diff options
Diffstat (limited to 'drivers/gpu/drm/xe/xe_lrc.c')
| -rw-r--r-- | drivers/gpu/drm/xe/xe_lrc.c | 194 |
1 files changed, 169 insertions, 25 deletions
diff --git a/drivers/gpu/drm/xe/xe_lrc.c b/drivers/gpu/drm/xe/xe_lrc.c index 7ad853b0788a..615bbc372ac6 100644 --- a/drivers/gpu/drm/xe/xe_lrc.c +++ b/drivers/gpu/drm/xe/xe_lrc.c @@ -5,8 +5,11 @@ #include "xe_lrc.h" +#include <linux/ascii85.h> + #include "instructions/xe_mi_commands.h" #include "instructions/xe_gfxpipe_commands.h" +#include "instructions/xe_gfx_state_commands.h" #include "regs/xe_engine_regs.h" #include "regs/xe_gpu_commands.h" #include "regs/xe_lrc_layout.h" @@ -23,13 +26,28 @@ #include "xe_sriov.h" #include "xe_vm.h" -#define LRC_VALID (1 << 0) -#define LRC_PRIVILEGE (1 << 8) -#define LRC_ADDRESSING_MODE_SHIFT 3 +#define LRC_VALID BIT_ULL(0) +#define LRC_PRIVILEGE BIT_ULL(8) +#define LRC_ADDRESSING_MODE GENMASK_ULL(4, 3) #define LRC_LEGACY_64B_CONTEXT 3 -#define ENGINE_CLASS_SHIFT 61 -#define ENGINE_INSTANCE_SHIFT 48 +#define LRC_ENGINE_CLASS GENMASK_ULL(63, 61) +#define LRC_ENGINE_INSTANCE GENMASK_ULL(53, 48) + +struct xe_lrc_snapshot { + struct xe_bo *lrc_bo; + void *lrc_snapshot; + unsigned long lrc_size, lrc_offset; + + u32 context_desc; + u32 head; + struct { + u32 internal; + u32 memory; + } tail; + u32 start_seqno; + u32 seqno; +}; static struct xe_device * lrc_to_xe(struct xe_lrc *lrc) @@ -97,7 +115,6 @@ static void set_offsets(u32 *regs, #define REG16(x) \ (((x) >> 9) | BIT(7) | BUILD_BUG_ON_ZERO(x >= 0x10000)), \ (((x) >> 2) & 0x7f) -#define END 0 { const u32 base = hwe->mmio_base; @@ -168,7 +185,7 @@ static const u8 gen12_xcs_offsets[] = { REG16(0x274), REG16(0x270), - END + 0 }; static const u8 dg2_xcs_offsets[] = { @@ -202,7 +219,7 @@ static const u8 dg2_xcs_offsets[] = { REG16(0x274), REG16(0x270), - END + 0 }; static const u8 gen12_rcs_offsets[] = { @@ -298,7 +315,7 @@ static const u8 gen12_rcs_offsets[] = { REG(0x084), NOP(1), - END + 0 }; static const u8 xehp_rcs_offsets[] = { @@ -339,7 +356,7 @@ static const u8 xehp_rcs_offsets[] = { LRI(1, 0), REG(0x0c8), - END + 0 }; static const u8 dg2_rcs_offsets[] = { @@ -382,7 +399,7 @@ static const u8 dg2_rcs_offsets[] = { LRI(1, 0), REG(0x0c8), - END + 0 }; static const u8 mtl_rcs_offsets[] = { @@ -425,7 +442,7 @@ static const u8 mtl_rcs_offsets[] = { LRI(1, 0), REG(0x0c8), - END + 0 }; #define XE2_CTX_COMMON \ @@ -471,7 +488,7 @@ static const u8 xe2_rcs_offsets[] = { LRI(1, 0), /* [0x47] */ REG(0x0c8), /* [0x48] R_PWR_CLK_STATE */ - END + 0 }; static const u8 xe2_bcs_offsets[] = { @@ -482,16 +499,15 @@ static const u8 xe2_bcs_offsets[] = { REG16(0x200), /* [0x42] BCS_SWCTRL */ REG16(0x204), /* [0x44] BLIT_CCTL */ - END + 0 }; static const u8 xe2_xcs_offsets[] = { XE2_CTX_COMMON, - END + 0 }; -#undef END #undef REG16 #undef REG #undef LRI @@ -527,9 +543,8 @@ static const u8 *reg_offsets(struct xe_device *xe, enum xe_engine_class class) static void set_context_control(u32 *regs, struct xe_hw_engine *hwe) { - regs[CTX_CONTEXT_CONTROL] = _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH) | - _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT) | - CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT; + regs[CTX_CONTEXT_CONTROL] = _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH | + CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT); /* TODO: Timestamp */ } @@ -637,7 +652,7 @@ static inline struct iosys_map __xe_lrc_##elem##_map(struct xe_lrc *lrc) \ iosys_map_incr(&map, __xe_lrc_##elem##_offset(lrc)); \ return map; \ } \ -static inline u32 __xe_lrc_##elem##_ggtt_addr(struct xe_lrc *lrc) \ +static inline u32 __maybe_unused __xe_lrc_##elem##_ggtt_addr(struct xe_lrc *lrc) \ { \ return xe_bo_ggtt_addr(lrc->bo) + __xe_lrc_##elem##_offset(lrc); \ } \ @@ -727,8 +742,9 @@ int xe_lrc_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe, lrc->bo = xe_bo_create_pin_map(xe, tile, vm, ring_size + xe_lrc_size(xe, hwe->class), ttm_bo_type_kernel, - XE_BO_CREATE_VRAM_IF_DGFX(tile) | - XE_BO_CREATE_GGTT_BIT); + XE_BO_FLAG_VRAM_IF_DGFX(tile) | + XE_BO_FLAG_GGTT | + XE_BO_FLAG_GGTT_INVALIDATE); if (IS_ERR(lrc->bo)) return PTR_ERR(lrc->bo); @@ -779,7 +795,7 @@ int xe_lrc_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe, xe_lrc_write_ctx_reg(lrc, PVC_CTX_ASID, vm->usm.asid); lrc->desc = LRC_VALID; - lrc->desc |= LRC_LEGACY_64B_CONTEXT << LRC_ADDRESSING_MODE_SHIFT; + lrc->desc |= FIELD_PREP(LRC_ADDRESSING_MODE, LRC_LEGACY_64B_CONTEXT); /* TODO: Priority */ /* While this appears to have something about privileged batches or @@ -789,8 +805,8 @@ int xe_lrc_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe, lrc->desc |= LRC_PRIVILEGE; if (GRAPHICS_VERx100(xe) < 1250) { - lrc->desc |= (u64)hwe->instance << ENGINE_INSTANCE_SHIFT; - lrc->desc |= (u64)hwe->class << ENGINE_CLASS_SHIFT; + lrc->desc |= FIELD_PREP(LRC_ENGINE_INSTANCE, hwe->instance); + lrc->desc |= FIELD_PREP(LRC_ENGINE_CLASS, hwe->class); } arb_enable = MI_ARB_ON_OFF | MI_ARB_ENABLE; @@ -1037,6 +1053,8 @@ static int dump_gfxpipe_command(struct drm_printer *p, MATCH(GPGPU_CSR_BASE_ADDRESS); MATCH(STATE_COMPUTE_MODE); MATCH3D(3DSTATE_BTD); + MATCH(STATE_SYSTEM_MEM_FENCE_ADDRESS); + MATCH(STATE_CONTEXT_DATA_BASE_ADDRESS); MATCH3D(3DSTATE_VF_STATISTICS); @@ -1061,6 +1079,7 @@ static int dump_gfxpipe_command(struct drm_printer *p, MATCH3D(3DSTATE_WM); MATCH3D(3DSTATE_CONSTANT_VS); MATCH3D(3DSTATE_CONSTANT_GS); + MATCH3D(3DSTATE_CONSTANT_PS); MATCH3D(3DSTATE_SAMPLE_MASK); MATCH3D(3DSTATE_CONSTANT_HS); MATCH3D(3DSTATE_CONSTANT_DS); @@ -1153,6 +1172,31 @@ static int dump_gfxpipe_command(struct drm_printer *p, } } +static int dump_gfx_state_command(struct drm_printer *p, + struct xe_gt *gt, + u32 *dw, + int remaining_dw) +{ + u32 numdw = instr_dw(*dw); + u32 opcode = REG_FIELD_GET(GFX_STATE_OPCODE, *dw); + + /* + * Make sure we haven't mis-parsed a number of dwords that exceeds the + * remaining size of the LRC. + */ + if (xe_gt_WARN_ON(gt, numdw > remaining_dw)) + numdw = remaining_dw; + + switch (*dw & (XE_INSTR_GFX_STATE | GFX_STATE_OPCODE)) { + MATCH(STATE_WRITE_INLINE); + + default: + drm_printf(p, "[%#010x] unknown GFX_STATE command (opcode=%#x), likely %d dwords\n", + *dw, opcode, numdw); + return numdw; + } +} + void xe_lrc_dump_default(struct drm_printer *p, struct xe_gt *gt, enum xe_engine_class hwe_class) @@ -1177,6 +1221,8 @@ void xe_lrc_dump_default(struct drm_printer *p, num_dw = dump_mi_command(p, gt, dw, remaining_dw); } else if ((*dw & XE_INSTR_CMD_TYPE) == XE_INSTR_GFXPIPE) { num_dw = dump_gfxpipe_command(p, gt, dw, remaining_dw); + } else if ((*dw & XE_INSTR_CMD_TYPE) == XE_INSTR_GFX_STATE) { + num_dw = dump_gfx_state_command(p, gt, dw, remaining_dw); } else { num_dw = min(instr_dw(*dw), remaining_dw); drm_printf(p, "[%#10x] Unknown instruction of type %#x, likely %d dwords\n", @@ -1300,3 +1346,101 @@ void xe_lrc_emit_hwe_state_instructions(struct xe_exec_queue *q, struct xe_bb *b bb->len += num_dw; } } + +struct xe_lrc_snapshot *xe_lrc_snapshot_capture(struct xe_lrc *lrc) +{ + struct xe_lrc_snapshot *snapshot = kmalloc(sizeof(*snapshot), GFP_NOWAIT); + + if (!snapshot) + return NULL; + + snapshot->context_desc = lower_32_bits(xe_lrc_ggtt_addr(lrc)); + snapshot->head = xe_lrc_ring_head(lrc); + snapshot->tail.internal = lrc->ring.tail; + snapshot->tail.memory = xe_lrc_read_ctx_reg(lrc, CTX_RING_TAIL); + snapshot->start_seqno = xe_lrc_start_seqno(lrc); + snapshot->seqno = xe_lrc_seqno(lrc); + snapshot->lrc_bo = xe_bo_get(lrc->bo); + snapshot->lrc_offset = xe_lrc_pphwsp_offset(lrc); + snapshot->lrc_size = lrc->bo->size - snapshot->lrc_offset; + snapshot->lrc_snapshot = NULL; + return snapshot; +} + +void xe_lrc_snapshot_capture_delayed(struct xe_lrc_snapshot *snapshot) +{ + struct xe_bo *bo; + struct iosys_map src; + + if (!snapshot) + return; + + bo = snapshot->lrc_bo; + snapshot->lrc_bo = NULL; + + snapshot->lrc_snapshot = kvmalloc(snapshot->lrc_size, GFP_KERNEL); + if (!snapshot->lrc_snapshot) + goto put_bo; + + dma_resv_lock(bo->ttm.base.resv, NULL); + if (!ttm_bo_vmap(&bo->ttm, &src)) { + xe_map_memcpy_from(xe_bo_device(bo), + snapshot->lrc_snapshot, &src, snapshot->lrc_offset, + snapshot->lrc_size); + ttm_bo_vunmap(&bo->ttm, &src); + } else { + kvfree(snapshot->lrc_snapshot); + snapshot->lrc_snapshot = NULL; + } + dma_resv_unlock(bo->ttm.base.resv); +put_bo: + xe_bo_put(bo); +} + +void xe_lrc_snapshot_print(struct xe_lrc_snapshot *snapshot, struct drm_printer *p) +{ + unsigned long i; + + if (!snapshot) + return; + + drm_printf(p, "\tHW Context Desc: 0x%08x\n", snapshot->context_desc); + drm_printf(p, "\tLRC Head: (memory) %u\n", snapshot->head); + drm_printf(p, "\tLRC Tail: (internal) %u, (memory) %u\n", + snapshot->tail.internal, snapshot->tail.memory); + drm_printf(p, "\tStart seqno: (memory) %d\n", snapshot->start_seqno); + drm_printf(p, "\tSeqno: (memory) %d\n", snapshot->seqno); + + if (!snapshot->lrc_snapshot) + return; + + drm_printf(p, "\t[HWSP].length: 0x%x\n", LRC_PPHWSP_SIZE); + drm_puts(p, "\t[HWSP].data: "); + for (i = 0; i < LRC_PPHWSP_SIZE; i += sizeof(u32)) { + u32 *val = snapshot->lrc_snapshot + i; + char dumped[ASCII85_BUFSZ]; + + drm_puts(p, ascii85_encode(*val, dumped)); + } + + drm_printf(p, "\n\t[HWCTX].length: 0x%lx\n", snapshot->lrc_size - LRC_PPHWSP_SIZE); + drm_puts(p, "\t[HWCTX].data: "); + for (; i < snapshot->lrc_size; i += sizeof(u32)) { + u32 *val = snapshot->lrc_snapshot + i; + char dumped[ASCII85_BUFSZ]; + + drm_puts(p, ascii85_encode(*val, dumped)); + } + drm_puts(p, "\n"); +} + +void xe_lrc_snapshot_free(struct xe_lrc_snapshot *snapshot) +{ + if (!snapshot) + return; + + kvfree(snapshot->lrc_snapshot); + if (snapshot->lrc_bo) + xe_bo_put(snapshot->lrc_bo); + kfree(snapshot); +} |