Merge pull request #2894 from sumanthkorikkar/bcc-tools-fixes

Bcc tools fixe for trace.py and argdist.py for bpf_probe_read_user.

src/cc/libbpf.c

@@ -1168,7 +1168,7 @@ int bpf_attach_raw_tracepoint(int progfd, const char *tp_name)
 
 bool bpf_has_kernel_btf(void)
 {
- return libbpf_find_vmlinux_btf_id("bpf_prog_put", 0);
+ return libbpf_find_vmlinux_btf_id("bpf_prog_put", 0) > 0;
 }
 
 int bpf_detach_kfunc(int prog_fd, char *func)

src/python/bcc/__init__.py

@@ -26,7 +26,7 @@
 from .libbcc import lib, bcc_symbol, bcc_symbol_option, bcc_stacktrace_build_id, _SYM_CB_TYPE
 from .table import Table, PerfEventArray
 from .perf import Perf
-from .utils import get_online_cpus, printb, _assert_is_bytes, ArgString
+from .utils import get_online_cpus, printb, _assert_is_bytes, ArgString, StrcmpRewrite
 from .version import __version__
 from .disassembler import disassemble_prog, decode_map
 

src/python/bcc/utils.py

@@ -15,6 +15,7 @@
 import sys
 import traceback
 import warnings
+import re
 
 from .libbcc import lib
 
@@ -97,3 +98,50 @@ def _assert_is_bytes(arg):
  return ArgString(arg).__bytes__()
  return arg
 
+class StrcmpRewrite(object):
+ @staticmethod
+ def _generate_streq_function(string, probe_read_func, streq_functions,
+ probeid):
+ fname = "streq_%d" % probeid
+ streq_functions += """
+static inline bool %s(char const *ignored, uintptr_t str) {
+ char needle[] = %s;
+ char haystack[sizeof(needle)];
+ %s(&haystack, sizeof(haystack), (void *)str);
+ for (int i = 0; i < sizeof(needle) - 1; ++i) {
+ if (needle[i] != haystack[i]) {
+ return false;
+ }
+ }
+ return true;
+}
+ """ % (fname, string, probe_read_func)
+ return fname, streq_functions
+
+ @staticmethod
+ def rewrite_expr(expr, bin_cmp, is_user, probe_user_list, streq_functions,
+ probeid):
+ if bin_cmp:
+ STRCMP_RE = 'STRCMP\\(\"([^"]+)\\",(.+?)\\)'
+ else:
+ STRCMP_RE = 'STRCMP\\(("[^"]+\\"),(.+?)\\)'
+ matches = re.finditer(STRCMP_RE, expr)
+ for match in matches:
+ string = match.group(1)
+ probe_read_func = "bpf_probe_read"
+ # if user probe or @user tag is specified, use
+ # bpf_probe_read_user for char* read
+ if is_user or \
+ match.group(2).strip() in probe_user_list:
+ probe_read_func = "bpf_probe_read_user"
+ fname, streq_functions = StrcmpRewrite._generate_streq_function(
+ string, probe_read_func,
+ streq_functions, probeid)
+ probeid += 1
+ expr = expr.replace("STRCMP", fname, 1)
+ rdict = {
+ "expr" : expr,
+ "streq_functions" : streq_functions,
+ "probeid" : probeid
+ }
+ return rdict

tools/argdist.py

@@ -9,7 +9,7 @@
 # Licensed under the Apache License, Version 2.0 (the "License")
 # Copyright (C) 2016 Sasha Goldshtein.
 
-from bcc import BPF, USDT
+from bcc import BPF, USDT, StrcmpRewrite
 from time import sleep, strftime
 import argparse
 import re
@@ -41,6 +41,10 @@ def _parse_signature(self):
  param_type = param[0:index + 1].strip()
  param_name = param[index + 1:].strip()
  self.param_types[param_name] = param_type
+ # Maintain list of user params. Then later decide to
+ # switch to bpf_probe_read or bpf_probe_read_user.
+ if "__user" in param_type.split():
+ self.probe_user_list.add(param_name)
 
  def _generate_entry(self):
  self.entry_probe_func = self.probe_func_name + "_entry"
@@ -182,6 +186,8 @@ def __init__(self, tool, type, specifier):
  self.pid = tool.args.pid
  self.cumulative = tool.args.cumulative or False
  self.raw_spec = specifier
+ self.probe_user_list = set()
+ self.bin_cmp = False
  self._validate_specifier()
 
  spec_and_label = specifier.split('#')
@@ -250,32 +256,16 @@ def _enable_usdt_probe(self):
  self.usdt_ctx.enable_probe(
  self.function, self.probe_func_name)
 
- def _generate_streq_function(self, string):
- fname = "streq_%d" % Probe.streq_index
- Probe.streq_index += 1
- self.streq_functions += """
-static inline bool %s(char const *ignored, char const *str) {
- char needle[] = %s;
- char haystack[sizeof(needle)];
- bpf_probe_read(&haystack, sizeof(haystack), (void *)str);
- for (int i = 0; i < sizeof(needle) - 1; ++i) {
- if (needle[i] != haystack[i]) {
- return false;
- }
- }
- return true;
-}
- """ % (fname, string)
- return fname
-
  def _substitute_exprs(self):
  def repl(expr):
  expr = self._substitute_aliases(expr)
- matches = re.finditer('STRCMP\\(("[^"]+\\")', expr)
- for match in matches:
- string = match.group(1)
- fname = self._generate_streq_function(string)
- expr = expr.replace("STRCMP", fname, 1)
+ rdict = StrcmpRewrite.rewrite_expr(expr,
+ self.bin_cmp, self.library,
+ self.probe_user_list, self.streq_functions,
+ Probe.streq_index)
+ expr = rdict["expr"]
+ self.streq_functions = rdict["streq_functions"]
+ Probe.streq_index = rdict["probeid"]
  return expr.replace("$retval", "PT_REGS_RC(ctx)")
  for i in range(0, len(self.exprs)):
  self.exprs[i] = repl(self.exprs[i])
@@ -305,9 +295,14 @@ def _generate_usdt_arg_assignment(self, i):
  def _generate_field_assignment(self, i):
  text = self._generate_usdt_arg_assignment(i)
  if self._is_string(self.expr_types[i]):
- return (text + " bpf_probe_read(&__key.v%d.s," +
+ if self.is_user or \
+ self.exprs[i] in self.probe_user_list:
+ probe_readfunc = "bpf_probe_read_user"
+ else:
+ probe_readfunc = "bpf_probe_read"
+ return (text + " %s(&__key.v%d.s," +
  " sizeof(__key.v%d.s), (void *)%s);\n") % \
- (i, i, self.exprs[i])
+ (probe_readfunc, i, i, self.exprs[i])
  else:
  return text + " __key.v%d = %s;\n" % \
  (i, self.exprs[i])

tools/argdist_example.txt

@@ -449,3 +449,12 @@ argdist -I 'kernel/sched/sched.h' \
  in kernel/sched/sched.h which is in kernel source tree and not in kernel-devel
  package. So this command needs to run at the kernel source tree root directory
  so that the added header file can be found by the compiler.
+
+argdist -C 'p::do_sys_open(int dfd, const char __user *filename, int flags,
+ umode_t mode):char*:filename:STRCMP("sample.txt", filename)'
+ Trace open of the file "sample.txt". It should be noted that 'filename'
+ passed to the do_sys_open is a char * user pointer. Hence parameter
+ 'filename' should be tagged with __user for kprobes (const char __user
+ *filename). This information distinguishes if the 'filename' should be
+ copied from userspace to the bpf stack or from kernel space to the bpf
+ stack.

tools/trace.py

@@ -11,7 +11,7 @@
 # Copyright (C) 2016 Sasha Goldshtein.
 
 from __future__ import print_function
-from bcc import BPF, USDT
+from bcc import BPF, USDT, StrcmpRewrite
 from functools import partial
 from time import sleep, strftime
 import time
@@ -65,6 +65,7 @@ def __init__(self, probe, string_size, kernel_stack, user_stack,
  self.string_size = string_size
  self.kernel_stack = kernel_stack
  self.user_stack = user_stack
+ self.probe_user_list = set()
  Probe.probe_count += 1
  self._parse_probe()
  self.probe_num = Probe.probe_count
@@ -260,47 +261,33 @@ def _parse_action(self, action):
  "$task" : "((struct task_struct *)bpf_get_current_task())"
  }
 
- def _generate_streq_function(self, string):
- fname = "streq_%d" % Probe.streq_index
- Probe.streq_index += 1
- self.streq_functions += """
-static inline bool %s(char const *ignored, uintptr_t str) {
- char needle[] = %s;
- char haystack[sizeof(needle)];
- bpf_probe_read(&haystack, sizeof(haystack), (void *)str);
- for (int i = 0; i < sizeof(needle) - 1; ++i) {
- if (needle[i] != haystack[i]) {
- return false;
- }
- }
- return true;
-}
- """ % (fname, string)
- return fname
-
  def _rewrite_expr(self, expr):
- if self.is_syscall_kprobe:
- for alias, replacement in Probe.aliases_indarg.items():
- expr = expr.replace(alias, replacement)
- else:
- for alias, replacement in Probe.aliases_arg.items():
- # For USDT probes, we replace argN values with the
- # actual arguments for that probe obtained using
- # bpf_readarg_N macros emitted at BPF construction.
- if self.probe_type == "u":
- continue
+ # Find the occurances of any arg[1-6]@user. Use it later to
+ # identify bpf_probe_read_user
+ for matches in re.finditer(r'(arg[1-6])(@user)', expr):
+ if matches.group(1).strip() not in self.probe_user_list:
+ self.probe_user_list.add(matches.group(1).strip())
+ # Remove @user occurrences from arg before resolving to its
+ # corresponding aliases.
+ expr = re.sub(r'(arg[1-6])@user', r'\1', expr)
+ rdict = StrcmpRewrite.rewrite_expr(expr,
+ self.bin_cmp, self.library,
+ self.probe_user_list, self.streq_functions,
+ Probe.streq_index)
+ expr = rdict["expr"]
+ self.streq_functions = rdict["streq_functions"]
+ Probe.streq_index = rdict["probeid"]
+ alias_to_check = Probe.aliases_indarg \
+ if self.is_syscall_kprobe \
+ else Probe.aliases_arg
+ # For USDT probes, we replace argN values with the
+ # actual arguments for that probe obtained using
+ # bpf_readarg_N macros emitted at BPF construction.
+ if not self.probe_type == "u":
+ for alias, replacement in alias_to_check.items():
  expr = expr.replace(alias, replacement)
  for alias, replacement in Probe.aliases_common.items():
  expr = expr.replace(alias, replacement)
- if self.bin_cmp:
- STRCMP_RE = 'STRCMP\\(\"([^"]+)\\"'
- else:
- STRCMP_RE = 'STRCMP\\(("[^"]+\\")'
- matches = re.finditer(STRCMP_RE, expr)
- for match in matches:
- string = match.group(1)
- fname = self._generate_streq_function(string)
- expr = expr.replace("STRCMP", fname, 1)
  return expr
 
  p_type = {"u": ct.c_uint, "d": ct.c_int, "lu": ct.c_ulong,
@@ -412,14 +399,24 @@ def _generate_field_assign(self, idx):
  text = (" %s %s = 0;\n" +
  " bpf_usdt_readarg(%s, ctx, &%s);\n") \
  % (arg_ctype, expr, expr[3], expr)
-
+ probe_read_func = "bpf_probe_read"
  if field_type == "s":
+ if self.library:
+ probe_read_func = "bpf_probe_read_user"
+ else:
+ alias_to_check = Probe.aliases_indarg \
+ if self.is_syscall_kprobe \
+ else Probe.aliases_arg
+ for arg, alias in alias_to_check.items():
+ if alias == expr and arg in self.probe_user_list:
+ probe_read_func = "bpf_probe_read_user"
+ break
  return text + """
  if (%s != 0) {
  void *__tmp = (void *)%s;
- bpf_probe_read(&__data.v%d, sizeof(__data.v%d), __tmp);
+ %s(&__data.v%d, sizeof(__data.v%d), __tmp);
  }
- """ % (expr, expr, idx, idx)
+ """ % (expr, expr, probe_read_func, idx, idx)
  if field_type in Probe.fmt_types:
  return text + " __data.v%d = (%s)%s;\n" % \
  (idx, Probe.c_type[field_type], expr)

tools/trace_example.txt

@@ -54,7 +54,7 @@ Event message filter is useful while you only interesting the specific event.
 Like the program open thousands file and you only want to see the "temp" file
 and print stack.
 
-# trace 'do_sys_open "%s", arg2' -UK -f temp
+# trace 'do_sys_open "%s", arg2@user' -UK -f temp
 PID TID COMM FUNC -
 9557 9557 a.out do_sys_open temp.1
  do_sys_open+0x1 [kernel]
@@ -71,7 +71,7 @@ PID TID COMM FUNC -
 
 Process name filter is porting from tools/opensnoop
 
-# trace 'do_sys_open "%s", arg2' -UK -n out
+# trace 'do_sys_open "%s", arg2@user' -UK -n out
 PID TID COMM FUNC -
 9557 9557 a.out do_sys_open temp.1
  do_sys_open+0x1 [kernel]
@@ -241,7 +241,7 @@ so it always includes this header file.
 As a final example, let's trace open syscalls for a specific process. By
 default, tracing is system-wide, but the -p switch overrides this:
 
-# trace -p 2740 'do_sys_open "%s", arg2' -T
+# trace -p 2740 'do_sys_open "%s", arg2@user' -T
 TIME PID COMM FUNC -
 05:36:16 15872 ls do_sys_open /etc/ld.so.cache
 05:36:16 15872 ls do_sys_open /lib64/libselinux.so.1
@@ -335,11 +335,14 @@ EXAMPLES:
 
 trace do_sys_open
  Trace the open syscall and print a default trace message when entered
-trace 'do_sys_open "%s", arg2'
- Trace the open syscall and print the filename being opened
-trace 'do_sys_open "%s", arg2' -n main
+trace 'do_sys_open "%s", arg2@user'
+ Trace the open syscall and print the filename being opened. @user is
+ added to arg2 in kprobes to ensure that char * should be copied from
+ the userspace stack to the bpf stack. If not specified, previous
+ behaviour is expected.
+trace 'do_sys_open "%s", arg2@user' -n main
  Trace the open syscall and only print event that process names containing "main"
-trace 'do_sys_open "%s", arg2' -f config
+trace 'do_sys_open "%s", arg2@user' -f config
  Trace the open syscall and print the filename being opened filtered by "config"
 trace 'sys_read (arg3 > 20000) "read %d bytes", arg3'
  Trace the read syscall and print a message for reads >20000 bytes