[ldv-project] [BUG] might sleep functions in atomic context in stmmac_resume()

[Giuseppe CAVALLARO]

Hello Alexey

On 5/17/2014 1:15 AM, Alexey Khoroshilov wrote:
> There are calls to might sleep functions in atomic context in
> stmmac_resume():
> - the first one is clk_prepare_enable(priv->stmmac_clk);
> - the second one is stmmac_hw_setup()
> -> init_dma_desc_rings()
>    -> stmmac_init_rx_buffers()
>      -> __netdev_alloc_skb(GFP_KERNEL)
>
> What is the purpose of spin_lock_irqsave(&priv->lock, flags) section?
> What does it protect against?

Some part of this driver, especially on PM stuff, has been reworked
time ago.
I have in plan to do further tests and investigations
also to understand and try to solve this kind of issues. I do not
want to see this kind of BUGs. Also I have some other patches on
other pending problem I will submit to be reviewed.
I hope to start doing this task in the incoming days.

BR
peppe

>
> Found by Linux Driver Verification project (linuxtesting.org).
>
> --
> Alexey Khoroshilov
> Linux Verification Center, ISPRAS
> web: http://linuxtesting.org
>
>