Disabling the adapter after each transfer adds additional delays
for each I2C transfer. Even if we don't wait for it to be disabled
anymore, on next transfer we will need to if we have several transfers
in a row.

Now during the transfer init we check if IC_TAR can be changed
dynamically, the status register for no activity and TX buffer being
empty. In this case we don't need to disable it

When a transfer fails the adapter will still be disabled - this is a
conservative approach. When transfers succeed, the adapter is left
enabled and it's configured so to disable interrupts.

Alternating register reads on 2 slaves:
perf stat -r4 chrt -f 10 ./i2c-test /dev/i2c-1 25000 0x40 0x6 0x1e 0x00

Before:
	8.638705161 seconds time elapsed                  ( +-  5.90% )
After:
	7.516821591 seconds time elapsed                  ( +-  0.11% )
Signed-off-by: NLucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: NJosé Roberto de Souza <jose.souza@intel.com>
Acked-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Tested-by: NChristian Ruppert <christian.ruppert@alitech.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

This adapter can be synthesized with dynamic tar update enabled or disabled.
When enabled it is not necessary to disable the adapter to change the slave
address in some situations, which saves some time per transaction.

There is no direct register to know if this feature is enabled but we can do it
indirectly by writing to the 10BIT_ADDR field in IC_CON: this field is
read only when dynamic tar update is enabled.
Signed-off-by: NLucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: NJosé Roberto de Souza <jose.souza@intel.com>
Acked-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Tested-by: NChristian Ruppert <christian.ruppert@alitech.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

These are used in 2 places and will be needed in more.
Signed-off-by: NLucas De Marchi <lucas.demarchi@intel.com>
Acked-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Tested-by: NChristian Ruppert <christian.ruppert@alitech.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

If we aren't going to continue using the controller we can just disable
it instead of waiting for it to complete. The biggest improvement here
is when a I2C transaction is completed and it doesn't block until
the adapter is disabled. When a new transfer is needed we will disable
and wait for its completion.

This way the adapter will continue changing its state in parallel to the
execution of the thread that requested the I2C transaction saving most
of the time 25~250 usec per I2C transaction.

A simple program doing a register read (1 byte write, 1 byte read)
alternating on 2 different slaves repeated 25k times for each and
measurements taken 4 times we get:

perf stat -r4 chrt -f 10 ./i2c-test /dev/i2c-1 25000 0x40 0x6 0x1e 0x00

Before:
	30.879317977 seconds time elapsed                 ( +- 14.83% )
After:
	8.638705161 seconds time elapsed                  ( +-  5.90% )
Signed-off-by: NJosé Roberto de Souza <jose.souza@intel.com>
Signed-off-by: NLucas De Marchi <lucas.demarchi@intel.com>
Acked-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Tested-by: NChristian Ruppert <christian.ruppert@alitech.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

Fast mode is the default speed of i2c-designware which can be overridden
by platform data or by "clock-frequency" device property. Even though
the ACPI 5.1 can pass device properties via _DSD method, shipping systems
define the connection speed between I2C host and each slave in their
I2cSerialBus resources. Which means speed is not defined per bus but per
slave.

As there is now support in i2c-core to find the bus speed from ACPI use
that to set up the bus speed prior registering the I2C adapter.
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

ACPI 5 specification doesn't have property for the I2C bus speed but
I2cSerialBus resource descriptor which define each controller-slave
connection define the maximum speed supported by that connection.

Thus finding the maximum safe speed for the bus is to walk through all
I2cSerialBus resources that are associated to I2C controller and use the
speed of slowest connection.

Add function i2c_acpi_find_bus_speed() to the i2c-core that adapter
drivers can call prior registering itself to core.

This implies two-step walk through the I2cSerialBus resources: call to
i2c_acpi_find_bus_speed() does the first scan and finds the safe bus
speed that adapter drivers can set up. Adapter driver registration does
the second scan when i2c-core creates the I2C slaves by calling the
i2c_acpi_register_devices(). In that way the bus speed is set in case
slave device probe gets called during registration and does communication.

Previous version commit 55d38d06 ("i2c: core: Add function for finding
the bus speed from ACPI") got reverted due merge conflicts from
commit 525e6fab ("i2c / ACPI: add support for ACPI reconfigure
notifications").

This version is a bit bigger than previous version but is still sharing
the lowest and complicated part of I2cSerialBus lookup routines with the
existing code.
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

I2C ACPI enumeration was originally implemented in another module under
drivers/acpi/ but was later moved into i2c-core with added support for
I2C ACPI operation region.

Rename these acpi_i2c_ prefixed functions, structures and defines in
i2c-core to i2c_acpi_ in order to have more consistent name space.

This is updated version from commit a7003b65 ("i2c: core: Cleanup I2C
ACPI namespace") that got reverted due merge conflicts from
commit 525e6fab ("i2c / ACPI: add support for ACPI reconfigure
notifications").
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

This patch enabled high speed mode. High speed mode can be turn on by
setting the clk_freq to 3400000. High speed HCNT and LCNT are needed
as there is no default value provided.
Signed-off-by: NWeifeng Voon <weifeng.voon@intel.com>
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

DW_IC_CON_MASTER, DW_IC_CON_SLAVE_DISABLE and DW_IC_CON_RESTART_EN are
common config that need to be set for i2c designware master. So, configure
it first without having to repeat inside the if else.
Signed-off-by: NWeifeng Voon <weifeng.voon@intel.com>
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

This patch enabled fast mode plus. The fast mode plus and fast speed
share the same HCNT and LCNT register. So, the fast mode plus will only
run when the HCNT and LCNT value is provided. Else, it will run at fast
speed as default.
Signed-off-by: NWeifeng Voon <weifeng.voon@intel.com>
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

I2C designware controller can run at fast mode plus and high speed. This
patch adds the capability to get the HCNT, LCNT configuration via
FPCN (fast plus) and HSCN (high speed) ACPI method.
Signed-off-by: NWeifeng Voon <weifeng.voon@intel.com>
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

I2c designware controller operate speed is configured in the register
IC_CON. Previously the operate speed is determined by a local variable
clk_freq. This patch will move the local variable clk_freq into struct
dw_i2c_dev. This change will ease the set and get of the clk_freq.
Signed-off-by: NWeifeng Voon <weifeng.voon@intel.com>
Signed-off-by: NJarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: NAndy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

Initialize booleon values with true instead of 1.
Signed-off-by: NJan Glauber <jglauber@cavium.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

Sort include files alphabetically to reduce probability of merge
conflicts.
Signed-off-by: NJan Glauber <jglauber@cavium.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

The register offsets are different between Octeon and ThunderX so move
them into the algorithm struct and get rid of the define.
Signed-off-by: NJan Glauber <jglauber@cavium.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

Add SMBUS alert interrupt support. For now only device tree is
supported for specifying the alert. In case of ACPI an error
is returned.
Signed-off-by: NJan Glauber <jglauber@cavium.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

The ThunderX SOC uses the same i2c block as the Octeon SOC.
The main difference is that on ThunderX the device is a PCI device
so device probing is done via PCI, interrupts are MSI-X. The
clock rates can be set via device tree or ACPI.
Signed-off-by: NJan Glauber <jglauber@cavium.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

Move common functionality into a separate file in preparation of the
re-use from the ThunderX i2c driver.

Functions are slightly re-ordered but no other changes are included.
Signed-off-by: NJan Glauber <jglauber@cavium.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

This is an intermediate commit in preparation of the driver split.
The module rename in this commit will be reverted in the next patch,
this is just done to make the series bisectible.
Signed-off-by: NJan Glauber <jglauber@cavium.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

No longer rely on the implicit matching with the i2c device name, use
an explicit compatible string instead.

Keep a direct pointer to the chip description instead of an index into
the chip description array.
Reviewed-by: NWolfram Sang <wsa@the-dreams.de>
Signed-off-by: NPeter Rosin <peda@axentia.se>

No longer rely on the implicit matching with the i2c device name, use
an explicit compatible string instead.
Reviewed-by: NWolfram Sang <wsa@the-dreams.de>
Signed-off-by: NPeter Rosin <peda@axentia.se>

The i2c mux core can then take appropriate action depending on if it is
used for an actual i2c mux, for a gate or for an arbitrator (the last
is the case for these drivers). This adds support for the new clearer
and more compact devicetree bindings that was added recently.
Reviewed-by: NWolfram Sang <wsa@the-dreams.de>
Signed-off-by: NPeter Rosin <peda@axentia.se>

Backwards compatibility is preserved; the subnodes are in practice
optional.

However, the mux core needs to know what subnode it should examine, so add
a couple of new flags for i2c_mux_alloc for this purpose.

The rule is that if the mux core finds a 'reg' property in the appropriate
subnode, e.g. if 'reg' exists in the 'i2c-mux' subnode, then the mux core
will assume that this is an old style entry and not an i2c-mux subnode
(correspondingly for arbitrators and gates with 'i2c-arb' and 'i2c-gate').
Reviewed-by: NWolfram Sang <wsa@the-dreams.de>
Signed-off-by: NPeter Rosin <peda@axentia.se>

This unifies usage with i2c_lock_bus and i2c_unlock_bus, and paves the
way for the next patch which looks a bit saner with this preparatory
work taken care of beforehand.
Signed-off-by: NPeter Rosin <peda@axentia.se>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

The at24 driver doesn't check if the chip is functional in its probe
function. This leads to instantiating devices that are not physically
present. For example the cape EEPROMs for BeagleBone Black are defined
in the device tree at four addresses on i2c2, but normally only one of
them is present.

If the userspace doesn't know the location in advance, it will need to
check if reading the nvmem attributes fails to determine which EEPROM
is actually there.

Try to read a single byte in probe() and bail-out with -ENODEV if the
read fails.
Signed-off-by: NBartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

The core will do this for us now.
Signed-off-by: NWolfram Sang <wsa-dev@sang-engineering.com>
Acked-by: NMasahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: NGuenter Roeck <linux@roeck-us.net>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

The core will do this for us now.
Signed-off-by: NWolfram Sang <wsa-dev@sang-engineering.com>
Reviewed-by: NGrygorii Strashko <grygorii.strashko@ti.com>
Acked-by: NPeter Korsgaard <peter@korsgaard.com>
Acked-by: NHeiko Stuebner <heiko@sntech.de>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Acked-by: NThierry Reding <treding@nvidia.com>
Acked-by: NRay Jui <ray.jui@broadcom.com>
Acked-by: NVladimir Zapolskiy <vz@mleia.com>
Acked-by: NLudovic Desroches <ludovic.desroches@atmel.com>
Acked-by: NKrzysztof Kozlowski <k.kozlowski@samsung.com>
Acked-by: NLaxman Dewangan <ldewangan@nvidia.com>
Signed-off-by: NWolfram Sang <wsa@the-dreams.de>

This is an entirely new driver instead of yet another set of patches
to sb_edac.c because:

1) Mapping from PCI devices to socket/memory controller is significantly
   different. Skylake scatters devices on a socket across a number of
   PCI buses.
2) There is an extra level of interleaving via the "mcroute" register
   that would be a little messy to squeeze into the old driver.
3) Validation is getting too expensive. Changes to sb_edac need to
   be checked against Sandy Bridge, Ivy Bridge, Haswell, Broadwell and
   Knights Landing.
Acked-by: NAristeu Rozanski <aris@redhat.com>
Acked-by: NBorislav Petkov <bp@suse.de>
Signed-off-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This reverts commit:

  fa7d81bb ("drm/fb-helper: Reduce READ_ONCE(master) to lockless_dereference")

As Peter explained:

  [...] lockless_dereference() is _stronger_ than READ_ONCE(), not weaker.

  [...]

  Also, clue is in the name: 'dereference', you don't actually dereference
  the pointer here, only load it.

My next patch breaks the compile without this revert, because it assumes
you want to deference and thus also need the struct type visible (which
it isn't here), so revert it.
Tested-by: NPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1470909022-687-1-git-send-email-johannes@sipsolutions.netSigned-off-by: NIngo Molnar <mingo@kernel.org>

As pointed out by Jamal, an action could be shared by
multiple filters, so we can't use list to chain them
any more after we get rid of the original tc_action.
Instead, we could just save pointers to these actions
in tcf_exts, since they are refcount'ed, so convert
the list to an array of pointers.

The "ugly" part is the action API still accepts list
as a parameter, I just introduce a helper function to
convert the array of pointers to a list, instead of
relying on the C99 feature to iterate the array.

Fixes: a85a970a ("net_sched: move tc_action into tcf_common")
Reported-by: NJamal Hadi Salim <jhs@mojatatu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: NCong Wang <xiyou.wangcong@gmail.com>
Acked-by: NJamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When packets enter the device they are classified to a priority group
(PG) buffer based on their PCP value. After their egress port and
traffic class are determined they are moved to the switch's shared
buffer and await transmission, if:

(Ingress{Port}.Usage < Thres && Ingress{Port,PG}.Usage < Thres &&
 Egress{Port}.Usage < Thres && Egress{Port,TC}.Usage < Thres)
||
(Ingress{Port}.Usage < Min || Ingress{Port,PG} < Min ||
 Egress{Port}.Usage < Min || Egress{Port,TC}.Usage < Min)

Packets scheduled to transmission through CPU port (trapped to CPU) use
traffic class 7, which has a zero maximum and minimum quotas. However,
when such packets arrive from PG 0 they are admitted to the shared
buffer as PG 0 has a non-zero minimum quota.

Allow all packets to be trapped to the CPU - regardless of the PG they
were classified to - by assigning a 10KB minimum quota for CPU port and
TC7.

Fixes: 8e8dfe9f ("mlxsw: spectrum: Add IEEE 802.1Qaz ETS support")
Reported-by: NTamir Winetroub <tamirw@mellanox.com>
Tested-by: NTamir Winetroub <tamirw@mellanox.com>
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Before destroying the 802.1Q FID we should first remove the VID-to-FID
mapping. This makes mlxsw_sp_fid_destroy() symmetric with regards to
mlxsw_sp_fid_create().

Fixes: 14d39461 ("mlxsw: spectrum: Use per-FID struct for the VLAN-aware bridge")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While going over the code I noticed we are missing two rollbacks in the
port's creation error path. Add them and adjust the place of one of them
in the port's removal sequence so that both are symmetric.

Fixes: 56ade8fe ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Ralue pack function needs to set op, otherwise it is 0 for add always.

Fixes: d5a1c749 ("mlxsw: reg: Add Router Algorithmic LPM Unicast Entry Register definition")
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

One of the conditions to generate an ICMP Redirect Message is that "the
packet is being forwarded out the same physical interface that it was
received from" (RFC 1812).

Therefore, we need to be able to trap such packets and let the kernel
decide what to do with them.

For each RIF, enable the loop-back filter, which will raise the LBERROR
trap whenever the ingress RIF equals the egress RIF.

Fixes: 99724c18 ("mlxsw: spectrum: Introduce support for router interfaces")
Reported-by: NIlan Tayari <ilant@mellanox.com>
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add the following traps:

1) MTU Error: Trap packets whose size is bigger than the egress RIF's
MTU. If DF bit isn't set, traffic will continue to be routed in slow
path.

2) TTL Error: Trap packets whose TTL expired. This allows traceroute to
work properly.

3) OSPF packets.

Fixes: 7b27ce7b ("mlxsw: spectrum: Add traps needed for router implementation")
Signed-off-by: NElad Raz <eladr@mellanox.com>
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit bbf2a475 ("mlxsw: spectrum: Initialize ports at the end of
init sequence") moved ports initialization to the end of the init
sequence, which means ports are the first to be removed during fini.

Since the FDB delayed work is still active when ports are removed it's
possible for it to process FDB notifications of inactive ports,
resulting in a warning message.

Fix that by marking ports as inactive only after unregistering them. The
NETDEV_UNREGISTER event will invoke bridge's driver port removal
sequence that will cause the FDB (and FDB notifications) to be flushed.

Fixes: bbf2a475 ("mlxsw: spectrum: Initialize ports at the end of init sequence")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After registering a netdevice it's possible for user space applications
to configure an IP address on it. From the driver's perspective, this
means a router interface (RIF) should be created for the PVID vPort.

Therefore, we must create the PVID vPort before registering the
netdevice.

Fixes: 99724c18 ("mlxsw: spectrum: Introduce support for router interfaces")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, when device configuration fails we emit errors to the kernel
log despite the fact we already get these from the EMAD transaction
layer, so remove them.

In addition to being unnecessary, removing these error messages will
allow us to reuse mlxsw_sp_port_add_vid() to create the PVID vPort
before registering the netdevice.

Fixes: 99724c18 ("mlxsw: spectrum: Introduce support for router interfaces")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When removing a VLAN filter from the device we shouldn't return upon the
first error we encounter, as otherwise we'll have resources that will
never be freed nor used.

Instead, we should keep trying to free as much resources as possible in
a best effort mode.

Remove the error message as well, since we already get these from the
EMAD transaction code.

Fixes: 99724c18 ("mlxsw: spectrum: Introduce support for router interfaces")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>