MPC850 SCC3/USB host conflict

MPC850 SCC3/USB host conflict

[Rolf Peukert]

Hi everybody!

We are installing Linux (2.4.4) on a board design using an MPC850
(XSPC850SRZT66C). We're planning to make the device work as USB host
controller, and to use SCC3 as an additional UART.

We're using Brad Parker's USB Host Controller Interface driver
m8xxhci.c, and the USB Host Start-of-Frame microcode patch available on
your site for the MPC823 (MPC823SOFGENMC). Timer 1 is configured to
generate the SOF clock, and PA6/Tout1 is connected to PC15/Dreq0. An
external 48 MHz oscillator on PA4 is used as USB clock.

The standard UART driver (8xx_io/uart.c) is configured to use SMC1 as
console, and SCC3 as an additional UART using PB29/30 (RXD3/TXD3), PC5
(CTS3), PC13 (RTS3) and PC4 (CD3). Ethernet is on SCC2.

As far as I can tell, these pin configurations don't conflict, but still
we are experiencing a conflict between the USB and UART drivers.

USB works fine when SCC3 UART is not enabled, as does SCC3 without USB.
But when both config options are selected, the kernel will hang during
bootup, shortly after the "Enable" bit in the USMOD register is set (in
function rh_port_reset).

With printk statements around that line,
...
         reset_bus_history();
     printk("rh_port_reset() before USMOD_EN\n");
         usbregs->usb_usmod |= USMOD_EN;
     printk("rh_port_reset() after USMOD_EN\n");
         wait_ms(20);
...

the output typically looks like this,
...
Freeing unused kernel memory: 60k init
rh_port_reset(1)
rh_port_reset() before USMOD_EN
rh_port

(with sometimes a few output characters more or less)

The microcode patch is installed during initialization, way before
rh_port_reset() is called.
We found out, that the Kernel will start up with a USB/SCC3
configuration, if the microcode patch is not installed. It will also
start up, if the microcode is installed, but timer 1 (SOF) is not
activated. In these cases, however, USB does not work.

So, we're left rather clueless and will be very grateful for any help.

Are there still possible I/O port conflicts we did not check?
What does the microcode patch actually do?
Is it possible to make USB (host controller mode) and SCC3 UART work
concurrently on the MPC850 at all?


Best Regards,
Rolf


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: MPC850 SCC3/USB host conflict

[Rolf Peukert]

Rolf Peukert wrote:
...

 > USB works fine when SCC3 UART is not enabled, as does SCC3 without
 > USB. But when both config options are selected, the kernel will hang
 > during bootup, ...

It turned out our problems were caused by m8xx_cpm_dpalloc() running out
of memory.
We didn't notice this for so long, because the return value of some
...dpalloc calls in uart.c and enet.c isn't checked. Even worse, our
version of m8xx_cpm_dpalloc didn't return the predefined error flag
CPM_DP_NOSPACE, but just a 0 upon failure (AFAIK the version in the
current kernel doesn't have this problem).

We solved the conflict by making more memory available for ...dpalloc,
as the USB microcode patch only needs the first 512 and last 256 bytes
reserved (CPM_DATAONLY_BASE is now 0x0200, CPM_DATAONLY_SIZE 0x0D00).

Best wishes,
Rolf


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/