Kernel Profiling
WAR STORY: Kernel Profiling http://sourceware.org/systemtap/wiki/WSKernelProfile?highlight=((WarStories))
http://sourceware.org/systemtap/examples/process/pf2.stp
In , you can revise the wildcards used in the probe to target all kernel functions. This can be useful if you are interested in indentifying what the kernel is performing over a specific time period.
However, doing so can cause considerable stress on the machine. In addition, this does not provide any indication of whether a specific function is being called too often during small time increments.
This section describes how to profile the kernel properly. does this by providing a list of the top ten kernel functions called within a specific time period, and how many times each function was called during that time.
kernelprof.stp
#! /usr/bin/env stap
global profile, pcount
probe timer.profile {
pcount <<< 1
fn = probefunc ()
if (fn != "") profile[fn] <<< 1
}
probe timer.ms(5000) {
printf ("\n--- %d samples recorded:\n", @count(pcount))
foreach (f in profile- limit 10) {
printf ("%-30s\t%6d\n", f, @count(profile[f]))
}
delete profile
delete pcount
}
records and outputs kernel functions called every 5 seconds. You can change this setting by editing probe timer.ms(5000) accordingly. contains an excerpt of the output over a 20-second period:
<xref linkend="kernelprof"/> Sample Output
[...]
--- 10002 samples recorded:
sys_recvfrom 1
memmove 1
__copy_from_user_ll 17
__copy_to_user_ll 15
mwait_idle 5868
link_path_walk 1
kfree 1
fget_light 1
audit_syscall_exit 1
__d_lookup 1
--- 10002 samples recorded:
sysfs_read_file 1
free_poll_entry 1
syscall_exit_work 1
profile_hit 1
do_page_fault 2
_read_lock 3
kmap_atomic 4
strncpy_from_user 1
find_vma_prepare 1
__copy_from_user_ll 12
--- 10000 samples recorded:
system_call 4
unix_stream_sendmsg 1
__d_path 1
do_page_fault 2
kmap_atomic 1
find_vma 1
__copy_to_user_ll 16
__copy_from_user_ll 10
mwait_idle 5759
memcpy_fromiovec 1
--- 10004 samples recorded:
syscall_exit 1
__d_path 1
fput 2
do_page_fault 4
strncpy_from_user 1
may_open 1
ide_outb 1
ide_outsw 1
__copy_to_user_ll 11
__copy_from_user_ll 14
[...]
allows you to identify interesting kernel-specific performance facts about the system, such as kernel-intensive workloads and idle times (in this case, mwait_idle. Note that does not count user-space functions, although it counts all kernel functions as part of the sample (even if the script cannot identify from which specific kernel function a call originated.