Re: Flexcan

[Marc Kleine-Budde <mkl@pengutronix.de>]

[-- Attachment #1: Type: text/plain, Size: 2575 bytes --]

On 04/02/2015 01:35 PM, Tom Evans wrote:
> On 2/04/2015 5:28 PM, Marc Kleine-Budde wrote:
>> On 04/02/2015 04:20 AM, Tom Evans wrote:
>>> I had to seriously rewrite the FlexCAN drivers for our i.MX53 because the
>>> Ethernet driver blocked the CAN driver for so long the CAN hardware receive
>>> FIFO overflowed. Linux doesn't even support setting different interrupt
>>> priorities on ARM (but my problem was with NAPI).
>>
>> Can you share your changes? I'll have some days to hack on a better
>> generic RX-FIFO implementation for the at91 and flexcan.
> 
> Sure. "git diff -p" based of the 3.4 kernel sources.

Thanks.

> Working from home over Easter using git under cygwin under Windows via 
> Thunderbird, so the CRLFs and line breaks could be messed up, sorry. 
> Patches inline and attached. If this lot is useless, tell me what to 
> type to do it properly and I'll resend on Tuesday.

Uhhh sounds scary - what about switching to Linux over the holidays? :D
As I don't want to push that patch upstream, but get more insight into
the various problem and solutions, this is alright.

BTW: you can send pathes directly via "git send-email".

> flexcan.c.patch.buffer
> 
> The problem with the FlexCan driver is that it takes the interrupt, then 
> throws to NAPI to actually read the hardware buffers. How messed up is 
> that? It isn't as if reading 16 bytes or so is going to take THAT long, 
> and as there's only six message buffers it overflows in 600us at 1MHz. 
> In the same kernel, the Ethernet driver DOESN'T throw to NAPI but 
> handles as many 1500-byte Ethernet frames as there are coming in in the 
> one ISR. Other things can delay NAPI as well. So I fixed it with a 
> sledgehammer. It reads the messages during the ISR to an internal ring 
> buffer (sized to 100, later to 128)

Sledgehammers are not the best tool for mainline kernel development, but
it's good to demolish old walls and get a clear view to head in the
hopefully right direction.

> flexcan.c.patch.flood
> 
> A disconnected CAN bus generate an interrupt storm. This patch fixes it 
> the same way it is apparently done in some 3.18 drivers. This may 
> already be in there.

We're heading for a different solution, limiting the bus errors.

Thanks,
Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]