume the mode is
correct and don't set the ATTR_KILL_S*ID bits. Doing so would trip the
BUG() in notify_change.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
---
fs/nfsd/vfs.c | 17 +++++++++++------
1 files changed, 11 insertions(+), 6 deletions(-)
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 70f2c86..3b5b8cf 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -369,14 +369,19 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
if (iap->ia_valid & ATTR_MODE) {
iap->ia_mode &= S_IALLUGO;
imode = iap->ia_mode |= (imode & ~S_IALLUGO);
+ if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid)
+ iap->ia_valid |= ATTR_KILL_PRIV;
+ } else {
+ /*
+ * Revoke setuid/setgid bit on chown/chgrp, unless the mode
+ * is being explicitly set
+ */
+ if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid)
+ iap->ia_valid |= ATTR_KILL_SUID | ATTR_KILL_PRIV;
+ if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid)
+ iap->ia_valid |= ATTR_KILL_SGID;
}
- /* Revoke setuid/setgid bit on chown/chgrp */
- if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid)
- iap->ia_valid |= ATTR_KILL_SUID | ATTR_KILL_PRIV;
- if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid)
- iap->ia_valid |= ATTR_KILL_SGID;
-
/* Change the attributes. */
iap->ia_valid |= ATTR_CTIME;
--
1.5.2.1
-
| Jesse Barnes | Re: PCI probing changes |
| Borislav Petkov | [PATCH] [KERNEL-DOC] kill warnings when building mandocs |
| Greg Kroah-Hartman | [PATCH 012/196] nozomi driver |
| Roland Dreier | Re: Integration of SCST in the mainstream Linux kernel |
git: | |
| Herbert Xu | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Gerrit Renker | [PATCH 27/37] dccp: Integration of dynamic feature activation - part 2 (server side) |
| Linus Torvalds | Re: [GIT]: Networking |
| Frans Pop | svc: failed to register lockdv1 RPC service (errno 97). |
