On Mon, Mar 02, 2015 at 11:07:20AM +0000, Arun Siluvery wrote:
> Some of the workarounds are to be applied during context save but before
> restore and some at the end of context save/restore but before executing
> the instructions in the ring. Workaround batch buffers are created for
> this purpose as they cannot be applied using normal means. HW executes
> them at specific stages during context save/restore.
>
> In this method we initialize batch buffer with w/a commands and its address
> is supplied using context offset pointers when a context is initialized.
>
> This patch introduces indirect and per-context batch buffers using which
> following workarounds are applied. These are required to fix issues
> observed with preemption related workloads.
>
> In Indirect context w/a batch buffer,
> +WaDisableCtxRestoreArbitration
> +WaFlushCoherentL3CacheLinesAtContextSwitch
> +WaClearSlmSpaceAtContextSwitch
>
> In Per context w/a batch buffer,
> +WaDisableCtxRestoreArbitration
> +WaRsRestoreWithPerCtxtBb
>
> v2: Use GTT address type for all privileged instructions, update as
> per dynamic pinning changes, minor simplifications, rename variables
> as follows to keep lines under 80 chars and rebase.
> s/indirect_ctx_wa_ringbuf/indirect_ctx_wa_bb
> s/per_ctx_wa_ringbuf/per_ctx_wa_bb
>
> v3: Modify WA BB initialization to Gen specific.
>
> v4: s/PIPE_CONTROL_FLUSH_RO_CACHES/PIPE_CONTROL_FLUSH_L3 (Ville)
> This patches modifies definitions of MI_LOAD_REGISTER_MEM and
> MI_LOAD_REGISTER_REG; Add GEN8 specific defines for these instructions
> so as to not break any future users of existing definitions (Michel)
>
> Change-Id: I0cedb536b7f6d9f10ba9e81ba625848e7bab603c
> Signed-off-by: Rafael Barbalho <[email protected]>
> Signed-off-by: Arun Siluvery <[email protected]>
> ---
> drivers/gpu/drm/i915/i915_drv.h | 3 +
> drivers/gpu/drm/i915/i915_reg.h | 28 ++++
> drivers/gpu/drm/i915/intel_lrc.c | 231
> +++++++++++++++++++++++++++++++-
> drivers/gpu/drm/i915/intel_ringbuffer.h | 3 +
> 4 files changed, 258 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index d42040f..86cdb52 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -774,6 +774,9 @@ struct intel_context {
>
> /* Execlists */
> bool rcs_initialized;
> + struct intel_ringbuffer *indirect_ctx_wa_bb;
> + struct intel_ringbuffer *per_ctx_wa_bb;
Why is this per-ctx and not per-engine? Also your patch splitting doesn't
make that much sense: Patch 1 only adds a static function without any
users (resulting in gcc being unhappy). Imo a better split would be:
- wire up wa batch/ring allocation/freing functions
- wire up the changes to the lrc initial reg state code
- one patch per w/a entry you add
Cheers, Daniel
> +
> struct {
> struct drm_i915_gem_object *state;
> struct intel_ringbuffer *ringbuf;
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index 55143cb..3048494 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -347,6 +347,26 @@
> #define MI_INVALIDATE_BSD (1<<7)
> #define MI_FLUSH_DW_USE_GTT (1<<2)
> #define MI_FLUSH_DW_USE_PPGTT (0<<2)
> +#define MI_ATOMIC(len) MI_INSTR(0x2F, (len-2))
> +#define MI_ATOMIC_MEMORY_TYPE_GGTT (1<<22)
> +#define MI_ATOMIC_INLINE_DATA (1<<18)
> +#define MI_ATOMIC_CS_STALL (1<<17)
> +#define MI_ATOMIC_RETURN_DATA_CTL (1<<16)
> +#define MI_ATOMIC_OP_MASK(op) ((op) << 8)
> +#define MI_ATOMIC_AND MI_ATOMIC_OP_MASK(0x01)
> +#define MI_ATOMIC_OR MI_ATOMIC_OP_MASK(0x02)
> +#define MI_ATOMIC_XOR MI_ATOMIC_OP_MASK(0x03)
> +#define MI_ATOMIC_MOVE MI_ATOMIC_OP_MASK(0x04)
> +#define MI_ATOMIC_INC MI_ATOMIC_OP_MASK(0x05)
> +#define MI_ATOMIC_DEC MI_ATOMIC_OP_MASK(0x06)
> +#define MI_ATOMIC_ADD MI_ATOMIC_OP_MASK(0x07)
> +#define MI_ATOMIC_SUB MI_ATOMIC_OP_MASK(0x08)
> +#define MI_ATOMIC_RSUB MI_ATOMIC_OP_MASK(0x09)
> +#define MI_ATOMIC_IMAX MI_ATOMIC_OP_MASK(0x0A)
> +#define MI_ATOMIC_IMIN MI_ATOMIC_OP_MASK(0x0B)
> +#define MI_ATOMIC_UMAX MI_ATOMIC_OP_MASK(0x0C)
> +#define MI_ATOMIC_UMIN MI_ATOMIC_OP_MASK(0x0D)
> +
> #define MI_BATCH_BUFFER MI_INSTR(0x30, 1)
> #define MI_BATCH_NON_SECURE (1)
> /* for snb/ivb/vlv this also means "batch in ppgtt" when ppgtt is enabled. */
> @@ -410,6 +430,7 @@
> #define DISPLAY_PLANE_A (0<<20)
> #define DISPLAY_PLANE_B (1<<20)
> #define GFX_OP_PIPE_CONTROL(len) ((0x3<<29)|(0x3<<27)|(0x2<<24)|(len-2))
> +#define PIPE_CONTROL_FLUSH_L3 (1<<27)
> #define PIPE_CONTROL_GLOBAL_GTT_IVB (1<<24) /*
> gen7+ */
> #define PIPE_CONTROL_MMIO_WRITE (1<<23)
> #define PIPE_CONTROL_STORE_DATA_INDEX (1<<21)
> @@ -426,6 +447,7 @@
> #define PIPE_CONTROL_INDIRECT_STATE_DISABLE (1<<9)
> #define PIPE_CONTROL_NOTIFY (1<<8)
> #define PIPE_CONTROL_FLUSH_ENABLE (1<<7) /* gen7+ */
> +#define PIPE_CONTROL_DC_FLUSH_ENABLE (1<<5)
> #define PIPE_CONTROL_VF_CACHE_INVALIDATE (1<<4)
> #define PIPE_CONTROL_CONST_CACHE_INVALIDATE (1<<3)
> #define PIPE_CONTROL_STATE_CACHE_INVALIDATE (1<<2)
> @@ -451,6 +473,10 @@
> #define MI_REPORT_PERF_COUNT_GGTT (1<<0)
> #define MI_LOAD_REGISTER_MEM MI_INSTR(0x29, 0)
> #define MI_LOAD_REGISTER_REG MI_INSTR(0x2A, 0)
> +#define MI_LOAD_REGISTER_MEM_GEN8 MI_INSTR(0x29, 2)
> +#define MI_LRM_USE_GLOBAL_GTT (1<<22)
> +#define MI_LRM_ASYNC_MODE_ENABLE (1<<21)
> +#define MI_LOAD_REGISTER_REG_GEN8 MI_INSTR(0x2A, 1)
> #define MI_RS_STORE_DATA_IMM MI_INSTR(0x2B, 0)
> #define MI_LOAD_URB_MEM MI_INSTR(0x2C, 0)
> #define MI_STORE_URB_MEM MI_INSTR(0x2D, 0)
> @@ -1520,6 +1546,8 @@ enum skl_disp_power_wells {
> #define GEN8_RC_SEMA_IDLE_MSG_DISABLE (1 << 12)
> #define GEN8_FF_DOP_CLOCK_GATE_DISABLE (1<<10)
>
> +#define GEN8_RS_PREEMPT_STATUS 0x215C
> +
> /* Fuse readout registers for GT */
> #define CHV_FUSE_GT (VLV_DISPLAY_BASE + 0x2168)
> #define CHV_FGT_EU_DIS_SS0_R0_SHIFT 16
> diff --git a/drivers/gpu/drm/i915/intel_lrc.c
> b/drivers/gpu/drm/i915/intel_lrc.c
> index ea37a56..326539b 100644
> --- a/drivers/gpu/drm/i915/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/intel_lrc.c
> @@ -202,6 +202,7 @@ enum {
> FAULT_AND_CONTINUE /* Unsupported */
> };
> #define GEN8_CTX_ID_SHIFT 32
> +#define CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT 0x17
>
> static int intel_lr_context_pin(struct intel_engine_cs *ring,
> struct intel_context *ctx);
> @@ -1139,6 +1140,169 @@ create_wa_bb(struct intel_engine_cs *ring, uint32_t
> bb_size)
> return ringbuf;
> }
>
> +static int gen8_init_indirectctx_bb(struct intel_engine_cs *ring,
> + struct intel_context *ctx)
> +{
> + unsigned long flags = 0;
> + u32 scratch_addr;
> + struct intel_ringbuffer *ringbuf = NULL;
> +
> + if (ring->scratch.obj == NULL) {
> + DRM_ERROR("scratch page not allocated for %s\n", ring->name);
> + return -EINVAL;
> + }
> +
> + ringbuf = create_wa_bb(ring, PAGE_SIZE);
> + if (!ringbuf)
> + return -ENOMEM;
> +
> + ctx->indirect_ctx_wa_bb = ringbuf;
> +
> + /* WaDisableCtxRestoreArbitration:bdw,chv */
> + intel_logical_ring_emit(ringbuf, MI_ARB_ON_OFF | MI_ARB_DISABLE);
> +
> + /* WaFlushCoherentL3CacheLinesAtContextSwitch:bdw,chv */
> + intel_logical_ring_emit(ringbuf, GFX_OP_PIPE_CONTROL(6));
> + intel_logical_ring_emit(ringbuf, PIPE_CONTROL_DC_FLUSH_ENABLE);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, 0);
> +
> + /* WaClearSlmSpaceAtContextSwitch:bdw,chv */
> + flags = PIPE_CONTROL_FLUSH_L3 |
> + PIPE_CONTROL_GLOBAL_GTT_IVB |
> + PIPE_CONTROL_CS_STALL |
> + PIPE_CONTROL_QW_WRITE;
> +
> + /* Actual scratch location is at 128 bytes offset */
> + scratch_addr = ring->scratch.gtt_offset + 2*CACHELINE_BYTES;
> + scratch_addr |= PIPE_CONTROL_GLOBAL_GTT;
> +
> + intel_logical_ring_emit(ringbuf, GFX_OP_PIPE_CONTROL(6));
> + intel_logical_ring_emit(ringbuf, flags);
> + intel_logical_ring_emit(ringbuf, scratch_addr);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, 0);
> +
> + /* Padding to align with cache line */
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, 0);
> +
> + /*
> + * No MI_BATCH_BUFFER_END is required in Indirect ctx BB because
> + * execution depends on the size defined in CTX_RCS_INDIRECT_CTX
> + */
> +
> + return 0;
> +}
> +
> +static int gen8_init_perctx_bb(struct intel_engine_cs *ring,
> + struct intel_context *ctx)
> +{
> + unsigned long flags = 0;
> + u32 scratch_addr;
> + struct intel_ringbuffer *ringbuf = NULL;
> +
> + if (ring->scratch.obj == NULL) {
> + DRM_ERROR("scratch page not allocated for %s\n", ring->name);
> + return -EINVAL;
> + }
> +
> + ringbuf = create_wa_bb(ring, PAGE_SIZE);
> + if (!ringbuf)
> + return -ENOMEM;
> +
> + ctx->per_ctx_wa_bb = ringbuf;
> +
> + /* Actual scratch location is at 128 bytes offset */
> + scratch_addr = ring->scratch.gtt_offset + 2*CACHELINE_BYTES;
> + scratch_addr |= PIPE_CONTROL_GLOBAL_GTT;
> +
> + /* WaDisableCtxRestoreArbitration:bdw,chv */
> + intel_logical_ring_emit(ringbuf, MI_ARB_ON_OFF | MI_ARB_ENABLE);
> +
> + /*
> + * As per Bspec, to workaround a known HW issue, SW must perform the
> + * below programming sequence prior to programming MI_BATCH_BUFFER_END.
> + *
> + * This is only applicable for Gen8.
> + */
> +
> + /* WaRsRestoreWithPerCtxtBb:bdw,chv */
> + intel_logical_ring_emit(ringbuf, MI_LOAD_REGISTER_IMM(1));
> + intel_logical_ring_emit(ringbuf, INSTPM);
> + intel_logical_ring_emit(ringbuf,
> + _MASKED_BIT_DISABLE(INSTPM_FORCE_ORDERING));
> +
> + flags = MI_ATOMIC_MEMORY_TYPE_GGTT |
> + MI_ATOMIC_INLINE_DATA |
> + MI_ATOMIC_CS_STALL |
> + MI_ATOMIC_RETURN_DATA_CTL |
> + MI_ATOMIC_MOVE;
> +
> + intel_logical_ring_emit(ringbuf, MI_ATOMIC(5) | flags);
> + intel_logical_ring_emit(ringbuf, scratch_addr);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf,
> + _MASKED_BIT_ENABLE(INSTPM_FORCE_ORDERING));
> + intel_logical_ring_emit(ringbuf,
> + _MASKED_BIT_ENABLE(INSTPM_FORCE_ORDERING));
> +
> + /*
> + * Bspec says MI_LOAD_REGISTER_MEM, MI_LOAD_REGISTER_REG and
> + * MI_BATCH_BUFFER_END instructions in this sequence need to be
> + * in the same cacheline.
> + */
> + while (((unsigned long) ringbuf->tail % CACHELINE_BYTES) != 0)
> + intel_logical_ring_emit(ringbuf, MI_NOOP);
> +
> + intel_logical_ring_emit(ringbuf,
> + MI_LOAD_REGISTER_MEM_GEN8 |
> + MI_LRM_USE_GLOBAL_GTT |
> + MI_LRM_ASYNC_MODE_ENABLE);
> + intel_logical_ring_emit(ringbuf, INSTPM);
> + intel_logical_ring_emit(ringbuf, 0);
> + intel_logical_ring_emit(ringbuf, scratch_addr);
> +
> + /*
> + * Bspec says there should not be any commands programmed
> + * between MI_LOAD_REGISTER_REG and MI_BATCH_BUFFER_END so
> + * do not add any new commands
> + */
> + intel_logical_ring_emit(ringbuf, MI_LOAD_REGISTER_REG_GEN8);
> + intel_logical_ring_emit(ringbuf, GEN8_RS_PREEMPT_STATUS);
> + intel_logical_ring_emit(ringbuf, GEN8_RS_PREEMPT_STATUS);
> +
> + intel_logical_ring_emit(ringbuf, MI_BATCH_BUFFER_END);
> +
> + return 0;
> +}
> +
> +static int intel_init_workaround_bb(struct intel_engine_cs *ring,
> + struct intel_context *ctx)
> +{
> + int ret;
> + struct drm_device *dev = ring->dev;
> +
> + WARN_ON(ring->id != RCS);
> +
> + if (IS_GEN8(dev)) {
> + ret = gen8_init_indirectctx_bb(ring, ctx);
> + if (ret)
> + return ret;
> +
> + ret = gen8_init_perctx_bb(ring, ctx);
> + if (ret)
> + return ret;
> + }
> +
> + return 0;
> +
> +}
> +
> static int gen8_init_common_ring(struct intel_engine_cs *ring)
> {
> struct drm_device *dev = ring->dev;
> @@ -1490,6 +1654,7 @@ static int logical_render_ring_init(struct drm_device
> *dev)
> else
> ring->init_hw = gen8_init_render_ring;
> ring->init_context = gen8_init_rcs_context;
> + ring->init_context_bb = intel_init_workaround_bb;
> ring->cleanup = intel_fini_pipe_control;
> ring->get_seqno = gen8_get_seqno;
> ring->set_seqno = gen8_set_seqno;
> @@ -1500,11 +1665,16 @@ static int logical_render_ring_init(struct drm_device
> *dev)
> ring->emit_bb_start = gen8_emit_bb_start;
>
> ring->dev = dev;
> +
> + ret = intel_init_pipe_control(ring);
> + if (ret)
> + return ret;
> +
> ret = logical_ring_init(dev, ring);
> if (ret)
> return ret;
>
> - return intel_init_pipe_control(ring);
> + return 0;
> }
>
> static int logical_bsd_ring_init(struct drm_device *dev)
> @@ -1784,15 +1954,29 @@ populate_lr_context(struct intel_context *ctx, struct
> drm_i915_gem_object *ctx_o
> reg_state[CTX_SECOND_BB_STATE] = ring->mmio_base + 0x118;
> reg_state[CTX_SECOND_BB_STATE+1] = 0;
> if (ring->id == RCS) {
> - /* TODO: according to BSpec, the register state context
> - * for CHV does not have these. OTOH, these registers do
> - * exist in CHV. I'm waiting for a clarification */
> reg_state[CTX_BB_PER_CTX_PTR] = ring->mmio_base + 0x1c0;
> - reg_state[CTX_BB_PER_CTX_PTR+1] = 0;
> +
> + if (ctx->per_ctx_wa_bb)
> + reg_state[CTX_BB_PER_CTX_PTR + 1] =
> + i915_gem_obj_ggtt_offset(
> + ctx->per_ctx_wa_bb->obj) | 0x01;
> + else
> + reg_state[CTX_BB_PER_CTX_PTR+1] = 0;
> +
> reg_state[CTX_RCS_INDIRECT_CTX] = ring->mmio_base + 0x1c4;
> - reg_state[CTX_RCS_INDIRECT_CTX+1] = 0;
> reg_state[CTX_RCS_INDIRECT_CTX_OFFSET] = ring->mmio_base +
> 0x1c8;
> - reg_state[CTX_RCS_INDIRECT_CTX_OFFSET+1] = 0;
> +
> + if (ctx->indirect_ctx_wa_bb) {
> + reg_state[CTX_RCS_INDIRECT_CTX + 1] =
> + i915_gem_obj_ggtt_offset(
> + ctx->indirect_ctx_wa_bb->obj) | 0x01;
> +
> + reg_state[CTX_RCS_INDIRECT_CTX_OFFSET + 1] =
> + CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT << 6;
> + } else {
> + reg_state[CTX_RCS_INDIRECT_CTX+1] = 0;
> + reg_state[CTX_RCS_INDIRECT_CTX_OFFSET+1] = 0;
> + }
> }
> reg_state[CTX_LRI_HEADER_1] = MI_LOAD_REGISTER_IMM(9);
> reg_state[CTX_LRI_HEADER_1] |= MI_LRI_FORCE_POSTED;
> @@ -1859,6 +2043,18 @@ void intel_lr_context_free(struct intel_context *ctx)
> drm_gem_object_unreference(&ctx_obj->base);
> }
> }
> +
> + if (ctx->indirect_ctx_wa_bb) {
> + intel_unpin_ringbuffer_obj(ctx->indirect_ctx_wa_bb);
> + intel_destroy_ringbuffer_obj(ctx->indirect_ctx_wa_bb);
> + kfree(ctx->indirect_ctx_wa_bb);
> + }
> +
> + if (ctx->per_ctx_wa_bb) {
> + intel_unpin_ringbuffer_obj(ctx->per_ctx_wa_bb);
> + intel_destroy_ringbuffer_obj(ctx->per_ctx_wa_bb);
> + kfree(ctx->per_ctx_wa_bb);
> + }
> }
>
> static uint32_t get_lr_context_size(struct intel_engine_cs *ring)
> @@ -1985,6 +2181,16 @@ int intel_lr_context_deferred_create(struct
> intel_context *ctx,
>
> }
>
> + if (ring->id == RCS && !ctx->rcs_initialized) {
> + if (ring->init_context_bb) {
> + ret = ring->init_context_bb(ring, ctx);
> + if (ret) {
> + DRM_ERROR("ring init context bb: %d\n", ret);
> + goto error;
> + }
> + }
> + }
> +
> ret = populate_lr_context(ctx, ctx_obj, ring, ringbuf);
> if (ret) {
> DRM_DEBUG_DRIVER("Failed to populate LRC: %d\n", ret);
> @@ -2013,6 +2219,17 @@ int intel_lr_context_deferred_create(struct
> intel_context *ctx,
> return 0;
>
> error:
> + if (ctx->indirect_ctx_wa_bb) {
> + intel_unpin_ringbuffer_obj(ctx->indirect_ctx_wa_bb);
> + intel_destroy_ringbuffer_obj(ctx->indirect_ctx_wa_bb);
> + kfree(ctx->indirect_ctx_wa_bb);
> + }
> + if (ctx->per_ctx_wa_bb) {
> + intel_unpin_ringbuffer_obj(ctx->per_ctx_wa_bb);
> + intel_destroy_ringbuffer_obj(ctx->per_ctx_wa_bb);
> + kfree(ctx->per_ctx_wa_bb);
> + }
> +
> if (is_global_default_ctx)
> intel_unpin_ringbuffer_obj(ringbuf);
> error_destroy_rbuf:
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h
> b/drivers/gpu/drm/i915/intel_ringbuffer.h
> index 39183fc..839d698 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> @@ -146,6 +146,9 @@ struct intel_engine_cs {
> int (*init_context)(struct intel_engine_cs *ring,
> struct intel_context *ctx);
>
> + int (*init_context_bb)(struct intel_engine_cs *ring,
> + struct intel_context *ctx);
> +
> void (*write_tail)(struct intel_engine_cs *ring,
> u32 value);
> int __must_check (*flush)(struct intel_engine_cs *ring,
> --
> 2.3.0
>
> _______________________________________________
> Intel-gfx mailing list
> [email protected]
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
