Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Support
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
B
bcc
Project overview
Project overview
Details
Activity
Releases
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Issues
0
Issues
0
List
Boards
Labels
Milestones
Merge Requests
0
Merge Requests
0
Analytics
Analytics
Repository
Value Stream
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Create a new issue
Commits
Issue Boards
Open sidebar
Kirill Smelkov
bcc
Commits
66bf2e8e
Commit
66bf2e8e
authored
Jul 31, 2016
by
Mark Drayton
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
offcputime: one symbol cache per process, improve pid/tid handling
parent
96483d48
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
41 additions
and
10 deletions
+41
-10
man/man8/offcputime.8
man/man8/offcputime.8
+12
-0
tools/offcputime.py
tools/offcputime.py
+25
-9
tools/offcputime_example.txt
tools/offcputime_example.txt
+4
-1
No files found.
man/man8/offcputime.8
View file @
66bf2e8e
...
@@ -40,6 +40,9 @@ Print output in folded stack format.
...
@@ -40,6 +40,9 @@ Print output in folded stack format.
\-p PID
\-p PID
Trace this process ID only (filtered in-kernel).
Trace this process ID only (filtered in-kernel).
.TP
.TP
\-t TID
Trace this thread ID only (filtered in-kernel).
.TP
\-u
\-u
Only trace user threads (no kernel threads).
Only trace user threads (no kernel threads).
.TP
.TP
...
@@ -52,6 +55,15 @@ Show stacks from user space only (no kernel space stacks).
...
@@ -52,6 +55,15 @@ Show stacks from user space only (no kernel space stacks).
\-K
\-K
Show stacks from kernel space only (no user space stacks).
Show stacks from kernel space only (no user space stacks).
.TP
.TP
\-d
Insert delimiter between kernel/user stacks.
.TP
\-f
Output folded format.
.TP
\-\-stack-storage-size STACK_STORAGE_SIZE
Change the number of unique stack traces that can be stored and displayed.
.TP
duration
duration
Duration to trace, in seconds.
Duration to trace, in seconds.
.SH EXAMPLES
.SH EXAMPLES
...
...
tools/offcputime.py
View file @
66bf2e8e
...
@@ -41,6 +41,7 @@ examples = """examples:
...
@@ -41,6 +41,7 @@ examples = """examples:
./offcputime 5 # trace for 5 seconds only
./offcputime 5 # trace for 5 seconds only
./offcputime -f 5 # 5 seconds, and output in folded format
./offcputime -f 5 # 5 seconds, and output in folded format
./offcputime -p 185 # only trace threads for PID 185
./offcputime -p 185 # only trace threads for PID 185
./offcputime -t 188 # only trace thread 188
./offcputime -u # only trace user threads (no kernel)
./offcputime -u # only trace user threads (no kernel)
./offcputime -k # only trace kernel threads (no user)
./offcputime -k # only trace kernel threads (no user)
./offcputime -U # only show user space stacks (no kernel)
./offcputime -U # only show user space stacks (no kernel)
...
@@ -51,8 +52,12 @@ parser = argparse.ArgumentParser(
...
@@ -51,8 +52,12 @@ parser = argparse.ArgumentParser(
formatter_class
=
argparse
.
RawDescriptionHelpFormatter
,
formatter_class
=
argparse
.
RawDescriptionHelpFormatter
,
epilog
=
examples
)
epilog
=
examples
)
thread_group
=
parser
.
add_mutually_exclusive_group
()
thread_group
=
parser
.
add_mutually_exclusive_group
()
thread_group
.
add_argument
(
"-p"
,
"--pid"
,
type
=
positive_int
,
# Note: this script provides --pid and --tid flags but their arguments are
help
=
"trace this PID only"
)
# referred to internally using kernel nomenclature: TGID and PID.
thread_group
.
add_argument
(
"-p"
,
"--pid"
,
metavar
=
"PID"
,
dest
=
"tgid"
,
help
=
"trace this PID only"
,
type
=
positive_int
)
thread_group
.
add_argument
(
"-t"
,
"--tid"
,
metavar
=
"TID"
,
dest
=
"pid"
,
help
=
"trace this TID only"
,
type
=
positive_int
)
thread_group
.
add_argument
(
"-u"
,
"--user-threads-only"
,
action
=
"store_true"
,
thread_group
.
add_argument
(
"-u"
,
"--user-threads-only"
,
action
=
"store_true"
,
help
=
"user threads only (no kernel threads)"
)
help
=
"user threads only (no kernel threads)"
)
thread_group
.
add_argument
(
"-k"
,
"--kernel-threads-only"
,
action
=
"store_true"
,
thread_group
.
add_argument
(
"-k"
,
"--kernel-threads-only"
,
action
=
"store_true"
,
...
@@ -68,12 +73,14 @@ parser.add_argument("-f", "--folded", action="store_true",
...
@@ -68,12 +73,14 @@ parser.add_argument("-f", "--folded", action="store_true",
help
=
"output folded format"
)
help
=
"output folded format"
)
parser
.
add_argument
(
"--stack-storage-size"
,
default
=
1024
,
parser
.
add_argument
(
"--stack-storage-size"
,
default
=
1024
,
type
=
positive_nonzero_int
,
type
=
positive_nonzero_int
,
help
=
"the number of unique stack traces that can be stored and "
\
help
=
"the number of unique stack traces that can be stored and "
"displayed (default 1024)"
)
"displayed (default 1024)"
)
parser
.
add_argument
(
"duration"
,
nargs
=
"?"
,
default
=
99999999
,
parser
.
add_argument
(
"duration"
,
nargs
=
"?"
,
default
=
99999999
,
type
=
positive_nonzero_int
,
type
=
positive_nonzero_int
,
help
=
"duration of trace, in seconds"
)
help
=
"duration of trace, in seconds"
)
args
=
parser
.
parse_args
()
args
=
parser
.
parse_args
()
if
args
.
pid
and
args
.
tgid
:
parser
.
error
(
"specify only one of -p and -t"
)
folded
=
args
.
folded
folded
=
args
.
folded
duration
=
int
(
args
.
duration
)
duration
=
int
(
args
.
duration
)
...
@@ -90,6 +97,7 @@ bpf_text = """
...
@@ -90,6 +97,7 @@ bpf_text = """
struct key_t {
struct key_t {
u32 pid;
u32 pid;
u32 tgid;
int user_stack_id;
int user_stack_id;
int kernel_stack_id;
int kernel_stack_id;
char name[TASK_COMM_LEN];
char name[TASK_COMM_LEN];
...
@@ -100,6 +108,7 @@ BPF_STACK_TRACE(stack_traces, STACK_STORAGE_SIZE)
...
@@ -100,6 +108,7 @@ BPF_STACK_TRACE(stack_traces, STACK_STORAGE_SIZE)
int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
u32 pid = prev->pid;
u32 pid = prev->pid;
u32 tgid = prev->tgid;
u64 ts, *tsp;
u64 ts, *tsp;
// record previous thread sleep time
// record previous thread sleep time
...
@@ -110,6 +119,7 @@ int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
...
@@ -110,6 +119,7 @@ int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
// get the current thread's start time
// get the current thread's start time
pid = bpf_get_current_pid_tgid();
pid = bpf_get_current_pid_tgid();
tgid = bpf_get_current_pid_tgid() >> 32;
tsp = start.lookup(&pid);
tsp = start.lookup(&pid);
if (tsp == 0) {
if (tsp == 0) {
return 0; // missed start or filtered
return 0; // missed start or filtered
...
@@ -128,6 +138,7 @@ int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
...
@@ -128,6 +138,7 @@ int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
struct key_t key = {};
struct key_t key = {};
key.pid = pid;
key.pid = pid;
key.tgid = tgid;
key.user_stack_id = USER_STACK_GET;
key.user_stack_id = USER_STACK_GET;
key.kernel_stack_id = KERNEL_STACK_GET;
key.kernel_stack_id = KERNEL_STACK_GET;
bpf_get_current_comm(&key.name, sizeof(key.name));
bpf_get_current_comm(&key.name, sizeof(key.name));
...
@@ -140,9 +151,12 @@ int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
...
@@ -140,9 +151,12 @@ int oncpu(struct pt_regs *ctx, struct task_struct *prev) {
# set thread filter
# set thread filter
thread_context
=
""
thread_context
=
""
if
args
.
pid
is
not
None
:
if
args
.
tgid
is
not
None
:
thread_context
=
"PID %s"
%
args
.
pid
thread_context
=
"PID %d"
%
args
.
tgid
thread_filter
=
'pid == %s'
%
args
.
pid
thread_filter
=
'tgid == %d'
%
args
.
tgid
elif
args
.
pid
is
not
None
:
thread_context
=
"TID %d"
%
args
.
pid
thread_filter
=
'pid == %d'
%
args
.
pid
elif
args
.
user_threads_only
:
elif
args
.
user_threads_only
:
thread_context
=
"user threads"
thread_context
=
"user threads"
thread_filter
=
'!(prev->flags & PF_KTHREAD)'
thread_filter
=
'!(prev->flags & PF_KTHREAD)'
...
@@ -224,6 +238,8 @@ for k, v in sorted(counts.items(), key=lambda counts: counts[1].value):
...
@@ -224,6 +238,8 @@ for k, v in sorted(counts.items(), key=lambda counts: counts[1].value):
has_enomem
=
True
has_enomem
=
True
continue
continue
# user stacks will be symbolized by tgid, not pid, to avoid the overhead
# of one symbol resolver per thread
user_stack
=
[]
if
k
.
user_stack_id
<
0
else
\
user_stack
=
[]
if
k
.
user_stack_id
<
0
else
\
stack_traces
.
walk
(
k
.
user_stack_id
)
stack_traces
.
walk
(
k
.
user_stack_id
)
kernel_stack
=
[]
if
k
.
kernel_stack_id
<
0
else
\
kernel_stack
=
[]
if
k
.
kernel_stack_id
<
0
else
\
...
@@ -234,7 +250,7 @@ for k, v in sorted(counts.items(), key=lambda counts: counts[1].value):
...
@@ -234,7 +250,7 @@ for k, v in sorted(counts.items(), key=lambda counts: counts[1].value):
user_stack
=
list
(
user_stack
)
user_stack
=
list
(
user_stack
)
kernel_stack
=
list
(
kernel_stack
)
kernel_stack
=
list
(
kernel_stack
)
line
=
[
k
.
name
.
decode
()]
+
\
line
=
[
k
.
name
.
decode
()]
+
\
[
b
.
sym
(
addr
,
k
.
p
id
)
for
addr
in
reversed
(
user_stack
)]
+
\
[
b
.
sym
(
addr
,
k
.
tg
id
)
for
addr
in
reversed
(
user_stack
)]
+
\
(
need_delimiter
and
[
"-"
]
or
[])
+
\
(
need_delimiter
and
[
"-"
]
or
[])
+
\
[
b
.
ksym
(
addr
)
for
addr
in
reversed
(
kernel_stack
)]
[
b
.
ksym
(
addr
)
for
addr
in
reversed
(
kernel_stack
)]
print
(
"%s %d"
%
(
";"
.
join
(
line
),
v
.
value
))
print
(
"%s %d"
%
(
";"
.
join
(
line
),
v
.
value
))
...
@@ -245,7 +261,7 @@ for k, v in sorted(counts.items(), key=lambda counts: counts[1].value):
...
@@ -245,7 +261,7 @@ for k, v in sorted(counts.items(), key=lambda counts: counts[1].value):
if
need_delimiter
:
if
need_delimiter
:
print
(
" --"
)
print
(
" --"
)
for
addr
in
user_stack
:
for
addr
in
user_stack
:
print
(
" %016x %s"
%
(
addr
,
b
.
sym
(
addr
,
k
.
p
id
)))
print
(
" %016x %s"
%
(
addr
,
b
.
sym
(
addr
,
k
.
tg
id
)))
print
(
" %-16s %s (%d)"
%
(
"-"
,
k
.
name
,
k
.
pid
))
print
(
" %-16s %s (%d)"
%
(
"-"
,
k
.
name
,
k
.
pid
))
print
(
" %d
\
n
"
%
v
.
value
)
print
(
" %d
\
n
"
%
v
.
value
)
...
...
tools/offcputime_example.txt
View file @
66bf2e8e
...
@@ -719,7 +719,7 @@ creating your "off-CPU time flame graphs".
...
@@ -719,7 +719,7 @@ creating your "off-CPU time flame graphs".
USAGE message:
USAGE message:
# ./offcputime -h
# ./offcputime -h
usage: offcputime.py [-h] [-p PID | -
k | -u] [-K | -U
] [-f]
usage: offcputime.py [-h] [-p PID | -
t TID | -u | -k] [-U | -K] [-d
] [-f]
[--stack-storage-size STACK_STORAGE_SIZE]
[--stack-storage-size STACK_STORAGE_SIZE]
[duration]
[duration]
...
@@ -731,6 +731,7 @@ positional arguments:
...
@@ -731,6 +731,7 @@ positional arguments:
optional arguments:
optional arguments:
-h, --help show this help message and exit
-h, --help show this help message and exit
-p PID, --pid PID trace this PID only
-p PID, --pid PID trace this PID only
-t TID, --tid TID trace this TID only
-u, --user-threads-only
-u, --user-threads-only
user threads only (no kernel threads)
user threads only (no kernel threads)
-k, --kernel-threads-only
-k, --kernel-threads-only
...
@@ -741,6 +742,7 @@ optional arguments:
...
@@ -741,6 +742,7 @@ optional arguments:
-K, --kernel-stacks-only
-K, --kernel-stacks-only
show stacks from kernel space only (no user space
show stacks from kernel space only (no user space
stacks)
stacks)
-d, --delimited insert delimiter between kernel/user stacks
-f, --folded output folded format
-f, --folded output folded format
--stack-storage-size STACK_STORAGE_SIZE
--stack-storage-size STACK_STORAGE_SIZE
the number of unique stack traces that can be stored
the number of unique stack traces that can be stored
...
@@ -751,6 +753,7 @@ examples:
...
@@ -751,6 +753,7 @@ examples:
./offcputime 5 # trace for 5 seconds only
./offcputime 5 # trace for 5 seconds only
./offcputime -f 5 # 5 seconds, and output in folded format
./offcputime -f 5 # 5 seconds, and output in folded format
./offcputime -p 185 # only trace threads for PID 185
./offcputime -p 185 # only trace threads for PID 185
./offcputime -t 188 # only trace thread 188
./offcputime -u # only trace user threads (no kernel)
./offcputime -u # only trace user threads (no kernel)
./offcputime -k # only trace kernel threads (no user)
./offcputime -k # only trace kernel threads (no user)
./offcputime -U # only show user space stacks (no kernel)
./offcputime -U # only show user space stacks (no kernel)
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment