[lkml]   [2009]   [Apr]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subject[PATCH 0/17] tracehook & user_regset for ARM
The various things listed under CONFIG_HAVE_ARCH_TRACEHOOK (arch/Kconfig)
are what each arch wants to do nowadays to fit with the generic code for
user debugging, core dumping, etc. The "big machine" arch's have done this
since 2.6.28 or earlier.

Christoph suggested helping ARM with the work would be a good way to
encourage all the "little machine" arch maintainers to catch up soon.
(Of the "little machine" and oddball arch's, so far only sh has it.)

The following patches bring ARM up to speed with HAVE_ARCH_TRACEHOOK (just
about). There are so many patches because I sliced them into many small
changes. Each patch is pretty short (some of them very tiny). The overall
diffstat from the whole series is attached in the "pull request" below.

This series is relative to ~2.6.30-rc3 (0c8454f). I expect it rebases
easily to whatever tree you might want to queue it on.

The immediate user-visible effects of the series are to enable the
/proc/pid/syscall feature, and to add VFP, WMMX, Crunch, and $tp register
data to core dumps.

AFAIK only the asm/syscall.h patch still needs work. The preliminary
version is only buggy in the way that /proc/pid/syscall will give bogus
answers for a task not really in a syscall, or for the non-EABI entry
styles. It's not unsafe or anything. It needs some attention from folks
who really know ARM to fill in the truly proper version of syscall_get_nr().

I only know how to run and test one ABI flavor, and only in qemu. I used
versatile_defconfig and ran it in qemu-system-arm -M versatilepb using NFS
root with the userland binaries from Fedora ARM.

I don't know how to simulate hardware that has iWMMXt or Crunch, nor if my
ARM userland handles those kernel configurations. So I've only
(cross-)compile-tested the iWMMXt and Crunch code. (It is however the
simplest of the user_regset code and pretty easy to eyeball-review.)

My testing is quite minimal. Booted, nothing went wrong, simple strace
uses still look sane, "cat /proc/self/syscall" looks right, core dump
contents look right. (Not knowing the arch at all, I don't actually know
how to put anything in the FPA or VFP registers so as to notice they are
right rather than just the right number of zeros.) The userland I have
does not have gdb (and qemu would take a week to build and run the gdb
testsuite if it did), so I didn't try to test any ptrace use beyond what
strace does.



The following changes since commit 0c8454f56623505a99463405fd7d5664adfbb094:
Rafael J. Wysocki (1):
PM/Hibernate: Fix waiting for image device to appear on resume

are available in the git repository at:

git:// arm/tracehook

Roland McGrath (17):
arm: arch_ptrace clean-up
arm: arch_ptrace indentation
arm: tracehook_report_syscall
arm: tracehook_signal_handler
arm: user_regset: general regs
arm: user_regset: FPU regs
arm: user_regset: VFP regs
arm: user_regset: VFP in core dumps
arm: user_regset: iWMMXt regs
arm: user_regset: iWMMXt in core dumps
arm: user_regset: Crunch regs
arm: user_regset: Crunch in core dumps
arm: user_regset: thread pointer in core dumps
arm: asm/syscall.h (unfinished)

arch/arm/Kconfig | 1 +
arch/arm/include/asm/elf.h | 1 +
arch/arm/include/asm/ptrace.h | 4 +-
arch/arm/include/asm/syscall.h | 65 +++++
arch/arm/include/asm/thread_info.h | 4 +
arch/arm/kernel/entry-common.S | 2 +-
arch/arm/kernel/ptrace.c | 526 ++++++++++++++++++++++++------------
arch/arm/kernel/signal.c | 7 +
include/linux/elf.h | 3 +
9 files changed, 433 insertions(+), 180 deletions(-)
create mode 100644 arch/arm/include/asm/syscall.h

 \ /
  Last update: 2009-04-25 02:11    [W:0.117 / U:0.056 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site