forked from Imagelibrary/binutils-gdb
Currently, on GNU/Linux, it is not possible to interrupt with Ctrl-C
programs that block or ignore SIGINT, with e.g., sigprocmask or
signal(SIGINT, SIG_IGN). You type Ctrl-C, but nothing happens.
Similarly, if a program uses sigwait to wait for SIGINT, and the
program receives a SIGINT, the SIGINT is _not_ intercepted by ptrace,
it goes straight to the inferior. These problems have been known for
years, and recorded in gdb/9425, gdb/14559.
This is a consequence of how GDB implements interrupting programs with
Ctrl-C -- when GDB spawns a new process, it makes the process use the
same terminal as GDB, and then makes the process's process group be
the foreground process group in GDB's terminal. This means that when
the process is running in the foreground, after e.g. "continue", when
the user types Ctrl-C, the kernel sends a SIGINT to the foreground
process group, which is the inferior. GDB then intercepts the SIGINT
via ptrace, the same way it intercepts any other signal, stops all the
other threads of the inferior if in all-stop, and presents the
"Program received SIGINT" stop to the user.
This patch paves the way to address gdb/9425, gdb/14559, by turning
Ctrl-C handling around such that the SIGINT always reaches GDB first,
not the inferior. That is done by making GDB put inferiors in their
own terminal/session created by GDB. I.e., GDB creates a
pseudo-terminal master/slave pair, makes the inferior run with the
slave as its terminal, and pumps output/input on the master end.
Because the inferior is run with its own session/terminal, GDB is free
to remain as the foreground process in its own terminal, which means
that the Ctrl-C SIGINT always reaches GDB first, instead of reaching
the inferior first, and then GDB reacting to the ptrace-intercepted
SIGINT. Because GDB gets the SIGINT first, GDB is then free to
handle it by interrupting the program any way it sees fit. A
following patch will then make GDB interrupt the program with SIGSTOP
instead of SIGINT, which always works even if the inferior
blocks/ignores SIGINT -- SIGSTOP can't be ignored. (In the future GDB
may even switch to PTRACE_INTERRUPT, though that's a project of its
own.)
Having the inferior in its own terminal also means that GDB is in
control of when inferior output is flushed to the screen. When
debugging with the CLI, this means that inferior output is now never
interpersed with GDB's output in an unreadable fashion. This will
also allow fixing the problem of inferior output really messing up the
screen in the TUI, forcing users to Ctrl-L to refresh the screen.
This patch does not address the TUI part, but it shouldn't be too hard
-- I wrote a quick&dirty prototype patch doing that a few years back,
so I know it works.
Implementation wise, here's what is happening:
- when GDB spawns an inferior, unless the user asked otherwise with
"tty /dev/tty", GDB creates a pty pair, and makes the slave end the
inferior's terminal. Note that starting an inferior on a given
terminal already exists, given the "tty" command. GDB records the
master and slave ends of the pty.
- GDB registers that new terminal's master end on the event loop.
When the master is written to, it means the inferior has written
some output on its terminal. The event loop wakes up and GDB
flushes the inferior output to its own terminal / to the screen.
- When target_terminal state is switched to "inferior", with
target_tarminal::inferiors(), GDB registers the stdin file
descriptor on the event loop with a callback that forwards input
typed on GDB's terminal to the inferior's tty.
- Similarly, when GDB receives a SIGWINCH signal, meaning GDB's
terminal was resized, GDB resizes the inferior's terminal too.
- GDB puts the inferior in its own session, and there's a "session
leader" process between GDB and the inferior. The latter is
because session leaders have special properties, one of which is,
if they exit, all progresses in the foreground process group in the
session get a SIGHUP. If the spawned inferior was the session
leader itself, if you were debugging an inferior that forks and
follow to the child, if the parent (the session leader) exits, then
the child would get a SIGHUP. Forking twice when launching an
inferior, and making the first child be the session leader, and the
second child the inferior avoids that problem.
- When the inferior exits or is killed, GDB sends a SIGHUP to the
session leader, waits for the leader to exit and then destroys the
terminal. The session leader's SIGHUP handler makes the session
leader pgrp be the foreground process group and then exits. This
sequence is important comparing to just closing the terminal and
letting the session leader terminate due to the SIGHUP the kernel
sends, because when the session leader exits, all processes in the
foreground process group get a SIGHUP, meaning that if the detached
process was still in the foreground, it would get a SIGHUP, and
likely die.
- The gdb.multi/multi-term-settings.exp was adjusted to test for
shared and not-shared terminal/session. Without the change, we get
failures:
FAIL: gdb.multi/multi-term-settings.exp: inf1_how=run: inf2_how=run: continue (expected SIGTTOU)
FAIL: gdb.multi/multi-term-settings.exp: inf1_how=run: inf2_how=run: stop with control-c (Quit)
Tested on GNU/Linux native, gdbserver and gdbserver + "maint target
set-non-stop on". Also build-tested tested on mingw32-w64, Solaris
11, and OpenBSD.
gdb/ChangeLog:
yyyy-mm-dd Pedro Alves <pedro@palves.net>
PR gdb/9425
PR gdb/14559
* fork-child.c (child_has_managed_tty_hook): New.
* inf-ptrace.c (inf_ptrace_me): If we created a managed tty, raise
SIGSTOP.
(inf_ptrace_handle_session_leader_fork): New.
(inf_ptrace_target::create_inferior): Pass it down as
handle_session_leader_fork callback.
* inf-ptrace.h (inf_ptrace_target) <handle_session_leader_fork>:
New virtual method.
* inferior.h (child_terminal_on_sigwinch): Declare.
* inflow.c: Include "gdbsupport/event-loop.h",
"gdbsupport/refcounted-object.h", "gdbsupport/gdb_wait.h",
"gdbsupport/managed-tty.h".
(USES_FORK_CHILD): Define, and wrap fork-child.c-related code with
it.
(struct run_terminal_info): New.
(struct terminal_info) <run_terminal>: Now a run_terminal_info.
<process_group>: Default to -1.
<save_from_tty>: New method.
(sigint_ours): Update comments.
(inferior_thisrun_terminal_pty_fd): New.
(input_fd_redirected): New.
(sharing_input_terminal): Adjust.
(gdb_tcgetattr, gdb_tcsetattr, make_raw, class scoped_raw_termios)
(child_terminal_flush_from_to, child_terminal_flush_stdout)
(inferior_stdout_event_handler, inferior_stdin_event_handler): New.
(child_terminal_inferior): Handle inferiors with gdb-managed ttys.
(child_terminal_save_inferior): Handle inferiors with gdb-managed
ttys. Use save_from_tty.
(child_terminal_ours_1): Handle inferiors with gdb-managed ttys.
(terminal_info::~terminal_info): Use delete instead of xfree.
(child_terminal_on_sigwinc): New.
(inflow_inferior_exit): Release terminal created by GDB.
(copy_terminal_info): Assert there's no run_terminal yet in TO
yet. Incref run_terminal after copying.
(child_terminal_info): Handle inferiors with gdb-managed ttys.
(new_tty_prefork): Allocate pseudo-terminal.
(created_managed_tty): New.
(new_tty): Remove __GO32__ and _WIN32 #ifdefs, not needed given
USES_FORK_CHILD.
(new_tty_postfork): Handle inferiors with gdb-managed ttys.
(show_debug_managed_tty): New.
(_initialize_inflow): Register "set/show debug managed-tty".
* linux-nat.c (waitpid_sigstop, waitpid_fork)
(linux_nat_target::handle_session_leader_fork): New.
* linux-nat.h (linux_nat_target) <handle_session_leader_fork>:
Declare override.
* nat/fork-inferior.c: Include
"gdbsupport/scoped_ignore_sigttou.h", "gdbsupport/managed-tty.h",
<sys/types.h> and <sys/wait.h>.
(session_leader_hup): New.
(fork_inferior): Add handle_session_leader_fork parameter. If the
inferior has a gdb-managed tty, don't use vfork, and fork twice,
with the first fork becoming the session leader. Call
handle_session_leader_fork.
* nat/fork-inferior.h (fork_inferior): Add
handle_session_leader_fork parameter and update comment.
(child_has_managed_tty_hook): Declare.
* terminal.h (created_managed_tty, child_gdb_owns_session):
Declare.
* tui/tui-win.c: Include "inferior.h".
(tui_async_resize_screen): Call child_terminal_on_sigwinch.
gdbsupport/ChangeLog:
yyyy-mm-dd Pedro Alves <pedro@palves.net>
PR gdb/9425
PR gdb/14559
* Makefile.am (libgdbsupport_a_SOURCES): Add managed-tty.cc.
* Makefile.in: Regenerate.
* managed-tty.cc: New.
* managed-tty.h: New.
gdbserver/ChangeLog:
yyyy-mm-dd Pedro Alves <pedro@palves.net>
PR gdb/9425
PR gdb/14559
* fork-child.cc (child_has_managed_tty_hook): New.
gdb/testsuite/ChangeLog:
yyyy-mm-dd Pedro Alves <pedro@palves.net>
PR gdb/9425
PR gdb/14559
* gdb.multi/multi-term-settings.exp (create_inferior): Document
"run-session", "run-share" and "run-tty" instead of "run" and
"tty". Adjust to handle "run-session" vs "run-share".
(coretest): Adjust to handle "run-session" vs "run-share".
(how_modes): Use "run-session", "run-share" and "run-tty" instead
of "run" and "tty".
Change-Id: I2569e189294044891e68a66401b381e4b999b19c
340 lines
11 KiB
C++
340 lines
11 KiB
C++
/* Native debugging support for GNU/Linux (LWP layer).
|
|
|
|
Copyright (C) 2000-2021 Free Software Foundation, Inc.
|
|
|
|
This file is part of GDB.
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>. */
|
|
|
|
#ifndef LINUX_NAT_H
|
|
#define LINUX_NAT_H
|
|
|
|
#include "nat/linux-nat.h"
|
|
#include "inf-ptrace.h"
|
|
#include "target.h"
|
|
#include <signal.h>
|
|
|
|
/* A prototype generic GNU/Linux target. A concrete instance should
|
|
override it with local methods. */
|
|
|
|
class linux_nat_target : public inf_ptrace_target
|
|
{
|
|
public:
|
|
linux_nat_target ();
|
|
~linux_nat_target () override = 0;
|
|
|
|
thread_control_capabilities get_thread_control_capabilities () override
|
|
{ return tc_schedlock; }
|
|
|
|
void create_inferior (const char *, const std::string &,
|
|
char **, int) override;
|
|
|
|
pid_t handle_session_leader_fork (pid_t sl_pid) override;
|
|
|
|
void attach (const char *, int) override;
|
|
|
|
void detach (inferior *, int) override;
|
|
|
|
void resume (ptid_t, int, enum gdb_signal) override;
|
|
|
|
ptid_t wait (ptid_t, struct target_waitstatus *, target_wait_flags) override;
|
|
|
|
void pass_signals (gdb::array_view<const unsigned char>) override;
|
|
|
|
enum target_xfer_status xfer_partial (enum target_object object,
|
|
const char *annex,
|
|
gdb_byte *readbuf,
|
|
const gdb_byte *writebuf,
|
|
ULONGEST offset, ULONGEST len,
|
|
ULONGEST *xfered_len) override;
|
|
|
|
void kill () override;
|
|
|
|
void mourn_inferior () override;
|
|
bool thread_alive (ptid_t ptid) override;
|
|
|
|
void update_thread_list () override;
|
|
|
|
std::string pid_to_str (ptid_t) override;
|
|
|
|
const char *thread_name (struct thread_info *) override;
|
|
|
|
struct address_space *thread_address_space (ptid_t) override;
|
|
|
|
bool stopped_by_watchpoint () override;
|
|
|
|
bool stopped_data_address (CORE_ADDR *) override;
|
|
|
|
bool stopped_by_sw_breakpoint () override;
|
|
bool supports_stopped_by_sw_breakpoint () override;
|
|
|
|
bool stopped_by_hw_breakpoint () override;
|
|
bool supports_stopped_by_hw_breakpoint () override;
|
|
|
|
void thread_events (int) override;
|
|
|
|
bool can_async_p () override;
|
|
bool is_async_p () override;
|
|
|
|
bool supports_non_stop () override;
|
|
bool always_non_stop_p () override;
|
|
|
|
int async_wait_fd () override;
|
|
void async (int) override;
|
|
|
|
void close () override;
|
|
|
|
void stop (ptid_t) override;
|
|
|
|
bool supports_multi_process () override;
|
|
|
|
bool supports_disable_randomization () override;
|
|
|
|
int core_of_thread (ptid_t ptid) override;
|
|
|
|
bool filesystem_is_local () override;
|
|
|
|
int fileio_open (struct inferior *inf, const char *filename,
|
|
int flags, int mode, int warn_if_slow,
|
|
int *target_errno) override;
|
|
|
|
gdb::optional<std::string>
|
|
fileio_readlink (struct inferior *inf,
|
|
const char *filename,
|
|
int *target_errno) override;
|
|
|
|
int fileio_unlink (struct inferior *inf,
|
|
const char *filename,
|
|
int *target_errno) override;
|
|
|
|
int insert_fork_catchpoint (int) override;
|
|
int remove_fork_catchpoint (int) override;
|
|
int insert_vfork_catchpoint (int) override;
|
|
int remove_vfork_catchpoint (int) override;
|
|
|
|
int insert_exec_catchpoint (int) override;
|
|
int remove_exec_catchpoint (int) override;
|
|
|
|
int set_syscall_catchpoint (int pid, bool needed, int any_count,
|
|
gdb::array_view<const int> syscall_counts) override;
|
|
|
|
char *pid_to_exec_file (int pid) override;
|
|
|
|
void post_startup_inferior (ptid_t) override;
|
|
|
|
void post_attach (int) override;
|
|
|
|
void follow_fork (bool, bool) override;
|
|
|
|
std::vector<static_tracepoint_marker>
|
|
static_tracepoint_markers_by_strid (const char *id) override;
|
|
|
|
/* Methods that are meant to overridden by the concrete
|
|
arch-specific target instance. */
|
|
|
|
virtual void low_resume (ptid_t ptid, int step, enum gdb_signal sig)
|
|
{ inf_ptrace_target::resume (ptid, step, sig); }
|
|
|
|
virtual bool low_stopped_by_watchpoint ()
|
|
{ return false; }
|
|
|
|
virtual bool low_stopped_data_address (CORE_ADDR *addr_p)
|
|
{ return false; }
|
|
|
|
/* The method to call, if any, when a new thread is attached. */
|
|
virtual void low_new_thread (struct lwp_info *)
|
|
{}
|
|
|
|
/* The method to call, if any, when a thread is destroyed. */
|
|
virtual void low_delete_thread (struct arch_lwp_info *lp)
|
|
{
|
|
gdb_assert (lp == NULL);
|
|
}
|
|
|
|
/* The method to call, if any, when a new fork is attached. */
|
|
virtual void low_new_fork (struct lwp_info *parent, pid_t child_pid)
|
|
{}
|
|
|
|
/* The method to call, if any, when a new clone event is detected. */
|
|
virtual void low_new_clone (struct lwp_info *parent, pid_t child_lwp)
|
|
{}
|
|
|
|
/* The method to call, if any, when a process is no longer
|
|
attached. */
|
|
virtual void low_forget_process (pid_t pid)
|
|
{}
|
|
|
|
/* Hook to call prior to resuming a thread. */
|
|
virtual void low_prepare_to_resume (struct lwp_info *)
|
|
{}
|
|
|
|
/* Convert a ptrace/host siginfo object, into/from the siginfo in
|
|
the layout of the inferiors' architecture. Returns true if any
|
|
conversion was done; false otherwise, in which case the caller
|
|
does a straight memcpy. If DIRECTION is 1, then copy from INF to
|
|
PTRACE. If DIRECTION is 0, copy from PTRACE to INF. */
|
|
virtual bool low_siginfo_fixup (siginfo_t *ptrace, gdb_byte *inf,
|
|
int direction)
|
|
{ return false; }
|
|
|
|
/* SIGTRAP-like breakpoint status events recognizer. The default
|
|
recognizes SIGTRAP only. */
|
|
virtual bool low_status_is_event (int status);
|
|
};
|
|
|
|
/* The final/concrete instance. */
|
|
extern linux_nat_target *linux_target;
|
|
|
|
struct arch_lwp_info;
|
|
|
|
/* Structure describing an LWP. This is public only for the purposes
|
|
of ALL_LWPS; target-specific code should generally not access it
|
|
directly. */
|
|
|
|
struct lwp_info
|
|
{
|
|
/* The process id of the LWP. This is a combination of the LWP id
|
|
and overall process id. */
|
|
ptid_t ptid;
|
|
|
|
/* If this flag is set, we need to set the event request flags the
|
|
next time we see this LWP stop. */
|
|
int must_set_ptrace_flags;
|
|
|
|
/* Non-zero if we sent this LWP a SIGSTOP (but the LWP didn't report
|
|
it back yet). */
|
|
int signalled;
|
|
|
|
/* Non-zero if this LWP is stopped. */
|
|
int stopped;
|
|
|
|
/* Non-zero if this LWP will be/has been resumed. Note that an LWP
|
|
can be marked both as stopped and resumed at the same time. This
|
|
happens if we try to resume an LWP that has a wait status
|
|
pending. We shouldn't let the LWP run until that wait status has
|
|
been processed, but we should not report that wait status if GDB
|
|
didn't try to let the LWP run. */
|
|
int resumed;
|
|
|
|
/* The last resume GDB requested on this thread. */
|
|
enum resume_kind last_resume_kind;
|
|
|
|
/* If non-zero, a pending wait status. */
|
|
int status;
|
|
|
|
/* When 'stopped' is set, this is where the lwp last stopped, with
|
|
decr_pc_after_break already accounted for. If the LWP is
|
|
running and stepping, this is the address at which the lwp was
|
|
resumed (that is, it's the previous stop PC). If the LWP is
|
|
running and not stepping, this is 0. */
|
|
CORE_ADDR stop_pc;
|
|
|
|
/* Non-zero if we were stepping this LWP. */
|
|
int step;
|
|
|
|
/* The reason the LWP last stopped, if we need to track it
|
|
(breakpoint, watchpoint, etc.). */
|
|
enum target_stop_reason stop_reason;
|
|
|
|
/* On architectures where it is possible to know the data address of
|
|
a triggered watchpoint, STOPPED_DATA_ADDRESS_P is non-zero, and
|
|
STOPPED_DATA_ADDRESS contains such data address. Otherwise,
|
|
STOPPED_DATA_ADDRESS_P is false, and STOPPED_DATA_ADDRESS is
|
|
undefined. Only valid if STOPPED_BY_WATCHPOINT is true. */
|
|
int stopped_data_address_p;
|
|
CORE_ADDR stopped_data_address;
|
|
|
|
/* Non-zero if we expect a duplicated SIGINT. */
|
|
int ignore_sigint;
|
|
|
|
/* If WAITSTATUS->KIND != TARGET_WAITKIND_SPURIOUS, the waitstatus
|
|
for this LWP's last event. This may correspond to STATUS above,
|
|
or to a local variable in lin_lwp_wait. */
|
|
struct target_waitstatus waitstatus;
|
|
|
|
/* Signal whether we are in a SYSCALL_ENTRY or
|
|
in a SYSCALL_RETURN event.
|
|
Values:
|
|
- TARGET_WAITKIND_SYSCALL_ENTRY
|
|
- TARGET_WAITKIND_SYSCALL_RETURN */
|
|
enum target_waitkind syscall_state;
|
|
|
|
/* The processor core this LWP was last seen on. */
|
|
int core;
|
|
|
|
/* Arch-specific additions. */
|
|
struct arch_lwp_info *arch_private;
|
|
|
|
/* Previous and next pointers in doubly-linked list of known LWPs,
|
|
sorted by reverse creation order. */
|
|
struct lwp_info *prev;
|
|
struct lwp_info *next;
|
|
};
|
|
|
|
/* The global list of LWPs, for ALL_LWPS. Unlike the threads list,
|
|
there is always at least one LWP on the list while the GNU/Linux
|
|
native target is active. */
|
|
extern struct lwp_info *lwp_list;
|
|
|
|
/* Does the current host support PTRACE_GETREGSET? */
|
|
extern enum tribool have_ptrace_getregset;
|
|
|
|
/* Iterate over each active thread (light-weight process). */
|
|
#define ALL_LWPS(LP) \
|
|
for ((LP) = lwp_list; \
|
|
(LP) != NULL; \
|
|
(LP) = (LP)->next)
|
|
|
|
/* Attempt to initialize libthread_db. */
|
|
void check_for_thread_db (void);
|
|
|
|
/* Called from the LWP layer to inform the thread_db layer that PARENT
|
|
spawned CHILD. Both LWPs are currently stopped. This function
|
|
does whatever is required to have the child LWP under the
|
|
thread_db's control --- e.g., enabling event reporting. Returns
|
|
true on success, false if the process isn't using libpthread. */
|
|
extern int thread_db_notice_clone (ptid_t parent, ptid_t child);
|
|
|
|
/* Return the number of signals used by the threads library. */
|
|
extern unsigned int lin_thread_get_thread_signal_num (void);
|
|
|
|
/* Return the i-th signal used by the threads library. */
|
|
extern int lin_thread_get_thread_signal (unsigned int i);
|
|
|
|
/* Find process PID's pending signal set from /proc/pid/status. */
|
|
void linux_proc_pending_signals (int pid, sigset_t *pending,
|
|
sigset_t *blocked, sigset_t *ignored);
|
|
|
|
/* For linux_stop_lwp see nat/linux-nat.h. */
|
|
|
|
/* Stop all LWPs, synchronously. (Any events that trigger while LWPs
|
|
are being stopped are left pending.) */
|
|
extern void linux_stop_and_wait_all_lwps (void);
|
|
|
|
/* Set resumed LWPs running again, as they were before being stopped
|
|
with linux_stop_and_wait_all_lwps. (LWPS with pending events are
|
|
left stopped.) */
|
|
extern void linux_unstop_all_lwps (void);
|
|
|
|
/* Update linux-nat internal state when changing from one fork
|
|
to another. */
|
|
void linux_nat_switch_fork (ptid_t new_ptid);
|
|
|
|
/* Store the saved siginfo associated with PTID in *SIGINFO.
|
|
Return 1 if it was retrieved successfully, 0 otherwise (*SIGINFO is
|
|
uninitialized in such case). */
|
|
int linux_nat_get_siginfo (ptid_t ptid, siginfo_t *siginfo);
|
|
|
|
#endif /* LINUX_NAT_H */
|