openwrt/staging/blogic.git
5 years agonet: phy: simplify genphy_config_eee_advert
Heiner Kallweit [Mon, 11 Feb 2019 21:16:13 +0000 (22:16 +0100)]
net: phy: simplify genphy_config_eee_advert

Use new function phy_modify_mmd_changed(), the result speaks for itself.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: Do not use async crypto for non-data records
Vakul Garg [Mon, 11 Feb 2019 11:31:05 +0000 (11:31 +0000)]
net/tls: Do not use async crypto for non-data records

Addition of tls1.3 support broke tls1.2 handshake when async crypto
accelerator is used. This is because the record type for non-data
records is not propagated to user application. Also when async
decryption happens, the decryption does not stop when two different
types of records get dequeued and submitted for decryption. To address
it, we decrypt tls1.2 non-data records in synchronous way. We check
whether the record we just processed has same type as the previous one
before checking for async condition and jumping to dequeue next record.

Fixes: 130b392c6cd6b ("net: tls: Add tls 1.3 support")
Signed-off-by: Vakul Garg <vakul.garg@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodevlink: use direct return of genlmsg_reply
Li RongQing [Mon, 11 Feb 2019 11:09:07 +0000 (19:09 +0800)]
devlink: use direct return of genlmsg_reply

This can remove redundant check

Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: use mvpp2_is_xlg() helper elsewhere
Russell King [Mon, 11 Feb 2019 10:23:15 +0000 (10:23 +0000)]
net: marvell: mvpp2: use mvpp2_is_xlg() helper elsewhere

There are several places which make the decision whether to access the
XLGMAC vs GMAC that only check for PHY_INTERFACE_MODE_10GKR and not its
XAUI variant.  Switch these to use the new helper so that we have
consistency through the driver.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: add mvpp2_is_xlg() helper
Russell King [Mon, 11 Feb 2019 10:23:10 +0000 (10:23 +0000)]
net: marvell: mvpp2: add mvpp2_is_xlg() helper

Add a mvpp2_is_xlg() helper to identify whether the interface mode
should be using the XLGMAC rather than the GMAC.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phylink: add phylink_init_eee() helper
Russell King [Mon, 11 Feb 2019 11:46:06 +0000 (11:46 +0000)]
net: phylink: add phylink_init_eee() helper

Provide phylink_init_eee() to allow MAC drivers to initialise PHY EEE
from within the ethtool set_eee() method.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phylink: only call mac_config() during resolve when link is up
Russell King [Mon, 11 Feb 2019 11:46:01 +0000 (11:46 +0000)]
net: phylink: only call mac_config() during resolve when link is up

There's little point calling mac_config() when the link is down.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoRevert "devlink: Add a generic wake_on_lan port parameter"
Vasundhara Volam [Mon, 11 Feb 2019 09:16:17 +0000 (14:46 +0530)]
Revert "devlink: Add a generic wake_on_lan port parameter"

This reverts commit b639583f9e36d044ac1b13090ae812266992cbac.

As per discussion with Jakub Kicinski and Michal Kubecek,
this will be better addressed by soon-too-come ethtool netlink
API with additional indication that given configuration request
is supposed to be persisted.

Also, remove the parameter support from bnxt_en driver.

Cc: Jiri Pirko <jiri@mellanox.com>
Cc: Michael Chan <michael.chan@broadcom.com>
Cc: Michal Kubecek <mkubecek@suse.cz>
Suggested-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-DMA-API'
David S. Miller [Tue, 12 Feb 2019 17:09:24 +0000 (12:09 -0500)]
Merge branch 'net-DMA-API'

Christoph Hellwig says:

====================
net: don't pass a NULL struct device to DMA API functions v2

We still have a few drivers which pass a NULL struct device pointer
to DMA API functions, which generally is a bad idea as the API
implementations rely on the device not only for ops selection, but
also the dma mask and various other attributes.

This series contains all easy conversions to pass a struct device,
besides that there also is some arch code that needs separate handling,
a driver that should not use the DMA API at all, and one that is
a complete basket case to be deal with separately.

Changes since v1:
 - fix an inverted ifdef in CAIF
 - update the smc911x changelog
 - split the series, this only contains the networking patches
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosmc911x: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:20:04 +0000 (14:20 +0100)]
smc911x: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Note that smc911x apparently is a PIO chip with an external DMA
handshake, and we probably use the wrong device here.  But at least
it matches the mapping side, which apparently works or at least
worked in the not too distant past.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agometh: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:20:03 +0000 (14:20 +0100)]
meth: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Also use GFP_KERNEL instead of GFP_ATOMIC as the gfp_t for the memory
allocation, as we aren't in interrupt context or under a lock.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomoxart_ether: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:20:02 +0000 (14:20 +0100)]
moxart_ether: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agopxa168_eth: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:20:01 +0000 (14:20 +0100)]
pxa168_eth: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Note that this driver seems to entirely lack dma_map_single error
handling, but that is left for another time.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agolantiq_etop: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:20:00 +0000 (14:20 +0100)]
lantiq_etop: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Note this driver seems to lack dma_unmap_* calls entirely, but fixing
that is left for another time.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomacb_main: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:19:59 +0000 (14:19 +0100)]
macb_main: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoau1000_eth: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:19:58 +0000 (14:19 +0100)]
au1000_eth: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: caif: pass struct device to DMA API functions
Christoph Hellwig [Mon, 11 Feb 2019 13:19:57 +0000 (14:19 +0100)]
net: caif: pass struct device to DMA API functions

The DMA API generally relies on a struct device to work properly, and
only barely works without one for legacy reasons.  Pass the easily
available struct device from the platform_device to remedy this.

Also use the proper Kconfig symbol to check for DMA API availability.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mlxsw-Several-updates'
David S. Miller [Tue, 12 Feb 2019 17:03:29 +0000 (12:03 -0500)]
Merge branch 'mlxsw-Several-updates'

Ido Schimmel says:

====================
mlxsw: Several updates

Patches #1-#3 contain misc updates for the mlxsw driver, one of which is
a fix following recent introduction of flow_rule infrastructure.

Patch #4 avoids double sourcing of lib.sh in forwarding selftests.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoselftests: mlxsw: avoid double sourcing of lib.sh
Jiri Pirko [Tue, 12 Feb 2019 16:29:54 +0000 (16:29 +0000)]
selftests: mlxsw: avoid double sourcing of lib.sh

Don't source lib.sh 2 times and make the script work with ifnames
passed on the command line.

Signed-off-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_flower: Fix VLAN modify action support
Ido Schimmel [Tue, 12 Feb 2019 16:29:53 +0000 (16:29 +0000)]
mlxsw: spectrum_flower: Fix VLAN modify action support

The driver does not support VLAN push and pop, but only VLAN modify.

Fixes: 738678817573 ("drivers: net: use flow action infrastructure")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_router: Drop unnecessary WARN_ON_ONCE()
Ido Schimmel [Tue, 12 Feb 2019 16:29:52 +0000 (16:29 +0000)]
mlxsw: spectrum_router: Drop unnecessary WARN_ON_ONCE()

In case the register access failed an error would be logged anyway, so
we can drop the warning.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Set LAG port collector only when active
Nir Dotan [Tue, 12 Feb 2019 16:29:51 +0000 (16:29 +0000)]
mlxsw: spectrum: Set LAG port collector only when active

The LAG port collecting (receive) function was mistakenly set when the
port was registered as a LAG member, while it should be set only when
the port collection state is set to true. Set LAG port to collecting
when it is set to distributing, as described in the IEEE link
aggregation standard coupled control mux machine state diagram.

Signed-off-by: Nir Dotan <nird@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 agoMerge branch 'net-smc-next'
David S. Miller [Tue, 12 Feb 2019 16:59:45 +0000 (11:59 -0500)]
Merge branch 'net-smc-next'

Ursula Braun says:

====================
net/smc: patches 2019-02-12

here are patches for SMC:
* patches 1 and 3 optimize SMC-R tx logic
* patch 2 is a cleanup without functional change
* patch 4 optimizes rx logic
* patches 5 and 6 improve robustness in link group and IB event handling
* patch 7 establishes Karsten Graul as another SMC maintainer
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMAINTAINERS: add Karsten as SMC maintainer
Ursula Braun [Tue, 12 Feb 2019 15:29:56 +0000 (16:29 +0100)]
MAINTAINERS: add Karsten as SMC maintainer

Add Karsten as additional maintainer for Shared Memory Communications
(SMC) Sockets.

Acked-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/smc: check port_idx of ib event
Karsten Graul [Tue, 12 Feb 2019 15:29:55 +0000 (16:29 +0100)]
net/smc: check port_idx of ib event

For robustness protect of higher port numbers than expected to avoid
setting bits behind our port_event_mask. In case of an DEVICE_FATAL
event all ports must be checked. The IB_EVENT_GID_CHANGE event is
provided in the global event handler, so handle it there. And handle a
QP_FATAL event instead of an DEVICE_FATAL event in the qp handler.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/smc: check connections in smc_lgr_free_work
Karsten Graul [Tue, 12 Feb 2019 15:29:54 +0000 (16:29 +0100)]
net/smc: check connections in smc_lgr_free_work

Remove the shortcut that smc_lgr_free() would skip the check for
existing connections when the link group is not in the link group list.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/smc: reduce amount of status updates to peer
Karsten Graul [Tue, 12 Feb 2019 15:29:53 +0000 (16:29 +0100)]
net/smc: reduce amount of status updates to peer

In smc_cdc_msg_recv_action() the received cdc message is evaluated.
To reduce the number of messaged triggered by this evaluation the logic
is streamlined. For the write_blocked condition we do not need to send
a response immediately. The remaining conditions can be put together
into one if clause.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/smc: no delay for free tx buffer wait
Karsten Graul [Tue, 12 Feb 2019 15:29:52 +0000 (16:29 +0100)]
net/smc: no delay for free tx buffer wait

When no free transfer buffers are available then a work to call
smc_tx_work() is scheduled. Set the schedule delay to zero, because for
the out-of-buffers condition the work can start immediately and will
block in the called function smc_wr_tx_get_free_slot(), waiting for free
buffers.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/smc: move wake up of close waiter
Karsten Graul [Tue, 12 Feb 2019 15:29:51 +0000 (16:29 +0100)]
net/smc: move wake up of close waiter

Move the call to smc_close_wake_tx_prepared() (which wakes up a possibly
waiting close processing that might wait for 'all data sent') to
smc_tx_sndbuf_nonempty() (which is the main function to send data).

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/smc: reset cursor update required flag
Karsten Graul [Tue, 12 Feb 2019 15:29:50 +0000 (16:29 +0100)]
net/smc: reset cursor update required flag

When an updated rx_cursor_confirmed field was sent to the peer then
reset the cons_curs_upd_req flag. And remove the duplicate reset and
cursor update in smc_tx_consumer_update().

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: remove double new line
Jakub Kicinski [Tue, 12 Feb 2019 08:18:02 +0000 (00:18 -0800)]
nfp: flower: remove double new line

Recent cls_flower offload rewrite added a double new line.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoflow_offload: Fix flow action infrastructure
Eli Britstein [Mon, 11 Feb 2019 07:52:59 +0000 (09:52 +0200)]
flow_offload: Fix flow action infrastructure

Implementation of macro "flow_action_for_each" introduced in
commit e3ab786b42535 ("flow_offload: add flow action infrastructure")
and used in commit 738678817573c ("drivers: net: use flow action
infrastructure") iterated the first item twice and did not reach the
last one. Fix it.

Fixes: e3ab786b42535 ("flow_offload: add flow action infrastructure")
Fixes: 738678817573c ("drivers: net: use flow action infrastructure")
Signed-off-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoDocumentation: fix some freescale dpio-driver.rst warnings
Randy Dunlap [Mon, 11 Feb 2019 06:32:42 +0000 (22:32 -0800)]
Documentation: fix some freescale dpio-driver.rst warnings

Fix markup warnings for one list by using correct list syntax.
Fix markup warnings for another list by using blank lines before the
list.

Documentation/networking/device_drivers/freescale/dpaa2/dpio-driver.rst:30: WARNING: Unexpected indentation.
Documentation/networking/device_drivers/freescale/dpaa2/dpio-driver.rst:143: WARNING: Unexpected indentation.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Stuart Yoder <stuyoder@gmail.com>
Cc: Laurentiu Tudor <laurentiu.tudor@nxp.com>
Cc: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Cc: netdev@vger.kernel.org
Cc: Madalin Bucur <madalin.bucur@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'devinfo-tweaks'
David S. Miller [Tue, 12 Feb 2019 04:39:56 +0000 (20:39 -0800)]
Merge branch 'devinfo-tweaks'

Jakub Kicinski says:

====================
devlink: minor tweaks to reported device info

This series contains two minor touch ups for devlink code. First
|| is corrected to && in the ethtool compat code. Next patch
decreases the stack allocation size.

On the nfp side after further discussions with the manufacturing
team we decided to realign the serial number contents slightly and
rename one of the other fields from "vendor" to "mfr", short for
"manufacture".

v2: - add patch 3 - move board maker as a generic attribute.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: devlink: include vendor/product info in serial number
Jakub Kicinski [Mon, 11 Feb 2019 03:35:31 +0000 (19:35 -0800)]
nfp: devlink: include vendor/product info in serial number

The manufacturing team requests we include vendor and product
in the serial number field, as the serial number itself is not
unique across manufacturing facilities and products.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: devlink: use the generic manufacture identifier instead of vendor
Jakub Kicinski [Mon, 11 Feb 2019 03:35:30 +0000 (19:35 -0800)]
nfp: devlink: use the generic manufacture identifier instead of vendor

Vendor may sound ambiguous, let's rename the fab string to
"board.manufacture" (which was just added as a generic identifier).

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodevlink: add a generic board.manufacture version name
Jakub Kicinski [Mon, 11 Feb 2019 03:35:29 +0000 (19:35 -0800)]
devlink: add a generic board.manufacture version name

At Jiri's suggestion add a generic "board.manufacture"
version identifier.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodevlink: don't allocate attrs on the stack
Jakub Kicinski [Mon, 11 Feb 2019 03:35:28 +0000 (19:35 -0800)]
devlink: don't allocate attrs on the stack

Number of devlink attributes has grown over 128, causing the
following warning:

../net/core/devlink.c: In function â€˜devlink_nl_cmd_region_read_dumpit’:
../net/core/devlink.c:3740:1: warning: the frame size of 1064 bytes is larger than 1024 bytes [-Wframe-larger-than=]
 }
  ^

Since the number of attributes is only going to grow allocate
the array dynamically.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodevlink: fix condition for compat device info
Jakub Kicinski [Mon, 11 Feb 2019 03:35:27 +0000 (19:35 -0800)]
devlink: fix condition for compat device info

We need the port to be both ethernet and have the rigth netdev,
not one or the other.

Fixes: ddb6e99e2db1 ("ethtool: add compat for devlink info")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodevlink: Add WARN_ON to catch errors of not cleaning devlink objects
Parav Pandit [Fri, 8 Feb 2019 21:15:00 +0000 (15:15 -0600)]
devlink: Add WARN_ON to catch errors of not cleaning devlink objects

Add WARN_ON to make sure that all sub objects of a devlink device are
cleanedup before freeing the devlink device.
This helps to catch any driver bugs.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: microchip: add switch offload forwarding support
Tristram Ha [Fri, 8 Feb 2019 04:05:58 +0000 (20:05 -0800)]
net: dsa: microchip: add switch offload forwarding support

The flag offload_fwd_mark is set as the switch can forward frames by
itself.

This can be considered a fix to a problem introduced in commit
c2e866911e254067 where the port membership are not set in sync.  The flag
offload_fwd_mark just needs to be set in tag_ksz.c to prevent the software
bridge from forwarding duplicate multicast frames.

Fixes: c2e866911e254067 ("microchip: break KSZ9477 DSA driver into two files")
Signed-off-by: Tristram Ha <Tristram.Ha@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-phy-register-modifying-helpers'
David S. Miller [Sun, 10 Feb 2019 20:53:19 +0000 (12:53 -0800)]
Merge branch 'net-phy-register-modifying-helpers'

Heiner Kallweit says:

====================
net: phy: add and use register modifying helpers returning 1 on change

Add and use register modifying helpers returning 1 on change.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: use phy_modify_changed in genphy_config_advert
Heiner Kallweit [Sun, 10 Feb 2019 18:59:57 +0000 (19:59 +0100)]
net: phy: use phy_modify_changed in genphy_config_advert

Use phy_modify_changed() to simplify the code.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: marvell10g: fix usage of new MMD modifying helpers
Heiner Kallweit [Sun, 10 Feb 2019 18:58:49 +0000 (19:58 +0100)]
net: phy: marvell10g: fix usage of new MMD modifying helpers

When replacing mv3310_modify() with phy_modify_mmd() we missed that
they behave differently, mv3310_modify() returns 1 on a changed
register value whilst phy_modify_mmd() returns 0. Fix this by replacing
phy_modify_mmd() with phy_modify_mmd_changed() where needed.

Fixes: b52c018ddccf ("net: phy: make use of new MMD accessors")
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: add register modifying helpers returning 1 on change
Heiner Kallweit [Sun, 10 Feb 2019 18:57:56 +0000 (19:57 +0100)]
net: phy: add register modifying helpers returning 1 on change

When modifying registers there are scenarios where we need to know
whether the register content actually changed. This patch adds
new helpers to not break users of the current ones, phy_modify() etc.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'tc-action-identifier-consistency'
David S. Miller [Sun, 10 Feb 2019 17:28:43 +0000 (09:28 -0800)]
Merge branch 'tc-action-identifier-consistency'

Eli Cohen says:

====================
Change tc action identifiers to be more consistent

This two patch series modifies TC actions identifiers to be more consistent and
also puts them in one place so new identifiers numbers can be chosen more
easily.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: Change TCA_ACT_* to TCA_ID_* to match that of TCA_ID_POLICE
Eli Cohen [Sun, 10 Feb 2019 12:25:00 +0000 (14:25 +0200)]
net: Change TCA_ACT_* to TCA_ID_* to match that of TCA_ID_POLICE

Modify the kernel users of the TCA_ACT_* macros to use TCA_ID_*. For
example, use TCA_ID_GACT instead of TCA_ACT_GACT. This will align with
TCA_ID_POLICE and also differentiates these identifier, used in struct
tc_action_ops type field, from other macros starting with TCA_ACT_.

To make things clearer, we name the enum defining the TCA_ID_*
identifiers and also change the "type" field of struct tc_action to
id.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: Move all TC actions identifiers to one place
Eli Cohen [Sun, 10 Feb 2019 12:24:59 +0000 (14:24 +0200)]
net: Move all TC actions identifiers to one place

Move all the TC identifiers to one place, to the same enum that defines
the identifier of police action. This makes it easier choose numbers for
new actions since they are now defined in one place. We preserve the
original values for binary compatibility. New IDs should be added inside
the enum.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: aquantia: add support for AQCS109
Nikita Yushchenko [Fri, 8 Feb 2019 21:12:23 +0000 (22:12 +0100)]
net: phy: aquantia: add support for AQCS109

Add support for the AQCS109. From software point of view,
it should be almost equivalent to AQR107.

v2:
- make Nikita the author
- document what I changed

Signed-off-by: Nikita Yushchenko <nikita.yoush@cogentembedded.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
[hkallweit1@gmail.com: use PHY_ID_MATCH_MODEL mascro]
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6xxx: SERDES support 2500BaseT via external PHY
Andrew Lunn [Fri, 8 Feb 2019 21:25:44 +0000 (22:25 +0100)]
net: dsa: mv88e6xxx: SERDES support 2500BaseT via external PHY

By using an external PHY, ports 9 and 10 can support 2500BaseT.
So set this link mode in the mask when validating.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoadd snmp counter document
yupeng [Sat, 9 Feb 2019 22:46:18 +0000 (14:46 -0800)]
add snmp counter document

add document for tcp retransmission, tcp fast open, syn cookies,
challenge ack, prune and several general counters

Signed-off-by: yupeng <yupeng0921@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: clear flow control modes in 10G mode
Russell King [Sat, 9 Feb 2019 16:06:51 +0000 (16:06 +0000)]
net: marvell: mvpp2: clear flow control modes in 10G mode

When mvpp2 configures the flow control modes in mvpp2_xlg_config() for
10G mode, it only ever set the flow control enable bits.  There is no
mechanism to clear these bits, which means that userspace is unable to
use standard APIs to disable flow control (the only way is to poke the
register directly.)

Fix the missing bit clearance to allow flow control to be disabled.
This means that, by default, as there is no negotiation in 10G modes
with mvpp2, flow control is now disabled rather than being rx-only.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: Add support for asking the PHY its abilities
Andrew Lunn [Sat, 9 Feb 2019 14:24:47 +0000 (15:24 +0100)]
net: phy: Add support for asking the PHY its abilities

Add support for runtime determination of what the PHY supports, by
adding a new function to the phy driver. The get_features call should
set the phydev->supported member with the features the PHY supports.
It is only called if phydrv->features is NULL.

This requires minor changes to pause. The PHY driver should not set
pause abilities, except for when it has odd cause capabilities, e.g.
pause cannot be disabled. With this change, phydev->supported already
contains the drivers abilities, including pause. So rather than
considering phydrv->features, look at the phydev->supported, and
enable pause if neither of the pause bits are already set.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
[hkallweit1@gmail.com: fixed small checkpatch complaint in one comment]
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: probe the PHY before determining the supported features
Andrew Lunn [Sat, 9 Feb 2019 13:46:30 +0000 (14:46 +0100)]
net: phy: probe the PHY before determining the supported features

We will soon support asking the PHY at runtime to determine what
features it supports, rather than forcing it to be compile time.
But we should probe the PHY first. So probe the phy driver earlier.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: remove unneeded masking of PHY register read results
Heiner Kallweit [Sat, 9 Feb 2019 08:46:53 +0000 (09:46 +0100)]
net: phy: remove unneeded masking of PHY register read results

PHY registers are only 16 bits wide, therefore, if the read was
successful, there's no need to mask out the higher 16 bits.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: Disable async decrytion for tls1.3
Vakul Garg [Sat, 9 Feb 2019 07:53:28 +0000 (07:53 +0000)]
net/tls: Disable async decrytion for tls1.3

Function tls_sw_recvmsg() dequeues multiple records from stream parser
and decrypts them. In case the decryption is done by async accelerator,
the records may get submitted for decryption while the previous ones may
not have been decryted yet. For tls1.3, the record type is known only
after decryption. Therefore, for tls1.3, tls_sw_recvmsg() may submit
records for decryption even if it gets 'handshake' records after 'data'
records. These intermediate 'handshake' records may do a key updation.
By the time new keys are given to ktls by userspace, it is possible that
ktls has already submitted some records i(which are encrypted with new
keys) for decryption using old keys. This would lead to decrypt failure.
Therefore, async decryption of records should be disabled for tls1.3.

Fixes: 130b392c6cd6b ("net: tls: Add tls 1.3 support")
Signed-off-by: Vakul Garg <vakul.garg@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: disregard "Clause 22 registers present" bit in get_phy_c45_devs_in_pkg
Heiner Kallweit [Fri, 8 Feb 2019 18:25:22 +0000 (19:25 +0100)]
net: phy: disregard "Clause 22 registers present" bit in get_phy_c45_devs_in_pkg

Bit 0 in register 1.5 doesn't represent a device but is a flag that
Clause 22 registers are present. Therefore disregard this bit when
populating the device list. If code needs this information it
should read register 1.5 directly instead of accessing the device
list.
Because this bit doesn't represent a device don't define a
MDIO_MMD_XYZ constant, just define a MDIO_DEVS_XYZ constant for
the flag in the device list bitmap.

v2:
- make masking of bit 0 more explicit
- improve commit message

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mvpp2-phylink-fixes'
David S. Miller [Sat, 9 Feb 2019 07:08:39 +0000 (23:08 -0800)]
Merge branch 'mvpp2-phylink-fixes'

Russell King says:

====================
mvpp2 phylink fixes

Having spent a while debugging issues with Sven Auhagen, it appears
that the mvpp2 network driver's phylink support isn't quite correct.

This series fixes that up, but, despite being tested locally, by
Sven, and by Antoine, I would prefer it to be applied to net-next
so that there is time for more people to test before it hits -rc or
stable backports.

The symptoms were that although PHYs would come up, the GMAC never
reported that the link was up, or in some cases it did report link
up but packets would not flow.  Various approaches were tried to
work around that, such as switching to in-band negotiation from
PHY mode, but ultimately the problem was in the way mvpp2 was being
programmed.

This series addresses that by, essentially, making mvpp2 follow the
same implementation pattern as mvneta: we configure the GMAC in three
stages:

1) the PHY interface mode
2) the negotiation advert
3) the negotiation style

Another issue is that mvpp2 was always taking the link down each time
its mac_config method was called: this is disruptive when the link is
already up, and we're just updating settings such as flow control.
There are some circumstances where we make the call despite there
being no changes (eg, when phylink is polling a GPIO or using a custom
link state function.)

This series depends on two previous patches already sent for net-next:
  net: marvell: mvpp2: fix lack of link interrupts
  net: marvell: mvpp2: use phy_interface_mode_is_8023z() helper

There is one last patch which deals with link status interrupts, which
I'll send separately because I think there's other considerations, but
that should not hold up this series of patches.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: fix AN restart
Russell King [Fri, 8 Feb 2019 15:35:59 +0000 (15:35 +0000)]
net: marvell: mvpp2: fix AN restart

phylink already limits which interface modes are able to call the
MACs AN restart function, but in any case, the commentry seems
incorrect: the AN restart bit does not automatically clear when
set.  This has been found via manual setting using devmem2, and
we can observe that the AN does indeed restart and complete, yet
the AN restart bit remains set.  Explicitly clear the AN restart
bit.

Tested-by: Sven Auhagen <sven.auhagen@voleatech.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: read correct pause bits
Russell King [Fri, 8 Feb 2019 15:35:54 +0000 (15:35 +0000)]
net: marvell: mvpp2: read correct pause bits

When reading the pause bits in mac_link_state, mvpp2 was reporting
the state of the "active pause" bits, which are set when the MAC is
in pause mode.  This is not what phylink wants - we want the
negotiated pause state.  Fix the definition so we read the correct
bits.

Tested-by: Sven Auhagen <sven.auhagen@voleatech.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: only reprogram what is necessary on mac_config
Russell King [Fri, 8 Feb 2019 15:35:49 +0000 (15:35 +0000)]
net: marvell: mvpp2: only reprogram what is necessary on mac_config

mac_config() can be called at any point, and the expected behaviour
from MAC drivers is to only reprogram when necessary - and certainly
avoid taking the link down on every call.

Unfortunately, mvpp2 does exactly that - it takes the link down, and
reprograms everything, and then releases the forced-link down.

This is bad, it can cause the link to bounce:

- SFP detects signal, disables LOS indication.
- SFP code calls into phylink, calling phylink_sfp_link_up() which
  triggers a resolve.
- phylink_resolve() calls phylink_get_mac_state() and finds the MAC
  reporting link up.
- phylink wants to configure the pause mode on the MAC, so calls
  phylink_mac_config()
- mvpp2 takes the link down temporarily, generating a MAC link down
  event followed by another MAC link event.
- phylink calls mac_link_up() and then processes the MAC link down
  event.
- phylink_resolve() gets called again, registers the link down, and
  calls mach_link_down() before re-running itself.
- phylink_resolve() starts again at step 3 above.  This sequence
  repeats.

GMAC versions prior to mvpp2 do not require the link to be taken down
except when certain link properties (eg, switching between SGMII and
1000base-X mode, or enabling/disabling in-band negotiation) are
changed.  Implement this for mvpp2.

Tested-by: Sven Auhagen <sven.auhagen@voleatech.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: fix stuck in-band SGMII negotiation
Russell King [Fri, 8 Feb 2019 15:35:43 +0000 (15:35 +0000)]
net: marvell: mvpp2: fix stuck in-band SGMII negotiation

It appears that the mvpp22 can get stuck with SGMII negotiation.  The
symptoms are that in-band negotiation never completes and the partner
(eg, PHY) never reports SGMII link up, or if it supports negotiation
bypass, goes into negotiation bypass mode (which will happen when the
PHY sees that the MAC is alive but gets no response.)

Triggering the PHY end of the link to re-negotiate results in the
bypass bit clearing on the PHY, and then re-setting - indicating that
the problem is at the mvpp22 GMAC end.

Asserting the GMAC reset and de-asserting it resolves the issue.
Arrange to assert the GMAC reset at probe time, and deassert it only
after we have configured the GMAC for the appropriate mode.  This
resolves the issue.

Tested-by: Sven Auhagen <sven.auhagen@voleatech.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: marvell: mvpp2: phylink compliance updates
Russell King [Fri, 8 Feb 2019 15:35:38 +0000 (15:35 +0000)]
net: marvell: mvpp2: phylink compliance updates

Sven Auhagen reported issues with negotiation on a couple of his
platforms using a mixture of SFP and PHYs in various different
modes.  Debugging to root cause proved difficult, but essentially
the problem comes down to the mvpp2 phylink implementation being
slightly at odds with what is expected.

phylink operates in three modes: phy, fixed-link, and in-band mode.

In the first two modes, the expected behaviour from a MAC driver is
that phylink resolves the operating mode and passes the mode to the
MAC driver for it to program, including when the link should be
brought up or taken down.  This is basically the same as the libphy
approach.  This does not negate the requirement to advertise a correct
control word for interface modes that have control words where that
can be reasonably controlled.

The second mode is in-band mode, where the MAC is expected to use the
in-band control word to determine the operating mode.

The mvneta driver implements the correct pattern required to support
this: configure the port interface type separately from the in-band
mode(s).  This is now specified in the phylink documentation patches.

mvpp2 was programming in-band mode for SGMII and the 802.3z modes no
what, and avoided forcing the link up in fixed/phy modes.  This caused
a problem with some boards where the PHY is by default programmed to
enter AN bypass mode, the PHY would report that the link was up, but
the mvpp2 never completed the exchange of control word.

Another issue that mvpp2 has is it sets SGMII AN format control word
for both SGMII and 802.3z modes. The format of the control word is
defined by MVPP2_GMAC_INBAND_AN_MASK, which should be set for SGMII
and clear for 802.3z. Available Marvell documentation for earlier
GMAC implementations does not make this clear, but this has been
ascertained via extensive testing on earlier GMAC implementations,
and then confirmed with a Macchiatobin Single Shot connected to a
Clearfog: when MVPP2_GMAC_INBAND_AN_MASK is set, the clearfog does
not receive the advertised pause mode settings.

Lastly, there is no flow control in the in-band control word in Cisco
SGMII, setting the flow control autonegotiation bit even with a PHY
that has the Marvell extension to send this information does not result
in the flow control being enabled at the MAC.  We need to do this
manually using the information provided via phylink.

Re-code mvpp2's mac_config() and mac_link_up() to follow this pattern.
This allows Sven Auhagen's board and Macchiatobin to reliably bring
the link up with the 88e1512 PHY with phylink operating in PHY mode
with COMPHY built as a module but the rest of the networking built-in,
and u-boot having brought up the interface.  in-band mode requires an
additional patch to resolve another problem.

Tested-by: Sven Auhagen <sven.auhagen@voleatech.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoethtool: Remove unnecessary null check in ethtool_rx_flow_rule_create
Nathan Chancellor [Fri, 8 Feb 2019 04:46:53 +0000 (21:46 -0700)]
ethtool: Remove unnecessary null check in ethtool_rx_flow_rule_create

net/core/ethtool.c:3023:19: warning: address of array
'ext_m_spec->h_dest' will always evaluate to 'true'
[-Wpointer-bool-conversion]
                if (ext_m_spec->h_dest) {
                ~~  ~~~~~~~~~~~~^~~~~~

h_dest is an array, it can't be null so remove this check.

Fixes: eca4205f9ec3 ("ethtool: add ethtool_rx_flow_spec to flow_rule structure translator")
Link: https://github.com/ClangBuiltLinux/linux/issues/353
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoixgbe: Use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 04:22:58 +0000 (22:22 -0600)]
ixgbe: Use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = kzalloc(size, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL);

Notice that, in this case, variable size is not necessary, hence
it is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoigc: Use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 04:19:45 +0000 (22:19 -0600)]
igc: Use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = kzalloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL)

Notice that, in this case, variable size is not necessary, hence
it is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoigb: use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 04:15:40 +0000 (22:15 -0600)]
igb: use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = alloc(size, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

size = struct_size(instance, entry, count);

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: don't double-read link status register if link is up
Heiner Kallweit [Thu, 7 Feb 2019 19:22:20 +0000 (20:22 +0100)]
net: phy: don't double-read link status register if link is up

The link status register latches link-down events. Therefore, if link
is reported as being up, there's no need for a second read.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agofm10k: use struct_size() in kzalloc()
Gustavo A. R. Silva [Fri, 8 Feb 2019 03:55:37 +0000 (21:55 -0600)]
fm10k: use struct_size() in kzalloc()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = kzalloc(size, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL);

Notice that, in this case, variable size is not necessary, hence
it is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: cmsg: use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 03:47:25 +0000 (21:47 -0600)]
nfp: flower: cmsg: use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    void *entry[];
};

size = sizeof(struct foo) + count * sizeof(void *);
instance = alloc(size, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = alloc(struct_size(instance, entry, count), GFP_KERNEL);

Notice that, in this case, variable size is not necessary, hence
it is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum_router: Use struct_size() in kzalloc()
Gustavo A. R. Silva [Fri, 8 Feb 2019 03:42:41 +0000 (21:42 -0600)]
mlxsw: spectrum_router: Use struct_size() in kzalloc()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = kzalloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL)

Notice that, in this case, variable alloc_size is not necessary, hence
it is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnx2x: Use struct_size() in kzalloc()
Gustavo A. R. Silva [Fri, 8 Feb 2019 03:29:10 +0000 (21:29 -0600)]
bnx2x: Use struct_size() in kzalloc()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = kzalloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL)

Notice that, in this case, variable fsz is not necessary, hence
it is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agowimax/i2400m: use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 03:22:17 +0000 (21:22 -0600)]
wimax/i2400m: use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    void *entry[];
};

size = sizeof(struct foo) + count * sizeof(void *);
instance = alloc(size, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

size = struct_size(instance, entry, count);

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agowan: wanxl: use struct_size() in kzalloc()
Gustavo A. R. Silva [Fri, 8 Feb 2019 03:16:48 +0000 (21:16 -0600)]
wan: wanxl: use struct_size() in kzalloc()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = alloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = alloc(struct_size(instance, entry, count), GFP_KERNEL)

Notice that, in this case, variable alloc_size is not necessary, hence
it is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: usb: cdc-phonet: use struct_size() in alloc_netdev()
Gustavo A. R. Silva [Fri, 8 Feb 2019 03:13:13 +0000 (21:13 -0600)]
net: usb: cdc-phonet: use struct_size() in alloc_netdev()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    void *entry[];
};

instance = alloc(sizeof(struct foo) + count * sizeof(void *));

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = alloc(struct_size(instance, entry, count));

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: use struct_size() in devm_kzalloc()
Gustavo A. R. Silva [Fri, 8 Feb 2019 01:16:03 +0000 (19:16 -0600)]
net: dsa: use struct_size() in devm_kzalloc()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = alloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = alloc(struct_size(instance, entry, count), GFP_KERNEL)

Notice that, in this case, variable size is not necessary, hence it is
removed.

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agompls_iptunnel: use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 01:10:52 +0000 (19:10 -0600)]
mpls_iptunnel: use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

instance = alloc(sizeof(struct foo) + count * sizeof(struct boo));

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = alloc(struct_size(instance, entry, count));

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 01:02:52 +0000 (19:02 -0600)]
net/sched: use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = alloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

size = struct_size(instance, entry, count);
instance = alloc(size, GFP_KERNEL)

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobridge: use struct_size() helper
Gustavo A. R. Silva [Fri, 8 Feb 2019 00:58:56 +0000 (18:58 -0600)]
bridge: use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = alloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

size = struct_size(instance, entry, count);

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'qed-SmartAN-query-support'
David S. Miller [Sat, 9 Feb 2019 06:39:01 +0000 (22:39 -0800)]
Merge branch 'qed-SmartAN-query-support'

Sudarsana Reddy Kalluru says:

====================
qed*: SmartAN query support

SmartAN feature detects the peer/cable capabilities and establishes the
link in the best possible configuration.
The patch series adds support for querying the capability. Please consider
applying it net-next.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoqede: Add ethtool interface for SmartAN query.
Sudarsana Reddy Kalluru [Thu, 7 Feb 2019 14:20:12 +0000 (06:20 -0800)]
qede: Add ethtool interface for SmartAN query.

The patch adds driver support to query SmartAN capability via ethtool.

Signed-off-by: Sudarsana Reddy Kalluru <skalluru@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Michal Kalderon <mkalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoqed: Add API for SmartAN query.
Sudarsana Reddy Kalluru [Thu, 7 Feb 2019 14:20:11 +0000 (06:20 -0800)]
qed: Add API for SmartAN query.

The patch adds driver interface to read the SmartAN capability from
management firmware.

Signed-off-by: Sudarsana Reddy Kalluru <skalluru@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Michal Kalderon <mkalderon@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-dsa-bcm_sf2-Add-support-for-CFP-statistics'
David S. Miller [Sat, 9 Feb 2019 06:11:07 +0000 (22:11 -0800)]
Merge branch 'net-dsa-bcm_sf2-Add-support-for-CFP-statistics'

Florian Fainelli says:

====================
net: dsa: bcm_sf2: Add support for CFP statistics

The Broadcom SF2 switch has a Compact Field Processor (CFP) which not
only can perform matching + action, but also counts the number of times
a rule has been hit. This is invaluable while debugging when/if rules
are not matched.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: bcm_sf2: Allow looping back CFP rules
Florian Fainelli [Wed, 6 Feb 2019 20:46:00 +0000 (12:46 -0800)]
net: dsa: bcm_sf2: Allow looping back CFP rules

When the source and destination port of a CFP rule match, we must set
the loopback bit enable to allow that, otherwise the frame is discarded.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: bcm_sf2: Add support for CFP statistics
Florian Fainelli [Wed, 6 Feb 2019 20:45:59 +0000 (12:45 -0800)]
net: dsa: bcm_sf2: Add support for CFP statistics

Return CFP policer statistics (Green, Yellow or Red) as part of the
standard ethtool statistics. This helps debug when CFP rules may not be
hit (0 counter).

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: bcm_sf2: Prepare for adding CFP statistics
Florian Fainelli [Wed, 6 Feb 2019 20:45:58 +0000 (12:45 -0800)]
net: dsa: bcm_sf2: Prepare for adding CFP statistics

In preparation for adding CFP statistics, we will need to overlay the
standard B53 statistics, so create specific bcm_sf2_sw_* functions to
call into their b53_common.c counterpart.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: bcm_sf2: Remove stats mutex
Florian Fainelli [Wed, 6 Feb 2019 20:45:57 +0000 (12:45 -0800)]
net: dsa: bcm_sf2: Remove stats mutex

We no longer need a dedicated statistics mutex since we leverage
b53_common for statistics now.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: consider latched link-down status in polling mode
Heiner Kallweit [Wed, 6 Feb 2019 18:39:52 +0000 (19:39 +0100)]
net: phy: consider latched link-down status in polling mode

The link status value latches link-down events. To get the current
status we read the register twice in genphy_update_link(). There's
a potential risk that we miss a link-down event in polling mode.
This may cause issues if the user e.g. connects his machine to a
different network.

On the other hand reading the latched value may cause issues in
interrupt mode. Following scenario:

- After boot link goes up
- phy_start() is called triggering an aneg restart, hence link goes
  down and link-down info is latched.
- After aneg has finished link goes up and triggers an interrupt.
  Interrupt handler reads link status, means it reads the latched
  "link is down" info. But there won't be another interrupt as long
  as link stays up, therefore phylib will never recognize that link
  is up.

Deal with both scenarios by reading the register twice in interrupt
mode only.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: macb: add sam9x60-macb compatibility string
Nicolas Ferre [Wed, 6 Feb 2019 17:56:10 +0000 (18:56 +0100)]
net: macb: add sam9x60-macb compatibility string

Add a new compatibility string for this product. It's using
at91sam9260-macb layout but has a newer hardware revision: it's safer
to use its own string.

Signed-off-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/macb: bindings doc: add sam9x60 binding
Nicolas Ferre [Wed, 6 Feb 2019 17:56:09 +0000 (18:56 +0100)]
net/macb: bindings doc: add sam9x60 binding

Add the compatibility sting documentation for sam9x60 10/100 interface.

Signed-off-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/macb: bindings doc/trivial: fix documentation for sama5d3 10/100 interface
Nicolas Ferre [Wed, 6 Feb 2019 17:56:08 +0000 (18:56 +0100)]
net/macb: bindings doc/trivial: fix documentation for sama5d3 10/100 interface

This removes a line left while adding the correct compatibility string for
sama5d3 10/100 interface. Now use the "atmel,sama5d3-macb" string.

Signed-off-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mlxsw-Implement-periodic-ERP-rehash'
David S. Miller [Fri, 8 Feb 2019 23:02:50 +0000 (15:02 -0800)]
Merge branch 'mlxsw-Implement-periodic-ERP-rehash'

Ido Schimmel says:

====================
mlxsw: Implement periodic ERP rehash

Currently, an ERP set is created for each region according to rules
inserted and order of their insertion. However that might lead to
suboptimal ERP sets and possible unnecessary spillage into C-TCAM.
This patchset aims to fix this problem and introduces periodical checking
of used ERP sets and in case a better ERP set is possible for the given
set of rules, it rehashes the region to use the better ERP set.

Patch 1 prepares devlink params infra in order to fix the
        init/fini sequences.
Patch 2 implements hints infra in objagg library.
Patch 3 fixes a typo
Patch 4 adds number of root objects directly into objagg stats.
Patches 5-7 do split of multiple structs in Spectrum TCAM code.
Patch 8 introduces initial implementation of ERP rehash logic,
         according to objagg hints.
Patch 9 adds hints priv passing trought the layers.
Patch 10 adds multi field into PAGT reg. (new patch)
Patch 11 implements actual region rules migration in TCAM code.
Patch 12 adds a devlink param so user is able to control
         rehash interval.
Patch 13 adds couple of tracepoints in order to track
         rehash procedures.
Patch 14 adds a simple selftest to test region rehash.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoselftests: mlxsw: spectrum-2: Add simple delta rehash test
Jiri Pirko [Thu, 7 Feb 2019 11:22:57 +0000 (11:22 +0000)]
selftests: mlxsw: spectrum-2: Add simple delta rehash test

Track the basic codepaths of delta rehash handling,
using mlxsw tracepoints.

Signed-off-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_acl: Add couple of vregion rehash tracepoints
Jiri Pirko [Thu, 7 Feb 2019 11:22:55 +0000 (11:22 +0000)]
mlxsw: spectrum_acl: Add couple of vregion rehash tracepoints

As vregion rehash is happening in delayed work, add some visibility to
the process using a few tracepoints.

Signed-off-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 "acl_region_rehash_interval" devlink param
Jiri Pirko [Thu, 7 Feb 2019 11:22:55 +0000 (11:22 +0000)]
mlxsw: spectrum: add "acl_region_rehash_interval" devlink param

Expose new driver-specific "acl_region_rehash_interval" devlink param
which would allow user to alter default ACL region rehash interval.

Signed-off-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_acl: Implement region migration according to hints
Jiri Pirko [Thu, 7 Feb 2019 11:22:54 +0000 (11:22 +0000)]
mlxsw: spectrum_acl: Implement region migration according to hints

If the hints are returned, the migration should be started. For that to
happen, there is a need to create a second physical region in TCAM with
new ERP set by passing the hints and then move chunk by chunk,
entry by entry.

During the transition, two lookups will occur. One in old region and
another in new region. The highest priority rule will be chosen.

In an unlikely case that the migration will fail and also rollback to
original region will fail the vregion will become in bad state.
Everything will work, only no future rehash will be possible. In a
follow-up work, this can be resolved by trying to resume the rollback
in delayed work and repair the vregion.

Signed-off-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 multi field to PAGT register
Jiri Pirko [Thu, 7 Feb 2019 11:22:53 +0000 (11:22 +0000)]
mlxsw: reg: Add multi field to PAGT register

For Spectrum-2 this allows parallel lookups in multiple regions.

Signed-off-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_acl: Pass hints priv all the way to ERP code
Jiri Pirko [Thu, 7 Feb 2019 11:22:52 +0000 (11:22 +0000)]
mlxsw: spectrum_acl: Pass hints priv all the way to ERP code

The hints priv comes from ERP code and it is possible to obtain it from
TCAM code. Add arg to appropriate functions so the hints
priv could be passed back down to ERP code. Pass NULL now as the
follow-up patches would pass an actual hints priv pointer.

Signed-off-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_acl: Implement basic ERP rehash hits creation
Jiri Pirko [Thu, 7 Feb 2019 11:22:51 +0000 (11:22 +0000)]
mlxsw: spectrum_acl: Implement basic ERP rehash hits creation

Introduce an initial implementation of rehash logic in ERP code.
Currently, the rehash is considered as needed only in case number of
roots in the hints is smaller than the number of roots actually in use.
In that case return hints pointer and let it be obtained through the
callpath through the Spectrum-2 TCAM op.

Signed-off-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_acl: Split entry struct into entry and ventry
Jiri Pirko [Thu, 7 Feb 2019 11:22:50 +0000 (11:22 +0000)]
mlxsw: spectrum_acl: Split entry struct into entry and ventry

Do the split of entry struct so the new entry struct is related to the
actual HW entry, whereas ventry struct is a SW abstration of that.
This split prepares possibility for ventry to hold 2 HW entries which
is needed for region ERP rehash flow.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>