Mike Rapoport (Microsoft)mm, arch: consolidate hugetlb CMA reservation9fac14510 months agoHistory
1016 lines · 27.5 KB · C
RawDownload
// SPDX-License-Identifier: GPL-2.0/* *  S390 version *    Copyright IBM Corp. 1999, 2012 *    Author(s): Hartmut Penner (hp@de.ibm.com), *               Martin Schwidefsky (schwidefsky@de.ibm.com) * *  Derived from "arch/i386/kernel/setup.c" *    Copyright (C) 1995, Linus Torvalds */ /* * This file handles the architecture-dependent parts of initialization */ #define pr_fmt(fmt) "setup: " fmt #include <linux/errno.h>#include <linux/export.h>#include <linux/sched.h>#include <linux/sched/task.h>#include <linux/cpu.h>#include <linux/kernel.h>#include <linux/memblock.h>#include <linux/mm.h>#include <linux/stddef.h>#include <linux/unistd.h>#include <linux/ptrace.h>#include <linux/random.h>#include <linux/user.h>#include <linux/tty.h>#include <linux/ioport.h>#include <linux/delay.h>#include <linux/init.h>#include <linux/initrd.h>#include <linux/root_dev.h>#include <linux/console.h>#include <linux/kernel_stat.h>#include <linux/dma-map-ops.h>#include <linux/device.h>#include <linux/notifier.h>#include <linux/pfn.h>#include <linux/ctype.h>#include <linux/reboot.h>#include <linux/topology.h>#include <linux/kexec.h>#include <linux/crash_dump.h>#include <linux/memory.h>#include <linux/start_kernel.h>#include <linux/hugetlb.h>#include <linux/kmemleak.h> #include <asm/archrandom.h>#include <asm/boot_data.h>#include <asm/machine.h>#include <asm/ipl.h>#include <asm/facility.h>#include <asm/smp.h>#include <asm/mmu_context.h>#include <asm/cpcmd.h>#include <asm/abs_lowcore.h>#include <asm/nmi.h>#include <asm/irq.h>#include <asm/page.h>#include <asm/ptrace.h>#include <asm/sections.h>#include <asm/ebcdic.h>#include <asm/diag.h>#include <asm/os_info.h>#include <asm/sclp.h>#include <asm/stacktrace.h>#include <asm/sysinfo.h>#include <asm/numa.h>#include <asm/alternative.h>#include <asm/nospec-branch.h>#include <asm/physmem_info.h>#include <asm/maccess.h>#include <asm/uv.h>#include <asm/asm-offsets.h>#include "entry.h" /* * Machine setup.. */unsigned int console_mode = 0;EXPORT_SYMBOL(console_mode); unsigned int console_devno = -1;EXPORT_SYMBOL(console_devno); unsigned int console_irq = -1;EXPORT_SYMBOL(console_irq); /* * Some code and data needs to stay below 2 GB, even when the kernel would be * relocated above 2 GB, because it has to use 31 bit addresses. * Such code and data is part of the .amode31 section. */char __amode31_ref *__samode31 = _samode31;char __amode31_ref *__eamode31 = _eamode31;char __amode31_ref *__stext_amode31 = _stext_amode31;char __amode31_ref *__etext_amode31 = _etext_amode31;struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table;struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table; /* * Control registers CR2, CR5 and CR15 are initialized with addresses * of tables that must be placed below 2G which is handled by the AMODE31 * sections. * Because the AMODE31 sections are relocated below 2G at startup, * the content of control registers CR2, CR5 and CR15 must be updated * with new addresses after the relocation. The initial initialization of * control registers occurs in head.S and then gets updated again after AMODE31 * relocation. We must access the relevant AMODE31 tables indirectly via * pointers placed in the .amode31.refs linker section. Those pointers get * updated automatically during AMODE31 relocation and always contain a valid * address within AMODE31 sections. */ static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64); static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = {	[1] = 0xffffffffffffffff}; static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = {	0x80000000, 0, 0, 0,	0x80000000, 0, 0, 0,	0x80000000, 0, 0, 0,	0x80000000, 0, 0, 0,	0x80000000, 0, 0, 0,	0x80000000, 0, 0, 0,	0x80000000, 0, 0, 0,	0x80000000, 0, 0, 0}; static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = {	0, 0, 0x89000000, 0,	0, 0, 0x8a000000, 0}; static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31;static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31;static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31;static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31; unsigned long __bootdata_preserved(max_mappable);struct physmem_info __bootdata(physmem_info); struct vm_layout __bootdata_preserved(vm_layout);EXPORT_SYMBOL(vm_layout);int __bootdata_preserved(__kaslr_enabled);unsigned int __bootdata_preserved(zlib_dfltcc_support);EXPORT_SYMBOL(zlib_dfltcc_support);u64 __bootdata_preserved(stfle_fac_list[16]);EXPORT_SYMBOL(stfle_fac_list);struct oldmem_data __bootdata_preserved(oldmem_data); char __bootdata(boot_rb)[PAGE_SIZE * 2];bool __bootdata(boot_earlyprintk);size_t __bootdata(boot_rb_off);char __bootdata(bootdebug_filter)[128];bool __bootdata(bootdebug); unsigned long __bootdata_preserved(VMALLOC_START);EXPORT_SYMBOL(VMALLOC_START); unsigned long __bootdata_preserved(VMALLOC_END);EXPORT_SYMBOL(VMALLOC_END); struct page *__bootdata_preserved(vmemmap);EXPORT_SYMBOL(vmemmap);unsigned long __bootdata_preserved(vmemmap_size); unsigned long __bootdata_preserved(MODULES_VADDR);unsigned long __bootdata_preserved(MODULES_END); /* An array with a pointer to the lowcore of every CPU. */struct lowcore *lowcore_ptr[NR_CPUS];EXPORT_SYMBOL(lowcore_ptr); /* * The Write Back bit position in the physaddr is given by the SLPC PCI. * Leaving the mask zero always uses write through which is safe */unsigned long mio_wb_bit_mask __ro_after_init; /* * This is set up by the setup-routine at boot-time * for S390 need to find out, what we have to setup * using address 0x10400 ... */ #include <asm/setup.h> /* * condev= and conmode= setup parameter. */ static int __init condev_setup(char *str){	int vdev; 	vdev = simple_strtoul(str, &str, 0);	if (vdev >= 0 && vdev < 65536) {		console_devno = vdev;		console_irq = -1;	}	return 1;} __setup("condev=", condev_setup); static void __init set_preferred_console(void){	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)		add_preferred_console("ttyS", 0, NULL);	else if (CONSOLE_IS_3270)		add_preferred_console("tty3270", 0, NULL);	else if (CONSOLE_IS_VT220)		add_preferred_console("ttysclp", 0, NULL);	else if (CONSOLE_IS_HVC)		add_preferred_console("hvc", 0, NULL);} static int __init conmode_setup(char *str){#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)	if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))                SET_CONSOLE_SCLP;#endif#if defined(CONFIG_TN3215_CONSOLE)	if (!strcmp(str, "3215"))		SET_CONSOLE_3215;#endif#if defined(CONFIG_TN3270_CONSOLE)	if (!strcmp(str, "3270"))		SET_CONSOLE_3270;#endif	set_preferred_console();        return 1;} __setup("conmode=", conmode_setup); static void __init conmode_default(void){	char query_buffer[1024];	char *ptr; 	if (machine_is_vm()) {		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);		ptr = strstr(query_buffer, "SUBCHANNEL =");		console_irq = simple_strtoul(ptr + 13, NULL, 16);		cpcmd("QUERY TERM", query_buffer, 1024, NULL);		ptr = strstr(query_buffer, "CONMODE");		/*		 * Set the conmode to 3215 so that the device recognition 		 * will set the cu_type of the console to 3215. If the		 * conmode is 3270 and we don't set it back then both		 * 3215 and the 3270 driver will try to access the console		 * device (3215 as console and 3270 as normal tty).		 */		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);		if (ptr == NULL) {#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)			SET_CONSOLE_SCLP;#endif			return;		}		if (str_has_prefix(ptr + 8, "3270")) {#if defined(CONFIG_TN3270_CONSOLE)			SET_CONSOLE_3270;#elif defined(CONFIG_TN3215_CONSOLE)			SET_CONSOLE_3215;#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)			SET_CONSOLE_SCLP;#endif		} else if (str_has_prefix(ptr + 8, "3215")) {#if defined(CONFIG_TN3215_CONSOLE)			SET_CONSOLE_3215;#elif defined(CONFIG_TN3270_CONSOLE)			SET_CONSOLE_3270;#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)			SET_CONSOLE_SCLP;#endif		}	} else if (machine_is_kvm()) {		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))			SET_CONSOLE_VT220;		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))			SET_CONSOLE_SCLP;		else			SET_CONSOLE_HVC;	} else {#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)		SET_CONSOLE_SCLP;#endif	}} #ifdef CONFIG_CRASH_DUMPstatic void __init setup_zfcpdump(void){	if (!is_ipl_type_dump())		return;	if (oldmem_data.start)		return;	strlcat(boot_command_line, " cio_ignore=all,!ipldev,!condev", COMMAND_LINE_SIZE);	console_loglevel = 2;}#elsestatic inline void setup_zfcpdump(void) {}#endif /* CONFIG_CRASH_DUMP */  /* * Reboot, halt and power_off stubs. They just call _machine_restart, * _machine_halt or _machine_power_off.  */ void machine_restart(char *command){	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)		/*		 * Only unblank the console if we are called in enabled		 * context or a bust_spinlocks cleared the way for us.		 */		console_unblank();	_machine_restart(command);} void machine_halt(void){	if (!in_interrupt() || oops_in_progress)		/*		 * Only unblank the console if we are called in enabled		 * context or a bust_spinlocks cleared the way for us.		 */		console_unblank();	_machine_halt();} void machine_power_off(void){	if (!in_interrupt() || oops_in_progress)		/*		 * Only unblank the console if we are called in enabled		 * context or a bust_spinlocks cleared the way for us.		 */		console_unblank();	_machine_power_off();} /* * Dummy power off function. */void (*pm_power_off)(void) = machine_power_off;EXPORT_SYMBOL_GPL(pm_power_off); void *restart_stack; unsigned long stack_alloc(void){	void *stack; 	stack = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,			       NUMA_NO_NODE, __builtin_return_address(0));	kmemleak_not_leak(stack);	return (unsigned long)stack;} void stack_free(unsigned long stack){	vfree((void *)stack);} static unsigned long __init stack_alloc_early(void){	unsigned long stack; 	stack = (unsigned long)memblock_alloc_or_panic(THREAD_SIZE, THREAD_SIZE);	return stack;} static void __init setup_lowcore(void){	struct lowcore *lc, *abs_lc; 	/*	 * Setup lowcore for boot cpu	 */	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));	if (!lc)		panic("%s: Failed to allocate %zu bytes align=%zx\n",		      __func__, sizeof(*lc), sizeof(*lc)); 	lc->pcpu = (unsigned long)per_cpu_ptr(&pcpu_devices, 0);	lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT;	lc->restart_psw.addr = __pa(restart_int_handler);	lc->external_new_psw.mask = PSW_KERNEL_BITS;	lc->external_new_psw.addr = (unsigned long) ext_int_handler;	lc->svc_new_psw.mask = PSW_KERNEL_BITS;	lc->svc_new_psw.addr = (unsigned long) system_call;	lc->program_new_psw.mask = PSW_KERNEL_BITS;	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;	lc->io_new_psw.mask = PSW_KERNEL_BITS;	lc->io_new_psw.addr = (unsigned long) io_int_handler;	lc->clock_comparator = clock_comparator_max;	lc->current_task = (unsigned long)&init_task;	lc->lpp = LPP_MAGIC;	lc->preempt_count = get_lowcore()->preempt_count;	nmi_alloc_mcesa_early(&lc->mcesad);	lc->sys_enter_timer = get_lowcore()->sys_enter_timer;	lc->exit_timer = get_lowcore()->exit_timer;	lc->user_timer = get_lowcore()->user_timer;	lc->system_timer = get_lowcore()->system_timer;	lc->steal_timer = get_lowcore()->steal_timer;	lc->last_update_timer = get_lowcore()->last_update_timer;	lc->last_update_clock = get_lowcore()->last_update_clock;	/*	 * Allocate the global restart stack which is the same for	 * all CPUs in case *one* of them does a PSW restart.	 */	restart_stack = (void *)(stack_alloc_early() + STACK_INIT_OFFSET);	lc->mcck_stack = stack_alloc_early() + STACK_INIT_OFFSET;	lc->async_stack = stack_alloc_early() + STACK_INIT_OFFSET;	lc->nodat_stack = stack_alloc_early() + STACK_INIT_OFFSET;	lc->kernel_stack = get_lowcore()->kernel_stack;	/*	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant	 * restart data to the absolute zero lowcore. This is necessary if	 * PSW restart is done on an offline CPU that has lowcore zero.	 */	lc->restart_stack = (unsigned long) restart_stack;	lc->restart_fn = (unsigned long) do_restart;	lc->restart_data = 0;	lc->restart_source = -1U;	lc->spinlock_lockval = arch_spin_lockval(0);	lc->spinlock_index = 0;	arch_spin_lock_setup(0);	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);	lc->preempt_count = PREEMPT_DISABLED;	lc->kernel_asce = get_lowcore()->kernel_asce;	lc->user_asce = get_lowcore()->user_asce; 	system_ctlreg_init_save_area(lc);	abs_lc = get_abs_lowcore();	abs_lc->restart_stack = lc->restart_stack;	abs_lc->restart_fn = lc->restart_fn;	abs_lc->restart_data = lc->restart_data;	abs_lc->restart_source = lc->restart_source;	abs_lc->restart_psw = lc->restart_psw;	abs_lc->restart_flags = RESTART_FLAG_CTLREGS;	abs_lc->program_new_psw = lc->program_new_psw;	abs_lc->mcesad = lc->mcesad;	put_abs_lowcore(abs_lc); 	set_prefix(__pa(lc));	lowcore_ptr[0] = lc;	if (abs_lowcore_map(0, lowcore_ptr[0], false))		panic("Couldn't setup absolute lowcore");} static struct resource code_resource = {	.name  = "Kernel code",	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,}; static struct resource data_resource = {	.name = "Kernel data",	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,}; static struct resource bss_resource = {	.name = "Kernel bss",	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,}; static struct resource __initdata *standard_resources[] = {	&code_resource,	&data_resource,	&bss_resource,}; static void __init setup_resources(void){	struct resource *res, *std_res, *sub_res;	phys_addr_t start, end;	int j;	u64 i; 	code_resource.start = __pa_symbol(_text);	code_resource.end = __pa_symbol(_etext) - 1;	data_resource.start = __pa_symbol(_etext);	data_resource.end = __pa_symbol(_edata) - 1;

Lines 1–500 of 1016.