[systemd-devel] systemd-213: regression with zram

[Alexander E. Patrakov]

[maintainers of zram CC:ed, in order to alert them of this 
incompatibility with a setup published by Gentoo and possibly picked up 
by others]

[Pacho Ramos CC:ed as a person who asked about zram here before and who 
can remove the non-working instruction listed at 
http://wiki.gentoo.org/wiki/Zram#Using_existing_tools ]

I wrote:

> I have upgraded systemd from 212 to 213 on two my Gentoo boxes, and see the same regression here: zram swap space does not get activated. It looks like systemd tries to activate swap before the RUN+=mkswap part of the udev rule finishes.
>
> Here are the relevant lines from my configuration files. Are they indeed supposed to work, or were working only by pure luck?
>
> $ cat /etc/modules-load.d/zram.conf
> zram
>
> $ cat /etc/modprobe.d/zram.conf
> options zram num_devices=4
>
> $ cat /etc/udev/rules.d/01-zram.rules
> KERNEL=="zram[0-9]", SUBSYSTEM=="block", DRIVER=="", ACTION=="add", ATTR{disksize}=="0", ATTR{disksize}="1536M", RUN+="/sbin/mkswap $env{DEVNAME}"
>
> $ grep zram /etc/fstab
> /dev/zram0 none            swap            pri=10,discard  0 0
> /dev/zram1 none            swap            pri=10,discard  0 0
> /dev/zram2 none            swap            pri=10,discard  0 0
> /dev/zram3 none            swap            pri=10,discard  0 0


09.06.2014 23:44, Kai Krakow wrote:
> I didn't mean to persuade you but zram looks a little bit broken to me with
> respect to configuration. So if zswap would be an option for you, it might
> be the way to go instead of trying to work around quirks that cannot be
> fixed easily. Of course, it is only an option if you also use a physical
> swap device.

I don't use a physical swap device.

But anyway, the issue is now fully diagnosed, see below. It is indeed 
zram-specific, and migrating to an alternative solution (such as 
physical swap) would not hide any bug in systemd.

The systemd commit that introduces the regression to the above setup is:

commit 3ebdb81ef088afd3b4c72b516beb5610f8c93a0d
Author: Kay Sievers <kay at vrfy.org>
Date:   Sun Apr 13 19:54:27 2014 -0700

     udev: serialize/synchronize block device event handling with file locks

As an effect of this commit, systemd-udevd holds an open file descriptor 
to the block device while processing rules for it. Thus, it writes to 
the "disksize" sysfs attribute, while also having the fd to /dev/zram0 
open. I could not trace by reading kernel sources why this fails to 
change the disk size, but this failure somehow feels logical. I should 
just stop setting the "disksize" attribute from an udev rule.

-- 
Alexander E. Patrakov