When doing output route lookups, we have to select the source address
if the user has not specified an explicit one.
First, if the route has an explicit preferred source address
specified, then we use that.
Otherwise we search the route's outgoing interface for a suitable
address.
This search can be precomputed and cached at route insertion time.
The only missing part is that we have to refresh this precomputed
value any time addresses are added or removed from the interface, and
this is accomplished by fib_update_nh_saddrs().
Signed-off-by: David S. Miller <davem@davemloft.net>
fib_semantic_match() requires that if the type doesn't signal an
automatic error, it must be of type RTN_UNICAST, RTN_LOCAL,
RTN_BROADCAST, RTN_ANYCAST, or RTN_MULTICAST.
Checking this every route lookup is pointless work.
Instead validate it during route insertion, via fib_create_info().
Also, there was nothing making sure the type value was less than
RTN_MAX, so add that missing check while we're here.
Signed-off-by: David S. Miller <davem@davemloft.net>
The only necessary parts are the src/dst addresses, the
interface indexes, the TOS, and the mark.
The rest is unnecessary bloat, which amounts to nearly
50 bytes on 64-bit.
Signed-off-by: David S. Miller <davem@davemloft.net>
rt->rt_iif is only ever inspected on input routes, for example DCCP
uses this to populate a route lookup flow key when generating replies
to another packet.
Therefore, setting it to anything other than zero on output routes
makes no sense.
Signed-off-by: David S. Miller <davem@davemloft.net>
We burn a lot of useless cycles, cpu store buffer traffic, and
memory operations memset()'ing the on-stack flow used to perform
output route lookups in __ip_route_output_key().
Only the first half of the flow object members even matter for
output route lookups in this context, specifically:
FIB rules matching cares about:
dst, src, tos, iif, oif, mark
FIB trie lookup cares about:
dst
FIB semantic match cares about:
tos, scope, oif
Therefore only initialize these specific members and elide the
memset entirely.
On Niagara2 this kills about ~300 cycles from the output route
lookup path.
Likely, we can take things further, since all callers of output
route lookups essentially throw away the on-stack flow they use.
So they don't care if we use it as a scratch-pad to compute the
final flow key.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
David noticed :
------------------
Eric, I was profiling the non-routing-cache case and something that
stuck out is the case of calling inet_getpeer() with create==0.
If an entry is not found, we have to redo the lookup under a spinlock
to make certain that a concurrent writer rebalancing the tree does
not "hide" an existing entry from us.
This makes the case of a create==0 lookup for a not-present entry
really expensive. It is on the order of 600 cpu cycles on my
Niagara2.
I added a hack to not do the relookup under the lock when create==0
and it now costs less than 300 cycles.
This is now a pretty common operation with the way we handle COW'd
metrics, so I think it's definitely worth optimizing.
-----------------
One solution is to use a seqlock instead of a spinlock to protect struct
inet_peer_base.
After a failed avl tree lookup, we can easily detect if a writer did
some changes during our lookup. Taking the lock and redo the lookup is
only necessary in this case.
Note: Add one private rcu_deref_locked() macro to place in one spot the
access to spinlock included in seqlock.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As reported by Eric:
[11483.697233] IP: [<c12b0638>] dst_release+0x18/0x60
...
[11483.697741] Call Trace:
[11483.697764] [<c12fc9d2>] udp_sendmsg+0x282/0x6e0
[11483.697790] [<c12a1c01>] ? memcpy_toiovec+0x51/0x70
[11483.697818] [<c12dbd90>] ? ip_generic_getfrag+0x0/0xb0
The pointer passed to dst_release() is -EINVAL, that's because
we leave an error pointer in the local variable "rt" by accident.
NULL it out to fix the bug.
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The patch to replace inet->cork with cork left out two spots in
__ip_append_data that can result in bogus packet construction.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
The route lookup code in icmp_send() is slightly tricky as a result of
having to handle all of the requirements of RFC 4301 host relookups.
Pull the route resolution into a seperate function, so that the error
handling and route reference counting is hopefully easier to see and
contained wholly within this new routine.
Signed-off-by: David S. Miller <davem@davemloft.net>
The UDP transmit path has been running under the socket lock
for a long time because of the corking feature. This means that
transmitting to the same socket in multiple threads does not
scale at all.
However, as most users don't actually use corking, the locking
can be removed in the common case.
This patch creates a lockless fast path where corking is not used.
Please note that this does create a slight inaccuracy in the
enforcement of socket send buffer limits. In particular, we
may exceed the socket limit by up to (number of CPUs) * (packet
size) because of the way the limit is computed.
As the primary purpose of socket buffers is to indicate congestion,
this should not be a great problem for now.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch converts UDP to use the new ip_finish_skb API. This
would then allows us to more easily use ip_make_skb which allows
UDP to run without a socket lock.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the helper ip_make_skb which is like ip_append_data
and ip_push_pending_frames all rolled into one, except that it does
not send the skb produced. The sending part is carried out by
ip_send_skb, which the transport protocol can call after it has
tweaked the skb.
It is meant to be called in cases where corking is not used should
have a one-to-one correspondence to sendmsg.
This patch also adds the helper ip_finish_skb which is meant to
be replace ip_push_pending_frames when corking is required.
Previously the protocol stack would peek at the socket write
queue and add its header to the first packet. With ip_finish_skb,
the protocol stack can directly operate on the final skb instead,
just like the non-corking case with ip_make_skb.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to allow simultaneous calls to ip_append_data on the same
socket, it must not modify any shared state in sk or inet (other
than those that are designed to allow that such as atomic counters).
This patch abstracts out write references to sk and inet_sk in
ip_append_data and its friends so that we may use the underlying
code in parallel.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
UFO doesn't really use the sk_sndmsg_* parameters so touching
them is pointless. It can't use them anyway since the whole
point of UFO is to use the original pages without copying.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ip_route_newports() is the only place in the entire kernel that
cares about the port members in the routing cache entry's lookup
flow key.
Therefore the only reason we store an entire flow inside of the
struct rtentry is for this one special case.
Rewrite ip_route_newports() such that:
1) The caller passes in the original port values, so we don't need
to use the rth->fl.fl_ip_{s,d}port values to remember them.
2) The lookup flow is constructed by hand instead of being copied
from the routing cache entry's flow.
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix a bug that undo_retrans is incorrectly decremented when undo_marker is
not set or undo_retrans is already 0. This happens when sender receives
more DSACK ACKs than packets retransmitted during the current
undo phase. This may also happen when sender receives DSACK after
the undo operation is completed or cancelled.
Fix another bug that undo_retrans is incorrectly incremented when
sender retransmits an skb and tcp_skb_pcount(skb) > 1 (TSO). This case
is rare but not impossible.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now, TCP_CHECK_TIMER is not used for debuging, it does nothing.
And, it has been there for several years, maybe 6 years.
Remove it to keep code clearer.
Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric W. Biederman reported a lockdep splat in inet_twsk_deschedule()
This is caused by inet_twsk_purge(), run from process context,
and commit 575f4cd5a5b6394577 (net: Use rcu lookups in inet_twsk_purge.)
removed the BH disabling that was necessary.
Add the BH disabling but fine grained, right before calling
inet_twsk_deschedule(), instead of whole function.
With help from Linus Torvalds and Eric W. Biederman
Reported-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Daniel Lezcano <daniel.lezcano@free.fr>
CC: Pavel Emelyanov <xemul@openvz.org>
CC: Arnaldo Carvalho de Melo <acme@redhat.com>
CC: stable <stable@kernel.org> (# 2.6.33+)
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 0dbaee3b37e118a (net: Abstract default ADVMSS behind an
accessor.) introduced a possible crash in tcp_connect_init(), when
dst->default_advmss() is called from dst_metric_advmss()
Reported-by: George Spelvin <linux@horizon.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The only troublesome bit here is __mkroute_output which wants
to override res->fi and res->type, compute those in local
variables instead.
Signed-off-by: David S. Miller <davem@davemloft.net>
GCC emits all kinds of crazy zero extensions when we go from signed
int, to unsigned short, etc. etc.
This transformation has to be legal because:
1) In tkey_extract_bits() in mask_pfx(), the values are used to
perform shifts, on which negative values are undefined by C.
2) In fib_table_lookup() we perform comparisons with unsigned
values, constants, and additions. None of which should
encounter negative values.
Signed-off-by: David S. Miller <davem@davemloft.net>
This allows avoiding multiple writes to the initial __refcnt.
The most simplest cases of wanting an initial reference of "1"
in ipv4 and ipv6 have been converted, the rest have been left
along and kept at the existing "0".
Signed-off-by: David S. Miller <davem@davemloft.net>
This also allows us to combine all the dst->flags settings and avoid
read/modify/write sequences to this struct member.
Signed-off-by: David S. Miller <davem@davemloft.net>
There's a lot of redundancy and unnecessary stack frames
in the output route creation path.
1) Make __mkroute_output() return error pointers.
2) Eliminate ip_mkroute_output() entirely, made possible by #1.
3) Call __mkroute_output() directly and handling the returning error
pointers in ip_route_output_slow().
Signed-off-by: David S. Miller <davem@davemloft.net>
Note that we do not generate the redirect netevent any longer,
because we don't create a new cached route.
Instead, once the new neighbour is bound to the cached route,
we emit a neigh update event instead.
Signed-off-by: David S. Miller <davem@davemloft.net>
The general idea is that if we learn new PMTU information, we
bump the peer genid.
This triggers the dst_ops->check() code to validate and if
necessary propagate the new PMTU value into the metrics.
Learned PMTU information self-expires.
This means that it is not necessary to kill a cached route
entry just because the PMTU information is too old.
As a consequence:
1) When the path appears unreachable (dst_ops->link_failure
or dst_ops->negative_advice) we unwind the PMTU state if
it is out of date, instead of killing the cached route.
A redirected route will still be invalidated in these
situations.
2) rt_check_expire(), rt_worker_func(), et al. are no longer
necessary at all.
Signed-off-by: David S. Miller <davem@davemloft.net>
NETDEV_NOTIFY_PEER is an explicit request by the driver to send a link
notification while NETDEV_UP/NETDEV_CHANGEADDR generate link
notifications as a sort of side effect.
In the later cases the sysctl option is present because link
notification events can have undesired effects e.g. if the link is
flapping. I don't think this applies in the case of an explicit
request from a driver.
This patch makes NETDEV_NOTIFY_PEER unconditional, if preferred we
could add a new sysctl for this case which defaults to on.
This change causes Xen post-migration ARP notifications (which cause
switches to relearn their MAC tables etc) to be sent by default.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 0c838ff1ade7 (ipv4: Consolidate all default route selection
implementations.) forgot to remove one rcu_read_unlock() from
fib_select_default().
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
