Heads up for 2.6.36 and 256.44

[siddly]

/usr/src/NVIDIA-Linux-x86_64-256.35/kernel/nv.c:426:5: error: unknown field
‘ioctl’ specified in initializer
/usr/src/NVIDIA-Linux-x86_64-256.35/kernel/nv.c:426:5: warning: initializati
on from incompatible pointer type
make[4]: *** [/usr/src/NVIDIA-Linux-x86_64-256.35/kernel/nv.o] Error 1
make[3]: *** [_module_/usr/src/NVIDIA-Linux-x86_64-256.35/kernel] Error 2
make[2]: *** [sub-make] Error 2
NVIDIA: left KBUILD.
nvidia.ko failed to build!
make[1]: *** [module] Error 1
make: *** [module] Error 2
-> Error.
ERROR: Unable to build the NVIDIA kernel module.
ERROR: Installation has failed.

Fixed with the following patch, I thought the "#if defined(HAVE_UNLOCKED_IOCTL)" would have taken care of it.
# diff -u NVIDIA-Linux-x86_64-256.44/kernel/nv.c nv.c
--- NVIDIA-Linux-x86_64-256.44/kernel/nv.c 2010-08-23 12:41:11.000000000 +0100
+++ nv.c 2010-08-23 15:38:58.042981847 +0100
@@ -423,7 +423,7 @@
static struct file_operations nv_fops = {
.owner = THIS_MODULE,
.poll = nv_kern_poll,
- .ioctl = nv_kern_ioctl,
+ .unlocked_ioctl = nv_kern_unlocked_ioctl,
#if defined(HAVE_UNLOCKED_IOCTL)
.unlocked_ioctl = nv_kern_unlocked_ioctl,
#endif

[zander]

Thank you for reporting this problem. I saw one earlier report of this and have made changes in the driver for future releases (likely not 256.xx initially, though). Note that the driver already plugs in an unlocked_ioctl handler.

[hvengel]

OK so what version of the driver should we be using with 2.6.36? Or are we stuck with hand patching the driver to get it to work?

[JackieBrown]

Originally Posted by hvengel

OK so what version of the driver should we be using with 2.6.36? Or are we stuck with hand patching the driver to get it to work?

When was 2.6.36 released?

[gonzo_s]

It is NOT released yet. Do not worry, people, NVIDIA is always in sync with kernel releases.

[siddly]

Originally Posted by JackieBrown

When was 2.6.36 released?

We are still at vanilla 2.6.36-rc2 and I'm running 2.6.36-rc2-git1 with driver 256.44, so 2.6.36 final is still some way off.
There are often API changes that effect NVidia drivers so it's good to note potential problems ahead of the final kernel release so that they can be fixed early or at least the developers are aware of what's coming in the next stable kernel.
So if you are using up to 2.6.35 there isn't a problem. I test and run vanilla kernels, NVidia Beta drivers and the latest VirtualBox to provide feedback to the developers. In my time I've found some humdingers, some where I was the only person to hit the problem. never had a data corruption or data loss.
Developers don't have all the hardware/software combinations, so testing and feedback stops the mass of users from suddenly hitting the buffer stop when the stable kernel hits the streets.

For backwards compatibility I had a problem building the module with 2.6.33.7-rt29 which I may have a look at again later. I didn't look closely at the error with 256.35 or 256.44, I ran with the distro's nv_drv.so module as I wanted to run some jack realtime audio tests with some SDR (Software Defined Radio) hardware and software.