Re: [Patch 3/4] ipc/mqueue: strengthen checks on mqueue creation

[Doug Ledford <dledford@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1552 bytes --]

On 5/1/2012 4:18 PM, KOSAKI Motohiro wrote:
>>> But ENOMEM is more inaccurate. It almostly is used for kmalloc failure.
>>
>> I chose ENOMEM for that particular error because above there we have
>> checked the passed in arguments to make sure that they don't violate our
>> allowances for max message or max message size.  If we violate either of
>> those items, we return EINVAL.  In this case, neither of the values is
>> invalid, it's just that together they make an overly large allocation.
>> I would see that as more helpful to a programmer than EINVAL when the
>> values are within the maximums allowed.  At least with ENOMEM the
>> programmer knows they have to reduce their combined message size and
>> message count in order to get things working.
> 
> Incorrect. When ENOMEM is returned, programmers can't know
> which problem was happen 1) kernel has real memory starvation
> or 2) queue limitation exceed was happen. The problem is, you
> introduced new overloaded error code for avoiding overload error code.
> It doesn't make sense. My question was, why can't you choose no
> overload error code if you want accurate one?

OK, then would EOVERFLOW suit things better?

All this reminds me that when this is taken into Linus' kernel, we need
to coordinate a man page update for the mq subsystem.


-- 
Doug Ledford <dledford@redhat.com>
              GPG KeyID: 0E572FDD
	      http://people.redhat.com/dledford

Infiniband specific RPMs available at
	      http://people.redhat.com/dledford/Infiniband


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 898 bytes --]