Re: [PATCH 05/29] drm/amdgpu: setup hw debug registers on driver initialization

[Felix Kuehling <felix.kuehling@xxxxxxx>]

On 2022-11-22 18:38, Felix Kuehling wrote:
> On 2022-10-31 12:23, Jonathan Kim wrote:
> > Add missing debug trap registers references and initialize all debug
> > registers on boot by clearing the hardware exception overrides and the
> > wave allocation ID index.
> >
> > For debug devices that only support single process debugging, enable
> > trap temporary setup by default.
> >
> > Debug devices that support multi-process debugging require trap
> > temporary setup to be disabled by default in order to satisfy microbench
> > performance when in non-debug mode.
>
> Where is this done? I don't think it's in the MQD setup because that 
> happens unconditionally on all GPUs.

If I understand it correctly, it's done by actually enabling the debug 
trap in patch 9 (for Aldebaran). For single-process debug devices, the 
debug trap is always on, as per this patch.

Maybe just add a reference to the Aldebaran patch to make it clearer.

Regards,
  Felix


> > The debugger requires that TTMPs 6 & 7 save the dispatch ID to map
> > waves onto dispatch during compute context inspection.
> > In order to correctly this up, set the special reserved CP bit by 
> > default
> > whenever the MQD is initailized.
>
> There is a word missing here. "In order to correctly _set_ this up ..."?
>
> This patch covers GFXv9 and 10. Will GFXv11 be handled separately?
>
> Regards,
>   Felix
>
>
> > Signed-off-by: Jonathan Kim <jonathan.kim@xxxxxxx>
> > ---
> >   drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c        | 26 +++++++
> >   drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c         | 30 ++++++++
> >   .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c  |  5 ++
> >   .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c   |  5 ++
> >   .../include/asic_reg/gc/gc_10_1_0_offset.h    | 14 ++++
> >   .../include/asic_reg/gc/gc_10_1_0_sh_mask.h   | 69 +++++++++++++++++++
> >   .../include/asic_reg/gc/gc_10_3_0_offset.h    | 10 +++
> >   .../include/asic_reg/gc/gc_10_3_0_sh_mask.h   |  4 ++
> >   8 files changed, 163 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c 
> > b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> > index af94ac580d3e..d49aff0b4ba3 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
> > @@ -4904,6 +4904,29 @@ static u32 
> > gfx_v10_0_init_pa_sc_tile_steering_override(struct amdgpu_device *ade
> >     #define DEFAULT_SH_MEM_BASES    (0x6000)
> >   +static void gfx_v10_0_debug_trap_config_init(struct amdgpu_device 
> > *adev,
> > +                uint32_t first_vmid,
> > +                uint32_t last_vmid)
> > +{
> > +    uint32_t data;
> > +    uint32_t trap_config_vmid_mask = 0;
> > +    int i;
> > +
> > +    /* Calculate trap config vmid mask */
> > +    for (i = first_vmid; i < last_vmid; i++)
> > +        trap_config_vmid_mask |= (1 << i);
> > +
> > +    data = REG_SET_FIELD(0, SPI_GDBG_TRAP_CONFIG,
> > +            VMID_SEL, trap_config_vmid_mask);
> > +    data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
> > +            TRAP_EN, 1);
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
> > +
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
> > +}
> > +
> >   static void gfx_v10_0_init_compute_vmid(struct amdgpu_device *adev)
> >   {
> >       int i;
> > @@ -4935,6 +4958,9 @@ static void gfx_v10_0_init_compute_vmid(struct 
> > amdgpu_device *adev)
> >           WREG32_SOC15_OFFSET(GC, 0, mmGDS_GWS_VMID0, i, 0);
> >           WREG32_SOC15_OFFSET(GC, 0, mmGDS_OA_VMID0, i, 0);
> >       }
> > +
> > +    gfx_v10_0_debug_trap_config_init(adev, 
> > adev->vm_manager.first_kfd_vmid,
> > +                    AMDGPU_NUM_VMID);
> >   }
> >     static void gfx_v10_0_init_gds_vmid(struct amdgpu_device *adev)
> > diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c 
> > b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> > index 0320be4a5fc6..a0e5ad342f13 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> > @@ -2337,6 +2337,29 @@ static void gfx_v9_0_setup_rb(struct 
> > amdgpu_device *adev)
> >       adev->gfx.config.num_rbs = hweight32(active_rbs);
> >   }
> >   +static void gfx_v9_0_debug_trap_config_init(struct amdgpu_device 
> > *adev,
> > +                uint32_t first_vmid,
> > +                uint32_t last_vmid)
> > +{
> > +    uint32_t data;
> > +    uint32_t trap_config_vmid_mask = 0;
> > +    int i;
> > +
> > +    /* Calculate trap config vmid mask */
> > +    for (i = first_vmid; i < last_vmid; i++)
> > +        trap_config_vmid_mask |= (1 << i);
> > +
> > +    data = REG_SET_FIELD(0, SPI_GDBG_TRAP_CONFIG,
> > +            VMID_SEL, trap_config_vmid_mask);
> > +    data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
> > +            TRAP_EN, 1);
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
> > +
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
> > +    WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
> > +}
> > +
> >   #define DEFAULT_SH_MEM_BASES    (0x6000)
> >   static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev)
> >   {
> > @@ -4609,6 +4632,13 @@ static int gfx_v9_0_late_init(void *handle)
> >       if (r)
> >           return r;
> >   +    if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
> > +        gfx_v9_4_2_debug_trap_config_init(adev,
> > +            adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID);
> > +    else
> > +        gfx_v9_0_debug_trap_config_init(adev,
> > +            adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID);
> > +
> >       return 0;
> >   }
> >   diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c 
> > b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
> > index d3e2b6a599a4..cb484ace17de 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
> > @@ -117,6 +117,11 @@ static void init_mqd(struct mqd_manager *mm, 
> > void **mqd,
> >               1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT |
> >               1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT;
> >   +    /* Set cp_hqd_hq_scheduler0 bit 14 to 1 to have the CP set up the
> > +     * DISPATCH_PTR.  This is required for the kfd debugger
> > +     */
> > +    m->cp_hqd_hq_scheduler0 = 1 << 14;
> > +
> >       if (q->format == KFD_QUEUE_FORMAT_AQL) {
> >           m->cp_hqd_aql_control =
> >               1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT;
> > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c 
> > b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
> > index 0778e587a2d6..86f1cf090246 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
> > @@ -164,6 +164,11 @@ static void init_mqd(struct mqd_manager *mm, 
> > void **mqd,
> >               1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT |
> >               1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT;
> >   +    /* Set cp_hqd_hq_scheduler0 bit 14 to 1 to have the CP set up the
> > +     * DISPATCH_PTR.  This is required for the kfd debugger
> > +     */
> > +    m->cp_hqd_hq_status0 = 1 << 14;
> > +
> >       if (q->format == KFD_QUEUE_FORMAT_AQL) {
> >           m->cp_hqd_aql_control =
> >               1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT;
> > diff --git 
> > a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h 
> > b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h
> > index 18d34bbceebe..7d384f86bd67 100644
> > --- a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h
> > +++ b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_offset.h
> > @@ -5190,6 +5190,20 @@
> >   #define mmSPI_WCL_PIPE_PERCENT_CS6_BASE_IDX 0
> >   #define mmSPI_WCL_PIPE_PERCENT_CS7 0x1f70
> >   #define mmSPI_WCL_PIPE_PERCENT_CS7_BASE_IDX 0
> > +#define mmSPI_GDBG_WAVE_CNTL 0x1f71
> > +#define mmSPI_GDBG_WAVE_CNTL_BASE_IDX 0
> > +#define mmSPI_GDBG_TRAP_CONFIG 0x1f72
> > +#define mmSPI_GDBG_TRAP_CONFIG_BASE_IDX 0
> > +#define mmSPI_GDBG_TRAP_MASK 0x1f73
> > +#define mmSPI_GDBG_TRAP_MASK_BASE_IDX 0
> > +#define mmSPI_GDBG_WAVE_CNTL2 0x1f74
> > +#define mmSPI_GDBG_WAVE_CNTL2_BASE_IDX 0
> > +#define mmSPI_GDBG_WAVE_CNTL3 0x1f75
> > +#define mmSPI_GDBG_WAVE_CNTL3_BASE_IDX 0
> > +#define mmSPI_GDBG_TRAP_DATA0 0x1f78
> > +#define mmSPI_GDBG_TRAP_DATA0_BASE_IDX 0
> > +#define mmSPI_GDBG_TRAP_DATA1 0x1f79
> > +#define mmSPI_GDBG_TRAP_DATA1_BASE_IDX 0
> >   #define mmSPI_COMPUTE_QUEUE_RESET 0x1f7b
> >   #define mmSPI_COMPUTE_QUEUE_RESET_BASE_IDX 0
> >   #define mmSPI_RESOURCE_RESERVE_CU_0 0x1f7c
> > diff --git 
> > a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_sh_mask.h 
> > b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_sh_mask.h
> > index 4127896ffcdf..08772ba845b0 100644
> > --- a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_sh_mask.h
> > +++ b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_1_0_sh_mask.h
> > @@ -19646,6 +19646,75 @@
> >   //SPI_WCL_PIPE_PERCENT_CS7
> >   #define SPI_WCL_PIPE_PERCENT_CS7__VALUE__SHIFT 0x0
> >   #define SPI_WCL_PIPE_PERCENT_CS7__VALUE_MASK 0x7FL
> > +//SPI_GDBG_WAVE_CNTL
> > +#define SPI_GDBG_WAVE_CNTL__STALL_RA__SHIFT 0x0
> > +#define SPI_GDBG_WAVE_CNTL__STALL_VMID__SHIFT 0x1
> > +#define SPI_GDBG_WAVE_CNTL__STALL_RA_MASK 0x00000001L
> > +#define SPI_GDBG_WAVE_CNTL__STALL_VMID_MASK 0x0001FFFEL
> > +//SPI_GDBG_TRAP_CONFIG
> > +#define SPI_GDBG_TRAP_CONFIG__ME_SEL__SHIFT 0x0
> > +#define SPI_GDBG_TRAP_CONFIG__PIPE_SEL__SHIFT 0x2
> > +#define SPI_GDBG_TRAP_CONFIG__QUEUE_SEL__SHIFT 0x4
> > +#define SPI_GDBG_TRAP_CONFIG__ME_MATCH__SHIFT 0x7
> > +#define SPI_GDBG_TRAP_CONFIG__PIPE_MATCH__SHIFT 0x8
> > +#define SPI_GDBG_TRAP_CONFIG__QUEUE_MATCH__SHIFT 0x9
> > +#define SPI_GDBG_TRAP_CONFIG__TRAP_EN__SHIFT 0xf
> > +#define SPI_GDBG_TRAP_CONFIG__VMID_SEL__SHIFT 0x10
> > +#define SPI_GDBG_TRAP_CONFIG__ME_SEL_MASK 0x00000003L
> > +#define SPI_GDBG_TRAP_CONFIG__PIPE_SEL_MASK 0x0000000CL
> > +#define SPI_GDBG_TRAP_CONFIG__QUEUE_SEL_MASK 0x00000070L
> > +#define SPI_GDBG_TRAP_CONFIG__ME_MATCH_MASK 0x00000080L
> > +#define SPI_GDBG_TRAP_CONFIG__PIPE_MATCH_MASK 0x00000100L
> > +#define SPI_GDBG_TRAP_CONFIG__QUEUE_MATCH_MASK 0x00000200L
> > +#define SPI_GDBG_TRAP_CONFIG__TRAP_EN_MASK 0x00008000L
> > +#define SPI_GDBG_TRAP_CONFIG__VMID_SEL_MASK 0xFFFF0000L
> > +//SPI_GDBG_TRAP_MASK
> > +#define SPI_GDBG_TRAP_MASK__EXCP_EN__SHIFT 0x0
> > +#define SPI_GDBG_TRAP_MASK__REPLACE__SHIFT 0x9
> > +#define SPI_GDBG_TRAP_MASK__EXCP_EN_MASK 0x01FFL
> > +#define SPI_GDBG_TRAP_MASK__REPLACE_MASK 0x0200L
> > +//SPI_GDBG_WAVE_CNTL2
> > +#define SPI_GDBG_WAVE_CNTL2__VMID_MASK__SHIFT 0x0
> > +#define SPI_GDBG_WAVE_CNTL2__MODE__SHIFT 0x10
> > +#define SPI_GDBG_WAVE_CNTL2__VMID_MASK_MASK 0x0000FFFFL
> > +#define SPI_GDBG_WAVE_CNTL2__MODE_MASK 0x00030000L
> > +//SPI_GDBG_WAVE_CNTL3
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_PS__SHIFT 0x0
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_VS__SHIFT 0x1
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_GS__SHIFT 0x2
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_HS__SHIFT 0x3
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CSG__SHIFT 0x4
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS0__SHIFT 0x5
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS1__SHIFT 0x6
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS2__SHIFT 0x7
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS3__SHIFT 0x8
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS4__SHIFT 0x9
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS5__SHIFT 0xa
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS6__SHIFT 0xb
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS7__SHIFT 0xc
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_DURATION__SHIFT 0xd
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_MULT__SHIFT 0x1c
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_PS_MASK 0x00000001L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_VS_MASK 0x00000002L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_GS_MASK 0x00000004L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_HS_MASK 0x00000008L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CSG_MASK 0x00000010L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS0_MASK 0x00000020L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS1_MASK 0x00000040L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS2_MASK 0x00000080L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS3_MASK 0x00000100L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS4_MASK 0x00000200L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS5_MASK 0x00000400L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS6_MASK 0x00000800L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_CS7_MASK 0x00001000L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_DURATION_MASK 0x0FFFE000L
> > +#define SPI_GDBG_WAVE_CNTL3__STALL_MULT_MASK 0x10000000L
> > +//SPI_GDBG_TRAP_DATA0
> > +#define SPI_GDBG_TRAP_DATA0__DATA__SHIFT 0x0
> > +#define SPI_GDBG_TRAP_DATA0__DATA_MASK 0xFFFFFFFFL
> > +//SPI_GDBG_TRAP_DATA1
> > +#define SPI_GDBG_TRAP_DATA1__DATA__SHIFT 0x0
> > +#define SPI_GDBG_TRAP_DATA1__DATA_MASK 0xFFFFFFFFL
> >   //SPI_COMPUTE_QUEUE_RESET
> >   #define SPI_COMPUTE_QUEUE_RESET__RESET__SHIFT 0x0
> >   #define SPI_COMPUTE_QUEUE_RESET__RESET_MASK 0x01L
> > diff --git 
> > a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_offset.h 
> > b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_offset.h
> > index 3973110f149c..d09f1a06f4bf 100644
> > --- a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_offset.h
> > +++ b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_offset.h
> > @@ -26,6 +26,8 @@
> >   #define mmSQ_DEBUG_STS_GLOBAL_BASE_IDX 0
> >   #define mmSQ_DEBUG_STS_GLOBAL2 0x10B0
> >   #define mmSQ_DEBUG_STS_GLOBAL2_BASE_IDX 0
> > +#define mmSQ_DEBUG 0x10B1
> > +#define mmSQ_DEBUG_BASE_IDX 0
> >     // addressBlock: gc_sdma0_sdma0dec
> >   // base address: 0x4980
> > @@ -4849,10 +4851,18 @@
> >   #define mmSPI_WCL_PIPE_PERCENT_CS3_BASE_IDX 0
> >   #define mmSPI_GDBG_WAVE_CNTL 0x1f71
> >   #define mmSPI_GDBG_WAVE_CNTL_BASE_IDX 0
> > +#define mmSPI_GDBG_TRAP_CONFIG 0x1f72
> > +#define mmSPI_GDBG_TRAP_CONFIG_BASE_IDX 0
> >   #define mmSPI_GDBG_TRAP_MASK 0x1f73
> >   #define mmSPI_GDBG_TRAP_MASK_BASE_IDX 0
> >   #define mmSPI_GDBG_WAVE_CNTL2 0x1f74
> >   #define mmSPI_GDBG_WAVE_CNTL2_BASE_IDX 0
> > +#define mmSPI_GDBG_WAVE_CNTL3 0x1f75
> > +#define mmSPI_GDBG_WAVE_CNTL3_BASE_IDX 0
> > +#define mmSPI_GDBG_TRAP_DATA0 0x1f78
> > +#define mmSPI_GDBG_TRAP_DATA0_BASE_IDX 0
> > +#define mmSPI_GDBG_TRAP_DATA1 0x1f79
> > +#define mmSPI_GDBG_TRAP_DATA1_BASE_IDX 0
> >   #define mmSPI_COMPUTE_QUEUE_RESET 0x1f7b
> >   #define mmSPI_COMPUTE_QUEUE_RESET_BASE_IDX 0
> >   #define mmSPI_RESOURCE_RESERVE_CU_0 0x1f7c
> > diff --git 
> > a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_sh_mask.h 
> > b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_sh_mask.h
> > index d4e8ff22ecb8..fc85aee010fe 100644
> > --- a/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_sh_mask.h
> > +++ b/drivers/gpu/drm/amd/include/asic_reg/gc/gc_10_3_0_sh_mask.h
> > @@ -47853,6 +47853,10 @@
> >       // addressBlock: sqind
> > +//SQ_DEBUG
> > +#define SQ_DEBUG__SINGLE_MEMOP_MASK 0x00000001L
> > +#define SQ_DEBUG__SINGLE_MEMOP__SHIFT 0x00000000
> > +
> >   //SQ_DEBUG_STS_GLOBAL
> >   #define SQ_DEBUG_STS_GLOBAL2__FIFO_LEVEL_GFX0_MASK 0x000000ffL
> >   #define SQ_DEBUG_STS_GLOBAL2__FIFO_LEVEL_GFX0__SHIFT 0x00000000