1 files changed, 159 insertions, 239 deletions

diff --git a/arch/s390/kernel/nmi.c b/arch/s390/kernel/nmi.c
index 53ed3884fe64..3da371c144eb 100644
--- a/arch/s390/kernel/nmi.c
+++ b/arch/s390/kernel/nmi.c
@@ -9,8 +9,10 @@
  */
 
 #include <linux/kernel_stat.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>
@@ -18,20 +20,19 @@
 #include <linux/time.h>
 #include <linux/module.h>
 #include <linux/sched/signal.h>
-
+#include <linux/kvm_host.h>
 #include <linux/export.h>
 #include <asm/lowcore.h>
+#include <asm/ctlreg.h>
+#include <asm/fpu.h>
 #include <asm/smp.h>
 #include <asm/stp.h>
 #include <asm/cputime.h>
 #include <asm/nmi.h>
 #include <asm/crw.h>
-#include <asm/switch_to.h>
-#include <asm/ctl_reg.h>
 #include <asm/asm-offsets.h>
 #include <asm/pai.h>
-
-#include <linux/kvm_host.h>
+#include <asm/vtime.h>
 
 struct mcck_struct {
 	unsigned int kill_task : 1;
@@ -42,19 +43,10 @@ struct mcck_struct {
 };
 
 static DEFINE_PER_CPU(struct mcck_struct, cpu_mcck);
-static struct kmem_cache *mcesa_cache;
-static unsigned long mcesa_origin_lc;
 
 static inline int nmi_needs_mcesa(void)
 {
-	return MACHINE_HAS_VX || MACHINE_HAS_GS;
-}
-
-static inline unsigned long nmi_get_mcesa_size(void)
-{
-	if (MACHINE_HAS_GS)
-		return MCESA_MAX_SIZE;
-	return MCESA_MIN_SIZE;
+	return cpu_has_vx() || cpu_has_gs();
 }
 
 /*
@@ -63,47 +55,34 @@ static inline unsigned long nmi_get_mcesa_size(void)
  * structure. The structure is required for machine check happening
  * early in the boot process.
  */
-static struct mcesa boot_mcesa __initdata __aligned(MCESA_MAX_SIZE);
+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 (MACHINE_HAS_GS)
+	if (cpu_has_gs())
 		*mcesad |= ilog2(MCESA_MAX_SIZE);
 }
 
-static void __init nmi_alloc_cache(void)
+int nmi_alloc_mcesa(u64 *mcesad)
 {
 	unsigned long size;
-
-	if (!nmi_needs_mcesa())
-		return;
-	size = nmi_get_mcesa_size();
-	if (size > MCESA_MIN_SIZE)
-		mcesa_origin_lc = ilog2(size);
-	/* create slab cache for the machine-check-extended-save-areas */
-	mcesa_cache = kmem_cache_create("nmi_save_areas", size, size, 0, NULL);
-	if (!mcesa_cache)
-		panic("Couldn't create nmi save area cache");
-}
-
-int __ref nmi_alloc_mcesa(u64 *mcesad)
-{
-	unsigned long origin;
+	void *origin;
 
 	*mcesad = 0;
 	if (!nmi_needs_mcesa())
 		return 0;
-	if (!mcesa_cache)
-		nmi_alloc_cache();
-	origin = (unsigned long) kmem_cache_alloc(mcesa_cache, GFP_KERNEL);
+	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((void *) origin);
-	*mcesad = __pa(origin) | mcesa_origin_lc;
+	kmemleak_not_leak(origin);
+	*mcesad = __pa(origin);
+	if (cpu_has_gs())
+		*mcesad |= ilog2(MCESA_MAX_SIZE);
 	return 0;
 }
 
@@ -111,12 +90,65 @@ void nmi_free_mcesa(u64 *mcesad)
 {
 	if (!nmi_needs_mcesa())
 		return;
-	kmem_cache_free(mcesa_cache, __va(*mcesad & MCESA_ORIGIN_MASK));
+	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 s390_handle_damage(void)
 {
+	struct lowcore *lc = get_lowcore();
+	union ctlreg0 cr0, cr0_new;
+	char message[100];
+	psw_t psw_save;
+	char *ptr;
+
 	smp_emergency_stop();
+	diag_amode31_ops.diag308_reset();
+	ptr = nmi_puts(message, "System stopped due to unrecoverable machine check, code: 0x");
+	u64_to_hex(ptr, lc->mcck_interruption_code);
+
+	/*
+	 * 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;
+	sclp_emergency_printk(message);
+
+	/*
+	 * 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();
 	while (1);
 }
@@ -126,19 +158,20 @@ 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)
+void s390_handle_mcck(void)
 {
 	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_mcck_disable();
+	local_mcck_save(mflags);
 	mcck = *this_cpu_ptr(&cpu_mcck);
 	memset(this_cpu_ptr(&cpu_mcck), 0, sizeof(mcck));
-	local_mcck_enable();
+	local_mcck_restore(mflags);
 
 	if (mcck.channel_report)
 		crw_handle_channel_report();
@@ -155,203 +188,80 @@ 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%016lx).\n", mcck.mcck_code);
 		printk(KERN_EMERG "mcck: task: %s, pid: %d.\n",
 		       current->comm, current->pid);
-		make_task_dead(SIGSEGV);
+		if (is_global_init(current))
+			panic("mcck: Attempting to kill init!\n");
+		do_send_sig_info(SIGKILL, SEND_SIG_PRIV, current, PIDTYPE_PID);
 	}
 }
 
-void noinstr s390_handle_mcck(struct pt_regs *regs)
-{
-	trace_hardirqs_off();
-	pai_kernel_enter(regs);
-	__s390_handle_mcck();
-	pai_kernel_exit(regs);
-	trace_hardirqs_on();
-}
-/*
- * returns 0 if all required registers are available
- * 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_validate_registers(union mci mci, int umode)
+static bool notrace nmi_registers_valid(union mci mci)
 {
-	struct mcesa *mcesa;
-	void *fpt_save_area;
 	union ctlreg2 cr2;
-	int kill_task;
-	u64 zero;
-
-	kill_task = 0;
-	zero = 0;
 
-	if (!mci.gr) {
-		/*
-		 * General purpose registers couldn't be restored and have
-		 * unknown contents. Stop system or terminate process.
-		 */
-		if (!umode)
-			s390_handle_damage();
-		kill_task = 1;
-	}
-	if (!mci.fp) {
-		/*
-		 * Floating point registers can't be restored. If the
-		 * kernel currently uses floating point registers the
-		 * system is stopped. If the process has its floating
-		 * pointer registers loaded it is terminated.
-		 */
-		if (S390_lowcore.fpu_flags & KERNEL_VXR_V0V7)
-			s390_handle_damage();
-		if (!test_cpu_flag(CIF_FPU))
-			kill_task = 1;
-	}
-	fpt_save_area = &S390_lowcore.floating_pt_save_area;
-	if (!mci.fc) {
-		/*
-		 * Floating point control register can't be restored.
-		 * If the kernel currently uses the floating pointer
-		 * registers and needs the FPC register the system is
-		 * stopped. If the process has its floating pointer
-		 * registers loaded it is terminated. Otherwise the
-		 * FPC is just validated.
-		 */
-		if (S390_lowcore.fpu_flags & KERNEL_FPC)
-			s390_handle_damage();
-		asm volatile(
-			"	lfpc	%0\n"
-			:
-			: "Q" (zero));
-		if (!test_cpu_flag(CIF_FPU))
-			kill_task = 1;
-	} else {
-		asm volatile(
-			"	lfpc	%0\n"
-			:
-			: "Q" (S390_lowcore.fpt_creg_save_area));
-	}
-
-	mcesa = __va(S390_lowcore.mcesad & MCESA_ORIGIN_MASK);
-	if (!MACHINE_HAS_VX) {
-		/* Validate floating point registers */
-		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)
-			: "memory");
-	} else {
-		/* Validate vector registers */
-		union ctlreg0 cr0;
-
-		/*
-		 * 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)) {
-			/*
-			 * Vector registers can't be restored. If the kernel
-			 * currently uses vector registers the system is
-			 * stopped. If the process has its vector registers
-			 * loaded it is terminated. Otherwise just validate
-			 * the registers.
-			 */
-			if (S390_lowcore.fpu_flags & KERNEL_VXR)
-				s390_handle_damage();
-			if (!test_cpu_flag(CIF_FPU))
-				kill_task = 1;
-		}
-		cr0.val = S390_lowcore.cregs_save_area[0];
-		cr0.afp = cr0.vx = 1;
-		__ctl_load(cr0.val, 0, 0);
-		asm volatile(
-			"	la	1,%0\n"
-			"	.word	0xe70f,0x1000,0x0036\n" /* vlm 0,15,0(1) */
-			"	.word	0xe70f,0x1100,0x0c36\n" /* vlm 16,31,256(1) */
-			:
-			: "Q" (*(struct vx_array *)mcesa->vector_save_area)
-			: "1");
-		__ctl_load(S390_lowcore.cregs_save_area[0], 0, 0);
-	}
-	/* Validate access registers */
-	asm volatile(
-		"	lam	0,15,0(%0)\n"
-		:
-		: "a" (&S390_lowcore.access_regs_save_area)
-		: "memory");
-	if (!mci.ar) {
-		/*
-		 * Access registers have unknown contents.
-		 * Terminating task.
-		 */
-		kill_task = 1;
-	}
-	/* Validate guarded storage registers */
-	cr2.val = S390_lowcore.cregs_save_area[2];
-	if (cr2.gse) {
-		if (!mci.gs) {
-			/*
-			 * 2 cases:
-			 * - 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.
-			 */
-			if (!test_cpu_flag(CIF_MCCK_GUEST))
-				kill_task = 1;
-		} else {
-			load_gs_cb((struct gs_cb *)mcesa->guarded_storage_save_area);
-		}
-	}
 	/*
-	 * The getcpu vdso syscall reads CPU number from the programmable
+	 * 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.
+	 * validity bit and load the CPU number into the TOD programmable field
+	 * unconditionally.
 	 */
 	set_tod_programmable_field(raw_smp_processor_id());
-	/* Validate clock comparator register */
-	set_clock_comparator(S390_lowcore.clock_comparator);
-
+	/*
+	 * Set the clock comparator register to the next expected value.
+	 */
+	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)
-		kill_task = 1;
-
-	return kill_task;
+		return false;
+	return true;
 }
-NOKPROBE_SYMBOL(s390_validate_registers);
+NOKPROBE_SYMBOL(nmi_registers_valid);
 
 /*
  * Backup the guest's machine check info to its description block
@@ -362,8 +272,7 @@ static void notrace s390_backup_mcck_info(struct pt_regs *regs)
 	struct sie_page *sie_page;
 
 	/* r14 contains the sie block, which was set in sie64a */
-	struct kvm_s390_sie_block *sie_block =
-			(struct kvm_s390_sie_block *) regs->gprs[14];
+	struct kvm_s390_sie_block *sie_block = phys_to_virt(regs->gprs[14]);
 
 	if (sie_block == NULL)
 		/* Something's seriously wrong, stop system. */
@@ -371,11 +280,10 @@ static void notrace s390_backup_mcck_info(struct pt_regs *regs)
 
 	sie_page = container_of(sie_block, struct sie_page, sie_block);
 	mcck_backup = &sie_page->mcck_info;
-	mcck_backup->mcic = S390_lowcore.mcck_interruption_code &
+	mcck_backup->mcic = get_lowcore()->mcck_interruption_code &
 				~(MCCK_CODE_CP | MCCK_CODE_EXT_DAMAGE);
-	mcck_backup->ext_damage_code = S390_lowcore.external_damage_code;
-	mcck_backup->failing_storage_address
-			= S390_lowcore.failing_storage_address;
+	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);
 
@@ -390,23 +298,25 @@ NOKPROBE_SYMBOL(s390_backup_mcck_info);
 /*
  * machine check handler.
  */
-int notrace s390_do_machine_check(struct pt_regs *regs)
+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;
+	irqentry_state_t irq_state;
 	union mci mci;
 	unsigned long mcck_dam_code;
 	int mcck_pending = 0;
 
-	nmi_enter();
+	irq_state = irqentry_nmi_enter(regs);
 
 	if (user_mode(regs))
 		update_timer_mcck();
 	inc_irq_stat(NMI_NMI);
-	mci.val = S390_lowcore.mcck_interruption_code;
+	mci.val = lc->mcck_interruption_code;
 	mcck = this_cpu_ptr(&cpu_mcck);
 
 	/*
@@ -449,7 +359,9 @@ int notrace s390_do_machine_check(struct pt_regs *regs)
 			s390_handle_damage();
 		}
 	}
-	if (s390_validate_registers(mci, user_mode(regs))) {
+	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.
@@ -472,13 +384,27 @@ int notrace s390_do_machine_check(struct pt_regs *regs)
 	}
 	if (mci.ed && mci.ec) {
 		/* External damage */
-		if (S390_lowcore.external_damage_code & (1U << ED_STP_SYNC))
+		if (lc->external_damage_code & (1U << ED_STP_SYNC))
 			mcck->stp_queue |= stp_sync_check();
-		if (S390_lowcore.external_damage_code & (1U << ED_STP_ISLAND))
+		if (lc->external_damage_code & (1U << ED_STP_ISLAND))
 			mcck->stp_queue |= stp_island_check();
 		mcck_pending = 1;
 	}
-
+	/*
+	 * 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 */
+		if (mci.se)
+			s390_handle_damage();
+		/* Storage key-error uncorrected */
+		if (mci.ke)
+			s390_handle_damage();
+		/* Storage degradation */
+		if (mci.ds && mci.fa)
+			s390_handle_damage();
+	}
 	if (mci.cp) {
 		/* Channel report word pending */
 		mcck->channel_report = 1;
@@ -503,24 +429,18 @@ int notrace s390_do_machine_check(struct pt_regs *regs)
 	}
 	clear_cpu_flag(CIF_MCCK_GUEST);
 
-	if (user_mode(regs) && mcck_pending) {
-		nmi_exit();
-		return 1;
-	}
-
 	if (mcck_pending)
 		schedule_mcck_handler();
 
-	nmi_exit();
-	return 0;
+	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;
 }
 early_initcall(machine_check_init);