2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * PACKET - implements raw packet sockets.
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Alan Cox, <gw4pts@gw4pts.ampr.org>
13 * Alan Cox : verify_area() now used correctly
14 * Alan Cox : new skbuff lists, look ma no backlogs!
15 * Alan Cox : tidied skbuff lists.
16 * Alan Cox : Now uses generic datagram routines I
17 * added. Also fixed the peek/read crash
18 * from all old Linux datagram code.
19 * Alan Cox : Uses the improved datagram code.
20 * Alan Cox : Added NULL's for socket options.
21 * Alan Cox : Re-commented the code.
22 * Alan Cox : Use new kernel side addressing
23 * Rob Janssen : Correct MTU usage.
24 * Dave Platt : Counter leaks caused by incorrect
25 * interrupt locking and some slightly
26 * dubious gcc output. Can you read
27 * compiler: it said _VOLATILE_
28 * Richard Kooijman : Timestamp fixes.
29 * Alan Cox : New buffers. Use sk->mac.raw.
30 * Alan Cox : sendmsg/recvmsg support.
31 * Alan Cox : Protocol setting support
32 * Alexey Kuznetsov : Untied from IPv4 stack.
33 * Cyrus Durgin : Fixed kerneld for kmod.
34 * Michal Ostrowski : Module initialization cleanup.
35 * Ulises Alonso : Frame number limit removal and
36 * packet_set_ring memory leak.
37 * Eric Biederman : Allow for > 8 byte hardware addresses.
38 * The convention is that longer addresses
39 * will simply extend the hardware address
40 * byte arrays at the end of sockaddr_ll
42 * Johann Baudy : Added TX RING.
44 * This program is free software; you can redistribute it and/or
45 * modify it under the terms of the GNU General Public License
46 * as published by the Free Software Foundation; either version
47 * 2 of the License, or (at your option) any later version.
51 #include <linux/types.h>
53 #include <linux/capability.h>
54 #include <linux/fcntl.h>
55 #include <linux/socket.h>
57 #include <linux/inet.h>
58 #include <linux/netdevice.h>
59 #include <linux/if_packet.h>
60 #include <linux/wireless.h>
61 #include <linux/kernel.h>
62 #include <linux/kmod.h>
63 #include <linux/slab.h>
64 #include <linux/vmalloc.h>
65 #include <net/net_namespace.h>
67 #include <net/protocol.h>
68 #include <linux/skbuff.h>
70 #include <linux/errno.h>
71 #include <linux/timer.h>
72 #include <asm/system.h>
73 #include <asm/uaccess.h>
74 #include <asm/ioctls.h>
76 #include <asm/cacheflush.h>
78 #include <linux/proc_fs.h>
79 #include <linux/seq_file.h>
80 #include <linux/poll.h>
81 #include <linux/module.h>
82 #include <linux/init.h>
83 #include <linux/mutex.h>
84 #include <linux/if_vlan.h>
85 #include <linux/virtio_net.h>
86 #include <linux/errqueue.h>
87 #include <linux/net_tstamp.h>
90 #include <net/inet_common.h>
95 - if device has no dev->hard_header routine, it adds and removes ll header
96 inside itself. In this case ll header is invisible outside of device,
97 but higher levels still should reserve dev->hard_header_len.
98 Some devices are enough clever to reallocate skb, when header
99 will not fit to reserved space (tunnel), another ones are silly
101 - packet socket receives packets with pulled ll header,
102 so that SOCK_RAW should push it back.
107 Incoming, dev->hard_header!=NULL
108 mac_header -> ll header
111 Outgoing, dev->hard_header!=NULL
112 mac_header -> ll header
115 Incoming, dev->hard_header==NULL
116 mac_header -> UNKNOWN position. It is very likely, that it points to ll
117 header. PPP makes it, that is wrong, because introduce
118 assymetry between rx and tx paths.
121 Outgoing, dev->hard_header==NULL
122 mac_header -> data. ll header is still not built!
126 If dev->hard_header==NULL we are unlikely to restore sensible ll header.
132 dev->hard_header != NULL
133 mac_header -> ll header
136 dev->hard_header == NULL (ll header is added by device, we cannot control it)
140 We should set nh.raw on output to correct posistion,
141 packet classifier depends on it.
144 /* Private packet socket structures. */
146 struct packet_mclist
{
147 struct packet_mclist
*next
;
152 unsigned char addr
[MAX_ADDR_LEN
];
154 /* identical to struct packet_mreq except it has
155 * a longer address field.
157 struct packet_mreq_max
{
159 unsigned short mr_type
;
160 unsigned short mr_alen
;
161 unsigned char mr_address
[MAX_ADDR_LEN
];
164 static int packet_set_ring(struct sock
*sk
, struct tpacket_req
*req
,
165 int closing
, int tx_ring
);
171 struct packet_ring_buffer
{
174 unsigned int frames_per_block
;
175 unsigned int frame_size
;
176 unsigned int frame_max
;
178 unsigned int pg_vec_order
;
179 unsigned int pg_vec_pages
;
180 unsigned int pg_vec_len
;
186 static int tpacket_snd(struct packet_sock
*po
, struct msghdr
*msg
);
188 static void packet_flush_mclist(struct sock
*sk
);
191 /* struct sock has to be the first member of packet_sock */
193 struct tpacket_stats stats
;
194 struct packet_ring_buffer rx_ring
;
195 struct packet_ring_buffer tx_ring
;
197 spinlock_t bind_lock
;
198 struct mutex pg_vec_lock
;
199 unsigned int running
:1, /* prot_hook is attached*/
203 int ifindex
; /* bound device */
205 struct packet_mclist
*mclist
;
207 enum tpacket_versions tp_version
;
208 unsigned int tp_hdrlen
;
209 unsigned int tp_reserve
;
210 unsigned int tp_loss
:1;
211 unsigned int tp_tstamp
;
212 struct packet_type prot_hook ____cacheline_aligned_in_smp
;
215 struct packet_skb_cb
{
216 unsigned int origlen
;
218 struct sockaddr_pkt pkt
;
219 struct sockaddr_ll ll
;
223 #define PACKET_SKB_CB(__skb) ((struct packet_skb_cb *)((__skb)->cb))
225 static inline __pure
struct page
*pgv_to_page(void *addr
)
227 if (is_vmalloc_addr(addr
))
228 return vmalloc_to_page(addr
);
229 return virt_to_page(addr
);
232 static void __packet_set_status(struct packet_sock
*po
, void *frame
, int status
)
235 struct tpacket_hdr
*h1
;
236 struct tpacket2_hdr
*h2
;
241 switch (po
->tp_version
) {
243 h
.h1
->tp_status
= status
;
244 flush_dcache_page(pgv_to_page(&h
.h1
->tp_status
));
247 h
.h2
->tp_status
= status
;
248 flush_dcache_page(pgv_to_page(&h
.h2
->tp_status
));
251 pr_err("TPACKET version not supported\n");
258 static int __packet_get_status(struct packet_sock
*po
, void *frame
)
261 struct tpacket_hdr
*h1
;
262 struct tpacket2_hdr
*h2
;
269 switch (po
->tp_version
) {
271 flush_dcache_page(pgv_to_page(&h
.h1
->tp_status
));
272 return h
.h1
->tp_status
;
274 flush_dcache_page(pgv_to_page(&h
.h2
->tp_status
));
275 return h
.h2
->tp_status
;
277 pr_err("TPACKET version not supported\n");
283 static void *packet_lookup_frame(struct packet_sock
*po
,
284 struct packet_ring_buffer
*rb
,
285 unsigned int position
,
288 unsigned int pg_vec_pos
, frame_offset
;
290 struct tpacket_hdr
*h1
;
291 struct tpacket2_hdr
*h2
;
295 pg_vec_pos
= position
/ rb
->frames_per_block
;
296 frame_offset
= position
% rb
->frames_per_block
;
298 h
.raw
= rb
->pg_vec
[pg_vec_pos
].buffer
+
299 (frame_offset
* rb
->frame_size
);
301 if (status
!= __packet_get_status(po
, h
.raw
))
307 static inline void *packet_current_frame(struct packet_sock
*po
,
308 struct packet_ring_buffer
*rb
,
311 return packet_lookup_frame(po
, rb
, rb
->head
, status
);
314 static inline void *packet_previous_frame(struct packet_sock
*po
,
315 struct packet_ring_buffer
*rb
,
318 unsigned int previous
= rb
->head
? rb
->head
- 1 : rb
->frame_max
;
319 return packet_lookup_frame(po
, rb
, previous
, status
);
322 static inline void packet_increment_head(struct packet_ring_buffer
*buff
)
324 buff
->head
= buff
->head
!= buff
->frame_max
? buff
->head
+1 : 0;
327 static inline struct packet_sock
*pkt_sk(struct sock
*sk
)
329 return (struct packet_sock
*)sk
;
332 static void packet_sock_destruct(struct sock
*sk
)
334 skb_queue_purge(&sk
->sk_error_queue
);
336 WARN_ON(atomic_read(&sk
->sk_rmem_alloc
));
337 WARN_ON(atomic_read(&sk
->sk_wmem_alloc
));
339 if (!sock_flag(sk
, SOCK_DEAD
)) {
340 pr_err("Attempt to release alive packet socket: %p\n", sk
);
344 sk_refcnt_debug_dec(sk
);
348 static const struct proto_ops packet_ops
;
350 static const struct proto_ops packet_ops_spkt
;
352 static int packet_rcv_spkt(struct sk_buff
*skb
, struct net_device
*dev
,
353 struct packet_type
*pt
, struct net_device
*orig_dev
)
356 struct sockaddr_pkt
*spkt
;
359 * When we registered the protocol we saved the socket in the data
360 * field for just this event.
363 sk
= pt
->af_packet_priv
;
366 * Yank back the headers [hope the device set this
367 * right or kerboom...]
369 * Incoming packets have ll header pulled,
372 * For outgoing ones skb->data == skb_mac_header(skb)
373 * so that this procedure is noop.
376 if (skb
->pkt_type
== PACKET_LOOPBACK
)
379 if (!net_eq(dev_net(dev
), sock_net(sk
)))
382 skb
= skb_share_check(skb
, GFP_ATOMIC
);
386 /* drop any routing info */
389 /* drop conntrack reference */
392 spkt
= &PACKET_SKB_CB(skb
)->sa
.pkt
;
394 skb_push(skb
, skb
->data
- skb_mac_header(skb
));
397 * The SOCK_PACKET socket receives _all_ frames.
400 spkt
->spkt_family
= dev
->type
;
401 strlcpy(spkt
->spkt_device
, dev
->name
, sizeof(spkt
->spkt_device
));
402 spkt
->spkt_protocol
= skb
->protocol
;
405 * Charge the memory to the socket. This is done specifically
406 * to prevent sockets using all the memory up.
409 if (sock_queue_rcv_skb(sk
, skb
) == 0)
420 * Output a raw packet to a device layer. This bypasses all the other
421 * protocol layers and you must therefore supply it with a complete frame
424 static int packet_sendmsg_spkt(struct kiocb
*iocb
, struct socket
*sock
,
425 struct msghdr
*msg
, size_t len
)
427 struct sock
*sk
= sock
->sk
;
428 struct sockaddr_pkt
*saddr
= (struct sockaddr_pkt
*)msg
->msg_name
;
429 struct sk_buff
*skb
= NULL
;
430 struct net_device
*dev
;
435 * Get and verify the address.
439 if (msg
->msg_namelen
< sizeof(struct sockaddr
))
441 if (msg
->msg_namelen
== sizeof(struct sockaddr_pkt
))
442 proto
= saddr
->spkt_protocol
;
444 return -ENOTCONN
; /* SOCK_PACKET must be sent giving an address */
447 * Find the device first to size check it
450 saddr
->spkt_device
[13] = 0;
453 dev
= dev_get_by_name_rcu(sock_net(sk
), saddr
->spkt_device
);
459 if (!(dev
->flags
& IFF_UP
))
463 * You may not queue a frame bigger than the mtu. This is the lowest level
464 * raw protocol and you must do your own fragmentation at this level.
468 if (len
> dev
->mtu
+ dev
->hard_header_len
+ VLAN_HLEN
)
472 size_t reserved
= LL_RESERVED_SPACE(dev
);
473 unsigned int hhlen
= dev
->header_ops
? dev
->hard_header_len
: 0;
476 skb
= sock_wmalloc(sk
, len
+ reserved
, 0, GFP_KERNEL
);
479 /* FIXME: Save some space for broken drivers that write a hard
480 * header at transmission time by themselves. PPP is the notable
481 * one here. This should really be fixed at the driver level.
483 skb_reserve(skb
, reserved
);
484 skb_reset_network_header(skb
);
486 /* Try to align data part correctly */
491 skb_reset_network_header(skb
);
493 err
= memcpy_fromiovec(skb_put(skb
, len
), msg
->msg_iov
, len
);
499 if (len
> (dev
->mtu
+ dev
->hard_header_len
)) {
500 /* Earlier code assumed this would be a VLAN pkt,
501 * double-check this now that we have the actual
505 skb_reset_mac_header(skb
);
507 if (ehdr
->h_proto
!= htons(ETH_P_8021Q
)) {
513 skb
->protocol
= proto
;
515 skb
->priority
= sk
->sk_priority
;
516 skb
->mark
= sk
->sk_mark
;
517 err
= sock_tx_timestamp(sk
, &skb_shinfo(skb
)->tx_flags
);
532 static inline unsigned int run_filter(const struct sk_buff
*skb
,
533 const struct sock
*sk
,
536 struct sk_filter
*filter
;
539 filter
= rcu_dereference(sk
->sk_filter
);
541 res
= SK_RUN_FILTER(filter
, skb
);
548 * This function makes lazy skb cloning in hope that most of packets
549 * are discarded by BPF.
551 * Note tricky part: we DO mangle shared skb! skb->data, skb->len
552 * and skb->cb are mangled. It works because (and until) packets
553 * falling here are owned by current CPU. Output packets are cloned
554 * by dev_queue_xmit_nit(), input packets are processed by net_bh
555 * sequencially, so that if we return skb to original state on exit,
556 * we will not harm anyone.
559 static int packet_rcv(struct sk_buff
*skb
, struct net_device
*dev
,
560 struct packet_type
*pt
, struct net_device
*orig_dev
)
563 struct sockaddr_ll
*sll
;
564 struct packet_sock
*po
;
565 u8
*skb_head
= skb
->data
;
566 int skb_len
= skb
->len
;
567 unsigned int snaplen
, res
;
569 if (skb
->pkt_type
== PACKET_LOOPBACK
)
572 sk
= pt
->af_packet_priv
;
575 if (!net_eq(dev_net(dev
), sock_net(sk
)))
580 if (dev
->header_ops
) {
581 /* The device has an explicit notion of ll header,
582 * exported to higher levels.
584 * Otherwise, the device hides details of its frame
585 * structure, so that corresponding packet head is
586 * never delivered to user.
588 if (sk
->sk_type
!= SOCK_DGRAM
)
589 skb_push(skb
, skb
->data
- skb_mac_header(skb
));
590 else if (skb
->pkt_type
== PACKET_OUTGOING
) {
591 /* Special case: outgoing packets have ll header at head */
592 skb_pull(skb
, skb_network_offset(skb
));
598 res
= run_filter(skb
, sk
, snaplen
);
604 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
605 (unsigned)sk
->sk_rcvbuf
)
608 if (skb_shared(skb
)) {
609 struct sk_buff
*nskb
= skb_clone(skb
, GFP_ATOMIC
);
613 if (skb_head
!= skb
->data
) {
614 skb
->data
= skb_head
;
621 BUILD_BUG_ON(sizeof(*PACKET_SKB_CB(skb
)) + MAX_ADDR_LEN
- 8 >
624 sll
= &PACKET_SKB_CB(skb
)->sa
.ll
;
625 sll
->sll_family
= AF_PACKET
;
626 sll
->sll_hatype
= dev
->type
;
627 sll
->sll_protocol
= skb
->protocol
;
628 sll
->sll_pkttype
= skb
->pkt_type
;
629 if (unlikely(po
->origdev
))
630 sll
->sll_ifindex
= orig_dev
->ifindex
;
632 sll
->sll_ifindex
= dev
->ifindex
;
634 sll
->sll_halen
= dev_parse_header(skb
, sll
->sll_addr
);
636 PACKET_SKB_CB(skb
)->origlen
= skb
->len
;
638 if (pskb_trim(skb
, snaplen
))
641 skb_set_owner_r(skb
, sk
);
645 /* drop conntrack reference */
648 spin_lock(&sk
->sk_receive_queue
.lock
);
649 po
->stats
.tp_packets
++;
650 skb
->dropcount
= atomic_read(&sk
->sk_drops
);
651 __skb_queue_tail(&sk
->sk_receive_queue
, skb
);
652 spin_unlock(&sk
->sk_receive_queue
.lock
);
653 sk
->sk_data_ready(sk
, skb
->len
);
657 po
->stats
.tp_drops
= atomic_inc_return(&sk
->sk_drops
);
660 if (skb_head
!= skb
->data
&& skb_shared(skb
)) {
661 skb
->data
= skb_head
;
669 static int tpacket_rcv(struct sk_buff
*skb
, struct net_device
*dev
,
670 struct packet_type
*pt
, struct net_device
*orig_dev
)
673 struct packet_sock
*po
;
674 struct sockaddr_ll
*sll
;
676 struct tpacket_hdr
*h1
;
677 struct tpacket2_hdr
*h2
;
680 u8
*skb_head
= skb
->data
;
681 int skb_len
= skb
->len
;
682 unsigned int snaplen
, res
;
683 unsigned long status
= TP_STATUS_LOSING
|TP_STATUS_USER
;
684 unsigned short macoff
, netoff
, hdrlen
;
685 struct sk_buff
*copy_skb
= NULL
;
688 struct skb_shared_hwtstamps
*shhwtstamps
= skb_hwtstamps(skb
);
690 if (skb
->pkt_type
== PACKET_LOOPBACK
)
693 sk
= pt
->af_packet_priv
;
696 if (!net_eq(dev_net(dev
), sock_net(sk
)))
699 if (dev
->header_ops
) {
700 if (sk
->sk_type
!= SOCK_DGRAM
)
701 skb_push(skb
, skb
->data
- skb_mac_header(skb
));
702 else if (skb
->pkt_type
== PACKET_OUTGOING
) {
703 /* Special case: outgoing packets have ll header at head */
704 skb_pull(skb
, skb_network_offset(skb
));
708 if (skb
->ip_summed
== CHECKSUM_PARTIAL
)
709 status
|= TP_STATUS_CSUMNOTREADY
;
713 res
= run_filter(skb
, sk
, snaplen
);
719 if (sk
->sk_type
== SOCK_DGRAM
) {
720 macoff
= netoff
= TPACKET_ALIGN(po
->tp_hdrlen
) + 16 +
723 unsigned maclen
= skb_network_offset(skb
);
724 netoff
= TPACKET_ALIGN(po
->tp_hdrlen
+
725 (maclen
< 16 ? 16 : maclen
)) +
727 macoff
= netoff
- maclen
;
730 if (macoff
+ snaplen
> po
->rx_ring
.frame_size
) {
731 if (po
->copy_thresh
&&
732 atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
<
733 (unsigned)sk
->sk_rcvbuf
) {
734 if (skb_shared(skb
)) {
735 copy_skb
= skb_clone(skb
, GFP_ATOMIC
);
737 copy_skb
= skb_get(skb
);
738 skb_head
= skb
->data
;
741 skb_set_owner_r(copy_skb
, sk
);
743 snaplen
= po
->rx_ring
.frame_size
- macoff
;
744 if ((int)snaplen
< 0)
748 spin_lock(&sk
->sk_receive_queue
.lock
);
749 h
.raw
= packet_current_frame(po
, &po
->rx_ring
, TP_STATUS_KERNEL
);
752 packet_increment_head(&po
->rx_ring
);
753 po
->stats
.tp_packets
++;
755 status
|= TP_STATUS_COPY
;
756 __skb_queue_tail(&sk
->sk_receive_queue
, copy_skb
);
758 if (!po
->stats
.tp_drops
)
759 status
&= ~TP_STATUS_LOSING
;
760 spin_unlock(&sk
->sk_receive_queue
.lock
);
762 skb_copy_bits(skb
, 0, h
.raw
+ macoff
, snaplen
);
764 switch (po
->tp_version
) {
766 h
.h1
->tp_len
= skb
->len
;
767 h
.h1
->tp_snaplen
= snaplen
;
768 h
.h1
->tp_mac
= macoff
;
769 h
.h1
->tp_net
= netoff
;
770 if ((po
->tp_tstamp
& SOF_TIMESTAMPING_SYS_HARDWARE
)
771 && shhwtstamps
->syststamp
.tv64
)
772 tv
= ktime_to_timeval(shhwtstamps
->syststamp
);
773 else if ((po
->tp_tstamp
& SOF_TIMESTAMPING_RAW_HARDWARE
)
774 && shhwtstamps
->hwtstamp
.tv64
)
775 tv
= ktime_to_timeval(shhwtstamps
->hwtstamp
);
776 else if (skb
->tstamp
.tv64
)
777 tv
= ktime_to_timeval(skb
->tstamp
);
779 do_gettimeofday(&tv
);
780 h
.h1
->tp_sec
= tv
.tv_sec
;
781 h
.h1
->tp_usec
= tv
.tv_usec
;
782 hdrlen
= sizeof(*h
.h1
);
785 h
.h2
->tp_len
= skb
->len
;
786 h
.h2
->tp_snaplen
= snaplen
;
787 h
.h2
->tp_mac
= macoff
;
788 h
.h2
->tp_net
= netoff
;
789 if ((po
->tp_tstamp
& SOF_TIMESTAMPING_SYS_HARDWARE
)
790 && shhwtstamps
->syststamp
.tv64
)
791 ts
= ktime_to_timespec(shhwtstamps
->syststamp
);
792 else if ((po
->tp_tstamp
& SOF_TIMESTAMPING_RAW_HARDWARE
)
793 && shhwtstamps
->hwtstamp
.tv64
)
794 ts
= ktime_to_timespec(shhwtstamps
->hwtstamp
);
795 else if (skb
->tstamp
.tv64
)
796 ts
= ktime_to_timespec(skb
->tstamp
);
799 h
.h2
->tp_sec
= ts
.tv_sec
;
800 h
.h2
->tp_nsec
= ts
.tv_nsec
;
801 if (vlan_tx_tag_present(skb
)) {
802 h
.h2
->tp_vlan_tci
= vlan_tx_tag_get(skb
);
803 status
|= TP_STATUS_VLAN_VALID
;
805 h
.h2
->tp_vlan_tci
= 0;
807 h
.h2
->tp_padding
= 0;
808 hdrlen
= sizeof(*h
.h2
);
814 sll
= h
.raw
+ TPACKET_ALIGN(hdrlen
);
815 sll
->sll_halen
= dev_parse_header(skb
, sll
->sll_addr
);
816 sll
->sll_family
= AF_PACKET
;
817 sll
->sll_hatype
= dev
->type
;
818 sll
->sll_protocol
= skb
->protocol
;
819 sll
->sll_pkttype
= skb
->pkt_type
;
820 if (unlikely(po
->origdev
))
821 sll
->sll_ifindex
= orig_dev
->ifindex
;
823 sll
->sll_ifindex
= dev
->ifindex
;
825 __packet_set_status(po
, h
.raw
, status
);
827 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1
831 end
= (u8
*)PAGE_ALIGN((unsigned long)h
.raw
+ macoff
+ snaplen
);
832 for (start
= h
.raw
; start
< end
; start
+= PAGE_SIZE
)
833 flush_dcache_page(pgv_to_page(start
));
837 sk
->sk_data_ready(sk
, 0);
840 if (skb_head
!= skb
->data
&& skb_shared(skb
)) {
841 skb
->data
= skb_head
;
849 po
->stats
.tp_drops
++;
850 spin_unlock(&sk
->sk_receive_queue
.lock
);
852 sk
->sk_data_ready(sk
, 0);
857 static void tpacket_destruct_skb(struct sk_buff
*skb
)
859 struct packet_sock
*po
= pkt_sk(skb
->sk
);
864 if (likely(po
->tx_ring
.pg_vec
)) {
865 ph
= skb_shinfo(skb
)->destructor_arg
;
866 BUG_ON(__packet_get_status(po
, ph
) != TP_STATUS_SENDING
);
867 BUG_ON(atomic_read(&po
->tx_ring
.pending
) == 0);
868 atomic_dec(&po
->tx_ring
.pending
);
869 __packet_set_status(po
, ph
, TP_STATUS_AVAILABLE
);
875 static int tpacket_fill_skb(struct packet_sock
*po
, struct sk_buff
*skb
,
876 void *frame
, struct net_device
*dev
, int size_max
,
877 __be16 proto
, unsigned char *addr
)
880 struct tpacket_hdr
*h1
;
881 struct tpacket2_hdr
*h2
;
884 int to_write
, offset
, len
, tp_len
, nr_frags
, len_max
;
885 struct socket
*sock
= po
->sk
.sk_socket
;
892 skb
->protocol
= proto
;
894 skb
->priority
= po
->sk
.sk_priority
;
895 skb
->mark
= po
->sk
.sk_mark
;
896 skb_shinfo(skb
)->destructor_arg
= ph
.raw
;
898 switch (po
->tp_version
) {
900 tp_len
= ph
.h2
->tp_len
;
903 tp_len
= ph
.h1
->tp_len
;
906 if (unlikely(tp_len
> size_max
)) {
907 pr_err("packet size is too long (%d > %d)\n", tp_len
, size_max
);
911 skb_reserve(skb
, LL_RESERVED_SPACE(dev
));
912 skb_reset_network_header(skb
);
914 data
= ph
.raw
+ po
->tp_hdrlen
- sizeof(struct sockaddr_ll
);
917 if (sock
->type
== SOCK_DGRAM
) {
918 err
= dev_hard_header(skb
, dev
, ntohs(proto
), addr
,
920 if (unlikely(err
< 0))
922 } else if (dev
->hard_header_len
) {
923 /* net device doesn't like empty head */
924 if (unlikely(tp_len
<= dev
->hard_header_len
)) {
925 pr_err("packet size is too short (%d < %d)\n",
926 tp_len
, dev
->hard_header_len
);
930 skb_push(skb
, dev
->hard_header_len
);
931 err
= skb_store_bits(skb
, 0, data
,
932 dev
->hard_header_len
);
936 data
+= dev
->hard_header_len
;
937 to_write
-= dev
->hard_header_len
;
941 offset
= offset_in_page(data
);
942 len_max
= PAGE_SIZE
- offset
;
943 len
= ((to_write
> len_max
) ? len_max
: to_write
);
945 skb
->data_len
= to_write
;
946 skb
->len
+= to_write
;
947 skb
->truesize
+= to_write
;
948 atomic_add(to_write
, &po
->sk
.sk_wmem_alloc
);
950 while (likely(to_write
)) {
951 nr_frags
= skb_shinfo(skb
)->nr_frags
;
953 if (unlikely(nr_frags
>= MAX_SKB_FRAGS
)) {
954 pr_err("Packet exceed the number of skb frags(%lu)\n",
959 page
= pgv_to_page(data
);
961 flush_dcache_page(page
);
963 skb_fill_page_desc(skb
, nr_frags
, page
, offset
, len
);
967 len
= ((to_write
> len_max
) ? len_max
: to_write
);
973 static int tpacket_snd(struct packet_sock
*po
, struct msghdr
*msg
)
976 struct net_device
*dev
;
978 int ifindex
, err
, reserve
= 0;
980 struct sockaddr_ll
*saddr
= (struct sockaddr_ll
*)msg
->msg_name
;
981 int tp_len
, size_max
;
986 mutex_lock(&po
->pg_vec_lock
);
990 ifindex
= po
->ifindex
;
995 if (msg
->msg_namelen
< sizeof(struct sockaddr_ll
))
997 if (msg
->msg_namelen
< (saddr
->sll_halen
998 + offsetof(struct sockaddr_ll
,
1001 ifindex
= saddr
->sll_ifindex
;
1002 proto
= saddr
->sll_protocol
;
1003 addr
= saddr
->sll_addr
;
1006 dev
= dev_get_by_index(sock_net(&po
->sk
), ifindex
);
1008 if (unlikely(dev
== NULL
))
1011 reserve
= dev
->hard_header_len
;
1014 if (unlikely(!(dev
->flags
& IFF_UP
)))
1017 size_max
= po
->tx_ring
.frame_size
1018 - (po
->tp_hdrlen
- sizeof(struct sockaddr_ll
));
1020 if (size_max
> dev
->mtu
+ reserve
)
1021 size_max
= dev
->mtu
+ reserve
;
1024 ph
= packet_current_frame(po
, &po
->tx_ring
,
1025 TP_STATUS_SEND_REQUEST
);
1027 if (unlikely(ph
== NULL
)) {
1032 status
= TP_STATUS_SEND_REQUEST
;
1033 skb
= sock_alloc_send_skb(&po
->sk
,
1034 LL_ALLOCATED_SPACE(dev
)
1035 + sizeof(struct sockaddr_ll
),
1038 if (unlikely(skb
== NULL
))
1041 tp_len
= tpacket_fill_skb(po
, skb
, ph
, dev
, size_max
, proto
,
1044 if (unlikely(tp_len
< 0)) {
1046 __packet_set_status(po
, ph
,
1047 TP_STATUS_AVAILABLE
);
1048 packet_increment_head(&po
->tx_ring
);
1052 status
= TP_STATUS_WRONG_FORMAT
;
1058 skb
->destructor
= tpacket_destruct_skb
;
1059 __packet_set_status(po
, ph
, TP_STATUS_SENDING
);
1060 atomic_inc(&po
->tx_ring
.pending
);
1062 status
= TP_STATUS_SEND_REQUEST
;
1063 err
= dev_queue_xmit(skb
);
1064 if (unlikely(err
> 0)) {
1065 err
= net_xmit_errno(err
);
1066 if (err
&& __packet_get_status(po
, ph
) ==
1067 TP_STATUS_AVAILABLE
) {
1068 /* skb was destructed already */
1073 * skb was dropped but not destructed yet;
1074 * let's treat it like congestion or err < 0
1078 packet_increment_head(&po
->tx_ring
);
1080 } while (likely((ph
!= NULL
) ||
1081 ((!(msg
->msg_flags
& MSG_DONTWAIT
)) &&
1082 (atomic_read(&po
->tx_ring
.pending
))))
1089 __packet_set_status(po
, ph
, status
);
1094 mutex_unlock(&po
->pg_vec_lock
);
1098 static inline struct sk_buff
*packet_alloc_skb(struct sock
*sk
, size_t prepad
,
1099 size_t reserve
, size_t len
,
1100 size_t linear
, int noblock
,
1103 struct sk_buff
*skb
;
1105 /* Under a page? Don't bother with paged skb. */
1106 if (prepad
+ len
< PAGE_SIZE
|| !linear
)
1109 skb
= sock_alloc_send_pskb(sk
, prepad
+ linear
, len
- linear
, noblock
,
1114 skb_reserve(skb
, reserve
);
1115 skb_put(skb
, linear
);
1116 skb
->data_len
= len
- linear
;
1117 skb
->len
+= len
- linear
;
1122 static int packet_snd(struct socket
*sock
,
1123 struct msghdr
*msg
, size_t len
)
1125 struct sock
*sk
= sock
->sk
;
1126 struct sockaddr_ll
*saddr
= (struct sockaddr_ll
*)msg
->msg_name
;
1127 struct sk_buff
*skb
;
1128 struct net_device
*dev
;
1130 unsigned char *addr
;
1131 int ifindex
, err
, reserve
= 0;
1132 struct virtio_net_hdr vnet_hdr
= { 0 };
1135 struct packet_sock
*po
= pkt_sk(sk
);
1136 unsigned short gso_type
= 0;
1139 * Get and verify the address.
1142 if (saddr
== NULL
) {
1143 ifindex
= po
->ifindex
;
1148 if (msg
->msg_namelen
< sizeof(struct sockaddr_ll
))
1150 if (msg
->msg_namelen
< (saddr
->sll_halen
+ offsetof(struct sockaddr_ll
, sll_addr
)))
1152 ifindex
= saddr
->sll_ifindex
;
1153 proto
= saddr
->sll_protocol
;
1154 addr
= saddr
->sll_addr
;
1158 dev
= dev_get_by_index(sock_net(sk
), ifindex
);
1162 if (sock
->type
== SOCK_RAW
)
1163 reserve
= dev
->hard_header_len
;
1166 if (!(dev
->flags
& IFF_UP
))
1169 if (po
->has_vnet_hdr
) {
1170 vnet_hdr_len
= sizeof(vnet_hdr
);
1173 if (len
< vnet_hdr_len
)
1176 len
-= vnet_hdr_len
;
1178 err
= memcpy_fromiovec((void *)&vnet_hdr
, msg
->msg_iov
,
1183 if ((vnet_hdr
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) &&
1184 (vnet_hdr
.csum_start
+ vnet_hdr
.csum_offset
+ 2 >
1186 vnet_hdr
.hdr_len
= vnet_hdr
.csum_start
+
1187 vnet_hdr
.csum_offset
+ 2;
1190 if (vnet_hdr
.hdr_len
> len
)
1193 if (vnet_hdr
.gso_type
!= VIRTIO_NET_HDR_GSO_NONE
) {
1194 switch (vnet_hdr
.gso_type
& ~VIRTIO_NET_HDR_GSO_ECN
) {
1195 case VIRTIO_NET_HDR_GSO_TCPV4
:
1196 gso_type
= SKB_GSO_TCPV4
;
1198 case VIRTIO_NET_HDR_GSO_TCPV6
:
1199 gso_type
= SKB_GSO_TCPV6
;
1201 case VIRTIO_NET_HDR_GSO_UDP
:
1202 gso_type
= SKB_GSO_UDP
;
1208 if (vnet_hdr
.gso_type
& VIRTIO_NET_HDR_GSO_ECN
)
1209 gso_type
|= SKB_GSO_TCP_ECN
;
1211 if (vnet_hdr
.gso_size
== 0)
1218 if (!gso_type
&& (len
> dev
->mtu
+ reserve
+ VLAN_HLEN
))
1222 skb
= packet_alloc_skb(sk
, LL_ALLOCATED_SPACE(dev
),
1223 LL_RESERVED_SPACE(dev
), len
, vnet_hdr
.hdr_len
,
1224 msg
->msg_flags
& MSG_DONTWAIT
, &err
);
1228 skb_set_network_header(skb
, reserve
);
1231 if (sock
->type
== SOCK_DGRAM
&&
1232 (offset
= dev_hard_header(skb
, dev
, ntohs(proto
), addr
, NULL
, len
)) < 0)
1235 /* Returns -EFAULT on error */
1236 err
= skb_copy_datagram_from_iovec(skb
, offset
, msg
->msg_iov
, 0, len
);
1239 err
= sock_tx_timestamp(sk
, &skb_shinfo(skb
)->tx_flags
);
1243 if (!gso_type
&& (len
> dev
->mtu
+ reserve
)) {
1244 /* Earlier code assumed this would be a VLAN pkt,
1245 * double-check this now that we have the actual
1248 struct ethhdr
*ehdr
;
1249 skb_reset_mac_header(skb
);
1250 ehdr
= eth_hdr(skb
);
1251 if (ehdr
->h_proto
!= htons(ETH_P_8021Q
)) {
1257 skb
->protocol
= proto
;
1259 skb
->priority
= sk
->sk_priority
;
1260 skb
->mark
= sk
->sk_mark
;
1262 if (po
->has_vnet_hdr
) {
1263 if (vnet_hdr
.flags
& VIRTIO_NET_HDR_F_NEEDS_CSUM
) {
1264 if (!skb_partial_csum_set(skb
, vnet_hdr
.csum_start
,
1265 vnet_hdr
.csum_offset
)) {
1271 skb_shinfo(skb
)->gso_size
= vnet_hdr
.gso_size
;
1272 skb_shinfo(skb
)->gso_type
= gso_type
;
1274 /* Header must be checked, and gso_segs computed. */
1275 skb_shinfo(skb
)->gso_type
|= SKB_GSO_DODGY
;
1276 skb_shinfo(skb
)->gso_segs
= 0;
1278 len
+= vnet_hdr_len
;
1285 err
= dev_queue_xmit(skb
);
1286 if (err
> 0 && (err
= net_xmit_errno(err
)) != 0)
1302 static int packet_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
1303 struct msghdr
*msg
, size_t len
)
1305 struct sock
*sk
= sock
->sk
;
1306 struct packet_sock
*po
= pkt_sk(sk
);
1307 if (po
->tx_ring
.pg_vec
)
1308 return tpacket_snd(po
, msg
);
1310 return packet_snd(sock
, msg
, len
);
1314 * Close a PACKET socket. This is fairly simple. We immediately go
1315 * to 'closed' state and remove our protocol entry in the device list.
1318 static int packet_release(struct socket
*sock
)
1320 struct sock
*sk
= sock
->sk
;
1321 struct packet_sock
*po
;
1323 struct tpacket_req req
;
1331 spin_lock_bh(&net
->packet
.sklist_lock
);
1332 sk_del_node_init_rcu(sk
);
1333 sock_prot_inuse_add(net
, sk
->sk_prot
, -1);
1334 spin_unlock_bh(&net
->packet
.sklist_lock
);
1336 spin_lock(&po
->bind_lock
);
1339 * Remove from protocol table
1343 __dev_remove_pack(&po
->prot_hook
);
1346 spin_unlock(&po
->bind_lock
);
1348 packet_flush_mclist(sk
);
1350 memset(&req
, 0, sizeof(req
));
1352 if (po
->rx_ring
.pg_vec
)
1353 packet_set_ring(sk
, &req
, 1, 0);
1355 if (po
->tx_ring
.pg_vec
)
1356 packet_set_ring(sk
, &req
, 1, 1);
1360 * Now the socket is dead. No more input will appear.
1367 skb_queue_purge(&sk
->sk_receive_queue
);
1368 sk_refcnt_debug_release(sk
);
1375 * Attach a packet hook.
1378 static int packet_do_bind(struct sock
*sk
, struct net_device
*dev
, __be16 protocol
)
1380 struct packet_sock
*po
= pkt_sk(sk
);
1382 * Detach an existing hook if present.
1387 spin_lock(&po
->bind_lock
);
1392 spin_unlock(&po
->bind_lock
);
1393 dev_remove_pack(&po
->prot_hook
);
1394 spin_lock(&po
->bind_lock
);
1398 po
->prot_hook
.type
= protocol
;
1399 po
->prot_hook
.dev
= dev
;
1401 po
->ifindex
= dev
? dev
->ifindex
: 0;
1406 if (!dev
|| (dev
->flags
& IFF_UP
)) {
1407 dev_add_pack(&po
->prot_hook
);
1411 sk
->sk_err
= ENETDOWN
;
1412 if (!sock_flag(sk
, SOCK_DEAD
))
1413 sk
->sk_error_report(sk
);
1417 spin_unlock(&po
->bind_lock
);
1423 * Bind a packet socket to a device
1426 static int packet_bind_spkt(struct socket
*sock
, struct sockaddr
*uaddr
,
1429 struct sock
*sk
= sock
->sk
;
1431 struct net_device
*dev
;
1438 if (addr_len
!= sizeof(struct sockaddr
))
1440 strlcpy(name
, uaddr
->sa_data
, sizeof(name
));
1442 dev
= dev_get_by_name(sock_net(sk
), name
);
1444 err
= packet_do_bind(sk
, dev
, pkt_sk(sk
)->num
);
1450 static int packet_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
1452 struct sockaddr_ll
*sll
= (struct sockaddr_ll
*)uaddr
;
1453 struct sock
*sk
= sock
->sk
;
1454 struct net_device
*dev
= NULL
;
1462 if (addr_len
< sizeof(struct sockaddr_ll
))
1464 if (sll
->sll_family
!= AF_PACKET
)
1467 if (sll
->sll_ifindex
) {
1469 dev
= dev_get_by_index(sock_net(sk
), sll
->sll_ifindex
);
1473 err
= packet_do_bind(sk
, dev
, sll
->sll_protocol
? : pkt_sk(sk
)->num
);
1481 static struct proto packet_proto
= {
1483 .owner
= THIS_MODULE
,
1484 .obj_size
= sizeof(struct packet_sock
),
1488 * Create a packet of type SOCK_PACKET.
1491 static int packet_create(struct net
*net
, struct socket
*sock
, int protocol
,
1495 struct packet_sock
*po
;
1496 __be16 proto
= (__force __be16
)protocol
; /* weird, but documented */
1499 if (!capable(CAP_NET_RAW
))
1501 if (sock
->type
!= SOCK_DGRAM
&& sock
->type
!= SOCK_RAW
&&
1502 sock
->type
!= SOCK_PACKET
)
1503 return -ESOCKTNOSUPPORT
;
1505 sock
->state
= SS_UNCONNECTED
;
1508 sk
= sk_alloc(net
, PF_PACKET
, GFP_KERNEL
, &packet_proto
);
1512 sock
->ops
= &packet_ops
;
1513 if (sock
->type
== SOCK_PACKET
)
1514 sock
->ops
= &packet_ops_spkt
;
1516 sock_init_data(sock
, sk
);
1519 sk
->sk_family
= PF_PACKET
;
1522 sk
->sk_destruct
= packet_sock_destruct
;
1523 sk_refcnt_debug_inc(sk
);
1526 * Attach a protocol block
1529 spin_lock_init(&po
->bind_lock
);
1530 mutex_init(&po
->pg_vec_lock
);
1531 po
->prot_hook
.func
= packet_rcv
;
1533 if (sock
->type
== SOCK_PACKET
)
1534 po
->prot_hook
.func
= packet_rcv_spkt
;
1536 po
->prot_hook
.af_packet_priv
= sk
;
1539 po
->prot_hook
.type
= proto
;
1540 dev_add_pack(&po
->prot_hook
);
1545 spin_lock_bh(&net
->packet
.sklist_lock
);
1546 sk_add_node_rcu(sk
, &net
->packet
.sklist
);
1547 sock_prot_inuse_add(net
, &packet_proto
, 1);
1548 spin_unlock_bh(&net
->packet
.sklist_lock
);
1555 static int packet_recv_error(struct sock
*sk
, struct msghdr
*msg
, int len
)
1557 struct sock_exterr_skb
*serr
;
1558 struct sk_buff
*skb
, *skb2
;
1562 skb
= skb_dequeue(&sk
->sk_error_queue
);
1568 msg
->msg_flags
|= MSG_TRUNC
;
1571 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1575 sock_recv_timestamp(msg
, sk
, skb
);
1577 serr
= SKB_EXT_ERR(skb
);
1578 put_cmsg(msg
, SOL_PACKET
, PACKET_TX_TIMESTAMP
,
1579 sizeof(serr
->ee
), &serr
->ee
);
1581 msg
->msg_flags
|= MSG_ERRQUEUE
;
1584 /* Reset and regenerate socket error */
1585 spin_lock_bh(&sk
->sk_error_queue
.lock
);
1587 if ((skb2
= skb_peek(&sk
->sk_error_queue
)) != NULL
) {
1588 sk
->sk_err
= SKB_EXT_ERR(skb2
)->ee
.ee_errno
;
1589 spin_unlock_bh(&sk
->sk_error_queue
.lock
);
1590 sk
->sk_error_report(sk
);
1592 spin_unlock_bh(&sk
->sk_error_queue
.lock
);
1601 * Pull a packet from our receive queue and hand it to the user.
1602 * If necessary we block.
1605 static int packet_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1606 struct msghdr
*msg
, size_t len
, int flags
)
1608 struct sock
*sk
= sock
->sk
;
1609 struct sk_buff
*skb
;
1611 struct sockaddr_ll
*sll
;
1612 int vnet_hdr_len
= 0;
1615 if (flags
& ~(MSG_PEEK
|MSG_DONTWAIT
|MSG_TRUNC
|MSG_CMSG_COMPAT
|MSG_ERRQUEUE
))
1619 /* What error should we return now? EUNATTACH? */
1620 if (pkt_sk(sk
)->ifindex
< 0)
1624 if (flags
& MSG_ERRQUEUE
) {
1625 err
= packet_recv_error(sk
, msg
, len
);
1630 * Call the generic datagram receiver. This handles all sorts
1631 * of horrible races and re-entrancy so we can forget about it
1632 * in the protocol layers.
1634 * Now it will return ENETDOWN, if device have just gone down,
1635 * but then it will block.
1638 skb
= skb_recv_datagram(sk
, flags
, flags
& MSG_DONTWAIT
, &err
);
1641 * An error occurred so return it. Because skb_recv_datagram()
1642 * handles the blocking we don't see and worry about blocking
1649 if (pkt_sk(sk
)->has_vnet_hdr
) {
1650 struct virtio_net_hdr vnet_hdr
= { 0 };
1653 vnet_hdr_len
= sizeof(vnet_hdr
);
1654 if (len
< vnet_hdr_len
)
1657 len
-= vnet_hdr_len
;
1659 if (skb_is_gso(skb
)) {
1660 struct skb_shared_info
*sinfo
= skb_shinfo(skb
);
1662 /* This is a hint as to how much should be linear. */
1663 vnet_hdr
.hdr_len
= skb_headlen(skb
);
1664 vnet_hdr
.gso_size
= sinfo
->gso_size
;
1665 if (sinfo
->gso_type
& SKB_GSO_TCPV4
)
1666 vnet_hdr
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV4
;
1667 else if (sinfo
->gso_type
& SKB_GSO_TCPV6
)
1668 vnet_hdr
.gso_type
= VIRTIO_NET_HDR_GSO_TCPV6
;
1669 else if (sinfo
->gso_type
& SKB_GSO_UDP
)
1670 vnet_hdr
.gso_type
= VIRTIO_NET_HDR_GSO_UDP
;
1671 else if (sinfo
->gso_type
& SKB_GSO_FCOE
)
1675 if (sinfo
->gso_type
& SKB_GSO_TCP_ECN
)
1676 vnet_hdr
.gso_type
|= VIRTIO_NET_HDR_GSO_ECN
;
1678 vnet_hdr
.gso_type
= VIRTIO_NET_HDR_GSO_NONE
;
1680 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1681 vnet_hdr
.flags
= VIRTIO_NET_HDR_F_NEEDS_CSUM
;
1682 vnet_hdr
.csum_start
= skb_checksum_start_offset(skb
);
1683 vnet_hdr
.csum_offset
= skb
->csum_offset
;
1684 } /* else everything is zero */
1686 err
= memcpy_toiovec(msg
->msg_iov
, (void *)&vnet_hdr
,
1693 * If the address length field is there to be filled in, we fill
1697 sll
= &PACKET_SKB_CB(skb
)->sa
.ll
;
1698 if (sock
->type
== SOCK_PACKET
)
1699 msg
->msg_namelen
= sizeof(struct sockaddr_pkt
);
1701 msg
->msg_namelen
= sll
->sll_halen
+ offsetof(struct sockaddr_ll
, sll_addr
);
1704 * You lose any data beyond the buffer you gave. If it worries a
1705 * user program they can ask the device for its MTU anyway.
1711 msg
->msg_flags
|= MSG_TRUNC
;
1714 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1718 sock_recv_ts_and_drops(msg
, sk
, skb
);
1721 memcpy(msg
->msg_name
, &PACKET_SKB_CB(skb
)->sa
,
1724 if (pkt_sk(sk
)->auxdata
) {
1725 struct tpacket_auxdata aux
;
1727 aux
.tp_status
= TP_STATUS_USER
;
1728 if (skb
->ip_summed
== CHECKSUM_PARTIAL
)
1729 aux
.tp_status
|= TP_STATUS_CSUMNOTREADY
;
1730 aux
.tp_len
= PACKET_SKB_CB(skb
)->origlen
;
1731 aux
.tp_snaplen
= skb
->len
;
1733 aux
.tp_net
= skb_network_offset(skb
);
1734 if (vlan_tx_tag_present(skb
)) {
1735 aux
.tp_vlan_tci
= vlan_tx_tag_get(skb
);
1736 aux
.tp_status
|= TP_STATUS_VLAN_VALID
;
1738 aux
.tp_vlan_tci
= 0;
1741 put_cmsg(msg
, SOL_PACKET
, PACKET_AUXDATA
, sizeof(aux
), &aux
);
1745 * Free or return the buffer as appropriate. Again this
1746 * hides all the races and re-entrancy issues from us.
1748 err
= vnet_hdr_len
+ ((flags
&MSG_TRUNC
) ? skb
->len
: copied
);
1751 skb_free_datagram(sk
, skb
);
1756 static int packet_getname_spkt(struct socket
*sock
, struct sockaddr
*uaddr
,
1757 int *uaddr_len
, int peer
)
1759 struct net_device
*dev
;
1760 struct sock
*sk
= sock
->sk
;
1765 uaddr
->sa_family
= AF_PACKET
;
1767 dev
= dev_get_by_index_rcu(sock_net(sk
), pkt_sk(sk
)->ifindex
);
1769 strncpy(uaddr
->sa_data
, dev
->name
, 14);
1771 memset(uaddr
->sa_data
, 0, 14);
1773 *uaddr_len
= sizeof(*uaddr
);
1778 static int packet_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
1779 int *uaddr_len
, int peer
)
1781 struct net_device
*dev
;
1782 struct sock
*sk
= sock
->sk
;
1783 struct packet_sock
*po
= pkt_sk(sk
);
1784 DECLARE_SOCKADDR(struct sockaddr_ll
*, sll
, uaddr
);
1789 sll
->sll_family
= AF_PACKET
;
1790 sll
->sll_ifindex
= po
->ifindex
;
1791 sll
->sll_protocol
= po
->num
;
1792 sll
->sll_pkttype
= 0;
1794 dev
= dev_get_by_index_rcu(sock_net(sk
), po
->ifindex
);
1796 sll
->sll_hatype
= dev
->type
;
1797 sll
->sll_halen
= dev
->addr_len
;
1798 memcpy(sll
->sll_addr
, dev
->dev_addr
, dev
->addr_len
);
1800 sll
->sll_hatype
= 0; /* Bad: we have no ARPHRD_UNSPEC */
1804 *uaddr_len
= offsetof(struct sockaddr_ll
, sll_addr
) + sll
->sll_halen
;
1809 static int packet_dev_mc(struct net_device
*dev
, struct packet_mclist
*i
,
1813 case PACKET_MR_MULTICAST
:
1814 if (i
->alen
!= dev
->addr_len
)
1817 return dev_mc_add(dev
, i
->addr
);
1819 return dev_mc_del(dev
, i
->addr
);
1821 case PACKET_MR_PROMISC
:
1822 return dev_set_promiscuity(dev
, what
);
1824 case PACKET_MR_ALLMULTI
:
1825 return dev_set_allmulti(dev
, what
);
1827 case PACKET_MR_UNICAST
:
1828 if (i
->alen
!= dev
->addr_len
)
1831 return dev_uc_add(dev
, i
->addr
);
1833 return dev_uc_del(dev
, i
->addr
);
1841 static void packet_dev_mclist(struct net_device
*dev
, struct packet_mclist
*i
, int what
)
1843 for ( ; i
; i
= i
->next
) {
1844 if (i
->ifindex
== dev
->ifindex
)
1845 packet_dev_mc(dev
, i
, what
);
1849 static int packet_mc_add(struct sock
*sk
, struct packet_mreq_max
*mreq
)
1851 struct packet_sock
*po
= pkt_sk(sk
);
1852 struct packet_mclist
*ml
, *i
;
1853 struct net_device
*dev
;
1859 dev
= __dev_get_by_index(sock_net(sk
), mreq
->mr_ifindex
);
1864 if (mreq
->mr_alen
> dev
->addr_len
)
1868 i
= kmalloc(sizeof(*i
), GFP_KERNEL
);
1873 for (ml
= po
->mclist
; ml
; ml
= ml
->next
) {
1874 if (ml
->ifindex
== mreq
->mr_ifindex
&&
1875 ml
->type
== mreq
->mr_type
&&
1876 ml
->alen
== mreq
->mr_alen
&&
1877 memcmp(ml
->addr
, mreq
->mr_address
, ml
->alen
) == 0) {
1879 /* Free the new element ... */
1885 i
->type
= mreq
->mr_type
;
1886 i
->ifindex
= mreq
->mr_ifindex
;
1887 i
->alen
= mreq
->mr_alen
;
1888 memcpy(i
->addr
, mreq
->mr_address
, i
->alen
);
1890 i
->next
= po
->mclist
;
1892 err
= packet_dev_mc(dev
, i
, 1);
1894 po
->mclist
= i
->next
;
1903 static int packet_mc_drop(struct sock
*sk
, struct packet_mreq_max
*mreq
)
1905 struct packet_mclist
*ml
, **mlp
;
1909 for (mlp
= &pkt_sk(sk
)->mclist
; (ml
= *mlp
) != NULL
; mlp
= &ml
->next
) {
1910 if (ml
->ifindex
== mreq
->mr_ifindex
&&
1911 ml
->type
== mreq
->mr_type
&&
1912 ml
->alen
== mreq
->mr_alen
&&
1913 memcmp(ml
->addr
, mreq
->mr_address
, ml
->alen
) == 0) {
1914 if (--ml
->count
== 0) {
1915 struct net_device
*dev
;
1917 dev
= __dev_get_by_index(sock_net(sk
), ml
->ifindex
);
1919 packet_dev_mc(dev
, ml
, -1);
1927 return -EADDRNOTAVAIL
;
1930 static void packet_flush_mclist(struct sock
*sk
)
1932 struct packet_sock
*po
= pkt_sk(sk
);
1933 struct packet_mclist
*ml
;
1939 while ((ml
= po
->mclist
) != NULL
) {
1940 struct net_device
*dev
;
1942 po
->mclist
= ml
->next
;
1943 dev
= __dev_get_by_index(sock_net(sk
), ml
->ifindex
);
1945 packet_dev_mc(dev
, ml
, -1);
1952 packet_setsockopt(struct socket
*sock
, int level
, int optname
, char __user
*optval
, unsigned int optlen
)
1954 struct sock
*sk
= sock
->sk
;
1955 struct packet_sock
*po
= pkt_sk(sk
);
1958 if (level
!= SOL_PACKET
)
1959 return -ENOPROTOOPT
;
1962 case PACKET_ADD_MEMBERSHIP
:
1963 case PACKET_DROP_MEMBERSHIP
:
1965 struct packet_mreq_max mreq
;
1967 memset(&mreq
, 0, sizeof(mreq
));
1968 if (len
< sizeof(struct packet_mreq
))
1970 if (len
> sizeof(mreq
))
1972 if (copy_from_user(&mreq
, optval
, len
))
1974 if (len
< (mreq
.mr_alen
+ offsetof(struct packet_mreq
, mr_address
)))
1976 if (optname
== PACKET_ADD_MEMBERSHIP
)
1977 ret
= packet_mc_add(sk
, &mreq
);
1979 ret
= packet_mc_drop(sk
, &mreq
);
1983 case PACKET_RX_RING
:
1984 case PACKET_TX_RING
:
1986 struct tpacket_req req
;
1988 if (optlen
< sizeof(req
))
1990 if (pkt_sk(sk
)->has_vnet_hdr
)
1992 if (copy_from_user(&req
, optval
, sizeof(req
)))
1994 return packet_set_ring(sk
, &req
, 0, optname
== PACKET_TX_RING
);
1996 case PACKET_COPY_THRESH
:
2000 if (optlen
!= sizeof(val
))
2002 if (copy_from_user(&val
, optval
, sizeof(val
)))
2005 pkt_sk(sk
)->copy_thresh
= val
;
2008 case PACKET_VERSION
:
2012 if (optlen
!= sizeof(val
))
2014 if (po
->rx_ring
.pg_vec
|| po
->tx_ring
.pg_vec
)
2016 if (copy_from_user(&val
, optval
, sizeof(val
)))
2021 po
->tp_version
= val
;
2027 case PACKET_RESERVE
:
2031 if (optlen
!= sizeof(val
))
2033 if (po
->rx_ring
.pg_vec
|| po
->tx_ring
.pg_vec
)
2035 if (copy_from_user(&val
, optval
, sizeof(val
)))
2037 po
->tp_reserve
= val
;
2044 if (optlen
!= sizeof(val
))
2046 if (po
->rx_ring
.pg_vec
|| po
->tx_ring
.pg_vec
)
2048 if (copy_from_user(&val
, optval
, sizeof(val
)))
2050 po
->tp_loss
= !!val
;
2053 case PACKET_AUXDATA
:
2057 if (optlen
< sizeof(val
))
2059 if (copy_from_user(&val
, optval
, sizeof(val
)))
2062 po
->auxdata
= !!val
;
2065 case PACKET_ORIGDEV
:
2069 if (optlen
< sizeof(val
))
2071 if (copy_from_user(&val
, optval
, sizeof(val
)))
2074 po
->origdev
= !!val
;
2077 case PACKET_VNET_HDR
:
2081 if (sock
->type
!= SOCK_RAW
)
2083 if (po
->rx_ring
.pg_vec
|| po
->tx_ring
.pg_vec
)
2085 if (optlen
< sizeof(val
))
2087 if (copy_from_user(&val
, optval
, sizeof(val
)))
2090 po
->has_vnet_hdr
= !!val
;
2093 case PACKET_TIMESTAMP
:
2097 if (optlen
!= sizeof(val
))
2099 if (copy_from_user(&val
, optval
, sizeof(val
)))
2102 po
->tp_tstamp
= val
;
2106 return -ENOPROTOOPT
;
2110 static int packet_getsockopt(struct socket
*sock
, int level
, int optname
,
2111 char __user
*optval
, int __user
*optlen
)
2115 struct sock
*sk
= sock
->sk
;
2116 struct packet_sock
*po
= pkt_sk(sk
);
2118 struct tpacket_stats st
;
2120 if (level
!= SOL_PACKET
)
2121 return -ENOPROTOOPT
;
2123 if (get_user(len
, optlen
))
2130 case PACKET_STATISTICS
:
2131 if (len
> sizeof(struct tpacket_stats
))
2132 len
= sizeof(struct tpacket_stats
);
2133 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
2135 memset(&po
->stats
, 0, sizeof(st
));
2136 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
2137 st
.tp_packets
+= st
.tp_drops
;
2141 case PACKET_AUXDATA
:
2142 if (len
> sizeof(int))
2148 case PACKET_ORIGDEV
:
2149 if (len
> sizeof(int))
2155 case PACKET_VNET_HDR
:
2156 if (len
> sizeof(int))
2158 val
= po
->has_vnet_hdr
;
2162 case PACKET_VERSION
:
2163 if (len
> sizeof(int))
2165 val
= po
->tp_version
;
2169 if (len
> sizeof(int))
2171 if (copy_from_user(&val
, optval
, len
))
2175 val
= sizeof(struct tpacket_hdr
);
2178 val
= sizeof(struct tpacket2_hdr
);
2185 case PACKET_RESERVE
:
2186 if (len
> sizeof(unsigned int))
2187 len
= sizeof(unsigned int);
2188 val
= po
->tp_reserve
;
2192 if (len
> sizeof(unsigned int))
2193 len
= sizeof(unsigned int);
2197 case PACKET_TIMESTAMP
:
2198 if (len
> sizeof(int))
2200 val
= po
->tp_tstamp
;
2204 return -ENOPROTOOPT
;
2207 if (put_user(len
, optlen
))
2209 if (copy_to_user(optval
, data
, len
))
2215 static int packet_notifier(struct notifier_block
*this, unsigned long msg
, void *data
)
2218 struct hlist_node
*node
;
2219 struct net_device
*dev
= data
;
2220 struct net
*net
= dev_net(dev
);
2223 sk_for_each_rcu(sk
, node
, &net
->packet
.sklist
) {
2224 struct packet_sock
*po
= pkt_sk(sk
);
2227 case NETDEV_UNREGISTER
:
2229 packet_dev_mclist(dev
, po
->mclist
, -1);
2233 if (dev
->ifindex
== po
->ifindex
) {
2234 spin_lock(&po
->bind_lock
);
2236 __dev_remove_pack(&po
->prot_hook
);
2239 sk
->sk_err
= ENETDOWN
;
2240 if (!sock_flag(sk
, SOCK_DEAD
))
2241 sk
->sk_error_report(sk
);
2243 if (msg
== NETDEV_UNREGISTER
) {
2245 po
->prot_hook
.dev
= NULL
;
2247 spin_unlock(&po
->bind_lock
);
2251 if (dev
->ifindex
== po
->ifindex
) {
2252 spin_lock(&po
->bind_lock
);
2253 if (po
->num
&& !po
->running
) {
2254 dev_add_pack(&po
->prot_hook
);
2258 spin_unlock(&po
->bind_lock
);
2268 static int packet_ioctl(struct socket
*sock
, unsigned int cmd
,
2271 struct sock
*sk
= sock
->sk
;
2276 int amount
= sk_wmem_alloc_get(sk
);
2278 return put_user(amount
, (int __user
*)arg
);
2282 struct sk_buff
*skb
;
2285 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
2286 skb
= skb_peek(&sk
->sk_receive_queue
);
2289 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
2290 return put_user(amount
, (int __user
*)arg
);
2293 return sock_get_timestamp(sk
, (struct timeval __user
*)arg
);
2295 return sock_get_timestampns(sk
, (struct timespec __user
*)arg
);
2305 case SIOCGIFBRDADDR
:
2306 case SIOCSIFBRDADDR
:
2307 case SIOCGIFNETMASK
:
2308 case SIOCSIFNETMASK
:
2309 case SIOCGIFDSTADDR
:
2310 case SIOCSIFDSTADDR
:
2312 return inet_dgram_ops
.ioctl(sock
, cmd
, arg
);
2316 return -ENOIOCTLCMD
;
2321 static unsigned int packet_poll(struct file
*file
, struct socket
*sock
,
2324 struct sock
*sk
= sock
->sk
;
2325 struct packet_sock
*po
= pkt_sk(sk
);
2326 unsigned int mask
= datagram_poll(file
, sock
, wait
);
2328 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
2329 if (po
->rx_ring
.pg_vec
) {
2330 if (!packet_previous_frame(po
, &po
->rx_ring
, TP_STATUS_KERNEL
))
2331 mask
|= POLLIN
| POLLRDNORM
;
2333 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
2334 spin_lock_bh(&sk
->sk_write_queue
.lock
);
2335 if (po
->tx_ring
.pg_vec
) {
2336 if (packet_current_frame(po
, &po
->tx_ring
, TP_STATUS_AVAILABLE
))
2337 mask
|= POLLOUT
| POLLWRNORM
;
2339 spin_unlock_bh(&sk
->sk_write_queue
.lock
);
2344 /* Dirty? Well, I still did not learn better way to account
2348 static void packet_mm_open(struct vm_area_struct
*vma
)
2350 struct file
*file
= vma
->vm_file
;
2351 struct socket
*sock
= file
->private_data
;
2352 struct sock
*sk
= sock
->sk
;
2355 atomic_inc(&pkt_sk(sk
)->mapped
);
2358 static void packet_mm_close(struct vm_area_struct
*vma
)
2360 struct file
*file
= vma
->vm_file
;
2361 struct socket
*sock
= file
->private_data
;
2362 struct sock
*sk
= sock
->sk
;
2365 atomic_dec(&pkt_sk(sk
)->mapped
);
2368 static const struct vm_operations_struct packet_mmap_ops
= {
2369 .open
= packet_mm_open
,
2370 .close
= packet_mm_close
,
2373 static void free_pg_vec(struct pgv
*pg_vec
, unsigned int order
,
2378 for (i
= 0; i
< len
; i
++) {
2379 if (likely(pg_vec
[i
].buffer
)) {
2380 if (is_vmalloc_addr(pg_vec
[i
].buffer
))
2381 vfree(pg_vec
[i
].buffer
);
2383 free_pages((unsigned long)pg_vec
[i
].buffer
,
2385 pg_vec
[i
].buffer
= NULL
;
2391 static inline char *alloc_one_pg_vec_page(unsigned long order
)
2393 char *buffer
= NULL
;
2394 gfp_t gfp_flags
= GFP_KERNEL
| __GFP_COMP
|
2395 __GFP_ZERO
| __GFP_NOWARN
| __GFP_NORETRY
;
2397 buffer
= (char *) __get_free_pages(gfp_flags
, order
);
2403 * __get_free_pages failed, fall back to vmalloc
2405 buffer
= vzalloc((1 << order
) * PAGE_SIZE
);
2411 * vmalloc failed, lets dig into swap here
2413 gfp_flags
&= ~__GFP_NORETRY
;
2414 buffer
= (char *)__get_free_pages(gfp_flags
, order
);
2419 * complete and utter failure
2424 static struct pgv
*alloc_pg_vec(struct tpacket_req
*req
, int order
)
2426 unsigned int block_nr
= req
->tp_block_nr
;
2430 pg_vec
= kcalloc(block_nr
, sizeof(struct pgv
), GFP_KERNEL
);
2431 if (unlikely(!pg_vec
))
2434 for (i
= 0; i
< block_nr
; i
++) {
2435 pg_vec
[i
].buffer
= alloc_one_pg_vec_page(order
);
2436 if (unlikely(!pg_vec
[i
].buffer
))
2437 goto out_free_pgvec
;
2444 free_pg_vec(pg_vec
, order
, block_nr
);
2449 static int packet_set_ring(struct sock
*sk
, struct tpacket_req
*req
,
2450 int closing
, int tx_ring
)
2452 struct pgv
*pg_vec
= NULL
;
2453 struct packet_sock
*po
= pkt_sk(sk
);
2454 int was_running
, order
= 0;
2455 struct packet_ring_buffer
*rb
;
2456 struct sk_buff_head
*rb_queue
;
2460 rb
= tx_ring
? &po
->tx_ring
: &po
->rx_ring
;
2461 rb_queue
= tx_ring
? &sk
->sk_write_queue
: &sk
->sk_receive_queue
;
2465 if (atomic_read(&po
->mapped
))
2467 if (atomic_read(&rb
->pending
))
2471 if (req
->tp_block_nr
) {
2472 /* Sanity tests and some calculations */
2474 if (unlikely(rb
->pg_vec
))
2477 switch (po
->tp_version
) {
2479 po
->tp_hdrlen
= TPACKET_HDRLEN
;
2482 po
->tp_hdrlen
= TPACKET2_HDRLEN
;
2487 if (unlikely((int)req
->tp_block_size
<= 0))
2489 if (unlikely(req
->tp_block_size
& (PAGE_SIZE
- 1)))
2491 if (unlikely(req
->tp_frame_size
< po
->tp_hdrlen
+
2494 if (unlikely(req
->tp_frame_size
& (TPACKET_ALIGNMENT
- 1)))
2497 rb
->frames_per_block
= req
->tp_block_size
/req
->tp_frame_size
;
2498 if (unlikely(rb
->frames_per_block
<= 0))
2500 if (unlikely((rb
->frames_per_block
* req
->tp_block_nr
) !=
2505 order
= get_order(req
->tp_block_size
);
2506 pg_vec
= alloc_pg_vec(req
, order
);
2507 if (unlikely(!pg_vec
))
2513 if (unlikely(req
->tp_frame_nr
))
2519 /* Detach socket from network */
2520 spin_lock(&po
->bind_lock
);
2521 was_running
= po
->running
;
2524 __dev_remove_pack(&po
->prot_hook
);
2529 spin_unlock(&po
->bind_lock
);
2534 mutex_lock(&po
->pg_vec_lock
);
2535 if (closing
|| atomic_read(&po
->mapped
) == 0) {
2537 spin_lock_bh(&rb_queue
->lock
);
2538 swap(rb
->pg_vec
, pg_vec
);
2539 rb
->frame_max
= (req
->tp_frame_nr
- 1);
2541 rb
->frame_size
= req
->tp_frame_size
;
2542 spin_unlock_bh(&rb_queue
->lock
);
2544 swap(rb
->pg_vec_order
, order
);
2545 swap(rb
->pg_vec_len
, req
->tp_block_nr
);
2547 rb
->pg_vec_pages
= req
->tp_block_size
/PAGE_SIZE
;
2548 po
->prot_hook
.func
= (po
->rx_ring
.pg_vec
) ?
2549 tpacket_rcv
: packet_rcv
;
2550 skb_queue_purge(rb_queue
);
2551 if (atomic_read(&po
->mapped
))
2552 pr_err("packet_mmap: vma is busy: %d\n",
2553 atomic_read(&po
->mapped
));
2555 mutex_unlock(&po
->pg_vec_lock
);
2557 spin_lock(&po
->bind_lock
);
2558 if (was_running
&& !po
->running
) {
2562 dev_add_pack(&po
->prot_hook
);
2564 spin_unlock(&po
->bind_lock
);
2569 free_pg_vec(pg_vec
, order
, req
->tp_block_nr
);
2574 static int packet_mmap(struct file
*file
, struct socket
*sock
,
2575 struct vm_area_struct
*vma
)
2577 struct sock
*sk
= sock
->sk
;
2578 struct packet_sock
*po
= pkt_sk(sk
);
2579 unsigned long size
, expected_size
;
2580 struct packet_ring_buffer
*rb
;
2581 unsigned long start
;
2588 mutex_lock(&po
->pg_vec_lock
);
2591 for (rb
= &po
->rx_ring
; rb
<= &po
->tx_ring
; rb
++) {
2593 expected_size
+= rb
->pg_vec_len
2599 if (expected_size
== 0)
2602 size
= vma
->vm_end
- vma
->vm_start
;
2603 if (size
!= expected_size
)
2606 start
= vma
->vm_start
;
2607 for (rb
= &po
->rx_ring
; rb
<= &po
->tx_ring
; rb
++) {
2608 if (rb
->pg_vec
== NULL
)
2611 for (i
= 0; i
< rb
->pg_vec_len
; i
++) {
2613 void *kaddr
= rb
->pg_vec
[i
].buffer
;
2616 for (pg_num
= 0; pg_num
< rb
->pg_vec_pages
; pg_num
++) {
2617 page
= pgv_to_page(kaddr
);
2618 err
= vm_insert_page(vma
, start
, page
);
2627 atomic_inc(&po
->mapped
);
2628 vma
->vm_ops
= &packet_mmap_ops
;
2632 mutex_unlock(&po
->pg_vec_lock
);
2636 static const struct proto_ops packet_ops_spkt
= {
2637 .family
= PF_PACKET
,
2638 .owner
= THIS_MODULE
,
2639 .release
= packet_release
,
2640 .bind
= packet_bind_spkt
,
2641 .connect
= sock_no_connect
,
2642 .socketpair
= sock_no_socketpair
,
2643 .accept
= sock_no_accept
,
2644 .getname
= packet_getname_spkt
,
2645 .poll
= datagram_poll
,
2646 .ioctl
= packet_ioctl
,
2647 .listen
= sock_no_listen
,
2648 .shutdown
= sock_no_shutdown
,
2649 .setsockopt
= sock_no_setsockopt
,
2650 .getsockopt
= sock_no_getsockopt
,
2651 .sendmsg
= packet_sendmsg_spkt
,
2652 .recvmsg
= packet_recvmsg
,
2653 .mmap
= sock_no_mmap
,
2654 .sendpage
= sock_no_sendpage
,
2657 static const struct proto_ops packet_ops
= {
2658 .family
= PF_PACKET
,
2659 .owner
= THIS_MODULE
,
2660 .release
= packet_release
,
2661 .bind
= packet_bind
,
2662 .connect
= sock_no_connect
,
2663 .socketpair
= sock_no_socketpair
,
2664 .accept
= sock_no_accept
,
2665 .getname
= packet_getname
,
2666 .poll
= packet_poll
,
2667 .ioctl
= packet_ioctl
,
2668 .listen
= sock_no_listen
,
2669 .shutdown
= sock_no_shutdown
,
2670 .setsockopt
= packet_setsockopt
,
2671 .getsockopt
= packet_getsockopt
,
2672 .sendmsg
= packet_sendmsg
,
2673 .recvmsg
= packet_recvmsg
,
2674 .mmap
= packet_mmap
,
2675 .sendpage
= sock_no_sendpage
,
2678 static const struct net_proto_family packet_family_ops
= {
2679 .family
= PF_PACKET
,
2680 .create
= packet_create
,
2681 .owner
= THIS_MODULE
,
2684 static struct notifier_block packet_netdev_notifier
= {
2685 .notifier_call
= packet_notifier
,
2688 #ifdef CONFIG_PROC_FS
2690 static void *packet_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2693 struct net
*net
= seq_file_net(seq
);
2696 return seq_hlist_start_head_rcu(&net
->packet
.sklist
, *pos
);
2699 static void *packet_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2701 struct net
*net
= seq_file_net(seq
);
2702 return seq_hlist_next_rcu(v
, &net
->packet
.sklist
, pos
);
2705 static void packet_seq_stop(struct seq_file
*seq
, void *v
)
2711 static int packet_seq_show(struct seq_file
*seq
, void *v
)
2713 if (v
== SEQ_START_TOKEN
)
2714 seq_puts(seq
, "sk RefCnt Type Proto Iface R Rmem User Inode\n");
2716 struct sock
*s
= sk_entry(v
);
2717 const struct packet_sock
*po
= pkt_sk(s
);
2720 "%pK %-6d %-4d %04x %-5d %1d %-6u %-6u %-6lu\n",
2722 atomic_read(&s
->sk_refcnt
),
2727 atomic_read(&s
->sk_rmem_alloc
),
2735 static const struct seq_operations packet_seq_ops
= {
2736 .start
= packet_seq_start
,
2737 .next
= packet_seq_next
,
2738 .stop
= packet_seq_stop
,
2739 .show
= packet_seq_show
,
2742 static int packet_seq_open(struct inode
*inode
, struct file
*file
)
2744 return seq_open_net(inode
, file
, &packet_seq_ops
,
2745 sizeof(struct seq_net_private
));
2748 static const struct file_operations packet_seq_fops
= {
2749 .owner
= THIS_MODULE
,
2750 .open
= packet_seq_open
,
2752 .llseek
= seq_lseek
,
2753 .release
= seq_release_net
,
2758 static int __net_init
packet_net_init(struct net
*net
)
2760 spin_lock_init(&net
->packet
.sklist_lock
);
2761 INIT_HLIST_HEAD(&net
->packet
.sklist
);
2763 if (!proc_net_fops_create(net
, "packet", 0, &packet_seq_fops
))
2769 static void __net_exit
packet_net_exit(struct net
*net
)
2771 proc_net_remove(net
, "packet");
2774 static struct pernet_operations packet_net_ops
= {
2775 .init
= packet_net_init
,
2776 .exit
= packet_net_exit
,
2780 static void __exit
packet_exit(void)
2782 unregister_netdevice_notifier(&packet_netdev_notifier
);
2783 unregister_pernet_subsys(&packet_net_ops
);
2784 sock_unregister(PF_PACKET
);
2785 proto_unregister(&packet_proto
);
2788 static int __init
packet_init(void)
2790 int rc
= proto_register(&packet_proto
, 0);
2795 sock_register(&packet_family_ops
);
2796 register_pernet_subsys(&packet_net_ops
);
2797 register_netdevice_notifier(&packet_netdev_notifier
);
2802 module_init(packet_init
);
2803 module_exit(packet_exit
);
2804 MODULE_LICENSE("GPL");
2805 MODULE_ALIAS_NETPROTO(PF_PACKET
);