ceph: force sending a cap update msg back to MDS for revoke op

JIRA: https://issues.redhat.com/browse/RHEL-48618
Conflicts: missing the commit 38d4640("ceph: print cluster fsid and
client global_id in all debug logs") and a tiny debug log conflict.

commit 31634d7
Author: Xiubo Li <xiubli@redhat.com>
Date:   Fri Jul 12 12:40:19 2024 +0800

    ceph: force sending a cap update msg back to MDS for revoke op

    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'.

    Currently if the cap->issued equals to the newcaps the check_caps()
    will do nothing, we should force flush the caps.

    Cc: stable@vger.kernel.org
    Link: https://tracker.ceph.com/issues/61782
    Signed-off-by: Xiubo Li <xiubli@redhat.com>
    Reviewed-by: Venky Shankar <vshankar@redhat.com>
    Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

Signed-off-by: Xiubo Li <xiubli@redhat.com>

Author: lxbsz

fs/ceph/caps.c

@@ -1963,6 +1963,8 @@ bool __ceph_should_report_size(struct ceph_inode_info *ci)
  *  CHECK_CAPS_AUTHONLY - we should only check the auth cap
  *  CHECK_CAPS_FLUSH - we should flush any dirty caps immediately, without
  *    further delay.
+ *  CHECK_CAPS_FLUSH_FORCE - we should flush any caps immediately, without
+ *    further delay.
  */
 void ceph_check_caps(struct ceph_inode_info *ci, int flags)
 {
@@ -2043,15 +2045,16 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags)
 	}
 
 	dout("check_caps %llx.%llx file_want %s used %s dirty %s flushing %s"
-	     " issued %s revoking %s retain %s %s%s%s\n", ceph_vinop(inode),
+	     " issued %s revoking %s retain %s %s%s%s%s\n", ceph_vinop(inode),
 	     ceph_cap_string(file_wanted),
 	     ceph_cap_string(used), ceph_cap_string(ci->i_dirty_caps),
 	     ceph_cap_string(ci->i_flushing_caps),
 	     ceph_cap_string(issued), ceph_cap_string(revoking),
 	     ceph_cap_string(retain),
 	     (flags & CHECK_CAPS_AUTHONLY) ? " AUTHONLY" : "",
 	     (flags & CHECK_CAPS_FLUSH) ? " FLUSH" : "",
-	     (flags & CHECK_CAPS_NOINVAL) ? " NOINVAL" : "");
+	     (flags & CHECK_CAPS_NOINVAL) ? " NOINVAL" : "",
+	     (flags & CHECK_CAPS_FLUSH_FORCE) ? " FLUSH_FORCE" : "");
 
 	/*
 	 * If we no longer need to hold onto old our caps, and we may
@@ -2126,6 +2129,11 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags)
 				queue_writeback = true;
 		}
 
+		if (flags & CHECK_CAPS_FLUSH_FORCE) {
+			dout("force to flush caps\n");
+			goto ack;
+		}
+
 		if (cap == ci->i_auth_cap &&
 		    (cap->issued & CEPH_CAP_FILE_WR)) {
 			/* request larger max_size from MDS? */
@@ -3418,6 +3426,8 @@ static void handle_cap_grant(struct inode *inode,
 	bool queue_invalidate = false;
 	bool deleted_inode = false;
 	bool fill_inline = false;
+	bool revoke_wait = false;
+	int flags = 0;
 
 	/*
 	 * If there is at least one crypto block then we'll trust
@@ -3612,16 +3622,18 @@ static void handle_cap_grant(struct inode *inode,
 		     ceph_cap_string(newcaps),
 		     ceph_cap_string(revoking));
 		if (S_ISREG(inode->i_mode) &&
-		    (revoking & used & CEPH_CAP_FILE_BUFFER))
+		    (revoking & used & CEPH_CAP_FILE_BUFFER)) {
 			writeback = true;  /* initiate writeback; will delay ack */
-		else if (queue_invalidate &&
+			revoke_wait = true;
+		} else if (queue_invalidate &&
 			 revoking == CEPH_CAP_FILE_CACHE &&
-			 (newcaps & CEPH_CAP_FILE_LAZYIO) == 0)
-			; /* do nothing yet, invalidation will be queued */
-		else if (cap == ci->i_auth_cap)
+			 (newcaps & CEPH_CAP_FILE_LAZYIO) == 0) {
+			revoke_wait = true; /* do nothing yet, invalidation will be queued */
+		} else if (cap == ci->i_auth_cap) {
 			check_caps = 1; /* check auth cap only */
-		else
+		} else {
 			check_caps = 2; /* check all caps */
+		}
 		/* If there is new caps, try to wake up the waiters */
 		if (~cap->issued & newcaps)
 			wake = true;
@@ -3647,8 +3659,9 @@ 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) {
+	if (!revoke_wait && le32_to_cpu(grant->op) == CEPH_CAP_OP_REVOKE) {
 		cap->mds_wanted = 0;
+		flags |= CHECK_CAPS_FLUSH_FORCE;
 		if (cap == ci->i_auth_cap)
 			check_caps = 1; /* check auth cap only */
 		else
@@ -3704,9 +3717,9 @@ static void handle_cap_grant(struct inode *inode,
 
 	mutex_unlock(&session->s_mutex);
 	if (check_caps == 1)
-		ceph_check_caps(ci, CHECK_CAPS_AUTHONLY | CHECK_CAPS_NOINVAL);
+		ceph_check_caps(ci, flags | CHECK_CAPS_AUTHONLY | CHECK_CAPS_NOINVAL);
 	else if (check_caps == 2)
-		ceph_check_caps(ci, CHECK_CAPS_NOINVAL);
+		ceph_check_caps(ci, flags | CHECK_CAPS_NOINVAL);
 }
 
 /*

fs/ceph/super.h

@@ -199,9 +199,10 @@ struct ceph_cap {
 	struct list_head caps_item;
 };
 
-#define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
-#define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
-#define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
+#define CHECK_CAPS_AUTHONLY     1  /* only check auth cap */
+#define CHECK_CAPS_FLUSH        2  /* flush any dirty caps */
+#define CHECK_CAPS_NOINVAL      4  /* don't invalidate pagecache */
+#define CHECK_CAPS_FLUSH_FORCE  8  /* force flush any caps */
 
 struct ceph_cap_flush {
 	u64 tid;