Remove the explicit call to mlx5_cmd_comp_handler on MLX5_EVENT_TYPE_CMD
and let command interface to register its own handler when its ready.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Remove the explicit call to mlx5_core_req_pages_handler on
MLX5_EVENT_TYPE_PAGE_REQUEST and let FW page logic to register its own
handler when its ready.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Remove the explicit call to mlx5_pps_event on MLX5_EVENT_TYPE_PPS_EVENT
and let clock logic to register its own handler when its ready.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Use atomic_notifier_chain to fire firmware events at internal mlx5 core
components such as eswitch/fpga/clock/FW tracer/etc.., this is to
avoid explicit calls from low level mlx5_core to upper components and to
simplify the mlx5_core API for future developments.
Simply provide register/unregister notifiers API and call the notifier
chain on firmware async events.
Example: to subscribe to a FW event:
struct mlx5_nb port_event;
MLX5_NB_INIT(&port_event, port_event_handler, PORT_CHANGE);
mlx5_eq_notifier_register(mdev, &port_event);
where:
- port_event_handler is the notifier block callback.
- PORT_EVENT is the suffix of MLX5_EVENT_TYPE_PORT_CHANGE.
The above will guarantee that port_event_handler will receive all FW
events of the type MLX5_EVENT_TYPE_PORT_CHANGE.
To receive all FW/HW events one can subscribe to
MLX5_EVENT_TYPE_NOTIFY_ANY.
The next few patches will start moving all mlx5 core components to use
this new API and cleanup mlx5_eq_async_int misx handler from component
explicit calls and specific logic.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add mlx5_eq_{create/destroy}_generic APIs and EQE access methods, for
mlx5 core consumers generic EQs.
This API will be used in downstream patch to move page fault (RDMA ODP)
EQ logic into mlx5_ib rdma driver, hence it will use a generic EQ.
Current mlx5 EQ allocation scheme:
On load mlx5 allocates 4 (for async) + #cores (for data completions)
MSIX vectors, mlx5 core will assign 3 MSIX vectors for internal async
EQs and will use all of the #cores MSIX vectors for completion EQs,
(One vector is going to be reserved for a generic EQ).
After this patch an external user (e.g mlx5_ib) of mlx5_core
can use this new API to create new generic EQs with the reserved msix
vector index for that eq.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
In mlx5 we have three types of usages for EQs,
1. Asynchronous EQs, used internally by mlx5 core for
a. FW command completions
b. FW page requests
c. one EQ for all other Asynchronous events
2. Completion EQs, used for CQ completion (we create one per core)
3. *Special type of EQ (page fault) used for RDMA on demand paging
(ODP).
*The 3rd type shouldn't be special at least in mlx5 core, it is yet
another async events EQ with specific use case, it will be removed in
the next two patches, and will completely move its logic to mlx5_ib,
as it is rdma specific.
In this patch we remove use case (eq type) specific fields from
struct mlx5_eq into a new eq type specific structures.
struct mlx5_eq_async;
truct mlx5_eq_comp;
struct mlx5_eq_pagefault;
Separate between their type specific flows.
In the future we will allow users to create there own generic EQs.
for now we will allow only one for ODP in next patches.
We will introduce event listeners registration API for those who
want to receive mlx5 async events.
After that mlx5 eq handling will be clean from feature/user specific
handling.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Move unnecessary EQ table structures and declaration from the
public include/linux/mlx5/driver.h into the private area of mlx5_core
and into eq.c/eq.h.
Introduce new mlx5 EQ APIs:
mlx5_comp_vectors_count(dev);
mlx5_comp_irq_get_affinity_mask(dev, vector);
And use them from mlx5_ib or mlx5e netdevice instead of direct access to
mlx5_core internal structures.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
irq_info and rmap are EQ properties of the driver, and only needed for
EQ objects, move them to the eq_table EQs database structure.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
eq->index is used only for completion EQs and is assigned to be
the completion eq index, it is used only when traversing the completion
eqs list, and it can be calculated dynamically, thus remove the
eq->index field.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Currently the cpu affinity hint mask for completion EQs is stored and
read from the wrong place, since reading and storing is done from the
same index, there is no actual issue with that, but internal irq_info
for completion EQs stars at MLX5_EQ_VEC_COMP_BASE offset in irq_info
array, this patch changes the code to use the correct offset to store
and read the IRQ affinity hint.
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Sometimes upper layers may want to prevent the destruction of a core
resource for a period of time while work on that resource is in
progress. Add API to support this.
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Fix wrong offsets of reserved fields in ifc file.
Issues found using pahole.
Signed-off-by: Gal Pressman <pressmangal@gmail.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Adapt XRC SRQ to the latest HW specification with fixed definition
around umem valid bits. The previous definition relied on a bit which
was taken for other purposes in legacy FW.
Fixes: bd37197554 ("net/mlx5: Update mlx5_ifc with DEVX UID bits")
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Pull rdma updates from Jason Gunthorpe:
"This has been a smaller cycle with many of the commits being smallish
code fixes and improvements across the drivers.
- Driver updates for bnxt_re, cxgb4, hfi1, hns, mlx5, nes, qedr, and
rxe
- Memory window support in hns
- mlx5 user API 'flow mutate/steering' allows accessing the full
packet mangling and matching machinery from user space
- Support inter-working with verbs API calls in the 'devx' mlx5 user
API, and provide options to use devx with less privilege
- Modernize the use of syfs and the device interface to use attribute
groups and cdev properly for uverbs, and clean up some of the core
code's device list management
- More progress on net namespaces for RDMA devices
- Consolidate driver BAR mmapping support into core code helpers and
rework how RDMA holds poitners to mm_struct for get_user_pages
cases
- First pass to use 'dev_name' instead of ib_device->name
- Device renaming for RDMA devices"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (242 commits)
IB/mlx5: Add support for extended atomic operations
RDMA/core: Fix comment for hw stats init for port == 0
RDMA/core: Refactor ib_register_device() function
RDMA/core: Fix unwinding flow in case of error to register device
ib_srp: Remove WARN_ON in srp_terminate_io()
IB/mlx5: Allow scatter to CQE without global signaled WRs
IB/mlx5: Verify that driver supports user flags
IB/mlx5: Support scatter to CQE for DC transport type
RDMA/drivers: Use core provided API for registering device attributes
RDMA/core: Allow existing drivers to set one sysfs group per device
IB/rxe: Remove unnecessary enum values
RDMA/umad: Use kernel API to allocate umad indexes
RDMA/uverbs: Use kernel API to allocate uverbs indexes
RDMA/core: Increase total number of RDMA ports across all devices
IB/mlx4: Add port and TID to MAD debug print
IB/mlx4: Enable debug print of SMPs
RDMA/core: Rename ports_parent to ports_kobj
RDMA/core: Do not expose unsupported counters
IB/mlx4: Refer to the device kobject instead of ports_parent
RDMA/nldev: Allow IB device rename through RDMA netlink
...
net/sched/cls_api.c has overlapping changes to a call to
nlmsg_parse(), one (from 'net') added rtm_tca_policy instead of NULL
to the 5th argument, and another (from 'net-next') added cb->extack
instead of NULL to the 6th argument.
net/ipv4/ipmr_base.c is a case of a bug fix in 'net' being done to
code which moved (to mr_table_dump)) in 'net-next'. Thanks to David
Ahern for the heads up.
Signed-off-by: David S. Miller <davem@davemloft.net>
Added "Per lane raw errors" capability bit in
Ports Capabilities Mask (PCAM) enhanced features
layout.
This bit determines if the fields "phy_raw_errors_laneX"
in "Physical Layer statistical" counters group are supported.
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Added FEC related fields to PPLM layout.
These fields are needed to set and query FEC policy
for different link speeds.
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Take struct mlx5_frag_buf out of mlx5_frag_buf_ctrl, as it is not
needed to manage and control the datapath of the fragmented buffers API.
struct mlx5_frag_buf contains control info to manage the allocation
and de-allocation of the fragmented buffer.
Its fields are not relevant for datapath, so here I take them out of the
struct mlx5_frag_buf_ctrl, except for the fragments array itself.
In addition, modified mlx5_fill_fbc to initialise the frags pointers
as well. This implies that the buffer must be allocated before the
function is called.
A set of type-specific *_get_byte_size() functions are replaced by
a generic one.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
If no-append flag is set, we will add a new FTE, instead of appending
the actions of the inserted rule when the same match already exists.
While here, move the has_flow_tag boolean indicator to be a flag too.
This patch doesn't change any functionality.
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanmox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Towards supporting multi-chains and priorities, split the FDB fast path
to multiple namespaces (sub namespaces), each with multiple priorities.
This patch adds a new flow steering type, FS_TYPE_PRIO_CHAINS, which is
like current FS_TYPE_PRIO, but may contain only namespaces, and those
will be in parallel to one another in terms of managing of the flow
tables connections inside them. Meaning, while searching for the next
or previous flow table to connect for a new table inside such namespace
we skip the parallel namespaces in the same level under the
FS_TYPE_PRIO_CHAINS prio we originated from.
We use this new type for splitting the fast path prio into multiple
parallel namespaces, each containing normal prios.
The prios inside them (and their tables) will be connected to one
another, but not from one parallel namespace to another, instead the
last prio in each namespace will be connected to the next prio in
the containing FDB namespace, which is the slow path prio.
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Acked-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
If set, the firmware supports creating of flow tables with encap
enabled while VFs are configured, if we already created one
(restriction still applies on the first creation).
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Currently, when a flow rule is created using the FS core layer, the caller
has to pass the entire flow counter object and not just the counter HW
handle (ID). This requires both the FS core and the caller to have
knowledge about the inner implementation of the FS layer flow counters
cache and limits the possible users.
Move to use the counter ID across the place when dealing with flows.
Doing this decoupling, now can we privatize the inner implementation
of the flow counters.
Signed-off-by: Mark Bloch <markb@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
There's no real reason for the e-switch logic to manage the creation of
counters for offloaded flows. The API already has the directive for the
caller to denote they want to attach a counter to the created flow.
As such, we go and move the management of flow counters to the mlx5e
tc offload logic. This also lets us remove an inelegant interface where
the FS layer had to provide a way to retrieve a counter from a flow rule.
Signed-off-by: Mark Bloch <markb@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
mlx5 updates for both net-next and rdma-next
* 'mlx5-next' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux: (21 commits)
net/mlx5: Expose DC scatter to CQE capability bit
net/mlx5: Update mlx5_ifc with DEVX UID bits
net/mlx5: Set uid as part of DCT commands
net/mlx5: Set uid as part of SRQ commands
net/mlx5: Set uid as part of SQ commands
net/mlx5: Set uid as part of RQ commands
net/mlx5: Set uid as part of QP commands
net/mlx5: Set uid as part of CQ commands
net/mlx5: Rename incorrect naming in IFC file
net/mlx5: Export packet reformat alloc/dealloc functions
net/mlx5: Pass a namespace for packet reformat ID allocation
net/mlx5: Expose new packet reformat capabilities
{net, RDMA}/mlx5: Rename encap to reformat packet
net/mlx5: Move header encap type to IFC header file
net/mlx5: Break encap/decap into two separated flow table creation flags
net/mlx5: Add support for more namespaces when allocating modify header
net/mlx5: Export modify header alloc/dealloc functions
net/mlx5: Add proper NIC TX steering flow tables support
net/mlx5: Cleanup flow namespace getter switch logic
net/mlx5: Add memic command opcode to command checker
...
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Extended atomic operations cmp&swp and fetch&add is a Mellanox
feature extending the standard atomic operation to use, varied
operand sizes, as apposed to normal atomic operation that use
an 8 byte operand only.
Extended atomics allows masking the results and arguments.
This patch configures QP to support extended atomic operation
with the maximum size possible, as exposed by HCA capabilities.
Signed-off-by: Yonatan Cohen <yonatanc@mellanox.com>
Reviewed-by: Guy Levi <guyle@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
dc_req_scat_data_cqe capability bit determines
if requester scatter to cqe is available for 64 bytes CQE over
DC transport type.
Signed-off-by: Yonatan Cohen <yonatanc@mellanox.com>
Reviewed-by: Guy Levi <guyle@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
mlx5e netdevice used to calculate fragment edges by a call to
mlx5_wq_cyc_get_frag_size(). This calculation did not give the correct
indication for queues smaller than a PAGE_SIZE, (broken by default on
PowerPC, where PAGE_SIZE == 64KB). Here it is replaced by the correct new
calls/API.
Since (TX/RX) Work Queues buffers are fragmented, here we introduce
changes to the API in core driver, so that it gets a stride index and
returns the index of last stride on same fragment, and an additional
wrapping function that returns the number of physically contiguous
strides that can be written contiguously to the work queue.
This obsoletes the following API functions, and their buggy
usage in EN driver:
* mlx5_wq_cyc_get_frag_size()
* mlx5_wq_cyc_ctr2fragix()
The new API improves modularity and hides the details of such
calculation for mlx5e netdevice and mlx5_ib rdma drivers.
New calculation is also more efficient, and improves performance
as follows:
Packet rate test: pktgen, UDP / IPv4, 64byte, single ring, 8K ring size.
Before: 16,477,619 pps
After: 17,085,793 pps
3.7% improvement
Fixes: 3a2f703312 ("net/mlx5: Use order-0 allocations for all WQ types")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
netdev has several interfaces that expect to call alloc_netdev_mqs from
the core code, with the driver only providing the arguments. This is
incompatible with the rdma_netdev interface that returns the netdev
directly.
Thus re-organize the API used by ipoib so that the verbs core code calls
alloc_netdev_mqs for the driver. This is done by allowing the drivers to
provide the allocation parameters via a 'get_params' callback and then
initializing an allocated netdev as a second step.
Fixes: cd565b4b51 ("IB/IPoIB: Support acceleration options callbacks")
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Denis Drozdov <denisd@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Saeed Mahameed says:
====================
mlx5-updates-2018-10-03
mlx5 core driver and ethernet netdev updates, please note there is a small
devlink releated update to allow extack argument to eswitch operations.
From Eli Britstein,
1) devlink: Add extack argument to the eswitch related operations
2) net/mlx5e: E-Switch, return extack messages for failures in the e-switch devlink callbacks
3) net/mlx5e: Add extack messages for TC offload failures
From Eran Ben Elisha,
4) mlx5e: Add counter for aRFS rule insertion failures
From Feras Daoud
5) Fast teardown support for mlx5 device
This change introduces the enhanced version of the "Force teardown" that
allows SW to perform teardown in a faster way without the need to reclaim
all the FW pages.
Fast teardown provides the following advantages:
1- Fix a FW race condition that could cause command timeout
2- Avoid moving to polling mode
3- Close the vport to prevent PCI ACK to be sent without been scatter
to memory
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Minor conflict in net/core/rtnetlink.c, David Ahern's bug fix in 'net'
overlapped the renaming of a netlink attribute in net-next.
Signed-off-by: David S. Miller <davem@davemloft.net>
Today mlx5 devices support two teardown modes:
1- Regular teardown
2- Force teardown
This change introduces the enhanced version of the "Force teardown" that
allows SW to perform teardown in a faster way without the need to reclaim
all the pages.
Fast teardown provides the following advantages:
1- Fix a FW race condition that could cause command timeout
2- Avoid moving to polling mode
3- Close the vport to prevent PCI ACK to be sent without been scatter
to memory
Signed-off-by: Feras Daoud <ferasda@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The system image guid is a read-only field which is used by the TC
offloads code to determine if two mlx5 devices belong to the same
ASIC while adding flows.
Read this once and save it on the core device rather than querying each
time an offloaded flow is added.
Signed-off-by: Alaa Hleihel <alaa@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
If the peer device was already unbound, then do not attempt to modify
it's resources, otherwise we will crash on dereferencing non-existing
device.
Fixes: 5c65c564c9 ("net/mlx5e: Support offloading TC NIC hairpin flows")
Signed-off-by: Alaa Hleihel <alaa@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add DEVX information to WQ, SRQ, CQ, TIR, TIS, QP,
RQ, XRCD, PD, MKEY and MCG.
Each object that is created/destroyed/modified via verbs will
be stamped with a UID based on its user context. This is already
done for DEVX objects commands.
This will enable the firmware to enforce the usage of kernel objects
from the DEVX flow by validating that the same UID is used and the
resources are really related to the same user.
The addition of *_valid fields are needed to distinguish
how various addresses are passed.
For non-DEVX callers, all those fields will be zero.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Set uid as part of DCT commands so that the firmware can manage the
DCT object in a secured way.
That will enable using a DCT that was created by verbs application
to be used by the DEVX flow in case the uid is equal.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Set uid as part of SRQ commands so that the firmware can manage the
SRQ object in a secured way.
That will enable using an SRQ that was created by verbs application
to be used by the DEVX flow in case the uid is equal.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Set uid as part of SQ commands so that the firmware can manage the
SQ object in a secured way.
That will enable using an SQ that was created by verbs application
to be used by the DEVX flow in case the uid is equal.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Set uid as part of RQ commands so that the firmware can manage the
RQ object in a secured way.
That will enable using an RQ that was created by verbs application
to be used by the DEVX flow in case the uid is equal.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Set uid as part of QP commands so that the firmware can manage the
QP object in a secured way.
That will enable using a QP that was created by verbs application to
be used by the DEVX flow in case the uid is equal.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Set uid as part of CQ commands so that the firmware can manage the CQ
object in a secured way.
The firmware should mark this CQ with the given uid so that it can
be used later on only by objects with the same uid.
Upon DEVX flows that use this CQ (e.g. create QP command), the
pointed CQ must have the same uid as of the issuer uid command.
When a command is issued with uid=0 it means that the issuer of the
command is trusted (i.e. kernel), in that case any pointed object
can be used regardless of its uid.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Changed "priv.clock.lock" lock from 'rw_lock' to 'seq_lock'
in order to improve packet rate performance.
Tested on Intel(R) Xeon(R) CPU E5-2660 v2 @ 2.20GHz.
Sent 64b packets between two peers connected by ConnectX-5,
and measured packet rate for the receiver in three modes:
no time-stamping (base rate)
time-stamping using rw_lock (old lock) for critical region
time-stamping using seq_lock (new lock) for critical region
Only the receiver time stamped its packets.
The measured packet rate improvements are:
Single flow (multiple TX rings to single RX ring):
without timestamping: 4.26 (M packets)/sec
with rw-lock (old lock): 4.1 (M packets)/sec
with seq-lock (new lock): 4.16 (M packets)/sec
1.46% improvement
Multiple flows (multiple TX rings to six RX rings):
without timestamping: 22 (M packets)/sec
with rw-lock (old lock): 11.7 (M packets)/sec
with seq-lock (new lock): 21.3 (M packets)/sec
82.05% improvement
The packet rate improvement is due to the lack of atomic operations
for the 'readers' by the seq-lock.
Since there are much more 'readers' than 'writers' contention
on this lock, almost all atomic operations are saved.
this results in a dramatic decrease in overall
cache misses.
Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Previous patch in series changed flow counter storage structure from
rb_tree to linked list in order to improve flow counter traversal
performance. The drawback of such solution is that flow counter lookup by
id becomes linear in complexity.
Store pointers to flow counters in idr in order to improve lookup
performance to logarithmic again. Idr is non-intrusive data structure and
doesn't require extending flow counter struct with new elements. This means
that idr can be used for lookup, while linked list from previous patch is
used for traversal, and struct mlx5_fc size is <= 2 cache lines.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Amir Vadai <amir@vadai.me>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In order to improve performance of flow counter stats query loop that
traverses all configured flow counters, replace rb_tree with double-linked
list. This change improves performance of traversing flow counters by
removing the tree traversal. (profiling data showed that call to rb_next
was most top CPU consumer)
However, lookup of flow flow counter in list becomes linear, instead of
logarithmic. This problem is fixed by next patch in series, which adds idr
for fast lookup. Idr is to be used because it is not an intrusive data
structure and doesn't require adding any new members to struct mlx5_fc,
which allows its control data part to stay <= 1 cache line in size.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Amir Vadai <amir@vadai.me>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In order to prevent flow counters stats work function from traversing whole
flow counters tree while searching for deleted flow counters, new list to
store deleted flow counters is added to struct mlx5_fc_stats. Lockless
NULL-terminated single linked list data type is used due to following
reasons:
- This use case only needs to add single element to list and
remove/iterate whole list. Lockless list doesn't require any additional
synchronization for these operations.
- First cache line of flow counter data structure only has space to store
single additional pointer, which precludes usage of double linked list.
Remove flow counter 'deleted' flag that is no longer needed.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Amir Vadai <amir@vadai.me>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
