[PATCH v2] ipc/sem: do not sleep with a spin lock held

[PATCH v2] ipc/sem: do not sleep with a spin lock held

[cgel.zte]

From: Minghao Chi <chi.minghao@zte.com.cn>

We can't call kvfree() with a spin lock held, so defer it.
Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo 
allocation")

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
---
changelog since v2:
+ Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo 
+ allocation")
 ipc/sem.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/ipc/sem.c b/ipc/sem.c
index 6693daf4fe11..0dbdb98fdf2d 100644
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -1964,6 +1964,7 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 	 */
 	un = lookup_undo(ulp, semid);
 	if (un) {
+		spin_unlock(&ulp->lock);
 		kvfree(new);
 		goto success;
 	}
@@ -1976,9 +1977,8 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 	ipc_assert_locked_object(&sma->sem_perm);
 	list_add(&new->list_id, &sma->list_id);
 	un = new;
-
-success:
 	spin_unlock(&ulp->lock);
+success:
 	sem_unlock(sma, -1);
 out:
 	return un;
-- 
2.25.1

[PATCH v2] ipc/sem: do not sleep with a spin lock held

[cgel.zte]

From: Minghao Chi <chi.minghao@zte.com.cn>

We can't call kvfree() with a spin lock held, so defer it.
Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
allocation")

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
---
changelog since v2:
+ Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
+ allocation")
 ipc/sem.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/ipc/sem.c b/ipc/sem.c
index 6693daf4fe11..0dbdb98fdf2d 100644
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -1964,6 +1964,7 @@ static struct sem_undo *find_alloc_undo(struct
ipc_namespace *ns, int semid)
 	 */
 	un = lookup_undo(ulp, semid);
 	if (un) {
+		spin_unlock(&ulp->lock);
 		kvfree(new);
 		goto success;
 	}
@@ -1976,9 +1977,8 @@ static struct sem_undo *find_alloc_undo(struct
ipc_namespace *ns, int semid)
 	ipc_assert_locked_object(&sma->sem_perm);
 	list_add(&new->list_id, &sma->list_id);
 	un = new;
-
-success:
 	spin_unlock(&ulp->lock);
+success:
 	sem_unlock(sma, -1);
 out:
 	return un;
-- 
2.25.1

[PATCH v2] ipc/sem: do not sleep with a spin lock held

[cgel.zte]

From: Minghao Chi <chi.minghao@zte.com.cn>

We can't call kvfree() with a spin lock held, so defer it.
Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
allocation")

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
---
changelog since v2:
+ Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
+ allocation")
 ipc/sem.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/ipc/sem.c b/ipc/sem.c
index 6693daf4fe11..0dbdb98fdf2d 100644
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -1964,6 +1964,7 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 	 */
 	un = lookup_undo(ulp, semid);
 	if (un) {
+		spin_unlock(&ulp->lock);
 		kvfree(new);
 		goto success;
 	}
@@ -1976,9 +1977,8 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 	ipc_assert_locked_object(&sma->sem_perm);
 	list_add(&new->list_id, &sma->list_id);
 	un = new;
-
-success:
 	spin_unlock(&ulp->lock);
+success:
 	sem_unlock(sma, -1);
 out:
 	return un;
--
2.25.1

Re: [PATCH v2] ipc/sem: do not sleep with a spin lock held

[Jiri Slaby]

On 23. 12. 21, 4:12, cgel.zte@gmail.com wrote:
> From: Minghao Chi <chi.minghao@zte.com.cn>
> 
> We can't call kvfree() with a spin lock held, so defer it.

Sorry, defer what?

There are attempts to fix kvfree instead, not sure which of these 
approaches (fix kvfree or its callers) won in the end?

> Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
> allocation")
> 
> Reported-by: Zeal Robot <zealci@zte.com.cn>
> Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
> ---
> changelog since v2:
> + Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
> + allocation")
>   ipc/sem.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/ipc/sem.c b/ipc/sem.c
> index 6693daf4fe11..0dbdb98fdf2d 100644
> --- a/ipc/sem.c
> +++ b/ipc/sem.c
> @@ -1964,6 +1964,7 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
>   	 */
>   	un = lookup_undo(ulp, semid);
>   	if (un) {
> +		spin_unlock(&ulp->lock);
>   		kvfree(new);
>   		goto success;
>   	}


-- 
js
suse labs

Re: [PATCH v2] ipc/sem: do not sleep with a spin lock held

[Manfred Spraul]

Hi Jiri,

On 1/3/22 10:27, Jiri Slaby wrote:
> On 23. 12. 21, 4:12, cgel.zte@gmail.com wrote:
>> From: Minghao Chi <chi.minghao@zte.com.cn>
>>
>> We can't call kvfree() with a spin lock held, so defer it.
>
> Sorry, defer what?
>
First drop the spinlock, then call kvfree().


> There are attempts to fix kvfree instead, not sure which of these 
> approaches (fix kvfree or its callers) won in the end?
>
Exactly. We have three options - but noone volunteered yet to decide:

- change ipc/sem.c [minimal change]

- change kvfree() to use vfree_atomic() [would also fix other changes 
that did s/kfree/kvfree/]

- Modify the vma handling so that it becomes safe to call vfree() while 
holding a spinlock. [perfect approach, but I'm concerned about side effects]


--

     Manfred

Re: [PATCH v2] ipc/sem: do not sleep with a spin lock held

[Shakeel Butt]

On Mon, Jan 3, 2022 at 9:18 AM Manfred Spraul <manfred@colorfullife.com> wrote:
>
> Hi Jiri,
>
> On 1/3/22 10:27, Jiri Slaby wrote:
> > On 23. 12. 21, 4:12, cgel.zte@gmail.com wrote:
> >> From: Minghao Chi <chi.minghao@zte.com.cn>
> >>
> >> We can't call kvfree() with a spin lock held, so defer it.
> >
> > Sorry, defer what?
> >
> First drop the spinlock, then call kvfree().
>
>
> > There are attempts to fix kvfree instead, not sure which of these
> > approaches (fix kvfree or its callers) won in the end?
> >
> Exactly. We have three options - but noone volunteered yet to decide:
>
> - change ipc/sem.c [minimal change]

Let's go with the minimal change for now which can easily be
cherry-picked for the stable tree. It seems other approaches need more
work/discussion.

>
> - change kvfree() to use vfree_atomic() [would also fix other changes
> that did s/kfree/kvfree/]
>
> - Modify the vma handling so that it becomes safe to call vfree() while
> holding a spinlock. [perfect approach, but I'm concerned about side effects]
>
>
> --
>
>      Manfred
>

Re: [PATCH v2] ipc/sem: do not sleep with a spin lock held

[Shakeel Butt]

On Wed, Dec 22, 2021 at 7:12 PM <cgel.zte@gmail.com> wrote:
>
> From: Minghao Chi <chi.minghao@zte.com.cn>
>
> We can't call kvfree() with a spin lock held, so defer it.
> Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
> allocation")
>
> Reported-by: Zeal Robot <zealci@zte.com.cn>
> Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>

Reviewed-by: Shakeel Butt <shakeelb@google.com>

Re: [PATCH v2] ipc/sem: do not sleep with a spin lock held

[Manfred Spraul]

On 1/4/22 19:20, Shakeel Butt wrote:
> On Wed, Dec 22, 2021 at 7:12 PM <cgel.zte@gmail.com> wrote:
>> From: Minghao Chi <chi.minghao@zte.com.cn>
>>
>> We can't call kvfree() with a spin lock held, so defer it.
>> Fixes: fc37a3b8b438 ("[PATCH] ipc sem: use kvmalloc for sem_undo
>> allocation")
>>
>> Reported-by: Zeal Robot <zealci@zte.com.cn>
>> Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
> Reviewed-by: Shakeel Butt <shakeelb@google.com>

Reviewed-by: Manfred Spraul <manfred@colorfullife.com>

--

     Manfred