Xen PVH and 'xl save' bug

Xen PVH and 'xl save' bug

[Konrad Rzeszutek Wilk]

This hadn't been tested and unsurprisingly we get:

[  190.091108] calling  vbd-51712+ @ 37, parent: none
[  190.091632] call vbd-51712+ returned 0 after 0 usecs
[  190.092167] PM: freeze of devices complete after 18.587 msecs
[  190.092690] suspending xenstore...
[  190.093272] PM: late freeze of devices complete after 0.062 msecs
[  190.093881] PM: noirq freeze of devices complete after 0.073 msecs
[  190.094539] PM: Calling mce_syscore_suspend+0x0/0x70
[  190.095043] PM: Calling timekeeping_suspend+0x0/0x180
[  190.095470] PM: Calling save_ioapic_entries+0x0/0xb0
[  190.095470] PM: Calling i8259A_suspend+0x0/0x30
[  190.095470] PM: Calling fw_suspend+0x0/0x20
[  190.095470] PM: Calling lapic_suspend+0x0/0x1b0

Entering kdb (current=0xffff88003c806090, pid 9) on processor 0 Oops: (null)
due to oops @ 0xffffffff81045e5b
dCPU: 0 PID: 9 Comm: migration/0 Not tainted 3.14.0-rc1upstream-00019-g4809e80-dirty #2
dtask: ffff88003c806090 ti: ffff88003c810000 task.ti: ffff88003c810000
dRIP: 0010:[<ffffffff81045e5b>]  [<ffffffff81045e5b>] xen_arch_pre_suspend+0x12b/0x170
dRSP: 0018:ffff88003c811d18  EFLAGS: 00010082
dRAX: ffffffffffffffda RBX: 000000000000777d RCX: ffffc90000342000
dRDX: 0000000000000000 RSI: 0000000000000000 RDI: ffffffffff570000
dRBP: ffff88003c811d38 R08: 0000000000000000 R09: 0000000000000001
dR10: 0000000000007ff0 R11: 000000000000057a R12: ffff88000777b000
dR13: 000000000000777d R14: ffff88003ca61d7c R15: ffff88003fc0ea01
dFS:  0000000000000000(0000) GS:ffff88003fc00000(0000) knlGS:0000000000000000
dCS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
dCR2: 00007fc373dd0dd0 CR3: 0000000001c0c000 CR4: 00000000000406f0
dDR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
dDR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
dStack:
 ffff88003c811d28 ffff88003ca61de8 0000000000000000 0000000000000003
 ffff88003c811d48 ffffffff813f30f3 ffff88003c811d78 ffffffff813f3045
 ffff88003c811d78 ffffffff810d7898 ffff88003ca61d58 0000000000000296
dCall Trace:
more> 

I hadn't yet digged in this, but will give more patches on this thread.

Re: Xen PVH and 'xl save' bug

[Mukesh Rathor]

On Mon, 3 Feb 2014 15:04:58 -0500
Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote:

I wouldn't call it a "bug". It's an unimplemented sub-feature... saying
bug could be misleading to some people who'd then expect it to work
under certain circumstances IMO :). Also, it may create impression PVH
is buggy :)... 

thanks
Mukesh

Re: Xen PVH and 'xl save' bug

[George Dunlap]

On Mon, Feb 3, 2014 at 8:53 PM, Mukesh Rathor <mukesh.rathor@oracle.com> wrote:
> On Mon, 3 Feb 2014 15:04:58 -0500
> Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote:
>
> I wouldn't call it a "bug". It's an unimplemented sub-feature... saying
> bug could be misleading to some people who'd then expect it to work
> under certain circumstances IMO :). Also, it may create impression PVH
> is buggy :)...

Yes, save/restore is a special case of migration (or vice versa),
which is not yet supported.

I suppose a seatbelt stopping xl from attempting a migrate might be
helpful, but not necessarily critical at this point.

 -George