Skip to contentSkip to conversation

bpf: Drop unprotected find_vpid() in favour of find_get_pid()

PATCH Superseded Lee Jones 4 participants 9 messages Aug 3, 2022 – Aug 10, 2022

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Lee Jones

author
#0 lore raw

PATCH v2 1/1 bpf: Drop unprotected find_vpid() in favour of find_get_pid()

The documentation for find_pid() clearly states:

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

---

v1 => v2:
* Commit log update - no code differences

 kernel/bpf/syscall.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +4 -1
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
 	const struct perf_event *event;
 	struct task_struct *task;
 	struct file *file;
+	struct pid *ppid;
 	int err;
 
 	if (CHECK_ATTR(BPF_TASK_FD_QUERY))
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
 	if (attr->task_fd_query.flags != 0)
 		return -EINVAL;
 
-	task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+	ppid = find_get_pid(pid);
+	task = get_pid_task(ppid, PIDTYPE_PID);
+	put_pid(ppid);
 	if (!task)
 		return -ENOENT;
 
Signature · 1 line
2.37.1.455.g008518b4e5-goog
Acked-by on v2 1/1
#1 lore raw
Quoted 1 line · Lee Jones

The documentation for find_pid() clearly states:

nit: typo find_vpid

Quoted 10 lines

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Quoted 14 lines
Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

jirka

Quoted 34 lines

---

v1 => v2:
* Commit log update - no code differences

 kernel/bpf/syscall.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +4 -1
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
 	const struct perf_event *event;
 	struct task_struct *task;
 	struct file *file;
+	struct pid *ppid;
 	int err;
 
 	if (CHECK_ATTR(BPF_TASK_FD_QUERY))
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
 	if (attr->task_fd_query.flags != 0)
 		return -EINVAL;
 
-	task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+	ppid = find_get_pid(pid);
+	task = get_pid_task(ppid, PIDTYPE_PID);
+	put_pid(ppid);
 	if (!task)
 		return -ENOENT;
 
Signature · 1 line
2.37.1.455.g008518b4e5-goog
#3 lore raw
In reply to Lee Jones
Quoted 47 lines · Lee Jones

The documentation for find_pid() clearly states:

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

---

v1 => v2:
* Commit log update - no code differences

 kernel/bpf/syscall.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +1 -0
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        const struct perf_event *event;
        struct task_struct *task;
        struct file *file;
+       struct pid *ppid;
        int err;
 
        if (CHECK_ATTR(BPF_TASK_FD_QUERY))

kernel/bpf/syscall.c static int bpf_task_fd_query(const union bpf_attr *attr, · line 4402

Quoted 7 lines
kernel/bpf/syscall.c +3 -1
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        if (attr->task_fd_query.flags != 0)
                return -EINVAL;
 
-       task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+       ppid = find_get_pid(pid);
+       task = get_pid_task(ppid, PIDTYPE_PID);
+       put_pid(ppid);

rcu_read_lock/unlock around this line
would be a cheaper and faster alternative than pid's refcount inc/dec.

author
#4 lore raw
Quoted 51 lines · Alexei Starovoitov
Quoted 50 lines · Lee Jones

The documentation for find_pid() clearly states:

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

---

v1 => v2:
* Commit log update - no code differences

 kernel/bpf/syscall.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +1 -0
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        const struct perf_event *event;
        struct task_struct *task;
        struct file *file;
+       struct pid *ppid;
        int err;
 
        if (CHECK_ATTR(BPF_TASK_FD_QUERY))
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        if (attr->task_fd_query.flags != 0)
                return -EINVAL;
 
Quoted 8 lines

kernel/bpf/syscall.c static int bpf_task_fd_query(const union bpf_attr *attr, · line 4402

Quoted 4 lines
kernel/bpf/syscall.c +3 -1
-       task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+       ppid = find_get_pid(pid);
+       task = get_pid_task(ppid, PIDTYPE_PID);
+       put_pid(ppid);

rcu_read_lock/unlock around this line
would be a cheaper and faster alternative than pid's refcount inc/dec.

This was already discussed here:

https://lore.kernel.org/all/YtsFT1yFtb7UW2Xu@krava/

Signature · 1 line
Lee Jones [李琼斯]
author
#2 lore raw
In reply to Jiri Olsa
Quoted 4 lines · Jiri Olsa
Quoted 1 line · Lee Jones

The documentation for find_pid() clearly states:

nit: typo find_vpid

Sorry missed this.

Will fix.

Quoted 24 lines
Quoted 9 lines

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Quoted 14 lines
Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")
Quoted 1 line

Thanks.

Signature · 1 line
Lee Jones [李琼斯]
#5 lore raw
In reply to Lee Jones
Quoted 63 lines · Lee Jones
Quoted 51 lines · Alexei Starovoitov
Quoted 50 lines · Lee Jones

The documentation for find_pid() clearly states:

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

---

v1 => v2:
* Commit log update - no code differences

 kernel/bpf/syscall.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +1 -0
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        const struct perf_event *event;
        struct task_struct *task;
        struct file *file;
+       struct pid *ppid;
        int err;
 
        if (CHECK_ATTR(BPF_TASK_FD_QUERY))
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        if (attr->task_fd_query.flags != 0)
                return -EINVAL;
 
Quoted 8 lines

kernel/bpf/syscall.c static int bpf_task_fd_query(const union bpf_attr *attr, · line 4402

Quoted 4 lines
kernel/bpf/syscall.c +3 -1
-       task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+       ppid = find_get_pid(pid);
+       task = get_pid_task(ppid, PIDTYPE_PID);
+       put_pid(ppid);

rcu_read_lock/unlock around this line
would be a cheaper and faster alternative than pid's refcount inc/dec.

Quoted 3 lines

This was already discussed here:

https://lore.kernel.org/all/YtsFT1yFtb7UW2Xu@krava/

Since several people thought about rcu_read_lock instead of your approach it means that it's preferred.
Sooner or later somebody will send a patch to optimize refcnt into rcu_read_lock.
So let's avoid the churn and do it now.

author
#6 lore raw
Quoted 66 lines · Alexei Starovoitov
Quoted 65 lines · Lee Jones
Quoted 51 lines · Alexei Starovoitov
Quoted 50 lines · Lee Jones

The documentation for find_pid() clearly states:

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

---

v1 => v2:
* Commit log update - no code differences

 kernel/bpf/syscall.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +1 -0
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        const struct perf_event *event;
        struct task_struct *task;
        struct file *file;
+       struct pid *ppid;
        int err;
 
        if (CHECK_ATTR(BPF_TASK_FD_QUERY))
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        if (attr->task_fd_query.flags != 0)
                return -EINVAL;
 
Quoted 8 lines

kernel/bpf/syscall.c static int bpf_task_fd_query(const union bpf_attr *attr, · line 4402

Quoted 4 lines
kernel/bpf/syscall.c +3 -1
-       task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+       ppid = find_get_pid(pid);
+       task = get_pid_task(ppid, PIDTYPE_PID);
+       put_pid(ppid);

rcu_read_lock/unlock around this line
would be a cheaper and faster alternative than pid's refcount inc/dec.

This was already discussed here:

Quoted 7 lines
Quoted 1 line

https://lore.kernel.org/all/YtsFT1yFtb7UW2Xu@krava/

Since several people thought about rcu_read_lock instead of your approach it means that it's preferred.
Sooner or later somebody will send a patch to optimize refcnt into rcu_read_lock.
So let's avoid the churn and do it now.

I'm not wed to either approach. Please discuss it with Yonghong and Jiri and I'll do whatever is agreed upon.

Signature · 1 line
Lee Jones [李琼斯]
#7 lore raw
Quoted 70 lines · Lee Jones
Quoted 68 lines · Alexei Starovoitov
Quoted 66 lines · Lee Jones
Quoted 51 lines · Alexei Starovoitov
Quoted 50 lines · Lee Jones

The documentation for find_pid() clearly states:

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

---

v1 => v2:
* Commit log update - no code differences

 kernel/bpf/syscall.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +1 -0
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        const struct perf_event *event;
        struct task_struct *task;
        struct file *file;
+       struct pid *ppid;
        int err;
 
        if (CHECK_ATTR(BPF_TASK_FD_QUERY))
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
        if (attr->task_fd_query.flags != 0)
                return -EINVAL;
 
Quoted 8 lines

kernel/bpf/syscall.c static int bpf_task_fd_query(const union bpf_attr *attr, · line 4402

Quoted 4 lines
kernel/bpf/syscall.c +3 -1
-       task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+       ppid = find_get_pid(pid);
+       task = get_pid_task(ppid, PIDTYPE_PID);
+       put_pid(ppid);

rcu_read_lock/unlock around this line
would be a cheaper and faster alternative than pid's refcount inc/dec.

This was already discussed here:

https://lore.kernel.org/all/YtsFT1yFtb7UW2Xu@krava/

Quoted 8 lines
Quoted 5 lines

Since several people thought about rcu_read_lock instead of your approach it means that it's preferred.
Sooner or later somebody will send a patch to optimize refcnt into rcu_read_lock.
So let's avoid the churn and do it now.

I'm not wed to either approach. Please discuss it with Yonghong and Jiri and I'll do whatever is agreed upon.

yea, I thought using rcu_read_lock would be better, but I did not have strong feelings against doing the pid's refcount inc/dec when Yonghong supported that.. now with Alexei it's 2 against 1 in favour of using rcu_read_lock ;-)

jirka

#8 lore raw
In reply to Lee Jones
Quoted 70 lines · Lee Jones
Quoted 68 lines · Alexei Starovoitov
Quoted 66 lines · Lee Jones
Quoted 51 lines · Alexei Starovoitov
Quoted 50 lines · Lee Jones

The documentation for find_pid() clearly states:

"Must be called with the tasklist_lock or rcu_read_lock() held."

Presently we do neither.

Let's use find_get_pid() which searches for the vpid, then takes a reference to it preventing early free, all within the safety of rcu_read_lock(). Once we have our reference we can safely make use of it up until the point it is put.

Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc John Fastabend <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Martin KaFai Lau <[email protected]>
Cc Song Liu <[email protected]>
Cc Yonghong Song <[email protected]>
Cc KP Singh <[email protected]>
Cc Stanislav Fomichev <[email protected]>
Cc Hao Luo <[email protected]>
Cc Jiri Olsa <[email protected]>
Fixes 41bdc4b40ed6f ("bpf: introduce bpf subcommand BPF_TASK_FD_QUERY")

---

v1 => v2:
* Commit log update - no code differences

  kernel/bpf/syscall.c | 5 ++++-
  1 file changed, 4 insertions(+), 1 deletion(-)
kernel/bpf/syscall.c +1 -0
index 83c7136c5788d..c20cff30581c4 100644
@@ -4385,6 +4385,7 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
         const struct perf_event *event;
         struct task_struct *task;
         struct file *file;
+       struct pid *ppid;
         int err;
 
         if (CHECK_ATTR(BPF_TASK_FD_QUERY))
@@ -4396,7 +4397,9 @@ static int bpf_task_fd_query(const union bpf_attr *attr,
         if (attr->task_fd_query.flags != 0)
                 return -EINVAL;
 
Quoted 8 lines

kernel/bpf/syscall.c static int bpf_task_fd_query(const union bpf_attr *attr, · line 4402

Quoted 4 lines
kernel/bpf/syscall.c +3 -1
-       task = get_pid_task(find_vpid(pid), PIDTYPE_PID);
+       ppid = find_get_pid(pid);
+       task = get_pid_task(ppid, PIDTYPE_PID);
+       put_pid(ppid);

rcu_read_lock/unlock around this line
would be a cheaper and faster alternative than pid's refcount inc/dec.

This was already discussed here:

https://lore.kernel.org/all/YtsFT1yFtb7UW2Xu@krava/

Quoted 8 lines
Quoted 5 lines

Since several people thought about rcu_read_lock instead of your approach it means that it's preferred.
Sooner or later somebody will send a patch to optimize refcnt into rcu_read_lock.
So let's avoid the churn and do it now.

I'm not wed to either approach. Please discuss it with Yonghong and Jiri and I'll do whatever is agreed upon.

Hi, Lee, Let us just do rcu_read_lock() approach then. I am okay with that. Thanks!

Quoted 1 line