Skip to contentSkip to conversation

kernel/sys.c: use RCU when accessing task_struct->real_parent

PATCH In review Alex Elder 3 participants 5 messages Aug 11 – Aug 12

lore
Share & tools
lore

Alex Elder

author
#0 lore raw

PATCH 1/1 kernel/sys.c: use RCU when accessing task_struct->real_parent

In the setpgid() syscall definition, a check is made to determine whether the target process is in the same thread group as the current process. The check accesses the target process's real_parent pointer directly, however that field is supposed to be accessed via RCU. Use rcu_dereference() to avoid this C=1 build warning:

kernel/sys.c:1144:32: warning: incorrect type in argument 1 (different address spaces)

---
 kernel/sys.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/sys.c +3 -1
index df69bd71de032..434fe6ad30e13 100644
@@ -1114,6 +1114,7 @@ COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 {
 	struct task_struct *p;
+	struct task_struct *real_parent;
 	struct task_struct *group_leader = current->group_leader;
 	struct pid *pids[PIDTYPE_MAX] = { 0 };
 	struct pid *pgrp;
@@ -1141,7 +1142,8 @@ SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 	if (!thread_group_leader(p))
 		goto out;
 
-	if (same_thread_group(p->real_parent, group_leader)) {
+	real_parent = rcu_dereference(p->real_parent);
+	if (same_thread_group(real_parent, group_leader)) {
 		err = -EPERM;
 		if (task_session(p) != task_session(group_leader))
 			goto out;

base-commit: 6b8c8af514d739d0335f5579b585e02babe8a727

Signature · 1 line
2.53.0
#1 lore raw
Quoted 7 lines · On Tue, 11 Aug 2026 14:05:00 -0500 Alex Elder <[email protected]> wrote:

In the setpgid() syscall definition, a check is made to determine whether the target process is in the same thread group as the current process. The check accesses the target process's real_parent pointer directly, however that field is supposed to be accessed via RCU. Use rcu_dereference() to avoid this C=1 build warning:

kernel/sys.c:1144:32: warning: incorrect type in argument 1 (different address spaces)

Oh.

Why don't all those other uses of ->real_parent produce this warning?

Quoted 21 lines
kernel/sys.c +3 -1
@@ -1114,6 +1114,7 @@ COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 {
 	struct task_struct *p;
+	struct task_struct *real_parent;
 	struct task_struct *group_leader = current->group_leader;
 	struct pid *pids[PIDTYPE_MAX] = { 0 };
 	struct pid *pgrp;
@@ -1141,7 +1142,8 @@ SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 	if (!thread_group_leader(p))
 		goto out;
 
-	if (same_thread_group(p->real_parent, group_leader)) {
+	real_parent = rcu_dereference(p->real_parent);
+	if (same_thread_group(real_parent, group_leader)) {
 		err = -EPERM;
 		if (task_session(p) != task_session(group_leader))
 			goto out;
author
#2 lore raw
Quoted 8 lines · Andrew Morton
Quoted 6 lines · On Tue, 11 Aug 2026 14:05:00 -0500 Alex Elder <[email protected]> wrote:

In the setpgid() syscall definition, a check is made to determine whether the target process is in the same thread group as the current process. The check accesses the target process's real_parent pointer directly, however that field is supposed to be accessed via RCU. Use rcu_dereference() to avoid this C=1 build warning:

Quoted 5 lines
Quoted 1 line

kernel/sys.c:1144:32: warning: incorrect type in argument 1 (different address spaces)

Oh.

Why don't all those other uses of ->real_parent produce this warning?

Looks like there are others; this was just the one that kept showing up in my builds. Maybe some cases simply don't need the rcu_dereference() (like initializing init_task.real_parent in init/init_task.c).

It's inconsistent. Some files use rcu_dereference() always for task_struct->real_parent:
- drivers/connector/cn_proc.c
- fs/binfmt_elf.c
- fs/nfs/dir.c
- fs/binfmt_elf_fdpic.c
- kernel/sched/core.c
- kernel/acct.c
- security/keys/keyctl.c
- security/selinux/hooks.c
- security/yama/yama_lsm.c

But others do not. And there are lots of other fields annotated with __rcu that don't always use rcu_dereference():
- task_struct->parent
- task_struct->sighand
- mm_struct->owner
and so on.

I could go dig a little deeper and solve this more comprehensively. It would probably have to be a long-running background effort. Do you think it's worth it?

If not, I don't mind just retracting this suggestion for now.

-Alex

Quoted 25 lines
Quoted 23 lines
kernel/sys.c +3 -1
@@ -1114,6 +1114,7 @@ COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
  SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
  {
  	struct task_struct *p;
+	struct task_struct *real_parent;
  	struct task_struct *group_leader = current->group_leader;
  	struct pid *pids[PIDTYPE_MAX] = { 0 };
  	struct pid *pgrp;
@@ -1141,7 +1142,8 @@ SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
  	if (!thread_group_leader(p))
  		goto out;
 
-	if (same_thread_group(p->real_parent, group_leader)) {
+	real_parent = rcu_dereference(p->real_parent);
+	if (same_thread_group(real_parent, group_leader)) {
  		err = -EPERM;
  		if (task_session(p) != task_session(group_leader))
  			goto out;
#3 lore raw
In reply to Alex Elder

Just in case, I am travelling without my work laptop until Aug 19, can't read the code and I rarely read emails...

Quoted 8 lines · Alex Elder

In the setpgid() syscall definition, a check is made to determine whether the target process is in the same thread group as the current process. The check accesses the target process's real_parent pointer directly, however that field is supposed to be accessed via RCU. Use rcu_dereference() to avoid this C=1 build warning:

kernel/sys.c:1144:32: warning: incorrect type in argument 1 (different address spaces)

This is the sparse warning... Do we really care? We have a lot more users of ->real_parent without rcu_dereference().

Note also that the "rcu" annotation of ->real_parent is misleading. If the task exits, task->real_parent points to nowhere. Same for ->group_leader.

This reminds me... months ago I was going to introduce the helper to access ->group_leader and move it to signal_struct. Then I was going to do the same with ->real_parent. I sent some preparations, but then I was distracted. I'll try to return to this after PTO.

Quoted 8 lines

SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) {

kernel/sys.c +1 -0
 	struct task_struct *p;
+	struct task_struct *real_parent;
 	struct task_struct *group_leader = current->group_leader;
 	struct pid *pids[PIDTYPE_MAX] = { 0 };
 	struct pid *pgrp;

kernel/sys.c SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) · line 1149

Quoted 9 lines
kernel/sys.c +2 -1
@@ -1141,7 +1142,8 @@ SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
 	if (!thread_group_leader(p))
 		goto out;
 
-	if (same_thread_group(p->real_parent, group_leader)) {
+	real_parent = rcu_dereference(p->real_parent);
+	if (same_thread_group(real_parent, group_leader)) {
 		err = -EPERM;
 		if (task_session(p) != task_session(group_leader))
 			goto out;

IIRC, this code runs under tasklist_lock (at least it should ;) ->real_parent is stable.

I'd prefer to leave this code as is (see above), but even if we really want to shut up sparse we don't need rcu_dereference() anyway, we are not going to dereference this pointer.

We have other helpers to read the "rcu" pointers, but I can't recall the names...

Oleg.

author
#4 lore raw
Quoted 11 lines · Oleg Nesterov

Just in case, I am travelling without my work laptop until Aug 19, can't read the code and I rarely read emails...

Quoted 7 lines · Alex Elder

In the setpgid() syscall definition, a check is made to determine whether the target process is in the same thread group as the current process. The check accesses the target process's real_parent pointer directly, however that field is supposed to be accessed via RCU. Use rcu_dereference() to avoid this C=1 build warning:

Quoted 8 lines
Quoted 1 line

kernel/sys.c:1144:32: warning: incorrect type in argument 1 (different address spaces)

This is the sparse warning... Do we really care? We have a lot more users of ->real_parent without rcu_dereference().

Note also that the "rcu" annotation of ->real_parent is misleading. If the task exits, task->real_parent points to nowhere. Same for ->group_leader.

Right, there are some cases where there is other protection as well (like the one you point out below).

Quoted 4 lines

This reminds me... months ago I was going to introduce the helper to access ->group_leader and move it to signal_struct. Then I was going to do the same with ->real_parent. I sent some preparations, but then I was distracted. I'll try to return to this after PTO.

Yes, this is a better solution overall. I sent this patch to address one warning I kept seeing in my builds, but Andrew pointed out there are lots of others that would also warn.

Quoted 21 lines
Quoted 8 lines
  SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
  {
  	struct task_struct *p;
kernel/sys.c +1 -0
+	struct task_struct *real_parent;
  	struct task_struct *group_leader = current->group_leader;
  	struct pid *pids[PIDTYPE_MAX] = { 0 };
  	struct pid *pgrp;

kernel/sys.c SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) · line 1149

Quoted 9 lines
kernel/sys.c +2 -1
@@ -1141,7 +1142,8 @@ SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
  	if (!thread_group_leader(p))
  		goto out;
 
-	if (same_thread_group(p->real_parent, group_leader)) {
+	real_parent = rcu_dereference(p->real_parent);
+	if (same_thread_group(real_parent, group_leader)) {
  		err = -EPERM;
  		if (task_session(p) != task_session(group_leader))
  			goto out;

IIRC, this code runs under tasklist_lock (at least it should ;) ->real_parent is stable.

Quoted 3 lines

I'd prefer to leave this code as is (see above), but even if we really want to shut up sparse we don't need rcu_dereference() anyway, we are not going to dereference this pointer.

OK that's perfectly fine with me. I retract this patch, and will look forward to you introducing helpers to address this in a more general way.

Thanks a lot.

-Alex

Quoted 5 lines

We have other helpers to read the "rcu" pointers, but I can't recall the names...

Oleg.