mlxsw: spectrum_qdisc: Offload root TBF as port shaper

The Spectrum ASIC allows configuration of maximum shaper on all levels of
the scheduling hierarchy: TCs, subgroups, groups and also ports. Curr

mlxsw: spectrum_qdisc: Offload root TBF as port shaper

The Spectrum ASIC allows configuration of maximum shaper on all levels of
the scheduling hierarchy: TCs, subgroups, groups and also ports. Currently,
TBF always configures a subgroup. But a user could reasonably express the
intent to configure port shaper by putting TBF to a root position, around
ETS / PRIO. Accept this usage and offload appropriately.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

show more ...

mlxsw: spectrum_qdisc: Make RED, TBF offloads classful

Permit offloading qdiscs below RED and TBF. In order to avoid having to
implement trivial propagating callbacks for get_prio_bitmap and
get_tcl

mlxsw: spectrum_qdisc: Make RED, TBF offloads classful

Permit offloading qdiscs below RED and TBF. In order to avoid having to
implement trivial propagating callbacks for get_prio_bitmap and
get_tclass_num, extend mlxsw_sp_qdisc_get_prio_bitmap() and
..._get_tclass_num() to handle the lack of the callback as a cue to forward
the request to the parent.

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>

show more ...

mlxsw: spectrum_qdisc: Validate qdisc topology

A following patch will enable offloading qdiscs that are deeper than
directly under root qdisc. Currently the topology validation consists of
demanding

mlxsw: spectrum_qdisc: Validate qdisc topology

A following patch will enable offloading qdiscs that are deeper than
directly under root qdisc. Currently the topology validation consists of
demanding a root qdisc position for ETS and PRIO. Since RED and TBF are
considered classless, this is enough. In order to prevent some nonsensical
combinations when RED and TBF become classful, introduce a more general
topology validator.

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>

show more ...

mlxsw: spectrum_qdisc: Clean stats recursively when priomap changes

On Spectrum, there are no per-TC TX counters. Instead, mlxsw uses per-prio
counters and aggregates them according to the priomap.

mlxsw: spectrum_qdisc: Clean stats recursively when priomap changes

On Spectrum, there are no per-TC TX counters. Instead, mlxsw uses per-prio
counters and aggregates them according to the priomap. Therefore when
priomap changes, the counter base values need to be reset to reflect the
change. Previously, this was only done for the sole child qdisc, but a
following patch makes RED and TBF classful. Thus apply the request to the
whole sub-tree.

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>

show more ...

mlxsw: spectrum_qdisc: Unify graft validation

Qdisc graft operations have so far been reported at PRIO, ETS and RED, with
RED events ignored, because RED was not considered a classful qdisc. A
follo

mlxsw: spectrum_qdisc: Unify graft validation

Qdisc graft operations have so far been reported at PRIO, ETS and RED, with
RED events ignored, because RED was not considered a classful qdisc. A
following patch will make mlxsw recognize RED and TBF as classful qdiscs,
and thus it is necessary to validate grafting at these qdiscs as well.
Rename the existing graft validator to make it clear that it is a generic
function, and invoke for RED and TBF graft events as well. Drop the
unnecessary PRIO helper and invoke the graft validator directly for PRIO as
well.

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>

show more ...

mlxsw: spectrum_qdisc: Destroy children in mlxsw_sp_qdisc_destroy()

Currently ETS and PRIO are the only offloaded classful qdiscs. Since they
are both similar, their destroy handler is the same, and

mlxsw: spectrum_qdisc: Destroy children in mlxsw_sp_qdisc_destroy()

Currently ETS and PRIO are the only offloaded classful qdiscs. Since they
are both similar, their destroy handler is the same, and it handles
children destruction itself. But now it is possible to do it generically
for any classful qdisc. Therefore promote the recursive destruction from
the ETS handler to mlxsw_sp_qdisc_destroy(), so that RED and TBF pick it up
in follow-up patches.

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>

show more ...

mlxsw: spectrum_qdisc: Extract two helpers for handling future FIFOs

Extract from __mlxsw_sp_qdisc_ets_replace() two helpers for handling of one
future FIFO resp. reinitializing the array of future

mlxsw: spectrum_qdisc: Extract two helpers for handling future FIFOs

Extract from __mlxsw_sp_qdisc_ets_replace() two helpers for handling of one
future FIFO resp. reinitializing the array of future FIFOs.

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>

show more ...

mlxsw: spectrum_qdisc: Query tclass / priomap instead of caching it

Currently when keeping track of qdiscs, mlxsw notes the TC and priomap
corresponding to each qdisc. That is fine currently, as the

mlxsw: spectrum_qdisc: Query tclass / priomap instead of caching it

Currently when keeping track of qdiscs, mlxsw notes the TC and priomap
corresponding to each qdisc. That is fine currently, as there only ever is
one level of qdiscs to update: the direct children of ETS / PRIO. However
as deeper structures are made offloadable, ETS would need to update these
values for the complete subtree, and interim qdiscs would need to remember
to propagate the value.

Instead, reverse the responsibility: child qdiscs can ask their parent what
their TC and priomap are. ETS / PRIO know the answer right away, or there
are defaults for when the root qdisc does not assign them (e.g. when RED is
used as root qdisc). When RED and TBF become classful, they will simply
forward the request up to their parent.

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>

show more ...

mlxsw: spectrum_qdisc: Introduce per-TC ECN counters

The Qdisc code in mlxsw used to report a number of packets ECN-marked on a
port. Because reporting a per-port value as a per-TC value was mislead

mlxsw: spectrum_qdisc: Introduce per-TC ECN counters

The Qdisc code in mlxsw used to report a number of packets ECN-marked on a
port. Because reporting a per-port value as a per-TC value was misleading,
this was removed in commit 8a29581eb001 ("mlxsw: spectrum: Move the
ECN-marked packet counter to ethtool").

On Spectrum-3, a per-TC number of ECN-marked packets is available in per-TC
congestion counter group. Add a new array for the ECN counter, fetch the
values from the per-TC congestion group, and pick the value indicated by
tclass_num as appropriate.

On Spectrum-1 and Spectrum-2, this per-TC value is not available, and
zeroes will be reported, as they currently are.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

show more ...

mlxsw: spectrum_qdisc: Offload RED qevent mark

The RED "mark" qevent can be offloaded under similar conditions as the RED
"early_drop" qevent. Therefore recognize its binding type in the
TC_SETUP_BL

mlxsw: spectrum_qdisc: Offload RED qevent mark

The RED "mark" qevent can be offloaded under similar conditions as the RED
"early_drop" qevent. Therefore recognize its binding type in the
TC_SETUP_BLOCK handler and translate to the right SPAN trigger, with the
right set of supported actions.

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>

show more ...

mlxsw: spectrum_qdisc: Track permissible actions per binding

One block can be bound to several qevents. The qevent type that the block
is bound to determines which actions make sense in a given cont

mlxsw: spectrum_qdisc: Track permissible actions per binding

One block can be bound to several qevents. The qevent type that the block
is bound to determines which actions make sense in a given context. In the
particular case of mlxsw, trap cannot be offloaded on a RED mark qevent,
because the trap contract specifies that the packet is dropped in the HW
datapath, and the HW trigger that the action is offloaded to is always
forwarding the packet (in addition to marking in).

Therefore keep track of which actions are permissible at each binding
block. When an attempt is made to bind a certain action at a binding point
where it is not supported, bounce the request.

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>

show more ...

mlxsw: spectrum_qdisc: Distinguish between ingress and egress triggers

The following patches will configure the MLXSW_SP_SPAN_TRIGGER_ECN
mirroring trigger. This trigger is considered "egress", unli

mlxsw: spectrum_qdisc: Distinguish between ingress and egress triggers

The following patches will configure the MLXSW_SP_SPAN_TRIGGER_ECN
mirroring trigger. This trigger is considered "egress", unlike the
previously-offloaded _EARLY_DROP. Add a helper to spectrum_span,
mlxsw_sp_span_trigger_is_ingress(), to classify triggers to ingress and
egress. Pass result of this instead of hardcoding true when calling
mlxsw_sp_span_analyzed_port_get()/_put().

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>

show more ...

mlxsw: spectrum_qdisc: Pass extack to mlxsw_sp_qevent_entry_configure()

This function will report a new failure in the following patches.
Pass extack so that the failure is explicable.

Signed-off-b

mlxsw: spectrum_qdisc: Pass extack to mlxsw_sp_qevent_entry_configure()

This function will report a new failure in the following patches.
Pass extack so that the failure is explicable.

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>

show more ...

mlxsw: spectrum_qdisc: Pass handle, not band number to find_class()

In mlxsw Qdisc offload, find_class() is an operation that yields a qdisc
offload descriptor given a parental qdisc descriptor and

mlxsw: spectrum_qdisc: Pass handle, not band number to find_class()

In mlxsw Qdisc offload, find_class() is an operation that yields a qdisc
offload descriptor given a parental qdisc descriptor and a class handle. In
__mlxsw_sp_qdisc_ets_graft() however, a band number is passed to that
function instead of a handle. This can lead to a trigger of a WARN_ON
with the following splat:

WARNING: CPU: 3 PID: 808 at drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c:1356 __mlxsw_sp_qdisc_ets_graft+0x115/0x130 [mlxsw_spectrum]
[...]
Call Trace:
mlxsw_sp_setup_tc_prio+0xe3/0x100 [mlxsw_spectrum]
qdisc_offload_graft_helper+0x35/0xa0
prio_graft+0x176/0x290 [sch_prio]
qdisc_graft+0xb3/0x540
tc_modify_qdisc+0x56a/0x8a0
rtnetlink_rcv_msg+0x12c/0x370
netlink_rcv_skb+0x49/0xf0
netlink_unicast+0x1f6/0x2b0
netlink_sendmsg+0x1fb/0x410
____sys_sendmsg+0x1f3/0x220
___sys_sendmsg+0x70/0xb0
__sys_sendmsg+0x54/0xa0
do_syscall_64+0x3a/0x70
entry_SYSCALL_64_after_hwframe+0x44/0xae

Since the parent handle is not passed with the offload information, compute
it from the band number and qdisc handle.

Fixes: 28052e618b04 ("mlxsw: spectrum_qdisc: Track children per qdisc")
Reported-by: Maksym Yaremchuk <maksymy@nvidia.com>
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>

show more ...

mlxsw: spectrum_qdisc: Index future FIFOs by band number

mlxsw used to hold an array of qdiscs indexed by the TC number. In the
previous patch, it was changed to allocate child qdiscs dynamically, a

mlxsw: spectrum_qdisc: Index future FIFOs by band number

mlxsw used to hold an array of qdiscs indexed by the TC number. In the
previous patch, it was changed to allocate child qdiscs dynamically, and
they are now indexed by band number. Follow suit with the array of future
FIFOs.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Allocate child qdiscs dynamically

Instead of keeping qdiscs in globally-preallocated arrays, introduce a
per-qdisc-kind value num_classes, and then allocate the necessary chil

mlxsw: spectrum_qdisc: Allocate child qdiscs dynamically

Instead of keeping qdiscs in globally-preallocated arrays, introduce a
per-qdisc-kind value num_classes, and then allocate the necessary child
qdiscs (if any) based on that value. Since now dynamic allocation is
involved, mlxsw_sp_qdisc_replace() gets messy enough that it is worth it to
split it to two cases: a new qdisc allocation and a change of existing
qdisc. (Note that the change also includes what TC formally calls replace,
if the qdisc kind is the same.)

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Guard all qdisc accesses with a lock

The FIFO handler currently guards accesses to the future FIFO tracking by
asserting RTNL. In the future, the changes to the qdisc state wi

mlxsw: spectrum_qdisc: Guard all qdisc accesses with a lock

The FIFO handler currently guards accesses to the future FIFO tracking by
asserting RTNL. In the future, the changes to the qdisc state will be more
thorough, so other qdiscs will need this guarding is as well. In order
to not further the RTNL infestation, instead convert to a custom lock that
will guard accesses to the qdisc state.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Track children per qdisc

mlxsw currently allows a two-level structure of qdiscs: the root and
possibly a number of children. In order to support offloading more general
qdisc

mlxsw: spectrum_qdisc: Track children per qdisc

mlxsw currently allows a two-level structure of qdiscs: the root and
possibly a number of children. In order to support offloading more general
qdisc trees, introduce to struct mlxsw_sp_qdisc a pointer to child qdiscs.
Refer to the child qdiscs through this pointer, instead of going through
the tclass_qdiscs in qdisc_state. Additionally introduce a field
num_classes, which holds number of given qdisc's children.

Also introduce a generic function for walking qdisc trees. Rewrite
mlxsw_sp_qdisc_find() and _find_by_handle() to use the generic walker.

For now, keep the qdisc_state.tclass_qdisc, and just point root_qdiscs's
children to this array. Following patches will make the allocation dynamic.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Promote backlog reduction to mlxsw_sp_qdisc_destroy()

When a qdisc is removed, it is necessary to update the backlog value at its
parent--unless the qdisc is at root position.

mlxsw: spectrum_qdisc: Promote backlog reduction to mlxsw_sp_qdisc_destroy()

When a qdisc is removed, it is necessary to update the backlog value at its
parent--unless the qdisc is at root position. RED, TBF and FIFO all do
that, each separately. Since all of them need to do this, just promote the
operation directly to mlxsw_sp_qdisc_destroy(), instead of deferring it to
individual destructors. Since FIFO dtor thus becomes trivial, remove it.

Add struct mlxsw_sp_qdisc.parent to point at the parent qdisc. This will be
handy later as deeper structures are offloaded. Use the parent qdisc to
find the chain of parents whose backlog value needs to be updated.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Track tclass_num as int, not u8

tclass_num is just a number, a value that would be ordinarily passed around
as an int. (Which is unlike a u8 prio_bitmap.) In several places,
t

mlxsw: spectrum_qdisc: Track tclass_num as int, not u8

tclass_num is just a number, a value that would be ordinarily passed around
as an int. (Which is unlike a u8 prio_bitmap.) In several places,
tclass_num already is an int. Convert the remaining instances.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Drop an always-true condition

The function mlxsw_sp_qdisc_compare() is invoked a couple lines above this
check, which will bounce any requests where this condition does not ho

mlxsw: spectrum_qdisc: Drop an always-true condition

The function mlxsw_sp_qdisc_compare() is invoked a couple lines above this
check, which will bounce any requests where this condition does not hold.
Therefore drop it.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Simplify mlxsw_sp_qdisc_compare()

The purpose of this function is to filter out events that are related to
qdiscs that are not offloaded, or are not offloaded anymore. But the

mlxsw: spectrum_qdisc: Simplify mlxsw_sp_qdisc_compare()

The purpose of this function is to filter out events that are related to
qdiscs that are not offloaded, or are not offloaded anymore. But the
function is unnecessarily thorough:

- mlxsw_sp_qdisc pointer is never NULL in the context where it is called
- Two qdiscs with the same handle will never have different types. Even
when replacing one qdisc with another in the same class, Linux will not
permit handle reuse unless the qdisc type also matches.

Simplify the function by omitting these two unnecessary conditions.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_qdisc: Drop one argument from check_params callback

The mlxsw_sp_qdisc argument is not used in any of the actual callbacks.
Drop it.

Signed-off-by: Petr Machata <petrm@nvidia.com>
R

mlxsw: spectrum_qdisc: Drop one argument from check_params callback

The mlxsw_sp_qdisc argument is not used in any of the actual callbacks.
Drop it.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_span: Add SPAN probability rate support

Currently, every packet that matches a mirroring trigger (e.g., received
packets, buffer dropped packets) is mirrored. Spectrum-2 and later AS

mlxsw: spectrum_span: Add SPAN probability rate support

Currently, every packet that matches a mirroring trigger (e.g., received
packets, buffer dropped packets) is mirrored. Spectrum-2 and later ASICs
support mirroring with probability, where every 1 in N matched packets
is mirrored.

Extend the API that creates the binding between the trigger and the SPAN
agent with a probability rate parameter, which is an attribute of the
trigger. Set it to '1' to maintain existing behavior.

Subsequent patches will use it to perform more sophisticated sampling,
by mirroring packets to the CPU with probability.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...

mlxsw: spectrum_span: Add SPAN session identifier support

When packets are mirrored to the CPU, the trap identifier with which the
packets are trapped is determined according to the session identifi

mlxsw: spectrum_span: Add SPAN session identifier support

When packets are mirrored to the CPU, the trap identifier with which the
packets are trapped is determined according to the session identifier of
the SPAN agent performing the mirroring. Packets that are trapped for
the same logical reason (e.g., buffer drops) should use the same session
identifier.

Currently, a single session is implicitly supported (identifier 0) and
is used for packets that are mirrored to the CPU due to buffer drops
(e.g., early drop).

Subsequent patches are going to mirror packets to the CPU due to
sampling, which will require a different session identifier.

Prepare for that by making the session identifier an attribute of the
SPAN agent.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

show more ...