Ricardo Martinez says:

====================
net: wwan: t7xx: PCIe driver for MediaTek M.2 modem

t7xx is the PCIe host device driver for Intel 5G 5000 M.2 solution which
is based on MediaTek's T700 modem to provide WWAN connectivity.
The driver uses the WWAN framework infrastructure to create the following
control ports and network interfaces:
* /dev/wwan0mbim0 - Interface conforming to the MBIM protocol.
  Applications like libmbim [1] or Modem Manager [2] from v1.16 onwards
  with [3][4] can use it to enable data communication towards WWAN.
* /dev/wwan0at0 - Interface that supports AT commands.
* wwan0 - Primary network interface for IP traffic.

The main blocks in t7xx driver are:
* PCIe layer - Implements probe, removal, and power management callbacks.
* Port-proxy - Provides a common interface to interact with different types
  of ports such as WWAN ports.
* Modem control & status monitor - Implements the entry point for modem
  initialization, reset and exit, as well as exception handling.
* CLDMA (Control Layer DMA) - Manages the HW used by the port layer to send
  control messages to the modem using MediaTek's CCCI (Cross-Core
  Communication Interface) protocol.
* DPMAIF (Data Plane Modem AP Interface) - Controls the HW that provides
  uplink and downlink queues for the data path. The data exchange takes
  place using circular buffers to share data buffer addresses and metadata
  to describe the packets.
* MHCCIF (Modem Host Cross-Core Interface) - Provides interrupt channels
  for bidirectional event notification such as handshake, exception, PM and
  port enumeration.

The compilation of the t7xx driver is enabled by the CONFIG_MTK_T7XX config
option which depends on CONFIG_WWAN.
This driver was originally developed by MediaTek. Intel adapted t7xx to
the WWAN framework, optimized and refactored the driver source code in close
collaboration with MediaTek. This will enable getting the t7xx driver on the
Approved Vendor List for interested OEM's and ODM's productization plans
with Intel 5G 5000 M.2 solution.

List of contributors:
Amir Hanania <amir.hanania@intel.com>
Andriy Shevchenko <andriy.shevchenko@linux.intel.com>
Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Dinesh Sharma <dinesh.sharma@intel.com>
Eliot Lee <eliot.lee@intel.com>
Haijun Liu <haijun.liu@mediatek.com>
M Chetan Kumar <m.chetan.kumar@intel.com>
Mika Westerberg <mika.westerberg@linux.intel.com>
Moises Veleta <moises.veleta@intel.com>
Pierre-louis Bossart <pierre-louis.bossart@intel.com>
Chiranjeevi Rapolu <chiranjeevi.rapolu@intel.com>
Ricardo Martinez <ricardo.martinez@linux.intel.com>
Madhusmita Sahu <madhusmita.sahu@intel.com>
Muralidharan Sethuraman <muralidharan.sethuraman@intel.com>
Soumya Prakash Mishra <Soumya.Prakash.Mishra@intel.com>
Sreehari Kancharla <sreehari.kancharla@intel.com>
Suresh Nagaraj <suresh.nagaraj@intel.com>

[1] https://www.freedesktop.org/software/libmbim/
[2] https://www.freedesktop.org/software/ModemManager/
[3] https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/merge_requests/582
[4] https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/merge_requests/523

V8:
- Rebase skb_data_area_size() patch (02).

V7:
- Delete unused macros.
- Avoid duplicated calls to le32_to_cpu().
- Fix 'out of bounds' compilation error.
- Rename port_number to port_count.
- Remove '!!' when the destination variable is a boolean.
- Remove unneeded spinlock around rx_length_th.
- Remove common field from union inside dpmaif_drb struct.
- Use 'goto' for the exit flow in t7xx_pci_enable_sleep().
- Merge CLDMA tgpd and rgpd structs.
- Add comments to clarify skb consumption by ports.
- Introduce skb_data_area_size() helper.
- Declare the port config array as constant.
- Update CLDMA_JUMBO_BUFF_SZ definition when ccci_header
  is introduced by port-proxy patch.
- Update Reviewed-by tags.
- Simplify t7xx_dpmaif_tx_send_skb() and make t7xx_dpmaif_add_skb_to_ring()
  report the tx queue full state early.

V6:
- Remove unneeded initializations and bit masks.
- Remove t7xx_common.h file.
- Add comment to circular linking in GPD list.
- Use min instead of min_t.
- Use int for local indexes instead of short or char.
- Update the commit message in CLDMA patch about dependencies on core patch.
- Add space between contributor name and email address.
- Rename registers with double negatives
  e.g. DIS_ASPM_LOWPWR_CLR_0 -> ENABLE_ASPM_LOWPWR.
- Fix a race condition in pci sleep resource locking.
- Initialize interrupts with t7xx_pcie_mac_set_int() instead of 'clear'.
- Remove duplicate spin_lock_init(&md->exp_lock).
- Remove .ndo_select_queue callback due to singular TX queue.
- Remove call to deprecated netif_rx_any_context().
- Fix include guard name in t7xx_hif_dpmaif.h.
- Remove unused q_num parameter in DPMAIF functions.
- Do not serialize the drb_wr_idx write in t7xx_dpmaif_add_skb_to_ring()
  and the read from t7xx_txq_drb_wr_available().
- Fix potential leak in t7xx_dpmaif_add_skb_to_ring().
- Unionize:
    DRB structs: msg and pd.
    PIT structs: msg and pd.
- Replace list_head & spinlock with skb_buff_head in dpmaif_tx_queue.
- Remove rx_length_th check in TX WWAN port flow.
- Remove wwan_remove_port() from the critical section in WWAN port uninit.
- Use skb_end_pointer() to avoid conditional compilation.
- Simplify the loop in t7xx_port_ctrl_tx() by checking the buffer offset
  instead of calculating the number of required packets.
- Remove the code for unused channel PORT_CH_STATUS_RX.
- Remove bit flags from ports. Ports can check chan_enable instead of the
  PORT_F_RX_ALLOW_DROP flag.
- Use INVALID_SEQ_NUM to identify the first seq number.
- Rename port_static to port_conf and ports_private to ports.
- Implement t7xx_port_send_skb() and t7xx_port_send_ctl_skb() in a layered
  approach to reduce duplicated code and simplify the CCCI header handling.
- Move wwan_port_rx() call from port-proxy to WWAN port.
- Rename t7xx_port_recv_skb() to t7xx_port_enqueue_skb().
- Move control message parsing logic from port-proxy to control port,
  preserve the endianness when parsing the message and make port-proxy
  export a function to enable/disable ports.
- Use flexible arrays for:
    port-proxy ports.
    payload data in t7xx_fsm_event, port_msg, and mtk_runtime_feature.

v5:
- Update Intel's copyright years to 2021-2022.
- Remove circular dependency between DPMAIF HW (07) and HIF (08).
- Keep separate patches for CLDMA (02) and Core (03)
  but improve the code split by decoupling CLDMA from
  modem ops and cleaning up t7xx_common.h.
- Rename ID_CLDMA0/ID_CLDMA1 to CLDMA_ID_AP/CLDMA_ID_MD.
- Consistently use CLDMA's ring_lock to protect tr_ring.
- Free resources first and then print messages.
- Implement suggested name changes.
- Do not explicitly include dev_printk.h.
- Remove redundant dev_err()s.
- Fix possible memory leak during probe.
- Remove infrastructure for legacy interrupts.
- Remove unused macros and variables, including those that
  can be replaced with constants.
- Remove PCIE_MAC_MSIX_MSK_SET macro which is duplicated code.
- Refactor __t7xx_pci_pm_suspend() for clarity.
- Refactor t7xx_cldma_rx_ring_init() and t7xx_cldma_tx_ring_init().
- Do not use & for function callbacks.
- Declare a structure to access skb->cb[] data.
- Use skb_put_data instead of memcpy.
- No need to use kfree_sensitive.
- Use dev_kfree_skb() instead of kfree_skb().
- Refactor t7xx_prepare_device_rt_data() to remove potential leaks,
  avoid unneeded memset and keep rt_data and packet_size updates
  inside the same 'if' block.
- Set port's rx_length_th back to 0 during uninit.
- Remove unneeded 'blocking' parameter from t7xx_cldma_send_skb().
- Return -EIO in t7xx_cldma_send_skb() if the queue is inactive.
- Refactor t7xx_cldma_qs_are_active() to use pci_device_is_present().
- Simplify t7xx_cldma_stop_q() and rename it to t7xx_cldma_stop_all_qs().
- Fix potential leaks in t7xx_cldma_init().
- Improve error handling in fsm_append_event and fsm_routine_starting().
- Propagate return codes from fsm_append_cmd() and t7xx_fsm_append_event().
- Refactor fsm_wait_for_event() to avoid unnecessary sleep.
- Create the WWAN ports and net device only after the modem is in
  the ready state.
- Refactor t7xx_port_proxy_recv_skb() and port_recv_skb().
- Rename t7xx_port_check_rx_seq_num() as t7xx_port_next_rx_seq_num()
  and fix the seq_num logic to handle overflows.
- Declare seq_nums as u16 instead of short.
- Use unsigned int for local indexes.
- Use min_t instead of the ternary operator.
- Refactor the loop in t7xx_dpmaif_rx_data_collect() to avoid a dead
  condition check.
- Use a bitmap (bat_bitmap) instead of an array to keep track of
  the DRB status. Used in t7xx_dpmaif_avail_pkt_bat_cnt().
- Refactor t7xx_dpmaif_tx_send_skb() to protect tx_submit_skb_cnt
  with spinlock and remove the misleading tx_drb_available variable.
- Consolidate bit operations before endianness conversion.
- Use C bit fields in dpmaif_drb_skb struct which is not HW related.
- Add back the que_started check in t7xx_select_tx_queue().
- Create a helper function to get the DRB count.
- Simplify the use of 'usage' during t7xx_ccmni_close().
- Enforce CCMNI MTU selection with BUILD_BUG_ON() instead of a comment.
- Remove t7xx_ccmni_ctrl->capability parameter which remains constant.

v4:
- Implement list_prev_entry_circular() and list_next_entry_circular() macros.
- Remove inline from all c files.
- Define ioread32_poll_timeout_atomic() helper macro.
- Fix return code for WWAN port tx op.
- Allow AT commands fragmentation same as MBIM commands.
- Introduce t7xx_common.h file in the first patch.
- Rename functions and variables as suggested in v3.
- Reduce code duplication by creating fsm_wait_for_event() helper function.
- Remove unneeded dev_err in t7xx_fsm_clr_event().
- Remove unused variable last_state from struct t7xx_fsm_ctl.
- Remove unused variable txq_select_times from struct dpmaif_ctrl.
- Replace ETXTBSY with EBUSY.
- Refactor t7xx_dpmaif_rx_buf_alloc() to remove an unneeded allocation.
- Fix potential leak at t7xx_dpmaif_rx_frag_alloc().
- Simplify return value handling at t7xx_dpmaif_rx_start().
- Add a helper to handle the common part of CCCI header initialization.
- Make sure interrupts are enabled during PM resume.
- Add a parameter to t7xx_fsm_append_cmd() to tell if it is in interrupt context.

v3:
- Avoid unneeded ping-pong changes between patches.
- Use t7xx_ prefix in functions.
- Use t7xx_ prefix in generic structs where mtk_ or ccci prefix was used.
- Update Authors/Contributors header.
- Remove skb pools used for control path.
- Remove skb pools used for RX data path.
- Do not use dedicated TX queue for ACK-only packets.
- Remove __packed attribute from GPD structs.
- Remove the infrastructure for test and debug ports.
- Use the skb control buffer to store metadata.
- Get the IP packet type from RX PIT.
- Merge variable declaration and simple assignments.
- Use preferred coding patterns.
- Remove global variables.
- Declare HW facing structure members as little endian.
- Rename goto tags to describe what is going to be done.
- Do not use variable length arrays.
- Remove unneeded blank lines source code and kdoc headers.
- Use C99 initialization format for port-proxy ports.
- Clean up comments.
- Review included headers.
- Better use of 100 column limit.
- Remove unneeded mb() in CLDMA.
- Remove unneeded spin locks and atomics.
- Handle read_poll_timeout error.
- Use dev_err_ratelimited() where required.
- Fix resource leak when requesting IRQs.
- Use generic DEFAULT_TX_QUEUE_LEN instead custom macro.
- Use ETH_DATA_LEN instead of defining WWAN_DEFAULT_MTU.
- Use sizeof() instead of defines when the size of structures is required.
- Remove unneeded code from netdev:
    No need to configure HW address length
    No need to implement .ndo_change_mtu
    Remove random address generation
- Code simplifications by using kernel provided functions and macros such as:
    module_pci_driver
    PTR_ERR_OR_ZERO
    for_each_set_bit
    pci_device_is_present
    skb_queue_purge
    list_prev_entry
    __ffs64

v2:
- Replace pdev->driver->name with dev_driver_string(&pdev->dev).
- Replace random_ether_addr() with eth_random_addr().
- Update kernel-doc comment for enum data_policy.
- Indicate the driver is 'Supported' instead of 'Maintained'.
- Fix the Signed-of-by and Co-developed-by tags in the patches.
- Added authors and contributors in the top comment of the src files.
====================

Adds maintainers and documentation for MediaTek t7xx 5G WWAN modem
device driver.
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Introduce the mechanism to lock/unlock the device 'deep sleep' mode.
When the PCIe link state is L1.2 or L2, the host side still can keep
the device is in D0 state from the host side point of view. At the same
time, if the device's 'deep sleep' mode is unlocked, the device will
go to 'deep sleep' while it is still in D0 state on the host side.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Enables runtime power management callbacks including runtime_suspend
and runtime_resume. Autosuspend is used to prevent overhead by frequent
wake-ups.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Eliot Lee <eliot.lee@intel.com>
Signed-off-by: Eliot Lee <eliot.lee@intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Implements suspend, resumes, freeze, thaw, poweroff, and restore
`dev_pm_ops` callbacks.

From the host point of view, the t7xx driver is one entity. But, the
device has several modules that need to be addressed in different ways
during power management (PM) flows.
The driver uses the term 'PM entities' to refer to the 2 DPMA and
2 CLDMA HW blocks that need to be managed during PM flows.
When a dev_pm_ops function is called, the PM entities list is iterated
and the matching function is called for each entry in the list.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Creates the Cross Core Modem Network Interface (CCMNI) which implements
the wwan_ops for registration with the WWAN framework, CCMNI also
implements the net_device_ops functions used by the network device.
Network device operations include open, close, start transmission, TX
timeout and change MTU.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Co-developed-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Data Path Modem AP Interface (DPMAIF) HIF layer provides methods
for initialization, ISR, control and event handling of TX/RX flows.

DPMAIF TX
Exposes the 'dmpaif_tx_send_skb' function which can be used by the
network device to transmit packets.
The uplink data management uses a Descriptor Ring Buffer (DRB).
First DRB entry is a message type that will be followed by 1 or more
normal DRB entries. Message type DRB will hold the skb information
and each normal DRB entry holds a pointer to the skb payload.

DPMAIF RX
The downlink buffer management uses Buffer Address Table (BAT) and
Packet Information Table (PIT) rings.
The BAT ring holds the address of skb data buffer for the HW to use,
while the PIT contains metadata about a whole network packet including
a reference to the BAT entry holding the data buffer address.
The driver reads the PIT and BAT entries written by the modem, when
reaching a threshold, the driver will reload the PIT and BAT rings.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Data Path Modem AP Interface (DPMAIF) HW layer provides HW abstraction
for the upper layer (DPMAIF HIF). It implements functions to do the HW
configuration, TX/RX control and interrupt handling.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Adds AT and MBIM ports to the port proxy infrastructure.
The initialization method is responsible for creating the corresponding
ports using the WWAN framework infrastructure. The implemented WWAN port
operations are start, stop, and TX.
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Control Port implements driver control messages such as modem-host
handshaking, controls port enumeration, and handles exception messages.

The handshaking process between the driver and the modem happens during
the init sequence. The process involves the exchange of a list of
supported runtime features to make sure that modem and host are ready
to provide proper feature lists including port enumeration. Further
features can be enabled and controlled in this handshaking process.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Port-proxy provides a common interface to interact with different types
of ports. Ports export their configuration via `struct t7xx_port` and
operate as defined by `struct port_ops`.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Co-developed-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Registers the t7xx device driver with the kernel. Setup all the core
components: PCIe layer, Modem Host Cross Core Interface (MHCCIF),
modem control operations, modem state machine, and build
infrastructure.

* PCIe layer code implements driver probe and removal.
* MHCCIF provides interrupt channels to communicate events
  such as handshake, PM and port enumeration.
* Modem control implements the entry point for modem init,
  reset and exit.
* The modem status monitor is a state machine used by modem control
  to complete initialization and stop. It is used also to propagate
  exception events reported by other components.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Cross Layer DMA (CLDMA) Hardware interface (HIF) enables the control
path of Host-Modem data transfers. CLDMA HIF layer provides a common
interface to the Port Layer.

CLDMA manages 8 independent RX/TX physical channels with data flow
control in HW queues. CLDMA uses ring buffers of General Packet
Descriptors (GPD) for TX/RX. GPDs can represent multiple or single
data buffers (DB).

CLDMA HIF initializes GPD rings, registers ISR handlers for CLDMA
interrupts, and initializes CLDMA HW registers.

CLDMA TX flow:
1. Port Layer write
2. Get DB address
3. Configure GPD
4. Triggering processing via HW register write

CLDMA RX flow:
1. CLDMA HW sends a RX "done" to host
2. Driver starts thread to safely read GPD
3. DB is sent to Port layer
4. Create a new buffer for GPD ring

Note: This patch does not enable compilation since it has dependencies
such as t7xx_pcie_mac_clear_int()/t7xx_pcie_mac_set_int() and
struct t7xx_pci_dev which are added by the core patch.
Signed-off-by: Haijun Liu <haijun.liu@mediatek.com>
Signed-off-by: Chandrashekar Devegowda <chandrashekar.devegowda@intel.com>
Co-developed-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Helper to calculate the linear data space in the skb.
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add macros to get the next or previous entries and wraparound if
needed. For example, calling list_next_entry_circular() on the last
element should return the first element in the list.
Signed-off-by: Ricardo Martinez <ricardo.martinez@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This cleanup patchset includes the following patches:

 - bump version strings, by Simon Wunderlich

 - remove unnecessary type castings, by Yu Zhe
Signed-off-by: David S. Miller <davem@davemloft.net>

Ido Schimmel says:

====================
mlxsw: A dedicated notifier block for router code

Petr says:

Currently all netdevice events are handled in the centralized notifier
handler maintained by spectrum.c. Since a number of events are involving
router code, spectrum.c needs to dispatch them to spectrum_router.c. The
spectrum module therefore needs to know more about the router code than it
should have, and there is are several API points through which the two
modules communicate.

In this patchset, move bulk of the router-related event handling to the
router code. Some of the knowledge has to stay: spectrum.c cannot veto
events that the router supports, and vice versa. But beyond that, the two
can ignore each other's details, which leads to more focused and simpler
code.

As a side effect, this fixes L3 HW stats support on tunnel netdevices.

The patch set progresses as follows:

- In patch #1, change spectrum code to not bounce L3 enslavement, which the
  router code supports.

- In patch #2, add a new do-nothing notifier block to the router code.

- In patches #3-#6, move router-specific event handling to the router
  module. In patch #7, clean up a comment.

- In patch #8, use the advantage that all router event handling is in the
  router code and clean up taking router lock.

- mlxsw supports L3 HW stats on tunnels as of this patchset. Patches #9 and
  #10 therefore add a selftest for L3 HW stats support on tunnels.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a selftest that uses an IPIP topology and tests that L3 HW stats
reflect the traffic in the tunnel.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The function get_l3_stats() from the test hw_stats_l3.sh will be useful for
any test that wishes to work with L3 stats. Furthermore, it is easy to
generalize to other HW stats suites (for when such are added). Therefore,
move the code to lib.sh, rewrite it to have the same interface as the other
stats-collecting functions, and generalize to take the name of the HW stats
suite to collect as an argument.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For notifications that the router needs to handle, router lock is taken.
Further, at least to determine whether an event is related to a tunnel
underlay, router lock also needs to be taken. Due to this, the router lock
is always taken for each unhandled event, and also for some handled events,
even if they are not related to underlay. Thus each event implies at least
one router lock, sometimes two.

Instead of deferring the locking to the leaf handlers, take the lock in the
router notifier handler always. This simplifies thinking about the locking
state, and in some cases saves one lock cycle.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The position of netdevice notifier registration no longer depends on the
router initialization, because the event handler no longer dispatches to
the router code. Update the comment at the registration to that effect.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The events related to IPIP tunnels are handled by the router code. Move the
handling from the central dispatcher in spectrum.c to the new notifier
handler in the router module.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The events NETDEV_PRE_CHANGEADDR, NETDEV_CHANGEADDR and NETDEV_CHANGEMTU
have implications for in-ASIC router interface objects, and as such are
handled in the router module. Move the handling from the central dispatcher
in spectrum.c to the new notifier handler in the router module.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

L3 HW stats are implemented in mlxsw as RIF counters, and therefore the
code resides in spectrum_router. Exclude the offload xstats events from the
mlxsw_sp_netdevice_event_is_router() predicate, and instead recreate the
glue code in the router module.

Previously, the order of dispatch was that for events on tunnels, a
dedicated handler was called, which however did not handle HW stats events.
But there is nothing special about tunnel devices as far as HW stats: there
is a RIF associated with the tunnel netdevice, and that RIF is where the
counter should be installed. Therefore now, HW stats events are tested
first, independent of netdevice type. The upshot is that as of this commit,
mlxsw supports L3 HW stats work on GRE tunnels.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Events involving VRF, as L3 concern, are handled in the router code, by the
helper mlxsw_sp_netdevice_vrf_event(). The handler is currently invoked
from the centralized dispatcher in spectrum.c. Instead, move the call to
the newly-introduced router-specific notifier handler.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently all netdevice events are handled in the centralized notifier
handler maintained by spectrum.c. Since a number of events are involving
router code, spectrum.c needs to dispatch them to spectrum_router.c. The
spectrum module therefore needs to know more about the router code than it
should have, and there is are several API points through which the two
modules communicate.

To simplify the notifier handlers, introduce a new notifier into the router
module.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Enslaving netdevices to VRF is currently handled through an
mlxsw_sp_is_vrf_event() conditional in mlxsw_sp_netdevice_event(). In the
following patch sets, VRF enslavement will be handled purely in the router
code. Therefore make handlers of NETDEV_PRECHANGEUPPER tolerant of
enslaving to VRF, so that they do not bounce the change.

For NETDEV_CHANGEUPPER, drop the WARN_ON(1) and bounce from
mlxsw_sp_netdevice_port_vlan_event(). This is the only handler that warned
and bounces even in the CHANGEUPPER code, other handler quietly do nothing
when they encounter an unfamiliar upper.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jakub Kicinski says:

====================
net: switch drivers to netif_napi_add_weight()

The minority of drivers pass a custom weight to netif_napi_add().
Switch those away to the new netif_napi_add_weight(). All drivers
(which can go thru net-next) calling netif_napi_add() will now
be calling it with NAPI_POLL_WEIGHT or 64.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

A handful of WAN drivers use custom napi weights,
switch them to the new API.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio netdev driver uses a custom napi weight, switch to the new
API for setting custom weight.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

r8152 uses a custom napi weight, switch to the new
API for setting custom weight.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Switch all Ethernet drivers which use custom napi weights
to the new API.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

caif_virtio uses a custom napi weight, switch to the new
API for setting custom weights.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

UM's netdev driver uses a custom napi weight, switch to the new
API for setting custom weight.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Vladimir Oltean says:

====================
Simplify migration of host filtered addresses in Felix driver

The purpose of this patch set is to remove the functions
dsa_port_walk_fdbs() and dsa_port_walk_mdbs() from the DSA core, which
were introduced when the Felix driver gained support for unicast
filtering on standalone ports. They get called when changing the tagging
protocol back and forth between "ocelot" and "ocelot-8021q".
I did not realize we could get away without having them.

The patch set was regression-tested using the local_termination.sh
selftest using both tagging protocols.
====================

Link: https://lore.kernel.org/r/20220505162213.307684-1-vladimir.oltean@nxp.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

All the users of these functions are gone, delete them before they gain
new ones.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The felix driver is the only user of dsa_port_walk_mdbs(), and there
isn't even a good reason for it, considering that the host MDB entries
are already saved by the ocelot switch lib in the ocelot->multicast list.

Rewrite the multicast entry migration procedure around the
ocelot->multicast list so we can delete dsa_port_walk_mdbs().
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

I just realized we don't need to migrate the host-filtered FDB entries
when the tagging protocol changes from "ocelot" to "ocelot-8021q".

Host-filtered addresses are learned towards the PGID_CPU "multicast"
port group, reserved by software, which contains BIT(ocelot->num_phys_ports).
That is the "special" port entry in the analyzer block for the CPU port
module.

In "ocelot" mode, the CPU port module's packets are redirected to the
NPI port.

In "ocelot-8021q" mode, felix_8021q_cpu_port_init() does something funny
anyway, and changes PGID_CPU to stop pointing at the CPU port module and
start pointing at the physical port where the DSA master is attached.

The fact that we can alter the destination of packets learned towards
PGID_CPU without altering the MAC table entries themselves means that it
is pointless to walk through the FDB entries, forget that they were
learned towards PGID_CPU, and re-learn them towards the "unicast" PGID
associated with the physical port connected to the DSA master. We can
let the PGID_CPU value change simply alter the destination of the
host-filtered unicast packets in one fell swoop.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ocelot_fdb_add() redirects FDB entries installed on the NPI port towards
the special reserved PGID_CPU used for host-filtered addresses. PGID_CPU
contains BIT(ocelot->num_phys_ports) in the destination port mask, which
is code name for the CPU port module.

Whereas felix_migrate_fdbs_to_*_port() uses the ocelot->num_phys_ports
PGID directly, and it appears that this works too. Even if this PGID is
set to zero, apparently its number is special and packets still reach
the CPU port module.

Nonetheless, in the end, these addresses end up in the same place
regardless of whether they go through an extra indirection layer or not.
Use PGID_CPU across to have more uniformity.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

This patch increases the polling rate used by the
mlxbf_gige driver on the MDIO bus.  The previous
polling rate was every 100us, and the new rate is
every 5us.  With this change the amount of time
spent waiting for the MDIO BUSY signal to de-assert
drops from ~100us to ~27us for each operation.
Signed-off-by: David Thompson <davthompson@nvidia.com>
Signed-off-by: Asmaa Mnebhi <asmaa@nvidia.com>
Link: https://lore.kernel.org/r/20220505162309.20050-1-davthompson@nvidia.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>