Messages in this thread | | | Date | Mon, 6 Sep 1999 04:08:01 -0700 (PDT) | From | Robert Dinse <> | Subject | spin_lock bug on Sparc SMP |
| |
I'm posting this one last report. At this point I'm building non-SMP kernels and letting 3 of the CPU's sit idle per David Miller's suggestion. I don't have money to go out and buy all new Ultra hardware (David's other suggestion) to replace the Hypersparc machines I presently have.
From a CPU occupancy standpoint I have no reason to, the news box runs at an average 9% occupancy and the web server 15% when all four CPU's are in service. I guess I'll either suffer along in single CPU mode, put up with instability, or find another OS unless someone else has an idea what is causing this.
To re-cap, on this particular lock-up, the machine in question is Quad CPU Ross Hypersparc SS-10 with 384MB of RAM, running Apache 1.3.3p3. Machine handles approximately 1 million hits per day and gets web pages from a shell server running SunOS 4.1.4 via NFS. The only extra card in the machine other than the CPU modules is a CG6 video board. It does have some local disk used for root, /usr, /var, /tmp, swap, etc, but web pages are on a different machine.
This crash the following message printed repeatively:
spin_lock (f0155f6c) CPU#3 stuck at f002dc7c, owner PC(f0020a40): CPU(0) . . .
f0155ca0 A _edata f0155f6c - No Exact Match f0157130 ? fixupretl
f002db98 T printk f002dc7c - No Exact Match f002ddcc T console_print
f0020a28 t srmmu_pgd_alloc f0020a40 - No Exact Match f0020bf0 t srmmu_set_pgdir
If anybody has any helpful suggestions besides buy new hardware, please let me know.
I'm afraid I don't have the old .Configure since I re-configured the OS for single CPU operation but it is essentially the same as the one from the previous post as I had kernels on both machines configured the same.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |