diff options
| author | Petr Baudis <pasky@suse.cz> | 2009-07-16 10:10:10 -0700 |
|---|---|---|
| committer | Ulrich Drepper <drepper@redhat.com> | 2009-07-16 10:10:10 -0700 |
| commit | 137028b4d7e50f71906c1656c27079eac5a1d085 (patch) | |
| tree | 61b121088c68e50ff2483cd15675729cf3c337e7 /nscd/mem.c | |
| parent | 50158f95525ca59459a90f2a7bc65ceb892a0807 (diff) | |
| download | glibc-137028b4d7e50f71906c1656c27079eac5a1d085.tar.gz glibc-137028b4d7e50f71906c1656c27079eac5a1d085.tar.xz glibc-137028b4d7e50f71906c1656c27079eac5a1d085.zip | |
Fix lock handling in memory hander of nscd.
The commit 20e498bd removes the pthread_mutex_rdlock() calls, but not the corresponding pthread_mutex_unlock() calls. Also, the database lock is never unlocked in one branch of the mempool_alloc() if. I think unreproducible random assert(dh->usable) crashes in prune_cache() were caused by this. But an easy way to make nscd threads hang with the broken locking was.
Diffstat (limited to 'nscd/mem.c')
| -rw-r--r-- | nscd/mem.c | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/nscd/mem.c b/nscd/mem.c index fcea6dbd03..80ea951146 100644 --- a/nscd/mem.c +++ b/nscd/mem.c @@ -566,9 +566,6 @@ mempool_alloc (struct database_dyn *db, size_t len, int data_alloc) } } - if (data_alloc) - pthread_rwlock_unlock (&db->lock); - if (! db->last_alloc_failed) { dbg_log (_("no more memory for database '%s'"), dbnames[db - dbs]); @@ -591,5 +588,8 @@ mempool_alloc (struct database_dyn *db, size_t len, int data_alloc) pthread_mutex_unlock (&db->memlock); + if (data_alloc) + pthread_rwlock_unlock (&db->lock); + return res; } |