Well this patch set does cause gdb to no longer be able to generate
stack traces from signal handlers, but that just means gdb needs to be
fixed. We will work on that next.
libgcc can unwind through signal handlers both with and without the patch.
David Daney
On 02/18/2010 04:13 PM, David Daney wrote:
This patch set creates a vdso and moves the signal
trampolines to it from their previous home on the stack.
In the original patch set:
http://www.linux-mips.org/cgi-bin/mesg.cgi?a=linux-mips&i=49EE3B0F.3040506%40caviumnetworks.com
I stated:
Tested with a 64-bit kernel on a Cavium Octeon cn3860 where I have the
following results from lmbench2:
Before:
n64 - Signal handler overhead: 14.517 microseconds
n32 - Signal handler overhead: 14.497 microseconds
o32 - Signal handler overhead: 16.637 microseconds
After:
n64 - Signal handler overhead: 7.935 microseconds
n32 - Signal handler overhead: 7.334 microseconds
o32 - Signal handler overhead: 8.628 microsecond
All that is still true.
Improvements from the first version:
* Compiles and runs in 32-bit kernels (on qemu at least).
* Updated for linux-queue based 2.6.33-rc8
David Daney (3):
MIPS: Add SYSCALL to uasm.
MIPS: Preliminary vdso.
MIPS: Move signal trampolines off of the stack.
arch/mips/include/asm/abi.h | 6 +-
arch/mips/include/asm/elf.h | 4 +
arch/mips/include/asm/mmu.h | 5 +-
arch/mips/include/asm/mmu_context.h | 2 +-
arch/mips/include/asm/processor.h | 11 +++-
arch/mips/include/asm/uasm.h | 1 +
arch/mips/include/asm/vdso.h | 29 +++++++++
arch/mips/kernel/Makefile | 2 +-
arch/mips/kernel/signal-common.h | 5 --
arch/mips/kernel/signal.c | 86 ++++++---------------------
arch/mips/kernel/signal32.c | 55 ++++-------------
arch/mips/kernel/signal_n32.c | 26 ++------
arch/mips/kernel/syscall.c | 6 ++-
arch/mips/kernel/vdso.c | 112 +++++++++++++++++++++++++++++++++++
arch/mips/mm/uasm.c | 19 +++++-
15 files changed, 226 insertions(+), 143 deletions(-)
create mode 100644 arch/mips/include/asm/vdso.h
create mode 100644 arch/mips/kernel/vdso.c
