Re: Sparc has slow disk access
Hi Ben, You don't use gcc-2.96 or gcc-3.0 for compiling sparc64 kernels. The package is egcs64. The 2.4.7 Debian kernels should be pretty recent, if you are still not comfortable compiling them. Linus has not taken the cache flush avoidance patches yet, it could give a noticable difference in this test. However the 2.4.7 debian kernel will still be a better comparison than a 2.2. Anton
Re: Sparc has slow disk access
Try the solaris test again, Binkey, and this time use a filesystem other than /tmp for the output. Try setting up a ufs-log file system and see if the performance doesn't exceed that of linux. I would think that Solaris tmpfs was faster than their filesystems since it wont hit disk unless memory pressure pushes things out to swap. btw last time I checked, Linux on ext2 beat Solaris on tmpfs at dbench (a filesystem benchmark) by a large margin on a dual ultra 60. Anton
Re: Sparc has slow disk access
Hi, That?s not good for the Sparc machine (and even worse for Solaris by the way but this is of less interest because the machine should run under Linux). Any hint if there some optimisations could be done? It would be worth upgrading the ultra to a recent 2.4 kernel from http://vger.samba.org. I found a 10% improvement in some IO benchmarks after the cache flush avoidance work went in. Anton
Re: 2.4.7 and sparc32... ;(((
Hi, hmm... the 2.4.7 boot messages say: Jul 22 22:33:31 twin kernel: 20306MB HIGHMEM available. Jul 22 22:33:31 twin kernel: On node 0 totalpages: 32080 Jul 22 22:33:31 twin kernel: zone(0): 36864 pages. Jul 22 22:33:31 twin kernel: zone(1): 0 pages. Jul 22 22:33:31 twin kernel: zone(2): 81747 pages. the box has 128 megs. I found a bug that could be causing your lockup, Ingo has already coded a patch which has made it into 2.4.8-pre1 :) Can you retest and let me know if you still have problems? Anton
Re: 2.4.7 and sparc32... ;(((
Hi, (who's anton? is he reading this list?) Yep :) yup; hung after a night on my SMP SS10, something about spinlocks (uh, forgot to write down the actual message), what I have written down and looked up in System.map is: CPU0 stuck at f0047e4c swap_out_mm CPU1 stuck at f003dedc exit_mmap owner 0:f003c568 handle_mm_fault Are you running with some highmem? (it should tell you in the boot log) If so I can see a few places that could cause this bug. It appears that we are scheduling with the page_table lock held. Anton
Re: 2.4.7 and sparc32... ;(((
Hi, Jul 22 22:33:31 twin kernel: 20306MB HIGHMEM available. Jul 22 22:33:31 twin kernel: On node 0 totalpages: 32080 Jul 22 22:33:31 twin kernel: zone(0): 36864 pages. Jul 22 22:33:31 twin kernel: zone(1): 0 pages. Jul 22 22:33:31 twin kernel: zone(2): 81747 pages. the box has 128 megs. Excellent it has highmem. My theory is that we end up sleeping in a kmap() inside handle_mm_fault. I'll should have a patch tomorrow, its getting late here in Canada but Im still on Australia time :) Anton
Re: reiserfs empirical study (very long)
The big endian patches change the code to use little endian ordering for all on-disk structures. IMO this is a mistake, and certainly costs a dear performance penalty, because on big endian processors, this method requires converting endianness both ways (reading and writing) for all meta data. Sane architectures (sparc64, ppc) have load/store with byte swap instructions and if reiserfs is using them you shouldnt see a performance penalty. cpu_to_le* etc make use of them. Anton
Re: Crash
data_access_exception: Shit SFSR[00801009] SFAR[fe320048], going. You need to pass this through ksymoops to be useful. What kernel version are you using? Anton
Re: Kernel troubles on sparc32
Hi, The canonical source for sparc kernels is from the CVS tree, rather than from linus' or alan's trees. Having said that, the error below occurs currently in the CVS tree. Check the archives of sparclinux@vger.kernel.org mailing list and you will find a solution for this compile problem. However, there are still other compile issues which prevent me from compiling CVS kernels on sparc32 right now. Anton is the primary maintainer of sparc32 right now, and he has been otherwise occupied for the past little while, hopefully he'll come back soon and make us compile again. :-) I have just finished changing jobs and am now removing my sparcs from storage :) Anton
Re: Sparc 20 and Kernel Oops on chmod()!
Hmm... That is interesting, care to share how to managed to find that this was the point of the crash from the data I sent you? Might be a handy thing to know. :) ksymoops -VOLK -m System.map oops Then match up the assembly with the C code :) Anton
Re: Sparc 20 and Kernel Oops on chmod()!
After some additional debugging and detective work, I figured out the catalyst for the crash... RAID partition autodetect by the kernel at boot time. Once I changed the types of the partitions on the raid disks to 83 (from fd), and rebooted, the system did not autodetect any raid arrays and came up fine with either 2.2.18 or 2.2.19 (w/new style raid patches applied [0]). Then I was able to start the raid up using /etc/init.d/raid2 and everything looks to be working just fine now. I've never tested the RAID autodetect stuff so its not surprising its broken :) Anton
Re: Problems with packages
i recently upgraded to woody. During installation i get this error on ghostcript fonts: Setting up gsfonts-x11 (0.12.1) ... /var/lib/dpkg/info/gsfonts-x11.postinst: line 12: 22181 Segmentation fault $currentcmd /usr/lib/X11/fonts/$dir Any try to remove or reinstall it gives the same error...any suggestion to get rid of it I didnt have a lot of time to look at this, but my feeling was it was compiled with an older version of glibc and that makes it unhappy. I found sid xmms was segving and a recompile fixed it. Anton
Re: Newbie questions
Do you get an oops report or does it lock up hard? I'd like to track this bug down since I havent had access to a 2 CPU machine for a while. No. it just stops. =^/ Is there a way for me to help tracking it ? If you have a keyboard, does the caps lock key work (ie the keyboard driver is working)? If so you could use the magic sysrq combination to dump task information. Can you stop-a on the keyboard or send a break over the serial console? If so then you can do ctrace, .registers, .locals at the OK prompt. If it locks up hard then debugging becomes a little more difficult. Some idea of the events up to the lockup will help. Anton
Re: Newbie questions
Hi, I've recently installed potato on a e4500 machine and noticed some strange behaviour. First, the bogoMIPS calculation is quite strange: Jan 26 17:00:21 levaca kernel: Calibrating delay loop... 799.53 BogoMIPS Jan 26 17:00:21 levaca last message repeated 7 times Jan 26 17:00:21 levaca kernel: Total of 9 processors activated (71.96 BogoMIPS) Oops, try this patch: --- linux/arch/sparc64/kernel/smp.c Tue Jan 2 17:27:24 2001 +++ linux_sparc32/arch/sparc64/kernel/smp.c Sat Jan 27 14:54:00 2001 @@ -290,8 +290,8 @@ } printk(Total of %d processors activated (%lu.%02lu BogoMIPS).\n, cpucount + 1, - (bogosum + 2500)/50, - ((bogosum + 2500)/5000)%100); + bogosum/(50/HZ), + (bogosum/(5000/HZ))%100); smp_activated = 1; smp_num_cpus = cpucount + 1; } Second, the machine hangs or boots when used heavily by period of aprox. 3hs. I'm crunching distributed.net rc5 keys, a task that uses full processing power, when these things happen. Do you get an oops report or does it lock up hard? I'd like to track this bug down since I havent had access to a 2 CPU machine for a while. and finaly, is there a way to obtain cpu temperature ? may I use SunOS's prtdiag for it ? There is an option to enable this stuff but I have no idea how well it works (I havent got a machine to test it myself): SUNW, envctrl support Anton
Re: make bzImage
Ok, thanks. I try to boot my Sparc10 with vmlinux-2.4.0; this is what happens: SILO boot: PROMLIB: obio_ranges 5 bootmem_init: Scan sp_banks, init_bootmem(spfn[217],bpfn[217],mlpfn[c000]) free_bootmem: base[0] size[100] reserve_bootmem: base[0] size[217000] reserve_bootmem: base[217000] size[1800] Booting Linux... Watchdog Reset Type help for more information There are a few problems with 2.4.0 on sparc32. I have some patches in the works but they arent quite ready yet. Anton
Re: netwatch bus error
It is probably doing an unaligned load or store.
How would I fix that?
Try this patch. Ben can this be added to the deb?
Anton
--- netdiag-0.7.orig/netwatch/netwatch.cWed Jul 7 23:31:26 1999
+++ netdiag-0.7/netwatch/netwatch.c Sat Oct 7 15:41:10 2000
@@ -2366,13 +2366,12 @@
}
void
-/* handle_ip (struct ip *argbuf, int length) */
- handle_ip (struct ip *buf, int length)
+ handle_ip (struct ip *argbuf, int length)
{
int wlen;
-/* struct ip bufstruct, *buf = bufstruct;
+ struct ip bufstruct, *buf = bufstruct;
- memcpy(buf, argbuf, sizeof(struct ip));*//* to fix unaligned accesses */
+ memcpy(buf, argbuf, sizeof(struct ip)); /* to fix unaligned accesses */
new = time (0);
wlen = ntohs (buf-tot_len);
selprob = 40;
Re: sun4 (SPARCserver 330)
does anyone have any info on the sun4 port? Whatever happened to it? what is the support in the kernel for sun4 at this moment? I ask because I have gotten hold of a SPARCserver 330. Is there anyone I should contact who would know (have info)? I have a sun4 machine running 2.2. There is currently no bootloader so it netboots the kernel. You will have to install the OS on another machine because there are currently no boot images with sun4 support. Anton
Re: Hardware support
Pretty much everything but E10k has been tested (sun4u). To be honest, a distributions hardware support is defined more by the kernel that what the distribution does. IOW, we support whatever the kernel supports. Our current (just released) 2.2 version is based on Linux Kernel 2.2.17. The E10k I booted a few weeks ago ran debian :) Of course we still have work to do to support its features (including the SSP console). Anton
Re: netwatch bus error
Any ideas why I'd be getting a bus error while running netwatch? Seems to die when it gets a packet :-( It is probably doing an unaligned load or store. Anton
Re: iptables not in potato?
to my surpise 'apt-get install iptables' on a potato box, fails with: E: Couldn't find package iptables. may I ask why? ;-) Not sure why, but I'd be interested to know of any problems with netfilter on sparc. Anton
Re: Potato: SPARCstation ELC white line on console
Having hunted high and low to find an answer as to why my ELC is showing a white line (the width of one character) on the left side of the console screen. Another curious thing is that the linux boot startup logo is munched. Can anybody shed some light on the subject? The white line is a bug in the bwtwo framebuffer. Not sure about the boot logo problem. Anton
Re: mozilla (previous was SV: Sound in Potato)
Would M17 exists on sparc (woody or whatever) ? Last time I check it is still M15. And it doesn't works for me. It does several things, fires up the profile manager which hangs. I hope that M17 would work (On an intel box M17 is much more stable than M15) The sparclinux builds appear on ftp.mozilla.org fairly soon after a release. Im even compiling them on a debian machine now :) Anton
Re: Gnome-Helix binaries for D2.2 USPARC
Woody's getting there now. I setup a brand spanking new (read uses all the new nifty buildd features) buildd last week, and so far it has pumped out 400 builds. Great stuff! I'll point one of my sparcs at woody. Does helix have an apt accesible source directory? If so, I can probably setup a chroot just to do autobuilds of it and make the deb's accessible. deb-src http://spidermonkey.helixcode.com/distributions/debian woody main Anton
Re: Gnome-Helix binaries for D2.2 USPARC
has someone built the binaries (or even an installer package) for the gnome-helix release? I'd be prepared to work on this. Ben: How is woody? Is the debian sparc unstable autobuilder up and running? Anton
Re: Power On After Power Failure...
I have a Sparc 20 running as a web server at work and is plugged
into a UPS along with two other computers. The UPS (Best Power Fortress)
has a serial cable that plugs into one of the other computers and warns it
of when there is a power loss and when to shutdown. I have this computer,
via LAN and ssh, tell the other computers on the UPS to also shutdown when
it is told by the UPS to shutdown. So, the Sparc receives a 'shutdown -h
1' when the AC power has failed and the UPS batteries are low.
This works just fine, and the Sparc shuts down cleanly. The
problem is that it will not start back up (after AC power has been
restored) until one presses the power button (either on the case or the
keyboard). Often power failures happen late at night and there is no one
to press the power button until the following morning, if not later.
We have a similar problem with shutting down an SS10/SS20 with a serial
console. In these cases we should drop into the prom and not issue a
shutdown...
OK I just fixed this in the CVS tree.
Your problem can be fixed if you fix your shutdown program. In kernel/sys.c
we have:
asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void * arg)
{
...
case LINUX_REBOOT_CMD_HALT:
notifier_call_chain(reboot_notifier_list, SYS_HALT, NULL);
printk(KERN_EMERG System halted.\n);
machine_halt();
do_exit(0);
break;
case LINUX_REBOOT_CMD_POWER_OFF:
notifier_call_chain(reboot_notifier_list, SYS_POWER_OFF, NULL);
printk(KERN_EMERG Power down.\n);
machine_power_off();
do_exit(0);
break;
So shutdown should be using LINUX_REBOOT_CMD_HALT.
Anton
Re: SMP SS10 freezes hard intermittently
I've got an SS10 with two CPUs, running a 2.2.15 kernel with SMP enabled. It locks up hard periodically, and won't respond to stop-A, a break on the serial port, the three-fingered salute, or, in fact, anything short of a power-cycle. When I was fixing sparc32 SMP in 2.2, I only had access to one cpu and machine type (viking supersparc on 4m/690). As such it tends to be stable :) ftoomsh# uptime 3:01am up 99 days, 13:59, 31 users, load average: 1.42, 1.38, 1.36 ftoomsh# cat /proc/cpuinfo cpu : Texas Instruments, Inc. - SuperSparc 50 fpu : SuperSparc on-chip FPU promlib : Version 3 Revision 2 prom: 2.14 type: sun4m ncpus probed: 4 ncpus active: 4 Cpu0Bogo: 49.86 Cpu1Bogo: 49.97 Cpu2Bogo: 49.97 Cpu3Bogo: 49.97 MMU type: TI Viking/MXCC contexts: 65536 CPU0: online CPU1: online CPU2: online CPU3: online There are too many things to fix in 2.2 (that I am fixing in 2.4) so random lockups are no surprise to me. 2.4 should be ready for sparc32 real soon. Cheers, Anton
Re: potato install on AX and AXi
Are the AX/AXi like the AXe (ps2 system)? Yep. They support both ps2 and sun keyboards. I assume this is with a Xsun. I think I have a solution to make the Xsun server try /dev/gpmdata initially before trying /dev/mouse. This will assume that you have -Rraw enabled on gpm which works for my Mach64 Ultra 5 (ps2 mouse though). I'll test that when I get home. binfmt_elf.o is actually 64bit userland support, which doesn't really work on 2.2.x anyway. The reason I did not compile it into the kernel is that it acts as more of a DoS (running 64bit programs) than anything useful :) binfmt_elf32 is compiled in obviously. Ahh of course. Just so long as kmod doesn't insert it automatically :) One more thing... why does setserial do stuff during bootup? Anton
Re: another take on the smp setup
In the make menuconfig for the 2.2.14 kernel, it has two entries which I think pertain to dual processor sparc 10 setups. These settings are: o Symetric Multi-processing support (does not work on sun4/sun4c) o Support for Sun4 machines (disables sun4[cdm] support) I said yes to the first option and no to the second. Is this the correct setting for a sparc 10 with two sm41 processors? Yes that is correct. Anton
Re: Kudos
Anyone had any luck with 2.3.99 kernels on sparc32? I'll be trying it maybe later today if i get a chance :) The latest 2.3.99 should work on sparc32. Please test if you can and send problems to myself or [EMAIL PROTECTED] Anton
frozen install on classic
Hi, I installed frozen on a classic using rescue/root floppy and the rest over http. Worked really well. Great stuff! I notice /dev/mouse is a symlink to /dev/gpmdata, however gpm does not support exporting a sunmouse device which the sparc X servers expect. Cheers, Anton
Re: Hardware compatibility...
Does anyone know of a link to the specifications of a Sun 4/330 model 547 server and will it run Linux of any kind? I know RH says it's not supported but I'm curious to know why and what exactly it it. The kernel supports 4/330, however the redhat/debian installers and silo do not as of yet. I installed to disk using another machine and now netboot it. Anton
Re: [constantin@fokus.gmd.de: [constantin@fokus.gmd.de: alt-ctrl-F[1-6] strange behaviour on Ultra1]]
I've hit similar behaviour on my 670MP. The problem is that X still 'has' the keyboard, and it's still in raw mode. The workaround is to login remotely or on the console, and run kbd_mode -a. Dunno if there's a fix. My 690MP has the same problem. I'll chase it up as soon as I have a non production sun4m. Anton
Re: utility for mapping oopsen to symbols in SPARC?
I keep hearing about koops (?) on the sparclinux list but can't seem to find it. Am I delusional, a poor speller, or something else entirely? You are probably looking for ksymoops. I usually run: ksymoops -VOLK -m System.map oops Cheers, Anton
Re: ANN: initial sparc potato boot floppies ready for testing
This is an initial shot at a set of boot floppies for sparc. Note that this is not uploaded for the dist yet since it is a) based on boot-floppies CVS and not the one in the potato archive, and b) uses a set of 2.2.14 kernel images that can't be uploaded until the 2.2.14 source is in the archive. On an ELC I get a data access exception at PC 0x4100 when loading the tftpboot.img. Anton
Re: Help! Kernel panics..
Logs at the bottom... (Looking at the logs, it looks like it happened yet again while I was at lunch, but it survived it this time. Weirdness.) Can you run this through ksymoops or else send me a copy of your System.map? Cheers, Anton
Re: kernel 2.0.37 compilation failure
Woaaah Apparently I'm very lucky, because the SLC here wasn't damaged. :-)) But I can confirm: X doesn't work if a kernel 2.2.x is running. (Tested up to 2.2.10) Text mode works fine if you ignore the big wipe stripe running down on the left side of the screen. ;-)) Don't worry, the SLC didn't die because of the bwtwo bug :) Its just old. So There is NO chance to get 2.0.37 running...?? (btw. 2.2.x would be even better- but I don't dare to ask ;-)) The bwtwo driver is pretty simple so shouldn't be too hard to fix (hint to anyone :) Anton
Re: kernel 2.0.37 compilation failure
What's broken about it, i've got a bwtwo in a IPC running 2.2.10 and it appears to work fine. X doesnt work. I was going to look into it but as soon as I decided to my SLC broke. Anton
Re: Sparc64 kernel on Sparcstation 10 ?????
I compiled successfully a sparc64 kernel on my SS10 but it does not boot. My question is, can I run a sparc64 kernel on a SS10? I need the kernel to enable solaris binary emulation. No you must run a 32 bit kernel on a SS10. The sparc64 solaris binary emulation needs to be ported to sparc32 for it to work. Anton
Re: New Kernels (Was: new bootdisks available for testing)
I suspect that there are serious problems with the 2.2.8 kernel, so you might want to go with the 2.2.7 ones. (I will try 2.2.9 RSN - Herbert Xu promised to upload an x86 package last sunday, but hasn't done so.) 2.2.7 hangs during booting on some sun4m machines. It would be a good idea to go with 2.2.9 if possible. Cheers, Anton
Re: error on instal sparc 1+
after booting from the CD, and starting the instal, finds the cd-rom, and hard drive, then gets caught in a loop of errors: esp0: disconnect, reseting bus esp0: restetting bus... esp0: SCSI bus reset interrupt The CD wouldn't happen to be a Sony CDU-8012 would it? A fix went in recently for a bug with this CD drive, however the debian images probably do not have it. Anton
Re: X on a SparcBook possible?
This SparcBook is running slink in text mode fine, although moving the cursor and srolling down the screen is painfully slow. (Why??) You are probably seeing the prom console which is very slow and only used if the framebuffer is not supported by the kernel. Anton
Re: Netscape 3.0 for Linux/Sparc
Anybody know any details about the problem ``SunOS apps sometimes crash''? I haven't had any problems recently. SunOS emulation was recently reenabled after many vfork bugs were fixed. Anton
Re: xwindow on sparc 1+
hi, I am trying to setup the xwindow for sparc 1+ with slink... I type startx and it starts, but it just sits there with a blank screen as black... I'm not sure if I have all the files or not... I have my xserver-xsun for my sparc and XFree86... I'm new to linux but have learned alot so far...not to get xwindow working.. Sounds like you are using a bwtwo. It is a known problem that should be fixed soon. Anton
Re: glibc 2.1 is now on master in my homedir
If that's the case, I've already tested the 2.2.4 and 2.2.5 from the cvs, which means it's still broke :) Sorry, I meant the broken vfork was added in 2.2.4 and was still broken in 2.2.5. With the latest cvs tree, vfork seems to work for me. The reason I say to use the cvs is that it is always more up to date for sparc than the Linus kernel, even tho some parts of it (like added drivers) aren't as stable. The differences should be manageable now that we have had the big sparc merge of 2.2.4. Anton
Re: glibc 2.1 is now on master in my homedir
Terrible that we will need to predepend on a kernel version for glibc 2.1.1 though. This also means we need to make a cvs source package for sparc kernels. IIRC, the vger cvs is also for powerpc so maybe we can have this for that arch aswell. From memory sparc vfork arrived in 2.2.4 and will be fixed in the mainstream kernel whenever the next merge is done. So assuming you don't ship these range of kernels all will be OK. There is no need to make a cvs source package unless you can't wait for the merge. Anton
Re: glibc 2.1 is now on master in my homedir
This isn't the problem. modprobe is locking up. Whenever a module is autoloaded, the system locks up. The last thing modprobe does is: SYS_66( Which should be a 'vfork()'. I'm not entirely sure what is going wrong here. In the kernel vfork calls exactly the same code as fork (although the entry code is different in sparc64). I would suspect a kernel problem, but the older version of libc works. sparc32 vfork has been broken until very recently. It should be fixed in the cvs tree. Anton
Re: Well... at least it's constant! Can not get passed SCSI bus reset
Please review the following boot sequence of a SPARC 630MP and if you can, shed some light as to why the unit wants to cycle through the SCSI bus reset process continuously. This is a self burned 2.1 debian CD image downloaded from the debian site. Detected scsi CD-ROM sr0 at scsi0, channel 0, id 6, lun 0 esp0: disconnect, resetting bus esp0: resetting bus... esp0: SCSI bus reset interrupt esp0: SCSI bus reset interrupt esp0: disconnect, resetting bus esp0: resetting bus... esp0: SCSI bus reset interrupt esp0: SCSI bus reset interrupt I have seen this before with the esp and certain SCSI devices. I'll look into it. Anton
Re: 2.1.132 compile errors.
Ahhh, ok. Wasn't sure what was up. the .125, .130 and .131 kernels compile fine for me from tsx-11. Oh well, i'll have to stick with .131. and play with .132 on my Pentiumpro box. :) I've been getting NMIs somewhat frequently with .125, especially, but not exclusively, when exiting X. Any advice on whether 130 or 131 might get rid of the NMIs? Or is it just a Sparc 2 thing? 2.1.132 (and the vger tree) has a patch for the cgsix framebuffer which should fix your problem. Anton
Re: 2.1.132 compile errors.
I'm getting errors trying to compile 2.1.132 on my SS5-85 and was wondering if anyone else had any trouble, or ideas on how to fix. Hi Chris, The mainstream kernel is out of sync for sparc. For the moment use the vger kernel (see ftp://vger.rutgers.edu/pub/linux/README.CVS) or else wait a while for davem and linus to sync up. Cheers, Anton
Re: The time has come ...
Jules the machine works well, although I do also occasionally see a Jules watchdog reset for no apparent reason. Speaking of which, is there any way to set it so that such resets reboot the box, instead of halting it? There is an option in the PROM to do this: ftoomsh-root# cat /proc/openprom/options/watchdog-reboot\? 'false' Anton
Re: The time has come ...
Alas :-( It doesn't fix it. The machine stil halts (actually, it watchdog resets) when X dies. Can you get a dump of .registers, .locals, ctrace? I don't have a cgsix so it is pretty hard to trace without something to go on :) You could set up a serial console on the sun if the screen is stuffed. Cheers, Anton
Re: The time has come ...
I updated my cvs snapshot a few hours ago, and there was still a bug. But applying the patch that you sent to the debian-sparc mailing list fixed the problem. Thanks! Great. The fix has been checked into the cvs tree. Cheers, Anton
Re: The time has come ...
Hi Jules, Are you using a cgsix framebuffer? The current kernels have a problem with the cgsix, which Davem narrowed down to the cg6_reset function. Do you get booted back to the PROM? If so can you get the results of: .locals .registers ctrace As well as your System.map Hmm. It does sound exactly like that. Unfortunately the PROM is rarely readable, since the CLUT has generally made background == foreground. However, I'll try to get those data. How come I never read about this on [EMAIL PROTECTED] Did I just miss it? I found it at http://dm.cobaltmicro.com/~davem/ultra-todo.html Let me know if this fixes your problem. Cheers, Anton --- linux/drivers/video/cgsixfb_old.c Thu Nov 26 15:04:28 1998 +++ linux/drivers/video/cgsixfb.c Thu Nov 26 15:07:37 1998 @@ -534,6 +534,7 @@ unsigned int rev, conf; struct cg6_tec *tec = fb-s.cg6.tec; struct cg6_fbc *fbc = fb-s.cg6.fbc; + u32 mode; /* Turn off stuff in the Transform Engine. */ tec-tec_matrix = 0; @@ -552,15 +553,20 @@ *(fb-s.cg6.fhc) = conf; } - /* Set things in the FBC. */ - fbc-mode = ~(CG6_FBC_BLIT_MASK | CG6_FBC_MODE_MASK | + /* Set things in the FBC. Bad things appear to happen if we do +* back to back store/loads on the mode register, so copy it +* out instead. */ + mode = fbc-mode; + mode = ~(CG6_FBC_BLIT_MASK | CG6_FBC_MODE_MASK | CG6_FBC_DRAW_MASK | CG6_FBC_BWRITE0_MASK | CG6_FBC_BWRITE1_MASK | CG6_FBC_BREAD_MASK | CG6_FBC_BDISP_MASK); - fbc-mode |= (CG6_FBC_BLIT_SRC | CG6_FBC_MODE_COLOR8 | + mode |= (CG6_FBC_BLIT_SRC | CG6_FBC_MODE_COLOR8 | CG6_FBC_DRAW_RENDER | CG6_FBC_BWRITE0_ENABLE | CG6_FBC_BWRITE1_DISABLE | CG6_FBC_BREAD_0 | CG6_FBC_BDISP_0); + fbc-mode = mode; + fbc-clip = 0; fbc-offx = 0; fbc-offy = 0;
Re: The time has come ...
I have a similar problem. I use X via startx. When I quit X, I get: Thanks for the report Paul. Davem should be adding a change into the CVS tree later on tonight. Let me know if it doesn't work. Cheers, Anton
Re: The time has come ...
I switched back to 2.1.125 and the situation is now: startx as root works OK, but when you end the X-session, the kernel halts itself. xdm works OK, but if you kill xdm (/etc/init.d/xdm stop) the kernel commits suicide :-( Are you using a cgsix framebuffer? The current kernels have a problem with the cgsix, which Davem narrowed down to the cg6_reset function. Do you get booted back to the PROM? If so can you get the results of: .locals .registers ctrace As well as your System.map Cheers, Anton
