Hi Xiaofan,
Based on your most recent information, I am pretty sure this a different
problem that my patch won't fix. I have a sneaking suspicion that you are
running into some issue with libusb. If you don't have access to a DSO, do
you have at least a DMM?
If you do, I think there is some value writing a simple test routine in
jlink_init after the USB handle is available that does the following:
jlink_jtag_handle = jlink_usb_open();
if (jlink_jtag_handle == 0)
{
blah...blah...blah...
}
/* Begin new test stuff here */
while (1)
{
jlink_simple_command(EMU_CMD_HW_RESET0);
sleep(1);
jlink_simple_command(EMU_CMD_HW_RESET1);
sleep(1);
}
You should see the reset line transition on your target once a second from
high to low on your DMM without skipping a beat. If it does this likely
will help you rule out libusb.
As for using USBMon, I can't disagree that it is hard to use and is
deficient compared to a true hardware protocol analyzer. I wouldn't waste
my time with that.
Gary
Gary Carlson
Gary Carlson, MSEE
Principal Engineer
Carlson-Minot Inc.
On 7/5/09 6:41 PM, "Xiaofan Chen" <[email protected]> wrote:
> On Mon, Jul 6, 2009 at 9:18 AM, Gary Carlson<[email protected]>
> wrote:
>> Hi Xiaofan,
>>
>> The session transcript that you provided below seems to indicate a different
>> problem then my patch would fix in your case. Was this for the Arch Linux
>> or all host systems? One of your earlier messages indicated success, so
>> knowing that helps.
>
> Your patch works for Ubuntu 9.04 (without the patch it works as well but
> sometimes got initialization error). I have not tested Fedora 10 but I believe
> it would work. It failed under Arch Linux and OpenSuse 11.1 (without
> your patches it did not work either. Your patch at lease helped to remove
> the initialization error).
>
> Take note all these 4 Linux distros are installed on the same computer.
> I also use the same USB port to connect the J-Link.
>
>> Judging from your log, you seem to be getting outright write errors out of
>> the USB port to the dongle.
>>
>> I obviously don't have your dongle, target, and computer in front of me or
>> know your technical background, but here are some starting suggestions of
>> things I would try to narrow the problem:
>>
>> 1) If you have a DSO with a differential probe you may want to look at the
>> data signals emitted onto the USB link. Are they reasonably clean at the
>> dongle?
>>
>> 2) As part of the previous suggestion, if you write a simple test loop and
>> emit one test command to the dongle at each startup, do you see signal
>> transitions at all or does the port seem to be dead during the failures? If
>> you send one command at a time and everything works normally every time,
>> what happens if you modify the test loop to send the same command in rapid
>> succession instead? Does the link then break?
>
> I was doing that at home so no DSOs. At work, I have DSOs but probably
> I would not want to do this type of things at work.
>
>> 3) If you have a good USB sniffer, have you done a closer inspection of what
>> is actually crossing the link? The simple test loop helps here so that you
>> are not overwhelmed by data.
>
> Maybe USBmon under Linux but it is a bit difficult to understand the log.
> On the other hand, I will try some old versions of OpenOCD. I remember
> it was working last time.
>
> On the other hand, Segger's Linux utility seems to work fine.
>
>> 4) Verify that your board is actually being reset during initialization by
>> monitoring the reset line into the processor using the DSO or reset LED (if
>> your board has one). Over 20 years of working on multitudes of different
>> processors and emulators I can't think of all the times that I nearly pulled
>> out my hair tracing odd problems only to find out that the board was not
>> reset like it should have been. If your link is working at all, the very
>> first thing you should see is the board tossed into reset by the dongle if
>> the first few USB writes are interpreted correctly.
>>
>> 5) What version of libusb do you have on your system(s)? If you are using
>> an older version, you may be running into bugs that have since been
>> corrected in the USB support library.
>
> I think Arch Linux and OpenSuse 11.1 have the latest version of libusb.
> Actually I normally run the latest version of libusb 0.1 and libusb 1.0
> (svn version) under my main Linux distro which is Ubuntu 9.04.
>
>> 6) What hardware version of USB does your host system have (USB 1.0
>> or USB 2.0)?
>
> It is a USB 2.0 high speed port. As I mentioned before, I
> am using the same computer (An Acer M1641, GeForce 7050 / nForce 620i
> ICP chipset, Intel Core 2 Duo E4700), same USB port.
>
>> Anyway some things to think about.
> Thanks for the hints.
_______________________________________________
Openocd-development mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/openocd-development
