Hello,
On Tue, Feb 11, 2020 at 02:16:01PM +0800, fugang.duan@xxxxxxx wrote:
> From: Fugang Duan <fugang.duan@xxxxxxx>
>
> There has oops as below happen on i.MX8MP EVK platform that has
> 6G bytes DDR memory.
>
> when (xmit->tail < xmit->head) && (xmit->head == 0),
> it setups one sg entry with sg->length is zero:
> sg_set_buf(sgl + 1, xmit->buf, xmit->head);
>
> if xmit->buf is allocated from >4G address space, and SDMA only
> support <4G address space, then dma_map_sg() will call swiotlb_map()
> to do bounce buffer copying and mapping.
>
> But swiotlb_map() don't allow sg entry's length is zero, otherwise
> report BUG_ON().
>
> So the patch is to correct the tx DMA scatter list.
>
> Oops:
> [...]
It's a bit sad this wasn't catched earlier. The fsl_lpuart driver got
this exact fix in 2016 (d704b2d32c39c256dea659e142a31b875a13c63b) and
obviously nobody looked left or right to check for other drivers being
affected. :-|
> Reported-by: Eagle Zhou <eagle.zhou@xxxxxxx>
> Tested-by: Eagle Zhou <eagle.zhou@xxxxxxx>
> Signed-off-by: Fugang Duan <fugang.duan@xxxxxxx>
Fixes: 7942f8577f2a ("serial: imx: TX DMA: clean up sg initialization")
(Also here nobody looked in detail: The commit log is obviously wrong as
the if condition isn't just inverted.)
Reviewed-by: Uwe Kleine-König <u.kleine-koenig@xxxxxxxxxxxxxx>
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | https://www.pengutronix.de/ |
