Date: Mon, 22 Sep 2003 12:09:17 +1000 From: Peter Jeremy <peter.jeremy@alcatel.com.au> To: Doug White <dwhite@gumbysoft.com> Cc: freebsd-stable@freebsd.org Subject: Re: kldload problem with 4.8-RELEASE-p7 on HP DL380 Message-ID: <20030922020917.GA91404@gsmx07.alcatel.com.au> In-Reply-To: <20030919092635.R45539@carver.gumbysoft.com> References: <20030919064556.GW91404@gsmx07.alcatel.com.au> <20030919092635.R45539@carver.gumbysoft.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2003-Sep-19 09:32:31 -0700, Doug White <dwhite@gumbysoft.com> wrote: >On Fri, 19 Sep 2003, Peter Jeremy wrote: > >> I have a HP DL380 with a DigiBoard Xem that I'm trying to use with >> FreeBSD 4.x. I've back-ported the digi device from -CURRENT as a >> kld but when I try to load it, the system wedges. Running exactly >> the same kernel on an older Compaq P1850 works. If I don't try to >> load digi.ko the system runs perfectly. > >Does it not work with the dgb driver? No, the dgb driver only supports the older ISA DigiBoards. >Can you try putting the digi in a different slot? From the traceback, its >sitting around in ataintr after going through intr_mux, so I wonder if >your backported digi and the ata driver are conflicting. That looks like it was the problem. IRQ15 was shared between ata1 (unused), bge1 (which I need) and digi0. I used the BIOS to change the IRQ for the DigiBoard and it seems much happier - it loads now but I still need to verify that it actually works. >Also have you tried building your driver into the kernel? When I first back-ported digi(4) (about 2 years ago), it had problems with assumptions about the system state and couldn't be built in. I just checked and it still dies (at least in -STABLE) because it needs to load firmware and linker_load_file() dies because the filesystem isn't setup at that point. It's easier to leave it as a kld than fix all these assumptions. >It seems to be stuck in a delay loop. Perhaps there's a bug in your >backport? -current and -stable are pretty different animals from an >interrupt perspective. It's been running for several years in older Proliants and the older kernel has the same failure in a DL380. It could be that there's a bug in the IRQ sharing that didn't bite me before but I think the general interupt behaviour is OK. >Also I'm suprised the ethernet on the dl380 isn't driving you nuts since >its the broken broadcom crap. So I haven't been imagining that the network performance is less than spectacular.... Running 100baseTX full-duplex, the throughput approximates what I would expect from 10baseT (and NFS has fits). Unfortunately, I didn't have a choice on the hardware. Luckily, network performance isn't an issue for us - as long as it can manage >10KB/sec it'll be OK for us. Thanks for your help. Peter
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030922020917.GA91404>