[v3.16.y] percpu: make this_cpu_generic_read() atomic w.r.t. interrupts

Message ID 20171018134022.25282-1-mark.rutland@arm.com
State New
Headers show
Series
  • [v3.16.y] percpu: make this_cpu_generic_read() atomic w.r.t. interrupts
Related show

Commit Message

Mark Rutland Oct. 18, 2017, 1:40 p.m.
Commit e88d62cd4b2f0b1ae55e9008e79c2794b1fc914d upstream.

As raw_cpu_generic_read() is a plain read from a raw_cpu_ptr() address,
it's possible (albeit unlikely) that the compiler will split the access
across multiple instructions.

In this_cpu_generic_read() we disable preemption but not interrupts
before calling raw_cpu_generic_read(). Thus, an interrupt could be taken
in the middle of the split load instructions. If a this_cpu_write() or
RMW this_cpu_*() op is made to the same variable in the interrupt
handling path, this_cpu_read() will return a torn value.

For native word types, we can avoid tearing using READ_ONCE(), but this
won't work in all cases (e.g. 64-bit types on most 32-bit platforms).
This patch reworks this_cpu_generic_read() to use READ_ONCE() where
possible, otherwise falling back to disabling interrupts.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Christoph Lameter <cl@linux.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Pranith Kumar <bobby.prani@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-arch@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>

[Mark: backport to v3.16.y]
Signed-off-by: Mark Rutland <mark.rutland@arm.com>

---
 include/linux/percpu.h | 29 +++++++++++++++++++++++++----
 1 file changed, 25 insertions(+), 4 deletions(-)

-- 
2.11.0

Comments

Greg KH Oct. 19, 2017, 9:24 a.m. | #1
On Wed, Oct 18, 2017 at 02:40:22PM +0100, Mark Rutland wrote:
> Commit e88d62cd4b2f0b1ae55e9008e79c2794b1fc914d upstream.

> 

> As raw_cpu_generic_read() is a plain read from a raw_cpu_ptr() address,

> it's possible (albeit unlikely) that the compiler will split the access

> across multiple instructions.

> 

> In this_cpu_generic_read() we disable preemption but not interrupts

> before calling raw_cpu_generic_read(). Thus, an interrupt could be taken

> in the middle of the split load instructions. If a this_cpu_write() or

> RMW this_cpu_*() op is made to the same variable in the interrupt

> handling path, this_cpu_read() will return a torn value.

> 

> For native word types, we can avoid tearing using READ_ONCE(), but this

> won't work in all cases (e.g. 64-bit types on most 32-bit platforms).

> This patch reworks this_cpu_generic_read() to use READ_ONCE() where

> possible, otherwise falling back to disabling interrupts.

> 

> Signed-off-by: Mark Rutland <mark.rutland@arm.com>

> Cc: Arnd Bergmann <arnd@arndb.de>

> Cc: Christoph Lameter <cl@linux.com>

> Cc: Peter Zijlstra <peterz@infradead.org>

> Cc: Pranith Kumar <bobby.prani@gmail.com>

> Cc: Tejun Heo <tj@kernel.org>

> Cc: Thomas Gleixner <tglx@linutronix.de>

> Cc: linux-arch@vger.kernel.org

> Cc: stable@vger.kernel.org

> Signed-off-by: Tejun Heo <tj@kernel.org>

> [Mark: backport to v3.16.y]


You sent 2 patches for 3.16.y, did you mean for one of these to be for
3.18.y?

thanks,

greg k-h
Mark Rutland Oct. 30, 2017, 2:40 p.m. | #2
On Thu, Oct 19, 2017 at 11:24:34AM +0200, Greg KH wrote:
> You sent 2 patches for 3.16.y, did you mean for one of these to be for

> 3.18.y?


Sorry; I got a bit muddled and thought I hadn't sent the first copy.
Please ignore the duplicate!

I'd ignored 3.18.y as EOL, and neither was intended for 3.18.y.

Sorry for the noise!

Mark.

Patch

diff --git a/include/linux/percpu.h b/include/linux/percpu.h
index 8419053d0f2e..37ff713a7c7f 100644
--- a/include/linux/percpu.h
+++ b/include/linux/percpu.h
@@ -276,12 +276,33 @@  do {									\
  *    used.
  */
 
-#define _this_cpu_generic_read(pcp)					\
-({	typeof(pcp) ret__;						\
+#define __this_cpu_generic_read_nopreempt(pcp)				\
+({									\
+	typeof(pcp) __ret;						\
 	preempt_disable();						\
-	ret__ = *this_cpu_ptr(&(pcp));					\
+	__ret = ACCESS_ONCE(*raw_cpu_ptr(&(pcp)));			\
 	preempt_enable();						\
-	ret__;								\
+	__ret;								\
+})
+
+#define __this_cpu_generic_read_noirq(pcp)				\
+({									\
+	typeof(pcp) __ret;						\
+	unsigned long __flags;						\
+	raw_local_irq_save(__flags);					\
+	__ret = *raw_cpu_ptr(&(pcp)); 					\
+	raw_local_irq_restore(__flags);					\
+	__ret;								\
+})
+
+#define _this_cpu_generic_read(pcp)					\
+({									\
+	typeof(pcp) __ret;						\
+	if (__native_word(pcp))						\
+		__ret = __this_cpu_generic_read_nopreempt(pcp);		\
+	else								\
+		__ret = __this_cpu_generic_read_noirq(pcp);		\
+	__ret;								\
 })
 
 #ifndef this_cpu_read