A fix for a potential stall on umount caused by the MDS dropping

our REQUEST_CLOSE message.  The code that handled this case was
 inadvertently disabled in 5.9, this patch removes it entirely and
 fixes the problem in a way that is consistent with ceph-fuse.
 -----BEGIN PGP SIGNATURE-----
 
 iQFHBAABCAAxFiEEydHwtzie9C7TfviiSn/eOAIR84sFAl+lookTHGlkcnlvbW92
 QGdtYWlsLmNvbQAKCRBKf944AhHziwFGB/43MBB+nG4wBHM58GybIi2NkS/TMmd5
 5D3GPmchWYbE1d3hzAJmAYUUZCIx8kh0TeWjPZzR0iEok+f9Zf8bjrGDEFRWWOZc
 OL+PMfZckhVS2W8dUkx9CsypnA9/Rx2i7y/XKDDiC3eumfkDMktTdSS4UNxZT5cg
 ElmfozOPdv7fRGNPZJiQnkgWdMRFkqiGsdL+9wgRP4qc8WOkipoFouw+gJ2lN3vK
 odcY3UGcmx4iuiBj0uXjiFy/MtdYuNLjJrtMmkkEBklxGgIP/1dTOMnV3ktMMYkT
 gRUNM7fz/HeZIXb1N6jFs2S/ai1uuS6wP7aTHGfi8W2xgQA5ukDLIbu/
 =Tbrl
 -----END PGP SIGNATURE-----

Merge tag 'ceph-for-5.10-rc3' of git://github.com/ceph/ceph-client

Pull ceph fix from Ilya Dryomov:
 "A fix for a potential stall on umount caused by the MDS dropping our
  REQUEST_CLOSE message. The code that handled this case was
  inadvertently disabled in 5.9, this patch removes it entirely and
  fixes the problem in a way that is consistent with ceph-fuse"

* tag 'ceph-for-5.10-rc3' of git://github.com/ceph/ceph-client:
  ceph: check session state after bumping session->s_seq
This commit is contained in:
Linus Torvalds 2020-11-06 15:46:39 -08:00
commit 659caaf65d
5 changed files with 39 additions and 18 deletions

View File

@ -4074,7 +4074,7 @@ void ceph_handle_caps(struct ceph_mds_session *session,
vino.snap, inode);
mutex_lock(&session->s_mutex);
session->s_seq++;
inc_session_sequence(session);
dout(" mds%d seq %lld cap seq %u\n", session->s_mds, session->s_seq,
(unsigned)seq);

View File

@ -4231,7 +4231,7 @@ static void handle_lease(struct ceph_mds_client *mdsc,
dname.len, dname.name);
mutex_lock(&session->s_mutex);
session->s_seq++;
inc_session_sequence(session);
if (!inode) {
dout("handle_lease no inode %llx\n", vino.ino);
@ -4385,28 +4385,48 @@ static void maybe_recover_session(struct ceph_mds_client *mdsc)
bool check_session_state(struct ceph_mds_session *s)
{
if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
dout("resending session close request for mds%d\n",
s->s_mds);
request_close_session(s);
return false;
}
if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
if (s->s_state == CEPH_MDS_SESSION_OPEN) {
switch (s->s_state) {
case CEPH_MDS_SESSION_OPEN:
if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
s->s_state = CEPH_MDS_SESSION_HUNG;
pr_info("mds%d hung\n", s->s_mds);
}
}
if (s->s_state == CEPH_MDS_SESSION_NEW ||
s->s_state == CEPH_MDS_SESSION_RESTARTING ||
s->s_state == CEPH_MDS_SESSION_CLOSED ||
s->s_state == CEPH_MDS_SESSION_REJECTED)
/* this mds is failed or recovering, just wait */
break;
case CEPH_MDS_SESSION_CLOSING:
/* Should never reach this when we're unmounting */
WARN_ON_ONCE(true);
fallthrough;
case CEPH_MDS_SESSION_NEW:
case CEPH_MDS_SESSION_RESTARTING:
case CEPH_MDS_SESSION_CLOSED:
case CEPH_MDS_SESSION_REJECTED:
return false;
}
return true;
}
/*
* If the sequence is incremented while we're waiting on a REQUEST_CLOSE reply,
* then we need to retransmit that request.
*/
void inc_session_sequence(struct ceph_mds_session *s)
{
lockdep_assert_held(&s->s_mutex);
s->s_seq++;
if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
int ret;
dout("resending session close request for mds%d\n", s->s_mds);
ret = request_close_session(s);
if (ret < 0)
pr_err("unable to close session to mds%d: %d\n",
s->s_mds, ret);
}
}
/*
* delayed work -- periodically trim expired leases, renew caps with mds
*/

View File

@ -480,6 +480,7 @@ struct ceph_mds_client {
extern const char *ceph_mds_op_name(int op);
extern bool check_session_state(struct ceph_mds_session *s);
void inc_session_sequence(struct ceph_mds_session *s);
extern struct ceph_mds_session *
__ceph_lookup_mds_session(struct ceph_mds_client *, int mds);

View File

@ -53,7 +53,7 @@ void ceph_handle_quota(struct ceph_mds_client *mdsc,
/* increment msg sequence number */
mutex_lock(&session->s_mutex);
session->s_seq++;
inc_session_sequence(session);
mutex_unlock(&session->s_mutex);
/* lookup inode */

View File

@ -873,7 +873,7 @@ void ceph_handle_snap(struct ceph_mds_client *mdsc,
ceph_snap_op_name(op), split, trace_len);
mutex_lock(&session->s_mutex);
session->s_seq++;
inc_session_sequence(session);
mutex_unlock(&session->s_mutex);
down_write(&mdsc->snap_rwsem);