drm/i915: Dump the whole context object.
Subsystems (2)
PATCH 1/1 drm/i915: Dump the whole context object.
As we've learned over time, the HW context is just a series of GPU commands that we're able to decode without intel_error_decode. Since many bugs recently have been implicated in the HW context state, it makes sense to dump the whole context object in a form which can be parsed.
Sample:
render ring --- HW Context = 0x042db000
ringbuffer (render ring) at 0x0160c000; HEAD points to: 0x0160c000
0x0160c000: 0x00000000: MI_NOOP 0x0160c004: 0x00000000: MI_NOOP 0x0160c008: 0x00000000: MI_NOOP 0x0160c00c: 0x00000000: MI_NOOP 0x0160c010: 0x00000000: MI_NOOP 0x0160c014: 0x00000000: MI_NOOP 0x0160c018: 0x00000000: MI_NOOP 0x0160c01c: 0x00000000: MI_NOOP
Unfortunately, our decoder isn't quite smart enough to deal with the variable length LRIs - but that is a tools problem.
--- drivers/gpu/drm/i915/i915_gpu_error.c | 16 ++-------------- 1 file changed, 2 insertions(+), 14 deletions(-)
drivers/gpu/drm/i915/i915_gpu_error.c +2 -14
@@ -452,16 +452,7 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *m,err_printf (m,"%s --- HW Context = 0x%08x\n" , dev_priv->ring[i].name, obj->gtt_offset); - offset =0 ; -for (elt =0 ; elt <PAGE_SIZE /16 ; elt +=4 ) { -err_printf (m,"[%04x] %08x %08x %08x %08x\n" , - offset, - obj->pages[0 ][elt], - obj->pages[0 ][elt+1 ], - obj->pages[0 ][elt+2 ], - obj->pages[0 ][elt+3 ]); - offset +=16 ; - } +print_error_obj (m, obj); } }
@@ -927,10 +918,7 @@ static void i915_gem_record_active_context(struct intel_ring_buffer *ring,list_for_each_entry (obj, &dev_priv->mm.bound_list, global_list) {if ((error->ccid &PAGE_MASK ) ==i915_gem_obj_ggtt_offset (obj)) { - ering->ctx =i915_error_object_create_sized (dev_priv, - obj, - &dev_priv->gtt.base, -1 ); + ering->ctx =i915_error_ggtt_object_create (dev_priv, obj);break ; } }
Signature · 1 line
1.9.1
Daniel Vetter
AppliedQuoted 2 lines · Ben Widawsky
As we've learned over time, the HW context is just a series of GPU commands that we're able to decode without
intel_error_decode. Since
Correct to "... without any changes in intel_error_decode. ..." since I presumed that's what you wanted to say.
Quoted 16 lines
many bugs recently have been implicated in the HW context state, it makes sense to dump the whole context object in a form which can be parsed.
Sample:
render ring --- HW Context =0x042db000
ringbuffer (render ring) at0x0160c000; HEAD points to:0x0160c0000x0160c000: 0x00000000: MI_NOOP 0x0160c004: 0x00000000: MI_NOOP 0x0160c008: 0x00000000: MI_NOOP 0x0160c00c: 0x00000000: MI_NOOP 0x0160c010: 0x00000000: MI_NOOP 0x0160c014: 0x00000000: MI_NOOP 0x0160c018: 0x00000000: MI_NOOP 0x0160c01c: 0x00000000: MI_NOOP
Quoted 4 lines
Unfortunately, our decoder isn't quite smart enough to deal with the variable length LRIs - but that is a tools problem.
Signed-off-by Ben Widawsky <[email protected]>
Queued for -next, thanks for the patch.
-Daniel
Quoted 45 lines
--- drivers/gpu/drm/i915/i915_gpu_error.c | 16 ++-------------- 1 file changed, 2 insertions(+), 14 deletions(-)drivers/gpu/drm/i915/i915_gpu_error.c +2
-14@@ -452,16 +452,7 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *m,err_printf (m,"%s --- HW Context = 0x%08x\n" , dev_priv->ring[i].name, obj->gtt_offset); - offset =0 ; -for (elt =0 ; elt <PAGE_SIZE /16 ; elt +=4 ) { -err_printf (m,"[%04x] %08x %08x %08x %08x\n" , - offset, - obj->pages[0 ][elt], - obj->pages[0 ][elt+1 ], - obj->pages[0 ][elt+2 ], - obj->pages[0 ][elt+3 ]); - offset +=16 ; - } +print_error_obj (m, obj); } }@@ -927,10 +918,7 @@ static void i915_gem_record_active_context(struct intel_ring_buffer *ring,list_for_each_entry (obj, &dev_priv->mm.bound_list, global_list) {if ((error->ccid &PAGE_MASK ) ==i915_gem_obj_ggtt_offset (obj)) { - ering->ctx =i915_error_object_create_sized (dev_priv, - obj, - &dev_priv->gtt.base, -1 ); + ering->ctx =i915_error_ggtt_object_create (dev_priv, obj);break ; } }Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch