Currently on stable trees we have support for netmem/devmem RX but not
TX. It is not safe to forward/redirect an RX unreadable netmem packet
into the device's TX path, as the device may call dma-mapping APIs on
dma addrs that should not be passed to it.
Fix this by preventing the xmit of unreadable skbs.
Tested by configuring tc redirect:
sudo tc qdisc add dev eth1 ingress
sudo tc filter add dev eth1 ingress protocol ip prio 1 flower ip_proto \
tcp src_ip 192.168.1.12 action mirred egress redirect dev eth1
Before, I see unreadable skbs in the driver's TX path passed to dma
mapping APIs.
After, I don't see unreadable skbs in the driver's TX path passed to dma
mapping APIs.
Fixes: 65249feb6b ("net: add support for skbs with unreadable frags")
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Mina Almasry <almasrymina@google.com>
Link: https://patch.msgid.link/20250306215520.1415465-1-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
napi_schedule() is expected to be called either:
* From an interrupt, where raised softirqs are handled on IRQ exit
* From a softirq disabled section, where raised softirqs are handled on
the next call to local_bh_enable().
* From a softirq handler, where raised softirqs are handled on the next
round in do_softirq(), or further deferred to a dedicated kthread.
Other bare tasks context may end up ignoring the raised NET_RX vector
until the next random softirq handling opportunity, which may not
happen before a while if the CPU goes idle afterwards with the tick
stopped.
Such "misuses" have been detected on several places thanks to messages
of the kind:
"NOHZ tick-stop error: local softirq work is pending, handler #08!!!"
For example:
__raise_softirq_irqoff
__napi_schedule
rtl8152_runtime_resume.isra.0
rtl8152_resume
usb_resume_interface.isra.0
usb_resume_both
__rpm_callback
rpm_callback
rpm_resume
__pm_runtime_resume
usb_autoresume_device
usb_remote_wakeup
hub_event
process_one_work
worker_thread
kthread
ret_from_fork
ret_from_fork_asm
And also:
* drivers/net/usb/r8152.c::rtl_work_func_t
* drivers/net/netdevsim/netdev.c::nsim_start_xmit
There is a long history of issues of this kind:
019edd01d1 ("ath10k: sdio: Add missing BH locking around napi_schdule()")
3300685893 ("idpf: disable local BH when scheduling napi for marker packets")
e3d5d70cb4 ("net: lan78xx: fix "softirq work is pending" error")
e55c27ed9c ("mt76: mt7615: add missing bh-disable around rx napi schedule")
c0182aa985 ("mt76: mt7915: add missing bh-disable around tx napi enable/schedule")
970be1dff2 ("mt76: disable BH around napi_schedule() calls")
019edd01d1 ("ath10k: sdio: Add missing BH locking around napi_schdule()")
30bfec4fec ("can: rx-offload: can_rx_offload_threaded_irq_finish(): add new function to be called from threaded interrupt")
e63052a5dd ("mlx5e: add add missing BH locking around napi_schdule()")
83a0c6e589 ("i40e: Invoke softirqs after napi_reschedule")
bd4ce941c8 ("mlx4: Invoke softirqs after napi_reschedule")
8cf699ec84 ("mlx4: do not call napi_schedule() without care")
ec13ee8014 ("virtio_net: invoke softirqs after __napi_schedule")
This shows that relying on the caller to arrange a proper context for
the softirqs to be handled while calling napi_schedule() is very fragile
and error prone. Also fixing them can also prove challenging if the
caller may be called from different kinds of contexts.
Therefore fix this from napi_schedule() itself with waking up ksoftirqd
when softirqs are raised from task contexts.
Reported-by: Paul Menzel <pmenzel@molgen.mpg.de>
Reported-by: Jakub Kicinski <kuba@kernel.org>
Reported-by: Francois Romieu <romieu@fr.zoreil.com>
Closes: https://lore.kernel.org/lkml/354a2690-9bbf-4ccb-8769-fa94707a9340@molgen.mpg.de/
Cc: Breno Leitao <leitao@debian.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250223221708.27130-1-frederic@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After the previous commit is finally safe to revert commit dbae2b0628
("net: skb: introduce and use a single page frag cache"): do it here.
The intended goal of such change was to counter a performance regression
introduced by commit 3226b158e6 ("net: avoid 32 x truesize
under-estimation for tiny skbs").
Unfortunately, the blamed commit introduces another regression for the
virtio_net driver. Such a driver calls napi_alloc_skb() with a tiny
size, so that the whole head frag could fit a 512-byte block.
The single page frag cache uses a 1K fragment for such allocation, and
the additional overhead, under small UDP packets flood, makes the page
allocator a bottleneck.
Thanks to commit bf9f1baa27 ("net: add dedicated kmem_cache for
typical/small skb->head"), this revert does not re-introduce the
original regression. Actually, in the relevant test on top of this
revert, I measure a small but noticeable positive delta, just above
noise level.
The revert itself required some additional mangling due to recent updates
in the affected code.
Suggested-by: Eric Dumazet <edumazet@google.com>
Fixes: dbae2b0628 ("net: skb: introduce and use a single page frag cache")
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The following sequence is basically illegal when dev was fetched
without lookup because dev_net(dev) might be different after holding
rtnl_net_lock():
net = dev_net(dev);
rtnl_net_lock(net);
Let's use rtnl_net_dev_lock() in unregister_netdev().
Note that there is no real bug in unregister_netdev() for now
because RTNL protects the scope even if dev_net(dev) is changed
before/after RTNL.
Fixes: 00fb982393 ("dev: Hold per-netns RTNL in (un)?register_netdev().")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250217191129.19967-4-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After the cited commit, dev_net(dev) is fetched before holding RTNL
and passed to __unregister_netdevice_notifier_net().
However, dev_net(dev) might be different after holding RTNL.
In the reported case [0], while removing a VF device, its netns was
being dismantled and the VF was moved to init_net.
So the following sequence is basically illegal when dev was fetched
without lookup:
net = dev_net(dev);
rtnl_net_lock(net);
Let's use a new helper rtnl_net_dev_lock() to fix the race.
It fetches dev_net_rcu(dev), bumps its net->passive, and checks if
dev_net_rcu(dev) is changed after rtnl_net_lock().
[0]:
BUG: KASAN: slab-use-after-free in notifier_call_chain (kernel/notifier.c:75 (discriminator 2))
Read of size 8 at addr ffff88810cefb4c8 by task test-bridge-lag/21127
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
Call Trace:
<TASK>
dump_stack_lvl (lib/dump_stack.c:123)
print_report (mm/kasan/report.c:379 mm/kasan/report.c:489)
kasan_report (mm/kasan/report.c:604)
notifier_call_chain (kernel/notifier.c:75 (discriminator 2))
call_netdevice_notifiers_info (net/core/dev.c:2011)
unregister_netdevice_many_notify (net/core/dev.c:11551)
unregister_netdevice_queue (net/core/dev.c:11487)
unregister_netdev (net/core/dev.c:11635)
mlx5e_remove (drivers/net/ethernet/mellanox/mlx5/core/en_main.c:6552 drivers/net/ethernet/mellanox/mlx5/core/en_main.c:6579) mlx5_core
auxiliary_bus_remove (drivers/base/auxiliary.c:230)
device_release_driver_internal (drivers/base/dd.c:1275 drivers/base/dd.c:1296)
bus_remove_device (./include/linux/kobject.h:193 drivers/base/base.h:73 drivers/base/bus.c:583)
device_del (drivers/base/power/power.h:142 drivers/base/core.c:3855)
mlx5_rescan_drivers_locked (./include/linux/auxiliary_bus.h:241 drivers/net/ethernet/mellanox/mlx5/core/dev.c:333 drivers/net/ethernet/mellanox/mlx5/core/dev.c:535 drivers/net/ethernet/mellanox/mlx5/core/dev.c:549) mlx5_core
mlx5_unregister_device (drivers/net/ethernet/mellanox/mlx5/core/dev.c:468) mlx5_core
mlx5_uninit_one (./include/linux/instrumented.h:68 ./include/asm-generic/bitops/instrumented-non-atomic.h:141 drivers/net/ethernet/mellanox/mlx5/core/main.c:1563) mlx5_core
remove_one (drivers/net/ethernet/mellanox/mlx5/core/main.c:965 drivers/net/ethernet/mellanox/mlx5/core/main.c:2019) mlx5_core
pci_device_remove (./include/linux/pm_runtime.h:129 drivers/pci/pci-driver.c:475)
device_release_driver_internal (drivers/base/dd.c:1275 drivers/base/dd.c:1296)
unbind_store (drivers/base/bus.c:245)
kernfs_fop_write_iter (fs/kernfs/file.c:338)
vfs_write (fs/read_write.c:587 (discriminator 1) fs/read_write.c:679 (discriminator 1))
ksys_write (fs/read_write.c:732)
do_syscall_64 (arch/x86/entry/common.c:52 (discriminator 1) arch/x86/entry/common.c:83 (discriminator 1))
entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:130)
RIP: 0033:0x7f6a4d5018b7
Fixes: 7fb1073300 ("net: Hold rtnl_net_lock() in (un)?register_netdevice_notifier_dev_net().")
Reported-by: Yael Chemla <ychemla@nvidia.com>
Closes: https://lore.kernel.org/netdev/146eabfe-123c-4970-901e-e961b4c09bc3@nvidia.com/
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250217191129.19967-3-kuniyu@amazon.com
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>
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>
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>
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>
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>
