Bug 200004 - domain_crash_sync called from entry.S
Summary: domain_crash_sync called from entry.S
Status: CLOSED RAWHIDE
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel-xen
Version: rawhide
Hardware: x86_64
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Xen Maintainance List
QA Contact: Brian Brock
URL:
Whiteboard:
Keywords:
: 200126 (view as bug list)
Depends On:
Blocks: 200124
TreeView+ depends on / blocked
 
Reported: 2006-07-24 20:08 UTC by Brian Brock
Modified: 2007-11-30 22:11 UTC (History)
3 users (show)

(edit)
Clone Of:
(edit)
Last Closed: 2006-07-31 20:51:14 UTC


Attachments (Terms of Use)
linux-2.6-xenguest-no-writenotify.patch (406 bytes, patch)
2006-07-27 10:13 UTC, Mark McLoughlin
no flags Details | Diff

Description Brian Brock 2006-07-24 20:08:00 UTC
2.6.17-1.2439.fc6xen x86-64
xen-3.0.2-18 x86-64

xenguest-install.py tracebacks.  `xm dmesg` shows a fault.

100% reproducible on at least affected system.

[xenguest-install.py traceback]
-(root@et-5:pts/0)-(0 jobs)-(349:41)-(Mon Jul 24:15:59:48)-
-(~:$)-> xenguest-install.py -n test8 -f /root/test8.img -s 20 -r 1024 -p -l
nfs:curly.devel.redhat.com:/vol/engineering/devarchive/redhat/nightly/rawhide-20060724/development/x86_64/os


Starting install...
libvir: Xen Daemon error : GET operation failed: No such domain test8
Traceback (most recent call last):
  File "/usr/sbin/xenguest-install.py", line 475, in ?
    main()
  File "/usr/sbin/xenguest-install.py", line 466, in main
    start_paravirt_install(name, ram, disk, mac, uuid, bridge, src, options.extra)
  File "/usr/sbin/xenguest-install.py", line 343, in start_paravirt_install
    dom = conn.createLinux(cfgxml, 0)
  File "/usr/lib64/python2.4/site-packages/libvirt.py", line 233, in createLinux
    if ret is None:raise libvirtError('virDomainCreateLinux() failed')
libvirt.libvirtError: virDomainCreateLinux() failed


[tail of xm dmesg output]
(XEN) Dom0 has maximum 16 VCPUs
(XEN) Initrd len 0x35c600, start at 0xffffffff805fe000
(XEN) Scrubbing Free RAM:
.......................................................................................................done.
(XEN) Xen trace buffers: disabled
(XEN) Xen is relinquishing VGA console.
(XEN) *** Serial input -> DOM0 (type 'CTRL-a' three times to switch input to Xen).
(XEN) (file=irq.c, line=459) Cannot bind IRQ 2 to guest. In use by 'cascade'.
(XEN) (file=irq.c, line=459) Cannot bind IRQ 2 to guest. In use by 'cascade'.
(XEN) domain_crash_sync called from entry.S
(XEN) Domain 2 (vcpu#0) crashed on cpu#14:
(XEN) ----[ Xen-3.0-unstable    Not tainted ]----
(XEN) CPU:    14
(XEN) RIP:    e033:[<ffffffff80200000>]
(XEN) RFLAGS: 0000000000010202   CONTEXT: guest
(XEN) rax: 0000000000000000   rbx: 0000000000000000   rcx: 0000000000000000
(XEN) rdx: 0000000000000000   rsi: ffffffff80ec3000   rdi: 0000000000000000
(XEN) rbp: 0000000000000000   rsp: ffffffff80ed2000   r8:  0000000000000000
(XEN) r9:  0000000000000000   r10: 0000000000000000   r11: 0000000000000000
(XEN) r12: 0000000000000000   r13: 0000000000000000   r14: 0000000000000000
(XEN) r15: 0000000000000000   cr0: 000000008005003b   cr3: 00000001d95d4000
(XEN) ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: e02b   cs: e033
(XEN) Guest stack trace from rsp=ffffffff80ed2000:
(XEN)   Fault while accessing guest memory.

Comment 1 Brian Brock 2006-07-24 20:15:04 UTC
Also, selinux disabled at boot with selinux=0

Comment 2 Brian Brock 2006-07-26 19:10:55 UTC
2.6.17-1.2449.fc6xen behaves similar, with the difference that the output in xm
dmesg differs slightly (stack empty):

(XEN) domain_crash_sync called from entry.S
(XEN) Domain 1 (vcpu#0) crashed on cpu#15:
(XEN) ----[ Xen-3.0-unstable    Not tainted ]----
(XEN) CPU:    15
(XEN) RIP:    e033:[<ffffffff80200000>]
(XEN) RFLAGS: 0000000000010202   CONTEXT: guest
(XEN) rax: 0000000000000000   rbx: 0000000000000000   rcx: 0000000000000000
(XEN) rdx: 0000000000000000   rsi: ffffffff80ec5000   rdi: 0000000000000000
(XEN) rbp: 0000000000000000   rsp: ffffffff80ed4000   r8:  0000000000000000
(XEN) r9:  0000000000000000   r10: 0000000000000000   r11: 0000000000000000
(XEN) r12: 0000000000000000   r13: 0000000000000000   r14: 0000000000000000
(XEN) r15: 0000000000000000   cr0: 000000008005003b   cr3: 00000001defd5000
(XEN) ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: e02b   cs: e033
(XEN) Guest stack trace from rsp=ffffffff80ed4000:
(XEN)   Stack empty.


Comment 3 Brian Brock 2006-07-26 19:24:54 UTC
Putting on the xen beta blocker list.

Seems similar to bug 200126 comment 1.  Note that this bug occurs on paravirt.

Comment 4 Jeremy Katz 2006-07-26 19:29:08 UTC
It's all guests -- the crash is before there's really a difference between FV
and PV.

Comment 5 Jeremy Katz 2006-07-26 19:29:23 UTC
*** Bug 200126 has been marked as a duplicate of this bug. ***

Comment 6 Brian Brock 2006-07-26 20:45:11 UTC
On i686 I see the xenguest-install.py traceback (identical), but nothing new in
`xm dmesg` output.

Comment 7 Brian Brock 2006-07-26 20:46:33 UTC
2.6.17-1.2454.fc6xen behaves identically.

Comment 8 Jeremy Katz 2006-07-26 20:55:15 UTC
The i686 guest creation crash is bug 200125.  And there was output when we were
building the HV with debug=y.  Of course, I only realized that after I hacked
libxenguest into itty bitty pieces :/

Comment 9 Rik van Riel 2006-07-27 05:03:26 UTC
As a data point, I can start domains just fine on x86-64 with kernel
2.6.17-1.2157_FC5xen0.

Comment 10 Mark McLoughlin 2006-07-27 09:14:23 UTC
It looks to be that linux-2.6-mm-tracking-dirty-pages.patch is causing the dom0
kernel to map the guests pages without VM_WRITE. This, in turn, is causing the
guest to fault on  its first instruction

Comment 11 Mark McLoughlin 2006-07-27 10:13:10 UTC
Created attachment 133139 [details]
linux-2.6-xenguest-no-writenotify.patch

Probably incorrect patch ...

Basic idea, though, is to catch these mappings in vma_wants_writenotify() using
the flags added by privcmd_mmap()

Comment 12 Jeremy Katz 2006-07-27 15:51:07 UTC
With 2.6.17-1.2461.fc6xen, this works again for me.

--> MODIFIED.  bbrock -- when you verify, can you close?

Comment 13 Brian Brock 2006-07-31 20:51:14 UTC
Verified in rawhide-20060731 with 2.6.17-1.2478.fc6xen and xen-3.0.2-22

install progresses to lang selection with similar xenguest-install.py command. 
`xm dmesg` doesn't show the fault listed above at that point

closing rawhide


Note You need to log in before you can comment on or make changes to this bug.