Merge git://git.kernel.org:/pub/scm/linux/kernel/git/davem/sparc

Pull sparc updates from David Miller:
 "Just some more random bits from Al, including a conversion over to
  generic extables"

* git://git.kernel.org:/pub/scm/linux/kernel/git/davem/sparc:
  sparc32: take ->thread.flags out
  sparc32: get rid of fake_swapper_regs
  sparc64: get rid of fake_swapper_regs
  sparc32: switch to generic extables
  sparc32: switch copy_user.S away from range exception table entries
  sparc32: get rid of range exception table entries in checksum_32.S
  sparc32: switch __bzero() away from range exception table entries
  sparc32: kill lookup_fault()
  sparc32: don't bother with lookup_fault() in __bzero()

20 files changed, 198 insertions, 642 deletions

diff --git a/arch/sparc/include/asm/elf_64.h b/arch/sparc/include/asm/elf_64.h
index 7e078bc73ef5..8fb09eec8c3e 100644
--- a/arch/sparc/include/asm/elf_64.h
+++ b/arch/sparc/include/asm/elf_64.h
@@ -8,7 +8,6 @@
 
 #include <asm/ptrace.h>
 #include <asm/processor.h>
-#include <asm/extable_64.h>
 #include <asm/spitfire.h>
 #include <asm/adi.h>
 
diff --git a/arch/sparc/include/asm/extable_64.h b/arch/sparc/include/asm/extable.h
index 5a0171907b7e..554a9dc376fc 100644
--- a/arch/sparc/include/asm/extable_64.h
+++ b/arch/sparc/include/asm/extable.h
@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 */
-#ifndef __ASM_EXTABLE64_H
-#define __ASM_EXTABLE64_H
+#ifndef __ASM_EXTABLE_H
+#define __ASM_EXTABLE_H
 /*
  * The exception table consists of pairs of addresses: the first is the
  * address of an instruction that is allowed to fault, and the second is
diff --git a/arch/sparc/include/asm/processor_32.h b/arch/sparc/include/asm/processor_32.h
index 3c4bc2189092..b6242f7771e9 100644
--- a/arch/sparc/include/asm/processor_32.h
+++ b/arch/sparc/include/asm/processor_32.h
@@ -50,16 +50,12 @@ struct thread_struct {
 	unsigned long   fsr;
 	unsigned long   fpqdepth;
 	struct fpq	fpqueue[16];
-	unsigned long flags;
 	mm_segment_t current_ds;
 };
 
-#define SPARC_FLAG_KTHREAD      0x1    /* task is a kernel thread */
-#define SPARC_FLAG_UNALIGNED    0x2    /* is allowed to do unaligned accesses */
-
 #define INIT_THREAD  { \
-	.flags = SPARC_FLAG_KTHREAD, \
 	.current_ds = KERNEL_DS, \
+	.kregs = (struct pt_regs *)(init_stack+THREAD_SIZE)-1 \
 }
 
 /* Do necessary setup to start up a newly executed thread. */
diff --git a/arch/sparc/include/asm/thread_info_64.h b/arch/sparc/include/asm/thread_info_64.h
index 42cd4cd3892e..8047a9caab2f 100644
--- a/arch/sparc/include/asm/thread_info_64.h
+++ b/arch/sparc/include/asm/thread_info_64.h
@@ -118,6 +118,7 @@ struct thread_info {
 	.task		=	&tsk,			\
 	.current_ds	=	ASI_P,			\
 	.preempt_count	=	INIT_PREEMPT_COUNT,	\
+	.kregs		=	(struct pt_regs *)(init_stack+THREAD_SIZE)-1 \
 }
 
 /* how to get the thread information struct from C */
diff --git a/arch/sparc/include/asm/uaccess.h b/arch/sparc/include/asm/uaccess.h
index dd85bc2c2cad..390094200fc4 100644
--- a/arch/sparc/include/asm/uaccess.h
+++ b/arch/sparc/include/asm/uaccess.h
@@ -1,6 +1,9 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #ifndef ___ASM_SPARC_UACCESS_H
 #define ___ASM_SPARC_UACCESS_H
+
+#include <asm/extable.h>
+
 #if defined(__sparc__) && defined(__arch64__)
 #include <asm/uaccess_64.h>
 #else
diff --git a/arch/sparc/include/asm/uaccess_32.h b/arch/sparc/include/asm/uaccess_32.h
index 0a2d3ebc4bb8..4a12346bb69c 100644
--- a/arch/sparc/include/asm/uaccess_32.h
+++ b/arch/sparc/include/asm/uaccess_32.h
@@ -13,9 +13,6 @@
 
 #include <asm/processor.h>
 
-#define ARCH_HAS_SORT_EXTABLE
-#define ARCH_HAS_SEARCH_EXTABLE
-
 /* Sparc is not segmented, however we need to be able to fool access_ok()
  * when doing system calls from kernel mode legitimately.
  *
@@ -40,36 +37,6 @@
 #define __access_ok(addr, size) (__user_ok((addr) & get_fs().seg, (size)))
 #define access_ok(addr, size) __access_ok((unsigned long)(addr), size)
 
-/*
- * The exception table consists of pairs of addresses: the first is the
- * address of an instruction that is allowed to fault, and the second is
- * the address at which the program should continue.  No registers are
- * modified, so it is entirely up to the continuation code to figure out
- * what to do.
- *
- * All the routines below use bits of fixup code that are out of line
- * with the main instruction path.  This means when everything is well,
- * we don't even have to jump over them.  Further, they do not intrude
- * on our cache or tlb entries.
- *
- * There is a special way how to put a range of potentially faulting
- * insns (like twenty ldd/std's with now intervening other instructions)
- * You specify address of first in insn and 0 in fixup and in the next
- * exception_table_entry you specify last potentially faulting insn + 1
- * and in fixup the routine which should handle the fault.
- * That fixup code will get
- * (faulting_insn_address - first_insn_in_the_range_address)/4
- * in %g2 (ie. index of the faulting instruction in the range).
- */
-
-struct exception_table_entry
-{
-        unsigned long insn, fixup;
-};
-
-/* Returns 0 if exception not found and fixup otherwise.  */
-unsigned long search_extables_range(unsigned long addr, unsigned long *g2);
-
 /* Uh, these should become the main single-value transfer routines..
  * They automatically use the right size if we just have the right
  * pointer type..
@@ -252,12 +219,7 @@ static inline unsigned long __clear_user(void __user *addr, unsigned long size)
 	unsigned long ret;
 
 	__asm__ __volatile__ (
-		".section __ex_table,#alloc\n\t"
-		".align 4\n\t"
-		".word 1f,3\n\t"
-		".previous\n\t"
 		"mov %2, %%o1\n"
-		"1:\n\t"
 		"call __bzero\n\t"
 		" mov %1, %%o0\n\t"
 		"mov %%o0, %0\n"
diff --git a/arch/sparc/include/asm/uaccess_64.h b/arch/sparc/include/asm/uaccess_64.h
index 698cf69f74e9..30eb4c6414d1 100644
--- a/arch/sparc/include/asm/uaccess_64.h
+++ b/arch/sparc/include/asm/uaccess_64.h
@@ -10,7 +10,6 @@
 #include <linux/string.h>
 #include <asm/asi.h>
 #include <asm/spitfire.h>
-#include <asm/extable_64.h>
 
 #include <asm/processor.h>
 
diff --git a/arch/sparc/kernel/head_32.S b/arch/sparc/kernel/head_32.S
index be30c8d4cc73..6044b82b9767 100644
--- a/arch/sparc/kernel/head_32.S
+++ b/arch/sparc/kernel/head_32.S
@@ -515,7 +515,7 @@ continue_boot:
 
 		/* I want a kernel stack NOW! */
 		set	init_thread_union, %g1
-		set	(THREAD_SIZE - STACKFRAME_SZ), %g2
+		set	(THREAD_SIZE - STACKFRAME_SZ - TRACEREG_SZ), %g2
 		add	%g1, %g2, %sp
 		mov	0, %fp			/* And for good luck */
 
diff --git a/arch/sparc/kernel/head_64.S b/arch/sparc/kernel/head_64.S
index c5ff2472b3d9..72a5bdc833ea 100644
--- a/arch/sparc/kernel/head_64.S
+++ b/arch/sparc/kernel/head_64.S
@@ -706,7 +706,7 @@ tlb_fixup_done:
 	wr	%g0, ASI_P, %asi
 	mov	1, %g1
 	sllx	%g1, THREAD_SHIFT, %g1
-	sub	%g1, (STACKFRAME_SZ + STACK_BIAS), %g1
+	sub	%g1, (STACKFRAME_SZ + STACK_BIAS + TRACEREG_SZ), %g1
 	add	%g6, %g1, %sp
 
 	/* Set per-cpu pointer initially to zero, this makes
diff --git a/arch/sparc/kernel/process_32.c b/arch/sparc/kernel/process_32.c
index b91e88058e0c..3b9794978e5b 100644
--- a/arch/sparc/kernel/process_32.c
+++ b/arch/sparc/kernel/process_32.c
@@ -216,16 +216,6 @@ void flush_thread(void)
 		clear_thread_flag(TIF_USEDFPU);
 #endif
 	}
-
-	/* This task is no longer a kernel thread. */
-	if (current->thread.flags & SPARC_FLAG_KTHREAD) {
-		current->thread.flags &= ~SPARC_FLAG_KTHREAD;
-
-		/* We must fixup kregs as well. */
-		/* XXX This was not fixed for ti for a while, worked. Unused? */
-		current->thread.kregs = (struct pt_regs *)
-		    (task_stack_page(current) + (THREAD_SIZE - TRACEREG_SZ));
-	}
 }
 
 static inline struct sparc_stackf __user *
@@ -313,7 +303,6 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, unsigned long arg,
 		extern int nwindows;
 		unsigned long psr;
 		memset(new_stack, 0, STACKFRAME_SZ + TRACEREG_SZ);
-		p->thread.flags |= SPARC_FLAG_KTHREAD;
 		p->thread.current_ds = KERNEL_DS;
 		ti->kpc = (((unsigned long) ret_from_kernel_thread) - 0x8);
 		childregs->u_regs[UREG_G1] = sp; /* function */
@@ -325,7 +314,6 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, unsigned long arg,
 	}
 	memcpy(new_stack, (char *)regs - STACKFRAME_SZ, STACKFRAME_SZ + TRACEREG_SZ);
 	childregs->u_regs[UREG_FP] = sp;
-	p->thread.flags &= ~SPARC_FLAG_KTHREAD;
 	p->thread.current_ds = USER_DS;
 	ti->kpc = (((unsigned long) ret_from_fork) - 0x8);
 	ti->kpsr = current->thread.fork_kpsr | PSR_PIL;
diff --git a/arch/sparc/kernel/setup_32.c b/arch/sparc/kernel/setup_32.c
index eea43a1aef1b..c8e0dd99f370 100644
--- a/arch/sparc/kernel/setup_32.c
+++ b/arch/sparc/kernel/setup_32.c
@@ -266,7 +266,6 @@ static __init void leon_patch(void)
 }
 
 struct tt_entry *sparc_ttable;
-static struct pt_regs fake_swapper_regs;
 
 /* Called from head_32.S - before we have setup anything
  * in the kernel. Be very careful with what you do here.
@@ -363,8 +362,6 @@ void __init setup_arch(char **cmdline_p)
 		(*(linux_dbvec->teach_debugger))();
 	}
 
-	init_task.thread.kregs = &fake_swapper_regs;
-
 	/* Run-time patch instructions to match the cpu model */
 	per_cpu_patch();
 
diff --git a/arch/sparc/kernel/setup_64.c b/arch/sparc/kernel/setup_64.c
index d87244197d5c..48abee4eee29 100644
--- a/arch/sparc/kernel/setup_64.c
+++ b/arch/sparc/kernel/setup_64.c
@@ -165,8 +165,6 @@ extern int root_mountflags;
 
 char reboot_command[COMMAND_LINE_SIZE];
 
-static struct pt_regs fake_swapper_regs = { { 0, }, 0, 0, 0, 0 };
-
 static void __init per_cpu_patch(void)
 {
 	struct cpuid_patch_entry *p;
@@ -661,8 +659,6 @@ void __init setup_arch(char **cmdline_p)
 	rd_image_start = ram_flags & RAMDISK_IMAGE_START_MASK;
 #endif
 
-	task_thread_info(&init_task)->kregs = &fake_swapper_regs;
-
 #ifdef CONFIG_IP_PNP
 	if (!ic_set_manually) {
 		phandle chosen = prom_finddevice("/chosen");
diff --git a/arch/sparc/kernel/unaligned_32.c b/arch/sparc/kernel/unaligned_32.c
index 83db94c0b431..ef5c5207c9ff 100644
--- a/arch/sparc/kernel/unaligned_32.c
+++ b/arch/sparc/kernel/unaligned_32.c
@@ -16,6 +16,7 @@
 #include <linux/uaccess.h>
 #include <linux/smp.h>
 #include <linux/perf_event.h>
+#include <linux/extable.h>
 
 #include <asm/setup.h>
 
@@ -213,10 +214,10 @@ static inline int ok_for_kernel(unsigned int insn)
 
 static void kernel_mna_trap_fault(struct pt_regs *regs, unsigned int insn)
 {
-	unsigned long g2 = regs->u_regs [UREG_G2];
-	unsigned long fixup = search_extables_range(regs->pc, &g2);
+	const struct exception_table_entry *entry;
 
-	if (!fixup) {
+	entry = search_exception_tables(regs->pc);
+	if (!entry) {
 		unsigned long address = compute_effective_address(regs, insn);
         	if(address < PAGE_SIZE) {
                 	printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference in mna handler");
@@ -232,9 +233,8 @@ static void kernel_mna_trap_fault(struct pt_regs *regs, unsigned int insn)
 	        die_if_kernel("Oops", regs);
 		/* Not reached */
 	}
-	regs->pc = fixup;
+	regs->pc = entry->fixup;
 	regs->npc = regs->pc + 4;
-	regs->u_regs [UREG_G2] = g2;
 }
 
 asmlinkage void kernel_unaligned_trap(struct pt_regs *regs, unsigned int insn)
@@ -274,103 +274,9 @@ asmlinkage void kernel_unaligned_trap(struct pt_regs *regs, unsigned int insn)
 	}
 }
 
-static inline int ok_for_user(struct pt_regs *regs, unsigned int insn,
-			      enum direction dir)
-{
-	unsigned int reg;
-	int size = ((insn >> 19) & 3) == 3 ? 8 : 4;
-
-	if ((regs->pc | regs->npc) & 3)
-		return 0;
-
-	/* Must access_ok() in all the necessary places. */
-#define WINREG_ADDR(regnum) \
-	((void __user *)(((unsigned long *)regs->u_regs[UREG_FP])+(regnum)))
-
-	reg = (insn >> 25) & 0x1f;
-	if (reg >= 16) {
-		if (!access_ok(WINREG_ADDR(reg - 16), size))
-			return -EFAULT;
-	}
-	reg = (insn >> 14) & 0x1f;
-	if (reg >= 16) {
-		if (!access_ok(WINREG_ADDR(reg - 16), size))
-			return -EFAULT;
-	}
-	if (!(insn & 0x2000)) {
-		reg = (insn & 0x1f);
-		if (reg >= 16) {
-			if (!access_ok(WINREG_ADDR(reg - 16), size))
-				return -EFAULT;
-		}
-	}
-#undef WINREG_ADDR
-	return 0;
-}
-
-static void user_mna_trap_fault(struct pt_regs *regs, unsigned int insn)
+asmlinkage void user_unaligned_trap(struct pt_regs *regs, unsigned int insn)
 {
 	send_sig_fault(SIGBUS, BUS_ADRALN,
 		       (void __user *)safe_compute_effective_address(regs, insn),
 		       0, current);
 }
-
-asmlinkage void user_unaligned_trap(struct pt_regs *regs, unsigned int insn)
-{
-	enum direction dir;
-
-	if(!(current->thread.flags & SPARC_FLAG_UNALIGNED) ||
-	   (((insn >> 30) & 3) != 3))
-		goto kill_user;
-	dir = decode_direction(insn);
-	if(!ok_for_user(regs, insn, dir)) {
-		goto kill_user;
-	} else {
-		int err, size = decode_access_size(insn);
-		unsigned long addr;
-
-		if(floating_point_load_or_store_p(insn)) {
-			printk("User FPU load/store unaligned unsupported.\n");
-			goto kill_user;
-		}
-
-		addr = compute_effective_address(regs, insn);
-		perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, addr);
-		switch(dir) {
-		case load:
-			err = do_int_load(fetch_reg_addr(((insn>>25)&0x1f),
-							 regs),
-					  size, (unsigned long *) addr,
-					  decode_signedness(insn));
-			break;
-
-		case store:
-			err = do_int_store(((insn>>25)&0x1f), size,
-					   (unsigned long *) addr, regs);
-			break;
-
-		case both:
-			/*
-			 * This was supported in 2.4. However, we question
-			 * the value of SWAP instruction across word boundaries.
-			 */
-			printk("Unaligned SWAP unsupported.\n");
-			err = -EFAULT;
-			break;
-
-		default:
-			unaligned_panic("Impossible user unaligned trap.");
-			goto out;
-		}
-		if (err)
-			goto kill_user;
-		else
-			advance(regs);
-		goto out;
-	}
-
-kill_user:
-	user_mna_trap_fault(regs, insn);
-out:
-	;
-}
diff --git a/arch/sparc/lib/checksum_32.S b/arch/sparc/lib/checksum_32.S
index 7488d130faf7..781e39b3c009 100644
--- a/arch/sparc/lib/checksum_32.S
+++ b/arch/sparc/lib/checksum_32.S
@@ -155,13 +155,6 @@ cpout:	retl						! get outta here
         .text;                                  \
         .align  4
 
-#define EXT(start,end)				\
-        .section __ex_table,ALLOC;		\
-        .align  4;                              \
-        .word   start, 0, end, cc_fault;         \
-        .text;                                  \
-        .align  4
-
 	/* This aligned version executes typically in 8.5 superscalar cycles, this
 	 * is the best I can do.  I say 8.5 because the final add will pair with
 	 * the next ldd in the main unrolled loop.  Thus the pipe is always full.
@@ -169,20 +162,20 @@ cpout:	retl						! get outta here
 	 * please check the fixup code below as well.
 	 */
 #define CSUMCOPY_BIGCHUNK_ALIGNED(src, dst, sum, off, t0, t1, t2, t3, t4, t5, t6, t7)	\
-	ldd	[src + off + 0x00], t0;							\
-	ldd	[src + off + 0x08], t2;							\
+	EX(ldd	[src + off + 0x00], t0);						\
+	EX(ldd	[src + off + 0x08], t2);						\
 	addxcc	t0, sum, sum;								\
-	ldd	[src + off + 0x10], t4;							\
+	EX(ldd	[src + off + 0x10], t4);						\
 	addxcc	t1, sum, sum;								\
-	ldd	[src + off + 0x18], t6;							\
+	EX(ldd	[src + off + 0x18], t6);						\
 	addxcc	t2, sum, sum;								\
-	std	t0, [dst + off + 0x00];							\
+	EX(std	t0, [dst + off + 0x00]);						\
 	addxcc	t3, sum, sum;								\
-	std	t2, [dst + off + 0x08];							\
+	EX(std	t2, [dst + off + 0x08]);						\
 	addxcc	t4, sum, sum;								\
-	std	t4, [dst + off + 0x10];							\
+	EX(std	t4, [dst + off + 0x10]);						\
 	addxcc	t5, sum, sum;								\
-	std	t6, [dst + off + 0x18];							\
+	EX(std	t6, [dst + off + 0x18]);						\
 	addxcc	t6, sum, sum;								\
 	addxcc	t7, sum, sum;
 
@@ -191,39 +184,39 @@ cpout:	retl						! get outta here
 	 * Viking MXCC into streaming mode.  Ho hum...
 	 */
 #define CSUMCOPY_BIGCHUNK(src, dst, sum, off, t0, t1, t2, t3, t4, t5, t6, t7)	\
-	ldd	[src + off + 0x00], t0;						\
-	ldd	[src + off + 0x08], t2;						\
-	ldd	[src + off + 0x10], t4;						\
-	ldd	[src + off + 0x18], t6;						\
-	st	t0, [dst + off + 0x00];						\
+	EX(ldd	[src + off + 0x00], t0);					\
+	EX(ldd	[src + off + 0x08], t2);					\
+	EX(ldd	[src + off + 0x10], t4);					\
+	EX(ldd	[src + off + 0x18], t6);					\
+	EX(st	t0, [dst + off + 0x00]);					\
 	addxcc	t0, sum, sum;							\
-	st	t1, [dst + off + 0x04];						\
+	EX(st	t1, [dst + off + 0x04]);					\
 	addxcc	t1, sum, sum;							\
-	st	t2, [dst + off + 0x08];						\
+	EX(st	t2, [dst + off + 0x08]);					\
 	addxcc	t2, sum, sum;							\
-	st	t3, [dst + off + 0x0c];						\
+	EX(st	t3, [dst + off + 0x0c]);					\
 	addxcc	t3, sum, sum;							\
-	st	t4, [dst + off + 0x10];						\
+	EX(st	t4, [dst + off + 0x10]);					\
 	addxcc	t4, sum, sum;							\
-	st	t5, [dst + off + 0x14];						\
+	EX(st	t5, [dst + off + 0x14]);					\
 	addxcc	t5, sum, sum;							\
-	st	t6, [dst + off + 0x18];						\
+	EX(st	t6, [dst + off + 0x18]);					\
 	addxcc	t6, sum, sum;							\
-	st	t7, [dst + off + 0x1c];						\
+	EX(st	t7, [dst + off + 0x1c]);					\
 	addxcc	t7, sum, sum;
 
 	/* Yuck, 6 superscalar cycles... */
 #define CSUMCOPY_LASTCHUNK(src, dst, sum, off, t0, t1, t2, t3)	\
-	ldd	[src - off - 0x08], t0;				\
-	ldd	[src - off - 0x00], t2;				\
+	EX(ldd	[src - off - 0x08], t0);			\
+	EX(ldd	[src - off - 0x00], t2);			\
 	addxcc	t0, sum, sum;					\
-	st	t0, [dst - off - 0x08];				\
+	EX(st	t0, [dst - off - 0x08]);			\
 	addxcc	t1, sum, sum;					\
-	st	t1, [dst - off - 0x04];				\
+	EX(st	t1, [dst - off - 0x04]);			\
 	addxcc	t2, sum, sum;					\
-	st	t2, [dst - off - 0x00];				\
+	EX(st	t2, [dst - off - 0x00]);			\
 	addxcc	t3, sum, sum;					\
-	st	t3, [dst - off + 0x04];
+	EX(st	t3, [dst - off + 0x04]);
 
 	/* Handle the end cruft code out of band for better cache patterns. */
 cc_end_cruft:
@@ -331,7 +324,6 @@ __csum_partial_copy_sparc_generic:
 	CSUMCOPY_BIGCHUNK(%o0,%o1,%g7,0x20,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3)
 	CSUMCOPY_BIGCHUNK(%o0,%o1,%g7,0x40,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3)
 	CSUMCOPY_BIGCHUNK(%o0,%o1,%g7,0x60,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3)
-10:	EXT(5b, 10b)			! note for exception handling
 	sub	%g1, 128, %g1		! detract from length
 	addx	%g0, %g7, %g7		! add in last carry bit
 	andcc	%g1, 0xffffff80, %g0	! more to csum?
@@ -356,8 +348,7 @@ cctbl:	CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x68,%g2,%g3,%g4,%g5)
 	CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x28,%g2,%g3,%g4,%g5)
 	CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x18,%g2,%g3,%g4,%g5)
 	CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x08,%g2,%g3,%g4,%g5)
-12:	EXT(cctbl, 12b)			! note for exception table handling
-	addx	%g0, %g7, %g7
+12:	addx	%g0, %g7, %g7
 	andcc	%o3, 0xf, %g0		! check for low bits set
 ccte:	bne	cc_end_cruft		! something left, handle it out of band
 	 andcc	%o3, 8, %g0		! begin checks for that code
@@ -367,7 +358,6 @@ ccdbl:	CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x00,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o
 	CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x20,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3)
 	CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x40,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3)
 	CSUMCOPY_BIGCHUNK_ALIGNED(%o0,%o1,%g7,0x60,%o4,%o5,%g2,%g3,%g4,%g5,%o2,%o3)
-11:	EXT(ccdbl, 11b)			! note for exception table handling
 	sub	%g1, 128, %g1		! detract from length
 	addx	%g0, %g7, %g7		! add in last carry bit
 	andcc	%g1, 0xffffff80, %g0	! more to csum?
diff --git a/arch/sparc/lib/copy_user.S b/arch/sparc/lib/copy_user.S
index dc72f2b970b7..954572c78539 100644
--- a/arch/sparc/lib/copy_user.S
+++ b/arch/sparc/lib/copy_user.S
@@ -21,98 +21,134 @@
 /* Work around cpp -rob */
 #define ALLOC #alloc
 #define EXECINSTR #execinstr
+
+#define EX_ENTRY(l1, l2)			\
+	.section __ex_table,ALLOC;		\
+	.align	4;				\
+	.word	l1, l2;				\
+	.text;
+
 #define EX(x,y,a,b) 				\
 98: 	x,y;					\
 	.section .fixup,ALLOC,EXECINSTR;	\
 	.align	4;				\
-99:	ba fixupretl;				\
-	 a, b, %g3;				\
-	.section __ex_table,ALLOC;		\
-	.align	4;				\
-	.word	98b, 99b;			\
-	.text;					\
-	.align	4
+99:	retl;					\
+	 a, b, %o0;				\
+	EX_ENTRY(98b, 99b)
 
 #define EX2(x,y,c,d,e,a,b) 			\
 98: 	x,y;					\
 	.section .fixup,ALLOC,EXECINSTR;	\
 	.align	4;				\
 99:	c, d, e;				\
-	ba fixupretl;				\
-	 a, b, %g3;				\
-	.section __ex_table,ALLOC;		\
-	.align	4;				\
-	.word	98b, 99b;			\
-	.text;					\
-	.align	4
+	retl;					\
+	 a, b, %o0;				\
+	EX_ENTRY(98b, 99b)
 
 #define EXO2(x,y) 				\
 98: 	x, y;					\
-	.section __ex_table,ALLOC;		\
-	.align	4;				\
-	.word	98b, 97f;			\
-	.text;					\
-	.align	4
+	EX_ENTRY(98b, 97f)
 
-#define EXT(start,end,handler)			\
-	.section __ex_table,ALLOC;		\
-	.align	4;				\
-	.word	start, 0, end, handler;		\
-	.text;					\
-	.align	4
+#define LD(insn, src, offset, reg, label)	\
+98:	insn [%src + (offset)], %reg;		\
+	.section .fixup,ALLOC,EXECINSTR;	\
+99:	ba	label;				\
+	 mov	offset, %g5;			\
+	EX_ENTRY(98b, 99b)
 
-/* Please do not change following macros unless you change logic used
- * in .fixup at the end of this file as well
- */
+#define ST(insn, dst, offset, reg, label)	\
+98:	insn %reg, [%dst + (offset)];		\
+	.section .fixup,ALLOC,EXECINSTR;	\
+99:	ba	label;				\
+	 mov	offset, %g5;			\
+	EX_ENTRY(98b, 99b)
 
 /* Both these macros have to start with exactly the same insn */
+/* left: g7 + (g1 % 128) - offset */
 #define MOVE_BIGCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \
-	ldd	[%src + (offset) + 0x00], %t0; \
-	ldd	[%src + (offset) + 0x08], %t2; \
-	ldd	[%src + (offset) + 0x10], %t4; \
-	ldd	[%src + (offset) + 0x18], %t6; \
-	st	%t0, [%dst + (offset) + 0x00]; \
-	st	%t1, [%dst + (offset) + 0x04]; \
-	st	%t2, [%dst + (offset) + 0x08]; \
-	st	%t3, [%dst + (offset) + 0x0c]; \
-	st	%t4, [%dst + (offset) + 0x10]; \
-	st	%t5, [%dst + (offset) + 0x14]; \
-	st	%t6, [%dst + (offset) + 0x18]; \
-	st	%t7, [%dst + (offset) + 0x1c];
-
+	LD(ldd, src, offset + 0x00, t0, bigchunk_fault)	\
+	LD(ldd, src, offset + 0x08, t2, bigchunk_fault)	\
+	LD(ldd, src, offset + 0x10, t4, bigchunk_fault)	\
+	LD(ldd, src, offset + 0x18, t6, bigchunk_fault)	\
+	ST(st, dst, offset + 0x00, t0, bigchunk_fault)	\
+	ST(st, dst, offset + 0x04, t1, bigchunk_fault)	\
+	ST(st, dst, offset + 0x08, t2, bigchunk_fault)	\
+	ST(st, dst, offset + 0x0c, t3, bigchunk_fault)	\
+	ST(st, dst, offset + 0x10, t4, bigchunk_fault)	\
+	ST(st, dst, offset + 0x14, t5, bigchunk_fault)	\
+	ST(st, dst, offset + 0x18, t6, bigchunk_fault)	\
+	ST(st, dst, offset + 0x1c, t7, bigchunk_fault)
+
+/* left: g7 + (g1 % 128) - offset */
 #define MOVE_BIGALIGNCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \
-	ldd	[%src + (offset) + 0x00], %t0; \
-	ldd	[%src + (offset) + 0x08], %t2; \
-	ldd	[%src + (offset) + 0x10], %t4; \
-	ldd	[%src + (offset) + 0x18], %t6; \
-	std	%t0, [%dst + (offset) + 0x00]; \
-	std	%t2, [%dst + (offset) + 0x08]; \
-	std	%t4, [%dst + (offset) + 0x10]; \
-	std	%t6, [%dst + (offset) + 0x18];
+	LD(ldd, src, offset + 0x00, t0, bigchunk_fault)	\
+	LD(ldd, src, offset + 0x08, t2, bigchunk_fault)	\
+	LD(ldd, src, offset + 0x10, t4, bigchunk_fault)	\
+	LD(ldd, src, offset + 0x18, t6, bigchunk_fault)	\
+	ST(std, dst, offset + 0x00, t0, bigchunk_fault)	\
+	ST(std, dst, offset + 0x08, t2, bigchunk_fault)	\
+	ST(std, dst, offset + 0x10, t4, bigchunk_fault)	\
+	ST(std, dst, offset + 0x18, t6, bigchunk_fault)
 
+	.section .fixup,#alloc,#execinstr
+bigchunk_fault:
+	sub	%g7, %g5, %o0
+	and	%g1, 127, %g1
+	retl
+	 add	%o0, %g1, %o0
+
+/* left: offset + 16 + (g1 % 16) */
 #define MOVE_LASTCHUNK(src, dst, offset, t0, t1, t2, t3) \
-	ldd	[%src - (offset) - 0x10], %t0; \
-	ldd	[%src - (offset) - 0x08], %t2; \
-	st	%t0, [%dst - (offset) - 0x10]; \
-	st	%t1, [%dst - (offset) - 0x0c]; \
-	st	%t2, [%dst - (offset) - 0x08]; \
-	st	%t3, [%dst - (offset) - 0x04];
+	LD(ldd, src, -(offset + 0x10), t0, lastchunk_fault)	\
+	LD(ldd, src, -(offset + 0x08), t2, lastchunk_fault)	\
+	ST(st, dst, -(offset + 0x10), t0, lastchunk_fault)	\
+	ST(st, dst, -(offset + 0x0c), t1, lastchunk_fault)	\
+	ST(st, dst, -(offset + 0x08), t2, lastchunk_fault)	\
+	ST(st, dst, -(offset + 0x04), t3, lastchunk_fault)
 
-#define MOVE_HALFCHUNK(src, dst, offset, t0, t1, t2, t3) \
-	lduh	[%src + (offset) + 0x00], %t0; \
-	lduh	[%src + (offset) + 0x02], %t1; \
-	lduh	[%src + (offset) + 0x04], %t2; \
-	lduh	[%src + (offset) + 0x06], %t3; \
-	sth	%t0, [%dst + (offset) + 0x00]; \
-	sth	%t1, [%dst + (offset) + 0x02]; \
-	sth	%t2, [%dst + (offset) + 0x04]; \
-	sth	%t3, [%dst + (offset) + 0x06];
+	.section .fixup,#alloc,#execinstr
+lastchunk_fault:
+	and	%g1, 15, %g1
+	retl
+	 sub	%g1, %g5, %o0
 
+/* left: o3 + (o2 % 16) - offset */
+#define MOVE_HALFCHUNK(src, dst, offset, t0, t1, t2, t3) \
+	LD(lduh, src, offset + 0x00, t0, halfchunk_fault)	\
+	LD(lduh, src, offset + 0x02, t1, halfchunk_fault)	\
+	LD(lduh, src, offset + 0x04, t2, halfchunk_fault)	\
+	LD(lduh, src, offset + 0x06, t3, halfchunk_fault)	\
+	ST(sth, dst, offset + 0x00, t0, halfchunk_fault)	\
+	ST(sth, dst, offset + 0x02, t1, halfchunk_fault)	\
+	ST(sth, dst, offset + 0x04, t2, halfchunk_fault)	\
+	ST(sth, dst, offset + 0x06, t3, halfchunk_fault)
+
+/* left: o3 + (o2 % 16) + offset + 2 */
 #define MOVE_SHORTCHUNK(src, dst, offset, t0, t1) \
-	ldub	[%src - (offset) - 0x02], %t0; \
-	ldub	[%src - (offset) - 0x01], %t1; \
-	stb	%t0, [%dst - (offset) - 0x02]; \
-	stb	%t1, [%dst - (offset) - 0x01];
+	LD(ldub, src, -(offset + 0x02), t0, halfchunk_fault)	\
+	LD(ldub, src, -(offset + 0x01), t1, halfchunk_fault)	\
+	ST(stb, dst, -(offset + 0x02), t0, halfchunk_fault)	\
+	ST(stb, dst, -(offset + 0x01), t1, halfchunk_fault)
+
+	.section .fixup,#alloc,#execinstr
+halfchunk_fault:
+	and	%o2, 15, %o2
+	sub	%o3, %g5, %o3
+	retl
+	 add	%o2, %o3, %o0
+
+/* left: offset + 2 + (o2 % 2) */
+#define MOVE_LAST_SHORTCHUNK(src, dst, offset, t0, t1) \
+	LD(ldub, src, -(offset + 0x02), t0, last_shortchunk_fault)	\
+	LD(ldub, src, -(offset + 0x01), t1, last_shortchunk_fault)	\
+	ST(stb, dst, -(offset + 0x02), t0, last_shortchunk_fault)	\
+	ST(stb, dst, -(offset + 0x01), t1, last_shortchunk_fault)
+
+	.section .fixup,#alloc,#execinstr
+last_shortchunk_fault:
+	and	%o2, 1, %o2
+	retl
+	 sub	%o2, %g5, %o0
 
 	.text
 	.align	4
@@ -182,8 +218,6 @@ __copy_user:	/* %o0=dst %o1=src %o2=len */
 	MOVE_BIGCHUNK(o1, o0, 0x20, o2, o3, o4, o5, g2, g3, g4, g5)
 	MOVE_BIGCHUNK(o1, o0, 0x40, o2, o3, o4, o5, g2, g3, g4, g5)
 	MOVE_BIGCHUNK(o1, o0, 0x60, o2, o3, o4, o5, g2, g3, g4, g5)
-80:
-	EXT(5b, 80b, 50f)
 	subcc	%g7, 128, %g7
 	add	%o1, 128, %o1
 	bne	5b
@@ -201,7 +235,6 @@ __copy_user:	/* %o0=dst %o1=src %o2=len */
 	jmpl	%o5 + %lo(copy_user_table_end), %g0
 	 add	%o0, %g7, %o0
 
-copy_user_table:
 	MOVE_LASTCHUNK(o1, o0, 0x60, g2, g3, g4, g5)
 	MOVE_LASTCHUNK(o1, o0, 0x50, g2, g3, g4, g5)
 	MOVE_LASTCHUNK(o1, o0, 0x40, g2, g3, g4, g5)
@@ -210,7 +243,6 @@ copy_user_table:
 	MOVE_LASTCHUNK(o1, o0, 0x10, g2, g3, g4, g5)
 	MOVE_LASTCHUNK(o1, o0, 0x00, g2, g3, g4, g5)
 copy_user_table_end:
-	EXT(copy_user_table, copy_user_table_end, 51f)
 	be	copy_user_last7
 	 andcc	%g1, 4, %g0
 
@@ -250,8 +282,6 @@ ldd_std:
 	MOVE_BIGALIGNCHUNK(o1, o0, 0x20, o2, o3, o4, o5, g2, g3, g4, g5)
 	MOVE_BIGALIGNCHUNK(o1, o0, 0x40, o2, o3, o4, o5, g2, g3, g4, g5)
 	MOVE_BIGALIGNCHUNK(o1, o0, 0x60, o2, o3, o4, o5, g2, g3, g4, g5)
-81:
-	EXT(ldd_std, 81b, 52f)
 	subcc	%g7, 128, %g7
 	add	%o1, 128, %o1
 	bne	ldd_std
@@ -290,8 +320,6 @@ cannot_optimize:
 10:
 	MOVE_HALFCHUNK(o1, o0, 0x00, g2, g3, g4, g5)
 	MOVE_HALFCHUNK(o1, o0, 0x08, g2, g3, g4, g5)
-82:
-	EXT(10b, 82b, 53f)
 	subcc	%o3, 0x10, %o3
 	add	%o1, 0x10, %o1
 	bne	10b
@@ -308,8 +336,6 @@ byte_chunk:
 	MOVE_SHORTCHUNK(o1, o0, -0x0c, g2, g3)
 	MOVE_SHORTCHUNK(o1, o0, -0x0e, g2, g3)
 	MOVE_SHORTCHUNK(o1, o0, -0x10, g2, g3)
-83:
-	EXT(byte_chunk, 83b, 54f)
 	subcc	%o3, 0x10, %o3
 	add	%o1, 0x10, %o1
 	bne	byte_chunk
@@ -325,16 +351,14 @@ short_end:
 	add	%o1, %o3, %o1
 	jmpl	%o5 + %lo(short_table_end), %g0
 	 andcc	%o2, 1, %g0
-84:
-	MOVE_SHORTCHUNK(o1, o0, 0x0c, g2, g3)
-	MOVE_SHORTCHUNK(o1, o0, 0x0a, g2, g3)
-	MOVE_SHORTCHUNK(o1, o0, 0x08, g2, g3)
-	MOVE_SHORTCHUNK(o1, o0, 0x06, g2, g3)
-	MOVE_SHORTCHUNK(o1, o0, 0x04, g2, g3)
-	MOVE_SHORTCHUNK(o1, o0, 0x02, g2, g3)
-	MOVE_SHORTCHUNK(o1, o0, 0x00, g2, g3)
+	MOVE_LAST_SHORTCHUNK(o1, o0, 0x0c, g2, g3)
+	MOVE_LAST_SHORTCHUNK(o1, o0, 0x0a, g2, g3)
+	MOVE_LAST_SHORTCHUNK(o1, o0, 0x08, g2, g3)
+	MOVE_LAST_SHORTCHUNK(o1, o0, 0x06, g2, g3)
+	MOVE_LAST_SHORTCHUNK(o1, o0, 0x04, g2, g3)
+	MOVE_LAST_SHORTCHUNK(o1, o0, 0x02, g2, g3)
+	MOVE_LAST_SHORTCHUNK(o1, o0, 0x00, g2, g3)
 short_table_end:
-	EXT(84b, short_table_end, 55f)
 	be	1f
 	 nop
 	EX(ldub	[%o1], %g2, add %g0, 1)
@@ -363,123 +387,8 @@ short_aligned_end:
 	.section .fixup,#alloc,#execinstr
 	.align	4
 97:
-	mov	%o2, %g3
-fixupretl:
 	retl
-	 mov	%g3, %o0
-
-/* exception routine sets %g2 to (broken_insn - first_insn)>>2 */
-50:
-/* This magic counts how many bytes are left when crash in MOVE_BIGCHUNK
- * happens. This is derived from the amount ldd reads, st stores, etc.
- * x = g2 % 12;
- * g3 = g1 + g7 - ((g2 / 12) * 32 + (x < 4) ? 0 : (x - 4) * 4);
- * o0 += (g2 / 12) * 32;
- */
-	cmp	%g2, 12
-	add	%o0, %g7, %o0
-	bcs	1f
-	 cmp	%g2, 24
-	bcs	2f
-	 cmp	%g2, 36
-	bcs	3f
-	 nop
-	sub	%g2, 12, %g2
-	sub	%g7, 32, %g7
-3:	sub	%g2, 12, %g2
-	sub	%g7, 32, %g7
-2:	sub	%g2, 12, %g2
-	sub	%g7, 32, %g7
-1:	cmp	%g2, 4
-	bcs,a	60f
-	 clr	%g2
-	sub	%g2, 4, %g2
-	sll	%g2, 2, %g2
-60:	and	%g1, 0x7f, %g3
-	sub	%o0, %g7, %o0
-	add	%g3, %g7, %g3
-	ba	fixupretl
-	 sub	%g3, %g2, %g3
-51:
-/* i = 41 - g2; j = i % 6;
- * g3 = (g1 & 15) + (i / 6) * 16 + (j < 4) ? (j + 1) * 4 : 16;
- * o0 -= (i / 6) * 16 + 16;
- */
-	neg	%g2
-	and	%g1, 0xf, %g1
-	add	%g2, 41, %g2
-	add	%o0, %g1, %o0
-1:	cmp	%g2, 6
-	bcs,a	2f
-	 cmp	%g2, 4
-	add	%g1, 16, %g1
-	b	1b
-	 sub	%g2, 6, %g2
-2:	bcc,a	2f
-	 mov	16, %g2
-	inc	%g2
-	sll	%g2, 2, %g2
-2:	add	%g1, %g2, %g3
-	ba	fixupretl
-	 sub	%o0, %g3, %o0
-52:
-/* g3 = g1 + g7 - (g2 / 8) * 32 + (g2 & 4) ? (g2 & 3) * 8 : 0;
-   o0 += (g2 / 8) * 32 */
-	andn	%g2, 7, %g4
-	add	%o0, %g7, %o0
-	andcc	%g2, 4, %g0
-	and	%g2, 3, %g2
-	sll	%g4, 2, %g4
-	sll	%g2, 3, %g2
-	bne	60b
-	 sub	%g7, %g4, %g7
-	ba	60b
-	 clr	%g2
-53:
-/* g3 = o3 + (o2 & 15) - (g2 & 8) - (g2 & 4) ? (g2 & 3) * 2 : 0;
-   o0 += (g2 & 8) */
-	and	%g2, 3, %g4
-	andcc	%g2, 4, %g0
-	and	%g2, 8, %g2
-	sll	%g4, 1, %g4
-	be	1f
-	 add	%o0, %g2, %o0
-	add	%g2, %g4, %g2
-1:	and	%o2, 0xf, %g3
-	add	%g3, %o3, %g3
-	ba	fixupretl
-	 sub	%g3, %g2, %g3
-54:
-/* g3 = o3 + (o2 & 15) - (g2 / 4) * 2 - (g2 & 2) ? (g2 & 1) : 0;
-   o0 += (g2 / 4) * 2 */
-	srl	%g2, 2, %o4
-	and	%g2, 1, %o5
-	srl	%g2, 1, %g2
-	add	%o4, %o4, %o4
-	and	%o5, %g2, %o5
-	and	%o2, 0xf, %o2
-	add	%o0, %o4, %o0
-	sub	%o3, %o5, %o3
-	sub	%o2, %o4, %o2
-	ba	fixupretl
-	 add	%o2, %o3, %g3
-55:
-/* i = 27 - g2;
-   g3 = (o2 & 1) + i / 4 * 2 + !(i & 3);
-   o0 -= i / 4 * 2 + 1 */
-	neg	%g2
-	and	%o2, 1, %o2
-	add	%g2, 27, %g2
-	srl	%g2, 2, %o5
-	andcc	%g2, 3, %g0
-	mov	1, %g2
-	add	%o5, %o5, %o5
-	be,a	1f
-	 clr	%g2
-1:	add	%g2, %o5, %g3
-	sub	%o0, %g3, %o0
-	ba	fixupretl
-	 add	%g3, %o2, %g3
+	 mov	%o2, %o0
 
 	.globl  __copy_user_end
 __copy_user_end:
diff --git a/arch/sparc/lib/memset.S b/arch/sparc/lib/memset.S
index f427f34b8b79..eaff68213fdf 100644
--- a/arch/sparc/lib/memset.S
+++ b/arch/sparc/lib/memset.S
@@ -19,7 +19,7 @@
 98: 	x,y;					\
 	.section .fixup,ALLOC,EXECINSTR;	\
 	.align	4;				\
-99:	ba 30f;					\
+99:	retl;					\
 	 a, b, %o0;				\
 	.section __ex_table,ALLOC;		\
 	.align	4;				\
@@ -27,35 +27,44 @@
 	.text;					\
 	.align	4
 
-#define EXT(start,end,handler) 			\
+#define STORE(source, base, offset, n)		\
+98: 	std source, [base + offset + n];	\
+	.section .fixup,ALLOC,EXECINSTR;	\
+	.align	4;				\
+99:	ba 30f;					\
+	 sub %o3, n - offset, %o3;		\
 	.section __ex_table,ALLOC;		\
 	.align	4;				\
-	.word	start, 0, end, handler;		\
+	.word	98b, 99b;			\
 	.text;					\
-	.align	4
+	.align	4;
+
+#define STORE_LAST(source, base, offset, n)	\
+	EX(std source, [base - offset - n],	\
+	   add %o1, offset + n);
 
 /* Please don't change these macros, unless you change the logic
  * in the .fixup section below as well.
  * Store 64 bytes at (BASE + OFFSET) using value SOURCE. */
-#define ZERO_BIG_BLOCK(base, offset, source)    \
-	std	source, [base + offset + 0x00]; \
-	std	source, [base + offset + 0x08]; \
-	std	source, [base + offset + 0x10]; \
-	std	source, [base + offset + 0x18]; \
-	std	source, [base + offset + 0x20]; \
-	std	source, [base + offset + 0x28]; \
-	std	source, [base + offset + 0x30]; \
-	std	source, [base + offset + 0x38];
+#define ZERO_BIG_BLOCK(base, offset, source)	\
+	STORE(source, base, offset, 0x00);	\
+	STORE(source, base, offset, 0x08);	\
+	STORE(source, base, offset, 0x10);	\
+	STORE(source, base, offset, 0x18);	\
+	STORE(source, base, offset, 0x20);	\
+	STORE(source, base, offset, 0x28);	\
+	STORE(source, base, offset, 0x30);	\
+	STORE(source, base, offset, 0x38);
 
 #define ZERO_LAST_BLOCKS(base, offset, source)	\
-	std	source, [base - offset - 0x38]; \
-	std	source, [base - offset - 0x30]; \
-	std	source, [base - offset - 0x28]; \
-	std	source, [base - offset - 0x20]; \
-	std	source, [base - offset - 0x18]; \
-	std	source, [base - offset - 0x10]; \
-	std	source, [base - offset - 0x08]; \
-	std	source, [base - offset - 0x00];
+	STORE_LAST(source, base, offset, 0x38);	\
+	STORE_LAST(source, base, offset, 0x30); \
+	STORE_LAST(source, base, offset, 0x28);	\
+	STORE_LAST(source, base, offset, 0x20);	\
+	STORE_LAST(source, base, offset, 0x18);	\
+	STORE_LAST(source, base, offset, 0x10);	\
+	STORE_LAST(source, base, offset, 0x08);	\
+	STORE_LAST(source, base, offset, 0x00);
 
 	.text
 	.align 4
@@ -68,8 +77,6 @@ __bzero_begin:
 	.globl	memset
 	EXPORT_SYMBOL(__bzero)
 	EXPORT_SYMBOL(memset)
-	.globl	__memset_start, __memset_end
-__memset_start:
 memset:
 	mov	%o0, %g1
 	mov	1, %g4
@@ -122,8 +129,6 @@ __bzero:
 	ZERO_BIG_BLOCK(%o0, 0x00, %g2)
 	subcc	%o3, 128, %o3
 	ZERO_BIG_BLOCK(%o0, 0x40, %g2)
-11:
-	EXT(10b, 11b, 20f)
 	bne	10b
 	 add	%o0, 128, %o0
 
@@ -138,11 +143,9 @@ __bzero:
 	jmp	%o4
 	 add	%o0, %o2, %o0
 
-12:
 	ZERO_LAST_BLOCKS(%o0, 0x48, %g2)
 	ZERO_LAST_BLOCKS(%o0, 0x08, %g2)
 13:
-	EXT(12b, 13b, 21f)
 	be	8f
 	 andcc	%o1, 4, %g0
 
@@ -182,37 +185,13 @@ __bzero:
 5:
 	retl
 	 clr	%o0
-__memset_end:
 
 	.section .fixup,#alloc,#execinstr
 	.align	4
-20:
-	cmp	%g2, 8
-	bleu	1f
-	 and	%o1, 0x7f, %o1
-	sub	%g2, 9, %g2
-	add	%o3, 64, %o3
-1:
-	sll	%g2, 3, %g2
-	add	%o3, %o1, %o0
-	b 30f
-	 sub	%o0, %g2, %o0
-21:
-	mov	8, %o0
-	and	%o1, 7, %o1
-	sub	%o0, %g2, %o0
-	sll	%o0, 3, %o0
-	b 30f
-	 add	%o0, %o1, %o0
 30:
-/* %o4 is faulting address, %o5 is %pc where fault occurred */
-	save	%sp, -104, %sp
-	mov	%i5, %o0
-	mov	%i7, %o1
-	call	lookup_fault
-	 mov	%i4, %o2
-	ret
-	 restore
+	and	%o1, 0x7f, %o1
+	retl
+	 add	%o3, %o1, %o0
 
 	.globl __bzero_end
 __bzero_end:
diff --git a/arch/sparc/mm/Makefile b/arch/sparc/mm/Makefile
index 68db1f859b02..871354aa3c00 100644
--- a/arch/sparc/mm/Makefile
+++ b/arch/sparc/mm/Makefile
@@ -8,7 +8,7 @@ ccflags-y := -Werror
 obj-$(CONFIG_SPARC64)   += ultra.o tlb.o tsb.o
 obj-y                   += fault_$(BITS).o
 obj-y                   += init_$(BITS).o
-obj-$(CONFIG_SPARC32)   += extable.o srmmu.o iommu.o io-unit.o
+obj-$(CONFIG_SPARC32)   += srmmu.o iommu.o io-unit.o
 obj-$(CONFIG_SPARC32)   += srmmu_access.o
 obj-$(CONFIG_SPARC32)   += hypersparc.o viking.o tsunami.o swift.o
 obj-$(CONFIG_SPARC32)   += leon_mm.o
diff --git a/arch/sparc/mm/extable.c b/arch/sparc/mm/extable.c
deleted file mode 100644
index 241b40641873..000000000000
--- a/arch/sparc/mm/extable.c
+++ /dev/null
@@ -1,107 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0
-/*
- * linux/arch/sparc/mm/extable.c
- */
-
-#include <linux/module.h>
-#include <linux/extable.h>
-#include <linux/uaccess.h>
-
-void sort_extable(struct exception_table_entry *start,
-		  struct exception_table_entry *finish)
-{
-}
-
-/* Caller knows they are in a range if ret->fixup == 0 */
-const struct exception_table_entry *
-search_extable(const struct exception_table_entry *base,
-	       const size_t num,
-	       unsigned long value)
-{
-	int i;
-
-	/* Single insn entries are encoded as:
-	 *	word 1:	insn address
-	 *	word 2:	fixup code address
-	 *
-	 * Range entries are encoded as:
-	 *	word 1: first insn address
-	 *	word 2: 0
-	 *	word 3: last insn address + 4 bytes
-	 *	word 4: fixup code address
-	 *
-	 * Deleted entries are encoded as:
-	 *	word 1: unused
-	 *	word 2: -1
-	 *
-	 * See asm/uaccess.h for more details.
-	 */
-
-	/* 1. Try to find an exact match. */
-	for (i = 0; i < num; i++) {
-		if (base[i].fixup == 0) {
-			/* A range entry, skip both parts. */
-			i++;
-			continue;
-		}
-
-		/* A deleted entry; see trim_init_extable */
-		if (base[i].fixup == -1)
-			continue;
-
-		if (base[i].insn == value)
-			return &base[i];
-	}
-
-	/* 2. Try to find a range match. */
-	for (i = 0; i < (num - 1); i++) {
-		if (base[i].fixup)
-			continue;
-
-		if (base[i].insn <= value && base[i + 1].insn > value)
-			return &base[i];
-
-		i++;
-	}
-
-        return NULL;
-}
-
-#ifdef CONFIG_MODULES
-/* We could memmove them around; easier to mark the trimmed ones. */
-void trim_init_extable(struct module *m)
-{
-	unsigned int i;
-	bool range;
-
-	for (i = 0; i < m->num_exentries; i += range ? 2 : 1) {
-		range = m->extable[i].fixup == 0;
-
-		if (within_module_init(m->extable[i].insn, m)) {
-			m->extable[i].fixup = -1;
-			if (range)
-				m->extable[i+1].fixup = -1;
-		}
-		if (range)
-			i++;
-	}
-}
-#endif /* CONFIG_MODULES */
-
-/* Special extable search, which handles ranges.  Returns fixup */
-unsigned long search_extables_range(unsigned long addr, unsigned long *g2)
-{
-	const struct exception_table_entry *entry;
-
-	entry = search_exception_tables(addr);
-	if (!entry)
-		return 0;
-
-	/* Inside range?  Fix g2 and return correct fixup */
-	if (!entry->fixup) {
-		*g2 = (addr - entry->insn) / 4;
-		return (entry + 1)->fixup;
-	}
-
-	return entry->fixup;
-}
diff --git a/arch/sparc/mm/fault_32.c b/arch/sparc/mm/fault_32.c
index 40ce087dfecf..de2031c2b2d7 100644
--- a/arch/sparc/mm/fault_32.c
+++ b/arch/sparc/mm/fault_32.c
@@ -23,6 +23,7 @@
 #include <linux/interrupt.h>
 #include <linux/kdebug.h>
 #include <linux/uaccess.h>
+#include <linux/extable.h>
 
 #include <asm/page.h>
 #include <asm/openprom.h>
@@ -54,54 +55,6 @@ static void __noreturn unhandled_fault(unsigned long address,
 	die_if_kernel("Oops", regs);
 }
 
-asmlinkage int lookup_fault(unsigned long pc, unsigned long ret_pc,
-			    unsigned long address)
-{
-	struct pt_regs regs;
-	unsigned long g2;
-	unsigned int insn;
-	int i;
-
-	i = search_extables_range(ret_pc, &g2);
-	switch (i) {
-	case 3:
-		/* load & store will be handled by fixup */
-		return 3;
-
-	case 1:
-		/* store will be handled by fixup, load will bump out */
-		/* for _to_ macros */
-		insn = *((unsigned int *) pc);
-		if ((insn >> 21) & 1)
-			return 1;
-		break;
-
-	case 2:
-		/* load will be handled by fixup, store will bump out */
-		/* for _from_ macros */
-		insn = *((unsigned int *) pc);
-		if (!((insn >> 21) & 1) || ((insn>>19)&0x3f) == 15)
-			return 2;
-		break;
-
-	default:
-		break;
-	}
-
-	memset(&regs, 0, sizeof(regs));
-	regs.pc = pc;
-	regs.npc = pc + 4;
-	__asm__ __volatile__(
-		"rd %%psr, %0\n\t"
-		"nop\n\t"
-		"nop\n\t"
-		"nop\n" : "=r" (regs.psr));
-	unhandled_fault(address, current, &regs);
-
-	/* Not reached */
-	return 0;
-}
-
 static inline void
 show_signal_msg(struct pt_regs *regs, int sig, int code,
 		unsigned long address, struct task_struct *tsk)
@@ -162,8 +115,6 @@ asmlinkage void do_sparc_fault(struct pt_regs *regs, int text_fault, int write,
 	struct vm_area_struct *vma;
 	struct task_struct *tsk = current;
 	struct mm_struct *mm = tsk->mm;
-	unsigned int fixup;
-	unsigned long g2;
 	int from_user = !(regs->psr & PSR_PS);
 	int code;
 	vm_fault_t fault;
@@ -281,30 +232,19 @@ bad_area_nosemaphore:
 
 	/* Is this in ex_table? */
 no_context:
-	g2 = regs->u_regs[UREG_G2];
 	if (!from_user) {
-		fixup = search_extables_range(regs->pc, &g2);
-		/* Values below 10 are reserved for other things */
-		if (fixup > 10) {
-			extern const unsigned int __memset_start[];
-			extern const unsigned int __memset_end[];
+		const struct exception_table_entry *entry;
 
+		entry = search_exception_tables(regs->pc);
 #ifdef DEBUG_EXCEPTIONS
-			printk("Exception: PC<%08lx> faddr<%08lx>\n",
-			       regs->pc, address);
-			printk("EX_TABLE: insn<%08lx> fixup<%08x> g2<%08lx>\n",
-				regs->pc, fixup, g2);
+		printk("Exception: PC<%08lx> faddr<%08lx>\n",
+		       regs->pc, address);
+		printk("EX_TABLE: insn<%08lx> fixup<%08x>\n",
+			regs->pc, entry->fixup);
 #endif
-			if ((regs->pc >= (unsigned long)__memset_start &&
-			     regs->pc < (unsigned long)__memset_end)) {
-				regs->u_regs[UREG_I4] = address;
-				regs->u_regs[UREG_I5] = regs->pc;
-			}
-			regs->u_regs[UREG_G2] = g2;
-			regs->pc = fixup;
-			regs->npc = regs->pc + 4;
-			return;
-		}
+		regs->pc = entry->fixup;
+		regs->npc = regs->pc + 4;
+		return;
 	}
 
 	unhandled_fault(address, tsk, regs);
diff --git a/arch/sparc/mm/mm_32.h b/arch/sparc/mm/mm_32.h
index ce750a99eea9..ee55f1080634 100644
--- a/arch/sparc/mm/mm_32.h
+++ b/arch/sparc/mm/mm_32.h
@@ -1,7 +1,5 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 /* fault_32.c - visible as they are called from assembler */
-asmlinkage int lookup_fault(unsigned long pc, unsigned long ret_pc,
-                            unsigned long address);
 asmlinkage void do_sparc_fault(struct pt_regs *regs, int text_fault, int write,
                                unsigned long address);