- 30 May, 2019 16 commits
-
-
Ivan Khoronzhuk authored
It's found while review and probably never happens, but real number of queues is set per device, and error path should be per device. So split error path based on usage_count. Signed-off-by:
Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Ioana Ciornei says: ==================== Decoupling PHYLINK from struct net_device Following two separate discussion threads in: https://www.spinics.net/lists/netdev/msg569087.html and: https://www.spinics.net/lists/netdev/msg570450.html Previous RFC patch set: https://www.spinics.net/lists/netdev/msg571995.html PHYLINK was reworked in order to accept multiple operation types, PHYLINK_NETDEV and PHYLINK_DEV, passed through a phylink_config structure alongside the corresponding struct device. One of the main concerns expressed in the RFC was that using notifiers to signal the corresponding phylink_mac_ops would break PHYLINK's API unity and that it would become harder to grep for its users. Using the current approach, we maintain a common API for all users. Also, printing useful information in PHYLINK, when decoupled from a net_device, is achieved using dev_err&co on the struct device received (in DSA's case is the device corresponding to the dsa_switch). PHYLIB (which PHYLINK uses) was reworked to the extent that it does not crash when connecting to a PHY and the net_device pointer is NULL. Lastly, DSA has been reworked in its way that it handles PHYs for ports that lack a net_device (CPU and DSA ports). For these, it was previously using PHYLIB and is now using the PHYLINK_DEV operation type. Previously, a driver that wanted to support PHY operations on CPU/DSA ports has to implement .adjust_link(). This patch set not only gives drivers the options to use PHYLINK uniformly but also urges them to convert to it. For compatibility, the old code is kept but it will be removed once all drivers switch over. The patchset was tested on the NXP LS1021A-TSN board having the following Ethernet layout: https://lkml.org/lkml/2019/5/5/279 The CPU port was moved from the internal RGMII fixed-link (enet2 -> switch port 4) to an external loopback Cat5 cable between the enet1 port and the front-facing swp2 SJA1105 port. In this mode, both the master and the CPU port have an attached PHY which detects link change events: [ 49.105426] fsl-gianfar soc:ethernet@2d50000 eth1: Link is Down [ 50.305486] sja1105 spi0.1: Link is Down [ 53.265596] fsl-gianfar soc:ethernet@2d50000 eth1: Link is Up - 1Gbps/Full - flow control off [ 54.466304] sja1105 spi0.1: Link is Up - 1Gbps/Full - flow control off Changes in v2: - fixed sparse warnings - updated 'Documentation/ABI/testing/sysfs-class-net-phydev' ==================== Signed-off-by:
-
Vladimir Oltean authored
PHYLIB and PHYLINK handle fixed-link interfaces differently. PHYLIB wraps them in a software PHY ("pseudo fixed link") phydev construct such that .adjust_link driver callbacks see an unified API. Whereas PHYLINK simply creates a phylink_link_state structure and passes it to .mac_config. At the time the driver was introduced, DSA was using PHYLIB for the CPU/cascade ports (the ones with no net devices) and PHYLINK for everything else. As explained below: commit aab9c406 Author: Florian Fainelli <f.fainelli@gmail.com> Date: Thu May 10 13:17:36 2018 -0700 net: dsa: Plug in PHYLINK support Drivers that utilize fixed links for user-facing ports (e.g: bcm_sf2) will need to implement phylink_mac_ops from now on to preserve functionality, since PHYLINK *does not* create a phy_device instance for fixed links. In the above patch, DSA guards the .phylink_mac_config callback against a NULL phydev pointer. Therefore, .adjust_link is not called in case of a fixed-link user port. This patch fixes the situation by converting the driver from using .adjust_link to .phylink_mac_config. This can be done now in a unified fashion for both slave and CPU/cascade ports because DSA now uses PHYLINK for all ports. Signed-off-by:Vladimir Oltean <olteanv@gmail.com> Signed-off-by:
Ioana Ciornei <ioana.ciornei@nxp.com> Reviewed-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
Ioana Ciornei authored
For DSA switches that do not have an .adjust_link callback, aka those who transitioned totally to the PHYLINK-compliant API, use PHYLINK to drive the CPU/DSA ports. The PHYLIB usage and .adjust_link are kept but deprecated, and users are asked to transition from it. The reason why we can't do anything for them is because PHYLINK does not wrap the fixed-link state behind a phydev object, so we cannot wrap .phylink_mac_config into .adjust_link unless we fabricate a phy_device structure. For these ports, the newly introduced PHYLINK_DEV operation type is used and the dsa_switch device structure is passed to PHYLINK for printing purposes. The handling of the PHYLINK_NETDEV and PHYLINK_DEV PHYLINK instances is common from the perspective of the driver. Signed-off-by:
-
Ioana Ciornei authored
In order to have a common handling of PHYLINK for the slave and non-user ports, the DSA core glue logic (between PHYLINK and the driver) must use an API that does not rely on a struct net_device. These will also be called by the CPU-port-handling code in a further patch. Signed-off-by:
-
Ioana Ciornei authored
With the latest addition to the PHYLINK infrastructure, we are faced with a decision on when to print necessary info using the struct net_device and when with the struct device. Add a series of macros that encapsulate this decision and replace all uses of netdev_err&co with phylink_err. Signed-off-by:
-
Ioana Ciornei authored
In the PHYLINK_DEV operation type, the PHYLINK infrastructure can work without an attached net_device. For printing usecases, instead, a struct device * should be passed to PHYLINK using the phylink_config structure. Also, netif_carrier_* calls ar guarded by the presence of a valid net_device. When using the PHYLINK_DEV operation type, we cannot check link status using the netif_carrier_ok() API so instead, keep an internal state of the MAC and call mac_link_{down,up} only when the link changed. Signed-off-by: -
Ioana Ciornei authored
The phylink_config structure will encapsulate a pointer to a struct device and the operation type requested for this instance of PHYLINK. This patch does not make any functional changes, it just transitions the PHYLINK internals and all its users to the new API. A pointer to a phylink_config structure will be passed to phylink_create() instead of the net_device directly. Also, the same phylink_config pointer will be passed back to all phylink_mac_ops callbacks instead of the net_device. Using this mechanism, a PHYLINK user can get the original net_device using a structure such as 'to_net_dev(config->dev)' or directly the structure containing the phylink_config using a container_of call. At the moment, only the PHYLINK_NETDEV is defined as a valid operation type for PHYLINK. In this mode, a valid reference to a struct device linked to the original net_device should be passed to PHYLINK through the phylink_config structure. This API changes is mainly driven by the necessity of adding a new operation type in PHYLINK that disconnects the phy_device from the net_device and also works when the net_device is lacking. Signed-off-by:
Maxime Chevallier <maxime.chevallier@bootlin.com> Tested-by:
-
Ioana Ciornei authored
This is a cosmetic patch that reduces the clutter in phylink_resolve around calling the .mac_link_up/.mac_link_down driver callbacks. In a further patch this logic will be extended to emit notifications in case a net device does not exist. Signed-off-by:
-
Ioana Ciornei authored
Export a phy_standalone device attribute that is meant to give the indication that this PHY lacks an attached_dev and its corresponding sysfs link. The attribute will be created only when the phy_attach_direct() function will be called with a NULL net_device. Signed-off-by:
-
Ioana Ciornei authored
In general, we don't want MAC drivers calling phy_attach_direct with the net_device being NULL. Add checks against this in all the functions calling it: phy_attach() and phy_connect_direct(). Signed-off-by:
Andrew Lunn <andrew@lunn.ch> Reviewed-by:
-
Ioana Ciornei authored
A prerequisite for PHYLIB to work in the absence of a struct net_device is to not access pointers to it. Changes are needed in the following areas: - Printing: In some places netdev_err was replaced with phydev_err. - Incrementing reference count to the parent MDIO bus driver: If there is no net device, then the reference count should definitely be incremented since there is no chance that it was an Ethernet driver who registered the MDIO bus. - Sysfs links are not created in case there is no attached_dev. - No netif_carrier_off is done if there is no attached_dev. Signed-off-by:
-
Vladimir Oltean authored
This is a cosmetic patch that wraps the operation of creating sysfs links between the netdev->phydev and the phydev->attached_dev. This is needed to keep the indentation level in check in a follow-up patch where this function will be guarded against the existence of a phydev->attached_dev. Signed-off-by:
-
Kevin 'ldir' Darbyshire-Bryant authored
ctinfo is a new tc filter action module. It is designed to restore information contained in firewall conntrack marks to other packet fields and is typically used on packet ingress paths. At present it has two independent sub-functions or operating modes, DSCP restoration mode & skb mark restoration mode. The DSCP restore mode: This mode copies DSCP values that have been placed in the firewall conntrack mark back into the IPv4/v6 diffserv fields of relevant packets. The DSCP restoration is intended for use and has been found useful for restoring ingress classifications based on egress classifications across links that bleach or otherwise change DSCP, typically home ISP Internet links. Restoring DSCP on ingress on the WAN link allows qdiscs such as but by no means limited to CAKE to shape inbound packets according to policies that are easier to set & mark on egress. Ingress classification is traditionally a challenging task since iptables rules haven't yet run and tc filter/eBPF programs are pre-NAT lookups, hence are unable to see internal IPv4 addresses as used on the typical home masquerading gateway. Thus marking the connection in some manner on egress for later restoration of classification on ingress is easier to implement. Parameters related to DSCP restore mode: dscpmask - a 32 bit mask of 6 contiguous bits and indicate bits of the conntrack mark field contain the DSCP value to be restored. statemask - a 32 bit mask of (usually) 1 bit length, outside the area specified by dscpmask. This represents a conditional operation flag whereby the DSCP is only restored if the flag is set. This is useful to implement a 'one shot' iptables based classification where the 'complicated' iptables rules are only run once to classify the connection on initial (egress) packet and subsequent packets are all marked/restored with the same DSCP. A mask of zero disables the conditional behaviour ie. the conntrack mark DSCP bits are always restored to the ip diffserv field (assuming the conntrack entry is found & the skb is an ipv4/ipv6 type) e.g. dscpmask 0xfc000000 statemask 0x01000000 |----0xFC----conntrack mark----000000---| | Bits 31-26 | bit 25 | bit24 |~~~ Bit 0| | DSCP | unused | flag |unused | |-----------------------0x01---000000---| | | | | ---| Conditional flag v only restore if set |-ip diffserv-| | 6 bits | |-------------| The skb mark restore mode (cpmark): This mode copies the firewall conntrack mark to the skb's mark field. It is completely the functional equivalent of the existing act_connmark action with the additional feature of being able to apply a mask to the restored value. Parameters related to skb mark restore mode: mask - a 32 bit mask applied to the firewall conntrack mark to mask out bits unwanted for restoration. This can be useful where the conntrack mark is being used for different purposes by different applications. If not specified and by default the whole mark field is copied (i.e. default mask of 0xffffffff) e.g. mask 0x00ffffff to mask out the top 8 bits being used by the aforementioned DSCP restore mode. |----0x00----conntrack mark----ffffff---| | Bits 31-24 | | | DSCP & flag| some value here | |---------------------------------------| | | v |------------skb mark-------------------| | | | | zeroed | | |---------------------------------------| Overall parameters: zone - conntrack zone control - action related control (reclassify | pipe | drop | continue | ok | goto chain <CHAIN_INDEX>) Signed-off-by:Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk> Reviewed-by:
Toke Høiland-Jørgensen <toke@redhat.com> Acked-by:
Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by:
-
Heiner Kallweit authored
MAC on the GBit versions supports 1000/Full only, however the PHY partially claims to support 1000/Half. So let's explicitly remove this mode. Signed-off-by:
Heiner Kallweit <hkallweit1@gmail.com> Reviewed-by:
-
Joergen Andreasen authored
Hardware offload of matchall classifier and police action are now supported via the tc command. Supported police parameters are: rate and burst. Example: Add: tc qdisc add dev eth3 handle ffff: ingress tc filter add dev eth3 parent ffff: prio 1 handle 2 \ matchall skip_sw \ action police rate 100Mbit burst 10000 Show: tc -s -d qdisc show dev eth3 tc -s -d filter show dev eth3 ingress Delete: tc filter del dev eth3 parent ffff: prio 1 tc qdisc del dev eth3 handle ffff: ingress Signed-off-by:
Joergen Andreasen <joergen.andreasen@microchip.com> Signed-off-by:
-
- 29 May, 2019 24 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queueDavid S. Miller authored
Jeff Kirsher says: ==================== 100GbE Intel Wired LAN Driver Updates 2019-05-29 This series contains updates to ice driver only. Bruce cleans up white space issues and fixes complaints about using bitop assignments using operands of different sizes. Anirudh cleans up code that is no longer needed now that the firmware supports the functionality. Adds support for ethtool selftestto the ice driver, which includes testing link, interrupts, eeprom, registers and packet loopback. Also, cleaned up duplicate code. Tony implements support for toggling receive VLAN filter via ethtool. Brett bumps up the minimum receive descriptor count per queue to resolve dropped packets. Refactored the interrupt tracking for the ice driver to resolve issues seen with the co-existence of features and SR-IOV, so instead of having a hardware IRQ tracker and a software IRQ tracker, simply use one tracker. Also adds a helper function to trigger software interrupts. Mitch changes how Malicious Driver Detection (MDD) events are handled, to ensure all VFs checked for MDD events and just log the event instead of disabling the VF, which was preventing proper release of resources if the VF is rebooted or the VF driver reloaded. Dave cleans up a redundant call to register LLDP MIB change events. Dan adds support to retrieve the current setting of firmware logging from the hardware to properly initialize the hardware structure. ==================== Signed-off-by: -
YueHaibing authored
Fix gcc build error while CONFIG_INET is not set drivers/net/ethernet/stmicro/stmmac/stmmac_selftests.o: In function `__stmmac_test_loopback': stmmac_selftests.c:(.text+0x8ec): undefined reference to `ip_send_check' stmmac_selftests.c:(.text+0xacc): undefined reference to `udp4_hwcsum' Add CONFIG_INET dependency to fix this. Reported-by:
Hulk Robot <hulkci@huawei.com> Fixes: 091810db ("net: stmmac: Introduce selftests support") Signed-off-by:
YueHaibing <yuehaibing@huawei.com> Signed-off-by:
-
Herbert Xu authored
This patch moves common code between rht_ptr and rht_ptr_exclusive into __rht_ptr. It also adds a new helper rht_ptr_rcu exclusively for the RCU case. This way rht_ptr becomes a lock-only construct so we can use the lighter rcu_dereference_protected primitive. Signed-off-by:
Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by:
-
Jisheng Zhang authored
Before the netdev is registered, calling netdev_info() will emit something as "(unnamed net device) (uninitialized)", looks confusing. Before this patch: [ 3.155028] stmmaceth f7b60000.ethernet (unnamed net_device) (uninitialized): device MAC address 52:1a:55:18:9e:9d After this patch: [ 3.155028] stmmaceth f7b60000.ethernet: device MAC address 52:1a:55:18:9e:9d Signed-off-by:
Jisheng Zhang <Jisheng.Zhang@synaptics.com> Signed-off-by:
-
Colin Ian King authored
There is a spelling mistake in a DP_INFO message. Fix it. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Acked-by: Michal Kalderon <michal.kalderon@marvell.com> Signed-off-by:
-
Brett Creeley authored
Add a new function ice_trigger_sw_intr to trigger interrupts. Signed-off-by:
Brett Creeley <brett.creeley@intel.com> Signed-off-by:
Anirudh Venkataramanan <anirudh.venkataramanan@intel.com> Tested-by:
Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com>
-
Md Fahad Iqbal Polash authored
Call ice_vsi_cfg_rss_lut_key only if RSS is enabled. Signed-off-by:
Md Fahad Iqbal Polash <md.fahad.iqbal.polash@intel.com> Signed-off-by:
-
Dan Nowlin authored
In order to initialize the current status of the FW logging, this patch adds ice_get_fw_log_cfg. The function retrieves the current setting of the FW logging from HW and updates the ice_hw structure accordingly. Signed-off-by:
Dan Nowlin <dan.nowlin@intel.com> Signed-off-by:
-
Anirudh Venkataramanan authored
Remove duplicate define for ICE_INVAL_Q_HANDLE. Move defines to the top of the file. Signed-off-by:
-
Dave Ertman authored
In the path for re-enabling FW LLDP engine, there is a call to register for LLDP MIB change events. This call is redundant, in that the call to ice_pf_dcb_cfg will already register the driver for these events. Also, the call as it stands now is too early in the flow before before DCB is configured. Remove the redundant call. Signed-off-by:
Dave Ertman <david.m.ertman@intel.com> Signed-off-by:
-
Mitch Williams authored
Change the message level of the MTU change log message from debug to info. Signed-off-by:
Mitch Williams <mitch.a.williams@intel.com> Signed-off-by:
-
Mitch Williams authored
Don't use the mdd_detected variable as an exit condition for this loop; the first VF to NOT have an MDD event will cause the loop to terminate. Instead just look at all of the VFs, but don't disable them. This prevents proper release of resources if the VFs are rebooted or the VF driver reloaded. Instead, just log a message and call out repeat offenders. To make it clear what we are doing, use a differently-named variable in the loop. Signed-off-by:
-
Brett Creeley authored
Currently we have two MSI-x (IRQ) trackers, one for OS requested MSI-x entries (sw_irq_tracker) and one for hardware MSI-x vectors (hw_irq_tracker). Generally the sw_irq_tracker has less entries than the hw_irq_tracker because the hw_irq_tracker has entries equal to the max allowed MSI-x per PF and the sw_irq_tracker is mainly the minimum (non SR-IOV portion of the vectors, kernel granted IRQs). All of the non SR-IOV portions of the driver (i.e. LAN queues, RDMA queues, OICR, etc.) take at least one of each type of tracker resource. SR-IOV only grabs entries from the hw_irq_tracker. There are a few issues with this approach that can be seen when doing any kind of device reconfiguration (i.e. ethtool -L, SR-IOV, etc.). One of them being, any time the driver creates an ice_q_vector and associates it to a LAN queue pair it will grab and use one entry from the hw_irq_tracker and one from the sw_irq_tracker. If the indices on these does not match it will cause a Tx timeout, which will cause a reset and then the indices will match up again and traffic will resume. The mismatched indices come from the trackers not being the same size and/or the search_hint in the two trackers not being equal. Another reason for the refactor is the co-existence of features with SR-IOV. If SR-IOV is enabled and the interrupts are taken from the end of the sw_irq_tracker then other features can no longer use this space because the hardware has now given the remaining interrupts to SR-IOV. This patch reworks how we track MSI-x vectors by removing the hw_irq_tracker completely and instead MSI-x resources needed for SR-IOV are determined all at once instead of per VF. This can be done because when creating VFs we know how many are wanted and how many MSI-x vectors each VF needs. This also allows us to start using MSI-x resources from the end of the PF's allowed MSI-x vectors so we are less likely to use entries needed for other features (i.e. RDMA, L2 Offload, etc). This patch also reworks the ice_res_tracker structure by removing the search_hint and adding a new member - "end". Instead of having a search_hint we will always search from 0. The new member, "end", will be used to manipulate the end of the ice_res_tracker (specifically sw_irq_tracker) during runtime based on MSI-x vectors needed by SR-IOV. In the normal case, the end of ice_res_tracker will be equal to the ice_res_tracker's num_entries. The sriov_base_vector member was added to the PF structure. It is used to represent the starting MSI-x index of all the needed MSI-x vectors for all SR-IOV VFs. Depending on how many MSI-x are needed, SR-IOV may have to take resources from the sw_irq_tracker. This is done by setting the sw_irq_tracker->end equal to the pf->sriov_base_vector. When all SR-IOV VFs are removed then the sw_irq_tracker->end is reset back to sw_irq_tracker->num_entries. The sriov_base_vector, along with the VF's number of MSI-x (pf->num_vf_msix), vf_id, and the base MSI-x index on the PF (pf->hw.func_caps.common_cap.msix_vector_first_id), is used to calculate the first HW absolute MSI-x index for each VF, which is used to write to the VPINT_ALLOC[_PCI] and GLINT_VECT2FUNC registers to program the VFs MSI-x PCI configuration bits. Also, the sriov_base_vector is used along with VF's num_vf_msix, vf_id, and q_vector->v_idx to determine the MSI-x register index (used for writing to GLINT_DYN_CTL) within the PF's space. Interrupt changes removed any references to hw_base_vector, hw_oicr_idx, and hw_irq_tracker. Only sw_base_vector, sw_oicr_idx, and sw_irq_tracker variables remain. Change all of these by removing the "sw_" prefix to help avoid confusion with these variables and their use. Signed-off-by:
-
Anirudh Venkataramanan authored
This patch adds a handler for ethtool selftest. Selftest includes testing link, interrupts, eeprom, registers and packet loopback. Signed-off-by:
-
Brett Creeley authored
ice_cfg_itr() sets the ITR granularity and default ITR values for the PF's interrupt vectors. For VF's this will be done in the AVF driver flow. Fix this by not calling ice_cfg_itr() for SR-IOV. Signed-off-by:
-
Brett Creeley authored
Currently we set the default number of Rx descriptors per queue to the system's page size divided by the number of bytes per descriptor. For 4K page size systems this is resulting in 128 Rx descriptors per queue. This is causing more dropped packets than desired in the default configuration. Fix this by setting the minimum default Rx descriptor count per queue to 512. Signed-off-by:
-
Bruce Allan authored
Some static analysis tools can complain when doing a bitop assignment using operands of different sizes. Fix that. Signed-off-by:
Bruce Allan <bruce.w.allan@intel.com> Signed-off-by:
-
Tony Nguyen authored
Implement the toggling of rx-vlan-filter; enable|disable VLAN pruning based on on|off, respectively. Signed-off-by:
Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by:
-
Anirudh Venkataramanan authored
Clear PXE mode AQ call (opcode 0x0110) is now supported in FW. So remove the direct register write to GLLAN_RCTL_0. Signed-off-by:
-
Bruce Allan authored
Fix a checkpatch "LINE_SPACING: Please don't use multiple blank lines" issue that has snuck in to the code. Signed-off-by:
-
David S. Miller authored
Sudarsana Reddy Kalluru says: ==================== qed*: Fix inifinite spinning of PTP poll thread. The patch series addresses an error scenario in the PTP Tx implementation. Please consider applying it to net-next. ==================== Signed-off-by: -
Sudarsana Reddy Kalluru authored
In PTP Tx implementation, driver kept scheduling a poll thread until the timestamp is available. In the error scenarios (e.g. app requesting the timestamp for non-ptp packet), this thread kept waiting for the timestamp forever. This patch add changes to report such scenario as an error and terminate the thread. Added a timeout of 2 seconds i.e., max time to wait for Tx timestamp. Added a stat value ptp_skip_txts for reporting the number of packets for which Tx timestamping is skipped. Signed-off-by:
Sudarsana Reddy Kalluru <skalluru@marvell.com> Signed-off-by:
Michal Kalderon <mkalderon@marvell.com> Signed-off-by:
-
Sudarsana Reddy Kalluru authored
PTP Tx implementation continuously polls for the availability of timestamp. Reducing the severity of a debug message in this path to avoid filling up the syslog buffer with this message, especially in the error scenarios. Signed-off-by:
-
Gustavo A. R. Silva authored
The strncpy() function is being deprecated. Replace it by the safer strscpy() and fix the following Coverity warning: "Calling strncpy with a maximum size argument of 16 bytes on destination array ifrr.ifr_ifrn.ifrn_name of size 16 bytes might leave the destination string unterminated." Notice that, unlike strncpy(), strscpy() always null-terminates the destination string. Addresses-Coverity-ID: 1445537 ("Buffer not null terminated") Signed-off-by:Gustavo A. R. Silva <gustavo@embeddedor.com> Signed-off-by:
-
