Arch Linux

luser

Member

Registered: 2016-08-27

Posts: 115

pstore puzzler

At this point, my question is essentially academic. I am not aware of any actual problem, but I don't understand why I am getting different results with different machines.

When one of my machines boots, it produces the following log messages:

$ journalctl -b --no-pager | grep pstore
Jul 26 10:40:59 archlinux kernel: pstore: using lz4hc compression
Jul 26 10:40:59 archlinux kernel: pstore: No lz4hc compression
Jul 26 10:40:59 archlinux kernel: pstore: Registered erst as persistent store backend

My other machines only show:

$ journalctl -b --no-pager | grep pstore
Jul 26 10:40:59 archlinux kernel: pstore: using lz4hc compression

/proc/config.gz seems to be the same on all machines:

$ zgrep PSTORE /proc/config.gz
CONFIG_CHROMEOS_PSTORE=m
CONFIG_PSTORE=y
CONFIG_PSTORE_DEFLATE_COMPRESS=m
CONFIG_PSTORE_LZO_COMPRESS=m
CONFIG_PSTORE_LZ4_COMPRESS=m
CONFIG_PSTORE_LZ4HC_COMPRESS=m
# CONFIG_PSTORE_842_COMPRESS is not set
CONFIG_PSTORE_COMPRESS=y
# CONFIG_PSTORE_DEFLATE_COMPRESS_DEFAULT is not set
# CONFIG_PSTORE_LZO_COMPRESS_DEFAULT is not set
# CONFIG_PSTORE_LZ4_COMPRESS_DEFAULT is not set
CONFIG_PSTORE_LZ4HC_COMPRESS_DEFAULT=y
CONFIG_PSTORE_COMPRESS_DEFAULT="lz4hc"
# CONFIG_PSTORE_CONSOLE is not set
# CONFIG_PSTORE_PMSG is not set
# CONFIG_PSTORE_FTRACE is not set
CONFIG_PSTORE_RAM=y

lz4 is installed on all machines. So, why do I get:

Jul 26 10:40:59 archlinux kernel: pstore: No lz4hc compression
Jul 26 10:40:59 archlinux kernel: pstore: Registered erst as persistent store backend

only on one machine? It's the newest of the three, so perhaps this is due to some hardware difference, but I'm asking about it because I don't understand why there's a difference and whether I have misconfigured something on the machine that won't use lz4hc for the pstore.

---

luser: an epithet applied by Windows users to linux users

loqs

Member

Registered: 2014-03-06

Posts: 19,113

Re: pstore puzzler

Is the lz4hc module in both systems initrd's?

luser

Member

Registered: 2016-08-27

Posts: 115

Re: pstore puzzler

Hi, thanks for responding. Hopefully, this addresses your question. Both machine's initrd's contain the shared library liblz4.so.1:

$ lsinitcpio /boot/initramfs-linux.img | grep lz4

usr/lib/liblz4.so.1

The machine that says there is no lz4hc compression does have a slightly different mkinitcpio.conf, because it refers to i915.conf:

/etc/modprobe.d/i915.conf

options i915 enable_guc=3

Both mkinitcpio files have default compression options.
Of course, if you wish, I can post them here.

---

luser: an epithet applied by Windows users to linux users

loqs

Member

Registered: 2014-03-06

Posts: 19,113

Re: pstore puzzler

I was wondering about the kernel module lz4hc.ko.xz being present in one initrd but not the other.  However is seems it was not present in either initrd so that is not the cause.

luser

Member

Registered: 2016-08-27

Posts: 115

Re: pstore puzzler

One difference between the machines is that the EFI partition is on a separate disk in the case where the system can't find the lz4hc module. Since lsinitcpio shows it's not embedded in the initramfs, I suppose it is instead being accessed from the root partition in:
/lib/modules/4.17.10-1-ARCH/kernel/crypto/lz4hc.ko.xz

My /boot partition is on an SSD, while root is on a spinning disk. I'm wondering could there be a latency issue causing this, or perhaps it's a configuration issue? What tests or inspections should I run to establish the cause?

---

luser: an epithet applied by Windows users to linux users