cc'ing the pcmcia maintainer and the author of
the patch for interrupt routing for TI bridges.
concerning a bug report about non working irq routing since 2.6.6-rc1
up until 2.6.12-rc1-mm3
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=270376
(lspci, dmesg)
hope you can help to resolve that issue.
i volountary test build kernels for bug reporter with proposed patches.
On Thu, 31 Mar 2005, Jefferson Cowart wrote:
(For reference my primary card is a 3C574. My secondary is a 3C589.)
..
========
(I put that card into a Windows XP laptop and it worked fine there so the
card itself is fine.)
Now I'm trying to see if I have these same problems on 2.5.5
First, Booting with the secondary card plugged in had the same failure as
above.
Next booting with the primary plugged in. It worked (as expected). Then I
was able to pull and re-insert the card without problems. However pulling
the primary and inserting the secondary killed the network (as expected).
Now I'm going back to 2.4.27 to figure out if my secondary card will work
there. It failed. I got a hard lock when I inserted the card. (Reboot button
needed.)
I'm 99.9% certain that my testing of 2.6.6-2.6.10 was using my primary nic,
but I'm booting 2.6.10 now to be sure. (I know the kernels I tested for you
yesterday used my primary nic.) As expected no networking.
It looks like we may have 2 different bugs here: one in the yenta_socket
driver starting somewhere after 2.6.5-bk1 and before 2.6.6-rc1 and one in
the 3c589 driver in I don't know what versions. (I seem to remember that
2.4.18 worked, but I'm not sure. If you want me to try that nic under and
older kernel let me know, however I think fixing the first bug is more
important.) If you need any further log info let me know.
indeed there is a change in the irq handling in the yenta code.
it went in after 2.6.5-bk2 and is in 2.6.6-rc1.
in the context could you send please the ouput of a pre
2.6.6-rc1 kernel on your box:
cat /proc/interrupts