2009-05-08 04:02:40 +08:00
|
|
|
/* drivers/net/pppolac.c
|
|
|
|
*
|
|
|
|
* Driver for PPP on L2TP Access Concentrator / PPPoLAC Socket (RFC 2661)
|
|
|
|
*
|
|
|
|
* Copyright (C) 2009 Google, Inc.
|
|
|
|
*
|
|
|
|
* This software is licensed under the terms of the GNU General Public
|
|
|
|
* License version 2, as published by the Free Software Foundation, and
|
|
|
|
* may be copied, distributed, and modified under those terms.
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* This driver handles L2TP data packets between a UDP socket and a PPP channel.
|
2011-04-15 15:22:09 -07:00
|
|
|
* The socket must keep connected, and only one session per socket is permitted.
|
|
|
|
* Sequencing of outgoing packets is controlled by LNS. Incoming packets with
|
|
|
|
* sequences are reordered within a sliding window of one second. Currently
|
|
|
|
* reordering only happens when a packet is received. It is done for simplicity
|
|
|
|
* since no additional locks or threads are required. This driver only works on
|
|
|
|
* IPv4 due to the lack of UDP encapsulation support in IPv6. */
|
2009-05-08 04:02:40 +08:00
|
|
|
|
|
|
|
#include <linux/module.h>
|
2011-04-15 15:22:09 -07:00
|
|
|
#include <linux/jiffies.h>
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
#include <linux/workqueue.h>
|
2009-05-08 04:02:40 +08:00
|
|
|
#include <linux/skbuff.h>
|
|
|
|
#include <linux/file.h>
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
#include <linux/netdevice.h>
|
2009-05-08 04:02:40 +08:00
|
|
|
#include <linux/net.h>
|
|
|
|
#include <linux/udp.h>
|
|
|
|
#include <linux/ppp_defs.h>
|
|
|
|
#include <linux/if_ppp.h>
|
|
|
|
#include <linux/if_pppox.h>
|
|
|
|
#include <linux/ppp_channel.h>
|
|
|
|
#include <net/tcp_states.h>
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
#include <asm/uaccess.h>
|
2009-05-08 04:02:40 +08:00
|
|
|
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
#define L2TP_CONTROL_BIT 0x80
|
|
|
|
#define L2TP_LENGTH_BIT 0x40
|
|
|
|
#define L2TP_SEQUENCE_BIT 0x08
|
|
|
|
#define L2TP_OFFSET_BIT 0x02
|
2009-05-08 04:02:40 +08:00
|
|
|
#define L2TP_VERSION 0x02
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
#define L2TP_VERSION_MASK 0x0F
|
2009-05-08 04:02:40 +08:00
|
|
|
|
|
|
|
#define PPP_ADDR 0xFF
|
|
|
|
#define PPP_CTRL 0x03
|
|
|
|
|
|
|
|
union unaligned {
|
|
|
|
__u32 u32;
|
|
|
|
} __attribute__((packed));
|
|
|
|
|
|
|
|
static inline union unaligned *unaligned(void *ptr)
|
|
|
|
{
|
|
|
|
return (union unaligned *)ptr;
|
|
|
|
}
|
|
|
|
|
2011-04-15 15:22:09 -07:00
|
|
|
struct meta {
|
|
|
|
__u32 sequence;
|
|
|
|
__u32 timestamp;
|
|
|
|
};
|
|
|
|
|
|
|
|
static inline struct meta *skb_meta(struct sk_buff *skb)
|
|
|
|
{
|
|
|
|
return (struct meta *)skb->cb;
|
|
|
|
}
|
|
|
|
|
|
|
|
/******************************************************************************/
|
|
|
|
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
static int pppolac_recv_core(struct sock *sk_udp, struct sk_buff *skb)
|
2009-05-08 04:02:40 +08:00
|
|
|
{
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
struct sock *sk = (struct sock *)sk_udp->sk_user_data;
|
|
|
|
struct pppolac_opt *opt = &pppox_sk(sk)->proto.lac;
|
2011-04-15 15:22:09 -07:00
|
|
|
struct meta *meta = skb_meta(skb);
|
|
|
|
__u32 now = jiffies;
|
2009-05-08 04:02:40 +08:00
|
|
|
__u8 bits;
|
|
|
|
__u8 *ptr;
|
|
|
|
|
2011-04-15 15:22:09 -07:00
|
|
|
/* Drop the packet if L2TP header is missing. */
|
2009-05-08 04:02:40 +08:00
|
|
|
if (skb->len < sizeof(struct udphdr) + 6)
|
|
|
|
goto drop;
|
|
|
|
|
|
|
|
/* Put it back if it is a control packet. */
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
if (skb->data[sizeof(struct udphdr)] & L2TP_CONTROL_BIT)
|
2018-04-13 12:56:57 +05:30
|
|
|
return 2;
|
2009-05-08 04:02:40 +08:00
|
|
|
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
/* Skip UDP header. */
|
2009-05-08 04:02:40 +08:00
|
|
|
skb_pull(skb, sizeof(struct udphdr));
|
|
|
|
|
|
|
|
/* Check the version. */
|
|
|
|
if ((skb->data[1] & L2TP_VERSION_MASK) != L2TP_VERSION)
|
|
|
|
goto drop;
|
|
|
|
bits = skb->data[0];
|
|
|
|
ptr = &skb->data[2];
|
|
|
|
|
|
|
|
/* Check the length if it is present. */
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
if (bits & L2TP_LENGTH_BIT) {
|
2009-05-08 04:02:40 +08:00
|
|
|
if ((ptr[0] << 8 | ptr[1]) != skb->len)
|
|
|
|
goto drop;
|
|
|
|
ptr += 2;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Skip all fields including optional ones. */
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
if (!skb_pull(skb, 6 + (bits & L2TP_SEQUENCE_BIT ? 4 : 0) +
|
|
|
|
(bits & L2TP_LENGTH_BIT ? 2 : 0) +
|
|
|
|
(bits & L2TP_OFFSET_BIT ? 2 : 0)))
|
2009-05-08 04:02:40 +08:00
|
|
|
goto drop;
|
|
|
|
|
|
|
|
/* Skip the offset padding if it is present. */
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
if (bits & L2TP_OFFSET_BIT &&
|
2009-05-08 04:02:40 +08:00
|
|
|
!skb_pull(skb, skb->data[-2] << 8 | skb->data[-1]))
|
|
|
|
goto drop;
|
|
|
|
|
|
|
|
/* Check the tunnel and the session. */
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
if (unaligned(ptr)->u32 != opt->local)
|
2009-05-08 04:02:40 +08:00
|
|
|
goto drop;
|
|
|
|
|
2011-04-15 15:22:09 -07:00
|
|
|
/* Check the sequence if it is present. */
|
|
|
|
if (bits & L2TP_SEQUENCE_BIT) {
|
|
|
|
meta->sequence = ptr[4] << 8 | ptr[5];
|
|
|
|
if ((__s16)(meta->sequence - opt->recv_sequence) < 0)
|
|
|
|
goto drop;
|
|
|
|
}
|
2009-05-08 04:02:40 +08:00
|
|
|
|
|
|
|
/* Skip PPP address and control if they are present. */
|
|
|
|
if (skb->len >= 2 && skb->data[0] == PPP_ADDR &&
|
|
|
|
skb->data[1] == PPP_CTRL)
|
|
|
|
skb_pull(skb, 2);
|
|
|
|
|
|
|
|
/* Fix PPP protocol if it is compressed. */
|
|
|
|
if (skb->len >= 1 && skb->data[0] & 1)
|
2017-07-10 11:28:18 +05:30
|
|
|
*(u8 *)skb_push(skb, 1) = 0;
|
2009-05-08 04:02:40 +08:00
|
|
|
|
2011-04-15 15:22:09 -07:00
|
|
|
/* Drop the packet if PPP protocol is missing. */
|
|
|
|
if (skb->len < 2)
|
|
|
|
goto drop;
|
|
|
|
|
|
|
|
/* Perform reordering if sequencing is enabled. */
|
|
|
|
atomic_set(&opt->sequencing, bits & L2TP_SEQUENCE_BIT);
|
|
|
|
if (bits & L2TP_SEQUENCE_BIT) {
|
|
|
|
struct sk_buff *skb1;
|
|
|
|
|
|
|
|
/* Insert the packet into receive queue in order. */
|
|
|
|
skb_set_owner_r(skb, sk);
|
|
|
|
skb_queue_walk(&sk->sk_receive_queue, skb1) {
|
|
|
|
struct meta *meta1 = skb_meta(skb1);
|
|
|
|
__s16 order = meta->sequence - meta1->sequence;
|
|
|
|
if (order == 0)
|
|
|
|
goto drop;
|
|
|
|
if (order < 0) {
|
|
|
|
meta->timestamp = meta1->timestamp;
|
|
|
|
skb_insert(skb1, skb, &sk->sk_receive_queue);
|
|
|
|
skb = NULL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (skb) {
|
|
|
|
meta->timestamp = now;
|
|
|
|
skb_queue_tail(&sk->sk_receive_queue, skb);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Remove packets from receive queue as long as
|
|
|
|
* 1. the receive buffer is full,
|
|
|
|
* 2. they are queued longer than one second, or
|
|
|
|
* 3. there are no missing packets before them. */
|
|
|
|
skb_queue_walk_safe(&sk->sk_receive_queue, skb, skb1) {
|
|
|
|
meta = skb_meta(skb);
|
|
|
|
if (atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
|
|
|
|
now - meta->timestamp < HZ &&
|
|
|
|
meta->sequence != opt->recv_sequence)
|
|
|
|
break;
|
|
|
|
skb_unlink(skb, &sk->sk_receive_queue);
|
|
|
|
opt->recv_sequence = (__u16)(meta->sequence + 1);
|
|
|
|
skb_orphan(skb);
|
|
|
|
ppp_input(&pppox_sk(sk)->chan, skb);
|
|
|
|
}
|
|
|
|
return NET_RX_SUCCESS;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Flush receive queue if sequencing is disabled. */
|
|
|
|
skb_queue_purge(&sk->sk_receive_queue);
|
2009-05-08 04:02:40 +08:00
|
|
|
skb_orphan(skb);
|
|
|
|
ppp_input(&pppox_sk(sk)->chan, skb);
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
return NET_RX_SUCCESS;
|
2009-05-08 04:02:40 +08:00
|
|
|
drop:
|
|
|
|
kfree_skb(skb);
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
return NET_RX_DROP;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int pppolac_recv(struct sock *sk_udp, struct sk_buff *skb)
|
|
|
|
{
|
2018-04-13 12:56:57 +05:30
|
|
|
int retval;
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
sock_hold(sk_udp);
|
2018-04-13 12:56:57 +05:30
|
|
|
retval = sk_receive_skb(sk_udp, skb, 0);
|
|
|
|
return (retval >> 1);
|
2009-05-08 04:02:40 +08:00
|
|
|
}
|
|
|
|
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
static struct sk_buff_head delivery_queue;
|
|
|
|
|
|
|
|
static void pppolac_xmit_core(struct work_struct *delivery_work)
|
|
|
|
{
|
|
|
|
mm_segment_t old_fs = get_fs();
|
|
|
|
struct sk_buff *skb;
|
|
|
|
|
|
|
|
set_fs(KERNEL_DS);
|
|
|
|
while ((skb = skb_dequeue(&delivery_queue))) {
|
|
|
|
struct sock *sk_udp = skb->sk;
|
|
|
|
struct kvec iov = {.iov_base = skb->data, .iov_len = skb->len};
|
2017-04-28 12:53:04 -06:00
|
|
|
struct msghdr msg = {
|
|
|
|
.msg_flags = MSG_NOSIGNAL | MSG_DONTWAIT,
|
|
|
|
};
|
2016-03-01 09:44:17 -08:00
|
|
|
|
|
|
|
iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, &iov, 1,
|
|
|
|
skb->len);
|
2015-12-08 18:26:39 +05:30
|
|
|
sk_udp->sk_prot->sendmsg(sk_udp, &msg, skb->len);
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
kfree_skb(skb);
|
|
|
|
}
|
|
|
|
set_fs(old_fs);
|
|
|
|
}
|
|
|
|
|
|
|
|
static DECLARE_WORK(delivery_work, pppolac_xmit_core);
|
|
|
|
|
2009-05-08 04:02:40 +08:00
|
|
|
static int pppolac_xmit(struct ppp_channel *chan, struct sk_buff *skb)
|
|
|
|
{
|
|
|
|
struct sock *sk_udp = (struct sock *)chan->private;
|
|
|
|
struct pppolac_opt *opt = &pppox_sk(sk_udp->sk_user_data)->proto.lac;
|
|
|
|
|
|
|
|
/* Install PPP address and control. */
|
|
|
|
skb_push(skb, 2);
|
|
|
|
skb->data[0] = PPP_ADDR;
|
|
|
|
skb->data[1] = PPP_CTRL;
|
|
|
|
|
|
|
|
/* Install L2TP header. */
|
2011-04-15 15:22:09 -07:00
|
|
|
if (atomic_read(&opt->sequencing)) {
|
2009-05-08 04:02:40 +08:00
|
|
|
skb_push(skb, 10);
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
skb->data[0] = L2TP_SEQUENCE_BIT;
|
2011-04-15 15:22:09 -07:00
|
|
|
skb->data[6] = opt->xmit_sequence >> 8;
|
|
|
|
skb->data[7] = opt->xmit_sequence;
|
2009-05-08 04:02:40 +08:00
|
|
|
skb->data[8] = 0;
|
|
|
|
skb->data[9] = 0;
|
2011-04-15 15:22:09 -07:00
|
|
|
opt->xmit_sequence++;
|
2009-05-08 04:02:40 +08:00
|
|
|
} else {
|
|
|
|
skb_push(skb, 6);
|
|
|
|
skb->data[0] = 0;
|
|
|
|
}
|
|
|
|
skb->data[1] = L2TP_VERSION;
|
|
|
|
unaligned(&skb->data[2])->u32 = opt->remote;
|
|
|
|
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
/* Now send the packet via the delivery queue. */
|
|
|
|
skb_set_owner_w(skb, sk_udp);
|
|
|
|
skb_queue_tail(&delivery_queue, skb);
|
|
|
|
schedule_work(&delivery_work);
|
2009-05-08 04:02:40 +08:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/******************************************************************************/
|
|
|
|
|
|
|
|
static struct ppp_channel_ops pppolac_channel_ops = {
|
|
|
|
.start_xmit = pppolac_xmit,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int pppolac_connect(struct socket *sock, struct sockaddr *useraddr,
|
|
|
|
int addrlen, int flags)
|
|
|
|
{
|
|
|
|
struct sock *sk = sock->sk;
|
|
|
|
struct pppox_sock *po = pppox_sk(sk);
|
|
|
|
struct sockaddr_pppolac *addr = (struct sockaddr_pppolac *)useraddr;
|
|
|
|
struct socket *sock_udp = NULL;
|
|
|
|
struct sock *sk_udp;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (addrlen != sizeof(struct sockaddr_pppolac) ||
|
|
|
|
!addr->local.tunnel || !addr->local.session ||
|
|
|
|
!addr->remote.tunnel || !addr->remote.session) {
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
lock_sock(sk);
|
|
|
|
error = -EALREADY;
|
|
|
|
if (sk->sk_state != PPPOX_NONE)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
sock_udp = sockfd_lookup(addr->udp_socket, &error);
|
|
|
|
if (!sock_udp)
|
|
|
|
goto out;
|
|
|
|
sk_udp = sock_udp->sk;
|
|
|
|
lock_sock(sk_udp);
|
|
|
|
|
|
|
|
/* Remove this check when IPv6 supports UDP encapsulation. */
|
|
|
|
error = -EAFNOSUPPORT;
|
|
|
|
if (sk_udp->sk_family != AF_INET)
|
|
|
|
goto out;
|
|
|
|
error = -EPROTONOSUPPORT;
|
|
|
|
if (sk_udp->sk_protocol != IPPROTO_UDP)
|
|
|
|
goto out;
|
|
|
|
error = -EDESTADDRREQ;
|
|
|
|
if (sk_udp->sk_state != TCP_ESTABLISHED)
|
|
|
|
goto out;
|
|
|
|
error = -EBUSY;
|
|
|
|
if (udp_sk(sk_udp)->encap_type || sk_udp->sk_user_data)
|
|
|
|
goto out;
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
if (!sk_udp->sk_bound_dev_if) {
|
|
|
|
struct dst_entry *dst = sk_dst_get(sk_udp);
|
|
|
|
error = -ENODEV;
|
|
|
|
if (!dst)
|
|
|
|
goto out;
|
|
|
|
sk_udp->sk_bound_dev_if = dst->dev->ifindex;
|
|
|
|
dst_release(dst);
|
|
|
|
}
|
2009-05-08 04:02:40 +08:00
|
|
|
|
|
|
|
po->chan.hdrlen = 12;
|
|
|
|
po->chan.private = sk_udp;
|
|
|
|
po->chan.ops = &pppolac_channel_ops;
|
2012-09-20 16:34:10 -07:00
|
|
|
po->chan.mtu = PPP_MRU - 80;
|
2009-05-08 04:02:40 +08:00
|
|
|
po->proto.lac.local = unaligned(&addr->local)->u32;
|
|
|
|
po->proto.lac.remote = unaligned(&addr->remote)->u32;
|
2011-04-15 15:22:09 -07:00
|
|
|
atomic_set(&po->proto.lac.sequencing, 1);
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
po->proto.lac.backlog_rcv = sk_udp->sk_backlog_rcv;
|
2009-05-08 04:02:40 +08:00
|
|
|
|
|
|
|
error = ppp_register_channel(&po->chan);
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
sk->sk_state = PPPOX_CONNECTED;
|
|
|
|
udp_sk(sk_udp)->encap_type = UDP_ENCAP_L2TPINUDP;
|
|
|
|
udp_sk(sk_udp)->encap_rcv = pppolac_recv;
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
sk_udp->sk_backlog_rcv = pppolac_recv_core;
|
2009-05-08 04:02:40 +08:00
|
|
|
sk_udp->sk_user_data = sk;
|
|
|
|
out:
|
|
|
|
if (sock_udp) {
|
|
|
|
release_sock(sk_udp);
|
|
|
|
if (error)
|
|
|
|
sockfd_put(sock_udp);
|
|
|
|
}
|
|
|
|
release_sock(sk);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int pppolac_release(struct socket *sock)
|
|
|
|
{
|
|
|
|
struct sock *sk = sock->sk;
|
|
|
|
|
|
|
|
if (!sk)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
lock_sock(sk);
|
|
|
|
if (sock_flag(sk, SOCK_DEAD)) {
|
|
|
|
release_sock(sk);
|
|
|
|
return -EBADF;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (sk->sk_state != PPPOX_NONE) {
|
|
|
|
struct sock *sk_udp = (struct sock *)pppox_sk(sk)->chan.private;
|
|
|
|
lock_sock(sk_udp);
|
2011-04-15 15:22:09 -07:00
|
|
|
skb_queue_purge(&sk->sk_receive_queue);
|
2009-05-08 04:02:40 +08:00
|
|
|
pppox_unbind_sock(sk);
|
|
|
|
udp_sk(sk_udp)->encap_type = 0;
|
|
|
|
udp_sk(sk_udp)->encap_rcv = NULL;
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
sk_udp->sk_backlog_rcv = pppox_sk(sk)->proto.lac.backlog_rcv;
|
|
|
|
sk_udp->sk_user_data = NULL;
|
2009-05-08 04:02:40 +08:00
|
|
|
release_sock(sk_udp);
|
|
|
|
sockfd_put(sk_udp->sk_socket);
|
|
|
|
}
|
|
|
|
|
|
|
|
sock_orphan(sk);
|
|
|
|
sock->sk = NULL;
|
|
|
|
release_sock(sk);
|
|
|
|
sock_put(sk);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/******************************************************************************/
|
|
|
|
|
|
|
|
static struct proto pppolac_proto = {
|
|
|
|
.name = "PPPOLAC",
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.obj_size = sizeof(struct pppox_sock),
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct proto_ops pppolac_proto_ops = {
|
|
|
|
.family = PF_PPPOX,
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.release = pppolac_release,
|
|
|
|
.bind = sock_no_bind,
|
|
|
|
.connect = pppolac_connect,
|
|
|
|
.socketpair = sock_no_socketpair,
|
|
|
|
.accept = sock_no_accept,
|
|
|
|
.getname = sock_no_getname,
|
|
|
|
.poll = sock_no_poll,
|
|
|
|
.ioctl = pppox_ioctl,
|
|
|
|
.listen = sock_no_listen,
|
|
|
|
.shutdown = sock_no_shutdown,
|
|
|
|
.setsockopt = sock_no_setsockopt,
|
|
|
|
.getsockopt = sock_no_getsockopt,
|
|
|
|
.sendmsg = sock_no_sendmsg,
|
|
|
|
.recvmsg = sock_no_recvmsg,
|
|
|
|
.mmap = sock_no_mmap,
|
|
|
|
};
|
|
|
|
|
2015-12-08 12:47:01 +05:30
|
|
|
static int pppolac_create(struct net *net, struct socket *sock, int kern)
|
2009-05-08 04:02:40 +08:00
|
|
|
{
|
|
|
|
struct sock *sk;
|
|
|
|
|
2015-12-08 12:47:01 +05:30
|
|
|
sk = sk_alloc(net, PF_PPPOX, GFP_KERNEL, &pppolac_proto, kern);
|
2009-05-08 04:02:40 +08:00
|
|
|
if (!sk)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
sock_init_data(sock, sk);
|
|
|
|
sock->state = SS_UNCONNECTED;
|
|
|
|
sock->ops = &pppolac_proto_ops;
|
|
|
|
sk->sk_protocol = PX_PROTO_OLAC;
|
|
|
|
sk->sk_state = PPPOX_NONE;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/******************************************************************************/
|
|
|
|
|
|
|
|
static struct pppox_proto pppolac_pppox_proto = {
|
|
|
|
.create = pppolac_create,
|
|
|
|
.owner = THIS_MODULE,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init pppolac_init(void)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = proto_register(&pppolac_proto, 0);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
error = register_pppox_proto(PX_PROTO_OLAC, &pppolac_pppox_proto);
|
|
|
|
if (error)
|
|
|
|
proto_unregister(&pppolac_proto);
|
ANDROID: net: PPPoPNS and PPPoLAC fixes.
net: Fix a bitmask in PPPoPNS and rename constants in PPPoPNS and PPPoLAC.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix a potential deadlock while releasing PPPoLAC/PPPoPNS socket.
PPP driver guarantees that no thread will be executing start_xmit() after
returning from ppp_unregister_channel(). To achieve this, a spinlock (downl)
is used. In pppolac_release(), ppp_unregister_channel() is called after sk_udp
is locked. At the same time, another thread might be running in pppolac_xmit()
with downl. Thus a deadlock will occur if the thread tries to lock sk_udp.
The same situation might happen on sk_raw in pppopns_release().
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Force PPPoLAC and PPPoPNS to bind an interface before creating PPP channel.
It is common to manipulate the routing table after configuring PPP device.
Since both PPPoLAC and PPPoPNS run over IP, care must be taken to make sure
that there is no loop in the routing table.
Although this can be done by adding a host route, it might still cause
problems when the interface is down for some reason.
To solve this, this patch forces both drivers to bind an interface before
creating PPP channel, so the system will not re-route the tunneling sockets
to another interface when the original one is down. Another benefit is that
now the host route is no longer required, so there is no need to remove it
when PPP channel is closed.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Avoid sleep-inside-spinlock in PPPoLAC and PPPoPNS.
Since recv() and xmit() are called with a spinlock held, routines which might
sleep cannot be used. This issue is solved by following changes:
Incoming packets are now processed in backlog handler, recv_core(), instead of
recv(). Since backlog handler is always executed with socket spinlock held, the
requirement of ppp_input() is still satisfied.
Outgoing packets are now processed in workqueue handler, xmit_core(), instead of
xmit(). Note that kernel_sendmsg() is no longer used to prevent touching dead
sockets.
In release(), lock_sock() and pppox_unbind_sock() ensure that no thread is in
recv_core() or xmit(). Then socket handlers are restored before release_sock(),
so no packets will leak in backlog queue.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
net: Fix msg_iovlen in PPPoLAC and PPPoPNS.
Although any positive value should work (which is always true in both drivers),
the correct value should be 1.
Signed-off-by: Chia-chi Yeh <chiachi@android.com>
2009-06-13 02:29:04 +08:00
|
|
|
else
|
|
|
|
skb_queue_head_init(&delivery_queue);
|
2009-05-08 04:02:40 +08:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __exit pppolac_exit(void)
|
|
|
|
{
|
|
|
|
unregister_pppox_proto(PX_PROTO_OLAC);
|
|
|
|
proto_unregister(&pppolac_proto);
|
|
|
|
}
|
|
|
|
|
|
|
|
module_init(pppolac_init);
|
|
|
|
module_exit(pppolac_exit);
|
|
|
|
|
|
|
|
MODULE_DESCRIPTION("PPP on L2TP Access Concentrator (PPPoLAC)");
|
|
|
|
MODULE_AUTHOR("Chia-chi Yeh <chiachi@android.com>");
|
|
|
|
MODULE_LICENSE("GPL");
|