Messages in this thread | | | Date | Sun, 07 Mar 1999 17:27:49 -0500 | From | Robert Gash <> | Subject | Re: 2.2.1-ac5/2.2.2 IDE GPF's [Was: Kernel hard reboot in 2.0.36] |
| |
Yes, it does appear that I didn't have overclocking setup properly, my CPU demanded 2.3v power to run properly at 450mhz. But for the sake of being complete in my efforts to debug this problem I have found this (and these notes stand for ALL speeds, 300mhz 2.0v core, 450mhz 2.3v core, etc., I have spent about 3 hours testing them): every time that I boot Linux from the harddrive it fails. Don't ask me why, I don't know. I am currently running a 2.2.2 kernel (stock, no patches) that I made a bootfloppy from. When I boot off of the floppy EVERYTHING works fine (except my cablemodem, but that's a driver problem). EIDE checks run dandy, everyone is happy. However, the SAME kernel, when installed on the harddrive (make install on the kernel) fails at the partition check. The floppy copy does NOT ever fail, the kernel loads (slow, as floppies are that way), then it just cranks right on through normal startup procedures.
Below is the message (took forever to write this damn thing down) that all of the 2.2 kernels give me when I boot them from my HD, the screen goes like this (none of this occurs w/ a bootfloppy ver of the same kernel):
Partition check hda: general protection fault: 0000 CPU: 0 EIP: 00010286 EFLAGS: 00010286 eax: ffffffff ebx: 00000000 ecx: 00000300 edx: ffffffff esi: 00000400 edi: 00000000 ebp: 00000300 esp: c0090f08 ds: 0018 es: 0018 ss: 0018 Process swapper (pid:1, process nr: 1, stackpage: c009d000) Stack: c0123919 00000300 00000000 00000400 c0123b77 00000300 00000000 00000400 00000300 c000d0e0 00000040 00000300 00000000 c0123f6c 00000300 00000000 00000400 c01f0cd4 c016ee27 00000300 00000000 00000400 00000300 00000000 Call Trace:[<c0123919>] [<c0123b77>] [<c0123f6c>] [<c016ee27>] [<c016f182>] [<c01bbd57>] [<c0106000>] [<c0106000>] [<c0106093>] [<c0106513>] Code 8b 12 39 58 04 75 f3 39 70 08 75 ee 66 39 48 0c 75 e8 89 c2
And that that point I must reboot. The biggest kicker is the face the EXACT SAME image, when booting the EXACT same way when loaded from a floppy works! It goes right through w/o any errors at all. I find this rather odd. This behavior is identical at all speeds, overclocked or not... I know the harddrives are good because I fsck'ed them and no errors showed up. Below is the output from the /proc/cpuinfo from 2.2.2 booted from a floppy (what I'm in right now), this is at 100mhz fsb and 450mhz CPU speed:
processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 6 model name : Celeron (Mendocino) stepping : 0 cpu MHz : 451.034056 cache size : 128 KB fdiv_bug : no hlt_bug : no sep_bug : no f00f_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx osfxsr bogomips : 448.92
I am unsure how much that information on the screen will help you, but I can assure everyone that that is exactly what happens (every time). I'll be happy to provide the kernel image if someone wants it. If you need any OTHER output from anything let me know and I'll be happy to provide it.
-R
I have also put this system through hard stress tests in both Windows and Linux, and none of them have failed. These tests included Gausian Blurr radius 10 on a 100mb file (600dpi), rc5des clients, prime95 (and mprime for Linux) for the torture test (ran overnight), played mp3's while compiling kernels and such. Everything I can think of to test the system has come out just fine (and fast), so I am really leaning to the fact there is a bug somewhere in LILO (if LILO could cause this) or the kernel IDE code (since this occurs at all speeds, overclocked or not). And I know the CPU is not hot because the heatsync is always cool to the touch (I have some monster heatsync to cool the CPU and 5 case fans).
Charles Cazabon wrote: > > Robert Gash <gashalot@cybermax.net> wrote: > > > To make a long story short: my machine would reboot every time it started > > services of some sort (most notable at the point where it does something with > [...] > > I should note that this is a > > Celeron 300A overclocked to 450mhz (100mhz FSB) running a core voltage of 2.2v > > However, I know this is not the problem > > I have to agree with others' opinions here; the overclocking is the most > likely culprit. > > There's a way to prove it, however; adjust your clock speed back to 300, > and your voltage back to spec (2.0V most likely) and run it for a while. > If this solves the problem, the overclocking is causing it. > > It is quite easy for a machine to be 'stable' in Windows (for small values of > 'stable') when a real operating system will crash it in minutes. I would > think that this is what is happening. > > The Celeron 300 used to be very overclockable; however, the success ratio > has been going down for several months as Intel cherry-picked most of the > exceptional dies to be sold as 333,366, and 400 MHz variants. The proof of > this is the fact that you have already had to raise your core voltage. > > Also, beware the dangers of electromigration. Running any silicon at a voltage > higher than specified by the manufacturer is very likely to cause > electromigration problems -- and when they develop, you throw the chip out > and start fresh. > > Charles > -- > ---------------------------------------------------- > Charles Cazabon <charlesc-linux@qcc.sk.ca> > Any opinions expressed are just that -- my opinions. > ----------------------------------------------------
-- Robert Gash | _____ __ Systems Administrator | / ___/_ __/ / ___ ______ _ ___ ___ __ Phone: (904) 281-2200 x3312 | / /__/ // / _ \/ -_) __/ ' \/ _ `/\ \ / Fax: (904) 296-4203 | \___/\_, /_.__/\__/_/ /_/_/_/\_,_//_\_\ gashalot@cybermax.net | /___/ , Inc.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |