During igc_down(), we call igc_nfc_rule_exit() which traverse the NFC
rule list disabling filters one by one. Later on in igc_down() flow
we issue an hardware reset which also clear all filters.  Since we
already reset the hardware, we don't actually need to disable each
filter manually. In order to simplify the code, this patch removes
igc_nfc_rule() altogether.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch changes adapter->nfc_rule_lock type from spin_lock to mutex
so we avoid unnecessary busy waiting on lock contention.

A closer look at the execution context of NFC rule API users shows that
all of them run in process context. The API users are: ethtool ops,
igc_configure(), called when interface is brought up by user or reset
workequeue thread, igc_down(), called when interface is brought down,
and igc_remove(), called when driver is unloaded.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

None of igc_disable_nfc_rule() callers actually check its returning
value. A closer look at why this function would fail shows that the
only situation is when we try to delete an Ethertype or MAC filter that
doesn't exist.

That situation is very unlikely so we can change igc_del_etype_filter()
and igc_del_mac_filter() logic to "if the filter doesn't exist, we are
done", and keep the logic in igc_disable_nfc_rule() callers simple.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

If we try to overwrite an existing rule with the same filter but
different action, we get EEXIST error as shown below.

$ ethtool -N eth0 flow-type ether dst <MACADDR> action 1 loc 10
$ ethtool -N eth0 flow-type ether dst <MACADDR> action 2 loc 10
rmgr: Cannot insert RX class rule: File exists

The second command is expected to overwrite the previous rule in location
10 and succeed.

This patch fixes igc_ethtool_check_nfc_rule() so it also checks the
rules location. In case they match, the rule under evaluation should not
be considered invalid.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

If we have RFC rules in adapter->nfc_rule_list when the IGC driver
is unloaded, all rules are leaked. This patch fixes the issue by
introducing the helper igc_flush_nfc_rules() and calling it in
igc_remove(). It also updates igc_set_features() so is reuses the
new helper instead of re-implementing it.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Current implementation of igc_ethtool_update_nfc_rule() is a bit
convoluted since it handles too many things: rule lookup, deletion
and addition. This patch breaks it into three functions so we simplify
the code and improve code reuse.

Code related to rule lookup is refactored out to a new function called
igc_get_nfc_rule().

Code related to rule addition is refactored out to a new function called
igc_add_nfc_rule(). This function enables the rule in hardware and adds
it to the adapter's list.

Code related to rule deletion is refactored out to a new function called
igc_del_nfc_rule(). This function disables the rule in hardware, removes
it from adapter's list, and deletes it.

As a byproduct of this refactoring, igc_enable_nfc_rule() and
igc_disable_nfc_rule() are moved to igc_main.c since they are not used
in igc_ethtool.c anymore, and igc_restore_nfc_rules() and igc_nfc_rule_
exit() are moved around to avoid forward declaration.

Also, since this patch already touches igc_ethtool_get_nfc_rule(), it
takes the opportunity to remove the 'match_flags' check. Empty flags
are not allowed to be added so no need to check that.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

When network interface is brought up, the driver re-enables the NFC
rules previously configured. However, this is done in reverse order
the rules were added and hardware filters are configured differently.

For example, consider the following rules:

$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:AA queue 0
$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:BB queue 1
$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:CC queue 2
$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:DD queue 3

RAL/RAH registers are configure so filter index 1 has address ending
with AA, filter index 2 has address ending in BB, and so on.

If we bring the interface down and up again, RAL/RAH registers are
configured so filter index 1 has address ending in DD, filter index 2
has CC, and so on. IOW, in reverse order we had before bringing the
interface down.

This issue can be fixed by traversing adapter->nfc_rule_list in
backwards when restoring the rules. Since hlist doesn't support
backwards traversal, this patch replaces it by list_head and fixes
igc_restore_nfc_rules() accordingly.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Multicast MAC addresses are valid address for NFC rules but
igc_add_mac_filter() is currently rejecting them. In fact, the I225
controller doesn't impose any constraint on the address value so this
patch gets rid of the address validation check in MAC filter APIs.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

When the 'loc' argument is passed in ethtool, the input rule overwrites
any rule present in that location. In this situation we must disable the
old rule otherwise it is left enabled in hardware. This patch fixes
the issue by always calling igc_disable_nfc_rule() when deleting the
old rule, no matter the value of 'input' argument.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Access to NFC rules stored in adapter->nfc_rule_list is protect by
adapter->nfc_rule_lock. The functions igc_ethtool_get_nfc_rule()
and igc_ethtool_get_nfc_rules() are missing to hold the lock while
accessing rule objects.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The 'sw_idx' field from 'struct igc_nfc_rule' is u16 type but it is
assigned an u32 value in igc_ethtool_init_nfc_rule(). This patch changes
'sw_idx' type to u32 so they match. Also, it makes more sense to call
this field 'location' since it holds the NFC rule location.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Current implementation of igc_ethtool_add_nfc_rule() is quite long and a
bit convoluted so this patch does a code refactoring to improve the
code.

Code related to NFC rule object initialization is refactored out to the
local helper function igc_ethtool_init_nfc_rule(). Likewise, code
related to NFC rule validation is refactored out to another local
helper, igc_ethtool_is_nfc_rule_valid().

RX_CLS_FLOW_DISC check is removed since it is redundant. The macro is
defined as the max value fsp->ring_cookie can have, so checking if
fsp->ring_cookie >= adapter->num_rx_queues is already sufficient.

Finally, some log messages are improved or added, and obvious comments
are removed.
Signed-off-by: NAndre Guedes <andre.guedes@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This reverts commit 5e3768a4.

On request from Russell King, this is a layering violation.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Upon adapter hotplug, cxgb4 registers ULD devices for all the ULDs that
are already loaded, ensuring that ULD's can enumerate the hotplugged
adapter without reloading the ULD.
Signed-off-by: NPotnuri Bharat Teja <bharat@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make FLOW_ACTION_HW_STATS_DONT_CARE be all bits, rather than none, so that
 drivers and __flow_action_hw_stats_check can use simple bitwise checks.

Pre-fill all actions with DONT_CARE in flow_rule_alloc(), rather than
 relying on implicit semantics of zero from kzalloc, so that callers which
 don't configure action stats themselves (i.e. netfilter) get the correct
 behaviour by default.

Only the kernel's internal API semantics change; the TC uAPI is unaffected.

v4: move DONT_CARE setting to flow_rule_alloc() for robustness and simplicity.

v3: set DONT_CARE in nft and ct offload.

v2: rebased on net-next, removed RFC tags.
Signed-off-by: NEdward Cree <ecree@solarflare.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This adds the driver for the MediaTek STAR Ethernet MAC currently used
on the MT8* SoC family. For now we only support full-duplex.
Signed-off-by: NBartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The Makefile formatting in the kernel tree usually doesn't use tabs,
so remove them before we add a second driver.
Signed-off-by: NBartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We'll soon by adding a second MediaTek Ethernet driver so modify the
Kconfig prompt.
Signed-off-by: NBartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This commit reduces the driver load time by using usec resolution
instead of msec when polling for hardware state change.

Also add back-off mechanism to handle cases where minimal sleep
time is not enough.
Signed-off-by: NJosh Triplett <josh@joshtriplett.org>
Signed-off-by: NSameeh Jubran <sameehj@amazon.com>
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

1. Use BIT macro instead of shift operator for code clarity
2. Replace multiple flag assignments to a single assignment of multiple
   flags in ena_com_add_single_rx_desc()
3. Move ENA_HASH_KEY_SIZE from ena_netdev.h to ena_com.h
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

1. Add leading and trailing spaces to several comments for better
   readability
2. Make tabs and spaces uniform in enum defines in ena_admin_defs.h
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

1. Reorder sanity checks in get_comp_ctxt() to make more sense
2. Reorder variables in ena_com_fill_hash_function() and
   ena_calc_io_queue_size() in reverse christmas tree.
3. Move around member initializations.
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

1. Remove unused definition of DRV_MODULE_VERSION
2. Remove {} from single line-of-code ifs
3. Remove unnecessary comments from ena_get/set_coalesce()
4. Remove unnecessary extra spaces and newlines
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

1. Join unnecessarily broken short lines in ena_com.c ena_netdev.c
2. Fix Indentations of broken lines
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

fix spelling and grammar mistakes in comments in ena_com.h,
ena_com.c and ena_netdev.c
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make all types of variables that convey the number and sizeof queues to
be u32, for consistency with the API between the driver and device via
ena_admin_defs.h:ena_admin_get_feat_resp.max_queue_ext fields. Current
code sometimes uses int and there are multiple assignments between these
variables with different types.
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename ena_update_tx/rx_rings_intr_moderation() to
ena_update_tx/rx_rings_nonadaptive_intr_moderation()
to distinguish between adaptive and non adaptive interrupt moderaion.
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Initialize prev_intr_delay_resolution with ena_dev->intr_delay_resolution
unconditionally, since it is initialized with
ENA_DEFAULT_INTR_DELAY_RESOLUTION in ena_probe(). This approach makes much
more sense than handling errors of not initializing it.

Also added unlikely to if condition.
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Default return value should be -EINVAL since the input
in this case was unexpected.
Also remove the now redundant check in the beginning
of the function.
Signed-off-by: NSameeh Jubran <sameehj@amazon.com>
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use u64 instead of unsigned long long for clarity
Signed-off-by: NShai Brandes <shaibran@amazon.com>
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename ena_com_free_desc to ena_com_free_q_entries to match
the LLQ mode.

In non-LLQ mode, an entry in an IO ring corresponds to a
a descriptor. In LLQ mode an entry may correspond to several
descriptors (per LLQ definition).
Signed-off-by: NIgor Chauskin <igorch@amazon.com>
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Newer ENA devices can write data to rx buffers with an offset
from the beginning of the buffer.

This commit adds support for this feature in the driver.
Signed-off-by: NSameeh Jubran <sameehj@amazon.com>
Signed-off-by: NArthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes an inconsistency between code and spec, which
was found while working on the QoS implementation.

When 8TCs are used, 2 is the maximum supported number of index bits.
In a 4TC mode, we do support 3, but we shouldn't really use the bytes,
which are intended for the 8TC mode.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for mqprio min_rate limiters.

A2 HW supports Weighted Strict Priority (WSP) arbitration for Tx Descriptor
Queue scheduling among TCs, which can be used for min_rate shaping.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch changes the order of arguments for TC weight/credit setter
functions.
Having the "value to be set" on the right is slightly more robust in
a sense that it's more natural for the humans, so it's a bit more
error-proof this way.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch changes the TC-queue mapping mechanism used on A2.
Configure the A2 HW in such a way that we can keep queue index mapping
exactly as it was on A1.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for automatic queue number downgrade.

On A2: this is a must have, because only TC0/TC1 support more than 4Q.
Other TCs support 4Qs maximum.
Thus, on A2 we must downgrade the number of queues per TC to 4, if more
than 2 TCs are requested.

On A1: this allows using 8TCs even on systems with cpu count >= 8, when
we have 8 queues by default.
We will just automatically switch to 8TCx4Q mode in this case.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds initial support for mqprio rate limiters (max_rate only).

Atlantic HW supports Rate-Shaping for time-sensitive traffic at per
Traffic Class (TC) granularity.
Target rate is defined by:
* nominal link rate (always 10G);
* rate factor (ratio between nominal rate and max allowed).
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch updates TCVEC2RING to accept nic_cfg, which is needed to be able
to use it from hw_atl.
The name is updated to reflect the changes.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for per-TC queue statistics.

By default (single TC), the output is the same as it used to be, e.g.:
     Queue[0] InPackets: 2
     Queue[0] OutPackets: 8
     Queue[0] Restarts: 0
     Queue[0] InJumboPackets: 0
     Queue[0] InLroPackets: 0
     Queue[0] InErrors: 0

If several TCs are enabled, then each queue statistics line is prefixed
with TC number, e.g.:
     TC0 Queue[0] InPackets: 6
     TC0 Queue[0] OutPackets: 11
Queue numbering is end-to-end, so:
     TC1 Queue[4] InPackets: 0
     TC1 Queue[4] OutPackets: 22
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>