Canbus Filter Matching Index

[e-zeki]

Hi,

I'm fairly new to arm platform. I'm trying to communicate between 2 MCU via Canbus.

with the latest HAL Library I set up 4 filters (2 for each fifo) id mask type , 32Bit.

I am receiving my messages from correct fifos. what I cant get is filter matching index.

If I set all my filters to FIFO0 then I get 2x Filter Bank Number(ex: Filter bank number = 4 / Matching index = 8)

But if I set the filters to FIFO1 then I get 0 no matter I do.

I read the refference manual for filtering but I couldn't find any detailed explanation for it

any help on this?

[Tesla DeLorean]

Seem to recall @Community member​ and I having a unhelpful discussion with ST about this issue some years back.

I don't think you can pull the FIFO out of order, so it would be just as efficient to pull the message number once you get the data in a buffer, and dispatch via a singular vector/jump table

[e-zeki]

So if I understand you correctly, this feature has issues by itself.

it works for FIFO0. I get different matching indexes from the filters setted to FIFO0 but I get all the messages from related FIFOs filters does work, only indexes are weird. Also indexes are different from what refference manual says.

[Lukasz Przenioslo]

Hi @e-zeki​ and @Community member​ 

I dont recall exactly what was the case, but after reading your comments the case sounds familiar- the FIFO0 caught all messages and FIFO1 did not catch anything. There was a bug in the HAL lib where FIFO is never set. Since the #define CAN_FIFO0 is equal to 0, and the variable is global, it works. CAN_FIFO1 is 1, so the requirement is never met. I remember I had to set the pRxMsg->FIFONumber variable to CAN_FIFO1 by hand in the interrupt routine.

Something more or less like that, sorry if cannot be of more help. I did not work with CAN peripheral in STM32 for some years.

Too bad this bug is still out there after so much time though...

[rammit]

I'm also currently working with the F7 bxCAN lib. I haven't gotten far enough to see the FIFO1 issue yet, but I think there is some discrepency between the manual and the library in terms of what a filter bank and a filter number is.

The manual says you can have up to 4 filter "numbers" per bank in 16-bit ID list mode.

But the hal_can.c file uses "FilterNumber" when really it is refering to a bank.

   /* Second 16-bit identifier and Second 16-bit mask */

   /* Or Third 16-bit identifier and Fourth 16-bit identifier */

   can_ip->sFilterRegister[sFilterConfig->FilterNumber].FR2 =

      ((0x0000FFFFU & (uint32_t)sFilterConfig->FilterMaskIdHigh) << 16U) |

       (0x0000FFFFU & (uint32_t)sFilterConfig->FilterIdHigh);

Sincerely,

Confused

[Lukasz Przenioslo]

Hi,
Yes, I also remember some numbering mismatch...

[rammit]

I was using Cube V1.14. Looks like they completely revamped it in v1.15. Looks good. They changed FilterNumber to FilterBank that I mentioned in my last post. @e-zeki​  Are you using 1.15?

[e-zeki]

I'm working on stm32F4 so it says v1.24.0 in Cube Repository. Before that I was using v1.16.0. Its totally different library than previous one. FMI and FIFO assign wasnt working in previous one. FMI still has issues in current one or I couldn't get it right.

[rammit]

@e-zeki​ I should have mentioned I'm using F7 Cube. I assume the F4 example code is the same because it is the same bxCAN peripheral. Not sure if you fixed your problem yet but I discovered that the latest example code does not handle FIFO1 interrupts, only FIFO0 interrupts (see main.h).

#define CANx_RX_IRQHandler            CAN1_RX0_IRQHandler

which calls:

 HAL_CAN_IRQHandler(&CanHandle);

which "handles" all interrupts by checking IER. That's not right.

But at least there's more useful code to extract now.