On Sun, Mar 6, 2011 at 6:33 PM, Loïc Minier <email address hidden> wrote:
> On Sun, Mar 06, 2011, Eric Miao wrote:
>> >> + kernel_suffix = 'linaro-imx5'
>> >
>> > The kernel suffix we use for i.MX51 boards is currently linaro-mx51;
>> > for which kernel is the above suffix? Is it a BSP kernel, and if so
>> > why does it have linaro in the name? If not, could we switch all the
>> > i.MX51 boards to use it instead of linaro-mx51? It would be nice to
>> > have a single kernel for all i.MX51 and i.MX53 boards.
>>
>> That's my intention too. The problem with upstream kernel is that
>> the RUNTIME PHYS_OFFSET is not yet there, thus preventing a
>> single kernel image built for i.MX51 and i.MX53, although work is in
>> smooth progress, and very hopefully we'll see it in 2.6.39.
>>
>> And for 11.05 release, we are using a kernel based on Freescale's
>> 2.6.35 BSP, which supports a single kernel for both i.MX51/53
>> (with those relevant patches backported)
>
> Ok; it seems we wont linux-linaro imx51 + imx53 support by 11.05? Or
> is there any chance that it happens?
Unless we use Freescale's BSP
>
> Does your BSP-based kernel add value to imx51 boards too, so that we
> should support installation of your BSP-based kernel on imx51 boards
> like mx51evk, or efikamx?
I think not, though we are still trying on that. Me still waiting for my
babbage to clear the Customs.
>
>> > Any reason this is called "imx5" instead of "mx5"? Is this to use the
>> > same name as SOC_IMX50, SOC_IMX51, SOC_IMX53? SOC_ is used a lot in
>> > i.MX mach-* subtrees, but not so much in other trees; maybe we should
>> > use the name from the mach-* subdirectories to name our suffixes?
>> > (linaro-mx5)
>> No specific reason for that. mx5 is perfectly fine.
>
> Ok; my vote would go for linaro-mx5 for linux-linaro based kernels when
> it starts supporting both imx51 and imx53; the BSP-based kernel should
> probably be named lp-mx5 or bsp-mx5, or at least not carry the "linaro"
> name since they are not really built straight from linux-linaro.
>
> Looking at the existing "lt-foo" hwpacks, I see:
> http://snapshots.linaro.org/11.05-daily/linaro-hwpacks/lt-u8500/latest/0/config/hwpacks/linaro-u8500
> uses linux-u8500 as package name, /boot/vmlinuz-2.6.35-1000-u8500 as
> filename, while:
> http://snapshots.linaro.org/11.05-daily/linaro-hwpacks/lt-s5pv310/latest/0/config/hwpacks/linaro-lt-s5pv310
> uses linux-image-2.6.37-1000-s5pv310 as package name (probably didn't
> upload a meta), /boot/vmlinuz-2.6.38-1000-s5pv310 as filename.
>
> Since these are PPA only, the name doesn't matter too much, but it
> would be nice if it was unambiguous/non-misleading and consistent; I'm
> sure John Rigby and Jamie Bennett can help ensure the package names are
> consistent, I will start discusson on this. I would find using
> "lt-$CommonSocName" the clearest. It's easy to update
> linaro-image-tools before 11.05 to whatever you pick, but it's a pain
> to change it once linaro-image-tools has been part of
> the 11.05 release with this name supported.
>
> So we can merge this "linaro-imx5" name right now, but I wouldn't mind
> if we had a bug tracking that this name needs to be fixed.
>
>> > Is there a reason why uboot_imx_file would ever be missing from the
>> > hwpack?
>>
>> We're currently using the u-boot from Freescale's BSP since the
>> support in upstream is not there yet. Until the mx53 loco patches
>> are upstreamed, we can definitely remove this.
> [...]
>> No, the u-boot.bin built from Freescale's BSP is the one padded with
>> the MBR. It's just painful to explain this. The u-boot.imx is generated
>> with a script in upstream, while there is no such thing in Freescale's
>> BSP (cuz it's basing on a earlier u-boot version). So instead, they're
>> generating a different u-boot.bin.
>
> Ok; I worked with Freescale u-boot tree a while ago, and I indeed
> remember they generate a padded u-boot.bin; I was confused because the
> config was also adding support for an optional "u-boot.imx", and this
> meant a mainline u-boot for me.
>
> It seems to me that support for mx53loco has been completed on top of
> the upstream u-boot tree; I've sent a message to John Rigby to request
> inclusion of the patches if that's possible. If not, I guess you could
> just create your own tree based on u-boot-linaro + these patches.
>
> This way, we wouldn't have to worry about the specificities of the
> Freescale u-boot tree anymore, and it's padded u-boot.bin.
>
> Is this ok? Or do we really need to support FSL's padded u-boot.bin?
> I would hope we can entirely avoid supporting this intermediate
> situation in the tool.
>
>
> If you're ok with all the above, could you merge my branch into yours
> and push yours again?
>
> --
> Loïc Minier
>
On Sun, Mar 6, 2011 at 6:33 PM, Loïc Minier <email address hidden> wrote:
> On Sun, Mar 06, 2011, Eric Miao wrote:
>> >> + kernel_suffix = 'linaro-imx5'
>> >
>> > The kernel suffix we use for i.MX51 boards is currently linaro-mx51;
>> > for which kernel is the above suffix? Is it a BSP kernel, and if so
>> > why does it have linaro in the name? If not, could we switch all the
>> > i.MX51 boards to use it instead of linaro-mx51? It would be nice to
>> > have a single kernel for all i.MX51 and i.MX53 boards.
>>
>> That's my intention too. The problem with upstream kernel is that
>> the RUNTIME PHYS_OFFSET is not yet there, thus preventing a
>> single kernel image built for i.MX51 and i.MX53, although work is in
>> smooth progress, and very hopefully we'll see it in 2.6.39.
>>
>> And for 11.05 release, we are using a kernel based on Freescale's
>> 2.6.35 BSP, which supports a single kernel for both i.MX51/53
>> (with those relevant patches backported)
>
> Ok; it seems we wont linux-linaro imx51 + imx53 support by 11.05? Or
> is there any chance that it happens?
Unless we use Freescale's BSP
>
> Does your BSP-based kernel add value to imx51 boards too, so that we
> should support installation of your BSP-based kernel on imx51 boards
> like mx51evk, or efikamx?
I think not, though we are still trying on that. Me still waiting for my
babbage to clear the Customs.
> snapshots. linaro. org/11. 05-daily/ linaro- hwpacks/ lt-u8500/ latest/ 0/config/ hwpacks/ linaro- u8500 2.6.35- 1000-u8500 as snapshots. linaro. org/11. 05-daily/ linaro- hwpacks/ lt-s5pv310/ latest/ 0/config/ hwpacks/ linaro- lt-s5pv310 2.6.37- 1000-s5pv310 as package name (probably didn't 2.6.38- 1000-s5pv310 as filename. non-misleading and consistent; I'm
>> > Any reason this is called "imx5" instead of "mx5"? Is this to use the
>> > same name as SOC_IMX50, SOC_IMX51, SOC_IMX53? SOC_ is used a lot in
>> > i.MX mach-* subtrees, but not so much in other trees; maybe we should
>> > use the name from the mach-* subdirectories to name our suffixes?
>> > (linaro-mx5)
>> No specific reason for that. mx5 is perfectly fine.
>
> Ok; my vote would go for linaro-mx5 for linux-linaro based kernels when
> it starts supporting both imx51 and imx53; the BSP-based kernel should
> probably be named lp-mx5 or bsp-mx5, or at least not carry the "linaro"
> name since they are not really built straight from linux-linaro.
>
> Looking at the existing "lt-foo" hwpacks, I see:
> http://
> uses linux-u8500 as package name, /boot/vmlinuz-
> filename, while:
> http://
> uses linux-image-
> upload a meta), /boot/vmlinuz-
>
> Since these are PPA only, the name doesn't matter too much, but it
> would be nice if it was unambiguous/
> sure John Rigby and Jamie Bennett can help ensure the package names are
> consistent, I will start discusson on this. I would find using
> "lt-$CommonSocName" the clearest. It's easy to update
> linaro-image-tools before 11.05 to whatever you pick, but it's a pain
> to change it once linaro-image-tools has been part of
> the 11.05 release with this name supported.
>
> So we can merge this "linaro-imx5" name right now, but I wouldn't mind
> if we had a bug tracking that this name needs to be fixed.
>
>> > Is there a reason why uboot_imx_file would ever be missing from the
>> > hwpack?
>>
>> We're currently using the u-boot from Freescale's BSP since the
>> support in upstream is not there yet. Until the mx53 loco patches
>> are upstreamed, we can definitely remove this.
> [...]
>> No, the u-boot.bin built from Freescale's BSP is the one padded with
>> the MBR. It's just painful to explain this. The u-boot.imx is generated
>> with a script in upstream, while there is no such thing in Freescale's
>> BSP (cuz it's basing on a earlier u-boot version). So instead, they're
>> generating a different u-boot.bin.
>
> Ok; I worked with Freescale u-boot tree a while ago, and I indeed
> remember they generate a padded u-boot.bin; I was confused because the
> config was also adding support for an optional "u-boot.imx", and this
> meant a mainline u-boot for me.
>
> It seems to me that support for mx53loco has been completed on top of
> the upstream u-boot tree; I've sent a message to John Rigby to request
> inclusion of the patches if that's possible. If not, I guess you could
> just create your own tree based on u-boot-linaro + these patches.
>
> This way, we wouldn't have to worry about the specificities of the
> Freescale u-boot tree anymore, and it's padded u-boot.bin.
>
> Is this ok? Or do we really need to support FSL's padded u-boot.bin?
> I would hope we can entirely avoid supporting this intermediate
> situation in the tool.
>
>
> If you're ok with all the above, could you merge my branch into yours
> and push yours again?
>
> --
> Loïc Minier
>