patches/old/add-sicode-to-proc-pid-stat.patch

From: Florian Mayer <fmayer@google.com>
Subject: procfs: add sicode to /proc/<PID>/stat.
Date: Fri, 9 Sep 2022 11:06:17 -0700

In order to enable additional debugging features, Android init needs a way
to distinguish MTE-related SEGVs (with si_code of SEGV_MTEAERR) from other
SEGVs.  This is not possible with current APIs, neither by the existing
information in /proc/<pid>/stat, nor via waitpid.

Tested with the following program

int main(int argc, char** argv) {
  int pid = fork();
  if (!pid) {
    if (strcmp(argv[1], "sigqueue") == 0) {
    union sigval value;
    value.sival_int = 0;
    sigqueue(getpid(), SIGSEGV, value);
    } else if (strcmp(argv[1], "raise") == 0) {
     raise(SIGSEGV);
    } else if (strcmp(argv[1], "kill") == 0) {
      kill(getpid(), SIGSEGV);
    } else if (strcmp(argv[1], "raisestop") == 0) {
      raise(SIGSTOP);
    } else if (strcmp(argv[1], "crash") == 0) {
      volatile int* x = (int*)(0x23);
      *x = 1;
    } else if (strcmp(argv[1], "mte") == 0) {
      volatile char* y = malloc(1);
      y += 100;
      *y = 1;
    }
  } else {
    printf("%d\n", pid);
    sleep(5);
    char buf[1024];
    sprintf(buf, "/proc/%d/stat", pid);
    int fd = open(buf, O_RDONLY);
    char statb[1024];
    read(fd, statb, sizeof(statb));
    printf("%s\n", statb);
  }
}

Link: https://lkml.kernel.org/r/20220909180617.374238-1-fmayer@google.com
Signed-off-by: Florian Mayer <fmayer@google.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner (Microsoft) <brauner@kernel.org>
Cc: "Eric W . Biederman" <ebiederm@xmission.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Collingbourne <pcc@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 Documentation/filesystems/proc.rst |    2 +
 fs/coredump.c                      |   17 +++++++++------
 fs/proc/array.c                    |   12 +++++++----
 include/linux/sched/signal.h       |    1 
 include/linux/sched/task.h         |    2 -
 kernel/exit.c                      |    5 ++--
 kernel/pid_namespace.c             |    4 ++-
 kernel/signal.c                    |   29 +++++++++++++++++----------
 8 files changed, 47 insertions(+), 25 deletions(-)

--- a/Documentation/filesystems/proc.rst~add-sicode-to-proc-pid-stat
+++ a/Documentation/filesystems/proc.rst
@@ -381,6 +381,8 @@ It's slow but very precise.
   env_end       address below which program environment is placed
   exit_code     the thread's exit_code in the form reported by the waitpid
 		system call
+  exit_sicode   if the process was stopped or terminated by a signal, the
+		signal's si_code. 0 otherwise
   ============= ===============================================================
 
 The /proc/PID/maps file contains the currently mapped memory regions and
--- a/fs/coredump.c~add-sicode-to-proc-pid-stat
+++ a/fs/coredump.c
@@ -353,7 +353,7 @@ out:
 	return ispipe;
 }
 
-static int zap_process(struct task_struct *start, int exit_code)
+static int zap_process(struct task_struct *start, int exit_code, int sicode)
 {
 	struct task_struct *t;
 	int nr = 0;
@@ -361,6 +361,7 @@ static int zap_process(struct task_struc
 	/* ignore all signals except SIGKILL, see prepare_signal() */
 	start->signal->flags = SIGNAL_GROUP_EXIT;
 	start->signal->group_exit_code = exit_code;
+	start->signal->group_exit_sicode = sicode;
 	start->signal->group_stop_count = 0;
 
 	for_each_thread(start, t) {
@@ -375,8 +376,8 @@ static int zap_process(struct task_struc
 	return nr;
 }
 
-static int zap_threads(struct task_struct *tsk,
-			struct core_state *core_state, int exit_code)
+static int zap_threads(struct task_struct *tsk, struct core_state *core_state,
+		       int exit_code, int sicode)
 {
 	struct signal_struct *signal = tsk->signal;
 	int nr = -EAGAIN;
@@ -384,7 +385,7 @@ static int zap_threads(struct task_struc
 	spin_lock_irq(&tsk->sighand->siglock);
 	if (!(signal->flags & SIGNAL_GROUP_EXIT) && !signal->group_exec_task) {
 		signal->core_state = core_state;
-		nr = zap_process(tsk, exit_code);
+		nr = zap_process(tsk, exit_code, sicode);
 		clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
 		tsk->flags |= PF_DUMPCORE;
 		atomic_set(&core_state->nr_threads, nr);
@@ -393,7 +394,8 @@ static int zap_threads(struct task_struc
 	return nr;
 }
 
-static int coredump_wait(int exit_code, struct core_state *core_state)
+static int coredump_wait(int exit_code, int sicode,
+			 struct core_state *core_state)
 {
 	struct task_struct *tsk = current;
 	int core_waiters = -EBUSY;
@@ -402,7 +404,7 @@ static int coredump_wait(int exit_code,
 	core_state->dumper.task = tsk;
 	core_state->dumper.next = NULL;
 
-	core_waiters = zap_threads(tsk, core_state, exit_code);
+	core_waiters = zap_threads(tsk, core_state, exit_code, sicode);
 	if (core_waiters > 0) {
 		struct core_thread *ptr;
 
@@ -565,7 +567,8 @@ void do_coredump(const kernel_siginfo_t
 		need_suid_safe = true;
 	}
 
-	retval = coredump_wait(siginfo->si_signo, &core_state);
+	retval =
+		coredump_wait(siginfo->si_signo, siginfo->si_code, &core_state);
 	if (retval < 0)
 		goto fail_creds;
 
--- a/fs/proc/array.c~add-sicode-to-proc-pid-stat
+++ a/fs/proc/array.c
@@ -474,6 +474,7 @@ static int do_task_stat(struct seq_file
 	unsigned long rsslim = 0;
 	unsigned long flags;
 	int exit_code = task->exit_code;
+	int exit_sicode = 0;
 
 	state = *get_task_state(task);
 	vsize = eip = esp = 0;
@@ -538,8 +539,10 @@ static int do_task_stat(struct seq_file
 			thread_group_cputime_adjusted(task, &utime, &stime);
 			gtime += sig->gtime;
 
-			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
+			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED)) {
 				exit_code = sig->group_exit_code;
+				exit_sicode = sig->group_exit_sicode;
+			}
 		}
 
 		sid = task_session_nr_ns(task, ns);
@@ -638,10 +641,11 @@ static int do_task_stat(struct seq_file
 	} else
 		seq_puts(m, " 0 0 0 0 0 0 0");
 
-	if (permitted)
+	if (permitted) {
 		seq_put_decimal_ll(m, " ", exit_code);
-	else
-		seq_puts(m, " 0");
+		seq_put_decimal_ll(m, " ", exit_sicode);
+	} else
+		seq_puts(m, " 0 0");
 
 	seq_putc(m, '\n');
 	if (mm)
--- a/include/linux/sched/signal.h~add-sicode-to-proc-pid-stat
+++ a/include/linux/sched/signal.h
@@ -109,6 +109,7 @@ struct signal_struct {
 
 	/* thread group exit support */
 	int			group_exit_code;
+	int			group_exit_sicode;
 	/* notify group_exec_task when notify_count is less or equal to 0 */
 	int			notify_count;
 	struct task_struct	*group_exec_task;
--- a/include/linux/sched/task.h~add-sicode-to-proc-pid-stat
+++ a/include/linux/sched/task.h
@@ -82,7 +82,7 @@ static inline void exit_thread(struct ta
 {
 }
 #endif
-extern __noreturn void do_group_exit(int);
+extern __noreturn void do_group_exit(int,int);
 
 extern void exit_files(struct task_struct *);
 extern void exit_itimers(struct task_struct *);
--- a/kernel/exit.c~add-sicode-to-proc-pid-stat
+++ a/kernel/exit.c
@@ -904,7 +904,7 @@ SYSCALL_DEFINE1(exit, int, error_code)
  * as well as by sys_exit_group (below).
  */
 void __noreturn
-do_group_exit(int exit_code)
+do_group_exit(int exit_code, int sicode)
 {
 	struct signal_struct *sig = current->signal;
 
@@ -923,6 +923,7 @@ do_group_exit(int exit_code)
 			exit_code = 0;
 		else {
 			sig->group_exit_code = exit_code;
+			sig->group_exit_sicode = sicode;
 			sig->flags = SIGNAL_GROUP_EXIT;
 			zap_other_threads(current);
 		}
@@ -940,7 +941,7 @@ do_group_exit(int exit_code)
  */
 SYSCALL_DEFINE1(exit_group, int, error_code)
 {
-	do_group_exit((error_code & 0xff) << 8);
+	do_group_exit((error_code & 0xff) << 8, 0);
 	/* NOTREACHED */
 	return 0;
 }
--- a/kernel/pid_namespace.c~add-sicode-to-proc-pid-stat
+++ a/kernel/pid_namespace.c
@@ -248,8 +248,10 @@ void zap_pid_ns_processes(struct pid_nam
 	}
 	__set_current_state(TASK_RUNNING);
 
-	if (pid_ns->reboot)
+	if (pid_ns->reboot) {
 		current->signal->group_exit_code = pid_ns->reboot;
+		current->signal->group_exit_sicode = 0;
+	}
 
 	acct_exit_ns(pid_ns);
 	return;
--- a/kernel/signal.c~add-sicode-to-proc-pid-stat
+++ a/kernel/signal.c
@@ -963,6 +963,7 @@ static bool prepare_signal(int sig, stru
 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
 			signal->group_stop_count = 0;
 			signal->group_exit_code = 0;
+			signal->group_exit_sicode = 0;
 		}
 	}
 
@@ -994,7 +995,8 @@ static inline bool wants_signal(int sig,
 	return task_curr(p) || !task_sigpending(p);
 }
 
-static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
+static void complete_signal(int sig, int code, struct task_struct *p,
+			    enum pid_type type)
 {
 	struct signal_struct *signal = p->signal;
 	struct task_struct *t;
@@ -1051,6 +1053,7 @@ static void complete_signal(int sig, str
 			 */
 			signal->flags = SIGNAL_GROUP_EXIT;
 			signal->group_exit_code = sig;
+			signal->group_exit_sicode = code;
 			signal->group_stop_count = 0;
 			t = p;
 			do {
@@ -1082,6 +1085,7 @@ static int __send_signal_locked(int sig,
 	struct sigqueue *q;
 	int override_rlimit;
 	int ret = 0, result;
+	int code = 0;
 
 	lockdep_assert_held(&t->sighand->siglock);
 
@@ -1129,7 +1133,7 @@ static int __send_signal_locked(int sig,
 			clear_siginfo(&q->info);
 			q->info.si_signo = sig;
 			q->info.si_errno = 0;
-			q->info.si_code = SI_USER;
+			code = q->info.si_code = SI_USER;
 			q->info.si_pid = task_tgid_nr_ns(current,
 							task_active_pid_ns(t));
 			rcu_read_lock();
@@ -1142,12 +1146,13 @@ static int __send_signal_locked(int sig,
 			clear_siginfo(&q->info);
 			q->info.si_signo = sig;
 			q->info.si_errno = 0;
-			q->info.si_code = SI_KERNEL;
+			code = q->info.si_code = SI_KERNEL;
 			q->info.si_pid = 0;
 			q->info.si_uid = 0;
 			break;
 		default:
 			copy_siginfo(&q->info, info);
+			code = info->si_code;
 			break;
 		}
 	} else if (!is_si_special(info) &&
@@ -1186,7 +1191,7 @@ out_set:
 		}
 	}
 
-	complete_signal(sig, t, type);
+	complete_signal(sig, code, t, type);
 ret:
 	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
 	return ret;
@@ -1960,6 +1965,7 @@ void sigqueue_free(struct sigqueue *q)
 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
 {
 	int sig = q->info.si_signo;
+	int code = q->info.si_code;
 	struct sigpending *pending;
 	struct task_struct *t;
 	unsigned long flags;
@@ -1995,7 +2001,7 @@ int send_sigqueue(struct sigqueue *q, st
 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
 	list_add_tail(&q->list, &pending->list);
 	sigaddset(&pending->signal, sig);
-	complete_signal(sig, t, type);
+	complete_signal(sig, code, t, type);
 	result = TRACE_SIGNAL_DELIVERED;
 out:
 	trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
@@ -2380,7 +2386,7 @@ int ptrace_notify(int exit_code, unsigne
  * %false if group stop is already cancelled or ptrace trap is scheduled.
  * %true if participated in group stop.
  */
-static bool do_signal_stop(int signr)
+static bool do_signal_stop(int signr, int sicode)
 	__releases(&current->sighand->siglock)
 {
 	struct signal_struct *sig = current->signal;
@@ -2415,8 +2421,10 @@ static bool do_signal_stop(int signr)
 		 * an intervening stop signal is required to cause two
 		 * continued events regardless of ptrace.
 		 */
-		if (!(sig->flags & SIGNAL_STOP_STOPPED))
+		if (!(sig->flags & SIGNAL_STOP_STOPPED)) {
 			sig->group_exit_code = signr;
+			sig->group_exit_sicode = sicode;
+		}
 
 		sig->group_stop_count = 0;
 
@@ -2701,7 +2709,7 @@ relock:
 		}
 
 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
-		    do_signal_stop(0))
+		    do_signal_stop(0, 0))
 			goto relock;
 
 		if (unlikely(current->jobctl &
@@ -2806,7 +2814,8 @@ relock:
 				spin_lock_irq(&sighand->siglock);
 			}
 
-			if (likely(do_signal_stop(ksig->info.si_signo))) {
+			if (likely(do_signal_stop(ksig->info.si_signo,
+						  ksig->info.si_code))) {
 				/* It released the siglock.  */
 				goto relock;
 			}
@@ -2854,7 +2863,7 @@ relock:
 		/*
 		 * Death signals, no core dump.
 		 */
-		do_group_exit(ksig->info.si_signo);
+		do_group_exit(ksig->info.si_signo, ksig->info.si_code);
 		/* NOTREACHED */
 	}
 	spin_unlock_irq(&sighand->siglock);
_