Re: [PATCH 06/32] drm/amdgpu: add gfx9 hw debug mode enable and disable calls

2023-02-16 Thread Felix Kuehling



On 2023-01-25 14:53, Jonathan Kim wrote:

Implement the per-device calls to enable or disable HW debug mode for
GFX9 prior to GFX9.4.1.

GFX9.4.1 and onward will require their own enable/disable sequence as
follow on patches.

When hardware debug mode setting is requested, waves will inherit
these settings in the Shader Processor Input's (SPI) Sequencer Global
Block (SQG). This means that the KGD must drain all waves from the SPI
into SQG (approximately 96 SPI clock cycles) prior to debug mode setting
to ensure that the order of operations that the debugger expects with
regards to debug mode setting transaction requests and wave inheritence
of that mode is upheld.

Also ensure that exception overrides are reset to their original state
prior to debug enable or disable.

v2: remove unnecessary static srbm lock renaming.
add comments to explain ignored arguments for debug trap enable and
disable.

Signed-off-by: Jonathan Kim 
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 93 +++
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 9 ++
drivers/gpu/drm/amd/amdkfd/kfd_debug.h | 3 +
3 files changed, 105 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c

index e92b93557c13..94a9fd9bd984 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -646,6 +646,97 @@ int kgd_gfx_v9_wave_control_execute(struct 
amdgpu_device *adev,

return 0;
}
+/*
+ * GFX9 helper for wave launch stall requirements on debug trap setting.
+ *
+ * vmid:
+ * Target VMID to stall/unstall.
+ *
+ * stall:
+ * 0-unstall wave launch (enable), 1-stall wave launch (disable).
+ * After wavefront launch has been stalled, allocated waves must 
drain from

+ * SPI in order for debug trap settings to take effect on those waves.
+ * This is roughly a ~96 clock cycle wait on SPI where a read on
+ * SPI_GDBG_WAVE_CNTL translates to ~32 clock cycles.
+ * KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY indicates the number of 
reads required.

+ *
+ * NOTE: We can afford to clear the entire STALL_VMID field on unstall
+ * because GFX9.4.1 cannot support multi-process debugging due to trap
+ * configuration and masking being limited to global scope. Always assume
+ * single process conditions.
+
+ */
+#define KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY 3
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+ uint32_t vmid,
+ bool stall)
+{
+ int i;
+ uint32_t data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+ if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1))
+ data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_VMID,
+ stall ? 1 << vmid : 0);
+ else
+ data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA,
+ stall ? 1 : 0);
+
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
+
+ if (!stall)
+ return;
+
+ for (i = 0; i < KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY; i++)
+ RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+}
+
+/**


This was flagged by the kernel test robot. Should just be /* because 
it's not a formal doc comment.



+ * restore_dbg_reisters is ignored here but is a general interface 
requirement


Typo: reisters -> registers



+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to 
manually
+ * initialize the debug mode registers after it has disabled GFX off 
during the

+ * debug session.
+ */
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+ bool restore_dbg_registers,
+ uint32_t vmid)
+{
+ mutex_lock(>grbm_idx_mutex);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+ mutex_unlock(>grbm_idx_mutex);
+
+ return 0;
+}
+
+/**


Same as above. With those fixed, the patch is

Reviewed-by: Felix Kuehling 


+ * keep_trap_enabled is ignored here but is a general interface 
requirement

+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+ bool keep_trap_enabled,
+ uint32_t vmid)
+{
+ mutex_lock(>grbm_idx_mutex);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+ WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+ kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+ mutex_unlock(>grbm_idx_mutex);
+
+ return 0;
+}
+
void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
uint32_t vmid, uint64_t page_table_base)
{
@@ -871,6 +962,8 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
.get_atc_vmid_pasid_mapping_info =
kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
.set_vm_context_page_table_base = 
kgd_gfx_v9_set_vm_context_page_table_base,

+ .enable_debug_trap = 

Re: [PATCH 06/32] drm/amdgpu: add gfx9 hw debug mode enable and disable calls

2023-01-28 Thread kernel test robot
Hi Jonathan,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on drm-tip/drm-tip]
[also build test WARNING on linus/master v6.2-rc5]
[cannot apply to drm-misc/drm-misc-next drm/drm-next drm-exynos/exynos-drm-next 
drm-intel/for-linux-next drm-intel/for-linux-next-fixes]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:
https://github.com/intel-lab-lkp/linux/commits/Jonathan-Kim/drm-amdkfd-add-debug-and-runtime-enable-interface/20230128-092952
base:   git://anongit.freedesktop.org/drm/drm-tip drm-tip
patch link:
https://lore.kernel.org/r/20230125195401.4183544-7-jonathan.kim%40amd.com
patch subject: [PATCH 06/32] drm/amdgpu: add gfx9 hw debug mode enable and 
disable calls
config: x86_64-allyesconfig 
(https://download.01.org/0day-ci/archive/20230129/202301291257.prqg0vpg-...@intel.com/config)
compiler: gcc-11 (Debian 11.3.0-8) 11.3.0
reproduce (this is a W=1 build):
# 
https://github.com/intel-lab-lkp/linux/commit/11bb8b2034cd92b687a2d5461298cc72d720d5c9
git remote add linux-review https://github.com/intel-lab-lkp/linux
git fetch --no-tags linux-review 
Jonathan-Kim/drm-amdkfd-add-debug-and-runtime-enable-interface/20230128-092952
git checkout 11bb8b2034cd92b687a2d5461298cc72d720d5c9
# save the config file
mkdir build_dir && cp config build_dir/.config
make W=1 O=build_dir ARCH=x86_64 olddefconfig
make W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash 
drivers/gpu/drm/amd/amdgpu/

If you fix the issue, kindly add following tag where applicable
| Reported-by: kernel test robot 

All warnings (new ones prefixed by >>):

>> drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c:694: warning: This comment 
>> starts with '/**', but isn't a kernel-doc comment. Refer 
>> Documentation/doc-guide/kernel-doc.rst
* restore_dbg_reisters is ignored here but is a general interface 
requirement
   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c:718: warning: This comment 
starts with '/**', but isn't a kernel-doc comment. Refer 
Documentation/doc-guide/kernel-doc.rst
* keep_trap_enabled is ignored here but is a general interface requirement


vim +694 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c

   692  
   693  /**
 > 694   * restore_dbg_reisters is ignored here but is a general interface 
 > requirement
   695   * for devices that support GFXOFF and where the RLC save/restore list
   696   * does not support hw registers for debugging i.e. the driver has to 
manually
   697   * initialize the debug mode registers after it has disabled GFX off 
during the
   698   * debug session.
   699   */
   700  uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
   701  bool restore_dbg_registers,
   702  uint32_t vmid)
   703  {
   704  mutex_lock(>grbm_idx_mutex);
   705  
   706  kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
   707  
   708  WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
   709  
   710  kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
   711  
   712  mutex_unlock(>grbm_idx_mutex);
   713  
   714  return 0;
   715  }
   716  

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests


[PATCH 06/32] drm/amdgpu: add gfx9 hw debug mode enable and disable calls

2023-01-25 Thread Jonathan Kim
Implement the per-device calls to enable or disable HW debug mode for
GFX9 prior to GFX9.4.1.

GFX9.4.1 and onward will require their own enable/disable sequence as
follow on patches.

When hardware debug mode setting is requested, waves will inherit
these settings in the Shader Processor Input's (SPI) Sequencer Global
Block (SQG). This means that the KGD must drain all waves from the SPI
into SQG (approximately 96 SPI clock cycles) prior to debug mode setting
to ensure that the order of operations that the debugger expects with
regards to debug mode setting transaction requests and wave inheritence
of that mode is upheld.

Also ensure that exception overrides are reset to their original state
prior to debug enable or disable.

v2: remove unnecessary static srbm lock renaming.
add comments to explain ignored arguments for debug trap enable and
disable.

Signed-off-by: Jonathan Kim 
---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 93 +++
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h |  9 ++
 drivers/gpu/drm/amd/amdkfd/kfd_debug.h|  3 +
 3 files changed, 105 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index e92b93557c13..94a9fd9bd984 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -646,6 +646,97 @@ int kgd_gfx_v9_wave_control_execute(struct amdgpu_device 
*adev,
return 0;
 }
 
+/*
+ * GFX9 helper for wave launch stall requirements on debug trap setting.
+ *
+ * vmid:
+ *   Target VMID to stall/unstall.
+ *
+ * stall:
+ *   0-unstall wave launch (enable), 1-stall wave launch (disable).
+ *   After wavefront launch has been stalled, allocated waves must drain from
+ *   SPI in order for debug trap settings to take effect on those waves.
+ *   This is roughly a ~96 clock cycle wait on SPI where a read on
+ *   SPI_GDBG_WAVE_CNTL translates to ~32 clock cycles.
+ *   KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY indicates the number of reads 
required.
+ *
+ *   NOTE: We can afford to clear the entire STALL_VMID field on unstall
+ *   because GFX9.4.1 cannot support multi-process debugging due to trap
+ *   configuration and masking being limited to global scope.  Always assume
+ *   single process conditions.
+
+ */
+#define KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY   3
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+   uint32_t vmid,
+   bool stall)
+{
+   int i;
+   uint32_t data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+   if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1))
+   data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_VMID,
+   stall ? 1 << vmid : 0);
+   else
+   data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA,
+   stall ? 1 : 0);
+
+   WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
+
+   if (!stall)
+   return;
+
+   for (i = 0; i < KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY; i++)
+   RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+}
+
+/**
+ * restore_dbg_reisters is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+   bool restore_dbg_registers,
+   uint32_t vmid)
+{
+   mutex_lock(>grbm_idx_mutex);
+
+   kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+   WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+   kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+   mutex_unlock(>grbm_idx_mutex);
+
+   return 0;
+}
+
+/**
+ * keep_trap_enabled is ignored here but is a general interface requirement
+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+   bool keep_trap_enabled,
+   uint32_t vmid)
+{
+   mutex_lock(>grbm_idx_mutex);
+
+   kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+   WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+   kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+   mutex_unlock(>grbm_idx_mutex);
+
+   return 0;
+}
+
 void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
uint32_t vmid,