Re: [PATCH 2.5.71-mm1] aio process hang on EINVAL

[jgmyers@netscape.com (John Myers)]

[-- Attachment #1: Type: text/plain, Size: 793 bytes --]

Joel Becker wrote:

>POSIX 1003.1 says this about aio_read() and aio_write():
>	If an error condition is encountered during queuing, the function
>	call shall return without having initiated or queued the request.
>
>If you intend to ever allow a POSIX wrapper to these interfaces (I have
>one, for instance), you need to return EINVAL, EBADF, and the like from
>io_submit().
>
No, you just declare that those errors happend "after queuing."

>A user has to handle EAGAIN, so io_submit() cannot return void, and you
>already have error handling logic here.
>  
>
EAGAIN error handling does not require contextual information about the 
operation being queued.  Error handling logic that knows about the 
context of the operation queued already has to exist in the 
io_getevents() processing.


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3941 bytes --]