Count dropped frames on CAN socket

Count dropped frames on CAN socket

[Austin Hendrix]

Hi linux-can,

I'm writing a program which receives CAN frames over SocketCAN. I've 
found on a couple of occasions that my program has had bugs that cause 
it not to empty the CAN socket receive buffer, and the kernel has 
dropped incoming CAN frames.

I had a look through the kernel source code, and found that the kernel's 
socket framework keeps track of dropped frames in the sk_drops field of 
the sock struct. Is there a way to retrieve that dropped frame counter 
from my program, so that it's easier for me to detect problems like this 
in the future? (I've looked in the docs and in /proc and haven't been 
able to find a way to retrieve the sk_drops counter, but it's quite 
possible that I missed something)


Thanks,

-Austin

Re: Count dropped frames on CAN socket

[Brian Silverman]

Hi Austin,

If you enable the (SOL_SOCKET, SO_RX_OVFL) option with setsockopt on a
CAN socket, you'll get a corresponding (cmsg_level == SOL_SOCKET,
cmsg_type == SO_RXQ_OVFL) piece of ancillary data from recvmsg. The
value is a uint32_t containing that exact sk_drops value. The
semantics for this value documented in socket(7) are wrong, but the
commit message adding it (3b885787ea41) gives a reasonable
description. In particular, the message doesn't show up at all until
at least one frame is dropped, and it's a total counter not a delta.

Brian

On Tue, Nov 15, 2016 at 6:32 PM, Austin Hendrix <namniart@gmail.com> wrote:
> Hi linux-can,
>
> I'm writing a program which receives CAN frames over SocketCAN. I've found
> on a couple of occasions that my program has had bugs that cause it not to
> empty the CAN socket receive buffer, and the kernel has dropped incoming CAN
> frames.
>
> I had a look through the kernel source code, and found that the kernel's
> socket framework keeps track of dropped frames in the sk_drops field of the
> sock struct. Is there a way to retrieve that dropped frame counter from my
> program, so that it's easier for me to detect problems like this in the
> future? (I've looked in the docs and in /proc and haven't been able to find
> a way to retrieve the sk_drops counter, but it's quite possible that I
> missed something)
>
>
> Thanks,
>
> -Austin
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-can" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Count dropped frames on CAN socket

[Oliver Hartkopp]

And here is how the dropcount stuff has been integrated in candump:

https://github.com/linux-can/can-utils/commit/3c019ea61169d7f08d10d1fece95433c629eb4a6

Regards,
Oliver


On 11/16/2016 01:08 AM, Brian Silverman wrote:
> Hi Austin,
>
> If you enable the (SOL_SOCKET, SO_RX_OVFL) option with setsockopt on a
> CAN socket, you'll get a corresponding (cmsg_level == SOL_SOCKET,
> cmsg_type == SO_RXQ_OVFL) piece of ancillary data from recvmsg. The
> value is a uint32_t containing that exact sk_drops value. The
> semantics for this value documented in socket(7) are wrong, but the
> commit message adding it (3b885787ea41) gives a reasonable
> description. In particular, the message doesn't show up at all until
> at least one frame is dropped, and it's a total counter not a delta.
>
> Brian
>
> On Tue, Nov 15, 2016 at 6:32 PM, Austin Hendrix <namniart@gmail.com> wrote:
>> Hi linux-can,
>>
>> I'm writing a program which receives CAN frames over SocketCAN. I've found
>> on a couple of occasions that my program has had bugs that cause it not to
>> empty the CAN socket receive buffer, and the kernel has dropped incoming CAN
>> frames.
>>
>> I had a look through the kernel source code, and found that the kernel's
>> socket framework keeps track of dropped frames in the sk_drops field of the
>> sock struct. Is there a way to retrieve that dropped frame counter from my
>> program, so that it's easier for me to detect problems like this in the
>> future? (I've looked in the docs and in /proc and haven't been able to find
>> a way to retrieve the sk_drops counter, but it's quite possible that I
>> missed something)
>>
>>
>> Thanks,
>>
>> -Austin
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-can" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-can" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Re: Count dropped frames on CAN socket

[Austin Hendrix]

Awesome, that's exactly what I was looking for.


Thanks!

-Austin


On 11/16/2016 01:57 AM, Oliver Hartkopp wrote:
> And here is how the dropcount stuff has been integrated in candump:
>
> https://github.com/linux-can/can-utils/commit/3c019ea61169d7f08d10d1fece95433c629eb4a6 
>
>
> Regards,
> Oliver
>
>
> On 11/16/2016 01:08 AM, Brian Silverman wrote:
>> Hi Austin,
>>
>> If you enable the (SOL_SOCKET, SO_RX_OVFL) option with setsockopt on a
>> CAN socket, you'll get a corresponding (cmsg_level == SOL_SOCKET,
>> cmsg_type == SO_RXQ_OVFL) piece of ancillary data from recvmsg. The
>> value is a uint32_t containing that exact sk_drops value. The
>> semantics for this value documented in socket(7) are wrong, but the
>> commit message adding it (3b885787ea41) gives a reasonable
>> description. In particular, the message doesn't show up at all until
>> at least one frame is dropped, and it's a total counter not a delta.
>>
>> Brian
>>
>> On Tue, Nov 15, 2016 at 6:32 PM, Austin Hendrix <namniart@gmail.com> 
>> wrote:
>>> Hi linux-can,
>>>
>>> I'm writing a program which receives CAN frames over SocketCAN. I've 
>>> found
>>> on a couple of occasions that my program has had bugs that cause it 
>>> not to
>>> empty the CAN socket receive buffer, and the kernel has dropped 
>>> incoming CAN
>>> frames.
>>>
>>> I had a look through the kernel source code, and found that the 
>>> kernel's
>>> socket framework keeps track of dropped frames in the sk_drops field 
>>> of the
>>> sock struct. Is there a way to retrieve that dropped frame counter 
>>> from my
>>> program, so that it's easier for me to detect problems like this in the
>>> future? (I've looked in the docs and in /proc and haven't been able 
>>> to find
>>> a way to retrieve the sk_drops counter, but it's quite possible that I
>>> missed something)
>>>
>>>
>>> Thanks,
>>>
>>> -Austin
>>>
>>> -- 
>>> To unsubscribe from this list: send the line "unsubscribe linux-can" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>> -- 
>> To unsubscribe from this list: send the line "unsubscribe linux-can" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>