diff options
Diffstat (limited to 'arch/s390/kernel/nmi.c')
| -rw-r--r-- | arch/s390/kernel/nmi.c | 594 |
1 files changed, 364 insertions, 230 deletions
diff --git a/arch/s390/kernel/nmi.c b/arch/s390/kernel/nmi.c index 504175ebf8b0..a55abbf65333 100644 --- a/arch/s390/kernel/nmi.c +++ b/arch/s390/kernel/nmi.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: GPL-2.0 /* * Machine check handler * @@ -5,59 +6,235 @@ * Author(s): Ingo Adlung <adlung@de.ibm.com>, * Martin Schwidefsky <schwidefsky@de.ibm.com>, * Cornelia Huck <cornelia.huck@de.ibm.com>, - * Heiko Carstens <heiko.carstens@de.ibm.com>, */ #include <linux/kernel_stat.h> +#include <linux/utsname.h> +#include <linux/cpufeature.h> #include <linux/init.h> #include <linux/errno.h> +#include <linux/entry-common.h> #include <linux/hardirq.h> +#include <linux/log2.h> +#include <linux/kprobes.h> +#include <linux/kmemleak.h> #include <linux/time.h> #include <linux/module.h> +#include <linux/sched/signal.h> +#include <linux/kvm_host.h> #include <asm/lowcore.h> +#include <asm/ctlreg.h> +#include <asm/fpu.h> #include <asm/smp.h> -#include <asm/etr.h> +#include <asm/stp.h> #include <asm/cputime.h> #include <asm/nmi.h> #include <asm/crw.h> +#include <asm/asm-offsets.h> +#include <asm/pai.h> +#include <asm/vtime.h> struct mcck_struct { - int kill_task; - int channel_report; - int warning; - unsigned long long mcck_code; + unsigned int kill_task : 1; + unsigned int channel_report : 1; + unsigned int warning : 1; + unsigned int stp_queue : 1; + unsigned long mcck_code; }; static DEFINE_PER_CPU(struct mcck_struct, cpu_mcck); -static void s390_handle_damage(char *msg) +static inline int nmi_needs_mcesa(void) { - smp_send_stop(); - disabled_wait((unsigned long) __builtin_return_address(0)); - while (1); + return cpu_has_vx() || cpu_has_gs(); } /* - * Main machine check handler function. Will be called with interrupts enabled - * or disabled and machine checks enabled or disabled. + * The initial machine check extended save area for the boot CPU. + * It will be replaced on the boot CPU reinit with an allocated + * structure. The structure is required for machine check happening + * early in the boot process. + */ +static struct mcesa boot_mcesa __aligned(MCESA_MAX_SIZE); + +void __init nmi_alloc_mcesa_early(u64 *mcesad) +{ + if (!nmi_needs_mcesa()) + return; + *mcesad = __pa(&boot_mcesa); + if (cpu_has_gs()) + *mcesad |= ilog2(MCESA_MAX_SIZE); +} + +int nmi_alloc_mcesa(u64 *mcesad) +{ + unsigned long size; + void *origin; + + *mcesad = 0; + if (!nmi_needs_mcesa()) + return 0; + size = cpu_has_gs() ? MCESA_MAX_SIZE : MCESA_MIN_SIZE; + origin = kmalloc(size, GFP_KERNEL); + if (!origin) + return -ENOMEM; + /* The pointer is stored with mcesa_bits ORed in */ + kmemleak_not_leak(origin); + *mcesad = __pa(origin); + if (cpu_has_gs()) + *mcesad |= ilog2(MCESA_MAX_SIZE); + return 0; +} + +void nmi_free_mcesa(u64 *mcesad) +{ + if (!nmi_needs_mcesa()) + return; + kfree(__va(*mcesad & MCESA_ORIGIN_MASK)); +} + +static __always_inline char *nmi_puts(char *dest, const char *src) +{ + while (*src) + *dest++ = *src++; + *dest = 0; + return dest; +} + +static __always_inline char *u64_to_hex(char *dest, u64 val) +{ + int i, num; + + for (i = 1; i <= 16; i++) { + num = (val >> (64 - 4 * i)) & 0xf; + if (num >= 10) + *dest++ = 'A' + num - 10; + else + *dest++ = '0' + num; + } + *dest = 0; + return dest; +} + +static notrace void nmi_print_info(void) +{ + struct lowcore *lc = get_lowcore(); + char message[100]; + char *ptr; + int i; + + ptr = nmi_puts(message, "Unrecoverable machine check, code: "); + ptr = u64_to_hex(ptr, lc->mcck_interruption_code); + ptr = nmi_puts(ptr, "\n"); + sclp_emergency_printk(message); + + ptr = nmi_puts(message, init_utsname()->release); + ptr = nmi_puts(ptr, "\n"); + sclp_emergency_printk(message); + + ptr = nmi_puts(message, arch_hw_string); + ptr = nmi_puts(ptr, "\n"); + sclp_emergency_printk(message); + + ptr = nmi_puts(message, "PSW: "); + ptr = u64_to_hex(ptr, lc->mcck_old_psw.mask); + ptr = nmi_puts(ptr, " "); + ptr = u64_to_hex(ptr, lc->mcck_old_psw.addr); + ptr = nmi_puts(ptr, " PFX: "); + ptr = u64_to_hex(ptr, (u64)get_lowcore()); + ptr = nmi_puts(ptr, "\n"); + sclp_emergency_printk(message); + + ptr = nmi_puts(message, "LBA: "); + ptr = u64_to_hex(ptr, lc->last_break_save_area); + ptr = nmi_puts(ptr, " EDC: "); + ptr = u64_to_hex(ptr, lc->external_damage_code); + ptr = nmi_puts(ptr, " FSA: "); + ptr = u64_to_hex(ptr, lc->failing_storage_address); + ptr = nmi_puts(ptr, "\n"); + sclp_emergency_printk(message); + + ptr = nmi_puts(message, "CRS:\n"); + sclp_emergency_printk(message); + ptr = message; + for (i = 0; i < 16; i++) { + ptr = u64_to_hex(ptr, lc->cregs_save_area[i].val); + ptr = nmi_puts(ptr, " "); + if ((i + 1) % 4 == 0) { + ptr = nmi_puts(ptr, "\n"); + sclp_emergency_printk(message); + ptr = message; + } + } + + ptr = nmi_puts(message, "GPRS:\n"); + sclp_emergency_printk(message); + ptr = message; + for (i = 0; i < 16; i++) { + ptr = u64_to_hex(ptr, lc->gpregs_save_area[i]); + ptr = nmi_puts(ptr, " "); + if ((i + 1) % 4 == 0) { + ptr = nmi_puts(ptr, "\n"); + sclp_emergency_printk(message); + ptr = message; + } + } + + ptr = nmi_puts(message, "System stopped\n"); + sclp_emergency_printk(message); +} + +static notrace void __noreturn s390_handle_damage(void) +{ + struct lowcore *lc = get_lowcore(); + union ctlreg0 cr0, cr0_new; + psw_t psw_save; + + smp_emergency_stop(); + diag_amode31_ops.diag308_reset(); + + /* + * Disable low address protection and make machine check new PSW a + * disabled wait PSW. Any additional machine check cannot be handled. + */ + local_ctl_store(0, &cr0.reg); + cr0_new = cr0; + cr0_new.lap = 0; + local_ctl_load(0, &cr0_new.reg); + psw_save = lc->mcck_new_psw; + psw_bits(lc->mcck_new_psw).io = 0; + psw_bits(lc->mcck_new_psw).ext = 0; + psw_bits(lc->mcck_new_psw).wait = 1; + nmi_print_info(); + + /* + * Restore machine check new PSW and control register 0 to original + * values. This makes possible system dump analysis easier. + */ + lc->mcck_new_psw = psw_save; + local_ctl_load(0, &cr0.reg); + disabled_wait(); +} +NOKPROBE_SYMBOL(s390_handle_damage); + +/* + * Main machine check handler function. Will be called with interrupts disabled + * and machine checks enabled. */ void s390_handle_mcck(void) { - unsigned long flags; struct mcck_struct mcck; + unsigned long mflags; /* * Disable machine checks and get the current state of accumulated * machine checks. Afterwards delete the old state and enable machine * checks again. */ - local_irq_save(flags); - local_mcck_disable(); - mcck = __get_cpu_var(cpu_mcck); - memset(&__get_cpu_var(cpu_mcck), 0, sizeof(struct mcck_struct)); - clear_thread_flag(TIF_MCCK_PENDING); - local_mcck_enable(); - local_irq_restore(flags); + local_mcck_save(mflags); + mcck = *this_cpu_ptr(&cpu_mcck); + memset(this_cpu_ptr(&cpu_mcck), 0, sizeof(mcck)); + local_mcck_restore(mflags); if (mcck.channel_report) crw_handle_channel_report(); @@ -74,171 +251,112 @@ void s390_handle_mcck(void) static int mchchk_wng_posted = 0; /* Use single cpu clear, as we cannot handle smp here. */ - __ctl_clear_bit(14, 24); /* Disable WARNING MCH */ + local_ctl_clear_bit(14, CR14_WARNING_SUBMASK_BIT); if (xchg(&mchchk_wng_posted, 1) == 0) kill_cad_pid(SIGPWR, 1); } + if (mcck.stp_queue) + stp_queue_work(); if (mcck.kill_task) { - local_irq_enable(); printk(KERN_EMERG "mcck: Terminating task because of machine " - "malfunction (code 0x%016llx).\n", mcck.mcck_code); + "malfunction (code 0x%016lx).\n", mcck.mcck_code); printk(KERN_EMERG "mcck: task: %s, pid: %d.\n", current->comm, current->pid); - do_exit(SIGSEGV); + if (is_global_init(current)) + panic("mcck: Attempting to kill init!\n"); + do_send_sig_info(SIGKILL, SEND_SIG_PRIV, current, PIDTYPE_PID); } } -EXPORT_SYMBOL_GPL(s390_handle_mcck); -/* - * returns 0 if all registers could be validated - * returns 1 otherwise +/** + * nmi_registers_valid - verify if registers are valid + * @mci: machine check interruption code + * + * Inspect a machine check interruption code and verify if all required + * registers are valid. For some registers the corresponding validity bit is + * ignored and the registers are set to the expected value. + * Returns true if all registers are valid, otherwise false. */ -static int notrace s390_revalidate_registers(struct mci *mci) +static bool notrace nmi_registers_valid(union mci mci) { - int kill_task; - u64 zero; - void *fpt_save_area, *fpt_creg_save_area; - - kill_task = 0; - zero = 0; + union ctlreg2 cr2; - if (!mci->gr) { - /* - * General purpose registers couldn't be restored and have - * unknown contents. Process needs to be terminated. - */ - kill_task = 1; - } - if (!mci->fp) { - /* - * Floating point registers can't be restored and - * therefore the process needs to be terminated. - */ - kill_task = 1; - } -#ifndef CONFIG_64BIT - asm volatile( - " ld 0,0(%0)\n" - " ld 2,8(%0)\n" - " ld 4,16(%0)\n" - " ld 6,24(%0)" - : : "a" (&S390_lowcore.floating_pt_save_area)); -#endif - - if (MACHINE_HAS_IEEE) { -#ifdef CONFIG_64BIT - fpt_save_area = &S390_lowcore.floating_pt_save_area; - fpt_creg_save_area = &S390_lowcore.fpt_creg_save_area; -#else - fpt_save_area = (void *) S390_lowcore.extended_save_area_addr; - fpt_creg_save_area = fpt_save_area + 128; -#endif - if (!mci->fc) { - /* - * Floating point control register can't be restored. - * Task will be terminated. - */ - asm volatile("lfpc 0(%0)" : : "a" (&zero), "m" (zero)); - kill_task = 1; - - } else - asm volatile("lfpc 0(%0)" : : "a" (fpt_creg_save_area)); - - asm volatile( - " ld 0,0(%0)\n" - " ld 1,8(%0)\n" - " ld 2,16(%0)\n" - " ld 3,24(%0)\n" - " ld 4,32(%0)\n" - " ld 5,40(%0)\n" - " ld 6,48(%0)\n" - " ld 7,56(%0)\n" - " ld 8,64(%0)\n" - " ld 9,72(%0)\n" - " ld 10,80(%0)\n" - " ld 11,88(%0)\n" - " ld 12,96(%0)\n" - " ld 13,104(%0)\n" - " ld 14,112(%0)\n" - " ld 15,120(%0)\n" - : : "a" (fpt_save_area)); - } - /* Revalidate access registers */ - asm volatile( - " lam 0,15,0(%0)" - : : "a" (&S390_lowcore.access_regs_save_area)); - if (!mci->ar) { - /* - * Access registers have unknown contents. - * Terminating task. - */ - kill_task = 1; - } - /* Revalidate control registers */ - if (!mci->cr) { - /* - * Control registers have unknown contents. - * Can't recover and therefore stopping machine. - */ - s390_handle_damage("invalid control registers."); - } else { -#ifdef CONFIG_64BIT - asm volatile( - " lctlg 0,15,0(%0)" - : : "a" (&S390_lowcore.cregs_save_area)); -#else - asm volatile( - " lctl 0,15,0(%0)" - : : "a" (&S390_lowcore.cregs_save_area)); -#endif - } /* - * We don't even try to revalidate the TOD register, since we simply - * can't write something sensible into that register. + * The getcpu vdso syscall reads the CPU number from the programmable + * field of the TOD clock. Disregard the TOD programmable register + * validity bit and load the CPU number into the TOD programmable field + * unconditionally. */ -#ifdef CONFIG_64BIT + set_tod_programmable_field(raw_smp_processor_id()); /* - * See if we can revalidate the TOD programmable register with its - * old contents (should be zero) otherwise set it to zero. + * Set the clock comparator register to the next expected value. */ - if (!mci->pr) - asm volatile( - " sr 0,0\n" - " sckpf" - : : : "0", "cc"); - else - asm volatile( - " l 0,0(%0)\n" - " sckpf" - : : "a" (&S390_lowcore.tod_progreg_save_area) - : "0", "cc"); -#endif - /* Revalidate clock comparator register */ - if (S390_lowcore.clock_comparator == -1) - set_clock_comparator(S390_lowcore.mcck_clock); - else - set_clock_comparator(S390_lowcore.clock_comparator); - /* Check if old PSW is valid */ - if (!mci->wp) - /* - * Can't tell if we come from user or kernel mode - * -> stopping machine. - */ - s390_handle_damage("old psw invalid."); + set_clock_comparator(get_lowcore()->clock_comparator); + if (!mci.gr || !mci.fp || !mci.fc) + return false; + /* + * The vector validity must only be checked if not running a + * KVM guest. For KVM guests the machine check is forwarded by + * KVM and it is the responsibility of the guest to take + * appropriate actions. The host vector or FPU values have been + * saved by KVM and will be restored by KVM. + */ + if (!mci.vr && !test_cpu_flag(CIF_MCCK_GUEST)) + return false; + if (!mci.ar) + return false; + /* + * Two cases for guarded storage registers: + * - machine check in kernel or userspace + * - machine check while running SIE (KVM guest) + * For kernel or userspace the userspace values of guarded storage + * control can not be recreated, the process must be terminated. + * For SIE the guest values of guarded storage can not be recreated. + * This is either due to a bug or due to GS being disabled in the + * guest. The guest will be notified by KVM code and the guests machine + * check handling must take care of this. The host values are saved by + * KVM and are not affected. + */ + cr2.reg = get_lowcore()->cregs_save_area[2]; + if (cr2.gse && !mci.gs && !test_cpu_flag(CIF_MCCK_GUEST)) + return false; + if (!mci.ms || !mci.pm || !mci.ia) + return false; + return true; +} +NOKPROBE_SYMBOL(nmi_registers_valid); - if (!mci->ms || !mci->pm || !mci->ia) - kill_task = 1; +/* + * Backup the guest's machine check info to its description block + */ +static void notrace s390_backup_mcck_info(struct pt_regs *regs) +{ + struct mcck_volatile_info *mcck_backup; + struct sie_page *sie_page; + + /* r14 contains the sie block, which was set in sie64a */ + struct kvm_s390_sie_block *sie_block = phys_to_virt(regs->gprs[14]); - return kill_task; + if (sie_block == NULL) + /* Something's seriously wrong, stop system. */ + s390_handle_damage(); + + sie_page = container_of(sie_block, struct sie_page, sie_block); + mcck_backup = &sie_page->mcck_info; + mcck_backup->mcic = get_lowcore()->mcck_interruption_code & + ~(MCCK_CODE_CP | MCCK_CODE_EXT_DAMAGE); + mcck_backup->ext_damage_code = get_lowcore()->external_damage_code; + mcck_backup->failing_storage_address = get_lowcore()->failing_storage_address; } +NOKPROBE_SYMBOL(s390_backup_mcck_info); #define MAX_IPD_COUNT 29 #define MAX_IPD_TIME (5 * 60 * USEC_PER_SEC) /* 5 minutes */ #define ED_STP_ISLAND 6 /* External damage STP island check */ #define ED_STP_SYNC 7 /* External damage STP sync check */ -#define ED_ETR_SYNC 12 /* External damage ETR sync check */ -#define ED_ETR_SWITCH 13 /* External damage ETR switch to local */ + +#define MCCK_CODE_NO_GUEST (MCCK_CODE_CP | MCCK_CODE_EXT_DAMAGE) /* * machine check handler. @@ -248,44 +366,41 @@ void notrace s390_do_machine_check(struct pt_regs *regs) static int ipd_count; static DEFINE_SPINLOCK(ipd_lock); static unsigned long long last_ipd; + struct lowcore *lc = get_lowcore(); struct mcck_struct *mcck; unsigned long long tmp; - struct mci *mci; - int umode; + irqentry_state_t irq_state; + union mci mci; + unsigned long mcck_dam_code; + int mcck_pending = 0; + + irq_state = irqentry_nmi_enter(regs); - nmi_enter(); + if (user_mode(regs)) + update_timer_mcck(); inc_irq_stat(NMI_NMI); - mci = (struct mci *) &S390_lowcore.mcck_interruption_code; - mcck = &__get_cpu_var(cpu_mcck); - umode = user_mode(regs); + mci.val = lc->mcck_interruption_code; + mcck = this_cpu_ptr(&cpu_mcck); - if (mci->sd) { - /* System damage -> stopping machine */ - s390_handle_damage("received system damage machine check."); - } - if (mci->pd) { - if (mci->b) { + /* + * Reinject the instruction processing damages' machine checks + * including Delayed Access Exception into the guest + * instead of damaging the host if they happen in the guest. + */ + if (mci.pd && !test_cpu_flag(CIF_MCCK_GUEST)) { + if (mci.b) { /* Processing backup -> verify if we can survive this */ u64 z_mcic, o_mcic, t_mcic; -#ifdef CONFIG_64BIT z_mcic = (1ULL<<63 | 1ULL<<59 | 1ULL<<29); o_mcic = (1ULL<<43 | 1ULL<<42 | 1ULL<<41 | 1ULL<<40 | 1ULL<<36 | 1ULL<<35 | 1ULL<<34 | 1ULL<<32 | 1ULL<<30 | 1ULL<<21 | 1ULL<<20 | 1ULL<<17 | 1ULL<<16); -#else - z_mcic = (1ULL<<63 | 1ULL<<59 | 1ULL<<57 | 1ULL<<50 | - 1ULL<<29); - o_mcic = (1ULL<<43 | 1ULL<<42 | 1ULL<<41 | 1ULL<<40 | - 1ULL<<36 | 1ULL<<35 | 1ULL<<34 | 1ULL<<32 | - 1ULL<<30 | 1ULL<<20 | 1ULL<<17 | 1ULL<<16); -#endif - t_mcic = *(u64 *)mci; + t_mcic = mci.val; if (((t_mcic & z_mcic) != 0) || ((t_mcic & o_mcic) != o_mcic)) { - s390_handle_damage("processing backup machine " - "check with damage."); + s390_handle_damage(); } /* @@ -300,76 +415,95 @@ void notrace s390_do_machine_check(struct pt_regs *regs) ipd_count = 1; last_ipd = tmp; if (ipd_count == MAX_IPD_COUNT) - s390_handle_damage("too many ipd retries."); + s390_handle_damage(); spin_unlock(&ipd_lock); } else { /* Processing damage -> stopping machine */ - s390_handle_damage("received instruction processing " - "damage machine check."); + s390_handle_damage(); } } - if (s390_revalidate_registers(mci)) { - if (umode) { - /* - * Couldn't restore all register contents while in - * user mode -> mark task for termination. - */ - mcck->kill_task = 1; - mcck->mcck_code = *(unsigned long long *) mci; - set_thread_flag(TIF_MCCK_PENDING); - } else { - /* - * Couldn't restore all register contents while in - * kernel mode -> stopping machine. - */ - s390_handle_damage("unable to revalidate registers."); - } + if (!nmi_registers_valid(mci)) { + if (!user_mode(regs)) + s390_handle_damage(); + /* + * Couldn't restore all register contents for the + * user space process -> mark task for termination. + */ + mcck->kill_task = 1; + mcck->mcck_code = mci.val; + mcck_pending = 1; } - if (mci->cd) { + + /* + * Backup the machine check's info if it happens when the guest + * is running. + */ + if (test_cpu_flag(CIF_MCCK_GUEST)) + s390_backup_mcck_info(regs); + + if (mci.cd) { /* Timing facility damage */ - s390_handle_damage("TOD clock damaged"); + s390_handle_damage(); } - if (mci->ed && mci->ec) { + if (mci.ed && mci.ec) { /* External damage */ - if (S390_lowcore.external_damage_code & (1U << ED_ETR_SYNC)) - etr_sync_check(); - if (S390_lowcore.external_damage_code & (1U << ED_ETR_SWITCH)) - etr_switch_to_local(); - if (S390_lowcore.external_damage_code & (1U << ED_STP_SYNC)) - stp_sync_check(); - if (S390_lowcore.external_damage_code & (1U << ED_STP_ISLAND)) - stp_island_check(); + if (lc->external_damage_code & (1U << ED_STP_SYNC)) + mcck->stp_queue |= stp_sync_check(); + if (lc->external_damage_code & (1U << ED_STP_ISLAND)) + mcck->stp_queue |= stp_island_check(); + mcck_pending = 1; } - if (mci->se) + /* + * Reinject storage related machine checks into the guest if they + * happen when the guest is running. + */ + if (!test_cpu_flag(CIF_MCCK_GUEST)) { /* Storage error uncorrected */ - s390_handle_damage("received storage error uncorrected " - "machine check."); - if (mci->ke) + if (mci.se) + s390_handle_damage(); /* Storage key-error uncorrected */ - s390_handle_damage("received storage key-error uncorrected " - "machine check."); - if (mci->ds && mci->fa) + if (mci.ke) + s390_handle_damage(); /* Storage degradation */ - s390_handle_damage("received storage degradation machine " - "check."); - if (mci->cp) { + if (mci.ds && mci.fa) + s390_handle_damage(); + } + if (mci.cp) { /* Channel report word pending */ mcck->channel_report = 1; - set_thread_flag(TIF_MCCK_PENDING); + mcck_pending = 1; } - if (mci->w) { + if (mci.w) { /* Warning pending */ mcck->warning = 1; - set_thread_flag(TIF_MCCK_PENDING); + mcck_pending = 1; } - nmi_exit(); + + /* + * If there are only Channel Report Pending and External Damage + * machine checks, they will not be reinjected into the guest + * because they refer to host conditions only. + */ + mcck_dam_code = (mci.val & MCIC_SUBCLASS_MASK); + if (test_cpu_flag(CIF_MCCK_GUEST) && + (mcck_dam_code & MCCK_CODE_NO_GUEST) != mcck_dam_code) { + /* Set exit reason code for host's later handling */ + *((long *)(regs->gprs[15] + __SF_SIE_REASON)) = -EINTR; + } + clear_cpu_flag(CIF_MCCK_GUEST); + + if (mcck_pending) + schedule_mcck_handler(); + + irqentry_nmi_exit(regs, irq_state); } +NOKPROBE_SYMBOL(s390_do_machine_check); static int __init machine_check_init(void) { - ctl_set_bit(14, 25); /* enable external damage MCH */ - ctl_set_bit(14, 27); /* enable system recovery MCH */ - ctl_set_bit(14, 24); /* enable warning MCH */ + system_ctl_set_bit(14, CR14_EXTERNAL_DAMAGE_SUBMASK_BIT); + system_ctl_set_bit(14, CR14_RECOVERY_SUBMASK_BIT); + system_ctl_set_bit(14, CR14_WARNING_SUBMASK_BIT); return 0; } -arch_initcall(machine_check_init); +early_initcall(machine_check_init); |
