fdfile leak; vmstat output error (signed problem?)

[Thomas Klausner <wiz%NetBSD.org@localhost>]

Hi!

I stumbled over the fd leak problem. My bulk build took much longer
than expected, so I tried to find out why and saw that it was swapping
madly.

From dmesg:
total memory = 2046 MB
avail memory = 1966 MB

From top:
Memory: 87M Act, 43M Inact, 6008K Wired, 6464K Exec, 26M File, 1320K Free

The kernel process didn't look that big:
    0 root     125    0     0K   14M schedule  13:33  0.00%  0.00% [system]

vmstat -m | grep fdfile said:
fdfile        64 26758557    0     4544 424667    0 424667 424667   0   inf    0
The patch from PR 38374 improves the situation (see there).

However, vmstat output didn't look correct either:
 procs    memory      page                       disks   faults      cpu
 r b w    avm    fre  flt  re  pi   po   fr   sr c0 c1   in   sy  cs us sy id
 1 0 0 538004    412 -6017 131 93  315  518  884  0  0  450 10602 2001 51 20 29

Look at the flt column.

That's on amd64/4.99.58 from Apr 6 around 2100 CEST.
 Thomas