Hi Wolfram,
Wolfram Sang wrote:
> Hi Wolfgang,
>
> On Mon, Nov 02, 2009 at 08:23:41PM +0100, Wolfgang Grandegger wrote:
>> Hi Wolfram,
>>
>> what is your plan now? Provide patches for mainline inclusion
>> (net-next-2.6) or for the Socket-CAN SVN trunk? Please don't send
>> patches for both. We can do the backport to the SVN trunk later.
>
> I changed my plans a bit. I'd now like to concentrate on MPC52xx only, bring
> that to mainline and add MPC512x support, when we the clocking issues have
> settled upstream. I still have all the patches for adding MPC512x anytime, and
> they just need one hook inside the driver, so no big reorganization is needed
> then. Are you okay with this, more incremental, approach?
Fine for me. But how do we select the clocks for the MSCAN on the
MPC5200? Currently it's implemented as shown below:
if (of_get_property(np, "clock-ipb", NULL))
clock_src = MSCAN_CLKSRC_BUS;
else
clock_src = MSCAN_CLKSRC_XTAL;
But we should have something compatible with the MPC512x, e.g.:
fsl,mscan-clock-source = "clk-ipb".
>> What's also missing is the proper bus-off recovery method for the
>> MCP512x. Unfortunately, it's also not yet OK for the MPC5200. It should
>> be implemented as listed below:
>>
>> - if possible, automatic bus-off recovery should be turned off, which
>> is possible on the MPC512x (but not the default for compatibility with
>> the MPC5200.
>>
>> - if bus-off recovery cannot be switched of, the driver should handle it
>> as follows:
>>
>> - if restart_ms == 0, the device should be stopped on bus-off to
>> suppress automatic recovery.
>>
>> - if restart_ms > 0, the hardware should do the automatic recovery and
>> a RESTARTED error message should be sent when it re-enters the
>> error-active state (or leaves the bus-off state).
>
> I am a bit puzzled here: Fu just showed me that it is already implemented like
> this? Or what exactly is missing?
Marc has implemented it for the AT91 CAN. Search for "restart_ms" and
"RESTARTED" in "at_can.c" to understand what I'm speaking about.
Wolfgang.
_______________________________________________
Socketcan-core mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/socketcan-core
