Re: [PATCH -mm 0/3] sysv ipc: increase msgmnb with the number of cpus

[Nadia Derbey <Nadia.Derbey@bull.net>]

Solofo.Ramangalahy@bull.net wrote:
> Andrew Morton writes:
>  > I'm afraid I've lost track of what's happening here.  Did we come up
>  > with an alternative to "magical positive-versus-negative number trick"?
> 
> Yes, several proposals:
> 1. /proc/sys/kernel/automatic-msgmnb
>    (attempt attached).
> 
> 2. a variation /sys/kernel/automatic/msgmn*
>    as a mean to alleviate the doubling of the interface.
> 
>  > Your patch #1 adds and uses recompute_msgmnb() without adding the
>  > declaration to a header file.  Your patch #2 does add the
>  > recompute_msgmnb() to a header file, so we have a window in which the
>  > build is broken, which is bad.
> 
> Sorry. Another fix needed to my quilt workflow.
> Thanks.
> 
>  > recompute_msgmnb() isn't a terribly good globally-visible identifier,
>  > btw.  It is nice to add some subsystem identifer as a prefix.  There's
>  > little chance of this symbol colliding with anything else, so this is a
>  > minor cosmetic thing in this case.
> 
> ok. msg_recompute_msgmnb seems to be the better name.
> 


Solofo,

I had a look at your patch, and noticed that it is mixing msgmni and 
msgmnb stuff.
Since the msgmni part is already in mainline, I think we need first to 
fix the interface issue for msgmni: I've got a patch ready.

And your patch, when reviwed, can then come on top of it.

Sending the patch right now.

Regards,
Nadia