2db9a4275c
TL;DR - GDB can hang if something refreshes the thread list out of the target while the target is running. GDB hangs inside td_ta_thr_iter. The fix is to not use that libthread_db function anymore. Long version: Running the testsuite against my all-stop-on-top-of-non-stop series is still exposing latent non-stop bugs. I was originally seeing this with the multi-create.exp test, back when we were still using libthread_db thread event breakpoints. The all-stop-on-top-of-non-stop series forces a thread list refresh each time GDB needs to start stepping over a breakpoint (to pause all threads). That test hits the thread event breakpoint often, resulting in a bunch of step-over operations, thus a bunch of thread list refreshes while some threads in the target are running. The commit adds a real non-stop mode test that triggers the issue, based on multi-create.exp, that does an explicit "info threads" when a breakpoint is hit. IOW, it does the same things the as-ns series was doing when testing multi-create.exp. The bug is a race, so it unfortunately takes several runs for the test to trigger it. In fact, even when setting the test running in a loop, it sometimes takes several minutes for it to trigger for me. The race is related to libthread_db's td_ta_thr_iter. This is libthread_db's entry point for walking the thread list of the inferior. Sometimes, when GDB refreshes the thread list from the target, libthread_db's td_ta_thr_iter can somehow see glibc's thread list as a cycle, and get stuck in an infinite loop. The issue is that when a thread exits, its thread control structure in glibc is moved from a "used" list to a "cache" list. These lists are simply circular linked lists where the "next/prev" pointers are embedded in the thread control structure itself. The "next" pointer of the last element of the list points back to the list's sentinel "head". There's only one set of "next/prev" pointers for both lists; thus a thread can only be in one of the lists at a time, not in both simultaneously. So when thread C exits, simplifying, the following happens. A-C are threads. stack_used and stack_cache are the list's heads. Before: stack_used -> A -> B -> C -> (&stack_used) stack_cache -> (&stack_cache) After: stack_used -> A -> B -> (&stack_used) stack_cache -> C -> (&stack_cache) td_ta_thr_iter starts by iterating at the list's head's next, and iterates until it sees a thread whose next pointer points to the list's head again. Thus in the before case above, C's next points to stack_used, indicating end of list. In the same case, the stack_cache list is empty. For each thread being iterated, td_ta_thr_iter reads the whole thread object out of the inferior. This includes the thread's "next" pointer. In the scenario above, it may happen that td_ta_thr_iter is iterating thread B and has already read B's thread structure just before thread C exits and its control structure moves to the cached list. Now, recall that td_ta_thr_iter is running in the context of GDB, and there's no locking between GDB and the inferior. From it's local copy of B, td_ta_thr_iter believes that the next thread after B is thread C, so it happilly continues iterating to C, a thread that has already exited, and is now in the stack cache list. After iterating C, td_ta_thr_iter finds the stack_cache head, which because it is not stack_used, td_ta_thr_iter assumes it's just another thread. After this, unless the reverse race triggers, GDB gets stuck in td_ta_thr_iter forever walking the stack_cache list, as no thread in thatlist has a next pointer that points back to stack_used (the terminating condition). Before fully understanding the issue, I tried adding cycle detection to GDB's td_ta_thr_iter callback. However, td_ta_thr_iter skips calling the callback in some cases, which means that it's possible that the callback isn't called at all, making it impossible for GDB to break the loop. I did manage to get GDB stuck in that state more than once. Fortunately, we can avoid the issue altogether. We don't really need td_ta_thr_iter for live debugging nowadays, given PTRACE_EVENT_CLONE. We already know how to map and lwp id to a thread id without iterating (thread_from_lwp), so use that more. gdb/ChangeLog: 2015-02-20 Pedro Alves <palves@redhat.com> * linux-nat.c (linux_handle_extended_wait): Call thread_db_notice_clone whenever a new clone LWP is detected. (linux_stop_and_wait_all_lwps, linux_unstop_all_lwps): New functions. * linux-nat.h (thread_db_attach_lwp): Delete declaration. (thread_db_notice_clone, linux_stop_and_wait_all_lwps) (linux_unstop_all_lwps): Declare. * linux-thread-db.c (struct thread_get_info_inout): Delete. (thread_get_info_callback): Delete. (thread_from_lwp): Use td_thr_get_info and record_thread. (thread_db_attach_lwp): Delete. (thread_db_notice_clone): New function. (try_thread_db_load_1): If /proc is mounted and shows the process'es task list, walk over all LWPs and call thread_from_lwp instead of relying on td_ta_thr_iter. (attach_thread): Don't call check_thread_signals here. Split the tail part of the function (which adds the thread to the core GDB thread list) to ... (record_thread): ... this function. Call check_thread_signals here. (thread_db_wait): Don't call thread_db_find_new_threads_1. Always call thread_from_lwp. (thread_db_update_thread_list): Rename to ... (thread_db_update_thread_list_org): ... this. (thread_db_update_thread_list): New function. (thread_db_find_thread_from_tid): Delete. (thread_db_get_ada_task_ptid): Simplify. * nat/linux-procfs.c: Include <sys/stat.h>. (linux_proc_task_list_dir_exists): New function. * nat/linux-procfs.h (linux_proc_task_list_dir_exists): Declare. gdb/gdbserver/ChangeLog: 2015-02-20 Pedro Alves <palves@redhat.com> * thread-db.c: Include "nat/linux-procfs.h". (thread_db_init): Skip listing new threads if the kernel supports PTRACE_EVENT_CLONE and /proc/PID/task/ is accessible. gdb/testsuite/ChangeLog: 2015-02-20 Pedro Alves <palves@redhat.com> * gdb.threads/multi-create-ns-info-thr.exp: New file. |
||
---|---|---|
.. | ||
.gitignore | ||
acinclude.m4 | ||
aclocal.m4 | ||
ax.c | ||
ax.h | ||
ChangeLog | ||
config.in | ||
configure | ||
configure.ac | ||
configure.srv | ||
debug.c | ||
debug.h | ||
dll.c | ||
dll.h | ||
event-loop.c | ||
event-loop.h | ||
gdb_proc_service.h | ||
gdbreplay.c | ||
gdbthread.h | ||
hostio-errno.c | ||
hostio.c | ||
hostio.h | ||
i387-fp.c | ||
i387-fp.h | ||
inferiors.c | ||
inferiors.h | ||
linux-aarch64-low.c | ||
linux-amd64-ipa.c | ||
linux-arm-low.c | ||
linux-bfin-low.c | ||
linux-cris-low.c | ||
linux-crisv32-low.c | ||
linux-i386-ipa.c | ||
linux-ia64-low.c | ||
linux-low.c | ||
linux-low.h | ||
linux-m32r-low.c | ||
linux-m68k-low.c | ||
linux-mips-low.c | ||
linux-nios2-low.c | ||
linux-ppc-low.c | ||
linux-s390-low.c | ||
linux-sh-low.c | ||
linux-sparc-low.c | ||
linux-tic6x-low.c | ||
linux-tile-low.c | ||
linux-x86-low.c | ||
linux-xtensa-low.c | ||
lynx-i386-low.c | ||
lynx-low.c | ||
lynx-low.h | ||
lynx-ppc-low.c | ||
Makefile.in | ||
mem-break.c | ||
mem-break.h | ||
notif.c | ||
notif.h | ||
nto-low.c | ||
nto-low.h | ||
nto-x86-low.c | ||
proc-service.c | ||
proc-service.list | ||
README | ||
regcache.c | ||
regcache.h | ||
remote-utils.c | ||
remote-utils.h | ||
server.c | ||
server.h | ||
spu-low.c | ||
symbol.c | ||
target.c | ||
target.h | ||
tdesc.c | ||
tdesc.h | ||
terminal.h | ||
thread-db.c | ||
tracepoint.c | ||
tracepoint.h | ||
utils.c | ||
utils.h | ||
win32-arm-low.c | ||
win32-i386-low.c | ||
win32-low.c | ||
win32-low.h | ||
wincecompat.c | ||
wincecompat.h | ||
x86-low.c | ||
x86-low.h | ||
xtensa-xtregs.c |
README for GDBserver & GDBreplay by Stu Grossman and Fred Fish Introduction: This is GDBserver, a remote server for Un*x-like systems. It can be used to control the execution of a program on a target system from a GDB on a different host. GDB and GDBserver communicate using the standard remote serial protocol implemented in remote.c, and various *-stub.c files. They communicate via either a serial line or a TCP connection. For more information about GDBserver, see the GDB manual. Usage (server (target) side): First, you need to have a copy of the program you want to debug put onto the target system. The program can be stripped to save space if needed, as GDBserver doesn't care about symbols. All symbol handling is taken care of by the GDB running on the host system. To use the server, you log on to the target system, and run the `gdbserver' program. You must tell it (a) how to communicate with GDB, (b) the name of your program, and (c) its arguments. The general syntax is: target> gdbserver COMM PROGRAM [ARGS ...] For example, using a serial port, you might say: target> gdbserver /dev/com1 emacs foo.txt This tells GDBserver to debug emacs with an argument of foo.txt, and to communicate with GDB via /dev/com1. GDBserver now waits patiently for the host GDB to communicate with it. To use a TCP connection, you could say: target> gdbserver host:2345 emacs foo.txt This says pretty much the same thing as the last example, except that we are going to communicate with the host GDB via TCP. The `host:2345' argument means that we are expecting to see a TCP connection from `host' to local TCP port 2345. (Currently, the `host' part is ignored.) You can choose any number you want for the port number as long as it does not conflict with any existing TCP ports on the target system. This same port number must be used in the host GDBs `target remote' command, which will be described shortly. Note that if you chose a port number that conflicts with another service, GDBserver will print an error message and exit. On some targets, GDBserver can also attach to running programs. This is accomplished via the --attach argument. The syntax is: target> gdbserver --attach COMM PID PID is the process ID of a currently running process. It isn't necessary to point GDBserver at a binary for the running process. Usage (host side): You need an unstripped copy of the target program on your host system, since GDB needs to examine it's symbol tables and such. Start up GDB as you normally would, with the target program as the first argument. (You may need to use the --baud option if the serial line is running at anything except 9600 baud.) Ie: `gdb TARGET-PROG', or `gdb --baud BAUD TARGET-PROG'. After that, the only new command you need to know about is `target remote'. It's argument is either a device name (usually a serial device, like `/dev/ttyb'), or a HOST:PORT descriptor. For example: (gdb) target remote /dev/ttyb communicates with the server via serial line /dev/ttyb, and: (gdb) target remote the-target:2345 communicates via a TCP connection to port 2345 on host `the-target', where you previously started up GDBserver with the same port number. Note that for TCP connections, you must start up GDBserver prior to using the `target remote' command, otherwise you may get an error that looks something like `Connection refused'. Building GDBserver: The supported targets as of November 2006 are: arm-*-linux* bfin-*-uclinux bfin-*-linux-uclibc crisv32-*-linux* cris-*-linux* i[34567]86-*-cygwin* i[34567]86-*-linux* i[34567]86-*-mingw* ia64-*-linux* m32r*-*-linux* m68*-*-linux* m68*-*-uclinux* mips*64*-*-linux* mips*-*-linux* powerpc[64]-*-linux* s390[x]-*-linux* sh-*-linux* spu*-*-* x86_64-*-linux* Configuring GDBserver you should specify the same machine for host and target (which are the machine that GDBserver is going to run on. This is not the same as the machine that GDB is going to run on; building GDBserver automatically as part of building a whole tree of tools does not currently work if cross-compilation is involved (we don't get the right CC in the Makefile, to start with)). Building GDBserver for your target is very straightforward. If you build GDB natively on a target which GDBserver supports, it will be built automatically when you build GDB. You can also build just GDBserver: % mkdir obj % cd obj % path-to-gdbserver-sources/configure % make If you prefer to cross-compile to your target, then you can also build GDBserver that way. In a Bourne shell, for example: % export CC=your-cross-compiler % path-to-gdbserver-sources/configure your-target-name % make Using GDBreplay: A special hacked down version of GDBserver can be used to replay remote debug log files created by GDB. Before using the GDB "target" command to initiate a remote debug session, use "set remotelogfile <filename>" to tell GDB that you want to make a recording of the serial or tcp session. Note that when replaying the session, GDB communicates with GDBreplay via tcp, regardless of whether the original session was via a serial link or tcp. Once you are done with the remote debug session, start GDBreplay and tell it the name of the log file and the host and port number that GDB should connect to (typically the same as the host running GDB): $ gdbreplay logfile host:port Then start GDB (preferably in a different screen or window) and use the "target" command to connect to GDBreplay: (gdb) target remote host:port Repeat the same sequence of user commands to GDB that you gave in the original debug session. GDB should not be able to tell that it is talking to GDBreplay rather than a real target, all other things being equal. Note that GDBreplay echos the command lines to stderr, as well as the contents of the packets it sends and receives. The last command echoed by GDBreplay is the next command that needs to be typed to GDB to continue the session in sync with the original session.