On Mon, 16 May 2011 12:17:15 -0700
"Luck, Tony" <tony.luck@xxxxxxxxx> wrote:
> This patch does some fast and loose translation from hex (0x12) in the kernel
> to decimal (18) in the user tools ... which doesn't look like it fits with
> either of these conventions.
All the patch does is to move an address calculation from the kernel "addr-0x12"
to user space by subtracting the constant offset from the addend value for the
relocation: "addend = _w(_w(relp->r_offset) - recval + mcount_adjust)".
The offset for ia64 is -18. I'd be very surprised if the outcome of the move
would yield a different end result. But as I took another look at the patch
I noticed a bug, for EM_IA64 we need to set mcount_adjust_64 and not
mcount_adjust_32 as it is a 64-bit architecture. New patch below.
I tried to do a test compile with/without the patch but with the latest git
kernels TRACING_SUPPORT depends on TRACE_IRQFLAGS_SUPPORT which is missing
for ia64. No luck ..
---
Subject: [PATCH] ia64 mcount offset calculation
From: Martin Schwidefsky <schwidefsky@xxxxxxxxxx>
Do the mcount offset adjustment in the recordmcount.pl/recordmcount.[ch]
at compile time and not in ftrace_call_adjust at run time.
Signed-off-by: Martin Schwidefsky <schwidefsky@xxxxxxxxxx>
---
arch/ia64/include/asm/ftrace.h | 7 +++++--
scripts/recordmcount.c | 7 ++++++-
scripts/recordmcount.pl | 1 +
3 files changed, 12 insertions(+), 3 deletions(-)
diff -urpN linux-2.6/arch/ia64/include/asm/ftrace.h linux-2.6-patched/arch/ia64/include/asm/ftrace.h
--- linux-2.6/arch/ia64/include/asm/ftrace.h 2011-03-15 02:20:32.000000000 +0100
+++ linux-2.6-patched/arch/ia64/include/asm/ftrace.h 2011-05-16 10:16:20.000000000 +0200
@@ -14,8 +14,11 @@ extern void _mcount(unsigned long pfs, u
static inline unsigned long ftrace_call_adjust(unsigned long addr)
{
- /* second bundle, insn 2 */
- return addr - 0x12;
+ /*
+ * addr is the address of the mcount call instruction.
+ * recordmcount does the necessary offset calculation.
+ */
+ return addr;
}
struct dyn_arch_ftrace {
diff -urpN linux-2.6/scripts/recordmcount.c linux-2.6-patched/scripts/recordmcount.c
--- linux-2.6/scripts/recordmcount.c 2011-05-16 10:16:20.000000000 +0200
+++ linux-2.6-patched/scripts/recordmcount.c 2011-05-16 10:16:20.000000000 +0200
@@ -307,7 +307,12 @@ do_file(char const *const fname)
case EM_ARM: reltype = R_ARM_ABS32;
altmcount = "__gnu_mcount_nc";
break;
- case EM_IA_64: reltype = R_IA64_IMM64; gpfx = '_'; break;
+ case EM_IA_64:
+ reltype = R_IA64_IMM64;
+ /* Adjust relocation to second bundle, insn 2 */
+ mcount_adjust_64 = -18;
+ gpfx = '_';
+ break;
case EM_MIPS: /* reltype: e_class */ gpfx = '_'; break;
case EM_PPC: reltype = R_PPC_ADDR32; gpfx = '_'; break;
case EM_PPC64: reltype = R_PPC64_ADDR64; gpfx = '_'; break;
diff -urpN linux-2.6/scripts/recordmcount.pl linux-2.6-patched/scripts/recordmcount.pl
--- linux-2.6/scripts/recordmcount.pl 2011-05-16 10:16:20.000000000 +0200
+++ linux-2.6-patched/scripts/recordmcount.pl 2011-05-16 10:16:20.000000000 +0200
@@ -278,6 +278,7 @@ if ($arch eq "x86_64") {
} elsif ($arch eq "ia64") {
$mcount_regex = "^\\s*([0-9a-fA-F]+):.*\\s_mcount\$";
+ $mcount_adjust = -18;
$type = "data8";
if ($is_module eq "0") {
--
blue skies,
Martin.
"Reality continues to ruin my life." - Calvin.
--
To unsubscribe from this list: send the line "unsubscribe linux-arch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
