Messages in this thread | | | Date | Sun, 28 Feb 2010 12:11:23 +1100 | From | Anton Blanchard <> | Subject | Re: [PATCHSET] workqueue: concurrency managed workqueue, take#4 |
| |
Hi Tejun,
I gave the workqueue patches a spin on PowerPC. I'm particularly interested from an OS jitter perspective, and that these patches wont introduce more jitter. It looks like we reach a steady state of worker threads and arent continually creating and destroying them which is good. This could be a big deal on compute CPUs (CPUs isolated via isol_cpus or cpusets).
A few things I've found so far:
1. NR_CPUS > 32 causes issues with the workqueue debugfs code:
kernel/workqueue.c:3314: warning: left shift count >= width of type kernel/workqueue.c:3323: warning: left shift count >= width of type kernel/workqueue.c:3323: warning: integer overflow in expression kernel/workqueue.c:3324: warning: enumeration values exceed range of largest integer kernel/workqueue.c: In function ‘wq_debugfs_decode_pos’: kernel/workqueue.c:3336: warning: right shift count is negative kernel/workqueue.c:3337: warning: right shift count is negative kernel/workqueue.c: In function ‘wq_debugfs_next_pos’: kernel/workqueue.c:3435: warning: left shift count is negative kernel/workqueue.c:3436: warning: left shift count is negative kernel/workqueue.c: In function ‘wq_debugfs_start’: kernel/workqueue.c:3455: warning: left shift count >= width of type
2. cifs needs to be converted:
fs/cifs/cifsfs.c: In function ‘exit_cifs’: fs/cifs/cifsfs.c:1067: error: ‘system_single_wq’ undeclared (first use in this function) fs/cifs/cifsfs.c:1067: error: (Each undeclared identifier is reported only once fs/cifs/cifsfs.c:1067: error: for each function it appears in.)
Anton -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |