MIPS: Octeon: Add kexec and kdump support

[ralf@linux-mips.org: Original patch by Maxim Uvarov <muvarov@gmail.com>
with plenty of further shining, polishing, debugging and testing by me.]

Signed-off-by: Maxim Uvarov <muvarov@gmail.com>
Cc: linux-mips@linux-mips.org
Cc: kexec@lists.infradead.org
Cc: horms@verge.net.au
Patchwork: https://patchwork.linux-mips.org/patch/1026/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

6 files changed, 335 insertions, 9 deletions

diff --git a/arch/mips/Makefile b/arch/mips/Makefile
index 654b1ad39f0..033395ec82f 100644
--- a/arch/mips/Makefile
+++ b/arch/mips/Makefile
@@ -192,6 +192,10 @@ endif
 #
 include $(srctree)/arch/mips/Kbuild.platforms
 
+ifdef CONFIG_PHYSICAL_START
+load-y                                  = $(CONFIG_PHYSICAL_START)
+endif
+
 cflags-y			+= -I$(srctree)/arch/mips/include/asm/mach-generic
 drivers-$(CONFIG_PCI)		+= arch/mips/pci/
 
diff --git a/arch/mips/cavium-octeon/executive/cvmx-bootmem.c b/arch/mips/cavium-octeon/executive/cvmx-bootmem.c
index fdf5f19bfdb..6d5ddbc112c 100644
--- a/arch/mips/cavium-octeon/executive/cvmx-bootmem.c
+++ b/arch/mips/cavium-octeon/executive/cvmx-bootmem.c
@@ -688,3 +688,8 @@ int64_t cvmx_bootmem_phy_named_block_alloc(uint64_t size, uint64_t min_addr,
 		cvmx_spinlock_unlock((cvmx_spinlock_t *)&(cvmx_bootmem_desc->lock));
 	return addr_allocated;
 }
+
+struct cvmx_bootmem_desc *cvmx_bootmem_get_desc(void)
+{
+	return cvmx_bootmem_desc;
+}
diff --git a/arch/mips/cavium-octeon/setup.c b/arch/mips/cavium-octeon/setup.c
index 67aa3b942f0..7c2b7aab77b 100644
--- a/arch/mips/cavium-octeon/setup.c
+++ b/arch/mips/cavium-octeon/setup.c
@@ -25,6 +25,7 @@
 #include <linux/serial_8250.h>
 #include <linux/of_fdt.h>
 #include <linux/libfdt.h>
+#include <linux/kexec.h>
 
 #include <asm/processor.h>
 #include <asm/reboot.h>
@@ -58,11 +59,208 @@ struct octeon_boot_descriptor *octeon_boot_desc_ptr;
 struct cvmx_bootinfo *octeon_bootinfo;
 EXPORT_SYMBOL(octeon_bootinfo);
 
+static unsigned long long RESERVE_LOW_MEM = 0ull;
+#ifdef CONFIG_KEXEC
+#ifdef CONFIG_SMP
+/*
+ * Wait for relocation code is prepared and send
+ * secondary CPUs to spin until kernel is relocated.
+ */
+static void octeon_kexec_smp_down(void *ignored)
+{
+	int cpu = smp_processor_id();
+
+	local_irq_disable();
+	set_cpu_online(cpu, false);
+	while (!atomic_read(&kexec_ready_to_reboot))
+		cpu_relax();
+
+	asm volatile (
+	"	sync						\n"
+	"	synci	($0)					\n");
+
+	relocated_kexec_smp_wait(NULL);
+}
+#endif
+
+#define OCTEON_DDR0_BASE    (0x0ULL)
+#define OCTEON_DDR0_SIZE    (0x010000000ULL)
+#define OCTEON_DDR1_BASE    (0x410000000ULL)
+#define OCTEON_DDR1_SIZE    (0x010000000ULL)
+#define OCTEON_DDR2_BASE    (0x020000000ULL)
+#define OCTEON_DDR2_SIZE    (0x3e0000000ULL)
+#define OCTEON_MAX_PHY_MEM_SIZE (16*1024*1024*1024ULL)
+
+static struct kimage *kimage_ptr;
+
+static void kexec_bootmem_init(uint64_t mem_size, uint32_t low_reserved_bytes)
+{
+	int64_t addr;
+	struct cvmx_bootmem_desc *bootmem_desc;
+
+	bootmem_desc = cvmx_bootmem_get_desc();
+
+	if (mem_size > OCTEON_MAX_PHY_MEM_SIZE) {
+		mem_size = OCTEON_MAX_PHY_MEM_SIZE;
+		pr_err("Error: requested memory too large,"
+		       "truncating to maximum size\n");
+	}
+
+	bootmem_desc->major_version = CVMX_BOOTMEM_DESC_MAJ_VER;
+	bootmem_desc->minor_version = CVMX_BOOTMEM_DESC_MIN_VER;
+
+	addr = (OCTEON_DDR0_BASE + RESERVE_LOW_MEM + low_reserved_bytes);
+	bootmem_desc->head_addr = 0;
+
+	if (mem_size <= OCTEON_DDR0_SIZE) {
+		__cvmx_bootmem_phy_free(addr,
+				mem_size - RESERVE_LOW_MEM -
+				low_reserved_bytes, 0);
+		return;
+	}
+
+	__cvmx_bootmem_phy_free(addr,
+			OCTEON_DDR0_SIZE - RESERVE_LOW_MEM -
+			low_reserved_bytes, 0);
+
+	mem_size -= OCTEON_DDR0_SIZE;
+
+	if (mem_size > OCTEON_DDR1_SIZE) {
+		__cvmx_bootmem_phy_free(OCTEON_DDR1_BASE, OCTEON_DDR1_SIZE, 0);
+		__cvmx_bootmem_phy_free(OCTEON_DDR2_BASE,
+				mem_size - OCTEON_DDR1_SIZE, 0);
+	} else
+		__cvmx_bootmem_phy_free(OCTEON_DDR1_BASE, mem_size, 0);
+}
+
+static int octeon_kexec_prepare(struct kimage *image)
+{
+	int i;
+	char *bootloader = "kexec";
+
+	octeon_boot_desc_ptr->argc = 0;
+	for (i = 0; i < image->nr_segments; i++) {
+		if (!strncmp(bootloader, (char *)image->segment[i].buf,
+				strlen(bootloader))) {
+			/*
+			 * convert command line string to array
+			 * of parameters (as bootloader does).
+			 */
+			int argc = 0, offt;
+			char *str = (char *)image->segment[i].buf;
+			char *ptr = strchr(str, ' ');
+			while (ptr && (OCTEON_ARGV_MAX_ARGS > argc)) {
+				*ptr = '\0';
+				if (ptr[1] != ' ') {
+					offt = (int)(ptr - str + 1);
+					octeon_boot_desc_ptr->argv[argc] =
+						image->segment[i].mem + offt;
+					argc++;
+				}
+				ptr = strchr(ptr + 1, ' ');
+			}
+			octeon_boot_desc_ptr->argc = argc;
+			break;
+		}
+	}
+
+	/*
+	 * Information about segments will be needed during pre-boot memory
+	 * initialization.
+	 */
+	kimage_ptr = image;
+	return 0;
+}
+
+static void octeon_generic_shutdown(void)
+{
+	int cpu, i;
+	struct cvmx_bootmem_desc *bootmem_desc;
+	void *named_block_array_ptr;
+
+	bootmem_desc = cvmx_bootmem_get_desc();
+	named_block_array_ptr =
+		cvmx_phys_to_ptr(bootmem_desc->named_block_array_addr);
+
+#ifdef CONFIG_SMP
+	/* disable watchdogs */
+	for_each_online_cpu(cpu)
+		cvmx_write_csr(CVMX_CIU_WDOGX(cpu_logical_map(cpu)), 0);
+#else
+	cvmx_write_csr(CVMX_CIU_WDOGX(cvmx_get_core_num()), 0);
+#endif
+	if (kimage_ptr != kexec_crash_image) {
+		memset(named_block_array_ptr,
+			0x0,
+			CVMX_BOOTMEM_NUM_NAMED_BLOCKS *
+			sizeof(struct cvmx_bootmem_named_block_desc));
+		/*
+		 * Mark all memory (except low 0x100000 bytes) as free.
+		 * It is the same thing that bootloader does.
+		 */
+		kexec_bootmem_init(octeon_bootinfo->dram_size*1024ULL*1024ULL,
+				0x100000);
+		/*
+		 * Allocate all segments to avoid their corruption during boot.
+		 */
+		for (i = 0; i < kimage_ptr->nr_segments; i++)
+			cvmx_bootmem_alloc_address(
+				kimage_ptr->segment[i].memsz + 2*PAGE_SIZE,
+				kimage_ptr->segment[i].mem - PAGE_SIZE,
+				PAGE_SIZE);
+	} else {
+		/*
+		 * Do not mark all memory as free. Free only named sections
+		 * leaving the rest of memory unchanged.
+		 */
+		struct cvmx_bootmem_named_block_desc *ptr =
+			(struct cvmx_bootmem_named_block_desc *)
+			named_block_array_ptr;
+
+		for (i = 0; i < bootmem_desc->named_block_num_blocks; i++)
+			if (ptr[i].size)
+				cvmx_bootmem_free_named(ptr[i].name);
+	}
+	kexec_args[2] = 1UL; /* running on octeon_main_processor */
+	kexec_args[3] = (unsigned long)octeon_boot_desc_ptr;
+#ifdef CONFIG_SMP
+	secondary_kexec_args[2] = 0UL; /* running on secondary cpu */
+	secondary_kexec_args[3] = (unsigned long)octeon_boot_desc_ptr;
+#endif
+}
+
+static void octeon_shutdown(void)
+{
+	octeon_generic_shutdown();
+#ifdef CONFIG_SMP
+	smp_call_function(octeon_kexec_smp_down, NULL, 0);
+	smp_wmb();
+	while (num_online_cpus() > 1) {
+		cpu_relax();
+		mdelay(1);
+	}
+#endif
+}
+
+static void octeon_crash_shutdown(struct pt_regs *regs)
+{
+	octeon_generic_shutdown();
+	default_machine_crash_shutdown(regs);
+}
+
+#endif /* CONFIG_KEXEC */
+
 #ifdef CONFIG_CAVIUM_RESERVE32
 uint64_t octeon_reserve32_memory;
 EXPORT_SYMBOL(octeon_reserve32_memory);
 #endif
 
+#ifdef CONFIG_KEXEC
+/* crashkernel cmdline parameter is parsed _after_ memory setup
+ * we also parse it here (workaround for EHB5200) */
+static uint64_t crashk_size, crashk_base;
+#endif
+
 static int octeon_uart;
 
 extern asmlinkage void handle_int(void);
@@ -417,6 +615,8 @@ void octeon_user_io_init(void)
 void __init prom_init(void)
 {
 	struct cvmx_sysinfo *sysinfo;
+	const char *arg;
+	char *p;
 	int i;
 	int argc;
 #ifdef CONFIG_CAVIUM_RESERVE32
@@ -568,6 +768,15 @@ void __init prom_init(void)
 	if (octeon_is_simulation())
 		MAX_MEMORY = 64ull << 20;
 
+	arg = strstr(arcs_cmdline, "mem=");
+	if (arg) {
+		MAX_MEMORY = memparse(arg + 4, &p);
+		if (MAX_MEMORY == 0)
+			MAX_MEMORY = 32ull << 30;
+		if (*p == '@')
+			RESERVE_LOW_MEM = memparse(p + 1, &p);
+	}
+
 	arcs_cmdline[0] = 0;
 	argc = octeon_boot_desc_ptr->argc;
 	for (i = 0; i < argc; i++) {
@@ -575,16 +784,30 @@ void __init prom_init(void)
 			cvmx_phys_to_ptr(octeon_boot_desc_ptr->argv[i]);
 		if ((strncmp(arg, "MEM=", 4) == 0) ||
 		    (strncmp(arg, "mem=", 4) == 0)) {
-			sscanf(arg + 4, "%llu", &MAX_MEMORY);
-			MAX_MEMORY <<= 20;
+			MAX_MEMORY = memparse(arg + 4, &p);
 			if (MAX_MEMORY == 0)
 				MAX_MEMORY = 32ull << 30;
+			if (*p == '@')
+				RESERVE_LOW_MEM = memparse(p + 1, &p);
 		} else if (strcmp(arg, "ecc_verbose") == 0) {
 #ifdef CONFIG_CAVIUM_REPORT_SINGLE_BIT_ECC
 			__cvmx_interrupt_ecc_report_single_bit_errors = 1;
 			pr_notice("Reporting of single bit ECC errors is "
 				  "turned on\n");
 #endif
+#ifdef CONFIG_KEXEC
+		} else if (strncmp(arg, "crashkernel=", 12) == 0) {
+			crashk_size = memparse(arg+12, &p);
+			if (*p == '@')
+				crashk_base = memparse(p+1, &p);
+			strcat(arcs_cmdline, " ");
+			strcat(arcs_cmdline, arg);
+			/*
+			 * To do: switch parsing to new style, something like:
+			 * parse_crashkernel(arg, sysinfo->system_dram_size,
+			 * 		  &crashk_size, &crashk_base);
+			 */
+#endif
 		} else if (strlen(arcs_cmdline) + strlen(arg) + 1 <
 			   sizeof(arcs_cmdline) - 1) {
 			strcat(arcs_cmdline, " ");
@@ -619,11 +842,18 @@ void __init prom_init(void)
 	_machine_restart = octeon_restart;
 	_machine_halt = octeon_halt;
 
+#ifdef CONFIG_KEXEC
+	_machine_kexec_shutdown = octeon_shutdown;
+	_machine_crash_shutdown = octeon_crash_shutdown;
+	_machine_kexec_prepare = octeon_kexec_prepare;
+#endif
+
 	octeon_user_io_init();
 	register_smp_ops(&octeon_smp_ops);
 }
 
 /* Exclude a single page from the regions obtained in plat_mem_setup. */
+#ifndef CONFIG_CRASH_DUMP
 static __init void memory_exclude_page(u64 addr, u64 *mem, u64 *size)
 {
 	if (addr > *mem && addr < *mem + *size) {
@@ -638,14 +868,21 @@ static __init void memory_exclude_page(u64 addr, u64 *mem, u64 *size)
 		*size -= PAGE_SIZE;
 	}
 }
+#endif /* CONFIG_CRASH_DUMP */
 
 void __init plat_mem_setup(void)
 {
 	uint64_t mem_alloc_size;
 	uint64_t total;
+	uint64_t crashk_end;
+#ifndef CONFIG_CRASH_DUMP
 	int64_t memory;
+	uint64_t kernel_start;
+	uint64_t kernel_size;
+#endif
 
 	total = 0;
+	crashk_end = 0;
 
 	/*
 	 * The Mips memory init uses the first memory location for
@@ -658,6 +895,17 @@ void __init plat_mem_setup(void)
 	if (mem_alloc_size > MAX_MEMORY)
 		mem_alloc_size = MAX_MEMORY;
 
+/* Crashkernel ignores bootmem list. It relies on mem=X@Y option */
+#ifdef CONFIG_CRASH_DUMP
+	add_memory_region(RESERVE_LOW_MEM, MAX_MEMORY, BOOT_MEM_RAM);
+	total += MAX_MEMORY;
+#else
+#ifdef CONFIG_KEXEC
+	if (crashk_size > 0) {
+		add_memory_region(crashk_base, crashk_size, BOOT_MEM_RAM);
+		crashk_end = crashk_base + crashk_size;
+	}
+#endif
 	/*
 	 * When allocating memory, we want incrementing addresses from
 	 * bootmem_alloc so the code in add_memory_region can merge
@@ -672,6 +920,9 @@ void __init plat_mem_setup(void)
 						CVMX_BOOTMEM_FLAG_NO_LOCKING);
 		if (memory >= 0) {
 			u64 size = mem_alloc_size;
+#ifdef CONFIG_KEXEC
+			uint64_t end;
+#endif
 
 			/*
 			 * exclude a page at the beginning and end of
@@ -684,20 +935,67 @@ void __init plat_mem_setup(void)
 			memory_exclude_page(CVMX_PCIE_BAR1_PHYS_BASE +
 					    CVMX_PCIE_BAR1_PHYS_SIZE,
 					    &memory, &size);
+#ifdef CONFIG_KEXEC
+			end = memory + mem_alloc_size;
 
 			/*
-			 * This function automatically merges address
-			 * regions next to each other if they are
-			 * received in incrementing order.
+			 * This function automatically merges address regions
+			 * next to each other if they are received in
+			 * incrementing order
 			 */
-			if (size)
-				add_memory_region(memory, size, BOOT_MEM_RAM);
+			if (memory < crashk_base && end >  crashk_end) {
+				/* region is fully in */
+				add_memory_region(memory,
+						  crashk_base - memory,
+						  BOOT_MEM_RAM);
+				total += crashk_base - memory;
+				add_memory_region(crashk_end,
+						  end - crashk_end,
+						  BOOT_MEM_RAM);
+				total += end - crashk_end;
+				continue;
+			}
+
+			if (memory >= crashk_base && end <= crashk_end)
+				/*
+				 * Entire memory region is within the new
+				 *  kernel's memory, ignore it.
+				 */
+				continue;
+
+			if (memory > crashk_base && memory < crashk_end &&
+			    end > crashk_end) {
+				/*
+				 * Overlap with the beginning of the region,
+				 * reserve the beginning.
+				  */
+				mem_alloc_size -= crashk_end - memory;
+				memory = crashk_end;
+			} else if (memory < crashk_base && end > crashk_base &&
+				   end < crashk_end)
+				/*
+				 * Overlap with the beginning of the region,
+				 * chop of end.
+				 */
+				mem_alloc_size -= end - crashk_base;
+#endif
+			add_memory_region(memory, mem_alloc_size, BOOT_MEM_RAM);
 			total += mem_alloc_size;
+			/* Recovering mem_alloc_size */
+			mem_alloc_size = 4 << 20;
 		} else {
 			break;
 		}
 	}
 	cvmx_bootmem_unlock();
+	/* Add the memory region for the kernel. */
+	kernel_start = (unsigned long) _text;
+	kernel_size = ALIGN(_end - _text, 0x100000);
+
+	/* Adjust for physical offset. */
+	kernel_start &= ~0xffffffff80000000ULL;
+	add_memory_region(kernel_start, kernel_size, BOOT_MEM_RAM);
+#endif /* CONFIG_CRASH_DUMP */
 
 #ifdef CONFIG_CAVIUM_RESERVE32
 	/*
diff --git a/arch/mips/include/asm/octeon/cvmx-bootmem.h b/arch/mips/include/asm/octeon/cvmx-bootmem.h
index 877845b84b1..42db2be663f 100644
--- a/arch/mips/include/asm/octeon/cvmx-bootmem.h
+++ b/arch/mips/include/asm/octeon/cvmx-bootmem.h
@@ -370,4 +370,6 @@ void cvmx_bootmem_lock(void);
  */
 void cvmx_bootmem_unlock(void);
 
+extern struct cvmx_bootmem_desc *cvmx_bootmem_get_desc(void);
+
 #endif /*   __CVMX_BOOTMEM_H__ */
diff --git a/arch/mips/kernel/crash_dump.c b/arch/mips/kernel/crash_dump.c
index d9ec3898f9f..35bed0d2342 100644
--- a/arch/mips/kernel/crash_dump.c
+++ b/arch/mips/kernel/crash_dump.c
@@ -3,8 +3,6 @@
 #include <linux/crash_dump.h>
 #include <asm/uaccess.h>
 
-unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
-
 static int __init parse_savemaxmem(char *p)
 {
 	if (p)
diff --git a/arch/mips/kernel/relocate_kernel.S b/arch/mips/kernel/relocate_kernel.S
index 0b108587f06..e4142c5f7c2 100644
--- a/arch/mips/kernel/relocate_kernel.S
+++ b/arch/mips/kernel/relocate_kernel.S
@@ -78,7 +78,19 @@ done:
 	LONG_S		zero,(t0)
 #endif
 
+#ifdef CONFIG_CPU_CAVIUM_OCTEON
+	/* We need to flush I-cache before jumping to new kernel.
+	 * Unfortunatelly, this code is cpu-specific.
+	 */
+	.set push
+	.set noreorder
+	syncw
+	syncw
+	synci		0($0)
+	.set pop
+#else
 	sync
+#endif
 	/* jump to kexec_start_address */
 	j		s1
 	END(relocate_new_kernel)
@@ -110,7 +122,14 @@ LEAF(kexec_smp_wait)
 1:	LONG_L		s0, (t0)
 	bne		s0, zero,1b
 
+#ifdef CONFIG_CPU_CAVIUM_OCTEON
+	.set push
+	.set noreorder
+	synci		0($0)
+	.set pop
+#else
 	sync
+#endif
 	j		s1
 	END(kexec_smp_wait)
 #endif