Description of problem: Kernel panic at load kernel modules with more 1G RAM in box. Version-Release number of selected component (if applicable): 2.4.21-9.0.1 How reproducible: run script find /lib/modules/`uname -r` -name "*.o" -exec insrmmod {} \; where insrmmod: #!/bin/sh echo Loading module $1... insmod $1 echo Module $1 has loaded echo Releasing module $1... rmmod $1 echo Module $1 has released Actual results: Panic Expected results: Additional info:
Could you please indicate which module load/release causes the panic and include the oops/panic console output? Thanks. -ernie
Panic are in random module and random points. Only one is equally panic in address 0xF....... simple output md74xx.o: Unable to handle kernel paging request at virtual address f8ab9560 piix.o: Unable to handle kernel paging request at virtual address f8a3f560 .... Please run this cript and you all see.
It is not reasonable to try loading all modules for a given release/architecture if the associated "unsupported" RPM has been installed (such as kernel-smp-unsupported-2.4.21-12.EL.i686.rpm) because some of the unsupported modules might cause crashes. Have you installed any of the "unsupported module" RPMs? If not, could you please attach the exact console output that occurs from one or two of your crashes? Thanks. -ernie
Created attachment 99240 [details] oops panic during module loading
Created attachment 99241 [details] dmesg from machine where panic was happened
these attaches for configuration: 1Gb memory kernel built from kernel-2.4.21-i686-config config noone unsupported modules has loaded
sorry, correction to previous post: kernel configuration made from original kernel-2.4.21-i686-config by additional check highmem (4GB) support.
Hello, Alexander. Red Hat supports kernels/modules from the binary RPMs that we ship (except for the modules in kernel-*unsupported*). For example, for the config specified by kernel-2.4.21-i686.config, we support the kernel/modules from kernel-2.4.21-9.EL.i686.rpm. We do not support customer built kernels because various changes to config settings can produce unreliable software, and further, when a customer builds kernels from our kernel-source-* RPMs, all of the unsupported modules (potentially with serious bugs) are also built. And finally, we do not provide any support for a tainted kernel such as is shown in the oops output in comment #4. If you can reproduce a crash with a supported module on a supported kernel, please indicate that in this bug report. Thanks. -ernie
ernie you read my first post? or you not want find and fix bug? if you do not want it - say it. You run attached script ? It`s bug provides with _DEFAULT_ kernel config and panic show at _SUPPORTED_ modules. If help you i can say - at some situations (i don`t find reason of it) at process VMA have area with end < start. How i see it after fork and COW. I see you find any possibilites for not work with this PR. why ?
Hello, Alex. I have tried a similar script on a 4-GB machine in our test lab (with a soon-to-be-released kernel version 2.4.21-9.0.3) and was not able to reproduce a crash. The difference between my script and yours is that it avoided loading/unloading any module that was already loaded before I started the test, and it also avoided loading/unloading aic7xxx_old.o, since that conflicts with aic7xxx.o (which is used on my test machine). My suspicion is that you must be trying to reload/unload a module that is necessary for your machine, or perhaps you're trying to load a module that is incompatible with your hardware. At this point, I simply don't know how to help you. As I mentioned before, Red Hat cannot provide support for customer-built or tainted kernels, which is the only oops output provided in this bug report. If you can somehow determine which module is giving you a problem, and you can show an untainted-kernel crash from loading or unloading that module, then we would be happy to debug the problem. Cheers. -ernie