linux-kernel - Re: [PATCH] serial: samsung: fix the inconsistency in spinlock

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, 19 Feb 2016 17:27:15 +0900
From:	Krzysztof Kozlowski <k.kozlowski@...sung.com>
To:	Anand Moon <linux.amoon@...il.com>
Cc:	Peter Hurley <peter@...leysoftware.com>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Jiri Slaby <jslaby@...e.com>, linux-serial@...r.kernel.org,
	"linux-samsung-soc@...r.kernel.org" 
	<linux-samsung-soc@...r.kernel.org>,
	Linux Kernel <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] serial: samsung: fix the inconsistency in spinlock

On 19.02.2016 17:23, Anand Moon wrote:
> Hi Krzysztof,
> 
> On 19 February 2016 at 13:14, Krzysztof Kozlowski
> <k.kozlowski@...sung.com> wrote:
>> On 19.02.2016 15:51, Anand Moon wrote:
>>> Hi Krzysztof,
>>>
>>> On 19 February 2016 at 11:39, Krzysztof Kozlowski
>>> <k.kozlowski@...sung.com> wrote:
>>>> 2016-02-19 4:14 GMT+09:00 Anand Moon <linux.amoon@...il.com>:
>>>>> Hi Peter,
>>>>>
>>>>> On 18 February 2016 at 23:18, Peter Hurley <peter@...leysoftware.com> wrote:
>>>>>> Hi Anand,
>>>>>>
>>>>>> On 02/18/2016 09:40 AM, Anand Moon wrote:
>>>>>>> From: Anand Moon <linux.amoon@...il.com>
>>>>>>>
>>>>>>> changes fix the correct order of the spin_lock_irqrestore/save.
>>>>>>>
>>>>>>> Signed-off-by: Anand Moon <linux.amoon@...il.com>
>>>>>>> ---
>>>>>>>  drivers/tty/serial/samsung.c | 4 ++--
>>>>>>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>>>>>>
>>>>>>> diff --git a/drivers/tty/serial/samsung.c b/drivers/tty/serial/samsung.c
>>>>>>> index d72cd73..96fe14d 100644
>>>>>>> --- a/drivers/tty/serial/samsung.c
>>>>>>> +++ b/drivers/tty/serial/samsung.c
>>>>>>> @@ -759,9 +759,9 @@ static irqreturn_t s3c24xx_serial_tx_chars(int irq, void *id)
>>>>>>>       }
>>>>>>>
>>>>>>>       if (uart_circ_chars_pending(xmit) < WAKEUP_CHARS) {
>>>>>>> -             spin_unlock(&port->lock);
>>>>>>> +             spin_unlock_irqrestore(&port->lock, flags);
>>>>>>>               uart_write_wakeup(port);
>>>>>>> -             spin_lock(&port->lock);
>>>>>>> +             spin_lock_irqsave(&port->lock, flags);
>>>>>>
>>>>>> This driver shouldn't be dropping the spin lock at for write wakeup.
>>>>>> If this is causing lock-ups in a line discipline, the line discipline
>>>>>> needs fixed.
>>>>>>
>>>>>
>>>>> Thanks for pointing out.
>>>>> Their is no lock up, just the inconstancy of the spin_lock.
>>>>> Then I will resend this patch dropping the spin_unlock/spin_lock
>>>>> around uart_write_wakeup.
>>>>> Is that ok with you.
>>>>
>>>> Anand, before doing that, can you check Peter's second sentence? I
>>>> mean the "If this is causing lock-ups in a line discipline, the line
>>>> discipline needs fixed.".
>>>> Don't drop the spin-locks "just because". I would be happy to see more
>>>> detailed explanation in changelog.
>>>>
>>>> Best regards,
>>>> Krzysztof
>>>
>>> Yes I understood the meaning of the sentence. Already the
>>> s3c24xx_serial_tx_chars function.
>>> holds the lock port->lock for safe IRQ execution.
>>
>> I am sorry but I don't get your explanation. I mentioned Peter's
>> thoughts about lockups after adding locking over uart_write(). However
>> you are referring to s3c24xx_serial_tx_chars() holding the spin lock...
>> I am missing the point...
>>
>> Best regards,
>> Krzysztof
>>
> 
> I should be sorry I could not explain you in technical terms.
> Interrupt routine already hold the port->lock
> 
> s3c24xx_serial_tx_chars
>      \
>      spin_lock_irqsave(&port->lock, flags);
>      \...
>     spin_unlock(&port->lock);
>      uart_write_wakeup(port);
>      spin_lock(&port->lock);
>      \
>      spin_unlock_irqrestore(&port->lock, flags);
> 

This is obvious.

> In my next patch I have tried to remove the spin_unlock/spin_lock over
> uart_write_wakeup(port);

Which may create lockups. Previously there was no port locking around
uart_write_wakeup. Now there will be. You are effectively adding locking
over uart_write_wakeup().
Again, we are back at the Peter's message - just check the damned lockups...

BR,
Krzysztof

BR