Re: [PATCH 1/2] dmaengine: sun6i: Fix use-after-free in slave sg LLI error cleanup

[Frank Li]

On Sun, Sep 27, 2026 at 08:39:50AM +0800, Slavin Liu wrote:
> sun6i_dma_prep_slave_sg() chains one pool-allocated LLI per
> scatterlist entry.  When a later dma_pool_alloc() fails (GFP_NOWAIT,
> so it does not wait for reclaim, making failure realistic under memory
> pressure), the err_lli_free path walks the chain to free what was
> built: the loop body hands the current LLI back to the pool with
> dma_pool_free(), and the for-loop increment then reads
> v_lli->p_lli_next and v_lli->v_lli_next from the memory that was just
> freed and may already have been reallocated.
>
> Cache both next fields of the current LLI before freeing it.
>
> Fixes: 4fbd804e009a ("dmaengine: sun6i: Fix memory leaks")
> Assisted-by: LLM
> Signed-off-by: Slavin Liu <bolin.liu@xxxxxxxxxx>
> ---

similar patch already in next branch

https://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine.git/commit/?id=f472919f4c0650d7fb22bb230eacaada0c6ee951

Frank

>  drivers/dma/sun6i-dma.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/dma/sun6i-dma.c b/drivers/dma/sun6i-dma.c
> index 7704b016aed8..025a43d0298d 100644
> --- a/drivers/dma/sun6i-dma.c
> +++ b/drivers/dma/sun6i-dma.c
> @@ -790,9 +790,14 @@ static struct dma_async_tx_descriptor *sun6i_dma_prep_slave_sg(
>  	return vchan_tx_prep(&vchan->vc, &txd->vd, flags);
>
>  err_lli_free:
> -	for (p_lli = txd->p_lli, v_lli = txd->v_lli; v_lli;
> -	     p_lli = v_lli->p_lli_next, v_lli = v_lli->v_lli_next)
> +	for (p_lli = txd->p_lli, v_lli = txd->v_lli; v_lli;) {
> +		dma_addr_t next_p_lli = v_lli->p_lli_next;
> +		struct sun6i_dma_lli *next_v_lli = v_lli->v_lli_next;
> +
>  		dma_pool_free(sdev->pool, v_lli, p_lli);
> +		p_lli = next_p_lli;
> +		v_lli = next_v_lli;
> +	}
>  	kfree(txd);
>  	return NULL;
>  }
> --
> 2.34.1
>