Messages in this thread | | | From | "Adam J. Richter" <> | Date | Fri, 24 Apr 1998 00:37:58 -0700 | Subject | 2.1.97 bug potpouri |
| |
I am experiencing a variety of bugs in 2.1.97. I have decided to catalog them on this list at this stage, just so people experiencing similar problems can notice common elements.
All of these bugs are from the same slightly hacked 2.1.97 kernel, compiled with SMP definied, and with only romfs and the initial ramdisk support compiled other. Other necessary modules are loaded by an initial ramdisk that runs from /sbin/init rather than /linuxrc and exits by mounting the new root on /mnt and doing "cd /mnt && exec /mnt/bin/chroot /mnt /sbin/init".
Here is a description of the machines and what I experience when I run 2.1.97 on each of the machine, and a description of the machines. To start with, here is the network topology:
|- asgard (modem)---------------------(modem) adamhome |- adam |- cdrom1 |- freya |- hilda |- other machines not running 2.1.97, internet connection 100Mbps ethernet,
adam.yggdrasil.com single 300MHz Pentium II + 128M SDRAM intel single cpu slot motherboard single scsi disk on a Diamond FirePort 40 (ncr53c8__) controller using ncr53c8xx driver tulip 100MHz ethernet card using tulip driver IDE cdrom Things work great, except for when I telnet in directly from my home machine (adamhome.yggdrasil.com), which runs 2.0.0. In that situation, my telnet connections hang every now and then if I do something that generates a little more than a screenful of output very quickly (e.g., starting a text editor, running "more"). The connection will hang for about 5-10 minutes and then will close with an error like "connection reset by peer." It is possible to make other direct telnet connections and have them work while this is going on, but they seem to hang more quickly once one connection has entered this hung state.
adamhome.yggdrasil.com notebook computer ti486 (386 clone) + 12MB RAM pcmcia modem umsdos root filesystem on IDE disk The ramdisk dies after mounting the root filesystem on /mnt, with the following messages: UMSDOS dentry-VIP-Beta 0.82-3 (compatability level 0.4, fast msdos) /mn/ kill_dentry: kill_dentry --linux--.---: inode is NULL emd_dir_lookup: <7>umsdos_setup_dir_inode: umsdos_emd_dir_lookup for inode=c0ba7400 returned 00000000 The "/mn/", makes me think that there is at least an off-by-one error is some manipulation of filenames (I would assume it wants "/mnt").
asgard.yggdrasil.com single Pentium Pro 180MHz + 128MB RAM intel venus single cpu motherboard Adaptec 2940 scsi controller with scsi disks At least one IDE disk too. Cyclades 8 port ISA smart serial card tulip ethernet card using tulip driver This machine routes networking connections from the modems on the multiport serial card to the local ethernet. I have just switched this machine to 2.1.97 for the evening as an experiment. It seems copasetic, although large sets of file transfers across the local ethernet from this machine seem slower when it is running 2.1.97. However, I get the following rather disconcerting error message about once every ten minutes: "(scsi0:-1:-1:-1) Data Parity Error during PCI address or PCI writephase".
cdrom1.yggdrasil.com Pentium 150MHz + 32MB RAM cd writer attached to ncr53c810 scsi controller using ncr53c8xx driver some ide disks tulip ethernet card using tulip driver This machine generates receives about sixteen CDROM images each day over the 100bT ethernet, sometimes from different machines at the same time, so it gives the networking code a good workout. Network transfers of even one CD image hang the connection under 2.1.97, even though it is still possible to ping the machine. Unlike the case with connection from my home machine to my office workstation, I cannot log into this machine again. Instead, I open the telnet connection and do not get a login prompt. There are two interesting messages that I get on this machine:
recvmsg bug: copied 46B9E551 seq 46B9EB05 (sometimes this is reapeated many times, but not always) and apm_bios: set display standby: Unrecognized device ID apm_bios: set display read: Unrecognized device ID
I am going to test this machine with APM disbabled.
freya.yggdrasil.com Single 300MHz Pentium II in Tyan dual PII motherboard + 128MB SDRAM 4 IDE disks a bunch of scsi disks on a Diamond FirePort 40 (ncr53c8__) controller using ncr53c8xx driver tulip ethernet card using tulip driver Even if comment out all entries in /etc/fstab that define swap partitions, I get flooded with "trying to free nonexistant swap-page" errors on the console.
hilda.yggdrasil.com 133MHz Pentium + 32MB RAM 1 scsi disk on ncr53c810 controller tulip ethernet card two 4-port dumb serial cards This machine seems to work fine under 2.1.97 so far. It has only been running it for about a half hour.
This is not a "help! rescue me!" message. Rather, I just think this collection of data may be useful to other people either experiencing similar problems that they are trying to isolare or working on the relevant sections of the kernel. Also, I hope that this will provide a sense of what remains to be done to produce a stable 2.2.
Finally, please do not take this message as too harsh of a criticism of 2.1.97. I love 2.1.97. From the SMP to the demand loading of modules to the faster copies and ieee-1284 parallel port probing, it is great stuff. I really cannot do justice to all of the contributions that have been made. So, for now, I'll just say thank you for your contributions one and all, and especially Linus! It's more great work.
Adam J. Richter __ ______________ 4880 Stevens Creek Blvd, Suite 205 adam@yggdrasil.com \ / San Jose, California 95129-1034 +1 408 261-6630 | g g d r a s i l United States of America fax +1 408 261-6631 "Free Software For The Rest Of Us."
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |