The DPAA2 DMA driver is an implementation of the dmadev APIs,
that provide means to initiate a DMA transaction from CPU.
Earlier this was part of RAW driver, but with DMA drivers
added as separate flavor of drivers, this driver is being
moved to DMA drivers.
Signed-off-by: Nipun Gupta <nipun.gupta@nxp.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
With DMA devices supported as a separate flavor of devices,
the DPAA2 QDMA driver is moved in the DMA devices.
This change removes the DPAA2 QDMA driver from raw devices.
Signed-off-by: Nipun Gupta <nipun.gupta@nxp.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Within ACL rule IPv6 address can be represented in different ways:
either as 4x4B fields, or as 2x8B fields.
Till now, only first format was supported.
Extend test-acl to support both formats, mainly for testing and
demonstrating purposes.
To control desired behavior '--ipv6' command-line option is extended
to accept an optional argument:
To be more precise:
'--ipv6' - use 4x4B fields format (default behavior)
'--ipv6=4B' - use 4x4B fields format (default behavior)
'--ipv6=8B' - use 2x8B fields format
Also replaced home brewed IPv4/IPv6 address parsing with inet_pton() calls.
Signed-off-by: Konstantin Ananyev <konstantin.v.ananyev@yandex.ru>
In theory ACL library allows fields with 8B long.
Though in practice they usually not used, not tested,
and as was revealed by Ido, this functionality is not working properly.
There are few places inside ACL build code-path that need to be addressed.
Bugzilla ID: 673
Fixes: dc276b5780 ("acl: new library")
Cc: stable@dpdk.org
Reported-by: Ido Goshen <ido@cgstowernetworks.com>
Signed-off-by: Konstantin Ananyev <konstantin.v.ananyev@yandex.ru>
Tested-by: Ido Goshen <ido@cgstowernetworks.com>
The AltiVec header file is defining "vector", except in C++ build.
The keyword "vector" may conflict easily.
As a rule, it is better to use the alternative keyword "__vector",
so we will be able to #undef vector after including AltiVec header.
Later it may become possible to #undef vector in rte_altivec.h
with a compatibility breakage.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Reviewed-by: David Christensen <drc@linux.vnet.ibm.com>
Current pmd_perf_autotest() in continuous mode tries
to enqueue MAX_TRAFFIC_BURST completely before starting
the test. Some drivers cannot accept complete
MAX_TRAFFIC_BURST even though rx+tx desc count can fit it.
This patch changes behaviour to stop enqueuing after few
retries.
Fixes: 002ade70e9 ("app/test: measure cycles per packet in Rx/Tx")
Cc: stable@dpdk.org
Signed-off-by: Rakesh Kudurumalla <rkudurumalla@marvell.com>
Enable GPU_REGISTERED flag in gpu/cuda driver in the memory list.
If a GPU memory address CPU mapped is freed before being
unmapped, CUDA driver unmaps it before freeing the memory.
Signed-off-by: Elena Agostini <eagostini@nvidia.com>
FreeBSD has updated its CPU macros to align more with the definitions
used on Linux[1]. Unfortunately, while this makes compatibility better
in future, it means we need to have both legacy and newer definition
support. Use a meson check to determine which set of macros are used.
[1] https://cgit.freebsd.org/src/commit/?id=e2650af157bc
Bugzilla ID: 1014
Fixes: c3568ea376 ("eal: restrict control threads to startup CPU affinity")
Fixes: b6be16acfe ("eal: fix control thread affinity with --lcores")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Tested-by: Daxue Gao <daxuex.gao@intel.com>
Forcing inlining in test_ring_enqueue and test_ring_dequeue can cause
the compiled code to grow extensively when compiled with no optimization
(-O0 or -Og). This is default in the meson's debug configuration. This
can collide with compiler bugs and cause issues during linking of unit
tests where the api_type or esize are non-const variables causing
inlining cascade. In perf tests this is not the case in perf-tests as
esize and api_type are const values.
One such case was discovered when porting DPDK to RISC-V. GCC 11.2 (and
no fix still in 12.1) is generating a short relative jump instruction
(J <offset>) for goto and for loops. When loop body grows extensively in
ring test, the target offset goes beyond supported offfset of +/- 1MB
from PC. This is an obvious bug in the GCC as RISC-V has a
two-instruction construct to jump to any absolute address (AUIPC+JALR).
However there is no reason to force inlining as the test code works
perfectly fine without it.
GCC has a bug report for a similar case (with conditionals):
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93062
Fixes: a9fe152363 ("test/ring: add custom element size functional tests")
Signed-off-by: Stanislaw Kardach <kda@semihalf.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Acked-by: Konstantin Ananyev <konstantin.v.ananyev@yandex.ru>
The lpm_process_event_pkt() can either process a packet using an
architecture specific (defined for X86/SSE, ARM/Neon and PPC64/Altivec)
path or a scalar one. The choice is however done using an ifdef
pre-processor macro. Because of that the scalar version was apparently
not widely exercised/compiled.
Due to some copy/paste errors, the scalar logic in
lpm_process_event_pkt() retained a "continue" statement where it should
utilize rfc1812_process() and return the port/BAD_PORT.
Fixes: 99fc91d180 ("examples/l3fwd: add event lpm main loop")
Signed-off-by: Stanislaw Kardach <kda@semihalf.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
Caught by ASan, if a secondary process tried to attach a device with an
incorrect driver name, devargs was leaked.
Fixes: 64051bb1f1 ("devargs: unify scratch buffer storage")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Calls to rte_memcpy for 1 < n < 16 could result in unaligned
loads/stores, which is undefined behaviour according to the C
standard, and strict aliasing violations.
The code was changed to use a packed structure that allows aliasing
(using the __may_alias__ attribute) to perform the load/store
operations. This results in code that has the same performance as the
original code and that is also C standards-compliant.
Fixes: af75078fec ("first public release")
Cc: stable@dpdk.org
Signed-off-by: Luc Pelletier <lucp.at.work@gmail.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Tested-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Establish unit test for testing thread api. Initial unit tests
for rte_thread_{get,set}_affinity_by_id().
Signed-off-by: Narcisa Vasile <navasile@linux.microsoft.com>
Signed-off-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
Implement functions for getting/setting thread affinity.
Threads can be pinned to specific cores by setting their
affinity attribute.
Windows error codes are translated to errno-style error codes.
The possible return values are chosen so that we have as
much semantic compatibility between platforms as possible.
note: convert_cpuset_to_affinity has a limitation that all cpus of
the set belong to the same processor group.
Signed-off-by: Narcisa Vasile <navasile@linux.microsoft.com>
Signed-off-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
Acked-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Provide a portable type-safe thread identifier.
Provide rte_thread_self for obtaining current thread identifier.
Signed-off-by: Narcisa Vasile <navasile@linux.microsoft.com>
Signed-off-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
Acked-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Acked-by: Konstantin Ananyev <konstantin.v.ananyev@yandex.ru>
Merge CRC32 hash calculation public API implementation for x86 and Arm.
Select the best available CRC32 algorithm when unsupported algorithm
on a given CPU architecture is requested by an application.
Previously, if an application directly includes `rte_crc_arm64.h`
without including `rte_hash_crc.h` it will fail to compile.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Split x86 and SW hash crc intrinsics into separate files.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Acked-by: Yipeng Wang <yipeng1.wang@intel.com>
If an event queue flush does not complete after a fixed number of tries,
remaining queues are flushed before retrying the one with incomplete
flush.
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Added API to set queue attributes at runtime and API to get weight and
affinity.
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Since mailbox is now accessed from multiple threads, use lock to
synchronize access.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Added test cases to test changing of queue QoS attributes priority,
weight and affinity at runtime.
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Extended eventdev queue QoS attributes to support weight and affinity.
If queues are of the same priority, events from the queue with highest
weight will be scheduled first. Affinity indicates the number of times,
the subsequent schedule calls from an event port will use the same event
queue. Schedule call selects another queue if current queue goes empty
or schedule count reaches affinity count.
To avoid ABI break, weight and affinity attributes are not yet added to
queue config structure and rely on PMD for managing it. New eventdev op
queue_attr_get can be used to get it from the PMD.
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Added a new eventdev API rte_event_queue_attr_set(), to set event queue
attributes at runtime from the values set during initialization using
rte_event_queue_setup(). PMD's supporting this feature should expose the
capability RTE_EVENT_DEV_CAP_RUNTIME_QUEUE_ATTR.
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Quiesce event ports used by the workers core on exit to free up
any outstanding resources.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Quiesce event ports used by the workers core on exit to free up
any outstanding resources.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Add function to quiesce any core specific resources consumed by
the event port.
When the application decides to migrate the event port to another lcore
or teardown the current lcore it may to call `rte_event_port_quiesce`
to make sure that all the data associated with the event port are released
from the lcore, this might also include any prefetched events.
While releasing the event port from the lcore, this function calls the
user-provided flush callback once per event.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Event ports are configured to implicitly release the scheduler contexts
currently held in the next call to rte_event_dequeue_burst().
A worker core might still hold a scheduling context during exit as the
next call to rte_event_dequeue_burst() is never made.
This might lead to deadlock based on the worker exit timing and when
there are very less number of flows.
Add a cleanup function to release any scheduling contexts held by the
worker by using RTE_EVENT_OP_RELEASE.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Event ports are configured to implicitly release the scheduler contexts
currently held in the next call to rte_event_dequeue_burst().
A worker core might still hold a scheduling context during exit, as the
next call to rte_event_dequeue_burst() is never made.
This might lead to deadlock based on the worker exit timing and when
there are very less number of flows.
Add clean up function to release any scheduling contexts held by the
worker by using RTE_EVENT_OP_RELEASE.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Event ports are configured to implicitly release the scheduler contexts
currently held in the next call to rte_event_dequeue_burst().
A worker core might still hold a scheduling context during exit, as the
next call to rte_event_dequeue_burst() is never made.
This might lead to deadlock based on the worker exit timing and when
there are very less number of flows.
Add clean up function to release any scheduling contexts held by the
worker by using RTE_EVENT_OP_RELEASE.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Event ports are configured to implicitly release the scheduler contexts
currently held in the next call to rte_event_dequeue_burst().
A worker core might still hold a scheduling context during exit, as the
next call to rte_event_dequeue_burst() is never made.
This might lead to deadlock based on the worker exit timing and when
there are very less number of flows.
Add clean up function to release any scheduling contexts held by the
worker by using RTE_EVENT_OP_RELEASE.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Event ports are configured to implicitly release the scheduler contexts
currently held in the next call to rte_event_dequeue_burst().
A worker core might still hold a scheduling context during exit, as the
next call to rte_event_dequeue_burst() is never made.
This might lead to deadlock based on the worker exit timing and when
there are very less number of flows.
Add clean up function to release any scheduling contexts held by the
worker by using RTE_EVENT_OP_RELEASE.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Remove rte_*_dev calls from signal handler callback as signal handlers
are supposed to be light weight.
Split ethdev teardown into Rx and Tx sections, wait for
workers to finish processing after disabling Rx to allow workers
to complete processing currently held packets.
Verified SW event device on ARM64 using the following command:
./build/app/dpdk-test-eventdev -l 7-23 -s 0xf00 --vdev=event_sw0
-a 0002:02:00.0 -- --prod_type_ethdev --nb_pkts=0 --verbose 2
--test=pipeline_queue --stlist=o --wlcores 16-23
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Move event post-processing to a separate function.
Do complete event post-processing in tear-down functions to prevent
incorrect memory free.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Add additional checks while performing RTE_EVENT_OP_RELEASE to
ensure that there are no pending SWTAGs and FLUSHEs in flight.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Adds telemetry support to get timer adapter info and timer adapter
statistics.
Signed-off-by: Ankur Dwivedi <adwivedi@marvell.com>
Reviewed-by: Jerin Jacob <jerinj@marvell.com>
Updated to allow overriding the default CQ depth of 32. Since there are
only 2048 DLB history list entries, increasing the CQ depth decreases
the number of available LDB ports to 2048/max_cq_depth. Resource query
will take this into account and return the correct maximum number of
LDB ports.
Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
Function roc_sso_hws_head_wait() expects a base as input pointer, and it
will itself get tag_op from the base. By passing tag_op instead of base
pointer to this function will add SSOW_LF_GWS_TAG register offset twice,
which will lead to accessing wrong register.
Fixes: 1f5b3d55c0 ("event/cnxk: store and reuse workslot status")
Cc: stable@dpdk.org
Signed-off-by: Volodymyr Fialko <vfialko@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Caught by covscan:
1. dpdk-21.11/lib/eventdev/rte_event_eth_rx_adapter.c:3279:
logical_vs_bitwise: "~(*__ctype_b_loc()[(int)*params] & 2048 /*
(unsigned short)_ISdigit */)" is always 1/true regardless of the values
of its operand. This occurs as the logical second operand of "||".
2. dpdk-21.11/lib/eventdev/rte_event_eth_rx_adapter.c:3279: remediation:
Did you intend to use "!" rather than "~"?
While isdigit return value should be compared as an int to 0,
prefer ! since all of this file uses this convention.
Fixes: 814d017093 ("eventdev/eth_rx: support telemetry")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Jay Jayatheerthan <jay.jayatheerthan@intel.com>
Latest Sphinx checks C language syntax more aggressively.
Fix the following warning by correcting C language syntax.
doc/guides/prog_guide/event_ethernet_rx_adapter.rst:243:
WARNING: Could not lex literal_block as "c". Highlighting skipped.
Fixes: 3c838062b9 ("eventdev: introduce event vector Rx capability")
Cc: stable@dpdk.org
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Jay Jayatheerthan <jay.jayatheerthan@intel.com>
In packets with ESP header, the inner IP will be encrypted, and
its fields cannot be used for RSS hashing. So, ESP packets
can be hashed only by the outer IP layer.
So, when using RSS on ESP packets, hashing may not be efficient,
because the fields used by the hash functions are only the outer IPs,
causing all traffic belonging to all tunnels between a given
pair of GWs to land on one core.
Adding the SPI hash field can extend the spreading of IPsec packets.
Signed-off-by: Raja Zidane <rzidane@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
When a meter with RSS action being used, there might be several
sub-flows using different sub-policies in the flow splitting stage.
If there's no green action, there's an error that will always use the
same sub-policy for all sub-flows, some resources will be
overwritten and cannot be released, leading assert during port close.
This patch fixes this issue by checking both green and yellow queue
index during getting a blank sub-policy, to avoid the incorrect
resource overwrite.
Fixes: b38a12272b ("net/mlx5: split meter color policy handling")
Cc: stable@dpdk.org
Signed-off-by: Shun Hao <shunh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Invoking bnxt_link_update_op() with wait_for_completion set would
result in the driver waiting for 10s in case the port link is down to
complete port initialization (dev_start_op()).
Change it by not waiting for the completion when invoking it in
dev_start_op()
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
In bnxt_free_all_filters(), all the filters attached to a vnic are removed.
But each of these filters hold a backreference ptr to the vnic and they
need to be reset to NULL now. Otherwise, during a normal testpmd quit, as
part of dev_close_op(), first bnxt_free_all_filters() is invoked in
dev_stop, followed by bnxt_free_filter_mem() from bnxt_uninit_resources(),
which finds a filter with a vnic back reference ptr and now
bnxt_hwrm_clean_up_l2_filter() also tries to remove the filter from the
vnic's filter list which was already done as part of
bnxt_free_all_filters().
Fixes: f0f6b5e6cf ("net/bnxt: fix reusing L2 filter")
Cc: stable@dpdk.org
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
If Firmware is still in reset process and returns the error
HWRM_ERR_CODE_HOT_RESET_PROGRESS, retry VER_GET command.
We have to do it in bnxt_handle_if_change_status().
Fixes: 0b53359123 ("net/bnxt: inform firmware about IF state changes")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Driver forces link down during port stop. But device is not obliged
link down in certain scenarios, even when forced. In that case,
subsequent link queries returns link as up.
Fixed to return link status as down when port is stopped.
Driver is already doing that for VF/NPAR/MH functions.
Fixes: c09f57b49c ("net/bnxt: add start/stop/link update operations")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Device is not obliged link down in certain scenarios, even
when forced. When FW does not allow any user other than the BMC
to shutdown the port, bnxt_get_hwrm_link_config() call always
returns link up. Force phy update always in that case,
else user configuration for speed/autoneg would not get applied
correctly.
Fixes: 7bc8e9a227 ("net/bnxt: support async link notification")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
The "active_fec_signal_mode" in HWRM_PORT_PHY_QCFG response
does not return correct value till the link is up. Driver cannot
rely on active_fec_signal_mode while setting autoneg speed.
While setting autoneg speed, driver is currently checking only
"auto_link_speed_mask". Fixed to check "auto_pam4_link_speed_mask"
as well. Also, while setting auto mode and setting speed mask,
driver will have to set both NRZ and PAM4 mask.
Fixes: c23f9ded03 ("net/bnxt: support 200G PAM4 link")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
The "active_fec_signal_mode" in HWRM_PORT_PHY_QCFG response is uint8_t.
So no need of endianness conversion while parsing response.
Also, signal_mode is the first 4bits of "active_fec_signal_mode".
Fixes: c23f9ded03 ("net/bnxt: support 200G PAM4 link")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
