diff mbox series

[1/2] drm/i915/mtl: Disable render power-gating during selftest only

Message ID 20250110140947.3471824-2-badal.nilawar@intel.com (mailing list archive)
State New
Headers show
Series Disable RPG during live selftest | expand

Commit Message

Badal Nilawar Jan. 10, 2025, 2:09 p.m. UTC 
The temporary wa ‘drm/i915/mtl: do not enable render power-gating on MTL’ disables RPG
globally. Since the issue seen during live self-tests disabling it during these tests.

Fixes: 25e7976db86b ("drm/i915/mtl: do not enable render power-gating on MTL")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Andi Shyti <andi.shyti@intel.com>
Cc: Andrzej Hajda <andrzej.hajda@intel.com>
Signed-off-by: Badal Nilawar <badal.nilawar@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_rc6.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

kernel test robot Jan. 10, 2025, 5:27 p.m. UTC | #1 
Hi Badal,

kernel test robot noticed the following build errors:

[auto build test ERROR on linus/master]
[also build test ERROR on drm-intel/for-linux-next drm-intel/for-linux-next-fixes drm-tip/drm-tip v6.13-rc6 next-20250110]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Badal-Nilawar/drm-i915-mtl-Disable-render-power-gating-during-selftest-only/20250110-214933
base:   linus/master
patch link:    https://lore.kernel.org/r/20250110140947.3471824-2-badal.nilawar%40intel.com
patch subject: [PATCH 1/2] drm/i915/mtl: Disable render power-gating during selftest only
config: x86_64-buildonly-randconfig-004-20250110 (https://download.01.org/0day-ci/archive/20250111/202501110110.QpyfeXt4-lkp@intel.com/config)
compiler: gcc-12 (Debian 12.2.0-14) 12.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250111/202501110110.QpyfeXt4-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202501110110.QpyfeXt4-lkp@intel.com/

All errors (new ones prefixed by >>):

   drivers/gpu/drm/i915/gt/intel_rc6.c: In function 'gen11_rc6_enable':
>> drivers/gpu/drm/i915/gt/intel_rc6.c:126:13: error: 'i915_selftest' undeclared (first use in this function)
     126 |         if (i915_selftest.live && IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))
         |             ^~~~~~~~~~~~~
   drivers/gpu/drm/i915/gt/intel_rc6.c:126:13: note: each undeclared identifier is reported only once for each function it appears in


vim +/i915_selftest +126 drivers/gpu/drm/i915/gt/intel_rc6.c

    55	
    56	static void gen11_rc6_enable(struct intel_rc6 *rc6)
    57	{
    58		struct intel_gt *gt = rc6_to_gt(rc6);
    59		struct intel_uncore *uncore = gt->uncore;
    60		struct intel_engine_cs *engine;
    61		enum intel_engine_id id;
    62		u32 pg_enable;
    63		int i;
    64	
    65		/*
    66		 * With GuCRC, these parameters are set by GuC
    67		 */
    68		if (!intel_uc_uses_guc_rc(&gt->uc)) {
    69			/* 2b: Program RC6 thresholds.*/
    70			intel_uncore_write_fw(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85);
    71			intel_uncore_write_fw(uncore, GEN10_MEDIA_WAKE_RATE_LIMIT, 150);
    72	
    73			intel_uncore_write_fw(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */
    74			intel_uncore_write_fw(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */
    75			for_each_engine(engine, rc6_to_gt(rc6), id)
    76				intel_uncore_write_fw(uncore, RING_MAX_IDLE(engine->mmio_base), 10);
    77	
    78			intel_uncore_write_fw(uncore, GUC_MAX_IDLE_COUNT, 0xA);
    79	
    80			intel_uncore_write_fw(uncore, GEN6_RC_SLEEP, 0);
    81	
    82			intel_uncore_write_fw(uncore, GEN6_RC6_THRESHOLD, 50000); /* 50/125ms per EI */
    83		}
    84	
    85		/*
    86		 * 2c: Program Coarse Power Gating Policies.
    87		 *
    88		 * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we
    89		 * use instead is a more conservative estimate for the maximum time
    90		 * it takes us to service a CS interrupt and submit a new ELSP - that
    91		 * is the time which the GPU is idle waiting for the CPU to select the
    92		 * next request to execute. If the idle hysteresis is less than that
    93		 * interrupt service latency, the hardware will automatically gate
    94		 * the power well and we will then incur the wake up cost on top of
    95		 * the service latency. A similar guide from plane_state is that we
    96		 * do not want the enable hysteresis to less than the wakeup latency.
    97		 *
    98		 * igt/gem_exec_nop/sequential provides a rough estimate for the
    99		 * service latency, and puts it under 10us for Icelake, similar to
   100		 * Broadwell+, To be conservative, we want to factor in a context
   101		 * switch on top (due to ksoftirqd).
   102		 */
   103		intel_uncore_write_fw(uncore, GEN9_MEDIA_PG_IDLE_HYSTERESIS, 60);
   104		intel_uncore_write_fw(uncore, GEN9_RENDER_PG_IDLE_HYSTERESIS, 60);
   105	
   106		/* 3a: Enable RC6
   107		 *
   108		 * With GuCRC, we do not enable bit 31 of RC_CTL,
   109		 * thus allowing GuC to control RC6 entry/exit fully instead.
   110		 * We will not set the HW ENABLE and EI bits
   111		 */
   112		if (!intel_guc_rc_enable(gt_to_guc(gt)))
   113			rc6->ctl_enable = GEN6_RC_CTL_RC6_ENABLE;
   114		else
   115			rc6->ctl_enable =
   116				GEN6_RC_CTL_HW_ENABLE |
   117				GEN6_RC_CTL_RC6_ENABLE |
   118				GEN6_RC_CTL_EI_MODE(1);
   119	
   120		/*
   121		 * BSpec 52698 - Render powergating must be off.
   122		 * FIXME BSpec is outdated, disabling powergating for MTL is just
   123		 * temporary wa and should be removed after fixing real cause
   124		 * of forcewake timeouts.
   125		 */
 > 126		if (i915_selftest.live && IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))
   127			pg_enable =
   128				GEN9_MEDIA_PG_ENABLE |
   129				GEN11_MEDIA_SAMPLER_PG_ENABLE;
   130		else
   131			pg_enable =
   132				GEN9_RENDER_PG_ENABLE |
   133				GEN9_MEDIA_PG_ENABLE |
   134				GEN11_MEDIA_SAMPLER_PG_ENABLE;
   135	
   136		if (GRAPHICS_VER(gt->i915) >= 12 && !IS_DG1(gt->i915)) {
   137			for (i = 0; i < I915_MAX_VCS; i++)
   138				if (HAS_ENGINE(gt, _VCS(i)))
   139					pg_enable |= (VDN_HCP_POWERGATE_ENABLE(i) |
   140						      VDN_MFX_POWERGATE_ENABLE(i));
   141		}
   142	
   143		intel_uncore_write_fw(uncore, GEN9_PG_ENABLE, pg_enable);
   144	}
   145
Rodrigo Vivi Jan. 10, 2025, 9:07 p.m. UTC | #2 
On Fri, Jan 10, 2025 at 07:39:46PM +0530, Badal Nilawar wrote:
> The temporary wa ‘drm/i915/mtl: do not enable render power-gating on MTL’ disables RPG
> globally. Since the issue seen during live self-tests disabling it during these tests.
> 
> Fixes: 25e7976db86b ("drm/i915/mtl: do not enable render power-gating on MTL")
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Cc: Andi Shyti <andi.shyti@intel.com>
> Cc: Andrzej Hajda <andrzej.hajda@intel.com>
> Signed-off-by: Badal Nilawar <badal.nilawar@intel.com>
> ---
>  drivers/gpu/drm/i915/gt/intel_rc6.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/i915/gt/intel_rc6.c b/drivers/gpu/drm/i915/gt/intel_rc6.c
> index 9378d5901c49..2dd68ca7ec0e 100644
> --- a/drivers/gpu/drm/i915/gt/intel_rc6.c
> +++ b/drivers/gpu/drm/i915/gt/intel_rc6.c
> @@ -123,7 +123,7 @@ static void gen11_rc6_enable(struct intel_rc6 *rc6)
>  	 * temporary wa and should be removed after fixing real cause
>  	 * of forcewake timeouts.
>  	 */
> -	if (IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))
> +	if (i915_selftest.live && IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))


hmmm... is there something we could do inside the selftest to mitigate that?
here it is crossing (abusing?!) a lot of layers to workaround a case...

also this starts to depend on some config flags, what the compilation bot test hit...

>  		pg_enable =
>  			GEN9_MEDIA_PG_ENABLE |
>  			GEN11_MEDIA_SAMPLER_PG_ENABLE;
> -- 
> 2.34.1
>
kernel test robot Jan. 11, 2025, 10:55 a.m. UTC | #3 
Hi Badal,

kernel test robot noticed the following build errors:

[auto build test ERROR on linus/master]
[also build test ERROR on drm-intel/for-linux-next drm-intel/for-linux-next-fixes drm-tip/drm-tip v6.13-rc6 next-20250110]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Badal-Nilawar/drm-i915-mtl-Disable-render-power-gating-during-selftest-only/20250110-214933
base:   linus/master
patch link:    https://lore.kernel.org/r/20250110140947.3471824-2-badal.nilawar%40intel.com
patch subject: [PATCH 1/2] drm/i915/mtl: Disable render power-gating during selftest only
config: i386-randconfig-014-20250111 (https://download.01.org/0day-ci/archive/20250111/202501111817.Kz3T41my-lkp@intel.com/config)
compiler: clang version 19.1.3 (https://github.com/llvm/llvm-project ab51eccf88f5321e7c60591c5546b254b6afab99)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250111/202501111817.Kz3T41my-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202501111817.Kz3T41my-lkp@intel.com/

All errors (new ones prefixed by >>):

>> drivers/gpu/drm/i915/gt/intel_rc6.c:126:6: error: use of undeclared identifier 'i915_selftest'
     126 |         if (i915_selftest.live && IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))
         |             ^
   1 error generated.


vim +/i915_selftest +126 drivers/gpu/drm/i915/gt/intel_rc6.c

    55	
    56	static void gen11_rc6_enable(struct intel_rc6 *rc6)
    57	{
    58		struct intel_gt *gt = rc6_to_gt(rc6);
    59		struct intel_uncore *uncore = gt->uncore;
    60		struct intel_engine_cs *engine;
    61		enum intel_engine_id id;
    62		u32 pg_enable;
    63		int i;
    64	
    65		/*
    66		 * With GuCRC, these parameters are set by GuC
    67		 */
    68		if (!intel_uc_uses_guc_rc(&gt->uc)) {
    69			/* 2b: Program RC6 thresholds.*/
    70			intel_uncore_write_fw(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85);
    71			intel_uncore_write_fw(uncore, GEN10_MEDIA_WAKE_RATE_LIMIT, 150);
    72	
    73			intel_uncore_write_fw(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */
    74			intel_uncore_write_fw(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */
    75			for_each_engine(engine, rc6_to_gt(rc6), id)
    76				intel_uncore_write_fw(uncore, RING_MAX_IDLE(engine->mmio_base), 10);
    77	
    78			intel_uncore_write_fw(uncore, GUC_MAX_IDLE_COUNT, 0xA);
    79	
    80			intel_uncore_write_fw(uncore, GEN6_RC_SLEEP, 0);
    81	
    82			intel_uncore_write_fw(uncore, GEN6_RC6_THRESHOLD, 50000); /* 50/125ms per EI */
    83		}
    84	
    85		/*
    86		 * 2c: Program Coarse Power Gating Policies.
    87		 *
    88		 * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we
    89		 * use instead is a more conservative estimate for the maximum time
    90		 * it takes us to service a CS interrupt and submit a new ELSP - that
    91		 * is the time which the GPU is idle waiting for the CPU to select the
    92		 * next request to execute. If the idle hysteresis is less than that
    93		 * interrupt service latency, the hardware will automatically gate
    94		 * the power well and we will then incur the wake up cost on top of
    95		 * the service latency. A similar guide from plane_state is that we
    96		 * do not want the enable hysteresis to less than the wakeup latency.
    97		 *
    98		 * igt/gem_exec_nop/sequential provides a rough estimate for the
    99		 * service latency, and puts it under 10us for Icelake, similar to
   100		 * Broadwell+, To be conservative, we want to factor in a context
   101		 * switch on top (due to ksoftirqd).
   102		 */
   103		intel_uncore_write_fw(uncore, GEN9_MEDIA_PG_IDLE_HYSTERESIS, 60);
   104		intel_uncore_write_fw(uncore, GEN9_RENDER_PG_IDLE_HYSTERESIS, 60);
   105	
   106		/* 3a: Enable RC6
   107		 *
   108		 * With GuCRC, we do not enable bit 31 of RC_CTL,
   109		 * thus allowing GuC to control RC6 entry/exit fully instead.
   110		 * We will not set the HW ENABLE and EI bits
   111		 */
   112		if (!intel_guc_rc_enable(gt_to_guc(gt)))
   113			rc6->ctl_enable = GEN6_RC_CTL_RC6_ENABLE;
   114		else
   115			rc6->ctl_enable =
   116				GEN6_RC_CTL_HW_ENABLE |
   117				GEN6_RC_CTL_RC6_ENABLE |
   118				GEN6_RC_CTL_EI_MODE(1);
   119	
   120		/*
   121		 * BSpec 52698 - Render powergating must be off.
   122		 * FIXME BSpec is outdated, disabling powergating for MTL is just
   123		 * temporary wa and should be removed after fixing real cause
   124		 * of forcewake timeouts.
   125		 */
 > 126		if (i915_selftest.live && IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))
   127			pg_enable =
   128				GEN9_MEDIA_PG_ENABLE |
   129				GEN11_MEDIA_SAMPLER_PG_ENABLE;
   130		else
   131			pg_enable =
   132				GEN9_RENDER_PG_ENABLE |
   133				GEN9_MEDIA_PG_ENABLE |
   134				GEN11_MEDIA_SAMPLER_PG_ENABLE;
   135	
   136		if (GRAPHICS_VER(gt->i915) >= 12 && !IS_DG1(gt->i915)) {
   137			for (i = 0; i < I915_MAX_VCS; i++)
   138				if (HAS_ENGINE(gt, _VCS(i)))
   139					pg_enable |= (VDN_HCP_POWERGATE_ENABLE(i) |
   140						      VDN_MFX_POWERGATE_ENABLE(i));
   141		}
   142	
   143		intel_uncore_write_fw(uncore, GEN9_PG_ENABLE, pg_enable);
   144	}
   145
diff mbox series

Patch

diff --git a/drivers/gpu/drm/i915/gt/intel_rc6.c b/drivers/gpu/drm/i915/gt/intel_rc6.c
index 9378d5901c49..2dd68ca7ec0e 100644
--- a/drivers/gpu/drm/i915/gt/intel_rc6.c
+++ b/drivers/gpu/drm/i915/gt/intel_rc6.c
@@ -123,7 +123,7 @@  static void gen11_rc6_enable(struct intel_rc6 *rc6)
 	 * temporary wa and should be removed after fixing real cause
 	 * of forcewake timeouts.
 	 */
-	if (IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))
+	if (i915_selftest.live && IS_GFX_GT_IP_RANGE(gt, IP_VER(12, 70), IP_VER(12, 74)))
 		pg_enable =
 			GEN9_MEDIA_PG_ENABLE |
 			GEN11_MEDIA_SAMPLER_PG_ENABLE;