Re: [PATCH v2 0/7] blktrace: last round of fixes

[Ingo Molnar <mingo@elte.hu>]

* Li Zefan <lizf@cn.fujitsu.com> wrote:

> This is the last round of fixes for blktrace, I hope.  :) 
> 
> Changelog v1 -> v2:
>   - fix a bug in "blktrace: make classic output more classic",
>     and rebase other patches.
>   - re-order patches.
>   - a new patch for filtering BLK_TN_MESSAGE
> 
> 
> [PATCH v2 1/6] trace: make 'mem' argument of trace_seq_putmem() const
> [PATCH v2 2/6] blktrace: make classic output more classic
> [PATCH v2 3/6] blktrace: fix blk_probes_ref chaos
> [PATCH v2 4/6] blktrace: fix memory leak when freeing struct blk_io_trace
> [PATCH v2 5/6] blktrace: extract duplidate code
> [PATCH v2 6/6] blktrace: print out BLK_TN_MESSAGE properly

thanks Li!

I think we can do the following ordering. First apply the fixes:

> [PATCH v2 1/6] trace: make 'mem' argument of trace_seq_putmem() const
> [PATCH v2 3/6] blktrace: fix blk_probes_ref chaos
> [PATCH v2 4/6] blktrace: fix memory leak when freeing struct blk_io_trace

These mostly relate to mixed use of ioctl and ftrace based block 
tracing - which is rare but indeed possible. The fixes and cleanups 
you did to refcounting look good.

These bits:

> [PATCH v2 2/6] blktrace: make classic output more classic
> [PATCH v2 5/6] blktrace: extract duplidate code
> [PATCH v2 6/6] blktrace: print out BLK_TN_MESSAGE properly

are optional improvements. The BLK_TN_MESSAGE filter addition - 
Jens, are you fine with that feature addition? (it certainly looks 
like a nice thing to me.)

Also, note that we'll get even more in-kernel filtering capabilities 
in blktrace if we convert the tracepoints to TRACE_EVENT().

Reviewed-by: Ingo Molnar <mingo@elte.hu>

	Ingo