Re: [PATCH 3/5] hvmemul_do_io: If the send to the ioreq server failed do not retry.

[Don Slutz <dslutz@verizon.com>]

On 01/30/15 05:37, Paul Durrant wrote:
>> -----Original Message-----
>> From: Don Slutz [mailto:dslutz@verizon.com]
>> Sent: 30 January 2015 00:52
>> To: xen-devel@lists.xen.org
>> Cc: Andrew Cooper; George Dunlap; Ian Campbell; Ian Jackson; Jan Beulich;
>> Keir (Xen.org); Stefano Stabellini; Wei Liu; Paul Durrant; Don Slutz
>> Subject: [PATCH 3/5] hvmemul_do_io: If the send to the ioreq server failed
>> do not retry.
>>
>> I.E. do just what no backing DM does.
>>
>> Signed-off-by: Don Slutz <dslutz@verizon.com>
>> ---
>>  xen/arch/x86/hvm/emulate.c | 4 ++++
>>  1 file changed, 4 insertions(+)
>>
>> diff --git a/xen/arch/x86/hvm/emulate.c b/xen/arch/x86/hvm/emulate.c
>> index 2ed4344..e9fc070 100644
>> --- a/xen/arch/x86/hvm/emulate.c
>> +++ b/xen/arch/x86/hvm/emulate.c
>> @@ -228,7 +228,11 @@ static int hvmemul_do_io(
>>          {
>>              rc = X86EMUL_RETRY;
>>              if ( !hvm_send_assist_req(&p) )
>> +            {
>> +                /* Since the send failed, do not retry */
>> +                rc = X86EMUL_OKAY;
> 
> I guess this is probably ok. It's a bit moot though as the only circumstance under which hvm_send_assist_req() will return 0 AFAICT, domain_crash() will have been called, or the vcpu is shutting down.
> 


Turns out this is not ok.  Jan pointed out that for guest suspend to
work correctly, you need to retry.

   -Don Slutz


>   Paul
> 
>>                  vio->io_state = HVMIO_none;
>> +            }
>>              else if ( p_data == NULL )
>>                  rc = X86EMUL_OKAY;
>>          }
>> --
>> 1.8.4
>