If you experience time outs with via-rhine (typically while copying large
files over the network), this patch is for you.
The problem with at least some of the Rhine chips is that they are skipping
one frame when they stop the Tx engine. This means that every time an abort
(due to excessive collisions) occurs, the driver will stall and need a
reset via netdev timeout [1]. The patch addresses that by setting the Tx
ring buffer pointer to where the driver thinks the chip should continue.
Note: this fix is not sufficient for all time out cases, but it should
improve the situation without having negative side effects for anyone. I
have a more comprehensive solution in my development tree, for now I am
curious how many of the time outs people keep reporting persist after this
one is fixed.
Patch is against Jeff's via-rhine.c LK 1.1.14. It should work with the
older version in 2.4.19-pre9 as well.
Roger
[1] The same problem will likely happen with any other Tx error (e.g.
underrun), abort is just the one I can easily produce here.
It also seems that the interrupt status is not always reliable. I
have one report of very mysterious behavior with a VT86C100A which
seems to produce all kinds of obviously bogus errors. That's
unrelated though and easier to study once the restart problems
are fixed.
This archive was generated by hypermail 2b29 : Fri Jun 07 2002 - 22:00:10 EST