drm/msm: Improved hang detection
Subsystems (4)
cover drm/msm: Improved hang detection
From: Rob Clark <[email protected]>
Try to detect when submit jobs are making forward progress and give them a bit more time.
Rob Clark (2):
drm/msm/adreno: Simplify read64/write64 helpers
drm/msm: Hangcheck progress detection
drivers/gpu/drm/msm/adreno/a4xx_gpu.c | 3 +- drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 27 ++++------ drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 4 +- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 58 +++++++++++++++------ drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c | 3 +- drivers/gpu/drm/msm/msm_drv.h | 8 ++- drivers/gpu/drm/msm/msm_gpu.c | 20 ++++++- drivers/gpu/drm/msm/msm_gpu.h | 17 +++--- drivers/gpu/drm/msm/msm_ringbuffer.h | 24 +++++++++ 9 files changed, 115 insertions(+), 49 deletions(-)
Signature · 1 line
2.37.3
PATCH 1/2 drm/msm/adreno: Simplify read64/write64 helpers
From: Rob Clark <[email protected]>
The _HI reg is always following the _LO reg, so no need to pass these offsets seprately.
--- drivers/gpu/drm/msm/adreno/a4xx_gpu.c | 3 +-- drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 27 ++++++++------------- drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 4 +-- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 24 ++++++------------ drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c | 3 +-- drivers/gpu/drm/msm/msm_gpu.h | 12 ++++----- 6 files changed, 27 insertions(+), 46 deletions(-)
drivers/gpu/drm/msm/adreno/a4xx_gpu.c +1 -2
@@ -606,8 +606,7 @@ static int a4xx_pm_suspend(struct msm_gpu *gpu) {static int a4xx_get_timestamp (struct msm_gpu *gpu,uint64_t *value) { - *value =gpu_read64 (gpu,REG_A4XX_RBBM_PERFCTR_CP_0_LO , -REG_A4XX_RBBM_PERFCTR_CP_0_HI ); + *value =gpu_read64 (gpu,REG_A4XX_RBBM_PERFCTR_CP_0_LO );return 0 ; }
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +10 -17
@@ -605,11 +605,9 @@ static int a5xx_ucode_init(struct msm_gpu *gpu)a5xx_ucode_check_version (a5xx_gpu, a5xx_gpu->pfp_bo); } -gpu_write64 (gpu,REG_A5XX_CP_ME_INSTR_BASE_LO , -REG_A5XX_CP_ME_INSTR_BASE_HI , a5xx_gpu->pm4_iova); +gpu_write64 (gpu,REG_A5XX_CP_ME_INSTR_BASE_LO , a5xx_gpu->pm4_iova); -gpu_write64 (gpu,REG_A5XX_CP_PFP_INSTR_BASE_LO , -REG_A5XX_CP_PFP_INSTR_BASE_HI , a5xx_gpu->pfp_iova); +gpu_write64 (gpu,REG_A5XX_CP_PFP_INSTR_BASE_LO , a5xx_gpu->pfp_iova);return 0 ; }
@@ -868,8 +866,7 @@ static int a5xx_hw_init(struct msm_gpu *gpu)* memory rendering at this point in time and we don't want to block off * part of the virtual memory space. */ -gpu_write64 (gpu,REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO , -REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_HI ,0x00000000 ); +gpu_write64 (gpu,REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO ,0x00000000 );gpu_write (gpu,REG_A5XX_RBBM_SECVID_TSB_TRUSTED_SIZE ,0x00000000 );/* Put the GPU into 64 bit by default */
@@ -908,8 +905,7 @@ static int a5xx_hw_init(struct msm_gpu *gpu)return ret;/* Set the ringbuffer address */ -gpu_write64 (gpu,REG_A5XX_CP_RB_BASE ,REG_A5XX_CP_RB_BASE_HI , - gpu->rb[0 ]->iova); +gpu_write64 (gpu,REG_A5XX_CP_RB_BASE , gpu->rb[0 ]->iova);/* * If the microcode supports the WHERE_AM_I opcode then we can use that
@@ -936,7 +932,7 @@ static int a5xx_hw_init(struct msm_gpu *gpu) }gpu_write64 (gpu,REG_A5XX_CP_RB_RPTR_ADDR , -REG_A5XX_CP_RB_RPTR_ADDR_HI ,shadowptr (a5xx_gpu, gpu->rb[0 ])); +shadowptr (a5xx_gpu, gpu->rb[0 ])); }else if (gpu->nr_rings >1 ) {/* Disable preemption if WHERE_AM_I isn't available */ a5xx_preempt_fini (gpu);
@@ -1239,9 +1235,9 @@ static void a5xx_fault_detect_irq(struct msm_gpu *gpu)gpu_read (gpu,REG_A5XX_RBBM_STATUS ),gpu_read (gpu,REG_A5XX_CP_RB_RPTR ),gpu_read (gpu,REG_A5XX_CP_RB_WPTR ), -gpu_read64 (gpu,REG_A5XX_CP_IB1_BASE ,REG_A5XX_CP_IB1_BASE_HI ), +gpu_read64 (gpu,REG_A5XX_CP_IB1_BASE ),gpu_read (gpu,REG_A5XX_CP_IB1_BUFSZ ), -gpu_read64 (gpu,REG_A5XX_CP_IB2_BASE ,REG_A5XX_CP_IB2_BASE_HI ), +gpu_read64 (gpu,REG_A5XX_CP_IB2_BASE ), gpu_read(gpu, REG_A5XX_CP_IB2_BUFSZ));/* Turn off the hangcheck timer to keep it from bothering us */
@@ -1427,8 +1423,7 @@ static int a5xx_pm_suspend(struct msm_gpu *gpu)static int a5xx_get_timestamp (struct msm_gpu *gpu,uint64_t *value) { - *value =gpu_read64 (gpu,REG_A5XX_RBBM_ALWAYSON_COUNTER_LO , -REG_A5XX_RBBM_ALWAYSON_COUNTER_HI ); + *value =gpu_read64 (gpu,REG_A5XX_RBBM_ALWAYSON_COUNTER_LO );return 0 ; }
@@ -1465,8 +1460,7 @@ static int a5xx_crashdumper_run(struct msm_gpu *gpu,if (IS_ERR_OR_NULL (dumper->ptr))return -EINVAL ; -gpu_write64 (gpu,REG_A5XX_CP_CRASH_SCRIPT_BASE_LO , -REG_A5XX_CP_CRASH_SCRIPT_BASE_HI , dumper->iova); +gpu_write64 (gpu,REG_A5XX_CP_CRASH_SCRIPT_BASE_LO , dumper->iova);gpu_write (gpu,REG_A5XX_CP_CRASH_DUMP_CNTL ,1 );
@@ -1666,8 +1660,7 @@ static u64 a5xx_gpu_busy(struct msm_gpu *gpu, unsigned long *out_sample_rate) {u64 busy_cycles; - busy_cycles =gpu_read64 (gpu,REG_A5XX_RBBM_PERFCTR_RBBM_0_LO , -REG_A5XX_RBBM_PERFCTR_RBBM_0_HI ); + busy_cycles =gpu_read64 (gpu,REG_A5XX_RBBM_PERFCTR_RBBM_0_LO ); *out_sample_rate =clk_get_rate (gpu->core_clk);return busy_cycles;
drivers/gpu/drm/msm/adreno/a5xx_preempt.c +1 -3
@@ -137,7 +137,6 @@ void a5xx_preempt_trigger(struct msm_gpu *gpu)/* Set the address of the incoming preemption record */ gpu_write64 (gpu,REG_A5XX_CP_CONTEXT_SWITCH_RESTORE_ADDR_LO , -REG_A5XX_CP_CONTEXT_SWITCH_RESTORE_ADDR_HI , a5xx_gpu->preempt_iova[ring->id]); a5xx_gpu->next_ring = ring;
@@ -211,8 +210,7 @@ void a5xx_preempt_hw_init(struct msm_gpu *gpu) }/* Write a 0 to signal that we aren't switching pagetables */ -gpu_write64 (gpu,REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_LO , -REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_HI ,0 ); +gpu_write64 (gpu,REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_LO ,0 );/* Reset the preemption state */ set_preempt_state (a5xx_gpu,PREEMPT_NONE );
drivers/gpu/drm/msm/adreno/a6xx_gpu.c +8 -16
@@ -247,8 +247,7 @@ static void a6xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)OUT_RING (ring, submit->seqno);trace_msm_gpu_submit_flush (submit, -gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO , -REG_A6XX_CP_ALWAYS_ON_COUNTER_HI )); +gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO ));a6xx_flush (gpu, ring); }
@@ -947,8 +946,7 @@ static int a6xx_ucode_init(struct msm_gpu *gpu) } } -gpu_write64 (gpu,REG_A6XX_CP_SQE_INSTR_BASE , -REG_A6XX_CP_SQE_INSTR_BASE +1 , a6xx_gpu->sqe_iova); +gpu_write64 (gpu,REG_A6XX_CP_SQE_INSTR_BASE , a6xx_gpu->sqe_iova);return 0 ; }
@@ -999,8 +997,7 @@ static int hw_init(struct msm_gpu *gpu)* memory rendering at this point in time and we don't want to block off * part of the virtual memory space. */ -gpu_write64 (gpu,REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO , -REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_HI ,0x00000000 ); +gpu_write64 (gpu,REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO ,0x00000000 );gpu_write (gpu,REG_A6XX_RBBM_SECVID_TSB_TRUSTED_SIZE ,0x00000000 );/* Turn on 64 bit addressing for all blocks */
@@ -1049,11 +1046,9 @@ static int hw_init(struct msm_gpu *gpu)if (!adreno_is_a650_family (adreno_gpu)) {/* Set the GMEM VA range [0x100000:0x100000 + gpu->gmem - 1] */ -gpu_write64 (gpu,REG_A6XX_UCHE_GMEM_RANGE_MIN_LO , -REG_A6XX_UCHE_GMEM_RANGE_MIN_HI ,0x00100000 ); +gpu_write64 (gpu,REG_A6XX_UCHE_GMEM_RANGE_MIN_LO ,0x00100000 );gpu_write64 (gpu,REG_A6XX_UCHE_GMEM_RANGE_MAX_LO , -REG_A6XX_UCHE_GMEM_RANGE_MAX_HI ,0x00100000 + adreno_gpu->gmem -1 ); }
@@ -1145,8 +1140,7 @@ static int hw_init(struct msm_gpu *gpu) goto out;/* Set the ringbuffer address */ -gpu_write64 (gpu,REG_A6XX_CP_RB_BASE ,REG_A6XX_CP_RB_BASE_HI , - gpu->rb[0 ]->iova); +gpu_write64 (gpu,REG_A6XX_CP_RB_BASE , gpu->rb[0 ]->iova);/* Targets that support extended APRIV can use the RPTR shadow from * hardware but all the other ones need to disable the feature. Targets
@@ -1178,7 +1172,6 @@ static int hw_init(struct msm_gpu *gpu) }gpu_write64 (gpu,REG_A6XX_CP_RB_RPTR_ADDR_LO , -REG_A6XX_CP_RB_RPTR_ADDR_HI ,shadowptr (a6xx_gpu, gpu->rb[0 ])); }
@@ -1499,9 +1492,9 @@ static void a6xx_fault_detect_irq(struct msm_gpu *gpu)gpu_read (gpu,REG_A6XX_RBBM_STATUS ),gpu_read (gpu,REG_A6XX_CP_RB_RPTR ),gpu_read (gpu,REG_A6XX_CP_RB_WPTR ), -gpu_read64 (gpu,REG_A6XX_CP_IB1_BASE ,REG_A6XX_CP_IB1_BASE_HI ), +gpu_read64 (gpu,REG_A6XX_CP_IB1_BASE ),gpu_read (gpu,REG_A6XX_CP_IB1_REM_SIZE ), -gpu_read64 (gpu,REG_A6XX_CP_IB2_BASE ,REG_A6XX_CP_IB2_BASE_HI ), +gpu_read64 (gpu,REG_A6XX_CP_IB2_BASE ), gpu_read(gpu, REG_A6XX_CP_IB2_REM_SIZE));/* Turn off the hangcheck timer to keep it from bothering us */
@@ -1712,8 +1705,7 @@ static int a6xx_get_timestamp(struct msm_gpu *gpu, uint64_t *value)/* Force the GPU power on so we can read this register */ a6xx_gmu_set_oob (&a6xx_gpu->gmu,GMU_OOB_PERFCOUNTER_SET ); - *value =gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO , -REG_A6XX_CP_ALWAYS_ON_COUNTER_HI ); + *value =gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO );a6xx_gmu_clear_oob (&a6xx_gpu->gmu,GMU_OOB_PERFCOUNTER_SET );
drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c +1 -2
@@ -147,8 +147,7 @@ static int a6xx_crashdumper_run(struct msm_gpu *gpu,/* Make sure all pending memory writes are posted */ wmb (); -gpu_write64 (gpu,REG_A6XX_CP_CRASH_SCRIPT_BASE_LO , -REG_A6XX_CP_CRASH_SCRIPT_BASE_HI , dumper->iova); +gpu_write64 (gpu,REG_A6XX_CP_CRASH_SCRIPT_BASE_LO , dumper->iova);gpu_write (gpu,REG_A6XX_CP_CRASH_DUMP_CNTL ,1 );
drivers/gpu/drm/msm/msm_gpu.h +6 -6
@@ -540,7 +540,7 @@ static inline void gpu_rmw(struct msm_gpu *gpu, u32 reg, u32 mask, u32 or)msm_rmw (gpu->mmio + (reg <<2 ), mask, or); } -static inline u64 gpu_read64 (struct msm_gpu *gpu,u32 lo,u32 hi) +static inline u64 gpu_read64 (struct msm_gpu *gpu,u32 reg) {u64 val;
@@ -558,17 +558,17 @@ static inline u64 gpu_read64(struct msm_gpu *gpu, u32 lo, u32 hi)* when the lo is read, so make sure to read the lo first to trigger * that */ - val = (u64 )msm_readl (gpu->mmio + (lo <<2 )); - val |= ((u64 )msm_readl (gpu->mmio + (hi <<2 )) <<32 ); + val = (u64 )msm_readl (gpu->mmio + (reg <<2 )); + val |= ((u64 )msm_readl (gpu->mmio + ((reg +1 ) <<2 )) <<32 );return val; } -static inline void gpu_write64 (struct msm_gpu *gpu,u32 lo,u32 hi,u64 val) +static inline void gpu_write64 (struct msm_gpu *gpu,u32 reg,u64 val) {/* Why not a writeq here? Read the screed above */ -msm_writel (lower_32_bits (val), gpu->mmio + (lo <<2 )); -msm_writel (upper_32_bits (val), gpu->mmio + (hi <<2 )); +msm_writel (lower_32_bits (val), gpu->mmio + (reg <<2 )); +msm_writel (upper_32_bits (val), gpu->mmio + ((reg +1 ) <<2 )); }int msm_gpu_pm_suspend (struct msm_gpu *gpu);
Signature · 1 line
2.37.3
Re: [PATCH 1/2] drm/msm/adreno: Simplify read64/write64 helpers
Quoted 6 lines · Rob Clark
From: Rob Clark <[email protected]>
The
_HIreg is always following the_LOreg, so no need to pass these offsets seprately.Signed-off-by Rob Clark <[email protected]>
Quoted 8 lines
--- drivers/gpu/drm/msm/adreno/a4xx_gpu.c | 3 +-- drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 27 ++++++++------------- drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 4 +-- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 24 ++++++------------ drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c | 3 +-- drivers/gpu/drm/msm/msm_gpu.h | 12 ++++----- 6 files changed, 27 insertions(+), 46 deletions(-)
Signature · 2 lines
With best wishes Dmitry
Re: [PATCH 1/2] drm/msm/adreno: Simplify read64/write64 helpers
In reply to Rob ClarkQuoted 267 lines · Rob Clark
From: Rob Clark <[email protected]>
The
_HIreg is always following the_LOreg, so no need to pass these offsets seprately.Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/adreno/a4xx_gpu.c | 3 +-- drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 27 ++++++++------------- drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 4 +-- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 24 ++++++------------ drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c | 3 +-- drivers/gpu/drm/msm/msm_gpu.h | 12 ++++----- 6 files changed, 27 insertions(+), 46 deletions(-)drivers/gpu/drm/msm/adreno/a4xx_gpu.c +1
-2@@ -606,8 +606,7 @@ static int a4xx_pm_suspend(struct msm_gpu *gpu) {static int a4xx_get_timestamp (struct msm_gpu *gpu,uint64_t *value) { - *value =gpu_read64 (gpu,REG_A4XX_RBBM_PERFCTR_CP_0_LO , -REG_A4XX_RBBM_PERFCTR_CP_0_HI ); + *value =gpu_read64 (gpu,REG_A4XX_RBBM_PERFCTR_CP_0_LO );return 0 ; }drivers/gpu/drm/msm/adreno/a5xx_gpu.c +10
-17@@ -605,11 +605,9 @@ static int a5xx_ucode_init(struct msm_gpu *gpu)a5xx_ucode_check_version (a5xx_gpu, a5xx_gpu->pfp_bo); } -gpu_write64 (gpu,REG_A5XX_CP_ME_INSTR_BASE_LO , -REG_A5XX_CP_ME_INSTR_BASE_HI , a5xx_gpu->pm4_iova); +gpu_write64 (gpu,REG_A5XX_CP_ME_INSTR_BASE_LO , a5xx_gpu->pm4_iova); -gpu_write64 (gpu,REG_A5XX_CP_PFP_INSTR_BASE_LO , -REG_A5XX_CP_PFP_INSTR_BASE_HI , a5xx_gpu->pfp_iova); +gpu_write64 (gpu,REG_A5XX_CP_PFP_INSTR_BASE_LO , a5xx_gpu->pfp_iova);return 0 ; }@@ -868,8 +866,7 @@ static int a5xx_hw_init(struct msm_gpu *gpu)* memory rendering at this point in time and we don't want to block off * part of the virtual memory space. */ -gpu_write64 (gpu,REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO , -REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_HI ,0x00000000 ); +gpu_write64 (gpu,REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO ,0x00000000 );gpu_write (gpu,REG_A5XX_RBBM_SECVID_TSB_TRUSTED_SIZE ,0x00000000 );/* Put the GPU into 64 bit by default */ @@ -908,8 +905,7 @@ static int a5xx_hw_init(struct msm_gpu *gpu)return ret;/* Set the ringbuffer address */ -gpu_write64 (gpu,REG_A5XX_CP_RB_BASE ,REG_A5XX_CP_RB_BASE_HI , - gpu->rb[0 ]->iova); +gpu_write64 (gpu,REG_A5XX_CP_RB_BASE , gpu->rb[0 ]->iova);/* * If the microcode supports the WHERE_AM_I opcode then we can use that @@ -936,7 +932,7 @@ static int a5xx_hw_init(struct msm_gpu *gpu) }gpu_write64 (gpu,REG_A5XX_CP_RB_RPTR_ADDR , -REG_A5XX_CP_RB_RPTR_ADDR_HI ,shadowptr (a5xx_gpu, gpu->rb[0 ])); +shadowptr (a5xx_gpu, gpu->rb[0 ])); }else if (gpu->nr_rings >1 ) {/* Disable preemption if WHERE_AM_I isn't available */ a5xx_preempt_fini (gpu);@@ -1239,9 +1235,9 @@ static void a5xx_fault_detect_irq(struct msm_gpu *gpu)gpu_read (gpu,REG_A5XX_RBBM_STATUS ),gpu_read (gpu,REG_A5XX_CP_RB_RPTR ),gpu_read (gpu,REG_A5XX_CP_RB_WPTR ), -gpu_read64 (gpu,REG_A5XX_CP_IB1_BASE ,REG_A5XX_CP_IB1_BASE_HI ), +gpu_read64 (gpu,REG_A5XX_CP_IB1_BASE ),gpu_read (gpu,REG_A5XX_CP_IB1_BUFSZ ), -gpu_read64 (gpu,REG_A5XX_CP_IB2_BASE ,REG_A5XX_CP_IB2_BASE_HI ), +gpu_read64 (gpu,REG_A5XX_CP_IB2_BASE ), gpu_read(gpu, REG_A5XX_CP_IB2_BUFSZ));/* Turn off the hangcheck timer to keep it from bothering us */ @@ -1427,8 +1423,7 @@ static int a5xx_pm_suspend(struct msm_gpu *gpu)static int a5xx_get_timestamp (struct msm_gpu *gpu,uint64_t *value) { - *value =gpu_read64 (gpu,REG_A5XX_RBBM_ALWAYSON_COUNTER_LO , -REG_A5XX_RBBM_ALWAYSON_COUNTER_HI ); + *value =gpu_read64 (gpu,REG_A5XX_RBBM_ALWAYSON_COUNTER_LO );return 0 ; }@@ -1465,8 +1460,7 @@ static int a5xx_crashdumper_run(struct msm_gpu *gpu,if (IS_ERR_OR_NULL (dumper->ptr))return -EINVAL ; -gpu_write64 (gpu,REG_A5XX_CP_CRASH_SCRIPT_BASE_LO , -REG_A5XX_CP_CRASH_SCRIPT_BASE_HI , dumper->iova); +gpu_write64 (gpu,REG_A5XX_CP_CRASH_SCRIPT_BASE_LO , dumper->iova);gpu_write (gpu,REG_A5XX_CP_CRASH_DUMP_CNTL ,1 );@@ -1666,8 +1660,7 @@ static u64 a5xx_gpu_busy(struct msm_gpu *gpu, unsigned long *out_sample_rate) {u64 busy_cycles; - busy_cycles =gpu_read64 (gpu,REG_A5XX_RBBM_PERFCTR_RBBM_0_LO , -REG_A5XX_RBBM_PERFCTR_RBBM_0_HI ); + busy_cycles =gpu_read64 (gpu,REG_A5XX_RBBM_PERFCTR_RBBM_0_LO ); *out_sample_rate =clk_get_rate (gpu->core_clk);return busy_cycles;drivers/gpu/drm/msm/adreno/a5xx_preempt.c +1
-3@@ -137,7 +137,6 @@ void a5xx_preempt_trigger(struct msm_gpu *gpu)/* Set the address of the incoming preemption record */ gpu_write64 (gpu,REG_A5XX_CP_CONTEXT_SWITCH_RESTORE_ADDR_LO , -REG_A5XX_CP_CONTEXT_SWITCH_RESTORE_ADDR_HI , a5xx_gpu->preempt_iova[ring->id]); a5xx_gpu->next_ring = ring;@@ -211,8 +210,7 @@ void a5xx_preempt_hw_init(struct msm_gpu *gpu) }/* Write a 0 to signal that we aren't switching pagetables */ -gpu_write64 (gpu,REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_LO , -REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_HI ,0 ); +gpu_write64 (gpu,REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_LO ,0 );/* Reset the preemption state */ set_preempt_state (a5xx_gpu,PREEMPT_NONE );drivers/gpu/drm/msm/adreno/a6xx_gpu.c +8
-16@@ -247,8 +247,7 @@ static void a6xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)OUT_RING (ring, submit->seqno);trace_msm_gpu_submit_flush (submit, -gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO , -REG_A6XX_CP_ALWAYS_ON_COUNTER_HI )); +gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO ));a6xx_flush (gpu, ring); }@@ -947,8 +946,7 @@ static int a6xx_ucode_init(struct msm_gpu *gpu) } } -gpu_write64 (gpu,REG_A6XX_CP_SQE_INSTR_BASE , -REG_A6XX_CP_SQE_INSTR_BASE +1 , a6xx_gpu->sqe_iova); +gpu_write64 (gpu,REG_A6XX_CP_SQE_INSTR_BASE , a6xx_gpu->sqe_iova);return 0 ; }@@ -999,8 +997,7 @@ static int hw_init(struct msm_gpu *gpu)* memory rendering at this point in time and we don't want to block off * part of the virtual memory space. */ -gpu_write64 (gpu,REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO , -REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_HI ,0x00000000 ); +gpu_write64 (gpu,REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO ,0x00000000 );gpu_write (gpu,REG_A6XX_RBBM_SECVID_TSB_TRUSTED_SIZE ,0x00000000 );/* Turn on 64 bit addressing for all blocks */ @@ -1049,11 +1046,9 @@ static int hw_init(struct msm_gpu *gpu)if (!adreno_is_a650_family (adreno_gpu)) {/* Set the GMEM VA range [0x100000:0x100000 + gpu->gmem - 1] */ -gpu_write64 (gpu,REG_A6XX_UCHE_GMEM_RANGE_MIN_LO , -REG_A6XX_UCHE_GMEM_RANGE_MIN_HI ,0x00100000 ); +gpu_write64 (gpu,REG_A6XX_UCHE_GMEM_RANGE_MIN_LO ,0x00100000 );gpu_write64 (gpu,REG_A6XX_UCHE_GMEM_RANGE_MAX_LO , -REG_A6XX_UCHE_GMEM_RANGE_MAX_HI ,0x00100000 + adreno_gpu->gmem -1 ); }@@ -1145,8 +1140,7 @@ static int hw_init(struct msm_gpu *gpu) goto out;/* Set the ringbuffer address */ -gpu_write64 (gpu,REG_A6XX_CP_RB_BASE ,REG_A6XX_CP_RB_BASE_HI , - gpu->rb[0 ]->iova); +gpu_write64 (gpu,REG_A6XX_CP_RB_BASE , gpu->rb[0 ]->iova);/* Targets that support extended APRIV can use the RPTR shadow from * hardware but all the other ones need to disable the feature. Targets @@ -1178,7 +1172,6 @@ static int hw_init(struct msm_gpu *gpu) }gpu_write64 (gpu,REG_A6XX_CP_RB_RPTR_ADDR_LO , -REG_A6XX_CP_RB_RPTR_ADDR_HI ,shadowptr (a6xx_gpu, gpu->rb[0 ])); }@@ -1499,9 +1492,9 @@ static void a6xx_fault_detect_irq(struct msm_gpu *gpu)gpu_read (gpu,REG_A6XX_RBBM_STATUS ),gpu_read (gpu,REG_A6XX_CP_RB_RPTR ),gpu_read (gpu,REG_A6XX_CP_RB_WPTR ), -gpu_read64 (gpu,REG_A6XX_CP_IB1_BASE ,REG_A6XX_CP_IB1_BASE_HI ), +gpu_read64 (gpu,REG_A6XX_CP_IB1_BASE ),gpu_read (gpu,REG_A6XX_CP_IB1_REM_SIZE ), -gpu_read64 (gpu,REG_A6XX_CP_IB2_BASE ,REG_A6XX_CP_IB2_BASE_HI ), +gpu_read64 (gpu,REG_A6XX_CP_IB2_BASE ), gpu_read(gpu, REG_A6XX_CP_IB2_REM_SIZE));/* Turn off the hangcheck timer to keep it from bothering us */ @@ -1712,8 +1705,7 @@ static int a6xx_get_timestamp(struct msm_gpu *gpu, uint64_t *value)/* Force the GPU power on so we can read this register */ a6xx_gmu_set_oob (&a6xx_gpu->gmu,GMU_OOB_PERFCOUNTER_SET ); - *value =gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO , -REG_A6XX_CP_ALWAYS_ON_COUNTER_HI ); + *value =gpu_read64 (gpu,REG_A6XX_CP_ALWAYS_ON_COUNTER_LO );a6xx_gmu_clear_oob (&a6xx_gpu->gmu,GMU_OOB_PERFCOUNTER_SET );drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c +1
-2@@ -147,8 +147,7 @@ static int a6xx_crashdumper_run(struct msm_gpu *gpu,/* Make sure all pending memory writes are posted */ wmb (); -gpu_write64 (gpu,REG_A6XX_CP_CRASH_SCRIPT_BASE_LO , -REG_A6XX_CP_CRASH_SCRIPT_BASE_HI , dumper->iova); +gpu_write64 (gpu,REG_A6XX_CP_CRASH_SCRIPT_BASE_LO , dumper->iova);gpu_write (gpu,REG_A6XX_CP_CRASH_DUMP_CNTL ,1 );drivers/gpu/drm/msm/msm_gpu.h +3
-4@@ -540,7 +540,7 @@ static inline void gpu_rmw(struct msm_gpu *gpu, u32 reg, u32 mask, u32 or)msm_rmw (gpu->mmio + (reg <<2 ), mask, or); } -static inline u64 gpu_read64 (struct msm_gpu *gpu,u32 lo,u32 hi) +static inline u64 gpu_read64 (struct msm_gpu *gpu,u32 reg) {u64 val;@@ -558,17 +558,17 @@ static inline u64 gpu_read64(struct msm_gpu *gpu, u32 lo, u32 hi)* when the lo is read, so make sure to read the lo first to trigger * that */ - val = (u64 )msm_readl (gpu->mmio + (lo <<2 )); - val |= ((u64 )msm_readl (gpu->mmio + (hi <<2 )) <<32 ); + val = (u64 )msm_readl (gpu->mmio + (reg <<2 )); + val |= ((u64 )msm_readl (gpu->mmio + ((reg +1 ) <<2 )) <<32 );return val; } -static inline void gpu_write64 (struct msm_gpu *gpu,u32 lo,u32 hi,u64 val)
drivers/gpu/drm/msm/msm_gpu.h static inline u64 gpu_read64(struct msm_gpu *gpu, u32 lo, u32 hi) · line 574
Quoted 10 lines
drivers/gpu/drm/msm/msm_gpu.h +3
-2+static inline void gpu_write64 (struct msm_gpu *gpu,u32 reg,u64 val) {/* Why not a writeq here? Read the screed above */ -msm_writel (lower_32_bits (val), gpu->mmio + (lo <<2 )); -msm_writel (upper_32_bits (val), gpu->mmio + (hi <<2 )); +msm_writel (lower_32_bits (val), gpu->mmio + (reg <<2 )); +msm_writel (upper_32_bits (val), gpu->mmio + ((reg +1 ) <<2 )); }int msm_gpu_pm_suspend (struct msm_gpu *gpu);
-Akhil.
PATCH 2/2 drm/msm: Hangcheck progress detection
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
If the hangcheck timer expires, check if the fw's position in the cmdstream has advanced (changed) since last timer expiration, and allow it up to three additional "extensions" to it's alotted time. The intention is to continue to catch "shader stuck in a loop" type hangs quickly, but allow more time for things that are actually making forward progress.
Because we need to sample the CP state twice to detect if there has not been progress, this also cuts the the timer's duration in half.
--- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 34 +++++++++++++++++++++++++++ drivers/gpu/drm/msm/msm_drv.h | 8 ++++++- drivers/gpu/drm/msm/msm_gpu.c | 20 +++++++++++++++- drivers/gpu/drm/msm/msm_gpu.h | 5 +++- drivers/gpu/drm/msm/msm_ringbuffer.h | 24 +++++++++++++++++++ 5 files changed, 88 insertions(+), 3 deletions(-)
drivers/gpu/drm/msm/adreno/a6xx_gpu.c +34 -0
@@ -1843,6 +1843,39 @@ static uint32_t a6xx_get_rptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring)return ring->memptrs->rptr =gpu_read (gpu,REG_A6XX_CP_RB_RPTR ); } +static bool a6xx_progress (struct msm_gpu *gpu,struct msm_ringbuffer *ring) +{ +struct msm_cp_state cp_state = { + .ib1_base =gpu_read64 (gpu,REG_A6XX_CP_IB1_BASE ), + .ib2_base =gpu_read64 (gpu,REG_A6XX_CP_IB2_BASE ), + .ib1_rem =gpu_read (gpu,REG_A6XX_CP_IB1_REM_SIZE ), + .ib2_rem =gpu_read (gpu,REG_A6XX_CP_IB2_REM_SIZE ), + }; +bool progress; + +/* +* Adjust the remaining data to account for what has already been +* fetched from memory, but not yet consumed by the SQE. +* +* This is not *technically* correct, the amount buffered could +* exceed the IB size due to hw prefetching ahead, but: +* +* (1) We aren't trying to find the exact position, just whether +* progress has been made +* (2) The CP_REG_TO_MEM at the end of a submit should be enough +* to prevent prefetching into an unrelated submit. (And +* either way, at some point the ROQ will be full.) +*/ + cp_state.ib1_rem +=gpu_read (gpu,REG_A6XX_CP_CSQ_IB1_STAT ) >>16 ; + cp_state.ib2_rem +=gpu_read (gpu,REG_A6XX_CP_CSQ_IB1_STAT ) >>16 ; + + progress = !!memcmp (&cp_state, &ring->last_cp_state,sizeof (cp_state)); + + ring->last_cp_state = cp_state; + +return progress; +} +static u32 a618_get_speed_bin (u32 fuse) {if (fuse ==0 )
@@ -1961,6 +1994,7 @@ static const struct adreno_gpu_funcs funcs = { .create_address_space = a6xx_create_address_space, .create_private_address_space = a6xx_create_private_address_space, .get_rptr = a6xx_get_rptr, + .progress = a6xx_progress, }, .get_timestamp = a6xx_get_timestamp, };
drivers/gpu/drm/msm/msm_drv.h +7 -1
@@ -226,7 +226,13 @@ struct msm_drm_private {struct drm_atomic_state *pm_state; -/* For hang detection, in ms */ +/** +* hangcheck_period: For hang detection, in ms +* +* Note that in practice, a submit/job will get at least two hangcheck +* periods, due to checking for progress being implemented as simply +* "have the CP position registers changed since last time?" +*/ unsigned int hangcheck_period;/**
drivers/gpu/drm/msm/msm_gpu.c +19 -1
@@ -500,6 +500,21 @@ static void hangcheck_timer_reset(struct msm_gpu *gpu) round_jiffies_up(jiffies + msecs_to_jiffies(priv->hangcheck_period))); } +static bool made_progress (struct msm_gpu *gpu,struct msm_ringbuffer *ring) +{ +if (ring->hangcheck_progress_retries >=DRM_MSM_HANGCHECK_PROGRESS_RETRIES ) +return false; + +if (!gpu->funcs->progress) +return false; + +if (!gpu->funcs->progress (gpu, ring)) +return false; + + ring->hangcheck_progress_retries++; +return true; +} +static void hangcheck_handler (struct timer_list *t) {struct msm_gpu *gpu =from_timer (gpu, t, hangcheck_timer);
@@ -511,9 +526,12 @@ static void hangcheck_handler(struct timer_list *t)if (fence != ring->hangcheck_fence) {/* some progress has been made.. ya! */ ring->hangcheck_fence = fence; - }else if (fence_before (fence, ring->fctx->last_fence)) { + ring->hangcheck_progress_retries =0 ; + }else if (fence_before (fence, ring->fctx->last_fence) && + !made_progress (gpu, ring)) {/* no progress and not done.. hung! */ ring->hangcheck_fence = fence; + ring->hangcheck_progress_retries =0 ;DRM_DEV_ERROR (dev->dev,"%s: hangcheck detected gpu lockup rb %d!\n" , gpu->name, ring->id); DRM_DEV_ERROR(dev->dev, "%s: completed fence: %u\n",
drivers/gpu/drm/msm/msm_gpu.h +4 -1
@@ -78,6 +78,8 @@ struct msm_gpu_funcs {struct msm_gem_address_space *(*create_private_address_space) (struct msm_gpu *gpu);uint32_t (*get_rptr)(struct msm_gpu *gpu,struct msm_ringbuffer *ring); + +bool (*progress)(struct msm_gpu *gpu,struct msm_ringbuffer *ring); };/* Additional state for iommu faults: */
@@ -236,7 +238,8 @@ struct msm_gpu {*/ #define DRM_MSM_INACTIVE_PERIOD 66/* in ms (roughly four frames) */ -#define DRM_MSM_HANGCHECK_DEFAULT_PERIOD 500/* in ms */ +#define DRM_MSM_HANGCHECK_DEFAULT_PERIOD 250/* in ms */ +#define DRM_MSM_HANGCHECK_PROGRESS_RETRIES 3struct timer_list hangcheck_timer;/* Fault info for most recent iova fault: */
drivers/gpu/drm/msm/msm_ringbuffer.h +24 -0
@@ -35,6 +35,11 @@ struct msm_rbmemptrs {volatile u64 ttbr0; }; +struct msm_cp_state { +uint64_t ib1_base, ib2_base; +uint32_t ib1_rem, ib2_rem; +}; +struct msm_ringbuffer {struct msm_gpu *gpu;int id;
@@ -64,6 +69,25 @@ struct msm_ringbuffer {uint64_t memptrs_iova;struct msm_fence_context *fctx; +/** +* hangcheck_progress_retries: +* +* The number of extra hangcheck duration cycles that we have given +* due to it appearing that the GPU is making forward progress. +* +* If the GPU appears to be making progress (ie. the CP has advanced +* in the command stream, we'll allow up to DRM_MSM_HANGCHECK_PROGRESS_RETRIES +* expirations of the hangcheck timer before killing the job. In other +* words we'll let the submit run for up to +* DRM_MSM_HANGCHECK_DEFAULT_PERIOD * DRM_MSM_HANGCHECK_PROGRESS_RETRIES +*/ +int hangcheck_progress_retries; + +/** +* last_cp_state: The state of the CP at the last call to gpu->progress() +*/ +struct msm_cp_state last_cp_state; +/* * preempt_lock protects preemption and serializes wptr updates against * preemption. Can be aquired from irq context.
Signature · 1 line
2.37.3
Re: [PATCH 2/2] drm/msm: Hangcheck progress detection
Quoted 44 lines · Rob Clark
From: Rob Clark <[email protected]>
If the hangcheck timer expires, check if the fw's position in the cmdstream has advanced (changed) since last timer expiration, and allow it up to three additional "extensions" to it's alotted time. The intention is to continue to catch "shader stuck in a loop" type hangs quickly, but allow more time for things that are actually making forward progress.
Because we need to sample the CP state twice to detect if there has not been progress, this also cuts the the timer's duration in half.
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 34 +++++++++++++++++++++++++++ drivers/gpu/drm/msm/msm_drv.h | 8 ++++++- drivers/gpu/drm/msm/msm_gpu.c | 20 +++++++++++++++- drivers/gpu/drm/msm/msm_gpu.h | 5 +++- drivers/gpu/drm/msm/msm_ringbuffer.h | 24 +++++++++++++++++++ 5 files changed, 88 insertions(+), 3 deletions(-)drivers/gpu/drm/msm/adreno/a6xx_gpu.c +15
-0@@ -1843,6 +1843,39 @@ static uint32_t a6xx_get_rptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring)return ring->memptrs->rptr =gpu_read (gpu,REG_A6XX_CP_RB_RPTR ); } +static bool a6xx_progress (struct msm_gpu *gpu,struct msm_ringbuffer *ring) +{ +struct msm_cp_state cp_state = { + .ib1_base =gpu_read64 (gpu,REG_A6XX_CP_IB1_BASE ), + .ib2_base =gpu_read64 (gpu,REG_A6XX_CP_IB2_BASE ), + .ib1_rem =gpu_read (gpu,REG_A6XX_CP_IB1_REM_SIZE ), + .ib2_rem =gpu_read (gpu,REG_A6XX_CP_IB2_REM_SIZE ), + }; +bool progress; + +/* +* Adjust the remaining data to account for what has already been +* fetched from memory, but not yet consumed by the SQE. +* +* This is not *technically* correct, the amount buffered could
drivers/gpu/drm/msm/adreno/a6xx_gpu.c static uint32_t a6xx_get_rptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring) · line 1870
Quoted 10 lines
drivers/gpu/drm/msm/adreno/a6xx_gpu.c +10
-0+* exceed the IB size due to hw prefetching ahead, but: +* +* (1) We aren't trying to find the exact position, just whether +* progress has been made +* (2) The CP_REG_TO_MEM at the end of a submit should be enough +* to prevent prefetching into an unrelated submit. (And +* either way, at some point the ROQ will be full.) +*/ + cp_state.ib1_rem +=gpu_read (gpu,REG_A6XX_CP_CSQ_IB1_STAT ) >>16 ; + cp_state.ib2_rem +=gpu_read (gpu,REG_A6XX_CP_CSQ_IB1_STAT ) >>16 ;
REG_A6XX_CP_CSQ_IB1_STAT -> REG_A6XX_CP_CSQ_IB2_STAT
With that, Reviewed-by: Akhil P Oommen <[email protected]>
-Akhil.
Quoted 106 lines
drivers/gpu/drm/msm/adreno/a6xx_gpu.c +9
-0+ + progress = !!memcmp (&cp_state, &ring->last_cp_state,sizeof (cp_state)); + + ring->last_cp_state = cp_state; + +return progress; +} +static u32 a618_get_speed_bin (u32 fuse) {if (fuse ==0 )@@ -1961,6 +1994,7 @@ static const struct adreno_gpu_funcs funcs = { .create_address_space = a6xx_create_address_space, .create_private_address_space = a6xx_create_private_address_space, .get_rptr = a6xx_get_rptr, + .progress = a6xx_progress, }, .get_timestamp = a6xx_get_timestamp, };drivers/gpu/drm/msm/msm_drv.h +7
-1@@ -226,7 +226,13 @@ struct msm_drm_private {struct drm_atomic_state *pm_state; -/* For hang detection, in ms */ +/** +* hangcheck_period: For hang detection, in ms +* +* Note that in practice, a submit/job will get at least two hangcheck +* periods, due to checking for progress being implemented as simply +* "have the CP position registers changed since last time?" +*/ unsigned int hangcheck_period;/** drivers/gpu/drm/msm/msm_gpu.c +19
-1@@ -500,6 +500,21 @@ static void hangcheck_timer_reset(struct msm_gpu *gpu) round_jiffies_up(jiffies + msecs_to_jiffies(priv->hangcheck_period))); } +static bool made_progress (struct msm_gpu *gpu,struct msm_ringbuffer *ring) +{ +if (ring->hangcheck_progress_retries >=DRM_MSM_HANGCHECK_PROGRESS_RETRIES ) +return false; + +if (!gpu->funcs->progress) +return false; + +if (!gpu->funcs->progress (gpu, ring)) +return false; + + ring->hangcheck_progress_retries++; +return true; +} +static void hangcheck_handler (struct timer_list *t) {struct msm_gpu *gpu =from_timer (gpu, t, hangcheck_timer);@@ -511,9 +526,12 @@ static void hangcheck_handler(struct timer_list *t)if (fence != ring->hangcheck_fence) {/* some progress has been made.. ya! */ ring->hangcheck_fence = fence; - }else if (fence_before (fence, ring->fctx->last_fence)) { + ring->hangcheck_progress_retries =0 ; + }else if (fence_before (fence, ring->fctx->last_fence) && + !made_progress (gpu, ring)) {/* no progress and not done.. hung! */ ring->hangcheck_fence = fence; + ring->hangcheck_progress_retries =0 ;DRM_DEV_ERROR (dev->dev,"%s: hangcheck detected gpu lockup rb %d!\n" , gpu->name, ring->id); DRM_DEV_ERROR(dev->dev, "%s: completed fence: %u\n",drivers/gpu/drm/msm/msm_gpu.h +4
-1@@ -78,6 +78,8 @@ struct msm_gpu_funcs {struct msm_gem_address_space *(*create_private_address_space) (struct msm_gpu *gpu);uint32_t (*get_rptr)(struct msm_gpu *gpu,struct msm_ringbuffer *ring); + +bool (*progress)(struct msm_gpu *gpu,struct msm_ringbuffer *ring); };/* Additional state for iommu faults: */ @@ -236,7 +238,8 @@ struct msm_gpu {*/ #define DRM_MSM_INACTIVE_PERIOD 66/* in ms (roughly four frames) */ -#define DRM_MSM_HANGCHECK_DEFAULT_PERIOD 500/* in ms */ +#define DRM_MSM_HANGCHECK_DEFAULT_PERIOD 250/* in ms */ +#define DRM_MSM_HANGCHECK_PROGRESS_RETRIES 3struct timer_list hangcheck_timer;/* Fault info for most recent iova fault: */ drivers/gpu/drm/msm/msm_ringbuffer.h +0
-0
drivers/gpu/drm/msm/msm_ringbuffer.h struct msm_rbmemptrs { · line 42
Quoted 8 lines
drivers/gpu/drm/msm/msm_ringbuffer.h +5
-0@@ -35,6 +35,11 @@ struct msm_rbmemptrs {volatile u64 ttbr0; }; +struct msm_cp_state { +uint64_t ib1_base, ib2_base; +uint32_t ib1_rem, ib2_rem; +}; +
I think we can add CP_RB_RPTR too here.
Quoted 29 lines
drivers/gpu/drm/msm/msm_ringbuffer.h +19
-0struct msm_ringbuffer {struct msm_gpu *gpu;int id;@@ -64,6 +69,25 @@ struct msm_ringbuffer {uint64_t memptrs_iova;struct msm_fence_context *fctx; +/** +* hangcheck_progress_retries: +* +* The number of extra hangcheck duration cycles that we have given +* due to it appearing that the GPU is making forward progress. +* +* If the GPU appears to be making progress (ie. the CP has advanced +* in the command stream, we'll allow up to DRM_MSM_HANGCHECK_PROGRESS_RETRIES +* expirations of the hangcheck timer before killing the job. In other +* words we'll let the submit run for up to +* DRM_MSM_HANGCHECK_DEFAULT_PERIOD * DRM_MSM_HANGCHECK_PROGRESS_RETRIES +*/ +int hangcheck_progress_retries; + +/** +* last_cp_state: The state of the CP at the last call to gpu->progress() +*/ +struct msm_cp_state last_cp_state; +/* * preempt_lock protects preemption and serializes wptr updates against * preemption. Can be aquired from irq context.
Re: [PATCH 2/2] drm/msm: Hangcheck progress detection
Quoted 49 lines · Akhil P Oommen
Quoted 47 lines · Rob Clark
From: Rob Clark <[email protected]>
If the hangcheck timer expires, check if the fw's position in the cmdstream has advanced (changed) since last timer expiration, and allow it up to three additional "extensions" to it's alotted time. The intention is to continue to catch "shader stuck in a loop" type hangs quickly, but allow more time for things that are actually making forward progress.
Because we need to sample the CP state twice to detect if there has not been progress, this also cuts the the timer's duration in half.
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 34 +++++++++++++++++++++++++++ drivers/gpu/drm/msm/msm_drv.h | 8 ++++++- drivers/gpu/drm/msm/msm_gpu.c | 20 +++++++++++++++- drivers/gpu/drm/msm/msm_gpu.h | 5 +++- drivers/gpu/drm/msm/msm_ringbuffer.h | 24 +++++++++++++++++++ 5 files changed, 88 insertions(+), 3 deletions(-)drivers/gpu/drm/msm/adreno/a6xx_gpu.c +18
-0@@ -1843,6 +1843,39 @@ static uint32_t a6xx_get_rptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring)return ring->memptrs->rptr =gpu_read (gpu,REG_A6XX_CP_RB_RPTR ); } +static bool a6xx_progress (struct msm_gpu *gpu,struct msm_ringbuffer *ring) +{ +struct msm_cp_state cp_state = { + .ib1_base =gpu_read64 (gpu,REG_A6XX_CP_IB1_BASE ), + .ib2_base =gpu_read64 (gpu,REG_A6XX_CP_IB2_BASE ), + .ib1_rem =gpu_read (gpu,REG_A6XX_CP_IB1_REM_SIZE ), + .ib2_rem =gpu_read (gpu,REG_A6XX_CP_IB2_REM_SIZE ), + }; +bool progress; + +/* +* Adjust the remaining data to account for what has already been +* fetched from memory, but not yet consumed by the SQE. +* +* This is not *technically* correct, the amount buffered could +* exceed the IB size due to hw prefetching ahead, but: +* +* (1) We aren't trying to find the exact position, just whether
Quoted 8 lines
drivers/gpu/drm/msm/adreno/a6xx_gpu.c static uint32_t a6xx_get_rptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring) · line 1870
Quoted 7 lines
drivers/gpu/drm/msm/adreno/a6xx_gpu.c +7
-0+* progress has been made +* (2) The CP_REG_TO_MEM at the end of a submit should be enough +* to prevent prefetching into an unrelated submit. (And +* either way, at some point the ROQ will be full.) +*/ + cp_state.ib1_rem +=gpu_read (gpu,REG_A6XX_CP_CSQ_IB1_STAT ) >>16 ; + cp_state.ib2_rem +=gpu_read (gpu,REG_A6XX_CP_CSQ_IB1_STAT ) >>16 ;
REG_A6XX_CP_CSQ_IB1_STAT->REG_A6XX_CP_CSQ_IB2_STAT
oops, will fix in v2
Quoted 112 lines
With that, Reviewed-by: Akhil P Oommen <[email protected]>
-Akhil.
Quoted 109 lines
drivers/gpu/drm/msm/adreno/a6xx_gpu.c +9
-0+ + progress = !!memcmp (&cp_state, &ring->last_cp_state,sizeof (cp_state)); + + ring->last_cp_state = cp_state; + +return progress; +} +static u32 a618_get_speed_bin (u32 fuse) {if (fuse ==0 )@@ -1961,6 +1994,7 @@ static const struct adreno_gpu_funcs funcs = { .create_address_space = a6xx_create_address_space, .create_private_address_space = a6xx_create_private_address_space, .get_rptr = a6xx_get_rptr, + .progress = a6xx_progress, }, .get_timestamp = a6xx_get_timestamp, };drivers/gpu/drm/msm/msm_drv.h +7
-1@@ -226,7 +226,13 @@ struct msm_drm_private {struct drm_atomic_state *pm_state; -/* For hang detection, in ms */ +/** +* hangcheck_period: For hang detection, in ms +* +* Note that in practice, a submit/job will get at least two hangcheck +* periods, due to checking for progress being implemented as simply +* "have the CP position registers changed since last time?" +*/ unsigned int hangcheck_period;/** drivers/gpu/drm/msm/msm_gpu.c +19
-1@@ -500,6 +500,21 @@ static void hangcheck_timer_reset(struct msm_gpu *gpu) round_jiffies_up(jiffies + msecs_to_jiffies(priv->hangcheck_period))); } +static bool made_progress (struct msm_gpu *gpu,struct msm_ringbuffer *ring) +{ +if (ring->hangcheck_progress_retries >=DRM_MSM_HANGCHECK_PROGRESS_RETRIES ) +return false; + +if (!gpu->funcs->progress) +return false; + +if (!gpu->funcs->progress (gpu, ring)) +return false; + + ring->hangcheck_progress_retries++; +return true; +} +static void hangcheck_handler (struct timer_list *t) {struct msm_gpu *gpu =from_timer (gpu, t, hangcheck_timer);@@ -511,9 +526,12 @@ static void hangcheck_handler(struct timer_list *t)if (fence != ring->hangcheck_fence) {/* some progress has been made.. ya! */ ring->hangcheck_fence = fence; - }else if (fence_before (fence, ring->fctx->last_fence)) { + ring->hangcheck_progress_retries =0 ; + }else if (fence_before (fence, ring->fctx->last_fence) && + !made_progress (gpu, ring)) {/* no progress and not done.. hung! */ ring->hangcheck_fence = fence; + ring->hangcheck_progress_retries =0 ;DRM_DEV_ERROR (dev->dev,"%s: hangcheck detected gpu lockup rb %d!\n" , gpu->name, ring->id); DRM_DEV_ERROR(dev->dev, "%s: completed fence: %u\n",drivers/gpu/drm/msm/msm_gpu.h +4
-1@@ -78,6 +78,8 @@ struct msm_gpu_funcs {struct msm_gem_address_space *(*create_private_address_space) (struct msm_gpu *gpu);uint32_t (*get_rptr)(struct msm_gpu *gpu,struct msm_ringbuffer *ring); + +bool (*progress)(struct msm_gpu *gpu,struct msm_ringbuffer *ring); };/* Additional state for iommu faults: */ @@ -236,7 +238,8 @@ struct msm_gpu {*/ #define DRM_MSM_INACTIVE_PERIOD 66/* in ms (roughly four frames) */ -#define DRM_MSM_HANGCHECK_DEFAULT_PERIOD 500/* in ms */ +#define DRM_MSM_HANGCHECK_DEFAULT_PERIOD 250/* in ms */ +#define DRM_MSM_HANGCHECK_PROGRESS_RETRIES 3struct timer_list hangcheck_timer;/* Fault info for most recent iova fault: */ drivers/gpu/drm/msm/msm_ringbuffer.h +0
-0@@ -35,6 +35,11 @@ struct msm_rbmemptrs {volatile u64 ttbr0; };
Quoted 6 lines
drivers/gpu/drm/msm/msm_ringbuffer.h struct msm_rbmemptrs { · line 42
Quoted 5 lines
drivers/gpu/drm/msm/msm_ringbuffer.h +5
-0+struct msm_cp_state { +uint64_t ib1_base, ib2_base; +uint32_t ib1_rem, ib2_rem; +}; +I think we can add
CP_RB_RPTRtoo here.
I originally was going to, but I figured if it is making progress in the RB, probably fences will be getting updated. And if not at least IB1 state should be changing, since we don't write all that much into RB other than CP_INDIRECT_BRANCH (ie. nothing that would take long enough to run for hangcheck timer to expire)
I guess I could at least add a comment.
BR,
-R
Quoted 30 lines
Quoted 29 lines
drivers/gpu/drm/msm/msm_ringbuffer.h +19
-0struct msm_ringbuffer {struct msm_gpu *gpu;int id;@@ -64,6 +69,25 @@ struct msm_ringbuffer {uint64_t memptrs_iova;struct msm_fence_context *fctx; +/** +* hangcheck_progress_retries: +* +* The number of extra hangcheck duration cycles that we have given +* due to it appearing that the GPU is making forward progress. +* +* If the GPU appears to be making progress (ie. the CP has advanced +* in the command stream, we'll allow up to DRM_MSM_HANGCHECK_PROGRESS_RETRIES +* expirations of the hangcheck timer before killing the job. In other +* words we'll let the submit run for up to +* DRM_MSM_HANGCHECK_DEFAULT_PERIOD * DRM_MSM_HANGCHECK_PROGRESS_RETRIES +*/ +int hangcheck_progress_retries; + +/** +* last_cp_state: The state of the CP at the last call to gpu->progress() +*/ +struct msm_cp_state last_cp_state; +/* * preempt_lock protects preemption and serializes wptr updates against * preemption. Can be aquired from irq context.