On 6/29/23 09:01, Patrick Donnelly wrote:
Patch looks good to me. Sorry I must nitpick the commit message wording:
If a client sends out a cap update dropping caps with the prior 'seq' just
before an incoming cap revoke request, then the client may drop the revoke
because it believes it's already released the requested capabilities.
This causes
the MDS to wait indefinitely for the client to respond to the revoke.
It's therefore always a good idea to ack the cap revoke request with
the bumped up 'seq'.
Updated it by sending out the V3.
Thanks
- Xiubo
On Tue, Jun 27, 2023 at 7:59 PM <xiubli@xxxxxxxxxx> wrote:
From: Xiubo Li <xiubli@xxxxxxxxxx>
If a client sends out a cap-update request with the old 'seq' just
before a pending cap revoke request, then the MDS might miscalculate
the 'seqs' and caps. It's therefore always a good idea to ack the
cap revoke request with the bumped up 'seq'.
Cc: stable@xxxxxxxxxxxxxxx
Cc: Patrick Donnelly <pdonnell@xxxxxxxxxx>
URL: https://tracker.ceph.com/issues/61782
Signed-off-by: Xiubo Li <xiubli@xxxxxxxxxx>
---
V2:
- Rephrased the commit comment for better understanding from Milind
fs/ceph/caps.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index 1052885025b3..eee2fbca3430 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -3737,6 +3737,15 @@ static void handle_cap_grant(struct inode *inode,
}
BUG_ON(cap->issued & ~cap->implemented);
+ /* don't let check_caps skip sending a response to MDS for revoke msgs */
+ if (le32_to_cpu(grant->op) == CEPH_CAP_OP_REVOKE) {
+ cap->mds_wanted = 0;
+ if (cap == ci->i_auth_cap)
+ check_caps = 1; /* check auth cap only */
+ else
+ check_caps = 2; /* check all caps */
+ }
+
if (extra_info->inline_version > 0 &&
extra_info->inline_version >= ci->i_inline_version) {
ci->i_inline_version = extra_info->inline_version;
--
2.40.1
