Re: [RFC PATCH] PCI: rockchip: fix system hang up if activate CONFIG_DEBUG_SHIRQ

From: jeffy
Date: Thu Aug 10 2017 - 05:40:51 EST


Hi Heiko,

On 08/10/2017 05:27 PM, Heiko Stuebner wrote:
Hi Shawn,

Am Donnerstag, 10. August 2017, 16:21:13 CEST schrieb Shawn Lin:
>With CONFIG_DEBUG_SHIRQ enabled, the irq tear down routine
>would still access the irq handler registed as a shard irq.
>Per the comment within the function of __free_irq, it says
>"It's a shared IRQ -- the driver ought to be prepared for
>an IRQ event to happen even now it's being freed". However
>when failing to probe the driver, it may disable the clock
>for accessing the register and the following check for shared
>irq state would call the irq handler which accesses the register
>w/o the clk enabled. That will hang the system forever.
The key point would be to fix the ordering. So you could also just use
devm_add_action to move the clock-disabling into a callback that
gets run at the appropriate time, as devm does the shutdown in the
exact opposite order this should fix your problem.

So until all clocks could be enabled, you would disable them manually
and after that rely on the devm_action to fire at the appropriate time.

ret = clk_prepare_enable(clk1);
if (ret < 0 )
return ret;

ret = clk_prepare_enable(clk2);
if (ret < 0) {
clk_disable_unprepare(clk1);
return ret;
}

devm_add_action_or_reset(dev, rk_pcie_disable_clocks);

right, and we should also move request irq after this, so the devres core can release them in reverse.

and we should double check would those clks be the only thing that irq handler required...

The rockchip crypto driver [0] shows how this could be done.


Heiko

[0]http://elixir.free-electrons.com/linux/latest/source/drivers/crypto/rockchip/rk3288_crypto.c#L307