aboutsummaryrefslogtreecommitdiff
path: root/arch/ia64/kernel
diff options
context:
space:
mode:
authorShaohua Li <shaohua.li@intel.com>2009-01-09 11:29:49 +0800
committerIngo Molnar <mingo@elte.hu>2009-01-14 12:11:31 +0100
commita14a07b8018b714e03a39ff2180c66e307ef4238 (patch)
tree7c4b9249f862969b80b74e5e11e56480b0e3d15c /arch/ia64/kernel
parentd3e75ff14bc1453c4762428395aac9953a023efc (diff)
ftrace, ia64: IA64 dynamic ftrace support
IA64 dynamic ftrace support. The original _mcount stub for each function is like: alloc r40=ar.pfs,12,8,0 mov r43=r0;; mov r42=b0 mov r41=r1 nop.i 0x0 br.call.sptk.many b0 = _mcount;; The patch convert it to below for nop: [MII] nop.m 0x0 mov r3=ip nop.i 0x0 [MLX] nop.m 0x0 nop.x 0x0;; This isn't completely nop, as there is one instuction 'mov r3=ip', but it should be light and harmless for code follow it. And below is for call [MII] nop.m 0x0 mov r3=ip nop.i 0x0 [MLX] nop.m 0x0 brl.many .;; In this way, only one instruction is changed to convert code between nop and call. This should meet dyn-ftrace's requirement. But this requires CPU support brl instruction, so dyn-ftrace isn't supported for old Itanium system. Assume there are quite few such old system running. Signed-off-by: Shaohua Li <shaohua.li@intel.com> Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'arch/ia64/kernel')
-rw-r--r--arch/ia64/kernel/Makefile5
-rw-r--r--arch/ia64/kernel/entry.S51
-rw-r--r--arch/ia64/kernel/ftrace.c206
3 files changed, 262 insertions, 0 deletions
diff --git a/arch/ia64/kernel/Makefile b/arch/ia64/kernel/Makefile
index c381ea95489..ab6e7ec0bba 100644
--- a/arch/ia64/kernel/Makefile
+++ b/arch/ia64/kernel/Makefile
@@ -2,6 +2,10 @@
# Makefile for the linux kernel.
#
+ifdef CONFIG_DYNAMIC_FTRACE
+CFLAGS_REMOVE_ftrace.o = -pg
+endif
+
extra-y := head.o init_task.o vmlinux.lds
obj-y := acpi.o entry.o efi.o efi_stub.o gate-data.o fsys.o ia64_ksyms.o irq.o irq_ia64.o \
@@ -28,6 +32,7 @@ obj-$(CONFIG_IA64_CYCLONE) += cyclone.o
obj-$(CONFIG_CPU_FREQ) += cpufreq/
obj-$(CONFIG_IA64_MCA_RECOVERY) += mca_recovery.o
obj-$(CONFIG_KPROBES) += kprobes.o jprobes.o
+obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o
obj-$(CONFIG_KEXEC) += machine_kexec.o relocate_kernel.o crash.o
obj-$(CONFIG_CRASH_DUMP) += crash_dump.o
obj-$(CONFIG_IA64_UNCACHED_ALLOCATOR) += uncached.o
diff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S
index c2f7d798e2a..e0be92a6abb 100644
--- a/arch/ia64/kernel/entry.S
+++ b/arch/ia64/kernel/entry.S
@@ -1406,6 +1406,56 @@ GLOBAL_ENTRY(unw_init_running)
END(unw_init_running)
#ifdef CONFIG_FUNCTION_TRACER
+#ifdef CONFIG_DYNAMIC_FTRACE
+GLOBAL_ENTRY(_mcount)
+ br ftrace_stub
+END(_mcount)
+
+.here:
+ br.ret.sptk.many b0
+
+GLOBAL_ENTRY(ftrace_caller)
+ alloc out0 = ar.pfs, 8, 0, 4, 0
+ mov out3 = r0
+ ;;
+ mov out2 = b0
+ add r3 = 0x20, r3
+ mov out1 = r1;
+ br.call.sptk.many b0 = ftrace_patch_gp
+ //this might be called from module, so we must patch gp
+ftrace_patch_gp:
+ movl gp=__gp
+ mov b0 = r3
+ ;;
+.global ftrace_call;
+ftrace_call:
+{
+ .mlx
+ nop.m 0x0
+ movl r3 = .here;;
+}
+ alloc loc0 = ar.pfs, 4, 4, 2, 0
+ ;;
+ mov loc1 = b0
+ mov out0 = b0
+ mov loc2 = r8
+ mov loc3 = r15
+ ;;
+ adds out0 = -MCOUNT_INSN_SIZE, out0
+ mov out1 = in2
+ mov b6 = r3
+
+ br.call.sptk.many b0 = b6
+ ;;
+ mov ar.pfs = loc0
+ mov b0 = loc1
+ mov r8 = loc2
+ mov r15 = loc3
+ br ftrace_stub
+ ;;
+END(ftrace_caller)
+
+#else
GLOBAL_ENTRY(_mcount)
movl r2 = ftrace_stub
movl r3 = ftrace_trace_function;;
@@ -1435,6 +1485,7 @@ GLOBAL_ENTRY(_mcount)
br ftrace_stub
;;
END(_mcount)
+#endif
GLOBAL_ENTRY(ftrace_stub)
mov r3 = b0
diff --git a/arch/ia64/kernel/ftrace.c b/arch/ia64/kernel/ftrace.c
new file mode 100644
index 00000000000..7fc8c961b1f
--- /dev/null
+++ b/arch/ia64/kernel/ftrace.c
@@ -0,0 +1,206 @@
+/*
+ * Dynamic function tracing support.
+ *
+ * Copyright (C) 2008 Shaohua Li <shaohua.li@intel.com>
+ *
+ * For licencing details, see COPYING.
+ *
+ * Defines low-level handling of mcount calls when the kernel
+ * is compiled with the -pg flag. When using dynamic ftrace, the
+ * mcount call-sites get patched lazily with NOP till they are
+ * enabled. All code mutation routines here take effect atomically.
+ */
+
+#include <linux/uaccess.h>
+#include <linux/ftrace.h>
+
+#include <asm/cacheflush.h>
+#include <asm/patch.h>
+
+/* In IA64, each function will be added below two bundles with -pg option */
+static unsigned char __attribute__((aligned(8)))
+ftrace_orig_code[MCOUNT_INSN_SIZE] = {
+ 0x02, 0x40, 0x31, 0x10, 0x80, 0x05, /* alloc r40=ar.pfs,12,8,0 */
+ 0xb0, 0x02, 0x00, 0x00, 0x42, 0x40, /* mov r43=r0;; */
+ 0x05, 0x00, 0xc4, 0x00, /* mov r42=b0 */
+ 0x11, 0x48, 0x01, 0x02, 0x00, 0x21, /* mov r41=r1 */
+ 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, /* nop.i 0x0 */
+ 0x08, 0x00, 0x00, 0x50 /* br.call.sptk.many b0 = _mcount;; */
+};
+
+struct ftrace_orig_insn {
+ u64 dummy1, dummy2, dummy3;
+ u64 dummy4:64-41+13;
+ u64 imm20:20;
+ u64 dummy5:3;
+ u64 sign:1;
+ u64 dummy6:4;
+};
+
+/* mcount stub will be converted below for nop */
+static unsigned char ftrace_nop_code[MCOUNT_INSN_SIZE] = {
+ 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, /* [MII] nop.m 0x0 */
+ 0x30, 0x00, 0x00, 0x60, 0x00, 0x00, /* mov r3=ip */
+ 0x00, 0x00, 0x04, 0x00, /* nop.i 0x0 */
+ 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, /* [MLX] nop.m 0x0 */
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, /* nop.x 0x0;; */
+ 0x00, 0x00, 0x04, 0x00
+};
+
+static unsigned char *ftrace_nop_replace(void)
+{
+ return ftrace_nop_code;
+}
+
+/*
+ * mcount stub will be converted below for call
+ * Note: Just the last instruction is changed against nop
+ * */
+static unsigned char __attribute__((aligned(8)))
+ftrace_call_code[MCOUNT_INSN_SIZE] = {
+ 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, /* [MII] nop.m 0x0 */
+ 0x30, 0x00, 0x00, 0x60, 0x00, 0x00, /* mov r3=ip */
+ 0x00, 0x00, 0x04, 0x00, /* nop.i 0x0 */
+ 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, /* [MLX] nop.m 0x0 */
+ 0xff, 0xff, 0xff, 0xff, 0x7f, 0x00, /* brl.many .;;*/
+ 0xf8, 0xff, 0xff, 0xc8
+};
+
+struct ftrace_call_insn {
+ u64 dummy1, dummy2;
+ u64 dummy3:48;
+ u64 imm39_l:16;
+ u64 imm39_h:23;
+ u64 dummy4:13;
+ u64 imm20:20;
+ u64 dummy5:3;
+ u64 i:1;
+ u64 dummy6:4;
+};
+
+static unsigned char *ftrace_call_replace(unsigned long ip, unsigned long addr)
+{
+ struct ftrace_call_insn *code = (void *)ftrace_call_code;
+ unsigned long offset = addr - (ip + 0x10);
+
+ code->imm39_l = offset >> 24;
+ code->imm39_h = offset >> 40;
+ code->imm20 = offset >> 4;
+ code->i = offset >> 63;
+ return ftrace_call_code;
+}
+
+static int
+ftrace_modify_code(unsigned long ip, unsigned char *old_code,
+ unsigned char *new_code, int do_check)
+{
+ unsigned char replaced[MCOUNT_INSN_SIZE];
+
+ /*
+ * Note: Due to modules and __init, code can
+ * disappear and change, we need to protect against faulting
+ * as well as code changing. We do this by using the
+ * probe_kernel_* functions.
+ *
+ * No real locking needed, this code is run through
+ * kstop_machine, or before SMP starts.
+ */
+
+ if (!do_check)
+ goto skip_check;
+
+ /* read the text we want to modify */
+ if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
+ return -EFAULT;
+
+ /* Make sure it is what we expect it to be */
+ if (memcmp(replaced, old_code, MCOUNT_INSN_SIZE) != 0)
+ return -EINVAL;
+
+skip_check:
+ /* replace the text with the new text */
+ if (probe_kernel_write(((void *)ip), new_code, MCOUNT_INSN_SIZE))
+ return -EPERM;
+ flush_icache_range(ip, ip + MCOUNT_INSN_SIZE);
+
+ return 0;
+}
+
+static int ftrace_make_nop_check(struct dyn_ftrace *rec, unsigned long addr)
+{
+ unsigned char __attribute__((aligned(8))) replaced[MCOUNT_INSN_SIZE];
+ unsigned long ip = rec->ip;
+
+ if (probe_kernel_read(replaced, (void *)ip, MCOUNT_INSN_SIZE))
+ return -EFAULT;
+ if (rec->flags & FTRACE_FL_CONVERTED) {
+ struct ftrace_call_insn *call_insn, *tmp_call;
+
+ call_insn = (void *)ftrace_call_code;
+ tmp_call = (void *)replaced;
+ call_insn->imm39_l = tmp_call->imm39_l;
+ call_insn->imm39_h = tmp_call->imm39_h;
+ call_insn->imm20 = tmp_call->imm20;
+ call_insn->i = tmp_call->i;
+ if (memcmp(replaced, ftrace_call_code, MCOUNT_INSN_SIZE) != 0)
+ return -EINVAL;
+ return 0;
+ } else {
+ struct ftrace_orig_insn *call_insn, *tmp_call;
+
+ call_insn = (void *)ftrace_orig_code;
+ tmp_call = (void *)replaced;
+ call_insn->sign = tmp_call->sign;
+ call_insn->imm20 = tmp_call->imm20;
+ if (memcmp(replaced, ftrace_orig_code, MCOUNT_INSN_SIZE) != 0)
+ return -EINVAL;
+ return 0;
+ }
+}
+
+int ftrace_make_nop(struct module *mod,
+ struct dyn_ftrace *rec, unsigned long addr)
+{
+ int ret;
+ char *new;
+
+ ret = ftrace_make_nop_check(rec, addr);
+ if (ret)
+ return ret;
+ new = ftrace_nop_replace();
+ return ftrace_modify_code(rec->ip, NULL, new, 0);
+}
+
+int ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
+{
+ unsigned long ip = rec->ip;
+ unsigned char *old, *new;
+
+ old= ftrace_nop_replace();
+ new = ftrace_call_replace(ip, addr);
+ return ftrace_modify_code(ip, old, new, 1);
+}
+
+/* in IA64, _mcount can't directly call ftrace_stub. Only jump is ok */
+int ftrace_update_ftrace_func(ftrace_func_t func)
+{
+ unsigned long ip;
+ unsigned long addr = ((struct fnptr *)ftrace_call)->ip;
+
+ if (func == ftrace_stub)
+ return 0;
+ ip = ((struct fnptr *)func)->ip;
+
+ ia64_patch_imm64(addr + 2, ip);
+
+ flush_icache_range(addr, addr + 16);
+ return 0;
+}
+
+/* run from kstop_machine */
+int __init ftrace_dyn_arch_init(void *data)
+{
+ *(unsigned long *)data = 0;
+
+ return 0;
+}