On Sep 6, 2006, at 4:40 PM, Hollis Blanchard wrote:

On Wed, 2006-09-06 at 14:56 -0500, Hollis Blanchard wrote:
On Wed, 2006-09-06 at 14:44 -0400, Amos Waterland wrote:
Using current xen.hg and linux.hg, I get this when I try to
run `xm vcpu-list':

(XEN) ****************************************
(XEN) Panic on CPU 0:
(XEN) paddr_to_maddr: Dom:0 paddr: 0xf5d5ecc0 bad type:0x3
(XEN) ****************************************
(XEN) Reboot in five seconds...

Note that `xm list' works fine. The 0xf5d5ecc0 address is consistent
across boots.

hmm, "xm vcpu-list" worked fine before the merge, I have not had a chance to look at the recent changes in that space. However, please note that the paddr is firmly in the I/O Hole of a 970 class machine (hence type 3) and you have no business xencomm'ing that address. I suspect that the paddr we are seeing is actually an from the stack of an app and the xencomm logic if faulty.

looks like Hollis has it figured out. :)

I see the same thing. The backtrace is:
#0  panic (fmt=0x456d10 "%s: Dom:%d paddr: 0x%lx bad type:0x%x\n")
    at console.c:610
#1 0x00000000004459bc in paddr_to_maddr (paddr=0xf6fb7278) at usercopy.c:65
#2  0x000000000044619c in xencomm_handle_is_null (ptr=0xf6fb7278)
    at usercopy.c:265
#3 0x0000000000401c58 in cpumask_to_xenctl_cpumap (xenctl_cpumap=0xfaa8,
    cpumask=0xb7e178) at domctl.c:36
#4 0x0000000000403348 in do_domctl (u_domctl={__pad = 0xfc20, p = 0x7d04000})
    at domctl.c:396
#5 0x00000000004373c4 in hcall_xen (num=0x24, regs=0xfd90) at hcalls.c:76
#6  0x00000000004374e0 in do_hcall (regs=0xfd90) at hcalls.c:104
#7  0x0000000000449d34 in ex_hcall_continued () at misc.h:31

OK, the problem was that the vcpuaffinity calls had a hidden handle that wasn't being mapped. I believe I've fixed this now with Linux changeset

BTW: this is a panic, only because of all this is at an early stage, normally it should just fail. BTW2: that reminds me, we do 0 bounds checking on the rest of memory, we should add that soon, I'll file a bug.


Xen-ppc-devel mailing list

Reply via email to