HV VHCA stats agent is responsible on running a preiodic rx/tx
packets/bytes stats update. Currently the supported format is version
MLX5_HV_VHCA_STATS_VERSION. Block ID 1 is dedicated for statistics data
transfer from the VF to the PF.
The reporter fetch the statistics data from all opened channels, fill it
in a buffer and send it to mlx5_hv_vhca_write_agent.
As the stats layer should include some metadata per block (sequence and
offset), the HV VHCA layer shall modify the buffer before actually send it
over block 1.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Control agent is responsible over of the control block (ID 0). It should
update the PF via this block about every capability change. In addition,
upon block 0 invalidate, it should activate all other supported agents
with data requests from the PF.
Upon agent create/destroy, the invalidate callback of the control agent
is being called in order to update the PF driver about this change.
The control agent is an integral part of HV VHCA and will be created
and destroy as part of the HV VHCA init/cleanup flow.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
HV VHCA is a layer which provides PF to VF communication channel based on
HyperV PCI config channel. It implements Mellanox's Inter VHCA control
communication protocol. The protocol contains control block in order to
pass messages between the PF and VF drivers, and data blocks in order to
pass actual data.
The infrastructure is agent based. Each agent will be responsible of
contiguous buffer blocks in the VHCA config space. This infrastructure will
bind agents to their blocks, and those agents can only access read/write
the buffer blocks assigned to them. Each agent will provide three
callbacks (control, invalidate, cleanup). Control will be invoked when
block-0 is invalidated with a command that concerns this agent. Invalidate
callback will be invoked if one of the blocks assigned to this agent was
invalidated. Cleanup will be invoked before the agent is being freed in
order to clean all of its open resources or deferred works.
Block-0 serves as the control block. All execution commands from the PF
will be written by the PF over this block. VF will ack on those by
writing on block-0 as well. Its format is described by struct
mlx5_hv_vhca_control_block layout.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add wrapper functions for HyperV PCIe read / write /
block_invalidate_register operations. This will be used as an
infrastructure in the downstream patch for software communication.
This will be enabled by default if CONFIG_PCI_HYPERV_INTERFACE is set.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To remove dependency on rtnl lock and prevent neigh update code from
accessing uninitialized flows when executing concurrently with tc, extend
mlx5e_tc_flow with 'init_done' completion. Modify helper
mlx5e_take_all_encap_flows() to wait for flow completion after obtaining
reference to it. Modify mlx5e_tc_encap_flows_del() and
mlx5e_tc_encap_flows_add() to skip flows that don't have OFFLOADED flag
set, which can happen if concurrent flow initialization failed.
This commit finishes neigh update refactoring for concurrent execution
started in previous change in this series.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In order to remove dependency on rtnl lock and allow neigh update workqueue
task to execute concurrently with tc, refactor mlx5e_rep_neigh_update() for
concurrent execution:
- Lock encap table when accessing encap entry to prevent concurrent
changes. To do this properly, the initial encap state check is moved from
mlx5e_rep_neigh_update() into mlx5e_rep_update_flows() to be performed
under encap_tbl_lock protection.
- Wait for encap to be fully initialized before accessing it by means of
'res_ready' completion.
- Add mlx5e_take_all_encap_flows() helper which is used to construct a
temporary list of flows and efi indexes that is used to access current
encap data in flow which can be attached to multiple encaps
simultaneously. Release the flows from temporary list after
encap_tbl_lock critical section. This is necessary because
mlx5e_flow_put() can't be called while holding encap_tbl_lock.
- Modify mlx5e_tc_encap_flows_add() and mlx5e_tc_encap_flows_del() to work
with user-provided list of flows built by mlx5e_take_all_encap_flows(),
instead of traversing encap flow list directly.
This is first step in complex neigh update refactoring, which is finished
by following commit in this series.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In order to remove dependency on rtnl lock and allow neigh used value
update workqueue task to execute concurrently with tc, refactor
mlx5e_tc_update_neigh_used_value() for concurrent execution:
- Lock encap table when accessing encap entry to prevent concurrent
changes.
- Save offloaded encap flows to temporary list and release them after encap
entry is updated. Add mlx5e_put_encap_flow_list() helper which is
intended to be shared with neigh update code in following patch in this
series. This is necessary because mlx5e_flow_put() can't be called while
holding encap_tbl_lock.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Rcu-ify mlx5e_neigh_hash_entry->encap_list by changing operations on encap
list to their rcu counterparts and extending encap structure with rcu_head
to free the encap instances after rcu grace period. Use rcu read lock when
traversing encap list. Implement helper mlx5e_get_next_valid_encap()
function that is used by mlx5e_tc_update_neigh_used_value() to safely
iterate over valid entries of nhe->encap_list.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
To remove dependency on rtnl lock, always take neigh update encap lock when
modifying neigh update hash table and list. Originally, this lock was only
used to synchronize with netevent handler function, which is called from bh
context and cannot use rtnl lock for synchronization. Take lock in encap
entry attach function to prevent concurrent modifications of neigh update
hash table and list.
Taking the encap lock when creating new nhe introduces a problem that we
need to allocate new entry with sleeping GFP_KERNEL flag while holding a
spinlock. However, since previous patch in this series has already
converted lookup in netevent handler function to user rcu read lock instead
of encap lock, we can safely convert the lock type to mutex.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
To remove dependency on rtnl lock and to allow unlocked iteration over list
of neigh hash entries, extend nhe with rcu. Change operations on neigh list
to their rcu counterparts and free neigh hash entry with rcu timeout.
Introduce mlx5e_get_next_nhe() helper that is used to iterate over rcu
neigh list with reference to nhe taken.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Neigh entry has reference counter, however it is only used when scheduling
neigh update event. In all other cases reference to neigh entry is not
taken while working with it. Neigh code relies on synchronization provided
by rtnl lock and uses encap list size as implicit reference counter.
To remove dependency on rtnl lock, always take reference to neigh entry
while using it. Remove neigh entry from hash table and delete it only when
reference counter reaches zero. This can result spurious neigh update
events, when there is an event on entry that has zero encaps attached.
However, such events are rare and properly handled by neigh update handler.
Extend encap entry with reference to neigh hash entry in order to be able
to directly release it when encap is detached, instead of lookup nhe by key
through hash table. Extend nhe with reference to device priv structure to
guarantee correctness when nhe is used with stack devices, bond setup, in
which case it is non-trivial to determine correct device when releasing the
nhe.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
As a preparation for following refactoring that removes rtnl lock
dependency from neigh hash entry handlers, extract code that enqueues neigh
update work into standalone function. This commit doesn't change
functionality.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Register supported packet traps (layer 2 drops only, currently) and
associated trap group with devlink during driver initialization.
The amount of traffic generated by these packet drop traps is capped at
10Kpps to ensure the CPU is not overwhelmed by incoming packets.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Discard trap groups are defined in a different enum so that they could
all share the same policer ID: MLXSW_REG_HTGT_TRAP_GROUP_MAX + 1.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Subsequent patches will add discard traps support in mlxsw. The driver
cannot configure such traps with a normal trap action, but needs to use
exception trap action, which also increments an error counter.
On the other hand, when these traps are initialized or set to drop
action, they should use the default drop action set by the firmware.
This guarantees that when the feature is disabled we get the exact same
behavior as before the feature was introduced.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Up until now the action of a trap was never changed during its lifetime.
This is going to change by subsequent patches that will allow devlink to
control the action of certain traps.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jason Gunthorpe says:
====================
This is a collection of general cleanups for ODP to clarify some of the
flows around umem creation and use of the interval tree.
====================
The branch is based on v5.3-rc5 due to dependencies
* odp_fixes:
RDMA/mlx5: Use odp instead of mr->umem in pagefault_mr
RDMA/mlx5: Use ib_umem_start instead of umem.address
RDMA/core: Make invalidate_range a device operation
RDMA/odp: Use kvcalloc for the dma_list and page_list
RDMA/odp: Check for overflow when computing the umem_odp end
RDMA/odp: Provide ib_umem_odp_release() to undo the allocs
RDMA/odp: Split creating a umem_odp from ib_umem_get
RDMA/odp: Make the three ways to create a umem_odp clear
RMDA/odp: Consolidate umem_odp initialization
RDMA/odp: Make it clearer when a umem is an implicit ODP umem
RDMA/odp: Iterate over the whole rbtree directly
RDMA/odp: Use the common interval tree library instead of generic
RDMA/mlx5: Fix MR npages calculation for IB_ACCESS_HUGETLB
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Use different namespaces for bypass and switchdev loopback because they
have different priorities and default table miss action requirement:
1. bypass: with multiple priorities support, and
MLX5_FLOW_TABLE_MISS_ACTION_DEF as the default table miss action;
2. switchdev loopback: with single priority support, and
MLX5_FLOW_TABLE_MISS_ACTION_SWITCH_DOMAIN as the default table miss
action.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Currently all the namespaces under the same steering domain share the same
default table miss action, however in some situations (e.g., RDMA RX)
different actions are required. This patch adds a per-namespace default
table miss action instead of using the miss action of the steering domain.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Now that the single target build descends into sub-directories in the
same way as the normal build, these dummy Makefiles are not needed
any more.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Fix documentation of mlx5_eq_enable/disable to cleanup compiler warnings.
drivers/net/ethernet/mellanox/mlx5/core//eq.c:334:
warning: Function parameter or member 'dev' not described in 'mlx5_eq_enable'
warning: Function parameter or member 'eq' not described in 'mlx5_eq_enable'
warning: Function parameter or member 'nb' not described in 'mlx5_eq_enable'
drivers/net/ethernet/mellanox/mlx5/core//eq.c:355:
warning: Function parameter or member 'dev' not described in 'mlx5_eq_disable'
warning: Function parameter or member 'eq' not described in 'mlx5_eq_disable'
warning: Function parameter or member 'nb' not described in 'mlx5_eq_disable'
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Previously, mlx5_cleanup_fc_stats() would cleanup the flow counter
pool beofre releasing all the counters to it, which would result in
flow counter bulks not getting freed. Resolve this by changing the
order in which elements of fc_stats are cleaned up, so that the flow
counter pool is cleaned up after all the counters are released.
Also move cleanup actions for freeing the bulk query memory and
destroying the idr to the end of mlx5_cleanup_fc_stats().
Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add support for report and recovery from error on completion on RQ by
setting the queue back to ready state. Handle only errors with a
syndrome indicating the RQ might enter error state and could be
recovered.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Just to be aligned with the MPWQE handlers, handle RX WQE with error
for legacy RQs in the top RX handlers, just before calling skb_from_cqe().
CQE error handling will now be called at the same stage regardless of
the RQ type or netdev mode NIC, Representor, IPoIB, etc ..
This will be useful for down stream patch to improve error CQE
handling.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add support for report and recovery from rx timeout. On driver open we
post NOP work request on the rx channels to trigger napi in order to
fillup the rx rings. In case napi wasn't scheduled due to a lost
interrupt, perform EQ recovery.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add support for report and recovery from error on completion on ICOSQ.
Deactivate RQ and flush, then deactivate ICOSQ. Set the queue back to
ready state (firmware) and reset the ICOSQ and the RQ (software
resources). Finally, activate the ICOSQ and the RQ.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Align ICOSQ open/close behaviour with RQ and SQ. Split open flow into
open and activate where open handles creation and activate enables the
queue. Do a symmetric thing in close flow: split into close and
deactivate.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Introduce helper functions for create and destroy reporters and update
channels. In the following patch, rx reporter is added and it will use
these helpers too.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The following patches in the set enhance the diagnostics info of tx
reporter. Therefore, it is better to pass a pointer to the SQ for
further data extraction.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Prepare for code sharing with rx reporter, which is added in the
following patches in the set. Introduce a generic error_ctx for
agnostic recovery despatch.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Change from mlx5e_tx_reporter_* to mlx5e_reporter_tx_*. In the following
patches in the set rx reporter is added, the new naming convention is
more uniformed.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
tc transparently maps the software priority number to hardware. Update
it to pass the major priority which is what most drivers expect. Update
drivers too so they do not need to lshift the priority field of the
flow_cls_common_offload object. The stmmac driver is an exception, since
this code assumes the tc software priority is fine, therefore, lshift it
just to be conservative.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit adds support for the new need_wakeup feature of AF_XDP. The
applications can opt-in by using the XDP_USE_NEED_WAKEUP bind() flag.
When this feature is enabled, some behavior changes:
RX side: If the Fill Ring is empty, instead of busy-polling, set the
flag to tell the application to kick the driver when it refills the Fill
Ring.
TX side: If there are pending completions or packets queued for
transmission, set the flag to tell the application that it can skip the
sendto() syscall and save time.
The performance testing was performed on a machine with the following
configuration:
- 24 cores of Intel Xeon E5-2620 v3 @ 2.40 GHz
- Mellanox ConnectX-5 Ex with 100 Gbit/s link
The results with retpoline disabled:
| without need_wakeup | with need_wakeup |
|----------------------|----------------------|
| one core | two cores | one core | two cores |
-------|----------|-----------|----------|-----------|
txonly | 20.1 | 33.5 | 29.0 | 34.2 |
rxdrop | 0.065 | 14.1 | 12.0 | 14.1 |
l2fwd | 0.032 | 7.3 | 6.6 | 7.2 |
"One core" means the application and NAPI run on the same core. "Two
cores" means they are pinned to different cores.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Two XSK tasks are performed during NAPI polling, that are not bound to
hardware interrupts: TXing packets and polling for frames in the Fill
Ring. They are special in a way that the hardware doesn't know about
these tasks, so it doesn't trigger interrupts if there is still some
work to be done, it's our driver's responsibility to ensure NAPI will be
rescheduled if needed.
Create a new function to handle these tasks and move the corresponding
code from mlx5e_napi_poll to the new function to improve modularity and
prepare for the changes in the following patch.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit replaces ndo_xsk_async_xmit with ndo_xsk_wakeup. This new
ndo provides the same functionality as before but with the addition of
a new flags field that is used to specifiy if Rx, Tx or both should be
woken up. The previous ndo only woke up Tx, as implied by the
name. The i40e and ixgbe drivers (which are all the supported ones)
are updated with this new interface.
This new ndo will be used by the new need_wakeup functionality of XDP
sockets that need to be able to wake up both Rx and Tx driver
processing.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cited patch deleted ethtool flash device support, as ethtool core can
fallback into devlink flash callback. However, this is supported only if
there is a devlink port registered over the corresponding netdevice.
As mlx5e do not have devlink port support over native netdevice, it broke
the ability to flash device via ethtool.
This patch re-add the ethtool callback to avoid user functionality breakage
when trying to flash device via ethtool.
Fixes: 9c8bca2637 ("mlx5: Move firmware flash implementation to devlink")
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add a missing spinlock around XSKICOSQ usage at the activation stage,
because there is a race between a configuration change and the
application calling sendto().
Fixes: db05815b36 ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In mlx4_en_config_rss_steer(), 'rss_map->indir_qp' is allocated through
kzalloc(). After that, mlx4_qp_alloc() is invoked to configure RSS
indirection. However, if mlx4_qp_alloc() fails, the allocated
'rss_map->indir_qp' is not deallocated, leading to a memory leak bug.
To fix the above issue, add the 'qp_alloc_err' label to free
'rss_map->indir_qp'.
Fixes: 4931c6ef04 ("net/mlx4_en: Optimized single ring steering")
Signed-off-by: Wenwen Wang <wenwen@cs.uga.edu>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Merging tip of mlx5-next in order to get changes related to adding
XRQ support to the DEVX interface needed prior to the following two
patches.
Signed-off-by: Doug Ledford <dledford@redhat.com>
