Bugzilla Bug 26452

getting the dreaded MODPROBE : Fatal : no /dev/sg etc etc when booting
up/shutting down, even though its using your be-quietwhenn-probing aliases patch
(think thats what is called)

Ok, but it is correct.  Previous patches did not behave like modutils - it
always suppressed output when called from devfsd.  New one works like modutils
in that it only suppress output if an alias exists.  Thus:

1)  If possible add all the output to this bug.

2)  Try to figure out if you really do not have an alias for /dev/sg, and
    if you do, let me know (check in /etc/modprobe.*).

I'm seeing a similar problem here with 2.6.0-test3, specifically relating to
ide devices (I have an all-scsi box, so no ide stuff compiled) at bootup, and
alsa? devices at shutdown.

At boot I see-

Calculating module dependencies...
FATAL: Module st not found
FATAL: Module ide_probe_mod not found
FATAL: Module ide_tape not found
FATAL: Module ide_probe_mod not found
FATAL: Module ide_disk not found
FATAL: Module ide_cd not found
FATAL: Module ide_tape not found
FATAL: Module ide_floppy not found

in /var/log/messages I see a bunch of other stuff too that is being suppressed
(? pardon my ignorance on this subject)

When I shut down, a bunch of "modprobe FATAL" warnings fly past, but nothing in
/var/log/messages.  They refer to /dev/snd/controlCxx and /dev/aloadCx (where x
is a number from 0-31 afaik - I have a cs46xx based card that supports hardware
mixing of 31 channels so I'm guessing that's where this number is coming from).

I've noted it in this thread
http://forums.gentoo.org/viewtopic.php?t=73649&postdays=0&postorder=asc&start=75

at least one other is having similar problems too, the IDE related ones at
least.

Created an attachment (id=15992) [details]
partial modprobe output

Making an attachment (part of my log files, there is seriously long strings of
output not being captured by my logger), also I get the long output from the
alsa mods on shutdown as noted by Jonathan, however the bootup stuff is a mixed
bag, some have aliases in /etc/modprobe* other dont (or shouldn't IIRC) I have
no idea even where to start unweaving this tangle....

Im having the same problem, usually with messages similar to Johnathan. After a
little investigation, it seems that the problem occurs during the execution of
/sbin/generate-modprobe.conf -- this script calls modutils modprobe commands
(ie /sbin/modprobe.old) which do not have the "quiet" feature applied to them.

I have two different ways to get around this, neither of which are really that
good.
1) Change the initialised value of quiet in the devfsd patch to 1, instead of
0, this will make modprobe always use quiet output, which could potentially be
bad.
2) Edit /sbin/modules-update and change the following line 
if [ -x /sbin/generate-modprobe.conf -a "${KERNEL_2_5}" = "yes" ]
to
if [ -x /sbin/generate-modprobe.conf -a "${KERNEL_2_5}" = "no" ]

Im not sure why modprobe.old is still called on the kernels > 2.5.48 but that
if statement is the script that does it. As I said above neither are really
that good.

I also get errors in reference to /dev/burner -- which is hdc on my system, and
now scsi actually gives me SCSI I/O errors when trying to run cdrecord
-scanbus, so not everything works properly still.

This is from modprobe included in modutils:

 # modprobe ide_tape
modprobe: Can't locate module ide_tape


The latest module-init-tools works much closer as modutils did, and
not like the prev versions that masked all errors.  Read the comments
to the patch ...

The problem is not with module-init-tools, but rather with the generated
/etc/modprobe.devfs, as it is just a converted /etc/modules.devfs.

To fix, we need to really have a look at /etc/modprobe.devfs as it is
now, and then convert all the invalid modules mentioned to the correct
2.6 ones.

I do not have the time currently, sorry.  Patches will be taken though :P

*** Bug 26607 has been marked as a duplicate of this bug. ***

*** Bug 26941 has been marked as a duplicate of this bug. ***

*** Bug 26571 has been marked as a duplicate of this bug. ***

Curiously enough, when delving through this problem, I notice that all the
modules
not being found (the FATAL's) 99.9% have an entry in
/etc/pam.d/console.perms...I wonder if it's possible that these are somehow
being set off indirectly? I dont even have aliases for some of the stuff that
went fatal, but with the exception of /dev/cpu and /dev/:0 ,every single other
fatal on this box can be found in console.perms, possibilities?

This problem of module-init-tools giving excessive output seems to be fixed in
version 0.9.15_pre1 in the ~x86 testing tree.  I get no modprobe errors on boot
and none on shutdown either.  Everything also looks normal in
/var/log/everything/current.  Anyone still have difficulties after uprading to
the new version?

Fixed in latest version.