em0 is still broken - watchdog timeouts, lockups ... workaround please

Joe Schmoe non_secure at yahoo.com
Sun Mar 28 16:34:18 PST 2004


System is a very vanilla 4.9-RELEASE with fxp and em
built into the kernel.  It is a dual processor machine
with two 550mhz p3s.  Its kernel config is very sparse
- all USB and pcmcia, etc. have been removed - the
only items of note are symbios scsi support, and
support for twe (my 3ware IDE raid card).

The string acpi/ACPI does not appear in my kernel
config file, thanks for asking.

Boot up the machine using em0 ... it continually spits
out:

Jan  4 14:56:50 die /kernel: em0: Link is up 100 Mbps
Full Duplex
Jan  4 15:04:18 die /kernel: em0: Link is Down
Jan  4 15:04:20 die /kernel: em0: Link is up 100 Mbps
Full Duplex
Jan  4 25:59:06 die /kernel: em0: Link is Down

and if I do any kind of file transfer with it, I will
get flooded with watchdog timeout errors.  Oh, and the
machine will lock up at that point as well - no longer
pingable.

I am happy for a workaround - there are several posts
in the archives that mention:

changing  '#define EM_REPORT_TX_EARLY 0' to '#define
EM_REPORT_TX_EARLY 
1' in /usr/src/sys/dev/em/if_em.h

HOWEVER, my if_em.h does not contain that line, so it
is not possible to change it.  Further, as this is
4.9-RELEASE, I also do not have that tunable sysctl
for em0 either that some folks suggest should be set
to zero.

So .. is there a workaround ?  And further, is Intel
dragging their feet on releasing some information on
this card, or blocking freebsd-licensed versions of
drivers that fix this problem ?  I am trying to think
of some plausible reason as to why this has been
broken _for years_ and never fixed.  Search the
archives and you will find people complaining about
this (specifically on dual cpu systems) as far back as
2002.

Help appreciated.

__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html


More information about the freebsd-hardware mailing list