Re: [PATCH 12/20] drm/i915/icl: Check for fused-off VDBOX and VEBOX instances

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





On 2/13/2018 10:07 PM, Mika Kuoppala wrote:
From: Oscar Mateo <oscar.mateo@xxxxxxxxx>

In Gen11, the Video Decode engines (aka VDBOX, aka VCS, aka BSD) and the
Video Enhancement engines (aka VEBOX, aka VECS) could be fused off. Also,
each VDBOX and VEBOX has its own power well, which only exist if the related
engine exists in the HW.

Unfortunately, we have a Catch-22 situation going on: we need to read an
MMIO register with the fuse info, but we cannot fully enable MMIO until
we read it (since we need the real engines to initialize the forcewake
domains).
We need to ensure BLITTER is initialized first and use low level functions fw_domains_get/put()
around raw read to know these engines status.
  We workaround this problem by reading the fuse after the MMIO
is partially ready, but before we initialize forcewake.

Bspec: 20680

v2: We were shifting incorrectly for vebox disable (Vinay)

v3: Assert mmio is ready and warn if we have attempted to initialize
     forcewake for fused-off engines (Paulo)

v4:
   - Use INTEL_GEN in new code (Tvrtko)
   - Shorter local variable (Tvrtko, Michal)
   - Keep "if (!...) continue" style (Tvrtko)
   - No unnecessary BUG_ON (Tvrtko)
   - WARN_ON and cleanup if wrong mask (Tvrtko, Michal)
   - Use I915_READ_FW (Michal)
   - Use I915_MAX_VCS/VECS macros (Michal)

v5: Rebased by Rodrigo fixing conflicts on top of:
     commit 33def1ff7b0 ("drm/i915: Simplify intel_engines_init")

v6: Fix v5. Remove info->num_rings. (by Oscar)

v7: Rebase (Rodrigo).

v8:
   - s/intel_device_info_fused_off_engines/intel_device_info_init_mmio (Chris)
   - Make vdbox_disable & vebox_disable local variables (Chris)

Cc: Paulo Zanoni <paulo.r.zanoni@xxxxxxxxx>
Cc: Vinay Belgaumkar <vinay.belgaumkar@xxxxxxxxx>
Cc: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx>
Cc: Michal Wajdeczko <michal.wajdeczko@xxxxxxxxx>
Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx>
Signed-off-by: Oscar Mateo <oscar.mateo@xxxxxxxxx>
---
  drivers/gpu/drm/i915/i915_drv.c          |  2 ++
  drivers/gpu/drm/i915/i915_drv.h          |  1 +
  drivers/gpu/drm/i915/i915_reg.h          |  5 +++
  drivers/gpu/drm/i915/intel_device_info.c | 54 ++++++++++++++++++++++++++++++++
  4 files changed, 62 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 9380c9f69b0f..43b2f620bca7 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1033,6 +1033,8 @@ static int i915_driver_init_mmio(struct drm_i915_private *dev_priv)
  	if (ret < 0)
  		goto err_bridge;
+ intel_device_info_init_mmio(dev_priv);
This should be called during intel_uncore_fw_domains_init after
        fw_domain_init(.., FW_DOMAIN_ID_BLITTER, ..);
+
  	intel_uncore_init(dev_priv);
intel_uc_init_mmio(dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 65e674668b2e..ba16c2025364 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3438,6 +3438,7 @@ void i915_unreserve_fence(struct drm_i915_fence_reg *fence);
  void i915_gem_revoke_fences(struct drm_i915_private *dev_priv);
  void i915_gem_restore_fences(struct drm_i915_private *dev_priv);
+void intel_device_info_init_mmio(struct drm_i915_private *dev_priv);
  void i915_gem_detect_bit_6_swizzle(struct drm_i915_private *dev_priv);
  void i915_gem_object_do_bit_17_swizzle(struct drm_i915_gem_object *obj,
  				       struct sg_table *pages);
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index b6cd725ff0b7..2b8d3a13dd27 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -2860,6 +2860,11 @@ enum i915_power_well_id {
  #define GEN10_EU_DISABLE3		_MMIO(0x9140)
  #define   GEN10_EU_DIS_SS_MASK		0xff
+#define GEN11_GT_VEBOX_VDBOX_DISABLE _MMIO(0x9140)
+#define GEN11_GT_VDBOX_DISABLE_MASK	0xff
+#define GEN11_GT_VEBOX_DISABLE_SHIFT	16
+#define GEN11_GT_VEBOX_DISABLE_MASK	(0xff << GEN11_GT_VEBOX_DISABLE_SHIFT)
+
  #define GEN6_BSD_SLEEP_PSMI_CONTROL	_MMIO(0x12050)
  #define   GEN6_BSD_SLEEP_MSG_DISABLE	(1 << 0)
  #define   GEN6_BSD_SLEEP_FLUSH_DISABLE	(1 << 2)
diff --git a/drivers/gpu/drm/i915/intel_device_info.c b/drivers/gpu/drm/i915/intel_device_info.c
index 9352f34e75c4..7c8779faf162 100644
--- a/drivers/gpu/drm/i915/intel_device_info.c
+++ b/drivers/gpu/drm/i915/intel_device_info.c
@@ -595,3 +595,57 @@ void intel_driver_caps_print(const struct intel_driver_caps *caps,
  {
  	drm_printf(p, "scheduler: %x\n", caps->scheduler);
  }
+
+/*
+ * Determine which engines are fused off in our particular hardware.
+ *
+ * This function needs to be called after the MMIO has been setup (as we need
+ * to read registers) but before uncore init (because the powerwell for the
+ * fused off engines doesn't exist, so we cannot initialize forcewake for them)
+ */
+void intel_device_info_init_mmio(struct drm_i915_private *dev_priv)
How about naming this function as intel_uncore_runtime_fw_domains_init such that fw_domain_init for each of the unfused engines in next patch get called as part of this. Additional effect is update of device info. Calling intel_uncore_runtime_fw_domains_init from intel_uncore_fw_domains_init and calls to intel_device_info_init*
from i915_driver_init_* paths seems to make code consistent.
+{
+	struct intel_device_info *info = mkwrite_device_info(dev_priv);
+	u8 vdbox_disable, vebox_disable;
+	u32 media_fuse;
+	int i;
+
+	if (INTEL_GEN(dev_priv) < 11)
+		return;
+
+	GEM_BUG_ON(!dev_priv->regs);
+
+	media_fuse = I915_READ_FW(GEN11_GT_VEBOX_VDBOX_DISABLE);
this should be
fw_domains_get(dev_priv, FORCEWAKE_BLITTER);
media_fuse = I915_READ_FW(GEN11, GT_VEBOX_VDBOX_DISABLE);
fw_domains_put(dev_priv, FORCEWAKE_BLITTER);

Earlier I had thought of calling ASSIGN_FW_DOMAINS_TABLE, ASSIGN_*_MMIO_VFUNCS before intel_uncore_fw_domains_init and use I915_READ here for reading the fuse. But that approach seems to expose the vfuncs and forcewake table before fw domains are initialized. Although we can get to know invalid access to read/write accessors before fw_domains get initialized, current ordering of fw_domains init followed by fw domain table/read/write vfuncs init seems right. So using fw_domains_get/put as suggested above
should be the way.

Chris, Tvrtko, Mika, do you agree?
+
+	vdbox_disable = media_fuse & GEN11_GT_VDBOX_DISABLE_MASK;
+	vebox_disable = (media_fuse & GEN11_GT_VEBOX_DISABLE_MASK) >>
+			GEN11_GT_VEBOX_DISABLE_SHIFT;
+
+	DRM_DEBUG_DRIVER("vdbox disable: %04x\n", vdbox_disable);
+	for (i = 0; i < I915_MAX_VCS; i++) {
+		if (!HAS_ENGINE(dev_priv, _VCS(i)))
+			continue;
+
+		if (!(BIT(i) & vdbox_disable))
+			continue;
+
+		info->ring_mask &= ~ENGINE_MASK(_VCS(i));
+		WARN_ON(dev_priv->uncore.fw_domains &
+			BIT(FW_DOMAIN_ID_MEDIA_VDBOX0 + i));
+		DRM_DEBUG_DRIVER("vcs%u fused off\n", i);
+	}
+
+	DRM_DEBUG_DRIVER("vebox disable: %04x\n", vebox_disable);
+	for (i = 0; i < I915_MAX_VECS; i++) {
+		if (!HAS_ENGINE(dev_priv, _VECS(i)))
+			continue;
+
+		if (!(BIT(i) & vebox_disable))
+			continue;
+
+		info->ring_mask &= ~ENGINE_MASK(_VECS(i));
+		WARN_ON(dev_priv->uncore.fw_domains &
+			BIT(FW_DOMAIN_ID_MEDIA_VEBOX0 + i));
+		DRM_DEBUG_DRIVER("vecs%u fused off\n", i);
+	}
+}

--
Thanks,
Sagar

_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx




[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux