[Xen-devel] question about ioreq server

[Wen Congyang <wency@xxxxxxxxxxxxxx>]

Hi, all

I am trying to rebase our colo codes to upstream recently. I meet the following
error in the test:
(XEN) Assertion 'consumer_is_xen(lchn)' failed at event_channel.c:1202
(XEN) ----[ Xen-4.5-unstable  x86_64  debug=y  Not tainted ]----
(XEN) CPU:    3
(XEN) RIP:    e008:[<ffff82d080107464>] notify_via_xen_event_channel+0xac/0x11a
(XEN) RFLAGS: 0000000000010246   CONTEXT: hypervisor
(XEN) rax: ffff830200d18000   rbx: ffff830200d19000   rcx: ffff830200d190b8
(XEN) rdx: ffff830200d18000   rsi: 0000000000000000   rdi: ffff830200d190bc
(XEN) rbp: ffff8302175ef398   rsp: ffff8302175ef378   r8:  00000000000c0000
(XEN) r9:  0000000000000004   r10: 0000000000020000   r11: 00000000f3044014
(XEN) r12: ffff830200d190b8   r13: 0000000000000000   r14: ffff830200d19000
(XEN) r15: 00000000f3044010   cr0: 0000000080050033   cr4: 00000000001526f0
(XEN) cr3: 0000000208e8b000   cr2: 0000000001b50004
(XEN) ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: 0000   cs: e008
(XEN) Xen stack trace from rsp=ffff8302175ef378:
(XEN)    ffff8302175ef3e8 ffff8302175ef438 ffff82c00020f000 ffff830208e397f0
(XEN)    ffff8302175ef3c8 ffff82d0801b5c43 ffff8302175ef438 0000000000000001    
                     
(XEN)    ffff82e00413dde0 0000000000000004 ffff8302175ef3e8 ffff82d0801b5cfe    
                     
(XEN)    0000000000000004 ffff83008a2bf000 ffff8302175ef488 ffff82d0801af231    
                     
(XEN)    0000000000000004 0000000000000010 ffff8302175ef4f8 00ff830000000004    
                     
(XEN)    0000000100000001 0000000000000000 0000000000000004 0000000200000010    
                   
(XEN)    00000000f3044010 0000000000000000 0000000400000001 0120000000000000
(XEN)    00000000000f3044 0000000000000004 0000000000000004 0000000000000010
(XEN)    ffff8302175efb88 ffff8302175ef980 ffff8302175ef4a8 ffff82d0801af3b1
(XEN)    ffff830200000000 ffff8302175ef980 ffff8302175ef538 ffff82d0801afe4e
(XEN)    ffff8302175ef980 ffff8302175ef550 ffff8302175ef4f0 ffff8302175ef4f8
(XEN)    ffff830200000002 00000001175ef508 ffff8302175ef510 00000000f3044010
(XEN)    0000000000000001 ffffd000211bf010 ffff8302175ef558 ffff8302175efb88
(XEN)    000000000000008b 0000000000000000 0000000000000000 ffff82d08027a320
(XEN)    ffff8302175ef548 ffff82d0801aff79 ffff8302175ef558 ffff82d08018d1db
(XEN)    ffff8302175efab8 ffff82d08018f2a6 ffff82d08018d1db ffff8302175efad0
(XEN)    ffff82d08018f700 ffffffffffd0d210 ffff8302175ef5d8 0000000000000018
(XEN)    0000000000000001 0000000000000000 0000000000000018 000000048027a300
(XEN)    ffff8302175ef588 ffff82d0801aff79 00000004175ef5d8 000000d08018d180
(XEN)    ffffd00000000001 000000080018f72a 00000002175e0000 ffffd00000000001
(XEN) Xen call trace:
(XEN)    [<ffff82d080107464>] notify_via_xen_event_channel+0xac/0x11a
(XEN)    [<ffff82d0801b5c43>] hvm_send_assist_req_to_ioreq_server+0x132/0x14c
(XEN)    [<ffff82d0801b5cfe>] hvm_send_assist_req+0x3e/0x45
(XEN)    [<ffff82d0801af231>] hvmemul_do_io+0x4dd/0x630
(XEN)    [<ffff82d0801af3b1>] hvmemul_do_mmio+0x2d/0x2f
(XEN)    [<ffff82d0801afe4e>] __hvmemul_read+0x227/0x29c
(XEN)    [<ffff82d0801aff79>] hvmemul_read+0x12/0x19
(XEN)    [<ffff82d08018d1db>] read_ulong+0xe/0x10
(XEN)    [<ffff82d08018f2a6>] x86_emulate+0x1745/0xf8ef
(XEN)    [<ffff82d0801ae6b0>] hvm_emulate_one+0x15e/0x25e
(XEN)    [<ffff82d0801bd94d>] handle_mmio+0x69/0x1f9
(XEN)    [<ffff82d0801b56bc>] hvm_hap_nested_page_fault+0x28a/0x489
(XEN)    [<ffff82d0801db48d>] vmx_vmexit_handler+0x1446/0x17c7
(XEN)    [<ffff82d0801e1ad1>] vmx_asm_vmexit_handler+0x41/0xc0
(XEN) 
(XEN) 
(XEN) ****************************************
(XEN) Panic on CPU 3:
(XEN) Assertion 'consumer_is_xen(lchn)' failed at event_channel.c:1202
(XEN) ****************************************
(XEN) 
(XEN) Reboot in five seconds...


COLO is very similar as remus, except that secondary vm is running.
This problem happens in the slave side. The things I do in slave side:
1. call libxl__xc_domain_restore_done() after migration to restore the secondary
   vm, and unpause it.
2. do a new checkpoint: suspend secondary vm and update secondary vm's state.
3. resume the secondary vm. The hypervisor crash now.

I study ioreq server related codes now. I am also happy if anyone can help me.

Thanks
Wen Congyang

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