|
From: J.P. L. <pi...@el...> - 2003-09-11 07:35:35
|
Hello, I own a Network Station 1000, model 8362-A53, with 64MB of RAM and no PC Card carrier installed. There is a module in the cache slot, however the boot monitor reports L2 cache failure in the "view hardware configuration" screen. It's running version 3.1.1.0 of the boot monitor. I'm new to the list, and by no means a kernel hacker--just a disclaimer. There are a few problems I've had with running Linux on the NS1000 that I'll describe in detail. Overall, I'm impressed with the quality of this machine, even for its low price (about $20-something on eBay, _after_ shipping), and appreciate the work you've all done to make Linux work on these systems. This message is a bit long (sorry!), so I've broken it into sections roughly reverse-ordered by importance. 2.4.21 will not boot ~~~~~~~~~~~~~~~~~~~~ I've been able to use both 2.4.19 kernels at http://networkstation.aytolacoruna.es/. However, the 2.4.21 kernel does not work. The system prints this: dev fn boot deviceid cmd_st bar0 bar1 bar2 0x0a 0 0x00 00371014 02000006 00000000 00000000 00000000 ibm 27-826= 60 0x0b 0 0x08 000a1014 04000147 00100021 00000000 00000000 ibm 27-823= 76 0x0c 0 0x10 20001022 02800005 00100001 04000000 00000000 amd pcnet 0x0e 0 0x20 89015333 02000003 00000000 00000000 00000000 s3 trio64v= 2/dx =20 reconfiguring frame buffer to 0xcc000000 =20 IBM NetworkStation 1000 =20 total dram: 00000000 04000000 loaded at: 00800000 008BE86C board data at: 008AE078 008B4A84 relocated to: 008B4B84 008BB590 zimage at: 0080A230 008ADC2E avail ram: 00400000 00800000 =20 Linux/PPC load: console=3DttyS0,9600 console=3Dtty0 root=3D/dev/nfs nfsro= ot=3D/NetworkStationV2/prodbase/linuxppc ip=3Ddhcp Uncompressing Linux...done. Now booting the kernel And it does not proceed any further. (Not even on the serial console, which is how I copied those messages.) I've managed to install Debian using one of the 2.4.19 kernels, and I've compiled my own 2.4.19 kernel with the 2003/06/25 patch on networkstation.sf.net. It seems the binary 2.4.19 kernels at aytolacoruna.es are older than the patch: booting my own kernel resulted in being able to change framebuffer modes and use XFree86 (version 4.1.0.1; the stock version that comes with Debian woody). I compiled a 2.4.21 kernel with the 2003/07/15 NS1000 patch, and it failed the same way the 2.4.21 kernel at aytolacoruna.es fails. Since I changed the configuration a bit from my 2.4.19 kernel, and also had preemptible kernel and NFS-swap patches added (the corresponding older versions of those patches did work with my 2.4.19 kernel), I re-applied the NS1000 patch to a clean 2.4.21 tree, and used the exact same .config from my 2.4.19 kernel (after running 'make oldconfig'); booting this new 2.4.21 kernel failed, too. It printed some slightly different information on boot. If it matters any, here's a diff (- is from the aytolacoruna.es kernel, and + is from mine): total dram: 00000000 04000000 -loaded at: 00800000 008BE86C +loaded at: 00800000 008E486C -board data at: 008AE078 008B4A84 +board data at: 008D4078 008DAA84 -relocated to: 008B4B84 008BB590 +relocated to: 008DAB84 008E1590 -zimage at: 0080A230 008ADC2E +zimage at: 0080A200 008D31C8 avail ram: 00400000 00800000 Any known problems booting 2.4.21 on the NS1000? Anyone else had this issue? Shifted Video ~~~~~~~~~~~~~ I noticed that in my custom-compiled 2.4.19 kernel, the video is shifted to the right by about 15-30 pixels (a wild guess) compared to the aytolacoruna.es 2.4.19 kernel and compared to using the monitor with a completely different system, in the same video mode (which is 1024x768/60Hz). I'd like to know if that's correctable without using the monitor controls. I seem to recall some talk on the lists about imprecise video timing problems; is that relevant? Bad Video (?) ~~~~~~~~~~~~~ I've also noticed that running XFree86, the image quality seems odd. I can't put my finger on it; perhaps it's slightly fuzzy, perhaps there's ghosting. I tried creating some contrasting vertical patterns in The Gimp to create ghosting, but the result wasn't very bad. The same monitor in the same mode seems to produce a better image on a different system. Is this a problem others have noticed? Could it be because of the NS1000's display hardware, and/or caused by the same problem that causes the image to shift to the right? I'm running with 16-bit color, if that makes a difference. Can't Reboot ~~~~~~~~~~~~ Linux can't seem to reboot the machine, both when going into runlevel 6, and when there's a kernel panic followed by a message to the effect of "Reboot in 180 seconds". Just a minor annoyance, of course. (I certainly don't feel bad about power-cycling a system with no moving parts.) With an aytolacoruna.es 2.4.19 kernel, it just hangs. With my own 2.4.19 kernel with preemptible kernel and NFS-swap patches, the kernel will print a message like: <4>Oops: kernel access of bad area, sig: 11 NIP: FFF019AC XER: 00000000 LR: FFF00178 SP: C100FE10 REGS: c100fd60 TRAP= : 0400 Not tainted MSR: 20000030 EE: 0 PR: 0 FP: 0 ME: 0 IR/DR: 11 TASK =3D c100e000[458] 'reboot' Last syscall: 88 last math c100e000 last altivec 00000000 [omitting 32-bit hex numbers are probably not useful to others] Call backtrace: C100FE38 note: reboot[458] exited with preempt_count 1 And print this repeatedly: RPC: sendmsg returned error 101 nfs: RPC call returned error 101 The exact result varies. (Before I run this into ksymoops and make a proper bug report, I should at least see which patch is causing the oops.) Problems with i386 X Clients ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Several programs compiled for i386 don't work correctly over the network: Eterm and aterm with tinted transparency enabled display bogus colors in the background; and xscreensaver with the fade option enabled crashes on activation. After some testing I realized these problems also occur with programs on i386 systems remotely using XFree86 on my Power Mac also running Debian. For now I've reported the xscreensaver bug to the Debian BTS, and will hopefully do so with aterm and Eterm (or a library they might share that causes the problem (?)). Just a heads-up for anyone else that's had this problem and couldn't figure out what was wrong. Documentation ~~~~~~~~~~~~~ Getting started with the NS1000 was a bit of a pain. There weren't very many sources of information about these machines in general, let alone the port of Linux to them. I hope to find the time/motivation to write some FAQ or HOWTO material, based on my experiences so far, for either the WikiWiki at http://howto.ipng.be/NetstationWiki/, and/or the Netstation HOWTO, and/or my (as of yet non-existent) own personal web site. I'm soliciting suggestions on content to write about and where I should send it to. Offtopic/Misc ~~~~~~~~~~~~~ Anyone know where I can buy those PC Card carriers? Is it possible to boot Linux from them and otherwise use them? That'd be great for securely booting the NCs (eg, not relying on a crackable NFS server) and using X11 with IPsec transport-mode. It's been a goal of mine to set up something like this for quite some time. Are there any other discussion lists relating to NS1000s? Does anyone have archives of this list? I Googled for a phrase from one message, and apparently SourceForge's own mail archive is the only public archive. About a week ago, when I was browsing the archive, links would often lead to pages that say "forum not found". Sometimes reloading the page a few hours later worked, sometimes it didn't. The problem seems to be cleared up at this point, though. --=20 - J.P. Larocque, <pi...@el...>, <pi...@th...> - The attachment is a PGP sig, not a "virus"; you don't have to open it. - Encrypted e-mail preferred; 0x0c14cdda; http://ely.ath.cx/~piranha/pgp - Join the Army, meet interesting people, kill them. |