Merge branch 'parisc-4.19-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux

Pull more parisc updates from Helge Deller:

 - fix boot failure of 64-bit kernel. It got broken by the unwind
   optimization commit in merge window.

 - fix 64-bit userspace support (static 64-bit applications only, e.g.
   we don't yet have 64-bit userspace support in glibc).

 - consolidate unwind initialization code.

 - add machine model description to stack trace.

* 'parisc-4.19-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Add hardware description to stack traces
  parisc: Fix boot failure of 64-bit kernel
  parisc: Consolidate unwind initialization calls
  parisc: Update comments in syscall.S regarding wide userland
  parisc: Fix ptraced 64-bit applications to call 64-bit syscalls
  parisc: Restore possibility to execute 64-bit applications

arch/parisc/include/asm/elf.h

@@ -235,6 +235,7 @@ typedef unsigned long elf_greg_t;
 #define SET_PERSONALITY(ex) \
 ({	\
 	set_personality((current->personality & ~PER_MASK) | PER_LINUX); \
+	clear_thread_flag(TIF_32BIT); \
 	current->thread.map_base = DEFAULT_MAP_BASE; \
 	current->thread.task_size = DEFAULT_TASK_SIZE; \
  })
@@ -243,9 +244,11 @@ typedef unsigned long elf_greg_t;
 
 #define COMPAT_SET_PERSONALITY(ex) \
 ({	\
-	set_thread_flag(TIF_32BIT); \
-	current->thread.map_base = DEFAULT_MAP_BASE32; \
-	current->thread.task_size = DEFAULT_TASK_SIZE32; \
+	if ((ex).e_ident[EI_CLASS] == ELFCLASS32) { \
+		set_thread_flag(TIF_32BIT); \
+		current->thread.map_base = DEFAULT_MAP_BASE32; \
+		current->thread.task_size = DEFAULT_TASK_SIZE32; \
+	} else clear_thread_flag(TIF_32BIT); \
  })
 
 /*

arch/parisc/include/asm/linkage.h

@@ -22,15 +22,6 @@
 name:		ASM_NL\
 	.export name
 
-#ifdef CONFIG_64BIT
-#define ENDPROC(name) \
-	END(name)
-#else
-#define ENDPROC(name) \
-	.type name, @function !\
-	END(name)
-#endif
-
 #define ENTRY_CFI(name, ...) \
 	ENTRY(name)	ASM_NL\
 	.proc		ASM_NL\

arch/parisc/include/asm/processor.h

@@ -256,11 +256,7 @@ on downward growing arches, it looks like this:
  * it in here from the current->personality
  */
 
-#ifdef CONFIG_64BIT
-#define USER_WIDE_MODE	(!test_thread_flag(TIF_32BIT))
-#else
-#define USER_WIDE_MODE	0
-#endif
+#define USER_WIDE_MODE	(!is_32bit_task())
 
 #define start_thread(regs, new_pc, new_sp) do {		\
 	elf_addr_t *sp = (elf_addr_t *)new_sp;		\

arch/parisc/include/asm/traps.h

@@ -2,7 +2,9 @@
 #ifndef __ASM_TRAPS_H
 #define __ASM_TRAPS_H
 
-#ifdef __KERNEL__
+#define PARISC_ITLB_TRAP	6 /* defined by architecture. Do not change. */
+
+#if !defined(__ASSEMBLY__)
 struct pt_regs;
 
 /* traps.c */

arch/parisc/include/asm/unwind.h

@@ -73,8 +73,10 @@ unwind_table_remove(struct unwind_table *table);
 
 void unwind_frame_init(struct unwind_frame_info *info, struct task_struct *t, 
 		       struct pt_regs *regs);
-void unwind_frame_init_from_blocked_task(struct unwind_frame_info *info, struct task_struct *t);
-void unwind_frame_init_running(struct unwind_frame_info *info, struct pt_regs *regs);
+void unwind_frame_init_from_blocked_task(struct unwind_frame_info *info,
+			struct task_struct *t);
+void unwind_frame_init_task(struct unwind_frame_info *info,
+			struct task_struct *task, struct pt_regs *regs);
 int unwind_once(struct unwind_frame_info *info);
 int unwind_to_user(struct unwind_frame_info *info);
 

arch/parisc/kernel/entry.S

@@ -36,6 +36,7 @@
 #include <asm/signal.h>
 #include <asm/unistd.h>
 #include <asm/ldcw.h>
+#include <asm/traps.h>
 #include <asm/thread_info.h>
 
 #include <linux/linkage.h>
@@ -692,7 +693,7 @@ ENTRY(fault_vector_20)
 	def		 3
 	extint		 4
 	def		 5
-	itlb_20		 6
+	itlb_20		 PARISC_ITLB_TRAP
 	def		 7
 	def		 8
 	def              9
@@ -735,7 +736,7 @@ ENTRY(fault_vector_11)
 	def		 3
 	extint		 4
 	def		 5
-	itlb_11		 6
+	itlb_11		 PARISC_ITLB_TRAP
 	def		 7
 	def		 8
 	def              9
@@ -776,7 +777,7 @@ END(fault_vector_11)
 	 * copy_thread moved args into task save area.
 	 */
 
-ENTRY_CFI(ret_from_kernel_thread)
+ENTRY(ret_from_kernel_thread)
 	/* Call schedule_tail first though */
 	BL	schedule_tail, %r2
 	nop
@@ -791,7 +792,7 @@ ENTRY_CFI(ret_from_kernel_thread)
 	copy	%r31, %r2
 	b	finish_child_return
 	nop
-ENDPROC_CFI(ret_from_kernel_thread)
+END(ret_from_kernel_thread)
 
 
 	/*
@@ -815,9 +816,8 @@ ENTRY_CFI(_switch_to)
 	LDREG	TASK_THREAD_INFO(%r25), %r25
 	bv	%r0(%r2)
 	mtctl   %r25,%cr30
-ENDPROC_CFI(_switch_to)
 
-ENTRY_CFI(_switch_to_ret)
+ENTRY(_switch_to_ret)
 	mtctl	%r0, %cr0		/* Needed for single stepping */
 	callee_rest
 	callee_rest_float
@@ -825,7 +825,7 @@ ENTRY_CFI(_switch_to_ret)
 	LDREG	-RP_OFFSET(%r30), %r2
 	bv	%r0(%r2)
 	copy	%r26, %r28
-ENDPROC_CFI(_switch_to_ret)
+ENDPROC_CFI(_switch_to)
 
 	/*
 	 * Common rfi return path for interruptions, kernel execve, and
@@ -886,14 +886,12 @@ ENTRY_CFI(syscall_exit_rfi)
 	STREG   %r19,PT_SR5(%r16)
 	STREG   %r19,PT_SR6(%r16)
 	STREG   %r19,PT_SR7(%r16)
-ENDPROC_CFI(syscall_exit_rfi)
 
-ENTRY_CFI(intr_return)
+ENTRY(intr_return)
 	/* check for reschedule */
 	mfctl   %cr30,%r1
 	LDREG   TI_FLAGS(%r1),%r19	/* sched.h: TIF_NEED_RESCHED */
 	bb,<,n	%r19,31-TIF_NEED_RESCHED,intr_do_resched /* forward */
-ENDPROC_CFI(intr_return)
 
 	.import do_notify_resume,code
 intr_check_sig:
@@ -1049,6 +1047,7 @@ intr_extint:
 
 	b	do_cpu_irq_mask
 	ldo	R%intr_return(%r2), %r2	/* return to intr_return, not here */
+ENDPROC_CFI(syscall_exit_rfi)
 
 
 	/* Generic interruptions (illegal insn, unaligned, page fault, etc) */
@@ -1068,21 +1067,12 @@ ENTRY_CFI(intr_save)		/* for os_hpmc */
 	save_specials	%r29
 
 	/* If this trap is a itlb miss, skip saving/adjusting isr/ior */
-
-	/*
-	 * FIXME: 1) Use a #define for the hardwired "6" below (and in
-	 *           traps.c.
-	 *        2) Once we start executing code above 4 Gb, we need
-	 *           to adjust iasq/iaoq here in the same way we
-	 *           adjust isr/ior below.
-	 */
-
-	cmpib,COND(=),n        6,%r26,skip_save_ior
+	cmpib,COND(=),n        PARISC_ITLB_TRAP,%r26,skip_save_ior
 
 
-	mfctl           %cr20, %r16 /* isr */
+	mfctl           %isr, %r16
 	nop		/* serialize mfctl on PA 2.0 to avoid 4 cycle penalty */
-	mfctl           %cr21, %r17 /* ior */
+	mfctl           %ior, %r17
 
 
 #ifdef CONFIG_64BIT
@@ -1094,22 +1084,34 @@ ENTRY_CFI(intr_save)		/* for os_hpmc */
 	extrd,u,*<>     %r8,PSW_W_BIT,1,%r0
 	depdi           0,1,2,%r17
 
-	/*
-	 * FIXME: This code has hardwired assumptions about the split
-	 *        between space bits and offset bits. This will change
-	 *        when we allow alternate page sizes.
-	 */
-
-	/* adjust isr/ior. */
-	extrd,u         %r16,63,SPACEID_SHIFT,%r1	/* get high bits from isr for ior */
-	depd            %r1,31,SPACEID_SHIFT,%r17	/* deposit them into ior */
-	depdi           0,63,SPACEID_SHIFT,%r16		/* clear them from isr */
+	/* adjust isr/ior: get high bits from isr and deposit in ior */
+	space_adjust	%r16,%r17,%r1
 #endif
 	STREG           %r16, PT_ISR(%r29)
 	STREG           %r17, PT_IOR(%r29)
 
+#if 0 && defined(CONFIG_64BIT)
+	/* Revisit when we have 64-bit code above 4Gb */
+	b,n		intr_save2
 
 skip_save_ior:
+	/* We have a itlb miss, and when executing code above 4 Gb on ILP64, we
+	 * need to adjust iasq/iaoq here in the same way we adjusted isr/ior
+	 * above.
+	 */
+	extrd,u,*	%r8,PSW_W_BIT,1,%r1
+	cmpib,COND(=),n	1,%r1,intr_save2
+	LDREG		PT_IASQ0(%r29), %r16
+	LDREG		PT_IAOQ0(%r29), %r17
+	/* adjust iasq/iaoq */
+	space_adjust	%r16,%r17,%r1
+	STREG           %r16, PT_IASQ0(%r29)
+	STREG           %r17, PT_IAOQ0(%r29)
+#else
+skip_save_ior:
+#endif
+
+intr_save2:
 	virt_map
 	save_general	%r29
 
@@ -1747,7 +1749,7 @@ fork_like fork
 fork_like vfork
 
 	/* Set the return value for the child */
-ENTRY_CFI(child_return)
+ENTRY(child_return)
 	BL	schedule_tail, %r2
 	nop
 finish_child_return:
@@ -1759,7 +1761,7 @@ finish_child_return:
 	reg_restore %r1
 	b	syscall_exit
 	copy	%r0,%r28
-ENDPROC_CFI(child_return)
+END(child_return)
 
 ENTRY_CFI(sys_rt_sigreturn_wrapper)
 	LDREG	TI_TASK-THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r26
@@ -1791,7 +1793,7 @@ ENTRY_CFI(sys_rt_sigreturn_wrapper)
 	LDREG	PT_GR28(%r1),%r28  /* reload original r28 for syscall_exit */
 ENDPROC_CFI(sys_rt_sigreturn_wrapper)
 
-ENTRY_CFI(syscall_exit)
+ENTRY(syscall_exit)
 	/* NOTE: Not all syscalls exit this way.  rt_sigreturn will exit
 	 * via syscall_exit_rfi if the signal was received while the process
 	 * was running.
@@ -1990,15 +1992,13 @@ syscall_do_resched:
 #else
 	nop
 #endif
-ENDPROC_CFI(syscall_exit)
+END(syscall_exit)
 
 
 #ifdef CONFIG_FUNCTION_TRACER
 
 	.import ftrace_function_trampoline,code
 	.align L1_CACHE_BYTES
-	.globl mcount
-	.type  mcount, @function
 ENTRY_CFI(mcount, caller)
 _mcount:
 	.export _mcount,data
@@ -2027,8 +2027,6 @@ ENDPROC_CFI(mcount)
 
 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
 	.align 8
-	.globl return_to_handler
-	.type  return_to_handler, @function
 ENTRY_CFI(return_to_handler, caller,frame=FRAME_SIZE)
 	.export parisc_return_to_handler,data
 parisc_return_to_handler:
@@ -2078,6 +2076,7 @@ ENDPROC_CFI(return_to_handler)
 /* void call_on_stack(unsigned long param1, void *func,
 		      unsigned long new_stack) */
 ENTRY_CFI(call_on_stack, FRAME=2*FRAME_SIZE,CALLS,SAVE_RP,SAVE_SP)
+ENTRY(_call_on_stack)
 	copy	%sp, %r1
 
 	/* Regarding the HPPA calling conventions for function pointers,

arch/parisc/kernel/processor.c

@@ -288,6 +288,8 @@ void __init collect_boot_cpu_data(void)
 		printk(KERN_INFO "model %s\n",
 			boot_cpu_data.pdc.sys_model_name);
 
+	dump_stack_set_arch_desc("%s", boot_cpu_data.pdc.sys_model_name);
+
 	boot_cpu_data.hversion =  boot_cpu_data.pdc.model.hversion;
 	boot_cpu_data.sversion =  boot_cpu_data.pdc.model.sversion;
 

arch/parisc/kernel/stacktrace.c

@@ -16,20 +16,7 @@ static void dump_trace(struct task_struct *task, struct stack_trace *trace)
 {
 	struct unwind_frame_info info;
 
-	/* initialize unwind info */
-	if (task == current) {
-		unsigned long sp;
-		struct pt_regs r;
-HERE:
-		asm volatile ("copy %%r30, %0" : "=r"(sp));
-		memset(&r, 0, sizeof(struct pt_regs));
-		r.iaoq[0] = (unsigned long)&&HERE;
-		r.gr[2] = (unsigned long)__builtin_return_address(0);
-		r.gr[30] = sp;
-		unwind_frame_init(&info, task, &r);
-	} else {
-		unwind_frame_init_from_blocked_task(&info, task);
-	}
+	unwind_frame_init_task(&info, task, NULL);
 
 	/* unwind stack and save entries in stack_trace struct */
 	trace->nr_entries = 0;

arch/parisc/kernel/sys_parisc.c

@@ -156,11 +156,6 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 	int do_color_align, last_mmap;
 	struct vm_unmapped_area_info info;
 
-#ifdef CONFIG_64BIT
-	/* This should only ever run for 32-bit processes.  */
-	BUG_ON(!test_thread_flag(TIF_32BIT));
-#endif
-
 	/* requested length too big for entire address space */
 	if (len > TASK_SIZE)
 		return -ENOMEM;

arch/parisc/kernel/syscall.S

@@ -108,12 +108,8 @@ linux_gateway_entry:
 	mtsp	%r0,%sr6			/* get kernel space into sr6 */
 
 #ifdef CONFIG_64BIT
-	/* for now we can *always* set the W bit on entry to the syscall
-	 * since we don't support wide userland processes.  We could
-	 * also save the current SM other than in r0 and restore it on
-	 * exit from the syscall, and also use that value to know
-	 * whether to do narrow or wide syscalls. -PB
-	 */
+	/* Store W bit on entry to the syscall in case it's a wide userland
+	 * process. */
 	ssm	PSW_SM_W, %r1
 	extrd,u	%r1,PSW_W_BIT,1,%r1
 	/* sp must be aligned on 4, so deposit the W bit setting into
@@ -227,8 +223,7 @@ linux_gateway_entry:
 	or,=	%r2,%r2,%r2
 	ldo	R%sys_call_table64(%r1), %r19
 #else
-	ldil	L%sys_call_table, %r1
-	ldo     R%sys_call_table(%r1), %r19
+	load32	sys_call_table, %r19
 #endif	
 	comiclr,>>	__NR_Linux_syscalls, %r20, %r0
 	b,n	.Lsyscall_nosys
@@ -331,8 +326,6 @@ tracesys_next:
 	 *  task->thread.regs.gr[20] above.
 	 */
 	copy	%ret0,%r20
-	ldil	L%sys_call_table,%r1
-	ldo     R%sys_call_table(%r1), %r19
 
 	ldo     -THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r1      /* get task ptr */
 	LDREG	TI_TASK(%r1), %r1
@@ -354,6 +347,23 @@ tracesys_next:
 	comiclr,>>	__NR_Linux_syscalls, %r20, %r0
 	b,n	.Ltracesys_nosys
 
+	/* Note!  We cannot use the syscall table that is mapped
+	nearby since the gateway page is mapped execute-only. */
+
+#ifdef CONFIG_64BIT
+	LDREG	TASK_PT_GR30(%r1), %r19		/* get users sp back */
+	extrd,u	%r19,63,1,%r2			/* W hidden in bottom bit */
+
+	ldil	L%sys_call_table, %r1
+	or,=	%r2,%r2,%r2
+	addil	L%(sys_call_table64-sys_call_table), %r1
+	ldo	R%sys_call_table(%r1), %r19
+	or,=	%r2,%r2,%r2
+	ldo	R%sys_call_table64(%r1), %r19
+#else
+	load32	sys_call_table, %r19
+#endif
+
 	LDREGX  %r20(%r19), %r19
 
 	/* If this is a sys_rt_sigreturn call, and the signal was received
@@ -464,16 +474,13 @@ tracesys_sigexit:
 lws_start:
 
 #ifdef CONFIG_64BIT
-	/* FIXME: If we are a 64-bit kernel just
-	 *        turn this on unconditionally.
-	 */
 	ssm	PSW_SM_W, %r1
 	extrd,u	%r1,PSW_W_BIT,1,%r1
 	/* sp must be aligned on 4, so deposit the W bit setting into
 	 * the bottom of sp temporarily */
 	or,ev	%r1,%r30,%r30
 
-	/* Clip LWS number to a 32-bit value always */
+	/* Clip LWS number to a 32-bit value for 32-bit processes */
 	depdi	0, 31, 32, %r20
 #endif	
 

arch/parisc/kernel/traps.c

@@ -45,7 +45,7 @@
 
 #include "../math-emu/math-emu.h"	/* for handle_fpe() */
 
-static void parisc_show_stack(struct task_struct *task, unsigned long *sp,
+static void parisc_show_stack(struct task_struct *task,
 	struct pt_regs *regs);
 
 static int printbinary(char *buf, unsigned long x, int nbits)
@@ -152,7 +152,7 @@ void show_regs(struct pt_regs *regs)
 		printk("%s IAOQ[1]: %pS\n", level, (void *) regs->iaoq[1]);
 		printk("%s RP(r2): %pS\n", level, (void *) regs->gr[2]);
 
-		parisc_show_stack(current, NULL, regs);
+		parisc_show_stack(current, regs);
 	}
 }
 
@@ -185,44 +185,19 @@ static void do_show_stack(struct unwind_frame_info *info)
 	printk(KERN_CRIT "\n");
 }
 
-static void parisc_show_stack(struct task_struct *task, unsigned long *sp,
+static void parisc_show_stack(struct task_struct *task,
 	struct pt_regs *regs)
 {
 	struct unwind_frame_info info;
-	struct task_struct *t;
 
-	t = task ? task : current;
-	if (regs) {
-		unwind_frame_init(&info, t, regs);
-		goto show_stack;
-	}
-
-	if (t == current) {
-		unsigned long sp;
-
-HERE:
-		asm volatile ("copy %%r30, %0" : "=r"(sp));
-		{
-			struct pt_regs r;
-
-			memset(&r, 0, sizeof(struct pt_regs));
-			r.iaoq[0] = (unsigned long)&&HERE;
-			r.gr[2] = (unsigned long)__builtin_return_address(0);
-			r.gr[30] = sp;
-
-			unwind_frame_init(&info, current, &r);
-		}
-	} else {
-		unwind_frame_init_from_blocked_task(&info, t);
-	}
+	unwind_frame_init_task(&info, task, regs);
 
-show_stack:
 	do_show_stack(&info);
 }
 
 void show_stack(struct task_struct *t, unsigned long *sp)
 {
-	return parisc_show_stack(t, sp, NULL);
+	parisc_show_stack(t, NULL);
 }
 
 int is_valid_bugaddr(unsigned long iaoq)
@@ -557,7 +532,7 @@ void notrace handle_interruption(int code, struct pt_regs *regs)
 		cpu_lpmc(5, regs);
 		return;
 
-	case  6:
+	case  PARISC_ITLB_TRAP:
 		/* Instruction TLB miss fault/Instruction page fault */
 		fault_address = regs->iaoq[0];
 		fault_space   = regs->iasq[0];

arch/parisc/kernel/unwind.c

@@ -209,6 +209,8 @@ static int unwind_special(struct unwind_frame_info *info, unsigned long pc, int
 	 * We have to use void * instead of a function pointer, because
 	 * function pointers aren't a pointer to the function on 64-bit.
 	 * Make them const so the compiler knows they live in .text
+	 * Note: We could use dereference_kernel_function_descriptor()
+	 * instead but we want to keep it simple here.
 	 */
 	extern void * const handle_interruption;
 	extern void * const ret_from_kernel_thread;
@@ -216,7 +218,7 @@ static int unwind_special(struct unwind_frame_info *info, unsigned long pc, int
 	extern void * const intr_return;
 	extern void * const _switch_to_ret;
 #ifdef CONFIG_IRQSTACKS
-	extern void * const call_on_stack;
+	extern void * const _call_on_stack;
 #endif /* CONFIG_IRQSTACKS */
 
 	if (pc == (unsigned long) &handle_interruption) {
@@ -251,7 +253,7 @@ static int unwind_special(struct unwind_frame_info *info, unsigned long pc, int
 	}
 
 #ifdef CONFIG_IRQSTACKS
-	if (pc == (unsigned long) &call_on_stack) {
+	if (pc == (unsigned long) &_call_on_stack) {
 		info->prev_sp = *(unsigned long *)(info->sp - FRAME_SIZE - REG_SZ);
 		info->prev_ip = *(unsigned long *)(info->sp - FRAME_SIZE - RP_OFFSET);
 		return 1;
@@ -403,9 +405,31 @@ void unwind_frame_init_from_blocked_task(struct unwind_frame_info *info, struct
 	kfree(r2);
 }
 
-void unwind_frame_init_running(struct unwind_frame_info *info, struct pt_regs *regs)
+#define get_parisc_stackpointer() ({ \
+	unsigned long sp; \
+	__asm__("copy %%r30, %0" : "=r"(sp)); \
+	(sp); \
+})
+
+void unwind_frame_init_task(struct unwind_frame_info *info,
+	struct task_struct *task, struct pt_regs *regs)
 {
-	unwind_frame_init(info, current, regs);
+	task = task ? task : current;
+
+	if (task == current) {
+		struct pt_regs r;
+
+		if (!regs) {
+			memset(&r, 0, sizeof(r));
+			r.iaoq[0] =  _THIS_IP_;
+			r.gr[2] = _RET_IP_;
+			r.gr[30] = get_parisc_stackpointer();
+			regs = &r;
+		}
+		unwind_frame_init(info, task, &r);
+	} else {
+		unwind_frame_init_from_blocked_task(info, task);
+	}
 }
 
 int unwind_once(struct unwind_frame_info *next_frame)
@@ -442,19 +466,12 @@ int unwind_to_user(struct unwind_frame_info *info)
 unsigned long return_address(unsigned int level)
 {
 	struct unwind_frame_info info;
-	struct pt_regs r;
-	unsigned long sp;
 
 	/* initialize unwind info */
-	asm volatile ("copy %%r30, %0" : "=r"(sp));
-	memset(&r, 0, sizeof(struct pt_regs));
-	r.iaoq[0] = _THIS_IP_;
-	r.gr[2] = _RET_IP_;
-	r.gr[30] = sp;
-	unwind_frame_init(&info, current, &r);
+	unwind_frame_init_task(&info, current, NULL);
 
 	/* unwind stack */
-	++level;
+	level += 2;
 	do {
 		if (unwind_once(&info) < 0 || info.ip == 0)
 			return 0;