The fsinfo provider, much like the cpuinfo and vminfo providers, places
a probe wherever the new fs kstats are incremented in the fop_*()
functions. Using a new vnode_t -> fileinfo_t translator, you''ll
have
the associated file whenever such a probe fires, allowing you to track
VFS operations by filename, mountpoint, etc. This is just a simple
counter (although reads and writes give you number of bytes as well).
For more complicated queries, you''ll have to instrument the associated
fop_*() routine directly using FBT.
Moving forward, we''re looking to expand this to include
"done" probes of
some fashion, so that you can also calculate the latency of various
operations.
- Eric
On Sat, Apr 15, 2006 at 12:47:57PM -0400, Matty wrote:>
> Howdy,
>
> I came across the following note while perusing the B38 release notes:
>
> http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6405662
>
> What kind of goodies will be added to the fsinfo provider?
>
> - Ryan
> --
> UNIX Administrator
> http://daemons.net/~matty
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
--
Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock