* [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups @ 2005-02-18 5:25 Matt Mackall 2005-02-18 5:25 ` [PATCH 1/6] netpoll: shorten carrier detect timeout Matt Mackall 2005-03-23 2:33 ` [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups David S. Miller 0 siblings, 2 replies; 8+ messages in thread From: Matt Mackall @ 2005-02-18 5:25 UTC (permalink / raw) To: David S. Miller; +Cc: netdev, Jeff Moyer This patch series against -rc4 fixes up some recursion deadlocks in netpoll and adds support for fallback to queueing. Various cleanups along the way. Holds up under load testing via ipt_LOG on a dual Opteron with tg3. ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 1/6] netpoll: shorten carrier detect timeout 2005-02-18 5:25 [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups Matt Mackall @ 2005-02-18 5:25 ` Matt Mackall 2005-02-18 5:25 ` [PATCH 2/6] netpoll: filter inlines Matt Mackall 2005-03-23 2:33 ` [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups David S. Miller 1 sibling, 1 reply; 8+ messages in thread From: Matt Mackall @ 2005-02-18 5:25 UTC (permalink / raw) To: David S. Miller; +Cc: netdev, Jeff Moyer Shorten carrier detect timeout to 4 seconds. Signed-off-by: Matt Mackall <mpm@selenic.com> Index: tiny-new/net/core/netpoll.c =================================================================== --- tiny-new.orig/net/core/netpoll.c 2004-11-17 00:05:28.000000000 -0800 +++ tiny-new/net/core/netpoll.c 2004-12-02 11:51:15.775256063 -0800 @@ -584,7 +584,7 @@ rtnl_shunlock(); atleast = jiffies + HZ/10; - atmost = jiffies + 10*HZ; + atmost = jiffies + 4*HZ; while (!netif_carrier_ok(ndev)) { if (time_after(jiffies, atmost)) { printk(KERN_NOTICE @@ -597,7 +597,7 @@ if (time_before(jiffies, atleast)) { printk(KERN_NOTICE "%s: carrier detect appears flaky," - " waiting 10 seconds\n", + " waiting 4 seconds\n", np->name); while (time_before(jiffies, atmost)) cond_resched(); ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 2/6] netpoll: filter inlines 2005-02-18 5:25 ` [PATCH 1/6] netpoll: shorten carrier detect timeout Matt Mackall @ 2005-02-18 5:25 ` Matt Mackall 2005-02-18 5:25 ` [PATCH 3/6] netpoll: add netpoll point to net_device Matt Mackall 0 siblings, 1 reply; 8+ messages in thread From: Matt Mackall @ 2005-02-18 5:25 UTC (permalink / raw) To: David S. Miller; +Cc: netdev, Jeff Moyer Add netpoll rx helpers Move skb_free for rx into __netpoll_rx Signed-off-by: Matt Mackall <mpm@selenic.com> Index: rc4bk2/net/core/netpoll.c =================================================================== --- rc4bk2.orig/net/core/netpoll.c 2005-02-14 10:34:12.000000000 -0800 +++ rc4bk2/net/core/netpoll.c 2005-02-14 17:10:34.000000000 -0800 @@ -368,7 +368,7 @@ netpoll_send_skb(np, send_skb); } -int netpoll_rx(struct sk_buff *skb) +int __netpoll_rx(struct sk_buff *skb) { int proto, len, ulen; struct iphdr *iph; @@ -440,12 +440,18 @@ (char *)(uh+1), ulen - sizeof(struct udphdr)); + kfree_skb(skb); return 1; } spin_unlock_irqrestore(&rx_list_lock, flags); out: - return atomic_read(&trapped); + if (atomic_read(&trapped)) { + kfree_skb(skb); + return 1; + } + + return 0; } int netpoll_parse_options(struct netpoll *np, char *opt) Index: rc4bk2/include/linux/netpoll.h =================================================================== --- rc4bk2.orig/include/linux/netpoll.h 2005-02-14 10:34:08.000000000 -0800 +++ rc4bk2/include/linux/netpoll.h 2005-02-14 17:10:34.000000000 -0800 @@ -30,7 +30,15 @@ int netpoll_trap(void); void netpoll_set_trap(int trap); void netpoll_cleanup(struct netpoll *np); -int netpoll_rx(struct sk_buff *skb); +int __netpoll_rx(struct sk_buff *skb); +#ifdef CONFIG_NETPOLL +static inline int netpoll_rx(struct sk_buff *skb) +{ + return skb->dev->netpoll_rx && __netpoll_rx(skb); +} +#else +#define netpoll_rx(a) 0 +#endif #endif Index: rc4bk2/net/core/dev.c =================================================================== --- rc4bk2.orig/net/core/dev.c 2005-02-14 10:34:12.000000000 -0800 +++ rc4bk2/net/core/dev.c 2005-02-14 17:10:34.000000000 -0800 @@ -1427,13 +1427,10 @@ struct softnet_data *queue; unsigned long flags; -#ifdef CONFIG_NETPOLL - if (skb->dev->netpoll_rx && netpoll_rx(skb)) { - kfree_skb(skb); + /* if netpoll wants it, pretend we never saw it */ + if (netpoll_rx(skb)) return NET_RX_DROP; - } -#endif - + if (!skb->stamp.tv_sec) net_timestamp(&skb->stamp); @@ -1629,12 +1626,9 @@ int ret = NET_RX_DROP; unsigned short type; -#ifdef CONFIG_NETPOLL - if (skb->dev->netpoll_rx && skb->dev->poll && netpoll_rx(skb)) { - kfree_skb(skb); + /* if we've gotten here through NAPI, check netpoll */ + if (skb->dev->poll && netpoll_rx(skb)) return NET_RX_DROP; - } -#endif if (!skb->stamp.tv_sec) net_timestamp(&skb->stamp); ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 3/6] netpoll: add netpoll point to net_device 2005-02-18 5:25 ` [PATCH 2/6] netpoll: filter inlines Matt Mackall @ 2005-02-18 5:25 ` Matt Mackall 2005-02-18 5:25 ` [PATCH 4/6] netpoll: fix ->poll() locking Matt Mackall 0 siblings, 1 reply; 8+ messages in thread From: Matt Mackall @ 2005-02-18 5:25 UTC (permalink / raw) To: David S. Miller; +Cc: netdev, Jeff Moyer Add struct netpoll pointer to struct netdevice Move netpoll rx flags to netpoll struct Stop traversing rx_list and get np pointer from skb->dev->np Remove now unneeded rx_list Signed-off-by: Matt Mackall <mpm@selenic.com> Index: rc4/include/linux/netdevice.h =================================================================== --- rc4.orig/include/linux/netdevice.h 2005-02-17 22:32:12.000000000 -0600 +++ rc4/include/linux/netdevice.h 2005-02-17 22:32:20.000000000 -0600 @@ -41,7 +41,7 @@ struct divert_blk; struct vlan_group; struct ethtool_ops; - +struct netpoll; /* source back-compat hooks */ #define SET_ETHTOOL_OPS(netdev,ops) \ ( (netdev)->ethtool_ops = (ops) ) @@ -471,7 +471,7 @@ int (*neigh_setup)(struct net_device *dev, struct neigh_parms *); int (*accept_fastpath)(struct net_device *, struct dst_entry*); #ifdef CONFIG_NETPOLL - int netpoll_rx; + struct netpoll *np; #endif #ifdef CONFIG_NET_POLL_CONTROLLER void (*poll_controller)(struct net_device *dev); Index: rc4/net/core/netpoll.c =================================================================== --- rc4.orig/net/core/netpoll.c 2005-02-17 22:32:19.000000000 -0600 +++ rc4/net/core/netpoll.c 2005-02-17 22:39:59.000000000 -0600 @@ -35,9 +35,6 @@ static int nr_skbs; static struct sk_buff *skbs; -static DEFINE_SPINLOCK(rx_list_lock); -static LIST_HEAD(rx_list); - static atomic_t trapped; static DEFINE_SPINLOCK(netpoll_poll_lock); @@ -84,13 +81,13 @@ queue = &__get_cpu_var(softnet_data); if (test_bit(__LINK_STATE_RX_SCHED, &np->dev->state) && !list_empty(&queue->poll_list)) { - np->dev->netpoll_rx |= NETPOLL_RX_DROP; + np->rx_flags |= NETPOLL_RX_DROP; atomic_inc(&trapped); np->dev->poll(np->dev, &budget); atomic_dec(&trapped); - np->dev->netpoll_rx &= ~NETPOLL_RX_DROP; + np->rx_flags &= ~NETPOLL_RX_DROP; } spin_unlock_irqrestore(&netpoll_poll_lock, flags); } @@ -279,18 +276,7 @@ int size, type = ARPOP_REPLY, ptype = ETH_P_ARP; u32 sip, tip; struct sk_buff *send_skb; - unsigned long flags; - struct list_head *p; - struct netpoll *np = NULL; - - spin_lock_irqsave(&rx_list_lock, flags); - list_for_each(p, &rx_list) { - np = list_entry(p, struct netpoll, rx_list); - if ( np->dev == skb->dev ) - break; - np = NULL; - } - spin_unlock_irqrestore(&rx_list_lock, flags); + struct netpoll *np = skb->dev->np; if (!np) return; @@ -373,10 +359,10 @@ int proto, len, ulen; struct iphdr *iph; struct udphdr *uh; - struct netpoll *np; - struct list_head *p; - unsigned long flags; + struct netpoll *np = skb->dev->np; + if (!np->rx_hook) + goto out; if (skb->dev->type != ARPHRD_ETHER) goto out; @@ -420,30 +406,19 @@ goto out; if (checksum_udp(skb, uh, ulen, iph->saddr, iph->daddr) < 0) goto out; + if (np->local_ip && np->local_ip != ntohl(iph->daddr)) + goto out; + if (np->remote_ip && np->remote_ip != ntohl(iph->saddr)) + goto out; + if (np->local_port && np->local_port != ntohs(uh->dest)) + goto out; - spin_lock_irqsave(&rx_list_lock, flags); - list_for_each(p, &rx_list) { - np = list_entry(p, struct netpoll, rx_list); - if (np->dev && np->dev != skb->dev) - continue; - if (np->local_ip && np->local_ip != ntohl(iph->daddr)) - continue; - if (np->remote_ip && np->remote_ip != ntohl(iph->saddr)) - continue; - if (np->local_port && np->local_port != ntohs(uh->dest)) - continue; - - spin_unlock_irqrestore(&rx_list_lock, flags); - - if (np->rx_hook) - np->rx_hook(np, ntohs(uh->source), - (char *)(uh+1), - ulen - sizeof(struct udphdr)); + np->rx_hook(np, ntohs(uh->source), + (char *)(uh+1), + ulen - sizeof(struct udphdr)); - kfree_skb(skb); - return 1; - } - spin_unlock_irqrestore(&rx_list_lock, flags); + kfree_skb(skb); + return 1; out: if (atomic_read(&trapped)) { @@ -574,6 +549,10 @@ np->name, np->dev_name); return -1; } + + np->dev = ndev; + ndev->np = np; + if (!ndev->poll_controller) { printk(KERN_ERR "%s: %s doesn't support polling, aborting.\n", np->name, np->dev_name); @@ -639,36 +618,22 @@ np->name, HIPQUAD(np->local_ip)); } - np->dev = ndev; - - if(np->rx_hook) { - unsigned long flags; - - np->dev->netpoll_rx = NETPOLL_RX_ENABLED; - - spin_lock_irqsave(&rx_list_lock, flags); - list_add(&np->rx_list, &rx_list); - spin_unlock_irqrestore(&rx_list_lock, flags); - } + if(np->rx_hook) + np->rx_flags = NETPOLL_RX_ENABLED; return 0; + release: + ndev->np = NULL; + np->dev = NULL; dev_put(ndev); return -1; } void netpoll_cleanup(struct netpoll *np) { - if (np->rx_hook) { - unsigned long flags; - - spin_lock_irqsave(&rx_list_lock, flags); - list_del(&np->rx_list); - spin_unlock_irqrestore(&rx_list_lock, flags); - } - if (np->dev) - np->dev->netpoll_rx = 0; + np->dev->np = NULL; dev_put(np->dev); np->dev = NULL; } Index: rc4/include/linux/netpoll.h =================================================================== --- rc4.orig/include/linux/netpoll.h 2005-02-17 22:32:19.000000000 -0600 +++ rc4/include/linux/netpoll.h 2005-02-17 22:39:59.000000000 -0600 @@ -16,11 +16,11 @@ struct netpoll { struct net_device *dev; char dev_name[16], *name; + int rx_flags; void (*rx_hook)(struct netpoll *, int, char *, int); u32 local_ip, remote_ip; u16 local_port, remote_port; unsigned char local_mac[6], remote_mac[6]; - struct list_head rx_list; }; void netpoll_poll(struct netpoll *np); @@ -35,7 +35,7 @@ #ifdef CONFIG_NETPOLL static inline int netpoll_rx(struct sk_buff *skb) { - return skb->dev->netpoll_rx && __netpoll_rx(skb); + return skb->dev->np && skb->dev->np->rx_flags && __netpoll_rx(skb); } #else #define netpoll_rx(a) 0 ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 4/6] netpoll: fix ->poll() locking 2005-02-18 5:25 ` [PATCH 3/6] netpoll: add netpoll point to net_device Matt Mackall @ 2005-02-18 5:25 ` Matt Mackall 2005-02-18 5:25 ` [PATCH 5/6] netpoll: add optional dropping and queueing support Matt Mackall 0 siblings, 1 reply; 8+ messages in thread From: Matt Mackall @ 2005-02-18 5:25 UTC (permalink / raw) To: David S. Miller; +Cc: netdev, Jeff Moyer Introduce a per-client poll lock and flag. The lock assures we never have more than one caller in dev->poll(). The flag provides recursion avoidance on UP where the lock disappears. Signed-off-by: Matt Mackall <mpm@selenic.com> Index: rc4/net/core/netpoll.c =================================================================== --- rc4.orig/net/core/netpoll.c 2005-02-17 22:39:59.000000000 -0600 +++ rc4/net/core/netpoll.c 2005-02-17 22:40:02.000000000 -0600 @@ -36,7 +36,6 @@ static struct sk_buff *skbs; static atomic_t trapped; -static DEFINE_SPINLOCK(netpoll_poll_lock); #define NETPOLL_RX_ENABLED 1 #define NETPOLL_RX_DROP 2 @@ -63,8 +62,15 @@ } /* - * Check whether delayed processing was scheduled for our current CPU, - * and then manually invoke NAPI polling to pump data off the card. + * Check whether delayed processing was scheduled for our NIC. If so, + * we attempt to grab the poll lock and use ->poll() to pump the card. + * If this fails, either we've recursed in ->poll() or it's already + * running on another CPU. + * + * Note: we don't mask interrupts with this lock because we're using + * trylock here and interrupts are already disabled in the softirq + * case. Further, we test the poll_owner to avoid recursion on UP + * systems where the lock doesn't exist. * * In cases where there is bi-directional communications, reading only * one message at a time can lead to packets being dropped by the @@ -74,13 +80,10 @@ static void poll_napi(struct netpoll *np) { int budget = 16; - unsigned long flags; - struct softnet_data *queue; - spin_lock_irqsave(&netpoll_poll_lock, flags); - queue = &__get_cpu_var(softnet_data); if (test_bit(__LINK_STATE_RX_SCHED, &np->dev->state) && - !list_empty(&queue->poll_list)) { + np->poll_owner != __smp_processor_id() && + spin_trylock(&np->poll_lock)) { np->rx_flags |= NETPOLL_RX_DROP; atomic_inc(&trapped); @@ -88,8 +91,8 @@ atomic_dec(&trapped); np->rx_flags &= ~NETPOLL_RX_DROP; + spin_unlock(&np->poll_lock); } - spin_unlock_irqrestore(&netpoll_poll_lock, flags); } void netpoll_poll(struct netpoll *np) @@ -194,6 +197,12 @@ return; } + /* avoid ->poll recursion */ + if(np->poll_owner == __smp_processor_id()) { + __kfree_skb(skb); + return; + } + spin_lock(&np->dev->xmit_lock); np->dev->xmit_lock_owner = smp_processor_id(); @@ -542,6 +551,9 @@ struct net_device *ndev = NULL; struct in_device *in_dev; + np->poll_lock = SPIN_LOCK_UNLOCKED; + np->poll_owner = -1; + if (np->dev_name) ndev = dev_get_by_name(np->dev_name); if (!ndev) { Index: rc4/include/linux/netpoll.h =================================================================== --- rc4.orig/include/linux/netpoll.h 2005-02-17 22:39:59.000000000 -0600 +++ rc4/include/linux/netpoll.h 2005-02-17 22:40:02.000000000 -0600 @@ -21,6 +21,8 @@ u32 local_ip, remote_ip; u16 local_port, remote_port; unsigned char local_mac[6], remote_mac[6]; + spinlock_t poll_lock; + int poll_owner; }; void netpoll_poll(struct netpoll *np); @@ -37,8 +39,27 @@ { return skb->dev->np && skb->dev->np->rx_flags && __netpoll_rx(skb); } + +static inline void netpoll_poll_lock(struct net_device *dev) +{ + if (dev->np) { + spin_lock(&dev->np->poll_lock); + dev->np->poll_owner = __smp_processor_id(); + } +} + +static inline void netpoll_poll_unlock(struct net_device *dev) +{ + if (dev->np) { + spin_unlock(&dev->np->poll_lock); + dev->np->poll_owner = -1; + } +} + #else #define netpoll_rx(a) 0 +#define netpoll_poll_lock(a) +#define netpoll_poll_unlock(a) #endif #endif Index: rc4/net/core/dev.c =================================================================== --- rc4.orig/net/core/dev.c 2005-02-17 22:39:59.000000000 -0600 +++ rc4/net/core/dev.c 2005-02-17 22:40:02.000000000 -0600 @@ -1775,8 +1775,10 @@ dev = list_entry(queue->poll_list.next, struct net_device, poll_list); + netpoll_poll_lock(dev); if (dev->quota <= 0 || dev->poll(dev, &budget)) { + netpoll_poll_unlock(dev); local_irq_disable(); list_del(&dev->poll_list); list_add_tail(&dev->poll_list, &queue->poll_list); @@ -1785,6 +1787,7 @@ else dev->quota = dev->weight; } else { + netpoll_poll_unlock(dev); dev_put(dev); local_irq_disable(); } ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 5/6] netpoll: add optional dropping and queueing support 2005-02-18 5:25 ` [PATCH 4/6] netpoll: fix ->poll() locking Matt Mackall @ 2005-02-18 5:25 ` Matt Mackall 2005-02-18 5:25 ` [PATCH 6/6] netpoll: handle xmit_lock recursion similarly Matt Mackall 0 siblings, 1 reply; 8+ messages in thread From: Matt Mackall @ 2005-02-18 5:25 UTC (permalink / raw) To: David S. Miller; +Cc: netdev, Jeff Moyer This adds a callback for packets we can't deliver immediately and a helper function for clients to queue such packets to the device post-interrupt. Netconsole is modified to use the queueing function for best-effort delivery. Signed-off-by: Matt Mackall <mpm@selenic.com> Index: rc4/drivers/net/netconsole.c =================================================================== --- rc4.orig/drivers/net/netconsole.c 2005-02-17 22:39:29.000000000 -0600 +++ rc4/drivers/net/netconsole.c 2005-02-17 22:40:05.000000000 -0600 @@ -60,6 +60,7 @@ .local_port = 6665, .remote_port = 6666, .remote_mac = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}, + .drop = netpoll_queue, }; static int configured = 0; Index: rc4/net/core/netpoll.c =================================================================== --- rc4.orig/net/core/netpoll.c 2005-02-17 22:40:02.000000000 -0600 +++ rc4/net/core/netpoll.c 2005-02-17 22:40:05.000000000 -0600 @@ -19,6 +19,7 @@ #include <linux/netpoll.h> #include <linux/sched.h> #include <linux/rcupdate.h> +#include <linux/workqueue.h> #include <net/tcp.h> #include <net/udp.h> #include <asm/unaligned.h> @@ -28,13 +29,18 @@ * message gets out even in extreme OOM situations. */ -#define MAX_SKBS 32 #define MAX_UDP_CHUNK 1460 +#define MAX_SKBS 32 +#define MAX_QUEUE_DEPTH (MAX_SKBS / 2) static DEFINE_SPINLOCK(skb_list_lock); static int nr_skbs; static struct sk_buff *skbs; +static DEFINE_SPINLOCK(queue_lock); +static int queue_depth; +static struct sk_buff *queue_head, *queue_tail; + static atomic_t trapped; #define NETPOLL_RX_ENABLED 1 @@ -46,6 +52,50 @@ static void zap_completion_queue(void); +static void queue_process(void *p) +{ + unsigned long flags; + struct sk_buff *skb; + + while (queue_head) { + spin_lock_irqsave(&queue_lock, flags); + + skb = queue_head; + queue_head = skb->next; + if (skb == queue_tail) + queue_head = NULL; + + queue_depth--; + + spin_unlock_irqrestore(&queue_lock, flags); + + dev_queue_xmit(skb); + } +} + +static DECLARE_WORK(send_queue, queue_process, NULL); + +void netpoll_queue(struct sk_buff *skb) +{ + unsigned long flags; + + if (queue_depth == MAX_QUEUE_DEPTH) { + __kfree_skb(skb); + return; + } + + spin_lock_irqsave(&queue_lock, flags); + if (!queue_head) + queue_head = skb; + else + queue_tail->next = skb; + queue_tail = skb; + queue_depth++; + spin_unlock_irqrestore(&queue_lock, flags); + + schedule_work(&send_queue); +} + static int checksum_udp(struct sk_buff *skb, struct udphdr *uh, unsigned short ulen, u32 saddr, u32 daddr) { @@ -199,7 +249,10 @@ /* avoid ->poll recursion */ if(np->poll_owner == __smp_processor_id()) { - __kfree_skb(skb); + if (np->drop) + np->drop(skb); + else + __kfree_skb(skb); return; } @@ -275,6 +328,8 @@ memcpy(eth->h_source, np->local_mac, 6); memcpy(eth->h_dest, np->remote_mac, 6); + skb->dev = np->dev; + netpoll_send_skb(np, skb); } Index: rc4/include/linux/netpoll.h =================================================================== --- rc4.orig/include/linux/netpoll.h 2005-02-17 22:40:02.000000000 -0600 +++ rc4/include/linux/netpoll.h 2005-02-17 22:40:05.000000000 -0600 @@ -18,6 +18,7 @@ char dev_name[16], *name; int rx_flags; void (*rx_hook)(struct netpoll *, int, char *, int); + void (*drop)(struct sk_buff *skb); u32 local_ip, remote_ip; u16 local_port, remote_port; unsigned char local_mac[6], remote_mac[6]; @@ -33,6 +34,7 @@ void netpoll_set_trap(int trap); void netpoll_cleanup(struct netpoll *np); int __netpoll_rx(struct sk_buff *skb); +void netpoll_queue(struct sk_buff *skb); #ifdef CONFIG_NETPOLL static inline int netpoll_rx(struct sk_buff *skb) ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 6/6] netpoll: handle xmit_lock recursion similarly 2005-02-18 5:25 ` [PATCH 5/6] netpoll: add optional dropping and queueing support Matt Mackall @ 2005-02-18 5:25 ` Matt Mackall 0 siblings, 0 replies; 8+ messages in thread From: Matt Mackall @ 2005-02-18 5:25 UTC (permalink / raw) To: David S. Miller; +Cc: netdev, Jeff Moyer Handle possible recursion on xmit_lock while we're at it. Signed-off-by: Matt Mackall <mpm@selenic.com> Index: rc4/net/core/netpoll.c =================================================================== --- rc4.orig/net/core/netpoll.c 2005-02-17 22:40:05.000000000 -0600 +++ rc4/net/core/netpoll.c 2005-02-17 22:40:07.000000000 -0600 @@ -247,8 +247,9 @@ return; } - /* avoid ->poll recursion */ - if(np->poll_owner == __smp_processor_id()) { + /* avoid recursion */ + if(np->poll_owner == __smp_processor_id() || + np->dev->xmit_lock_owner == __smp_processor_id()) { if (np->drop) np->drop(skb); else ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups 2005-02-18 5:25 [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups Matt Mackall 2005-02-18 5:25 ` [PATCH 1/6] netpoll: shorten carrier detect timeout Matt Mackall @ 2005-03-23 2:33 ` David S. Miller 1 sibling, 0 replies; 8+ messages in thread From: David S. Miller @ 2005-03-23 2:33 UTC (permalink / raw) To: Matt Mackall; +Cc: netdev, jmoyer On Thu, 17 Feb 2005 23:25:18 -0600 Matt Mackall <mpm@selenic.com> wrote: > This patch series against -rc4 fixes up some recursion deadlocks in > netpoll and adds support for fallback to queueing. Various cleanups > along the way. > > Holds up under load testing via ipt_LOG on a dual Opteron with tg3. I've applied all 6 patches to my tree. Sorry for taking so long Matt. ^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2005-03-23 2:33 UTC | newest] Thread overview: 8+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2005-02-18 5:25 [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups Matt Mackall 2005-02-18 5:25 ` [PATCH 1/6] netpoll: shorten carrier detect timeout Matt Mackall 2005-02-18 5:25 ` [PATCH 2/6] netpoll: filter inlines Matt Mackall 2005-02-18 5:25 ` [PATCH 3/6] netpoll: add netpoll point to net_device Matt Mackall 2005-02-18 5:25 ` [PATCH 4/6] netpoll: fix ->poll() locking Matt Mackall 2005-02-18 5:25 ` [PATCH 5/6] netpoll: add optional dropping and queueing support Matt Mackall 2005-02-18 5:25 ` [PATCH 6/6] netpoll: handle xmit_lock recursion similarly Matt Mackall 2005-03-23 2:33 ` [PATCH 0/6] netpoll: recursion fixes, queueing, and cleanups David S. Miller
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).