ucounts: Move kfree() out of critical zone protected by ucounts_lock

Posted by mengensun88@gmail.com 1 year ago

From: MengEn Sun <mengensun@tencent.com>

Although kfree is a non-sleep function, it is possible to enter a
long chain of calls probabilistically, so it looks better to move
kfree from alloc_ucounts() out of the critical zone of ucounts_lock.

Reviewed-by: YueHong Wu <yuehongwu@tencent.com>
Signed-off-by: MengEn Sun <mengensun@tencent.com>
---
 kernel/ucount.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/ucount.c b/kernel/ucount.c
index f950b5e5..86c5f1c 100644
--- a/kernel/ucount.c
+++ b/kernel/ucount.c
@@ -164,8 +164,8 @@ struct ucounts *get_ucounts(struct ucounts *ucounts)
 struct ucounts *alloc_ucounts(struct user_namespace *ns, kuid_t uid)
 {
 	struct hlist_head *hashent = ucounts_hashentry(ns, uid);
-	struct ucounts *ucounts, *new;
 	bool wrapped;
+	struct ucounts *ucounts, *new = NULL;
 
 	spin_lock_irq(&ucounts_lock);
 	ucounts = find_ucounts(ns, uid, hashent);
@@ -182,17 +182,17 @@ struct ucounts *alloc_ucounts(struct user_namespace *ns, kuid_t uid)
 
 		spin_lock_irq(&ucounts_lock);
 		ucounts = find_ucounts(ns, uid, hashent);
-		if (ucounts) {
-			kfree(new);
-		} else {
+		if (!ucounts) {
 			hlist_add_head(&new->node, hashent);
 			get_user_ns(new->ns);
 			spin_unlock_irq(&ucounts_lock);
 			return new;
 		}
 	}
+
 	wrapped = !get_ucounts_or_wrap(ucounts);
 	spin_unlock_irq(&ucounts_lock);
+	kfree(new);
 	if (wrapped) {
 		put_ucounts(ucounts);
 		return NULL;
-- 
1.8.3.1

Re: ucounts: Move kfree() out of critical zone protected by ucounts_lock

Posted by Kuan-Wei Chiu 1 year ago

Hi MengEn,

On Fri, Dec 06, 2024 at 12:13:47PM +0800, mengensun88@gmail.com wrote:
> From: MengEn Sun <mengensun@tencent.com>
> 
> Although kfree is a non-sleep function, it is possible to enter a
> long chain of calls probabilistically, so it looks better to move
> kfree from alloc_ucounts() out of the critical zone of ucounts_lock.
> 
> Reviewed-by: YueHong Wu <yuehongwu@tencent.com>
> Signed-off-by: MengEn Sun <mengensun@tencent.com>
> ---
>  kernel/ucount.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
>
It's a common convention to prefix your subject line with [PATCH] to
let others know it's a patch. Additionally, when resending a patch
without changes, consider adding RESEND to the subject line, e.g.,
[PATCH RESEND] ucounts: Move kfree() out of critical zone protected by
ucounts_lock.

See: https://www.kernel.org/doc/html/v6.12/process/submitting-patches.html#include-patch-in-the-subject

Regards,
Kuan-Wei

Re: ucounts: Move kfree() out of critical zone protected by ucounts_lock

Posted by MengEn Sun 1 year ago

Hi Kuan Wei

On Fri, 6 Dec 2024 13:55:51 +0800 visitorckw@gmail.com wrote:
> Hi MengEn,
> 
> On Fri, Dec 06, 2024 at 12:13:47PM +0800, mengensun88@gmail.com wrote:
> > From: MengEn Sun <mengensun@tencent.com>
> > 
> > Although kfree is a non-sleep function, it is possible to enter a
> > long chain of calls probabilistically, so it looks better to move
> > kfree from alloc_ucounts() out of the critical zone of ucounts_lock.
> > 
> > Reviewed-by: YueHong Wu <yuehongwu@tencent.com>
> > Signed-off-by: MengEn Sun <mengensun@tencent.com>
> > ---
> >  kernel/ucount.c | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> >
> It's a common convention to prefix your subject line with [PATCH] to
> let others know it's a patch. Additionally, when resending a patch
> without changes, consider adding RESEND to the subject line, e.g.,
> [PATCH RESEND] ucounts: Move kfree() out of critical zone protected by
> ucounts_lock.

Thank you very much for your reply. I will make some changes here.

> 
> See: https://www.kernel.org/doc/html/v6.12/process/submitting-patches.html#include-patch-in-the-subject

I will study this document carefully.

> 
> Regards,
> Kuan-Wei

Regards,
Meng En