s390/dump: rework CPU register dump code

To collect the CPU registers of the crashed system allocated a single
page with memblock_alloc_base and use it as a copy buffer. Replace the
stop-and-store-status sigp with a store-status-at-address sigp in
smp_save_dump_cpus() and smp_store_status(). In both cases the target
CPU is already stopped and store-status-at-address avoids the detour
via the absolute zero page.

For kexec simplify s390_reset_system and call store_status() before
the prefix register of the boot CPU has been set to zero. Use STPX
to store the prefix register and remove dump_prefix_page.
Acked-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/fpu/internal.h

@@ -12,21 +12,13 @@
 #include <asm/ctl_reg.h>
 #include <asm/fpu/types.h>
 
-static inline void save_vx_regs_safe(__vector128 *vxrs)
+static inline void save_vx_regs(__vector128 *vxrs)
 {
-	unsigned long cr0, flags;
-
-	flags = arch_local_irq_save();
-	__ctl_store(cr0, 0, 0);
-	__ctl_set_bit(0, 17);
-	__ctl_set_bit(0, 18);
 	asm volatile(
 		"	la	1,%0\n"
 		"	.word	0xe70f,0x1000,0x003e\n"	/* vstm 0,15,0(1) */
 		"	.word	0xe70f,0x1100,0x0c3e\n"	/* vstm 16,31,256(1) */
 		: "=Q" (*(struct vx_array *) vxrs) : : "1");
-	__ctl_load(cr0, 0, 0);
-	arch_local_irq_restore(flags);
 }
 
 static inline void convert_vx_to_fp(freg_t *fprs, __vector128 *vxrs)

arch/s390/include/asm/ipl.h

@@ -87,7 +87,6 @@ struct ipl_parameter_block {
  * IPL validity flags
  */
 extern u32 ipl_flags;
-extern u32 dump_prefix_page;
 
 struct dump_save_areas {
 	struct save_area_ext **areas;
@@ -176,7 +175,7 @@ enum diag308_rc {
 
 extern int diag308(unsigned long subcode, void *addr);
 extern void diag308_reset(void);
-extern void store_status(void);
+extern void store_status(void (*fn)(void *), void *data);
 extern void lgr_info_log(void);
 
 #endif /* _ASM_S390_IPL_H */

arch/s390/include/asm/reset.h

@@ -15,6 +15,5 @@ struct reset_call {
 
 extern void register_reset_call(struct reset_call *reset);
 extern void unregister_reset_call(struct reset_call *reset);
-extern void s390_reset_system(void (*fn_pre)(void),
-			      void (*fn_post)(void *), void *data);
+extern void s390_reset_system(void);
 #endif /* _ASM_S390_RESET_H */

arch/s390/include/asm/smp.h

@@ -18,6 +18,7 @@
 extern struct mutex smp_cpu_state_mutex;
 extern unsigned int smp_cpu_mt_shift;
 extern unsigned int smp_cpu_mtid;
+extern __vector128 __initdata boot_cpu_vector_save_area[__NUM_VXRS];
 
 extern int __cpu_up(unsigned int cpu, struct task_struct *tidle);
 
@@ -55,7 +56,6 @@ static inline int smp_store_status(int cpu) { return 0; }
 static inline int smp_vcpu_scheduled(int cpu) { return 1; }
 static inline void smp_yield_cpu(int cpu) { }
 static inline void smp_fill_possible_mask(void) { }
-static inline void smp_save_dump_cpus(void) { }
 
 #endif /* CONFIG_SMP */
 

arch/s390/kernel/early.c

@@ -335,6 +335,14 @@ static __init void detect_machine_facilities(void)
 	}
 }
 
+static inline void save_vector_registers(void)
+{
+#ifdef CONFIG_CRASH_DUMP
+	if (test_facility(129))
+		save_vx_regs(boot_cpu_vector_save_area);
+#endif
+}
+
 static int __init disable_vector_extension(char *str)
 {
 	S390_lowcore.machine_flags &= ~MACHINE_FLAG_VX;
@@ -451,6 +459,7 @@ void __init startup_init(void)
 	detect_diag9c();
 	detect_diag44();
 	detect_machine_facilities();
+	save_vector_registers();
 	setup_topology();
 	sclp_early_detect();
 	lockdep_on();

arch/s390/kernel/ipl.c

@@ -2039,10 +2039,7 @@ static void do_reset_calls(void)
 		reset->fn();
 }
 
-u32 dump_prefix_page;
-
-void s390_reset_system(void (*fn_pre)(void),
-		       void (*fn_post)(void *), void *data)
+void s390_reset_system(void)
 {
 	struct _lowcore *lc;
 
@@ -2051,9 +2048,6 @@ void s390_reset_system(void (*fn_pre)(void),
 	/* Stack for interrupt/machine check handler */
 	lc->panic_stack = S390_lowcore.panic_stack;
 
-	/* Save prefix page address for dump case */
-	dump_prefix_page = (u32)(unsigned long) lc;
-
 	/* Disable prefixing */
 	set_prefix(0);
 
@@ -2077,14 +2071,5 @@ void s390_reset_system(void (*fn_pre)(void),
 	S390_lowcore.subchannel_id = 0;
 	S390_lowcore.subchannel_nr = 0;
 
-	/* Store status at absolute zero */
-	store_status();
-
-	/* Call function before reset */
-	if (fn_pre)
-		fn_pre();
 	do_reset_calls();
-	/* Call function after reset */
-	if (fn_post)
-		fn_post(data);
 }

arch/s390/kernel/machine_kexec.c

@@ -34,40 +34,6 @@ extern const unsigned long long relocate_kernel_len;
 
 #ifdef CONFIG_CRASH_DUMP
 
-/*
- * Initialize CPU ELF notes
- */
-static void setup_regs(void)
-{
-	struct save_area *sa, *sa_0;
-	unsigned long prefix;
-	int cpu, this_cpu;
-
-	/* setup_regs is called with the prefix register = 0 */
-	sa_0 = (struct save_area *) __LC_FPREGS_SAVE_AREA;
-
-	/* Get status of this CPU out of absolute zero */
-	prefix = (unsigned long) S390_lowcore.prefixreg_save_area;
-	sa = (struct save_area *)(prefix + __LC_FPREGS_SAVE_AREA);
-	memcpy(sa, sa_0, sizeof(struct save_area));
-	if (MACHINE_HAS_VX) {
-		struct _lowcore *lc = (struct _lowcore *) prefix;
-		save_vx_regs_safe((void *) lc->vector_save_area_addr);
-	}
-
-	/* Get status of the other CPUs */
-	this_cpu = smp_find_processor_id(stap());
-	for_each_online_cpu(cpu) {
-		if (cpu == this_cpu)
-			continue;
-		if (smp_store_status(cpu))
-			continue;
-		prefix = (unsigned long) S390_lowcore.prefixreg_save_area;
-		sa = (struct save_area *)(prefix + __LC_FPREGS_SAVE_AREA);
-		memcpy(sa, sa_0, sizeof(struct save_area));
-	}
-}
-
 /*
  * PM notifier callback for kdump
  */
@@ -99,14 +65,66 @@ static int __init machine_kdump_pm_init(void)
 arch_initcall(machine_kdump_pm_init);
 
 /*
- * Start kdump: We expect here that a store status has been done on our CPU
+ * Reset the system, copy boot CPU registers to absolute zero,
+ * and jump to the kdump image
  */
 static void __do_machine_kdump(void *image)
 {
-	int (*start_kdump)(int) = (void *)((struct kimage *) image)->start;
+	int (*start_kdump)(int);
+	unsigned long prefix;
+
+	/* store_status() saved the prefix register to lowcore */
+	prefix = (unsigned long) S390_lowcore.prefixreg_save_area;
+
+	/* Now do the reset  */
+	s390_reset_system();
+
+	/*
+	 * Copy dump CPU store status info to absolute zero.
+	 * This need to be done *after* s390_reset_system set the
+	 * prefix register of this CPU to zero
+	 */
+	memcpy((void *) __LC_FPREGS_SAVE_AREA,
+	       (void *)(prefix + __LC_FPREGS_SAVE_AREA), 512);
 
 	__load_psw_mask(PSW_MASK_BASE | PSW_DEFAULT_KEY | PSW_MASK_EA | PSW_MASK_BA);
+	start_kdump = (void *)((struct kimage *) image)->start;
 	start_kdump(1);
+
+	/* Die if start_kdump returns */
+	disabled_wait((unsigned long) __builtin_return_address(0));
+}
+
+/*
+ * Start kdump: create a LGR log entry, store status of all CPUs and
+ * branch to __do_machine_kdump.
+ */
+static noinline void __machine_kdump(void *image)
+{
+	int this_cpu, cpu;
+
+	lgr_info_log();
+	/* Get status of the other CPUs */
+	this_cpu = smp_find_processor_id(stap());
+	for_each_online_cpu(cpu) {
+		if (cpu == this_cpu)
+			continue;
+		if (smp_store_status(cpu))
+			continue;
+	}
+	/* Store status of the boot CPU */
+	if (MACHINE_HAS_VX)
+		save_vx_regs((void *) &S390_lowcore.vector_save_area);
+	/*
+	 * To create a good backchain for this CPU in the dump store_status
+	 * is passed the address of a function. The address is saved into
+	 * the PSW save area of the boot CPU and the function is invoked as
+	 * a tail call of store_status. The backchain in the dump will look
+	 * like this:
+	 *   restart_int_handler ->  __machine_kexec -> __do_machine_kdump
+	 * The call to store_status() will not return.
+	 */
+	store_status(__do_machine_kdump, image);
 }
 #endif
 
@@ -229,10 +247,14 @@ static void __do_machine_kexec(void *data)
 	relocate_kernel_t data_mover;
 	struct kimage *image = data;
 
+	s390_reset_system();
 	data_mover = (relocate_kernel_t) page_to_phys(image->control_code_page);
 
 	/* Call the moving routine */
 	(*data_mover)(&image->head, image->start);
+
+	/* Die if kexec returns */
+	disabled_wait((unsigned long) __builtin_return_address(0));
 }
 
 /*
@@ -245,14 +267,10 @@ static void __machine_kexec(void *data)
 	tracing_off();
 	debug_locks_off();
 #ifdef CONFIG_CRASH_DUMP
-	if (((struct kimage *) data)->type == KEXEC_TYPE_CRASH) {
-
-		lgr_info_log();
-		s390_reset_system(setup_regs, __do_machine_kdump, data);
-	} else
+	if (((struct kimage *) data)->type == KEXEC_TYPE_CRASH)
+		__machine_kdump(data);
 #endif
-		s390_reset_system(NULL, __do_machine_kexec, data);
-	disabled_wait((unsigned long) __builtin_return_address(0));
+	__do_machine_kexec(data);
 }
 
 /*

arch/s390/kernel/reipl.S

@@ -9,12 +9,11 @@
 #include <asm/sigp.h>
 
 #
-# store_status
+# Issue "store status" for the current CPU to its prefix page
+# and call passed function afterwards
 #
-# Prerequisites to run this function:
-# - Prefix register is set to zero
-# - Original prefix register is stored in "dump_prefix_page"
-# - Lowcore protection is off
+# r2 = Function to be called after store status
+# r3 = Parameter for function
 #
 ENTRY(store_status)
 	/* Save register one and load save area base */
@@ -53,23 +52,23 @@ ENTRY(store_status)
 	/* CPU timer */
 	lghi	%r1,__LC_CPU_TIMER_SAVE_AREA
 	stpt	0(%r1)
-	/* Saved prefix register */
+	/* Store prefix register */
 	lghi	%r1,__LC_PREFIX_SAVE_AREA
-	larl	%r2,dump_prefix_page
-	mvc	0(4,%r1),0(%r2)
+	stpx	0(%r1)
 	/* Clock comparator - seven bytes */
 	lghi	%r1,__LC_CLOCK_COMP_SAVE_AREA
-	larl	%r2,.Lclkcmp
-	stckc	0(%r2)
-	mvc	1(7,%r1),1(%r2)
+	larl	%r4,.Lclkcmp
+	stckc	0(%r4)
+	mvc	1(7,%r1),1(%r4)
 	/* Program status word */
 	lghi	%r1,__LC_PSW_SAVE_AREA
-	epsw	%r2,%r3
-	st	%r2,0(%r1)
-	st	%r3,4(%r1)
-	larl	%r2,store_status
+	epsw	%r4,%r5
+	st	%r4,0(%r1)
+	st	%r5,4(%r1)
 	stg	%r2,8(%r1)
-	br	%r14
+	lgr	%r1,%r2
+	lgr	%r2,%r3
+	br	%r1
 
 	.section .bss
 	.align	8
@@ -84,9 +83,11 @@ ENTRY(store_status)
 ENTRY(do_reipl_asm)
 		basr	%r13,0
 .Lpg0:		lpswe	.Lnewpsw-.Lpg0(%r13)
-.Lpg1:		brasl	%r14,store_status
+.Lpg1:		lgr	%r3,%r2
+		larl	%r2,.Lstatus
+		brasl	%r14,store_status
 
-		lctlg	%c6,%c6,.Lall-.Lpg0(%r13)
+.Lstatus:	lctlg	%c6,%c6,.Lall-.Lpg0(%r13)
 		lgr	%r1,%r2
 		mvc	__LC_PGM_NEW_PSW(16),.Lpcnew-.Lpg0(%r13)
 		stsch	.Lschib-.Lpg0(%r13)

arch/s390/kernel/setup.c

@@ -865,11 +865,13 @@ void __init setup_arch(char **cmdline_p)
 
 	check_initrd();
 	reserve_crashkernel();
+#ifdef CONFIG_CRASH_DUMP
 	/*
 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
 	 * Therefore CPU and device initialization should be done afterwards.
 	 */
 	smp_save_dump_cpus();
+#endif
 
 	setup_resources();
 	setup_vmcoreinfo();

arch/s390/kernel/smp.c

@@ -80,6 +80,10 @@ EXPORT_SYMBOL(smp_cpu_mt_shift);
 unsigned int smp_cpu_mtid;
 EXPORT_SYMBOL(smp_cpu_mtid);
 
+#ifdef CONFIG_CRASH_DUMP
+__vector128 __initdata boot_cpu_vector_save_area[__NUM_VXRS];
+#endif
+
 static unsigned int smp_max_threads __initdata = -1U;
 
 static int __init early_nosmt(char *s)
@@ -105,8 +109,7 @@ DEFINE_MUTEX(smp_cpu_state_mutex);
 /*
  * Signal processor helper functions.
  */
-static inline int __pcpu_sigp_relax(u16 addr, u8 order, unsigned long parm,
-				    u32 *status)
+static inline int __pcpu_sigp_relax(u16 addr, u8 order, unsigned long parm)
 {
 	int cc;
 
@@ -538,53 +541,24 @@ EXPORT_SYMBOL(smp_ctl_clear_bit);
 
 #ifdef CONFIG_CRASH_DUMP
 
-static void __init __smp_store_cpu_state(struct save_area_ext *sa_ext,
-					 u16 address, int is_boot_cpu)
-{
-	void *lc = (void *)(unsigned long) store_prefix();
-	unsigned long vx_sa;
-
-	if (is_boot_cpu) {
-		/* Copy the registers of the boot CPU. */
-		copy_oldmem_kernel(&sa_ext->sa, (void *) __LC_FPREGS_SAVE_AREA,
-				   sizeof(sa_ext->sa));
-		if (MACHINE_HAS_VX)
-			save_vx_regs_safe(sa_ext->vx_regs);
-		return;
-	}
-	/* Get the registers of a non-boot cpu. */
-	__pcpu_sigp_relax(address, SIGP_STOP_AND_STORE_STATUS, 0, NULL);
-	memcpy_real(&sa_ext->sa, lc + __LC_FPREGS_SAVE_AREA, sizeof(sa_ext->sa));
-	if (!MACHINE_HAS_VX)
-		return;
-	/* Get the VX registers */
-	vx_sa = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
-	if (!vx_sa)
-		panic("could not allocate memory for VX save area\n");
-	__pcpu_sigp_relax(address, SIGP_STORE_ADDITIONAL_STATUS, vx_sa, NULL);
-	memcpy(sa_ext->vx_regs, (void *) vx_sa, sizeof(sa_ext->vx_regs));
-	memblock_free(vx_sa, PAGE_SIZE);
-}
-
 int smp_store_status(int cpu)
 {
-	unsigned long vx_sa;
-	struct pcpu *pcpu;
+	struct pcpu *pcpu = pcpu_devices + cpu;
+	unsigned long pa;
 
-	pcpu = pcpu_devices + cpu;
-	if (__pcpu_sigp_relax(pcpu->address, SIGP_STOP_AND_STORE_STATUS,
-			      0, NULL) != SIGP_CC_ORDER_CODE_ACCEPTED)
+	pa = __pa(&pcpu->lowcore->floating_pt_save_area);
+	if (__pcpu_sigp_relax(pcpu->address, SIGP_STORE_STATUS_AT_ADDRESS,
+			      pa) != SIGP_CC_ORDER_CODE_ACCEPTED)
 		return -EIO;
 	if (!MACHINE_HAS_VX)
 		return 0;
-	vx_sa = __pa(pcpu->lowcore->vector_save_area_addr);
-	__pcpu_sigp_relax(pcpu->address, SIGP_STORE_ADDITIONAL_STATUS,
-			  vx_sa, NULL);
+	pa = __pa(pcpu->lowcore->vector_save_area_addr);
+	if (__pcpu_sigp_relax(pcpu->address, SIGP_STORE_ADDITIONAL_STATUS,
+			      pa) != SIGP_CC_ORDER_CODE_ACCEPTED)
+		return -EIO;
 	return 0;
 }
 
-#endif /* CONFIG_CRASH_DUMP */
-
 /*
  * Collect CPU state of the previous, crashed system.
  * There are four cases:
@@ -593,7 +567,7 @@ int smp_store_status(int cpu)
  *    The state for all CPUs except the boot CPU needs to be collected
  *    with sigp stop-and-store-status. The boot CPU state is located in
  *    the absolute lowcore of the memory stored in the HSA. The zcore code
- *    will allocate the save area and copy the boot CPU state from the HSA.
+ *    will copy the boot CPU state from the HSA.
  * 2) stand-alone kdump for SCSI (zfcp dump with swapped memory)
  *    condition: OLDMEM_BASE != NULL && ipl_info.type == IPL_TYPE_FCP_DUMP
  *    The state for all CPUs except the boot CPU needs to be collected
@@ -611,21 +585,49 @@ int smp_store_status(int cpu)
  *    This case does not exist for s390 anymore, setup_arch explicitly
  *    deactivates the elfcorehdr= kernel parameter
  */
+static __init void smp_save_cpu_vxrs(struct save_area_ext *sa_ext, u16 addr,
+				     bool is_boot_cpu, unsigned long page)
+{
+	__vector128 *vxrs = (__vector128 *) page;
+
+	if (is_boot_cpu)
+		vxrs = boot_cpu_vector_save_area;
+	else
+		__pcpu_sigp_relax(addr, SIGP_STORE_ADDITIONAL_STATUS, page);
+	memcpy(&sa_ext->vx_regs, vxrs, sizeof(sa_ext->vx_regs));
+}
+
+static __init void smp_save_cpu_regs(struct save_area_ext *sa_ext, u16 addr,
+				     bool is_boot_cpu, unsigned long page)
+{
+	void *regs = (void *) page;
+
+	if (is_boot_cpu)
+		copy_oldmem_kernel(regs, (void *) __LC_FPREGS_SAVE_AREA, 512);
+	else
+		__pcpu_sigp_relax(addr, SIGP_STORE_STATUS_AT_ADDRESS, page);
+	memcpy(&sa_ext->sa, regs, sizeof(sa_ext->sa));
+}
+
 void __init smp_save_dump_cpus(void)
 {
-#ifdef CONFIG_CRASH_DUMP
 	int addr, cpu, boot_cpu_addr, max_cpu_addr;
 	struct save_area_ext *sa_ext;
+	unsigned long page;
 	bool is_boot_cpu;
 
 	if (!(OLDMEM_BASE || ipl_info.type == IPL_TYPE_FCP_DUMP))
 		/* No previous system present, normal boot. */
 		return;
+	/* Allocate a page as dumping area for the store status sigps */
+	page = memblock_alloc_base(PAGE_SIZE, PAGE_SIZE, 1UL << 31);
+	if (!page)
+		panic("could not allocate memory for save area\n");
 	/* Set multi-threading state to the previous system. */
 	pcpu_set_smt(sclp.mtid_prev);
 	max_cpu_addr = SCLP_MAX_CORES << sclp.mtid_prev;
 	for (cpu = 0, addr = 0; addr <= max_cpu_addr; addr++) {
-		if (__pcpu_sigp_relax(addr, SIGP_SENSE, 0, NULL) ==
+		if (__pcpu_sigp_relax(addr, SIGP_SENSE, 0) ==
 		    SIGP_CC_NOT_OPERATIONAL)
 			continue;
 		cpu += 1;
@@ -634,7 +636,7 @@ void __init smp_save_dump_cpus(void)
 	dump_save_areas.count = cpu;
 	boot_cpu_addr = stap();
 	for (cpu = 0, addr = 0; addr <= max_cpu_addr; addr++) {
-		if (__pcpu_sigp_relax(addr, SIGP_SENSE, 0, NULL) ==
+		if (__pcpu_sigp_relax(addr, SIGP_SENSE, 0) ==
 		    SIGP_CC_NOT_OPERATIONAL)
 			continue;
 		sa_ext = (void *) memblock_alloc(sizeof(*sa_ext), 8);
@@ -643,16 +645,24 @@ void __init smp_save_dump_cpus(void)
 			panic("could not allocate memory for save area\n");
 		is_boot_cpu = (addr == boot_cpu_addr);
 		cpu += 1;
-		if (is_boot_cpu && !OLDMEM_BASE)
-			/* Skip boot CPU for standard zfcp dump. */
-			continue;
-		/* Get state for this CPU. */
-		__smp_store_cpu_state(sa_ext, addr, is_boot_cpu);
+		if (MACHINE_HAS_VX)
+			/* Get the vector registers */
+			smp_save_cpu_vxrs(sa_ext, addr, is_boot_cpu, page);
+		/*
+		 * For a zfcp dump OLDMEM_BASE == NULL and the registers
+		 * of the boot CPU are stored in the HSA. To retrieve
+		 * these registers an SCLP request is required which is
+		 * done by drivers/s390/char/zcore.c:init_cpu_info()
+		 */
+		if (!is_boot_cpu || OLDMEM_BASE)
+			/* Get the CPU registers */
+			smp_save_cpu_regs(sa_ext, addr, is_boot_cpu, page);
 	}
+	memblock_free(page, PAGE_SIZE);
 	diag308_reset();
 	pcpu_set_smt(0);
-#endif /* CONFIG_CRASH_DUMP */
 }
+#endif /* CONFIG_CRASH_DUMP */
 
 void smp_cpu_set_polarization(int cpu, int val)
 {
@@ -676,7 +686,7 @@ static struct sclp_core_info *smp_get_core_info(void)
 		for (address = 0;
 		     address < (SCLP_MAX_CORES << smp_cpu_mt_shift);
 		     address += (1U << smp_cpu_mt_shift)) {
-			if (__pcpu_sigp_relax(address, SIGP_SENSE, 0, NULL) ==
+			if (__pcpu_sigp_relax(address, SIGP_SENSE, 0) ==
 			    SIGP_CC_NOT_OPERATIONAL)
 				continue;
 			info->core[info->configured].core_id =

drivers/s390/char/zcore.c

@@ -129,8 +129,6 @@ static int __init init_cpu_info(void)
 		TRACE("could not copy from HSA\n");
 		return -EIO;
 	}
-	if (MACHINE_HAS_VX)
-		save_vx_regs_safe(sa_ext->vx_regs);
 	return 0;
 }
 

drivers/s390/cio/cio.c

@@ -917,7 +917,7 @@ void reipl_ccw_dev(struct ccw_dev_id *devid)
 {
 	struct subchannel_id uninitialized_var(schid);
 
-	s390_reset_system(NULL, NULL, NULL);
+	s390_reset_system();
 	if (reipl_find_schid(devid, &schid) != 0)
 		panic("IPL Device not found\n");
 	do_reipl_asm(*((__u32*)&schid));