- 06 Nov, 2020 40 commits
-
-
Jakub Kicinski authored
Ido Schimmel says: ==================== nexthop: Add support for nexthop objects offload This patch set adds support for nexthop objects offload with a dummy implementation over netdevsim. mlxsw support will be added later. The general idea is very similar to route offload in that notifications are sent whenever nexthop objects are changed. A listener can veto the change and the error will be communicated to user space with extack. To keep listeners as simple as possible, they not only receive notifications for the nexthop object that is changed, but also for all the other objects affected by this change. For example, when a single nexthop is replaced, a replace notification is sent for the single nexthop, but also for all the nexthop groups this nexthop is member in. This relieves listeners from the need to track such dependencies. To simplify things further for listeners, the notification info does not contain the raw nexthop data structures (e.g., 'struct nexthop'), but less complex data structures into which the raw data structures are parsed into. Tested with a new selftest over netdevsim and with fib_nexthops.sh: Tests passed: 164 Tests failed: 0 Patch set overview: Patches #1-#4 introduce the aforementioned data structures and convert existing listeners (i.e., the VXLAN driver) to use them. Patches #5-#6 add a new RTNH_F_TRAP flag and the ability to set it and RTNH_F_OFFLOAD on nexthops. This flag is used by netdevsim for testing purposes and will also be used by mlxsw. These flags are consistent with the existing RTM_F_OFFLOAD and RTM_F_TRAP flags. Patches #7-#14 gradually add the new nexthop notifications. Patches #15-#18 add a dummy implementation for nexthop offload over netdevsim and a selftest to exercise both good and bad flows. Changes since RFC [1]: Patch #1: s/is_encap/has_encap/ Patch #3: Add a blank line in __nh_notifier_single_info_init() Patch #5: Reword commit message Patch #6: s/nexthop_hw_flags_set/nexthop_set_hw_flags/ Patch #7: Reword commit message Patch #11: Allocate extack on the stack Follow-up patch sets: selftests: forwarding: Add nexthop objects tests mlxsw: Preparations for nexthop objects support - part 1/2 mlxsw: Preparations for nexthop objects support - part 2/2 mlxsw: Add support for nexthop objects mlxsw: Add support for blackhole nexthops mlxsw: Update adjacency index more efficiently [1] https://lore.kernel.org/netdev/20200908091037.2709823-1-idosch@idosch.org/ ==================== Link: https://lore.kernel.org/r/20201104133040.1125369-1-idosch@idosch.orgSigned-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Ido Schimmel authored
Test various aspects of the nexthop offload API on top of the netdevsim implementation. Both good and bad flows are tested. Signed-off-by:
Ido Schimmel <idosch@nvidia.com> Signed-off-by:
-
Ido Schimmel authored
Previous patches added the ability to program nexthop objects. Therefore, no longer forbid the programming of routes that point to such objects. Signed-off-by:
David Ahern <dsahern@gmail.com> Signed-off-by:
-
Ido Schimmel authored
Implement dummy nexthop "offload" in the driver by storing currently "programmed" nexthops in a hash table. Each nexthop in the hash table is marked with "trap" indication and increments the nexthops resource occupancy. This will later allow us to test the nexthop offload API on top of netdevsim. Signed-off-by:
-
Ido Schimmel authored
The Spectrum ASIC has a dedicated table where nexthops (i.e., adjacency entries) are populated. The size of this table can be controlled via devlink-resource. Add such a resource to netdevsim so that its occupancy will reflect the number of nexthop objects currently programmed to the device. By limiting the size of the resource, error paths could be exercised and tested. Example output: # devlink resource show netdevsim/netdevsim10 netdevsim/netdevsim10: name IPv4 size unlimited unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none resources: name fib size unlimited occ 4 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none name fib-rules size unlimited occ 3 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none name IPv6 size unlimited unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none resources: name fib size unlimited occ 1 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none name fib-rules size unlimited occ 2 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none name nexthops size unlimited occ 0 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none Signed-off-by: -
Ido Schimmel authored
Remove in-kernel route notifications when the configuration of their nexthop changes. These notifications are unnecessary because the route still uses the same nexthop ID. A separate notification for the nexthop change itself is now sent in the nexthop notification chain. Signed-off-by:
-
Ido Schimmel authored
When registering a new notifier to the nexthop notification chain, replay all the existing nexthops to the new notifier so that it will have a complete picture of the available nexthops. Signed-off-by:
-
Ido Schimmel authored
This will be used by the next patch which extends the function to replay all the existing nexthops to the notifier block being registered. Device drivers will be able to pass extack to the function since it is passed to them upon reload from devlink. Signed-off-by:
-
Ido Schimmel authored
When a single nexthop is deleted, the configuration of all the groups using the nexthop is effectively modified. In this case, emit a notification in the nexthop notification chain for each modified group so that listeners would not need to keep track of which nexthops are member in which groups. In the rare cases where the notification fails, emit an error to the kernel log. This is done by allocating extack on the stack and printing the error logged by the listener that rejected the notification. Changes since RFC: * Allocate extack on the stack Signed-off-by:
-
Ido Schimmel authored
When a single nexthop is replaced, the configuration of all the groups using the nexthop is effectively modified. In this case, emit a notification in the nexthop notification chain for each modified group so that listeners would not need to keep track of which nexthops are member in which groups. The notification can only be emitted after the new configuration (i.e., 'struct nh_info') is pointed at by the old shell (i.e., 'struct nexthop'). Before that the configuration of the nexthop groups is still the same as before the replacement. Signed-off-by:
-
Ido Schimmel authored
The notification is emitted after all the validation checks were performed, but before the new configuration (i.e., 'struct nh_info') is pointed at by the old shell (i.e., 'struct nexthop'). This prevents the need to perform rollback in case the notification is vetoed. The next patch will also emit a replace notification for all the nexthop groups in which the nexthop is used. Signed-off-by:
-
Ido Schimmel authored
Emit a notification in the nexthop notification chain when an existing nexthop group is replaced. The notification is emitted after all the validation checks were performed, but before the new configuration (i.e., 'struct nh_grp') is pointed at by the old shell (i.e., 'struct nexthop'). This prevents the need to perform rollback in case the notification is vetoed. Signed-off-by:
-
Ido Schimmel authored
Emit a notification in the nexthop notification chain when a new nexthop is added (not replaced). The nexthop can either be a new group or a single nexthop. The notification is sent after the nexthop is inserted into the red-black tree, as listeners might need to callback into the nexthop code with the nexthop ID in order to mark the nexthop as offloaded. A 'REPLACE' notification is emitted instead of 'ADD' as the distinction between the two is not important for in-kernel listeners. In case the listener is not familiar with the encoded nexthop ID, it can simply treat it as a new one. This is also consistent with the route offload API. Changes since RFC: * Reword commit message Signed-off-by:
-
Ido Schimmel authored
Add a function that can be called by device drivers to set "offload" or "trap" indication on nexthops following nexthop notifications. Changes since RFC: * s/nexthop_hw_flags_set/nexthop_set_hw_flags/ Signed-off-by:
-
Ido Schimmel authored
The flag indicates to user space that the nexthop is not programmed to forward packets in hardware, but rather to trap them to the CPU. This is needed, for example, when the MAC of the nexthop neighbour is not resolved and packets should reach the CPU to trigger neighbour resolution. The flag will be used in subsequent patches by netdevsim to test nexthop objects programming to device drivers and in the future by mlxsw as well. Changes since RFC: * Reword commit message Signed-off-by:
-
Ido Schimmel authored
Convert the sole listener of the nexthop notification chain (the VXLAN driver) to the new notification info. Signed-off-by:
-
Ido Schimmel authored
Prepare the new notification information so that it could be passed to listeners in the new patch. Changes since RFC: * Add a blank line in __nh_notifier_single_info_init() Signed-off-by:
-
Ido Schimmel authored
The next patch will add extack to the notification info. This allows listeners to veto notifications and communicate the reason to user space. Signed-off-by:
-
Ido Schimmel authored
Add data structures that will be used for nexthop replace and delete notifications in the previously introduced nexthop notification chain. New data structures are added instead of passing the existing nexthop code structures directly for several reasons. First, the existing structures encode a lot of bookkeeping information which is irrelevant for listeners of the notification chain. Second, the existing structures can be changed without worrying about introducing regressions in listeners since they are not accessed directly by them. Third, listeners of the notification chain do not need to each parse the relatively complex nexthop code structures. They are passing the required information in a simplified way. Note that a single 'has_encap' bit is added instead of the actual encapsulation information since current listeners do not support such nexthops. Changes since RFC: * s/is_encap/has_encap/ Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxJakub Kicinski authored
Saeed Mahameed says: ==================== mlx5-updates-2020-11-03 This series includes updates to mlx5 software steering component. 1) Few improvements in the DR area, such as removing unneeded checks, renaming to better general names, refactor in some places, etc. 2) Software steering (DR) Memory management improvements This patch series contains SW Steering memory management improvements: using buddy allocator instead of an existing bucket allocator, and several other optimizations. The buddy system is a memory allocation and management algorithm that manages memory in power of two increments. The algorithm is well-known and well-described, such as here: https://en.wikipedia.org/wiki/Buddy_memory_allocation Linux uses this algorithm for managing and allocating physical pages, as described here: https://www.kernel.org/doc/gorman/html/understand/understand009.html In our case, although the algorithm in principal is similar to the Linux physical page allocator, the "building blocks" and the circumstances are different: in SW steering, buddy allocator doesn't really allocates a memory, but rather manages ICM (Interconnect Context Memory) that was previously allocated and registered. The ICM memory that is used in SW steering is always power of 2 (order), so buddy system is a good fit for this. Patches in this series: [PATH 4] net/mlx5: DR, Add buddy allocator utilities This patch adds a modified implementation of a well-known buddy allocator, adjusted for SW steering needs: the algorithm in principal is similar to the Linux physical page allocator, but in our case buddy allocator doesn't really allocate a memory, but rather manages ICM memory that was previously allocated and registered. [PATH 5] net/mlx5: DR, Handle ICM memory via buddy allocation instead of bucket management This patch changes ICM management of SW steering to use buddy-system mechanism Instead of the previous bucket management. [PATH 6] net/mlx5: DR, Sync chunks only during free This patch makes syncing happen only when freeing memory chunks. [PATH 7] net/mlx5: DR, ICM memory pools sync optimization This patch adds tracking of pool's "hot" memory and makes the check whether steering sync is required much shorter and faster. [PATH 8] net/mlx5: DR, Free buddy ICM memory if it is unused This patch adds tracking buddy's used ICM memory, and frees the buddy if all its memory becomes unused. 3) Misc code cleanups * tag 'mlx5-updates-2020-11-03' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux: net: mlx5: Replace in_irq() usage net/mlx5: Cleanup kernel-doc warnings net/mlx4: Cleanup kernel-doc warnings net/mlx5e: Validate stop_room size upon user input net/mlx5: DR, Free unused buddy ICM memory net/mlx5: DR, ICM memory pools sync optimization net/mlx5: DR, Sync chunks only during free net/mlx5: DR, Handle ICM memory via buddy allocation instead of buckets net/mlx5: DR, Add buddy allocator utilities net/mlx5: DR, Rename matcher functions to be more HW agnostic net/mlx5: DR, Rename builders HW specific names net/mlx5: DR, Remove unused member of action struct ==================== Link: https://lore.kernel.org/r/20201105201242.21716-1-saeedm@nvidia.comSigned-off-by:
-
Hayes Wang authored
Support ECM mode based on cdc_ether with relative mii functions, when CONFIG_USB_RTL8152 is not set, or the device is not supported by r8152 driver. Both r8152 and r8153_ecm would check the return value of rtl8152_get_version() in porbe(). If rtl8152_get_version() return none zero value, the r8152 is used for the device with vendor mode. Otherwise, the r8153_ecm is used for the device with ECM mode. Signed-off-by:
Hayes Wang <hayeswang@realtek.com> Link: https://lore.kernel.org/r/1394712342-15778-392-Taiwan-albertk@realtek.comSigned-off-by:
-
Loic Poulain authored
This patch adds a new network driver implementing MHI transport for network packets. Packets can be in any format, though QMAP (rmnet) is the usual protocol (flow control + PDN mux). It support two MHI devices, IP_HW0 which is, the path to the IPA (IP accelerator) on qcom modem, And IP_SW0 which is the software driven IP path (to modem CPU). Signed-off-by:
Loic Poulain <loic.poulain@linaro.org> Reviewed-by:
Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> Link: https://lore.kernel.org/r/1604424234-24446-2-git-send-email-loic.poulain@linaro.orgSigned-off-by:
-
Loic Poulain authored
This function can be used by client driver to determine whether it's possible to queue new elements in a channel ring. Signed-off-by:
-
Jakub Kicinski authored
Ioana Ciornei says: ==================== net: phy: add support for shared interrupts (part 1) This patch set aims to actually add support for shared interrupts in phylib and not only for multi-PHY devices. While we are at it, streamline the interrupt handling in phylib. For a bit of context, at the moment, there are multiple phy_driver ops that deal with this subject: - .config_intr() - Enable/disable the interrupt line. - .ack_interrupt() - Should quiesce any interrupts that may have been fired. It's also used by phylib in conjunction with .config_intr() to clear any pending interrupts after the line was disabled, and before it is going to be enabled. - .did_interrupt() - Intended for multi-PHY devices with a shared IRQ line and used by phylib to discern which PHY from the package was the one that actually fired the interrupt. - .handle_interrupt() - Completely overrides the default interrupt handling logic from phylib. The PHY driver is responsible for checking if any interrupt was fired by the respective PHY and choose accordingly if it's the one that should trigger the link state machine. From my point of view, the interrupt handling in phylib has become somewhat confusing with all these callbacks that actually read the same PHY register - the interrupt status. A more streamlined approach would be to just move the responsibility to write an interrupt handler to the driver (as any other device driver does) and make .handle_interrupt() the only way to deal with interrupts. Another advantage with this approach would be that phylib would gain support for shared IRQs between different PHY (not just multi-PHY devices), something which at the moment would require extending every PHY driver anyway in order to implement their .did_interrupt() callback and duplicate the same logic as in .ack_interrupt(). The disadvantage of making .did_interrupt() mandatory would be that we are slightly changing the semantics of the phylib API and that would increase confusion instead of reducing it. What I am proposing is the following: - As a first step, make the .ack_interrupt() callback optional so that we do not break any PHY driver amid the transition. - Every PHY driver gains a .handle_interrupt() implementation that, for the most part, would look like below: irq_status = phy_read(phydev, INTR_STATUS); if (irq_status < 0) { phy_error(phydev); return IRQ_NONE; } if (!(irq_status & irq_mask)) return IRQ_NONE; phy_trigger_machine(phydev); return IRQ_HANDLED; - Remove each PHY driver's implementation of the .ack_interrupt() by actually taking care of quiescing any pending interrupts before enabling/after disabling the interrupt line. - Finally, after all drivers have been ported, remove the .ack_interrupt() and .did_interrupt() callbacks from phy_driver. This patch set is part 1 and it addresses the changes needed in phylib and 7 PHY drivers. The rest can be found on my Github branch here: https://github.com/IoanaCiornei/linux/commits/phylib-shared-irq I do not have access to most of these PHY's, therefore I Cc-ed the latest contributors to the individual PHY drivers in order to have access, hopefully, to more regression testing. Changes in v2: - Rework the .handle_interrupt() implementation for each driver so that only the enabled interrupts are taken into account when IRQ_NONE/IRQ_HANDLED it returned. The main idea is so that we avoid falsely blaming a device for triggering an interrupt when this is not the case. The only devices for which I was unable to make this adjustment were the BCM8706, BCM8727, BCMAC131 and BCM5241 since I do not have access to their datasheets. - I also updated the pseudo-code added in the cover-letter so that it's more clear how a .handle_interrupt() callback should look like. ==================== Tested-by:Andrew Lunn <andrew@lunn.ch> Link: https://lore.kernel.org/r/20201101125114.1316879-1-ciorneiioana@gmail.comSigned-off-by:
-
Ioana Ciornei authored
In preparation of removing the .ack_interrupt() callback, we must replace its occurrences (aka phy_clear_interrupt), from the 2 places where it is called from (phy_enable_interrupts and phy_disable_interrupts), with equivalent functionality. This means that clearing interrupts now becomes something that the PHY driver is responsible of doing, before enabling interrupts and after clearing them. Make this driver follow the new contract. Cc: Martin Blumenstingl <martin.blumenstingl@googlemail.com> Cc: Willy Liu <willy.liu@realtek.com> Signed-off-by:
Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by:
-
Ioana Ciornei authored
In an attempt to actually support shared IRQs in phylib, we now move the responsibility of triggering the phylib state machine or just returning IRQ_NONE, based on the IRQ status register, to the PHY driver. Having 3 different IRQ handling callbacks (.handle_interrupt(), .did_interrupt() and .ack_interrupt() ) is confusing so let the PHY driver implement directly an IRQ handler like any other device driver. Make this driver follow the new convention. Cc: Martin Blumenstingl <martin.blumenstingl@googlemail.com> Cc: Willy Liu <willy.liu@realtek.com> Signed-off-by:
-
Ioana Ciornei authored
It seems there are cases where the interrupts are handled by another entity (ie an IRQ controller embedded inside the PHY) and do not need any other interraction from phylib. For this kind of PHYs, like the RTL8366RB, add the genphy_handle_interrupt_no_ack() function which just triggers the link state machine. Signed-off-by:
-
Ioana Ciornei authored
In preparation of removing the .ack_interrupt() callback, we must replace its occurrences (aka phy_clear_interrupt), from the 2 places where it is called from (phy_enable_interrupts and phy_disable_interrupts), with equivalent functionality. This means that clearing interrupts now becomes something that the PHY driver is responsible of doing, before enabling interrupts and after clearing them. Make this driver follow the new contract. Signed-off-by:
-
Ioana Ciornei authored
In an attempt to actually support shared IRQs in phylib, we now move the responsibility of triggering the phylib state machine or just returning IRQ_NONE, based on the IRQ status register, to the PHY driver. Having 3 different IRQ handling callbacks (.handle_interrupt(), .did_interrupt() and .ack_interrupt() ) is confusing so let the PHY driver implement directly an IRQ handler like any other device driver. Make this driver follow the new convention. Signed-off-by:
-
Ioana Ciornei authored
In preparation of removing the .ack_interrupt() callback, we must replace its occurrences (aka phy_clear_interrupt), from the 2 places where it is called from (phy_enable_interrupts and phy_disable_interrupts), with equivalent functionality. This means that clearing interrupts now becomes something that the PHY driver is responsible of doing, before enabling interrupts and after clearing them. Make this driver follow the new contract. Signed-off-by:
-
Ioana Ciornei authored
In an attempt to actually support shared IRQs in phylib, we now move the responsibility of triggering the phylib state machine or just returning IRQ_NONE, based on the IRQ status register, to the PHY driver. Having 3 different IRQ handling callbacks (.handle_interrupt(), .did_interrupt() and .ack_interrupt() ) is confusing so let the PHY driver implement directly an IRQ handler like any other device driver. Make this driver follow the new convention. Signed-off-by:
-
Ioana Ciornei authored
In preparation of removing the .ack_interrupt() callback, we must replace its occurrences (aka phy_clear_interrupt), from the 2 places where it is called from (phy_enable_interrupts and phy_disable_interrupts), with equivalent functionality. This means that clearing interrupts now becomes something that the PHY driver is responsible of doing, before enabling interrupts and after clearing them. Make this driver follow the new contract. Cc: Michael Walle <michael@walle.cc> Cc: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
Ioana Ciornei authored
In an attempt to actually support shared IRQs in phylib, we now move the responsibility of triggering the phylib state machine or just returning IRQ_NONE, based on the IRQ status register, to the PHY driver. Having 3 different IRQ handling callbacks (.handle_interrupt(), .did_interrupt() and .ack_interrupt() ) is confusing so let the PHY driver implement directly an IRQ handler like any other device driver. Make this driver follow the new convention. Cc: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
Michael Walle <michael@walle.cc> Signed-off-by:
-
Ioana Ciornei authored
In preparation of removing the .ack_interrupt() callback, we must replace its occurrences (aka phy_clear_interrupt), from the 2 places where it is called from (phy_enable_interrupts and phy_disable_interrupts), with equivalent functionality. This means that clearing interrupts now becomes something that the PHY driver is responsible of doing, before enabling interrupts and after clearing them. Make this driver follow the new contract. Cc: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by:
-
Ioana Ciornei authored
In an attempt to actually support shared IRQs in phylib, we now move the responsibility of triggering the phylib state machine or just returning IRQ_NONE, based on the IRQ status register, to the PHY driver. Having 3 different IRQ handling callbacks (.handle_interrupt(), .did_interrupt() and .ack_interrupt() ) is confusing so let the PHY driver implement directly an IRQ handler like any other device driver. Make this driver follow the new convention. Cc: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by:
-
Ioana Ciornei authored
In preparation of removing the .ack_interrupt() callback, we must replace its occurrences (aka phy_clear_interrupt), from the 2 places where it is called from (phy_enable_interrupts and phy_disable_interrupts), with equivalent functionality. This means that clearing interrupts now becomes something that the PHY driver is responsible of doing, before enabling interrupts and after clearing them. Make this driver follow the new contract. Cc: Antoine Tenart <atenart@kernel.org> Signed-off-by:
-
Ioana Ciornei authored
In an attempt to actually support shared IRQs in phylib, we now move the responsibility of triggering the phylib state machine or just returning IRQ_NONE, based on the IRQ status register, to the PHY driver. Having 3 different IRQ handling callbacks (.handle_interrupt(), .did_interrupt() and .ack_interrupt() ) is confusing so let the PHY driver implement directly an IRQ handler like any other device driver. Make this driver follow the new convention. Also, remove the .did_interrupt() callback since it's not anymore used. Signed-off-by:
-
Ioana Ciornei authored
According to the comment describing the phy_mac_interrupt() function, it it intended to be used by MAC drivers which have noticed a link change thus its use in the mscc PHY driver is improper and, most probably, was added just because phy_trigger_machine() was not exported. Now that we have acces to trigger the link state machine, use directly the phy_trigger_machine() function to notify a link change detected by the PHY driver. Signed-off-by:
Vladimir Oltean <olteanv@gmail.com> Signed-off-by:
-
Ioana Ciornei authored
In preparation of removing the .ack_interrupt() callback, we must replace its occurrences (aka phy_clear_interrupt), from the 2 places where it is called from (phy_enable_interrupts and phy_disable_interrupts), with equivalent functionality. This means that clearing interrupts now becomes something that the PHY driver is responsible of doing, before enabling interrupts and after clearing them. Make this driver follow the new contract. Cc: Oleksij Rempel <o.rempel@pengutronix.de> Cc: Michael Walle <michael@walle.cc> Signed-off-by:
Oleksij Rempel <o.rempel@pengutronix.de> Signed-off-by:
-
Ioana Ciornei authored
In an attempt to actually support shared IRQs in phylib, we now move the responsibility of triggering the phylib state machine or just returning IRQ_NONE, based on the IRQ status register, to the PHY driver. Having 3 different IRQ handling callbacks (.handle_interrupt(), .did_interrupt() and .ack_interrupt() ) is confusing so let the PHY driver implement directly an IRQ handler like any other device driver. Make this driver follow the new convention. Cc: Oleksij Rempel <o.rempel@pengutronix.de> Cc: Michael Walle <michael@walle.cc> Signed-off-by:
-
