All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH] drm/i915: Fix workarounds on Gen2-3
Date: Fri, 18 Nov 2022 21:58:42 +0200	[thread overview]
Message-ID: <Y3fj8rJpqMhzsW35@intel.com> (raw)
In-Reply-To: <20221118115249.2683946-1-tvrtko.ursulin@linux.intel.com>

On Fri, Nov 18, 2022 at 11:52:49AM +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> 
> In 3653727560d0 ("drm/i915: Simplify internal helper function signature")
> I broke the old platforms by not noticing engine workaround init does not
> initialize the list on old platforms. Fix it by always initializing which
> already does the right thing by mostly not doing anything if there aren't
> any workarounds on the list.

Was going to give this a quick smoke test on my 865 but I can't even
reproduce the original issue on it.

Turns out the 64bit compiler is too smart:
0000000000000000 <wa_list_apply>:
       0:       8b 77 20                mov    0x20(%rdi),%esi
       3:       85 f6                   test   %esi,%esi
       5:       75 01                   jne    8 <wa_list_apply+0x8>
       7:       c3                      ret
       8:       41 57                   push   %r15
       a:       41 56                   push   %r14
       c:       41 55                   push   %r13
       e:       41 54                   push   %r12
      10:       55                      push   %rbp
      11:       53                      push   %rbx
      12:       48 83 ec 10             sub    $0x10,%rsp
      16:       48 89 fd                mov    %rdi,%rbp
      19:       4c 8b 2f                mov    (%rdi),%r13

So it has moved the wal->count check to be the very first thing,
even before even doing any stack setup.

The 32bit compiler is somewhat less smart:
00000000 <wa_list_apply>:
       0:       55                      push   %ebp
       1:       89 e5                   mov    %esp,%ebp
       3:       57                      push   %edi
       4:       56                      push   %esi
       5:       53                      push   %ebx
       6:       83 ec 10                sub    $0x10,%esp
       9:       89 45 f0                mov    %eax,-0x10(%ebp)
       c:       8b 58 10                mov    0x10(%eax),%ebx
       f:       8b 38                   mov    (%eax),%edi
      11:       85 db                   test   %ebx,%ebx
      13:       89 7d e8                mov    %edi,-0x18(%ebp)
      16:       8b 7f 0c                mov    0xc(%edi),%edi
      19:       75 0d                   jne    28 <wa_list_apply+0x28>

Not only does it do all that potentially pointless stack
setup, but then it has decided to do a bunch of stuff
with wal->gt before the jne.

That presumably explains why CI is still green despite blb/pnv.

Hmm. Now a different 32bit build also failed to hit this:
00000003 <wa_list_apply>:
       3:       55                      push   %ebp
       4:       89 e5                   mov    %esp,%ebp
       6:       57                      push   %edi
       7:       56                      push   %esi
       8:       53                      push   %ebx
       9:       83 ec 14                sub    $0x14,%esp
       c:       89 45 f0                mov    %eax,-0x10(%ebp)
       f:       8b 58 10                mov    0x10(%eax),%ebx
      12:       85 db                   test   %ebx,%ebx
      14:       75 08                   jne    1e <wa_list_apply+0x1b>

So this time it moved the wal->gt stuff to some later point.
Same compiler, different .config. Not sure which knob is
causing the difference here.

> 
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Fixes: 3653727560d0 ("drm/i915: Simplify internal helper function signature")
> Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
>  drivers/gpu/drm/i915/gt/intel_workarounds.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> index 213160f29ec3..4d7a01b45e09 100644
> --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> @@ -2991,7 +2991,7 @@ general_render_compute_wa_init(struct intel_engine_cs *engine, struct i915_wa_li
>  static void
>  engine_init_workarounds(struct intel_engine_cs *engine, struct i915_wa_list *wal)
>  {
> -	if (I915_SELFTEST_ONLY(GRAPHICS_VER(engine->i915) < 4))
> +	if (GRAPHICS_VER(engine->i915) < 4)
>  		return;
>  
>  	engine_fake_wa_init(engine, wal);
> @@ -3016,9 +3016,6 @@ void intel_engine_init_workarounds(struct intel_engine_cs *engine)
>  {
>  	struct i915_wa_list *wal = &engine->wa_list;
>  
> -	if (GRAPHICS_VER(engine->i915) < 4)
> -		return;
> -
>  	wa_init_start(wal, engine->gt, "engine", engine->name);
>  	engine_init_workarounds(engine, wal);
>  	wa_init_finish(wal);
> -- 
> 2.34.1

-- 
Ville Syrjälä
Intel

WARNING: multiple messages have this Message-ID (diff)
From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>,
	Intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org,
	Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Subject: Re: [PATCH] drm/i915: Fix workarounds on Gen2-3
Date: Fri, 18 Nov 2022 21:58:42 +0200	[thread overview]
Message-ID: <Y3fj8rJpqMhzsW35@intel.com> (raw)
In-Reply-To: <20221118115249.2683946-1-tvrtko.ursulin@linux.intel.com>

On Fri, Nov 18, 2022 at 11:52:49AM +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> 
> In 3653727560d0 ("drm/i915: Simplify internal helper function signature")
> I broke the old platforms by not noticing engine workaround init does not
> initialize the list on old platforms. Fix it by always initializing which
> already does the right thing by mostly not doing anything if there aren't
> any workarounds on the list.

Was going to give this a quick smoke test on my 865 but I can't even
reproduce the original issue on it.

Turns out the 64bit compiler is too smart:
0000000000000000 <wa_list_apply>:
       0:       8b 77 20                mov    0x20(%rdi),%esi
       3:       85 f6                   test   %esi,%esi
       5:       75 01                   jne    8 <wa_list_apply+0x8>
       7:       c3                      ret
       8:       41 57                   push   %r15
       a:       41 56                   push   %r14
       c:       41 55                   push   %r13
       e:       41 54                   push   %r12
      10:       55                      push   %rbp
      11:       53                      push   %rbx
      12:       48 83 ec 10             sub    $0x10,%rsp
      16:       48 89 fd                mov    %rdi,%rbp
      19:       4c 8b 2f                mov    (%rdi),%r13

So it has moved the wal->count check to be the very first thing,
even before even doing any stack setup.

The 32bit compiler is somewhat less smart:
00000000 <wa_list_apply>:
       0:       55                      push   %ebp
       1:       89 e5                   mov    %esp,%ebp
       3:       57                      push   %edi
       4:       56                      push   %esi
       5:       53                      push   %ebx
       6:       83 ec 10                sub    $0x10,%esp
       9:       89 45 f0                mov    %eax,-0x10(%ebp)
       c:       8b 58 10                mov    0x10(%eax),%ebx
       f:       8b 38                   mov    (%eax),%edi
      11:       85 db                   test   %ebx,%ebx
      13:       89 7d e8                mov    %edi,-0x18(%ebp)
      16:       8b 7f 0c                mov    0xc(%edi),%edi
      19:       75 0d                   jne    28 <wa_list_apply+0x28>

Not only does it do all that potentially pointless stack
setup, but then it has decided to do a bunch of stuff
with wal->gt before the jne.

That presumably explains why CI is still green despite blb/pnv.

Hmm. Now a different 32bit build also failed to hit this:
00000003 <wa_list_apply>:
       3:       55                      push   %ebp
       4:       89 e5                   mov    %esp,%ebp
       6:       57                      push   %edi
       7:       56                      push   %esi
       8:       53                      push   %ebx
       9:       83 ec 14                sub    $0x14,%esp
       c:       89 45 f0                mov    %eax,-0x10(%ebp)
       f:       8b 58 10                mov    0x10(%eax),%ebx
      12:       85 db                   test   %ebx,%ebx
      14:       75 08                   jne    1e <wa_list_apply+0x1b>

So this time it moved the wal->gt stuff to some later point.
Same compiler, different .config. Not sure which knob is
causing the difference here.

> 
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Fixes: 3653727560d0 ("drm/i915: Simplify internal helper function signature")
> Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
>  drivers/gpu/drm/i915/gt/intel_workarounds.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> index 213160f29ec3..4d7a01b45e09 100644
> --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> @@ -2991,7 +2991,7 @@ general_render_compute_wa_init(struct intel_engine_cs *engine, struct i915_wa_li
>  static void
>  engine_init_workarounds(struct intel_engine_cs *engine, struct i915_wa_list *wal)
>  {
> -	if (I915_SELFTEST_ONLY(GRAPHICS_VER(engine->i915) < 4))
> +	if (GRAPHICS_VER(engine->i915) < 4)
>  		return;
>  
>  	engine_fake_wa_init(engine, wal);
> @@ -3016,9 +3016,6 @@ void intel_engine_init_workarounds(struct intel_engine_cs *engine)
>  {
>  	struct i915_wa_list *wal = &engine->wa_list;
>  
> -	if (GRAPHICS_VER(engine->i915) < 4)
> -		return;
> -
>  	wa_init_start(wal, engine->gt, "engine", engine->name);
>  	engine_init_workarounds(engine, wal);
>  	wa_init_finish(wal);
> -- 
> 2.34.1

-- 
Ville Syrjälä
Intel

  parent reply	other threads:[~2022-11-18 19:58 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-11-18 11:52 [Intel-gfx] [PATCH] drm/i915: Fix workarounds on Gen2-3 Tvrtko Ursulin
2022-11-18 11:52 ` Tvrtko Ursulin
2022-11-18 12:15 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for " Patchwork
2022-11-18 12:38 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2022-11-18 17:14 ` [Intel-gfx] [PATCH] " Matt Roper
2022-11-21  8:45   ` Tvrtko Ursulin
2022-11-18 19:58 ` Ville Syrjälä [this message]
2022-11-18 19:58   ` Ville Syrjälä
2022-11-19  3:59 ` [Intel-gfx] ✓ Fi.CI.IGT: success for " Patchwork

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=Y3fj8rJpqMhzsW35@intel.com \
    --to=ville.syrjala@linux.intel.com \
    --cc=Intel-gfx@lists.freedesktop.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=tvrtko.ursulin@linux.intel.com \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.