X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=datapath%2Fvport.c;h=cdf615a47b8b135c631988ba6d52e9666391bab3;hb=33ef7695b71a71f2d0bb0e00aab46855afb8c2a6;hp=13039505c687f978d94d98d83552deb1ae0e8fe5;hpb=ecbb6953b37f81f5ea1979aa4e5d8e11df0929ad;p=openvswitch diff --git a/datapath/vport.c b/datapath/vport.c index 13039505..cdf615a4 100644 --- a/datapath/vport.c +++ b/datapath/vport.c @@ -9,6 +9,7 @@ #include #include #include +#include #include #include #include @@ -34,17 +35,6 @@ static int n_vport_types; static struct hlist_head *dev_table; #define VPORT_HASH_BUCKETS 1024 -/* We limit the number of times that we pass through vport_send() to - * avoid blowing out the stack in the event that we have a loop. There is - * a separate counter for each CPU for both interrupt and non-interrupt - * context in order to keep the limit deterministic for a given packet. */ -struct percpu_loop_counter { - int count[2]; -}; - -static struct percpu_loop_counter *vport_loop_counter; -#define VPORT_MAX_LOOPS 5 - /* Both RTNL lock and vport_mutex need to be held when updating dev_table. * * If you use vport_locate and then perform some operations, you need to hold @@ -67,8 +57,7 @@ static DEFINE_MUTEX(vport_mutex); * Acquire global vport lock. See above comment about locking requirements * and specific function definitions. May sleep. */ -void -vport_lock(void) +void vport_lock(void) { mutex_lock(&vport_mutex); } @@ -78,8 +67,7 @@ vport_lock(void) * * Release lock acquired with vport_lock. */ -void -vport_unlock(void) +void vport_unlock(void) { mutex_unlock(&vport_mutex); } @@ -98,8 +86,7 @@ vport_unlock(void) * Called at module load time to initialize the vport subsystem and any * compiled in vport types. */ -int -vport_init(void) +int vport_init(void) { int err; int i; @@ -118,12 +105,6 @@ vport_init(void) goto error_dev_table; } - vport_loop_counter = alloc_percpu(struct percpu_loop_counter); - if (!vport_loop_counter) { - err = -ENOMEM; - goto error_ops_list; - } - for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) { struct vport_ops *new_ops = base_vport_ops_list[i]; @@ -142,16 +123,13 @@ vport_init(void) return 0; -error_ops_list: - kfree(vport_ops_list); error_dev_table: kfree(dev_table); error: return err; } -static void -vport_del_all(void) +static void vport_del_all(void) { int i; @@ -177,8 +155,7 @@ vport_del_all(void) * Called at module exit time to shutdown the vport subsystem and any * initialized vport types. */ -void -vport_exit(void) +void vport_exit(void) { int i; @@ -189,13 +166,11 @@ vport_exit(void) vport_ops_list[i]->exit(); } - free_percpu(vport_loop_counter); kfree(vport_ops_list); kfree(dev_table); } -static int -do_vport_add(struct odp_vport_add *vport_config) +static int do_vport_add(struct odp_vport_add *vport_config) { struct vport *vport; int err = 0; @@ -207,7 +182,7 @@ do_vport_add(struct odp_vport_add *vport_config) vport = vport_locate(vport_config->devname); if (vport) { - err = -EEXIST; + err = -EBUSY; goto out; } @@ -233,8 +208,7 @@ out: * on device type). This function is for userspace callers and assumes no * locks are held. */ -int -vport_user_add(const struct odp_vport_add __user *uvport_config) +int vport_user_add(const struct odp_vport_add __user *uvport_config) { struct odp_vport_add vport_config; @@ -245,8 +219,7 @@ vport_user_add(const struct odp_vport_add __user *uvport_config) } #ifdef CONFIG_COMPAT -int -compat_vport_user_add(struct compat_odp_vport_add *ucompat) +int compat_vport_user_add(struct compat_odp_vport_add *ucompat) { struct compat_odp_vport_add compat; struct odp_vport_add vport_config; @@ -262,8 +235,7 @@ compat_vport_user_add(struct compat_odp_vport_add *ucompat) } #endif -static int -do_vport_mod(struct odp_vport_mod *vport_config) +static int do_vport_mod(struct odp_vport_mod *vport_config) { struct vport *vport; int err; @@ -296,8 +268,7 @@ out: * dependent on device type). This function is for userspace callers and * assumes no locks are held. */ -int -vport_user_mod(const struct odp_vport_mod __user *uvport_config) +int vport_user_mod(const struct odp_vport_mod __user *uvport_config) { struct odp_vport_mod vport_config; @@ -308,8 +279,7 @@ vport_user_mod(const struct odp_vport_mod __user *uvport_config) } #ifdef CONFIG_COMPAT -int -compat_vport_user_mod(struct compat_odp_vport_mod *ucompat) +int compat_vport_user_mod(struct compat_odp_vport_mod *ucompat) { struct compat_odp_vport_mod compat; struct odp_vport_mod vport_config; @@ -335,8 +305,7 @@ compat_vport_user_mod(struct compat_odp_vport_mod *ucompat) * reasons, such as lack of memory. This function is for userspace callers and * assumes no locks are held. */ -int -vport_user_del(const char __user *udevname) +int vport_user_del(const char __user *udevname) { char devname[IFNAMSIZ]; struct vport *vport; @@ -395,8 +364,7 @@ out: * Retrieves transmit, receive, and error stats for the given device. This * function is for userspace callers and assumes no locks are held. */ -int -vport_user_stats_get(struct odp_vport_stats_req __user *ustats_req) +int vport_user_stats_get(struct odp_vport_stats_req __user *ustats_req) { struct odp_vport_stats_req stats_req; struct vport *vport; @@ -438,8 +406,7 @@ out: * -EOPNOTSUPP. This function is for userspace callers and assumes no locks * are held. */ -int -vport_user_stats_set(struct odp_vport_stats_req __user *ustats_req) +int vport_user_stats_set(struct odp_vport_stats_req __user *ustats_req) { struct odp_vport_stats_req stats_req; struct vport *vport; @@ -476,8 +443,7 @@ out: * Retrieves the Ethernet address of the given device. This function is for * userspace callers and assumes no locks are held. */ -int -vport_user_ether_get(struct odp_vport_ether __user *uvport_ether) +int vport_user_ether_get(struct odp_vport_ether __user *uvport_ether) { struct odp_vport_ether vport_ether; struct vport *vport; @@ -520,8 +486,7 @@ out: * -EOPNOTSUPP. This function is for userspace callers and assumes no locks * are held. */ -int -vport_user_ether_set(struct odp_vport_ether __user *uvport_ether) +int vport_user_ether_set(struct odp_vport_ether __user *uvport_ether) { struct odp_vport_ether vport_ether; struct vport *vport; @@ -557,8 +522,7 @@ out: * Retrieves the MTU of the given device. This function is for userspace * callers and assumes no locks are held. */ -int -vport_user_mtu_get(struct odp_vport_mtu __user *uvport_mtu) +int vport_user_mtu_get(struct odp_vport_mtu __user *uvport_mtu) { struct odp_vport_mtu vport_mtu; struct vport *vport; @@ -598,8 +562,7 @@ out: * MTU, in which case the result will always be -EOPNOTSUPP. This function is * for userspace callers and assumes no locks are held. */ -int -vport_user_mtu_set(struct odp_vport_mtu __user *uvport_mtu) +int vport_user_mtu_set(struct odp_vport_mtu __user *uvport_mtu) { struct odp_vport_mtu vport_mtu; struct vport *vport; @@ -627,8 +590,7 @@ out: return err; } -static struct hlist_head * -hash_bucket(const char *name) +static struct hlist_head *hash_bucket(const char *name) { unsigned int hash = full_name_hash(name, strlen(name)); return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)]; @@ -643,8 +605,7 @@ hash_bucket(const char *name) * and held while using the found port. See the locking comments at the * top of the file. */ -struct vport * -vport_locate(const char *name) +struct vport *vport_locate(const char *name) { struct hlist_head *bucket = hash_bucket(name); struct vport *vport; @@ -668,14 +629,12 @@ out: return vport; } -static void -register_vport(struct vport *vport) +static void register_vport(struct vport *vport) { hlist_add_head(&vport->hash_node, hash_bucket(vport_get_name(vport))); } -static void -unregister_vport(struct vport *vport) +static void unregister_vport(struct vport *vport) { hlist_del(&vport->hash_node); } @@ -691,8 +650,7 @@ unregister_vport(struct vport *vport) * vport_priv(). vports that are no longer needed should be released with * vport_free(). */ -struct vport * -vport_alloc(int priv_size, const struct vport_ops *ops) +struct vport *vport_alloc(int priv_size, const struct vport_ops *ops) { struct vport *vport; size_t alloc_size; @@ -727,8 +685,7 @@ vport_alloc(int priv_size, const struct vport_ops *ops) * * Frees a vport allocated with vport_alloc() when it is no longer needed. */ -void -vport_free(struct vport *vport) +void vport_free(struct vport *vport) { if (vport->ops->flags & VPORT_F_GEN_STATS) free_percpu(vport->percpu_stats); @@ -747,8 +704,7 @@ vport_free(struct vport *vport) * Creates a new vport with the specified configuration (which is dependent * on device type). Both RTNL and vport locks must be held. */ -struct vport * -vport_add(const char *name, const char *type, const void __user *config) +struct vport *vport_add(const char *name, const char *type, const void __user *config) { struct vport *vport; int err = 0; @@ -785,8 +741,7 @@ out: * Modifies an existing device with the specified configuration (which is * dependent on device type). Both RTNL and vport locks must be held. */ -int -vport_mod(struct vport *vport, const void __user *config) +int vport_mod(struct vport *vport, const void __user *config) { ASSERT_RTNL(); ASSERT_VPORT(); @@ -806,8 +761,7 @@ vport_mod(struct vport *vport, const void __user *config) * a datapath. It is possible to fail for reasons such as lack of memory. * Both RTNL and vport locks must be held. */ -int -vport_del(struct vport *vport) +int vport_del(struct vport *vport) { ASSERT_RTNL(); ASSERT_VPORT(); @@ -829,14 +783,10 @@ vport_del(struct vport *vport) * attached to a vport before it is connected to a datapath and must not be * modified while connected. RTNL lock and the appropriate DP mutex must be held. */ -int -vport_attach(struct vport *vport, struct dp_port *dp_port) +int vport_attach(struct vport *vport, struct dp_port *dp_port) { ASSERT_RTNL(); - if (dp_port->vport) - return -EBUSY; - if (vport_get_dp_port(vport)) return -EBUSY; @@ -848,7 +798,6 @@ vport_attach(struct vport *vport, struct dp_port *dp_port) return err; } - dp_port->vport = vport; rcu_assign_pointer(vport->dp_port, dp_port); return 0; @@ -862,8 +811,7 @@ vport_attach(struct vport *vport, struct dp_port *dp_port) * Detaches a vport from a datapath. May fail for a variety of reasons, * including lack of memory. RTNL lock and the appropriate DP mutex must be held. */ -int -vport_detach(struct vport *vport) +int vport_detach(struct vport *vport) { struct dp_port *dp_port; @@ -873,7 +821,6 @@ vport_detach(struct vport *vport) if (!dp_port) return -EINVAL; - dp_port->vport = NULL; rcu_assign_pointer(vport->dp_port, NULL); if (vport->ops->detach) @@ -892,8 +839,7 @@ vport_detach(struct vport *vport) * MTU, in which case the result will always be -EOPNOTSUPP. RTNL lock must * be held. */ -int -vport_set_mtu(struct vport *vport, int mtu) +int vport_set_mtu(struct vport *vport, int mtu) { ASSERT_RTNL(); @@ -927,8 +873,7 @@ vport_set_mtu(struct vport *vport, int mtu) * setting the Ethernet address, in which case the result will always be * -EOPNOTSUPP. RTNL lock must be held. */ -int -vport_set_addr(struct vport *vport, const unsigned char *addr) +int vport_set_addr(struct vport *vport, const unsigned char *addr) { ASSERT_RTNL(); @@ -952,8 +897,7 @@ vport_set_addr(struct vport *vport, const unsigned char *addr) * support setting the stats, in which case the result will always be * -EOPNOTSUPP. RTNL lock must be held. */ -int -vport_set_stats(struct vport *vport, struct odp_vport_stats *stats) +int vport_set_stats(struct vport *vport, struct odp_vport_stats *stats) { ASSERT_RTNL(); @@ -977,8 +921,7 @@ vport_set_stats(struct vport *vport, struct odp_vport_stats *stats) * Retrieves the name of the given device. Either RTNL lock or rcu_read_lock * must be held for the entire duration that the name is in use. */ -const char * -vport_get_name(const struct vport *vport) +const char *vport_get_name(const struct vport *vport) { return vport->ops->get_name(vport); } @@ -991,8 +934,7 @@ vport_get_name(const struct vport *vport) * Retrieves the type of the given device. Either RTNL lock or rcu_read_lock * must be held for the entire duration that the type is in use. */ -const char * -vport_get_type(const struct vport *vport) +const char *vport_get_type(const struct vport *vport) { return vport->ops->type; } @@ -1006,8 +948,7 @@ vport_get_type(const struct vport *vport) * rcu_read_lock must be held for the entire duration that the Ethernet address * is in use. */ -const unsigned char * -vport_get_addr(const struct vport *vport) +const unsigned char *vport_get_addr(const struct vport *vport) { return vport->ops->get_addr(vport); } @@ -1021,8 +962,7 @@ vport_get_addr(const struct vport *vport) * lock or rcu_read_lock must be held for the entire duration that the datapath * port is being accessed. */ -struct dp_port * -vport_get_dp_port(const struct vport *vport) +struct dp_port *vport_get_dp_port(const struct vport *vport) { return rcu_dereference(vport->dp_port); } @@ -1035,8 +975,7 @@ vport_get_dp_port(const struct vport *vport) * Retrieves the associated kobj or null if no kobj. The returned kobj is * valid for as long as the vport exists. */ -struct kobject * -vport_get_kobj(const struct vport *vport) +struct kobject *vport_get_kobj(const struct vport *vport) { if (vport->ops->get_kobj) return vport->ops->get_kobj(vport); @@ -1052,8 +991,7 @@ vport_get_kobj(const struct vport *vport) * * Retrieves transmit, receive, and error stats for the given device. */ -int -vport_get_stats(struct vport *vport, struct odp_vport_stats *stats) +int vport_get_stats(struct vport *vport, struct odp_vport_stats *stats) { struct odp_vport_stats dev_stats; struct odp_vport_stats *dev_statsp = NULL; @@ -1114,12 +1052,20 @@ vport_get_stats(struct vport *vport, struct odp_vport_stats *stats) for_each_possible_cpu(i) { const struct vport_percpu_stats *percpu_stats; + struct vport_percpu_stats local_stats; + unsigned seqcount; percpu_stats = per_cpu_ptr(vport->percpu_stats, i); - stats->rx_bytes += percpu_stats->rx_bytes; - stats->rx_packets += percpu_stats->rx_packets; - stats->tx_bytes += percpu_stats->tx_bytes; - stats->tx_packets += percpu_stats->tx_packets; + + do { + seqcount = read_seqcount_begin(&percpu_stats->seqlock); + local_stats = *percpu_stats; + } while (read_seqcount_retry(&percpu_stats->seqlock, seqcount)); + + stats->rx_bytes += local_stats.rx_bytes; + stats->rx_packets += local_stats.rx_packets; + stats->tx_bytes += local_stats.tx_bytes; + stats->tx_packets += local_stats.tx_packets; } err = 0; @@ -1138,8 +1084,7 @@ out: * Retrieves the flags of the given device. Either RTNL lock or rcu_read_lock * must be held. */ -unsigned -vport_get_flags(const struct vport *vport) +unsigned vport_get_flags(const struct vport *vport) { return vport->ops->get_dev_flags(vport); } @@ -1152,8 +1097,7 @@ vport_get_flags(const struct vport *vport) * Checks whether the given device is running. Either RTNL lock or * rcu_read_lock must be held. */ -int -vport_is_running(const struct vport *vport) +int vport_is_running(const struct vport *vport) { return vport->ops->is_running(vport); } @@ -1166,8 +1110,7 @@ vport_is_running(const struct vport *vport) * Retrieves the RFC2863 operstate of the given device. Either RTNL lock or * rcu_read_lock must be held. */ -unsigned char -vport_get_operstate(const struct vport *vport) +unsigned char vport_get_operstate(const struct vport *vport) { return vport->ops->get_operstate(vport); } @@ -1182,8 +1125,7 @@ vport_get_operstate(const struct vport *vport) * port is returned. Returns a negative index on error. Either RTNL lock or * rcu_read_lock must be held. */ -int -vport_get_ifindex(const struct vport *vport) +int vport_get_ifindex(const struct vport *vport) { const struct dp_port *dp_port; @@ -1210,8 +1152,7 @@ vport_get_ifindex(const struct vport *vport) * Returns a negative index on error. Either RTNL lock or rcu_read_lock must * be held. */ -int -vport_get_iflink(const struct vport *vport) +int vport_get_iflink(const struct vport *vport) { if (vport->ops->get_iflink) return vport->ops->get_iflink(vport); @@ -1229,8 +1170,7 @@ vport_get_iflink(const struct vport *vport) * Retrieves the MTU of the given device. Either RTNL lock or rcu_read_lock * must be held. */ -int -vport_get_mtu(const struct vport *vport) +int vport_get_mtu(const struct vport *vport) { return vport->ops->get_mtu(vport); } @@ -1245,8 +1185,7 @@ vport_get_mtu(const struct vport *vport) * skb->data should point to the Ethernet header. The caller must have already * called compute_ip_summed() to initialize the checksumming fields. */ -void -vport_receive(struct vport *vport, struct sk_buff *skb) +void vport_receive(struct vport *vport, struct sk_buff *skb) { struct dp_port *dp_port = vport_get_dp_port(vport); @@ -1261,10 +1200,12 @@ vport_receive(struct vport *vport, struct sk_buff *skb) struct vport_percpu_stats *stats; local_bh_disable(); - stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id()); + + write_seqcount_begin(&stats->seqlock); stats->rx_packets++; stats->rx_bytes += skb->len; + write_seqcount_end(&stats->seqlock); local_bh_enable(); } @@ -1275,6 +1216,16 @@ vport_receive(struct vport *vport, struct sk_buff *skb) dp_process_received_packet(dp_port, skb); } +static inline unsigned packet_length(const struct sk_buff *skb) +{ + unsigned length = skb->len - ETH_HLEN; + + if (skb->protocol == htons(ETH_P_8021Q)) + length -= VLAN_HLEN; + + return length; +} + /** * vport_send - send a packet on a device * @@ -1284,43 +1235,41 @@ vport_receive(struct vport *vport, struct sk_buff *skb) * Sends the given packet and returns the length of data sent. Either RTNL * lock or rcu_read_lock must be held. */ -int -vport_send(struct vport *vport, struct sk_buff *skb) +int vport_send(struct vport *vport, struct sk_buff *skb) { - int *loop_count; + int mtu; int sent; - loop_count = &per_cpu_ptr(vport_loop_counter, get_cpu())->count[!!in_interrupt()]; - (*loop_count)++; - - if (likely(*loop_count <= VPORT_MAX_LOOPS)) { - sent = vport->ops->send(vport, skb); - } else { + mtu = vport_get_mtu(vport); + if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) { if (net_ratelimit()) - printk(KERN_WARNING "%s: dropping packet that has looped more than %d times\n", - dp_name(vport_get_dp_port(vport)->dp), VPORT_MAX_LOOPS); - - sent = 0; - kfree_skb(skb); - vport_record_error(vport, VPORT_E_TX_DROPPED); + printk(KERN_WARNING "%s: dropped over-mtu packet: %d > %d\n", + dp_name(vport_get_dp_port(vport)->dp), packet_length(skb), mtu); + goto error; } - (*loop_count)--; - put_cpu(); + sent = vport->ops->send(vport, skb); if (vport->ops->flags & VPORT_F_GEN_STATS && sent > 0) { struct vport_percpu_stats *stats; local_bh_disable(); - stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id()); + + write_seqcount_begin(&stats->seqlock); stats->tx_packets++; stats->tx_bytes += sent; + write_seqcount_end(&stats->seqlock); local_bh_enable(); } return sent; + +error: + kfree_skb(skb); + vport_record_error(vport, VPORT_E_TX_DROPPED); + return 0; } /** @@ -1332,8 +1281,7 @@ vport_send(struct vport *vport, struct sk_buff *skb) * If using the vport generic stats layer indicate that an error of the given * type has occured. */ -void -vport_record_error(struct vport *vport, enum vport_err_type err_type) +void vport_record_error(struct vport *vport, enum vport_err_type err_type) { if (vport->ops->flags & VPORT_F_GEN_STATS) {