drm/amdgpu: request reg_val_offs each kiq read reg
Subsystems (2)
PATCH 1/1 drm/amdgpu: request reg_val_offs each kiq read reg
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ
2. client-A poll the seqno-0
3. client-B start to read REG-1 through KIQ
4. client-B poll the seqno-1
5. the kiq complete these two read operation
6. client-A to read the register at the wb buffer and get REG-1 value
Therefore, use amdgpu_device_wb_get() to request reg_val_offs for each kiq read register.
v2: fix the error remove
--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -1
@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +11 -8
@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ; ring->use_doorbell = true;
@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,void amdgpu_gfx_kiq_free_ring (struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }
@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring);spin_unlock_irqrestore (&kiq->ring_lock, flags);
@@ -707,7 +707,10 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)if (cnt >MAX_KIQ_REG_TRY ) goto failed_kiq_read; -return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read:pr_err ("failed to read reg:%x\n" , reg);
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0 -1
@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/*
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +3 -2
@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val,uint32_t mask);
@@ -265,7 +266,7 @@ struct amdgpu_ring {#define amdgpu_ring_emit_hdp_flush (r) (r)->funcs->emit_hdp_flush((r))#define amdgpu_ring_emit_switch_buffer (r) (r)->funcs->emit_switch_buffer((r))#define amdgpu_ring_emit_cntxcntl (r, d) (r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg (r, d) (r)->funcs->emit_rreg((r), (d)) +#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), (d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m))
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4 -3
@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq;
@@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +4 -3
@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq;
@@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +18 -9
@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/
@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring);spin_unlock_irqrestore (&kiq->ring_lock, flags);
@@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)if (cnt >MAX_KIQ_REG_TRY ) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read:pr_err ("failed to read gpu clock\n" );
@@ -5482,10 +5491,10 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size>>2 ) - offset + cur; } -static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg) +static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/
@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
Signature · 6 lines
2.17.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Christian König
maintainerAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 69 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ
2. client-A poll the seqno-0
3. client-B start to read REG-1 through KIQ
4. client-B poll the seqno-1
5. the kiq complete these two read operation
6. client-A to read the register at the wb buffer and get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-6@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ; ring->use_doorbell = true;@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,void amdgpu_gfx_kiq_free_ring (struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );
Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the failed_kiq_read label also prints an error.
With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.
Quoted 173 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +7
-2+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring);spin_unlock_irqrestore (&kiq->ring_lock, flags);@@ -707,7 +707,10 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)if (cnt >MAX_KIQ_REG_TRY ) goto failed_kiq_read; -return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read:pr_err ("failed to read reg:%x\n" , reg);drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +3
-2@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val,uint32_t mask);@@ -265,7 +266,7 @@ struct amdgpu_ring {#define amdgpu_ring_emit_hdp_flush (r) (r)->funcs->emit_hdp_flush((r))#define amdgpu_ring_emit_switch_buffer (r) (r)->funcs->emit_switch_buffer((r))#define amdgpu_ring_emit_cntxcntl (r, d) (r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg (r, d) (r)->funcs->emit_rreg((r), (d)) +#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), (d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m))drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +4
-3@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq;@@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +4
-3@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq;@@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +16
-7@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring);spin_unlock_irqrestore (&kiq->ring_lock, flags);@@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)if (cnt >MAX_KIQ_REG_TRY ) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read:pr_err ("failed to read gpu clock\n" );@@ -5482,10 +5491,10 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size>>2 ) - offset + cur; } -static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg) +static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) · line 5511
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2
-2amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?
Regards
Shaoyun.liu
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Am 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );
Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the failed_kiq_read label also prints an error.
With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.
Quoted 194 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ struct
amdgpu_ring{
#defineamdgpu_ring_emit_hdp_flush(r)(r)->funcs->emit_hdp_flush((r)) #defineamdgpu_ring_emit_switch_buffer(r)(r)->funcs->emit_switch_buffer((r)) #defineamdgpu_ring_emit_cntxcntl(r, d)
(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct
amdgpu_kiq*kiq =&adev->gfx.kiq; @@ -7606,9 +7607,9 @@
static voidgfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)
amdgpu_ring_write(ring, reg);
amdgpu_ring_write(ring, 0);
amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct
amdgpu_kiq*kiq =&adev->gfx.kiq; @@ -6395,9 +6396,9 @@
static voidgfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)
amdgpu_ring_write(ring, reg);
amdgpu_ring_write(ring, 0);
amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +9
-3@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11@@ static
uint64_tgfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +5
-2-return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsigne
ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
}-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
Quoted 1 line
uint32_treg,
_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CShaoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTPnmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Christian König
maintainerHi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 299 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 193 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +9
-3@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11@@ static
uint64_tgfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +5
-2-return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 6 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CShaoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTPnmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
[AMD Official Use Only - Internal Distribution Only]
Usually doesn't means we can not do it . I feel this proposal is the simplest and clean . But anyway this is just my suggestion.
Regards
Shaoyun.liu
-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: Wednesday, April 22, 2020 10:52 AM
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 302 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 193 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +15
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 10 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Christian König
maintainerYeah, we discussed that approach as well. But both Yintian and I agreed that this is the simpler solution.
Regards,
Christian.
Am 22.04.20 um 16:59 schrieb Liu, Shaoyun:
Quoted 16 lines
[AMD Official Use Only - Internal Distribution Only]
Usually doesn't means we can not do it . I feel this proposal is the simplest and clean . But anyway this is just my suggestion.
Regards
Shaoyun.liu-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: Wednesday, April 22, 2020 10:52 AM
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regHi Shaoyun,
Quoted 320 lines
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 312 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 194 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +15
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,Quoted 1 line
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Hi Shaoyun
There is one rare corner case which will raise problem when using ring buffer to store value.
It is assumed there are only total four slots at KIQ ring buffer.
And these four slots are fulfilled with command to read registers. Slot-A Slot-B Slot-C Slot-D
And they are waiting for the sequence fences to be signaled. Here, there is one new command to write register to be submitted
1. Slot-A under msleep not to read register
2. Slot-B under msleep not to read register
3. Slot-C under msleep not to read register.
4. Slot-D happen to find the sequence signaled and here the new write command will overwrite the Slot-A contents.
Best Regards
Yintian Tao
-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: 2020年4月22日 22:52
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 302 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 193 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +15
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 10 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
[AMD Official Use Only - Internal Distribution Only]
This is the issue you try to solve with your second patch (protect kiq overrun) . For current patch , if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain .
Shaoyun.liu
-----Original Message-----
From: Tao, Yintian <[email protected]>
Sent: Wednesday, April 22, 2020 11:00 AM
To: Koenig, Christian <[email protected]>; Liu, Shaoyun <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
There is one rare corner case which will raise problem when using ring buffer to store value.
It is assumed there are only total four slots at KIQ ring buffer.
And these four slots are fulfilled with command to read registers. Slot-A Slot-B Slot-C Slot-D
And they are waiting for the sequence fences to be signaled. Here, there is one new command to write register to be submitted
1. Slot-A under msleep not to read register 2. Slot-B under msleep not to read register 3. Slot-C under msleep not to read register.
4. Slot-D happen to find the sequence signaled and here the new write command will overwrite the Slot-A contents.
Best Regards
Yintian Tao
-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: 2020年4月22日 22:52
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 303 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 194 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +16
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct +amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 10 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Hi Shaoyun
No, the second patch can't solve this rare case because only Slot-D is signaled and the Slot-A can be overwritten. The second patch think the sequence is signaled the Slot-A buffer can be freed.
if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain . [yttao]: If we wan to really make the kiq operation be atomic then we have to do the things below:spin_lock_irqsave(&kiq->ring_lock, flags);
.....
Fulfill the command buffer
.....
if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
goto failed_kiq_write;
might_sleep();
while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); /* here will break atomic and we need directly use udealy*/
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
}spin_lock_irqrestore(&kiq->ring_lock, flags);
Best Regards
Yintian Tao
-----Original Message-----
From: Liu, Shaoyun <[email protected]>
Sent: 2020年4月22日 23:35
To: Tao, Yintian <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
[AMD Official Use Only - Internal Distribution Only]
This is the issue you try to solve with your second patch (protect kiq overrun) . For current patch , if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain .
Shaoyun.liu
-----Original Message-----
From: Tao, Yintian <[email protected]>
Sent: Wednesday, April 22, 2020 11:00 AM
To: Koenig, Christian <[email protected]>; Liu, Shaoyun <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
There is one rare corner case which will raise problem when using ring buffer to store value.
It is assumed there are only total four slots at KIQ ring buffer.
And these four slots are fulfilled with command to read registers. Slot-A Slot-B Slot-C Slot-D
And they are waiting for the sequence fences to be signaled. Here, there is one new command to write register to be submitted
1. Slot-A under msleep not to read register 2. Slot-B under msleep not to read register 3. Slot-C under msleep not to read register.
4. Slot-D happen to find the sequence signaled and here the new write command will overwrite the Slot-A contents.
Best Regards
Yintian Tao
-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: 2020年4月22日 22:52
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 303 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 194 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +16
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct +amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 10 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Add more
Especially for the multi-VF environment, we have to wait through msleep() instead udeay. Because the max udelay time is 15VF * 6ms(world-switch) = 90ms.
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Tao, Yintian
Sent: 2020年4月22日 23:43
To: Liu, Shaoyun <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
No, the second patch can't solve this rare case because only Slot-D is signaled and the Slot-A can be overwritten. The second patch think the sequence is signaled the Slot-A buffer can be freed.
if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain . [yttao]: If we wan to really make the kiq operation be atomic then we have to do the things below:spin_lock_irqsave(&kiq->ring_lock, flags); .....
Fulfill the command buffer
.....
if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
goto failed_kiq_write;
might_sleep();
while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); /* here will break atomic and we need directly use udealy*/
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
}spin_lock_irqrestore(&kiq->ring_lock, flags);
Best Regards
Yintian Tao
-----Original Message-----
From: Liu, Shaoyun <[email protected]>
Sent: 2020年4月22日 23:35
To: Tao, Yintian <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
[AMD Official Use Only - Internal Distribution Only]
This is the issue you try to solve with your second patch (protect kiq overrun) . For current patch , if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain .
Shaoyun.liu
-----Original Message-----
From: Tao, Yintian <[email protected]>
Sent: Wednesday, April 22, 2020 11:00 AM
To: Koenig, Christian <[email protected]>; Liu, Shaoyun <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
There is one rare corner case which will raise problem when using ring buffer to store value.
It is assumed there are only total four slots at KIQ ring buffer.
And these four slots are fulfilled with command to read registers. Slot-A Slot-B Slot-C Slot-D
And they are waiting for the sequence fences to be signaled. Here, there is one new command to write register to be submitted
1. Slot-A under msleep not to read register 2. Slot-B under msleep not to read register 3. Slot-C under msleep not to read register.
4. Slot-D happen to find the sequence signaled and here the new write command will overwrite the Slot-A contents.
Best Regards
Yintian Tao
-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: 2020年4月22日 22:52
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 303 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 194 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +16
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct +amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 10 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7Cyintian.tao%40amd.com%7C28627a9c362245c1969d08d7e6d3d568%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637231670346872304&sdata=o0%2FdGt1sqir2eVGTf1JdamDSNvJnbMY5WvyxWa1mUUU%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
[AMD Official Use Only - Internal Distribution Only]
OK, I see, the submission it self be signaled so the ring space for this submission will be re-use by other submission , but the CPU still not read the out put value yet.
Thanks
Shaoyun.liu
-----Original Message-----
From: Tao, Yintian <[email protected]>
Sent: Wednesday, April 22, 2020 11:47 AM
To: Tao, Yintian <[email protected]>; Liu, Shaoyun <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Add more
Especially for the multi-VF environment, we have to wait through msleep() instead udeay. Because the max udelay time is 15VF * 6ms(world-switch) = 90ms.
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Tao, Yintian
Sent: 2020年4月22日 23:43
To: Liu, Shaoyun <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
No, the second patch can't solve this rare case because only Slot-D is signaled and the Slot-A can be overwritten. The second patch think the sequence is signaled the Slot-A buffer can be freed.
if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain . [yttao]: If we wan to really make the kiq operation be atomic then we have to do the things below:spin_lock_irqsave(&kiq->ring_lock, flags); .....
Fulfill the command buffer
.....
if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
goto failed_kiq_write;
might_sleep();
while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); /* here will break atomic and we need directly use udealy*/
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
}spin_lock_irqrestore(&kiq->ring_lock, flags);
Best Regards
Yintian Tao
-----Original Message-----
From: Liu, Shaoyun <[email protected]>
Sent: 2020年4月22日 23:35
To: Tao, Yintian <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
[AMD Official Use Only - Internal Distribution Only]
This is the issue you try to solve with your second patch (protect kiq overrun) . For current patch , if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain .
Shaoyun.liu
-----Original Message-----
From: Tao, Yintian <[email protected]>
Sent: Wednesday, April 22, 2020 11:00 AM
To: Koenig, Christian <[email protected]>; Liu, Shaoyun <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
There is one rare corner case which will raise problem when using ring buffer to store value.
It is assumed there are only total four slots at KIQ ring buffer.
And these four slots are fulfilled with command to read registers. Slot-A Slot-B Slot-C Slot-D
And they are waiting for the sequence fences to be signaled. Here, there is one new command to write register to be submitted
1. Slot-A under msleep not to read register 2. Slot-B under msleep not to read register 3. Slot-C under msleep not to read register.
4. Slot-D happen to find the sequence signaled and here the new write command will overwrite the Slot-A contents.
Best Regards
Yintian Tao
-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: 2020年4月22日 22:52
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 303 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 194 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +16
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct +amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 10 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7Cyintian.tao%40amd.com%7C28627a9c362245c1969d08d7e6d3d568%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637231670346872304&sdata=o0%2FdGt1sqir2eVGTf1JdamDSNvJnbMY5WvyxWa1mUUU%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Hi Shaoyun
Yes, you are right. It is the rare corner case.
Best Regards
Yintian Tao
-----Original Message-----
From: Liu, Shaoyun <[email protected]>
Sent: 2020年4月22日 23:51
To: Tao, Yintian <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
[AMD Official Use Only - Internal Distribution Only]
OK, I see, the submission it self be signaled so the ring space for this submission will be re-use by other submission , but the CPU still not read the out put value yet.
Thanks
Shaoyun.liu
-----Original Message-----
From: Tao, Yintian <[email protected]>
Sent: Wednesday, April 22, 2020 11:47 AM
To: Tao, Yintian <[email protected]>; Liu, Shaoyun <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Add more
Especially for the multi-VF environment, we have to wait through msleep() instead udeay. Because the max udelay time is 15VF * 6ms(world-switch) = 90ms.
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Tao, Yintian
Sent: 2020年4月22日 23:43
To: Liu, Shaoyun <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
No, the second patch can't solve this rare case because only Slot-D is signaled and the Slot-A can be overwritten. The second patch think the sequence is signaled the Slot-A buffer can be freed.
if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain . [yttao]: If we wan to really make the kiq operation be atomic then we have to do the things below:spin_lock_irqsave(&kiq->ring_lock, flags); .....
Fulfill the command buffer
.....
if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
goto failed_kiq_write;
might_sleep();
while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); /* here will break atomic and we need directly use udealy*/
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
}spin_lock_irqrestore(&kiq->ring_lock, flags);
Best Regards
Yintian Tao
-----Original Message-----
From: Liu, Shaoyun <[email protected]>
Sent: 2020年4月22日 23:35
To: Tao, Yintian <[email protected]>; Koenig, Christian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
[AMD Official Use Only - Internal Distribution Only]
This is the issue you try to solve with your second patch (protect kiq overrun) . For current patch , if you store the output value in each ring buffer itself , each kiq operation will be atomic and self contain .
Shaoyun.liu
-----Original Message-----
From: Tao, Yintian <[email protected]>
Sent: Wednesday, April 22, 2020 11:00 AM
To: Koenig, Christian <[email protected]>; Liu, Shaoyun <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: RE: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun
There is one rare corner case which will raise problem when using ring buffer to store value.
It is assumed there are only total four slots at KIQ ring buffer.
And these four slots are fulfilled with command to read registers. Slot-A Slot-B Slot-C Slot-D
And they are waiting for the sequence fences to be signaled. Here, there is one new command to write register to be submitted
1. Slot-A under msleep not to read register 2. Slot-B under msleep not to read register 3. Slot-C under msleep not to read register.
4. Slot-D happen to find the sequence signaled and here the new write command will overwrite the Slot-A contents.
Best Regards
Yintian Tao
-----Original Message-----
From: Koenig, Christian <[email protected]>
Sent: 2020年4月22日 22:52
To: Liu, Shaoyun <[email protected]>; Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: request reg_val_offs each kiq read reg
Hi Shaoyun,
the ring buffer is usually filled with command and not read results.
Allocating extra space would only work if we use the special NOP command and that is way more complicated and fragile than just using the wb functions which where made for this stuff.
Regards,
Christian.
Am 22.04.20 um 16:48 schrieb Liu, Shaoyun:
Quoted 303 lines
[AMD Official Use Only - Internal Distribution Only]
Hi ,Yintian & Christian
I still don't understand why we need this complicated change here . Why can not just allocate few more extra space in the ring for each read and use the space to store the output value ?Regards
Shaoyun.liu-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Christian König
Sent: Wednesday, April 22, 2020 8:42 AM
To: Tao, Yintian <[email protected]>; Liu, Monk <[email protected]>; Kuehling, Felix <[email protected]>
Cc: [email protected]
Subject: Re: [PATCH] drm/amdgpu: requestreg_val_offseach kiq read regAm 22.04.20 um 14:36 schrieb Yintian Tao:
Quoted 70 lines
According to the current kiq read register method, there will be race condition when using KIQ to read register if multiple clients want to read at same time just like the expample below:
1. client-A start to read REG-0 throguh KIQ 2. client-A poll the seqno-0 3. client-B start to read REG-1 through KIQ 4. client-B poll the seqno-1 5. the kiq complete these two read operation 6. client-A to read the register at the wb buffer and
get REG-1 valueTherefore, use
amdgpu_device_wb_get()to requestreg_val_offsfor each kiq read register.v2: fix the error remove
Signed-off-by Yintian Tao <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 19 ++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 7 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 27 ++++++++++++++++-------- 7 files changed, 41 insertions(+), 27 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-1@@ -526,7 +526,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,/* * Writeback */ -#define AMDGPU_MAX_WB 128/* Reserve at most 128 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 256/* Reserve at most 256 WB slots for amdgpu-owned rings. */ struct amdgpu_wb {struct amdgpu_bo *wb_obj;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-4@@ -304,10 +304,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,spin_lock_init (&kiq->ring_lock); - r =amdgpu_device_wb_get (adev, &kiq->reg_val_offs); -if (r) -return r; - ring->adev =NULL ; ring->ring_obj =NULL ;ring->use_doorbell = true;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +0
-2@@ -331,7 +327,6 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring) { -amdgpu_device_wb_free (ring->adev, ring->adev->gfx.kiq.reg_val_offs);amdgpu_ring_fini (ring); }@@ -672,15 +667,20 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) · line 679
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-0+uint32_t seq, reg_val_offs =0 , value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" );Typo here, this should probably read "too many kiq readers".
But I don't think we need the error message here anyway, the
failed_kiq_readlabel also prints an error.With that fixed the patch is Reviewed-by: Christian König <[email protected]>.
Thanks,
Christian.Quoted 194 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +3
-1+ goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 ); -amdgpu_ring_emit_rreg (ring, reg); +amdgpu_ring_emit_rreg (ring, reg, reg_val_offs);amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -707,7 +707,10@@
uint32_tamdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (cnt >MAX_KIQ_REG_TRY)
gotofailed_kiq_read;drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +4
-1-return adev->wb.wb[kiq->reg_val_offs]; +mb (); + value = adev->wb.wb[reg_val_offs]; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read reg:%x\n", reg); diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.hb/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 634746829024..ee698f0246d8 100644drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +0
-1@@ -103,7 +103,6 @@ struct amdgpu_kiq {struct amdgpu_ring ring;struct amdgpu_irq_src irq;const struct kiq_pm4_funcs *pmf; -uint32_t reg_val_offs; };/* drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-1@@ -181,7 +181,8 @@ struct amdgpu_ring_funcs {void (*end_use)(struct amdgpu_ring *ring);void (*emit_switch_buffer) (struct amdgpu_ring *ring);void (*emit_cntxcntl) (struct amdgpu_ring *ring,uint32_t flags); -void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg); +void (*emit_rreg)(struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs);void (*emit_wreg)(struct amdgpu_ring *ring,uint32_t reg,uint32_t val);void (*emit_reg_wait)(struct amdgpu_ring *ring,uint32_t reg, uint32_t val, uint32_t mask); @@ -265,7 +266,7 @@ structamdgpu_ring { #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d)(r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r,
d) (r)->funcs->emit_rreg((r), (d))drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2
-0+#define amdgpu_ring_emit_rreg (r, d, o) (r)->funcs->emit_rreg((r), +(d), (o))#define amdgpu_ring_emit_wreg (r, d, v) (r)->funcs->emit_wreg((r), (d), (v))#define amdgpu_ring_emit_reg_wait (r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))#define amdgpu_ring_emit_reg_write_reg_wait (r, d0, d1, v, m)(r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) diff --git
a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index0a03e2ad5d95..0a3d3bc9d441100644drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-1@@ -7594,7 +7594,8 @@ static void gfx_v10_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start,amdgpu_ring_write (ring, v |FRAME_CMD (start ?0 :1 )); } -static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v10_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -7606,9 +7607,9 @@ static void gfx_v10_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v10_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
indexfc6c2f2bc76c..ccfa65c3cb5a100644drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-1@@ -6383,7 +6383,8 @@ static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size >>2 ) - offset + cur; } -static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring, uint32_t reg) +static void gfx_v8_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev;struct amdgpu_kiq *kiq = &adev->gfx.kiq; @@ -6395,9 +6396,9 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) amdgpu_ring_write(ring, reg); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +2
-2- kiq->reg_val_offs * 4)); + reg_val_offs * 4));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); } static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg, diff --gita/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.cb/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index84fcf842316d..0b706f908d9a100644drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +16
-5@@ -4043,13 +4043,19 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev) {signed long r, cnt =0 ;unsigned long flags; -uint32_t seq; +uint32_t seq, reg_val_offs =0 ; +uint64_t value =0 ;struct amdgpu_kiq *kiq = &adev->gfx.kiq;struct amdgpu_ring *ring = &kiq->ring;BUG_ON (!ring->funcs->emit_rreg);spin_lock_irqsave (&kiq->ring_lock, flags); +if (amdgpu_device_wb_get (adev, ®_val_offs)) { +spin_unlock_irqrestore (&kiq->ring_lock, flags); +pr_err ("critical bug! too more kiq readers\n" ); + goto failed_kiq_read; + }amdgpu_ring_alloc (ring,32 );amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 9 | /* src: register*/@@ -4059,9 +4065,9 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_fence_emit_polling (ring, &seq);amdgpu_ring_commit (ring); spin_unlock_irqrestore(&kiq->ring_lock, flags); @@ -4088,8 +4094,11 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct +amdgpu_device *adev) if (cnt > MAX_KIQ_REG_TRY) goto failed_kiq_read; -return (uint64_t )adev->wb.wb[kiq->reg_val_offs] | - (uint64_t )adev->wb.wb[kiq->reg_val_offs +1 ] <<32ULL ; +mb (); + value = (uint64_t )adev->wb.wb[reg_val_offs] | + (uint64_t )adev->wb.wb[reg_val_offs +1 ] <<32ULL ; +amdgpu_device_wb_free (adev, reg_val_offs); +return value; failed_kiq_read: pr_err("failed to read gpu clock\n"); @@ -5482,10 +5491,10 @@static void gfx_v9_0_ring_emit_patch_cond_exec(struct
amdgpu_ring*ring, unsignering->ring[offset] = (ring->ring_size>>2) - offset + cur; }-static void
gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +4
-3+static void gfx_v9_0_ring_emit_rreg (struct amdgpu_ring *ring,uint32_t reg, +uint32_t reg_val_offs) {struct amdgpu_device *adev = ring->adev; -struct amdgpu_kiq *kiq = &adev->gfx.kiq;amdgpu_ring_write (ring,PACKET3 (PACKET3_COPY_DATA ,4 )); amdgpu_ring_write(ring, 0 | /* src: register*/@@ -5494,9 +5503,9 @@ static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)amdgpu_ring_write (ring, reg);amdgpu_ring_write (ring,0 );amdgpu_ring_write (ring,lower_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 ));amdgpu_ring_write (ring,upper_32_bits (adev->wb.gpu_addr + - kiq->reg_val_offs *4 )); + reg_val_offs *4 )); }
Quoted 10 lines
Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0
-0static void gfx_v9_0_ring_emit_wreg(struct amdgpu_ring *ring,
uint32_treg,_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
s.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7CSh aoyun.Liu%40amd.com%7C9d960d5be3f141f44f0608d7e6ba8962%7C3dd8961fe4884 e608e11a82d994e183d%7C0%7C0%7C637231561176171819&sdata=OnDtxn%2FTP nmzigVKJEAF37fQdUcfDMWB1xcbrQbDcIc%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=02%7C01%7Cyintian.tao%40amd.com%7C28627a9c362245c1969d08d7e6d3d568%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637231670346872304&sdata=o0%2FdGt1sqir2eVGTf1JdamDSNvJnbMY5WvyxWa1mUUU%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx