| Message ID | 20200310004911.1723239-4-matthew.d.roper@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Gen11 workarounds | expand |
Quoting Matt Roper (2020-03-10 00:49:07) > The bspec description for this workaround tells us to program > 0xFFFF_FFFF into both FBC_RT_BASE_ADDR_REGISTER_* registers, but we've > previously found that this leads to failures in CI. Our suspicion is > that the failures are caused by this valid turning on the "address valid > bit" even though we're intentionally supplying an invalid address. > Experimentation has shown that setting all bits _except_ for the > RT_VALID bit seems to avoid these failures. > > v2: > - Mask off the RT_VALID bit. Experimentation with CI trybot indicates > that this is necessary to avoid reset failures on BCS. What reset failures? > Bspec: 11388 > Bspec: 33451 > Signed-off-by: Matt Roper <matthew.d.roper@intel.com> > --- > drivers/gpu/drm/i915/gt/intel_workarounds.c | 6 ++++++ > drivers/gpu/drm/i915/i915_reg.h | 1 + > 2 files changed, 7 insertions(+) > > diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c > index 3e352e2a5b79..1cf931dde0ca 100644 > --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c > +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c > @@ -575,6 +575,12 @@ static void icl_ctx_workarounds_init(struct intel_engine_cs *engine, > /* allow headerless messages for preemptible GPGPU context */ > WA_SET_BIT_MASKED(GEN10_SAMPLER_MODE, > GEN11_SAMPLER_ENABLE_HEADLESS_MSG); > + > + /* Wa_1604278689:icl,ehl */ > + wa_write_masked_or(wal, IVB_FBC_RT_BASE_UPPER, > + 0, /* write-only register; skip validation */ > + 0xFFFFFFFF); > + wa_write(wal, IVB_FBC_RT_BASE, 0xFFFFFFFF & ~ILK_FBC_RT_VALID); Disable first. -Chris
On Tue, Mar 10, 2020 at 04:37:52PM +0000, Chris Wilson wrote: > Quoting Matt Roper (2020-03-10 00:49:07) > > The bspec description for this workaround tells us to program > > 0xFFFF_FFFF into both FBC_RT_BASE_ADDR_REGISTER_* registers, but we've > > previously found that this leads to failures in CI. Our suspicion is > > that the failures are caused by this valid turning on the "address valid > > bit" even though we're intentionally supplying an invalid address. > > Experimentation has shown that setting all bits _except_ for the > > RT_VALID bit seems to avoid these failures. > > > > v2: > > - Mask off the RT_VALID bit. Experimentation with CI trybot indicates > > that this is necessary to avoid reset failures on BCS. > > What reset failures? It was back in August, so I think the logs have expired from CI now: https://patchwork.freedesktop.org/series/65276/#rev1 Matt > > > Bspec: 11388 > > Bspec: 33451 > > Signed-off-by: Matt Roper <matthew.d.roper@intel.com> > > --- > > drivers/gpu/drm/i915/gt/intel_workarounds.c | 6 ++++++ > > drivers/gpu/drm/i915/i915_reg.h | 1 + > > 2 files changed, 7 insertions(+) > > > > diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c > > index 3e352e2a5b79..1cf931dde0ca 100644 > > --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c > > +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c > > @@ -575,6 +575,12 @@ static void icl_ctx_workarounds_init(struct intel_engine_cs *engine, > > /* allow headerless messages for preemptible GPGPU context */ > > WA_SET_BIT_MASKED(GEN10_SAMPLER_MODE, > > GEN11_SAMPLER_ENABLE_HEADLESS_MSG); > > + > > + /* Wa_1604278689:icl,ehl */ > > + wa_write_masked_or(wal, IVB_FBC_RT_BASE_UPPER, > > + 0, /* write-only register; skip validation */ > > + 0xFFFFFFFF); > > + wa_write(wal, IVB_FBC_RT_BASE, 0xFFFFFFFF & ~ILK_FBC_RT_VALID); > > Disable first. > -Chris
diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c index 3e352e2a5b79..1cf931dde0ca 100644 --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c @@ -575,6 +575,12 @@ static void icl_ctx_workarounds_init(struct intel_engine_cs *engine, /* allow headerless messages for preemptible GPGPU context */ WA_SET_BIT_MASKED(GEN10_SAMPLER_MODE, GEN11_SAMPLER_ENABLE_HEADLESS_MSG); + + /* Wa_1604278689:icl,ehl */ + wa_write_masked_or(wal, IVB_FBC_RT_BASE_UPPER, + 0, /* write-only register; skip validation */ + 0xFFFFFFFF); + wa_write(wal, IVB_FBC_RT_BASE, 0xFFFFFFFF & ~ILK_FBC_RT_VALID); } static void tgl_ctx_workarounds_init(struct intel_engine_cs *engine, diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 79ae9654dac9..92ae96cf5b64 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -3285,6 +3285,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) /* Framebuffer compression for Ivybridge */ #define IVB_FBC_RT_BASE _MMIO(0x7020) +#define IVB_FBC_RT_BASE_UPPER _MMIO(0x7024) #define IPS_CTL _MMIO(0x43408) #define IPS_ENABLE (1 << 31)
The bspec description for this workaround tells us to program 0xFFFF_FFFF into both FBC_RT_BASE_ADDR_REGISTER_* registers, but we've previously found that this leads to failures in CI. Our suspicion is that the failures are caused by this valid turning on the "address valid bit" even though we're intentionally supplying an invalid address. Experimentation has shown that setting all bits _except_ for the RT_VALID bit seems to avoid these failures. v2: - Mask off the RT_VALID bit. Experimentation with CI trybot indicates that this is necessary to avoid reset failures on BCS. Bspec: 11388 Bspec: 33451 Signed-off-by: Matt Roper <matthew.d.roper@intel.com> --- drivers/gpu/drm/i915/gt/intel_workarounds.c | 6 ++++++ drivers/gpu/drm/i915/i915_reg.h | 1 + 2 files changed, 7 insertions(+)