skb_defer_free_flush() runs from softirq context,
we have the opportunity to refill the napi_alloc_cache,
and/or use kmem_cache_free_bulk() when this cache is full.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A cpu can observe sd->defer_count reaching 128,
and call smp_call_function_single_async()
Problem is that the remote CPU can clear sd->defer_count
before the IPI is run/acknowledged.
Other cpus can queue more packets and also decide
to call smp_call_function_single_async() while the pending
IPI was not yet delivered.
This is a common issue with smp_call_function_single_async().
Callers must ensure correct synchronization and serialization.
I triggered this issue while experimenting smaller threshold.
Performing the call to smp_call_function_single_async()
under sd->defer_lock protection did not solve the problem.
Commit 5a18ceca63 ("smp: Allow smp_call_function_single_async()
to insert locked csd") replaced an informative WARN_ON_ONCE()
with a return of -EBUSY, which is often ignored.
Test of CSD_FLAG_LOCK presence is racy anyway.
Fixes: 68822bdf76 ("net: generalize skb freeing deferral to per-cpu lists")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow the gro_max_size to exceed a value larger than 65536.
There weren't really any external limitations that prevented this other
than the fact that IPv4 only supports a 16 bit length field. Since we have
the option of adding a hop-by-hop header for IPv6 we can allow IPv6 to
exceed this value and for IPv4 and non-TCP flows we can cap things at 65536
via a constant rather than relying on gro_max_size.
[edumazet] limit GRO_MAX_SIZE to (8 * 65535) to avoid overflows.
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code for gso_max_size was added originally to allow for debugging and
workaround of buggy devices that couldn't support TSO with blocks 64K in
size. The original reason for limiting it to 64K was because that was the
existing limits of IPv4 and non-jumbogram IPv6 length fields.
With the addition of Big TCP we can remove this limit and allow the value
to potentially go up to UINT_MAX and instead be limited by the tso_max_size
value.
So in order to support this we need to go through and clean up the
remaining users of the gso_max_size value so that the values will cap at
64K for non-TCPv6 flows. In addition we can clean up the GSO_MAX_SIZE value
so that 64K becomes GSO_LEGACY_MAX_SIZE and UINT_MAX will now be the upper
limit for GSO_MAX_SIZE.
v6: (edumazet) fixed a compile error if CONFIG_IPV6=n,
in a new sk_trim_gso_size() helper.
netif_set_tso_max_size() caps the requested TSO size
with GSO_MAX_SIZE.
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a followup of previous patch.
Dumping the stack trace is a good start, but printing
basic skb information is probably better.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I have a syzbot report that managed to get a crash in skb_checksum_help()
If syzbot can trigger these BUG(), it makes sense to replace
them with more friendly WARN_ON_ONCE() since skb_checksum_help()
can instead return an error code.
Note that syzbot will still crash there, until real bug is fixed.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Drivers should call the TSO setting helper, GSO is controllable
by user space.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Up until commit 46e6b992c2 ("rtnetlink: allow GSO maximums to
be set on device creation") the gso_max_segs and gso_max_size
of a device were not controlled from user space.
The quoted commit added the ability to control them because of
the following setup:
netns A | netns B
veth<->veth eth0
If eth0 has TSO limitations and user wants to efficiently forward
traffic between eth0 and the veths they should copy the TSO
limitations of eth0 onto the veths. This would happen automatically
for macvlans or ipvlan but veth users are not so lucky (given the
loose coupling).
Unfortunately the commit in question allowed users to also override
the limits on real HW devices.
It may be useful to control the max GSO size and someone may be using
that ability (not that I know of any user), so create a separate set
of knobs to reliably record the TSO limitations. Validate the user
requests.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
To make later patches smaller create a helper for inheriting
the TSO limitations of a lower device. The TSO in the name
is not an accident, subsequent patches will replace GSO
with TSO in more names.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most drivers should not have to worry about selecting the right
weight for their NAPI instances and pass NAPI_POLL_WEIGHT.
It'd be best if we didn't require the argument at all and selected
the default internally.
This change prepares the ground for such reshuffling, allowing
for a smooth transition. The following API should remain after
the next release cycle:
netif_napi_add()
netif_napi_add_weight()
netif_napi_add_tx()
netif_napi_add_tx_weight()
Where the _weight() variants take an explicit weight argument.
I opted for a _weight() suffix rather than a __ prefix, because
we use __ in places to mean that caller needs to also issue a
synchronize_net() call.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20220502232703.396351-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Inline dev_queue_xmit() and dev_queue_xmit_accel(), they both are small
proxy functions doing nothing but redirecting the control flow to
__dev_queue_xmit().
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I missed a stray return; in net_rx_action(), which very well
is taken whenever trigger_rx_softirq() has been called on
a cpu that is no longer receiving network packets,
or receiving too few of them.
Fixes: 68822bdf76 ("net: generalize skb freeing deferral to per-cpu lists")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/20220427204147.1310161-1-eric.dumazet@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The macro dev_core_stats_##FIELD##_inc() disables preemption and invokes
netdev_core_stats_alloc() to return a per-CPU pointer.
netdev_core_stats_alloc() will allocate memory on its first invocation
which breaks on PREEMPT_RT because it requires non-atomic context for
memory allocation.
This can be avoided by enabling preemption in netdev_core_stats_alloc()
assuming the caller always disables preemption.
It might be better to replace local_inc() with this_cpu_inc() now that
dev_core_stats_##FIELD##_inc() gained a preempt-disable section and does
not rely on already disabled preemption. This results in less
instructions on x86-64:
local_inc:
| incl %gs:__preempt_count(%rip) # __preempt_count
| movq 488(%rdi), %rax # _1->core_stats, _22
| testq %rax, %rax # _22
| je .L585 #,
| add %gs:this_cpu_off(%rip), %rax # this_cpu_off, tcp_ptr__
| .L586:
| testq %rax, %rax # _27
| je .L587 #,
| incq (%rax) # _6->a.counter
| .L587:
| decl %gs:__preempt_count(%rip) # __preempt_count
this_cpu_inc(), this patch:
| movq 488(%rdi), %rax # _1->core_stats, _5
| testq %rax, %rax # _5
| je .L591 #,
| .L585:
| incq %gs:(%rax) # _18->rx_dropped
Use unsigned long as type for the counter. Use this_cpu_inc() to
increment the counter. Use a plain read of the counter.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/YmbO0pxgtKpCw4SY@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Logic added in commit f35f821935 ("tcp: defer skb freeing after socket
lock is released") helped bulk TCP flows to move the cost of skbs
frees outside of critical section where socket lock was held.
But for RPC traffic, or hosts with RFS enabled, the solution is far from
being ideal.
For RPC traffic, recvmsg() has to return to user space right after
skb payload has been consumed, meaning that BH handler has no chance
to pick the skb before recvmsg() thread. This issue is more visible
with BIG TCP, as more RPC fit one skb.
For RFS, even if BH handler picks the skbs, they are still picked
from the cpu on which user thread is running.
Ideally, it is better to free the skbs (and associated page frags)
on the cpu that originally allocated them.
This patch removes the per socket anchor (sk->defer_list) and
instead uses a per-cpu list, which will hold more skbs per round.
This new per-cpu list is drained at the end of net_action_rx(),
after incoming packets have been processed, to lower latencies.
In normal conditions, skbs are added to the per-cpu list with
no further action. In the (unlikely) cases where the cpu does not
run net_action_rx() handler fast enough, we use an IPI to raise
NET_RX_SOFTIRQ on the remote cpu.
Also, we do not bother draining the per-cpu list from dev_cpu_dead()
This is because skbs in this list have no requirement on how fast
they should be freed.
Note that we can add in the future a small per-cpu cache
if we see any contention on sd->defer_lock.
Tested on a pair of hosts with 100Gbit NIC, RFS enabled,
and /proc/sys/net/ipv4/tcp_rmem[2] tuned to 16MB to work around
page recycling strategy used by NIC driver (its page pool capacity
being too small compared to number of skbs/pages held in sockets
receive queues)
Note that this tuning was only done to demonstrate worse
conditions for skb freeing for this particular test.
These conditions can happen in more general production workload.
10 runs of one TCP_STREAM flow
Before:
Average throughput: 49685 Mbit.
Kernel profiles on cpu running user thread recvmsg() show high cost for
skb freeing related functions (*)
57.81% [kernel] [k] copy_user_enhanced_fast_string
(*) 12.87% [kernel] [k] skb_release_data
(*) 4.25% [kernel] [k] __free_one_page
(*) 3.57% [kernel] [k] __list_del_entry_valid
1.85% [kernel] [k] __netif_receive_skb_core
1.60% [kernel] [k] __skb_datagram_iter
(*) 1.59% [kernel] [k] free_unref_page_commit
(*) 1.16% [kernel] [k] __slab_free
1.16% [kernel] [k] _copy_to_iter
(*) 1.01% [kernel] [k] kfree
(*) 0.88% [kernel] [k] free_unref_page
0.57% [kernel] [k] ip6_rcv_core
0.55% [kernel] [k] ip6t_do_table
0.54% [kernel] [k] flush_smp_call_function_queue
(*) 0.54% [kernel] [k] free_pcppages_bulk
0.51% [kernel] [k] llist_reverse_order
0.38% [kernel] [k] process_backlog
(*) 0.38% [kernel] [k] free_pcp_prepare
0.37% [kernel] [k] tcp_recvmsg_locked
(*) 0.37% [kernel] [k] __list_add_valid
0.34% [kernel] [k] sock_rfree
0.34% [kernel] [k] _raw_spin_lock_irq
(*) 0.33% [kernel] [k] __page_cache_release
0.33% [kernel] [k] tcp_v6_rcv
(*) 0.33% [kernel] [k] __put_page
(*) 0.29% [kernel] [k] __mod_zone_page_state
0.27% [kernel] [k] _raw_spin_lock
After patch:
Average throughput: 73076 Mbit.
Kernel profiles on cpu running user thread recvmsg() looks better:
81.35% [kernel] [k] copy_user_enhanced_fast_string
1.95% [kernel] [k] _copy_to_iter
1.95% [kernel] [k] __skb_datagram_iter
1.27% [kernel] [k] __netif_receive_skb_core
1.03% [kernel] [k] ip6t_do_table
0.60% [kernel] [k] sock_rfree
0.50% [kernel] [k] tcp_v6_rcv
0.47% [kernel] [k] ip6_rcv_core
0.45% [kernel] [k] read_tsc
0.44% [kernel] [k] _raw_spin_lock_irqsave
0.37% [kernel] [k] _raw_spin_lock
0.37% [kernel] [k] native_irq_return_iret
0.33% [kernel] [k] __inet6_lookup_established
0.31% [kernel] [k] ip6_protocol_deliver_rcu
0.29% [kernel] [k] tcp_rcv_established
0.29% [kernel] [k] llist_reverse_order
v2: kdoc issue (kernel bots)
do not defer if (alloc_cpu == smp_processor_id()) (Paolo)
replace the sk_buff_head with a single-linked list (Jakub)
add a READ_ONCE()/WRITE_ONCE() for the lockless read of sd->defer_list
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/20220422201237.416238-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch fixes issue:
* If we install tc filters with act_skbedit in clsact hook.
It doesn't work, because netdev_core_pick_tx() overwrites
queue_mapping.
$ tc filter ... action skbedit queue_mapping 1
And this patch is useful:
* We can use FQ + EDT to implement efficient policies. Tx queues
are picked by xps, ndo_select_queue of netdev driver, or skb hash
in netdev_core_pick_tx(). In fact, the netdev driver, and skb
hash are _not_ under control. xps uses the CPUs map to select Tx
queues, but we can't figure out which task_struct of pod/containter
running on this cpu in most case. We can use clsact filters to classify
one pod/container traffic to one Tx queue. Why ?
In containter networking environment, there are two kinds of pod/
containter/net-namespace. One kind (e.g. P1, P2), the high throughput
is key in these applications. But avoid running out of network resource,
the outbound traffic of these pods is limited, using or sharing one
dedicated Tx queues assigned HTB/TBF/FQ Qdisc. Other kind of pods
(e.g. Pn), the low latency of data access is key. And the traffic is not
limited. Pods use or share other dedicated Tx queues assigned FIFO Qdisc.
This choice provides two benefits. First, contention on the HTB/FQ Qdisc
lock is significantly reduced since fewer CPUs contend for the same queue.
More importantly, Qdisc contention can be eliminated completely if each
CPU has its own FIFO Qdisc for the second kind of pods.
There must be a mechanism in place to support classifying traffic based on
pods/container to different Tx queues. Note that clsact is outside of Qdisc
while Qdisc can run a classifier to select a sub-queue under the lock.
In general recording the decision in the skb seems a little heavy handed.
This patch introduces a per-CPU variable, suggested by Eric.
The xmit.skip_txqueue flag is firstly cleared in __dev_queue_xmit().
- Tx Qdisc may install that skbedit actions, then xmit.skip_txqueue flag
is set in qdisc->enqueue() though tx queue has been selected in
netdev_tx_queue_mapping() or netdev_core_pick_tx(). That flag is cleared
firstly in __dev_queue_xmit(), is useful:
- Avoid picking Tx queue with netdev_tx_queue_mapping() in next netdev
in such case: eth0 macvlan - eth0.3 vlan - eth0 ixgbe-phy:
For example, eth0, macvlan in pod, which root Qdisc install skbedit
queue_mapping, send packets to eth0.3, vlan in host. In __dev_queue_xmit() of
eth0.3, clear the flag, does not select tx queue according to skb->queue_mapping
because there is no filters in clsact or tx Qdisc of this netdev.
Same action taked in eth0, ixgbe in Host.
- Avoid picking Tx queue for next packet. If we set xmit.skip_txqueue
in tx Qdisc (qdisc->enqueue()), the proper way to clear it is clearing it
in __dev_queue_xmit when processing next packets.
For performance reasons, use the static key. If user does not config the NET_EGRESS,
the patch will not be compiled.
+----+ +----+ +----+
| P1 | | P2 | | Pn |
+----+ +----+ +----+
| | |
+-----------+-----------+
|
| clsact/skbedit
| MQ
v
+-----------+-----------+
| q0 | q1 | qn
v v v
HTB/FQ HTB/FQ ... FIFO
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Jonathan Lemon <jonathan.lemon@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Alexander Lobakin <alobakin@pm.me>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Talal Ahmad <talalahmad@google.com>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Cc: Antoine Tenart <atenart@kernel.org>
Cc: Wei Wang <weiwan@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
As David Ahern suggested, the reasons for skb drops should be more
general and not be code based.
Therefore, rename SKB_DROP_REASON_PTYPE_ABSENT to
SKB_DROP_REASON_UNHANDLED_PROTO, which is used for the cases of no
L3 protocol handler, no L4 protocol handler, version extensions, etc.
From previous discussion, now we have the aim to make these reasons
more abstract and users based, avoiding code based.
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Increment rx_otherhost_dropped counter when packet dropped due to
mismatched dest MAC addr.
An example when this drop can occur is when manually crafting raw
packets that will be consumed by a user space application via a tap
device. For testing purposes local traffic was generated using trafgen
for the client and netcat to start a server
Tested: Created 2 netns, sent 1 packet using trafgen from 1 to the other
with "{eth(daddr=$INCORRECT_MAC...}", verified that iproute2 showed the
counter was incremented. (Also had to modify iproute2 to show the stat,
additional patch for that coming next.)
Signed-off-by: Jeffrey Ji <jeffreyji@google.com>
Reviewed-by: Brian Vazquez <brianvv@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220406172600.1141083-1-jeffreyjilinux@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There's a number of functions and static variables used
under net/core/ but not from the outside. We currently
dump most of them into netdevice.h. That bad for many
reasons:
- netdevice.h is very cluttered, hard to figure out
what the APIs are;
- netdevice.h is very long;
- we have to touch netdevice.h more which causes expensive
incremental builds.
Create a header under net/core/ and move some declarations.
The new header is also a bit of a catch-all but that's
fine, if we create more specific headers people will
likely over-think where their declaration fit best.
And end up putting them in netdevice.h, again.
More work should be done on splitting netdevice.h into more
targeted headers, but that'd be more time consuming so small
steps.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We have a bunch of functions which are only used under
net/core/ yet they get exported. Remove the exports.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This allows hardware flow offloading from Ethernet to WLAN on MT7622 SoC
Co-developed-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: David S. Miller <davem@davemloft.net>
In [1], Will raised a potential issue that the cfg80211 code,
which does (from a locking perspective)
rtnl_lock()
wiphy_lock()
rtnl_unlock()
might be suspectible to ABBA deadlocks, because rtnl_unlock()
calls netdev_run_todo(), which might end up calling rtnl_lock()
again, which could then deadlock (see the comment in the code
added here for the scenario).
Some back and forth and thinking ensued, but clearly this can't
happen if the net_todo_list is empty at the rtnl_unlock() here.
Clearly, the code here cannot actually put an entry on it, and
all other users of rtnl_unlock() will empty it since that will
always go through netdev_run_todo(), emptying the list.
So the only other way to get there would be to add to the list
and then unlock the RTNL without going through rtnl_unlock(),
which is only possible through __rtnl_unlock(). However, this
isn't exported and not used in many places, and none of them
seem to be able to unregister before using it.
Therefore, add a WARN_ON() in the code to ensure this invariant
won't be broken, so that the cfg80211 (or any similar) code
stays safe.
[1] https://lore.kernel.org/r/Yjzpo3TfZxtKPMAG@google.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20220404113847.0ee02e4a70da.Ic73d206e217db20fd22dcec14fe5442ca732804b@changeid
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The lockdep annotation lockdep_assert_softirq_will_run() expects that
either hard or soft interrupts are disabled because both guaranty that
the "raised" soft-interrupts will be processed once the context is left.
This triggers in flush_smp_call_function_from_idle() but it this case it
explicitly calls do_softirq() in case of pending softirqs.
Revert the "softirq will run" annotation in ____napi_schedule() and move
the check back to __netif_rx() as it was. Keep the IRQ-off assert in
____napi_schedule() because this is always required.
Fixes: fbd9a2ceba ("net: Add lockdep asserts to ____napi_schedule().")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Link: https://lore.kernel.org/r/YjhD3ZKWysyw8rc6@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Network drivers can call to netif_get_num_default_rss_queues to get the
default number of receive queues to use. Right now, this default number
is min(8, num_online_cpus()).
Instead, as suggested by Jakub, use the number of physical cores divided
by 2 as a way to avoid wasting CPU resources and to avoid using both CPU
threads, but still allowing to scale for high-end processors with many
cores.
As an exception, select 2 queues for processors with 2 cores, because
otherwise it won't take any advantage of RSS despite being SMP capable.
Tested: Processor Intel Xeon E5-2620 (2 sockets, 6 cores/socket, 2
threads/core). NIC Broadcom NetXtreme II BCM57810 (10GBps). Ran some
tests with `perf stat iperf3 -R`, with parallelisms of 1, 8 and 24,
getting the following results:
- Number of queues: 6 (instead of 8)
- Network throughput: not affected
- CPU usage: utilized 0.05-0.12 CPUs more than before (having 24 CPUs
this is only 0.2-0.5% higher)
- Reduced the number of context switches by 7-50%, being more noticeable
when using a higher number of parallel threads.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Link: https://lore.kernel.org/r/20220315091832.13873-1-ihuguet@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
____napi_schedule() needs to be invoked with disabled interrupts due to
__raise_softirq_irqoff (in order not to corrupt the per-CPU list).
____napi_schedule() needs also to be invoked from an interrupt context
so that the raised-softirq is processed while the interrupt context is
left.
Add lockdep asserts for both conditions.
While this is the second time the irq/softirq check is needed, provide a
generic lockdep_assert_softirq_will_run() which is used by both caller.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before adding yet another possibly contended atomic_long_t,
it is time to add per-cpu storage for existing ones:
dev->tx_dropped, dev->rx_dropped, and dev->rx_nohandler
Because many devices do not have to increment such counters,
allocate the per-cpu storage on demand, so that dev_get_stats()
does not have to spend considerable time folding zero counters.
Note that some drivers have abused these counters which
were supposed to be only used by core networking stack.
v4: should use per_cpu_ptr() in dev_get_stats() (Jakub)
v3: added a READ_ONCE() in netdev_core_stats_alloc() (Paolo)
v2: add a missing include (reported by kernel test robot <lkp@intel.com>)
Change in netdev_core_stats_alloc() (Jakub)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: jeffreyji <jeffreyji@google.com>
Reviewed-by: Brian Vazquez <brianvv@google.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/20220311051420.2608812-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
netdev_name_node_alt_create() and netdev_name_node_alt_destroy()
are only called by rtnetlink, so no need for exports.
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20220310223952.558779-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add reason for skb drops to __netif_receive_skb_core() when packet_type
not found to handle the skb. For this purpose, the drop reason
SKB_DROP_REASON_PTYPE_ABSENT is introduced. Take ether packets for
example, this case mainly happens when L3 protocol is not supported.
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace kfree_skb() used in sch_handle_ingress() with
kfree_skb_reason(). Following drop reasons are introduced:
SKB_DROP_REASON_TC_INGRESS
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace kfree_skb() used in do_xdp_generic() with kfree_skb_reason().
The drop reason SKB_DROP_REASON_XDP is introduced for this case.
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace kfree_skb() used in enqueue_to_backlog() with
kfree_skb_reason(). The skb rop reason SKB_DROP_REASON_CPU_BACKLOG is
introduced for the case of failing to enqueue the skb to the per CPU
backlog queue. The further reason can be backlog queue full or RPS
flow limition, and I think we needn't to make further distinctions.
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add reasons for skb drops to __dev_xmit_skb() by replacing
kfree_skb_list() with kfree_skb_list_reason(). The drop reason of
SKB_DROP_REASON_QDISC_DROP is introduced for qdisc enqueue fails.
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace kfree_skb() used in sch_handle_egress() with kfree_skb_reason().
The drop reason SKB_DROP_REASON_TC_EGRESS is introduced. Considering
the code path of tc egerss, we make it distinct with the drop reason
of SKB_DROP_REASON_QDISC_DROP in the next commit.
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit
baebdf48c3 ("net: dev: Makes sure netif_rx() can be invoked in any context.")
the function netif_rx() can be used in preemptible/thread context as
well as in interrupt context.
Use netif_rx().
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
The previous patches handled the delivery_time in the ingress path
before the routing decision is made. This patch can postpone clearing
delivery_time in a skb until knowing it is delivered locally and also
set the (rcv) timestamp if needed. This patch moves the
skb_clear_delivery_time() from dev.c to ip_local_deliver_finish()
and ip6_input_finish().
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The previous patches handled the delivery_time before sch_handle_ingress().
This patch can now set the skb->mono_delivery_time to flag the skb->tstamp
is used as the mono delivery_time (EDT) instead of the (rcv) timestamp
and also clear it with skb_clear_delivery_time() after
sch_handle_ingress(). This will make the bpf_redirect_*()
to keep the mono delivery_time and used by a qdisc (fq) of
the egress-ing interface.
A latter patch will postpone the skb_clear_delivery_time() until the
stack learns that the skb is being delivered locally and that will
make other kernel forwarding paths (ip[6]_forward) able to keep
the delivery_time also. Thus, like the previous patches on using
the skb->mono_delivery_time bit, calling skb_clear_delivery_time()
is not limited within the CONFIG_NET_INGRESS to avoid too many code
churns among this set.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In __skb_tstamp_tx(), it may clone the egress skb and queues the clone to
the sk_error_queue. The outgoing skb may have the mono delivery_time
while the (rcv) timestamp is expected for the clone, so the
skb->mono_delivery_time bit needs to be cleared from the clone.
This patch adds the skb->mono_delivery_time clearing to the existing
__net_timestamp() and use it in __skb_tstamp_tx().
The __net_timestamp() fast path usage in dev.c is changed to directly
call ktime_get_real() since the mono_delivery_time bit is not set at
that point.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A latter patch will set the skb->mono_delivery_time to flag the skb->tstamp
is used as the mono delivery_time (EDT) instead of the (rcv) timestamp.
skb_clear_tstamp() will then keep this delivery_time during forwarding.
This patch is to make the network tapping (with af_packet) to handle
the delivery_time stored in skb->tstamp.
Regardless of tapping at the ingress or egress, the tapped skb is
received by the af_packet socket, so it is ingress to the af_packet
socket and it expects the (rcv) timestamp.
When tapping at egress, dev_queue_xmit_nit() is used. It has already
expected skb->tstamp may have delivery_time, so it does
skb_clone()+net_timestamp_set() to ensure the cloned skb has
the (rcv) timestamp before passing to the af_packet sk.
This patch only adds to clear the skb->mono_delivery_time
bit in net_timestamp_set().
When tapping at ingress, it currently expects the skb->tstamp is either 0
or the (rcv) timestamp. Meaning, the tapping at ingress path
has already expected the skb->tstamp could be 0 and it will get
the (rcv) timestamp by ktime_get_real() when needed.
There are two cases for tapping at ingress:
One case is af_packet queues the skb to its sk_receive_queue.
The skb is either not shared or new clone created. The newly
added skb_clear_delivery_time() is called to clear the
delivery_time (if any) and set the (rcv) timestamp if
needed before the skb is queued to the sk_receive_queue.
Another case, the ingress skb is directly copied to the rx_ring
and tpacket_get_timestamp() is used to get the (rcv) timestamp.
The newly added skb_tstamp() is used in tpacket_get_timestamp()
to check the skb->mono_delivery_time bit before returning skb->tstamp.
As mentioned earlier, the tapping@ingress has already expected
the skb may not have the (rcv) timestamp (because no sk has asked
for it) and has handled this case by directly calling ktime_get_real().
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Offloading switch device drivers may be able to collect statistics of the
traffic taking place in the HW datapath that pertains to a certain soft
netdevice, such as VLAN. Add the necessary infrastructure to allow exposing
these statistics to the offloaded netdevice in question. The API was shaped
by the following considerations:
- Collection of HW statistics is not free: there may be a finite number of
counters, and the act of counting may have a performance impact. It is
therefore necessary to allow toggling whether HW counting should be done
for any particular SW netdevice.
- As the drivers are loaded and removed, a particular device may get
offloaded and unoffloaded again. At the same time, the statistics values
need to stay monotonic (modulo the eventual 64-bit wraparound),
increasing only to reflect traffic measured in the device.
To that end, the netdevice keeps around a lazily-allocated copy of struct
rtnl_link_stats64. Device drivers then contribute to the values kept
therein at various points. Even as the driver goes away, the struct stays
around to maintain the statistics values.
- Different HW devices may be able to count different things. The
motivation behind this patch in particular is exposure of HW counters on
Nvidia Spectrum switches, where the only practical approach to counting
traffic on offloaded soft netdevices currently is to use router interface
counters, and count L3 traffic. Correspondingly that is the statistics
suite added in this patch.
Other devices may be able to measure different kinds of traffic, and for
that reason, the APIs are built to allow uniform access to different
statistics suites.
- Because soft netdevices and offloading drivers are only loosely bound, a
netdevice uses a notifier chain to communicate with the drivers. Several
new notifiers, NETDEV_OFFLOAD_XSTATS_*, have been added to carry messages
to the offloading drivers.
- Devices can have various conditions for when a particular counter is
available. As the device is configured and reconfigured, the device
offload may become or cease being suitable for counter binding. A
netdevice can use a notifier type NETDEV_OFFLOAD_XSTATS_REPORT_USED to
ping offloading drivers and determine whether anyone currently implements
a given statistics suite. This information can then be propagated to user
space.
When the driver decides to unoffload a netdevice, it can use a
newly-added function, netdev_offload_xstats_report_delta(), to record
outstanding collected statistics, before destroying the HW counter.
This patch adds a helper, call_netdevice_notifiers_info_robust(), for
dispatching a notifier with the possibility of unwind when one of the
consumers bails. Given the wish to eventually get rid of the global
notifier block altogether, this helper only invokes the per-netns notifier
block.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I missed the obvious case where netif_ix() is invoked from hard-IRQ
context.
Disabling bottom halves is only needed in process context. This ensures
that the code remains on the current CPU and that the soft-interrupts
are processed at local_bh_enable() time.
In hard- and soft-interrupt context this is already the case and the
soft-interrupts will be processed once the context is left (at irq-exit
time).
Disable bottom halves if neither hard-interrupts nor soft-interrupts are
disabled. Update the kernel-doc, mention that interrupts must be enabled
if invoked from process context.
Fixes: baebdf48c3 ("net: dev: Makes sure netif_rx() can be invoked in any context.")
Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Link: https://lore.kernel.org/r/Yg05duINKBqvnxUc@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After recent patches, and in particular commits
faab39f63c ("net: allow out-of-order netdev unregistration") and
e5f80fcf86 ("ipv6: give an IPv6 dev to blackhole_netdev")
we no longer need the barrier implemented in rtnl_lock_unregistering().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the list of devices has N elements, netdev_wait_allrefs_any()
is called N times, and linkwatch_forget_dev() is called N*(N-1)/2 times.
Fix this by calling linkwatch_forget_dev() only once per device.
Fixes: faab39f63c ("net: allow out-of-order netdev unregistration")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220218065430.2613262-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sprinkle for each loops to allow netdevices to be unregistered
out of order, as their refs are released.
This prevents problems caused by dependencies between netdevs
which want to release references in their ->priv_destructor.
See commit d6ff94afd9 ("vlan: move dev_put into vlan_dev_uninit")
for example.
Eric has removed the only known ordering requirement in
commit c002496bab ("Merge branch 'ipv6-loopback'")
so let's try this and see if anything explodes...
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/20220215225310.3679266-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In prep for unregistering netdevs out of order move the netdev
state validation and change outside of the loop.
While at it modernize this code and use WARN() instead of
pr_err() + dump_stack().
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/20220215225310.3679266-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Disabling interrupts and in the RPS case locking input_pkt_queue is
split into local_irq_disable() and optional spin_lock().
This breaks on PREEMPT_RT because the spinlock_t typed lock can not be
acquired with disabled interrupts.
The sections in which the lock is acquired is usually short in a sense that it
is not causing long und unbounded latiencies. One exception is the
skb_flow_limit() invocation which may invoke a BPF program (and may
require sleeping locks).
By moving local_irq_disable() + spin_lock() into rps_lock(), we can keep
interrupts disabled on !PREEMPT_RT and enabled on PREEMPT_RT kernels.
Without RPS on a PREEMPT_RT kernel, the needed synchronisation happens
as part of local_bh_disable() on the local CPU.
____napi_schedule() is only invoked if sd is from the local CPU. Replace
it with __napi_schedule_irqoff() which already disables interrupts on
PREEMPT_RT as needed. Move this call to rps_ipi_queued() and rename the
function to napi_schedule_rps as suggested by Jakub.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dave suggested a while ago (eleven years by now) "Let's make netif_rx()
work in all contexts and get rid of netif_rx_ni()". Eric agreed and
pointed out that modern devices should use netif_receive_skb() to avoid
the overhead.
In the meantime someone added another variant, netif_rx_any_context(),
which behaves as suggested.
netif_rx() must be invoked with disabled bottom halves to ensure that
pending softirqs, which were raised within the function, are handled.
netif_rx_ni() can be invoked only from process context (bottom halves
must be enabled) because the function handles pending softirqs without
checking if bottom halves were disabled or not.
netif_rx_any_context() invokes on the former functions by checking
in_interrupts().
netif_rx() could be taught to handle both cases (disabled and enabled
bottom halves) by simply disabling bottom halves while invoking
netif_rx_internal(). The local_bh_enable() invocation will then invoke
pending softirqs only if the BH-disable counter drops to zero.
Eric is concerned about the overhead of BH-disable+enable especially in
regard to the loopback driver. As critical as this driver is, it will
receive a shortcut to avoid the additional overhead which is not needed.
Add a local_bh_disable() section in netif_rx() to ensure softirqs are
handled if needed.
Provide __netif_rx() which does not disable BH and has a lockdep assert
to ensure that interrupts are disabled. Use this shortcut in the
loopback driver and in drivers/net/*.c.
Make netif_rx_ni() and netif_rx_any_context() invoke netif_rx() so they
can be removed once they are no more users left.
Link: https://lkml.kernel.org/r/20100415.020246.218622820.davem@davemloft.net
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The preempt_disable() () section was introduced in commit
cece1945bf ("net: disable preemption before call smp_processor_id()")
and adds it in case this function is invoked from preemtible context and
because get_cpu() later on as been added.
The get_cpu() usage was added in commit
b0e28f1eff ("net: netif_rx() must disable preemption")
because ip_dev_loopback_xmit() invoked netif_rx() with enabled preemption
causing a warning in smp_processor_id(). The function netif_rx() should
only be invoked from an interrupt context which implies disabled
preemption. The commit
e30b38c298 ("ip: Fix ip_dev_loopback_xmit()")
was addressing this and replaced netif_rx() with in netif_rx_ni() in
ip_dev_loopback_xmit().
Based on the discussion on the list, the former patch (b0e28f1eff)
should not have been applied only the latter (e30b38c298).
Remove get_cpu() and preempt_disable() since the function is supposed to
be invoked from context with stable per-CPU pointers. Bottom halves have
to be disabled at this point because the function may raise softirqs
which need to be processed.
Link: https://lkml.kernel.org/r/20100415.013347.98375530.davem@davemloft.net
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
