We find that when traffic is light, a small number of packets will have
wrong metadata (e.g. packet type), however this issue will not happen
when traffic is heavy.
The root cause is some fields in fm10k_rx_desc are read at the wrong time,
since the descriptor (being 16-bytes big) is not read as a single atomic
operation. When the input speed is slower than software's capability,
fm10k scalar Rx function accesses descriptors at about the same time
as HW writes them, so the scenario can occur: some fields like pkt_info
in fm10k_rx_desc are read before HW writeback but some fields like DD bit
are read after HW writeback, this will lead to the later packet parsing
function using incorrect value.
This patch fixes this issue by reading and parsing Rx descriptor only after
first checking that the DD bit is set.
Fixes: 4b61d3bfa941 ("fm10k: add receive and tranmit")
Fixes: c82dd0a7bfa5 ("fm10k: add scatter receive")
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Modify the PMD_REGISTER_DRIVER macro, adding a name argument to it. The
addition of a name argument creates a token that can be used for subsequent
macros in the creation of unique symbol names to export additional bits of
information for use by the pmdinfogen tool. For example:
PMD_REGISTER_DRIVER(ena_driver, ena);
registers the ena_driver struct as it always did, and creates a symbol
const char this_pmd_name0[] __attribute__((used)) = "ena";
which pmdinfogen can search for and extract. The subsequent macro
DRIVER_REGISTER_PCI_TABLE(ena, ena_pci_id_map);
creates a symbol const char ena_pci_tbl_export[] __attribute__((used)) =
"ena_pci_id_map";
Which allows pmdinfogen to find the pci table of this driver
Using this pattern, we can export arbitrary bits of information.
pmdinfo uses this information to extract hardware support from an object
file and create a json string to make hardware support info discoverable
later.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Remy Horton <remy.horton@intel.com>
For all drivers that currently implement xstats, the id field in the
rte_eth_stats_name structure equals the entry's array index. This
patch eliminates the redundant id field as a direct index lookup is
faster than a search for the matching id field.
Suggested-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: Remy Horton <remy.horton@intel.com>
When app tries to change promisc/allmulti setting, fm10k will check if a
valid glort is acquired, and exit without doing anything if not.
For VFs, this glort value is not necessary, and so the check can be
removed. This saves having unnecessary failures of the API call, as well as
saving the time taken for the mailbox communication between VF and PF in
the case when the glort check passes.
Fixes: df02ba864695 ("fm10k: support promiscuous mode")
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Following the discussions from:
http://dpdk.org/ml/archives/dev/2015-July/021721.htmlhttp://dpdk.org/ml/archives/dev/2016-April/038143.html
The value of these flags is 0, making them useless. Today, no example
application checks them on Rx, and only few drivers sets them and
silently give wrong packets to the application, which should not happen.
This patch removes the unused flags from rte_mbuf and their use in the
drivers. The i40e and fm10k are kept as they are today and should be
fixed to drop bad packets. The enic driver is managed by its maintainer
in another patch.
Fixes: c22265f6 ("mbuf: add new packet flags for i40e")
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the test-pmd
and proc_info applications to use the new xstats API, and removes
deprecated code associated with the old API.
Signed-off-by: Remy Horton <remy.horton@intel.com>
The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the fm10k driver
to use the new API that seperates name string and value queries.
Signed-off-by: Remy Horton <remy.horton@intel.com>
This patch adds missing DEPDIRS to avoid any library referring to
symbols they are not linked against.
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Removed comparison against $CC in Makefiles as
in cross-compiling mode CC can be a different string
instead of string "gcc"
Suggested-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Some architectures (ex: Power8) have a cache line size of 128 bytes,
so the drivers should not expect that prefetching the second part of
the mbuf with rte_prefetch0(&m->cacheline1) is valid.
This commit add helpers that can be used by drivers to prefetch the
rx or tx part of the mbuf, whatever the cache line size.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Reviewed-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
When building a chain of mbufs for a multi-segment packet, the
packet_type field resides at the end of the chain. It should be
copied forward to the head of the list.
Also, uses RTE_LIBRTE_FM10K_RX_OLFLAGS_ENABLE to guard packet-type
computation. The mbuf fields are not copied when this define is not set.
Fixes: fe65e1e1ce61 ("fm10k: add vector scatter Rx")
Signed-off-by: Michael Frasca <michael.frasca@oracle.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
The link speed configuration is now done with bitmaps so 100G speed
requires only a new bit flag.
The actual link speed is a number so its size must be increased from
16-bit to 32-bit.
Signed-off-by: Marc Sune <marcdevel@gmail.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Tested-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Tested-by: Matej Vido <vido@cesnet.cz>
The speed capabilities of a device can be retrieved with
rte_eth_dev_info_get().
The new field speed_capa is initialized in the drivers without
taking care of device characteristics in this patch.
When the capabilities of a driver are accurate, the table in
overview.rst must be filled.
Signed-off-by: Marc Sune <marcdevel@gmail.com>
Define and use ETH_LINK_UP and ETH_LINK_DOWN where appropriate.
Signed-off-by: Marc Sune <marcdevel@gmail.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
If the provided configuration does not call for RSS, then RSS is
explicitly disabled. Without this change, the device continues to
operate under the previous RSS configuration.
Fixes: 57033cdf8fdc ("fm10k: add PF RSS")
Signed-off-by: Michael Frasca <michael.frasca@oracle.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Add a new API rte_eth_dev_get_supported_ptypes to query what packet types
can be filled by a given device. The device should be already started or
its PMD RX burst function already decided, since the packet types supported
may vary depending on RX function.
Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
This patch enables reading sglort (global resource tag) info into the
mbuf for RX and inserting an FTAG (Fabric Tag) at the beginning of the
packet for TX. The vlan_tci_outer field selected from rte_mbuf structure
for sglort is not used in fm10k now.
In FTAG based forwarding mode, the switch will forward packets according
to glort info in FTAG rather than mac and vlan table.
To activate this feature, user needs to pass a devargs parameter to eal
for fm10k device like "-w 0000:84:00.0,enable_ftag=1". Currently this
feature is supported only on PF, because FM10K_PFVTCTL register is
read-only for VF.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
Remove the unused element request_lport_map in struct fm10k_mac_ops.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Some cleanups to better reflect the code that was actually pushed out to
the upstream Linux community.
Among the above cleanups, a few macros such as FM10K_RXINT_TIMER_SHIFT are
removed, but they are needed in dpdk/fm10k, so we have to put all these
necessary macros into fm10k_osdep.h.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Per comments from an upstream kernel patch, and looking at how TLV
LE_STRUCT code works, we actually want these structures to be 4byte
aligned, not 1byte aligned.
In practice, 1byte alignment has worked so far because all our
structures end up being a multiple of 4. But if a future TLV
structure were added that had a u8 or similar sticking on the end things
would break. Fix this by using 4byte alignment which will prevent the
TLV LE_STRUCT code from breaking. Update the comment explaining that we
need 4byte alignment of our structures.
Fixes: 925c862cbc21 ("fm10k/base: pack TLV overlay structures")
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
The comment for fm10k_iov_msg_lport_state_pf was changed during
review of kernel driver, and the new wording is slightly clearer.
Re-write the comment in base code based on this new wording.
Fix a number of mailbox comment issues with function header comments,
lower-case acronyms (i.e. FIFO, TLV), incorrect function names in
DEBUGFUNC(), duplicate comments and a stubbed-out header comment for
fm10k_sm_mbx_init.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
The vid variable name is shorthand for VLAN ID, so we should use this in
comments explaining what is happening.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
The Linux Kernel provides the OS a call "pcie_get_minimum_link" which
can crawl the PCIe tree and determine the actual minimum link speed of a
device which is a more general check than provided by
is_slot_appropriate. Thus, the kernel driver does not use or want the
is_slot_appropriate function call. Add a NO_IS_SLOT_APPROPRIATE_CHECK
definition which can be defined to remove the code.
If left undefined (the default) then the code will all be active and no
driver changes should be necessary.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Use memcpy instead of copying MAC address byte-by-byte.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Using the BIT macro can simplify the bit-shifting operation and make the
code look clean. Similar to how this is handled in the i40e base code,
define a macro for it in DPDK, so it can be used here too.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
"else" is not generally useful after a break or return.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Recommended line length maximum is 80 characters
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Add comments which properly explain the undocumented use of bits in
TDLEN register prior to VF initializing it to the correct value. Note
that the mechanism is entirely software-defined and explain its purpose
to help reduce confusion in the future.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
VF drivers must detect how many queues are available. Previously, the
driver assumed that each VF has at minimum 1 queue. This assumption is
incorrect, since it is possible that the PF has not yet assigned the
queues to the VF by the time the VF checks.
To resolve this, we added a check first to ensure that the first queue
is, in fact, owned by the VF at init_hw_vf time.
However, the code flow did not reset hw->mac.max_queues to 0.
In some cases, such as during reinit flows, we call init_hw_vf
without clearing the previous value of hw->mac.max_queues. Due to this,
when init_hw_vf errors out, if its error code is not properly handled
the VF driver may still believe it has queues which no longer belong to
it. Fix this by clearing the hw->mac.max_queues on exit due to errors.
Fixes: 8b8264bdb90d ("fm10k/base: check VF has a queue")
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Use bitshift instead of a divisor, because this is faster, and
eliminates any need for a '0' check. In our case, this even works
out because default Gen3 will be 0.
Because of this, we are also able to remove the check for non-zero value
in the VF code path since that will already be the default Gen3 case.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Make functions that are only referenced locally static.
Wrap fm10k_msg_data fm10k_iov_msg_data_pf[] in the new ifndef
NO_DEFAULT_SRIOV_MSG_HANDLERS so that drivers with custom SR-IOV
message handlers can strip it.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Since the resultant data type of the mac_update.mac_upper field is u16,
it does not make sense to typecast u8 variables to u32 first.
Fixes: 7223d200c227 ("fm10k: add base driver")
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
The new share code makes fm10k_msg_update_pvid_pf function static, so we
can not refer to it now in fm10k_ethdev.c. The registered PF handler is
almost the same as the default PF handler, removing it has no impact on
mailbox.
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Tested-by: Heng Ding <hengx.ding@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Using SSE instructions to parse error flags in HW Rx descriptor,
then set corresponding bits of mbuf.
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Acked-by: Cunming Liang <cunming.liang@intel.com>
When the TX function tries to free a bunch of mbufs, it will free
them one by one. This change will scan the free list and merge the
requests in case they belongs to same pool, then free once, which
will reduce cycles on freeing mbufs.
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Acked-by: Shaopeng He <shaopeng.he@intel.com>
fm10k switch core uses source MAC + VID + SGLORT to do
look up in MAC table. If no match, an exception interrupt
will be sent to the switch manager. Too much of this kind
of exception interrupts cause switch manager side high CPU
usage.
To reproduce this issue, one DPDK testpmd runs on a server
with one fm10k NIC, mac forwards test traffic from one of
fm10k ports to another port. The CPU usage for the switch
manager will go up to about 20% for test traffic rate at
10G bps, comparing to near 0% for no test traffic.
This patch fixes this issue. A default SGLORT is assigned
to each TX queue. This default value works for non-VMDq mode
and current VMDq example. For advanced VMDq usage, e.g.
different source MAC address for different TX queue, FTAG
forwarding function could be used to change this default
SGLORT value.
Fixes: 9ae6068c86da ("fm10k: add dev start/stop")
Signed-off-by: Shaopeng He <shaopeng.he@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
In FM10K, a single PCIe port can derive out a few logical ports,
like SRIOV PF/VF devices, VMDQ objects. To better manage them, FM10K
silicon assigns a Unique GLORT ID to each logical port.
When a logical port sends a broadcast packet, the silicon will flood
it to all logical ports, including the one that sent the broadcast packet.
To prevent this, silicon has an rxq register to store the glort id of
the logical port that queue binds to.
FM10K has a switch core inside, which has a loopback suppression
mechanism in the switch level. Switch level loopback suppression mostly
works for the ether port traffic.
This patch assigns a SGLORT for each RX queue, and enables PCIe port
level loopback suppression.
Signed-off-by: Shaopeng He <shaopeng.he@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
When the PF establishes a connection with Switch Manager(SM), it receives
a logical port range from SM, and registers certain logical ports from
that range. Then a default VID will be sent back from the SM.
This whole transaction - finishing with the default VID being set -
needs to be completed before dev_init returns. If not, the interrupt
setting will subsequently be changed in dev_start according to the RX
queue number, and that can cause this transaction to fail.
Signed-off-by: Shaopeng He <shaopeng.he@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Acked-by: Michael Qiu <michael.qiu@intel.com>
Interrupt mode framework has per-queue enable/disable functions.
Implement these two functions for fm10k driver.
Signed-off-by: Shaopeng He <shaopeng.he@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Acked-by: Michael Qiu <michael.qiu@intel.com>
Previous dev_stop function stops the rx/tx queues. This patch adds logic
to disable rx queue interrupt, clean the datapath event and queue/vector
map.
Signed-off-by: Shaopeng He <shaopeng.he@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Acked-by: Michael Qiu <michael.qiu@intel.com>
In interrupt mode, each rx queue can have one interrupt to notify the
application when packets are available in that queue. Some queues
also can share one interrupt.
Currently, fm10k needs one separate interrupt for mailbox. So, only those
drivers which support multiple interrupt vectors e.g. vfio-pci can work
in fm10k interrupt mode.
This patch uses the RXINT/INT_MAP registers to map interrupt causes
(rx queue and other events) to vectors, and enable these interrupts
through kernel drivers like vfio-pci.
Signed-off-by: Shaopeng He <shaopeng.he@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Acked-by: Michael Qiu <michael.qiu@intel.com>
rx_descriptor_done is used by interrupt mode example application
(l3fwd-power) to check rxd DD bit to decide the RX trend,
then l3fwd-power will adjust the cpu frequency according to
the result.
Signed-off-by: Shaopeng He <shaopeng.he@intel.com>
Acked-by: Jing Chen <jing.d.chen@intel.com>
Acked-by: Michael Qiu <michael.qiu@intel.com>
In fm10k, PF, VF, VMDQ or queues binding to flow director rule can
be considered as a logical port. Original implementation only creates
a single port for all cases. This change creates 128 logical ports;
first 64 for PF and VMDQ, second 64 for flow director.
Registers DGLORTDEC/DGLORTMAP define rules for how to classify packets
into different queues. Currently only PF and VMDQ cases are considered.
This change add rules for flow director.
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Acked-by: Shaopeng He <shaopeng.he@intel.com>
In fm10k_recv_scattered_pkts function, a packet is stored in a linked list,
offload flags such as PKT_RX_VLAN_PKT should be set in the first segment.
Fixes: 6b59a3bc82b1 ("fm10k: fix VLAN in Rx mbuf")
Signed-off-by: Wang Xiao W <xiao.w.wang@intel.com>
Acked-by: Shaopeng He <shaopeng.he@intel.com>
fix the error reported by checkpatch:
"ERROR: return is not a function, parentheses are not required"
remove parentheses in return like:
"return (logical expressions)"
remove parentheses in return a function like:
"return (rte_mempool_lookup(...))"
Fixes: 6307b909b8e0 ("lib: remove extra parenthesis after return")
Signed-off-by: Huawei Xie <huawei.xie@intel.com>
After introducing vPMD feature, fm10k driver will select best
Rx/Tx in running time. Original implementation selects Rx/Tx
silently without notifications.
This patch adds debug info to notify user what actual Rx/Tx
func are used.
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Steps to reproduce the bug:
1. All Rx offloading is disabled and start the device, then
Vector Rx is used.
2. Stop the device. Re-configure to enable hw_ip_checksum = 1,
start the device again.
3. In this case, assume regular Rx should be used since Vector
Rx doesn't support ip checksum offload. But actually Vector
Rx is used and cause checksum won't be done by hardware.
The reason is after re-configuring, driver misses an "else" in
func fm10k_set_rx_function(). Then Rx func in last round are
used.
Fixes: 77a8ab47eb38 ("fm10k: select best Rx function")
Reported-by: Xiao Wang <xiao.w.wang@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Acked-by: Xiao Wang <xiao.w.wang@intel.com>
