[ofw] RE: IBAT allow simultaneous device access

[Smith, Stan]

Fab Tillier wrote:
>> Fab & community,
>>   Are you good with this patch to allow simultaneous IBAT device
>> access from user mode?
>
> Looks very reasonable to me, I'm surprised we've never run into this
> issue before...

OK, thanks; will commit.
Case addressed is some (>2) MPI ranks @ a single node all attempting to access the IBAT device.
We were seeing E_PENDING being returned.

>
> -Fab
>
>>
>> Thanks,
>>
>> Stan.
>>
>>
>> --- C:/Documents and Settings/scsmith/Local Settings/Temp/ibat.cpp-
>> revBASE.svn000.tmp.cpp       Mon Aug 17 14:33:47 2009
>> +++ C:/Documents and Settings/scsmith/My Documents/openIB-
>> windows/SVN/gen1/branches/WOF2-1/core/ibat/user/ibat.cpp      Mon
>> Aug 17 14:01:18 2009 @@ -98,7 +98,7 @@
>>          return E_INVALIDARG;
>>
>>      H hIbatDev = CreateFileW( IBAT_WIN32_NAME,
>> -        MAXIMUM_ALLOWED, 0, NULL,
>> +        MAXIMUM_ALLOWED,(FILE_SHARE_READ|FILE_SHARE_WRITE), NULL,
>>          OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL );
>>      if( hIbatDev == INVALID_HANDLE_VALUE )
>>          return HRESULT_FROM_WIN32( GetLastError() ); @@ -306,7
>>          +306,7 @@ return E_NOTIMPL;
>>
>>      H hIbatDev = CreateFileW( IBAT_WIN32_NAME,
>> -        MAXIMUM_ALLOWED, 0, NULL,
>> +        MAXIMUM_ALLOWED,(FILE_SHARE_READ|FILE_SHARE_WRITE), NULL,
>>          OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL );
>>      if( hIbatDev == INVALID_HANDLE_VALUE )
>>          return HRESULT_FROM_WIN32( GetLastError() );