mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Shashank Sharma <shashank.sharma@amd.com>
To: Felix Kuehling <felix.kuehling@amd.com>,
	"Joshi, Mukul" <Mukul.Joshi@amd.com>,
	"Yadav, Arvind" <Arvind.Yadav@amd.com>,
	"Koenig, Christian" <Christian.Koenig@amd.com>,
	"Deucher, Alexander" <Alexander.Deucher@amd.com>,
	"Pan, Xinhui" <Xinhui.Pan@amd.com>,
	"airlied@gmail.com" <airlied@gmail.com>,
	"daniel@ffwll.ch" <daniel@ffwll.ch>
Cc: "dri-devel@lists.freedesktop.org"
	<dri-devel@lists.freedesktop.org>,
	"amd-gfx@lists.freedesktop.org" <amd-gfx@lists.freedesktop.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2 1/1] drm/amdkfd: Fix unaligned doorbell absolute offset for gfx8
Date: Fri, 29 Sep 2023 10:39:28 +0200	[thread overview]
Message-ID: <e33eb526-c0b7-c404-72d2-390ce7fb0108@amd.com> (raw)
In-Reply-To: <2f56333d-4549-3118-5cd0-3ad2be025b66@amd.com>


On 28/09/2023 20:53, Felix Kuehling wrote:
> On 2023-09-28 11:38, Shashank Sharma wrote:
>> Hello Felix, Mukul,
>>
>> On 28/09/2023 17:30, Felix Kuehling wrote:
>>> On 2023-09-28 10:30, Joshi, Mukul wrote:
>>>> [AMD Official Use Only - General]
>>>>
>>>>> -----Original Message-----
>>>>> From: Yadav, Arvind <Arvind.Yadav@amd.com>
>>>>> Sent: Thursday, September 28, 2023 5:54 AM
>>>>> To: Koenig, Christian <Christian.Koenig@amd.com>; Deucher, Alexander
>>>>> <Alexander.Deucher@amd.com>; Sharma, Shashank
>>>>> <Shashank.Sharma@amd.com>; Kuehling, Felix <Felix.Kuehling@amd.com>;
>>>>> Joshi, Mukul <Mukul.Joshi@amd.com>; Pan, Xinhui <Xinhui.Pan@amd.com>;
>>>>> airlied@gmail.com; daniel@ffwll.ch
>>>>> Cc: amd-gfx@lists.freedesktop.org; 
>>>>> dri-devel@lists.freedesktop.org; linux-
>>>>> kernel@vger.kernel.org; Yadav, Arvind <Arvind.Yadav@amd.com>; Koenig,
>>>>> Christian <Christian.Koenig@amd.com>
>>>>> Subject: [PATCH v2 1/1] drm/amdkfd: Fix unaligned doorbell 
>>>>> absolute offset
>>>>> for gfx8
>>>>>
>>>>> This patch is to adjust the absolute doorbell offset against the 
>>>>> doorbell id
>>>>> considering the doorbell size of 32/64 bit.
>>>>>
>>>>> v2:
>>>>> - Addressed the review comment from Felix.
>>>>>
>>>>> Cc: Christian Koenig <christian.koenig@amd.com>
>>>>> Cc: Alex Deucher <alexander.deucher@amd.com>
>>>>> Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
>>>>> Signed-off-by: Arvind Yadav <Arvind.Yadav@amd.com>
>>>>> ---
>>>>>   drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 9 ++++++++-
>>>>>   1 file changed, 8 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
>>>>> b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
>>>>> index 0d3d538b64eb..c54c4392d26e 100644
>>>>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
>>>>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
>>>>> @@ -407,7 +407,14 @@ static int allocate_doorbell(struct
>>>>> qcm_process_device *qpd,
>>>>>
>>>>>        q->properties.doorbell_off = amdgpu_doorbell_index_on_bar(dev-
>>>>>> adev,
>>>>>                                                                  qpd-
>>>>>> proc_doorbells,
>>>>> -                                                               q-
>>>>>> doorbell_id);
>>>>> +                                                               0);
>>>>> +
>>>> It looks like amdgpu_doorbell_index_on_bar() works only for 64-bit 
>>>> doorbells.
>>>> Shouldn't it work for both 32-bit and 64-bit doorbells considering 
>>>> this is common
>>>> doorbell manager code?
>>
>>
>> Yes, You are right that the calculations to find a particular 
>> doorbell in the doorbell page considers a doorbell width of 64-bit.
>>
>>>
>>> I could see this argument going either way. KFD is the only one that 
>>> cares about managing doorbells for user mode queues on GFXv8 GPUs. 
>>> This is not a use case that amdgpu cares about. So I'm OK with KFD 
>>> doing its own address calculations to make sure doorbells continue 
>>> to work on GFXv8.
>>>
>>> It may not be worth adding complexity to the common doorbell manager 
>>> code to support legacy GPUs with 32-bit doorbells.
>>
>>
>> I was thinking about adding an additional input parameter which will 
>> indicate if the doorbell width is 32-bit vs 64-bit (like 
>> is_doorbell_64_bit), and doorbell manager can alter the multiplier 
>> while calculating the final offset. Please let me know if that will 
>> work for both the cases.
>
> Yes, that would work for KFD because we already have the doorbell size 
> in our device-info structure. Instead of making it a boolean flag, you 
> could make it a doorbell_size parameter, in byte or dword units to 
> simplify the pointer math.


Sounds good, will do that and send an update.

- Shashank

>
> Regards,
>   Felix
>
>
>>
>> - Shashank
>>
>>>
>>>
>>> Regards,
>>>   Felix
>>>
>>>
>>>>
>>>> Thanks,
>>>> Mukul
>>>>
>>>>> +     /* Adjust the absolute doorbell offset against the doorbell id
>>>>> considering
>>>>> +      * the doorbell size of 32/64 bit.
>>>>> +      */
>>>>> +     q->properties.doorbell_off += q->doorbell_id *
>>>>> + dev->kfd->device_info.doorbell_size / 4;
>>>>> +
>>>>>        return 0;
>>>>>   }
>>>>>
>>>>> -- 
>>>>> 2.34.1

      reply	other threads:[~2023-09-29  8:40 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-09-28  9:54 [PATCH v2 0/1] " Arvind Yadav
2023-09-28  9:54 ` [PATCH v2 1/1] " Arvind Yadav
2023-09-28 14:30   ` Joshi, Mukul
2023-09-28 15:30     ` Felix Kuehling
2023-09-28 15:38       ` Shashank Sharma
2023-09-28 18:53         ` Felix Kuehling
2023-09-29  8:39           ` Shashank Sharma [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=e33eb526-c0b7-c404-72d2-390ce7fb0108@amd.com \
    --to=shashank.sharma@amd.com \
    --cc=Alexander.Deucher@amd.com \
    --cc=Arvind.Yadav@amd.com \
    --cc=Christian.Koenig@amd.com \
    --cc=Mukul.Joshi@amd.com \
    --cc=Xinhui.Pan@amd.com \
    --cc=airlied@gmail.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=daniel@ffwll.ch \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=felix.kuehling@amd.com \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®