[PATCH net] bonding: Fix header_ops type confusion

[PATCH net] bonding: Fix header_ops type confusion

[戸田晃太]

In bond_setup_by_slave(), the slave’s header_ops are unconditionally
copied into the bonding device. As a result, the bonding device may invoke
the slave-specific header operations on itself, causing
netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
as the slave's private-data type.

This type-confusion bug can lead to out-of-bounds writes into the skb,
resulting in memory corruption.

This patch adds two members to struct bonding, bond_header_ops and
header_slave_dev, to avoid type-confusion while keeping track of the
slave's header_ops.

Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Co-Developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Reviewed-by: Paolo Abeni <pabeni@redhat.com>
Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
---
 drivers/net/bonding/bond_main.c | 61
++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 include/net/bonding.h           |  5 +++++
 2 files changed, 65 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 8ea183da8d53..690f3e0971d0 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1619,14 +1619,65 @@ static void bond_compute_features(struct bonding *bond)
     netdev_change_features(bond_dev);
 }

+static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
+        unsigned short type, const void *daddr,
+        const void *saddr, unsigned int len)
+{
+    struct bonding *bond = netdev_priv(dev);
+    struct net_device *slave_dev;
+
+    slave_dev = bond->header_slave_dev;
+
+    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
+}
+
+static void bond_header_cache_update(struct hh_cache *hh, const
struct net_device *dev,
+        const unsigned char *haddr)
+{
+    const struct bonding *bond = netdev_priv(dev);
+    struct net_device *slave_dev;
+
+    slave_dev = bond->header_slave_dev;
+
+    if (!slave_dev->header_ops || !slave_dev->header_ops->cache_update)
+        return;
+
+    slave_dev->header_ops->cache_update(hh, slave_dev, haddr);
+}
+
 static void bond_setup_by_slave(struct net_device *bond_dev,
                 struct net_device *slave_dev)
 {
+    struct bonding *bond = netdev_priv(bond_dev);
     bool was_up = !!(bond_dev->flags & IFF_UP);

     dev_close(bond_dev);

-    bond_dev->header_ops        = slave_dev->header_ops;
+    /* Some functions are given dev as an argument
+     * while others not. When dev is not given, we cannot
+     * find out what is the slave device through struct bonding
+     * (the private data of bond_dev). Therefore, we need a raw
+     * header_ops variable instead of its pointer to const header_ops
+     * and assign slave's functions directly.
+     * For the other case, we set the wrapper functions that pass
+     * slave_dev to the wrapped functions.
+     */
+    bond->bond_header_ops.create = bond_hard_header;
+    bond->bond_header_ops.cache_update = bond_header_cache_update;
+    if (slave_dev->header_ops) {
+        bond->bond_header_ops.parse = slave_dev->header_ops->parse;
+        bond->bond_header_ops.cache = slave_dev->header_ops->cache;
+        bond->bond_header_ops.validate = slave_dev->header_ops->validate;
+        bond->bond_header_ops.parse_protocol =
slave_dev->header_ops->parse_protocol;
+    } else {
+        bond->bond_header_ops.parse = NULL;
+        bond->bond_header_ops.cache = NULL;
+        bond->bond_header_ops.validate = NULL;
+        bond->bond_header_ops.parse_protocol = NULL;
+    }
+
+    bond->header_slave_dev      = slave_dev;
+    bond_dev->header_ops        = &bond->bond_header_ops;

     bond_dev->type            = slave_dev->type;
     bond_dev->hard_header_len   = slave_dev->hard_header_len;
@@ -2676,6 +2727,14 @@ static int bond_release_and_destroy(struct
net_device *bond_dev,
     struct bonding *bond = netdev_priv(bond_dev);
     int ret;

+    /* If slave_dev is the earliest registered one, we must clear
+     * the variables related to header_ops to avoid dangling pointer.
+     */
+    if (bond->header_slave_dev == slave_dev) {
+        bond->header_slave_dev = NULL;
+        bond_dev->header_ops = NULL;
+    }
+
     ret = __bond_release_one(bond_dev, slave_dev, false, true);
     if (ret == 0 && !bond_has_slaves(bond) &&
         bond_dev->reg_state != NETREG_UNREGISTERING) {
diff --git a/include/net/bonding.h b/include/net/bonding.h
index 95f67b308c19..cf8206187ce9 100644
--- a/include/net/bonding.h
+++ b/include/net/bonding.h
@@ -215,6 +215,11 @@ struct bond_ipsec {
  */
 struct bonding {
     struct   net_device *dev; /* first - useful for panic debug */
+    struct   net_device *header_slave_dev;  /* slave net_device for
header_ops */
+    /* maintained as a non-const variable
+     * because bond's header_ops should change depending on slaves.
+     */
+    struct   header_ops bond_header_ops;
     struct   slave __rcu *curr_active_slave;
     struct   slave __rcu *current_arp_slave;
     struct   slave __rcu *primary_slave;

Re: [PATCH net] bonding: Fix header_ops type confusion

[Eric Dumazet]

On Sun, May 25, 2025 at 10:08 PM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
>
> In bond_setup_by_slave(), the slave’s header_ops are unconditionally
> copied into the bonding device. As a result, the bonding device may invoke
> the slave-specific header operations on itself, causing
> netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
> as the slave's private-data type.
>
> This type-confusion bug can lead to out-of-bounds writes into the skb,
> resulting in memory corruption.
>
> This patch adds two members to struct bonding, bond_header_ops and
> header_slave_dev, to avoid type-confusion while keeping track of the
> slave's header_ops.
>
> Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
> Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> Co-Developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> Reviewed-by: Paolo Abeni <pabeni@redhat.com>
> Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> ---
>  drivers/net/bonding/bond_main.c | 61
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
>  include/net/bonding.h           |  5 +++++
>  2 files changed, 65 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> index 8ea183da8d53..690f3e0971d0 100644
> --- a/drivers/net/bonding/bond_main.c
> +++ b/drivers/net/bonding/bond_main.c
> @@ -1619,14 +1619,65 @@ static void bond_compute_features(struct bonding *bond)
>      netdev_change_features(bond_dev);
>  }
>
> +static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
> +        unsigned short type, const void *daddr,
> +        const void *saddr, unsigned int len)
> +{
> +    struct bonding *bond = netdev_priv(dev);
> +    struct net_device *slave_dev;
> +
> +    slave_dev = bond->header_slave_dev;
> +
> +    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
> +}
> +
> +static void bond_header_cache_update(struct hh_cache *hh, const
> struct net_device *dev,
> +        const unsigned char *haddr)
> +{
> +    const struct bonding *bond = netdev_priv(dev);
> +    struct net_device *slave_dev;
> +
> +    slave_dev = bond->header_slave_dev;

I do not see any barrier ?

> +
> +    if (!slave_dev->header_ops || !slave_dev->header_ops->cache_update)
> +        return;
> +
> +    slave_dev->header_ops->cache_update(hh, slave_dev, haddr);
> +}
> +
>  static void bond_setup_by_slave(struct net_device *bond_dev,
>                  struct net_device *slave_dev)
>  {
> +    struct bonding *bond = netdev_priv(bond_dev);
>      bool was_up = !!(bond_dev->flags & IFF_UP);
>
>      dev_close(bond_dev);
>
> -    bond_dev->header_ops        = slave_dev->header_ops;
> +    /* Some functions are given dev as an argument
> +     * while others not. When dev is not given, we cannot
> +     * find out what is the slave device through struct bonding
> +     * (the private data of bond_dev). Therefore, we need a raw
> +     * header_ops variable instead of its pointer to const header_ops
> +     * and assign slave's functions directly.
> +     * For the other case, we set the wrapper functions that pass
> +     * slave_dev to the wrapped functions.
> +     */
> +    bond->bond_header_ops.create = bond_hard_header;
> +    bond->bond_header_ops.cache_update = bond_header_cache_update;
> +    if (slave_dev->header_ops) {
> +        bond->bond_header_ops.parse = slave_dev->header_ops->parse;
> +        bond->bond_header_ops.cache = slave_dev->header_ops->cache;
> +        bond->bond_header_ops.validate = slave_dev->header_ops->validate;
> +        bond->bond_header_ops.parse_protocol =
> slave_dev->header_ops->parse_protocol;

All these updates probably need WRITE_ONCE(), and corresponding
READ_ONCE() on reader sides, at a very minimum ...

RCU would even be better later.


> +    } else {
> +        bond->bond_header_ops.parse = NULL;
> +        bond->bond_header_ops.cache = NULL;
> +        bond->bond_header_ops.validate = NULL;
> +        bond->bond_header_ops.parse_protocol = NULL;
> +    }
> +
> +    bond->header_slave_dev      = slave_dev;
> +    bond_dev->header_ops        = &bond->bond_header_ops;
>
>      bond_dev->type            = slave_dev->type;
>      bond_dev->hard_header_len   = slave_dev->hard_header_len;
> @@ -2676,6 +2727,14 @@ static int bond_release_and_destroy(struct
> net_device *bond_dev,
>      struct bonding *bond = netdev_priv(bond_dev);
>      int ret;
>
> +    /* If slave_dev is the earliest registered one, we must clear
> +     * the variables related to header_ops to avoid dangling pointer.
> +     */
> +    if (bond->header_slave_dev == slave_dev) {
> +        bond->header_slave_dev = NULL;
> +        bond_dev->header_ops = NULL;
> +    }
> +
>      ret = __bond_release_one(bond_dev, slave_dev, false, true);
>      if (ret == 0 && !bond_has_slaves(bond) &&
>          bond_dev->reg_state != NETREG_UNREGISTERING) {
> diff --git a/include/net/bonding.h b/include/net/bonding.h
> index 95f67b308c19..cf8206187ce9 100644
> --- a/include/net/bonding.h
> +++ b/include/net/bonding.h
> @@ -215,6 +215,11 @@ struct bond_ipsec {
>   */
>  struct bonding {
>      struct   net_device *dev; /* first - useful for panic debug */
> +    struct   net_device *header_slave_dev;  /* slave net_device for
> header_ops */
> +    /* maintained as a non-const variable
> +     * because bond's header_ops should change depending on slaves.
> +     */
> +    struct   header_ops bond_header_ops;
>      struct   slave __rcu *curr_active_slave;
>      struct   slave __rcu *current_arp_slave;
>      struct   slave __rcu *primary_slave;

Re: [PATCH net] bonding: Fix header_ops type confusion

[戸田晃太]

Thank you for your review.

2025年5月26日(月) 17:23 Eric Dumazet <edumazet@google.com>:
>
> On Sun, May 25, 2025 at 10:08 PM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > In bond_setup_by_slave(), the slave’s header_ops are unconditionally
> > copied into the bonding device. As a result, the bonding device may invoke
> > the slave-specific header operations on itself, causing
> > netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
> > as the slave's private-data type.
> >
> > This type-confusion bug can lead to out-of-bounds writes into the skb,
> > resulting in memory corruption.
> >
> > This patch adds two members to struct bonding, bond_header_ops and
> > header_slave_dev, to avoid type-confusion while keeping track of the
> > slave's header_ops.
> >
> > Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
> > Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > Co-Developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > Reviewed-by: Paolo Abeni <pabeni@redhat.com>
> > Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > ---
> >  drivers/net/bonding/bond_main.c | 61
> > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> >  include/net/bonding.h           |  5 +++++
> >  2 files changed, 65 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> > index 8ea183da8d53..690f3e0971d0 100644
> > --- a/drivers/net/bonding/bond_main.c
> > +++ b/drivers/net/bonding/bond_main.c
> > @@ -1619,14 +1619,65 @@ static void bond_compute_features(struct bonding *bond)
> >      netdev_change_features(bond_dev);
> >  }
> >
> > +static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
> > +        unsigned short type, const void *daddr,
> > +        const void *saddr, unsigned int len)
> > +{
> > +    struct bonding *bond = netdev_priv(dev);
> > +    struct net_device *slave_dev;
> > +
> > +    slave_dev = bond->header_slave_dev;
> > +
> > +    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
> > +}
> > +
> > +static void bond_header_cache_update(struct hh_cache *hh, const
> > struct net_device *dev,
> > +        const unsigned char *haddr)
> > +{
> > +    const struct bonding *bond = netdev_priv(dev);
> > +    struct net_device *slave_dev;
> > +
> > +    slave_dev = bond->header_slave_dev;
>
> I do not see any barrier ?
>
> > +
> > +    if (!slave_dev->header_ops || !slave_dev->header_ops->cache_update)
> > +        return;
> > +
> > +    slave_dev->header_ops->cache_update(hh, slave_dev, haddr);
> > +}
> > +
> >  static void bond_setup_by_slave(struct net_device *bond_dev,
> >                  struct net_device *slave_dev)
> >  {
> > +    struct bonding *bond = netdev_priv(bond_dev);
> >      bool was_up = !!(bond_dev->flags & IFF_UP);
> >
> >      dev_close(bond_dev);
> >
> > -    bond_dev->header_ops        = slave_dev->header_ops;
> > +    /* Some functions are given dev as an argument
> > +     * while others not. When dev is not given, we cannot
> > +     * find out what is the slave device through struct bonding
> > +     * (the private data of bond_dev). Therefore, we need a raw
> > +     * header_ops variable instead of its pointer to const header_ops
> > +     * and assign slave's functions directly.
> > +     * For the other case, we set the wrapper functions that pass
> > +     * slave_dev to the wrapped functions.
> > +     */
> > +    bond->bond_header_ops.create = bond_hard_header;
> > +    bond->bond_header_ops.cache_update = bond_header_cache_update;
> > +    if (slave_dev->header_ops) {
> > +        bond->bond_header_ops.parse = slave_dev->header_ops->parse;
> > +        bond->bond_header_ops.cache = slave_dev->header_ops->cache;
> > +        bond->bond_header_ops.validate = slave_dev->header_ops->validate;
> > +        bond->bond_header_ops.parse_protocol =
> > slave_dev->header_ops->parse_protocol;
>
> All these updates probably need WRITE_ONCE(), and corresponding
> READ_ONCE() on reader sides, at a very minimum ...
>
> RCU would even be better later.
>
I believe that locking is not necessary in this patch. The update of
`header_ops` only happens when a slave is newly enslaved to a bond.
Under such circumstances, members of `header_ops` are not called in
parallel with updating. Therefore, there is no possibility of race
conditions occurring.
>
> > +    } else {
> > +        bond->bond_header_ops.parse = NULL;
> > +        bond->bond_header_ops.cache = NULL;
> > +        bond->bond_header_ops.validate = NULL;
> > +        bond->bond_header_ops.parse_protocol = NULL;
> > +    }
> > +
> > +    bond->header_slave_dev      = slave_dev;
> > +    bond_dev->header_ops        = &bond->bond_header_ops;
> >
> >      bond_dev->type            = slave_dev->type;
> >      bond_dev->hard_header_len   = slave_dev->hard_header_len;
> > @@ -2676,6 +2727,14 @@ static int bond_release_and_destroy(struct
> > net_device *bond_dev,
> >      struct bonding *bond = netdev_priv(bond_dev);
> >      int ret;
> >
> > +    /* If slave_dev is the earliest registered one, we must clear
> > +     * the variables related to header_ops to avoid dangling pointer.
> > +     */
> > +    if (bond->header_slave_dev == slave_dev) {
> > +        bond->header_slave_dev = NULL;
> > +        bond_dev->header_ops = NULL;
> > +    }
> > +
> >      ret = __bond_release_one(bond_dev, slave_dev, false, true);
> >      if (ret == 0 && !bond_has_slaves(bond) &&
> >          bond_dev->reg_state != NETREG_UNREGISTERING) {
> > diff --git a/include/net/bonding.h b/include/net/bonding.h
> > index 95f67b308c19..cf8206187ce9 100644
> > --- a/include/net/bonding.h
> > +++ b/include/net/bonding.h
> > @@ -215,6 +215,11 @@ struct bond_ipsec {
> >   */
> >  struct bonding {
> >      struct   net_device *dev; /* first - useful for panic debug */
> > +    struct   net_device *header_slave_dev;  /* slave net_device for
> > header_ops */
> > +    /* maintained as a non-const variable
> > +     * because bond's header_ops should change depending on slaves.
> > +     */
> > +    struct   header_ops bond_header_ops;
> >      struct   slave __rcu *curr_active_slave;
> >      struct   slave __rcu *current_arp_slave;
> >      struct   slave __rcu *primary_slave;

Re: [PATCH net] bonding: Fix header_ops type confusion

[Eric Dumazet]

On Wed, May 28, 2025 at 7:36 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
>
> Thank you for your review.
>
> 2025年5月26日(月) 17:23 Eric Dumazet <edumazet@google.com>:
> >
> > On Sun, May 25, 2025 at 10:08 PM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> > >
> > > In bond_setup_by_slave(), the slave’s header_ops are unconditionally
> > > copied into the bonding device. As a result, the bonding device may invoke
> > > the slave-specific header operations on itself, causing
> > > netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
> > > as the slave's private-data type.
> > >
> > > This type-confusion bug can lead to out-of-bounds writes into the skb,
> > > resulting in memory corruption.
> > >
> > > This patch adds two members to struct bonding, bond_header_ops and
> > > header_slave_dev, to avoid type-confusion while keeping track of the
> > > slave's header_ops.
> > >
> > > Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
> > > Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > Co-Developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > Reviewed-by: Paolo Abeni <pabeni@redhat.com>
> > > Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > ---
> > >  drivers/net/bonding/bond_main.c | 61
> > > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> > >  include/net/bonding.h           |  5 +++++
> > >  2 files changed, 65 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> > > index 8ea183da8d53..690f3e0971d0 100644
> > > --- a/drivers/net/bonding/bond_main.c
> > > +++ b/drivers/net/bonding/bond_main.c
> > > @@ -1619,14 +1619,65 @@ static void bond_compute_features(struct bonding *bond)
> > >      netdev_change_features(bond_dev);
> > >  }
> > >
> > > +static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
> > > +        unsigned short type, const void *daddr,
> > > +        const void *saddr, unsigned int len)
> > > +{
> > > +    struct bonding *bond = netdev_priv(dev);
> > > +    struct net_device *slave_dev;
> > > +
> > > +    slave_dev = bond->header_slave_dev;
> > > +
> > > +    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
> > > +}
> > > +
> > > +static void bond_header_cache_update(struct hh_cache *hh, const
> > > struct net_device *dev,
> > > +        const unsigned char *haddr)
> > > +{
> > > +    const struct bonding *bond = netdev_priv(dev);
> > > +    struct net_device *slave_dev;
> > > +
> > > +    slave_dev = bond->header_slave_dev;
> >
> > I do not see any barrier ?
> >
> > > +
> > > +    if (!slave_dev->header_ops || !slave_dev->header_ops->cache_update)
> > > +        return;
> > > +
> > > +    slave_dev->header_ops->cache_update(hh, slave_dev, haddr);
> > > +}
> > > +
> > >  static void bond_setup_by_slave(struct net_device *bond_dev,
> > >                  struct net_device *slave_dev)
> > >  {
> > > +    struct bonding *bond = netdev_priv(bond_dev);
> > >      bool was_up = !!(bond_dev->flags & IFF_UP);
> > >
> > >      dev_close(bond_dev);
> > >
> > > -    bond_dev->header_ops        = slave_dev->header_ops;
> > > +    /* Some functions are given dev as an argument
> > > +     * while others not. When dev is not given, we cannot
> > > +     * find out what is the slave device through struct bonding
> > > +     * (the private data of bond_dev). Therefore, we need a raw
> > > +     * header_ops variable instead of its pointer to const header_ops
> > > +     * and assign slave's functions directly.
> > > +     * For the other case, we set the wrapper functions that pass
> > > +     * slave_dev to the wrapped functions.
> > > +     */
> > > +    bond->bond_header_ops.create = bond_hard_header;
> > > +    bond->bond_header_ops.cache_update = bond_header_cache_update;
> > > +    if (slave_dev->header_ops) {
> > > +        bond->bond_header_ops.parse = slave_dev->header_ops->parse;
> > > +        bond->bond_header_ops.cache = slave_dev->header_ops->cache;
> > > +        bond->bond_header_ops.validate = slave_dev->header_ops->validate;
> > > +        bond->bond_header_ops.parse_protocol =
> > > slave_dev->header_ops->parse_protocol;
> >
> > All these updates probably need WRITE_ONCE(), and corresponding
> > READ_ONCE() on reader sides, at a very minimum ...
> >
> > RCU would even be better later.
> >
> I believe that locking is not necessary in this patch. The update of
> `header_ops` only happens when a slave is newly enslaved to a bond.
> Under such circumstances, members of `header_ops` are not called in
> parallel with updating. Therefore, there is no possibility of race
> conditions occurring.

bond_dev can certainly be live, and packets can flow.

I have seen enough syzbot reports hinting at this precise issue.

Re: [PATCH net] bonding: Fix header_ops type confusion

[戸田晃太]

2025年5月29日(木) 0:10 Eric Dumazet <edumazet@google.com>:

>
> On Wed, May 28, 2025 at 7:36 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Thank you for your review.
> >
> > 2025年5月26日(月) 17:23 Eric Dumazet <edumazet@google.com>:
> > >
> > > On Sun, May 25, 2025 at 10:08 PM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> > > >
> > > > In bond_setup_by_slave(), the slave’s header_ops are unconditionally
> > > > copied into the bonding device. As a result, the bonding device may invoke
> > > > the slave-specific header operations on itself, causing
> > > > netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
> > > > as the slave's private-data type.
> > > >
> > > > This type-confusion bug can lead to out-of-bounds writes into the skb,
> > > > resulting in memory corruption.
> > > >
> > > > This patch adds two members to struct bonding, bond_header_ops and
> > > > header_slave_dev, to avoid type-confusion while keeping track of the
> > > > slave's header_ops.
> > > >
> > > > Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
> > > > Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > > Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > > Co-Developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > > Reviewed-by: Paolo Abeni <pabeni@redhat.com>
> > > > Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > > ---
> > > >  drivers/net/bonding/bond_main.c | 61
> > > > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> > > >  include/net/bonding.h           |  5 +++++
> > > >  2 files changed, 65 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> > > > index 8ea183da8d53..690f3e0971d0 100644
> > > > --- a/drivers/net/bonding/bond_main.c
> > > > +++ b/drivers/net/bonding/bond_main.c
> > > > @@ -1619,14 +1619,65 @@ static void bond_compute_features(struct bonding *bond)
> > > >      netdev_change_features(bond_dev);
> > > >  }
> > > >
> > > > +static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
> > > > +        unsigned short type, const void *daddr,
> > > > +        const void *saddr, unsigned int len)
> > > > +{
> > > > +    struct bonding *bond = netdev_priv(dev);
> > > > +    struct net_device *slave_dev;
> > > > +
> > > > +    slave_dev = bond->header_slave_dev;
> > > > +
> > > > +    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
> > > > +}
> > > > +
> > > > +static void bond_header_cache_update(struct hh_cache *hh, const
> > > > struct net_device *dev,
> > > > +        const unsigned char *haddr)
> > > > +{
> > > > +    const struct bonding *bond = netdev_priv(dev);
> > > > +    struct net_device *slave_dev;
> > > > +
> > > > +    slave_dev = bond->header_slave_dev;
> > >
> > > I do not see any barrier ?
> > >
> > > > +
> > > > +    if (!slave_dev->header_ops || !slave_dev->header_ops->cache_update)
> > > > +        return;
> > > > +
> > > > +    slave_dev->header_ops->cache_update(hh, slave_dev, haddr);
> > > > +}
> > > > +
> > > >  static void bond_setup_by_slave(struct net_device *bond_dev,
> > > >                  struct net_device *slave_dev)
> > > >  {
> > > > +    struct bonding *bond = netdev_priv(bond_dev);
> > > >      bool was_up = !!(bond_dev->flags & IFF_UP);
> > > >
> > > >      dev_close(bond_dev);
> > > >
> > > > -    bond_dev->header_ops        = slave_dev->header_ops;
> > > > +    /* Some functions are given dev as an argument
> > > > +     * while others not. When dev is not given, we cannot
> > > > +     * find out what is the slave device through struct bonding
> > > > +     * (the private data of bond_dev). Therefore, we need a raw
> > > > +     * header_ops variable instead of its pointer to const header_ops
> > > > +     * and assign slave's functions directly.
> > > > +     * For the other case, we set the wrapper functions that pass
> > > > +     * slave_dev to the wrapped functions.
> > > > +     */
> > > > +    bond->bond_header_ops.create = bond_hard_header;
> > > > +    bond->bond_header_ops.cache_update = bond_header_cache_update;
> > > > +    if (slave_dev->header_ops) {
> > > > +        bond->bond_header_ops.parse = slave_dev->header_ops->parse;
> > > > +        bond->bond_header_ops.cache = slave_dev->header_ops->cache;
> > > > +        bond->bond_header_ops.validate = slave_dev->header_ops->validate;
> > > > +        bond->bond_header_ops.parse_protocol =
> > > > slave_dev->header_ops->parse_protocol;
> > >
> > > All these updates probably need WRITE_ONCE(), and corresponding
> > > READ_ONCE() on reader sides, at a very minimum ...
> > >
> > > RCU would even be better later.
> > >
> > I believe that locking is not necessary in this patch. The update of
> > `header_ops` only happens when a slave is newly enslaved to a bond.
> > Under such circumstances, members of `header_ops` are not called in
> > parallel with updating. Therefore, there is no possibility of race
> > conditions occurring.
>
> bond_dev can certainly be live, and packets can flow.
>
> I have seen enough syzbot reports hinting at this precise issue.

Hi Eric, Thank you for reviewing the patch.

At the beginning of `bond_setup_by_slave`, `dev_close(bond_dev)` is called,
meaning bond_dev is down and no packets can flow during the update of
`bond_header_ops`.

The syzbot report (you mentioned in the conversation in security@) indicating
`dev->header_ops` becoming NULL should be resolved by this patch.
I couldn't find any other related syzbot reports.

Re: [PATCH net] bonding: Fix header_ops type confusion

[小池悠生]

Hello, Eric and other maintainers,

I'm deeply sorry to have left the patch suggestion for this long period.
I became extremely busy, and that took its toll on my health, causing
me to take sick leave for nearly half a year (And my colleague Kota
had been waiting for me to come back).
As fortunately I've recovered and returned to work, I hope to move
forward with this matter as well.

Recalling the issue Eric raised, I understand it was a concern about
potential race conditions arising from the `bond_header_ops` and
`header_slave_dev` added to the `struct bonding`. For example, one
could imagine a situation where `header_slave_dev` is rewritten to a
different type, and at that exact moment a function from the old
`bond_header_ops` gets called, or vice versa.

However, I am actually skeptical that this can happen. The reason is
that `bond_setup_by_slave` is only called when there are no slaves at
all:

```
bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev,
             struct netlink_ext_ack *extack)
...
  if (!bond_has_slaves(bond)) {
...
      if (slave_dev->type != ARPHRD_ETHER)
        bond_setup_by_slave(bond_dev, slave_dev);
```

In other words, in order to trigger a race condition, one would need
to remove the slave once and make the slave list empty first. However,
as shown below, in `bond_release_and_destroy`, when the slave list
becomes empty, it appears that the bond interface itself is removed.
This makes it seem impossible to "quickly remove a slave and
re-register it.":

```
static int bond_slave_netdev_event(unsigned long event,
           struct net_device *slave_dev)
...
  switch (event) {
  case NETDEV_UNREGISTER:
    if (bond_dev->type != ARPHRD_ETHER)
      bond_release_and_destroy(bond_dev, slave_dev);
...
}
...
/* First release a slave and then destroy the bond if no more slaves are left.
 * Must be under rtnl_lock when this function is called.
 */
static int bond_release_and_destroy(struct net_device *bond_dev,
            struct net_device *slave_dev)
{
  struct bonding *bond = netdev_priv(bond_dev);
  int ret;

  ret = __bond_release_one(bond_dev, slave_dev, false, true);
  if (ret == 0 && !bond_has_slaves(bond) &&
      bond_dev->reg_state != NETREG_UNREGISTERING) {
    bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
    netdev_info(bond_dev, "Destroying bond\n");
    bond_remove_proc_entry(bond);
    unregister_netdevice(bond_dev);
  }
  return ret;
}
```

Moreover, as noted in the comments, these functions are executed under
the netlink-side lock. Therefore, my conclusion is that a race
condition cannot actually occur. I also think that the fact that, even
before our patch, these code paths had almost no explicit locking
anywhere serves as circumstantial evidence for this view. As Kota
said, as far as I saw, the past syzkaller-bot's report is seemingly
only NULL pointer dereference due to the root cause we reported, and
this patch should fix them.

That said, even so, I agree that the kind of countermeasures Eric
suggests are worth applying if they do not cause problems in terms of
execution speed or code size. However, I am concerned that addressing
this with READ_ONCE or RCU would imply a somewhat large amount of
rewriting.
`header_ops` is designed to allow various types of devices to be
handled in an object-oriented way, and as such it is used throughout
many parts of the Linux kernel. Using READ_ONCE or RCU every time
header_ops is accessed simply because we are worried about a race
condition in bond’s header_ops seems to imply changes like the
following, for example:

```
diff --git a/net/core/neighbour.c b/net/core/neighbour.c
index 92dc1f1788de..d9aad38746ad 100644
--- a/net/core/neighbour.c
+++ b/net/core/neighbour.c
@@ -1538,7 +1538,7 @@ static void neigh_hh_init(struct neighbour *n)
  * hh_cache entry.
  */
  if (!hh->hh_len)
- dev->header_ops->cache(n, hh, prot);
+ READ_ONCE(dev->header_ops->cache)(n, hh, prot);

  write_unlock_bh(&n->lock);
 }
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3131,35 +3131,41 @@ static inline int dev_hard_header(struct
sk_buff *skb, struct net_device *dev,
   const void *daddr, const void *saddr,
   unsigned int len)
 {
- if (!dev->header_ops || !dev->header_ops->create)
+ int (*create)(struct sk_buff *skb, struct net_device *dev,
+      unsigned short type, const void *daddr,
+      const void *saddr, unsigned int len);
+ if (!dev->header_ops || !(create = READ_ONCE(dev->header_ops->create)))
  return 0;

- return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
+ return create(skb, dev, type, daddr, saddr, len);
 }

 static inline int dev_parse_header(const struct sk_buff *skb,
    unsigned char *haddr)
 {
+ int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
  const struct net_device *dev = skb->dev;

- if (!dev->header_ops || !dev->header_ops->parse)
+ if (!dev->header_ops || !(parse = READ_ONCE(dev->header_ops->parse)))
  return 0;
- return dev->header_ops->parse(skb, haddr);
+ return parse(skb, haddr);
 }
... (and so on)
```

It looks like we would end up rewriting on the order of a dozen or so
places with this kind of pattern, but from the perspective of the
maintainers (or in terms of Linux kernel culture), would this be
considered an acceptable change?
If this differs from what you intended, please correct me.

Best regards,
Yuki Koike

2025年5月29日(木) 0:10 Eric Dumazet <edumazet@google.com>:
>
> On Wed, May 28, 2025 at 7:36 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Thank you for your review.
> >
> > 2025年5月26日(月) 17:23 Eric Dumazet <edumazet@google.com>:
> > >
> > > On Sun, May 25, 2025 at 10:08 PM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> > > >
> > > > In bond_setup_by_slave(), the slave’s header_ops are unconditionally
> > > > copied into the bonding device. As a result, the bonding device may invoke
> > > > the slave-specific header operations on itself, causing
> > > > netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
> > > > as the slave's private-data type.
> > > >
> > > > This type-confusion bug can lead to out-of-bounds writes into the skb,
> > > > resulting in memory corruption.
> > > >
> > > > This patch adds two members to struct bonding, bond_header_ops and
> > > > header_slave_dev, to avoid type-confusion while keeping track of the
> > > > slave's header_ops.
> > > >
> > > > Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
> > > > Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > > Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > > Co-Developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > > Reviewed-by: Paolo Abeni <pabeni@redhat.com>
> > > > Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > > ---
> > > >  drivers/net/bonding/bond_main.c | 61
> > > > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> > > >  include/net/bonding.h           |  5 +++++
> > > >  2 files changed, 65 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> > > > index 8ea183da8d53..690f3e0971d0 100644
> > > > --- a/drivers/net/bonding/bond_main.c
> > > > +++ b/drivers/net/bonding/bond_main.c
> > > > @@ -1619,14 +1619,65 @@ static void bond_compute_features(struct bonding *bond)
> > > >      netdev_change_features(bond_dev);
> > > >  }
> > > >
> > > > +static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
> > > > +        unsigned short type, const void *daddr,
> > > > +        const void *saddr, unsigned int len)
> > > > +{
> > > > +    struct bonding *bond = netdev_priv(dev);
> > > > +    struct net_device *slave_dev;
> > > > +
> > > > +    slave_dev = bond->header_slave_dev;
> > > > +
> > > > +    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
> > > > +}
> > > > +
> > > > +static void bond_header_cache_update(struct hh_cache *hh, const
> > > > struct net_device *dev,
> > > > +        const unsigned char *haddr)
> > > > +{
> > > > +    const struct bonding *bond = netdev_priv(dev);
> > > > +    struct net_device *slave_dev;
> > > > +
> > > > +    slave_dev = bond->header_slave_dev;
> > >
> > > I do not see any barrier ?
> > >
> > > > +
> > > > +    if (!slave_dev->header_ops || !slave_dev->header_ops->cache_update)
> > > > +        return;
> > > > +
> > > > +    slave_dev->header_ops->cache_update(hh, slave_dev, haddr);
> > > > +}
> > > > +
> > > >  static void bond_setup_by_slave(struct net_device *bond_dev,
> > > >                  struct net_device *slave_dev)
> > > >  {
> > > > +    struct bonding *bond = netdev_priv(bond_dev);
> > > >      bool was_up = !!(bond_dev->flags & IFF_UP);
> > > >
> > > >      dev_close(bond_dev);
> > > >
> > > > -    bond_dev->header_ops        = slave_dev->header_ops;
> > > > +    /* Some functions are given dev as an argument
> > > > +     * while others not. When dev is not given, we cannot
> > > > +     * find out what is the slave device through struct bonding
> > > > +     * (the private data of bond_dev). Therefore, we need a raw
> > > > +     * header_ops variable instead of its pointer to const header_ops
> > > > +     * and assign slave's functions directly.
> > > > +     * For the other case, we set the wrapper functions that pass
> > > > +     * slave_dev to the wrapped functions.
> > > > +     */
> > > > +    bond->bond_header_ops.create = bond_hard_header;
> > > > +    bond->bond_header_ops.cache_update = bond_header_cache_update;
> > > > +    if (slave_dev->header_ops) {
> > > > +        bond->bond_header_ops.parse = slave_dev->header_ops->parse;
> > > > +        bond->bond_header_ops.cache = slave_dev->header_ops->cache;
> > > > +        bond->bond_header_ops.validate = slave_dev->header_ops->validate;
> > > > +        bond->bond_header_ops.parse_protocol =
> > > > slave_dev->header_ops->parse_protocol;
> > >
> > > All these updates probably need WRITE_ONCE(), and corresponding
> > > READ_ONCE() on reader sides, at a very minimum ...
> > >
> > > RCU would even be better later.
> > >
> > I believe that locking is not necessary in this patch. The update of
> > `header_ops` only happens when a slave is newly enslaved to a bond.
> > Under such circumstances, members of `header_ops` are not called in
> > parallel with updating. Therefore, there is no possibility of race
> > conditions occurring.
>
> bond_dev can certainly be live, and packets can flow.
>
> I have seen enough syzbot reports hinting at this precise issue.

Re: [PATCH net] bonding: Fix header_ops type confusion

[戸田晃太]

Hello, Eric and other maintainers,

I hope you’re doing well. I’m following up on our email, sent during
the holiday season, in case it got buried.

When you have a moment, could you please let us know if you had a
chance to review it?

Thank you in advance, and I look forward to your response.

Best regards,
Kota Toda

2025年12月22日(月) 17:20 小池悠生 <yuki.koike@gmo-cybersecurity.com>:
>
> Hello, Eric and other maintainers,
>
> I'm deeply sorry to have left the patch suggestion for this long period.
> I became extremely busy, and that took its toll on my health, causing
> me to take sick leave for nearly half a year (And my colleague Kota
> had been waiting for me to come back).
> As fortunately I've recovered and returned to work, I hope to move
> forward with this matter as well.
>
> Recalling the issue Eric raised, I understand it was a concern about
> potential race conditions arising from the `bond_header_ops` and
> `header_slave_dev` added to the `struct bonding`. For example, one
> could imagine a situation where `header_slave_dev` is rewritten to a
> different type, and at that exact moment a function from the old
> `bond_header_ops` gets called, or vice versa.
>
> However, I am actually skeptical that this can happen. The reason is
> that `bond_setup_by_slave` is only called when there are no slaves at
> all:
>
> ```
> bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev,
>              struct netlink_ext_ack *extack)
> ...
>   if (!bond_has_slaves(bond)) {
> ...
>       if (slave_dev->type != ARPHRD_ETHER)
>         bond_setup_by_slave(bond_dev, slave_dev);
> ```
>
> In other words, in order to trigger a race condition, one would need
> to remove the slave once and make the slave list empty first. However,
> as shown below, in `bond_release_and_destroy`, when the slave list
> becomes empty, it appears that the bond interface itself is removed.
> This makes it seem impossible to "quickly remove a slave and
> re-register it.":
>
> ```
> static int bond_slave_netdev_event(unsigned long event,
>            struct net_device *slave_dev)
> ...
>   switch (event) {
>   case NETDEV_UNREGISTER:
>     if (bond_dev->type != ARPHRD_ETHER)
>       bond_release_and_destroy(bond_dev, slave_dev);
> ...
> }
> ...
> /* First release a slave and then destroy the bond if no more slaves are left.
>  * Must be under rtnl_lock when this function is called.
>  */
> static int bond_release_and_destroy(struct net_device *bond_dev,
>             struct net_device *slave_dev)
> {
>   struct bonding *bond = netdev_priv(bond_dev);
>   int ret;
>
>   ret = __bond_release_one(bond_dev, slave_dev, false, true);
>   if (ret == 0 && !bond_has_slaves(bond) &&
>       bond_dev->reg_state != NETREG_UNREGISTERING) {
>     bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
>     netdev_info(bond_dev, "Destroying bond\n");
>     bond_remove_proc_entry(bond);
>     unregister_netdevice(bond_dev);
>   }
>   return ret;
> }
> ```
>
> Moreover, as noted in the comments, these functions are executed under
> the netlink-side lock. Therefore, my conclusion is that a race
> condition cannot actually occur. I also think that the fact that, even
> before our patch, these code paths had almost no explicit locking
> anywhere serves as circumstantial evidence for this view. As Kota
> said, as far as I saw, the past syzkaller-bot's report is seemingly
> only NULL pointer dereference due to the root cause we reported, and
> this patch should fix them.
>
> That said, even so, I agree that the kind of countermeasures Eric
> suggests are worth applying if they do not cause problems in terms of
> execution speed or code size. However, I am concerned that addressing
> this with READ_ONCE or RCU would imply a somewhat large amount of
> rewriting.
> `header_ops` is designed to allow various types of devices to be
> handled in an object-oriented way, and as such it is used throughout
> many parts of the Linux kernel. Using READ_ONCE or RCU every time
> header_ops is accessed simply because we are worried about a race
> condition in bond’s header_ops seems to imply changes like the
> following, for example:
>
> ```
> diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> index 92dc1f1788de..d9aad38746ad 100644
> --- a/net/core/neighbour.c
> +++ b/net/core/neighbour.c
> @@ -1538,7 +1538,7 @@ static void neigh_hh_init(struct neighbour *n)
>   * hh_cache entry.
>   */
>   if (!hh->hh_len)
> - dev->header_ops->cache(n, hh, prot);
> + READ_ONCE(dev->header_ops->cache)(n, hh, prot);
>
>   write_unlock_bh(&n->lock);
>  }
> --- a/include/linux/netdevice.h
> +++ b/include/linux/netdevice.h
> @@ -3131,35 +3131,41 @@ static inline int dev_hard_header(struct
> sk_buff *skb, struct net_device *dev,
>    const void *daddr, const void *saddr,
>    unsigned int len)
>  {
> - if (!dev->header_ops || !dev->header_ops->create)
> + int (*create)(struct sk_buff *skb, struct net_device *dev,
> +      unsigned short type, const void *daddr,
> +      const void *saddr, unsigned int len);
> + if (!dev->header_ops || !(create = READ_ONCE(dev->header_ops->create)))
>   return 0;
>
> - return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
> + return create(skb, dev, type, daddr, saddr, len);
>  }
>
>  static inline int dev_parse_header(const struct sk_buff *skb,
>     unsigned char *haddr)
>  {
> + int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
>   const struct net_device *dev = skb->dev;
>
> - if (!dev->header_ops || !dev->header_ops->parse)
> + if (!dev->header_ops || !(parse = READ_ONCE(dev->header_ops->parse)))
>   return 0;
> - return dev->header_ops->parse(skb, haddr);
> + return parse(skb, haddr);
>  }
> ... (and so on)
> ```
>
> It looks like we would end up rewriting on the order of a dozen or so
> places with this kind of pattern, but from the perspective of the
> maintainers (or in terms of Linux kernel culture), would this be
> considered an acceptable change?
> If this differs from what you intended, please correct me.
>
> Best regards,
> Yuki Koike
>
> 2025年5月29日(木) 0:10 Eric Dumazet <edumazet@google.com>:
> >
> > On Wed, May 28, 2025 at 7:36 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> > >
> > > Thank you for your review.
> > >
> > > 2025年5月26日(月) 17:23 Eric Dumazet <edumazet@google.com>:
> > > >
> > > > On Sun, May 25, 2025 at 10:08 PM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> > > > >
> > > > > In bond_setup_by_slave(), the slave’s header_ops are unconditionally
> > > > > copied into the bonding device. As a result, the bonding device may invoke
> > > > > the slave-specific header operations on itself, causing
> > > > > netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
> > > > > as the slave's private-data type.
> > > > >
> > > > > This type-confusion bug can lead to out-of-bounds writes into the skb,
> > > > > resulting in memory corruption.
> > > > >
> > > > > This patch adds two members to struct bonding, bond_header_ops and
> > > > > header_slave_dev, to avoid type-confusion while keeping track of the
> > > > > slave's header_ops.
> > > > >
> > > > > Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
> > > > > Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > > > Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > > > Co-Developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> > > > > Reviewed-by: Paolo Abeni <pabeni@redhat.com>
> > > > > Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> > > > > ---
> > > > >  drivers/net/bonding/bond_main.c | 61
> > > > > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> > > > >  include/net/bonding.h           |  5 +++++
> > > > >  2 files changed, 65 insertions(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> > > > > index 8ea183da8d53..690f3e0971d0 100644
> > > > > --- a/drivers/net/bonding/bond_main.c
> > > > > +++ b/drivers/net/bonding/bond_main.c
> > > > > @@ -1619,14 +1619,65 @@ static void bond_compute_features(struct bonding *bond)
> > > > >      netdev_change_features(bond_dev);
> > > > >  }
> > > > >
> > > > > +static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
> > > > > +        unsigned short type, const void *daddr,
> > > > > +        const void *saddr, unsigned int len)
> > > > > +{
> > > > > +    struct bonding *bond = netdev_priv(dev);
> > > > > +    struct net_device *slave_dev;
> > > > > +
> > > > > +    slave_dev = bond->header_slave_dev;
> > > > > +
> > > > > +    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
> > > > > +}
> > > > > +
> > > > > +static void bond_header_cache_update(struct hh_cache *hh, const
> > > > > struct net_device *dev,
> > > > > +        const unsigned char *haddr)
> > > > > +{
> > > > > +    const struct bonding *bond = netdev_priv(dev);
> > > > > +    struct net_device *slave_dev;
> > > > > +
> > > > > +    slave_dev = bond->header_slave_dev;
> > > >
> > > > I do not see any barrier ?
> > > >
> > > > > +
> > > > > +    if (!slave_dev->header_ops || !slave_dev->header_ops->cache_update)
> > > > > +        return;
> > > > > +
> > > > > +    slave_dev->header_ops->cache_update(hh, slave_dev, haddr);
> > > > > +}
> > > > > +
> > > > >  static void bond_setup_by_slave(struct net_device *bond_dev,
> > > > >                  struct net_device *slave_dev)
> > > > >  {
> > > > > +    struct bonding *bond = netdev_priv(bond_dev);
> > > > >      bool was_up = !!(bond_dev->flags & IFF_UP);
> > > > >
> > > > >      dev_close(bond_dev);
> > > > >
> > > > > -    bond_dev->header_ops        = slave_dev->header_ops;
> > > > > +    /* Some functions are given dev as an argument
> > > > > +     * while others not. When dev is not given, we cannot
> > > > > +     * find out what is the slave device through struct bonding
> > > > > +     * (the private data of bond_dev). Therefore, we need a raw
> > > > > +     * header_ops variable instead of its pointer to const header_ops
> > > > > +     * and assign slave's functions directly.
> > > > > +     * For the other case, we set the wrapper functions that pass
> > > > > +     * slave_dev to the wrapped functions.
> > > > > +     */
> > > > > +    bond->bond_header_ops.create = bond_hard_header;
> > > > > +    bond->bond_header_ops.cache_update = bond_header_cache_update;
> > > > > +    if (slave_dev->header_ops) {
> > > > > +        bond->bond_header_ops.parse = slave_dev->header_ops->parse;
> > > > > +        bond->bond_header_ops.cache = slave_dev->header_ops->cache;
> > > > > +        bond->bond_header_ops.validate = slave_dev->header_ops->validate;
> > > > > +        bond->bond_header_ops.parse_protocol =
> > > > > slave_dev->header_ops->parse_protocol;
> > > >
> > > > All these updates probably need WRITE_ONCE(), and corresponding
> > > > READ_ONCE() on reader sides, at a very minimum ...
> > > >
> > > > RCU would even be better later.
> > > >
> > > I believe that locking is not necessary in this patch. The update of
> > > `header_ops` only happens when a slave is newly enslaved to a bond.
> > > Under such circumstances, members of `header_ops` are not called in
> > > parallel with updating. Therefore, there is no possibility of race
> > > conditions occurring.
> >
> > bond_dev can certainly be live, and packets can flow.
> >
> > I have seen enough syzbot reports hinting at this precise issue.

Re: [PATCH net] bonding: Fix header_ops type confusion

[Eric Dumazet]

On Thu, Jan 15, 2026 at 11:33 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
>
> Hello, Eric and other maintainers,
>
> I hope you’re doing well. I’m following up on our email, sent during
> the holiday season, in case it got buried.
>
> When you have a moment, could you please let us know if you had a
> chance to review it?
>
> Thank you in advance, and I look forward to your response.
>

I think it would be nice to provide an actual stack trace of the bug,
on a recent kernel tree.

We had recent patches dealing with dev->hard_header_len changes.

Re: [PATCH net] bonding: Fix header_ops type confusion

[戸田晃太]

Thanks for your quick response.

The following information is based on Linux kernel version 6.12.65,
the latest release in the 6.12 tree.
The kernel config is identical to that of the kernelCTF instance
(available at: https://storage.googleapis.com/kernelctf-build/releases/lts-6.12.65/.config)


This type confusion occurs in several locations, including,
for example, `ipgre_header` (`header_ops->create`),
where the private data of the network device is incorrectly cast as
`struct ip_tunnel *`.

```
static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
      unsigned short type,
      const void *daddr, const void *saddr, unsigned int len)
{
  struct ip_tunnel *t = netdev_priv(dev);
  struct iphdr *iph;
  struct gre_base_hdr *greh;
...
```

When a bond interface is given to this function,
it should not reference the private data as `struct ip_tunnel *`,
because the bond interface uses the private data as `struct bonding *`.
(quickly confirmed by seeing drivers/net/bonding/bond_netlink.c:909)

```
struct rtnl_link_ops bond_link_ops __read_mostly = {
    .kind            = "bond",
    .priv_size        = sizeof(struct bonding),
...
```

The stack trace below is the backtrace of all stack frame during a
call to `ipgre_header`.

```
ipgre_header at net/ipv4/ip_gre.c:890
dev_hard_header at ./include/linux/netdevice.h:3156
packet_snd at net/packet/af_packet.c:3082
packet_sendmsg at net/packet/af_packet.c:3162
sock_sendmsg_nosec at net/socket.c:729
__sock_sendmsg at net/socket.c:744
__sys_sendto at net/socket.c:2213
__do_sys_sendto at net/socket.c:2225
__se_sys_sendto at net/socket.c:2221
__x64_sys_sendto at net/socket.c:2221
do_syscall_x64 at arch/x86/entry/common.c:47
do_syscall_64 at arch/x86/entry/common.c:78
entry_SYSCALL_64 at arch/x86/entry/entry_64.S:121
```

This causes memory corruption during subsequent operations.

The following stack trace shows a General Protection Fault triggered
when sending a packet
to a bonding interface that has an IPv4 GRE interface as a slave.

```
[    1.712329] Oops: general protection fault, probably for
non-canonical address 0xdead0000cafebabe: 0000 [#1] SMP NOPTI
[    1.712972] CPU: 0 UID: 1000 PID: 205 Comm: exp Not tainted 6.12.65 #1
[    1.713344] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS Arch Linux 1.17.0-2-2 04/01/2014
[    1.713890] RIP: 0010:skb_release_data+0x8a/0x1c0
[    1.714162] Code: c0 00 00 00 49 03 86 c8 00 00 00 0f b6 10 f6 c2
01 74 48 48 8b 70 28 48 85 f6 74 3f 41 0f b6 5d 00 83 e3 10 40 f6 c6
01 75 24 <48> 8b 06 ba 01 00 00 00 4c 89 f7 48 8b 00 ff d0 0f 1f 00 41
8b6
[    1.715276] RSP: 0018:ffffc900007cfcc0 EFLAGS: 00010246
[    1.715583] RAX: ffff888106fe12c0 RBX: 0000000000000010 RCX: 0000000000000000
[    1.716036] RDX: 0000000000000017 RSI: dead0000cafebabe RDI: ffff8881059c4a00
[    1.716504] RBP: ffffc900007cfe10 R08: 0000000000000010 R09: 0000000000000000
[    1.716955] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
[    1.717429] R13: ffff888106fe12c0 R14: ffff8881059c4a00 R15: ffff888106e57000
[    1.717866] FS:  0000000038e54380(0000) GS:ffff88813bc00000(0000)
knlGS:0000000000000000
[    1.718350] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    1.718703] CR2: 00000000004bf480 CR3: 00000001009ec001 CR4: 0000000000772ef0
[    1.719109] PKRU: 55555554
[    1.719297] Call Trace:
[    1.719461]  <TASK>
[    1.719611]  sk_skb_reason_drop+0x58/0x120
[    1.719891]  packet_sendmsg+0xbcb/0x18f0
[    1.720166]  ? pcpu_alloc_area+0x186/0x260
[    1.720421]  __sys_sendto+0x1e2/0x1f0
[    1.720691]  __x64_sys_sendto+0x24/0x30
[    1.720948]  do_syscall_64+0x58/0x120
[    1.721174]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
[    1.721509] RIP: 0033:0x42860d
[    1.721713] Code: c3 ff ff ff ff 64 89 02 eb b9 0f 1f 00 f3 0f 1e
fa 80 3d 5d 4a 09 00 00 41 89 ca 74 20 45 31 c9 45 31 c0 b8 2c 00 00
00 0f 05 <48> 3d 00 f0 ff ff 77 6b c3 66 2e 0f 1f 84 00 00 00 00 00 55
489
[    1.722837] RSP: 002b:00007fff597e95e8 EFLAGS: 00000246 ORIG_RAX:
000000000000002c
[    1.723315] RAX: ffffffffffffffda RBX: 00000000000003e8 RCX: 000000000042860d
[    1.723721] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000310
[    1.724103] RBP: 00007fff597e9880 R08: 0000000000000000 R09: 0000000000000000
[    1.724565] R10: 0000000000000000 R11: 0000000000000246 R12: 00007fff597e99f8
[    1.725010] R13: 00007fff597e9a08 R14: 00000000004b7828 R15: 0000000000000001
[    1.725441]  </TASK>
[    1.725594] Modules linked in:
[    1.725790] ---[ end trace 0000000000000000 ]---
[    1.726057] RIP: 0010:skb_release_data+0x8a/0x1c0
[    1.726339] Code: c0 00 00 00 49 03 86 c8 00 00 00 0f b6 10 f6 c2
01 74 48 48 8b 70 28 48 85 f6 74 3f 41 0f b6 5d 00 83 e3 10 40 f6 c6
01 75 24 <48> 8b 06 ba 01 00 00 00 4c 89 f7 48 8b 00 ff d0 0f 1f 00 41
8b6
[    1.727285] RSP: 0018:ffffc900007cfcc0 EFLAGS: 00010246
[    1.727623] RAX: ffff888106fe12c0 RBX: 0000000000000010 RCX: 0000000000000000
[    1.728052] RDX: 0000000000000017 RSI: dead0000cafebabe RDI: ffff8881059c4a00
[    1.728467] RBP: ffffc900007cfe10 R08: 0000000000000010 R09: 0000000000000000
[    1.728908] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
[    1.729323] R13: ffff888106fe12c0 R14: ffff8881059c4a00 R15: ffff888106e57000
[    1.729744] FS:  0000000038e54380(0000) GS:ffff88813bc00000(0000)
knlGS:0000000000000000
[    1.730236] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    1.730597] CR2: 00000000004bf480 CR3: 00000001009ec001 CR4: 0000000000772ef0
[    1.730988] PKRU: 55555554
```


2026年1月15日(木) 20:07 Eric Dumazet <edumazet@google.com>:
>
> On Thu, Jan 15, 2026 at 11:33 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Hello, Eric and other maintainers,
> >
> > I hope you’re doing well. I’m following up on our email, sent during
> > the holiday season, in case it got buried.
> >
> > When you have a moment, could you please let us know if you had a
> > chance to review it?
> >
> > Thank you in advance, and I look forward to your response.
> >
>
> I think it would be nice to provide an actual stack trace of the bug,
> on a recent kernel tree.
>
> We had recent patches dealing with dev->hard_header_len changes.

Re: [PATCH net] bonding: Fix header_ops type confusion

[Eric Dumazet]

On Mon, Jan 19, 2026 at 6:36 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
>
> Thanks for your quick response.
>
> The following information is based on Linux kernel version 6.12.65,
> the latest release in the 6.12 tree.
> The kernel config is identical to that of the kernelCTF instance
> (available at: https://storage.googleapis.com/kernelctf-build/releases/lts-6.12.65/.config)
>
>
> This type confusion occurs in several locations, including,
> for example, `ipgre_header` (`header_ops->create`),
> where the private data of the network device is incorrectly cast as
> `struct ip_tunnel *`.
>
> ```
> static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
>       unsigned short type,
>       const void *daddr, const void *saddr, unsigned int len)
> {
>   struct ip_tunnel *t = netdev_priv(dev);
>   struct iphdr *iph;
>   struct gre_base_hdr *greh;
> ...
> ```
>
> When a bond interface is given to this function,
> it should not reference the private data as `struct ip_tunnel *`,
> because the bond interface uses the private data as `struct bonding *`.
> (quickly confirmed by seeing drivers/net/bonding/bond_netlink.c:909)
>
> ```
> struct rtnl_link_ops bond_link_ops __read_mostly = {
>     .kind            = "bond",
>     .priv_size        = sizeof(struct bonding),
> ...
> ```
>
> The stack trace below is the backtrace of all stack frame during a
> call to `ipgre_header`.
>
> ```
> ipgre_header at net/ipv4/ip_gre.c:890
> dev_hard_header at ./include/linux/netdevice.h:3156
> packet_snd at net/packet/af_packet.c:3082
> packet_sendmsg at net/packet/af_packet.c:3162
> sock_sendmsg_nosec at net/socket.c:729
> __sock_sendmsg at net/socket.c:744
> __sys_sendto at net/socket.c:2213
> __do_sys_sendto at net/socket.c:2225
> __se_sys_sendto at net/socket.c:2221
> __x64_sys_sendto at net/socket.c:2221
> do_syscall_x64 at arch/x86/entry/common.c:47
> do_syscall_64 at arch/x86/entry/common.c:78
> entry_SYSCALL_64 at arch/x86/entry/entry_64.S:121
> ```
>
> This causes memory corruption during subsequent operations.
>
> The following stack trace shows a General Protection Fault triggered
> when sending a packet
> to a bonding interface that has an IPv4 GRE interface as a slave.
>
> ```
> [    1.712329] Oops: general protection fault, probably for
> non-canonical address 0xdead0000cafebabe: 0000 [#1] SMP NOPTI
> [    1.712972] CPU: 0 UID: 1000 PID: 205 Comm: exp Not tainted 6.12.65 #1
> [    1.713344] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
> BIOS Arch Linux 1.17.0-2-2 04/01/2014
> [    1.713890] RIP: 0010:skb_release_data+0x8a/0x1c0
> [    1.714162] Code: c0 00 00 00 49 03 86 c8 00 00 00 0f b6 10 f6 c2
> 01 74 48 48 8b 70 28 48 85 f6 74 3f 41 0f b6 5d 00 83 e3 10 40 f6 c6
> 01 75 24 <48> 8b 06 ba 01 00 00 00 4c 89 f7 48 8b 00 ff d0 0f 1f 00 41
> 8b6
> [    1.715276] RSP: 0018:ffffc900007cfcc0 EFLAGS: 00010246
> [    1.715583] RAX: ffff888106fe12c0 RBX: 0000000000000010 RCX: 0000000000000000
> [    1.716036] RDX: 0000000000000017 RSI: dead0000cafebabe RDI: ffff8881059c4a00
> [    1.716504] RBP: ffffc900007cfe10 R08: 0000000000000010 R09: 0000000000000000
> [    1.716955] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
> [    1.717429] R13: ffff888106fe12c0 R14: ffff8881059c4a00 R15: ffff888106e57000
> [    1.717866] FS:  0000000038e54380(0000) GS:ffff88813bc00000(0000)
> knlGS:0000000000000000
> [    1.718350] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [    1.718703] CR2: 00000000004bf480 CR3: 00000001009ec001 CR4: 0000000000772ef0
> [    1.719109] PKRU: 55555554
> [    1.719297] Call Trace:
> [    1.719461]  <TASK>
> [    1.719611]  sk_skb_reason_drop+0x58/0x120
> [    1.719891]  packet_sendmsg+0xbcb/0x18f0
> [    1.720166]  ? pcpu_alloc_area+0x186/0x260
> [    1.720421]  __sys_sendto+0x1e2/0x1f0
> [    1.720691]  __x64_sys_sendto+0x24/0x30
> [    1.720948]  do_syscall_64+0x58/0x120
> [    1.721174]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
> [    1.721509] RIP: 0033:0x42860d
> [    1.721713] Code: c3 ff ff ff ff 64 89 02 eb b9 0f 1f 00 f3 0f 1e
> fa 80 3d 5d 4a 09 00 00 41 89 ca 74 20 45 31 c9 45 31 c0 b8 2c 00 00
> 00 0f 05 <48> 3d 00 f0 ff ff 77 6b c3 66 2e 0f 1f 84 00 00 00 00 00 55
> 489
> [    1.722837] RSP: 002b:00007fff597e95e8 EFLAGS: 00000246 ORIG_RAX:
> 000000000000002c
> [    1.723315] RAX: ffffffffffffffda RBX: 00000000000003e8 RCX: 000000000042860d
> [    1.723721] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000310
> [    1.724103] RBP: 00007fff597e9880 R08: 0000000000000000 R09: 0000000000000000
> [    1.724565] R10: 0000000000000000 R11: 0000000000000246 R12: 00007fff597e99f8
> [    1.725010] R13: 00007fff597e9a08 R14: 00000000004b7828 R15: 0000000000000001
> [    1.725441]  </TASK>
> [    1.725594] Modules linked in:
> [    1.725790] ---[ end trace 0000000000000000 ]---
> [    1.726057] RIP: 0010:skb_release_data+0x8a/0x1c0
> [    1.726339] Code: c0 00 00 00 49 03 86 c8 00 00 00 0f b6 10 f6 c2
> 01 74 48 48 8b 70 28 48 85 f6 74 3f 41 0f b6 5d 00 83 e3 10 40 f6 c6
> 01 75 24 <48> 8b 06 ba 01 00 00 00 4c 89 f7 48 8b 00 ff d0 0f 1f 00 41
> 8b6
> [    1.727285] RSP: 0018:ffffc900007cfcc0 EFLAGS: 00010246
> [    1.727623] RAX: ffff888106fe12c0 RBX: 0000000000000010 RCX: 0000000000000000
> [    1.728052] RDX: 0000000000000017 RSI: dead0000cafebabe RDI: ffff8881059c4a00
> [    1.728467] RBP: ffffc900007cfe10 R08: 0000000000000010 R09: 0000000000000000
> [    1.728908] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
> [    1.729323] R13: ffff888106fe12c0 R14: ffff8881059c4a00 R15: ffff888106e57000
> [    1.729744] FS:  0000000038e54380(0000) GS:ffff88813bc00000(0000)
> knlGS:0000000000000000
> [    1.730236] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [    1.730597] CR2: 00000000004bf480 CR3: 00000001009ec001 CR4: 0000000000772ef0
> [    1.730988] PKRU: 55555554
> ```
>

OK thanks.

I will repeat my original feedback : I do not see any barriers in the
patch you sent.

Assuming bond_setup_by_slave() can be called multiple times during one
master lifetime, I do not think your patch is enough.

Also, please clarify what happens with stacks of two or more bonding devices ?

Re: [PATCH net] bonding: Fix header_ops type confusion

[戸田晃太]

Here is the patch with the barriers added, based on v6.12.67.

However, as Yuki said, we are wondering if this would be considered an
acceptable change
from the perspective of the maintainers (or in terms of Linux kernel
culture). This is because
the patch adds `READ_ONCE` to several locations outside of bonding subsystem.
Please let me know if you have any concerns regarding this point.

>  Also, please clarify what happens with stacks of two or more bonding devices ?

To clarify, currently the `header_ops` of the bottom-most
interface are used regardless of the number of bonding layers.
This patch changes it so that `&bond->bond_header_ops` is used
as the bond device's `header_ops`, regardless of the stack depth.

```
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index f17a170d1..5ecc64e38 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1616,14 +1616,70 @@ static void bond_compute_features(struct bonding *bond)
     netdev_change_features(bond_dev);
 }

+static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
+        unsigned short type, const void *daddr,
+        const void *saddr, unsigned int len)
+{
+    struct bonding *bond = netdev_priv(dev);
+    struct net_device *slave_dev;
+
+    slave_dev = bond->header_slave_dev;
+
+    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
+}
+
+static void bond_header_cache_update(struct hh_cache *hh, const
+struct net_device *dev,
+        const unsigned char *haddr)
+{
+    const struct bonding *bond = netdev_priv(dev);
+    void (*cache_update)(struct hh_cache *hh,
+        const struct net_device *dev,
+        const unsigned char *haddr);
+    struct net_device *slave_dev;
+
+    slave_dev = bond->header_slave_dev;
+
+    if (!slave_dev->header_ops || !(cache_update =
READ_ONCE(slave_dev->header_ops->cache_update)))
+        return;
+
+    cache_update(hh, slave_dev, haddr);
+}
+
+
 static void bond_setup_by_slave(struct net_device *bond_dev,
                 struct net_device *slave_dev)
 {
+    struct bonding *bond = netdev_priv(bond_dev);
     bool was_up = !!(bond_dev->flags & IFF_UP);

     dev_close(bond_dev);

-    bond_dev->header_ops        = slave_dev->header_ops;
+    /* Some functions are given dev as an argument
+    * while others not. When dev is not given, we cannot
+    * find out what is the slave device through struct bonding
+    * (the private data of bond_dev). Therefore, we need a raw
+    * header_ops variable instead of its pointer to const header_ops
+    * and assign slave's functions directly.
+    * For the other case, we set the wrapper functions that pass
+    * slave_dev to the wrapped functions.
+    */
+    bond->bond_header_ops.create = bond_hard_header;
+    bond->bond_header_ops.cache_update = bond_header_cache_update;
+    if (slave_dev->header_ops) {
+        WRITE_ONCE(bond->bond_header_ops.parse, slave_dev->header_ops->parse);
+        WRITE_ONCE(bond->bond_header_ops.cache, slave_dev->header_ops->cache);
+        WRITE_ONCE(bond->bond_header_ops.validate,
slave_dev->header_ops->validate);
+        WRITE_ONCE(bond->bond_header_ops.parse_protocol,
slave_dev->header_ops->parse_protocol);
+    } else {
+        WRITE_ONCE(bond->bond_header_ops.parse, NULL);
+        WRITE_ONCE(bond->bond_header_ops.cache, NULL);
+        WRITE_ONCE(bond->bond_header_ops.validate, NULL);
+        WRITE_ONCE(bond->bond_header_ops.parse_protocol, NULL);
+    }
+
+    bond->header_slave_dev      = slave_dev;
+    bond_dev->header_ops        = &bond->bond_header_ops;

     bond_dev->type            = slave_dev->type;
     bond_dev->hard_header_len   = slave_dev->hard_header_len;
@@ -2682,6 +2738,14 @@ static int bond_release_and_destroy(struct
net_device *bond_dev,
     struct bonding *bond = netdev_priv(bond_dev);
     int ret;

+    /* If slave_dev is the earliest registered one, we must clear
+    * the variables related to header_ops to avoid dangling pointer.
+    */
+    if (bond->header_slave_dev == slave_dev) {
+        bond->header_slave_dev = NULL;
+        bond_dev->header_ops = NULL;
+    }
+
     ret = __bond_release_one(bond_dev, slave_dev, false, true);
     if (ret == 0 && !bond_has_slaves(bond) &&
         bond_dev->reg_state != NETREG_UNREGISTERING) {
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 77a99c8ab..0d2c1c852 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3150,35 +3150,41 @@ static inline int dev_hard_header(struct
sk_buff *skb, struct net_device *dev,
                   const void *daddr, const void *saddr,
                   unsigned int len)
 {
-    if (!dev->header_ops || !dev->header_ops->create)
+    int (*create)(struct sk_buff *skb, struct net_device *dev,
+        unsigned short type, const void *daddr,
+        const void *saddr, unsigned int len);
+    if (!dev->header_ops || !(create = READ_ONCE(dev->header_ops->create)))
         return 0;

-    return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
+    return create(skb, dev, type, daddr, saddr, len);
 }

 static inline int dev_parse_header(const struct sk_buff *skb,
                    unsigned char *haddr)
 {
+    int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
     const struct net_device *dev = skb->dev;

-    if (!dev->header_ops || !dev->header_ops->parse)
+    if (!dev->header_ops || !(parse = READ_ONCE(dev->header_ops->parse)))
         return 0;
-    return dev->header_ops->parse(skb, haddr);
+    return parse(skb, haddr);
 }

 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
 {
+    __be16    (*parse_protocol)(const struct sk_buff *skb);
     const struct net_device *dev = skb->dev;

-    if (!dev->header_ops || !dev->header_ops->parse_protocol)
+    if (!dev->header_ops || !(parse_protocol =
READ_ONCE(dev->header_ops->parse_protocol)))
         return 0;
-    return dev->header_ops->parse_protocol(skb);
+    return parse_protocol(skb);
 }

 /* ll_header must have at least hard_header_len allocated */
 static inline bool dev_validate_header(const struct net_device *dev,
                        char *ll_header, int len)
 {
+    bool    (*validate)(const char *ll_header, unsigned int len);
     if (likely(len >= dev->hard_header_len))
         return true;
     if (len < dev->min_header_len)
@@ -3189,15 +3195,15 @@ static inline bool dev_validate_header(const
struct net_device *dev,
         return true;
     }

-    if (dev->header_ops && dev->header_ops->validate)
-        return dev->header_ops->validate(ll_header, len);
+    if (dev->header_ops && (validate = READ_ONCE(dev->header_ops->validate)))
+        return validate(ll_header, len);

     return false;
 }

 static inline bool dev_has_header(const struct net_device *dev)
 {
-    return dev->header_ops && dev->header_ops->create;
+    return dev->header_ops && READ_ONCE(dev->header_ops->create);
 }

 /*
diff --git a/include/net/bonding.h b/include/net/bonding.h
index 95f67b308..c37800e17 100644
--- a/include/net/bonding.h
+++ b/include/net/bonding.h
@@ -215,6 +215,11 @@ struct bond_ipsec {
  */
 struct bonding {
     struct   net_device *dev; /* first - useful for panic debug */
+    struct   net_device *header_slave_dev;  /* slave net_device for
header_ops */
+    /* maintained as a non-const variable
+    * because bond's header_ops should change depending on slaves.
+    */
+    struct   header_ops bond_header_ops;
     struct   slave __rcu *curr_active_slave;
     struct   slave __rcu *current_arp_slave;
     struct   slave __rcu *primary_slave;
diff --git a/include/net/cfg802154.h b/include/net/cfg802154.h
index 76d2cd2e2..dec638763 100644
--- a/include/net/cfg802154.h
+++ b/include/net/cfg802154.h
@@ -522,7 +522,7 @@ wpan_dev_hard_header(struct sk_buff *skb, struct
net_device *dev,
 {
     struct wpan_dev *wpan_dev = dev->ieee802154_ptr;

-    return wpan_dev->header_ops->create(skb, dev, daddr, saddr, len);
+    return READ_ONCE(wpan_dev->header_ops->create)(skb, dev, daddr,
saddr, len);
 }
 #endif

diff --git a/net/core/neighbour.c b/net/core/neighbour.c
index 96786016d..ff948e35e 100644
--- a/net/core/neighbour.c
+++ b/net/core/neighbour.c
@@ -1270,7 +1270,7 @@ static void neigh_update_hhs(struct neighbour *neigh)
         = NULL;

     if (neigh->dev->header_ops)
-        update = neigh->dev->header_ops->cache_update;
+        update = READ_ONCE(neigh->dev->header_ops->cache_update);

     if (update) {
         hh = &neigh->hh;
@@ -1540,7 +1540,7 @@ static void neigh_hh_init(struct neighbour *n)
      * hh_cache entry.
      */
     if (!hh->hh_len)
-        dev->header_ops->cache(n, hh, prot);
+        READ_ONCE(dev->header_ops->cache)(n, hh, prot);

     write_unlock_bh(&n->lock);
 }
@@ -1556,7 +1556,7 @@ int neigh_resolve_output(struct neighbour
*neigh, struct sk_buff *skb)
         struct net_device *dev = neigh->dev;
         unsigned int seq;

-        if (dev->header_ops->cache && !READ_ONCE(neigh->hh.hh_len))
+        if (READ_ONCE(dev->header_ops->cache) && !READ_ONCE(neigh->hh.hh_len))
             neigh_hh_init(neigh);

         do {
diff --git a/net/ipv4/arp.c b/net/ipv4/arp.c
index 7822b2144..421bea6eb 100644
--- a/net/ipv4/arp.c
+++ b/net/ipv4/arp.c
@@ -278,7 +278,7 @@ static int arp_constructor(struct neighbour *neigh)
             memcpy(neigh->ha, dev->broadcast, dev->addr_len);
         }

-        if (dev->header_ops->cache)
+        if (READ_ONCE(dev->header_ops->cache))
             neigh->ops = &arp_hh_ops;
         else
             neigh->ops = &arp_generic_ops;
diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c
index d961e6c2d..d81f509ec 100644
--- a/net/ipv6/ndisc.c
+++ b/net/ipv6/ndisc.c
@@ -361,7 +361,7 @@ static int ndisc_constructor(struct neighbour *neigh)
             neigh->nud_state = NUD_NOARP;
             memcpy(neigh->ha, dev->broadcast, dev->addr_len);
         }
-        if (dev->header_ops->cache)
+        if (READ_ONCE(dev->header_ops->cache))
             neigh->ops = &ndisc_hh_ops;
         else
             neigh->ops = &ndisc_generic_ops;
```

2026年1月19日(月) 18:30 Eric Dumazet <edumazet@google.com>:
>
> On Mon, Jan 19, 2026 at 6:36 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Thanks for your quick response.
> >
> > The following information is based on Linux kernel version 6.12.65,
> > the latest release in the 6.12 tree.
> > The kernel config is identical to that of the kernelCTF instance
> > (available at: https://storage.googleapis.com/kernelctf-build/releases/lts-6.12.65/.config)
> >
> >
> > This type confusion occurs in several locations, including,
> > for example, `ipgre_header` (`header_ops->create`),
> > where the private data of the network device is incorrectly cast as
> > `struct ip_tunnel *`.
> >
> > ```
> > static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
> >       unsigned short type,
> >       const void *daddr, const void *saddr, unsigned int len)
> > {
> >   struct ip_tunnel *t = netdev_priv(dev);
> >   struct iphdr *iph;
> >   struct gre_base_hdr *greh;
> > ...
> > ```
> >
> > When a bond interface is given to this function,
> > it should not reference the private data as `struct ip_tunnel *`,
> > because the bond interface uses the private data as `struct bonding *`.
> > (quickly confirmed by seeing drivers/net/bonding/bond_netlink.c:909)
> >
> > ```
> > struct rtnl_link_ops bond_link_ops __read_mostly = {
> >     .kind            = "bond",
> >     .priv_size        = sizeof(struct bonding),
> > ...
> > ```
> >
> > The stack trace below is the backtrace of all stack frame during a
> > call to `ipgre_header`.
> >
> > ```
> > ipgre_header at net/ipv4/ip_gre.c:890
> > dev_hard_header at ./include/linux/netdevice.h:3156
> > packet_snd at net/packet/af_packet.c:3082
> > packet_sendmsg at net/packet/af_packet.c:3162
> > sock_sendmsg_nosec at net/socket.c:729
> > __sock_sendmsg at net/socket.c:744
> > __sys_sendto at net/socket.c:2213
> > __do_sys_sendto at net/socket.c:2225
> > __se_sys_sendto at net/socket.c:2221
> > __x64_sys_sendto at net/socket.c:2221
> > do_syscall_x64 at arch/x86/entry/common.c:47
> > do_syscall_64 at arch/x86/entry/common.c:78
> > entry_SYSCALL_64 at arch/x86/entry/entry_64.S:121
> > ```
> >
> > This causes memory corruption during subsequent operations.
> >
> > The following stack trace shows a General Protection Fault triggered
> > when sending a packet
> > to a bonding interface that has an IPv4 GRE interface as a slave.
> >
> > ```
> > [    1.712329] Oops: general protection fault, probably for
> > non-canonical address 0xdead0000cafebabe: 0000 [#1] SMP NOPTI
> > [    1.712972] CPU: 0 UID: 1000 PID: 205 Comm: exp Not tainted 6.12.65 #1
> > [    1.713344] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
> > BIOS Arch Linux 1.17.0-2-2 04/01/2014
> > [    1.713890] RIP: 0010:skb_release_data+0x8a/0x1c0
> > [    1.714162] Code: c0 00 00 00 49 03 86 c8 00 00 00 0f b6 10 f6 c2
> > 01 74 48 48 8b 70 28 48 85 f6 74 3f 41 0f b6 5d 00 83 e3 10 40 f6 c6
> > 01 75 24 <48> 8b 06 ba 01 00 00 00 4c 89 f7 48 8b 00 ff d0 0f 1f 00 41
> > 8b6
> > [    1.715276] RSP: 0018:ffffc900007cfcc0 EFLAGS: 00010246
> > [    1.715583] RAX: ffff888106fe12c0 RBX: 0000000000000010 RCX: 0000000000000000
> > [    1.716036] RDX: 0000000000000017 RSI: dead0000cafebabe RDI: ffff8881059c4a00
> > [    1.716504] RBP: ffffc900007cfe10 R08: 0000000000000010 R09: 0000000000000000
> > [    1.716955] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
> > [    1.717429] R13: ffff888106fe12c0 R14: ffff8881059c4a00 R15: ffff888106e57000
> > [    1.717866] FS:  0000000038e54380(0000) GS:ffff88813bc00000(0000)
> > knlGS:0000000000000000
> > [    1.718350] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > [    1.718703] CR2: 00000000004bf480 CR3: 00000001009ec001 CR4: 0000000000772ef0
> > [    1.719109] PKRU: 55555554
> > [    1.719297] Call Trace:
> > [    1.719461]  <TASK>
> > [    1.719611]  sk_skb_reason_drop+0x58/0x120
> > [    1.719891]  packet_sendmsg+0xbcb/0x18f0
> > [    1.720166]  ? pcpu_alloc_area+0x186/0x260
> > [    1.720421]  __sys_sendto+0x1e2/0x1f0
> > [    1.720691]  __x64_sys_sendto+0x24/0x30
> > [    1.720948]  do_syscall_64+0x58/0x120
> > [    1.721174]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
> > [    1.721509] RIP: 0033:0x42860d
> > [    1.721713] Code: c3 ff ff ff ff 64 89 02 eb b9 0f 1f 00 f3 0f 1e
> > fa 80 3d 5d 4a 09 00 00 41 89 ca 74 20 45 31 c9 45 31 c0 b8 2c 00 00
> > 00 0f 05 <48> 3d 00 f0 ff ff 77 6b c3 66 2e 0f 1f 84 00 00 00 00 00 55
> > 489
> > [    1.722837] RSP: 002b:00007fff597e95e8 EFLAGS: 00000246 ORIG_RAX:
> > 000000000000002c
> > [    1.723315] RAX: ffffffffffffffda RBX: 00000000000003e8 RCX: 000000000042860d
> > [    1.723721] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000310
> > [    1.724103] RBP: 00007fff597e9880 R08: 0000000000000000 R09: 0000000000000000
> > [    1.724565] R10: 0000000000000000 R11: 0000000000000246 R12: 00007fff597e99f8
> > [    1.725010] R13: 00007fff597e9a08 R14: 00000000004b7828 R15: 0000000000000001
> > [    1.725441]  </TASK>
> > [    1.725594] Modules linked in:
> > [    1.725790] ---[ end trace 0000000000000000 ]---
> > [    1.726057] RIP: 0010:skb_release_data+0x8a/0x1c0
> > [    1.726339] Code: c0 00 00 00 49 03 86 c8 00 00 00 0f b6 10 f6 c2
> > 01 74 48 48 8b 70 28 48 85 f6 74 3f 41 0f b6 5d 00 83 e3 10 40 f6 c6
> > 01 75 24 <48> 8b 06 ba 01 00 00 00 4c 89 f7 48 8b 00 ff d0 0f 1f 00 41
> > 8b6
> > [    1.727285] RSP: 0018:ffffc900007cfcc0 EFLAGS: 00010246
> > [    1.727623] RAX: ffff888106fe12c0 RBX: 0000000000000010 RCX: 0000000000000000
> > [    1.728052] RDX: 0000000000000017 RSI: dead0000cafebabe RDI: ffff8881059c4a00
> > [    1.728467] RBP: ffffc900007cfe10 R08: 0000000000000010 R09: 0000000000000000
> > [    1.728908] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
> > [    1.729323] R13: ffff888106fe12c0 R14: ffff8881059c4a00 R15: ffff888106e57000
> > [    1.729744] FS:  0000000038e54380(0000) GS:ffff88813bc00000(0000)
> > knlGS:0000000000000000
> > [    1.730236] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > [    1.730597] CR2: 00000000004bf480 CR3: 00000001009ec001 CR4: 0000000000772ef0
> > [    1.730988] PKRU: 55555554
> > ```
> >
>
> OK thanks.
>
> I will repeat my original feedback : I do not see any barriers in the
> patch you sent.
>
> Assuming bond_setup_by_slave() can be called multiple times during one
> master lifetime, I do not think your patch is enough.
>
> Also, please clarify what happens with stacks of two or more bonding devices ?

Re: [PATCH net] bonding: Fix header_ops type confusion

[Eric Dumazet]

On Wed, Jan 28, 2026 at 11:46 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
>
> Here is the patch with the barriers added, based on v6.12.67.
>
> However, as Yuki said, we are wondering if this would be considered an
> acceptable change
> from the perspective of the maintainers (or in terms of Linux kernel
> culture). This is because
> the patch adds `READ_ONCE` to several locations outside of bonding subsystem.
> Please let me know if you have any concerns regarding this point.
>
> >  Also, please clarify what happens with stacks of two or more bonding devices ?
>
> To clarify, currently the `header_ops` of the bottom-most
> interface are used regardless of the number of bonding layers.
> This patch changes it so that `&bond->bond_header_ops` is used
> as the bond device's `header_ops`, regardless of the stack depth.

Could you try to cook a patch series perhaps ?

The READ_ONCE()/WRITE_ONCE() on dev->header_ops->cache could be done separately.

Thanks.

[PATCH net 0/2] net: bonding: fix type-confusion in bonding header_ops

[戸田晃太]

In bond_setup_by_slave(), the slave’s header_ops are unconditionally
copied into the bonding device. As a result, the bonding device may invoke
the slave-specific header operations on itself, causing
netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
as the slave's private-data type.

This type-confusion bug can lead to out-of-bounds writes into the skb,
resulting in memory corruption.

Patch 1 stores the slave's header_ops in struct bonding and sets
wrapper callbacks in bond_In bond_setup_by_slave(), the slave’s
header_ops are unconditionally
copied into the bonding device. As a result, the bonding device may invoke
the slave-specific header operations on itself, causing
netdev_priv(bond_dev) (a struct bonding) to be incorrectly interpreted
as the slave's private-data type.

This type-confusion bug can lead to out-of-bounds writes into the skb,
resulting in memory corruption.

Patch 1 stores the slave's header_ops in struct bonding and sets
wrapper callbacks in bond_dev->header_ops.

Patch 2 uses READ_ONCE when loading header_ops callbacks
to avoid races with concurrent updates.

Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Co-developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Reviewed-by: Paolo Abeni <pabeni@redhat.com>
Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
--

Kota Toda (2):
  net: bonding: fix type-confusion in bonding header_ops
  net: add READ_ONCE for header_ops callbacks

 drivers/net/bonding/bond_main.c | 66 ++++++++++++++++++++++++++++++++-
 include/linux/netdevice.h       | 24 +++++++-----
 include/net/bonding.h           |  5 +++
 include/net/cfg802154.h         |  2 +-
 net/core/neighbour.c            |  6 +--
 net/ipv4/arp.c                  |  2 +-
 net/ipv6/ndisc.c                |  2 +-
 7 files changed, 91 insertions(+), 16 deletions(-)

--
2.53.0


2026年2月3日(火) 2:11 Eric Dumazet <edumazet@google.com>:

>
> On Wed, Jan 28, 2026 at 11:46 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Here is the patch with the barriers added, based on v6.12.67.
> >
> > However, as Yuki said, we are wondering if this would be considered an
> > acceptable change
> > from the perspective of the maintainers (or in terms of Linux kernel
> > culture). This is because
> > the patch adds `READ_ONCE` to several locations outside of bonding subsystem.
> > Please let me know if you have any concerns regarding this point.
> >
> > >  Also, please clarify what happens with stacks of two or more bonding devices ?
> >
> > To clarify, currently the `header_ops` of the bottom-most
> > interface are used regardless of the number of bonding layers.
> > This patch changes it so that `&bond->bond_header_ops` is used
> > as the bond device's `header_ops`, regardless of the stack depth.
>
> Could you try to cook a patch series perhaps ?
>
> The READ_ONCE()/WRITE_ONCE() on dev->header_ops->cache could be done separately.
>
> Thanks.dev->header_ops.

Patch 2 uses READ_ONCE when loading header_ops callbacks
to avoid races with concurrent updates.

Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Co-developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Reviewed-by: Paolo Abeni <pabeni@redhat.com>
Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
--

Kota Toda (2):
  net: bonding: fix type-confusion in bonding header_ops
  net: add READ_ONCE for header_ops callbacks

 drivers/net/bonding/bond_main.c | 66 ++++++++++++++++++++++++++++++++-
 include/linux/netdevice.h       | 24 +++++++-----
 include/net/bonding.h           |  5 +++
 include/net/cfg802154.h         |  2 +-
 net/core/neighbour.c            |  6 +--
 net/ipv4/arp.c                  |  2 +-
 net/ipv6/ndisc.c                |  2 +-
 7 files changed, 91 insertions(+), 16 deletions(-)

--
2.53.0


2026年2月3日(火) 2:11 Eric Dumazet <edumazet@google.com>:
>
> On Wed, Jan 28, 2026 at 11:46 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Here is the patch with the barriers added, based on v6.12.67.
> >
> > However, as Yuki said, we are wondering if this would be considered an
> > acceptable change
> > from the perspective of the maintainers (or in terms of Linux kernel
> > culture). This is because
> > the patch adds `READ_ONCE` to several locations outside of bonding subsystem.
> > Please let me know if you have any concerns regarding this point.
> >
> > >  Also, please clarify what happens with stacks of two or more bonding devices ?
> >
> > To clarify, currently the `header_ops` of the bottom-most
> > interface are used regardless of the number of bonding layers.
> > This patch changes it so that `&bond->bond_header_ops` is used
> > as the bond device's `header_ops`, regardless of the stack depth.
>
> Could you try to cook a patch series perhaps ?
>
> The READ_ONCE()/WRITE_ONCE() on dev->header_ops->cache could be done separately.
>
> Thanks.

[PATCH net 1/2] net: bonding: fix type-confusion in bonding header_ops

[戸田晃太]

Add two members to struct bonding, bond_header_ops and
header_slave_dev, to avoid type-confusion while keeping track of the
slave's header_ops.

Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Co-developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Reviewed-by: Paolo Abeni <pabeni@redhat.com>
Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
---
 drivers/net/bonding/bond_main.c | 66 ++++++++++++++++++++++++++++++++-
 include/net/bonding.h           |  5 +++
 2 files changed, 70 insertions(+), 1 deletion(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index f17a170d1..5ecc64e38 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1616,14 +1616,70 @@ static void bond_compute_features(struct bonding *bond)
     netdev_change_features(bond_dev);
 }

+static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
+        unsigned short type, const void *daddr,
+        const void *saddr, unsigned int len)
+{
+    struct bonding *bond = netdev_priv(dev);
+    struct net_device *slave_dev;
+
+    slave_dev = bond->header_slave_dev;
+
+    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
+}
+
+static void bond_header_cache_update(struct hh_cache *hh, const
+struct net_device *dev,
+        const unsigned char *haddr)
+{
+    const struct bonding *bond = netdev_priv(dev);
+    void (*cache_update)(struct hh_cache *hh,
+        const struct net_device *dev,
+        const unsigned char *haddr);
+    struct net_device *slave_dev;
+
+    slave_dev = bond->header_slave_dev;
+
+    if (!slave_dev->header_ops || !(cache_update =
READ_ONCE(slave_dev->header_ops->cache_update)))
+        return;
+
+    cache_update(hh, slave_dev, haddr);
+}
+
+
 static void bond_setup_by_slave(struct net_device *bond_dev,
                 struct net_device *slave_dev)
 {
+    struct bonding *bond = netdev_priv(bond_dev);
     bool was_up = !!(bond_dev->flags & IFF_UP);

     dev_close(bond_dev);

-    bond_dev->header_ops        = slave_dev->header_ops;
+    /* Some functions are given dev as an argument
+    * while others not. When dev is not given, we cannot
+    * find out what is the slave device through struct bonding
+    * (the private data of bond_dev). Therefore, we need a raw
+    * header_ops variable instead of its pointer to const header_ops
+    * and assign slave's functions directly.
+    * For the other case, we set the wrapper functions that pass
+    * slave_dev to the wrapped functions.
+    */
+    bond->bond_header_ops.create = bond_hard_header;
+    bond->bond_header_ops.cache_update = bond_header_cache_update;
+    if (slave_dev->header_ops) {
+        WRITE_ONCE(bond->bond_header_ops.parse, slave_dev->header_ops->parse);
+        WRITE_ONCE(bond->bond_header_ops.cache, slave_dev->header_ops->cache);
+        WRITE_ONCE(bond->bond_header_ops.validate,
slave_dev->header_ops->validate);
+        WRITE_ONCE(bond->bond_header_ops.parse_protocol,
slave_dev->header_ops->parse_protocol);
+    } else {
+        WRITE_ONCE(bond->bond_header_ops.parse, NULL);
+        WRITE_ONCE(bond->bond_header_ops.cache, NULL);
+        WRITE_ONCE(bond->bond_header_ops.validate, NULL);
+        WRITE_ONCE(bond->bond_header_ops.parse_protocol, NULL);
+    }
+
+    bond->header_slave_dev      = slave_dev;
+    bond_dev->header_ops        = &bond->bond_header_ops;

     bond_dev->type            = slave_dev->type;
     bond_dev->hard_header_len   = slave_dev->hard_header_len;
@@ -2682,6 +2738,14 @@ static int bond_release_and_destroy(struct
net_device *bond_dev,
     struct bonding *bond = netdev_priv(bond_dev);
     int ret;

+    /* If slave_dev is the earliest registered one, we must clear
+    * the variables related to header_ops to avoid dangling pointer.
+    */
+    if (bond->header_slave_dev == slave_dev) {
+        bond->header_slave_dev = NULL;
+        bond_dev->header_ops = NULL;
+    }
+
     ret = __bond_release_one(bond_dev, slave_dev, false, true);
     if (ret == 0 && !bond_has_slaves(bond) &&
         bond_dev->reg_state != NETREG_UNREGISTERING) {
diff --git a/include/net/bonding.h b/include/net/bonding.h
index 95f67b308..c37800e17 100644
--- a/include/net/bonding.h
+++ b/include/net/bonding.h
@@ -215,6 +215,11 @@ struct bond_ipsec {
  */
 struct bonding {
     struct   net_device *dev; /* first - useful for panic debug */
+    struct   net_device *header_slave_dev;  /* slave net_device for
header_ops */
+    /* maintained as a non-const variable
+    * because bond's header_ops should change depending on slaves.
+    */
+    struct   header_ops bond_header_ops;
     struct   slave __rcu *curr_active_slave;
     struct   slave __rcu *current_arp_slave;
     struct   slave __rcu *primary_slave;
--
2.53.0


2026年2月3日(火) 2:11 Eric Dumazet <edumazet@google.com>:
>
> On Wed, Jan 28, 2026 at 11:46 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Here is the patch with the barriers added, based on v6.12.67.
> >
> > However, as Yuki said, we are wondering if this would be considered an
> > acceptable change
> > from the perspective of the maintainers (or in terms of Linux kernel
> > culture). This is because
> > the patch adds `READ_ONCE` to several locations outside of bonding subsystem.
> > Please let me know if you have any concerns regarding this point.
> >
> > >  Also, please clarify what happens with stacks of two or more bonding devices ?
> >
> > To clarify, currently the `header_ops` of the bottom-most
> > interface are used regardless of the number of bonding layers.
> > This patch changes it so that `&bond->bond_header_ops` is used
> > as the bond device's `header_ops`, regardless of the stack depth.
>
> Could you try to cook a patch series perhaps ?
>
> The READ_ONCE()/WRITE_ONCE() on dev->header_ops->cache could be done separately.
>
> Thanks.

Re: [PATCH net 1/2] net: bonding: fix type-confusion in bonding header_ops

[Kuniyuki Iwashima]

Hi,

Looks like patches were copy-and-pasted to email client since
diff is corrupted (\t is replaced with \s, line is wrapped, etc).

You may want to check your email client config or simply use
git send-email like

 $ git send-email --annotate --cover-letter --thread --no-chain-reply-to \
   --subject-prefix "PATCH v3 net" \
   --to "NAME <EMAIL@ADDRESS>" \
   --cc "NAME <EMAIL@ADDRESS>" \
   --cc netdev@vger.kernel.org \
   --dry-run HEAD~2

Then, make sure to CC maintainers listed by ./scripts/get_maintainer.pl

There are a few more rules to follow, so please read these doc
for next submission.

  Documentation/process/submitting-patches.rst
  Documentation/process/maintainer-netdev.rst


From: 戸田晃太 <kota.toda@gmo-cybersecurity.com>
Date: Thu, 5 Feb 2026 19:57:00 +0900
> Add two members to struct bonding, bond_header_ops and
> header_slave_dev, to avoid type-confusion while keeping track of the
> slave's header_ops.
> 
> Fixes: 1284cd3a2b740 (bonding: two small fixes for IPoIB support)
> Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
> Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> Co-developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
> Reviewed-by: Paolo Abeni <pabeni@redhat.com>
> Reported-by: Kota Toda <kota.toda@gmo-cybersecurity.com>

nit: Reported-by is not needed when it's identical with the author.


> ---
>  drivers/net/bonding/bond_main.c | 66 ++++++++++++++++++++++++++++++++-
>  include/net/bonding.h           |  5 +++
>  2 files changed, 70 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> index f17a170d1..5ecc64e38 100644
> --- a/drivers/net/bonding/bond_main.c
> +++ b/drivers/net/bonding/bond_main.c
> @@ -1616,14 +1616,70 @@ static void bond_compute_features(struct bonding *bond)
>      netdev_change_features(bond_dev);
>  }
> 
> +static int bond_hard_header(struct sk_buff *skb, struct net_device *dev,
> +        unsigned short type, const void *daddr,
> +        const void *saddr, unsigned int len)
> +{
> +    struct bonding *bond = netdev_priv(dev);
> +    struct net_device *slave_dev;
> +
> +    slave_dev = bond->header_slave_dev;
> +
> +    return dev_hard_header(skb, slave_dev, type, daddr, saddr, len);
> +}
> +
> +static void bond_header_cache_update(struct hh_cache *hh, const
> +struct net_device *dev,
> +        const unsigned char *haddr)
> +{
> +    const struct bonding *bond = netdev_priv(dev);
> +    void (*cache_update)(struct hh_cache *hh,
> +        const struct net_device *dev,
> +        const unsigned char *haddr);
> +    struct net_device *slave_dev;
> +
> +    slave_dev = bond->header_slave_dev;
> +
> +    if (!slave_dev->header_ops

Is slave_dev always non-NULL ?


> || !(cache_update =
> READ_ONCE(slave_dev->header_ops->cache_update)))

READ_ONCE() seems necessary for header_ops.

and please run ./scripts/checkpatch.pl, which will
complain about the coding style setting var in if().


> +        return;
> +
> +    cache_update(hh, slave_dev, haddr);
> +}
> +
> +
>  static void bond_setup_by_slave(struct net_device *bond_dev,
>                  struct net_device *slave_dev)
>  {
> +    struct bonding *bond = netdev_priv(bond_dev);
>      bool was_up = !!(bond_dev->flags & IFF_UP);
> 
>      dev_close(bond_dev);
> 
> -    bond_dev->header_ops        = slave_dev->header_ops;
> +    /* Some functions are given dev as an argument
> +    * while others not. When dev is not given, we cannot
> +    * find out what is the slave device through struct bonding
> +    * (the private data of bond_dev). Therefore, we need a raw
> +    * header_ops variable instead of its pointer to const header_ops
> +    * and assign slave's functions directly.
> +    * For the other case, we set the wrapper functions that pass
> +    * slave_dev to the wrapped functions.
> +    */
> +    bond->bond_header_ops.create = bond_hard_header;
> +    bond->bond_header_ops.cache_update = bond_header_cache_update;
> +    if (slave_dev->header_ops) {
> +        WRITE_ONCE(bond->bond_header_ops.parse, slave_dev->header_ops->parse);

Are WRITE_ONCE()s needed ?

RCU readers will see NULL bond_dev->header_ops at this
point, no ?

If they see non-NULL header_ops, then it will not be changed
because only bond_release_and_destroy() clears it and
we wait one RCU grace period after issuing NETDEV_UNREGISTER.


> +        WRITE_ONCE(bond->bond_header_ops.cache, slave_dev->header_ops->cache);
> +        WRITE_ONCE(bond->bond_header_ops.validate,
> slave_dev->header_ops->validate);
> +        WRITE_ONCE(bond->bond_header_ops.parse_protocol,
> slave_dev->header_ops->parse_protocol);
> +    } else {
> +        WRITE_ONCE(bond->bond_header_ops.parse, NULL);
> +        WRITE_ONCE(bond->bond_header_ops.cache, NULL);
> +        WRITE_ONCE(bond->bond_header_ops.validate, NULL);
> +        WRITE_ONCE(bond->bond_header_ops.parse_protocol, NULL);
> +    }
> +
> +    bond->header_slave_dev      = slave_dev;
> +    bond_dev->header_ops        = &bond->bond_header_ops;

Rather WRITE_ONCE() seems necessary here.


> 
>      bond_dev->type            = slave_dev->type;
>      bond_dev->hard_header_len   = slave_dev->hard_header_len;
> @@ -2682,6 +2738,14 @@ static int bond_release_and_destroy(struct
> net_device *bond_dev,
>      struct bonding *bond = netdev_priv(bond_dev);
>      int ret;
> 
> +    /* If slave_dev is the earliest registered one, we must clear
> +    * the variables related to header_ops to avoid dangling pointer.
> +    */
> +    if (bond->header_slave_dev == slave_dev) {
> +        bond->header_slave_dev = NULL;
> +        bond_dev->header_ops = NULL;

Ditto.

Also, before clearing header_ops, there would be a small (race)
window where

  bond->header_slave_dev == NULL &&
  bond_dev->header_ops == &bond->bond_header_ops

is true and this would trigger null-deref in bond_header_cache_update().

This is Eric's point, I think.


> +    }
> +
>      ret = __bond_release_one(bond_dev, slave_dev, false, true);
>      if (ret == 0 && !bond_has_slaves(bond) &&
>          bond_dev->reg_state != NETREG_UNREGISTERING) {
> diff --git a/include/net/bonding.h b/include/net/bonding.h
> index 95f67b308..c37800e17 100644
> --- a/include/net/bonding.h
> +++ b/include/net/bonding.h
> @@ -215,6 +215,11 @@ struct bond_ipsec {
>   */
>  struct bonding {
>      struct   net_device *dev; /* first - useful for panic debug */
> +    struct   net_device *header_slave_dev;  /* slave net_device for
> header_ops */
> +    /* maintained as a non-const variable
> +    * because bond's header_ops should change depending on slaves.
> +    */
> +    struct   header_ops bond_header_ops;
>      struct   slave __rcu *curr_active_slave;
>      struct   slave __rcu *current_arp_slave;
>      struct   slave __rcu *primary_slave;
> --

[PATCH net 2/2] net: add READ_ONCE for header_ops callbacks

[戸田晃太]

Use READ_ONCE when loading header_ops callbacks to avoid races with
concurrent updates.

Signed-off-by: Kota Toda <kota.toda@gmo-cybersecurity.com>
Signed-off-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
Co-developed-by: Yuki Koike <yuki.koike@gmo-cybersecurity.com>
---
 include/linux/netdevice.h | 24 +++++++++++++++---------
 include/net/cfg802154.h   |  2 +-
 net/core/neighbour.c      |  6 +++---
 net/ipv4/arp.c            |  2 +-
 net/ipv6/ndisc.c          |  2 +-
 5 files changed, 21 insertions(+), 15 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 77a99c8ab..0d2c1c852 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3150,35 +3150,41 @@ static inline int dev_hard_header(struct
sk_buff *skb, struct net_device *dev,
                   const void *daddr, const void *saddr,
                   unsigned int len)
 {
-    if (!dev->header_ops || !dev->header_ops->create)
+    int (*create)(struct sk_buff *skb, struct net_device *dev,
+        unsigned short type, const void *daddr,
+        const void *saddr, unsigned int len);
+    if (!dev->header_ops || !(create = READ_ONCE(dev->header_ops->create)))
         return 0;

-    return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
+    return create(skb, dev, type, daddr, saddr, len);
 }

 static inline int dev_parse_header(const struct sk_buff *skb,
                    unsigned char *haddr)
 {
+    int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
     const struct net_device *dev = skb->dev;

-    if (!dev->header_ops || !dev->header_ops->parse)
+    if (!dev->header_ops || !(parse = READ_ONCE(dev->header_ops->parse)))
         return 0;
-    return dev->header_ops->parse(skb, haddr);
+    return parse(skb, haddr);
 }

 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
 {
+    __be16    (*parse_protocol)(const struct sk_buff *skb);
     const struct net_device *dev = skb->dev;

-    if (!dev->header_ops || !dev->header_ops->parse_protocol)
+    if (!dev->header_ops || !(parse_protocol =
READ_ONCE(dev->header_ops->parse_protocol)))
         return 0;
-    return dev->header_ops->parse_protocol(skb);
+    return parse_protocol(skb);
 }

 /* ll_header must have at least hard_header_len allocated */
 static inline bool dev_validate_header(const struct net_device *dev,
                        char *ll_header, int len)
 {
+    bool    (*validate)(const char *ll_header, unsigned int len);
     if (likely(len >= dev->hard_header_len))
         return true;
     if (len < dev->min_header_len)
@@ -3189,15 +3195,15 @@ static inline bool dev_validate_header(const
struct net_device *dev,
         return true;
     }

-    if (dev->header_ops && dev->header_ops->validate)
-        return dev->header_ops->validate(ll_header, len);
+    if (dev->header_ops && (validate = READ_ONCE(dev->header_ops->validate)))
+        return validate(ll_header, len);

     return false;
 }

 static inline bool dev_has_header(const struct net_device *dev)
 {
-    return dev->header_ops && dev->header_ops->create;
+    return dev->header_ops && READ_ONCE(dev->header_ops->create);
 }

 /*
diff --git a/include/net/cfg802154.h b/include/net/cfg802154.h
index 76d2cd2e2..dec638763 100644
--- a/include/net/cfg802154.h
+++ b/include/net/cfg802154.h
@@ -522,7 +522,7 @@ wpan_dev_hard_header(struct sk_buff *skb, struct
net_device *dev,
 {
     struct wpan_dev *wpan_dev = dev->ieee802154_ptr;

-    return wpan_dev->header_ops->create(skb, dev, daddr, saddr, len);
+    return READ_ONCE(wpan_dev->header_ops->create)(skb, dev, daddr,
saddr, len);
 }
 #endif

diff --git a/net/core/neighbour.c b/net/core/neighbour.c
index 96786016d..ff948e35e 100644
--- a/net/core/neighbour.c
+++ b/net/core/neighbour.c
@@ -1270,7 +1270,7 @@ static void neigh_update_hhs(struct neighbour *neigh)
         = NULL;

     if (neigh->dev->header_ops)
-        update = neigh->dev->header_ops->cache_update;
+        update = READ_ONCE(neigh->dev->header_ops->cache_update);

     if (update) {
         hh = &neigh->hh;
@@ -1540,7 +1540,7 @@ static void neigh_hh_init(struct neighbour *n)
      * hh_cache entry.
      */
     if (!hh->hh_len)
-        dev->header_ops->cache(n, hh, prot);
+        READ_ONCE(dev->header_ops->cache)(n, hh, prot);

     write_unlock_bh(&n->lock);
 }
@@ -1556,7 +1556,7 @@ int neigh_resolve_output(struct neighbour
*neigh, struct sk_buff *skb)
         struct net_device *dev = neigh->dev;
         unsigned int seq;

-        if (dev->header_ops->cache && !READ_ONCE(neigh->hh.hh_len))
+        if (READ_ONCE(dev->header_ops->cache) && !READ_ONCE(neigh->hh.hh_len))
             neigh_hh_init(neigh);

         do {
diff --git a/net/ipv4/arp.c b/net/ipv4/arp.c
index 7822b2144..421bea6eb 100644
--- a/net/ipv4/arp.c
+++ b/net/ipv4/arp.c
@@ -278,7 +278,7 @@ static int arp_constructor(struct neighbour *neigh)
             memcpy(neigh->ha, dev->broadcast, dev->addr_len);
         }

-        if (dev->header_ops->cache)
+        if (READ_ONCE(dev->header_ops->cache))
             neigh->ops = &arp_hh_ops;
         else
             neigh->ops = &arp_generic_ops;
diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c
index d961e6c2d..d81f509ec 100644
--- a/net/ipv6/ndisc.c
+++ b/net/ipv6/ndisc.c
@@ -361,7 +361,7 @@ static int ndisc_constructor(struct neighbour *neigh)
             neigh->nud_state = NUD_NOARP;
             memcpy(neigh->ha, dev->broadcast, dev->addr_len);
         }
-        if (dev->header_ops->cache)
+        if (READ_ONCE(dev->header_ops->cache))
             neigh->ops = &ndisc_hh_ops;
         else
             neigh->ops = &ndisc_generic_ops;
--
2.53.0


2026年2月3日(火) 2:11 Eric Dumazet <edumazet@google.com>:
>
> On Wed, Jan 28, 2026 at 11:46 AM 戸田晃太 <kota.toda@gmo-cybersecurity.com> wrote:
> >
> > Here is the patch with the barriers added, based on v6.12.67.
> >
> > However, as Yuki said, we are wondering if this would be considered an
> > acceptable change
> > from the perspective of the maintainers (or in terms of Linux kernel
> > culture). This is because
> > the patch adds `READ_ONCE` to several locations outside of bonding subsystem.
> > Please let me know if you have any concerns regarding this point.
> >
> > >  Also, please clarify what happens with stacks of two or more bonding devices ?
> >
> > To clarify, currently the `header_ops` of the bottom-most
> > interface are used regardless of the number of bonding layers.
> > This patch changes it so that `&bond->bond_header_ops` is used
> > as the bond device's `header_ops`, regardless of the stack depth.
>
> Could you try to cook a patch series perhaps ?
>
> The READ_ONCE()/WRITE_ONCE() on dev->header_ops->cache could be done separately.
>
> Thanks.