Skip to contentSkip to conversation

mtd: ixp4xx: Unrequire CONFIG_MTD_CFI_BE_BYTE_SWAP

PATCH Superseded Aaron Sierra 2 participants 3 messages Sep 23, 2014 – Sep 29, 2014 1 message not in the archive Memory technology devices (MTD)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

PATCH v2 1/1 mtd: ixp4xx: Unrequire CONFIG_MTD_CFI_BE_BYTE_SWAP

The .swap member of the map_info structure is provided for situations where a mapping must always be big or little endian regardless of the endianness of the system performing the access.

---
v2: Account for 32-bit swapping

 drivers/mtd/maps/ixp4xx.c | 35 ++++++++++++-----------------------
 1 file changed, 12 insertions(+), 23 deletions(-)
drivers/mtd/maps/ixp4xx.c +12 -23
index 6a589f1..541727e 100644
@@ -26,6 +26,7 @@
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/map.h>
 #include <linux/mtd/partitions.h>
+#include <linux/mtd/cfi_endian.h>
 
 #include <asm/io.h>
 #include <asm/mach/flash.h>
@@ -48,43 +49,30 @@
  *     | C | D | 2
  *     +---+---+
  * This means that on LE systems each 16 bit word must be swapped. Note that
- * this requires CONFIG_MTD_CFI_BE_BYTE_SWAP to be enabled to 'unswap' the CFI
- * data and other flash commands which are always in D7-D0.
+ * this is accounted for by defining map_info.swap to be CFI_BIG_ENDIAN to
+ * 'unswap' the CFI commands and using cpu_to_be16/be16_to_cpu to 'unswap'
+ * the data.
  */
-#ifndef __ARMEB__
-#ifndef CONFIG_MTD_CFI_BE_BYTE_SWAP
-#  error CONFIG_MTD_CFI_BE_BYTE_SWAP required
+
+#ifdef __ARMEB__
+#define fixup_32(io) (io)
+#else
+#define fixup_32(io) (void __iomem *)((unsigned long)(io) ^ 2)
 #endif
 
 static inline u16 flash_read16(void __iomem *addr)
 {
-	return be16_to_cpu(__raw_readw((void __iomem *)((unsigned long)addr ^ 0x2)));
+	return be16_to_cpu(__raw_readw(fixup_32(addr)));
 }
 
 static inline void flash_write16(u16 d, void __iomem *addr)
 {
-	__raw_writew(cpu_to_be16(d), (void __iomem *)((unsigned long)addr ^ 0x2));
+	__raw_writew(cpu_to_be16(d), fixup_32(addr));
 }
 
 #define	BYTE0(h)	((h) & 0xFF)
 #define	BYTE1(h)	(((h) >> 8) & 0xFF)
 
-#else
-
-static inline u16 flash_read16(const void __iomem *addr)
-{
-	return __raw_readw(addr);
-}
-
-static inline void flash_write16(u16 d, void __iomem *addr)
-{
-	__raw_writew(d, addr);
-}
-
-#define	BYTE0(h)	(((h) >> 8) & 0xFF)
-#define	BYTE1(h)	((h) & 0xFF)
-#endif
-
 static map_word ixp4xx_read16(struct map_info *map, unsigned long ofs)
 {
 	map_word val;
@@ -200,6 +188,7 @@ static int ixp4xx_flash_probe(struct platform_device *dev)
 	 * Tell the MTD layer we're not 1:1 mapped so that it does
 	 * not attempt to do a direct access on us.
 	 */
+	info->map.swap = CFI_BIG_ENDIAN;
 	info->map.phys = NO_XIP;
 	info->map.size = resource_size(dev->resource);
 
Signature · 1 line
1.9.1
#2 lore raw

PATCH v2 1/1 mtd: ixp4xx: Unrequire CONFIG_MTD_CFI_BE_BYTE_SWAP

Quoted 3 lines · Aaron Sierra

The .swap member of the map_info structure is provided for situations where a mapping must always be big or little endian regardless of the endianness of the system performing the access.

I like the idea but the patch doesn't work :-(

Without the patch:
IXP4XX-Flash.0: Found 1 x16 devices at 0x0 in 16-bit bank. Manufacturer ID 0x000089 Chip ID 0x008922 Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Using buffer write method
Using auto-unlock on power-up/resume
cfi_cmdset_0001: Erase suspend on write enabled
Searching for RedBoot partition table in IXP4XX-Flash.0 at offset 0x1fe0000
5 RedBoot partitions found on MTD device IXP4XX-Flash.0

With the patch applied:
IXP4XX-Flash.0: Found 1 x16 devices at 0x0 in 16-bit bank. Manufacturer ID 0x000089 Chip ID 0x008922 Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Using buffer write method
Using auto-unlock on power-up/resume
cfi_cmdset_0001: Erase suspend on write enabled
Searching for RedBoot partition table in IXP4XX-Flash.0 at offset 0x1fe0000
No RedBoot partition table detected in IXP4XX-Flash.0

Quoted 13 lines

+++ b/drivers/mtd/maps/ixp4xx.c

drivers/mtd/maps/ixp4xx.c +5 -3
-#ifndef __ARMEB__
-#ifndef CONFIG_MTD_CFI_BE_BYTE_SWAP
-#  error CONFIG_MTD_CFI_BE_BYTE_SWAP required
+
+#ifdef __ARMEB__
+#define fixup_32(io) (io)
+#else
+#define fixup_32(io) (void __iomem *)((unsigned long)(io) ^ 2)
 #endif
 
 static inline u16 flash_read16(void __iomem *addr)

drivers/mtd/maps/ixp4xx.c line 71

Quoted 10 lines
drivers/mtd/maps/ixp4xx.c +2 -2
 {
-	return be16_to_cpu(__raw_readw((void __iomem *)((unsigned long)addr ^ 0x2)));
+	return be16_to_cpu(__raw_readw(fixup_32(addr)));
 }
 
 static inline void flash_write16(u16 d, void __iomem *addr)
 {
-	__raw_writew(cpu_to_be16(d), (void __iomem *)((unsigned long)addr ^ 0x2));
+	__raw_writew(cpu_to_be16(d), fixup_32(addr));
 }

The above change seems fine.

Quoted 4 lines

#define BYTE0(h) ((h) & 0xFF)
#define BYTE1(h) (((h) >> 8) & 0xFF)

-#else

...

drivers/mtd/maps/ixp4xx.c line 86

Quoted 3 lines
drivers/mtd/maps/ixp4xx.c +0 -3
-#define	BYTE0(h)	(((h) >> 8) & 0xFF)
-#define	BYTE1(h)	((h) & 0xFF)
-#endif

This is used by ixp4xx_copy_from(). I don't exactly know what is MTD layer going to do with map.swap = CFI_BIG_ENDIAN, but I think the natural thing (= big endian on this CPU) is to copy data in big-endian order, and then maybe (in LE mode) swap it.

In fact, it detects those RedBoot partitions when I use the original BE BYTE0/BYTE1 macros (this second set, deleted by your patch). Tested in BE mode only for now, can't give it more time at the moment but will do soon.

CONFIG_MTD_CFI=y
CONFIG_MTD_CFI_ADV_OPTIONS=y
CONFIG_MTD_CFI_NOSWAP=y
# CONFIG_MTD_CFI_BE_BYTE_SWAP is not set
# CONFIG_MTD_CFI_LE_BYTE_SWAP is not set
CONFIG_MTD_CFI_GEOMETRY=y
CONFIG_MTD_CFI_I1=y
# CONFIG_MTD_CFI_I2 is not set
# CONFIG_MTD_CFI_I4 is not set
# CONFIG_MTD_CFI_I8 is not set
CONFIG_MTD_CFI_INTELEXT=y
# CONFIG_MTD_CFI_AMDSTD is not set
# CONFIG_MTD_CFI_STAA is not set
CONFIG_MTD_CFI_UTIL=y

Signature · 4 lines
Krzysztof Halasa

Research Institute for Automation and Measurements PIAP
Al. Jerozolimskie 202, 02-486 Warsaw, Poland
#3 lore raw

PATCH v2 1/1 mtd: ixp4xx: Unrequire CONFIG_MTD_CFI_BE_BYTE_SWAP

----- Original Message -----

Quoted 39 lines

From: "Krzysztof Ha?asa" <[email protected]>
Sent: Monday, September 29, 2014 2:50:08 AM

Quoted 3 lines · Aaron Sierra

The .swap member of the map_info structure is provided for situations where a mapping must always be big or little endian regardless of the endianness of the system performing the access.

I like the idea but the patch doesn't work :-(

Without the patch:
IXP4XX-Flash.0: Found 1 x16 devices at 0x0 in 16-bit bank. Manufacturer ID 0x000089 Chip ID 0x008922
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Using buffer write method
Using auto-unlock on power-up/resume
cfi_cmdset_0001: Erase suspend on write enabled
Searching for RedBoot partition table in IXP4XX-Flash.0 at offset 0x1fe0000 5 RedBoot partitions found on MTD device IXP4XX-Flash.0

With the patch applied:
IXP4XX-Flash.0: Found 1 x16 devices at 0x0 in 16-bit bank. Manufacturer ID 0x000089 Chip ID 0x008922
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Intel/Sharp Extended Query Table at 0x010A
Using buffer write method
Using auto-unlock on power-up/resume
cfi_cmdset_0001: Erase suspend on write enabled
Searching for RedBoot partition table in IXP4XX-Flash.0 at offset 0x1fe0000 No RedBoot partition table detected in IXP4XX-Flash.0

Quoted 2 lines
Quoted 1 line

+++ b/drivers/mtd/maps/ixp4xx.c

[ snip agreed code ]

Quoted 6 lines
Quoted 5 lines

#define BYTE0(h) ((h) & 0xFF)
#define BYTE1(h) (((h) >> 8) & 0xFF)

-#else

...

Quoted 8 lines

drivers/mtd/maps/ixp4xx.c line 86

Quoted 3 lines
drivers/mtd/maps/ixp4xx.c +0 -3
-#define	BYTE0(h)	(((h) >> 8) & 0xFF)
-#define	BYTE1(h)	((h) & 0xFF)
-#endif

This is used by ixp4xx_copy_from(). I don't exactly know what is MTD layer going to do with map.swap = CFI_BIG_ENDIAN, but I think the natural thing (= big endian on this CPU) is to copy data in big-endian order, and then maybe (in LE mode) swap it.

Your guess regarding .swap is correct. The swapping is performed by inline functions defined in include/linux/mtd/cfi.h via cfiX_to_cpu/cpu_to_cfiX macros.

Quoted 5 lines

In fact, it detects those RedBoot partitions when I use the original BE BYTE0/BYTE1 macros (this second set, deleted by your patch). Tested in BE mode only for now, can't give it more time at the moment but will do soon.

Thanks Krzysztof, I see what I did wrong. I will submit a new patch that defines BYTE0/BYTE1 to work for both endian modes.

-Aaron