2019-01-07 09:18:50

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: [BUG] tty: serial: mxs-auart: possible concurrency use-after-free bugs in mxs_auart_dma_exit_channel()

On Mon, Jan 07, 2019 at 04:47:43PM +0800, Jia-Ju Bai wrote:
> The driver functions mxs_auart_settermios(), dma_rx_callback() and dma_tx_callback() can be concurrently executed.
>
> In Linux 4.19:
>
> mxs_auart_settermios
> mxs_auart_dma_exit
> mxs_auart_dma_exit_channel
> line 918: kfree(s->tx_dma_buf);
> line 919: kfree(s->rx_dma_buf);
>
> dma_rx_callback
> line 862: tty_insert_flip_string(port, s->rx_dma_buf, count);
> mxs_auart_dma_prep_rx
> line 890: sg_init_one(sgl, s->rx_dma_buf, UART_XMIT_SIZE);
>
> dma_tx_callback
> mxs_auart_tx_chars
> line 590: void *buffer = s->tx_dma_buf;
> mxs_auart_dma_tx
> line 566: sg_init_one(sgl, s->tx_dma_buf, size);
>
> Thus, possible concurrency use-after-free bugs may occur.
>
> These possible bugs are found by a static analysis tool written by myself and my manual code review.

Care to send a patch to fix up this issue?

thanks,

greg k-h


2019-01-07 09:06:33

by Jia-Ju Bai

[permalink] [raw]
Subject: Re: [BUG] tty: serial: mxs-auart: possible concurrency use-after-free bugs in mxs_auart_dma_exit_channel()



On 2019/1/7 16:52, Greg KH wrote:
> On Mon, Jan 07, 2019 at 04:47:43PM +0800, Jia-Ju Bai wrote:
>> The driver functions mxs_auart_settermios(), dma_rx_callback() and dma_tx_callback() can be concurrently executed.
>>
>> In Linux 4.19:
>>
>> mxs_auart_settermios
>> mxs_auart_dma_exit
>> mxs_auart_dma_exit_channel
>> line 918: kfree(s->tx_dma_buf);
>> line 919: kfree(s->rx_dma_buf);
>>
>> dma_rx_callback
>> line 862: tty_insert_flip_string(port, s->rx_dma_buf, count);
>> mxs_auart_dma_prep_rx
>> line 890: sg_init_one(sgl, s->rx_dma_buf, UART_XMIT_SIZE);
>>
>> dma_tx_callback
>> mxs_auart_tx_chars
>> line 590: void *buffer = s->tx_dma_buf;
>> mxs_auart_dma_tx
>> line 566: sg_init_one(sgl, s->tx_dma_buf, size);
>>
>> Thus, possible concurrency use-after-free bugs may occur.
>>
>> These possible bugs are found by a static analysis tool written by myself and my manual code review.
> Care to send a patch to fix up this issue?

I would like to, but I do not know how to fix these bugs properly...
There is no lock and lock-related function call in
drivers/tty/serial/mxs-auart.c.
Thus, we may need to introduce a new lock in this source file.
What is your opinion?


Best wishes,
Jia-Ju Bai