The function name ixgbe_vmdq_mode_check is not right.
This function checks if Virtualization Technology is
enabled. It's for both VMDq and IOV.
Others may be misled by the current name.
Fixes: fe3a45fd41 ("ixgbe: add VMDq support")
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
In function "reassemble_packets()", the statement "end = secondlast;"
is redundant since there is another assignment "start = end = NULL;"
3 lines below. BTW, I removed the redundant braces in the conditional
statement "if (end->data_len > rxq->crc_len)".
Signed-off-by: Yong Wang <wang.yong19@zte.com.cn>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
When freeing up last mbuf, start->nb_segs should be decremented
by one. See also ixgbe process.
Fixes: 0e0da28cd8 ("i40e: add vector scatter Rx")
Signed-off-by: Chenghu Yao <yao.chenghu@zte.com.cn>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
When using the same file descriptor for both rx and tx, the
eth_dev_stop function would close the same fd twice. This
change prevents that from happening.
Fixes: 364e08f2bb ("af_packet: add PMD for AF_PACKET-based virtual devices")
Signed-off-by: Timmons C. Player <timmons.player@spirent.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Add support for PCI IDs which was removed as a part of
commit 0e7bd0b2d6 ("net/bnxt: remove support for few devices")
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
The only reason why bulk alloc disabled for the rings with
more than (IXGBE_MAX_RING_DESC - RTE_PMD_IXGBE_RX_MAX_BURST)
descriptors is the possible out-of-bound access to the dma
memory. But it's the artificial limit and can be easily
avoided by allocating of RTE_PMD_IXGBE_RX_MAX_BURST more
descriptors in memory. This will not interfere the HW and,
as soon as all rings' memory zeroized, Rx functions will
work correctly.
This change allows to use vectorized Rx functions with
4096 descriptors in Rx ring which is important to achieve
zero packet drop rate in high-load installations.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
The only reason why bulk alloc disabled for the rings with
more than (I40E_MAX_RING_DESC - RTE_PMD_I40E_RX_MAX_BURST)
descriptors is the possible out-of-bound access to the dma
memory. But it's the artificial limit and can be easily
avoided by allocating of RTE_PMD_I40E_RX_MAX_BURST more
descriptors in memory. This will not interfere the HW and,
as soon as all rings' memory zeroized, Rx functions will
work correctly.
This change allows to use vectorized Rx functions with
4096 descriptors in Rx ring which is important to achieve
zero packet drop rate in high-load installations.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Add PCI device ID for ConnectX-5 and enable multi-packet send for PF and VF
along with changing documentation and release note.
Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
If value of number of rxq/txq is different than
RTE_ETHDEV_QUEUE_STAT_CNTRS, limit per queue
stats/xstats to minimum of the two.
Fixes: 7634c5f915 ("net/qede: add queue statistics")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Allocate double the number of fastpath status block index
since the PF RX/TX queues are not sharing the status block.
This is an interim solution till other parts of the code
is modified to handle the same.
Fixes: f1e4b6c0ac ("net/qede: fix status block index for VF queues")
Signed-off-by: Harish Patil <harish.patil@qlogic.com>
- Fix minimum RX buffer size to 1024B
- Force enable scatter/gather mode if given RX buf size is lesser than MTU
- Adjust RX buffer size to cache-line size with overhead included
Fixes: bec0228816 ("net/qede: support scatter gather")
Fixes: 2ea6f76aff ("qede: add core driver")
Signed-off-by: Harish Patil <harish.patil@qlogic.com>
This patch adds i40e_flow_flush_tunnel_filter
function to flush all tunnel filters, including
filters in SW and HW.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_flush_ethertype_filter
function to flush all ethertype filters, including
filters in SW and HW.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_flush function to flush all
filters for users. And flow director flush function
is involved first.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_destroy_tunnel_filter
function to destroy a tunnel filter for users.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_destroy_ethertype_filter
function to destroy a ethertype filter for users.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_destroy function to destroy
a flow for users.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_create function to create a
rule. It will check if a flow matches ethertype filter
or flow director filter or tunnel filter, if the flow
matches some kind of filter, then set the filter to HW.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_parse_tunnel_filter to check
if a rule is a tunnel rule according to items of the
flow pattern, and the function also gets the tunnel info.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_parse_fdir_filter to check
if a rule is a flow director rule according to the
flow pattern, and the function also gets the flow
director info.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
This patch adds i40e_flow_validation function to check if
a flow is valid according to the flow pattern.
i40e_flow_parse_ethertype_filter is added first, it also
gets the ethertype info.
i40e_flow.c is added to handle all generic filter events.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Add support of restoring ethertype filter in case filter
dropped accidentally, as all filters need to be added and
removed by user obviously for generic filter API.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Currently there's no flow director filter stored in SW. This
patch stores flow director filters in SW with cuckoo hash,
also adds protection if a flow director filter has been added.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Currently there's no tunnel filter stored in SW.
This patch stores tunnel filter in SW with cuckoo
hash, also adds protection if a tunnel filter has
been added.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Currently there's no ethertype filter stored in SW.
This patch stores ethertype filter with cuckoo hash
in SW, also adds protection if an ethertype filter
has been added.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Dseg pointer is not initialized when the first segment is inlined
causing a segmentation fault in such situation.
Fixes: 2a66cf3789 ("net/mlx5: support inline send")
Signed-off-by: Shahaf Shuler <shahafs@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Too much data is uselessly written to the Tx doorbell, which since v16.11
may also cause Tx queues to behave erratically and crash applications.
This regression was seen on VF devices when the BlueFlame buffer size is
zero (txq->bf_buf_size) due to the following change:
- txq->bf_offset ^= txq->bf_buf_size;
+ txq->bf_offset ^= (1 << txq->bf_buf_size);
Fixes: 1d88ba1719 ("net/mlx5: refactor Tx data path")
Fixes: d5793daefe ("net/mlx5: reduce memory overhead for BF handling")
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Prefetching completion queue entries is inefficient because too few CPU
cycles are spent before their use, which results into cache misses anyway.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Use fewer instructions to copy the first two bytes of Ethernet headers to
work queue elements.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Gather function prototypes at the beginning of the file.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Let compiler automatically use the vector capabilities of the target
machine to optimize instructions.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Define a single work queue element type that encompasses them all. It
includes control, Ethernet segment and raw data all grouped in a single
place.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Prepare the code to write the Work Queue Element with vectorized
instructions.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Elad Persiko <eladpe@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
This patch includes changes for DCBX like:
- Return empty parameters for oper-params query when negotiation is not
complete
- Use the ieee specific mask value for reading the ethtype value in the
ieee dcbx mode
- Endian-ness conversion is not needed for priority<->TC field, as the
data is already being read/written by ecore in the bigendian way
- While writing the ets config, base driver incorrectly merges the input
values with the operational values. The values should be either set
or unset
- CEE selection field must be set regardless CEE/IEEE mode
- Fail the dcbx query for VF interfaces
- Semantic changes
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Add support for 8.14.x.x firmware.
The new firmware adds support for external PHY BCM8485x; configures
fixed link speed with transceiver/cable not supporting negotiation;
supports engine swap; supports overriding PCIe preset equalization
value; checks pause too long for ports and reads die temperature
every second for shutdown threshold.
It includes change in FLR flow when there is a SW initiated FLR.
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Add PCI IDs for new asic type (defined as CHIP_NUM_AH_xxx).
It supports 50G, 40G, 25G and 10G speeds.
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
As part of device probe, check if management FW crash dump logs are
available. If available, then log an warning message and update the
epoch value too. A new struct ecore_mdump_info is added to populate
dump info including the new "reason" field by reading shared memory
region.
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Add new option called external PBL (page base list) to ecore_chain_alloc
for future use. Mark chain as external if external PBL is provided.
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Add macros PTR_LO and PTR_HI (for converting pointer to HI and LOW bits for
passing to FW hsi handles).
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
In order to improve Tx switching performance an additional HSI parameter
called same-as-last-id is introduced. This resource allows FW caching the
txqs packet properties. Driver needs to set same-as-last-id to be equal to
the qzone.
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Receive indication that VFs are malicious and pass it to the
caller/clients and stop serving those VF's additional resource requests.
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Add a request for PF Function Level Reset (FLR) before a load request
Fix the location of the PF FLR initiation to be after ecore_get_hw_info()
(which invokes ecore_hw_info_port_num())
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
This commits adds:
- Type of service
- Next protocol ID
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Introduce initial software for rte_flow rules.
VLAN, VXLAN are still not supported.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Previously the capability bitmap for Rx offloads was mistakenly taken
from Tx capability bitmap field. This patch fixes the problem.
Signed-off-by: Jakub Palider <jpa@semihalf.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>
When VF receives a message from PF, it should check the return
value. But in i40evf_execute_vf_cmd the value is ignored and not
returned to the caller.
Fixes: 95cd21f45d ("i40evf: allocate virtchnl commands buffer per VF")
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Flows redirected to a specific queue do not have a valid RSS hash result
and the related mbuf flag must not be set.
Fixes: ecf60761fc ("net/mlx5: return RSS hash result in mbuf")
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Removed unused "reg_info" global variable from ixgbe driver.
cat build/app/testpmd.map | grep "Allocating common symbols" -A 15
Allocating common symbols
Common symbol size file
reg_info 0x18 build/lib/librte_pmd_ixgbe.a(ixgbe_ethdev.o)
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.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>
Apparently VF over Legacy PF doesn't work, as VF would fail after
getting the initial rejection message [instead of sending an additional
one where it asks for a FW override and see if it works].
Fixes: 22d07d939c ("net/qede/base: update")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Add a check and fail the VF's probe request if VF is already in
VF_ACQUIRED state.
Fixes: 22d07d939c ("net/qede/base: update")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Rx queue access is still done prior to the index being validated by PF.
Hence move Rx queue and status block validation check before accessing
Rx queue to prevent malicious VFs from using out-of-bound queue indices.
Fixes: 98bc693e19 ("net/qede/base: change queue start")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Fix to return error code ECORE_INVAL instead of 0 when EQ elements
is too large as done elsewhere in this function.
Fixes: 22d07d939c ("net/qede/base: update")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
There are certain conditions under which VF would infinitely send
ACQUIRE messages, as it will fail to understand that PF has rejected
the ACQUIRE request. Fix to reject multiple acquisition requests by VF.
Fixes: 22d07d939c ("net/qede/base: update")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
Remove unmeaningful function ID value in print.
Don't print the number of Multicast filters as part of Acquire response,
as this is an obsolete field which isn't enforced by PF.
Fixes: 86a2265e59 ("qede: add SRIOV support")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
The vendor_id and device_id are used to determine device type. If you
don't have them, then check for determining device type fails and is
always set to default device type.
Fixes: ec94dbc573 ("qede: add base driver")
Signed-off-by: Rasesh Mody <rasesh.mody@cavium.com>
HW can support up to 128 queues based on the NIC config/personality.
But most of the testing is done with 32 queues for PF and 16 for VF
device across different qede devices, so change here is to advertise
only those many instead of returning max queues supported by HW.
Fixes: 2ea6f76aff ("qede: add core driver")
Signed-off-by: Harish Patil <harish.patil@qlogic.com>
qede PMD does not support SR-IOV PF driver functionality, so
max_vfs is set to 0 to imply the same.
Fixes: 2ea6f76aff ("qede: add core driver")
Signed-off-by: Harish Patil <harish.patil@qlogic.com>
This patch includes the following:
- Fix missing hash_key_size advertisement
- Fix RSS hash query function
- Update RSS offload flag
- Accept user provided RSS configuration params via rx_adv_conf
in dev_configure()
- Decouple RSS configuration from common qed_update_vport() and
instead make use of existing RSS APIs for default RSS configuration
Fixes: 6d9e26c42c ("net/qede: get RSS hash configuration")
Fixes: 9c5d0a669f ("net/qede: fix RSS")
Signed-off-by: Harish Patil <harish.patil@qlogic.com>
In qede_mac_addr_add() a check is added to differentiate between
unicast/multicast mac to prevent a multicast mac from being wrongly added
to unicast filter table. Secondly, two separate lists will be used to keep
track of unicast/multicast mac filters to prevent duplicate filter
programming. The other change is to remove filter_config from struct
qed_eth_ops_pass and invoke the base APIs directly. This avoids the need
to have multiple structs and function calls.
Fixes: 2ea6f76aff ("qede: add core driver")
Signed-off-by: Harish Patil <harish.patil@qlogic.com>
Update the version of shared codes to cid-ixgbe.2016.11.21.tar.gz,
All files in net/ixgbe/base are developed by another team and
DPDK PMD uses them accordingly.
Signed-off-by: Wei Dai <wei.dai@intel.com>
This patch updates the configuration of PHY from Inphi (www.inphi.com)
to flush the register write with a reg read.
The Inphi PHY is configured in ixgbe_setup_mac_link_sfp_x550a.
The Inphi PHY setup flow has been updated to read configuration reg,
write only linear/non-linear, and then read (write flush).
Signed-off-by: Wei Dai <wei.dai@intel.com>
This register was incorrect when compared to the data sheet.
Even though the driver doesn't currently use this register,
it is better to fix it upstream.
Fixes: af75078fec ("first public release")
Signed-off-by: Wei Dai <wei.dai@intel.com>
