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

[Rick Lindsley]

    I've done some disk-io (compiled some software) and the values
    now are even lower:
    
Interesting, yes, you now have 60 fewer "I/O's in flight."

    Where exactly is the code in the kernel, that produces the diskstats file,
    so I can try grepping through it?

The I/O's in flight are incremented in drive_stat_acct() (see ll_rw_blk.c)
and decremented in attempt_merge() and end_that_request_last().  My gut feel
is that end_that_request_last() is getting called more often than expected
via some interesting path, but I've nothing right now to back that up.

Rick