https://issues.apache.org/bugzilla/show_bug.cgi?id=53690
Priority: P3
Bug ID: 53690
Keywords: PatchAvailable
Assignee: [email protected]
Summary: Calls to semctl() trigger "uninitialized memory"
warning from valgrind
Severity: normal
Classification: Unclassified
OS: Linux
Reporter: [email protected]
Hardware: All
Status: NEW
Version: 2.4.2
Component: Core
Product: Apache httpd-2
The warnings looks like this:
==977== Syscall param semctl(IPC_SET, arg.buf) points to uninitialised byte(s)
==977== at 0x62FA8EA: semctl (in /lib64/libc-2.5.so)
==977== by 0x18705C: ap_unixd_set_proc_mutex_perms (unixd.c:251)
==977== by 0x1870E0: ap_unixd_set_global_mutex_perms (unixd.c:284)
==977== by 0x16FC19: ap_global_mutex_create (util_mutex.c:444)
==977== by 0xAE3D17A: rewritelock_create (mod_rewrite.c:2594)
==977== by 0xAE41730: post_config (mod_rewrite.c:4316)
==977== by 0x15F4B3: ap_run_post_config (config.c:105)
==977== by 0x1365C5: main (main.c:696)
==977== Address 0x7ff000430 is on thread 1's stack
==977== Uninitialised value was created by a stack allocation
==977== at 0x186FB1: ap_unixd_set_proc_mutex_perms (unixd.c:227)
Because the semid_ds structure on Linux (as well as FreeBSD and, likely, other
Unixes) contains some undocumented fields (such as __unused1 or sem_pad2),
explicitly setting each one is not portable -- far simpler to just request,
that the entire structure be zeroed at the declaration time:
--- os/unix/unixd.c 2011-12-18 13:02:21.000000000 -0500
+++ os/unix/unixd.c 2012-08-09 18:40:55.000000000 -0400
@@ -242,5 +242,5 @@
#endif
union semun ick;
- struct semid_ds buf;
+ struct semid_ds buf = { 0 };
apr_os_proc_mutex_get(&ospmutex, pmutex);
While the warning is benign, it is better to suppress it, then to needlessly
worry people attempting to use valgrind to debug some other problem -- the
fewer such false alarms, the better.
Earlier releases of httpd have the same issue (I've seen it in 2.2.22 myself).
--
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
