Am 16.01.24 um 08:24 schrieb Yangyu Chen:
Some platforms may not have any memory in ZONE_DMA32 and use IOMMU to allow
32-bit-DMA-only device to work. Forcing GFP_DMA32 on dummy_read_page will
fail in such platforms. Only use DMA32 when it must to get the bug
resolved.
Well that makes no sense.
If a platform doesn't have a ZONE_DMA32 then GFP_DMA32 is just ignored
as far as I know.
This patch here won't work since the use_dma32 flag is a per device flag
which can't be used for the global initialization.
Otherwise this can randomly fail depending on if a DMA32 device
initializes first or after some device with larger addressing
capabilities. This configuration is quite common on older motherboards
with both integrated and dedicated graphics.
Regards,
Christian.
Signed-off-by: Yangyu Chen <cyy@xxxxxxxxxxxx>
---
drivers/gpu/drm/ttm/ttm_device.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_device.c b/drivers/gpu/drm/ttm/ttm_device.c
index d48b39132b32..62f16fb72428 100644
--- a/drivers/gpu/drm/ttm/ttm_device.c
+++ b/drivers/gpu/drm/ttm/ttm_device.c
@@ -63,7 +63,7 @@ static void ttm_global_release(void)
mutex_unlock(&ttm_global_mutex);
}
-static int ttm_global_init(void)
+static int ttm_global_init(bool use_dma32)
{
struct ttm_global *glob = &ttm_glob;
unsigned long num_pages, num_dma32;
@@ -95,7 +95,8 @@ static int ttm_global_init(void)
ttm_pool_mgr_init(num_pages);
ttm_tt_mgr_init(num_pages, num_dma32);
- glob->dummy_read_page = alloc_page(__GFP_ZERO | GFP_DMA32);
+ glob->dummy_read_page = use_dma32 ? alloc_page(__GFP_ZERO | GFP_DMA32) :
+ alloc_page(__GFP_ZERO);
if (unlikely(glob->dummy_read_page == NULL)) {
ret = -ENOMEM;
@@ -200,7 +201,7 @@ int ttm_device_init(struct ttm_device *bdev, const struct ttm_device_funcs *func
if (WARN_ON(vma_manager == NULL))
return -EINVAL;
- ret = ttm_global_init();
+ ret = ttm_global_init(use_dma32);
if (ret)
return ret;
