linux-2.6-microblaze.git
2 years agoRevert "net: dsa: setup master before ports"
Vladimir Oltean [Tue, 12 Apr 2022 09:44:26 +0000 (12:44 +0300)]
Revert "net: dsa: setup master before ports"

This reverts commit 11fd667dac315ea3f2469961f6d2869271a46cae.

dsa_slave_change_mtu() updates the MTU of the DSA master and of the
associated CPU port, but only if it detects a change to the master MTU.

The blamed commit in the Fixes: tag below addressed a regression where
dsa_slave_change_mtu() would return early and not do anything due to
ds->ops->port_change_mtu() not being implemented.

However, that commit also had the effect that the master MTU got set up
to the correct value by dsa_master_setup(), but the associated CPU port's
MTU did not get updated. This causes breakage for drivers that rely on
the ->port_change_mtu() DSA call to account for the tagging overhead on
the CPU port, and don't set up the initial MTU during the setup phase.

Things actually worked before because they were in a fragile equilibrium
where dsa_slave_change_mtu() was called before dsa_master_setup() was.
So dsa_slave_change_mtu() could actually detect a change and update the
CPU port MTU too.

Restore the code to the way things used to work by reverting the reorder
of dsa_tree_setup_master() and dsa_tree_setup_ports(). That change did
not have a concrete motivation going for it anyway, it just looked
better.

Fixes: 066dfc429040 ("Revert "net: dsa: stop updating master MTU from master.c"")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomacvlan: Fix leaking skb in source mode with nodst option
Martin Willi [Tue, 12 Apr 2022 09:34:57 +0000 (11:34 +0200)]
macvlan: Fix leaking skb in source mode with nodst option

The MACVLAN receive handler clones skbs to all matching source MACVLAN
interfaces, before it passes the packet along to match on destination
based MACVLANs.

When using the MACVLAN nodst mode, passing the packet to destination based
MACVLANs is omitted and the handler returns with RX_HANDLER_CONSUMED.
However, the passed skb is not freed, leaking for any packet processed
with the nodst option.

Properly free the skb when consuming packets to fix that leak.

Fixes: 427f0c8c194b ("macvlan: Add nodst option to macvlan type source")
Signed-off-by: Martin Willi <martin@strongswan.org>
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 [Tue, 12 Apr 2022 16:48:53 +0000 (09:48 -0700)]
Merge git://git./linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

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

1) Fix cgroupv2 from the input path, from Florian Westphal.

2) Fix incorrect return value of nft_parse_register(), from Antoine Tenart.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nf_tables: nft_parse_register can return a negative value
  netfilter: nft_socket: make cgroup match work in input too
====================

Link: https://lore.kernel.org/r/20220412094246.448055-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonetfilter: nf_tables: nft_parse_register can return a negative value
Antoine Tenart [Tue, 12 Apr 2022 08:14:59 +0000 (10:14 +0200)]
netfilter: nf_tables: nft_parse_register can return a negative value

Since commit 6e1acfa387b9 ("netfilter: nf_tables: validate registers
coming from userspace.") nft_parse_register can return a negative value,
but the function prototype is still returning an unsigned int.

Fixes: 6e1acfa387b9 ("netfilter: nf_tables: validate registers coming from userspace.")
Signed-off-by: Antoine Tenart <atenart@kernel.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agoMerge branch 'net-lan966x-lan966x-fixes'
Jakub Kicinski [Tue, 12 Apr 2022 03:47:18 +0000 (20:47 -0700)]
Merge branch 'net-lan966x-lan966x-fixes'

Horatiu Vultur says:

====================
net: lan966x: lan966x fixes

This contains different fixes for lan966x in different areas like PTP, MAC,
Switchdev and IGMP processing.
====================

Link: https://lore.kernel.org/r/20220409184143.1204786-1-horatiu.vultur@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: lan966x: Stop processing the MAC entry is port is wrong.
Horatiu Vultur [Sat, 9 Apr 2022 18:41:43 +0000 (20:41 +0200)]
net: lan966x: Stop processing the MAC entry is port is wrong.

Currently when getting a new MAC is learn, the HW generates an
interrupt. So then the SW will check the new entry and checks if it
arrived on a correct port. If it didn't just generate a warning.
But this could still crash the system. Therefore stop processing that
entry when an issue is seen.

Fixes: 5ccd66e01cbef8 ("net: lan966x: add support for interrupts from analyzer")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: lan966x: Fix when a port's upper is changed.
Horatiu Vultur [Sat, 9 Apr 2022 18:41:42 +0000 (20:41 +0200)]
net: lan966x: Fix when a port's upper is changed.

On lan966x it is not allowed to have foreign interfaces under a bridge
which already contains lan966x ports. So when a port leaves the bridge
it would call switchdev_bridge_port_unoffload which eventually will
notify the other ports that bridge left the vlan group but that is not
true because the bridge is still part of the vlan group.

Therefore when a port leaves the bridge, stop generating replays because
already the HW cleared after itself and the other ports don't need to do
anything else.

Fixes: cf2f60897e921e ("net: lan966x: Add support to offload the forwarding.")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: lan966x: Fix IGMP snooping when frames have vlan tag
Horatiu Vultur [Sat, 9 Apr 2022 18:41:41 +0000 (20:41 +0200)]
net: lan966x: Fix IGMP snooping when frames have vlan tag

In case an IGMP frame has a vlan tag, then the function
lan966x_hw_offload couldn't figure out that is a IGMP frame. Therefore
the SW thinks that the frame was already forward by the HW which is not
true.
Extend lan966x_hw_offload to pop the vlan tag if are any and then check
for IGMP frames.

Fixes: 47aeea0d57e80c ("net: lan966x: Implement the callback SWITCHDEV_ATTR_ID_BRIDGE_MC_DISABLED ")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: lan966x: Update lan966x_ptp_get_nominal_value
Horatiu Vultur [Sat, 9 Apr 2022 18:41:40 +0000 (20:41 +0200)]
net: lan966x: Update lan966x_ptp_get_nominal_value

The clk_per_cfg register represents the value added to the system clock
for each clock cycle. The issue is that the default value is wrong,
meaning that in case the DUT was a grandmaster then everone in the
network was too slow. In case there was a grandmaster, then there is no
issue because the DUT will configure clk_per_cfg register based on the
master frequency.

Fixes: d096459494a887 ("net: lan966x: Add support for ptp clocks")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosctp: Initialize daddr on peeled off socket
Petr Malat [Sat, 9 Apr 2022 06:36:11 +0000 (08:36 +0200)]
sctp: Initialize daddr on peeled off socket

Function sctp_do_peeloff() wrongly initializes daddr of the original
socket instead of the peeled off socket, which makes getpeername()
return zeroes instead of the primary address. Initialize the new socket
instead.

Fixes: d570ee490fb1 ("[SCTP]: Correctly set daddr for IPv6 sockets during peeloff")
Signed-off-by: Petr Malat <oss@malat.biz>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Link: https://lore.kernel.org/r/20220409063611.673193-1-oss@malat.biz
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 [Tue, 12 Apr 2022 01:34:43 +0000 (18:34 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-04-08

Alexander fixes a use after free issue with aRFS for ice driver.

Mateusz reverts a commit that introduced issues related to device
resets for iavf driver.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  Revert "iavf: Fix deadlock occurrence during resetting VF interface"
  ice: arfs: fix use-after-free when freeing @rx_cpu_rmap
====================

Link: https://lore.kernel.org/r/20220408163411.2415552-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'net-smc-fixes-2022-04-08'
Jakub Kicinski [Tue, 12 Apr 2022 01:28:07 +0000 (18:28 -0700)]
Merge branch 'net-smc-fixes-2022-04-08'

Karsten Graul says:

====================
net/smc: fixes 2022-04-08

Patch 1 fixes two usages of snprintf() with non null-terminated
string which results into an out-of-bounds read.
Pach 2 fixes a syzbot finding where a pointer check was missed
before the call to dev_name().
Patch 3 fixes a crash when already released memory is used as
a function pointer.
====================

Link: https://lore.kernel.org/r/20220408151035.1044701-1-kgraul@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/smc: Fix af_ops of child socket pointing to released memory
Karsten Graul [Fri, 8 Apr 2022 15:10:35 +0000 (17:10 +0200)]
net/smc: Fix af_ops of child socket pointing to released memory

Child sockets may inherit the af_ops from the parent listen socket.
When the listen socket is released then the af_ops of the child socket
points to released memory.
Solve that by restoring the original af_ops for child sockets which
inherited the parent af_ops. And clear any inherited user_data of the
parent socket.

Fixes: 8270d9c21041 ("net/smc: Limit backlog connections")
Reviewed-by: Wenjia Zhang <wenjia@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/smc: Fix NULL pointer dereference in smc_pnet_find_ib()
Karsten Graul [Fri, 8 Apr 2022 15:10:34 +0000 (17:10 +0200)]
net/smc: Fix NULL pointer dereference in smc_pnet_find_ib()

dev_name() was called with dev.parent as argument but without to
NULL-check it before.
Solve this by checking the pointer before the call to dev_name().

Fixes: af5f60c7e3d5 ("net/smc: allow PCI IDs as ib device names in the pnet table")
Reported-by: syzbot+03e3e228510223dabd34@syzkaller.appspotmail.com
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/smc: use memcpy instead of snprintf to avoid out of bounds read
Karsten Graul [Fri, 8 Apr 2022 15:10:33 +0000 (17:10 +0200)]
net/smc: use memcpy instead of snprintf to avoid out of bounds read

Using snprintf() to convert not null-terminated strings to null
terminated strings may cause out of bounds read in the source string.
Therefore use memcpy() and terminate the target string with a null
afterwards.

Fixes: fa0866625543 ("net/smc: add support for user defined EIDs")
Fixes: 3c572145c24e ("net/smc: add generic netlink support for system EID")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: macb: Restart tx only if queue pointer is lagging
Tomas Melin [Thu, 7 Apr 2022 16:16:59 +0000 (19:16 +0300)]
net: macb: Restart tx only if queue pointer is lagging

commit 4298388574da ("net: macb: restart tx after tx used bit read")
added support for restarting transmission. Restarting tx does not work
in case controller asserts TXUBR interrupt and TQBP is already at the end
of the tx queue. In that situation, restarting tx will immediately cause
assertion of another TXUBR interrupt. The driver will end up in an infinite
interrupt loop which it cannot break out of.

For cases where TQBP is at the end of the tx queue, instead
only clear TX_USED interrupt. As more data gets pushed to the queue,
transmission will resume.

This issue was observed on a Xilinx Zynq-7000 based board.
During stress test of the network interface,
driver would get stuck on interrupt loop within seconds or minutes
causing CPU to stall.

Signed-off-by: Tomas Melin <tomas.melin@vaisala.com>
Tested-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Reviewed-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20220407161659.14532-1-tomas.melin@vaisala.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoipv6: fix panic when forwarding a pkt with no in6 dev
Nicolas Dichtel [Fri, 8 Apr 2022 14:03:42 +0000 (16:03 +0200)]
ipv6: fix panic when forwarding a pkt with no in6 dev

kongweibin reported a kernel panic in ip6_forward() when input interface
has no in6 dev associated.

The following tc commands were used to reproduce this panic:
tc qdisc del dev vxlan100 root
tc qdisc add dev vxlan100 root netem corrupt 5%

CC: stable@vger.kernel.org
Fixes: ccd27f05ae7b ("ipv6: fix 'disable_policy' for fwd packets")
Reported-by: kongweibin <kongweibin2@huawei.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodpaa_eth: Fix missing of_node_put in dpaa_get_ts_info()
Lv Ruyi [Fri, 8 Apr 2022 09:49:41 +0000 (09:49 +0000)]
dpaa_eth: Fix missing of_node_put in dpaa_get_ts_info()

Both of of_get_parent() and of_parse_phandle() return node pointer with
refcount incremented, use of_node_put() on it to decrease refcount
when done.

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Lv Ruyi <lv.ruyi@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonetfilter: nft_socket: make cgroup match work in input too
Florian Westphal [Sat, 9 Apr 2022 11:20:19 +0000 (13:20 +0200)]
netfilter: nft_socket: make cgroup match work in input too

cgroupv2 helper function ignores the already-looked up sk
and uses skb->sk instead.

Just pass sk from the calling function instead; this will
make cgroup matching work for udp and tcp in input even when
edemux did not set skb->sk already.

Fixes: e0bb96db96f8 ("netfilter: nft_socket: add support for cgroupsv2")
Signed-off-by: Florian Westphal <fw@strlen.de>
Tested-by: Topi Miettinen <toiwoton@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agonet/sched: taprio: Check if socket flags are valid
Benedikt Spranger [Fri, 8 Apr 2022 09:47:45 +0000 (11:47 +0200)]
net/sched: taprio: Check if socket flags are valid

A user may set the SO_TXTIME socket option to ensure a packet is send
at a given time. The taprio scheduler has to confirm, that it is allowed
to send a packet at that given time, by a check against the packet time
schedule. The scheduler drop the packet, if the gates are closed at the
given send time.

The check, if SO_TXTIME is set, may fail since sk_flags are part of an
union and the union is used otherwise. This happen, if a socket is not
a full socket, like a request socket for example.

Add a check to verify, if the union is used for sk_flags.

Fixes: 4cfd5779bd6e ("taprio: Add support for txtime-assist mode")
Signed-off-by: Benedikt Spranger <b.spranger@linutronix.de>
Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: stmmac: fix altr_tse_pcs function when using a fixed-link
Dinh Nguyen [Thu, 7 Apr 2022 13:25:21 +0000 (08:25 -0500)]
net: ethernet: stmmac: fix altr_tse_pcs function when using a fixed-link

When using a fixed-link, the altr_tse_pcs driver crashes
due to null-pointer dereference as no phy_device is provided to
tse_pcs_fix_mac_speed function. Fix this by adding a check for
phy_dev before calling the tse_pcs_fix_mac_speed() function.

Also clean up the tse_pcs_fix_mac_speed function a bit. There is
no need to check for splitter_base and sgmii_adapter_base
because the driver will fail if these 2 variables are not
derived from the device tree.

Fixes: fb3bbdb85989 ("net: ethernet: Add TSE PCS support to dwmac-socfpga")
Signed-off-by: Dinh Nguyen <dinguyen@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: phy: LAN87xx: remove genphy_softreset in config_aneg
Arun Ramadoss [Thu, 7 Apr 2022 04:46:10 +0000 (10:16 +0530)]
net: phy: LAN87xx: remove genphy_softreset in config_aneg

When the T1 phy master/slave state is changed, at the end of config_aneg
function genphy_softreset is called. After the reset all the registers
configured during the config_init are restored to default value.
To avoid this, removed the genphy_softreset call.

v1->v2
------
Added the author in cc

Fixes: 8a1b415d70b7 ("net: phy: added ethtool master-slave configuration support")
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: atlantic: Avoid out-of-bounds indexing
Kai-Heng Feng [Fri, 8 Apr 2022 02:22:04 +0000 (10:22 +0800)]
net: atlantic: Avoid out-of-bounds indexing

UBSAN warnings are observed on atlantic driver:
[ 294.432996] UBSAN: array-index-out-of-bounds in /build/linux-Qow4fL/linux-5.15.0/drivers/net/ethernet/aquantia/atlantic/aq_nic.c:484:48
[ 294.433695] index 8 is out of range for type 'aq_vec_s *[8]'

The ring is dereferenced right before breaking out the loop, to prevent
that from happening, only use the index in the loop to fix the issue.

BugLink: https://bugs.launchpad.net/bugs/1958770
Tested-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Reviewed-by: Igor Russkikh <irusskikh@marvell.com>
Link: https://lore.kernel.org/r/20220408022204.16815-1-kai.heng.feng@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: felix: suppress -EPROBE_DEFER errors
Michael Walle [Fri, 8 Apr 2022 10:15:21 +0000 (12:15 +0200)]
net: dsa: felix: suppress -EPROBE_DEFER errors

The DSA master might not have been probed yet in which case the probe of
the felix switch fails with -EPROBE_DEFER:
[    4.435305] mscc_felix 0000:00:00.5: Failed to register DSA switch: -517

It is not an error. Use dev_err_probe() to demote this particular error
to a debug message.

Fixes: 56051948773e ("net: dsa: ocelot: add driver for Felix switch family")
Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20220408101521.281886-1-michael@walle.cc
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/sched: fix initialization order when updating chain 0 head
Marcelo Ricardo Leitner [Thu, 7 Apr 2022 14:29:23 +0000 (11:29 -0300)]
net/sched: fix initialization order when updating chain 0 head

Currently, when inserting a new filter that needs to sit at the head
of chain 0, it will first update the heads pointer on all devices using
the (shared) block, and only then complete the initialization of the new
element so that it has a "next" element.

This can lead to a situation that the chain 0 head is propagated to
another CPU before the "next" initialization is done. When this race
condition is triggered, packets being matched on that CPU will simply
miss all other filters, and will flow through the stack as if there were
no other filters installed. If the system is using OVS + TC, such
packets will get handled by vswitchd via upcall, which results in much
higher latency and reordering. For other applications it may result in
packet drops.

This is reproducible with a tc only setup, but it varies from system to
system. It could be reproduced with a shared block amongst 10 veth
tunnels, and an ingress filter mirroring packets to another veth.
That's because using the last added veth tunnel to the shared block to
do the actual traffic, it makes the race window bigger and easier to
trigger.

The fix is rather simple, to just initialize the next pointer of the new
filter instance (tp) before propagating the head change.

The fixes tag is pointing to the original code though this issue should
only be observed when using it unlocked.

Fixes: 2190d1d0944f ("net: sched: introduce helpers to work with filter chains")
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Link: https://lore.kernel.org/r/b97d5f4eaffeeb9d058155bcab63347527261abf.1649341369.git.marcelo.leitner@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosctp: use the correct skb for security_sctp_assoc_request
Xin Long [Thu, 7 Apr 2022 13:24:22 +0000 (09:24 -0400)]
sctp: use the correct skb for security_sctp_assoc_request

Yi Chen reported an unexpected sctp connection abort, and it occurred when
COOKIE_ECHO is bundled with DATA Fragment by SCTP HW GSO. As the IP header
is included in chunk->head_skb instead of chunk->skb, it failed to check
IP header version in security_sctp_assoc_request().

According to Ondrej, SELinux only looks at IP header (address and IPsec
options) and XFRM state data, and these are all included in head_skb for
SCTP HW GSO packets. So fix it by using head_skb when calling
security_sctp_assoc_request() in processing COOKIE_ECHO.

v1->v2:
  - As Ondrej noticed, chunk->head_skb should also be used for
    security_sctp_assoc_established() in sctp_sf_do_5_1E_ca().

Fixes: e215dab1c490 ("security: call security_sctp_assoc_request in sctp_sf_do_5_1D_ce")
Reported-by: Yi Chen <yiche@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Link: https://lore.kernel.org/r/71becb489e51284edf0c11fc15246f4ed4cef5b6.1649337862.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: i2c: Fix initialization error flow
Vadim Pasternak [Thu, 7 Apr 2022 07:07:03 +0000 (10:07 +0300)]
mlxsw: i2c: Fix initialization error flow

Add mutex_destroy() call in driver initialization error flow.

Fixes: 6882b0aee180f ("mlxsw: Introduce support for I2C bus")
Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/20220407070703.2421076-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: mdio: don't defer probe forever if PHY IRQ provider is missing
Vladimir Oltean [Thu, 7 Apr 2022 16:55:38 +0000 (19:55 +0300)]
net: mdio: don't defer probe forever if PHY IRQ provider is missing

When a driver for an interrupt controller is missing, of_irq_get()
returns -EPROBE_DEFER ad infinitum, causing
fwnode_mdiobus_phy_device_register(), and ultimately, the entire
of_mdiobus_register() call, to fail. In turn, any phy_connect() call
towards a PHY on this MDIO bus will also fail.

This is not what is expected to happen, because the PHY library falls
back to poll mode when of_irq_get() returns a hard error code, and the
MDIO bus, PHY and attached Ethernet controller work fine, albeit
suboptimally, when the PHY library polls for link status. However,
-EPROBE_DEFER has special handling given the assumption that at some
point probe deferral will stop, and the driver for the supplier will
kick in and create the IRQ domain.

Reasons for which the interrupt controller may be missing:

- It is not yet written. This may happen if a more recent DT blob (with
  an interrupt-parent for the PHY) is used to boot an old kernel where
  the driver didn't exist, and that kernel worked with the
  vintage-correct DT blob using poll mode.

- It is compiled out. Behavior is the same as above.

- It is compiled as a module. The kernel will wait for a number of
  seconds specified in the "deferred_probe_timeout" boot parameter for
  user space to load the required module. The current default is 0,
  which times out at the end of initcalls. It is possible that this
  might cause regressions unless users adjust this boot parameter.

The proposed solution is to use the driver_deferred_probe_check_state()
helper function provided by the driver core, which gives up after some
-EPROBE_DEFER attempts, taking "deferred_probe_timeout" into consideration.
The return code is changed from -EPROBE_DEFER into -ENODEV or
-ETIMEDOUT, depending on whether the kernel is compiled with support for
modules or not.

Fixes: 66bdede495c7 ("of_mdio: Fix broken PHY IRQ in case of probe deferral")
Suggested-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20220407165538.4084809-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoRevert "iavf: Fix deadlock occurrence during resetting VF interface"
Mateusz Palczewski [Thu, 24 Mar 2022 13:19:15 +0000 (14:19 +0100)]
Revert "iavf: Fix deadlock occurrence during resetting VF interface"

This change caused a regression with resetting while changing network
namespaces. By clearing the IFF_UP flag, the kernel now thinks it has
fully closed the device.

This reverts commit 0cc318d2e8408bc0ffb4662a0c3e5e57005ac6ff.

Fixes: 0cc318d2e840 ("iavf: Fix deadlock occurrence during resetting VF interface")
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: arfs: fix use-after-free when freeing @rx_cpu_rmap
Alexander Lobakin [Mon, 4 Apr 2022 16:15:09 +0000 (18:15 +0200)]
ice: arfs: fix use-after-free when freeing @rx_cpu_rmap

The CI testing bots triggered the following splat:

[  718.203054] BUG: KASAN: use-after-free in free_irq_cpu_rmap+0x53/0x80
[  718.206349] Read of size 4 at addr ffff8881bd127e00 by task sh/20834
[  718.212852] CPU: 28 PID: 20834 Comm: sh Kdump: loaded Tainted: G S      W IOE     5.17.0-rc8_nextqueue-devqueue-02643-g23f3121aca93 #1
[  718.219695] Hardware name: Intel Corporation S2600WFT/S2600WFT, BIOS SE5C620.86B.02.01.0012.070720200218 07/07/2020
[  718.223418] Call Trace:
[  718.227139]
[  718.230783]  dump_stack_lvl+0x33/0x42
[  718.234431]  print_address_description.constprop.9+0x21/0x170
[  718.238177]  ? free_irq_cpu_rmap+0x53/0x80
[  718.241885]  ? free_irq_cpu_rmap+0x53/0x80
[  718.245539]  kasan_report.cold.18+0x7f/0x11b
[  718.249197]  ? free_irq_cpu_rmap+0x53/0x80
[  718.252852]  free_irq_cpu_rmap+0x53/0x80
[  718.256471]  ice_free_cpu_rx_rmap.part.11+0x37/0x50 [ice]
[  718.260174]  ice_remove_arfs+0x5f/0x70 [ice]
[  718.263810]  ice_rebuild_arfs+0x3b/0x70 [ice]
[  718.267419]  ice_rebuild+0x39c/0xb60 [ice]
[  718.270974]  ? asm_sysvec_apic_timer_interrupt+0x12/0x20
[  718.274472]  ? ice_init_phy_user_cfg+0x360/0x360 [ice]
[  718.278033]  ? delay_tsc+0x4a/0xb0
[  718.281513]  ? preempt_count_sub+0x14/0xc0
[  718.284984]  ? delay_tsc+0x8f/0xb0
[  718.288463]  ice_do_reset+0x92/0xf0 [ice]
[  718.292014]  ice_pci_err_resume+0x91/0xf0 [ice]
[  718.295561]  pci_reset_function+0x53/0x80
<...>
[  718.393035] Allocated by task 690:
[  718.433497] Freed by task 20834:
[  718.495688] Last potentially related work creation:
[  718.568966] The buggy address belongs to the object at ffff8881bd127e00
                which belongs to the cache kmalloc-96 of size 96
[  718.574085] The buggy address is located 0 bytes inside of
                96-byte region [ffff8881bd127e00ffff8881bd127e60)
[  718.579265] The buggy address belongs to the page:
[  718.598905] Memory state around the buggy address:
[  718.601809]  ffff8881bd127d00: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
[  718.604796]  ffff8881bd127d80: 00 00 00 00 00 00 00 00 00 00 fc fc fc fc fc fc
[  718.607794] >ffff8881bd127e00: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
[  718.610811]                    ^
[  718.613819]  ffff8881bd127e80: 00 00 00 00 00 00 00 00 00 00 00 00 fc fc fc fc
[  718.617107]  ffff8881bd127f00: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc

This is due to that free_irq_cpu_rmap() is always being called
*after* (devm_)free_irq() and thus it tries to work with IRQ descs
already freed. For example, on device reset the driver frees the
rmap right before allocating a new one (the splat above).
Make rmap creation and freeing function symmetrical with
{request,free}_irq() calls i.e. do that on ifup/ifdown instead
of device probe/remove/resume. These operations can be performed
independently from the actual device aRFS configuration.
Also, make sure ice_vsi_free_irq() clears IRQ affinity notifiers
only when aRFS is disabled -- otherwise, CPU rmap sets and clears
its own and they must not be touched manually.

Fixes: 28bf26724fdb0 ("ice: Implement aRFS")
Co-developed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Tested-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agovxlan: fix error return code in vxlan_fdb_append
Hongbin Wang [Thu, 7 Apr 2022 02:46:22 +0000 (22:46 -0400)]
vxlan: fix error return code in vxlan_fdb_append

When kmalloc and dst_cache_init failed,
should return ENOMEM rather than ENOBUFS.

Signed-off-by: Hongbin Wang <wh_bin@126.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_dissector: fix false-positive __read_overflow2_field() warning
Jakub Kicinski [Wed, 6 Apr 2022 21:15:21 +0000 (14:15 -0700)]
flow_dissector: fix false-positive __read_overflow2_field() warning

Bounds checking is unhappy that we try to copy both Ethernet
addresses but pass pointer to the first one. Luckily destination
address is the first field so pass the pointer to the entire header,
whatever.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoveth: Ensure eth header is in skb's linear part
Guillaume Nault [Wed, 6 Apr 2022 14:18:54 +0000 (16:18 +0200)]
veth: Ensure eth header is in skb's linear part

After feeding a decapsulated packet to a veth device with act_mirred,
skb_headlen() may be 0. But veth_xmit() calls __dev_forward_skb(),
which expects at least ETH_HLEN byte of linear data (as
__dev_forward_skb2() calls eth_type_trans(), which pulls ETH_HLEN bytes
unconditionally).

Use pskb_may_pull() to ensure veth_xmit() respects this constraint.

kernel BUG at include/linux/skbuff.h:2328!
RIP: 0010:eth_type_trans+0xcf/0x140
Call Trace:
 <IRQ>
 __dev_forward_skb2+0xe3/0x160
 veth_xmit+0x6e/0x250 [veth]
 dev_hard_start_xmit+0xc7/0x200
 __dev_queue_xmit+0x47f/0x520
 ? skb_ensure_writable+0x85/0xa0
 ? skb_mpls_pop+0x98/0x1c0
 tcf_mirred_act+0x442/0x47e [act_mirred]
 tcf_action_exec+0x86/0x140
 fl_classify+0x1d8/0x1e0 [cls_flower]
 ? dma_pte_clear_level+0x129/0x1a0
 ? dma_pte_clear_level+0x129/0x1a0
 ? prb_fill_curr_block+0x2f/0xc0
 ? skb_copy_bits+0x11a/0x220
 __tcf_classify+0x58/0x110
 tcf_classify_ingress+0x6b/0x140
 __netif_receive_skb_core.constprop.0+0x47d/0xfd0
 ? __iommu_dma_unmap_swiotlb+0x44/0x90
 __netif_receive_skb_one_core+0x3d/0xa0
 netif_receive_skb+0x116/0x170
 be_process_rx+0x22f/0x330 [be2net]
 be_poll+0x13c/0x370 [be2net]
 __napi_poll+0x2a/0x170
 net_rx_action+0x22f/0x2f0
 __do_softirq+0xca/0x2a8
 __irq_exit_rcu+0xc1/0xe0
 common_interrupt+0x83/0xa0

Fixes: e314dbdc1c0d ("[NET]: Virtual ethernet device driver.")
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobonding: Update layer2 and layer2+3 hash formula documentation
Gal Pressman [Wed, 6 Apr 2022 13:54:20 +0000 (16:54 +0300)]
bonding: Update layer2 and layer2+3 hash formula documentation

When using layer2 or layer2+3 hash, only the 5th byte of the MAC
addresses is used.

Signed-off-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/sched: flower: fix parsing of ethertype following VLAN header
Vlad Buslov [Wed, 6 Apr 2022 11:22:41 +0000 (14:22 +0300)]
net/sched: flower: fix parsing of ethertype following VLAN header

A tc flower filter matching TCA_FLOWER_KEY_VLAN_ETH_TYPE is expected to
match the L2 ethertype following the first VLAN header, as confirmed by
linked discussion with the maintainer. However, such rule also matches
packets that have additional second VLAN header, even though filter has
both eth_type and vlan_ethtype set to "ipv4". Looking at the code this
seems to be mostly an artifact of the way flower uses flow dissector.
First, even though looking at the uAPI eth_type and vlan_ethtype appear
like a distinct fields, in flower they are all mapped to the same
key->basic.n_proto. Second, flow dissector skips following VLAN header as
no keys for FLOW_DISSECTOR_KEY_CVLAN are set and eventually assigns the
value of n_proto to last parsed header. With these, such filters ignore any
headers present between first VLAN header and first "non magic"
header (ipv4 in this case) that doesn't result
FLOW_DISSECT_RET_PROTO_AGAIN.

Fix the issue by extending flow dissector VLAN key structure with new
'vlan_eth_type' field that matches first ethertype following previously
parsed VLAN header. Modify flower classifier to set the new
flow_dissector_key_vlan->vlan_eth_type with value obtained from
TCA_FLOWER_KEY_VLAN_ETH_TYPE/TCA_FLOWER_KEY_CVLAN_ETH_TYPE uAPIs.

Link: https://lore.kernel.org/all/Yjhgi48BpTGh6dig@nanopsycho/
Fixes: 9399ae9a6cb2 ("net_sched: flower: Add vlan support")
Fixes: d64efd0926ba ("net/sched: flower: Add supprt for matching on QinQ vlan headers")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodt-bindings: net: ave: Use unevaluatedProperties
Kunihiko Hayashi [Tue, 5 Apr 2022 07:53:01 +0000 (16:53 +0900)]
dt-bindings: net: ave: Use unevaluatedProperties

This refers common bindings, so this is preferred for
unevaluatedProperties instead of additionalProperties.

Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodt-bindings: net: ave: Clean up clocks, resets, and their names using compatible...
Kunihiko Hayashi [Tue, 5 Apr 2022 07:53:00 +0000 (16:53 +0900)]
dt-bindings: net: ave: Clean up clocks, resets, and their names using compatible string

Instead of "oneOf:" choices, use "allOf:" and "if:" to define clocks,
resets, and their names that can be taken by the compatible string.

The order of clock-names and reset-names doesn't change here.

Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge tag 'net-5.18-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Fri, 8 Apr 2022 05:01:47 +0000 (19:01 -1000)]
Merge tag 'net-5.18-rc2' of git://git./linux/kernel/git/netdev/net

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

  Current release - new code bugs:

   - mctp: correct mctp_i2c_header_create result

   - eth: fungible: fix reference to __udivdi3 on 32b builds

   - eth: micrel: remove latencies support lan8814

  Previous releases - regressions:

   - bpf: resolve to prog->aux->dst_prog->type only for BPF_PROG_TYPE_EXT

   - vrf: fix packet sniffing for traffic originating from ip tunnels

   - rxrpc: fix a race in rxrpc_exit_net()

   - dsa: revert "net: dsa: stop updating master MTU from master.c"

   - eth: ice: fix MAC address setting

  Previous releases - always broken:

   - tls: fix slab-out-of-bounds bug in decrypt_internal

   - bpf: support dual-stack sockets in bpf_tcp_check_syncookie

   - xdp: fix coalescing for page_pool fragment recycling

   - ovs: fix leak of nested actions

   - eth: sfc:
      - add missing xdp queue reinitialization
      - fix using uninitialized xdp tx_queue

   - eth: ice:
      - clear default forwarding VSI during VSI release
      - fix broken IFF_ALLMULTI handling
      - synchronize_rcu() when terminating rings

   - eth: qede: confirm skb is allocated before using

   - eth: aqc111: fix out-of-bounds accesses in RX fixup

   - eth: slip: fix NPD bug in sl_tx_timeout()"

* tag 'net-5.18-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (61 commits)
  drivers: net: slip: fix NPD bug in sl_tx_timeout()
  bpf: Adjust bpf_tcp_check_syncookie selftest to test dual-stack sockets
  bpf: Support dual-stack sockets in bpf_tcp_check_syncookie
  myri10ge: fix an incorrect free for skb in myri10ge_sw_tso
  net: usb: aqc111: Fix out-of-bounds accesses in RX fixup
  qede: confirm skb is allocated before using
  net: ipv6mr: fix unused variable warning with CONFIG_IPV6_PIMSM_V2=n
  net: phy: mscc-miim: reject clause 45 register accesses
  net: axiemac: use a phandle to reference pcs_phy
  dt-bindings: net: add pcs-handle attribute
  net: axienet: factor out phy_node in struct axienet_local
  net: axienet: setup mdio unconditionally
  net: sfc: fix using uninitialized xdp tx_queue
  rxrpc: fix a race in rxrpc_exit_net()
  net: openvswitch: fix leak of nested actions
  net: ethernet: mv643xx: Fix over zealous checking of_get_mac_address()
  net: openvswitch: don't send internal clone attribute to the userspace.
  net: micrel: Fix KS8851 Kconfig
  ice: clear cmd_type_offset_bsz for TX rings
  ice: xsk: fix VSI state check in ice_xsk_wakeup()
  ...

2 years agoMerge tag 'hyperv-fixes-signed-20220407' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 7 Apr 2022 16:35:34 +0000 (06:35 -1000)]
Merge tag 'hyperv-fixes-signed-20220407' of git://git./linux/kernel/git/hyperv/linux

Pull hyperv fixes from Wei Liu:

 - Correctly propagate coherence information for VMbus devices (Michael
   Kelley)

 - Disable balloon and memory hot-add on ARM64 temporarily (Boqun Feng)

 - Use barrier to prevent reording when reading ring buffer (Michael
   Kelley)

 - Use virt_store_mb in favour of smp_store_mb (Andrea Parri)

 - Fix VMbus device object initialization (Andrea Parri)

 - Deactivate sysctl_record_panic_msg on isolated guest (Andrea Parri)

 - Fix a crash when unloading VMbus module (Guilherme G. Piccoli)

* tag 'hyperv-fixes-signed-20220407' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux:
  Drivers: hv: vmbus: Replace smp_store_mb() with virt_store_mb()
  Drivers: hv: balloon: Disable balloon and hot-add accordingly
  Drivers: hv: balloon: Support status report for larger page sizes
  Drivers: hv: vmbus: Prevent load re-ordering when reading ring buffer
  PCI: hv: Propagate coherence from VMbus device to PCI device
  Drivers: hv: vmbus: Propagate VMbus coherence to each VMbus device
  Drivers: hv: vmbus: Fix potential crash on module unload
  Drivers: hv: vmbus: Fix initialization of device object in vmbus_device_register()
  Drivers: hv: vmbus: Deactivate sysctl_record_panic_msg by default in isolated guests

2 years agoMerge tag 'random-5.18-rc2-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 7 Apr 2022 16:02:55 +0000 (06:02 -1000)]
Merge tag 'random-5.18-rc2-for-linus' of git://git./linux/kernel/git/crng/random

Pull random number generator fixes from Jason Donenfeld:

 - Another fixup to the fast_init/crng_init split, this time in how much
   entropy is being credited, from Jan Varho.

 - As discussed, we now opportunistically call try_to_generate_entropy()
   in /dev/urandom reads, as a replacement for the reverted commit. I
   opted to not do the more invasive wait_for_random_bytes() change at
   least for now, preferring to do something smaller and more obvious
   for the time being, but maybe that can be revisited as things evolve
   later.

 - Userspace can use FUSE or userfaultfd or simply move a process to
   idle priority in order to make a read from the random device never
   complete, which breaks forward secrecy, fixed by overwriting
   sensitive bytes early on in the function.

 - Jann Horn noticed that /dev/urandom reads were only checking for
   pending signals if need_resched() was true, a bug going back to the
   genesis commit, now fixed by always checking for signal_pending() and
   calling cond_resched(). This explains various noticeable signal
   delivery delays I've seen in programs over the years that do long
   reads from /dev/urandom.

 - In order to be more like other devices (e.g. /dev/zero) and to
   mitigate the impact of fixing the above bug, which has been around
   forever (users have never really needed to check the return value of
   read() for medium-sized reads and so perhaps many didn't), we now
   move signal checking to the bottom part of the loop, and do so every
   PAGE_SIZE-bytes.

* tag 'random-5.18-rc2-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random:
  random: check for signals every PAGE_SIZE chunk of /dev/[u]random
  random: check for signal_pending() outside of need_resched() check
  random: do not allow user to keep crng key around on stack
  random: opportunistically initialize on /dev/urandom reads
  random: do not split fast init input in add_hwgenerator_randomness()

2 years agoMerge tag 'ata-5.18-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal...
Linus Torvalds [Thu, 7 Apr 2022 15:56:54 +0000 (05:56 -1000)]
Merge tag 'ata-5.18-rc2' of git://git./linux/kernel/git/dlemoal/libata

Pull ata fixes from Damien Le Moal:

 - Fix a compilation warning due to an uninitialized variable in
   ata_sff_lost_interrupt(), from me.

 - Fix invalid internal command tag handling in the sata_dwc_460ex
   driver, from Christian.

 - Disable READ LOG DMA EXT with Samsung 840 EVO SSDs as this command
   causes the drives to hang, from Christian.

 - Change the config option CONFIG_SATA_LPM_POLICY back to its original
   name CONFIG_SATA_LPM_MOBILE_POLICY to avoid potential problems with
   users losing their configuration (as discussed during the merge
   window), from Mario.

* tag 'ata-5.18-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal/libata:
  ata: ahci: Rename CONFIG_SATA_LPM_POLICY configuration item back
  ata: libata-core: Disable READ LOG DMA EXT for Samsung 840 EVOs
  ata: sata_dwc_460ex: Fix crash due to OOB write
  ata: libata-sff: Fix compilation warning in ata_sff_lost_interrupt()

2 years agodrivers: net: slip: fix NPD bug in sl_tx_timeout()
Duoming Zhou [Tue, 5 Apr 2022 13:22:06 +0000 (21:22 +0800)]
drivers: net: slip: fix NPD bug in sl_tx_timeout()

When a slip driver is detaching, the slip_close() will act to
cleanup necessary resources and sl->tty is set to NULL in
slip_close(). Meanwhile, the packet we transmit is blocked,
sl_tx_timeout() will be called. Although slip_close() and
sl_tx_timeout() use sl->lock to synchronize, we don`t judge
whether sl->tty equals to NULL in sl_tx_timeout() and the
null pointer dereference bug will happen.

   (Thread 1)                 |      (Thread 2)
                              | slip_close()
                              |   spin_lock_bh(&sl->lock)
                              |   ...
...                           |   sl->tty = NULL //(1)
sl_tx_timeout()               |   spin_unlock_bh(&sl->lock)
  spin_lock(&sl->lock);       |
  ...                         |   ...
  tty_chars_in_buffer(sl->tty)|
    if (tty->ops->..) //(2)   |
    ...                       |   synchronize_rcu()

We set NULL to sl->tty in position (1) and dereference sl->tty
in position (2).

This patch adds check in sl_tx_timeout(). If sl->tty equals to
NULL, sl_tx_timeout() will goto out.

Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Reviewed-by: Jiri Slaby <jirislaby@kernel.org>
Link: https://lore.kernel.org/r/20220405132206.55291-1-duoming@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Jakub Kicinski [Thu, 7 Apr 2022 04:58:49 +0000 (21:58 -0700)]
Merge https://git./linux/kernel/git/bpf/bpf

Alexei Starovoitov says:

====================
pull-request: bpf 2022-04-06

We've added 8 non-merge commits during the last 8 day(s) which contain
a total of 9 files changed, 139 insertions(+), 36 deletions(-).

The main changes are:

1) rethook related fixes, from Jiri and Masami.

2) Fix the case when tracing bpf prog is attached to struct_ops, from Martin.

3) Support dual-stack sockets in bpf_tcp_check_syncookie, from Maxim.

* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
  bpf: Adjust bpf_tcp_check_syncookie selftest to test dual-stack sockets
  bpf: Support dual-stack sockets in bpf_tcp_check_syncookie
  bpf: selftests: Test fentry tracing a struct_ops program
  bpf: Resolve to prog->aux->dst_prog->type only for BPF_PROG_TYPE_EXT
  rethook: Fix to use WRITE_ONCE() for rethook:: Handler
  selftests/bpf: Fix warning comparing pointer to 0
  bpf: Fix sparse warnings in kprobe_multi_resolve_syms
  bpftool: Explicit errno handling in skeletons
====================

Link: https://lore.kernel.org/r/20220407031245.73026-1-alexei.starovoitov@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agorandom: check for signals every PAGE_SIZE chunk of /dev/[u]random
Jason A. Donenfeld [Wed, 6 Apr 2022 00:36:16 +0000 (02:36 +0200)]
random: check for signals every PAGE_SIZE chunk of /dev/[u]random

In 1448769c9cdb ("random: check for signal_pending() outside of
need_resched() check"), Jann pointed out that we previously were only
checking the TIF_NOTIFY_SIGNAL and TIF_SIGPENDING flags if the process
had TIF_NEED_RESCHED set, which meant in practice, super long reads to
/dev/[u]random would delay signal handling by a long time. I tried this
using the below program, and indeed I wasn't able to interrupt a
/dev/urandom read until after several megabytes had been read. The bug
he fixed has always been there, and so code that reads from /dev/urandom
without checking the return value of read() has mostly worked for a long
time, for most sizes, not just for <= 256.

Maybe it makes sense to keep that code working. The reason it was so
small prior, ignoring the fact that it didn't work anyway, was likely
because /dev/random used to block, and that could happen for pretty
large lengths of time while entropy was gathered. But now, it's just a
chacha20 call, which is extremely fast and is just operating on pure
data, without having to wait for some external event. In that sense,
/dev/[u]random is a lot more like /dev/zero.

Taking a page out of /dev/zero's read_zero() function, it always returns
at least one chunk, and then checks for signals after each chunk. Chunk
sizes there are of length PAGE_SIZE. Let's just copy the same thing for
/dev/[u]random, and check for signals and cond_resched() for every
PAGE_SIZE amount of data. This makes the behavior more consistent with
expectations, and should mitigate the impact of Jann's fix for the
age-old signal check bug.

---- test program ----

  #include <unistd.h>
  #include <signal.h>
  #include <stdio.h>
  #include <sys/random.h>

  static unsigned char x[~0U];

  static void handle(int) { }

  int main(int argc, char *argv[])
  {
    pid_t pid = getpid(), child;
    signal(SIGUSR1, handle);
    if (!(child = fork())) {
      for (;;)
        kill(pid, SIGUSR1);
    }
    pause();
    printf("interrupted after reading %zd bytes\n", getrandom(x, sizeof(x), 0));
    kill(child, SIGTERM);
    return 0;
  }

Cc: Jann Horn <jannh@google.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2 years agobpf: Adjust bpf_tcp_check_syncookie selftest to test dual-stack sockets
Maxim Mikityanskiy [Wed, 6 Apr 2022 12:41:13 +0000 (15:41 +0300)]
bpf: Adjust bpf_tcp_check_syncookie selftest to test dual-stack sockets

The previous commit fixed support for dual-stack sockets in
bpf_tcp_check_syncookie. This commit adjusts the selftest to verify the
fixed functionality.

Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Arthur Fabre <afabre@cloudflare.com>
Link: https://lore.kernel.org/bpf/20220406124113.2795730-2-maximmi@nvidia.com
2 years agobpf: Support dual-stack sockets in bpf_tcp_check_syncookie
Maxim Mikityanskiy [Wed, 6 Apr 2022 12:41:12 +0000 (15:41 +0300)]
bpf: Support dual-stack sockets in bpf_tcp_check_syncookie

bpf_tcp_gen_syncookie looks at the IP version in the IP header and
validates the address family of the socket. It supports IPv4 packets in
AF_INET6 dual-stack sockets.

On the other hand, bpf_tcp_check_syncookie looks only at the address
family of the socket, ignoring the real IP version in headers, and
validates only the packet size. This implementation has some drawbacks:

1. Packets are not validated properly, allowing a BPF program to trick
   bpf_tcp_check_syncookie into handling an IPv6 packet on an IPv4
   socket.

2. Dual-stack sockets fail the checks on IPv4 packets. IPv4 clients end
   up receiving a SYNACK with the cookie, but the following ACK gets
   dropped.

This patch fixes these issues by changing the checks in
bpf_tcp_check_syncookie to match the ones in bpf_tcp_gen_syncookie. IP
version from the header is taken into account, and it is validated
properly with address family.

Fixes: 399040847084 ("bpf: add helper to check for a valid SYN cookie")
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Acked-by: Arthur Fabre <afabre@cloudflare.com>
Link: https://lore.kernel.org/bpf/20220406124113.2795730-1-maximmi@nvidia.com
2 years agomyri10ge: fix an incorrect free for skb in myri10ge_sw_tso
Xiaomeng Tong [Wed, 6 Apr 2022 03:55:56 +0000 (11:55 +0800)]
myri10ge: fix an incorrect free for skb in myri10ge_sw_tso

All remaining skbs should be released when myri10ge_xmit fails to
transmit a packet. Fix it within another skb_list_walk_safe.

Signed-off-by: Xiaomeng Tong <xiam0nd.tong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: usb: aqc111: Fix out-of-bounds accesses in RX fixup
Marcin Kozlowski [Wed, 6 Apr 2022 08:05:37 +0000 (10:05 +0200)]
net: usb: aqc111: Fix out-of-bounds accesses in RX fixup

aqc111_rx_fixup() contains several out-of-bounds accesses that can be
triggered by a malicious (or defective) USB device, in particular:

 - The metadata array (desc_offset..desc_offset+2*pkt_count) can be out of bounds,
   causing OOB reads and (on big-endian systems) OOB endianness flips.
 - A packet can overlap the metadata array, causing a later OOB
   endianness flip to corrupt data used by a cloned SKB that has already
   been handed off into the network stack.
 - A packet SKB can be constructed whose tail is far beyond its end,
   causing out-of-bounds heap data to be considered part of the SKB's
   data.

Found doing variant analysis. Tested it with another driver (ax88179_178a), since
I don't have a aqc111 device to test it, but the code looks very similar.

Signed-off-by: Marcin Kozlowski <marcinguy@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoqede: confirm skb is allocated before using
Jamie Bainbridge [Wed, 6 Apr 2022 11:19:19 +0000 (21:19 +1000)]
qede: confirm skb is allocated before using

qede_build_skb() assumes build_skb() always works and goes straight
to skb_reserve(). However, build_skb() can fail under memory pressure.
This results in a kernel panic because the skb to reserve is NULL.

Add a check in case build_skb() failed to allocate and return NULL.

The NULL return is handled correctly in callers to qede_build_skb().

Fixes: 8a8633978b842 ("qede: Add build_skb() support.")
Signed-off-by: Jamie Bainbridge <jamie.bainbridge@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ipv6mr: fix unused variable warning with CONFIG_IPV6_PIMSM_V2=n
Florian Westphal [Wed, 6 Apr 2022 10:04:45 +0000 (12:04 +0200)]
net: ipv6mr: fix unused variable warning with CONFIG_IPV6_PIMSM_V2=n

net/ipv6/ip6mr.c:1656:14: warning: unused variable 'do_wrmifwhole'

Move it to the CONFIG_IPV6_PIMSM_V2 scope where its used.

Fixes: 4b340a5a726d ("net: ip6mr: add support for passing full packet on wrong mif")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net...
David S. Miller [Wed, 6 Apr 2022 14:03:50 +0000 (15:03 +0100)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-04-05

Maciej Fijalkowski says:

We were solving issues around AF_XDP busy poll's not-so-usual scenarios,
such as very big busy poll budgets applied to very small HW rings. This
set carries the things that were found during that work that apply to
net tree.

One thing that was fixed for all in-tree ZC drivers was missing on ice
side all the time - it's about syncing RCU before destroying XDP
resources. Next one fixes the bit that is checked in ice_xsk_wakeup and
third one avoids false setting of DD bits on Tx descriptors.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoDrivers: hv: vmbus: Replace smp_store_mb() with virt_store_mb()
Andrea Parri (Microsoft) [Mon, 28 Mar 2022 15:44:57 +0000 (17:44 +0200)]
Drivers: hv: vmbus: Replace smp_store_mb() with virt_store_mb()

Following the recommendation in Documentation/memory-barriers.txt for
virtual machine guests.

Fixes: 8b6a877c060ed ("Drivers: hv: vmbus: Replace the per-CPU channel lists with a global array of channels")
Signed-off-by: Andrea Parri (Microsoft) <parri.andrea@gmail.com>
Link: https://lore.kernel.org/r/20220328154457.100872-1-parri.andrea@gmail.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
2 years agoDrivers: hv: balloon: Disable balloon and hot-add accordingly
Boqun Feng [Fri, 25 Mar 2022 02:32:12 +0000 (10:32 +0800)]
Drivers: hv: balloon: Disable balloon and hot-add accordingly

Currently there are known potential issues for balloon and hot-add on
ARM64:

* Unballoon requests from Hyper-V should only unballoon ranges
that are guest page size aligned, otherwise guests cannot handle
because it's impossible to partially free a page. This is a
problem when guest page size > 4096 bytes.

* Memory hot-add requests from Hyper-V should provide the NUMA
node id of the added ranges or ARM64 should have a functional
memory_add_physaddr_to_nid(), otherwise the node id is missing
for add_memory().

These issues require discussions on design and implementation. In the
meanwhile, post_status() is working and essential to guest monitoring.
Therefore instead of disabling the entire hv_balloon driver, the
ballooning (when page size > 4096 bytes) and hot-add are disabled
accordingly for now. Once the issues are fixed, they can be re-enable in
these cases.

Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Link: https://lore.kernel.org/r/20220325023212.1570049-3-boqun.feng@gmail.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
2 years agoDrivers: hv: balloon: Support status report for larger page sizes
Boqun Feng [Fri, 25 Mar 2022 02:32:11 +0000 (10:32 +0800)]
Drivers: hv: balloon: Support status report for larger page sizes

DM_STATUS_REPORT expects the numbers of pages in the unit of 4k pages
(HV_HYP_PAGE) instead of guest pages, so to make it work when guest page
sizes are larger than 4k, convert the numbers of guest pages into the
numbers of HV_HYP_PAGEs.

Note that the numbers of guest pages are still used for tracing because
tracing is internal to the guest kernel.

Reported-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Link: https://lore.kernel.org/r/20220325023212.1570049-2-boqun.feng@gmail.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
2 years agorandom: check for signal_pending() outside of need_resched() check
Jann Horn [Tue, 5 Apr 2022 16:39:31 +0000 (18:39 +0200)]
random: check for signal_pending() outside of need_resched() check

signal_pending() checks TIF_NOTIFY_SIGNAL and TIF_SIGPENDING, which
signal that the task should bail out of the syscall when possible. This
is a separate concept from need_resched(), which checks
TIF_NEED_RESCHED, signaling that the task should preempt.

In particular, with the current code, the signal_pending() bailout
probably won't work reliably.

Change this to look like other functions that read lots of data, such as
read_zero().

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2 years agorandom: do not allow user to keep crng key around on stack
Jason A. Donenfeld [Tue, 5 Apr 2022 14:40:51 +0000 (16:40 +0200)]
random: do not allow user to keep crng key around on stack

The fast key erasure RNG design relies on the key that's used to be used
and then discarded. We do this, making judicious use of
memzero_explicit().  However, reads to /dev/urandom and calls to
getrandom() involve a copy_to_user(), and userspace can use FUSE or
userfaultfd, or make a massive call, dynamically remap memory addresses
as it goes, and set the process priority to idle, in order to keep a
kernel stack alive indefinitely. By probing
/proc/sys/kernel/random/entropy_avail to learn when the crng key is
refreshed, a malicious userspace could mount this attack every 5 minutes
thereafter, breaking the crng's forward secrecy.

In order to fix this, we just overwrite the stack's key with the first
32 bytes of the "free" fast key erasure output. If we're returning <= 32
bytes to the user, then we can still return those bytes directly, so
that short reads don't become slower. And for long reads, the difference
is hopefully lost in the amortization, so it doesn't change much, with
that amortization helping variously for medium reads.

We don't need to do this for get_random_bytes() and the various
kernel-space callers, and later, if we ever switch to always batching,
this won't be necessary either, so there's no need to change the API of
these functions.

Cc: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jann Horn <jannh@google.com>
Fixes: c92e040d575a ("random: add backtracking protection to the CRNG")
Fixes: 186873c549df ("random: use simpler fast key erasure flow on per-cpu keys")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2 years agonet: phy: mscc-miim: reject clause 45 register accesses
Michael Walle [Tue, 5 Apr 2022 12:02:33 +0000 (14:02 +0200)]
net: phy: mscc-miim: reject clause 45 register accesses

The driver doesn't support clause 45 register access yet, but doesn't
check if the access is a c45 one either. This leads to spurious register
reads and writes. Add the check.

Fixes: 542671fe4d86 ("net: phy: mscc-miim: Add MDIO driver")
Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'axienet-broken-link'
David S. Miller [Wed, 6 Apr 2022 12:54:52 +0000 (13:54 +0100)]
Merge branch 'axienet-broken-link'

Andy Chiu says:

====================
Fix broken link on Xilinx's AXI Ethernet in SGMII mode

The Ethernet driver use phy-handle to reference the PCS/PMA PHY. This
could be a problem if one wants to configure an external PHY via phylink,
since it use the same phandle to get the PHY. To fix this, introduce a
dedicated pcs-handle to point to the PCS/PMA PHY and deprecate the use
of pointing it with phy-handle. A similar use case of pcs-handle can be
seen on dpaa2 as well.

--- patch v5 ---
 - Re-apply the v4 patch on the net tree.
 - Describe the pcs-handle DT binding at ethernet-controller level.
--- patch v6 ---
 - Remove "preferrably" to clearify usage of pcs_handle.
--- patch v7 ---
 - Rebase the patch on latest net/master
--- patch v8 ---
 - Rebase the patch on net-next/master
 - Add "reviewed-by" tag in PATCH 3/4: dt-bindings: net: add pcs-handle
   attribute
 - Remove "fix" tag in last commit message since this is not a critical
   bug and will not be back ported to stable.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: axiemac: use a phandle to reference pcs_phy
Andy Chiu [Tue, 5 Apr 2022 09:19:29 +0000 (17:19 +0800)]
net: axiemac: use a phandle to reference pcs_phy

In some SGMII use cases where both a fixed link external PHY and the
internal PCS/PMA PHY need to be configured, we should explicitly use a
phandle "pcs-phy" to get the reference to the PCS/PMA PHY. Otherwise, the
driver would use "phy-handle" in the DT as the reference to both the
external and the internal PCS/PMA PHY.

In other cases where the core is connected to a SFP cage, we could still
point phy-handle to the intenal PCS/PMA PHY, and let the driver connect
to the SFP module, if exist, via phylink.

Signed-off-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Robert Hancock <robert.hancock@calian.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodt-bindings: net: add pcs-handle attribute
Andy Chiu [Tue, 5 Apr 2022 09:19:28 +0000 (17:19 +0800)]
dt-bindings: net: add pcs-handle attribute

Document the new pcs-handle attribute to support connecting to an
external PHY. For Xilinx's AXI Ethernet, this is used when the core
operates in SGMII or 1000Base-X modes and links through the internal
PCS/PMA PHY.

Signed-off-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: axienet: factor out phy_node in struct axienet_local
Andy Chiu [Tue, 5 Apr 2022 09:19:27 +0000 (17:19 +0800)]
net: axienet: factor out phy_node in struct axienet_local

the struct member `phy_node` of struct axienet_local is not used by the
driver anymore after initialization. It might be a remnent of old code
and could be removed.

Signed-off-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Robert Hancock <robert.hancock@calian.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: axienet: setup mdio unconditionally
Andy Chiu [Tue, 5 Apr 2022 09:19:26 +0000 (17:19 +0800)]
net: axienet: setup mdio unconditionally

The call to axienet_mdio_setup should not depend on whether "phy-node"
pressents on the DT. Besides, since `lp->phy_node` is used if PHY is in
SGMII or 100Base-X modes, move it into the if statement. And the next patch
will remove `lp->phy_node` from driver's private structure and do an
of_node_put on it right away after use since it is not used elsewhere.

Signed-off-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Greentime Hu <greentime.hu@sifive.com>
Reviewed-by: Robert Hancock <robert.hancock@calian.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: sfc: fix using uninitialized xdp tx_queue
Taehee Yoo [Tue, 5 Apr 2022 08:45:44 +0000 (08:45 +0000)]
net: sfc: fix using uninitialized xdp tx_queue

In some cases, xdp tx_queue can get used before initialization.
1. interface up/down
2. ring buffer size change

When CPU cores are lower than maximum number of channels of sfc driver,
it creates new channels only for XDP.

When an interface is up or ring buffer size is changed, all channels
are initialized.
But xdp channels are always initialized later.
So, the below scenario is possible.
Packets are received to rx queue of normal channels and it is acted
XDP_TX and tx_queue of xdp channels get used.
But these tx_queues are not initialized yet.
If so, TX DMA or queue error occurs.

In order to avoid this problem.
1. initializes xdp tx_queues earlier than other rx_queue in
efx_start_channels().
2. checks whether tx_queue is initialized or not in efx_xdp_tx_buffers().

Splat looks like:
   sfc 0000:08:00.1 enp8s0f1np1: TX queue 10 spurious TX completion id 250
   sfc 0000:08:00.1 enp8s0f1np1: resetting (RECOVER_OR_ALL)
   sfc 0000:08:00.1 enp8s0f1np1: MC command 0x80 inlen 100 failed rc=-22
   (raw=22) arg=789
   sfc 0000:08:00.1 enp8s0f1np1: has been disabled

Fixes: f28100cb9c96 ("sfc: fix lack of XDP TX queues - error XDP TX failed (-22)")
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agorxrpc: fix a race in rxrpc_exit_net()
Eric Dumazet [Mon, 4 Apr 2022 18:34:39 +0000 (11:34 -0700)]
rxrpc: fix a race in rxrpc_exit_net()

Current code can lead to the following race:

CPU0                                                 CPU1

rxrpc_exit_net()
                                                     rxrpc_peer_keepalive_worker()
                                                       if (rxnet->live)

  rxnet->live = false;
  del_timer_sync(&rxnet->peer_keepalive_timer);

                                                             timer_reduce(&rxnet->peer_keepalive_timer, jiffies + delay);

  cancel_work_sync(&rxnet->peer_keepalive_work);

rxrpc_exit_net() exits while peer_keepalive_timer is still armed,
leading to use-after-free.

syzbot report was:

ODEBUG: free active (active state 0) object type: timer_list hint: rxrpc_peer_keepalive_timeout+0x0/0xb0
WARNING: CPU: 0 PID: 3660 at lib/debugobjects.c:505 debug_print_object+0x16e/0x250 lib/debugobjects.c:505
Modules linked in:
CPU: 0 PID: 3660 Comm: kworker/u4:6 Not tainted 5.17.0-syzkaller-13993-g88e6c0207623 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: netns cleanup_net
RIP: 0010:debug_print_object+0x16e/0x250 lib/debugobjects.c:505
Code: ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 0f 85 af 00 00 00 48 8b 14 dd 00 1c 26 8a 4c 89 ee 48 c7 c7 00 10 26 8a e8 b1 e7 28 05 <0f> 0b 83 05 15 eb c5 09 01 48 83 c4 18 5b 5d 41 5c 41 5d 41 5e c3
RSP: 0018:ffffc9000353fb00 EFLAGS: 00010082
RAX: 0000000000000000 RBX: 0000000000000003 RCX: 0000000000000000
RDX: ffff888029196140 RSI: ffffffff815efad8 RDI: fffff520006a7f52
RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000000
R10: ffffffff815ea4ae R11: 0000000000000000 R12: ffffffff89ce23e0
R13: ffffffff8a2614e0 R14: ffffffff816628c0 R15: dffffc0000000000
FS:  0000000000000000(0000) GS:ffff8880b9c00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fe1f2908924 CR3: 0000000043720000 CR4: 00000000003506f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 <TASK>
 __debug_check_no_obj_freed lib/debugobjects.c:992 [inline]
 debug_check_no_obj_freed+0x301/0x420 lib/debugobjects.c:1023
 kfree+0xd6/0x310 mm/slab.c:3809
 ops_free_list.part.0+0x119/0x370 net/core/net_namespace.c:176
 ops_free_list net/core/net_namespace.c:174 [inline]
 cleanup_net+0x591/0xb00 net/core/net_namespace.c:598
 process_one_work+0x996/0x1610 kernel/workqueue.c:2289
 worker_thread+0x665/0x1080 kernel/workqueue.c:2436
 kthread+0x2e9/0x3a0 kernel/kthread.c:376
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:298
 </TASK>

Fixes: ace45bec6d77 ("rxrpc: Fix firewall route keepalive")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Marc Dionne <marc.dionne@auristor.com>
Cc: linux-afs@lists.infradead.org
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: openvswitch: fix leak of nested actions
Ilya Maximets [Mon, 4 Apr 2022 15:43:45 +0000 (17:43 +0200)]
net: openvswitch: fix leak of nested actions

While parsing user-provided actions, openvswitch module may dynamically
allocate memory and store pointers in the internal copy of the actions.
So this memory has to be freed while destroying the actions.

Currently there are only two such actions: ct() and set().  However,
there are many actions that can hold nested lists of actions and
ovs_nla_free_flow_actions() just jumps over them leaking the memory.

For example, removal of the flow with the following actions will lead
to a leak of the memory allocated by nf_ct_tmpl_alloc():

  actions:clone(ct(commit),0)

Non-freed set() action may also leak the 'dst' structure for the
tunnel info including device references.

Under certain conditions with a high rate of flow rotation that may
cause significant memory leak problem (2MB per second in reporter's
case).  The problem is also hard to mitigate, because the user doesn't
have direct control over the datapath flows generated by OVS.

Fix that by iterating over all the nested actions and freeing
everything that needs to be freed recursively.

New build time assertion should protect us from this problem if new
actions will be added in the future.

Unfortunately, openvswitch module doesn't use NLA_F_NESTED, so all
attributes has to be explicitly checked.  sample() and clone() actions
are mixing extra attributes into the user-provided action list.  That
prevents some code generalization too.

Fixes: 34ae932a4036 ("openvswitch: Make tunnel set action attach a metadata dst")
Link: https://mail.openvswitch.org/pipermail/ovs-dev/2022-March/392922.html
Reported-by: Stéphane Graber <stgraber@ubuntu.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Acked-by: Aaron Conole <aconole@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoata: ahci: Rename CONFIG_SATA_LPM_POLICY configuration item back
Mario Limonciello [Wed, 6 Apr 2022 01:57:51 +0000 (10:57 +0900)]
ata: ahci: Rename CONFIG_SATA_LPM_POLICY configuration item back

CONFIG_SATA_LPM_MOBILE_POLICY was renamed to CONFIG_SATA_LPM_POLICY in
commit 4dd4d3deb502 ("ata: ahci: Rename CONFIG_SATA_LPM_MOBILE_POLICY
configuration item").

This can potentially cause problems as users would invisibly lose
configuration policy defaults when they built the new kernel. To
avoid such problems, switch back to the old name (even if it's wrong).

Suggested-by: Christoph Hellwig <hch@infradead.org>
Suggested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
2 years agonet: ethernet: mv643xx: Fix over zealous checking of_get_mac_address()
Andrew Lunn [Tue, 5 Apr 2022 00:04:04 +0000 (02:04 +0200)]
net: ethernet: mv643xx: Fix over zealous checking of_get_mac_address()

There is often not a MAC address available in an EEPROM accessible by
Linux with Marvell devices. Instead the bootload has the MAC address
and directly programs it into the hardware. So don't consider an error
from of_get_mac_address() has fatal. However, the check was added for
the case where there is a MAC address in an the EEPROM, but the EEPROM
has not probed yet, and -EPROBE_DEFER is returned. In that case the
error should be returned. So make the check specific to this error
code.

Cc: Mauri Sandberg <maukka@ext.kapsi.fi>
Reported-by: Thomas Walther <walther-it@gmx.de>
Fixes: 42404d8f1c01 ("net: mv643xx_eth: process retval from of_get_mac_address")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20220405000404.3374734-1-andrew@lunn.ch
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: openvswitch: don't send internal clone attribute to the userspace.
Ilya Maximets [Mon, 4 Apr 2022 10:41:50 +0000 (12:41 +0200)]
net: openvswitch: don't send internal clone attribute to the userspace.

'OVS_CLONE_ATTR_EXEC' is an internal attribute that is used for
performance optimization inside the kernel.  It's added by the kernel
while parsing user-provided actions and should not be sent during the
flow dump as it's not part of the uAPI.

The issue doesn't cause any significant problems to the ovs-vswitchd
process, because reported actions are not really used in the
application lifecycle and only supposed to be shown to a human via
ovs-dpctl flow dump.  However, the action list is still incorrect
and causes the following error if the user wants to look at the
datapath flows:

  # ovs-dpctl add-dp system@ovs-system
  # ovs-dpctl add-flow "<flow match>" "clone(ct(commit),0)"
  # ovs-dpctl dump-flows
  <flow match>, packets:0, bytes:0, used:never,
    actions:clone(bad length 4, expected -1 for: action0(01 00 00 00),
                  ct(commit),0)

With the fix:

  # ovs-dpctl dump-flows
  <flow match>, packets:0, bytes:0, used:never,
    actions:clone(ct(commit),0)

Additionally fixed an incorrect attribute name in the comment.

Fixes: b233504033db ("openvswitch: kernel datapath clone action")
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Acked-by: Aaron Conole <aconole@redhat.com>
Link: https://lore.kernel.org/r/20220404104150.2865736-1-i.maximets@ovn.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: micrel: Fix KS8851 Kconfig
Horatiu Vultur [Tue, 5 Apr 2022 06:59:36 +0000 (08:59 +0200)]
net: micrel: Fix KS8851 Kconfig

KS8851 selects MICREL_PHY, which depends on PTP_1588_CLOCK_OPTIONAL, so
make KS8851 also depend on PTP_1588_CLOCK_OPTIONAL.

Fixes kconfig warning and build errors:

WARNING: unmet direct dependencies detected for MICREL_PHY
  Depends on [m]: NETDEVICES [=y] && PHYLIB [=y] && PTP_1588_CLOCK_OPTIONAL [=m]
    Selected by [y]:
      - KS8851 [=y] && NETDEVICES [=y] && ETHERNET [=y] && NET_VENDOR_MICREL [=y] && SPI [=y]

ld.lld: error: undefined symbol: ptp_clock_register referenced by micrel.c
net/phy/micrel.o:(lan8814_probe) in archive drivers/built-in.a
ld.lld: error: undefined symbol: ptp_clock_index referenced by micrel.c
net/phy/micrel.o:(lan8814_ts_info) in archive drivers/built-in.a

Reported-by: kernel test robot <lkp@intel.com>
Fixes: ece19502834d ("net: phy: micrel: 1588 support for LAN8814 phy")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://lore.kernel.org/r/20220405065936.4105272-1-horatiu.vultur@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Jakub Kicinski [Tue, 5 Apr 2022 20:04:03 +0000 (13:04 -0700)]
Merge git://git./linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

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

1) Incorrect comparison in bitmask .reduce, from Jeremy Sowden.

2) Missing GFP_KERNEL_ACCOUNT for dynamically allocated objects,
   from Vasily Averin.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nf_tables: memcg accounting for dynamically allocated objects
  netfilter: bitwise: fix reduce comparisons
====================

Link: https://lore.kernel.org/r/20220405100923.7231-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Linus Torvalds [Tue, 5 Apr 2022 17:40:52 +0000 (10:40 -0700)]
Merge tag 'for_linus' of git://git./linux/kernel/git/mst/vhost

Pull virtio fixes from Michael Tsirkin:
 "Fixes and cleanups:

   - A couple of mlx5 fixes related to cvq

   - A couple of reverts dropping useless code (code that used it got
     reverted earlier)"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
  vdpa: mlx5: synchronize driver status with CVQ
  vdpa: mlx5: prevent cvq work from hogging CPU
  Revert "virtio_config: introduce a new .enable_cbs method"
  Revert "virtio: use virtio_device_ready() in virtio_device_restore()"

2 years agox86/speculation: Restore speculation related MSRs during S3 resume
Pawan Gupta [Tue, 5 Apr 2022 00:35:45 +0000 (17:35 -0700)]
x86/speculation: Restore speculation related MSRs during S3 resume

After resuming from suspend-to-RAM, the MSRs that control CPU's
speculative execution behavior are not being restored on the boot CPU.

These MSRs are used to mitigate speculative execution vulnerabilities.
Not restoring them correctly may leave the CPU vulnerable.  Secondary
CPU's MSRs are correctly being restored at S3 resume by
identify_secondary_cpu().

During S3 resume, restore these MSRs for boot CPU when restoring its
processor state.

Fixes: 772439717dbf ("x86/bugs/intel: Set proper CPU features and setup RDS")
Reported-by: Neelima Krishnan <neelima.krishnan@intel.com>
Signed-off-by: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>
Tested-by: Neelima Krishnan <neelima.krishnan@intel.com>
Acked-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agox86/pm: Save the MSR validity status at context setup
Pawan Gupta [Tue, 5 Apr 2022 00:34:19 +0000 (17:34 -0700)]
x86/pm: Save the MSR validity status at context setup

The mechanism to save/restore MSRs during S3 suspend/resume checks for
the MSR validity during suspend, and only restores the MSR if its a
valid MSR.  This is not optimal, as an invalid MSR will unnecessarily
throw an exception for every suspend cycle.  The more invalid MSRs,
higher the impact will be.

Check and save the MSR validity at setup.  This ensures that only valid
MSRs that are guaranteed to not throw an exception will be attempted
during suspend.

Fixes: 7a9c2dd08ead ("x86/pm: Introduce quirk framework to save/restore extra MSR registers around suspend/resume")
Suggested-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>
Reviewed-by: Dave Hansen <dave.hansen@linux.intel.com>
Acked-by: Borislav Petkov <bp@suse.de>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoice: clear cmd_type_offset_bsz for TX rings
Maciej Fijalkowski [Thu, 17 Mar 2022 18:36:29 +0000 (19:36 +0100)]
ice: clear cmd_type_offset_bsz for TX rings

Currently when XDP rings are created, each descriptor gets its DD bit
set, which turns out to be the wrong approach as it can lead to a
situation where more descriptors get cleaned than it was supposed to,
e.g. when AF_XDP busy poll is run with a large batch size. In this
situation, the driver would request for more buffers than it is able to
handle.

Fix this by not setting the DD bits in ice_xdp_alloc_setup_rings(). They
should be initialized to zero instead.

Fixes: 9610bd988df9 ("ice: optimize XDP_TX workloads")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Shwetha Nagaraju <shwetha.nagaraju@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: xsk: fix VSI state check in ice_xsk_wakeup()
Maciej Fijalkowski [Thu, 17 Mar 2022 18:36:28 +0000 (19:36 +0100)]
ice: xsk: fix VSI state check in ice_xsk_wakeup()

ICE_DOWN is dedicated for pf->state. Check for ICE_VSI_DOWN being set on
vsi->state in ice_xsk_wakeup().

Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Shwetha Nagaraju <shwetha.nagaraju@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: synchronize_rcu() when terminating rings
Maciej Fijalkowski [Thu, 17 Mar 2022 18:36:27 +0000 (19:36 +0100)]
ice: synchronize_rcu() when terminating rings

Unfortunately, the ice driver doesn't respect the RCU critical section that
XSK wakeup is surrounded with. To fix this, add synchronize_rcu() calls to
paths that destroy resources that might be in use.

This was addressed in other AF_XDP ZC enabled drivers, for reference see
for example commit b3873a5be757 ("net/i40e: Fix concurrency issues
between config flow and XSK")

Fixes: efc2214b6047 ("ice: Add support for XDP")
Fixes: 2d4238f55697 ("ice: Add support for AF_XDP")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Shwetha Nagaraju <shwetha.nagaraju@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoMerge tag 'for-5.18-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Tue, 5 Apr 2022 15:59:37 +0000 (08:59 -0700)]
Merge tag 'for-5.18-rc1-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - prevent deleting subvolume with active swapfile

 - fix qgroup reserve limit calculation overflow

 - remove device count in superblock and its item in one transaction so
   they cant't get out of sync

 - skip defragmenting an isolated sector, this could cause some extra IO

 - unify handling of mtime/permissions in hole punch with fallocate

 - zoned mode fixes:
     - remove assert checking for only single mode, we have the
       DUP mode implemented
     - fix potential lockdep warning while traversing devices
       when checking for zone activation

* tag 'for-5.18-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: prevent subvol with swapfile from being deleted
  btrfs: do not warn for free space inode in cow_file_range
  btrfs: avoid defragging extents whose next extents are not targets
  btrfs: fix fallocate to use file_modified to update permissions consistently
  btrfs: remove device item and update super block in the same transaction
  btrfs: fix qgroup reserve overflow the qgroup limit
  btrfs: zoned: remove left over ASSERT checking for single profile
  btrfs: zoned: traverse devices under chunk_mutex in btrfs_can_activate_zone

2 years agorandom: opportunistically initialize on /dev/urandom reads
Jason A. Donenfeld [Tue, 5 Apr 2022 13:57:05 +0000 (15:57 +0200)]
random: opportunistically initialize on /dev/urandom reads

In 6f98a4bfee72 ("random: block in /dev/urandom"), we tried to make a
successful try_to_generate_entropy() call *required* if the RNG was not
already initialized. Unfortunately, weird architectures and old
userspaces combined in TCG test harnesses, making that change still not
realistic, so it was reverted in 0313bc278dac ("Revert "random: block in
/dev/urandom"").

However, rather than making a successful try_to_generate_entropy() call
*required*, we can instead make it *best-effort*.

If try_to_generate_entropy() fails, it fails, and nothing changes from
the current behavior. If it succeeds, then /dev/urandom becomes safe to
use for free. This way, we don't risk the regression potential that led
to us reverting the required-try_to_generate_entropy() call before.

Practically speaking, this means that at least on x86, /dev/urandom
becomes safe. Probably other architectures with working cycle counters
will also become safe. And architectures with slow or broken cycle
counters at least won't be affected at all by this change.

So it may not be the glorious "all things are unified!" change we were
hoping for initially, but practically speaking, it makes a positive
impact.

Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2 years agoipv6: Fix stats accounting in ip6_pkt_drop
David Ahern [Mon, 4 Apr 2022 15:09:08 +0000 (09:09 -0600)]
ipv6: Fix stats accounting in ip6_pkt_drop

VRF devices are the loopbacks for VRFs, and a loopback can not be
assigned to a VRF. Accordingly, the condition in ip6_pkt_drop should
be '||' not '&&'.

Fixes: 1d3fd8a10bed ("vrf: Use orig netdev to count Ip6InNoRoutes and a fresh route lookup when sending dest unreach")
Reported-by: Pudak, Filip <Filip.Pudak@windriver.com>
Reported-by: Xiao, Jiguang <Jiguang.Xiao@windriver.com>
Signed-off-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20220404150908.2937-1-dsahern@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoMerge branch 'ice-bug-fixes'
Paolo Abeni [Tue, 5 Apr 2022 10:50:28 +0000 (12:50 +0200)]
Merge branch 'ice-bug-fixes'

Tony Nguyen says:

====================
ice bug fixes

Alice Michael says:

There were a couple of bugs that have been found and
fixed by Anatolii in the ice driver.  First he fixed
a bug on ring creation by setting the default value
for the teid.  Anatolli also fixed a bug with deleting
queues in ice_vc_dis_qs_msg based on their enablement.
---
v2: Remove empty lines between tags

The following are changes since commit 458f5d92df4807e2a7c803ed928369129996bf96:
  sfc: Do not free an empty page_ring
and are available in the git repository at:
  git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue 100GbE
====================

Link: https://lore.kernel.org/r/20220404183548.3422851-1-anthony.l.nguyen@intel.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoice: Do not skip not enabled queues in ice_vc_dis_qs_msg
Anatolii Gerasymenko [Mon, 4 Apr 2022 18:35:48 +0000 (11:35 -0700)]
ice: Do not skip not enabled queues in ice_vc_dis_qs_msg

Disable check for queue being enabled in ice_vc_dis_qs_msg, because
there could be a case when queues were created, but were not enabled.
We still need to delete those queues.

Normal workflow for VF looks like:
Enable path:
VIRTCHNL_OP_ADD_ETH_ADDR (opcode 10)
VIRTCHNL_OP_CONFIG_VSI_QUEUES (opcode 6)
VIRTCHNL_OP_ENABLE_QUEUES (opcode 8)

Disable path:
VIRTCHNL_OP_DISABLE_QUEUES (opcode 9)
VIRTCHNL_OP_DEL_ETH_ADDR (opcode 11)

The issue appears only in stress conditions when VF is enabled and
disabled very fast.
Eventually there will be a case, when queues are created by
VIRTCHNL_OP_CONFIG_VSI_QUEUES, but are not enabled by
VIRTCHNL_OP_ENABLE_QUEUES.
In turn, these queues are not deleted by VIRTCHNL_OP_DISABLE_QUEUES,
because there is a check whether queues are enabled in
ice_vc_dis_qs_msg.

When we bring up the VF again, we will see the "Failed to set LAN Tx queue
context" error during VIRTCHNL_OP_CONFIG_VSI_QUEUES step. This
happens because old 16 queues were not deleted and VF requests to create
16 more, but ice_sched_get_free_qparent in ice_ena_vsi_txq would fail to
find a parent node for first newly requested queue (because all nodes
are allocated to 16 old queues).

Testing Hints:

Just enable and disable VF fast enough, so it would be disabled before
reaching VIRTCHNL_OP_ENABLE_QUEUES.

while true; do
        ip link set dev ens785f0v0 up
        sleep 0.065 # adjust delay value for you machine
        ip link set dev ens785f0v0 down
done

Fixes: 77ca27c41705 ("ice: add support for virtchnl_queue_select.[tx|rx]_queues bitmap")
Signed-off-by: Anatolii Gerasymenko <anatolii.gerasymenko@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Alice Michael <alice.michael@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoice: Set txq_teid to ICE_INVAL_TEID on ring creation
Anatolii Gerasymenko [Mon, 4 Apr 2022 18:35:47 +0000 (11:35 -0700)]
ice: Set txq_teid to ICE_INVAL_TEID on ring creation

When VF is freshly created, but not brought up, ring->txq_teid
value is by default set to 0.
But 0 is a valid TEID. On some platforms the Root Node of
Tx scheduler has a TEID = 0. This can cause issues as shown below.

The proper way is to set ring->txq_teid to ICE_INVAL_TEID (0xFFFFFFFF).

Testing Hints:
echo 1 > /sys/class/net/ens785f0/device/sriov_numvfs
ip link set dev ens785f0v0 up
ip link set dev ens785f0v0 down

If we have freshly created VF and quickly turn it on and off, so there
would be no time to reach VIRTCHNL_OP_CONFIG_VSI_QUEUES stage, then
VIRTCHNL_OP_DISABLE_QUEUES stage will fail with error:
[  639.531454] disable queue 89 failed 14
[  639.532233] Failed to disable LAN Tx queues, error: ICE_ERR_AQ_ERROR
[  639.533107] ice 0000:02:00.0: Failed to stop Tx ring 0 on VSI 5

The reason for the fail is that we are trying to send AQ command to
delete queue 89, which has never been created and receive an "invalid
argument" error from firmware.

As this queue has never been created, it's teid and ring->txq_teid
have default value 0.
ice_dis_vsi_txq has a check against non-existent queues:

node = ice_sched_find_node_by_teid(pi->root, q_teids[i]);
if (!node)
continue;

But on some platforms the Root Node of Tx scheduler has a teid = 0.
Hence, ice_sched_find_node_by_teid finds a node with teid = 0 (it is
pi->root), and we go further to submit an erroneous request to firmware.

Fixes: 37bb83901286 ("ice: Move common functions out of ice_main.c part 7/7")
Signed-off-by: Anatolii Gerasymenko <anatolii.gerasymenko@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Alice Michael <alice.michael@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agodpaa2-ptp: Fix refcount leak in dpaa2_ptp_probe
Miaoqian Lin [Mon, 4 Apr 2022 12:53:36 +0000 (12:53 +0000)]
dpaa2-ptp: Fix refcount leak in dpaa2_ptp_probe

This node pointer is returned by of_find_compatible_node() with
refcount incremented. Calling of_node_put() to aovid the refcount leak.

Fixes: d346c9e86d86 ("dpaa2-ptp: reuse ptp_qoriq driver")
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Link: https://lore.kernel.org/r/20220404125336.13427-1-linmq006@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonetfilter: nf_tables: memcg accounting for dynamically allocated objects
Vasily Averin [Sat, 2 Apr 2022 09:50:37 +0000 (12:50 +0300)]
netfilter: nf_tables: memcg accounting for dynamically allocated objects

nft_*.c files whose NFT_EXPR_STATEFUL flag is set on need to
use __GFP_ACCOUNT flag for objects that are dynamically
allocated from the packet path.

Such objects are allocated inside nft_expr_ops->init() callbacks
executed in task context while processing netlink messages.

In addition, this patch adds accounting to nft_set_elem_expr_clone()
used for the same purposes.

Signed-off-by: Vasily Averin <vvs@openvz.org>
Acked-by: Roman Gushchin <roman.gushchin@linux.dev>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
2 years agosctp: count singleton chunks in assoc user stats
Jamie Bainbridge [Sun, 3 Apr 2022 23:47:48 +0000 (09:47 +1000)]
sctp: count singleton chunks in assoc user stats

Singleton chunks (INIT, HEARTBEAT PMTU probes, and SHUTDOWN-
COMPLETE) are not counted in SCTP_GET_ASOC_STATS "sas_octrlchunks"
counter available to the assoc owner.

These are all control chunks so they should be counted as such.

Add counting of singleton chunks so they are properly accounted for.

Fixes: 196d67593439 ("sctp: Add support to per-association statistics via a new SCTP_GET_ASSOC_STATS call")
Signed-off-by: Jamie Bainbridge <jamie.bainbridge@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Link: https://lore.kernel.org/r/c9ba8785789880cf07923b8a5051e174442ea9ee.1649029663.git.jamie.bainbridge@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agorandom: do not split fast init input in add_hwgenerator_randomness()
Jan Varho [Mon, 4 Apr 2022 16:42:30 +0000 (19:42 +0300)]
random: do not split fast init input in add_hwgenerator_randomness()

add_hwgenerator_randomness() tries to only use the required amount of input
for fast init, but credits all the entropy, rather than a fraction of
it. Since it's hard to determine how much entropy is left over out of a
non-unformly random sample, either give it all to fast init or credit
it, but don't attempt to do both. In the process, we can clean up the
injection code to no longer need to return a value.

Signed-off-by: Jan Varho <jan.varho@gmail.com>
[Jason: expanded commit message]
Fixes: 73c7733f122e ("random: do not throw away excess input to crng_fast_load")
Cc: stable@vger.kernel.org # 5.17+, requires af704c856e88
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2 years agosfc: Do not free an empty page_ring
Martin Habets [Mon, 4 Apr 2022 10:48:51 +0000 (11:48 +0100)]
sfc: Do not free an empty page_ring

When the page_ring is not used page_ptr_mask is 0.
Do not dereference page_ring[0] in this case.

Fixes: 2768935a4660 ("sfc: reuse pages to avoid DMA mapping/unmapping costs")
Reported-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agostmmac: dwmac-loongson: change loongson_dwmac_driver from global to static
Tom Rix [Sun, 3 Apr 2022 14:02:02 +0000 (10:02 -0400)]
stmmac: dwmac-loongson: change loongson_dwmac_driver from global to static

Smatch reports this issue
dwmac-loongson.c:208:19: warning: symbol
  'loongson_dwmac_driver' was not declared.
  Should it be static?

loongson_dwmac_driver is only used in dwmac-loongson.c.
File scope variables used only in one file should
be static. Change loongson_dwmac_driver's
storage-class-specifier from global to static.

Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'bnxt_en-fixes'
David S. Miller [Mon, 4 Apr 2022 11:44:50 +0000 (12:44 +0100)]
Merge branch 'bnxt_en-fixes'

Michael Chan says:

====================
bnxt_en: XDP redirect fixes

This series includes 3 fixes related to the XDP redirect code path in
the driver.  The first one adds locking when the number of TX XDP rings
is less than the number of CPUs.  The second one adjusts the maximum MTU
that can support XDP with enough tail room in the buffer.  The 3rd one
fixes a race condition between TX ring shutdown and the XDP redirect path.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobnxt_en: Prevent XDP redirect from running when stopping TX queue
Ray Jui [Sat, 2 Apr 2022 00:21:12 +0000 (20:21 -0400)]
bnxt_en: Prevent XDP redirect from running when stopping TX queue

Add checks in the XDP redirect callback to prevent XDP from running when
the TX ring is undergoing shutdown.

Also remove redundant checks in the XDP redirect callback to validate the
txr and the flag that indicates the ring supports XDP. The modulo
arithmetic on 'tx_nr_rings_xdp' already guarantees the derived TX
ring is an XDP ring.  txr is also guaranteed to be valid after checking
BNXT_STATE_OPEN and within RCU grace period.

Fixes: f18c2b77b2e4 ("bnxt_en: optimized XDP_REDIRECT support")
Reviewed-by: Vladimir Olovyannikov <vladimir.olovyannikov@broadcom.com>
Signed-off-by: Ray Jui <ray.jui@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobnxt_en: reserve space inside receive page for skb_shared_info
Andy Gospodarek [Sat, 2 Apr 2022 00:21:11 +0000 (20:21 -0400)]
bnxt_en: reserve space inside receive page for skb_shared_info

Insufficient space was being reserved in the page used for packet
reception, so the interface MTU could be set too large to still have
room for the contents of the packet when doing XDP redirect.  This
resulted in the following message when redirecting a packet between
3520 and 3822 bytes with an MTU of 3822:

[311815.561880] XDP_WARN: xdp_update_frame_from_buff(line:200): Driver BUG: missing reserved tailroom

Fixes: f18c2b77b2e4 ("bnxt_en: optimized XDP_REDIRECT support")
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobnxt_en: Synchronize tx when xdp redirects happen on same ring
Pavan Chebbi [Sat, 2 Apr 2022 00:21:10 +0000 (20:21 -0400)]
bnxt_en: Synchronize tx when xdp redirects happen on same ring

If there are more CPUs than the number of TX XDP rings, multiple XDP
redirects can select the same TX ring based on the CPU on which
XDP redirect is called.  Add locking when needed and use static
key to decide whether to take the lock.

Fixes: f18c2b77b2e4 ("bnxt_en: optimized XDP_REDIRECT support")
Signed-off-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>
2 years agoqed: fix ethtool register dump
Manish Chopra [Fri, 1 Apr 2022 18:53:04 +0000 (11:53 -0700)]
qed: fix ethtool register dump

To fix a coverity complain, commit d5ac07dfbd2b
("qed: Initialize debug string array") removed "sw-platform"
(one of the common global parameters) from the dump as this
was used in the dump with an uninitialized string, however
it did not reduce the number of common global parameters
which caused the incorrect (unable to parse) register dump

this patch fixes it with reducing NUM_COMMON_GLOBAL_PARAMS
bye one.

Cc: stable@vger.kernel.org
Cc: Tim Gardner <tim.gardner@canonical.com>
Cc: "David S. Miller" <davem@davemloft.net>
Fixes: d5ac07dfbd2b ("qed: Initialize debug string array")
Signed-off-by: Prabhakar Kushwaha <pkushwaha@marvell.com>
Signed-off-by: Alok Prasad <palok@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Manish Chopra <manishc@marvell.com>
Reviewed-by: Tim Gardner <tim.gardner@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'micrel-lan8814-remove-latencies'
David S. Miller [Mon, 4 Apr 2022 11:40:42 +0000 (12:40 +0100)]
Merge branch 'micrel-lan8814-remove-latencies'

Horatiu Vultur says:

====================
net: phy: micrel: Remove latencies support lan8814

Remove the latencies support both from the PHY driver and from the DT.
The IP already has some default latencies values which can be used to get
decent results. It has the following values(defined in ns):
rx-1000mbit: 429
tx-1000mbit: 201
rx-100mbit:  2346
tx-100mbit:  705

v0->v1:
- fix the split of the patches, there was a compiling error between patch 2 and
  patch 3.

---
But to get better results the following values needs to be set:
rx-1000mbit: 459
tx-1000mbit: 171
rx-100mbit:  1706
tx-100mbit:  1345

We are proposing to use ethtool to set these latencies, the RFC can be found
here[1]
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: phy: micrel: Remove DT option lan8814,ignore-ts
Horatiu Vultur [Fri, 1 Apr 2022 11:05:22 +0000 (13:05 +0200)]
net: phy: micrel: Remove DT option lan8814,ignore-ts

When the PHY and the MAC are capable of doing timestamping, the PHY has
priority. Therefore the DT option lan8814,ignore-ts was added such that
the PHY will not expose a PHC so then the timestamping was done in the
MAC. This is not the correct approach of doing it, therefore remove
this.

Fixes: ece19502834d84 ("net: phy: micrel: 1588 support for LAN8814 phy")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: phy: micrel: Remove latency from driver
Horatiu Vultur [Fri, 1 Apr 2022 11:05:21 +0000 (13:05 +0200)]
net: phy: micrel: Remove latency from driver

Based on the discussions here[1], the PHY driver is the wrong place
to set the latencies, therefore remove them.

[1] https://lkml.org/lkml/2022/3/4/325

Fixes: ece19502834d84 ("net: phy: micrel: 1588 support for LAN8814 phy")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodt-bindings: net: micrel: Revert latency support and timestamping check
Horatiu Vultur [Fri, 1 Apr 2022 11:05:20 +0000 (13:05 +0200)]
dt-bindings: net: micrel: Revert latency support and timestamping check

Revert latency support from binding.
Based on the discussion[1], the DT is the wrong place to have the
lantecies for the PHY.

[1] https://lkml.org/lkml/2022/3/4/325

Fixes: 2358dd3fd325fc ("dt-bindings: net: micrel: Configure latency values and timestamping check for LAN8814 phy")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoata: libata-core: Disable READ LOG DMA EXT for Samsung 840 EVOs
Christian Lamparter [Sat, 19 Mar 2022 20:11:03 +0000 (21:11 +0100)]
ata: libata-core: Disable READ LOG DMA EXT for Samsung 840 EVOs

Samsung' 840 EVO with the latest firmware (EXT0DB6Q) locks up with
the a message: "READ LOG DMA EXT failed, trying PIO" during boot.

Initially this was discovered because it caused a crash
with the sata_dwc_460ex controller on a WD MyBook Live DUO.

The reporter "Tice Rex" which has the unique opportunity that he
has two Samsung 840 EVO SSD! One with the older firmware "EXT0BB0Q"
which booted fine and didn't expose "READ LOG DMA EXT". But the
newer/latest firmware "EXT0DB6Q" caused the headaches.

BugLink: https://github.com/openwrt/openwrt/issues/9505
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
2 years agoata: sata_dwc_460ex: Fix crash due to OOB write
Christian Lamparter [Sat, 19 Mar 2022 20:11:02 +0000 (21:11 +0100)]
ata: sata_dwc_460ex: Fix crash due to OOB write

the driver uses libata's "tag" values from in various arrays.
Since the mentioned patch bumped the ATA_TAG_INTERNAL to 32,
the value of the SATA_DWC_QCMD_MAX needs to account for that.

Otherwise ATA_TAG_INTERNAL usage cause similar crashes like
this as reported by Tice Rex on the OpenWrt Forum and
reproduced (with symbols) here:

| BUG: Kernel NULL pointer dereference at 0x00000000
| Faulting instruction address: 0xc03ed4b8
| Oops: Kernel access of bad area, sig: 11 [#1]
| BE PAGE_SIZE=4K PowerPC 44x Platform
| CPU: 0 PID: 362 Comm: scsi_eh_1 Not tainted 5.4.163 #0
| NIP:  c03ed4b8 LR: c03d27e8 CTR: c03ed36c
| REGS: cfa59950 TRAP: 0300   Not tainted  (5.4.163)
| MSR:  00021000 <CE,ME>  CR: 42000222  XER: 00000000
| DEAR: 00000000 ESR: 00000000
| GPR00: c03d27e8 cfa59a08 cfa55fe0 00000000 0fa46bc0 [...]
| [..]
| NIP [c03ed4b8] sata_dwc_qc_issue+0x14c/0x254
| LR [c03d27e8] ata_qc_issue+0x1c8/0x2dc
| Call Trace:
| [cfa59a08] [c003f4e0] __cancel_work_timer+0x124/0x194 (unreliable)
| [cfa59a78] [c03d27e8] ata_qc_issue+0x1c8/0x2dc
| [cfa59a98] [c03d2b3c] ata_exec_internal_sg+0x240/0x524
| [cfa59b08] [c03d2e98] ata_exec_internal+0x78/0xe0
| [cfa59b58] [c03d30fc] ata_read_log_page.part.38+0x1dc/0x204
| [cfa59bc8] [c03d324c] ata_identify_page_supported+0x68/0x130
| [...]

This is because sata_dwc_dma_xfer_complete() NULLs the
dma_pending's next neighbour "chan" (a *dma_chan struct) in
this '32' case right here (line ~735):
> hsdevp->dma_pending[tag] = SATA_DWC_DMA_PENDING_NONE;

Then the next time, a dma gets issued; dma_dwc_xfer_setup() passes
the NULL'd hsdevp->chan to the dmaengine_slave_config() which then
causes the crash.

With this patch, SATA_DWC_QCMD_MAX is now set to ATA_MAX_QUEUE + 1.
This avoids the OOB. But please note, there was a worthwhile discussion
on what ATA_TAG_INTERNAL and ATA_MAX_QUEUE is. And why there should not
be a "fake" 33 command-long queue size.

Ideally, the dw driver should account for the ATA_TAG_INTERNAL.
In Damien Le Moal's words: "... having looked at the driver, it
is a bigger change than just faking a 33rd "tag" that is in fact
not a command tag at all."

Fixes: 28361c403683c ("libata: add extra internal command")
Cc: stable@kernel.org # 4.18+
BugLink: https://github.com/openwrt/openwrt/issues/9505
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
2 years agoata: libata-sff: Fix compilation warning in ata_sff_lost_interrupt()
Damien Le Moal [Tue, 22 Mar 2022 07:01:37 +0000 (16:01 +0900)]
ata: libata-sff: Fix compilation warning in ata_sff_lost_interrupt()

When returning false, ata_sff_altstatus() does not return any status
value, resulting in a compilation warning in ata_sff_lost_interrupt()
("uninitialized symbol 'status'"). Fix this by initializing the local
variable "status" to 0.

Fixes: 03c0e84f9c1e ("ata: libata-sff: refactor ata_sff_altstatus()")
Cc: stable@vger.kernel.org
Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>