Skip to contentSkip to conversation

net, mellanox mlx4 Fix compile warnings

PATCH In review Prarit Bhargava 6 participants 12 messages Sep 13, 2013 – Sep 17, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net, mellanox mlx4 Fix compile warnings

Fix unitialized variable warnings.

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_CQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2551:16: error: ‘cq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]

  atomic_dec(&cq->mtt->ref_count);
                ^

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_SRQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2734:17: error: ‘srq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
atomic_dec(&srq->mtt->ref_count);

---
 .../net/ethernet/mellanox/mlx4/resource_tracker.c  |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c +2 -2
index dd68763..d703838 100644
@@ -2563,7 +2563,7 @@ int mlx4_HW2SW_CQ_wrapper(struct mlx4_dev *dev, int slave,
 {
 	int err;
 	int cqn = vhcr->in_modifier;
-	struct res_cq *cq;
+	struct res_cq *uninitialized_var(cq);
 
 	err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED, &cq);
 	if (err)
@@ -2746,7 +2746,7 @@ int mlx4_HW2SW_SRQ_wrapper(struct mlx4_dev *dev, int slave,
 {
 	int err;
 	int srqn = vhcr->in_modifier;
-	struct res_srq *srq;
+	struct res_srq *uninitialized_var(srq);
 
 	err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED, &srq);
 	if (err)
Signature · 1 line
1.7.9.3
#1 lore raw
Quoted 1 line · Prarit Bhargava

Fix unitialized variable warnings.

Hi,

I'd like Jack, who is in charge on this code to take a look, added him.

Or.

Quoted 46 lines

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_CQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2551:16: error: ‘cq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]

  atomic_dec(&cq->mtt->ref_count);
                ^

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_SRQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2734:17: error: ‘srq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
atomic_dec(&srq->mtt->ref_count);

---
 .../net/ethernet/mellanox/mlx4/resource_tracker.c  |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c +2 -3
index dd68763..d703838 100644
@@ -2563,7 +2563,7 @@ int mlx4_HW2SW_CQ_wrapper(struct mlx4_dev *dev, int slave,
 {
        int err;
        int cqn = vhcr->in_modifier;
-       struct res_cq *cq;
+       struct res_cq *uninitialized_var(cq);
 
        err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED, &cq);
        if (err)
@@ -2746,7 +2746,7 @@ int mlx4_HW2SW_SRQ_wrapper(struct mlx4_dev *dev, int slave,
 {
        int err;
        int srqn = vhcr->in_modifier;
-       struct res_srq *srq;
+       struct res_srq *uninitialized_var(srq);
 
        err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED, &srq);
        if (err)
--

1.7.9.3

--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
Quoted 4 lines · On Sat, 14 Sep 2013 22:10:19 +0300 Or Gerlitz <[email protected]> wrote:
Quoted 4 lines

+ struct res_cq *uninitialized_var(cq);

err = cq_res_start_move_to(dev, slave, cqn,
RES_CQ_ALLOCATED, &cq);

I have no objection. However, I don't know if the compiler is being too clever here or too stupid. The cq variable is initialized in cq_res_start_move_to(), but the compiler is ignoring this -- or maybe it is simply not assuming that cq will in fact be initialized by the called procedure?

In any event, this change cannot hurt.

-Jack

#3 lore raw

hell what!

Quoted 21 lines · Jack Morgenstein
Quoted 4 lines · On Sat, 14 Sep 2013 22:10:19 +0300 Or Gerlitz <[email protected]> wrote:
Quoted 4 lines

+ struct res_cq *uninitialized_var(cq);

err = cq_res_start_move_to(dev, slave, cqn,
RES_CQ_ALLOCATED, &cq);

I have no objection. However, I don't know if the compiler is being too clever here or too stupid. The cq variable is initialized in cq_res_start_move_to(), but the compiler is ignoring this -- or maybe it is simply not assuming that cq will in fact be initialized by the called procedure?

In any event, this change cannot hurt.

-Jack
--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#4 lore raw

sorry for the inconvenince, i'm just practise the use of mutt, sorry again.

Quoted 24 lines · ZHAO Gang

hell what!

Quoted 21 lines · Jack Morgenstein
Quoted 4 lines · On Sat, 14 Sep 2013 22:10:19 +0300 Or Gerlitz <[email protected]> wrote:
Quoted 4 lines

+ struct res_cq *uninitialized_var(cq);

err = cq_res_start_move_to(dev, slave, cqn,
RES_CQ_ALLOCATED, &cq);

I have no objection. However, I don't know if the compiler is being too clever here or too stupid. The cq variable is initialized in cq_res_start_move_to(), but the compiler is ignoring this -- or maybe it is simply not assuming that cq will in fact be initialized by the called procedure?

In any event, this change cannot hurt.

-Jack
--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#5 lore raw
In reply to Jack Morgenstein
Quoted 5 lines
Quoted 2 lines · On Sat, 14 Sep 2013 22:10:19 +0300 Or Gerlitz <[email protected]> wrote:
Quoted 2 lines

+ struct res_cq *uninitialized_var(cq);

Quoted 8 lines
Quoted 2 lines
Quoted 2 lines

err = cq_res_start_move_to(dev, slave, cqn,
RES_CQ_ALLOCATED, &cq);

I have no objection. However, I don't know if the compiler is being too clever here or too stupid. The cq variable is initialized in cq_res_start_move_to(), but the compiler is ignoring this -- or maybe it is simply not assuming that cq will in fact be initialized by the called procedure?

Possibly cq_res_start_move_to() got inlined - and the analysis of the inlined code found a path where it wasn't initialised.

Most likely one of the paths where the return value is non-zero.

Quoted 1 line

In any event, this change cannot hurt.

It could hide a real 'used but not initialised' error later on...

David

#6 lore raw
Quoted 14 lines · David Laight
Quoted 5 lines
Quoted 2 lines · On Sat, 14 Sep 2013 22:10:19 +0300 Or Gerlitz <[email protected]> wrote:
Quoted 2 lines

+ struct res_cq *uninitialized_var(cq);

Quoted 8 lines
Quoted 2 lines
Quoted 2 lines

err = cq_res_start_move_to(dev, slave, cqn,
RES_CQ_ALLOCATED, &cq);

I have no objection. However, I don't know if the compiler is being too clever here or too stupid. The cq variable is initialized in cq_res_start_move_to(), but the compiler is ignoring this -- or maybe it is simply not assuming that cq will in fact be initialized by the called procedure?

Quoted 8 lines

Possibly cq_res_start_move_to() got inlined - and the analysis of the inlined code found a path where it wasn't initialised.

Most likely one of the paths where the return value is non-zero.

Quoted 1 line

In any event, this change cannot hurt.

It could hide a real 'used but not initialised' error later on...

Hi David,

I certainly don't profess to know the code better than you do of course, but in terms of risk we're no better in than the original code here, and it seems that other places in the kernel the resolution has been to either init to NULL or wrap with uninitialized_var()...

P.

#7 lore raw
In reply to David Laight
Quoted 5 lines · On Mon, 16 Sep 2013 09:59:41 +0100 "David Laight" <[email protected]>…
Quoted 1 line

In any event, this change cannot hurt.

It could hide a real 'used but not initialised' error later on...

David

In this case, it doesn't. If cq_res_start_move_to returns 0, the "cq" pointer points to the desired cq object. In all instances where the cq object is not found, you get an error return, and the cq pointer in the caller is never de-referenced.

Therefore, the compiler warning is a false positive, and the "uninitialized_var" is just a workaround to satisfy the compiler.

Same for the SRQ case.

-Jack

maintainer
#8 lore raw
In reply to Prarit Bhargava

From: Prarit Bhargava <[email protected]>
Date: Fri, 13 Sep 2013 08:30:38 -0400

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c int mlx4_HW2SW_CQ_wrapper(struct mlx4_dev *dev, int slave, · line 2569

Quoted 9 lines
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c +1 -1
@@ -2563,7 +2563,7 @@ int mlx4_HW2SW_CQ_wrapper(struct mlx4_dev *dev, int slave,
 {
 	int err;
 	int cqn = vhcr->in_modifier;
-	struct res_cq *cq;
+	struct res_cq *uninitialized_var(cq);
 
 	err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED, &cq);
 	if (err)

This gross annotation is not necessary.

Make cq_res_start_move_to() return 'cq' as an error pointer instead of setting 'cq' by reference.

#9 lore raw

Fix unitialized variable warnings.

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_CQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2551:16: error: ‘cq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]

  atomic_dec(&cq->mtt->ref_count);
                ^

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_SRQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2734:17: error: ‘srq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
atomic_dec(&srq->mtt->ref_count);

[v2]: davem suggested making cq_res_start_move_to() return 'cq' as an error pointer instead of setting 'cq' by reference. I also did the same for srq.

---
 .../net/ethernet/mellanox/mlx4/resource_tracker.c  |   46 ++++++++++----------
 1 file changed, 24 insertions(+), 22 deletions(-)
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c +24 -22
index dd68763..343206b 100644
@@ -1073,8 +1073,9 @@ static int eq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
 	return err;
 }
 
-static int cq_res_start_move_to(struct mlx4_dev *dev, int slave, int cqn,
-				enum res_cq_states state, struct res_cq **cq)
+static struct res_cq *cq_res_start_move_to(struct mlx4_dev *dev,
+						  int slave, int cqn,
+						  enum res_cq_states state)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
@@ -1117,18 +1118,19 @@ static int cq_res_start_move_to(struct mlx4_dev *dev, int slave, int cqn,
 			r->com.from_state = r->com.state;
 			r->com.to_state = state;
 			r->com.state = RES_CQ_BUSY;
-			if (cq)
-				*cq = r;
+		} else {
+			r = ERR_PTR(err);
 		}
 	}
 
 	spin_unlock_irq(mlx4_tlock(dev));
 
-	return err;
+	return r;
 }
 
-static int srq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
-				 enum res_cq_states state, struct res_srq **srq)
+static struct res_srq *srq_res_start_move_to(struct mlx4_dev *dev, int slave,
+					     int index,
+					     enum res_cq_states state)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
@@ -1167,14 +1169,14 @@ static int srq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
 			r->com.from_state = r->com.state;
 			r->com.to_state = state;
 			r->com.state = RES_SRQ_BUSY;
-			if (srq)
-				*srq = r;
+		} else {
+			r = ERR_PTR(err);
 		}
 	}
 
 	spin_unlock_irq(mlx4_tlock(dev));
 
-	return err;
+	return r;
 }
 
 static void res_abort_move(struct mlx4_dev *dev, int slave,
@@ -2530,9 +2532,9 @@ int mlx4_SW2HW_CQ_wrapper(struct mlx4_dev *dev, int slave,
 	struct res_cq *cq;
 	struct res_mtt *mtt;
 
-	err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_HW, &cq);
-	if (err)
-		return err;
+	cq = cq_res_start_move_to(dev, slave, cqn, RES_CQ_HW);
+	if (IS_ERR(cq))
+		return PTR_ERR(cq);
 	err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
 	if (err)
 		goto out_move;
@@ -2565,9 +2567,9 @@ int mlx4_HW2SW_CQ_wrapper(struct mlx4_dev *dev, int slave,
 	int cqn = vhcr->in_modifier;
 	struct res_cq *cq;
 
-	err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED, &cq);
-	if (err)
-		return err;
+	cq = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED);
+	if (IS_ERR(cq))
+		return PTR_ERR(cq);
 	err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
 	if (err)
 		goto out_move;
@@ -2709,9 +2711,9 @@ int mlx4_SW2HW_SRQ_wrapper(struct mlx4_dev *dev, int slave,
 	if (srqn != (be32_to_cpu(srqc->state_logsize_srqn) & 0xffffff))
 		return -EINVAL;
 
-	err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_HW, &srq);
-	if (err)
-		return err;
+	srq = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED);
+	if (IS_ERR(srq))
+		return PTR_ERR(srq);
 	err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
 	if (err)
 		goto ex_abort;
@@ -2748,9 +2750,9 @@ int mlx4_HW2SW_SRQ_wrapper(struct mlx4_dev *dev, int slave,
 	int srqn = vhcr->in_modifier;
 	struct res_srq *srq;
 
-	err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED, &srq);
-	if (err)
-		return err;
+	srq = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED);
+	if (IS_ERR(srq))
+		return PTR_ERR(srq);
 	err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
 	if (err)
 		goto ex_abort;
Signature · 1 line
1.7.9.3
#10 lore raw
Quoted 4 lines · Prarit Bhargava

Fix unitialized variable warnings.

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_CQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2551:16: error: ‘cq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]

Quoted 8 lines
  atomic_dec(&cq->mtt->ref_count);
                ^

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_SRQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2734:17: error: ‘srq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
atomic_dec(&srq->mtt->ref_count);

[v2]: davem suggested making cq_res_start_move_to() return 'cq' as an error pointer instead of setting 'cq' by reference. I also did the same for srq.

Pravit, as I wrote you earlier on this thread, Jack from our team maintains this piece (SRIOV resource tracker) of the mlx4 core driver code, so I am adding him. Dave, as many of us here he might be OOO for the coming ten days for holiday vacation, so would ask you to please wait patiently for his ack/nak...

Or.

Quoted 127 lines
---
 .../net/ethernet/mellanox/mlx4/resource_tracker.c  |   46 ++++++++++----------
 1 file changed, 24 insertions(+), 22 deletions(-)
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c +24 -23
index dd68763..343206b 100644
@@ -1073,8 +1073,9 @@ static int eq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
        return err;
 }
 
-static int cq_res_start_move_to(struct mlx4_dev *dev, int slave, int cqn,
-                               enum res_cq_states state, struct res_cq **cq)
+static struct res_cq *cq_res_start_move_to(struct mlx4_dev *dev,
+                                                 int slave, int cqn,
+                                                 enum res_cq_states state)
 {
        struct mlx4_priv *priv = mlx4_priv(dev);
        struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
@@ -1117,18 +1118,19 @@ static int cq_res_start_move_to(struct mlx4_dev *dev, int slave, int cqn,
                        r->com.from_state = r->com.state;
                        r->com.to_state = state;
                        r->com.state = RES_CQ_BUSY;
-                       if (cq)
-                               *cq = r;
+               } else {
+                       r = ERR_PTR(err);
                }
        }
 
        spin_unlock_irq(mlx4_tlock(dev));
 
-       return err;
+       return r;
 }
 
-static int srq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
-                                enum res_cq_states state, struct res_srq **srq)
+static struct res_srq *srq_res_start_move_to(struct mlx4_dev *dev, int slave,
+                                            int index,
+                                            enum res_cq_states state)
 {
        struct mlx4_priv *priv = mlx4_priv(dev);
        struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
@@ -1167,14 +1169,14 @@ static int srq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
                        r->com.from_state = r->com.state;
                        r->com.to_state = state;
                        r->com.state = RES_SRQ_BUSY;
-                       if (srq)
-                               *srq = r;
+               } else {
+                       r = ERR_PTR(err);
                }
        }
 
        spin_unlock_irq(mlx4_tlock(dev));
 
-       return err;
+       return r;
 }
 
 static void res_abort_move(struct mlx4_dev *dev, int slave,
@@ -2530,9 +2532,9 @@ int mlx4_SW2HW_CQ_wrapper(struct mlx4_dev *dev, int slave,
        struct res_cq *cq;
        struct res_mtt *mtt;
 
-       err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_HW, &cq);
-       if (err)
-               return err;
+       cq = cq_res_start_move_to(dev, slave, cqn, RES_CQ_HW);
+       if (IS_ERR(cq))
+               return PTR_ERR(cq);
        err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
        if (err)
                goto out_move;
@@ -2565,9 +2567,9 @@ int mlx4_HW2SW_CQ_wrapper(struct mlx4_dev *dev, int slave,
        int cqn = vhcr->in_modifier;
        struct res_cq *cq;
 
-       err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED, &cq);
-       if (err)
-               return err;
+       cq = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED);
+       if (IS_ERR(cq))
+               return PTR_ERR(cq);
        err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
        if (err)
                goto out_move;
@@ -2709,9 +2711,9 @@ int mlx4_SW2HW_SRQ_wrapper(struct mlx4_dev *dev, int slave,
        if (srqn != (be32_to_cpu(srqc->state_logsize_srqn) & 0xffffff))
                return -EINVAL;
 
-       err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_HW, &srq);
-       if (err)
-               return err;
+       srq = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED);
+       if (IS_ERR(srq))
+               return PTR_ERR(srq);
        err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
        if (err)
                goto ex_abort;
@@ -2748,9 +2750,9 @@ int mlx4_HW2SW_SRQ_wrapper(struct mlx4_dev *dev, int slave,
        int srqn = vhcr->in_modifier;
        struct res_srq *srq;
 
-       err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED, &srq);
-       if (err)
-               return err;
+       srq = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED);
+       if (IS_ERR(srq))
+               return PTR_ERR(srq);
        err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
        if (err)
                goto ex_abort;
--

1.7.9.3

--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

maintainer
#11 lore raw

From: Or Gerlitz <[email protected]>
Date: Tue, 17 Sep 2013 22:25:57 +0300

Quoted 11 lines · Prarit Bhargava
Quoted 10 lines

Fix unitialized variable warnings.

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_CQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2551:16: error: ‘cq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]

  atomic_dec(&cq->mtt->ref_count);
                ^

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function ‘mlx4_HW2SW_SRQ_wrapper’:
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:2734:17: error: ‘srq’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
atomic_dec(&srq->mtt->ref_count);

Quoted 8 lines
Quoted 2 lines

[v2]: davem suggested making cq_res_start_move_to() return 'cq' as an error pointer instead of setting 'cq' by reference. I also did the same for srq.

Pravit, as I wrote you earlier on this thread, Jack from our team maintains this piece (SRIOV resource tracker) of the mlx4 core driver code, so I am adding him. Dave, as many of us here he might be OOO for the coming ten days for holiday vacation, so would ask you to please wait patiently for his ack/nak...

I already stated that I wanted this issue fixed differently.

Specifically I said that these functions should return error pointers, instead of trying to return an integer error whilst setting the cq pointer by reference.

This is a method which is idiomatic and common across the kernel, and is the preferred way to handle this kind of situation.