Red Hat Bugzilla – Bug 63928
duplicate microcode device nodes in /dev/cpu/
Last modified: 2007-11-30 17:10:30 EST
It seems wrong to have multiple aliases (/dev/cpu/0..15/microcode) for the same
device node (major 10, minor 184). Why is it done this way? Would it be better
to have unique microcode device nodes for each processor instead of just using
/dev/cpu/microcode? Is this something I should change in the official devices.txt?
(wearing my LANANA hat)
I couldn't hazard a guess. The init script which calls microcode_ctl
on my system is in kernel-utils, maybe there's some idea there.
I've never figured out why we deviated away from upstream here.
Our microcode_ctl has a patch to change reading the file from /dev/cpu/microcode
to /dev/cpu/0/microcode. It makes no sense to me as
a) We only do it for cpu #0
b) The driver patches every CPU in the system, so multiple nodes don't make sense.
We could drop that patch and fix up udev in FC4 / RHEL4.
as of tomorrow, microcode_ctl in rawhide will use the upstream /dev/cpu/microcode.
Harald, please adjust udev accordingly.
If udev-050-1 was the 'adjusted' udev for this bug, then the commented
line in /etc/makedev.d/linux-2.6.x:
#c $ROOT 10 184 1 1 cpu/microcode
should be uncommented.
Likewise, the uncommented line in /etc/makedev.d/redhat:
c $ROOT 10 184 0 16 cpu/%d/microcode
should be commented.
*** Bug 144915 has been marked as a duplicate of this bug. ***
*** Bug 144887 has been marked as a duplicate of this bug. ***
*** Bug 147657 has been marked as a duplicate of this bug. ***
*** Bug 145861 has been marked as a duplicate of this bug. ***
*** Bug 146292 has been marked as a duplicate of this bug. ***
This fix is not in RHEL4-U1-beta.
microcode_ctl has been changed to use /dev/cpu/microcode,
but shipped MAKEDEV-3.15-2 doesn't contain the corresponding fix in Comment#4.
Can you get this fixed in RHEL4-U1?