Zefir Kurtisi reported kernel panic with an openwrt specific patch.
However, it turns out that mainline has a similar bug waiting to happen.

Once NF_HOOK() returns the skb is in undefined state and must not be
used.   Moreover, the okfn must consume the skb to support async
processing (NF_QUEUE).

Current okfn in this spot doesn't consume it and caller assumes that
NF_HOOK return value tells us if skb was freed or not, but thats wrong.

It "works" because no in-tree user registers a NFPROTO_BRIDGE hook at
LOCAL_IN that returns STOLEN or NF_QUEUE verdicts.

Once we add NF_QUEUE support for nftables bridge this will break --
NF_QUEUE holds the skb for async processing, caller will erronoulsy
return RX_HANDLER_PASS and on reinject netfilter will access free'd skb.

Fix this by pushing skb up the stack in the okfn instead.

NB: It also seems dubious to use LOCAL_IN while bypassing PRE_ROUTING
completely in this case but this is how its been forever so it seems
preferable to not change this.

Cc: Felix Fietkau <nbd@openwrt.org>
Cc: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Tested-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next



Johan Hedberg says:

====================
pull request: bluetooth-next 2016-03-12

Here's the last bluetooth-next pull request for the 4.6 kernel.

 - New USB ID for AR3012 in btusb
 - New BCM2E55 ACPI ID
 - Buffer overflow fix for the Add Advertising command
 - Support for a new Bluetooth LE limited privacy mode
 - Fix for firmware activation in btmrvl_sdio
 - Cleanups to mac802154 & 6lowpan code

Please let me know if there are any issues pulling. Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Andrew Lunn says:

====================
DSA cleanup and fixes

The RFC patchset for re-architecturing DSA probing contains a few
standalone patches, either clean up or fixes. This pulls them out for
submission.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

The fixed phys delete function simply removed the fixed phy from the
internal linked list and freed the memory. It however did not
unregister the associated phy device. This meant it was still possible
to find the phy device on the mdio bus.

Make fixed_phy_del() an internal function and add a
fixed_phy_unregister() to unregisters the phy device and then uses
fixed_phy_del() to free resources.

Modify DSA to use this new API function, so we don't leak phys.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

All ports types can have a fixed PHY associated with it. Remove the
check which limits removal to only CPU and DSA ports.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

The phy is disconnected from the slave in dsa_slave_destroy(). Don't
destroy fixed link phys until after this, since there can be fixed
linked phys connected to ports.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

When the phy is disconnected, the parent pointer to the netdev it was
attached to is set to NULL. The code then tries to suspend the phy,
but dsa_slave_fixed_link_update needs the parent pointer to determine
which switch the phy is connected to. So it dereferenced a NULL
pointer. Check for this condition.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

All the drivers support multiple chips, but mv88e6123_61_65 is the
only one that reflects this in its naming. Change it to be consistent
with the other drivers.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Sergei Shtylyov says:

====================
of_mdio: use IS_ERR_OR_NULL() and PTR_ERR_OR_ZERO()

   Here's the set of 3 patches against DaveM's 'net-next.git' repo. They deal
with some error checks in the device tree MDIO code...
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

PTR_ERR_OR_ZERO() is open coded in of_phy_register_fixed_link(), so just
call it directly...

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

IS_ERR_OR_NULL() is open coded in of_mdiobus_register_phy(), so just call
it directly...

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mdio_device_create() never returns NULL, thus checking for it in
of_mdiobus_register_device() is pointless...

Suggested-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

David Daney says:

====================
net/phy: Improvements to Cavium Thunder MDIO code.

Changes from v1:

 - In 1/3 Add back check for non-OF objects in bgx_init_of_phy().  It
   is probably not necessary, but better safe than sorry...

The firmware on many Cavium Thunder systems configures the MDIO bus
hardware to be probed as a PCI device.  In order to use the MDIO bus
drivers in this configuration, we must add PCI probing to the driver.

There are two parts to this set of three patches:

 1) Cleanup the PHY probing code in thunder_bgx.c to handle the case
    where there is no PHY attached to a port, as well as being more
    robust in the face of driver loading order by use of
    -EPROBE_DEFER.

 2) Split mdio-octeon.c into two drivers, one with platform probing,
 and the other with PCI probing.  Common code is shared between the
 two.

Tested on several different Thunder and OCTEON systems, also compile
tested on x86_64.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

The Cavium Thunder SoCs have multiple MIDO buses that are part of a
single PCI device.  To model this in the device tree we call the PCI
parent device a "cavium,thunder-8890-mdio-nexus", it has several
children, one for each MDIO bus.

The MDIO bus hardware is identical to that found in the OCTEON SoCs,
so we use that code for things that are not part of the PCI driver
probe/remove

Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

A follow-on patch uses PCI probing to find the Thunder MDIO hardware.
In preparation for this, split out the common code into a new file
mdio-cavium.c, which will be used by both the existing OCTEON driver,
and the new Thunder PCI based driver.

As part of the refactoring simplify the struct cavium_mdiobus by
removing fields that are only ever used in the probe function and can
just as well be local variables.

Use readq/writeq in preference to readq_relaxed/writeq_relaxed as the
relaxed form was an optimization for an early chip revision, and the
MDIO drivers are not performance bottlenecks that need optimization in
the first place.

Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Remove the call to force the octeon-mdio driver to be loaded.  Allow
the standard driver loading mechanisms to load the PHY drivers, and
use -EPROBE_DEFER to cause the BGX driver to be probed only after the
PHY drivers are available.

Reorder the setting of MAC addresses and PHY probing to allow BGX
LMACs with no attached PHY to still be assigned a MAC address.

Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The mvneta_percpu_notifier() hotplug callback lacks handling of the
CPU_DOWN_FAILED case. That means, if CPU_DOWN_PREPARE failes, the
driver is not well configured on the CPU.

Add handling for CPU_DOWN_FAILED[_FROZEN] hotplug notifier transition
to setup the driver.

Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: netdev@vger.kernel.org
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix a bug introduced in e06a03bd

 (fsl/fman: fix the pause_time test)
When pause_time is set to '0' - pause frames are disabled and
there's no need to apply dTSEC-A003 Errata workaround.

Signed-off-by: Igal Liberman <igal.liberman@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Some new development in PHYLIB added new function pointers to the struct
phy_driver, document these.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Per RFC4898, they count segments sent/received
containing a positive length data segment (that includes
retransmission segments carrying data).  Unlike
tcpi_segs_out/in, tcpi_data_segs_out/in excludes segments
carrying no data (e.g. pure ack).

The patch also updates the segs_in in tcp_fastopen_add_skb()
so that segs_in >= data_segs_in property is kept.

Together with retransmission data, tcpi_data_segs_out
gives a better signal on the rxmit rate.

v6: Rebase on the latest net-next

v5: Eric pointed out that checking skb->len is still needed in
tcp_fastopen_add_skb() because skb can carry a FIN without data.
Hence, instead of open coding segs_in and data_segs_in, tcp_segs_in()
helper is used.  Comment is added to the fastopen case to explain why
segs_in has to be reset and tcp_segs_in() has to be called before
__skb_pull().

v4: Add comment to the changes in tcp_fastopen_add_skb()
and also add remark on this case in the commit message.

v3: Add const modifier to the skb parameter in tcp_segs_in()

v2: Rework based on recent fix by Eric:
commit a9d99ce2

 ("tcp: fix tcpi_segs_in after connection establishment")

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Chris Rapier <rapier@psc.edu>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Marcelo Ricardo Leitner <mleitner@redhat.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

A recent bug fix rearranged the code in vmxnet3_tq_xmit() in a
way that left the error handling for oversized headers unlock
a lock that had not been taken yet. Gcc warns about the incorrect
use of the 'flags' variable because of that:

drivers/net/vmxnet3/vmxnet3_drv.c: In function 'vmxnet3_tq_xmit.constprop':
include/linux/spinlock.h:246:3: error: 'flags' may be used uninitialized in this function [-Werror=maybe-uninitialized]

This changes the error handling path to 'goto' the end of the function
beyond the lock/unlock pair.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: cec05562

 ("vmxnet3: avoid calling pskb_may_pull with interrupts disabled")
Signed-off-by: David S. Miller <davem@davemloft.net>

Arnd Bergmann says:

====================
net: gcc-6.0 warning fixes

I've just installed gcc-6.0 to see what kinds of new warnings
we get. It turns out that it's actually really useful once I
disabled -Wunused-const-variable, and all of the warnings it
found in network drivers seem valid.

Sorry for the bad timing in the merge window, but I figured
it would be better to send the fixes as I found the bugs
rather than waiting for the next cycle. The first three
look appropriate for stable backports.

The other two only fix a gcc warning about incorrect whitespace,
probably not worth backporting those.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

gcc points out code that is not indented the way it is
interpreted:

net/caif/cfpkt_skbuff.c: In function 'cfpkt_setlen':
net/caif/cfpkt_skbuff.c:289:4: error: statement is indented as if it were guarded by... [-Werror=misleading-indentation]
    return cfpkt_getlen(pkt);
    ^~~~~~
net/caif/cfpkt_skbuff.c:286:3: note: ...this 'else' clause, but it is not
   else
   ^~~~

It is clear from the context that not returning here would be
a bug, as we'd end up passing a negative length into a function
that takes a u16 length, so it is not missing curly braces
here, and I'm assuming that the indentation is the only part
that's wrong about it.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

A cleanup patch in linux-3.18 moved around some code in the ath9k
driver and left some code to be indented in a misleading way,
made worse by the addition of some new code for p2p mode, as
discovered by a new gcc-6 warning:

drivers/net/wireless/ath/ath9k/init.c: In function 'ath9k_set_hw_capab':
drivers/net/wireless/ath/ath9k/init.c:851:4: warning: statement is indented as if it were guarded by... [-Wmisleading-indentation]
    hw->wiphy->iface_combinations = if_comb;
    ^~
drivers/net/wireless/ath/ath9k/init.c:847:3: note: ...this 'if' clause, but it is not
   if (ath9k_is_chanctx_enabled())
   ^~

The code is in fact correct, but the indentation is not, so I'm
reformatting it as it should have been after the original cleanup.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 499afacc ("ath9k: Isolate ath9k_use_chanctx module parameter")
Fixes: eb61f9f6

 ("ath9k: advertise p2p dev support when chanctx")
Signed-off-by: David S. Miller <davem@davemloft.net>

Code that was added back in 2.6.38 has an obvious overflow
when accessing a static array, and at the time it was added
only a code comment was put in front of it as a reminder
to have it reviewed properly.

This has not happened, but gcc-6 now points to the specific
overflow:

drivers/net/wireless/ath/ath9k/eeprom.c: In function 'ath9k_hw_get_gain_boundaries_pdadcs':
drivers/net/wireless/ath/ath9k/eeprom.c:483:44: error: array subscript is above array bounds [-Werror=array-bounds]
     maxPwrT4[i] = data_9287[idxL].pwrPdg[i][4];
                   ~~~~~~~~~~~~~~~~~~~~~~~~~^~~

It turns out that the correct array length exists in the local
'intercepts' variable of this function, so we can just use that
instead of hardcoding '4', so this patch changes all three
instances to use that variable. The other two instances were
already correct, but it's more consistent this way.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 940cd2c1

 ("ath9k_hw: merge the ar9287 version of ath9k_hw_get_gain_boundaries_pdadcs")
Signed-off-by: David S. Miller <davem@davemloft.net>

gcc-6 finds an out of bounds access in the fst_add_one function
when calculating the end of the mmio area:

drivers/net/wan/farsync.c: In function 'fst_add_one':
drivers/net/wan/farsync.c:418:53: error: index 2 denotes an offset greater than size of 'u8[2][8192] {aka unsigned char[2][8192]}' [-Werror=array-bounds]
 #define BUF_OFFSET(X)   (BFM_BASE + offsetof(struct buf_window, X))
                                                     ^
include/linux/compiler-gcc.h:158:21: note: in definition of macro '__compiler_offsetof'
  __builtin_offsetof(a, b)
                     ^
drivers/net/wan/farsync.c:418:37: note: in expansion of macro 'offsetof'
 #define BUF_OFFSET(X)   (BFM_BASE + offsetof(struct buf_window, X))
                                     ^~~~~~~~
drivers/net/wan/farsync.c:2519:36: note: in expansion of macro 'BUF_OFFSET'
                                  + BUF_OFFSET ( txBuffer[i][NUM_TX_BUFFER][0]);
                                    ^~~~~~~~~~

The warning is correct, but not critical because this appears
to be a write-only variable that is set by each WAN driver but
never accessed afterwards.

I'm taking the minimal fix here, using the correct pointer by
pointing 'mem_end' to the last byte inside of the register area
as all other WAN drivers do, rather than the first byte outside of
it. An alternative would be to just remove the mem_end member
entirely.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

The implementation of QP paravirtualization back in linux-3.7 included
some code that looks very dubious, and gcc-6 has grown smart enough
to warn about it:

drivers/net/ethernet/mellanox/mlx4/resource_tracker.c: In function 'verify_qp_parameters':
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:3154:5: error: statement is indented as if it were guarded by... [-Werror=misleading-indentation]
     if (optpar & MLX4_QP_OPTPAR_ALT_ADDR_PATH) {
     ^~
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c:3144:4: note: ...this 'if' clause, but it is not
    if (slave != mlx4_master_func_num(dev))

>From looking at the context, I'm reasonably sure that the indentation
is correct but that it should have contained curly braces from the
start, as the update_gid() function in the same patch correctly does.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 54679e14

 ("mlx4: Implement QP paravirtualization and maintain phys_pkey_cache for smp_snoop")
Signed-off-by: David S. Miller <davem@davemloft.net>

The device_reset() function may fail, so we have to check
its return value, e.g. to make deferred probing work correctly.
gcc warns about it because of the warn_unused_result attribute:

drivers/net/ethernet/mediatek/mtk_eth_soc.c: In function 'mtk_probe':
drivers/net/ethernet/mediatek/mtk_eth_soc.c:1679:2: error: ignoring return value of 'device_reset', declared with attribute warn_unused_result [-Werror=unused-result]

This adds the trivial error check to propagate the return value
to the generic platform device probe code.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Device drivers should not mess with the DMA mask directly,
but instead call dma_set_mask() etc if needed.

In case of the mtk_eth_soc driver, the mask already gets set
correctly when the device is created, and setting it again
is against the documented API.

This removes the incorrect setting.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

dma_alloc_coherent() expects a dma_addr_t pointer as its argument,
not an 'unsigned int', and gcc correctly warns about broken
code in the mtk_init_fq_dma function:

drivers/net/ethernet/mediatek/mtk_eth_soc.c: In function 'mtk_init_fq_dma':
drivers/net/ethernet/mediatek/mtk_eth_soc.c:463:13: error: passing argument 3 of 'dma_alloc_coherent' from incompatible pointer type [-Werror=incompatible-pointer-types]

This changes the type of the local variable to dma_addr_t.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

The syzkaller fuzzer hit the following use-after-free:

  Call Trace:
   [<ffffffff8175ea0e>] __asan_report_load8_noabort+0x3e/0x40 mm/kasan/report.c:295
   [<ffffffff851cc31a>] __sys_recvmmsg+0x6fa/0x7f0 net/socket.c:2261
   [<     inline     >] SYSC_recvmmsg net/socket.c:2281
   [<ffffffff851cc57f>] SyS_recvmmsg+0x16f/0x180 net/socket.c:2270
   [<ffffffff86332bb6>] entry_SYSCALL_64_fastpath+0x16/0x7a
  arch/x86/entry/entry_64.S:185

And, as Dmitry rightly assessed, that is because we can drop the
reference and then touch it when the underlying recvmsg calls return
some packets and then hit an error, which will make recvmmsg to set
sock->sk->sk_err, oops, fix it.

Reported-and-Tested-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Sasha Levin <sasha.levin@oracle.com>
Fixes: a2e27255 ("net: Introduce recvmmsg socket syscall")
http://lkml.kernel.org/r/20160122211644.GC2470@redhat.com


Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Sunil Goutham says:

====================
net: thunderx: Performance enhancement changes

Below patches attempts to improve performance by reducing
no of atomic operations while allocating new receive buffers
and reducing cache misses by adjusting nicvf structure elements.

Changes from v1:
 No changes, resubmitting a fresh as per David's suggestion.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Adjusted nicvf structure such that all elements used in hot
path like napi, xmit e.t.c fall into same cache line. This reduced
no of cache misses and resulted in ~2% increase in no of packets
handled on a core.

Also modified elements with :1 notation to boolean, to be
consistent with other element definitions.

Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Instead of calling get_page() for every receive buffer carved out
of page, set page's usage count at the end, to reduce no of atomic
calls.

Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Expand headroom further in order to be able to fit the larger IPv6
header. Prior to this patch this caused a skb under panic for certain
tipc packets when using IPv6 UDP bearer(s).

Signed-off-by: Richard Alpe <richard.alpe@ericsson.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Gregory CLEMENT says:

====================
API set for HW Buffer management

This is the sixth version of the API set for HW Buffer management (that was
initially submitted here:
http://thread.gmane.org/gmane.linux.kernel/2125152

).

This version is just a rebasing onto the last net-next. I also added
the Tested-by flag from Sebastian Careba : "The patch set applies
successfully and it works well, no more Samba issues any longer".

For the record in the previous versions I made the following changes:
v4 -> v5:
- Add a field with the size of the buffer of the pool was added. It
  then allow to fix some misused size in the mvneta_bm code when using
  the new framework.

- Add a new patch from Marcin for sram allowing to require
  non-bufferable access to the memory. It was needed for the hardware
  buffer management of the mvneta.

- Fix the build issue notified by the 0-day builder when building the
  drivers as module.

v3 -> v4
- Fix build issue when HWBM is not selected

v2 -> v3
- Make a HWBM and a SWBM version of the mvneta_rx() function in order
  to reduce the the conditional code. Kept a condition inside the
  mvneta_poll because specializing this function would have means
  duplicating 95% of the code.

- Put back the register_netdev() call at the end of the mvneta_probe()
  function. In order to have a unique ID for each port, just used a
  global variable in the driver.

- Added a fix from Marcin in the "net: mvneta: bm: add support for
  hardware buffer management" patch: "when dropping packets, only
  buffer pointers passed from BM to descriptors have to be returned to
  the pool. In submitted version after closing the port and
  mvneta_rxq_deinit(), it was very likely that a lot of fake buffers
  are added to the pool, because all descriptors took part in
  iteration."

- Removed the select MVNETA_BM from the Kconfig, it will let the user
  the choice to use not use it if they want.

v1 -> v2
- The hardware buffer management helpers are no more built by default
  and now depend on a hidden config symbol which has to be selected
  by the driver if needed
- The hwbm_pool_refill() and hwbm_pool_add() now receive a gfp_t as
  argument allowing the caller to specify the flag it needs.
- buf_num is now tested to ensure there is no wrapping
- A spinlock has been added to protect the hwbm_pool_add() function in
  SMP or irq context.
- used pr_warn instead of pr_debug in case of errors.
- fixed the mvneta implementation by returning the buffer to the pool
  at various place instead of ignoring it.
- Squashed "bus: mvenus-mbus: Fix size test for
   mvebu_mbus_get_dram_win_info" into bus: mvebu-mbus: provide api for
   obtaining IO and DRAM window information.
- Added my signed-otf-by on all the patches as submitter of the series.
- Renamed the dts patches with the pattern "ARM: dts: platform:"
- Removed the patch "ARM: mvebu: enable SRAM support in
  mvebu_v7_defconfig" of this series and already applied it
- Modified the order of the patches.

In order to ease the test the branch mvneta-BM-framework-v6 is
available at git@github.com:MISL-EBU-System-SW/mainline-public.git.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Now that the hardware buffer management framework had been introduced,
let's use it.

Tested-by: Sebastian Careba <nitroshift@yahoo.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This basic implementation allows to share code between driver using
hardware buffer management. As the code is hardware agnostic, there is
few helpers, most of the optimization brought by the an HW BM has to be
done at driver level.

Tested-by: Sebastian Careba <nitroshift@yahoo.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Buffer manager (BM) is a dedicated hardware unit that can be used by all
ethernet ports of Armada XP and 38x SoC's. It allows to offload CPU on RX
path by sparing DRAM access on refilling buffer pool, hardware-based
filling of descriptor ring data and better memory utilization due to HW
arbitration for using 'short' pools for small packets.

Tests performed with A388 SoC working as a network bridge between two
packet generators showed increase of maximum processed 64B packets by
~20k (~555k packets with BM enabled vs ~535 packets without BM). Also
when pushing 1500B-packets with a line rate achieved, CPU load decreased
from around 25% without BM to 20% with BM.

BM comprise up to 4 buffer pointers' (BP) rings kept in DRAM, which
are called external BP pools - BPPE. Allocating and releasing buffer
pointers (BP) to/from BPPE is performed indirectly by write/read access
to a dedicated internal SRAM, where internal BP pools (BPPI) are placed.
BM hardware controls status of BPPE automatically, as well as assigning
proper buffers to RX descriptors. For more details please refer to
Functional Specification of Armada XP or 38x SoC.

In order to enable support for a separate hardware block, common for all
ports, a new driver has to be implemented ('mvneta_bm'). It provides
initialization sequence of address space, clocks, registers, SRAM,
empty pools' structures and also obtaining optional configuration
from DT (please refer to device tree binding documentation). mvneta_bm
exposes also a necessary API to mvneta driver, as well as a dedicated
structure with BM information (bm_priv), whose presence is used as a
flag notifying of BM usage by port. It has to be ensured that mvneta_bm
probe is executed prior to the ones in ports' driver. In case BM is not
used or its probe fails, mvneta falls back to use software buffer
management.

A sequence executed in mvneta_probe function is modified in order to have
an access to needed resources before possible port's BM initialization is
done. According to port-pools mapping provided by DT appropriate registers
are configured and the buffer pools are filled. RX path is modified
accordingly. Becaues the hardware allows a wide variety of configuration
options, following assumptions are made:
* using BM mechanisms can be selectively disabled/enabled basing
  on DT configuration among the ports
* 'long' pool's single buffer size is tied to port's MTU
* using 'long' pool by port is obligatory and it cannot be shared
* using 'short' pool for smaller packets is optional
* one 'short' pool can be shared among all ports

This commit enables hardware buffer management operation cooperating with
existing mvneta driver. New device tree binding documentation is added and
the one of mvneta is updated accordingly.

[gregory.clement@free-electrons.com: removed the suspend/resume part]

Signed-off-by: Marcin Wojtas <mw@semihalf.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This commit enables finding appropriate mbus window and obtaining its
target id and attribute for given physical address in two separate
routines, both for IO and DRAM windows. This functionality
is needed for Armada XP/38x Network Controller's Buffer Manager and
PnC configuration.

[gregory.clement@free-electrons.com: Fix size test for
mvebu_mbus_get_dram_win_info]

Signed-off-by: Marcin Wojtas <mw@semihalf.com>
[DRAM window information reference in LKv3.10]
Signed-off-by: Evan Wang <xswang@marvell.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>