[U-Boot-Users] [PATCH] make the machid configurable via the environment (Was: [PATCH] Changed load address and entry point for uImage to ZRELADDR + 6MB)
Hello Nicolas,
Nicolas Pitre wrote:
But does U-Boot properly provide machine information to the kernel these days? I have been and still being burned by U-Boot hardcoding the _wrong_ machine ID, and therefore custom built kernels from k.o simply won't boot on those boards without hacking the kernel source.
With "our" U-Boot you can override the compiled-in machid with an environment variable. I didn't implement this patch but I think it's not that difficult. I will check tomorrow if I can easily get a patch out of our scm.
It took a bit longer, but here it comes. This is on top of todays u-boot.
I didn't tested this patch in that version, but this is how it is done in our version of U-Boot (that is considerable older).
Switching the machid is also very convenient because we have modules that can be plugged on different base boards and so can use different machids.
Best regards Uwe
--->8--- From: Uwe Kleine-König Uwe.Kleine-Koenig@digi.com
make the machid configurable via the environment
If the variable "machid" exists, let do_bootm_linux use that instead of bd->bi_arch_number.
Signed-off-by: Uwe Kleine-König Uwe.Kleine-Koenig@digi.com --- lib_arm/armlinux.c | 10 +++++++++- 1 files changed, 9 insertions(+), 1 deletions(-)
diff --git a/lib_arm/armlinux.c b/lib_arm/armlinux.c index 6d32a41..62185f0 100644 --- a/lib_arm/armlinux.c +++ b/lib_arm/armlinux.c @@ -78,6 +78,8 @@ void do_bootm_linux (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[], void (*theKernel)(int zero, int arch, uint params); image_header_t *hdr = &header; bd_t *bd = gd->bd; + int machid = bd->bi_arch_number; + char *s;
#ifdef CONFIG_CMDLINE_TAG char *commandline = getenv ("bootargs"); @@ -85,6 +87,12 @@ void do_bootm_linux (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[],
theKernel = (void (*)(int, int, uint))ntohl(hdr->ih_ep);
+ s = getenv ("machid"); + if (s) { + machid = simple_strtoul (s, NULL, 16); + printf ("Using machid 0x%x from environment\n", machid); + } + /* * Check if there is an initrd image */ @@ -260,7 +268,7 @@ void do_bootm_linux (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[],
cleanup_before_linux ();
- theKernel (0, bd->bi_arch_number, bd->bi_boot_params); + theKernel (0, machid, bd->bi_boot_params); }
Nicolas Pitre wrote:
But does U-Boot properly provide machine information to the kernel these days? I have been and still being burned by U-Boot hardcoding the _wrong_ machine ID, and therefore custom built kernels from k.o simply won't boot on those boards without hacking the kernel source.
With "our" U-Boot you can override the compiled-in machid with an environment variable. I didn't implement this patch but I think it's not that difficult. I will check tomorrow if I can easily get a patch out of our scm.
I proposed this patch for a long time, but it was rejected by Wolfgang.
I am seeing multiple companies implementing small CPU modules incorporating CPU + FPGA, and the customers will adopt the FPGA code for their own need, and then they need to create a new machine id.
Having such a patch will allow people in this situation to use the U-Boot on the board without recompilation.
Best Regards Ulf Samuelsson
On Wed, 30 Jan 2008, Ulf Samuelsson wrote:
Nicolas Pitre wrote:
But does U-Boot properly provide machine information to the kernel these days? I have been and still being burned by U-Boot hardcoding the _wrong_ machine ID, and therefore custom built kernels from k.o simply won't boot on those boards without hacking the kernel source.
With "our" U-Boot you can override the compiled-in machid with an environment variable. I didn't implement this patch but I think it's not that difficult. I will check tomorrow if I can easily get a patch out of our scm.
I proposed this patch for a long time, but it was rejected by Wolfgang.
And may Wolfgang explain just why?
I am seeing multiple companies implementing small CPU modules incorporating CPU + FPGA, and the customers will adopt the FPGA code for their own need, and then they need to create a new machine id.
Having such a patch will allow people in this situation to use the U-Boot on the board without recompilation.
Yes, and it will considerably reduce the grundge against U-Boot from kernel people, myself included.
The current situation with U-Boot simply encourage people to put stupid hacks in the kernel instead in order to work around this U-Boot artificial limitation, which is not bringing any sympathy for better U-Boot support in the kernel build system amongst other things.
So I'm asking for Wolfgang to seriously reconsider accepting this patch.
Nicolas
On Wed, Jan 30, 2008 at 10:33:46AM +0100, Ulf Samuelsson wrote:
Nicolas Pitre wrote:
But does U-Boot properly provide machine information to the kernel these days? I have been and still being burned by U-Boot hardcoding the _wrong_ machine ID, and therefore custom built kernels from k.o simply won't boot on those boards without hacking the kernel source.
With "our" U-Boot you can override the compiled-in machid with an environment variable. I didn't implement this patch but I think it's not that difficult. I will check tomorrow if I can easily get a patch out of our scm.
I proposed this patch for a long time, but it was rejected by Wolfgang.
One solution to that problem is to have someone who maintains a separate tree for uboot which does have the feature in. Then it becomes a matter of community choice whether they use the official tree or the tree with sensible machine-id configurability.
That way, those who want the configurability can have it, and Wolfgang can avoid having his tree polluted with useful features he doesn't want.
In message 20080130154740.GA29330@flint.arm.linux.org.uk you wrote:
I proposed this patch for a long time, but it was rejected by Wolfgang.
One solution to that problem is to have someone who maintains a separate tree for uboot which does have the feature in. Then it becomes a matter of community choice whether they use the official tree or the tree with sensible machine-id configurability.
That way, those who want the configurability can have it, and Wolfgang can avoid having his tree polluted with useful features he doesn't want.
Russel, are you by chance volunteering to become the custodian for the ARM related parts of U-Boot? Just let me know...
BTW: do you actually know what you are commenting about? Did you look up Ulf's proposal and the reject message in the U-Boot mailing list archive?
Best regards,
Wolfgang Denk
On Thu, Jan 31, 2008 at 01:05:44AM +0100, Wolfgang Denk wrote:
In message 20080130154740.GA29330@flint.arm.linux.org.uk you wrote:
I proposed this patch for a long time, but it was rejected by Wolfgang.
One solution to that problem is to have someone who maintains a separate tree for uboot which does have the feature in. Then it becomes a matter of community choice whether they use the official tree or the tree with sensible machine-id configurability.
That way, those who want the configurability can have it, and Wolfgang can avoid having his tree polluted with useful features he doesn't want.
Russel, are you by chance volunteering to become the custodian for the ARM related parts of U-Boot? Just let me know...
Wulfgage, no I am not.
BTW: do you actually know what you are commenting about? Did you look up Ulf's proposal and the reject message in the U-Boot mailing list archive?
Duh. I think everyone knows the answer to those questions who've seen my past comments on this subject - and it's plainly obvious what the answer to the second question is.
In message 000201c8634b$546ee570$8901a8c0@atmel.com you wrote:
I proposed this patch for a long time, but it was rejected by Wolfgang.
Either you did it differently, or I changed my mind.
Best regards,
Wolfgang Denk
In message 20080130080849.GA25875@digi.com you wrote:
From: Uwe Kleine-K=F6nig Uwe.Kleine-Koenig@digi.com
make the machid configurable via the environment
If the variable "machid" exists, let do_bootm_linux use that instead of bd->bi_arch_number.
Signed-off-by: Uwe Kleine-K=F6nig Uwe.Kleine-Koenig@digi.com
Acked-by: Wolfgang Denk wd@denx.de
Best regards,
Wolfgang Denk
In message 20080130080849.GA25875@digi.com you wrote:
make the machid configurable via the environment
If the variable "machid" exists, let do_bootm_linux use that instead of bd->bi_arch_number.
Signed-off-by: Uwe Kleine-K=F6nig Uwe.Kleine-Koenig@digi.com
lib_arm/armlinux.c | 10 +++++++++- 1 files changed, 9 insertions(+), 1 deletions(-)
Applied, thanks.
Best regards,
Wolfgang Denk
participants (5)
-
Nicolas Pitre -
Russell King - ARM Linux -
Ulf Samuelsson -
Uwe Kleine-König -
Wolfgang Denk