Re: [PATCH net-next 1/5] ionic: Small improvements in devcmd retry logic

[Eric Joyner <eric.joyner@amd.com>]

On 5/1/2026 4:37 PM, Jakub Kicinski wrote:
> Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.
> 
> 
> On Thu, 30 Apr 2026 20:15:51 -0700 Eric Joyner wrote:
>> If the timeout time is hit when the last attempt returned EAGAIN, the
>> driver returns -ETIMEDOUT. This causes the -EAGAIN result to be lost.
>> Fix this by returning -EAGAIN if the timeout time is hit and the
>> previous result matches.
>>
>> Also, reduce the sleep between the write to done and doorbell
>> registers. The msleep(1000) was initially added in an arbitrary
>> manner. However, this long of a sleep is problematic because
>> it reduces the number of retries when -EAGAIN is returned, which
>> may result in the devmcd giving up early due to the timeout. Fix
>> this by reducing the sleep to msleep(50).
> 
> It would be useful to explain the significance of the return value
> to the callers. Are you trying to return EAGAIN to user space?
> Or some kernel-internal function catches this?

It looks like the -EAGAIN return value used here would have been used by
"[PATCH net 1/7] ionic: Allow the first devcmd to trigger deferred probe" [1];
this change got put in this series because we decided it was more of a net-next
change rather than a net change, but didn't realize these two were more
connected than what we saw at first glance.

1: https://lore.kernel.org/netdev/20260429210007.40015-2-eric.joyner@amd.com/

 - Eric