Interrupt handler API changed in 2.6.19

[Altera_Forum]

If you are developing kernel drivers, 

please note, the interrupt handler API changed in 2.6.19 . Read this, http://lwn.net/articles/202449/ (http://lwn.net/articles/202449/)  I am trying uClinux-dist-test-20070108. We will need to come up with the new interrupt API.

[Altera_Forum]

Have anyone tried to get 2.6.19.x to work on Nios2 yet ? 

I&#39;ve done a port which is mostly OK, but I have experienced a couple of issues not present in 2.6.18: * /proc/<pid>/cmdline is empty for all processes. This could possibly be releated to the new kernel_execve call (used to start init). * Top (from busybox) is using up to 100% CPU. Probably some /proc problems (top monitors /proc/stat).

[Altera_Forum]

I am working on uClinux-dist-test-20070108, which uses 2.6.19 . I haven&#39;t got through yet. 

Could you attach your patches on the wiki, http://nioswiki.jot.com/wikihome/operating...s/kernelpatches (http://nioswiki.jot.com/wikihome/operatingsystems/kernelpatches)

[Altera_Forum]

Sure. I&#39;ll see if I can generate a diff with my changes from 2.6.18 to 2.6.19.

[Altera_Forum]

The patch is uploaded to: 

http://nioswiki.jot.com/wikihome/operating...9-nios2.diff.gz (http://nioswiki.jot.com/wikihome/operatingsystems/%c2%b5clinux/kernelpatches/linux-2.6.18_to_2.6.19-nios2.diff.gz)  It&#39;s generated against my own sources, so I can not guarantee that it will apply cleanly against 2.6.19 + the nios2-patch (2.6.18) on the wiki, or that all required changes are in there. But at least you&#39;ll get an idea about what changes that have to be done.

[Altera_Forum]

Thank you very much. 

It would be very helpful. http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/smile.gif  Hippo

[Altera_Forum]

Hi atlenisse, 

 I&#39;ve adapted your patches, and got the same issues of /proc//cmdline, environ... But /proc//status, stat are correct. I wonder if there are problems on task->mm. Do you have any found?  Hippo

[Altera_Forum]

Hello Hippo, 

 I&#39;m sorry but I haven&#39;t had the time to look into this since I initially did the port (before X-mas). I chickened out and rolled back to 2.6.18...  One thing I noticed is that the proc-behaviour of processes started from init/sh is exactly as for kernel-processes (started by kthread_run, e.x kjournald), so I suspected that this issue could be related to the new kernel_execve call (starting with v2.6.19 init is started with kernel_execve() instead of using the execve syscall). But if it is so, this issue would (possibly) also apply to other architectures, and I have not seen any bug reports anywhere.

[Altera_Forum]

I traced proc fs. 

It seems to be problem of access_process_vm() in linux-2.6.x/mm/nommu.c . several nommu stuff updated in 2.6.19 .

[Altera_Forum]

The empty /proc/pid/cmdline issue was resolved with a patch, 

http://mailman.uclinux.org/pipermail/uclin...ary/041734.html (http://mailman.uclinux.org/pipermail/uclinux-dev/2007-january/041734.html)

[Altera_Forum]

Nice! This fixed the cmdline-issue, but top is still acting weird. I&#39;ll post a question about this on the uClinux list.