[PATCH net] pds_core: keep the health thread stopped during reset

[PATCH net] pds_core: keep the health thread stopped during reset

[Nikhil P. Rao]

Commit d9407ff11809 ("pds_core: Prevent health thread from running
during reset/remove") stops the health thread with cancel_work_sync()
before a reset, but a devcmd timeout during pdsc_fw_down() re-queues
health_work, so pdsc_health_thread() runs again mid-reset and double
allocates the core DMA queues via pdsc_fw_up().

Only the reset path is affected. On remove, PDSC_S_STOPPING_DRIVER gates
the health thread.

Track the health thread's stopped state so a work item queued after the
cancel bails out.

Fixes: d9407ff11809 ("pds_core: Prevent health thread from running during reset/remove")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Closes: https://sashiko.dev/#/patchset/20260629200358.2626129-1-nikhil.rao%40amd.com?part=2
Signed-off-by: Nikhil P. Rao <nikhil.rao@amd.com>
---
 drivers/net/ethernet/amd/pds_core/core.c | 3 +++
 drivers/net/ethernet/amd/pds_core/core.h | 1 +
 drivers/net/ethernet/amd/pds_core/main.c | 2 ++
 3 files changed, 6 insertions(+)

diff --git a/drivers/net/ethernet/amd/pds_core/core.c b/drivers/net/ethernet/amd/pds_core/core.c
index 38a2446571af..c4cba7194da0 100644
--- a/drivers/net/ethernet/amd/pds_core/core.c
+++ b/drivers/net/ethernet/amd/pds_core/core.c
@@ -635,6 +635,9 @@ void pdsc_health_thread(struct work_struct *work)
 	unsigned long mask;
 	bool healthy;
 
+	if (READ_ONCE(pdsc->health_stopped))
+		return;
+
 	mutex_lock(&pdsc->config_lock);
 
 	/* Don't do a check when in a transition state */
diff --git a/drivers/net/ethernet/amd/pds_core/core.h b/drivers/net/ethernet/amd/pds_core/core.h
index b7fe9ad73349..a1d41329209f 100644
--- a/drivers/net/ethernet/amd/pds_core/core.h
+++ b/drivers/net/ethernet/amd/pds_core/core.h
@@ -171,6 +171,7 @@ struct pdsc {
 	struct timer_list wdtimer;
 	unsigned int wdtimer_period;
 	struct work_struct health_work;
+	bool health_stopped;
 	struct devlink_health_reporter *fw_reporter;
 	u32 fw_recoveries;
 
diff --git a/drivers/net/ethernet/amd/pds_core/main.c b/drivers/net/ethernet/amd/pds_core/main.c
index 22db78343eb0..32e1d7069969 100644
--- a/drivers/net/ethernet/amd/pds_core/main.c
+++ b/drivers/net/ethernet/amd/pds_core/main.c
@@ -463,6 +463,7 @@ static void pdsc_stop_health_thread(struct pdsc *pdsc)
 	if (pdsc->pdev->is_virtfn)
 		return;
 
+	WRITE_ONCE(pdsc->health_stopped, true);
 	timer_shutdown_sync(&pdsc->wdtimer);
 	if (pdsc->health_work.func)
 		cancel_work_sync(&pdsc->health_work);
@@ -473,6 +474,7 @@ static void pdsc_restart_health_thread(struct pdsc *pdsc)
 	if (pdsc->pdev->is_virtfn)
 		return;
 
+	WRITE_ONCE(pdsc->health_stopped, false);
 	timer_setup(&pdsc->wdtimer, pdsc_wdtimer_cb, 0);
 	mod_timer(&pdsc->wdtimer, jiffies + 1);
 }
-- 
2.43.0

Re: [PATCH net] pds_core: keep the health thread stopped during reset

[Jakub Kicinski]

On Tue, 21 Jul 2026 16:21:41 +0000 Nikhil P. Rao wrote:
> Commit d9407ff11809 ("pds_core: Prevent health thread from running
> during reset/remove") stops the health thread with cancel_work_sync()
> before a reset, but a devcmd timeout during pdsc_fw_down() re-queues
> health_work, so pdsc_health_thread() runs again mid-reset and double
> allocates the core DMA queues via pdsc_fw_up().

Would using disable_work instead of cancel_work fix this without
the extra state? The patch reads a bit racy, IIUC the cancel work
can't be trusted, which in turn puts in question calling it in the
first place.