Herbert Xu says:
====================
rhashtable: Multiple rehashing
This series introduces multiple rehashing.
Recall that the original implementation in br_multicast used
two list pointers per hash node and therefore is limited to at
most one rehash at a time since you need one list pointer for
the old table and one for the new table.
Thanks to Josh Triplett's suggestion of using a single list pointer
we're no longer limited by that. So it is perfectly OK to have
an arbitrary number of tables in existence at any one time.
The reader and removal simply has to walk from the oldest table
to the newest table in order not to miss anything. Insertion
without lookup are just as easy as we simply go to the last table
that we can find and add the entry there.
However, insertion with uniqueness lookup is more complicated
because we need to ensure that two simultaneous insertions of the
same key do not both succeed. To achieve this, all insertions
including those without lookups are required to obtain the bucket
lock from the oldest hash table that is still alive. This is
determined by having the rehasher (there is only one rehashing
thread in the system) keep a pointer of where it is up to. If
a bucket has already been rehashed then it is dead, i.e., there
cannot be any more insertions to it, otherwise it is considered
alive. This guarantees that the same key cannot be inserted
in two different tables in parallel.
Patch 1 is actually a bug fix for the walker.
Patch 2-5 eliminates unnecessary out-of-line copies of jhash.
Patch 6 makes rhashtable_shrink shrink to fit.
Patch 7 introduces multiple rehashing. This means that if we
decide to grow then we will grow regardless of whether the previous
one has finished. However, this is still asynchronous meaning
that if insertions come fast enough we may still end up with a
table that is overutilised.
Patch 8 adds support for GFP_ATOMIC allocations of struct bucket_table.
Finally patch 9 enables immediate rehashing. This is done either
when the table reaches 100% utilisation, or when the chain length
exceeds 16 (the latter can be disabled on request, e.g., for
nft_hash.
With these patches the system should no longer have any trouble
dealing with fast insertions on a small table. In the worst
case you end up with a list of tables that's log N in length
while the rehasher catches up.
v3 restores rhashtable_shrink and fixes a number of bugs in the
multiple rehashing patches (7 and 9).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch reintroduces immediate rehash during insertion. If
we find during insertion that the table is full or the chain
length exceeds a set limit (currently 16 but may be disabled
with insecure_elasticity) then we will force an immediate rehash.
The rehash will contain an expansion if the table utilisation
exceeds 75%.
If this rehash fails then the insertion will fail. Otherwise the
insertion will be reattempted in the new hash table.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the ability to allocate bucket table with GFP_ATOMIC
instead of GFP_KERNEL. This is needed when we perform an immediate
rehash during insertion.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the missing bits to allow multiple rehashes. The
read-side as well as remove already handle this correctly. So it's
only the rehasher and insertion that need modification to handle
this.
Note that this patch doesn't actually enable it so for now rehashing
is still only performed by the worker thread.
This patch also disables the explicit expand/shrink interface because
the table is meant to expand and shrink automatically, and continuing
to export these interfaces unnecessarily complicates the life of the
rehasher since the rehash process is now composed of two parts.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch changes rhashtable_shrink to shrink to the smallest
size possible rather than halving the table. This is needed
because with multiple rehashing we will defer shrinking until
all other rehashing is done, meaning that when we do shrink
we may be able to shrink a lot.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes the explicit jhash value for the hashfn parameter
of rhashtable. The default is now jhash so removing the setting
makes no difference apart from making one less copy of jhash in
the kernel.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes the explicit jhash value for the hashfn parameter
of rhashtable. As the key length is a multiple of 4, this means that
we will actually end up using jhash2.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since every current rhashtable user uses jhash as their hash
function, the fact that jhash is an inline function causes each
user to generate a copy of its code.
This function provides a solution to this problem by allowing
hashfn to be unset. In which case rhashtable will automatically
set it to jhash. Furthermore, if the key length is a multiple
of 4, we will switch over to jhash2.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
When rht_key_hashfn is called from rhashtable itself and params
is equal to ht->p, there is no point in checking params.key_len
and falling back to ht->p.key_len.
For some reason gcc couldn't figure out that params is the same
as ht->p. So let's help it by only checking params.key_len when
it's a constant.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The walker is a lockless reader so it too needs an smp_rmb before
reading the future_tbl field in order to see any new tables that
may contain elements that we should have walked over.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=m6GG
-----END PGP SIGNATURE-----
Merge tag 'linux-can-next-for-4.1-20150323' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next
Marc Kleine-Budde says:
====================
pull-request: can-next 2015-03-23
this is a pull request of 6 patches for net-next/master.
A patch by Florian Westphal, converts the skb->destructor to use
sock_efree() instead of own destructor. Ahmed S. Darwish's patch
converts the kvaser_usb driver to use unregister_candev(). A patch by
me removes a return from a void function in the m_can driver. Yegor
Yefremov contributes a patch for combined rx/tx LED trigger support. A
sparse warning in the esd_usb2 driver was fixes by Thomas Körper. Ben
Dooks converts the at91_can driver to use endian agnostic IO accessors.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
Netfilter updates for net-next
The following patchset contains Netfilter updates for net-next.
Basically, more incremental updates for br_netfilter from Florian
Westphal, small nf_tables updates (including one fix for rb-tree
locking) and small two-liner to add extra validation for the REJECT6
target.
More specifically, they are:
1) Use the conntrack status flags from br_netfilter to know that DNAT is
happening. Patch for Florian Westphal.
2) nf_bridge->physoutdev == NULL already indicates that the traffic is
bridged, so let's get rid of the BRNF_BRIDGED flag. Also from Florian.
3) Another patch to prepare voidization of seq_printf/seq_puts/seq_putc,
from Joe Perches.
4) Consolidation of nf_tables_newtable() error path.
5) Kill nf_bridge_pad used by br_netfilter from ip_fragment(),
from Florian Westphal.
6) Access rb-tree root node inside the lock and remove unnecessary
locking from the get path (we already hold nfnl_lock there), from
Patrick McHardy.
7) You cannot use a NFT_SET_ELEM_INTERVAL_END when the set doesn't
support interval, also from Patrick.
8) Enforce IP6T_F_PROTO from ip6t_REJECT to make sure the core is
actually restricting matches to TCP.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce TP_STATUS_CSUM_VALID tp_status flag to tell the
af_packet user that at least the transport header checksum
has been already validated.
For now, the flag may be set for incoming packets only.
Signed-off-by: Alexander Drozdov <al.drozdov@gmail.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is just an optimization. We don't need the value of status variable
if the packet is filtered.
Signed-off-by: Alexander Drozdov <al.drozdov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dan Carpenter's static checker warned that in vxlan_stop we are checking
if 'vs' can be NULL while later we simply derreference it.
As after commit 56ef9c909b ("vxlan: Move socket initialization to
within rtnl scope") 'vs' just cannot be NULL in vxlan_stop() anymore, as
the interface won't go up if the socket initialization fails. So we are
good to just remove the check and make it consistent.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When I updated the code to address a possible null pointer dereference in
resize I ended up reverting an exception handling fix for the suffix length
in the event that inflate or halve failed. This change is meant to correct
that by reverting the earlier fix and instead simply getting the parent
again after inflate has been completed to avoid the possible null pointer
issue.
Fixes: ddb4b9a13 ("fib_trie: Address possible NULL pointer dereference in resize")
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Always use software checksumming, since the hardware does not have any
checksum offload support.
This significantly improves local TCP tx performance.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The start-of-frame and end-of-frame bits were accidentally swapped.
In the current code it does not make any difference, since they are
always used together.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We implement the SO_SNDLOWAT etc not to be settable and return
ENOPROTOOPT per 1003.1g 7. Move the comment to appropriate
position and update the reference.
Signed-off-by: YOSHIFUJI Hideaki <hideaki.yoshifuji@miraclelinux.com>
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet says:
====================
tcp listener refactoring part 15
I am trying to make the final patch pushing request socks into ehash
as small as possible. In this patch series, I made various adjustments
for the SYNACK generation, allowing me to reach 1 Mpps SYNACK in my
stress test (still hitting LISTENER spinlock of course, and the syn_wait
spinlock)
I also converted the ICMP handlers a bit ahead of time :
They no longer need to get the LISTENER socket, and can use
only a lookup in ehash table. No big deal if we ignore ICMP
for requests socks before the final steps.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
dccp_v6_err() can restrict lookups to ehash table, and not to listeners.
Note this patch creates the infrastructure, but this means that ICMP
messages for request sockets are ignored until complete conversion.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
dccp_v4_err() can restrict lookups to ehash table, and not to listeners.
Note this patch creates the infrastructure, but this means that ICMP
messages for request sockets are ignored until complete conversion.
New dccp_req_err() helper is exported so that we can use it in IPv6
in following patch.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_v6_err() can restrict lookups to ehash table, and not to listeners.
Note this patch creates the infrastructure, but this means that ICMP
messages for request sockets are ignored until complete conversion.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_v4_err() can restrict lookups to ehash table, and not to listeners.
Note this patch creates the infrastructure, but this means that ICMP
messages for request sockets are ignored until complete conversion.
New tcp_req_err() helper is exported so that we can use it in IPv6
in following patch.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a low hanging fruit, as we'll get rid of syn_wait_lock eventually.
We hold syn_wait_lock for such small sections, that it makes no sense to use
a read/write lock. A spin lock is simply faster.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
listener can be source of false sharing. request sock has some
useful information like : ireq->ir_iif, ireq->ir_num, ireq->ireq_net
This patch does not solve the major problem of having to read
sk->sk_protocol which is sharing a cache line with sk->sk_wmem_alloc.
(This same field is read later in ip_build_and_send_pkt())
One idea would be to move sk_protocol close to sk_family
(using 8 bits instead of 16 for sk_family seems enough)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is not needed, and req->sk_listener points to the listener anyway.
request_sock argument can be const.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cache listen_sock_qlen() to limit false sharing, and read
rskq_defer_accept once as it might change under us.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tilman Schmidt says:
====================
isdn/gigaset: restructure modem response parser
This series of patches restructures the Gigaset ISDN driver's
modem response parser to improve code readability and conform
better to the device's specification and actual behaviour.
Could you please merge these through net-next?
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Restructure the control structure of the modem response parser
to improve readability and error handling.
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
Separate literal string handling from main parser loop.
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
Factor out queueing of modem response events into helper function
add_cid_event().
Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
make it same as the netdev_switch_port_bridge_setlink/dellink
api (ie traverse lowerdevs to get to the switch port).
removes "WARN_ON(!ops->ndo_switch_parent_id_get)" because
direct bridge ports can be stacked netdevices (like bonds
and team of switch ports) which may not implement this ndo.
v2 to v3:
- remove changes to bond and team. Bring back the
transparently following lowerdevs like i initially
had for setlink/getlink
(http://www.spinics.net/lists/netdev/msg313436.html)
dave and scott feldman also seem to prefer it be that
way and move to non-transparent way of doing things
if we see a problem down the lane.
v3 to v4:
- fix ret initialization
v4 to v5:
- return err on first failure (scott feldman)
v5 to v6:
- change variable name (err) and initialize to
-EOPNOTSUPP (scott feldman).
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Acked-by: Scott Feldman <sfeldma@gmail.com>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb->priority can be set for two purposes:
1) With respect to IP TOS field, which is computed by a mask.
Ususally used for priority qdisc's (pfifo, prio etc.), on TX
side (we only have ingress qdisc on RX side).
2) Used as a classid or flowid, works in the same way with tc
classid. What's more, this can even override the classid
of tc filters.
For case 1), it has been respected within its netns, I don't
see any point of keeping it for another netns, especially
when packets will be forwarded to Rx path (no matter from TX
path or RX path).
For case 2) we care, our applications run inside a netns,
and we classify the packets by our own filters outside,
If some application sets this priority, it could bypass
our filters, therefore clear it when moving out of a netns,
it makes no sense to bypass tc filters out of its netns.
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tadeusz Struk says:
====================
Add support for async socket operations
After the iocb parameter has been removed from sendmsg() and recvmsg() ops
the socket layer, and the network stack no longer support async operations.
This patch set adds support for asynchronous operations on sockets back.
Changes in v3:
* As sugested by Al Viro instead of adding new functions aio_sendmsg
and aio_recvmsg, added a ptr to iocb into the kernel-side msghdr structure.
This way no change to aio.c is required.
Changes in v2:
* removed redundant total_size param from aio_sendmsg and aio_recvmsg functions
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The way the algif_skcipher works currently is that on sendmsg/sendpage it
builds an sgl for the input data and then on read/recvmsg it sends the job
for encryption putting the user to sleep till the data is processed.
This way it can only handle one job at a given time.
This patch changes it to be asynchronous by adding AIO support.
Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change __raw accesors to endian agnostic versions to allow the driver
to work properly on big endian systems.
Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The hnd field of the structs does not need to be __le32: the
device just returns the value without using it itself.
Signed-off-by: Thomas Körper <thomas.koerper@esd.eu>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Add <ifname>-rxtx trigger, that will be activated both for tx
as rx events. This trigger mimics "activity" LED for Ethernet
devices.
Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Use can-dev's unregister_candev() instead of directly calling
networking unregister_netdev(). While both are functionally
equivalent, unregister_candev() might do extra stuff in the
future than just calling networking layer unregistration code.
Signed-off-by: Ahmed S. Darwish <ahmed.darwish@valeo.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
It is identical to the can destructor.
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Make sure IP6T_F_PROTO is set to enforce layer 4 protocol matching from
the ip6_tables core.
Suggested-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Fix a race condition and unnecessary locking:
* the root rb_node must only be accessed under the lock in nft_rbtree_lookup()
* the lock is not needed in lookup functions in netlink context
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
