Hi Mun, On 20 November 2016 at 11:11, Mun Gwan-gyeong <elong...@gmail.com> wrote: > When the allocation fails on u_vector_init(), it returns 0 > This fixes u_vector_init failure path on anv_block_pool_init > > CID 1394319 > > Signed-off-by: Mun Gwan-gyeong <elong...@gmail.com> > --- > src/intel/vulkan/anv_allocator.c | 6 ++++-- > 1 file changed, 4 insertions(+), 2 deletions(-) > > diff --git a/src/intel/vulkan/anv_allocator.c > b/src/intel/vulkan/anv_allocator.c > index f472213..30bbd69 100644 > --- a/src/intel/vulkan/anv_allocator.c > +++ b/src/intel/vulkan/anv_allocator.c > @@ -269,8 +269,10 @@ anv_block_pool_init(struct anv_block_pool *pool, > if (ftruncate(pool->fd, BLOCK_POOL_MEMFD_SIZE) == -1) > return; > > - u_vector_init(&pool->mmap_cleanups, > - round_to_power_of_two(sizeof(struct anv_mmap_cleanup)), > 128); > + if (!u_vector_init(&pool->mmap_cleanups, > + round_to_power_of_two(sizeof(struct anv_mmap_cleanup)), > + 128)) > + return; > Seems like we don't do any proper teardown in the error paths in this function. Can you address that first ? As-is even though we return here, we'll crash shortly after since the error is not propagated through anv_CreateDevice(). Having a look at the function it doesn't check anv_bo_init_new() either.
At the same time we can drop the (always true) return type for anv_queue_init() and radv_queue_init() (in src/amd/vulkan). Most of the above will be fixes, so please include (in the commit message): Cc: "13.0" <mesa-sta...@lists.freedesktop.org> Keeping Jason in the CC list will be nice. Emil _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
