Re: [Xen-devel] [xen-unstable test] 36728: regressions - FAIL

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 26.03.15 at 05:36, <Ian.Jackson@xxxxxxxxxxxxx> wrote:
> flight 36728 xen-unstable real [real]
> http://www.chiark.greenend.org.uk/~xensrcts/logs/36728/ 
> 
> Regressions :-(
> 
> Tests which did not succeed and are blocking,
> including tests which could not be run:
>  test-amd64-amd64-xl-credit2  12 guest-localmigrate        fail REGR. vs. 
> 36514
>  test-amd64-amd64-xl          12 guest-localmigrate        fail REGR. vs. 
> 36514
>  test-amd64-amd64-xl-multivcpu 12 guest-localmigrate       fail REGR. vs. 
> 36514

Considering

(XEN) mm.c:759:d2v0 Bad L1 flags 400000
(XEN) mm.c:5028:d2v0 ptwr_emulate: could not get_page_from_l1e()
(XEN) Pagetable walk from ffff88001e81afd8:
(XEN)  L4[0x110] = 00000000be59b067 ffffffffffffffff
(XEN)  L3[0x000] = 00000000be59a067 ffffffffffffffff
(XEN)  L2[0x0f4] = 00000000be590067 ffffffffffffffff 
(XEN)  L1[0x01a] = 80000000bf915025 000000000001e81a
(XEN) domain_crash_sync called from entry.S: fault at ffff82d080235094 
create_bounce_frame+0x74/0x148
(XEN) Domain 2 (vcpu#0) crashed on cpu#3:

these are almost certainly the result of d639e6a05a ("x86: allow
64-bit PV guest kernels to suppress user mode exposure of M2P"),
unless this is a kernel side issue that somehow managed to pass
the push gate there.

However, the registers and stack dumped don't show anything
that looks like a PTE with flag bit 22 (_PAGE_GNTTAB) set. Nor
does said change have anything to do with that flag. (In each
case there is one value on the stack that looks like a valid PTE
one, but which doesn't have said flag set.)

What's additionally odd is that the L4, L3, and L2 entries dumped
don't have a valid physical address associated with them.

IOW I'm a little lost for the moment, as I can't immediately see a
connection, not even a remote one. I'll see that I can find time
later today or tomorrow to think some more about his, but in the
meantime unless others have a good idea I'll wait for the bisector
to confirm the suspicion, or in the worst case revert tomorrow
just in case (as I'll be gone for the coming two weeks).

Jan


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