net: dsa: bcm_sf2: Fix bcm_sf2_reg_rgmii_cntrl() call for non-RGMII port

We cannot call bcm_sf2_reg_rgmii_cntrl() for a port that is not RGMII,
yet we do that in bcm_sf2_sw_mac_link_up() irrespective of the port's
interface. Move that read until we have properly qualified the PHY
interface mode. This avoids triggering a warning on 7278 platforms that
have GMII ports.

Fixes: 55cfeb396965 ("net: dsa: bcm_sf2: add function finding RGMII register")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

MAINTAINERS: s390/net: add netdev list

Discussions for network-related code should include the netdev list.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethernet: mtk_eth_soc: Fix DIM support for MT7628/88

When updating to latest mainline for some testing on the GARDENA smart
gateway based on the MT7628, I noticed that ethernet does not work any
more. Commit e9229ffd550b ("net: ethernet: mtk_eth_soc: implement
dynamic interrupt moderation") introduced this problem, as it missed the
RX_DIM & TX_DIM configuration for this SoC variant. This patch fixes
this by calling mtk_dim_rx() & mtk_dim_tx() in this case as well.

Signed-off-by: Stefan Roese <sr@denx.de>
Fixes: e9229ffd550b ("net: ethernet: mtk_eth_soc: implement dynamic interrupt moderation")
Cc: Felix Fietkau <nbd@nbd.name>
Cc: John Crispin <john@phrozen.org>
Cc: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
Cc: Reto Schneider <code@reto-schneider.ch>
Cc: Reto Schneider <reto.schneider@husqvarnagroup.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'stmmac-fixes'

Joakim Zhang says:

====================
net: fixes for stmmac

Two clock fixes for stmmac driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: fix system hang if change mac address after interface ifdown

Fix system hang with below sequences:
~# ifconfig ethx down
~# ifconfig ethx hw ether xx:xx:xx:xx:xx:xx

After ethx down, stmmac all clocks gated off and then register access causes
system hang.

Fixes: 5ec55823438e ("net: stmmac: add clocks management for gmac driver")
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: correct clocks enabled in stmmac_vlan_rx_kill_vid()

This should be a mistake to fix conflicts when removing RFC tag to
repost the patch.

Fixes: 5ec55823438e ("net: stmmac: add clocks management for gmac driver")
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/qla3xxx: fix schedule while atomic in ql_sem_spinlock

When calling the 'ql_sem_spinlock', the driver has already acquired the
spin lock, so the driver should not call 'ssleep' in atomic context.

This bug can be fixed by using 'mdelay' instead of 'ssleep'.

The KASAN's log reveals it:

[    3.238124 ] BUG: scheduling while atomic: swapper/0/1/0x00000002
[    3.238748 ] 2 locks held by swapper/0/1:
[    3.239151 ]  #0: ffff88810177b240 (&dev->mutex){....}-{3:3}, at:
__device_driver_lock+0x41/0x60
[    3.240026 ]  #1: ffff888107c60e28 (&qdev->hw_lock){....}-{2:2}, at:
ql3xxx_probe+0x2aa/0xea0
[    3.240873 ] Modules linked in:
[    3.241187 ] irq event stamp: 460854
[    3.241541 ] hardirqs last  enabled at (460853): [<ffffffff843051bf>]
_raw_spin_unlock_irqrestore+0x4f/0x70
[    3.242245 ] hardirqs last disabled at (460854): [<ffffffff843058ca>]
_raw_spin_lock_irqsave+0x2a/0x70
[    3.242245 ] softirqs last  enabled at (446076): [<ffffffff846002e4>]
__do_softirq+0x2e4/0x4b1
[    3.242245 ] softirqs last disabled at (446069): [<ffffffff811ba5e0>]
irq_exit_rcu+0x100/0x110
[    3.242245 ] Preemption disabled at:
[    3.242245 ] [<ffffffff828ca5ba>] ql3xxx_probe+0x2aa/0xea0
[    3.242245 ] Kernel panic - not syncing: scheduling while atomic
[    3.242245 ] CPU: 2 PID: 1 Comm: swapper/0 Not tainted
5.13.0-rc1-00145
-gee7dc339169-dirty #16
[    3.242245 ] Call Trace:
[    3.242245 ]  dump_stack+0xba/0xf5
[    3.242245 ]  ? ql3xxx_probe+0x1f0/0xea0
[    3.242245 ]  panic+0x15a/0x3f2
[    3.242245 ]  ? vprintk+0x76/0x150
[    3.242245 ]  ? ql3xxx_probe+0x2aa/0xea0
[    3.242245 ]  __schedule_bug+0xae/0xe0
[    3.242245 ]  __schedule+0x72e/0xa00
[    3.242245 ]  schedule+0x43/0xf0
[    3.242245 ]  schedule_timeout+0x28b/0x500
[    3.242245 ]  ? del_timer_sync+0xf0/0xf0
[    3.242245 ]  ? msleep+0x2f/0x70
[    3.242245 ]  msleep+0x59/0x70
[    3.242245 ]  ql3xxx_probe+0x307/0xea0
[    3.242245 ]  ? _raw_spin_unlock_irqrestore+0x3a/0x70
[    3.242245 ]  ? pci_device_remove+0x110/0x110
[    3.242245 ]  local_pci_probe+0x45/0xa0
[    3.242245 ]  pci_device_probe+0x12b/0x1d0
[    3.242245 ]  really_probe+0x2a9/0x610
[    3.242245 ]  driver_probe_device+0x90/0x1d0
[    3.242245 ]  ? mutex_lock_nested+0x1b/0x20
[    3.242245 ]  device_driver_attach+0x68/0x70
[    3.242245 ]  __driver_attach+0x124/0x1b0
[    3.242245 ]  ? device_driver_attach+0x70/0x70
[    3.242245 ]  bus_for_each_dev+0xbb/0x110
[    3.242245 ]  ? rdinit_setup+0x45/0x45
[    3.242245 ]  driver_attach+0x27/0x30
[    3.242245 ]  bus_add_driver+0x1eb/0x2a0
[    3.242245 ]  driver_register+0xa9/0x180
[    3.242245 ]  __pci_register_driver+0x82/0x90
[    3.242245 ]  ? yellowfin_init+0x25/0x25
[    3.242245 ]  ql3xxx_driver_init+0x23/0x25
[    3.242245 ]  do_one_initcall+0x7f/0x3d0
[    3.242245 ]  ? rdinit_setup+0x45/0x45
[    3.242245 ]  ? rcu_read_lock_sched_held+0x4f/0x80
[    3.242245 ]  kernel_init_freeable+0x2aa/0x301
[    3.242245 ]  ? rest_init+0x2c0/0x2c0
[    3.242245 ]  kernel_init+0x18/0x190
[    3.242245 ]  ? rest_init+0x2c0/0x2c0
[    3.242245 ]  ? rest_init+0x2c0/0x2c0
[    3.242245 ]  ret_from_fork+0x1f/0x30
[    3.242245 ] Dumping ftrace buffer:
[    3.242245 ]    (ftrace buffer empty)
[    3.242245 ] Kernel Offset: disabled
[    3.242245 ] Rebooting in 1 seconds.

Reported-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: encx24j600: fix kernel-doc syntax in file headers

The opening comment mark '/**' is used for highlighting the beginning of
kernel-doc comments.
The header for drivers/net/ethernet/microchip/encx24j600 files follows
this syntax, but the content inside does not comply with kernel-doc.

This line was probably not meant for kernel-doc parsing, but is parsed
due to the presence of kernel-doc like comment syntax(i.e, '/**'), which
causes unexpected warning from kernel-doc.
For e.g., running scripts/kernel-doc -none
drivers/net/ethernet/microchip/encx24j600_hw.h emits:
warning: expecting prototype for h(). Prototype was for _ENCX24J600_HW_H() instead

Provide a simple fix by replacing such occurrences with general comment
format, i.e. '/*', to prevent kernel-doc from parsing it.

Signed-off-by: Aditya Srivastava <yashsri421@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: fix large MTU request from VF

Check that the MTU value requested by the VF is in the supported
range of MTUs before attempting to set the VF large packet enable,
otherwise reject the request. This also avoids unnecessary
register updates in the case of the 82599 controller.

Fixes: 872844ddb9e4 ("ixgbe: Enable jumbo frames support w/ SR-IOV")
Co-developed-by: Piotr Skajewski <piotrx.skajewski@intel.com>
Signed-off-by: Piotr Skajewski <piotrx.skajewski@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Co-developed-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
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>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: Add .gitignore for nci test suite

Building the nci test suite produces a binary, nci_dev, that git then
tries to track. Add a .gitignore file to tell git to ignore this binary.

Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hso: bail out on interrupt URB allocation failure

Commit 31db0dbd7244 ("net: hso: check for allocation failure in
hso_create_bulk_serial_device()") recently started returning an error
when the driver fails to allocate resources for the interrupt endpoint
and tiocmget functionality.

For consistency let's bail out from probe also if the URB allocation
fails.

Signed-off-by: Johan Hovold <johan@kernel.org>
Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-bindings: net: renesas,ether: Update Sergei's email address

Update Sergei's email address, as per commit 534a8bf0ccdd7b3f
("MAINTAINERS: switch to my private email for Renesas Ethernet
drivers").

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Sergei Shtylyov <sergei.shtylyov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: avoid accessing registers when clearing filters

Hardware register having the server TID base can contain
invalid values when adapter is in bad state (for example,
due to AER fatal error). Reading these invalid values in the
register can lead to out-of-bound memory access. So, fix
by using the saved server TID base when clearing filters.

Fixes: b1a79360ee86 ("cxgb4: Delete all hash and TCAM filters before resource cleanup")
Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'mlx5-fixes-2021-05-18' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5 fixes 2021-05-18

This series introduces some fixes to mlx5 driver.

Please pull and let me know if there is any problem.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

ethtool: stats: Fix a copy-paste error

data->ctrl_stats should be memset with correct size.

Fixes: bfad2b979ddc ("ethtool: add interface to read standard MAC Ctrl stats")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlx5e: add add missing BH locking around napi_schdule()

It's not correct to call napi_schedule() in pure process
context. Because we use __raise_softirq_irqoff() we require
callers to be in a context which will eventually lead to
softirq handling (hardirq, bh disabled, etc.).

With code as is users will see:

 NOHZ tick-stop error: Non-RCU local softirq work is pending, handler #08!!!

Fixes: a8dd7ac12fc3 ("net/mlx5e: Generalize RQ activation")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: Set term table as an unmanaged flow table

Termination tables are restricted to have the default miss action and
cannot be set to forward to another table in case of a miss.
If the fs prio of the termination table is not the last one in the
list, fs_core will attempt to attach it to another table.

Set the unmanaged ft flag when creating the termination table ft
and select the tc offload prio for it to prevent fs_core from selecting
the forwarding to next ft miss action and use the default one.

In addition, set the flow that forwards to the termination table to
ignore ft level restrictions since the ft level is not set by fs_core
for unamanged fts.

Fixes: 249ccc3c95bd ("net/mlx5e: Add support for offloading traffic from uplink to uplink")
Signed-off-by: Ariel Levkovich <lariel@nvidia.com>

net/mlx5: Don't overwrite HCA capabilities when setting MSI-X count

During driver probe of device that has dynamic MSI-X feature enabled,
the following error is printed in some FW flavour (not released yet).

 mlx5_core 0000:06:00.0: firmware version: 4.7.4387
 mlx5_core 0000:06:00.0: 126.016 Gb/s available PCIe bandwidth (8.0 GT/s PCIe x16 link)
 mlx5_core 0000:06:00.0: mlx5_cmd_check:777:(pid 70599): SET_HCA_CAP(0x109) op_mod(0x0) failed, status bad parameter(0x3), syndrome (0x0)
 mlx5_core 0000:06:00.0: set_hca_cap:622:(pid 70599): handle_hca_cap failed
 mlx5_core 0000:06:00.0: mlx5_function_setup:1045:(pid 70599): set_hca_cap failed
 mlx5_core 0000:06:00.0: probe_one:1465:(pid 70599): mlx5_init_one failed with error code -22
 mlx5_core: probe of 0000:06:00.0 failed with error -22

In order to make the setting capability of MSI-X future proof, let's
query the current capabilities first.

Fixes: 604774add516 ("net/mlx5: Dynamically assign MSI-X vectors count")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

{net,vdpa}/mlx5: Configure interface MAC into mpfs L2 table

net/mlx5: Expose MPFS configuration API

MPFS is the multi physical function switch that bridges traffic between
the physical port and any physical functions associated with it. The
driver is required to add or remove MAC entries to properly forward
incoming traffic to the correct physical function.

We export the API to control MPFS so that other drivers, such as
mlx5_vdpa are able to add MAC addresses of their network interfaces.

The MAC address of the vdpa interface must be configured into the MPFS L2
address. Failing to do so could cause, in some NIC configurations, failure
to forward packets to the vdpa network device instance.

Fix this by adding calls to update the MPFS table.

CC: <mst@redhat.com>
CC: <jasowang@redhat.com>
CC: <virtualization@lists.linux-foundation.org>
Fixes: 1a86b377aa21 ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix error path of updating netdev queues

Avoid division by zero in the error flow. In the driver TC number can be
either 1 or 8. When TC count is set to 1, driver zero netdev->num_tc.
Hence, need to convert it back from 0 to 1 in the error flow.

Fixes: fa3748775b92 ("net/mlx5e: Handle errors from netif_set_real_num_{tx,rx}_queues")
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Reject mirroring on source port change encap rules

Rules with MLX5_ESW_DEST_CHAIN_WITH_SRC_PORT_CHANGE dest flag are
translated to destination FT in eswitch. Currently it is not possible to
mirror such rules because firmware doesn't support mixing FT and Vport
destinations in single rule when one of them adds encapsulation. Since the
only use case for MLX5_ESW_DEST_CHAIN_WITH_SRC_PORT_CHANGE destination is
support for tunnel endpoints on VF and trying to offload such rule with
mirror action causes either crash in fs_core or firmware error with
syndrome 0xff6a1d, reject all such rules in mlx5 TC layer.

Fixes: 10742efc20a4 ("net/mlx5e: VF tunnel TX traffic offloading")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix multipath lag activation

When handling FIB_EVENT_ENTRY_REPLACE event for a new multipath route,
lag activation can be missed if a stale (struct lag_mp)->mfi pointer
exists, which was associated with an older multipath route that had been
removed.

Normally, when a route is removed, it triggers mlx5_lag_fib_event(),
which handles FIB_EVENT_ENTRY_DEL and clears mfi pointer. But, if
mlx5_lag_check_prereq() condition isn't met, for example when eswitch is
in legacy mode, the fib event is skipped and mfi pointer becomes stale.

Fix by resetting mfi pointer to NULL every time mlx5_lag_mp_init() is
called.

Fixes: 544fe7c2e654 ("net/mlx5e: Activate HW multipath and handle port affinity based on FIB events")
Signed-off-by: Dima Chumak <dchumak@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: reset XPS on error flow if netdev isn't registered yet

mlx5e_attach_netdev can be called prior to registering the netdevice:
Example stack:

ipoib_new_child_link ->
ipoib_intf_init->
rdma_init_netdev->
mlx5_rdma_setup_rn->

mlx5e_attach_netdev->
mlx5e_num_channels_changed ->
mlx5e_set_default_xps_cpumasks ->
netif_set_xps_queue ->
__netif_set_xps_queue -> kmalloc

If any later stage fails at any point after mlx5e_num_channels_changed()
returns, XPS allocated maps will never be freed as they
are only freed during netdev unregistration, which will never happen for
yet to be registered netdevs.

Fixes: 3909a12e7913 ("net/mlx5e: Fix configuration of XPS cpumasks and netdev queues in corner cases")
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>

net/mlx5e: Make sure fib dev exists in fib event

For unreachable route entry the fib dev does not exists.

Fixes: 8914add2c9e5 ("net/mlx5e: Handle FIB events to update tunnel endpoint device")
Reported-by: Dennis Afanasev <dennis.afanasev@stateless.net>
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix null deref accessing lag dev

It could be the lag dev is null so stop processing the event.
In bond_enslave() the active/backup slave being set before setting the
upper dev so first event is without an upper dev.
After setting the upper dev with bond_master_upper_dev_link() there is
a second event and in that event we have an upper dev.

Fixes: 7e51891a237f ("net/mlx5e: Use netdev events to set/del egress acl forward-to-vport rule")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix nullptr in mlx5e_tc_add_fdb_flow()

The result of __dev_get_by_index() is not checked for NULL, which then
passed to mlx5e_attach_encap() and gets dereferenced.

Also, in case of a successful lookup, the net_device reference count is
not incremented, which may result in net_device pointer becoming invalid
at any time during mlx5e_attach_encap() execution.

Fix by using dev_get_by_index(), which does proper reference counting on
the net_device pointer. Also, handle nullptr return value when mirred
device is not found.

It's safe to call dev_put() on the mirred net_device pointer, right
after mlx5e_attach_encap() call, because it's not being saved/copied
down the call chain.

Fixes: 3c37745ec614 ("net/mlx5e: Properly deal with encap flows add/del under neigh update")
Addresses-Coverity: ("Dereference null return value")
Signed-off-by: Dima Chumak <dchumak@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: SF, Fix show state inactive when its inactivated

When a SF is inactivated and when it is in a TEARDOWN_REQUEST
state, driver still returns its state as active. This is incorrect.
Fix it by treating TEARDOWN_REQEUST as inactive state. When a SF
is still attached to the driver, on user request to reactivate EINVAL
error is returned. Inform user about it with better code EBUSY and
informative error message.

Fixes: 6a3273217469 ("net/mlx5: SF, Port function state change support")
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Vu Pham <vuhuong@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: Fix err prints and return when creating termination table

Fix print to print correct error code and not using IS_ERR() which
will just result in always printing 1.
Also return real err instead of always -EOPNOTSUPP.

Fixes: 10caabdaad5a ("net/mlx5e: Use termination table for VLAN push actions")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: Set reformat action when needed for termination rules

For remote mirroring, after the tunnel packets are received, they are
decapsulated and sent to representor, then re-encapsulated and sent
out over another tunnel. So reformat action is set only when the
destination is required to do encapsulation.

Fixes: 249ccc3c95bd ("net/mlx5e: Add support for offloading traffic from uplink to uplink")
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Ariel Levkovich <lariel@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix nullptr in add_vlan_push_action()

The result of dev_get_by_index_rcu() is not checked for NULL and then
gets dereferenced immediately.

Also, the RCU lock must be held by the caller of dev_get_by_index_rcu(),
which isn't satisfied by the call stack.

Fix by handling nullptr return value when iflink device is not found.
Add RCU locking around dev_get_by_index_rcu() to avoid possible adverse
effects while iterating over the net_device's hlist.

It is safe not to increment reference count of the net_device pointer in
case of a successful lookup, because it's already handled by VLAN code
during VLAN device registration (see register_vlan_dev and
netdev_upper_dev_link).

Fixes: 278748a95aa3 ("net/mlx5e: Offload TC e-switch rules with egress VLAN device")
Addresses-Coverity: ("Dereference null return value")
Signed-off-by: Dima Chumak <dchumak@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

{net, RDMA}/mlx5: Fix override of log_max_qp by other device

mlx5_core_dev holds pointer to static profile, hence when the
log_max_qp of the profile is override by some device, then it
effect all other mlx5 devices that share the same profile.
Fix it by having a profile instance for every mlx5 device.

Fixes: 883371c453b9 ("net/mlx5: Check FW limitations on log_max_qp before setting it")
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Merge branch 'hns3-fixes'

Huazhong Tan says:

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

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

Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: check the return of skb_checksum_help()

Currently skb_checksum_help()'s return is ignored, but it may
return error when it fails to allocate memory when linearizing.

So adds checking for the return of skb_checksum_help().

Fixes: 76ad4f0ee747("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Fixes: 3db084d28dc0("net: hns3: Fix for vxlan tx checksum bug")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: fix user's coalesce configuration lost issue

Currently, when adaptive is on, the user's coalesce configuration
may be overwritten by the dynamic one. The reason is that user's
configurations are saved in struct hns3_enet_tqp_vector whose
value maybe changed by the dynamic algorithm. To fix it, use
struct hns3_nic_priv instead of struct hns3_enet_tqp_vector to
save and get the user's configuration.

BTW, operations of storing and restoring coalesce info in the reset
process are unnecessary now, so remove them as well.

Fixes: 434776a5fae2 ("net: hns3: add ethtool_ops.set_coalesce support to PF")
Fixes: 7e96adc46633 ("net: hns3: add ethtool_ops.get_coalesce support to PF")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: put off calling register_netdev() until client initialize complete

Currently, the netdevice is registered before client initializing
complete. So there is a timewindow between netdevice available
and usable. In this case, if user try to change the channel number
or ring param, it may cause the hns3_set_rx_cpu_rmap() being called
twice, and report bug.

[47199.416502] hns3 0000:35:00.0 eth1: set channels: tqp_num=1, rxfh=0
[47199.430340] hns3 0000:35:00.0 eth1: already uninitialized
[47199.438554] hns3 0000:35:00.0: rss changes from 4 to 1
[47199.511854] hns3 0000:35:00.0: Channels changed, rss_size from 4 to 1, tqps from 4 to 1
[47200.163524] ------------[ cut here ]------------
[47200.171674] kernel BUG at lib/cpu_rmap.c:142!
[47200.177847] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[47200.185259] Modules linked in: hclge(+) hns3(-) hns3_cae(O) hns_roce_hw_v2 hnae3 vfio_iommu_type1 vfio_pci vfio_virqfd vfio pv680_mii(O) [last unloaded: hclge]
[47200.205912] CPU: 1 PID: 8260 Comm: ethtool Tainted: G           O      5.11.0-rc3+ #1
[47200.215601] Hardware name:  , xxxxxx 02/04/2021
[47200.223052] pstate: 60400009 (nZCv daif +PAN -UAO -TCO BTYPE=--)
[47200.230188] pc : cpu_rmap_add+0x38/0x40
[47200.237472] lr : irq_cpu_rmap_add+0x84/0x140
[47200.243291] sp : ffff800010e93a30
[47200.247295] x29: ffff800010e93a30 x28: ffff082100584880
[47200.254155] x27: 0000000000000000 x26: 0000000000000000
[47200.260712] x25: 0000000000000000 x24: 0000000000000004
[47200.267241] x23: ffff08209ba03000 x22: ffff08209ba038c0
[47200.273789] x21: 000000000000003f x20: ffff0820e2bc1680
[47200.280400] x19: ffff0820c970ec80 x18: 00000000000000c0
[47200.286944] x17: 0000000000000000 x16: ffffb43debe4a0d0
[47200.293456] x15: fffffc2082990600 x14: dead000000000122
[47200.300059] x13: ffffffffffffffff x12: 000000000000003e
[47200.306606] x11: ffff0820815b8080 x10: ffff53e411988000
[47200.313171] x9 : 0000000000000000 x8 : ffff0820e2bc1700
[47200.319682] x7 : 0000000000000000 x6 : 000000000000003f
[47200.326170] x5 : 0000000000000040 x4 : ffff800010e93a20
[47200.332656] x3 : 0000000000000004 x2 : ffff0820c970ec80
[47200.339168] x1 : ffff0820e2bc1680 x0 : 0000000000000004
[47200.346058] Call trace:
[47200.349324]  cpu_rmap_add+0x38/0x40
[47200.354300]  hns3_set_rx_cpu_rmap+0x6c/0xe0 [hns3]
[47200.362294]  hns3_reset_notify_init_enet+0x1cc/0x340 [hns3]
[47200.370049]  hns3_change_channels+0x40/0xb0 [hns3]
[47200.376770]  hns3_set_channels+0x12c/0x2a0 [hns3]
[47200.383353]  ethtool_set_channels+0x140/0x250
[47200.389772]  dev_ethtool+0x714/0x23d0
[47200.394440]  dev_ioctl+0x4cc/0x640
[47200.399277]  sock_do_ioctl+0x100/0x2a0
[47200.404574]  sock_ioctl+0x28c/0x470
[47200.409079]  __arm64_sys_ioctl+0xb4/0x100
[47200.415217]  el0_svc_common.constprop.0+0x84/0x210
[47200.422088]  do_el0_svc+0x28/0x34
[47200.426387]  el0_svc+0x28/0x70
[47200.431308]  el0_sync_handler+0x1a4/0x1b0
[47200.436477]  el0_sync+0x174/0x180
[47200.441562] Code: 11000405 79000c45 f8247861 d65f03c0 (d4210000)
[47200.448869] ---[ end trace a01efe4ce42e5f34 ]---

The process is like below:
excuting hns3_client_init
|
register_netdev()
|                           hns3_set_channels()
|                           |
hns3_set_rx_cpu_rmap()      hns3_reset_notify_uninit_enet()
|                               |
|                            quit without calling function
|                            hns3_free_rx_cpu_rmap for flag
|                            HNS3_NIC_STATE_INITED is unset.
|                           |
|                           hns3_reset_notify_init_enet()
|                               |
set HNS3_NIC_STATE_INITED    call hns3_set_rx_cpu_rmap()-- crash

Fix it by calling register_netdev() at the end of function
hns3_client_init().

Fixes: 08a100689d4b ("net: hns3: re-organize vector handle")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: fix incorrect resp_msg issue

In hclge_mbx_handler(), if there are two consecutive mailbox
messages that requires resp_msg, the resp_msg is not cleared
after processing the first message, which will cause the resp_msg
data of second message incorrect.

Fix it by clearing the resp_msg before processing every mailbox
message.

Fixes: bb5790b71bad ("net: hns3: refactor mailbox response scheme between PF and VF")
Signed-off-by: Jiaran Zhang <zhangjiaran@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: lan78xx: advertise tx software timestamping support

lan78xx already calls skb_tx_timestamp() in its lan78xx_start_xmit().
Override .get_ts_info to also advertise this capability
(SOF_TIMESTAMPING_TX_SOFTWARE) via ethtool.

Signed-off-by: Markus Blöchl <markus.bloechl@ipetronik.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: simplify the finalize work queue

This patch is to use "struct work_struct" for the finalize work queue
instead of "struct tipc_net_work", as it can get the "net" and "addr"
from tipc_net's other members and there is no need to add extra net
and addr in tipc_net by defining "struct tipc_net_work".

Note that it's safe to get net from tn->bcl as bcl is always released
after the finalize work queue is done.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mdiobus: get rid of a BUG_ON()

We spotted a bug recently during a review where a driver was
unregistering a bus that wasn't registered, which would trigger this
BUG_ON().  Let's handle that situation more gracefully, and just print
a warning and return.

Reported-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'gve-fixes'

David Awogbemila says:

====================
GVE bug fixes

This patch series includes fixes to some bugs in the gve driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

gve: Correct SKB queue index validation.

SKBs with skb_get_queue_mapping(skb) == tx_cfg.num_queues should also be
considered invalid.

Fixes: f5cedc84a30d ("gve: Add transmit and receive support")
Signed-off-by: David Awogbemila <awogbemila@google.com>
Acked-by: Willem de Brujin <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gve: Upgrade memory barrier in poll routine

As currently written, if the driver checks for more work (via
gve_tx_poll or gve_rx_poll) before the device posts work and the
irq doorbell is not unmasked
(via iowrite32be(GVE_IRQ_ACK | GVE_IRQ_EVENT, ...)) before the device
attempts to raise an interrupt, an interrupt is lost and this could
potentially lead to the traffic being completely halted. For
example, if a tx queue has already been stopped, the driver won't get
the chance to complete work and egress will be halted.

We need a full memory barrier in the poll
routine to ensure that the irq doorbell is unmasked before the driver
checks for more work.

Fixes: f5cedc84a30d ("gve: Add transmit and receive support")
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Acked-by: Willem de Brujin <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gve: Add NULL pointer checks when freeing irqs.

When freeing notification blocks, we index priv->msix_vectors.
If we failed to allocate priv->msix_vectors (see abort_with_msix_vectors)
this could lead to a NULL pointer dereference if the driver is unloaded.

Fixes: 893ce44df565 ("gve: Add basic driver framework for Compute Engine Virtual NIC")
Signed-off-by: David Awogbemila <awogbemila@google.com>
Acked-by: Willem de Brujin <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gve: Update mgmt_msix_idx if num_ntfy changes

If we do not get the expected number of vectors from
pci_enable_msix_range, we update priv->num_ntfy_blks but not
priv->mgmt_msix_idx. This patch fixes this so that priv->mgmt_msix_idx
is updated accordingly.

Fixes: f5cedc84a30d ("gve: Add transmit and receive support")
Signed-off-by: David Awogbemila <awogbemila@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gve: Check TX QPL was actually assigned

Correctly check the TX QPL was assigned and unassigned if
other steps in the allocation fail.

Fixes: f5cedc84a30d (gve: Add transmit and receive support)
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netlink: disable IRQs for netlink_lock_table()

Syzbot reports that in mac80211 we have a potential deadlock
between our "local->stop_queue_reasons_lock" (spinlock) and
netlink's nl_table_lock (rwlock). This is because there's at
least one situation in which we might try to send a netlink
message with this spinlock held while it is also possible to
take the spinlock from a hardirq context, resulting in the
following deadlock scenario reported by lockdep:

       CPU0                    CPU1
       ----                    ----
  lock(nl_table_lock);
                               local_irq_disable();
                               lock(&local->queue_stop_reason_lock);
                               lock(nl_table_lock);
  <Interrupt>
    lock(&local->queue_stop_reason_lock);

This seems valid, we can take the queue_stop_reason_lock in
any kind of context ("CPU0"), and call ieee80211_report_ack_skb()
with the spinlock held and IRQs disabled ("CPU1") in some
code path (ieee80211_do_stop() via ieee80211_free_txskb()).

Short of disallowing netlink use in scenarios like these
(which would be rather complex in mac80211's case due to
the deep callchain), it seems the only fix for this is to
disable IRQs while nl_table_lock is held to avoid hitting
this scenario, this disallows the "CPU0" portion of the
reported deadlock.

Note that the writer side (netlink_table_grab()) already
disables IRQs for this lock.

Unfortunately though, this seems like a huge hammer, and
maybe the whole netlink table locking should be reworked.

Reported-by: syzbot+69ff9dff50dcfe14ddd4@syzkaller.appspotmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: remove device from smcd_dev_list after failed device_add()

If the device_add() for a smcd_dev fails, there's no cleanup step that
rolls back the earlier list_add(). The device subsequently gets freed,
and we end up with a corrupted list.

Add some error handling that removes the device from the list.

Fixes: c6ba7c9ba43d ("net/smc: add base infrastructure for SMC-D and ISM")
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: init notify_work earlier to avoid uninitialized use

If bond_kobj_init() or later kzalloc() in bond_alloc_slave() fail,
then we call kobject_put() on the slave->kobj. This in turn calls
the release function slave_kobj_release() which will always try to
cancel_delayed_work_sync(&slave->notify_work), which shouldn't be
done on an uninitialized work struct.

Always initialize the work struct earlier to avoid problems here.

Syzbot bisected this down to a completely pointless commit, some
fault injection may have been at work here that caused the alloc
failure in the first place, which may interact badly with bisect.

Reported-by: syzbot+bfda097c12a00c8cae67@syzkaller.appspotmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

MAINTAINERS: net: remove stale website link

The http://www.linuxfoundation.org/en/Net does not contain networking
subsystem description ("Nothing found").

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: wait and exit until all work queues are done

On some host, a crash could be triggered simply by repeating these
commands several times:

  # modprobe tipc
  # tipc bearer enable media udp name UDP1 localip 127.0.0.1
  # rmmod tipc

  [] BUG: unable to handle kernel paging request at ffffffffc096bb00
  [] Workqueue: events 0xffffffffc096bb00
  [] Call Trace:
  []  ? process_one_work+0x1a7/0x360
  []  ? worker_thread+0x30/0x390
  []  ? create_worker+0x1a0/0x1a0
  []  ? kthread+0x116/0x130
  []  ? kthread_flush_work_fn+0x10/0x10
  []  ? ret_from_fork+0x35/0x40

When removing the TIPC module, the UDP tunnel sock will be delayed to
release in a work queue as sock_release() can't be done in rtnl_lock().
If the work queue is schedule to run after the TIPC module is removed,
kernel will crash as the work queue function cleanup_beareri() code no
longer exists when trying to invoke it.

To fix it, this patch introduce a member wq_count in tipc_net to track
the numbers of work queues in schedule, and  wait and exit until all
work queues are done in tipc_exit_net().

Fixes: d0f91938bede ("tipc: add ip/udp media type")
Reported-by: Shuang Li <shuali@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mld: fix panic in mld_newpack()

mld_newpack() doesn't allow to allocate high order page,
only order-0 allocation is allowed.
If headroom size is too large, a kernel panic could occur in skb_put().

Test commands:
    ip netns del A
    ip netns del B
    ip netns add A
    ip netns add B
    ip link add veth0 type veth peer name veth1
    ip link set veth0 netns A
    ip link set veth1 netns B

    ip netns exec A ip link set lo up
    ip netns exec A ip link set veth0 up
    ip netns exec A ip -6 a a 2001:db8:0::1/64 dev veth0
    ip netns exec B ip link set lo up
    ip netns exec B ip link set veth1 up
    ip netns exec B ip -6 a a 2001:db8:0::2/64 dev veth1
    for i in {1..99}
    do
        let A=$i-1
        ip netns exec A ip link add ip6gre$i type ip6gre \
local 2001:db8:$A::1 remote 2001:db8:$A::2 encaplimit 100
        ip netns exec A ip -6 a a 2001:db8:$i::1/64 dev ip6gre$i
        ip netns exec A ip link set ip6gre$i up

        ip netns exec B ip link add ip6gre$i type ip6gre \
local 2001:db8:$A::2 remote 2001:db8:$A::1 encaplimit 100
        ip netns exec B ip -6 a a 2001:db8:$i::2/64 dev ip6gre$i
        ip netns exec B ip link set ip6gre$i up
    done

Splat looks like:
kernel BUG at net/core/skbuff.c:110!
invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI
CPU: 0 PID: 7 Comm: kworker/0:1 Not tainted 5.12.0+ #891
Workqueue: ipv6_addrconf addrconf_dad_work
RIP: 0010:skb_panic+0x15d/0x15f
Code: 92 fe 4c 8b 4c 24 10 53 8b 4d 70 45 89 e0 48 c7 c7 00 ae 79 83
41 57 41 56 41 55 48 8b 54 24 a6 26 f9 ff <0f> 0b 48 8b 6c 24 20 89
34 24 e8 4a 4e 92 fe 8b 34 24 48 c7 c1 20
RSP: 0018:ffff88810091f820 EFLAGS: 00010282
RAX: 0000000000000089 RBX: ffff8881086e9000 RCX: 0000000000000000
RDX: 0000000000000089 RSI: 0000000000000008 RDI: ffffed1020123efb
RBP: ffff888005f6eac0 R08: ffffed1022fc0031 R09: ffffed1022fc0031
R10: ffff888117e00187 R11: ffffed1022fc0030 R12: 0000000000000028
R13: ffff888008284eb0 R14: 0000000000000ed8 R15: 0000000000000ec0
FS:  0000000000000000(0000) GS:ffff888117c00000(0000)
knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f8b801c5640 CR3: 0000000033c2c006 CR4: 00000000003706f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 ? ip6_mc_hdr.isra.26.constprop.46+0x12a/0x600
 ? ip6_mc_hdr.isra.26.constprop.46+0x12a/0x600
 skb_put.cold.104+0x22/0x22
 ip6_mc_hdr.isra.26.constprop.46+0x12a/0x600
 ? rcu_read_lock_sched_held+0x91/0xc0
 mld_newpack+0x398/0x8f0
 ? ip6_mc_hdr.isra.26.constprop.46+0x600/0x600
 ? lock_contended+0xc40/0xc40
 add_grhead.isra.33+0x280/0x380
 add_grec+0x5ca/0xff0
 ? mld_sendpack+0xf40/0xf40
 ? lock_downgrade+0x690/0x690
 mld_send_initial_cr.part.34+0xb9/0x180
 ipv6_mc_dad_complete+0x15d/0x1b0
 addrconf_dad_completed+0x8d2/0xbb0
 ? lock_downgrade+0x690/0x690
 ? addrconf_rs_timer+0x660/0x660
 ? addrconf_dad_work+0x73c/0x10e0
 addrconf_dad_work+0x73c/0x10e0

Allowing high order page allocation could fix this problem.

Fixes: 72e09ad107e7 ("ipv6: avoid high order allocations")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

isdn: mISDN: netjet: Fix crash in nj_probe:

'nj_setup' in netjet.c might fail with -EIO and in this case
'card->irq' is initialized and is bigger than zero. A subsequent call to
'nj_release' will free the irq that has not been requested.

Fix this bug by deleting the previous assignment to 'card->irq' and just
keep the assignment before 'request_irq'.

The KASAN's log reveals it:

[    3.354615 ] WARNING: CPU: 0 PID: 1 at kernel/irq/manage.c:1826
free_irq+0x100/0x480
[    3.355112 ] Modules linked in:
[    3.355310 ] CPU: 0 PID: 1 Comm: swapper/0 Not tainted
5.13.0-rc1-00144-g25a1298726e #13
[    3.355816 ] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS
rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
[    3.356552 ] RIP: 0010:free_irq+0x100/0x480
[    3.356820 ] Code: 6e 08 74 6f 4d 89 f4 e8 5e ac 09 00 4d 8b 74 24 18
4d 85 f6 75 e3 e8 4f ac 09 00 8b 75 c8 48 c7 c7 78 c1 2e 85 e8 e0 cf f5
ff <0f> 0b 48 8b 75 c0 4c 89 ff e8 72 33 0b 03 48 8b 43 40 4c 8b a0 80
[    3.358012 ] RSP: 0000:ffffc90000017b48 EFLAGS: 00010082
[    3.358357 ] RAX: 0000000000000000 RBX: ffff888104dc8000 RCX:
0000000000000000
[    3.358814 ] RDX: ffff8881003c8000 RSI: ffffffff8124a9e6 RDI:
00000000ffffffff
[    3.359272 ] RBP: ffffc90000017b88 R08: 0000000000000000 R09:
0000000000000000
[    3.359732 ] R10: ffffc900000179f0 R11: 0000000000001d04 R12:
0000000000000000
[    3.360195 ] R13: ffff888107dc6000 R14: ffff888107dc6928 R15:
ffff888104dc80a8
[    3.360652 ] FS:  0000000000000000(0000) GS:ffff88817bc00000(0000)
knlGS:0000000000000000
[    3.361170 ] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    3.361538 ] CR2: 0000000000000000 CR3: 000000000582e000 CR4:
00000000000006f0
[    3.362003 ] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[    3.362175 ] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[    3.362175 ] Call Trace:
[    3.362175 ]  nj_release+0x51/0x1e0
[    3.362175 ]  nj_probe+0x450/0x950
[    3.362175 ]  ? pci_device_remove+0x110/0x110
[    3.362175 ]  local_pci_probe+0x45/0xa0
[    3.362175 ]  pci_device_probe+0x12b/0x1d0
[    3.362175 ]  really_probe+0x2a9/0x610
[    3.362175 ]  driver_probe_device+0x90/0x1d0
[    3.362175 ]  ? mutex_lock_nested+0x1b/0x20
[    3.362175 ]  device_driver_attach+0x68/0x70
[    3.362175 ]  __driver_attach+0x124/0x1b0
[    3.362175 ]  ? device_driver_attach+0x70/0x70
[    3.362175 ]  bus_for_each_dev+0xbb/0x110
[    3.362175 ]  ? rdinit_setup+0x45/0x45
[    3.362175 ]  driver_attach+0x27/0x30
[    3.362175 ]  bus_add_driver+0x1eb/0x2a0
[    3.362175 ]  driver_register+0xa9/0x180
[    3.362175 ]  __pci_register_driver+0x82/0x90
[    3.362175 ]  ? w6692_init+0x38/0x38
[    3.362175 ]  nj_init+0x36/0x38
[    3.362175 ]  do_one_initcall+0x7f/0x3d0
[    3.362175 ]  ? rdinit_setup+0x45/0x45
[    3.362175 ]  ? rcu_read_lock_sched_held+0x4f/0x80
[    3.362175 ]  kernel_init_freeable+0x2aa/0x301
[    3.362175 ]  ? rest_init+0x2c0/0x2c0
[    3.362175 ]  kernel_init+0x18/0x190
[    3.362175 ]  ? rest_init+0x2c0/0x2c0
[    3.362175 ]  ? rest_init+0x2c0/0x2c0
[    3.362175 ]  ret_from_fork+0x1f/0x30
[    3.362175 ] Kernel panic - not syncing: panic_on_warn set ...
[    3.362175 ] CPU: 0 PID: 1 Comm: swapper/0 Not tainted
5.13.0-rc1-00144-g25a1298726e #13
[    3.362175 ] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS
rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
[    3.362175 ] Call Trace:
[    3.362175 ]  dump_stack+0xba/0xf5
[    3.362175 ]  ? free_irq+0x100/0x480
[    3.362175 ]  panic+0x15a/0x3f2
[    3.362175 ]  ? __warn+0xf2/0x150
[    3.362175 ]  ? free_irq+0x100/0x480
[    3.362175 ]  __warn+0x108/0x150
[    3.362175 ]  ? free_irq+0x100/0x480
[    3.362175 ]  report_bug+0x119/0x1c0
[    3.362175 ]  handle_bug+0x3b/0x80
[    3.362175 ]  exc_invalid_op+0x18/0x70
[    3.362175 ]  asm_exc_invalid_op+0x12/0x20
[    3.362175 ] RIP: 0010:free_irq+0x100/0x480
[    3.362175 ] Code: 6e 08 74 6f 4d 89 f4 e8 5e ac 09 00 4d 8b 74 24 18
4d 85 f6 75 e3 e8 4f ac 09 00 8b 75 c8 48 c7 c7 78 c1 2e 85 e8 e0 cf f5
ff <0f> 0b 48 8b 75 c0 4c 89 ff e8 72 33 0b 03 48 8b 43 40 4c 8b a0 80
[    3.362175 ] RSP: 0000:ffffc90000017b48 EFLAGS: 00010082
[    3.362175 ] RAX: 0000000000000000 RBX: ffff888104dc8000 RCX:
0000000000000000
[    3.362175 ] RDX: ffff8881003c8000 RSI: ffffffff8124a9e6 RDI:
00000000ffffffff
[    3.362175 ] RBP: ffffc90000017b88 R08: 0000000000000000 R09:
0000000000000000
[    3.362175 ] R10: ffffc900000179f0 R11: 0000000000001d04 R12:
0000000000000000
[    3.362175 ] R13: ffff888107dc6000 R14: ffff888107dc6928 R15:
ffff888104dc80a8
[    3.362175 ]  ? vprintk+0x76/0x150
[    3.362175 ]  ? free_irq+0x100/0x480
[    3.362175 ]  nj_release+0x51/0x1e0
[    3.362175 ]  nj_probe+0x450/0x950
[    3.362175 ]  ? pci_device_remove+0x110/0x110
[    3.362175 ]  local_pci_probe+0x45/0xa0
[    3.362175 ]  pci_device_probe+0x12b/0x1d0
[    3.362175 ]  really_probe+0x2a9/0x610
[    3.362175 ]  driver_probe_device+0x90/0x1d0
[    3.362175 ]  ? mutex_lock_nested+0x1b/0x20
[    3.362175 ]  device_driver_attach+0x68/0x70
[    3.362175 ]  __driver_attach+0x124/0x1b0
[    3.362175 ]  ? device_driver_attach+0x70/0x70
[    3.362175 ]  bus_for_each_dev+0xbb/0x110
[    3.362175 ]  ? rdinit_setup+0x45/0x45
[    3.362175 ]  driver_attach+0x27/0x30
[    3.362175 ]  bus_add_driver+0x1eb/0x2a0
[    3.362175 ]  driver_register+0xa9/0x180
[    3.362175 ]  __pci_register_driver+0x82/0x90
[    3.362175 ]  ? w6692_init+0x38/0x38
[    3.362175 ]  nj_init+0x36/0x38
[    3.362175 ]  do_one_initcall+0x7f/0x3d0
[    3.362175 ]  ? rdinit_setup+0x45/0x45
[    3.362175 ]  ? rcu_read_lock_sched_held+0x4f/0x80
[    3.362175 ]  kernel_init_freeable+0x2aa/0x301
[    3.362175 ]  ? rest_init+0x2c0/0x2c0
[    3.362175 ]  kernel_init+0x18/0x190
[    3.362175 ]  ? rest_init+0x2c0/0x2c0
[    3.362175 ]  ? rest_init+0x2c0/0x2c0
[    3.362175 ]  ret_from_fork+0x1f/0x30
[    3.362175 ] Dumping ftrace buffer:
[    3.362175 ]    (ftrace buffer empty)
[    3.362175 ] Kernel Offset: disabled
[    3.362175 ] Rebooting in 1 seconds..

Reported-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'bnxt_en-fixes'

Michael Chan says:

====================
bnxt_en: 2 bug fixes.

The first one fixes a bug to properly identify some recently added HyperV
device IDs.  The second one fixes device context memory set up on systems
with 64K page size.

Please queue these for -stable as well.  Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

bnxt_en: Fix context memory setup for 64K page size.

There was a typo in the code that checks for 64K BNXT_PAGE_SHIFT in
bnxt_hwrm_set_pg_attr().  Fix it and make the code more understandable
with a new macro BNXT_SET_CTX_PAGE_ATTR().

Fixes: 1b9394e5a2ad ("bnxt_en: Configure context memory on new devices.")
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnxt_en: Include new P5 HV definition in VF check.

Otherwise, some of the recently added HyperV VF IDs would not be
recognized as VF devices and they would not initialize properly.

Fixes: 7fbf359bb2c1 ("bnxt_en: Add PCI IDs for Hyper-V VF devices.")
Reviewed-by: Edwin Peer <edwin.peer@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>

net: bnx2: Fix error return code in bnx2_init_board()

Fix to return -EPERM from the error handling case instead of 0, as done
elsewhere in this function.

Fixes: b6016b767397 ("[BNX2]: New Broadcom gigabit network driver.")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

NFC: nci: fix memory leak in nci_allocate_device

nfcmrvl_disconnect fails to free the hci_dev field in struct nci_dev.
Fix this by freeing hci_dev in nci_free_device.

BUG: memory leak
unreferenced object 0xffff888111ea6800 (size 1024):
  comm "kworker/1:0", pid 19, jiffies 4294942308 (age 13.580s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 00 60 fd 0c 81 88 ff ff  .........`......
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<000000004bc25d43>] kmalloc include/linux/slab.h:552 [inline]
    [<000000004bc25d43>] kzalloc include/linux/slab.h:682 [inline]
    [<000000004bc25d43>] nci_hci_allocate+0x21/0xd0 net/nfc/nci/hci.c:784
    [<00000000c59cff92>] nci_allocate_device net/nfc/nci/core.c:1170 [inline]
    [<00000000c59cff92>] nci_allocate_device+0x10b/0x160 net/nfc/nci/core.c:1132
    [<00000000006e0a8e>] nfcmrvl_nci_register_dev+0x10a/0x1c0 drivers/nfc/nfcmrvl/main.c:153
    [<000000004da1b57e>] nfcmrvl_probe+0x223/0x290 drivers/nfc/nfcmrvl/usb.c:345
    [<00000000d506aed9>] usb_probe_interface+0x177/0x370 drivers/usb/core/driver.c:396
    [<00000000bc632c92>] really_probe+0x159/0x4a0 drivers/base/dd.c:554
    [<00000000f5009125>] driver_probe_device+0x84/0x100 drivers/base/dd.c:740
    [<000000000ce658ca>] __device_attach_driver+0xee/0x110 drivers/base/dd.c:846
    [<000000007067d05f>] bus_for_each_drv+0xb7/0x100 drivers/base/bus.c:431
    [<00000000f8e13372>] __device_attach+0x122/0x250 drivers/base/dd.c:914
    [<000000009cf68860>] bus_probe_device+0xc6/0xe0 drivers/base/bus.c:491
    [<00000000359c965a>] device_add+0x5be/0xc30 drivers/base/core.c:3109
    [<00000000086e4bd3>] usb_set_configuration+0x9d9/0xb90 drivers/usb/core/message.c:2164
    [<00000000ca036872>] usb_generic_driver_probe+0x8c/0xc0 drivers/usb/core/generic.c:238
    [<00000000d40d36f6>] usb_probe_device+0x5c/0x140 drivers/usb/core/driver.c:293
    [<00000000bc632c92>] really_probe+0x159/0x4a0 drivers/base/dd.c:554

Reported-by: syzbot+19bcfc64a8df1318d1c3@syzkaller.appspotmail.com
Fixes: 11f54f228643 ("NFC: nci: Add HCI over NCI protocol support")
Signed-off-by: Dongliang Mu <mudongliangabcd@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hso: check for allocation failure in hso_create_bulk_serial_device()

In current kernels, small allocations never actually fail so this
patch shouldn't affect runtime.

Originally this error handling code written with the idea that if
the "serial->tiocmget" allocation failed, then we would continue
operating instead of bailing out early.  But in later years we added
an unchecked dereference on the next line.

serial->tiocmget->serial_state_notification = kzalloc();
        ^^^^^^^^^^^^^^^^^^

Since these allocations are never going fail in real life, this is
mostly a philosophical debate, but I think bailing out early is the
correct behavior that the user would want.  And generally it's safer to
bail as soon an error happens.

Fixes: af0de1303c4e ("usb: hso: obey DMA rules in tiocmget")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: skb_linearize the head skb when reassembling msgs

It's not a good idea to append the frag skb to a skb's frag_list if
the frag_list already has skbs from elsewhere, such as this skb was
created by pskb_copy() where the frag_list was cloned (all the skbs
in it were skb_get'ed) and shared by multiple skbs.

However, the new appended frag skb should have been only seen by the
current skb. Otherwise, it will cause use after free crashes as this
appended frag skb are seen by multiple skbs but it only got skb_get
called once.

The same thing happens with a skb updated by pskb_may_pull() with a
skb_cloned skb. Li Shuang has reported quite a few crashes caused
by this when doing testing over macvlan devices:

  [] kernel BUG at net/core/skbuff.c:1970!
  [] Call Trace:
  []  skb_clone+0x4d/0xb0
  []  macvlan_broadcast+0xd8/0x160 [macvlan]
  []  macvlan_process_broadcast+0x148/0x150 [macvlan]
  []  process_one_work+0x1a7/0x360
  []  worker_thread+0x30/0x390

  [] kernel BUG at mm/usercopy.c:102!
  [] Call Trace:
  []  __check_heap_object+0xd3/0x100
  []  __check_object_size+0xff/0x16b
  []  simple_copy_to_iter+0x1c/0x30
  []  __skb_datagram_iter+0x7d/0x310
  []  __skb_datagram_iter+0x2a5/0x310
  []  skb_copy_datagram_iter+0x3b/0x90
  []  tipc_recvmsg+0x14a/0x3a0 [tipc]
  []  ____sys_recvmsg+0x91/0x150
  []  ___sys_recvmsg+0x7b/0xc0

  [] kernel BUG at mm/slub.c:305!
  [] Call Trace:
  []  <IRQ>
  []  kmem_cache_free+0x3ff/0x400
  []  __netif_receive_skb_core+0x12c/0xc40
  []  ? kmem_cache_alloc+0x12e/0x270
  []  netif_receive_skb_internal+0x3d/0xb0
  []  ? get_rx_page_info+0x8e/0xa0 [be2net]
  []  be_poll+0x6ef/0xd00 [be2net]
  []  ? irq_exit+0x4f/0x100
  []  net_rx_action+0x149/0x3b0

  ...

This patch is to fix it by linearizing the head skb if it has frag_list
set in tipc_buf_append(). Note that we choose to do this before calling
skb_unshare(), as __skb_linearize() will avoid skb_copy(). Also, we can
not just drop the frag_list either as the early time.

Fixes: 45c8b7b175ce ("tipc: allow non-linear first fragment buffer")
Reported-by: Li Shuang <shuali@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: cdc_eem: fix URL to CDC EEM 1.0 spec

The old URL is no longer accessible.

Signed-off-by: Jonathan Davies <jonathan.davies@nutanix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'lockless-qdisc-packet-stuck'

Yunsheng Lin says:

====================
ix packet stuck problem for lockless qdisc

This patchset fixes the packet stuck problem mentioned in [1].

Patch 1: Add STATE_MISSED flag to fix packet stuck problem.
Patch 2: Fix a tx_action rescheduling problem after STATE_MISSED
         flag is added in patch 1.
Patch 3: Fix the significantly higher CPU consumption problem when
         multiple threads are competing on a saturated outgoing
         device.

V8: Change function name as suggested by Jakub and fix some typo
    in patch 3, adjust commit log in patch 2, and add Acked-by
    from Jakub.
V7: Fix netif_tx_wake_queue() data race noted by Jakub.
V6: Some performance optimization in patch 1 suggested by Jakub
    and drop NET_XMIT_DROP checking in patch 3.
V5: add patch 3 to fix the problem reported by Michal Kubecek.
V4: Change STATE_NEED_RESCHEDULE to STATE_MISSED and add patch 2.

[1]. https://lkml.org/lkml/2019/10/9/42
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: fix tx action reschedule issue with stopped queue

The netdev qeueue might be stopped when byte queue limit has
reached or tx hw ring is full, net_tx_action() may still be
rescheduled if STATE_MISSED is set, which consumes unnecessary
cpu without dequeuing and transmiting any skb because the
netdev queue is stopped, see qdisc_run_end().

This patch fixes it by checking the netdev queue state before
calling qdisc_run() and clearing STATE_MISSED if netdev queue is
stopped during qdisc_run(), the net_tx_action() is rescheduled
again when netdev qeueue is restarted, see netif_tx_wake_queue().

As there is time window between netif_xmit_frozen_or_stopped()
checking and STATE_MISSED clearing, between which STATE_MISSED
may set by net_tx_action() scheduled by netif_tx_wake_queue(),
so set the STATE_MISSED again if netdev queue is restarted.

Fixes: 6b3ba9146fe6 ("net: sched: allow qdiscs to handle locking")
Reported-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: fix tx action rescheduling issue during deactivation

Currently qdisc_run() checks the STATE_DEACTIVATED of lockless
qdisc before calling __qdisc_run(), which ultimately clear the
STATE_MISSED when all the skb is dequeued. If STATE_DEACTIVATED
is set before clearing STATE_MISSED, there may be rescheduling
of net_tx_action() at the end of qdisc_run_end(), see below:

CPU0(net_tx_atcion)  CPU1(__dev_xmit_skb)  CPU2(dev_deactivate)
          .                   .                     .
          .            set STATE_MISSED             .
          .           __netif_schedule()            .
          .                   .           set STATE_DEACTIVATED
          .                   .                qdisc_reset()
          .                   .                     .
          .<---------------   .              synchronize_net()
clear __QDISC_STATE_SCHED  |  .                     .
          .                |  .                     .
          .                |  .            some_qdisc_is_busy()
          .                |  .               return *false*
          .                |  .                     .
  test STATE_DEACTIVATED   |  .                     .
__qdisc_run() *not* called |  .                     .
          .                |  .                     .
   test STATE_MISS         |  .                     .
 __netif_schedule()--------|  .                     .
          .                   .                     .
          .                   .                     .

__qdisc_run() is not called by net_tx_atcion() in CPU0 because
CPU2 has set STATE_DEACTIVATED flag during dev_deactivate(), and
STATE_MISSED is only cleared in __qdisc_run(), __netif_schedule
is called at the end of qdisc_run_end(), causing tx action
rescheduling problem.

qdisc_run() called by net_tx_action() runs in the softirq context,
which should has the same semantic as the qdisc_run() called by
__dev_xmit_skb() protected by rcu_read_lock_bh(). And there is a
synchronize_net() between STATE_DEACTIVATED flag being set and
qdisc_reset()/some_qdisc_is_busy in dev_deactivate(), we can safely
bail out for the deactived lockless qdisc in net_tx_action(), and
qdisc_reset() will reset all skb not dequeued yet.

So add the rcu_read_lock() explicitly to protect the qdisc_run()
and do the STATE_DEACTIVATED checking in net_tx_action() before
calling qdisc_run_begin(). Another option is to do the checking in
the qdisc_run_end(), but it will add unnecessary overhead for
non-tx_action case, because __dev_queue_xmit() will not see qdisc
with STATE_DEACTIVATED after synchronize_net(), the qdisc with
STATE_DEACTIVATED can only be seen by net_tx_action() because of
__netif_schedule().

The STATE_DEACTIVATED checking in qdisc_run() is to avoid race
between net_tx_action() and qdisc_reset(), see:
commit d518d2ed8640 ("net/sched: fix race between deactivation
and dequeue for NOLOCK qdisc"). As the bailout added above for
deactived lockless qdisc in net_tx_action() provides better
protection for the race without calling qdisc_run() at all, so
remove the STATE_DEACTIVATED checking in qdisc_run().

After qdisc_reset(), there is no skb in qdisc to be dequeued, so
clear the STATE_MISSED in dev_reset_queue() too.

Fixes: 6b3ba9146fe6 ("net: sched: allow qdiscs to handle locking")
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
V8: Clearing STATE_MISSED before calling __netif_schedule() has
    avoid the endless rescheduling problem, but there may still
    be a unnecessary rescheduling, so adjust the commit log.
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: fix packet stuck problem for lockless qdisc

Lockless qdisc has below concurrent problem:
    cpu0                 cpu1
     .                     .
q->enqueue                 .
     .                     .
qdisc_run_begin()          .
     .                     .
dequeue_skb()              .
     .                     .
sch_direct_xmit()          .
     .                     .
     .                q->enqueue
     .             qdisc_run_begin()
     .            return and do nothing
     .                     .
qdisc_run_end()            .

cpu1 enqueue a skb without calling __qdisc_run() because cpu0
has not released the lock yet and spin_trylock() return false
for cpu1 in qdisc_run_begin(), and cpu0 do not see the skb
enqueued by cpu1 when calling dequeue_skb() because cpu1 may
enqueue the skb after cpu0 calling dequeue_skb() and before
cpu0 calling qdisc_run_end().

Lockless qdisc has below another concurrent problem when
tx_action is involved:

cpu0(serving tx_action)     cpu1             cpu2
          .                   .                .
          .              q->enqueue            .
          .            qdisc_run_begin()       .
          .              dequeue_skb()         .
          .                   .            q->enqueue
          .                   .                .
          .             sch_direct_xmit()      .
          .                   .         qdisc_run_begin()
          .                   .       return and do nothing
          .                   .                .
 clear __QDISC_STATE_SCHED    .                .
 qdisc_run_begin()            .                .
 return and do nothing        .                .
          .                   .                .
          .            qdisc_run_end()         .

This patch fixes the above data race by:
1. If the first spin_trylock() return false and STATE_MISSED is
   not set, set STATE_MISSED and retry another spin_trylock() in
   case other CPU may not see STATE_MISSED after it releases the
   lock.
2. reschedule if STATE_MISSED is set after the lock is released
   at the end of qdisc_run_end().

For tx_action case, STATE_MISSED is also set when cpu1 is at the
end if qdisc_run_end(), so tx_action will be rescheduled again
to dequeue the skb enqueued by cpu2.

Clear STATE_MISSED before retrying a dequeuing when dequeuing
returns NULL in order to reduce the overhead of the second
spin_trylock() and __netif_schedule() calling.

Also clear the STATE_MISSED before calling __netif_schedule()
at the end of qdisc_run_end() to avoid doing another round of
dequeuing in the pfifo_fast_dequeue().

The performance impact of this patch, tested using pktgen and
dummy netdev with pfifo_fast qdisc attached:

 threads  without+this_patch   with+this_patch      delta
    1        2.61Mpps            2.60Mpps           -0.3%
    2        3.97Mpps            3.82Mpps           -3.7%
    4        5.62Mpps            5.59Mpps           -0.5%
    8        2.78Mpps            2.77Mpps           -0.3%
   16        2.22Mpps            2.22Mpps           -0.0%

Fixes: 6b3ba9146fe6 ("net: sched: allow qdiscs to handle locking")
Acked-by: Jakub Kicinski <kuba@kernel.org>
Tested-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tls splice: check SPLICE_F_NONBLOCK instead of MSG_DONTWAIT

In tls_sw_splice_read, checkout MSG_* is inappropriate, should use
SPLICE_*, update tls_wait_data to accept nonblock arguments instead
of flags for recvmsg and splice.

Fixes: c46234ebb4d1 ("tls: RX path for ktls")
Signed-off-by: Jim Ma <majinjing3@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Revert "net:tipc: Fix a double free in tipc_sk_mcast_rcv"

This reverts commit 6bf24dc0cc0cc43b29ba344b66d78590e687e046.
Above fix is not correct and caused memory leak issue.

Fixes: 6bf24dc0cc0c ("net:tipc: Fix a double free in tipc_sk_mcast_rcv")
Acked-by: Jon Maloy <jmaloy@redhat.com>
Acked-by: Tung Nguyen <tung.q.nguyen@dektech.com.au>
Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

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

The following patchset contains Netfilter fixes for net:

1) Remove the flowtable hardware refresh state, fall back to the
   existing hardware pending state instead, from Roi Dayan.

2) Fix crash in pipapo avx2 lookup when FPU is in used from user
   context, from Stefano Brivio.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: nft_set_pipapo_avx2: Add irq_fpu_usable() check, fallback to non-AVX2 version

Arturo reported this backtrace:

[709732.358791] WARNING: CPU: 3 PID: 456 at arch/x86/kernel/fpu/core.c:128 kernel_fpu_begin_mask+0xae/0xe0
[709732.358793] Modules linked in: binfmt_misc nft_nat nft_chain_nat nf_nat nft_counter nft_ct nf_tables nf_conntrack_netlink nfnetlink 8021q garp stp mrp llc vrf intel_rapl_msr intel_rapl_common skx_edac nfit libnvdimm ipmi_ssif x86_pkg_temp_thermal intel_powerclamp coretemp crc32_pclmul mgag200 ghash_clmulni_intel drm_kms_helper cec aesni_intel drm libaes crypto_simd cryptd glue_helper mei_me dell_smbios iTCO_wdt evdev intel_pmc_bxt iTCO_vendor_support dcdbas pcspkr rapl dell_wmi_descriptor wmi_bmof sg i2c_algo_bit watchdog mei acpi_ipmi ipmi_si button nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 ipmi_devintf ipmi_msghandler ip_tables x_tables autofs4 ext4 crc16 mbcache jbd2 dm_mod raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor sd_mod t10_pi crc_t10dif crct10dif_generic raid6_pq libcrc32c crc32c_generic raid1 raid0 multipath linear md_mod ahci libahci tg3 libata xhci_pci libphy xhci_hcd ptp usbcore crct10dif_pclmul crct10dif_common bnxt_en crc32c_intel scsi_mod
[709732.358941]  pps_core i2c_i801 lpc_ich i2c_smbus wmi usb_common
[709732.358957] CPU: 3 PID: 456 Comm: jbd2/dm-0-8 Not tainted 5.10.0-0.bpo.5-amd64 #1 Debian 5.10.24-1~bpo10+1
[709732.358959] Hardware name: Dell Inc. PowerEdge R440/04JN2K, BIOS 2.9.3 09/23/2020
[709732.358964] RIP: 0010:kernel_fpu_begin_mask+0xae/0xe0
[709732.358969] Code: ae 54 24 04 83 e3 01 75 38 48 8b 44 24 08 65 48 33 04 25 28 00 00 00 75 33 48 83 c4 10 5b c3 65 8a 05 5e 21 5e 76 84 c0 74 92 <0f> 0b eb 8e f0 80 4f 01 40 48 81 c7 00 14 00 00 e8 dd fb ff ff eb
[709732.358972] RSP: 0018:ffffbb9700304740 EFLAGS: 00010202
[709732.358976] RAX: 0000000000000001 RBX: 0000000000000003 RCX: 0000000000000001
[709732.358979] RDX: ffffbb9700304970 RSI: ffff922fe1952e00 RDI: 0000000000000003
[709732.358981] RBP: ffffbb9700304970 R08: ffff922fc868a600 R09: ffff922fc711e462
[709732.358984] R10: 000000000000005f R11: ffff922ff0b27180 R12: ffffbb9700304960
[709732.358987] R13: ffffbb9700304b08 R14: ffff922fc664b6c8 R15: ffff922fc664b660
[709732.358990] FS:  0000000000000000(0000) GS:ffff92371fec0000(0000) knlGS:0000000000000000
[709732.358993] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[709732.358996] CR2: 0000557a6655bdd0 CR3: 000000026020a001 CR4: 00000000007706e0
[709732.358999] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[709732.359001] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[709732.359003] PKRU: 55555554
[709732.359005] Call Trace:
[709732.359009]  <IRQ>
[709732.359035]  nft_pipapo_avx2_lookup+0x4c/0x1cba [nf_tables]
[709732.359046]  ? sched_clock+0x5/0x10
[709732.359054]  ? sched_clock_cpu+0xc/0xb0
[709732.359061]  ? record_times+0x16/0x80
[709732.359068]  ? plist_add+0xc1/0x100
[709732.359073]  ? psi_group_change+0x47/0x230
[709732.359079]  ? skb_clone+0x4d/0xb0
[709732.359085]  ? enqueue_task_rt+0x22b/0x310
[709732.359098]  ? bnxt_start_xmit+0x1e8/0xaf0 [bnxt_en]
[709732.359102]  ? packet_rcv+0x40/0x4a0
[709732.359121]  nft_lookup_eval+0x59/0x160 [nf_tables]
[709732.359133]  nft_do_chain+0x350/0x500 [nf_tables]
[709732.359152]  ? nft_lookup_eval+0x59/0x160 [nf_tables]
[709732.359163]  ? nft_do_chain+0x364/0x500 [nf_tables]
[709732.359172]  ? fib4_rule_action+0x6d/0x80
[709732.359178]  ? fib_rules_lookup+0x107/0x250
[709732.359184]  nft_nat_do_chain+0x8a/0xf2 [nft_chain_nat]
[709732.359193]  nf_nat_inet_fn+0xea/0x210 [nf_nat]
[709732.359202]  nf_nat_ipv4_out+0x14/0xa0 [nf_nat]
[709732.359207]  nf_hook_slow+0x44/0xc0
[709732.359214]  ip_output+0xd2/0x100
[709732.359221]  ? __ip_finish_output+0x210/0x210
[709732.359226]  ip_forward+0x37d/0x4a0
[709732.359232]  ? ip4_key_hashfn+0xb0/0xb0
[709732.359238]  ip_sublist_rcv_finish+0x4f/0x60
[709732.359243]  ip_sublist_rcv+0x196/0x220
[709732.359250]  ? ip_rcv_finish_core.isra.22+0x400/0x400
[709732.359255]  ip_list_rcv+0x137/0x160
[709732.359264]  __netif_receive_skb_list_core+0x29b/0x2c0
[709732.359272]  netif_receive_skb_list_internal+0x1a6/0x2d0
[709732.359280]  gro_normal_list.part.156+0x19/0x40
[709732.359286]  napi_complete_done+0x67/0x170
[709732.359298]  bnxt_poll+0x105/0x190 [bnxt_en]
[709732.359304]  ? irqentry_exit+0x29/0x30
[709732.359309]  ? asm_common_interrupt+0x1e/0x40
[709732.359315]  net_rx_action+0x144/0x3c0
[709732.359322]  __do_softirq+0xd5/0x29c
[709732.359329]  asm_call_irq_on_stack+0xf/0x20
[709732.359332]  </IRQ>
[709732.359339]  do_softirq_own_stack+0x37/0x40
[709732.359346]  irq_exit_rcu+0x9d/0xa0
[709732.359353]  common_interrupt+0x78/0x130
[709732.359358]  asm_common_interrupt+0x1e/0x40
[709732.359366] RIP: 0010:crc_41+0x0/0x1e [crc32c_intel]
[709732.359370] Code: ff ff f2 4d 0f 38 f1 93 a8 fe ff ff f2 4c 0f 38 f1 81 b0 fe ff ff f2 4c 0f 38 f1 8a b0 fe ff ff f2 4d 0f 38 f1 93 b0 fe ff ff <f2> 4c 0f 38 f1 81 b8 fe ff ff f2 4c 0f 38 f1 8a b8 fe ff ff f2 4d
[709732.359373] RSP: 0018:ffffbb97008dfcd0 EFLAGS: 00000246
[709732.359377] RAX: 000000000000002a RBX: 0000000000000400 RCX: ffff922fc591dd50
[709732.359379] RDX: ffff922fc591dea0 RSI: 0000000000000a14 RDI: ffffffffc00dddc0
[709732.359382] RBP: 0000000000001000 R08: 000000000342d8c3 R09: 0000000000000000
[709732.359384] R10: 0000000000000000 R11: ffff922fc591dff0 R12: ffffbb97008dfe58
[709732.359386] R13: 000000000000000a R14: ffff922fd2b91e80 R15: ffff922fef83fe38
[709732.359395]  ? crc_43+0x1e/0x1e [crc32c_intel]
[709732.359403]  ? crc32c_pcl_intel_update+0x97/0xb0 [crc32c_intel]
[709732.359419]  ? jbd2_journal_commit_transaction+0xaec/0x1a30 [jbd2]
[709732.359425]  ? irq_exit_rcu+0x3e/0xa0
[709732.359447]  ? kjournald2+0xbd/0x270 [jbd2]
[709732.359454]  ? finish_wait+0x80/0x80
[709732.359470]  ? commit_timeout+0x10/0x10 [jbd2]
[709732.359476]  ? kthread+0x116/0x130
[709732.359481]  ? kthread_park+0x80/0x80
[709732.359488]  ? ret_from_fork+0x1f/0x30
[709732.359494] ---[ end trace 081a19978e5f09f5 ]---

that is, nft_pipapo_avx2_lookup() uses the FPU running from a softirq
that interrupted a kthread, also using the FPU.

That's exactly the reason why irq_fpu_usable() is there: use it, and
if we can't use the FPU, fall back to the non-AVX2 version of the
lookup operation, i.e. nft_pipapo_lookup().

Reported-by: Arturo Borrero Gonzalez <arturo@netfilter.org>
Cc: <stable@vger.kernel.org> # 5.6.x
Fixes: 7400b063969b ("nft_set_pipapo: Introduce AVX2-based lookup implementation")
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: flowtable: Remove redundant hw refresh bit

Offloading conns could fail for multiple reasons and a hw refresh bit is
set to try to reoffload it in next sw packet.
But it could be in some cases and future points that the hw refresh bit
is not set but a refresh could succeed.
Remove the hw refresh bit and do offload refresh if requested.
There won't be a new work entry if a work is already pending
anyway as there is the hw pending bit.

Fixes: 8b3646d6e0c4 ("net/sched: act_ct: Support refreshing the flow table entries")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

openvswitch: meter: fix race when getting now_ms.

We have observed meters working unexpected if traffic is 3+Gbit/s
with multiple connections.

now_ms is not pretected by meter->lock, we may get a negative
long_delta_ms when another cpu updated meter->used, then:
    delta_ms = (u32)long_delta_ms;
which will be a large value.

    band->bucket += delta_ms * band->rate;
then we get a wrong band->bucket.

OpenVswitch userspace datapath has fixed the same issue[1] some
time ago, and we port the implementation to kernel datapath.

[1] https://patchwork.ozlabs.org/project/openvswitch/patch/20191025114436.9746-1-i.maximets@ovn.org/

Fixes: 96fbc13d7e77 ("openvswitch: Add meter infrastructure")
Signed-off-by: Tao Liu <thomas.liu@ucloud.cn>
Suggested-by: Ilya Maximets <i.maximets@ovn.org>
Reviewed-by: Ilya Maximets <i.maximets@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: korina: Fix return value check in korina_probe()

In case of error, the function devm_platform_ioremap_resource_byname()
returns ERR_PTR() and never returns NULL. The NULL test in the return
value check should be replaced with IS_ERR().

Fixes: b4cd249a8cc0 ("net: korina: Use devres functions")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4/ch_ktls: Clear resources when pf4 device is removed

This patch maintain the list of active tids and clear all the active
connection resources when DETACH notification comes.

Fixes: a8c16e8ed624f ("crypto/chcr: move nic TLS functionality to drivers/net")
Signed-off-by: Ayush Sawal <ayush.sawal@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mdio: octeon: Fix some double free issues

'bus->mii_bus' has been allocated with 'devm_mdiobus_alloc_size()' in the
probe function. So it must not be freed explicitly or there will be a
double free.

Remove the incorrect 'mdiobus_free' in the error handling path of the
probe function and in remove function.

Suggested-By: Andrew Lunn <andrew@lunn.ch>
Fixes: 35d2aeac9810 ("phy: mdio-octeon: Use devm_mdiobus_alloc_size()")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mdio: thunder: Fix a double free issue in the .remove function

'bus->mii_bus' have been allocated with 'devm_mdiobus_alloc_size()' in the
probe function. So it must not be freed explicitly or there will be a
double free.

Remove the incorrect 'mdiobus_free' in the remove function.

Fixes: 379d7ac7ca31 ("phy: mdio-thunder: Add driver for Cavium Thunder SoC MDIO buses.")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

chelsio/chtls: unlock on error in chtls_pt_recvmsg()

This error path needs to release some memory and call release_sock(sk);
before returning.

Fixes: 6919a8264a32 ("Crypto/chtls: add/delete TLS header in driver")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

MAINTAINERS: nfc: include linux-nfc mailing list

Keep all NFC related patches in existing linux-nfc@lists.01.org mailing
list.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Acked-by: Mark Greer <mgreer@animalcreek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

MAINTAINERS: nfc: add Krzysztof Kozlowski as maintainer

The NFC subsystem is orphaned.  I am happy to spend some cycles to
review the patches, send pull requests and in general keep the NFC
subsystem running.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Acked-by: Mark Greer <mgreer@animalcreek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

MAINTAINERS: nfc: drop Clément Perrochaud from NXP-NCI

Emails to Clément Perrochaud bounce with permanent error "user does not
exist", so remove Clément Perrochaud from NXP-NCI driver maintainers
entry.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Acked-by: Mark Greer <mgreer@animalcreek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'linux-can-fixes-for-5.13-20210512' of git://git./linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2021-05-12

this is a pull request of a single patch for net/master.

The patch is by Norbert Slusarek and it fixes a race condition in the
CAN ISO-TP socket between isotp_bind() and isotp_setsockopt().
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

ptp: ocp: Fix a resource leak in an error handling path

If an error occurs after a successful 'pci_ioremap_bar()' call, it must be
undone by a corresponding 'pci_iounmap()' call, as already done in the
remove function.

Fixes: a7e1abad13f3 ("ptp: Add clock driver for the OpenCompute TimeCard.")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

octeontx2-pf: fix a buffer overflow in otx2_set_rxfh_context()

This function is called from ethtool_set_rxfh() and "*rss_context"
comes from the user.  Add some bounds checking to prevent memory
corruption.

Fixes: 81a4362016e7 ("octeontx2-pf: Add RSS multi group support")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'fec-fixes'

Joakim Zhang says:

====================
net: fixes for fec driver

Two small fixes for fec driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: fec: add defer probe for of_get_mac_address

If MAC address read from nvmem efuse by calling .of_get_mac_address(),
but nvmem efuse is registered later than the driver, then it
return -EPROBE_DEFER value. So modify the driver to support
defer probe when read MAC address from nvmem efuse.

Signed-off-by: Fugang Duan <fugang.duan@nxp.com>
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fec: fix the potential memory leak in fec_enet_init()

If the memory allocated for cbd_base is failed, it should
free the memory allocated for the queues, otherwise it causes
memory leak.

And if the memory allocated for the queues is failed, it can
return error directly.

Fixes: 59d0f7465644 ("net: fec: init multi queue date structure")
Signed-off-by: Fugang Duan <fugang.duan@nxp.com>
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: packetmmap: fix only tx timestamp on request

The packetmmap tx ring should only return timestamps if requested via
setsockopt PACKET_TIMESTAMP, as documented. This allows compatibility
with non-timestamp aware user-space code which checks
tp_status == TP_STATUS_AVAILABLE; not expecting additional timestamp
flags to be set in tp_status.

Fixes: b9c32fb27170 ("packet: if hw/sw ts enabled in rx/tx ring, report which ts we got")
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Signed-off-by: Richard Sanger <rsanger@wand.net.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnxt_en: Fix and improve .ndo_features_check().

Jakub Kicinski pointed out that we need to handle ipv6 extension headers
and to explicitly check for supported tunnel types in
.ndo_features_check().

For ipv6 extension headers, the hardware supports up to 2 ext. headers
and each must be <= 64 bytes.  For tunneled packets, the supported
packets are UDP with supported VXLAN and Geneve ports, GRE, and IPIP.

v3: More improvements based on Alexander Duyck's valuable feedback -
    Remove the jump lable in bnxt_features_check() and restructure it
    so that the TCP/UDP is check is consolidated in bnxt_exthdr_check().

v2: Add missing step to check inner ipv6 header for UDP and GRE tunnels.
    Check TCP/UDP next header after skipping ipv6 ext headers for
    non-tunneled packets and for inner ipv6.
    (Both feedback from Alexander Duyck)

Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Fixes: 1698d600b361 ("bnxt_en: Implement .ndo_features_check().")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: really orphan skbs tied to closing sk

If the owing socket is shutting down - e.g. the sock reference
count already dropped to 0 and only sk_wmem_alloc is keeping
the sock alive, skb_orphan_partial() becomes a no-op.

When forwarding packets over veth with GRO enabled, the above
causes refcount errors.

This change addresses the issue with a plain skb_orphan() call
in the critical scenario.

Fixes: 9adc89af724f ("net: let skb_orphan_partial wake-up waiters.")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

can: isotp: prevent race between isotp_bind() and isotp_setsockopt()

A race condition was found in isotp_setsockopt() which allows to
change socket options after the socket was bound.
For the specific case of SF_BROADCAST support, this might lead to possible
use-after-free because can_rx_unregister() is not called.

Checking for the flag under the socket lock in isotp_bind() and taking
the lock in isotp_setsockopt() fixes the issue.

Fixes: 921ca574cd38 ("can: isotp: add SF_BROADCAST support for functional addressing")
Link: https://lore.kernel.org/r/trinity-e6ae9efa-9afb-4326-84c0-f3609b9b8168-1620773528307@3c-app-gmx-bs06
Reported-by: Norbert Slusarek <nslusarek@gmx.net>
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Signed-off-by: Norbert Slusarek <nslusarek@gmx.net>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

net: ipa: memory region array is variable size

IPA configuration data includes an array of memory region
descriptors.  That was a fixed-size array at one time, but
at some point we started defining it such that it was only
as big as required for a given platform.  The actual number
of entries in the array is recorded in the configuration data
along with the array.

A loop in ipa_mem_config() still assumes the array has entries
for all defined memory region IDs.  As a result, this loop can
go past the end of the actual array and attempt to write
"canary" values based on nonsensical data.

Fix this, by stashing the number of entries in the array, and
using that rather than IPA_MEM_COUNT in the initialization loop
found in ipa_mem_config().

The only remaining use of IPA_MEM_COUNT is in a validation check
to ensure configuration data doesn't have too many entries.
That's fine for now.

Fixes: 3128aae8c439a ("net: ipa: redefine struct ipa_mem_data")
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: fix ptp support config breakage

When IONIC=y and PTP_1588_CLOCK=m were set in the .config file
the driver link failed with undefined references.

We add the dependancy
depends on PTP_1588_CLOCK || !PTP_1588_CLOCK
to clear this up.

If PTP_1588_CLOCK=m, the depends limits IONIC to =m (or disabled).
If PTP_1588_CLOCK is disabled, IONIC can be any of y/m/n.

Fixes: 61db421da31b ("ionic: link in the new hw timestamp code")
Reported-by: kernel test robot <lkp@intel.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Allen Hubbe <allenbh@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

mptcp: fix data stream corruption

Maxim reported several issues when forcing a TCP transparent proxy
to use the MPTCP protocol for the inbound connections. He also
provided a clean reproducer.

The problem boils down to 'mptcp_frag_can_collapse_to()' assuming
that only MPTCP will use the given page_frag.

If others - e.g. the plain TCP protocol - allocate page fragments,
we can end-up re-using already allocated memory for mptcp_data_frag.

Fix the issue ensuring that the to-be-expanded data fragment is
located at the current page frag end.

v1 -> v2:
 - added missing fixes tag (Mat)

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/178
Reported-and-tested-by: Maxim Galaganov <max@internet.ru>
Fixes: 18b683bff89d ("mptcp: queue data for mptcp level retransmission")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2021-05-11

The following pull-request contains BPF updates for your *net* tree.

We've added 13 non-merge commits during the last 8 day(s) which contain
a total of 21 files changed, 817 insertions(+), 382 deletions(-).

The main changes are:

1) Fix multiple ringbuf bugs in particular to prevent writable mmap of
   read-only pages, from Andrii Nakryiko & Thadeu Lima de Souza Cascardo.

2) Fix verifier alu32 known-const subregister bound tracking for bitwise
   operations and/or/xor, from Daniel Borkmann.

3) Reject trampoline attachment for functions with variable arguments,
   and also add a deny list of other forbidden functions, from Jiri Olsa.

4) Fix nested bpf_bprintf_prepare() calls used by various helpers by
   switching to per-CPU buffers, from Florent Revest.

5) Fix kernel compilation with BTF debug info on ppc64 due to pahole
   missing TCP-CC functions like cubictcp_init, from Martin KaFai Lau.

6) Add a kconfig entry to provide an option to disallow unprivileged
   BPF by default, from Daniel Borkmann.

7) Fix libbpf compilation for older libelf when GELF_ST_VISIBILITY()
   macro is not available, from Arnaldo Carvalho de Melo.

8) Migrate test_tc_redirect to test_progs framework as prep work
   for upcoming skb_change_head() fix & selftest, from Jussi Maki.

9) Fix a libbpf segfault in add_dummy_ksym_var() if BTF is not
   present, from Ian Rogers.

10) Fix tx_only micro-benchmark in xdpsock BPF sample with proper frame
    size, from Magnus Karlsson.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'mac80211-for-net-2021-05-11' of git://git./linux/kernel/git/jberg/mac80211

Johannes Berg says:

====================
pull-request: mac80211 2021-05-11

So exciting times, for the first pull request for fixes I
have a bunch of security things that have been under embargo
for a while - see more details in the tag below, and at the
patch posting message I linked to.

I organized with Kalle to just have a single set of fixes
for mac80211 and ath10k/ath11k, we don't know about any of
the other vendors (the mac80211 + already released firmware
is sufficient to fix iwlwifi.)

Please pull and let me know if there's any problem.

Several security issues in the 802.11 implementations were found by
Mathy Vanhoef (New York University Abu Dhabi), and this contains the
fixes developed for mac80211 and specifically Qualcomm drivers, I'm
sending this together (as agreed with Kalle) to have just a single
set of patches for now. We don't know about other vendors though.

More details in the patch posting:
https://lore.kernel.org/r/20210511180259.159598-1-johannes@sipsolutions.net
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Fix MAC WoL not working if PHY does not support WoL

Both get and set WoL will check device_can_wakeup(), if MAC supports PMT, it
will set device wakeup capability. After commit 1d8e5b0f3f2c ("net: stmmac:
Support WOL with phy"), device wakeup capability will be overwrite in
stmmac_init_phy() according to phy's Wol feature. If phy doesn't support WoL,
then MAC will lose wakeup capability. To fix this issue, only overwrite device
wakeup capability when MAC doesn't support PMT.

For STMMAC now driver checks MAC's WoL capability if MAC supports PMT, if
not support, driver will check PHY's WoL capability.

Fixes: 1d8e5b0f3f2c ("net: stmmac: Support WOL with phy")
Reviewed-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bpf: Limit static tcp-cc functions in the .BTF_ids list to x86

During the discussion in [0]. It was pointed out that static functions
in ppc64 is prefixed with ".". For example, the 'readelf -s vmlinux.ppc':

  89326: c000000001383280    24 NOTYPE  LOCAL  DEFAULT   31 cubictcp_init
  89327: c000000000c97c50   168 FUNC    LOCAL  DEFAULT    2 .cubictcp_init

The one with FUNC type is ".cubictcp_init" instead of "cubictcp_init".
The "." seems to be done by arch/powerpc/include/asm/ppc_asm.h.

This caused that pahole cannot generate the BTF for these tcp-cc kernel
functions because pahole only captures the FUNC type and "cubictcp_init"
is not. It then failed the kernel compilation in ppc64.

This behavior is only reported in ppc64 so far. I tried arm64, s390,
and sparc64 and did not observe this "." prefix and NOTYPE behavior.

Since the kfunc call is only supported in the x86_64 and x86_32 JIT,
this patch limits those tcp-cc functions to x86 only to avoid unnecessary
compilation issue in other ARCHs. In the future, we can examine if it
is better to change all those functions from static to extern.

  [0] https://lore.kernel.org/bpf/4e051459-8532-7b61-c815-f3435767f8a0@kernel.org/

Fixes: e78aea8b2170 ("bpf: tcp: Put some tcp cong functions in allowlist for bpf-tcp-cc")
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Michal Suchánek <msuchanek@suse.de>
Cc: Jiri Slaby <jslaby@suse.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: https://lore.kernel.org/bpf/20210508005011.3863757-1-kafai@fb.com

selftests/bpf: Rewrite test_tc_redirect.sh as prog_tests/tc_redirect.c

As discussed in [0], this ports test_tc_redirect.sh to the test_progs
framework and removes the old test.

This makes it more in line with rest of the tests and makes it possible
to run this test case with vmtest.sh and under the bpf CI.

The upcoming skb_change_head() helper fix in [0] is depending on it and
extending the test case to redirect a packet from L3 device to veth.

  [0] https://lore.kernel.org/bpf/20210427135550.807355-1-joamaki@gmail.com

Signed-off-by: Jussi Maki <joamaki@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20210505085925.783985-1-joamaki@gmail.com

libbpf: Provide GELF_ST_VISIBILITY() define for older libelf

Where that macro isn't available.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/YJaspEh0qZr4LYOc@kernel.org

bpf: Fix nested bpf_bprintf_prepare with more per-cpu buffers

The bpf_seq_printf, bpf_trace_printk and bpf_snprintf helpers share one
per-cpu buffer that they use to store temporary data (arguments to
bprintf). They "get" that buffer with try_get_fmt_tmp_buf and "put" it
by the end of their scope with bpf_bprintf_cleanup.

If one of these helpers gets called within the scope of one of these
helpers, for example: a first bpf program gets called, uses
bpf_trace_printk which calls raw_spin_lock_irqsave which is traced by
another bpf program that calls bpf_snprintf, then the second "get"
fails. Essentially, these helpers are not re-entrant. They would return
-EBUSY and print a warning message once.

This patch triples the number of bprintf buffers to allow three levels
of nesting. This is very similar to what was done for tracepoints in
"9594dc3c7e7 bpf: fix nested bpf tracepoints with per-cpu data"

Fixes: d9c9e4db186a ("bpf: Factorize bpf_trace_printk and bpf_seq_printf")
Reported-by: syzbot+63122d0bc347f18c1884@syzkaller.appspotmail.com
Signed-off-by: Florent Revest <revest@chromium.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210511081054.2125874-1-revest@chromium.org

bpf: Add deny list of btf ids check for tracing programs

The recursion check in __bpf_prog_enter and __bpf_prog_exit
leaves some (not inlined) functions unprotected:

In __bpf_prog_enter:
  - migrate_disable is called before prog->active is checked

In __bpf_prog_exit:
  - migrate_enable,rcu_read_unlock_strict are called after
    prog->active is decreased

When attaching trampoline to them we get panic like:

  traps: PANIC: double fault, error_code: 0x0
  double fault: 0000 [#1] SMP PTI
  RIP: 0010:__bpf_prog_enter+0x4/0x50
  ...
  Call Trace:
   <IRQ>
   bpf_trampoline_6442466513_0+0x18/0x1000
   migrate_disable+0x5/0x50
   __bpf_prog_enter+0x9/0x50
   bpf_trampoline_6442466513_0+0x18/0x1000
   migrate_disable+0x5/0x50
   __bpf_prog_enter+0x9/0x50
   bpf_trampoline_6442466513_0+0x18/0x1000
   migrate_disable+0x5/0x50
   __bpf_prog_enter+0x9/0x50
   bpf_trampoline_6442466513_0+0x18/0x1000
   migrate_disable+0x5/0x50
   ...

Fixing this by adding deny list of btf ids for tracing
programs and checking btf id during program verification.
Adding above functions to this list.

Suggested-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210429114712.43783-1-jolsa@kernel.org

bpf: Add kconfig knob for disabling unpriv bpf by default

Add a kconfig knob which allows for unprivileged bpf to be disabled by default.
If set, the knob sets /proc/sys/kernel/unprivileged_bpf_disabled to value of 2.

This still allows a transition of 2 -> {0,1} through an admin. Similarly,
this also still keeps 1 -> {1} behavior intact, so that once set to permanently
disabled, it cannot be undone aside from a reboot.

We've also added extra2 with max of 2 for the procfs handler, so that an admin
still has a chance to toggle between 0 <-> 2.

Either way, as an additional alternative, applications can make use of CAP_BPF
that we added a while ago.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/74ec548079189e4e4dffaeb42b8987bb3c852eee.1620765074.git.daniel@iogearbox.net