Re: failed to remove key (0, ce:ce:1e:27:bb:e0) from hardware (-110) (ETIMEDOUT)

From: James Prestwood
Date: Fri Oct 11 2024 - 12:36:43 EST


Hi Paul,

On 10/11/24 5:48 AM, Paul Menzel wrote:
Dear James,


Am 04.09.24 um 16:09 schrieb James Prestwood:

On 9/4/24 3:45 AM, Paul Menzel wrote:

Linux 6.11-rc6+ logged the warning below when resuming from ACPI S3 (or unloading and loading the `ath10k_core`/`ath10k_pci` modules) having been connected to an AVM network:

    wlp58s0: failed to remove key (0, ce:ce:1e:27:bb:e0) from hardware (-110)

Error code 110 is the value for ETIMEDOUT. I saw James patch [1], and applied it, and the error is still there (as expected).

My patch won't actually fix the timeout, I just lowered the time that ath10k would wait before it continued which fixed some incompatibility on the AP side of things. Based on your logs though, it appears you already got disconnected before the failure to remove the key...

I am still applying your patch manually on the current master branch, and it looks like it hasn’t still been reviewed. Maybe it’s due to the RFC tag. Do you plan on moving it ahead?

There was another thread, which I cant seem to find where Kalle had mentioned the RFC patch. IIRC Kalle had said the proper fix is in the firmware, and would rather not take this patch upstream if we can get the firmware addressed. And I would have to agree this is the best approach if this is an option. This RFC patch was merely a workaround for some odd behavior with Cisco APs that one of our customers was using, where the APs would actually reject roams if they took too long, hence why reducing the timeout "fixed" the problem.

Last I checked your situation was different, your device got disconnected _then_ failed to remove the key. If this is the case my RFC patch isn't going to really change much, apart from fail in 1 second vs 3 seconds.

Thanks,

James



Kind regards,

Paul


[1]: https://lore.kernel.org/all/20240814164507.996303-1-prestwoj@xxxxxxxxx/