[PATCH v2 0/3] io.h, logic_pio: Allow barriers for inX() and outX() be overridden

[PATCH v2 0/3] io.h, logic_pio: Allow barriers for inX() and outX() be overridden

[John Garry]

Since commits a7851aa54c0c ("io: change outX() to have their own IO
barrier overrides") and 87fe2d543f81 ("io: change inX() to have their own
IO barrier overrides"), the outX() and inX() functions have memory
barriers which can be overridden per-arch.

However, under CONFIG_INDIRECT_PIO, logic_pio defines its own version of
inX() and outX(), which still use readb et al. For these, the barrier
after a raw read is weaker than it otherwise would be. 

This series generates consistent behaviour for logic_pio, by defining
generic _inX() and _outX() in asm-generic/io.h, and using those in
logic_pio. Generic _inX() and _outX() have per-arch overrideable
barriers.

The topic was discussed there originally:
https://lore.kernel.org/lkml/2e80d7bc-32a0-cc40-00a9-8a383a1966c2@huawei.com/

A small tidy-up patch is included.

I hope that series can go through the arm-soc tree, as with other recent
logic_pio changes.

Hi Arnd,

I added your tag, but please let me know if you have any issue with the
updated change in patch #1.

cheers

- Differences to v1
	- fix x86 clang build by adding extra build swicth for _{in,out}X
	- added Arnd's RB tag

John Garry (3):
  io: Provide _inX() and _outX()
  logic_pio: Improve macro argument name
  logic_pio: Use _inX() and _outX()

 include/asm-generic/io.h | 64 +++++++++++++++++++++++++++++++++---------------
 lib/logic_pio.c          | 22 ++++++++---------
 2 files changed, 55 insertions(+), 31 deletions(-)

-- 
2.16.4


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 1/3] io: Provide _inX() and _outX()

[John Garry]

Since commit a7851aa54c0c ("io: change outX() to have their own IO barrier
overrides") and commit 87fe2d543f81 ("io: change inX() to have their own
IO barrier overrides"), the outX and inX functions have memory barriers
which can be overridden.

However, the generic logic_pio lib has continued to use readl/writel et al
for IO port accesses, which has weaker barriers on arm64.

Provide generic _inX() and _outX(), which can be used by logic pio.

For consistency, we check for !defined({in,out}X) && !defined(_{in,out}X),
for defining _{in,out}X, while a check for just !defined({in,out}X) should
suffice.

Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: John Garry <john.garry@huawei.com>
---
 include/asm-generic/io.h | 64 +++++++++++++++++++++++++++++++++---------------
 1 file changed, 44 insertions(+), 20 deletions(-)

diff --git a/include/asm-generic/io.h b/include/asm-generic/io.h
index d39ac997dda8..3a7871130112 100644
--- a/include/asm-generic/io.h
+++ b/include/asm-generic/io.h
@@ -448,17 +448,15 @@ static inline void writesq(volatile void __iomem *addr, const void *buffer,
 #define IO_SPACE_LIMIT 0xffff
 #endif
 
-#include <linux/logic_pio.h>
-
 /*
  * {in,out}{b,w,l}() access little endian I/O. {in,out}{b,w,l}_p() can be
  * implemented on hardware that needs an additional delay for I/O accesses to
  * take effect.
  */
 
-#ifndef inb
-#define inb inb
-static inline u8 inb(unsigned long addr)
+#if !defined(inb) && !defined(_inb)
+#define _inb _inb
+static inline u16 _inb(unsigned long addr)
 {
 	u8 val;
 
@@ -469,9 +467,9 @@ static inline u8 inb(unsigned long addr)
 }
 #endif
 
-#ifndef inw
-#define inw inw
-static inline u16 inw(unsigned long addr)
+#if !defined(inw) && !defined(_inw)
+#define _inw _inw
+static inline u16 _inw(unsigned long addr)
 {
 	u16 val;
 
@@ -482,9 +480,9 @@ static inline u16 inw(unsigned long addr)
 }
 #endif
 
-#ifndef inl
-#define inl inl
-static inline u32 inl(unsigned long addr)
+#if !defined(inl) && !defined(_inl)
+#define _inl _inl
+static inline u16 _inl(unsigned long addr)
 {
 	u32 val;
 
@@ -495,9 +493,9 @@ static inline u32 inl(unsigned long addr)
 }
 #endif
 
-#ifndef outb
-#define outb outb
-static inline void outb(u8 value, unsigned long addr)
+#if !defined(outb) && !defined(_outb)
+#define _outb _outb
+static inline void _outb(u8 value, unsigned long addr)
 {
 	__io_pbw();
 	__raw_writeb(value, PCI_IOBASE + addr);
@@ -505,9 +503,9 @@ static inline void outb(u8 value, unsigned long addr)
 }
 #endif
 
-#ifndef outw
-#define outw outw
-static inline void outw(u16 value, unsigned long addr)
+#if !defined(outw) && !defined(_outw)
+#define _outw _outw
+static inline void _outw(u16 value, unsigned long addr)
 {
 	__io_pbw();
 	__raw_writew(cpu_to_le16(value), PCI_IOBASE + addr);
@@ -515,9 +513,9 @@ static inline void outw(u16 value, unsigned long addr)
 }
 #endif
 
-#ifndef outl
-#define outl outl
-static inline void outl(u32 value, unsigned long addr)
+#if !defined(outl) && !defined(_outl)
+#define _outl _outl
+static inline void _outl(u32 value, unsigned long addr)
 {
 	__io_pbw();
 	__raw_writel(cpu_to_le32(value), PCI_IOBASE + addr);
@@ -525,6 +523,32 @@ static inline void outl(u32 value, unsigned long addr)
 }
 #endif
 
+#include <linux/logic_pio.h>
+
+#ifndef inb
+#define inb _inb
+#endif
+
+#ifndef inw
+#define inw _inw
+#endif
+
+#ifndef inl
+#define inl _inl
+#endif
+
+#ifndef outb
+#define outb _outb
+#endif
+
+#ifndef outw
+#define outw _outw
+#endif
+
+#ifndef outl
+#define outl _outl
+#endif
+
 #ifndef inb_p
 #define inb_p inb_p
 static inline u8 inb_p(unsigned long addr)
-- 
2.16.4


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 2/3] logic_pio: Improve macro argument name

[John Garry]

Macro argument "bw" is used for building byte, word, and long-based
functions. Use "bwl" instead, to include long.

Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: John Garry <john.garry@huawei.com>
---
 lib/logic_pio.c | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

diff --git a/lib/logic_pio.c b/lib/logic_pio.c
index f511a99bb389..21dc731bec88 100644
--- a/lib/logic_pio.c
+++ b/lib/logic_pio.c
@@ -229,13 +229,13 @@ unsigned long logic_pio_trans_cpuaddr(resource_size_t addr)
 }
 
 #if defined(CONFIG_INDIRECT_PIO) && defined(PCI_IOBASE)
-#define BUILD_LOGIC_IO(bw, type)					\
-type logic_in##bw(unsigned long addr)					\
+#define BUILD_LOGIC_IO(bwl, type)					\
+type logic_in##bwl(unsigned long addr)					\
 {									\
 	type ret = (type)~0;						\
 									\
 	if (addr < MMIO_UPPER_LIMIT) {					\
-		ret = read##bw(PCI_IOBASE + addr);			\
+		ret = read##bwl(PCI_IOBASE + addr);			\
 	} else if (addr >= MMIO_UPPER_LIMIT && addr < IO_SPACE_LIMIT) { \
 		struct logic_pio_hwaddr *entry = find_io_range(addr);	\
 									\
@@ -248,10 +248,10 @@ type logic_in##bw(unsigned long addr)					\
 	return ret;							\
 }									\
 									\
-void logic_out##bw(type value, unsigned long addr)			\
+void logic_out##bwl(type value, unsigned long addr)			\
 {									\
 	if (addr < MMIO_UPPER_LIMIT) {					\
-		write##bw(value, PCI_IOBASE + addr);			\
+		write##bwl(value, PCI_IOBASE + addr);			\
 	} else if (addr >= MMIO_UPPER_LIMIT && addr < IO_SPACE_LIMIT) {	\
 		struct logic_pio_hwaddr *entry = find_io_range(addr);	\
 									\
@@ -263,11 +263,11 @@ void logic_out##bw(type value, unsigned long addr)			\
 	}								\
 }									\
 									\
-void logic_ins##bw(unsigned long addr, void *buffer,			\
-		   unsigned int count)					\
+void logic_ins##bwl(unsigned long addr, void *buffer,			\
+		    unsigned int count)					\
 {									\
 	if (addr < MMIO_UPPER_LIMIT) {					\
-		reads##bw(PCI_IOBASE + addr, buffer, count);		\
+		reads##bwl(PCI_IOBASE + addr, buffer, count);		\
 	} else if (addr >= MMIO_UPPER_LIMIT && addr < IO_SPACE_LIMIT) {	\
 		struct logic_pio_hwaddr *entry = find_io_range(addr);	\
 									\
@@ -280,11 +280,11 @@ void logic_ins##bw(unsigned long addr, void *buffer,			\
 									\
 }									\
 									\
-void logic_outs##bw(unsigned long addr, const void *buffer,		\
-		    unsigned int count)					\
+void logic_outs##bwl(unsigned long addr, const void *buffer,		\
+		     unsigned int count)				\
 {									\
 	if (addr < MMIO_UPPER_LIMIT) {					\
-		writes##bw(PCI_IOBASE + addr, buffer, count);		\
+		writes##bwl(PCI_IOBASE + addr, buffer, count);		\
 	} else if (addr >= MMIO_UPPER_LIMIT && addr < IO_SPACE_LIMIT) {	\
 		struct logic_pio_hwaddr *entry = find_io_range(addr);	\
 									\
-- 
2.16.4


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 3/3] logic_pio: Use _inX() and _outX()

[John Garry]

Use _inX() and _outX(), which include memory barriers which may be
overridden per arch.

Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: John Garry <john.garry@huawei.com>
---
 lib/logic_pio.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/logic_pio.c b/lib/logic_pio.c
index 21dc731bec88..f32fe481b492 100644
--- a/lib/logic_pio.c
+++ b/lib/logic_pio.c
@@ -235,7 +235,7 @@ type logic_in##bwl(unsigned long addr)					\
 	type ret = (type)~0;						\
 									\
 	if (addr < MMIO_UPPER_LIMIT) {					\
-		ret = read##bwl(PCI_IOBASE + addr);			\
+		ret = _in##bwl(addr);					\
 	} else if (addr >= MMIO_UPPER_LIMIT && addr < IO_SPACE_LIMIT) { \
 		struct logic_pio_hwaddr *entry = find_io_range(addr);	\
 									\
@@ -251,7 +251,7 @@ type logic_in##bwl(unsigned long addr)					\
 void logic_out##bwl(type value, unsigned long addr)			\
 {									\
 	if (addr < MMIO_UPPER_LIMIT) {					\
-		write##bwl(value, PCI_IOBASE + addr);			\
+		_out##bwl(value, addr);				\
 	} else if (addr >= MMIO_UPPER_LIMIT && addr < IO_SPACE_LIMIT) {	\
 		struct logic_pio_hwaddr *entry = find_io_range(addr);	\
 									\
-- 
2.16.4


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v2 0/3] io.h, logic_pio: Allow barriers for inX() and outX() be overridden

[Wei Xu]

Hi John,

On 2020/3/28 0:06, John Garry wrote:
> Since commits a7851aa54c0c ("io: change outX() to have their own IO
> barrier overrides") and 87fe2d543f81 ("io: change inX() to have their own
> IO barrier overrides"), the outX() and inX() functions have memory
> barriers which can be overridden per-arch.
> 
> However, under CONFIG_INDIRECT_PIO, logic_pio defines its own version of
> inX() and outX(), which still use readb et al. For these, the barrier
> after a raw read is weaker than it otherwise would be. 
> 
> This series generates consistent behaviour for logic_pio, by defining
> generic _inX() and _outX() in asm-generic/io.h, and using those in
> logic_pio. Generic _inX() and _outX() have per-arch overrideable
> barriers.
> 
> The topic was discussed there originally:
> https://lore.kernel.org/lkml/2e80d7bc-32a0-cc40-00a9-8a383a1966c2@huawei.com/
> 
> A small tidy-up patch is included.
> 
> I hope that series can go through the arm-soc tree, as with other recent
> logic_pio changes.
> 
> Hi Arnd,
> 
> I added your tag, but please let me know if you have any issue with the
> updated change in patch #1.
> 
> cheers
> 
> - Differences to v1
> 	- fix x86 clang build by adding extra build swicth for _{in,out}X
> 	- added Arnd's RB tag
> 
> John Garry (3):
>   io: Provide _inX() and _outX()
>   logic_pio: Improve macro argument name
>   logic_pio: Use _inX() and _outX()
> 
>  include/asm-generic/io.h | 64 +++++++++++++++++++++++++++++++++---------------
>  lib/logic_pio.c          | 22 ++++++++---------
>  2 files changed, 55 insertions(+), 31 deletions(-)
> 

Thanks!
Series applied to the hisilicon SoC tree.

Best Regards,
Wei

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel