platform/kernel/linux-rpi.git
2 years agonet: phy: Add support for 1PPS out and external timestamps
Jonathan Lemon [Wed, 22 Jun 2022 05:04:54 +0000 (22:04 -0700)]
net: phy: Add support for 1PPS out and external timestamps

The perout function is used to generate a 1PPS signal, synchronized
to the PHC.  This is accomplished by a using the hardware oneshot
functionality, which is reset by a timer.

The external timestamp function is set up for a 1PPS input pulse,
and uses a timer to poll for temestamps.

Both functions use the SYNC_OUT/SYNC_IN1 pin, so cannot run
simultaneously.

Co-developed-by: Lasse Johnsen <l@ssejohnsen.me>
Signed-off-by: Lasse Johnsen <l@ssejohnsen.me>
Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: broadcom: Add PTP support for some Broadcom PHYs.
Jonathan Lemon [Wed, 22 Jun 2022 05:04:53 +0000 (22:04 -0700)]
net: phy: broadcom: Add PTP support for some Broadcom PHYs.

This adds PTP support for BCM54210E Broadcom PHYs, in particular,
the BCM54213PE, as used in the Rasperry PI CM4.  It has only been
tested on that hardware.

Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: broadcom: Add Broadcom PTP hooks to bcm-phy-lib
Jonathan Lemon [Wed, 22 Jun 2022 05:04:52 +0000 (22:04 -0700)]
net: phy: broadcom: Add Broadcom PTP hooks to bcm-phy-lib

Add 'struct bcm_ptp_private' to bcm54xx_phy_priv which points to
an optional PTP structure attached to the PHY.  This is allocated
on probe if PHY PTP support is configured, and if the driver supports
PTP for the specified PHY.

Add the bcm_ptp_probe(), bcm_ptp_config_init() and bcn_ptp_stop()
API functions to the bcm-phy library.

Signed-off-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'net-dsa-mv88e6xxx-get-rid-of-speed_max'
Jakub Kicinski [Fri, 24 Jun 2022 03:26:47 +0000 (20:26 -0700)]
Merge branch 'net-dsa-mv88e6xxx-get-rid-of-speed_max'

Russell King says:

====================
net: dsa: mv88e6xxx: get rid of SPEED_MAX

This series does two things:

1. it gets rid of mv88e6065_port_set_speed_duplex() which is completely
   unused (do we support this device? I couldn't find it in the tables
   in chip.c) This has a max speed of 200Mbps which we don't support.

2. get rid of the SPEED_MAX constant, which is used to configure a DSA
   or CPU port to their maximum speed during initialisation. We no
   longer need this as we can derive the maximum port speed from the
   mac_capabilities instead.

The reason for making this change is in preparation for phylink to be
used by DSA for CPU ports. This omission has come back to bite us with
the conversion of DSA drivers to phylink_pcs, since phylink_pcs won't
get used unless phylink is being used. Particularly with this driver,
it is very common for DT descriptions to omit the fixed-link details
which means "use maximum speed".

It will eventually be necessary to hoist the selection of "max speed"
into the DSA layer (trivial) and also have a way for the DSA driver
to tell the DSA layer which interface it should be using for these
ports.
====================

Link: https://lore.kernel.org/r/YrGQBssOvQBZiDS4@shell.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: mv88e6xxx: get rid of SPEED_MAX setting
Russell King [Tue, 21 Jun 2022 09:37:40 +0000 (10:37 +0100)]
net: dsa: mv88e6xxx: get rid of SPEED_MAX setting

Currently, all the device specific speed setting functions convert
SPEED_MAX to the actual speed of the port. Rather than having each
of the mv88e6xxx chip specifics handling SPEED_MAX, derive it from
the mac_capabilities instead.

This is only needed for CPU and DSA ports, so move the logic up into
mv88e6xxx_setup_port() - which allows us to kill off all users of
SPEED_MAX throughout the driver.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: mv88e6xxx: remove mv88e6065 dead code
Russell King (Oracle) [Tue, 21 Jun 2022 09:37:35 +0000 (10:37 +0100)]
net: dsa: mv88e6xxx: remove mv88e6065 dead code

Remove mv88e6065_port_set_speed_duplex() - this is never called, and
thus is completely redundant.

Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Jakub Kicinski [Thu, 23 Jun 2022 19:33:24 +0000 (12:33 -0700)]
Merge git://git./linux/kernel/git/netdev/net

No conflicts.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'net-5.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 23 Jun 2022 14:01:01 +0000 (09:01 -0500)]
Merge tag 'net-5.19-rc4' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Paolo Abeni:
 "Including fixes from bpf and netfilter.

  Current release - regressions:

   - netfilter: cttimeout: fix slab-out-of-bounds read in
     cttimeout_net_exit

Current release - new code bugs:

   - bpf: ftrace: keep address offset in ftrace_lookup_symbols

   - bpf: force cookies array to follow symbols sorting

  Previous releases - regressions:

   - ipv4: ping: fix bind address validity check

   - tipc: fix use-after-free read in tipc_named_reinit

   - eth: veth: add updating of trans_start

  Previous releases - always broken:

   - sock: redo the psock vs ULP protection check

   - netfilter: nf_dup_netdev: fix skb_under_panic

   - bpf: fix request_sock leak in sk lookup helpers

   - eth: igb: fix a use-after-free issue in igb_clean_tx_ring

   - eth: ice: prohibit improper channel config for DCB

   - eth: at803x: fix null pointer dereference on AR9331 phy

   - eth: virtio_net: fix xdp_rxq_info bug after suspend/resume

  Misc:

   - eth: hinic: replace memcpy() with direct assignment"

* tag 'net-5.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (47 commits)
  net: openvswitch: fix parsing of nw_proto for IPv6 fragments
  sock: redo the psock vs ULP protection check
  Revert "net/tls: fix tls_sk_proto_close executed repeatedly"
  virtio_net: fix xdp_rxq_info bug after suspend/resume
  igb: Make DMA faster when CPU is active on the PCIe link
  net: dsa: qca8k: reduce mgmt ethernet timeout
  net: dsa: qca8k: reset cpu port on MTU change
  MAINTAINERS: Add a maintainer for OCP Time Card
  hinic: Replace memcpy() with direct assignment
  Revert "drivers/net/ethernet/neterion/vxge: Fix a use-after-free bug in vxge-main.c"
  net: phy: smsc: Disable Energy Detect Power-Down in interrupt mode
  ice: ethtool: Prohibit improper channel config for DCB
  ice: ethtool: advertise 1000M speeds properly
  ice: Fix switchdev rules book keeping
  ice: ignore protocol field in GTP offload
  netfilter: nf_dup_netdev: add and use recursion counter
  netfilter: nf_dup_netdev: do not push mac header a second time
  selftests: netfilter: correct PKTGEN_SCRIPT_PATHS in nft_concat_range.sh
  net/tls: fix tls_sk_proto_close executed repeatedly
  erspan: do not assume transport header is always set
  ...

2 years agoMerge tag 'mmc-v5.19-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Thu, 23 Jun 2022 13:55:37 +0000 (08:55 -0500)]
Merge tag 'mmc-v5.19-rc2' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:

 - mtk-sd: Fix dma hang issues

 - sdhci-pci-o2micro: Fix card detect by dealing with debouncing

* tag 'mmc-v5.19-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: mediatek: wait dma stop bit reset to 0
  mmc: sdhci-pci-o2micro: Fix card detect by dealing with debouncing

2 years agoMerge tag 'sound-5.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 23 Jun 2022 13:44:00 +0000 (08:44 -0500)]
Merge tag 'sound-5.19-rc4' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "All small changes, mostly device-specific:

   - A regression fix for PCM WC-page allocation on x86

   - A regression fix for i915 audio component binding

   - Fixes for (longstanding) beep handling bug

   - Runtime PM fixes for Intel LPE HDMI audio

   - A couple of pending FireWire fixes

   - Usual HD-audio and USB-audio quirks, new Intel dspconf entries"

* tag 'sound-5.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda/realtek: Add quirk for Clevo NS50PU
  ALSA: hda: Fix discovery of i915 graphics PCI device
  ALSA: hda/via: Fix missing beep setup
  ALSA: hda/conexant: Fix missing beep setup
  ALSA: memalloc: Drop x86-specific hack for WC allocations
  ALSA: hda/realtek: Add quirk for Clevo PD70PNT
  ALSA: x86: intel_hdmi_audio: use pm_runtime_resume_and_get()
  ALSA: x86: intel_hdmi_audio: enable pm_runtime and set autosuspend delay
  ALSA: hda: intel-nhlt: remove use of __func__ in dev_dbg
  ALSA: hda: intel-dspcfg: use SOF for UpExtreme and UpExtreme11 boards
  firewire: convert sysfs sprintf/snprintf family to sysfs_emit
  firewire: cdev: fix potential leak of kernel stack due to uninitialized value
  ALSA: hda/realtek: Apply fixup for Lenovo Yoga Duet 7 properly
  ALSA: hda/realtek - ALC897 headset MIC no sound
  ALSA: usb-audio: US16x08: Move overflow check before array access
  ALSA: hda/realtek: Add mute LED quirk for HP Omen laptop

2 years agonfp: add 'ethtool --identify' support
Sixiang Chen [Wed, 22 Jun 2022 08:39:38 +0000 (10:39 +0200)]
nfp: add 'ethtool --identify' support

Add support for ethtool -p|--identify
by enabling blinking of the panel LED if supported by the NIC firmware.

Signed-off-by: Sixiang Chen <sixiang.chen@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20220622083938.291548-1-simon.horman@corigine.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: openvswitch: fix parsing of nw_proto for IPv6 fragments
Rosemarie O'Riorden [Tue, 21 Jun 2022 20:48:45 +0000 (16:48 -0400)]
net: openvswitch: fix parsing of nw_proto for IPv6 fragments

When a packet enters the OVS datapath and does not match any existing
flows installed in the kernel flow cache, the packet will be sent to
userspace to be parsed, and a new flow will be created. The kernel and
OVS rely on each other to parse packet fields in the same way so that
packets will be handled properly.

As per the design document linked below, OVS expects all later IPv6
fragments to have nw_proto=44 in the flow key, so they can be correctly
matched on OpenFlow rules. OpenFlow controllers create pipelines based
on this design.

This behavior was changed by the commit in the Fixes tag so that
nw_proto equals the next_header field of the last extension header.
However, there is no counterpart for this change in OVS userspace,
meaning that this field is parsed differently between OVS and the
kernel. This is a problem because OVS creates actions based on what is
parsed in userspace, but the kernel-provided flow key is used as a match
criteria, as described in Documentation/networking/openvswitch.rst. This
leads to issues such as packets incorrectly matching on a flow and thus
the wrong list of actions being applied to the packet. Such changes in
packet parsing cannot be implemented without breaking the userspace.

The offending commit is partially reverted to restore the expected
behavior.

The change technically made sense and there is a good reason that it was
implemented, but it does not comply with the original design of OVS.
If in the future someone wants to implement such a change, then it must
be user-configurable and disabled by default to preserve backwards
compatibility with existing OVS versions.

Cc: stable@vger.kernel.org
Fixes: fa642f08839b ("openvswitch: Derive IP protocol number for IPv6 later frags")
Link: https://docs.openvswitch.org/en/latest/topics/design/#fragments
Signed-off-by: Rosemarie O'Riorden <roriorden@redhat.com>
Acked-by: Eelco Chaudron <echaudro@redhat.com>
Link: https://lore.kernel.org/r/20220621204845.9721-1-roriorden@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agosock: redo the psock vs ULP protection check
Jakub Kicinski [Mon, 20 Jun 2022 19:13:53 +0000 (12:13 -0700)]
sock: redo the psock vs ULP protection check

Commit 8a59f9d1e3d4 ("sock: Introduce sk->sk_prot->psock_update_sk_prot()")
has moved the inet_csk_has_ulp(sk) check from sk_psock_init() to
the new tcp_bpf_update_proto() function. I'm guessing that this
was done to allow creating psocks for non-inet sockets.

Unfortunately the destruction path for psock includes the ULP
unwind, so we need to fail the sk_psock_init() itself.
Otherwise if ULP is already present we'll notice that later,
and call tcp_update_ulp() with the sk_proto of the ULP
itself, which will most likely result in the ULP looping
its callbacks.

Fixes: 8a59f9d1e3d4 ("sock: Introduce sk->sk_prot->psock_update_sk_prot()")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Tested-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/r/20220620191353.1184629-2-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoRevert "net/tls: fix tls_sk_proto_close executed repeatedly"
Jakub Kicinski [Mon, 20 Jun 2022 19:13:52 +0000 (12:13 -0700)]
Revert "net/tls: fix tls_sk_proto_close executed repeatedly"

This reverts commit 69135c572d1f84261a6de2a1268513a7e71753e2.

This commit was just papering over the issue, ULP should not
get ->update() called with its own sk_prot. Each ULP would
need to add this check.

Fixes: 69135c572d1f ("net/tls: fix tls_sk_proto_close executed repeatedly")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/r/20220620191353.1184629-1-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoMerge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next...
Jakub Kicinski [Thu, 23 Jun 2022 02:38:58 +0000 (19:38 -0700)]
Merge branch '40GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
40GbE Intel Wired LAN Driver Updates 2022-06-21

This series contains updates to i40e driver only.

Mateusz adds support for using the speed option in ethtool.

Minghao Chi removes unneeded synchronize_irq() calls.

Bernard Zhao removes unneeded NULL check.

* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  intel/i40e: delete if NULL check before dev_kfree_skb
  i40e: Remove unnecessary synchronize_irq() before free_irq()
  i40e: Add support for ethtool -s <interface> speed <speed in Mb>
====================

Link: https://lore.kernel.org/r/20220621225930.632741-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agovirtio_net: fix xdp_rxq_info bug after suspend/resume
Stephan Gerhold [Tue, 21 Jun 2022 11:48:44 +0000 (13:48 +0200)]
virtio_net: fix xdp_rxq_info bug after suspend/resume

The following sequence currently causes a driver bug warning
when using virtio_net:

  # ip link set eth0 up
  # echo mem > /sys/power/state (or e.g. # rtcwake -s 10 -m mem)
  <resume>
  # ip link set eth0 down

  Missing register, driver bug
  WARNING: CPU: 0 PID: 375 at net/core/xdp.c:138 xdp_rxq_info_unreg+0x58/0x60
  Call trace:
   xdp_rxq_info_unreg+0x58/0x60
   virtnet_close+0x58/0xac
   __dev_close_many+0xac/0x140
   __dev_change_flags+0xd8/0x210
   dev_change_flags+0x24/0x64
   do_setlink+0x230/0xdd0
   ...

This happens because virtnet_freeze() frees the receive_queue
completely (including struct xdp_rxq_info) but does not call
xdp_rxq_info_unreg(). Similarly, virtnet_restore() sets up the
receive_queue again but does not call xdp_rxq_info_reg().

Actually, parts of virtnet_freeze_down() and virtnet_restore_up()
are almost identical to virtnet_close() and virtnet_open(): only
the calls to xdp_rxq_info_(un)reg() are missing. This means that
we can fix this easily and avoid such problems in the future by
just calling virtnet_close()/open() from the freeze/restore handlers.

Aside from adding the missing xdp_rxq_info calls the only difference
is that the refill work is only cancelled if netif_running(). However,
this should not make any functional difference since the refill work
should only be active if the network interface is actually up.

Fixes: 754b8a21a96d ("virtio_net: setup xdp_rxq_info")
Signed-off-by: Stephan Gerhold <stephan.gerhold@kernkonzept.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20220621114845.3650258-1-stephan.gerhold@kernkonzept.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net...
Jakub Kicinski [Thu, 23 Jun 2022 01:59:29 +0000 (18:59 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-06-21

This series contains updates to ice driver only.

Marcin fixes GTP filters by allowing ignoring of the inner ethertype field.

Wojciech adds VSI handle tracking in order to properly distinguish similar
filters for removal.

Anatolii removes ability to set 1000baseT and 1000baseX fields
concurrently which caused link issues. He also disallows setting
channels to less than the number of Traffic Classes which would cause
NULL pointer dereference.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  ice: ethtool: Prohibit improper channel config for DCB
  ice: ethtool: advertise 1000M speeds properly
  ice: Fix switchdev rules book keeping
  ice: ignore protocol field in GTP offload
====================

Link: https://lore.kernel.org/r/20220621224756.631765-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoraw: remove unused variables from raw6_icmp_error()
Eric Dumazet [Wed, 22 Jun 2022 03:23:03 +0000 (03:23 +0000)]
raw: remove unused variables from raw6_icmp_error()

saddr and daddr are set but not used.

Fixes: ba44f8182ec2 ("raw: use more conventional iterators")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/r/20220622032303.159394-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoigb: Make DMA faster when CPU is active on the PCIe link
Kai-Heng Feng [Tue, 21 Jun 2022 22:10:56 +0000 (15:10 -0700)]
igb: Make DMA faster when CPU is active on the PCIe link

Intel I210 on some Intel Alder Lake platforms can only achieve ~750Mbps
Tx speed via iperf. The RR2DCDELAY shows around 0x2xxx DMA delay, which
will be significantly lower when 1) ASPM is disabled or 2) SoC package
c-state stays above PC3. When the RR2DCDELAY is around 0x1xxx the Tx
speed can reach to ~950Mbps.

According to the I210 datasheet "8.26.1 PCIe Misc. Register - PCIEMISC",
"DMA Idle Indication" doesn't seem to tie to DMA coalesce anymore, so
set it to 1b for "DMA is considered idle when there is no Rx or Tx AND
when there are no TLPs indicating that CPU is active detected on the
PCIe link (such as the host executes CSR or Configuration register read
or write operation)" and performing Tx should also fall under "active
CPU on PCIe link" case.

In addition to that, commit b6e0c419f040 ("igb: Move DMA Coalescing init
code to separate function.") seems to wrongly changed from enabling
E1000_PCIEMISC_LX_DECISION to disabling it, also fix that.

Fixes: b6e0c419f040 ("igb: Move DMA Coalescing init code to separate function.")
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20220621221056.604304-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: Add support for AQR113C EPHY
Bhadram Varka [Tue, 21 Jun 2022 03:40:27 +0000 (09:10 +0530)]
net: phy: Add support for AQR113C EPHY

Add support multi-gigabit and single-port Ethernet
PHY transceiver (AQR113C).

Signed-off-by: Bhadram Varka <vbhadram@nvidia.com>
Link: https://lore.kernel.org/r/20220621034027.56508-1-vbhadram@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: pcs: lynx: use mdiodev accessors
Russell King (Oracle) [Tue, 21 Jun 2022 08:53:02 +0000 (09:53 +0100)]
net: pcs: lynx: use mdiodev accessors

Use the recently introduced mdiodev accessors for the lynx PCS.

Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://lore.kernel.org/r/E1o3Zd8-002yHI-G2@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: reduce mgmt ethernet timeout
Christian Marangi [Tue, 21 Jun 2022 15:16:33 +0000 (17:16 +0200)]
net: dsa: qca8k: reduce mgmt ethernet timeout

The current mgmt ethernet timeout is set to 100ms. This value is too
big and would slow down any mdio command in case the mgmt ethernet
packet have some problems on the receiving part.
Reduce it to just 5ms to handle case when some operation are done on the
master port that would cause the mgmt ethernet to not work temporarily.

Fixes: 5950c7c0a68c ("net: dsa: qca8k: add support for mgmt read/write in Ethernet packet")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20220621151633.11741-1-ansuelsmth@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: reset cpu port on MTU change
Christian Marangi [Tue, 21 Jun 2022 15:11:22 +0000 (17:11 +0200)]
net: dsa: qca8k: reset cpu port on MTU change

It was discovered that the Documentation lacks of a fundamental detail
on how to correctly change the MAX_FRAME_SIZE of the switch.

In fact if the MAX_FRAME_SIZE is changed while the cpu port is on, the
switch panics and cease to send any packet. This cause the mgmt ethernet
system to not receive any packet (the slow fallback still works) and
makes the device not reachable. To recover from this a switch reset is
required.

To correctly handle this, turn off the cpu ports before changing the
MAX_FRAME_SIZE and turn on again after the value is applied.

Fixes: f58d2598cf70 ("net: dsa: qca8k: implement the port MTU callbacks")
Cc: stable@vger.kernel.org
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20220621151122.10220-1-ansuelsmth@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoisdn: mISDN: hfcsusb: drop unexpected word "the" in the comments
Jiang Jian [Tue, 21 Jun 2022 11:45:29 +0000 (19:45 +0800)]
isdn: mISDN: hfcsusb: drop unexpected word "the" in the comments

there is an unexpected word "the" in the comments that need to be dropped

file: ./drivers/isdn/hardware/mISDN/hfcsusb.c
line: 1560
 /* set USB_SIZE_I to match the the wMaxPacketSize for ISO transfers */
changed to
 /* set USB_SIZE_I to match the wMaxPacketSize for ISO transfers */

Signed-off-by: Jiang Jian <jiangjian@cdjrlc.com>
Link: https://lore.kernel.org/r/20220621114529.108079-1-jiangjian@cdjrlc.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: ipa: remove unexpected word "the"
Jiang Jian [Tue, 21 Jun 2022 08:50:01 +0000 (16:50 +0800)]
net: ipa: remove unexpected word "the"

there is an unexpected word "the" in the comments that need to be removed

Signed-off-by: Jiang Jian <jiangjian@cdjrlc.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20220621085001.61320-1-jiangjian@cdjrlc.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agocxgb4vf: remove unexpected word "the"
Jiang Jian [Tue, 21 Jun 2022 08:45:37 +0000 (16:45 +0800)]
cxgb4vf: remove unexpected word "the"

there is an unexpected word "the" in the comments that need to be removed

Signed-off-by: Jiang Jian <jiangjian@cdjrlc.com>
Link: https://lore.kernel.org/r/20220621084537.58402-1-jiangjian@cdjrlc.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMAINTAINERS: Add a maintainer for OCP Time Card
Vadim Fedorenko [Tue, 21 Jun 2022 23:31:31 +0000 (02:31 +0300)]
MAINTAINERS: Add a maintainer for OCP Time Card

I've been contributing and reviewing patches for ptp_ocp driver for
some time and I'm taking care of it's github mirror. On Jakub's
suggestion, I would like to step forward and become a maintainer for
this driver. This patch adds a dedicated entry to MAINTAINERS.

Signed-off-by: Vadim Fedorenko <vadfed@fb.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/r/20220621233131.21240-1-vfedorenko@novek.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoamt: remove unnecessary (void*) conversions
Yu Zhe [Tue, 21 Jun 2022 02:16:48 +0000 (10:16 +0800)]
amt: remove unnecessary (void*) conversions

Remove unnecessary void* type castings.

Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
Acked-by: Taehee Yoo <ap420073@gmail.com>
Link: https://lore.kernel.org/r/20220621021648.2544-1-yuzhe@nfschina.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'linux-kselftest-fixes-5.19-rc4' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Wed, 22 Jun 2022 19:08:06 +0000 (14:08 -0500)]
Merge tag 'linux-kselftest-fixes-5.19-rc4' of git://git./linux/kernel/git/shuah/linux-kselftest

Pull Kselftest fixes from Shuah Khan:
 "Compile time fixes and run-time resources leaks:

   - Fix clang cross compilation

   - Fix resource leak when return error

   - fix compile error for dma_map_benchmark

   - Fix regression - make use of GUP_TEST_FILE macro"

* tag 'linux-kselftest-fixes-5.19-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  selftests: make use of GUP_TEST_FILE macro
  selftests: vm: Fix resource leak when return error
  selftests dma: fix compile error for dma_map_benchmark
  selftests: Fix clang cross compilation

2 years agohinic: Replace memcpy() with direct assignment
Kees Cook [Thu, 16 Jun 2022 05:23:12 +0000 (22:23 -0700)]
hinic: Replace memcpy() with direct assignment

Under CONFIG_FORTIFY_SOURCE=y and CONFIG_UBSAN_BOUNDS=y, Clang is bugged
here for calculating the size of the destination buffer (0x10 instead of
0x14). This copy is a fixed size (sizeof(struct fw_section_info_st)), with
the source and dest being struct fw_section_info_st, so the memcpy should
be safe, assuming the index is within bounds, which is UBSAN_BOUNDS's
responsibility to figure out.

Avoid the whole thing and just do a direct assignment. This results in
no change to the executable code.

[This is a duplicate of commit 2c0ab32b73cf ("hinic: Replace memcpy()
 with direct assignment") which was applied to net-next.]

Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Tom Rix <trix@redhat.com>
Cc: llvm@lists.linux.dev
Link: https://github.com/ClangBuiltLinux/linux/issues/1592
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org> # build
Link: https://lore.kernel.org/r/20220616052312.292861-1-keescook@chromium.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoALSA: hda/realtek: Add quirk for Clevo NS50PU
Tim Crawford [Wed, 22 Jun 2022 15:00:17 +0000 (09:00 -0600)]
ALSA: hda/realtek: Add quirk for Clevo NS50PU

Fixes headset detection on Clevo NS50PU.

Signed-off-by: Tim Crawford <tcrawford@system76.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220622150017.9897-1-tcrawford@system76.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
2 years agoMerge tag '9p-for-5.19-rc4' of https://github.com/martinetd/linux
Linus Torvalds [Wed, 22 Jun 2022 13:09:49 +0000 (08:09 -0500)]
Merge tag '9p-for-5.19-rc4' of https://github.com/martinetd/linux

Pull 9pfs fixes from Dominique Martinet:
 "A couple of fid refcount and fscache fixes:

   - fid refcounting was incorrect in some corner cases and would leak
     resources, only freed at umount time. The first three commits fix
     three such cases

   - 'cache=loose' or fscache was broken when trying to write a partial
     page to a file with no read permission since the rework a few
     releases ago.

     The fix taken here is just to restore old behavior of using the
     special 'writeback_fid' for such reads, which is open as root/RDWR
     and such not get complains that we try to read on a WRONLY fid.

     Long-term it'd be nice to get rid of this and not issue the read at
     all (skip cache?) in such cases, but that direction hasn't
     progressed"

* tag '9p-for-5.19-rc4' of https://github.com/martinetd/linux:
  9p: fix EBADF errors in cached mode
  9p: Fix refcounting during full path walks for fid lookups
  9p: fix fid refcount leak in v9fs_vfs_get_link
  9p: fix fid refcount leak in v9fs_vfs_atomic_open_dotl

2 years agoRevert "drivers/net/ethernet/neterion/vxge: Fix a use-after-free bug in vxge-main.c"
Jakub Kicinski [Mon, 20 Jun 2022 19:12:37 +0000 (12:12 -0700)]
Revert "drivers/net/ethernet/neterion/vxge: Fix a use-after-free bug in vxge-main.c"

This reverts commit 8fc74d18639a2402ca52b177e990428e26ea881f.

BAR0 is the main (only?) register bank for this device. We most
obviously can't unmap it before the netdev is unregistered.
This was pointed out in review but the patch got reposted and
merged, anyway.

The author of the patch was only testing it with a QEMU model,
which I presume does not emulate enough for the netdev to be brought
up (author's replies are not visible in lore because they kept sending
their emails in HTML).

Link: https://lore.kernel.org/all/20220616085059.680dc215@kernel.org/
Fixes: 8fc74d18639a ("drivers/net/ethernet/neterion/vxge: Fix a use-after-free bug in vxge-main.c")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'af_unix-per-netns-socket-hash'
David S. Miller [Wed, 22 Jun 2022 11:59:43 +0000 (12:59 +0100)]
Merge branch 'af_unix-per-netns-socket-hash'

Kuniyuki Iwashima says:

====================
af_unix: Introduce per-netns socket hash table.

This series replaces unix_socket_table with a per-netns hash table and
reduces lock contention and time on iterating over the list.

Note the 3rd-6th patches can be a single patch, but for ease of review,
they are split into small changes without breakage.

Changes:
  v3:
    6th:
      * Remove unix_table_locks from comments.
      * Remove missed spin_unlock(&unix_table_locks) in
        unix_lookup_by_ino() (kernel test robot)

  v2: https://lore.kernel.org/netdev/20220620185151.65294-1-kuniyu@amazon.com/
    3rd:
      * Update changelog
      * Remove holes from per-netns hash table structure
      * Use kvmalloc_array() instead of kmalloc() (Eric Dumazet)
      * Remove unnecessary parts in af_unix_init() (Eric Dumazet)
      * Move `err_sysctl` label into ifdef block (kernel test robot)
      * Remove struct netns_unix from struct net if CONFIG_UNIX is disabled
    4th:
      * Use spin_lock_nested() (kernel test robot)

  v1: https://lore.kernel.org/netdev/20220616234714.4291-1-kuniyu@amazon.com/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoaf_unix: Remove unix_table_locks.
Kuniyuki Iwashima [Tue, 21 Jun 2022 17:19:13 +0000 (10:19 -0700)]
af_unix: Remove unix_table_locks.

unix_table_locks are to protect the global hash table, unix_socket_table.
The previous commit removed it, so let's clean up the unnecessary locks.

Here is a test result on EC2 c5.9xlarge where 10 processes run concurrently
in different netns and bind 100,000 sockets for each.

  without this series : 1m 38s
  with this series    :    11s

It is ~10x faster because the global hash table is split into 10 netns in
this case.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoaf_unix: Put a socket into a per-netns hash table.
Kuniyuki Iwashima [Tue, 21 Jun 2022 17:19:12 +0000 (10:19 -0700)]
af_unix: Put a socket into a per-netns hash table.

This commit replaces the global hash table with a per-netns one and removes
the global one.

We now link a socket in each netns's hash table so we can save some netns
comparisons when iterating through a hash bucket.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoaf_unix: Acquire/Release per-netns hash table's locks.
Kuniyuki Iwashima [Tue, 21 Jun 2022 17:19:11 +0000 (10:19 -0700)]
af_unix: Acquire/Release per-netns hash table's locks.

This commit adds extra spin_lock/spin_unlock() for a per-netns
hash table inside the existing ones for unix_table_locks.

As of this commit, sockets are still linked in the global hash
table.  After putting sockets in a per-netns hash table and
removing the old one in the next patch, we remove the global
locks in the last patch.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoaf_unix: Define a per-netns hash table.
Kuniyuki Iwashima [Tue, 21 Jun 2022 17:19:10 +0000 (10:19 -0700)]
af_unix: Define a per-netns hash table.

This commit adds a per netns hash table for AF_UNIX, which size is fixed
as UNIX_HASH_SIZE for now.

The first implementation defines a per-netns hash table as a single array
of lock and list:

struct unix_hashbucket {
spinlock_t lock;
struct hlist_head head;
};

struct netns_unix {
struct unix_hashbucket *hash;
...
};

But, Eric pointed out memory cost that the structure has holes because of
sizeof(spinlock_t), which is 4 (or more if LOCKDEP is enabled). [0]  It
could be expensive on a host with thousands of netns and few AF_UNIX
sockets.  For this reason, a per-netns hash table uses two dense arrays.

struct unix_table {
spinlock_t *locks;
struct hlist_head *buckets;
};

struct netns_unix {
struct unix_table table;
...
};

Note the length of the list has a significant impact rather than lock
contention, so having shared locks can be an option.  But, per-netns
locks and lists still perform better than the global locks and per-netns
lists. [1]

Also, this patch adds a change so that struct netns_unix disappears from
struct net if CONFIG_UNIX is disabled.

[0]: https://lore.kernel.org/netdev/CANn89iLVxO5aqx16azNU7p7Z-nz5NrnM5QTqOzueVxEnkVTxyg@mail.gmail.com/
[1]: https://lore.kernel.org/netdev/20220617175215.1769-1-kuniyu@amazon.com/

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoaf_unix: Include the whole hash table size in UNIX_HASH_SIZE.
Kuniyuki Iwashima [Tue, 21 Jun 2022 17:19:09 +0000 (10:19 -0700)]
af_unix: Include the whole hash table size in UNIX_HASH_SIZE.

Currently, the size of AF_UNIX hash table is UNIX_HASH_SIZE * 2,
the first half for bind()ed sockets and the second half for unbound
ones.  UNIX_HASH_SIZE * 2 is used to define the table and iterate
over it.

In some places, we use ARRAY_SIZE(unix_socket_table) instead of
UNIX_HASH_SIZE * 2.  However, we cannot use it anymore because we
will allocate the hash table dynamically.  Then, we would have to
add UNIX_HASH_SIZE * 2 in many places, which would be troublesome.

This patch adapts the UNIX_HASH_SIZE definition to include bound
and unbound sockets and defines a new UNIX_HASH_MOD macro to ease
calculations.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoaf_unix: Clean up some sock_net() uses.
Kuniyuki Iwashima [Tue, 21 Jun 2022 17:19:08 +0000 (10:19 -0700)]
af_unix: Clean up some sock_net() uses.

Some functions define a net pointer only for one-shot use.  Others call
sock_net() redundantly even when a net pointer is available.  Let's fix
these and make the code simpler.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mlxsw-unified-bridge-conversion-part-2'
David S. Miller [Wed, 22 Jun 2022 11:56:58 +0000 (12:56 +0100)]
Merge branch 'mlxsw-unified-bridge-conversion-part-2'

Ido Schimmel says:

====================
mlxsw: Unified bridge conversion - part 2/6

This is the second part of the conversion of mlxsw to the unified bridge
model. Part 1 was merged in commit 4336487e30c3 ("Merge branch
'mlxsw-unified-bridge-conversion-part-1'") which includes details about
the new model and the motivation behind the conversion.

This patchset does not begin the conversion, but rather prepares the code
base for it.

Patchset overview:

Patch #1 removes an unnecessary field from one of the FID families.

Patches #2-#7 make various improvements in the layer 2 multicast code,
making it more receptive towards upcoming changes.

Patches #8-#10 prepare the CONFIG_PROFILE command for the unified bridge
model. This command will be used to enable the new model in the last
patchset.

Patches #11-#13 perform small changes in the FID code, preparing it for
upcoming changes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_fid: Implement missing operations for rFID and dummy FID
Amit Cohen [Tue, 21 Jun 2022 08:33:45 +0000 (11:33 +0300)]
mlxsw: spectrum_fid: Implement missing operations for rFID and dummy FID

rFID and dummy FID do not support FID->VNI mapping. Currently, these
families do not implement the vni_{set, clear}() operations. Instead, there
is a check if these functions are implemented.

Similarly, 'SFMR.nve_tunnel_flood_ptr' is not relevant for rFID and dummy
FID, therefore, these families do not implement
nve_flood_index_{set, clear}().

Align the behavior to other unsupported operations, implement the functions
and just return an error or warn. Then, checks like '!ops->vni_set' can be
removed.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_fid: Use 'fid->fid_offset' when setting VNI
Amit Cohen [Tue, 21 Jun 2022 08:33:44 +0000 (11:33 +0300)]
mlxsw: spectrum_fid: Use 'fid->fid_offset' when setting VNI

The previous patch added 'fid_offset' field to FID structure. Now, this
field can be used when VNI is set using SFMR register. Currently
'fid_offset' is set to zero, instead, use the new field which is now set
to zero and in the future will be changed.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_fid: Save 'fid_offset' as part of FID structure
Amit Cohen [Tue, 21 Jun 2022 08:33:43 +0000 (11:33 +0300)]
mlxsw: spectrum_fid: Save 'fid_offset' as part of FID structure

SFMR register contains a 'fid_offset' field which is used when flooding
tables of type FID offset are used.

Currently, the driver sets this field to zero, as flooding tables of type
FID are used.

Using unified bridge model, the driver will use FID offset flooding
tables. As preparation, add 'fid_offset' to 'struct mlxsw_sp_fid'. Then,
use this field instead of passing zero to the function that configures
SFMR.

Set the new field as part of 'ops->setup()', for that, implement this
function for dummy FID and rFID.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: pci: Query resources before and after issuing 'CONFIG_PROFILE' command
Amit Cohen [Tue, 21 Jun 2022 08:33:42 +0000 (11:33 +0300)]
mlxsw: pci: Query resources before and after issuing 'CONFIG_PROFILE' command

Currently, as part of mlxsw_pci_init(), resources are queried from firmware
before issuing the 'CONFIG_PROFILE' command.

There are resources whose size depend on the enablement of the unified
bridge model that is performed via 'CONFIG_PROFILE' command. As a
preparation for unified bridge model, add an additional query after issuing
this command. Both queries are required as KVD sizes are read from
firmware and then are configured as part of 'CONFIG_PROFILE' command.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: cmd: Increase 'config_profile.flood_mode' length
Amit Cohen [Tue, 21 Jun 2022 08:33:41 +0000 (11:33 +0300)]
mlxsw: cmd: Increase 'config_profile.flood_mode' length

Currently, the length of 'config_profile.flood_mode' is defined as 2
bits, while the correct length is 3 bits.

As preparation for unified bridge model, which will use the whole field
length, fix it and increase the field to the correct size.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: Add enumerator for 'config_profile.flood_mode'
Amit Cohen [Tue, 21 Jun 2022 08:33:40 +0000 (11:33 +0300)]
mlxsw: Add enumerator for 'config_profile.flood_mode'

Currently magic constant is used for setting 'flood_mode' as part of
config profile.

As preparation for unified bridge model, which will require different
'flood_mode', add a dedicated enumerator for this field and use it as
part of 'struct mlxsw_config_profile'.

Add the relevant value for unified bridge model.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_switchdev: Handle error in mlxsw_sp_bridge_mdb_mc_enable_sync()
Amit Cohen [Tue, 21 Jun 2022 08:33:39 +0000 (11:33 +0300)]
mlxsw: spectrum_switchdev: Handle error in mlxsw_sp_bridge_mdb_mc_enable_sync()

The above mentioned function calls two functions which return values, but
it ignores them.

Check if the return value is an error, handle it in such case and return
an error back to the caller.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_switchdev: Convert mlxsw_sp_mc_write_mdb_entry() to return int
Amit Cohen [Tue, 21 Jun 2022 08:33:38 +0000 (11:33 +0300)]
mlxsw: spectrum_switchdev: Convert mlxsw_sp_mc_write_mdb_entry() to return int

The function returns 'false' upon failure and 'true' upon success.
Convert it to the usual scheme of returning integer error codes and
align the callers.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_switchdev: Add error path in mlxsw_sp_port_mc_disabled_set()
Amit Cohen [Tue, 21 Jun 2022 08:33:37 +0000 (11:33 +0300)]
mlxsw: spectrum_switchdev: Add error path in mlxsw_sp_port_mc_disabled_set()

The above mentioned function just returns an error in case that
mlxsw_sp_bridge_ports_flood_table_set() fails. That means that the previous
configurations are not cleaned.

Fix it by adding error path to clean the configurations in case of error.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_switchdev: Simplify mlxsw_sp_port_mc_disabled_set()
Amit Cohen [Tue, 21 Jun 2022 08:33:36 +0000 (11:33 +0300)]
mlxsw: spectrum_switchdev: Simplify mlxsw_sp_port_mc_disabled_set()

The above mentioned function is called for each port in bridge when the
multicast behavior is changed. Currently, it updates all the relevant
MID entries only for the first call, but the update of flooding per port
is done only for the port that it was called for it.

To simplify this behavior, it is possible to handle flooding for all the
ports in the first call, then, the other calls will do nothing. For
that, new functions are required to set flooding for all ports, no
'struct mlxsw_sp_port' is required.

This issue was found while extending this function for unified bridge
model, the above mentioned change will ease the extending later.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_switchdev: Do not set 'multicast_enabled' twice
Amit Cohen [Tue, 21 Jun 2022 08:33:35 +0000 (11:33 +0300)]
mlxsw: spectrum_switchdev: Do not set 'multicast_enabled' twice

The function mlxsw_sp_port_mc_disabled_set() sets
'bridge_device->multicast_enabled' twice. Remove the unnecessary setting.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_switchdev: Pass 'struct mlxsw_sp' to mlxsw_sp_bridge_mdb_mc_enable_sync()
Amit Cohen [Tue, 21 Jun 2022 08:33:34 +0000 (11:33 +0300)]
mlxsw: spectrum_switchdev: Pass 'struct mlxsw_sp' to mlxsw_sp_bridge_mdb_mc_enable_sync()

Currently the above mentioned function gets 'struct mlxsw_sp_port',
while the only use of this structure is to get 'mlxsw_sp_port->mlxsw_sp'.

Instead, pass 'struct mlxsw_sp'.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: Remove lag_vid_valid indication
Amit Cohen [Tue, 21 Jun 2022 08:33:33 +0000 (11:33 +0300)]
mlxsw: Remove lag_vid_valid indication

Currently 'struct mlxsw_sp_fid_family' has a field which indicates if
'lag_vid' is valid for use in SFD register.

This is a leftover from using .1Q FIDs instead of emulating them using
.1D FIDs.

Currently when .1Q FIDs are emulated using .1D FIDs, this field is true
for both families, so there is no reason to maintain it.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Jakub Kicinski [Wed, 22 Jun 2022 05:41:40 +0000 (22:41 -0700)]
Merge git://git./linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

1) Use get_random_u32() instead of prandom_u32_state() in nft_meta
   and nft_numgen, from Florian Westphal.

2) Incorrect list head in nfnetlink_cttimeout in recent update coming
   from previous development cycle. Also from Florian.

3) Incorrect path to pktgen scripts for nft_concat_range.sh selftest.
   From Jie2x Zhou.

4) Two fixes for the for nft_fwd and nft_dup egress support, from Florian.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nf_dup_netdev: add and use recursion counter
  netfilter: nf_dup_netdev: do not push mac header a second time
  selftests: netfilter: correct PKTGEN_SCRIPT_PATHS in nft_concat_range.sh
  netfilter: cttimeout: fix slab-out-of-bounds read typo in cttimeout_net_exit
  netfilter: use get_random_u32 instead of prandom
====================

Link: https://lore.kernel.org/r/20220621085618.3975-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: dp83td510: add SQI support
Oleksij Rempel [Mon, 20 Jun 2022 11:56:01 +0000 (13:56 +0200)]
net: phy: dp83td510: add SQI support

Convert MSE (mean-square error) values to SNR and split it SQI (Signal Quality
Indicator) ranges. The used ranges are taken from "OPEN ALLIANCE - Advanced
diagnostic features for 100BASE-T1 automotive Ethernet PHYs"
specification.

Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20220620115601.2035452-1-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: smsc: Disable Energy Detect Power-Down in interrupt mode
Lukas Wunner [Mon, 20 Jun 2022 11:04:50 +0000 (13:04 +0200)]
net: phy: smsc: Disable Energy Detect Power-Down in interrupt mode

Simon reports that if two LAN9514 USB adapters are directly connected
without an intermediate switch, the link fails to come up and link LEDs
remain dark.  The issue was introduced by commit 1ce8b37241ed ("usbnet:
smsc95xx: Forward PHY interrupts to PHY driver to avoid polling").

The PHY suffers from a known erratum wherein link detection becomes
unreliable if Energy Detect Power-Down is used.  In poll mode, the
driver works around the erratum by briefly disabling EDPD for 640 msec
to detect a neighbor, then re-enabling it to save power.

In interrupt mode, no interrupt is signaled if EDPD is used by both link
partners, so it must not be enabled at all.

We'll recoup the power savings by enabling SUSPEND1 mode on affected
LAN95xx chips in a forthcoming commit.

Fixes: 1ce8b37241ed ("usbnet: smsc95xx: Forward PHY interrupts to PHY driver to avoid polling")
Reported-by: Simon Han <z.han@kunbus.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Link: https://lore.kernel.org/r/439a3f3168c2f9d44b5fd9bb8d2b551711316be6.1655714438.git.lukas@wunner.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: smsc: Deduplicate interrupt acknowledgement upon phy_init_hw()
Lukas Wunner [Mon, 20 Jun 2022 09:28:39 +0000 (11:28 +0200)]
net: phy: smsc: Deduplicate interrupt acknowledgement upon phy_init_hw()

Since commit 4c0d2e96ba05 ("net: phy: consider that suspend2ram may cut
off PHY power"), phy_init_hw() invokes both, the ->config_init() and
->config_intr() callbacks.

In the SMSC PHY driver, the latter acknowledges stale interrupts, hence
there's no longer a need to acknowledge them in the former as well.

There are no other callers of ->config_init() besides phy_init_hw().

Drop the redundant code.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
Link: https://lore.kernel.org/r/0254edf48bddc96c6248c4414043a3699e94614a.1655716767.git.lukas@wunner.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoice: ethtool: Prohibit improper channel config for DCB
Anatolii Gerasymenko [Mon, 6 Jun 2022 07:01:21 +0000 (09:01 +0200)]
ice: ethtool: Prohibit improper channel config for DCB

Do not allow setting less channels, than Traffic Classes there are
via ethtool. There must be at least one channel per Traffic Class.

If you set less channels, than Traffic Classes there are, then during
ice_vsi_rebuild there would be allocated only the requested amount
of tx/rx rings in ice_vsi_alloc_arrays. But later in ice_vsi_setup_q_map
there would be requested at least one channel per Traffic Class. This
results in setting num_rxq > alloc_rxq and num_txq > alloc_txq.
Later, there would be a NULL pointer dereference in
ice_vsi_map_rings_to_vectors, because we go beyond of rx_rings or
tx_rings arrays.

Change ice_set_channels() to return error if you try to allocate less
channels, than Traffic Classes there are.
Change ice_vsi_setup_q_map() and ice_vsi_setup_q_map_mqprio() to return
status code instead of void.
Add error handling for ice_vsi_setup_q_map() and
ice_vsi_setup_q_map_mqprio() in ice_vsi_init() and ice_vsi_cfg_tc().

[53753.889983] INFO: Flow control is disabled for this traffic class (0) on this vsi.
[53763.984862] BUG: unable to handle kernel NULL pointer dereference at 0000000000000028
[53763.992915] PGD 14b45f5067 P4D 0
[53763.996444] Oops: 0002 [#1] SMP NOPTI
[53764.000312] CPU: 12 PID: 30661 Comm: ethtool Kdump: loaded Tainted: GOE    --------- -  - 4.18.0-240.el8.x86_64 #1
[53764.011825] Hardware name: Intel Corporation WilsonCity/WilsonCity, BIOS WLYDCRB1.SYS.0020.P21.2012150710 12/15/2020
[53764.022584] RIP: 0010:ice_vsi_map_rings_to_vectors+0x7e/0x120 [ice]
[53764.029089] Code: 41 0d 0f b7 b7 12 05 00 00 0f b6 d0 44 29 de 44 0f b7 c6 44 01 c2 41 39 d0 7d 2d 4c 8b 47 28 44 0f b7 ce 83 c6 01 4f 8b 04 c8 <49> 89 48 28 4                           c 8b 89 b8 01 00 00 4d 89 08 4c 89 81 b8 01 00 00 44
[53764.048379] RSP: 0018:ff550dd88ea47b20 EFLAGS: 00010206
[53764.053884] RAX: 0000000000000002 RBX: 0000000000000004 RCX: ff385ea42fa4a018
[53764.061301] RDX: 0000000000000006 RSI: 0000000000000005 RDI: ff385e9baeedd018
[53764.068717] RBP: 0000000000000010 R08: 0000000000000000 R09: 0000000000000004
[53764.076133] R10: 0000000000000002 R11: 0000000000000004 R12: 0000000000000000
[53764.083553] R13: 0000000000000000 R14: ff385e658fdd9000 R15: ff385e9baeedd018
[53764.090976] FS:  000014872c5b5740(0000) GS:ff385e847f100000(0000) knlGS:0000000000000000
[53764.099362] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[53764.105409] CR2: 0000000000000028 CR3: 0000000a820fa002 CR4: 0000000000761ee0
[53764.112851] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[53764.120301] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[53764.127747] PKRU: 55555554
[53764.130781] Call Trace:
[53764.133564]  ice_vsi_rebuild+0x611/0x870 [ice]
[53764.138341]  ice_vsi_recfg_qs+0x94/0x100 [ice]
[53764.143116]  ice_set_channels+0x1a8/0x3e0 [ice]
[53764.147975]  ethtool_set_channels+0x14e/0x240
[53764.152667]  dev_ethtool+0xd74/0x2a10
[53764.156665]  ? __mod_lruvec_state+0x44/0x110
[53764.161280]  ? __mod_lruvec_state+0x44/0x110
[53764.165893]  ? page_add_file_rmap+0x15/0x170
[53764.170518]  ? inet_ioctl+0xd1/0x220
[53764.174445]  ? netdev_run_todo+0x5e/0x290
[53764.178808]  dev_ioctl+0xb5/0x550
[53764.182485]  sock_do_ioctl+0xa0/0x140
[53764.186512]  sock_ioctl+0x1a8/0x300
[53764.190367]  ? selinux_file_ioctl+0x161/0x200
[53764.195090]  do_vfs_ioctl+0xa4/0x640
[53764.199035]  ksys_ioctl+0x60/0x90
[53764.202722]  __x64_sys_ioctl+0x16/0x20
[53764.206845]  do_syscall_64+0x5b/0x1a0
[53764.210887]  entry_SYSCALL_64_after_hwframe+0x65/0xca

Fixes: 87324e747fde ("ice: Implement ethtool ops for channels")
Signed-off-by: Anatolii Gerasymenko <anatolii.gerasymenko@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agointel/i40e: delete if NULL check before dev_kfree_skb
Bernard Zhao [Wed, 11 May 2022 06:54:51 +0000 (23:54 -0700)]
intel/i40e: delete if NULL check before dev_kfree_skb

dev_kfree_skb check if the input parameter NULL and do the right
thing, there is no need to check again.
This change is to cleanup the code a bit.

Signed-off-by: Bernard Zhao <zhaojunkui2008@126.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoi40e: Remove unnecessary synchronize_irq() before free_irq()
Minghao Chi [Wed, 18 May 2022 21:31:11 +0000 (14:31 -0700)]
i40e: Remove unnecessary synchronize_irq() before free_irq()

Calling synchronize_irq() right before free_irq() is quite useless. On one
hand the IRQ can easily fire again before free_irq() is entered, on the
other hand free_irq() itself calls synchronize_irq() internally (in a race
condition free way), before any state associated with the IRQ is freed.

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoi40e: Add support for ethtool -s <interface> speed <speed in Mb>
Mateusz Palczewski [Wed, 11 May 2022 10:21:28 +0000 (12:21 +0200)]
i40e: Add support for ethtool -s <interface> speed <speed in Mb>

Add ability to change speed through ethtool -s <interface> speed <speed in Mb>
Driver advertises all link modes that support requested speed.
Autoneg must be set off e.g.:
ethtool -s <interface> autoneg off speed <speed in Mb>.

Add helper function that translate speed in Mb to
enum i40e_aq_link_speed and compare it to supported speeds from
given ethtool_link_ksettings. Add in i40e_set_link_ksettings hold
for requested speed and set copy_ks.base.speed to safe_ks.base.speed
to be sure that user is not changing unsupported setting.
In i40e_speed_to_link_speed compare requested speed with supported
speeds. Set speed to requested speed.

Signed-off-by: Norbert Zulinski <norbertx.zulinski@intel.com>
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
Tested-by: Dave Switzer <david.switzer@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: ethtool: advertise 1000M speeds properly
Anatolii Gerasymenko [Mon, 20 Jun 2022 07:47:05 +0000 (09:47 +0200)]
ice: ethtool: advertise 1000M speeds properly

In current implementation ice_update_phy_type enables all link modes
for selected speed. This approach doesn't work for 1000M speeds,
because both copper (1000baseT) and optical (1000baseX) standards
cannot be enabled at once.

Fix this, by adding the function `ice_set_phy_type_from_speed()`
for 1000M speeds.

Fixes: 48cb27f2fd18 ("ice: Implement handlers for ethtool PHY/link operations")
Signed-off-by: Anatolii Gerasymenko <anatolii.gerasymenko@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: Fix switchdev rules book keeping
Wojciech Drewek [Mon, 23 May 2022 09:41:55 +0000 (11:41 +0200)]
ice: Fix switchdev rules book keeping

Adding two filters with same matching criteria ends up with
one rule in hardware with act = ICE_FWD_TO_VSI_LIST.
In order to remove them properly we have to keep the
information about vsi handle which is used in VSI bitmap
(ice_adv_fltr_mgmt_list_entry::vsi_list_info::vsi_map).

Fixes: 0d08a441fb1a ("ice: ndo_setup_tc implementation for PF")
Reported-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: Wojciech Drewek <wojciech.drewek@intel.com>
Tested-by: Sandeep Penigalapati <sandeep.penigalapati@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: ignore protocol field in GTP offload
Marcin Szycik [Mon, 9 May 2022 19:01:18 +0000 (21:01 +0200)]
ice: ignore protocol field in GTP offload

Commit 34a897758efe ("ice: Add support for inner etype in switchdev")
added the ability to match on inner ethertype. A side effect of that change
is that it is now impossible to add some filters for protocols which do not
contain inner ethtype field. tc requires the protocol field to be specified
when providing certain other options, e.g. src_ip. This is a problem in
case of GTP - when user wants to specify e.g. src_ip, they also need to
specify protocol in tc command (otherwise tc fails with: Illegal "src_ip").
Because GTP is a tunnel, the protocol field is treated as inner protocol.
GTP does not contain inner ethtype field and the filter cannot be added.

To fix this, ignore the ethertype field in case of GTP filters.

Fixes: 9a225f81f540 ("ice: Support GTP-U and GTP-C offload in switchdev")
Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Tested-by: Sandeep Penigalapati <sandeep.penigalapati@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoMerge tag 'efi-urgent-for-v5.19-1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 21 Jun 2022 17:20:11 +0000 (12:20 -0500)]
Merge tag 'efi-urgent-for-v5.19-1' of git://git./linux/kernel/git/efi/efi

Pull EFI fixes from Ard Biesheuvel:

 - remove pointless include of asm/efi.h, which does not exist on ia64

 - fix DXE service marshalling prototype for mixed mode

* tag 'efi-urgent-for-v5.19-1' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi:
  efi/x86: libstub: Fix typo in __efi64_argmap* name
  efi: sysfb_efi: remove unnecessary <asm/efi.h> include

2 years agoMerge tag 'certs-20220621' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells...
Linus Torvalds [Tue, 21 Jun 2022 17:13:53 +0000 (12:13 -0500)]
Merge tag 'certs-20220621' of git://git./linux/kernel/git/dhowells/linux-fs

Pull signature checking selftest from David Howells:
 "The signature checking code, as used by module signing, kexec, etc.,
  is non-FIPS compliant as there is no selftest.

  For a kernel to be FIPS-compliant, signature checking would have to be
  tested before being used, and the box would need to panic if it's not
  available (probably reasonable as simply disabling signature checking
  would prevent you from loading any driver modules).

  Deal with this by adding a minimal test.

  This is split into two patches: the first moves load_certificate_list()
  to the same place as the X.509 code to make it more accessible
  internally; the second adds a selftest"

* tag 'certs-20220621' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs:
  certs: Add FIPS selftests
  certs: Move load_certificate_list() to be with the asymmetric keys code

2 years agoMerge tag 'for-5.19-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Tue, 21 Jun 2022 17:06:04 +0000 (12:06 -0500)]
Merge tag 'for-5.19-rc3-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - print more error messages for invalid mount option values

 - prevent remount with v1 space cache for subpage filesystem

 - fix hang during unmount when block group reclaim task is running

* tag 'for-5.19-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: add error messages to all unrecognized mount options
  btrfs: prevent remounting to v1 space cache for subpage mount
  btrfs: fix hang during unmount when block group reclaim task is running

2 years agoafs: Fix dynamic root getattr
David Howells [Tue, 21 Jun 2022 14:59:57 +0000 (15:59 +0100)]
afs: Fix dynamic root getattr

The recent patch to make afs_getattr consult the server didn't account
for the pseudo-inodes employed by the dynamic root-type afs superblock
not having a volume or a server to access, and thus an oops occurs if
such a directory is stat'd.

Fix this by checking to see if the vnode->volume pointer actually points
anywhere before following it in afs_getattr().

This can be tested by stat'ing a directory in /afs.  It may be
sufficient just to do "ls /afs" and the oops looks something like:

        BUG: kernel NULL pointer dereference, address: 0000000000000020
        ...
        RIP: 0010:afs_getattr+0x8b/0x14b
        ...
        Call Trace:
         <TASK>
         vfs_statx+0x79/0xf5
         vfs_fstatat+0x49/0x62

Fixes: 2aeb8c86d499 ("afs: Fix afs_getattr() to refetch file status if callback break occurred")
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Marc Dionne <marc.dionne@auristor.com>
Tested-by: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/165408450783.1031787.7941404776393751186.stgit@warthog.procyon.org.uk/
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoefi/x86: libstub: Fix typo in __efi64_argmap* name
Evgeniy Baskov [Thu, 9 Jun 2022 11:37:30 +0000 (14:37 +0300)]
efi/x86: libstub: Fix typo in __efi64_argmap* name

The actual name of the DXE services function used
is set_memory_space_attributes(), not set_memory_space_descriptor().

Change EFI mixed mode helper macro name to match the function name.

Fixes: 31f1a0edff78 ("efi/x86: libstub: Make DXE calls mixed mode safe")
Signed-off-by: Evgeniy Baskov <baskov@ispras.ru>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
2 years agoefi: sysfb_efi: remove unnecessary <asm/efi.h> include
Javier Martinez Canillas [Fri, 10 Jun 2022 09:00:05 +0000 (11:00 +0200)]
efi: sysfb_efi: remove unnecessary <asm/efi.h> include

Nothing defined in the header is used by drivers/firmware/efi/sysfb_efi.c
but also, including it can lead to build errors when built on arches that
don't have an asm/efi.h header file.

This can happen for example if a driver that is built when COMPILE_TEST is
enabled selects the SYSFB symbol, e.g. on powerpc with allyesconfig:

drivers/firmware/efi/sysfb_efi.c:29:10: fatal error: asm/efi.h: No such file or directory
   29 | #include <asm/efi.h>
      |          ^~~~~~~~~~~

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
2 years agocerts: Add FIPS selftests
David Howells [Wed, 18 May 2022 16:15:34 +0000 (17:15 +0100)]
certs: Add FIPS selftests

Add some selftests for signature checking when FIPS mode is enabled.  These
need to be done before we start actually using the signature checking for
things and must panic the kernel upon failure.

Note that the tests must not check the blacklist lest this provide a way to
prevent a kernel from booting by installing a hash of a test key in the
appropriate UEFI table.

Reported-by: Simo Sorce <simo@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Simo Sorce <simo@redhat.com>
Reviewed-by: Herbert Xu <herbert@gondor.apana.org.au>
cc: keyrings@vger.kernel.org
cc: linux-crypto@vger.kernel.org
Link: https://lore.kernel.org/r/165515742832.1554877.2073456606206090838.stgit@warthog.procyon.org.uk/
2 years agocerts: Move load_certificate_list() to be with the asymmetric keys code
David Howells [Wed, 18 May 2022 22:48:09 +0000 (23:48 +0100)]
certs: Move load_certificate_list() to be with the asymmetric keys code

Move load_certificate_list(), which loads a series of binary X.509
certificates from a blob and inserts them as keys into a keyring, to be
with the asymmetric keys code that it drives.

This makes it easier to add FIPS selftest code in which we need to load up
a private keyring for the tests to use.

Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Simo Sorce <simo@redhat.com>
Reviewed-by: Herbert Xu <herbert@gondor.apana.org.au>
cc: keyrings@vger.kernel.org
cc: linux-crypto@vger.kernel.org
Link: https://lore.kernel.org/r/165515742145.1554877.13488098107542537203.stgit@warthog.procyon.org.uk/
2 years agoMerge branch 'net-dsa-microchip-common-spi-probe-for-the-ksz-series-switches-part-1'
Paolo Abeni [Tue, 21 Jun 2022 13:26:47 +0000 (15:26 +0200)]
Merge branch 'net-dsa-microchip-common-spi-probe-for-the-ksz-series-switches-part-1'

Arun Ramadoss says:

====================
net: dsa: microchip: common spi probe for the ksz series switches - part 1

This patch series aims to refactor the ksz_switch_register routine to have the
common flow for the ksz series switch. At present ksz8795.c & ksz9477.c have
its own dsa_switch_ops and switch detect functionality.
In ksz_switch_register, ksz_dev_ops is assigned based on the function parameter
passed by the individual ksz8/ksz9477 switch register function. And then switch
detect is performed based on the ksz_dev_ops.detect hook.  This patch modifies
the ksz_switch_register such a way that switch detect is performed first, based
on the chip ksz_dev_ops is assigned to ksz_device structure. It ensures the
common flow for the existing as well as LAN937x switches.
In the next series of patch, it will move ksz_dsa_ops and dsa_switch_ops
from ksz8795.c and ksz9477.c to ksz_common.c and have the common spi
probe all the ksz based switches.

Changes in v1
- Splitted the patch series into two.
- Replaced all occurrence of REG_PORT_STATUS_0 and PORT_FIBER_MODE to
  KSZ8_PORT_STATUS_0 and KSZ8_PORT_FIBER_MODE.
- Separated the tag protocol and phy read/write patch into two.
- Assigned the DSA_TAG_PROTO_NONE as the default value for get_tag_protocol hook.
- Reduced the indentation level by using the if(!dev->dev_ops->mirror_add).
- Added the stp_ctrl_reg as a member in ksz_chip_data and removed the member
  in ksz_dev_ops.
- Removed the r_dyn_mac_table, r_sta_mac_table and w_sta_mac_table from the
  ksz_dev_ops since it is used only in the ksz8795.c.

Changes in RFC v2
- Fixed the compilation issue.
- Reduced the patch set to 15.
====================

Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20220617084255.19376-1-arun.ramadoss@microchip.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: move get_phy_flags & mtu to ksz_common
Arun Ramadoss [Fri, 17 Jun 2022 08:42:55 +0000 (14:12 +0530)]
net: dsa: microchip: move get_phy_flags & mtu to ksz_common

This patch assigns the get_phy_flags & mtu  hook of ksz8795 and ksz9477
in dsa_switch_ops to ksz_common.  For get_phy_flags hooks,checks whether
the chip is ksz8863/kss8793 then it returns error for port1.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: update fdb add/del/dump in ksz_common
Arun Ramadoss [Fri, 17 Jun 2022 08:42:54 +0000 (14:12 +0530)]
net: dsa: microchip: update fdb add/del/dump in ksz_common

This patch makes the dsa_switch_hook for fdbs to use ksz_common.c file.
And from ksz_common, individual switches fdb functions are called using
the dev->dev_ops. And removed the r_dyn_mac_table, r_sta_mac_table and
w_sta_mac_table from ksz_dev_ops as it is used only in ksz8795.c

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: update the ksz_port_mdb_add/del
Arun Ramadoss [Fri, 17 Jun 2022 08:42:53 +0000 (14:12 +0530)]
net: dsa: microchip: update the ksz_port_mdb_add/del

ksz_mdb_add/del in ksz_common.c is specific for the ksz8795.c file. The
ksz9477 has its separate ksz9477_port_mdb_add/del functions.  This patch
moves the ksz8795 specific mdb functionality from ksz_common to ksz8795.
And this dsa_switch_ops hooks for ksz8795/ksz9477 are invoked through
the ksz_port_mdb_add/del.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: update the ksz_phylink_get_caps
Arun Ramadoss [Fri, 17 Jun 2022 08:42:52 +0000 (14:12 +0530)]
net: dsa: microchip: update the ksz_phylink_get_caps

This patch assigns the phylink_get_caps in ksz8795 and ksz9477 to
ksz_phylink_get_caps. And update their mac_capabilities in the
respective ksz_dev_ops.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: get P_STP_CTRL in ksz_port_stp_state by ksz_dev_ops
Arun Ramadoss [Fri, 17 Jun 2022 08:42:51 +0000 (14:12 +0530)]
net: dsa: microchip: get P_STP_CTRL in ksz_port_stp_state by ksz_dev_ops

At present, P_STP_CTRL register value is passed as parameter to
ksz_port_stp_state from the individual dsa_switch_ops hooks. This patch
update the function to retrieve the register value through the
ksz_chip_data member.
And add the static to ksz_update_port_member since it is not called
outside the ksz_common.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: move the port mirror to ksz_common
Arun Ramadoss [Fri, 17 Jun 2022 08:42:50 +0000 (14:12 +0530)]
net: dsa: microchip: move the port mirror to ksz_common

This patch updates the common port mirror add/del dsa_switch_ops in
ksz_common.c. The individual switches implementation is executed based
on the ksz_dev_ops function pointers.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: move vlan functionality to ksz_common
Arun Ramadoss [Fri, 17 Jun 2022 08:42:49 +0000 (14:12 +0530)]
net: dsa: microchip: move vlan functionality to ksz_common

This patch moves the vlan dsa_switch_ops such as vlan_add, vlan_del and
vlan_filtering from the individual files ksz8795.c, ksz9477.c to
ksz_common.c file.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: ksz9477: use ksz_read_phy16 & ksz_write_phy16
Arun Ramadoss [Fri, 17 Jun 2022 08:42:48 +0000 (14:12 +0530)]
net: dsa: microchip: ksz9477: use ksz_read_phy16 & ksz_write_phy16

ksz8795 and ksz9477 implementation on phy read/write hooks are
different. This patch modifies the ksz9477 implementation same as
ksz8795 by updating the ksz9477_dev_ops structure.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: move tag_protocol to ksz_common
Arun Ramadoss [Fri, 17 Jun 2022 08:42:47 +0000 (14:12 +0530)]
net: dsa: microchip: move tag_protocol to ksz_common

This patch move the dsa hook get_tag_protocol to ksz_common file. And
the tag_protocol is returned based on the dev->chip_id.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: move switch chip_id detection to ksz_common
Arun Ramadoss [Fri, 17 Jun 2022 08:42:46 +0000 (14:12 +0530)]
net: dsa: microchip: move switch chip_id detection to ksz_common

KSZ87xx and KSZ88xx have chip_id representation at reg location 0. And
KSZ9477 compatible switch and LAN937x switch have same chip_id detection
at location 0x01 and 0x02. To have the common switch detect
functionality for ksz switches, ksz_switch_detect function is
introduced.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: dsa: microchip: ksz9477: cleanup the ksz9477_switch_detect
Arun Ramadoss [Fri, 17 Jun 2022 08:42:45 +0000 (14:12 +0530)]
net: dsa: microchip: ksz9477: cleanup the ksz9477_switch_detect

The ksz9477_switch_detect performs the detecting the chip id from the
location 0x00 and also check gigabit compatibility check & number of
ports based on the register global_options0. To prepare the common ksz
switch detect function, routine other than chip id read is moved to
ksz9477_switch_init.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoALSA: hda: Fix discovery of i915 graphics PCI device
Takashi Iwai [Tue, 21 Jun 2022 12:00:44 +0000 (14:00 +0200)]
ALSA: hda: Fix discovery of i915 graphics PCI device

It's been reported that the recent fix for skipping the
component-binding with D-GPU caused a regression on some systems; it
resulted in the completely missing component binding with i915 GPU.

The problem was the use of pci_get_class() function.  It matches with
the full PCI class bits, while we want to match only partially the PCI
base class bits.  So, when a system has an i915 graphics device with
the PCI class 0380, it won't hit because we're looking for only the
PCI class 0300.

This patch fixes i915_gfx_present() to look up each PCI device and
match with PCI base class explicitly instead of pci_get_class().

Fixes: c9db8a30d9f0 ("ALSA: hda/i915 - skip acomp init if no matching display")
Reviewed-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Tested-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Cc: <stable@vger.kernel.org>
Link: https://bugzilla.opensuse.org/show_bug.cgi?id=1200611
Link: https://lore.kernel.org/r/87bkunztec.wl-tiwai@suse.de
Link: https://lore.kernel.org/r/20220621120044.11573-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
2 years agonfp: compose firmware file name with new hwinfo "nffw.partno"
Yu Xiao [Mon, 20 Jun 2022 10:39:12 +0000 (12:39 +0200)]
nfp: compose firmware file name with new hwinfo "nffw.partno"

During initialization of the NFP driver, a file name for loading
application firmware is composed using the NIC's AMDA information and port
type (count and speed). E.g.: "nic_AMDA0145-1012_2x10.nffw".

In practice there may be many variants for each NIC type, and many of the
variants relate to assembly components which do not concern the driver and
application firmware implementation. Yet the current scheme leads to a
different application firmware file name for each variant, because they
have different AMDA information.

To reduce proliferation of content-duplicated application firmware images
or symlinks, the NIC's management firmware will only expose differences
between variants that need different application firmware via a newly
introduced hwinfo, "nffw.partno".

Use of the existing hwinfo, "assembly.partno", is maintained in order to
support for NICs with management firmware that does not expose
"nffw.partno".

Signed-off-by: Yu Xiao <yu.xiao@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20220620103912.46164-1-simon.horman@corigine.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoraw: complete rcu conversion
Eric Dumazet [Mon, 20 Jun 2022 10:05:09 +0000 (03:05 -0700)]
raw: complete rcu conversion

raw_diag_dump() can use rcu_read_lock() instead of read_lock()

Now the hashinfo lock is only used from process context,
in write mode only, we can convert it to a spinlock,
and we do not need to block BH anymore.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220620100509.3493504-1-eric.dumazet@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: warn if mac header was not set
Eric Dumazet [Mon, 20 Jun 2022 09:30:17 +0000 (02:30 -0700)]
net: warn if mac header was not set

Make sure skb_mac_header(), skb_mac_offset() and skb_mac_header_len() uses
are not fooled if the mac header has not been set.

These checks are enabled if CONFIG_DEBUG_NET=y

This commit will likely expose existing bugs in linux networking stacks.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220620093017.3366713-1-eric.dumazet@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonetfilter: nf_dup_netdev: add and use recursion counter
Florian Westphal [Mon, 20 Jun 2022 14:17:31 +0000 (16:17 +0200)]
netfilter: nf_dup_netdev: add and use recursion counter

Now that the egress function can be called from egress hook, we need
to avoid recursive calls into the nf_tables traverser, else crash.

Fixes: f87b9464d152 ("netfilter: nft_fwd_netdev: Support egress hook")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agonetfilter: nf_dup_netdev: do not push mac header a second time
Florian Westphal [Mon, 20 Jun 2022 14:17:30 +0000 (16:17 +0200)]
netfilter: nf_dup_netdev: do not push mac header a second time

Eric reports skb_under_panic when using dup/fwd via bond+egress hook.
Before pushing mac header, we should make sure that we're called from
ingress to put back what was pulled earlier.

In egress case, the MAC header is already there; we should leave skb
alone.

While at it be more careful here: skb might have been altered and
headroom reduced, so add a skb_cow() before so that headroom is
increased if necessary.

nf_do_netdev_egress() assumes skb ownership (it normally ends with
a call to dev_queue_xmit), so we must free the packet on error.

Fixes: f87b9464d152 ("netfilter: nft_fwd_netdev: Support egress hook")
Reported-by: Eric Garver <eric@garver.life>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agoselftests: netfilter: correct PKTGEN_SCRIPT_PATHS in nft_concat_range.sh
Jie2x Zhou [Thu, 16 Jun 2022 07:40:46 +0000 (15:40 +0800)]
selftests: netfilter: correct PKTGEN_SCRIPT_PATHS in nft_concat_range.sh

Before change:
make -C netfilter
 TEST: performance
   net,port                                                      [SKIP]
   perf not supported
   port,net                                                      [SKIP]
   perf not supported
   net6,port                                                     [SKIP]
   perf not supported
   port,proto                                                    [SKIP]
   perf not supported
   net6,port,mac                                                 [SKIP]
   perf not supported
   net6,port,mac,proto                                           [SKIP]
   perf not supported
   net,mac                                                       [SKIP]
   perf not supported

After change:
   net,mac                                                       [ OK ]
     baseline (drop from netdev hook):               2061098pps
     baseline hash (non-ranged entries):             1606741pps
     baseline rbtree (match on first field only):    1191607pps
     set with  1000 full, ranged entries:            1639119pps
ok 8 selftests: netfilter: nft_concat_range.sh

Fixes: 611973c1e06f ("selftests: netfilter: Introduce tests for sets with range concatenation")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Jie2x Zhou <jie2x.zhou@intel.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Mon, 20 Jun 2022 14:35:04 +0000 (09:35 -0500)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Eight fixes, all in drivers (ufs, scsi_debug, storvsc, iscsi, ibmvfc).

  Apart from the ufs command clearing updates, these are mostly minor
  and obvious fixes"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: ibmvfc: Store vhost pointer during subcrq allocation
  scsi: ibmvfc: Allocate/free queue resource only during probe/remove
  scsi: storvsc: Correct reporting of Hyper-V I/O size limits
  scsi: ufs: Fix a race between the interrupt handler and the reset handler
  scsi: ufs: Support clearing multiple commands at once
  scsi: ufs: Simplify ufshcd_clear_cmd()
  scsi: iscsi: Exclude zero from the endpoint ID range
  scsi: scsi_debug: Fix zone transition to full condition

2 years agoMerge tag 'perf-tools-fixes-for-v5.19-2022-06-19' of git://git.kernel.org/pub/scm...
Linus Torvalds [Mon, 20 Jun 2022 14:31:03 +0000 (09:31 -0500)]
Merge tag 'perf-tools-fixes-for-v5.19-2022-06-19' of git://git./linux/kernel/git/acme/linux

Pull perf tool fixes from Arnaldo Carvalho de Melo:

 - Don't set data source if it's not a memory operation in ARM SPE
   (Statistical Profiling Extensions).

 - Fix handling of exponent floating point values in perf stat
   expressions.

 - Don't leak fd on failure on libperf open.

 - Fix 'perf test' CPU topology test for PPC guest systems.

 - Fix undefined behaviour on breakpoint account 'perf test' entry.

 - Record only user callchains on the "Check ARM64 callgraphs are
   complete in FP mode" 'perf test' entry.

 - Fix "perf stat CSV output linter" test on s390.

 - Sync batch of kernel headers with tools/perf/.

* tag 'perf-tools-fixes-for-v5.19-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  tools headers UAPI: Sync linux/prctl.h with the kernel sources
  perf metrics: Ensure at least 1 id per metric
  tools headers arm64: Sync arm64's cputype.h with the kernel sources
  tools headers UAPI: Sync x86's asm/kvm.h with the kernel sources
  perf arm-spe: Don't set data source if it's not a memory operation
  perf expr: Allow exponents on floating point values
  perf test topology: Use !strncmp(right platform) to fix guest PPC comparision check
  perf test: Record only user callchains on the "Check Arm64 callgraphs are complete in fp mode" test
  perf beauty: Update copy of linux/socket.h with the kernel sources
  perf test: Fix variable length array undefined behavior in bp_account
  libperf evsel: Open shouldn't leak fd on failure
  perf test: Fix "perf stat CSV output linter" test on s390
  perf unwind: Fix uninitialized variable

2 years agoMerge tag 'slab-for-5.19-fixup' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 20 Jun 2022 14:28:51 +0000 (09:28 -0500)]
Merge tag 'slab-for-5.19-fixup' of git://git./linux/kernel/git/vbabka/slab

Pull slab fixes from Vlastimil Babka:

 - A slub fix for PREEMPT_RT locking semantics from Sebastian.

 - A slub fix for state corruption due to a possible race scenario from
   Jann.

* tag 'slab-for-5.19-fixup' of git://git.kernel.org/pub/scm/linux/kernel/git/vbabka/slab:
  mm/slub: add missing TID updates on slab deactivation
  mm/slub: Move the stackdepot related allocation out of IRQ-off section.

2 years agoudmabuf: add back sanity check
Gerd Hoffmann [Mon, 20 Jun 2022 07:15:47 +0000 (09:15 +0200)]
udmabuf: add back sanity check

Check vm_fault->pgoff before using it.  When we removed the warning, we
also removed the check.

Fixes: 7b26e4e2119d ("udmabuf: drop WARN_ON() check.")
Reported-by: zdi-disclosures@trendmicro.com
Suggested-by: Linus Torvalds <torvalds@linuxfoundation.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoALSA: hda/via: Fix missing beep setup
Takashi Iwai [Mon, 20 Jun 2022 10:40:08 +0000 (12:40 +0200)]
ALSA: hda/via: Fix missing beep setup

Like the previous fix for Conexant codec, the beep_nid has to be set
up before calling snd_hda_gen_parse_auto_config(); otherwise it'd miss
the path setup.

Fix the call order for addressing the missing beep setup.

Fixes: 0e8f9862493a ("ALSA: hda/via - Simplify control management")
Cc: <stable@vger.kernel.org>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216152
Link: https://lore.kernel.org/r/20220620104008.1994-2-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
2 years agoALSA: hda/conexant: Fix missing beep setup
Takashi Iwai [Mon, 20 Jun 2022 10:40:07 +0000 (12:40 +0200)]
ALSA: hda/conexant: Fix missing beep setup

Currently the Conexant codec driver sets up the beep NID after calling
snd_hda_gen_parse_auto_config().  It turned out that this results in
the insufficient setup for the beep control, as the generic parser
handles the fake path in snd_hda_gen_parse_auto_config() only if the
beep_nid is set up beforehand.

For dealing with the beep widget properly, call cx_auto_parse_beep()
before snd_hda_gen_parse_auto_config() call.

Fixes: 51e19ca5f755 ("ALSA: hda/conexant - Clean up beep code")
Cc: <stable@vger.kernel.org>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216152
Link: https://lore.kernel.org/r/20220620104008.1994-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
2 years agoMerge branch 'mlxsw-unified-bridge-conversion-part-1'
David S. Miller [Mon, 20 Jun 2022 09:03:34 +0000 (10:03 +0100)]
Merge branch 'mlxsw-unified-bridge-conversion-part-1'

Ido Schimmel says:

====================
mlxsw: Unified bridge conversion - part 1/6

This set starts converting mlxsw to the unified bridge model and mainly
adds new device registers and extends existing ones that will be used in
follow-up patchsets.

High-level summary
==================

The unified bridge model is a new way of managing low-level device
objects such as filtering identifiers (FIDs). The conversion moves a lot
of logic out of the device's firmware towards the driver, but its main
selling point is that it allows to overcome various scalability issues
related to the amount of entries that need to be programmed to the
device.

The only (intended) user visible changes of the conversion are
improvement in resource utilization and ability to support more router
interfaces (RIFs) in Spectrum-{2,3}.

Details
=======

Commit 50853808ff4a ("Merge branch
'mlxsw-Prepare-for-VLAN-aware-bridge-w-VxLAN'") converted mlxsw to
emulate 802.1Q FIDs (represent VLANs in a VLAN-aware bridge) using
802.1D FIDs (represent VLAN-unaware bridges). This was necessary because
at that time VNI could not be assigned to 802.1Q FIDs, which effectively
meant that mlxsw could not support VXLAN with VLAN-aware bridges.

The downside of this approach is that multiple {Port,VID}->FID entries
are required in order to classify incoming traffic to a FID, as opposed
to a single VID->FID entry that can be used with actual 802.1Q FIDs.

For example, if 10 ports are members in the same VLAN-aware bridge and
the same 100 VLANs are configured on each port, then only 100 VID->FID
entries are required with 802.1Q FIDs, whereas 1000 {Port,VID}->FID
entries are required with emulated 802.1Q FIDs.

The above limitation is the result of various assumptions that were made
in the design of the API that was exposed to software. In the unified
bridge model the API is much more "raw" and therefore avoids these
assumptions, allowing software to configure the device in a more
efficient manner.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Add support for VLAN RIF as part of RITR register
Amit Cohen [Sun, 19 Jun 2022 10:29:21 +0000 (13:29 +0300)]
mlxsw: reg: Add support for VLAN RIF as part of RITR register

Router interfaces (RIFs) constructed on top of VLAN-aware bridges are of
"VLAN" type, whereas RIFs constructed on top of VLAN-unaware bridges of
"FID" type.

In other words, the RIF type is derived from the underlying FID type.
VLAN RIFs are used on top of 802.1Q FIDs, whereas FID RIFs are used on
top of 802.1D FIDs.

Currently 802.1Q FIDs are emulated using 802.1D FIDs, and therefore VLAN
RIFs are emulated using FID RIFs.

As part of converting the driver to use unified bridge, 802.1Q FIDs and
VLAN RIFs will be used.

Add the relevant fields to RITR register, add pack() function for VLAN
RIF and rename one field to fit the internal name.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>