Re: [9fans] [7f Bugfix] Plan9 FS Realtek Driver

Re: [9fans] [7f Bugfix] Plan9 FS Realtek Driver

[Russ Cox]

Thanks for the report.  Geoff Collyer sent in
a slightly different fix that works within the compat.h
macros, and that fix is now on sources.

Russ

Re: [9fans] [7f Bugfix] Plan9 FS Realtek Driver

[north_]

> Thanks for the report.
My pleasure

> Geoff Collyer sent in
> a slightly different fix that works within the compat.h
> macros, and that fix is now on sources.
Yah, I was so excited that I had created a solid fix for
the driver that I didn't take the time to elaborate and see
if I could go further via compat.h, etc. I'm just glad
that my fix works well. Finally, I am on my way to being
as good with hardware as I am with software!
Nice job, Geoff.
Don

[9fans] [7f Bugfix] Plan9 FS Realtek Driver

[north_]

Hello, all!
    I hope everyone is enjoying their winter as much as I am! Our
small coat of snow is making for great photography, here.
    The following summarizes my general work on a bug that renders
the realtek ethernet card unusable on the plan9 file server
kernel. The bug, in the realtek driver, causes any packet read
from the network to be discarded.
    I was finally able to run an entire plan9 network (cpu, auth,
term, fs) due to the amiable donation of three workstations from
Countrywide Financial. Despite this much appreciated gift, I
found myself unable to run a file server due to a lack of ethernet
cards (and a distrust of the realtek 8139). After reading through
the fs source code, I realized that the realtek *should* work
after a few quick fixes to the driver. Once these fixes were
solidified in a kernel build, I have found the card to (surprisingly)
be very stable over the past week. Latency is low, data corruption
has not been evident, etc, so, I feel very confident that this fix
will work well with anyone utilizing the realtek. Aside from normal
fs usage, I (of course) had to pull the entire ISO over the network
to create the initial file system. I used this as a base stress test.
Large files and large clusters of small files were also shuttled
back and forth to test data corruption. Normal network traffic tests
show no problems, either. Thanks to all of the Bell Labs crew that
gave me pointers on hardware.
    The bugfix can be located at:
        http://www.7f.no-ip.com/bugfix/7f20021203-1.tgz
Don
One Twenty Seven Research and Development