linux-kernel - Re: [PATCH 27/31] usb/gadget/snps_udc_core: Move timer initialization earlier

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 for Android: free password hash cracker in your pocket

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

Date:   Wed, 6 Sep 2017 16:23:51 -0700
From:   Kees Cook <keescook@...omium.org>
To:     Michal Nazarewicz <mina86@...a86.com>
Cc:     Thomas Gleixner <tglx@...utronix.de>,
        Felipe Balbi <balbi@...nel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Raviteja Garimella <raviteja.garimella@...adcom.com>,
        "Gustavo A. R. Silva" <garsilva@...eddedor.com>,
        linux-usb@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 27/31] usb/gadget/snps_udc_core: Move timer initialization earlier

On Sun, Sep 3, 2017 at 2:12 PM, Michal Nazarewicz <mina86@...a86.com> wrote:
> On Thu, Aug 31 2017, Kees Cook wrote:
>> With timer initialization made earlier at the start, there is no reason
>> to make del_timer_sync() calls conditionally, there by removing the
>> assignments and tests of the .data field.
>>
>> Cc: Felipe Balbi <balbi@...nel.org>
>> Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
>> Cc: Raviteja Garimella <raviteja.garimella@...adcom.com>
>> Cc: Michal Nazarewicz <mina86@...a86.com>
>> Cc: "Gustavo A. R. Silva" <garsilva@...eddedor.com>
>> Cc: linux-usb@...r.kernel.org
>> Signed-off-by: Kees Cook <keescook@...omium.org>
>> ---
>>  drivers/usb/gadget/udc/snps_udc_core.c | 16 +++++-----------
>>  1 file changed, 5 insertions(+), 11 deletions(-)
>>
>> diff --git a/drivers/usb/gadget/udc/snps_udc_core.c b/drivers/usb/gadget/udc/snps_udc_core.c
>> index 5460e5ba1c3c..1607e901e16b 100644
>> --- a/drivers/usb/gadget/udc/snps_udc_core.c
>> +++ b/drivers/usb/gadget/udc/snps_udc_core.c
>> @@ -3067,14 +3067,12 @@ void udc_remove(struct udc *dev)
>>       stop_timer++;
>>       if (timer_pending(&udc_timer))
>>               wait_for_completion(&on_exit);
>> -     if (udc_timer.data)
>> -             del_timer_sync(&udc_timer);
>> +     del_timer_sync(&udc_timer);
>>       /* remove pollstall timer */
>>       stop_pollstall_timer++;
>>       if (timer_pending(&udc_pollstall_timer))
>>               wait_for_completion(&on_pollstall_exit);
>> -     if (udc_pollstall_timer.data)
>> -             del_timer_sync(&udc_pollstall_timer);
>> +     del_timer_sync(&udc_pollstall_timer);
>>       udc = NULL;
>>  }
>>  EXPORT_SYMBOL_GPL(udc_remove);
>> @@ -3164,9 +3162,9 @@ int udc_probe(struct udc *dev)
>>       u32             reg;
>>       int             retval;
>>
>> -     /* mark timer as not initialized */
>> -     udc_timer.data = 0;
>> -     udc_pollstall_timer.data = 0;
>> +     /* timer init */
>> +     setup_timer(&udc_timer, udc_timer_function, 0);
>> +     setup_timer(&udc_pollstall_timer, udc_pollstall_timer_function, 0);
>>
>>       /* device struct setup */
>>       dev->gadget.ops = &udc_ops;
>> @@ -3206,10 +3204,6 @@ int udc_probe(struct udc *dev)
>>       if (retval)
>>               goto finished;
>>
>> -     /* timer init */
>> -     setup_timer(&udc_timer, udc_timer_function, 1);
>> -     setup_timer(&udc_pollstall_timer, udc_pollstall_timer_function, 1);
>> -
>>       /* set SD */
>>       reg = readl(&dev->regs->ctl);
>>       reg |= AMD_BIT(UDC_DEVCTL_SD);
>
>
> Stupid question, is the check in udc_remove even necessary?
>
> udc_probe is called from udc_plat_probe:
>
>         if (udc_probe(udc)) {
>                 ret = -ENODEV;
>                 goto exit_dma;
>         }
>
> If the call fails, udc_plat_probe cleans up after itself and noticeably
> *does not* call udc_remove.  As far as I understand, if probe callback
> fails, remove callback is *not* called.  Meanwhile, udc_remove is called
> from the remove callback which is udc_plat_remove.  So, udc_remove can
> be called only if udc_probe succeeds.
>
> It seems to me that a better patch is:
>
> --- a/drivers/usb/gadget/udc/snps_udc_core.c
> +++ b/drivers/usb/gadget/udc/snps_udc_core.c
> @@ -3067,14 +3067,12 @@ void udc_remove(struct udc *dev)
>         stop_timer++;
>         if (timer_pending(&udc_timer))
>                 wait_for_completion(&on_exit);
> -       if (udc_timer.data)
> -               del_timer_sync(&udc_timer);
> +       del_timer_sync(&udc_timer);
>         /* remove pollstall timer */
>         stop_pollstall_timer++;
>         if (timer_pending(&udc_pollstall_timer))
>                 wait_for_completion(&on_pollstall_exit);
> -       if (udc_pollstall_timer.data)
> -               del_timer_sync(&udc_pollstall_timer);
> +       del_timer_sync(&udc_pollstall_timer);
>         udc = NULL;
>  }
>  EXPORT_SYMBOL_GPL(udc_remove);
> @@ -3164,9 +3162,9 @@ int udc_probe(struct udc *dev)
>         u32             reg;
>         int             retval;
>
> -       /* mark timer as not initialized */
> -       udc_timer.data = 0;
> -       udc_pollstall_timer.data = 0;
> -
>         /* device struct setup */
>         dev->gadget.ops = &udc_ops;
>

Okay, sounds good. I'll adjust the patch. Thanks!

-Kees

-- 
Kees Cook
Pixel Security