Re: [Xen-devel] Xen 4.6 --with-ovmf, on UEFI Dom0 host up/running, but logging repeated "d1v0 Over-allocation for domain" ?

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Fri, Mar 18, 2016 at 08:55:16AM -0700, PGNet Dev wrote:
> On 03/18/2016 08:48 AM, Wei Liu wrote:> The log message isn't necessarily
> indication for a bug.
> >
> >OVMF is huge, so it pushes up total memory consumption of the guest.  It
> >is normal that you see that message when using OVMF during boot time.
> >
> >What is unclear to me is why your guest keeps trying to ask for more
> >pages after it boots up.  Is your guest trying to balloon in pages?
> 
> Re: ballooning. I don't _think_ so, but not sure how to verify that.
> 

In Dom0:

  xenstore-ls -f /local/domain/$guest_domid

and paste it here.

> Currently, my DomO's cmd line is
> 
>       xen_commandline        : dom0_mem=4096M,max:4096M dom0_max_vcpus=1
> dom0_vcpus_pin=true cpuidle=1 cpufreq=xen clocksource=hpet iommu=verbose
> sched=credit vga=gfx-1920x1080x16 com1=115200,8n1,pci console=com1,vga
> console_timestamps sync_console console_to_ring conring_size=64
> earlyprintk=vga,keep log_buf_len=16M sched_debug apic_verbosity=verbose
> debug loglvl=debug guest_loglvl=debug
> 
> Yes, I'm sure it's subject to improvement.
> 
> On the guest,
> 
>       Kernel command line: BOOT_IMAGE=/vmlinuz root=LABEL=ROOT 
> resume=LABEL=SWAP
> kbdtype=us headless text quiet nofb apparmor=0 edd=off splash=silent noshell
> showopts net.ifnames=0 console=ttyS0,19200n8 systemd.log_level=info
> systemd.log_target=kmsg
> 

First observation is that the log message starts with d1v0, which means
domain 1 vcpu 0, so Dom0 is irrelevant in that case.

The guest kernel command line looks normal.

You can also verify if your guest is trying to balloon in memory by
looking at sysfs knobs.

According to Linux kernel documentation, the relevant knobs live under

/sys/devices/system/xen_memory0/*

> >Do you see the same message repeated when using seabios?
> 
> Haven't yet checked.
> 
> Atm, I've only UEFI guests, setup.  To test, I'll just have to spin up a
> non-UEFI guest.  A bit later ...

Yes, this would be helpful for identifying the issue.

Wei.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel