net/devlink/leftover.c / net/core/devlink.c: 565b4824c39f ("devlink: change port event netdev notifier from per-net to global") f05bd8ebeb69 ("devlink: move code to a dedicated directory") 687125b5799c ("devlink: split out core code")
https://lore.kernel.org/all/20230208094657.379f2b1a@canb.auug.org.au/
Jiri Olsa [Thu, 9 Feb 2023 14:37:35 +0000 (15:37 +0100)]
tools/resolve_btfids: Pass HOSTCFLAGS as EXTRA_CFLAGS to prepare targets
Thorsten reported build issue with command line that defined extra
HOSTCFLAGS that were not passed into 'prepare' targets, but were
used to build resolve_btfids objects.
This results in build fail when these objects are linked together:
/usr/bin/ld: /build.../tools/bpf/resolve_btfids//libbpf/libbpf.a(libbpf-in.o):
relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a PIE \
object; recompile with -fPIE
Fixing this by passing HOSTCFLAGS in EXTRA_CFLAGS as part of
HOST_OVERRIDES variable for prepare targets.
- eth: ice: fix out-of-bounds KASAN warning in virtchnl
Current release - new code bugs:
- eth: mtk_eth_soc: enable special tag when any MAC uses DSA
Previous releases - always broken:
- core: fix sk->sk_txrehash default
- neigh: make sure used and confirmed times are valid
- mptcp: be careful on subflow status propagation on errors
- xfrm: prevent potential spectre v1 gadget in xfrm_xlate32_attr()
- phylink: move phy_device_free() to correctly release phy device
- eth: mlx5:
- fix crash unsetting rx-vlan-filter in switchdev mode
- fix hang on firmware reset
- serialize module cleanup with reload and remove"
* tag 'net-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (57 commits)
selftests: forwarding: lib: quote the sysctl values
net: mscc: ocelot: fix all IPv6 getting trapped to CPU when PTP timestamping is used
rds: rds_rm_zerocopy_callback() use list_first_entry()
net: txgbe: Update support email address
selftests: Fix failing VXLAN VNI filtering test
selftests: mptcp: stop tests earlier
selftests: mptcp: allow more slack for slow test-case
mptcp: be careful on subflow status propagation on errors
mptcp: fix locking for in-kernel listener creation
mptcp: fix locking for setsockopt corner-case
mptcp: do not wait for bare sockets' timeout
net: ethernet: mtk_eth_soc: fix DSA TX tag hwaccel for switch port 0
nfp: ethtool: fix the bug of setting unsupported port speed
txhash: fix sk->sk_txrehash default
net: ethernet: mtk_eth_soc: fix wrong parameters order in __xdp_rxq_info_reg()
net: ethernet: mtk_eth_soc: enable special tag when any MAC uses DSA
net: sched: sch: Fix off by one in htb_activate_prios()
igc: Add ndo_tx_timeout support
net: mana: Fix accessing freed irq affinity_hint
hv_netvsc: Allocate memory in netvsc_dma_map() with GFP_ATOMIC
...
Linus Torvalds [Thu, 9 Feb 2023 17:09:13 +0000 (09:09 -0800)]
Merge tag 'for-linus-2023020901' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid
Pull HID fixes from Benjamin Tissoires:
- fix potential infinite loop with a badly crafted HID device (Xin
Zhao)
- fix regression from 6.1 in USB logitech devices potentially making
their mouse wheel not working (Bastien Nocera)
- clean up in AMD sensors, which fixes a long time resume bug (Mario
Limonciello)
- few device small fixes and quirks
* tag 'for-linus-2023020901' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
HID: Ignore battery for ELAN touchscreen 29DF on HP
HID: amd_sfh: if no sensors are enabled, clean up
HID: logitech: Disable hi-res scrolling on USB
HID: core: Fix deadloop in hid_apply_multiplier.
HID: Ignore battery for Elan touchscreen on Asus TP420IA
HID: elecom: add support for TrackBall 056E:011C
Kees Cook [Thu, 5 Jan 2023 23:46:01 +0000 (15:46 -0800)]
net/i40e: Replace 0-length array with flexible array
Zero-length arrays are deprecated[1]. Replace struct i40e_lump_tracking's
"list" 0-length array with a flexible array. Detected with GCC 13,
using -fstrict-flex-arrays=3:
In function 'i40e_put_lump',
inlined from 'i40e_clear_interrupt_scheme' at drivers/net/ethernet/intel/i40e/i40e_main.c:5145:2:
drivers/net/ethernet/intel/i40e/i40e_main.c:278:27: warning: array subscript <unknown> is outside array bounds of 'u16[0]' {aka 'short unsigned int[]'} [-Warray-bounds=]
278 | pile->list[i] = 0;
| ~~~~~~~~~~^~~
drivers/net/ethernet/intel/i40e/i40e.h: In function 'i40e_clear_interrupt_scheme':
drivers/net/ethernet/intel/i40e/i40e.h:179:13: note: while referencing 'list'
179 | u16 list[0];
| ^~~~
Jan Sokolowski [Mon, 9 Jan 2023 14:11:20 +0000 (15:11 +0100)]
i40e: use ERR_PTR error print in i40e messages
In i40e_status removal patches, i40e_status conversion
to strings was removed in order to easily refactor
the code to use standard errornums. This however made it
more difficult for read error logs.
Use %pe formatter to print error messages in human-readable
format.
Jan Sokolowski [Mon, 9 Jan 2023 14:11:18 +0000 (15:11 +0100)]
i40e: use int for i40e_status
To prepare for removal of i40e_status, change the variables
from i40e_status to int. This eases the transition when values
are changed to return standard int error codes over enum i40e_status.
As such changes often also change variable orders, a cleanup
is also applied here to make variables conform to RCT and
some lines are also reformatted where applicable.
Jan Sokolowski [Mon, 9 Jan 2023 14:11:17 +0000 (15:11 +0100)]
i40e: Remove string printing for i40e_status
Remove the i40e_stat_str() function which prints the string
representation of the i40e_status error code. With upcoming changes
moving away from i40e_status, there will be no need for this function
Vladimir Oltean [Tue, 7 Feb 2023 18:31:17 +0000 (20:31 +0200)]
net: mscc: ocelot: fix all IPv6 getting trapped to CPU when PTP timestamping is used
While running this selftest which usually passes:
~/selftests/drivers/net/dsa# ./local_termination.sh eno0 swp0
TEST: swp0: Unicast IPv4 to primary MAC address [ OK ]
TEST: swp0: Unicast IPv4 to macvlan MAC address [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address, promisc [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address, allmulti [ OK ]
TEST: swp0: Multicast IPv4 to joined group [ OK ]
TEST: swp0: Multicast IPv4 to unknown group [ OK ]
TEST: swp0: Multicast IPv4 to unknown group, promisc [ OK ]
TEST: swp0: Multicast IPv4 to unknown group, allmulti [ OK ]
TEST: swp0: Multicast IPv6 to joined group [ OK ]
TEST: swp0: Multicast IPv6 to unknown group [ OK ]
TEST: swp0: Multicast IPv6 to unknown group, promisc [ OK ]
TEST: swp0: Multicast IPv6 to unknown group, allmulti [ OK ]
if I start PTP timestamping then run it again (debug prints added by me),
the unknown IPv6 MC traffic is seen by the CPU port even when it should
have been dropped:
~/selftests/drivers/net/dsa# ptp4l -i swp0 -2 -P -m
ptp4l[225.410]: selected /dev/ptp1 as PTP clock
[ 225.445746] mscc_felix 0000:00:00.5: ocelot_l2_ptp_trap_add: port 0 adding L2 PTP trap
[ 225.453815] mscc_felix 0000:00:00.5: ocelot_ipv4_ptp_trap_add: port 0 adding IPv4 PTP event trap
[ 225.462703] mscc_felix 0000:00:00.5: ocelot_ipv4_ptp_trap_add: port 0 adding IPv4 PTP general trap
[ 225.471768] mscc_felix 0000:00:00.5: ocelot_ipv6_ptp_trap_add: port 0 adding IPv6 PTP event trap
[ 225.480651] mscc_felix 0000:00:00.5: ocelot_ipv6_ptp_trap_add: port 0 adding IPv6 PTP general trap
ptp4l[225.488]: port 1: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[225.488]: port 0: INITIALIZING to LISTENING on INIT_COMPLETE
^C
~/selftests/drivers/net/dsa# ./local_termination.sh eno0 swp0
TEST: swp0: Unicast IPv4 to primary MAC address [ OK ]
TEST: swp0: Unicast IPv4 to macvlan MAC address [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address, promisc [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address, allmulti [ OK ]
TEST: swp0: Multicast IPv4 to joined group [ OK ]
TEST: swp0: Multicast IPv4 to unknown group [ OK ]
TEST: swp0: Multicast IPv4 to unknown group, promisc [ OK ]
TEST: swp0: Multicast IPv4 to unknown group, allmulti [ OK ]
TEST: swp0: Multicast IPv6 to joined group [ OK ]
TEST: swp0: Multicast IPv6 to unknown group [FAIL]
reception succeeded, but should have failed
TEST: swp0: Multicast IPv6 to unknown group, promisc [ OK ]
TEST: swp0: Multicast IPv6 to unknown group, allmulti [ OK ]
The PGID_MCIPV6 is configured correctly to not flood to the CPU,
I checked that.
Furthermore, when I disable back PTP RX timestamping (ptp4l doesn't do
that when it exists), packets are RX filtered again as they should be:
~/selftests/drivers/net/dsa# hwstamp_ctl -i swp0 -r 0
[ 218.202854] mscc_felix 0000:00:00.5: ocelot_l2_ptp_trap_del: port 0 removing L2 PTP trap
[ 218.212656] mscc_felix 0000:00:00.5: ocelot_ipv4_ptp_trap_del: port 0 removing IPv4 PTP event trap
[ 218.222975] mscc_felix 0000:00:00.5: ocelot_ipv4_ptp_trap_del: port 0 removing IPv4 PTP general trap
[ 218.233133] mscc_felix 0000:00:00.5: ocelot_ipv6_ptp_trap_del: port 0 removing IPv6 PTP event trap
[ 218.242251] mscc_felix 0000:00:00.5: ocelot_ipv6_ptp_trap_del: port 0 removing IPv6 PTP general trap
current settings:
tx_type 1
rx_filter 12
new settings:
tx_type 1
rx_filter 0
~/selftests/drivers/net/dsa# ./local_termination.sh eno0 swp0
TEST: swp0: Unicast IPv4 to primary MAC address [ OK ]
TEST: swp0: Unicast IPv4 to macvlan MAC address [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address, promisc [ OK ]
TEST: swp0: Unicast IPv4 to unknown MAC address, allmulti [ OK ]
TEST: swp0: Multicast IPv4 to joined group [ OK ]
TEST: swp0: Multicast IPv4 to unknown group [ OK ]
TEST: swp0: Multicast IPv4 to unknown group, promisc [ OK ]
TEST: swp0: Multicast IPv4 to unknown group, allmulti [ OK ]
TEST: swp0: Multicast IPv6 to joined group [ OK ]
TEST: swp0: Multicast IPv6 to unknown group [ OK ]
TEST: swp0: Multicast IPv6 to unknown group, promisc [ OK ]
TEST: swp0: Multicast IPv6 to unknown group, allmulti [ OK ]
So it's clear that something in the PTP RX trapping logic went wrong.
Looking a bit at the code, I can see that there are 4 typos, which
populate "ipv4" VCAP IS2 key filter fields for IPv6 keys.
VCAP IS2 keys of type OCELOT_VCAP_KEY_IPV4 and OCELOT_VCAP_KEY_IPV6 are
handled by is2_entry_set(). OCELOT_VCAP_KEY_IPV4 looks at
&filter->key.ipv4, and OCELOT_VCAP_KEY_IPV6 at &filter->key.ipv6.
Simply put, when we populate the wrong key field, &filter->key.ipv6
fields "proto.mask" and "proto.value" remain all zeroes (or "don't care").
So is2_entry_set() will enter the "else" of this "if" condition:
if (msk == 0xff && (val == IPPROTO_TCP || val == IPPROTO_UDP))
and proceed to ignore the "proto" field. The resulting rule will match
on all IPv6 traffic, trapping it to the CPU.
This is the reason why the local_termination.sh selftest sees it,
because control traps are stronger than the PGID_MCIPV6 used for
flooding (from the forwarding data path).
But the problem is in fact much deeper. We trap all IPv6 traffic to the
CPU, but if we're bridged, we set skb->offload_fwd_mark = 1, so software
forwarding will not take place and IPv6 traffic will never reach its
destination.
The fix is simple - correct the typos.
I was intentionally inaccurate in the commit message about the breakage
occurring when any PTP timestamping is enabled. In fact it only happens
when L4 timestamping is requested (HWTSTAMP_FILTER_PTP_V2_EVENT or
HWTSTAMP_FILTER_PTP_V2_L4_EVENT). But ptp4l requests a larger RX
timestamping filter than it needs for "-2": HWTSTAMP_FILTER_PTP_V2_EVENT.
I wanted people skimming through git logs to not think that the bug
doesn't affect them because they only use ptp4l in L2 mode.
Pietro Borrello [Tue, 7 Feb 2023 18:26:34 +0000 (18:26 +0000)]
rds: rds_rm_zerocopy_callback() use list_first_entry()
rds_rm_zerocopy_callback() uses list_entry() on the head of a list
causing a type confusion.
Use list_first_entry() to actually access the first element of the
rs_zcookie_queue list.
Jakub Kicinski [Thu, 9 Feb 2023 05:35:38 +0000 (21:35 -0800)]
Merge tag 'ipsec-2023-02-08' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec
Steffen Klassert says:
====================
ipsec 2023-02-08
1) Fix policy checks for nested IPsec tunnels when using
xfrm interfaces. From Benedict Wong.
2) Fix netlink message expression on 32=>64-bit
messages translators. From Anastasia Belova.
3) Prevent potential spectre v1 gadget in xfrm_xlate32_attr.
From Eric Dumazet.
4) Always consistently use time64_t in xfrm_timer_handler.
From Eric Dumazet.
5) Fix KCSAN reported bug: Multiple cpus can update use_time
at the same time. From Eric Dumazet.
6) Fix SCP copy from IPv4 to IPv6 on interfamily tunnel.
From Christian Hopps.
* tag 'ipsec-2023-02-08' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec:
xfrm: fix bug with DSCP copy to v6 from v4 tunnel
xfrm: annotate data-race around use_time
xfrm: consistently use time64_t in xfrm_timer_handler()
xfrm/compat: prevent potential spectre v1 gadget in xfrm_xlate32_attr()
xfrm: compat: change expression for switch in xfrm_xlate64
Fix XFRM-I support for nested ESP tunnels
====================
Jakub Kicinski [Thu, 9 Feb 2023 05:32:19 +0000 (21:32 -0800)]
Merge tag 'linux-can-next-for-6.3-20230208' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next
Marc Kleine-Budde says:
====================
can-next 2023-02-08
The 1st patch is by Oliver Hartkopp and cleans up the CAN_RAW's
raw_setsockopt() for CAN_RAW_FD_FRAMES.
The 2nd patch is by me and fixes the compilation if
CONFIG_CAN_CALC_BITTIMING is disabled. (Problem introduced in last
pull request to next-next.)
* tag 'linux-can-next-for-6.3-20230208' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next:
can: bittiming: can_calc_bittiming(): add missing parameter to no-op function
can: raw: use temp variable instead of rolling back config
====================
Jakub Kicinski [Thu, 9 Feb 2023 05:00:54 +0000 (21:00 -0800)]
Merge tag 'mlx5-next-netdev-deadlock' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Saeed Mahameed says:
====================
mlx5-next-netdev-deadlock
This series from Jiri solves a deadlock when removing a network namespace
with mlx5 devlink instance being in it.
The deadlock is between:
1) mlx5_ib->unregister_netdevice_notifier()
AND
2) mlx5_core->devlink_reload->cleanup_net()
To slove this introduced mlx5 netdev added/removed events to track uplink
netdev to be used for register_netdevice_notifier_dev_net() purposes.
* tag 'mlx5-next-netdev-deadlock' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux:
RDMA/mlx5: Track netdev to avoid deadlock during netdev notifier unregister
net/mlx5e: Propagate an internal event in case uplink netdev changes
net/mlx5e: Fix trap event handling
net/mlx5: Introduce CQE error syndrome
====================
Resolve this by converting to register_netdevice_notifier_dev_net()
which does not take pernet_ops_rwsem and moves the notifier block around
according to netdev it takes as arg.
Use previously introduced netdev added/removed events to track uplink
netdev to be used for register_netdevice_notifier_dev_net() purposes.
Jiri Pirko [Tue, 1 Nov 2022 14:27:43 +0000 (15:27 +0100)]
net/mlx5e: Propagate an internal event in case uplink netdev changes
Whenever uplink netdev is set/cleared, propagate newly introduced event
to inform notifier blocks netdev was added/removed.
Move the set() helper to core.c from header, introduce clear() and
netdev_added_event_replay() helpers. The last one is going to be called
from rdma driver, so export it.
Jiri Pirko [Thu, 24 Nov 2022 12:05:53 +0000 (13:05 +0100)]
net/mlx5e: Fix trap event handling
Current code does not return correct return value from event handler.
Fix it by returning NOTIFY_* and propagate err over newly introduce ctx
structure.
Jakub Kicinski [Thu, 9 Feb 2023 03:23:44 +0000 (19:23 -0800)]
Merge tag 'mlx5-fixes-2023-02-07' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5 fixes 2023-02-07
This series provides bug fixes to mlx5 driver.
* tag 'mlx5-fixes-2023-02-07' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
net/mlx5: Serialize module cleanup with reload and remove
net/mlx5: fw_tracer, Zero consumer index when reloading the tracer
net/mlx5: fw_tracer, Clear load bit when freeing string DBs buffers
net/mlx5: Expose SF firmware pages counter
net/mlx5: Store page counters in a single array
net/mlx5e: IPoIB, Show unknown speed instead of error
net/mlx5e: Fix crash unsetting rx-vlan-filter in switchdev mode
net/mlx5: Bridge, fix ageing of peer FDB entries
net/mlx5: DR, Fix potential race in dr_rule_create_rule_nic
net/mlx5e: Update rx ring hw mtu upon each rx-fcs flag change
====================
Rong Tao [Wed, 8 Feb 2023 01:04:41 +0000 (09:04 +0800)]
samples/bpf: Add openat2() enter/exit tracepoint to syscall_tp sample
Commit fe3300897cbf("samples: bpf: fix syscall_tp due to unused syscall")
added openat() syscall tracepoints. This patch adds support for
openat2() as well.
can: bittiming: can_calc_bittiming(): add missing parameter to no-op function
In commit 286c0e09e8e0 ("can: bittiming: can_changelink() pass extack
down callstack") a new parameter was added to can_calc_bittiming(),
however the static inline no-op (which is used if
CONFIG_CAN_CALC_BITTIMING is disabled) wasn't converted.
Add the new parameter to the static inline no-op of
can_calc_bittiming().
Oliver Hartkopp [Fri, 3 Feb 2023 09:08:07 +0000 (10:08 +0100)]
can: raw: use temp variable instead of rolling back config
Introduce a temporary variable to check for an invalid configuration
attempt from user space. Before this patch the value was copied to
the real config variable and rolled back in the case of an error.
bpf/docs: Update design QA to be consistent with kfunc lifecycle docs
Cong pointed out that there are some inconsistencies between the BPF design
QA and the lifecycle expectations documentation we added for kfuncs. Let's
update the QA file to be consistent with the kfunc docs, and add references
where it makes sense. Also document that modules may export kfuncs now.
v3:
- Grammar nit + ack from David
v2:
- Fix repeated word (s/defined defined/defined/)
David S. Miller [Wed, 8 Feb 2023 09:48:53 +0000 (09:48 +0000)]
Merge branch 'taprio-auto-qmaxsdu-new-tx'
Vladimir Oltean says:
====================
taprio automatic queueMaxSDU and new TXQ selection procedure
This patch set addresses 2 design limitations in the taprio software scheduler:
1. Software scheduling fundamentally prioritizes traffic incorrectly,
in a way which was inspired from Intel igb/igc drivers and does not
follow the inputs user space gives (traffic classes and TC to TXQ
mapping). Patch 05/15 handles this, 01/15 - 04/15 are preparations
for this work.
2. Software scheduling assumes that the gate for a traffic class closes
as soon as the next interval begins. But this isn't true.
If consecutive schedule entries have that traffic class gate open,
there is no "gate close" event and taprio should keep dequeuing from
that TC without interruptions. Patches 06/15 - 15/15 handle this.
Patch 10/15 is a generic Qdisc change required for this to work.
Future development directions which depend on this patch set are:
- Propagating the automatic queueMaxSDU calculation down to offloading
device drivers, instead of letting them calculate this, as
vsc9959_tas_guard_bands_update() does today.
- A software data path for tc-taprio with preemptible traffic and
Hold/Release events.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:40 +0000 (15:54 +0200)]
net/sched: taprio: don't segment unnecessarily
Improve commit 497cc00224cf ("taprio: Handle short intervals and large
packets") to only perform segmentation when skb->len exceeds what
taprio_dequeue() expects.
In practice, this will make the biggest difference when a traffic class
gate is always open in the schedule. This is because the max_frm_len
will be U32_MAX, and such large skb->len values as Kurt reported will be
sent just fine unsegmented.
What I don't seem to know how to handle is how to make sure that the
segmented skbs themselves are smaller than the maximum frame size given
by the current queueMaxSDU[tc]. Nonetheless, we still need to drop
those, otherwise the Qdisc will hang.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:39 +0000 (15:54 +0200)]
net/sched: taprio: split segmentation logic from qdisc_enqueue()
The majority of the taprio_enqueue()'s function is spent doing TCP
segmentation, which doesn't look right to me. Compilers shouldn't have a
problem in inlining code no matter how we write it, so move the
segmentation logic to a separate function.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:38 +0000 (15:54 +0200)]
net/sched: taprio: automatically calculate queueMaxSDU based on TC gate durations
taprio today has a huge problem with small TC gate durations, because it
might accept packets in taprio_enqueue() which will never be sent by
taprio_dequeue().
Since not much infrastructure was available, a kludge was added in
commit 497cc00224cf ("taprio: Handle short intervals and large
packets"), which segmented large TCP segments, but the fact of the
matter is that the issue isn't specific to large TCP segments (and even
worse, the performance penalty in segmenting those is absolutely huge).
In commit a54fc09e4cba ("net/sched: taprio: allow user input of per-tc
max SDU"), taprio gained support for queueMaxSDU, which is precisely the
mechanism through which packets should be dropped at qdisc_enqueue() if
they cannot be sent.
After that patch, it was necessary for the user to manually limit the
maximum MTU per TC. This change adds the necessary logic for taprio to
further limit the values specified (or not specified) by the user to
some minimum values which never allow oversized packets to be sent.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:37 +0000 (15:54 +0200)]
net/sched: keep the max_frm_len information inside struct sched_gate_list
I have one practical reason for doing this and one concerning correctness.
The practical reason has to do with a follow-up patch, which aims to mix
2 sources of max_sdu (one coming from the user and the other automatically
calculated based on TC gate durations @current link speed). Among those
2 sources of input, we must always select the smaller max_sdu value, but
this can change at various link speeds. So the max_sdu coming from the
user must be kept separated from the value that is operationally used
(the minimum of the 2), because otherwise we overwrite it and forget
what the user asked us to do.
To solve that, this patch proposes that struct sched_gate_list contains
the operationally active max_frm_len, and q->max_sdu contains just what
was requested by the user.
The reason having to do with correctness is based on the following
observation: the admin sched_gate_list becomes operational at a given
base_time in the future. Until then, it is inactive and applies no
shaping, all gates are open, etc. So the queueMaxSDU dropping shouldn't
apply either (this is a mechanism to ensure that packets smaller than
the largest gate duration for that TC don't hang the port; clearly it
makes little sense if the gates are always open).
Vladimir Oltean [Tue, 7 Feb 2023 13:54:36 +0000 (15:54 +0200)]
net/sched: taprio: warn about missing size table
Vinicius intended taprio to take the L1 overhead into account when
estimating packet transmission time through user input, specifically
through the qdisc size table (man tc-stab).
Without the overhead being specified, transmission times will be
underestimated and will cause late transmissions. For an offloading
driver, it might even cause TX hangs if there is no open gate large
enough to send the maximum sized packets for that TC (including L1
overhead). Properly knowing the L1 overhead will ensure that we are able
to auto-calculate the queueMaxSDU per traffic class just right, and
avoid these hangs due to head-of-line blocking.
We can't make the stab mandatory due to existing setups, but we can warn
the user that it's important with a warning netlink extack.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:35 +0000 (15:54 +0200)]
net/sched: make stab available before ops->init() call
Some qdiscs like taprio turn out to be actually pretty reliant on a well
configured stab, to not underestimate the skb transmission time (by
properly accounting for L1 overhead).
In a future change, taprio will need the stab, if configured by the
user, to be available at ops->init() time. It will become even more
important in upcoming work, when the overhead will be used for the
queueMaxSDU calculation that is passed to an offloading driver.
However, rcu_assign_pointer(sch->stab, stab) is called right after
ops->init(), making it unavailable, and I don't really see a good reason
for that.
Move it earlier, which nicely seems to simplify the error handling path
as well.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:34 +0000 (15:54 +0200)]
net/sched: taprio: calculate guard band against actual TC gate close time
taprio_dequeue_from_txq() looks at the entry->end_time to determine
whether the skb will overrun its traffic class gate, as if at the end of
the schedule entry there surely is a "gate close" event for it. Hint:
maybe there isn't.
For each schedule entry, introduce an array of kernel times which
actually tracks when in the future will there be an *actual* gate close
event for that traffic class, and use that in the guard band overrun
calculation.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:33 +0000 (15:54 +0200)]
net/sched: taprio: calculate budgets per traffic class
Currently taprio assumes that the budget for a traffic class expires at
the end of the current interval as if the next interval contains a "gate
close" event for this traffic class.
This is, however, an unfounded assumption. Allow schedule entry
intervals to be fused together for a particular traffic class by
calculating the budget until the gate *actually* closes.
This means we need to keep budgets per traffic class, and we also need
to update the budget consumption procedure.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:32 +0000 (15:54 +0200)]
net/sched: taprio: rename close_time to end_time
There is a confusion in terms in taprio which makes what is called
"close_time" to be actually used for 2 things:
1. determining when an entry "closes" such that transmitted skbs are
never allowed to overrun that time (?!)
2. an aid for determining when to advance and/or restart the schedule
using the hrtimer
It makes more sense to call this so-called "close_time" "end_time",
because it's not clear at all to me what "closes". Future patches will
hopefully make better use of the term "to close".
Vladimir Oltean [Tue, 7 Feb 2023 13:54:31 +0000 (15:54 +0200)]
net/sched: taprio: calculate tc gate durations
Current taprio code operates on a very simplistic (and incorrect)
assumption: that egress scheduling for a traffic class can only take
place for the duration of the current interval, or i.o.w., it assumes
that at the end of each schedule entry, there is a "gate close" event
for all traffic classes.
As an example, traffic sent with the schedule below will be jumpy, even
though all 8 TC gates are open, so there is absolutely no "gate close"
event (effectively a transition from BIT(tc)==1 to BIT(tc)==0 in
consecutive schedule entries):
This qdisc simply does not have what it takes in terms of logic to
*actually* compute the durations of traffic classes. Also, it does not
recognize the need to use this information on a per-traffic-class basis:
it always looks at entry->interval and entry->close_time.
This change proposes that each schedule entry has an array called
tc_gate_duration[tc]. This holds the information: "for how long will
this traffic class gate remain open, starting from *this* schedule
entry". If the traffic class gate is always open, that value is equal to
the cycle time of the schedule.
We'll also need to keep track, for the purpose of queueMaxSDU[tc]
calculation, what is the maximum time duration for a traffic class
having an open gate. This gives us directly what is the maximum sized
packet that this traffic class will have to accept. For everything else
it has to qdisc_drop() it in qdisc_enqueue().
Vladimir Oltean [Tue, 7 Feb 2023 13:54:30 +0000 (15:54 +0200)]
net/sched: taprio: give higher priority to higher TCs in software dequeue mode
Current taprio software implementation is haunted by the shadow of the
igb/igc hardware model. It iterates over child qdiscs in increasing
order of TXQ index, therefore giving higher xmit priority to TXQ 0 and
lower to TXQ N. According to discussions with Vinicius, that is the
default (perhaps even unchangeable) prioritization scheme used for the
NICs that taprio was first written for (igb, igc), and we have a case of
two bugs canceling out, resulting in a functional setup on igb/igc, but
a less sane one on other NICs.
To the best of my understanding, taprio should prioritize based on the
traffic class, so it should really dequeue starting with the highest
traffic class and going down from there. We get to the TXQ using the
tc_to_txq[] netdev property.
TXQs within the same TC have the same (strict) priority, so we should
pick from them as fairly as we can. We can achieve that by implementing
something very similar to q->curband from multiq_dequeue().
Since igb/igc really do have TXQ 0 of higher hardware priority than
TXQ 1 etc, we need to preserve the behavior for them as well. We really
have no choice, because in txtime-assist mode, taprio is essentially a
software scheduler towards offloaded child tc-etf qdiscs, so the TXQ
selection really does matter (not all igb TXQs support ETF/SO_TXTIME,
says Kurt Kanzenbach).
To preserve the behavior, we need a capability bit so that taprio can
determine if it's running on igb/igc, or on something else. Because igb
doesn't offload taprio at all, we can't piggyback on the
qdisc_offload_query_caps() call from taprio_enable_offload(), but
instead we need a separate call which is also made for software
scheduling.
Introduce two static keys to minimize the performance penalty on systems
which only have igb/igc NICs, and on systems which only have other NICs.
For mixed systems, taprio will have to dynamically check whether to
dequeue using one prioritization algorithm or using the other.
Simplify taprio_dequeue_from_txq() by noticing that we can goto one call
earlier than the previous skb_found label. This is possible because
we've unified the treatment of the child->ops->dequeue(child) return
call, we always try other TXQs now, instead of abandoning the root
dequeue completely if we failed in the peek() case.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:28 +0000 (15:54 +0200)]
net/sched: taprio: refactor one skb dequeue from TXQ to separate function
Future changes will refactor the TXQ selection procedure, and a lot of
stuff will become messy, the indentation of the bulk of the dequeue
procedure would increase, etc.
Break out the bulk of the function into a new one, which knows the TXQ
(child qdisc) we should perform a dequeue from.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:27 +0000 (15:54 +0200)]
net/sched: taprio: continue with other TXQs if one dequeue() failed
This changes the handling of an unlikely condition to not stop dequeuing
if taprio failed to dequeue the peeked skb in taprio_dequeue().
I've no idea when this can happen, but the only side effect seems to be
that the atomic_sub_return() call right above will have consumed some
budget. This isn't a big deal, since either that made us remain without
any budget (and therefore, we'd exit on the next peeked skb anyway), or
we could send some packets from other TXQs.
I'm making this change because in a future patch I'll be refactoring the
dequeue procedure to simplify it, and this corner case will have to go
away.
Vladimir Oltean [Tue, 7 Feb 2023 13:54:26 +0000 (15:54 +0200)]
net/sched: taprio: delete peek() implementation
There isn't any code in the network stack which calls taprio_peek().
We only see qdisc->ops->peek() being called on child qdiscs of other
classful qdiscs, never from the generic qdisc code. Whereas taprio is
never a child qdisc, it is always root.
This snippet of a comment from qdisc_peek_dequeued() seems to confirm:
/* we can reuse ->gso_skb because peek isn't called for root qdiscs */
Since I've been known to be wrong many times though, I'm not completely
removing it, but leaving a stub function in place which emits a warning.
David S. Miller [Wed, 8 Feb 2023 09:39:34 +0000 (09:39 +0000)]
Merge branch 'mptcp-fixes'
Matthieu Baerts says:
====================
mptcp: fixes for v6.2
Patch 1 clears resources earlier if there is no more reasons to keep
MPTCP sockets alive.
Patches 2 and 3 fix some locking issues visible in some rare corner
cases: the linked issues should be quite hard to reproduce.
Patch 4 makes sure subflows are correctly cleaned after the end of a
connection.
Patch 5 and 6 improve the selftests stability when running in a slow
environment by transfering data for a longer period on one hand and by
stopping the tests when all expected events have been observed on the
other hand.
All these patches fix issues introduced before v6.2.
====================
Matthieu Baerts [Tue, 7 Feb 2023 13:04:18 +0000 (14:04 +0100)]
selftests: mptcp: stop tests earlier
These 'endpoint' tests from 'mptcp_join.sh' selftest start a transfer in
the background and check the status during this transfer.
Once the expected events have been recorded, there is no reason to wait
for the data transfer to finish. It can be stopped earlier to reduce the
execution time by more than half.
For these tests, the exchanged data were not verified. Errors, if any,
were ignored but that's fine, plenty of other tests are looking at that.
It is then OK to mute stderr now that we are sure errors will be printed
(and still ignored) because the transfer is stopped before the end.
Paolo Abeni [Tue, 7 Feb 2023 13:04:17 +0000 (14:04 +0100)]
selftests: mptcp: allow more slack for slow test-case
A test-case is frequently failing on some extremely slow VMs.
The mptcp transfer completes before the script is able to do
all the required PM manipulation.
Address the issue in the simplest possible way, making the
transfer even more slow.
Additionally dump more info in case of failures, to help debugging
similar problems in the future and init dump_stats var.
Fixes: e274f7154008 ("selftests: mptcp: add subflow limits test-cases") Cc: [email protected] Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/323 Signed-off-by: Paolo Abeni <[email protected]> Reviewed-by: Matthieu Baerts <[email protected]> Signed-off-by: Matthieu Baerts <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Paolo Abeni [Tue, 7 Feb 2023 13:04:16 +0000 (14:04 +0100)]
mptcp: be careful on subflow status propagation on errors
Currently the subflow error report callback unconditionally
propagates the fallback subflow status to the owning msk.
If the msk is already orphaned, the above prevents the code
from correctly tracking the msk moving to the TCP_CLOSE state
and doing the appropriate cleanup.
All the above causes increasing memory usage over time and
sporadic self-tests failures.
There is a great deal of infrastructure trying to propagate
correctly the fallback subflow status to the owning mptcp socket,
e.g. via mptcp_subflow_eof() and subflow_sched_work_if_closed():
in the error propagation path we need only to cope with unorphaned
sockets.
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/339 Fixes: 15cc10453398 ("mptcp: deliver ssk errors to msk") Cc: [email protected] Signed-off-by: Paolo Abeni <[email protected]> Reviewed-by: Matthieu Baerts <[email protected]> Signed-off-by: Matthieu Baerts <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Paolo Abeni [Tue, 7 Feb 2023 13:04:15 +0000 (14:04 +0100)]
mptcp: fix locking for in-kernel listener creation
For consistency, in mptcp_pm_nl_create_listen_socket(), we need to
call the __mptcp_nmpc_socket() under the msk socket lock.
Note that as a side effect, mptcp_subflow_create_socket() needs a
'nested' lockdep annotation, as it will acquire the subflow (kernel)
socket lock under the in-kernel listener msk socket lock.
The current lack of locking is almost harmless, because the relevant
socket is not exposed to the user space, but in future we will add
more complexity to the mentioned helper, let's play safe.
Paolo Abeni [Tue, 7 Feb 2023 13:04:14 +0000 (14:04 +0100)]
mptcp: fix locking for setsockopt corner-case
We need to call the __mptcp_nmpc_socket(), and later subflow socket
access under the msk socket lock, or e.g. a racing connect() could
change the socket status under the hood, with unexpected results.
Paolo Abeni [Tue, 7 Feb 2023 13:04:13 +0000 (14:04 +0100)]
mptcp: do not wait for bare sockets' timeout
If the peer closes all the existing subflows for a given
mptcp socket and later the application closes it, the current
implementation let it survive until the timewait timeout expires.
While the above is allowed by the protocol specification it
consumes resources for almost no reason and additionally
causes sporadic self-tests failures.
Let's move the mptcp socket to the TCP_CLOSE state when there are
no alive subflows at close time, so that the allocated resources
will be freed immediately.
David S. Miller [Wed, 8 Feb 2023 09:16:07 +0000 (09:16 +0000)]
Merge branch 'micrel-lan8841-support'
Horatiu Vultur says:
====================
net: micrel: Add support for lan8841 PHY
Add support for lan8841 PHY.
The first patch add the support for lan8841 PHY which can run at
10/100/1000Mbit. It also has support for other features, but they are not
added in this series.
The second patch updates the documentation for the dt-bindings which is
similar to the ksz9131.
v3->v4:
- add space between defines and function names
- inside lan8841_config_init use only ret variable
v2->v3:
- reuse ksz9131_config_init
- allow only open-drain configuration
- change from single patch to a patch series
Horatiu Vultur [Tue, 7 Feb 2023 10:52:11 +0000 (11:52 +0100)]
net: micrel: Add support for lan8841 PHY
The LAN8841 is completely integrated triple-speed (10BASE-T/ 100BASE-TX/
1000BASE-T) Ethernet physical layer transceivers for transmission and
reception of data on standard CAT-5, as well as CAT-5e and CAT-6,
unshielded twisted pair (UTP) cables.
The LAN8841 offers the industry-standard GMII/MII as well as the RGMII.
Some of the features of the PHY are:
- Wake on LAN
- Auto-MDIX
- IEEE 1588-2008 (V2)
- LinkMD Capable diagnosis
Currently the patch offers support only for link configuration.
Vladimir Oltean [Tue, 7 Feb 2023 10:30:27 +0000 (12:30 +0200)]
net: ethernet: mtk_eth_soc: fix DSA TX tag hwaccel for switch port 0
Arınç reports that on his MT7621AT Unielec U7621-06 board and MT7623NI
Bananapi BPI-R2, packets received by the CPU over mt7530 switch port 0
(of which this driver acts as the DSA master) are not processed
correctly by software. More precisely, they arrive without a DSA tag
(in packet or in the hwaccel area - skb_metadata_dst()), so DSA cannot
demux them towards the switch's interface for port 0. Traffic from other
ports receives a skb_metadata_dst() with the correct port and is demuxed
properly.
Looking at mtk_poll_rx(), it becomes apparent that this driver uses the
skb vlan hwaccel area:
as a temporary storage for the VLAN hwaccel tag, or the DSA hwaccel tag.
If this is a DSA master it's a DSA hwaccel tag, and finally clears up
the skb VLAN hwaccel header.
I'm guessing that the problem is the (mis)use of API.
skb_vlan_tag_present() looks like this:
So if both vlan_proto and vlan_tci are zeroes, skb_vlan_tag_present()
returns precisely false. I don't know for sure what is the format of the
DSA hwaccel tag, but I surely know that lowermost 3 bits of vlan_proto
are 0 when receiving from port 0:
unsigned int port = vlan_proto & GENMASK(2, 0);
If the RX descriptor has no other bits set to non-zero values in
RX_DMA_VTAG, then the call to __vlan_hwaccel_put_tag() will not, in
fact, make the subsequent skb_vlan_tag_present() return true, because
it's implemented like this:
What we need to do to fix this problem (assuming this is the problem) is
to stop using skb->vlan_all as temporary storage for driver affairs, and
just create some local variables that serve the same purpose, but
hopefully better. Instead of calling skb_vlan_tag_present(), let's look
at a boolean has_hwaccel_tag which we set to true when the RX DMA
descriptors have something. Disambiguate based on netdev_uses_dsa()
whether this is a VLAN or DSA hwaccel tag, and only call
__vlan_hwaccel_put_tag() if we're certain it's a VLAN tag.
Arınç confirms that the treatment works, so this validates the
assumption.
Yu Xiao [Tue, 7 Feb 2023 10:16:50 +0000 (11:16 +0100)]
nfp: ethtool: fix the bug of setting unsupported port speed
Unsupported port speed can be set and cause error. Now fixing it
and return an error if setting unsupported speed.
This fix depends on the following, which was included in v6.2-rc1:
commit a61474c41e8c ("nfp: ethtool: support reporting link modes").
Fixes: 7c698737270f ("nfp: add support for .set_link_ksettings()") Signed-off-by: Yu Xiao <[email protected]> Signed-off-by: Simon Horman <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Fixes: 26859240e4ee ("txhash: Add socket option to control TX hash rethink behavior") Signed-off-by: Kevin Yang <[email protected]> Signed-off-by: Eric Dumazet <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Tariq Toukan [Mon, 6 Feb 2023 20:47:03 +0000 (22:47 +0200)]
net: ethernet: mtk_eth_soc: fix wrong parameters order in __xdp_rxq_info_reg()
Parameters 'queue_index' and 'napi_id' are passed in a swapped order.
Fix it here.
Fixes: 23233e577ef9 ("net: ethernet: mtk_eth_soc: rely on page_pool for single page buffers") Signed-off-by: Tariq Toukan <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Arınç ÜNAL [Sun, 5 Feb 2023 17:53:31 +0000 (20:53 +0300)]
net: ethernet: mtk_eth_soc: enable special tag when any MAC uses DSA
The special tag is only enabled when the first MAC uses DSA. However, it
must be enabled when any MAC uses DSA. Change the check accordingly.
This fixes hardware DSA untagging not working on the second MAC of the
MT7621 and MT7623 SoCs, and likely other SoCs too. Therefore, remove the
check that disables hardware DSA untagging for the second MAC of the MT7621
and MT7623 SoCs.
Fixes: a1f47752fd62 ("net: ethernet: mtk_eth_soc: disable hardware DSA untagging for second MAC") Co-developed-by: Richard van Schagen <[email protected]> Signed-off-by: Richard van Schagen <[email protected]> Signed-off-by: Arınç ÜNAL <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Jakub Kicinski [Wed, 8 Feb 2023 06:04:44 +0000 (22:04 -0800)]
Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2023-02-06 (ice)
This series contains updates to ice driver only.
Ani removes WQ_MEM_RECLAIM flag from workqueue to resolve
check_flush_dependency warning.
Michal fixes KASAN out-of-bounds warning.
Brett corrects behaviour for port VLAN Rx filters to prevent receiving
of unintended traffic.
Dan Carpenter fixes possible off by one issue.
Zhang Changzhong adjusts error path for switch recipe to prevent memory
leak.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
ice: switch: fix potential memleak in ice_add_adv_recipe()
ice: Fix off by one in ice_tc_forward_to_queue()
ice: Fix disabling Rx VLAN filtering with port VLAN enabled
ice: fix out-of-bounds KASAN warning in virtchnl
ice: Do not use WQ_MEM_RECLAIM flag for workqueue
====================
Sasha Neftin [Mon, 6 Feb 2023 23:58:18 +0000 (15:58 -0800)]
igc: Add ndo_tx_timeout support
On some platforms, 100/1000/2500 speeds seem to have sometimes problems
reporting false positive tx unit hang during stressful UDP traffic. Likely
other Intel drivers introduce responses to a tx hang. Update the 'tx hang'
comparator with the comparison of the head and tail of ring pointers and
restore the tx_timeout_factor to the previous value (one).
This can be test by using netperf or iperf3 applications.
Example:
iperf3 -s -p 5001
iperf3 -c 192.168.0.2 --udp -p 5001 --time 600 -b 0
Jakub Kicinski [Wed, 8 Feb 2023 05:40:40 +0000 (21:40 -0800)]
Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue
Tony Nguyen says:
====================
ice: various virtualization cleanups
Jacob Keller says:
This series contains a variety of refactors and cleanups in the VF code for
the ice driver. Its primary focus is cleanup and simplification of the VF
operations and addition of a few new operations that will be required by
Scalable IOV, as well as some other refactors needed for the handling of VF
subfunctions.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
ice: remove unnecessary virtchnl_ether_addr struct use
ice: introduce .irq_close VF operation
ice: introduce clear_reset_state operation
ice: convert vf_ops .vsi_rebuild to .create_vsi
ice: introduce ice_vf_init_host_cfg function
ice: add a function to initialize vf entry
ice: Pull common tasks into ice_vf_post_vsi_rebuild
ice: move ice_vf_vsi_release into ice_vf_lib.c
ice: move vsi_type assignment from ice_vsi_alloc to ice_vsi_cfg
ice: refactor VSI setup to use parameter structure
ice: drop unnecessary VF parameter from several VSI functions
ice: fix function comment referring to ice_vsi_alloc
ice: Add more usage of existing function ice_get_vf_vsi(vf)
====================
Haiyang Zhang [Mon, 6 Feb 2023 21:28:49 +0000 (13:28 -0800)]
net: mana: Fix accessing freed irq affinity_hint
After calling irq_set_affinity_and_hint(), the cpumask pointer is
saved in desc->affinity_hint, and will be used later when reading
/proc/irq/<num>/affinity_hint. So the cpumask variable needs to be
persistent. Otherwise, we are accessing freed memory when reading
the affinity_hint file.
Also, need to clear affinity_hint before free_irq(), otherwise there
is a one-time warning and stack trace during module unloading:
Jakub Kicinski [Wed, 8 Feb 2023 04:50:30 +0000 (20:50 -0800)]
Merge tag 'linux-can-fixes-for-6.2-20230207' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can
Marc Kleine-Budde says:
====================
can 2023-02-07
The patch is from Devid Antonio Filoni and fixes an address claiming
problem in the J1939 CAN protocol.
* tag 'linux-can-fixes-for-6.2-20230207' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can:
can: j1939: do not wait 250 ms if the same addr was already claimed
====================
James Hershaw [Mon, 6 Feb 2023 15:48:36 +0000 (16:48 +0100)]
nfp: flower: add check for flower VF netdevs for get/set_eeprom
Move the nfp_net_get_port_mac_by_hwinfo() check to ahead in the
get/set_eeprom() functions to in order to check for a VF netdev, which
this function does not support.
It is debatable if this is a fix or an enhancement, and we have chosen
to go for the latter. It does address a problem introduced by
commit 74b4f1739d4e ("nfp: flower: change get/set_eeprom logic and enable for flower reps").
However, the ethtool->len == 0 check avoids the problem manifesting as a
run-time bug (NULL pointer dereference of app).
Jakub Kicinski [Wed, 8 Feb 2023 04:18:51 +0000 (20:18 -0800)]
Merge branch 'mlxsw-misc-devlink-changes'
Petr Machata says:
====================
mlxsw: Misc devlink changes
This patchset adjusts mlxsw to recent devlink changes in net-next.
Patch #1 removes a devl_param_driverinit_value_set() call that was
unnecessary, but now additionally triggers a WARN_ON.
Patches #2-#4 are non-functional preparations for the following patches.
Patch #5 fixes a use-after-free that is triggered while changing network
namespaces.
Patch #6 makes mlxsw consistent with netdevsim by having mlxsw register
its devlink instance before its sub-objects. It helps us avoid a warning
described in the commit message.
====================
Ido Schimmel [Mon, 6 Feb 2023 15:39:23 +0000 (16:39 +0100)]
mlxsw: core: Register devlink instance before sub-objects
Recent changes made it possible to register the devlink instance before
its sub-objects and under the instance lock. Among other things, it
allows us to avoid warnings such as this one [1]. The warning is
generated because a buggy firmware is generating a health event during
driver initialization, before the devlink instance is registered.
Move the registration of the devlink instance to the beginning of the
initialization flow to avoid such problems.
A similar change was implemented in netdevsim in commit 82a3aef2e6af
("netdevsim: move devlink registration under the instance lock").
Ido Schimmel [Mon, 6 Feb 2023 15:39:22 +0000 (16:39 +0100)]
mlxsw: spectrum_acl_tcam: Move devlink param to TCAM code
Cited commit added 'DEVLINK_CMD_PARAM_DEL' notifications whenever the
network namespace of the devlink instance is changed. Specifically, the
notifications are generated after calling reload_down(), but before
calling reload_up(). At this stage, the data structures accessed while
reading the value of the "acl_region_rehash_interval" devlink parameter
are uninitialized, resulting in a use-after-free [1].
Fix by moving the registration and unregistration of the devlink
parameter to the TCAM code where it is actually used. This means that
the parameter is unregistered during reload_down() and then
re-registered during reload_up(), avoiding the use-after-free between
these two operations.
Reproducer:
# ip netns add test123
# devlink dev reload pci/0000:06:00.0 netns test123
[1]
BUG: KASAN: use-after-free in mlxsw_sp_acl_tcam_vregion_rehash_intrvl_get+0xb2/0xd0
Read of size 4 at addr ffff888162fd37d8 by task devlink/1323
[...]
Call Trace:
<TASK>
dump_stack_lvl+0x95/0xbd
print_report+0x181/0x4a1
kasan_report+0xdb/0x200
mlxsw_sp_acl_tcam_vregion_rehash_intrvl_get+0xb2/0xd0
mlxsw_sp_params_acl_region_rehash_intrvl_get+0x32/0x80
devlink_nl_param_fill.constprop.0+0x29a/0x11e0
devlink_param_notify.constprop.0+0xb9/0x250
devlink_notify_unregister+0xbc/0x470
devlink_reload+0x1aa/0x440
devlink_nl_cmd_reload+0x559/0x11b0
genl_family_rcv_msg_doit.isra.0+0x1f8/0x2e0
genl_rcv_msg+0x558/0x7f0
netlink_rcv_skb+0x170/0x440
genl_rcv+0x2d/0x40
netlink_unicast+0x53f/0x810
netlink_sendmsg+0x961/0xe80
__sys_sendto+0x2a4/0x420
__x64_sys_sendto+0xe5/0x1c0
do_syscall_64+0x38/0x80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
Fixes: 7d7e9169a3ec ("devlink: move devlink reload notifications back in between _down() and _up() calls") Signed-off-by: Ido Schimmel <[email protected]> Reviewed-by: Jiri Pirko <[email protected]> Signed-off-by: Petr Machata <[email protected]> Reviewed-by: Jacob Keller <[email protected]> Signed-off-by: Jakub Kicinski <[email protected]>
Danielle Ratson [Mon, 6 Feb 2023 15:39:18 +0000 (16:39 +0100)]
mlxsw: spectrum: Remove pointless call to devlink_param_driverinit_value_set()
The "acl_region_rehash_interval" devlink parameter is a "runtime"
parameter, making the call to devl_param_driverinit_value_set()
pointless. Before cited commit the function simply returned an error
(that was not checked), but now it emits a WARNING [1].
Vladimir Oltean [Mon, 6 Feb 2023 09:45:30 +0000 (11:45 +0200)]
net: enetc: add support for MAC Merge layer
Add PF driver support for viewing and changing the MAC Merge sublayer
parameters, and seeing the verification state machine's current state.
The verification handshake with the link partner is driven by hardware.
Shay Drory [Wed, 14 Dec 2022 20:16:23 +0000 (22:16 +0200)]
net/mlx5: Serialize module cleanup with reload and remove
Currently, remove and reload flows can run in parallel to module cleanup.
This design is error prone. For example: aux_drivers callbacks are called
from both cleanup and remove flows with different lockings, which can
cause a deadlock[1].
Hence, serialize module cleanup with reload and remove.
Shay Drory [Wed, 25 Jan 2023 15:39:36 +0000 (17:39 +0200)]
net/mlx5: fw_tracer, Zero consumer index when reloading the tracer
When tracer is reloaded, the device will log the traces at the
beginning of the log buffer. Also, driver is reading the log buffer in
chunks in accordance to the consumer index.
Hence, zero consumer index when reloading the tracer.
Shay Drory [Mon, 9 Jan 2023 13:27:40 +0000 (15:27 +0200)]
net/mlx5: fw_tracer, Clear load bit when freeing string DBs buffers
Whenever the driver is reading the string DBs into buffers, the driver
is setting the load bit, but the driver never clears this bit.
As a result, in case load bit is on and the driver query the device for
new string DBs, the driver won't read again the string DBs.
Fix it by clearing the load bit when query the device for new string
DBs.
Fixes: 2d69356752ff ("net/mlx5: Add support for fw live patch event") Signed-off-by: Shay Drory <[email protected]> Reviewed-by: Moshe Shemesh <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
Maher Sanalla [Sun, 22 Jan 2023 21:24:56 +0000 (23:24 +0200)]
net/mlx5: Expose SF firmware pages counter
Currently, each core device has VF pages counter which stores number of
fw pages used by its VFs and SFs.
The current design led to a hang when performing firmware reset on DPU,
where the DPU PFs stalled in sriov unload flow due to waiting on release
of SFs pages instead of waiting on only VFs pages.
Thus, Add a separate counter for SF firmware pages, which will prevent
the stall scenario described above.
Dragos Tatulea [Mon, 23 Jan 2023 14:06:32 +0000 (16:06 +0200)]
net/mlx5e: IPoIB, Show unknown speed instead of error
ethtool is returning an error for unknown speeds for the IPoIB interface:
$ ethtool ib0
netlink error: failed to retrieve link settings
netlink error: Invalid argument
netlink error: failed to retrieve link settings
netlink error: Invalid argument
Settings for ib0:
Link detected: no
After this change, ethtool will return success and show "unknown speed":
$ ethtool ib0
Settings for ib0:
Supported ports: [ ]
Supported link modes: Not reported
Supported pause frame use: No
Supports auto-negotiation: No
Supported FEC modes: Not reported
Advertised link modes: Not reported
Advertised pause frame use: No
Advertised auto-negotiation: No
Advertised FEC modes: Not reported
Speed: Unknown!
Duplex: Full
Auto-negotiation: off
Port: Other
PHYAD: 0
Transceiver: internal
Link detected: no
Fixes: eb234ee9d541 ("net/mlx5e: IPoIB, Add support for get_link_ksettings in ethtool") Signed-off-by: Dragos Tatulea <[email protected]> Reviewed-by: Gal Pressman <[email protected]> Reviewed-by: Tariq Toukan <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
Amir Tzin [Sun, 8 Jan 2023 13:54:46 +0000 (15:54 +0200)]
net/mlx5e: Fix crash unsetting rx-vlan-filter in switchdev mode
Moving to switchdev mode with rx-vlan-filter on and then setting it off
causes the kernel to crash since fs->vlan is freed during nic profile
cleanup flow.
RX VLAN filtering is not supported in switchdev mode so unset it when
changing to switchdev and restore its value when switching back to
legacy.
Vlad Buslov [Thu, 26 Jan 2023 13:47:12 +0000 (14:47 +0100)]
net/mlx5: Bridge, fix ageing of peer FDB entries
SWITCHDEV_FDB_ADD_TO_BRIDGE event handler that updates FDB entry 'lastuse'
field is only executed for eswitch that owns the entry. However, if peer
entry processed packets at least once it will have hardware counter 'used'
value greater than entry 'lastuse' from that point on, which will cause FDB
entry not being aged out.
net/mlx5: DR, Fix potential race in dr_rule_create_rule_nic
Selecting builder should be protected by the lock to prevent the case
where a new rule sets a builder in the nic_matcher while the previous
rule is still using the nic_matcher.
Fixing this issue and cleaning the error flow.
Fixes: b9b81e1e9382 ("net/mlx5: DR, For short chains of STEs, avoid allocating ste_arr dynamically") Signed-off-by: Yevgeny Kliteynik <[email protected]> Reviewed-by: Alex Vesker <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
Adham Faris [Sun, 8 Jan 2023 16:09:32 +0000 (18:09 +0200)]
net/mlx5e: Update rx ring hw mtu upon each rx-fcs flag change
rq->hw_mtu is used in function en_rx.c/mlx5e_skb_from_cqe_mpwrq_linear()
to catch oversized packets. If FCS is concatenated to the end of the
packet then the check should be updated accordingly.
Rx rings initialization (mlx5e_init_rxq_rq()) invoked for every new set
of channels, as part of mlx5e_safe_switch_params(), unknowingly if it
runs with default configuration or not. Current rq->hw_mtu
initialization assumes default configuration and ignores
params->scatter_fcs_en flag state.
Fix this, by accounting for params->scatter_fcs_en flag state during
rq->hw_mtu initialization.
In addition, updating rq->hw_mtu value during ingress traffic might
lead to packets drop and oversize_pkts_sw_drop counter increase with no
good reason. Hence we remove this optimization and switch the set of
channels with a new one, to make sure we don't get false positives on
the oversize_pkts_sw_drop counter.
Fixes: 102722fc6832 ("net/mlx5e: Add support for RXFCS feature flag") Signed-off-by: Adham Faris <[email protected]> Reviewed-by: Tariq Toukan <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
====================
sched: cpumask: improve on cpumask_local_spread() locality
cpumask_local_spread() currently checks local node for presence of i'th
CPU, and then if it finds nothing makes a flat search among all non-local
CPUs. We can do it better by checking CPUs per NUMA hops.
This has significant performance implications on NUMA machines, for example
when using NUMA-aware allocated memory together with NUMA-aware IRQ
affinity hints.
Performance tests from patch 8 of this series for mellanox network
driver show:
TCP multi-stream, using 16 iperf3 instances pinned to 16 cores (with aRFS on).
Active cores: 64,65,72,73,80,81,88,89,96,97,104,105,112,113,120,121
+-------------------------+-----------+------------------+------------------+
| | BW (Gbps) | TX side CPU util | RX side CPU util |
+-------------------------+-----------+------------------+------------------+
| Baseline | 52.3 | 6.4 % | 17.9 % |
+-------------------------+-----------+------------------+------------------+
| Applied on TX side only | 52.6 | 5.2 % | 18.5 % |
+-------------------------+-----------+------------------+------------------+
| Applied on RX side only | 94.9 | 11.9 % | 27.2 % |
+-------------------------+-----------+------------------+------------------+
| Applied on both sides | 95.1 | 8.4 % | 27.3 % |
+-------------------------+-----------+------------------+------------------+
Bottleneck in RX side is released, reached linerate (~1.8x speedup).
~30% less cpu util on TX.
====================
Yury Norov [Sat, 21 Jan 2023 04:24:36 +0000 (20:24 -0800)]
lib/cpumask: update comment for cpumask_local_spread()
Now that we have an iterator-based alternative for a very common case
of using cpumask_local_spread for all cpus in a row, it's worth to
mention that in comment to cpumask_local_spread().
Tariq Toukan [Sat, 21 Jan 2023 04:24:35 +0000 (20:24 -0800)]
net/mlx5e: Improve remote NUMA preferences used for the IRQ affinity hints
In the IRQ affinity hints, replace the binary NUMA preference (local /
remote) with the improved for_each_numa_hop_cpu() API that minds the
actual distances, so that remote NUMAs with short distance are preferred
over farther ones.
This has significant performance implications when using NUMA-aware
allocated memory (follow [1] and derivatives for example).
[1]
drivers/net/ethernet/mellanox/mlx5/core/en_main.c :: mlx5e_open_channel()
int cpu = cpumask_first(mlx5_comp_irq_get_affinity_mask(priv->mdev, ix));
Performance tests:
TCP multi-stream, using 16 iperf3 instances pinned to 16 cores (with aRFS on).
Active cores: 64,65,72,73,80,81,88,89,96,97,104,105,112,113,120,121
+-------------------------+-----------+------------------+------------------+
| | BW (Gbps) | TX side CPU util | RX side CPU util |
+-------------------------+-----------+------------------+------------------+
| Baseline | 52.3 | 6.4 % | 17.9 % |
+-------------------------+-----------+------------------+------------------+
| Applied on TX side only | 52.6 | 5.2 % | 18.5 % |
+-------------------------+-----------+------------------+------------------+
| Applied on RX side only | 94.9 | 11.9 % | 27.2 % |
+-------------------------+-----------+------------------+------------------+
| Applied on both sides | 95.1 | 8.4 % | 27.3 % |
+-------------------------+-----------+------------------+------------------+
Bottleneck in RX side is released, reached linerate (~1.8x speedup).
~30% less cpu util on TX.
* CPU util on active cores only.
Setups details (similar for both sides):
NIC: ConnectX6-DX dual port, 100 Gbps each.
Single port used in the tests.
$ lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 256
On-line CPU(s) list: 0-255
Thread(s) per core: 2
Core(s) per socket: 64
Socket(s): 2
NUMA node(s): 16
Vendor ID: AuthenticAMD
CPU family: 25
Model: 1
Model name: AMD EPYC 7763 64-Core Processor
Stepping: 1
CPU MHz: 2594.804
BogoMIPS: 4890.73
Virtualization: AMD-V
L1d cache: 32K
L1i cache: 32K
L2 cache: 512K
L3 cache: 32768K
NUMA node0 CPU(s): 0-7,128-135
NUMA node1 CPU(s): 8-15,136-143
NUMA node2 CPU(s): 16-23,144-151
NUMA node3 CPU(s): 24-31,152-159
NUMA node4 CPU(s): 32-39,160-167
NUMA node5 CPU(s): 40-47,168-175
NUMA node6 CPU(s): 48-55,176-183
NUMA node7 CPU(s): 56-63,184-191
NUMA node8 CPU(s): 64-71,192-199
NUMA node9 CPU(s): 72-79,200-207
NUMA node10 CPU(s): 80-87,208-215
NUMA node11 CPU(s): 88-95,216-223
NUMA node12 CPU(s): 96-103,224-231
NUMA node13 CPU(s): 104-111,232-239
NUMA node14 CPU(s): 112-119,240-247
NUMA node15 CPU(s): 120-127,248-255
..
The recently introduced sched_numa_hop_mask() exposes cpumasks of CPUs
reachable within a given distance budget, wrap the logic for iterating over
all (distance, mask) values inside an iterator macro.
projects / linux.git / log