const struct intel_engine_cs *engine,
u32 head);
+static int lrc_ring_mi_mode(const struct intel_engine_cs *engine)
+{
+ if (INTEL_GEN(engine->i915) >= 12)
+ return 0x60;
+ else if (INTEL_GEN(engine->i915) >= 9)
+ return 0x54;
+ else if (engine->class == RENDER_CLASS)
+ return 0x58;
+ else
+ return -1;
+}
+
+static int lrc_ring_gpr0(const struct intel_engine_cs *engine)
+{
+ if (INTEL_GEN(engine->i915) >= 12)
+ return 0x74;
+ else if (INTEL_GEN(engine->i915) >= 9)
+ return 0x68;
+ else if (engine->class == RENDER_CLASS)
+ return 0xd8;
+ else
+ return -1;
+}
+
+static int lrc_ring_wa_bb_per_ctx(const struct intel_engine_cs *engine)
+{
+ if (INTEL_GEN(engine->i915) >= 12)
+ return 0x12;
+ else if (INTEL_GEN(engine->i915) >= 9 || engine->class == RENDER_CLASS)
+ return 0x18;
+ else
+ return -1;
+}
+
+static int lrc_ring_indirect_ptr(const struct intel_engine_cs *engine)
+{
+ int x;
+
+ x = lrc_ring_wa_bb_per_ctx(engine);
+ if (x < 0)
+ return x;
+
+ return x + 2;
+}
+
+static int lrc_ring_indirect_offset(const struct intel_engine_cs *engine)
+{
+ int x;
+
+ x = lrc_ring_indirect_ptr(engine);
+ if (x < 0)
+ return x;
+
+ return x + 2;
+}
+
+static int lrc_ring_cmd_buf_cctl(const struct intel_engine_cs *engine)
+{
+ if (engine->class != RENDER_CLASS)
+ return -1;
+
+ if (INTEL_GEN(engine->i915) >= 12)
+ return 0xb6;
+ else if (INTEL_GEN(engine->i915) >= 11)
+ return 0xaa;
+ else
+ return -1;
+}
+
+static u32
+lrc_ring_indirect_offset_default(const struct intel_engine_cs *engine)
+{
+ switch (INTEL_GEN(engine->i915)) {
+ default:
+ MISSING_CASE(INTEL_GEN(engine->i915));
+ fallthrough;
+ case 12:
+ return GEN12_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
+ case 11:
+ return GEN11_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
+ case 10:
+ return GEN10_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
+ case 9:
+ return GEN9_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
+ case 8:
+ return GEN8_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
+ }
+}
+
+static void
+lrc_ring_setup_indirect_ctx(u32 *regs,
+ const struct intel_engine_cs *engine,
+ u32 ctx_bb_ggtt_addr,
+ u32 size)
+{
+ GEM_BUG_ON(!size);
+ GEM_BUG_ON(!IS_ALIGNED(size, CACHELINE_BYTES));
+ GEM_BUG_ON(lrc_ring_indirect_ptr(engine) == -1);
+ regs[lrc_ring_indirect_ptr(engine) + 1] =
+ ctx_bb_ggtt_addr | (size / CACHELINE_BYTES);
+
+ GEM_BUG_ON(lrc_ring_indirect_offset(engine) == -1);
+ regs[lrc_ring_indirect_offset(engine) + 1] =
+ lrc_ring_indirect_offset_default(engine) << 6;
+}
+
static u32 intel_context_get_runtime(const struct intel_context *ce)
{
/*
#define REG16(x) \
(((x) >> 9) | BIT(7) | BUILD_BUG_ON_ZERO(x >= 0x10000)), \
(((x) >> 2) & 0x7f)
-#define END(x) 0, (x)
+#define END(total_state_size) 0, (total_state_size)
{
const u32 base = engine->mmio_base;
if (flags & POSTED)
*regs |= MI_LRI_FORCE_POSTED;
if (INTEL_GEN(engine->i915) >= 11)
- *regs |= MI_LRI_CS_MMIO;
+ *regs |= MI_LRI_LRM_CS_MMIO;
regs++;
GEM_BUG_ON(!count);
NOP(6),
LRI(1, 0),
REG(0x0c8),
+ NOP(3 + 9 + 1),
+
+ LRI(51, POSTED),
+ REG16(0x588),
+ REG16(0x588),
+ REG16(0x588),
+ REG16(0x588),
+ REG16(0x588),
+ REG16(0x588),
+ REG(0x028),
+ REG(0x09c),
+ REG(0x0c0),
+ REG(0x178),
+ REG(0x17c),
+ REG16(0x358),
+ REG(0x170),
+ REG(0x150),
+ REG(0x154),
+ REG(0x158),
+ REG16(0x41c),
+ REG16(0x600),
+ REG16(0x604),
+ REG16(0x608),
+ REG16(0x60c),
+ REG16(0x610),
+ REG16(0x614),
+ REG16(0x618),
+ REG16(0x61c),
+ REG16(0x620),
+ REG16(0x624),
+ REG16(0x628),
+ REG16(0x62c),
+ REG16(0x630),
+ REG16(0x634),
+ REG16(0x638),
+ REG16(0x63c),
+ REG16(0x640),
+ REG16(0x644),
+ REG16(0x648),
+ REG16(0x64c),
+ REG16(0x650),
+ REG16(0x654),
+ REG16(0x658),
+ REG16(0x65c),
+ REG16(0x660),
+ REG16(0x664),
+ REG16(0x668),
+ REG16(0x66c),
+ REG16(0x670),
+ REG16(0x674),
+ REG16(0x678),
+ REG16(0x67c),
+ REG(0x068),
+ REG(0x084),
+ NOP(1),
- END(80)
+ END(192)
};
#undef END
write_sequnlock_irqrestore(&engine->stats.lock, flags);
}
-static int lrc_ring_mi_mode(const struct intel_engine_cs *engine)
-{
- if (INTEL_GEN(engine->i915) >= 12)
- return 0x60;
- else if (INTEL_GEN(engine->i915) >= 9)
- return 0x54;
- else if (engine->class == RENDER_CLASS)
- return 0x58;
- else
- return -1;
-}
-
static void
execlists_check_context(const struct intel_context *ce,
const struct intel_engine_cs *engine)
if (engine->pinned_default_state)
memcpy(regs, /* skip restoring the vanilla PPHWSP */
- engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE,
+ engine->pinned_default_state + LRC_STATE_OFFSET,
engine->context_size - PAGE_SIZE);
execlists_init_reg_state(regs, ce, engine, ce->ring, false);
return *csb & (GEN8_CTX_STATUS_IDLE_ACTIVE | GEN8_CTX_STATUS_PREEMPTED);
}
-static inline void flush_hwsp(const struct i915_request *rq)
-{
- mb();
- clflush((void *)READ_ONCE(rq->hwsp_seqno));
- mb();
-}
-
static void process_csb(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
* We rely on the hardware being strongly
* ordered, that the breadcrumb write is
* coherent (visible from the CPU) before the
- * user interrupt and CSB is processed.
+ * user interrupt is processed. One might assume
+ * that the breadcrumb write being before the
+ * user interrupt and the CS event for the context
+ * switch would therefore be before the CS event
+ * itself...
*/
if (GEM_SHOW_DEBUG() &&
!i915_request_completed(*execlists->active)) {
const u32 *regs __maybe_unused =
rq->context->lrc_reg_state;
- /*
- * Flush the breadcrumb before crying foul.
- *
- * Since we have hit this on icl and seen the
- * breadcrumb advance as we print out the debug
- * info (so the problem corrected itself without
- * lasting damage), and we know that icl suffers
- * from missing global observation points in
- * execlists, presume that affects even more
- * coherency.
- */
- flush_hwsp(rq);
-
+ ENGINE_TRACE(engine,
+ "context completed before request!\n");
ENGINE_TRACE(engine,
"ring:{start:0x%08x, head:%04x, tail:%04x, ctl:%08x, mode:%08x}\n",
ENGINE_READ(engine, RING_START),
regs[CTX_RING_START],
regs[CTX_RING_HEAD],
regs[CTX_RING_TAIL]);
-
- /* Still? Declare it caput! */
- if (!i915_request_completed(rq) &&
- !reset_in_progress(execlists))
- GEM_BUG_ON("context completed before request");
}
execlists_schedule_out(*execlists->active++);
static void execlists_context_unpin(struct intel_context *ce)
{
- check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE,
+ check_redzone((void *)ce->lrc_reg_state - LRC_STATE_OFFSET,
ce->engine);
i915_gem_object_unpin_map(ce->state->obj);
}
+static u32 *
+gen12_emit_timestamp_wa(const struct intel_context *ce, u32 *cs)
+{
+ *cs++ = MI_LOAD_REGISTER_MEM_GEN8 |
+ MI_SRM_LRM_GLOBAL_GTT |
+ MI_LRI_LRM_CS_MMIO;
+ *cs++ = i915_mmio_reg_offset(GEN8_RING_CS_GPR(0, 0));
+ *cs++ = i915_ggtt_offset(ce->state) + LRC_STATE_OFFSET +
+ CTX_TIMESTAMP * sizeof(u32);
+ *cs++ = 0;
+
+ *cs++ = MI_LOAD_REGISTER_REG |
+ MI_LRR_SOURCE_CS_MMIO |
+ MI_LRI_LRM_CS_MMIO;
+ *cs++ = i915_mmio_reg_offset(GEN8_RING_CS_GPR(0, 0));
+ *cs++ = i915_mmio_reg_offset(RING_CTX_TIMESTAMP(0));
+
+ *cs++ = MI_LOAD_REGISTER_REG |
+ MI_LRR_SOURCE_CS_MMIO |
+ MI_LRI_LRM_CS_MMIO;
+ *cs++ = i915_mmio_reg_offset(GEN8_RING_CS_GPR(0, 0));
+ *cs++ = i915_mmio_reg_offset(RING_CTX_TIMESTAMP(0));
+
+ return cs;
+}
+
+static u32 *
+gen12_emit_restore_scratch(const struct intel_context *ce, u32 *cs)
+{
+ GEM_BUG_ON(lrc_ring_gpr0(ce->engine) == -1);
+
+ *cs++ = MI_LOAD_REGISTER_MEM_GEN8 |
+ MI_SRM_LRM_GLOBAL_GTT |
+ MI_LRI_LRM_CS_MMIO;
+ *cs++ = i915_mmio_reg_offset(GEN8_RING_CS_GPR(0, 0));
+ *cs++ = i915_ggtt_offset(ce->state) + LRC_STATE_OFFSET +
+ (lrc_ring_gpr0(ce->engine) + 1) * sizeof(u32);
+ *cs++ = 0;
+
+ return cs;
+}
+
+static u32 *
+gen12_emit_cmd_buf_wa(const struct intel_context *ce, u32 *cs)
+{
+ GEM_BUG_ON(lrc_ring_cmd_buf_cctl(ce->engine) == -1);
+
+ *cs++ = MI_LOAD_REGISTER_MEM_GEN8 |
+ MI_SRM_LRM_GLOBAL_GTT |
+ MI_LRI_LRM_CS_MMIO;
+ *cs++ = i915_mmio_reg_offset(GEN8_RING_CS_GPR(0, 0));
+ *cs++ = i915_ggtt_offset(ce->state) + LRC_STATE_OFFSET +
+ (lrc_ring_cmd_buf_cctl(ce->engine) + 1) * sizeof(u32);
+ *cs++ = 0;
+
+ *cs++ = MI_LOAD_REGISTER_REG |
+ MI_LRR_SOURCE_CS_MMIO |
+ MI_LRI_LRM_CS_MMIO;
+ *cs++ = i915_mmio_reg_offset(GEN8_RING_CS_GPR(0, 0));
+ *cs++ = i915_mmio_reg_offset(RING_CMD_BUF_CCTL(0));
+
+ return cs;
+}
+
+static u32 *
+gen12_emit_indirect_ctx_rcs(const struct intel_context *ce, u32 *cs)
+{
+ cs = gen12_emit_timestamp_wa(ce, cs);
+ cs = gen12_emit_cmd_buf_wa(ce, cs);
+ cs = gen12_emit_restore_scratch(ce, cs);
+
+ return cs;
+}
+
+static u32 *
+gen12_emit_indirect_ctx_xcs(const struct intel_context *ce, u32 *cs)
+{
+ cs = gen12_emit_timestamp_wa(ce, cs);
+ cs = gen12_emit_restore_scratch(ce, cs);
+
+ return cs;
+}
+
+static inline u32 context_wa_bb_offset(const struct intel_context *ce)
+{
+ return PAGE_SIZE * ce->wa_bb_page;
+}
+
+static u32 *context_indirect_bb(const struct intel_context *ce)
+{
+ void *ptr;
+
+ GEM_BUG_ON(!ce->wa_bb_page);
+
+ ptr = ce->lrc_reg_state;
+ ptr -= LRC_STATE_OFFSET; /* back to start of context image */
+ ptr += context_wa_bb_offset(ce);
+
+ return ptr;
+}
+
+static void
+setup_indirect_ctx_bb(const struct intel_context *ce,
+ const struct intel_engine_cs *engine,
+ u32 *(*emit)(const struct intel_context *, u32 *))
+{
+ u32 * const start = context_indirect_bb(ce);
+ u32 *cs;
+
+ cs = emit(ce, start);
+ GEM_BUG_ON(cs - start > I915_GTT_PAGE_SIZE / sizeof(*cs));
+ while ((unsigned long)cs % CACHELINE_BYTES)
+ *cs++ = MI_NOOP;
+
+ lrc_ring_setup_indirect_ctx(ce->lrc_reg_state, engine,
+ i915_ggtt_offset(ce->state) +
+ context_wa_bb_offset(ce),
+ (cs - start) * sizeof(*cs));
+}
+
static void
__execlists_update_reg_state(const struct intel_context *ce,
const struct intel_engine_cs *engine,
i915_oa_init_reg_state(ce, engine);
}
+
+ if (ce->wa_bb_page) {
+ u32 *(*fn)(const struct intel_context *ce, u32 *cs);
+
+ fn = gen12_emit_indirect_ctx_xcs;
+ if (ce->engine->class == RENDER_CLASS)
+ fn = gen12_emit_indirect_ctx_rcs;
+
+ /* Mutually exclusive wrt to global indirect bb */
+ GEM_BUG_ON(engine->wa_ctx.indirect_ctx.size);
+ setup_indirect_ctx_bb(ce, engine, fn);
+ }
}
static int
return PTR_ERR(vaddr);
ce->lrc_desc = lrc_descriptor(ce, engine) | CTX_DESC_FORCE_RESTORE;
- ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE;
+ ce->lrc_reg_state = vaddr + LRC_STATE_OFFSET;
__execlists_update_reg_state(ce, engine, ce->ring->tail);
return 0;
return 0;
}
-static u32 intel_lr_indirect_ctx_offset(const struct intel_engine_cs *engine)
-{
- u32 indirect_ctx_offset;
-
- switch (INTEL_GEN(engine->i915)) {
- default:
- MISSING_CASE(INTEL_GEN(engine->i915));
- /* fall through */
- case 12:
- indirect_ctx_offset =
- GEN12_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
- break;
- case 11:
- indirect_ctx_offset =
- GEN11_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
- break;
- case 10:
- indirect_ctx_offset =
- GEN10_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
- break;
- case 9:
- indirect_ctx_offset =
- GEN9_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
- break;
- case 8:
- indirect_ctx_offset =
- GEN8_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT;
- break;
- }
-
- return indirect_ctx_offset;
-}
-
-
static void init_common_reg_state(u32 * const regs,
const struct intel_engine_cs *engine,
const struct intel_ring *ring,
}
static void init_wa_bb_reg_state(u32 * const regs,
- const struct intel_engine_cs *engine,
- u32 pos_bb_per_ctx)
+ const struct intel_engine_cs *engine)
{
const struct i915_ctx_workarounds * const wa_ctx = &engine->wa_ctx;
if (wa_ctx->per_ctx.size) {
const u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma);
- regs[pos_bb_per_ctx] =
+ GEM_BUG_ON(lrc_ring_wa_bb_per_ctx(engine) == -1);
+ regs[lrc_ring_wa_bb_per_ctx(engine) + 1] =
(ggtt_offset + wa_ctx->per_ctx.offset) | 0x01;
}
if (wa_ctx->indirect_ctx.size) {
- const u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma);
-
- regs[pos_bb_per_ctx + 2] =
- (ggtt_offset + wa_ctx->indirect_ctx.offset) |
- (wa_ctx->indirect_ctx.size / CACHELINE_BYTES);
-
- regs[pos_bb_per_ctx + 4] =
- intel_lr_indirect_ctx_offset(engine) << 6;
+ lrc_ring_setup_indirect_ctx(regs, engine,
+ i915_ggtt_offset(wa_ctx->vma) +
+ wa_ctx->indirect_ctx.offset,
+ wa_ctx->indirect_ctx.size);
}
}
init_common_reg_state(regs, engine, ring, inhibit);
init_ppgtt_reg_state(regs, vm_alias(ce->vm));
- init_wa_bb_reg_state(regs, engine,
- INTEL_GEN(engine->i915) >= 12 ?
- GEN12_CTX_BB_PER_CTX_PTR :
- CTX_BB_PER_CTX_PTR);
+ init_wa_bb_reg_state(regs, engine);
__reset_stop_ring(regs, engine);
}
* The second page of the context object contains some registers which
* must be set up prior to the first execution.
*/
- execlists_init_reg_state(vaddr + LRC_STATE_PN * PAGE_SIZE,
+ execlists_init_reg_state(vaddr + LRC_STATE_OFFSET,
ce, engine, ring, inhibit);
ret = 0;
if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM))
context_size += I915_GTT_PAGE_SIZE; /* for redzone */
+ if (INTEL_GEN(engine->i915) == 12) {
+ ce->wa_bb_page = context_size / PAGE_SIZE;
+ context_size += PAGE_SIZE;
+ }
+
ctx_obj = i915_gem_object_create_shmem(engine->i915, context_size);
if (IS_ERR(ctx_obj))
return PTR_ERR(ctx_obj);
return;
local_irq_disable();
- for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
- struct intel_engine_cs *sibling = ve->siblings[n];
+ for (n = 0; n < ve->num_siblings; n++) {
+ struct intel_engine_cs *sibling = READ_ONCE(ve->siblings[n]);
struct ve_node * const node = &ve->nodes[sibling->id];
struct rb_node **parent, *rb;
bool first;
+ if (!READ_ONCE(ve->request))
+ break; /* already handled by a sibling's tasklet */
+
if (unlikely(!(mask & sibling->mask))) {
if (!RB_EMPTY_NODE(&node->rb)) {
spin_lock(&sibling->active.lock);