From: Jonathan Sambrook (jonathan.sambrook_at_dsvr.co.uk)
Date: Wed 12 Mar 2003 - 16:04:19 GMT
At 17:05 on Wed 12/03/03, sam_at_vilain.net masquerading as 'Sam Vilain' wrote:
> On Wed, 12 Mar 2003 12:00, Jacques Gelinas wrote:
> > > If you've got a setup/operating environment whichtriggers the oops,
> > > you might like to test my hypothesis by applying the attached patch
> > > after applying ctx16.
> >
> > ip_info is a pointer. While the content may change between the two
> > lines, the pointer itself can't, so your patch does nothing.
>
> Is there any reason that the sys_release_ip_info/sys_assign_ip_info pair
> don't take task_struct pointers like sys_release_s_info/sys_assign_s_info?
> This would allow you to put the semaphore in a more meaningful place.
struct sock's ip_info member
> Putting on my System Engineer's hat, I'd also humbly suggest that the
> sys_alloc_*_info functions explicitly take a pointer rather than relying
> on `current'. If nothing else, for consistency & because it doesn't lose
> you any execution speed. Globals suck, even if you think it's never going
> to matter.
> --
> Sam Vilain, sam_at_vilain.net
>
> "I like a man who grins when he fights."
> - Winston Churchill -
--
Jonathan Sambrook
Software Developer
Designer Servers