David S. Miller [Sat, 13 Feb 2021 00:31:46 +0000 (16:31 -0800)]
Merge branch 'mptcp-genl-events'
Mat Martineau says:
====================
mptcp: Add genl events for connection info
This series from the MPTCP tree adds genl multicast events that are
important for implementing a userspace path manager. In MPTCP, a path
manager is responsible for adding or removing additional subflows on
each MPTCP connection. The in-kernel path manager (already part of the
kernel) is a better fit for many server use cases, but the additional
flexibility of userspace path managers is often useful for client
devices.
Patches 1, 2, 4, 5, and 6 do some refactoring to streamline the netlink
event implementation in the final patch.
Patch 3 improves the timeliness of subflow destruction to ensure the
'subflow closed' event will be sent soon enough.
Patch 7 allows use of the GENL_UNS_ADMIN_PERM flag on genl mcast groups
to mandate CAP_NET_ADMIN, which is important to protect token information
in the MPTCP events. This is a genetlink change.
Patch 8 adds the MPTCP netlink events.
====================
Florian Westphal [Sat, 13 Feb 2021 00:00:01 +0000 (16:00 -0800)]
mptcp: add netlink event support
Allow userspace (mptcpd) to subscribe to mptcp genl multicast events.
This implementation reuses the same event API as the mptcp kernel fork
to ease integration of existing tools, e.g. mptcpd.
Supported events include:
1. start and close of an mptcp connection
2. start and close of subflows (joins)
3. announce and withdrawals of addresses
4. subflow priority (backup/non-backup) change.
Florian Westphal [Fri, 12 Feb 2021 23:59:55 +0000 (15:59 -0800)]
mptcp: split __mptcp_close_ssk helper
Prepare for subflow close events:
When mptcp connection is torn down its enough to send the mptcp socket
close notification rather than a subflow close event for all of the
subflows followed by the mptcp close event.
This splits the helper: mptcp_close_ssk() will emit the close
notification, __mptcp_close_ssk will not.
Matthieu Baerts [Fri, 12 Feb 2021 23:20:30 +0000 (15:20 -0800)]
selftests: mptcp: fail if not enough SYN/3rd ACK
If we receive less MPCapable SYN or 3rd ACK than expected, we now mark
the test as failed.
On the other hand, if we receive more, we keep the warning but we add a
hint that it is probably due to retransmissions and that's why we don't
mark the test as failed.
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/148 Signed-off-by: Matthieu Baerts <[email protected]> Signed-off-by: Mat Martineau <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Matthieu Baerts [Fri, 12 Feb 2021 23:20:28 +0000 (15:20 -0800)]
selftests: mptcp: fix ACKRX debug message
Info from received MPCapable SYN were printed instead of the ones from
received MPCapable 3rd ACK.
Fixes: fed61c4b584c ("selftests: mptcp: make 2nd net namespace use tcp syn cookies unconditionally") Signed-off-by: Matthieu Baerts <[email protected]> Signed-off-by: Mat Martineau <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Paolo Abeni [Fri, 12 Feb 2021 23:20:27 +0000 (15:20 -0800)]
selftests: mptcp: dump more info on errors
Even if that may sound completely unlikely, the mptcp implementation
is not perfect, yet.
When the self-tests report an error we usually need more information
of what the scripts currently report. iproute allow provides
some additional goodies since a few releases, let's dump them.
Peng Li [Fri, 12 Feb 2021 03:21:05 +0000 (11:21 +0800)]
net: hns3: refactor out hclge_set_vf_vlan_common()
To improve code readability and maintainability, separate
the command handling part and the status parsing part from
bloated hclge_set_vf_vlan_common().
Peng Li [Fri, 12 Feb 2021 03:21:02 +0000 (11:21 +0800)]
net: hns3: refactor out hclgevf_cmd_convert_err_code()
To improve code readability and maintainability, refactor
hclgevf_cmd_convert_err_code() with an array of imp_errcode
and common_errno mapping, instead of a bloated switch/case.
Peng Li [Fri, 12 Feb 2021 03:21:01 +0000 (11:21 +0800)]
net: hns3: refactor out hclge_cmd_convert_err_code()
To improve code readability and maintainability, refactor
hclge_cmd_convert_err_code() with an array of imp_errcode
and common_errno mapping, instead of a bloated switch/case.
ixgbe: store the result of ixgbe_rx_offset() onto ixgbe_ring
Output of ixgbe_rx_offset() is based on ethtool's priv flag setting, which
when changed, causes PF reset (disables napi, frees irqs, loads
different Rx mem model, etc.). This means that within napi its result is
constant and there is no reason to call it per each processed frame.
Add new 'rx_offset' field to ixgbe_ring that is meant to hold the
ixgbe_rx_offset() result and use it within ixgbe_clean_rx_irq().
Furthermore, use it within ixgbe_alloc_mapped_page().
Last but not least, un-inline the function of interest as it lives in .c
file so let compiler do the decision about the inlining.
ice: store the result of ice_rx_offset() onto ice_ring
Output of ice_rx_offset() is based on ethtool's priv flag setting, which
when changed, causes PF reset (disables napi, frees irqs, loads
different Rx mem model, etc.). This means that within napi its result is
constant and there is no reason to call it per each processed frame.
Add new 'rx_offset' field to ice_ring that is meant to hold the
ice_rx_offset() result and use it within ice_clean_rx_irq().
Furthermore, use it within ice_alloc_mapped_page().
i40e: store the result of i40e_rx_offset() onto i40e_ring
Output of i40e_rx_offset() is based on ethtool's priv flag setting,
which when changed, causes PF reset (disables napi, frees irqs, loads
different Rx mem model, etc.). This means that within napi its result is
constant and there is no reason to call it per each processed frame.
Add new 'rx_offset' field to i40e_ring that is meant to hold the
i40e_rx_offset() result and use it within i40e_clean_rx_irq().
Furthermore, use it within i40e_alloc_mapped_page().
Last but not least, un-inline the function of interest so that compiler
makes the decision about inlining as it lives in .c file.
Whole zero-copy variant of clean Rx IRQ is executed when xsk_pool is
attached to rx_ring and it can happen only when XDP program is present
on interface. Therefore it is safe to assume that program is always
!NULL and there is no need for checking it in ice_run_xdp_zc.
dev_validate_mtu checks that mtu value specified by user is not less
than min mtu and not greater than max allowed mtu. It is being done
before calling the ndo_change_mtu exposed by driver, so remove these
redundant checks in ice_change_mtu.
Similar thing has been done in i40e, as there is no real need for having
the sk_buff pointer in each rx_buf. Non-eop frames can be simply handled
on that pointer moved upwards to rx_ring.
i40e_is_non_eop had a leftover comment and unused skb argument which was
used for placing the skb onto rx_buf in case when current buffer was
non-eop one. This is not relevant anymore as commit e72e56597ba1
("i40e/i40evf: Moves skb from i40e_rx_buffer to i40e_ring") pulled the
non-complete skb handling out of rx_bufs up to rx_ring. Therefore,
let's adjust the function arguments that i40e_is_non_eop takes.
Furthermore, since there is already a function responsible for bumping
the ntc, make use of that and drop that logic from i40e_is_non_eop so
that the scope of this function is limited to what the name actually
states.
Philipp Borgers [Mon, 25 Jan 2021 15:07:44 +0000 (16:07 +0100)]
mac80211: add STBC encoding to ieee80211_parse_tx_radiotap
This patch adds support for STBC encoding to the radiotap tx parse
function. Prior to this change adding the STBC flag to the radiotap
header did not encode frames with STBC.
Felix Fietkau [Wed, 27 Jan 2021 05:57:35 +0000 (06:57 +0100)]
mac80211: minstrel_ht: remove sample rate switching code for constrained devices
This was added to mitigate the effects of too much sampling on devices that
use a static global fallback table instead of configurable multi-rate retry.
Now that the sampling algorithm is improved, this code path no longer performs
any better than the standard probing on affected devices.
Felix Fietkau [Wed, 27 Jan 2021 05:57:33 +0000 (06:57 +0100)]
mac80211: minstrel_ht: significantly redesign the rate probing strategy
The biggest flaw in current minstrel_ht is the fact that it needs way too
many probing packets to be able to quickly find the best rate.
Depending on the wifi hardware and operating mode, this can significantly
reduce throughput when not operating at the highest available data rate.
In order to be able to significantly reduce the amount of rate sampling,
we need a much smarter selection of probing rates.
The new approach introduced by this patch maintains a limited set of
available rates to be tested during a statistics window.
They are split into distinct categories:
- MINSTREL_SAMPLE_TYPE_INC - incremental rate upgrade:
Pick the next rate group and find the first rate that is faster than
the current max. throughput rate
- MINSTREL_SAMPLE_TYPE_JUMP - random testing of higher rates:
Pick a random rate from the next group that is faster than the current
max throughput rate. This allows faster adaptation when the link changes
significantly
- MINSTREL_SAMPLE_TYPE_SLOW - test a rate between max_prob, max_tp2 and
max_tp in order to reduce the gap between them
In order to prioritize sampling, every 6 attempts are split into 3x INC,
2x JUMP, 1x SLOW.
Available rates are checked and refilled on every stats window update.
With this approach, we finally get a very small delta in throughput when
comparing setting the optimal data rate as a fixed rate vs normal rate
control operation.
Felix Fietkau [Wed, 27 Jan 2021 05:57:32 +0000 (06:57 +0100)]
mac80211: minstrel_ht: reduce the need to sample slower rates
In order to more gracefully be able to fall back to lower rates without too
much throughput fluctuations, initialize all untested rates below tested ones
to the maximum probabilty of higher rates.
Usually this leads to untested lower rates getting initialized with a
probability value of 100%, making them better candidates for fallback without
having to rely on random probing
Luca Coelho [Thu, 4 Feb 2021 13:44:39 +0000 (15:44 +0200)]
cfg80211: initialize reg_rule in __freq_reg_info()
Sparse started warning on this function because we can potentially
return an uninitialized value. The reason is that if the caller
passes a min_bw value that is higher then the last value in bws[], we
will not go into the loop and reg_rule will remain initialized. This
cannot happen because the only caller of this function uses either 1
or 20 in min_bw, but the function will be more robust if we
pre-initialize the value.
Colin Ian King [Fri, 5 Feb 2021 17:53:52 +0000 (17:53 +0000)]
mac80211: fix potential overflow when multiplying to u32 integers
The multiplication of the u32 variables tx_time and estimated_retx is
performed using a 32 bit multiplication and the result is stored in
a u64 result. This has a potential u32 overflow issue, so avoid this
by casting tx_time to a u64 to force a 64 bit multiply.
Markus Theil [Sat, 6 Feb 2021 11:51:12 +0000 (12:51 +0100)]
mac80211: enable QoS support for nl80211 ctrl port
This patch unifies sending control port frames
over nl80211 and AF_PACKET sockets a little more.
Before this patch, EAPOL frames got QoS prioritization
only when using AF_PACKET sockets.
__ieee80211_select_queue only selects a QoS-enabled queue
for control port frames, when the control port protocol
is set correctly on the skb. For the AF_PACKET path this
works, but the nl80211 path used ETH_P_802_3.
Another check for injected frames in wme.c then prevented
the QoS TID to be copied in the frame.
In order to fix this, get rid of the frame injection marking
for nl80211 ctrl port and set the correct ethernet protocol.
Please note:
An erlier version of this path tried to prevent
frame aggregation for control port frames in order to speed up
the initial connection setup a little. This seemed to cause
issues on my older Intel dvm-based hardware, and was therefore
removed again. Future commits which try to reintroduce this
have to check carefully how hw behaves with aggregated and
non-aggregated traffic for the same TID.
My NIC: Intel(R) Centrino(R) Ultimate-N 6300 AGN, REV=0x74
Matteo Croce [Mon, 8 Feb 2021 11:33:56 +0000 (12:33 +0100)]
cfg80211: remove unused callback
The ieee80211 class registers a callback which actually does nothing.
Given that the callback is optional, and all its accesses are protected
by a NULL check, remove it entirely.
Po-Hao Huang [Tue, 9 Feb 2021 07:07:51 +0000 (15:07 +0800)]
rtw88: replace tx tasklet with work queue
Replace tasklet so we can do tx scheduling in parallel. Since throughput
is delay-sensitive in most cases, we allocate a dedicated, high priority
wq for our needs.
Po-Hao Huang [Tue, 9 Feb 2021 07:07:50 +0000 (15:07 +0800)]
rtw88: add napi support
Use napi to reduce overhead on rx interrupts.
Driver used to interrupt kernel for every Rx packet, this could
affect both system and network performance. NAPI is a mechanism that
uses polling when processing huge amount of traffic, by doing this
the number of interrupts can be decreased.
Network performance can also benefit from this patch. Since TCP
connection is bidirectional and acks are required for every several
packets. These ack packets occupie the PCI bus bandwidth and could
lead to performance degradation.
When napi is used, GRO receive is enabled by default in the mac80211
stack. So mac80211 won't pass every RX TCP packets to the kernel TCP
network stack immediately. Instead an aggregated large length TCP packet
will be delivered.
This reduces the tx acks sent and gains rx performance. After the patch,
the Rx throughput increases about 25Mbps in 11ac.
Po-Hao Huang [Tue, 9 Feb 2021 07:07:49 +0000 (15:07 +0800)]
rtw88: add rts condition
Since we set the IEEE80211_HW_HAS_RATE_CONTROL flag, so use_rts in
ieee80211_tx_info will never be set in the ieee80211_xmit_fast path.
Add length check for skb to decide whether rts is needed.
Po-Hao Huang [Tue, 9 Feb 2021 07:07:48 +0000 (15:07 +0800)]
rtw88: add dynamic rrsr configuration
Register rrsr determines the response rate we send.
In field tests, using rate higher than current tx rate could lead
to difficulty for the receiving end to receive management/control
frames. Calculate current modulation level by tx rate then cross out
rate higher than those.
Luca Coelho [Thu, 11 Feb 2021 20:30:55 +0000 (22:30 +0200)]
iwlwifi: remove incorrect comment in pnvm
We use this driver as a backport that also runs on older kernels (as
part of the backports project). So we use some checks to backport or
prevent code from compiling in incompatible kernel version.
When I took one of the PNVM patches from the backport, I accidentally
left the comment that a certain part of the code doesn't work in older
kernels. This obviously should never be valid for the mainline.
Remove this comment.
David S. Miller [Fri, 12 Feb 2021 03:08:07 +0000 (19:08 -0800)]
Merge branch 'sock-rx-qmap'
Tariq Toukan says:
====================
Compile-flag for sock RX queue mapping
Socket's RX queue mapping logic is useful also for non-XPS use cases.
This series breaks the dependency between the two, introducing a new
kernel config flag SOCK_RX_QUEUE_MAPPING.
Here we select this new kernel flag from TLS_DEVICE, as well as XPS.
====================
Tariq Toukan [Thu, 11 Feb 2021 11:35:52 +0000 (13:35 +0200)]
net/tls: Select SOCK_RX_QUEUE_MAPPING from TLS_DEVICE
Compile-in the socket RX queue mapping field and logic when TLS_DEVICE
is enabled. This allows device drivers to pick the recorded socket's
RX queue and use it for streams distribution.
Tariq Toukan [Thu, 11 Feb 2021 11:35:51 +0000 (13:35 +0200)]
net/sock: Add kernel config SOCK_RX_QUEUE_MAPPING
Use a new config SOCK_RX_QUEUE_MAPPING to compile-in the socket
RX queue field and logic, instead of the XPS config.
This breaks dependency in XPS, and allows selecting it from non-XPS
use cases, as we do in the next patch.
In addition, use the new flag to wrap the logic in sk_rx_queue_get()
and protect access to the sk_rx_queue_mapping field, while keeping
the function exposed unconditionally, just like sk_rx_queue_set()
and sk_rx_queue_clear().
Arjun Roy [Thu, 11 Feb 2021 21:21:07 +0000 (13:21 -0800)]
tcp: Sanitize CMSG flags and reserved args in tcp_zerocopy_receive.
Explicitly define reserved field and require it and any subsequent
fields to be zero-valued for now. Additionally, limit the valid CMSG
flags that tcp_zerocopy_receive accepts.
Cong Wang [Thu, 11 Feb 2021 19:34:10 +0000 (11:34 -0800)]
net: fix dev_ifsioc_locked() race condition
dev_ifsioc_locked() is called with only RCU read lock, so when
there is a parallel writer changing the mac address, it could
get a partially updated mac address, as shown below:
Close this race condition by guarding them with a RW semaphore,
like netdev_get_name(). We can not use seqlock here as it does not
allow blocking. The writers already take RTNL anyway, so this does
not affect the slow path. To avoid bothering existing
dev_set_mac_address() callers in drivers, introduce a new wrapper
just for user-facing callers on ioctl and rtnetlink paths.
Note, bonding also changes slave mac addresses but that requires
a separate patch due to the complexity of bonding code.
The condition should be skipped if CPU ID equal to nthreads.
The patch doesn't fix any actual issue since
nthreads = min_t(unsigned int, num_present_cpus(), MVPP2_MAX_THREADS).
On all current Armada platforms, the number of CPU's is
less than MVPP2_MAX_THREADS.
Fixes: e531f76757eb ("net: mvpp2: handle cases where more CPUs are available than s/w threads") Reported-by: Russell King <[email protected]> Signed-off-by: Stefan Chulski <[email protected]> Reviewed-by: Russell King <[email protected]> Signed-off-by: David S. Miller <[email protected]>
This series adds switchdev support for AM65 CPSW NUSS driver to support
multi port CPSW present on J721e and AM64 SoCs.
It adds devlink hook to switch b/w switch mode and multi mac mode.
v2:
Rebased on latest net-next
Update patch 1/4 with rationale for using devlink
====================
J721e, J7200 and AM64 have multi port switches which can work in multi
mac mode and in switch mode. Add documentation explaining how to use
different modes.
J721e, J7200 and AM64 have multi port switches which can work in multi
mac mode and in switch mode. Add support for configuring this HW in
switch mode using devlink and switchdev notifiers.
Support is similar to existing CPSW switchdev implementation of TI's 32 bit
platform like AM33/AM43/AM57.
To enable switch mode:
devlink dev param set platform/8000000.ethernet name switch_mode value true cmode runtime
All configuration is implemented via switchdev API and notifiers.
Supported:
- SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS
- SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS
- SWITCHDEV_ATTR_ID_PORT_STP_STATE
- SWITCHDEV_OBJ_ID_PORT_VLAN
- SWITCHDEV_OBJ_ID_PORT_MDB
- SWITCHDEV_OBJ_ID_HOST_MDB
AM65 NUSS ethernet switch on K3 devices can be configured to work either
in independent mac mode where each port acts as independent network
interface (multi mac) or switch mode.
Add devlink hooks to provide a way to switch b/w these modes.
Rationale to use devlink instead of defaulting to bridge mode is that
SoC use cases require to support multiple independent MAC ports with no
switching so that users can use software bridges with multi-mac
configuration (e.g: to support LAG, HSR/PRP, etc). Also, switching
between multi mac and switch mode requires significant Port and ALE
reconfiguration, therefore is easier to be made as part of mode change
devlink hooks. It also allows to keep user interface similar to what
was implemented for the previous generation of TI CPSW IP
(on AM33/AM43/AM57 SoCs).
V2 of my BCM4908 Ethernet patchset was applied to the net-next.git and
it was later that is received some extra reviews. I'm sending patches
that handle pointed out issues.
David: earler I missed that V2 was applied and I sent V3 and V4 of my
inital patchset. Sorry for that. I think it's the best to ignore V3 and
V4 I sent and proceed with this fixes patchset instead.
====================
Here's the main bluetooth-next pull request for 5.12:
- Add support for advertising monitor offliading using Microsoft
vendor extensions
- Add firmware download support for MediaTek MT7921U USB devices
- Suspend-related fixes for Qualcomm devices
- Add support for Intel GarfieldPeak controller
- Various other smaller fixes & cleanups
Please let me know if there are any issues pulling. Thanks.
====================
David S. Miller [Thu, 11 Feb 2021 22:55:04 +0000 (14:55 -0800)]
Merge branch 'marvell-cn10k'
Geetha sowjanya says:
====================
Add Marvell CN10K support
The current admin function (AF) driver and the netdev driver supports
OcteonTx2 silicon variants. The same OcteonTx2's
Resource Virtualization Unit (RVU) is carried forward to the next-gen
silicon ie OcteonTx3, with some changes and feature enhancements.
This patch set adds support for OcteonTx3 (CN10K) silicon and gets
the drivers to the same level as OcteonTx2. No new OcteonTx3 specific
features are added.
Changes cover below HW level differences
- PCIe BAR address changes wrt shared mailbox memory region
- Receive buffer freeing to HW
- Transmit packet's descriptor submission to HW
- Programmable HW interface identifiers (channels)
- Increased MTU support
- A Serdes MAC block (RPM) configuration
v5-v6
Rebased on top of latest net-next branch.
v4-v5
Fixed sparse warnings.
v3-v4
Fixed compiler warnings.
v2-v3
Reposting as a single thread.
Rebased on top latest net-next branch.
Hariprasad Kelam [Thu, 11 Feb 2021 15:58:34 +0000 (21:28 +0530)]
octeontx2-af: cn10k: MAC internal loopback support
MAC on CN10K silicon support loopback for selftest or debug purposes.
This patch does necessary configuration to loopback packets upon receiving
request from LMAC mapped RVU PF's netdev via mailbox.
Also MAC (CGX) on OcteonTx2 silicon variants and MAC (RPM) on
OcteonTx3 CN10K are different and loopback needs to be configured
differently. Upper layer interface between RVU AF and PF netdev is
kept same. Based on silicon variant appropriate fn() pointer is
called to config the MAC.
Hariprasad Kelam [Thu, 11 Feb 2021 15:58:33 +0000 (21:28 +0530)]
octeontx2-af: cn10k: Add RPM Rx/Tx stats support
RPM supports below list of counters as an extension to existing counters
* class based flow control pause frames
* vlan/jabber/fragmented packets
* fcs/alignment/oversized error packets
This patch adds support to display supported RPM counters via debugfs
and define new mbox rpm_stats to read all support counters.
Rakesh Babu [Thu, 11 Feb 2021 15:58:32 +0000 (21:28 +0530)]
octeontx2-af: cn10k: Add RPM LMAC pause frame support
Flow control configuration is different for CGX(Octeontx2)
and RPM(CN10K) functional blocks. This patch adds the necessary
changes for RPM to support 802.3 pause frames configuration on
cn10k platforms.
Hariprasad Kelam [Thu, 11 Feb 2021 15:58:31 +0000 (21:28 +0530)]
octeontx2-pf: cn10k: Get max mtu supported from admin function
CN10K supports max MTU of 16K on LMAC links and 64k on LBK
links and Octeontx2 silicon supports 9K mtu on both links.
Get the same from nix_get_hw_info mbox message in netdev probe.
This patch also calculates receive buffer size required based
on the MTU set.
Hariprasad Kelam [Thu, 11 Feb 2021 15:58:30 +0000 (21:28 +0530)]
octeontx2-af: cn10K: Add MTU configuration
OcteonTx3 CN10K silicon supports bigger MTU when compared
to 9216 MTU supported by OcteonTx2 silicon variants. Lookback
interface supports upto 64K and RPM LMAC interfaces support
upto 16K.
This patch does the necessary configuration and adds support
for PF/VF drivers to retrieve max packet size supported via mbox
This patch also configures tx link credit by considering supported
fifo size and max packet length for Octeontx3 silicon.
This patch also removes platform specific name from the driver name.
octeontx2-af: cn10k: Add support for programmable channels
NIX uses unique channel numbers to identify the packet sources/sinks
like CGX,LBK and SDP. The channel numbers assigned to each block are
hardwired in CN9xxx silicon.
The fixed channel numbers in CN9xxx are:
0x0 | a << 8 | b - LBK(0..3)_CH(0..63)
0x0 | a << 8 - Reserved
0x700 | a - SDP_CH(0..255)
0x800 | a << 8 | b << 4 | c - CGX(0..7)_LMAC(0..3)_CH(0..15)
All the channels in the above fixed enumerator(with maximum
number of blocks) are not required since some chips
have less number of blocks.
For CN10K silicon the channel numbers need to be programmed by
software in each block with the base channel number and range of
channels. This patch calculates and assigns the channel numbers
to efficiently distribute the channel number range(0-4095) among
all the blocks. The assignment is made based on the actual number of
blocks present and also contiguously leaving no holes.
The channel numbers remaining after the math are used as new CPT
replay channels present in CN10K. Also since channel numbers are
not fixed the transmit channel link number needed by AF consumers
is calculated by AF and sent along with nix_lf_alloc mailbox response.
Hariprasad Kelam [Thu, 11 Feb 2021 15:58:28 +0000 (21:28 +0530)]
octeontx2-af: cn10k: Add RPM MAC support
OcteonTx2's next gen platform the CN10K has RPM MAC which has a
different serdes when compared to CGX MAC. Though the underlying
HW is different, the CSR interface has been designed largely inline
with CGX MAC, with few exceptions though. So we are using the same
CGX driver for RPM MAC as well and will have a different set of APIs
for RPM where ever necessary.
This patch adds initial support for CN10K's RPM MAC i.e. the driver
registration, communication with firmware etc. For communication with
firmware, RPM provides a different IRQ when compared to CGX.
The CGX and RPM blocks support different features. Currently few
features like ptp, flowcontrol and higig are not supported by RPM. This
patch adds new mailbox message "CGX_FEATURES_GET" to get the list of
features supported by underlying MAC.
RPM has different implementations for RX/TX stats. Unlike CGX,
bar offset of stat registers are different. This patch adds
support to access the same and dump the values in debugfs.
Geetha sowjanya [Thu, 11 Feb 2021 15:58:27 +0000 (21:28 +0530)]
octeontx2-pf: cn10k: Use LMTST lines for NPA/NIX operations
This patch adds support to use new LMTST lines for NPA batch free
and burst SQE flush. Adds new dev_hw_ops structure to hold platform
specific functions and create new files cn10k.c and cn10k.h.
Geetha sowjanya [Thu, 11 Feb 2021 15:58:26 +0000 (21:28 +0530)]
octeontx2-pf: cn10k: Map LMTST region
On CN10K platform transmit/receive buffer alloc and free from/to hardware
had changed to support burst operation. Whereas pervious silicon's only
support single buffer free at a time.
To Support the same firmware allocates a DRAM region for each PF/VF for
storing LMTLINES. These LMTLINES are used for NPA batch free and for
flushing SQE to the hardware.
PF/VF LMTST region is accessed via BAR4. PFs LMTST region is followed
by its VFs mbox memory. The size of region varies from 2KB to 256KB based
on number of LMTLINES configured.
This patch adds support for
- Mapping PF/VF LMTST region.
- Reserves 0-71 (RX + TX + XDP) LMTST lines for NPA batch
free operation.
- Reserves 72-512 LMTST lines for NIX SQE flush.
Geetha sowjanya [Thu, 11 Feb 2021 15:58:25 +0000 (21:28 +0530)]
octeontx2-pf: cn10k: Initialise NIX context
On CN10K platform NIX RQ and SQ context structure got changed.
This patch uses new mbox message "NIX_CN10K_AQ_ENQ" for NIX
context initialization on CN10K platform.
This patch also updates the nix_rx_parse_s and nix_sqe_sg_s
structures to add packet steering bit feilds.
Geetha sowjanya [Thu, 11 Feb 2021 15:58:24 +0000 (21:28 +0530)]
octeontx2-af: cn10k: Update NIX and NPA context in debugfs
On CN10K platform NPA and NIX context structure bit fields
had changed to support new features like bandwidth steering etc.
This patch dumps approprate context for CN10K platform.
NIX hardware context structure got changed to accommodate new
features like bandwidth steering, L3/L4 outer/inner checksum
enable/disable etc., on CN10K platform.
This patch defines new mbox message NIX_CN10K_AQ_INST for new
NIX context initialization.
This patch also updates the NPA context structures to accommodate
bit field changes made for CN10K platform.
This patch also removes Big endian bit fields from existing
structures as its support got deprecated in current and upcoming silicons.
Firmware allocates memory regions for PFs and VFs in DRAM.
The PFs memory region is used for AF-PF and PF-VF mailbox.
This mbox facilitate communication between AF-PF and PF-VF.
On CN10K platform:
The DRAM region allocated to PF is enumerated as PF BAR4 memory.
PF BAR4 contains AF-PF mbox region followed by its VFs mbox region.
AF-PF mbox region base address is configured at RVU_AF_PFX_BAR4_ADDR
PF-VF mailbox base address is configured at
RVU_PF(x)_VF_MBOX_ADDR = RVU_AF_PF()_BAR4_ADDR+64KB. PF access its
mbox region via BAR4, whereas VF accesses PF-VF DRAM mailboxes via
BAR2 indirect access.
On CN9XX platform:
Mailbox region in DRAM is divided into two parts AF-PF mbox region and
PF-VF mbox region i.e all PFs mbox region is contiguous similarly all
VFs.
The base address of the AF-PF mbox region is configured at
RVU_AF_PF_BAR4_ADDR.
AF-PF1 mbox address can be calculated as RVU_AF_PF_BAR4_ADDR * mbox
size.
The base address of PF-VF mbox region for each PF is configure at
RVU_AF_PF(0..15)_VF_BAR4_ADDR.PF access its mbox region via BAR4 and its
VF mbox regions from RVU_PF_VF_BAR4_ADDR register, whereas VF access its
mbox region via BAR4.
This patch changes mbox initialization to support both CN9XX and CN10K
platform.
The patch also adds new hw_cap flag to setting hw features like TSO etc
and removes platform specific name from the PF/VF driver name to make it
appropriate for all supported platforms
This patch also removes platform specific name from the PF/VF driver name
to make it appropriate for all supported platforms
octeontx2-af: cn10k: Add mbox support for CN10K platform
Firmware allocates memory regions for PFs and VFs in DRAM.
The PFs memory region is used for AF-PF and PF-VF mailbox.
This mbox facilitates communication between AF-PF and PF-VF.
On CN10K platform:
The DRAM region allocated to PF is enumerated as PF BAR4 memory.
PF BAR4 contains AF-PF mbox region followed by its VFs mbox region.
AF-PF mbox region base address is configured at RVU_AF_PFX_BAR4_ADDR
PF-VF mailbox base address is configured at
RVU_PF(x)_VF_MBOX_ADDR = RVU_AF_PF()_BAR4_ADDR+64KB. PF access its
mbox region via BAR4, whereas VF accesses PF-VF DRAM mailboxes via
BAR2 indirect access.
On CN9XX platform:
Mailbox region in DRAM is divided into two parts AF-PF mbox region and
PF-VF mbox region i.e all PFs mbox region is contiguous similarly all
VFs.
The base address of the AF-PF mbox region is configured at
RVU_AF_PF_BAR4_ADDR.
AF-PF1 mbox address can be calculated as RVU_AF_PF_BAR4_ADDR * mbox
size.
The base address of PF-VF mbox region for each PF is configure at
RVU_AF_PF(0..15)_VF_BAR4_ADDR.PF access its mbox region via BAR4 and its
VF mbox regions from RVU_PF_VF_BAR4_ADDR register, whereas VF access its
mbox region via BAR4.
This patch changes mbox initialization to support both CN9XX and CN10K
platform.
This patch also adds CN10K PTP subsystem and device IDs to ptp
driver id table.
David S. Miller [Thu, 11 Feb 2021 22:50:24 +0000 (14:50 -0800)]
Merge branch 'mvpp2-tx-flow-control'
Stefan Chulski says:
====================
net: mvpp2: Add TX Flow Control support
Armada hardware has a pause generation mechanism in GOP (MAC).
The GOP generate flow control frames based on an indication programmed in Ports Control 0 Register. There is a bit per port.
However assertion of the PortX Pause bits in the ports control 0 register only sends a one time pause.
To complement the function the GOP has a mechanism to periodically send pause control messages based on periodic counters.
This mechanism ensures that the pause is effective as long as the Appropriate PortX Pause is asserted.
Problem is that Packet Processor that actually can drop packets due to lack of resources not connected to the GOP flow control generation mechanism.
To solve this issue Armada has firmware running on CM3 CPU dedicated for Flow Control support.
Firmware monitors Packet Processor resources and asserts XON/XOFF by writing to Ports Control 0 Register.
MSS shared SRAM memory used to communicate between CM3 firmware and PP2 driver.
During init PP2 driver informs firmware about used BM pools, RXQs, congestion and depletion thresholds.
The pause frames are generated whenever congestion or depletion in resources is detected.
The back pressure is stopped when the resource reaches a sufficient level.
So the congestion/depletion and sufficient level implement a hysteresis that reduces the XON/XOFF toggle frequency.
Packet Processor v23 hardware introduces support for RX FIFO fill level monitor.
Patch "add PPv23 version definition" to differ between v23 and v22 hardware.
Patch "add TX FC firmware check" verifies that CM3 firmware supports Flow Control monitoring.
v8 --> v9
- Replace generic pool allocation with devm_ioremap_resource
v7 --> v8
- Reorder "always compare hw-version vs MVPP21" and "add PPv23 version definition" commits
- Typo fixes
- Remove condition fix from "add RXQ flow control configurations"
v6 --> v7
- Reduce patch set from 18 to 15 patches
- Documentation change combined into a single patch
- RXQ and BM size change combined into a single patch
- Ring size change check moved into "add RXQ flow control configurations" commit
v5 --> v6
- No change
v4 --> v5
- Add missed Signed-off
- Fix warnings in patches 3 and 12
- Add revision requirement to warning message
- Move mss_spinlock into RXQ flow control configurations patch
- Improve FCA RXQ non occupied descriptor threshold commit message
v3 --> v4
- Remove RFC tag
v2 --> v3
- Remove inline functions
- Add PPv2.3 description into marvell-pp2.txt
- Improve mvpp2_interrupts_mask/unmask procedure
- Improve FC enable/disable procedure
- Add priv->sram_pool check
- Remove gen_pool_destroy call
- Reduce Flow Control timer to x100 faster
v1 --> v2
- Add memory requirements information
- Add EPROBE_DEFER if of_gen_pool_get return NULL
- Move Flow control configuration to mvpp2_mac_link_up callback
====================
projects / linux.git / log