linux/drivers/net/tun.c

1696 lines
38 KiB
C
Raw Normal View History

/*
* TUN - Universal TUN/TAP device driver.
* Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
*
* 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.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
*/
/*
* Changes:
*
* Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
* Add TUNSETLINK ioctl to set the link encapsulation
*
* Mark Smith <markzzzsmith@yahoo.com.au>
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
* Use random_ether_addr() for tap MAC address.
*
* Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
* Fixes in packet dropping, queue length setting and queue wakeup.
* Increased default tx queue length.
* Added ethtool API.
* Minor cleanups
*
* Daniel Podlejski <underley@underley.eu.org>
* Modifications for 2.3.99-pre5 kernel.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#define DRV_NAME "tun"
#define DRV_VERSION "1.6"
#define DRV_DESCRIPTION "Universal TUN/TAP device driver"
#define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/major.h>
#include <linux/slab.h>
#include <linux/poll.h>
#include <linux/fcntl.h>
#include <linux/init.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/miscdevice.h>
#include <linux/ethtool.h>
#include <linux/rtnetlink.h>
#include <linux/compat.h>
#include <linux/if.h>
#include <linux/if_arp.h>
#include <linux/if_ether.h>
#include <linux/if_tun.h>
#include <linux/crc32.h>
#include <linux/nsproxy.h>
#include <linux/virtio_net.h>
#include <linux/rcupdate.h>
[NET]: Make the device list and device lookups per namespace. This patch makes most of the generic device layer network namespace safe. This patch makes dev_base_head a network namespace variable, and then it picks up a few associated variables. The functions: dev_getbyhwaddr dev_getfirsthwbytype dev_get_by_flags dev_get_by_name __dev_get_by_name dev_get_by_index __dev_get_by_index dev_ioctl dev_ethtool dev_load wireless_process_ioctl were modified to take a network namespace argument, and deal with it. vlan_ioctl_set and brioctl_set were modified so their hooks will receive a network namespace argument. So basically anthing in the core of the network stack that was affected to by the change of dev_base was modified to handle multiple network namespaces. The rest of the network stack was simply modified to explicitly use &init_net the initial network namespace. This can be fixed when those components of the network stack are modified to handle multiple network namespaces. For now the ifindex generator is left global. Fundametally ifindex numbers are per namespace, or else we will have corner case problems with migration when we get that far. At the same time there are assumptions in the network stack that the ifindex of a network device won't change. Making the ifindex number global seems a good compromise until the network stack can cope with ifindex changes when you change namespaces, and the like. Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-09-17 18:56:21 +00:00
#include <net/net_namespace.h>
#include <net/netns/generic.h>
#include <net/rtnetlink.h>
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
#include <net/sock.h>
#include <asm/uaccess.h>
/* Uncomment to enable debugging */
/* #define TUN_DEBUG 1 */
#ifdef TUN_DEBUG
static int debug;
#define tun_debug(level, tun, fmt, args...) \
do { \
if (tun->debug) \
netdev_printk(level, tun->dev, fmt, ##args); \
} while (0)
#define DBG1(level, fmt, args...) \
do { \
if (debug == 2) \
printk(level fmt, ##args); \
} while (0)
#else
#define tun_debug(level, tun, fmt, args...) \
do { \
if (0) \
netdev_printk(level, tun->dev, fmt, ##args); \
} while (0)
#define DBG1(level, fmt, args...) \
do { \
if (0) \
printk(level fmt, ##args); \
} while (0)
#endif
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
#define FLT_EXACT_COUNT 8
struct tap_filter {
unsigned int count; /* Number of addrs. Zero means disabled */
u32 mask[2]; /* Mask of the hashed addrs */
unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
};
struct tun_file {
atomic_t count;
struct tun_struct *tun;
struct net *net;
};
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
struct tun_sock;
struct tun_struct {
struct tun_file *tfile;
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
unsigned int flags;
uid_t owner;
gid_t group;
struct net_device *dev;
netdev_features_t set_features;
#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
NETIF_F_TSO6|NETIF_F_UFO)
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
struct fasync_struct *fasync;
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
struct tap_filter txflt;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
struct socket socket;
net: sock_def_readable() and friends RCU conversion sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-04-29 11:01:49 +00:00
struct socket_wq wq;
int vnet_hdr_sz;
#ifdef TUN_DEBUG
int debug;
#endif
};
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
struct tun_sock {
struct sock sk;
struct tun_struct *tun;
};
static inline struct tun_sock *tun_sk(struct sock *sk)
{
return container_of(sk, struct tun_sock, sk);
}
static int tun_attach(struct tun_struct *tun, struct file *file)
{
struct tun_file *tfile = file->private_data;
int err;
ASSERT_RTNL();
netif_tx_lock_bh(tun->dev);
err = -EINVAL;
if (tfile->tun)
goto out;
err = -EBUSY;
if (tun->tfile)
goto out;
err = 0;
tfile->tun = tun;
tun->tfile = tfile;
tun->socket.file = file;
netif_carrier_on(tun->dev);
dev_hold(tun->dev);
sock_hold(tun->socket.sk);
atomic_inc(&tfile->count);
out:
netif_tx_unlock_bh(tun->dev);
return err;
}
static void __tun_detach(struct tun_struct *tun)
{
/* Detach from net device */
netif_tx_lock_bh(tun->dev);
netif_carrier_off(tun->dev);
tun->tfile = NULL;
tun->socket.file = NULL;
netif_tx_unlock_bh(tun->dev);
/* Drop read queue */
skb_queue_purge(&tun->socket.sk->sk_receive_queue);
/* Drop the extra count on the net device */
dev_put(tun->dev);
}
static void tun_detach(struct tun_struct *tun)
{
rtnl_lock();
__tun_detach(tun);
rtnl_unlock();
}
static struct tun_struct *__tun_get(struct tun_file *tfile)
{
struct tun_struct *tun = NULL;
if (atomic_inc_not_zero(&tfile->count))
tun = tfile->tun;
return tun;
}
static struct tun_struct *tun_get(struct file *file)
{
return __tun_get(file->private_data);
}
static void tun_put(struct tun_struct *tun)
{
struct tun_file *tfile = tun->tfile;
if (atomic_dec_and_test(&tfile->count))
tun_detach(tfile->tun);
}
/* TAP filtering */
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
static void addr_hash_set(u32 *mask, const u8 *addr)
{
int n = ether_crc(ETH_ALEN, addr) >> 26;
mask[n >> 5] |= (1 << (n & 31));
}
static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
{
int n = ether_crc(ETH_ALEN, addr) >> 26;
return mask[n >> 5] & (1 << (n & 31));
}
static int update_filter(struct tap_filter *filter, void __user *arg)
{
struct { u8 u[ETH_ALEN]; } *addr;
struct tun_filter uf;
int err, alen, n, nexact;
if (copy_from_user(&uf, arg, sizeof(uf)))
return -EFAULT;
if (!uf.count) {
/* Disabled */
filter->count = 0;
return 0;
}
alen = ETH_ALEN * uf.count;
addr = kmalloc(alen, GFP_KERNEL);
if (!addr)
return -ENOMEM;
if (copy_from_user(addr, arg + sizeof(uf), alen)) {
err = -EFAULT;
goto done;
}
/* The filter is updated without holding any locks. Which is
* perfectly safe. We disable it first and in the worst
* case we'll accept a few undesired packets. */
filter->count = 0;
wmb();
/* Use first set of addresses as an exact filter */
for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
nexact = n;
/* Remaining multicast addresses are hashed,
* unicast will leave the filter disabled. */
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
memset(filter->mask, 0, sizeof(filter->mask));
for (; n < uf.count; n++) {
if (!is_multicast_ether_addr(addr[n].u)) {
err = 0; /* no filter */
goto done;
}
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
addr_hash_set(filter->mask, addr[n].u);
}
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
/* For ALLMULTI just set the mask to all ones.
* This overrides the mask populated above. */
if ((uf.flags & TUN_FLT_ALLMULTI))
memset(filter->mask, ~0, sizeof(filter->mask));
/* Now enable the filter */
wmb();
filter->count = nexact;
/* Return the number of exact filters */
err = nexact;
done:
kfree(addr);
return err;
}
/* Returns: 0 - drop, !=0 - accept */
static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
{
/* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
* at this point. */
struct ethhdr *eh = (struct ethhdr *) skb->data;
int i;
/* Exact match */
for (i = 0; i < filter->count; i++)
if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
return 1;
/* Inexact match (multicast only) */
if (is_multicast_ether_addr(eh->h_dest))
return addr_hash_test(filter->mask, eh->h_dest);
return 0;
}
/*
* Checks whether the packet is accepted or not.
* Returns: 0 - drop, !=0 - accept
*/
static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
{
if (!filter->count)
return 1;
return run_filter(filter, skb);
}
/* Network device part of the driver */
static const struct ethtool_ops tun_ethtool_ops;
/* Net device detach from fd. */
static void tun_net_uninit(struct net_device *dev)
{
struct tun_struct *tun = netdev_priv(dev);
struct tun_file *tfile = tun->tfile;
/* Inform the methods they need to stop using the dev.
*/
if (tfile) {
net: sock_def_readable() and friends RCU conversion sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-04-29 11:01:49 +00:00
wake_up_all(&tun->wq.wait);
if (atomic_dec_and_test(&tfile->count))
__tun_detach(tun);
}
}
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
static void tun_free_netdev(struct net_device *dev)
{
struct tun_struct *tun = netdev_priv(dev);
sk_release_kernel(tun->socket.sk);
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
}
/* Net device open. */
static int tun_net_open(struct net_device *dev)
{
netif_start_queue(dev);
return 0;
}
/* Net device close. */
static int tun_net_close(struct net_device *dev)
{
netif_stop_queue(dev);
return 0;
}
/* Net device start xmit */
static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct tun_struct *tun = netdev_priv(dev);
tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
/* Drop packet if interface is not attached */
if (!tun->tfile)
goto drop;
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
/* Drop if the filter does not like it.
* This is a noop if the filter is disabled.
* Filter can be enabled only for the TAP devices. */
if (!check_filter(&tun->txflt, skb))
goto drop;
if (tun->socket.sk->sk_filter &&
sk_filter(tun->socket.sk, skb))
goto drop;
if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) {
if (!(tun->flags & TUN_ONE_QUEUE)) {
/* Normal queueing mode. */
/* Packet scheduler handles dropping of further packets. */
netif_stop_queue(dev);
/* We won't see all dropped packets individually, so overrun
* error is more appropriate. */
dev->stats.tx_fifo_errors++;
} else {
/* Single queue mode.
* Driver handles dropping of all packets itself. */
goto drop;
}
}
/* Orphan the skb - required as we might hang on to it
* for indefinite time. */
skb_orphan(skb);
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
/* Enqueue packet */
skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb);
/* Notify and wake up reader process */
if (tun->flags & TUN_FASYNC)
kill_fasync(&tun->fasync, SIGIO, POLL_IN);
net: sock_def_readable() and friends RCU conversion sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-04-29 11:01:49 +00:00
wake_up_interruptible_poll(&tun->wq.wait, POLLIN |
POLLRDNORM | POLLRDBAND);
return NETDEV_TX_OK;
drop:
dev->stats.tx_dropped++;
kfree_skb(skb);
return NETDEV_TX_OK;
}
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
static void tun_net_mclist(struct net_device *dev)
{
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
/*
* This callback is supposed to deal with mc filter in
* _rx_ path and has nothing to do with the _tx_ path.
* In rx path we always accept everything userspace gives us.
*/
}
#define MIN_MTU 68
#define MAX_MTU 65535
static int
tun_net_change_mtu(struct net_device *dev, int new_mtu)
{
if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
return -EINVAL;
dev->mtu = new_mtu;
return 0;
}
static netdev_features_t tun_net_fix_features(struct net_device *dev,
netdev_features_t features)
{
struct tun_struct *tun = netdev_priv(dev);
return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
}
#ifdef CONFIG_NET_POLL_CONTROLLER
static void tun_poll_controller(struct net_device *dev)
{
/*
* Tun only receives frames when:
* 1) the char device endpoint gets data from user space
* 2) the tun socket gets a sendmsg call from user space
* Since both of those are syncronous operations, we are guaranteed
* never to have pending data when we poll for it
* so theres nothing to do here but return.
* We need this though so netpoll recognizes us as an interface that
* supports polling, which enables bridge devices in virt setups to
* still use netconsole
*/
return;
}
#endif
static const struct net_device_ops tun_netdev_ops = {
.ndo_uninit = tun_net_uninit,
.ndo_open = tun_net_open,
.ndo_stop = tun_net_close,
.ndo_start_xmit = tun_net_xmit,
.ndo_change_mtu = tun_net_change_mtu,
.ndo_fix_features = tun_net_fix_features,
#ifdef CONFIG_NET_POLL_CONTROLLER
.ndo_poll_controller = tun_poll_controller,
#endif
};
static const struct net_device_ops tap_netdev_ops = {
.ndo_uninit = tun_net_uninit,
.ndo_open = tun_net_open,
.ndo_stop = tun_net_close,
.ndo_start_xmit = tun_net_xmit,
.ndo_change_mtu = tun_net_change_mtu,
.ndo_fix_features = tun_net_fix_features,
.ndo_set_rx_mode = tun_net_mclist,
.ndo_set_mac_address = eth_mac_addr,
.ndo_validate_addr = eth_validate_addr,
#ifdef CONFIG_NET_POLL_CONTROLLER
.ndo_poll_controller = tun_poll_controller,
#endif
};
/* Initialize net device. */
static void tun_net_init(struct net_device *dev)
{
struct tun_struct *tun = netdev_priv(dev);
switch (tun->flags & TUN_TYPE_MASK) {
case TUN_TUN_DEV:
dev->netdev_ops = &tun_netdev_ops;
/* Point-to-Point TUN Device */
dev->hard_header_len = 0;
dev->addr_len = 0;
dev->mtu = 1500;
/* Zero header length */
dev->type = ARPHRD_NONE;
dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
break;
case TUN_TAP_DEV:
dev->netdev_ops = &tap_netdev_ops;
/* Ethernet TAP Device */
ether_setup(dev);
dev->priv_flags &= ~IFF_TX_SKB_SHARING;
eth_hw_addr_random(dev);
dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
break;
}
}
/* Character device part */
/* Poll */
static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
{
struct tun_file *tfile = file->private_data;
struct tun_struct *tun = __tun_get(tfile);
struct sock *sk;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
unsigned int mask = 0;
if (!tun)
return POLLERR;
sk = tun->socket.sk;
tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
net: sock_def_readable() and friends RCU conversion sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-04-29 11:01:49 +00:00
poll_wait(file, &tun->wq.wait, wait);
if (!skb_queue_empty(&sk->sk_receive_queue))
mask |= POLLIN | POLLRDNORM;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
if (sock_writeable(sk) ||
(!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
sock_writeable(sk)))
mask |= POLLOUT | POLLWRNORM;
if (tun->dev->reg_state != NETREG_REGISTERED)
mask = POLLERR;
tun_put(tun);
return mask;
}
/* prepad is the amount to reserve at front. len is length after that.
* linear is a hint as to how much to copy (usually headers). */
static struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
size_t prepad, size_t len,
size_t linear, int noblock)
{
struct sock *sk = tun->socket.sk;
struct sk_buff *skb;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
int err;
sock_update_classid(sk);
/* Under a page? Don't bother with paged skb. */
if (prepad + len < PAGE_SIZE || !linear)
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
linear = len;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
&err);
if (!skb)
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
return ERR_PTR(err);
skb_reserve(skb, prepad);
skb_put(skb, linear);
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
skb->data_len = len - linear;
skb->len += len - linear;
return skb;
}
/* Get packet from user space buffer */
static ssize_t tun_get_user(struct tun_struct *tun,
const struct iovec *iv, size_t count,
int noblock)
{
struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
struct sk_buff *skb;
size_t len = count, align = NET_SKB_PAD;
struct virtio_net_hdr gso = { 0 };
int offset = 0;
if (!(tun->flags & TUN_NO_PI)) {
if ((len -= sizeof(pi)) > count)
return -EINVAL;
if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
return -EFAULT;
offset += sizeof(pi);
}
if (tun->flags & TUN_VNET_HDR) {
if ((len -= tun->vnet_hdr_sz) > count)
return -EINVAL;
if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
return -EFAULT;
if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
if (gso.hdr_len > len)
return -EINVAL;
offset += tun->vnet_hdr_sz;
}
if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
align += NET_IP_ALIGN;
if (unlikely(len < ETH_HLEN ||
(gso.hdr_len && gso.hdr_len < ETH_HLEN)))
return -EINVAL;
}
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
if (IS_ERR(skb)) {
if (PTR_ERR(skb) != -EAGAIN)
tun->dev->stats.rx_dropped++;
return PTR_ERR(skb);
}
if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
tun->dev->stats.rx_dropped++;
kfree_skb(skb);
return -EFAULT;
}
if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
if (!skb_partial_csum_set(skb, gso.csum_start,
gso.csum_offset)) {
tun->dev->stats.rx_frame_errors++;
kfree_skb(skb);
return -EINVAL;
}
}
switch (tun->flags & TUN_TYPE_MASK) {
case TUN_TUN_DEV:
if (tun->flags & TUN_NO_PI) {
switch (skb->data[0] & 0xf0) {
case 0x40:
pi.proto = htons(ETH_P_IP);
break;
case 0x60:
pi.proto = htons(ETH_P_IPV6);
break;
default:
tun->dev->stats.rx_dropped++;
kfree_skb(skb);
return -EINVAL;
}
}
skb_reset_mac_header(skb);
skb->protocol = pi.proto;
skb->dev = tun->dev;
break;
case TUN_TAP_DEV:
skb->protocol = eth_type_trans(skb, tun->dev);
break;
}
if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
pr_debug("GSO!\n");
switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
case VIRTIO_NET_HDR_GSO_TCPV4:
skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
break;
case VIRTIO_NET_HDR_GSO_TCPV6:
skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
break;
case VIRTIO_NET_HDR_GSO_UDP:
skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
break;
default:
tun->dev->stats.rx_frame_errors++;
kfree_skb(skb);
return -EINVAL;
}
if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
skb_shinfo(skb)->gso_size = gso.gso_size;
if (skb_shinfo(skb)->gso_size == 0) {
tun->dev->stats.rx_frame_errors++;
kfree_skb(skb);
return -EINVAL;
}
/* Header must be checked, and gso_segs computed. */
skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
skb_shinfo(skb)->gso_segs = 0;
}
netif_rx_ni(skb);
tun->dev->stats.rx_packets++;
tun->dev->stats.rx_bytes += len;
return count;
}
static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
unsigned long count, loff_t pos)
{
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
struct file *file = iocb->ki_filp;
struct tun_struct *tun = tun_get(file);
ssize_t result;
if (!tun)
return -EBADFD;
tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count);
result = tun_get_user(tun, iv, iov_length(iv, count),
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
file->f_flags & O_NONBLOCK);
tun_put(tun);
return result;
}
/* Put packet to the user space buffer */
static ssize_t tun_put_user(struct tun_struct *tun,
struct sk_buff *skb,
const struct iovec *iv, int len)
{
struct tun_pi pi = { 0, skb->protocol };
ssize_t total = 0;
if (!(tun->flags & TUN_NO_PI)) {
if ((len -= sizeof(pi)) < 0)
return -EINVAL;
if (len < skb->len) {
/* Packet will be striped */
pi.flags |= TUN_PKT_STRIP;
}
if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
return -EFAULT;
total += sizeof(pi);
}
if (tun->flags & TUN_VNET_HDR) {
struct virtio_net_hdr gso = { 0 }; /* no info leak */
if ((len -= tun->vnet_hdr_sz) < 0)
return -EINVAL;
if (skb_is_gso(skb)) {
struct skb_shared_info *sinfo = skb_shinfo(skb);
/* This is a hint as to how much should be linear. */
gso.hdr_len = skb_headlen(skb);
gso.gso_size = sinfo->gso_size;
if (sinfo->gso_type & SKB_GSO_TCPV4)
gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
else if (sinfo->gso_type & SKB_GSO_TCPV6)
gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
else if (sinfo->gso_type & SKB_GSO_UDP)
gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
else {
pr_err("unexpected GSO type: "
"0x%x, gso_size %d, hdr_len %d\n",
sinfo->gso_type, gso.gso_size,
gso.hdr_len);
print_hex_dump(KERN_ERR, "tun: ",
DUMP_PREFIX_NONE,
16, 1, skb->head,
min((int)gso.hdr_len, 64), true);
WARN_ON_ONCE(1);
return -EINVAL;
}
if (sinfo->gso_type & SKB_GSO_TCP_ECN)
gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
} else
gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
if (skb->ip_summed == CHECKSUM_PARTIAL) {
gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
gso.csum_start = skb_checksum_start_offset(skb);
gso.csum_offset = skb->csum_offset;
} else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
} /* else everything is zero */
if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
sizeof(gso))))
return -EFAULT;
total += tun->vnet_hdr_sz;
}
len = min_t(int, skb->len, len);
skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
total += skb->len;
tun->dev->stats.tx_packets++;
tun->dev->stats.tx_bytes += len;
return total;
}
static ssize_t tun_do_read(struct tun_struct *tun,
struct kiocb *iocb, const struct iovec *iv,
ssize_t len, int noblock)
{
DECLARE_WAITQUEUE(wait, current);
struct sk_buff *skb;
ssize_t ret = 0;
tun_debug(KERN_INFO, tun, "tun_chr_read\n");
if (unlikely(!noblock))
add_wait_queue(&tun->wq.wait, &wait);
while (len) {
current->state = TASK_INTERRUPTIBLE;
/* Read frames from the queue */
if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) {
if (noblock) {
ret = -EAGAIN;
break;
}
if (signal_pending(current)) {
ret = -ERESTARTSYS;
break;
}
if (tun->dev->reg_state != NETREG_REGISTERED) {
ret = -EIO;
break;
}
/* Nothing to read, let's sleep */
schedule();
continue;
}
netif_wake_queue(tun->dev);
ret = tun_put_user(tun, skb, iv, len);
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
kfree_skb(skb);
break;
}
current->state = TASK_RUNNING;
if (unlikely(!noblock))
remove_wait_queue(&tun->wq.wait, &wait);
return ret;
}
static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
unsigned long count, loff_t pos)
{
struct file *file = iocb->ki_filp;
struct tun_file *tfile = file->private_data;
struct tun_struct *tun = __tun_get(tfile);
ssize_t len, ret;
if (!tun)
return -EBADFD;
len = iov_length(iv, count);
if (len < 0) {
ret = -EINVAL;
goto out;
}
ret = tun_do_read(tun, iocb, iv, len, file->f_flags & O_NONBLOCK);
ret = min_t(ssize_t, ret, len);
out:
tun_put(tun);
return ret;
}
static void tun_setup(struct net_device *dev)
{
struct tun_struct *tun = netdev_priv(dev);
tun->owner = -1;
tun->group = -1;
dev->ethtool_ops = &tun_ethtool_ops;
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
dev->destructor = tun_free_netdev;
}
/* Trivial set of netlink ops to allow deleting tun or tap
* device with netlink.
*/
static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
{
return -EINVAL;
}
static struct rtnl_link_ops tun_link_ops __read_mostly = {
.kind = DRV_NAME,
.priv_size = sizeof(struct tun_struct),
.setup = tun_setup,
.validate = tun_validate,
};
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
static void tun_sock_write_space(struct sock *sk)
{
struct tun_struct *tun;
net: sock_def_readable() and friends RCU conversion sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-04-29 11:01:49 +00:00
wait_queue_head_t *wqueue;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
if (!sock_writeable(sk))
return;
if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
return;
net: sock_def_readable() and friends RCU conversion sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-04-29 11:01:49 +00:00
wqueue = sk_sleep(sk);
if (wqueue && waitqueue_active(wqueue))
wake_up_interruptible_sync_poll(wqueue, POLLOUT |
POLLWRNORM | POLLWRBAND);
tun = tun_sk(sk)->tun;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
}
static void tun_sock_destruct(struct sock *sk)
{
free_netdev(tun_sk(sk)->tun->dev);
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
}
static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
struct msghdr *m, size_t total_len)
{
struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
return tun_get_user(tun, m->msg_iov, total_len,
m->msg_flags & MSG_DONTWAIT);
}
static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
struct msghdr *m, size_t total_len,
int flags)
{
struct tun_struct *tun = container_of(sock, struct tun_struct, socket);
int ret;
if (flags & ~(MSG_DONTWAIT|MSG_TRUNC))
return -EINVAL;
ret = tun_do_read(tun, iocb, m->msg_iov, total_len,
flags & MSG_DONTWAIT);
if (ret > total_len) {
m->msg_flags |= MSG_TRUNC;
ret = flags & MSG_TRUNC ? ret : total_len;
}
return ret;
}
static int tun_release(struct socket *sock)
{
if (sock->sk)
sock_put(sock->sk);
return 0;
}
/* Ops structure to mimic raw sockets with tun */
static const struct proto_ops tun_socket_ops = {
.sendmsg = tun_sendmsg,
.recvmsg = tun_recvmsg,
.release = tun_release,
};
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
static struct proto tun_proto = {
.name = "tun",
.owner = THIS_MODULE,
.obj_size = sizeof(struct tun_sock),
};
static int tun_flags(struct tun_struct *tun)
{
int flags = 0;
if (tun->flags & TUN_TUN_DEV)
flags |= IFF_TUN;
else
flags |= IFF_TAP;
if (tun->flags & TUN_NO_PI)
flags |= IFF_NO_PI;
if (tun->flags & TUN_ONE_QUEUE)
flags |= IFF_ONE_QUEUE;
if (tun->flags & TUN_VNET_HDR)
flags |= IFF_VNET_HDR;
return flags;
}
static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
char *buf)
{
struct tun_struct *tun = netdev_priv(to_net_dev(dev));
return sprintf(buf, "0x%x\n", tun_flags(tun));
}
static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
char *buf)
{
struct tun_struct *tun = netdev_priv(to_net_dev(dev));
return sprintf(buf, "%d\n", tun->owner);
}
static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
char *buf)
{
struct tun_struct *tun = netdev_priv(to_net_dev(dev));
return sprintf(buf, "%d\n", tun->group);
}
static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
{
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
struct sock *sk;
struct tun_struct *tun;
struct net_device *dev;
int err;
dev = __dev_get_by_name(net, ifr->ifr_name);
if (dev) {
const struct cred *cred = current_cred();
if (ifr->ifr_flags & IFF_TUN_EXCL)
return -EBUSY;
if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
tun = netdev_priv(dev);
else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
tun = netdev_priv(dev);
else
return -EINVAL;
if (((tun->owner != -1 && cred->euid != tun->owner) ||
(tun->group != -1 && !in_egroup_p(tun->group))) &&
!capable(CAP_NET_ADMIN))
return -EPERM;
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6 * git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (1623 commits) netxen: update copyright netxen: fix tx timeout recovery netxen: fix file firmware leak netxen: improve pci memory access netxen: change firmware write size tg3: Fix return ring size breakage netxen: build fix for INET=n cdc-phonet: autoconfigure Phonet address Phonet: back-end for autoconfigured addresses Phonet: fix netlink address dump error handling ipv6: Add IFA_F_DADFAILED flag net: Add DEVTYPE support for Ethernet based devices mv643xx_eth.c: remove unused txq_set_wrr() ucc_geth: Fix hangs after switching from full to half duplex ucc_geth: Rearrange some code to avoid forward declarations phy/marvell: Make non-aneg speed/duplex forcing work for 88E1111 PHYs drivers/net/phy: introduce missing kfree drivers/net/wan: introduce missing kfree net: force bridge module(s) to be GPL Subject: [PATCH] appletalk: Fix skb leak when ipddp interface is not loaded ... Fixed up trivial conflicts: - arch/x86/include/asm/socket.h converted to <asm-generic/socket.h> in the x86 tree. The generic header has the same new #define's, so that works out fine. - drivers/net/tun.c fix conflict between 89f56d1e9 ("tun: reuse struct sock fields") that switched over to using 'tun->socket.sk' instead of the redundantly available (and thus removed) 'tun->sk', and 2b980dbd ("lsm: Add hooks to the TUN driver") which added a new 'tun->sk' use. Noted in 'next' by Stephen Rothwell.
2009-09-14 17:37:28 +00:00
err = security_tun_dev_attach(tun->socket.sk);
if (err < 0)
return err;
err = tun_attach(tun, file);
if (err < 0)
return err;
}
else {
char *name;
unsigned long flags = 0;
if (!capable(CAP_NET_ADMIN))
return -EPERM;
err = security_tun_dev_create();
if (err < 0)
return err;
/* Set dev type */
if (ifr->ifr_flags & IFF_TUN) {
/* TUN device */
flags |= TUN_TUN_DEV;
name = "tun%d";
} else if (ifr->ifr_flags & IFF_TAP) {
/* TAP device */
flags |= TUN_TAP_DEV;
name = "tap%d";
} else
return -EINVAL;
if (*ifr->ifr_name)
name = ifr->ifr_name;
dev = alloc_netdev(sizeof(struct tun_struct), name,
tun_setup);
if (!dev)
return -ENOMEM;
dev_net_set(dev, net);
dev->rtnl_link_ops = &tun_link_ops;
tun = netdev_priv(dev);
tun->dev = dev;
tun->flags = flags;
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
tun->txflt.count = 0;
tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
err = -ENOMEM;
sk = sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
if (!sk)
goto err_free_dev;
sk_change_net(sk, net);
net: sock_def_readable() and friends RCU conversion sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we need two atomic operations (and associated dirtying) per incoming packet. RCU conversion is pretty much needed : 1) Add a new structure, called "struct socket_wq" to hold all fields that will need rcu_read_lock() protection (currently: a wait_queue_head_t and a struct fasync_struct pointer). [Future patch will add a list anchor for wakeup coalescing] 2) Attach one of such structure to each "struct socket" created in sock_alloc_inode(). 3) Respect RCU grace period when freeing a "struct socket_wq" 4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct socket_wq" 5) Change sk_sleep() function to use new sk->sk_wq instead of sk->sk_sleep 6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside a rcu_read_lock() section. 7) Change all sk_has_sleeper() callers to : - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock) - Use wq_has_sleeper() to eventually wakeup tasks. - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock) 8) sock_wake_async() is modified to use rcu protection as well. 9) Exceptions : macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq" instead of dynamically allocated ones. They dont need rcu freeing. Some cleanups or followups are probably needed, (possible sk_callback_lock conversion to a spinlock for example...). Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-04-29 11:01:49 +00:00
tun->socket.wq = &tun->wq;
init_waitqueue_head(&tun->wq.wait);
tun->socket.ops = &tun_socket_ops;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
sock_init_data(&tun->socket, sk);
sk->sk_write_space = tun_sock_write_space;
sk->sk_sndbuf = INT_MAX;
tun_sk(sk)->tun = tun;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
security_tun_dev_post_create(sk);
tun_net_init(dev);
dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
TUN_USER_FEATURES;
dev->features = dev->hw_features;
err = register_netdevice(tun->dev);
if (err < 0)
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
goto err_free_sk;
if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
device_create_file(&tun->dev->dev, &dev_attr_owner) ||
device_create_file(&tun->dev->dev, &dev_attr_group))
pr_err("Failed to create tun sysfs files\n");
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
sk->sk_destruct = tun_sock_destruct;
err = tun_attach(tun, file);
if (err < 0)
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
goto failed;
}
tun_debug(KERN_INFO, tun, "tun_set_iff\n");
if (ifr->ifr_flags & IFF_NO_PI)
tun->flags |= TUN_NO_PI;
else
tun->flags &= ~TUN_NO_PI;
if (ifr->ifr_flags & IFF_ONE_QUEUE)
tun->flags |= TUN_ONE_QUEUE;
else
tun->flags &= ~TUN_ONE_QUEUE;
if (ifr->ifr_flags & IFF_VNET_HDR)
tun->flags |= TUN_VNET_HDR;
else
tun->flags &= ~TUN_VNET_HDR;
/* Make sure persistent devices do not get stuck in
* xoff state.
*/
if (netif_running(tun->dev))
netif_wake_queue(tun->dev);
strcpy(ifr->ifr_name, tun->dev->name);
return 0;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
err_free_sk:
tun_free_netdev(dev);
err_free_dev:
free_netdev(dev);
failed:
return err;
}
static int tun_get_iff(struct net *net, struct tun_struct *tun,
struct ifreq *ifr)
{
tun_debug(KERN_INFO, tun, "tun_get_iff\n");
strcpy(ifr->ifr_name, tun->dev->name);
ifr->ifr_flags = tun_flags(tun);
return 0;
}
/* This is like a cut-down ethtool ops, except done via tun fd so no
* privs required. */
static int set_offload(struct tun_struct *tun, unsigned long arg)
{
netdev_features_t features = 0;
if (arg & TUN_F_CSUM) {
features |= NETIF_F_HW_CSUM;
arg &= ~TUN_F_CSUM;
if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
if (arg & TUN_F_TSO_ECN) {
features |= NETIF_F_TSO_ECN;
arg &= ~TUN_F_TSO_ECN;
}
if (arg & TUN_F_TSO4)
features |= NETIF_F_TSO;
if (arg & TUN_F_TSO6)
features |= NETIF_F_TSO6;
arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
}
if (arg & TUN_F_UFO) {
features |= NETIF_F_UFO;
arg &= ~TUN_F_UFO;
}
}
/* This gives the user a way to test for new features in future by
* trying to set them. */
if (arg)
return -EINVAL;
tun->set_features = features;
netdev_update_features(tun->dev);
return 0;
}
static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
unsigned long arg, int ifreq_len)
{
struct tun_file *tfile = file->private_data;
struct tun_struct *tun;
void __user* argp = (void __user*)arg;
struct sock_fprog fprog;
struct ifreq ifr;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
int sndbuf;
int vnet_hdr_sz;
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
int ret;
if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
if (copy_from_user(&ifr, argp, ifreq_len))
return -EFAULT;
if (cmd == TUNGETFEATURES) {
/* Currently this just means: "what IFF flags are valid?".
* This is needed because we never checked for invalid flags on
* TUNSETIFF. */
return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
IFF_VNET_HDR,
(unsigned int __user*)argp);
}
rtnl_lock();
tun = __tun_get(tfile);
if (cmd == TUNSETIFF && !tun) {
ifr.ifr_name[IFNAMSIZ-1] = '\0';
ret = tun_set_iff(tfile->net, file, &ifr);
if (ret)
goto unlock;
if (copy_to_user(argp, &ifr, ifreq_len))
ret = -EFAULT;
goto unlock;
}
ret = -EBADFD;
if (!tun)
goto unlock;
tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %d\n", cmd);
ret = 0;
switch (cmd) {
case TUNGETIFF:
ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
if (ret)
break;
if (copy_to_user(argp, &ifr, ifreq_len))
ret = -EFAULT;
break;
case TUNSETNOCSUM:
/* Disable/Enable checksum */
/* [unimplemented] */
tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
arg ? "disabled" : "enabled");
break;
case TUNSETPERSIST:
/* Disable/Enable persist mode */
if (arg)
tun->flags |= TUN_PERSIST;
else
tun->flags &= ~TUN_PERSIST;
tun_debug(KERN_INFO, tun, "persist %s\n",
arg ? "enabled" : "disabled");
break;
case TUNSETOWNER:
/* Set owner of the device */
tun->owner = (uid_t) arg;
tun_debug(KERN_INFO, tun, "owner set to %d\n", tun->owner);
break;
case TUNSETGROUP:
/* Set group of the device */
tun->group= (gid_t) arg;
tun_debug(KERN_INFO, tun, "group set to %d\n", tun->group);
break;
case TUNSETLINK:
/* Only allow setting the type when the interface is down */
if (tun->dev->flags & IFF_UP) {
tun_debug(KERN_INFO, tun,
"Linktype set failed because interface is up\n");
ret = -EBUSY;
} else {
tun->dev->type = (int) arg;
tun_debug(KERN_INFO, tun, "linktype set to %d\n",
tun->dev->type);
ret = 0;
}
break;
#ifdef TUN_DEBUG
case TUNSETDEBUG:
tun->debug = arg;
break;
#endif
case TUNSETOFFLOAD:
ret = set_offload(tun, arg);
break;
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
case TUNSETTXFILTER:
/* Can be set only for TAPs */
ret = -EINVAL;
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
break;
ret = update_filter(&tun->txflt, (void __user *)arg);
break;
case SIOCGIFHWADDR:
/* Get hw address */
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
ifr.ifr_hwaddr.sa_family = tun->dev->type;
if (copy_to_user(argp, &ifr, ifreq_len))
ret = -EFAULT;
break;
case SIOCSIFHWADDR:
tun: Fix/rewrite packet filtering logic Please see the following thread to get some context on this http://marc.info/?l=linux-netdev&m=121564433018903&w=2 Basically the issue is that current multi-cast filtering stuff in the TUN/TAP driver is seriously broken. Original patch went in without proper review and ACK. It was broken and confusing to start with and subsequent patches broke it completely. To give you an idea of what's broken here are some of the issues: - Very confusing comments throughout the code that imply that the character device is a network interface in its own right, and that packets are passed between the two nics. Which is completely wrong. - Wrong set of ioctls is used for setting up filters. They look like shortcuts for manipulating state of the tun/tap network interface but in reality manipulate the state of the TX filter. - ioctls that were originally used for setting address of the the TX filter got "fixed" and now set the address of the network interface itself. Which made filter totaly useless. - Filtering is done too late. Instead of filtering early on, to avoid unnecessary wakeups, filtering is done in the read() call. The list goes on and on :) So the patch cleans all that up. It introduces simple and clean interface for setting up TX filters (TUNSETTXFILTER + tun_filter spec) and does filtering before enqueuing the packets. TX filtering is useful in the scenarios where TAP is part of a bridge, in which case it gets all broadcast, multicast and potentially other packets when the bridge is learning. So for example Ethernet tunnelling app may want to setup TX filters to avoid tunnelling multicast traffic. QEMU and other hypervisors can push RX filtering that is currently done in the guest into the host context therefore saving wakeups and unnecessary data transfer. Signed-off-by: Max Krasnyansky <maxk@qualcomm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2008-07-15 05:18:19 +00:00
/* Set hw address */
tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
ifr.ifr_hwaddr.sa_data);
ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
break;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
case TUNGETSNDBUF:
sndbuf = tun->socket.sk->sk_sndbuf;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
ret = -EFAULT;
break;
case TUNSETSNDBUF:
if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
ret = -EFAULT;
break;
}
tun->socket.sk->sk_sndbuf = sndbuf;
tun: Limit amount of queued packets per device Unlike a normal socket path, the tuntap device send path does not have any accounting. This means that the user-space sender may be able to pin down arbitrary amounts of kernel memory by continuing to send data to an end-point that is congested. Even when this isn't an issue because of limited queueing at most end points, this can also be a problem because its only response to congestion is packet loss. That is, when those local queues at the end-point fills up, the tuntap device will start wasting system time because it will continue to send data there which simply gets dropped straight away. Of course one could argue that everybody should do congestion control end-to-end, unfortunately there are people in this world still hooked on UDP, and they don't appear to be going away anywhere fast. In fact, we've always helped them by performing accounting in our UDP code, the sole purpose of which is to provide congestion feedback other than through packet loss. This patch attempts to apply the same bandaid to the tuntap device. It creates a pseudo-socket object which is used to account our packets just as a normal socket does for UDP. Of course things are a little complex because we're actually reinjecting traffic back into the stack rather than out of the stack. The stack complexities however should have been resolved by preceding patches. So this one can simply start using skb_set_owner_w. For now the accounting is essentially disabled by default for backwards compatibility. In particular, we set the cap to INT_MAX. This is so that existing applications don't get confused by the sudden arrival EAGAIN errors. In future we may wish (or be forced to) do this by default. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-02-06 05:25:32 +00:00
break;
case TUNGETVNETHDRSZ:
vnet_hdr_sz = tun->vnet_hdr_sz;
if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
ret = -EFAULT;
break;
case TUNSETVNETHDRSZ:
if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
ret = -EFAULT;
break;
}
if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
ret = -EINVAL;
break;
}
tun->vnet_hdr_sz = vnet_hdr_sz;
break;
case TUNATTACHFILTER:
/* Can be set only for TAPs */
ret = -EINVAL;
if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
break;
ret = -EFAULT;
if (copy_from_user(&fprog, argp, sizeof(fprog)))
break;
ret = sk_attach_filter(&fprog, tun->socket.sk);
break;
case TUNDETACHFILTER:
/* Can be set only for TAPs */
ret = -EINVAL;
if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
break;
ret = sk_detach_filter(tun->socket.sk);
break;
default:
ret = -EINVAL;
break;
}
unlock:
rtnl_unlock();
if (tun)
tun_put(tun);
return ret;
}
static long tun_chr_ioctl(struct file *file,
unsigned int cmd, unsigned long arg)
{
return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
}
#ifdef CONFIG_COMPAT
static long tun_chr_compat_ioctl(struct file *file,
unsigned int cmd, unsigned long arg)
{
switch (cmd) {
case TUNSETIFF:
case TUNGETIFF:
case TUNSETTXFILTER:
case TUNGETSNDBUF:
case TUNSETSNDBUF:
case SIOCGIFHWADDR:
case SIOCSIFHWADDR:
arg = (unsigned long)compat_ptr(arg);
break;
default:
arg = (compat_ulong_t)arg;
break;
}
/*
* compat_ifreq is shorter than ifreq, so we must not access beyond
* the end of that structure. All fields that are used in this
* driver are compatible though, we don't need to convert the
* contents.
*/
return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
}
#endif /* CONFIG_COMPAT */
static int tun_chr_fasync(int fd, struct file *file, int on)
{
struct tun_struct *tun = tun_get(file);
int ret;
if (!tun)
return -EBADFD;
tun_debug(KERN_INFO, tun, "tun_chr_fasync %d\n", on);
if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
goto out;
if (on) {
ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
if (ret)
goto out;
tun->flags |= TUN_FASYNC;
} else
tun->flags &= ~TUN_FASYNC;
ret = 0;
out:
tun_put(tun);
return ret;
}
static int tun_chr_open(struct inode *inode, struct file * file)
{
struct tun_file *tfile;
DBG1(KERN_INFO, "tunX: tun_chr_open\n");
tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
if (!tfile)
return -ENOMEM;
atomic_set(&tfile->count, 0);
tfile->tun = NULL;
tfile->net = get_net(current->nsproxy->net_ns);
file->private_data = tfile;
return 0;
}
static int tun_chr_close(struct inode *inode, struct file *file)
{
struct tun_file *tfile = file->private_data;
struct tun_struct *tun;
tun = __tun_get(tfile);
if (tun) {
struct net_device *dev = tun->dev;
tun_debug(KERN_INFO, tun, "tun_chr_close\n");
__tun_detach(tun);
/* If desirable, unregister the netdevice. */
if (!(tun->flags & TUN_PERSIST)) {
rtnl_lock();
if (dev->reg_state == NETREG_REGISTERED)
unregister_netdevice(dev);
rtnl_unlock();
}
}
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
tun = tfile->tun;
if (tun)
sock_put(tun->socket.sk);
tun: Only free a netdev when all tun descriptors are closed The commit c70f182940f988448f3c12a209d18b1edc276e33 ("tun: Fix races between tun_net_close and free_netdev") fixed a race where an asynchronous deletion of a tun device can hose a poll(2) on a tun fd attached to that device. However, this came at the cost of moving the tun wait queue into the tun file data structure. The problem with this is that it imposes restrictions on when and where the tun device can access the wait queue since the tun file may change at any time due to detaching and reattaching. In particular, now that we need to use the wait queue on the receive path it becomes difficult to properly synchronise this with the detachment of the tun device. This patch solves the original race in a different way. Since the race is only because the underlying memory gets freed, we can prevent it simply by ensuring that we don't do that until all tun descriptors ever attached to the device (even if they have since be detached because they may still be sitting in poll) have been closed. This is done by using reference counting the attached tun file descriptors. The refcount in tun->sk has been reappropriated for this purpose since it was already being used for that, albeit from the opposite angle. Note that we no longer zero tfile->tun since tun_get will return NULL anyway after the refcount on tfile hits zero. Instead it represents whether this device has ever been attached to a device. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-04-18 14:15:52 +00:00
put_net(tfile->net);
kfree(tfile);
return 0;
}
static const struct file_operations tun_fops = {
.owner = THIS_MODULE,
.llseek = no_llseek,
.read = do_sync_read,
.aio_read = tun_chr_aio_read,
.write = do_sync_write,
.aio_write = tun_chr_aio_write,
.poll = tun_chr_poll,
.unlocked_ioctl = tun_chr_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = tun_chr_compat_ioctl,
#endif
.open = tun_chr_open,
.release = tun_chr_close,
.fasync = tun_chr_fasync
};
static struct miscdevice tun_miscdev = {
.minor = TUN_MINOR,
.name = "tun",
.nodename = "net/tun",
.fops = &tun_fops,
};
/* ethtool interface */
static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
{
cmd->supported = 0;
cmd->advertising = 0;
ethtool_cmd_speed_set(cmd, SPEED_10);
cmd->duplex = DUPLEX_FULL;
cmd->port = PORT_TP;
cmd->phy_address = 0;
cmd->transceiver = XCVR_INTERNAL;
cmd->autoneg = AUTONEG_DISABLE;
cmd->maxtxpkt = 0;
cmd->maxrxpkt = 0;
return 0;
}
static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
{
struct tun_struct *tun = netdev_priv(dev);
strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
strlcpy(info->version, DRV_VERSION, sizeof(info->version));
switch (tun->flags & TUN_TYPE_MASK) {
case TUN_TUN_DEV:
strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
break;
case TUN_TAP_DEV:
strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
break;
}
}
static u32 tun_get_msglevel(struct net_device *dev)
{
#ifdef TUN_DEBUG
struct tun_struct *tun = netdev_priv(dev);
return tun->debug;
#else
return -EOPNOTSUPP;
#endif
}
static void tun_set_msglevel(struct net_device *dev, u32 value)
{
#ifdef TUN_DEBUG
struct tun_struct *tun = netdev_priv(dev);
tun->debug = value;
#endif
}
static const struct ethtool_ops tun_ethtool_ops = {
.get_settings = tun_get_settings,
.get_drvinfo = tun_get_drvinfo,
.get_msglevel = tun_get_msglevel,
.set_msglevel = tun_set_msglevel,
.get_link = ethtool_op_get_link,
};
static int __init tun_init(void)
{
int ret = 0;
pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
pr_info("%s\n", DRV_COPYRIGHT);
ret = rtnl_link_register(&tun_link_ops);
if (ret) {
pr_err("Can't register link_ops\n");
goto err_linkops;
}
ret = misc_register(&tun_miscdev);
if (ret) {
pr_err("Can't register misc device %d\n", TUN_MINOR);
goto err_misc;
}
return 0;
err_misc:
rtnl_link_unregister(&tun_link_ops);
err_linkops:
return ret;
}
static void tun_cleanup(void)
{
misc_deregister(&tun_miscdev);
rtnl_link_unregister(&tun_link_ops);
}
/* Get an underlying socket object from tun file. Returns error unless file is
* attached to a device. The returned object works like a packet socket, it
* can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
* holding a reference to the file for as long as the socket is in use. */
struct socket *tun_get_socket(struct file *file)
{
struct tun_struct *tun;
if (file->f_op != &tun_fops)
return ERR_PTR(-EINVAL);
tun = tun_get(file);
if (!tun)
return ERR_PTR(-EBADFD);
tun_put(tun);
return &tun->socket;
}
EXPORT_SYMBOL_GPL(tun_get_socket);
module_init(tun_init);
module_exit(tun_cleanup);
MODULE_DESCRIPTION(DRV_DESCRIPTION);
MODULE_AUTHOR(DRV_COPYRIGHT);
MODULE_LICENSE("GPL");
MODULE_ALIAS_MISCDEV(TUN_MINOR);
driver core: add devname module aliases to allow module on-demand auto-loading This adds: alias: devname:<name> to some common kernel modules, which will allow the on-demand loading of the kernel module when the device node is accessed. Ideally all these modules would be compiled-in, but distros seems too much in love with their modularization that we need to cover the common cases with this new facility. It will allow us to remove a bunch of pretty useless init scripts and modprobes from init scripts. The static device node aliases will be carried in the module itself. The program depmod will extract this information to a file in the module directory: $ cat /lib/modules/2.6.34-00650-g537b60d-dirty/modules.devname # Device nodes to trigger on-demand module loading. microcode cpu/microcode c10:184 fuse fuse c10:229 ppp_generic ppp c108:0 tun net/tun c10:200 dm_mod mapper/control c10:235 Udev will pick up the depmod created file on startup and create all the static device nodes which the kernel modules specify, so that these modules get automatically loaded when the device node is accessed: $ /sbin/udevd --debug ... static_dev_create_from_modules: mknod '/dev/cpu/microcode' c10:184 static_dev_create_from_modules: mknod '/dev/fuse' c10:229 static_dev_create_from_modules: mknod '/dev/ppp' c108:0 static_dev_create_from_modules: mknod '/dev/net/tun' c10:200 static_dev_create_from_modules: mknod '/dev/mapper/control' c10:235 udev_rules_apply_static_dev_perms: chmod '/dev/net/tun' 0666 udev_rules_apply_static_dev_perms: chmod '/dev/fuse' 0666 A few device nodes are switched to statically allocated numbers, to allow the static nodes to work. This might also useful for systems which still run a plain static /dev, which is completely unsafe to use with any dynamic minor numbers. Note: The devname aliases must be limited to the *common* and *single*instance* device nodes, like the misc devices, and never be used for conceptually limited systems like the loop devices, which should rather get fixed properly and get a control node for losetup to talk to, instead of creating a random number of device nodes in advance, regardless if they are ever used. This facility is to hide the mess distros are creating with too modualized kernels, and just to hide that these modules are not compiled-in, and not to paper-over broken concepts. Thanks! :) Cc: Greg Kroah-Hartman <gregkh@suse.de> Cc: David S. Miller <davem@davemloft.net> Cc: Miklos Szeredi <miklos@szeredi.hu> Cc: Chris Mason <chris.mason@oracle.com> Cc: Alasdair G Kergon <agk@redhat.com> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Ian Kent <raven@themaw.net> Signed-Off-By: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2010-05-20 16:07:20 +00:00
MODULE_ALIAS("devname:net/tun");