Re: [Qemu-devel] [PATCH v3 00/11] simplify usage of tracepoints, and connect them to logging

["Denis V. Lunev" <den@openvz.org>]

On 11/03/2015 08:33 PM, Paolo Bonzini wrote:
>
> On 03/11/2015 18:10, Stefan Hajnoczi wrote:
>> On Thu, Oct 29, 2015 at 11:44:06PM +0300, Denis V. Lunev wrote:
>>> This series does three things:
>>>
>>> 1) add a "-trace [enable=]foo" option to enable one or more
>>> trace events, and a "-trace help" option to show the list of
>>> tracepoints (patches 4-5)
>>>
>>> 2) change the stderr tracing backend so that it prints to the -D
>>> log file, and enable it by default.  "-trace file=..." is now a
>>> synonym of -D if the log backend is enabled (patches 7-8)
>>>
>>> 3) add a "-d trace:foo" option that is a synonym for "-trace
>>> foo"; this makes the new functionality more discoverable to
>>> people used to "-d", makes it available for user-mode emulation
>>> (which does not have -trace), and is somewhat nice if you want to
>>> enable both tracepoints and some other "-d" flag (patch 9).  When
>>> globbing it is also less susceptible to unwanted shell
>>> expansion.
>>>
>>> For example, you can trace block device I/O and save the result
>>> to a file just by adding "-trace bdrv_aio_*,file=trace.txt", or
>>> correlate it to guest PCs with "-d
>>> exec,nochain,trace:bdrv_aio_*".
>>>
>>> Opinions?  I would like to have this in 2.5 if there is
>>> agreement.
>>>
>>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by:
>>> Denis V. Lunev <den@openvz.org>
>>>
>>> Changes from v2: - compilation fix was extended to patch 2 and 3
>>> - replaced Reviewed-by with Acked-by by request from Christian
>>>
>>> Changes from v1: - small cleanup to vl.c is added as patch (4) -
>>> compilation is fixed in patch (2) - moved qemu-log.c to
>>> util/log.c to fix linking of qemu-io/qemu-nbd
>>>
>>> Denis V. Lunev (2): trace: no need to call trace_backend_init in
>>> different branches now log: move qemu-log.c into util/ directory
>>>
>>> Paolo Bonzini (9): trace: fix documentation trace: split
>>> trace_init_events out of trace_init_backends trace: split
>>> trace_init_file out of trace_init_backends trace: add "-trace
>>> enable=..." trace: add "-trace help" log: do not unnecessarily
>>> include qom/cpu.h trace: convert stderr backend to log trace:
>>> switch default backend to "log" log: add "-d trace:PATTERN"
>>>
>>> Makefile.objs                                   |  1 -
>>> bsd-user/main.c                                 |  1 + configure
>>> |  6 +- cpu-exec.c                                      |  1 +
>>> exec.c                                          |  1 +
>>> hw/acpi/cpu_hotplug.c                           |  1 +
>>> hw/timer/a9gtimer.c                             |  1 +
>>> include/exec/log.h                              | 60
>>> ++++++++++++++++ include/qemu/log.h
>>> | 60 +--------------- linux-user/main.c
>>> |  1 + qemu-io.c                                       |  2 +-
>>> qemu-options.hx                                 | 22 ++++--
>>> qom/cpu.c                                       |  1 +
>>> scripts/tracetool/backend/{stderr.py => log.py} |  9 +--
>>> target-alpha/translate.c                        |  1 +
>>> target-arm/translate.c                          |  1 +
>>> target-cris/translate.c                         |  1 +
>>> target-i386/seg_helper.c                        |  1 +
>>> target-i386/smm_helper.c                        |  1 +
>>> target-i386/translate.c                         |  1 +
>>> target-lm32/helper.c                            |  1 +
>>> target-lm32/translate.c                         |  1 +
>>> target-m68k/translate.c                         |  1 +
>>> target-microblaze/helper.c                      |  1 +
>>> target-microblaze/translate.c                   |  1 +
>>> target-mips/helper.c                            |  1 +
>>> target-mips/translate.c                         |  1 +
>>> target-moxie/translate.c                        |  1 +
>>> target-openrisc/translate.c                     |  1 +
>>> target-ppc/mmu-hash32.c                         |  1 +
>>> target-ppc/mmu-hash64.c                         |  1 +
>>> target-ppc/mmu_helper.c                         |  1 +
>>> target-ppc/translate.c                          |  1 +
>>> target-s390x/translate.c                        |  1 +
>>> target-sh4/helper.c                             |  1 +
>>> target-sh4/translate.c                          |  1 +
>>> target-sparc/int32_helper.c                     |  1 +
>>> target-sparc/int64_helper.c                     |  1 +
>>> target-sparc/translate.c                        |  1 +
>>> target-tilegx/translate.c                       |  1 +
>>> target-tricore/translate.c                      |  1 +
>>> target-unicore32/translate.c                    |  1 +
>>> target-xtensa/translate.c                       |  1 + tcg/tcg.c
>>> |  1 + trace/control.c                                 | 95
>>> ++++++++++++++++++------- trace/control.h
>>> | 42 ++++++++++- trace/simple.c
>>> |  6 +- trace/simple.h                                  |  4 +-
>>> translate-all.c                                 |  1 +
>>> util/Makefile.objs                              |  1 + qemu-log.c
>>> => util/log.c                        | 19 ++++- vl.c
>>> | 38 +++++----- 52 files changed, 274 insertions(+), 129
>>> deletions(-) create mode 100644 include/exec/log.h rename
>>> scripts/tracetool/backend/{stderr.py => log.py} (78%) rename
>>> qemu-log.c => util/log.c (90%)
>>>
>>> -- 2.1.4
>>>
>>>
>> Thanks, applied to my tracing tree:
>> https://github.com/stefanha/qemu/commits/tracing
> Thanks Denis and Stefan!  I hope this will make it simpler to track
> user problems, and will also encourage people to use tracepoints more.
>
> Paolo

There 2 small notes here:
- there is small patch with 'all' alias for all tracepoints which
   could also be useful in the scope of this work:
   [PATCH v2 1/1] allow to enable all tracepoints via alias all

- it would be great if patches from my log patchset will be merged.
   I'll port them tomorrow on top of this set (if will not work on
   findings in locking patchset)
   Switch from inline call to #define and QMP/QAPI logging seems
   really necessary. timestamp could be enabled or not but it is
   better to use unified approach for logging timestamps and
   tracing timestamps.

Den