| Age | Commit message (Collapse) | Author |
|---|
|
When the XDP program returns an invalid action, we should free the RX
buffer.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
It is possible for one CPU to perform TX hashing (see netdev_pick_tx)
between the 8 ENETC TX rings, and the TX hashing to select TX queue 1.
At the same time, it is possible for the other CPU to already use TX
ring 1 for XDP (either XDP_TX or XDP_REDIRECT). Since there is no mutual
exclusion between XDP and the network stack, we run into an issue
because the ENETC TX procedure is not reentrant.
The obvious approach would be to just make XDP take the lock of the
network stack's TX queue corresponding to the ring it's about to enqueue
in.
For XDP_REDIRECT, this is quite straightforward, a lock at the beginning
and end of enetc_xdp_xmit() should do the trick.
But for XDP_TX, it's a bit more complicated. For one, we do TX batching
all by ourselves for frames with the XDP_TX verdict. This is something
we would like to keep the way it is, for performance reasons. But
batching means that the network stack's lock should be kept from the
first enqueued XDP_TX frame and until we ring the doorbell. That is
mostly fine, except for cases when in the same NAPI loop we have mixed
XDP_TX and XDP_REDIRECT frames. So if enetc_xdp_xmit() gets called while
we are holding the lock from the RX NAPI, then bam, deadlock. The naive
answer could be 'just flush the XDP_TX frames first, then release the
network stack's TX queue lock, then call xdp_do_flush_map()'. But even
xdp_do_redirect() is capable of flushing the batched XDP_REDIRECT
frames, so unless we unlock/relock the TX queue around xdp_do_redirect(),
there simply isn't any clean way to protect XDP_TX from concurrent
network stack .ndo_start_xmit() on another CPU.
So we need to take a different approach, and that is to reserve two
rings for the sole use of XDP. We leave TX rings
0..ndev->real_num_tx_queues-1 to be handled by the network stack, and we
pick them from the end of the priv->tx_ring array.
We make an effort to keep the mapping done by enetc_alloc_msix() which
decides which CPU handles the TX completions of which TX ring in its
NAPI poll. So the XDP TX ring of CPU 0 is handled by TX ring 6, and the
XDP TX ring of CPU 1 is handled by TX ring 7.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Now that commit d6a2829e82cf ("net: enetc: increase RX ring default
size") has increased the RX ring size, it is quite easy to congest the
TX rings when the traffic is predominantly XDP_TX, as the RX ring is
quite a bit larger than the TX one.
Since we bit the bullet and did the expensive thing already (larger RX
rings consume more memory pages), it seems quite foolish to keep the TX
rings small. So make them equally sized with TX.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
xdp_do_redirect already contains:
-> dev_map_enqueue
-> __xdp_enqueue
-> bq_enqueue
-> bq_xmit_all // if we have more than 16 frames
So the logic from enetc will never be hit, because ENETC_DEFAULT_TX_WORK
is 128.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When the code path below fails:
enetc_clean_rx_ring_xdp // XDP_PASS
-> enetc_build_skb
-> enetc_map_rx_buff_to_skb
-> build_skb
enetc_clean_rx_ring_xdp will 'break', but that 'break' instruction isn't
strong enough to actually break the NAPI poll loop, just the switch/case
statement for XDP actions. So we increment rx_frm_cnt and go to the next
frames minding our own business.
Instead let's do what the skb NAPI poll function does, and break the
loop now, waiting for the memory pressure to go away. Otherwise the next
calls to build_skb() are likely to fail too.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When receiving a frame with errors, currently we do nothing with it (we
don't construct an skb or an xdp_buff), we just exit the NAPI poll loop.
Let's put the buffer back into the RX ring (similar to XDP_DROP).
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
enetc_put_xdp_buff has nothing to do with XDP, frankly, it is just a
helper to populate the recycle end of the shadow RX BD ring
(next_to_alloc) with a given buffer.
On the other hand, enetc_put_rx_buff plays more tricks than its name
would suggest.
So let's rename enetc_put_rx_buff into enetc_flip_rx_buff to reflect the
half-page buffer reuse tricks that it employs, and enetc_put_xdp_buff
into enetc_put_rx_buff which suggests a more garden-variety operation.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Later in enetc_clean_tx_ring we have:
/* Scrub the swbd here so we don't have to do that
* when we reuse it during xmit
*/
memset(tx_swbd, 0, sizeof(*tx_swbd));
So these assignments are unnecessary.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue
Tony Nguyen says:
====================
1GbE Intel Wired LAN Driver Updates 2021-04-16
This series contains updates to igb and igc drivers.
Ederson adjusts Tx buffer distributions in Qav mode to improve
TSN-aware traffic for igb. He also enable PPS support and auxiliary PHC
functions for igc.
Grzegorz checks that the MTA register was properly written and
retries if not for igb.
Sasha adds reporting of EEE low power idle counters to ethtool and fixes
a return value being overwritten through looping for igc.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add support for PHY/MAC/Ctrl/RMON stats.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Most of the names seem to strongly correlate with names from
the standard and RFC. Whether ..+good_frames are indeed Frames..OK
I'm the least sure of.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
mlxsw has nicely grouped stats, add support for standard uAPI.
I'm guessing the register access part. Compile tested only.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2021-04-16
This patchset introduces updates to mlx5e netdev driver.
1) Tariq refactors TLS offloads and adds resiliency against RX resync
failures
2) Maxim reduces code duplications by unifying channels reset flow
regardless if channels are closed or open
3) Aya Enhances TX/RX health reporters diagnostics to expose the
internal clock time-stamping format
4) Moshe adds support for ethtool extended link state, to show the reason
for link down
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Gianfar used to enable all 8 Rx queues (DMA rings) per
ethernet device, even though the controller can only
support 2 interrupt lines at most. This meant that
multiple Rx queues would have to be grouped per NAPI poll
routine, and the CPU would have to split the budget and
service them in a round robin manner. The overhead of
this scheme proved to outweight the potential benefits.
The alternative was to introduce the "Single Queue" polling
mode, supporting one Rx queue per NAPI, which became the
default packet processing option and helped improve the
performance of the driver.
MQ_POLLING also relies on undocumeted device tree properties
to specify how to map the 8 Rx and Tx queues to a given
interrupt line (aka "interrupt group"). Using module parameters
to enable this mode wasn't an option either. Long story short,
MQ_POLLING became obsolete, now it is just dead code, and no
one asked for it so far.
For the Tx queues, multi-queue support (more than 1 Tx queue
per CPU) could be revisited by adding tc MQPRIO support, but
again, one has to consider that there are only 2 interrupt lines.
So the NAPI poll routine would have to service multiple Tx rings.
Signed-off-by: Claudiu Manoil <claudiu.manoil@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The recent patch that tied enabling of veth NAPI to the GRO flag also has
the nice side effect that a veth device can be the target of an
XDP_REDIRECT without an XDP program needing to be loaded on the peer
device. However, the patch adding this extra NAPI mode didn't actually
change the check in veth_xdp_xmit() to also look at the new NAPI pointer,
so let's fix that.
Fixes: 6788fa154546 ("veth: allow enabling NAPI even without XDP")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
IPA initialization includes loading some firmware. This step is
done either by the modem or by the AP under Trust Zone. If the
AP loads firmware, the name of the firmware file is currently
hard-coded ("ipa_fws.mdt").
Add the ability to specify the relative path of the firmware file to
use in a property in the Device Tree IPA node. If the property is
not found (or if any other error occurs attempting to get it), fall
back to using a default relative path.
Use the "old" fixed name as the default. Rename the symbol that
represents this default to emphasize its purpose.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In page_to_skb(), if we have enough tailroom to save skb_shared_info, we
can use build_skb to create skb directly. No need to alloc for
additional space. And it can save a 'frags slot', which is very friendly
to GRO.
Here, if the payload of the received package is too small (less than
GOOD_COPY_LEN), we still choose to copy it directly to the space got by
napi_alloc_skb. So we can reuse these pages.
Testing Machine:
The four queues of the network card are bound to the cpu1.
Test command:
for ((i=0;i<5;++i)); do sockperf tp --ip 192.168.122.64 -m 1000 -t 150& done
The size of the udp package is 1000, so in the case of this patch, there
will always be enough tailroom to use build_skb. The sent udp packet
will be discarded because there is no port to receive it. The irqsoftd
of the machine is 100%, we observe the received quantity displayed by
sar -n DEV 1:
no build_skb: 956864.00 rxpck/s
build_skb: 1158465.00 rxpck/s
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Suggested-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The MHI WWWAN control driver allows MHI QCOM-based modems to expose
different modem control protocols/ports via the WWAN framework, so that
userspace modem tools or daemon (e.g. ModemManager) can control WWAN
config and state (APN config, SMS, provider selection...). A QCOM-based
modem can expose one or several of the following protocols:
- AT: Well known AT commands interactive protocol (microcom, minicom...)
- MBIM: Mobile Broadband Interface Model (libmbim, mbimcli)
- QMI: QCOM MSM/Modem Interface (libqmi, qmicli)
- QCDM: QCOM Modem diagnostic interface (libqcdm)
- FIREHOSE: XML-based protocol for Modem firmware management
(qmi-firmware-update)
Note that this patch is mostly a rework of the earlier MHI UCI
tentative that was a generic interface for accessing MHI bus from
userspace. As suggested, this new version is WWAN specific and is
dedicated to only expose channels used for controlling a modem, and
for which related opensource userpace support exist.
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This change introduces initial support for a WWAN framework. Given the
complexity and heterogeneity of existing WWAN hardwares and interfaces,
there is no strict definition of what a WWAN device is and how it should
be represented. It's often a collection of multiple devices that perform
the global WWAN feature (netdev, tty, chardev, etc).
One usual way to expose modem controls and configuration is via high
level protocols such as the well known AT command protocol, MBIM or
QMI. The USB modems started to expose them as character devices, and
user daemons such as ModemManager learnt to use them.
This initial version adds the concept of WWAN port, which is a logical
pipe to a modem control protocol. The protocols are rawly exposed to
user via character device, allowing straigthforward support in existing
tools (ModemManager, ofono...). The WWAN core takes care of the generic
part, including character device management, and relies on port driver
operations to receive/submit protocol data.
Since the different devices exposing protocols for a same WWAN hardware
do not necessarily know about each others (e.g. two different USB
interfaces, PCI/MHI channel devices...) and can be created/removed in
different orders, the WWAN core ensures that all WAN ports contributing
to the 'whole' WWAN feature are grouped under the same virtual WWAN
device, relying on the provided parent device (e.g. mhi controller,
USB device). It's a 'trick' I copied from Johannes's earlier WWAN
subsystem proposal.
This initial version is purposely minimalist, it's essentially moving
the generic part of the previously proposed mhi_wwan_ctrl driver inside
a common WWAN framework, but the implementation is open and flexible
enough to allow extension for further drivers.
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add parser entries for different IPv4 IHL values.
Each entry will set the L4 header offset according to the IPv4 IHL field.
L3 header offset will set during the parsing of the IPv4 protocol.
Because of missed parser support for IP header length > 20, RX IPv4 checksum HW offload fails
and skb->ip_summed set to CHECKSUM_NONE(checksum done by Network stack).
This patch adds RX IPv4 checksum HW offload capability for frames with IP header length > 20.
v1 --> v2
- Improve commit message.
Suggested-by: Dana Vardi <danat@marvell.com>
Signed-off-by: Stefan Chulski <stefanc@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The vendor mode is not always at config #1, so it is necessary to
set the correct configuration number.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Support new firmware type and method for RTL8156 series.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Support RTL8153C, RTL8153D, RTL8156A, and RTL8156B. The RTL8156A
and RTL8156B are the 2.5G ethernet.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The different chips may have different requests when changing mtu.
Therefore, add a new help function of rtl_ops to change mtu. Besides,
reset the tx/rx after changing mtu.
Additionally, add mtu_to_size() and size_to_mtu() macros to simplify
the code.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use bits operations to record and check the firmware.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Set the maximum inter frame gap time (144ns) for speed 10M/half and
100M/half. It improves the performance for those speeds. And, there
is no effect for the other speeds.
For 10M/half and 100M/half, the fast inter frame gap time let the
device couldn't use the feature of the aggregation effectively,
because the transfer would be completed fastly. Therefore, use the
maximum value to improve the effect of the aggregation. However, you
may not feel the improvement for fast CPUs, because they compensate
for the effect of the aggregation.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The intention is for the loop to timeout if the body does not succeed.
The current logic calls time_is_before_jiffies(timeout) which is false
until after the timeout, so the loop body never executes.
Fix by using readl_poll_timeout as a more standard and less error-prone
solution.
Fixes: ba37b7caf1ed ("net: ethernet: mtk_eth_soc: add support for initializing the PPE")
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
Cc: Felix Fietkau <nbd@nbd.name>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Tx queue cleanup happens in interrupt handler on same core as rx queue
processing. Both can take considerable amount of processing in high
packet-per-second scenarios.
Sending big amounts of packets can stall the rx processing which is
unfair and also can lead to out-of-memory condition since
__dev_kfree_skb_irq queues the skbs for later kfree in softirq which
is not allowed to happen with heavy load in interrupt handler.
This puts tx cleanup in its own napi and enables threaded napi to
allow the rx/tx queue processing to happen on different cores.
The ability to sustain equal amounts of tx/rx traffic increased:
from 280Kpps to 1130Kpps on Threadripper 3960X with upcoming
Mikrotik 10/25G NIC,
from 520Kpps to 850Kpps on Intel i3-3320 with Mikrotik RB44Ge adapter.
Signed-off-by: Gatis Peisenieks <gatis@mikrotik.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
As explained in bugfix commit 6ab4c3117aec ("net: bridge: don't notify
switchdev for local FDB addresses") as well as in this discussion:
https://lore.kernel.org/netdev/20210117193009.io3nungdwuzmo5f7@skbuf/
the switchdev notifiers for FDB entries managed to have a zero-day bug,
which was that drivers would not know what to do with local FDB entries,
because they were not told that they are local. The bug fix was to
simply not notify them of those addresses.
Let us now add the 'is_local' bit to bridge FDB entries, and make all
drivers ignore these entries by their own choice.
Co-developed-by: Tobias Waldekranz <tobias@waldekranz.com>
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Expose EEE Tx and Rx low power idle counters via ethtool
A EEE TX or RX LPI event occurs when the transmitter or the receiver
enters EEE (IEEE802.3az) LPI state.
ethtool --statistics <iface>
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
drivers/net/ethernet/intel/igc/igc_i225.c:235 igc_write_nvm_srwr()
warn: loop overwrites return value 'ret_val'
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
The i225 device offers a number of special PTP Hardware Clock features on
the Software Defined Pins (SDPs) - much like i210, which is used as
inspiration for this patch. It enables two possible functions, namely
time stamping external events and periodic output signals.
The assignment of PHC functions to the four SDP can be freely chosen by
the user.
For the external events time stamping, when the SDP (configured as input
by user) level changes, an interrupt is generated and the kernel
Precision Time Protocol (PTP) is informed.
For the periodic output signals, the i225 is configured to generate them
(so the SDP level will change periodically) and the driver also has to
keep updating the time of the next level change. However, this work is
not necessary for some frequencies as the i225 takes care of them
(namely, anything with a half-cycle of 500ms, 250ms, 125ms or < 70ms).
While i225 allows up to four timers to be used to source the time used
on the external events or output signals, this patch uses only one of
those timers. Main reason is to keep it simple, as it's not clear how
these extra timers would be exposed to users. Note that currently a NIC
can expose a single PTP device.
Signed-off-by: Ederson de Souza <ederson.desouza@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
The i225 device can produce one interrupt on the full second, much
like i210 - from where this patch is inspired.
This patch sets up the full second interruption on the i225 and when
receiving it, it sends a PPS event to PTP (Precision Time Protocol)
kernel subsystem.
The PTP subsystem exposes the PPS events via ioctl and sysfs, and one
can use the `testptp` tool (tools/testing/selftests/ptp) to check that
the events are being generated.
Signed-off-by: Ederson de Souza <ederson.desouza@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Add new function which checks MTA_REGISTER if its filled correctly.
If not then writes again to same register.
There is possibility that i210 and i211 could not accept
MTA_REGISTER settings, specially when you add and remove
many of multicast addresses in short time.
Without this patch there is possibility that multicast settings will be
not always set correctly in hardware.
Signed-off-by: Grzegorz Siwik <grzegorz.siwik@intel.com>
Tested-by: Dave Switzer <david.switzer@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Add ts_format to 'Common Config' section of the TX/RX devlink reporters
diagnostics info. Possible values for ts_format: 'RT' or 'FRC'
which stands for: Real Time and Free Running Counters correspondingly.
Signed-off-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Wrap 1PPS initialization in a helper for a cleaner init flow.
Signed-off-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
In case the interface was set up but cannot establish the link, ethtool
will print more information to help the user troubleshoot the state.
For example, no link due to missing cable:
$ ethtool eth1
...
Link detected: no (No cable)
Beside the general extended state, drivers can pass additional
information about the link state using the sub-state field. For example:
$ ethtool eth1
...
Link detected: no (Autoneg, No partner detected)
The extended state is available only for specific cases, in other cases
ethtool with print only "Link detected: no" as before
Signed-off-by: Moshe Tal <moshet@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
The bulk size is larger than 16K so use kvzalloc().
The bulk bitmask upper size limit is 16K so use kvcalloc().
Signed-off-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
mlx5e_safe_switch_channels accepts new_chs as a parameter and opens new
channels in place, then copying them to priv->channels. It requires all
the callers to allocate space for this temporary storage of the new
channels.
This commit cleans up the API by replacing new_chs with new_params, a
meaningful subset of new_chs to be filled by the caller. The temporary
space for the new channels is allocated inside mlx5e_safe_switch_params
(a new name for mlx5e_safe_switch_channels). An extra copy of params is
made, but since it's control flow, it's not critical.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
This commit extends mlx5e_safe_switch_channels() to support on-the-fly
configuration changes, when the channels are open, but don't need to be
recreated. Such flows exist when a parameter being changed doesn't
affect how the queues are created, or when the queues can be modified
while remaining active.
Before this commit, such flows were handled as special cases on the
caller site. This commit adds this functionality to
mlx5e_safe_switch_channels(), allowing the caller to pass a boolean
indicating whether it's required to recreate the channels or it's
allowed to skip it. The logic of switching channel parameters is now
completely encapsulated into mlx5e_safe_switch_channels().
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
This commit uses new functionality of mlx5e_safe_switch_channels
introduced by the previous commit to reduce the amount of repeating
similar code all over the driver.
It's very common in mlx5e to call mlx5e_safe_switch_channels when the
channels are open, but assign parameters and run hardware commands
manually when the channels are closed.
After the previous commit it's no longer needed to do such manual things
every time, so this commit removes unneeded code and relies on the new
functionality of mlx5e_safe_switch_channels. Some of the places are
refactored and simplified, where more complex flows are used to change
configuration on the fly, without recreating the channels (the logic is
rewritten in a more robust way, with a reset required by default and a
list of exceptions).
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
mlx5e_safe_switch_channels is used to modify channel parameters and/or
hardware configuration in a safe way, so that if anything goes wrong,
everything reverts to the old configuration and remains in a consistent
state.
However, this function only works when the channels are open. When the
caller needs to modify some parameters, first it has to check that the
channels are open, otherwise it has to assign parameters directly, and
such boilerplate repeats in many different places.
This commit prepares for the refactoring of such places by allowing
mlx5e_safe_switch_channels to work when the channels are closed. In this
case it will assign the new parameters and run the preactivate hook.
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
When the TLS logic finds a tcp seq match for a kTLS RX resync
request, it calls the driver callback function mlx5e_ktls_resync()
to handle it and communicate it to the device.
Errors might occur during mlx5e_ktls_resync(), however, they are not
reported to the stack. Moreover, there is no error handling in the
stack for these errors.
In this patch, the driver obtains responsibility on errors handling,
adding queue and retry mechanisms to these resyncs.
We maintain a linked list of resync matches, and try posting them
to the async ICOSQ in the NAPI context.
Only possible failure that demands driver handling is ICOSQ being full.
By relying on the NAPI mechanism, we make sure that the entries in list
will be handled when ICOSQ completions arrive and make some room
available.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
When TLS is supported, WQE ctrl segment of every transmitted packet
is updated with the (possibly empty, for non-TLS packets) TISN field.
Take this one-liner function into the header file and inline it,
to save the overhead of a function call per packet.
While here, remove unused function parameter.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
When TLS is supported and enabled, every transmitted packet is tested
to identify if TLS offload is required.
Take the early-return condition into an inline function, to save
the overhead of a function call for non-TLS packets.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Socket parameter is not used in accel_rule_init(), remove it.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Maintaining an SQ state bit to indicate TLS support
has no real need, a simple and fast test [1] for the SKB is
almost equally good.
[1] !skb->sk || !tls_is_sk_tx_device_offloaded(skb->sk)
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
i210 has a total of 24KB of transmit packet buffer. When in Qav mode,
this buffer is divided into four pieces, one for each Tx queue.
Currently, 8KB are given to each of the two SR queues and 4KB are given
to each of the two SP queues.
However, it was noticed that such distribution can make best effort
traffic (which would usually go to the SP queues when Qav is enabled, as
the SR queues would be used by ETF or CBS qdiscs for TSN-aware traffic)
perform poorly. Using iperf3 to measure, one could see the performance
of best effort traffic drop by nearly a third (from 935Mbps to 578Mbps),
with no TSN traffic competing.
This patch redistributes the 24KB to each queue equally: 6KB each. On
tests, there was no notable performance reduction of best effort traffic
performance when there was no TSN traffic competing.
Below, more details about the data collected:
All experiments were run using the following qdisc setup:
qdisc taprio 100: root refcnt 9 tc 4 map 3 3 3 2 3 0 0 3 3 3 3 3 3 3 3 3
queues offset 0 count 1 offset 1 count 1 offset 2 count 1 offset 3 count 1
clockid TAI base-time 0 cycle-time 10000000 cycle-time-extension 0
index 0 cmd S gatemask 0xf interval 10000000
qdisc etf 8045: parent 100:1 clockid TAI delta 1000000 offload on
deadline_mode off skip_sock_check off
TSN traffic, when enabled, had this characteristics:
Packet size: 1500 bytes
Transmission interval: 125us
----------------------------------
Without this patch:
----------------------------------
- TCP data:
- No TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.35 GBytes 578 Mbits/sec 0
- With TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.07 GBytes 460 Mbits/sec 1
- TCP data limiting iperf3 buffer size to 4K:
- No TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.35 GBytes 579 Mbits/sec 0
- With TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.08 GBytes 462 Mbits/sec 0
- TCP data limiting iperf3 buffer size to 192 bytes (smallest size without
serious performance degradation):
- No TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.34 GBytes 577 Mbits/sec 0
- With TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.07 GBytes 461 Mbits/sec 1
- UDP data at 1000Mbit/sec:
- No TSN traffic:
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-20.00 sec 1.36 GBytes 586 Mbits/sec 0.000 ms 0/1011407 (0%)
- With TSN traffic:
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-20.00 sec 1.05 GBytes 451 Mbits/sec 0.000 ms 0/778672 (0%)
----------------------------------
With this patch:
----------------------------------
- TCP data:
- No TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 2.17 GBytes 932 Mbits/sec 0
- With TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.50 GBytes 646 Mbits/sec 1
- TCP data limiting iperf3 buffer size to 4K:
- No TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 2.17 GBytes 931 Mbits/sec 0
- With TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.50 GBytes 645 Mbits/sec 0
- TCP data limiting iperf3 buffer size to 192 bytes (smallest size without
serious performance degradation):
- No TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 2.17 GBytes 932 Mbits/sec 1
- With TSN traffic:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-20.00 sec 1.50 GBytes 645 Mbits/sec 0
- UDP data at 1000Mbit/sec:
- No TSN traffic:
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-20.00 sec 2.23 GBytes 956 Mbits/sec 0.000 ms 0/1650226 (0%)
- With TSN traffic:
[ ID] Interval Transfer Bitrate Jitter Lost/Total Datagrams
[ 5] 0.00-20.00 sec 1.51 GBytes 649 Mbits/sec 0.000 ms 0/1120264 (0%)
Signed-off-by: Ederson de Souza <ederson.desouza@intel.com>
Tested-by: Tony Brelinski <tonyx.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Report corrected bits.
v2: catch reg access errors (Saeed)
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Report what appears to be the standard block counts:
- 30.5.1.1.17 aFECCorrectedBlocks
- 30.5.1.1.18 aFECUncorrectableBlocks
Don't report the per-lane symbol counts, if those really
count symbols they are not what the standard calls for
(even if symbols seem like the most useful thing to count.)
Fingers crossed that fec_corrected_errors is not in symbols.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
