Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: DMA: atmel_serial: Opening and closing the serial device
 repeatedly causes kmalloc-32 slab leak
To:     Alexandre Belloni <alexandre.belloni@bootlin.com>,
        richard.genoud@gmail.com
Cc:     Ludovic Desroches <ludovic.desroches@microchip.com>,
        Nicolas Ferre <nicolas.ferre@microchip.com>,
        Maxime Ripard <maxime.ripard@bootlin.com>,
        "linux-arm-kernel@lists.infradead.org" 
        <linux-arm-kernel@lists.infradead.org>, dmaengine@vger.kernel.org,
        Linux Kernel <linux-kernel@vger.kernel.org>,
        "linux-serial@vger.kernel.org" <linux-serial@vger.kernel.org>,
        Mario Forner <m.forner@be4energy.com>,
        Vinod Koul <vkoul@kernel.org>
References: <22061488.b0eNpyQjWt@linux-7rm0>
 <83753d21-f3c8-c8dd-75d7-741cb597d1a3@sorico.fr>
 <20181127095843.GF19871@piout.net>
From:   Richard Genoud <richard@sorico.fr>
Message-ID: <febddc12-200d-139f-18f0-fcb055951324@sorico.fr>
Date:   Tue, 27 Nov 2018 11:46:56 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.3.0
MIME-Version: 1.0
In-Reply-To: <20181127095843.GF19871@piout.net>
Content-Type: text/plain; charset=utf-8
Content-Language: fr
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Le 27/11/2018 à 10:58, Alexandre Belloni a écrit :
> Hello Richard,
> 
> On 27/11/2018 10:51:13+0100, richard.genoud@gmail.com wrote:
>> Hi all,
>>
>> I reproduced the memory leak on my board (at91sam9g35-cm) with a 4.20-rc3.
>>
>> It triggered an OOM after a couple of hours running a code like this:
>> #include <sys/types.h>
>> #include <sys/stat.h>
>> #include <fcntl.h>
>> #include <unistd.h>
>>
>>
>> int main(int argc, char **argv)
>> {
>> 	int fd;
>> 	do {
>> 		fd = open("/dev/ttyS1", O_RDONLY);
>> 		close(fd);
>> 	} while (true);
>> 	return 0;
>> }
>>
>> As Mario pointed out, this only happens when atmel,use-dma-{r,t}x are
>> used in the device-tree.
>>
>> Adding:
>> CONFIG_DEBUG_SLAB=y
>> CONFIG_DEBUG_SLAB_LEAK=y
>> Doesn't show anything suspect in /proc/slab_allocators
>>
>> From what I found until now, it's something done in :
>> dma_request_slave_channel();
>> that leaks kmalloc-32
>> Mabe I missed something, but it seems that everything DMA related is
>> deallocated in atmel_release_{tx,rx}_dma().
>>
>> Is this ringing a bell ?
>>
> 
> Yes, this is known issue and it has yet to be worked on.
> 

After a talk on freenode, Alex found the problem.
A patch is on its way.

Thanks !