2004-03-28 11:34:26

by Paul Mackerras

Subject: [PATCH] further early_param fixes

Andrew & Tom,

Tom's patches to add early_param also added some code to read the
command line from nvram in setup_arch on PReP boxes. This broke
powermacs, though, since init_prep_nvram() calls the
ppc_md.nvram_read_val pointer, which is initialized later on
powermac. In any case we don't want to call the prep nvram code on
powermacs. This patch fixes it, and also eliminates a compile warning
about early_adb_sync.

Paul.

diff -urN linux-2.6.5-rc2-mm4.orig/arch/ppc/kernel/setup.c linux-2.6.5-rc2-mm4/arch/ppc/kernel/setup.c
--- linux-2.6.5-rc2-mm4.orig/arch/ppc/kernel/setup.c 2004-03-28 15:46:22.621884568 +1000
+++ linux-2.6.5-rc2-mm4/arch/ppc/kernel/setup.c 2004-03-28 16:26:20.297905848 +1000
@@ -470,10 +470,11 @@

#ifdef CONFIG_ADB
/* Allow us to say that ADB probing will be done synchronously. */
-static void __init early_adb_sync(char **ign)
+static int __init early_adb_sync(char *ign)
{
extern int __adb_probe_sync;
__adb_probe_sync = 1;
+ return 0;
}
__early_param("adb_sync", early_adb_sync);
#endif /* CONFIG_ADB */
@@ -663,13 +664,15 @@

/* See if we need to grab the command line params from PPCBUG. */
#ifdef CONFIG_PPCBUG_NVRAM
- /* Read in NVRAM data */
- init_prep_nvram();
-
- if (cmd_line[0] == '\0') {
- char *bootargs = prep_nvram_get_var("bootargs");
- if (bootargs != NULL)
- strlcpy(cmd_line, bootargs, sizeof(cmd_line));
+ if (_machine == _MACH_prep) {
+ /* Read in NVRAM data */
+ init_prep_nvram();
+
+ if (cmd_line[0] == '\0') {
+ char *bootargs = prep_nvram_get_var("bootargs");
+ if (bootargs != NULL)
+ strlcpy(cmd_line, bootargs, sizeof(cmd_line));
+ }
}
#endif

2004-03-29 15:04:19

by Tom Rini

Subject: Re: [PATCH] further early_param fixes

On Sun, Mar 28, 2004 at 08:23:28PM +1000, Paul Mackerras wrote:

> Andrew & Tom,
>
> Tom's patches to add early_param also added some code to read the
> command line from nvram in setup_arch on PReP boxes. This broke
> powermacs, though, since init_prep_nvram() calls the
> ppc_md.nvram_read_val pointer, which is initialized later on
> powermac. In any case we don't want to call the prep nvram code on
> powermacs.

But we do want to call it on other configs where _machine == 0. So how
about:

diff -urN linux-2.6.5-rc2-mm4.orig/arch/ppc/kernel/setup.c linux-2.6.5-rc2-mm4/arch/ppc/kernel/setup.c
--- linux-2.6.5-rc2-mm4.orig/arch/ppc/kernel/setup.c 2004-03-28 15:46:22.621884568 +1000
+++ linux-2.6.5-rc2-mm4/arch/ppc/kernel/setup.c 2004-03-28 16:26:20.297905848 +1000
@@ -663,13 +664,18 @@

/* See if we need to grab the command line params from PPCBUG. */
#ifdef CONFIG_PPCBUG_NVRAM
- /* Read in NVRAM data */
- init_prep_nvram();
-
- if (cmd_line[0] == '\0') {
- char *bootargs = prep_nvram_get_var("bootargs");
- if (bootargs != NULL)
- strlcpy(cmd_line, bootargs, sizeof(cmd_line));
+#ifdef CONFIG_PPC_PMAC
+ if (_machine == _MACH_prep)
+#endif
+ {
+ /* Read in NVRAM data */
+ init_prep_nvram();
+
+ if (cmd_line[0] == '\0') {
+ char *bootargs = prep_nvram_get_var("bootargs");
+ if (bootargs != NULL)
+ strlcpy(cmd_line, bootargs, sizeof(cmd_line));
+ }
}
#endif


> This patch fixes it, and also eliminates a compile warning
> about early_adb_sync.

Whoops, good catch.

--
Tom Rini
http://gate.crashing.org/~trini/