Re: [Xen-devel] [xen-4.0-testing test] 2032: regressions - FAIL

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Fri, 2010-08-20 at 08:58 +0100, Ian Campbell wrote:
> On Thu, 2010-08-19 at 09:00 +0100, Ian Campbell wrote:
> > 
> > e73f4 is the merge of 2.6.32.19, e6b9b is the merge of 2.6.32.18,
> > there are no proper xen.git changesets between the two -- only the
> > contents of 2.6.32.19, which does include some Xen save/restore
> > "fixes".
> 
> 2.6.32.19 has a known issue relating to the new stack guard page which
> has prompted a quick 2.6.32.20-rc1 with only a couple of patches.
> 
> The issue is supposedly PAE and/or HIGHPTE specific but I wonder if it
> also has an impact under Xen in other configurations?
> 
> The save/restore issue we are seeing seems to relate to an inability to
> lock memory for a hypercall and d7824370e26325c881b665350ce64fb0a4fde24a
> (the second fix in 2.6.32.20-rc) specifically mentions fixing up a
> user-visible mlock change relating to mlock.
> 
> I'll try and confirm that, if I can ever manage to get a consistent
> repro.

Reverting
        ab832422673d1774c4ce3941f2ac87743d73bded mm: fix missing page table 
unmap for stack guard page failure case
        7e281afe24330aeea86113ac241eabdac8ba2311 mm: keep a guard page below a 
grow-down stack segment
resolves the save/restore issue for me.

However cherry-picking the 2.6.32.20 fixes
        11ac552477e32835cb6970bf0a70c210807f5673 mm: fix page table unmap for 
stack guard page properly
        d7824370e26325c881b665350ce64fb0a4fde24a mm: fix up some user-visible 
effects of the stack guard page
does not do the trick.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel