[Mesa-dev] [PATCH 4/8] i965: Store mcs buffer size

Lionel Landwerlin lionel.g.landwerlin at intel.com
Mon Nov 7 10:22:46 UTC 2016


On 07/11/16 10:07, Pohjolainen, Topi wrote:
> On Thu, Nov 03, 2016 at 10:39:39AM +0000, Lionel Landwerlin wrote:
>> From: Ben Widawsky <benjamin.widawsky at intel.com>
>>
>> libdrm may round up the allocation requested by mesa. As a result, accesses
>> through the gtt may end up accessing memory which does not belong to mesa. The
>> problem is described in the following commit:
>> commit 7ae870211ddc40ef6ed209a322c3a721214bb737
>> Author: Eric Anholt <eric at anholt.net>
>> Date:   Mon Apr 14 16:52:43 2014 -0700
>>
>>      i965: Fix buffer overruns in MSAA MCS buf This size field is an alternate
>>
>> In that patch this was solved by making sure we only 1'd the logical size of the
>> buffer. This patch becomes necessary because the miptree data structure is going
>> to go away in the upcoming patch and we won't have access to the total_height
>> field anymore.
>>
>> v2: drop setting the size in intel_hiz_miptree_buf_create() (Lionel)
>>
>> Signed-off-by: Ben Widawsky <benjamin.widawsky at intel.com> (v1)
>> Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin at intel.com> (v2)
>> ---
>>   src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 4 ++--
>>   src/mesa/drivers/dri/i965/intel_mipmap_tree.h | 9 +++++++++
>>   2 files changed, 11 insertions(+), 2 deletions(-)
>>
>> diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> index c2bff17..3d1bdb1 100644
>> --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> @@ -1503,8 +1503,7 @@ intel_miptree_init_mcs(struct brw_context *brw,
>>         return;
>>      }
>>      void *data = mt->mcs_buf->bo->virtual;
>> -   memset(data, init_value,
>> -          mt->mcs_buf->mt->total_height * mt->mcs_buf->mt->pitch);
> If I read the previous patch right, this is already needed there as
> mcs_buf->mt is left NULL?

Thanks, that's wrong indeed.
I think it makes sense to squash patch 3 & 4.

>
>> +   memset(data, init_value, mt->mcs_buf->size);
>>      drm_intel_bo_unmap(mt->mcs_buf->bo);
>>      mt->fast_clear_state = INTEL_FAST_CLEAR_STATE_CLEAR;
>>   }
>> @@ -1545,6 +1544,7 @@ intel_mcs_miptree_buf_create(struct brw_context *brw,
>>   
>>      buf->bo = temp_mt->bo;
>>      buf->offset = temp_mt->offset;
>> +   buf->size = temp_mt->total_height * temp_mt->pitch;
>>      buf->pitch = temp_mt->pitch;
>>      buf->qpitch = temp_mt->qpitch;
>>   
>> diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.h b/src/mesa/drivers/dri/i965/intel_mipmap_tree.h
>> index 0b49dc2..0b4b353 100644
>> --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.h
>> +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.h
>> @@ -350,6 +350,15 @@ struct intel_miptree_aux_buffer
>>       */
>>      uint32_t offset;
>>   
>> +   /*
>> +    * Size of the MCS surface.
>> +    *
>> +    * This is needed when doing any gtt mapped operations on the buffer (which
>> +    * will be Y-tiled). It is possible that it will not be the same as bo->size
>> +    * when the drm allocator rounds up the requested size.
>> +    */
>> +   size_t size;
>> +
>>      /**
>>       * Pitch in bytes.
>>       *
>> -- 
>> 2.10.2
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev




More information about the mesa-dev mailing list