From: Kui-Feng Lee <kuifeng@meta.com>
To: <bpf@vger.kernel.org>, <ast@kernel.org>, <andrii@kernel.org>,
<kernel-team@meta.com>, <song@kernel.org>, <yhs@meta.com>
Cc: Kui-Feng Lee <kuifeng@meta.com>,
Nathan Slingerland <slinger@meta.com>, Yonghong Song <yhs@fb.com>
Subject: [PATCH bpf-next v2 1/2] bpf: keep a reference to the mm, in case the task is dead.
Date: Fri, 16 Dec 2022 14:18:54 -0800 [thread overview]
Message-ID: <20221216221855.4122288-2-kuifeng@meta.com> (raw)
In-Reply-To: <20221216221855.4122288-1-kuifeng@meta.com>
Fix the system crash that happens when a task iterator travel through
vma of tasks.
In task iterators, we used to access mm by following the pointer on
the task_struct; however, the death of a task will clear the pointer,
even though we still hold the task_struct. That can cause an
unexpected crash for a null pointer when an iterator is visiting a
task that dies during the visit. Keeping a reference of mm on the
iterator ensures we always have a valid pointer to mm.
Co-developed-by: Song Liu <song@kernel.org>
Signed-off-by: Song Liu <song@kernel.org>
Signed-off-by: Kui-Feng Lee <kuifeng@meta.com>
Reported-by: Nathan Slingerland <slinger@meta.com>
Acked-by: Yonghong Song <yhs@fb.com>
---
kernel/bpf/task_iter.c | 39 +++++++++++++++++++++++++++------------
1 file changed, 27 insertions(+), 12 deletions(-)
diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
index c2a2182ce570..c4ab9d6cdbe9 100644
--- a/kernel/bpf/task_iter.c
+++ b/kernel/bpf/task_iter.c
@@ -438,6 +438,7 @@ struct bpf_iter_seq_task_vma_info {
*/
struct bpf_iter_seq_task_common common;
struct task_struct *task;
+ struct mm_struct *mm;
struct vm_area_struct *vma;
u32 tid;
unsigned long prev_vm_start;
@@ -456,16 +457,19 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
enum bpf_task_vma_iter_find_op op;
struct vm_area_struct *curr_vma;
struct task_struct *curr_task;
+ struct mm_struct *curr_mm;
u32 saved_tid = info->tid;
/* If this function returns a non-NULL vma, it holds a reference to
- * the task_struct, and holds read lock on vma->mm->mmap_lock.
+ * the task_struct, holds a refcount on mm->mm_users, and holds
+ * read lock on vma->mm->mmap_lock.
* If this function returns NULL, it does not hold any reference or
* lock.
*/
if (info->task) {
curr_task = info->task;
curr_vma = info->vma;
+ curr_mm = info->mm;
/* In case of lock contention, drop mmap_lock to unblock
* the writer.
*
@@ -504,13 +508,15 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
* 4.2) VMA2 and VMA2' covers different ranges, process
* VMA2'.
*/
- if (mmap_lock_is_contended(curr_task->mm)) {
+ if (mmap_lock_is_contended(curr_mm)) {
info->prev_vm_start = curr_vma->vm_start;
info->prev_vm_end = curr_vma->vm_end;
op = task_vma_iter_find_vma;
- mmap_read_unlock(curr_task->mm);
- if (mmap_read_lock_killable(curr_task->mm))
+ mmap_read_unlock(curr_mm);
+ if (mmap_read_lock_killable(curr_mm)) {
+ mmput(curr_mm);
goto finish;
+ }
} else {
op = task_vma_iter_next_vma;
}
@@ -535,42 +541,47 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
op = task_vma_iter_find_vma;
}
- if (!curr_task->mm)
+ curr_mm = get_task_mm(curr_task);
+ if (!curr_mm)
goto next_task;
- if (mmap_read_lock_killable(curr_task->mm))
+ if (mmap_read_lock_killable(curr_mm)) {
+ mmput(curr_mm);
goto finish;
+ }
}
switch (op) {
case task_vma_iter_first_vma:
- curr_vma = find_vma(curr_task->mm, 0);
+ curr_vma = find_vma(curr_mm, 0);
break;
case task_vma_iter_next_vma:
- curr_vma = find_vma(curr_task->mm, curr_vma->vm_end);
+ curr_vma = find_vma(curr_mm, curr_vma->vm_end);
break;
case task_vma_iter_find_vma:
/* We dropped mmap_lock so it is necessary to use find_vma
* to find the next vma. This is similar to the mechanism
* in show_smaps_rollup().
*/
- curr_vma = find_vma(curr_task->mm, info->prev_vm_end - 1);
+ curr_vma = find_vma(curr_mm, info->prev_vm_end - 1);
/* case 1) and 4.2) above just use curr_vma */
/* check for case 2) or case 4.1) above */
if (curr_vma &&
curr_vma->vm_start == info->prev_vm_start &&
curr_vma->vm_end == info->prev_vm_end)
- curr_vma = find_vma(curr_task->mm, curr_vma->vm_end);
+ curr_vma = find_vma(curr_mm, curr_vma->vm_end);
break;
}
if (!curr_vma) {
/* case 3) above, or case 2) 4.1) with vma->next == NULL */
- mmap_read_unlock(curr_task->mm);
+ mmap_read_unlock(curr_mm);
+ mmput(curr_mm);
goto next_task;
}
info->task = curr_task;
info->vma = curr_vma;
+ info->mm = curr_mm;
return curr_vma;
next_task:
@@ -579,6 +590,7 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
put_task_struct(curr_task);
info->task = NULL;
+ info->mm = NULL;
info->tid++;
goto again;
@@ -587,6 +599,7 @@ task_vma_seq_get_next(struct bpf_iter_seq_task_vma_info *info)
put_task_struct(curr_task);
info->task = NULL;
info->vma = NULL;
+ info->mm = NULL;
return NULL;
}
@@ -658,7 +671,9 @@ static void task_vma_seq_stop(struct seq_file *seq, void *v)
*/
info->prev_vm_start = ~0UL;
info->prev_vm_end = info->vma->vm_end;
- mmap_read_unlock(info->task->mm);
+ mmap_read_unlock(info->mm);
+ mmput(info->mm);
+ info->mm = NULL;
put_task_struct(info->task);
info->task = NULL;
}
--
2.30.2
next prev parent reply other threads:[~2022-12-16 22:21 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-12-16 22:18 [PATCH bpf-next v2 0/2] bpf: fix the crash caused by task iterators over vma Kui-Feng Lee
2022-12-16 22:18 ` Kui-Feng Lee [this message]
2022-12-16 22:18 ` [PATCH bpf-next v2 2/2] selftests/bpf: add a test for iter/task_vma for short-lived processes Kui-Feng Lee
2022-12-28 22:20 ` [PATCH bpf-next v2 0/2] bpf: fix the crash caused by task iterators over vma patchwork-bot+netdevbpf
2023-01-04 21:16 ` Kui-Feng Lee
2023-01-04 21:21 ` Alexei Starovoitov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20221216221855.4122288-2-kuifeng@meta.com \
--to=kuifeng@meta.com \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=kernel-team@meta.com \
--cc=slinger@meta.com \
--cc=song@kernel.org \
--cc=yhs@fb.com \
--cc=yhs@meta.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox