Merge branch 'sparc-vdso'

sparc: VDSO improvements

I started out on these changes with the goal of improving perf
annotations when the VDSO is in use.  Due to lack of inlining the
helper functions are typically hit when profiling instead of
__vdso_gettimeoday() or __vdso_vclock_gettime().

The only symbols available by default are the dyanmic symbols,
which therefore doesn't cover the helper functions.

So the perf output looks terrible, because the symbols cannot be
resolved and all show up as "Unknown".

The sparc VDSO code forces no inlining because of the way the
simplistic %tick register read code patching works.  So fixing that
was the first order of business.  Tricks were taken from how x86
implements alternates.  The crucial factor is that if you want to
refer to locations (for the original and patch instruction(s)) you
have to do so in a way that is resolvable at link time even for a
shared object.  So you have to do this by storing PC-relative
values, and not in executable sections.

Next, we sanitize the Makefile so that the cflags et al. make more
sense.  And LDFLAGS are applied actually to invocations of LD instead
of CC.

We also add some sanity checking, specifically in a post-link check
that makes sure we don't have any unexpected unresolved symbols in the
VDSO.  This is essential because the dynamic linker cannot resolve
symbols in the VDSO because it cannot write to it.

Finally some very minor optimizations are preformed to the
vclock_gettime.c code.  One thing which is tricky with this code on
sparc is that struct timeval and struct timespec are layed out
differently on 64-bit.  This is because, unlike other architectures,
sparc defined suseconds_t as 'int' even on 64-bit.  This is why we
have all of the "union" tstv_t" business and the weird assignments
in __vdso_gettimeofday().

Performance wise we do gain some cycle shere, specifically here
are cycle counts for a user application calling gettimeofday():

	no-VDSO		VDSO-orig	VDSO-new
================================================
64-bit	853 cycles	112 cycles	125 cycles
32-bit	849 cycles	134 cycles	141 cycles

These results are with current glibc sources.

To get better we'd need to implement this in assembler, and I might
just do that at some point.

Signed-off-by: David S. Miller <davem@davemloft.net>

10 files changed, 121 insertions, 107 deletions

diff --git a/arch/sparc/include/asm/thread_info_64.h b/arch/sparc/include/asm/thread_info_64.h
index 7fb676360928..20255471e653 100644
--- a/arch/sparc/include/asm/thread_info_64.h
+++ b/arch/sparc/include/asm/thread_info_64.h
@@ -121,8 +121,12 @@ struct thread_info {
 }
 
 /* how to get the thread information struct from C */
+#ifndef BUILD_VDSO
 register struct thread_info *current_thread_info_reg asm("g6");
 #define current_thread_info()	(current_thread_info_reg)
+#else
+extern struct thread_info *current_thread_info(void);
+#endif
 
 /* thread information allocation */
 #if PAGE_SHIFT == 13
diff --git a/arch/sparc/include/asm/vdso.h b/arch/sparc/include/asm/vdso.h
index 93b628731a5e..56836eb01787 100644
--- a/arch/sparc/include/asm/vdso.h
+++ b/arch/sparc/include/asm/vdso.h
@@ -8,10 +8,10 @@
 struct vdso_image {
 	void *data;
 	unsigned long size;   /* Always a multiple of PAGE_SIZE */
+
+	unsigned long tick_patch, tick_patch_len;
+
 	long sym_vvar_start;  /* Negative offset to the vvar area */
-	long sym_vread_tick; /* Start of vread_tick section */
-	long sym_vread_tick_patch_start; /* Start of tick read */
-	long sym_vread_tick_patch_end;   /* End of tick read */
 };
 
 #ifdef CONFIG_SPARC64
diff --git a/arch/sparc/kernel/time_64.c b/arch/sparc/kernel/time_64.c
index f0eba72aa1ad..5f356dc8e178 100644
--- a/arch/sparc/kernel/time_64.c
+++ b/arch/sparc/kernel/time_64.c
@@ -53,8 +53,6 @@
 
 DEFINE_SPINLOCK(rtc_lock);
 
-unsigned int __read_mostly vdso_fix_stick;
-
 #ifdef CONFIG_SMP
 unsigned long profile_pc(struct pt_regs *regs)
 {
@@ -838,7 +836,6 @@ void __init time_init_early(void)
 		} else {
 			init_tick_ops(&tick_operations);
 			clocksource_tick.archdata.vclock_mode = VCLOCK_TICK;
-			vdso_fix_stick = 1;
 		}
 	} else {
 		init_tick_ops(&stick_operations);
diff --git a/arch/sparc/vdso/Makefile b/arch/sparc/vdso/Makefile
index dc85570d8839..a6e18ca4cc18 100644
--- a/arch/sparc/vdso/Makefile
+++ b/arch/sparc/vdso/Makefile
@@ -33,10 +33,8 @@ targets += $(vdso_img_sodbg) $(vdso_img-y:%=vdso%.so)
 
 CPPFLAGS_vdso.lds += -P -C
 
-VDSO_LDFLAGS_vdso.lds = -m64 -Wl,-soname=linux-vdso.so.1 \
-			-Wl,--no-undefined \
-			-Wl,-z,max-page-size=8192 -Wl,-z,common-page-size=8192 \
-			$(DISABLE_LTO)
+VDSO_LDFLAGS_vdso.lds = -m elf64_sparc -soname linux-vdso.so.1 --no-undefined \
+			-z max-page-size=8192 -z common-page-size=8192
 
 $(obj)/vdso64.so.dbg: $(obj)/vdso.lds $(vobjs) FORCE
 	$(call if_changed,vdso)
@@ -54,13 +52,14 @@ $(obj)/vdso-image-%.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
 # Don't omit frame pointers for ease of userspace debugging, but do
 # optimize sibling calls.
 #
-CFL := $(PROFILING) -mcmodel=medlow -fPIC -O2 -fasynchronous-unwind-tables \
-       -m64 -ffixed-g2 -ffixed-g3 -fcall-used-g4 -fcall-used-g5 -ffixed-g6 \
-       -ffixed-g7 $(filter -g%,$(KBUILD_CFLAGS)) \
-       $(call cc-option, -fno-stack-protector) -fno-omit-frame-pointer \
-       -foptimize-sibling-calls -DBUILD_VDSO
+CFL := $(PROFILING) -mcmodel=medlow -fPIC -O2 -fasynchronous-unwind-tables -m64 \
+       $(filter -g%,$(KBUILD_CFLAGS)) $(call cc-option, -fno-stack-protector) \
+       -fno-omit-frame-pointer -foptimize-sibling-calls \
+       -DDISABLE_BRANCH_PROFILING -DBUILD_VDSO
 
-$(vobjs): KBUILD_CFLAGS += $(CFL)
+SPARC_REG_CFLAGS = -ffixed-g4 -ffixed-g5 -fcall-used-g5 -fcall-used-g7
+
+$(vobjs): KBUILD_CFLAGS := $(filter-out $(GCC_PLUGINS_CFLAGS) $(SPARC_REG_CFLAGS),$(KBUILD_CFLAGS)) $(CFL)
 
 #
 # vDSO code runs in userspace and -pg doesn't help with profiling anyway.
@@ -73,7 +72,7 @@ $(obj)/%.so: $(obj)/%.so.dbg
 	$(call if_changed,objcopy)
 
 CPPFLAGS_vdso32.lds = $(CPPFLAGS_vdso.lds)
-VDSO_LDFLAGS_vdso32.lds = -m32 -Wl,-m,elf32_sparc,-soname=linux-gate.so.1
+VDSO_LDFLAGS_vdso32.lds = -m elf32_sparc -soname linux-gate.so.1
 
 #This makes sure the $(obj) subdirectory exists even though vdso32/
 #is not a kbuild sub-make subdirectory
@@ -91,7 +90,8 @@ KBUILD_CFLAGS_32 := $(filter-out -m64,$(KBUILD_CFLAGS))
 KBUILD_CFLAGS_32 := $(filter-out -mcmodel=medlow,$(KBUILD_CFLAGS_32))
 KBUILD_CFLAGS_32 := $(filter-out -fno-pic,$(KBUILD_CFLAGS_32))
 KBUILD_CFLAGS_32 := $(filter-out $(GCC_PLUGINS_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 += -m32 -msoft-float -fpic -mno-app-regs -ffixed-g7
+KBUILD_CFLAGS_32 := $(filter-out $(SPARC_REG_CFLAGS),$(KBUILD_CFLAGS_32))
+KBUILD_CFLAGS_32 += -m32 -msoft-float -fpic
 KBUILD_CFLAGS_32 += $(call cc-option, -fno-stack-protector)
 KBUILD_CFLAGS_32 += $(call cc-option, -foptimize-sibling-calls)
 KBUILD_CFLAGS_32 += -fno-omit-frame-pointer
@@ -109,12 +109,13 @@ $(obj)/vdso32.so.dbg: FORCE \
 # The DSO images are built using a special linker script.
 #
 quiet_cmd_vdso = VDSO    $@
-      cmd_vdso = $(CC) -nostdlib -o $@ \
+      cmd_vdso = $(LD) -nostdlib -o $@ \
 		       $(VDSO_LDFLAGS) $(VDSO_LDFLAGS_$(filter %.lds,$(^F))) \
-		       -Wl,-T,$(filter %.lds,$^) $(filter %.o,$^)
+		       -T $(filter %.lds,$^) $(filter %.o,$^) && \
+		sh $(srctree)/$(src)/checkundef.sh '$(OBJDUMP)' '$@'
 
-VDSO_LDFLAGS = -fPIC -shared $(call cc-ldoption, -Wl$(comma)--hash-style=sysv) \
-	$(call cc-ldoption, -Wl$(comma)--build-id) -Wl,-Bsymbolic
+VDSO_LDFLAGS = -shared $(call ld-option, --hash-style=both) \
+	$(call ld-option, --build-id) -Bsymbolic
 GCOV_PROFILE := n
 
 #
diff --git a/arch/sparc/vdso/checkundef.sh b/arch/sparc/vdso/checkundef.sh
new file mode 100644
index 000000000000..2d85876ffc32
--- /dev/null
+++ b/arch/sparc/vdso/checkundef.sh
@@ -0,0 +1,10 @@
+#!/bin/sh
+objdump="$1"
+file="$2"
+$objdump -t "$file" | grep '*UUND*' | grep -v '#scratch' > /dev/null 2>&1
+if [ $? -eq 1 ]; then
+    exit 0
+else
+    echo "$file: undefined symbols found" >&2
+    exit 1
+fi
diff --git a/arch/sparc/vdso/vclock_gettime.c b/arch/sparc/vdso/vclock_gettime.c
index 75dca9aab737..7b539ceebe13 100644
--- a/arch/sparc/vdso/vclock_gettime.c
+++ b/arch/sparc/vdso/vclock_gettime.c
@@ -12,11 +12,6 @@
  * Copyright (c) 2017 Oracle and/or its affiliates. All rights reserved.
  */
 
-/* Disable profiling for userspace code: */
-#ifndef	DISABLE_BRANCH_PROFILING
-#define	DISABLE_BRANCH_PROFILING
-#endif
-
 #include <linux/kernel.h>
 #include <linux/time.h>
 #include <linux/string.h>
@@ -26,13 +21,6 @@
 #include <asm/clocksource.h>
 #include <asm/vvar.h>
 
-#undef	TICK_PRIV_BIT
-#ifdef	CONFIG_SPARC64
-#define	TICK_PRIV_BIT	(1UL << 63)
-#else
-#define	TICK_PRIV_BIT	(1ULL << 63)
-#endif
-
 #ifdef	CONFIG_SPARC64
 #define SYSCALL_STRING							\
 	"ta	0x6d;"							\
@@ -60,24 +48,22 @@
  * Compute the vvar page's address in the process address space, and return it
  * as a pointer to the vvar_data.
  */
-static notrace noinline struct vvar_data *
-get_vvar_data(void)
+notrace static __always_inline struct vvar_data *get_vvar_data(void)
 {
 	unsigned long ret;
 
 	/*
-	 * vdso data page is the first vDSO page so grab the return address
+	 * vdso data page is the first vDSO page so grab the PC
 	 * and move up a page to get to the data page.
 	 */
-	ret = (unsigned long)__builtin_return_address(0);
+	__asm__("rd %%pc, %0" : "=r" (ret));
 	ret &= ~(8192 - 1);
 	ret -= 8192;
 
 	return (struct vvar_data *) ret;
 }
 
-static notrace long
-vdso_fallback_gettime(long clock, struct timespec *ts)
+notrace static long vdso_fallback_gettime(long clock, struct timespec *ts)
 {
 	register long num __asm__("g1") = __NR_clock_gettime;
 	register long o0 __asm__("o0") = clock;
@@ -88,8 +74,7 @@ vdso_fallback_gettime(long clock, struct timespec *ts)
 	return o0;
 }
 
-static notrace __always_inline long
-vdso_fallback_gettimeofday(struct timeval *tv, struct timezone *tz)
+notrace static long vdso_fallback_gettimeofday(struct timeval *tv, struct timezone *tz)
 {
 	register long num __asm__("g1") = __NR_gettimeofday;
 	register long o0 __asm__("o0") = (long) tv;
@@ -101,38 +86,43 @@ vdso_fallback_gettimeofday(struct timeval *tv, struct timezone *tz)
 }
 
 #ifdef	CONFIG_SPARC64
-static notrace noinline u64
-vread_tick(void) {
+notrace static __always_inline u64 vread_tick(void)
+{
 	u64	ret;
 
-	__asm__ __volatile__("rd	%%asr24, %0 \n"
-			     ".section	.vread_tick_patch, \"ax\" \n"
-			     "rd	%%tick, %0 \n"
-			     ".previous \n"
-			     : "=&r" (ret));
-	return ret & ~TICK_PRIV_BIT;
+	__asm__ __volatile__("1:\n\t"
+			     "rd		%%tick, %0\n\t"
+			     ".pushsection	.tick_patch, \"a\"\n\t"
+			     ".word		1b - ., 1f - .\n\t"
+			     ".popsection\n\t"
+			     ".pushsection	.tick_patch_replacement, \"ax\"\n\t"
+			     "1:\n\t"
+			     "rd		%%asr24, %0\n\t"
+			     ".popsection\n"
+			     : "=r" (ret));
+	return ret;
 }
 #else
-static notrace noinline u64
-vread_tick(void)
+notrace static __always_inline u64 vread_tick(void)
 {
-	unsigned int lo, hi;
-
-	__asm__ __volatile__("rd	%%asr24, %%g1\n\t"
-			     "srlx	%%g1, 32, %1\n\t"
-			     "srl	%%g1, 0, %0\n"
-			     ".section	.vread_tick_patch, \"ax\" \n"
-			     "rd	%%tick, %%g1\n"
-			     ".previous \n"
-			     : "=&r" (lo), "=&r" (hi)
-			     :
-			     : "g1");
-	return lo | ((u64)hi << 32);
+	register unsigned long long ret asm("o4");
+
+	__asm__ __volatile__("1:\n\t"
+			     "rd		%%tick, %L0\n\t"
+			     "srlx		%L0, 32, %H0\n\t"
+			     ".pushsection	.tick_patch, \"a\"\n\t"
+			     ".word		1b - ., 1f - .\n\t"
+			     ".popsection\n\t"
+			     ".pushsection	.tick_patch_replacement, \"ax\"\n\t"
+			     "1:\n\t"
+			     "rd		%%asr24, %L0\n\t"
+			     ".popsection\n"
+			     : "=r" (ret));
+	return ret;
 }
 #endif
 
-static notrace inline u64
-vgetsns(struct vvar_data *vvar)
+notrace static __always_inline u64 vgetsns(struct vvar_data *vvar)
 {
 	u64 v;
 	u64 cycles;
@@ -142,13 +132,12 @@ vgetsns(struct vvar_data *vvar)
 	return v * vvar->clock.mult;
 }
 
-static notrace noinline int
-do_realtime(struct vvar_data *vvar, struct timespec *ts)
+notrace static __always_inline int do_realtime(struct vvar_data *vvar,
+					       struct timespec *ts)
 {
 	unsigned long seq;
 	u64 ns;
 
-	ts->tv_nsec = 0;
 	do {
 		seq = vvar_read_begin(vvar);
 		ts->tv_sec = vvar->wall_time_sec;
@@ -157,18 +146,18 @@ do_realtime(struct vvar_data *vvar, struct timespec *ts)
 		ns >>= vvar->clock.shift;
 	} while (unlikely(vvar_read_retry(vvar, seq)));
 
-	timespec_add_ns(ts, ns);
+	ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
+	ts->tv_nsec = ns;
 
 	return 0;
 }
 
-static notrace noinline int
-do_monotonic(struct vvar_data *vvar, struct timespec *ts)
+notrace static __always_inline int do_monotonic(struct vvar_data *vvar,
+						struct timespec *ts)
 {
 	unsigned long seq;
 	u64 ns;
 
-	ts->tv_nsec = 0;
 	do {
 		seq = vvar_read_begin(vvar);
 		ts->tv_sec = vvar->monotonic_time_sec;
@@ -177,13 +166,14 @@ do_monotonic(struct vvar_data *vvar, struct timespec *ts)
 		ns >>= vvar->clock.shift;
 	} while (unlikely(vvar_read_retry(vvar, seq)));
 
-	timespec_add_ns(ts, ns);
+	ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
+	ts->tv_nsec = ns;
 
 	return 0;
 }
 
-static notrace noinline int
-do_realtime_coarse(struct vvar_data *vvar, struct timespec *ts)
+notrace static int do_realtime_coarse(struct vvar_data *vvar,
+				      struct timespec *ts)
 {
 	unsigned long seq;
 
@@ -195,8 +185,8 @@ do_realtime_coarse(struct vvar_data *vvar, struct timespec *ts)
 	return 0;
 }
 
-static notrace noinline int
-do_monotonic_coarse(struct vvar_data *vvar, struct timespec *ts)
+notrace static int do_monotonic_coarse(struct vvar_data *vvar,
+				       struct timespec *ts)
 {
 	unsigned long seq;
 
diff --git a/arch/sparc/vdso/vdso-layout.lds.S b/arch/sparc/vdso/vdso-layout.lds.S
index f2c83abaca12..ed36d49e1617 100644
--- a/arch/sparc/vdso/vdso-layout.lds.S
+++ b/arch/sparc/vdso/vdso-layout.lds.S
@@ -73,11 +73,8 @@ SECTIONS
 
 	.text		: { *(.text*) }			:text	=0x90909090,
 
-	.vread_tick_patch : {
-		vread_tick_patch_start = .;
-		*(.vread_tick_patch)
-		vread_tick_patch_end = .;
-	}
+	.tick_patch 	  : { *(.tick_patch) }		:text
+	.tick_patch_insns : { *(.tick_patch_insns) }	:text
 
 	/DISCARD/ : {
 		*(.discard)
diff --git a/arch/sparc/vdso/vdso2c.c b/arch/sparc/vdso/vdso2c.c
index 9f5b1cd6d51d..ab7504176a7f 100644
--- a/arch/sparc/vdso/vdso2c.c
+++ b/arch/sparc/vdso/vdso2c.c
@@ -63,9 +63,6 @@ enum {
 	sym_vvar_start,
 	sym_VDSO_FAKE_SECTION_TABLE_START,
 	sym_VDSO_FAKE_SECTION_TABLE_END,
-	sym_vread_tick,
-	sym_vread_tick_patch_start,
-	sym_vread_tick_patch_end
 };
 
 struct vdso_sym {
@@ -81,9 +78,6 @@ struct vdso_sym required_syms[] = {
 	[sym_VDSO_FAKE_SECTION_TABLE_END] = {
 		"VDSO_FAKE_SECTION_TABLE_END", 0
 	},
-	[sym_vread_tick] = {"vread_tick", 1},
-	[sym_vread_tick_patch_start] = {"vread_tick_patch_start", 1},
-	[sym_vread_tick_patch_end] = {"vread_tick_patch_end", 1}
 };
 
 __attribute__((format(printf, 1, 2))) __attribute__((noreturn))
diff --git a/arch/sparc/vdso/vdso2c.h b/arch/sparc/vdso/vdso2c.h
index 808decb0f7be..4df005cf98c0 100644
--- a/arch/sparc/vdso/vdso2c.h
+++ b/arch/sparc/vdso/vdso2c.h
@@ -17,10 +17,11 @@ static void BITSFUNC(go)(void *raw_addr, size_t raw_len,
 	unsigned long mapping_size;
 	int i;
 	unsigned long j;
-
-	ELF(Shdr) *symtab_hdr = NULL, *strtab_hdr;
+	ELF(Shdr) *symtab_hdr = NULL, *strtab_hdr, *secstrings_hdr,
+		*patch_sec = NULL;
 	ELF(Ehdr) *hdr = (ELF(Ehdr) *)raw_addr;
 	ELF(Dyn) *dyn = 0, *dyn_end = 0;
+	const char *secstrings;
 	INT_BITS syms[NSYMS] = {};
 
 	ELF(Phdr) *pt = (ELF(Phdr) *)(raw_addr + GET_BE(&hdr->e_phoff));
@@ -63,11 +64,18 @@ static void BITSFUNC(go)(void *raw_addr, size_t raw_len,
 	}
 
 	/* Walk the section table */
+	secstrings_hdr = raw_addr + GET_BE(&hdr->e_shoff) +
+		GET_BE(&hdr->e_shentsize)*GET_BE(&hdr->e_shstrndx);
+	secstrings = raw_addr + GET_BE(&secstrings_hdr->sh_offset);
 	for (i = 0; i < GET_BE(&hdr->e_shnum); i++) {
 		ELF(Shdr) *sh = raw_addr + GET_BE(&hdr->e_shoff) +
 			GET_BE(&hdr->e_shentsize) * i;
 		if (GET_BE(&sh->sh_type) == SHT_SYMTAB)
 			symtab_hdr = sh;
+
+		if (!strcmp(secstrings + GET_BE(&sh->sh_name),
+			    ".tick_patch"))
+			patch_sec = sh;
 	}
 
 	if (!symtab_hdr)
@@ -134,6 +142,12 @@ static void BITSFUNC(go)(void *raw_addr, size_t raw_len,
 	fprintf(outfile, "const struct vdso_image %s_builtin = {\n", name);
 	fprintf(outfile, "\t.data = raw_data,\n");
 	fprintf(outfile, "\t.size = %lu,\n", mapping_size);
+	if (patch_sec) {
+		fprintf(outfile, "\t.tick_patch = %lu,\n",
+			(unsigned long)GET_BE(&patch_sec->sh_offset));
+		fprintf(outfile, "\t.tick_patch_len = %lu,\n",
+			(unsigned long)GET_BE(&patch_sec->sh_size));
+	}
 	for (i = 0; i < NSYMS; i++) {
 		if (required_syms[i].export && syms[i])
 			fprintf(outfile, "\t.sym_%s = %" PRIi64 ",\n",
diff --git a/arch/sparc/vdso/vma.c b/arch/sparc/vdso/vma.c
index 5eaff3c1aa0c..8874a27d8adc 100644
--- a/arch/sparc/vdso/vma.c
+++ b/arch/sparc/vdso/vma.c
@@ -16,6 +16,8 @@
 #include <linux/linkage.h>
 #include <linux/random.h>
 #include <linux/elf.h>
+#include <asm/cacheflush.h>
+#include <asm/spitfire.h>
 #include <asm/vdso.h>
 #include <asm/vvar.h>
 #include <asm/page.h>
@@ -40,7 +42,25 @@ static struct vm_special_mapping vdso_mapping32 = {
 
 struct vvar_data *vvar_data;
 
-#define	SAVE_INSTR_SIZE	4
+struct tick_patch_entry {
+	s32 orig, repl;
+};
+
+static void stick_patch(const struct vdso_image *image)
+{
+	struct tick_patch_entry *p, *p_end;
+
+	p = image->data + image->tick_patch;
+	p_end = (void *)p + image->tick_patch_len;
+	while (p < p_end) {
+		u32 *instr = (void *)&p->orig + p->orig;
+		u32 *repl = (void *)&p->repl + p->repl;
+
+		*instr = *repl;
+		flushi(instr);
+		p++;
+	}
+}
 
 /*
  * Allocate pages for the vdso and vvar, and copy in the vdso text from the
@@ -68,21 +88,8 @@ int __init init_vdso_image(const struct vdso_image *image,
 	if (!cpp)
 		goto oom;
 
-	if (vdso_fix_stick) {
-		/*
-		 * If the system uses %tick instead of %stick, patch the VDSO
-		 * with instruction reading %tick instead of %stick.
-		 */
-		unsigned int j, k = SAVE_INSTR_SIZE;
-		unsigned char *data = image->data;
-
-		for (j = image->sym_vread_tick_patch_start;
-		     j < image->sym_vread_tick_patch_end; j++) {
-
-			data[image->sym_vread_tick + k] = data[j];
-			k++;
-		}
-	}
+	if (tlb_type != spitfire)
+		stick_patch(image);
 
 	for (i = 0; i < cnpages; i++) {
 		cp = alloc_page(GFP_KERNEL);