[Mesa-dev] [PATCH 2/2] radv: do not update vertex descriptors if the allocation failed

[Bas Nieuwenhuizen]

On Fri, Sep 8, 2017 at 3:36 PM, Samuel Pitoiset
<samuel.pitoiset at gmail.com> wrote:
>
>
> On 09/08/2017 03:33 PM, Bas Nieuwenhuizen wrote:
>>
>> While this is a fix, I'm not sure if it is a fix we want without more
>> fixes. If we end up not finishing the flush_state, we could end up
>> emitting a draw packet without having emitted the shader which sounds
>> susceptible to hangs to me.
>
>
> That's true. How about skipping the draw call by returning a boolean in
> radv_cmd_buffer_flush_state() ? That way we don't emit any packets if the
> allocation fails.

That plus also skipping if an error has been set for the command
buffer (for earlier flushes) would be best I think.
>
>
>>
>> On Fri, Sep 8, 2017 at 3:15 PM, Samuel Pitoiset
>> <samuel.pitoiset at gmail.com> wrote:
>>>
>>> A return code error is stored in the command buffer and should
>>> be returned to the user via EndCommandBuffer().
>>>
>>> Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
>>> ---
>>>   src/amd/vulkan/radv_cmd_buffer.c | 12 ++++++++----
>>>   1 file changed, 8 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/src/amd/vulkan/radv_cmd_buffer.c
>>> b/src/amd/vulkan/radv_cmd_buffer.c
>>> index a79b513ffc..07d9be9bf4 100644
>>> --- a/src/amd/vulkan/radv_cmd_buffer.c
>>> +++ b/src/amd/vulkan/radv_cmd_buffer.c
>>> @@ -1578,7 +1578,7 @@ static void radv_emit_primitive_reset_state(struct
>>> radv_cmd_buffer *cmd_buffer,
>>>          }
>>>   }
>>>
>>> -static void
>>> +static bool
>>>   radv_cmd_buffer_update_vertex_descriptors(struct radv_cmd_buffer
>>> *cmd_buffer)
>>>   {
>>>          struct radv_device *device = cmd_buffer->device;
>>> @@ -1594,8 +1594,9 @@ radv_cmd_buffer_update_vertex_descriptors(struct
>>> radv_cmd_buffer *cmd_buffer)
>>>                  uint64_t va;
>>>
>>>                  /* allocate some descriptor state for vertex buffers */
>>> -               radv_cmd_buffer_upload_alloc(cmd_buffer, count * 16, 256,
>>> -                                            &vb_offset, &vb_ptr);
>>> +               if (!radv_cmd_buffer_upload_alloc(cmd_buffer, count * 16,
>>> 256,
>>> +                                                 &vb_offset, &vb_ptr))
>>> +                       return false;
>>>
>>>                  for (i = 0; i < count; i++) {
>>>                          uint32_t *desc = &((uint32_t *)vb_ptr)[i * 4];
>>> @@ -1625,6 +1626,8 @@ radv_cmd_buffer_update_vertex_descriptors(struct
>>> radv_cmd_buffer *cmd_buffer)
>>>                                             AC_UD_VS_VERTEX_BUFFERS, va);
>>>          }
>>>          cmd_buffer->state.vb_dirty = false;
>>> +
>>> +       return true;
>>>   }
>>>
>>>   static void
>>> @@ -1639,7 +1642,8 @@ radv_cmd_buffer_flush_state(struct radv_cmd_buffer
>>> *cmd_buffer,
>>>          MAYBE_UNUSED unsigned cdw_max =
>>> radeon_check_space(cmd_buffer->device->ws,
>>>
>>> cmd_buffer->cs, 4096);
>>>
>>> -       radv_cmd_buffer_update_vertex_descriptors(cmd_buffer);
>>> +       if (!radv_cmd_buffer_update_vertex_descriptors(cmd_buffer))
>>> +               return;
>>>
>>>          if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_PIPELINE)
>>>                  radv_emit_graphics_pipeline(cmd_buffer, pipeline);
>>> --
>>> 2.14.1
>>>
>>> _______________________________________________
>>> mesa-dev mailing list
>>> mesa-dev at lists.freedesktop.org
>>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev