linux-2.6-microblaze.git
5 years agoMerge branch 'net-phy-aquantia-improve-and-extend-driver'
David S. Miller [Sat, 23 Feb 2019 22:12:10 +0000 (14:12 -0800)]
Merge branch 'net-phy-aquantia-improve-and-extend-driver'

Heiner Kallweit says:

====================
net: phy: aquantia: improve and extend driver

This series improves and extends the Aquantia PHY driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: aquantia: use genphy_c45_read_status
Heiner Kallweit [Fri, 22 Feb 2019 22:52:32 +0000 (23:52 +0100)]
net: phy: aquantia: use genphy_c45_read_status

Use new function genphy_c45_read_status(). 1000BaseT link partner
advertisement needs to be read from vendor registers.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: add genphy_c45_read_status
Heiner Kallweit [Fri, 22 Feb 2019 22:51:44 +0000 (23:51 +0100)]
net: phy: add genphy_c45_read_status

Similar to genphy_read_status() for Clause 22 add a generic read_status
function for Clause 45.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: don't change modes we don't care about in genphy_c45_read_lpa
Heiner Kallweit [Fri, 22 Feb 2019 22:50:49 +0000 (23:50 +0100)]
net: phy: don't change modes we don't care about in genphy_c45_read_lpa

Because 1000BaseT isn't covered by Clause 45, the 1000BaseT flags in
phydev->lp_advertising may have been set based on vendor registers
already. genphy_c45_read_lpa() would clear these flags as of today.
Therefore switch to mii_lpa_mod_linkmode_lpa_t.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: aquantia: add support for auto-negotiation configuration
Andrew Lunn [Fri, 22 Feb 2019 22:49:54 +0000 (23:49 +0100)]
net: phy: aquantia: add support for auto-negotiation configuration

Make use of the generic c45 code, plus code specific to the Aquantia
phy for 1000BaseT negotiation.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: aquantia: remove false 5G and 10G speed ability for AQCS109
Heiner Kallweit [Fri, 22 Feb 2019 22:48:14 +0000 (23:48 +0100)]
net: phy: aquantia: remove false 5G and 10G speed ability for AQCS109

AQCS109 belongs to a family of PHY's where certain members don't
support 5G or 10G. However for all members of the family the chip
reports 10G and 5G capability. Therefore remove the not supported
modes for AQCS109.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mlx5-updates-2019-02-21' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Sat, 23 Feb 2019 21:56:25 +0000 (13:56 -0800)]
Merge tag 'mlx5-updates-2019-02-21' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2019-02-21

This series adds some misc updates to mlx5 driver,

1) Eli Britstein, Introduces tunnel entropy control from PCMR register
and fixes GRE key by controlling port tunnel entropy calculation.

2) Eran Ben Elisha, provides some mlx5 fixes to the latest tx devlink health
reporting mechanism.

3) Huy Nguyen, Added the support for ndo bridge_setlink to allow
   VEPA/VEB E-Switch legacy mode configurations.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mlxsw-Add-support-for-new-port-types-and-speeds-for-Spectrum-2'
David S. Miller [Sat, 23 Feb 2019 21:54:36 +0000 (13:54 -0800)]
Merge branch 'mlxsw-Add-support-for-new-port-types-and-speeds-for-Spectrum-2'

Ido Schimmel says:

====================
mlxsw: Add support for new port types and speeds for Spectrum-2

Shalom says:

This patchset adds support for new port types and speeds for Spectrum-2.

Patch #1 + #2 removes an unsupported PTYS field and a duplicate link
mode entry.

Patch #3 queries port's connector type from firmware instead of deriving
it from port admin state.

Patch #4 renames functions which relate to port type-speed to be
Spectrum-1 specific.

Patch #5 defines port type-speed operations and applies it for
Spectrum-1.

Patch #6 + #7 are small renaming and cosmetic changes.

Patch #8 adds new port type-speed fields for PTYS register. These new
fields extend the existing ones in order to support more types and
speeds.

Patch #9 adds Spectrum-2 support for port type-speed operations.

Patch #10 adds Spectrum-2 new port types and speeds.

For Spectrum-2, the user must configure all the types per speed if he /
she wants a specific speed to be advertised. For example, if the user
wants to advertise 100Gbps 4-lanes speed, the following ethtool bits
should be advertised:

  Supported ethtool bits for 100Gbps 4-lanes:
      0x1000000000      100000baseKR4 Full
      0x2000000000      100000baseSR4 Full
      0x4000000000      100000baseCR4 Full
      0x8000000000      100000baseLR4_ER4 Full

  Command for advertising 100Gbps 4-lanes:
      ethtool -s enp3s0np1 advertise 0xF000000000
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Add Spectrum-2 ASIC support for new port types and speeds
Shalom Toledo [Fri, 22 Feb 2019 13:56:46 +0000 (13:56 +0000)]
mlxsw: spectrum: Add Spectrum-2 ASIC support for new port types and speeds

Add Spectrum-2 ASIC support for the following new port types and speeds:
  * 50Gbps 1-lane
  * 100Gbps 2-lanes
  * 200Gbps 4-lanes

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Add Spectrum-2 ASIC port type-speed operations
Shalom Toledo [Fri, 22 Feb 2019 13:56:45 +0000 (13:56 +0000)]
mlxsw: spectrum: Add Spectrum-2 ASIC port type-speed operations

Add Spectrum-2 ASIC port type-speed operations.

Since multiple ethtool link modes are represented using a single bit in the
ASIC, the driver forces the user to configure all types per a specific
speed. For example, if the user wants to advertise 100Gbps 4-lanes speed,
he should advertise all the types of 100Gbps 4-lanes speed that are
supported by the ASIC as shown below:

  Supported ethtool bits for 100Gbps 4-lanes:
      0x1000000000      100000baseKR4 Full
      0x2000000000      100000baseSR4 Full
      0x4000000000      100000baseCR4 Full
      0x8000000000      100000baseLR4_ER4 Full

  Command for advertising 100Gbps 4-lanes:
      ethtool -s enp3s0np1 advertise 0xF000000000

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: reg: Add new port type-speed fields for PTYS register
Shalom Toledo [Fri, 22 Feb 2019 13:56:44 +0000 (13:56 +0000)]
mlxsw: reg: Add new port type-speed fields for PTYS register

PTYS register introduces a new layout for port type-speed fields. These
fields extend the existing ones in order to handle more types and speeds.
For example, the new 200Gbps speed.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: reg: 80 columns wrapping change
Shalom Toledo [Fri, 22 Feb 2019 13:56:42 +0000 (13:56 +0000)]
mlxsw: reg: 80 columns wrapping change

80 columns wrapping change in mlxsw_reg_ptys_eth_unpack function.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: reg: Rename p_eth_proto_adm to full name p_eth_proto_admin
Shalom Toledo [Fri, 22 Feb 2019 13:56:41 +0000 (13:56 +0000)]
mlxsw: reg: Rename p_eth_proto_adm to full name p_eth_proto_admin

Rename p_eth_proto_adm to p_eth_proto_admin in mlxsw_reg_ptys_eth_unpack
function.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Add port type-speed operations
Shalom Toledo [Fri, 22 Feb 2019 13:56:40 +0000 (13:56 +0000)]
mlxsw: spectrum: Add port type-speed operations

Add port type-speed operations in order to have different operations for
different ASICs. For now, both ASICs use the same pointer.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Rename port type-speed functions to ASIC specific
Shalom Toledo [Fri, 22 Feb 2019 13:56:39 +0000 (13:56 +0000)]
mlxsw: spectrum: Rename port type-speed functions to ASIC specific

Rename port speed-type functions to be Spectrum-1 ASIC specific.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Query port connector type from firmware
Shalom Toledo [Fri, 22 Feb 2019 13:56:38 +0000 (13:56 +0000)]
mlxsw: spectrum: Query port connector type from firmware

Instead of deriving the port connector type from port admin state, query it
from firmware.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Remove unsupported eth_proto_lp_advertise field in PTYS
Shalom Toledo [Fri, 22 Feb 2019 13:56:37 +0000 (13:56 +0000)]
mlxsw: spectrum: Remove unsupported eth_proto_lp_advertise field in PTYS

Remove eth_proto_lp_advertise field in PTYS register since it is not
supported by the firmware.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Remove duplicate port link mode entry
Shalom Toledo [Fri, 22 Feb 2019 13:56:36 +0000 (13:56 +0000)]
mlxsw: spectrum: Remove duplicate port link mode entry

Remove duplicate port link mode entry from mlxsw_sp_port_link_mode.

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agokcm: Remove unnecessary SLAB_PANIC for kmem_cache_create() in kcm_init
YueHaibing [Fri, 22 Feb 2019 06:15:30 +0000 (14:15 +0800)]
kcm: Remove unnecessary SLAB_PANIC for kmem_cache_create() in kcm_init

There has check NULL on kmem_cache_create on failure in kcm_init,
no need use SLAB_PANIC to panic the system.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-Wformat-fixes'
David S. Miller [Sat, 23 Feb 2019 21:44:58 +0000 (13:44 -0800)]
Merge branch 'net-Wformat-fixes'

Florian Fainelli says:

====================
-Wformat fixes

This is a collection of some -Wformat fixes found during build, nothing
critical, but nice to have for people turning on more warnings with
their builds.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoveth: Fix -Wformat-truncation
Florian Fainelli [Fri, 22 Feb 2019 04:09:29 +0000 (20:09 -0800)]
veth: Fix -Wformat-truncation

Provide a precision hint to snprintf() in order to eliminate a
-Wformat-truncation warning provided below. A maximum of 11 characters
is allowed to reach a maximum of 32 - 1 characters given a possible
maximum value of queues using up to UINT_MAX which occupies 10
characters. Incidentally 11 is the number of characters for
"xdp_packets" which is the largest string we append.

drivers/net/veth.c: In function 'veth_get_strings':
drivers/net/veth.c:118:47: warning: '%s' directive output may be
truncated writing up to 31 bytes into a region of size between 12 and 21
[-Wformat-truncation=]
     snprintf(p, ETH_GSTRING_LEN, "rx_queue_%u_%s",
                                               ^~
drivers/net/veth.c:118:5: note: 'snprintf' output between 12 and 52
bytes into a destination of size 32
     snprintf(p, ETH_GSTRING_LEN, "rx_queue_%u_%s",
     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
       i, veth_rq_stats_desc[j].desc);
       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoe1000e: Fix -Wformat-truncation warnings
Florian Fainelli [Fri, 22 Feb 2019 04:09:28 +0000 (20:09 -0800)]
e1000e: Fix -Wformat-truncation warnings

Provide precision hints to snprintf() since we know the destination
buffer size of the RX/TX ring names are IFNAMSIZ + 5 - 1. This fixes the
following warnings:

drivers/net/ethernet/intel/e1000e/netdev.c: In function
'e1000_request_msix':
drivers/net/ethernet/intel/e1000e/netdev.c:2109:13: warning: 'snprintf'
output may be truncated before the last format character
[-Wformat-truncation=]
     "%s-rx-0", netdev->name);
             ^
drivers/net/ethernet/intel/e1000e/netdev.c:2107:3: note: 'snprintf'
output between 6 and 21 bytes into a destination of size 20
   snprintf(adapter->rx_ring->name,
   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     sizeof(adapter->rx_ring->name) - 1,
     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     "%s-rx-0", netdev->name);
     ~~~~~~~~~~~~~~~~~~~~~~~~
drivers/net/ethernet/intel/e1000e/netdev.c:2125:13: warning: 'snprintf'
output may be truncated before the last format character
[-Wformat-truncation=]
     "%s-tx-0", netdev->name);
             ^
drivers/net/ethernet/intel/e1000e/netdev.c:2123:3: note: 'snprintf'
output between 6 and 21 bytes into a destination of size 20
   snprintf(adapter->tx_ring->name,
   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     sizeof(adapter->tx_ring->name) - 1,
     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     "%s-tx-0", netdev->name);
     ~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6xxx: Fix -Wformat-security warnings
Florian Fainelli [Fri, 22 Feb 2019 04:09:27 +0000 (20:09 -0800)]
net: dsa: mv88e6xxx: Fix -Wformat-security warnings

We are not specifying an explicit format argument but instead passing a
string litteral which causes these two warnings to show up:

drivers/net/dsa/mv88e6xxx/chip.c: In function
'mv88e6xxx_irq_poll_setup':
drivers/net/dsa/mv88e6xxx/chip.c:483:2: warning: format not a string
literal and no format arguments [-Wformat-security]
  chip->kworker = kthread_create_worker(0, dev_name(chip->dev));
  ^~~~
drivers/net/dsa/mv88e6xxx/ptp.c: In function 'mv88e6xxx_ptp_setup':
drivers/net/dsa/mv88e6xxx/ptp.c:403:4: warning: format not a string
literal and no format arguments [-Wformat-security]
    dev_name(chip->dev));
    ^~~~~~~~
  LD [M]  drivers/net/dsa/mv88e6xxx/mv88e6xxx.o

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Avoid -Wformat-truncation warnings
Florian Fainelli [Fri, 22 Feb 2019 04:09:26 +0000 (20:09 -0800)]
mlxsw: spectrum: Avoid -Wformat-truncation warnings

Give precision identifiers to the two snprintf() formatting the priority
and TC strings to avoid producing these two warnings:

drivers/net/ethernet/mellanox/mlxsw/spectrum.c: In function
'mlxsw_sp_port_get_prio_strings':
drivers/net/ethernet/mellanox/mlxsw/spectrum.c:2132:37: warning: '%d'
directive output may be truncated writing between 1 and 3 bytes into a
region of size between 0 and 31 [-Wformat-truncation=]
   snprintf(*p, ETH_GSTRING_LEN, "%s_%d",
                                     ^~
drivers/net/ethernet/mellanox/mlxsw/spectrum.c:2132:3: note: 'snprintf'
output between 3 and 36 bytes into a destination of size 32
   snprintf(*p, ETH_GSTRING_LEN, "%s_%d",
   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     mlxsw_sp_port_hw_prio_stats[i].str, prio);
     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/net/ethernet/mellanox/mlxsw/spectrum.c: In function
'mlxsw_sp_port_get_tc_strings':
drivers/net/ethernet/mellanox/mlxsw/spectrum.c:2143:37: warning: '%d'
directive output may be truncated writing between 1 and 11 bytes into a
region of size between 0 and 31 [-Wformat-truncation=]
   snprintf(*p, ETH_GSTRING_LEN, "%s_%d",
                                     ^~
drivers/net/ethernet/mellanox/mlxsw/spectrum.c:2143:3: note: 'snprintf'
output between 3 and 44 bytes into a destination of size 32
   snprintf(*p, ETH_GSTRING_LEN, "%s_%d",
   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     mlxsw_sp_port_hw_tc_stats[i].str, tc);
     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobpfilter: re-add header search paths to tools include to fix build error
Masahiro Yamada [Fri, 22 Feb 2019 03:23:19 +0000 (12:23 +0900)]
bpfilter: re-add header search paths to tools include to fix build error

I thought header search paths to tools/include(/uapi) were unneeded,
but it looks like a build error occurs depending on the compiler.

Commit 303a339f30a9 ("bpfilter: remove extra header search paths for
bpfilter_umh") reintroduced the build error fixed by commit ae40832e53c3
("bpfilter: fix a build err").

Apology for the breakage, and thanks to Guenter for reporting this.

Fixes: 303a339f30a9 ("bpfilter: remove extra header search paths for bpfilter_umh")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-phy-at803x-Update-delays-for-RGMII-modes'
David S. Miller [Fri, 22 Feb 2019 23:30:03 +0000 (15:30 -0800)]
Merge branch 'net-phy-at803x-Update-delays-for-RGMII-modes'

Vinod Koul says:

====================
net: phy: at803x: Update delays for RGMII modes

Peter[1] reported that patch cd28d1d6e52e: ("net: phy: at803x: Disable
phy delay for RGMII mode") caused regression on am335x-evmsk board.
This board expects the Phy delay to be enabled but specified RGMII mode
which refers to delays being disabled. So fix this by disabling delay only
for RGMII mode and enable for RGMII_ID and RGMII_TXID/RXID modes.

While at it, as pointed by Dave, don't inline the helpers.

[1]: https://www.spinics.net/lists/netdev/msg550749.html

Changes in v4:
 - fix log & comments nbased on Marc's feedback
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: at803x: disable delay only for RGMII mode
Vinod Koul [Thu, 21 Feb 2019 10:23:15 +0000 (15:53 +0530)]
net: phy: at803x: disable delay only for RGMII mode

Per "Documentation/devicetree/bindings/net/ethernet.txt" RGMII mode
should not have delay in PHY whereas RGMII_ID and RGMII_RXID/RGMII_TXID
can have delay in PHY.

So disable the delay only for RGMII mode and enable for other modes.
Also treat the default case as disabled delays.

Fixes: cd28d1d6e52e: ("net: phy: at803x: Disable phy delay for RGMII mode")
Reported-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Reviewed-by: Niklas Cassel <niklas.cassel@linaro.org>
Tested-by: Peter Ujfalusi <peter.ujflausi@ti.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: at803x: don't inline helpers
Vinod Koul [Thu, 21 Feb 2019 10:23:14 +0000 (15:53 +0530)]
net: phy: at803x: don't inline helpers

Some helpers were declared with the "inline" function specifier.
It is preferable to let the compiler pick the right optimizations,
so drop the specifier for at803x_disable_rx_delay() and
at803x_disable_tx_delay()

Reviewed-by: Niklas Cassel <niklas.cassel@linaro.org>
Tested-by: Peter Ujfalusi <peter.ujflausi@ti.com>
Reviewed-by: Marc Gonzalez <marc.w.gonzalez@free.fr>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet_sched: initialize net pointer inside tcf_exts_init()
Cong Wang [Thu, 21 Feb 2019 05:37:42 +0000 (21:37 -0800)]
net_sched: initialize net pointer inside tcf_exts_init()

For tcindex filter, it is too late to initialize the
net pointer in tcf_exts_validate(), as tcf_exts_get_net()
requires a non-NULL net pointer. We can just move its
initialization into tcf_exts_init(), which just requires
an additional parameter.

This makes the code in tcindex_alloc_perfect_hash()
prettier.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx5: Support ndo bridge_setlink and getlink
Huy Nguyen [Wed, 20 Feb 2019 03:40:34 +0000 (21:40 -0600)]
net/mlx5: Support ndo bridge_setlink and getlink

Allow enabling VEPA mode on the HCA's port in legacy devlink mode.

Example:
bridge link set dev ens1f0 hwmode vepa
will turn on VEPA mode on the netdev ens1f0.

Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: E-Switch, Add support for VEPA in legacy mode.
Huy Nguyen [Mon, 21 Jan 2019 22:22:05 +0000 (16:22 -0600)]
net/mlx5: E-Switch, Add support for VEPA in legacy mode.

In Virtual Ethernet Port Aggregator (VEPA) mode, the packet skips
the system internal virtual switch and forwards to external network
switch. In Mellanox HCA case, the virtual switch is the HCA's Eswitch.

To support this, an new FDB flow table are created with level 0 and
linked to the existing FDB flow table in legacy mode. By default,
VEPA is turned off and this FDB flow table is empty. When VEPA is
turned on, two rules are created. One rule to forward on uplink vport
traffic to the legacy FDB. The other rule forward all other traffic
to uplink vport.

Other design alternatives were not chosen as explained below:
1. Create a forward rule in ACL flow table (most efficient design).
This approach is the not chosen because firmware does not support
forward rule to uplink vport (0xffff) for ACL flow table.
2. Add additional source port criteria in all the FDB rules to make the
FDB rules to be received rules only. This approach is not chosen because
it is not efficient as there can many rules in the FDB and VEPA mode
cannot be controlled per vport.
3. Add a highest prioirty flow group in the existing legacy FDB Flow
Table instead of a new flow table. This approoach does not work because the
new flow group has the same match criteria as the promiscuous flow group
and mlx5_add_flow_rules does not allow specifying flow group.

Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Fix mlx5e_tx_reporter_create return value
Eran Ben Elisha [Wed, 20 Feb 2019 10:33:01 +0000 (12:33 +0200)]
net/mlx5e: Fix mlx5e_tx_reporter_create return value

If reporter is ERR_PTR or NULL, error code shall be returned. At all other
cases it shall return success. Fix that.

Fixes: de8650a82071 ("net/mlx5e: Add tx reporter support")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Fix return status of TX reporter timeout recover
Eran Ben Elisha [Wed, 20 Feb 2019 10:10:48 +0000 (12:10 +0200)]
net/mlx5e: Fix return status of TX reporter timeout recover

In case of lost interrupt recover, we shall return success. Fix that.

Fixes: 7d91126b1aea ("net/mlx5e: Add tx timeout support for mlx5e tx reporter")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Reported-by: Maria Pasechnik <mariap@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Re-add support for TX timeout when TX reporter is not valid
Eran Ben Elisha [Tue, 19 Feb 2019 13:26:04 +0000 (15:26 +0200)]
net/mlx5e: Re-add support for TX timeout when TX reporter is not valid

When TX reporter was introduced, it took ownership over TX timeout error
handling. this introduced a regression in case TX reporter is not valid
(NET_DEVLINK is not set, or devlink_health_reporter_create failure).

Fix mlx5e_tx_reporter_timeout function so it can be called at all times.

In addition, remove a warning print that indicates that a TX timeout won't
be handled in case of no valid TX reporter.

Fixes: 7d91126b1aea ("net/mlx5e: Add tx timeout support for mlx5e tx reporter")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Fix warn print in case of TX reporter creation failure
Eran Ben Elisha [Tue, 19 Feb 2019 13:05:12 +0000 (15:05 +0200)]
net/mlx5e: Fix warn print in case of TX reporter creation failure

Print warning message in case of TX reporter creation failure, only if the
return value is ERR_PTR type. NULL pointer return indicates that
NET_DEVLINK is not set, and the warning print can be skipped.

Fixes: de8650a82071 ("net/mlx5e: Add tx reporter support")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Fix GRE key by controlling port tunnel entropy calculation
Eli Britstein [Mon, 14 Jan 2019 08:07:44 +0000 (10:07 +0200)]
net/mlx5e: Fix GRE key by controlling port tunnel entropy calculation

Flow entropy is calculated on the inner packet headers and used for
flow distribution in processing, routing etc. For GRE-type
encapsulations the entropy value is placed in the eight LSB of the key
field in the GRE header as defined in NVGRE RFC 7637. For UDP based
encapsulations the entropy value is placed in the source port of the
UDP header.
The hardware may support entropy calculation specifically for GRE and
for all tunneling protocols. With commit df2ef3bff193 ("net/mlx5e: Add
GRE protocol offloading") GRE is offloaded, but the hardware is
configured by default to calculate flow entropy so packets transmitted
on the wire have a wrong key. To support UDP based tunnels (i.e VXLAN),
GRE (i.e. no flow entropy) and NVGRE (i.e. with flow entropy) the
hardware behaviour must be controlled by the driver.

Ensure port entropy calculation is enabled for offloaded VXLAN tunnels
and disable port entropy calculation in the presence of offloaded GRE
tunnels by monitoring the presence of entropy enabling tunnels (i.e
VXLAN) and entropy disabing tunnels (i.e GRE).

Fixes: df2ef3bff193 ("net/mlx5e: Add GRE protocol offloading")
Signed-off-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: Introduce tunnel entropy control in PCMR register
Eli Britstein [Sun, 20 Jan 2019 20:33:19 +0000 (22:33 +0200)]
net/mlx5: Introduce tunnel entropy control in PCMR register

When using the device packet encapsulation offload, the device
calculates an entropy value, representing the inner packet headers. The
entropy field is placed inside the outer packet headers. For UDP-type
encapsulations, the entropy is placed in the source port field of the
UDP header. For GRE-type encapsulations, the entropy is placed in the 8
LSB of the key field in the GRE header. If the device does not recognize
the encapsulation type, the entropy is not placed in the packet.

Entropy setting can be controlled using PCMR register. if encapsulation
offload is not used force_entropy_cap should be set to 0x0. Entropy
setting is enabled/disabled using entropy_calc, and could be
additionally enabled/disabled for GRE encapsulation by entropy_gre_calc.

As a pre-step to automatically control the tunnel entropy, introduce
the entropy fields in the PCMR register with no functional change.

Signed-off-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: Use read-modify-write when changing PCMR register values
Eli Britstein [Mon, 21 Jan 2019 07:00:42 +0000 (09:00 +0200)]
net/mlx5: Use read-modify-write when changing PCMR register values

Currently changing a PCMR field is done by setting the field in a
zeroed buffer, zeroing other unrelated fields.
Fix this behaviour by modifying only the required field after first
reading the current register values, as a pre-step towards using more
fields in PCMR register.

Signed-off-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agoMerge tag 'wireless-drivers-next-for-davem-2019-02-22' of git://git.kernel.org/pub...
David S. Miller [Fri, 22 Feb 2019 20:56:24 +0000 (12:56 -0800)]
Merge tag 'wireless-drivers-next-for-davem-2019-02-22' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for 5.1

Most likely the last set of patches for 5.1. WPA3 support to ath10k
and qtnfmac. FTM support to iwlwifi and ath10k. And of course other
new features and bugfixes.

wireless-drivers was merged due to dependency in mt76.

Major changes:

iwlwifi

* HE radiotap

* FTM (Fine Timing Measurement) initiator and responder implementation

* bump supported firmware API to 46

* VHT extended NSS support

* new PCI IDs for 9260 and 22000 series

ath10k

* change QMI interface to support the new (and backwards incompatible)
  interface from HL3.1 and used in recent HL2.0 branch firmware
  releases

* support WPA3 with WCN3990

* support for mac80211 airtime fairness based on transmit rate
  estimation, the firmware needs to support WMI_SERVICE_PEER_STATS to
  enable this

* report transmit airtime to mac80211 with firmwares having
  WMI_SERVICE_REPORT_AIRTIME feature, this to have more accurate
  airtime fairness based on real transmit time (instead of just
  estimated from transmit rate)

* support Fine Timing Measurement (FTM) responder role

* add dynamic VLAN support with firmware having WMI_SERVICE_PER_PACKET_SW_ENCRYPT

* switch to use SPDX license identifiers

ath

* add new country codes for US

brcmfmac

* support monitor frames with the hardware/ucode header

qtnfmac

* enable WPA3 SAE and OWE support

mt76

* beacon support for USB devices (mesh+ad-hoc only)

rtlwifi

* convert to use SPDX license identifiers

libertas_tf

* get the MAC address before registering the device
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'AF_PACKET-transport_offset-fix'
David S. Miller [Fri, 22 Feb 2019 20:55:32 +0000 (12:55 -0800)]
Merge branch 'AF_PACKET-transport_offset-fix'

Maxim Mikityanskiy says:

====================
AF_PACKET transport_offset fix

This patch series contains the implementation of the RFC that was posted
on this mailing list previously:
https://www.spinics.net/lists/netdev/msg541709.html

It fixes having incorrect skb->transport_header values in cases when
dissect fails. Having correct values set by the kernel fixes mlx5
operation and allows to remove some unnecessary code flows in mlx5.

v2 changes:

- Rebase against the fresh net-next.
- Don't return bool from skb_probe_transport_header (and don't rename
  the function).
- WARN_ON_ONCE and error path in case of GSO without the L4 header.
====================

Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx5e: Trust kernel regarding transport offset
Maxim Mikityanskiy [Thu, 21 Feb 2019 12:40:05 +0000 (12:40 +0000)]
net/mlx5e: Trust kernel regarding transport offset

After AF_PACKET is fixed to calculate the transport header offset
correctly, trust the value set by the kernel. If the offset wasn't set,
it means there is no transport header in the packet.

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx5e: Remove the wrong assumption about transport offset
Maxim Mikityanskiy [Thu, 21 Feb 2019 12:40:04 +0000 (12:40 +0000)]
net/mlx5e: Remove the wrong assumption about transport offset

skb_transport_offset() == 0 is not a special value. The only special
value is when skb->transport_header is ~0U, and it's checked by
skb_transport_header_was_set().

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/packet: Remove redundant skb->protocol set
Maxim Mikityanskiy [Thu, 21 Feb 2019 12:40:02 +0000 (12:40 +0000)]
net/packet: Remove redundant skb->protocol set

c72219b75f introduced tpacket_set_protocol that parses the Ethernet L2
header and sets skb->protocol if it's unset. It is no longer needed
since the introduction of packet_parse_headers. In case of SOCK_RAW and
unset skb->protocol, packet_parse_headers asks the driver to tell the
protocol number, and it's implemented for all Ethernet devices. As the
old function supported only Ethernet, no functionality is lost.

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/packet: Ask driver for protocol if not provided by user
Maxim Mikityanskiy [Thu, 21 Feb 2019 12:40:01 +0000 (12:40 +0000)]
net/packet: Ask driver for protocol if not provided by user

If a socket was created with socket(AF_PACKET, SOCK_RAW, 0), the
protocol number is unavailable. Try to ask the driver to extract it from
the L2 header in order for skb_try_probe_transport_header to succeed.

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/ethernet: Add parse_protocol header_ops support
Maxim Mikityanskiy [Thu, 21 Feb 2019 12:39:59 +0000 (12:39 +0000)]
net/ethernet: Add parse_protocol header_ops support

The previous commit introduced parse_protocol callback which should
extract the protocol number from the L2 header. Make all Ethernet
devices support it.

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: Introduce parse_protocol header_ops callback
Maxim Mikityanskiy [Thu, 21 Feb 2019 12:39:58 +0000 (12:39 +0000)]
net: Introduce parse_protocol header_ops callback

Introduce a new optional header_ops callback called parse_protocol and a
wrapper function dev_parse_header_protocol, similar to dev_parse_header.

The new callback's purpose is to extract the protocol number from the L2
header, the format of which is known to the driver, but not to the upper
layers of the stack.

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: Don't set transport offset to invalid value
Maxim Mikityanskiy [Thu, 21 Feb 2019 12:39:57 +0000 (12:39 +0000)]
net: Don't set transport offset to invalid value

If the socket was created with socket(AF_PACKET, SOCK_RAW, 0),
skb->protocol will be unset, __skb_flow_dissect() will fail, and
skb_probe_transport_header() will fall back to the offset_hint, making
the resulting skb_transport_offset incorrect.

If, however, there is no transport header in the packet,
transport_header shouldn't be set to an arbitrary value.

Fix it by leaving the transport offset unset if it couldn't be found, to
be explicit rather than to fill it with some wrong value. It changes the
behavior, but if some code relied on the old behavior, it would be
broken anyway, as the old one is incorrect.

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mac80211-next-for-davem-2019-02-22' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Fri, 22 Feb 2019 20:52:23 +0000 (12:52 -0800)]
Merge tag 'mac80211-next-for-davem-2019-02-22' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
This time we have, of note:
 * the massive patch series for multi-BSSID support, I ended up
   applying that through a side branch to record some details
 * CSA improvements
 * HE (802.11ax) updates to Draft 3.3
 * strongly typed element iteration/etc. to make such code more
   readable - this came up in particular in multi-BSSID
 * rhashtable conversion patches from Herbert
Along, as usual, with various fixes and improvements.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'dsa-vlan'
David S. Miller [Fri, 22 Feb 2019 19:53:32 +0000 (11:53 -0800)]
Merge branch 'dsa-vlan'

Florian Fainelli says:

====================
net: dsa: VLAN devices w/ filtering

This patch series supports having VLAN devices on top of DSA/switch
ports while the switch has VLAN filtering globally turned on (as is the
case with Broadcom switches). Whether the switch does global or per-port
VLAN filtering, having VLAN entries for these VLAN devices is
beneficial.

We take care of a few possibly problematic cases:

- adding a VLAN device while there is an existing VLAN entry created by
  a VLAN aware bridge. The entire bridge's VLAN database and not just
  the specific bridge port is being checked to be safe and conserative

- adding a bridge VLAN entry when there is an existing VLAN device
  created is also not possible because that would lead to the bridge
  being able to manipulate the VLAN device's VID/attributes under its feet

- enslaving a VLAN device into a VLAN aware bridge since that duplicates
  functionality already offered by the VLAN aware bridge

Here are the different test cases that were run to exercise this:

ip addr flush dev gphy
ip link add dev br0 type bridge
echo 1 > /sys/class/net/br0/bridge/vlan_filtering
ip link set dev gphy master br0
udhcpc -i br0

vconfig add rgmii_1 100
ifconfig rgmii_1.100 192.168.100.10
ping -c 2 192.168.100.1

vconfig add br0 42
bridge vlan add vid 42 dev gphy
bridge vlan add vid 42 dev br0 self
ifconfig br0.42 192.168.42.2
ping -c 2 192.168.42.1

ip link del rgmii_1.100
vconfig add rgmii_1 100
ifconfig rgmii_1.100 192.168.100.10
ping -c 2 192.168.100.1
echo 0 > /sys/class/net/br0/bridge/vlan_filtering
ping -c 2 192.168.100.1

ip link del rgmii_1.100
echo 1 > /sys/class/net/br0/bridge/vlan_filtering

vconfig add rgmii_1 100
brctl addif br0 rgmii_1
bridge vlan add vid 100 dev rgmii_1

vconfig rem rgmii_1.100
bridge vlan add vid 100 dev rgmii_1
vconfig add rgmii_1 100

bridge vlan del vid 100 dev rgmii_1
vconfig add rgmii_1 100
brctl addif br0 rgmii_1.100
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Add ndo_vlan_rx_{add, kill}_vid implementation
Florian Fainelli [Wed, 20 Feb 2019 22:35:39 +0000 (14:35 -0800)]
net: dsa: Add ndo_vlan_rx_{add, kill}_vid implementation

In order to properly support VLAN filtering being enabled/disabled on a
bridge, while having other ports being non bridge port members, we need
to support the ndo_vlan_rx_{add,kill}_vid callbacks in order to make
sure the non-bridge ports can continue receiving VLAN tags, even when
the switch is globally configured to do ingress/egress VID checking.

Since we can call dsa_port_vlan_{add,del} with a bridge_dev pointer
NULL, we now need to check that in these two functions.

We specifically deal with two possibly problematic cases:

- creating a bridge VLAN entry while there is an existing VLAN device
  claiming that same VID

- creating a VLAN device while there is an existing bridge VLAN entry
  with that VID

Those are both resolved with returning -EBUSY back to user-space.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Deny enslaving VLAN devices into VLAN aware bridge
Florian Fainelli [Wed, 20 Feb 2019 22:35:38 +0000 (14:35 -0800)]
net: dsa: Deny enslaving VLAN devices into VLAN aware bridge

VLAN devices on top of a DSA network device which is already part of a
bridge and with said bridge being VLAN aware should not be allowed to be
enslaved into that bridge. For one, this duplicates functionality
offered by the VLAN aware bridge which supports tagged and untagged VLAN
frames processing and it would make things needlessly complex to e.g.:
propagate FDB/MDB accordingly.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoisdn_common: Mark expected switch fall-throughs
Gustavo A. R. Silva [Wed, 20 Feb 2019 20:25:26 +0000 (14:25 -0600)]
isdn_common: Mark expected switch fall-throughs

In preparation to enabling -Wimplicit-fallthrough, mark switch
cases where we are expecting to fall through.

This patch fixes the following warnings:

drivers/isdn/i4l/isdn_common.c: In function â€˜isdn_wildmat’:
drivers/isdn/i4l/isdn_common.c:173:5: warning: this statement may fall through [-Wimplicit-fallthrough=]
    p++;
    ~^~
drivers/isdn/i4l/isdn_common.c:174:3: note: here
   default:
   ^~~~~~~
  CC [M]  drivers/leds/leds-lp8788.o
  CC [M]  drivers/gpu/drm/amd/amdgpu/../powerplay/smumgr/smu10_smumgr.o
drivers/isdn/i4l/isdn_common.c: In function â€˜isdn_status_callback’:
drivers/isdn/i4l/isdn_common.c:729:6: warning: this statement may fall through [-Wimplicit-fallthrough=]
   if (divert_if)
      ^
drivers/isdn/i4l/isdn_common.c:732:2: note: here
  default:
  ^~~~~~~

Warning level 3 was used: -Wimplicit-fallthrough=3

Notice that, in this particular case, the code comment is modified
in accordance with what GCC is expecting to find.

This patch is part of the ongoing efforts to enable
-Wimplicit-fallthrough.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: aquantia: Use get_features for the PHYs abilities
Andrew Lunn [Wed, 20 Feb 2019 06:46:22 +0000 (07:46 +0100)]
net: phy: aquantia: Use get_features for the PHYs abilities

Use the new PHY driver call to get the PHYs supported features.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
[hkallweit1@gmail.com: removed new config_init callback from patch]
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agorhashtable: Remove obsolete rhashtable_walk_init function
Herbert Xu [Thu, 14 Feb 2019 14:03:27 +0000 (22:03 +0800)]
rhashtable: Remove obsolete rhashtable_walk_init function

The rhashtable_walk_init function has been obsolete for more than
two years.  This patch finally converts its last users over to
rhashtable_walk_enter and removes it.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: Use rhashtable_lookup_get_insert_fast instead of racy code
Herbert Xu [Thu, 14 Feb 2019 14:03:26 +0000 (22:03 +0800)]
mac80211: Use rhashtable_lookup_get_insert_fast instead of racy code

The code in mesh_path_add tries to handle the case where a duplicate
entry is added to the rhashtable by doing a lookup after a failed
insertion.  It also tries to handle races by repeating the insertion
should the lookup fail.

This is now unnecessary as we have rhashtable API functions that can
directly return the mathcing object.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agoMerge remote-tracking branch 'net-next/master' into mac80211-next
Johannes Berg [Fri, 22 Feb 2019 12:48:13 +0000 (13:48 +0100)]
Merge remote-tracking branch 'net-next/master' into mac80211-next

Merge net-next to resolve a conflict and to get the mac80211
rhashtable fixes so further patches can be applied on top.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: update HE IEs to D3.3
Liad Kaufman [Wed, 6 Feb 2019 11:17:20 +0000 (13:17 +0200)]
mac80211: update HE IEs to D3.3

Update element names and new fields according to D3.3 of
the HE spec.

Signed-off-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: allow sending vendor events unicast
Johannes Berg [Wed, 6 Feb 2019 11:17:19 +0000 (13:17 +0200)]
cfg80211: allow sending vendor events unicast

Sometimes, we may want to transport higher bandwidth data
through vendor events, and in that case sending it multicast
is a bad idea. Allow vendor events to be unicast.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: ignore quiet mode in probe
Sara Sharon [Wed, 6 Feb 2019 11:17:17 +0000 (13:17 +0200)]
mac80211: ignore quiet mode in probe

Some buggy APs keep the CSA IE in probes after the channel
switch was completed and can silence us for no good reason.
Apply quiet mode only from beacons. If there is real channel
switch going on, we will see the beacon anyway.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: allow CSA to self with immediate quiet
Sara Sharon [Wed, 6 Feb 2019 11:17:16 +0000 (13:17 +0200)]
mac80211: allow CSA to self with immediate quiet

Currently, due to some buggy APs that continue to include
CSA IEs after the switch, we ignore CSA to same channel.
However, some other APs may do CSA to self in order to have
immediate quiet. Allow it. Do it only for beacons.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: notify driver on subsequent CSA beacons
Sara Sharon [Wed, 6 Feb 2019 11:17:15 +0000 (13:17 +0200)]
mac80211: notify driver on subsequent CSA beacons

Some drivers may want to track further the CSA beacons, for example
to compensate for buggy APs that change the beacon count or quiet
mode during CSA flow.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: fix position of vendor_data read
Liad Kaufman [Wed, 6 Feb 2019 11:17:13 +0000 (13:17 +0200)]
mac80211: fix position of vendor_data read

The ieee80211_vendor_radiotap was read from the beginning
of the skb->data regardless of the existence of other
elements in radiotap that would cause it to move to another
position. Fix this by taking into account where it really
should be.

Signed-off-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agoradiotap: add 0-length PSDU "not captured" type
Johannes Berg [Wed, 6 Feb 2019 11:17:11 +0000 (13:17 +0200)]
radiotap: add 0-length PSDU "not captured" type

This type was defined in radiotap but we didn't add it to the
header file, add it now.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: abort CSA if beacon does not include CSA IEs
Sara Sharon [Wed, 6 Feb 2019 11:17:10 +0000 (13:17 +0200)]
mac80211: abort CSA if beacon does not include CSA IEs

In case we receive a beacon without CSA IE while we are in
the middle of channel switch - abort the operation.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: support max channel switch time element
Sara Sharon [Wed, 6 Feb 2019 11:17:09 +0000 (13:17 +0200)]
mac80211: support max channel switch time element

2018 REVmd of the spec introduces the max channel switch time
element which is optionally included in beacons/probes when there
is a channel switch / extended channel switch element.
The value represents the maximum delay between the time the AP
transmitted the last beacon in current channel and the expected
time of the first beacon in the new channel, in TU.

Parse the value and pass it to the driver.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: Report Association Request frame IEs in association events
Jouni Malinen [Fri, 15 Feb 2019 00:14:33 +0000 (02:14 +0200)]
cfg80211: Report Association Request frame IEs in association events

This extends the NL80211_CMD_ASSOCIATE event case to report
NL80211_ATTR_REQ_IE similarly to what is already done with the
NL80211_CMD_CONNECT events if the driver provides this information. In
practice, this adds (Re)Association Request frame information element
reporting to mac80211 drivers for the cases where user space SME is
used.

This provides more information for user space to figure out which
capabilities were negotiated for the association. For example, this can
be used to determine whether HT, VHT, or HE is used.

Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: pmsr: use eth_broadcast_addr() to assign broadcast address
Mao Wenan [Sat, 16 Feb 2019 09:47:10 +0000 (17:47 +0800)]
cfg80211: pmsr: use eth_broadcast_addr() to assign broadcast address

This patch is to use eth_broadcast_addr() to assign broadcast address
insetad of memset().

Signed-off-by: Mao Wenan <maowenan@huawei.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agorocker: Add missing break for PRE_BRIDGE_FLAGS
Florian Fainelli [Fri, 22 Feb 2019 02:26:46 +0000 (18:26 -0800)]
rocker: Add missing break for PRE_BRIDGE_FLAGS

A missing break keyword should have been added after adding support for
PRE_BRIDGE_FLAGS.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 93700458ff63 ("rocker: Check Handle PORT_PRE_BRIDGE_FLAGS")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'hns3-next'
David S. Miller [Fri, 22 Feb 2019 00:29:05 +0000 (16:29 -0800)]
Merge branch 'hns3-next'

Huazhong Tan says:

====================
code optimizations & bugfixes for HNS3 driver

This patchset includes bugfixes and code optimizations for
the HNS3 ethernet controller driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: clear command queue's registers when unloading VF driver
Huazhong Tan [Wed, 20 Feb 2019 02:32:51 +0000 (10:32 +0800)]
net: hns3: clear command queue's registers when unloading VF driver

According to the hardware's description, the driver should clear
the command queue's registers when uloading VF driver. Otherwise,
these existing value may lead the IMP get into a wrong state.

Fixes: fedd0c15d288 ("net: hns3: Add HNS3 VF IMP(Integrated Management Proc) cmd interface")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: uninitialize command queue while unloading PF driver
Huazhong Tan [Wed, 20 Feb 2019 02:32:50 +0000 (10:32 +0800)]
net: hns3: uninitialize command queue while unloading PF driver

According to the hardware's description, the driver should clear
the command queue's registers when uloading driver. Otherwise,
these existing value may lead the IMP get into a wrong state.

Also this patch adds hclge_cmd_uninit() to do the command queue
uninitialization which includes clearing registers and freeing
memory.

Fixes: 68c0a5c70614 ("net: hns3: Add HNS3 IMP(Integrated Mgmt Proc) Cmd Interface Support")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: Record VF vlan tables
liuzhongzhu [Wed, 20 Feb 2019 02:32:49 +0000 (10:32 +0800)]
net: hns3: Record VF vlan tables

Record the vlan tables that the VF sends to the chip.
After the VF exception, the PF actively clears the VF to chip config.

Signed-off-by: liuzhongzhu <liuzhongzhu@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: Record VF unicast and multicast tables
liuzhongzhu [Wed, 20 Feb 2019 02:32:48 +0000 (10:32 +0800)]
net: hns3: Record VF unicast and multicast tables

Record the unicast and multicast tables that the VF sends to the chip.
After the VF exception, the PF actively clears the VF to chip config.

Signed-off-by: liuzhongzhu <liuzhongzhu@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix 6th bit of ppp mpf abnormal errors
Weihang Li [Wed, 20 Feb 2019 02:32:47 +0000 (10:32 +0800)]
net: hns3: fix 6th bit of ppp mpf abnormal errors

This patch modify print message of 6th bit of ppp mpf abnormal errors,
there is a extra letter e in it.

Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: enable 8~11th bit of mac common msi-x error
Weihang Li [Wed, 20 Feb 2019 02:32:46 +0000 (10:32 +0800)]
net: hns3: enable 8~11th bit of mac common msi-x error

These bits are enabled now and have been test.

Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: some bugfix of ppu(rcb) ras errors
Weihang Li [Wed, 20 Feb 2019 02:32:45 +0000 (10:32 +0800)]
net: hns3: some bugfix of ppu(rcb) ras errors

The 3rd and 4th of PPU(RCB) PF Abnormal is RAS errors instead of MSI-X
like other bits. This patch adds process of handling and logging this
two bits. Otherwise, this patch modifies print message of 28th and 29th
bit of PPU MPF Abnormal errors, which keep same with other errors now.

Fixes: f69b10b317f9 ("net: hns3: handle hw errors of PPU(RCB)")
Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: modify print message of ssu common ecc errors
Weihang Li [Wed, 20 Feb 2019 02:32:44 +0000 (10:32 +0800)]
net: hns3: modify print message of ssu common ecc errors

This patch add information of specific bit in log to be consistent
with other type of errors, so that we can know which memory of ssu
has occurred a ecc ras errors.

Signed-off-by: Weihang Li <liweihang@hisilicon.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix port info query issue for copper port
Jian Shen [Wed, 20 Feb 2019 02:32:43 +0000 (10:32 +0800)]
net: hns3: fix port info query issue for copper port

In original codes, for copper port which doesn't connect to phy,
it always returns -EOPNOTSUPP when query port information. This
patch fixes it by return the port information of MAC.

Fixes: 5f373b158523 ("net: hns3: Fix speed/duplex information loss problem when executing ethtool ethx cmd of VF")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: convert mac advertize and supported from u32 to link mode
Jian Shen [Wed, 20 Feb 2019 02:32:42 +0000 (10:32 +0800)]
net: hns3: convert mac advertize and supported from u32 to link mode

The link mode with bits has been up to more than 31 for some MAC
and phy. Convert to using a linkmode bitmap, which can support all
link modes.

Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: Check variable is valid before assigning it to another
Yonglong Liu [Wed, 20 Feb 2019 02:32:41 +0000 (10:32 +0800)]
net: hns3: Check variable is valid before assigning it to another

In hnae3_register_ae_dev(), ae_algo->ops is assigned to ae_dev->ops
before check that ae_algo->ops is valid.

And in hnae3_register_ae_algo(), missing check for ae_algo->ops.

This patch fixes them.

Signed-off-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: add pointer checking at the beginning of the exported functions.
Yonglong Liu [Wed, 20 Feb 2019 02:32:40 +0000 (10:32 +0800)]
net: hns3: add pointer checking at the beginning of the exported functions.

These functions are exported, add pointer checking at the beginning
can make them more safe.

Signed-off-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mlxsw-Support-for-shared-buffers-in-Spectrum-2'
David S. Miller [Thu, 21 Feb 2019 23:57:46 +0000 (15:57 -0800)]
Merge branch 'mlxsw-Support-for-shared-buffers-in-Spectrum-2'

Ido Schimmel says:

====================
mlxsw: Support for shared buffers in Spectrum-2

Petr says:

Spectrum-2 will be configured with a different set of pools than
Spectrum-1, their sizes will be larger, and the individual quotas will
be different as well. It is therefore necessary to make the shared
buffer module aware of this dependence on chip type, and adjust the
individual tables.

In patch #1, introduce a structure for keeping per-chip immutable and
default values.

In patch #2, structures for keeping current values of SBPM and SBPR
(pool configuration and port-pool quota) are allocated dynamically to
support varying pool counts.

In patches #3 to #7, uses of individual shared buffer configuration
tables are migrated from global definitions to fields in struct
mlxsw_sp_sb_vals, which was introduced above.

Up until this point, the actual configuration is still the one suitable
for Spectrum-1. In patch #8 Spectrum-2 configuration is added.

In patch #9, port headroom configuration is changed to take into account
current recommended value for a 100-Gbps port, and the split factor.

In patch #10, requests for overlarge headroom are rejected. This avoids
potential chip freeze should such overlarge requests be made.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Reject overlarge headroom size requests
Petr Machata [Wed, 20 Feb 2019 19:32:29 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Reject overlarge headroom size requests

cap_max_headroom_size holds maximum headroom size supported.
Overstepping that limit might under certain conditions lead to ASIC
freeze.

Query and store the value, and add mlxsw_sp_sb_max_headroom_cells() for
obtaining the stored value. In __mlxsw_sp_port_headroom_set(), reject
requests where the total port buffer is larger than the advertised
maximum.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Update port headroom configuration
Petr Machata [Wed, 20 Feb 2019 19:32:27 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Update port headroom configuration

The recommendation for headroom size for 100Gbps port and 100m cable is
101.6KB, reduced accordingly for split ports. The closest higher number
evenly divisible by cell size for both Spectrum-1 and Spectrum-2, and
such that the number of cells can be further divided by maximum split
factor of 4, is 102528 bytes, or 25632 bytes per lane.

Update mlxsw_sp_port_pb_init() to compute the headroom taking into
account this recommended per-lane value and number of lanes actually
dedicated to a given port.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Add Spectrum-2 shared buffer configuration
Petr Machata [Wed, 20 Feb 2019 19:32:25 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Add Spectrum-2 shared buffer configuration

Customize the tables related to shared buffer configuration to match the
current recommendation for Spectrum-2 systems.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Keep mlxsw_sp_sb_mm in sb_vals
Petr Machata [Wed, 20 Feb 2019 19:32:23 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Keep mlxsw_sp_sb_mm in sb_vals

The SBMM register configures the shared buffer quota for MC packets
according to Switch-Priority. The default configuration depends on the
chip type. Therefore keep the table and length in struct
mlxsw_sp_sb_vals. Redirect the references from the global definitions to
the fields.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Keep mlxsw_sp_sb_cm in sb_vals
Petr Machata [Wed, 20 Feb 2019 19:32:22 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Keep mlxsw_sp_sb_cm in sb_vals

The SBCM register configures shared buffer quota according to
port-priority resp. port-TC. The default configuration depends on the
chip type. Therefore keep the tables and their lengths in struct
mlxsw_sp_sb_vals. Redirect the references from the global definitions to
the fields.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Keep mlxsw_sp_sb_prs in mlxsw_sp_sb_vals
Petr Machata [Wed, 20 Feb 2019 19:32:20 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Keep mlxsw_sp_sb_prs in mlxsw_sp_sb_vals

The SBPR register configures shared buffer pools. The default
configuration depends on the chip type. Therefore keep it in struct
mlxsw_sp_sb_vals. Redirect the one reference from the global array to
the field.

Because the pool descriptor ID is implicit in the ordering of array
members, both this array and the pool descriptor array have the same
length. Therefore reuse mlxsw_sp_sb.pool_dess_len for the purpose of
determining the length of SBPR array.

Drop the now useless MLXSW_SP_SB_PRS_LEN.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Keep mlxsw_sp_sb_pms in mlxsw_sp_sb_vals
Petr Machata [Wed, 20 Feb 2019 19:32:18 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Keep mlxsw_sp_sb_pms in mlxsw_sp_sb_vals

The SBPM register can be used to configure quotas for packets ingressing
from a certain pool to a certain port, and egressing from a certain pool
to a certain port. The default configuration depends on the chip type.
Therefore keep it in struct mlxsw_sp_sb_vals. Redirect the one reference
from the global array to the field.

Because the pool descriptor ID is implicit in the ordering of array
members, both this array and the pool descriptor array have the same
length. Therefore reuse mlxsw_sp_sb.pool_dess_len for the purpose of
determining the length of SBPM array.

Drop the now useless MLXSW_SP_SB_PMS_LEN.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Keep pool descriptors in mlxsw_sp_sb_vals
Petr Machata [Wed, 20 Feb 2019 19:32:16 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Keep pool descriptors in mlxsw_sp_sb_vals

Keep the table of pool descriptors and its length in struct
mlxsw_sp_sb_vals so that it can be specialized per chip type. Redirect
all users from the global definitions to the mlxsw_sp_sb fields.

Give mlxsw_sp_pool_count() an extra mlxsw_sp parameter so that it can
access the descriptor table.

Drop the now unnecessary MLXSW_SP_SB_POOL_DESS_LEN.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_buffers: Allocate prs & pms dynamically
Petr Machata [Wed, 20 Feb 2019 19:32:14 +0000 (19:32 +0000)]
mlxsw: spectrum_buffers: Allocate prs & pms dynamically

Spectrum-2 will be configured with a different set of pools than
Spectrum-1. The size of prs and pms buffers will therefore depend on the
chip type of the device.

Therefore, instead of reserving an array directly in a structure
definition, allocate the buffer in mlxsw_sp_sb_port{,s}_init().

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Add struct mlxsw_sp_sb_vals
Petr Machata [Wed, 20 Feb 2019 19:32:12 +0000 (19:32 +0000)]
mlxsw: spectrum: Add struct mlxsw_sp_sb_vals

Spectrum-2 will be configured with a different shared buffer
configuration than Spectrum-1. Therefore introduce a structure for
keeping the chip-specific default and immutable configuration.

Configuration mutable in runtime will still be kept in struct
mlxsw_sp_sb.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-stmmac-Performance-improvements-in-Multi-Queue'
David S. Miller [Thu, 21 Feb 2019 23:42:34 +0000 (15:42 -0800)]
Merge branch 'net-stmmac-Performance-improvements-in-Multi-Queue'

Jose Abreu says:

====================
net: stmmac: Performance improvements in Multi-Queue

Tested in XGMAC2 and GMAC5.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: dwxgmac2: Also use TBU interrupt to clean TX path
Jose Abreu [Tue, 19 Feb 2019 09:38:49 +0000 (10:38 +0100)]
net: stmmac: dwxgmac2: Also use TBU interrupt to clean TX path

TBU interrupt is a normal interrupt and can be used to trigger the
cleaning of TX path. Lets check if it's active in DMA interrupt handler.

While at it, refactor a little bit the function:
- Don't check if RI is enabled because at function exit we will
  only clear the interrupts that are enabled so, no event will
  be missed.

In my tests withe XGMAC2 this increased performance.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: dwmac4: Also use TBU interrupt to clean TX path
Jose Abreu [Tue, 19 Feb 2019 09:38:48 +0000 (10:38 +0100)]
net: stmmac: dwmac4: Also use TBU interrupt to clean TX path

TBU interrupt is a normal interrupt and can be used to trigger the
cleaning of TX path. Lets check if it's active in DMA interrupt handler.

While at it, refactor a little bit the function:
- Don't check if RI is enabled because at function exit we will
  only clear the interrupts that are enabled so, no event will be
  missed.

In my tests with GMAC5 this increased performance.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Fix NAPI poll in TX path when in multi-queue
Jose Abreu [Tue, 19 Feb 2019 09:38:47 +0000 (10:38 +0100)]
net: stmmac: Fix NAPI poll in TX path when in multi-queue

Commit 8fce33317023 introduced the concept of NAPI per-channel and
independent cleaning of TX path.

This is currently breaking performance in some cases. The scenario
happens when all packets are being received in Queue 0 but the TX is
performed in Queue != 0.

Fix this by using different NAPI instances per each TX and RX queue, as
suggested by Florian.

Changes from v2:
- Only force restart transmission if there are pending packets
Changes from v1:
- Pass entire ring size to TX clean path (Florian)

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-Get-rid-of-switchdev_port_attr_get'
David S. Miller [Thu, 21 Feb 2019 22:55:14 +0000 (14:55 -0800)]
Merge branch 'net-Get-rid-of-switchdev_port_attr_get'

Florian Fainelli says:

====================
net: Get rid of switchdev_port_attr_get()

This patch series splits the removal of the switchdev_ops that was
proposed a few times before and first tackles the easy part which is the
removal of the single call to switchdev_port_attr_get() within the
bridge code.

As suggestd by Ido, this patch series adds a
SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS which is used in the same
context as the caller of switchdev_port_attr_set(), so not deferred, and
then the operation is carried out in deferred context with setting a
support bridge port flag.

Follow-up patches will do the switchdev_ops removal after introducing
the proper helpers for the switchdev blocking notifier to work across
stacked devices (unlike the previous submissions).

David this does depend on Russell's "[PATCH net-next v5 0/3] net: dsa:
mv88e6xxx: fix IPv6".

Changes in v3:

- rebased against net-next/master after Russell's IPv6 changes to DSA
- ignore prepare/commit phase for PRE_BRIDGE_FLAGS since we don't
  want to trigger the WARN() in net/switchdev/switchdev.c in the commit
  phase

Changes in v2:

- differentiate callers not supporting switchdev_port_attr_set() from
  the driver not being able to support specific bridge flags

- pass "mask" instead of "flags" for the PRE_BRIDGE_FLAGS check

- skip prepare phase for PRE_BRIDGE_FLAGS

- corrected documentation a bit more

- tested bridge_vlan_aware.sh with veth/VRF
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: Get rid of switchdev_port_attr_get()
Florian Fainelli [Thu, 21 Feb 2019 00:58:26 +0000 (16:58 -0800)]
net: Get rid of switchdev_port_attr_get()

With the bridge no longer calling switchdev_port_attr_get() to obtain
the supported bridge port flags from a driver but instead trying to set
the bridge port flags directly and relying on driver to reject
unsupported configurations, we can effectively get rid of
switchdev_port_attr_get() entirely since this was the only place where
it was called.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: Remove SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS_SUPPORT
Florian Fainelli [Thu, 21 Feb 2019 00:58:25 +0000 (16:58 -0800)]
net: Remove SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS_SUPPORT

Now that we have converted the bridge code and the drivers to check for
bridge port(s) flags at the time we try to set them, there is no need
for a get() -> set() sequence anymore and
SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS_SUPPORT therefore becomes unused.

Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: bridge: Stop calling switchdev_port_attr_get()
Florian Fainelli [Thu, 21 Feb 2019 00:58:24 +0000 (16:58 -0800)]
net: bridge: Stop calling switchdev_port_attr_get()

Now that all switchdev drivers have been converted to check the
SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS flags and report flags that they
do not support accordingly, we can migrate the bridge code to try to set
that attribute first, check the results and then do the actual setting.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>