Re: [Intel-gfx] [PATCH v2 4/5] drm/i915/guc: Add GuC CT specific debug print wrappers

[John Harrison <john.c.harrison@intel.com>]

On 1/9/2023 01:39, Tvrtko Ursulin wrote:
> On 06/01/2023 18:57, John Harrison wrote:
>> On 12/6/2022 03:06, Tvrtko Ursulin wrote:
>>> On 05/12/2022 18:44, Michal Wajdeczko wrote:
>>>> On 05.12.2022 14:16, Tvrtko Ursulin wrote:
>>>>> On 02/12/2022 20:14, John Harrison wrote:
>>>>> [snip]
>>>>>
>>>>>> Random meaningless (to me) message that is apparently a display 
>>>>>> thing:
>>>>>> drm_dbg_kms(&dev_priv->drm, "disabling %s\n", pll->info->name);
>>>>>> i915 0000:00:02.0: [drm:intel_disable_shared_dpll [i915]] disabling
>>>>>> PORT PLL B
>>>>>
>>>>> Plan is to not touch outside gt/.
>> For some unexplicable reason that means it is almost impossible to 
>> see the actual problems in most CI dmesg logs because they are 
>> swamped with irrelevant display messages that cannot be filtered out. 
>> For example, I recently manually grep'd out all the display spam from 
>> a bug report log. The dmesg file went from 12MB to 700KB. That is a 
>> significant problem that makes bug triage way harder than it needs to 
>> be.
>
> I didn't get this part, how it would reduce the amount of spam by 
> adding new macros? Anyway, that's something to split out and discuss 
> with display folks.
It will allow someone to trivially filter out everything with that tag. 
Which then makes it orders of magnitude easy to scan through the log to 
see what happened.

>
>>> Maybe as a way forward work could be split? If John wants to deal 
>>> with gt_xxx macros, avoid touching GuC (putting his original 
>>> motivation aside) and you want to convert the gt/uc folder? Assuming 
>>> John you are okay with "GuC:" and "CT:" prefixes.
>> Meaning just repost patch #1 only and expand to more intel_gt_* 
>> files? Sure, if someone will actually reply to that patch with some 
>> kind of r-b first so I know I'm not still wasting my time on a huge 
>> re-write that will to be redone multiple times when someone objects 
>> to the use of a colon or the lack of spaces, braces or whatever.
>
> First patch looks good to me (ack in principle) apart that Michal 
> found one potential null pointer dereference if I understood it right. 
> That other comment about the ratelimited call is maybe okay to leave 
> for later, *if* it will be a single instance, otherwise needs a gt 
> logger as well. I can r-b once you re-send with the first issue fixed.
I've already fixed those two issues locally. I'm not going to touch the 
TRACE macros.

Okay. I'll extend it further and repost.

John.


>
> Regards,
>
> Tvrtko