As there are now no remaining users of arch_fast_hash(), lets kill
it entirely.

This basically reverts commit 71ae8aac ("lib: introduce arch
optimized hash library") and follow-up work, that is f.e., commit
23721754 ("lib: hash: follow-up fixups for arch hash"),
commit e3fec2f7 ("lib: Add missing arch generic-y entries for
asm-generic/hash.h") and last but not least commit 6a02652d
("perf tools: Fix include for non x86 architectures").

Cc: Francesco Fusco <fusco@ntop.org>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch effectively reverts commit 500f8087 ("net: ovs: use CRC32
accelerated flow hash if available"), and other remaining arch_fast_hash()
users such as from nfsd via commit 6282cd56 ("NFSD: Don't hand out
delegations for 30 seconds after recalling them.") where it has been used
as a hash function for bloom filtering.

While we think that these users are actually not much of concern, it has
been requested to remove the arch_fast_hash() library bits that arose
from [1] entirely as per recent discussion [2]. The main argument is that
using it as a hash may introduce bias due to its linearity (see avalanche
criterion) and thus makes it less clear (though we tried to document that)
when this security/performance trade-off is actually acceptable for a
general purpose library function.

Lets therefore avoid any further confusion on this matter and remove it to
prevent any future accidental misuse of it. For the time being, this is
going to make hashing of flow keys a bit more expensive in the ovs case,
but future work could reevaluate a different hashing discipline.

  [1] https://patchwork.ozlabs.org/patch/299369/
  [2] https://patchwork.ozlabs.org/patch/418756/

Cc: Neil Brown <neilb@suse.de>
Cc: Francesco Fusco <fusco@ntop.org>
Cc: Jesse Gross <jesse@nicira.com>
Cc: Thomas Graf <tgraf@suug.ch>
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For netlink, we shouldn't be using arch_fast_hash() as a hashing
discipline, but rather jhash() instead.

Since netlink sockets can be opened by any user, a local attacker
would be able to easily create collisions with the DPDK-derived
arch_fast_hash(), which trades off performance for security by
using crc32 CPU instructions on x86_64.

While it might have a legimite use case in other places, it should
be avoided in netlink context, though. As rhashtable's API is very
flexible, we could later on still decide on other hashing disciplines,
if legitimate.

Reference: http://thread.gmane.org/gmane.linux.kernel/1844123
Fixes: e341694e ("netlink: Convert netlink_lookup() to use RCU protected hash table")
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Acked-by: NThomas Graf <tgraf@suug.ch>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Tilman Schmidt says:

====================
ISDN patches for net-next

Here's a series of patches for the Gigaset ISDN driver and one for
the ISDN CAPI subsystem.  Please merge as appropriate.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Utility function command_2_index is always called with arguments of
type u8. Adapt its declaration accordingly.
Signed-off-by: NTilman Schmidt <tilman@imap.cc>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Kernel CAPI has been the recommended ISDN subsystem for the Gigaset
driver since kernel release 2.6.34.2. It provides full backwards
compatibility to the old I4L subsystem thanks to the capidrv module.
I4L has been marked as deprecated for more than seven years.
Signed-off-by: NTilman Schmidt <tilman@imap.cc>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

No need to pass a member of the cardstate structure as a separate
argument if the entire structure is already passed.
Signed-off-by: NTilman Schmidt <tilman@imap.cc>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace the flag-controlled retry loop by explicit goto statements
in the error branches to make the control structure easier to
understand.
Signed-off-by: NTilman Schmidt <tilman@imap.cc>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Function gigaset_skb_sent was declared twice, identically, in gigaset.h.
Signed-off-by: NTilman Schmidt <tilman@imap.cc>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit 908344cd ("tipc: fix bug in multicast congestion handling")
introduced a race in the broadcast link wakeup functionality.

This patch eliminates this broadcast link wakeup race caused by
operation on the wakeup list without proper locking. If this race
hit and corrupted the list all subsequent wakeup messages would be
lost, resulting in a considerable memory leak.
Signed-off-by: NRichard Alpe <richard.alpe@ericsson.com>
Signed-off-by: NErik Hugne <erik.hugne@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for setting/getting rss hash key using ethtool.

v2:
respin patch to support RSS hash function changes.
Signed-off-by: NGovindarajulu Varadarajan <_govind@gmx.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Alexander Duyck says:

====================
net: Alloc NAPI page frags from their own pool

This patch series implements a means of allocating page fragments without
the need for the local_irq_save/restore in __netdev_alloc_frag.  By doing
this I am able to decrease packet processing time by 11ns per packet in my
test environment.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch replaces the calls to netdev_alloc_skb_ip_align in the
copybreak paths.

Cc: Gary Zambrano <zambrano@broadcom.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Ariel Elior <ariel.elior@qlogic.com>
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This replaces most of the calls to netdev_alloc_skb_ip_align in the Realtek
drivers.  The one instance I didn't replace in 8139cp.c is because it was
called as a part of init and as such is not always accessed from the
softirq context.

Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In order to use napi_alloc_skb I needed to pass a pointer to struct adapter
instead of struct pci_dev.  This allowed me to access &adapter->napi.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This change replaces calls to netdev_alloc_skb_ip_align with
napi_alloc_skb.  The advantage of napi_alloc_skb is currently the fact that
the page allocation doesn't make use of any irq disable calls.

There are few spots where I couldn't replace the calls as the buffer
allocation routine is called as a part of init which is outside of the
softirq context.

Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This change pulls the core functionality out of __netdev_alloc_skb and
places them in a new function named __alloc_rx_skb.  The reason for doing
this is to make these bits accessible to a new function __napi_alloc_skb.
In addition __alloc_rx_skb now has a new flags value that is used to
determine which page frag pool to allocate from.  If the SKB_ALLOC_NAPI
flag is set then the NAPI pool is used.  The advantage of this is that we
do not have to use local_irq_save/restore when accessing the NAPI pool from
NAPI context.

In my test setup I saw at least 11ns of savings using the napi_alloc_skb
function versus the netdev_alloc_skb function, most of this being due to
the fact that we didn't have to call local_irq_save/restore.

The main use case for napi_alloc_skb would be for things such as copybreak
or page fragment based receive paths where an skb is allocated after the
data has been received instead of before.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch splits the netdev_alloc_frag function up so that it can be used
on one of two page frag pools instead of being fixed on the
netdev_alloc_cache.  By doing this we can add a NAPI specific function
__napi_alloc_frag that accesses a pool that is only used from softirq
context.  The advantage to this is that we do not need to call
local_irq_save/restore which can be a significant savings.

I also took the opportunity to refactor the core bits that were placed in
__alloc_page_frag.  First I updated the allocation to do either a 32K
allocation or an order 0 page.  This is based on the changes in commmit
d9b2938a where it was found that latencies could be reduced in case of
failures.  Then I also rewrote the logic to work from the end of the page to
the start.  By doing this the size value doesn't have to be used unless we
have run out of space for page fragments.  Finally I cleaned up the atomic
bits so that we just do an atomic_sub_and_test and if that returns true then
we set the page->_count via an atomic_set.  This way we can remove the extra
conditional for the atomic_read since it would have led to an atomic_inc in
the case of success anyway.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Acked-by: NAlexei Starovoitov <ast@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

More iov_iter work for the networking from Al Viro.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use MODULE_VERSION() now that dummy driver has a version.
Signed-off-by: NFlavio Leitner <fbl@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To cancel nesting, this function is more convenient.
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The disable_irq_nosync function, not the disable_irq function, must be
used to disable the DMA channel interrupt from within the interrupt
service routine. Change the disable_irq call to disable_irq_nosync.
Signed-off-by: NTom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

On i.MX6SX sabreauto board, when there have no phy daughter board connection,
there have kernel crash by NULL pointer:

fec 2188000.ethernet eth0: could not attach to PHY
Unable to handle kernel NULL pointer dereference at virtual address 00000220
pgd = 80004000
[00000220] *pgd=00000000
Internal error: Oops: 5 [#1] PREEMPT SMP ARM
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.14.24-01042-g27eaeea-dirty #405
task: d8078000 ti: d8076000 task.ti: d8076000
PC is at mutex_lock+0x10/0x54
LR is at phy_start+0x14/0x68
pc : [<806ad4e4>]    lr : [<803b0f90>]    psr: 60000113
sp : d8077d80  ip : 00000000  fp : d83cc000
r10: 0000100c  r9 : d83cc800  r8 : 00000000
r7 : d83bcd0c  r6 : 00000200  r5 : 00000220  r4 : 00000220
r3 : 00000000  r2 : 00000000  r1 : d83bcd90  r0 : 00000220
Flags: nZCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment kernel
Control: 10c5387d  Table: 8000404a  DAC: 00000015
Process swapper/0 (pid: 1, stack limit = 0xd8076240)
Stack: (0xd8077d80 to 0xd8078000)
7d80: 00000000 803b0f90 00000001 00000000 d83bc800 803be034 00000007 805c3fb4
7da0: 00000003 80d4e0bc 805efcb8 fffffff1 fffffff0 00000000 00000000 d8077dfc
7dc0: 0000000d 80d6ce80 80d126b0 800499c8 d83bc800 d83bc800 806f0f40 d83bc82c
7de0: 00000000 00000000 80d6ce80 80d126b0 0000016b 80540250 d8076008 d83bc800
7e00: 0000016b d83bc800 00001003 00000001 00001002 805404d4 d83bc800 00000120
7e20: 00001002 00001002 00000000 805405d4 d83bc800 00000001 80d126c0 00001002
7e40: 80dbc5dc 80d02024 00000000 806ae360 00000002 d6128420 d6127198 12400000
7e60: 00000000 00000000 00000002 d61271e8 00000000 12400000 d801674c 800e49f0
7e80: d6127198 d6124e58 00000000 80238848 d61271c4 00000000 00000001 d8016700
7ea0: 80dd2e00 80d752c0 80d752c0 80cfdaec 0000010c 80239430 806c2e90 d800f080
7ec0: d800f380 804e46b4 ffffffbc 80d15cb0 00000007 80d752c0 80d752c0 80d01e94
7ee0: 0000010c d8076030 00000000 800088cc 80dbaba4 80bd411c d80a6f00 806b1e04
7f00: 00000000 00000000 00000000 80125b84 00000000 80d2c56c 60000113 00000001
7f20: ef7ff9df 806c80cc 0000010c 80043f5c 80c95eb8 00000007 ef7ffa1d 00000007
7f40: 80d2c55c 80d15cb0 00000007 80d752c0 80d752c0 80ccc50c 0000010c 80d0a114
7f60: 80d0a10c 80cccc04 00000007 00000007 80ccc50c 806ae410 00000000 8004cb84
7f80: 80d17bc0 00000000 806a4bd4 00000000 00000000 00000000 00000000 00000000
7fa0: 00000000 806a4bdc 00000000 8000e5f8 00000000 00000000 00000000 00000000
7fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
7fe0: 00000000 00000000 00000000 00000000 00000013 00000000 1e79a7bb e5337f77
[<806ad4e4>] (mutex_lock) from [<803b0f90>] (phy_start+0x14/0x68)
[<803b0f90>] (phy_start) from [<803be034>] (fec_enet_open+0x448/0x5dc)
[<803be034>] (fec_enet_open) from [<80540250>] (__dev_open+0xa8/0x110)
[<80540250>] (__dev_open) from [<805404d4>] (__dev_change_flags+0x88/0x170)
[<805404d4>] (__dev_change_flags) from [<805405d4>] (dev_change_flags+0x18/0x48)
[<805405d4>] (dev_change_flags) from [<80d02024>] (ip_auto_config+0x190/0xf94)
[<80d02024>] (ip_auto_config) from [<800088cc>] (do_one_initcall+0xe8/0x144)
[<800088cc>] (do_one_initcall) from [<80cccc04>] (kernel_init_freeable+0x104/0x1c8)
[<80cccc04>] (kernel_init_freeable) from [<806a4bdc>] (kernel_init+0x8/0xec)
[<806a4bdc>] (kernel_init) from [<8000e5f8>] (ret_from_fork+0x14/0x3c)
Code: e92d4010 e3a03000 e1a04000 ee073fba (e1903f9f)

Add phydev check to fix the issue.
Signed-off-by: NFugang Duan <B38611@freescale.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The commit fb9962f3 ("tipc: ensure all name sequences are properly
protected with its lock") involves below errors:

net/tipc/name_table.c:980 tipc_purge_publications() error: double lock 'spin_lock:&seq->lock'
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Hook a ndo_set_mac_address callback, update the internal Ethernet MAC in
the netdevice structure, and finally write that address down to the
UniMAC registers. If the interface is down, and most likely clock gated,
we do not update the registers but just the local copy, such that next
ndo_open() call will effectively write down the address.
Signed-off-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Roopa Prabhu says:

====================
remove bridge mode BRIDGE_MODE_SWDEV

BRIDGE_MODE_SWDEV was introduced to indicate switchdev offloads
for bridging from user space (In other words to call into the hw switch
port driver directly). But user can use existing BRIDGE_FLAGS_SELF
to call into the hw switch port driver today. swdev mode is not required
anymore. So, this patch removes it.

v4 - v5
    incorporate comments
    - Define BRIDGE_MODE_UNDEF to handle cases where mode is not defined
    - reverse the order of patches
    - include patch comments in all patches
====================
Signed-off-by: NRoopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch removes bridge mode swdev.
Users can use BRIDGE_FLAGS_SELF to indicate swdev offload
if needed.
Signed-off-by: NRoopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: NScott Feldman <sfeldma@gmail.com>
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove use of 'swdev' mode in rocker. rocker dev offloads
can use the BRIDGE_FLAGS_SELF to indicate offload to hardware.
Signed-off-by: NRoopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: NScott Feldman <sfeldma@gmail.com>
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds mode BRIDGE_MODE_UNDEF for cases where mode is not needed.
Signed-off-by: NRoopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: NScott Feldman <sfeldma@gmail.com>
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Merge tag 'master-2014-12-08' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
pull request: wireless-next 2014-12-08

Please pull this last batch of pending wireless updates for the 3.19 tree...

For the wireless bits, Johannes says:

"This time I have Felix's no-status rate control work, which will allow
drivers to work better with rate control even if they don't have perfect
status reporting. In addition to this, a small hwsim fix from Patrik,
one of the regulatory patches from Arik, and a number of cleanups and
fixes I did myself.

Of note is a patch where I disable CFG80211_WEXT so that compatibility
is no longer selectable - this is intended as a wake-up call for anyone
who's still using it, and is still easily worked around (it's a one-line
patch) before we fully remove the code as well in the future."

For the Bluetooth bits, Johan says:

"Here's one more bluetooth-next pull request for 3.19:

 - Minor cleanups for ieee802154 & mac802154
 - Fix for the kernel warning with !TASK_RUNNING reported by Kirill A.
   Shutemov
 - Support for another ath3k device
 - Fix for tracking link key based security level
 - Device tree bindings for btmrvl + a state update fix
 - Fix for wrong ACL flags on LE links"

And...

"In addition to the previous one this contains two more cleanups to
mac802154 as well as support for some new HCI features from the
Bluetooth 4.2 specification.

From the original request:

'Here's what should be the last bluetooth-next pull request for 3.19.
It's rather large but the majority of it is the Low Energy Secure
Connections feature that's part of the Bluetooth 4.2 specification. The
specification went public only this week so we couldn't publish the
corresponding code before that. The code itself can nevertheless be
considered fairly mature as it's been in development for over 6 months
and gone through several interoperability test events.

Besides LE SC the pull request contains an important fix for command
complete events for mgmt sockets which also fixes some leaks of hci_conn
objects when powering off or unplugging Bluetooth adapters.

A smaller feature that's part of the pull request is service discovery
support. This is like normal device discovery except that devices not
matching specific UUIDs or strong enough RSSI are filtered out.

Other changes that the pull request contains are firmware dump support
to the btmrvl driver, firmware download support for Broadcom BCM20702A0
variants, as well as some coding style cleanups in 6lowpan &
ieee802154/mac802154 code.'"

For the NFC bits, Samuel says:

"With this one we get:

- NFC digital improvements for DEP support: Chaining, NACK and ATN
  support added.

- NCI improvements: Support for p2p target, SE IO operand addition,
  SE operands extensions to support proprietary implementations, and
  a few fixes.

- NFC HCI improvements: OPEN_PIPE and NOTIFY_ALL_CLEARED support,
  and SE IO operand addition.

- A bunch of minor improvements and fixes for STMicro st21nfcb and
  st21nfca"

For the iwlwifi bits, Emmanuel says:

"Major works are CSA and TDLS. On top of that I have a new
firmware API for scan and a few rate control improvements.
Johannes find a few tricks to improve our CPU utilization
and adds support for a new spin of 7265 called 7265D.
Along with this a few random things that don't stand out."

And...

"I deprecate here -8.ucode since -9 has been published long ago.
Along with that I have a new activity, we have now better
a infrastructure for firmware debugging. This will allow to
have configurable probes insides the firmware.
Luca continues his work on NetDetect, this feature is now
complete. All the rest is minor fixes here and there."

For the Atheros bits, Kalle says:

"Only ath10k changes this time and no major changes. Most visible are:

o new debugfs interface for runtime firmware debugging (Yanbo)

o fix shared WEP (Sujith)

o don't rebuild whenever kernel version changes (Johannes)

o lots of refactoring to make it easier to add new hw support (Michal)

There's also smaller fixes and improvements with no point of listing
here."

In addition, there are a few last minute updates to ath5k,
ath9k, brcmfmac, brcmsmac, mwifiex, rt2x00, rtlwifi, and wil6210.
Also included is a pull of the wireless tree to pick-up the fixes
originally included in "pull request: wireless 2014-12-03"...

Please let me know if there are problems!
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

PTR_ALIGN macro after skb_reserve is redundant, because skb_reserve
function adjusts the alignment of skb->data.
Signed-off-by: NMitsuhiro Kimura <mitsuhiro.kimura.kc@renesas.com>
Signed-off-by: NYoshihiro Kaneko <ykaneko0929@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Both of 'boguscnt' and 'quota' have nearly meaning as the condition of
the reception loop.
In order to cut down redundant processing, this patch changes excess
judgement.
Signed-off-by: NMitsuhiro Kimura <mitsuhiro.kimura.kc@renesas.com>
Signed-off-by: NYoshihiro Kaneko <ykaneko0929@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

the queue length of sd->input_pkt_queue has been put into qlen,
and impossible to change, since hold the lock
Signed-off-by: NLi RongQing <roy.qing.li@gmail.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2014-12-09

This series contains updates to i40e and i40evf.

Jeff (me) provides a single patch to convert a macro to a static inline
function based on feedback from Joe Perches on a previous patch.

Shannon provides the remaining twelve patches against i40e.  Almost all
of Shannon's patches cleanup/fix NVM issues varying in range from
adding more detail to debug messages, to removing dead code, to fixing
NVM state transitions after an error.  Change the handy decoder interface
for admin queue return code to help catch and properly report the condition
as a useful errno rather than returning a misleading '0'.  Added a range
check to avoid any possible array index-out-of-bound issues.

v2:
 - fixed up patch 05 in the series to use the ARRAY_SIZE() macro as suggested
   by Sergei Shtylyov
 - fix up patch 13 to remove unnecessary parens in the return statement
   as suggested by Sergei Shtylyov
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Marc Kleine-Budde says:

====================
pull-request: can-next 2014-12-07

this is a pull request of 8 patches for net-next/master.

Andri Yngvason contributes 4 patches in which the CAN state change
handling is consolidated and unified among the sja1000, mscan and
flexcan driver. The three patches by Jeremiah Mahler fix spelling
mistakes and eliminate the banner[] variable in various parts. And a
patch by me that switches on sparse endianess checking by default.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 95bd09eb ("tcp: TSO packets automatic sizing") tried to
control TSO size, but did this at the wrong place (sendmsg() time)

At sendmsg() time, we might have a pessimistic view of flow rate,
and we end up building very small skbs (with 2 MSS per skb).

This is bad because :

 - It sends small TSO packets even in Slow Start where rate quickly
   increases.
 - It tends to make socket write queue very big, increasing tcp_ack()
   processing time, but also increasing memory needs, not necessarily
   accounted for, as fast clones overhead is currently ignored.
 - Lower GRO efficiency and more ACK packets.

Servers with a lot of small lived connections suffer from this.

Lets instead fill skbs as much as possible (64KB of payload), but split
them at xmit time, when we have a precise idea of the flow rate.
skb split is actually quite efficient.

Patch looks bigger than necessary, because TCP Small Queue decision now
has to take place after the eventual split.

As Neal suggested, introduce a new tcp_tso_autosize() helper, so that
tcp_tso_should_defer() can be synchronized on same goal.

Rename tp->xmit_size_goal_segs to tp->gso_segs, as this variable
contains number of mss that we can put in GSO packet, and is not
related to the autosizing goal anymore.

Tested:

40 ms rtt link

nstat >/dev/null
netperf -H remote -l -2000000 -- -s 1000000
nstat | egrep "IpInReceives|IpOutRequests|TcpOutSegs|IpExtOutOctets"

Before patch :

Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/s

 87380 2000000 2000000    0.36         44.22
IpInReceives                    600                0.0
IpOutRequests                   599                0.0
TcpOutSegs                      1397               0.0
IpExtOutOctets                  2033249            0.0

After patch :

Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

 87380 2000000 2000000    0.36       44.27
IpInReceives                    221                0.0
IpOutRequests                   232                0.0
TcpOutSegs                      1397               0.0
IpExtOutOctets                  2013953            0.0
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Acked-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

no users left
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

no callers other than itself.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

... making both non-draining.  That means that tcp_recvmsg() becomes
non-draining.  And _that_ would break iscsit_do_rx_data() unless we
	a) make sure tcp_recvmsg() is uniformly non-draining (it is)
	b) make sure it copes with arbitrary (including shifted)
iov_iter (it does, all it uses is iov_iter primitives)
	c) make iscsit_do_rx_data() initialize ->msg_iter only once.

Fortunately, (c) is doable with minimal work and we are rid of one
the two places where kernel send/recvmsg users would be unhappy with
non-draining behaviour.

Actually, that makes all but one of ->recvmsg() instances iov_iter-clean.
The exception is skcipher_recvmsg() and it also isn't hard to convert
to primitives (iov_iter_get_pages() is needed there).  That'll wait
a bit - there's some interplay with ->sendmsg() path for that one.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>