linux-kernel - Re: Disabling an interrupt in the handler locks the system up

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 21 Oct 2016 21:47:05 +0200
From:   Mason <slash.tmp@...e.fr>
To:     Marc Zyngier <marc.zyngier@....com>
Cc:     Thomas Gleixner <tglx@...utronix.de>,
        Jason Cooper <jason@...edaemon.net>,
        LKML <linux-kernel@...r.kernel.org>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>,
        Sebastian Frias <sf84@...oste.net>
Subject: Re: Disabling an interrupt in the handler locks the system up

On 21/10/2016 21:14, Marc Zyngier wrote:

> Mason wrote:
> 
>> On 21/10/2016 19:46, Marc Zyngier wrote:
>>
>>> On 21/10/16 17:37, Mason wrote:
>>>  
>>>> On my platform, one HW block pulls the interrupt line high
>>>> as long as it remains idle, and low when it is busy.
>>>>
>>>> The device tree node is:
>>>>
>>>> 		test@...22 {
>>>> 			compatible = "vendor,testme";
>>>> 			interrupts = <23 IRQ_TYPE_LEVEL_HIGH>;
>>>> 		};  
>>>
>>> I assume that this is for the sake of the discussion, and that you do
>>> not actually intend to put together such a monstrosity.  
>>
>> It's just missing a reg properties to be a valid node, right?
> 
> If connecting a device that signals its interrupt as level low to an
> input line configured as level high doesn't strike you as a major
> issue, nothing will. At that point, you can put anything you want in
> your DT.

If I understand correctly, you are saying that I should have
specified IRQ_TYPE_LEVEL_LOW, instead of IRQ_TYPE_LEVEL_HIGH?

If the HW outputs 1 when idle, and 0 when busy, that
is level low? (Sorry if this is obvious, I'm absolutely
clueless in this subject matter.)


>>>> I wrote a minimal driver which registers the irq.
>>>> And in the interrupt handler, I disable said irq.
>>>>
>>>> Since the irq is IRQ_TYPE_LEVEL_HIGH, it will fire as soon as
>>>> it is registered (because the block is idle).
>>>>
>>>> Here is the code I've been running, request_irq doesn't return.  
>>>
>>> [...]
>>>   
>>>> And here's what I get when I try to load the module:
>>>> (I'm using the default CONFIG_RCU_CPU_STALL_TIMEOUT=21)  
>>>
>>> [...]
>>>   
>>>> Are we not supposed to disable the irq in the handler?  
>>>
>>> You can. It then depends on what your interrupt controller does to
>>> actually ensure that the interrupt is disabled. Only you can trace it on
>>> your HW to find out.  
>>
>> I'm using an upstream driver on v4.9-rc1
>>
>> http://lxr.free-electrons.com/source/drivers/irqchip/irq-tango.c
>>
>> Given that the system locks up, is it possible there is a bug
>> in the driver?
> 
> That's possible.
> 
>> Which call-back handles enabling/disabling interrupts?
> 
> How about irq_unmask/irq_mask?

I tried following the source from disable_irq_nosync()
as far down as I could.

disable_irq_nosync ->
__disable_irq_nosync ->
__disable_irq ->
irq_disable -> ??

http://lxr.free-electrons.com/source/kernel/irq/chip.c#L232

I don't know if desc->irq_data.chip->irq_disable is defined
by the driver I'm using?

I don't know how the trail goes to irq_mask?

Regards.