i don't know if the original 8.1 kernel is okay, but the updated kernel pass my testing.. and about DMA, only cdrom using ide DMA, so it seems not very important to me. just disable it and don't let it crsah my server:)
> I had similar problems with 8.1 too, so now i'm back to 8.0 kernel and > everything works fine. > What about the DMA change, do you need the latest kernel ? > Stef > > Le Mercredi 29 Mai 2002 09:49, vous avez ?crit : >> > Have you tried kernel-secure, or an older kernel (for example the >> > one provided with 8.0) ? >> > >> > Try to disable DMA in /etc/sysconfig/harddisk* (as suggested by John >> > Allen). >> >> oh..yes. old kernel is okay. mdk 8.1 is ok (sorry not try mdk 8.0) >> red-hat 7.2 is okay. red-hat 7.3 is okay. >> DMA must disable. >> if DMA enable, >> under my test with a scratched CD, only red-7.2 can survive. >> mdk 8.1 will lock cd-rom forever until reboot. >> mdk 8.2 and red-hat 7.3 will hang. >> >> mdk kernel(include cooker) seems not as new as red-hat in some >> respects. it seems red-hat kernel already solve the problem. >> (original 2.4.18 + serverworks chipset + smp + highmem = freeze. >> it can be checked at www.kerenel.org email list archive, and this >> is my problem.) >> >> and redhat's megaraid driver is much faster. >> in my hp tc4100 server with hp net-raid 1M raid card, >> testing with "hdparm -t /dev/sda" >> redhat 7.3 got about 70MB/sec >> redhat 7.2 and mandrake 8.2 got only about 30MB/sec >> >> > Stef >> > >> > Le Mardi 28 Mai 2002 18:05, [EMAIL PROTECTED] a ?crit : >> >> hi: >> >> i test hp-tc4100 with red-hat 7.3 today. >> >> it is okay after upgrade to 2.4.18-4. >> >> (2.4.18-3 kernel panic after several hours. just like red-hat >> >> said.) >> >> >> >> i also test the latest cooker 2.4.18-16 >> >> it crashed as usual. >> >> so the problem seems at mandrake kernel. >> >> hope 2.4.19 will fix it. >> >> >> >> ========================================================================>> >>>===>>===tbsky : 27:05 09:34 : Incident created hi: >> >>our >> >> company buy a hp lc2000 server and try to install mdk 8.2 on it. it >> >> can not pass testing and finally we run mdk 8.1 at it. >> >> this time we buy a "certified server" -> hp tc4100. >> >> the machine arrived today and it fail again. >> >> this machine comes with 2 cpu and 1g ram. >> >> to use the memory i must enable highmem(eg: enterprise kernel). >> >> and with highmem enabled we got crash under testing. >> >> i had report this before to: >> >> >> >> [EMAIL PROTECTED] >> >> [EMAIL PROTECTED] >> >> >> >> and i saw people at list has same problem. but nobody at mandrake >> >> answer these questions. >> >> >> >> i also wrote several times to >> >> >> >> [EMAIL PROTECTED] >> >> >> >> although mdk said that email list is not for supporting, but whom >> >> should i write if the hardware database is wrong?? >> >> >> >> i'll use paid support tomorrow, but before that, i would like to >> >> post it last time for anyone who is interested..below is my >> >> experience with hp lc2000 last month. and now tc4100 kill me >> >> again. >> >> >> >> ************************************************************************>> >>>***>>* >> >> ************************************************************************>> >>>***>>* >> >> ************************************************************************>> >>>***>>* >> >> >> >> Hello: >> >> i m glad that i got ur reply :) let me describe my problem >> >> and testing procedure in details...i think the problem is >> >> causing by kernel, since redhat 7.2 is fine... and i run >> >> memtest86 to test memories, which are fine.. >> >> >> >> 1.) OSB4 DMA bug. it will cause system hang if u leave the ide dma >> >> enabled, >> >> and u try to read a scratched cd which the cdrom can not read >> >> well. redhat 7.2 is fine with dma enabled. >> >> >> >> 2.) xwindow say mtrr can not use and leave messages below in the >> >> syslog: >> >> mtrr: Serverworks LE detected. Write-combining disabled. >> >> redhat 7.2 is fine. >> >> >> >> 3.) irq conflict in the linux boot message. at bios boot screen i >> >> saw >> >> vga using irq 10, lan using irq 9. and cat /proc/pci give me >> >> the same result. but boot message say device 00:04:00(lan) want >> >> irq 10, which already has irq 9. >> >> so i try to lock lan to irq 10 at bios. bios boot screen is >> >> fine, vga using irq9, lan using irq 10, cat /proc/pci give me >> >> the same. but boot message now tell me device 00:04:00 want irq >> >> 9, which already has irq 10... :( >> >> >> >> 4.) unstable with kernel-enterprise. i test my system with ctcs >> >> that >> >> comes with mandrake 8.2. and using ctcs default, so it will run >> >> "kernel making","memory testing" (with memtest86),and "disk >> >> testing" (with badblocks) simultaneously. the system will >> >> freeze about 4 hours. the system didn't hang, just freeze, so >> >> no more new process. i can switch virtual consoles. i login 3 >> >> virtual consoles before testing, one run ctcs,one run top, and >> >> one stay at shell prompt. >> >> and when freezing: >> >> the ctcs clock is still ticking at console 1. >> >> the top process is hang at console 2. looking at time at >> >> up-left corner, i can know when the system froze. >> >> i can type command at console 3, the shell prompt, but it will >> >> hang after i hit "enter" if need a new process. >> >> >> >> there is a clue but i don't understand. during testing of ctcs, >> >> sometimes there is error below in the syslog: >> >> >> >> "kernel: Warning: dev (04:01) tty->count(5) != #fd'(4) in >> >> tty-open" >> >> >> >> the message appear about every 20-30 minutes. and after about >> >> 10 messages, the system froze. >> >> redhat 7.2 with ctcs comes with MDK 8.2 pass the testing for 12 >> >> hours... >> >> >> >> Regards, >> >> tbsky >> >> >> >> >> Hello: >> >> >> i install mandrake 8.2 on a HP lc2000 u3 srever with 1GB ram and >> >> >> 2 cpu, >> >> >> it is unstable with enterprise kernel. i found a link which is >> >> >> the same as my problem: >> >> >> >> >> >> http://marc.theaimsgroup.com/?l=linux-kernel&m=101770689615279&w=2 >> >> >> >> >> >> i also try the kernel in cooker but the same. >> >> >> and when i start X under 8.2, >> >> >> i have no experience with serverworks chipset before. >> >> >> does serverworks support linux? should i prevent it in the >> >> >> future? >> >> > >> >> > Yes. ServerWorks chipsets are supported. I have about 20 Dell PE >> >> > 2xx0 (ServerWorks chipset based) servers, two cpus, mostly 1G >> >> > RAM. All working fine, distrib kernel from mdk 8.1 and/or from >> >> > mdk 8.2. One diffence: i use kernel-secure (it's based on >> >> > kernel-enterprise). >> >> > >> >> > IMHO problem is motherboard/memmory not in chipset. All (two) my >> >> > HP server are shit hardware, mandrake have healthy problems on >> >> > it. >> >> > >> >> > Zdenek Mazanec >> >> >> >> ========================================================================>> >>>===>>===
