Vadim Fedorenko ays:

====================
Fix PMTU for ESP-in-UDP encapsulation

Bug 213669 uncovered regression in PMTU discovery for UDP-encapsulated
routes and some incorrect usage in udp tunnel fields. This series fixes
problems and also adds such case for selftests

v3:
 - update checking logic to account SCTP use case
v2:
 - remove refactor code that was in first patch
 - move checking logic to __udp{4,6}_lib_err_encap
 - add more tests, especially routed configuration
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The case of ESP in UDP encapsulation was not covered before. Add
cases of local changes of MTU and difference on routed path.
Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit d26796ae ("udp: check udp sock encap_type in __udp_lib_err")
added checks for encapsulated sockets but it broke cases when there is
no implementation of encap_err_lookup for encapsulation, i.e. ESP in
UDP encapsulation. Fix it by calling encap_err_lookup only if socket
implements this method otherwise treat it as legal socket.

Fixes: d26796ae ("udp: check udp sock encap_type in __udp_lib_err")
Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

syzbot reported a call trace:

  BUG: KASAN: use-after-free in sctp_auth_shkey_hold+0x22/0xa0 net/sctp/auth.c:112
  Call Trace:
   sctp_auth_shkey_hold+0x22/0xa0 net/sctp/auth.c:112
   sctp_set_owner_w net/sctp/socket.c:131 [inline]
   sctp_sendmsg_to_asoc+0x152e/0x2180 net/sctp/socket.c:1865
   sctp_sendmsg+0x103b/0x1d30 net/sctp/socket.c:2027
   inet_sendmsg+0x99/0xe0 net/ipv4/af_inet.c:821
   sock_sendmsg_nosec net/socket.c:703 [inline]
   sock_sendmsg+0xcf/0x120 net/socket.c:723

This is an use-after-free issue caused by not updating asoc->shkey after
it was replaced in the key list asoc->endpoint_shared_keys, and the old
key was freed.

This patch is to fix by also updating active_key for asoc when old key is
being replaced with a new one. Note that this issue doesn't exist in
sctp_auth_del_key_id(), as it's not allowed to delete the active_key
from the asoc.

Fixes: 1b1e0bc9 ("sctp: add refcnt support for sh_key")
Reported-by: syzbot+b774577370208727d12b@syzkaller.appspotmail.com
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When registering the MDIO bus for a r8169 device, we use the PCI
bus/device specifier as a (seemingly) unique device identifier.
However the very same BDF number can be used on another PCI segment,
which makes the driver fail probing:

[ 27.544136] r8169 0002:07:00.0: enabling device (0000 -> 0003)
[ 27.559734] sysfs: cannot create duplicate filename '/class/mdio_bus/r8169-700'
....
[ 27.684858] libphy: mii_bus r8169-700 failed to register
[ 27.695602] r8169: probe of 0002:07:00.0 failed with error -22

Add the segment number to the device name to make it more unique.

This fixes operation on ARM N1SDP boards, with two boards connected
together to form an SMP system, and all on-board devices showing up
twice, just on different PCI segments. A similar issue would occur on
large systems with many PCI slots and multiple RTL8169 NICs.

Fixes: f1e911d5 ("r8169: add basic phylib support")
Signed-off-by: Sayanta Pattanayak <sayanta.pattanayak@arm.com>
[Andre: expand commit message, use pci_domain_nr()]
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Acked-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When receiving a packet with multiple fragments, hardware may still
touch the first fragment until the entire packet has been received. The
driver therefore keeps the first fragment mapped for DMA until end of
packet has been asserted, and delays its dma_sync call until then.

The driver tries to fit multiple receive buffers on one page. When using
3K receive buffers (e.g. using Jumbo frames and legacy-rx is turned
off/build_skb is being used) on an architecture with 4K pages, the
driver allocates an order 1 compound page and uses one page per receive
buffer. To determine the correct offset for a delayed DMA sync of the
first fragment of a multi-fragment packet, the driver then cannot just
use PAGE_MASK on the DMA address but has to construct a mask based on
the actual size of the backing page.

Using PAGE_MASK in the 3K RX buffer/4K page architecture configuration
will always sync the first page of a compound page. With the SWIOTLB
enabled this can lead to corrupted packets (zeroed out first fragment,
re-used garbage from another packet) and various consequences, such as
slow/stalling data transfers and connection resets. For example, testing
on a link with MTU exceeding 3058 bytes on a host with SWIOTLB enabled
(e.g. "iommu=soft swiotlb=262144,force") TCP transfers quickly fizzle
out without this patch.

Cc: stable@vger.kernel.org
Fixes: 0c5661ec ("ixgbe: fix crash in build_skb Rx code path")
Signed-off-by: Markus Boehme <markubo@amazon.com>
Tested-by: Tony Brelinski <tonyx.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This reverts commit 6206b798.

That patch added additional spin_{un}lock_bh(), which was harmless
but pointless. The orginal code path has guaranteed the pair of
spin_{un}lock_bh().

We'd better revert it before we find the exact root cause of the
bug_on mentioned in that patch.

Fixes: 6206b798 ("qed: fix possible unpaired spin_{un}lock_bh in _qed_mcp_cmd_and_union()")
Cc: David S. Miller <davem@davemloft.net>
Cc: Prabhakar Kushwaha <pkushwaha@marvell.com>
Signed-off-by: Jia He <justin.he@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

While running the self-tests on a KASAN enabled kernel, I observed a
slab-out-of-bounds splat very similar to the one reported in
commit 821bbf79 ("ipv6: Fix KASAN: slab-out-of-bounds Read in
 fib6_nh_flush_exceptions").

We additionally need to take care of fib6_metrics initialization
failure when the caller provides an nh.

The fix is similar, explicitly free the route instead of calling
fib6_info_release on a half-initialized object.

Fixes: f88d8ea6 ("ipv6: Plumb support for nexthop object in a fib6_info")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Set SUPPORTED_FIBRE to mac_dev->if_support. It allows proper usage of
PHYs with optical/fiber support.
Signed-off-by: Maxim Kochetkov <fido_max@inbox.ru>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently tcf_skbmod_act() assumes that packets use Ethernet as their L2
protocol, which is not always the case.  As an example, for CAN devices:

	$ ip link add dev vcan0 type vcan
	$ ip link set up vcan0
	$ tc qdisc add dev vcan0 root handle 1: htb
	$ tc filter add dev vcan0 parent 1: protocol ip prio 10 \
		matchall action skbmod swap mac

Doing the above silently corrupts all the packets.  Do not perform skbmod
actions for non-Ethernet packets.

Fixes: 86da71b5 ("net_sched: Introduce skbmod action")
Reviewed-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fixes my earlier patch which broke vlan unaware bridges.

The IVL bit now only gets set for vid's larger than 1.

Fixes: 11d8d98c ("mt7530 fix mt7530_fdb_write vid missing ivl bit")
Signed-off-by: Eric Woudstra <ericwouds@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Subbaraya Sundeep says:

====================
octeontx2-af: Introduce DMAC based switching

With this patch set packets can be switched between
all CGX mapped PFs and VFs in the system based on
the DMAC addresses. To implement this:
AF allocates high priority rules from top entry(0) in MCAM.
Rules are allocated for all the CGX mapped PFs and VFs though
they are not active and with no NIXLFs attached.
Rules for a PF/VF will be enabled only after they are brought up.
Two rules one for TX and one for RX are allocated for each PF/VF.

A packet sent from a PF/VF with a destination mac of another
PF/VF will be hit by TX rule and sent to LBK channel 63. The
same returned packet will be hit by RX rule whose action is
to forward packet to PF/VF with that destination mac.

Implementation of this for 98xx is tricky since there are
two NIX blocks and till now a PF/VF can install rule for
an NIX0/1 interface only if it is mapped to corresponding NIX0/1 block.
Hence Tx rules are modified such that TX interface in MCAM
entry can be either NIX0-TX or NIX1-TX.

Testing:

1. Create two VFs over PF1(on NIX0) and assign two VFs to two VMs
2. Assign ip addresses to two VFs in VMs and PF2(on NIX1) in host.
3. Assign static arp entries in two VMs and PF2.
4. Ping between VMs and host PF2.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Guangbin Huang says:

====================
net: hns3: fixes for -net

This series includes some bugfixes for the HNS3 ethernet driver.
====================

Link: https://lore.kernel.org/r/1626685988-25869-1-git-send-email-huangguangbin2@huawei.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Currently, VF doesn't enable rx VLAN offload when initializating,
and PF does it for VFs. If user disable the rx VLAN offload for
VF with ethtool -K, and reload the VF driver, it may cause the
rx VLAN offload state being inconsistent between hardware and
software.

Fixes it by enabling rx VLAN offload when VF initializing.

Fixes: e2cb1dec ("net: hns3: Add HNS3 VF HCL(Hardware Compatibility Layer) Support")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

For hardware limitation, port VLAN filter is port level, and
effective for all the functions of the port. So if not support
port VLAN bypass, it's necessary to disable the port VLAN filter,
in order to support function level VLAN filter control.

Fixes: 2ba30662 ("net: hns3: add support for modify VLAN filter state")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

When VF need response from PF, VF will wait (1us - 1s) to receive
the response, or it will wait timeout and the VF action fails.
If VF do not receive response in 1st action because timeout,
the 2nd action may receive response for the 1st action, and get
incorrect response data.VF must reciveve the right response from
PF,or it will cause unexpected error.

This patch adds match_id to check mailbox response from PF to VF,
to make sure VF get the right response:
1. The message sent from VF was labelled with match_id which was a
unique 16-bit non-zero value.
2. The response sent from PF will label with match_id which got from
the request.
3. The VF uses the match_id to match request and response message.

This scheme depends on PF driver supports match_id, if PF driver doesn't
support then VF will uses the original scheme.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Currently, the mailbox synchronous communication between VF and PF use
the following fields to maintain communication:
1. Origin_mbx_msg which was combined by message code and subcode, used
to match request and response.
2. Received_resp which means whether received response.

There may possible mismatches of the following situation:
1. VF sends message A with code=1 subcode=1.
2. PF was blocked about 500ms when processing the message A.
3. VF will detect message A timeout because it can't get the response
within 500ms.
4. VF sends message B with code=1 subcode=1 which equal message A.
5. PF processes the first message A and send the response message to
VF.
6. VF will identify the response matched the message B because the
code/subcode is the same. This will lead to mismatch of request and
response.

To fix the above bug, we use the following scheme:
1. The message sent from VF was labelled with match_id which was a
unique 16-bit non-zero value.
2. The response sent from PF will label with match_id which got from
the request.
3. The VF uses the match_id to match request and response message.

As for PF driver, it only needs to copy the match_id from request to
response.

Fixes: dde1a86e ("net: hns3: Add mailbox support to PF driver")
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

This simple script:

ip link add br0 type bridge
ip link set swp2 master br0
ip link set br0 address 00:01:02:03:04:05
ip link del br0

produces this result on a DSA switch:

[  421.306399] br0: port 1(swp2) entered blocking state
[  421.311445] br0: port 1(swp2) entered disabled state
[  421.472553] device swp2 entered promiscuous mode
[  421.488986] device swp2 left promiscuous mode
[  421.493508] br0: port 1(swp2) entered disabled state
[  421.886107] sja1105 spi0.1: port 1 failed to delete 00:01:02:03:04:05 vid 1 from fdb: -ENOENT
[  421.894374] sja1105 spi0.1: port 1 failed to delete 00:01:02:03:04:05 vid 0 from fdb: -ENOENT
[  421.943982] br0: port 1(swp2) entered blocking state
[  421.949030] br0: port 1(swp2) entered disabled state
[  422.112504] device swp2 entered promiscuous mode

A very simplified view of what happens is:

(1) the bridge port is created, and the bridge device inherits its MAC
    address

(2) when joining, the bridge port (DSA) requests a replay of the
    addition of all FDB entries towards this bridge port and towards the
    bridge device itself. In fact, DSA calls br_fdb_replay() twice:

	br_fdb_replay(br, brport_dev);
	br_fdb_replay(br, br);

    DSA uses reference counting for the FDB entries. So the MAC address
    of the bridge is simply kept with refcount 2. When the bridge port
    leaves under normal circumstances, everything cancels out since the
    replay of the FDB entry deletion is also done twice per VLAN.

(3) when the bridge MAC address changes, switchdev is notified of the
    deletion of the old address and of the insertion of the new one.
    But the old address does not really go away, since it had refcount
    2, and the new address is added "only" with refcount 1.

(4) when the bridge port leaves now, it will replay a deletion of the
    FDB entries pointing towards the bridge twice. Then DSA will
    complain that it can't delete something that no longer exists.

It is clear that the problem is that the FDB entries towards the bridge
are replayed too many times, so let's fix that problem.

Fixes: 63c51453 ("net: dsa: replay the local bridge FDB entries pointing to the bridge dev too")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20210719093916.4099032-1-vladimir.oltean@nxp.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Update maintainers for MediaTek switch driver with Deng Qingfang who has
contributed many high-quality patches (interrupt, VLAN, GPIO, and etc.)
and will help maintenance.
Signed-off-by: Landen Chao <landen.chao@mediatek.com>
Signed-off-by: DENG Qingfang <dqfext@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/49e1aa8aac58dcbf1b5e036d09b3fa3bbb1d94d0.1626751861.git.landen.chao@mediatek.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

After cited commit, sysctl_tcp_fastopen_blackhole_timeout is no longer
a global variable.

Fixes: 3733be14 ("ipv4: Namespaceify tcp_fastopen_blackhole_timeout knob")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Wei Wang <weiwan@google.com>
Link: https://lore.kernel.org/r/20210719092028.3016745-1-eric.dumazet@gmail.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

skb_set_owner_w() should set sk not to old skb but to new nskb.

Fixes: 5796015f ("ipv6: allocate enough headroom in ip6_finish_output2()")
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Link: https://lore.kernel.org/r/70c0744f-89ae-1869-7e3e-4fa292158f4b@virtuozzo.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

As of now any communication between CGXs PFs and
their VFs within the system is possible only by
external switches sending packets back to the
system. This patch adds internal switching support.
Broadcast packet replication is not covered here.
RVU admin function (AF) maintains MAC addresses
of all interfaces in the system. When switching is
enabled, MCAM entries are allocated to install rules
such that packets with DMAC matching any of the
internal interface MAC addresses is punted back
into the system via the loopback channel.
On the receive side the default unicast rules
are modified to not check for ingress channel.
So any packet with matching DMAC irrespective of
which interface it is coming from will be forwarded
to the respective PF/VF interface.
The transmit side rules and default unicast rules
are updated if user changes MAC address of an interface.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

AF till now only manages the allocation and freeing of
MCAM rules for other PF/VFs in system. To implement
L2 switching between all CGX mapped PF and VFs, AF
requires MCAM entries for DMAC rules for each PF and VF.
This patch modifies AF driver such that AF can also
allocate MCAM rules and install rules for other
PFs and VFs. All the checks like channel verification
for RX rules and PF_FUNC verification for TX rules are
relaxed in case AF is allocating or installing rules.
Also all the entry and counter to owner mappings are
set to NPC_MCAM_INVALID_MAP when they are free indicating
those are not allocated to AF nor PF/VFs.
This patch also ensures that AF allocated and installed
entries are displayed in debugfs.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For enabling VF-VF switching the packets egressing
out of CGX mapped VFs needed to be sent to LBK
so that same packets are received back to the system.
But the LBK link also needs to be enabled in addition
to a VF's mapped CGX_LMAC link otherwise hardware
raises send error interrupt indicating selected LBK
link is not enabled in NIX_AF_TL3_TL2X_LINKX_CFG register.
Hence this patch enables all LBK links in
TL3_TL2_LINKX_CFG registers.
Also to enable packet flow between PFs/VFs of NIX0
to PFs/VFs of NIX1(in 98xx silicon) the NPC TX DMAC
rules has to be installed such that rules must be hit
for any TX interface i.e., NIX0-TX or NIX1-TX provided
DMAC match creteria is met. Hence this patch changes the
behavior such that MCAM is programmed to match with any
NIX0/1-TX interface for TX rules.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tfo_active_disable_stamp is read and written locklessly.
We need to annotate these accesses appropriately.

Then, we need to perform the atomic_inc(tfo_active_disable_times)
after the timestamp has been updated, and thus add barriers
to make sure tcp_fastopen_active_should_disable() wont read
a stale timestamp.

Fixes: cf1ef3f0 ("net/tcp_fastopen: Disable active side TFO in certain scenarios")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Wei Wang <weiwan@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Wei Wang <weiwan@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Joakim Zhang says:

====================
dt-bindings: net: dwmac-imx: convert

This patch set intends to convert imx dwmac binding to schema, and fixes
found by dt_binding_check and dtbs_check.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch changs interrupt order which found by dtbs_check.

$ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- dtbs_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/net/nxp,dwmac-imx.yaml
arch/arm64/boot/dts/freescale/imx8mp-evk.dt.yaml: ethernet@30bf0000: interrupt-names:0: 'macirq' was expected
arch/arm64/boot/dts/freescale/imx8mp-evk.dt.yaml: ethernet@30bf0000: interrupt-names:1: 'eth_wake_irq' was expected

According to Documentation/devicetree/bindings/net/snps,dwmac.yaml, we
should list interrupt in it's order.
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In order to automate the verification of DT nodes covert imx-dwmac to
nxp,dwmac-imx.yaml, and pass below checking.

$ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- dt_binding_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/net/nxp,dwmac-imx.yaml
$ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- dtbs_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/net/nxp,dwmac-imx.yaml
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add missing DWMAC IP version in snps,dwmac.yaml which found by below
command, as NXP i.MX8 families support SNPS DWMAC 5.10a IP.

$ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- dt_binding_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/net/nxp,dwmac-imx.yaml
Documentation/devicetree/bindings/net/nxp,dwmac-imx.example.dt.yaml:
ethernet@30bf0000: compatible: None of ['nxp,imx8mp-dwmac-eqos', 'snps,dwmac-5.10a'] are valid under the given schema
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Building on ARCH=arc causes a "redefined" warning, so rename this
driver's CACHE_LINE_MASK to avoid the warning.

../drivers/net/ethernet/hisilicon/hip04_eth.c:134: warning: "CACHE_LINE_MASK" redefined
  134 | #define CACHE_LINE_MASK   0x3F
In file included from ../include/linux/cache.h:6,
                 from ../include/linux/printk.h:9,
                 from ../include/linux/kernel.h:19,
                 from ../include/linux/list.h:9,
                 from ../include/linux/module.h:12,
                 from ../drivers/net/ethernet/hisilicon/hip04_eth.c:7:
../arch/arc/include/asm/cache.h:17: note: this is the location of the previous definition
   17 | #define CACHE_LINE_MASK  (~(L1_CACHE_BYTES - 1))

Fixes: d413779c ("net: hisilicon: Add an tx_desc to adapt HI13X1_GMAC")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Jiangfeng Xiao <xiaojiangfeng@huawei.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Michael Chan says:

====================
bnxt_en: Bug fixes

Most of the fixes in this series have to do with error recovery.  They
include error path handling when the error recovery has to abort, and
the rediscovery of capabilities (PTP and RoCE) after firmware reset
that may result in capability changes.

Two other fixes are to reject invalid ETS settings and to validate
VLAN protocol in the RX path.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The current PTP initialization logic does not account for firmware
reset that may cause PTP capability to change.  The valid pointer
bp->ptp_cfg is used to indicate that the device is capable of PTP
and that it has been initialized.  So we must clean up bp->ptp_cfg
and free it if the firmware after reset does not support PTP.

Fixes: 93cb62d9 ("bnxt_en: Enable hardware PTP support")
Cc: Richard Cochran <richardcochran@gmail.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The device needs to be in ifup state for PTP to function, so move
bnxt_ptp_init() to bnxt_open().  This means that the PHC will be
registered during bnxt_open().

This also makes firmware reset work correctly.  PTP configurations
may change after firmware upgrade or downgrade.  bnxt_open() will
be called after firmware reset, so it will work properly.

bnxt_ptp_start() is now incorporated into bnxt_ptp_init().  We now
also need to call bnxt_ptp_clear() in bnxt_close().

Fixes: 93cb62d9 ("bnxt_en: Enable hardware PTP support")
Cc: Richard Cochran <richardcochran@gmail.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnxt_half_open_nic() is called during during ethtool self test and is
protected by rtnl_lock.  Firmware reset can be happening at the same
time.  Only critical portions of the entire firmware reset sequence
are protected by the rtnl_lock.  It is possible that bnxt_half_open_nic()
can be called when the firmware reset sequence is aborting.  In that
case, bnxt_half_open_nic() needs to check if the ABORT_ERR flag is set
and abort if it is.  The ethtool self test will fail but the NIC will be
brought to a consistent IF_DOWN state.

Without this patch, if bnxt_half_open_nic() were to continue in this
error state, it may crash like this:

  bnxt_en 0000:82:00.1 enp130s0f1np1: FW reset in progress during close, FW reset will be aborted
  Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
  ...
  Process ethtool (pid: 333327, stack limit = 0x0000000046476577)
  Call trace:
  bnxt_alloc_mem+0x444/0xef0 [bnxt_en]
  bnxt_half_open_nic+0x24/0xb8 [bnxt_en]
  bnxt_self_test+0x2dc/0x390 [bnxt_en]
  ethtool_self_test+0xe0/0x1f8
  dev_ethtool+0x1744/0x22d0
  dev_ioctl+0x190/0x3e0
  sock_ioctl+0x238/0x480
  do_vfs_ioctl+0xc4/0x758
  ksys_ioctl+0x84/0xb8
  __arm64_sys_ioctl+0x28/0x38
  el0_svc_handler+0xb0/0x180
  el0_svc+0x8/0xc

Fixes: a1301f08 ("bnxt_en: Check abort error state in bnxt_open_nic().")
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Only pass supported VLAN protocol IDs for stripped VLAN tags to the
stack.  The stack will hit WARN() if the protocol ID is unsupported.

Existing firmware sets up the chip to strip 0x8100, 0x88a8, 0x9100.
Only the 1st two protocols are supported by the kernel.

Fixes: a196e96b ("bnxt_en: clean up VLAN feature bit handling")
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When bnxt_open() fails in the firmware reset path, the driver needs to
gracefully abort, but it is executing code that should be invoked only
in the success path.  Define a function to abort FW reset and
consolidate all error paths to call this new function.

Fixes: dab62e7c ("bnxt_en: Implement faster recovery for firmware fatal error.")
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In the BNXT_FW_RESET_STATE_POLL_VF state in bnxt_fw_reset_task() after all
VFs have unregistered, we need to check for BNXT_STATE_ABORT_ERR after
we acquire the rtnl_lock.  If the flag is set, we need to abort.

Fixes: 230d1f0d ("bnxt_en: Handle firmware reset.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The capabilities can change after firmware upgrade/downgrade, so we
should get the up-to-date RoCE capabilities everytime bnxt_ulp_probe()
is called.

Fixes: 2151fe08 ("bnxt_en: Handle RESET_NOTIFY async event from firmware.")
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ETS proportions are presented to HWRM_QUEUE_COS2BW_CFG as minimum
bandwidth constraints. Thus, zero is a legal value for a given TC.
However, if all the other TCs sum up to 100%, then at least one
hardware queue will starve, resulting in guaranteed TX timeouts.
Reject such nonsensical configurations.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If device is already disabled in reset path and PCI io error is
detected before the device could be enabled, driver could
call pci_disable_device() for already disabled device. Fix this
problem by calling pci_disable_device() only if the device is already
enabled.

Fixes: 6316ea6d ("bnxt_en: Enable AER support.")
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>