diff mbox series

[v2,2/3] powerpc: use NMI IPI for smp_send_stop

Message ID 20171023080507.21974-3-npiggin@gmail.com (mailing list archive)
State Superseded
Headers show
Series avoid secondary hold spinloop when possible | expand

Commit Message

Nicholas Piggin Oct. 23, 2017, 8:05 a.m. UTC 
Use the NMI IPI rather than smp_call_function for smp_send_stop.
Have stopped CPUs hard disable interrupts rather than just soft
disable.

This function is used in crash/panic/shutdown paths to bring other
CPUs down as quickly and reliably as possible, and minimizing their
potential to cause trouble.

Avoiding the Linux smp_call_function infrastructure and (if supported)
using true NMI IPIs makes this more robust.

Also use spin loop primitives in the stop callback, mainly to help
processing speed of the active thread speed in the simulator.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/kernel/smp.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

Comments

kernel test robot Oct. 23, 2017, 9:05 p.m. UTC | #1 
Hi Nicholas,

[auto build test ERROR on powerpc/next]
[also build test ERROR on v4.14-rc6 next-20171018]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Nicholas-Piggin/avoid-secondary-hold-spinloop-when-possible/20171023-173012
base:   https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git next
config: powerpc-ge_imp3a_defconfig (attached as .config)
compiler: powerpc-linux-gnu-gcc (Debian 6.1.1-9) 6.1.1 20160705
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=powerpc 

All errors (new ones prefixed by >>):

   arch/powerpc/kernel/smp.o: In function `smp_send_stop':
>> smp.c:(.text+0x774): undefined reference to `smp_send_nmi_ipi'

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation
kernel test robot Oct. 23, 2017, 11:15 p.m. UTC | #2 
Hi Nicholas,

[auto build test ERROR on powerpc/next]
[also build test ERROR on v4.14-rc6 next-20171018]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Nicholas-Piggin/avoid-secondary-hold-spinloop-when-possible/20171023-173012
base:   https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git next
config: powerpc-iss476-smp_defconfig (attached as .config)
compiler: powerpc-linux-gnu-gcc (Debian 6.1.1-9) 6.1.1 20160705
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=powerpc 

All errors (new ones prefixed by >>):

   arch/powerpc/kernel/smp.o: In function `smp_send_stop':
>> arch/powerpc/kernel/smp.c:563: undefined reference to `smp_send_nmi_ipi'

vim +563 arch/powerpc/kernel/smp.c

   560	
   561	void smp_send_stop(void)
   562	{
 > 563		smp_send_nmi_ipi(NMI_IPI_ALL_OTHERS, stop_this_cpu, 1000000);
   564	}
   565	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation
Nicholas Piggin Oct. 31, 2017, 7:43 a.m. UTC | #3 
On Tue, 24 Oct 2017 07:15:30 +0800
kbuild test robot <lkp@intel.com> wrote:

> Hi Nicholas,
> 
> [auto build test ERROR on powerpc/next]
> [also build test ERROR on v4.14-rc6 next-20171018]
> [if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
> 
> url:    https://github.com/0day-ci/linux/commits/Nicholas-Piggin/avoid-secondary-hold-spinloop-when-possible/20171023-173012
> base:   https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git next
> config: powerpc-iss476-smp_defconfig (attached as .config)
> compiler: powerpc-linux-gnu-gcc (Debian 6.1.1-9) 6.1.1 20160705
> reproduce:
>         wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
>         chmod +x ~/bin/make.cross
>         # save the attached .config to linux build tree
>         make.cross ARCH=powerpc 
> 
> All errors (new ones prefixed by >>):
> 
>    arch/powerpc/kernel/smp.o: In function `smp_send_stop':
> >> arch/powerpc/kernel/smp.c:563: undefined reference to `smp_send_nmi_ipi'  
> 
> vim +563 arch/powerpc/kernel/smp.c
> 
>    560	
>    561	void smp_send_stop(void)
>    562	{
>  > 563		smp_send_nmi_ipi(NMI_IPI_ALL_OTHERS, stop_this_cpu, 1000000);  
>    564	}
>    565	

Good catch, 0day. Thanks.

Rather than have all architectures select NMI_IPI just for this, I
think I'll make it depend on whether NMI_IPI was selected at all.
Otherwise just keep using the normal smp_call_function().

Thanks,
Nick
diff mbox series

Patch

diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index e0a4c1f82e25..ce891030d925 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -547,19 +547,20 @@  void crash_send_ipi(void (*crash_ipi_callback)(struct pt_regs *))
 }
 #endif
 
-static void stop_this_cpu(void *dummy)
+static void __noreturn stop_this_cpu(struct pt_regs *regs)
 {
 	/* Remove this CPU */
 	set_cpu_online(smp_processor_id(), false);
 
-	local_irq_disable();
+	hard_irq_disable();
+	spin_begin();
 	while (1)
-		;
+		spin_cpu_relax();
 }
 
 void smp_send_stop(void)
 {
-	smp_call_function(stop_this_cpu, NULL, 0);
+	smp_send_nmi_ipi(NMI_IPI_ALL_OTHERS, stop_this_cpu, 1000000);
 }
 
 struct thread_info *current_set[NR_CPUS];