reredirect is a utility for taking an existing running program and
attaching its outputs (standard output and error output) to files or
another process.
Using reredirect, you can log output of a already launched process, redirect
debug output of a background process to /dev/null or to a pager as if you
launched it with > or |.
Simple usage is:
reredirect -m FILE PID
It will redirect outputs of PID to FILE. It is also possible to redirect standard output and error output in different files:
reredirect -o FILE1 -e FILE2 PID
-m option is just a shortcut to -o FILE -e FILE.
After launched, reredirect, give you command to restore state of PID. It will looks like:
reredirect -N -O 5 -E 3 5453
-O and -E act as -o and -e but with already opened file descriptors in
PID. They only used to restore previous state of PID.
Without -N, reredirect keep previous outputs opened which allow you to
restore them. This will produce file descriptor leak if you are you call
reredirect multiple times. You can use -N to close and forget previous
outputs.
This package also provide an utility called relink that allow to redirect
outputs to current terminal. When relink exit (with Ctrl+C for exemple)
original state is restored and command is detached.
For exemple:
relink 5453
relink 5453 | grep usefull_line
relink maintains stderr from original command to stderr. So you can do things
like:
relink 5453 > /dev/null
Internally, relink is just a small shell script that create necessary context
and call reredirect as necessary. It use "named pipes". Using "named pipes",
you can redirect output of your target to another command (as a normal pipe):
First create a named pipe:
mkfifo /tmp/myfifo
Run reredirect to redirect your target to /tmp/myfifo:
reredirect -m /tmp/myfifo PID
Launch a command on this named pipe:
less < /tmp/myfifo
tee my_log < /tmp/myfifo
cat -n < /tmp/myfifo
Note that relink only redirect outputs. The target process keep its original
terminal. So if you type Ctrl+Z or CTRL+C, they are not sent to target process.
If you want to do that, you check reptyr command from Nelson Elhage.
Sometime, I work with complex projects and I want to log subparts of compilation output in different files. I use this trick:
target:
@FIFO=$$(mktemp -u); mkfifo $$FIFO; tee my_file.log < $$FIFO & ./redirect -m $$FIFO $$PPID > ./restore_$$PPID.cmd
@echo Call sub makefile here
@sh ./restore_$$PPID.cmd
@echo No more in log file
reredirect is Linux-only. It uses ptrace to attach to the target and control it at the syscall level, so it is highly dependent on Linux's particular syscall API, syscalls, and terminal ioctl()s. A port to Solaris or BSD may be technically feasible, but would probably require significant re-architecting to abstract out the platform-specific bits.
reredirect works on i386, x86_64, and ARM. Ports to other architectures should be straightforward, and should in most cases be as simple as adding an arch/ARCH.h file and adding a clause to the ifdef ladder in ptrace.c.
redirect depends on the ptrace system call to attach to the remote program. On
Ubuntu Maverick and higher, this ability is disabled by default for security
reasons. You can enable it temporarily by doing
# echo 0 > /proc/sys/kernel/yama/ptrace_scope
as root, or permanently by editing the file /etc/sysctl.d/10-ptrace.conf, which also contains more information about exactly what this setting accomplishes.
Reredirect act as a debugger to take control of running process (it use ptrace
syscall). Once it took control of runnign process, it use classical calls to
dup, and dup2 to change targets of file descriptors 1 and 2.
Basicly, to redirect to file, this pseudo code is executed:
orig_fd = 1;
save_fd = dup(1);
new_fd = open(file, O_WRONLY | O_CREAT, 0666);
dup2(new_fd, orig_fd);
close(new_fd);
and to restore state:
ret = dup2(save_fd, orig_fd);
close(save_fd);
reredirect was mainly written by Jérôme Pouiller jezz@sysmic.org. You can contact hom for any questions or bug reports.
reredirect (and especially all ptrace layer) is based on reptyr programm. reptyr was written by Nelson Elhage nelhage@nelhage.com.