[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <260d8ab9-c986-9dab-a447-ebb55df302c9@amd.com>
Date: Wed, 29 Oct 2025 12:17:33 -0700
From: Lizhi Hou <lizhi.hou@....com>
To: Markus Elfring <Markus.Elfring@....de>, <dri-devel@...ts.freedesktop.org>,
	Jeffrey Hugo <quic_jhugo@...cinc.com>, Maciej Falkowski
	<maciej.falkowski@...ux.intel.com>, Oded Gabbay <ogabbay@...nel.org>
CC: LKML <linux-kernel@...r.kernel.org>, Mario Limonciello
	<mario.limonciello@....com>, Max Zhen <max.zhen@....com>, Sonal Santan
	<sonal.santan@....com>
Subject: Re: accel/amdxdna: Fix incorrect command state for timed out job
On 10/29/25 11:54, Markus Elfring wrote:
>>> https://elixir.bootlin.com/linux/v6.18-rc3/C/ident/amdxdna_cmd_set_state
>> Sure. amdxdna_cmd_set_state() updates the return code to command buffer. So application which issues the command will be able to get the return code.
>>
>> The function return value "ret" is used by mailbox receiving kernel thread to deal with the error.
> I miss a clearer answer for the indicated function call incidence.
>
> Can it be helpful to determine the state value before it would be passed to a concrete call?
aie2_sched_resp_handler() is called either after get the firmware 
response through mailbox or the request is timed out/ canceled. So in 
this handler, it based on the response to set the state field in command 
buffer.
What do you mean here for "determine the state value before it would be 
passed to a concrete call?". What is your concern here? Maybe you can 
provide a simple patch if you think there is anything can be improved?
Thanks,
Lizhi
>
> Regards,
> Markus
Powered by blists - more mailing lists
 
