Re: [BUG 4.2-rc8] Interrupt occurs while apply_alternatives() is patching the handler

[Thomas Gleixner <tglx@linutronix.de>]

On Thu, 3 Sep 2015, Borislav Petkov wrote:
> On Wed, Sep 02, 2015 at 11:11:55AM +0200, Thomas Gleixner wrote:
> >  static void __init_or_module add_nops(void *insns, unsigned int len)
> >  {
> > +	unsigned long flags;
> > +
> > +	local_irq_save(flags);
> >  	while (len > 0) {
> 
> I guess you want to optimize the len==0 case to not disable interrupts
> needlessly:
> 
> 	if (!len)
> 		return;
> 
> 	local_irq_save(flags);
> 	while (len > 0)
> 		...

Nah. I rather put the local_irq_save into optimize_nops(). All other
callers of add_nops() are operating on a buffer and use text_poke
after that. Aside of that optimize_nops() is missing a sync_core().

Updated patch below.

Thanks,

	tglx

----------------->
Subject: x86/alternatives: Make optimize_nops() interrupt safe and synced
From: Thomas Gleixner <tglx@linutronix.de>
Date: Thu, 03 Sep 2015 12:34:55 +0200

optimize_nops() is buggy in two aspects:

- It's not disabling interrupts across the modification
- It's lacking a sync_core() call

Fixes: 4fd4b6e5537c 'x86/alternatives: Use optimized NOPs for padding'
Reported-by: "Richard W.M. Jones" <rjones@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/kernel/alternative.c |    5 +++++
 1 file changed, 5 insertions(+)

Index: tip/arch/x86/kernel/alternative.c
===================================================================
--- tip.orig/arch/x86/kernel/alternative.c
+++ tip/arch/x86/kernel/alternative.c
@@ -338,10 +338,15 @@ done:
 
 static void __init_or_module optimize_nops(struct alt_instr *a, u8 *instr)
 {
+	unsigned long flags;
+
 	if (instr[0] != 0x90)
 		return;
 
+	local_irq_save(flags);
 	add_nops(instr + (a->instrlen - a->padlen), a->padlen);
+	sync_core();
+	local_irq_restore(flags);
 
 	DUMP_BYTES(instr, a->instrlen, "%p: [%d:%d) optimized NOPs: ",
 		   instr, a->instrlen - a->padlen, a->padlen);