Mempool is a generic allocator that is not necessarily used
for device IO operations and its memory for DMA.
Add MEMPOOL_F_NON_IO flag to mark such mempools automatically
a) if their objects are not contiguous;
b) if IOVA is not available for any object.
Other components can inspect this flag
in order to optimize their memory management.
Discussion: https://mails.dpdk.org/archives/dev/2021-August/216654.html
Signed-off-by: Dmitry Kozlyuk <dkozlyuk@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Data path performance can benefit if the PMD knows which memory it will
need to handle in advance, before the first mbuf is sent to the PMD.
It is impractical, however, to consider all allocated memory for this
purpose. Most often mbuf memory comes from mempools that can come and
go. PMD can enumerate existing mempools on device start, but it also
needs to track creation and destruction of mempools after the forwarding
starts but before an mbuf from the new mempool is sent to the device.
Add an API to register callback for mempool life cycle events:
* rte_mempool_event_callback_register()
* rte_mempool_event_callback_unregister()
Currently tracked events are:
* RTE_MEMPOOL_EVENT_READY (after populating a mempool)
* RTE_MEMPOOL_EVENT_DESTROY (before freeing a mempool)
Provide a unit test for the new API.
The new API is internal, because it is primarily demanded by PMDs that
may need to deal with any mempools and do not control their creation,
while an application, on the other hand, knows which mempools it creates
and doesn't care about internal mempools PMDs might create.
Signed-off-by: Dmitry Kozlyuk <dkozlyuk@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Fix spelling error which is causing reports of other patches failing.
Fixes: 69daa9e502 ("net/cnxk: support inline security setup for cn10k")
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Removing the rawdev based octeontx2-ep driver as the dependent
common/octeontx2 will soon be going away. Moreover this driver is no
longer required as the net/octeontx_ep driver is sufficient.
Signed-off-by: Radha Mohan Chintakuntla <radhac@marvell.com>
Removing the rawdev based octeontx2-dma driver as the dependent
common/octeontx2 will be soon be going away. Also a new DMA driver will
be coming in this place once the rte_dmadev library is in.
Signed-off-by: Radha Mohan Chintakuntla <radhac@marvell.com>
Add a test case to validate the functionality of drivers' burst capacity
API implementations.
Signed-off-by: Kevin Laatz <kevin.laatz@intel.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
For DMA devices which support the fill operation, run unit tests to
verify fill behaviour is correct.
Signed-off-by: Kevin Laatz <kevin.laatz@intel.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Add a series of tests to inject bad copy operations into a dmadev to
test the error handling and reporting capabilities. Various combinations
of errors in various positions in a burst are tested, as are errors in
bursts with fence flag set, and multiple errors in a single burst.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Due to HW or driver limitations, not all dmadevs may support full error
handling e.g. safely managing and reporting an invalid address to a copy
operation. The skeleton dmadev, for example, being pure software will
always seg-fault if passed an invalid address. To indicate the
availability of safe error handling by a device, we add a capability
flag for it.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Add unit tests for various combinations of use for dmadev, copying
bursts of packets in various formats, e.g.
1. enqueuing two smaller bursts and completing them as one burst
2. enqueuing one burst and gathering completions in smaller bursts
3. using completed_status() function to gather completions rather than
just completed()
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
When running the dmadev_autotest, run the suite of copy tests on the
skeleton driver created for API testing too, rather than just destroying
the driver instances once the API tests are complete. This helps to
sanity check the tests themselves are reasonable.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
For each dmadev instance, perform some basic copy tests to validate that
functionality.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Run basic sanity tests for configuring, starting and stopping a dmadev
instance to help validate drivers. This also provides the framework for
future tests for data-path operation.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Add a function and wrapper macro to iterate over all DMA devices.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Add a burst capacity check API to the dmadev library. This API is useful to
applications which need to how many descriptors can be enqueued in the
current batch. For example, it could be used to determine whether all
segments of a multi-segment packet can be enqueued in the same batch or not
(to avoid half-offload of the packet).
Signed-off-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Add in a function to check if a device or vchan has completed all jobs
assigned to it, without gathering in the results. This is primarily for
use in testing, to allow the hardware to be in a known-state prior to
gathering completions.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
This patch add dmadev API test which based on 'dma_skeleton' vdev. The
test cases could be executed using 'dmadev_autotest' command in test
framework.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Skeleton dmadevice driver, on the lines of rawdev skeleton, is for
showcasing of the dmadev library.
Design of skeleton involves a virtual device which is plugged into VDEV
bus on initialization.
Also, enable compilation of dmadev skeleton drivers.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
This patch add data plane API for dmadev.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
This patch add control plane API for dmadev.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
The 'dmadev' is a generic type of DMA device.
This patch introduce the 'dmadev' device allocation functions.
The infrastructure is prepared to welcome drivers in drivers/dma/
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
As stated in the API, dynamic field and flags should be created with no
additional flag (simply in the API for future changes).
Fix the dynamic flag register helper which was not enforcing it and add
unit tests.
Fixes: 4958ca3a44 ("mbuf: support dynamic fields and flags")
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Ray Kinsella <mdr@ashroe.eu>
If we do not enforce valid flags are passed by an application, this
application might face issues in the future when we add more flags.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Ray Kinsella <mdr@ashroe.eu>
If we do not enforce valid flags are passed by an application, this
application might face issues in the future when we add more flags.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Ray Kinsella <mdr@ashroe.eu>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
When a user runs the dpdk-telemetry script and fails to connect because
the socket path does not exist, run a scan for possible sockets that
could be connected to and inform the user of the command needed to
connect to those.
For example:
$ ./dpdk-telemetry.py -i4
Connecting to /run/user/1000/dpdk/rte/dpdk_telemetry.v2:4
Error connecting to /run/user/1000/dpdk/rte/dpdk_telemetry.v2:4
Other DPDK telemetry sockets found:
- dpdk_telemetry.v2 # Connect with './dpdk-telemetry.py'
- dpdk_telemetry.v2:2 # Connect with './dpdk-telemetry.py -i 2'
- dpdk_telemetry.v2:1 # Connect with './dpdk-telemetry.py -i 1'
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Ciara Power <ciara.power@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
For processes run using "in-memory" mode sharing the same runtime dir,
we add support for connecting to the separate instance sockets created
using ":1", ":2" etc. via new "-i" or "--instance" argument. Add details
on connecting to separate instances to the telemetry howto document.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Ciara Power <ciara.power@intel.com>
Tested-by: Conor Walsh <conor.walsh@intel.com>
When running using in-memory mode, multiple processes can use the same
runtime dir, leading to conflicts with the telemetry sockets in that
directory. We can resolve this by appending a suffix to each socket
beyond the first, with the suffix being an increasing counter value.
Each process uses the first unused socket counter value.
Fixes: 6dd571fd07 ("telemetry: introduce new functionality")
Reported-by: David Marchand <david.marchand@redhat.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Ciara Power <ciara.power@intel.com>
Acked-by: Kevin Traynor <ktraynor@redhat.com>
Tested-by: Conor Walsh <conor.walsh@intel.com>
Telemetry interface should be exposed for primary processes only, since
secondary processes will conflict on socket creation, and since all
data in secondary process is generally available to primary. For
example, all device stats for ethdevs, cryptodevs, etc. will all be
common across processes.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Ciara Power <ciara.power@intel.com>
Tested-by: Conor Walsh <conor.walsh@intel.com>
For native build, enabling building the highest cpu_instruction_set
supported by the build host, including the new POWER10.
For cross compile, verifying that the compiler supports the
cpu_instruction_set specified in the cross-file
Signed-off-by: Thinh Tran <thinhtr@linux.vnet.ibm.com>
Reviewed-by: David Christensen <drc@linux.vnet.ibm.com>
When checking the results of the rte_atomic128_cmp_exchange() function,
current code compares the values of a uint32_t and a uint64_t variable.
If the number of lcores used by the test is large, or the value of the
iteration count N is increased, the variable size mismatch can cause a
false test failure. Modify the comparison to compare uint64_t values.
Fixes: fa3253c534 ("test/atomic: add 128-bit atomic compare exchange test")
Cc: stable@dpdk.org
Signed-off-by: David Christensen <drc@linux.vnet.ibm.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Current implementation of rte_ipv4_fragment_packet() doesn’t take
into account offset and flag values of the given packet, but blindly
assumes they are always zero (original packet is not fragmented).
According to RFC791, fragment and flag values for new fragment
should take into account values provided in the original IPv4 packet.
Fixes: 4c38e5532a ("ip_frag: refactor IPv4 fragmentation into a proper library")
Cc: stable@dpdk.org
Signed-off-by: Huichao Cai <chcchc88@163.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Removed offload flag PKT_RX_EIP_CKSUM_BAD. PKT_RX_OUTER_IP_CKSUM_BAD
should be used as a replacement.
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Use correct define as a name array size.
The change breaks ABI and therefore cannot be backported to
stable branches.
Fixes: 38c9817ee1 ("mempool: adjust name size in related data types")
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: David Marchand <david.marchand@redhat.com>
Earlier, JSON message length was limited to 1024 which would not
allow data more than this size. Removed this limitation by creating
output buffer based on requested data length.
Fixes: 52af6ccb2b ("telemetry: add utility functions for creating JSON")
Cc: stable@dpdk.org
Signed-off-by: Gowrishankar Muthukrishnan <gmuthukrishn@marvell.com>
Acked-by: Ciara Power <ciara.power@intel.com>
The in-memory option is not supported on FreeBSD so print a warning and
ignore the flag when it is specified for BSD apps. The lack of support
is due to the different way in which memory is managed on FreeBSD using
the contigmem driver rather than via a hugetlbfs filesystem.
Fixes: 14de8734c4 ("eal: add --in-memory option")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
The driver code had wrong assumption that all the addresses to ring buffers
in the secondary process are the same as those in the primary process. This
is not always correct as the channels could be mapped to different
addresses in the secondary process.
Fix this by keeping track of all the mapped addresses from the primary
process in the shared uio_res, and have second process map to the same
addresses.
Fixes: 831dba47bd ("bus/vmbus: add Hyper-V virtual bus support")
Cc: stable@dpdk.org
Reported-by: Jonathan Erb <jonathan.erb@banduracyber.com>
Signed-off-by: Long Li <longli@microsoft.com>
Acked-by: Stephen Hemminger <sthemmin@microsoft.com>
This function is public since commit 8f0e4d6a78 ("net: export IPv6
header extensions skip function") (2018), and is used by vmxnet3 driver.
Promote it as stable.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: David Marchand <david.marchand@redhat.com>
Acked-by: Ray Kinsella <mdr@ashroe.eu>
The guides should be referenced locally with RST syntax
:doc: (beginning of page) or :ref: (specific chapter).
The links to doc.dpdk.org/guides/ are removed.
The links to the doc.dpdk.org/api/ are acceptable,
but should not point to a specific version, so one is fixed.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: David Marchand <david.marchand@redhat.com>
Sort the definitions for extended features (leaf 0) to enhance
readability.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Caught while checking CPUID related stuff in OVS.
According to [1], for Structured Extended Feature Flags Enumeration Leaf
(EAX = 0x07H, ECX = 0):
- BMI1 is associated to EBX, bit 3 (was incorrectly 2),
- SMEP is associated to EBX, bit 7 (was incorrectly 6),
- BMI2 is associated to EBX, bit 8 (was incorrectly 7),
- ERMS is associated to EBX, bit 9 (was incorrectly 8),
1: https://software.intel.com/sites/default/files/managed/c5/15/architecture-instruction-set-extensions-programming-reference.pdf
Fixes: af75078fec ("first public release")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
get_hugepage_dir() was implemented in such a way that a --huge-dir
option had to exactly match the mountpoint, but there's no reason for
this restriction: DPDK might not be the only user of hugepages, and
shouldn't assume it owns an entire mountpoint. For example, if I have
/dev/hugepages/myapp, and /dev/hugepages/dpdk, I should be able to
specify:
--huge-dir=/dev/hugepages/dpdk/
and have DPDK only use that sub-directory.
Fix the implementation to allow a sub-directory within a suitable
hugetlbfs mountpoint to be specified, preferring the closest match.
Signed-off-by: John Levon <john.levon@nutanix.com>
Acked-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
This tool was used in the past to execute unit tests with make.
Meson has its own list of unit test, and its own way of calling them.
Since the switch to meson only builds, there is nothing depending on the
script in DPDK itself.
No CI seems to call it, time to put it to rest.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Aaron Conole <aconole@redhat.com>
The CI reported rare (and cryptic) failures like:
RTE>>service_autotest
+ ------------------------------------------------------- +
+ Test Suite : service core test suite
+ ------------------------------------------------------- +
+ TestCase [ 0] : unregister_all succeeded
+ TestCase [ 1] : service_name succeeded
+ TestCase [ 2] : service_get_by_name succeeded
Service dummy_service Summary
dummy_service: stats 1 calls 0 cycles 0 avg: 0
Service dummy_service Summary
dummy_service: stats 0 calls 0 cycles 0 avg: 0
+ TestCase [ 3] : service_dump succeeded
+ TestCase [ 4] : service_attr_get failed
+ TestCase [ 5] : service_lcore_attr_get succeeded
+ TestCase [ 6] : service_probe_capability succeeded
+ TestCase [ 7] : service_start_stop succeeded
+ TestCase [ 8] : service_lcore_add_del succeeded
+ TestCase [ 9] : service_lcore_start_stop succeeded
+ TestCase [10] : service_lcore_en_dis_able succeeded
+ TestCase [11] : service_mt_unsafe_poll succeeded
+ TestCase [12] : service_mt_safe_poll succeeded
perf test for MT Safe: 42.7 cycles per call
+ TestCase [13] : service_app_lcore_mt_safe succeeded
perf test for MT Unsafe: 73.3 cycles per call
+ TestCase [14] : service_app_lcore_mt_unsafe succeeded
+ TestCase [15] : service_may_be_active succeeded
+ TestCase [16] : service_active_two_cores succeeded
+ ------------------------------------------------------- +
+ Test Suite Summary : service core test suite
+ ------------------------------------------------------- +
+ Tests Total : 17
+ Tests Skipped : 0
+ Tests Executed : 17
+ Tests Unsupported: 0
+ Tests Passed : 16
+ Tests Failed : 1
+ ------------------------------------------------------- +
Test Failed
RTE>>
stderr:
EAL: Detected CPU lcores: 16
EAL: Detected NUMA nodes: 2
EAL: Detected static linkage of DPDK
EAL: Multi-process socket /var/run/dpdk/service_autotest/mp_socket
EAL: Selected IOVA mode 'PA'
EAL: No available 1048576 kB hugepages reported
EAL: VFIO support initialized
EAL: Device 0000:03:00.0 is not NUMA-aware, defaulting socket to 0
APP: HPET is not enabled, using TSC as default timer
EAL: Test assert service_attr_get line 340 failed: attr_get() call didn't
get call count (zero)
According to API, trying to stop a service lcore is not possible if this
lcore is the only one associated to a service.
Doing this will result in a -EBUSY return code from
rte_service_lcore_stop() which the service_attr_get subtest was not
checking.
This left the service lcore running, and a race existed with the main
lcore on checking the service attributes which triggered this CI
failure.
To fix this, dissociate the service lcore with current service.
Once fixed this first issue, a race still exists, because the
wait_slcore_inactive helper added in a previous fix was not
paired with a check that the service lcore _did_ stop.
Add missing check on rte_service_lcore_may_be_active.
Fixes: 4d55194d76 ("service: add attribute get function")
Fixes: 52bb6be259 ("test/service: fix race condition on stopping lcore")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Aaron Conole <aconole@redhat.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
Change the inaccurate comment of 'set pass flag' to 'clear pass flag'
as the '*pass_test = 0' code actually implements clearing.
Fixes: f038a81e1c ("service: add unit tests")
Cc: stable@dpdk.org
Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
This patch adds tests for inner IP and inner L4 checksum
verification in IPsec mode.
Signed-off-by: Archana Muniganti <marchana@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
