Re: [PATCH] x86/paravirt: Add missing noinstr to arch_local*() helpers

From: Peter Zijlstra <peterz@infradead.org>
To: "Jürgen Groß" <jgross@suse.com>
Cc: syzbot <syzbot+8db9e1ecde74e590a657@syzkaller.appspotmail.com>,
	fenghua.yu@intel.com, Marco Elver <elver@google.com>,
	"Luck, Tony" <tony.luck@intel.com>,
	"Paul E. McKenney" <paulmck@kernel.org>,
	yu-cheng.yu@intel.com, Dave Hansen <dave.hansen@linux.intel.com>,
	syzkaller-bugs <syzkaller-bugs@googlegroups.com>,
	LKML <linux-kernel@vger.kernel.org>,
	kasan-dev <kasan-dev@googlegroups.com>,
	virtualization@lists.linux-foundation.org,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	"H. Peter Anvin" <hpa@zytor.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Wei Liu <wei.liu@kernel.org>,
	the arch/x86 maintainers <x86@kernel.org>
Subject: Re: [PATCH] x86/paravirt: Add missing noinstr to arch_local*() helpers
Date: Tue, 11 Aug 2020 10:12:05 +0200	[thread overview]
Message-ID: <20200811081205.GV3982@worktop.programming.kicks-ass.net> (raw)
In-Reply-To: <a2dffeeb-04f0-8042-b39a-b839c4800d6f@suse.com>

On Tue, Aug 11, 2020 at 09:57:55AM +0200, Jürgen Groß wrote:
> On 11.08.20 09:41, Peter Zijlstra wrote:
> > On Fri, Aug 07, 2020 at 05:19:03PM +0200, Marco Elver wrote:
> > 
> > > My hypothesis here is simply that kvm_wait() may be called in a place
> > > where we get the same case I mentioned to Peter,
> > > 
> > > 	raw_local_irq_save(); /* or other IRQs off without tracing */
> > > 	...
> > > 	kvm_wait() /* IRQ state tracing gets confused */
> > > 	...
> > > 	raw_local_irq_restore();
> > > 
> > > and therefore, using raw variants in kvm_wait() works. It's also safe
> > > because it doesn't call any other libraries that would result in corrupt
> > 
> > Yes, this is definitely an issue.
> > 
> > Tracing, we also musn't call into tracing when using raw_local_irq_*().
> > Because then we re-intoduce this same issue all over again.
> > 
> > Both halt() and safe_halt() are more paravirt calls, but given we're in
> > a KVM paravirt call already, I suppose we can directly use native_*()
> > here.
> > 
> > Something like so then... I suppose, but then the Xen variants need TLC
> > too.
> 
> Just to be sure I understand you correct:
> 
> You mean that xen_qlock_kick() and xen_qlock_wait() and all functions
> called by those should gain the "notrace" attribute, right?
> 
> I am not sure why the kick variants need it, though. IMO those are
> called only after the lock has been released, so they should be fine
> without notrace.

The issue happens when someone uses arch_spinlock_t under
raw_local_irq_*().

> And again: we shouldn't forget the Hyper-V variants.

Bah, my grep failed :/ Also *groan*, that's calling apic->send_IPI().

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization