[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [Xen on Arm BUG] Alignment trap: not handling instruction



On Thu, 2013-05-02 at 15:34 +0100, Sander Bogaert wrote:
> On 2 May 2013 15:44, Ian Campbell <Ian.Campbell@xxxxxxxxxx> wrote:
> > On Thu, 2013-05-02 at 12:46 +0100, Sander Bogaert wrote:
> >> Hi,
> >>
> >> just a small bug report, not sure if it's useful. I was just
> >> installing some packages, not doing anything fancy, and I got:
> >>
> >> [ 1113.248819] Alignment trap: not handling instruction e8513f00 at 
> >> [<b6e88a52>]
> >> [ 1113.254644] Unhandled fault: alignment exception (0x001) at 0x20746e6f
> >
> > That looks like a kernel message? I assume it was running under Xen.
> 
> Yes it was.
> 
> >
> > 0xb.... is a userspace address, so some bit of userspace is up to
> > "something".
> 
> I'm still learning of course, I missed this. So basically
> Xen is mapped on 0x5000000 - ...
> Dom0 kernel on 0x8000000 - ...
> 
> If an address isn't in those ( physical ) ranges then it's userspace.
> Is that correct?

Nope ;-)

User vs kernel space is a virtual address distinction. The usual
arrangement for 32-bit Linux (including arm) is a 3:1 split, so 3G of
user and 1G of kernel, meaning kernels a re from 0xc0000000 onwards.

Also remember that physical addresses mean different things to the
hypervisor (physical == real physical) and guests (physical ==
intermediate physical address, due to second stage paging).

> > I'd have to reread the bits of the ARM and virt extensions to know if
> > this was our doing ;-)
> 
> I was just assuming since it's a trap but it could indeed be one Xen
> isn't set up to receive. Since I'm using a recent linux kernel and
> wasn't doing anything funny, just normal usage, I was thinking the
> 'odd factor' was Xen underneath. I'll figure out which instruction
> e8513f00 is.

In principal what to do about alignment traps (including whether to
receive them at all) is all the guests decision, that's not to say Xen
isn't doing something which causes the kernel to get it wrong.

Ian.


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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.