From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0; Q312461) Description of problem: booted from kernel-bigmem-2.4.17-0.6 2 8G 4xPIII machines died within hours of eachother, one had this in the logs: Jan 30 23:55:18 tigris kernel: kernel BUG at /usr/src/build/65535- i686/BUILD/kernel-2.4.17/linux/include/asm/spinlock.h:90! Jan 30 23:55:18 tigris kernel: invalid operand: 0000 Version-Release number of selected component (if applicable): How reproducible: Sometimes Steps to Reproduce: 1.It just happens after uptime of a few days and withmost of the 8G ram used for file system buffers and only 2G swap which we think is all unused (it is whenever we look). We also find unless we artifically limit the RAM with MEM=something less than 8G the machine won't even bootup, but this is with all 2.4 bigmem/enterprise kernels we have tried, compiled ourselves OR prebuilt 2. 3. Additional info:
Does this behave any better with current rawhide kernels (2.4.18-0.x)?
We have been trying 2.4.18-0.4bigmem with an uptime of 14 days so far. The MEM= problem we also mentioned was fixed with a BIOS update from DELL but this did not fix uptime-crashing problems. Anyway, possibly 2.4.18 has fixed it, if we have no further issues in 4-6 weeks we will presume it has and install that kernel on the other boxes. Thanks.