| Message ID | 1345711214-11969-1-git-send-email-sbabic@denx.de |
|---|---|
| State | Superseded |
| Delegated to: | Tom Rini |
| Headers | show |
On 08/23/2012 10:40 AM, Stefano Babic wrote: > If an u-boot image is not found, SPL thinks to load a bare > u-boot.bin image with a maximum size of 200KB. > Use CONFIG_SYS_MONITOR_LEN instead. > > Signed-off-by: Stefano Babic <sbabic@denx.de> > CC: Tom Rini <trini@ti.com> > --- > > Note: this is based on Tom's series > "ARM: SPL: Make more generic, merge DaVinci and OMAP" > > Tested with V2 version on a MX35. > > common/spl/spl.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/common/spl/spl.c b/common/spl/spl.c > index 7d15460..827ff1c 100644 > --- a/common/spl/spl.c > +++ b/common/spl/spl.c > @@ -100,7 +100,11 @@ void spl_parse_image_header(const struct image_header *header) > debug("mkimage signature not found - ih_magic = %x\n", > header->ih_magic); > /* Let's assume U-Boot will not be more than 200 KB */ > +#ifdef CONFIG_SYS_MONITOR_LEN > + spl_image.size = CONFIG_SYS_MONITOR_LEN; > +#else > spl_image.size = 200 * 1024; > +#endif Yes, I noticed this 200 KiB setting as well. But it seems that this ".size" variable is not referenced at all. Or am I missing something? If this is correct, then we should probably remove setting it completely. What do you think? Thanks, Stefan
On 23/08/2012 10:51, Stefan Roese wrote: > On 08/23/2012 10:40 AM, Stefano Babic wrote: >> If an u-boot image is not found, SPL thinks to load a bare >> u-boot.bin image with a maximum size of 200KB. >> Use CONFIG_SYS_MONITOR_LEN instead. >> >> Signed-off-by: Stefano Babic <sbabic@denx.de> >> CC: Tom Rini <trini@ti.com> >> --- >> Hi Stefan, >> Note: this is based on Tom's series >> "ARM: SPL: Make more generic, merge DaVinci and OMAP" >> >> Tested with V2 version on a MX35. >> >> common/spl/spl.c | 4 ++++ >> 1 file changed, 4 insertions(+) >> >> diff --git a/common/spl/spl.c b/common/spl/spl.c >> index 7d15460..827ff1c 100644 >> --- a/common/spl/spl.c >> +++ b/common/spl/spl.c >> @@ -100,7 +100,11 @@ void spl_parse_image_header(const struct image_header *header) >> debug("mkimage signature not found - ih_magic = %x\n", >> header->ih_magic); >> /* Let's assume U-Boot will not be more than 200 KB */ >> +#ifdef CONFIG_SYS_MONITOR_LEN >> + spl_image.size = CONFIG_SYS_MONITOR_LEN; >> +#else >> spl_image.size = 200 * 1024; >> +#endif > > Yes, I noticed this 200 KiB setting as well. But it seems that this > ".size" variable is not referenced at all. Or am I missing something? Your are missing something: drivers/mmc/spl_mmc.c: 55 /* convert size to sectors - round up */ 56 image_size_sectors = (spl_image.size + MMCSD_SECTOR_SIZE - 1) / 57 MMCSD_SECTOR_SIZE; And I have an example where only a part of u-boot is copied. Cheers, Stefano
Hi Stefano, On 08/23/2012 11:23 AM, Stefano Babic wrote: >>> diff --git a/common/spl/spl.c b/common/spl/spl.c >>> index 7d15460..827ff1c 100644 >>> --- a/common/spl/spl.c >>> +++ b/common/spl/spl.c >>> @@ -100,7 +100,11 @@ void spl_parse_image_header(const struct image_header *header) >>> debug("mkimage signature not found - ih_magic = %x\n", >>> header->ih_magic); >>> /* Let's assume U-Boot will not be more than 200 KB */ >>> +#ifdef CONFIG_SYS_MONITOR_LEN >>> + spl_image.size = CONFIG_SYS_MONITOR_LEN; >>> +#else >>> spl_image.size = 200 * 1024; >>> +#endif >> >> Yes, I noticed this 200 KiB setting as well. But it seems that this >> ".size" variable is not referenced at all. Or am I missing something? > > Your are missing something: > > drivers/mmc/spl_mmc.c: > > 55 /* convert size to sectors - round up */ > 56 image_size_sectors = (spl_image.size + MMCSD_SECTOR_SIZE - 1) / > 57 MMCSD_SECTOR_SIZE; > > And I have an example where only a part of u-boot is copied. Ahh, I see. Thanks. I would prefer the following approach then, to move the #ifdef out of the C code: #ifndef CONFIG_SYS_MONITOR_LEN #define CONFIG_SYS_MONITOR_LEN (200 * 1024) #endif and then use CONFIG_SYS_MONITOR_LEN unconditionally. Thanks, Stefan
On 23/08/2012 11:28, Stefan Roese wrote: > Hi Stefano, > > On 08/23/2012 11:23 AM, Stefano Babic wrote: >>>> diff --git a/common/spl/spl.c b/common/spl/spl.c >>>> index 7d15460..827ff1c 100644 >>>> --- a/common/spl/spl.c >>>> +++ b/common/spl/spl.c >>>> @@ -100,7 +100,11 @@ void spl_parse_image_header(const struct image_header *header) >>>> debug("mkimage signature not found - ih_magic = %x\n", >>>> header->ih_magic); >>>> /* Let's assume U-Boot will not be more than 200 KB */ >>>> +#ifdef CONFIG_SYS_MONITOR_LEN >>>> + spl_image.size = CONFIG_SYS_MONITOR_LEN; >>>> +#else >>>> spl_image.size = 200 * 1024; >>>> +#endif >>> >>> Yes, I noticed this 200 KiB setting as well. But it seems that this >>> ".size" variable is not referenced at all. Or am I missing something? >> >> Your are missing something: >> >> drivers/mmc/spl_mmc.c: >> >> 55 /* convert size to sectors - round up */ >> 56 image_size_sectors = (spl_image.size + MMCSD_SECTOR_SIZE - 1) / >> 57 MMCSD_SECTOR_SIZE; >> >> And I have an example where only a part of u-boot is copied. > > Ahh, I see. Thanks. > > I would prefer the following approach then, to move the #ifdef out of > the C code: > > #ifndef CONFIG_SYS_MONITOR_LEN > #define CONFIG_SYS_MONITOR_LEN (200 * 1024) > #endif > > and then use CONFIG_SYS_MONITOR_LEN unconditionally. Fine with me - I will fix in V2. Regards, Stefano
diff --git a/common/spl/spl.c b/common/spl/spl.c index 7d15460..827ff1c 100644 --- a/common/spl/spl.c +++ b/common/spl/spl.c @@ -100,7 +100,11 @@ void spl_parse_image_header(const struct image_header *header) debug("mkimage signature not found - ih_magic = %x\n", header->ih_magic); /* Let's assume U-Boot will not be more than 200 KB */ +#ifdef CONFIG_SYS_MONITOR_LEN + spl_image.size = CONFIG_SYS_MONITOR_LEN; +#else spl_image.size = 200 * 1024; +#endif spl_image.entry_point = CONFIG_SYS_TEXT_BASE; spl_image.load_addr = CONFIG_SYS_TEXT_BASE; spl_image.os = IH_OS_U_BOOT;
If an u-boot image is not found, SPL thinks to load a bare u-boot.bin image with a maximum size of 200KB. Use CONFIG_SYS_MONITOR_LEN instead. Signed-off-by: Stefano Babic <sbabic@denx.de> CC: Tom Rini <trini@ti.com> --- Note: this is based on Tom's series "ARM: SPL: Make more generic, merge DaVinci and OMAP" Tested with V2 version on a MX35. common/spl/spl.c | 4 ++++ 1 file changed, 4 insertions(+)