diff mbox

[U-Boot,v5,06/12] ums: always initialize mmc before ums_disk_init()

Message ID 1398712412-15503-7-git-send-email-m.zalega@samsung.com
State Accepted
Delegated to: Marek Vasut
Headers show

Commit Message

Mateusz Zalega April 28, 2014, 7:13 p.m. UTC 
In cases when MMC hadn't been initialized before, ie. by the user or other
subsystem, it was still uninitialized while UMS media capacity check,
leading to broken ums command.

UMS has to initialize resources it uses.

Tested on Samsung Goni.

Signed-off-by: Mateusz Zalega <m.zalega@samsung.com>
Tested-by: Mateusz Zalega <m.zalega@samsung.com>
Acked-by: Lukasz Majewski <l.majewski@samsung.com>
Cc: Minkyu Kang <mk7.kang@samsung.com>
---
Changes since v3:
- Added const attribute of ums_disk_init introduced type warnings. It would
  take a lot of changes to MMC core code to fix it, so it was dropped, even
  though ums_disk_init shouldn't and doesn't change struct mmc it is passed.
Changes since v4:
- added slightly more detailed commit message 
---
 board/samsung/common/ums.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

Comments

Stephen Warren May 1, 2014, 8:03 p.m. UTC | #1 
On 04/28/2014 01:13 PM, Mateusz Zalega wrote:
> In cases when MMC hadn't been initialized before, ie. by the user or other
> subsystem, it was still uninitialized while UMS media capacity check,
> leading to broken ums command.
> 
> UMS has to initialize resources it uses.

It's probably better to just switch from find_mmc_device() to
get_device(), which does the initialization internally. patch "ums: use
get_device() not find_mmc_device()" does this.

http://patchwork.ozlabs.org/patch/344328/

That said, at this point, it's probably simplest to just apply this
series, and I'll rebase my series on top of this in a day or two.
Łukasz Majewski May 5, 2014, 8:05 a.m. UTC | #2 
Hi Mateusz,

> In cases when MMC hadn't been initialized before, ie. by the user or
> other subsystem, it was still uninitialized while UMS media capacity
> check, leading to broken ums command.
> 
> UMS has to initialize resources it uses.
> 
> Tested on Samsung Goni.
> 
> Signed-off-by: Mateusz Zalega <m.zalega@samsung.com>
> Tested-by: Mateusz Zalega <m.zalega@samsung.com>
> Acked-by: Lukasz Majewski <l.majewski@samsung.com>
> Cc: Minkyu Kang <mk7.kang@samsung.com>

Applied to u-boot-dfu. THanks,


> ---
> Changes since v3:
> - Added const attribute of ums_disk_init introduced type warnings. It
> would take a lot of changes to MMC core code to fix it, so it was
> dropped, even though ums_disk_init shouldn't and doesn't change
> struct mmc it is passed. Changes since v4:
> - added slightly more detailed commit message 
> ---
>  board/samsung/common/ums.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/board/samsung/common/ums.c b/board/samsung/common/ums.c
> index dc155ad..cebabe9 100644
> --- a/board/samsung/common/ums.c
> +++ b/board/samsung/common/ums.c
> @@ -66,11 +66,9 @@ static struct ums *ums_disk_init(struct mmc *mmc)
>  
>  struct ums *ums_init(unsigned int dev_num)
>  {
> -	struct mmc *mmc = NULL;
> +	struct mmc *mmc = find_mmc_device(dev_num);
>  
> -	mmc = find_mmc_device(dev_num);
> -	if (!mmc)
> +	if (!mmc || mmc_init(mmc))
>  		return NULL;
> -
>  	return ums_disk_init(mmc);
>  }
diff mbox

Patch

diff --git a/board/samsung/common/ums.c b/board/samsung/common/ums.c
index dc155ad..cebabe9 100644
--- a/board/samsung/common/ums.c
+++ b/board/samsung/common/ums.c
@@ -66,11 +66,9 @@  static struct ums *ums_disk_init(struct mmc *mmc)
 
 struct ums *ums_init(unsigned int dev_num)
 {
-	struct mmc *mmc = NULL;
+	struct mmc *mmc = find_mmc_device(dev_num);
 
-	mmc = find_mmc_device(dev_num);
-	if (!mmc)
+	if (!mmc || mmc_init(mmc))
 		return NULL;
-
 	return ums_disk_init(mmc);
 }