grant table unmap failure makes guest unreapable and causes xen oops

[Kip Macy <kip.macy@gmail.com>]

I just hit this so I don't fully understand it yet, but it looks like
there may be some race condition with grant_table unmap requests and
garbage collection of domain memory on crashed guests.

My centos4 domU isn't finding its init (this may be the breakage in
file-backed VBDs that Mark mentioned - it was finding it a couple of
days ago) and thus calls HYPERVISOR_crash:

Freeing unused kernel memory: 92k freed
Kernel panic - not syncing: No init found.  Try passing init= option to kernel.

[root@rs0 ~]# xm list
Name              Id  Mem(MB)  CPU VCPU(s)  State  Time(s)  Console
Domain-0           0      251            0         1       r----     1013.3
rhel4_0               1          0            3         1        ----c
         0.5     9601

The following errors show up on the console:

(XEN) (file=grant_table.c, line=500) Bad handle (0).
(XEN) (file=grant_table.c, line=500) Bad handle (49152).
(XEN) (file=grant_table.c, line=500) Bad handle (49792).
(XEN) (file=grant_table.c, line=500) Bad handle (0).
(XEN) (file=grant_table.c, line=500) Bad handle (61440).

And the guest never goes away.

[root@rs0 ~]# xm destroy 1
[root@rs0 ~]# xm list
Name              Id  Mem(MB)  CPU VCPU(s)  State  Time(s)  Console
Domain-0          0          251        0             1     r----   1208.2
rhel4_0               1             0        3              1    
----c        0.5     9601

restarting xend here is interesting:

[root@rs0 ~]# xend start
DBMap>introduceDomain> 1 69067 <EventChannel dom1:0:14 dom2:1:2>
/domain/4042ebcc-778d-4488-a0bd-6152c42ba98b
Traceback (most recent call last):
<snip>
RuntimeError: (9, 'Bad file descriptor')

Message from syslogd@rs0 at Wed Jul  6 10:59:17 2005 ...
rs0 xenstored: xenstored corruption: connection id 0: err Bad address:
Unknown error 14 (Bad address)
Exception starting xend: (9, 'Bad file descriptor')

On the console we see:
(XEN) (file=/build/kmacy/xen/xen-unstable.hg/xen/include/asm/mm.h,
line=187) Error pfn be9: rd=ffbf8a80, od=ffbf8a80, caf=00000000,
taf=f0000001
(XEN) (file=/build/kmacy/xen/xen-unstable.hg/xen/include/asm/mm.h,
line=187) Error pfn 10dcb: rd=ffbf8a80, od=00000000, caf=00000000,
taf=f0000000
[ERR] corruptxenstored corruption: connection id 0: err Bad address:
Unknown error 14 (Bad address)

*NOW* comes the fun part:
[root@rs0 ~]# /sbin/shutdown -r now

Broadcast message from root (pts/1) (Wed Jul  6 11:01:00 2005):

The system is going down for reboot NOW!
INIT: Sending processes the TERM signal
(XEN) CPU:    0
(XEN) EIP:    e008:[<ff10b882>]
(XEN) EFLAGS: 00210202   CONTEXT: hypervisor
(XEN) eax: 0000000a   ebx: 00000000   ecx: 00000000   edx: 00000003
(XEN) esi: 00000001   edi: ffbf2700   ebp: ffbf1004   esp: ff103e04
(XEN) cr0: 8005003b   cr3: 181cd000
(XEN) ds: e010   es: e010   fs: 0000   gs: 0000   ss: e010   cs: e008
(XEN) Xen stack trace from esp=ff103e04:
(XEN)    00000001 00000052 00000000 00000400 fec6b000 ff1a9900 fc400000 00000f00
(XEN)    00000000 00000000 00000001 000a0067 fc400f00 ff1a9900
ff1a7080 [ff1269de]
(XEN)    ff1a7080 ff1a9900 000000a0 00000000 fec6b000 32ff0001 ff1a9900 0000067c
(XEN)    00000261 000a0067 fec72000 [ff12a110] 000a0067 ff1a9900
00000000 [ff13b9ef]
(XEN)    181cd000 ff103fb4 00000001 c7910985 ff1a9b80 fec72000
ff1a9900 [ff12a266]
(XEN)    ff1a9900 fec72000 ff1b2000 [ff12ba14] fec71000 ff1a9b80
ff103fb4 ff1a9b90
(XEN)    c7910985 fe31e440 00000000 00000000 00000000 00000000
ff1a9900 [ff12d55e]
(XEN)    ff1a9900 00000000 0000000c 00200286 ff103fb4 ff1a9900
[ff13b9ef] 181cd000
(XEN)    001446c9 00000000 c7910984 ff1a9900 00018ef0 18ef0061
[ff12eb6a] 00018ef0
(XEN)    ffffffff 00000010 ff1a9900 00000007 c873e000 00010000 c0568ee0 000002db
(XEN)    32db0001 ff103fb4 ff1a9b80 ff1a9900 00000000 fe3f8b6c ff1a9900 ff103fb4
(XEN)    c7910984 ffbf3080 [ff13e867] 00000000 00000000 00000000
00000000 00000001
(XEN)    00000005 00000020 ee000000 ffbf3080 ffbf3bf8 ffbf3080 ffbf3080 ffbf3080
(XEN)    00007ff0 c8623284 b6e69000 [ff14a8f3] c8683eec 00000001
00000000 00007ff0
(XEN)    c8623284 b6e69000 0000001a 000e0003 c0115b33 00000061 00200282 c8683eec
(XEN)    00000069 0000007b 0000007b 00000000 00000000 00000000 ffbf3080
(XEN) Xen call trace from esp=ff103e04:
(XEN)    [<ff1269de>] [<ff12a110>] [<ff13b9ef>] [<ff12a266>]
[<ff12ba14>] [<ff12d55e>]
(XEN)    [<ff13b9ef>] [<ff12eb6a>] [<ff13e867>] [<ff14a8f3>]

****************************************
Panic on CPU0:
CPU0 FATAL PAGE FAULT
[error_code=0000]
Faulting linear address: 00000004
****************************************

Reboot in five seconds...


Line 910 of "grant_table.c" starts at address 0xff10b87f
<gnttab_check_unmap+175>
   and ends at 0xff10b888 <gnttab_check_unmap+184>.
<...>
             ( readonly ? 1 : (!(map->ref_and_flags & GNTMAP_readonly))))
        {
            ref = (map->ref_and_flags >> MAPTRACK_REF_SHIFT);
            act = &rgt->active[ref];      <- line 910
 
            spin_lock(&rgt->lock);

            if ( act->frame != frame )
<...>
0xff10b882 <gnttab_check_unmap+178>:    mov    0x4(%ecx),%eax
0xff1269de <put_page_from_l1e+270>:     test   %eax,%eax
0xff12a110 <revalidate_l1+176>: jmp    0xff12a090 <revalidate_l1+48>
0xff13b9ef <__flush_tlb_mask+239>:      mov    0x44(%ebx),%eax
0xff12a266 <ptwr_flush+246>:    mov    %edi,(%esp)
0xff12d55e <do_mmuext_op+1150>: jmp    0xff12d17a <do_mmuext_op+154>
0xff13b9ef <__flush_tlb_mask+239>:      mov    0x44(%ebx),%eax
0xff12eb6a <ptwr_do_page_fault+506>:    mov    %eax,%esi
0xff13e867 <do_page_fault+423>: test   %eax,%eax
0xff14a8f3 <hypercall+83>:      mov    %eax,0x18(%esp)