Namespace related cleaning

 * make cred_to_ucred static
 * remove unused sock_rmalloc function
Signed-off-by: NStephen Hemminger <stephen@networkplumber.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While we're at it and introduced CGROUP_NET_CLASSID, lets also make
NETPRIO_CGROUP more consistent with the rest of cgroups and rename it
into CONFIG_CGROUP_NET_PRIO so that for networking, we now have
CONFIG_CGROUP_NET_{PRIO,CLASSID}. This not only makes the CONFIG
option consistent among networking cgroups, but also among cgroups
CONFIG conventions in general as the vast majority has a prefix of
CONFIG_CGROUP_<SUBSYS>.
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: cgroups@vger.kernel.org
Acked-by: NLi Zefan <lizefan@huawei.com>
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

Zefan Li requested [1] to perform the following cleanup/refactoring:

- Split cgroupfs classid handling into net core to better express a
  possible more generic use.

- Disable module support for cgroupfs bits as the majority of other
  cgroupfs subsystems do not have that, and seems to be not wished
  from cgroup side. Zefan probably might want to follow-up for netprio
  later on.

- By this, code can be further reduced which previously took care of
  functionality built when compiled as module.

cgroupfs bits are being placed under net/core/netclassid_cgroup.c, so
that we are consistent with {netclassid,netprio}_cgroup naming that is
under net/core/ as suggested by Zefan.

No change in functionality, but only code refactoring that is being
done here.

 [1] http://patchwork.ozlabs.org/patch/304825/Suggested-by: NLi Zefan <lizefan@huawei.com>
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Cc: Zefan Li <lizefan@huawei.com>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: cgroups@vger.kernel.org
Acked-by: NLi Zefan <lizefan@huawei.com>
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

unix_dgram_recvmsg() will hold the readlock of the socket until recv
is complete.

In the same time, we may try to setsockopt(SO_PEEK_OFF) which will hang until
unix_dgram_recvmsg() will complete (which can take a while) without allowing
us to break out of it, triggering a hung task spew.

Instead, allow set_peek_off to fail, this way userspace will not hang.
Signed-off-by: NSasha Levin <sasha.levin@oracle.com>
Acked-by: NPavel Emelyanov <xemul@parallels.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

What sk_reset_txq() does is just calls function sk_tx_queue_reset(),
and sk_reset_txq() is used only in sock.h, by dst_negative_advice().
Let dst_negative_advice() calls sk_tx_queue_reset() directly so we
can remove unneeded sk_reset_txq().
Signed-off-by: NZHAO Gang <gamerh2o@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While working on virtio_net new allocation strategy to increase
payload/truesize ratio, we found that refactoring sk_page_frag_refill()
was needed.

This patch splits sk_page_frag_refill() into two parts, adding
skb_page_frag_refill() which can be used without a socket.

While we are at it, add a minimum frag size of 32 for
sk_page_frag_refill()

Michael will either use netdev_alloc_frag() from softirq context,
or skb_page_frag_refill() from process context in refill_work()
 (GFP_KERNEL allocations)
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Michael Dalton <mwdalton@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Steinar reported FQ pacing was not working for UDP flows.

It looks like the initial sk->sk_pacing_rate value of 0 was
a wrong choice. We should init it to ~0U (unlimited)

Then, TCA_FQ_FLOW_DEFAULT_RATE should be removed because it makes
no real sense. The default rate is really unlimited, and we
need to avoid a zero divide.
Reported-by: NSteinar H. Gunderson <sesse@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As mentioned in commit afe4fd06 ("pkt_sched: fq: Fair Queue packet
scheduler"), this patch adds a new socket option.

SO_MAX_PACING_RATE offers the application the ability to cap the
rate computed by transport layer. Value is in bytes per second.

u32 val = 1000000;
setsockopt(sockfd, SOL_SOCKET, SO_MAX_PACING_RATE, &val, sizeof(val));

To be effectively paced, a flow must use FQ packet scheduler.

Note that a packet scheduler takes into account the headers for its
computations. The effective payload rate depends on MSS and retransmits
if any.

I chose to make this pacing rate a SOL_SOCKET option instead of a
TCP one because this can be used by other protocols.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Adding paged frags skbs to af_unix sockets introduced a performance
regression on large sends because of additional page allocations, even
if each skb could carry at least 100% more payload than before.

We can instruct sock_alloc_send_pskb() to attempt high order
allocations.

Most of the time, it does a single page allocation instead of 8.

I added an additional parameter to sock_alloc_send_pskb() to
let other users to opt-in for this new feature on followup patches.

Tested:

Before patch :

$ netperf -t STREAM_STREAM
STREAM STREAM TEST
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

 2304  212992  212992    10.00    46861.15

After patch :

$ netperf -t STREAM_STREAM
STREAM STREAM TEST
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

 2304  212992  212992    10.00    57981.11
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Eliezer renames several *ll_poll to *busy_poll, but forgets
CONFIG_NET_LL_RX_POLL, so in case of confusion, rename it too.

Cc: Eliezer Tamir <eliezer.tamir@linux.intel.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: NCong Wang <amwang@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 547669d4 ("tcp: xps: fix reordering issues") added
unexpected reorders in case netem is used in a MQ setup for high
performance test bed.

ETH=eth0
tc qd del dev $ETH root 2>/dev/null
tc qd add dev $ETH root handle 1: mq
for i in `seq 1 32`
do
 tc qd add dev $ETH parent 1:$i netem delay 100ms
done

As all tcp packets are orphaned by netem, TCP stack believes it can
set skb->ooo_okay on all packets.

In order to allow producers to send more packets, we want to
keep sk_wmem_alloc from reaching sk_sndbuf limit.

We can do that by accounting one byte per skb in netem queues,
so that TCP stack is not fooled too much.

Tested:

With above MQ/netem setup, scaling number of concurrent flows gives
linear results and no reorders/retransmits

lpq83:~# for n in 1 10 20 30 40 50 60 70 80 90 100
 do echo -n "n:$n " ; ./super_netperf $n -H 10.7.7.84; done
n:1 198.46
n:10 2002.69
n:20 4000.98
n:30 6006.35
n:40 8020.93
n:50 10032.3
n:60 12081.9
n:70 13971.3
n:80 16009.7
n:90 17117.3
n:100 17425.5
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch moves the private error queue delivery function from the
af_packet code to the core socket method. In this way, network layers
only needing the error queue for transmit time stamping can share common
code.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename LL_SO to BUSY_POLL_SO
Rename sysctl_net_ll_{read,poll} to sysctl_busy_{read,poll}
Fix up users of these variables.
Fix documentation for sysctl.

a patch for the socket.7  man page will follow separately,
because of limitations of my mail setup.
Signed-off-by: NEliezer Tamir <eliezer.tamir@linux.intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename the file and correct all the places where it is included.
Signed-off-by: NEliezer Tamir <eliezer.tamir@linux.intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the kernel (compiled with CONFIG_PREEMPT=n) is performing the
rename of a network interface, it can end up waiting for a workqueue
to complete. If userland is able to invoke a SIOCGIFNAME ioctl or a
SO_BINDTODEVICE getsockopt in between, the kernel will deadlock due to
the fact that read_secklock_begin() will spin forever waiting for the
writer process (the one doing the interface rename) to update the
devnet_rename_seq sequence.

This patch fixes the problem by adding a helper (netdev_get_name())
and using it in the code handling the SIOCGIFNAME ioctl and
SO_BINDTODEVICE setsockopt.

The netdev_get_name() helper uses raw_seqcount_begin() to avoid
spinning forever, waiting for devnet_rename_seq->sequence to become
even. cond_resched() is used in the contended case, before retrying
the access to give the writer process a chance to finish.

The use of raw_seqcount_begin() will incur some unneeded work in the
reader process in the contended case, but this is better than
deadlocking the system.
Signed-off-by: NNicolas Schichan <nschichan@freebox.fr>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

select/poll busy-poll support.

Split sysctl value into two separate ones, one for read and one for poll.
updated Documentation/sysctl/net.txt

Add a new poll flag POLL_LL. When this flag is set, sock_poll will call
sk_poll_ll if possible. sock_poll sets this flag in its return value
to indicate to select/poll when a socket that can busy poll is found.

When poll/select have nothing to report, call the low-level
sock_poll again until we are out of time or we find something.

Once the system call finds something, it stops setting POLL_LL, so it can
return the result to the user ASAP.
Signed-off-by: NEliezer Tamir <eliezer.tamir@linux.intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

adds a socket option for low latency polling.
This allows overriding the global sysctl value with a per-socket one.
Unexport sysctl_net_ll_poll since for now it's not needed in modules.
Signed-off-by: NEliezer Tamir <eliezer.tamir@linux.intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Adds an ndo_ll_poll method and the code that supports it.
This method can be used by low latency applications to busy-poll
Ethernet device queues directly from the socket code.
sysctl_net_ll_poll controls how many microseconds to poll.
Default is zero (disabled).
Individual protocol support will be added by subsequent patches.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: NEliezer Tamir <eliezer.tamir@linux.intel.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Tested-by: NWillem de Bruijn <willemb@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The three arrays of strings: af_family_key_strings,
af_family_slock_key_strings and af_family_clock_key_strings have not
VSOCK's string
Signed-off-by: NFederico Vaga <federico.vaga@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We have seen multiple NULL dereferences in __inet6_lookup_established()

After analysis, I found that inet6_sk() could be NULL while the
check for sk_family == AF_INET6 was true.

Bug was added in linux-2.6.29 when RCU lookups were introduced in UDP
and TCP stacks.

Once an IPv6 socket, using SLAB_DESTROY_BY_RCU is inserted in a hash
table, we no longer can clear pinet6 field.

This patch extends logic used in commit fcbdf09d
("net: fix nulls list corruptions in sk_prot_alloc")

TCP/UDP/UDPLite IPv6 protocols provide their own .clear_sk() method
to make sure we do not clear pinet6 field.

At socket clone phase, we do not really care, as cloning the parent (non
NULL) pinet6 is not adding a fatal race.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The callers always pass current to sock_update_netprio().
Signed-off-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The callers always pass current to sock_update_classid().
Signed-off-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, when a socket receives something on the error queue it only wakes up
the socket on select if it is in the "read" list, that is the socket has
something to read. It is useful also to wake the socket if it is in the error
list, which would enable software to wait on error queue packets without waking
up for regular data on the socket. The main use case is for receiving
timestamped transmit packets which return the timestamp to the socket via the
error queue. This enables an application to select on the socket for the error
queue only instead of for the regular traffic.

-v2-
* Added the SO_SELECT_ERR_QUEUE socket option to every architechture specific file
* Modified every socket poll function that checks error queue
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Cc: Jeffrey Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Matthew Vick <matthew.vick@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The cgroup code has been surrounded by ifdef CONFIG_NET_CLS_CGROUP
and CONFIG_NETPRIO_CGROUP.
Signed-off-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The mem cgroup socket limit is only used if the config option is
enabled. Found with sparse
Signed-off-by: NStephen Hemminger <stephen@networkplumber.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

proc_net_remove is only used to remove proc entries
that under /proc/net,it's not a general function for
removing proc entries of netns. if we want to remove
some proc entries which under /proc/net/stat/, we still
need to call remove_proc_entry.

this patch use remove_proc_entry to replace proc_net_remove.
we can remove proc_net_remove after this patch.
Signed-off-by: NGao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Right now, some modules such as bonding use proc_create
to create proc entries under /proc/net/, and other modules
such as ipv4 use proc_net_fops_create.

It looks a little chaos.this patch changes all of
proc_net_fops_create to proc_create. we can remove
proc_net_fops_create after this patch.
Signed-off-by: NGao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As in del_timer() there has already placed a timer_pending() function
to check whether the timer to be deleted is pending or not, it's
unnecessary to check timer pending state again before del_timer() is
called.
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Definitions and macros for implementing soreusport.
Signed-off-by: NTom Herbert <therbert@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While a privileged program can open a raw socket, attach some
restrictive filter and drop its privileges (or send the socket to an
unprivileged program through some Unix socket), the filter can still
be removed or modified by the unprivileged program. This commit adds a
socket option to lock the filter (SO_LOCK_FILTER) preventing any
modification of a socket filter program.

This is similar to OpenBSD BIOCLOCK ioctl on bpf sockets, except even
root is not allowed change/drop the filter.

The state of the lock can be read with getsockopt(). No error is
triggered if the state is not changed. -EPERM is returned when a user
tries to remove the lock or to change/remove the filter while the lock
is active. The check is done directly in sk_attach_filter() and
sk_detach_filter() and does not affect only setsockopt() syscall.
Signed-off-by: NVincent Bernat <bernat@luffy.cx>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Using a seqlock for devnet_rename_seq is not a good idea,
as device_rename() can sleep.

As we hold RTNL, we dont need a protection for writers,
and only need a seqcount so that readers can catch a change done
by a writer.

Bug added in commit c91f6df2 (sockopt: Change getsockopt() of
SO_BINDTODEVICE to return an interface name)
Reported-by: NDave Jones <davej@redhat.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Brian Haley <brian.haley@hp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of having the getsockopt() of SO_BINDTODEVICE return an index, which
will then require another call like if_indextoname() to get the actual interface
name, have it return the name directly.

This also matches the existing man page description on socket(7) which mentions
the argument being an interface name.

If the value has not been set, zero is returned and optlen will be set to zero
to indicate there is no interface name present.

Added a seqlock to protect this code path, and dev_ifname(), from someone
changing the device name via dev_change_name().

v2: Added seqlock protection while copying device name.

v3: Fixed word wrap in patch.
Signed-off-by: NBrian Haley <brian.haley@hp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Allow an unpriviled user who has created a user namespace, and then
created a network namespace to effectively use the new network
namespace, by reducing capable(CAP_NET_ADMIN) and
capable(CAP_NET_RAW) calls to be ns_capable(net->user_ns,
CAP_NET_ADMIN), or capable(net->user_ns, CAP_NET_RAW) calls.

Settings that merely control a single network device are allowed.
Either the network device is a logical network device where
restrictions make no difference or the network device is hardware NIC
that has been explicity moved from the initial network namespace.

In general policy and network stack state changes are allowed
while resource control is left unchanged.

Allow ethtool ioctls.

Allow binding to network devices.
Allow setting the socket mark.
Allow setting the socket priority.

Allow setting the network device alias via sysfs.
Allow setting the mtu via sysfs.
Allow changing the network device flags via sysfs.
Allow setting the network device group via sysfs.

Allow the following network device ioctls.
SIOCGMIIPHY
SIOCGMIIREG
SIOCSIFNAME
SIOCSIFFLAGS
SIOCSIFMETRIC
SIOCSIFMTU
SIOCSIFHWADDR
SIOCSIFSLAVE
SIOCADDMULTI
SIOCDELMULTI
SIOCSIFHWBROADCAST
SIOCSMIIREG
SIOCBONDENSLAVE
SIOCBONDRELEASE
SIOCBONDSETHWADDR
SIOCBONDCHANGEACTIVE
SIOCBRADDIF
SIOCBRDELIF
SIOCSHWTSTAMP
Signed-off-by: N"Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The SO_ATTACH_FILTER option is set only. I propose to add the get
ability by using SO_ATTACH_FILTER in getsockopt. To be less
irritating to eyes the SO_GET_FILTER alias to it is declared. This
ability is required by checkpoint-restore project to be able to
save full state of a socket.

There are two issues with getting filter back.

First, kernel modifies the sock_filter->code on filter load, thus in
order to return the filter element back to user we have to decode it
into user-visible constants. Fortunately the modification in question
is interconvertible.

Second, the BPF_S_ALU_DIV_K code modifies the command argument k to
speed up the run-time division by doing kernel_k = reciprocal(user_k).
Bad news is that different user_k may result in same kernel_k, so we
can't get the original user_k back. Good news is that we don't have
to do it. What we need to is calculate a user2_k so, that

  reciprocal(user2_k) == reciprocal(user_k) == kernel_k

i.e. if it's re-loaded back the compiled again value will be exactly
the same as it was. That said, the user2_k can be calculated like this

  user2_k = reciprocal(kernel_k)

with an exception, that if kernel_k == 0, then user2_k == 1.

The optlen argument is treated like this -- when zero, kernel returns
the amount of sock_fprog elements in filter, otherwise it should be
large enough for the sock_fprog array.

changes since v1:
* Declared SO_GET_FILTER in all arch headers
* Added decode of vlan-tag codes
Signed-off-by: NPavel Emelyanov <xemul@parallels.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

sock_update_classid() assumes that the update operation always are
applied on the current task. sock_update_classid() needs to know on
which tasks to work on in order to be able to migrate task between
cgroups using the struct cgroup_subsys attach() callback.
Signed-off-by: NDaniel Wagner <daniel.wagner@bmw-carit.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Joe Perches <joe@perches.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: <netdev@vger.kernel.org>
Cc: <cgroups@vger.kernel.org>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As Eric pointed out:
"Hey task_cls_classid() has its own rcu protection since commit
3fb5a991 (cls_cgroup: Fix rcu lockdep warning)

So we can safely revert Paul commit (1144182a)
(We no longer need rcu_read_lock/unlock here)"
Signed-off-by: NDaniel Wagner <daniel.wagner@bmw-carit.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: netdev@vger.kernel.org
Cc: cgroups@vger.kernel.org
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The SO_BINDTODEVICE option is the only SOL_SOCKET one that can be set, but
cannot be get via sockopt API. The only way we can find the device id a
socket is bound to is via sock-diag interface. But the diag works only on
hashed sockets, while the opt in question can be set for yet unhashed one.

That said, in order to know what device a socket is bound to (we do want
to know this in checkpoint-restore project) I propose to make this option
getsockopt-able and report the respective device index.

Another solution to the problem might be to teach the sock-diag reporting
info on unhashed sockets. Should I go this way instead?
Signed-off-by: NPavel Emelyanov <xemul@parallels.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It seems sk_init() has no value today and even does strange things :

# grep . /proc/sys/net/core/?mem_*
/proc/sys/net/core/rmem_default:212992
/proc/sys/net/core/rmem_max:131071
/proc/sys/net/core/wmem_default:212992
/proc/sys/net/core/wmem_max:131071

We can remove it completely.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reviewed-by: NShan Wei <davidshan@tencent.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Its possible to use RAW sockets to get a crash in
tcp_set_keepalive() / sk_reset_timer()

Fix is to make sure socket is a SOCK_STREAM one.
Reported-by: NDave Jones <davej@redhat.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.

This page is used to build fragments for skbs.

Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)

But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page

Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.

This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.

(up to 32768 bytes per frag, thats order-3 pages on x86)

This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.

Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536

Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NVijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>