Using the latest MAKEDEV-3.23-7.i386 I get the following output when booting: Starting udev: don't know how to make device "loop1" don't know how to make device "loop2" don't know how to make device "loop3" don't know how to make device "loop4" don't know how to make device "loop5" don't know how to make device "loop6" don't know how to make device "loop7" don't know how to make device "lp1" don't know how to make device "lp2" don't know how to make device "lp3" The machine continues to boot fine though.
Using nvidia driver(livina rpm), I get following messages, also. don't know how to make device "nvidia1" don't know how to make device "nvidia2" don't know how to make device "nvidia3" $ cat /etc/udev/makedev.d/60-nvidia.nodes nvidia0 nvidia1 nvidia2 nvidia3 nvidiactl
Looks like the recent patch breaks matching between device minor number and "%d" token in device name when parsing configuration files in "/etc/makedev.d" directory. Before, "%d" was matched against the minor number. Now it's matched against 0 only. Assume we have the following configuration: c $CONSOLE 195 0 1 255 nvidia%d Before the version 3.23-7 this was interpreted the following way: for each N in range 0..254 device "nvidiaN" has major 195 and minor N. In version 3.23-7 this is interpreted the different way: for each N in range 0..254 device "nvidia0" has major 195 and minor N. This becomes evident when invoking "MAKEDEV -v -x nvidia0": it creates "/dev/nvidia0" 255 times with minor numers from 0 to 254 in sequence. As the end result, we have "/dev/nvidia0" with minor 254. And when invoking "MAKEDEV -v -x nvidia1", it complains that "nvidia1" is not found. As a temporary workaround, I've downgraded MAKEDEV package to the previous version 3.23-6 until the problem with "%d" matching is fixed.
(In reply to comment #0) > Using the latest MAKEDEV-3.23-7.i386 I get the following output when booting: > > Starting udev: don't know how to make device "loop1" > don't know how to make device "loop2" > don't know how to make device "loop3" > don't know how to make device "loop4" > don't know how to make device "loop5" > don't know how to make device "loop6" > don't know how to make device "loop7" > don't know how to make device "lp1" > don't know how to make device "lp2" > don't know how to make device "lp3" > > The machine continues to boot fine though. Same result
looks to be the same as BZ 466560
*** Bug 466560 has been marked as a duplicate of this bug. ***
I am seeing the same w/ current (2008-10-21) rawhide (udev-127-2.fc10.i386, MAKEDEV-3.23-7.i386)
This bug still lingers after several rawhide updates. If it isn't gonna get fixed, how about someone just take the source for 3.23-6, call it 3.23-8, and rebuild it so it is safe to update again without paying special attention to avoid getting the utterly busted 3.23-7 version?
Created attachment 321125 [details] Difference between the patch that got actually applied and the latest version of the patch The problem is that what got applied to MAKEDEV git is apparently not the latest patch I've posted. I'm attaching interdiff between what was applied and the latest version. Among some speedups (e.g. using obstack) or cleanups, it contains a fix for this bug (the part where num2++ is used if entry->modulo == 0 instead of incrementing num1) and fixes for the configuration files.
This should be fixed in MAKEDEV-3.24-1. Thanks for the patch, Jakub.
Yep, works for me today. No error messages and plenty of /dev/loop device nodes.
The newest 3.24 version is also working fine for me.