diff --git a/Documentation/atomic_ops.txt b/Documentation/atomic_ops.txt
index b19fc34efdb17921af43bda0000b13dc82640451..c9d1cacb439590a02dbfb1282cbe0d3abe3a9b0d 100644
--- a/Documentation/atomic_ops.txt
+++ b/Documentation/atomic_ops.txt
@@ -542,6 +542,10 @@ The routines xchg() and cmpxchg() must provide the same exact
 memory-barrier semantics as the atomic and bit operations returning
 values.
 
+Note: If someone wants to use xchg(), cmpxchg() and their variants,
+linux/atomic.h should be included rather than asm/cmpxchg.h, unless
+the code is in arch/* and can take care of itself.
+
 Spinlocks and rwlocks have memory barrier expectations as well.
 The rule to follow is simple:
 
diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
index 98d172b04f71815a1105b304ea5cadfa552687a2..a9b9460de0d63b2369062ce1fc9dee90909878ee 100644
--- a/drivers/net/ethernet/sfc/mcdi.c
+++ b/drivers/net/ethernet/sfc/mcdi.c
@@ -9,7 +9,7 @@
 
 #include <linux/delay.h>
 #include <linux/moduleparam.h>
-#include <asm/cmpxchg.h>
+#include <linux/atomic.h>
 #include "net_driver.h"
 #include "nic.h"
 #include "io.h"
diff --git a/drivers/phy/phy-rcar-gen2.c b/drivers/phy/phy-rcar-gen2.c
index 6e0d9fa8e1d13f8d1ad0eec73b1040f594637df7..c7a05996d5c1a841a65e515e746d1008d3d47489 100644
--- a/drivers/phy/phy-rcar-gen2.c
+++ b/drivers/phy/phy-rcar-gen2.c
@@ -17,8 +17,7 @@
 #include <linux/phy/phy.h>
 #include <linux/platform_device.h>
 #include <linux/spinlock.h>
-
-#include <asm/cmpxchg.h>
+#include <linux/atomic.h>
 
 #define USBHS_LPSTS			0x02
 #define USBHS_UGCTRL			0x80
diff --git a/drivers/staging/speakup/selection.c b/drivers/staging/speakup/selection.c
index 98af3b1f2d2a13efea585a11429bc00a48451582..aa5ab6c80ed4ffefc12aaca9c1337e403371d478 100644
--- a/drivers/staging/speakup/selection.c
+++ b/drivers/staging/speakup/selection.c
@@ -7,7 +7,7 @@
 #include <linux/workqueue.h>
 #include <linux/tty.h>
 #include <linux/tty_flip.h>
-#include <asm/cmpxchg.h>
+#include <linux/atomic.h>
 
 #include "speakup.h"