Fix initrd/dtb interaction

The original code would wrongly relocate the blob to be right before the initrd if it existed. The blob *must* be within CFG_BOOTMAPSZ, if it is defined. So we make two changes:

1) flag the blob for relocation whenever its address is above BOOTMAPSZ

2) If the blob is being relocated, relocate it before kbd, not initrd

---

Fixed the merge problem. I thought I had based off of the fdt tree, but it was a patch from gvb which was in the mainline and not my tree, so I'm confused. Either way, it's fixed now.

 common/cmd_bootm.c |   23 ++++++++++++++---------
 1 files changed, 14 insertions(+), 9 deletions(-)

index e19f83e..8249dce 100644
@@ -924,6 +924,15 @@ do_bootm_linux (cmd_tbl_t *cmdtp, int flag,
 		initrd_end = 0;
 	}
 
+#ifdef CFG_BOOTMAPSZ
+	/*
+	 * The blob must be within CFG_BOOTMAPSZ,
+	 * so we flag it to be copied if it is
+	 */
+	if (of_flat_tree >= (char *)CFG_BOOTMAPSZ)
+		of_data = of_flat_tree;
+#endif
+
 #if defined(CONFIG_OF_LIBFDT)
 	/* move of_flat_tree if needed */
 	if (of_data) {

@@ -931,11 +940,9 @@ do_bootm_linux (cmd_tbl_t *cmdtp, int flag,
 		ulong of_start, of_len;
 
 		of_len = be32_to_cpu(fdt_totalsize(of_data));
-		/* position on a 4K boundary before the initrd/kbd */
-		if (initrd_start)
-			of_start = initrd_start - of_len;
-		else
-			of_start  = (ulong)kbd - of_len;
+
+		/* position on a 4K boundary before the kbd */
+		of_start  = (ulong)kbd - of_len;
 		of_start &= ~(4096 - 1);	/* align on page */
 		debug ("## device tree at 0x%08lX ... 0x%08lX (len=%ld=0x%lX)\n",
 			of_data, of_data + of_len - 1, of_len, of_len);

@@ -983,11 +990,9 @@ do_bootm_linux (cmd_tbl_t *cmdtp, int flag,
 	if (of_data) {
 		ulong of_start, of_len;
 		of_len = ((struct boot_param_header *)of_data)->totalsize;
+
 		/* provide extra 8k pad */
-		if (initrd_start)
-			of_start = initrd_start - of_len - 8192;
-		else
-			of_start  = (ulong)kbd - of_len - 8192;
+		of_start  = (ulong)kbd - of_len - 8192;
 		of_start &= ~(4096 - 1);	/* align on page */
 		debug ("## device tree at 0x%08lX ... 0x%08lX (len=%ld=0x%lX)\n",
 			of_data, of_data + of_len - 1, of_len, of_len);

1.5.0.2.230.gfbe3d-dirty

Applied, thanks. [May take a while until it's visible on the public server, though; got some more stuff in the queue here.]

Best regards,

Wolfgang Denk

DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Status quo. Latin for "the mess we're in."

Hi Andy,

I get the following error with your patch on platforms not using device tree at all (like 4xx sequoia...):

cmd_bootm.c: In function 'do_bootm_linux':
cmd_bootm.c:932: error: 'of_flat_tree' undeclared (first use in this function) cmd_bootm.c:932: error: (Each undeclared identifier is reported only once cmd_bootm.c:932: error: for each function it appears in.) cmd_bootm.c:933: error: 'of_data' undeclared (first use in this function) make[1]: *** [cmd_bootm.o] Error 1

Would be great if you could generate a quick patch for that.

Thanks.

Best regards,
Stefan

=====================================================================
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-0 Fax: +49-8142-66989-80 Email: office at denx.de
=====================================================================