diff mbox

[U-Boot,V2] cfb_console: force natural alignment of 32-bit fields in gzipped .bmp files

Message ID 1393621572-10306-1-git-send-email-eric.nelson@boundarydevices.com
State Superseded
Delegated to: Tom Rini
Headers show

Commit Message

Eric Nelson Feb. 28, 2014, 9:06 p.m. UTC
.bmp files contain 32-bit integers aligned at offsets of +2, +6,
et cetera within the bmp_header structure (see include/bmp_layout.h).

Support for gzip-compressed .bmp files is present in the cfb_console
display subsystem by uncompressing them prior to use.

This patch forces the in-memory header to be aligned properly
for these compressed images by extracting them to a 2-byte
offset in the memory returned by malloc. Since malloc will always
return a 4-byte aligned value, this forces the .bmp header
fields to be naturally aligned on 4-byte addresses.

Refer to these files for more details:
	doc/README.displaying-bmps
	doc/README.arm-unaligned-accesses

Signed-off-by: Eric Nelson <eric.nelson@boundarydevices.com>
---
 drivers/video/cfb_console.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

Comments

Wolfgang Denk Feb. 28, 2014, 9:44 p.m. UTC | #1
Dear Eric Nelson,

your message contains no thread information - there is neither a
In-reply-to: nor any References: header.  This is bad!

Please also make sure to keep the Subject: test short; the recommended
style for the commit message is "50/72 formatting", i. e. the Subject
should not exceed 50 (...60) characters.

In message <1393621572-10306-1-git-send-email-eric.nelson@boundarydevices.com> you wrote:
> .bmp files contain 32-bit integers aligned at offsets of +2, +6,
> et cetera within the bmp_header structure (see include/bmp_layout.h).
> 
> Support for gzip-compressed .bmp files is present in the cfb_console
> display subsystem by uncompressing them prior to use.
> 
> This patch forces the in-memory header to be aligned properly
> for these compressed images by extracting them to a 2-byte
> offset in the memory returned by malloc. Since malloc will always
> return a 4-byte aligned value, this forces the .bmp header
> fields to be naturally aligned on 4-byte addresses.
> 
> Refer to these files for more details:
> 	doc/README.displaying-bmps
> 	doc/README.arm-unaligned-accesses
> 
> Signed-off-by: Eric Nelson <eric.nelson@boundarydevices.com>
> ---
>  drivers/video/cfb_console.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)

So what exactly has changed since version 1?  Please note that the
change log is mandatory!

> diff --git a/drivers/video/cfb_console.c b/drivers/video/cfb_console.c
> index 6db4073..2209148 100644
> --- a/drivers/video/cfb_console.c
> +++ b/drivers/video/cfb_console.c
> @@ -1473,7 +1473,10 @@ int video_display_bitmap(ulong bmp_image, int x, int y)
>  			printf("Error: malloc in gunzip failed!\n");
>  			return 1;
>  		}
> -		if (gunzip(dst, CONFIG_SYS_VIDEO_LOGO_MAX_SIZE,
> +		/* NB: we need to force offset of +2
> +		 * See doc/README.displaying-bmps
> +		 */

Incorrect multiline comment style, please fix.

I asked you before to fix that, please make sure not to ignore
review comments!

Best regards,

Wolfgang Denk
Eric Nelson Feb. 28, 2014, 9:51 p.m. UTC | #2
Sorry Wolfgang,

On 02/28/2014 02:44 PM, Wolfgang Denk wrote:
> Dear Eric Nelson,
>
> your message contains no thread information - there is neither a
> In-reply-to: nor any References: header.  This is bad!
>
> Please also make sure to keep the Subject: test short; the recommended
> style for the commit message is "50/72 formatting", i. e. the Subject
> should not exceed 50 (...60) characters.
>

Okay. I was trying to be more explicit here.

> In message <1393621572-10306-1-git-send-email-eric.nelson@boundarydevices.com> you wrote:
>> .bmp files contain 32-bit integers aligned at offsets of +2, +6,
>> et cetera within the bmp_header structure (see include/bmp_layout.h).
>>
>> Support for gzip-compressed .bmp files is present in the cfb_console
>> display subsystem by uncompressing them prior to use.
>>
>> This patch forces the in-memory header to be aligned properly
>> for these compressed images by extracting them to a 2-byte
>> offset in the memory returned by malloc. Since malloc will always
>> return a 4-byte aligned value, this forces the .bmp header
>> fields to be naturally aligned on 4-byte addresses.
>>
>> Refer to these files for more details:
>> 	doc/README.displaying-bmps
>> 	doc/README.arm-unaligned-accesses
>>
>> Signed-off-by: Eric Nelson <eric.nelson@boundarydevices.com>
>> ---
>>   drivers/video/cfb_console.c | 7 +++++--
>>   1 file changed, 5 insertions(+), 2 deletions(-)
>
> So what exactly has changed since version 1?  Please note that the
> change log is mandatory!
>

Got it.

>> diff --git a/drivers/video/cfb_console.c b/drivers/video/cfb_console.c
>> index 6db4073..2209148 100644
>> --- a/drivers/video/cfb_console.c
>> +++ b/drivers/video/cfb_console.c
>> @@ -1473,7 +1473,10 @@ int video_display_bitmap(ulong bmp_image, int x, int y)
>>   			printf("Error: malloc in gunzip failed!\n");
>>   			return 1;
>>   		}
>> -		if (gunzip(dst, CONFIG_SYS_VIDEO_LOGO_MAX_SIZE,
>> +		/* NB: we need to force offset of +2
>> +		 * See doc/README.displaying-bmps
>> +		 */
>
> Incorrect multiline comment style, please fix.
>
> I asked you before to fix that, please make sure not to ignore
> review comments!
>
Not intentional... Will fix in V3.

Regards,


Eric
diff mbox

Patch

diff --git a/drivers/video/cfb_console.c b/drivers/video/cfb_console.c
index 6db4073..2209148 100644
--- a/drivers/video/cfb_console.c
+++ b/drivers/video/cfb_console.c
@@ -1473,7 +1473,10 @@  int video_display_bitmap(ulong bmp_image, int x, int y)
 			printf("Error: malloc in gunzip failed!\n");
 			return 1;
 		}
-		if (gunzip(dst, CONFIG_SYS_VIDEO_LOGO_MAX_SIZE,
+		/* NB: we need to force offset of +2
+		 * See doc/README.displaying-bmps
+		 */
+		if (gunzip(dst+2, CONFIG_SYS_VIDEO_LOGO_MAX_SIZE-2,
 			   (uchar *) bmp_image,
 			   &len) != 0) {
 			printf("Error: no valid bmp or bmp.gz image at %lx\n",
@@ -1489,7 +1492,7 @@  int video_display_bitmap(ulong bmp_image, int x, int y)
 		/*
 		 * Set addr to decompressed image
 		 */
-		bmp = (bmp_image_t *) dst;
+		bmp = (bmp_image_t *)(dst+2);
 
 		if (!((bmp->header.signature[0] == 'B') &&
 		      (bmp->header.signature[1] == 'M'))) {