- 26 Feb, 2020 10 commits
-
-
Maxim Mikityanskiy authored
The current steps that are performed when the trust state changes, if the channels are active: 1. The trust state is changed in hardware. 2. The new inline mode is calculated. 3. If the new inline mode is different, the channels are recreated using the new inline mode. This approach has some issues: 1. There is a time gap between changing trust state in hardware and starting sending enough inline headers (the latter happens after recreation of channels). It leads to failed transmissions and error CQEs. 2. If the new channels fail to open, we'll be left with the old ones, but the hardware will be configured for the new trust state, so the interval when we can see TX errors never ends. This patch fixes the issues above by moving the trust state change into the preactivate hook that runs during the recreation of the channels when no channels are active, so it eliminates the gap of partially applied configuration. If the inline mode doesn't change with the change of the trust state, the channels won't be recreated, just like before this patch. Signed-off-by:
Maxim Mikityanskiy <maximmi@mellanox.com> Reviewed-by:
Tariq Toukan <tariqt@mellanox.com> Signed-off-by:
Saeed Mahameed <saeedm@mellanox.com>
-
Maxim Mikityanskiy authored
Sometimes the preactivate hook of mlx5e_safe_switch_channels needs more parameters than just struct mlx5e_priv *. For such cases, a new parameter (void *context) is added to preactivate hooks. Some of the existing normal functions are currently used as preactivate callbacks. To avoid adding an extra unused parameter, they are wrapped in an automatic way using the MLX5E_DEFINE_PREACTIVATE_WRAPPER_CTX macro. Signed-off-by:
-
Maxim Mikityanskiy authored
Currently mlx5e_switch_priv_channels expects that the preactivate hook doesn't fail, however, it can fail, because it may set hardware parameters. This commit addresses this issue and provides a way to recover from failures of the preactivate hook: the old channels are not closed until the point where nothing can fail anymore, so in case preactivate fails, the driver can roll back the old channels and activate them again. Signed-off-by:
-
Maxim Mikityanskiy authored
The number of queues is now updated by mlx5e_update_netdev_queues in a centralized way, when no channels are active. Remove an extra occurrence of netif_set_real_num_tx_queues to prepare it for the next commit. Signed-off-by:
-
Maxim Mikityanskiy authored
Currently, mlx5e notifies the kernel about the number of queues and sets the default XPS cpumasks when channels are activated. This implementation has several corner cases, in which the kernel may not be updated on time, or XPS cpumasks may be reset when not directly touched by the user. This commit fixes these corner cases to match the following expected behavior: 1. The number of queues always corresponds to the number of channels configured. 2. XPS cpumasks are set to driver's defaults on netdev attach. 3. XPS cpumasks set by user are not reset, unless the number of channels changes. If the number of channels changes, they are reset to driver's defaults. (In general case, when the number of channels increases or decreases, it's not possible to guess how to convert the current XPS cpumasks to work with the new number of channels, so we let the user reconfigure it if they change the number of channels.) XPS cpumasks are no longer stored per channel. Only one temporary cpumask is used. The old stored cpumasks didn't reflect the user's changes and were not used after applying them. A scratchpad area is added to struct mlx5e_priv. As cpumask_var_t requires allocation, and the preactivate hook can't fail, we need to preallocate the temporary cpumask in advance. It's stored in the scratchpad. Fixes: 149e566f ("net/mlx5e: Expand XPS cpumask to cover all online cpus") Signed-off-by:
-
Maxim Mikityanskiy authored
mlx5e_ethtool_set_channels updates the indirection table before switching to the new channels. If the switch fails, the indirection table is new, but the channels are old, which is wrong. Fix it by using the preactivate hook of mlx5e_safe_switch_channels to update the indirection table at the stage when nothing can fail anymore. As the code that updates the indirection table is now encapsulated into a new function, use that function in the attach flow when the driver has to reduce the number of channels, and prepare the code for the next commit. Fixes: 85082dba ("net/mlx5e: Correctly handle RSS indirection table when changing number of channels") Signed-off-by:
-
Maxim Mikityanskiy authored
mlx5e_safe_switch_channels accepts a callback to be called before activating new channels. It is intended to configure some hardware parameters in cases where channels are recreated because some configuration has changed. Recently, this callback has started being used to update the driver's internal MLX5E_STATE_XDP_OPEN flag, and the following patches also intend to use this callback for software preparations. This patch renames the hw_modify callback to preactivate, so that the name fits better. Signed-off-by:
-
Maxim Mikityanskiy authored
As a preparation for one of the following commits, create a function to encapsulate the code that notifies the kernel about the new amount of RX and TX queues. The code will be called multiple times in the next commit. Signed-off-by:
-
Tariq Toukan authored
The LRO boolean state in params->lro_en must not be set in case the NIC is not capable. Enforce this check and remove the TODO comment. Signed-off-by:
-
Eran Ben Elisha authored
We shall always extract channel index out of the txq, regardless of the relation between txq_ix and num channels. The extraction is always valid, as if txq is smaller than number of channels, txq_ix == priv->txq2sq[txq_ix]->ch_ix. By doing so, we can remove an if clause from the select queue method, and have one flow for all packets. Signed-off-by:
Eran Ben Elisha <eranbe@mellanox.com> Reviewed-by:
-
- 25 Feb, 2020 11 commits
-
-
David S. Miller authored
Jiri Pirko says: ==================== mlxsw: Implement ACL-dropped packets identification mlxsw hardware allows to insert a ACL-drop action with a value defined by user that would be later on passed with a dropped packet. To implement this, use the existing TC action cookie and pass it to the driver. As the cookie format coming down from TC and the mlxsw HW cookie format is different, do the mapping of these two using idr and rhashtable. The cookie is passed up from the HW through devlink_trap_report() to drop_monitor code. A new metadata type is used for that. Example: $ tc qdisc add dev enp0s16np1 clsact $ tc filter add dev enp0s16np1 ingress protocol ip pref 10 flower skip_sw dst_ip 192.168.1.2 action drop cookie 3b45fa38c8 ^^^^^^^^^^ $ devlink trap set pci/0000:00:10.0 trap acl action trap $ dropwatch Initializing null lookup method dropwatch> set hw true setting hardware drops monitoring to 1 dropwatch> set alertmode packet Setting alert mode Alert mode successfully set dropwatch> start Enabling monitoring... Kernel monitoring activated. Issue Ctrl-C to stop monitoring drop at: ingress_flow_action_drop (acl_drops) origin: hardware input port ifindex: 30 input port name: enp0s16np1 cookie: 3b45fa38c8 <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< timestamp: Fri Jan 24 17:10:53 2020 715387671 nsec protocol: 0x800 length: 98 original length: 98 This way the user may insert multiple drop rules and monitor the dropped packets with the information of which action caused the drop. ==================== Signed-off-by:David S. Miller <davem@davemloft.net>
-
Jiri Pirko authored
Extend existing devlink trap test to include metadata type for flow action cookie. Signed-off-by:
Jiri Pirko <jiri@mellanox.com> Signed-off-by:
Ido Schimmel <idosch@mellanox.com> Signed-off-by:
-
Jiri Pirko authored
Add new trap ACL which reports flow action cookie in a metadata. Allow used to setup the cookie using debugfs file. Signed-off-by:
Jakub Kicinski <kuba@kernel.org> Signed-off-by:
-
Jiri Pirko authored
Use the cookie index received along with the packet to lookup original flow_offload cookie binary and pass it down to devlink_trap_report(). Add "fa_cookie" metadata to the ACL trap. Signed-off-by:
-
Jiri Pirko authored
In case the received packet comes in due to one of ACL discard traps, take the user_def_val_orig_pkt_len field from CQE and store it in skb->cb as ACL cookie index. Signed-off-by:
-
Jiri Pirko authored
Track cookies coming down to driver by flow_offload. Assign a cookie_index to each unique cookie binary. Use previously defined "Trap with userdef" flex action to ask HW to pass cookie_index alongside with the dropped packets. Signed-off-by:
-
Jiri Pirko authored
Expose "Trap action with userdef". It is the same as already defined "Trap action" with a difference that it would ask the policy engine to pass arbitrary value (userdef) alongside with received packets. This would be later on used to carry cookie index. Signed-off-by:
-
Jiri Pirko authored
Add cookie argument to devlink_trap_report() allowing driver to pass in the user cookie. Pass on the cookie down to drop monitor code. Signed-off-by:
-
Jiri Pirko authored
If driver passed along the cookie, push it through Netlink. Signed-off-by:
-
Jiri Pirko authored
Allow driver to indicate cookie metadata for registered traps. Signed-off-by:
-
Jiri Pirko authored
Extend struct flow_action_entry in order to hold TC action cookie specified by user inserting the action. Signed-off-by:
-
- 24 Feb, 2020 19 commits
-
-
David S. Miller authored
Merge tag 'mac80211-next-for-net-next-2020-02-24' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next Johannes Berg says: ==================== A new set of changes: * lots of small documentation fixes, from Jérôme Pouiller * beacon protection (BIGTK) support from Jouni Malinen * some initial code for TID configuration, from Tamizh chelvam * I reverted some new API before it's actually used, because it's wrong to mix controlled port and preauth * a few other cleanups/fixes ==================== Signed-off-by:
-
Colin Ian King authored
There is a spelling mistake in a pr_err message. Fix it. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Ahmad Fatoum authored
The "stm32_pwr_wakeup" is optional per the binding and the driver handles its absence gracefully. Request it with platform_get_irq_byname_optional, so its absence doesn't needlessly clutter the log. Signed-off-by:
Ahmad Fatoum <a.fatoum@pengutronix.de> Signed-off-by:
-
Ahmad Fatoum authored
The specification of a "eth-ck" and a "ptp_ref" clock is optional per the binding and the driver handles them gracefully. Demote the output to an info message accordingly. Signed-off-by:
-
David S. Miller authored
Jeremy Linton says: ==================== Add ACPI bindings to the genet This patch series allows the BCM GENET, as used on the RPi4, to attach when booted in an ACPI environment. The DSDT entry to trigger this is seen below. Of note, the first patch adds a small extension to the mdio layer which allows drivers to find the mii_bus without firmware assistance. The fifth patch in the set retrieves the MAC address from the umac registers rather than carrying it directly in the DSDT. This of course requires the firmware to pre-program it, so we continue to fall back on a random one if it appears to be garbage. v1 -> v2: fail on missing phy-mode property replace phy-mode internal property read string with device_get_phy_mode() equivalent rework mac address detection logic so that it merges the acpi/DT case into device_get_mac_address() allowing _DSD mac address properties. some commit messages justifying why phy_find_first() isn't the worst choice for this driver. + Device (ETH0) + { + Name (_HID, "BCM6E4E") + Name (_UID, 0) + Name (_CCA, 0x0) + Method (_STA) + { + Return (0xf) + } + Method (_CRS, 0x0, Serialized) + { + Name (RBUF, ResourceTemplate () + { + Memory32Fixed (ReadWrite, 0xFd580000, 0x10000, ) + Interrupt (ResourceConsumer, Level, ActiveHigh, Exclusive) { 0xBD } + Interrupt (ResourceConsumer, Level, ActiveHigh, Exclusive) { 0xBE } + }) + Return (RBUF) + } + Name (_DSD, Package () { + ToUUID("daffd814-6eba-4d8c-8a91-bc9bbf4aa301"), + Package () { + Package () { "phy-mode", "rgmii-rxid" }, + } + }) + } ==================== Tested-by:Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
Jeremy Linton authored
If one types "failed to get enet clock" or similar into google there are ~370k hits. The vast majority are people debugging problems unrelated to this adapter, or bragging about their rpi's. Further, the DT clock bindings here are optional. Given that its not a fatal situation with common DT based systems, lets reduce the severity so people aren't seeing failure messages in everyday operation. Signed-off-by:
Jeremy Linton <jeremy.linton@arm.com> Reviewed-by:
Nicolas Saenz Julienne <nsaenzjulienne@suse.de> Acked-by:
-
Jeremy Linton authored
ARM/ACPI machines should utilize self describing hardware when possible. The MAC address on the BCMGENET can be read from the adapter if a full featured firmware has already programmed it. Lets try using the address already programmed, if it appears to be valid. It should be noted that while we move the macaddr logic below the clock and power logic in the driver, none of that code will ever be active in an ACPI environment as the device will be attached to the acpi power domain, and brought to full power with all clocks enabled immediately before the device probe routine is called. One side effect of the above tweak is that while its now possible to read the MAC address via _DSD properties, it should be avoided. Signed-off-by:
-
Jeremy Linton authored
The rpi4 is capable of booting in ACPI mode with the latest edk2-platform commits. As such it would be helpful if the genet platform device were usable. To achieve this we add a new MODULE_DEVICE_TABLE, and convert a few dt specific methods to their generic device_ calls. Until the next patch, ACPI based machines will fallback on random mac addresses. Signed-off-by:
-
Jeremy Linton authored
The unimac mdio driver falls back to scanning the entire bus if its given an appropriate mask. In ACPI mode we expect that the system is well behaved and conforms to recent versions of the specification. We then utilize phy_find_first(), and phy_connect_direct() to find and attach to the discovered phy during net_device open. While its apparently possible to build a genet based device with multiple phys on a single mdio bus, this works for current machines. Further, this driver makes a number of assumptions about the platform device, mac, mdio and phy all being 1:1. Lastly, It also avoids having to create references across the ACPI namespace hierarchy. Signed-off-by:
-
Jeremy Linton authored
The DT phy mode is similar to what we want for ACPI lets factor it out of the of path, and change the of_ call to device_. Signed-off-by:
-
Jeremy Linton authored
It appears most ethernet drivers follow one of two main strategies for mdio bus/phy management. A monolithic model where the net driver itself creates, probes and uses the phy, and one where an external mdio/phy driver instantiates the mdio bus/phy and the net driver only attaches to a known phy. Usually in this latter model the phys are discovered via DT relationships or simply phy name/address hardcoding. This is a shame because modern well behaved mdio buses are self describing and can be probed. The mdio layer itself is fully capable of this, yet there isn't a clean way for a standalone net driver to attach and enumerate the discovered devices. This is because outside of of_mdio_find_bus() there isn't a straightforward way to acquire the mii_bus pointer. So, lets add a mdio_find_bus which can return the mii_bus based only on its name. Signed-off-by:
-
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by:Gustavo A. R. Silva <gustavo@embeddedor.com> Signed-off-by:
-
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
David S. Miller authored
Vladimir Oltean says: ==================== Remainder for "DT bindings for Felix DSA switch on LS1028A" This series is the remainder of patchset [0] which has been merged through Shawn Guo's devicetree tree. It contains changes to the PHY mode validation in the Felix driver ("gmii" to "internal") and the documentation for the DT bindings. [0]: https://patchwork.ozlabs.org/cover/1242716/ ==================== Signed-off-by: -
Vladimir Oltean authored
This patch adds the required documentation for the embedded L2 switch inside the NXP LS1028A chip. I've submitted it in the legacy format instead of yaml schema, because DSA itself has not yet been converted to yaml, and this driver defines no custom bindings. Signed-off-by:
Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Vladimir Oltean authored
phy-mode = "gmii" is confusing because it may mean that the port supports the 8-bit-wide parallel data interface pinout, which it doesn't. It may also be confusing because one of the "gmii" internal ports is actually overclocked to run at 2.5Gbps (even though, yes, as far as the switch MAC is concerned, it still thinks it's gigabit). So use the phy-mode = "internal" property to describe the internal ports inside the NXP LS1028A chip (the ones facing the ENETC). The change should be fine, because the device tree bindings document is yet to be introduced, and there are no stable DT blobs in use. Signed-off-by:
Michael Walle <michael@walle.cc> Signed-off-by:
-
