2017-04-20 23:50:34

by Robert Lippert

Subject: [PATCH] ipmi/watchdog: fix wdog hang on panic waiting for ipmi response

Commit c49c097610fe ("ipmi: Don't call receive handler in the
panic context") means that the panic_recv_free is not called during a
panic and the atomic count does not drop to 0.

Fix this by only expecting one decrement of the atomic variable
which comes from panic_smi_free.

Signed-off-by: Robert Lippert <[email protected]>
---

drivers/char/ipmi/ipmi_watchdog.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/char/ipmi/ipmi_watchdog.c b/drivers/char/ipmi/ipmi_watchdog.c
index 5ca24d9b101b..d165af8abe36 100644
--- a/drivers/char/ipmi/ipmi_watchdog.c
+++ b/drivers/char/ipmi/ipmi_watchdog.c
@@ -516,7 +516,7 @@ static void panic_halt_ipmi_heartbeat(void)
msg.cmd = IPMI_WDOG_RESET_TIMER;
msg.data = NULL;
msg.data_len = 0;
- atomic_add(2, &panic_done_count);
+ atomic_add(1, &panic_done_count);
rv = ipmi_request_supply_msgs(watchdog_user,
(struct ipmi_addr *) &addr,
0,
@@ -526,7 +526,7 @@ static void panic_halt_ipmi_heartbeat(void)
&panic_halt_heartbeat_recv_msg,
1);
if (rv)
- atomic_sub(2, &panic_done_count);
+ atomic_sub(1, &panic_done_count);
}

static struct ipmi_smi_msg panic_halt_smi_msg = {
@@ -550,12 +550,12 @@ static void panic_halt_ipmi_set_timeout(void)
/* Wait for the messages to be free. */
while (atomic_read(&panic_done_count) != 0)
ipmi_poll_interface(watchdog_user);
- atomic_add(2, &panic_done_count);
+ atomic_add(1, &panic_done_count);
rv = i_ipmi_set_timeout(&panic_halt_smi_msg,
&panic_halt_recv_msg,
&send_heartbeat_now);
if (rv) {
- atomic_sub(2, &panic_done_count);
+ atomic_sub(1, &panic_done_count);
printk(KERN_WARNING PFX
"Unable to extend the watchdog timeout.");
} else {
--
2.12.2.816.g2cccc81164-goog

2017-04-28 19:53:18

by Corey Minyard

Subject: Re: [PATCH] ipmi/watchdog: fix wdog hang on panic waiting for ipmi response

On 04/20/2017 06:49 PM, Robert Lippert wrote:
> Commit c49c097610fe ("ipmi: Don't call receive handler in the
> panic context") means that the panic_recv_free is not called during a
> panic and the atomic count does not drop to 0.
>
> Fix this by only expecting one decrement of the atomic variable
> which comes from panic_smi_free.

Thanks, queued for the next kernel release.

-corey

> Signed-off-by: Robert Lippert <[email protected]>
> ---
>
> drivers/char/ipmi/ipmi_watchdog.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/char/ipmi/ipmi_watchdog.c b/drivers/char/ipmi/ipmi_watchdog.c
> index 5ca24d9b101b..d165af8abe36 100644
> --- a/drivers/char/ipmi/ipmi_watchdog.c
> +++ b/drivers/char/ipmi/ipmi_watchdog.c
> @@ -516,7 +516,7 @@ static void panic_halt_ipmi_heartbeat(void)
> msg.cmd = IPMI_WDOG_RESET_TIMER;
> msg.data = NULL;
> msg.data_len = 0;
> - atomic_add(2, &panic_done_count);
> + atomic_add(1, &panic_done_count);
> rv = ipmi_request_supply_msgs(watchdog_user,
> (struct ipmi_addr *) &addr,
> 0,
> @@ -526,7 +526,7 @@ static void panic_halt_ipmi_heartbeat(void)
> &panic_halt_heartbeat_recv_msg,
> 1);
> if (rv)
> - atomic_sub(2, &panic_done_count);
> + atomic_sub(1, &panic_done_count);
> }
>
> static struct ipmi_smi_msg panic_halt_smi_msg = {
> @@ -550,12 +550,12 @@ static void panic_halt_ipmi_set_timeout(void)
> /* Wait for the messages to be free. */
> while (atomic_read(&panic_done_count) != 0)
> ipmi_poll_interface(watchdog_user);
> - atomic_add(2, &panic_done_count);
> + atomic_add(1, &panic_done_count);
> rv = i_ipmi_set_timeout(&panic_halt_smi_msg,
> &panic_halt_recv_msg,
> &send_heartbeat_now);
> if (rv) {
> - atomic_sub(2, &panic_done_count);
> + atomic_sub(1, &panic_done_count);
> printk(KERN_WARNING PFX
> "Unable to extend the watchdog timeout.");
> } else {