On Thu, Jun 6, 2024 at 10:27 AM Xuan Zhuo [off-list ref] wrote:

On Thu, 6 Jun 2024 10:20:21 +0200, Alexander Potapenko [off-list ref] wrote:

quoted

quoted

Could you try this?

Hi Xuan,

What kernel revision does this patch apply to? I tried it against
v6.10-rc2, and only the first hunk applied.
However this seems to fix the problem, at least the kernel boots without
warnings now.

Sorry, I have some changes locally.

If the hunk #1 is applied, then it is ok.

Do you think we need more test? Or I post an new patch directly.

I think we're good to go (just double-checked that). Assuming you're
sending the following patch:

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 2a972752ff1bc..9d3a9942c8c82 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c

@@ -3121,8 +3121,10 @@ dma_addr_t

virtqueue_dma_map_single_attrs(struct virtqueue *_vq, void *ptr,
 {
        struct vring_virtqueue *vq = to_vvq(_vq);

-       if (!vq->use_dma_api)
+       if (!vq->use_dma_api) {
+               kmsan_handle_dma(virt_to_page(ptr),
offset_in_page(ptr), size, dir);
                return (dma_addr_t)virt_to_phys(ptr);
+       }

        return dma_map_single_attrs(vring_dma_dev(vq), ptr, size, dir, attrs);
 }

, feel free to add:

Tested-by: Alexander Potapenko <glider@google.com>