Tony Nguyen says:
====================
libeth: add libeth_xdp helper lib
Alexander Lobakin says:
Time to add XDP helpers infra to libeth to greatly simplify adding
XDP to idpf and iavf, as well as improve and extend XDP in ice and
i40e. Any vendor is free to reuse helpers. If this happens, I'm fine
with moving the folder of out intel/.
The helpers greatly simplify building xdp_buff, running a prog,
handling the verdict, implement XDP_TX, .ndo_xdp_xmit, XDP buffer
completion. Same applies to XSk (with XSk xmit instead of
.ndo_xdp_xmit, plus stuff like XSk wakeup).
They are entirely generic with no HW definitions or assumptions.
HW-specific stuff like parsing Rx desc / filling Tx desc is passed
from the driver as inline callbacks.
For now, key assumptions that optimize performance / avoid code
bloat, but might not fit every driver in driver/net/:
* netmem holding the buffers are always order-0;
* driver has separate XDP Tx queues, doesn't use stack queues for
that. For best efficiency, you may want to have nr_cpu_ids XDP
queues, but less (queue sharing) is also supported;
* XDP Tx queues are interrupt-less and use "lazy" cleaning only
when there are less than 1/4 free Tx descriptors of the queue
size;
* main target platforms are 64-bit, although 32-bit is also fully
supported, but the code might be not as optimized for them.
Library code already supports multi-buffer for all kinds of Tx and
both header split and no split for Rx and Tx. Frags can come from
devmem/io_uring etc., direct `struct page *` is used only for header
buffers for which it's always true.
Drivers are free to pass their own Rx hints and XSK xmit hints ops.
XDP_TX and ndo_xdp_xmit use onstack bulk for the frames to be sent
and send them by batches of 16 buffers. This eats ~280 bytes on the
stack, but gives good boosts and allow to greatly optimize the main
sending function leaving it without any error/exception paths.
XSk xmit fills Tx descriptors in the loop unrolled by 8. This was
proven to improve perf on ice and i40e. XDP_TX and ndo_xdp_xmit
doesn't use unrolling as I wasn't able to get any improvements in
those scenenarios from this, while +1 Kb for their sending functions
for nothing doesn't sound reasonable.
XSk wakeup, instead of traditionally used "SW interrupts" provided
by NICs, uses IPI to schedule NAPI on the CPU corresponding to the
given queue pair. It gives better control over CPU distribution and
in general performs way better than "SW interrupts", plus allows us
to not pass any HW-specific callbacks there.
The code is built the way that all callbacks passed from drivers
get inlined; in general, most of hotpath gets inlined. Everything
slow/exception lands to .c files in the libeth folder, doesn't
create copies in the drivers themselves and doesn't overloat
hotpath.
Sure, inlining means that hotpath will be compiled into every driver
that uses the lib, but the core code is written in one place, so no
copying of bugs happens. Fixed once -- works everywhere.
The last commit might look like sorta hack, but it gives really good
boosts and decreases object code size, plus there are checks that
all those wider accesses are fully safe, so I don't feel anything
bad about it.
An example of using libeth_xdp can be found either on my GitHub or
on the mailing lists here ("XDP for idpf"). Macros for building
driver XDP functions lead to that some implementations (XDP_TX,
ndo_xdp_xmit etc.) consist of really only a few lines.
* '200GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
libeth: xdp, xsk: access adjacent u32s as u64 where applicable
libeth: xsk: add XSkFQ refill and XSk wakeup helpers
libeth: xsk: add XSk Rx processing support
libeth: xsk: add XSk xmit functions
libeth: xsk: add XSk XDP_TX sending helpers
libeth: xdp: add RSS hash hint and XDP features setup helpers
libeth: xdp: add templates for building driver-side callbacks
libeth: xdp: add XDP prog run and verdict result handling
libeth: xdp: add helpers for preparing/processing &libeth_xdp_buff
libeth: xdp: add XDPSQ cleanup timers
libeth: xdp: add XDPSQ locking helpers
libeth: xdp: add XDPSQE completion helpers
libeth: xdp: add .ndo_xdp_xmit() helpers
libeth: xdp: add XDP_TX buffers sending
libeth: support native XDP and register memory model
libeth: convert to netmem
libeth, libie: clean symbol exports up a little
====================
Link: https://patch.msgid.link/20250616201639.710420-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This is the netmem counterpart of page_pool_dev_alloc_pages() which
uses the default GFP flags for RX.
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250616141441.1243044-4-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Allow drivers that have moved over to netmem to do fragment coalescing.
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250616141441.1243044-3-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The only generic interface to execute asynchronously in the BH context is
tasklet; however, it's marked deprecated and has some design flaws. To
replace tasklets, BH workqueue support was recently added. A BH workqueue
behaves similarly to regular workqueues except that the queued work items
are executed in the BH context.
This patch converts TCP Small Queues implementation from tasklet to BH
workqueue.
Semantically, this is an equivalent conversion and there shouldn't be any
user-visible behavior changes. While workqueue's queueing and execution
paths are a bit heavier than tasklet's, unless the work item is being queued
every packet, the difference hopefully shouldn't matter.
My experience with the networking stack is very limited and this patch
definitely needs attention from someone who actually understands networking.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Cc: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/aFBeJ38AS1ZF3Dq5@slm.duckdns.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Locally-generated MC packets have so far not been subject to MC routing.
Instead an MC-enabled installation would maintain the MC routing tables,
and separately from that the list of interfaces to send packets to as part
of the VXLAN FDB and MDB.
In a previous patch, a ip_mr_output() and ip6_mr_output() routines were
added for IPv4 and IPv6. All locally generated MC traffic is now passed
through these functions. For reasons of backward compatibility, an SKB
(IPCB / IP6CB) flag guards the actual MC routing.
This patch adds logic to set the flag, and the UAPI to enable the behavior.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://patch.msgid.link/d899655bb7e9b2521ee8c793e67056b9fd02ba12.1750113335.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Multicast routing is today handled in the input path. Locally generated MC
packets don't hit the IPMR code today. Thus if a VXLAN remote address is
multicast, the driver needs to set an OIF during route lookup. Thus MC
routing configuration needs to be kept in sync with the VXLAN FDB and MDB.
Ideally, the VXLAN packets would be routed by the MC routing code instead.
To that end, this patch adds support to route locally generated multicast
packets. The newly-added routines do largely what ip6_mr_input() and
ip6_mr_forward() do: make an MR cache lookup to find where to send the
packets, and use ip6_output() to send each of them. When no cache entry is
found, the packet is punted to the daemon for resolution.
Similarly to the IPv4 case in a previous patch, the new logic is contingent
on a newly-added IP6CB flag being set.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Link: https://patch.msgid.link/3bcc034a3ab4d3c291072fff38f78d7fbbeef4e6.1750113335.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ip6tunnel_xmit() erases the contents of the SKB control block. In order to
be able to set particular IP6CB flags on the SKB, add a corresponding
parameter, and propagate it to udp_tunnel6_xmit_skb() as well.
In one of the following patches, VXLAN driver will use this facility to
mark packets as subject to IPv6 multicast routing.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://patch.msgid.link/acb4f9f3e40c3a931236c3af08a720b017fbfbfb.1750113335.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The function always returns zero, thus the return value does not carry any
signal. Just make it void.
Most callers already ignore the return value. However:
- Refold arguments of the call from sctp_v6_xmit() so that they fit into
the 80-column limit.
- tipc_udp_xmit() initializes err from the return value, but that should
already be always zero at that point. So there's no practical change, but
elision of the assignment prompts a couple more tweaks to clean up the
function.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://patch.msgid.link/7facacf9d8ca3ca9391a4aee88160913671b868d.1750113335.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Multicast routing is today handled in the input path. Locally generated MC
packets don't hit the IPMR code today. Thus if a VXLAN remote address is
multicast, the driver needs to set an OIF during route lookup. Thus MC
routing configuration needs to be kept in sync with the VXLAN FDB and MDB.
Ideally, the VXLAN packets would be routed by the MC routing code instead.
To that end, this patch adds support to route locally generated multicast
packets. The newly-added routines do largely what ip_mr_input() and
ip_mr_forward() do: make an MR cache lookup to find where to send the
packets, and use ip_mc_output() to send each of them. When no cache entry
is found, the packet is punted to the daemon for resolution.
However, an installation that uses a VXLAN underlay netdevice for which it
also has matching MC routes, would get a different routing with this patch.
Previously, the MC packets would be delivered directly to the underlay
port, whereas now they would be MC-routed. In order to avoid this change in
behavior, introduce an IPCB flag. Only if the flag is set will
ip_mr_output() actually engage, otherwise it reverts to ip_mc_output().
This code is based on work by Roopa Prabhu and Nikolay Aleksandrov.
Signed-off-by: Roopa Prabhu <roopa@nvidia.com>
Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Benjamin Poirier <bpoirier@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/0aadbd49330471c0f758d54afb05eb3b6e3a6b65.1750113335.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
iptunnel_xmit() erases the contents of the SKB control block. In order to
be able to set particular IPCB flags on the SKB, add a corresponding
parameter, and propagate it to udp_tunnel_xmit_skb() as well.
In one of the following patches, VXLAN driver will use this facility to
mark packets as subject to IP multicast routing.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Acked-by: Antonio Quartulli <antonio@openvpn.net>
Link: https://patch.msgid.link/89c9daf9f2dc088b6b92ccebcc929f51742de91f.1750113335.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The header currently tests the VLAN core with an explicit pair of 'if
defined' checks:
#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
Instead, use IS_ENABLED() which is the kernel way to test whether an
option is configured as builtin/module.
This is purely cosmetic – no functional changes.
Reviewed-by: Alex Lazar <alazar@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
Link: https://patch.msgid.link/20250616132626.1749331-4-gal@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When CONFIG_VLAN_8021Q=n, a set of stub helpers are used, three of these
helpers use BUG() unconditionally.
This code should not be reached, as callers of these functions should
always check for is_vlan_dev() first, but the usage of BUG() is not
recommended, replace it with WARN_ON() instead.
Reviewed-by: Alex Lazar <alazar@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
Link: https://patch.msgid.link/20250616132626.1749331-3-gal@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a stub implementation of is_vlan_dev() that returns false when
VLAN support is not compiled in (CONFIG_VLAN_8021Q=n).
This allows us to compile-out VLAN-dependent dead code when it is not
needed.
This also resolves the following compilation error when:
* CONFIG_VLAN_8021Q=n
* CONFIG_OBJTOOL=y
* CONFIG_OBJTOOL_WERROR=y
drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.o: error: objtool: parse_mirred.isra.0+0x370: mlx5e_tc_act_vlan_add_push_action() missing __noreturn in .c/.h or NORETURN() in noreturns.h
The error occurs because objtool cannot determine that unreachable BUG()
(which doesn't return) calls in VLAN code paths are actually dead code
when VLAN support is disabled.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Link: https://patch.msgid.link/20250616132626.1749331-2-gal@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Trivial fix to a couple of outdated netmem comments. No code changes,
just more accurately describing current code.
Signed-off-by: Mina Almasry <almasrymina@google.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20250615203511.591438-1-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add support for legacy Broadcom FCS tags, which are similar to
DSA_TAG_PROTO_BRCM_LEGACY.
BCM5325 and BCM5365 switches require including the original FCS value and
length, as opposed to BCM63xx switches.
Adding the original FCS value and length to DSA_TAG_PROTO_BRCM_LEGACY would
impact performance of BCM63xx switches, so it's better to create a new tag.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250614080000.1884236-3-noltari@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Now that we have removed the RFC3517/RFC6675 hints,
tcp_clear_retrans_hints_partial() is empty, and can be removed.
Suggested-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250615001435.2390793-4-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Now that obsolete RFC3517/RFC6675 TCP loss detection has been removed,
we can remove the somewhat complex and intrusive code to maintain its
hint state: lost_skb_hint and lost_cnt_hint.
This commit makes tcp_clear_retrans_hints_partial() empty. We will
remove tcp_clear_retrans_hints_partial() and its call sites in the
next commit.
Suggested-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250615001435.2390793-3-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pavel Begunkov says:
====================
io_uring cmd for tx timestamps (part)
Apply the networking helpers for the io_uring timestamp API.
====================
Link: https://patch.msgid.link/cover.1750065793.git.asml.silence@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Shradha Gupta says:
====================
Allow dyn MSI-X vector allocation of MANA
In this patchset we want to enable the MANA driver to be able to
allocate MSI-X vectors in PCI dynamically.
The first patch exports pci_msix_prepare_desc() in PCI to be able to
correctly prepare descriptors for dynamically added MSI-X vectors.
The second patch adds the support of dynamic vector allocation in
pci-hyperv PCI controller by enabling the MSI_FLAG_PCI_MSIX_ALLOC_DYN
flag and using the pci_msix_prepare_desc() exported in first patch.
The third patch adds a detailed description of the irq_setup(), to
help understand the function design better.
The fourth patch is a preparation patch for mana changes to support
dynamic IRQ allocation. It contains changes in irq_setup() to allow
skipping first sibling CPU sets, in case certain IRQs are already
affinitized to them.
The fifth patch has the changes in MANA driver to be able to allocate
MSI-X vectors dynamically. If the support does not exist it defaults to
older behavior.
* 'shradha_v6.16-rc1' of https://github.com/shradhagupta6/linux:
net: mana: Allocate MSI-X vectors dynamically
net: mana: Allow irq_setup() to skip cpus for affinity
net: mana: explain irq_setup() algorithm
PCI: hv: Allow dynamic MSI-X vector allocation
PCI/MSI: Export pci_msix_prepare_desc() for dynamic MSI-X allocations
====================
Link: https://patch.msgid.link/1749650984-9193-1-git-send-email-shradhagupta@linux.microsoft.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, the MANA driver allocates MSI-X vectors statically based on
MANA_MAX_NUM_QUEUES and num_online_cpus() values and in some cases ends
up allocating more vectors than it needs. This is because, by this time
we do not have a HW channel and do not know how many IRQs should be
allocated.
To avoid this, we allocate 1 MSI-X vector during the creation of HWC and
after getting the value supported by hardware, dynamically add the
remaining MSI-X vectors.
Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
For supporting dynamic MSI-X vector allocation by PCI controllers, enabling
the flag MSI_FLAG_PCI_MSIX_ALLOC_DYN is not enough, msix_prepare_msi_desc()
to prepare the MSI descriptor is also needed.
Export pci_msix_prepare_desc() to allow PCI controllers to support dynamic
MSI-X vector allocation.
Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Saurabh Sengar <ssengar@linux.microsoft.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Remove now unused function phy_driver_is_genphy_10g().
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/49b0589a-9604-4ee9-add5-28fbbbe2c2f3@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use new flag phydev->is_genphy_driven to simplify this function.
Note that this includes a minor functional change:
Now this function returns true if ANY of the genphy drivers
is bound to the PHY device.
We have only one user in DSA driver mt7530, and there the
functional change doesn't matter.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/c9ac3a7d-262a-425d-9153-97fe3ca6280a@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In order to get rid of phy_driver_is_genphy() and
phy_driver_is_genphy_10g(), as first step add and use a flag
phydev->is_genphy_driven.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/3f3ad6dc-402e-4915-8d5a-2306b6d5562b@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Number of RVU PFs on CN20K silicon have increased to 96 from maximum
of 32 that were supported on earlier silicons. Every RVU PF and VF is
identified by HW using a 16bit PF_FUNC value. Due to the change in
Max number of PFs in CN20K, the bit encoding of this PF_FUNC has changed.
This patch handles the change by using helper functions(using silicon
check) to use PF,VF masks and shifts to support both new silicon CN20K,
OcteonTx series. These helper functions are used in different modules.
Also moved the NIX AF register offset macros to other files which
will be posted in coming patches.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Sai Krishna <saikrishnag@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Link: https://patch.msgid.link/1749639716-13868-2-git-send-email-sbhatta@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
To collaborate with hardware servicing events, upon receiving the special
EQE notification from the HW channel, remove the devices on this bus.
Then, after a waiting period based on the device specs, rescan the parent
bus to recover the devices.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Shradha Gupta <shradhagupta@linux.microsoft.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/1749834034-18498-1-git-send-email-haiyangz@linux.microsoft.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move netpoll_print_options() from net/core/netpoll.c to
drivers/net/netconsole.c and make it static. This function is only used
by netconsole, so there's no need to export it or keep it in the public
netpoll API.
This reduces the netpoll API surface and improves code locality
by keeping netconsole-specific functionality within the netconsole
driver.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20250613-rework-v3-4-0752bf2e6912@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move netpoll_parse_ip_addr() and netpoll_parse_options() from the generic
netpoll module to the netconsole module where they are actually used.
These functions were originally placed in netpoll but are only consumed by
netconsole. This refactoring improves code organization by:
- Removing unnecessary exported symbols from netpoll
- Making netpoll_parse_options() static (no longer needs global visibility)
- Reducing coupling between netpoll and netconsole modules
The functions remain functionally identical - this is purely a code
reorganization to better reflect their actual usage patterns. Here are
the changes:
1) Move both functions from netpoll to netconsole
2) Add static to netpoll_parse_options()
3) Removed the EXPORT_SYMBOL()
PS: This diff does not change the function format, so, it is easy to
review, but, checkpatch will not be happy. A follow-up patch will
address the current issues reported by checkpatch.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20250613-rework-v3-3-0752bf2e6912@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move np_info(), np_err(), and np_notice() macros from internal
implementation to the public netpoll header file to make them
available for use by netpoll consumers.
These logging macros provide consistent formatting for netpoll-related
messages by automatically prefixing log output with the netpoll instance
name.
The goal is to use the exact same format that is being displayed today,
instead of creating something netconsole-specific.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20250613-rework-v3-2-0752bf2e6912@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Since commit 97714695ef ("net: netconsole: Defer netpoll cleanup to
avoid lock release during list traversal"), netconsole no longer uses
__netpoll_cleanup(). With no remaining users, remove this function
from the exported netpoll API.
The function remains available internally within netpoll for use by
netpoll_cleanup().
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20250613-rework-v3-1-0752bf2e6912@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Several of the tcp_ao events are only called when CONFIG_TCP_AO is
defined. As each event can take up to 5K regardless if they are used or
not, it's best not to define them when they are not used. Add #ifdef
around these events when they are not used.
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250612094616.4222daf0@batman.local.home
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On 64-bit systems, writing/reading one u64 is faster than two u32s even
when they're are adjacent in a struct. The compilers won't guarantee
they will combine those; I observed both successful and unsuccessful
attempts with both GCC and Clang, and it's not easy to say what it
depends on.
There's a few places in libeth_xdp winning up to several percent from
combined access (both performance and object code size, especially
when unrolling). Add __LIBETH_WORD_ACCESS and use it there on LE.
Drivers are free to optimize HW-specific callbacks under the same
definition.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
XSkFQ refill is pretty generic across the drivers minus FQ descriptor
filling and can easily be unified with one inline callback.
XSk wakeup is usually not, but here, instead of commonly used
"SW interrupts", I picked firing an IPI. In most tests, it showed better
performance; it also provides better control for userspace on which CPU
will handle the xmit, as SW interrupts honor IRQ affinity no matter
which core produces XSk xmit descs (while XDPSQs are associated 1:1
with cores having the same ID).
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Add XSk counterparts for preparing XSk &libeth_xdp_buff (adding head and
frags), running the program, and handling the verdict, inc. XDP_PASS.
Shortcuts in comparison with regular Rx: frags and all verdicts except
XDP_REDIRECT are under unlikely() and out of line; no checks for XDP
program presence as it's always true for XSk.
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> # optimizations
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reuse core sending functions to send XSk xmit frames.
Both metadata and no metadata pools/driver are supported. libeth_xdp
also provides generic XSk metadata ops, currently with the checksum
offload only and for cases when HW doesn't require supplying L3/L4
checksum offsets. Drivers are free to pass their own ops.
&libeth_xdp_tx_bulk is not used here as it would be redundant;
pool->tx_descs are accessed directly.
Fake "libeth_xsktmo" is needed to hide implementation details from the
drivers when they want to use the generic ops: the original struct is
defined in the same file where dev->xsk_tx_metadata_ops gets set to
avoid duplication of slowpath; at the same time; XSk xmit functions
use local "fast" copy to inline XMO callbacks.
Tx descriptor filling loop is unrolled by 8.
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> # optimizations
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Add Xsk counterparts for XDP_TX buffer sending and completion.
The same base structures and functions used from the libeth_xdp core,
with adjustments to that XSk Rx always operates on &xdp_buff_xsk for
both head and frags. And unlike regular Rx, here unlikely() are used
for frags, as the header split gives no benefits for XSk Rx, at
least for now.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
End the XDP section by adding helpers to setup XDP features, flipping
.ndo_xdp_xmit() support at runtime (in case when it's not always on),
and calculating the queue clean/refill threshold.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Defining driver-specific functions to pass to libeth_xdp functions can
induce boilerplates and/or look a bit cryptic with all those layers of
indirection. On the other hand, this indirection is needed to allow
compilers to uninline big functions even when passed to __always_inline
helpers (too much inlining also hurts performance in some cases), plus
to reuse some XDP helpers in XSk code.
Add macros to quickly build them, with the detailed kdoc. They take
names of the actual callbacks for filling a Tx descriptor and other
purely HW-specific things and wrap them appropriately.
LIBETH_XDP_DEFINE_{BEGIN,END}() is needed for GCC 8+ unfortunately to
let the drivers control which functions will be static and which global
without hitting `-Wold-style-declaration`.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Running a prog and handling the verdicts, up to napi_gro_receive()
is also pretty generic code not really differing between vendors
(except for Tx descriptor filling and Rx descriptor parsing).
Define a couple inlines to do that. The inline callbacks a driver
needs to pass is mentioned above: Tx descriptor filling for XDP_TX,
populating skb with the descriptor data for XDP_PASS, finalizing
XDPSQs after the polling loop for XDP_TX (kicking the HW to start
sending).
The populate callback passes only &libeth_xdp_buff assuming buff::desc
pointer is enough, plus you can always get the corresponding Rx queue
structure via container_of(buff::rxq). If not, a driver can extend
the buff with more fields directly on the stack without touching
libeth_xdp definitions.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Add convenience helpers to build an &xdp_buff. This means: general
initialization before the NAPI loop, adding head, adding frags etc.
libeth_xdp_process_buff() is the same what everybody have in their
drivers:
dma_sync_for_cpu();
if (!frag) {
add_head();
prefetch();
} else {
add_frag();
}
Note that I don't use net_prefetch(), sticking to the original
prefetch(). In none of my tests prefetching 128 bytes yielded better
perf than 64 bytes. That might differ if the headers are huge enough,
but then additional tunneling etc. overhead takes place, you either
way won't win a lot.
&libeth_xdp_stash is for cases when you exit the polling loop without
finishing building the buff. If that happens, you need to store the
buffer in the queue structure until the next loop and then restore it.
It makes no sense to place a whole full &xdp_buff there. Define a
minimal structure, which would store only the fields essential to
restore it.
I was able to pack it into 16 bytes, which is only 8 bytes bigger
than `struct sk_buff *skb` on x64.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
When XDP Tx queues are not interrupt-driven but use lazy cleaning,
i.e. only when there are less than `threshold` free descriptors left,
we also need cleanup timers to avoid &xdp_buff and &xdp_frame stall
for too long, especially with Page Pool (it warns every about inflight
pages every 60 second).
Let's say we sent 256 frames and don't need to send more, but we clean
only when the number of pending items >= 384. In that case, those 256
will stall until 128 more are sent. For this, add simple helpers to
run a timer which will clean the queue regardless, after 1 second of
the last send.
The timer is triggered when finalizing the queue. As long as there is
regular active traffic, the timer doesn't fire.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Unfortunately, it's not always possible to allocate
max(num_rxqs, nr_cpu_ids) even on hi-end NICs.
To mitigate this, add simple locking helpers to libeth_xdp.
As long as XDPSQs are not shared, the whole functionality is gated
behind a static lock. Otherwise, each bulk flush locks the queue for
the time of cleaning and filling the descriptors.
As long as this particular queue is not used by more than 1 CPU,
the impact is minimal (runtime check for boolean twice per 16+
descriptors).
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> # static key
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Similarly to libeth_tx_complete(), add libeth_xdp_complete_tx() to
handle XDP_TX and xmit buffers. Both use bulk return under the hood.
Also add out of line libeth_tx_complete_any() which handles both
regular and XDP frames (if libeth_xdp is loaded), for example,
to call on queue destroy, where we don't need inlining but
convenience.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Add helpers for implementing .ndo_xdp_xmit().
Same as for XDP_TX, accumulate up to 16 DMA-mapped frames on the stack,
then flush. If DMA mapping is failed for some reason, don't try mapping
further frames, but still flush what was already prepared.
DMA address of a head frame is stored in its headroom, assuming it
has enough of it for an 8 (or 4) byte value.
In addition to @prep and @xmit driver callbacks in XDP_TX, xmit also
needs @finalize to kick the XDPSQ after filling.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Start adding XDP-specific code to libeth, namely handling XDP_TX buffers
(only sending).
The idea is that we accumulate up to 16 buffers on the stack, then,
if either the limit is reached or the polling is finished, flush them
at once with only one XDPSQ cleaning (if needed). The main sending
function will be aware of the sending budget and already have all the
info to send the buffers, so it can't fail.
Drivers need to provide 2 inline callbacks to the main sending function:
for cleaning an XDPSQ and for filling descriptors; the library code
takes care of the rest.
Note that unlike the generic code, multi-buffer support is not wrapped
here with unlikely() to not hurt header split setups.
&libeth_xdp_buff is a simple extension over &xdp_buff which has a direct
pointer to the corresponding Rx descriptor (and, luckily, precisely 1 CL
size and 16-byte alignment on x86_64).
Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> # xmit logic
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Expand libeth's Page Pool functionality by adding native XDP support.
This means picking the appropriate headroom and DMA direction.
Also, register all the created &page_pools as XDP memory models.
A driver then can call xdp_rxq_info_attach_page_pool() when registering
its RxQ info.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Back when the libeth Rx core was initially written, devmem was a draft
and netmem_ref didn't exist in the mainline. Now that it's here, make
libeth MP-agnostic before introducing any new code or any new library
users.
When it's known that the created PP/FQ is for header buffers, use faster
"unsafe" underscored netmem <--> virt accessors as netmem_is_net_iov()
is always false in that case, but consumes some cycles (bit test +
true branch).
Reviewed-by: Mina Almasry <almasrymina@google.com>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
phys_port_id_show, phys_port_name_show and phys_switch_id_show would
return -EOPNOTSUPP if the netdev didn't implement the corresponding
method.
There is no point in creating these files if they are unsupported.
Put these attributes in netdev_phys_group and implement the is_visible
method. make phys_(port_id, port_name, switch_id) invisible if the netdev
dosen't implement the corresponding method.
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250612142707.4644-1-yajun.deng@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
