you could try mdadm_udev hook instead, the new package just replaced the udev hook with the old mdassemble hook.

I can confirm this bug. I have to manually boot my system now.
(eg mdadm --assemble --scan and ln -sf /dev/md<nr> /dev/disk/by-label/<label> )

Does the udev hook fix the problem? I didn't change anything. mdadm hook uses static mdassemble, mdadm_udev uses mdadm with udev.

The mdadm_udev hook does not exist.
"udev" is present in my HOOKS list.
Running mdassemble by hand in the shell does not reassemble the devices.
mdadm --assemble --scan does BUT does not create /dev/disk/by-label/* links
If you need more info, please ask.

mdadm /lib/initcpio/install/mdadm_udev
pacman -Qo /lib/initcpio/install/mdadm_udev

What I just don't like is that upgrading my system renders it unbootable.
Can this be fixed automatically?
I will try this fix tomorrow (currently not at home, rebooting this system remotely is scary)

Sure I have no clue why your systems don't work anymore,
The udev way also made systems unbootable, so I restored the old way of assembling.
Could you give me more input on what raid are you using, mdadm.conf and mkinitcpio.conf?

My mdadm.conf contains:
DEVICE partitions
MAILADDR <emailaddr>

My mkinitcpio.conf contains:
MODULES="raid1 raid456"
BINARIES="/sbin/mdadm /sbin/mdassemble"
FILES=""
HOOKS="base udev autodetect pata scsi sata filesystems usbinput mdadm"
COMPRESSION="xz"

I am using rai1 & raid6:

[jaap@server ~]$ cat /proc/mdstat
Personalities : [raid1] [raid6] [raid5] [raid4]
md124 : active raid1 sda1[0] sdf1[4] sdd1[3] sdc1[2] sdb1[1]
48756 blocks super 1.2 [6/5] [UUUUU_]

md125 : active raid6 sda2[0] sdf2[6] sdd2[3] sdc2[2] sdb2[1]
7995392 blocks super 1.2 level 6, 512k chunk, algorithm 2 [6/5] [UUUUU_]

md126 : active raid6 sda3[0] sdf3[6] sdd3[3] sdc3[2] sdb3[1]
280670208 blocks super 1.2 level 6, 512k chunk, algorithm 2 [6/5] [UUUUU_]

md127 : active raid6 sda4[0] sdf4[6] sdd4[3] sdc4[2] sdb4[1]
7525181440 blocks super 1.2 level 6, 512k chunk, algorithm 2 [6/5] [UUUUU_]

unused devices: <none>
[jaap@server ~]$

BTW I have not seen any segfault during startup, but maybe I am not looking good enough.
It just does not assemble the MD arrays.

Ok the mdadm hook cannot work for you, because you don't define arrays in mdadm.conf.
run: mdadm -D --scan >>/etc/mdadm.conf
to add the arrays.

I have never added the arrays to mdadm.conf and it has always worked (for all my machines). Please add a option to mkinicpio which activates/restores the old behaviour....

How long do you run this setup? Have you defined the arrays on bootloader commandline?

I run this kind of setup for over 2 years....
I have not defined md boot (kernel) commandline parameters.
Maybe I should look at the diff from 3.2.2-4 to 3.2.2-3 and suggest a patch?

This is strange the change to udev happened in june this year, it makes me wonder how it was working before.
The old way to assemble md arrays was only using mdadm.conf file with arrays defined or define it on commandline.
I think the best solution for you is to change to mdadm_udev in mkinitcpio.conf.

Ok, I will try the mdadm_udev (install) option.
Maybe I will bisect it exactly later.

No need to bisect, there happened not much in the last releases.
I just divided the both assemble methods.
It doesn't make sense to assemble arrays with 2 ways in 1 hook.

The mdadm_udev hook works for me. System boots itself again.

I just upgraded my mdadm package to 3.2.2-4 and followed the above instructions for placing the mdadm_udev hook in /etc/mkinitpcio.conf. I can confirm that the solution works and I am no longer getting a Segmentation Fault error upon boot.

So can we close this now?