Description of problem: I have a pretty much reproducable kernel crash which spits out the following message to the console before dying: drivers/block/cfq-iosched.c:1070: spin_is_locked on uninitialized spinlock d9d133b8. (The last number varies). Reproduction as follows: 1) Plug in an USB stick, mount it 2) Write something to it 3) umount and remove Sometimes the system hangs a short time (max 30 seconds) after the removal. Sometimes the system stays up for hours, and then hangs when trying to log out of gnome. Sadly the line above is all that is printed of the oops. Happens on an i686 kernel (athlon), I have similar effects on ppc, but no kernel dump for lack of a serial console. Version-Release number of selected component (if applicable): kernel-2.6.11-1.1240_FC4 How reproducible: Pretty often. Steps to Reproduce: 1. See above. 2. 3. Actual results: Kernel crashes Expected results: normal operation Additional info:
can you paste lspci & lsmod output please ?
Created attachment 113402 [details] lsmod output
Created attachment 113403 [details] lspci -vv output
todays rawhide kernel has a bunch of extra debugging info enabled, can you try and repeat it with that please ?
I have posted my results with a rawhide kernel over in bug #155472
Result with 2.6.11-1.1251_FC4: Kernel panic - not syncing: drivers/block/cfq-iosched.c:1070: spin_is_locked on uninitialized spinlock d9228474. (Not tainted) [<c01208f8>] panic+0x45/0x1e2 [<c02995d4>] cfq_exit_io_context+0x130/0x486 [<c0292154>] exit_io_context+0x41/0x49 [<c0124b1d>] do_exit+0x4c/0x510 [<c0188f01>] sys_stat64+0x23/0x28 [<c01250ab>] do_group_exit+0x9e/0x1f5 [<c0103a61>] syscall_call+0x7/0xb
I get the exact same kernel panic on my laptop whilst performing a shutdown -h now on kernel 2.6.11-1.14_FC3 but this never panicked on the previous kernel of 2.6.10-1.770_FC3
this should have been fixed in 1254 build of rawhide. FC3 users need http://people.redhat.com/davej/kernels/Fedora/FC3
I appreciate the link to FC3 RPMs. Currently, I get my rpm files via a "home grown" yum-like update script that mirrors them from: mirror3.cs.wisc.edu: /pub/mirrors/linux/download.fedora.redhat.com/pub/fedora/linux/core/updates/3 and they've always seemed to be reliable in the past; if the latest officially released FC3 RPMs are on redhat.com , why are they not mirrored at wisc.edu ?? Are the RPMs in your davej directory "official" ?? or are they "test" code ?? Should I be getting all non-kernel RPMs from wisc.edu and only the kernel stuff from people.redhat.com/davej/kern... Thanks ... dazed and confused...
Greg: The RPMs in davej's directory are usually prerelease kernels. If you always want to test the very very latest stuff *and have time to report the breakages* then you could use them all the time (of course they might eat your system - that's life on the edge). Fixes to problems will appear there first sometimes months before the next "regular" kernel is released and it is feedback on these davej kernels that is used to decide what to release to the mainstream. There are lots of davej kernels. A brief aside - if too few people test (and report the problems) in davej kernels then a kernel which has problems for people who excercise different parts of the kernel to the people who did test it can slip through and be released as a regular kernel. Summary The davej kernels are meant for testing and not for mass consumption. Good testing kernels (especially ones that fix bugs without introducing side effects) are eventually released for mass consumption. If you are seeing a problem that might be fixed in a yet-to-be-released davej kernel then the developers will ask you to test the davej kernel to save you waiting for a regular release and to get feedback quicker. (This whole thing is offtopic for bugzilla)
I tried yesterday with -1261, and was able to plug/write/unplug my USB-sticks without making the kernel barf.
also with todays 1275 ?
-1275 seems to work fine, too.