* Re: [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages()
@ 2020-08-20 2:18 linmiaohe
2020-08-20 21:21 ` Kees Cook
0 siblings, 1 reply; 6+ messages in thread
From: linmiaohe @ 2020-08-20 2:18 UTC (permalink / raw)
To: Kees Cook; +Cc: akpm, linux-mm, linux-kernel, linux-security-module
Kees Cook <keescook@chromium.org> wrote:
>On Mon, Aug 17, 2020 at 07:59:33AM -0400, Miaohe Lin wrote:
>> There is no need to check if this process has the right to modify the
>> specified process when they are same.
>>
>> Signed-off-by: Hongxiang Lou <louhongxiang@huawei.com>
>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>
>NAK, please don't do this -- the ptrace and security hooks already do these kinds of self-introspection checks, and I'd like to keep a central place to perform these kinds of checks.
>
Many thanks for your reply.
We also avoid get_task_struct/ put_task_struct pair of atomic ops, rcu_lock, task_lock and so on this way.
>Is there a specific problem you've encountered that this fixes?
>
I'am sorry but there's no specific problem. I do this mainly to skip the unnecessary ptrace and security hooks.
>--
>Kees Cook
Thanks again.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages()
2020-08-20 2:18 [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages() linmiaohe
@ 2020-08-20 21:21 ` Kees Cook
0 siblings, 0 replies; 6+ messages in thread
From: Kees Cook @ 2020-08-20 21:21 UTC (permalink / raw)
To: linmiaohe; +Cc: akpm, linux-mm, linux-kernel, linux-security-module
On Thu, Aug 20, 2020 at 02:18:21AM +0000, linmiaohe wrote:
> Kees Cook <keescook@chromium.org> wrote:
> >On Mon, Aug 17, 2020 at 07:59:33AM -0400, Miaohe Lin wrote:
> >> There is no need to check if this process has the right to modify the
> >> specified process when they are same.
> >>
> >> Signed-off-by: Hongxiang Lou <louhongxiang@huawei.com>
> >> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> >
> >NAK, please don't do this -- the ptrace and security hooks already do these kinds of self-introspection checks, and I'd like to keep a central place to perform these kinds of checks.
> >
>
> Many thanks for your reply.
> We also avoid get_task_struct/ put_task_struct pair of atomic ops, rcu_lock, task_lock and so on this way.
>
> >Is there a specific problem you've encountered that this fixes?
> >
>
> I'am sorry but there's no specific problem. I do this mainly to skip the unnecessary ptrace and security hooks.
Cool. Let's keep this as-is so we continue to have centralized
instrumentation of these things in the LSM. :)
Thanks for your attention to performance!
--
Kees Cook
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages()
@ 2020-08-18 2:25 linmiaohe
0 siblings, 0 replies; 6+ messages in thread
From: linmiaohe @ 2020-08-18 2:25 UTC (permalink / raw)
To: Matthew Wilcox; +Cc: akpm, linux-mm, linux-kernel
Matthew Wilcox <willy@infradead.org> wrote:
>On Mon, Aug 17, 2020 at 07:59:33AM -0400, Miaohe Lin wrote:
>> There is no need to check if this process has the right to modify the
>> specified process when they are same.
>
>We should probably also skip the security hook call if a process is modifying its own pages.
>
>How about this instead?
>
It sounds good, thanks for your good advice. Would you like a tag like suggest-by?
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages()
@ 2020-08-17 11:59 Miaohe Lin
2020-08-17 12:32 ` Matthew Wilcox
2020-08-19 18:04 ` Kees Cook
0 siblings, 2 replies; 6+ messages in thread
From: Miaohe Lin @ 2020-08-17 11:59 UTC (permalink / raw)
To: akpm; +Cc: linux-mm, linux-kernel, linmiaohe
There is no need to check if this process has the right to modify the
specified process when they are same.
Signed-off-by: Hongxiang Lou <louhongxiang@huawei.com>
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
---
mm/migrate.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/mm/migrate.c b/mm/migrate.c
index 34a842a8eb6a..342c1ce0b433 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1903,7 +1903,7 @@ static int kernel_move_pages(pid_t pid, unsigned long nr_pages,
* Check if this process has the right to modify the specified
* process. Use the regular "ptrace_may_access()" checks.
*/
- if (!ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS)) {
+ if (pid && !ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS)) {
rcu_read_unlock();
err = -EPERM;
goto out;
--
2.19.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages()
2020-08-17 11:59 Miaohe Lin
@ 2020-08-17 12:32 ` Matthew Wilcox
2020-08-19 18:04 ` Kees Cook
1 sibling, 0 replies; 6+ messages in thread
From: Matthew Wilcox @ 2020-08-17 12:32 UTC (permalink / raw)
To: Miaohe Lin; +Cc: akpm, linux-mm, linux-kernel
On Mon, Aug 17, 2020 at 07:59:33AM -0400, Miaohe Lin wrote:
> There is no need to check if this process has the right to modify the
> specified process when they are same.
We should probably also skip the security hook call if a process is
modifying its own pages.
How about this instead?
diff --git a/mm/migrate.c b/mm/migrate.c
index f37729673558..84dc1f42161d 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1797,33 +1797,22 @@ static int do_pages_stat(struct mm_struct *mm, unsigned long nr_pages,
return nr_pages ? -EFAULT : 0;
}
-/*
- * Move a list of pages in the address space of the currently executing
- * process.
- */
-static int kernel_move_pages(pid_t pid, unsigned long nr_pages,
- const void __user * __user *pages,
- const int __user *nodes,
- int __user *status, int flags)
+static struct mm_struct *find_mm_struct(pid_t pid, nodemask_t *mem_nodes)
{
struct task_struct *task;
struct mm_struct *mm;
- int err;
- nodemask_t task_nodes;
-
- /* Check flags */
- if (flags & ~(MPOL_MF_MOVE|MPOL_MF_MOVE_ALL))
- return -EINVAL;
- if ((flags & MPOL_MF_MOVE_ALL) && !capable(CAP_SYS_NICE))
- return -EPERM;
+ if (!pid) {
+ mmget(current->mm);
+ *mem_nodes = cpuset_mems_allowed(current);
+ return current->mm;
+ }
- /* Find the mm_struct */
rcu_read_lock();
- task = pid ? find_task_by_vpid(pid) : current;
+ task = find_task_by_vpid(pid);
if (!task) {
rcu_read_unlock();
- return -ESRCH;
+ return ERR_PTR(-ESRCH);
}
get_task_struct(task);
@@ -1833,22 +1822,49 @@ static int kernel_move_pages(pid_t pid, unsigned long nr_pages,
*/
if (!ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS)) {
rcu_read_unlock();
- err = -EPERM;
+ mm = ERR_PTR(-EPERM);
goto out;
}
rcu_read_unlock();
- err = security_task_movememory(task);
- if (err)
+ mm = ERR_PTR(security_task_movememory(task));
+ if (IS_ERR(mm))
goto out;
- task_nodes = cpuset_mems_allowed(task);
+ *mem_nodes = cpuset_mems_allowed(task);
mm = get_task_mm(task);
+out:
put_task_struct(task);
if (!mm)
+ mm = ERR_PTR(-EINVAL);
+ return mm;
+}
+
+/*
+ * Move a list of pages in the address space of the specified
+ * process (pid of 0 means current).
+ */
+static int kernel_move_pages(pid_t pid, unsigned long nr_pages,
+ const void __user * __user *pages,
+ const int __user *nodes,
+ int __user *status, int flags)
+{
+ struct mm_struct *mm;
+ int err;
+ nodemask_t task_nodes;
+
+ /* Check flags */
+ if (flags & ~(MPOL_MF_MOVE|MPOL_MF_MOVE_ALL))
return -EINVAL;
+ if ((flags & MPOL_MF_MOVE_ALL) && !capable(CAP_SYS_NICE))
+ return -EPERM;
+
+ mm = find_mm_struct(pid, &task_nodes);
+ if (IS_ERR(mm))
+ return PTR_ERR(mm);
+
if (nodes)
err = do_pages_move(mm, task_nodes, nr_pages, pages,
nodes, status, flags);
@@ -1857,10 +1873,6 @@ static int kernel_move_pages(pid_t pid, unsigned long nr_pages,
mmput(mm);
return err;
-
-out:
- put_task_struct(task);
- return err;
}
SYSCALL_DEFINE6(move_pages, pid_t, pid, unsigned long, nr_pages,
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages()
2020-08-17 11:59 Miaohe Lin
2020-08-17 12:32 ` Matthew Wilcox
@ 2020-08-19 18:04 ` Kees Cook
1 sibling, 0 replies; 6+ messages in thread
From: Kees Cook @ 2020-08-19 18:04 UTC (permalink / raw)
To: Miaohe Lin; +Cc: akpm, linux-mm, linux-kernel, linux-security-module
On Mon, Aug 17, 2020 at 07:59:33AM -0400, Miaohe Lin wrote:
> There is no need to check if this process has the right to modify the
> specified process when they are same.
>
> Signed-off-by: Hongxiang Lou <louhongxiang@huawei.com>
> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> ---
> mm/migrate.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 34a842a8eb6a..342c1ce0b433 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1903,7 +1903,7 @@ static int kernel_move_pages(pid_t pid, unsigned long nr_pages,
> * Check if this process has the right to modify the specified
> * process. Use the regular "ptrace_may_access()" checks.
> */
> - if (!ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS)) {
> + if (pid && !ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS)) {
> rcu_read_unlock();
> err = -EPERM;
> goto out;
NAK, please don't do this -- the ptrace and security hooks already do
these kinds of self-introspection checks, and I'd like to keep a central
place to perform these kinds of checks.
Is there a specific problem you've encountered that this fixes?
--
Kees Cook
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2020-08-20 21:21 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-08-20 2:18 [PATCH] mm/migrate: Avoid possible unnecessary ptrace_may_access() call in kernel_move_pages() linmiaohe
2020-08-20 21:21 ` Kees Cook
-- strict thread matches above, loose matches on Subject: below --
2020-08-18 2:25 linmiaohe
2020-08-17 11:59 Miaohe Lin
2020-08-17 12:32 ` Matthew Wilcox
2020-08-19 18:04 ` Kees Cook
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).