[RFC] [PATCH] Make lio_listio set errno to EIO on requests with invalid aio_lio_opcode

[Suzuki]

Roland McGrath wrote:
> There is no obligation to check for this kind of error, given the standard.
According to the standards, lio_listio should return EIO with invalid 
opcodes.

> At any rate, the EINVAL will indeed be diagnosed when the request is
> examined in the service thread, as I read the code. 

Yes, the service thread does check it and sets EINVAL for the iocb, not 
for lio_listio(). But, the lio_listio() would succeed, without returning 
an error. It should have returned EIO, if it has some requests with 
invalid opcode ( or opcodes other than LIO_{READ,WRITE,NOP}.

> If you think this is
> not happening, please supply a test program to demonstrate the problem.
> 
> 
> Thanks,
> Roland