| Age | Commit message (Collapse) | Author |
|---|
|
The decision whether to report serdes statistics currently depends on
the cached C_Mode value for the port, read at probe time or updated by
configuration. However, port 4 can be in "automedia" mode when it is
used as a serdes port, meaning it switches between the internal PHY and
the serdes, changing the read-only C_Mode value depending on which
first gains link. Consequently, the C_Mode value read at probe does not
accurately reflect whether the port has the serdes associated with it.
In "net: dsa: mv88e6xxx: add mv88e6352_g2_scratch_port_has_serdes()",
we added a way to read the hardware configuration to determine which
port has the serdes associated with it. Use this to determine which
port reports the serdes statistics.
Reviewed-by: Marek Behún <kabel@kernel.org>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Now that the mv88e6xxx chip drivers are supplying the supported
interfaces and MAC capabilities, switch the driver to use the generic
phylink validation implementation by removing our own validation
implementations. This causes DSA to call phylink_generic_validate()
on our behalf.
Reviewed-by: Marek Behún <kabel@kernel.org>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Populate the supported interfaces and MAC capabilities for the
Marvell MV88E6xxx DSA switches in preparation to using these for the
validation functionality.
Patch co-authored by Marek.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Marek Behún <kabel@kernel.org> [ fixed 6341 and 6393x ]
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Read the hardware configuration to determine which port is attached
to the serdes.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Given that standalone ports are now configured to bypass the ATU and
forward all frames towards the upstream port, extend the ATU bypass to
multichip systems.
Load VID 0 (standalone) into the VTU with the policy bit set. Since
VID 4095 (bridged) is already loaded, we now know that all VIDs in use
are always available in all VTUs. Therefore, we can safely enable
802.1Q on DSA ports.
Setting the DSA ports' VTU policy to TRAP means that all incoming
frames on VID 0 will be classified as MGMT - as a result, the ATU is
bypassed on all subsequent switches.
With this isolation in place, we are able to support configurations
that are simultaneously very quirky and very useful. Quirky because it
involves looping cables between local switchports like in this
example:
CPU
| .------.
.---0---. | .----0----.
| sw0 | | | sw1 |
'-1-2-3-' | '-1-2-3-4-'
$ @ '---' $ @ % %
We have three physically looped pairs ($, @, and %).
This is very useful because it allows us to run the kernel's
kselftests for the bridge on mv88e6xxx hardware.
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This chip has support for the same per-port policy actions found in
later versions of LinkStreet devices.
Fixes: f3a2cd326e44 ("net: dsa: mv88e6xxx: introduce .port_set_policy")
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
A VTU entry with policy enabled is used in combination with a port's
VTU policy setting to override normal switching behavior for frames
assigned to the entry's VID.
A typical example is to Treat all frames in a particular VLAN as
control traffic, and trap them to the CPU. In which case the relevant
user port's VTU policy would be set to TRAP.
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Clear MapDA on standalone ports to bypass any ATU lookup that might
point the packet in the wrong direction. This means that all packets
are flooded using the PVT config. So make sure that standalone ports
are only allowed to communicate with the local upstream port.
Here is a scenario in which this is needed:
CPU
| .----.
.---0---. | .--0--.
| sw0 | | | sw1 |
'-1-2-3-' | '-1-2-'
'---'
- sw0p1 and sw1p1 are bridged
- sw0p2 and sw1p2 are in standalone mode
- Learning must be enabled on sw0p3 in order for hardware forwarding
to work properly between bridged ports
1. A packet with SA :aa comes in on sw1p2
1a. Egresses sw1p0
1b. Ingresses sw0p3, ATU adds an entry for :aa towards port 3
1c. Egresses sw0p0
2. A packet with DA :aa comes in on sw0p2
2a. If an ATU lookup is done at this point, the packet will be
incorrectly forwarded towards sw0p3. With this change in place,
the ATU is bypassed and the packet is forwarded in accordance
with the PVT, which only contains the CPU port.
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If the physical clock supports cross timestamping (it has the
getcrosststamp() function), provide a wrapper in the virtual clock to
enable cross timestamping.
This adds support for the PTP_SYS_OFFSET_PRECISE ioctl.
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Cc: Yangbo Lu <yangbo.lu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If the physical clock has the gettimex64() function, provide a
gettimex64() wrapper in the virtual clock to enable more accurate
and stable synchronization.
This adds support for the PTP_SYS_OFFSET_EXTENDED ioctl.
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Cc: Yangbo Lu <yangbo.lu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Increase the maximum frequency offset of virtual clocks to 50% to enable
faster slewing corrections.
This value cannot be represented as scaled ppm when long has 32 bits,
but that is already the case for other drivers, even those that provide
the adjfine() function, i.e. 32-bit applications are expected to check
for the limit.
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Cc: Yangbo Lu <yangbo.lu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When unregistering a physical clock which has some virtual clocks,
unregister the virtual clocks with it.
This fixes the following oops, which can be triggered by unloading
a driver providing a PTP clock when it has enabled virtual clocks:
BUG: unable to handle page fault for address: ffffffffc04fc4d8
Oops: 0000 [#1] PREEMPT SMP NOPTI
RIP: 0010:ptp_vclock_read+0x31/0xb0
Call Trace:
timecounter_read+0xf/0x50
ptp_vclock_refresh+0x2c/0x50
? ptp_clock_release+0x40/0x40
ptp_aux_kworker+0x17/0x30
kthread_worker_fn+0x9b/0x240
? kthread_should_park+0x30/0x30
kthread+0xe2/0x110
? kthread_complete_and_exit+0x20/0x20
ret_from_fork+0x22/0x30
Fixes: 73f37068d540 ("ptp: support ptp physical/virtual clocks conversion")
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Cc: Yangbo Lu <yangbo.lu@nxp.com>
Cc: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Populate the supported interfaces and MAC capabilities for the xrs700x
family of DSA switches and remove the old validate implementation to
allow DSA to use phylink_generic_validate() for this switch driver.
According to commit ee00b24f32eb ("net: dsa: add Arrow SpeedChips
XRS700x driver") the switch supports one RMII port and up to three
RGMII ports. This commit assumes that port 0 is the RMII port and the
remainder are RGMII.
This commit also results in the Autoneg bit being set in the ethtool
link modes, which wasn't in the original; if this switch supports
RGMII to a 10/100/1G PHY, then surely we want to allow Autoneg on the
PHY.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Populate the supported interfaces and MAC capabilities for the QCA8K
DSA switch and remove the old validate implementation to allow DSA to
use phylink_generic_validate() for this switch driver.
In making this change, we bring consistency to the ethtool linkmodes
that phylink's validate step produces, thereby following the expected
behaviour as the phylink documentation has explained. Specifically, the
ethtool 1000baseX_Full capability is now permitted for all interface
modes, as it is a property of the PHY driver whether 1000baseX fiber
connections can be supported.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Populate the supported interfaces and MAC capabilities for the
Microchip KSZ8795 DSA switch and remove the old validate implementation
to allow DSA to use phylink_generic_validate() for this switch driver.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Populate the supported interfaces and MAC capabilities for the bcm_sf2
DSA switch and remove the old validate implementation to allow DSA to
use phylink_generic_validate() for this switch driver.
The exclusion of Gigabit linkmodes for MII and Reverse MII links is
handled within phylink_generic_validate() in phylink, so there is no
need to make them conditional on the interface mode in the driver.
Thanks to Florian Fainelli for suggesting how to populate the supported
interfaces.
Link: https://lore.kernel.org/r/3b3fed98-0c82-99e9-dc72-09fe01c2bcf3@gmail.com
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Populate the supported interfaces and MAC capabilities for the AR9331
DSA switch and remove the old validate implementation to allow DSA to
use phylink_generic_validate() for this switch driver.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Increase the receive buffer size used for data received from the
modem to 32KB, to improve download performance by allowing much
greater aggregation.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Allow RX endpoints to have differing receive buffer sizes. Define
the receive buffer size in the configuration data, and use that
rather than IPA_RX_BUFFER_SIZE when configuring the endpoint.
Add verification in ipa_endpoint_data_valid_one() that the receive
buffer specified for AP RX endpoints is both big enough to handle at
least one full packet, and not so big in an aggregating endpoint
that its size can't be represented when programming the hardware.
Move aggr_byte_limit_max() up in "ipa_endpoint.c" so it can be used
earlier in the file without a forward-reference.
Initially we'll just keep the 8KB receive buffer size already in use
for all AP RX endpoints..
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Replace acpi_bus_get_device() that is going to be dropped with
acpi_fetch_acpi_dev().
No intentional functional impact.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/r/11918902.O9o76ZdvQC@kreacher
Link: https://lore.kernel.org/r/11920660.O9o76ZdvQC@kreacher
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Introduce qca8k_bulk_read/write() function to use mgmt Ethernet way to
read/write packet in bulk. Make use of this new function in the fdb
function and while at it reduce the reg for fdb_read from 4 to 3 as the
max bit for the ARL(fdb) table is 83 bits.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
mgmt Ethernet packet can read/write up to 16byte at times. The len reg
is limited to 15 (0xf). The switch actually sends and accepts data in 4
different steps of len values.
Len steps:
- 0: nothing
- 1-4: first 4 byte
- 5-6: first 12 byte
- 7-15: all 16 byte
In the alloc skb function we check if the len is 16 and we fix it to a
len of 15. It the read/write function interest to extract the real asked
data. The tagger handler will always copy the fully 16byte with a READ
command. This is useful for some big regs like the fdb reg that are
more than 4byte of data. This permits to introduce a bulk function that
will send and request the entire entry in one go.
Write function is changed and it does now require to pass the pointer to
val to also handle array val.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
From Documentation, we can cache lo and hi the same way we do with the
page. This massively reduce the mdio write as 3/4 of the time as we only
require to write the lo or hi part for a mdio write.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There can be multiple qca8k switch on the same system. Move the static
qca8k_current_page to qca8k_priv and make it specific for each switch.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use mgmt Ethernet also for phy read/write if availabale. Use a different
seq number to make sure we receive the correct packet.
On any error, we fallback to the legacy mdio read/write.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The switch can autocast MIB counter using Ethernet packet.
Add support for this and provide a handler for the tagger.
The switch will send packet with MIB counter for each port, the switch
will use completion API to wait for the correct packet to be received
and will complete the task only when each packet is received.
Although the handler will drop all the other packet, we still have to
consume each MIB packet to complete the request. This is done to prevent
mixed data with concurrent ethtool request.
connect_tag_protocol() is used to add the handler to the tag_qca tagger,
master_state_change() use the MIB lock to make sure no MIB Ethernet is
in progress.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add qca8k side support for mgmt read/write in Ethernet packet.
qca8k supports some specially crafted Ethernet packet that can be used
for mgmt read/write instead of the legacy method uart/internal mdio.
This add support for the qca8k side to craft the packet and enqueue it.
Each port and the qca8k_priv have a special struct to put data in it.
The completion API is used to wait for the packet to be received back
with the requested data.
The various steps are:
1. Craft the special packet with the qca hdr set to mgmt read/write
mode.
2. Set the lock in the dedicated mgmt struct.
3. Increment the seq number and set it in the mgmt pkt
4. Reinit the completion.
5. Enqueue the packet.
6. Wait the packet to be received.
7. Use the data set by the tagger to complete the mdio operation.
If the completion timeouts or the ack value is not true, the legacy
mdio way is used.
It has to be considered that in the initial setup mdio is still used and
mdio is still used until DSA is ready to accept and tag packet.
tag_proto_connect() is used to fill the required handler for the tagger
to correctly parse and elaborate the special Ethernet mdio packet.
Locking is added to qca8k_master_change() to make sure no mgmt Ethernet
are in progress.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
MDIO/MIB Ethernet require the master port and the tagger availabale to
correctly work. Use the new api master_state_change to track when master
is operational or not and set a bool in qca8k_priv.
We cache the first cached master available and we check if other cpu
port are operational when the cached one goes down.
This cached master will later be used by mdio read/write and mib request to
correctly use the working function.
qca8k implementation for MDIO/MIB Ethernet is bad. CPU port0 is the only
one that answers with the ack packet or sends MIB Ethernet packets. For
this reason the master_state_change ignore CPU port6 and only checks
CPU port0 if it's operational and enables this mode.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Ideally the size would depend on the link speed, but the recycle
ring is created when the interface is brought up before the driver
knows the link speed. So size it for the maximum speed of a given NIC.
PowerPC is only supported on SFN7xxx and SFN8xxx NICs.
With this patch on a 40G NIC the number of calls to alloc_pages and
friends went down from about 18% to under 2%.
On a 10G NIC the number of calls to alloc_pages and friends went down
from about 15% to 0 (perf did not capture any calls during the 60
second test).
On a 100G NIC the number of calls to alloc_pages and friends went down
from about 23% to 4%.
Reported-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20220131111054.cp4f6foyinaarwbn@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
According to Realtek RTL8168h supports the same L1.2 control as RTL8125.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/4784d5ce-38ac-046a-cbfa-5fdd9773f820@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Implement the function get_ts_info in ethtool_ops which is needed to get
the HW capabilities for timestamping.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When doing 2-step timestamping the HW will generate an interrupt when it
managed to timestamp a frame. It is the SW responsibility to read it
from the FIFO.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Update both the extraction and injection to do timestamping of the
frames. The extraction is always doing the timestamping while for
injection is doing the timestamping only if it is configured.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implement the ioctl callbacks SIOCSHWTSTAMP and SIOCGHWTSTAMP to allow
to configure the ports to enable/disable timestamping for TX. The RX
timestamping is always enabled. The HW is capable to run both 1-step
timestamping and 2-step timestamping.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The lan966x has 3 PHC. Enable each of them, for now all the
timestamping is happening on the first PHC.
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add the registers that will be used to configure the PHC in the HW.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
-queue
Tony Nguyen says:
====================
10GbE Intel Wired LAN Driver Updates 2022-01-31
Alexander Lobakin says:
This is an interpolation of [0] to other Intel Ethernet drivers
(and is (re)based on its code).
The main aim is to keep XDP metadata not only in case with
build_skb(), but also when we do napi_alloc_skb() + memcpy().
All Intel drivers suffers from the same here:
- metadata gets lost on XDP_PASS in legacy-rx;
- excessive headroom allocation on XSK Rx to skbs;
- metadata gets lost on XSK Rx to skbs.
Those get especially actual in XDP Hints upcoming.
I couldn't have addressed the first one for all Intel drivers due to
that they don't reserve any headroom for now in legacy-rx mode even
with XDP enabled. This is hugely wrong, but requires quite a bunch
of work and a separate series. Luckily, ice doesn't suffer from
that.
igc has 1 and 3 already fixed in [0].
[0] https://lore.kernel.org/netdev/163700856423.565980.10162564921347693758.stgit@firesoul
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
sh_eth_{suspend|resume}() initialize their local variable 'ret' to 0 but
this value is never really used, thus we can kill those intializers...
Signed-off-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Link: https://lore.kernel.org/r/f09d7c64-4a2b-6973-09a4-10d759ed0df4@omp.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
By profiling, discovered that ena device driver allocates skb by
build_skb() and frees by napi_skb_cache_put(). Because the driver
does not use napi skb cache in allocation path, napi skb cache is
periodically filled and flushed. This is waste of napi skb cache.
As ena_alloc_skb() is called only in napi, Use napi_build_skb()
and napi_alloc_skb() when allocating skb.
This patch was tested on aws a1.metal instance.
[ jwiedmann.dev@gmail.com: Use napi_alloc_skb() instead of
netdev_alloc_skb_ip_align() to keep things consistent. ]
Signed-off-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Acked-by: Shay Agroskin <shayagr@amazon.com>
Link: https://lore.kernel.org/r/YfUAkA9BhyOJRT4B@ip-172-31-19-208.ap-northeast-1.compute.internal
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Change qed_mcp_cmd() to use msleep() (by setting QED_MB_FLAG_CAN_SLEEP
flag) and add new nosleep() version of the api. These api are used to
issue cmds to management fw and the change affects how driver
behaves while waiting for a response/resource.
All sleepable callers of the existing api now use msleep() version. For
non-sleepable callers, the new nosleep() version is explicitly used.
Signed-off-by: Venkata Sudheer Kumar Bhavaraju <vbhavaraju@marvell.com>
Signed-off-by: Alok Prasad <palok@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Link: https://lore.kernel.org/r/20220131005235.1647881-1-vbhavaraju@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
For now, if the XDP prog returns XDP_PASS on XSK, the metadata
will be lost as it doesn't get copied to the skb.
Copy it along with the frame headers. Account its size on skb
allocation, and when copying just treat it as a part of the frame
and do a pull after to "move" it to the "reserved" zone.
net_prefetch() xdp->data_meta and align the copy size to speed-up
memcpy() a little and better match ixgbe_construct_skb().
Fixes: d0bcacd0a130 ("ixgbe: add AF_XDP zero-copy Rx support")
Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Sandeep Penigalapati <sandeep.penigalapati@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
{__,}napi_alloc_skb() allocates and reserves additional NET_SKB_PAD
+ NET_IP_ALIGN for any skb.
OTOH, ixgbe_construct_skb_zc() currently allocates and reserves
additional `xdp->data - xdp->data_hard_start`, which is
XDP_PACKET_HEADROOM for XSK frames.
There's no need for that at all as the frame is post-XDP and will
go only to the networking stack core.
Pass the size of the actual data only to __napi_alloc_skb() and
don't reserve anything. This will give enough headroom for stack
processing.
Fixes: d0bcacd0a130 ("ixgbe: add AF_XDP zero-copy Rx support")
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Sandeep Penigalapati <sandeep.penigalapati@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
To not dereference bi->xdp each time in ixgbe_construct_skb_zc(),
pass bi->xdp as an argument instead of bi. We can also call
xsk_buff_free() outside of the function as well as assign bi->xdp
to NULL, which seems to make it closer to its name.
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Tested-by: Sandeep Penigalapati <sandeep.penigalapati@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
{__,}napi_alloc_skb() allocates and reserves additional NET_SKB_PAD
+ NET_IP_ALIGN for any skb.
OTOH, igc_construct_skb_zc() currently allocates and reserves
additional `xdp->data_meta - xdp->data_hard_start`, which is about
XDP_PACKET_HEADROOM for XSK frames.
There's no need for that at all as the frame is post-XDP and will
go only to the networking stack core.
Pass the size of the actual data only (+ meta) to
__napi_alloc_skb() and don't reserve anything. This will give
enough headroom for stack processing.
Also, net_prefetch() xdp->data_meta and align the copy size to
speed-up memcpy() a little and better match igc_construct_skb().
Fixes: fc9df2a0b520 ("igc: Enable RX via AF_XDP zero-copy")
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
For now, if the XDP prog returns XDP_PASS on XSK, the metadata will
be lost as it doesn't get copied to the skb.
Copy it along with the frame headers. Account its size on skb
allocation, and when copying just treat it as a part of the frame
and do a pull after to "move" it to the "reserved" zone.
net_prefetch() xdp->data_meta and align the copy size to speed-up
memcpy() a little and better match ice_construct_skb().
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
{__,}napi_alloc_skb() allocates and reserves additional NET_SKB_PAD
+ NET_IP_ALIGN for any skb.
OTOH, ice_construct_skb_zc() currently allocates and reserves
additional `xdp->data - xdp->data_hard_start`, which is
XDP_PACKET_HEADROOM for XSK frames.
There's no need for that at all as the frame is post-XDP and will
go only to the networking stack core.
Pass the size of the actual data only to __napi_alloc_skb() and
don't reserve anything. This will give enough headroom for stack
processing.
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
In "legacy-rx" mode represented by ice_construct_skb(), we can
still use XDP (and XDP metadata), but after XDP_PASS the metadata
will be lost as it doesn't get copied to the skb.
Copy it along with the frame headers. Account its size on skb
allocation, and when copying just treat it as a part of the frame
and do a pull after to "move" it to the "reserved" zone.
Point net_prefetch() to xdp->data_meta instead of data. This won't
change anything when the meta is not here, but will save some cache
misses otherwise.
Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
For now, if the XDP prog returns XDP_PASS on XSK, the metadata will
be lost as it doesn't get copied to the skb.
Copy it along with the frame headers. Account its size on skb
allocation, and when copying just treat it as a part of the frame
and do a pull after to "move" it to the "reserved" zone.
net_prefetch() xdp->data_meta and align the copy size to speed-up
memcpy() a little and better match i40e_construct_skb().
Fixes: 0a714186d3c0 ("i40e: add AF_XDP zero-copy Rx support")
Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
{__,}napi_alloc_skb() allocates and reserves additional NET_SKB_PAD
+ NET_IP_ALIGN for any skb.
OTOH, i40e_construct_skb_zc() currently allocates and reserves
additional `xdp->data - xdp->data_hard_start`, which is
XDP_PACKET_HEADROOM for XSK frames.
There's no need for that at all as the frame is post-XDP and will
go only to the networking stack core.
Pass the size of the actual data only to __napi_alloc_skb() and
don't reserve anything. This will give enough headroom for stack
processing.
Fixes: 0a714186d3c0 ("i40e: add AF_XDP zero-copy Rx support")
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
|
|
Reuse the dropped page in RX path to save page allocation
overhead.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
