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@xxxxxxxxxxxxxxxxxx>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx>
> ---
> 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
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
