ceph: don't let check_caps skip sending responses for revoke msgs
Subsystems (2)
PATCH v2 1/1 ceph: don't let check_caps skip sending responses for revoke msgs
From: Xiubo Li <[email protected]>
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: [email protected]
Cc: Patrick Donnelly <[email protected]>
URL: https://tracker.ceph.com/issues/61782
---
V2:
- Rephrased the commit comment for better understanding from Milind
fs/ceph/caps.c | 9 +++++++++ 1 file changed, 9 insertions(+)
fs/ceph/caps.c +9 -0
@@ -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;
Signature · 1 line
2.40.1
Looks good to me.
Quoted 44 lines · On Wed, Jun 28, 2023 at 5:29 AM <[email protected]> wrote:
From: Xiubo Li <[email protected]>
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: [email protected]
Cc: Patrick Donnelly <[email protected]>
URL: https://tracker.ceph.com/issues/61782Signed-off-by Xiubo Li <[email protected]>---
V2:
- Rephrased the commit comment for better understanding from Milindfs/ceph/caps.c | 9 +++++++++ 1 file changed, 9 insertions(+)fs/ceph/caps.c +9
-1@@ -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
Signature · 1 line
Milind
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'.
Quoted 44 lines · On Tue, Jun 27, 2023 at 7:59 PM <[email protected]> wrote:
From: Xiubo Li <[email protected]>
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: [email protected]
Cc: Patrick Donnelly <[email protected]>
URL: https://tracker.ceph.com/issues/61782Signed-off-by Xiubo Li <[email protected]>---
V2:
- Rephrased the commit comment for better understanding from Milindfs/ceph/caps.c | 9 +++++++++ 1 file changed, 9 insertions(+)fs/ceph/caps.c +9
-1@@ -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
Signature · 5 lines
Patrick Donnelly, Ph.D. He / Him / His Red Hat Partner Engineer IBM, Inc. GPG: 19F28A586F808C2402351B93C3301A3E258DD79D
Quoted 9 lines · Patrick Donnelly
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
Quoted 45 lines · On Tue, Jun 27, 2023 at 7:59 PM <[email protected]> wrote:
Quoted 43 lines
From: Xiubo Li <[email protected]>
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: [email protected]
Cc: Patrick Donnelly <[email protected]>
URL: https://tracker.ceph.com/issues/61782Signed-off-by Xiubo Li <[email protected]>---
V2:
- Rephrased the commit comment for better understanding from Milindfs/ceph/caps.c | 9 +++++++++ 1 file changed, 9 insertions(+)fs/ceph/caps.c +9
-1@@ -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