NFP can support several physical ports per PF device. Depending on
firmware info, one or more eth_dev objects will need to be created.
This patch adds the call to create just one eth_dev by now with future
commits supporting the multiport option. Once the eth_dev has been
created, probe function invokes pmd initialization with the new eth_dev.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
nfp_net_init is where a dpdk port related to a eth_dev is initialized.
NFP VF vNICs use VF PCI BARs as they come after SRIOV is enabled. But for
NFP PF vNIC just a subset of PF PCI BARs are used.
This patch adds support for mapping the right PCI BAR subsets for the PF
vNIC. It uses the NSPU API functions introduced previously for configuring
NFP expansion bars.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
NFP vNICs use a subset of PCI device BARs. vNIC rx/tx bars point to
NFP hardware queues unit. Unlike vNIC config bar, the NFP address is
always the same so the NFP expansion bar configuration always uses
the same hardcoded physical address.
This patch adds a NSPU API function for getting vNIC rx/tx bars
mapped through a expansion bar using that specific physical address.
The PMD will use the PCI bar offset returned for mapping the vNIC
rx/tx bars.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
NFP vNICs use a subset of PCI device BARs. vNIC config bar depends on
firmware symbol defining how to map it through a NFP expansion bar.
This patch adds a NSPU API function for getting a vNIC config bar
mapped through a expansion bar giving a firmware symbol. The PMD will
use the PCI bar offset returned for accessing the vNIC bar.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
PMD will use this function for uploading the firmware. First, a
symbol resolution is done for finding out if there is a firmware
already there. If not, a NFP reset is called before using NSPU
fw upload code.
PMD PF probe function is now using this logic.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Firmware has symbols helping to configure things like number of
PF ports, vNIC BARs addresses inside NFP memories, or ethernet
link state. Different firmware apps have different things to map
and likely different internal NFP addresses to use.
Host drivers can use the NSPU interface for getting symbol data
regarding different hardware configurations. Once the driver has
the information about a specific object, a mapping is required
configuring an NFP expansion bar creating a device PCI bar window.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Using NSPU interface for fw upload. Firmware file needs to be
installed in specific path inside system firmware directory.
NSPU buffer is used for writing the firmware before sending the
command.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
NSPU interface declares a buffer controlled by the NFP NSP service
processor. It is possible to send commands to the NSP using the NSPU
and this buffer for data related to the command. A command can imply
buffer read, buffer write, both or none.
Initial command for resetting the firmware is added as well which
does not require the buffer at all.
Commands will allow firmware upload, symbol resolution and ethernet
link configuration. Future commands will allow specific offloads like
flow offloads and eBPF offload.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Configuring the NFP PMD for using the PF requires access through the
NSPU interface for device configuration. This patch adds a specific probe
function for the PF which uses the NSPU interface. Just basic NSPU access
is done by now reading the NSPU ABI version.
NSPU ABI version needs to be greater than 0.19.
No ethernet port is created yet.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Working with the PF requires access to the NFP for basic configuration.
NSP is the NFP Service Processor helping with hardware and firmware
configuration. NSPU is the NSP user space interface for working with the
NSP.
Configuration through NSPU allows to create PCI BAR windows for accessing
different NFP hardware units, including the BAR window for the NSPU
interface access itself. NFP expansion bar registers are used for creating
those PCI BAR windows. NSPU uses a specific expansion bar which is
reprogrammed for accessing/doing different things.
Other expansion bars will be configured later for configuring the PF vNIC
bars, a subset of PF PCI BARs.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
We need to support how firmware metadata was handled until now and also
the new API, since NFP NFD 3.0 firmware versions. The new metadata API
adds flexibility for working with different metadata types and, mainly,
to allow adding metadata from different firmware components independently.
Although this patch just supports one type handled by the PMD, future uses
regarding firmware apps will extend this support.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
A DPDK app could, whatever the reason, send packets with size 0.
The PMD is not sending those packets, which does make sense,
but the problem is the mbuf is not released either. That leads
to mbufs not being available, because the app trusts the
PMD will do it.
Although this is a problem related to app wrong behavior, we
should harden the PMD in this regard. Not sending a packet with
size 0 could be problematic, needing special handling inside the
PMD xmit function. It could be a burst of those packets, which can
be easily handled, but it could also be a single packet in a burst,
what is harder to handle.
It would be simpler to just send that kind of packets, which will
likely be dropped by the hw at some point. The main problem is how
the fw/hw handles the DMA, because a dma read to a hypothetical 0x0
address could trigger an IOMMU error. It turns out, it is safe to
send a descriptor with packet size 0 to the hardware: the DMA never
happens, from the PCIe point of view.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
intr_vec was not properly configured. This is not a problem when
just one queue is supported but it fails with multiqueue.
Some minor refactoring also done for hardware interrupt configuration.
Fixes: ea121b28316d ("net/nfp: add Rx interrupts")
Cc: stable@dpdk.org
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Redirection table was not being updated properly.
There is also a problem when configuring RSS.
Fixes: 934e4c60fbff ("nfp: add RSS")
Cc: stable@dpdk.org
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
If not a valid mac present in configuration bar, PMD creates a random
one. It needs to be passed to the NIC.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
At some places, the log2() function is used despite this function
works on float. This introduces a dependency to the math lib but
most of the time it is not required because we want an integer log2.
Add a new helper to do this job and fix nfp driver.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Change the rte_eth_dev_callback_process function to return int,
and add a void *ret_param parameter.
The new parameter is used by ixgbe and i40e instead of abusing
the user data of the callback.
Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Fixing typos across dpdk source code using codespell utility.
Skipped the ethdev driver's base code fixes to keep the base
code intact.
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
Instead of many PMD define their own macro, define a generic one in
ethdev and use that in PMDs.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Allain Legacy <allain.legacy@windriver.com>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
vfio is the kernel framework used by the vfio-pci kernel driver.
DPDK drivers do not rely solely on vfio, but rather on vfio-pci to gain
access to pci resources.
Fixes: 0880c40113ef ("drivers: advertise kmod dependencies in pmdinfo")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
If segments are used, just mbufs previously linked to head descriptor
of a mbuf chain are released. Other Tx descriptor are used for the
mbuf chain but they keep their linked mbufs without releasing them.
It is not a fatal issue because sooner or later those descriptors will
be head descriptors or just used for a single mbuf packet, then those
linked mbufs will be released.
However, this leads to apps needing bigger mbufs pools and some
confusion about memory requirements. Indeed, because larger pools, some
performance impact could also be expected due to cache misses.
With this patch all Tx descriptors will release linked mbufs inside the
xmit function, and rte_pktmbuf_seg_free is used instead of
rte_pktmbuf_free.
Fixes: 142854c62134 ("nfp: fix freeing multi-mbuf packets")
Cc: stable@dpdk.org
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
This driver can use the library function rte_eth_dma_zone_reserve()
instead of duplicating the code.
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
The patch change the prototype of callback function
(rte_intr_callback_fn) by removing the unnecessary parameter.
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Current code enables RX interrupts even if this it not
requested.
Fixes: ea121b28316d ("net/nfp: add Rx interrupts")
Cc: stable@dpdk.org
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Chained mbufs hold data_len as the length of that particular mbuf
and pkt_len as the full packet length including all the chained
mbufs. It is not clear from the mbuf definition if pkt_len should
be set for all the mbufs in a chain, but code there for handling
mbufs suggests just the first mbuf requires to have pkt_len set.
NFP PMD was assuming pkt_len is set in all the chained mbufs and
unit tests for gather dma were building mbufs with pkt_len always
set. This patch gets rid of that assumption.
Fixes: b812daadad0d ("nfp: add Rx and Tx")
Cc: stable@dpdk.org
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
When LSO, not doing this can led to firmware disruption. It does
not show as error because TCP ends up sending data again later on.
Fixes: 9ba3d0ae2090 ("net/nfp: add TSO support")
Cc: stable@dpdk.org
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Before this patch, the management of dependencies between directories
had several issues:
- the generation of .depdirs, done at configuration is slow: it can take
more than one minute on some slow targets (usually ~10s on a standard
PC without -j).
- for instance, it is possible to express a dependency like:
- app/foo depends on lib/librte_foo
- and lib/librte_foo depends on app/bar
But this won't work because the directories are traversed with a
depth-first algorithm, so we have to choose between doing 'app' before
or after 'lib'.
- the script depdirs-rule.sh is too complex.
- we cannot use "make -d" for debug, because the output of make is used for
the generation of .depdirs.
This patch moves the DEPDIRS-* variables in the upper Makefile, making
the dependencies much easier to calculate. A DEPDIRS variable is still
used to process library dependencies in LDLIBS.
After this commit, "make config" is almost immediate.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Tested-by: Robin Jarry <robin.jarry@6wind.com>
Tested-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
The check of queue_id is done in all drivers implementing
rte_eth_rx_queue_count(). Factorize this check in the generic function.
Note that the nfp driver was doing the check differently, which could
induce crashes if the queue index was too big.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
dev_flags is wrongly overwritten with RTE_ETH_DEV_DETACHABLE value
in drivers after rte_eth_copy_pci_info().
Previous values of the dev_flags set in rte_eth_copy_pci_info(),
like RTE_ETH_DEV_INTR_LSC, are get lost. That will fail the device
configuration.
Fix by preventing dev_flags overwritten.
Fixes: 22dda618c00c ("pci: separate detaching ethernet ports from PCI devices")
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Signed-off-by: Jeff Guo <jia.guo@intel.com>
Replace the raw I/O device memory read/write access with eal
abstraction for I/O device memory read/write access to fix
portability issues across different architectures.
CC: Alejandro Lucero <alejandro.lucero@netronome.com>
Signed-off-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Fix typo when checking that no VLAN offload flags are passed at port
initialization.
By the way, also fix a typo in the log.
Fixes: d4a27a3b092a ("nfp: add basic features")
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Alejandro Lucero <alejandro.lucero@netronome.com>
This patch implements NFP PMD support for TSO but it also requires
a firmware advertising the capability.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Having those modulo operations implies costly instructions execution,
what can be avoided with conditionals and unlikely clauses.
This change makes the software ring read and write indexes to be now
always within the ring size which has to be handled properly. The main
problem is when write pointer wraps and being less than the read pointer.
This happened before, but just with indexes type size (uint32_t) wrapping,
and in that case the processor does the right thing no requiring special
handling by software.
This work has also led to discovering redundant pointers in the driver,
which have been removed.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
NFP supports more speeds than just 40 and 100GB, which were
what was advertised before.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Previous reported speed was hardcoded because there was not firmware
support for getting this information. This change needs also to support
old firmware versions, but instead of the previous hardcoded report, no
speed is reported to the user avoiding to give the wrong speed when link
is not configured to 40G.
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Because macros for TCP and UDP related to offload cksums have
same values, this was not a main problem. But better to use the
right ones.
Fixes: d4a27a3b092a ("nfp: add basic features")
Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Since commit b1fb53a39d88 ("ethdev: remove some PCI specific handling"),
rte_eth_dev_info_get() relies on dev->data->drv_name to report the driver
name to caller.
Having the pmds set driver_info->driver_name in the pmds is useless,
since ethdev overwrites it right after.
The only thing the pmd must do is:
- for pci drivers, call rte_eth_copy_pci_info() which then sets
data->drv_name
- for vdev drivers, manually set data->drv_name
At this stage, virtio-user does not properly report a driver name (fixed in
next commit).
Signed-off-by: David Marchand <david.marchand@6wind.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Reviewed-by: Jan Blunck <jblunck@infradead.org>
Attaching and detaching ethernet ports from an application
is not the same thing as physically removing a PCI device,
so clarify the flags indicating support. All PCI devices
are assumed to be physically removable, so no flag is
necessary in the PCI layer.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This makes struct rte_eth_dev independent of struct rte_pci_device by
replacing it with a pointer to the generic struct rte_device.
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Only the drivers itself can decide if it could fill PCI information fields
of dev_info.
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Add a new macro RTE_PMD_REGISTER_KMOD_DEP() that allows a driver to
declare the list of kernel modules required to run properly.
Today, most PCI drivers require uio/vfio.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Today, all logs whose level is lower than INFO are dropped at
compile-time. This prevents from enabling debug logs at runtime using
--log-level=8.
The rationale was to remove debug logs from the data path at
compile-time, avoiding a test at run-time.
This patch changes the behavior of RTE_LOG() to avoid the compile-time
optimization, and introduces the RTE_LOG_DP() macro that has the same
behavior than the previous RTE_LOG(), for the rare cases where debug
logs are in the data path.
So it is now possible to enable debug logs at run-time by just
specifying --log-level=8. Some drivers still have special compile-time
options to enable more debug log. Maintainers may consider to
remove/reduce them.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
add cb_arg parameter to the _rte_eth_dev_callback_process function.
Adding a parameter to this function allows passing information
to the application when an eth device event occurs such as
a VF to PF message.
This allows the application to decide if a particular function
is permitted.
Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Signed-off-by: Alex Zelezniak <alexz@att.com>
