Patchwork [U-Boot] cfi: fix the incomplete erased status check in buffer write

login
register
mail settings
Submitter htbegin
Date Feb. 19, 2012, 11:48 a.m.
Message ID <1329652117-14084-1-git-send-email-hotforest@gmail.com>
Download mbox | patch
Permalink /patch/142055/
State Accepted
Commit 85c344e5f98408c0bcf988a6c6fca68c1f3c2015
Delegated to: Stefan Roese
Headers show

Comments

htbegin - Feb. 19, 2012, 11:48 a.m.
Without the fix, flash_write_cfibuffer will return timeout error
instead of not-erased error when all bytes in the first word of
a sector are 0xFF and other words are not.

Signed-off-by: htbegin <hotforest@gmail.com>
Cc: Stefan Roese <sr@denx.de>
---
 drivers/mtd/cfi_flash.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
Stefan Roese - March 6, 2012, 2:51 p.m.
Hi "htbegin",

first sorry for the late review.

On Sunday 19 February 2012 12:48:37 htbegin wrote:
> Without the fix, flash_write_cfibuffer will return timeout error
> instead of not-erased error when all bytes in the first word of
> a sector are 0xFF and other words are not.
> 
> Signed-off-by: htbegin <hotforest@gmail.com>

Please use your real name here in the s-o-b line.

> ---
>  drivers/mtd/cfi_flash.c |    4 ++--
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/mtd/cfi_flash.c b/drivers/mtd/cfi_flash.c
> index 722c3fc..35294bc 100644
> --- a/drivers/mtd/cfi_flash.c
> +++ b/drivers/mtd/cfi_flash.c
> @@ -873,7 +873,7 @@ static int flash_write_cfibuffer (flash_info_t * info,
> ulong dest, uchar * cp, void *src = cp;
>  	void *dst = (void *)dest;
>  	void *dst2 = dst;
> -	int flag = 0;
> +	int flag = 1;
>  	uint offset = 0;
>  	unsigned int shift;
>  	uchar write_cmd;
> @@ -898,7 +898,7 @@ static int flash_write_cfibuffer (flash_info_t * info,
> ulong dest, uchar * cp,
> 
>  	cnt = len >> shift;
> 
> -	while ((cnt-- > 0) && (flag == 0)) {
> +	while ((cnt-- > 0) && (flag == 1)) {
>  		switch (info->portwidth) {
>  		case FLASH_CFI_8BIT:
>  			flag = ((flash_read8(dst2) & flash_read8(src)) ==

I checked your patch, and it seems to me that you have spotted a problem. But 
your problem description above doesn't seem to be totally correct. From my 
understanding, the loop checking for not erased flash aborts directly once 
0xff is found. So that the check doesn't continue till the first non-0xff word 
as it should be.

Can you confirm this? Then please re-send you patch with a new patch 
description and a correct name in the Signed-off-by line.

Thanks.

Best regards,
Stefan

--
DENX Software Engineering GmbH,      MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich,  Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-0 Fax: (+49)-8142-66989-80 Email: office@denx.de

Patch

diff --git a/drivers/mtd/cfi_flash.c b/drivers/mtd/cfi_flash.c
index 722c3fc..35294bc 100644
--- a/drivers/mtd/cfi_flash.c
+++ b/drivers/mtd/cfi_flash.c
@@ -873,7 +873,7 @@  static int flash_write_cfibuffer (flash_info_t * info, ulong dest, uchar * cp,
 	void *src = cp;
 	void *dst = (void *)dest;
 	void *dst2 = dst;
-	int flag = 0;
+	int flag = 1;
 	uint offset = 0;
 	unsigned int shift;
 	uchar write_cmd;
@@ -898,7 +898,7 @@  static int flash_write_cfibuffer (flash_info_t * info, ulong dest, uchar * cp,
 
 	cnt = len >> shift;
 
-	while ((cnt-- > 0) && (flag == 0)) {
+	while ((cnt-- > 0) && (flag == 1)) {
 		switch (info->portwidth) {
 		case FLASH_CFI_8BIT:
 			flag = ((flash_read8(dst2) & flash_read8(src)) ==