Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1657323 > unrolled thread

[PATCH V3] rxe: Fix a sleep-in-atomic bug in post_one_send

Back to article view | Back to linux.kernel

Contents

  [PATCH V3] rxe: Fix a sleep-in-atomic bug in post_one_send Jia-Ju Bai <baijiaju1990@163.com> - 2017-06-05 09:40 +0200
    Re: [PATCH V3] rxe: Fix a sleep-in-atomic bug in post_one_send Yuval Shaia <yuval.shaia@oracle.com> - 2017-06-05 09:50 +0200
    Re: [PATCH V3] rxe: Fix a sleep-in-atomic bug in post_one_send Jia-Ju Bai <baijiaju1990@163.com> - 2017-06-05 10:40 +0200
      Re: [PATCH V3] rxe: Fix a sleep-in-atomic bug in post_one_send Moni Shoua <monis@mellanox.com> - 2017-06-05 11:30 +0200
    Re: [PATCH V3] rxe: Fix a sleep-in-atomic bug in post_one_send Moni Shoua <monis@mellanox.com> - 2017-06-05 10:40 +0200

#1657323 — [PATCH V3] rxe: Fix a sleep-in-atomic bug in post_one_send

The driver may sleep under a spin lock, and the function call path is:
post_one_send (acquire the lock by spin_lock_irqsave)
  init_send_wqe
    copy_from_user --> may sleep

To fix it, the lock is released before copy_from_user, and the lock is
acquired again after this function. The parameter "flags" is used to
restore and save the irq status.


Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
---
V3:
* It corrects the mistakes of remaining legacy code in V2. 
  (Thank Ram for pointing it out)

V2:
* The parameter "flags" is added to restore and save the irq status.
  Thank Leon for good advice.

 drivers/infiniband/sw/rxe/rxe_verbs.c |   13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index 83d709e..5293d15 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -721,11 +721,11 @@ static void init_send_wr(struct rxe_qp *qp, struct rxe_send_wr *wr,
 
 static int init_send_wqe(struct rxe_qp *qp, struct ib_send_wr *ibwr,
 			 unsigned int mask, unsigned int length,
-			 struct rxe_send_wqe *wqe)
+			 struct rxe_send_wqe *wqe, unsigned long *flags)
 {
 	int num_sge = ibwr->num_sge;
 	struct ib_sge *sge;
-	int i;
+	int i, err;
 	u8 *p;
 
 	init_send_wr(qp, &wqe->wr, ibwr);
@@ -740,8 +740,11 @@ static int init_send_wqe(struct rxe_qp *qp, struct ib_send_wr *ibwr,
 
 		sge = ibwr->sg_list;
 		for (i = 0; i < num_sge; i++, sge++) {
-			if (qp->is_user && copy_from_user(p, (__user void *)
-					    (uintptr_t)sge->addr, sge->length))
+			spin_unlock_irqrestore(&qp->sq.sq_lock, *flags);
+			err = copy_from_user(p, (__user void *)
+					(uintptr_t)sge->addr, sge->length);
+			spin_lock_irqsave(&qp->sq.sq_lock, *flags);
+			if (qp->is_user && err)
 				return -EFAULT;
 
 			else if (!qp->is_user)
@@ -794,7 +797,7 @@ static int post_one_send(struct rxe_qp *qp, struct ib_send_wr *ibwr,
 
 	send_wqe = producer_addr(sq->queue);
 
-	err = init_send_wqe(qp, ibwr, mask, length, send_wqe);
+	err = init_send_wqe(qp, ibwr, mask, length, send_wqe, &flags);
 	if (unlikely(err))
 		goto err1;
 
-- 
1.7.9.5

[toc] | [next] | [standalone]

#1657327

On Mon, Jun 05, 2017 at 03:39:02PM +0800, Jia-Ju Bai wrote:
> The driver may sleep under a spin lock, and the function call path is:
> post_one_send (acquire the lock by spin_lock_irqsave)
>   init_send_wqe
>     copy_from_user --> may sleep
> 
> To fix it, the lock is released before copy_from_user, and the lock is
> acquired again after this function. The parameter "flags" is used to
> restore and save the irq status.
> 
> 
> Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
> ---
> V3:
> * It corrects the mistakes of remaining legacy code in V2. 
>   (Thank Ram for pointing it out)
> 
> V2:
> * The parameter "flags" is added to restore and save the irq status.
>   Thank Leon for good advice.
> 

My mistake for not mentioning it specifically but in addition to moving
such version history after the "---" you should also add (manually) a
terminating "---" after the block.

>  drivers/infiniband/sw/rxe/rxe_verbs.c |   13 ++++++++-----
>  1 file changed, 8 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
> index 83d709e..5293d15 100644
> --- a/drivers/infiniband/sw/rxe/rxe_verbs.c
> +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
> @@ -721,11 +721,11 @@ static void init_send_wr(struct rxe_qp *qp, struct rxe_send_wr *wr,
>  
>  static int init_send_wqe(struct rxe_qp *qp, struct ib_send_wr *ibwr,
>  			 unsigned int mask, unsigned int length,
> -			 struct rxe_send_wqe *wqe)
> +			 struct rxe_send_wqe *wqe, unsigned long *flags)
>  {
>  	int num_sge = ibwr->num_sge;
>  	struct ib_sge *sge;
> -	int i;
> +	int i, err;
>  	u8 *p;
>  
>  	init_send_wr(qp, &wqe->wr, ibwr);
> @@ -740,8 +740,11 @@ static int init_send_wqe(struct rxe_qp *qp, struct ib_send_wr *ibwr,
>  
>  		sge = ibwr->sg_list;
>  		for (i = 0; i < num_sge; i++, sge++) {
> -			if (qp->is_user && copy_from_user(p, (__user void *)
> -					    (uintptr_t)sge->addr, sge->length))
> +			spin_unlock_irqrestore(&qp->sq.sq_lock, *flags);
> +			err = copy_from_user(p, (__user void *)
> +					(uintptr_t)sge->addr, sge->length);
> +			spin_lock_irqsave(&qp->sq.sq_lock, *flags);
> +			if (qp->is_user && err)
>  				return -EFAULT;
>  
>  			else if (!qp->is_user)
> @@ -794,7 +797,7 @@ static int post_one_send(struct rxe_qp *qp, struct ib_send_wr *ibwr,
>  
>  	send_wqe = producer_addr(sq->queue);
>  
> -	err = init_send_wqe(qp, ibwr, mask, length, send_wqe);
> +	err = init_send_wqe(qp, ibwr, mask, length, send_wqe, &flags);
>  	if (unlikely(err))
>  		goto err1;
>  
> -- 
> 1.7.9.5
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

[toc] | [prev] | [next] | [standalone]

#1657388

On 06/05/2017 04:30 PM, Moni Shoua wrote:
>> -                       if (qp->is_user&&  copy_from_user(p, (__user void *)
>> -                                           (uintptr_t)sge->addr, sge->length))
>> +                       spin_unlock_irqrestore(&qp->sq.sq_lock, *flags);
>> +                       err = copy_from_user(p, (__user void *)
>> +                                       (uintptr_t)sge->addr, sge->length);
>> +                       spin_lock_irqsave(&qp->sq.sq_lock, *flags);
>> +                       if (qp->is_user&&  err)
>>                                  return -EFAULT;
> qp-_is_user is always false in this function (flow starts from
> rxe_post_send_kernel) so this line is a dead code
> In fact, this patch seems to add a serious bug when it uses
> copy_from_user() from a non user pointer.
> Do you agree?
I agree.
So, it is fine to me to remove this line, as you said in the former email:

>  Second, I think that there is no flow that leads to this function
>  when qp->is user is true so maybe the correct action is to remove this
>  line completely
>  if (qp->is_user&&  copy_from_user(p, (__user void *)

[toc] | [prev] | [next] | [standalone]

#1657457

> I agree.
> So, it is fine to me to remove this line, as you said in the former email:
>
Thanks.
Can you please send a patch like that?

[toc] | [prev] | [next] | [standalone]

#1657390

> -                       if (qp->is_user && copy_from_user(p, (__user void *)
> -                                           (uintptr_t)sge->addr, sge->length))
> +                       spin_unlock_irqrestore(&qp->sq.sq_lock, *flags);
> +                       err = copy_from_user(p, (__user void *)
> +                                       (uintptr_t)sge->addr, sge->length);
> +                       spin_lock_irqsave(&qp->sq.sq_lock, *flags);
> +                       if (qp->is_user && err)
>                                 return -EFAULT;
qp-_is_user is always false in this function (flow starts from
rxe_post_send_kernel) so this line is a dead code
In fact, this patch seems to add a serious bug when it uses
copy_from_user() from a non user pointer.
Do you agree?

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web