B.A.T.M.A.N June 2016

b.a.t.m.a.n@lists.open-mesh.org

14 participants
56 discussions

[B.A.T.M.A.N.] [PATCHv2] batman-adv: Remove unused primary_if and bat_priv variables
by Linus Lüssing 20 Jun '16

20 Jun '16

Acked-by: Simon Wunderlich <sw(a)simonwunderlich.de> Fixes: 29b9256e6631 ("batman-adv: consider outgoing interface in OGM sending") Signed-off-by: Linus Lüssing <linus.luessing(a)c0d3.blue> --- v2: * Remove now unused bat_priv variable, too net/batman-adv/bat_iv_ogm.c | 19 ++++--------------- 1 file changed, 4 insertions(+), 15 deletions(-) diff --git a/net/batman-adv/bat_iv_ogm.c b/net/batman-adv/bat_iv_ogm.c index 19b0abd..6af4462 100644 --- a/net/batman-adv/bat_iv_ogm.c +++ b/net/batman-adv/bat_iv_ogm.c @@ -528,36 +528,25 @@ static void batadv_iv_ogm_send_to_if(struct batadv_forw_packet *forw_packet, static void batadv_iv_ogm_emit(struct batadv_forw_packet *forw_packet) { struct net_device *soft_iface; - struct batadv_priv *bat_priv; - struct batadv_hard_iface *primary_if = NULL; if (!forw_packet->if_incoming) { pr_err("Error - can't forward packet: incoming iface not specified\n"); - goto out; + return; } soft_iface = forw_packet->if_incoming->soft_iface; - bat_priv = netdev_priv(soft_iface); if (WARN_ON(!forw_packet->if_outgoing)) - goto out; + return; if (WARN_ON(forw_packet->if_outgoing->soft_iface != soft_iface)) - goto out; + return; if (forw_packet->if_incoming->if_status != BATADV_IF_ACTIVE) - goto out; - - primary_if = batadv_primary_if_get_selected(bat_priv); - if (!primary_if) - goto out; + return; /* only for one specific outgoing interface */ batadv_iv_ogm_send_to_if(forw_packet, forw_packet->if_outgoing); - -out: - if (primary_if) - batadv_hardif_put(primary_if); } /** -- 2.1.4

3 2

[B.A.T.M.A.N.] [PATCH v5 1/4] Revert "batman-adv: Fix hardif remove/add race"
by Sven Eckelmann 20 Jun '16

20 Jun '16

The description given in the commit is misleading and is at least not true for sysfs. The sysfs file structure should instead be deleted immediately on rtnl notifications and be underlying objects should be removed later when all kobject_put were done for the object. This reverts commit 6b0485c758be ("batman-adv: Fix hardif remove/add race") Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- v5: - add this revert --- net/batman-adv/debugfs.c | 23 ----------------------- net/batman-adv/debugfs.h | 1 - net/batman-adv/hard-interface.c | 19 ------------------- net/batman-adv/sysfs.c | 17 ----------------- net/batman-adv/sysfs.h | 2 -- 5 files changed, 62 deletions(-) diff --git a/net/batman-adv/debugfs.c b/net/batman-adv/debugfs.c index aea4133..1d68b6e 100644 --- a/net/batman-adv/debugfs.c +++ b/net/batman-adv/debugfs.c @@ -44,7 +44,6 @@ #include "translation-table.h" static struct dentry *batadv_debugfs; -static atomic_t batadv_rename = ATOMIC_INIT(0); static int batadv_algorithms_open(struct inode *inode, struct file *file) { @@ -348,28 +347,6 @@ void batadv_debugfs_del_hardif(struct batadv_hard_iface *hard_iface) } } -/** - * batadv_debugfs_rename_hardif - rename the base directory - * @hard_iface: hard interface which is renamed. - * - * The interface may be removed and then quickly added back - * again. Rename the old instance to something temporary and unique, - * so avoiding a name space clash if it does reappear before the deferred - * work completes the removal. - */ -void batadv_debugfs_rename_hardif(struct batadv_hard_iface *hard_iface) -{ - char new_name[32]; - - snprintf(new_name, sizeof(*new_name) - 1, "tmp-%d", - atomic_inc_return(&batadv_rename)); - - if (batadv_debugfs && hard_iface->debug_dir) { - debugfs_rename(batadv_debugfs, hard_iface->debug_dir, - batadv_debugfs, new_name); - } -} - int batadv_debugfs_add_meshif(struct net_device *dev) { struct batadv_priv *bat_priv = netdev_priv(dev); diff --git a/net/batman-adv/debugfs.h b/net/batman-adv/debugfs.h index e7d19c1..1ab4e2e 100644 --- a/net/batman-adv/debugfs.h +++ b/net/batman-adv/debugfs.h @@ -34,7 +34,6 @@ int batadv_debugfs_add_meshif(struct net_device *dev); void batadv_debugfs_del_meshif(struct net_device *dev); int batadv_debugfs_add_hardif(struct batadv_hard_iface *hard_iface); void batadv_debugfs_del_hardif(struct batadv_hard_iface *hard_iface); -void batadv_debugfs_rename_hardif(struct batadv_hard_iface *hard_iface); #else diff --git a/net/batman-adv/hard-interface.c b/net/batman-adv/hard-interface.c index ccc7641..1f90808 100644 --- a/net/batman-adv/hard-interface.c +++ b/net/batman-adv/hard-interface.c @@ -706,23 +706,6 @@ out: return NULL; } -/** - * batadv_hardif_rename - rename the sysfs and debugfs - * @hard_iface: The hard interface to rename - * - * The sysfs and debugfs files cannot be removed until all users close - * them. So the removal is deferred into a work queue. This however - * means if the interface comes back before the work queue runs, the - * files are still there, and so the create gives an EEXISTS error. To - * avoid this, rename them to a tempory name. - */ -static void batadv_hardif_rename(struct batadv_hard_iface *hard_iface) -{ - batadv_sysfs_rename_hardif(&hard_iface->hardif_obj, - hard_iface->net_dev); - batadv_debugfs_rename_hardif(hard_iface); -} - static void batadv_hardif_remove_interface(struct batadv_hard_iface *hard_iface) { ASSERT_RTNL(); @@ -736,8 +719,6 @@ static void batadv_hardif_remove_interface(struct batadv_hard_iface *hard_iface) return; hard_iface->if_status = BATADV_IF_TO_BE_REMOVED; - batadv_hardif_rename(hard_iface); - queue_work(batadv_event_workqueue, &hard_iface->cleanup_work); } diff --git a/net/batman-adv/sysfs.c b/net/batman-adv/sysfs.c index 9b2c28f..fe9ca94 100644 --- a/net/batman-adv/sysfs.c +++ b/net/batman-adv/sysfs.c @@ -48,8 +48,6 @@ #include "packet.h" #include "soft-interface.h" -static atomic_t batadv_rename = ATOMIC_INIT(0); - static struct net_device *batadv_kobj_to_netdev(struct kobject *obj) { struct device *dev = container_of(obj->parent, struct device, kobj); @@ -1048,21 +1046,6 @@ out: return -ENOMEM; } -void batadv_sysfs_rename_hardif(struct kobject **hardif_obj, - struct net_device *dev) -{ - char new_name[32]; - int err; - - snprintf(new_name, sizeof(*new_name) - 1, "tmp-%d", - atomic_inc_return(&batadv_rename)); - - err = kobject_rename(*hardif_obj, new_name); - if (err) - batadv_err(dev, "Can't rename sysfs dir: %s/%s: %d\n", - dev->name, new_name, err); -} - void batadv_sysfs_del_hardif(struct kobject **hardif_obj) { kobject_put(*hardif_obj); diff --git a/net/batman-adv/sysfs.h b/net/batman-adv/sysfs.h index 64d3722..c76021b 100644 --- a/net/batman-adv/sysfs.h +++ b/net/batman-adv/sysfs.h @@ -48,8 +48,6 @@ void batadv_sysfs_del_meshif(struct net_device *dev); int batadv_sysfs_add_hardif(struct kobject **hardif_obj, struct net_device *dev); void batadv_sysfs_del_hardif(struct kobject **hardif_obj); -void batadv_sysfs_rename_hardif(struct kobject **hardif_obj, - struct net_device *dev); int batadv_sysfs_add_vlan(struct net_device *dev, struct batadv_softif_vlan *vlan); void batadv_sysfs_del_vlan(struct batadv_priv *bat_priv, -- 2.8.1

2 7

[B.A.T.M.A.N.] [PATCH next] batman-adv: remove NET_XMIT_POLICED
by Sven Eckelmann 14 Jun '16

14 Jun '16

From: Florian Westphal <fw(a)strlen.de> BATMAN uses it as an intermediate return value to signal forwarding vs. buffering, but it did not return POLICED to callers outside of BATMAN. Signed-off-by: Florian Westphal <fw(a)strlen.de> Signed-off-by: David S. Miller <davem(a)davemloft.net> [sven(a)narfation.org: rebased on top of TX path rewrite] Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- Taken from upstream commit 99860208bc62 ("sched: remove NET_XMIT_POLICED"). Please make sure you first apply * https://patchwork.open-mesh.org/patch/16362/ * https://patchwork.open-mesh.org/patch/16363/ * https://patchwork.open-mesh.org/patch/16364/ --- net/batman-adv/send.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/net/batman-adv/send.c b/net/batman-adv/send.c index 49836da..729deec 100644 --- a/net/batman-adv/send.c +++ b/net/batman-adv/send.c @@ -20,6 +20,7 @@ #include <linux/atomic.h> #include <linux/byteorder/generic.h> +#include <linux/errno.h> #include <linux/etherdevice.h> #include <linux/fs.h> #include <linux/if.h> @@ -156,7 +157,7 @@ int batadv_send_unicast_skb(struct sk_buff *skb, * host, NULL can be passed as recv_if and no interface alternating is * attempted. * - * Return: -1 on failure (and the skb is not consumed), NET_XMIT_POLICED if the + * Return: -1 on failure (and the skb is not consumed), -EINPROGRESS if the * skb is buffered for later transmit or the NET_XMIT status returned by the * lower routine if the packet has been passed down. * @@ -191,7 +192,7 @@ int batadv_send_skb_to_orig(struct sk_buff *skb, * network coding fails, then send the packet as usual. */ if (recv_if && batadv_nc_skb_forward(skb, neigh_node)) - ret = NET_XMIT_POLICED; + ret = -EINPROGRESS; else ret = batadv_send_unicast_skb(skb, neigh_node); -- 2.8.1

1 0

Re: [B.A.T.M.A.N.] [PATCH 3.16 046/114] batman-adv: Fix broadcast/ogm queue limit on a removed interface
by Linus Lüssing 13 Jun '16

13 Jun '16

On Mon, Jun 13, 2016 at 07:36:37PM +0100, Ben Hutchings wrote: > 3.16.36-rc1 review patch. If anyone has any objections, please let me know. > Hi Ben, This one looks weird. The version you added for 3.2.81-rc1 looked better. Cheers, Linus > ------------------ > > From: Linus Lüssing <linus.luessing(a)c0d3.blue> > > commit c4fdb6cff2aa0ae740c5f19b6f745cbbe786d42f upstream. > > When removing a single interface while a broadcast or ogm packet is > still pending then we will free the forward packet without releasing the > queue slots again. > > This patch is supposed to fix this issue. > > Fixes: 6d5808d4ae1b ("batman-adv: Add missing hardif_free_ref in forw_packet_free") > Signed-off-by: Linus Lüssing <linus.luessing(a)c0d3.blue> > [sven(a)narfation.org: fix conflicts with current version] > Signed-off-by: Sven Eckelmann <sven(a)narfation.org> > Signed-off-by: Marek Lindner <mareklindner(a)neomailbox.ch> > Signed-off-by: Antonio Quartulli <a(a)unstable.cc> > Signed-off-by: Ben Hutchings <ben(a)decadent.org.uk> > --- > net/batman-adv/send.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > --- a/net/batman-adv/send.c > +++ b/net/batman-adv/send.c > @@ -638,6 +638,12 @@ batadv_purge_outstanding_packets(struct > > if (pending) { > hlist_del(&forw_packet->list); > + if (!forw_packet->own) > + atomic_inc(&bat_priv->bcast_queue_left); > + > + if (!forw_packet->own) > + atomic_inc(&bat_priv->batman_queue_left); > + > batadv_forw_packet_free(forw_packet); > } > } >

1 0

[B.A.T.M.A.N.] [PATCH v9 00/14] netns and netlink support
by Simon Wunderlich 13 Jun '16

13 Jun '16

This patchset completes netns support, by disabling debugfs entries when not in the default name space, and correctly handling interface stack loops when the parent is in a different name space. It additionally adds netlink support for most of the information found in debugfs, and is netns aware. This patchset is based on Andrews v7 series and contains an update/rebase of the v8 sent by myself on May 23rd, with the following changes: * removed useless initialization "struct net_device *soft_iface = NULL;" in: - batadv_tt_global_dump - batadv_hardif_neigh_dump - batadv_orig_dump - batadv_gw_dump - batadv_bla_claim_dump - batadv_bla_backbone_dump * add kerneldoc about rcu_read_lock requirement to: - batadv_tt_global_dump_entry - batadv_iv_ogm_neigh_dump_hardif - batadv_iv_ogm_orig_dump_entry - batadv_v_neigh_dump_hardif - batadv_v_orig_dump_entry * simplify code to add BATADV_ATTR_LAST_SEEN_MSECS in batadv_tt_local_dump_entry * rebased on top of current master and fix conflicts with batadv_algo_ops restructuring It should apply clean on todays master. The batctl patches will be sent separately. Cheers, Simon Andrew Lunn (5): batman-adv: Handle parent interfaces in a different netns batman-adv: Suppress debugfs entries for netns's batman-adv: add B.A.T.M.A.N. Dump gateways via netlink batman-adv: add B.A.T.M.A.N. Dump BLA claims via netlink batman-adv: Indicate netlink socket can be used with netns. Matthias Schiffer (6): batman-adv: netlink: add routing_algo query batman-adv: netlink: hardif query batman-adv: netlink: add translation table query batman-adv: netlink: add originator and neighbor table queries batman-adv: add B.A.T.M.A.N. IV bat_{orig, neigh}_dump implementations batman-adv: add B.A.T.M.A.N. V bat_{orig, neigh}_dump implementations Simon Wunderlich (1): batman-adv: add backbone table netlink support Sven Eckelmann (2): batman-adv: Provide TTVN in the mesh_info netlink msg batman-adv: Provide bla group in the mesh_info netlink msg compat-include/linux/netlink.h | 18 ++ compat.h | 7 + include/uapi/linux/batman_adv.h | 94 ++++++++ net/batman-adv/bat_algo.c | 68 ++++++ net/batman-adv/bat_algo.h | 3 + net/batman-adv/bat_iv_ogm.c | 366 ++++++++++++++++++++++++++++++++ net/batman-adv/bat_v.c | 341 +++++++++++++++++++++++++++++ net/batman-adv/bridge_loop_avoidance.c | 333 +++++++++++++++++++++++++++++ net/batman-adv/bridge_loop_avoidance.h | 17 +- net/batman-adv/debugfs.c | 18 ++ net/batman-adv/gateway_client.c | 148 +++++++++++++ net/batman-adv/gateway_client.h | 2 + net/batman-adv/hard-interface.c | 50 ++++- net/batman-adv/netlink.c | 217 ++++++++++++++++++- net/batman-adv/netlink.h | 6 + net/batman-adv/originator.c | 160 ++++++++++++++ net/batman-adv/originator.h | 4 + net/batman-adv/packet.h | 36 ---- net/batman-adv/translation-table.c | 377 +++++++++++++++++++++++++++++++++ net/batman-adv/translation-table.h | 4 + net/batman-adv/types.h | 9 + 21 files changed, 2230 insertions(+), 48 deletions(-) -- 2.8.1

2 15

[B.A.T.M.A.N.] [PATCH v4 1/5] batman-adv: Define module rtnl link name
by Sven Eckelmann 13 Jun '16

13 Jun '16

The batman-adv module can automatically be loaded when operations over the rtnl link are triggered. This requires only the correct rtnl link name in the module header. Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- v4: - rebase on top of current master v3: - rebased on top of current master to fix conflicts with newest patches v2: - rebased on top of current master to fix conflicts with newest patches --- net/batman-adv/main.c | 1 + 1 file changed, 1 insertion(+) diff --git a/net/batman-adv/main.c b/net/batman-adv/main.c index fe4c5e2..f61479b 100644 --- a/net/batman-adv/main.c +++ b/net/batman-adv/main.c @@ -638,3 +638,4 @@ MODULE_AUTHOR(BATADV_DRIVER_AUTHOR); MODULE_DESCRIPTION(BATADV_DRIVER_DESC); MODULE_SUPPORTED_DEVICE(BATADV_DRIVER_DEVICE); MODULE_VERSION(BATADV_SOURCE_VERSION); +MODULE_ALIAS_RTNL_LINK("batadv"); -- 2.8.1

2 6

[B.A.T.M.A.N.] [PATCH] batman-adv: Document optional batadv_algo_ops
by Sven Eckelmann 13 Jun '16

13 Jun '16

Some operations in batadv_algo_ops are optional and marked as such in the kerneldoc. But some of them miss the "(optional)" in their kerneldoc. These have to also be marked to give an implementor of an algorithm the correct background information without looking in the code calling these function pointers. Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- net/batman-adv/types.h | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/net/batman-adv/types.h b/net/batman-adv/types.h index 43db7b6..0eeb68f 100644 --- a/net/batman-adv/types.h +++ b/net/batman-adv/types.h @@ -1392,6 +1392,7 @@ struct batadv_forw_packet { /** * struct batadv_algo_iface_ops - mesh algorithm callbacks (interface specific) * @activate: start routing mechanisms when hard-interface is brought up + * (optional) * @enable: init routing info when hard-interface is enabled * @disable: de-init routing info when hard-interface is disabled * @update_mac: (re-)init mac addresses of the protocol information @@ -1409,6 +1410,7 @@ struct batadv_algo_iface_ops { /** * struct batadv_algo_neigh_ops - mesh algorithm callbacks (neighbour specific) * @hardif_init: called on creation of single hop entry + * (optional) * @cmp: compare the metrics of two neighbors for their respective outgoing * interfaces * @is_similar_or_better: check if neigh1 is equally similar or better than @@ -1431,11 +1433,11 @@ struct batadv_algo_neigh_ops { /** * struct batadv_algo_orig_ops - mesh algorithm callbacks (originator specific) * @free: free the resources allocated by the routing algorithm for an orig_node - * object + * object (optional) * @add_if: ask the routing algorithm to apply the needed changes to the - * orig_node due to a new hard-interface being added into the mesh + * orig_node due to a new hard-interface being added into the mesh (optional) * @del_if: ask the routing algorithm to apply the needed changes to the - * orig_node due to an hard-interface being removed from the mesh + * orig_node due to an hard-interface being removed from the mesh (optional) * @print: print the originator table (optional) */ struct batadv_algo_orig_ops { -- 2.8.1

2 1

[B.A.T.M.A.N.] [PATCH] batctl: Fix format string for raw tp_meter output
by Sven Eckelmann 13 Jun '16

13 Jun '16

The tp_meter output for raw bytes is not casting the values to float. It must therefore use PRIu64 to print the results. Reported-by: Simon Wunderlich <sw(a)simonwunderlich.de> Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- tp_meter.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tp_meter.c b/tp_meter.c index d980ebd..43c19da 100644 --- a/tp_meter.c +++ b/tp_meter.c @@ -524,8 +524,8 @@ int tp_meter(char *mesh_iface, int argc, char **argv) (float)throughput / (1<<10), (float)throughput * 8 / 1000); else - printf("%lu Bytes/s (%lu Bps)\n", - throughput, throughput * 8); + printf("%" PRIu64 " Bytes/s (%" PRIu64 " Bps)\n", + throughput, throughput * 8); ret = 0; break; -- 2.8.1

2 1

[B.A.T.M.A.N.] [PATCH] batman-adv: Fix hardif remove/add race
by Andrew Lunn 13 Jun '16

13 Jun '16

A hard interface can be removed and then added back in quick succession. This is particularly true for hdlc interface when changing the protocol. It is not possible it synchronously remove the sysfs and debugfs entries for the hard interface when it is removed because the files may be open. Thus removal is deferred. The files may thus already exist in sysfs and debugfs when the hard interface is re-added, and the operations fail. To fix this race, synchronously rename the debugfs and sysfs files to a unique temporary name, thus making the name available when the interface comes back, yet keeps open files still available. Signed-off-by: Andrew Lunn <andrew(a)lunn.ch> --- net/batman-adv/debugfs.c | 23 +++++++++++++++++++++++ net/batman-adv/debugfs.h | 1 + net/batman-adv/hard-interface.c | 19 +++++++++++++++++++ net/batman-adv/sysfs.c | 17 +++++++++++++++++ net/batman-adv/sysfs.h | 2 ++ 5 files changed, 62 insertions(+) diff --git a/net/batman-adv/debugfs.c b/net/batman-adv/debugfs.c index 50545b0..8b852aa 100644 --- a/net/batman-adv/debugfs.c +++ b/net/batman-adv/debugfs.c @@ -55,6 +55,7 @@ #include "translation-table.h" static struct dentry *batadv_debugfs; +static atomic_t batadv_rename = ATOMIC_INIT(0); #ifdef CONFIG_BATMAN_ADV_DEBUG #define BATADV_LOG_BUFF_MASK (batadv_log_buff_len - 1) @@ -570,6 +571,28 @@ void batadv_debugfs_del_hardif(struct batadv_hard_iface *hard_iface) } } +/** + * batadv_debugfs_rename_hardif - rename the base directory + * @hard_iface: hard interface which is renamed. + * + * The interface may be removed and then quickly added back + * again. Rename the old instance to something temporary and unique, + * so avoiding a name space clash if it does reappear before the deferred + * work completes the removal. + */ +void batadv_debugfs_rename_hardif(struct batadv_hard_iface *hard_iface) +{ + char new_name[32]; + + snprintf(new_name, sizeof(*new_name) - 1, "tmp-%d", + atomic_inc_return(&batadv_rename)); + + if (batadv_debugfs && hard_iface->debug_dir) { + debugfs_rename(batadv_debugfs, hard_iface->debug_dir, + batadv_debugfs, new_name); + } +} + int batadv_debugfs_add_meshif(struct net_device *dev) { struct batadv_priv *bat_priv = netdev_priv(dev); diff --git a/net/batman-adv/debugfs.h b/net/batman-adv/debugfs.h index 1ab4e2e..e7d19c1 100644 --- a/net/batman-adv/debugfs.h +++ b/net/batman-adv/debugfs.h @@ -34,6 +34,7 @@ int batadv_debugfs_add_meshif(struct net_device *dev); void batadv_debugfs_del_meshif(struct net_device *dev); int batadv_debugfs_add_hardif(struct batadv_hard_iface *hard_iface); void batadv_debugfs_del_hardif(struct batadv_hard_iface *hard_iface); +void batadv_debugfs_rename_hardif(struct batadv_hard_iface *hard_iface); #else diff --git a/net/batman-adv/hard-interface.c b/net/batman-adv/hard-interface.c index bf4ee24..08b62d9 100644 --- a/net/batman-adv/hard-interface.c +++ b/net/batman-adv/hard-interface.c @@ -747,6 +747,23 @@ out: return NULL; } +/** + * batadv_hardif_rename - rename the sysfs and debugfs + * @hard_iface: The hard interface to rename + * + * The sysfs and debugfs files cannot be removed until all users close + * them. So the removal is differed into a work queue. This however + * means if the interface comes back before the work queue runs, the + * files are still there, and so the create gives an EEXISTS error. To + * avoid this, rename them to a tempory name. + */ +static void batadv_hardif_rename(struct batadv_hard_iface *hard_iface) +{ + batadv_sysfs_rename_hardif(&hard_iface->hardif_obj, + hard_iface->net_dev); + batadv_debugfs_rename_hardif(hard_iface); +} + static void batadv_hardif_remove_interface(struct batadv_hard_iface *hard_iface) { ASSERT_RTNL(); @@ -760,6 +777,8 @@ static void batadv_hardif_remove_interface(struct batadv_hard_iface *hard_iface) return; hard_iface->if_status = BATADV_IF_TO_BE_REMOVED; + batadv_hardif_rename(hard_iface); + queue_work(batadv_event_workqueue, &hard_iface->cleanup_work); } diff --git a/net/batman-adv/sysfs.c b/net/batman-adv/sysfs.c index 233abcf..37b0aae 100644 --- a/net/batman-adv/sysfs.c +++ b/net/batman-adv/sysfs.c @@ -47,6 +47,8 @@ #include "packet.h" #include "soft-interface.h" +static atomic_t batadv_rename = ATOMIC_INIT(0); + static struct net_device *batadv_kobj_to_netdev(struct kobject *obj) { struct device *dev = container_of(obj->parent, struct device, kobj); @@ -1045,6 +1047,21 @@ out: return -ENOMEM; } +void batadv_sysfs_rename_hardif(struct kobject **hardif_obj, + struct net_device *dev) +{ + char new_name[32]; + int err; + + snprintf(new_name, sizeof(*new_name) - 1, "tmp-%d", + atomic_inc_return(&batadv_rename)); + + err = kobject_rename(*hardif_obj, new_name); + if (err) + batadv_err(dev, "Can't rename sysfs dir: %s/%s: %d\n", + dev->name, new_name, err); +} + void batadv_sysfs_del_hardif(struct kobject **hardif_obj) { kobject_put(*hardif_obj); diff --git a/net/batman-adv/sysfs.h b/net/batman-adv/sysfs.h index c76021b..64d3722 100644 --- a/net/batman-adv/sysfs.h +++ b/net/batman-adv/sysfs.h @@ -48,6 +48,8 @@ void batadv_sysfs_del_meshif(struct net_device *dev); int batadv_sysfs_add_hardif(struct kobject **hardif_obj, struct net_device *dev); void batadv_sysfs_del_hardif(struct kobject **hardif_obj); +void batadv_sysfs_rename_hardif(struct kobject **hardif_obj, + struct net_device *dev); int batadv_sysfs_add_vlan(struct net_device *dev, struct batadv_softif_vlan *vlan); void batadv_sysfs_del_vlan(struct batadv_priv *bat_priv, -- 2.8.0.rc3

2 1

[B.A.T.M.A.N.] [PATCH] batman-adv: Don't propagate negative dev_queue_xmit return values
by Sven Eckelmann 10 Jun '16

10 Jun '16

batadv_send_skb_packet used by batadv_send_skb_to_orig and its return value is given directly to callers of batadv_send_skb_packet. batadv_send_skb_to_orig -> batadv_send_unicast_skb -> batadv_send_skb_packet -> dev_queue_xmit These callers of batadv_send_skb_to_orig expect that the skb isn't consumed when they receive a -1. But dev_queue_xmit may still have consumed it and still returned -1. Thus the free for the skb would be called twice. Fixes: e3b8acbff9c8 ("batman-adv: return netdev status in the TX path") Signed-off-by: Sven Eckelmann <sven(a)narfation.org> --- Antonio, this is not really tested. Could you please review it and tell me if I may have missed something. net/batman-adv/send.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/net/batman-adv/send.c b/net/batman-adv/send.c index 49836da..d76ccb2 100644 --- a/net/batman-adv/send.c +++ b/net/batman-adv/send.c @@ -72,6 +72,7 @@ int batadv_send_skb_packet(struct sk_buff *skb, { struct batadv_priv *bat_priv; struct ethhdr *ethhdr; + int ret; bat_priv = netdev_priv(hard_iface->soft_iface); @@ -109,8 +110,15 @@ int batadv_send_skb_packet(struct sk_buff *skb, /* dev_queue_xmit() returns a negative result on error. However on * congestion and traffic shaping, it drops and returns NET_XMIT_DROP * (which is > 0). This will not be treated as an error. + * + * a negative value cannot be returned because it could be interepreted + * as not consumed skb by callers of batadv_send_skb_to_orig. */ - return dev_queue_xmit(skb); + ret = dev_queue_xmit(skb); + if (ret < 0) + ret = NET_XMIT_DROP; + + return ret; send_skb_err: kfree_skb(skb); return NET_XMIT_DROP; -- 2.8.1

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

B.A.T.M.A.N June 2016