bnx(4) thread consumes 100%+ of CPU.

["John D. Baker" <jdbaker%mylinuxisp.com@localhost>]

I have an IBM System x3650 with 2 quad-core CPUs running amd64-7.1_STABLE.
It has lately started becoming sluggish.  The culprit is the system
process service thread for the "bnx0" interface (the one in use).
Partial output of 'top -nt':

$ top -nt
load averages:  2.30,  1.36,  1.40;               up 0+16:39:03        15:36:48
121 threads: 33 idle, 1 runnable, 78 sleeping, 9 on CPU
CPU states:  0.0% user,  0.0% nice, 15.7% system, 12.2% interrupt, 71.9% idle
Memory: 27G Act, 13G Inact, 16M Wired, 9800K Exec, 41G File, 229M Free
Swap: 48G Total, 48G Free

  PID   LID USERNAME PRI STATE      TIME   WCPU    CPU NAME      COMMAND
    0    60 root      25 CPU/3      4:55 99.17% 99.17% bnx0      [system]
    0     3 root     222 CPU/0      1:20 26.22% 26.22% softnet/0 [system]
    0    98 root     124 syncer/4  52:12 15.67% 15.67% ioflush   [system]
   34     1 sysop     83 select/5   1:09 11.47% 11.47% -         sshd
  423     5 root      85 nfsd/6     0:21  7.03%  7.03% slave     nfsd
    0    59 root      96 ipmicm/5   0:40  3.66%  3.66% ipmi      [system]
    0    32 root     220 IDLE/3     0:11  3.12%  3.12% softclk/3 [system]
[...]

Ordinarily, the bnx0 service thread never shows up, or is significantly
down the list, even when operating as a semi-diskless system.

It may run for weeks with no problem, but it has recently done this
twice in as many days.

So far, the only way to restore normal operation is to reboot the machine.
Bouncing the interface with "ifconfig bnx0 down ; ifconfig bnx0 up" has
no discernable effect.

Driver issue?  Hardware failing?  I could switch to its second interface,
"bnx1" and see if it does the same thing.

-- 
|/"\ John D. Baker, KN5UKS               NetBSD     Darwin/MacOS X
|\ / jdbaker[snail]mylinuxisp[flyspeck]com    OpenBSD            FreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645