IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
strace -k does not produce a reliable output on all supported
configurations yet, even basic strace-k.test is known to fail
on some of them.
* strace.c (usage): Document -k option as experimental.
* strace.1: Likewise.
* NEWS: Likewise.
captured stacktrace is printed in unwind_tcb_fin if tcp->queue is not
empty. This should happen before printing detached message, so
unwind_tcb_fin is moved to the top of droptcb.
This is implicitly suggested by Dmitry Levin in patch review process.
Signed-off-by: Masatake YAMATO <yamato@redhat.com>
free_libunwind_ui is expected to release all unwind related resources
attached to tcp.
* strace.c (droptcb): Move delete_mmap_cache call ...
* unwind.c (free_libunwind_ui): ... to here.
Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Print the stack trace of the traced process after each system call when
-k option is specified. It is implemented using libunwind to unwind the
stack and to obtain the function name pointed by the IP.
Based on the code that was originally taken from strace-plus
of Philip J. Guo.
* configure.ac: Add --with-libunwind option. Check libunwind support.
* Makefile.am: Add libunwind support.
* defs.h (struct tcb) [USE_LIBUNWIND]: Append libunwind specific fields.
[USE_LIBUNWIND] (stack_trace_enabled, alloc_mmap_cache,
delete_mmap_cache, print_stacktrace): New prototypes.
* mem.c (print_mmap, sys_munmap, sys_mprotect): Add libunwind support.
* process.c (sys_execve): Likewise.
* strace.c (usage, alloctcb, droptcb, init): Likewise.
* syscall.c (trace_syscall_exiting): Likewise.
* unwind.c: New file.
* strace.1: Document -k option.
Time spent in system time is not useful where a syscall depends on some
non-CPU resource, eg. typically open() or stat() to a network drive.
This patch adds a new flag (-w) to produce a summary of the time
difference between beginning and end of the system call (ie. latency)
This functionality has been useful to profile slow processes that
are not CPU-bound.
Signed-off-by: Mark Hills <mark.hills@framestore.com>
I found that I use it quite often. Lets make it so that
after cut-n-pasting it into a file, there is no need
to edit the result (e.g. no need to remove C comment
chars from every line.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Add generic tests for fopen64 and fputs_unlocked functions to fix build
with Bionic libc that does not provide them.
* configure.ac (AC_CHECK_FUNCS): Add fopen64 and fputs_unlocked.
* strace.c [_LARGEFILE64_SOURCE]: Use fopen instead of fopen64
if !HAVE_FOPEN64.
Use fputs instead of fputs_unlocked if !HAVE_FPUTS_UNLOCKED.
* vsprintf.c: Use fputs instead of fputs_unlocked
if !HAVE_FPUTS_UNLOCKED.
Reported-by: Elliott Hughes <enh@google.com>
There is only one place left in the code where strace guesses whether
libc provides LFS64 functions and structures. The most natural thing to
do there is to check for _LARGEFILE64_SOURCE - the macro provided by
glibc. Other libc implementations that provide nondegenerate LFS64
interfaces are expected to define this macro as well.
* defs.h (_LFS64_LARGEFILE): Remove.
* strace.c: Use _LARGEFILE64_SOURCE instead of _LFS64_LARGEFILE.
This eliminates some rare bugs, such as post-execve SIGTRAP
generation when we attach to a process, and it manages to finish
execve'ing before we set TRACEEXEC option to suppress that.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
popen_pid = vfork() does work correctly, but for a subtle reason
that wrong assignment of 0 happens in the child _first_,
and _then_ correct value overwrites it in the parent.
(And in a hyphothetical system where vfork = fork,
popen_pid wouldn't be shared, so it will also be ok.)
However, it's not necessary to be difficult.
This change makes it so that assignment is done only in parent.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
The cleanup sequence in error cases, and on normal code path
was nearly the same, no point in duplicating it.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
"strace_child = pid" assignment was racing in NOMMU case
because of vfork (no way to know which process would win).
After this change, strace_child is set to nonzero only
in one process after [v]fork.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
We can use tcb::pid == 0 as an indicator of free tcb,
and we already have strace_child variable which holds
pid of our child, if any.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
I stumbeld over this case when I used "exec stace ...".
* strace.c (trace): Do not exit if we see an unknown pid in wait4.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
* strace.c (detach): Use goto's instead of excessive nesting.
Drop sigstop_expected and interrupt_done variables.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
* strace.c (detach): If TCB_IGNORE_ONE_SIGSTOP is set,
di not PTRACE_DETACH - wait for SIGSTOP first.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
* strace.c (detach): Warn if we see ECHILD from waitpid.
Explain in comments that we don't normally expect !WIFSTOPPED
either, and also that PTRACE_CONT failure isn't expected
(the "break" there is a "I'm confused, bailing out" code style)..
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
* strace.c (detach): Change return type from int to void.
Improve error meesages: show PID, tell exactly which operation fails.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
__WALL is available and working at least since 2.4 kernels: 10 years ago.
* strace (detach): Remove workarounds for ancient kernels lacking __WALL.
Add missing EINTR check.
(trace): Remove workarounds for ancient kernels lacking __WALL.
Remove redundant double error printout.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
V3: split SEIZE/!SEIZE code paths to reduce confusion.
Extensively comment every possible case.
Verified that all tests/detach* tests work in both SEIZE and !SEIZE
cases.
* strace.c (detach): If PTRACE_SEIZE API is in use, stop the tracee
using PTRACE_INTERRUPT instead of sending it a SIGSTOP.
In a subsequent waitpid loop, correctly wait and suppress SIGSTOP
on detach if PTRACE_INTERRUPT wasn't used, or wait for any ptrace
stop and detach without suppressing signals.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Current implementation of -qq does not allow suppressing exit status
messages without suppressing signal delivery messages, which is not
good. There is a traditional "-e signal=none" syntax that can be used
to suppress all signal delivery messages.
This partially reverts commit v4.7-222-g01997cf.
* strace.c (trace): Do not suppress signal delivery messages with -qq.
* strace.1: Update documentation about -qq option.
On NOMMU systems in "strace PROG" case, we have no way to attach to
the tracee before it calls execve unless in -D mode. That is, the
first execve call is very likely to be missed, and setting
hide_log_until_execve just results to empty log.
* strace.c (init) [NOMMU_SYSTEM]: Do not set hide_log_until_execve
unless in -D mode.
* strace.c [HAVE_PRCTL]: Include sys/prctl.h.
(startup_child) [HAVE_PRCTL && PR_SET_PTRACER && PR_SET_PTRACER_ANY]:
In -D mode, allow tracing the process that is going to become the
tracee.
Tested with "./strace [-D] [-q] [-bexecve] env true",
all cases seem to work.
* defs.h: Declare new variable: bool hide_log_until_execve.
* strace.c: Define hide_log_until_execve.
Rename skip_startup_execve to skip_one_b_execve.
(startup_child): Do not set skip_one_b_execve = 1 here.
(init): If "strace PROG" case (as opposed to "strace -pPID"),
set skip_one_b_execve and hide_log_until_execve to 1.
(trace): Don't print signal messages if hide_log_until_execve == 1.
* syscall.c (trace_syscall_entering):
Skip syscall printing if hide_log_until_execve == 1.
Reset hide_log_until_execve if we enter execve syscall.
(trace_syscall_exiting): Skip syscall printing if hide_log_until_execve == 1.
Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Even with the -q flag specified, tracing output is still mixed
with messages about signals and process exit status, which is
often irrelevant. Allow the -q option to be repeated to force
the suppression of signals / exit status info too.
* defs.h: Change 'qflag' from 'bool' to 'unsigned int'.
* strace.1: Document ability to repeat '-q' option.
* strace.c: Allow '-q' to be repeated to quieten process
exit status and signal messages.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
strace already has a mechanism to use fopen64 for output when the 64-bit
version of fopen is available on 32-bit architectures. Apply this
mechanism for other three functions to make strace fully adopted for
64-bit types.
* strace.c (struct_stat, stat_file, struct_dirent, read_dir,
struct_rlimit, set_rlimit): New macros.
(startup_attach): Use read_dir.
(startup_child): Use struct_stat and stat_file.
(main): Use struct_rlimit and set_rlimit.
While at it, added a small optimization of not remembering
the path twice if it happens to be the same.
text data bss dec hex filename
245111 680 10860 256651 3ea8b strace_old
245075 680 9804 255559 3e647 strace
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
To not waste an option letter for just one trick,
extend -b to take a parameter:
"on which syscalls do you want to detach?".
Currently supports only execve.
While at it, fixed (by removing non-Linux and stale info)
and extended manpage text about -f.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
There is no need to print literal strings through "%s".
Only untrusted strings such as filenames need that.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
With -D, strdup'ing of pathname is necessary only on NOMMU.
Don't set skip_startup_execve to 1 if NOMMU and not in daemonized mode
(try "strace [-D] -b env echo HI" to see whether we detach on correct execve).
Fix test_ptrace_FOO shortcuts on NOMMU to always assume success
and _properly_ set all variables.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This is a preparatory mass replace patch with no code changes.
The future change will need to typedef sysent to sysent0,
which results in compile failures when "struct sysent" string
gets mangled into "struct sysent0".
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
By adding tcp->s_ent pointer tot syscall table entry,
we can replace sysent[tcp->scno] references by tcp->s_ent.
More importantly, we may ensure that tcp->s_ent is always valid,
regardless of tcp->scno value. This allows us to drop
SCNO_IS_VALID(tcp->scno) checks before we access syscall
table entry.
We can optimize (qual_flags[tcp->scno] & QUAL_foo) checks
with a similar technique.
Resulting code shrink:
text data bss dec hex filename
245975 700 19072 265747 40e13 strace.t3/strace
245703 700 19072 265475 40d03 strace.t4/strace
* count.c (count_syscall): Use cheaper SCNO_IN_RANGE() check.
* defs.h: Add "int qual_flg" and "const struct sysent *s_ent"
to struct tcb. Remove "int u_nargs" from it.
Add UNDEFINED_SCNO constant which will mark undefined scnos
in tcp->qual_flg.
* pathtrace.c (pathtrace_match): Drop SCNO_IS_VALID check.
Use tcp->s_ent instead of sysent[tcp->scno].
* process.c (sys_prctl): Use tcp->s_ent->nargs instead of tcp->u_nargs.
(sys_waitid): Likewise.
* strace.c (init): Add compile-time check that DEFAULT_QUAL_FLAGS
constant is consistent with init code.
* syscall.c (decode_socket_subcall): Use tcp->s_ent->nargs
instead of tcp->u_nargs. Set tcp->qual_flg and tcp->s_ent.
(decode_ipc_subcall): Likewise.
(printargs): Use tcp->s_ent->nargs instead of tcp->u_nargs.
(printargs_lu): Likewise.
(printargs_ld): Likewise.
(get_scno): [MIPS,ALPHA] Use cheaper SCNO_IN_RANGE() check.
If !SCNO_IS_VALID, set tcp->s_ent and tcp->qual_flg to default values.
(internal_fork): Use tcp->s_ent instead of sysent[tcp->scno].
(syscall_fixup_for_fork_exec): Remove SCNO_IS_VALID check.
Use tcp->s_ent instead of sysent[tcp->scno].
(get_syscall_args): Likewise.
(get_error): Drop SCNO_IS_VALID check where it is redundant.
(dumpio): Drop SCNO_IS_VALID check where it is redundant.
Use tcp->s_ent instead of sysent[tcp->scno].
(trace_syscall_entering): Use (tcp->qual_flg & UNDEFINED_SCNO) instead
of SCNO_IS_VALID check. Use tcp->s_ent instead of sysent[tcp->scno].
Drop SCNO_IS_VALID check where it is redundant.
Print undefined syscall name with undefined_scno_name(tcp).
(trace_syscall_exiting): Likewise.
* util.c (setbpt): Use tcp->s_ent instead of sysent[tcp->scno].
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
pathname[] was getting destroyed, execve of garbage pathname
failing, and to top it off, the tracer's stack was also
smashed and trecer segfaulted.
* strace.c (exec_or_die): New function.
(startup_child): Don't use pathname[] contents after vfork,
make a malloced copy instead. Explain "NOMMU + -D bug"
and how we work around it.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
And while using it, I discovered that -D doesn't work too well
on NOMMU. Added a comment about it.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
