Age | Commit message (Collapse) | Author |
|
commit dd6d25e1cb573750dedd0d1b04e9cd0b6418cd36 from
git@git.assembla.com:cavium/WindRiver.linux.git
Add support for HIGIG2 protocol header to the kpu parser profile
Add a NPC_LT_LA_HIGIG2_ETHER for a combined header of HIGIG2 and
Ethernet. Add a NPC_LT_LA_IH_NIX_HIGIG2_ETHER for a combined
header of nix_ih, HIGIG2 and Ethernet.
Also add 2 upper flags in LA to indicate the presence of nix_ih
and HIGIG2.
Change-Id: I652ed86ab1f06f5b7f90ac4e985e1c0c854108b8
Signed-off-by: Hao Zheng <haoz@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17841
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit a108847efc4ad4a07bb7f5a2612391047e979620 from
git@git.assembla.com:cavium/WindRiver.linux.git
For the PHY host-side, LIO3 uses the Marvell proprietary PCS types
25GBASE-R2 and 50GBASE-R4; each is combined (at the PHY line-side) with its
counterpart (25G/50G Ethernet Consortium) standard PCS type (25GBASE-R or
50GBASE-R2).
ethtool is not aware of and does not support Marvell proprietary PCS types;
it only knows about standard PCS types. So the PF driver must pass (to
ethtool) information about LIO3's PHY line-side (not PHY host-side).
To that end, replace the entries of the cgx_link_mode[] lookup table that
correspond to the proprietary PCS types with their standard counterparts.
Also, set the bits of the macro OTX2_ETHTOOL_SUPPORTED_MODES that
correspond to the proprietary modes/types.
Change-Id: I993906364747abb4f75ae5ce235bf6af96ee70d0
Signed-off-by: Felix Manlunas <fmanlunas@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17825
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
(cherry picked from commit c0b60c39f136e42619e9a8602b52e299af1359d1)
Reviewed-on: https://sj1git1.cavium.com/17834
[Kevin: Just some minor context mods in order to port to linux-yocto]
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 1f56892993fb7d9b4aa13ed8481e3422e3d87ab1 from
git@git.assembla.com:cavium/WindRiver.linux.git
In case of PTP enable/disable event, PF should inform its VFs
about the event. So that VFs can adjust the shifted data offset
(viz 8 bytes on PTP enable).
Change-Id: I83d45441ee18c53bad3b68249c32cdc273cf4fc9
Signed-off-by: Harman Kalra <hkalra@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17618
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 3d41081109f29fdf84517b3cd1220852f783b382 from
git@git.assembla.com:cavium/WindRiver.linux.git
It used to be that the intended 1 second mbox response timeout was not very
precise such that the effective timeout was longer than 1 second. Recently
code was checked in to make the timeout very precise.
It turns out that the old, imprecise, longer-than-1-second timeout was
beneficial. Now that it's been replaced with a precise 1-second timeout,
we sometimes see these alarming error messages from ethtool:
Cannot set new settings: Input/output error
Cannot set private flags: Input/output error
These turn out to be false alarms because the operations (of changing
Ethernet speed or setting private flags) actually take effect.
One second is not enough time for ATF to detect the arrival of a CGX_CMD,
to execute the command, and to send back a response.
Increase the timeout to 2 seconds for ample safety margin.
Change-Id: Ia663fdf55a7a1417fc2d8b9ee751cab0903a7710
Signed-off-by: Felix Manlunas <fmanlunas@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17678
Tested-by: sa_ip-sw-jenkins <sa_ip-sw-jenkins@marvell.com>
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
(cherry picked from commit b842f3df534a85e70811b4490900a10d17f46f99)
Reviewed-on: https://sj1git1.cavium.com/17824
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 671aef6d85d317e805ac179c3d6d2983835e9ea1 from
git@git.assembla.com:cavium/WindRiver.linux.git
Use BLKTYPE instead of BLKADDR in register offset
macros so that NIX1 block registers can also be
accessed when required.
Change-Id: I4f32885fe2fa28beb3483f84c3b13d4eb5818c04
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17664
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 72df259bd2f894a62c3c78e97883afe90940243c from
git@git.assembla.com:cavium/WindRiver.linux.git
Switch headers are designed to support better flow control and
loadbalancing etc. When switch headers like EDSA, Higig2 etc are
present in ingress or egress pkts default iKPU index (or PKIND)
used by NPC to parse pkts will not work as there are additional
headers appended to the pkt. Hence a separate Pkind is chosen on
Rx and/or Tx sides to tell to KPU to parse the pkts accordingly.
Changes to NPC KPU parse profile to parse headers using these iKPU
indices will be pushed as a separate patch
Change-Id: I1796bafadb507a528d41f96376bf44eb33cc4bb3
Signed-off-by: hariprasad <hkelam@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17520
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 5fe1527404347eba09db0d56b92c5cbf0a2f3092 from
git@git.assembla.com:cavium/WindRiver.linux.git
In otx2_get_link_mode_info(), the signed integer literal "1" is left-
shifted by a value obtained by the table lookup
"cgx_link_mode[bit_position]". The result is then implicitly promoted from
signed int to u64 before it gets bitwise-ORed and assigned to
"ethtool_link_mode" (whose type is u64).
This is problematic if the result of the table lookup (which is used as the
number of bit positions to left shift) is greater than or equal to 31 (one
less than the bitwidth of an int). Such a situation will lead to a wrong
value for "ethtool_link_mode", causing ethtool to show wrong or incomplete
information in the list of supported link modes.
Fix it by adding the "ULL" suffix to "1".
Change-Id: I411befd4ad23df2566bf5e9cd6141faeeb53a26c
Signed-off-by: Felix Manlunas <fmanlunas@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17524
Tested-by: sa_ip-sw-jenkins <sa_ip-sw-jenkins@marvell.com>
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
(cherry picked from commit 8b1d4774140e9b1ff99a49e2b6ab8b610631c2e9)
Reviewed-on: https://sj1git1.cavium.com/17537
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 751d5a0b04b7af458779149ae33cf44071e9f905 from
git@git.assembla.com:cavium/WindRiver.linux.git
This patch fixes cleaning SQ/CQ queue interrupt bits and
LBK VF interface state after changing Queue length or
number of TX/RX queues using ethtool.
Change-Id: Ic0b98cc4eee22ce32c2247a8c1ffc89c3dbea6aa
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16691
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 9a12d900323a132d38f0f238a971c31458ac93cb from
git@git.assembla.com:cavium/WindRiver.linux.git
msleep(1) in "mbox_wait_rsp()" often sleep longer i.e, ~10m.
Replacing with usleep_range() as it builts on hrtimer,
the wakeup will be very precise.
Change-Id: I8c674f507a7d36422b1f7f73b344e25ce53a6eef
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16690
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit debcb5884a13363a4d387ffa7cac2dd56f5ac568 from
git@git.assembla.com:cavium/WindRiver.linux.git
Functions which install default mcam rules like unicast, enable
entries as well. This patch modifies that behavior to enable
only when they were enabled before. This simplifies logic such
that traffic is received only after rvu_mbox_handler_nix_lf_start_rx
is called.
Change-Id: I936097892c63f5210dd62e4328a2b89db45fc16f
Signed-off-by: hariprasad <hkelam@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17493
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 68c4f101435f82b4071116ff7457c51ffcdd90a8 from
git@git.assembla.com:cavium/WindRiver.linux.git
Irrespective of vf interface state always send request to AF.
Change-Id: I0cae60a18cef19cd95990c4a17ec61429fc2ce2e
Signed-off-by: hariprasad <hkelam@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17412
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit a58cab5aa7d440e7e9608c55a4814e862e93a55d from
git@git.assembla.com:cavium/WindRiver.linux.git
Update rvu pfvf mac address in below cases
* when fwdata has invalid mac
* pf/vf request change in mac address
Change-Id: I7d9da0e5aa810175c1aa6438c787e9f525c66942
Signed-off-by: hariprasad <hkelam@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17411
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit b58685bf513cb05f282dcfb7817664ac711ee4c0 from
git@git.assembla.com:cavium/WindRiver.linux.git
Get MSIX base address from firmware via fwdata structure
instead of from CSR as the later will cause issue when kernel crashes
and Kexec/Kdump loads secondary kernel and AF driver in the secondary
kernel reads IOVA written by primary.
Change-Id: I5a54638e2909e208537c68d4725f8f159201fc9d
Signed-off-by: Rakesh Babu <rsaladi2@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/17057
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 367da0d198068dd4b1310beb78ec4eb62729f7af from
git@git.assembla.com:cavium/WindRiver.linux.git
OCTEONTX2 NPC KPU parser configuration can be adapted to parse custom
packet frames. Such configuration is called a "KPU profile". It defines
both the KPU state machine as well as the LD/LTYPE values used in MCAM
classification and visible in packet metadata.
For performance and interoperability reasons it is best to leave a
common set of LD/LTYPE flags static for all users of the HW.
This commit reserves top 2 LTYPE values in each LD to allow custom
creating profiles with custom protocols that would be interoperable with
standard profile users.
Change-Id: I31f907850da114c44c25db6b1190f6deddf7471a
Signed-off-by: Stanislaw Kardach <skardach@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16546
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 73be0de715d6027c9e32fb91c659e7caae032ec5 from
git@git.assembla.com:cavium/WindRiver.linux.git
Update address for global CGX RX_STATS by passing proper lmac value.
Change-Id: I9a00411b993a11a6c7328af9b519aa809f659444
Signed-off-by: hariprasad <hkelam@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16569
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit b86183aa6f19548eab924c7155f4a35005965798 from
git@git.assembla.com:cavium/WindRiver.linux.git
Keysize of default profile is vaildated using a
variable which gets initialized later. Hence get
key size from register for the validation.
Change-Id: Ib44ebfc1c9c9c322ebc0e0228fe0b613333214d6
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16520
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit e3f9d403368f5339a345cfa0098449f465300564 from
git@git.assembla.com:cavium/WindRiver.linux.git
Initialize mcam resources like key width after a
profile is loaded because loaded profile may
change key width.
Change-Id: Ia372ab9357efe6e1cb879ba3e6b8b5764b2eeb88
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16397
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 087085463bd075d4b14b8220bf202f8877ea203b from
git@git.assembla.com:cavium/WindRiver.linux.git
AF driver need not bail out when no CGX devices are
found. LBK devices can be used in that case.
Change-Id: Ib480e9be99da3fed6d9affcc0ba137731f8c4116
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16285
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit f511c212d5166756041444359251afff023f208b from
git@git.assembla.com:cavium/WindRiver.linux.git
Changing the MTU for high to low during packet transmission may
lead to SMQ meta-descriptor enqueue error due to miss match in packet
length and configured SMQ MAXLEN.
Hence, reset the PF/VF inorder to flush SMQ before configuring new MTU.
Change-Id: Icaa98653781907d9297431fc6a0b1d466d09b3e2
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16447
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit e4804915b954374aeaa1ab524d3c9b40bd66d5e6 from
git@git.assembla.com:cavium/WindRiver.linux.git
Explicitly disable backpressure in CQ and Aura contexts
during lf teardown to deassert backpressure on link as
clearing CQ.ena or AURA.ena is not sufficient.
Also enable pause frame config everytime lf is allocated to
override previous configurations.
Change-Id: Ibe50c11e6e7992e56979bbb9757e2f3cc3c3f519
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16217
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 90c64b10f51d921844ada550493594877724686a from
git@git.assembla.com:cavium/WindRiver.linux.git
This patch adds pci driver shutdown support device
teardown on system reboot.
Change-Id: I820ee817c1b8ed399d3e309b5060334c03c5f5e5
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/16060
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit f18e37bd1cf6ffe13eff7c78154e222eeb50c676 from
git@git.assembla.com:cavium/WindRiver.linux.git
Support vlan offload in receive path by configuring NIX
to strip outer vlan.
Change-Id: I3b6c6be28fe9f1804843b6cc30d18afd06d76ed5
Signed-off-by: hariprasad <hkelam@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15967
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 8f3dbde036b2e6b24d68aa66cf8f980be417d987 from
git@git.assembla.com:cavium/WindRiver.linux.git
As long as ISR is not cleared upcoming interrupts
will be lost hence clear the ISR first and queue
work for processing.
Change-Id: Ifd68bbbe16b91bf5640d8c7b19478f24c205da6e
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15919
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 7f3d06ad12b9f5cd4cd1bcb81d6f946817401949 from
git@git.assembla.com:cavium/WindRiver.linux.git
Add change to match NVGRE as ltype for RSS in sync with KPU profile.
Change-Id: I7ee72266dab30d60bb9d9a281ee49f73ee5d2657
Signed-off-by: Kiran Kumar K <kirankumark@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15626
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit ea8daebe069245b205b7bbf875113b6144edc97f from
git@git.assembla.com:cavium/WindRiver.linux.git
Support changing cgx link mode via ethtool.
Example usage: ethtool -s ethX advertise <hex_value>
Change-Id: I2e77f87c8ca0a6e684c446723f823fd51780f71e
Signed-off-by: Christina Jacob <cjacob@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15607
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
[Kevin: Change the definition of OTX2_ETHTOOL_ALL_MODES
due to the __ETHTOOL_LINK_MODE_LAST chagne.]
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit aaca7982f8d0883ff2b8c17737cb18607f4cd99e from
git@git.assembla.com:cavium/WindRiver.linux.git
AF driver sets mac address for all PFs/VFs.
Setting MAC address again in otx2_open enables
RX traffic before setting up NAPI hence removed
it.
Change-Id: Icd6182d4428b81100c5d3d55dcddd2f2da870caa
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15665
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit eb65f5377da6cccad942ccb4c18819d8e3dc8a8a from
git@git.assembla.com:cavium/WindRiver.linux.git
Changes for supporting cgx link mode change using ethtool.
Change-Id: I91cda92064ded056b074725016795f17aa114167
Signed-off-by: Christina Jacob <cjacob@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15606
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 889b9c4ab3ab62a5f56abe05d4e9b7eb64722a0f from
git@git.assembla.com:cavium/WindRiver.linux.git
clean up speed change support via ethtool.
Change-Id: I3c9f606f04c67c01421660bb3759331aa5fd80b0
Signed-off-by: Christina Jacob <cjacob@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15605
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit f03662a20e69b916ae13ac28841381962814b358 from
git@git.assembla.com:cavium/WindRiver.linux.git
This patch fixes a problem where the catchall entry in KPU2 misses
the corresponding action entry.
Change-Id: I5525b5029623007af087170ca30e24592fcf8e6e
Signed-off-by: Hao Zheng <haoz@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15602
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit e849a34c43680a5dbe0bf18bc1317e0d290de3e2 from
git@git.assembla.com:cavium/WindRiver.linux.git
update kpu profile so that nvgre has a separate ltype LD_NVGRE
instead of combined with LD_GRE
Change-Id: I3ef48896470cfed83c203e430915252c359db210
Signed-off-by: Hao Zheng <haoz@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15508
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
|
|
[ Upstream commit 5aafeb74b5bb65b34cc87c7623f9fa163a34fa3b ]
Running old skge driver on PowerPC causes checksum errors
because hardware reported 1's complement checksum is in little-endian
byte order.
Reported-by: Benoit <benoit.sansoni@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
|
|
commit 6c958e5015dd3b1307f8e4382141648c14eb3b70 from
git@git.assembla.com:cavium/WindRiver.linux.git
This patch allows to attach Marvell PTP Driver to the PTP device
on t98.
Change-Id: I7dad35443c8ad96c6529f4dd2eb6c0bdf9753013
Signed-off-by: Tomasz Michalec <tomasz@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15215
Reviewed-by: Chandrakala Chavva <cchavva@marvell.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 446d57d2e083ee57defc821d23c52687cc880d75 from
git@git.assembla.com:cavium/WindRiver.linux.git
Ensure that intf_down flag is updated by using a barrier
before enabling traffic.
Change-Id: I61e003346030b69278d56b4d2a7bba5891470284
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15093
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit b785026db2f2feec5d76d7a63421c41bc50294da from
git@git.assembla.com:cavium/WindRiver.linux.git
Upon enabling SQ, HW immediately fetches 2 SQB pointers from
the mapped Aura, these will not be freed back to Pool once SQ
is disabled. So while cleanup if we do alloc pointer from Aura
and free them to kernel then 2 SQB pointers will be leaked ie
never freed. Hence Aura alloc pointer cannot be used, this patch
fixes this issue by saving the SQB pointers freed to Pool in the
driver itself and uses that list while cleanup.
Change-Id: I0ecd2a71f55d0cbf16b9d5c57ec3bc983bcb1240
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/15089
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 203b8920a96e69616f69b4966f1b97bfb3245ea9 from
git@git.assembla.com:cavium/WindRiver.linux.git
This commit adds the mac address that is programmed from
application into the RVU pfvf structure. The mac address
retrieval will return the address from RVU pfvf structure.
Change-Id: Ia9f4685b49a794dd678ed14f0e0e91d3b89de1d4
Signed-off-by: Vidhya Vidhyaraman <Vidhya.Raman@cavium.com>
Reviewed-on: https://sj1git1.cavium.com/14399
Tested-by: sa_ip-sw-jenkins <sa_ip-sw-jenkins@marvell.com>
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
(cherry picked from commit 232b8af019a9b0cedcf1c82ac6592282af2c74fb)
Reviewed-on: https://sj1git1.cavium.com/15064
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 369c58581073495eae851f4813a5133cc37ef92f from
git@git.assembla.com:cavium/WindRiver.linux.git
NPC parser changes to combine NPC_LT_LB_STAG and NPC_LT_LB_QINQ
to NPC_LT_LB_STAG_QINQ, to facilitate formulating a single MCAM
entry to match on packets with one or stacked vlan tags.
Change-Id: I5fa7dab5c99f2ee964123b6c81f7af695d72ba40
Signed-off-by: Hao Zheng <haoz@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14808
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 13d8c3dd95622989a2f9499b64dbf2dfa3189f44 from
git@git.assembla.com:cavium/WindRiver.linux.git
Fix compilation issue introduced by patch
'commit e833c99ef4dd ("octeontx2-af: Transmit packets during SMQ flush")'
Change-Id: Ice1473e779d04263577e73d5e83e2d925642cdc9
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14606
Tested-by: sa_ip-sw-jenkins <sa_ip-sw-jenkins@marvell.com>
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Reviewed-on: https://sj1git1.cavium.com/14930
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit d2a41a1a926cee01f0a2ac31a626afab33df0475 from
git@git.assembla.com:cavium/WindRiver.linux.git
NPC reports a layer error if none of the control flags are
set in TCP packet and software drops the packets.
Which is not RFC compliant.
This patch fixes the issue by forwarding the packets
without dropping.
Change-Id: I1a2d371603124abdb2d57543478307bc13c83114
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14568
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 27cfc1441d82840fe65d3ccb00b3ae8eeda8e489 from
git@git.assembla.com:cavium/WindRiver.linux.git
Marvell SOHO switches support distributed switch architecture by
inserting a DSA tag right after ethernet header with a tpid of
0xDADA. Marvell Falcon switches support distributed switch
architecture by inserting either a extended DSA tag of 8 bytes
or a eDSA tag of 16 bytes right after the ethernet SMAC. These
tags don't have a tpid field.
This patch provides parser support for the tags mentioned above
optionally followed by a VLAN tag.
For extended DSA and eDSA tags, a special PKIND of 62 is used,
as these tags don't contain a tpid field.
Change-Id: I1167f2a1b47793432ac4649e1f85f9621fac3bc4
Signed-off-by: Hao Zheng <haoz@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/13970
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit accd6b7e1f2e7078aeaa9442862ca2165504ab2a from
git@git.assembla.com:cavium/WindRiver.linux.git
NIX SMQ flush may not complete if no send queue is
transmitting packets. Hence packets need to be
transmitted from a send queue when SMQ flush is
initiated. This patch adds a SMQVF driver which
is a stripped down version of VF netdev driver.
The driver exports a function which can be called
by AF driver during SMQ flush to transmit packets.
A new VF of AF with pcifunc 17 is used for this
purpose. So 17th AF VF should not attached to
userspace.
Change-Id: I6592f7615f33f423de2d464f7441864443f0dec4
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14560
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 010491289485c563d340d4782e33124bb00a346b from
git@git.assembla.com:cavium/WindRiver.linux.git
Enabling MCAM rule depends on whether the NIXLF
is initialized or not. MCAM rules for TX can
always be enabled because packet hitting a MCAM
rule in TX path implies that it was transmitted
by a initialized NIXLF.
Change-Id: I59f33ec042b852253240a3c29fec0136066ad6ec
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14559
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit f31c3d4bd2b2c467b38940f56a361ea2f4002270 from
git@git.assembla.com:cavium/WindRiver.linux.git
nix_smq_flush function is available to flush all
metadescriptors/packets from SMQ through PSE and
the send data path. Hence use it wherever
required.
Change-Id: I9b2c85971b95780475182e8a5bd8d82a7ee9dd42
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14558
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 062ac62b16496d2d2d215d9712df8c9f7f225a44 from
git@git.assembla.com:cavium/WindRiver.linux.git
Helper function is available to check given
device is LBK VF or not hence use it wherever
required.
Change-Id: Id0d548f51cd747fddbab7fde577d7968432b480e
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14476
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 16eaaa4a89729695d764d9a49203e9c257b11148 from
git@git.assembla.com:cavium/WindRiver.linux.git
Only even number of AF VFs are enabled currently
and last VF is disabled if number of VFs is odd.
Instead of disabling last VF altogether this patch
enables it. There is no harm in doing so since last
VF will remain with no pair for sending or receiving
traffic.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Change-Id: Ifdc0b4d90805a2c27c654f9147408b7f804d17de
Reviewed-on: https://sj1git1.cavium.com/14475
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 1a16c0f852a40bba93a6f3170992fba342ca811f from
git@git.assembla.com:cavium/WindRiver.linux.git
Message responses may time out at sender if there
is no receipent hence make the time out message as
dev_dbg.
Change-Id: Id68d7294f2ebe52d5530692703da2bb7c44c825a
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14491
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 59a200836fd89dd2322910d0d700c4459770ee6f from
git@git.assembla.com:cavium/WindRiver.linux.git
Added a debug message for MSIX vector alloc and initialization
failure.
Change-Id: Id8adef75b0f37150555ef39609303bdea524eb52
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14427
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit b17c623daff142f98b51c0234355bd33cc5f310e from
git@git.assembla.com:cavium/WindRiver.linux.git
Freeing buffer pointers to Aura needs a 128bit atomic store
to two 64bit registers. A regular STP can get fissioned into
two separate 64bit stores which will be ignored by Aura and
buffer pointer will be lost. Hence use a post increment STP.
Change-Id: I0e91250f16cfe875acdb1d2c1d051ddf90cf75de
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14426
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|
|
commit 41bf34f01baaa784960a3e5801a414eeea92ad0d from
git@git.assembla.com:cavium/WindRiver.linux.git
This patch fixes multiple issues
- netif_carrier_off() should be done before disabling
transmit at CGX/NPC level.
- Fixed SQE free count checking in otx2_xmit()
- In NAPI, added a check to wakeup a TXQ if incase it was stopped
earlier due to queue full.
- In otx2_open() 'pfvf->intf_down' should be cleared before
enabling Rx at CGX/NPC level. Otherwise if in between CQ IRQ
is raised then in NAPI CQ interrupt will be disabled forever.
- In otx2_stop(), RQ and SQ sizes should not be cleared, otherwise
upon interface DOWN and UP, RQ size will fallback to 256 which
is not supported on 96xx A0. Also if user has changed RQ/SQ size
via ethtool intf DOWN and UP will reset the sizes back to default.
- Fixed time and packet coalescing settings, user given value was
getting shifted due to wrong type used while clamping it to within
supported ranges.
Change-Id: I152ce646909e64ac6ff858ba0d253f3d59ea3d8d
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/14394
Reviewed-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Tested-by: Sunil Kovvuri Goutham <Sunil.Goutham@cavium.com>
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
|