ralink: Introduce fw_passed_dtb to arch/mips/ralink

This patch adds fw_passed_dtb to arch/mips/ralink to support CONFIG_MIPS_RAW_APPENDED_DTB. Furthermore it adds a check that __dtb_start is not the same address as __dtb_end.

---

@@ -66,13 +66,21 @@
 
 void __init plat_mem_setup(void)
 {
+	void *dtb;
+
 	set_io_port_base(KSEG1);
 
 	/*
 	 * Load the builtin devicetree. This causes the chosen node to be
 	 * parsed resulting in our memory appearing
 	 */
-	__dt_setup_arch(__dtb_start);
+	if (fw_passed_dtb) /* used by CONFIG_MIPS_APPENDED_RAW_DTB as well */
+		dtb = (void *)fw_passed_dtb;
+	else if (__dtb_start != __dtb_end)
+		dtb = (void *)__dtb_start;
+
+	if (dtb)
+		__dt_setup_arch(dtb);
 
 	of_scan_flat_dt(early_init_dt_find_memory, NULL);
 	if (memory_dtb)

An empty __dtb_start to __dtb_end section might result in initial_boot_params being null for arch/mips/ralink. This showed that the boot process hangs indefinitely in of_scan_flat_dt().

---

@@ -628,6 +628,9 @@
 				     void *data),
 			   void *data)
 {
+	if (!initial_boot_params)
+		return;
+
 	const void *blob = initial_boot_params;
 	const char *pathp;
 	int offset, rc = 0, depth = -1;

Am Montag, 21. November 2016, 20:21:03 CET schrieb Sergei Shtylyov:

Dear Sergei,

Thanks for that note. In fact it returns whatever it provides. Question is, what would be the right value.

int offset, rc = 0, depth = -1;

rc defaults to 0 but something like EINVAL would better reflect that something is wrong.

Best regards
Tobias

An empty __dtb_start to __dtb_end section might result in initial_boot_params being null for arch/mips/ralink. This showed that the boot process hangs indefinitely in of_scan_flat_dt().

---

@@ -628,6 +628,9 @@
 				     void *data),
 			   void *data)
 {
+	if (!initial_boot_params)
+		return 0;
+
 	const void *blob = initial_boot_params;
 	const char *pathp;
 	int offset, rc = 0, depth = -1;

---

Dear Sergei,

After checking the use of "of_scan_flat_dt()" I revised the patch to return 0 as any other value would most likely break code in:

Best regards
Tobias

Hello.

It still causes a compiler warning -- you can't mix code and declarations
in C90.

MBR, Sergei

An empty __dtb_start to __dtb_end section might result in initial_boot_params being null for arch/mips/ralink. This showed that the boot process hangs indefinitely in of_scan_flat_dt().

---

@@ -738,9 +738,12 @@
 	const char *pathp;
 	int offset, rc = 0, depth = -1;
 
-        for (offset = fdt_next_node(blob, -1, &depth);
-             offset >= 0 && depth >= 0 && !rc;
-             offset = fdt_next_node(blob, offset, &depth)) {
+	if (!blob)
+		return 0;
+
+	for (offset = fdt_next_node(blob, -1, &depth);
+	     offset >= 0 && depth >= 0 && !rc;
+	     offset = fdt_next_node(blob, offset, &depth)) {
 
 		pathp = fdt_get_name(blob, offset, NULL);
 		if (*pathp == '/')

Dear Sergei,

Missed that warning completely during compilation of a testable image for my device. I regenerated the patch based on your input (for 4.9-rc6 this time) and based the check on the local blob variable this time.

Haven't seen any warnings this time.

Hope it's correct that I reference the new patch version each time in the subject line.

Best regards
Tobias

Btw.: Last e-mail I wanted to list occurrences EINVAL would break existing code. One is kernel/prom.c in arch/microblaze.

This patch adds fw_passed_dtb to arch/mips/ralink to support CONFIG_MIPS_RAW_APPENDED_DTB. Furthermore it adds a check that __dtb_start is not the same address as __dtb_end.

---

@@ -66,13 +66,21 @@
 
 void __init plat_mem_setup(void)
 {
+	void *dtb = NULL;
+
 	set_io_port_base(KSEG1);
 
 	/*
 	 * Load the builtin devicetree. This causes the chosen node to be
-	 * parsed resulting in our memory appearing
+	 * parsed resulting in our memory appearing. fw_passed_dtb is used
+	 * by CONFIG_MIPS_APPENDED_RAW_DTB as well.
 	 */
-	__dt_setup_arch(__dtb_start);
+	if (fw_passed_dtb)
+		dtb = (void *)fw_passed_dtb;
+	else if (__dtb_start != __dtb_end)
+		dtb = (void *)__dtb_start;
+
+	__dt_setup_arch(dtb);
 
 	of_scan_flat_dt(early_init_dt_find_memory, NULL);
 	if (memory_dtb)

---
This version has been cleaned up based on feedback [1] of John Crispin for the LEDE Project.

[1] https://github.com/lede-project/source/pull/582#discussion_r90778573

Best regards
Tobias

there should be a chunk here explaining the changes between v1 and v2 ... anyhow