Hi to everyone,
I forward this message to the LPC2000 list, as I'm still waiting for approval on LPC2400 list.

----- Forwarded Message ----
From: Andrea Barberio
To: l...
Sent: Thu, May 27, 2010 12:28:28 PM
Subject: Serial driver

Hi to everyone,
this is my first post to the mailing list. I am working on an LPC2478 branded by EmbeddedArtists,
which uses uClinux with a 2.6.21 linux kernel, and I'm having issues with the serial ports.
My application needs to communicate with a serial device (only RX/TX required), and the board
makes available UART #0 (used for the serial console, working fine), UART #1 (full modem port)
and UART #2 (RX/TX only, it's mutually exclusive to UART #1 if I understand correctly).

With a simple open() on /dev/ttyS1, UART #1 simply screams of pain and floods the terminal with
the following message:

"too much work for irq7".

I've searched around and I found it's a common problem. I also contacted the technical support from
the vendor, but the only "solution" I came across is the one at the following link:

http://tech.groups.yahoo.com/group/lpc2400_uClinux/message/692

which enables UART #2 instead of UART #1, by patching both the bootloader and the 8250 driver.
This solution doesn't work for me, even after a lot of work and further patches. Since I was spending
too much time, I decided to work around it for some time using an FTDI usb-serial converter, while
keeping searching a solution on the real serial port. USB-serial wasn't a solution, not even as
workaround, as the usb-serial converter doesn't work as expected. It works fine with the same
application code and converter on a PC with Linux, but doesn't work on uClinux/LPC2478: using a
serial analyzer it seems like a parity problem.

So, I had to move back to the original problem: the real serial port. Has anyone suggestions or
tests/patches to try to solve it? I really need that serial port to work, and I think it's impossible
that no one can use it on uClinux.

Thanks in advance,
Andrea Barberio Start a New Thread

--- In l..., Andrea Barberio wrote:
serial analyzer it seems like a parity problem.
>
> So, I had to move back to the original problem: the real serial port. Has anyone suggestions or
> tests/patches to try to solve it? I really need that serial port to work, and I think it's impossible
> that no one can use it on uClinux.
>
> Thanks in advance,
> Andrea Barberio
>

Let's assume the serial port really is causing excessive interrupts. Why would that be? Since this port has full modem control, are the control signals in a state that would result in an interrupt?

Have you tried actually controlling all of the modem signals? What about disabling those interrupts in the driver?

Richard

Start a New Thread

> From: rtstofer

> > So, I had to move back to the original problem: the real serial port. Has anyone suggestions or
> > tests/patches to try to solve it? I really need that serial port to work, and I think it's impossible
< > that no one can use it on uClinux.

> Let's assume the serial port really is causing excessive interrupts. Why would that be? Since this
> port has full modem control, are the control signals in a state that would result in an interrupt?
>
> Have you tried actually controlling all of the modem signals? What about disabling those interrupts in the driver?
Sorry for my late reply, but I've been sick for over a week. As of the LPC2478's
manual, UART #1 can't be used on 32-bit databus boards (which is my case), so
I don't actually have a full modem port available. The only available ports are
UART #0 (working fine as serial console with uClinux) and UART #2 (rt/tx only).
I _think_ there's no actual interrupt triggering, as the port is disconnected and no
pins, with the exclusion of tx and rx, are enabled by the jumpers on the board.

I didn't try disabling interrupts, but I will try that right now.

Andrea

Start a New Thread

I've found the solution, as the port was enabled and it wasn't a driver issue.
I just needed to enable the (by default) disabled UART #2 and UART #3 in the
linux kernel configuration, by enabling the following config directives:

CONFIG_MACH_LPC32XX_UART2_ENABLE

CONFIG_MACH_LPC32XX_UART3_ENABLE

This happens to be true also on LPC3250, for the interested ones.

A new problem arises now. I'm observing weird behaviour with a serial analyzer.
I am communicating with odd parity, but the parity bit is never set, rendering invalid
the traffic between the LPC2478 and the connected hardware. This doesn't seem
a code issue, because the same C program runs fine on a Linux/x86 PC, while it
fails on the LPC2478/uClinux system.

What may be happening?

Regards,
Andrea

________________________________
From: Andrea Barberio
To: l...
Sent: Mon, June 7, 2010 9:32:08 AM
Subject: Re: [lpc2000] Re: Serial driver

> From: rtstofer

> > So, I had to move back to the original problem: the real serial port. Has anyone suggestions or
> > tests/patches to try to solve it? I really need that serial port to work, and I think it's impossible
< > that no one can use it on uClinux.

> Let's assume the serial port really is causing excessive interrupts. Why would that be? Since this
> port has full modem control, are the control signals in a state that would result in an interrupt?
>
> Have you tried actually controlling all of the modem signals? What about disabling those interrupts in the driver?

Sorry for my late reply, but I've been sick for over a week. As of the LPC2478's
manual, UART #1 can't be used on 32-bit databus boards (which is my case), so
I don't actually have a full modem port available. The only available ports are
UART #0 (working fine as serial console with uClinux) and UART #2 (rt/tx only).
I _think_ there's no actual interrupt triggering, as the port is disconnected and no
pins, with the exclusion of tx and rx, are enabled by the jumpers on the board.

I didn't try disabling interrupts, but I will try that right now.

Andrea Start a New Thread