| Age | Commit message (Collapse) | Author |
|---|
|
It has been observed that on certain chipsets a peer can be assigned
peer_id=0. For reception of non-aggregated MPDUs this is fine as
ath11k_dp_rx_h_find_peer() has a fallback case where it locates the peer
based upon the source MAC address. On an aggregated link, the mpdu_start
header is only populated by hardware on the first sub-MSDU. This causes
the peer resolution to be skipped for the subsequent MSDUs and the
encryption type of these frames to be set to an incorrect value,
resulting in these MSDUs being dropped by ieee80211.
ath11k_pci 0000:03:00.0: data rx skb 000000002f4b704d len 1534 peer xx:xx:xx:xx:xx:xx 0 ucast sn 3063 he160 rate_idx 9 vht_nss 2 freq 5240 band 1 flag 0x40d1a fcs-err 0 mic-err 0 amsdu-more 0 peer_id 0 first_msdu 1 last_msdu 0
ath11k_pci 0000:03:00.0: data rx skb 0000000038acd580 len 1534 peer (null) 0 ucast sn 3063 he160 rate_idx 9 vht_nss 2 freq 5240 band 1 flag 0x40d00 fcs-err 0 mic-err 0 amsdu-more 0 peer_id 0 first_msdu 0 last_msdu 1
Remove the null peer_id checks in ath11k_dp_rx_h_find_peer() and
ath11k_hal_rx_parse_mon_status_tlv(), allowing peers with an assigned ID
of 0 to be resolved.
Tested-on: QCA2066 hw2.1 PCI WLAN.HSP.1.1-03926.13-QCAHSPSWPL_V2_SILICONZ_CE-2.52297.9
Fixes: 2167fa606c0f ("ath11k: Add support for RX decapsulation offload")
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Signed-off-by: Matthew Leach <matthew.leach@collabora.com>
Reviewed-by: P Praneesh <praneesh.p@oss.qualcomm.com>
Link: https://patch.msgid.link/20260424-ath11k-null-peerid-workaround-v4-1-252b224d3cf6@collabora.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
After 4 missed beacons since last Rx, the firmware will send an NDP to the
AP. If the NDP is ACK'ed, it'll reset the missed_beacons_since_last_rx
counter.
Disconnecting after 4 beacons doesn't give enough time to the firmware
to send the NDP.
Wait until we get 6 missed beacons since last Rx before disconnecting.
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Link: https://patch.msgid.link/20260515151352.c4ed0d849f98.Iefa2e8be9edfc74683997eea60bb53c2002f31f0@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
|
|
Clearly, from a user perspective, it must be valid to configure
WoWLAN (which can include network detection) and then suspend
while not connected to a network, or even without an interface
at all (WoWLAN config is handled on a per-wiphy basis). Since
mac80211 doesn't distinguish these cases and simply calls the
driver to suspend whenever WoWLAN is configured, the driver has
to cleanly handle the case where it's called for WoWLAN but no
(BSS) interface exists.
Remove the WARN_ON(), move the print so it doesn't get done in
this case, and keep returning 1 to disconnect everything.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260515151352.0c55d1135409.I54f8be0e2aa28cfb1cb1dcf3b2d2d8fe75b4397b@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
|
|
CNVI_PMU_STEP_FLOW is a PRPH register, not a UMAC PRPH register.
Use iwl_read_prph() instead of iwl_read_umac_prph() to read it
correctly.
Signed-off-by: Moriya Itzchaki <moriya.itzchaki@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260515151352.3a69fa2dbda7.I8d96635a9c06a835b05a10b6d66c8a9299676246@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
|
|
On old devices such as 7265D, rates are still encoded in version 1
format, which doesn't use the CCK/OFDM rate index (0-3/0-7) but
rather their PLCP value (e.g. 10 for 1 Mbps CCK rate.)
While introducing v3 rates, I changed the driver from internally
handling v1 rates and converting to v2, to internally handling v3
and converting to v1 or v2 according to the firmware. I accordingly
changed the code in iwl_mvm_mac80211_idx_to_hwrate() to no longer
have different values for different APIs. This was correct.
However, I later reverted this part of the change, because it was
reported that I had broken beacon rates, causing a FW assert/crash.
This caused TX_CMD rates to be set incorrectly, potentially causing
a warning when reported back from the device as having been used.
Fix this (hopefully correctly now) by handling beacon rates in the
TX_CMD that's embedded in the beacon template command separately.
Restore iwl_mvm_mac80211_idx_to_hwrate() to return only the rate
index, not PLCP value, fixing the real TX_CMD.
Cc: stable@vger.kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260515151351.7407e293dff7.I4ea1a17f8fe99c933d3f3e30d077cf4246125c3e@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
|
|
In iwl_mld_remove_link, the link->fw_id is saved at the beginning of the
function so we have it after we freed the link.
But the link pointer can be NULL, and is not checked when the fw_id is
stored.
Fix it by simply freeing the link at the end of the function.
fFixes: 0e66a39f4f0e ("wifi: iwlwifi: fix potential use after free in iwl_mld_remove_link()")
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Link: https://patch.msgid.link/20260515151351.371f40fc6711.I6a82cfe9655564e9c5731af91c36493b26b1208e@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
|
|
When iwlwifi firmware crashes (e.g., NMI_INTERRUPT_UNKNOWN on Intel
BE201/Wi-Fi 7), iwl_mld_nic_error() sets mld->fw_status.in_hw_restart
to true. However, iwl_mld_tx_from_txq() does not check this flag before
dequeuing frames from mac80211 and pushing them to the transport layer.
Since the firmware is dead, iwl_trans_tx() returns -EIO for each frame,
which then gets freed immediately. Under high-throughput conditions
(e.g., Tailscale UDP traffic or active SSH sessions), this creates a
tight dequeue-send-fail-free loop that wastes CPU cycles and generates
rapid skb allocation churn, leading to memory pressure from slab
fragmentation.
The RX path already has this guard (iwl_mld_rx_mpdu checks
in_hw_restart at rx.c:1906), and so does the TXQ allocation worker
(iwl_mld_add_txqs_wk at tx.c:156). Add the same guard to
iwl_mld_tx_from_txq() to stop all TX during firmware restart.
Frames left in mac80211's TXQs are naturally drained after restart
completes, when queue reallocation triggers iwl_mld_tx_from_txq()
via iwl_mld_add_txq_list(), or when new upper-layer traffic invokes
wake_tx_queue.
Tested on ASUS Zenbook 14 UX3405CA with Intel BE201 (Wi-Fi 7) on
kernel 6.19.5 where the firmware crashes approximately every 10-15
minutes under Tailscale traffic.
Fixes: d1e879ec600f ("wifi: iwlwifi: add iwlmld sub-driver")
Cc: stable@vger.kernel.org
Signed-off-by: Sheroz Juraev <goodmartiandev@gmail.com>
Link: https://patch.msgid.link/20260315081221.2678478-1-goodmartiandev@gmail.com
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
|
|
When the TLC notification disables AMSDU for a TID, the MLD driver sets
max_tid_amsdu_len to the sentinel value 1. The TSO segmentation path in
iwl_mld_tx_tso_segment() checks for zero but not for this sentinel,
allowing it to reach the num_subframes calculation:
num_subframes = (max_tid_amsdu_len + pad) / (subf_len + pad)
= (1 + 2) / (1534 + 2) = 0
This zero propagates to iwl_tx_tso_segment() which sets:
gso_size = num_subframes * mss = 0
Calling skb_gso_segment() with gso_size=0 creates over 32000 tiny
segments from a single GSO skb. This floods the TX ring with ~1024
micro-frames (the rest are purged), creating a massive burst of TX
completion events that can lead to memory corruption and a subsequent
use-after-free in TCP's retransmit queue (refcount underflow in
tcp_shifted_skb, NULL deref in tcp_rack_detect_loss).
The MVM driver is immune because it checks mvmsta->amsdu_enabled before
reaching the num_subframes calculation. The MLD driver has no equivalent
bitmap check and relies solely on max_tid_amsdu_len, which does not
catch the sentinel value.
Fix this by detecting the sentinel value (max_tid_amsdu_len == 1) at the
existing check and falling back to non-AMSDU TSO segmentation. Also add
a WARN_ON_ONCE guard after the num_subframes division as defense-in-depth
to catch any future code paths that produce zero through a different
mechanism.
Suggested-by: Miriam Rachel Korenblit <miriam.rachel.korenblit@intel.com>
Fixes: d1e879ec600f ("wifi: iwlwifi: add iwlmld sub-driver")
Signed-off-by: Cole Leavitt <cole@unwrap.rs>
Link: https://patch.msgid.link/20260405054145.1064152-3-cole@unwrap.rs
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
|
|
Reserve NET_SKB_PAD + NET_IP_ALIGN bytes of headroom for received packets
to avoid skb head reallocation when pushing protocol headers into the skb.
Tested-by: Xuegang Lu <xuegang.lu@airoha.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://patch.msgid.link/20260513-airoha-rx-headroom-v1-1-bd87798e422d@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
cgx_speed_mbps has 13 elements but RESP_LINKSTAT_SPEED can yield values
0-15. If it returns a value >= 13, this causes an out-of-bounds array
access. Add a bounds check and default to speed 0 if the index is out of
range.
Fixes: 61071a871ea6 ("octeontx2-af: Forward CGX link notifications to PFs")
Cc: Sunil Goutham <sgoutham@marvell.com>
Cc: Linu Cherian <lcherian@marvell.com>
Cc: Geetha sowjanya <gakula@marvell.com>
Cc: hariprasad <hkelam@marvell.com>
Cc: Subbaraya Sundeep <sbhatta@marvell.com>
Cc: Andrew Lunn <andrew+netdev@lunn.ch>
Cc: stable <stable@kernel.org>
Signed-off-by: Sam Daly <sam@samdaly.ie>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://patch.msgid.link/2026051352-refined-demise-e88d@gregkh
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
ifb_dev_init() allocates dp->tx_private to dev->num_tx_queues
entries via kzalloc_objs(*txp, dev->num_tx_queues). Both IFB
per-queue RX and TX stats live in those entries: ifb_xmit() updates
txp->rx_stats using the skb queue mapping, ifb_ri_tasklet() updates
txp->tx_stats, and ifb_stats64() aggregates both over
dev->num_tx_queues.
The ethtool stats callbacks instead size and walk the per-queue
stats with dev->real_num_rx_queues and dev->real_num_tx_queues. With
an asymmetric device where the RX queue count exceeds the TX queue
count, for example:
ip link add name ifb10 numtxqueues 1 numrxqueues 8 type ifb
ethtool -S ifb10
ifb_get_ethtool_stats() indexes past the tx_private allocation and
copies adjacent slab data through ETHTOOL_GSTATS.
Use dev->num_tx_queues consistently for the stats strings, the
stats count, and the stats data walks. This reports one RX stats
group and one TX stats group for each backing ifb_q_private entry,
which is the queue set IFB can actually populate.
Reproduced under UML+KASAN at v7.1-rc2:
BUG: KASAN: slab-out-of-bounds in ifb_fill_stats_data+0x3c/0xae
Read of size 8 at addr 0000000062dbd228 by task ethtool/36
ifb_fill_stats_data+0x3c/0xae
ifb_get_ethtool_stats+0xc0/0x129
__dev_ethtool+0x1ca5/0x363c
dev_ethtool+0x123/0x1b3
dev_ioctl+0x56c/0x744
sock_do_ioctl+0x15f/0x1b2
sock_ioctl+0x4d5/0x50a
sys_ioctl+0xd8b/0xde9
With the patch applied, the same UML+KASAN repro is silent and
ethtool -S ifb10 reports only the stats backed by the single
allocated tx_private entry.
Fixes: a21ee5b2fcb8 ("net: ifb: support ethtools stats")
Cc: stable@vger.kernel.org
Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com>
Link: https://patch.msgid.link/20260514013739.3549624-1-michael.bommarito@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When setting vports max TX speed during LAG activation or bond state
changes, the code iterates over all eswitch vports. However, some
vports may not be enabled yet.
Skip vports that are not enabled to avoid sending FW commands for
uninitialized vports. Save the LAG aggregated speed in the vport
struct so it can be applied when the vport is enabled later.
Fixes: 50f1d188c580 ("net/mlx5: Propagate LAG effective max_tx_speed to vports")
Signed-off-by: Or Har-Toov <ohartoov@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260513063640.334132-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
After IPsec policy/state TX rules are added, any TC flow rule, which
forwards packets to uplink, is modified to forward to IPsec TX tables.
As these tables are destroyed dynamically, whenever there is no
reference to them, the destinations of this kind of rules must be
restored to uplink, unless there is no destination for that rule.
The flow rules FLOW_ACTION_ACCEPT, DROP, TRAP, GOTO and SAMPLE do not
have a destination port, and thus out_count = 0.
At cleanup time of the rules in mlx5_esw_ipsec_modify_flow_dests
we call mlx5_eswitch_restore_ipsec_rule but as the above types
do not have a destination we get an underflow of out_count, as
the port is passed, which is esw_attr->out_count - 1.
This change avoids calling mlx5_eswitch_restore_ipsec_rule when
there are no output destinations and thus avoids the underflow.
Fixes: d1569537a837 ("net/mlx5e: Modify and restore TC rules for IPSec TX rules")
Signed-off-by: Jeroen Massar <jmassar@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260513063302.333761-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
If mlx5e_rx_res_rss_set_rxfh() fails during mlx5e_create_rxfh_context(),
the RSS context is not cleaned up.
This leaves a stale entry in 'res->rss[rss_idx]' that occupies a context
slot.
Destroy the RSS context before returning the error.
Fixes: 6c2509d44636 ("net/mlx5e: Add error flow for ethtool -X command")
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260513062737.333259-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
We have macros to retrieve endpoint numbers now.
Use them.
Signed-off-by: Oliver Neukum <oneukum@suse.com>
Link: https://patch.msgid.link/20260513132916.39896-1-oneukum@suse.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
rvu_rep_rsrc_init() allocates queue memory before calling
otx2_init_hw_resources(). When hardware resource setup fails,
otx2_init_hw_resources() already unwinds the partially initialized
SQ, CQ, and aura state before returning an error. The representor
error path then calls otx2_free_hw_resources() again and can free
the same resources a second time.
Fix this by splitting the cleanup labels so that a failure from
otx2_init_hw_resources() only releases queue memory. Keep the
otx2_free_hw_resources() call for failures that happen after
hardware resource initialization completed successfully.
The bug was first flagged by an experimental analysis tool we are
developing for kernel memory-management bugs while analyzing
v6.13-rc1. The tool is still under development and is not yet publicly
available. Manual inspection confirms that the bug is still
present in v7.1-rc3.
Runtime validation was not performed because reproducing this path
requires OcteonTX2 representor hardware.
Fixes: 3937b7308d4f ("octeontx2-pf: Create representor netdev")
Cc: stable@vger.kernel.org # v6.13+
Signed-off-by: Zilin Guan <zilin@seu.edu.cn>
Signed-off-by: Dawei Feng <dawei.feng@seu.edu.cn>
Reviewed-by: Geetha sowjanya <gakula@marvell.com>
Link: https://patch.msgid.link/20260513151320.213260-1-dawei.feng@seu.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
ksz_phylink_get_caps() is a bit different from other generic methods.
It has a dev_ops->get_caps() call in the middle of the function, and it
does other stuff before (set some supported_interfaces) and after (set
lpi_interfaces from supported_interfaces).
Whereas the dev_ops->get_caps() methods set mac_capabilities and
(optionally) logically OR the supported_interfaces with that of the PCS.
The idea is that this can be expressed simpler, and avoid a indirect
function call to dev_ops->get_caps(). If we tail-call the common
ksz_phylink_get_caps() from individual phylink_get_caps() methods, we do
reorder the settings, but in an inconsequential way (the transfer from
supported_interfaces to lpi_interfaces still sees a complete list of the
supported_interfaces).
Remove the no longer used get_caps() callbacl the ksz_dev_ops.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Bastien Curutchet (Schneider Electric) <bastien.curutchet@bootlin.com>
Link: https://patch.msgid.link/20260512-clean-ksz-2nd-series-v1-7-c00f6ce037fa@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Mirror operations are handled through a common function that redirects
the treatment to ksz_dev_ops callbacks. This layer of indirection isn't
needed since we now have a dsa_switch_ops for each switch family.
Remove this indirection layer for KSZ switches, by connecting the
ksz_dev_ops :: mirror_add() and mirror_del() operations directly to
dsa_switch_ops.
Remove the now unused mirror callbacks from ksz_dev_ops.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Bastien Curutchet (Schneider Electric) <bastien.curutchet@bootlin.com>
Link: https://patch.msgid.link/20260512-clean-ksz-2nd-series-v1-6-c00f6ce037fa@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
FDB and MDB operations are handled through a common function that
redirects the treatment to ksz_dev_ops callbacks. This layer of
indirection isn't needed since we now have a dsa_switch_ops for each kind
of switch.
Remove one indirection layer for KSZ switches, by connecting the
ksz_dev_ops :: fdb_dump(), fdb_add(), fdb_del(), mdb_add() and mdb_del()
operations directly to dsa_switch_ops.
Remove the FDB and MDB operations from ksz_dev_ops.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Bastien Curutchet (Schneider Electric) <bastien.curutchet@bootlin.com>
Link: https://patch.msgid.link/20260512-clean-ksz-2nd-series-v1-5-c00f6ce037fa@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
VLAN operations are handled through a common function that redirects the
treatment to ksz_dev_ops callbacks. This level of indirection isn't
needed since we now have a dsa_switch_ops for each kind of switch.
Remove this useless layer of indirection by connecting directly the VLAN
operations to the relevant dsa_switch_ops.
Adapt their prototypes to match dsa_switch_ops expectations.
Remove the now unused VLAN callbacks from ksz_dev_ops.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Bastien Curutchet (Schneider Electric) <bastien.curutchet@bootlin.com>
Link: https://patch.msgid.link/20260512-clean-ksz-2nd-series-v1-4-c00f6ce037fa@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
MTU changing is done through a common function that redirects the
treatment to a specific ksz_dev_ops callback. This layer of indirection
isn't needed since we now have a dsa_switch_ops struct for each switch
family.
Remove this indirection layer in MTU changing for KSZ switches, by
directly connecting the ksz_dev_ops :: change_mtu() implementations to
dsa_switch_ops.
Remove the no longer used change_mtu() callback from ksz_dev_ops
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Bastien Curutchet (Schneider Electric) <bastien.curutchet@bootlin.com>
Link: https://patch.msgid.link/20260512-clean-ksz-2nd-series-v1-3-c00f6ce037fa@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
dsa_switch_ops :: set_ageing_time() goes through ksz_set_ageing_time(),
further dispatched through ksz_dev_ops :: set_ageing_time(). Only
ksz9477 and lan937x provide an implementation for this, so remove the
(optional) method from ksz8463_switch_ops, ksz87xx_switch_ops,
ksz88xx_switch_ops. Also, hook up ksz9477 and lan937x dsa_switch_ops
directly to their respective implementations.
Every switch family provides a dsa_switch_ops :: port_fast_age()
implementation, which is dispatched through ksz_dev_ops ::
flush_dyn_mac_table(). Remove the dev_ops indirection and connect the
flush_dyn_mac_table() methods directly to their respective dsa_switch_ops.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Bastien Curutchet (Schneider Electric) <bastien.curutchet@bootlin.com>
Link: https://patch.msgid.link/20260512-clean-ksz-2nd-series-v1-2-c00f6ce037fa@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
ksz_dev_ops::phylink_mac_link_up() isn't used anywhere.
Remove it.
Signed-off-by: Bastien Curutchet (Schneider Electric) <bastien.curutchet@bootlin.com>
Link: https://patch.msgid.link/20260512-clean-ksz-2nd-series-v1-1-c00f6ce037fa@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
mlx5e_tx_reporter_timeout_recover() accesses sq->netdev after
mlx5e_safe_reopen_channels() has torn down and freed the channel (and
its embedded SQs). Replace the three sq->netdev references with
priv->netdev which is safe because priv outlives channel teardown.
The netdev_err() call already used priv->netdev for this reason; make
the trylock/unlock and health_channel_eq_recover calls consistent.
This fixes the following KASAN splat:
BUG: KASAN: use-after-free in mlx5e_tx_reporter_timeout_recover+0x1dd/0x360 [mlx5_core]
Read of size 8 at addr ffff889860ed0b28 by task kworker/u113:2/5277
Call Trace:
mlx5e_tx_reporter_timeout_recover+0x1dd/0x360 [mlx5_core]
devlink_health_reporter_recover+0xa2/0x150
devlink_health_report+0x254/0x7c0
mlx5e_reporter_tx_timeout+0x297/0x380 [mlx5_core]
mlx5e_tx_timeout_work+0x109/0x170 [mlx5_core]
process_one_work+0x677/0xf20
worker_thread+0x51f/0xd90
kthread+0x3a5/0x810
ret_from_fork+0x208/0x400
ret_from_fork_asm+0x1a/0x30
Fixes: 83ac0304a2d7 ("net/mlx5e: Fix deadlocks between devlink and netdev instance locks")
Cc: stable@vger.kernel.org
Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Matt Fleming <mfleming@cloudflare.com>
Link: https://patch.msgid.link/20260513112226.140512-1-matt@readmodwrite.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
ovpn updates dev->dstats from both process and softirq contexts. In
particular, TCP paths may run from socket callbacks, workqueues or
strparser work, while UDP receive and ovpn's ndo_start_xmit path may
update the same per-device dstats from BH context.
Add ovpn device drop-stat helpers that disable BHs around
dev_dstats_rx_dropped() and dev_dstats_tx_dropped(), and use them for
drop accounting.
The successful RX dev_dstats_rx_add() update is already covered by the
BH-disabled section around gro_cells_receive(). For the successful TCP
TX dev_dstats_tx_add() update, replace the existing preempt-disabled
section with a BH-disabled one.
Fixes: 11851cbd60ea ("ovpn: implement TCP transport")
Signed-off-by: Ralf Lici <ralf@mandelbit.com>
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
|
|
Cross-merge networking fixes after downstream PR (net-7.1-rc4).
No conflicts, or adjacent changes.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
While deleting an existing ovpn interface, there is a very
narrow window where adding a new peer via netlink may cause
the netdevice to hang and prevent its unregistration.
It may happen during ovpn_dellink(), when all existing peers are
freed and the device is queued for deregistration, but a
CMD_PEER_NEW message comes in adding a new peer that takes again
a reference to the netdev.
At this point there is no way to release the device because we are
under the assumption that all peers were already released.
Fix the race condition by releasing all peers in ndo_uninit(),
when the netdevice has already been removed from the netdev
list.
Also ovpn_peer_add() has now an extra check that forces the
function to bail out if the device reg_state is not REGISTERED.
This way any incoming CMD_PEER_NEW racing with the interface
deletion routine will simply stop before adding the peer.
Note that the above check happens while holding the netdev_lock
to prevent racing netdev state changes.
ovpn_dellink() is now empty and can be removed.
Reported-by: Hyunwoo Kim <imv4bel@gmail.com>
Closes: https://lore.kernel.org/netdev/aaVgJ16edTfQkYbx@v4bel/
Suggested-by: Sabrina Dubroca <sd@queasysnail.net>
Fixes: 80747caef33d ("ovpn: introduce the ovpn_peer object")
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
|
|
ovpn_nl_peer_new_doit()'s error path calls ovpn_peer_release() directly
rather than ovpn_peer_put(), bypassing the kref. The accompanying
comment ("peer was not yet hashed, thus it is not used in any context")
holds for UDP but not for TCP.
For UDP, the ovpn_socket union uses the .ovpn arm and never points back
at a peer; UDP encap_recv looks up peers via the not-yet-populated
hashtables, so the new peer is unreachable until ovpn_peer_add()
publishes it.
For TCP, ovpn_socket_new() sets ovpn_sock->peer and
ovpn_tcp_socket_attach() publishes ovpn_sock via rcu_assign_sk_user_data().
From that moment until ovpn_socket_release() detaches in the error path,
the TCP fd is fully wired: userspace recvmsg / sendmsg / close / poll
on the fd, as well as the strparser-driven ovpn_tcp_rcv() path, can
reach the peer through sk_user_data -> ovpn_sock->peer and bump its
refcount via ovpn_peer_hold().
ovpn_tcp_socket_wait_finish() (called inside ovpn_socket_release())
drains strparser and the tx work, but does not synchronize with
userspace syscall callers that already hold a peer reference. If
ovpn_nl_peer_modify() or ovpn_peer_add() returns an error while such
a caller is in flight - notably an ovpn_tcp_recvmsg() blocked in
__skb_recv_datagram() on peer->tcp.user_queue - the direct
ovpn_peer_release() destroys the peer while the caller still holds
the reference, and the eventual ovpn_peer_put() from that caller
operates on freed memory.
Replace the direct destructor call with ovpn_peer_put() so the kref
correctly defers destruction until the last reference is dropped.
In the common case where no concurrent user is present, behaviour is
unchanged: the kref hits zero immediately and ovpn_peer_release_kref()
runs the same destructor.
With this conversion ovpn_peer_release() has no callers outside peer.c
- ovpn_peer_release_kref() in the same translation unit is the only
remaining user - so make it static and drop its declaration from
peer.h.
Fixes: 11851cbd60ea ("ovpn: implement TCP transport")
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: David Carlier <devnexen@gmail.com>
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
|
|
ovpn_tcp_close() loads the ovpn_socket via rcu_dereference_sk_user_data()
under rcu_read_lock(), takes a reference on sock->peer, caches the peer
pointer in a local, and drops the read lock. It then passes sock->peer
(rather than the cached local) to ovpn_peer_del(), re-dereferencing the
ovpn_socket after the RCU read section has ended.
Unlike ovpn_tcp_sendmsg(), which uses the same "load under RCU, use
after unlock" pattern but is protected by lock_sock() held across the
function, ovpn_tcp_close() runs without the socket lock: inet_release()
invokes sk_prot->close() without taking lock_sock first.
ovpn_socket_release() can therefore complete its kref_put -> detach ->
synchronize_rcu -> kfree(sock) sequence concurrently, in the window
after ovpn_tcp_close() drops rcu_read_lock() but before it dereferences
sock->peer. The synchronize_rcu() in ovpn_socket_release() protects
readers that use the dereferenced pointer inside the RCU read section,
not those that escape the pointer to a local and use it afterwards.
A reproducer follows the pattern of commit 94560267d6c4 ("ovpn: tcp -
don't deref NULL sk_socket member after tcp_close()"): trigger a peer
removal (keepalive expiration or netlink OVPN_CMD_DEL_PEER) at the same
moment userspace closes the TCP fd. That commit fixed the detach-side
of the same race window; this one fixes the close-side at a different
victim.
Tighten the entry block to read sock->peer exactly once into the cached
peer local, and route all subsequent uses (the hold check, the
ovpn_peer_del() call, and the prot->close() invocation) through that
local. sock->peer is only ever written once in ovpn_socket_new() under
lock_sock(), before rcu_assign_sk_user_data() publishes the ovpn_socket,
and is never reassigned afterwards - but the previous multi-read pattern
made that invariant implicit rather than explicit. The same multi-read
shape exists in ovpn_tcp_recvmsg(), ovpn_tcp_sendmsg(),
ovpn_tcp_data_ready() and ovpn_tcp_write_space(); those will be cleaned
up via a dedicated helper in a follow-up net-next series.
Fixes: 11851cbd60ea ("ovpn: implement TCP transport")
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: David Carlier <devnexen@gmail.com>
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
|
|
At this time the driver is not listing any speeds
it supports. This should be ETHTOOL_LINK_MODE_100baseT1_Full_BIT
for DP83TC811. Add the missing call for phylib to read the abilities.
Fixes: b753a9faaf9a ("net: phy: DP83TC811: Introduce support for the DP83TC811 phy")
Suggested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Sven Schuchmann <schuchmann@schleissheimer.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20260512071949.6218-1-schuchmann@schleissheimer.de
[pabeni@redhat.com: dropped revision history]
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Increase the RQT uniform spread factor from 2 to 4 so that each channel
gets more indirection table entries and traffic is spread more evenly.
For num_channels > 64 imbalance drops from up to ~50% to up to ~25%.
For 64 or fewer channels the 256 entry minimum already provides at least
4x coverage and the table size is unchanged by this commit.
This satisfies the minimum 4x coverage requirement validated by the
generic RSS selftest commit 9e3d4dae9832 ("selftests: drv-net: rss:
validate min RSS table size").
The 4x spread factor is best-effort and the table size is always capped by
the device's log_max_rqt_size capability.
Signed-off-by: Yael Chemla <ychemla@nvidia.com>
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260511172719.330490-6-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
mlx5e_ethtool_set_channels() rejected channel count changes that
required a different RQT size when the default context indirection
table was user-configured. This restriction was introduced by
commit ee3572409f74 ("net/mlx5e: RSS, Block changing channels number
when RXFH is configured").
Lift the restriction. Validate the resize upfront with
ethtool_rxfh_indir_can_resize(), then fold or unfold the table
in-place via ethtool_rxfh_indir_resize() inside state_lock, before
mlx5e_safe_switch_params(), so the preactivate callback sees the
correct table content when it programs the HW.
Signed-off-by: Yael Chemla <ychemla@nvidia.com>
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260511172719.330490-5-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When the channel count changes and the RQT size changes with it, a
problem arise for non-default RSS contexts. The driver-side indirection
table grows actual_table_size without filling the new entries; stale
entries from a prior larger configuration may be re-exposed, causing
mlx5e_calc_indir_rqns() to WARN on an out-of-range index.
Replace mlx5e_rss_params_indir_modify_actual_size() with
mlx5e_rss_ctx_resize(), which fills new entries by replicating
the existing pattern, matching what ethtool_rxfh_ctxs_resize() does
for the same case. And restrict the loop to non-default contexts.
Call ethtool_rxfh_ctxs_can_resize() before acquiring state_lock to
validate that all non-default contexts can be resized, and
ethtool_rxfh_ctxs_resize() after releasing it to fold or unfold their
indirection tables. Both functions acquire rss_lock internally and
cannot be called under state_lock. RTNL, held by all set_channels
callers, serialises context creation and deletion making the pre-lock
check safe.
Guard both ethtool calls on mlx5e_rx_res_rss_cnt() > 1: skip the
validation and resize when no non-default contexts exist. This
naturally covers representors and IPoIB, which share
mlx5e_ethtool_set_channels() but cannot have non-default RSS contexts.
Signed-off-by: Yael Chemla <ychemla@nvidia.com>
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260511172719.330490-4-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Set rxfh_indir_space to the maximum indirection table size the driver
can support: the next power of two above MLX5E_MAX_NUM_CHANNELS times
MLX5E_UNIFORM_SPREAD_RQT_FACTOR.
Without this, ethtool_rxfh_ctxs_can_resize() returns -EINVAL, blocking
non-default RSS contexts from tracking indirection table size changes
when the channel count changes.
Signed-off-by: Yael Chemla <ychemla@nvidia.com>
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260511172719.330490-3-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
mlx5e_ethtool_set_channels() and mlx5e_rxfh_hfunc_check() rejected
channel counts that would produce an indirection table larger than 256
entries when the XOR8 hash function was active. This check was
introduced in commit 49e6c9387051 ("net/mlx5e: RSS, Block XOR hash
with over 128 channels").
XOR8 yields an 8-bit hash, so in practice only up to 256 entries in the
indirection table can be reached due to limited entropy. However, this
does not provide a strong justification for prohibiting larger
indirection tables. Remove the limitation.
Signed-off-by: Yael Chemla <ychemla@nvidia.com>
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260511172719.330490-2-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
free_txsa() is an RCU callback running in softirq context, but calls
crypto_free_aead() which can invoke vunmap() internally on hardware
crypto drivers (e.g. hisi_sec2), triggering a kernel crash.
Use rcu_work to defer the cleanup to a workqueue, for the same reasons
as the analogous fix to free_rxsa() in the previous patch.
Fixes: c09440f7dcb3 ("macsec: introduce IEEE 802.1AE driver")
Signed-off-by: Jinliang Zheng <alexjlzheng@tencent.com>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/20260511153102.2640368-4-alexjlzheng@tencent.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
crypto_free_aead() can internally invoke vunmap() (e.g. via
dma_free_attrs() in hardware crypto drivers such as hisi_sec2).
vunmap() must not be called from softirq context, but free_rxsa()
is an RCU callback that runs in softirq, leading to a kernel crash:
vunmap+0x4c/0x70
__iommu_dma_free+0xd0/0x138
dma_free_attrs+0xf4/0x100
sec_aead_exit+0x64/0xb8 [hisi_sec2]
crypto_destroy_tfm+0x98/0x110
free_rxsa+0x28/0x50 [macsec]
rcu_do_batch+0x184/0x460
rcu_core+0xf4/0x1f8
handle_softirqs+0x118/0x330
Use rcu_work to defer the cleanup to a workqueue. rcu_work dispatches
the worker asynchronously after the RCU grace period, so no thread
blocks waiting, and concurrent releases of multiple SAs naturally
share the same grace period.
Fixes: c09440f7dcb3 ("macsec: introduce IEEE 802.1AE driver")
Signed-off-by: Jinliang Zheng <alexjlzheng@tencent.com>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/20260511153102.2640368-3-alexjlzheng@tencent.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Introduce a dedicated ordered workqueue, macsec_wq, which will be used
by subsequent patches to defer SA crypto cleanup (crypto_free_aead and
related teardown) out of softirq context.
Using a dedicated workqueue instead of system_wq allows macsec_exit()
to drain exactly the work items belonging to this module via
destroy_workqueue(), without interfering with unrelated work items on
system_wq or causing unexpected delays elsewhere.
rcu_barrier() in macsec_exit() ensures all in-flight rcu_work callbacks
have enqueued their work items before destroy_workqueue() drains and
destroys the queue, making the two-step teardown correct and complete.
The same sequence is kept in the error path of macsec_init() as a
precaution, to mirror macsec_exit() and stay safe if work ever becomes
queueable before this point in the future.
While at it, rename the error labels in macsec_init() from the
resource-named style (rtnl:, notifier:, wq:) to the err_xxx: style
(err_rtnl:, err_notifier:, err_destroy_wq:) to align with the broader
kernel convention.
Signed-off-by: Jinliang Zheng <alexjlzheng@tencent.com>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/20260511153102.2640368-2-alexjlzheng@tencent.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
There is netdev_lock_ops() before the NETDEV_REGISTER notifier
in register_netdevice(), so use the non-locking functions
in net_failover_slave_register().
failover_slave_register() in failover_existing_slave_register() adds lock
and unlock ops too.
Call Trace:
<TASK>
__schedule+0x30d/0x7a0
schedule+0x27/0x90
schedule_preempt_disabled+0x15/0x30
__mutex_lock.constprop.0+0x538/0x9e0
__mutex_lock_slowpath+0x13/0x20
mutex_lock+0x3b/0x50
dev_set_mtu+0x40/0xe0
net_failover_slave_register+0x24/0x280
failover_slave_register+0x103/0x1b0
failover_event+0x15e/0x210
? dropmon_net_event+0xac/0xe0
notifier_call_chain+0x5e/0xe0
raw_notifier_call_chain+0x16/0x30
call_netdevice_notifiers_info+0x52/0xa0
register_netdevice+0x5f4/0x7c0
register_netdev+0x1e/0x40
_mlx5e_probe+0xe2/0x370 [mlx5_core]
mlx5e_probe+0x59/0x70 [mlx5_core]
? __pfx_mlx5e_probe+0x10/0x10 [mlx5_core]
Fixes: 4c975fd70002 ("net: hold instance lock during NETDEV_REGISTER/UP")
Signed-off-by: Faicker Mo <faicker.mo@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
find_skb() is the netconsole-specific entry into the netpoll skb
pool: every other netpoll consumer (bonding, team, vlan, bridge,
macvlan, dsa) builds its own sk_buff and never touches the pool.
With netpoll_send_udp() (its only caller) now living in netconsole,
find_skb() can join it.
Move find_skb() into drivers/net/netconsole.c as a file-static
helper, drop EXPORT_SYMBOL_GPL(find_skb) and remove its prototype
from include/linux/netpoll.h. find_skb() drains TX completions via
netpoll_zap_completion_queue(), which is already exported in the
NETDEV_INTERNAL namespace, so netconsole picks up
MODULE_IMPORT_NS("NETDEV_INTERNAL") to consume it.
The skb pool's lifecycle (np->skb_pool, np->refill_wq, refill_skbs(),
refill_skbs_work_handler(), skb_pool_flush()) stays in netpoll: it
is initialised in __netpoll_setup() and torn down in
__netpoll_cleanup(), both of which remain netpoll's responsibility.
The refill work queued via schedule_work(&np->refill_wq) from the
moved find_skb() runs refill_skbs_work_handler() in netpoll without
any further plumbing.
This is pure code motion: the function body is unchanged and its
sole caller (netpoll_send_udp(), already moved by an earlier patch)
keeps invoking it the same way. Pre-existing concerns about
find_skb() running from NMI/printk context (zap_completion_queue()
re-entry, skb_pool spinlocks, GFP_ATOMIC allocation, fallback skb
sizing vs. MAX_SKB_SIZE, PREEMPT_RT semantics of __kfree_skb()) are
inherited as-is and are not addressed here; they predate this
series and are out of scope. Fixing them is left for follow-up
work.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260512-netconsole_split-v2-9-1191d14ad66d@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
netpoll_udp_checksum() computes the UDP checksum for netconsole's
packets. Move it into drivers/net/netconsole.c as a file-static
helper; drop its EXPORT_SYMBOL_GPL and remove the prototype from
include/linux/netpoll.h.
This was the last csum_ipv6_magic() consumer in net/core/netpoll.c,
so drop the now-stale <net/ip6_checksum.h> include there. Pull it
into netconsole.c so the moved code keeps building.
It was also the last udp_hdr() consumer in net/core/netpoll.c. The
file no longer needs anything from <net/udp.h> (the UDP socket-layer
helpers); MAX_SKB_SIZE only needs struct udphdr, which is provided
by the lighter <linux/udp.h>. Swap the include accordingly.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260512-netconsole_split-v2-7-1191d14ad66d@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
push_udp() builds the UDP header (and triggers the checksum) for
netconsole's UDP packets. Move it into drivers/net/netconsole.c as
a file-static helper; drop its EXPORT_SYMBOL_GPL and remove the
prototype from include/linux/netpoll.h.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260512-netconsole_split-v2-6-1191d14ad66d@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
push_eth() builds the Ethernet header for netconsole's UDP packets.
Move it into drivers/net/netconsole.c as a file-static helper; drop
its EXPORT_SYMBOL_GPL and remove the prototype from
include/linux/netpoll.h.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260512-netconsole_split-v2-5-1191d14ad66d@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
push_ipv4() builds the IPv4 header for netconsole's UDP packets.
Move it into drivers/net/netconsole.c as a file-static helper; drop
its EXPORT_SYMBOL_GPL and remove the prototype from
include/linux/netpoll.h.
put_unaligned() is no longer used in net/core/netpoll.c, so drop
the now-stale <linux/unaligned.h> include from there. Pull it into
netconsole.c so the moved code keeps building.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260512-netconsole_split-v2-4-1191d14ad66d@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
push_ipv6() builds the IPv6 header for netconsole's UDP packets.
Its only caller, netpoll_send_udp(), now lives in netconsole, so
the helper can move there as a file-static function. Drop its
EXPORT_SYMBOL_GPL and remove the prototype from
include/linux/netpoll.h.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260512-netconsole_split-v2-3-1191d14ad66d@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Move netpoll_send_udp() from net/core/netpoll.c into
drivers/net/netconsole.c as a static helper, drop EXPORT_SYMBOL(),
and remove the prototype from include/linux/netpoll.h.
netconsole was the only in-tree caller of this entry point. Every
other netpoll consumer (bonding, team, vlan, bridge, macvlan, dsa)
already builds its own sk_buff and hands it to netpoll_send_skb(),
so the netpoll send-side interface is now skb-only.
The helpers it depends on (find_skb(), push_ipv6(), push_ipv4(),
push_udp(), push_eth(), netpoll_udp_checksum()) were exposed in
the previous patches and stay in net/core/netpoll.c for now.
Subsequent patches move each of them into netconsole one at a time
and drop the corresponding EXPORT_SYMBOL_GPL.
Pull <linux/ip.h>, <linux/ipv6.h> and <linux/udp.h> into netconsole.c
so the moved code can name the header structures.
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260512-netconsole_split-v2-2-1191d14ad66d@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
simple_strtoul() is deprecated as it has no error checking. Replace it
with kstrtouint() which returns an error code on invalid input, and add
appropriate error handling.
Also add a NULL check before parsing flags, since strsep() can set id
to NULL if the input has fewer tokens than expected.
Preserve the original behavior for a trailing colon by checking *id
before parsing flags, so an empty string results in flags = 0 rather
than an error.
Signed-off-by: Sajal Gupta <sajal2005gupta@gmail.com>
Link: https://patch.msgid.link/20260509095518.2640-1-sajal2005gupta@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
ipvlan_process_multicast() runs from process context, there is no
risk of stack overflow if we call netif_receive_skb() instead
of netif_rx().
This avoids some overhead adding/removing skbs to/from a per-cpu
backlog and raising/processing NET_RX softirqs.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260512042019.3300975-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This commit prevents tail-drop when a qdisc is present and the ptr_ring
becomes full. Once the ring reaches capacity after a produce attempt,
the netdev queue is stopped instead of dropping subsequent packets.
If no qdisc is present, the previous tail-drop behavior is preserved.
If producing an entry fails anyway due to a race, tun_net_xmit() drops
the packet. Such races are expected because LLTX is enabled and the
transmit path operates without the usual locking.
The __tun_wake_queue() function of the consumer races with the producer
for waking/stopping the netdev queue, which could result in a stalled
queue. Therefore, an smp_mb__after_atomic() is introduced that pairs
with the smp_mb() of the consumer. It follows the principle of store
buffering described in tools/memory-model/Documentation/recipes.txt:
- The producer in tun_net_xmit() first sets __QUEUE_STATE_DRV_XOFF,
followed by an smp_mb__after_atomic() (= smp_mb()), and then reads the
ring with __ptr_ring_check_produce().
- The consumer in __tun_wake_queue() first writes zero to the ring in
__ptr_ring_consume(), followed by an smp_mb(), and then reads the queue
status with netif_tx_queue_stopped().
=> Following the aforementioned principle, it is impossible for the
producer to see a full ring (and therefore not wake the queue on the
re-check) while the consumer simultaneously fails to see a stopped
queue (and therefore also does not wake it).
Benchmarks:
The benchmarks show a slight regression in raw transmission performance
when using two sending threads. Packet loss also occurs only in the
two-thread sending case; no packet loss was observed with a single
sending thread.
Test setup:
AMD Ryzen 5 5600X at 4.3 GHz, 3200 MHz RAM, isolated QEMU threads;
Average over 50 runs @ 100,000,000 packets. SRSO and spectre v2
mitigations disabled.
Note for tap+vhost-net:
XDP drop program active in VM -> ~2.5x faster; slower for tap due to
more syscalls (high utilization of entry_SYSRETQ_unsafe_stack in perf)
+--------------------------+--------------+----------------+----------+
| 1 thread | Stock | Patched with | diff |
| sending | | fq_codel qdisc | |
+------------+-------------+--------------+----------------+----------+
| TAP | Received | 1.132 Mpps | 1.123 Mpps | -0.8% |
| +-------------+--------------+----------------+----------+
| | Lost/s | 3.765 Mpps | 0 pps | |
+------------+-------------+--------------+----------------+----------+
| TAP | Received | 3.857 Mpps | 3.901 Mpps | +1.1% |
| +-------------+--------------+----------------+----------+
| +vhost-net | Lost/s | 0.802 Mpps | 0 pps | |
+------------+-------------+--------------+----------------+----------+
+--------------------------+--------------+----------------+----------+
| 2 threads | Stock | Patched with | diff |
| sending | | fq_codel qdisc | |
+------------+-------------+--------------+----------------+----------+
| TAP | Received | 1.115 Mpps | 1.081 Mpps | -3.0% |
| +-------------+--------------+----------------+----------+
| | Lost/s | 8.490 Mpps | 391 pps | |
+------------+-------------+--------------+----------------+----------+
| TAP | Received | 3.664 Mpps | 3.555 Mpps | -3.0% |
| +-------------+--------------+----------------+----------+
| +vhost-net | Lost/s | 5.330 Mpps | 938 pps | |
+------------+-------------+--------------+----------------+----------+
Co-developed-by: Tim Gebauer <tim.gebauer@tu-dortmund.de>
Signed-off-by: Tim Gebauer <tim.gebauer@tu-dortmund.de>
Signed-off-by: Simon Schippers <simon.schippers@tu-dortmund.de>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://patch.msgid.link/20260510151529.43895-5-simon.schippers@tu-dortmund.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add tun_wake_queue() to tun.c and export it for use by vhost-net. The
function validates that the file belongs to a tun/tap device and that
the tfile exists, dereferences the tun_struct under RCU, and delegates
to __tun_wake_queue().
vhost_net_buf_produce() now calls tun_wake_queue() after a successful
batched consume of the ring to allow the netdev subqueue to be woken up.
The point is to allow the queue to be stopped when it gets full, which
is required for traffic shaping - implemented by the following
"avoid ptr_ring tail-drop when a qdisc is present".
Without the corresponding queue stopping, this patch alone causes no
throughput regression for a tap+vhost-net setup sending to a qemu VM:
3.857 Mpps to 3.891 Mpps.
Details: AMD Ryzen 5 5600X at 4.3 GHz, 3200 MHz RAM, isolated QEMU
threads, XDP drop program active in VM, pktgen sender; Avg over
50 runs @ 100,000,000 packets. SRSO and spectre v2 mitigations disabled.
Co-developed-by: Tim Gebauer <tim.gebauer@tu-dortmund.de>
Signed-off-by: Tim Gebauer <tim.gebauer@tu-dortmund.de>
Signed-off-by: Simon Schippers <simon.schippers@tu-dortmund.de>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://patch.msgid.link/20260510151529.43895-3-simon.schippers@tu-dortmund.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
