Prev: [PATCH] perf: fix cmpxchg warning in perf_event_amd.c
Next: v4 x86/mrst: add more timer config options
From: Mathieu Desnoyers on 18 May 2010 17:00
* Srikar Dronamraju (srikar(a)linux.vnet.ibm.com) wrote:
> Uprobes Patches
>
> Changelog from v3:
> - Reverted to background page replacement as suggested by Peter Zijlstra.

Why ?

I'm not sure we reached any concensus about the need for a background page
replacement.

Thanks,

Mathieu

> - Dso in 'perf probe' can be either be a short name or a absolute path.
> - Addressed comments from Masami, Frederic, Steven on traceevents and perf
>
> Changelog from v2:
> - Addressed comments from Oleg, including removal of interrupt context
> handlers, reverting background page replacement in favour of
> access_process_vm().
>
> - Provides perf interface for uprobes.
>
> Changelog from v1:
> - Added trace_event interface for uprobes.
> - Addressed comments from Andrew Morton and Randy Dunlap.
>
> For previous postings: please refer: http://lkml.org/lkml/2010/5/6/273
> and http://lkml.org/lkml/2010/3/20/107
>
> This patchset implements Uprobes which enables you to dynamically break
> into any routine in a user space application and collect information
> non-disruptively.
>
> This patchset is a rework based on suggestions from discussions on lkml
> in January and March this year (http://lkml.org/lkml/2010/1/11/92,
> http://lkml.org/lkml/2010/1/27/19, http://lkml.org/lkml/2010/3/20/107
> and http://lkml.org/lkml/2010/3/31/199 ). This implementation of
> uprobes doesnt depend on utrace.
>
> When a uprobe is registered, Uprobes makes a copy of the probed
> instruction, replaces the first byte(s) of the probed instruction with a
> breakpoint instruction. (Uprobes uses background page replacement
> mechanism and ensures that the breakpoint affects only that process.)
>
> When a CPU hits the breakpoint instruction, Uprobes gets notified of
> trap and finds the associated uprobe. It then executes the associated
> handler. Uprobes single-steps its copy of the probed instruction and
> resumes execution of the probed process at the instruction following the
> probepoint. Instruction copies to be single-stepped are stored in a
> per-process "execution out of line (XOL) area". Currently XOL area is
> allocated as one page vma.
>
> Advantages of uprobes over conventional debugging include:
>
> 1. Non-disruptive.
> Unlike current ptrace based mechanisms, uprobes tracing wouldnt
> involve signals, stopping threads and context switching between the
> tracer and tracee.
>
> 2. Much better handling of multithreaded programs because of XOL.
> Current ptrace based mechanisms use single stepping inline, i.e they
> copy back the original instruction on hitting a breakpoint. In such
> mechanisms tracers have to stop all the threads on a breakpoint hit or
> tracers will not be able to handle all hits to the location of
> interest. Uprobes uses execution out of line, where the instruction to
> be traced is analysed at the time of breakpoint insertion and a copy
> of instruction is stored at a different location. On breakpoint hit,
> uprobes jumps to that copied location and singlesteps the same
> instruction and does the necessary fixups post singlestepping.
>
> 3. Multiple tracers for an application.
> Multiple uprobes based tracer could work in unison to trace an
> application. There could one tracer that could be interested in
> generic events for a particular set of process. While there could be
> another tracer that is just interested in one specific event of a
> particular process thats part of the previous set of process.
>
> 4. Corelating events from kernels and userspace.
> Uprobes could be used with other tools like kprobes, tracepoints or as
> part of higher level tools like perf to give a consolidated set of
> events from kernel and userspace. In future we could look at a single
> backtrace showing application, library and kernel calls.
>
> Here is the list of TODO Items.
>
> - Rebase to -tip tree. (targeted for v5)
> - Allowing probes across fork.
> - Allowing probes per-executable/per dso.
> - Allow multiple probes to share a probepoint.
> - Return probes.
> - Support for other architectures.
> - Uprobes booster.
> - Merge uprobes and kprobes trace_event.
> - replace macro with bits in inat table.
>
> The current patchset is based on 2.6.34.
>
> Please do provide your valuable comments.
>
> Thanks in advance.
> Srikar
>
> Srikar Dronamraju (10):
> 1. X86 instruction analysis: Move Macro W to insn.h
> 2. mm: Move replace_page() to mm/memory.c
> 3. user_bkpt: User Space Breakpoint Assistance Layer
> 4. user_bkpt: X86 details for User space breakpoint assistance
> 5. user_bkpt: Slot allocation for Execution out of line
> 6. uprobes: Uprobes Implementation
> 7. uprobes: X86 details for Uprobes
> 8. samples: Uprobes samples
> 9. uprobes: Uprobes Documentation patch
> 10. trace: Share common code for uprobes/kprobes traceevents
> 11. trace: uprobes trace_event interface
> 12. perf: Dont adjust symbols if lookup is by name.
> 13. perf: perf interface for uprobes.
>
> Documentation/uprobes.txt | 236 ++++++++++
> arch/Kconfig | 31 ++
> arch/x86/Kconfig | 2 +
> arch/x86/include/asm/insn.h | 7 +
> arch/x86/include/asm/thread_info.h | 2 +
> arch/x86/include/asm/user_bkpt.h | 43 ++
> arch/x86/kernel/Makefile | 3 +
> arch/x86/kernel/kprobes.c | 7 -
> arch/x86/kernel/signal.c | 17 +
> arch/x86/kernel/uprobes.c | 77 ++++
> arch/x86/kernel/user_bkpt.c | 572 ++++++++++++++++++++++++
> fs/exec.c | 4 +
> include/linux/mm.h | 4 +
> include/linux/mm_types.h | 4 +
> include/linux/sched.h | 4 +
> include/linux/uprobes.h | 169 ++++++++
> include/linux/user_bkpt.h | 305 +++++++++++++
> include/linux/user_bkpt_xol.h | 40 ++
> kernel/Makefile | 3 +
> kernel/fork.c | 20 +
> kernel/trace/Kconfig | 13 +
> kernel/trace/Makefile | 1 +
> kernel/trace/trace.h | 11 +
> kernel/trace/trace_kprobe.c | 76 +----
> kernel/trace/trace_probe.h | 111 +++++
> kernel/trace/trace_uprobe.c | 839 ++++++++++++++++++++++++++++++++++++
> kernel/uprobes.c | 681 +++++++++++++++++++++++++++++
> kernel/user_bkpt.c | 589 +++++++++++++++++++++++++
> kernel/user_bkpt_xol.c | 302 +++++++++++++
> mm/ksm.c | 112 -----
> mm/memory.c | 120 +++++
> samples/Kconfig | 7 +
> samples/Makefile | 2 +-
> samples/uprobes/Makefile | 17 +
> samples/uprobes/uprobe_example.c | 83 ++++
> tools/perf/builtin-probe.c | 38 ++-
> tools/perf/builtin-top.c | 20 -
> tools/perf/util/event.c | 20 +
> tools/perf/util/event.h | 1 +
> tools/perf/util/probe-event.c | 237 +++++++++--
> tools/perf/util/probe-event.h | 9 +-
> tools/perf/util/probe-finder.h | 1 +
> tools/perf/util/symbol.c | 6 +-
> 43 files changed, 4576 insertions(+), 270 deletions(-)

--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Srikar Dronamraju on 19 May 2010 02:10
> * Srikar Dronamraju (srikar(a)linux.vnet.ibm.com) wrote:
> > Uprobes Patches
> >
> > Changelog from v3:
> > - Reverted to background page replacement as suggested by Peter Zijlstra.
>
> Why ?
>
> I'm not sure we reached any concensus about the need for a background page
> replacement.
>
> Thanks,
>
> Mathieu
>

I was tested with access_process_vm(previous patchset) and the current
(background page replacement) and the results are the same.

However Peter Zijlstra's contention still stands that we might be
relying on a undocumented feature in the hardware and the flipping the
pages isnt that hard or expensive.

Even Linus, (http://lkml.org/lkml/2010/1/27/87) had shown inclination
towards background page replacement.

Also when uprobes implements global tracing support (i.e probing a
particular symbol in a dso across processes), it has to rely on
background page replacement.

Hence I based this patchset on background page replacement rather than
on access_process_vm. Later On, if there is a consensus on using
access_process_vm, we can make the corresponding changes.

--
Thanks and Regards
Srikar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Mathieu Desnoyers on 19 May 2010 10:20
* Srikar Dronamraju (srikar(a)linux.vnet.ibm.com) wrote:
> > * Srikar Dronamraju (srikar(a)linux.vnet.ibm.com) wrote:
> > > Uprobes Patches
> > >
> > > Changelog from v3:
> > > - Reverted to background page replacement as suggested by Peter Zijlstra.
> >
> > Why ?
> >
> > I'm not sure we reached any concensus about the need for a background page
> > replacement.
> >
> > Thanks,
> >
> > Mathieu
> >
>
> I was tested with access_process_vm(previous patchset) and the current
> (background page replacement) and the results are the same.
>
> However Peter Zijlstra's contention still stands that we might be
> relying on a undocumented feature in the hardware

The access_process_vm scheme is in many ways similar to what kprobes has been
doing for years. So I would not rely on that as a primary argument against
the access_process_vm approach.

> and the flipping the
> pages isnt that hard or expensive.
>
> Even Linus, (http://lkml.org/lkml/2010/1/27/87) had shown inclination
> towards background page replacement.
>
> Also when uprobes implements global tracing support (i.e probing a
> particular symbol in a dso across processes), it has to rely on
> background page replacement.

Ah OK. If you have to use page replacement for global tracing, I see that as a
good argument for using page replacement everywhere.

>
> Hence I based this patchset on background page replacement rather than
> on access_process_vm. Later On, if there is a consensus on using
> access_process_vm, we can make the corresponding changes.

Well, page replacement seems like a good way to support global tracing, so I
doubt that we'll ever revert back to access_process_vm.

Thanks,

Mathieu

>
> --
> Thanks and Regards
> Srikar

--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Peter Zijlstra on 19 May 2010 10:20
On Wed, 2010-05-19 at 11:38 +0530, Srikar Dronamraju wrote:
>
> Also when uprobes implements global tracing support (i.e probing a
> particular symbol in a dso across processes), it has to rely on
> background page replacement.

Uhm, how so?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Srikar Dronamraju on 19 May 2010 13:40
* Peter Zijlstra <peterz(a)infradead.org> [2010-05-19 16:17:40]:

> On Wed, 2010-05-19 at 11:38 +0530, Srikar Dronamraju wrote:
> >
> > Also when uprobes implements global tracing support (i.e probing a
> > particular symbol in a dso across processes), it has to rely on
> > background page replacement.
>
> Uhm, how so?

For global tracing we cant use access_process_vm. The equivalent to
access_process_vm for global tracing would be to use copy_to_user.

However I am sure people wouldnt be happy using copy_to_user on a page
that other processes could have mapped.

So that leaves us with background page replacement method.
I do agree that we will not be using the same routine. But the technique
should be something similar(or do you disagree?).

Or do you suggest any alternative methods for global tracing?

--
Thanks and Regards
Srikar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 |  Next  |  Last
Pages: 1 2
Prev: [PATCH] perf: fix cmpxchg warning in perf_event_amd.c
Next: v4 x86/mrst: add more timer config options