Re: [xen-unstable test] 152067: regressions - trouble: fail/pass/starved

[Jan Beulich <jbeulich@xxxxxxxx>]

On 22.07.2020 02:37, osstest service owner wrote:
> flight 152067 xen-unstable real [real]
> http://logs.test-lab.xenproject.org/osstest/logs/152067/
> 
> Regressions :-(
> 
> Tests which did not succeed and are blocking,
> including tests which could not be run:
>  test-amd64-amd64-dom0pvh-xl-intel 18 guest-localmigrate/x10 fail REGR. vs. 
> 152045

Jul 21 16:20:58.985209 [  530.412043] libxl-save-help: page allocation failure: 
order:4, mode:0x60c0c0(GFP_KERNEL|__GFP_COMP|__GFP_ZERO), nodemask=(null)

My first reaction to this would be to ask if Dom0 was given too little
memory here. Or of course there could be a memory leak somewhere. But
the system isn't entirely out of memory (about 7Mb left), so perhaps
the "order:4" aspect here also plays a meaningful role. Hence ...

Jul 21 16:21:00.390810 [  530.412448] Call Trace:
Jul 21 16:21:00.402721 [  530.412499]  dump_stack+0x72/0x8c
Jul 21 16:21:00.402801 [  530.412541]  warn_alloc.cold.140+0x68/0xe8
Jul 21 16:21:00.402841 [  530.412585]  __alloc_pages_slowpath+0xc73/0xcb0
Jul 21 16:21:00.414737 [  530.412640]  ? __do_page_fault+0x249/0x4d0
Jul 21 16:21:00.414786 [  530.412681]  __alloc_pages_nodemask+0x235/0x250
Jul 21 16:21:00.426555 [  530.412734]  kmalloc_order+0x13/0x60
Jul 21 16:21:00.426619 [  530.412774]  kmalloc_order_trace+0x18/0xa0
Jul 21 16:21:00.426671 [  530.412816]  alloc_empty_pages.isra.15+0x24/0x60
Jul 21 16:21:00.438447 [  530.412867]  
privcmd_ioctl_mmap_batch.isra.18+0x303/0x320
Jul 21 16:21:00.438507 [  530.412918]  ? vmacache_find+0xb0/0xb0
Jul 21 16:21:00.450475 [  530.412957]  privcmd_ioctl+0x253/0xa9b

... perhaps we ought to consider re-working this code path to avoid
order > 0 allocations (may be as simple as switching to vmalloc(),
but I say this without having looked at the code).

Jan