On Thu, Jun 18, 2009 at 5:37 AM, Mikael Kjerrman<mikael.kjerr...@gmail.com> wrote: > Thanks for the help. > BTW this is what I see, [ ... ] > genunix`rm_assize+0xa0 > procfs`prgetpsinfo+0x3d0 > procfs`pr_read_psinfo+0x38 > genunix`fop_read+0x20 > genunix`pread+0x244 > unix`syscall_trap+0xac > 57 > [ ... ] > unix`page_exists+0x54 > genunix`vmu_calculate_seg+0x350 > genunix`vmu_calculate_proc+0x284 > genunix`vmu_calculate+0xf8 > genunix`vm_getusage+0x24c > unix`syscall_trap+0xac > 753
Okay, so prstat(1M) is spending some time in rm_assize(), which is what I was expecting to see, but it's spending the bulk of its time in vm_getusage(). Whoever's running prstat(1M) is using one of the options that aggregates resource utilization per-user (-a), per-project (-J), per-task (-T) or per-zone (-Z). The way things are now, this is a very expensive operation. Unfortunately, the only advice to give here is, "Don't do that." (This is something else I'd like to fix, but I haven't gotten to it yet.) Chad _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
