Re: [PATCH] X86: reboot-notify additions

[Ingo Molnar <mingo@elte.hu>]

* Cliff Wickman <cpw@sgi.com> wrote:

> > > +#define SYS_INSANE	0x0004	/* Notify of system error/panic/oops */
> > > +/* For the SYS_INSANE case, no locks should be taken by the called-back
> > > + * function.  The kernel is ready for an immediate reboot.
> > > + */

please use standard comment blocks, something like this:

  /*
   * Notify of system error/panic/oops
   *
   * No locks should be taken by the called-back function.
   * The kernel is ready for an immediate reboot.
   */
  #define SYS_INSANE	0x0004

> > > +++ linux/kernel/kexec.c
> > > @@ -1001,6 +1001,9 @@ asmlinkage long sys_kexec_load(unsigned 
> > >  		if (result)
> > >  			goto out;
> > >  	}
> > > +
> > > +	blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);

this puts an extra blocking call into a kexec critical path.

>  			struct pt_regs fixed_regs;
> +        		struct raw_notifier_head rh;
> +        		rh.head = reboot_notifier_list.head;
> +        		raw_notifier_call_chain(&rh, SYS_EMERGENCY, NULL);

that's very nasty. Now callbacks which put themselves on a blocking list 
will be called without locking.

>  void emergency_restart(void)
>  {
> +	struct raw_notifier_head rh;
> +
> +	rh.head = reboot_notifier_list.head;
> +	raw_notifier_call_chain(&rh, SYS_EMERGENCY, NULL);

ditto.

this patch is still far from being acceptable.

	Ingo