Received-SPF: pass (google.com: domain of linux-kernel+bounces-99241-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Message-ID: <06e02957-4cfe-4dba-a7a3-fa08e7a199c3@amd.com>
Date: Mon, 11 Mar 2024 17:27:40 +0100
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] drm/amdgpu: add ring buffer information in devcoredump
Content-Language: en-US
To: "Khatri, Sunil" <sukhatri@amd.com>, Sunil Khatri <sunil.khatri@amd.com>,
 Alex Deucher <alexander.deucher@amd.com>,
 Shashank Sharma <shashank.sharma@amd.com>
Cc: amd-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org,
 linux-kernel@vger.kernel.org
References: <20240311122212.13713-1-sunil.khatri@amd.com>
 <9e1dbcbc-f00e-417c-995e-d9c539292e03@amd.com>
 <34ed7275-e1a5-4d93-8ba1-ef3c597eaef3@amd.com>
From: =?UTF-8?Q?Christian_K=C3=B6nig?= <christian.koenig@amd.com>
In-Reply-To: <34ed7275-e1a5-4d93-8ba1-ef3c597eaef3@amd.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk
MIME-Version: 1.0
X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1
X-MS-Exchange-AntiSpam-MessageData-0:
	=?utf-8?B?b012eXV5blQrS1dNTTdOTHZOV2d4aGZLU0hvTjVtMnl5ZlkyUjVmMDdTNlhU?=
 =?utf-8?B?Q2VaRHFwbGIySHhtc0hDVUVVdzVUVFZSTVU0WWVzbVBwL3NxQzlyTlZ4ZHZO?=
 =?utf-8?B?Tm9mR1EyY3ZyTzFWcFQvejYzVWVOSEFJTnJwRnRQVlM1OC9XMExFSzF0aHlC?=
 =?utf-8?B?aXRjRm13QXFOSXNTK2dDdVNjRHFLSEtvOWFPeTB1WXZLTXZ5YWM3Ry9Ldi93?=
 =?utf-8?B?SkF4bzhWdXNvaUc2enJ1aU9Jei9ERnFVZEF5Q3RIYzZ1bHlidjRicnZsdXk1?=
 =?utf-8?B?WWRBVGlMNytDd1VheURXaWN2VFoxL3dWeFg0VWZ4aFBxZ3lqckJCMk51NW9h?=
 =?utf-8?B?MXhjVDltUndpMk9qM2ZHMjJZbmJnYWJJNVV5aGVrUVNFajl6WEp5RlJSWU1l?=
 =?utf-8?B?cCtEWWtJallHa2Z3VTJiUlpubVIvWVpSYnlLL1RqdkxNdXFObkNGbkE1MGFH?=
 =?utf-8?B?aGp4WFY2KzdLTnppcTVjTmxISlJwOUo1MjFCU05zRGc5YTY5QUs2QmFIZWhK?=
 =?utf-8?B?TDBFWVNSRUVQNjg0R0Y1d2FTdWlTT1piWWFUTDBUSTAvWXJna2VGU2YwbUdI?=
 =?utf-8?B?ajlMZi9xMFJoY3p5TnM3bVgxQVlWa3QzRTJIa3J4N3pDR29UR2JFWjVvODEw?=
 =?utf-8?B?VU1jeHdrS2IrVGY4bjFtQzNMWjNUOFdTTzRpTVhEV0tzejkrV1p5R2xaUzVq?=
 =?utf-8?B?SUM0aFNjUHBzd1NaZjdtMURVdUZkRzhHNG1lNmlIbTE4TFZHWmtMSzNyenZN?=
 =?utf-8?B?UlpWbVRUV2Zkamg4WW5LM1pDWnhKbVc5Z0hHVjVUWlNVUXk2dW1meERkTk9t?=
 =?utf-8?B?OUZLazFlSWJXZTZEUk9FMVkwWXo3RzhsZnlNQUdCYkh5TkdrR1cvTVZzei8v?=
 =?utf-8?B?WER3TWN0dThEd2M1SVFHaHc5L0g3aW91U3dKSzFncHZFYlVWVW9hMG1wc1Bp?=
 =?utf-8?B?aWJQWjNkZkMrUDZaRXZUd3YvQ0xhM2E0NUdaUTJzd3p5LzBjRVlvdFFPUzlN?=
 =?utf-8?B?THI4VE5sNlVna0JsMVdoeFNXVWVLZElYVEs1dWU0OUxkMjlYWk1rNENWSk9q?=
 =?utf-8?B?QXA0S2VCUFFJUzBLRThWVUVWRlJ3bWsvSm9ialdtYTJmY1Z1WVBrL0FwMDVE?=
 =?utf-8?B?alo0NGVadzVzWTBUMDdJTEw0RjFsYks5S3B3U3lIekIrQnpoZ3M0TWEvdEly?=
 =?utf-8?B?QjdIa3RPdUJXRTFTOWZHTjdMeWZXT2swQVlkU09CZTZYd21aRUkxVWxMVmNZ?=
 =?utf-8?B?K2Z2cGQ0UDBBWmZiMU1ML0d1bkF3cmhucGJrcERSQUZOVW9xY0ZvRTArdUlr?=
 =?utf-8?B?MUhMMHVLOTE4a1FnVEFnd002Q09DNkYyZ2VrRGR5UE9mWlUrTXVRMGovK3ZC?=
 =?utf-8?B?YSsrWXA0L2Z2RUhneFFsNjNuWmlQZ0JmTzRXUnBuVHRMTmlTajlXVE9wR1A0?=
 =?utf-8?B?RjIrM1ZMMERicFZ1VXEwMHQwMU1iVG02RGUzM1JIZUdPVTRaOXdPSWQ4Q01D?=
 =?utf-8?B?VGF3Qkc4YkRocmVEUWZxRC9IYXVYNGVPUUN5QVgxbHludXUra2FrQkJJK1JO?=
 =?utf-8?B?UGNrNi8yYWxpTVJtOU1HOHZzY2VQRXM1VTZJZGhuMjNxdmdITDVDT1dTRXRq?=
 =?utf-8?B?TW5DRHRTNHRkclc1dklPam5nSjdBUGlqaFlFdmVSb0pkSzNvVm1qMXZBUWFt?=
 =?utf-8?B?OUl2bTRad0NJSlNPQWM0em0xbkM2T2xvVlJBT3VRTnFWRjhhOTg5bDEzNGZn?=
 =?utf-8?B?TVRIK085VHVwa09ZQ3lya2VCNFRHUytUZWw3MVoxaTRrMGFRM3FBQ25PTU9h?=
 =?utf-8?B?Y1FZYXRTbWJENE53eTl6YndhejR2R3M1ZGVnWkpGdEY2TDhGakZUQjl2SDVX?=
 =?utf-8?B?SklMTkRXa2QvYVRaSlJWT0R0aEo0Y3gyL0g0YnliczlBVzRUVFd0cmh6UGo3?=
 =?utf-8?B?WitNd0M5MkxxejAycTB3VHlnTjF4NDRKVlZMWmkyTm54V21CbVVxNExnTks0?=
 =?utf-8?B?b3M3SW1HSDgyRDlWT1lvZkVmVzIzUzdqZG5tTjZha1ZUd25EQUhZbUFRS29Z?=
 =?utf-8?B?SDgzRG9sWko1T3VkVnRpUUtLM3FVTTJIQWp4SDBmQ2NvQXI1Qk91MnFWMmtm?=
 =?utf-8?Q?2t6c=3D?=
X-OriginatorOrg: amd.com
X-MS-Exchange-CrossTenant-Network-Message-Id: 51be8f33-a4a4-489b-2d14-08dc41e82f9b
X-MS-Exchange-CrossTenant-AuthSource: PH7PR12MB5685.namprd12.prod.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 11 Mar 2024 16:27:46.9611
 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-CrossTenant-Id: 3dd8961f-e488-4e60-8e11-a82d994e183d
X-MS-Exchange-CrossTenant-MailboxType: HOSTED
X-MS-Exchange-CrossTenant-UserPrincipalName: DlzUIm5+yl/bPasZAq1ZRuKGo13IG0kSojGm+Jr5IltHYgszPmwcKwZ/phbcnBZE
X-MS-Exchange-Transport-CrossTenantHeadersStamped: DS0PR12MB7899

Am 11.03.24 um 15:48 schrieb Khatri, Sunil:
>
> On 3/11/2024 7:29 PM, Christian König wrote:
>>
>>
>> Am 11.03.24 um 13:22 schrieb Sunil Khatri:
>>> Add relevant ringbuffer information such as
>>> rptr, wptr, ring name, ring size and also
>>> the ring contents for each ring on a gpu reset.
>>>
>>> Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
>>> ---
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c | 21 +++++++++++++++++++++
>>>   1 file changed, 21 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
>>> index 6d059f853adc..1992760039da 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
>>> @@ -215,6 +215,27 @@ amdgpu_devcoredump_read(char *buffer, loff_t 
>>> offset, size_t count,
>>>                  fault_info->status);
>>>       }
>>>   +    drm_printf(&p, "Ring buffer information\n");
>>> +    for (int i = 0; i < coredump->adev->num_rings; i++) {
>>> +        int j = 0;
>>> +        struct amdgpu_ring *ring = coredump->adev->rings[i];
>>> +
>>> +        drm_printf(&p, "ring name: %s\n", ring->name);
>>> +        drm_printf(&p, "Rptr: 0x%llx Wptr: 0x%llx\n",
>>> +               amdgpu_ring_get_rptr(ring) & ring->buf_mask,
>>> +               amdgpu_ring_get_wptr(ring) & ring->buf_mask);
>>
>> Don't apply the mask here. We do have some use cases where the rptr 
>> and wptr are outside the ring buffer.
> Sure i will remove the mask.
>>
>>> +        drm_printf(&p, "Ring size in dwords: %d\n",
>>> +               ring->ring_size / 4);
>>
>> Rather print the mask as additional value here.
> Does that help adding the mask value ?

I think it should help as a reminder that rptr & wptr needs to be masked 
to become valid indexes.

Some hw generations have really crude workarounds where we have to 
allocate an extra page after the ring buffer because the hw is buddy and 
sometimes tries to read command from there as well.

So when we see a hang with some rptr and wptr values which don't fit 
into the mask we will know that the hw has another issue in that direction.

Regards,
Christian.

>>
>>> +        drm_printf(&p, "Ring contents\n");
>>> +        drm_printf(&p, "Offset \t Value\n");
>>> +
>>> +        while (j < ring->ring_size) {
>>> +            drm_printf(&p, "0x%x \t 0x%x\n", j, ring->ring[j/4]);
>>> +            j += 4;
>>> +        }
>>
>>> +        drm_printf(&p, "Ring dumped\n");
>>
>> That seems superfluous.
>
> Noted
>
>
> Regards
> Sunil
>
>>
>> Regards,
>> Christian.
>>
>>> +    }
>>> +
>>>       if (coredump->reset_vram_lost)
>>>           drm_printf(&p, "VRAM is lost due to GPU reset!\n");
>>>       if (coredump->adev->reset_info.num_regs) {
>>