linux-kernel - Re: [PATCH] x86/smp: Add irq_enter/exit() in smp_reschedule

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 17 Oct 2016 10:22:50 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Wanpeng Li <kernellwp@...il.com>
Cc:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Wanpeng Li <wanpeng.li@...mail.com>,
        Ingo Molnar <mingo@...nel.org>, Mike Galbraith <efault@....de>,
        Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH] x86/smp: Add irq_enter/exit() in
 smp_reschedule_interrupt()

On Mon, Oct 17, 2016 at 12:19:43PM +0800, Wanpeng Li wrote:
> 2016-10-16 21:39 GMT+08:00 Peter Zijlstra <peterz@...radead.org>:

> >>   [<ffffffff9d492b95>] do_trace_write_msr+0x135/0x140
> >>   [<ffffffff9d06f860>] native_write_msr+0x20/0x30
> >>   [<ffffffff9d065fad>] native_apic_msr_eoi_write+0x1d/0x30
> >>   [<ffffffff9d05bd1d>] smp_reschedule_interrupt+0x1d/0x30
> >>   [<ffffffff9d8daec6>] reschedule_interrupt+0x96/0xa0

> >>  __visible void smp_reschedule_interrupt(struct pt_regs *regs)
> >>  {
> >> +     irq_enter();
> >>       ack_APIC_irq();
> >>       __smp_reschedule_interrupt();
> >> +     irq_exit();
> >
> > Urgh, I really hate this...
> >
> > So now we're making a very frequent interrupt slower because of debug
> > code :/
> 
> Do you have a better idea? :)

Something like the below avoids all that. Paravirt will still need fixing.

The thing is, many many smp_reschedule_interrupt() invocations don't
actually execute anything much at all and are only send to tickle the
return to user path (which does the actual preemption).

Having to do the whole irq_enter/irq_exit dance just for this unlikely
debug case totally blows.

---
 arch/x86/include/asm/apic.h |  2 +-
 arch/x86/include/asm/msr.h  | 15 +++++++++++++++
 2 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/apic.h b/arch/x86/include/asm/apic.h
index f5aaf6c83222..b97bfeed6456 100644
--- a/arch/x86/include/asm/apic.h
+++ b/arch/x86/include/asm/apic.h
@@ -196,7 +196,7 @@ static inline void native_apic_msr_write(u32 reg, u32 v)
 
 static inline void native_apic_msr_eoi_write(u32 reg, u32 v)
 {
-	wrmsr(APIC_BASE_MSR + (APIC_EOI >> 4), APIC_EOI_ACK, 0);
+	wrmsr_notrace(APIC_BASE_MSR + (APIC_EOI >> 4), APIC_EOI_ACK, 0);
 }
 
 static inline u32 native_apic_msr_read(u32 reg)
diff --git a/arch/x86/include/asm/msr.h b/arch/x86/include/asm/msr.h
index b5fee97813cd..45c080449d5b 100644
--- a/arch/x86/include/asm/msr.h
+++ b/arch/x86/include/asm/msr.h
@@ -127,6 +127,16 @@ notrace static inline void native_write_msr(unsigned int msr,
 }
 
 /* Can be uninlined because referenced by paravirt */
+notrace static inline void native_write_msr_notrace(unsigned int msr,
+					    unsigned low, unsigned high)
+{
+	asm volatile("1: wrmsr\n"
+		     "2:\n"
+		     _ASM_EXTABLE_HANDLE(1b, 2b, ex_handler_wrmsr_unsafe)
+		     : : "c" (msr), "a"(low), "d" (high) : "memory");
+}
+
+/* Can be uninlined because referenced by paravirt */
 notrace static inline int native_write_msr_safe(unsigned int msr,
 					unsigned low, unsigned high)
 {
@@ -228,6 +238,11 @@ static inline void wrmsr(unsigned msr, unsigned low, unsigned high)
 	native_write_msr(msr, low, high);
 }
 
+static inline void wrmsr_notrace(unsigned msr, unsigned low, unsigned high)
+{
+	native_write_msr_notrace(msr, low, high);
+}
+
 #define rdmsrl(msr, val)			\
 	((val) = native_read_msr((msr)))