Re: [RFC][PATCH] binfmt: turn MAX_ARG_PAGES into a sysctl tunable

From: Peter Zijlstra
Date: Sat Jun 17 2006 - 02:41:39 EST


On Fri, 2006-06-16 at 20:40 +0000, Pavel Machek wrote:
> Hi!
>
> > From: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
> >
> > Some folks find 128KB of env+arg space too little. Solaris provides them with
> > 1MB. Manually changing MAX_ARG_PAGES worked for them so far, however they
> > would like to run the supported vendor kernel.
> >
> > In the interrest of not penalizing everybody with the overhead of just
> > setting it larger, provide a sysctl to change it.
>
> I very muh like that idea.
>
> > Compiles and boots on i386.
> >
> > Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
>
> ....but does is also work if you keep changing that value in the tight
> loop while trying to work with the system?
>

Yes, I've added a bprm local copy of max_arg_pages:

> @@ -20,9 +13,10 @@ struct pt_regs;
> > /*
> > * This structure is used to hold the arguments that are used when loading binaries.
> > */
> > -struct linux_binprm{
> > +struct linux_binprm {
> > char buf[BINPRM_BUF_SIZE];
> > - struct page *page[MAX_ARG_PAGES];
> > + struct page **page;
> > + int max_arg_pages;
> > struct mm_struct *mm;
> > unsigned long p; /* current top of mem */
> > int sh_bang;

That is set once from the sysctl variable:

> > @@ -1153,14 +1164,20 @@ int do_execve(char * filename,
> > if (!bprm)
> > goto out_ret;
> >
> > + bprm->max_arg_pages = max_arg_pages;
> > + bprm->page = kzalloc(bprm->max_arg_pages*sizeof(struct page*),
> > + GFP_KERNEL);
> > + if (!bprm->page)
> > + goto out_kfree;
> > +

And is thereafter used and never again changed.

This should be enough to guard against your scenario I recon.

Peter


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/