David Wei says:
====================
io_uring zero copy rx
This patchset contains net/ patches needed by a new io_uring request
implementing zero copy rx into userspace pages, eliminating a kernel
to user copy.
We configure a page pool that a driver uses to fill a hw rx queue to
hand out user pages instead of kernel pages. Any data that ends up
hitting this hw rx queue will thus be dma'd into userspace memory
directly, without needing to be bounced through kernel memory. 'Reading'
data out of a socket instead becomes a _notification_ mechanism, where
the kernel tells userspace where the data is. The overall approach is
similar to the devmem TCP proposal.
This relies on hw header/data split, flow steering and RSS to ensure
packet headers remain in kernel memory and only desired flows hit a hw
rx queue configured for zero copy. Configuring this is outside of the
scope of this patchset.
We share netdev core infra with devmem TCP. The main difference is that
io_uring is used for the uAPI and the lifetime of all objects are bound
to an io_uring instance. Data is 'read' using a new io_uring request
type. When done, data is returned via a new shared refill queue. A zero
copy page pool refills a hw rx queue from this refill queue directly. Of
course, the lifetime of these data buffers are managed by io_uring
rather than the networking stack, with different refcounting rules.
This patchset is the first step adding basic zero copy support. We will
extend this iteratively with new features e.g. dynamically allocated
zero copy areas, THP support, dmabuf support, improved copy fallback,
general optimisations and more.
In terms of netdev support, we're first targeting Broadcom bnxt. Patches
aren't included since Taehee Yoo has already sent a more comprehensive
patchset adding support in [1]. Google gve should already support this,
and Mellanox mlx5 support is WIP pending driver changes.
===========
Performance
===========
Note: Comparison with epoll + TCP_ZEROCOPY_RECEIVE isn't done yet.
Test setup:
* AMD EPYC 9454
* Broadcom BCM957508 200G
* Kernel v6.11 base [2]
* liburing fork [3]
* kperf fork [4]
* 4K MTU
* Single TCP flow
With application thread + net rx softirq pinned to _different_ cores:
+-------------------------------+
| epoll | io_uring |
|-----------|-------------------|
| 82.2 Gbps | 116.2 Gbps (+41%) |
+-------------------------------+
Pinned to _same_ core:
+-------------------------------+
| epoll | io_uring |
|-----------|-------------------|
| 62.6 Gbps | 80.9 Gbps (+29%) |
+-------------------------------+
=====
Links
=====
Broadcom bnxt support:
[1]: https://lore.kernel.org/20241003160620.1521626-8-ap420073@gmail.com
Linux kernel branch including io_uring bits:
[2]: https://github.com/isilence/linux.git zcrx/v13
liburing for testing:
[3]: https://github.com/isilence/liburing.git zcrx/next
kperf for testing:
[4]: https://git.kernel.dk/kperf.git
====================
Link: https://patch.msgid.link/20250204215622.695511-1-dw@davidwei.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Devmem TCP needs a hook in unregister_netdevice_many_notify() to upkeep
the set tracking queues it's bound to, i.e. ->bound_rxqs. Instead of
devmem sticking directly out of the genetic path, add a mp function.
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250204215622.695511-8-dw@davidwei.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add READ_ONCE() around reads of skb->dev->reg_state, because
this field can be changed from other threads/cpus.
Instead of calling dev_kfree_skb_irq() and kfree_skb()
while interrupts are masked and locks held,
use a temporary list and use __skb_queue_purge_reason()
Use SKB_DROP_REASON_DEV_READY drop reason to better
describe why these skbs are dropped.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Link: https://patch.msgid.link/20250204144825.316785-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Drivers should always disable a NAPI instance before removing it.
If they don't the instance may be queued for polling.
Since commit 86e25f40aa ("net: napi: Add napi_config")
we also remove the NAPI from the busy polling hash table
in napi_disable(), so not disabling would leave a stale
entry there.
Use of busy polling is relatively uncommon so bugs may be lurking
in the drivers. Add an explicit warning.
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250203215816.1294081-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After the blamed commits below, some UDP tunnel use dstats for
accounting. On the xmit path, all the UDP-base tunnels ends up
using iptunnel_xmit_stats() for stats accounting, and the latter
assumes the relevant (tunnel) network device uses tstats.
The end result is some 'funny' stat report for the mentioned UDP
tunnel, e.g. when no packet is actually dropped and a bunch of
packets are transmitted:
gnv2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue \
state UNKNOWN mode DEFAULT group default qlen 1000
link/ether ee:7d:09:87:90:ea brd ff:ff:ff:ff:ff:ff
RX: bytes packets errors dropped missed mcast
14916 23 0 15 0 0
TX: bytes packets errors dropped carrier collsns
0 1566 0 0 0 0
Address the issue ensuring the same binary layout for the overlapping
fields of dstats and tstats. While this solution is a bit hackish, is
smaller and with no performance pitfall compared to other alternatives
i.e. supporting both dstat and tstat in iptunnel_xmit_stats() or
reverting the blamed commit.
With time we should possibly move all the IP-based tunnel (and virtual
devices) to dstats.
Fixes: c77200c074 ("bareudp: Handle stats using NETDEV_PCPU_STAT_DSTATS.")
Fixes: 6fa6de3022 ("geneve: Handle stats using NETDEV_PCPU_STAT_DSTATS.")
Fixes: be226352e8 ("vxlan: Handle stats using NETDEV_PCPU_STAT_DSTATS.")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Link: https://patch.msgid.link/2e1c444cf0f63ae472baff29862c4c869be17031.1738432804.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch reverts following changes:
83419b61d1 net: reduce RTNL hold duration in unregister_netdevice_many_notify() (part 2)
ae646f1a0b net: reduce RTNL hold duration in unregister_netdevice_many_notify() (part 1)
cfa579f666 net: no longer hold RTNL while calling flush_all_backlogs()
This caused issues in layers holding a private mutex:
cleanup_net()
rtnl_lock();
mutex_lock(subsystem_mutex);
unregister_netdevice();
rtnl_unlock(); // LOCKDEP violation
rtnl_lock();
I will revisit this in next cycle, opt-in for the new behavior
from safe contexts only.
Fixes: cfa579f666 ("net: no longer hold RTNL while calling flush_all_backlogs()")
Fixes: ae646f1a0b ("net: reduce RTNL hold duration in unregister_netdevice_many_notify() (part 1)")
Fixes: 83419b61d1 ("net: reduce RTNL hold duration in unregister_netdevice_many_notify() (part 2)")
Reported-by: syzbot+5b9196ecf74447172a9a@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/6789d55f.050a0220.20d369.004e.GAE@google.com/
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250129142726.747726-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Device-bound programs are used to support RX metadata kfuncs. These
kfuncs are driver-specific and rely on the driver context to read the
metadata. This means they can't work in generic XDP mode. However, there
is no check to disallow such programs from being attached in generic
mode, in which case the metadata kfuncs will be called in an invalid
context, leading to crashes.
Fix this by adding a check to disallow attaching device-bound programs
in generic mode.
Fixes: 2b3486bc2d ("bpf: Introduce device-bound XDP programs")
Reported-by: Marcus Wichelmann <marcus.wichelmann@hetzner-cloud.de>
Closes: https://lore.kernel.org/r/dae862ec-43b5-41a0-8edf-46c59071cdda@hetzner-cloud.de
Tested-by: Marcus Wichelmann <marcus.wichelmann@hetzner-cloud.de>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://patch.msgid.link/20250127131344.238147-1-toke@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Page ppol tried to cache the NAPI ID in page pool info to avoid
having a dependency on the life cycle of the NAPI instance.
Since commit under Fixes the NAPI ID is not populated until
napi_enable() and there's a good chance that page pool is
created before NAPI gets enabled.
Protect the NAPI pointer with the existing page pool mutex,
the reading path already holds it. napi_id itself we need
to READ_ONCE(), it's protected by netdev_lock() which are
not holding in page pool.
Before this patch napi IDs were missing for mlx5:
# ./cli.py --spec netlink/specs/netdev.yaml --dump page-pool-get
[{'id': 144, 'ifindex': 2, 'inflight': 3072, 'inflight-mem': 12582912},
{'id': 143, 'ifindex': 2, 'inflight': 5568, 'inflight-mem': 22806528},
{'id': 142, 'ifindex': 2, 'inflight': 5120, 'inflight-mem': 20971520},
{'id': 141, 'ifindex': 2, 'inflight': 4992, 'inflight-mem': 20447232},
...
After:
[{'id': 144, 'ifindex': 2, 'inflight': 3072, 'inflight-mem': 12582912,
'napi-id': 565},
{'id': 143, 'ifindex': 2, 'inflight': 4224, 'inflight-mem': 17301504,
'napi-id': 525},
{'id': 142, 'ifindex': 2, 'inflight': 4288, 'inflight-mem': 17563648,
'napi-id': 524},
...
Fixes: 86e25f40aa ("net: napi: Add napi_config")
Reviewed-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20250123231620.1086401-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Record the pending configuration in net_device struct.
ethtool core duplicates the current config and the specific
handlers (for now just ringparam) can modify it.
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Link: https://patch.msgid.link/20250119020518.1962249-4-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Separate the HDS config from the ethtool state struct.
The HDS config contains just simple parameters, not state.
Having it as a separate struct will make it easier to clone / copy
and also long term potentially make it per-queue.
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Link: https://patch.msgid.link/20250119020518.1962249-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch addresses issues with filter counting in block (tcf_block),
particularly for software bypass scenarios, by introducing a more
accurate mechanism using useswcnt.
Previously, filtercnt and skipswcnt were introduced by:
Commit 2081fd3445 ("net: sched: cls_api: add filter counter") and
Commit f631ef39d8 ("net: sched: cls_api: add skip_sw counter")
filtercnt tracked all tp (tcf_proto) objects added to a block, and
skipswcnt counted tp objects with the skipsw attribute set.
The problem is: a single tp can contain multiple filters, some with skipsw
and others without. The current implementation fails in the case:
When the first filter in a tp has skipsw, both skipswcnt and filtercnt
are incremented, then adding a second filter without skipsw to the same
tp does not modify these counters because tp->counted is already set.
This results in bypass software behavior based solely on skipswcnt
equaling filtercnt, even when the block includes filters without
skipsw. Consequently, filters without skipsw are inadvertently bypassed.
To address this, the patch introduces useswcnt in block to explicitly count
tp objects containing at least one filter without skipsw. Key changes
include:
Whenever a filter without skipsw is added, its tp is marked with usesw
and counted in useswcnt. tc_run() now uses useswcnt to determine software
bypass, eliminating reliance on filtercnt and skipswcnt.
This refined approach prevents software bypass for blocks containing
mixed filters, ensuring correct behavior in tc_run().
Additionally, as atomic operations on useswcnt ensure thread safety and
tp->lock guards access to tp->usesw and tp->counted, the broader lock
down_write(&block->cb_lock) is no longer required in tc_new_tfilter(),
and this resolves a performance regression caused by the filter counting
mechanism during parallel filter insertions.
The improvement can be demonstrated using the following script:
# cat insert_tc_rules.sh
tc qdisc add dev ens1f0np0 ingress
for i in $(seq 16); do
taskset -c $i tc -b rules_$i.txt &
done
wait
Each of rules_$i.txt files above includes 100000 tc filter rules to a
mlx5 driver NIC ens1f0np0.
Without this patch:
# time sh insert_tc_rules.sh
real 0m50.780s
user 0m23.556s
sys 4m13.032s
With this patch:
# time sh insert_tc_rules.sh
real 0m17.718s
user 0m7.807s
sys 3m45.050s
Fixes: 047f340b36 ("net: sched: make skip_sw actually skip software")
Reported-by: Shuang Li <shuali@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Reviewed-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
After 1b23cdbd2b ("net: protect netdev->napi_list with netdev_lock()")
it makes sense to iterate through dev->napi_list while holding
the device lock.
Also call synchronize_net() at most one time.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250117232113.1612899-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Basically, dev_ifsioc() operates on the passed single netns (except
for netdev notifier chains with lower/upper devices for which we will
need more changes).
Let's hold rtnl_net_lock() for dev_ifsioc().
Now that NETDEV_CHANGENAME is always triggered under rtnl_net_lock()
of the device's netns. (do_setlink() and dev_ifsioc())
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250115095545.52709-4-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
devnet_rename_sem is no longer used since commit
0840556e5a ("net: Protect dev->name by seqlock.").
Also, RTNL serialises dev_change_name().
Let's remove devnet_rename_sem.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250115095545.52709-3-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The cited commit forgot to add netdev_rename_lock in one of the
error paths in dev_change_name().
Let's hold netdev_rename_lock before restoring the old dev->name.
Fixes: 0840556e5a ("net: Protect dev->name by seqlock.")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250115095545.52709-2-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet says:
====================
net: reduce RTNL pressure in unregister_netdevice()
One major source of RTNL contention resides in unregister_netdevice()
Due to RCU protection of various network structures, and
unregister_netdevice() being a synchronous function,
it is calling potentially slow functions while holding RTNL.
I think we can release RTNL in two points, so that three
slow functions are called while RTNL can be used
by other threads.
v1: https://lore.kernel.org/netdev/20250107130906.098fc8d6@kernel.org/T/#m398c95f5778e1ff70938e079d3c4c43c050ad2a6
====================
Link: https://patch.msgid.link/20250114205531.967841-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
One synchronize_net() call is currently done while holding RTNL.
This is source of RTNL contention in workloads adding and deleting
many network namespaces per second, because synchronize_rcu()
and synchronize_rcu_expedited() can use 60+ ms in some cases.
For cleanup_net() use, temporarily release RTNL
while calling the last synchronize_net().
This should be safe, because devices are no longer visible
to other threads after unlist_netdevice() call
and setting dev->reg_state to NETREG_UNREGISTERING.
In any case, the new netdev_lock() / netdev_unlock()
infrastructure that we are adding should allow
to fix potential issues, with a combination
of a per-device mutex and dev->reg_state awareness.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jesse Brandeburg <jbrandeburg@cloudflare.com>
Link: https://patch.msgid.link/20250114205531.967841-6-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Two synchronize_net() calls are currently done while holding RTNL.
This is source of RTNL contention in workloads adding and deleting
many network namespaces per second, because synchronize_rcu()
and synchronize_rcu_expedited() can use 60+ ms in some cases.
For cleanup_net() use, temporarily release RTNL
while calling the last synchronize_net().
This should be safe, because devices are no longer visible
to other threads at this point.
In any case, the new netdev_lock() / netdev_unlock()
infrastructure that we are adding should allow
to fix potential issues, with a combination
of a per-device mutex and dev->reg_state awareness.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jesse Brandeburg <jbrandeburg@cloudflare.com>
Link: https://patch.msgid.link/20250114205531.967841-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
flush_all_backlogs() is called from unregister_netdevice_many_notify()
as part of netdevice dismantles.
This is currently called under RTNL, and can last up to 50 ms
on busy hosts.
There is no reason to hold RTNL at this stage, if our caller
is cleanup_net() : netns are no more visible, devices
are in NETREG_UNREGISTERING state and no other thread
could mess our state while RTNL is temporarily released.
In order to provide isolation, this patch provides a separate
'net_todo_list' for cleanup_net().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jesse Brandeburg <jbrandeburg@cloudflare.com>
Link: https://patch.msgid.link/20250114205531.967841-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
flush_all_backlogs() uses per-cpu and static data to hold its
temporary data, on the assumption it is called under RTNL
protection.
Following patch in the series will break this assumption.
Use instead a dynamically allocated piece of memory.
In the unlikely case the allocation fails,
use a boot-time allocated memory.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jesse Brandeburg <jbrandeburg@cloudflare.com>
Link: https://patch.msgid.link/20250114205531.967841-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
cleanup_net() is the single thread responsible
for netns dismantles, and a serious bottleneck.
Before we can get per-netns RTNL, make sure
all synchronize_net() called from this thread
are using rcu_synchronize_expedited().
v3: deal with CONFIG_NET_NS=n
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jesse Brandeburg <jbrandeburg@cloudflare.com>
Link: https://patch.msgid.link/20250114205531.967841-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Take netdev_lock() in netif_napi_set_irq(). All NAPI "control fields"
are now protected by that lock (most of the other ones are set during
napi add/del). The napi_hash_node is fully protected by the hash
spin lock, but close enough for the kdoc...
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-10-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Now that NAPI instances can't come and go without holding
netdev->lock we can trivially switch from rtnl_lock() to
netdev_lock() for setting netdev->threaded via sysfs.
Note that since we do not lock netdev_lock around sysfs
calls in the core we don't have to "trylock" like we do
with rtnl_lock.
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-9-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In prep for dropping rtnl_lock, start locking netdev->lock in netlink
genl ops. We need to be using netdev->up instead of flags & IFF_UP.
We can remove the RCU lock protection for the NAPI since NAPI list
is protected by netdev->lock already.
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-8-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Wrap napi_enable() / napi_disable() with netdev_lock().
Provide the "already locked" flavor of the API.
iavf needs the usual adjustment. A number of drivers call
napi_enable() under a spin lock, so they have to be modified
to take netdev_lock() first, then spin lock then call
napi_enable_locked().
Protecting napi_enable() implies that napi->napi_id is protected
by netdev_lock().
Acked-by: Francois Romieu <romieu@fr.zoreil.com> # via-velocity
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-7-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Hold netdev->lock when NAPIs are getting added or removed.
This will allow safe access to NAPI instances of a net_device
without rtnl_lock.
Create a family of helpers which assume the lock is already taken.
Switch iavf to them, as it makes extensive use of netdev->lock,
already.
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-6-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Some uAPI (netdev netlink) hide net_device's sub-objects while
the interface is down to ensure uniform behavior across drivers.
To remove the rtnl_lock dependency from those uAPIs we need a way
to safely tell if the device is down or up.
Add an indication of whether device is open or closed, protected
by netdev->lock. The semantics are the same as IFF_UP, but taking
netdev_lock around every write to ->flags would be a lot of code
churn.
We don't want to blanket the entire open / close path by netdev_lock,
because it will prevent us from applying it to specific structures -
core helpers won't be able to take that lock from any function
called by the drivers on open/close paths.
So the state of the flag is "pessimistic", as in it may report false
negatives, but never false positives.
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-5-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add helpers for accessing netdevs under netdev_lock().
There's some careful handling needed to find the device and lock it
safely, without it getting unregistered, and without taking rtnl_lock
(the latter being the whole point of the new locking, after all).
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-4-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Protect writes to netdev->reg_state with netdev_lock().
From now on holding netdev_lock() is sufficient to prevent
the net_device from getting unregistered, so code which
wants to hold just a single netdev around no longer needs
to hold rtnl_lock.
We do not protect the NETREG_UNREGISTERED -> NETREG_RELEASED
transition. We'd need to move mutex_destroy(netdev->lock)
to .release, but the real reason is that trying to stop
the unregistration process mid-way would be unsafe / crazy.
Taking references on such devices is not safe, either.
So the intended semantics are to lock REGISTERED devices.
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250115035319.559603-3-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When a single buffer XDP is attached, NIC should guarantee only single
page packets will be received.
tcp-data-split feature splits packets into header and payload. single
buffer XDP can't handle it properly.
So attaching single buffer XDP should be disallowed when tcp-data-split
is enabled.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Link: https://patch.msgid.link/20250114142852.3364986-6-ap420073@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When tcp-data-split is UNKNOWN mode, drivers arbitrarily handle it.
For example, bnxt_en driver automatically enables if at least one of
LRO/GRO/JUMBO is enabled.
If tcp-data-split is UNKNOWN and LRO is enabled, a driver returns
ENABLES of tcp-data-split, not UNKNOWN.
So, `ethtool -g eth0` shows tcp-data-split is enabled.
The problem is in the setting situation.
In the ethnl_set_rings(), it first calls get_ringparam() to get the
current driver's config.
At that moment, if driver's tcp-data-split config is UNKNOWN, it returns
ENABLE if LRO/GRO/JUMBO is enabled.
Then, it sets values from the user and driver's current config to
kernel_ethtool_ringparam.
Last it calls .set_ringparam().
The driver, especially bnxt_en driver receives
ETHTOOL_TCP_DATA_SPLIT_ENABLED.
But it can't distinguish whether it is set by the user or just the
current config.
When user updates ring parameter, the new hds_config value is updated
and current hds_config value is stored to old_hdsconfig.
Driver's .set_ringparam() callback can distinguish a passed
tcp-data-split value is came from user explicitly.
If .set_ringparam() is failed, hds_config is rollbacked immediately.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Link: https://patch.msgid.link/20250114142852.3364986-2-ap420073@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
init_dummy_netdev_core() used to cater to net_devices which
did not come from alloc_netdev_mqs(). Since that's no longer
supported remove the init logic which duplicates alloc_netdev_mqs().
While at it rename back to init_dummy_netdev().
Reviewed-by: Joe Damato <jdamato@fastly.com>
Link: https://patch.msgid.link/20250113003456.3904110-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
init_dummy_netdev() can initialize statically declared or embedded
net_devices. Such netdevs did not come from alloc_netdev_mqs().
After recent work by Breno, there are the only two cases where
we have do that.
Switch those cases to alloc_netdev_mqs() and delete init_dummy_netdev().
Dealing with static netdevs is not worth the maintenance burden.
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Link: https://patch.msgid.link/20250113003456.3904110-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are no module callers of dev_get_by_napi_id(),
and commit d1cacd7477 ("netdev: prevent accessing NAPI instances
from another namespace") proves that getting NAPI by id
needs to be done with care. So hide dev_get_by_napi_id().
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Link: https://patch.msgid.link/20250110004924.3212260-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Netlink code depends on NAPI instances being sorted by ID on
the netdev list for dump continuation. We need to be able to
find the position on the list where we left off if dump does
not fit in a single skb, and in the meantime NAPI instances
can come and go.
This was trivially true when we were assigning a new ID to every
new NAPI instance. Since we added the NAPI config API, we try
to retain the ID previously used for the same queue, but still
add the new NAPI instance at the start of the list.
This is fine if we reset the entire netdev and all NAPIs get
removed and added back. If driver replaces a NAPI instance
during an operation like DEVMEM queue reset, or recreates
a subset of NAPI instances in other ways we may end up with
broken ordering, and therefore Netlink dumps with either
missing or duplicated entries.
At this stage the problem is theoretical. Only two drivers
support queue API, bnxt and gve. gve recreates NAPIs during
queue reset, but it doesn't support NAPI config.
bnxt supports NAPI config but doesn't recreate instances
during reset.
We need to save the ID in the config as soon as it is assigned
because otherwise the new NAPI will not know what ID it will
get at enable time, at the time it is being added.
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In commit 66e4c8d950 ("net: warn if transport header was not set")
I added a debug check in skb_transport_header() to detect
if a caller expects the transport_header to be set to a meaningful
value by a prior code path.
Unfortunately, __netif_receive_skb_core() resets the transport header
to the same value than the network header, defeating this check
in receive paths.
Pretending the transport and network headers are the same
is usually wrong.
This patch removes this reset for CONFIG_DEBUG_NET=y builds
to let fuzzers and CI find bugs.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250107144342.499759-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The NAPI IDs were not fully exposed to user space prior to the netlink
API, so they were never namespaced. The netlink API must ensure that
at the very least NAPI instance belongs to the same netns as the owner
of the genl sock.
napi_by_id() can become static now, but it needs to move because of
dev_get_by_napi_id().
Cc: stable@vger.kernel.org
Fixes: 1287c1ae0f ("netdev-genl: Support setting per-NAPI config values")
Fixes: 27f91aaf49 ("netdev-genl: Add netlink framework functions for napi")
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Link: https://patch.msgid.link/20250106180137.1861472-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(un)?register_netdevice_notifier_dev_net() hold RTNL before triggering
the notifier for all netdev in the netns.
Let's convert the RTNL to rtnl_net_lock().
Note that move_netdevice_notifiers_dev_net() is assumed to be (but not
yet) protected by per-netns RTNL of both src and dst netns; we need to
convert wireless and hyperv drivers that call dev_change_net_namespace().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250106070751.63146-4-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(un)?register_netdevice_notifier_net() hold RTNL before triggering the
notifier for all netdev in the netns.
Let's convert the RTNL to rtnl_net_lock().
Note that the per-netns netdev notifier is protected by per-netns RTNL.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250106070751.63146-3-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In commit d7811e623d ("[NET]: Drop tx lock in dev_watchdog_up")
dev_watchdog_up() became a simple wrapper for __netdev_watchdog_up()
Herbert also said : "In 2.6.19 we can eliminate the unnecessary
__dev_watchdog_up and replace it with dev_watchdog_up."
This patch consolidates things to have only two functions, with
a common prefix.
- netdev_watchdog_up(), exported for the sake of one freescale driver.
This replaces __netdev_watchdog_up() and dev_watchdog_up().
- netdev_watchdog_down(), static to net/sched/sch_generic.c
This replaces dev_watchdog_down().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Link: https://patch.msgid.link/20250105090924.1661822-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Let's hold per-netns RTNL of dev_net(dev) in register_netdev()
and unregister_netdev().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The blamed commit disabled hardware offoad of IPv6 packets with
extension headers on devices that advertise NETIF_F_IPV6_CSUM,
based on the definition of that feature in skbuff.h:
* * - %NETIF_F_IPV6_CSUM
* - Driver (device) is only able to checksum plain
* TCP or UDP packets over IPv6. These are specifically
* unencapsulated packets of the form IPv6|TCP or
* IPv6|UDP where the Next Header field in the IPv6
* header is either TCP or UDP. IPv6 extension headers
* are not supported with this feature. This feature
* cannot be set in features for a device with
* NETIF_F_HW_CSUM also set. This feature is being
* DEPRECATED (see below).
The change causes skb_warn_bad_offload to fire for BIG TCP
packets.
[ 496.310233] WARNING: CPU: 13 PID: 23472 at net/core/dev.c:3129 skb_warn_bad_offload+0xc4/0xe0
[ 496.310297] ? skb_warn_bad_offload+0xc4/0xe0
[ 496.310300] skb_checksum_help+0x129/0x1f0
[ 496.310303] skb_csum_hwoffload_help+0x150/0x1b0
[ 496.310306] validate_xmit_skb+0x159/0x270
[ 496.310309] validate_xmit_skb_list+0x41/0x70
[ 496.310312] sch_direct_xmit+0x5c/0x250
[ 496.310317] __qdisc_run+0x388/0x620
BIG TCP introduced an IPV6_TLV_JUMBO IPv6 extension header to
communicate packet length, as this is an IPv6 jumbogram. But, the
feature is only enabled on devices that support BIG TCP TSO. The
header is only present for PF_PACKET taps like tcpdump, and not
transmitted by physical devices.
For this specific case of extension headers that are not
transmitted, return to the situation before the blamed commit
and support hardware offload.
ipv6_has_hopopt_jumbo() tests not only whether this header is present,
but also that it is the only extension header before a terminal (L4)
header.
Fixes: 04c20a9356 ("net: skip offload for NETIF_F_IPV6_CSUM if ipv6 header contains extension")
Reported-by: syzbot <syzkaller@googlegroups.com>
Reported-by: Eric Dumazet <edumazet@google.com>
Closes: https://lore.kernel.org/netdev/CANn89iK1hdC3Nt8KPhOtTF8vCPc1AHDCtse_BTNki1pWxAByTQ@mail.gmail.com/
Signed-off-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250101164909.1331680-1-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
To make the system page pool usable as a source for allocating XDP
frames, we need to register it with xdp_reg_mem_model(), so that page
return works correctly. This is done in preparation for using the system
page_pool to convert XDP_PASS XSk frames to skbs; for the same reason,
make the per-cpu variable non-static so we can access it from other
source files as well (but w/o exporting).
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://patch.msgid.link/20241203173733.3181246-7-aleksander.lobakin@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In lots of places, bpf_prog pointer is used only for tracing or other
stuff that doesn't modify the structure itself. Same for net_device.
Address at least some of them and add `const` attributes there. The
object code didn't change, but that may prevent unwanted data
modifications and also allow more helpers to have const arguments.
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Make napi_hash_lock IRQ safe. It is used during the control path, and is
taken and released in napi_hash_add and napi_hash_del, which will
typically be called by calls to napi_enable and napi_disable.
This change avoids a deadlock in pcnet32 (and other any other drivers
which follow the same pattern):
CPU 0:
pcnet32_open
spin_lock_irqsave(&lp->lock, ...)
napi_enable
napi_hash_add <- before this executes, CPU 1 proceeds
spin_lock(napi_hash_lock)
[...]
spin_unlock_irqrestore(&lp->lock, flags);
CPU 1:
pcnet32_close
napi_disable
napi_hash_del
spin_lock(napi_hash_lock)
< INTERRUPT >
pcnet32_interrupt
spin_lock(lp->lock) <- DEADLOCK
Changing the napi_hash_lock to be IRQ safe prevents the IRQ from firing
on CPU 1 until napi_hash_lock is released, preventing the deadlock.
Cc: stable@vger.kernel.org
Fixes: 86e25f40aa ("net: napi: Add napi_config")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Closes: https://lore.kernel.org/netdev/85dd4590-ea6b-427d-876a-1d8559c7ad82@roeck-us.net/
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Joe Damato <jdamato@fastly.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241202182103.363038-1-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The napi_suspend_irqs routine bootstraps irq suspension by elongating
the defer timeout to irq_suspend_timeout.
The napi_resume_irqs routine effectively cancels irq suspension by
forcing the napi to be scheduled immediately.
Signed-off-by: Martin Karsten <mkarsten@uwaterloo.ca>
Co-developed-by: Joe Damato <jdamato@fastly.com>
Signed-off-by: Joe Damato <jdamato@fastly.com>
Tested-by: Joe Damato <jdamato@fastly.com>
Tested-by: Martin Karsten <mkarsten@uwaterloo.ca>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Link: https://patch.msgid.link/20241109050245.191288-3-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a per-NAPI IRQ suspension parameter, which can be get/set with
netdev-genl.
This patch doesn't change any behavior but prepares the code for other
changes in the following commits which use irq_suspend_timeout as a
timeout for IRQ suspension.
Signed-off-by: Martin Karsten <mkarsten@uwaterloo.ca>
Co-developed-by: Joe Damato <jdamato@fastly.com>
Signed-off-by: Joe Damato <jdamato@fastly.com>
Tested-by: Joe Damato <jdamato@fastly.com>
Tested-by: Martin Karsten <mkarsten@uwaterloo.ca>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Link: https://patch.msgid.link/20241109050245.191288-2-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
As documented in skbuff.h, devices with NETIF_F_IPV6_CSUM capability
can only checksum TCP and UDP over IPv6 if the IP header does not
contains extension.
This is enforced for UDP packets emitted from user-space to an IPv6
address as they go through ip6_make_skb(), which calls
__ip6_append_data() where a check is done on the header size before
setting CHECKSUM_PARTIAL.
But the introduction of UDP encapsulation with fou6 added a code-path
where it is possible to get an skb with a partial UDP checksum and an
IPv6 header with extension:
* fou6 adds a UDP header with a partial checksum if the inner packet
does not contains a valid checksum.
* ip6_tunnel adds an IPv6 header with a destination option extension
header if encap_limit is non-zero (the default value is 4).
The thread linked below describes in more details how to reproduce the
problem with GRE-in-UDP tunnel.
Add a check on the network header size in skb_csum_hwoffload_help() to
make sure no IPv6 packet with extension header is handed to a network
device with NETIF_F_IPV6_CSUM capability.
Link: https://lore.kernel.org/netdev/26548921.1r3eYUQgxm@benoit.monin/T/#u
Fixes: aa3463d65e ("fou: Add encap ops for IPv6 tunnels")
Signed-off-by: Benoît Monin <benoit.monin@gmx.fr>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/5fbeecfc311ea182aa1d1c771725ab8b4cac515e.1729778144.git.benoit.monin@gmx.fr
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a persistent NAPI config area for NAPI configuration to the core.
Drivers opt-in to setting the persistent config for a NAPI by passing an
index when calling netif_napi_add_config.
napi_config is allocated in alloc_netdev_mqs, freed in free_netdev
(after the NAPIs are deleted).
Drivers which call netif_napi_add_config will have persistent per-NAPI
settings: NAPI IDs, gro_flush_timeout, and defer_hard_irq settings.
Per-NAPI settings are saved in napi_disable and restored in napi_enable.
Co-developed-by: Martin Karsten <mkarsten@uwaterloo.ca>
Signed-off-by: Martin Karsten <mkarsten@uwaterloo.ca>
Signed-off-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241011184527.16393-6-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Allow per-NAPI gro_flush_timeout setting.
The existing sysfs parameter is respected; writes to sysfs will write to
all NAPI structs for the device and the net_device gro_flush_timeout
field. Reads from sysfs will read from the net_device field.
The ability to set gro_flush_timeout on specific NAPI instances will be
added in a later commit, via netdev-genl.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20241011184527.16393-4-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add defer_hard_irqs to napi_struct in preparation for per-NAPI
settings.
The existing sysfs parameter is respected; writes to sysfs will write to
all NAPI structs for the device and the net_device defer_hard_irq field.
Reads from sysfs show the net_device field.
The ability to set defer_hard_irqs on specific NAPI instances will be
added in a later commit, via netdev-genl.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20241011184527.16393-2-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
hook into netif_set_real_num_tx_queues() to cleanup any shaper
configured on top of the to-be-destroyed TX queues.
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Link: https://patch.msgid.link/6da4ee03cae2b2a757d7b59e88baf09cc94c5ef1.1728460186.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Introduce the basic infrastructure to implement the net-shaper
core functionality. Each network devices carries a net-shaper cache,
the NL get() operation fetches the data from such cache.
The cache is initially empty, will be fill by the set()/group()
operation implemented later and is destroyed at device cleanup time.
The net_shaper_fill_handle(), net_shaper_ctx_init(), and
net_shaper_generic_pre() implementations handle generic index type
attributes, despite the current caller always pass a constant value
to avoid more noise in later patches using them with different
attributes.
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Link: https://patch.msgid.link/ddd10fd645a9367803ad02fca4a5664ea5ace170.1728460186.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
One path takes care of SKB_GSO_DODGY, assuming
skb->len is bigger than hdr_len.
virtio_net_hdr_to_skb() does not fully dissect TCP headers,
it only make sure it is at least 20 bytes.
It is possible for an user to provide a malicious 'GSO' packet,
total length of 80 bytes.
- 20 bytes of IPv4 header
- 60 bytes TCP header
- a small gso_size like 8
virtio_net_hdr_to_skb() would declare this packet as a normal
GSO packet, because it would see 40 bytes of payload,
bigger than gso_size.
We need to make detect this case to not underflow
qdisc_skb_cb(skb)->pkt_len.
Fixes: 1def9238d4 ("net_sched: more precise pkt_len computation")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Commit 24ab059d2e ("net: check dev->gso_max_size in gso_features_check()")
added a dev->gso_max_size test to gso_features_check() in order to fall
back to GSO when needed.
This was added as it was noticed that some drivers could misbehave if TSO
packets get too big. However, the check doesn't respect dev->gso_ipv4_max_size
limit. For instance, a device could be configured with BIG TCP for IPv4,
but not IPv6.
Therefore, add a netif_get_gso_max_size() equivalent to netif_get_gro_max_size()
and use the helper to respect both limits before falling back to GSO engine.
Fixes: 24ab059d2e ("net: check dev->gso_max_size in gso_features_check()")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20240923212242.15669-2-daniel@iogearbox.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
- Core:
- Remove a global lock in the affinity setting code
The lock protects a cpumask for intermediate results and the lock
causes a bottleneck on simultaneous start of multiple virtual
machines. Replace the lock and the static cpumask with a per CPU
cpumask which is nicely serialized by raw spinlock held when
executing this code.
- Provide support for giving a suffix to interrupt domain names.
That's required to support devices with subfunctions so that the
domain names are distinct even if they originate from the same
device node.
- The usual set of cleanups and enhancements all over the place
- Drivers:
- Support for longarch AVEC interrupt chip
- Refurbishment of the Armada driver so it can be extended for new
variants.
- The usual set of cleanups and enhancements all over the place
-----BEGIN PGP SIGNATURE-----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=JlJ2
-----END PGP SIGNATURE-----
Merge tag 'irq-core-2024-09-16' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull irq updates from Thomas Gleixner:
"Core:
- Remove a global lock in the affinity setting code
The lock protects a cpumask for intermediate results and the lock
causes a bottleneck on simultaneous start of multiple virtual
machines. Replace the lock and the static cpumask with a per CPU
cpumask which is nicely serialized by raw spinlock held when
executing this code.
- Provide support for giving a suffix to interrupt domain names.
That's required to support devices with subfunctions so that the
domain names are distinct even if they originate from the same
device node.
- The usual set of cleanups and enhancements all over the place
Drivers:
- Support for longarch AVEC interrupt chip
- Refurbishment of the Armada driver so it can be extended for new
variants.
- The usual set of cleanups and enhancements all over the place"
* tag 'irq-core-2024-09-16' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (73 commits)
genirq: Use cpumask_intersects()
genirq/cpuhotplug: Use cpumask_intersects()
irqchip/apple-aic: Only access system registers on SoCs which provide them
irqchip/apple-aic: Add a new "Global fast IPIs only" feature level
irqchip/apple-aic: Skip unnecessary enabling of use_fast_ipi
dt-bindings: apple,aic: Document A7-A11 compatibles
irqdomain: Use IS_ERR_OR_NULL() in irq_domain_trim_hierarchy()
genirq/msi: Use kmemdup_array() instead of kmemdup()
genirq/proc: Change the return value for set affinity permission error
genirq/proc: Use irq_move_pending() in show_irq_affinity()
genirq/proc: Correctly set file permissions for affinity control files
genirq: Get rid of global lock in irq_do_set_affinity()
genirq: Fix typo in struct comment
irqchip/loongarch-avec: Add AVEC irqchip support
irqchip/loongson-pch-msi: Prepare get_pch_msi_handle() for AVECINTC
irqchip/loongson-eiointc: Rename CPUHP_AP_IRQ_LOONGARCH_STARTING
LoongArch: Architectural preparation for AVEC irqchip
LoongArch: Move irqchip function prototypes to irq-loongson.h
irqchip/loongson-pch-msi: Switch to MSI parent domains
softirq: Remove unused 'action' parameter from action callback
...
For device memory TCP, we expect the skb headers to be available in host
memory for access, and we expect the skb frags to be in device memory
and unaccessible to the host. We expect there to be no mixing and
matching of device memory frags (unaccessible) with host memory frags
(accessible) in the same skb.
Add a skb->devmem flag which indicates whether the frags in this skb
are device memory frags or not.
__skb_fill_netmem_desc() now checks frags added to skbs for net_iov,
and marks the skb as skb->devmem accordingly.
Add checks through the network stack to avoid accessing the frags of
devmem skbs and avoid coalescing devmem skbs with non devmem skbs.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Kaiyuan Zhang <kaiyuanz@google.com>
Signed-off-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20240910171458.219195-9-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Make skb_frag_page() fail in the case where the frag is not backed
by a page, and fix its relevant callers to handle this case.
Signed-off-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20240910171458.219195-8-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a netdev_dmabuf_binding struct which represents the
dma-buf-to-netdevice binding. The netlink API will bind the dma-buf to
rx queues on the netdevice. On the binding, the dma_buf_attach
& dma_buf_map_attachment will occur. The entries in the sg_table from
mapping will be inserted into a genpool to make it ready
for allocation.
The chunks in the genpool are owned by a dmabuf_chunk_owner struct which
holds the dma-buf offset of the base of the chunk and the dma_addr of
the chunk. Both are needed to use allocations that come from this chunk.
We create a new type that represents an allocation from the genpool:
net_iov. We setup the net_iov allocation size in the
genpool to PAGE_SIZE for simplicity: to match the PAGE_SIZE normally
allocated by the page pool and given to the drivers.
The user can unbind the dmabuf from the netdevice by closing the netlink
socket that established the binding. We do this so that the binding is
automatically unbound even if the userspace process crashes.
The binding and unbinding leaves an indicator in struct netdev_rx_queue
that the given queue is bound, and the binding is actuated by resetting
the rx queue using the queue API.
The netdev_dmabuf_binding struct is refcounted, and releases its
resources only when all the refs are released.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Kaiyuan Zhang <kaiyuanz@google.com>
Signed-off-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com> # excluding netlink
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20240910171458.219195-4-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
dev_pick_tx_cpu_id() has been introduced with two users by
commit a4ea8a3dac ("net: Add generic ndo_select_queue functions").
The use in AF_PACKET has been removed in 2019 by
commit b71b5837f8 ("packet: rework packet_pick_tx_queue() to use common code selection")
The other user was a Netlogic XLP driver, removed in 2021 by
commit 47ac6f567c ("staging: Remove Netlogic XLP network driver").
It's relatively unlikely that any modern driver will need an
.ndo_select_queue implementation which picks purely based on CPU ID
and skips XPS, delete dev_pick_tx_cpu_id()
Found by code inspection.
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20240906161059.715546-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
list_head can be initialized automatically with LIST_HEAD()
instead of calling INIT_LIST_HEAD(). Here we can simplify
the code.
Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Link: https://patch.msgid.link/20240904093243.3345012-6-lihongbo22@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
"Interface can't change network namespaces" is rather an attribute,
not a feature, and it can't be changed via Ethtool.
Make it a "cold" private flag instead of a netdev_feature and free
one more bit.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Make dev->priv_flags `u32` back and define bits higher than 31 as
bitfield booleans as per Jakub's suggestion. This simplifies code
which accesses these bits with no optimization loss (testb both
before/after), allows to not extend &netdev_priv_flags each time,
but also scales better as bits > 63 in the future would only add
a new u64 to the structure with no complications, comparing to
that extending ::priv_flags would require converting it to a bitmap.
Note that I picked `unsigned long :1` to not lose any potential
optimizations comparing to `bool :1` etc.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The deprecated helper strcpy() performs no bounds checking on the
destination buffer. This could result in linear overflows beyond
the end of the buffer, leading to all kinds of misbehaviors.
The safe replacement is strscpy() [1].
Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strcpy [1]
Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----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=O8tu
-----END PGP SIGNATURE-----
Merge tag 'nf-next-24-08-23' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next
Pablo Neira Ayuso says:
====================
Netfilter updates for net-next
The following batch contains Netfilter updates for net-next:
Patch #1 fix checksum calculation in nfnetlink_queue with SCTP,
segment GSO packet since skb_zerocopy() does not support
GSO_BY_FRAGS, from Antonio Ojea.
Patch #2 extend nfnetlink_queue coverage to handle SCTP packets,
from Antonio Ojea.
Patch #3 uses consume_skb() instead of kfree_skb() in nfnetlink,
from Donald Hunter.
Patch #4 adds a dedicate commit list for sets to speed up
intra-transaction lookups, from Florian Westphal.
Patch #5 skips removal of element from abort path for the pipapo
backend, ditching the shadow copy of this datastructure
is sufficient.
Patch #6 moves nf_ct_netns_get() out of nf_conncount_init() to
let users of conncoiunt decide when to enable conntrack,
this is needed by openvswitch, from Xin Long.
Patch #7 pass context to all nft_parse_register_load() in
preparation for the next patch.
Patches #8 and #9 reject loads from uninitialized registers from
control plane to remove register initialization from
datapath. From Florian Westphal.
* tag 'nf-next-24-08-23' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next:
netfilter: nf_tables: don't initialize registers in nft_do_chain()
netfilter: nf_tables: allow loads only when register is initialized
netfilter: nf_tables: pass context structure to nft_parse_register_load
netfilter: move nf_ct_netns_get out of nf_conncount_init
netfilter: nf_tables: do not remove elements if set backend implements .abort
netfilter: nf_tables: store new sets in dedicated list
netfilter: nfnetlink: convert kfree_skb to consume_skb
selftests: netfilter: nft_queue.sh: sctp coverage
netfilter: nfnetlink_queue: unbreak SCTP traffic
====================
Link: https://patch.msgid.link/20240822221939.157858-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When net devices propagate xdp configurations to slave devices,
we will need to perform a memory provider check to ensure we're
not binding xdp to a device using unreadable netmem.
Currently the ->ndo_bpf calls in a few places. Adding checks to all
these places would not be ideal.
Refactor all the ->ndo_bpf calls into one place where we can add this
check in the future.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Mina Almasry <almasrymina@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Link topologies containing multiple network PHYs attached to the same
net_device can be found when using a PHY as a media converter for use
with an SFP connector, on which an SFP transceiver containing a PHY can
be used.
With the current model, the transceiver's PHY can't be used for
operations such as cable testing, timestamping, macsec offload, etc.
The reason being that most of the logic for these configuration, coming
from either ethtool netlink or ioctls tend to use netdev->phydev, which
in multi-phy systems will reference the PHY closest to the MAC.
Introduce a numbering scheme allowing to enumerate PHY devices that
belong to any netdev, which can in turn allow userspace to take more
precise decisions with regard to each PHY's configuration.
The numbering is maintained per-netdev, in a phy_device_list.
The numbering works similarly to a netdevice's ifindex, with
identifiers that are only recycled once INT_MAX has been reached.
This prevents races that could occur between PHY listing and SFP
transceiver removal/insertion.
The identifiers are assigned at phy_attach time, as the numbering
depends on the netdevice the phy is attached to. The PHY index can be
re-used for PHYs that are persistent.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Tested-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
When soft interrupt actions are called, they are passed a pointer to the
struct softirq action which contains the action's function pointer.
This pointer isn't useful, as the action callback already knows what
function it is. And since each callback handles a specific soft interrupt,
the callback also knows which soft interrupt number is running.
No soft interrupt action callback actually uses this parameter, so remove
it from the function pointer signature. This clarifies that soft interrupt
actions are global routines and makes it slightly cheaper to call them.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Link: https://lore.kernel.org/all/20240815171549.3260003-1-csander@purestorage.com
when packet is enqueued with nfqueue and GSO is enabled, checksum
calculation has to take into account the protocol, as SCTP uses a
32 bits CRC checksum.
Enter skb_gso_segment() path in case of SCTP GSO packets because
skb_zerocopy() does not support for GSO_BY_FRAGS.
Joint work with Pablo.
Signed-off-by: Antonio Ojea <aojea@google.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Add missing __percpu qualifier to a (void *) cast to fix
dev.c:10863:45: warning: cast removes address space '__percpu' of expression
sparse warning. Also remove now unneeded __force sparse directives.
Found by GCC's named address space checks.
There were no changes in the resulting object file.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Link: https://patch.msgid.link/20240814070748.943671-1-ubizjak@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
UDP segmentation offload inherently depends on checksum offload. It should
not be possible to disable checksum offload while leaving USO enabled.
Enforce this dependency in code.
There is a single tx-udp-segmentation feature flag to indicate support for
both IPv4/6, hence the devices wishing to support USO must offer checksum
offload for both IP versions.
Fixes: 10154dbded ("udp: Allow GSO transmit from devices with no checksum offload")
Suggested-by: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20240808-udp-gso-egress-from-tunnel-v4-1-f5c5b4149ab9@cloudflare.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In fact, this structure contains a flexible array at the end, but
historically its size, alignment etc., is calculated manually.
There are several instances of the structure embedded into other
structures, but also there's ongoing effort to remove them and we
could in the meantime declare &net_device properly.
Declare the array explicitly, use struct_size() and store the array
size inside the structure, so that __counted_by() can be applied.
Don't use PTR_ALIGN(), as SLUB itself tries its best to ensure the
allocated buffer is aligned to what the user expects.
Also, change its alignment from %NETDEV_ALIGN to the cacheline size
as per several suggestions on the netdev ML.
bloat-o-meter for vmlinux:
free_netdev 445 440 -5
netdev_freemem 24 - -24
alloc_netdev_mqs 1481 1450 -31
On x86_64 with several NICs of different vendors, I was never able to
get a &net_device pointer not aligned to the cacheline size after the
change.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kees Cook <kees@kernel.org>
Link: https://patch.msgid.link/20240710113036.2125584-1-leitao@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
While this is not needed to serialise the ethtool entry points (which
are all under RTNL), drivers may have cause to asynchronously access
dev->ethtool->rss_ctx; taking dev->ethtool->rss_lock allows them to
do this safely without needing to take the RTNL.
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Link: https://patch.msgid.link/7f9c15eb7525bf87af62c275dde3a8570ee8bf0a.1719502240.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a new API to create/modify/remove RSS contexts, that passes in the
newly-chosen context ID (not as a pointer) rather than leaving the
driver to choose it on create. Also pass in the ctx, allowing drivers
to easily use its private data area to store their hardware-specific
state.
Keep the existing .set_rxfh API for now as a fallback, but deprecate it
for custom contexts (rss_context != 0).
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Link: https://patch.msgid.link/45f1fe61df2163c091ec394c9f52000c8b16cc3b.1719502240.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Each context stores the RXFH settings (indir, key, and hfunc) as well
as optionally some driver private data.
Delete any still-existing contexts at netdev unregister time.
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Link: https://patch.msgid.link/cbd1c402cec38f2e03124f2ab65b4ae4e08bd90d.1719502240.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net_dev->ethtool is a pointer to new struct ethtool_netdev_state, which
currently contains only the wol_enabled field.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Link: https://patch.msgid.link/293a562278371de7534ed1eb17531838ca090633.1719502239.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cross-merge networking fixes after downstream PR.
No conflicts.
Adjacent changes:
e3f02f32a0 ("ionic: fix kernel panic due to multi-buffer handling")
d9c0420999 ("ionic: Mark error paths in the data path as unlikely")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The XDP redirect process is two staged:
- bpf_prog_run_xdp() is invoked to run a eBPF program which inspects the
packet and makes decisions. While doing that, the per-CPU variable
bpf_redirect_info is used.
- Afterwards xdp_do_redirect() is invoked and accesses bpf_redirect_info
and it may also access other per-CPU variables like xskmap_flush_list.
At the very end of the NAPI callback, xdp_do_flush() is invoked which
does not access bpf_redirect_info but will touch the individual per-CPU
lists.
The per-CPU variables are only used in the NAPI callback hence disabling
bottom halves is the only protection mechanism. Users from preemptible
context (like cpu_map_kthread_run()) explicitly disable bottom halves
for protections reasons.
Without locking in local_bh_disable() on PREEMPT_RT this data structure
requires explicit locking.
PREEMPT_RT has forced-threaded interrupts enabled and every
NAPI-callback runs in a thread. If each thread has its own data
structure then locking can be avoided.
Create a struct bpf_net_context which contains struct bpf_redirect_info.
Define the variable on stack, use bpf_net_ctx_set() to save a pointer to
it, bpf_net_ctx_clear() removes it again.
The bpf_net_ctx_set() may nest. For instance a function can be used from
within NET_RX_SOFTIRQ/ net_rx_action which uses bpf_net_ctx_set() and
NET_TX_SOFTIRQ which does not. Therefore only the first invocations
updates the pointer.
Use bpf_net_ctx_get_ri() as a wrapper to retrieve the current struct
bpf_redirect_info. The returned data structure is zero initialized to
ensure nothing is leaked from stack. This is done on first usage of the
struct. bpf_net_ctx_set() sets bpf_redirect_info::kern_flags to 0 to
note that initialisation is required. First invocation of
bpf_net_ctx_get_ri() will memset() the data structure and update
bpf_redirect_info::kern_flags.
bpf_redirect_info::nh is excluded from memset because it is only used
once BPF_F_NEIGH is set which also sets the nh member. The kern_flags is
moved past nh to exclude it from memset.
The pointer to bpf_net_context is saved task's task_struct. Using
always the bpf_net_context approach has the advantage that there is
almost zero differences between PREEMPT_RT and non-PREEMPT_RT builds.
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Eduard Zingerman <eddyz87@gmail.com>
Cc: Hao Luo <haoluo@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: KP Singh <kpsingh@kernel.org>
Cc: Martin KaFai Lau <martin.lau@linux.dev>
Cc: Song Liu <song@kernel.org>
Cc: Stanislav Fomichev <sdf@google.com>
Cc: Yonghong Song <yonghong.song@linux.dev>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://patch.msgid.link/20240620132727.660738-15-bigeasy@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
softnet_data::process_queue is a per-CPU variable and relies on disabled
BH for its locking. Without per-CPU locking in local_bh_disable() on
PREEMPT_RT this data structure requires explicit locking.
softnet_data::input_queue_head can be updated lockless. This is fine
because this value is only update CPU local by the local backlog_napi
thread.
Add a local_lock_t to softnet_data and use local_lock_nested_bh() for locking
of process_queue. This change adds only lockdep coverage and does not
alter the functional behaviour for !PREEMPT_RT.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://patch.msgid.link/20240620132727.660738-11-bigeasy@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The backlog_napi locking (previously RPS) relies on explicit locking if
either RPS or backlog NAPI is enabled. If both are disabled then locking
was achieved by disabling interrupts except on PREEMPT_RT. PREEMPT_RT
was excluded because the needed synchronisation was already provided
local_bh_disable().
Since the introduction of backlog NAPI and making it mandatory for
PREEMPT_RT the ifdef within backlog_lock.*() is obsolete and can be
removed.
Remove the ifdefs in backlog_lock.*().
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://patch.msgid.link/20240620132727.660738-10-bigeasy@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Softirq is preemptible on PREEMPT_RT. Without a per-CPU lock in
local_bh_disable() there is no guarantee that only one device is
transmitting at a time.
With preemption and multiple senders it is possible that the per-CPU
`recursion' counter gets incremented by different threads and exceeds
XMIT_RECURSION_LIMIT leading to a false positive recursion alert.
The `more' member is subject to similar problems if set by one thread
for one driver and wrongly used by another driver within another thread.
Instead of adding a lock to protect the per-CPU variable it is simpler
to make xmit per-task. Sending and receiving skbs happens always
in thread context anyway.
Having a lock to protected the per-CPU counter would block/ serialize two
sending threads needlessly. It would also require a recursive lock to
ensure that the owner can increment the counter further.
Make the softnet_data.xmit a task_struct member on PREEMPT_RT. Add
needed wrapper.
Cc: Ben Segall <bsegall@google.com>
Cc: Daniel Bristot de Oliveira <bristot@redhat.com>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Valentin Schneider <vschneid@redhat.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://patch.msgid.link/20240620132727.660738-9-bigeasy@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
skb does not include enough information to find out receiving
sockets/services and netns/containers on packet drops. In theory
skb->dev tells about netns, but it can get cleared/reused, e.g. by TCP
stack for OOO packet lookup. Similarly, skb->sk often identifies a local
sender, and tells nothing about a receiver.
Allow passing an extra receiving socket to the tracepoint to improve
the visibility on receiving drops.
Signed-off-by: Yan Zhai <yan@cloudflare.com>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently have dev_get_tstats64() for collecting per-cpu stats of
type pcpu_sw_netstats ("tstats"). However, tstats doesn't allow for
accounting tx/rx drops. We do have a stats variant that does have stats
for dropped packets: struct pcpu_dstats, but there are no core helpers
for using those stats.
The VRF driver uses dstats, by providing its own collation/fetch
functions to do so.
This change adds a common implementation for dstats-type collection,
used when pcpu_stat_type == NETDEV_PCPU_STAT_DSTAT. This is based on the
VRF driver's existing stats collator (plus the unused tx_drops stat from
there). We will switch the VRF driver to use this in the next change.
Signed-off-by: Jeremy Kerr <jk@codeconstruct.com.au>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240607-dstats-v3-2-cc781fe116f7@codeconstruct.com.au
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cross-merge networking fixes after downstream PR.
No conflicts.
Adjacent changes:
drivers/net/ethernet/pensando/ionic/ionic_txrx.c
d9c0420999 ("ionic: Mark error paths in the data path as unlikely")
491aee894a ("ionic: fix kernel panic in XDP_TX action")
net/ipv6/ip6_fib.c
b4cb4a1391 ("net: use unrcu_pointer() helper")
b01e1c0307 ("ipv6: fix possible race in __fib6_drop_pcpu_from()")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
mono_delivery_time was added to check if skb->tstamp has delivery
time in mono clock base (i.e. EDT) otherwise skb->tstamp has
timestamp in ingress and delivery_time at egress.
Renaming the bitfield from mono_delivery_time to tstamp_type is for
extensibilty for other timestamps such as userspace timestamp
(i.e. SO_TXTIME) set via sock opts.
As we are renaming the mono_delivery_time to tstamp_type, it makes
sense to start assigning tstamp_type based on enum defined
in this commit.
Earlier we used bool arg flag to check if the tstamp is mono in
function skb_set_delivery_time, Now the signature of the functions
accepts tstamp_type to distinguish between mono and real time.
Also skb_set_delivery_type_by_clockid is a new function which accepts
clockid to determine the tstamp_type.
In future tstamp_type:1 can be extended to support userspace timestamp
by increasing the bitfield.
Signed-off-by: Abhishek Chauhan <quic_abchauha@quicinc.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20240509211834.3235191-2-quic_abchauha@quicinc.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Change dev_change_proto_down() and dev_change_proto_down_reason()
to write once on dev->proto_down and dev->proto_down_reason.
Then rtnl_fill_proto_down() can use READ_ONCE() annotations
and run locklessly.
rtnl_proto_down_size() should assume worst case,
because readng dev->proto_down_reason multiple
times would be racy without RTNL in the future.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In the following patch we want to read dev->allmulti
and dev->promiscuity locklessly from rtnl_fill_ifinfo()
In this patch I change __dev_set_promiscuity() and
__dev_set_allmulti() to write these fields (and dev->flags)
only if they succeed, with WRITE_ONCE() annotations.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
rtnl_fill_ifinfo() can read dev->tx_queue_len locklessly,
granted we add corresponding READ_ONCE()/WRITE_ONCE() annotations.
Add missing READ_ONCE(dev->tx_queue_len) in teql_enqueue()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
We will convert ioctl(SIOCGARP) to RCU, and then we need to copy
dev->name which is currently protected by rtnl_lock().
This patch does the following:
1) Add seqlock netdev_rename_lock to protect dev->name
2) Add netdev_copy_name() that copies dev->name to buffer
under netdev_rename_lock
3) Use netdev_copy_name() in netdev_get_name() and drop
devnet_rename_sem
Suggested-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/netdev/CANn89iJEWs7AYSJqGCUABeVqOCTkErponfZdT5kV-iD=-SajnQ@mail.gmail.com/
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/20240430015813.71143-7-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
sysctl_skb_defer_max is used in TCP fast path,
move it to net_hodata.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240429134025.1233626-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This came while reviewing commit c4e86b4363 ("net: add two more
call_rcu_hurry()").
Paolo asked if adding one synchronize_rcu() would help.
While synchronize_rcu() does not help, making sure to call
rcu_barrier() before msleep(wait) is definitely helping
to make sure lazy call_rcu() are completed.
Instead of waiting ~100 seconds in my tests, the ref_tracker
splats occurs one time only, and netdev_wait_allrefs_any()
latency is reduced to the strict minimum.
Ideally we should audit our call_rcu() users to make sure
no refcount (or cascading call_rcu()) is held too long,
because rcu_barrier() is quite expensive.
Fixes: 0e4be9e57e ("net: use exponential backoff in netdev_wait_allrefs")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/all/28bbf698-befb-42f6-b561-851c67f464aa@kernel.org/T/#m76d73ed6b03cd930778ac4d20a777f22a08d6824
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is impossible to use init_dummy_netdev together with alloc_netdev()
as the 'setup' argument.
This is because alloc_netdev() initializes some fields in the net_device
structure, and later init_dummy_netdev() memzero them all. This causes
some problems as reported here:
https://lore.kernel.org/all/20240322082336.49f110cc@kernel.org/
Split the init_dummy_netdev() function in two. Create a new function called
init_dummy_netdev_core() that does not memzero the net_device structure.
Then have init_dummy_netdev() memzero-ing and calling
init_dummy_netdev_core(), keeping the old behaviour.
init_dummy_netdev_core() is the new function that could be called as an
argument for alloc_netdev().
Also, create a helper to allocate and initialize dummy net devices,
leveraging init_dummy_netdev_core() as the setup argument. This function
basically simplify the allocation of dummy devices, by allocating and
initializing it. Freeing the device continue to be done through
free_netdev()
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For dummy devices, exit earlier at free_netdev() instead of executing
the whole function. This is necessary, because dummy devices are
special, and shouldn't have the second part of the function executed.
Otherwise reg_state, which is NETREG_DUMMY, will be overwritten and
there will be no way to identify that this is a dummy device. Also, this
device do not need the final put_device(), since dummy devices are not
registered (through register_netdevice()), where the device reference is
increased (at netdev_register_kobject()/device_add()).
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix bad grammar in description of init_dummy_netdev() function. This
topic showed up in the review of the "allocate dummy device dynamically"
patch set.
Suggested-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the last member in struct rps_dev_flow which should be
protected locklessly. So finish it.
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As we can see, rflow->filter can be written/read concurrently, so
lockless access is needed.
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Removing one unnecessary reader protection and add another writer
protection to finish the locklessly proctection job.
Note: the removed READ_ONCE() is not needed because we only have to protect
the locklessly reader in the different context (rps_may_expire_flow()).
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the mirred action is used on a classful egress qdisc and a packet is
mirrored or redirected to self we hit a qdisc lock deadlock.
See trace below.
[..... other info removed for brevity....]
[ 82.890906]
[ 82.890906] ============================================
[ 82.890906] WARNING: possible recursive locking detected
[ 82.890906] 6.8.0-05205-g77fadd89fe2d-dirty #213 Tainted: G W
[ 82.890906] --------------------------------------------
[ 82.890906] ping/418 is trying to acquire lock:
[ 82.890906] ffff888006994110 (&sch->q.lock){+.-.}-{3:3}, at:
__dev_queue_xmit+0x1778/0x3550
[ 82.890906]
[ 82.890906] but task is already holding lock:
[ 82.890906] ffff888006994110 (&sch->q.lock){+.-.}-{3:3}, at:
__dev_queue_xmit+0x1778/0x3550
[ 82.890906]
[ 82.890906] other info that might help us debug this:
[ 82.890906] Possible unsafe locking scenario:
[ 82.890906]
[ 82.890906] CPU0
[ 82.890906] ----
[ 82.890906] lock(&sch->q.lock);
[ 82.890906] lock(&sch->q.lock);
[ 82.890906]
[ 82.890906] *** DEADLOCK ***
[ 82.890906]
[..... other info removed for brevity....]
Example setup (eth0->eth0) to recreate
tc qdisc add dev eth0 root handle 1: htb default 30
tc filter add dev eth0 handle 1: protocol ip prio 2 matchall \
action mirred egress redirect dev eth0
Another example(eth0->eth1->eth0) to recreate
tc qdisc add dev eth0 root handle 1: htb default 30
tc filter add dev eth0 handle 1: protocol ip prio 2 matchall \
action mirred egress redirect dev eth1
tc qdisc add dev eth1 root handle 1: htb default 30
tc filter add dev eth1 handle 1: protocol ip prio 2 matchall \
action mirred egress redirect dev eth0
We fix this by adding an owner field (CPU id) to struct Qdisc set after
root qdisc is entered. When the softirq enters it a second time, if the
qdisc owner is the same CPU, the packet is dropped to break the loop.
Reported-by: Mingshuai Ren <renmingshuai@huawei.com>
Closes: https://lore.kernel.org/netdev/20240314111713.5979-1-renmingshuai@huawei.com/
Fixes: 3bcb846ca4 ("net: get rid of spin_trylock() in net_tx_action()")
Fixes: e578d9c025 ("net: sched: use counter to break reclassify loops")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Victor Nogueira <victor@mojatatu.com>
Reviewed-by: Pedro Tammela <pctammela@mojatatu.com>
Tested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20240415210728.36949-1-victor@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link topologies containing multiple network PHYs attached to the same
net_device can be found when using a PHY as a media converter for use
with an SFP connector, on which an SFP transceiver containing a PHY can
be used.
With the current model, the transceiver's PHY can't be used for
operations such as cable testing, timestamping, macsec offload, etc.
The reason being that most of the logic for these configuration, coming
from either ethtool netlink or ioctls tend to use netdev->phydev, which
in multi-phy systems will reference the PHY closest to the MAC.
Introduce a numbering scheme allowing to enumerate PHY devices that
belong to any netdev, which can in turn allow userspace to take more
precise decisions with regard to each PHY's configuration.
The numbering is maintained per-netdev, in a phy_device_list.
The numbering works similarly to a netdevice's ifindex, with
identifiers that are only recycled once INT_MAX has been reached.
This prevents races that could occur between PHY listing and SFP
transceiver removal/insertion.
The identifiers are assigned at phy_attach time, as the numbering
depends on the netdevice the phy is attached to. The PHY index can be
re-used for PHYs that are persistent.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
For systems that use CPU isolation (via nohz_full), creating or destroying
a socket with SO_TIMESTAMP, SO_TIMESTAMPNS or SO_TIMESTAMPING with flag
SOF_TIMESTAMPING_RX_SOFTWARE will cause a static key to be enabled/disabled.
This in turn causes undesired IPIs to isolated CPUs.
So enable the static key unconditionally, if CPU isolation is enabled,
thus avoiding the IPIs.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/ZgrUiLLtbEUf9SFn@tpad
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
process_backlog() can batch increments of sd->input_queue_head,
saving some memory bandwidth.
Also add READ_ONCE()/WRITE_ONCE() annotations around
sd->input_queue_head accesses.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
input_queue_tail_incr_save() is incrementing the sd queue_tail
and save it in the flow last_qtail.
Two issues here :
- no lock protects the write on last_qtail, we should use appropriate
annotations.
- We can perform this write after releasing the per-cpu backlog lock,
to decrease this lock hold duration (move away the cache line miss)
Also move input_queue_head_incr() and rps helpers to include/net/rps.h,
while adding rps_ prefix to better reflect their role.
v2: Fixed a build issue (Jakub and kernel build bots)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can remove a goto and a label by reversing a condition.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If under extreme cpu backlog pressure enqueue_to_backlog() has
to drop a packet, it could do this without dirtying a cache line
and potentially slowing down the target cpu.
Move sd->dropped into a separate cache line, and make it atomic.
In non pressure mode, this field is not touched, no need to consume
valuable space in a hot cache line.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the device attached to the packet given to enqueue_to_backlog()
is not running, we drop the packet.
But we accidentally increase sd->dropped, giving false signals
to admins: sd->dropped should be reserved to cpu backlog pressure,
not to temporary glitches at device dismantles.
While we are at it, perform the netif_running() test before
we get the rps lock, and use REASON_DEV_READY
drop reason instead of NOT_SPECIFIED.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no reason to consume a full cacheline to store system_page_pool.
We can eventually move it to softnet_data later for full locality control.
Fixes: 2b0cfa6e49 ("net: add generic percpu page_pool allocator")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Lorenzo Bianconi <lorenzo@kernel.org>
Cc: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Link: https://lore.kernel.org/r/20240328173448.2262593-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
TC filters come in 3 variants:
- no flag (try to process in hardware, but fallback to software))
- skip_hw (do not process filter by hardware)
- skip_sw (do not process filter by software)
However skip_sw is implemented so that the skip_sw
flag can first be checked, after it has been matched.
IMHO it's common when using skip_sw, to use it on all rules.
So if all filters in a block is skip_sw filters, then
we can bail early, we can thus avoid having to match
the filters, just to check for the skip_sw flag.
This patch adds a bypass, for when only TC skip_sw rules
are used. The bypass is guarded by a static key, to avoid
harming other workloads.
There are 3 ways that a packet from a skip_sw ruleset, can
end up in the kernel path. Although the send packets to a
non-existent chain way is only improved a few percents, then
I believe it's worth optimizing the trap and fall-though
use-cases.
+----------------------------+--------+--------+--------+
| Test description | Pre- | Post- | Rel. |
| | kpps | kpps | chg. |
+----------------------------+--------+--------+--------+
| basic forwarding + notrack | 3589.3 | 3587.9 | 1.00x |
| switch to eswitch mode | 3081.8 | 3094.7 | 1.00x |
| add ingress qdisc | 3042.9 | 3063.6 | 1.01x |
| tc forward in hw / skip_sw |37024.7 |37028.4 | 1.00x |
| tc forward in sw / skip_hw | 3245.0 | 3245.3 | 1.00x |
+----------------------------+--------+--------+--------+
| tests with only skip_sw rules below: |
+----------------------------+--------+--------+--------+
| 1 non-matching rule | 2694.7 | 3058.7 | 1.14x |
| 1 n-m rule, match trap | 2611.2 | 3323.1 | 1.27x |
| 1 n-m rule, goto non-chain | 2886.8 | 2945.9 | 1.02x |
| 5 non-matching rules | 1958.2 | 3061.3 | 1.56x |
| 5 n-m rules, match trap | 1911.9 | 3327.0 | 1.74x |
| 5 n-m rules, goto non-chain| 2883.1 | 2947.5 | 1.02x |
| 10 non-matching rules | 1466.3 | 3062.8 | 2.09x |
| 10 n-m rules, match trap | 1444.3 | 3317.9 | 2.30x |
| 10 n-m rules,goto non-chain| 2883.1 | 2939.5 | 1.02x |
| 25 non-matching rules | 838.5 | 3058.9 | 3.65x |
| 25 n-m rules, match trap | 824.5 | 3323.0 | 4.03x |
| 25 n-m rules,goto non-chain| 2875.8 | 2944.7 | 1.02x |
| 50 non-matching rules | 488.1 | 3054.7 | 6.26x |
| 50 n-m rules, match trap | 484.9 | 3318.5 | 6.84x |
| 50 n-m rules,goto non-chain| 2884.1 | 2939.7 | 1.02x |
+----------------------------+--------+--------+--------+
perf top (25 n-m skip_sw rules - pre patch):
20.39% [kernel] [k] __skb_flow_dissect
16.43% [kernel] [k] rhashtable_jhash2
10.58% [kernel] [k] fl_classify
10.23% [kernel] [k] fl_mask_lookup
4.79% [kernel] [k] memset_orig
2.58% [kernel] [k] tcf_classify
1.47% [kernel] [k] __x86_indirect_thunk_rax
1.42% [kernel] [k] __dev_queue_xmit
1.36% [kernel] [k] nft_do_chain
1.21% [kernel] [k] __rcu_read_lock
perf top (25 n-m skip_sw rules - post patch):
5.12% [kernel] [k] __dev_queue_xmit
4.77% [kernel] [k] nft_do_chain
3.65% [kernel] [k] dev_gro_receive
3.41% [kernel] [k] check_preemption_disabled
3.14% [kernel] [k] mlx5e_skb_from_cqe_mpwrq_nonlinear
2.88% [kernel] [k] __netif_receive_skb_core.constprop.0
2.49% [kernel] [k] mlx5e_xmit
2.15% [kernel] [k] ip_forward
1.95% [kernel] [k] mlx5e_tc_restore_tunnel
1.92% [kernel] [k] vlan_gro_receive
Test setup:
DUT: Intel Xeon D-1518 (2.20GHz) w/ Nvidia/Mellanox ConnectX-6 Dx 2x100G
Data rate measured on switch (Extreme X690), and DUT connected as
a router on a stick, with pktgen and pktsink as VLANs.
Pktgen-dpdk was in range 36.6-37.7 Mpps 64B packets across all tests.
Full test data at https://files.fiberby.net/ast/2024/tc_skip_sw/v2_tests/
Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
System page_pools are percpu and one instance can be used only on
one CPU.
%NUMA_NO_NODE is fine for allocating pages, as the PP core always
allocates local pages in this case. But for the struct &page_pool
itself, this node ID means they are allocated on the boot CPU,
which may belong to a different node than the target CPU.
Pin system page_pools to the corresponding nodes when creating,
so that all the allocated data will always be local. Use
cpu_to_mem() to account memless nodes.
Nodes != 0 win some Kpps when testing with xdp-trafficgen.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/20240325160635.3215855-1-aleksander.lobakin@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The rps_lock.*() functions use the inner lock of a sk_buff_head for
locking. This lock is used if RPS is enabled, otherwise the list is
accessed lockless and disabling interrupts is enough for the
synchronisation because it is only accessed CPU local. Not only the list
is protected but also the NAPI state protected.
With the addition of backlog threads, the lock is also needed because of
the cross CPU access even without RPS. The clean up of the defer_list
list is also done via backlog threads (if enabled).
It has been suggested to rename the locking function since it is no
longer just RPS.
Rename the rps_lock*() functions to backlog_lock*().
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The defer_list is a per-CPU list which is used to free skbs outside of
the socket lock and on the CPU on which they have been allocated.
The list is processed during NAPI callbacks so ideally the list is
cleaned up.
Should the amount of skbs on the list exceed a certain water mark then
the softirq is triggered remotely on the target CPU by invoking a remote
function call. The raise of the softirqs via a remote function call
leads to waking the ksoftirqd on PREEMPT_RT which is undesired.
The backlog-NAPI threads already provide the infrastructure which can be
utilized to perform the cleanup of the defer_list.
The NAPI state is updated with the input_pkt_queue.lock acquired. It
order not to break the state, it is needed to also wake the backlog-NAPI
thread with the lock held. This requires to acquire the use the lock in
rps_lock_irq*() if the backlog-NAPI threads are used even with RPS
disabled.
Move the logic of remotely starting softirqs to clean up the defer_list
into kick_defer_list_purge(). Make sure a lock is held in
rps_lock_irq*() if backlog-NAPI threads are used. Schedule backlog-NAPI
for defer_list cleanup if backlog-NAPI is available.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Backlog NAPI is a per-CPU NAPI struct only (with no device behind it)
used by drivers which don't do NAPI them self, RPS and parts of the
stack which need to avoid recursive deadlocks while processing a packet.
The non-NAPI driver use the CPU local backlog NAPI. If RPS is enabled
then a flow for the skb is computed and based on the flow the skb can be
enqueued on a remote CPU. Scheduling/ raising the softirq (for backlog's
NAPI) on the remote CPU isn't trivial because the softirq is only
scheduled on the local CPU and performed after the hardirq is done.
In order to schedule a softirq on the remote CPU, an IPI is sent to the
remote CPU which schedules the backlog-NAPI on the then local CPU.
On PREEMPT_RT interrupts are force-threaded. The soft interrupts are
raised within the interrupt thread and processed after the interrupt
handler completed still within the context of the interrupt thread. The
softirq is handled in the context where it originated.
With force-threaded interrupts enabled, ksoftirqd is woken up if a
softirq is raised from hardirq context. This is the case if it is raised
from an IPI. Additionally there is a warning on PREEMPT_RT if the
softirq is raised from the idle thread.
This was done for two reasons:
- With threaded interrupts the processing should happen in thread
context (where it originated) and ksoftirqd is the only thread for
this context if raised from hardirq. Using the currently running task
instead would "punish" a random task.
- Once ksoftirqd is active it consumes all further softirqs until it
stops running. This changed recently and is no longer the case.
Instead of keeping the backlog NAPI in ksoftirqd (in force-threaded/
PREEMPT_RT setups) I am proposing NAPI-threads for backlog.
The "proper" setup with threaded-NAPI is not doable because the threads
are not pinned to an individual CPU and can be modified by the user.
Additionally a dummy network device would have to be assigned. Also
CPU-hotplug has to be considered if additional CPUs show up.
All this can be probably done/ solved but the smpboot-threads already
provide this infrastructure.
Sending UDP packets over loopback expects that the packet is processed
within the call. Delaying it by handing it over to the thread hurts
performance. It is not beneficial to the outcome if the context switch
happens immediately after enqueue or after a while to process a few
packets in a batch.
There is no need to always use the thread if the backlog NAPI is
requested on the local CPU. This restores the loopback throuput. The
performance drops mostly to the same value after enabling RPS on the
loopback comparing the IPI and the tread result.
Create NAPI-threads for backlog if request during boot. The thread runs
the inner loop from napi_threaded_poll(), the wait part is different. It
checks for NAPI_STATE_SCHED (the backlog NAPI can not be disabled).
The NAPI threads for backlog are optional, it has to be enabled via the boot
argument "thread_backlog_napi". It is mandatory for PREEMPT_RT to avoid the
wakeup of ksoftirqd from the IPI.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
A NAPI thread is scheduled by first setting NAPI_STATE_SCHED bit. If
successful (the bit was not yet set) then the NAPI_STATE_SCHED_THREADED
is set but only if thread's state is not TASK_INTERRUPTIBLE (is
TASK_RUNNING) followed by task wakeup.
If the task is idle (TASK_INTERRUPTIBLE) then the
NAPI_STATE_SCHED_THREADED bit is not set. The thread is no relying on
the bit but always leaving the wait-loop after returning from schedule()
because there must have been a wakeup.
The smpboot-threads implementation for per-CPU threads requires an
explicit condition and does not support "if we get out of schedule()
then there must be something to do".
Removing this optimisation simplifies the following integration.
Set NAPI_STATE_SCHED_THREADED unconditionally on wakeup and rely on it
in the wait path by removing the `woken' condition.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
NAPI threads can keep polling packets under load. Currently it is only
calling cond_resched() before repolling, but it is not sufficient to
clear out the holdout of RCU tasks, which prevent BPF tracing programs
from detaching for long period. This can be reproduced easily with
following set up:
ip netns add test1
ip netns add test2
ip -n test1 link add veth1 type veth peer name veth2 netns test2
ip -n test1 link set veth1 up
ip -n test1 link set lo up
ip -n test2 link set veth2 up
ip -n test2 link set lo up
ip -n test1 addr add 192.168.1.2/31 dev veth1
ip -n test1 addr add 1.1.1.1/32 dev lo
ip -n test2 addr add 192.168.1.3/31 dev veth2
ip -n test2 addr add 2.2.2.2/31 dev lo
ip -n test1 route add default via 192.168.1.3
ip -n test2 route add default via 192.168.1.2
for i in `seq 10 210`; do
for j in `seq 10 210`; do
ip netns exec test2 iptables -I INPUT -s 3.3.$i.$j -p udp --dport 5201
done
done
ip netns exec test2 ethtool -K veth2 gro on
ip netns exec test2 bash -c 'echo 1 > /sys/class/net/veth2/threaded'
ip netns exec test1 ethtool -K veth1 tso off
Then run an iperf3 client/server and a bpftrace script can trigger it:
ip netns exec test2 iperf3 -s -B 2.2.2.2 >/dev/null&
ip netns exec test1 iperf3 -c 2.2.2.2 -B 1.1.1.1 -u -l 1500 -b 3g -t 100 >/dev/null&
bpftrace -e 'kfunc:__napi_poll{@=count();} interval:s:1{exit();}'
Report RCU quiescent states periodically will resolve the issue.
Fixes: 29863d41bb ("net: implement threaded-able napi poll loop support")
Reviewed-by: Jesper Dangaard Brouer <hawk@kernel.org>
Signed-off-by: Yan Zhai <yan@cloudflare.com>
Acked-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: Jesper Dangaard Brouer <hawk@kernel.org>
Link: https://lore.kernel.org/r/4c3b0d3f32d3b18949d75b18e5e1d9f13a24f025.1710877680.git.yan@cloudflare.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
dev->state can be read in rx and tx fast paths.
netif_running() which needs dev->state is called from
- enqueue_to_backlog() [RX path]
- __dev_direct_xmit() [TX path]
Fixes: 43a71cd66b ("net-device: reorganize net_device fast path variables")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Coco Li <lixiaoyan@google.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20240314200845.3050179-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
rps_sock_flow_table and rps_cpu_mask are used in fast path.
Move them to net_hotdata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-19-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move RPS related structures and helpers from include/linux/netdevice.h
and include/net/sock.h to a new include file.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-18-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
dev_rx_weight is read from process_backlog().
Move it to net_hotdata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-10-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
dev_tx_weight is used in tx fast path.
Move it to net_hotdata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-9-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
netdev_max_backlog is used in rx fat path.
Move it to net_hodata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-6-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ptype_all is used in rx/tx fast paths.
Move it to net_hotdata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
netdev_tstamp_prequeue is used in rx path.
Move it to net_hotdata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
netdev_budget and netdev_budget are used in rx path (net_rx_action())
Move them into net_hotdata for better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Older versions of GCC really want to know the full definition
of the type involved in rcu_assign_pointer().
struct dpll_pin is defined in a local header, net/core can't
reach it. Move all the netdev <> dpll code into dpll, where
the type is known. Otherwise we'd need multiple function calls
to jump between the compilation units.
This is the same problem the commit under fixes was trying to address,
but with rcu_assign_pointer() not rcu_dereference().
Some of the exports are not needed, networking core can't
be a module, we only need exports for the helpers used by
drivers.
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Link: https://lore.kernel.org/all/35a869c8-52e8-177-1d4d-e57578b99b6@linux-m68k.org/
Fixes: 640f41ed33 ("dpll: fix build failure due to rcu_dereference_check() on unknown type")
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20240305013532.694866-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In the following patch, inet_base_seq() will no longer be called
with RTNL held.
Add READ_ONCE()/WRITE_ONCE() annotations in dev_base_seq_inc()
and inet_base_seq().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the network driver is relying in the net core to do stats allocation,
then we want to dev_get_tstats64() instead of netdev_stats_to_stats64(),
since there are per-cpu stats that needs to be taken in consideration.
This will also simplify the drivers in regard to statistics. Once the
driver sets NETDEV_PCPU_STAT_TSTATS, it doesn't not need to allocate the
stacks, neither it needs to set `.ndo_get_stats64 = dev_get_tstats64`
for the generic stats collection function anymore.
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
skb_defer_free_flush() is currently called from net_rx_action()
and napi_threaded_poll().
We should also call it from __napi_busy_loop() otherwise
there is the risk the percpu queue can grow until an IPI
is forced from skb_attempt_defer_free() adding a latency spike.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Samiullah Khawaja <skhawaja@google.com>
Acked-by: Stanislav Fomichev <sdf@google.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20240227210105.3815474-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This fixes a possible UAF in if_nlmsg_size(),
which can run without RTNL.
Add rcu protection to "struct dpll_pin"
Move netdev_dpll_pin() from netdevice.h to dpll.h to
decrease name pollution.
Note: This looks possible to no longer acquire RTNL in
netdev_dpll_pin_assign() later in net-next.
v2: do not force rcu_read_lock() in rtnl_dpll_pin_size() (Jiri Pirko)
Fixes: 5f18426928 ("netdev: expose DPLL pin handle for netdevice")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Cc: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20240223123208.3543319-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We want to use RCU protection instead of RTNL
for inet6_fill_ifinfo().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We want to be able to run rtnl_fill_ifinfo() under RCU protection
instead of RTNL in the future.
This patch prepares dev_get_iflink() and nla_put_iflink()
to run either with RTNL or RCU held.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use global percpu page_pool_recycle_stats counter for system page_pool
allocator instead of allocating a separate percpu variable for each
(also percpu) page pool instance.
Reviewed-by: Toke Hoiland-Jorgensen <toke@redhat.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://lore.kernel.org/r/87f572425e98faea3da45f76c3c68815c01a20ee.1708075412.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
