INTERACTIVE_CONSOLE for misc-embedded.c and watchdog issue

Hi Tom,

The following patch against misc-embedded.c adds an INTERACTIVE_CONSOLE
#define to guard reading from console (causes unecessary delay in some
situations).

Its an adaptation of misc.c's define, a difference being that platform
headers define "NO_INTERACTIVE_CONSOLE" if required.

What do you think?

Another issue is the watchdog - if it is active all 8xx systems
should fail to boot because load_kernel() (and the gzip functions)
don't calm the dog.

Our internal kernels always have sprinkled "service_wdt()" calls
around "load_kernel()".

Adding hooks for the wdt to be service there (and possibly other places)
is terrible, but is there any other solution so early?

Setting up a HW timer with early setup code might be possible, 
but sounds quite complicated.

Of course, with U-Boot all of this is a non-issue, but for simple
bootloaders (which I suppose are quite common) it is a nasty problem.

diff --git a/arch/ppc/boot/simple/misc-embedded.c b/arch/ppc/boot/simple/misc-embedded.c
index 3865f3f..85e3e24 100644
--- a/arch/ppc/boot/simple/misc-embedded.c
+++ b/arch/ppc/boot/simple/misc-embedded.c

@@ -23,6 +23,10 @@
 
 #include "nonstdio.h"
 
+#ifndef NO_INTERACTIVE_CONSOLE
+#define INTERACTIVE_CONSOLE
+#endif
+
 /* The linker tells us where the image is. */
 extern char __image_begin, __image_end;
 extern char __ramdisk_begin, __ramdisk_end;

@@ -82,8 +86,11 @@ embed_config(bd_t **bdp)
 unsigned long
 load_kernel(unsigned long load_addr, int num_words, unsigned long cksum, bd_t *bp)
 {
+#ifdef INTERACTIVE_CONSOLE
 	char *cp, ch;
-	int timer = 0, zimage_size;
+	int timer = 0;
+#endif
+	int zimage_size;
 	unsigned long initrd_size;
 
 	/* First, capture the embedded board information.  Then

@@ -187,6 +194,8 @@ load_kernel(unsigned long load_addr, int
 #endif
 	while ( *cp )
 		putc(*cp++);
+
+#ifdef INTERACTIVE_CONSOLE
 	while (timer++ < 5*1000) {
 		if (tstc()) {
 			while ((ch = getc()) != '\n' && ch != '\r') {

@@ -211,6 +220,7 @@ load_kernel(unsigned long load_addr, int
 		udelay(1000);  /* 1 msec */
 	}
 	*cp = 0;
+#endif
 	puts("\nUncompressing Linux...");
 
 	gunzip(0, 0x400000, zimage_start, &zimage_size);

On Thu, Nov 17, 2005 at 01:17:30PM -0200, Marcelo Tosatti wrote:

The following patch against misc-embedded.c adds an INTERACTIVE_CONSOLE
#define to guard reading from console (causes unecessary delay in some
situations).

Its an adaptation of misc.c's define, a difference being that platform
headers define "NO_INTERACTIVE_CONSOLE" if required.

What do you think?

Please put a comment above the #ifndef stating where the define should
be.  But I have a feeling that's going to be somewhere under
arch/ppc/platforms/ or include/asm-ppc/ and I don't know if that's a
good idea...

I've got a vauge idea on how we can handle this a little bit more
cleanly, in the merge tree.

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

Hi Tom,

On Mon, Nov 28, 2005 at 04:27:38PM -0700, Tom Rini wrote:

On Thu, Nov 17, 2005 at 01:17:30PM -0200, Marcelo Tosatti wrote:

quoted

The following patch against misc-embedded.c adds an INTERACTIVE_CONSOLE
#define to guard reading from console (causes unecessary delay in some
situations).

Its an adaptation of misc.c's define, a difference being that platform
headers define "NO_INTERACTIVE_CONSOLE" if required.

What do you think?

Please put a comment above the #ifndef stating where the define should
be.  But I have a feeling that's going to be somewhere under
arch/ppc/platforms/ or include/asm-ppc/ and I don't know if that's a
good idea...

arch/ppc/platforms/ surely. You have a problem with that?

I've got a vauge idea on how we can handle this a little bit more
cleanly, in the merge tree.

How's that? Should I cancel the patch, then?