diff mbox series

[bpf-next] bpf: increment and use correct thread iterator

Message ID 20201203215907.975053-1-jonathan.lemon@gmail.com (mailing list archive)
State Superseded
Delegated to: BPF
Headers show
Series [bpf-next] bpf: increment and use correct thread iterator | expand

Checks

Context Check Description
netdev/cover_letter success Link
netdev/fixes_present success Link
netdev/patch_count success Link
netdev/tree_selection success Clearly marked for bpf-next
netdev/subject_prefix success Link
netdev/source_inline success Was 0 now: 0
netdev/verify_signedoff success Link
netdev/module_param success Was 0 now: 0
netdev/build_32bit success Errors and warnings before: 1 this patch: 1
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/verify_fixes fail Link
netdev/checkpatch warning WARNING: From:/Signed-off-by: email address mismatch: 'From: Jonathan Lemon <bsd@fb.com>' != 'Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>' WARNING: Unknown commit id '67b6b863e6ab', maybe rebased or not pulled? WARNING: Unknown commit id 'a650da2ee52a', maybe rebased or not pulled?
netdev/build_allmodconfig_warn success Errors and warnings before: 1 this patch: 1
netdev/header_inline success Link
netdev/stable success Stable not CCed

Commit Message

Jonathan Lemon Dec. 3, 2020, 9:59 p.m. UTC 
From: Jonathan Lemon <bsd@fb.com>

If unable to obtain the file structure for the current task,
proceed to the next task number after the one returned from
task_seq_get_next(), instead of the next task number from the
original iterator.

Use thread_group_leader() instead of comparing tgid vs pid, which
might may be racy.

Only obtain the task reference count at the end of the RCU section
instead of repeatedly obtaining/releasing it when iterathing though
a thread group.

This patch fixes a recurring RCU stall seen from task_file_seq_next().

Fixes: a650da2ee52a ("bpf: Add task and task/file iterator targets")
Fixes: 67b6b863e6ab ("bpf: Avoid iterating duplicated files for task_file iterator")

Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
---
 kernel/bpf/task_iter.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

Comments

Alexei Starovoitov Dec. 4, 2020, 1:58 a.m. UTC | #1 
On Thu, Dec 03, 2020 at 01:59:07PM -0800, Jonathan Lemon wrote:
> From: Jonathan Lemon <bsd@fb.com>
> 
> If unable to obtain the file structure for the current task,
> proceed to the next task number after the one returned from
> task_seq_get_next(), instead of the next task number from the
> original iterator.
> 
> Use thread_group_leader() instead of comparing tgid vs pid, which
> might may be racy.
> 
> Only obtain the task reference count at the end of the RCU section
> instead of repeatedly obtaining/releasing it when iterathing though
> a thread group.
> 
> This patch fixes a recurring RCU stall seen from task_file_seq_next().
> 
> Fixes: a650da2ee52a ("bpf: Add task and task/file iterator targets")
> Fixes: 67b6b863e6ab ("bpf: Avoid iterating duplicated files for task_file iterator")

There are no such sha-s in bpf-next tree. Pls fix.
diff mbox series

Patch

diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
index 0458a40edf10..66a52fcf589a 100644
--- a/kernel/bpf/task_iter.c
+++ b/kernel/bpf/task_iter.c
@@ -33,17 +33,17 @@  static struct task_struct *task_seq_get_next(struct pid_namespace *ns,
 	pid = find_ge_pid(*tid, ns);
 	if (pid) {
 		*tid = pid_nr_ns(pid, ns);
-		task = get_pid_task(pid, PIDTYPE_PID);
+		task = pid_task(pid, PIDTYPE_PID);
 		if (!task) {
 			++*tid;
 			goto retry;
-		} else if (skip_if_dup_files && task->tgid != task->pid &&
+		} else if (skip_if_dup_files && !thread_group_leader(task) &&
 			   task->files == task->group_leader->files) {
-			put_task_struct(task);
 			task = NULL;
 			++*tid;
 			goto retry;
 		}
+		get_task_struct(task);
 	}
 	rcu_read_unlock();
 
@@ -164,7 +164,7 @@  task_file_seq_get_next(struct bpf_iter_seq_task_file_info *info)
 		curr_files = get_files_struct(curr_task);
 		if (!curr_files) {
 			put_task_struct(curr_task);
-			curr_tid = ++(info->tid);
+			curr_tid = curr_tid + 1;
 			info->fd = 0;
 			goto again;
 		}