[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20250201205127.GA1191798@lichtman.org>
Date: Sat, 1 Feb 2025 20:51:27 +0000
From: Nir Lichtman <nir@...htman.org>
To: viro@...iv.linux.org.uk, brauner@...nel.org, jack@...e.cz,
kees@...nel.org, ebiederm@...ssion.com,
linux-fsdevel@...r.kernel.org, linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: [PATCH] exec: improve clarity of comment regarding pid save-asides
Problem: Current comment regarding saving aside the old pid and old vpid
is very vague, especially when considering that it is unexpected that
execve can actually kill the current process.
Solution: Improve the description of the comment explaining more
in-depth the reasoning behind.
Suggested-by: Kees Cook <kees@...nel.org>
Signed-off-by: Nir Lichtman <nir@...htman.org>
---
fs/exec.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/fs/exec.c b/fs/exec.c
index 506cd411f4ac..343c435b00ee 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1792,7 +1792,14 @@ static int exec_binprm(struct linux_binprm *bprm)
pid_t old_pid, old_vpid;
int ret, depth;
- /* Need to fetch pid before load_binary changes it */
+ /*
+ * Need to save aside the current thread pid and vpid
+ * since if the current thread is not a thread group leader
+ * the logic in de_thread kills the current thread and all
+ * other threads in the current thread group, except the leader.
+ * The new program will execute in the leader, with the leader pid
+ * ("man 2 clone" CLONE_THREAD flag for more info)
+ */
old_pid = current->pid;
rcu_read_lock();
old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
--
2.39.5
Powered by blists - more mailing lists