kernel-ark/include/linux/skbuff.h

2232 lines
61 KiB
C
Raw Normal View History

/*
* Definitions for the 'struct sk_buff' memory handlers.
*
* Authors:
* Alan Cox, <gw4pts@gw4pts.ampr.org>
* Florian La Roche, <rzsfl@rz.uni-sb.de>
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#ifndef _LINUX_SKBUFF_H
#define _LINUX_SKBUFF_H
#include <linux/kernel.h>
#include <linux/kmemcheck.h>
#include <linux/compiler.h>
#include <linux/time.h>
#include <linux/cache.h>
#include <asm/atomic.h>
#include <asm/types.h>
#include <linux/spinlock.h>
#include <linux/net.h>
#include <linux/textsearch.h>
#include <net/checksum.h>
#include <linux/rcupdate.h>
#include <linux/dmaengine.h>
#include <linux/hrtimer.h>
/* Don't change this without changing skb_csum_unnecessary! */
#define CHECKSUM_NONE 0
#define CHECKSUM_UNNECESSARY 1
#define CHECKSUM_COMPLETE 2
#define CHECKSUM_PARTIAL 3
#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
~(SMP_CACHE_BYTES - 1))
#define SKB_WITH_OVERHEAD(X) \
((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
#define SKB_MAX_ORDER(X, ORDER) \
SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
/* A. Checksumming of received packets by device.
*
* NONE: device failed to checksum this packet.
* skb->csum is undefined.
*
* UNNECESSARY: device parsed packet and wouldbe verified checksum.
* skb->csum is undefined.
* It is bad option, but, unfortunately, many of vendors do this.
* Apparently with secret goal to sell you new device, when you
* will add new protocol to your host. F.e. IPv6. 8)
*
* COMPLETE: the most generic way. Device supplied checksum of _all_
* the packet as seen by netif_rx in skb->csum.
* NOTE: Even if device supports only some protocols, but
* is able to produce some skb->csum, it MUST use COMPLETE,
* not UNNECESSARY.
*
* PARTIAL: identical to the case for output below. This may occur
* on a packet received directly from another Linux OS, e.g.,
* a virtualised Linux kernel on the same host. The packet can
* be treated in the same way as UNNECESSARY except that on
* output (i.e., forwarding) the checksum must be filled in
* by the OS or the hardware.
*
* B. Checksumming on output.
*
* NONE: skb is checksummed by protocol or csum is not required.
*
* PARTIAL: device is required to csum packet as seen by hard_start_xmit
* from skb->csum_start to the end and to record the checksum
* at skb->csum_start + skb->csum_offset.
*
* Device must show its capabilities in dev->features, set
* at device setup time.
* NETIF_F_HW_CSUM - it is clever device, it is able to checksum
* everything.
* NETIF_F_NO_CSUM - loopback or reliable single hop media.
* NETIF_F_IP_CSUM - device is dumb. It is able to csum only
* TCP/UDP over IPv4. Sigh. Vendors like this
* way by an unknown reason. Though, see comment above
* about CHECKSUM_UNNECESSARY. 8)
* NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
*
* Any questions? No questions, good. --ANK
*/
struct net_device;
struct scatterlist;
struct pipe_inode_info;
#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
struct nf_conntrack {
atomic_t use;
};
#endif
#ifdef CONFIG_BRIDGE_NETFILTER
struct nf_bridge_info {
atomic_t use;
struct net_device *physindev;
struct net_device *physoutdev;
unsigned int mask;
unsigned long data[32 / sizeof(unsigned long)];
};
#endif
struct sk_buff_head {
/* These two members must be first. */
struct sk_buff *next;
struct sk_buff *prev;
__u32 qlen;
spinlock_t lock;
};
struct sk_buff;
/* To allow 64K frame to be packed as single skb without frag_list */
#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
typedef struct skb_frag_struct skb_frag_t;
struct skb_frag_struct {
struct page *page;
#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
__u32 page_offset;
__u32 size;
#else
__u16 page_offset;
__u16 size;
#endif
};
#define HAVE_HW_TIME_STAMP
/**
* struct skb_shared_hwtstamps - hardware time stamps
* @hwtstamp: hardware time stamp transformed into duration
* since arbitrary point in time
* @syststamp: hwtstamp transformed to system time base
*
* Software time stamps generated by ktime_get_real() are stored in
* skb->tstamp. The relation between the different kinds of time
* stamps is as follows:
*
* syststamp and tstamp can be compared against each other in
* arbitrary combinations. The accuracy of a
* syststamp/tstamp/"syststamp from other device" comparison is
* limited by the accuracy of the transformation into system time
* base. This depends on the device driver and its underlying
* hardware.
*
* hwtstamps can only be compared against other hwtstamps from
* the same device.
*
* This structure is attached to packets as part of the
* &skb_shared_info. Use skb_hwtstamps() to get a pointer.
*/
struct skb_shared_hwtstamps {
ktime_t hwtstamp;
ktime_t syststamp;
};
/* Definitions for tx_flags in struct skb_shared_info */
enum {
/* generate hardware time stamp */
SKBTX_HW_TSTAMP = 1 << 0,
/* generate software time stamp */
SKBTX_SW_TSTAMP = 1 << 1,
/* device driver is going to provide hardware time stamp */
SKBTX_IN_PROGRESS = 1 << 2,
/* ensure the originating sk reference is available on driver level */
SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
};
/* This data is invariant across clones and lives at
* the end of the header data, ie. at skb->end.
*/
struct skb_shared_info {
unsigned short nr_frags;
unsigned short gso_size;
/* Warning: this field is not always filled in (UFO)! */
unsigned short gso_segs;
unsigned short gso_type;
__be32 ip6_frag_id;
__u8 tx_flags;
struct sk_buff *frag_list;
struct skb_shared_hwtstamps hwtstamps;
/*
* Warning : all fields before dataref are cleared in __alloc_skb()
*/
atomic_t dataref;
/* Intermediate layers must ensure that destructor_arg
* remains valid until skb destructor */
void * destructor_arg;
/* must be last field, see pskb_expand_head() */
skb_frag_t frags[MAX_SKB_FRAGS];
};
/* We divide dataref into two halves. The higher 16 bits hold references
* to the payload part of skb->data. The lower 16 bits hold references to
[SKBUFF]: Keep track of writable header len of headerless clones Currently NAT (and others) that want to modify cloned skbs copy them, even if in the vast majority of cases its not necessary because the skb is a clone made by TCP and the portion NAT wants to modify is actually writable because TCP release the header reference before cloning. The problem is that there is no clean way for NAT to find out how long the writable header area is, so this patch introduces skb->hdr_len to hold this length. When a headerless skb is cloned skb->hdr_len is set to the current headroom, for regular clones it is copied from the original. A new function skb_clone_writable(skb, len) returns whether the skb is writable up to len bytes from skb->data. To avoid enlarging the skb the mac_len field is reduced to 16 bit and the new hdr_len field is put in the remaining 16 bit. I've done a few rough benchmarks of NAT (not with this exact patch, but a very similar one). As expected it saves huge amounts of system time in case of sendfile, bringing it down to basically the same amount as without NAT, with sendmsg it only helps on loopback, probably because of the large MTU. Transmit a 1GB file using sendfile/sendmsg over eth0/lo with and without NAT: - sendfile eth0, no NAT: sys 0m0.388s - sendfile eth0, NAT: sys 0m1.835s - sendfile eth0: NAT + path: sys 0m0.370s (~ -80%) - sendfile lo, no NAT: sys 0m0.258s - sendfile lo, NAT: sys 0m2.609s - sendfile lo, NAT + patch: sys 0m0.260s (~ -90%) - sendmsg eth0, no NAT: sys 0m2.508s - sendmsg eth0, NAT: sys 0m2.539s - sendmsg eth0, NAT + patch: sys 0m2.445s (no change) - sendmsg lo, no NAT: sys 0m2.151s - sendmsg lo, NAT: sys 0m3.557s - sendmsg lo, NAT + patch: sys 0m2.159s (~ -40%) I expect other users can see a similar performance improvement, packet mangling iptables targets, ipip and ip_gre come to mind .. Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-25 11:35:20 +00:00
* the entire skb->data. A clone of a headerless skb holds the length of
* the header in skb->hdr_len.
*
* All users must obey the rule that the skb->data reference count must be
* greater than or equal to the payload reference count.
*
* Holding a reference to the payload part means that the user does not
* care about modifications to the header part of skb->data.
*/
#define SKB_DATAREF_SHIFT 16
#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
enum {
SKB_FCLONE_UNAVAILABLE,
SKB_FCLONE_ORIG,
SKB_FCLONE_CLONE,
};
enum {
SKB_GSO_TCPV4 = 1 << 0,
[IPV6]: Added GSO support for TCPv6 This patch adds GSO support for IPv6 and TCPv6. This is based on a patch by Ananda Raju <Ananda.Raju@neterion.com>. His original description is: This patch enables TSO over IPv6. Currently Linux network stacks restricts TSO over IPv6 by clearing of the NETIF_F_TSO bit from "dev->features". This patch will remove this restriction. This patch will introduce a new flag NETIF_F_TSO6 which will be used to check whether device supports TSO over IPv6. If device support TSO over IPv6 then we don't clear of NETIF_F_TSO and which will make the TCP layer to create TSO packets. Any device supporting TSO over IPv6 will set NETIF_F_TSO6 flag in "dev->features" along with NETIF_F_TSO. In case when user disables TSO using ethtool, NETIF_F_TSO will get cleared from "dev->features". So even if we have NETIF_F_TSO6 we don't get TSO packets created by TCP layer. SKB_GSO_TCPV4 renamed to SKB_GSO_TCP to make it generic GSO packet. SKB_GSO_UDPV4 renamed to SKB_GSO_UDP as UFO is not a IPv4 feature. UFO is supported over IPv6 also The following table shows there is significant improvement in throughput with normal frames and CPU usage for both normal and jumbo. -------------------------------------------------- | | 1500 | 9600 | | ------------------|-------------------| | | thru CPU | thru CPU | -------------------------------------------------- | TSO OFF | 2.00 5.5% id | 5.66 20.0% id | -------------------------------------------------- | TSO ON | 2.63 78.0 id | 5.67 39.0% id | -------------------------------------------------- Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2006-06-30 20:37:03 +00:00
SKB_GSO_UDP = 1 << 1,
/* This indicates the skb is from an untrusted source. */
SKB_GSO_DODGY = 1 << 2,
/* This indicates the tcp segment has CWR set. */
[IPV6]: Added GSO support for TCPv6 This patch adds GSO support for IPv6 and TCPv6. This is based on a patch by Ananda Raju <Ananda.Raju@neterion.com>. His original description is: This patch enables TSO over IPv6. Currently Linux network stacks restricts TSO over IPv6 by clearing of the NETIF_F_TSO bit from "dev->features". This patch will remove this restriction. This patch will introduce a new flag NETIF_F_TSO6 which will be used to check whether device supports TSO over IPv6. If device support TSO over IPv6 then we don't clear of NETIF_F_TSO and which will make the TCP layer to create TSO packets. Any device supporting TSO over IPv6 will set NETIF_F_TSO6 flag in "dev->features" along with NETIF_F_TSO. In case when user disables TSO using ethtool, NETIF_F_TSO will get cleared from "dev->features". So even if we have NETIF_F_TSO6 we don't get TSO packets created by TCP layer. SKB_GSO_TCPV4 renamed to SKB_GSO_TCP to make it generic GSO packet. SKB_GSO_UDPV4 renamed to SKB_GSO_UDP as UFO is not a IPv4 feature. UFO is supported over IPv6 also The following table shows there is significant improvement in throughput with normal frames and CPU usage for both normal and jumbo. -------------------------------------------------- | | 1500 | 9600 | | ------------------|-------------------| | | thru CPU | thru CPU | -------------------------------------------------- | TSO OFF | 2.00 5.5% id | 5.66 20.0% id | -------------------------------------------------- | TSO ON | 2.63 78.0 id | 5.67 39.0% id | -------------------------------------------------- Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2006-06-30 20:37:03 +00:00
SKB_GSO_TCP_ECN = 1 << 3,
SKB_GSO_TCPV6 = 1 << 4,
SKB_GSO_FCOE = 1 << 5,
};
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
#if BITS_PER_LONG > 32
#define NET_SKBUFF_DATA_USES_OFFSET 1
#endif
#ifdef NET_SKBUFF_DATA_USES_OFFSET
typedef unsigned int sk_buff_data_t;
#else
typedef unsigned char *sk_buff_data_t;
#endif
/**
* struct sk_buff - socket buffer
* @next: Next buffer in list
* @prev: Previous buffer in list
* @sk: Socket we are owned by
* @tstamp: Time we arrived
* @dev: Device we arrived on/are leaving by
* @transport_header: Transport layer header
* @network_header: Network layer header
* @mac_header: Link layer header
* @_skb_refdst: destination entry (with norefcount bit)
* @sp: the security path, used for xfrm
* @cb: Control buffer. Free for use by every layer. Put private vars here
* @len: Length of actual data
* @data_len: Data length
* @mac_len: Length of link layer header
[SKBUFF]: Keep track of writable header len of headerless clones Currently NAT (and others) that want to modify cloned skbs copy them, even if in the vast majority of cases its not necessary because the skb is a clone made by TCP and the portion NAT wants to modify is actually writable because TCP release the header reference before cloning. The problem is that there is no clean way for NAT to find out how long the writable header area is, so this patch introduces skb->hdr_len to hold this length. When a headerless skb is cloned skb->hdr_len is set to the current headroom, for regular clones it is copied from the original. A new function skb_clone_writable(skb, len) returns whether the skb is writable up to len bytes from skb->data. To avoid enlarging the skb the mac_len field is reduced to 16 bit and the new hdr_len field is put in the remaining 16 bit. I've done a few rough benchmarks of NAT (not with this exact patch, but a very similar one). As expected it saves huge amounts of system time in case of sendfile, bringing it down to basically the same amount as without NAT, with sendmsg it only helps on loopback, probably because of the large MTU. Transmit a 1GB file using sendfile/sendmsg over eth0/lo with and without NAT: - sendfile eth0, no NAT: sys 0m0.388s - sendfile eth0, NAT: sys 0m1.835s - sendfile eth0: NAT + path: sys 0m0.370s (~ -80%) - sendfile lo, no NAT: sys 0m0.258s - sendfile lo, NAT: sys 0m2.609s - sendfile lo, NAT + patch: sys 0m0.260s (~ -90%) - sendmsg eth0, no NAT: sys 0m2.508s - sendmsg eth0, NAT: sys 0m2.539s - sendmsg eth0, NAT + patch: sys 0m2.445s (no change) - sendmsg lo, no NAT: sys 0m2.151s - sendmsg lo, NAT: sys 0m3.557s - sendmsg lo, NAT + patch: sys 0m2.159s (~ -40%) I expect other users can see a similar performance improvement, packet mangling iptables targets, ipip and ip_gre come to mind .. Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-25 11:35:20 +00:00
* @hdr_len: writable header length of cloned skb
* @csum: Checksum (must include start/offset pair)
* @csum_start: Offset from skb->head where checksumming should start
* @csum_offset: Offset from csum_start where checksum should be stored
* @local_df: allow local fragmentation
* @cloned: Head may be cloned (check refcnt to be sure)
* @nohdr: Payload reference only, must not modify header
* @pkt_type: Packet class
* @fclone: skbuff clone status
* @ip_summed: Driver fed us an IP checksum
* @priority: Packet queueing priority
* @users: User count - see {datagram,tcp}.c
* @protocol: Packet protocol from driver
* @truesize: Buffer size
* @head: Head of buffer
* @data: Data head pointer
* @tail: Tail pointer
* @end: End pointer
* @destructor: Destruct function
* @mark: Generic packet mark
* @nfct: Associated connection, if any
* @ipvs_property: skbuff is owned by ipvs
* @peeked: this packet has been seen already, so stats have been
* done for it, don't do them again
* @nf_trace: netfilter packet trace flag
* @nfctinfo: Relationship of this skb to the connection
* @nfct_reasm: netfilter conntrack re-assembly pointer
* @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
* @skb_iif: ifindex of device we arrived on
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
* @rxhash: the packet hash computed on receive
* @queue_mapping: Queue mapping for multiqueue devices
* @tc_index: Traffic control index
* @tc_verd: traffic control verdict
* @ndisc_nodetype: router type (from link layer)
* @dma_cookie: a cookie to one of several possible DMA operations
* done by skb DMA functions
* @secmark: security marking
* @vlan_tci: vlan tag control information
*/
struct sk_buff {
/* These two members must be first. */
struct sk_buff *next;
struct sk_buff *prev;
ktime_t tstamp;
struct sock *sk;
struct net_device *dev;
/*
* This is the control buffer. It is free to use for every
* layer. Please put your private variables there. If you
* want to keep them across layers you have to do a skb_clone()
* first. This is owned by whoever has the skb queued ATM.
*/
char cb[48] __aligned(8);
unsigned long _skb_refdst;
#ifdef CONFIG_XFRM
struct sec_path *sp;
#endif
unsigned int len,
[SKBUFF]: Keep track of writable header len of headerless clones Currently NAT (and others) that want to modify cloned skbs copy them, even if in the vast majority of cases its not necessary because the skb is a clone made by TCP and the portion NAT wants to modify is actually writable because TCP release the header reference before cloning. The problem is that there is no clean way for NAT to find out how long the writable header area is, so this patch introduces skb->hdr_len to hold this length. When a headerless skb is cloned skb->hdr_len is set to the current headroom, for regular clones it is copied from the original. A new function skb_clone_writable(skb, len) returns whether the skb is writable up to len bytes from skb->data. To avoid enlarging the skb the mac_len field is reduced to 16 bit and the new hdr_len field is put in the remaining 16 bit. I've done a few rough benchmarks of NAT (not with this exact patch, but a very similar one). As expected it saves huge amounts of system time in case of sendfile, bringing it down to basically the same amount as without NAT, with sendmsg it only helps on loopback, probably because of the large MTU. Transmit a 1GB file using sendfile/sendmsg over eth0/lo with and without NAT: - sendfile eth0, no NAT: sys 0m0.388s - sendfile eth0, NAT: sys 0m1.835s - sendfile eth0: NAT + path: sys 0m0.370s (~ -80%) - sendfile lo, no NAT: sys 0m0.258s - sendfile lo, NAT: sys 0m2.609s - sendfile lo, NAT + patch: sys 0m0.260s (~ -90%) - sendmsg eth0, no NAT: sys 0m2.508s - sendmsg eth0, NAT: sys 0m2.539s - sendmsg eth0, NAT + patch: sys 0m2.445s (no change) - sendmsg lo, no NAT: sys 0m2.151s - sendmsg lo, NAT: sys 0m3.557s - sendmsg lo, NAT + patch: sys 0m2.159s (~ -40%) I expect other users can see a similar performance improvement, packet mangling iptables targets, ipip and ip_gre come to mind .. Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-25 11:35:20 +00:00
data_len;
__u16 mac_len,
hdr_len;
union {
__wsum csum;
struct {
__u16 csum_start;
__u16 csum_offset;
};
};
__u32 priority;
kmemcheck_bitfield_begin(flags1);
__u8 local_df:1,
cloned:1,
ip_summed:2,
nohdr:1,
nfctinfo:3;
__u8 pkt_type:3,
fclone:2,
ipvs_property:1,
peeked:1,
nf_trace:1;
kmemcheck_bitfield_end(flags1);
__be16 protocol;
void (*destructor)(struct sk_buff *skb);
[NETFILTER]: Add nf_conntrack subsystem. The existing connection tracking subsystem in netfilter can only handle ipv4. There were basically two choices present to add connection tracking support for ipv6. We could either duplicate all of the ipv4 connection tracking code into an ipv6 counterpart, or (the choice taken by these patches) we could design a generic layer that could handle both ipv4 and ipv6 and thus requiring only one sub-protocol (TCP, UDP, etc.) connection tracking helper module to be written. In fact nf_conntrack is capable of working with any layer 3 protocol. The existing ipv4 specific conntrack code could also not deal with the pecularities of doing connection tracking on ipv6, which is also cured here. For example, these issues include: 1) ICMPv6 handling, which is used for neighbour discovery in ipv6 thus some messages such as these should not participate in connection tracking since effectively they are like ARP messages 2) fragmentation must be handled differently in ipv6, because the simplistic "defrag, connection track and NAT, refrag" (which the existing ipv4 connection tracking does) approach simply isn't feasible in ipv6 3) ipv6 extension header parsing must occur at the correct spots before and after connection tracking decisions, and there were no provisions for this in the existing connection tracking design 4) ipv6 has no need for stateful NAT The ipv4 specific conntrack layer is kept around, until all of the ipv4 specific conntrack helpers are ported over to nf_conntrack and it is feature complete. Once that occurs, the old conntrack stuff will get placed into the feature-removal-schedule and we will fully kill it off 6 months later. Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp> Signed-off-by: Harald Welte <laforge@netfilter.org> Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-11-10 00:38:16 +00:00
#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
struct nf_conntrack *nfct;
[NETFILTER]: Add nf_conntrack subsystem. The existing connection tracking subsystem in netfilter can only handle ipv4. There were basically two choices present to add connection tracking support for ipv6. We could either duplicate all of the ipv4 connection tracking code into an ipv6 counterpart, or (the choice taken by these patches) we could design a generic layer that could handle both ipv4 and ipv6 and thus requiring only one sub-protocol (TCP, UDP, etc.) connection tracking helper module to be written. In fact nf_conntrack is capable of working with any layer 3 protocol. The existing ipv4 specific conntrack code could also not deal with the pecularities of doing connection tracking on ipv6, which is also cured here. For example, these issues include: 1) ICMPv6 handling, which is used for neighbour discovery in ipv6 thus some messages such as these should not participate in connection tracking since effectively they are like ARP messages 2) fragmentation must be handled differently in ipv6, because the simplistic "defrag, connection track and NAT, refrag" (which the existing ipv4 connection tracking does) approach simply isn't feasible in ipv6 3) ipv6 extension header parsing must occur at the correct spots before and after connection tracking decisions, and there were no provisions for this in the existing connection tracking design 4) ipv6 has no need for stateful NAT The ipv4 specific conntrack layer is kept around, until all of the ipv4 specific conntrack helpers are ported over to nf_conntrack and it is feature complete. Once that occurs, the old conntrack stuff will get placed into the feature-removal-schedule and we will fully kill it off 6 months later. Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp> Signed-off-by: Harald Welte <laforge@netfilter.org> Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-11-10 00:38:16 +00:00
struct sk_buff *nfct_reasm;
#endif
#ifdef CONFIG_BRIDGE_NETFILTER
struct nf_bridge_info *nf_bridge;
#endif
int skb_iif;
#ifdef CONFIG_NET_SCHED
__u16 tc_index; /* traffic control index */
#ifdef CONFIG_NET_CLS_ACT
__u16 tc_verd; /* traffic control verdict */
#endif
#endif
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
__u32 rxhash;
kmemcheck_bitfield_begin(flags2);
__u16 queue_mapping:16;
#ifdef CONFIG_IPV6_NDISC_NODETYPE
net: deliver skbs on inactive slaves to exact matches Currently, the accelerated receive path for VLAN's will drop packets if the real device is an inactive slave and is not one of the special pkts tested for in skb_bond_should_drop(). This behavior is different then the non-accelerated path and for pkts over a bonded vlan. For example, vlanx -> bond0 -> ethx will be dropped in the vlan path and not delivered to any packet handlers at all. However, bond0 -> vlanx -> ethx and bond0 -> ethx will be delivered to handlers that match the exact dev, because the VLAN path checks the real_dev which is not a slave and netif_recv_skb() doesn't drop frames but only delivers them to exact matches. This patch adds a sk_buff flag which is used for tagging skbs that would previously been dropped and allows the skb to continue to skb_netif_recv(). Here we add logic to check for the deliver_no_wcard flag and if it is set only deliver to handlers that match exactly. This makes both paths above consistent and gives pkt handlers a way to identify skbs that come from inactive slaves. Without this patch in some configurations skbs will be delivered to handlers with exact matches and in others be dropped out right in the vlan path. I have tested the following 4 configurations in failover modes and load balancing modes. # bond0 -> ethx # vlanx -> bond0 -> ethx # bond0 -> vlanx -> ethx # bond0 -> ethx | vlanx -> -- Signed-off-by: John Fastabend <john.r.fastabend@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-06-03 09:30:11 +00:00
__u8 ndisc_nodetype:2,
deliver_no_wcard:1;
#else
__u8 deliver_no_wcard:1;
#endif
kmemcheck_bitfield_end(flags2);
/* 0/14 bit hole */
#ifdef CONFIG_NET_DMA
dma_cookie_t dma_cookie;
#endif
#ifdef CONFIG_NETWORK_SECMARK
__u32 secmark;
#endif
net: Generalize socket rx gap / receive queue overflow cmsg Create a new socket level option to report number of queue overflows Recently I augmented the AF_PACKET protocol to report the number of frames lost on the socket receive queue between any two enqueued frames. This value was exported via a SOL_PACKET level cmsg. AFter I completed that work it was requested that this feature be generalized so that any datagram oriented socket could make use of this option. As such I've created this patch, It creates a new SOL_SOCKET level option called SO_RXQ_OVFL, which when enabled exports a SOL_SOCKET level cmsg that reports the nubmer of times the sk_receive_queue overflowed between any two given frames. It also augments the AF_PACKET protocol to take advantage of this new feature (as it previously did not touch sk->sk_drops, which this patch uses to record the overflow count). Tested successfully by me. Notes: 1) Unlike my previous patch, this patch simply records the sk_drops value, which is not a number of drops between packets, but rather a total number of drops. Deltas must be computed in user space. 2) While this patch currently works with datagram oriented protocols, it will also be accepted by non-datagram oriented protocols. I'm not sure if thats agreeable to everyone, but my argument in favor of doing so is that, for those protocols which aren't applicable to this option, sk_drops will always be zero, and reporting no drops on a receive queue that isn't used for those non-participating protocols seems reasonable to me. This also saves us having to code in a per-protocol opt in mechanism. 3) This applies cleanly to net-next assuming that commit 977750076d98c7ff6cbda51858bb5a5894a9d9ab (my af packet cmsg patch) is reverted Signed-off-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-10-12 20:26:31 +00:00
union {
__u32 mark;
__u32 dropcount;
};
__u16 vlan_tci;
sk_buff_data_t transport_header;
sk_buff_data_t network_header;
sk_buff_data_t mac_header;
/* These elements must be at the end, see alloc_skb() for details. */
sk_buff_data_t tail;
sk_buff_data_t end;
unsigned char *head,
*data;
unsigned int truesize;
atomic_t users;
};
#ifdef __KERNEL__
/*
* Handling routines are only of interest to the kernel
*/
#include <linux/slab.h>
#include <asm/system.h>
/*
* skb might have a dst pointer attached, refcounted or not.
* _skb_refdst low order bit is set if refcount was _not_ taken
*/
#define SKB_DST_NOREF 1UL
#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
/**
* skb_dst - returns skb dst_entry
* @skb: buffer
*
* Returns skb dst_entry, regardless of reference taken or not.
*/
static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
{
/* If refdst was not refcounted, check we still are in a
* rcu_read_lock section
*/
WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
!rcu_read_lock_held() &&
!rcu_read_lock_bh_held());
return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
}
/**
* skb_dst_set - sets skb dst
* @skb: buffer
* @dst: dst entry
*
* Sets skb dst, assuming a reference was taken on dst and should
* be released by skb_dst_drop()
*/
static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
{
skb->_skb_refdst = (unsigned long)dst;
}
/**
* skb_dst_set_noref - sets skb dst, without a reference
* @skb: buffer
* @dst: dst entry
*
* Sets skb dst, assuming a reference was not taken on dst
* skb_dst_drop() should not dst_release() this dst
*/
static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
{
WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
}
/**
* skb_dst_is_noref - Test if skb dst isnt refcounted
* @skb: buffer
*/
static inline bool skb_dst_is_noref(const struct sk_buff *skb)
{
return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
}
static inline struct rtable *skb_rtable(const struct sk_buff *skb)
{
return (struct rtable *)skb_dst(skb);
}
extern void kfree_skb(struct sk_buff *skb);
extern void consume_skb(struct sk_buff *skb);
extern void __kfree_skb(struct sk_buff *skb);
extern struct sk_buff *__alloc_skb(unsigned int size,
gfp_t priority, int fclone, int node);
static inline struct sk_buff *alloc_skb(unsigned int size,
gfp_t priority)
{
return __alloc_skb(size, priority, 0, -1);
}
static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
gfp_t priority)
{
return __alloc_skb(size, priority, 1, -1);
}
extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
extern struct sk_buff *skb_clone(struct sk_buff *skb,
gfp_t priority);
extern struct sk_buff *skb_copy(const struct sk_buff *skb,
gfp_t priority);
extern struct sk_buff *pskb_copy(struct sk_buff *skb,
gfp_t gfp_mask);
extern int pskb_expand_head(struct sk_buff *skb,
int nhead, int ntail,
gfp_t gfp_mask);
extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
unsigned int headroom);
extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
int newheadroom, int newtailroom,
gfp_t priority);
extern int skb_to_sgvec(struct sk_buff *skb,
struct scatterlist *sg, int offset,
int len);
extern int skb_cow_data(struct sk_buff *skb, int tailbits,
struct sk_buff **trailer);
extern int skb_pad(struct sk_buff *skb, int pad);
#define dev_kfree_skb(a) consume_skb(a)
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-18 22:46:41 +00:00
extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
int getfrag(void *from, char *to, int offset,
int len,int odd, struct sk_buff *skb),
void *from, int length);
struct skb_seq_state {
__u32 lower_offset;
__u32 upper_offset;
__u32 frag_idx;
__u32 stepped_offset;
struct sk_buff *root_skb;
struct sk_buff *cur_skb;
__u8 *frag_data;
};
extern void skb_prepare_seq_read(struct sk_buff *skb,
unsigned int from, unsigned int to,
struct skb_seq_state *st);
extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
struct skb_seq_state *st);
extern void skb_abort_seq_read(struct skb_seq_state *st);
extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
unsigned int to, struct ts_config *config,
struct ts_state *state);
extern __u32 __skb_get_rxhash(struct sk_buff *skb);
static inline __u32 skb_get_rxhash(struct sk_buff *skb)
{
if (!skb->rxhash)
skb->rxhash = __skb_get_rxhash(skb);
return skb->rxhash;
}
#ifdef NET_SKBUFF_DATA_USES_OFFSET
static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
{
return skb->head + skb->end;
}
#else
static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
{
return skb->end;
}
#endif
/* Internal */
#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
{
return &skb_shinfo(skb)->hwtstamps;
}
/**
* skb_queue_empty - check if a queue is empty
* @list: queue head
*
* Returns true if the queue is empty, false otherwise.
*/
static inline int skb_queue_empty(const struct sk_buff_head *list)
{
return list->next == (struct sk_buff *)list;
}
/**
* skb_queue_is_last - check if skb is the last entry in the queue
* @list: queue head
* @skb: buffer
*
* Returns true if @skb is the last buffer on the list.
*/
static inline bool skb_queue_is_last(const struct sk_buff_head *list,
const struct sk_buff *skb)
{
return skb->next == (struct sk_buff *)list;
}
tcp: Try to restore large SKBs while SACK processing During SACK processing, most of the benefits of TSO are eaten by the SACK blocks that one-by-one fragment SKBs to MSS sized chunks. Then we're in problems when cleanup work for them has to be done when a large cumulative ACK comes. Try to return back to pre-split state already while more and more SACK info gets discovered by combining newly discovered SACK areas with the previous skb if that's SACKed as well. This approach has a number of benefits: 1) The processing overhead is spread more equally over the RTT 2) Write queue has less skbs to process (affect everything which has to walk in the queue past the sacked areas) 3) Write queue is consistent whole the time, so no other parts of TCP has to be aware of this (this was not the case with some other approach that was, well, quite intrusive all around). 4) Clean_rtx_queue can release most of the pages using single put_page instead of previous PAGE_SIZE/mss+1 calls In case a hole is fully filled by the new SACK block, we attempt to combine the next skb too which allows construction of skbs that are even larger than what tso split them to and it handles hole per on every nth patterns that often occur during slow start overshoot pretty nicely. Though this to be really useful also a retransmission would have to get lost since cumulative ACKs advance one hole at a time in the most typical case. TODO: handle upwards only merging. That should be rather easy when segment is fully sacked but I'm leaving that as future work item (it won't make very large difference anyway since this current approach already covers quite a lot of normal cases). I was earlier thinking of some sophisticated way of tracking timestamps of the first and the last segment but later on realized that it won't be that necessary at all to store the timestamp of the last segment. The cases that can occur are basically either: 1) ambiguous => no sensible measurement can be taken anyway 2) non-ambiguous is due to reordering => having the timestamp of the last segment there is just skewing things more off than does some good since the ack got triggered by one of the holes (besides some substle issues that would make determining right hole/skb even harder problem). Anyway, it has nothing to do with this change then. I choose to route some abnormal looking cases with goto noop, some could be handled differently (eg., by stopping the walking at that skb but again). In general, they either shouldn't happen at all or are rare enough to make no difference in practice. In theory this change (as whole) could cause some macroscale regression (global) because of cache misses that are taken over the round-trip time but it gets very likely better because of much less (local) cache misses per other write queue walkers and the big recovery clearing cumulative ack. Worth to note that these benefits would be very easy to get also without TSO/GSO being on as long as the data is in pages so that we can merge them. Currently I won't let that happen because DSACK splitting at fragment that would mess up pcounts due to sk_can_gso in tcp_set_skb_tso_segs. Once DSACKs fragments gets avoided, we have some conditions that can be made less strict. TODO: I will probably have to convert the excessive pointer passing to struct sacktag_state... :-) My testing revealed that considerable amount of skbs couldn't be shifted because they were cloned (most likely still awaiting tx reclaim)... [The rest is considering future work instead since I got repeatably EFAULT to tcpdump's recvfrom when I added pskb_expand_head to deal with clones, so I separated that into another, later patch] ...To counter that, I gave up on the fifth advantage: 5) When growing previous SACK block, less allocs for new skbs are done, basically a new alloc is needed only when new hole is detected and when the previous skb runs out of frags space ...which now only happens of if reclaim is fast enough to dispose the clone before the SACK block comes in (the window is RTT long), otherwise we'll have to alloc some. With clones being handled I got these numbers (will be somewhat worse without that), taken with fine-grained mibs: TCPSackShifted 398 TCPSackMerged 877 TCPSackShiftFallback 320 TCPSACKCOLLAPSEFALLBACKGSO 0 TCPSACKCOLLAPSEFALLBACKSKBBITS 0 TCPSACKCOLLAPSEFALLBACKSKBDATA 0 TCPSACKCOLLAPSEFALLBACKBELOW 0 TCPSACKCOLLAPSEFALLBACKFIRST 1 TCPSACKCOLLAPSEFALLBACKPREVBITS 318 TCPSACKCOLLAPSEFALLBACKMSS 1 TCPSACKCOLLAPSEFALLBACKNOHEAD 0 TCPSACKCOLLAPSEFALLBACKSHIFT 0 TCPSACKCOLLAPSENOOPSEQ 0 TCPSACKCOLLAPSENOOPSMALLPCOUNT 0 TCPSACKCOLLAPSENOOPSMALLLEN 0 TCPSACKCOLLAPSEHOLE 12 Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-11-25 05:20:15 +00:00
/**
* skb_queue_is_first - check if skb is the first entry in the queue
* @list: queue head
* @skb: buffer
*
* Returns true if @skb is the first buffer on the list.
*/
static inline bool skb_queue_is_first(const struct sk_buff_head *list,
const struct sk_buff *skb)
{
return skb->prev == (struct sk_buff *)list;
tcp: Try to restore large SKBs while SACK processing During SACK processing, most of the benefits of TSO are eaten by the SACK blocks that one-by-one fragment SKBs to MSS sized chunks. Then we're in problems when cleanup work for them has to be done when a large cumulative ACK comes. Try to return back to pre-split state already while more and more SACK info gets discovered by combining newly discovered SACK areas with the previous skb if that's SACKed as well. This approach has a number of benefits: 1) The processing overhead is spread more equally over the RTT 2) Write queue has less skbs to process (affect everything which has to walk in the queue past the sacked areas) 3) Write queue is consistent whole the time, so no other parts of TCP has to be aware of this (this was not the case with some other approach that was, well, quite intrusive all around). 4) Clean_rtx_queue can release most of the pages using single put_page instead of previous PAGE_SIZE/mss+1 calls In case a hole is fully filled by the new SACK block, we attempt to combine the next skb too which allows construction of skbs that are even larger than what tso split them to and it handles hole per on every nth patterns that often occur during slow start overshoot pretty nicely. Though this to be really useful also a retransmission would have to get lost since cumulative ACKs advance one hole at a time in the most typical case. TODO: handle upwards only merging. That should be rather easy when segment is fully sacked but I'm leaving that as future work item (it won't make very large difference anyway since this current approach already covers quite a lot of normal cases). I was earlier thinking of some sophisticated way of tracking timestamps of the first and the last segment but later on realized that it won't be that necessary at all to store the timestamp of the last segment. The cases that can occur are basically either: 1) ambiguous => no sensible measurement can be taken anyway 2) non-ambiguous is due to reordering => having the timestamp of the last segment there is just skewing things more off than does some good since the ack got triggered by one of the holes (besides some substle issues that would make determining right hole/skb even harder problem). Anyway, it has nothing to do with this change then. I choose to route some abnormal looking cases with goto noop, some could be handled differently (eg., by stopping the walking at that skb but again). In general, they either shouldn't happen at all or are rare enough to make no difference in practice. In theory this change (as whole) could cause some macroscale regression (global) because of cache misses that are taken over the round-trip time but it gets very likely better because of much less (local) cache misses per other write queue walkers and the big recovery clearing cumulative ack. Worth to note that these benefits would be very easy to get also without TSO/GSO being on as long as the data is in pages so that we can merge them. Currently I won't let that happen because DSACK splitting at fragment that would mess up pcounts due to sk_can_gso in tcp_set_skb_tso_segs. Once DSACKs fragments gets avoided, we have some conditions that can be made less strict. TODO: I will probably have to convert the excessive pointer passing to struct sacktag_state... :-) My testing revealed that considerable amount of skbs couldn't be shifted because they were cloned (most likely still awaiting tx reclaim)... [The rest is considering future work instead since I got repeatably EFAULT to tcpdump's recvfrom when I added pskb_expand_head to deal with clones, so I separated that into another, later patch] ...To counter that, I gave up on the fifth advantage: 5) When growing previous SACK block, less allocs for new skbs are done, basically a new alloc is needed only when new hole is detected and when the previous skb runs out of frags space ...which now only happens of if reclaim is fast enough to dispose the clone before the SACK block comes in (the window is RTT long), otherwise we'll have to alloc some. With clones being handled I got these numbers (will be somewhat worse without that), taken with fine-grained mibs: TCPSackShifted 398 TCPSackMerged 877 TCPSackShiftFallback 320 TCPSACKCOLLAPSEFALLBACKGSO 0 TCPSACKCOLLAPSEFALLBACKSKBBITS 0 TCPSACKCOLLAPSEFALLBACKSKBDATA 0 TCPSACKCOLLAPSEFALLBACKBELOW 0 TCPSACKCOLLAPSEFALLBACKFIRST 1 TCPSACKCOLLAPSEFALLBACKPREVBITS 318 TCPSACKCOLLAPSEFALLBACKMSS 1 TCPSACKCOLLAPSEFALLBACKNOHEAD 0 TCPSACKCOLLAPSEFALLBACKSHIFT 0 TCPSACKCOLLAPSENOOPSEQ 0 TCPSACKCOLLAPSENOOPSMALLPCOUNT 0 TCPSACKCOLLAPSENOOPSMALLLEN 0 TCPSACKCOLLAPSEHOLE 12 Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-11-25 05:20:15 +00:00
}
/**
* skb_queue_next - return the next packet in the queue
* @list: queue head
* @skb: current buffer
*
* Return the next packet in @list after @skb. It is only valid to
* call this if skb_queue_is_last() evaluates to false.
*/
static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
const struct sk_buff *skb)
{
/* This BUG_ON may seem severe, but if we just return then we
* are going to dereference garbage.
*/
BUG_ON(skb_queue_is_last(list, skb));
return skb->next;
}
tcp: Try to restore large SKBs while SACK processing During SACK processing, most of the benefits of TSO are eaten by the SACK blocks that one-by-one fragment SKBs to MSS sized chunks. Then we're in problems when cleanup work for them has to be done when a large cumulative ACK comes. Try to return back to pre-split state already while more and more SACK info gets discovered by combining newly discovered SACK areas with the previous skb if that's SACKed as well. This approach has a number of benefits: 1) The processing overhead is spread more equally over the RTT 2) Write queue has less skbs to process (affect everything which has to walk in the queue past the sacked areas) 3) Write queue is consistent whole the time, so no other parts of TCP has to be aware of this (this was not the case with some other approach that was, well, quite intrusive all around). 4) Clean_rtx_queue can release most of the pages using single put_page instead of previous PAGE_SIZE/mss+1 calls In case a hole is fully filled by the new SACK block, we attempt to combine the next skb too which allows construction of skbs that are even larger than what tso split them to and it handles hole per on every nth patterns that often occur during slow start overshoot pretty nicely. Though this to be really useful also a retransmission would have to get lost since cumulative ACKs advance one hole at a time in the most typical case. TODO: handle upwards only merging. That should be rather easy when segment is fully sacked but I'm leaving that as future work item (it won't make very large difference anyway since this current approach already covers quite a lot of normal cases). I was earlier thinking of some sophisticated way of tracking timestamps of the first and the last segment but later on realized that it won't be that necessary at all to store the timestamp of the last segment. The cases that can occur are basically either: 1) ambiguous => no sensible measurement can be taken anyway 2) non-ambiguous is due to reordering => having the timestamp of the last segment there is just skewing things more off than does some good since the ack got triggered by one of the holes (besides some substle issues that would make determining right hole/skb even harder problem). Anyway, it has nothing to do with this change then. I choose to route some abnormal looking cases with goto noop, some could be handled differently (eg., by stopping the walking at that skb but again). In general, they either shouldn't happen at all or are rare enough to make no difference in practice. In theory this change (as whole) could cause some macroscale regression (global) because of cache misses that are taken over the round-trip time but it gets very likely better because of much less (local) cache misses per other write queue walkers and the big recovery clearing cumulative ack. Worth to note that these benefits would be very easy to get also without TSO/GSO being on as long as the data is in pages so that we can merge them. Currently I won't let that happen because DSACK splitting at fragment that would mess up pcounts due to sk_can_gso in tcp_set_skb_tso_segs. Once DSACKs fragments gets avoided, we have some conditions that can be made less strict. TODO: I will probably have to convert the excessive pointer passing to struct sacktag_state... :-) My testing revealed that considerable amount of skbs couldn't be shifted because they were cloned (most likely still awaiting tx reclaim)... [The rest is considering future work instead since I got repeatably EFAULT to tcpdump's recvfrom when I added pskb_expand_head to deal with clones, so I separated that into another, later patch] ...To counter that, I gave up on the fifth advantage: 5) When growing previous SACK block, less allocs for new skbs are done, basically a new alloc is needed only when new hole is detected and when the previous skb runs out of frags space ...which now only happens of if reclaim is fast enough to dispose the clone before the SACK block comes in (the window is RTT long), otherwise we'll have to alloc some. With clones being handled I got these numbers (will be somewhat worse without that), taken with fine-grained mibs: TCPSackShifted 398 TCPSackMerged 877 TCPSackShiftFallback 320 TCPSACKCOLLAPSEFALLBACKGSO 0 TCPSACKCOLLAPSEFALLBACKSKBBITS 0 TCPSACKCOLLAPSEFALLBACKSKBDATA 0 TCPSACKCOLLAPSEFALLBACKBELOW 0 TCPSACKCOLLAPSEFALLBACKFIRST 1 TCPSACKCOLLAPSEFALLBACKPREVBITS 318 TCPSACKCOLLAPSEFALLBACKMSS 1 TCPSACKCOLLAPSEFALLBACKNOHEAD 0 TCPSACKCOLLAPSEFALLBACKSHIFT 0 TCPSACKCOLLAPSENOOPSEQ 0 TCPSACKCOLLAPSENOOPSMALLPCOUNT 0 TCPSACKCOLLAPSENOOPSMALLLEN 0 TCPSACKCOLLAPSEHOLE 12 Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-11-25 05:20:15 +00:00
/**
* skb_queue_prev - return the prev packet in the queue
* @list: queue head
* @skb: current buffer
*
* Return the prev packet in @list before @skb. It is only valid to
* call this if skb_queue_is_first() evaluates to false.
*/
static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
const struct sk_buff *skb)
{
/* This BUG_ON may seem severe, but if we just return then we
* are going to dereference garbage.
*/
BUG_ON(skb_queue_is_first(list, skb));
return skb->prev;
}
/**
* skb_get - reference buffer
* @skb: buffer to reference
*
* Makes another reference to a socket buffer and returns a pointer
* to the buffer.
*/
static inline struct sk_buff *skb_get(struct sk_buff *skb)
{
atomic_inc(&skb->users);
return skb;
}
/*
* If users == 1, we are the only owner and are can avoid redundant
* atomic change.
*/
/**
* skb_cloned - is the buffer a clone
* @skb: buffer to check
*
* Returns true if the buffer was generated with skb_clone() and is
* one of multiple shared copies of the buffer. Cloned buffers are
* shared data so must not be written to under normal circumstances.
*/
static inline int skb_cloned(const struct sk_buff *skb)
{
return skb->cloned &&
(atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
}
/**
* skb_header_cloned - is the header a clone
* @skb: buffer to check
*
* Returns true if modifying the header part of the buffer requires
* the data to be copied.
*/
static inline int skb_header_cloned(const struct sk_buff *skb)
{
int dataref;
if (!skb->cloned)
return 0;
dataref = atomic_read(&skb_shinfo(skb)->dataref);
dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
return dataref != 1;
}
/**
* skb_header_release - release reference to header
* @skb: buffer to operate on
*
* Drop a reference to the header part of the buffer. This is done
* by acquiring a payload reference. You must not read from the header
* part of skb->data after this.
*/
static inline void skb_header_release(struct sk_buff *skb)
{
BUG_ON(skb->nohdr);
skb->nohdr = 1;
atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
}
/**
* skb_shared - is the buffer shared
* @skb: buffer to check
*
* Returns true if more than one person has a reference to this
* buffer.
*/
static inline int skb_shared(const struct sk_buff *skb)
{
return atomic_read(&skb->users) != 1;
}
/**
* skb_share_check - check if buffer is shared and if so clone it
* @skb: buffer to check
* @pri: priority for memory allocation
*
* If the buffer is shared the buffer is cloned and the old copy
* drops a reference. A new clone with a single reference is returned.
* If the buffer is not shared the original buffer is returned. When
* being called from interrupt status or with spinlocks held pri must
* be GFP_ATOMIC.
*
* NULL is returned on a memory allocation failure.
*/
static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
gfp_t pri)
{
might_sleep_if(pri & __GFP_WAIT);
if (skb_shared(skb)) {
struct sk_buff *nskb = skb_clone(skb, pri);
kfree_skb(skb);
skb = nskb;
}
return skb;
}
/*
* Copy shared buffers into a new sk_buff. We effectively do COW on
* packets to handle cases where we have a local reader and forward
* and a couple of other messy ones. The normal one is tcpdumping
* a packet thats being forwarded.
*/
/**
* skb_unshare - make a copy of a shared buffer
* @skb: buffer to check
* @pri: priority for memory allocation
*
* If the socket buffer is a clone then this function creates a new
* copy of the data, drops a reference count on the old copy and returns
* the new copy with the reference count at 1. If the buffer is not a clone
* the original buffer is returned. When called with a spinlock held or
* from interrupt state @pri must be %GFP_ATOMIC
*
* %NULL is returned on a memory allocation failure.
*/
static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
gfp_t pri)
{
might_sleep_if(pri & __GFP_WAIT);
if (skb_cloned(skb)) {
struct sk_buff *nskb = skb_copy(skb, pri);
kfree_skb(skb); /* Free our shared copy */
skb = nskb;
}
return skb;
}
/**
* skb_peek - peek at the head of an &sk_buff_head
* @list_: list to peek at
*
* Peek an &sk_buff. Unlike most other operations you _MUST_
* be careful with this one. A peek leaves the buffer on the
* list and someone else may run off with it. You must hold
* the appropriate locks or have a private queue to do this.
*
* Returns %NULL for an empty list or a pointer to the head element.
* The reference count is not incremented and the reference is therefore
* volatile. Use with caution.
*/
static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
{
struct sk_buff *list = ((struct sk_buff *)list_)->next;
if (list == (struct sk_buff *)list_)
list = NULL;
return list;
}
/**
* skb_peek_tail - peek at the tail of an &sk_buff_head
* @list_: list to peek at
*
* Peek an &sk_buff. Unlike most other operations you _MUST_
* be careful with this one. A peek leaves the buffer on the
* list and someone else may run off with it. You must hold
* the appropriate locks or have a private queue to do this.
*
* Returns %NULL for an empty list or a pointer to the tail element.
* The reference count is not incremented and the reference is therefore
* volatile. Use with caution.
*/
static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
{
struct sk_buff *list = ((struct sk_buff *)list_)->prev;
if (list == (struct sk_buff *)list_)
list = NULL;
return list;
}
/**
* skb_queue_len - get queue length
* @list_: list to measure
*
* Return the length of an &sk_buff queue.
*/
static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
{
return list_->qlen;
}
/**
* __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
* @list: queue to initialize
*
* This initializes only the list and queue length aspects of
* an sk_buff_head object. This allows to initialize the list
* aspects of an sk_buff_head without reinitializing things like
* the spinlock. It can also be used for on-stack sk_buff_head
* objects where the spinlock is known to not be used.
*/
static inline void __skb_queue_head_init(struct sk_buff_head *list)
{
list->prev = list->next = (struct sk_buff *)list;
list->qlen = 0;
}
/*
* This function creates a split out lock class for each invocation;
* this is needed for now since a whole lot of users of the skb-queue
* infrastructure in drivers have different locking usage (in hardirq)
* than the networking core (in softirq only). In the long run either the
* network layer or drivers should need annotation to consolidate the
* main types of usage into 3 classes.
*/
static inline void skb_queue_head_init(struct sk_buff_head *list)
{
spin_lock_init(&list->lock);
__skb_queue_head_init(list);
}
static inline void skb_queue_head_init_class(struct sk_buff_head *list,
struct lock_class_key *class)
{
skb_queue_head_init(list);
lockdep_set_class(&list->lock, class);
}
/*
* Insert an sk_buff on a list.
*
* The "__skb_xxxx()" functions are the non-atomic ones that
* can only be called with interrupts disabled.
*/
extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
static inline void __skb_insert(struct sk_buff *newsk,
struct sk_buff *prev, struct sk_buff *next,
struct sk_buff_head *list)
{
newsk->next = next;
newsk->prev = prev;
next->prev = prev->next = newsk;
list->qlen++;
}
static inline void __skb_queue_splice(const struct sk_buff_head *list,
struct sk_buff *prev,
struct sk_buff *next)
{
struct sk_buff *first = list->next;
struct sk_buff *last = list->prev;
first->prev = prev;
prev->next = first;
last->next = next;
next->prev = last;
}
/**
* skb_queue_splice - join two skb lists, this is designed for stacks
* @list: the new list to add
* @head: the place to add it in the first list
*/
static inline void skb_queue_splice(const struct sk_buff_head *list,
struct sk_buff_head *head)
{
if (!skb_queue_empty(list)) {
__skb_queue_splice(list, (struct sk_buff *) head, head->next);
head->qlen += list->qlen;
}
}
/**
* skb_queue_splice - join two skb lists and reinitialise the emptied list
* @list: the new list to add
* @head: the place to add it in the first list
*
* The list at @list is reinitialised
*/
static inline void skb_queue_splice_init(struct sk_buff_head *list,
struct sk_buff_head *head)
{
if (!skb_queue_empty(list)) {
__skb_queue_splice(list, (struct sk_buff *) head, head->next);
head->qlen += list->qlen;
__skb_queue_head_init(list);
}
}
/**
* skb_queue_splice_tail - join two skb lists, each list being a queue
* @list: the new list to add
* @head: the place to add it in the first list
*/
static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
struct sk_buff_head *head)
{
if (!skb_queue_empty(list)) {
__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
head->qlen += list->qlen;
}
}
/**
* skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
* @list: the new list to add
* @head: the place to add it in the first list
*
* Each of the lists is a queue.
* The list at @list is reinitialised
*/
static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
struct sk_buff_head *head)
{
if (!skb_queue_empty(list)) {
__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
head->qlen += list->qlen;
__skb_queue_head_init(list);
}
}
/**
* __skb_queue_after - queue a buffer at the list head
* @list: list to use
* @prev: place after this buffer
* @newsk: buffer to queue
*
* Queue a buffer int the middle of a list. This function takes no locks
* and you must therefore hold required locks before calling it.
*
* A buffer cannot be placed on two lists at the same time.
*/
static inline void __skb_queue_after(struct sk_buff_head *list,
struct sk_buff *prev,
struct sk_buff *newsk)
{
__skb_insert(newsk, prev, prev->next, list);
}
extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
struct sk_buff_head *list);
static inline void __skb_queue_before(struct sk_buff_head *list,
struct sk_buff *next,
struct sk_buff *newsk)
{
__skb_insert(newsk, next->prev, next, list);
}
/**
* __skb_queue_head - queue a buffer at the list head
* @list: list to use
* @newsk: buffer to queue
*
* Queue a buffer at the start of a list. This function takes no locks
* and you must therefore hold required locks before calling it.
*
* A buffer cannot be placed on two lists at the same time.
*/
extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
static inline void __skb_queue_head(struct sk_buff_head *list,
struct sk_buff *newsk)
{
__skb_queue_after(list, (struct sk_buff *)list, newsk);
}
/**
* __skb_queue_tail - queue a buffer at the list tail
* @list: list to use
* @newsk: buffer to queue
*
* Queue a buffer at the end of a list. This function takes no locks
* and you must therefore hold required locks before calling it.
*
* A buffer cannot be placed on two lists at the same time.
*/
extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
static inline void __skb_queue_tail(struct sk_buff_head *list,
struct sk_buff *newsk)
{
__skb_queue_before(list, (struct sk_buff *)list, newsk);
}
/*
* remove sk_buff from list. _Must_ be called atomically, and with
* the list known..
*/
extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
{
struct sk_buff *next, *prev;
list->qlen--;
next = skb->next;
prev = skb->prev;
skb->next = skb->prev = NULL;
next->prev = prev;
prev->next = next;
}
/**
* __skb_dequeue - remove from the head of the queue
* @list: list to dequeue from
*
* Remove the head of the list. This function does not take any locks
* so must be used with appropriate locks held only. The head item is
* returned or %NULL if the list is empty.
*/
extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
{
struct sk_buff *skb = skb_peek(list);
if (skb)
__skb_unlink(skb, list);
return skb;
}
/**
* __skb_dequeue_tail - remove from the tail of the queue
* @list: list to dequeue from
*
* Remove the tail of the list. This function does not take any locks
* so must be used with appropriate locks held only. The tail item is
* returned or %NULL if the list is empty.
*/
extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
{
struct sk_buff *skb = skb_peek_tail(list);
if (skb)
__skb_unlink(skb, list);
return skb;
}
static inline int skb_is_nonlinear(const struct sk_buff *skb)
{
return skb->data_len;
}
static inline unsigned int skb_headlen(const struct sk_buff *skb)
{
return skb->len - skb->data_len;
}
static inline int skb_pagelen(const struct sk_buff *skb)
{
int i, len = 0;
for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
len += skb_shinfo(skb)->frags[i].size;
return len + skb_headlen(skb);
}
static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
struct page *page, int off, int size)
{
skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
frag->page = page;
frag->page_offset = off;
frag->size = size;
skb_shinfo(skb)->nr_frags = i + 1;
}
extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
int off, int size);
#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
#ifdef NET_SKBUFF_DATA_USES_OFFSET
static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
{
return skb->head + skb->tail;
}
static inline void skb_reset_tail_pointer(struct sk_buff *skb)
{
skb->tail = skb->data - skb->head;
}
static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
{
skb_reset_tail_pointer(skb);
skb->tail += offset;
}
#else /* NET_SKBUFF_DATA_USES_OFFSET */
static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
{
return skb->tail;
}
static inline void skb_reset_tail_pointer(struct sk_buff *skb)
{
skb->tail = skb->data;
}
static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
{
skb->tail = skb->data + offset;
}
#endif /* NET_SKBUFF_DATA_USES_OFFSET */
/*
* Add data to an sk_buff
*/
extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
{
unsigned char *tmp = skb_tail_pointer(skb);
SKB_LINEAR_ASSERT(skb);
skb->tail += len;
skb->len += len;
return tmp;
}
extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
{
skb->data -= len;
skb->len += len;
return skb->data;
}
extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
{
skb->len -= len;
BUG_ON(skb->len < skb->data_len);
return skb->data += len;
}
static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
{
return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
}
extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
{
if (len > skb_headlen(skb) &&
!__pskb_pull_tail(skb, len - skb_headlen(skb)))
return NULL;
skb->len -= len;
return skb->data += len;
}
static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
{
return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
}
static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
{
if (likely(len <= skb_headlen(skb)))
return 1;
if (unlikely(len > skb->len))
return 0;
return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
}
/**
* skb_headroom - bytes at buffer head
* @skb: buffer to check
*
* Return the number of bytes of free space at the head of an &sk_buff.
*/
static inline unsigned int skb_headroom(const struct sk_buff *skb)
{
return skb->data - skb->head;
}
/**
* skb_tailroom - bytes at buffer end
* @skb: buffer to check
*
* Return the number of bytes of free space at the tail of an sk_buff
*/
static inline int skb_tailroom(const struct sk_buff *skb)
{
return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
}
/**
* skb_reserve - adjust headroom
* @skb: buffer to alter
* @len: bytes to move
*
* Increase the headroom of an empty &sk_buff by reducing the tail
* room. This is only allowed for an empty buffer.
*/
static inline void skb_reserve(struct sk_buff *skb, int len)
{
skb->data += len;
skb->tail += len;
}
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
#ifdef NET_SKBUFF_DATA_USES_OFFSET
static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
return skb->head + skb->transport_header;
}
static inline void skb_reset_transport_header(struct sk_buff *skb)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
skb->transport_header = skb->data - skb->head;
}
static inline void skb_set_transport_header(struct sk_buff *skb,
const int offset)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
skb_reset_transport_header(skb);
skb->transport_header += offset;
}
static inline unsigned char *skb_network_header(const struct sk_buff *skb)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
return skb->head + skb->network_header;
}
static inline void skb_reset_network_header(struct sk_buff *skb)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
skb->network_header = skb->data - skb->head;
}
static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
skb_reset_network_header(skb);
skb->network_header += offset;
}
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
return skb->head + skb->mac_header;
}
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
static inline int skb_mac_header_was_set(const struct sk_buff *skb)
{
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
return skb->mac_header != ~0U;
}
static inline void skb_reset_mac_header(struct sk_buff *skb)
{
skb->mac_header = skb->data - skb->head;
}
static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
{
skb_reset_mac_header(skb);
skb->mac_header += offset;
}
#else /* NET_SKBUFF_DATA_USES_OFFSET */
static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
{
return skb->transport_header;
}
static inline void skb_reset_transport_header(struct sk_buff *skb)
{
skb->transport_header = skb->data;
}
static inline void skb_set_transport_header(struct sk_buff *skb,
const int offset)
{
skb->transport_header = skb->data + offset;
}
static inline unsigned char *skb_network_header(const struct sk_buff *skb)
{
return skb->network_header;
}
static inline void skb_reset_network_header(struct sk_buff *skb)
{
skb->network_header = skb->data;
}
static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
{
skb->network_header = skb->data + offset;
}
static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
{
return skb->mac_header;
}
static inline int skb_mac_header_was_set(const struct sk_buff *skb)
{
return skb->mac_header != NULL;
}
static inline void skb_reset_mac_header(struct sk_buff *skb)
{
skb->mac_header = skb->data;
}
static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
{
skb->mac_header = skb->data + offset;
}
[SK_BUFF]: Use offsets for skb->{mac,network,transport}_header on 64bit architectures With this we save 8 bytes per network packet, leaving a 4 bytes hole to be used in further shrinking work, likely with the offsetization of other pointers, such as ->{data,tail,end}, at the cost of adds, that were minimized by the usual practice of setting skb->{mac,nh,n}.raw to a local variable that is then accessed multiple times in each function, it also is not more expensive than before with regards to most of the handling of such headers, like setting one of these headers to another (transport to network, etc), or subtracting, adding to/from it, comparing them, etc. Now we have this layout for sk_buff on a x86_64 machine: [acme@mica net-2.6.22]$ pahole vmlinux sk_buff struct sk_buff { struct sk_buff * next; /* 0 8 */ struct sk_buff * prev; /* 8 8 */ struct rb_node rb; /* 16 24 */ struct sock * sk; /* 40 8 */ ktime_t tstamp; /* 48 8 */ struct net_device * dev; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct net_device * input_dev; /* 64 8 */ sk_buff_data_t transport_header; /* 72 4 */ sk_buff_data_t network_header; /* 76 4 */ sk_buff_data_t mac_header; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dst_entry * dst; /* 88 8 */ struct sec_path * sp; /* 96 8 */ char cb[48]; /* 104 48 */ /* cacheline 2 boundary (128 bytes) was 24 bytes ago*/ unsigned int len; /* 152 4 */ unsigned int data_len; /* 156 4 */ unsigned int mac_len; /* 160 4 */ union { __wsum csum; /* 4 */ __u32 csum_offset; /* 4 */ }; /* 164 4 */ __u32 priority; /* 168 4 */ __u8 local_df:1; /* 172 1 */ __u8 cloned:1; /* 172 1 */ __u8 ip_summed:2; /* 172 1 */ __u8 nohdr:1; /* 172 1 */ __u8 nfctinfo:3; /* 172 1 */ __u8 pkt_type:3; /* 173 1 */ __u8 fclone:2; /* 173 1 */ __u8 ipvs_property:1; /* 173 1 */ /* XXX 2 bits hole, try to pack */ __be16 protocol; /* 174 2 */ void (*destructor)(struct sk_buff *); /* 176 8 */ struct nf_conntrack * nfct; /* 184 8 */ /* --- cacheline 3 boundary (192 bytes) --- */ struct sk_buff * nfct_reasm; /* 192 8 */ struct nf_bridge_info *nf_bridge; /* 200 8 */ __u16 tc_index; /* 208 2 */ __u16 tc_verd; /* 210 2 */ dma_cookie_t dma_cookie; /* 212 4 */ __u32 secmark; /* 216 4 */ __u32 mark; /* 220 4 */ unsigned int truesize; /* 224 4 */ atomic_t users; /* 228 4 */ unsigned char * head; /* 232 8 */ unsigned char * data; /* 240 8 */ unsigned char * tail; /* 248 8 */ /* --- cacheline 4 boundary (256 bytes) --- */ unsigned char * end; /* 256 8 */ }; /* size: 264, cachelines: 5 */ /* sum members: 260, holes: 1, sum holes: 4 */ /* bit holes: 1, sum bit holes: 2 bits */ /* last cacheline: 8 bytes */ On 32 bits nothing changes, and pointers continue to be used with the compiler turning all this abstraction layer into dust. But there are some sk_buff validation tricks that are now possible, humm... :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-04-11 04:22:35 +00:00
#endif /* NET_SKBUFF_DATA_USES_OFFSET */
static inline int skb_transport_offset(const struct sk_buff *skb)
{
return skb_transport_header(skb) - skb->data;
}
static inline u32 skb_network_header_len(const struct sk_buff *skb)
{
return skb->transport_header - skb->network_header;
}
static inline int skb_network_offset(const struct sk_buff *skb)
{
return skb_network_header(skb) - skb->data;
}
static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
{
return pskb_may_pull(skb, skb_network_offset(skb) + len);
}
/*
* CPUs often take a performance hit when accessing unaligned memory
* locations. The actual performance hit varies, it can be small if the
* hardware handles it or large if we have to take an exception and fix it
* in software.
*
* Since an ethernet header is 14 bytes network drivers often end up with
* the IP header at an unaligned offset. The IP header can be aligned by
* shifting the start of the packet by 2 bytes. Drivers should do this
* with:
*
* skb_reserve(skb, NET_IP_ALIGN);
*
* The downside to this alignment of the IP header is that the DMA is now
* unaligned. On some architectures the cost of an unaligned DMA is high
* and this cost outweighs the gains made by aligning the IP header.
*
* Since this trade off varies between architectures, we allow NET_IP_ALIGN
* to be overridden.
*/
#ifndef NET_IP_ALIGN
#define NET_IP_ALIGN 2
#endif
/*
* The networking layer reserves some headroom in skb data (via
* dev_alloc_skb). This is used to avoid having to reallocate skb data when
* the header has to grow. In the default case, if the header has to grow
* 32 bytes or less we avoid the reallocation.
*
* Unfortunately this headroom changes the DMA alignment of the resulting
* network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
* on some architectures. An architecture can override this value,
* perhaps setting it to a cacheline in size (since that will maintain
* cacheline alignment of the DMA). It must be a power of 2.
*
* Various parts of the networking layer expect at least 32 bytes of
* headroom, you should not reduce this.
*
* Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
* to reduce average number of cache lines per packet.
* get_rps_cpus() for example only access one 64 bytes aligned block :
* NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
*/
#ifndef NET_SKB_PAD
#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
#endif
extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
{
if (unlikely(skb->data_len)) {
WARN_ON(1);
return;
}
skb->len = len;
skb_set_tail_pointer(skb, len);
}
extern void skb_trim(struct sk_buff *skb, unsigned int len);
static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
{
if (skb->data_len)
return ___pskb_trim(skb, len);
__skb_trim(skb, len);
return 0;
}
static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
{
return (len < skb->len) ? __pskb_trim(skb, len) : 0;
}
/**
* pskb_trim_unique - remove end from a paged unique (not cloned) buffer
* @skb: buffer to alter
* @len: new length
*
* This is identical to pskb_trim except that the caller knows that
* the skb is not cloned so we should never get an error due to out-
* of-memory.
*/
static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
{
int err = pskb_trim(skb, len);
BUG_ON(err);
}
/**
* skb_orphan - orphan a buffer
* @skb: buffer to orphan
*
* If a buffer currently has an owner then we call the owner's
* destructor function and make the @skb unowned. The buffer continues
* to exist but is no longer charged to its former owner.
*/
static inline void skb_orphan(struct sk_buff *skb)
{
if (skb->destructor)
skb->destructor(skb);
skb->destructor = NULL;
skb->sk = NULL;
}
/**
* __skb_queue_purge - empty a list
* @list: list to empty
*
* Delete all buffers on an &sk_buff list. Each buffer is removed from
* the list and one reference dropped. This function does not take the
* list lock and the caller must hold the relevant locks to use it.
*/
extern void skb_queue_purge(struct sk_buff_head *list);
static inline void __skb_queue_purge(struct sk_buff_head *list)
{
struct sk_buff *skb;
while ((skb = __skb_dequeue(list)) != NULL)
kfree_skb(skb);
}
/**
* __dev_alloc_skb - allocate an skbuff for receiving
* @length: length to allocate
* @gfp_mask: get_free_pages mask, passed to alloc_skb
*
* Allocate a new &sk_buff and assign it a usage count of one. The
* buffer has unspecified headroom built in. Users should allocate
* the headroom they think they need without accounting for the
* built in space. The built in space is used for optimisations.
*
* %NULL is returned if there is no free memory.
*/
static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
gfp_t gfp_mask)
{
struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
if (likely(skb))
skb_reserve(skb, NET_SKB_PAD);
return skb;
}
extern struct sk_buff *dev_alloc_skb(unsigned int length);
extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
unsigned int length, gfp_t gfp_mask);
/**
* netdev_alloc_skb - allocate an skbuff for rx on a specific device
* @dev: network device to receive on
* @length: length to allocate
*
* Allocate a new &sk_buff and assign it a usage count of one. The
* buffer has unspecified headroom built in. Users should allocate
* the headroom they think they need without accounting for the
* built in space. The built in space is used for optimisations.
*
* %NULL is returned if there is no free memory. Although this function
* allocates memory it can be called from an interrupt.
*/
static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
unsigned int length)
{
return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
}
static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
unsigned int length)
{
struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
if (NET_IP_ALIGN && skb)
skb_reserve(skb, NET_IP_ALIGN);
return skb;
}
extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
/**
* netdev_alloc_page - allocate a page for ps-rx on a specific device
* @dev: network device to receive on
*
* Allocate a new page node local to the specified device.
*
* %NULL is returned if there is no free memory.
*/
static inline struct page *netdev_alloc_page(struct net_device *dev)
{
return __netdev_alloc_page(dev, GFP_ATOMIC);
}
static inline void netdev_free_page(struct net_device *dev, struct page *page)
{
__free_page(page);
}
[SKBUFF]: Keep track of writable header len of headerless clones Currently NAT (and others) that want to modify cloned skbs copy them, even if in the vast majority of cases its not necessary because the skb is a clone made by TCP and the portion NAT wants to modify is actually writable because TCP release the header reference before cloning. The problem is that there is no clean way for NAT to find out how long the writable header area is, so this patch introduces skb->hdr_len to hold this length. When a headerless skb is cloned skb->hdr_len is set to the current headroom, for regular clones it is copied from the original. A new function skb_clone_writable(skb, len) returns whether the skb is writable up to len bytes from skb->data. To avoid enlarging the skb the mac_len field is reduced to 16 bit and the new hdr_len field is put in the remaining 16 bit. I've done a few rough benchmarks of NAT (not with this exact patch, but a very similar one). As expected it saves huge amounts of system time in case of sendfile, bringing it down to basically the same amount as without NAT, with sendmsg it only helps on loopback, probably because of the large MTU. Transmit a 1GB file using sendfile/sendmsg over eth0/lo with and without NAT: - sendfile eth0, no NAT: sys 0m0.388s - sendfile eth0, NAT: sys 0m1.835s - sendfile eth0: NAT + path: sys 0m0.370s (~ -80%) - sendfile lo, no NAT: sys 0m0.258s - sendfile lo, NAT: sys 0m2.609s - sendfile lo, NAT + patch: sys 0m0.260s (~ -90%) - sendmsg eth0, no NAT: sys 0m2.508s - sendmsg eth0, NAT: sys 0m2.539s - sendmsg eth0, NAT + patch: sys 0m2.445s (no change) - sendmsg lo, no NAT: sys 0m2.151s - sendmsg lo, NAT: sys 0m3.557s - sendmsg lo, NAT + patch: sys 0m2.159s (~ -40%) I expect other users can see a similar performance improvement, packet mangling iptables targets, ipip and ip_gre come to mind .. Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-25 11:35:20 +00:00
/**
* skb_clone_writable - is the header of a clone writable
* @skb: buffer to check
* @len: length up to which to write
*
* Returns true if modifying the header part of the cloned buffer
* does not requires the data to be copied.
*/
static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
[SKBUFF]: Keep track of writable header len of headerless clones Currently NAT (and others) that want to modify cloned skbs copy them, even if in the vast majority of cases its not necessary because the skb is a clone made by TCP and the portion NAT wants to modify is actually writable because TCP release the header reference before cloning. The problem is that there is no clean way for NAT to find out how long the writable header area is, so this patch introduces skb->hdr_len to hold this length. When a headerless skb is cloned skb->hdr_len is set to the current headroom, for regular clones it is copied from the original. A new function skb_clone_writable(skb, len) returns whether the skb is writable up to len bytes from skb->data. To avoid enlarging the skb the mac_len field is reduced to 16 bit and the new hdr_len field is put in the remaining 16 bit. I've done a few rough benchmarks of NAT (not with this exact patch, but a very similar one). As expected it saves huge amounts of system time in case of sendfile, bringing it down to basically the same amount as without NAT, with sendmsg it only helps on loopback, probably because of the large MTU. Transmit a 1GB file using sendfile/sendmsg over eth0/lo with and without NAT: - sendfile eth0, no NAT: sys 0m0.388s - sendfile eth0, NAT: sys 0m1.835s - sendfile eth0: NAT + path: sys 0m0.370s (~ -80%) - sendfile lo, no NAT: sys 0m0.258s - sendfile lo, NAT: sys 0m2.609s - sendfile lo, NAT + patch: sys 0m0.260s (~ -90%) - sendmsg eth0, no NAT: sys 0m2.508s - sendmsg eth0, NAT: sys 0m2.539s - sendmsg eth0, NAT + patch: sys 0m2.445s (no change) - sendmsg lo, no NAT: sys 0m2.151s - sendmsg lo, NAT: sys 0m3.557s - sendmsg lo, NAT + patch: sys 0m2.159s (~ -40%) I expect other users can see a similar performance improvement, packet mangling iptables targets, ipip and ip_gre come to mind .. Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-06-25 11:35:20 +00:00
{
return !skb_header_cloned(skb) &&
skb_headroom(skb) + len <= skb->hdr_len;
}
static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
int cloned)
{
int delta = 0;
if (headroom < NET_SKB_PAD)
headroom = NET_SKB_PAD;
if (headroom > skb_headroom(skb))
delta = headroom - skb_headroom(skb);
if (delta || cloned)
return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
GFP_ATOMIC);
return 0;
}
/**
* skb_cow - copy header of skb when it is required
* @skb: buffer to cow
* @headroom: needed headroom
*
* If the skb passed lacks sufficient headroom or its data part
* is shared, data is reallocated. If reallocation fails, an error
* is returned and original skb is not changed.
*
* The result is skb with writable area skb->head...skb->tail
* and at least @headroom of space at head.
*/
static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
{
return __skb_cow(skb, headroom, skb_cloned(skb));
}
/**
* skb_cow_head - skb_cow but only making the head writable
* @skb: buffer to cow
* @headroom: needed headroom
*
* This function is identical to skb_cow except that we replace the
* skb_cloned check by skb_header_cloned. It should be used when
* you only need to push on some header and do not need to modify
* the data.
*/
static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
{
return __skb_cow(skb, headroom, skb_header_cloned(skb));
}
/**
* skb_padto - pad an skbuff up to a minimal size
* @skb: buffer to pad
* @len: minimal length
*
* Pads up a buffer to ensure the trailing bytes exist and are
* blanked. If the buffer already contains sufficient data it
* is untouched. Otherwise it is extended. Returns zero on
* success. The skb is freed on error.
*/
static inline int skb_padto(struct sk_buff *skb, unsigned int len)
{
unsigned int size = skb->len;
if (likely(size >= len))
return 0;
return skb_pad(skb, len - size);
}
static inline int skb_add_data(struct sk_buff *skb,
char __user *from, int copy)
{
const int off = skb->len;
if (skb->ip_summed == CHECKSUM_NONE) {
int err = 0;
__wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
copy, 0, &err);
if (!err) {
skb->csum = csum_block_add(skb->csum, csum, off);
return 0;
}
} else if (!copy_from_user(skb_put(skb, copy), from, copy))
return 0;
__skb_trim(skb, off);
return -EFAULT;
}
static inline int skb_can_coalesce(struct sk_buff *skb, int i,
struct page *page, int off)
{
if (i) {
struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
return page == frag->page &&
off == frag->page_offset + frag->size;
}
return 0;
}
static inline int __skb_linearize(struct sk_buff *skb)
{
return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
}
/**
* skb_linearize - convert paged skb to linear one
* @skb: buffer to linarize
*
* If there is no free memory -ENOMEM is returned, otherwise zero
* is returned and the old skb data released.
*/
static inline int skb_linearize(struct sk_buff *skb)
{
return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
}
/**
* skb_linearize_cow - make sure skb is linear and writable
* @skb: buffer to process
*
* If there is no free memory -ENOMEM is returned, otherwise zero
* is returned and the old skb data released.
*/
static inline int skb_linearize_cow(struct sk_buff *skb)
{
return skb_is_nonlinear(skb) || skb_cloned(skb) ?
__skb_linearize(skb) : 0;
}
/**
* skb_postpull_rcsum - update checksum for received skb after pull
* @skb: buffer to update
* @start: start of data before pull
* @len: length of data pulled
*
* After doing a pull on a received packet, you need to call this to
* update the CHECKSUM_COMPLETE checksum, or set ip_summed to
* CHECKSUM_NONE so that it can be recomputed from scratch.
*/
static inline void skb_postpull_rcsum(struct sk_buff *skb,
const void *start, unsigned int len)
{
if (skb->ip_summed == CHECKSUM_COMPLETE)
skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
}
unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
/**
* pskb_trim_rcsum - trim received skb and update checksum
* @skb: buffer to trim
* @len: new length
*
* This is exactly the same as pskb_trim except that it ensures the
* checksum of received packets are still valid after the operation.
*/
static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
{
if (likely(len >= skb->len))
return 0;
if (skb->ip_summed == CHECKSUM_COMPLETE)
skb->ip_summed = CHECKSUM_NONE;
return __pskb_trim(skb, len);
}
#define skb_queue_walk(queue, skb) \
for (skb = (queue)->next; \
prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
skb = skb->next)
#define skb_queue_walk_safe(queue, skb, tmp) \
for (skb = (queue)->next, tmp = skb->next; \
skb != (struct sk_buff *)(queue); \
skb = tmp, tmp = skb->next)
#define skb_queue_walk_from(queue, skb) \
for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
skb = skb->next)
#define skb_queue_walk_from_safe(queue, skb, tmp) \
for (tmp = skb->next; \
skb != (struct sk_buff *)(queue); \
skb = tmp, tmp = skb->next)
#define skb_queue_reverse_walk(queue, skb) \
for (skb = (queue)->prev; \
prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
skb = skb->prev)
static inline bool skb_has_frag_list(const struct sk_buff *skb)
{
return skb_shinfo(skb)->frag_list != NULL;
}
static inline void skb_frag_list_init(struct sk_buff *skb)
{
skb_shinfo(skb)->frag_list = NULL;
}
static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
{
frag->next = skb_shinfo(skb)->frag_list;
skb_shinfo(skb)->frag_list = frag;
}
#define skb_walk_frags(skb, iter) \
for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
int *peeked, int *err);
extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
int noblock, int *err);
extern unsigned int datagram_poll(struct file *file, struct socket *sock,
struct poll_table_struct *wait);
extern int skb_copy_datagram_iovec(const struct sk_buff *from,
int offset, struct iovec *to,
int size);
extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
int hlen,
struct iovec *iov);
extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
int offset,
const struct iovec *from,
int from_offset,
int len);
extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
int offset,
const struct iovec *to,
int to_offset,
int size);
extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
extern void skb_free_datagram_locked(struct sock *sk,
struct sk_buff *skb);
extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
unsigned int flags);
extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
int len, __wsum csum);
extern int skb_copy_bits(const struct sk_buff *skb, int offset,
void *to, int len);
extern int skb_store_bits(struct sk_buff *skb, int offset,
const void *from, int len);
extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
int offset, u8 *to, int len,
__wsum csum);
extern int skb_splice_bits(struct sk_buff *skb,
unsigned int offset,
struct pipe_inode_info *pipe,
unsigned int len,
unsigned int flags);
extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
extern void skb_split(struct sk_buff *skb,
struct sk_buff *skb1, const u32 len);
tcp: Try to restore large SKBs while SACK processing During SACK processing, most of the benefits of TSO are eaten by the SACK blocks that one-by-one fragment SKBs to MSS sized chunks. Then we're in problems when cleanup work for them has to be done when a large cumulative ACK comes. Try to return back to pre-split state already while more and more SACK info gets discovered by combining newly discovered SACK areas with the previous skb if that's SACKed as well. This approach has a number of benefits: 1) The processing overhead is spread more equally over the RTT 2) Write queue has less skbs to process (affect everything which has to walk in the queue past the sacked areas) 3) Write queue is consistent whole the time, so no other parts of TCP has to be aware of this (this was not the case with some other approach that was, well, quite intrusive all around). 4) Clean_rtx_queue can release most of the pages using single put_page instead of previous PAGE_SIZE/mss+1 calls In case a hole is fully filled by the new SACK block, we attempt to combine the next skb too which allows construction of skbs that are even larger than what tso split them to and it handles hole per on every nth patterns that often occur during slow start overshoot pretty nicely. Though this to be really useful also a retransmission would have to get lost since cumulative ACKs advance one hole at a time in the most typical case. TODO: handle upwards only merging. That should be rather easy when segment is fully sacked but I'm leaving that as future work item (it won't make very large difference anyway since this current approach already covers quite a lot of normal cases). I was earlier thinking of some sophisticated way of tracking timestamps of the first and the last segment but later on realized that it won't be that necessary at all to store the timestamp of the last segment. The cases that can occur are basically either: 1) ambiguous => no sensible measurement can be taken anyway 2) non-ambiguous is due to reordering => having the timestamp of the last segment there is just skewing things more off than does some good since the ack got triggered by one of the holes (besides some substle issues that would make determining right hole/skb even harder problem). Anyway, it has nothing to do with this change then. I choose to route some abnormal looking cases with goto noop, some could be handled differently (eg., by stopping the walking at that skb but again). In general, they either shouldn't happen at all or are rare enough to make no difference in practice. In theory this change (as whole) could cause some macroscale regression (global) because of cache misses that are taken over the round-trip time but it gets very likely better because of much less (local) cache misses per other write queue walkers and the big recovery clearing cumulative ack. Worth to note that these benefits would be very easy to get also without TSO/GSO being on as long as the data is in pages so that we can merge them. Currently I won't let that happen because DSACK splitting at fragment that would mess up pcounts due to sk_can_gso in tcp_set_skb_tso_segs. Once DSACKs fragments gets avoided, we have some conditions that can be made less strict. TODO: I will probably have to convert the excessive pointer passing to struct sacktag_state... :-) My testing revealed that considerable amount of skbs couldn't be shifted because they were cloned (most likely still awaiting tx reclaim)... [The rest is considering future work instead since I got repeatably EFAULT to tcpdump's recvfrom when I added pskb_expand_head to deal with clones, so I separated that into another, later patch] ...To counter that, I gave up on the fifth advantage: 5) When growing previous SACK block, less allocs for new skbs are done, basically a new alloc is needed only when new hole is detected and when the previous skb runs out of frags space ...which now only happens of if reclaim is fast enough to dispose the clone before the SACK block comes in (the window is RTT long), otherwise we'll have to alloc some. With clones being handled I got these numbers (will be somewhat worse without that), taken with fine-grained mibs: TCPSackShifted 398 TCPSackMerged 877 TCPSackShiftFallback 320 TCPSACKCOLLAPSEFALLBACKGSO 0 TCPSACKCOLLAPSEFALLBACKSKBBITS 0 TCPSACKCOLLAPSEFALLBACKSKBDATA 0 TCPSACKCOLLAPSEFALLBACKBELOW 0 TCPSACKCOLLAPSEFALLBACKFIRST 1 TCPSACKCOLLAPSEFALLBACKPREVBITS 318 TCPSACKCOLLAPSEFALLBACKMSS 1 TCPSACKCOLLAPSEFALLBACKNOHEAD 0 TCPSACKCOLLAPSEFALLBACKSHIFT 0 TCPSACKCOLLAPSENOOPSEQ 0 TCPSACKCOLLAPSENOOPSMALLPCOUNT 0 TCPSACKCOLLAPSENOOPSMALLLEN 0 TCPSACKCOLLAPSEHOLE 12 Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-11-25 05:20:15 +00:00
extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
int shiftlen);
extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
int len, void *buffer)
{
int hlen = skb_headlen(skb);
if (hlen - offset >= len)
return skb->data + offset;
if (skb_copy_bits(skb, offset, buffer, len) < 0)
return NULL;
return buffer;
}
static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
void *to,
const unsigned int len)
{
memcpy(to, skb->data, len);
}
static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
const int offset, void *to,
const unsigned int len)
{
memcpy(to, skb->data + offset, len);
}
static inline void skb_copy_to_linear_data(struct sk_buff *skb,
const void *from,
const unsigned int len)
{
memcpy(skb->data, from, len);
}
static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
const int offset,
const void *from,
const unsigned int len)
{
memcpy(skb->data + offset, from, len);
}
extern void skb_init(void);
static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
{
return skb->tstamp;
}
/**
* skb_get_timestamp - get timestamp from a skb
* @skb: skb to get stamp from
* @stamp: pointer to struct timeval to store stamp in
*
* Timestamps are stored in the skb as offsets to a base timestamp.
* This function converts the offset back to a struct timeval and stores
* it in stamp.
*/
static inline void skb_get_timestamp(const struct sk_buff *skb,
struct timeval *stamp)
{
*stamp = ktime_to_timeval(skb->tstamp);
}
static inline void skb_get_timestampns(const struct sk_buff *skb,
struct timespec *stamp)
{
*stamp = ktime_to_timespec(skb->tstamp);
}
static inline void __net_timestamp(struct sk_buff *skb)
{
skb->tstamp = ktime_get_real();
}
static inline ktime_t net_timedelta(ktime_t t)
{
return ktime_sub(ktime_get_real(), t);
}
static inline ktime_t net_invalid_timestamp(void)
{
return ktime_set(0, 0);
}
extern void skb_timestamping_init(void);
#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
extern void skb_clone_tx_timestamp(struct sk_buff *skb);
extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
{
}
static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
{
return false;
}
#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
/**
* skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
*
* @skb: clone of the the original outgoing packet
* @hwtstamps: hardware time stamps
*
*/
void skb_complete_tx_timestamp(struct sk_buff *skb,
struct skb_shared_hwtstamps *hwtstamps);
/**
* skb_tstamp_tx - queue clone of skb with send time stamps
* @orig_skb: the original outgoing packet
* @hwtstamps: hardware time stamps, may be NULL if not available
*
* If the skb has a socket associated, then this function clones the
* skb (thus sharing the actual data and optional structures), stores
* the optional hardware time stamping information (if non NULL) or
* generates a software time stamp (otherwise), then queues the clone
* to the error queue of the socket. Errors are silently ignored.
*/
extern void skb_tstamp_tx(struct sk_buff *orig_skb,
struct skb_shared_hwtstamps *hwtstamps);
static inline void sw_tx_timestamp(struct sk_buff *skb)
{
if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
!(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
skb_tstamp_tx(skb, NULL);
}
/**
* skb_tx_timestamp() - Driver hook for transmit timestamping
*
* Ethernet MAC Drivers should call this function in their hard_xmit()
* function as soon as possible after giving the sk_buff to the MAC
* hardware, but before freeing the sk_buff.
*
* @skb: A socket buffer.
*/
static inline void skb_tx_timestamp(struct sk_buff *skb)
{
skb_clone_tx_timestamp(skb);
sw_tx_timestamp(skb);
}
extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
static inline int skb_csum_unnecessary(const struct sk_buff *skb)
{
return skb->ip_summed & CHECKSUM_UNNECESSARY;
}
/**
* skb_checksum_complete - Calculate checksum of an entire packet
* @skb: packet to process
*
* This function calculates the checksum over the entire packet plus
* the value of skb->csum. The latter can be used to supply the
* checksum of a pseudo header as used by TCP/UDP. It returns the
* checksum.
*
* For protocols that contain complete checksums such as ICMP/TCP/UDP,
* this function can be used to verify that checksum on received
* packets. In that case the function should return zero if the
* checksum is correct. In particular, this function will return zero
* if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
* hardware has already verified the correctness of the checksum.
*/
static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
{
return skb_csum_unnecessary(skb) ?
0 : __skb_checksum_complete(skb);
}
#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
static inline void nf_conntrack_put(struct nf_conntrack *nfct)
{
if (nfct && atomic_dec_and_test(&nfct->use))
nf_conntrack_destroy(nfct);
}
static inline void nf_conntrack_get(struct nf_conntrack *nfct)
{
if (nfct)
atomic_inc(&nfct->use);
}
[NETFILTER]: Add nf_conntrack subsystem. The existing connection tracking subsystem in netfilter can only handle ipv4. There were basically two choices present to add connection tracking support for ipv6. We could either duplicate all of the ipv4 connection tracking code into an ipv6 counterpart, or (the choice taken by these patches) we could design a generic layer that could handle both ipv4 and ipv6 and thus requiring only one sub-protocol (TCP, UDP, etc.) connection tracking helper module to be written. In fact nf_conntrack is capable of working with any layer 3 protocol. The existing ipv4 specific conntrack code could also not deal with the pecularities of doing connection tracking on ipv6, which is also cured here. For example, these issues include: 1) ICMPv6 handling, which is used for neighbour discovery in ipv6 thus some messages such as these should not participate in connection tracking since effectively they are like ARP messages 2) fragmentation must be handled differently in ipv6, because the simplistic "defrag, connection track and NAT, refrag" (which the existing ipv4 connection tracking does) approach simply isn't feasible in ipv6 3) ipv6 extension header parsing must occur at the correct spots before and after connection tracking decisions, and there were no provisions for this in the existing connection tracking design 4) ipv6 has no need for stateful NAT The ipv4 specific conntrack layer is kept around, until all of the ipv4 specific conntrack helpers are ported over to nf_conntrack and it is feature complete. Once that occurs, the old conntrack stuff will get placed into the feature-removal-schedule and we will fully kill it off 6 months later. Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp> Signed-off-by: Harald Welte <laforge@netfilter.org> Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-11-10 00:38:16 +00:00
static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
{
if (skb)
atomic_inc(&skb->users);
}
static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
{
if (skb)
kfree_skb(skb);
}
#endif
#ifdef CONFIG_BRIDGE_NETFILTER
static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
{
if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
kfree(nf_bridge);
}
static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
{
if (nf_bridge)
atomic_inc(&nf_bridge->use);
}
#endif /* CONFIG_BRIDGE_NETFILTER */
static inline void nf_reset(struct sk_buff *skb)
{
#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
nf_conntrack_put(skb->nfct);
skb->nfct = NULL;
nf_conntrack_put_reasm(skb->nfct_reasm);
skb->nfct_reasm = NULL;
#endif
#ifdef CONFIG_BRIDGE_NETFILTER
nf_bridge_put(skb->nf_bridge);
skb->nf_bridge = NULL;
#endif
}
/* Note: This doesn't put any conntrack and bridge info in dst. */
static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
{
#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
dst->nfct = src->nfct;
nf_conntrack_get(src->nfct);
dst->nfctinfo = src->nfctinfo;
dst->nfct_reasm = src->nfct_reasm;
nf_conntrack_get_reasm(src->nfct_reasm);
#endif
#ifdef CONFIG_BRIDGE_NETFILTER
dst->nf_bridge = src->nf_bridge;
nf_bridge_get(src->nf_bridge);
#endif
}
static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
{
#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
nf_conntrack_put(dst->nfct);
nf_conntrack_put_reasm(dst->nfct_reasm);
#endif
#ifdef CONFIG_BRIDGE_NETFILTER
nf_bridge_put(dst->nf_bridge);
#endif
__nf_copy(dst, src);
}
#ifdef CONFIG_NETWORK_SECMARK
static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
{
to->secmark = from->secmark;
}
static inline void skb_init_secmark(struct sk_buff *skb)
{
skb->secmark = 0;
}
#else
static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
{ }
static inline void skb_init_secmark(struct sk_buff *skb)
{ }
#endif
static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
{
skb->queue_mapping = queue_mapping;
}
static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
{
return skb->queue_mapping;
}
static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
{
to->queue_mapping = from->queue_mapping;
}
static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
{
skb->queue_mapping = rx_queue + 1;
}
static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
{
return skb->queue_mapping - 1;
}
static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
{
return skb->queue_mapping != 0;
}
extern u16 skb_tx_hash(const struct net_device *dev,
const struct sk_buff *skb);
#ifdef CONFIG_XFRM
static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
{
return skb->sp;
}
#else
static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
{
return NULL;
}
#endif
static inline int skb_is_gso(const struct sk_buff *skb)
{
return skb_shinfo(skb)->gso_size;
}
static inline int skb_is_gso_v6(const struct sk_buff *skb)
{
return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
}
extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
static inline bool skb_warn_if_lro(const struct sk_buff *skb)
{
/* LRO sets gso_size but not gso_type, whereas if GSO is really
* wanted then gso_type will be set. */
struct skb_shared_info *shinfo = skb_shinfo(skb);
if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
unlikely(shinfo->gso_type == 0)) {
__skb_warn_lro_forwarding(skb);
return true;
}
return false;
}
static inline void skb_forward_csum(struct sk_buff *skb)
{
/* Unfortunately we don't support this one. Any brave souls? */
if (skb->ip_summed == CHECKSUM_COMPLETE)
skb->ip_summed = CHECKSUM_NONE;
}
/**
* skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
* @skb: skb to check
*
* fresh skbs have their ip_summed set to CHECKSUM_NONE.
* Instead of forcing ip_summed to CHECKSUM_NONE, we can
* use this helper, to document places where we make this assertion.
*/
static inline void skb_checksum_none_assert(struct sk_buff *skb)
{
#ifdef DEBUG
BUG_ON(skb->ip_summed != CHECKSUM_NONE);
#endif
}
bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
#endif /* __KERNEL__ */
#endif /* _LINUX_SKBUFF_H */