On Fri, 2006-09-08 at 15:06 -0400, Jimi Xenidis wrote:
> On Sep 8, 2006, at 10:38 AM, Maria Butrico wrote:
> >
> > I had run my domains and killed them and my console was still  
> > running.  I also have as you know multi patch, your 128m patch and
> > the patch I sent this morning to xen-ppc.
> >
> > I saw this on the machine/xen console (we have so many consoles we  
> > need to be specific)
> >
> > (XEN) paddr_to_maddr: Dom:0 paddr: 0xc000000 type: REMOTE
> > (XEN) WARN at usercopy.c:62
> > (XEN) [000000000000F530] 000000000044C5EC .paddr_to_maddr+0x118/0x1a0
> > (XEN) [000000000000F5F0] 000000000044CBC4 .xencomm_copy_to_guest+0x220/0x2f4
> > (XEN) [000000000000F6F0] 00000000004390D8 .arch_do_dom0_op+0x140/0x4c4
> > (XEN) [000000000000F810] 0000000000404220 .do_dom0_op+0x193c/0x19b4
> > (XEN) [000000000000FC00] 000000000043CE28 .hcall_xen+0xcc/0xfc
> > (XEN) [000000000000FC90] 000000000043CF4C .do_hcall+0x54/0x84
> > (XEN) [000000000000FD20] 0000000000450348 ex_hcall_continued+0xe4/0xf4
> > (XEN) [000000000043B948] 800100017C0802A6
> > (XEN) SP (600000004bffffd8) is not in xen space
> >
> This was seen before (by another person that did not post to the  
> list) and used to be a panic.

Please post all problems and questions to the list! I don't know how
else to ask politely.

> I demoted it to a warn, because we are looking for stuff but I don't  
> think anything will be done with that address.

What do you mean you don't think anything will be done to that address?
It's in copy_to_guest(), so I think that address will be overwritten.

> Basically, any page not belonging to Dom0 can be mapped by Dom0 be  
> cause she's the boss, these pages are either IO or REMOTE (and IO  
> will panic).
> However, I do not believe we should ever be performing guest copies  
> of IO or REMOTE.
> 0xc000000 is 192M
> Hollis, I believe this is an off by one error in the xencomm
> somewhere.
> any ideas?

Not just from that description... :)

How reproducible is it? When it does occur, what are the steps to
reproduce? Please try replacing the WARN with panic, then attach with
gdb so we can at least see which dom0 op it was.

(Note that this tree is backlevel; dom0_ops are gone now. Not that I
think your problem is fixed, but it would be more convenient to debug
with a current tree.)

Hollis Blanchard
IBM Linux Technology Center

Xen-ppc-devel mailing list

Reply via email to