[PATCH] Fix builtin command line interaction with firmware

[PATCH] Fix builtin command line interaction with firmware

[Amos Waterland]

It seems that prom_init's early_cmdline_parse is broken on at least
Apple 970 xserves and IBM JS20 blades with SLOF.  The firmware of these
machines returns -1 and 1 respectively when getprop is called for the
bootargs property of /chosen, causing Linux to ignore its builtin
command line in favor of a null string.  This patch makes Linux use its
builtin command line if getprop returns an error or a null string.

Signed-off-by: Amos Waterland <apw@us.ibm.com>

---

 arch/powerpc/kernel/prom_init.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

0f0caf0342f39b4437dcb3e9f2f1d62bfb4e3e7d
diff --git a/arch/powerpc/kernel/prom_init.c b/arch/powerpc/kernel/prom_init.c
index f70bd09..2e63a90 100644
--- a/arch/powerpc/kernel/prom_init.c
+++ b/arch/powerpc/kernel/prom_init.c
@@ -574,7 +574,7 @@ static void __init early_cmdline_parse(v
 	if ((long)_prom->chosen > 0)
 		l = prom_getprop(_prom->chosen, "bootargs", p, COMMAND_LINE_SIZE-1);
 #ifdef CONFIG_CMDLINE
-	if (l == 0) /* dbl check */
+	if (l <= 0 || p[0] == '\0') /* dbl check */
 		strlcpy(RELOC(prom_cmd_line),
 			RELOC(CONFIG_CMDLINE), sizeof(prom_cmd_line));
 #endif /* CONFIG_CMDLINE */
-- 
1.0.4

Re: [PATCH] Fix builtin command line interaction with firmware

[Segher Boessenkool]

> It seems that prom_init's early_cmdline_parse is broken on at least
> Apple 970 xserves and IBM JS20 blades with SLOF.  The firmware of  
> these
> machines returns -1 and 1 respectively when getprop is called for the
> bootargs property of /chosen, causing Linux to ignore its builtin
> command line in favor of a null string.  This patch makes Linux use  
> its
> builtin command line if getprop returns an error or a null string.

If the user gave no boot arguments on the boot command, -1 and 1
are the only two valid return values even, and -1 is only valid
if no "boot" or "load" command was used to load the kernel.
Makes me wonder what those XServe's do ;-)

(0 isn't valid as the property is encoded with encode-string, which
puts a trailing 0 byte on the empty string.  Doesn't hurt to check
for it anyway of course ;-) ).

> Signed-off-by: Amos Waterland <apw@us.ibm.com>

Looks good, thanks!

Acked-by: Segher Boessenkool <segher@kernel.crashing.org>


>
> ---
>
>  arch/powerpc/kernel/prom_init.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
>
> 0f0caf0342f39b4437dcb3e9f2f1d62bfb4e3e7d
> diff --git a/arch/powerpc/kernel/prom_init.c b/arch/powerpc/kernel/ 
> prom_init.c
> index f70bd09..2e63a90 100644
> --- a/arch/powerpc/kernel/prom_init.c
> +++ b/arch/powerpc/kernel/prom_init.c
> @@ -574,7 +574,7 @@ static void __init early_cmdline_parse(v
>  	if ((long)_prom->chosen > 0)
>  		l = prom_getprop(_prom->chosen, "bootargs", p,  
> COMMAND_LINE_SIZE-1);
>  #ifdef CONFIG_CMDLINE
> -	if (l == 0) /* dbl check */
> +	if (l <= 0 || p[0] == '\0') /* dbl check */
>  		strlcpy(RELOC(prom_cmd_line),
>  			RELOC(CONFIG_CMDLINE), sizeof(prom_cmd_line));
>  #endif /* CONFIG_CMDLINE */
> -- 
> 1.0.4
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-dev