cfi driver fixes

cfi driver fixes

[Sascha Hauer]

The following fixes some problems with the cfi flash driver for
32 bit flashes.

----------------------------------------------------------------
Sascha Hauer (2):
      cfi flash: Fix loop count calculation
      cfi flash: Fix alignment problem

 drivers/nor/cfi_flash_intel.c |   20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 1/2] cfi flash: Fix loop count calculation

[Sascha Hauer]

'reduce the number of loops by the width of the port' means
a simple len / width. Do not try to be clever by shifting
and doing it wrong.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 drivers/nor/cfi_flash_intel.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nor/cfi_flash_intel.c b/drivers/nor/cfi_flash_intel.c
index 6318cfe..f0cbf72 100644
--- a/drivers/nor/cfi_flash_intel.c
+++ b/drivers/nor/cfi_flash_intel.c
@@ -68,7 +68,7 @@ static int intel_flash_write_cfibuffer (struct flash_info *info, ulong dest, con
 		return retcode;
 
 	/* reduce the number of loops by the width of the port	*/
-	cnt = len >> (info->portwidth - 1);
+	cnt = len / width;
 
 	flash_write_cmd(info, sector, 0, (u32)cnt - 1);
 	while (cnt-- > 0) {
-- 
1.7.9.5


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 2/2] cfi flash: Fix alignment problem

[Sascha Hauer]

The intel cfi buffer write has a problem with writing when
the alignment of the buffer in memory is smaller than the
flash bus width.

This patch fixes a alignment problem which may show during this
scenario:
- 32 or 64 attached NOR flash
- flashing an image directly from network to the nor flash

The involved network driver is "smc9111.c".

The data that comes from the network stack and should be written into
the flash isn't 32 bit aligned (at least with this network driver).
This is probably due to the 48 bit wide ethernet addresses.

However the "cfi_flash.c" driver doesn't handle this situation, and
accesses the not-aligned address with a 32 bit pointer.

This patch fixes the problem by reducing the access width if an
aligment problem between source and destination is found.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 drivers/nor/cfi_flash_intel.c |   18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/drivers/nor/cfi_flash_intel.c b/drivers/nor/cfi_flash_intel.c
index f0cbf72..32e581a 100644
--- a/drivers/nor/cfi_flash_intel.c
+++ b/drivers/nor/cfi_flash_intel.c
@@ -57,6 +57,9 @@ static int intel_flash_write_cfibuffer (struct flash_info *info, ulong dest, con
 	int retcode;
 	void *src = (void*)cp;
 	void *dst = (void *)dest;
+	/* reduce width due to possible alignment problems */
+	const unsigned long ptr = (unsigned long)dest | (unsigned long)cp | info->portwidth;
+	const int width = ptr & -ptr;
 
 	sector = find_sector (info, dest);
 	flash_write_cmd (info, sector, 0, FLASH_CMD_CLEAR_STATUS);
@@ -70,20 +73,25 @@ static int intel_flash_write_cfibuffer (struct flash_info *info, ulong dest, con
 	/* reduce the number of loops by the width of the port	*/
 	cnt = len / width;
 
-	flash_write_cmd(info, sector, 0, (u32)cnt - 1);
+	flash_write_cmd(info, sector, 0, cnt - 1);
 	while (cnt-- > 0) {
-		if (bankwidth_is_1(info)) {
+		switch (width) {
+		case 1:
 			flash_write8(flash_read8(src), dst);
 			src += 1, dst += 1;
-		} else if (bankwidth_is_2(info)) {
+			break;
+		case 2:
 			flash_write16(flash_read16(src), dst);
 			src += 2, dst += 2;
-		} else if (bankwidth_is_4(info)) {
+			break;
+		case 4:
 			flash_write32(flash_read32(src), dst);
 			src += 4, dst += 4;
-		} else if (bankwidth_is_8(info)) {
+			break;
+		case 8:
 			flash_write64(flash_read64(src), dst);
 			src += 8, dst += 8;
+			break;
 		}
 	}
 
-- 
1.7.9.5


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: [PATCH 1/2] cfi flash: Fix loop count calculation

[Eric Bénard]

Le Thu,  5 Apr 2012 10:59:14 +0200,
Sascha Hauer <s.hauer@pengutronix.de> a écrit :

> 'reduce the number of loops by the width of the port' means
> a simple len / width. Do not try to be clever by shifting
> and doing it wrong.
> 
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Tested-by: Eric Bénard <eric@eukrea.com>
> ---
>  drivers/nor/cfi_flash_intel.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/nor/cfi_flash_intel.c b/drivers/nor/cfi_flash_intel.c
> index 6318cfe..f0cbf72 100644
> --- a/drivers/nor/cfi_flash_intel.c
> +++ b/drivers/nor/cfi_flash_intel.c
> @@ -68,7 +68,7 @@ static int intel_flash_write_cfibuffer (struct flash_info *info, ulong dest, con
>  		return retcode;
>  
>  	/* reduce the number of loops by the width of the port	*/
> -	cnt = len >> (info->portwidth - 1);
> +	cnt = len / width;
>  
>  	flash_write_cmd(info, sector, 0, (u32)cnt - 1);
>  	while (cnt-- > 0) {



_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox