Skip to content

B
bcc
Commit 7df65da7 authored by Sasha Goldshtein's avatar Sasha Goldshtein
Browse files
Options

Added small list of automatic includes to make the -I switch less necessary

parent 32c89978
Show whitespace changes
Inline Side-by-side
Showing with 72 additions and 55 deletions
man/man8/argdist.8
View file @ 7df65da7
......@@ -45,7 +45,8 @@ any filtering. See SPECIFIER SYNTAX below.
One or more header files that should be included in the BPF program. This
enables the use of structure definitions, enumerations, and constants that
are available in these headers. You should provide the same path you would
include in the BPF program, e.g. 'linux/blkdev.h' or 'linux/time.h'.
include in the BPF program, e.g. 'linux/blkdev.h' or 'linux/time.h'. Note: in
many cases, argdist will deduce the necessary header files automatically.
.SH SPECIFIER SYNTAX
The general specifier syntax is as follows:
......@@ -106,47 +107,47 @@ this is the probe specifier.
.TP
Print a histogram of allocation sizes passed to kmalloc:
#
.B argdist.py -H 'p::__kmalloc(u64 size):u64:size'
.B argdist -H 'p::__kmalloc(u64 size):u64:size'
.TP
Print a count of how many times process 1005 called malloc with an allocation size of 16 bytes:
#
.B argdist.py -p 1005 -C 'p:c:malloc(size_t size):size_t:size:size==16'
.B argdist -p 1005 -C 'p:c:malloc(size_t size):size_t:size:size==16'
.TP
Snoop on all strings returned by gets():
#
.B argdist.py -C 'r:c:gets():char*:$retval'
.B argdist -C 'r:c:gets():char*:$retval'
.TP
Print a histogram of read sizes that were longer than 1ms:
#
.B argdist.py -H 'r::__vfs_read(void *file, void *buf, size_t count):size_t:$entry(count):$latency > 1000000'
.B argdist -H 'r::__vfs_read(void *file, void *buf, size_t count):size_t:$entry(count):$latency > 1000000'
.TP
Print frequency counts of how many times writes were issued to a particular file descriptor number, in process 1005:
#
.B argdist.py -p 1005 -C 'p:c:write(int fd):int:fd'
.B argdist -p 1005 -C 'p:c:write(int fd):int:fd'
.TP
Print a histogram of error codes returned by read() in process 1005:
#
.B argdist.py -p 1005 -H 'r:c:read()'
.B argdist -p 1005 -H 'r:c:read()'
.TP
Print a histogram of buffer sizes passed to write() across all processes, where the file descriptor was 1 (STDOUT):
#
.B argdist.py -H 'p:c:write(int fd, const void *buf, size_t count):size_t:count:fd==1'
.B argdist -H 'p:c:write(int fd, const void *buf, size_t count):size_t:count:fd==1'
.TP
Count fork() calls in libc across all processes, grouped by pid:
#
.B argdist.py -C 'p:c:fork():int:$PID;fork per process'
.B argdist -C 'p:c:fork():int:$PID;fork per process'
.TP
Print histograms of sleep() and nanosleep() parameter values:
#
.B argdist.py -I 'linux/time.h' -H 'p:c:sleep(u32 seconds):u32:seconds' 'p:c:nanosleep(struct timespec *req):long:req->tv_nsec'
.B argdist -H 'p:c:sleep(u32 seconds):u32:seconds' 'p:c:nanosleep(struct timespec *req):long:req->tv_nsec'
.TP
Spy on writes to STDOUT performed by process 2780, up to a string size of 120 characters:
#
.B argdist.py -p 2780 -z 120 -C 'p:c:write(int fd, char* buf, size_t len):char*:buf:fd==1'
.B argdist -p 2780 -z 120 -C 'p:c:write(int fd, char* buf, size_t len):char*:buf:fd==1'
.TP
Group files being read from and the read sizes from __vfs_read:
#
.B argdist.py -I 'linux/fs.h' -C 'p::__vfs_read(struct file *file, void *buf, size_t count):char*,size_t:file->f_path.dentry->d_iname,count:file->f_path.dentry->d_iname[0]!=0'
.B argdist -C 'p::__vfs_read(struct file *file, void *buf, size_t count):char*,size_t:file->f_path.dentry->d_iname,count:file->f_path.dentry->d_iname[0]!=0'
.SH SOURCE
This is from bcc.
.IP
......
tools/argdist.py
View file @ 7df65da7
#!/usr/bin/env python
#
# argdist.py Trace a function and display a distribution of its
# argdist Trace a function and display a distribution of its
# parameter values as a histogram or frequency count.
#
# USAGE: argdist.py [-h] [-p PID] [-z STRING_SIZE] [-i INTERVAL]
# USAGE: argdist [-h] [-p PID] [-z STRING_SIZE] [-i INTERVAL]
# [-n COUNT] [-v] [-T TOP]
# [-C specifier [specifier ...]]
# [-H specifier [specifier ...]]
......@@ -33,6 +33,23 @@ int PROBENAME(struct pt_regs *ctx SIGNATURE)
"""
next_probe_index = 0
aliases = { "$PID": "bpf_get_current_pid_tgid()" }
auto_includes = {
"linux/time.h" : ["time"],
"linux/fs.h" : ["fs", "file"],
"linux/blkdev.h" : ["bio", "request"],
"linux/slab.h" : ["alloc"]
}
@staticmethod
def generate_auto_includes(specifiers):
headers = ""
for header, keywords in Specifier.auto_includes.items():
for keyword in keywords:
for specifier in specifiers:
if keyword in specifier:
headers += "#include <%s>\n" \
% header
return headers
def _substitute_aliases(self, expr):
if expr is None:
......@@ -443,52 +460,51 @@ Where:
EXAMPLES:
argdist.py -H 'p::__kmalloc(u64 size):u64:size'
argdist -H 'p::__kmalloc(u64 size):u64:size'
Print a histogram of allocation sizes passed to kmalloc
argdist.py -p 1005 -C 'p:c:malloc(size_t size):size_t:size:size==16'
argdist -p 1005 -C 'p:c:malloc(size_t size):size_t:size:size==16'
Print a frequency count of how many times process 1005 called malloc
with an allocation size of 16 bytes
argdist.py -C 'r:c:gets():char*:(char*)$retval#snooped strings'
argdist -C 'r:c:gets():char*:(char*)$retval#snooped strings'
Snoop on all strings returned by gets()
argdist.py -H 'r::__kmalloc(size_t size):u64:$latency/$entry(size)#ns per byte'
argdist -H 'r::__kmalloc(size_t size):u64:$latency/$entry(size)#ns per byte'
Print a histogram of nanoseconds per byte from kmalloc allocations
argdist.py -I 'linux/slab.h' \\
-C 'p::__kmalloc(size_t size, gfp_t flags):size_t:size:flags&GFP_ATOMIC'
argdist -C 'p::__kmalloc(size_t size, gfp_t flags):size_t:size:flags&GFP_ATOMIC'
Print frequency count of kmalloc allocation sizes that have GFP_ATOMIC
argdist.py -p 1005 -C 'p:c:write(int fd):int:fd' -T 5
argdist -p 1005 -C 'p:c:write(int fd):int:fd' -T 5
Print frequency counts of how many times writes were issued to a
particular file descriptor number, in process 1005, but only show
the top 5 busiest fds
argdist.py -p 1005 -H 'r:c:read()'
argdist -p 1005 -H 'r:c:read()'
Print a histogram of results (sizes) returned by read() in process 1005
argdist.py -C 'r::__vfs_read():u32:$PID:$latency > 100000'
argdist -C 'r::__vfs_read():u32:$PID:$latency > 100000'
Print frequency of reads by process where the latency was >0.1ms
argdist.py -H 'r::__vfs_read(void *file, void *buf, size_t count):size_t:$entry(count):$latency > 1000000'
argdist -H 'r::__vfs_read(void *file, void *buf, size_t count):size_t:$entry(count):$latency > 1000000'
Print a histogram of read sizes that were longer than 1ms
argdist.py -H \\
argdist -H \\
'p:c:write(int fd, const void *buf, size_t count):size_t:count:fd==1'
Print a histogram of buffer sizes passed to write() across all
processes, where the file descriptor was 1 (STDOUT)
argdist.py -C 'p:c:fork()#fork calls'
argdist -C 'p:c:fork()#fork calls'
Count fork() calls in libc across all processes
Can also use funccount.py, which is easier and more flexible
argdist.py -I 'linux/time.h' -H \\
argdist -H \\
'p:c:sleep(u32 seconds):u32:seconds' \\
'p:c:nanosleep(struct timespec *req):long:req->tv_nsec'
Print histograms of sleep() and nanosleep() parameter values
argdist.py -p 2780 -z 120 \\
argdist -p 2780 -z 120 \\
-C 'p:c:write(int fd, char* buf, size_t len):char*:buf:fd==1'
Spy on writes to STDOUT performed by process 2780, up to a string size
of 120 characters
......@@ -536,6 +552,7 @@ struct __string_t { char s[%d]; };
""" % args.string_size
for include in (args.include or []):
bpf_source += "#include <%s>\n" % include
bpf_source += Specifier.generate_auto_includes(map(lambda s: s.raw_spec, specifiers))
for specifier in specifiers:
bpf_source += specifier.generate_text()
......
tools/argdist_examples.txt
View file @ 7df65da7
......@@ -10,7 +10,7 @@ various functions.
For example, suppose you want to find what allocation sizes are common in
your application:
# ./argdist.py -p 2420 -C 'p:c:malloc(size_t size):size_t:size'
# ./argdist -p 2420 -C 'p:c:malloc(size_t size):size_t:size'
[01:42:29]
p:c:malloc(size_t size):size_t:size
COUNT EVENT
......@@ -43,7 +43,7 @@ probed and its value was 16, repeatedly.
Now, suppose you wanted a histogram of buffer sizes passed to the write()
function across the system:
# ./argdist.py -H 'p:c:write(int fd, void *buf, size_t len):size_t:len'
# ./argdist -H 'p:c:write(int fd, void *buf, size_t len):size_t:len'
[01:45:22]
p:c:write(int fd, void *buf, size_t len):size_t:len
len : count distribution
......@@ -81,7 +81,7 @@ bytes, medium writes of 32-63 bytes, and larger writes of 64-127 bytes.
But these are writes across the board -- what if you wanted to focus on writes
to STDOUT?
# ./argdist.py -H 'p:c:write(int fd, void *buf, size_t len):size_t:len:fd==1'
# ./argdist -H 'p:c:write(int fd, void *buf, size_t len):size_t:len:fd==1'
[01:47:17]
p:c:write(int fd, void *buf, size_t len):size_t:len:fd==1
len : count distribution
......@@ -120,7 +120,7 @@ You can also use argdist to trace kernel functions. For example, suppose you
wanted a histogram of kernel allocation (kmalloc) sizes across the system,
printed twice with 3 second intervals:
# ./argdist.py -i 3 -n 2 -H 'p::__kmalloc(size_t size):size_t:size'
# ./argdist -i 3 -n 2 -H 'p::__kmalloc(size_t size):size_t:size'
[01:50:00]
p::__kmalloc(size_t size):size_t:size
size : count distribution
......@@ -143,7 +143,7 @@ p::__kmalloc(size_t size):size_t:size
Occasionally, numeric information isn't enough and you want to capture strings.
What are the strings printed by puts() across the system?
# ./argdist.py -i 10 -n 1 -C 'p:c:puts(char *str):char*:str'
# ./argdist -i 10 -n 1 -C 'p:c:puts(char *str):char*:str'
[01:53:54]
p:c:puts(char *str):char*:str
COUNT EVENT
......@@ -156,7 +156,7 @@ What about reads? You could trace gets() across the system and print the
strings input by the user (note how "r" is used instead of "p" to attach a
probe to the function's return):
# ./argdist.py -i 10 -n 1 -C 'r:c:gets():char*:(char*)$retval:$retval!=0'
# ./argdist -i 10 -n 1 -C 'r:c:gets():char*:(char*)$retval:$retval!=0'
[02:12:23]
r:c:gets():char*:$retval:$retval!=0
COUNT EVENT
......@@ -166,7 +166,7 @@ r:c:gets():char*:$retval:$retval!=0
Similarly, we could get a histogram of the error codes returned by read():
# ./argdist.py -i 10 -c 1 -H 'r:c:read()'
# ./argdist -i 10 -c 1 -H 'r:c:read()'
[02:15:36]
r:c:read()
retval : count distribution
......@@ -188,7 +188,7 @@ recursive) and the parameters it had on entry. For example, we can identify
which processes are performing slow synchronous filesystem reads -- say,
longer than 0.1ms (100,000ns):
# ./argdist.py -C 'r::__vfs_read():u32:$PID:$latency > 100000'
# ./argdist -C 'r::__vfs_read():u32:$PID:$latency > 100000'
[01:08:48]
r::__vfs_read():u32:$PID:$latency > 100000
COUNT EVENT
......@@ -207,7 +207,7 @@ Occasionally, entry parameter values are also interesting. For example, you
might be curious how long it takes malloc() to allocate memory -- nanoseconds
per byte allocated. Let's go:
# ./argdist.py -H 'r:c:malloc(size_t size):u64:$latency/$entry(size);ns per byte' -n 1 -i 10
# ./argdist -H 'r:c:malloc(size_t size):u64:$latency/$entry(size);ns per byte' -n 1 -i 10
[01:11:13]
ns per byte : count distribution
0 -> 1 : 0 | |
......@@ -232,7 +232,7 @@ multiple microseconds per byte.
You could also group results by more than one field. For example, __kmalloc
takes an additional flags parameter that describes how to allocate memory:
# ./argdist.py -I 'linux/slab.h' -C 'p::__kmalloc(size_t size, gfp_t flags):gfp_t,size_t:flags,size'
# ./argdist -C 'p::__kmalloc(size_t size, gfp_t flags):gfp_t,size_t:flags,size'
[03:42:29]
p::__kmalloc(size_t size, gfp_t flags):gfp_t,size_t:flags,size
COUNT EVENT
......@@ -265,7 +265,7 @@ certain kinds of allocations or visually group them together.
Here's a final example that finds how many write() system calls are performed
by each process on the system:
# argdist.py -C 'p:c:write():int:$PID;write per process' -n 2
# argdist -C 'p:c:write():int:$PID;write per process' -n 2
[06:47:18]
write by process
COUNT EVENT
......@@ -282,8 +282,8 @@ write by process
USAGE message:
# argdist.py -h
usage: argdist.py [-h] [-p PID] [-z STRING_SIZE] [-i INTERVAL] [-n COUNT] [-v]
# argdist -h
usage: argdist [-h] [-p PID] [-z STRING_SIZE] [-i INTERVAL] [-n COUNT] [-v]
[-T TOP] [-H [specifier [specifier ...]]]
[-C [specifier [specifier ...]]] [-I [header [header ...]]]
......@@ -326,52 +326,51 @@ Where:
EXAMPLES:
argdist.py -H 'p::__kmalloc(u64 size):u64:size'
argdist -H 'p::__kmalloc(u64 size):u64:size'
Print a histogram of allocation sizes passed to kmalloc
argdist.py -p 1005 -C 'p:c:malloc(size_t size):size_t:size:size==16'
argdist -p 1005 -C 'p:c:malloc(size_t size):size_t:size:size==16'
Print a frequency count of how many times process 1005 called malloc
with an allocation size of 16 bytes
argdist.py -C 'r:c:gets():char*:$retval#snooped strings'
argdist -C 'r:c:gets():char*:$retval#snooped strings'
Snoop on all strings returned by gets()
argdist.py -H 'r::__kmalloc(size_t size):u64:$latency/$entry(size)#ns per byte'
argdist -H 'r::__kmalloc(size_t size):u64:$latency/$entry(size)#ns per byte'
Print a histogram of nanoseconds per byte from kmalloc allocations
argdist.py -I 'linux/slab.h' \
-C 'p::__kmalloc(size_t size, gfp_t flags):size_t:size:flags&GFP_ATOMIC'
argdist -C 'p::__kmalloc(size_t size, gfp_t flags):size_t:size:flags&GFP_ATOMIC'
Print frequency count of kmalloc allocation sizes that have GFP_ATOMIC
argdist.py -p 1005 -C 'p:c:write(int fd):int:fd' -T 5
argdist -p 1005 -C 'p:c:write(int fd):int:fd' -T 5
Print frequency counts of how many times writes were issued to a
particular file descriptor number, in process 1005, but only show
the top 5 busiest fds
argdist.py -p 1005 -H 'r:c:read()'
argdist -p 1005 -H 'r:c:read()'
Print a histogram of error codes returned by read() in process 1005
argdist.py -C 'r::__vfs_read():u32:$PID:$latency > 100000'
argdist -C 'r::__vfs_read():u32:$PID:$latency > 100000'
Print frequency of reads by process where the latency was >0.1ms
argdist.py -H 'r::__vfs_read(void *file, void *buf, size_t count):size_t:$entry(count):$latency > 1000000'
argdist -H 'r::__vfs_read(void *file, void *buf, size_t count):size_t:$entry(count):$latency > 1000000'
Print a histogram of read sizes that were longer than 1ms
argdist.py -H \
argdist -H \
'p:c:write(int fd, const void *buf, size_t count):size_t:count:fd==1'
Print a histogram of buffer sizes passed to write() across all
processes, where the file descriptor was 1 (STDOUT)
argdist.py -C 'p:c:fork()#fork calls'
argdist -C 'p:c:fork()#fork calls'
Count fork() calls in libc across all processes
Can also use funccount.py, which is easier and more flexible
argdist.py -I 'linux/time.h' -H \
argdist -H \
'p:c:sleep(u32 seconds):u32:seconds' \
'p:c:nanosleep(struct timespec *req):long:req->tv_nsec'
Print histograms of sleep() and nanosleep() parameter values
argdist.py -p 2780 -z 120 \
argdist -p 2780 -z 120 \
-C 'p:c:write(int fd, char* buf, size_t len):char*:buf:fd==1'
Spy on writes to STDOUT performed by process 2780, up to a string size
of 120 characters
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7