- 14 Jul, 2023 40 commits
-
-
Petr Machata authored
The fid_get callback is called to allocate a FID for the newly-created RIF. In a following patch, the fid_get implementation for VLANs will be modified to take the VLAN ID from the parameters instead of deducing it from the netdevice. To that end, propagate the RIF parameters to the fid_get callback. Signed-off-by:
Petr Machata <petrm@nvidia.com> Reviewed-by:
Danielle Ratson <danieller@nvidia.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Petr Machata authored
Currently the reason for rejection of PVID manipulation is dumped to syslog, and a generic -EBUSY is returned to the userspace. But switchdev_handle_port_obj_add(), through which we get to mlxsw_sp_port_vlans_add(), handles extack just fine, and we can pass the message this way. This improves visibility into reasons why the request to change PVID was rejected. Before the change: # bridge vlan add dev br vid 2 self pvid untagged RTNETLINK answers: Device or resource busy (plus a syslog line) After the change: # bridge vlan add dev br vid 2 self pvid untagged Error: mlxsw_spectrum: Can't change PVID, it's used by router interface. Note that this particular error message is going away in the following patches. However the ability to pass error messages through extack will be useful more broadly for communicating in particular reasons why a RIF failed to be created. Signed-off-by:
-
David S. Miller authored
Sabrina Dubroca says: ==================== net: add MACsec offload selftests Patch 1 adds MACsec offload to netdevsim (unchanged from v2). Patch 2 adds a corresponding selftest to the rtnetlink testsuite. ==================== Signed-off-by: -
Sabrina Dubroca authored
Like the IPsec offload test, this requires netdevsim. Signed-off-by:
Sabrina Dubroca <sd@queasysnail.net> Signed-off-by:
-
Sabrina Dubroca authored
When the kernel is compiled with MACsec support, add the NETIF_F_HW_MACSEC feature to netdevsim devices and implement macsec_ops. To allow easy testing of failure from the device, support is limited to 3 SecY's per netdevsim device, and 1 RXSC per SecY. v2: - nsim_macsec_add_secy, return -ENOSPC if secy_count isn't full but we can't find an empty slot (Simon Horman) - add sci_to_cpu to make sparse happy (Simon Horman) - remove set but not used secy variable (kernel test robot and Simon Horman) Signed-off-by:
Simon Horman <simon.horman@corigine.com> Signed-off-by:
-
Jiri Pirko authored
The checks in question were introduced by: commit 6b4db2e5 ("devlink: Fix use-after-free after a failed reload"). That fixed an issue of reload with mlxsw driver. Back then, that was a valid fix, because there was a limitation in place that prevented drivers from registering/unregistering params when devlink instance was registered. It was possible to do the fix differently by changing drivers to register/unregister params in appropriate places making sure the ops operate only on memory which is allocated and initialized. But that, as a dependency, would require to remove the limitation mentioned above. Eventually, this limitation was lifted by: commit 1d18bb1a ("devlink: allow registering parameters after the instance") Also, the alternative fix (which also fixed another issue) was done by: commit 74cbc3c0 ("mlxsw: spectrum_acl_tcam: Move devlink param to TCAM code"). Therefore, the checks are no longer relevant. Each driver should make sure to have the params registered only when the memory the ops are working with is allocated and initialized. So remove the checks. Signed-off-by:
Jiri Pirko <jiri@nvidia.com> Reviewed-by:
Ido Schimmel <idosch@nvidia.com> Reviewed-by:
Jakub Kicinski <kuba@kernel.org> Signed-off-by:
-
David S. Miller authored
Russell King says: ==================== Convert mv88e6xxx to phylink_pcs This series (previously posted with further patches on the 26 June as RFC) converts mv88e6xxx to phylink_pcs, and thus moves it from being a pre-March 2020 legacy driver. The first four patches lay the ground-work for the conversion by adding four new methods to the phylink_pcs operations structure: pcs_enable() - called when the PCS is going to start to be used pcs_disable() - called when the PCS is no longer being used pcs_pre_config() - called before the MAC configuration method pcs_post_config() - called after the MAC configuration method Both of these are necessary for some of the mv88e639x workarounds. We also add the ability to inform phylink of a change to the PCS state without involving the MAC later, by providing phylink_pcs_change() which takes a phylink_pcs structure rather than a phylink structure. phylink maintains which instance the PCS is conencted to, so internally it can do the right thing when the PCS is in-use. Then we provide some additional mdiobus and mdiodev accessors that we will be using in the new PCS drivers. The changes for mv88e6xxx follow, and the first one needs to be explicitly pointed out - we (Andrew and myself) have both decided that all possible approaches to maintaining backwards compatibility with DT have been exhaused - everyone has some objection to everything that has been proposed. So, after many years of trying, we have decided that this is just an impossibility, and with this patch, we are now intentionally and knowingly breaking any DT that does not specify the CPU and DSA port fixed-link parameters. Hence why Andrew has recently been submitting DT update patches. It is regrettable that it has come to this. Following this, we start preparing 88e6xxx for phylink_pcs conversion by padding the mac_select_pcs() DSA method, and the internal hooks to create and tear-down PCS instances. Rather than bloat the already very large mv88e6xxx_ops structure, I decided that it would be better that the new internal chip specific PCS methods are all grouped within their own structure - and this structure can be declared in the PCS drivers themselves. Then we have the actual conversion patches, one for each family of PCS. Lastly, we clean up the driver after conversion, removing all the now redundant code. ==================== Signed-off-by: -
Russell King (Oracle) authored
Now that mv88e6xxx is completely converted to using phylink_pcs support, we have no need for the serdes methods. Remove all this infrastructure. Also remove the __maybe_unused from mv88e6xxx_pcs_select(). Signed-off-by:
Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by:
-
Russell King (Oracle) authored
Convert the 88E6390, 88E6390X, and 88E6393X family of switches to use the phylink_pcs infrastructure. Signed-off-by:
-
Russell King authored
Convert the 88E6352 SERDES code to use the phylink_pcs infrastructure. Signed-off-by:
-
Russell King (Oracle) authored
Convert the 88E6185 SERDES code to use the phylink_pcs infrastructure. Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Russell King (Oracle) authored
Rename and export the PCS state decoding function so our PCS can make use of the functionality provided by this. Reviewed-by:
-
Russell King (Oracle) authored
Add infrastructure for phylink_pcs to the mv88e6xxx driver. This involves adding a mac_select_pcs() hook so we can pass the PCS to phylink at the appropriate time, and a PCS initialisation function. As the various chip implementations are converted to use phylink_pcs, they are no longer reliant on the legacy phylink behaviour. We detect this by the use of this infrastructure, or the lack of any serdes. Signed-off-by:
-
Russell King (Oracle) authored
As we now always use a fixed-link for DSA and CPU ports, we no longer need the hack in the Marvell code to make this work. Remove it. This is especially important with the conversion of DSA drivers to phylink_pcs, as the PCS code only gets called if we are using phylink for the port. Signed-off-by:
-
Russell King (Oracle) authored
Add the following unlocked accessors to complete the set: __mdiobus_modify() __mdiodev_read() __mdiodev_write() __mdiodev_modify() __mdiodev_modify_changed() which we will need for Marvell DSA PCS conversion. Reviewed-by:
-
Russell King (Oracle) authored
Add a function, phylink_pcs_change() which can be used by PCs drivers to notify phylink about changes to the PCS link state. Signed-off-by:
-
Russell King (Oracle) authored
Add hooks that are called before and after the mac_config() call, which will be needed to deal with errata workarounds for the Marvell 88e639x DSA switches. Reviewed-by:
-
Russell King (Oracle) authored
Add phylink PCS enable/disable callbacks that will allow us to place IEEE 802.3 register compliant PCS in power-down mode while not being used. Signed-off-by:
-
Mengyuan Lou authored
Implement ethtool_ops get_wol and set_wol. Implement Wake-on-LAN support. Wol requires hardware board support which use sub id to identify. Magic packets are checked by fw, for now just support WAKE_MAGIC. Signed-off-by:
Mengyuan Lou <mengyuanlou@net-swift.com> Signed-off-by:
-
Mark Brown authored
We now have a regmap cache which uses a maple tree to store the register state, this is a more modern data structure and the regmap level code using it makes a number of assumptions better tuned for modern hardware than those made by the rbtree cache type that the at9331 driver uses. Switch the ar9331 driver to use the more modern data structure. This should have minimal practical impact, it's mainly code modernisation. Signed-off-by:
Mark Brown <broonie@kernel.org> Signed-off-by:
-
David S. Miller authored
Guillaume Nault says: ==================== net: Mark the sk parameter of routing functions as 'const'. The sk_getsecid security hook prevents the use of a const sk pointer in several routing functions. Since this hook should only read sk data, make its sk argument const (patch 1), then constify the sk parameter of various routing functions (patches 2-4). Build-tested with make allmodconfig. ==================== Signed-off-by: -
Guillaume Nault authored
Make it explicit that this function doesn't modify the socket passed as parameter. Signed-off-by:
Guillaume Nault <gnault@redhat.com> Reviewed-by:
-
Guillaume Nault authored
icmpv6_flow_init(), ip6_datagram_flow_key_init() and ip6_mc_hdr() don't need to modify their sk argument. Make that explicit using const. Signed-off-by:
David Ahern <dsahern@kernel.org> Signed-off-by:
-
Guillaume Nault authored
These functions don't need to modify the socket, so let's allow the callers to pass a const struct sock *. Signed-off-by:
-
Guillaume Nault authored
The sk_getsecid hook shouldn't need to modify its socket argument. Make it const so that callers of security_sk_classify_flow() can use a const struct sock *. Signed-off-by:
-
Jakub Kicinski authored
Bartosz Golaszewski says: ==================== net: stmmac: replace boolean fields in plat_stmmacenet_data with flags As suggested by Jose Abreu: let's drop all 12 boolean fields in plat_stmmacenet_data and replace them with a common bitfield. ==================== Link: https://lore.kernel.org/r/20230710090001.303225-1-brgl@bgdev.plSigned-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Reviewed-by:
Andrew Halaney <ahalaney@redhat.com> Link: https://lore.kernel.org/r/20230710090001.303225-13-brgl@bgdev.plReviewed-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
Drop the boolean field of the plat_stmmacenet_data structure in favor of a simple bitfield flag. Signed-off-by:
-
Bartosz Golaszewski authored
struct plat_stmmacenet_data contains several boolean fields that could be easily replaced with a common integer 'flags' bitfield and bit defines. Start the process with the has_integrated_pcs field. Signed-off-by:
-
Ziyang Chen authored
The dev_rx_discards counter will increment by one when an interface is toggled up and down. The main reason is that the driver first sends a `NFP_NET_CFG_CTRL_ENABLE` configuration packet to the NIC to perform port initialisation when an interface is set up. But there is a race between physical link up and free list queue initialization which may lead to the configuration packet being discarded. To address this problem a new bit NFP_NET_CFG_CTRL_FREELIST_EN is added to perform free list initialisation on the NIC. The FREELIST_EN should be sent in advance to initialize free list queue. When a port is set to down, FREELIST_EN should be sent after CTRL_ENABLE to avoid packet discards. Signed-off-by:
Ziyang Chen <ziyang.chen@corigine.com> Acked-by:
Louis Peens <louis.peens@corigine.com> Link: https://lore.kernel.org/r/20230712123551.13858-1-louis.peens@corigine.comSigned-off-by:
-
Jakub Kicinski authored
Matthieu Baerts says: ==================== selftests: mptcp: join: pass args in new env vars In this series, Geliang did some refactoring in the mptcp_join.sh file. Patch 1 reduces the scope of some global env vars, only used by some tests: easier to deal with. Patch 2 uses a dedicated env var for fastclose case instead of re-using addr_nr_ns2 with embedded info, clearer. Patch 3 is similar but for the fullmesh case. Patch 4 moves a positional but optional argument of run_tests() to an env var like it has already been done with the other args, cleaner. ==================== Link: https://lore.kernel.org/r/20230712-upstream-net-next-20230712-selftests-mptcp-use-local-env-v1-0-f1c8b62fbf95@tessares.netSigned-off-by:
-
