Re: a domain VTx with the VNIF does hang.

[Hirofumi Tsujimura <tsujimura.hirof@jp.fujitsu.com>]

The patch which fix a inter-domain communication using rx-copy path
was released in the changeset 11737.
I confirmed that the problem was fixed in the changeset 11751.

Thank you.



Hirofumi Tsujimura wrote:
> Hi all, my name is Hirofumi Tsujimura.
> We are porting and testing a PV-on-HVM in the IPF.
> This is a first time to send mail.
> 
> I probably found the problem when I tested the VNIF.
> My operation for the test is following.
>   1. create a domain VTx and attach the VNIF in it.
>   2. create a domain U.
>   3. send a packet to the domain VTx from the domain U with ping
>      command.
>      Then, the domain VTx does hang.
>      And, it outputs as follows.
>      The domain U and a domain 0 doesn't do hang.
> 
>     EAX: c03c7000 EBX: 00000286 ECX: 00000286 EDX: da676000
>     ESI: 000004e2 EDI: 0000000a EBP: c03c7fb8 DS: 007b ES: 007b
>     CR0: 8005003b CR2: b7ee2044 CR3: 1a6af000 CR4: 000006d0
>       [<c0118f92>] __printk_ratelimit+0x8f/0x95
>       [<e08a6316>] xennet_get_responses+0xca/0x31c [xen_vnif]
>       [<c0103dc2>] do_IRQ+0x70/0x7a
>       [<e08a6851>] netif_poll+0x1c/0x60v [xen_vnif]
>       [<c0103dc2>] net_rx_action+0x35/0x7d
>       [<c011c47d>] __do_softirq+0x38/0x3f
>       [<c0103e94>] do_softirq+0x38/0x3f
>       =======================
>       [<c0103dc2>] do_IRQ+0x70/0x7a
>       [<c0102c1a>] common_interrupt+0x27/0x2b
>       [<c014b5d8>] kmem_cache_free+0x27/0x2b
>       [<c013dac4>] free_pgtables+0x31/0x63
>       [<c01421ab>] exit_mmap+0x6a/0xb6
>       [<c0115f53>] mmput+0x1f/0x76
>       [<c011a58d>] do_exit+0x173/0x32b
>       [<c011a7f7>] sys_exit_group+0x0/0xd
>       [<c01029f3>] sysenter_past_esp+0x54/0x75
>     printk: 64425470 messages suppressed.
>     netfront: rx->offset: 0, size: 4294967295
>     printk: 64601036 messages suppressed.
>     netfront: rx->offset: 0, size: 4294967295
> *** Please look at screenshots about the detail of the log. ***
> 
> The test environment is as follows.
>   - Architecture  : x86
>   - Xen change set: 11683
>   - Domain 0      : kernel2.6.16.29-xen
>   - Domain U      : kernel2.6.16.29-xen
>   - Domain VTx    : kernel2.6.16
> 
> Probably, a domain VTx with the VNIF does Hang when it recieves a packet
> from some domain U or from some domain VTx with the VNIF.
> 
> 
> Additionally, I similarly tested the VNIF in the following environments
> today(On October 3 2006).
>   - Architecture  : x86
>   - Xen change set: 11686
>   - Domain 0      : kernel2.6.16.29-xen
>   - Domain U      : kernel2.6.16.29-xen
>   - Domain VTx    : kernel2.6.16
> 
> As a result, a kernel of domain VTx with the VNIF does panic when I use
> "xm network-attach domainVTx eth1" in a domain 0.
> Even if that command succeeds, it similarly does hang when it recieves a
> packet.
> 
> What's happend?
> Thank you.
> 
> --
> FUJITSU COMPUTER TECHNOLOGIES LIMITED
> Development Dept. 5  Division 1
> Hirofumi Tsujimura
> E-MAIL: tsujimura.hirof@jp.fujitsu.com


--
FUJITSU COMPUTER TECHNOLOGIES LIMITED
Development Dept. 5  Division 1
Hirofumi Tsujimura
E-MAIL: tsujimura.hirof@jp.fujitsu.com