udev-096

[Matthew Burgess]

Version increment.

The binary firmware helper is replaced by the usual simple shell script. Udevsend is removed from the tree.

[b3nt@…]

The firmware helper will need a small change to the relevant udev.rule. In 27-firmware.rules, firmware_helper should be changed to firmware.sh.

[Matthew Burgess]

Now version 094. From RELEASE-NOTES:

The built-in MODALIAS key and substitution is removed.

There's also updates to the "writing udev rules" docs and added support for gsf & gfs2 support.

[dbn.lists@…]

New version. From the release announcement:

​http://sourceforge.net/mailarchive/forum.php?thread_id=18680481&forum_id=3157

 udev 095
 ========
 %e is finally gone.
 
 Added support for swapping network interface names, by temporarily
 renaming the device and wait for the target name to become free.

[alexander@…]

Version increment.

Summary of changes from v095 to v096 ============================================

Kay Sievers:

Makefiles: fix .PHONY for man page target allow longer devpath values path_id: prepare for new sysfs layout

i.e. nothing important while we stay with linux-2.6.16.x, and nothing bad too.

[dnicholson@…]

Alexander, do you still want to build the bug.c event recorder? I recall a conversation about this a while back. Possibly, you just wanted to remove the big warning from the udev and udev_retry bootscripts.

Any input?

[alexander@…]

The bug.c file is (and in fact always was) useless. Remove.

[dnicholson@…]

udev-096 changes

[dnicholson@…]

Attached a diff for udev-096. A few points:

1. Using the udev-config tarball in Alexander's home directory. I'm going to merge his changes into the svn if no one objects. There are almost no changes except a small bug he uncovered with device-mapper devices.

2. Removed bug.c. The bootscript needs to be changed to reflect this.

3. The write_cd_aliases script is not installed. Another system is in the udev-config tarball in contrib/. Should this be mentioned? Bootscripts page doesn't mention it's contrib/ directory. Alexander is working on a patch for Ch. 7.4 for creating persistent cd symlinks. This will fix #1818 and provide an example of writing a custom udev rule.

4. I'm not crazy about this line. "Udev's configuration is far from ideal by default, so install the configuration files here:". Any suggestions? Do people like it the way it is?

[alexander@…]

All changes look good, but are not acceptable as a whole, because this will break BLFS expectations that /dev/cdrom exists. Please attach the tarball whith your changes to this ticket.

Note that I will have to release yet another tarball tomorrow because of for CDROM permissions.

[alexander@…]

More udev changes

[alexander@…]

I added more textual changes, to be appl;ied after udev-096.diff. They expect a new udev tarball, though, for the CD-ROM permissions to be correct. A new file is needed:

# /etc/udev/rules.d/81-cdrom.rules: Set CD-ROM permissions.

ACTION=="add", SUBSYSTEM=="block", ENV{ID_TYPE}=="cd", GROUP="cdrom"

[alexander@…]

Desired changes to udev config

[alexander@…]

The attached diff implements CDROM-related changes and a device mapper fix. Dan also wrote to me privately that he wants to do some whitespace changes.

[alexander@…]

Udev's configuration is far from ideal by default, so install the configuration files here

Why not say this:

Udev has to be configured in order to work properly, and does not come with any sort of "default" configuration. Install the LFS-specific configuration files:

[dnicholson@…]

I like the previous comment. I'll make that change.

Would you prefer if I rolled a tarball (haven't changed anything but whitespace), or just check the changes into svn? Whatever, I'm attaching the tarball and checking in the old changes.

Why does the CHANGELOG file have listings all with n/a? I would think it would be the authors name followed by the date. Also, dates should probably be in the standard format (xxxx-xx-xx). I guess I can change that later.

[dnicholson@…]

Whitespace cleaned rules without above changes from Alexander

[alexander@…]

Indeed, you are basing your work on the correct tarball. Now apply my changes (udev-config.diff), test (e.g., make sure that xterm runs from non-root well) and release.

[dnicholson@…]

Current bootscripts have removed the sleep's from after the udevsettle commands. This doesn't seem correct. I suggest changing to:

/sbin/udevsettle --timeout=6

[alexander@…]

You are not the first to make this mistake. These are completely different timeouts. Read the code.

6 seconds is the "dark" timeout: we waited 6 seconds in order to be sure that uevents don't reappear. It is only a part of the debugging hook.

The --timeout parameter controls the "white" timeout: suppose that some uevent child doesn't die in --timeout seconds (default: 180). Then, in order to avoid hanging the complete system at boot due to a single misbehaving udev rule, udevsettle decides to continue with partially-populated /dev in hope that root fixes this before the next reboot.

[dnicholson@…]

I got a "midair collision" trying to write the part below. And now I see the sleep was part of the debug, so I'll leave the bootscripts alone.

OK, udevsettle --timeout doesn't do what I expected. I thought that was time to wait after it thought the queue was empty. Do we want the sleeps? Maybe I got this wrong and the removal was intentional. I vaguely recall a conversation about how the USB events were always going to leak, so there's no way to fix that with sleep.

System is working fine. I'll run through a couple more tests, but everything seems good for committing.

[alexander@…]

Do we want the sleeps?

No. We ignore all leaks as unfixable bugs present in all distros but relevant only to those who have / on ramdisk and /usr on USB flash.

[dnicholson@…]

OK. I'm going to commit the changes to the udev page. Would you like to commit the Ch. 7 changes? I think the new text looks really good and is informative.

Later I'll roll a new udev-config tarball and get it into downloads.linuxfromscratch.org.

[alexander@…]

Please do all needed commits yourself, because you have all the pieces together (including the ability to reboot, I am currently busy with the LiveCD and thus can't test my changes outside of qemu). Note that the new text depends upon the new udev version and new udev-config tarball.

[dnicholson@…]

I'm closing this ticket. I'll commit your changes to Ch. 7 in a little while. I've gotta leave right now. Fix is in r7659.

Also, someone needs to move the udev-config tarball to downloads.linuxfromscratch.org. I don't have permissions.