Module: xenomai-forge
Branch: master
Commit: 9dff85814cd6a4d3f8109a66ca841330bdad413d
URL:
http://git.xenomai.org/?p=xenomai-forge.git;a=commit;h=9dff85814cd6a4d3f8109a66ca841330bdad413d
Author: Philippe Gerum <r...@xenomai.org>
Date: Tue Nov 22 10:39:22 2011 +0100
alchemy/buffer: fix deadlock handling on the write side
---
lib/alchemy/buffer.c | 18 ++++++++++++++++++
1 files changed, 18 insertions(+), 0 deletions(-)
diff --git a/lib/alchemy/buffer.c b/lib/alchemy/buffer.c
index 07f34f8..a246925 100644
--- a/lib/alchemy/buffer.c
+++ b/lib/alchemy/buffer.c
@@ -381,6 +381,24 @@ ssize_t rt_buffer_write_until(RT_BUFFER *bf,
}
wait->size = len;
+ /*
+ * Check whether readers are already waiting for
+ * receiving data, while we are about to wait for
+ * sending some. In such a case, we have the converse
+ * pathological use of the buffer. We must kick
+ * readers to allow for a short read to prevent a
+ * deadlock.
+ *
+ * XXX: instead of broadcasting a general wake up
+ * event, we could be smarter and wake up only the
+ * number of waiters required to consume the amount of
+ * data we want to send, but this does not seem worth
+ * the burden: this is an error condition, we just
+ * have to mitigate its effect, avoiding a deadlock.
+ */
+ if (bcb->fillsz > 0 && syncobj_pend_count(&bcb->sobj))
+ syncobj_flush(&bcb->sobj, SYNCOBJ_BROADCAST);
+
ret = syncobj_wait_drain(&bcb->sobj, timespec, &syns);
if (ret) {
if (ret == -EIDRM)
_______________________________________________
Xenomai-git mailing list
Xenomai-git@gna.org
https://mail.gna.org/listinfo/xenomai-git
