diff mbox

[U-Boot,v2] common: env_nand: Ensure that we have nand_info[0] prior to use

Message ID 1471276604-25544-1-git-send-email-trini@konsulko.com
State Superseded
Delegated to: Tom Rini
Headers show

Commit Message

Tom Rini Aug. 15, 2016, 3:56 p.m. UTC
Now that nand_info[] is an array of pointers we need to ensure that it's
been populated prior to use.  We may for example have ENV in NAND set in
configurations that run on boards with and without NAND (where default
env is fine enough, such as omap3_beagle and beagleboard (NAND) vs
beagle xM (no NAND)).

Fixes: b616d9b0a708 ("nand: Embed mtd_info in struct nand_chip")
Cc: Scott Wood <oss@buserror.net>
Signed-off-by: Tom Rini <trini@konsulko.com>
---
Changes in v2:
- Oops, move check on the saveenv side in to erase_and_write_env
---
 common/env_nand.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

Comments

Crystal Wood Aug. 15, 2016, 4:48 p.m. UTC | #1
On Mon, 2016-08-15 at 11:56 -0400, Tom Rini wrote:
> Now that nand_info[] is an array of pointers we need to ensure that it's
> been populated prior to use.  We may for example have ENV in NAND set in
> configurations that run on boards with and without NAND (where default
> env is fine enough, such as omap3_beagle and beagleboard (NAND) vs
> beagle xM (no NAND)).
> 
> Fixes: b616d9b0a708 ("nand: Embed mtd_info in struct nand_chip")
> Cc: Scott Wood <oss@buserror.net>
> Signed-off-by: Tom Rini <trini@konsulko.com>
> ---
> Changes in v2:
> - Oops, move check on the saveenv side in to erase_and_write_env
> ---
>  common/env_nand.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/common/env_nand.c b/common/env_nand.c
> index fc99a5e3fc0d..96a1020b5e79 100644
> --- a/common/env_nand.c
> +++ b/common/env_nand.c
> @@ -163,6 +163,9 @@ static int erase_and_write_env(const struct env_location
> *location,
>  {
>  	int ret = 0;
>  
> +	if (!nand_info[0])
> +		return 1;
> +
>  	printf("Erasing %s...\n", location->name);
>  	if (nand_erase_opts(nand_info[0], &location->erase_opts))
>  		return 1;
> @@ -244,10 +247,12 @@ static int readenv(size_t offset, u_char *buf)
>  {
>  	size_t end = offset + CONFIG_ENV_RANGE;
>  	size_t amount_loaded = 0;
> -	size_t blocksize, len;
> +	size_t blocksize = 0, len;
>  	u_char *char_ptr;
>  
> -	blocksize = nand_info[0]->erasesize;
> +	if (nand_info[0])
> +		blocksize = nand_info[0]->erasesize;
> +
>  	if (!blocksize)
>  		return 1;
>  

Messing around with blocksize rather than directly returning is awkward -- it
makes it look as if you could proceed with blocksize zero until you read the
next statement.  And is there ever a case where blocksize can be zero, now
that uninitialized NAND devices will simply have a NULL pointer?

-Scott
Tom Rini Aug. 15, 2016, 4:54 p.m. UTC | #2
On Mon, Aug 15, 2016 at 11:48:31AM -0500, Scott Wood wrote:
> On Mon, 2016-08-15 at 11:56 -0400, Tom Rini wrote:
> > Now that nand_info[] is an array of pointers we need to ensure that it's
> > been populated prior to use.  We may for example have ENV in NAND set in
> > configurations that run on boards with and without NAND (where default
> > env is fine enough, such as omap3_beagle and beagleboard (NAND) vs
> > beagle xM (no NAND)).
> > 
> > Fixes: b616d9b0a708 ("nand: Embed mtd_info in struct nand_chip")
> > Cc: Scott Wood <oss@buserror.net>
> > Signed-off-by: Tom Rini <trini@konsulko.com>
> > ---
> > Changes in v2:
> > - Oops, move check on the saveenv side in to erase_and_write_env
> > ---
> >  common/env_nand.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> > 
> > diff --git a/common/env_nand.c b/common/env_nand.c
> > index fc99a5e3fc0d..96a1020b5e79 100644
> > --- a/common/env_nand.c
> > +++ b/common/env_nand.c
> > @@ -163,6 +163,9 @@ static int erase_and_write_env(const struct env_location
> > *location,
> >  {
> >  	int ret = 0;
> >  
> > +	if (!nand_info[0])
> > +		return 1;
> > +
> >  	printf("Erasing %s...\n", location->name);
> >  	if (nand_erase_opts(nand_info[0], &location->erase_opts))
> >  		return 1;
> > @@ -244,10 +247,12 @@ static int readenv(size_t offset, u_char *buf)
> >  {
> >  	size_t end = offset + CONFIG_ENV_RANGE;
> >  	size_t amount_loaded = 0;
> > -	size_t blocksize, len;
> > +	size_t blocksize = 0, len;
> >  	u_char *char_ptr;
> >  
> > -	blocksize = nand_info[0]->erasesize;
> > +	if (nand_info[0])
> > +		blocksize = nand_info[0]->erasesize;
> > +
> >  	if (!blocksize)
> >  		return 1;
> >  
> 
> Messing around with blocksize rather than directly returning is awkward -- it
> makes it look as if you could proceed with blocksize zero until you read the
> next statement.  And is there ever a case where blocksize can be zero, now
> that uninitialized NAND devices will simply have a NULL pointer?

OK, I can see how that reads awkwardly now.  And I suspect that no,
blocksize could never be valid and zero.  I'll re-work for a v3.
diff mbox

Patch

diff --git a/common/env_nand.c b/common/env_nand.c
index fc99a5e3fc0d..96a1020b5e79 100644
--- a/common/env_nand.c
+++ b/common/env_nand.c
@@ -163,6 +163,9 @@  static int erase_and_write_env(const struct env_location *location,
 {
 	int ret = 0;
 
+	if (!nand_info[0])
+		return 1;
+
 	printf("Erasing %s...\n", location->name);
 	if (nand_erase_opts(nand_info[0], &location->erase_opts))
 		return 1;
@@ -244,10 +247,12 @@  static int readenv(size_t offset, u_char *buf)
 {
 	size_t end = offset + CONFIG_ENV_RANGE;
 	size_t amount_loaded = 0;
-	size_t blocksize, len;
+	size_t blocksize = 0, len;
 	u_char *char_ptr;
 
-	blocksize = nand_info[0]->erasesize;
+	if (nand_info[0])
+		blocksize = nand_info[0]->erasesize;
+
 	if (!blocksize)
 		return 1;