diff mbox series

[2/2] drm/i915: Handle catastrophic error on engine reset

Message ID 20190412161629.26005-2-mika.kuoppala@linux.intel.com (mailing list archive)
State New, archived
Headers show
Series [1/2] drm/i915: Shortcut readiness to reset check | expand

Commit Message

Mika Kuoppala April 12, 2019, 4:16 p.m. UTC
If cat error is set, we need to clear it by acking it. Further,
if it is set, we must not do a normal request for reset.

v2: avoid goto (Chris)

Bspec: 12567
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_reg.h   |  6 ++++--
 drivers/gpu/drm/i915/i915_reset.c | 32 ++++++++++++++++++-------------
 2 files changed, 23 insertions(+), 15 deletions(-)

Comments

Chris Wilson April 12, 2019, 4:24 p.m. UTC | #1
Quoting Mika Kuoppala (2019-04-12 17:16:29)
> If cat error is set, we need to clear it by acking it. Further,
> if it is set, we must not do a normal request for reset.
> 
> v2: avoid goto (Chris)
> 
> Bspec: 12567
> Cc: Chris Wilson <chris@chris-wilson.co.uk>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
>  drivers/gpu/drm/i915/i915_reg.h   |  6 ++++--
>  drivers/gpu/drm/i915/i915_reset.c | 32 ++++++++++++++++++-------------
>  2 files changed, 23 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index 8ad2f0a03f28..c1c0f7ab03e9 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -2446,8 +2446,10 @@ enum i915_power_well_id {
>  #define RING_HWS_PGA(base)     _MMIO((base) + 0x80)
>  #define RING_HWS_PGA_GEN6(base)        _MMIO((base) + 0x2080)
>  #define RING_RESET_CTL(base)   _MMIO((base) + 0xd0)
> -#define   RESET_CTL_REQUEST_RESET  (1 << 0)
> -#define   RESET_CTL_READY_TO_RESET (1 << 1)
> +#define   RESET_CTL_CAT_ERROR     REG_BIT(2)
> +#define   RESET_CTL_READY_TO_RESET REG_BIT(1)
> +#define   RESET_CTL_REQUEST_RESET  REG_BIT(0)
> +
>  #define RING_SEMA_WAIT_POLL(base) _MMIO((base) + 0x24c)
>  
>  #define HSW_GTT_CACHE_EN       _MMIO(0x4024)
> diff --git a/drivers/gpu/drm/i915/i915_reset.c b/drivers/gpu/drm/i915/i915_reset.c
> index d874a62103e5..75d66c47db48 100644
> --- a/drivers/gpu/drm/i915/i915_reset.c
> +++ b/drivers/gpu/drm/i915/i915_reset.c
> @@ -491,25 +491,31 @@ static int gen8_engine_reset_prepare(struct intel_engine_cs *engine)
>  {
>         struct intel_uncore *uncore = engine->uncore;
>         const i915_reg_t reg = RING_RESET_CTL(engine->mmio_base);
> -       u32 ctl;
> +       u32 ctl, ack = 0, mask = 0, request = 0;
>         int ret;
>  
>         ctl = intel_uncore_read_fw(uncore, reg);
> -       if (ctl & RESET_CTL_READY_TO_RESET)
> +
> +       if (ctl & RESET_CTL_CAT_ERROR) {
> +               request |= RESET_CTL_CAT_ERROR;
> +               mask |= RESET_CTL_CAT_ERROR;
> +               /* HAS#396813: Avoid reset request if cat error */

Comment before would be our usual practice.

With the if blocks, we can just assign the values, no need for |=, just
remember to add ack here.

> +       } else if (!(ctl & RESET_CTL_READY_TO_RESET)) {
> +               request |= RESET_CTL_REQUEST_RESET;
> +               mask |= RESET_CTL_READY_TO_RESET;
> +               ack |= RESET_CTL_READY_TO_RESET;
> +       } else {
>                 return 0;
> +       }
>  
> -       intel_uncore_write_fw(uncore,
> -                             reg,
> -                             _MASKED_BIT_ENABLE(RESET_CTL_REQUEST_RESET));
> -
> -       ret = __intel_wait_for_register_fw(uncore,
> -                                          reg,
> -                                          RESET_CTL_READY_TO_RESET,
> -                                          RESET_CTL_READY_TO_RESET,
> -                                          700, 0,
> -                                          NULL);
> +       intel_uncore_write_fw(uncore, reg, _MASKED_BIT_ENABLE(request));
> +
> +       ret = __intel_wait_for_register_fw(uncore, reg, mask, ack,
> +                                          700, 0, NULL);
>         if (ret)
> -               DRM_ERROR("%s: reset request timeout\n", engine->name);
> +               DRM_ERROR("%s: reset request 0x%08x timeout 0x%08x\n",
> +                         engine->name, request,
> +                         intel_uncore_read_fw(uncore, reg));

"%s reset request timed out: {request:%08x, RESET_CTL:%08x}"
makes more sense to me -- I was struggling with the intermix of
request/timeout and the values.
-Chris
diff mbox series

Patch

diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 8ad2f0a03f28..c1c0f7ab03e9 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -2446,8 +2446,10 @@  enum i915_power_well_id {
 #define RING_HWS_PGA(base)	_MMIO((base) + 0x80)
 #define RING_HWS_PGA_GEN6(base)	_MMIO((base) + 0x2080)
 #define RING_RESET_CTL(base)	_MMIO((base) + 0xd0)
-#define   RESET_CTL_REQUEST_RESET  (1 << 0)
-#define   RESET_CTL_READY_TO_RESET (1 << 1)
+#define   RESET_CTL_CAT_ERROR	   REG_BIT(2)
+#define   RESET_CTL_READY_TO_RESET REG_BIT(1)
+#define   RESET_CTL_REQUEST_RESET  REG_BIT(0)
+
 #define RING_SEMA_WAIT_POLL(base) _MMIO((base) + 0x24c)
 
 #define HSW_GTT_CACHE_EN	_MMIO(0x4024)
diff --git a/drivers/gpu/drm/i915/i915_reset.c b/drivers/gpu/drm/i915/i915_reset.c
index d874a62103e5..75d66c47db48 100644
--- a/drivers/gpu/drm/i915/i915_reset.c
+++ b/drivers/gpu/drm/i915/i915_reset.c
@@ -491,25 +491,31 @@  static int gen8_engine_reset_prepare(struct intel_engine_cs *engine)
 {
 	struct intel_uncore *uncore = engine->uncore;
 	const i915_reg_t reg = RING_RESET_CTL(engine->mmio_base);
-	u32 ctl;
+	u32 ctl, ack = 0, mask = 0, request = 0;
 	int ret;
 
 	ctl = intel_uncore_read_fw(uncore, reg);
-	if (ctl & RESET_CTL_READY_TO_RESET)
+
+	if (ctl & RESET_CTL_CAT_ERROR) {
+		request |= RESET_CTL_CAT_ERROR;
+		mask |= RESET_CTL_CAT_ERROR;
+		/* HAS#396813: Avoid reset request if cat error */
+	} else if (!(ctl & RESET_CTL_READY_TO_RESET)) {
+		request |= RESET_CTL_REQUEST_RESET;
+		mask |= RESET_CTL_READY_TO_RESET;
+		ack |= RESET_CTL_READY_TO_RESET;
+	} else {
 		return 0;
+	}
 
-	intel_uncore_write_fw(uncore,
-			      reg,
-			      _MASKED_BIT_ENABLE(RESET_CTL_REQUEST_RESET));
-
-	ret = __intel_wait_for_register_fw(uncore,
-					   reg,
-					   RESET_CTL_READY_TO_RESET,
-					   RESET_CTL_READY_TO_RESET,
-					   700, 0,
-					   NULL);
+	intel_uncore_write_fw(uncore, reg, _MASKED_BIT_ENABLE(request));
+
+	ret = __intel_wait_for_register_fw(uncore, reg, mask, ack,
+					   700, 0, NULL);
 	if (ret)
-		DRM_ERROR("%s: reset request timeout\n", engine->name);
+		DRM_ERROR("%s: reset request 0x%08x timeout 0x%08x\n",
+			  engine->name, request,
+			  intel_uncore_read_fw(uncore, reg));
 
 	return ret;
 }