linux-kernel - Re: [PATCH v1 41/63] Input: touchscreen: Atmel: Enable IRQ_DISABLE

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]

Message-ID: <20190816172615.GJ121898@dtor-ws>
Date:   Fri, 16 Aug 2019 10:26:15 -0700
From:   Dmitry Torokhov <dmitry.torokhov@...il.com>
To:     Jiada Wang <jiada_wang@...tor.com>
Cc:     nick@...anahar.org, linux-input@...r.kernel.org,
        linux-kernel@...r.kernel.org, george_davis@...tor.com
Subject: Re: [PATCH v1 41/63] Input: touchscreen: Atmel: Enable
 IRQ_DISABLE_UNLAZY flag for interrupt

On Fri, Aug 16, 2019 at 05:35:36PM +0900, Jiada Wang wrote:
> From: Bhuvanesh Surachari <bhuvanesh_surachari@...tor.com>
> 
> The de-/serializer driver has defined only irq_mask "ds90ub927_irq_mask" and
> irq_unmask "ds90ub927_irq_unmask" callback functions. And de-/serializer
> driver doesn't implement the irq_disable and irq_enable callback functions.
> Hence inorder to invoke irq_mask callback function when disable_irq_nosync is
> called the IRQ_DISABLE_UNLAZY interrupt flag should be set. If not the
> disable_irq_nosync will just increment the depth field in the irq
> descriptor only once as shown below.
> 
> disable_irq_nosync
>  __disable_irq_nosync
>   __disable_irq (desc->depth++)
>    irq_disable
>     if irq_disable present -----------> if IRQ_DISABLE_UNLAZYflag set
>              |                  no                  |
>          yes |                                  yes |
>              |                                      |
>      desc->irq_data.chip->irq_disable   desc->irq_data.chip->irq_unmask
>                                          (ds90ub927_irq_mask)
>                                           disable_irq
>                                            __disable_irq_nosync
>                                             __disable_irq
> (desc->depth++)
> But the enable_irq will try to decrement the depth field twice which generates
> the backtrace stating "Unbalanced enable for irq 293". This is because there is
> no IRQ_DISABLE_UNLAZY flag check while calling irq_unmask callback function
> of the "ds90ub927_irq_unmask" de-/serializer via enable_irq.
> 
> enable_irq
>  __enable_irq (desc->depth--)
>   irq_enable
>    if irq_enable present -------------> desc->irq_data.chip->irq_unmask
>               |                no        (ds90ub927_irq_unmask)
>           yes |                            enable_irq
>               |                             __enable_irq (desc->depth--)
>     (desc->irq_data.chip->irq_enable)

I'd prefer if we instead did not use the disable_irq_nosync() in the
driver.

Thanks.

-- 
Dmitry