summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/nvgpu/include/nvgpu/dma.h
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/nvgpu/include/nvgpu/dma.h')
-rw-r--r--drivers/gpu/nvgpu/include/nvgpu/dma.h5
1 files changed, 5 insertions, 0 deletions
diff --git a/drivers/gpu/nvgpu/include/nvgpu/dma.h b/drivers/gpu/nvgpu/include/nvgpu/dma.h
index 43cff215..1c6474e7 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/dma.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/dma.h
@@ -197,6 +197,11 @@ void nvgpu_dma_free(struct gk20a *g, struct nvgpu_mem *mem);
197 * Note this is different than mapping it into the CPU. This memory can be 197 * Note this is different than mapping it into the CPU. This memory can be
198 * either placed in VIDMEM or SYSMEM, which ever is more convenient for the 198 * either placed in VIDMEM or SYSMEM, which ever is more convenient for the
199 * driver. 199 * driver.
200 *
201 * Note: currently a bug exists in the nvgpu_dma_alloc_map*() routines: you
202 * cannot use nvgpu_gmmu_map() on said buffer - it will overwrite the necessary
203 * information for the DMA unmap routines to actually unmap the buffer. You
204 * will either leak mappings or see GMMU faults.
200 */ 205 */
201int nvgpu_dma_alloc_map(struct vm_gk20a *vm, size_t size, 206int nvgpu_dma_alloc_map(struct vm_gk20a *vm, size_t size,
202 struct nvgpu_mem *mem); 207 struct nvgpu_mem *mem);