This patch addresses a macvlan leak issue in the i40e driver caused by
concurrent access to vsi->mac_filter_hash. The leak occurs when multiple
threads attempt to modify the mac_filter_hash simultaneously, leading to
inconsistent state and potential memory leaks.
To fix this, we now wrap the calls to i40e_del_mac_filter() and zeroing
vf->default_lan_addr.addr with spin_lock/unlock_bh(&vsi->mac_filter_hash_lock),
ensuring atomic operations and preventing concurrent access.
Additionally, we add lockdep_assert_held(&vsi->mac_filter_hash_lock) in
i40e_add_mac_filter() to help catch similar issues in the future.
Reproduction steps:
1. Spawn VFs and configure port vlan on them.
2. Trigger concurrent macvlan operations (e.g., adding and deleting
portvlan and/or mac filters).
3. Observe the potential memory leak and inconsistent state in the
mac_filter_hash.
This synchronization ensures the integrity of the mac_filter_hash and prevents
the described leak.
Fixes: fed0d9f132 ("i40e: Fix VF's MAC Address change on VM")
Reviewed-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Increasing MSI-X value on a VF leads to invalid memory operations. This
is caused by not reallocating some arrays.
Reproducer:
modprobe ice
echo 0 > /sys/bus/pci/devices/$PF_PCI/sriov_drivers_autoprobe
echo 1 > /sys/bus/pci/devices/$PF_PCI/sriov_numvfs
echo 17 > /sys/bus/pci/devices/$VF0_PCI/sriov_vf_msix_count
Default MSI-X is 16, so 17 and above triggers this issue.
KASAN reports:
BUG: KASAN: slab-out-of-bounds in ice_vsi_alloc_ring_stats+0x38d/0x4b0 [ice]
Read of size 8 at addr ffff8888b937d180 by task bash/28433
(...)
Call Trace:
(...)
? ice_vsi_alloc_ring_stats+0x38d/0x4b0 [ice]
kasan_report+0xed/0x120
? ice_vsi_alloc_ring_stats+0x38d/0x4b0 [ice]
ice_vsi_alloc_ring_stats+0x38d/0x4b0 [ice]
ice_vsi_cfg_def+0x3360/0x4770 [ice]
? mutex_unlock+0x83/0xd0
? __pfx_ice_vsi_cfg_def+0x10/0x10 [ice]
? __pfx_ice_remove_vsi_lkup_fltr+0x10/0x10 [ice]
ice_vsi_cfg+0x7f/0x3b0 [ice]
ice_vf_reconfig_vsi+0x114/0x210 [ice]
ice_sriov_set_msix_vec_count+0x3d0/0x960 [ice]
sriov_vf_msix_count_store+0x21c/0x300
(...)
Allocated by task 28201:
(...)
ice_vsi_cfg_def+0x1c8e/0x4770 [ice]
ice_vsi_cfg+0x7f/0x3b0 [ice]
ice_vsi_setup+0x179/0xa30 [ice]
ice_sriov_configure+0xcaa/0x1520 [ice]
sriov_numvfs_store+0x212/0x390
(...)
To fix it, use ice_vsi_rebuild() instead of ice_vf_reconfig_vsi(). This
causes the required arrays to be reallocated taking the new queue count
into account (ice_vsi_realloc_stat_arrays()). Set req_txq and req_rxq
before ice_vsi_rebuild(), so that realloc uses the newly set queue
count.
Additionally, ice_vsi_rebuild() does not remove VSI filters
(ice_fltr_remove_all()), so ice_vf_init_host_cfg() is no longer
necessary.
Reported-by: Jacob Keller <jacob.e.keller@intel.com>
Fixes: 2a2cb4c6c1 ("ice: replace ice_vf_recreate_vsi() with ice_vf_reconfig_vsi()")
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Triggering the reset while in switchdev mode causes
errors[1]. Rules are already removed by this time
because switch content is flushed in case of the reset.
This means that rules were deleted from HW but SW
still thinks they exist so when we get
SWITCHDEV_FDB_DEL_TO_DEVICE notification we try to
delete not existing rule.
We can avoid these errors by clearing the rules
early in the reset flow before they are removed from HW.
Switchdev API will get notified that the rule was removed
so we won't get SWITCHDEV_FDB_DEL_TO_DEVICE notification.
Remove unnecessary ice_clear_sw_switch_recipes.
[1]
ice 0000:01:00.0: Failed to delete FDB forward rule, err: -2
ice 0000:01:00.0: Failed to delete FDB guard rule, err: -2
Fixes: 7c945a1a8e ("ice: Switchdev FDB events support")
Reviewed-by: Mateusz Polchlopek <mateusz.polchlopek@intel.com>
Signed-off-by: Wojciech Drewek <wojciech.drewek@intel.com>
Tested-by: Sujai Buvaneswaran <sujai.buvaneswaran@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
netif_is_ice() works by checking the pointer to netdev ops. However, it
only checks for the default ice_netdev_ops, not ice_netdev_safe_mode_ops,
so in Safe Mode it always returns false, which is unintuitive. While it
doesn't look like netif_is_ice() is currently being called anywhere in Safe
Mode, this could change and potentially lead to unexpected behaviour.
Fixes: df006dd4b1 ("ice: Add initial support framework for LAG")
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Reviewed-by: Brett Creeley <brett.creeley@amd.com>
Tested-by: Sujai Buvaneswaran <sujai.buvaneswaran@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
If DDP package is missing or corrupted, the driver should enter Safe Mode.
Instead, an error is returned and probe fails.
To fix this, don't exit init if ice_init_ddp_config() returns an error.
Repro:
* Remove or rename DDP package (/lib/firmware/intel/ice/ddp/ice.pkg)
* Load ice
Fixes: cc5776fe18 ("ice: Enable switching default Tx scheduler topology")
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Reviewed-by: Brett Creeley <brett.creeley@amd.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Starting with 6.12 commit 85585b4bc8 ("selftests: add ncdevmem, netcat
for devmem TCP") kselftest-all creates additional outputs that
kselftest-clean does not cleanup:
$ make defconfig
$ make kselftest-all
$ make kselftest-clean
$ git clean -ndxf | grep tools/net
Would remove tools/net/ynl/lib/__pycache__/
Would remove tools/net/ynl/lib/ynl.a
Would remove tools/net/ynl/lib/ynl.d
Would remove tools/net/ynl/lib/ynl.o
Make kselftest-clean remove the newly added net/ynl outputs.
Fixes: 85585b4bc8 ("selftests: add ncdevmem, netcat for devmem TCP")
Signed-off-by: Greg Thelen <gthelen@google.com>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Link: https://patch.msgid.link/20241005215600.852260-1-gthelen@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jonas Gorski says:
====================
net: dsa: b53: assorted jumbo frame fixes
While investigating the capabilities of BCM63XX's integrated switch and
its DMA engine, I noticed a few issues in b53's jumbo frame code.
Mostly a confusion of MTU vs frame length, but also a few missing cases
for 100M switches.
Tested on BCM63XX and BCM53115 with intel 1G and realtek 1G NICs,
which support MTUs of 9000 or slightly above, but significantly less
than the 9716/9720 supported by BCM53115, so I couldn't verify the
actual maximum frame length.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
---
====================
Link: https://patch.msgid.link/20241004-b53_jumbo_fixes-v1-0-ce1e54aa7b3c@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
All modern chips support and need the 10_100 bit set for supporting jumbo
frames on 10/100 ports, so instead of enabling it only for 583XX enable
it for everything except bcm63xx, where the bit is writeable, but does
nothing.
Tested on BCM53115, where jumbo frames were dropped at 10/100 speeds
without the bit set.
Fixes: 6ae5834b98 ("net: dsa: b53: add MTU configuration support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
While BCM5325/5365 do not support jumbo frames, they do support slightly
oversized frames, so do not error out if requesting a supported MTU for
them.
Fixes: 6ae5834b98 ("net: dsa: b53: add MTU configuration support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
BCM5325/BCM5365 do not support jumbo frames, so we should not report a
jumbo frame mtu for them. But they do support so called "oversized"
frames up to 1536 bytes long by default, so report an appropriate MTU.
Fixes: 6ae5834b98 ("net: dsa: b53: add MTU configuration support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
JMS_MAX_SIZE is the ethernet frame length, not the MTU, which is payload
without ethernet headers.
According to the datasheets maximum supported frame length for most
gigabyte swithes is 9720 bytes, so convert that to the expected MTU when
using VLAN tagged frames.
Fixes: 6ae5834b98 ("net: dsa: b53: add MTU configuration support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
JMS_MIN_SIZE is the full ethernet frame length, while mtu is just the
data payload size. Comparing these two meant that mtus between 1500 and
1518 did not trigger enabling jumbo frames.
So instead compare the set mtu ETH_DATA_LEN, which is equal to
JMS_MIN_SIZE - ETH_HLEN - ETH_FCS_LEN;
Also do a check that the requested mtu is actually greater than the
minimum length, else we do not need to enable jumbo frames.
In practice this only introduced a very small range of mtus that did not
work properly. Newer chips allow 2000 byte large frames by default, and
older chips allow 1536 bytes long, which is equivalent to an mtu of
1514. So effectivly only mtus of 1515~1517 were broken.
Fixes: 6ae5834b98 ("net: dsa: b53: add MTU configuration support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Usage of devm_alloc_etherdev_mqs() conflicts with
am65_cpsw_nuss_cleanup_ndev() as the same struct net_device instances
get unregistered twice. Switch to alloc_etherdev_mqs() and make sure
am65_cpsw_nuss_cleanup_ndev() unregisters and frees those net_device
instances properly.
With this, it is finally possible to rmmod the driver without oopsing
the kernel.
Fixes: 93a7653031 ("net: ethernet: ti: introduce am65x/j721e gigabit eth subsystem driver")
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Reviewed-by: Roger Quadros <roger@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In am65_cpsw_nuss_remove(), move the call to am65_cpsw_unregister_devlink()
after am65_cpsw_nuss_cleanup_ndev() to avoid triggering the
WARN_ON(devlink_port->type != DEVLINK_PORT_TYPE_NOTSET) in
devl_port_unregister(). Makes it coherent with usage in
m65_cpsw_nuss_register_ndevs()'s cleanup path.
Fixes: 58356eb31d ("net: ti: am65-cpsw-nuss: Add devlink support")
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Reviewed-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Commit c938ab4da0 ("net: phy: Manual remove LEDs to ensure correct
ordering") correctly fixed a problem with using devm_ but missed
removing the LED entry from the LEDs list.
This cause kernel panic on specific scenario where the port for the PHY
is torn down and up and the kmod for the PHY is removed.
On setting the port down the first time, the assosiacted LEDs are
correctly unregistered. The associated kmod for the PHY is now removed.
The kmod is now added again and the port is now put up, the associated LED
are registered again.
On putting the port down again for the second time after these step, the
LED list now have 4 elements. With the first 2 already unregistered
previously and the 2 new one registered again.
This cause a kernel panic as the first 2 element should have been
removed.
Fix this by correctly removing the element when LED is unregistered.
Reported-by: Daniel Golle <daniel@makrotopia.org>
Tested-by: Daniel Golle <daniel@makrotopia.org>
Cc: stable@vger.kernel.org
Fixes: c938ab4da0 ("net: phy: Manual remove LEDs to ensure correct ordering")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20241004182759.14032-1-ansuelsmth@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Luiz Augusto von Dentz says:
====================
bluetooth pull request for net:
- RFCOMM: FIX possible deadlock in rfcomm_sk_state_change
- hci_conn: Fix UAF in hci_enhanced_setup_sync
- btusb: Don't fail external suspend requests
* tag 'for-net-2024-10-04' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth:
Bluetooth: btusb: Don't fail external suspend requests
Bluetooth: hci_conn: Fix UAF in hci_enhanced_setup_sync
Bluetooth: RFCOMM: FIX possible deadlock in rfcomm_sk_state_change
====================
Link: https://patch.msgid.link/20241004210124.4010321-1-luiz.dentz@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We have recently noticed the exact same KASAN splat as in commit
6cd4a78d96 ("net: do not leave a dangling sk pointer, when socket
creation fails"). The problem is that commit did not fully address the
problem, as some pf->create implementations do not use sk_common_release
in their error paths.
For example, we can use the same reproducer as in the above commit, but
changing ping to arping. arping uses AF_PACKET socket and if packet_create
fails, it will just sk_free the allocated sk object.
While we could chase all the pf->create implementations and make sure they
NULL the freed sk object on error from the socket, we can't guarantee
future protocols will not make the same mistake.
So it is easier to just explicitly NULL the sk pointer upon return from
pf->create in __sock_create. We do know that pf->create always releases the
allocated sk object on error, so if the pointer is not NULL, it is
definitely dangling.
Fixes: 6cd4a78d96 ("net: do not leave a dangling sk pointer, when socket creation fails")
Signed-off-by: Ignat Korchagin <ignat@cloudflare.com>
Cc: stable@vger.kernel.org
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241003170151.69445-1-ignat@cloudflare.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The kernel may crash when deleting a genetlink family if there are still
listeners for that family:
Oops: Kernel access of bad area, sig: 11 [#1]
...
NIP [c000000000c080bc] netlink_update_socket_mc+0x3c/0xc0
LR [c000000000c0f764] __netlink_clear_multicast_users+0x74/0xc0
Call Trace:
__netlink_clear_multicast_users+0x74/0xc0
genl_unregister_family+0xd4/0x2d0
Change the unsafe loop on the list to a safe one, because inside the
loop there is an element removal from this list.
Fixes: b8273570f8 ("genetlink: fix netns vs. netlink table locking (2)")
Cc: stable@vger.kernel.org
Signed-off-by: Anastasia Kovaleva <a.kovaleva@yadro.com>
Reviewed-by: Dmitry Bogdanov <d.bogdanov@yadro.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20241003104431.12391-1-a.kovaleva@yadro.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 4e0a1d8b06
("Bluetooth: btusb: Don't suspend when there are connections")
introduces a check for connections to prevent auto-suspend but that
actually ignored the fact the .suspend callback can be called for
external suspend requests which
Documentation/driver-api/usb/power-management.rst states the following:
'External suspend calls should never be allowed to fail in this way,
only autosuspend calls. The driver can tell them apart by applying
the :c:func:`PMSG_IS_AUTO` macro to the message argument to the
``suspend`` method; it will return True for internal PM events
(autosuspend) and False for external PM events.'
In addition to that align system suspend with USB suspend by using
hci_suspend_dev since otherwise the stack would be expecting events
such as advertising reports which may not be delivered while the
transport is suspended.
Fixes: 4e0a1d8b06 ("Bluetooth: btusb: Don't suspend when there are connections")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Tested-by: Kiran K <kiran.k@intel.com>
rfcomm_sk_state_change attempts to use sock_lock so it must never be
called with it locked but rfcomm_sock_ioctl always attempt to lock it
causing the following trace:
======================================================
WARNING: possible circular locking dependency detected
6.8.0-syzkaller-08951-gfe46a7dd189e #0 Not tainted
------------------------------------------------------
syz-executor386/5093 is trying to acquire lock:
ffff88807c396258 (sk_lock-AF_BLUETOOTH-BTPROTO_RFCOMM){+.+.}-{0:0}, at: lock_sock include/net/sock.h:1671 [inline]
ffff88807c396258 (sk_lock-AF_BLUETOOTH-BTPROTO_RFCOMM){+.+.}-{0:0}, at: rfcomm_sk_state_change+0x5b/0x310 net/bluetooth/rfcomm/sock.c:73
but task is already holding lock:
ffff88807badfd28 (&d->lock){+.+.}-{3:3}, at: __rfcomm_dlc_close+0x226/0x6a0 net/bluetooth/rfcomm/core.c:491
Reported-by: syzbot+d7ce59b06b3eb14fd218@syzkaller.appspotmail.com
Tested-by: syzbot+d7ce59b06b3eb14fd218@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=d7ce59b06b3eb14fd218
Fixes: 3241ad820d ("[Bluetooth] Add timestamp support to L2CAP, RFCOMM and SCO")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
PSE controllers like the TPS23881 can forcefully turn off their
configuration state. In such cases, the is_enabled() and get_status()
callbacks will report the PSE as disabled, while admin_state_enabled
will show it as enabled. This mismatch can lead the user to attempt
to enable it, but no action is taken as admin_state_enabled remains set.
The solution is to disable the PSE before enabling it, ensuring the
actual status matches admin_state_enabled.
Fixes: d83e13761d ("net: pse-pd: Use regulator framework within PSE framework")
Signed-off-by: Kory Maincent <kory.maincent@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20241002121706.246143-1-kory.maincent@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, the second bridge command overwrites the first one.
Fix this by adding this VID to the interface behind $swp2.
The one_bridge_two_pvids() test intends to check that there is no
leakage of traffic between bridge ports which have a single VLAN - the
PVID VLAN.
Because of a typo, port $swp1 is configured with a PVID twice (second
command overwrites first), and $swp2 isn't configured at all (and since
the bridge vlan_default_pvid property is set to 0, this port will not
have a PVID at all, so it will drop all untagged and priority-tagged
traffic).
So, instead of testing the configuration that was intended, we are
testing a different one, where one port has PVID 2 and the other has
no PVID. This incorrect version of the test should also pass, but is
ineffective for its purpose, so fix the typo.
This typo has an impact on results of the test,
potentially leading to wrong conclusions regarding
the functionality of a network device.
The tests results:
TEST: Switch ports in VLAN-aware bridge with different PVIDs:
Unicast non-IP untagged [ OK ]
Multicast non-IP untagged [ OK ]
Broadcast non-IP untagged [ OK ]
Unicast IPv4 untagged [ OK ]
Multicast IPv4 untagged [ OK ]
Unicast IPv6 untagged [ OK ]
Multicast IPv6 untagged [ OK ]
Unicast non-IP VID 1 [ OK ]
Multicast non-IP VID 1 [ OK ]
Broadcast non-IP VID 1 [ OK ]
Unicast IPv4 VID 1 [ OK ]
Multicast IPv4 VID 1 [ OK ]
Unicast IPv6 VID 1 [ OK ]
Multicast IPv6 VID 1 [ OK ]
Unicast non-IP VID 4094 [ OK ]
Multicast non-IP VID 4094 [ OK ]
Broadcast non-IP VID 4094 [ OK ]
Unicast IPv4 VID 4094 [ OK ]
Multicast IPv4 VID 4094 [ OK ]
Unicast IPv6 VID 4094 [ OK ]
Multicast IPv6 VID 4094 [ OK ]
Fixes: 476a4f05d9 ("selftests: forwarding: add a no_forwarding.sh test")
Reviewed-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Kacper Ludwinski <kac.ludwinski@icloud.com>
Link: https://patch.msgid.link/20241002051016.849-1-kac.ludwinski@icloud.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Nick Child says:
====================
ibmvnic: Fix for send scrq direct
This is a v2 of a patchset (now just patch) which addresses a
bug in a new feature which is causing major link UP issues with
certain physical cards.
For a full summary of the issue:
1. During vnic initialization we get the following values from vnic
server regarding "Transmit / Receive Descriptor Requirement" (see
PAPR Table 584. CAPABILITIES Commands):
- LSO Tx frame = 0x0F , header offsets + L2, L3, L4 headers required
- CSO Tx frame = 0x0C , header offsets + L2 header required
- standard frame = 0x0C , header offsets + L2 header required
2. Assume we are dealing with only "standard frames" from now on (no
CSO, no LSO)
3. When using 100G backing device, we don't hand vnic server any header
information and TX is successful
4. When using 25G backing device, we don't hand vnic server any header
information and TX fails and we get "Adapter Error" transport events.
The obvious issue here is that vnic client should be respecting the 0X0C
header requirement for standard frames. But 100G cards will also give
0x0C despite the fact that we know TX works if we ignore it. That being
said, we still must respect values given from the managing server. Will
need to work with them going forward to hopefully get 100G cards to
return 0x00 for this bitstring so the performance gains of using
send_subcrq_direct can be continued.
====================
Link: https://patch.msgid.link/20241001163200.1802522-1-nnac123@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Previously, the TX header requirement for standard frames was ignored.
This requirement is a bitstring sent from the VIOS which maps to the
type of header information needed during TX. If no header information,
is needed then send subcrq direct can be used (which can be more
performant).
This bitstring was previously ignored for standard packets (AKA non LSO,
non CSO) due to the belief that the bitstring was over-cautionary. It
turns out that there are some configurations where the backing device
does need header information for transmission of standard packets. If
the information is not supplied then this causes continuous "Adapter
error" transport events. Therefore, this bitstring should be respected
and observed before considering the use of send subcrq direct.
Fixes: 74839f7a82 ("ibmvnic: Introduce send sub-crq direct")
Signed-off-by: Nick Child <nnac123@linux.ibm.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20241001163200.1802522-2-nnac123@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The blamed commit introduced an unexpected regression in the sja1105
driver. Packets from VLAN-unaware bridge ports get received correctly,
but the protocol stack can't seem to decode them properly.
For ds->untag_bridge_pvid users (thus also sja1105), the blamed commit
did introduce a functional change: dsa_switch_rcv() used to call
dsa_untag_bridge_pvid(), which looked like this:
err = br_vlan_get_proto(br, &proto);
if (err)
return skb;
/* Move VLAN tag from data to hwaccel */
if (!skb_vlan_tag_present(skb) && skb->protocol == htons(proto)) {
skb = skb_vlan_untag(skb);
if (!skb)
return NULL;
}
and now it calls dsa_software_vlan_untag() which has just this:
/* Move VLAN tag from data to hwaccel */
if (!skb_vlan_tag_present(skb)) {
skb = skb_vlan_untag(skb);
if (!skb)
return NULL;
}
thus lacks any skb->protocol == bridge VLAN protocol check. That check
is deferred until a later check for skb->vlan_proto (in the hwaccel area).
The new code is problematic because, for VLAN-untagged packets,
skb_vlan_untag() blindly takes the 4 bytes starting with the EtherType
and turns them into a hwaccel VLAN tag. This is what breaks the protocol
stack.
It would be tempting to "make it work as before" and only call
skb_vlan_untag() for those packets with the skb->protocol actually
representing a VLAN.
But the premise of the newly introduced dsa_software_vlan_untag() core
function is not wrong. Drivers set ds->untag_bridge_pvid or
ds->untag_vlan_aware_bridge_pvid presumably because they send all
traffic to the CPU reception path as VLAN-tagged. So why should we spend
any additional CPU cycles assuming that the packet may be VLAN-untagged?
And why does the sja1105 driver opt into ds->untag_bridge_pvid if it
doesn't always deliver packets to the CPU as VLAN-tagged?
The answer to the latter question is indeed more interesting: it doesn't
need to. This got done in commit 884be12f85 ("net: dsa: sja1105: add
support for imprecise RX"), because I thought it would be needed, but I
didn't realize that it doesn't actually make a difference.
As explained in the commit message of the blamed patch, ds->untag_bridge_pvid
only makes a difference in the VLAN-untagged receive path of a bridge port.
However, in that operating mode, tag_sja1105.c makes use of VLAN tags
with the ETH_P_SJA1105 TPID, and it decodes and consumes these VLAN tags
as if they were DSA tags (aka tag_8021q operation). Even if commit
884be12f85 ("net: dsa: sja1105: add support for imprecise RX") added
this logic in sja1105_bridge_vlan_add():
/* Always install bridge VLANs as egress-tagged on the CPU port. */
if (dsa_is_cpu_port(ds, port))
flags = 0;
that was for _bridge_ VLANs, which are _not_ committed to hardware
in VLAN-unaware mode (aka the mode where ds->untag_bridge_pvid does
anything at all). Even prior to that change, the tag_8021q VLANs
were always installed as egress-tagged on the CPU port, see
dsa_switch_tag_8021q_vlan_add():
u16 flags = 0; // egress-tagged, non-PVID
if (dsa_port_is_user(dp))
flags |= BRIDGE_VLAN_INFO_UNTAGGED |
BRIDGE_VLAN_INFO_PVID;
err = dsa_port_do_tag_8021q_vlan_add(dp, info->vid,
flags);
if (err)
return err;
Whether the sja1105 driver needs the new flag, ds->untag_vlan_aware_bridge_pvid,
rather than ds->untag_bridge_pvid, is a separate discussion. To fix the
current bug in VLAN-unaware bridge mode, I would argue that the sja1105
driver should not request something it doesn't need, rather than
complicating the core DSA helper. Whereas before the blamed commit, this
setting was harmless, now it has caused breakage.
Fixes: 93e4649efa ("net: dsa: provide a software untagging function on RX for VLAN-aware bridges")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20241001140206.50933-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2024-09-30 (ice, idpf)
This series contains updates to ice and idpf drivers:
For ice:
Michal corrects setting of dst VSI on LAN filters and adds clearing of
port VLAN configuration during reset.
Gui-Dong Han corrects failures to decrement refcount in some error
paths.
Przemek resolves a memory leak in ice_init_tx_topology().
Arkadiusz prevents setting of DPLL_PIN_STATE_SELECTABLE to an improper
value.
Dave stops clearing of VLAN tracking bit to allow for VLANs to be properly
restored after reset.
For idpf:
Ahmed sets uninitialized dyn_ctl_intrvl_s value.
Josh corrects use and reporting of mailbox size.
Larysa corrects order of function calls during de-initialization.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
idpf: deinit virtchnl transaction manager after vport and vectors
idpf: use actual mbx receive payload length
idpf: fix VF dynamic interrupt ctl register initialization
ice: fix VLAN replay after reset
ice: disallow DPLL_PIN_STATE_SELECTABLE for dpll output pins
ice: fix memleak in ice_init_tx_topology()
ice: clear port vlan config during reset
ice: Fix improper handling of refcount in ice_sriov_set_msix_vec_count()
ice: Fix improper handling of refcount in ice_dpll_init_rclk_pins()
ice: set correct dst VSI in only LAN filters
====================
Link: https://patch.msgid.link/20240930223601.3137464-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
David Howells says:
====================
rxrpc: Miscellaneous fixes
Here some miscellaneous fixes for AF_RXRPC:
(1) Fix a race in the I/O thread vs UDP socket setup.
(2) Fix an uninitialised variable.
====================
Link: https://patch.msgid.link/20241001132702.3122709-1-dhowells@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In rxrpc_open_socket(), it sets up the socket and then sets up the I/O
thread that will handle it. This is a problem, however, as there's a gap
between the two phases in which a packet may come into rxrpc_encap_rcv()
from the UDP packet but we oops when trying to wake the not-yet created I/O
thread.
As a quick fix, just make rxrpc_encap_rcv() discard the packet if there's
no I/O thread yet.
A better, but more intrusive fix would perhaps be to rearrange things such
that the socket creation is done by the I/O thread.
Fixes: a275da62e8 ("rxrpc: Create a per-local endpoint receive queue and I/O thread")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: yuxuanzhe@outlook.com
cc: Marc Dionne <marc.dionne@auristor.com>
cc: Simon Horman <horms@kernel.org>
cc: linux-afs@lists.infradead.org
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241001132702.3122709-2-dhowells@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Neal Cardwell says:
====================
tcp: 3 fixes for retrans_stamp and undo logic
Geumhwan Yu <geumhwan.yu@samsung.com> recently reported and diagnosed
a regression in TCP loss recovery undo logic in the case where a TCP
connection enters fast recovery, is unable to retransmit anything due to
TSQ, and then receives an ACK allowing forward progress. The sender should
be able to undo the spurious loss recovery in this case, but was not doing
so. The first patch fixes this regression.
Running our suite of packetdrill tests with the first fix, the tests
highlighted two other small bugs in the way retrans_stamp is updated in
some rare corner cases. The second two patches fix those other two small
bugs.
Thanks to Geumhwan Yu for the bug report!
====================
Link: https://patch.msgid.link/20241001200517.2756803-1-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix tcp_rcv_synrecv_state_fastopen() to not zero retrans_stamp
if retransmits are outstanding.
tcp_fastopen_synack_timer() sets retrans_stamp, so typically we'll
need to zero retrans_stamp here to prevent spurious
retransmits_timed_out(). The logic to zero retrans_stamp is from this
2019 commit:
commit cd736d8b67 ("tcp: fix retrans timestamp on passive Fast Open")
However, in the corner case where the ACK of our TFO SYNACK carried
some SACK blocks that caused us to enter TCP_CA_Recovery then that
non-zero retrans_stamp corresponds to the active fast recovery, and we
need to leave retrans_stamp with its current non-zero value, for
correct ETIMEDOUT and undo behavior.
Fixes: cd736d8b67 ("tcp: fix retrans timestamp on passive Fast Open")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241001200517.2756803-4-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix tcp_enter_recovery() so that if there are no retransmits out then
we zero retrans_stamp when entering fast recovery. This is necessary
to fix two buggy behaviors.
Currently a non-zero retrans_stamp value can persist across multiple
back-to-back loss recovery episodes. This is because we generally only
clears retrans_stamp if we are completely done with loss recoveries,
and get to tcp_try_to_open() and find !tcp_any_retrans_done(sk). This
behavior causes two bugs:
(1) When a loss recovery episode (CA_Loss or CA_Recovery) is followed
immediately by a new CA_Recovery, the retrans_stamp value can persist
and can be a time before this new CA_Recovery episode starts. That
means that timestamp-based undo will be using the wrong retrans_stamp
(a value that is too old) when comparing incoming TS ecr values to
retrans_stamp to see if the current fast recovery episode can be
undone.
(2) If there is a roughly minutes-long sequence of back-to-back fast
recovery episodes, one after another (e.g. in a shallow-buffered or
policed bottleneck), where each fast recovery successfully makes
forward progress and recovers one window of sequence space (but leaves
at least one retransmit in flight at the end of the recovery),
followed by several RTOs, then the ETIMEDOUT check may be using the
wrong retrans_stamp (a value set at the start of the first fast
recovery in the sequence). This can cause a very premature ETIMEDOUT,
killing the connection prematurely.
This commit changes the code to zero retrans_stamp when entering fast
recovery, when this is known to be safe (no retransmits are out in the
network). That ensures that when starting a fast recovery episode, and
it is safe to do so, retrans_stamp is set when we send the fast
retransmit packet. That addresses both bug (1) and bug (2) by ensuring
that (if no retransmits are out when we start a fast recovery) we use
the initial fast retransmit of this fast recovery as the time value
for undo and ETIMEDOUT calculations.
This makes intuitive sense, since the start of a new fast recovery
episode (in a scenario where no lost packets are out in the network)
means that the connection has made forward progress since the last RTO
or fast recovery, and we should thus "restart the clock" used for both
undo and ETIMEDOUT logic.
Note that if when we start fast recovery there *are* retransmits out
in the network, there can still be undesirable (1)/(2) issues. For
example, after this patch we can still have the (1) and (2) problems
in cases like this:
+ round 1: sender sends flight 1
+ round 2: sender receives SACKs and enters fast recovery 1,
retransmits some packets in flight 1 and then sends some new data as
flight 2
+ round 3: sender receives some SACKs for flight 2, notes losses, and
retransmits some packets to fill the holes in flight 2
+ fast recovery has some lost retransmits in flight 1 and continues
for one or more rounds sending retransmits for flight 1 and flight 2
+ fast recovery 1 completes when snd_una reaches high_seq at end of
flight 1
+ there are still holes in the SACK scoreboard in flight 2, so we
enter fast recovery 2, but some retransmits in the flight 2 sequence
range are still in flight (retrans_out > 0), so we can't execute the
new retrans_stamp=0 added here to clear retrans_stamp
It's not yet clear how to fix these remaining (1)/(2) issues in an
efficient way without breaking undo behavior, given that retrans_stamp
is currently used for undo and ETIMEDOUT. Perhaps the optimal (but
expensive) strategy would be to set retrans_stamp to the timestamp of
the earliest outstanding retransmit when entering fast recovery. But
at least this commit makes things better.
Note that this does not change the semantics of retrans_stamp; it
simply makes retrans_stamp accurate in some cases where it was not
before:
(1) Some loss recovery, followed by an immediate entry into a fast
recovery, where there are no retransmits out when entering the fast
recovery.
(2) When a TFO server has a SYNACK retransmit that sets retrans_stamp,
and then the ACK that completes the 3-way handshake has SACK blocks
that trigger a fast recovery. In this case when entering fast recovery
we want to zero out the retrans_stamp from the TFO SYNACK retransmit,
and set the retrans_stamp based on the timestamp of the fast recovery.
We introduce a tcp_retrans_stamp_cleanup() helper, because this
two-line sequence already appears in 3 places and is about to appear
in 2 more as a result of this bug fix patch series. Once this bug fix
patches series in the net branch makes it into the net-next branch
we'll update the 3 other call sites to use the new helper.
This is a long-standing issue. The Fixes tag below is chosen to be the
oldest commit at which the patch will apply cleanly, which is from
Linux v3.5 in 2012.
Fixes: 1fbc340514 ("tcp: early retransmit: tcp_enter_recovery()")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241001200517.2756803-3-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix the TCP loss recovery undo logic in tcp_packet_delayed() so that
it can trigger undo even if TSQ prevents a fast recovery episode from
reaching tcp_retransmit_skb().
Geumhwan Yu <geumhwan.yu@samsung.com> recently reported that after
this commit from 2019:
commit bc9f38c832 ("tcp: avoid unconditional congestion window undo
on SYN retransmit")
...and before this fix we could have buggy scenarios like the
following:
+ Due to reordering, a TCP connection receives some SACKs and enters a
spurious fast recovery.
+ TSQ prevents all invocations of tcp_retransmit_skb(), because many
skbs are queued in lower layers of the sending machine's network
stack; thus tp->retrans_stamp remains 0.
+ The connection receives a TCP timestamp ECR value echoing a
timestamp before the fast recovery, indicating that the fast
recovery was spurious.
+ The connection fails to undo the spurious fast recovery because
tp->retrans_stamp is 0, and thus tcp_packet_delayed() returns false,
due to the new logic in the 2019 commit: commit bc9f38c832 ("tcp:
avoid unconditional congestion window undo on SYN retransmit")
This fix tweaks the logic to be more similar to the
tcp_packet_delayed() logic before bc9f38c832, except that we take
care not to be fooled by the FLAG_SYN_ACKED code path zeroing out
tp->retrans_stamp (the bug noted and fixed by Yuchung in
bc9f38c832).
Note that this returns the high-level behavior of tcp_packet_delayed()
to again match the comment for the function, which says: "Nothing was
retransmitted or returned timestamp is less than timestamp of the
first retransmission." Note that this comment is in the original
2005-04-16 Linux git commit, so this is evidently long-standing
behavior.
Fixes: bc9f38c832 ("tcp: avoid unconditional congestion window undo on SYN retransmit")
Reported-by: Geumhwan Yu <geumhwan.yu@samsung.com>
Diagnosed-by: Geumhwan Yu <geumhwan.yu@samsung.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241001200517.2756803-2-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Abhishek Chauhan says:
====================
Fix AQR PMA capabilities
Patch 1:-
AQR115c reports incorrect PMA capabilities which includes
10G/5G and also incorrectly disables capabilities like autoneg
and 10Mbps support.
AQR115c as per the Marvell databook supports speeds up to 2.5Gbps
with autonegotiation.
Patch 2:-
Remove the use of phy_set_max_speed in phy driver as the
function is mainly used in MAC driver to set the max
speed.
Instead use get_features to fix up Phy PMA capabilities for
AQR111, AQR111B0, AQR114C and AQCS109
====================
Link: https://patch.msgid.link/20241001224626.2400222-1-quic_abchauha@quicinc.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
