Re: [Qemu-discuss] Qemu-COLO bug report

[Zhang, Chen]

> -----Original Message-----
> From: wenzt [mailto:address@hidden
> Sent: Monday, April 15, 2019 10:32 AM
> To: 'Lukas Straub' <address@hidden>
> Cc: Zhang, Chen <address@hidden>; 'qemu-discuss' <qemu-
> address@hidden>
> Subject: Re: [Qemu-discuss] Qemu-COLO bug report
> 
> Thank you for quick response !
> 
> About lose network connection, I give the wrong parameters in qemu cmdline
> for SVM, sorry.

Sorry for slow reply.
Can you share the detail of the wrong parameters? In case other people have the 
same issue.

> 
> About : address@hidden:colo_compare_main : secondary:
> unsupported.
> I got between checkpoint:
> address@hidden:colo_send_message Send 'vmstate-send' message
> address@hidden:colo_send_message Send 'vmstate-size' message
> address@hidden:colo_receive_message Receive 'vmstate-received'
> message
> address@hidden:colo_compare_main : secondary: unsupported
> packet in < address@hidden:colo_receive_message Receive
> 'vmstate-loaded'
> message
> {"timestamp": {"seconds": 1555071933, "microseconds": 805979}, "event":
> "RESUME"}
> address@hidden:colo_vm_state_change Change 'stop' => 'run'
> 
> I use tcpdump  -s 0 port 9004 -w dump.pcap,  9004 is secondary_in port on
> colo-compare object.
> I find out that data in pri_rs.buf  same as DATA section in ETH packet rather
> than HEAD section(what we expected).
> 
> Did you get the same issue ?

Why you dump the secondary_in and find out pri_rs.buf? pri_rs.buf should used 
by primary_in.
Maybe I misunderstand your means, please give more detail.

> 
> Regards,
> Zhengtao
> 
> > Moving the assert after the mutex lock does indeed fix this Bug for me
> 

Thanks Lukas, I think you can submit a patch for this issue. But if you are not 
convenient, I can do it too.

Thanks
Zhang Chen

> > Regards,
> > Lukas Straub
> 
> 
> diff --git a/net/colo-compare.c b/net/colo-compare.c index
> bf10526..fcb4911 100644
> --- a/net/colo-compare.c
> +++ b/net/colo-compare.c
> @@ -813,9 +813,8 @@ static void colo_compare_handle_event(void *opaque)
>          break;
>      }
> 
> -    assert(event_unhandled_count > 0);
> -
>      qemu_mutex_lock(&event_mtx);
> +    assert(event_unhandled_count > 0);
>      event_unhandled_count--;
>      qemu_cond_broadcast(&event_complete_cond);
>      qemu_mutex_unlock(&event_mtx);
>