Log in

No account? Create an account
Pete Zaitcev's Journal -- Day [entries|friends|calendar]
Pete Zaitcev

[ userinfo | livejournal userinfo ]
[ calendar | livejournal calendar ]

[26 Dec 2004|01:21am]
Bad news on USB-in-2.4 front. Marcelo's 2.4.29-pre3 does not want to work on the PE750 I use for testing. This was the case at least since 2.4.27, and I have no idea where it began. The kernel throws symptoms of memory corruption with random oopses which have vague relationship with USB: they always happen after usb is started. However, the RHEL 3 works without any problems. I spent about three days diffing and patching only to find out that after all no changes in drivers/usb/ between 2.4.21-20.EL and 2.4.29-pre3 are responsible. So, it's not USB per se, but something else which affects USB. Both RHEL and Marcelo trees run without ACPI. I tried Early USB Handoff, too.

The patch between working and non-working is 3.5MB. Another bad part is that the box requires latest SATA to boot (with ____request_resource), so stepping down to older releases of Marcelo tree is going to be difficult. But I'm just about ready to start doing it.

I started to poke at the box because I found a bug in my last batch of 2.4 changes, codenamed "Little rework of usbserial". But now without a test box I cannot get them rolled out. It is possible to throw them out and plead for testing, but I know better. So, I'm in a bind: I need a test SMP box to work with Marcelo tree before I can do anything at all, and I have work piling up.

I am pondering breaking out the ancient P5 SMP which I packed for eBaying.

It's rather irritating that nobody else noticed the Dell PE750 disaster with 2.4, but I guess everyone either runs RHEL 3 on it, or switched to 2.6 long ago, because the thing won't boot stock trees due to the SATA. How unfortunate that I cannot just off-load this headache on the faceless community and just expect it fixed.
post comment

[ viewing | December 26th, 2004 ]
[ go | previous day|next day ]