Udev-091

[alexander@…]

Version increment. This version includes the "udevsettle" program that is supposed to be used instead of our custom waiting loop.

[alexander@…]

Modified udev script

[alexander@…]

New udev_retry script

[alexander@…]

Modified bug reporter

[alexander@…]

Modified udev script that uses the new udevsettle program, the new udev_retry script that retries failed uevents (to be started after mountfs, i.e. at S45), and the modified bug.c program are attached. Now all USB-related leaks are ignored because they do happen and nothing can be done about them. Kernel issue.

[alexander@…]

Avtually not sure if the "mkdir -p /dev/.udev/queue" is still needed. If it is needed, the attached scripts are buggy. But they work for me, and the need for mkdir will certainly show up as a red bugreport.

[bryan@linuxfromscratch.org]

Just so this doesn't get forgotten: For udev >=089, we should add extras/path_id to the EXTRAS variable in the udev commands. It's a script in the same vein as ata_id, edd_id, scsi_id, usb_id, and vol_id.

We don't install any udev rules that use any of these programs today, but the sample 60-persistent-storage.rules file in the udev source tree uses all of them. (Not that that's the only way to do persistent device links, though.) In that file, path_id is used to create /dev/disk/by-path/ symlinks ("shortest physical device path").

[archaic@…]

Alex has been working on this on one of my systems and I'm in the process of packaging up his work into a digestable form for the list and book. We tested id capability rules and persistent rules with 4 cd/dvd devices. Once I get those unplugged, I can plug the 3 scsi disks back in and run a couple more tests.

[Matthew Burgess]

Now udev-091:

Some more keys require the correct use of '==' and '=' depending on the kind of operation beeing an assignment or a match. Rules with invalid operations are skipped and logged to syslog. Please test with udevtest if the parsing of your rules throws errors and fix possibly broken rules.

[Chris Schwemmer]

The two following lines need fixing with udev-091:

diff /etc/udev/rules.d/25-lfs.rules /sources/udev-config-6.rules
157c155
< ACTION=="add", SUBSYSTEM=="scsi" , SYSFS{type}=="0|7|14", \
---
> ACTION=="add", SUBSYSTEM="scsi" , SYSFS{type}=="0|7|14", \
159c157
< ACTION=="add", SUBSYSTEM=="scsi" , SYSFS{type}=="1", \
---
> ACTION=="add", SUBSYSTEM="scsi" , SYSFS{type}=="1", \

(note the extra '=' after SUBSYSTEM)

[alexander@…]

I never understood the purpose of those rules. They were copied from Jim's udev-cross-lfs tarball, and they got there from SuSE.

Proposal: remove them unless Jim can explain their purpose without references to SuSE (in which case they indeed need fixing, not removing).

[archaic@…]

Fixed in r7569.

[archaic@…]

Correction. The update is fixed in r7569. The comments about SUBSYSTEM= vs SUBSYSTEM== are valid, but so is Alex's comment about the need for those rules. Keeping open pending discussion.

[archaic@…]

Opened ticket #1785 to address rule file changes. Closing this ticket as fixed.

[alexander@…]

Reopening: the new bug.c file is not in the book yet. Please fix ASAP to avoid USB-related SPAM.

[archaic@…]

Fixed in r7584.