Syslinux - Feb 2018 - problem booting from virtual disk, looking for pointers to the code

Hi,

I'm booting an qemu VM from a virtual disk.  In the success case the virtual
console shows something like this:

SYSLINUX 4.05 0x54f93f16 Copyright (C) 1994-2011 H. Peter Anvin et al
Loading vmlinuz.........
Loading initramfs.img...................................ready.
[ 0.000000] Initializing cgroup subsys cpuset
<etc>


In the failure case I'm seeing something like this:

SYSLINUX 4.05 0x54f93f16 Copyright (C) 1994-2011 H. Peter Anvin et al
Loading vmlinuz.........
Loading initramfs.img..........


and then it takes up to ~100 seconds for the VM to start running.  I suspect 
most of that time was spent trying to load this file.

My assumption is that this is likely an IO problem on the host. Would that make 
sense?

Also, can someone point me to the code that emits the dots while loading the
file?

Thanks,
Chris

On Thu, Feb 22, 2018 at 11:23 AM, Chris Friesen via Syslinux
<syslinux at zytor.com> wrote:
> Hi,
>
> I'm booting an qemu VM from a virtual disk.  In the success case the
virtual
> console shows something like this:
>
> SYSLINUX 4.05 0x54f93f16 Copyright (C) 1994-2011 H. Peter Anvin et al
> Loading vmlinuz.........
> Loading initramfs.img...................................ready.
> [ 0.000000] Initializing cgroup subsys cpuset
> <etc>
>
>
> In the failure case I'm seeing something like this:
>
> SYSLINUX 4.05 0x54f93f16 Copyright (C) 1994-2011 H. Peter Anvin et al
> Loading vmlinuz.........
> Loading initramfs.img..........
>
>
> and then it takes up to ~100 seconds for the VM to start running.  I
suspect
> most of that time was spent trying to load this file.
>
> My assumption is that this is likely an IO problem on the host. Would that
> make sense?
It's most likely although CPU contention can cause similar symptoms.
An 8vCPU VM needs 8 idle cores to have execution time.  How many vCPUs
(vCPU = vSockets * vCoresPerSocket)?  Physical cores (not counting CPU
threading)?
> Also, can someone point me to the code that emits the dots while loading
the
> file?
core/runkernel.inc calls dot_pause in core/abort.inc

-- 
-Gene

On 02/22/2018 11:23 AM, Gene Cumm wrote:
> On Thu, Feb 22, 2018 at 11:23 AM, Chris Friesen via Syslinux
> <syslinux at zytor.com> wrote:
>> In the failure case I'm seeing something like this:
>>
>> SYSLINUX 4.05 0x54f93f16 Copyright (C) 1994-2011 H. Peter Anvin et al
>> Loading vmlinuz.........
>> Loading initramfs.img..........
>>
>>
>> and then it takes up to ~100 seconds for the VM to start running.  I
suspect
>> most of that time was spent trying to load this file.
>>
>> My assumption is that this is likely an IO problem on the host. Would
that
>> make sense?
>
> It's most likely although CPU contention can cause similar symptoms.
> An 8vCPU VM needs 8 idle cores to have execution time.  How many vCPUs
> (vCPU = vSockets * vCoresPerSocket)?  Physical cores (not counting CPU
> threading)?
At the time of the problem there are 10 VMs running (including the problematic 
one) with one vCPU per VM.  There are 10 physical cores available.  We had just 
recovered the host after maintenance, so there were a log of VMs being restarted
around the same time.
>> Also, can someone point me to the code that emits the dots while
loading the
>> file?
>
> core/runkernel.inc calls dot_pause in core/abort.inc
Found it.  runkernel.inc passes the address of dot_pause to pm_load_high() which
actually loads the file and calls dot_pause().

Chris