[PATCH 0/8] signal: Cleanup of the signal->flags

[PATCH 0/8] signal: Cleanup of the signal->flags

[Eric W. Biederman]

The special case of SIGKILL during coredumps is very fragile today and
while reading through the code I realized I have almost broken it twice.
So this simplifies that special case, removes SIGNAL_GROUP_COREDUMP
which has become unnecessary with the addition of signal->core_state,
and this removes the helper signal_group_exit which is misnamed and
is not used properly.

If you squint very hard there might be a user space visible difference
in behavior somewhere but I don't think there is one in practice.

These patches are on top of:
https://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git/ signal-for-v5.17

After these patches have been reviewed it is my plan to apply them to my
signal-for-v5.17 branch.

Eric W. Biederman (8):
      signal: Make SIGKILL during coredumps an explicit special case
      signal: Drop signals received after a fatal signal has been processed
      signal: Have the oom killer detect coredumps using signal->core_state
      signal: During coredumps set SIGNAL_GROUP_EXIT in zap_process
      signal: Remove SIGNAL_GROUP_COREDUMP
      coredump: Stop setting signal->group_exit_task
      signal: Rename group_exit_task group_exec_task
      signal: Remove the helper signal_group_exit

 fs/coredump.c                | 20 +++++++++-----------
 fs/exec.c                    | 10 +++++-----
 include/linux/sched/signal.h | 18 +++---------------
 kernel/exit.c                | 12 ++++++++----
 kernel/signal.c              | 24 ++++++++++++++++--------
 mm/oom_kill.c                |  2 +-
 6 files changed, 42 insertions(+), 44 deletions(-)

Eric

[PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Simplify the code that allows SIGKILL during coredumps to terminate
the coredump.  As far as I can tell I have avoided breaking it
by dumb luck.

Historically with all of the other threads stopping in exit_mm the
wants_signal loop in complete_signal would find the dumper task and
then complete_signal would wake the dumper task with signal_wake_up.

After moving the coredump_task_exit above the setting of PF_EXITING in
commit 92307383082d ("coredump: Don't perform any cleanups before
dumping core") wants_signal will consider all of the threads in a
multi-threaded process for waking up, not just the core dumping task.

Luckily complete_signal short circuits SIGKILL during a coredump marks
every thread with SIGKILL and signal_wake_up.  This code is arguably
buggy however as it tries to skip creating a group exit when is already
present, and it fails that a coredump is in progress.

Ever since commit 06af8679449d ("coredump: Limit what can interrupt
coredumps") was added dump_interrupted needs not just TIF_SIGPENDING
set on the dumper task but also SIGKILL set in it's pending bitmap.
This means that if the code is ever fixed not to short-circuit and
kill a process after it has already been killed the special case
for SIGKILL during a coredump will be broken.

Sort all of this out by making the coredump special case more special,
and perform all of the work in prepare_signal and leave the rest of
the signal delivery path out of it.

In prepare_signal when the process coredumping is sent SIGKILL find
the task performing the coredump and use sigaddset and signal_wake_up
to ensure that task reports fatal_signal_pending.

Return false from prepare_signal to tell the rest of the signal
delivery path to ignore the signal.

Update wait_for_dump_helpers to perform a wait_event_killable wait
so that if signal_pending gets set spuriously the wait will not
be interrupted unless fatal_signal_pending is true.

I have tested this and verified I did not break SIGKILL during
coredumps by accident (before or after this change).  I actually
thought I had and I had to figure out what I had misread that kept
SIGKILL during coredumps working.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c   |  4 ++--
 kernel/signal.c | 11 +++++++++--
 2 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index a6b3c196cdef..7b91fb32dbb8 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -448,7 +448,7 @@ static void coredump_finish(bool core_dumped)
 static bool dump_interrupted(void)
 {
 	/*
-	 * SIGKILL or freezing() interrupt the coredumping. Perhaps we
+	 * SIGKILL or freezing() interrupted the coredumping. Perhaps we
 	 * can do try_to_freeze() and check __fatal_signal_pending(),
 	 * but then we need to teach dump_write() to restart and clear
 	 * TIF_SIGPENDING.
@@ -471,7 +471,7 @@ static void wait_for_dump_helpers(struct file *file)
 	 * We actually want wait_event_freezable() but then we need
 	 * to clear TIF_SIGPENDING and improve dump_interrupted().
 	 */
-	wait_event_interruptible(pipe->rd_wait, pipe->readers == 1);
+	wait_event_killable(pipe->rd_wait, pipe->readers == 1);
 
 	pipe_lock(pipe);
 	pipe->readers--;
diff --git a/kernel/signal.c b/kernel/signal.c
index 8272cac5f429..7e305a8ec7c2 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -907,8 +907,15 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
 	sigset_t flush;
 
 	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
-		if (!(signal->flags & SIGNAL_GROUP_EXIT))
-			return sig == SIGKILL;
+		struct core_state *core_state = signal->core_state;
+		if (core_state) {
+			if (sig == SIGKILL) {
+				struct task_struct *dumper = core_state->dumper.task;
+				sigaddset(&dumper->pending.signal, SIGKILL);
+				signal_wake_up(dumper, 1);
+			}
+			return false;
+		}
 		/*
 		 * The process is in the middle of dying, nothing to do.
 		 */
-- 
2.29.2

[PATCH 2/8] signal: Drop signals received after a fatal signal has been processed

[Eric W. Biederman]

In 403bad72b67d ("coredump: only SIGKILL should interrupt the
coredumping task") Oleg modified the kernel to drop all signals that
come in during a coredump except SIGKILL, and suggested that it might
be a good idea to generalize that to other cases after the process has
received a fatal signal.

Semantically it does not make sense to perform any signal delivery
after the process has already been killed.

When a signal is sent while a process is dying today the signal is
placed in the signal queue by __send_signal and a single task of the
process is woken up with signal_wake_up, if there are any tasks that
have not set PF_EXITING.

Take things one step farther and have prepare_signal report that all
signals that come after a process has been killed should be ignored.
While retaining the historical exception of allowing SIGKILL to
interrupt coredumps.

Remove the SIGNAL_GROUP_EXIT test from complete_signal, as it is no
longer possible for signal processing to reach complete_signal when
SIGNAL_GROUP_EXIT is true.

Update the comment in fs/coredump.c to make it clear coredumps are
special in being able to receive SIGKILL.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c   | 2 +-
 kernel/signal.c | 5 ++---
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index 7b91fb32dbb8..a9c25f20118f 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -352,7 +352,7 @@ static int zap_process(struct task_struct *start, int exit_code, int flags)
 	struct task_struct *t;
 	int nr = 0;
 
-	/* ignore all signals except SIGKILL, see prepare_signal() */
+	/* Allow SIGKILL, see prepare_signal() */
 	start->signal->flags = SIGNAL_GROUP_COREDUMP | flags;
 	start->signal->group_exit_code = exit_code;
 	start->signal->group_stop_count = 0;
diff --git a/kernel/signal.c b/kernel/signal.c
index 7e305a8ec7c2..cdccbacac685 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -914,11 +914,11 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
 				sigaddset(&dumper->pending.signal, SIGKILL);
 				signal_wake_up(dumper, 1);
 			}
-			return false;
 		}
 		/*
-		 * The process is in the middle of dying, nothing to do.
+		 * The process is in the middle of dying, drop the signal.
 		 */
+		return false;
 	} else if (sig_kernel_stop(sig)) {
 		/*
 		 * This is a stop signal.  Remove SIGCONT from all queues.
@@ -1039,7 +1039,6 @@ static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
 	 * then start taking the whole group down immediately.
 	 */
 	if (sig_fatal(p, sig) &&
-	    !(signal->flags & SIGNAL_GROUP_EXIT) &&
 	    !sigismember(&t->real_blocked, sig) &&
 	    (sig == SIGKILL || !p->ptrace)) {
 		/*
-- 
2.29.2

[PATCH 3/8] signal: Have the oom killer detect coredumps using signal->core_state

[Eric W. Biederman]

In preparation for removing the flag SIGNAL_GROUP_COREDUMP change
__task_will_free_mem to test signal->core_state instead of the flag
SIGNAL_GROUP_COREDUMP.

Both fields are protected by siglock and both live in signal_struct so
there are no real tradeoffs here, just a change to which field is
being tested.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 mm/oom_kill.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 1ddabefcfb5a..5c92aad8ca1a 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -793,7 +793,7 @@ static inline bool __task_will_free_mem(struct task_struct *task)
 	 * coredump_task_exit(), so the oom killer cannot assume that
 	 * the process will promptly exit and release memory.
 	 */
-	if (sig->flags & SIGNAL_GROUP_COREDUMP)
+	if (sig->core_state)
 		return false;
 
 	if (sig->flags & SIGNAL_GROUP_EXIT)
-- 
2.29.2

[PATCH 4/8] signal: During coredumps set SIGNAL_GROUP_EXIT in zap_process

[Eric W. Biederman]

There are only a few places that test SIGNAL_GROUP_EXIT and
are not also already testing SIGNAL_GROUP_COREDUMP.

This will not affect the callers of signal_group_exit as zap_process
also sets group_exit_task so signal_group_exit will continue to return
true at the same times.

This does not affect wait_task_zombie as the none of the threads
wind up in EXIT_ZOMBIE state during a coredump.

This does not affect oom_kill.c:__task_will_free_mem as
sig->core_state is tested and handled before SIGNAL_GROUP_EXIT is
tested for.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index a9c25f20118f..5e5a90de7be3 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -347,13 +347,13 @@ static int format_corename(struct core_name *cn, struct coredump_params *cprm,
 	return ispipe;
 }
 
-static int zap_process(struct task_struct *start, int exit_code, int flags)
+static int zap_process(struct task_struct *start, int exit_code)
 {
 	struct task_struct *t;
 	int nr = 0;
 
 	/* Allow SIGKILL, see prepare_signal() */
-	start->signal->flags = SIGNAL_GROUP_COREDUMP | flags;
+	start->signal->flags = SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP;
 	start->signal->group_exit_code = exit_code;
 	start->signal->group_stop_count = 0;
 
@@ -378,7 +378,7 @@ static int zap_threads(struct task_struct *tsk,
 	if (!signal_group_exit(tsk->signal)) {
 		tsk->signal->core_state = core_state;
 		tsk->signal->group_exit_task = tsk;
-		nr = zap_process(tsk, exit_code, 0);
+		nr = zap_process(tsk, exit_code);
 		clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
 		tsk->flags |= PF_DUMPCORE;
 		atomic_set(&core_state->nr_threads, nr);
-- 
2.29.2

[PATCH 5/8] signal: Remove SIGNAL_GROUP_COREDUMP

[Eric W. Biederman]

After the previous cleanups "signal->core_state" is set whenever
SIGNAL_GROUP_COREDUMP is set and "signal->core_state" is tested
whenver the code wants to know if a coredump is in progress.  The
remaining tests of SIGNAL_GROUP_COREDUMP also test to see if
SIGNAL_GROUP_EXIT is set.  Similarly the only place that sets
SIGNAL_GROUP_COREDUMP also sets SIGNAL_GROUP_EXIT.

Which makes SIGNAL_GROUP_COREDUMP unecessary and redundant so
stop setting SIGNAL_GROUP_COREDUMP, stop testing SIGNAL_GROUP_COREDUMP
and remove it's definition makeing the code slightly simpler.

With the setting of SIGNAL_GROUP_COREDUMP gone coredump_finish no
longer needs to clear SIGNAL_GROUP_COREDUMP out of signal->flags
by setting SIGNAL_GROUP_EXIT.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c                | 3 +--
 include/linux/sched/signal.h | 3 +--
 kernel/signal.c              | 2 +-
 3 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index 5e5a90de7be3..2c9d16d4b57a 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -353,7 +353,7 @@ static int zap_process(struct task_struct *start, int exit_code)
 	int nr = 0;
 
 	/* Allow SIGKILL, see prepare_signal() */
-	start->signal->flags = SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP;
+	start->signal->flags = SIGNAL_GROUP_EXIT;
 	start->signal->group_exit_code = exit_code;
 	start->signal->group_stop_count = 0;
 
@@ -427,7 +427,6 @@ static void coredump_finish(bool core_dumped)
 	if (core_dumped && !__fatal_signal_pending(current))
 		current->signal->group_exit_code |= 0x80;
 	current->signal->group_exit_task = NULL;
-	current->signal->flags = SIGNAL_GROUP_EXIT;
 	next = current->signal->core_state->dumper.next;
 	current->signal->core_state = NULL;
 	spin_unlock_irq(&current->sighand->siglock);
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index fa26d2a58413..ecc10e148799 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -256,7 +256,6 @@ struct signal_struct {
 #define SIGNAL_STOP_STOPPED	0x00000001 /* job control stop in effect */
 #define SIGNAL_STOP_CONTINUED	0x00000002 /* SIGCONT since WCONTINUED reap */
 #define SIGNAL_GROUP_EXIT	0x00000004 /* group exit in progress */
-#define SIGNAL_GROUP_COREDUMP	0x00000008 /* coredump in progress */
 /*
  * Pending notifications to parent.
  */
@@ -272,7 +271,7 @@ struct signal_struct {
 static inline void signal_set_stop_flags(struct signal_struct *sig,
 					 unsigned int flags)
 {
-	WARN_ON(sig->flags & (SIGNAL_GROUP_EXIT|SIGNAL_GROUP_COREDUMP));
+	WARN_ON(sig->flags & SIGNAL_GROUP_EXIT);
 	sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
 }
 
diff --git a/kernel/signal.c b/kernel/signal.c
index cdccbacac685..9eb3e2c1f9f7 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -906,7 +906,7 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
 	struct task_struct *t;
 	sigset_t flush;
 
-	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
+	if (signal->flags & SIGNAL_GROUP_EXIT) {
 		struct core_state *core_state = signal->core_state;
 		if (core_state) {
 			if (sig == SIGKILL) {
-- 
2.29.2

[PATCH 6/8] coredump: Stop setting signal->group_exit_task

[Eric W. Biederman]

Currently the coredump code sets group_exit_task so that
signal_group_exit() will return true during a coredump.  Now that the
coredump code always sets SIGNAL_GROUP_EXIT there is no longer a need
to set signal->group_exit_task.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index 2c9d16d4b57a..ef56595a0d87 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -377,7 +377,6 @@ static int zap_threads(struct task_struct *tsk,
 	spin_lock_irq(&tsk->sighand->siglock);
 	if (!signal_group_exit(tsk->signal)) {
 		tsk->signal->core_state = core_state;
-		tsk->signal->group_exit_task = tsk;
 		nr = zap_process(tsk, exit_code);
 		clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
 		tsk->flags |= PF_DUMPCORE;
@@ -426,7 +425,6 @@ static void coredump_finish(bool core_dumped)
 	spin_lock_irq(&current->sighand->siglock);
 	if (core_dumped && !__fatal_signal_pending(current))
 		current->signal->group_exit_code |= 0x80;
-	current->signal->group_exit_task = NULL;
 	next = current->signal->core_state->dumper.next;
 	current->signal->core_state = NULL;
 	spin_unlock_irq(&current->sighand->siglock);
-- 
2.29.2

[PATCH 7/8] signal: Rename group_exit_task group_exec_task

[Eric W. Biederman]

The only remaining user of group_exit_task is exec.  Rename the field
so that it is clear which part of the code uses it.

Update the comment above the definition of group_exec_task
to document how it is currently used.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/exec.c                    |  8 ++++----
 include/linux/sched/signal.h | 12 ++++--------
 kernel/exit.c                |  4 ++--
 3 files changed, 10 insertions(+), 14 deletions(-)

diff --git a/fs/exec.c b/fs/exec.c
index 59cac7c18178..9d2925811011 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1054,7 +1054,7 @@ static int de_thread(struct task_struct *tsk)
 		return -EAGAIN;
 	}
 
-	sig->group_exit_task = tsk;
+	sig->group_exec_task = tsk;
 	sig->notify_count = zap_other_threads(tsk);
 	if (!thread_group_leader(tsk))
 		sig->notify_count--;
@@ -1082,7 +1082,7 @@ static int de_thread(struct task_struct *tsk)
 			write_lock_irq(&tasklist_lock);
 			/*
 			 * Do this under tasklist_lock to ensure that
-			 * exit_notify() can't miss ->group_exit_task
+			 * exit_notify() can't miss ->group_exec_task
 			 */
 			sig->notify_count = -1;
 			if (likely(leader->exit_state))
@@ -1149,7 +1149,7 @@ static int de_thread(struct task_struct *tsk)
 		release_task(leader);
 	}
 
-	sig->group_exit_task = NULL;
+	sig->group_exec_task = NULL;
 	sig->notify_count = 0;
 
 no_thread_group:
@@ -1162,7 +1162,7 @@ static int de_thread(struct task_struct *tsk)
 killed:
 	/* protects against exit_notify() and __exit_signal() */
 	read_lock(&tasklist_lock);
-	sig->group_exit_task = NULL;
+	sig->group_exec_task = NULL;
 	sig->notify_count = 0;
 	read_unlock(&tasklist_lock);
 	return -EAGAIN;
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index ecc10e148799..d3248aba5183 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -109,13 +109,9 @@ struct signal_struct {
 
 	/* thread group exit support */
 	int			group_exit_code;
-	/* overloaded:
-	 * - notify group_exit_task when ->count is equal to notify_count
-	 * - everyone except group_exit_task is stopped during signal delivery
-	 *   of fatal signals, group_exit_task processes the signal.
-	 */
+	/* notify group_exec_task when notify_count is less or equal to 0 */
 	int			notify_count;
-	struct task_struct	*group_exit_task;
+	struct task_struct	*group_exec_task;
 
 	/* thread group stop support, overloads group_exit_code too */
 	int			group_stop_count;
@@ -275,11 +271,11 @@ static inline void signal_set_stop_flags(struct signal_struct *sig,
 	sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
 }
 
-/* If true, all threads except ->group_exit_task have pending SIGKILL */
+/* If true, all threads except ->group_exec_task have pending SIGKILL */
 static inline int signal_group_exit(const struct signal_struct *sig)
 {
 	return	(sig->flags & SIGNAL_GROUP_EXIT) ||
-		(sig->group_exit_task != NULL);
+		(sig->group_exec_task != NULL);
 }
 
 extern void flush_signals(struct task_struct *);
diff --git a/kernel/exit.c b/kernel/exit.c
index 6c4b04531f17..527c5e4430ae 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -116,7 +116,7 @@ static void __exit_signal(struct task_struct *tsk)
 		 * then notify it:
 		 */
 		if (sig->notify_count > 0 && !--sig->notify_count)
-			wake_up_process(sig->group_exit_task);
+			wake_up_process(sig->group_exec_task);
 
 		if (tsk == sig->curr_target)
 			sig->curr_target = next_thread(tsk);
@@ -697,7 +697,7 @@ static void exit_notify(struct task_struct *tsk, int group_dead)
 
 	/* mt-exec, de_thread() is waiting for group leader */
 	if (unlikely(tsk->signal->notify_count < 0))
-		wake_up_process(tsk->signal->group_exit_task);
+		wake_up_process(tsk->signal->group_exec_task);
 	write_unlock_irq(&tasklist_lock);
 
 	list_for_each_entry_safe(p, n, &dead, ptrace_entry) {
-- 
2.29.2

[PATCH 8/8] signal: Remove the helper signal_group_exit

[Eric W. Biederman]

This helper is misleading.  It tests for an ongoing exec as well as
the process having received a fatal signal.

Sometimes it is appropriate to treat an on-going exec differently than
a process that is shutting down due to a fatal signal.  In particular
taking the fast path out of exit_signals instead of retargeting
signals is not appropriate during exec, and not changing the the exit
code in do_group_exit during exec.

Removing the helper so that both cases must be coded for explicitly
makes it more obvious what is going on as both cases must be coded for
explicitly.

While removing the helper fix the two cases where I have observed
using signal_group_helper resulted in the wrong result.

For the unset exit_code in do_group_exit during an exec I use 0 as I
think that is what group_exit_code has been set to most of the time.
During a thread group stop group_exit_code is set to the stop signal
and when the thread group receives SIGCONT group_exit_code is reset to
0.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c                | 5 +++--
 fs/exec.c                    | 2 +-
 include/linux/sched/signal.h | 7 -------
 kernel/exit.c                | 8 ++++++--
 kernel/signal.c              | 8 +++++---
 5 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index ef56595a0d87..09302a6a0d80 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -372,11 +372,12 @@ static int zap_process(struct task_struct *start, int exit_code)
 static int zap_threads(struct task_struct *tsk,
 			struct core_state *core_state, int exit_code)
 {
+	struct signal_struct *signal = tsk->signal;
 	int nr = -EAGAIN;
 
 	spin_lock_irq(&tsk->sighand->siglock);
-	if (!signal_group_exit(tsk->signal)) {
-		tsk->signal->core_state = core_state;
+	if (!(signal->flags & SIGNAL_GROUP_EXIT) && !signal->group_exec_task) {
+		signal->core_state = core_state;
 		nr = zap_process(tsk, exit_code);
 		clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
 		tsk->flags |= PF_DUMPCORE;
diff --git a/fs/exec.c b/fs/exec.c
index 9d2925811011..82db656ca709 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1045,7 +1045,7 @@ static int de_thread(struct task_struct *tsk)
 	 * Kill all other threads in the thread group.
 	 */
 	spin_lock_irq(lock);
-	if (signal_group_exit(sig)) {
+	if ((sig->flags & SIGNAL_GROUP_EXIT) || sig->group_exec_task) {
 		/*
 		 * Another group action in progress, just
 		 * return so that the signal is processed.
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index d3248aba5183..b6ecb9fc4cd2 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -271,13 +271,6 @@ static inline void signal_set_stop_flags(struct signal_struct *sig,
 	sig->flags = (sig->flags & ~SIGNAL_STOP_MASK) | flags;
 }
 
-/* If true, all threads except ->group_exec_task have pending SIGKILL */
-static inline int signal_group_exit(const struct signal_struct *sig)
-{
-	return	(sig->flags & SIGNAL_GROUP_EXIT) ||
-		(sig->group_exec_task != NULL);
-}
-
 extern void flush_signals(struct task_struct *);
 extern void ignore_signals(struct task_struct *);
 extern void flush_signal_handlers(struct task_struct *, int force_default);
diff --git a/kernel/exit.c b/kernel/exit.c
index 527c5e4430ae..e7104f803be0 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -907,15 +907,19 @@ do_group_exit(int exit_code)
 
 	BUG_ON(exit_code & 0x80); /* core dumps don't get here */
 
-	if (signal_group_exit(sig))
+	if (sig->flags & SIGNAL_GROUP_EXIT)
 		exit_code = sig->group_exit_code;
+	else if (sig->group_exec_task)
+		exit_code = 0;
 	else if (!thread_group_empty(current)) {
 		struct sighand_struct *const sighand = current->sighand;
 
 		spin_lock_irq(&sighand->siglock);
-		if (signal_group_exit(sig))
+		if (sig->flags & SIGNAL_GROUP_EXIT)
 			/* Another thread got here before we took the lock.  */
 			exit_code = sig->group_exit_code;
+		else if (sig->group_exec_task)
+			exit_code = 0;
 		else {
 			sig->group_exit_code = exit_code;
 			sig->flags = SIGNAL_GROUP_EXIT;
diff --git a/kernel/signal.c b/kernel/signal.c
index 9eb3e2c1f9f7..860d844542b2 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2392,7 +2392,8 @@ static bool do_signal_stop(int signr)
 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
 
 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
-		    unlikely(signal_group_exit(sig)))
+		    unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
+		    unlikely(sig->group_exec_task))
 			return false;
 		/*
 		 * There is no group stop already in progress.  We must
@@ -2699,7 +2700,8 @@ bool get_signal(struct ksignal *ksig)
 		enum pid_type type;
 
 		/* Has this task already been marked for death? */
-		if (signal_group_exit(signal)) {
+		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
+		     signal->group_exec_task) {
 			ksig->info.si_signo = signr = SIGKILL;
 			sigdelset(&current->pending.signal, SIGKILL);
 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
@@ -2955,7 +2957,7 @@ void exit_signals(struct task_struct *tsk)
 	 */
 	cgroup_threadgroup_change_begin(tsk);
 
-	if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
+	if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
 		tsk->flags |= PF_EXITING;
 		cgroup_threadgroup_change_end(tsk);
 		return;
-- 
2.29.2

[PATCH 00/17] exit: Making task exiting a first class concept

[Eric W. Biederman]

The changes below contain some cleanups and the work to make implement
first class asynchronous task exit.  Most of the cleanups are necessary
for this work but a couple of them (removing profile_task_exit and the
extra setting of PT_SEIZED in ptrace_attach) are included because I
stumbled over them and they are worth applying but they aren't
interesting enough to me to make be in their own patchset.

The core of this set of changes is the addition of
schedule_task_exit_locked.  Ptrace is cleaned up to avoid a conflict in
task->exit_code.  Then the existing task exit code is gradually moved
into the final shape of schedule_task_exit_locked.

This is the fundamental building block I need to fix alpha, m68k,
nios2 and any other architecture that does not always save all of
their registers except when entering into a ptrace context.

This is about half the work to allow coredump signals to use
short-circuit delivery.

With coredumps signals available for short-circuit delivery the
SA_IMMUTABLE hack can be replace by something clean.

The counting of the number of threads that have not been killed to
always set SIGNAL_GROUP_EXIT when a process exits and the coredump
signal short-circuit delivery is a foundation for updating the
SECCOMP_RET_KILL_THREAD implementation such that it can decide if it
should coredump without races.

I have most of those changes pretty much ready I just need to get these
changes finalized reviewed first.  At this point they are looking at
v5.18 material.

These patches are on top of:
https://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git/ signal-for-v5.17

After these patches have been reviewed it is my plan to apply them to my
signal-for-v5.17 branch.  Any and all feedback is welcome.

Eric W. Biederman (17):
      exit: Remove profile_task_exit & profile_munmap
      exit: Coredumps reach do_group_exit
      exit: Fix the exit_code for wait_task_zombie
      exit: Use the correct exit_code in /proc/<pid>/stat
      taskstats: Cleanup the use of task->exit_code
      ptrace: Remove second setting of PT_SEIZED in ptrace_attach
      ptrace: Remove unused regs argument from ptrace_report_syscall
      ptrace/m68k: Stop open coding ptrace_report_syscall
      ptrace: Move setting/clearing ptrace_message into prace_stop
      ptrace: Return the signal to continue with from ptrace_stop
      ptrace: Separate task->ptrace_code out from task->exit_code
      signal: Compute the process exit_code in get_signal
      signal: Make individual tasks exiting a first class concept
      signal: Remove zap_other_threads
      signal: Add JOBCTL_WILL_EXIT to mark exiting tasks
      signal: Record the exit_code when an exit is scheduled
      signal: Always set SIGNAL_GROUP_EXIT on process exit

 arch/m68k/kernel/ptrace.c    |  12 +----
 fs/coredump.c                |  17 +++---
 fs/exec.c                    |  12 +++--
 fs/proc/array.c              |   9 +++-
 include/linux/profile.h      |  26 ---------
 include/linux/ptrace.h       |   5 +-
 include/linux/sched.h        |   1 +
 include/linux/sched/jobctl.h |   2 +
 include/linux/sched/signal.h |   6 ++-
 include/linux/tracehook.h    |  21 ++++----
 kernel/exit.c                |  29 +++++-----
 kernel/fork.c                |   2 +
 kernel/profile.c             |  50 ------------------
 kernel/ptrace.c              |  14 +++--
 kernel/signal.c              | 122 +++++++++++++++++++++++--------------------
 kernel/tsacct.c              |   7 ++-
 mm/mmap.c                    |   1 -
 17 files changed, 134 insertions(+), 202 deletions(-)

[PATCH 01/17] exit: Remove profile_task_exit & profile_munmap

[Eric W. Biederman]

When I say remove I mean remove.  All profile_task_exit and
profile_munmap do is call a blocking notifier chain.  The helpers
profile_task_register and profile_task_unregister are not called
anywhere in the tree.  Which means this is all dead code.

So remove the dead code and make it easier to read do_exit.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 include/linux/profile.h | 26 ---------------------
 kernel/exit.c           |  1 -
 kernel/profile.c        | 50 -----------------------------------------
 mm/mmap.c               |  1 -
 4 files changed, 78 deletions(-)

diff --git a/include/linux/profile.h b/include/linux/profile.h
index fd18ca96f557..f7eb2b57d890 100644
--- a/include/linux/profile.h
+++ b/include/linux/profile.h
@@ -31,11 +31,6 @@ static inline int create_proc_profile(void)
 }
 #endif
 
-enum profile_type {
-	PROFILE_TASK_EXIT,
-	PROFILE_MUNMAP
-};
-
 #ifdef CONFIG_PROFILING
 
 extern int prof_on __read_mostly;
@@ -66,23 +61,14 @@ static inline void profile_hit(int type, void *ip)
 struct task_struct;
 struct mm_struct;
 
-/* task is in do_exit() */
-void profile_task_exit(struct task_struct * task);
-
 /* task is dead, free task struct ? Returns 1 if
  * the task was taken, 0 if the task should be freed.
  */
 int profile_handoff_task(struct task_struct * task);
 
-/* sys_munmap */
-void profile_munmap(unsigned long addr);
-
 int task_handoff_register(struct notifier_block * n);
 int task_handoff_unregister(struct notifier_block * n);
 
-int profile_event_register(enum profile_type, struct notifier_block * n);
-int profile_event_unregister(enum profile_type, struct notifier_block * n);
-
 #else
 
 #define prof_on 0
@@ -117,19 +103,7 @@ static inline int task_handoff_unregister(struct notifier_block * n)
 	return -ENOSYS;
 }
 
-static inline int profile_event_register(enum profile_type t, struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-static inline int profile_event_unregister(enum profile_type t, struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-#define profile_task_exit(a) do { } while (0)
 #define profile_handoff_task(a) (0)
-#define profile_munmap(a) do { } while (0)
 
 #endif /* CONFIG_PROFILING */
 
diff --git a/kernel/exit.c b/kernel/exit.c
index e7104f803be0..b5c35b520fda 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -737,7 +737,6 @@ void __noreturn do_exit(long code)
 
 	WARN_ON(blk_needs_flush_plug(tsk));
 
-	profile_task_exit(tsk);
 	kcov_task_exit(tsk);
 
 	coredump_task_exit(tsk);
diff --git a/kernel/profile.c b/kernel/profile.c
index eb9c7f0f5ac5..9355cc934a96 100644
--- a/kernel/profile.c
+++ b/kernel/profile.c
@@ -135,14 +135,7 @@ int __ref profile_init(void)
 
 /* Profile event notifications */
 
-static BLOCKING_NOTIFIER_HEAD(task_exit_notifier);
 static ATOMIC_NOTIFIER_HEAD(task_free_notifier);
-static BLOCKING_NOTIFIER_HEAD(munmap_notifier);
-
-void profile_task_exit(struct task_struct *task)
-{
-	blocking_notifier_call_chain(&task_exit_notifier, 0, task);
-}
 
 int profile_handoff_task(struct task_struct *task)
 {
@@ -151,11 +144,6 @@ int profile_handoff_task(struct task_struct *task)
 	return (ret == NOTIFY_OK) ? 1 : 0;
 }
 
-void profile_munmap(unsigned long addr)
-{
-	blocking_notifier_call_chain(&munmap_notifier, 0, (void *)addr);
-}
-
 int task_handoff_register(struct notifier_block *n)
 {
 	return atomic_notifier_chain_register(&task_free_notifier, n);
@@ -168,44 +156,6 @@ int task_handoff_unregister(struct notifier_block *n)
 }
 EXPORT_SYMBOL_GPL(task_handoff_unregister);
 
-int profile_event_register(enum profile_type type, struct notifier_block *n)
-{
-	int err = -EINVAL;
-
-	switch (type) {
-	case PROFILE_TASK_EXIT:
-		err = blocking_notifier_chain_register(
-				&task_exit_notifier, n);
-		break;
-	case PROFILE_MUNMAP:
-		err = blocking_notifier_chain_register(
-				&munmap_notifier, n);
-		break;
-	}
-
-	return err;
-}
-EXPORT_SYMBOL_GPL(profile_event_register);
-
-int profile_event_unregister(enum profile_type type, struct notifier_block *n)
-{
-	int err = -EINVAL;
-
-	switch (type) {
-	case PROFILE_TASK_EXIT:
-		err = blocking_notifier_chain_unregister(
-				&task_exit_notifier, n);
-		break;
-	case PROFILE_MUNMAP:
-		err = blocking_notifier_chain_unregister(
-				&munmap_notifier, n);
-		break;
-	}
-
-	return err;
-}
-EXPORT_SYMBOL_GPL(profile_event_unregister);
-
 #if defined(CONFIG_SMP) && defined(CONFIG_PROC_FS)
 /*
  * Each cpu has a pair of open-addressed hashtables for pending
diff --git a/mm/mmap.c b/mm/mmap.c
index bfb0ea164a90..70318c2a47c3 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2928,7 +2928,6 @@ EXPORT_SYMBOL(vm_munmap);
 SYSCALL_DEFINE2(munmap, unsigned long, addr, size_t, len)
 {
 	addr = untagged_addr(addr);
-	profile_munmap(addr);
 	return __vm_munmap(addr, len, true);
 }
 
-- 
2.29.2

[PATCH 02/17] exit: Coredumps reach do_group_exit

[Eric W. Biederman]

The comment about coredumps not reaching do_group_exit and the
corresponding BUG_ON are bogus.

What happens and has happened for years is that get_signal calls
do_coredump (which sets SIGNAL_GROUP_EXIT and group_exit_code) and
then do_group_exit passing the signal number.  Then do_group_exit
ignores the exit_code it is passed and uses signal->group_exit_code
from the coredump.

The comment and BUG_ON were correct when they were added during the
2.5 development cycle, but became obsolete and incorrect when
get_signal was changed to fall through to do_group_exit after
do_coredump in 2.6.10-rc2.

So remove the stale comment and BUG_ON

Fixes: 63bd6144f191 ("[PATCH] Invalid BUG_ONs in signal.c")
History-Tree: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 kernel/exit.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/kernel/exit.c b/kernel/exit.c
index b5c35b520fda..34c43037450f 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -904,8 +904,6 @@ do_group_exit(int exit_code)
 {
 	struct signal_struct *sig = current->signal;
 
-	BUG_ON(exit_code & 0x80); /* core dumps don't get here */
-
 	if (sig->flags & SIGNAL_GROUP_EXIT)
 		exit_code = sig->group_exit_code;
 	else if (sig->group_exec_task)
-- 
2.29.2

[PATCH 03/17] exit: Fix the exit_code for wait_task_zombie

[Eric W. Biederman]

The function wait_task_zombie is defined to always returns the process not
thread exit status.  Unfortunately when process group exit support
was added to wait_task_zombie the WNOWAIT case was overlooked.

Usually tsk->exit_code and tsk->signal->group_exit_code will be in sync
so fixing this is bug probably has no effect in practice.  But fix
it anyway so that people aren't scratching their heads about why
the two code paths are different.

History-Tree: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Fixes: 2c66151cbc2c ("[PATCH] sys_exit() threading improvements, BK-curr")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 kernel/exit.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/exit.c b/kernel/exit.c
index 34c43037450f..7121db37c411 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -1011,7 +1011,8 @@ static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
 		return 0;
 
 	if (unlikely(wo->wo_flags & WNOWAIT)) {
-		status = p->exit_code;
+		status = (p->signal->flags & SIGNAL_GROUP_EXIT)
+			? p->signal->group_exit_code : p->exit_code;
 		get_task_struct(p);
 		read_unlock(&tasklist_lock);
 		sched_annotate_sleep();
-- 
2.29.2

[PATCH 04/17] exit: Use the correct exit_code in /proc/<pid>/stat

[Eric W. Biederman]

Since do_proc_statt was modified to return process wide values instead
of per task values the exit_code calculation has never been updated.
Update it now to return the process wide exit_code when it is requested
and available.

History-Tree: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Fixes: bf719d26a5c1 ("[PATCH] distinct tgid/tid CPU usage")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/proc/array.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/fs/proc/array.c b/fs/proc/array.c
index ff869a66b34e..43a7abde9e42 100644
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -468,6 +468,7 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 	u64 cgtime, gtime;
 	unsigned long rsslim = 0;
 	unsigned long flags;
+	int exit_code = task->exit_code;
 
 	state = *get_task_state(task);
 	vsize = eip = esp = 0;
@@ -531,6 +532,9 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 			maj_flt += sig->maj_flt;
 			thread_group_cputime_adjusted(task, &utime, &stime);
 			gtime += sig->gtime;
+
+			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
+				exit_code = sig->group_exit_code;
 		}
 
 		sid = task_session_nr_ns(task, ns);
@@ -630,7 +634,7 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 		seq_puts(m, " 0 0 0 0 0 0 0");
 
 	if (permitted)
-		seq_put_decimal_ll(m, " ", task->exit_code);
+		seq_put_decimal_ll(m, " ", exit_code);
 	else
 		seq_puts(m, " 0");
 
-- 
2.29.2

[PATCH 05/17] taskstats: Cleanup the use of task->exit_code

[Eric W. Biederman]

In the function bacct_add_task the code reading task->exit_code was
introduced in commit f3cef7a99469 ("[PATCH] csa: basic accounting over
taskstats"), and it is not entirely clear what the taskstats interface
is trying to return as only returning the exit_code of the first task
in a process doesn't make a lot of sense.

As best as I can figure the intent is to return task->exit_code after
a task exits.  The field is returned with per task fields, so the
exit_code of the entire process is not wanted.  Only the value of the
first task is returned so this is not a useful way to get the per task
ptrace stop code.  The ordinary case of returning this value is
returning after a task exits, which also precludes use for getting
a ptrace value.

It is common to for the first task of a process to also be the last
task of a process so this field may have done something reasonable by
accident in testing.

Make ac_exitcode a reliable per task value by always returning it for
every exited task.

Setting ac_exitcode in a sensible mannter makes it possible to continue
to provide this value going forward.

Cc: Balbir Singh <bsingharora@gmail.com>
Fixes: f3cef7a99469 ("[PATCH] csa: basic accounting over taskstats")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 kernel/tsacct.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/kernel/tsacct.c b/kernel/tsacct.c
index f00de83d0246..1d261fbe367b 100644
--- a/kernel/tsacct.c
+++ b/kernel/tsacct.c
@@ -38,11 +38,10 @@ void bacct_add_tsk(struct user_namespace *user_ns,
 	stats->ac_btime = clamp_t(time64_t, btime, 0, U32_MAX);
 	stats->ac_btime64 = btime;
 
-	if (thread_group_leader(tsk)) {
+	if (tsk->flags & PF_EXITING)
 		stats->ac_exitcode = tsk->exit_code;
-		if (tsk->flags & PF_FORKNOEXEC)
-			stats->ac_flag |= AFORK;
-	}
+	if (thread_group_leader(tsk) && (tsk->flags & PF_FORKNOEXEC))
+		stats->ac_flag |= AFORK;
 	if (tsk->flags & PF_SUPERPRIV)
 		stats->ac_flag |= ASU;
 	if (tsk->flags & PF_DUMPCORE)
-- 
2.29.2

[PATCH 06/17] ptrace: Remove second setting of PT_SEIZED in ptrace_attach

[Eric W. Biederman]

The code is totally redundant remove it.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 kernel/ptrace.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index f8589bf8d7dc..eea265082e97 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -419,8 +419,6 @@ static int ptrace_attach(struct task_struct *task, long request,
 	if (task->ptrace)
 		goto unlock_tasklist;
 
-	if (seize)
-		flags |= PT_SEIZED;
 	task->ptrace = flags;
 
 	ptrace_link(task, current);
-- 
2.29.2

[PATCH 07/17] ptrace: Remove unused regs argument from ptrace_report_syscall

[Eric W. Biederman]

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 include/linux/tracehook.h | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/include/linux/tracehook.h b/include/linux/tracehook.h
index 2564b7434b4d..88c007ab5ebc 100644
--- a/include/linux/tracehook.h
+++ b/include/linux/tracehook.h
@@ -54,8 +54,7 @@ struct linux_binprm;
 /*
  * ptrace report for syscall entry and exit looks identical.
  */
-static inline int ptrace_report_syscall(struct pt_regs *regs,
-					unsigned long message)
+static inline int ptrace_report_syscall(unsigned long message)
 {
 	int ptrace = current->ptrace;
 
@@ -102,7 +101,7 @@ static inline int ptrace_report_syscall(struct pt_regs *regs,
 static inline __must_check int tracehook_report_syscall_entry(
 	struct pt_regs *regs)
 {
-	return ptrace_report_syscall(regs, PTRACE_EVENTMSG_SYSCALL_ENTRY);
+	return ptrace_report_syscall(PTRACE_EVENTMSG_SYSCALL_ENTRY);
 }
 
 /**
@@ -127,7 +126,7 @@ static inline void tracehook_report_syscall_exit(struct pt_regs *regs, int step)
 	if (step)
 		user_single_step_report(regs);
 	else
-		ptrace_report_syscall(regs, PTRACE_EVENTMSG_SYSCALL_EXIT);
+		ptrace_report_syscall(PTRACE_EVENTMSG_SYSCALL_EXIT);
 }
 
 /**
-- 
2.29.2

[PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Eric W. Biederman]

The generic function ptrace_report_syscall does a little more
than syscall_trace on m68k.  The function ptrace_report_syscall
stops early if PT_TRACED is not set, it sets ptrace_message,
and returns the result of fatal_signal_pending.

Setting ptrace_message to a passed in value of 0 is effectively not
setting ptrace_message, making that additional work a noop.

Returning the result of fatal_signal_pending and letting the caller
ignore the result becomes a noop in this change.

When a process is ptraced, the flag PT_PTRACED is always set in
current->ptrace.  Testing for PT_PTRACED in ptrace_report_syscall is
just an optimization to fail early if the process is not ptraced.
Later on in ptrace_notify, ptrace_stop will test current->ptrace under
tasklist_lock and skip performing any work if the task is not ptraced.

Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 arch/m68k/kernel/ptrace.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)

diff --git a/arch/m68k/kernel/ptrace.c b/arch/m68k/kernel/ptrace.c
index 94b3b274186d..aa3a0b8d07e9 100644
--- a/arch/m68k/kernel/ptrace.c
+++ b/arch/m68k/kernel/ptrace.c
@@ -273,17 +273,7 @@ long arch_ptrace(struct task_struct *child, long request,
 
 asmlinkage void syscall_trace(void)
 {
-	ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD)
-				 ? 0x80 : 0));
-	/*
-	 * this isn't the same as continuing with a signal, but it will do
-	 * for normal use.  strace only continues with a signal if the
-	 * stopping signal is not SIGTRAP.  -brl
-	 */
-	if (current->exit_code) {
-		send_sig(current->exit_code, current, 1);
-		current->exit_code = 0;
-	}
+	ptrace_report_syscall(0);
 }
 
 #if defined(CONFIG_COLDFIRE) || !defined(CONFIG_MMU)
-- 
2.29.2

[PATCH 09/17] ptrace: Move setting/clearing ptrace_message into ptrace_stop

[Eric W. Biederman]

Today ptrace_message is easy to overlook as it not a core part of
ptrace_stop.  It has been overlooked so much that there are places
that set ptrace_message and don't clear it, and places that never set
it.  So if you get an unlucky sequence of events the ptracer may be
able to read a ptrace_message that does not apply to the current
ptrace stop.

Move setting of ptrace_message into ptrace_stop so that it always gets
set before the stop, and always gets cleared after the stop.  This
prevents non-sense from being reported to userspace and makes
ptrace_message more visible in the ptrace API so that kernel
developers can see it.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 include/linux/ptrace.h    |  5 ++---
 include/linux/tracehook.h |  6 ++----
 kernel/signal.c           | 19 +++++++++++--------
 3 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
index 8aee2945ff08..06f27736c6f8 100644
--- a/include/linux/ptrace.h
+++ b/include/linux/ptrace.h
@@ -60,7 +60,7 @@ extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned
 extern void ptrace_disable(struct task_struct *);
 extern int ptrace_request(struct task_struct *child, long request,
 			  unsigned long addr, unsigned long data);
-extern void ptrace_notify(int exit_code);
+extern void ptrace_notify(int exit_code, unsigned long message);
 extern void __ptrace_link(struct task_struct *child,
 			  struct task_struct *new_parent,
 			  const struct cred *ptracer_cred);
@@ -155,8 +155,7 @@ static inline bool ptrace_event_enabled(struct task_struct *task, int event)
 static inline void ptrace_event(int event, unsigned long message)
 {
 	if (unlikely(ptrace_event_enabled(current, event))) {
-		current->ptrace_message = message;
-		ptrace_notify((event << 8) | SIGTRAP);
+		ptrace_notify((event << 8) | SIGTRAP, message);
 	} else if (event == PTRACE_EVENT_EXEC) {
 		/* legacy EXEC report via SIGTRAP */
 		if ((current->ptrace & (PT_PTRACED|PT_SEIZED)) == PT_PTRACED)
diff --git a/include/linux/tracehook.h b/include/linux/tracehook.h
index 88c007ab5ebc..5e60af8a11fc 100644
--- a/include/linux/tracehook.h
+++ b/include/linux/tracehook.h
@@ -61,8 +61,7 @@ static inline int ptrace_report_syscall(unsigned long message)
 	if (!(ptrace & PT_PTRACED))
 		return 0;
 
-	current->ptrace_message = message;
-	ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0));
+	ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0), message);
 
 	/*
 	 * this isn't the same as continuing with a signal, but it will do
@@ -74,7 +73,6 @@ static inline int ptrace_report_syscall(unsigned long message)
 		current->exit_code = 0;
 	}
 
-	current->ptrace_message = 0;
 	return fatal_signal_pending(current);
 }
 
@@ -143,7 +141,7 @@ static inline void tracehook_report_syscall_exit(struct pt_regs *regs, int step)
 static inline void tracehook_signal_handler(int stepping)
 {
 	if (stepping)
-		ptrace_notify(SIGTRAP);
+		ptrace_notify(SIGTRAP, 0);
 }
 
 /**
diff --git a/kernel/signal.c b/kernel/signal.c
index 802acca0207b..75bb062d8534 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2197,7 +2197,8 @@ static void do_notify_parent_cldstop(struct task_struct *tsk,
  * If we actually decide not to stop at all because the tracer
  * is gone, we keep current->exit_code unless clear_code.
  */
-static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
+static void ptrace_stop(int exit_code, int why, int clear_code,
+			unsigned long message, kernel_siginfo_t *info)
 	__releases(&current->sighand->siglock)
 	__acquires(&current->sighand->siglock)
 {
@@ -2243,6 +2244,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t
 	 */
 	smp_wmb();
 
+	current->ptrace_message = message;
 	current->last_siginfo = info;
 	current->exit_code = exit_code;
 
@@ -2321,6 +2323,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t
 	 */
 	spin_lock_irq(&current->sighand->siglock);
 	current->last_siginfo = NULL;
+	current->ptrace_message = 0;
 
 	/* LISTENING can be set only during STOP traps, clear it */
 	current->jobctl &= ~JOBCTL_LISTENING;
@@ -2333,7 +2336,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t
 	recalc_sigpending_tsk(current);
 }
 
-static void ptrace_do_notify(int signr, int exit_code, int why)
+static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
 {
 	kernel_siginfo_t info;
 
@@ -2344,17 +2347,17 @@ static void ptrace_do_notify(int signr, int exit_code, int why)
 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
 
 	/* Let the debugger run.  */
-	ptrace_stop(exit_code, why, 1, &info);
+	ptrace_stop(exit_code, why, 1, message, &info);
 }
 
-void ptrace_notify(int exit_code)
+void ptrace_notify(int exit_code, unsigned long message)
 {
 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
 	if (unlikely(current->task_works))
 		task_work_run();
 
 	spin_lock_irq(&current->sighand->siglock);
-	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
+	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
 	spin_unlock_irq(&current->sighand->siglock);
 }
 
@@ -2510,10 +2513,10 @@ static void do_jobctl_trap(void)
 			signr = SIGTRAP;
 		WARN_ON_ONCE(!signr);
 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
-				 CLD_STOPPED);
+				 CLD_STOPPED, 0);
 	} else {
 		WARN_ON_ONCE(!signr);
-		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
+		ptrace_stop(signr, CLD_STOPPED, 0, 0, NULL);
 		current->exit_code = 0;
 	}
 }
@@ -2567,7 +2570,7 @@ static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
 	 * comment in dequeue_signal().
 	 */
 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
-	ptrace_stop(signr, CLD_TRAPPED, 0, info);
+	ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
 
 	/* We're back.  Did the debugger cancel the sig?  */
 	signr = current->exit_code;
-- 
2.29.2

[PATCH 10/17] ptrace: Return the signal to continue with from ptrace_stop

[Eric W. Biederman]

The signal a task should continue with after a ptrace stop is
inconsistently read, cleared, and sent.  Solve this by reading and
clearing the signal to be sent in ptrace_stop.

In an ideal world everything except ptrace_signal would share a common
implementation of continuing with the signal, so ptracers could count
on the signal they ask to continue with actually being delivered.  For
now retain bug compatibility and just return with the signal number
the ptracer requested the code continue with.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 include/linux/ptrace.h    |  2 +-
 include/linux/tracehook.h | 10 +++++-----
 kernel/signal.c           | 31 ++++++++++++++++++-------------
 3 files changed, 24 insertions(+), 19 deletions(-)

diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
index 06f27736c6f8..323c9950e705 100644
--- a/include/linux/ptrace.h
+++ b/include/linux/ptrace.h
@@ -60,7 +60,7 @@ extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned
 extern void ptrace_disable(struct task_struct *);
 extern int ptrace_request(struct task_struct *child, long request,
 			  unsigned long addr, unsigned long data);
-extern void ptrace_notify(int exit_code, unsigned long message);
+extern int ptrace_notify(int exit_code, unsigned long message);
 extern void __ptrace_link(struct task_struct *child,
 			  struct task_struct *new_parent,
 			  const struct cred *ptracer_cred);
diff --git a/include/linux/tracehook.h b/include/linux/tracehook.h
index 5e60af8a11fc..2fd0bfe866c0 100644
--- a/include/linux/tracehook.h
+++ b/include/linux/tracehook.h
@@ -57,21 +57,21 @@ struct linux_binprm;
 static inline int ptrace_report_syscall(unsigned long message)
 {
 	int ptrace = current->ptrace;
+	int signr;
 
 	if (!(ptrace & PT_PTRACED))
 		return 0;
 
-	ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0), message);
+	signr = ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0),
+			      message);
 
 	/*
 	 * this isn't the same as continuing with a signal, but it will do
 	 * for normal use.  strace only continues with a signal if the
 	 * stopping signal is not SIGTRAP.  -brl
 	 */
-	if (current->exit_code) {
-		send_sig(current->exit_code, current, 1);
-		current->exit_code = 0;
-	}
+	if (signr)
+		send_sig(signr, current, 1);
 
 	return fatal_signal_pending(current);
 }
diff --git a/kernel/signal.c b/kernel/signal.c
index 75bb062d8534..9903ff12e581 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2194,15 +2194,17 @@ static void do_notify_parent_cldstop(struct task_struct *tsk,
  * That makes it a way to test a stopped process for
  * being ptrace-stopped vs being job-control-stopped.
  *
- * If we actually decide not to stop at all because the tracer
- * is gone, we keep current->exit_code unless clear_code.
+ * Returns the signal the ptracer requested the code resume
+ * with.  If the code did not stop because the tracer is gone,
+ * the stop signal remains unchanged unless clear_code.
  */
-static void ptrace_stop(int exit_code, int why, int clear_code,
+static int ptrace_stop(int exit_code, int why, int clear_code,
 			unsigned long message, kernel_siginfo_t *info)
 	__releases(&current->sighand->siglock)
 	__acquires(&current->sighand->siglock)
 {
 	bool gstop_done = false;
+	bool read_code = true;
 
 	if (arch_ptrace_stop_needed()) {
 		/*
@@ -2311,8 +2313,9 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
 
 		/* tasklist protects us from ptrace_freeze_traced() */
 		__set_current_state(TASK_RUNNING);
+		read_code = false;
 		if (clear_code)
-			current->exit_code = 0;
+			exit_code = 0;
 		read_unlock(&tasklist_lock);
 	}
 
@@ -2322,8 +2325,10 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
 	 * any signal-sending on another CPU that wants to examine it.
 	 */
 	spin_lock_irq(&current->sighand->siglock);
+	if (read_code) exit_code = current->exit_code;
 	current->last_siginfo = NULL;
 	current->ptrace_message = 0;
+	current->exit_code = 0;
 
 	/* LISTENING can be set only during STOP traps, clear it */
 	current->jobctl &= ~JOBCTL_LISTENING;
@@ -2334,9 +2339,10 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
 	 * This sets TIF_SIGPENDING, but never clears it.
 	 */
 	recalc_sigpending_tsk(current);
+	return exit_code;
 }
 
-static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
+static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
 {
 	kernel_siginfo_t info;
 
@@ -2347,18 +2353,21 @@ static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long me
 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
 
 	/* Let the debugger run.  */
-	ptrace_stop(exit_code, why, 1, message, &info);
+	return ptrace_stop(exit_code, why, 1, message, &info);
 }
 
-void ptrace_notify(int exit_code, unsigned long message)
+int ptrace_notify(int exit_code, unsigned long message)
 {
+	int signr;
+
 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
 	if (unlikely(current->task_works))
 		task_work_run();
 
 	spin_lock_irq(&current->sighand->siglock);
-	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
+	signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
 	spin_unlock_irq(&current->sighand->siglock);
+	return signr;
 }
 
 /**
@@ -2517,7 +2526,6 @@ static void do_jobctl_trap(void)
 	} else {
 		WARN_ON_ONCE(!signr);
 		ptrace_stop(signr, CLD_STOPPED, 0, 0, NULL);
-		current->exit_code = 0;
 	}
 }
 
@@ -2570,15 +2578,12 @@ static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
 	 * comment in dequeue_signal().
 	 */
 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
-	ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
+	signr = ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
 
 	/* We're back.  Did the debugger cancel the sig?  */
-	signr = current->exit_code;
 	if (signr == 0)
 		return signr;
 
-	current->exit_code = 0;
-
 	/*
 	 * Update the siginfo structure if the signal has
 	 * changed.  If the debugger wanted something
-- 
2.29.2

[PATCH 11/17] ptrace: Separate task->ptrace_code out from task->exit_code

[Eric W. Biederman]

A process can be marked for death by setting SIGNAL_GROUP_EXIT and
group_exit_code, long before do_exit is called.  Unfortunately because
of PTRACE_EVENT_EXIT residing in do_exit this same tactic can not be
used for task death.

Correct this by adding a new task field task->ptrace_code that holds
the code for ptrace stops.  This allows task->exit_code to be set to
the exit code long before the PTRACE_EVENT_EXIT ptrace stop.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/proc/array.c       |  3 +++
 include/linux/sched.h |  1 +
 kernel/exit.c         |  2 +-
 kernel/ptrace.c       | 12 ++++++------
 kernel/signal.c       | 18 +++++++++---------
 5 files changed, 20 insertions(+), 16 deletions(-)

diff --git a/fs/proc/array.c b/fs/proc/array.c
index 43a7abde9e42..3042015c11ad 100644
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -519,6 +519,9 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 		cgtime = sig->cgtime;
 		rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
 
+		if (task_is_traced(task) && !(task->jobctl & JOBCTL_LISTENING))
+			exit_code = task->ptrace_code;
+
 		/* add up live thread stats at the group level */
 		if (whole) {
 			struct task_struct *t = task;
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 52f2fdffa3ab..c3d732bf7833 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1174,6 +1174,7 @@ struct task_struct {
 	/* Ptrace state: */
 	unsigned long			ptrace_message;
 	kernel_siginfo_t		*last_siginfo;
+	int				ptrace_code;
 
 	struct task_io_accounting	ioac;
 #ifdef CONFIG_PSI
diff --git a/kernel/exit.c b/kernel/exit.c
index 7121db37c411..aedefe5eb0eb 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -1134,7 +1134,7 @@ static int *task_stopped_code(struct task_struct *p, bool ptrace)
 {
 	if (ptrace) {
 		if (task_is_traced(p) && !(p->jobctl & JOBCTL_LISTENING))
-			return &p->exit_code;
+			return &p->ptrace_code;
 	} else {
 		if (p->signal->flags & SIGNAL_STOP_STOPPED)
 			return &p->signal->group_exit_code;
diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index eea265082e97..8bbd73ab9a34 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -172,7 +172,7 @@ void __ptrace_unlink(struct task_struct *child)
 
 static bool looks_like_a_spurious_pid(struct task_struct *task)
 {
-	if (task->exit_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP))
+	if (task->ptrace_code != ((PTRACE_EVENT_EXEC << 8) | SIGTRAP))
 		return false;
 
 	if (task_pid_vnr(task) == task->ptrace_message)
@@ -573,7 +573,7 @@ static int ptrace_detach(struct task_struct *child, unsigned int data)
 	 * tasklist_lock avoids the race with wait_task_stopped(), see
 	 * the comment in ptrace_resume().
 	 */
-	child->exit_code = data;
+	child->ptrace_code = data;
 	__ptrace_detach(current, child);
 	write_unlock_irq(&tasklist_lock);
 
@@ -863,11 +863,11 @@ static int ptrace_resume(struct task_struct *child, long request,
 	}
 
 	/*
-	 * Change ->exit_code and ->state under siglock to avoid the race
-	 * with wait_task_stopped() in between; a non-zero ->exit_code will
+	 * Change ->ptrace_code and ->state under siglock to avoid the race
+	 * with wait_task_stopped() in between; a non-zero ->ptrace_code will
 	 * wrongly look like another report from tracee.
 	 *
-	 * Note that we need siglock even if ->exit_code == data and/or this
+	 * Note that we need siglock even if ->ptrace_code == data and/or this
 	 * status was not reported yet, the new status must not be cleared by
 	 * wait_task_stopped() after resume.
 	 *
@@ -878,7 +878,7 @@ static int ptrace_resume(struct task_struct *child, long request,
 	need_siglock = data && !thread_group_empty(current);
 	if (need_siglock)
 		spin_lock_irq(&child->sighand->siglock);
-	child->exit_code = data;
+	child->ptrace_code = data;
 	wake_up_state(child, __TASK_TRACED);
 	if (need_siglock)
 		spin_unlock_irq(&child->sighand->siglock);
diff --git a/kernel/signal.c b/kernel/signal.c
index 9903ff12e581..fd3c404de8b6 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2168,7 +2168,7 @@ static void do_notify_parent_cldstop(struct task_struct *tsk,
  		info.si_status = tsk->signal->group_exit_code & 0x7f;
  		break;
  	case CLD_TRAPPED:
- 		info.si_status = tsk->exit_code & 0x7f;
+		info.si_status = tsk->ptrace_code & 0x7f;
  		break;
  	default:
  		BUG();
@@ -2198,7 +2198,7 @@ static void do_notify_parent_cldstop(struct task_struct *tsk,
  * with.  If the code did not stop because the tracer is gone,
  * the stop signal remains unchanged unless clear_code.
  */
-static int ptrace_stop(int exit_code, int why, int clear_code,
+static int ptrace_stop(int code, int why, int clear_code,
 			unsigned long message, kernel_siginfo_t *info)
 	__releases(&current->sighand->siglock)
 	__acquires(&current->sighand->siglock)
@@ -2248,7 +2248,7 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
 
 	current->ptrace_message = message;
 	current->last_siginfo = info;
-	current->exit_code = exit_code;
+	current->ptrace_code = code;
 
 	/*
 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
@@ -2315,7 +2315,7 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
 		__set_current_state(TASK_RUNNING);
 		read_code = false;
 		if (clear_code)
-			exit_code = 0;
+			code = 0;
 		read_unlock(&tasklist_lock);
 	}
 
@@ -2325,10 +2325,10 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
 	 * any signal-sending on another CPU that wants to examine it.
 	 */
 	spin_lock_irq(&current->sighand->siglock);
-	if (read_code) exit_code = current->exit_code;
+	if (read_code) code = current->ptrace_code;
 	current->last_siginfo = NULL;
 	current->ptrace_message = 0;
-	current->exit_code = 0;
+	current->ptrace_code = 0;
 
 	/* LISTENING can be set only during STOP traps, clear it */
 	current->jobctl &= ~JOBCTL_LISTENING;
@@ -2339,7 +2339,7 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
 	 * This sets TIF_SIGPENDING, but never clears it.
 	 */
 	recalc_sigpending_tsk(current);
-	return exit_code;
+	return code;
 }
 
 static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
@@ -2501,11 +2501,11 @@ static bool do_signal_stop(int signr)
  *
  * When PT_SEIZED, it's used for both group stop and explicit
  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
- * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
+ * accompanying siginfo.  If stopped, lower eight bits of ptrace_code contain
  * the stop signal; otherwise, %SIGTRAP.
  *
  * When !PT_SEIZED, it's used only for group stop trap with stop signal
- * number as exit_code and no siginfo.
+ * number as ptrace_code and no siginfo.
  *
  * CONTEXT:
  * Must be called with @current->sighand->siglock held, which may be
-- 
2.29.2

[PATCH 12/17] signal: Compute the process exit_code in get_signal

[Eric W. Biederman]

In prepartion for moving the work of sys_exit and sys_group_exit into
get_signal compute exit_code in get_signal, make PF_SIGNALED depend on
the exit_code and pass the exit_code to do_group_exit.

Anytime there is a group exit the exit_code may differ from the signal
number.

To match the historical precedent as best I can make the exit_code 0
during exec. (The exit_code field would not have been set but probably
would have been left at a value of 0).

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 kernel/signal.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/kernel/signal.c b/kernel/signal.c
index fd3c404de8b6..2a24cca00ca1 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2707,6 +2707,7 @@ bool get_signal(struct ksignal *ksig)
 	for (;;) {
 		struct k_sigaction *ka;
 		enum pid_type type;
+		int exit_code;
 
 		/* Has this task already been marked for death? */
 		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
@@ -2716,6 +2717,10 @@ bool get_signal(struct ksignal *ksig)
 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
 				&sighand->action[SIGKILL - 1]);
 			recalc_sigpending();
+			if (signal->flags & SIGNAL_GROUP_EXIT)
+				exit_code = signal->group_exit_code;
+			else
+				exit_code = 0;
 			goto fatal;
 		}
 
@@ -2837,15 +2842,17 @@ bool get_signal(struct ksignal *ksig)
 			continue;
 		}
 
+		/*
+		 * Anything else is fatal, maybe with a core dump.
+		 */
+		exit_code = signr;
 	fatal:
 		spin_unlock_irq(&sighand->siglock);
 		if (unlikely(cgroup_task_frozen(current)))
 			cgroup_leave_frozen(true);
 
-		/*
-		 * Anything else is fatal, maybe with a core dump.
-		 */
-		current->flags |= PF_SIGNALED;
+		if (exit_code & 0x7f)
+			current->flags |= PF_SIGNALED;
 
 		if (sig_kernel_coredump(signr)) {
 			if (print_fatal_signals)
@@ -2873,7 +2880,7 @@ bool get_signal(struct ksignal *ksig)
 		/*
 		 * Death signals, no core dump.
 		 */
-		do_group_exit(ksig->info.si_signo);
+		do_group_exit(exit_code);
 		/* NOTREACHED */
 	}
 	spin_unlock_irq(&sighand->siglock);
-- 
2.29.2

[PATCH 13/17] signal: Make individual tasks exiting a first class concept

[Eric W. Biederman]

Add a helper schedule_task_exit_locked that is equivalent to
asynchronously calling exit(2) except for not having an exit code.

This is a generalization of what happens in de_thread, zap_process,
prepare_signal, complete_signal, and zap_other_threads when individual
tasks are asked to shutdown.

The various code paths optimize away the setting sigaddset and
signal_wake_up based on different conditions.  Neither sigaddset nor
signal_wake_up are needed if the task has already started running
do_exit.  So skip the work if PF_POSTCOREDUMP is set.  Which is the
earliest any of the original hand rolled implementations used.

Update get_signal to detect either signal group exit or a single task
exit by testing for __fatal_signal_pending.  This works because the
all of the tasks in group exits are killed with
schedule_task_exit_locked.

For clarity the code in get_signal has been updated to call do_exit
instead of do_group_exit when a single task is exiting.

While this schedule_task_exit_locked is a generalization of what
happens in prepare_signal I do not change prepare_signal to use
schedule_task_exit_locked to deliver SIGKILL to a coredumping process.
This keeps all of the specialness delivering a signal to a coredumping
process limited to prepare_signal and the coredump code itself.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c                |  7 ++-----
 include/linux/sched/signal.h |  2 ++
 kernel/signal.c              | 36 +++++++++++++++++++++---------------
 3 files changed, 25 insertions(+), 20 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index 09302a6a0d80..9559e29daada 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -358,12 +358,9 @@ static int zap_process(struct task_struct *start, int exit_code)
 	start->signal->group_stop_count = 0;
 
 	for_each_thread(start, t) {
-		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
-		if (t != current && !(t->flags & PF_POSTCOREDUMP)) {
-			sigaddset(&t->pending.signal, SIGKILL);
-			signal_wake_up(t, 1);
+		schedule_task_exit_locked(t);
+		if (t != current && !(t->flags & PF_POSTCOREDUMP))
 			nr++;
-		}
 	}
 
 	return nr;
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index b6ecb9fc4cd2..7c62b7c29cc0 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -427,6 +427,8 @@ static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
 	signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
 }
 
+void schedule_task_exit_locked(struct task_struct *task);
+
 void task_join_group_stop(struct task_struct *task);
 
 #ifdef TIF_RESTORE_SIGMASK
diff --git a/kernel/signal.c b/kernel/signal.c
index 2a24cca00ca1..cbfb9020368e 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1056,9 +1056,7 @@ static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
 			signal->group_stop_count = 0;
 			t = p;
 			do {
-				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
-				sigaddset(&t->pending.signal, SIGKILL);
-				signal_wake_up(t, 1);
+				schedule_task_exit_locked(t);
 			} while_each_thread(p, t);
 			return;
 		}
@@ -1363,6 +1361,16 @@ int force_sig_info(struct kernel_siginfo *info)
 	return force_sig_info_to_task(info, current, HANDLER_CURRENT);
 }
 
+void schedule_task_exit_locked(struct task_struct *task)
+{
+	task_clear_jobctl_pending(task, JOBCTL_PENDING_MASK);
+	/* Only bother with threads that might be alive */
+	if (!(task->flags & PF_POSTCOREDUMP)) {
+		sigaddset(&task->pending.signal, SIGKILL);
+		signal_wake_up(task, 1);
+	}
+}
+
 /*
  * Nuke all other threads in the group.
  */
@@ -1374,16 +1382,9 @@ int zap_other_threads(struct task_struct *p)
 	p->signal->group_stop_count = 0;
 
 	while_each_thread(p, t) {
-		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
 		count++;
-
-		/* Don't bother with already dead threads */
-		if (t->exit_state)
-			continue;
-		sigaddset(&t->pending.signal, SIGKILL);
-		signal_wake_up(t, 1);
+		schedule_task_exit_locked(t);
 	}
-
 	return count;
 }
 
@@ -2706,12 +2707,12 @@ bool get_signal(struct ksignal *ksig)
 
 	for (;;) {
 		struct k_sigaction *ka;
+		bool group_exit = true;
 		enum pid_type type;
 		int exit_code;
 
 		/* Has this task already been marked for death? */
-		if ((signal->flags & SIGNAL_GROUP_EXIT) ||
-		     signal->group_exec_task) {
+		if (__fatal_signal_pending(current)) {
 			ksig->info.si_signo = signr = SIGKILL;
 			sigdelset(&current->pending.signal, SIGKILL);
 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
@@ -2719,8 +2720,10 @@ bool get_signal(struct ksignal *ksig)
 			recalc_sigpending();
 			if (signal->flags & SIGNAL_GROUP_EXIT)
 				exit_code = signal->group_exit_code;
-			else
+			else {
 				exit_code = 0;
+				group_exit = false;
+			}
 			goto fatal;
 		}
 
@@ -2880,7 +2883,10 @@ bool get_signal(struct ksignal *ksig)
 		/*
 		 * Death signals, no core dump.
 		 */
-		do_group_exit(exit_code);
+		if (group_exit)
+			do_group_exit(exit_code);
+		else
+			do_exit(exit_code);
 		/* NOTREACHED */
 	}
 	spin_unlock_irq(&sighand->siglock);
-- 
2.29.2

[PATCH 14/17] signal: Remove zap_other_threads

[Eric W. Biederman]

The two callers of zap_other_threads want different things. The
function do_group_exit wants to set the exit code and it does not care
about the number of threads exiting.  In de_thread the current thread
is not exiting so there is not really an exit code.

Since schedule_task_exit_locked factors out the tricky bits stop
sharing the loop in zap_other_threads between de_thread and
do_group_exit.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/exec.c                    | 12 +++++++++---
 include/linux/sched/signal.h |  1 -
 kernel/exit.c                |  9 ++++++++-
 kernel/signal.c              | 17 -----------------
 4 files changed, 17 insertions(+), 22 deletions(-)

diff --git a/fs/exec.c b/fs/exec.c
index 82db656ca709..b9f646fddc51 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1037,6 +1037,7 @@ static int de_thread(struct task_struct *tsk)
 	struct signal_struct *sig = tsk->signal;
 	struct sighand_struct *oldsighand = tsk->sighand;
 	spinlock_t *lock = &oldsighand->siglock;
+	struct task_struct *t;
 
 	if (thread_group_empty(tsk))
 		goto no_thread_group;
@@ -1055,9 +1056,14 @@ static int de_thread(struct task_struct *tsk)
 	}
 
 	sig->group_exec_task = tsk;
-	sig->notify_count = zap_other_threads(tsk);
-	if (!thread_group_leader(tsk))
-		sig->notify_count--;
+	sig->group_stop_count = 0;
+	sig->notify_count = 0;
+	__for_each_thread(sig, t) {
+		if (t == tsk)
+			continue;
+		sig->notify_count++;
+		schedule_task_exit_locked(t);
+	}
 
 	while (sig->notify_count) {
 		__set_current_state(TASK_KILLABLE);
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index 7c62b7c29cc0..eed54f9ea2fc 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -343,7 +343,6 @@ extern void force_sig(int);
 extern void force_fatal_sig(int);
 extern void force_exit_sig(int);
 extern int send_sig(int, struct task_struct *, int);
-extern int zap_other_threads(struct task_struct *p);
 extern struct sigqueue *sigqueue_alloc(void);
 extern void sigqueue_free(struct sigqueue *);
 extern int send_sigqueue(struct sigqueue *, struct pid *, enum pid_type);
diff --git a/kernel/exit.c b/kernel/exit.c
index aedefe5eb0eb..27bc0ccfea78 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -918,9 +918,16 @@ do_group_exit(int exit_code)
 		else if (sig->group_exec_task)
 			exit_code = 0;
 		else {
+			struct task_struct *t;
+
 			sig->group_exit_code = exit_code;
 			sig->flags = SIGNAL_GROUP_EXIT;
-			zap_other_threads(current);
+			sig->group_stop_count = 0;
+			__for_each_thread(sig, t) {
+				if (t == current)
+					continue;
+				schedule_task_exit_locked(t);
+			}
 		}
 		spin_unlock_irq(&sighand->siglock);
 	}
diff --git a/kernel/signal.c b/kernel/signal.c
index cbfb9020368e..b0201e05be40 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1371,23 +1371,6 @@ void schedule_task_exit_locked(struct task_struct *task)
 	}
 }
 
-/*
- * Nuke all other threads in the group.
- */
-int zap_other_threads(struct task_struct *p)
-{
-	struct task_struct *t = p;
-	int count = 0;
-
-	p->signal->group_stop_count = 0;
-
-	while_each_thread(p, t) {
-		count++;
-		schedule_task_exit_locked(t);
-	}
-	return count;
-}
-
 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
 					   unsigned long *flags)
 {
-- 
2.29.2

[PATCH 15/17] signal: Add JOBCTL_WILL_EXIT to mark exiting tasks

[Eric W. Biederman]

Mark tasks that need to exit with JOBCTL_WILL_EXIT instead of reusing
the per thread SIGKILL.

This removes the double meaning of the per thread SIGKILL and makes it
possible to detect when a task has already been scheduled for exiting
and to skip unnecessary work if the task is already scheduled to exit.

A jobctl flag was choosen for this purpose because jobctl changes are
protected by siglock, and updates are already careful not to change or
clear other bits in jobctl.  Protection by a lock when changing the
value is necessary as JOBCTL_WILL_EXIT will not be limited to being
set by the current task.  That task->jobctl is protected by siglock is
convenient as siglock is already held everywhere I want to set or reset
JOBCTL_WILL_EXIT.

Teach wants_signal and retarget_shared_pending to use
JOBCTL_TASK_EXITING to detect threads that have an exit pending and so
will not be processing any more signals.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c                |  6 ++++--
 include/linux/sched/jobctl.h |  2 ++
 include/linux/sched/signal.h |  2 +-
 kernel/exit.c                |  4 ++--
 kernel/signal.c              | 19 +++++++++----------
 5 files changed, 18 insertions(+), 15 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index 9559e29daada..4e82ee51633d 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -352,7 +352,6 @@ static int zap_process(struct task_struct *start, int exit_code)
 	struct task_struct *t;
 	int nr = 0;
 
-	/* Allow SIGKILL, see prepare_signal() */
 	start->signal->flags = SIGNAL_GROUP_EXIT;
 	start->signal->group_exit_code = exit_code;
 	start->signal->group_stop_count = 0;
@@ -376,9 +375,11 @@ static int zap_threads(struct task_struct *tsk,
 	if (!(signal->flags & SIGNAL_GROUP_EXIT) && !signal->group_exec_task) {
 		signal->core_state = core_state;
 		nr = zap_process(tsk, exit_code);
+		atomic_set(&core_state->nr_threads, nr);
+		 /* Allow SIGKILL, see prepare_signal() */
 		clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
 		tsk->flags |= PF_DUMPCORE;
-		atomic_set(&core_state->nr_threads, nr);
+		tsk->jobctl &= ~JOBCTL_WILL_EXIT;
 	}
 	spin_unlock_irq(&tsk->sighand->siglock);
 	return nr;
@@ -425,6 +426,7 @@ static void coredump_finish(bool core_dumped)
 		current->signal->group_exit_code |= 0x80;
 	next = current->signal->core_state->dumper.next;
 	current->signal->core_state = NULL;
+	current->jobctl |= JOBCTL_WILL_EXIT;
 	spin_unlock_irq(&current->sighand->siglock);
 
 	while ((curr = next) != NULL) {
diff --git a/include/linux/sched/jobctl.h b/include/linux/sched/jobctl.h
index fa067de9f1a9..9887d737ccfb 100644
--- a/include/linux/sched/jobctl.h
+++ b/include/linux/sched/jobctl.h
@@ -19,6 +19,7 @@ struct task_struct;
 #define JOBCTL_TRAPPING_BIT	21	/* switching to TRACED */
 #define JOBCTL_LISTENING_BIT	22	/* ptracer is listening for events */
 #define JOBCTL_TRAP_FREEZE_BIT	23	/* trap for cgroup freezer */
+#define JOBCTL_WILL_EXIT_BIT	31	/* task will exit */
 
 #define JOBCTL_STOP_DEQUEUED	(1UL << JOBCTL_STOP_DEQUEUED_BIT)
 #define JOBCTL_STOP_PENDING	(1UL << JOBCTL_STOP_PENDING_BIT)
@@ -28,6 +29,7 @@ struct task_struct;
 #define JOBCTL_TRAPPING		(1UL << JOBCTL_TRAPPING_BIT)
 #define JOBCTL_LISTENING	(1UL << JOBCTL_LISTENING_BIT)
 #define JOBCTL_TRAP_FREEZE	(1UL << JOBCTL_TRAP_FREEZE_BIT)
+#define JOBCTL_WILL_EXIT	(1UL << JOBCTL_WILL_EXIT_BIT)
 
 #define JOBCTL_TRAP_MASK	(JOBCTL_TRAP_STOP | JOBCTL_TRAP_NOTIFY)
 #define JOBCTL_PENDING_MASK	(JOBCTL_STOP_PENDING | JOBCTL_TRAP_MASK)
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index eed54f9ea2fc..989bb665f107 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -373,7 +373,7 @@ static inline int signal_pending(struct task_struct *p)
 
 static inline int __fatal_signal_pending(struct task_struct *p)
 {
-	return unlikely(sigismember(&p->pending.signal, SIGKILL));
+	return unlikely(p->jobctl & JOBCTL_WILL_EXIT);
 }
 
 static inline int fatal_signal_pending(struct task_struct *p)
diff --git a/kernel/exit.c b/kernel/exit.c
index 27bc0ccfea78..7a7a0cbac28e 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -906,7 +906,7 @@ do_group_exit(int exit_code)
 
 	if (sig->flags & SIGNAL_GROUP_EXIT)
 		exit_code = sig->group_exit_code;
-	else if (sig->group_exec_task)
+	else if (current->jobctl & JOBCTL_WILL_EXIT)
 		exit_code = 0;
 	else if (!thread_group_empty(current)) {
 		struct sighand_struct *const sighand = current->sighand;
@@ -915,7 +915,7 @@ do_group_exit(int exit_code)
 		if (sig->flags & SIGNAL_GROUP_EXIT)
 			/* Another thread got here before we took the lock.  */
 			exit_code = sig->group_exit_code;
-		else if (sig->group_exec_task)
+		else if (current->jobctl & JOBCTL_WILL_EXIT)
 			exit_code = 0;
 		else {
 			struct task_struct *t;
diff --git a/kernel/signal.c b/kernel/signal.c
index b0201e05be40..6179e34ce666 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -153,7 +153,8 @@ static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
 
 static bool recalc_sigpending_tsk(struct task_struct *t)
 {
-	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
+	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE |
+			  JOBCTL_WILL_EXIT)) ||
 	    PENDING(&t->pending, &t->blocked) ||
 	    PENDING(&t->signal->shared_pending, &t->blocked) ||
 	    cgroup_task_frozen(t)) {
@@ -911,7 +912,7 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
 		if (core_state) {
 			if (sig == SIGKILL) {
 				struct task_struct *dumper = core_state->dumper.task;
-				sigaddset(&dumper->pending.signal, SIGKILL);
+				dumper->jobctl |= JOBCTL_WILL_EXIT;
 				signal_wake_up(dumper, 1);
 			}
 		}
@@ -985,7 +986,7 @@ static inline bool wants_signal(int sig, struct task_struct *p)
 	if (sigismember(&p->blocked, sig))
 		return false;
 
-	if (p->flags & PF_EXITING)
+	if (p->jobctl & JOBCTL_WILL_EXIT)
 		return false;
 
 	if (sig == SIGKILL)
@@ -1363,10 +1364,9 @@ int force_sig_info(struct kernel_siginfo *info)
 
 void schedule_task_exit_locked(struct task_struct *task)
 {
-	task_clear_jobctl_pending(task, JOBCTL_PENDING_MASK);
-	/* Only bother with threads that might be alive */
-	if (!(task->flags & PF_POSTCOREDUMP)) {
-		sigaddset(&task->pending.signal, SIGKILL);
+	if (!(task->jobctl & JOBCTL_WILL_EXIT)) {
+		task_clear_jobctl_pending(task, JOBCTL_PENDING_MASK);
+		task->jobctl |= JOBCTL_WILL_EXIT;
 		signal_wake_up(task, 1);
 	}
 }
@@ -2695,9 +2695,8 @@ bool get_signal(struct ksignal *ksig)
 		int exit_code;
 
 		/* Has this task already been marked for death? */
-		if (__fatal_signal_pending(current)) {
+		if (current->jobctl & JOBCTL_WILL_EXIT) {
 			ksig->info.si_signo = signr = SIGKILL;
-			sigdelset(&current->pending.signal, SIGKILL);
 			trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
 				&sighand->action[SIGKILL - 1]);
 			recalc_sigpending();
@@ -2935,7 +2934,7 @@ static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
 
 	t = tsk;
 	while_each_thread(tsk, t) {
-		if (t->flags & PF_EXITING)
+		if (t->jobctl & JOBCTL_WILL_EXIT)
 			continue;
 
 		if (!has_pending_signals(&retarget, &t->blocked))
-- 
2.29.2

[PATCH 16/17] signal: Record the exit_code when an exit is scheduled

[Eric W. Biederman]

With ptrace_stop no longer using task->exit_code it is safe
to set task->exit_code when an exit is scheduled.

Use the bit JOBCTL_WILL_EXIT to detect when a exit is first scheduled
and only set exit_code the first time.  Only use the code provided
to do_exit if the task has not yet been schedled to exit.

In get_signal and do_grup_exit when JOBCTL_WILL_EXIT is set read the
recored exit_code from current->exit_code, instead of assuming
exit_code will always be 0.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c                |  2 +-
 fs/exec.c                    |  2 +-
 include/linux/sched/signal.h |  2 +-
 kernel/exit.c                | 12 ++++++++----
 kernel/signal.c              |  7 ++++---
 5 files changed, 15 insertions(+), 10 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index 4e82ee51633d..c54b502bf648 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -357,7 +357,7 @@ static int zap_process(struct task_struct *start, int exit_code)
 	start->signal->group_stop_count = 0;
 
 	for_each_thread(start, t) {
-		schedule_task_exit_locked(t);
+		schedule_task_exit_locked(t, exit_code);
 		if (t != current && !(t->flags & PF_POSTCOREDUMP))
 			nr++;
 	}
diff --git a/fs/exec.c b/fs/exec.c
index b9f646fddc51..3203605e54cb 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1062,7 +1062,7 @@ static int de_thread(struct task_struct *tsk)
 		if (t == tsk)
 			continue;
 		sig->notify_count++;
-		schedule_task_exit_locked(t);
+		schedule_task_exit_locked(t, 0);
 	}
 
 	while (sig->notify_count) {
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index 989bb665f107..e8034ecaee84 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -426,7 +426,7 @@ static inline void ptrace_signal_wake_up(struct task_struct *t, bool resume)
 	signal_wake_up_state(t, resume ? __TASK_TRACED : 0);
 }
 
-void schedule_task_exit_locked(struct task_struct *task);
+void schedule_task_exit_locked(struct task_struct *task, int exit_code);
 
 void task_join_group_stop(struct task_struct *task);
 
diff --git a/kernel/exit.c b/kernel/exit.c
index 7a7a0cbac28e..e95500e2d27c 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -735,6 +735,11 @@ void __noreturn do_exit(long code)
 	struct task_struct *tsk = current;
 	int group_dead;
 
+	spin_lock_irq(&tsk->sighand->siglock);
+	schedule_task_exit_locked(tsk, code);
+	code = tsk->exit_code;
+	spin_unlock_irq(&tsk->sighand->siglock);
+
 	WARN_ON(blk_needs_flush_plug(tsk));
 
 	kcov_task_exit(tsk);
@@ -773,7 +778,6 @@ void __noreturn do_exit(long code)
 		tty_audit_exit();
 	audit_free(tsk);
 
-	tsk->exit_code = code;
 	taskstats_exit(tsk, group_dead);
 
 	exit_mm();
@@ -907,7 +911,7 @@ do_group_exit(int exit_code)
 	if (sig->flags & SIGNAL_GROUP_EXIT)
 		exit_code = sig->group_exit_code;
 	else if (current->jobctl & JOBCTL_WILL_EXIT)
-		exit_code = 0;
+		exit_code = current->exit_code;
 	else if (!thread_group_empty(current)) {
 		struct sighand_struct *const sighand = current->sighand;
 
@@ -916,7 +920,7 @@ do_group_exit(int exit_code)
 			/* Another thread got here before we took the lock.  */
 			exit_code = sig->group_exit_code;
 		else if (current->jobctl & JOBCTL_WILL_EXIT)
-			exit_code = 0;
+			exit_code = current->exit_code;
 		else {
 			struct task_struct *t;
 
@@ -926,7 +930,7 @@ do_group_exit(int exit_code)
 			__for_each_thread(sig, t) {
 				if (t == current)
 					continue;
-				schedule_task_exit_locked(t);
+				schedule_task_exit_locked(t, exit_code);
 			}
 		}
 		spin_unlock_irq(&sighand->siglock);
diff --git a/kernel/signal.c b/kernel/signal.c
index 6179e34ce666..e8fac8a3c935 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1057,7 +1057,7 @@ static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
 			signal->group_stop_count = 0;
 			t = p;
 			do {
-				schedule_task_exit_locked(t);
+				schedule_task_exit_locked(t, sig);
 			} while_each_thread(p, t);
 			return;
 		}
@@ -1362,11 +1362,12 @@ int force_sig_info(struct kernel_siginfo *info)
 	return force_sig_info_to_task(info, current, HANDLER_CURRENT);
 }
 
-void schedule_task_exit_locked(struct task_struct *task)
+void schedule_task_exit_locked(struct task_struct *task, int exit_code)
 {
 	if (!(task->jobctl & JOBCTL_WILL_EXIT)) {
 		task_clear_jobctl_pending(task, JOBCTL_PENDING_MASK);
 		task->jobctl |= JOBCTL_WILL_EXIT;
+		task->exit_code = exit_code;
 		signal_wake_up(task, 1);
 	}
 }
@@ -2703,7 +2704,7 @@ bool get_signal(struct ksignal *ksig)
 			if (signal->flags & SIGNAL_GROUP_EXIT)
 				exit_code = signal->group_exit_code;
 			else {
-				exit_code = 0;
+				exit_code = current->exit_code;
 				group_exit = false;
 			}
 			goto fatal;
-- 
2.29.2

[PATCH 17/17] signal: Always set SIGNAL_GROUP_EXIT on process exit

[Eric W. Biederman]

Track how many threads have not started exiting and when
the last thread starts exiting set SIGNAL_GROUP_EXIT.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 fs/coredump.c                |  4 ----
 include/linux/sched/signal.h |  1 +
 kernel/exit.c                |  8 +-------
 kernel/fork.c                |  2 ++
 kernel/signal.c              | 10 +++++++---
 5 files changed, 11 insertions(+), 14 deletions(-)

diff --git a/fs/coredump.c b/fs/coredump.c
index c54b502bf648..029d0f98aa90 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -352,10 +352,6 @@ static int zap_process(struct task_struct *start, int exit_code)
 	struct task_struct *t;
 	int nr = 0;
 
-	start->signal->flags = SIGNAL_GROUP_EXIT;
-	start->signal->group_exit_code = exit_code;
-	start->signal->group_stop_count = 0;
-
 	for_each_thread(start, t) {
 		schedule_task_exit_locked(t, exit_code);
 		if (t != current && !(t->flags & PF_POSTCOREDUMP))
diff --git a/include/linux/sched/signal.h b/include/linux/sched/signal.h
index e8034ecaee84..bd9435e934a1 100644
--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -94,6 +94,7 @@ struct signal_struct {
 	refcount_t		sigcnt;
 	atomic_t		live;
 	int			nr_threads;
+	int			quick_threads;
 	struct list_head	thread_head;
 
 	wait_queue_head_t	wait_chldexit;	/* for wait4() */
diff --git a/kernel/exit.c b/kernel/exit.c
index e95500e2d27c..be867a12de65 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -924,14 +924,8 @@ do_group_exit(int exit_code)
 		else {
 			struct task_struct *t;
 
-			sig->group_exit_code = exit_code;
-			sig->flags = SIGNAL_GROUP_EXIT;
-			sig->group_stop_count = 0;
-			__for_each_thread(sig, t) {
-				if (t == current)
-					continue;
+			__for_each_thread(sig, t)
 				schedule_task_exit_locked(t, exit_code);
-			}
 		}
 		spin_unlock_irq(&sighand->siglock);
 	}
diff --git a/kernel/fork.c b/kernel/fork.c
index 6f0293cb29c9..d973189a4014 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1644,6 +1644,7 @@ static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
 		return -ENOMEM;
 
 	sig->nr_threads = 1;
+	sig->quick_threads = 1;
 	atomic_set(&sig->live, 1);
 	refcount_set(&sig->sigcnt, 1);
 
@@ -2383,6 +2384,7 @@ static __latent_entropy struct task_struct *copy_process(
 			__this_cpu_inc(process_counts);
 		} else {
 			current->signal->nr_threads++;
+			current->signal->quick_threads++;
 			atomic_inc(&current->signal->live);
 			refcount_inc(&current->signal->sigcnt);
 			task_join_group_stop(p);
diff --git a/kernel/signal.c b/kernel/signal.c
index e8fac8a3c935..9bd835fcb1dc 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1052,9 +1052,6 @@ static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
 			 * running and doing things after a slower
 			 * thread has the fatal signal pending.
 			 */
-			signal->flags = SIGNAL_GROUP_EXIT;
-			signal->group_exit_code = sig;
-			signal->group_stop_count = 0;
 			t = p;
 			do {
 				schedule_task_exit_locked(t, sig);
@@ -1365,10 +1362,17 @@ int force_sig_info(struct kernel_siginfo *info)
 void schedule_task_exit_locked(struct task_struct *task, int exit_code)
 {
 	if (!(task->jobctl & JOBCTL_WILL_EXIT)) {
+		struct signal_struct *signal = task->signal;
 		task_clear_jobctl_pending(task, JOBCTL_PENDING_MASK);
 		task->jobctl |= JOBCTL_WILL_EXIT;
 		task->exit_code = exit_code;
 		signal_wake_up(task, 1);
+		signal->quick_threads--;
+		if (signal->quick_threads == 0) {
+			signal->flags = SIGNAL_GROUP_EXIT;
+			signal->group_exit_code = exit_code;
+			signal->group_stop_count = 0;
+		}
 	}
 }
 
-- 
2.29.2

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Dmitry Osipenko]

14.12.2021 01:53, Eric W. Biederman пишет:
> Simplify the code that allows SIGKILL during coredumps to terminate
> the coredump.  As far as I can tell I have avoided breaking it
> by dumb luck.
> 
> Historically with all of the other threads stopping in exit_mm the
> wants_signal loop in complete_signal would find the dumper task and
> then complete_signal would wake the dumper task with signal_wake_up.
> 
> After moving the coredump_task_exit above the setting of PF_EXITING in
> commit 92307383082d ("coredump: Don't perform any cleanups before
> dumping core") wants_signal will consider all of the threads in a
> multi-threaded process for waking up, not just the core dumping task.
> 
> Luckily complete_signal short circuits SIGKILL during a coredump marks
> every thread with SIGKILL and signal_wake_up.  This code is arguably
> buggy however as it tries to skip creating a group exit when is already
> present, and it fails that a coredump is in progress.
> 
> Ever since commit 06af8679449d ("coredump: Limit what can interrupt
> coredumps") was added dump_interrupted needs not just TIF_SIGPENDING
> set on the dumper task but also SIGKILL set in it's pending bitmap.
> This means that if the code is ever fixed not to short-circuit and
> kill a process after it has already been killed the special case
> for SIGKILL during a coredump will be broken.
> 
> Sort all of this out by making the coredump special case more special,
> and perform all of the work in prepare_signal and leave the rest of
> the signal delivery path out of it.
> 
> In prepare_signal when the process coredumping is sent SIGKILL find
> the task performing the coredump and use sigaddset and signal_wake_up
> to ensure that task reports fatal_signal_pending.
> 
> Return false from prepare_signal to tell the rest of the signal
> delivery path to ignore the signal.
> 
> Update wait_for_dump_helpers to perform a wait_event_killable wait
> so that if signal_pending gets set spuriously the wait will not
> be interrupted unless fatal_signal_pending is true.
> 
> I have tested this and verified I did not break SIGKILL during
> coredumps by accident (before or after this change).  I actually
> thought I had and I had to figure out what I had misread that kept
> SIGKILL during coredumps working.
> 
> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> ---
>  fs/coredump.c   |  4 ++--
>  kernel/signal.c | 11 +++++++++--
>  2 files changed, 11 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/coredump.c b/fs/coredump.c
> index a6b3c196cdef..7b91fb32dbb8 100644
> --- a/fs/coredump.c
> +++ b/fs/coredump.c
> @@ -448,7 +448,7 @@ static void coredump_finish(bool core_dumped)
>  static bool dump_interrupted(void)
>  {
>  	/*
> -	 * SIGKILL or freezing() interrupt the coredumping. Perhaps we
> +	 * SIGKILL or freezing() interrupted the coredumping. Perhaps we
>  	 * can do try_to_freeze() and check __fatal_signal_pending(),
>  	 * but then we need to teach dump_write() to restart and clear
>  	 * TIF_SIGPENDING.
> @@ -471,7 +471,7 @@ static void wait_for_dump_helpers(struct file *file)
>  	 * We actually want wait_event_freezable() but then we need
>  	 * to clear TIF_SIGPENDING and improve dump_interrupted().
>  	 */
> -	wait_event_interruptible(pipe->rd_wait, pipe->readers == 1);
> +	wait_event_killable(pipe->rd_wait, pipe->readers == 1);
>  
>  	pipe_lock(pipe);
>  	pipe->readers--;
> diff --git a/kernel/signal.c b/kernel/signal.c
> index 8272cac5f429..7e305a8ec7c2 100644
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -907,8 +907,15 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
>  	sigset_t flush;
>  
>  	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
> -		if (!(signal->flags & SIGNAL_GROUP_EXIT))
> -			return sig == SIGKILL;
> +		struct core_state *core_state = signal->core_state;
> +		if (core_state) {
> +			if (sig == SIGKILL) {
> +				struct task_struct *dumper = core_state->dumper.task;
> +				sigaddset(&dumper->pending.signal, SIGKILL);
> +				signal_wake_up(dumper, 1);
> +			}
> +			return false;
> +		}
>  		/*
>  		 * The process is in the middle of dying, nothing to do.
>  		 */
> 

Hi,

This patch breaks userspace, in particular it breaks gst-plugin-scanner
of GStreamer which hangs now on next-20211224. IIUC, this tool builds a
registry of good/working GStreamer plugins by loading them and
blacklisting those that don't work (crash). Before the hang I see
systemd-coredump process running, taking snapshot of gst-plugin-scanner
and then gst-plugin-scanner gets stuck.

Bisection points at this patch, reverting it restores
gst-plugin-scanner. Systemd-coredump still running, but there is no hang
anymore and everything works properly as before.

I'm seeing this problem on ARM32 and haven't checked other arches.
Please fix, thanks in advance.

Re: [PATCH 01/17] exit: Remove profile_task_exit & profile_munmap

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Dmitry Osipenko <digetx@gmail.com> writes:

> 14.12.2021 01:53, Eric W. Biederman пишет:
>> Simplify the code that allows SIGKILL during coredumps to terminate
>> the coredump.  As far as I can tell I have avoided breaking it
>> by dumb luck.
>> 
>> Historically with all of the other threads stopping in exit_mm the
>> wants_signal loop in complete_signal would find the dumper task and
>> then complete_signal would wake the dumper task with signal_wake_up.
>> 
>> After moving the coredump_task_exit above the setting of PF_EXITING in
>> commit 92307383082d ("coredump: Don't perform any cleanups before
>> dumping core") wants_signal will consider all of the threads in a
>> multi-threaded process for waking up, not just the core dumping task.
>> 
>> Luckily complete_signal short circuits SIGKILL during a coredump marks
>> every thread with SIGKILL and signal_wake_up.  This code is arguably
>> buggy however as it tries to skip creating a group exit when is already
>> present, and it fails that a coredump is in progress.
>> 
>> Ever since commit 06af8679449d ("coredump: Limit what can interrupt
>> coredumps") was added dump_interrupted needs not just TIF_SIGPENDING
>> set on the dumper task but also SIGKILL set in it's pending bitmap.
>> This means that if the code is ever fixed not to short-circuit and
>> kill a process after it has already been killed the special case
>> for SIGKILL during a coredump will be broken.
>> 
>> Sort all of this out by making the coredump special case more special,
>> and perform all of the work in prepare_signal and leave the rest of
>> the signal delivery path out of it.
>> 
>> In prepare_signal when the process coredumping is sent SIGKILL find
>> the task performing the coredump and use sigaddset and signal_wake_up
>> to ensure that task reports fatal_signal_pending.
>> 
>> Return false from prepare_signal to tell the rest of the signal
>> delivery path to ignore the signal.
>> 
>> Update wait_for_dump_helpers to perform a wait_event_killable wait
>> so that if signal_pending gets set spuriously the wait will not
>> be interrupted unless fatal_signal_pending is true.
>> 
>> I have tested this and verified I did not break SIGKILL during
>> coredumps by accident (before or after this change).  I actually
>> thought I had and I had to figure out what I had misread that kept
>> SIGKILL during coredumps working.
>> 
>> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
>> ---
>>  fs/coredump.c   |  4 ++--
>>  kernel/signal.c | 11 +++++++++--
>>  2 files changed, 11 insertions(+), 4 deletions(-)
>> 
>> diff --git a/fs/coredump.c b/fs/coredump.c
>> index a6b3c196cdef..7b91fb32dbb8 100644
>> --- a/fs/coredump.c
>> +++ b/fs/coredump.c
>> @@ -448,7 +448,7 @@ static void coredump_finish(bool core_dumped)
>>  static bool dump_interrupted(void)
>>  {
>>  	/*
>> -	 * SIGKILL or freezing() interrupt the coredumping. Perhaps we
>> +	 * SIGKILL or freezing() interrupted the coredumping. Perhaps we
>>  	 * can do try_to_freeze() and check __fatal_signal_pending(),
>>  	 * but then we need to teach dump_write() to restart and clear
>>  	 * TIF_SIGPENDING.
>> @@ -471,7 +471,7 @@ static void wait_for_dump_helpers(struct file *file)
>>  	 * We actually want wait_event_freezable() but then we need
>>  	 * to clear TIF_SIGPENDING and improve dump_interrupted().
>>  	 */
>> -	wait_event_interruptible(pipe->rd_wait, pipe->readers == 1);
>> +	wait_event_killable(pipe->rd_wait, pipe->readers == 1);
>>  
>>  	pipe_lock(pipe);
>>  	pipe->readers--;
>> diff --git a/kernel/signal.c b/kernel/signal.c
>> index 8272cac5f429..7e305a8ec7c2 100644
>> --- a/kernel/signal.c
>> +++ b/kernel/signal.c
>> @@ -907,8 +907,15 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
>>  	sigset_t flush;
>>  
>>  	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
>> -		if (!(signal->flags & SIGNAL_GROUP_EXIT))
>> -			return sig == SIGKILL;
>> +		struct core_state *core_state = signal->core_state;
>> +		if (core_state) {
>> +			if (sig == SIGKILL) {
>> +				struct task_struct *dumper = core_state->dumper.task;
>> +				sigaddset(&dumper->pending.signal, SIGKILL);
>> +				signal_wake_up(dumper, 1);
>> +			}
>> +			return false;
>> +		}
>>  		/*
>>  		 * The process is in the middle of dying, nothing to do.
>>  		 */
>> 
>
> Hi,
>
> This patch breaks userspace, in particular it breaks gst-plugin-scanner
> of GStreamer which hangs now on next-20211224. IIUC, this tool builds a
> registry of good/working GStreamer plugins by loading them and
> blacklisting those that don't work (crash). Before the hang I see
> systemd-coredump process running, taking snapshot of gst-plugin-scanner
> and then gst-plugin-scanner gets stuck.
>
> Bisection points at this patch, reverting it restores
> gst-plugin-scanner. Systemd-coredump still running, but there is no hang
> anymore and everything works properly as before.
>
> I'm seeing this problem on ARM32 and haven't checked other arches.
> Please fix, thanks in advance.

That is weird.

Doubly weird really because this should only change the case where
coredumps are interrupted by SIGKILL.

What distro are you running?  I would like to match things as closely
as I can.  So I can reproduce the issue so I can figure out what
is wrong so I can fix it.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Linus Torvalds]

On Mon, Dec 13, 2021 at 2:54 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>
>
>         if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
> -               if (!(signal->flags & SIGNAL_GROUP_EXIT))
> -                       return sig == SIGKILL;
> +               struct core_state *core_state = signal->core_state;
> +               if (core_state) {

This change is very confusing.

Also, why does it do that 'signal->core_state->dumper.task', when we
already know that it's the same as 'signal->group_exit_task'?

The only thing that sets 'signal->core_state' also sets
'signal->group_exit_task', and the call chain has set both to the same
task.

So the code is odd and makes little sense.

But what's even more odd is how it

 (a) sends the SIGKILL to somebody else

 (b) does *NOT* send SIGKILL to itself

Now, (a) is explained in the commit message. The intent is to signal
the core dumper.

But (b) looks like a fundamental change in semantics. The target of
the SIGKILL is still running, might be in some loop in the kernel that
wants to be interrupted by a fatal signal, and you expressly disabled
the code that would send that fatal signal.

If I send SIGKILL to thread A, then that SIGKILL had *better* be
delivered. To thread A, which may be in a "mutex_lock_killable()" or
whatever else.

The fact that thread B may be in the process of trying to dump core
doesn't change that at all, as far as I can see.

So I think this patch is fundamentally buggy and wrong. Or at least
needs much more explanation of why you'd not send SIGKILL to the
target thread.

               Linus

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Linus Torvalds <torvalds@linux-foundation.org> writes:

> On Mon, Dec 13, 2021 at 2:54 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>>
>>
>>         if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
>> -               if (!(signal->flags & SIGNAL_GROUP_EXIT))
>> -                       return sig == SIGKILL;
>> +               struct core_state *core_state = signal->core_state;
>> +               if (core_state) {
>
> This change is very confusing.
>
> Also, why does it do that 'signal->core_state->dumper.task', when we
> already know that it's the same as 'signal->group_exit_task'?
>
> The only thing that sets 'signal->core_state' also sets
> 'signal->group_exit_task', and the call chain has set both to the same
> task.
>
> So the code is odd and makes little sense.

As you say signal->group_exit_task, and core_state->dumper.task point to
the same task.  So it may be a little silly when viewed independently of
everything else to use core_state->dumper.task instead of
group_exit_task as it is an extra cache line dereference.

The thing is signal->group_exit_task is only used by coredumps currently
as a flag to tell signal_group_exit to return true.  It is exec that
actually uses signal->group_exit_task in conjunction with
signal->notify_count to wake itself up.

Using a pointer as a flag and not for it's value.  Having different
semantics for who sets the pointer.  All of those are weird enough
I just want to make signal->group_exit_task to go away.

By using core_state->dumper.task I was able to make
signal->group_exit_task exclusive to the exec case in the following
changes, and to rename it signal->group_exec_task so no one gets
confused what the field is for.

> But what's even more odd is how it
>
>  (a) sends the SIGKILL to somebody else
>
>  (b) does *NOT* send SIGKILL to itself
>
> Now, (a) is explained in the commit message. The intent is to signal
> the core dumper.

Which is the a specific thread of the target process, and it is
the only thread running of the target process.

> But (b) looks like a fundamental change in semantics. The target of
> the SIGKILL is still running, might be in some loop in the kernel that
> wants to be interrupted by a fatal signal, and you expressly disabled
> the code that would send that fatal signal.
>
> If I send SIGKILL to thread A, then that SIGKILL had *better* be
> delivered. To thread A, which may be in a "mutex_lock_killable()" or
> whatever else.
>
> The fact that thread B may be in the process of trying to dump core
> doesn't change that at all, as far as I can see.
>
> So I think this patch is fundamentally buggy and wrong. Or at least
> needs much more explanation of why you'd not send SIGKILL to the
> target thread.

If you look at zap_threads.  You can observe that it takes the siglock,
sets SIGNAL_GROUP_COREDUMP, and sets signal->core_state and in
zap_process makes SIGKILL pending is the per-task sigset, and calls
signal_wake_up on every task.

This case in prepare_signal happens after that.  After every task
has been told to die, and __fatal_signal_pending is true for all of
them if they have not reached do_exit yet.



If you look in zap_threads you will see that the core dumping thread
clears TIF_SIGPENDING, and in general makes fatal_signal_pending false
for itself.  But keep in mind that this thread because it is dumping
core is already on the path to do_exit.  It has already processed a
fatal signal.


So in the special case I only worry about the dumping task as it is the
only task after zap_threads that does not have fatal_signal_pending.


This is different than the ordinary case of delivering SIGKILL
where complete_signal makes SIGKILL pending in the per-task sigset
of every task in the process.


Currently I suspect changing wait_event_uninterruptible to
wait_event_killable, is causing problems.

Or perhaps there is some reason tasks that have already entered do_exit
need to have fatal_signal_pending set. (The will have
fatal_signal_pending set up until they enter get_signal which calls
do_group_exit which calls do_exit).

Which is why I am trying to reproduce the reported failure so I can get
the kernel to tell me what is going on.  If this is not resolved quickly
I won't send you this change, and I will pull it out of linux-next.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Dmitry Osipenko <digetx@gmail.com> writes:

> 14.12.2021 01:53, Eric W. Biederman пишет:
>> Simplify the code that allows SIGKILL during coredumps to terminate
>> the coredump.  As far as I can tell I have avoided breaking it
>> by dumb luck.
>> 
>> Historically with all of the other threads stopping in exit_mm the
>> wants_signal loop in complete_signal would find the dumper task and
>> then complete_signal would wake the dumper task with signal_wake_up.
>> 
>> After moving the coredump_task_exit above the setting of PF_EXITING in
>> commit 92307383082d ("coredump: Don't perform any cleanups before
>> dumping core") wants_signal will consider all of the threads in a
>> multi-threaded process for waking up, not just the core dumping task.
>> 
>> Luckily complete_signal short circuits SIGKILL during a coredump marks
>> every thread with SIGKILL and signal_wake_up.  This code is arguably
>> buggy however as it tries to skip creating a group exit when is already
>> present, and it fails that a coredump is in progress.
>> 
>> Ever since commit 06af8679449d ("coredump: Limit what can interrupt
>> coredumps") was added dump_interrupted needs not just TIF_SIGPENDING
>> set on the dumper task but also SIGKILL set in it's pending bitmap.
>> This means that if the code is ever fixed not to short-circuit and
>> kill a process after it has already been killed the special case
>> for SIGKILL during a coredump will be broken.
>> 
>> Sort all of this out by making the coredump special case more special,
>> and perform all of the work in prepare_signal and leave the rest of
>> the signal delivery path out of it.
>> 
>> In prepare_signal when the process coredumping is sent SIGKILL find
>> the task performing the coredump and use sigaddset and signal_wake_up
>> to ensure that task reports fatal_signal_pending.
>> 
>> Return false from prepare_signal to tell the rest of the signal
>> delivery path to ignore the signal.
>> 
>> Update wait_for_dump_helpers to perform a wait_event_killable wait
>> so that if signal_pending gets set spuriously the wait will not
>> be interrupted unless fatal_signal_pending is true.
>> 
>> I have tested this and verified I did not break SIGKILL during
>> coredumps by accident (before or after this change).  I actually
>> thought I had and I had to figure out what I had misread that kept
>> SIGKILL during coredumps working.
>> 
>> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
>> ---
>>  fs/coredump.c   |  4 ++--
>>  kernel/signal.c | 11 +++++++++--
>>  2 files changed, 11 insertions(+), 4 deletions(-)
>> 
>> diff --git a/fs/coredump.c b/fs/coredump.c
>> index a6b3c196cdef..7b91fb32dbb8 100644
>> --- a/fs/coredump.c
>> +++ b/fs/coredump.c
>> @@ -448,7 +448,7 @@ static void coredump_finish(bool core_dumped)
>>  static bool dump_interrupted(void)
>>  {
>>  	/*
>> -	 * SIGKILL or freezing() interrupt the coredumping. Perhaps we
>> +	 * SIGKILL or freezing() interrupted the coredumping. Perhaps we
>>  	 * can do try_to_freeze() and check __fatal_signal_pending(),
>>  	 * but then we need to teach dump_write() to restart and clear
>>  	 * TIF_SIGPENDING.
>> @@ -471,7 +471,7 @@ static void wait_for_dump_helpers(struct file *file)
>>  	 * We actually want wait_event_freezable() but then we need
>>  	 * to clear TIF_SIGPENDING and improve dump_interrupted().
>>  	 */
>> -	wait_event_interruptible(pipe->rd_wait, pipe->readers == 1);
>> +	wait_event_killable(pipe->rd_wait, pipe->readers == 1);
>>  
>>  	pipe_lock(pipe);
>>  	pipe->readers--;
>> diff --git a/kernel/signal.c b/kernel/signal.c
>> index 8272cac5f429..7e305a8ec7c2 100644
>> --- a/kernel/signal.c
>> +++ b/kernel/signal.c
>> @@ -907,8 +907,15 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
>>  	sigset_t flush;
>>  
>>  	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
>> -		if (!(signal->flags & SIGNAL_GROUP_EXIT))
>> -			return sig == SIGKILL;
>> +		struct core_state *core_state = signal->core_state;
>> +		if (core_state) {
>> +			if (sig == SIGKILL) {
>> +				struct task_struct *dumper = core_state->dumper.task;
>> +				sigaddset(&dumper->pending.signal, SIGKILL);
>> +				signal_wake_up(dumper, 1);
>> +			}
>> +			return false;
>> +		}
>>  		/*
>>  		 * The process is in the middle of dying, nothing to do.
>>  		 */
>> 
>
> Hi,
>
> This patch breaks userspace, in particular it breaks gst-plugin-scanner
> of GStreamer which hangs now on next-20211224. IIUC, this tool builds a
> registry of good/working GStreamer plugins by loading them and
> blacklisting those that don't work (crash). Before the hang I see
> systemd-coredump process running, taking snapshot of gst-plugin-scanner
> and then gst-plugin-scanner gets stuck.
>
> Bisection points at this patch, reverting it restores
> gst-plugin-scanner. Systemd-coredump still running, but there is no hang
> anymore and everything works properly as before.
>
> I'm seeing this problem on ARM32 and haven't checked other arches.
> Please fix, thanks in advance.


I have not yet been able to figure out how to run gst-pluggin-scanner in
a way that triggers this yet.  In truth I can't figure out how to
run gst-pluggin-scanner in a useful way.

I am going to set up some unit tests and see if I can reproduce your
hang another way, but if you could give me some more information on what
you are doing to trigger this I would appreciate it.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Dmitry Osipenko]

05.01.2022 22:58, Eric W. Biederman пишет:
> Dmitry Osipenko <digetx@gmail.com> writes:
> 
>> 14.12.2021 01:53, Eric W. Biederman пишет:
>>> Simplify the code that allows SIGKILL during coredumps to terminate
>>> the coredump.  As far as I can tell I have avoided breaking it
>>> by dumb luck.
>>>
>>> Historically with all of the other threads stopping in exit_mm the
>>> wants_signal loop in complete_signal would find the dumper task and
>>> then complete_signal would wake the dumper task with signal_wake_up.
>>>
>>> After moving the coredump_task_exit above the setting of PF_EXITING in
>>> commit 92307383082d ("coredump: Don't perform any cleanups before
>>> dumping core") wants_signal will consider all of the threads in a
>>> multi-threaded process for waking up, not just the core dumping task.
>>>
>>> Luckily complete_signal short circuits SIGKILL during a coredump marks
>>> every thread with SIGKILL and signal_wake_up.  This code is arguably
>>> buggy however as it tries to skip creating a group exit when is already
>>> present, and it fails that a coredump is in progress.
>>>
>>> Ever since commit 06af8679449d ("coredump: Limit what can interrupt
>>> coredumps") was added dump_interrupted needs not just TIF_SIGPENDING
>>> set on the dumper task but also SIGKILL set in it's pending bitmap.
>>> This means that if the code is ever fixed not to short-circuit and
>>> kill a process after it has already been killed the special case
>>> for SIGKILL during a coredump will be broken.
>>>
>>> Sort all of this out by making the coredump special case more special,
>>> and perform all of the work in prepare_signal and leave the rest of
>>> the signal delivery path out of it.
>>>
>>> In prepare_signal when the process coredumping is sent SIGKILL find
>>> the task performing the coredump and use sigaddset and signal_wake_up
>>> to ensure that task reports fatal_signal_pending.
>>>
>>> Return false from prepare_signal to tell the rest of the signal
>>> delivery path to ignore the signal.
>>>
>>> Update wait_for_dump_helpers to perform a wait_event_killable wait
>>> so that if signal_pending gets set spuriously the wait will not
>>> be interrupted unless fatal_signal_pending is true.
>>>
>>> I have tested this and verified I did not break SIGKILL during
>>> coredumps by accident (before or after this change).  I actually
>>> thought I had and I had to figure out what I had misread that kept
>>> SIGKILL during coredumps working.
>>>
>>> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
>>> ---
>>>  fs/coredump.c   |  4 ++--
>>>  kernel/signal.c | 11 +++++++++--
>>>  2 files changed, 11 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/fs/coredump.c b/fs/coredump.c
>>> index a6b3c196cdef..7b91fb32dbb8 100644
>>> --- a/fs/coredump.c
>>> +++ b/fs/coredump.c
>>> @@ -448,7 +448,7 @@ static void coredump_finish(bool core_dumped)
>>>  static bool dump_interrupted(void)
>>>  {
>>>  	/*
>>> -	 * SIGKILL or freezing() interrupt the coredumping. Perhaps we
>>> +	 * SIGKILL or freezing() interrupted the coredumping. Perhaps we
>>>  	 * can do try_to_freeze() and check __fatal_signal_pending(),
>>>  	 * but then we need to teach dump_write() to restart and clear
>>>  	 * TIF_SIGPENDING.
>>> @@ -471,7 +471,7 @@ static void wait_for_dump_helpers(struct file *file)
>>>  	 * We actually want wait_event_freezable() but then we need
>>>  	 * to clear TIF_SIGPENDING and improve dump_interrupted().
>>>  	 */
>>> -	wait_event_interruptible(pipe->rd_wait, pipe->readers == 1);
>>> +	wait_event_killable(pipe->rd_wait, pipe->readers == 1);
>>>  
>>>  	pipe_lock(pipe);
>>>  	pipe->readers--;
>>> diff --git a/kernel/signal.c b/kernel/signal.c
>>> index 8272cac5f429..7e305a8ec7c2 100644
>>> --- a/kernel/signal.c
>>> +++ b/kernel/signal.c
>>> @@ -907,8 +907,15 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
>>>  	sigset_t flush;
>>>  
>>>  	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
>>> -		if (!(signal->flags & SIGNAL_GROUP_EXIT))
>>> -			return sig == SIGKILL;
>>> +		struct core_state *core_state = signal->core_state;
>>> +		if (core_state) {
>>> +			if (sig == SIGKILL) {
>>> +				struct task_struct *dumper = core_state->dumper.task;
>>> +				sigaddset(&dumper->pending.signal, SIGKILL);
>>> +				signal_wake_up(dumper, 1);
>>> +			}
>>> +			return false;
>>> +		}
>>>  		/*
>>>  		 * The process is in the middle of dying, nothing to do.
>>>  		 */
>>>
>>
>> Hi,
>>
>> This patch breaks userspace, in particular it breaks gst-plugin-scanner
>> of GStreamer which hangs now on next-20211224. IIUC, this tool builds a
>> registry of good/working GStreamer plugins by loading them and
>> blacklisting those that don't work (crash). Before the hang I see
>> systemd-coredump process running, taking snapshot of gst-plugin-scanner
>> and then gst-plugin-scanner gets stuck.
>>
>> Bisection points at this patch, reverting it restores
>> gst-plugin-scanner. Systemd-coredump still running, but there is no hang
>> anymore and everything works properly as before.
>>
>> I'm seeing this problem on ARM32 and haven't checked other arches.
>> Please fix, thanks in advance.
> 
> 
> I have not yet been able to figure out how to run gst-pluggin-scanner in
> a way that triggers this yet.  In truth I can't figure out how to
> run gst-pluggin-scanner in a useful way.
> 
> I am going to set up some unit tests and see if I can reproduce your
> hang another way, but if you could give me some more information on what
> you are doing to trigger this I would appreciate it.

Thanks, Eric. The distro is Arch Linux, but it's a development
environment where I'm running latest GStreamer from git master. I'll try
to figure out the reproduction steps and get back to you.

Re: [PATCH 01/17] exit: Remove profile_task_exit & profile_munmap

[Al Viro]

On Mon, Jan 03, 2022 at 03:32:56PM -0600, Eric W. Biederman wrote:
> When I say remove I mean remove.  All profile_task_exit and
> profile_munmap do is call a blocking notifier chain.  The helpers
> profile_task_register and profile_task_unregister are not called
> anywhere in the tree.  Which means this is all dead code.
> 
> So remove the dead code and make it easier to read do_exit.

How about doing the same to profile_handoff_task() and
task_handoff_register()/task_handoff_unregister(),
while we are at it?  Combined diff would be this:

diff --git a/include/linux/profile.h b/include/linux/profile.h
index fd18ca96f5574..6aa64730298a0 100644
--- a/include/linux/profile.h
+++ b/include/linux/profile.h
@@ -31,11 +31,6 @@ static inline int create_proc_profile(void)
 }
 #endif
 
-enum profile_type {
-	PROFILE_TASK_EXIT,
-	PROFILE_MUNMAP
-};
-
 #ifdef CONFIG_PROFILING
 
 extern int prof_on __read_mostly;
@@ -63,26 +58,6 @@ static inline void profile_hit(int type, void *ip)
 		profile_hits(type, ip, 1);
 }
 
-struct task_struct;
-struct mm_struct;
-
-/* task is in do_exit() */
-void profile_task_exit(struct task_struct * task);
-
-/* task is dead, free task struct ? Returns 1 if
- * the task was taken, 0 if the task should be freed.
- */
-int profile_handoff_task(struct task_struct * task);
-
-/* sys_munmap */
-void profile_munmap(unsigned long addr);
-
-int task_handoff_register(struct notifier_block * n);
-int task_handoff_unregister(struct notifier_block * n);
-
-int profile_event_register(enum profile_type, struct notifier_block * n);
-int profile_event_unregister(enum profile_type, struct notifier_block * n);
-
 #else
 
 #define prof_on 0
@@ -107,30 +82,6 @@ static inline void profile_hit(int type, void *ip)
 	return;
 }
 
-static inline int task_handoff_register(struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-static inline int task_handoff_unregister(struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-static inline int profile_event_register(enum profile_type t, struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-static inline int profile_event_unregister(enum profile_type t, struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-#define profile_task_exit(a) do { } while (0)
-#define profile_handoff_task(a) (0)
-#define profile_munmap(a) do { } while (0)
-
 #endif /* CONFIG_PROFILING */
 
 #endif /* _LINUX_PROFILE_H */
diff --git a/kernel/exit.c b/kernel/exit.c
index f702a6a63686e..5086a5e9d02de 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -765,7 +765,6 @@ void __noreturn do_exit(long code)
 		preempt_count_set(PREEMPT_ENABLED);
 	}
 
-	profile_task_exit(tsk);
 	kcov_task_exit(tsk);
 
 	coredump_task_exit(tsk);
diff --git a/kernel/fork.c b/kernel/fork.c
index 3244cc56b697d..496c0b6c8cb83 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -754,9 +754,7 @@ void __put_task_struct(struct task_struct *tsk)
 	delayacct_tsk_free(tsk);
 	put_signal_struct(tsk->signal);
 	sched_core_free(tsk);
-
-	if (!profile_handoff_task(tsk))
-		free_task(tsk);
+	free_task(tsk);
 }
 EXPORT_SYMBOL_GPL(__put_task_struct);
 
diff --git a/kernel/profile.c b/kernel/profile.c
index eb9c7f0f5ac52..37640a0bd8a3c 100644
--- a/kernel/profile.c
+++ b/kernel/profile.c
@@ -133,79 +133,6 @@ int __ref profile_init(void)
 	return -ENOMEM;
 }
 
-/* Profile event notifications */
-
-static BLOCKING_NOTIFIER_HEAD(task_exit_notifier);
-static ATOMIC_NOTIFIER_HEAD(task_free_notifier);
-static BLOCKING_NOTIFIER_HEAD(munmap_notifier);
-
-void profile_task_exit(struct task_struct *task)
-{
-	blocking_notifier_call_chain(&task_exit_notifier, 0, task);
-}
-
-int profile_handoff_task(struct task_struct *task)
-{
-	int ret;
-	ret = atomic_notifier_call_chain(&task_free_notifier, 0, task);
-	return (ret == NOTIFY_OK) ? 1 : 0;
-}
-
-void profile_munmap(unsigned long addr)
-{
-	blocking_notifier_call_chain(&munmap_notifier, 0, (void *)addr);
-}
-
-int task_handoff_register(struct notifier_block *n)
-{
-	return atomic_notifier_chain_register(&task_free_notifier, n);
-}
-EXPORT_SYMBOL_GPL(task_handoff_register);
-
-int task_handoff_unregister(struct notifier_block *n)
-{
-	return atomic_notifier_chain_unregister(&task_free_notifier, n);
-}
-EXPORT_SYMBOL_GPL(task_handoff_unregister);
-
-int profile_event_register(enum profile_type type, struct notifier_block *n)
-{
-	int err = -EINVAL;
-
-	switch (type) {
-	case PROFILE_TASK_EXIT:
-		err = blocking_notifier_chain_register(
-				&task_exit_notifier, n);
-		break;
-	case PROFILE_MUNMAP:
-		err = blocking_notifier_chain_register(
-				&munmap_notifier, n);
-		break;
-	}
-
-	return err;
-}
-EXPORT_SYMBOL_GPL(profile_event_register);
-
-int profile_event_unregister(enum profile_type type, struct notifier_block *n)
-{
-	int err = -EINVAL;
-
-	switch (type) {
-	case PROFILE_TASK_EXIT:
-		err = blocking_notifier_chain_unregister(
-				&task_exit_notifier, n);
-		break;
-	case PROFILE_MUNMAP:
-		err = blocking_notifier_chain_unregister(
-				&munmap_notifier, n);
-		break;
-	}
-
-	return err;
-}
-EXPORT_SYMBOL_GPL(profile_event_unregister);
-
 #if defined(CONFIG_SMP) && defined(CONFIG_PROC_FS)
 /*
  * Each cpu has a pair of open-addressed hashtables for pending
diff --git a/mm/mmap.c b/mm/mmap.c
index bfb0ea164a90a..70318c2a47c39 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2928,7 +2928,6 @@ EXPORT_SYMBOL(vm_munmap);
 SYSCALL_DEFINE2(munmap, unsigned long, addr, size_t, len)
 {
 	addr = untagged_addr(addr);
-	profile_munmap(addr);
 	return __vm_munmap(addr, len, true);
 }
 

Re: [PATCH 01/17] exit: Remove profile_task_exit & profile_munmap

[Eric W. Biederman]

Al Viro <viro@zeniv.linux.org.uk> writes:

> On Mon, Jan 03, 2022 at 03:32:56PM -0600, Eric W. Biederman wrote:
>> When I say remove I mean remove.  All profile_task_exit and
>> profile_munmap do is call a blocking notifier chain.  The helpers
>> profile_task_register and profile_task_unregister are not called
>> anywhere in the tree.  Which means this is all dead code.
>> 
>> So remove the dead code and make it easier to read do_exit.
>
> How about doing the same to profile_handoff_task() and
> task_handoff_register()/task_handoff_unregister(),
> while we are at it?  Combined diff would be this:

A very good idea.  I have added this incremental patch to my queue.

Eric

From: "Eric W. Biederman" <ebiederm@xmission.com>
Date: Sat, 8 Jan 2022 10:03:24 -0600
Subject: [PATCH] exit: Remove profile_handoff_task

All profile_handoff_task does is notify the task_free_notifier chain.
The helpers task_handoff_register and task_handoff_unregister are used
to add and delete entries from that chain and are never called.

So remove the dead code and make it much easier to read and reason
about __put_task_struct.

Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 include/linux/profile.h | 19 -------------------
 kernel/fork.c           |  4 +---
 kernel/profile.c        | 23 -----------------------
 3 files changed, 1 insertion(+), 45 deletions(-)

diff --git a/include/linux/profile.h b/include/linux/profile.h
index f7eb2b57d890..11db1ec516e2 100644
--- a/include/linux/profile.h
+++ b/include/linux/profile.h
@@ -61,14 +61,6 @@ static inline void profile_hit(int type, void *ip)
 struct task_struct;
 struct mm_struct;
 
-/* task is dead, free task struct ? Returns 1 if
- * the task was taken, 0 if the task should be freed.
- */
-int profile_handoff_task(struct task_struct * task);
-
-int task_handoff_register(struct notifier_block * n);
-int task_handoff_unregister(struct notifier_block * n);
-
 #else
 
 #define prof_on 0
@@ -93,17 +85,6 @@ static inline void profile_hit(int type, void *ip)
 	return;
 }
 
-static inline int task_handoff_register(struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-static inline int task_handoff_unregister(struct notifier_block * n)
-{
-	return -ENOSYS;
-}
-
-#define profile_handoff_task(a) (0)
 
 #endif /* CONFIG_PROFILING */
 
diff --git a/kernel/fork.c b/kernel/fork.c
index 6f0293cb29c9..494539ecb6d3 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -754,9 +754,7 @@ void __put_task_struct(struct task_struct *tsk)
 	delayacct_tsk_free(tsk);
 	put_signal_struct(tsk->signal);
 	sched_core_free(tsk);
-
-	if (!profile_handoff_task(tsk))
-		free_task(tsk);
+	free_task(tsk);
 }
 EXPORT_SYMBOL_GPL(__put_task_struct);
 
diff --git a/kernel/profile.c b/kernel/profile.c
index 9355cc934a96..37640a0bd8a3 100644
--- a/kernel/profile.c
+++ b/kernel/profile.c
@@ -133,29 +133,6 @@ int __ref profile_init(void)
 	return -ENOMEM;
 }
 
-/* Profile event notifications */
-
-static ATOMIC_NOTIFIER_HEAD(task_free_notifier);
-
-int profile_handoff_task(struct task_struct *task)
-{
-	int ret;
-	ret = atomic_notifier_call_chain(&task_free_notifier, 0, task);
-	return (ret == NOTIFY_OK) ? 1 : 0;
-}
-
-int task_handoff_register(struct notifier_block *n)
-{
-	return atomic_notifier_chain_register(&task_free_notifier, n);
-}
-EXPORT_SYMBOL_GPL(task_handoff_register);
-
-int task_handoff_unregister(struct notifier_block *n)
-{
-	return atomic_notifier_chain_unregister(&task_free_notifier, n);
-}
-EXPORT_SYMBOL_GPL(task_handoff_unregister);
-
 #if defined(CONFIG_SMP) && defined(CONFIG_PROC_FS)
 /*
  * Each cpu has a pair of open-addressed hashtables for pending
-- 
2.29.2

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Dmitry Osipenko <digetx@gmail.com> writes:

> 05.01.2022 22:58, Eric W. Biederman пишет:
>> 
>> I have not yet been able to figure out how to run gst-pluggin-scanner in
>> a way that triggers this yet.  In truth I can't figure out how to
>> run gst-pluggin-scanner in a useful way.
>> 
>> I am going to set up some unit tests and see if I can reproduce your
>> hang another way, but if you could give me some more information on what
>> you are doing to trigger this I would appreciate it.
>
> Thanks, Eric. The distro is Arch Linux, but it's a development
> environment where I'm running latest GStreamer from git master. I'll try
> to figure out the reproduction steps and get back to you.

Thank you.

Until I can figure out why this is causing problems I have dropped the
following two patches from my queue:
 signal: Make SIGKILL during coredumps an explicit special case
 signal: Drop signals received after a fatal signal has been processed

I have replaced them with the following two patches that just do what
is needed for the rest of the code in the series:
 signal: Have prepare_signal detect coredumps using
 signal: Make coredump handling explicit in complete_signal

Perversely my failure to change the SIGKILL handling when coredumps are
happening proves to me that I need to change the SIGKILL handling when
coredumps are happening to make the code more maintainable.

Eric

[PATCH 1/2] signal: Have prepare_signal detect coredumps using signal->core_state

[Eric W. Biederman]

In preparation for removing the flag SIGNAL_GROUP_COREDUMP, change
prepare_signal to test signal->core_state instead of the flag
SIGNAL_GROUP_COREDUMP.

Both fields are protected by siglock and both live in signal_struct so
there are no real tradeoffs here, just a change to which field is
being tested.

Link: https://lkml.kernel.org/r/20211213225350.27481-1-ebiederm@xmission.com
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 kernel/signal.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/signal.c b/kernel/signal.c
index 8272cac5f429..f95a4423519d 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -906,8 +906,8 @@ static bool prepare_signal(int sig, struct task_struct *p, bool force)
 	struct task_struct *t;
 	sigset_t flush;
 
-	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
-		if (!(signal->flags & SIGNAL_GROUP_EXIT))
+	if ((signal->flags & SIGNAL_GROUP_EXIT) || signal->core_state) {
+		if (signal->core_state)
 			return sig == SIGKILL;
 		/*
 		 * The process is in the middle of dying, nothing to do.
-- 
2.29.2

[PATCH 2/2] signal: Make coredump handling explicit in complete_signal

[Eric W. Biederman]

Ever since commit 6cd8f0acae34 ("coredump: ensure that SIGKILL always
kills the dumping thread") it has been possible for a SIGKILL received
during a coredump to set SIGNAL_GROUP_EXIT and trigger a process
shutdown (for a second time).

Update the logic to explicitly allow coredumps so that coredumps can
set SIGNAL_GROUP_EXIT and shutdown like an ordinary process.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 kernel/signal.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/signal.c b/kernel/signal.c
index f95a4423519d..0706c1345a71 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1032,7 +1032,7 @@ static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
 	 * then start taking the whole group down immediately.
 	 */
 	if (sig_fatal(p, sig) &&
-	    !(signal->flags & SIGNAL_GROUP_EXIT) &&
+	    (signal->core_state || !(signal->flags & SIGNAL_GROUP_EXIT)) &&
 	    !sigismember(&t->real_blocked, sig) &&
 	    (sig == SIGKILL || !p->ptrace)) {
 		/*
-- 
2.29.2

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Heiko Carstens]

On Tue, Jan 04, 2022 at 01:47:05PM -0600, Eric W. Biederman wrote:
> Currently I suspect changing wait_event_uninterruptible to
> wait_event_killable, is causing problems.
> 
> Or perhaps there is some reason tasks that have already entered do_exit
> need to have fatal_signal_pending set. (The will have
> fatal_signal_pending set up until they enter get_signal which calls
> do_group_exit which calls do_exit).
> 
> Which is why I am trying to reproduce the reported failure so I can get
> the kernel to tell me what is going on.  If this is not resolved quickly
> I won't send you this change, and I will pull it out of linux-next.

It would have been good if you would have removed this from linux-next
already.

Anyway, now I also had to spend quite some time to bisect why several test
suites just hang with linux-next. It's probably because of holidays that
you didn't get more bug reports.

On s390

- ltp
- elfutils selftests
- seccomp kernel selftests

hang with linux-next.

I bisected the problem to this patch using elfutils selftests:

git clone git://sourceware.org/git/elfutils.git
cd elfutils
autoreconf -fi
./configure --enable-maintainer-mode --disable-debuginfod
make -j $(nproc) > /dev/null
cd tests
make -j $(nproc) check

Note: I actually didn't verify if this also causes ltp+seccomp selftests
      to hang. I just assume it is the case.

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Geert Uytterhoeven]

On Mon, Jan 3, 2022 at 10:33 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
> The generic function ptrace_report_syscall does a little more
> than syscall_trace on m68k.  The function ptrace_report_syscall
> stops early if PT_TRACED is not set, it sets ptrace_message,
> and returns the result of fatal_signal_pending.
>
> Setting ptrace_message to a passed in value of 0 is effectively not
> setting ptrace_message, making that additional work a noop.
>
> Returning the result of fatal_signal_pending and letting the caller
> ignore the result becomes a noop in this change.
>
> When a process is ptraced, the flag PT_PTRACED is always set in
> current->ptrace.  Testing for PT_PTRACED in ptrace_report_syscall is
> just an optimization to fail early if the process is not ptraced.
> Later on in ptrace_notify, ptrace_stop will test current->ptrace under
> tasklist_lock and skip performing any work if the task is not ptraced.
>
> Cc: Geert Uytterhoeven <geert@linux-m68k.org>
> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

As this depends on the removal of a parameter from
ptrace_report_syscall() earlier in this series:
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Al Viro]

On Mon, Jan 10, 2022 at 04:26:57PM +0100, Geert Uytterhoeven wrote:
> On Mon, Jan 3, 2022 at 10:33 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
> > The generic function ptrace_report_syscall does a little more
> > than syscall_trace on m68k.  The function ptrace_report_syscall
> > stops early if PT_TRACED is not set, it sets ptrace_message,
> > and returns the result of fatal_signal_pending.
> >
> > Setting ptrace_message to a passed in value of 0 is effectively not
> > setting ptrace_message, making that additional work a noop.
> >
> > Returning the result of fatal_signal_pending and letting the caller
> > ignore the result becomes a noop in this change.
> >
> > When a process is ptraced, the flag PT_PTRACED is always set in
> > current->ptrace.  Testing for PT_PTRACED in ptrace_report_syscall is
> > just an optimization to fail early if the process is not ptraced.
> > Later on in ptrace_notify, ptrace_stop will test current->ptrace under
> > tasklist_lock and skip performing any work if the task is not ptraced.
> >
> > Cc: Geert Uytterhoeven <geert@linux-m68k.org>
> > Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> 
> As this depends on the removal of a parameter from
> ptrace_report_syscall() earlier in this series:
> Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>

FWIW, I would suggest taking it a bit further: make syscall_trace_enter()
and syscall_trace_leave() in m68k ptrace.c unconditional, replace the
calls of syscall_trace() in entry.S with syscall_trace_enter() and
syscall_trace_leave() resp. and remove syscall_trace().

Geert, do you see any problems with that?  The only difference is that
current->ptrace_message would be set to 1 for ptrace stop on entry and
2 - on leave.  Currently m68k just has it 0 all along.

It is user-visible (the whole point is to let the tracer see which
stop it is - entry or exit one), so somebody using PTRACE_GETEVENTMSG
on syscall stops would start seeing 1 or 2 instead of "0 all along".
That's how it works on all other architectures (including m68k-nommu),
and I doubt that anything in userland will get broken.

Behaviour of PTRACE_GETEVENTMSG for other stops (fork, etc.) remains
as-is, of course.

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Al Viro]

On Mon, Jan 10, 2022 at 04:20:03PM +0000, Al Viro wrote:

> Geert, do you see any problems with that?  The only difference is that
> current->ptrace_message would be set to 1 for ptrace stop on entry and
> 2 - on leave.  Currently m68k just has it 0 all along.
> 
> It is user-visible (the whole point is to let the tracer see which
> stop it is - entry or exit one), so somebody using PTRACE_GETEVENTMSG
> on syscall stops would start seeing 1 or 2 instead of "0 all along".
> That's how it works on all other architectures (including m68k-nommu),
> and I doubt that anything in userland will get broken.
> 
> Behaviour of PTRACE_GETEVENTMSG for other stops (fork, etc.) remains
> as-is, of course.

Actually, the current behaviour is "report what the last PTRACE_GETEVENTMSG
has reported, whatever kind of stop that used to be for".  So I very much
doubt that anything could break there.

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Geert Uytterhoeven]

Hi Al,

CC Michael/m68k,

On Mon, Jan 10, 2022 at 5:20 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
> On Mon, Jan 10, 2022 at 04:26:57PM +0100, Geert Uytterhoeven wrote:
> > On Mon, Jan 3, 2022 at 10:33 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
> > > The generic function ptrace_report_syscall does a little more
> > > than syscall_trace on m68k.  The function ptrace_report_syscall
> > > stops early if PT_TRACED is not set, it sets ptrace_message,
> > > and returns the result of fatal_signal_pending.
> > >
> > > Setting ptrace_message to a passed in value of 0 is effectively not
> > > setting ptrace_message, making that additional work a noop.
> > >
> > > Returning the result of fatal_signal_pending and letting the caller
> > > ignore the result becomes a noop in this change.
> > >
> > > When a process is ptraced, the flag PT_PTRACED is always set in
> > > current->ptrace.  Testing for PT_PTRACED in ptrace_report_syscall is
> > > just an optimization to fail early if the process is not ptraced.
> > > Later on in ptrace_notify, ptrace_stop will test current->ptrace under
> > > tasklist_lock and skip performing any work if the task is not ptraced.
> > >
> > > Cc: Geert Uytterhoeven <geert@linux-m68k.org>
> > > Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> >
> > As this depends on the removal of a parameter from
> > ptrace_report_syscall() earlier in this series:
> > Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
>
> FWIW, I would suggest taking it a bit further: make syscall_trace_enter()
> and syscall_trace_leave() in m68k ptrace.c unconditional, replace the
> calls of syscall_trace() in entry.S with syscall_trace_enter() and
> syscall_trace_leave() resp. and remove syscall_trace().
>
> Geert, do you see any problems with that?  The only difference is that
> current->ptrace_message would be set to 1 for ptrace stop on entry and
> 2 - on leave.  Currently m68k just has it 0 all along.
>
> It is user-visible (the whole point is to let the tracer see which
> stop it is - entry or exit one), so somebody using PTRACE_GETEVENTMSG
> on syscall stops would start seeing 1 or 2 instead of "0 all along".
> That's how it works on all other architectures (including m68k-nommu),
> and I doubt that anything in userland will get broken.
>
> Behaviour of PTRACE_GETEVENTMSG for other stops (fork, etc.) remains
> as-is, of course.

In fact Michael did so in "[PATCH v7 1/2] m68k/kernel - wire up
syscall_trace_enter/leave for m68k"[1], but that's still stuck...

[1] https://lore.kernel.org/r/1624924520-17567-2-git-send-email-schmitzmic@gmail.com/

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Al Viro]

On Mon, Jan 10, 2022 at 06:54:57PM +0100, Geert Uytterhoeven wrote:

> In fact Michael did so in "[PATCH v7 1/2] m68k/kernel - wire up
> syscall_trace_enter/leave for m68k"[1], but that's still stuck...
> 
> [1] https://lore.kernel.org/r/1624924520-17567-2-git-send-email-schmitzmic@gmail.com/

Looks sane, but I'd split it in two - switch to calling syscall_trace_{enter,leave}
and then handling the return values...

The former would keep the current behaviour (modulo reporting enter vs. leave
via PTRACE_GETEVENTMSG), the latter would allow syscall number change by tracer
and/or handling of seccomp/audit/whatnot.

For exit+signal work the former would suffice, and IMO it would be a good idea
to put that one into a shared branch to be pulled both by seccomp and by signal
series.  Would reduce the conflicts...

Objections?

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Eric W. Biederman]

Al Viro <viro@zeniv.linux.org.uk> writes:

> On Mon, Jan 10, 2022 at 06:54:57PM +0100, Geert Uytterhoeven wrote:
>
>> In fact Michael did so in "[PATCH v7 1/2] m68k/kernel - wire up
>> syscall_trace_enter/leave for m68k"[1], but that's still stuck...
>> 
>> [1] https://lore.kernel.org/r/1624924520-17567-2-git-send-email-schmitzmic@gmail.com/
>
> Looks sane, but I'd split it in two - switch to calling syscall_trace_{enter,leave}
> and then handling the return values...
>
> The former would keep the current behaviour (modulo reporting enter vs. leave
> via PTRACE_GETEVENTMSG), the latter would allow syscall number change by tracer
> and/or handling of seccomp/audit/whatnot.
>
> For exit+signal work the former would suffice, and IMO it would be a good idea
> to put that one into a shared branch to be pulled both by seccomp and by signal
> series.  Would reduce the conflicts...
>
> Objections?

I have the version that Geert ack'ed queued up for v5.17 in my
signal-for-v5.17 branch, along with a couple others prior fixes in this
series of changes where it was clear they were just obviously correct
bug fixes.  No need to delay the removal of profiling bits for example.

I would love to see the m68k perform syscall_trace_{enter,leave} but
just getting as far as ptrace_report_syscall will be enough to avoid any
dependencies on my side.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Olivier Langlois]

On Mon, 2022-01-10 at 15:11 -0600, Eric W. Biederman wrote:
> 
> 
> I have been able to confirm that changing wait_event_interruptible to
> wait_event_killable was the culprit.  Something about the way
> systemd-coredump handles coredumps is not compatible with
> wait_event_killable.

This is my experience too that systemd-coredump is doing something
unexpected. When I tested the patch:
https://lore.kernel.org/lkml/cover.1629655338.git.olivier@trillion01.com/

to make sure that the patch worked, sending coredumps to systemd-
coredump was making systemd-coredump, well, core dump... Not very
useful...

Sending the dumps through a pipe to anything else than systemd-coredump
was working fine.

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Michael Schmitz]

Hi Geert,

Am 11.01.2022 um 06:54 schrieb Geert Uytterhoeven:
> Hi Al,
>
> CC Michael/m68k,
>
> On Mon, Jan 10, 2022 at 5:20 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
>> On Mon, Jan 10, 2022 at 04:26:57PM +0100, Geert Uytterhoeven wrote:
>>> On Mon, Jan 3, 2022 at 10:33 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>>>> The generic function ptrace_report_syscall does a little more
>>>> than syscall_trace on m68k.  The function ptrace_report_syscall
>>>> stops early if PT_TRACED is not set, it sets ptrace_message,
>>>> and returns the result of fatal_signal_pending.
>>>>
>>>> Setting ptrace_message to a passed in value of 0 is effectively not
>>>> setting ptrace_message, making that additional work a noop.
>>>>
>>>> Returning the result of fatal_signal_pending and letting the caller
>>>> ignore the result becomes a noop in this change.
>>>>
>>>> When a process is ptraced, the flag PT_PTRACED is always set in
>>>> current->ptrace.  Testing for PT_PTRACED in ptrace_report_syscall is
>>>> just an optimization to fail early if the process is not ptraced.
>>>> Later on in ptrace_notify, ptrace_stop will test current->ptrace under
>>>> tasklist_lock and skip performing any work if the task is not ptraced.
>>>>
>>>> Cc: Geert Uytterhoeven <geert@linux-m68k.org>
>>>> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
>>>
>>> As this depends on the removal of a parameter from
>>> ptrace_report_syscall() earlier in this series:
>>> Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
>>
>> FWIW, I would suggest taking it a bit further: make syscall_trace_enter()
>> and syscall_trace_leave() in m68k ptrace.c unconditional, replace the
>> calls of syscall_trace() in entry.S with syscall_trace_enter() and
>> syscall_trace_leave() resp. and remove syscall_trace().
>>
>> Geert, do you see any problems with that?  The only difference is that
>> current->ptrace_message would be set to 1 for ptrace stop on entry and
>> 2 - on leave.  Currently m68k just has it 0 all along.
>>
>> It is user-visible (the whole point is to let the tracer see which
>> stop it is - entry or exit one), so somebody using PTRACE_GETEVENTMSG
>> on syscall stops would start seeing 1 or 2 instead of "0 all along".
>> That's how it works on all other architectures (including m68k-nommu),
>> and I doubt that anything in userland will get broken.
>>
>> Behaviour of PTRACE_GETEVENTMSG for other stops (fork, etc.) remains
>> as-is, of course.
>
> In fact Michael did so in "[PATCH v7 1/2] m68k/kernel - wire up
> syscall_trace_enter/leave for m68k"[1], but that's still stuck...
>
> [1] https://lore.kernel.org/r/1624924520-17567-2-git-send-email-schmitzmic@gmail.com/

That patch (for reasons I never found out) did interact badly with 
Christoph Hellwig's 'remove set_fs' patches (and Al's signal fixes which 
Christoph's patches are based upon). Caused format errors under memory 
stress tests quite reliably, on my 030 hardware.

Probably needs a fresh look - the signal return path got changed by Al's 
patches IIRC, and I might have relied on offsets to data on the stack 
that are no longer correct with these patches. Or there's a race between 
the syscall trap and signal handling when returning from interrupt 
context ...

Still school hols over here so I won't have much peace and quiet until 
February.

Cheers,

	Michael


>
> Gr{oetje,eeting}s,
>
>                         Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds
>

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Dmitry Osipenko]

08.01.2022 21:13, Eric W. Biederman пишет:
> Dmitry Osipenko <digetx@gmail.com> writes:
> 
>> 05.01.2022 22:58, Eric W. Biederman пишет:
>>>
>>> I have not yet been able to figure out how to run gst-pluggin-scanner in
>>> a way that triggers this yet.  In truth I can't figure out how to
>>> run gst-pluggin-scanner in a useful way.
>>>
>>> I am going to set up some unit tests and see if I can reproduce your
>>> hang another way, but if you could give me some more information on what
>>> you are doing to trigger this I would appreciate it.
>>
>> Thanks, Eric. The distro is Arch Linux, but it's a development
>> environment where I'm running latest GStreamer from git master. I'll try
>> to figure out the reproduction steps and get back to you.
> 
> Thank you.
> 
> Until I can figure out why this is causing problems I have dropped the
> following two patches from my queue:
>  signal: Make SIGKILL during coredumps an explicit special case
>  signal: Drop signals received after a fatal signal has been processed
> 
> I have replaced them with the following two patches that just do what
> is needed for the rest of the code in the series:
>  signal: Have prepare_signal detect coredumps using
>  signal: Make coredump handling explicit in complete_signal
> 
> Perversely my failure to change the SIGKILL handling when coredumps are
> happening proves to me that I need to change the SIGKILL handling when
> coredumps are happening to make the code more maintainable.

Eric, thank you again. I started to look at the reproduction steps and
haven't completed it yet. Turned out the problem affects only older
NVIDIA Tegra2 Cortex-A9 CPU that lacks support of ARM NEON instructions
set, hence the problem isn't visible on x86 and other CPUs out of the
box. I'll need to check whether the problem could be simulated on all
arches or maybe it's specific to VFP exception handling of ARM32.

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Dmitry Osipenko <digetx@gmail.com> writes:

> 08.01.2022 21:13, Eric W. Biederman пишет:
>> Dmitry Osipenko <digetx@gmail.com> writes:
>> 
>>> 05.01.2022 22:58, Eric W. Biederman пишет:
>>>>
>>>> I have not yet been able to figure out how to run gst-pluggin-scanner in
>>>> a way that triggers this yet.  In truth I can't figure out how to
>>>> run gst-pluggin-scanner in a useful way.
>>>>
>>>> I am going to set up some unit tests and see if I can reproduce your
>>>> hang another way, but if you could give me some more information on what
>>>> you are doing to trigger this I would appreciate it.
>>>
>>> Thanks, Eric. The distro is Arch Linux, but it's a development
>>> environment where I'm running latest GStreamer from git master. I'll try
>>> to figure out the reproduction steps and get back to you.
>> 
>> Thank you.
>> 
>> Until I can figure out why this is causing problems I have dropped the
>> following two patches from my queue:
>>  signal: Make SIGKILL during coredumps an explicit special case
>>  signal: Drop signals received after a fatal signal has been processed
>> 
>> I have replaced them with the following two patches that just do what
>> is needed for the rest of the code in the series:
>>  signal: Have prepare_signal detect coredumps using
>>  signal: Make coredump handling explicit in complete_signal
>> 
>> Perversely my failure to change the SIGKILL handling when coredumps are
>> happening proves to me that I need to change the SIGKILL handling when
>> coredumps are happening to make the code more maintainable.
>
> Eric, thank you again. I started to look at the reproduction steps and
> haven't completed it yet. Turned out the problem affects only older
> NVIDIA Tegra2 Cortex-A9 CPU that lacks support of ARM NEON instructions
> set, hence the problem isn't visible on x86 and other CPUs out of the
> box. I'll need to check whether the problem could be simulated on all
> arches or maybe it's specific to VFP exception handling of ARM32.

It sounds like the gstreamer plugins only fail on certain hardware on
arm32, and things don't hang in coredumps unless the plugins fail.
That does make things tricky to minimize.

I have just verified that the known problematic code is not
in linux-next for Jan 11 2022.

If folks as they have time can double check linux-next and verify all is
well I would appreciate it.  I don't expect that there are problems but
sometimes one problem hides another.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Olivier Langlois <olivier@trillion01.com> writes:

> On Mon, 2022-01-10 at 15:11 -0600, Eric W. Biederman wrote:
>> 
>> 
>> I have been able to confirm that changing wait_event_interruptible to
>> wait_event_killable was the culprit.  Something about the way
>> systemd-coredump handles coredumps is not compatible with
>> wait_event_killable.
>
> This is my experience too that systemd-coredump is doing something
> unexpected. When I tested the patch:
> https://lore.kernel.org/lkml/cover.1629655338.git.olivier@trillion01.com/
>
> to make sure that the patch worked, sending coredumps to systemd-
> coredump was making systemd-coredump, well, core dump... Not very
> useful...

Oh.  Wow....

> Sending the dumps through a pipe to anything else than systemd-coredump
> was working fine.

Interesting.

I need to read through the pipe code and see how all of that works.  For
writing directly to disk only ignoring killable interruptions are the
usual semantics.  Ordinary pipe code has different semantics, and I
suspect that is what is tripping things up.

As for systemd-coredump it does whatever it does and I suspect some
versions of systemd-coredump are simply not robust if a coredump stops
unexpectedly.

The good news is the pipe code is simple enough, it will be possible to
completely read through that code.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

"Eric W. Biederman" <ebiederm@xmission.com> writes:

> Olivier Langlois <olivier@trillion01.com> writes:
>
>> On Mon, 2022-01-10 at 15:11 -0600, Eric W. Biederman wrote:
>>> 
>>> 
>>> I have been able to confirm that changing wait_event_interruptible to
>>> wait_event_killable was the culprit.  Something about the way
>>> systemd-coredump handles coredumps is not compatible with
>>> wait_event_killable.
>>
>> This is my experience too that systemd-coredump is doing something
>> unexpected. When I tested the patch:
>> https://lore.kernel.org/lkml/cover.1629655338.git.olivier@trillion01.com/
>>
>> to make sure that the patch worked, sending coredumps to systemd-
>> coredump was making systemd-coredump, well, core dump... Not very
>> useful...
>
> Oh.  Wow....
>
>> Sending the dumps through a pipe to anything else than systemd-coredump
>> was working fine.
>
> Interesting.
>
> I need to read through the pipe code and see how all of that works.  For
> writing directly to disk only ignoring killable interruptions are the
> usual semantics.  Ordinary pipe code has different semantics, and I
> suspect that is what is tripping things up.
>
> As for systemd-coredump it does whatever it does and I suspect some
> versions of systemd-coredump are simply not robust if a coredump stops
> unexpectedly.
>
> The good news is the pipe code is simple enough, it will be possible to
> completely read through that code.

My bug, obvious in hindsight is that "try_to_wait_up(TASK_INTERRUPTIBLE)"
does not work on a task that is in sleeping in TASK_KILLABLE.
That looks fixable in wait_for_dump_helpers it just won't be as easy
as changing wait_event_interruptible to wait_event_killable.

To prevent short pipe write from causing short writes during a coredump
I believe all we need to do handle -ERSTARTSYS with TIF_NOTIFY_SIGNAL.
Something like what I have below.

Until wait_for_dump_helpers is sorted out the coredump won't wait for
the dump helper the way it should, but otherwise things should work.

diff --git a/fs/coredump.c b/fs/coredump.c
index 7dece20b162b..0db1baf91420 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -796,6 +796,10 @@ static int __dump_emit(struct coredump_params *cprm, const void *addr, int nr)
 	if (dump_interrupted())
 		return 0;
 	n = __kernel_write(file, addr, nr, &pos);
+	while ((n == -ERESTARTSYS) && test_thread_flag(TIF_NOTIFY_SIGNAL)) {
+		tracehook_notify_signal();
+		n = __kernel_write(file, addr, nr, &pos);
+	}
 	if (n != nr)
 		return 0;
 	file->f_pos = pos;

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Linus Torvalds]

On Tue, Jan 11, 2022 at 10:51 AM Eric W. Biederman
<ebiederm@xmission.com> wrote:
>
> +       while ((n == -ERESTARTSYS) && test_thread_flag(TIF_NOTIFY_SIGNAL)) {
> +               tracehook_notify_signal();
> +               n = __kernel_write(file, addr, nr, &pos);
> +       }

This reads horribly wrongly to me.

That "tracehook_notify_signal()" thing *has* to be renamed before we
have anything like this that otherwise looks like "this will just loop
forever".

I'm pretty sure we've discussed that "tracehook" thing before - the
whole header file is misnamed, and most of the functions in theer are
too.

As an ugly alternative, open-code it, so that it's clear that "yup,
that clears the TIF_NOTIFY_SIGNAL flag".

             Linus

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Finn Thain]

On Tue, 11 Jan 2022, Michael Schmitz wrote:

> Am 11.01.2022 um 06:54 schrieb Geert Uytterhoeven:
> > Hi Al,
> >
> > CC Michael/m68k,
> >
> > On Mon, Jan 10, 2022 at 5:20 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
> >> On Mon, Jan 10, 2022 at 04:26:57PM +0100, Geert Uytterhoeven wrote:
> >>> On Mon, Jan 3, 2022 at 10:33 PM Eric W. Biederman <ebiederm@xmission.com>
> >>> wrote:
> >>>> The generic function ptrace_report_syscall does a little more
> >>>> than syscall_trace on m68k.  The function ptrace_report_syscall
> >>>> stops early if PT_TRACED is not set, it sets ptrace_message,
> >>>> and returns the result of fatal_signal_pending.
> >>>>
> >>>> Setting ptrace_message to a passed in value of 0 is effectively not
> >>>> setting ptrace_message, making that additional work a noop.
> >>>>
> >>>> Returning the result of fatal_signal_pending and letting the caller
> >>>> ignore the result becomes a noop in this change.
> >>>>
> >>>> When a process is ptraced, the flag PT_PTRACED is always set in
> >>>> current->ptrace.  Testing for PT_PTRACED in ptrace_report_syscall is
> >>>> just an optimization to fail early if the process is not ptraced.
> >>>> Later on in ptrace_notify, ptrace_stop will test current->ptrace under
> >>>> tasklist_lock and skip performing any work if the task is not ptraced.
> >>>>
> >>>> Cc: Geert Uytterhoeven <geert@linux-m68k.org>
> >>>> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> >>>
> >>> As this depends on the removal of a parameter from
> >>> ptrace_report_syscall() earlier in this series:
> >>> Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
> >>
> >> FWIW, I would suggest taking it a bit further: make syscall_trace_enter()
> >> and syscall_trace_leave() in m68k ptrace.c unconditional, replace the
> >> calls of syscall_trace() in entry.S with syscall_trace_enter() and
> >> syscall_trace_leave() resp. and remove syscall_trace().
> >>
> >> Geert, do you see any problems with that?  The only difference is that
> >> current->ptrace_message would be set to 1 for ptrace stop on entry and
> >> 2 - on leave.  Currently m68k just has it 0 all along.
> >>
> >> It is user-visible (the whole point is to let the tracer see which
> >> stop it is - entry or exit one), so somebody using PTRACE_GETEVENTMSG
> >> on syscall stops would start seeing 1 or 2 instead of "0 all along".
> >> That's how it works on all other architectures (including m68k-nommu),
> >> and I doubt that anything in userland will get broken.
> >>
> >> Behaviour of PTRACE_GETEVENTMSG for other stops (fork, etc.) remains
> >> as-is, of course.
> >
> > In fact Michael did so in "[PATCH v7 1/2] m68k/kernel - wire up
> > syscall_trace_enter/leave for m68k"[1], but that's still stuck...
> >
> > [1]
> > https://lore.kernel.org/r/1624924520-17567-2-git-send-email-schmitzmic@gmail.com/
> 
> That patch (for reasons I never found out) did interact badly with 
> Christoph Hellwig's 'remove set_fs' patches (and Al's signal fixes which 
> Christoph's patches are based upon). Caused format errors under memory 
> stress tests quite reliably, on my 030 hardware.
> 

Those patches have since been merged, BTW.

> Probably needs a fresh look - the signal return path got changed by Al's 
> patches IIRC, and I might have relied on offsets to data on the stack 
> that are no longer correct with these patches. Or there's a race between 
> the syscall trap and signal handling when returning from interrupt 
> context ...
> 
> Still school hols over here so I won't have much peace and quiet until 
> February.
> 

So the patch works okay with Aranym 68040 but not Motorola 68030? Since 
there is at least one known issue affecting both Motorola 68030 and Hatari 
68030, perhaps this patch is not the problem. In anycase, Al's suggestion 
to split the patch into two may help in that testing two smaller patches 
might narrow down the root cause.

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Michael Schmitz]

Hi Finn,

Am 12.01.2022 um 11:42 schrieb Finn Thain:
> On Tue, 11 Jan 2022, Michael Schmitz wrote:
>>> In fact Michael did so in "[PATCH v7 1/2] m68k/kernel - wire up
>>> syscall_trace_enter/leave for m68k"[1], but that's still stuck...
>>>
>>> [1]
>>> https://lore.kernel.org/r/1624924520-17567-2-git-send-email-schmitzmic@gmail.com/
>>
>> That patch (for reasons I never found out) did interact badly with
>> Christoph Hellwig's 'remove set_fs' patches (and Al's signal fixes which
>> Christoph's patches are based upon). Caused format errors under memory
>> stress tests quite reliably, on my 030 hardware.
>>
>
> Those patches have since been merged, BTW.

Yes, that's why I advised caution with mine.

>
>> Probably needs a fresh look - the signal return path got changed by Al's
>> patches IIRC, and I might have relied on offsets to data on the stack
>> that are no longer correct with these patches. Or there's a race between
>> the syscall trap and signal handling when returning from interrupt
>> context ...
>>
>> Still school hols over here so I won't have much peace and quiet until
>> February.
>>
>
> So the patch works okay with Aranym 68040 but not Motorola 68030? Since

Correct - I seem to recall we also tested those on your 040 and there 
was no regression there, but I may be misremembering that.

> there is at least one known issue affecting both Motorola 68030 and Hatari
> 68030, perhaps this patch is not the problem. In anycase, Al's suggestion

I hadn't ever made that connection, but it might be another explanation, 
yes.

> to split the patch into two may help in that testing two smaller patches
> might narrow down the root cause.

That's certainly true.

What's the other reason these patches are still stuck, Geert? Did we 
ever settle the dispute about what return code ought to abort a syscall 
(in the seccomp context)?

Cheers,

	Michael

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Finn Thain]

On Wed, 12 Jan 2022, Michael Schmitz wrote:

> 
> I seem to recall we also tested those on your 040 and there was no 
> regression there, but I may be misremembering that.
> 

I abandoned that regression testing exercise when unpatched mainline 
kernels began failing on that machine. I'm in the process of setting up a 
different 68040 machine.

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Michael Schmitz]

Hi Finn,

Am 12.01.2022 um 16:32 schrieb Finn Thain:
> On Wed, 12 Jan 2022, Michael Schmitz wrote:
>
>>
>> I seem to recall we also tested those on your 040 and there was no
>> regression there, but I may be misremembering that.
>>
>
> I abandoned that regression testing exercise when unpatched mainline
> kernels began failing on that machine. I'm in the process of setting up a
> different 68040 machine.
>

Thanks for refreshing my memory!

Splitting my first patch as suggested by Al in order to defer handling 
of the syscall_trace_enter() return code would achieve what Geert 
suggested (eliminate m68k syscall_trace() altogether) without risk of 
regression. This would need to replace Eric's patch 8.

Do you want me to send such a version based on my old patch series, or 
would you rather prepare that yourself, Eric?

Cheers,

	Michael

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Geert Uytterhoeven]

Hi Michael,

On Wed, Jan 12, 2022 at 1:20 AM Michael Schmitz <schmitzmic@gmail.com> wrote:
> Am 12.01.2022 um 11:42 schrieb Finn Thain:
> > On Tue, 11 Jan 2022, Michael Schmitz wrote:
> >>> In fact Michael did so in "[PATCH v7 1/2] m68k/kernel - wire up
> >>> syscall_trace_enter/leave for m68k"[1], but that's still stuck...
> >>>
> >>> [1]
> >>> https://lore.kernel.org/r/1624924520-17567-2-git-send-email-schmitzmic@gmail.com/
> >>
> >> That patch (for reasons I never found out) did interact badly with
> >> Christoph Hellwig's 'remove set_fs' patches (and Al's signal fixes which
> >> Christoph's patches are based upon). Caused format errors under memory
> >> stress tests quite reliably, on my 030 hardware.
> >>
> >
> > Those patches have since been merged, BTW.
>
> Yes, that's why I advised caution with mine.
>
> >
> >> Probably needs a fresh look - the signal return path got changed by Al's
> >> patches IIRC, and I might have relied on offsets to data on the stack
> >> that are no longer correct with these patches. Or there's a race between
> >> the syscall trap and signal handling when returning from interrupt
> >> context ...
> >>
> >> Still school hols over here so I won't have much peace and quiet until
> >> February.
> >>
> >
> > So the patch works okay with Aranym 68040 but not Motorola 68030? Since
>
> Correct - I seem to recall we also tested those on your 040 and there
> was no regression there, but I may be misremembering that.
>
> > there is at least one known issue affecting both Motorola 68030 and Hatari
> > 68030, perhaps this patch is not the problem. In anycase, Al's suggestion
>
> I hadn't ever made that connection, but it might be another explanation,
> yes.
>
> > to split the patch into two may help in that testing two smaller patches
> > might narrow down the root cause.
>
> That's certainly true.
>
> What's the other reason these patches are still stuck, Geert? Did we
> ever settle the dispute about what return code ought to abort a syscall
> (in the seccomp context)?

IIRC, some (self)tests were still failing?

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH 08/17] ptrace/m68k: Stop open coding ptrace_report_syscall

[Michael Schmitz]

Hi Geert,

Am 12.01.2022 um 20:55 schrieb Geert Uytterhoeven:
> Hi Michael,
>
>> What's the other reason these patches are still stuck, Geert? Did we
>> ever settle the dispute about what return code ought to abort a syscall
>> (in the seccomp context)?
>
> IIRC, some (self)tests were still failing?

Too true - but I don't think my way of building the testsuite was 
entirely according to the book. And I'm not sure I ran the testsuite 
with more than one of the return code options. In all honesty, I had 
been waiting for Adrian Glaubitz to test the patches with his seccomp 
library port instead of relying on the testsuite.

Still, reason enough to split off the removal of syscall_trace() from 
the seccomp stuff if it helps with Eric's patch series.

Cheers,

	Michael


>
> Gr{oetje,eeting}s,
>
>                         Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds
>

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Linus Torvalds <torvalds@linux-foundation.org> writes:

> On Tue, Jan 11, 2022 at 10:51 AM Eric W. Biederman
> <ebiederm@xmission.com> wrote:
>>
>> +       while ((n == -ERESTARTSYS) && test_thread_flag(TIF_NOTIFY_SIGNAL)) {
>> +               tracehook_notify_signal();
>> +               n = __kernel_write(file, addr, nr, &pos);
>> +       }
>
> This reads horribly wrongly to me.
>
> That "tracehook_notify_signal()" thing *has* to be renamed before we
> have anything like this that otherwise looks like "this will just loop
> forever".
>
> I'm pretty sure we've discussed that "tracehook" thing before - the
> whole header file is misnamed, and most of the functions in theer are
> too.
>
> As an ugly alternative, open-code it, so that it's clear that "yup,
> that clears the TIF_NOTIFY_SIGNAL flag".

A cleaner alternative looks like to modify the pipe code to use
wake_up_XXX instead of wake_up_interruptible_XXX and then have code
that does pipe_write_killable instead of pipe_write_interruptible.

There is also a question of how all of this should interact with the
freezer, as I think changing from interruptible to killable means that
the coredumps became unfreezable.

I am busily simmering this on my back burner and I hope I can come up
with something sensible.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Olivier Langlois]

On Fri, 2022-01-14 at 18:12 -0600, Eric W. Biederman wrote:
> Linus Torvalds <torvalds@linux-foundation.org> writes:
> 
> > On Tue, Jan 11, 2022 at 10:51 AM Eric W. Biederman
> > <ebiederm@xmission.com> wrote:
> > > 
> > > +       while ((n == -ERESTARTSYS) &&
> > > test_thread_flag(TIF_NOTIFY_SIGNAL)) {
> > > +               tracehook_notify_signal();
> > > +               n = __kernel_write(file, addr, nr, &pos);
> > > +       }
> > 
> > This reads horribly wrongly to me.
> > 
> > That "tracehook_notify_signal()" thing *has* to be renamed before
> > we
> > have anything like this that otherwise looks like "this will just
> > loop
> > forever".
> > 
> > I'm pretty sure we've discussed that "tracehook" thing before - the
> > whole header file is misnamed, and most of the functions in theer
> > are
> > too.
> > 
> > As an ugly alternative, open-code it, so that it's clear that "yup,
> > that clears the TIF_NOTIFY_SIGNAL flag".
> 
> A cleaner alternative looks like to modify the pipe code to use
> wake_up_XXX instead of wake_up_interruptible_XXX and then have code
> that does pipe_write_killable instead of pipe_write_interruptible.

Do not forget that the problem might not be limited to the pipe FS as
Oleg Nesterov pointed out here:

https://lore.kernel.org/io-uring/20210614141032.GA13677@redhat.com/

This is why I did like your patch fixing __dump_emit. If the only
problem is the tracehook_notify_signal() function unclear name, that
should be addressed instead of trying to fix the problem in a different
way.
> 
> There is also a question of how all of this should interact with the
> freezer, as I think changing from interruptible to killable means
> that
> the coredumps became unfreezable.
> 
> I am busily simmering this on my back burner and I hope I can come up
> with something sensible.

IMHO, fixing the problem on the emit function side has the merit of
being future proof if something else than io_uring in the future would
raise the TIF_NOTIFY_SIGNAL flag

but I am wondering why no one commented anything about my proposal of
cancelling io_uring before generating the core dump therefore stopping
it to flip TIF_NOTIFY_SIGNAL while the core dump is generated.

Is there something wrong with my proposed approach?
https://lore.kernel.org/lkml/cover.1629655338.git.olivier@trillion01.com/

It did flawlessly created many dozens of io_uring app core dumps in the
last months for me...

Olivier

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Olivier Langlois <olivier@trillion01.com> writes:

> On Fri, 2022-01-14 at 18:12 -0600, Eric W. Biederman wrote:
>> Linus Torvalds <torvalds@linux-foundation.org> writes:
>> 
>> > On Tue, Jan 11, 2022 at 10:51 AM Eric W. Biederman
>> > <ebiederm@xmission.com> wrote:
>> > > 
>> > > +       while ((n == -ERESTARTSYS) &&
>> > > test_thread_flag(TIF_NOTIFY_SIGNAL)) {
>> > > +               tracehook_notify_signal();
>> > > +               n = __kernel_write(file, addr, nr, &pos);
>> > > +       }
>> > 
>> > This reads horribly wrongly to me.
>> > 
>> > That "tracehook_notify_signal()" thing *has* to be renamed before
>> > we
>> > have anything like this that otherwise looks like "this will just
>> > loop
>> > forever".
>> > 
>> > I'm pretty sure we've discussed that "tracehook" thing before - the
>> > whole header file is misnamed, and most of the functions in theer
>> > are
>> > too.
>> > 
>> > As an ugly alternative, open-code it, so that it's clear that "yup,
>> > that clears the TIF_NOTIFY_SIGNAL flag".
>> 
>> A cleaner alternative looks like to modify the pipe code to use
>> wake_up_XXX instead of wake_up_interruptible_XXX and then have code
>> that does pipe_write_killable instead of pipe_write_interruptible.
>
> Do not forget that the problem might not be limited to the pipe FS as
> Oleg Nesterov pointed out here:
>
> https://lore.kernel.org/io-uring/20210614141032.GA13677@redhat.com/
>
> This is why I did like your patch fixing __dump_emit. If the only
> problem is the tracehook_notify_signal() function unclear name, that
> should be addressed instead of trying to fix the problem in a different
> way.

It might be that the fix is to run a portion of the exit_to_userspace
loop that does:

	if (ti_work & (_TIF_SIGPENDING | _TIF_NOTIFY_SIGNAL))
		handle_signal_work(regs, ti_work);

I am deep in brainstorm mode trying to find something that comes out
clean.

Oleg is right that while to be POSIX compliant and otherwise compatible
with traditional unix behavior sleeps in filesystems need to be
uninterruptible.  NFS has not always provided that compatibility.


>> There is also a question of how all of this should interact with the
>> freezer, as I think changing from interruptible to killable means
>> that
>> the coredumps became unfreezable.
>> 
>> I am busily simmering this on my back burner and I hope I can come up
>> with something sensible.
>
> IMHO, fixing the problem on the emit function side has the merit of
> being future proof if something else than io_uring in the future would
> raise the TIF_NOTIFY_SIGNAL flag
>
> but I am wondering why no one commented anything about my proposal of
> cancelling io_uring before generating the core dump therefore stopping
> it to flip TIF_NOTIFY_SIGNAL while the core dump is generated.
>
> Is there something wrong with my proposed approach?
> https://lore.kernel.org/lkml/cover.1629655338.git.olivier@trillion01.com/
>
> It did flawlessly created many dozens of io_uring app core dumps in the
> last months for me...

From my perspective I am not at all convinced that io_uring is the only
culprit.

Beyond that the purpose of a coredump is to snapshot the process as it
is, before anything is shutdown so that someone can examine the coredump
and figure out what failed.  Running around changing the state of the
process has a very real chance of hiding what is going wrong.

Further your change requires that there be a place for io_uring to clean
things up.  Given that fundamentally that seems like the wrong thing to
me I am not interested in making it easy to what looks like the wrong
thing.

All of this may be perfection being the enemy of the good (especially as
your io_uring magic happens as a special case in do_coredump).  My work
in this area is to remove hacks so I can be convinced the code works
100% of the time so unfortunately I am not interested in pick up a
change that is only good enough.  Someone else like Andrew Morton might
be.


None of that changes the fact that tracehook_notify_signal needs to be
renamed.  That effects your approach and my proof of concept approach.
So renaming tracehook_notify_signal just needs to be done.

Eric

io_uring truncating coredumps

[Eric W. Biederman]

Subject updated to reflect the current discussion.

> Linus Torvalds <torvalds@linux-foundation.org> writes:

> But I really think it's wrong.
> 
> You're trying to work around a problem the wrong way around. If a task
> is dead, and is dumping core, then signals just shouldn't matter in
> the first place, and thus the whole "TASK_INTERRUPTIBLE vs
> TASK_UNINTERRUPTIBLE" really shouldn't be an issue. The fact that it
> is an issue means there's something wrong in signaling, not in the
> pipe code.
> 
> So I really think that's where the fix should be - on the signal delivery side.

Thinking about it from the perspective of not delivering the wake-ups
fixing io_uring and coredumps in a non-hacky way looks comparatively
simple.  The function task_work_add just needs to not wake anything up
after a process has started dying.

Something like the patch below.

The only tricky part I can see is making certain there are not any races
between task_work_add and do_coredump depending on task_work_add not
causing signal_pending to return true.

diff --git a/kernel/task_work.c b/kernel/task_work.c
index fad745c59234..5f941e377268 100644
--- a/kernel/task_work.c
+++ b/kernel/task_work.c
@@ -44,6 +44,9 @@ int task_work_add(struct task_struct *task, struct callback_head *work,
 		work->next = head;
 	} while (cmpxchg(&task->task_works, head, work) != head);
 
+	if (notify && (task->signal->flags & SIGNAL_GROUP_EXIT))
+		return 0;
+
 	switch (notify) {
 	case TWA_NONE:
 		break;

Eric

Re: io_uring truncating coredumps

[Linus Torvalds]

On Mon, Jan 17, 2022 at 8:47 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>
> Thinking about it from the perspective of not delivering the wake-ups
> fixing io_uring and coredumps in a non-hacky way looks comparatively
> simple.  The function task_work_add just needs to not wake anything up
> after a process has started dying.
>
> Something like the patch below.

Hmm. Yes, I think this is the right direction.

That said, I think it should not add the work at all, and return
-ESRCH, the exact same way that it does for that work_exited
condition.

Because it's basically the same thing: the task is dead and shouldn't
do more work. In fact, task_work_run() is the thing that sets it to
&work_exited as it sees PF_EXITING, so it feels to me that THAT is
actually the issue here - we react to PF_EXITING too late. We react to
it *after* we've already added the work, and then we do that "no more
work" logic only after we've accepted those late work entries?

So my gut feel is that task_work_add() should just also test PF_EXITING.

And in fact, my gut feel is that PF_EXITING is too late anyway (it
happens after core-dumping, no?)

But I guess that thing may be on purpose, and maybe the act of dumping
core itself wants to do more work, and so that isn't an option?

So I don't think your patch is "right" as-is, and it all worries me,
but yes, I think this area is very much the questionable one.

I think that work stopping and the io_uring shutdown should probably
move earlier in the exit queue, but as mentioned above, maybe the work
addition boundary in particular really wants to be late because the
exit process itself still uses task works? ;(

              Linus

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Dmitry Osipenko]

11.01.2022 20:20, Eric W. Biederman пишет:
> Dmitry Osipenko <digetx@gmail.com> writes:
> 
>> 08.01.2022 21:13, Eric W. Biederman пишет:
>>> Dmitry Osipenko <digetx@gmail.com> writes:
>>>
>>>> 05.01.2022 22:58, Eric W. Biederman пишет:
>>>>>
>>>>> I have not yet been able to figure out how to run gst-pluggin-scanner in
>>>>> a way that triggers this yet.  In truth I can't figure out how to
>>>>> run gst-pluggin-scanner in a useful way.
>>>>>
>>>>> I am going to set up some unit tests and see if I can reproduce your
>>>>> hang another way, but if you could give me some more information on what
>>>>> you are doing to trigger this I would appreciate it.
>>>>
>>>> Thanks, Eric. The distro is Arch Linux, but it's a development
>>>> environment where I'm running latest GStreamer from git master. I'll try
>>>> to figure out the reproduction steps and get back to you.
>>>
>>> Thank you.
>>>
>>> Until I can figure out why this is causing problems I have dropped the
>>> following two patches from my queue:
>>>  signal: Make SIGKILL during coredumps an explicit special case
>>>  signal: Drop signals received after a fatal signal has been processed
>>>
>>> I have replaced them with the following two patches that just do what
>>> is needed for the rest of the code in the series:
>>>  signal: Have prepare_signal detect coredumps using
>>>  signal: Make coredump handling explicit in complete_signal
>>>
>>> Perversely my failure to change the SIGKILL handling when coredumps are
>>> happening proves to me that I need to change the SIGKILL handling when
>>> coredumps are happening to make the code more maintainable.
>>
>> Eric, thank you again. I started to look at the reproduction steps and
>> haven't completed it yet. Turned out the problem affects only older
>> NVIDIA Tegra2 Cortex-A9 CPU that lacks support of ARM NEON instructions
>> set, hence the problem isn't visible on x86 and other CPUs out of the
>> box. I'll need to check whether the problem could be simulated on all
>> arches or maybe it's specific to VFP exception handling of ARM32.
> 
> It sounds like the gstreamer plugins only fail on certain hardware on
> arm32, and things don't hang in coredumps unless the plugins fail.
> That does make things tricky to minimize.
> 
> I have just verified that the known problematic code is not
> in linux-next for Jan 11 2022.
> 
> If folks as they have time can double check linux-next and verify all is
> well I would appreciate it.  I don't expect that there are problems but
> sometimes one problem hides another.

Hello Eric,

I reproduced the trouble on x86_64.

Here are the reproduction steps, using ArchLinux and linux-next-20211224:

```
sudo pacman -S base-devel git mesa glu meson wget
git clone https://github.com/grate-driver/gstreamer.git
cd gstreamer
git checkout sigill
meson --prefix=/usr -Dgst-plugins-base:playback=enabled -Dgst-devtools:validate=disabled build
cd build
sudo ninja install
wget https://www.peach.themazzone.com/big_buck_bunny_720p_h264.mov
rm -r ~/.cache/gstreamer-1.0
gst-play-1.0 ./big_buck_bunny_720p_h264.mov
```

The SIGILL, thrown by [1], causes the hang. There is no hang using v5.16.1 kernel. 

[1] https://github.com/grate-driver/gstreamer/commit/006f9a2ee6dcf7b31c9b5413815d6054d82a3b2f

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Eric W. Biederman]

Dmitry Osipenko <digetx@gmail.com> writes:

> 11.01.2022 20:20, Eric W. Biederman пишет:
>> Dmitry Osipenko <digetx@gmail.com> writes:
>> 
>>> 08.01.2022 21:13, Eric W. Biederman пишет:
>>>> Dmitry Osipenko <digetx@gmail.com> writes:
>>>>
>>>>> 05.01.2022 22:58, Eric W. Biederman пишет:
>>>>>>
>>>>>> I have not yet been able to figure out how to run gst-pluggin-scanner in
>>>>>> a way that triggers this yet.  In truth I can't figure out how to
>>>>>> run gst-pluggin-scanner in a useful way.
>>>>>>
>>>>>> I am going to set up some unit tests and see if I can reproduce your
>>>>>> hang another way, but if you could give me some more information on what
>>>>>> you are doing to trigger this I would appreciate it.
>>>>>
>>>>> Thanks, Eric. The distro is Arch Linux, but it's a development
>>>>> environment where I'm running latest GStreamer from git master. I'll try
>>>>> to figure out the reproduction steps and get back to you.
>>>>
>>>> Thank you.
>>>>
>>>> Until I can figure out why this is causing problems I have dropped the
>>>> following two patches from my queue:
>>>>  signal: Make SIGKILL during coredumps an explicit special case
>>>>  signal: Drop signals received after a fatal signal has been processed
>>>>
>>>> I have replaced them with the following two patches that just do what
>>>> is needed for the rest of the code in the series:
>>>>  signal: Have prepare_signal detect coredumps using
>>>>  signal: Make coredump handling explicit in complete_signal
>>>>
>>>> Perversely my failure to change the SIGKILL handling when coredumps are
>>>> happening proves to me that I need to change the SIGKILL handling when
>>>> coredumps are happening to make the code more maintainable.
>>>
>>> Eric, thank you again. I started to look at the reproduction steps and
>>> haven't completed it yet. Turned out the problem affects only older
>>> NVIDIA Tegra2 Cortex-A9 CPU that lacks support of ARM NEON instructions
>>> set, hence the problem isn't visible on x86 and other CPUs out of the
>>> box. I'll need to check whether the problem could be simulated on all
>>> arches or maybe it's specific to VFP exception handling of ARM32.
>> 
>> It sounds like the gstreamer plugins only fail on certain hardware on
>> arm32, and things don't hang in coredumps unless the plugins fail.
>> That does make things tricky to minimize.
>> 
>> I have just verified that the known problematic code is not
>> in linux-next for Jan 11 2022.
>> 
>> If folks as they have time can double check linux-next and verify all is
>> well I would appreciate it.  I don't expect that there are problems but
>> sometimes one problem hides another.
>
> Hello Eric,
>
> I reproduced the trouble on x86_64.
>
> Here are the reproduction steps, using ArchLinux and linux-next-20211224:
>
> ```
> sudo pacman -S base-devel git mesa glu meson wget
> git clone https://github.com/grate-driver/gstreamer.git
> cd gstreamer
> git checkout sigill
> meson --prefix=/usr -Dgst-plugins-base:playback=enabled -Dgst-devtools:validate=disabled build
> cd build
> sudo ninja install
> wget https://www.peach.themazzone.com/big_buck_bunny_720p_h264.mov
> rm -r ~/.cache/gstreamer-1.0
> gst-play-1.0 ./big_buck_bunny_720p_h264.mov
> ```
>
> The SIGILL, thrown by [1], causes the hang. There is no hang using v5.16.1 kernel. 
>
> [1] https://github.com/grate-driver/gstreamer/commit/006f9a2ee6dcf7b31c9b5413815d6054d82a3b2f

Thank you.

I will verify this works before I add my updated version to
my signal-for-v5.18 branch.

Have you by any chance tried a newer version of linux-next without
commit fbc11520b58a ("signal: Make SIGKILL during coredumps an explicit
special case") in it?

If not I will double check that my pulling the commit out does not break
in the case you have documented.

Eric

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Dmitry Osipenko]

18.01.2022 20:52, Eric W. Biederman пишет:
> Dmitry Osipenko <digetx@gmail.com> writes:
> 
>> 11.01.2022 20:20, Eric W. Biederman пишет:
>>> Dmitry Osipenko <digetx@gmail.com> writes:
>>>
>>>> 08.01.2022 21:13, Eric W. Biederman пишет:
>>>>> Dmitry Osipenko <digetx@gmail.com> writes:
>>>>>
>>>>>> 05.01.2022 22:58, Eric W. Biederman пишет:
>>>>>>>
>>>>>>> I have not yet been able to figure out how to run gst-pluggin-scanner in
>>>>>>> a way that triggers this yet.  In truth I can't figure out how to
>>>>>>> run gst-pluggin-scanner in a useful way.
>>>>>>>
>>>>>>> I am going to set up some unit tests and see if I can reproduce your
>>>>>>> hang another way, but if you could give me some more information on what
>>>>>>> you are doing to trigger this I would appreciate it.
>>>>>>
>>>>>> Thanks, Eric. The distro is Arch Linux, but it's a development
>>>>>> environment where I'm running latest GStreamer from git master. I'll try
>>>>>> to figure out the reproduction steps and get back to you.
>>>>>
>>>>> Thank you.
>>>>>
>>>>> Until I can figure out why this is causing problems I have dropped the
>>>>> following two patches from my queue:
>>>>>  signal: Make SIGKILL during coredumps an explicit special case
>>>>>  signal: Drop signals received after a fatal signal has been processed
>>>>>
>>>>> I have replaced them with the following two patches that just do what
>>>>> is needed for the rest of the code in the series:
>>>>>  signal: Have prepare_signal detect coredumps using
>>>>>  signal: Make coredump handling explicit in complete_signal
>>>>>
>>>>> Perversely my failure to change the SIGKILL handling when coredumps are
>>>>> happening proves to me that I need to change the SIGKILL handling when
>>>>> coredumps are happening to make the code more maintainable.
>>>>
>>>> Eric, thank you again. I started to look at the reproduction steps and
>>>> haven't completed it yet. Turned out the problem affects only older
>>>> NVIDIA Tegra2 Cortex-A9 CPU that lacks support of ARM NEON instructions
>>>> set, hence the problem isn't visible on x86 and other CPUs out of the
>>>> box. I'll need to check whether the problem could be simulated on all
>>>> arches or maybe it's specific to VFP exception handling of ARM32.
>>>
>>> It sounds like the gstreamer plugins only fail on certain hardware on
>>> arm32, and things don't hang in coredumps unless the plugins fail.
>>> That does make things tricky to minimize.
>>>
>>> I have just verified that the known problematic code is not
>>> in linux-next for Jan 11 2022.
>>>
>>> If folks as they have time can double check linux-next and verify all is
>>> well I would appreciate it.  I don't expect that there are problems but
>>> sometimes one problem hides another.
>>
>> Hello Eric,
>>
>> I reproduced the trouble on x86_64.
>>
>> Here are the reproduction steps, using ArchLinux and linux-next-20211224:
>>
>> ```
>> sudo pacman -S base-devel git mesa glu meson wget
>> git clone https://github.com/grate-driver/gstreamer.git
>> cd gstreamer
>> git checkout sigill
>> meson --prefix=/usr -Dgst-plugins-base:playback=enabled -Dgst-devtools:validate=disabled build
>> cd build
>> sudo ninja install
>> wget https://www.peach.themazzone.com/big_buck_bunny_720p_h264.mov
>> rm -r ~/.cache/gstreamer-1.0
>> gst-play-1.0 ./big_buck_bunny_720p_h264.mov
>> ```
>>
>> The SIGILL, thrown by [1], causes the hang. There is no hang using v5.16.1 kernel. 
>>
>> [1] https://github.com/grate-driver/gstreamer/commit/006f9a2ee6dcf7b31c9b5413815d6054d82a3b2f
> 
> Thank you.
> 
> I will verify this works before I add my updated version to
> my signal-for-v5.18 branch.
> 
> Have you by any chance tried a newer version of linux-next without
> commit fbc11520b58a ("signal: Make SIGKILL during coredumps an explicit
> special case") in it?
> 
> If not I will double check that my pulling the commit out does not break
> in the case you have documented.

Recent linux-next works fine.

Re: [PATCH 1/8] signal: Make SIGKILL during coredumps an explicit special case

[Olivier Langlois]

On Mon, 2022-01-17 at 10:09 -0600, Eric W. Biederman wrote:
> Olivier Langlois <olivier@trillion01.com> writes:
> From my perspective I am not at all convinced that io_uring is the
> only
> culprit.
> 
> Beyond that the purpose of a coredump is to snapshot the process as
> it
> is, before anything is shutdown so that someone can examine the
> coredump
> and figure out what failed.  Running around changing the state of the
> process has a very real chance of hiding what is going wrong.
> 
> Further your change requires that there be a place for io_uring to
> clean
> things up.  Given that fundamentally that seems like the wrong thing
> to
> me I am not interested in making it easy to what looks like the wrong
> thing.
> 
> All of this may be perfection being the enemy of the good (especially
> as
> your io_uring magic happens as a special case in do_coredump).  My
> work
> in this area is to remove hacks so I can be convinced the code works
> 100% of the time so unfortunately I am not interested in pick up a
> change that is only good enough.  Someone else like Andrew Morton
> might
> be.
> 
> 
Fair enough.

You do bring good points but I am not so sure about the second one
considering that the coredump is meant to be a snapshot and if io_uring
still runs, the state may change as the dump is generated anyway.

I'll follow with interest what you finally come up with but my mindset
when I wrote the patch was that there does not seem to be any benefit
keeping io_uring active while coredumping and it has the potential to
create nasty issues.

I did stumble into core file truncation problem.

Pavel got that when modifying io_uring code:
https://lore.kernel.org/all/1b519092-2ebf-3800-306d-c354c24a9ad1@gmail.com/

and I find very likely that keeping io_uring active while coredumping
might create new nasty but subtle issues down the road...

Greetings,
Olivier

[PATCH 00/13] Removing tracehook.h

[Eric W. Biederman]

While working on cleaning up do_exit I have been having to deal with the
code in tracehook.h.  Unfortunately the code in tracehook.h does not
make sense as organized.

This set of changes reorganizes things so that tracehook.h no longer
exists, and so that it's current contents are organized in a fashion
that is a little easier to understand.

The biggest change is that I lean into the fact that get_signal
always calls task_work_run and removes the logic that tried to
be smart and decouple task_work_run and get_signal as it has proven
to not be effective.

This is a conservative change and I am not changing the how things
like signal_pending operate (although it is probably justified).

A new header resume_user_mode.h is added to hold resume_user_mode_work
which was previously known as tracehook_notify_resume.

Eric W. Biederman (13):
      ptrace: Move ptrace_report_syscall into ptrace.h
      ptrace/arm: Rename tracehook_report_syscall report_syscall
      ptrace: Create ptrace_report_syscall_{entry,exit} in ptrace.h
      ptrace: Remove arch_syscall_{enter,exit}_tracehook
      ptrace: Remove tracehook_signal_handler
      task_work: Remove unnecessary include from posix_timers.h
      task_work: Introduce task_work_pending
      task_work: Call tracehook_notify_signal from get_signal on all architectures
      task_work: Decouple TIF_NOTIFY_SIGNAL and task_work
      signal: Move set_notify_signal and clear_notify_signal into sched/signal.h
      resume_user_mode: Remove #ifdef TIF_NOTIFY_RESUME in set_notify_resume
      resume_user_mode: Move to resume_user_mode.h
      tracehook: Remove tracehook.h

 MAINTAINERS                          |   1 -
 arch/Kconfig                         |   5 +-
 arch/alpha/kernel/ptrace.c           |   5 +-
 arch/alpha/kernel/signal.c           |   4 +-
 arch/arc/kernel/ptrace.c             |   5 +-
 arch/arc/kernel/signal.c             |   4 +-
 arch/arm/kernel/ptrace.c             |  12 +-
 arch/arm/kernel/signal.c             |   4 +-
 arch/arm64/kernel/ptrace.c           |  14 +--
 arch/arm64/kernel/signal.c           |   4 +-
 arch/csky/kernel/ptrace.c            |   5 +-
 arch/csky/kernel/signal.c            |   4 +-
 arch/h8300/kernel/ptrace.c           |   5 +-
 arch/h8300/kernel/signal.c           |   4 +-
 arch/hexagon/kernel/process.c        |   4 +-
 arch/hexagon/kernel/signal.c         |   1 -
 arch/hexagon/kernel/traps.c          |   6 +-
 arch/ia64/kernel/process.c           |   4 +-
 arch/ia64/kernel/ptrace.c            |   6 +-
 arch/ia64/kernel/signal.c            |   1 -
 arch/m68k/kernel/ptrace.c            |   6 +-
 arch/m68k/kernel/signal.c            |   4 +-
 arch/microblaze/kernel/ptrace.c      |   5 +-
 arch/microblaze/kernel/signal.c      |   4 +-
 arch/mips/kernel/ptrace.c            |   5 +-
 arch/mips/kernel/signal.c            |   4 +-
 arch/nds32/include/asm/syscall.h     |   2 +-
 arch/nds32/kernel/ptrace.c           |   5 +-
 arch/nds32/kernel/signal.c           |   4 +-
 arch/nios2/kernel/ptrace.c           |   5 +-
 arch/nios2/kernel/signal.c           |   4 +-
 arch/openrisc/kernel/ptrace.c        |   5 +-
 arch/openrisc/kernel/signal.c        |   4 +-
 arch/parisc/kernel/ptrace.c          |   7 +-
 arch/parisc/kernel/signal.c          |   4 +-
 arch/powerpc/kernel/ptrace/ptrace.c  |   8 +-
 arch/powerpc/kernel/signal.c         |   4 +-
 arch/riscv/kernel/ptrace.c           |   5 +-
 arch/riscv/kernel/signal.c           |   4 +-
 arch/s390/include/asm/entry-common.h |   1 -
 arch/s390/kernel/ptrace.c            |   1 -
 arch/s390/kernel/signal.c            |   5 +-
 arch/sh/kernel/ptrace_32.c           |   5 +-
 arch/sh/kernel/signal_32.c           |   4 +-
 arch/sparc/kernel/ptrace_32.c        |   5 +-
 arch/sparc/kernel/ptrace_64.c        |   5 +-
 arch/sparc/kernel/signal32.c         |   1 -
 arch/sparc/kernel/signal_32.c        |   4 +-
 arch/sparc/kernel/signal_64.c        |   4 +-
 arch/um/kernel/process.c             |   4 +-
 arch/um/kernel/ptrace.c              |   5 +-
 arch/x86/kernel/ptrace.c             |   1 -
 arch/x86/kernel/signal.c             |   5 +-
 arch/x86/mm/tlb.c                    |   1 +
 arch/xtensa/kernel/ptrace.c          |   5 +-
 arch/xtensa/kernel/signal.c          |   4 +-
 block/blk-cgroup.c                   |   2 +-
 fs/coredump.c                        |   1 -
 fs/exec.c                            |   1 -
 fs/io-wq.c                           |   6 +-
 fs/io_uring.c                        |  11 +-
 fs/proc/array.c                      |   1 -
 fs/proc/base.c                       |   1 -
 include/asm-generic/syscall.h        |   2 +-
 include/linux/entry-common.h         |  47 +-------
 include/linux/entry-kvm.h            |   2 +-
 include/linux/posix-timers.h         |   1 -
 include/linux/ptrace.h               |  78 ++++++++++++
 include/linux/resume_user_mode.h     |  64 ++++++++++
 include/linux/sched/signal.h         |  17 +++
 include/linux/task_work.h            |   5 +
 include/linux/tracehook.h            | 226 -----------------------------------
 include/uapi/linux/ptrace.h          |   2 +-
 kernel/entry/common.c                |  19 +--
 kernel/entry/kvm.c                   |   9 +-
 kernel/exit.c                        |   3 +-
 kernel/livepatch/transition.c        |   1 -
 kernel/seccomp.c                     |   1 -
 kernel/signal.c                      |  23 ++--
 kernel/task_work.c                   |   4 +-
 kernel/time/posix-cpu-timers.c       |   1 +
 mm/memcontrol.c                      |   2 +-
 security/apparmor/domain.c           |   1 -
 security/selinux/hooks.c             |   1 -
 84 files changed, 317 insertions(+), 462 deletions(-)

Eric

[PATCH 00/13] Removing tracehook.h

[Eric W. Biederman]

While working on cleaning up do_exit I have been having to deal with the
code in tracehook.h.  Unfortunately the code in tracehook.h does not
make sense as organized.

This set of changes reorganizes things so that tracehook.h no longer
exists, and so that it's current contents are organized in a fashion
that is a little easier to understand.

The biggest change is that I lean into the fact that get_signal
always calls task_work_run and removes the logic that tried to
be smart and decouple task_work_run and get_signal as it has proven
to not be effective.

This is a conservative change and I am not changing the how things
like signal_pending operate (although it is probably justified).

A new header resume_user_mode.h is added to hold resume_user_mode_work
which was previously known as tracehook_notify_resume.

Eric W. Biederman (13):
      ptrace: Move ptrace_report_syscall into ptrace.h
      ptrace/arm: Rename tracehook_report_syscall report_syscall
      ptrace: Create ptrace_report_syscall_{entry,exit} in ptrace.h
      ptrace: Remove arch_syscall_{enter,exit}_tracehook
      ptrace: Remove tracehook_signal_handler
      task_work: Remove unnecessary include from posix_timers.h
      task_work: Introduce task_work_pending
      task_work: Call tracehook_notify_signal from get_signal on all architectures
      task_work: Decouple TIF_NOTIFY_SIGNAL and task_work
      signal: Move set_notify_signal and clear_notify_signal into sched/signal.h
      resume_user_mode: Remove #ifdef TIF_NOTIFY_RESUME in set_notify_resume
      resume_user_mode: Move to resume_user_mode.h
      tracehook: Remove tracehook.h

 MAINTAINERS                          |   1 -
 arch/Kconfig                         |   5 +-
 arch/alpha/kernel/ptrace.c           |   5 +-
 arch/alpha/kernel/signal.c           |   4 +-
 arch/arc/kernel/ptrace.c             |   5 +-
 arch/arc/kernel/signal.c             |   4 +-
 arch/arm/kernel/ptrace.c             |  12 +-
 arch/arm/kernel/signal.c             |   4 +-
 arch/arm64/kernel/ptrace.c           |  14 +--
 arch/arm64/kernel/signal.c           |   4 +-
 arch/csky/kernel/ptrace.c            |   5 +-
 arch/csky/kernel/signal.c            |   4 +-
 arch/h8300/kernel/ptrace.c           |   5 +-
 arch/h8300/kernel/signal.c           |   4 +-
 arch/hexagon/kernel/process.c        |   4 +-
 arch/hexagon/kernel/signal.c         |   1 -
 arch/hexagon/kernel/traps.c          |   6 +-
 arch/ia64/kernel/process.c           |   4 +-
 arch/ia64/kernel/ptrace.c            |   6 +-
 arch/ia64/kernel/signal.c            |   1 -
 arch/m68k/kernel/ptrace.c            |   6 +-
 arch/m68k/kernel/signal.c            |   4 +-
 arch/microblaze/kernel/ptrace.c      |   5 +-
 arch/microblaze/kernel/signal.c      |   4 +-
 arch/mips/kernel/ptrace.c            |   5 +-
 arch/mips/kernel/signal.c            |   4 +-
 arch/nds32/include/asm/syscall.h     |   2 +-
 arch/nds32/kernel/ptrace.c           |   5 +-
 arch/nds32/kernel/signal.c           |   4 +-
 arch/nios2/kernel/ptrace.c           |   5 +-
 arch/nios2/kernel/signal.c           |   4 +-
 arch/openrisc/kernel/ptrace.c        |   5 +-
 arch/openrisc/kernel/signal.c        |   4 +-
 arch/parisc/kernel/ptrace.c          |   7 +-
 arch/parisc/kernel/signal.c          |   4 +-
 arch/powerpc/kernel/ptrace/ptrace.c  |   8 +-
 arch/powerpc/kernel/signal.c         |   4 +-
 arch/riscv/kernel/ptrace.c           |   5 +-
 arch/riscv/kernel/signal.c           |   4 +-
 arch/s390/include/asm/entry-common.h |   1 -
 arch/s390/kernel/ptrace.c            |   1 -
 arch/s390/kernel/signal.c            |   5 +-
 arch/sh/kernel/ptrace_32.c           |   5 +-
 arch/sh/kernel/signal_32.c           |   4 +-
 arch/sparc/kernel/ptrace_32.c        |   5 +-
 arch/sparc/kernel/ptrace_64.c        |   5 +-
 arch/sparc/kernel/signal32.c         |   1 -
 arch/sparc/kernel/signal_32.c        |   4 +-
 arch/sparc/kernel/signal_64.c        |   4 +-
 arch/um/kernel/process.c             |   4 +-
 arch/um/kernel/ptrace.c              |   5 +-
 arch/x86/kernel/ptrace.c             |   1 -
 arch/x86/kernel/signal.c             |   5 +-
 arch/x86/mm/tlb.c                    |   1 +
 arch/xtensa/kernel/ptrace.c          |   5 +-
 arch/xtensa/kernel/signal.c          |   4 +-
 block/blk-cgroup.c                   |   2 +-
 fs/coredump.c                        |   1 -
 fs/exec.c                            |   1 -
 fs/io-wq.c                           |   6 +-
 fs/io_uring.c                        |  11 +-
 fs/proc/array.c                      |   1 -
 fs/proc/base.c                       |   1 -
 include/asm-generic/syscall.h        |   2 +-
 include/linux/entry-common.h         |  47 +-------
 include/linux/entry-kvm.h            |   2 +-
 include/linux/posix-timers.h         |   1 -
 include/linux/ptrace.h               |  78 ++++++++++++
 include/linux/resume_user_mode.h     |  64 ++++++++++
 include/linux/sched/signal.h         |  17 +++
 include/linux/task_work.h            |   5 +
 include/linux/tracehook.h            | 226 -----------------------------------
 include/uapi/linux/ptrace.h          |   2 +-
 kernel/entry/common.c                |  19 +--
 kernel/entry/kvm.c                   |   9 +-
 kernel/exit.c                        |   3 +-
 kernel/livepatch/transition.c        |   1 -
 kernel/seccomp.c                     |   1 -
 kernel/signal.c                      |  23 ++--
 kernel/task_work.c                   |   4 +-
 kernel/time/posix-cpu-timers.c       |   1 +
 mm/memcontrol.c                      |   2 +-
 security/apparmor/domain.c           |   1 -
 security/selinux/hooks.c             |   1 -
 84 files changed, 317 insertions(+), 462 deletions(-)

Eric

Re: [PATCH 00/13] Removing tracehook.h

[Linus Torvalds]

On Tue, Mar 8, 2022 at 4:16 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>
> While working on cleaning up do_exit I have been having to deal with the
> code in tracehook.h.  Unfortunately the code in tracehook.h does not
> make sense as organized. [...]

Thanks, that odd naming has tripped me up several times, this looks
like an improvement.

                  Linus

Re: [PATCH 00/13] Removing tracehook.h

[Jens Axboe]

On 3/8/22 5:13 PM, Eric W. Biederman wrote:
> 
> While working on cleaning up do_exit I have been having to deal with the
> code in tracehook.h.  Unfortunately the code in tracehook.h does not
> make sense as organized.
> 
> This set of changes reorganizes things so that tracehook.h no longer
> exists, and so that it's current contents are organized in a fashion
> that is a little easier to understand.
> 
> The biggest change is that I lean into the fact that get_signal
> always calls task_work_run and removes the logic that tried to
> be smart and decouple task_work_run and get_signal as it has proven
> to not be effective.
> 
> This is a conservative change and I am not changing the how things
> like signal_pending operate (although it is probably justified).
> 
> A new header resume_user_mode.h is added to hold resume_user_mode_work
> which was previously known as tracehook_notify_resume.

This is a nice cleanup. I did bother me adding the TIF_NOTIFY_SIGNAL
bits and work hooks to something unrelated, but that's where other
things resided then. This makes it a lot better.

-- 
Jens Axboe

[PATCH 0/2] ptrace: Making the ptrace changes atomic

[Eric W. Biederman]

While working on cleaning up the exit path it have had occasion to look
at what guarantees are provided for setting and reading the fields that
are provided in task_struct for ptraces.  Namely exit_code,
ptrace_message, and last_siginfo. It turns out as the ptrace interface
in the kernel was extended in the kernel the old existing interfaces
in the kernel were just wrapped and not properly updated to handle
the new functionality.  This lead to races and inconsistencies.

This fixes the reason for the races and inconsistencies by moving the
work of maintaining the ptrace fields into ptrace_stop.

The inconsistency that results in some ptrace_stop points continuing
with a signal while others will not I have left alone as it appears to
be part of our userspace ABI, and changing that risks breaking
userspace.

Eric W. Biederman (2):
      ptrace: Move setting/clearing ptrace_message into ptrace_stop
      ptrace: Return the signal to continue with from ptrace_stop

 include/linux/ptrace.h      | 17 +++++++----------
 include/uapi/linux/ptrace.h |  2 +-
 kernel/signal.c             | 40 ++++++++++++++++++++++++----------------
 3 files changed, 32 insertions(+), 27 deletions(-)

Eric

[PATCH 1/2] ptrace: Move setting/clearing ptrace_message into ptrace_stop

[Eric W. Biederman]

Today ptrace_message is easy to overlook as it not a core part of
ptrace_stop.  It has been overlooked so much that there are places
that set ptrace_message and don't clear it, and places that never set
it.  So if you get an unlucky sequence of events the ptracer may be
able to read a ptrace_message that does not apply to the current
ptrace stop.

Move setting of ptrace_message into ptrace_stop so that it always gets
set before the stop, and always gets cleared after the stop.  This
prevents non-sense from being reported to userspace and makes
ptrace_message more visible in the ptrace helper functions so that
kernel developers can see it.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 include/linux/ptrace.h      |  9 +++------
 include/uapi/linux/ptrace.h |  2 +-
 kernel/signal.c             | 21 ++++++++++++---------
 3 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
index 5310f43e4762..3e6b46e2b7be 100644
--- a/include/linux/ptrace.h
+++ b/include/linux/ptrace.h
@@ -60,7 +60,7 @@ extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned
 extern void ptrace_disable(struct task_struct *);
 extern int ptrace_request(struct task_struct *child, long request,
 			  unsigned long addr, unsigned long data);
-extern void ptrace_notify(int exit_code);
+extern void ptrace_notify(int exit_code, unsigned long message);
 extern void __ptrace_link(struct task_struct *child,
 			  struct task_struct *new_parent,
 			  const struct cred *ptracer_cred);
@@ -155,8 +155,7 @@ static inline bool ptrace_event_enabled(struct task_struct *task, int event)
 static inline void ptrace_event(int event, unsigned long message)
 {
 	if (unlikely(ptrace_event_enabled(current, event))) {
-		current->ptrace_message = message;
-		ptrace_notify((event << 8) | SIGTRAP);
+		ptrace_notify((event << 8) | SIGTRAP, message);
 	} else if (event == PTRACE_EVENT_EXEC) {
 		/* legacy EXEC report via SIGTRAP */
 		if ((current->ptrace & (PT_PTRACED|PT_SEIZED)) == PT_PTRACED)
@@ -424,8 +423,7 @@ static inline int ptrace_report_syscall(unsigned long message)
 	if (!(ptrace & PT_PTRACED))
 		return 0;
 
-	current->ptrace_message = message;
-	ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0));
+	ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0), message);
 
 	/*
 	 * this isn't the same as continuing with a signal, but it will do
@@ -437,7 +435,6 @@ static inline int ptrace_report_syscall(unsigned long message)
 		current->exit_code = 0;
 	}
 
-	current->ptrace_message = 0;
 	return fatal_signal_pending(current);
 }
 
diff --git a/include/uapi/linux/ptrace.h b/include/uapi/linux/ptrace.h
index b7af92e07d1f..195ae64a8c87 100644
--- a/include/uapi/linux/ptrace.h
+++ b/include/uapi/linux/ptrace.h
@@ -114,7 +114,7 @@ struct ptrace_rseq_configuration {
 
 /*
  * These values are stored in task->ptrace_message
- * by ptrace_report_syscall_* to describe the current syscall-stop.
+ * by ptrace_stop to describe the current syscall-stop.
  */
 #define PTRACE_EVENTMSG_SYSCALL_ENTRY	1
 #define PTRACE_EVENTMSG_SYSCALL_EXIT	2
diff --git a/kernel/signal.c b/kernel/signal.c
index c2dee5420567..a49ac7149256 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2191,7 +2191,8 @@ static void do_notify_parent_cldstop(struct task_struct *tsk,
  * If we actually decide not to stop at all because the tracer
  * is gone, we keep current->exit_code unless clear_code.
  */
-static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
+static void ptrace_stop(int exit_code, int why, int clear_code,
+			unsigned long message, kernel_siginfo_t *info)
 	__releases(&current->sighand->siglock)
 	__acquires(&current->sighand->siglock)
 {
@@ -2237,6 +2238,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t
 	 */
 	smp_wmb();
 
+	current->ptrace_message = message;
 	current->last_siginfo = info;
 	current->exit_code = exit_code;
 
@@ -2315,6 +2317,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t
 	 */
 	spin_lock_irq(&current->sighand->siglock);
 	current->last_siginfo = NULL;
+	current->ptrace_message = 0;
 
 	/* LISTENING can be set only during STOP traps, clear it */
 	current->jobctl &= ~JOBCTL_LISTENING;
@@ -2327,7 +2330,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t
 	recalc_sigpending_tsk(current);
 }
 
-static void ptrace_do_notify(int signr, int exit_code, int why)
+static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
 {
 	kernel_siginfo_t info;
 
@@ -2338,17 +2341,17 @@ static void ptrace_do_notify(int signr, int exit_code, int why)
 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
 
 	/* Let the debugger run.  */
-	ptrace_stop(exit_code, why, 1, &info);
+	ptrace_stop(exit_code, why, 1, message, &info);
 }
 
-void ptrace_notify(int exit_code)
+void ptrace_notify(int exit_code, unsigned long message)
 {
 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
 	if (unlikely(task_work_pending(current)))
 		task_work_run();
 
 	spin_lock_irq(&current->sighand->siglock);
-	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
+	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
 	spin_unlock_irq(&current->sighand->siglock);
 }
 
@@ -2504,10 +2507,10 @@ static void do_jobctl_trap(void)
 			signr = SIGTRAP;
 		WARN_ON_ONCE(!signr);
 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
-				 CLD_STOPPED);
+				 CLD_STOPPED, 0);
 	} else {
 		WARN_ON_ONCE(!signr);
-		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
+		ptrace_stop(signr, CLD_STOPPED, 0, 0, NULL);
 		current->exit_code = 0;
 	}
 }
@@ -2561,7 +2564,7 @@ static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
 	 * comment in dequeue_signal().
 	 */
 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
-	ptrace_stop(signr, CLD_TRAPPED, 0, info);
+	ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
 
 	/* We're back.  Did the debugger cancel the sig?  */
 	signr = current->exit_code;
@@ -2891,7 +2894,7 @@ static void signal_delivered(struct ksignal *ksig, int stepping)
 	if (current->sas_ss_flags & SS_AUTODISARM)
 		sas_ss_reset(current);
 	if (stepping)
-		ptrace_notify(SIGTRAP);
+		ptrace_notify(SIGTRAP, 0);
 }
 
 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
-- 
2.29.2

[PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Eric W. Biederman]

The signal a task should continue with after a ptrace stop is
inconsistently read, cleared, and sent.  Solve this by reading and
clearing the signal to be sent in ptrace_stop.

In an ideal world everything except ptrace_signal would share a common
implementation of continuing with the signal, so ptracers could count
on the signal they ask to continue with actually being delivered.  For
now retain bug compatibility and just return with the signal number
the ptracer requested the code continue with.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
---
 include/linux/ptrace.h | 12 ++++++------
 kernel/signal.c        | 31 ++++++++++++++++++-------------
 2 files changed, 24 insertions(+), 19 deletions(-)

diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
index 3e6b46e2b7be..15b3d176b6b4 100644
--- a/include/linux/ptrace.h
+++ b/include/linux/ptrace.h
@@ -60,7 +60,7 @@ extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned
 extern void ptrace_disable(struct task_struct *);
 extern int ptrace_request(struct task_struct *child, long request,
 			  unsigned long addr, unsigned long data);
-extern void ptrace_notify(int exit_code, unsigned long message);
+extern int ptrace_notify(int exit_code, unsigned long message);
 extern void __ptrace_link(struct task_struct *child,
 			  struct task_struct *new_parent,
 			  const struct cred *ptracer_cred);
@@ -419,21 +419,21 @@ extern void sigaction_compat_abi(struct k_sigaction *act, struct k_sigaction *oa
 static inline int ptrace_report_syscall(unsigned long message)
 {
 	int ptrace = current->ptrace;
+	int signr;
 
 	if (!(ptrace & PT_PTRACED))
 		return 0;
 
-	ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0), message);
+	signr = ptrace_notify(SIGTRAP | ((ptrace & PT_TRACESYSGOOD) ? 0x80 : 0),
+			      message);
 
 	/*
 	 * this isn't the same as continuing with a signal, but it will do
 	 * for normal use.  strace only continues with a signal if the
 	 * stopping signal is not SIGTRAP.  -brl
 	 */
-	if (current->exit_code) {
-		send_sig(current->exit_code, current, 1);
-		current->exit_code = 0;
-	}
+	if (signr)
+		send_sig(signr, current, 1);
 
 	return fatal_signal_pending(current);
 }
diff --git a/kernel/signal.c b/kernel/signal.c
index a49ac7149256..e53ee84b9021 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2188,15 +2188,17 @@ static void do_notify_parent_cldstop(struct task_struct *tsk,
  * That makes it a way to test a stopped process for
  * being ptrace-stopped vs being job-control-stopped.
  *
- * If we actually decide not to stop at all because the tracer
- * is gone, we keep current->exit_code unless clear_code.
+ * Returns the signal the ptracer requested the code resume
+ * with.  If the code did not stop because the tracer is gone,
+ * the stop signal remains unchanged unless clear_code.
  */
-static void ptrace_stop(int exit_code, int why, int clear_code,
+static int ptrace_stop(int exit_code, int why, int clear_code,
 			unsigned long message, kernel_siginfo_t *info)
 	__releases(&current->sighand->siglock)
 	__acquires(&current->sighand->siglock)
 {
 	bool gstop_done = false;
+	bool read_code = true;
 
 	if (arch_ptrace_stop_needed()) {
 		/*
@@ -2305,8 +2307,9 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
 
 		/* tasklist protects us from ptrace_freeze_traced() */
 		__set_current_state(TASK_RUNNING);
+		read_code = false;
 		if (clear_code)
-			current->exit_code = 0;
+			exit_code = 0;
 		read_unlock(&tasklist_lock);
 	}
 
@@ -2316,8 +2319,10 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
 	 * any signal-sending on another CPU that wants to examine it.
 	 */
 	spin_lock_irq(&current->sighand->siglock);
+	if (read_code) exit_code = current->exit_code;
 	current->last_siginfo = NULL;
 	current->ptrace_message = 0;
+	current->exit_code = 0;
 
 	/* LISTENING can be set only during STOP traps, clear it */
 	current->jobctl &= ~JOBCTL_LISTENING;
@@ -2328,9 +2333,10 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
 	 * This sets TIF_SIGPENDING, but never clears it.
 	 */
 	recalc_sigpending_tsk(current);
+	return exit_code;
 }
 
-static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
+static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
 {
 	kernel_siginfo_t info;
 
@@ -2341,18 +2347,21 @@ static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long me
 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
 
 	/* Let the debugger run.  */
-	ptrace_stop(exit_code, why, 1, message, &info);
+	return ptrace_stop(exit_code, why, 1, message, &info);
 }
 
-void ptrace_notify(int exit_code, unsigned long message)
+int ptrace_notify(int exit_code, unsigned long message)
 {
+	int signr;
+
 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
 	if (unlikely(task_work_pending(current)))
 		task_work_run();
 
 	spin_lock_irq(&current->sighand->siglock);
-	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
+	signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
 	spin_unlock_irq(&current->sighand->siglock);
+	return signr;
 }
 
 /**
@@ -2511,7 +2520,6 @@ static void do_jobctl_trap(void)
 	} else {
 		WARN_ON_ONCE(!signr);
 		ptrace_stop(signr, CLD_STOPPED, 0, 0, NULL);
-		current->exit_code = 0;
 	}
 }
 
@@ -2564,15 +2572,12 @@ static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
 	 * comment in dequeue_signal().
 	 */
 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
-	ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
+	signr = ptrace_stop(signr, CLD_TRAPPED, 0, 0, info);
 
 	/* We're back.  Did the debugger cancel the sig?  */
-	signr = current->exit_code;
 	if (signr == 0)
 		return signr;
 
-	current->exit_code = 0;
-
 	/*
 	 * Update the siginfo structure if the signal has
 	 * changed.  If the debugger wanted something
-- 
2.29.2

Re: [PATCH 1/2] ptrace: Move setting/clearing ptrace_message into ptrace_stop

[Oleg Nesterov]

On 03/15, Eric W. Biederman wrote:
>
> there are places
> that set ptrace_message and don't clear it, and places that never set
> it.

Yes, I too never understood this.

So I obviously like this change. The only problem (as usual) is that we
can never know if something depends on this old (and strange) behaviour.

Acked-by: Oleg Nesterov <oleg@redhat.com>

Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Oleg Nesterov]

Not sure I understand this patch, I can't apply it. I guess I need to
clone your tree first, will do later.

Just one question right now,

On 03/15, Eric W. Biederman wrote:
>
> +static int ptrace_stop(int exit_code, int why, int clear_code,
>  			unsigned long message, kernel_siginfo_t *info)
>  	__releases(&current->sighand->siglock)
>  	__acquires(&current->sighand->siglock)
>  {
>  	bool gstop_done = false;
> +	bool read_code = true;
>  
>  	if (arch_ptrace_stop_needed()) {
>  		/*
> @@ -2305,8 +2307,9 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
>  
>  		/* tasklist protects us from ptrace_freeze_traced() */
>  		__set_current_state(TASK_RUNNING);
> +		read_code = false;
>  		if (clear_code)
> -			current->exit_code = 0;
> +			exit_code = 0;
>  		read_unlock(&tasklist_lock);
>  	}
>  
> @@ -2316,8 +2319,10 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
>  	 * any signal-sending on another CPU that wants to examine it.
>  	 */
>  	spin_lock_irq(&current->sighand->siglock);
> +	if (read_code) exit_code = current->exit_code;

style ;)

>  	current->last_siginfo = NULL;
>  	current->ptrace_message = 0;
> +	current->exit_code = 0;

OK, I like it... but can't we remove the ugly "int clear_code" arg?

Oleg.

Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Eric W. Biederman]

Oleg Nesterov <oleg@redhat.com> writes:

> Not sure I understand this patch, I can't apply it. I guess I need to
> clone your tree first, will do later.
>
> Just one question right now,
>
> On 03/15, Eric W. Biederman wrote:
>>
>> +static int ptrace_stop(int exit_code, int why, int clear_code,
>>  			unsigned long message, kernel_siginfo_t *info)
>>  	__releases(&current->sighand->siglock)
>>  	__acquires(&current->sighand->siglock)
>>  {
>>  	bool gstop_done = false;
>> +	bool read_code = true;
>>  
>>  	if (arch_ptrace_stop_needed()) {
>>  		/*
>> @@ -2305,8 +2307,9 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
>>  
>>  		/* tasklist protects us from ptrace_freeze_traced() */
>>  		__set_current_state(TASK_RUNNING);
>> +		read_code = false;
>>  		if (clear_code)
>> -			current->exit_code = 0;
>> +			exit_code = 0;
>>  		read_unlock(&tasklist_lock);
>>  	}
>>  
>> @@ -2316,8 +2319,10 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
>>  	 * any signal-sending on another CPU that wants to examine it.
>>  	 */
>>  	spin_lock_irq(&current->sighand->siglock);
>> +	if (read_code) exit_code = current->exit_code;
>
> style ;)
>
>>  	current->last_siginfo = NULL;
>>  	current->ptrace_message = 0;
>> +	current->exit_code = 0;
>
> OK, I like it... but can't we remove the ugly "int clear_code" arg?

The flag clear_code controls what happens if a ptrace_stop does not
stop.  In particular clear_code means do not continue with
a signal if we can not stop.

For do_jobctl_trap calling ptrace_stop it clearly does not matter.

For ptrace_signal it would be a change in behavior, that would
cause the signal not to be delivered.

For ptrace_do_notify we don't have a signal to deliver unless userspace
gives us one so clear_code makes sense at that point.

Which is a long way of saying that no we can not remove clear_stop
because the behavior it is used to implement makes sense and userspace
can reasonably depend upon it.

Eric

Re: [PATCH 1/2] ptrace: Move setting/clearing ptrace_message into ptrace_stop

[Kees Cook]

On Tue, Mar 15, 2022 at 06:21:08PM -0500, Eric W. Biederman wrote:
> 
> Today ptrace_message is easy to overlook as it not a core part of
> ptrace_stop.  It has been overlooked so much that there are places
> that set ptrace_message and don't clear it, and places that never set
> it.  So if you get an unlucky sequence of events the ptracer may be
> able to read a ptrace_message that does not apply to the current
> ptrace stop.
> 
> Move setting of ptrace_message into ptrace_stop so that it always gets
> set before the stop, and always gets cleared after the stop.  This
> prevents non-sense from being reported to userspace and makes
> ptrace_message more visible in the ptrace helper functions so that
> kernel developers can see it.
> 
> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

This looks good to me. Did you happen to run the seccomp selftests
before/after these changes?

Reviewed-by: Kees Cook <keescook@chromium.org>

-- 
Kees Cook

Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Kees Cook]

On Tue, Mar 15, 2022 at 06:22:26PM -0500, Eric W. Biederman wrote:
> 
> The signal a task should continue with after a ptrace stop is
> inconsistently read, cleared, and sent.  Solve this by reading and
> clearing the signal to be sent in ptrace_stop.
> 
> In an ideal world everything except ptrace_signal would share a common
> implementation of continuing with the signal, so ptracers could count
> on the signal they ask to continue with actually being delivered.  For
> now retain bug compatibility and just return with the signal number
> the ptracer requested the code continue with.
> 
> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> ---
>  include/linux/ptrace.h | 12 ++++++------
>  kernel/signal.c        | 31 ++++++++++++++++++-------------
>  2 files changed, 24 insertions(+), 19 deletions(-)
> 
> diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
> index 3e6b46e2b7be..15b3d176b6b4 100644
> --- a/include/linux/ptrace.h
> +++ b/include/linux/ptrace.h
> @@ -60,7 +60,7 @@ extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned
>  extern void ptrace_disable(struct task_struct *);
>  extern int ptrace_request(struct task_struct *child, long request,
>  			  unsigned long addr, unsigned long data);
> -extern void ptrace_notify(int exit_code, unsigned long message);
> +extern int ptrace_notify(int exit_code, unsigned long message);
> [...]
> -static void ptrace_stop(int exit_code, int why, int clear_code,
> +static int ptrace_stop(int exit_code, int why, int clear_code,
>  			unsigned long message, kernel_siginfo_t *info)
> [...]
> -static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
> +static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
> [...]
> -void ptrace_notify(int exit_code, unsigned long message)
> +int ptrace_notify(int exit_code, unsigned long message)

Just for robustness, how about marking the functions that have switched
from void to int return as __must_check (or at least just ptrace_notify)?

With that and the style nit Oleg already mentioned, yeah, this looks
good too.

Reviewed-by: Kees Cook <keescook@chromium.org>

-- 
Kees Cook

Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Eric W. Biederman]

Oleg Nesterov <oleg@redhat.com> writes:

> Not sure I understand this patch, I can't apply it. I guess I need to
> clone your tree first, will do later.

Yes this series is on top of:
  https://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git kill-tracehook-for-v5.18

Nothing particularly interesting happens in my kill tracehook series
but I do get rid of tracehook.h and so everything gets moved and
renamed.

Eric

Re: [PATCH 1/2] ptrace: Move setting/clearing ptrace_message into ptrace_stop

[Eric W. Biederman]

Kees Cook <keescook@chromium.org> writes:

> On Tue, Mar 15, 2022 at 06:21:08PM -0500, Eric W. Biederman wrote:
>> 
>> Today ptrace_message is easy to overlook as it not a core part of
>> ptrace_stop.  It has been overlooked so much that there are places
>> that set ptrace_message and don't clear it, and places that never set
>> it.  So if you get an unlucky sequence of events the ptracer may be
>> able to read a ptrace_message that does not apply to the current
>> ptrace stop.
>> 
>> Move setting of ptrace_message into ptrace_stop so that it always gets
>> set before the stop, and always gets cleared after the stop.  This
>> prevents non-sense from being reported to userspace and makes
>> ptrace_message more visible in the ptrace helper functions so that
>> kernel developers can see it.
>> 
>> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
>
> This looks good to me. Did you happen to run the seccomp selftests
> before/after these changes?

I did not.  This is a pure ptrace change.  Do you see a way that seccomp
could be involved?

Eric

Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Eric W. Biederman]

Kees Cook <keescook@chromium.org> writes:

> On Tue, Mar 15, 2022 at 06:22:26PM -0500, Eric W. Biederman wrote:
>> 
>> The signal a task should continue with after a ptrace stop is
>> inconsistently read, cleared, and sent.  Solve this by reading and
>> clearing the signal to be sent in ptrace_stop.
>> 
>> In an ideal world everything except ptrace_signal would share a common
>> implementation of continuing with the signal, so ptracers could count
>> on the signal they ask to continue with actually being delivered.  For
>> now retain bug compatibility and just return with the signal number
>> the ptracer requested the code continue with.
>> 
>> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
>> ---
>>  include/linux/ptrace.h | 12 ++++++------
>>  kernel/signal.c        | 31 ++++++++++++++++++-------------
>>  2 files changed, 24 insertions(+), 19 deletions(-)
>> 
>> diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
>> index 3e6b46e2b7be..15b3d176b6b4 100644
>> --- a/include/linux/ptrace.h
>> +++ b/include/linux/ptrace.h
>> @@ -60,7 +60,7 @@ extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned
>>  extern void ptrace_disable(struct task_struct *);
>>  extern int ptrace_request(struct task_struct *child, long request,
>>  			  unsigned long addr, unsigned long data);
>> -extern void ptrace_notify(int exit_code, unsigned long message);
>> +extern int ptrace_notify(int exit_code, unsigned long message);
>> [...]
>> -static void ptrace_stop(int exit_code, int why, int clear_code,
>> +static int ptrace_stop(int exit_code, int why, int clear_code,
>>  			unsigned long message, kernel_siginfo_t *info)
>> [...]
>> -static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
>> +static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
>> [...]
>> -void ptrace_notify(int exit_code, unsigned long message)
>> +int ptrace_notify(int exit_code, unsigned long message)
>
> Just for robustness, how about marking the functions that have switched
> from void to int return as __must_check (or at least just ptrace_notify)?

We can't.  There are historical cases that simply don't check if a
signal should be sent after the function, and they exist for every
function that is modified.

If we can modify the code so that everyone is checking the return value
than certainly, but that just doesn't happen to reflect how this
ptrace helper is being used today.

> With that and the style nit Oleg already mentioned, yeah, this looks
> good too.

Alright style nit fixed.

> Reviewed-by: Kees Cook <keescook@chromium.org>

Eric

Re: [PATCH 1/2] ptrace: Move setting/clearing ptrace_message into ptrace_stop

[Kees Cook]

On Fri, Mar 18, 2022 at 09:44:30AM -0500, Eric W. Biederman wrote:
> Kees Cook <keescook@chromium.org> writes:
> 
> > On Tue, Mar 15, 2022 at 06:21:08PM -0500, Eric W. Biederman wrote:
> >> 
> >> Today ptrace_message is easy to overlook as it not a core part of
> >> ptrace_stop.  It has been overlooked so much that there are places
> >> that set ptrace_message and don't clear it, and places that never set
> >> it.  So if you get an unlucky sequence of events the ptracer may be
> >> able to read a ptrace_message that does not apply to the current
> >> ptrace stop.
> >> 
> >> Move setting of ptrace_message into ptrace_stop so that it always gets
> >> set before the stop, and always gets cleared after the stop.  This
> >> prevents non-sense from being reported to userspace and makes
> >> ptrace_message more visible in the ptrace helper functions so that
> >> kernel developers can see it.
> >> 
> >> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> >
> > This looks good to me. Did you happen to run the seccomp selftests
> > before/after these changes?
> 
> I did not.  This is a pure ptrace change.  Do you see a way that seccomp
> could be involved?

Sorry, that wasn't clear: seccomp includes a number of ptrace tests as
well, especially involving handling process death, messages, and
signals. I'll give it a spin; so far it seems fine.

-- 
Kees Cook

Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Kees Cook]

On Fri, Mar 18, 2022 at 09:52:46AM -0500, Eric W. Biederman wrote:
> Kees Cook <keescook@chromium.org> writes:
> 
> > On Tue, Mar 15, 2022 at 06:22:26PM -0500, Eric W. Biederman wrote:
> >> 
> >> The signal a task should continue with after a ptrace stop is
> >> inconsistently read, cleared, and sent.  Solve this by reading and
> >> clearing the signal to be sent in ptrace_stop.
> >> 
> >> In an ideal world everything except ptrace_signal would share a common
> >> implementation of continuing with the signal, so ptracers could count
> >> on the signal they ask to continue with actually being delivered.  For
> >> now retain bug compatibility and just return with the signal number
> >> the ptracer requested the code continue with.
> >> 
> >> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> >> ---
> >>  include/linux/ptrace.h | 12 ++++++------
> >>  kernel/signal.c        | 31 ++++++++++++++++++-------------
> >>  2 files changed, 24 insertions(+), 19 deletions(-)
> >> 
> >> diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
> >> index 3e6b46e2b7be..15b3d176b6b4 100644
> >> --- a/include/linux/ptrace.h
> >> +++ b/include/linux/ptrace.h
> >> @@ -60,7 +60,7 @@ extern int ptrace_writedata(struct task_struct *tsk, char __user *src, unsigned
> >>  extern void ptrace_disable(struct task_struct *);
> >>  extern int ptrace_request(struct task_struct *child, long request,
> >>  			  unsigned long addr, unsigned long data);
> >> -extern void ptrace_notify(int exit_code, unsigned long message);
> >> +extern int ptrace_notify(int exit_code, unsigned long message);
> >> [...]
> >> -static void ptrace_stop(int exit_code, int why, int clear_code,
> >> +static int ptrace_stop(int exit_code, int why, int clear_code,
> >>  			unsigned long message, kernel_siginfo_t *info)
> >> [...]
> >> -static void ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
> >> +static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
> >> [...]
> >> -void ptrace_notify(int exit_code, unsigned long message)
> >> +int ptrace_notify(int exit_code, unsigned long message)
> >
> > Just for robustness, how about marking the functions that have switched
> > from void to int return as __must_check (or at least just ptrace_notify)?
> 
> We can't.  There are historical cases that simply don't check if a
> signal should be sent after the function, and they exist for every
> function that is modified.

This seems at least worth documenting with a comment, otherwise we're
just trading one kind of "weirdness" (setting/clearing
current->exit_code) with another (ignoring the signal returned by
ptrace_notify()).

I see only two cases that would need comments:

static inline void ptrace_event(int event, unsigned long message)
{
        if (unlikely(ptrace_event_enabled(current, event))) {
                ptrace_notify((event << 8) | SIGTRAP, message);
        } else if (event == PTRACE_EVENT_EXEC) {
                /* legacy EXEC report via SIGTRAP */
                if ((current->ptrace & (PT_PTRACED|PT_SEIZED)) == PT_PTRACED)
                        send_sig(SIGTRAP, current, 0);
        }
}

static void signal_delivered(struct ksignal *ksig, int stepping)
{
	...
        if (stepping)
                ptrace_notify(SIGTRAP, 0);
}


-- 
Kees Cook

Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

[Oleg Nesterov]

On 03/17, Eric W. Biederman wrote:
>
> Oleg Nesterov <oleg@redhat.com> writes:
>
> > OK, I like it... but can't we remove the ugly "int clear_code" arg?
>
> The flag clear_code controls what happens if a ptrace_stop does not
> stop.  In particular clear_code means do not continue with
> a signal if we can not stop.
>
> For do_jobctl_trap calling ptrace_stop it clearly does not matter.
>
> For ptrace_signal it would be a change in behavior, that would
> cause the signal not to be delivered.

Well I meant that "clear_code" should be false, iirc only
ptrace_report_syscall() should be updated to void the spurious
send_sig() if debugger exits... Nevermind, pleasee forget, this is
not as trivial as I thought.

Acked-by: Oleg Nesterov <oleg@redhat.com>

[GIT PULL] ptrace: Cleanups for v5.18

[Eric W. Biederman]

Linus,

Please pull the ptrace-cleanups-for-v5.18 tag from the git tree:

  git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git ptrace-cleanups-for-v5.18

  HEAD: dcbc65aac28360df5f5a3b613043ccc0e81da3cf ptrace: Remove duplicated include in ptrace.c

This set of changes removes tracehook.h, moves modification of all of
the ptrace fields inside of siglock to remove races, adds a missing
permission check to ptrace.c

The removal of tracehook.h is quite significant as it has been a major
source of confusion in recent years.  Much of that confusion was
around task_work and TIF_NOTIFY_SIGNAL (which I have now decoupled
making the semantics clearer).

For people who don't know tracehook.h is a vestiage of an attempt to
implement uprobes like functionality that was never fully merged, and
was later superseeded by uprobes when uprobes was merged.  For many
years now we have been removing what tracehook functionaly a little
bit at a time.  To the point where now anything left in tracehook.h is
some weird strange thing that is difficult to understand.

Eric W. Biederman (15):
      ptrace: Move ptrace_report_syscall into ptrace.h
      ptrace/arm: Rename tracehook_report_syscall report_syscall
      ptrace: Create ptrace_report_syscall_{entry,exit} in ptrace.h
      ptrace: Remove arch_syscall_{enter,exit}_tracehook
      ptrace: Remove tracehook_signal_handler
      task_work: Remove unnecessary include from posix_timers.h
      task_work: Introduce task_work_pending
      task_work: Call tracehook_notify_signal from get_signal on all architectures
      task_work: Decouple TIF_NOTIFY_SIGNAL and task_work
      signal: Move set_notify_signal and clear_notify_signal into sched/signal.h
      resume_user_mode: Remove #ifdef TIF_NOTIFY_RESUME in set_notify_resume
      resume_user_mode: Move to resume_user_mode.h
      tracehook: Remove tracehook.h
      ptrace: Move setting/clearing ptrace_message into ptrace_stop
      ptrace: Return the signal to continue with from ptrace_stop

Jann Horn (1):
      ptrace: Check PTRACE_O_SUSPEND_SECCOMP permission on PTRACE_SEIZE

Yang Li (1):
      ptrace: Remove duplicated include in ptrace.c


 MAINTAINERS                          |   1 -
 arch/Kconfig                         |   5 +-
 arch/alpha/kernel/ptrace.c           |   5 +-
 arch/alpha/kernel/signal.c           |   4 +-
 arch/arc/kernel/ptrace.c             |   5 +-
 arch/arc/kernel/signal.c             |   4 +-
 arch/arm/kernel/ptrace.c             |  12 +-
 arch/arm/kernel/signal.c             |   4 +-
 arch/arm64/kernel/ptrace.c           |  14 +--
 arch/arm64/kernel/signal.c           |   4 +-
 arch/csky/kernel/ptrace.c            |   5 +-
 arch/csky/kernel/signal.c            |   4 +-
 arch/h8300/kernel/ptrace.c           |   5 +-
 arch/h8300/kernel/signal.c           |   4 +-
 arch/hexagon/kernel/process.c        |   4 +-
 arch/hexagon/kernel/signal.c         |   1 -
 arch/hexagon/kernel/traps.c          |   6 +-
 arch/ia64/kernel/process.c           |   4 +-
 arch/ia64/kernel/ptrace.c            |   6 +-
 arch/ia64/kernel/signal.c            |   1 -
 arch/m68k/kernel/ptrace.c            |   5 +-
 arch/m68k/kernel/signal.c            |   4 +-
 arch/microblaze/kernel/ptrace.c      |   5 +-
 arch/microblaze/kernel/signal.c      |   4 +-
 arch/mips/kernel/ptrace.c            |   5 +-
 arch/mips/kernel/signal.c            |   4 +-
 arch/nds32/include/asm/syscall.h     |   2 +-
 arch/nds32/kernel/ptrace.c           |   5 +-
 arch/nds32/kernel/signal.c           |   4 +-
 arch/nios2/kernel/ptrace.c           |   5 +-
 arch/nios2/kernel/signal.c           |   4 +-
 arch/openrisc/kernel/ptrace.c        |   5 +-
 arch/openrisc/kernel/signal.c        |   4 +-
 arch/parisc/kernel/ptrace.c          |   7 +-
 arch/parisc/kernel/signal.c          |   4 +-
 arch/powerpc/kernel/ptrace/ptrace.c  |   8 +-
 arch/powerpc/kernel/signal.c         |   4 +-
 arch/riscv/kernel/ptrace.c           |   5 +-
 arch/riscv/kernel/signal.c           |   4 +-
 arch/s390/include/asm/entry-common.h |   1 -
 arch/s390/kernel/ptrace.c            |   1 -
 arch/s390/kernel/signal.c            |   5 +-
 arch/sh/kernel/ptrace_32.c           |   5 +-
 arch/sh/kernel/signal_32.c           |   4 +-
 arch/sparc/kernel/ptrace_32.c        |   5 +-
 arch/sparc/kernel/ptrace_64.c        |   5 +-
 arch/sparc/kernel/signal32.c         |   1 -
 arch/sparc/kernel/signal_32.c        |   4 +-
 arch/sparc/kernel/signal_64.c        |   4 +-
 arch/um/kernel/process.c             |   4 +-
 arch/um/kernel/ptrace.c              |   5 +-
 arch/x86/kernel/ptrace.c             |   1 -
 arch/x86/kernel/signal.c             |   5 +-
 arch/x86/mm/tlb.c                    |   1 +
 arch/xtensa/kernel/ptrace.c          |   5 +-
 arch/xtensa/kernel/signal.c          |   4 +-
 block/blk-cgroup.c                   |   2 +-
 fs/coredump.c                        |   1 -
 fs/exec.c                            |   1 -
 fs/io-wq.c                           |   6 +-
 fs/io_uring.c                        |  11 +-
 fs/proc/array.c                      |   1 -
 fs/proc/base.c                       |   1 -
 include/asm-generic/syscall.h        |   2 +-
 include/linux/entry-common.h         |  47 +-------
 include/linux/entry-kvm.h            |   2 +-
 include/linux/posix-timers.h         |   1 -
 include/linux/ptrace.h               |  81 ++++++++++++-
 include/linux/resume_user_mode.h     |  64 ++++++++++
 include/linux/sched/signal.h         |  17 +++
 include/linux/task_work.h            |   5 +
 include/linux/tracehook.h            | 226 -----------------------------------
 include/uapi/linux/ptrace.h          |   2 +-
 kernel/entry/common.c                |  19 +--
 kernel/entry/kvm.c                   |   9 +-
 kernel/exit.c                        |   3 +-
 kernel/livepatch/transition.c        |   1 -
 kernel/ptrace.c                      |  47 +++++---
 kernel/seccomp.c                     |   1 -
 kernel/signal.c                      |  62 +++++-----
 kernel/task_work.c                   |   4 +-
 kernel/time/posix-cpu-timers.c       |   1 +
 mm/memcontrol.c                      |   2 +-
 security/apparmor/domain.c           |   1 -
 security/selinux/hooks.c             |   1 -
 85 files changed, 372 insertions(+), 495 deletions(-)

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

Re: [GIT PULL] ptrace: Cleanups for v5.18

[Jens Axboe]

On 3/28/22 5:56 PM, Eric W. Biederman wrote:
> 
> Linus,
> 
> Please pull the ptrace-cleanups-for-v5.18 tag from the git tree:
> 
>   git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git ptrace-cleanups-for-v5.18
> 
>   HEAD: dcbc65aac28360df5f5a3b613043ccc0e81da3cf ptrace: Remove duplicated include in ptrace.c
> 
> This set of changes removes tracehook.h, moves modification of all of
> the ptrace fields inside of siglock to remove races, adds a missing
> permission check to ptrace.c
> 
> The removal of tracehook.h is quite significant as it has been a major
> source of confusion in recent years.  Much of that confusion was
> around task_work and TIF_NOTIFY_SIGNAL (which I have now decoupled
> making the semantics clearer).
> 
> For people who don't know tracehook.h is a vestiage of an attempt to
> implement uprobes like functionality that was never fully merged, and
> was later superseeded by uprobes when uprobes was merged.  For many
> years now we have been removing what tracehook functionaly a little
> bit at a time.  To the point where now anything left in tracehook.h is
> some weird strange thing that is difficult to understand.

FWIW, the notify/task_work/io_uring changes look good to me. Thanks for
cleaning this up, Eric.

-- 
Jens Axboe

Re: [GIT PULL] ptrace: Cleanups for v5.18

[Linus Torvalds]

On Mon, Mar 28, 2022 at 4:56 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>
> The removal of tracehook.h is quite significant as it has been a major
> source of confusion in recent years.  Much of that confusion was
> around task_work and TIF_NOTIFY_SIGNAL (which I have now decoupled
> making the semantics clearer).

Hmm. I love removing tracehook.c, but this looks like it hasn't been
in linux-next.

The header file changes messes with other changes, and we have

  kernel/sched/fair.c:2884:9: error: implicit declaration of function
‘init_task_work’; did you mean ‘init_irq_work’?
[-Werror=implicit-function-declaration]
   2884 |         init_task_work(&p->numa_work, task_numa_work);
        |         ^~~~~~~~~~~~~~

as a result (also a few other things in that same file).

Now, this is trivial to fix - just add an include for
<linux/task_work.h> from that file - and that's the right thing to do
anyway.

But I'm a bit unhappy that this was either not tested in linux-next,
or if it was, I wasn't notified about the semantic in the pull
request.

So I've pulled this, and fixed up things in my merge, but I'm a bit
worried that there might be other situations like this where some
header file is no longer included and it was included implicitly
before through that disgusting tracehook.h header..

I *hope* it was just the scheduler header file updates that ended up
having this effect, and nothing else is affected.

Let's see if the test robots start complaining about non-x86
architecture-specific stuff that I don't build test.

              Linus

Re: [GIT PULL] ptrace: Cleanups for v5.18

[pr-tracker-bot]

The pull request you sent on Mon, 28 Mar 2022 18:56:46 -0500:

> git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git ptrace-cleanups-for-v5.18

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/1930a6e739c4b4a654a69164dbe39e554d228915

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/prtracker.html

Re: [GIT PULL] ptrace: Cleanups for v5.18

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 657 bytes --]

Hi Linus,

On Mon, 28 Mar 2022 17:33:52 -0700 Linus Torvalds <torvalds@linux-foundation.org> wrote:
>
> On Mon, Mar 28, 2022 at 4:56 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
> >
> > The removal of tracehook.h is quite significant as it has been a major
> > source of confusion in recent years.  Much of that confusion was
> > around task_work and TIF_NOTIFY_SIGNAL (which I have now decoupled
> > making the semantics clearer).  
> 
> Hmm. I love removing tracehook.c, but this looks like it hasn't been
> in linux-next.

See https://lore.kernel.org/lkml/20220316165612.4f50faad@canb.auug.org.au/

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [GIT PULL] ptrace: Cleanups for v5.18

[Linus Torvalds]

On Mon, Mar 28, 2022 at 5:53 PM Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>
> See https://lore.kernel.org/lkml/20220316165612.4f50faad@canb.auug.org.au/

Ok, so it was known, just not reported to me.

And the good news seems to be that at least there isn't anything
hiding elsewhere.

                Linus

Re: [GIT PULL] ptrace: Cleanups for v5.18

[Eric W. Biederman]

Linus Torvalds <torvalds@linux-foundation.org> writes:

> On Mon, Mar 28, 2022 at 4:56 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>>
>> The removal of tracehook.h is quite significant as it has been a major
>> source of confusion in recent years.  Much of that confusion was
>> around task_work and TIF_NOTIFY_SIGNAL (which I have now decoupled
>> making the semantics clearer).
>
> Hmm. I love removing tracehook.c, but this looks like it hasn't been
> in linux-next.
>
> The header file changes messes with other changes, and we have
>
>   kernel/sched/fair.c:2884:9: error: implicit declaration of function
> ‘init_task_work’; did you mean ‘init_irq_work’?
> [-Werror=implicit-function-declaration]
>    2884 |         init_task_work(&p->numa_work, task_numa_work);
>         |         ^~~~~~~~~~~~~~
>
> as a result (also a few other things in that same file).
>
> Now, this is trivial to fix - just add an include for
> <linux/task_work.h> from that file - and that's the right thing to do
> anyway.
>
> But I'm a bit unhappy that this was either not tested in linux-next,
> or if it was, I wasn't notified about the semantic in the pull
> request.
>
> So I've pulled this, and fixed up things in my merge, but I'm a bit
> worried that there might be other situations like this where some
> header file is no longer included and it was included implicitly
> before through that disgusting tracehook.h header..
>
> I *hope* it was just the scheduler header file updates that ended up
> having this effect, and nothing else is affected.
>
> Let's see if the test robots start complaining about non-x86
> architecture-specific stuff that I don't build test.

Sorry for not mentioning that.  I had tracked it down.  It was
fundamentally in the scheduler headers changes removing an include of
task_work.h, so it didn't feel like there was anything I could do in my
tree.  I asked Ingo if he could fix his tree and unfortunately forgot
about it.

For the record there were also a couple of other pretty trivial
conflicts, the removal of nds32, some block_cgroup header where
an adjacent line was modified to what I was changing.  But it thankfully
looks like none of those caused you any problems.

Sorry about all of that I am about that.  I am running pretty weak this
last couple of days as a cold has been running through the household.


Dumb question because this seems to burning a few extra creativity
points.  Is there any way to create a signed tag and a branch with the
same name?  Or in general is there a good way to manage topic branches
and then tag them at the end before I send them?

Having a tag and a branch with the same name seems to completely confuse
git and it just tells me no I won't push anything to another git tree,
because what you are asking me to do is ambiguous.  So now I am having
to come up with two names for each topic branch, even if I only push the
tags upstream.

I feel like there is a best practice on how to manage tags and topic
branches and I just haven't seen it yet.

Eric

Re: [GIT PULL] ptrace: Cleanups for v5.18

[Linus Torvalds]

On Mon, Mar 28, 2022 at 8:38 PM Eric W. Biederman <ebiederm@xmission.com> wrote:
>
> Dumb question because this seems to burning a few extra creativity
> points.  Is there any way to create a signed tag and a branch with the
> same name?

Oh, absolutely.

But you may find it annoying, because as you noticed:

> Having a tag and a branch with the same name seems to completely confuse
> git and it just tells me no I won't push anything to another git tree,
> because what you are asking me to do is ambiguous.

Not at all.

git is not at all confused by the situation, git is in fact very aware
of it indeed.

But as git then tells you, exactly *because* it is aware that you have
picked the same name for both a branch and a tag, it will keep warning
you about the ambiguity of said name (but after warning, will
generally then preferentially use the tag of that name over the branch
of the same name).

So if you have both a branch and a tag named 'xyz', you generally need
to disambiguate them when you use them. That will make git happy,
because now it's not ambiguous any more.

(Technical detail: some git operations work on specific namespaces, so
"git branch -d xyz" should never remove a _tag_ called 'xyz', and as
such the name isn't ambiguous in the context of that git command)

And disambiguating them is simple, but I suspect you'll find it's
annoying enough that you simply don't want to use the same name for
tags and branches.

The full name of a tag 'x' is actually 'refs/tags/x', and the full
unambiguous name of a branch 'x' is 'refs/heads/x'.

So technically a tag and a branch can never _really_ have the same
name, because internally they have longer unambiguous names.

You would almost never actually use that full name, it's mostly an
internal git thing. Because even if you have ambiguous branch and tag
names, you'd then still shorten it to just 'tags/x' and 'heads/x'
respectively.

Git has other "namespaces" for these refs, too, notably
'refs/remotes/'. In fact, I guess you could make up your own namespace
too if you really wanted, although I don't think anybody really has
ever had a good reason for it.

(There is also the non-namespaced special refs like HEAD and
FETCH_HEAD and MERGE_HEAD for "current branch", "what you fetched" and
"what you are merging")

So you *can* do this:

    # create and check out the branch 'xyz'
    $ git checkout -b xyz master

    # create a tag called 'xyz', but to confuse everybody, make it
point somewhere else
    $ git tag xyz master~3

    # look at what 'xyz' means:
    $ git rev-parse xyz
    warning: refname 'xyz' is ambiguous.
    cffb2b72d3ed47f5093d128bd44d9ce136b6b5af

    # Hmm. it warned about it being ambiguous
    $ git rev-parse heads/xyz
    1930a6e739c4b4a654a69164dbe39e554d228915

    # Ok, it clearly took the tag, not the branch:
    $ git rev-parse tags/xyz
    cffb2b72d3ed47f5093d128bd44d9ce136b6b5af

so as you can see, yes, you can work with a tag and a branch that have
the same 'short name', but having to disambiguate them all the time
will likely just drive you mad.

And it's worth pointing out that the name does not imply a
relationship. So the branch called 'xyz' (ie refs/heads/xyz) has
absolutely *no* relationship to the tag called 'xyz' (ie
refs/tags/xyz) except for that ambiguous short name. So updating the
branch - by perhaps committing more to it - will in no way affect the
tag.

Also note that branches and tags are both "just refs" as far as git is
concerned, but git *does* give some semantic meaning to the
namespaces.

So the branch namespace (it those 'refs/heads/*' things) are things
that get updated automatically as you make new commits.

In contrast, the tag namespace is something you *can* update, but it's
considered odd, and if you want to overwrite an existing tag you
generally need to do something special (eg "git tag -f" to force
overwriting a new tag rather than telling you that you already have
one).

So in a very real way, to git a ref is a ref is a ref, with very
little to no real *technical* distinction. They are just ways to point
to the SHA1 hash of an object. But there are basically some common
semantic rules that are based on the namespaces, and all those git
operations that use certain namespaces by default.

            Linus

Re: [GIT PULL] ptrace: Cleanups for v5.18

[Linus Torvalds]

On Mon, Mar 28, 2022 at 9:49 PM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> So if you have both a branch and a tag named 'xyz', you generally need
> to disambiguate them when you use them. That will make git happy,
> because now it's not ambiguous any more.

On a similar but very different issue: this is not the only kind of
naming ambiguity you can have.

For example, try this insane thing:

    git tag Makefile

that creates a tag called 'Makefile' (pointing to your current HEAD, of course).

Now, guess what happens when you then do

    git log Makefile

that's right - git once again notices that you are doing something ambiguous.

In fact, git will be *so* unhappy about this kind of ambiguity that
(unlike the 'tag vs branch' case) it will not prefer one version of
reality over the other, and simply consider this to be a fatal error,
and say

    fatal: ambiguous argument 'Makefile': both revision and filename
    Use '--' to separate paths from revisions, like this:
    'git <command> [<revision>...] -- [<file>...]'

and as a result you will hopefully go "Oh, I didn't mean to have that
tag at all" and just remove the bogus tagname.

Because you probably made it by mistake.

But you *can* choose to say

   git log Makefile --

or

   git log refs/tags/Makefile

to make it clear that no, 'Makefile' is not the pathname in the
repository, you really mean the tag 'Makefile'.

Or use

  git log -- Makefile

or

    git log ./Makefile

to say "yeah, I meant the _pathname_ 'Makefile', not the tag".

Or, if you just like playing games, do

    git log Makefile -- Makefile

if you want to track the history of the path 'Makefile' starting at
the tag 'Makefile'.

But don't actually do this for real. There's a reason git notices these things.

Using ambiguous branch names (whether ambiguous with tag-names or with
filenames) is a pain that is not worth it in practice. Git will
notice, and git will allow you to work around it, but it's just a
BadIdea(tm) despite that.

But you probably want to be aware of issues like this if you script
things around git, though.

That "--" form in particular is generally the one you want to use for
scripting, if you want to allow people to do anything they damn well
please. It's the "Give them rope" syntax.

Of course, a much more common reason for "--" for pathname separation,
and one you may already be familiar with, is that you want to see the
history of a pathname that is not *currently* a pathname, but was one
at some point in the past.

So in my current kernel tree, doing

    $ git log arch/nds32/

will actually cause an error, because of "unknown revision or path not
in the working tree".

But if you really want to see the history of that now-deleted
architecture, you do

    $ git log -- arch/nds32/

and it's all good.

This concludes today's sermon on git,

             Linus