parisc: Fix and improve kernel stack unwinding

This patchset fixes and improves stack unwinding a lot:
1. Show backward stack traces with up to 30 callsites
2. Add callinfo to ENTRY_CFI() such that every assembler function will get an
   entry in the unwind table
3. Use constants instead of numbers in call_on_stack()
4. Do not depend on CONFIG_KALLSYMS to generate backtraces.
5. Speed up backtrace generation

Make sure you have this patch to GNU as installed:
https://sourceware.org/ml/binutils/2018-07/msg00474.html
Without this patch, unwind info in the kernel is often wrong for various
functions.

Signed-off-by: Helge Deller <deller@gmx.de>

6 files changed, 72 insertions, 205 deletions

diff --git a/arch/parisc/kernel/entry.S b/arch/parisc/kernel/entry.S
index 1b4732e20137..c7508f5717fb 100644
--- a/arch/parisc/kernel/entry.S
+++ b/arch/parisc/kernel/entry.S
@@ -766,7 +766,6 @@ END(fault_vector_11)
 #endif
 	/* Fault vector is separately protected and *must* be on its own page */
 	.align		PAGE_SIZE
-ENTRY(end_fault_vector)
 
 	.import		handle_interruption,code
 	.import		do_cpu_irq_mask,code
@@ -778,7 +777,6 @@ ENTRY(end_fault_vector)
 	 */
 
 ENTRY_CFI(ret_from_kernel_thread)
-
 	/* Call schedule_tail first though */
 	BL	schedule_tail, %r2
 	nop
@@ -817,8 +815,9 @@ ENTRY_CFI(_switch_to)
 	LDREG	TASK_THREAD_INFO(%r25), %r25
 	bv	%r0(%r2)
 	mtctl   %r25,%cr30
+ENDPROC_CFI(_switch_to)
 
-_switch_to_ret:
+ENTRY_CFI(_switch_to_ret)
 	mtctl	%r0, %cr0		/* Needed for single stepping */
 	callee_rest
 	callee_rest_float
@@ -826,7 +825,7 @@ _switch_to_ret:
 	LDREG	-RP_OFFSET(%r30), %r2
 	bv	%r0(%r2)
 	copy	%r26, %r28
-ENDPROC_CFI(_switch_to)
+ENDPROC_CFI(_switch_to_ret)
 
 	/*
 	 * Common rfi return path for interruptions, kernel execve, and
@@ -887,12 +886,14 @@ 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)
 
-intr_return:
+ENTRY_CFI(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:
@@ -1048,7 +1049,6 @@ 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) */
@@ -1999,12 +1999,9 @@ ENDPROC_CFI(syscall_exit)
 	.align L1_CACHE_BYTES
 	.globl mcount
 	.type  mcount, @function
-ENTRY(mcount)
+ENTRY_CFI(mcount, caller)
 _mcount:
 	.export _mcount,data
-	.proc
-	.callinfo caller,frame=0
-	.entry
 	/*
 	 * The 64bit mcount() function pointer needs 4 dwords, of which the
 	 * first two are free.  We optimize it here and put 2 instructions for
@@ -2026,18 +2023,13 @@ ftrace_stub:
 	.dword mcount
 	.dword 0 /* code in head.S puts value of global gp here */
 #endif
-	.exit
-	.procend
-ENDPROC(mcount)
+ENDPROC_CFI(mcount)
 
 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
 	.align 8
 	.globl return_to_handler
 	.type  return_to_handler, @function
-ENTRY_CFI(return_to_handler)
-	.proc
-	.callinfo caller,frame=FRAME_SIZE
-	.entry
+ENTRY_CFI(return_to_handler, caller,frame=FRAME_SIZE)
 	.export parisc_return_to_handler,data
 parisc_return_to_handler:
 	copy %r3,%r1
@@ -2076,8 +2068,6 @@ parisc_return_to_handler:
 	bv	%r0(%rp)
 #endif
 	LDREGM -FRAME_SIZE(%sp),%r3
-	.exit
-	.procend
 ENDPROC_CFI(return_to_handler)
 
 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
@@ -2087,31 +2077,30 @@ ENDPROC_CFI(return_to_handler)
 #ifdef CONFIG_IRQSTACKS
 /* void call_on_stack(unsigned long param1, void *func,
 		      unsigned long new_stack) */
-ENTRY_CFI(call_on_stack)
+ENTRY_CFI(call_on_stack, FRAME=2*FRAME_SIZE,CALLS,SAVE_RP,SAVE_SP)
 	copy	%sp, %r1
 
 	/* Regarding the HPPA calling conventions for function pointers,
 	   we assume the PIC register is not changed across call.  For
 	   CONFIG_64BIT, the argument pointer is left to point at the
 	   argument region allocated for the call to call_on_stack. */
+
+	/* Switch to new stack.  We allocate two frames.  */
+	ldo	2*FRAME_SIZE(%arg2), %sp
 # ifdef CONFIG_64BIT
-	/* Switch to new stack.  We allocate two 128 byte frames.  */
-	ldo	256(%arg2), %sp
 	/* Save previous stack pointer and return pointer in frame marker */
-	STREG	%rp, -144(%sp)
+	STREG	%rp, -FRAME_SIZE-RP_OFFSET(%sp)
 	/* Calls always use function descriptor */
 	LDREG	16(%arg1), %arg1
 	bve,l	(%arg1), %rp
-	STREG	%r1, -136(%sp)
-	LDREG	-144(%sp), %rp
+	STREG	%r1, -FRAME_SIZE-REG_SZ(%sp)
+	LDREG	-FRAME_SIZE-RP_OFFSET(%sp), %rp
 	bve	(%rp)
-	LDREG	-136(%sp), %sp
+	LDREG	-FRAME_SIZE-REG_SZ(%sp), %sp
 # else
-	/* Switch to new stack.  We allocate two 64 byte frames.  */
-	ldo	128(%arg2), %sp
 	/* Save previous stack pointer and return pointer in frame marker */
-	STREG	%r1, -68(%sp)
-	STREG	%rp, -84(%sp)
+	STREG	%r1, -FRAME_SIZE-REG_SZ(%sp)
+	STREG	%rp, -FRAME_SIZE-RP_OFFSET(%sp)
 	/* Calls use function descriptor if PLABEL bit is set */
 	bb,>=,n	%arg1, 30, 1f
 	depwi	0,31,2, %arg1
@@ -2119,9 +2108,9 @@ ENTRY_CFI(call_on_stack)
 1:
 	be,l	0(%sr4,%arg1), %sr0, %r31
 	copy	%r31, %rp
-	LDREG	-84(%sp), %rp
+	LDREG	-FRAME_SIZE-RP_OFFSET(%sp), %rp
 	bv	(%rp)
-	LDREG	-68(%sp), %sp
+	LDREG	-FRAME_SIZE-REG_SZ(%sp), %sp
 # endif /* CONFIG_64BIT */
 ENDPROC_CFI(call_on_stack)
 #endif /* CONFIG_IRQSTACKS */
diff --git a/arch/parisc/kernel/pacache.S b/arch/parisc/kernel/pacache.S
index 97451e67d35b..f33bf2d306d6 100644
--- a/arch/parisc/kernel/pacache.S
+++ b/arch/parisc/kernel/pacache.S
@@ -44,10 +44,6 @@
 	.align	16
 
 ENTRY_CFI(flush_tlb_all_local)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	/*
 	 * The pitlbe and pdtlbe instructions should only be used to
 	 * flush the entire tlb. Also, there needs to be no intervening
@@ -189,18 +185,11 @@ fdtdone:
 
 2:      bv		%r0(%r2)
 	nop
-
-	.exit
-	.procend
 ENDPROC_CFI(flush_tlb_all_local)
 
 	.import cache_info,data
 
 ENTRY_CFI(flush_instruction_cache_local)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	load32		cache_info, %r1
 
 	/* Flush Instruction Cache */
@@ -256,18 +245,11 @@ fisync:
 	mtsm		%r22			/* restore I-bit */
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_instruction_cache_local)
 
 
 	.import cache_info, data
 ENTRY_CFI(flush_data_cache_local)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	load32		cache_info, %r1
 
 	/* Flush Data Cache */
@@ -324,9 +306,6 @@ fdsync:
 	mtsm		%r22			/* restore I-bit */
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_data_cache_local)
 
 /* Macros to serialize TLB purge operations on SMP.  */
@@ -362,10 +341,6 @@ ENDPROC_CFI(flush_data_cache_local)
 /* Clear page using kernel mapping.  */
 
 ENTRY_CFI(clear_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 #ifdef CONFIG_64BIT
 
 	/* Unroll the loop.  */
@@ -424,18 +399,11 @@ ENTRY_CFI(clear_page_asm)
 #endif
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(clear_page_asm)
 
 /* Copy page using kernel mapping.  */
 
 ENTRY_CFI(copy_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 #ifdef CONFIG_64BIT
 	/* PA8x00 CPUs can consume 2 loads or 1 store per cycle.
 	 * Unroll the loop by hand and arrange insn appropriately.
@@ -542,9 +510,6 @@ ENTRY_CFI(copy_page_asm)
 #endif
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(copy_page_asm)
 
 /*
@@ -598,10 +563,6 @@ ENDPROC_CFI(copy_page_asm)
 	 */
 
 ENTRY_CFI(copy_user_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	/* Convert virtual `to' and `from' addresses to physical addresses.
 	   Move `from' physical address to non shadowed register.  */
 	ldil		L%(__PAGE_OFFSET), %r1
@@ -750,16 +711,9 @@ ENTRY_CFI(copy_user_page_asm)
 
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(copy_user_page_asm)
 
 ENTRY_CFI(clear_user_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	tophys_r1	%r26
 
 	ldil		L%(TMPALIAS_MAP_START), %r28
@@ -836,16 +790,9 @@ ENTRY_CFI(clear_user_page_asm)
 
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(clear_user_page_asm)
 
 ENTRY_CFI(flush_dcache_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%(TMPALIAS_MAP_START), %r28
 #ifdef CONFIG_64BIT
 #if (TMPALIAS_MAP_START >= 0x80000000)
@@ -903,16 +850,9 @@ ENTRY_CFI(flush_dcache_page_asm)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_dcache_page_asm)
 
 ENTRY_CFI(flush_icache_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%(TMPALIAS_MAP_START), %r28
 #ifdef CONFIG_64BIT
 #if (TMPALIAS_MAP_START >= 0x80000000)
@@ -977,16 +917,9 @@ ENTRY_CFI(flush_icache_page_asm)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_icache_page_asm)
 
 ENTRY_CFI(flush_kernel_dcache_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%dcache_stride, %r1
 	ldw		R%dcache_stride(%r1), %r23
 
@@ -1020,16 +953,9 @@ ENTRY_CFI(flush_kernel_dcache_page_asm)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_kernel_dcache_page_asm)
 
 ENTRY_CFI(purge_kernel_dcache_page_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%dcache_stride, %r1
 	ldw		R%dcache_stride(%r1), %r23
 
@@ -1062,16 +988,9 @@ ENTRY_CFI(purge_kernel_dcache_page_asm)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(purge_kernel_dcache_page_asm)
 
 ENTRY_CFI(flush_user_dcache_range_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%dcache_stride, %r1
 	ldw		R%dcache_stride(%r1), %r23
 	ldo		-1(%r23), %r21
@@ -1083,16 +1002,9 @@ ENTRY_CFI(flush_user_dcache_range_asm)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_user_dcache_range_asm)
 
 ENTRY_CFI(flush_kernel_dcache_range_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%dcache_stride, %r1
 	ldw		R%dcache_stride(%r1), %r23
 	ldo		-1(%r23), %r21
@@ -1105,16 +1017,9 @@ ENTRY_CFI(flush_kernel_dcache_range_asm)
 	syncdma
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_kernel_dcache_range_asm)
 
 ENTRY_CFI(purge_kernel_dcache_range_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%dcache_stride, %r1
 	ldw		R%dcache_stride(%r1), %r23
 	ldo		-1(%r23), %r21
@@ -1127,16 +1032,9 @@ ENTRY_CFI(purge_kernel_dcache_range_asm)
 	syncdma
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(purge_kernel_dcache_range_asm)
 
 ENTRY_CFI(flush_user_icache_range_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%icache_stride, %r1
 	ldw		R%icache_stride(%r1), %r23
 	ldo		-1(%r23), %r21
@@ -1148,16 +1046,9 @@ ENTRY_CFI(flush_user_icache_range_asm)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_user_icache_range_asm)
 
 ENTRY_CFI(flush_kernel_icache_page)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%icache_stride, %r1
 	ldw		R%icache_stride(%r1), %r23
 
@@ -1191,16 +1082,9 @@ ENTRY_CFI(flush_kernel_icache_page)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(flush_kernel_icache_page)
 
 ENTRY_CFI(flush_kernel_icache_range_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	ldil		L%icache_stride, %r1
 	ldw		R%icache_stride(%r1), %r23
 	ldo		-1(%r23), %r21
@@ -1212,8 +1096,6 @@ ENTRY_CFI(flush_kernel_icache_range_asm)
 	sync
 	bv		%r0(%r2)
 	nop
-	.exit
-	.procend
 ENDPROC_CFI(flush_kernel_icache_range_asm)
 
 	__INIT
@@ -1223,10 +1105,6 @@ ENDPROC_CFI(flush_kernel_icache_range_asm)
 	 */
 	.align	256
 ENTRY_CFI(disable_sr_hashing_asm)
-	.proc
-	.callinfo NO_CALLS
-	.entry
-
 	/*
 	 * Switch to real mode
 	 */
@@ -1308,9 +1186,6 @@ srdis_done:
 
 2:      bv		%r0(%r2)
 	nop
-	.exit
-
-	.procend
 ENDPROC_CFI(disable_sr_hashing_asm)
 
 	.end
diff --git a/arch/parisc/kernel/process.c b/arch/parisc/kernel/process.c
index b931745815e0..eb39e7e380d7 100644
--- a/arch/parisc/kernel/process.c
+++ b/arch/parisc/kernel/process.c
@@ -302,7 +302,7 @@ get_wchan(struct task_struct *p)
 		ip = info.ip;
 		if (!in_sched_functions(ip))
 			return ip;
-	} while (count++ < 16);
+	} while (count++ < MAX_UNWIND_ENTRIES);
 	return 0;
 }
 
diff --git a/arch/parisc/kernel/real2.S b/arch/parisc/kernel/real2.S
index cc9963421a19..2b16d8d6598f 100644
--- a/arch/parisc/kernel/real2.S
+++ b/arch/parisc/kernel/real2.S
@@ -35,12 +35,6 @@ real32_stack:
 real64_stack:
 	.block	8192
 
-#ifdef CONFIG_64BIT
-#  define REG_SZ 8
-#else
-#  define REG_SZ 4
-#endif
-
 #define N_SAVED_REGS 9
 
 save_cr_space:
diff --git a/arch/parisc/kernel/traps.c b/arch/parisc/kernel/traps.c
index 4309ad31a874..318815212518 100644
--- a/arch/parisc/kernel/traps.c
+++ b/arch/parisc/kernel/traps.c
@@ -172,7 +172,7 @@ static void do_show_stack(struct unwind_frame_info *info)
 	int i = 1;
 
 	printk(KERN_CRIT "Backtrace:\n");
-	while (i <= 16) {
+	while (i <= MAX_UNWIND_ENTRIES) {
 		if (unwind_once(info) < 0 || info->ip == 0)
 			break;
 
diff --git a/arch/parisc/kernel/unwind.c b/arch/parisc/kernel/unwind.c
index a4b430f440a9..5cdf13069dd9 100644
--- a/arch/parisc/kernel/unwind.c
+++ b/arch/parisc/kernel/unwind.c
@@ -13,7 +13,6 @@
 #include <linux/init.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
-#include <linux/kallsyms.h>
 #include <linux/sort.h>
 
 #include <linux/uaccess.h>
@@ -117,7 +116,8 @@ unwind_table_init(struct unwind_table *table, const char *name,
 	for (; start <= end; start++) {
 		if (start < end && 
 		    start->region_end > (start+1)->region_start) {
-			printk("WARNING: Out of order unwind entry! %p and %p\n", start, start+1);
+			pr_warn("Out of order unwind entry! %px and %px\n",
+				start, start+1);
 		}
 
 		start->region_start += base_addr;
@@ -203,25 +203,60 @@ int __init unwind_init(void)
 	return 0;
 }
 
-#ifdef CONFIG_64BIT
-#define get_func_addr(fptr) fptr[2]
-#else
-#define get_func_addr(fptr) fptr[0]
-#endif
-
 static int unwind_special(struct unwind_frame_info *info, unsigned long pc, int frame_size)
 {
-	extern void handle_interruption(int, struct pt_regs *);
-	static unsigned long *hi = (unsigned long *)&handle_interruption;
-
-	if (pc == get_func_addr(hi)) {
+	/*
+	 * 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
+	 */
+	extern void * const handle_interruption;
+	extern void * const ret_from_kernel_thread;
+	extern void * const syscall_exit;
+	extern void * const intr_return;
+	extern void * const _switch_to_ret;
+#ifdef CONFIG_IRQSTACKS
+	extern void * const call_on_stack;
+#endif /* CONFIG_IRQSTACKS */
+
+	if (pc == (unsigned long) &handle_interruption) {
 		struct pt_regs *regs = (struct pt_regs *)(info->sp - frame_size - PT_SZ_ALGN);
 		dbg("Unwinding through handle_interruption()\n");
 		info->prev_sp = regs->gr[30];
 		info->prev_ip = regs->iaoq[0];
+		return 1;
+	}
+
+	if (pc == (unsigned long) &ret_from_kernel_thread ||
+	    pc == (unsigned long) &syscall_exit) {
+		info->prev_sp = info->prev_ip = 0;
+		return 1;
+	}
+
+	if (pc == (unsigned long) &intr_return) {
+		struct pt_regs *regs;
+
+		dbg("Found intr_return()\n");
+		regs = (struct pt_regs *)(info->sp - PT_SZ_ALGN);
+		info->prev_sp = regs->gr[30];
+		info->prev_ip = regs->iaoq[0];
+		info->rp = regs->gr[2];
+		return 1;
+	}
+
+	if (pc == (unsigned long) &_switch_to_ret) {
+		info->prev_sp = info->sp - CALLEE_SAVE_FRAME_SIZE;
+		info->prev_ip = *(unsigned long *)(info->prev_sp - RP_OFFSET);
+		return 1;
+	}
 
+#ifdef CONFIG_IRQSTACKS
+	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;
 	}
+#endif
 
 	return 0;
 }
@@ -238,34 +273,8 @@ static void unwind_frame_regs(struct unwind_frame_info *info)
 	if (e == NULL) {
 		unsigned long sp;
 
-		dbg("Cannot find unwind entry for 0x%lx; forced unwinding\n", info->ip);
-
-#ifdef CONFIG_KALLSYMS
-		/* Handle some frequent special cases.... */
-		{
-			char symname[KSYM_NAME_LEN];
-			char *modname;
-
-			kallsyms_lookup(info->ip, NULL, NULL, &modname,
-				symname);
-
-			dbg("info->ip = 0x%lx, name = %s\n", info->ip, symname);
-
-			if (strcmp(symname, "_switch_to_ret") == 0) {
-				info->prev_sp = info->sp - CALLEE_SAVE_FRAME_SIZE;
-				info->prev_ip = *(unsigned long *)(info->prev_sp - RP_OFFSET);
-				dbg("_switch_to_ret @ %lx - setting "
-				    "prev_sp=%lx prev_ip=%lx\n", 
-				    info->ip, info->prev_sp, 
-				    info->prev_ip);
-				return;
-			} else if (strcmp(symname, "ret_from_kernel_thread") == 0 ||
-				   strcmp(symname, "syscall_exit") == 0) {
-				info->prev_ip = info->prev_sp = 0;
-				return;
-			}
-		}
-#endif
+		dbg("Cannot find unwind entry for %pS; forced unwinding\n",
+			(void *) info->ip);
 
 		/* Since we are doing the unwinding blind, we don't know if
 		   we are adjusting the stack correctly or extracting the rp