Skip to contentSkip to conversation

native: Properly update lockcnt on recursive call from secondary mode

PATCH New Jan Kiszka 1 participant 1 message Nov 5, 2009

lore
Share & tools
lore

Jan Kiszka

author
#0 lore raw

PATCH 1/1 native: Properly update lockcnt on recursive call from secondary mode

We cannot assume that only non-recursive rt_mutex_acquire calls will happen; the application may be buggy and come from secondary mode. Make sure to update the user space recursion counter correctly in any case so that at least the mutex is kept in a valid state.

---

Can be pulled from 'for-upstream' as well.

 src/skins/native/mutex.c |   14 ++++++++++++++
 1 files changed, 14 insertions(+), 0 deletions(-)
src/skins/native/mutex.c +14 -0
index 29843c7..98844ef 100644
@@ -96,6 +96,20 @@ static int rt_mutex_acquire_inner(RT_MUTEX *mutex, RTIME timeout, xntmode_t mode
 
 		if (timeout == TM_NONBLOCK && mode == XN_RELATIVE)
 			return -EWOULDBLOCK;
+	} else if (xnsynch_fast_owner_check(mutex->fastlock, cur) == 0) {
+		/*
+		 * The application is buggy as it jumped to secondary mode
+		 * while holding the mutex. Nevertheless, we have to keep the
+		 * mutex state consistent.
+		 *
+		 * We make no efforts to migrate or warn here. There is
+		 * CONFIG_XENO_OPT_DEBUG_SYNCH_RELAX to catch such bugs.
+		 */
+		if (mutex->lockcnt == UINT_MAX)
+			return -EAGAIN;
+
+		mutex->lockcnt++;
+		return 0;
 	}
 #endif /* CONFIG_XENO_FASTSYNCH */
 
Signature · 1 line
1.6.0.2