[EMAIL PROTECTED] wrote on Tue, 24 Apr 2007 01:23 -0700:
> Following patch adds the sysfs magic.
> Can you test the following patch and see if it helps? It should do the
> trick although I havent been able to test, which actually brings me to
> the next question.
> Is HEAD broken from a build perspective? I can't get it to build at
> all. I havent spent too much time trying to fix it though :(
> It seems to be related to the recent merge of walt3 to head.
[..]
> Index: src/kernel/linux-2.6/devpvfs2-req.c
[..]
+ class_device_create(pvfs2_dev_class, NULL,
+ MKDEV(pvfs2_dev_major, 0), NULL,
+ PVFS2_REQDEVICE_NAME);
+ ret = devfs_mk_cdev(MKDEV(pvfs2_dev_major, 0),
+ S_IFCHR|S_IRUSR|S_IWUSR,
+ PVFS2_REQDEVICE_NAME);
Looks good in spirit, but the kernel has already moved on. For
versions >= 2.6.19 we're apparently supposed to use device_create(),
which takes more or less the same arguments. But I don't understand
the difference; only what you have does not auto-make the chardev
through udev (in a different project). The udev/sysfs characters
seem to rewrite the entire API every couple of versions. And devfs
has long been dead.
As much as I think this is the right thing to do, I think we'll have
such a huge pile of compatibility code that it is best not to add
sysfs support. Plus we'll always need the userspace code to support
non-udev installations. If someone does do the massive rewrite to
make the kernel module acceptable to upstream, then we can put in
the sysfs API du jour.
I've not tried HEAD myself since 8 mar. But should do a pull into
the OSD tree to stay current. Only the state machine fixes generate
many conflicts, so I've been waiting for some "quiet time" to get
around to it. :)
-- Pete
_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
