System stuck during boot for ~5 minutes hanging at random

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

Hi,

My system is stuck during boot for quite some time due to random fast init and random crng init hanging:

fedeliallalinea · Posted: Sun Apr 05, 2020 4:06 pm Post subject:

You tried also some of others proposed by toralf (I use haveged)?
_________________
Questions are guaranteed in life; Answers aren't.

mike155 · Posted: Sun Apr 05, 2020 4:37 pm Post subject:

Please install haveged AND enable it.

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

Hi,

please see my inital post. I installed and enabled haveged. Did not do the trick.

mike155 · Posted: Mon Apr 06, 2020 12:36 pm Post subject:

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

1. done - no change

2.

https://pastebin.com/LuFrWWsW

3.

https://pastebin.com/VjpY6fWV

4.

http://dpaste.com/1QT2PES

mike155 · Posted: Mon Apr 06, 2020 3:54 pm Post subject:

dmesg shows 2 issues, which are most probably independent of one another:

it takes 150 seconds to initialize the random number generator:

Code:

[ 149.762497] random: crng init done
it takes 430 seconds to initialize the Adaptec SmartRAID driver:

Code:

[ 434.259870] smartpqi 0000:0c:00.0: Online Firmware Activation enabled

The second issue is the one that hurts you. Let's try to solve that first.

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

hmm....you are probably right. At first I thought that the Adaptec HBA is the issue but concentrated on random (which also takes far too long

)

In the meanwhile I updated the HBA's firmware to the latest one but that did not change anything in boot behaviour. Not sure what else I can try.

mike155 · Posted: Mon Apr 06, 2020 4:29 pm Post subject:

Some of your kernel option look suspicious. AMD ACPI2Platform is disabled, NUMA is disabled, IOMMU is disabled, etc.

Please read https://wiki.gentoo.org/wiki/Ryzen and adjust your kernel settings. It might also be necessary to increase CONFIG_NR_CPUS to 32.

EDIT (not related to your boot issues): don't use -O3 in your CFLAGS! -O3 is known to break several packages and it's not recommended.

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

are these options also necessary as this is a virtual machine? The Adaptec HBA is passed through to the VM as additional info.

mike155 · Posted: Mon Apr 06, 2020 4:52 pm Post subject:

You're talking about a virtual machine? Really?

Your output of dmesg doesn't look like a virtual machine! 16 CPUs, multiple ethernet adapters, many scsi devices, ... What are you doing?

I'm afraid I can't help you.

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

Yes I am.

This is quite a special setup. I have multiple VMs runnig on that host ( 16C/32T Epyc2 CPU, 128GB RAM). I am passing through 2 NVMe disks and the Adaptec HBA - no more specialties than that.

Neither of that can explain my issues.

toralf · Posted: Mon Apr 06, 2020 7:06 pm Post subject:

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

@toralf: The Hypervisor used here is esxi 6.7. As far as I know VMWare does not pass through a hardware number generator.

toralf · Posted: Tue Apr 07, 2020 7:17 am Post subject:

chrisk2305 · Tux's lil' helper Joined: 05 Sep 2007 Posts: 110

All the threads I find on the internet suggest to install haveged and/ or rng-tools. I also tried "Trust CPU for rng" Kernel option but the issue persists. I am stuck here.