Re: [2.5.70] possible problem with /dev/diskstats

[Rick Lindsley]

    Documentation/iostats.txt says about diskstats:
    [SNIP]
    Field  9 -- # of I/Os currently in progress
        The only field that should go to zero. Incremented as requests are
        given to appropriate request_queue_t and decremented as they finish.
    [SNIP]
    
    But here is a cat /proc/diskstats:
       3    0 hda 948 317 16216 4294408142 90 333 848 7309 4294967294 7309215 4280372198
                                                           ~~~~~~~~~~

Yes, I've had a couple of other reports of this.  I suspect there is
a path by which an "I/O" appears to have been completed while none was
begun.  I've only noticed this on my hda drive as well.  What I didn't
notice was exactly when this behavior began, which may have been useful
in tracking down the problem.  You'll notice a couple of other values
there in the 4.2 billion range that probably suffer from a similar
(or maybe the same) issue.  I haven't seen this on SCSI drives yet,
just hda drives, which suggests there may be something about that I/O
path that bears some closer scrutiny.

Rick