diff options
| author | Nicholas Piggin <npiggin@gmail.com> | 2023-05-13 22:12:23 +1000 |
|---|---|---|
| committer | Reza Arbab <arbab@linux.ibm.com> | 2023-06-06 13:32:12 -0500 |
| commit | bd0595f7f3827d2a402819c9b215a9ceba254972 (patch) | |
| tree | 27e31e9b767a92d25c92ec88b578cadd6b41429c | |
| parent | 1e7dd419df98a948506c918d66007a05a7aebce0 (diff) | |
| download | skiboot-bd0595f7f3827d2a402819c9b215a9ceba254972.zip skiboot-bd0595f7f3827d2a402819c9b215a9ceba254972.tar.gz skiboot-bd0595f7f3827d2a402819c9b215a9ceba254972.tar.bz2 | |
ipmi: Fix potential infinite loop in sync_msg polling
Current gcc with -Os happens to generate code that re-loads the variable
in the loop, but that could change without notice, and with -O2 it does
infinite loop if sync_msg is !NULL, because it is not declared volatile
and there is no compiler barrier in the loop.
Add the usual cpu_relax() there to provide the compiler barrier.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Stewart Smith <stewart@flamingspork.com>
Signed-off-by: Reza Arbab <arbab@linux.ibm.com>
| -rw-r--r-- | core/ipmi.c | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/core/ipmi.c b/core/ipmi.c index bbc1a7b..59aa95f 100644 --- a/core/ipmi.c +++ b/core/ipmi.c @@ -170,7 +170,8 @@ void ipmi_queue_msg_sync(struct ipmi_msg *msg) } lock(&sync_lock); - while (sync_msg); + while (sync_msg) + cpu_relax(); sync_msg = msg; if (msg->backend->disable_retry && !opal_booting()) msg->backend->disable_retry(msg); |