Currently we face a severe problem on Braswell that manifests as invalid ppGTT accesses. The code tries to maintain the PDP (page directory pointers) inside the context in two ways, direct write into the context and a pipelined LRI update. The direct write into the context is fundamentally racy as it is unserialised with any access (read or write) the GPU is doing. By asserting that Braswell is not used with vGPU (currently an unsupported platform) we can eliminate the dangerous direct write into the context image and solely use the pipelined update. However, the LRI of the PDP fouls up the GPU, causing it to freeze and take out the machine with "forcewake ack timeouts". This seems possible to workaround by preventing the GPU from sleeping (via means of disabling the power-state management interface, i.e. forcing each ring to remain awake) around the update. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108656 References: https://bugs.freedesktop.org/show_bug.cgi?id=108714 Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> Cc: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx> --- drivers/gpu/drm/i915/intel_lrc.c | 143 ++++++++++++++++--------------- 1 file changed, 76 insertions(+), 67 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index c60a1c2233e0..071022acbb4d 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -363,31 +363,12 @@ execlists_context_schedule_out(struct i915_request *rq, unsigned long status) trace_i915_request_out(rq); } -static void -execlists_update_context_pdps(struct i915_hw_ppgtt *ppgtt, u32 *reg_state) -{ - ASSIGN_CTX_PDP(ppgtt, reg_state, 3); - ASSIGN_CTX_PDP(ppgtt, reg_state, 2); - ASSIGN_CTX_PDP(ppgtt, reg_state, 1); - ASSIGN_CTX_PDP(ppgtt, reg_state, 0); -} - static u64 execlists_update_context(struct i915_request *rq) { - struct i915_hw_ppgtt *ppgtt = rq->gem_context->ppgtt; struct intel_context *ce = rq->hw_context; - u32 *reg_state = ce->lrc_reg_state; - - reg_state[CTX_RING_TAIL+1] = intel_ring_set_tail(rq->ring, rq->tail); - /* - * True 32b PPGTT with dynamic page allocation: update PDP - * registers and point the unallocated PDPs to scratch page. - * PML4 is allocated during ppgtt init, so this is not needed - * in 48-bit mode. - */ - if (!i915_vm_is_48bit(&ppgtt->vm)) - execlists_update_context_pdps(ppgtt, reg_state); + ce->lrc_reg_state[CTX_RING_TAIL + 1] = + intel_ring_set_tail(rq->ring, rq->tail); /* * Make sure the context image is complete before we submit it to HW. @@ -1233,6 +1214,64 @@ execlists_context_pin(struct intel_engine_cs *engine, return __execlists_context_pin(engine, ctx, ce); } +static int emit_pdps(struct i915_request *rq) +{ + const struct intel_engine_cs * const engine = rq->engine; + struct i915_hw_ppgtt * const ppgtt = rq->gem_context->ppgtt; + const unsigned int num_rings = INTEL_INFO(rq->i915)->num_rings; + struct intel_engine_cs *other; + enum intel_engine_id id; + u32 *cs; + int i; + + if (!(ppgtt->pd_dirty_rings & intel_engine_flag(engine))) + return 0; + + cs = intel_ring_begin(rq, 4 * (GEN8_3LVL_PDPES + num_rings) + 8); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + /* + * Force the GPU (not just the local engine/powerwell!) to remain awake, + * or else we may kill the machine with "timed out waiting for + * forcewake ack request". + */ + + *cs++ = MI_LOAD_REGISTER_IMM(num_rings); + for_each_engine(other, rq->i915, id) { + *cs++ = i915_mmio_reg_offset(RING_PSMI_CTL(other->mmio_base)); + *cs++ = _MASKED_BIT_ENABLE(GEN6_PSMI_SLEEP_MSG_DISABLE); + } + + *cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES); + for (i = GEN8_3LVL_PDPES; i--; ) { + const dma_addr_t pd_daddr = i915_page_dir_dma_addr(ppgtt, i); + + *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(engine, i)); + *cs++ = upper_32_bits(pd_daddr); + *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(engine, i)); + *cs++ = lower_32_bits(pd_daddr); + } + *cs++ = MI_NOOP; + + /* Posting read to flush the mmio before letting the GPU sleep again */ + *cs++ = MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT; + *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(engine, 0)); + *cs++ = i915_ggtt_offset(engine->i915->gt.scratch); + *cs++ = 0; + + *cs++ = MI_LOAD_REGISTER_IMM(num_rings); + for_each_engine(other, rq->i915, id) { + *cs++ = i915_mmio_reg_offset(RING_PSMI_CTL(other->mmio_base)); + *cs++ = _MASKED_BIT_DISABLE(GEN6_PSMI_SLEEP_MSG_DISABLE); + } + + intel_ring_advance(rq, cs); + + ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine); + return 0; +} + static int execlists_request_alloc(struct i915_request *request) { int ret; @@ -1249,13 +1288,22 @@ static int execlists_request_alloc(struct i915_request *request) if (ret) return ret; - /* Note that after this point, we have committed to using + /* + * Note that after this point, we have committed to using * this request as it is being used to both track the * state of engine initialisation and liveness of the * golden renderstate above. Think twice before you try * to cancel/unwind this request now. */ + if (!i915_vm_is_48bit(&request->gem_context->ppgtt->vm)) { + GEM_BUG_ON(intel_vgpu_active(request->i915)); + + ret = emit_pdps(request); + if (ret) + return ret; + } + request->reserved_space -= EXECLISTS_REQUEST_SIZE; return 0; } @@ -1825,56 +1873,11 @@ static void execlists_reset_finish(struct intel_engine_cs *engine) atomic_read(&execlists->tasklet.count)); } -static int intel_logical_ring_emit_pdps(struct i915_request *rq) -{ - struct i915_hw_ppgtt *ppgtt = rq->gem_context->ppgtt; - struct intel_engine_cs *engine = rq->engine; - const int num_lri_cmds = GEN8_3LVL_PDPES * 2; - u32 *cs; - int i; - - cs = intel_ring_begin(rq, num_lri_cmds * 2 + 2); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - *cs++ = MI_LOAD_REGISTER_IMM(num_lri_cmds); - for (i = GEN8_3LVL_PDPES - 1; i >= 0; i--) { - const dma_addr_t pd_daddr = i915_page_dir_dma_addr(ppgtt, i); - - *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(engine, i)); - *cs++ = upper_32_bits(pd_daddr); - *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(engine, i)); - *cs++ = lower_32_bits(pd_daddr); - } - - *cs++ = MI_NOOP; - intel_ring_advance(rq, cs); - - return 0; -} - static int gen8_emit_bb_start(struct i915_request *rq, u64 offset, u32 len, const unsigned int flags) { u32 *cs; - int ret; - - /* Don't rely in hw updating PDPs, specially in lite-restore. - * Ideally, we should set Force PD Restore in ctx descriptor, - * but we can't. Force Restore would be a second option, but - * it is unsafe in case of lite-restore (because the ctx is - * not idle). PML4 is allocated during ppgtt init so this is - * not needed in 48-bit.*/ - if ((intel_engine_flag(rq->engine) & rq->gem_context->ppgtt->pd_dirty_rings) && - !i915_vm_is_48bit(&rq->gem_context->ppgtt->vm) && - !intel_vgpu_active(rq->i915)) { - ret = intel_logical_ring_emit_pdps(rq); - if (ret) - return ret; - - rq->gem_context->ppgtt->pd_dirty_rings &= ~intel_engine_flag(rq->engine); - } cs = intel_ring_begin(rq, 6); if (IS_ERR(cs)) @@ -1907,6 +1910,7 @@ static int gen8_emit_bb_start(struct i915_request *rq, *cs++ = MI_ARB_ON_OFF | MI_ARB_DISABLE; *cs++ = MI_NOOP; + intel_ring_advance(rq, cs); return 0; @@ -2537,6 +2541,11 @@ static void execlists_init_reg_state(u32 *regs, * other PDP Descriptors are ignored. */ ASSIGN_CTX_PML4(ctx->ppgtt, regs); + } else { + ASSIGN_CTX_PDP(ctx->ppgtt, regs, 3); + ASSIGN_CTX_PDP(ctx->ppgtt, regs, 2); + ASSIGN_CTX_PDP(ctx->ppgtt, regs, 1); + ASSIGN_CTX_PDP(ctx->ppgtt, regs, 0); } if (rcs) { -- 2.20.0.rc1 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
