path: root/fs/dlm/user.c
diff options
authorDavid Teigland <teigland@redhat.com>2007-03-28 09:56:46 -0500
committerSteven Whitehouse <swhiteho@redhat.com>2007-05-01 09:11:00 +0100
commitef0c2bb05f40f9a0cd2deae63e199bfa62faa7fa (patch)
treedf73645f93cfec29fe5b854ff5990a69b03d5c1d /fs/dlm/user.c
parent032067270295cfca11975c0f7b467244aa170c14 (diff)
[DLM] overlapping cancel and unlock
Full cancel and force-unlock support. In the past, cancel and force-unlock wouldn't work if there was another operation in progress on the lock. Now, both cancel and unlock-force can overlap an operation on a lock, meaning there may be 2 or 3 operations in progress on a lock in parallel. This support is important not only because cancel and force-unlock are explicit operations that an app can use, but both are used implicitly when a process exits while holding locks. Summary of changes: - add-to and remove-from waiters functions were rewritten to handle situations with more than one remote operation outstanding on a lock - validate_unlock_args detects when an overlapping cancel/unlock-force can be sent and when it needs to be delayed until a request/lookup reply is received - processing request/lookup replies detects when cancel/unlock-force occured during the op, and carries out the delayed cancel/unlock-force - manipulation of the "waiters" (remote operation) state of a lock moved under the standard rsb mutex that protects all the other lock state - the two recovery routines related to locks on the waiters list changed according to the way lkb's are now locked before accessing waiters state - waiters recovery detects when lkb's being recovered have overlapping cancel/unlock-force, and may not recover such locks - revert_lock (cancel) returns a value to distinguish cases where it did nothing vs cases where it actually did a cancel; the cancel completion ast should only be done when cancel did something - orphaned locks put on new list so they can be found later for purging - cancel must be called on a lock when making it an orphan - flag user locks (ENDOFLIFE) at the end of their useful life (to the application) so we can return an error for any further cancel/unlock-force - we weren't setting COMP/BAST ast flags if one was already set, so we'd lose either a completion or blocking ast - clear an unread bast on a lock that's become unlocked Signed-off-by: David Teigland <teigland@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Diffstat (limited to 'fs/dlm/user.c')
1 files changed, 44 insertions, 33 deletions
diff --git a/fs/dlm/user.c b/fs/dlm/user.c
index 27a75ce571c..c978c67b1ef 100644
--- a/fs/dlm/user.c
+++ b/fs/dlm/user.c
@@ -1,5 +1,5 @@
- * Copyright (C) 2006 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2006-2007 Red Hat, Inc. All rights reserved.
* This copyrighted material is made available to anyone wishing to use,
* modify, copy, or redistribute it subject to the terms and conditions
@@ -128,35 +128,30 @@ static void compat_output(struct dlm_lock_result *res,
+/* we could possibly check if the cancel of an orphan has resulted in the lkb
+ being removed and then remove that lkb from the orphans list and free it */
void dlm_user_add_ast(struct dlm_lkb *lkb, int type)
struct dlm_ls *ls;
struct dlm_user_args *ua;
struct dlm_user_proc *proc;
- int remove_ownqueue = 0;
+ int eol = 0, ast_type;
- /* dlm_clear_proc_locks() sets ORPHAN/DEAD flag on each
- lkb before dealing with it. We need to check this
- flag before taking ls_clear_proc_locks mutex because if
- it's set, dlm_clear_proc_locks() holds the mutex. */
- if (lkb->lkb_flags & (DLM_IFL_ORPHAN | DLM_IFL_DEAD)) {
- /* log_print("user_add_ast skip1 %x", lkb->lkb_flags); */
+ if (lkb->lkb_flags & (DLM_IFL_ORPHAN | DLM_IFL_DEAD))
- }
ls = lkb->lkb_resource->res_ls;
/* If ORPHAN/DEAD flag is set, it means the process is dead so an ast
can't be delivered. For ORPHAN's, dlm_clear_proc_locks() freed
- lkb->ua so we can't try to use it. */
+ lkb->ua so we can't try to use it. This second check is necessary
+ for cases where a completion ast is received for an operation that
+ began before clear_proc_locks did its cancel/unlock. */
- if (lkb->lkb_flags & (DLM_IFL_ORPHAN | DLM_IFL_DEAD)) {
- /* log_print("user_add_ast skip2 %x", lkb->lkb_flags); */
+ if (lkb->lkb_flags & (DLM_IFL_ORPHAN | DLM_IFL_DEAD))
goto out;
- }
DLM_ASSERT(lkb->lkb_astparam, dlm_print_lkb(lkb););
ua = (struct dlm_user_args *)lkb->lkb_astparam;
@@ -166,28 +161,42 @@ void dlm_user_add_ast(struct dlm_lkb *lkb, int type)
goto out;
- if (!(lkb->lkb_ast_type & (AST_COMP | AST_BAST))) {
+ ast_type = lkb->lkb_ast_type;
+ lkb->lkb_ast_type |= type;
+ if (!ast_type) {
list_add_tail(&lkb->lkb_astqueue, &proc->asts);
- lkb->lkb_ast_type |= type;
- /* noqueue requests that fail may need to be removed from the
- proc's locks list, there should be a better way of detecting
- this situation than checking all these things... */
- if (type == AST_COMP && lkb->lkb_grmode == DLM_LOCK_IV &&
- ua->lksb.sb_status == -EAGAIN && !list_empty(&lkb->lkb_ownqueue))
- remove_ownqueue = 1;
- /* unlocks or cancels of waiting requests need to be removed from the
- proc's unlocking list, again there must be a better way... */
- if (ua->lksb.sb_status == -DLM_EUNLOCK ||
+ if (type == AST_COMP && (ast_type & AST_COMP))
+ log_debug(ls, "ast overlap %x status %x %x",
+ lkb->lkb_id, ua->lksb.sb_status, lkb->lkb_flags);
+ /* Figure out if this lock is at the end of its life and no longer
+ available for the application to use. The lkb still exists until
+ the final ast is read. A lock becomes EOL in three situations:
+ 1. a noqueue request fails with EAGAIN
+ 2. an unlock completes with EUNLOCK
+ 3. a cancel of a waiting request completes with ECANCEL
+ An EOL lock needs to be removed from the process's list of locks.
+ And we can't allow any new operation on an EOL lock. This is
+ not related to the lifetime of the lkb struct which is managed
+ entirely by refcount. */
+ if (type == AST_COMP &&
+ lkb->lkb_grmode == DLM_LOCK_IV &&
+ ua->lksb.sb_status == -EAGAIN)
+ eol = 1;
+ else if (ua->lksb.sb_status == -DLM_EUNLOCK ||
(ua->lksb.sb_status == -DLM_ECANCEL &&
lkb->lkb_grmode == DLM_LOCK_IV))
- remove_ownqueue = 1;
+ eol = 1;
+ if (eol) {
+ lkb->lkb_ast_type &= ~AST_BAST;
+ lkb->lkb_flags |= DLM_IFL_ENDOFLIFE;
+ }
/* We want to copy the lvb to userspace when the completion
ast is read if the status is 0, the lock has an lvb and
@@ -204,11 +213,13 @@ void dlm_user_add_ast(struct dlm_lkb *lkb, int type)
- if (remove_ownqueue) {
+ if (eol) {
- list_del_init(&lkb->lkb_ownqueue);
+ if (!list_empty(&lkb->lkb_ownqueue)) {
+ list_del_init(&lkb->lkb_ownqueue);
+ dlm_put_lkb(lkb);
+ }
- dlm_put_lkb(lkb);