[534] in SIPB_Linux_Development
Re: Linix Demo at See It, Try It, Get It
daemon@ATHENA.MIT.EDU (Erik Nygren)
Tue Mar 29 12:36:24 1994
To: mlbarrow@MIT.EDU (Michael L. Barrow)
Cc: Erik Nygren <nygren@MIT.EDU>, joanne@MIT.EDU, dennis@MIT.EDU,
chris@MIT.EDU, rita@MIT.EDU, jwl@MIT.EDU, mlc@MIT.EDU,
linux-dev@MIT.EDU
In-Reply-To: Your message of Tue, 29 Mar 94 10:13:46 -0500.
<9403291513.AA28601@sunshine.MIT.EDU>
Date: Tue, 29 Mar 94 12:35:49 EST
From: Erik Nygren <nygren@MIT.EDU>
>> I will get back to you about this stuff, but I would like to know why you
>> make the following request in regards to the AT-1700:
>> > * You said you'll have a machine we can use to install Linux on.
>> > What are its specs? (disk space needs to be ~150 MB empty partition,
>> > ram needs to be >= 8MB, video card can be almost any standard thing,
>> > network card is preferably a 3c509 (but can be anything that
>> > is compatable and preferably not an AT-1700).
I'm sure it is possible to get the AT-1700 to work and I'm sure there
are many people out there who DO have it working. It just takes far
more effort and is more risky than using a 3c509. Things broke even
further with Linux 1.0 for some people and started working better for
others. For the sake of a demo, it would be much easier to set up
with something that is more likely to work.
The 3c509 works out of the box with Linux 1.0. The transmit timeout
problems that used to plague it have been fixed and everything seems
to work pretty much flawlessly. My machine has now been up for 11 days
with about 600 ftp server connections and about 2400 web server connections
without having any problems (except one stray transmit timeout a few
days ago, probably due to noise on the line). In the past,
a single connection to the ftp server would generate lots of transmit
timeouts and the kernel would hang once every few days (due to this I think).
Thanks to Donald Becker for fixing these problems.
The following is based on second hand information and may or not
still be true:
On the other hand, the AT-1700 support actually seems to have gotten
worse in Linux 1.0. A number of bug fixes were included (including
one to support IRQ's above 9). There are people who have gotten it to
work with few problems. However, I have recieved at least three
reports from people who have tried the newer driver and have had it do
very bad things. These people have sent some mail to Donald Becker
but I have not heard of any fixes yet. Basically, the driver installs
and works well the first time. However, after a few hours of use it
hangs. When the machine is power cycled (the at-1700's have always
needed to be power cycled and not simply reset when they hang), the
card is not recognized by Linux on the next bootup. For some of the
people, no amount of power cycling, twiddling with setup17 and the
IRQ's and ports, reinstalling the driver, or anything will cause Linux
to recognize the card again. One person reported that reverting to
the old kernel caused things to somewhat work, but others have
reported that this doesn't help either. Another person reported that
simply reinstalling LILO (the dual booter) seems to fix the problem
for one session, but things then go back to not working after the next
reboot until LILO is again reinstalled (?!)
One person reports that changing the BIOS signature checking
stuff (as was done with the older driver in the Linux locker)
causes things to work, but I don't know if this is the case
universally.
BTW, I don't think there's a big performance difference between the two.
The qualitative performance seems to be about the same (the 3c509
driver does not do predictive interrupts, so it really has no real
speed advantage).
--- Erik Nygren
