1 files changed, 360 insertions, 43 deletions

diff --git a/arch/powerpc/kernel/prom_init.c b/arch/powerpc/kernel/prom_init.c
index c016033ba78..cc584865b3d 100644
--- a/arch/powerpc/kernel/prom_init.c
+++ b/arch/powerpc/kernel/prom_init.c
@@ -43,6 +43,7 @@
 #include <asm/btext.h>
 #include <asm/sections.h>
 #include <asm/machdep.h>
+#include <asm/opal.h>
 
 #include <linux/linux_logo.h>
 
@@ -139,7 +140,9 @@ struct mem_map_entry {
 
 typedef u32 cell_t;
 
-extern void __start(unsigned long r3, unsigned long r4, unsigned long r5);
+extern void __start(unsigned long r3, unsigned long r4, unsigned long r5,
+		    unsigned long r6, unsigned long r7, unsigned long r8,
+		    unsigned long r9);
 
 #ifdef CONFIG_PPC64
 extern int enter_prom(struct prom_args *args, unsigned long entry);
@@ -185,6 +188,7 @@ static unsigned long __initdata prom_tce_alloc_end;
 #define PLATFORM_LPAR		0x0001
 #define PLATFORM_POWERMAC	0x0400
 #define PLATFORM_GENERIC	0x0500
+#define PLATFORM_OPAL		0x0600
 
 static int __initdata of_platform;
 
@@ -644,7 +648,7 @@ static void __init early_cmdline_parse(void)
 	}
 }
 
-#ifdef CONFIG_PPC_PSERIES
+#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
 /*
  * There are two methods for telling firmware what our capabilities are.
  * Newer machines have an "ibm,client-architecture-support" method on the
@@ -1020,7 +1024,7 @@ static unsigned long __init alloc_up(unsigned long size, unsigned long align)
 	}
 	if (addr == 0)
 		return 0;
-	RELOC(alloc_bottom) = addr;
+	RELOC(alloc_bottom) = addr + size;
 
 	prom_debug(" -> %x\n", addr);
 	prom_debug("  alloc_bottom : %x\n", RELOC(alloc_bottom));
@@ -1274,6 +1278,284 @@ static void __init prom_init_mem(void)
 	prom_printf("  ram_top      : %x\n", RELOC(ram_top));
 }
 
+static void __init prom_close_stdin(void)
+{
+	struct prom_t *_prom = &RELOC(prom);
+	ihandle val;
+
+	if (prom_getprop(_prom->chosen, "stdin", &val, sizeof(val)) > 0)
+		call_prom("close", 1, 0, val);
+}
+
+#ifdef CONFIG_PPC_POWERNV
+
+static u64 __initdata prom_opal_size;
+static u64 __initdata prom_opal_align;
+static int __initdata prom_rtas_start_cpu;
+static u64 __initdata prom_rtas_data;
+static u64 __initdata prom_rtas_entry;
+
+#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
+static u64 __initdata prom_opal_base;
+static u64 __initdata prom_opal_entry;
+#endif
+
+/* XXX Don't change this structure without updating opal-takeover.S */
+static struct opal_secondary_data {
+	s64				ack;	/*  0 */
+	u64				go;	/*  8 */
+	struct opal_takeover_args	args;	/* 16 */
+} opal_secondary_data;
+
+extern char opal_secondary_entry;
+
+static void prom_query_opal(void)
+{
+	long rc;
+
+	/* We must not query for OPAL presence on a machine that
+	 * supports TNK takeover (970 blades), as this uses the same
+	 * h-call with different arguments and will crash
+	 */
+	if (PHANDLE_VALID(call_prom("finddevice", 1, 1,
+				    ADDR("/tnk-memory-map")))) {
+		prom_printf("TNK takeover detected, skipping OPAL check\n");
+		return;
+	}
+
+	prom_printf("Querying for OPAL presence... ");
+	rc = opal_query_takeover(&RELOC(prom_opal_size),
+				 &RELOC(prom_opal_align));
+	prom_debug("(rc = %ld) ", rc);
+	if (rc != 0) {
+		prom_printf("not there.\n");
+		return;
+	}
+	RELOC(of_platform) = PLATFORM_OPAL;
+	prom_printf(" there !\n");
+	prom_debug("  opal_size  = 0x%lx\n", RELOC(prom_opal_size));
+	prom_debug("  opal_align = 0x%lx\n", RELOC(prom_opal_align));
+	if (RELOC(prom_opal_align) < 0x10000)
+		RELOC(prom_opal_align) = 0x10000;
+}
+
+static int prom_rtas_call(int token, int nargs, int nret, int *outputs, ...)
+{
+	struct rtas_args rtas_args;
+	va_list list;
+	int i;
+
+	rtas_args.token = token;
+	rtas_args.nargs = nargs;
+	rtas_args.nret  = nret;
+	rtas_args.rets  = (rtas_arg_t *)&(rtas_args.args[nargs]);
+	va_start(list, outputs);
+	for (i = 0; i < nargs; ++i)
+		rtas_args.args[i] = va_arg(list, rtas_arg_t);
+	va_end(list);
+
+	for (i = 0; i < nret; ++i)
+		rtas_args.rets[i] = 0;
+
+	opal_enter_rtas(&rtas_args, RELOC(prom_rtas_data),
+			RELOC(prom_rtas_entry));
+
+	if (nret > 1 && outputs != NULL)
+		for (i = 0; i < nret-1; ++i)
+			outputs[i] = rtas_args.rets[i+1];
+	return (nret > 0)? rtas_args.rets[0]: 0;
+}
+
+static void __init prom_opal_hold_cpus(void)
+{
+	int i, cnt, cpu, rc;
+	long j;
+	phandle node;
+	char type[64];
+	u32 servers[8];
+	struct prom_t *_prom = &RELOC(prom);
+	void *entry = (unsigned long *)&RELOC(opal_secondary_entry);
+	struct opal_secondary_data *data = &RELOC(opal_secondary_data);
+
+	prom_debug("prom_opal_hold_cpus: start...\n");
+	prom_debug("    - entry       = 0x%x\n", entry);
+	prom_debug("    - data        = 0x%x\n", data);
+
+	data->ack = -1;
+	data->go = 0;
+
+	/* look for cpus */
+	for (node = 0; prom_next_node(&node); ) {
+		type[0] = 0;
+		prom_getprop(node, "device_type", type, sizeof(type));
+		if (strcmp(type, RELOC("cpu")) != 0)
+			continue;
+
+		/* Skip non-configured cpus. */
+		if (prom_getprop(node, "status", type, sizeof(type)) > 0)
+			if (strcmp(type, RELOC("okay")) != 0)
+				continue;
+
+		cnt = prom_getprop(node, "ibm,ppc-interrupt-server#s", servers,
+			     sizeof(servers));
+		if (cnt == PROM_ERROR)
+			break;
+		cnt >>= 2;
+		for (i = 0; i < cnt; i++) {
+			cpu = servers[i];
+			prom_debug("CPU %d ... ", cpu);
+			if (cpu == _prom->cpu) {
+				prom_debug("booted !\n");
+				continue;
+			}
+			prom_debug("starting ... ");
+
+			/* Init the acknowledge var which will be reset by
+			 * the secondary cpu when it awakens from its OF
+			 * spinloop.
+			 */
+			data->ack = -1;
+			rc = prom_rtas_call(RELOC(prom_rtas_start_cpu), 3, 1,
+					    NULL, cpu, entry, data);
+			prom_debug("rtas rc=%d ...", rc);
+
+			for (j = 0; j < 100000000 && data->ack == -1; j++) {
+				HMT_low();
+				mb();
+			}
+			HMT_medium();
+			if (data->ack != -1)
+				prom_debug("done, PIR=0x%x\n", data->ack);
+			else
+				prom_debug("timeout !\n");
+		}
+	}
+	prom_debug("prom_opal_hold_cpus: end...\n");
+}
+
+static void prom_opal_takeover(void)
+{
+	struct opal_secondary_data *data = &RELOC(opal_secondary_data);
+	struct opal_takeover_args *args = &data->args;
+	u64 align = RELOC(prom_opal_align);
+	u64 top_addr, opal_addr;
+
+	args->k_image	= (u64)RELOC(_stext);
+	args->k_size	= _end - _stext;
+	args->k_entry	= 0;
+	args->k_entry2	= 0x60;
+
+	top_addr = _ALIGN_UP(args->k_size, align);
+
+	if (RELOC(prom_initrd_start) != 0) {
+		args->rd_image = RELOC(prom_initrd_start);
+		args->rd_size = RELOC(prom_initrd_end) - args->rd_image;
+		args->rd_loc = top_addr;
+		top_addr = _ALIGN_UP(args->rd_loc + args->rd_size, align);
+	}
+
+	/* Pickup an address for the HAL. We want to go really high
+	 * up to avoid problem with future kexecs. On the other hand
+	 * we don't want to be all over the TCEs on P5IOC2 machines
+	 * which are going to be up there too. We assume the machine
+	 * has plenty of memory, and we ask for the HAL for now to
+	 * be just below the 1G point, or above the initrd
+	 */
+	opal_addr = _ALIGN_DOWN(0x40000000 - RELOC(prom_opal_size), align);
+	if (opal_addr < top_addr)
+		opal_addr = top_addr;
+	args->hal_addr = opal_addr;
+
+	/* Copy the command line to the kernel image */
+	strlcpy(RELOC(boot_command_line), RELOC(prom_cmd_line),
+		COMMAND_LINE_SIZE);
+
+	prom_debug("  k_image    = 0x%lx\n", args->k_image);
+	prom_debug("  k_size     = 0x%lx\n", args->k_size);
+	prom_debug("  k_entry    = 0x%lx\n", args->k_entry);
+	prom_debug("  k_entry2   = 0x%lx\n", args->k_entry2);
+	prom_debug("  hal_addr   = 0x%lx\n", args->hal_addr);
+	prom_debug("  rd_image   = 0x%lx\n", args->rd_image);
+	prom_debug("  rd_size    = 0x%lx\n", args->rd_size);
+	prom_debug("  rd_loc     = 0x%lx\n", args->rd_loc);
+	prom_printf("Performing OPAL takeover,this can take a few minutes..\n");
+	prom_close_stdin();
+	mb();
+	data->go = 1;
+	for (;;)
+		opal_do_takeover(args);
+}
+
+/*
+ * Allocate room for and instantiate OPAL
+ */
+static void __init prom_instantiate_opal(void)
+{
+	phandle opal_node;
+	ihandle opal_inst;
+	u64 base, entry;
+	u64 size = 0, align = 0x10000;
+	u32 rets[2];
+
+	prom_debug("prom_instantiate_opal: start...\n");
+
+	opal_node = call_prom("finddevice", 1, 1, ADDR("/ibm,opal"));
+	prom_debug("opal_node: %x\n", opal_node);
+	if (!PHANDLE_VALID(opal_node))
+		return;
+
+	prom_getprop(opal_node, "opal-runtime-size", &size, sizeof(size));
+	if (size == 0)
+		return;
+	prom_getprop(opal_node, "opal-runtime-alignment", &align,
+		     sizeof(align));
+
+	base = alloc_down(size, align, 0);
+	if (base == 0) {
+		prom_printf("OPAL allocation failed !\n");
+		return;
+	}
+
+	opal_inst = call_prom("open", 1, 1, ADDR("/ibm,opal"));
+	if (!IHANDLE_VALID(opal_inst)) {
+		prom_printf("opening opal package failed (%x)\n", opal_inst);
+		return;
+	}
+
+	prom_printf("instantiating opal at 0x%x...", base);
+
+	if (call_prom_ret("call-method", 4, 3, rets,
+			  ADDR("load-opal-runtime"),
+			  opal_inst,
+			  base >> 32, base & 0xffffffff) != 0
+	    || (rets[0] == 0 && rets[1] == 0)) {
+		prom_printf(" failed\n");
+		return;
+	}
+	entry = (((u64)rets[0]) << 32) | rets[1];
+
+	prom_printf(" done\n");
+
+	reserve_mem(base, size);
+
+	prom_debug("opal base     = 0x%x\n", base);
+	prom_debug("opal align    = 0x%x\n", align);
+	prom_debug("opal entry    = 0x%x\n", entry);
+	prom_debug("opal size     = 0x%x\n", (long)size);
+
+	prom_setprop(opal_node, "/ibm,opal", "opal-base-address",
+		     &base, sizeof(base));
+	prom_setprop(opal_node, "/ibm,opal", "opal-entry-address",
+		     &entry, sizeof(entry));
+
+#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
+	RELOC(prom_opal_base) = base;
+	RELOC(prom_opal_entry) = entry;
+#endif
+	prom_debug("prom_instantiate_opal: end...\n");
+}
+
+#endif /* CONFIG_PPC_POWERNV */
 
 /*
  * Allocate room for and instantiate RTAS
@@ -1297,10 +1579,8 @@ static void __init prom_instantiate_rtas(void)
 		return;
 
 	base = alloc_down(size, PAGE_SIZE, 0);
-	if (base == 0) {
-		prom_printf("RTAS allocation failed !\n");
-		return;
-	}
+	if (base == 0)
+		prom_panic("Could not allocate memory for RTAS\n");
 
 	rtas_inst = call_prom("open", 1, 1, ADDR("/rtas"));
 	if (!IHANDLE_VALID(rtas_inst)) {
@@ -1326,6 +1606,12 @@ static void __init prom_instantiate_rtas(void)
 	prom_setprop(rtas_node, "/rtas", "linux,rtas-entry",
 		     &entry, sizeof(entry));
 
+#ifdef CONFIG_PPC_POWERNV
+	/* PowerVN takeover hack */
+	RELOC(prom_rtas_data) = base;
+	RELOC(prom_rtas_entry) = entry;
+	prom_getprop(rtas_node, "start-cpu", &RELOC(prom_rtas_start_cpu), 4);
+#endif
 	prom_debug("rtas base     = 0x%x\n", base);
 	prom_debug("rtas entry    = 0x%x\n", entry);
 	prom_debug("rtas size     = 0x%x\n", (long)size);
@@ -1543,7 +1829,7 @@ static void __init prom_hold_cpus(void)
 		*acknowledge = (unsigned long)-1;
 
 		if (reg != _prom->cpu) {
-			/* Primary Thread of non-boot cpu */
+			/* Primary Thread of non-boot cpu or any thread */
 			prom_printf("starting cpu hw idx %lu... ", reg);
 			call_prom("start-cpu", 3, 0, node,
 				  secondary_hold, reg);
@@ -1652,15 +1938,6 @@ static void __init prom_init_stdout(void)
 		prom_setprop(val, path, "linux,boot-display", NULL, 0);
 }
 
-static void __init prom_close_stdin(void)
-{
-	struct prom_t *_prom = &RELOC(prom);
-	ihandle val;
-
-	if (prom_getprop(_prom->chosen, "stdin", &val, sizeof(val)) > 0)
-		call_prom("close", 1, 0, val);
-}
-
 static int __init prom_find_machine_type(void)
 {
 	struct prom_t *_prom = &RELOC(prom);
@@ -1671,7 +1948,7 @@ static int __init prom_find_machine_type(void)
 	int x;
 #endif
 
-	/* Look for a PowerMac */
+	/* Look for a PowerMac or a Cell */
 	len = prom_getprop(_prom->root, "compatible",
 			   compat, sizeof(compat)-1);
 	if (len > 0) {
@@ -1697,7 +1974,11 @@ static int __init prom_find_machine_type(void)
 		}
 	}
 #ifdef CONFIG_PPC64
-	/* If not a mac, try to figure out if it's an IBM pSeries or any other
+	/* Try to detect OPAL */
+	if (PHANDLE_VALID(call_prom("finddevice", 1, 1, ADDR("/ibm,opal"))))
+		return PLATFORM_OPAL;
+
+	/* Try to figure out if it's an IBM pSeries or any other
 	 * PAPR compliant platform. We assume it is if :
 	 *  - /device_type is "chrp" (please, do NOT use that for future
 	 *    non-IBM designs !
@@ -1830,11 +2111,13 @@ static void __init *make_room(unsigned long *mem_start, unsigned long *mem_end,
 		if (room > DEVTREE_CHUNK_SIZE)
 			room = DEVTREE_CHUNK_SIZE;
 		if (room < PAGE_SIZE)
-			prom_panic("No memory for flatten_device_tree (no room)");
+			prom_panic("No memory for flatten_device_tree "
+				   "(no room)\n");
 		chunk = alloc_up(room, 0);
 		if (chunk == 0)
-			prom_panic("No memory for flatten_device_tree (claim failed)");
-		*mem_end = RELOC(alloc_top);
+			prom_panic("No memory for flatten_device_tree "
+				   "(claim failed)\n");
+		*mem_end = chunk + room;
 	}
 
 	ret = (void *)*mem_start;
@@ -1922,7 +2205,7 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
 	unsigned long soff;
 	unsigned char *valp;
 	static char pname[MAX_PROPERTY_NAME];
-	int l, room;
+	int l, room, has_phandle = 0;
 
 	dt_push_token(OF_DT_BEGIN_NODE, mem_start, mem_end);
 
@@ -2006,19 +2289,26 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
 		valp = make_room(mem_start, mem_end, l, 4);
 		call_prom("getprop", 4, 1, node, RELOC(pname), valp, l);
 		*mem_start = _ALIGN(*mem_start, 4);
+
+		if (!strcmp(RELOC(pname), RELOC("phandle")))
+			has_phandle = 1;
 	}
 
-	/* Add a "linux,phandle" property. */
-	soff = dt_find_string(RELOC("linux,phandle"));
-	if (soff == 0)
-		prom_printf("WARNING: Can't find string index for"
-			    " <linux-phandle> node %s\n", path);
-	else {
-		dt_push_token(OF_DT_PROP, mem_start, mem_end);
-		dt_push_token(4, mem_start, mem_end);
-		dt_push_token(soff, mem_start, mem_end);
-		valp = make_room(mem_start, mem_end, 4, 4);
-		*(u32 *)valp = node;
+	/* Add a "linux,phandle" property if no "phandle" property already
+	 * existed (can happen with OPAL)
+	 */
+	if (!has_phandle) {
+		soff = dt_find_string(RELOC("linux,phandle"));
+		if (soff == 0)
+			prom_printf("WARNING: Can't find string index for"
+				    " <linux-phandle> node %s\n", path);
+		else {
+			dt_push_token(OF_DT_PROP, mem_start, mem_end);
+			dt_push_token(4, mem_start, mem_end);
+			dt_push_token(soff, mem_start, mem_end);
+			valp = make_room(mem_start, mem_end, 4, 4);
+			*(u32 *)valp = node;
+		}
 	}
 
 	/* do all our children */
@@ -2042,7 +2332,7 @@ static void __init flatten_device_tree(void)
 
 	/*
 	 * Check how much room we have between alloc top & bottom (+/- a
-	 * few pages), crop to 4Mb, as this is our "chuck" size
+	 * few pages), crop to 1MB, as this is our "chunk" size
 	 */
 	room = RELOC(alloc_top) - RELOC(alloc_bottom) - 0x4000;
 	if (room > DEVTREE_CHUNK_SIZE)
@@ -2053,7 +2343,7 @@ static void __init flatten_device_tree(void)
 	mem_start = (unsigned long)alloc_up(room, PAGE_SIZE);
 	if (mem_start == 0)
 		prom_panic("Can't allocate initial device-tree chunk\n");
-	mem_end = RELOC(alloc_top);
+	mem_end = mem_start + room;
 
 	/* Get root of tree */
 	root = call_prom("peer", 1, 1, (phandle)0);
@@ -2502,6 +2792,7 @@ static void __init prom_check_initrd(unsigned long r3, unsigned long r4)
 #endif /* CONFIG_BLK_DEV_INITRD */
 }
 
+
 /*
  * We enter here early on, when the Open Firmware prom is still
  * handling exceptions and the MMU hash table for us.
@@ -2551,6 +2842,7 @@ unsigned long __init prom_init(unsigned long r3, unsigned long r4,
 	 * between pSeries SMP and pSeries LPAR
 	 */
 	RELOC(of_platform) = prom_find_machine_type();
+	prom_printf("Detected machine type: %x\n", RELOC(of_platform));
 
 #ifndef CONFIG_RELOCATABLE
 	/* Bail if this is a kdump kernel. */
@@ -2563,7 +2855,7 @@ unsigned long __init prom_init(unsigned long r3, unsigned long r4,
 	 */
 	prom_check_initrd(r3, r4);
 
-#ifdef CONFIG_PPC_PSERIES
+#if defined(CONFIG_PPC_PSERIES) || defined(CONFIG_PPC_POWERNV)
 	/*
 	 * On pSeries, inform the firmware about our capabilities
 	 */
@@ -2609,14 +2901,33 @@ unsigned long __init prom_init(unsigned long r3, unsigned long r4,
 #endif
 
 	/*
-	 * On non-powermacs, try to instantiate RTAS and puts all CPUs
-	 * in spin-loops. PowerMacs don't have a working RTAS and use
-	 * a different way to spin CPUs
+	 * On non-powermacs, try to instantiate RTAS. PowerMacs don't
+	 * have a usable RTAS implementation.
 	 */
-	if (RELOC(of_platform) != PLATFORM_POWERMAC) {
+	if (RELOC(of_platform) != PLATFORM_POWERMAC &&
+	    RELOC(of_platform) != PLATFORM_OPAL)
 		prom_instantiate_rtas();
+
+#ifdef CONFIG_PPC_POWERNV
+	/* Detect HAL and try instanciating it & doing takeover */
+	if (RELOC(of_platform) == PLATFORM_PSERIES_LPAR) {
+		prom_query_opal();
+		if (RELOC(of_platform) == PLATFORM_OPAL) {
+			prom_opal_hold_cpus();
+			prom_opal_takeover();
+		}
+	} else if (RELOC(of_platform) == PLATFORM_OPAL)
+		prom_instantiate_opal();
+#endif
+
+	/*
+	 * On non-powermacs, put all CPUs in spin-loops.
+	 *
+	 * PowerMacs use a different mechanism to spin CPUs
+	 */
+	if (RELOC(of_platform) != PLATFORM_POWERMAC &&
+	    RELOC(of_platform) != PLATFORM_OPAL)
 		prom_hold_cpus();
-	}
 
 	/*
 	 * Fill in some infos for use by the kernel later on
@@ -2683,7 +2994,13 @@ unsigned long __init prom_init(unsigned long r3, unsigned long r4,
 	reloc_got2(-offset);
 #endif
 
-	__start(hdr, kbase, 0);
+#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
+	/* OPAL early debug gets the OPAL base & entry in r8 and r9 */
+	__start(hdr, kbase, 0, 0, 0,
+		RELOC(prom_opal_base), RELOC(prom_opal_entry));
+#else
+	__start(hdr, kbase, 0, 0, 0, 0, 0);
+#endif
 
 	return 0;
 }