[RFC PATCH] thunderbolt: icm: report switch runtime resume timeout

[RFC PATCH] thunderbolt: icm: report switch runtime resume timeout

[Pengpeng Hou]

icm_runtime_resume_switch() waits for firmware to signal switch runtime-
resume completion. If the wait expires, it only prints a debug message
and still returns success to the runtime PM path.

Return -ETIMEDOUT when the completion does not arrive so runtime PM
callers can observe the failed switch resume instead of using the device
as if resume had completed.

Signed-off-by: Pengpeng Hou <pengpeng@iscas.ac.cn>
---
 drivers/thunderbolt/icm.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/thunderbolt/icm.c b/drivers/thunderbolt/icm.c
index 10fefac3b1d9..46d491627970 100644
--- a/drivers/thunderbolt/icm.c
+++ b/drivers/thunderbolt/icm.c
@@ -2179,6 +2179,7 @@ static int icm_runtime_resume_switch(struct tb_switch *sw)
 		if (!wait_for_completion_timeout(&sw->rpm_complete,
 						 msecs_to_jiffies(500))) {
 			dev_dbg(&sw->dev, "runtime resuming timed out\n");
+			return -ETIMEDOUT;
 		}
 	}
 	return 0;
-- 
2.50.1 (Apple Git-155)

Re: [RFC PATCH] thunderbolt: icm: report switch runtime resume timeout

[Mika Westerberg]

Hi,

On Tue, Jun 23, 2026 at 02:18:11PM +0800, Pengpeng Hou wrote:
> icm_runtime_resume_switch() waits for firmware to signal switch runtime-
> resume completion. If the wait expires, it only prints a debug message
> and still returns success to the runtime PM path.

IIRC that's intentional. Otherwise it would have logged it as warn or so.
This can happen if the device is removed while we were runtime suspended.

Do you see any problem with this?