i40e: Fix adding unsupported cloud filters

[ Upstream commit 4e20655e503e3a478cd1682bf25e3202dd823da8 ]

If a VF tries to add unsupported cloud filter through virtchnl
then i40e_add_del_cloud_filter(_big_buf) returns -ENOTSUPP but
this error code is stored in 'ret' instead of 'aq_ret' that
is used as error code sent back to VF. In this scenario where
one of the mentioned functions fails the value of 'aq_ret'
is zero so the VF will incorrectly receive a 'success'.

Use 'aq_ret' to store return value and remove 'ret' local
variable. Additionally fix the issue when filter allocation
fails, in this case no notification is sent back to the VF.

Fixes: e284fc280473 ("i40e: Add and delete cloud filter")
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20231121211338.3348677-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

amd-xgbe: propagate the correct speed and duplex status

[ Upstream commit 7a2323ac24a50311f64a3a9b54ed5bef5821ecae ]

xgbe_get_link_ksettings() does not propagate correct speed and duplex
information to ethtool during cable unplug. Due to which ethtool reports
incorrect values for speed and duplex.

Address this by propagating correct information.

Fixes: 7c12aa08779c ("amd-xgbe: Move the PHY support into amd-xgbe")
Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

amd-xgbe: handle the corner-case during tx completion

[ Upstream commit 7121205d5330c6a3cb3379348886d47c77b78d06 ]

The existing implementation uses software logic to accumulate tx
completions until the specified time (1ms) is met and then poll them.
However, there exists a tiny gap which leads to a race between
resetting and checking the tx_activate flag. Due to this the tx
completions are not reported to upper layer and tx queue timeout
kicks-in restarting the device.

To address this, introduce a tx cleanup mechanism as part of the
periodic maintenance process.

Fixes: c5aa9e3b8156 ("amd-xgbe: Initial AMD 10GbE platform driver")
Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

amd-xgbe: handle corner-case during sfp hotplug

[ Upstream commit 676ec53844cbdf2f47e68a076cdff7f0ec6cbe3f ]

Force the mode change for SFI in Fixed PHY configurations. Fixed PHY
configurations needs PLL to be enabled while doing mode set. When the
SFP module isn't connected during boot, driver assumes AN is ON and
attempts auto-negotiation. However, if the connected SFP comes up in
Fixed PHY configuration the link will not come up as PLL isn't enabled
while the initial mode set command is issued. So, force the mode change
for SFI in Fixed PHY configuration to fix link issues.

Fixes: e57f7a3feaef ("amd-xgbe: Prepare for working with more than one type of phy")
Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: veth: fix ethtool stats reporting

[ Upstream commit 818ad9cc90d4a7165caaee7e32800c50d0564ec3 ]

Fix a possible misalignment between page_pool stats and tx xdp_stats
reported in veth_get_ethtool_stats routine.
The issue can be reproduced configuring the veth pair with the
following tx/rx queues:

$ip link add v0 numtxqueues 2 numrxqueues 4 type veth peer name v1 \
 numtxqueues 1 numrxqueues 1

and loading a simple XDP program on v0 that just returns XDP_PASS.
In this case on v0 the page_pool stats overwrites tx xdp_stats for queue 1.
Fix the issue incrementing pp_idx of dev->real_num_tx_queues * VETH_TQ_STATS_LEN
since we always report xdp_stats for all tx queues in ethtool.

Fixes: 4fc418053ec7 ("net: veth: add page_pool stats")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/c5b5d0485016836448453f12846c7c4ab75b094a.1700593593.git.lorenzo@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-pf: Fix ntuple rule creation to direct packet to VF with higher Rx queue than its PF

[ Upstream commit 4aa1d8f89b10cdc25a231dabf808d8935e0b137a ]

It is possible to add a ntuple rule which would like to direct packet to
a VF whose number of queues are greater/less than its PF's queue numbers.
For example a PF can have 2 Rx queues but a VF created on that PF can have
8 Rx queues. As of today, ntuple rule will reject rule because it is
checking the requested queue number against PF's number of Rx queues.
As a part of this fix if the action of a ntuple rule is to move a packet
to a VF's queue then the check is removed. Also, a debug information is
printed to aware user that it is user's responsibility to cross check if
the requested queue number on that VF is a valid one.

Fixes: f0a1913f8a6f ("octeontx2-pf: Add support for ethtool ntuple filters")
Signed-off-by: Suman Ghosh <sumang@marvell.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20231121165624.3664182-1-sumang@marvell.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm/xen: fix xen_vcpu_info allocation alignment

[ Upstream commit 7bf9a6b46549852a37e6d07e52c601c3c706b562 ]

xen_vcpu_info is a percpu area than needs to be mapped by Xen.
Currently, it could cross a page boundary resulting in Xen being unable
to map it:

[    0.567318] kernel BUG at arch/arm64/xen/../../arm/xen/enlighten.c:164!
[    0.574002] Internal error: Oops - BUG: 00000000f2000800 [#1] PREEMPT SMP

Fix the issue by using __alloc_percpu and requesting alignment for the
memory allocation.

Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>
Link: https://lore.kernel.org/r/alpine.DEB.2.22.394.2311221501340.2053963@ubuntu-linux-20-04-desktop
Fixes: 24d5373dda7c ("arm/xen: Use alloc_percpu rather than __alloc_percpu")
Reviewed-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: mm: Fix "rodata=on" when CONFIG_RODATA_FULL_DEFAULT_ENABLED=y

[ Upstream commit acfa60dbe03802d6afd28401aa47801270e82021 ]

When CONFIG_RODATA_FULL_DEFAULT_ENABLED=y, passing "rodata=on" on the
kernel command-line (rather than "rodata=full") should turn off the
"full" behaviour, leaving writable linear aliases of read-only kernel
memory. Unfortunately, the option has no effect in this situation and
the only way to disable the "rodata=full" behaviour is to disable rodata
protection entirely by passing "rodata=off".

Fix this by parsing the "on" and "off" options in the arch code,
additionally enforcing that 'rodata_full' cannot be set without also
setting 'rodata_enabled', allowing us to simplify a couple of checks
in the process.

Fixes: 2e8cff0a0eee ("arm64: fix rodata=full")
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Reviewed-by: "Russell King (Oracle)" <rmk+kernel@armlinux.org.uk>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20231117131422.29663-1-will@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/ipl: add missing IPL_TYPE_ECKD_DUMP case to ipl_init()

[ Upstream commit 673752a839694133a328610fcbc54f3d59ae87f3 ]

Add missing IPL_TYPE_ECKD_DUMP case to ipl_init() creating
ECKD ipl device attribute group similar to IPL_TYPE_ECKD case.
Commit e2d2a2968f2a ("s390/ipl: add eckd dump support") should
have had it from the beginning.

Fixes: e2d2a2968f2a ("s390/ipl: add eckd dump support")
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Reviewed-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/smc: avoid data corruption caused by decline

[ Upstream commit e6d71b437abc2f249e3b6a1ae1a7228e09c6e563 ]

We found a data corruption issue during testing of SMC-R on Redis
applications.

The benchmark has a low probability of reporting a strange error as
shown below.

"Error: Protocol error, got "\xe2" as reply type byte"

Finally, we found that the retrieved error data was as follows:

0xE2 0xD4 0xC3 0xD9 0x04 0x00 0x2C 0x20 0xA6 0x56 0x00 0x16 0x3E 0x0C
0xCB 0x04 0x02 0x01 0x00 0x00 0x20 0x00 0x00 0x00 0x00 0x00 0x00 0x00
0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0xE2

It is quite obvious that this is a SMC DECLINE message, which means that
the applications received SMC protocol message.
We found that this was caused by the following situations:

client                  server
        ¦  clc proposal
        ------------->
        ¦  clc accept
        <-------------
        ¦  clc confirm
        ------------->
wait llc confirm
send llc confirm
        ¦failed llc confirm
        ¦   x------
(after 2s)timeout
                        wait llc confirm rsp

wait decline

(after 1s) timeout
                        (after 2s) timeout
        ¦   decline
        -------------->
        ¦   decline
        <--------------

As a result, a decline message was sent in the implementation, and this
message was read from TCP by the already-fallback connection.

This patch double the client timeout as 2x of the server value,
With this simple change, the Decline messages should never cross or
collide (during Confirm link timeout).

This issue requires an immediate solution, since the protocol updates
involve a more long-term solution.

Fixes: 0fb0b02bd6fd ("net/smc: adapt SMC client code to use the LLC flow")
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Reviewed-by: Wen Gu <guwen@linux.alibaba.com>
Reviewed-by: Wenjia Zhang <wenjia@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: usb: ax88179_178a: fix failed operations during ax88179_reset

[ Upstream commit 0739af07d1d947af27c877f797cb82ceee702515 ]

Using generic ASIX Electronics Corp. AX88179 Gigabit Ethernet device,
the following test cycle has been implemented:
    - power on
    - check logs
    - shutdown
    - after detecting the system shutdown, disconnect power
    - after approximately 60 seconds of sleep, power is restored
Running some cycles, sometimes error logs like this appear:
    kernel: ax88179_178a 2-9:1.0 (unnamed net_device) (uninitialized): Failed to write reg index 0x0001: -19
    kernel: ax88179_178a 2-9:1.0 (unnamed net_device) (uninitialized): Failed to read reg index 0x0001: -19
    ...
These failed operation are happening during ax88179_reset execution, so
the initialization could not be correct.

In order to avoid this, we need to increase the delay after reset and
clock initial operations. By using these larger values, many cycles
have been run and no failed operations appear.

It would be better to check some status register to verify when the
operation has finished, but I do not have found any available information
(neither in the public datasheets nor in the manufacturer's driver). The
only available information for the necessary delays is the maufacturer's
driver (original values) but the proposed values are not enough for the
tested devices.

Fixes: e2ca90c276e1f ("ax88179_178a: ASIX AX88179_178A USB 3.0/2.0 to gigabit ethernet adapter driver")
Reported-by: Herb Wei <weihao.bj@ieisystem.com>
Tested-by: Herb Wei <weihao.bj@ieisystem.com>
Signed-off-by: Jose Ignacio Tornos Martinez <jtornosm@redhat.com>
Link: https://lore.kernel.org/r/20231120120642.54334-1-jtornosm@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/panel: boe-tv101wum-nl6: Fine tune Himax83102-j02 panel HFP and HBP

[ Upstream commit cea7008190ad65b4aaae6e94667a358d2c10a696 ]

The refresh reported by modetest is 60.46Hz, and the actual measurement
is 60.01Hz, which is outside the expected tolerance. Adjust hporch and
pixel clock to fix it. After repair, modetest and actual measurement were
all 60.01Hz.

Modetest refresh = Pixel CLK/ htotal* vtotal, but measurement frame rate
is HS->LP cycle time(Vblanking). Measured frame rate is not only affecte
by Htotal/Vtotal/pixel clock, also affected by Lane-num/PixelBit/LineTime
/DSI CLK. Assume that the DSI controller could not make the mode that we
requested(presumably it's PLL couldn't generate the exact pixel clock?).
If you use a different DSI controller, you may need to readjust these
parameters. Now this panel looks like it's only used by me on the MTK
platform, so let's change this set of parameters.

Fixes: 1bc2ef065f13 ("drm/panel: Support for Starry-himax83102-j02 TDDI MIPI-DSI panel")
Signed-off-by: Cong Yang <yangcong5@huaqin.corp-partner.google.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20231120020109.3216343-1-yangcong5@huaqin.corp-partner.google.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ipv4: Correct/silence an endian warning in __ip_do_redirect

[ Upstream commit c0e2926266af3b5acf28df0a8fc6e4d90effe0bb ]

net/ipv4/route.c:783:46: warning: incorrect type in argument 2 (different base types)
net/ipv4/route.c:783:46:    expected unsigned int [usertype] key
net/ipv4/route.c:783:46:    got restricted __be32 [usertype] new_gw

Fixes: 969447f226b4 ("ipv4: use new_gw for redirect neigh lookup")
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
Link: https://lore.kernel.org/r/20231119141759.420477-1-chentao@kylinos.cn
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

HID: fix HID device resource race between HID core and debugging support

[ Upstream commit fc43e9c857b7aa55efba9398419b14d9e35dcc7d ]

hid_debug_events_release releases resources bound to the HID device instance.
hid_device_release releases the underlying HID device instance potentially
before hid_debug_events_release has completed releasing debug resources bound
to the same HID device instance.

Reference count to prevent the HID device instance from being torn down
preemptively when HID debugging support is used. When count reaches zero,
release core resources of HID device instance using hiddev_free.

The crash:

[  120.728477][ T4396] kernel BUG at lib/list_debug.c:53!
[  120.728505][ T4396] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[  120.739806][ T4396] Modules linked in: bcmdhd dhd_static_buf 8822cu pcie_mhi r8168
[  120.747386][ T4396] CPU: 1 PID: 4396 Comm: hidt_bridge Not tainted 5.10.110 #257
[  120.754771][ T4396] Hardware name: Rockchip RK3588 EVB4 LP4 V10 Board (DT)
[  120.761643][ T4396] pstate: 60400089 (nZCv daIf +PAN -UAO -TCO BTYPE=--)
[  120.768338][ T4396] pc : __list_del_entry_valid+0x98/0xac
[  120.773730][ T4396] lr : __list_del_entry_valid+0x98/0xac
[  120.779120][ T4396] sp : ffffffc01e62bb60
[  120.783126][ T4396] x29: ffffffc01e62bb60 x28: ffffff818ce3a200
[  120.789126][ T4396] x27: 0000000000000009 x26: 0000000000980000
[  120.795126][ T4396] x25: ffffffc012431000 x24: ffffff802c6d4e00
[  120.801125][ T4396] x23: ffffff8005c66f00 x22: ffffffc01183b5b8
[  120.807125][ T4396] x21: ffffff819df2f100 x20: 0000000000000000
[  120.813124][ T4396] x19: ffffff802c3f0700 x18: ffffffc01d2cd058
[  120.819124][ T4396] x17: 0000000000000000 x16: 0000000000000000
[  120.825124][ T4396] x15: 0000000000000004 x14: 0000000000003fff
[  120.831123][ T4396] x13: ffffffc012085588 x12: 0000000000000003
[  120.837123][ T4396] x11: 00000000ffffbfff x10: 0000000000000003
[  120.843123][ T4396] x9 : 455103d46b329300 x8 : 455103d46b329300
[  120.849124][ T4396] x7 : 74707572726f6320 x6 : ffffffc0124b8cb5
[  120.855124][ T4396] x5 : ffffffffffffffff x4 : 0000000000000000
[  120.861123][ T4396] x3 : ffffffc011cf4f90 x2 : ffffff81fee7b948
[  120.867122][ T4396] x1 : ffffffc011cf4f90 x0 : 0000000000000054
[  120.873122][ T4396] Call trace:
[  120.876259][ T4396]  __list_del_entry_valid+0x98/0xac
[  120.881304][ T4396]  hid_debug_events_release+0x48/0x12c
[  120.886617][ T4396]  full_proxy_release+0x50/0xbc
[  120.891323][ T4396]  __fput+0xdc/0x238
[  120.895075][ T4396]  ____fput+0x14/0x24
[  120.898911][ T4396]  task_work_run+0x90/0x148
[  120.903268][ T4396]  do_exit+0x1bc/0x8a4
[  120.907193][ T4396]  do_group_exit+0x8c/0xa4
[  120.911458][ T4396]  get_signal+0x468/0x744
[  120.915643][ T4396]  do_signal+0x84/0x280
[  120.919650][ T4396]  do_notify_resume+0xd0/0x218
[  120.924262][ T4396]  work_pending+0xc/0x3f0

[ Rahul Rameshbabu <sergeantsagara@protonmail.com>: rework changelog ]
Fixes: cd667ce24796 ("HID: use debugfs for events/reports dumping")
Signed-off-by: Charles Yi <be286@163.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Sasha Levin <sashal@kernel.org>

accel/ivpu/37xx: Fix hangs related to MMIO reset

[ Upstream commit 3f7c0634926daf48cd2f6db6c1197a1047074088 ]

There is no need to call MMIO reset using VPU_37XX_BUTTRESS_VPU_IP_RESET
register. IP will be reset by FLR or by entering d0i3. Also IP reset
during power_up is not needed as the VPU is already in reset.

Removing MMIO reset improves stability as it a partial device reset
that is not safe in some corner cases.

This change also brings back ivpu_boot_pwr_domain_disable() that
helps to properly power down VPU when it is hung by a buggy workload.

Signed-off-by: Jacek Lawrynowicz <jacek.lawrynowicz@linux.intel.com>
Fixes: 828d63042aec ("accel/ivpu: Don't enter d0i3 during FLR")
Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231115111004.1304092-1-jacek.lawrynowicz@linux.intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

accel/ivpu: Do not initialize parameters on power up

[ Upstream commit f956bf2080862cfc97412e1eaa08689bc9838d20 ]

Initialize HW specific parameters only once. We do not have to do this
on every power_up (performed during initialization and on resume). Move
corresponding code to ->info_init()

Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231020104501.697763-6-stanislaw.gruszka@linux.intel.com
Stable-dep-of: 3f7c0634926d ("accel/ivpu/37xx: Fix hangs related to MMIO reset")
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Fix dev's rx stats for bpf_redirect_peer traffic

[ Upstream commit 024ee930cb3c9ae49e4266aee89cfde0ebb407e1 ]

Traffic redirected by bpf_redirect_peer() (used by recent CNIs like Cilium)
is not accounted for in the RX stats of supported devices (that is, veth
and netkit), confusing user space metrics collectors such as cAdvisor [0],
as reported by Youlun.

Fix it by calling dev_sw_netstats_rx_add() in skb_do_redirect(), to update
RX traffic counters. Devices that support ndo_get_peer_dev _must_ use the
@tstats per-CPU counters (instead of @lstats, or @dstats).

To make this more fool-proof, error out when ndo_get_peer_dev is set but
@tstats are not selected.

  [0] Specifically, the "container_network_receive_{byte,packet}s_total"
      counters are affected.

Fixes: 9aa1206e8f48 ("bpf: Add redirect_peer helper")
Reported-by: Youlun Zhang <zhangyoulun@bytedance.com>
Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://lore.kernel.org/r/20231114004220.6495-6-daniel@iogearbox.net
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: Move {l,t,d}stats allocation to core and convert veth & vrf

[ Upstream commit 34d21de99cea9cb17967874313e5b0262527833c ]

Move {l,t,d}stats allocation to the core and let netdevs pick the stats
type they need. That way the driver doesn't have to bother with error
handling (allocation failure checking, making sure free happens in the
right spot, etc) - all happening in the core.

Co-developed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Cc: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20231114004220.6495-3-daniel@iogearbox.net
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Stable-dep-of: 024ee930cb3c ("bpf: Fix dev's rx stats for bpf_redirect_peer traffic")
Signed-off-by: Sasha Levin <sashal@kernel.org>

net, vrf: Move dstats structure to core

[ Upstream commit 79e0c5be8c73a674c92bd4ba77b75f4f8c91d32e ]

Just move struct pcpu_dstats out of the vrf into the core, and streamline
the field names slightly, so they better align with the {t,l}stats ones.

No functional change otherwise. A conversion of the u64s to u64_stats_t
could be done at a separate point in future. This move is needed as we are
moving the {t,l,d}stats allocation/freeing to the core.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20231114004220.6495-2-daniel@iogearbox.net
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Stable-dep-of: 024ee930cb3c ("bpf: Fix dev's rx stats for bpf_redirect_peer traffic")
Signed-off-by: Sasha Levin <sashal@kernel.org>

PM: tools: Fix sleepgraph syntax error

[ Upstream commit b85e2dab33ce467e8dcf1cb6c0c587132ff17f56 ]

The sleepgraph tool currently fails:

  File "/usr/bin/sleepgraph", line 4155
    or re.match('psci: CPU(?P<cpu>[0-9]*) killed.*', msg)):
                                                         ^
SyntaxError: unmatched ')'

Fixes: 34ea427e01ea ("PM: tools: sleepgraph: Recognize "CPU killed" messages")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/rockchip: vop: Fix color for RGB888/BGR888 format on VOP full

[ Upstream commit bb0a05acd6121ff0e810b44fdc24dbdfaa46b642 ]

Use of DRM_FORMAT_RGB888 and DRM_FORMAT_BGR888 on e.g. RK3288, RK3328
and RK3399 result in wrong colors being displayed.

The issue can be observed using modetest:

  modetest -s <connector_id>@<crtc_id>:1920x1080-60@RG24
  modetest -s <connector_id>@<crtc_id>:1920x1080-60@BG24

Vendor 4.4 kernel apply an inverted rb swap for these formats on VOP
full framework (IP version 3.x) compared to VOP little framework (2.x).

Fix colors by applying different rb swap for VOP full framework (3.x)
and VOP little framework (2.x) similar to vendor 4.4 kernel.

Fixes: 85a359f25388 ("drm/rockchip: Add BGR formats to VOP")
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Tested-by: Diederik de Haas <didi.debian@cknow.org>
Reviewed-by: Christopher Obbard <chris.obbard@collabora.com>
Tested-by: Christopher Obbard <chris.obbard@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20231026191500.2994225-1-jonas@kwiboo.se
Signed-off-by: Sasha Levin <sashal@kernel.org>

libfs: getdents() should return 0 after reaching EOD

[ Upstream commit 796432efab1e372d404e7a71cc6891a53f105051 ]

The new directory offset helpers don't conform with the convention
of getdents() returning no more entries once a directory file
descriptor has reached the current end-of-directory.

To address this, copy the logic from dcache_readdir() to mark the
open directory file descriptor once EOD has been reached. Seeking
resets the mark.

Reported-by: Tavian Barnes <tavianator@tavianator.com>
Closes: https://lore.kernel.org/linux-fsdevel/20231113180616.2831430-1-tavianator@tavianator.com/
Fixes: 6faddda69f62 ("libfs: Add directory operations for stable offsets")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Link: https://lore.kernel.org/r/170043792492.4628.15646203084646716134.stgit@bazille.1015granger.net
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

block: update the stable_writes flag in bdev_add

[ Upstream commit 1898efcdbed32bb1c67269c985a50bab0dbc9493 ]

Propagate the per-queue stable_write flags into each bdev inode in bdev_add.
This makes sure devices that require stable writes have it set for I/O
on the block device node as well.

Note that this doesn't cover the case of a flag changing on a live device
yet.  We should handle that as well, but I plan to cover it as part of a
more general rework of how changing runtime paramters on block devices
works.

Fixes: 1cb039f3dc16 ("bdi: replace BDI_CAP_STABLE_WRITES with a queue and a sb flag")
Reported-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20231025141020.192413-3-hch@lst.de
Tested-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

filemap: add a per-mapping stable writes flag

[ Upstream commit 762321dab9a72760bf9aec48362f932717c9424d ]

folio_wait_stable waits for writeback to finish before modifying the
contents of a folio again, e.g. to support check summing of the data
in the block integrity code.

Currently this behavior is controlled by the SB_I_STABLE_WRITES flag
on the super_block, which means it is uniform for the entire file system.
This is wrong for the block device pseudofs which is shared by all
block devices, or file systems that can use multiple devices like XFS
witht the RT subvolume or btrfs (although btrfs currently reimplements
folio_wait_stable anyway).

Add a per-address_space AS_STABLE_WRITES flag to control the behavior
in a more fine grained way.  The existing SB_I_STABLE_WRITES is kept
to initialize AS_STABLE_WRITES to the existing default which covers
most cases.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20231025141020.192413-2-hch@lst.de
Tested-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Stable-dep-of: 1898efcdbed3 ("block: update the stable_writes flag in bdev_add")
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/i915: do not clean GT table on error path

[ Upstream commit 0561794b6b642b84b879bf97061c4b4fa692839e ]

The only task of intel_gt_release_all is to zero gt table. Calling
it on error path prevents intel_gt_driver_late_release_all (called from
i915_driver_late_release) to cleanup GTs, causing leakage.
After i915_driver_late_release GT array is not used anymore so
it does not need cleaning at all.

Sample leak report:

BUG i915_request (...): Objects remaining in i915_request on __kmem_cache_shutdown()
...
Object 0xffff888113420040 @offset=64
Allocated in __i915_request_create+0x75/0x610 [i915] age=18339 cpu=1 pid=1454
 kmem_cache_alloc+0x25b/0x270
 __i915_request_create+0x75/0x610 [i915]
 i915_request_create+0x109/0x290 [i915]
 __engines_record_defaults+0xca/0x440 [i915]
 intel_gt_init+0x275/0x430 [i915]
 i915_gem_init+0x135/0x2c0 [i915]
 i915_driver_probe+0x8d1/0xdc0 [i915]

v2: removed whole intel_gt_release_all

Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8489
Fixes: bec68cc9ea42 ("drm/i915: Prepare for multiple GTs")
Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231115-dont_clean_gt_on_error_path-v2-1-54250125470a@intel.com
(cherry picked from commit e899505533852bf1da133f2f4c9a9655ff77f7e5)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ata: pata_isapnp: Add missing error check for devm_ioport_map()

[ Upstream commit a6925165ea82b7765269ddd8dcad57c731aa00de ]

Add missing error return check for devm_ioport_map() and return the
error if this function call fails.

Fixes: 0d5ff566779f ("libata: convert to iomap")
Signed-off-by: Chen Ni <nichen@iscas.ac.cn>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-pf: Fix memory leak during interface down

[ Upstream commit 5f228d7c8a539714c1e9b7e7534f76bb7979f268 ]

During 'ifconfig <netdev> down' one RSS memory was not getting freed.
This patch fixes the same.

Fixes: 81a4362016e7 ("octeontx2-pf: Add RSS multi group support")
Signed-off-by: Suman Ghosh <sumang@marvell.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wireguard: use DEV_STATS_INC()

[ Upstream commit 93da8d75a66568ba4bb5b14ad2833acd7304cd02 ]

wg_xmit() can be called concurrently, KCSAN reported [1]
some device stats updates can be lost.

Use DEV_STATS_INC() for this unlikely case.

[1]
BUG: KCSAN: data-race in wg_xmit / wg_xmit

read-write to 0xffff888104239160 of 8 bytes by task 1375 on cpu 0:
wg_xmit+0x60f/0x680 drivers/net/wireguard/device.c:231
__netdev_start_xmit include/linux/netdevice.h:4918 [inline]
netdev_start_xmit include/linux/netdevice.h:4932 [inline]
xmit_one net/core/dev.c:3543 [inline]
dev_hard_start_xmit+0x11b/0x3f0 net/core/dev.c:3559
...

read-write to 0xffff888104239160 of 8 bytes by task 1378 on cpu 1:
wg_xmit+0x60f/0x680 drivers/net/wireguard/device.c:231
__netdev_start_xmit include/linux/netdevice.h:4918 [inline]
netdev_start_xmit include/linux/netdevice.h:4932 [inline]
xmit_one net/core/dev.c:3543 [inline]
dev_hard_start_xmit+0x11b/0x3f0 net/core/dev.c:3559
...

v2: also change wg_packet_consume_data_done() (Hangbin Liu)
    and wg_packet_purge_staged_packets()

Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: wangxun: fix kernel panic due to null pointer

[ Upstream commit 8ba2c459668cfe2aaacc5ebcd35b4b9ef8643013 ]

When the device uses a custom subsystem vendor ID, the function
wx_sw_init() returns before the memory of 'wx->mac_table' is allocated.
The null pointer will causes the kernel panic.

Fixes: 79625f45ca73 ("net: wangxun: Move MAC address handling to libwx")
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/panel: simple: Fix Innolux G101ICE-L01 timings

[ Upstream commit 3f9a91b6c00e655d27bd785dcda1742dbdc31bda ]

The Innolux G101ICE-L01 datasheet [1] page 17 table
6.1 INPUT SIGNAL TIMING SPECIFICATIONS
indicates that maximum vertical blanking time is 40 lines.
Currently the driver uses 29 lines.

Fix it, and since this panel is a DE panel, adjust the timings
to make them less hostile to controllers which cannot do 1 px
HSA/VSA, distribute the delays evenly between all three parts.

[1] https://www.data-modul.com/sites/default/files/products/G101ICE-L01-C2-specification-12042389.pdf

Fixes: 1e29b840af9f ("drm/panel: simple: Add Innolux G101ICE-L01 panel")
Signed-off-by: Marek Vasut <marex@denx.de>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20231008223256.279196-1-marex@denx.de
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/panel: simple: Fix Innolux G101ICE-L01 bus flags

[ Upstream commit 06fc41b09cfbc02977acd9189473593a37d82d9b ]

Add missing .bus_flags = DRM_BUS_FLAG_DE_HIGH to this panel description,
ones which match both the datasheet and the panel display_timing flags .

Fixes: 1e29b840af9f ("drm/panel: simple: Add Innolux G101ICE-L01 panel")
Signed-off-by: Marek Vasut <marex@denx.de>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20231008223315.279215-1-marex@denx.de
Signed-off-by: Sasha Levin <sashal@kernel.org>

fs: Pass AT_GETATTR_NOSEC flag to getattr interface function

[ Upstream commit 8a924db2d7b5eb69ba08b1a0af46e9f1359a9bdf ]

When vfs_getattr_nosec() calls a filesystem's getattr interface function
then the 'nosec' should propagate into this function so that
vfs_getattr_nosec() can again be called from the filesystem's gettattr
rather than vfs_getattr(). The latter would add unnecessary security
checks that the initial vfs_getattr_nosec() call wanted to avoid.
Therefore, introduce the getattr flag GETATTR_NOSEC and allow to pass
with the new getattr_flags parameter to the getattr interface function.
In overlayfs and ecryptfs use this flag to determine which one of the
two functions to call.

In a recent code change introduced to IMA vfs_getattr_nosec() ended up
calling vfs_getattr() in overlayfs, which in turn called
security_inode_getattr() on an exiting process that did not have
current->fs set anymore, which then caused a kernel NULL pointer
dereference. With this change the call to security_inode_getattr() can
be avoided, thus avoiding the NULL pointer dereference.

Reported-by: <syzbot+a67fc5321ffb4b311c98@syzkaller.appspotmail.com>
Fixes: db1d1e8b9867 ("IMA: use vfs_getattr_nosec to get the i_version")
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: <linux-fsdevel@vger.kernel.org>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Amir Goldstein <amir73il@gmail.com>
Cc: Tyler Hicks <code@tyhicks.com>
Cc: Mimi Zohar <zohar@linux.ibm.com>
Suggested-by: Christian Brauner <brauner@kernel.org>
Co-developed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Link: https://lore.kernel.org/r/20231002125733.1251467-1-stefanb@linux.vnet.ibm.com
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/panel: auo,b101uan08.3: Fine tune the panel power sequence

[ Upstream commit 6965809e526917b73c8f9178173184dcf13cec4b ]

For "auo,b101uan08.3" this panel, it is stipulated in the panel spec that
MIPI needs to keep the LP11 state before the lcm_reset pin is pulled high.

Fixes: 56ad624b4cb5 ("drm/panel: support for auo, b101uan08.3 wuxga dsi video mode panel")
Signed-off-by: Xuxin Xiong <xuxinxiong@huaqin.corp-partner.google.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20231114044205.613421-1-xuxinxiong@huaqin.corp-partner.google.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

blk-cgroup: avoid to warn !rcu_read_lock_held() in blkg_lookup()

[ Upstream commit 35a99d6557cacbc177314735342f77a2dda41872 ]

So far, all callers either holds spin lock or rcu read explicitly, and
most of the caller has added WARN_ON_ONCE(!rcu_read_lock_held()) or
lockdep_assert_held(&disk->queue->queue_lock).

Remove WARN_ON_ONCE(!rcu_read_lock_held()) from blkg_lookup() for
killing the false positive warning from blkg_conf_prep().

Reported-by: Changhui Zhong <czhong@redhat.com>
Fixes: 83462a6c971c ("blkcg: Drop unnecessary RCU read [un]locks from blkg_conf_prep/finish()")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20231117023527.3188627-3-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

afs: Make error on cell lookup failure consistent with OpenAFS

[ Upstream commit 2a4ca1b4b77850544408595e2433f5d7811a9daa ]

When kafs tries to look up a cell in the DNS or the local config, it will
translate a lookup failure into EDESTADDRREQ whereas OpenAFS translates it
into ENOENT.  Applications such as West expect the latter behaviour and
fail if they see the former.

This can be seen by trying to mount an unknown cell:

   # mount -t afs %example.com:cell.root /mnt
   mount: /mnt: mount(2) system call failed: Destination address required.

Fixes: 4d673da14533 ("afs: Support the AFS dynamic root")
Reported-by: Markus Suvanto <markus.suvanto@gmail.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216637
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Jeffrey Altman <jaltman@auristor.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

afs: Fix afs_server_list to be cleaned up with RCU

[ Upstream commit e6bace7313d61e31f2b16fa3d774fd8cb3cb869e ]

afs_server_list is accessed with the rcu_read_lock() held from
volume->servers, so it needs to be cleaned up correctly.

Fix this by using kfree_rcu() instead of kfree().

Fixes: 8a070a964877 ("afs: Detect cell aliases 1 - Cells with root volumes")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

rxrpc: Defer the response to a PING ACK until we've parsed it

[ Upstream commit 1a01319feef7047aa2ba400ffa3e047776aa29ca ]

Defer the generation of a PING RESPONSE ACK in response to a PING ACK until
we've parsed the PING ACK so that we pick up any changes to the packet
queue so that we can update ackinfo.

This is also applied to an ACK generated in response to an ACK with the
REQUEST_ACK flag set.

Note that whilst the problem was added in commit 248f219cb8bc, it didn't
really matter at that point because the ACK was proposed in softirq mode
and generated asynchronously later in process context, taking the latest
values at the time.  But this fix is only needed since the move to parse
incoming packets in an I/O thread rather than in softirq and generate the
ACK at point of proposal (b0346843b1076b34a0278ff601f8f287535cb064).

Fixes: 248f219cb8bc ("rxrpc: Rewrite the data and ack handling code")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: linux-afs@lists.infradead.org
cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

rxrpc: Fix RTT determination to use any ACK as a source

[ Upstream commit 3798680f2fbbe0ca3ab6138b34e0d161c36497ee ]

Fix RTT determination to be able to use any type of ACK as the response
from which RTT can be calculated provided its ack.serial is non-zero and
matches the serial number of an outgoing DATA or ACK packet.  This
shouldn't be limited to REQUESTED-type ACKs as these can have other types
substituted for them for things like duplicate or out-of-order packets.

Fixes: 4700c4d80b7b ("rxrpc: Fix loss of RTT samples due to interposed ACK")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: linux-afs@lists.infradead.org
cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/ism: ism driver implies smc protocol

[ Upstream commit d565fa4300d9ebd5ba3bbd259ce841f8dab609d6 ]

Since commit a72178cfe855 ("net/smc: Fix dependency of SMC on ISM")
you can build the ism code without selecting the SMC network protocol.
That leaves some ism functions be reported as unused. Move these
functions under the conditional compile with CONFIG_SMC.

Also codify the suggestion to also configure the SMC protocol in ism's
Kconfig - but with an "imply" rather than a "select" as SMC depends on
other config options and allow for a deliberate decision not to build
SMC. Also, mention that in ISM's help.

Fixes: a72178cfe855 ("net/smc: Fix dependency of SMC on ISM")
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Closes: https://lore.kernel.org/netdev/afd142a2-1fa0-46b9-8b2d-7652d41d3ab8@infradead.org/
Signed-off-by: Gerd Bayer <gbayer@linux.ibm.com>
Reviewed-by: Wenjia Zhang <wenjia@linux.ibm.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/msm/dsi: use the correct VREG_CTRL_1 value for 4nm cphy

[ Upstream commit b3e0f94d15700ac8e8c1c2355834f5d5c753c41d ]

Use the same value as the downstream driver. This change is needed for CPHY
mode to work correctly.

Fixes: 8b034e677111 ("drm/msm/dsi: add support for DSI-PHY on SM8550")
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/566987/
Link: https://lore.kernel.org/r/20231110000216.29979-1-jonathan@marek.ca
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/fair: Fix the decision for load balance

[ Upstream commit 6d7e4782bcf549221b4ccfffec2cf4d1a473f1a3 ]

should_we_balance is called for the decision to do load-balancing.
When sched ticks invoke this function, only one CPU should return
true. However, in the current code, two CPUs can return true. The
following situation, where b means busy and i means idle, is an
example, because CPU 0 and CPU 2 return true.

        [0, 1] [2, 3]
         b  b   i  b

This fix checks if there exists an idle CPU with busy sibling(s)
after looking for a CPU on an idle core. If some idle CPUs with busy
siblings are found, just the first one should do load-balancing.

Fixes: b1bfeab9b002 ("sched/fair: Consider the idle state of the whole core for load balance")
Signed-off-by: Keisuke Nishimura <keisuke.nishimura@inria.fr>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Chen Yu <yu.c.chen@intel.com>
Reviewed-by: Shrikanth Hegde <sshegde@linux.vnet.ibm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20231031133821.1570861-1-keisuke.nishimura@inria.fr
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/eevdf: Fix vruntime adjustment on reweight

[ Upstream commit eab03c23c2a162085b13200d7942fc5a00b5ccc8 ]

vruntime of the (on_rq && !0-lag) entity needs to be adjusted when
it gets re-weighted, and the calculations can be simplified based
on the fact that re-weight won't change the w-average of all the
entities. Please check the proofs in comments.

But adjusting vruntime can also cause position change in RB-tree
hence require re-queue to fix up which might be costly. This might
be avoided by deferring adjustment to the time the entity actually
leaves tree (dequeue/pick), but that will negatively affect task
selection and probably not good enough either.

Fixes: 147f3efaa241 ("sched/fair: Implement an EEVDF-like scheduling policy")
Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20231107090510.71322-2-wuyun.abel@bytedance.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

hv/hv_kvp_daemon: Some small fixes for handling NM keyfiles

[ Upstream commit c3803203bc5ec910a3eb06172cf6fb368e0e4390 ]

Some small fixes:
 - lets make sure we are not adding ipv4 addresses in ipv6 section in
   keyfile and vice versa.
 - ADDR_FAMILY_IPV6 is a bit in addr_family. Test that bit instead of
   checking the whole value of addr_family.
 - Some trivial fixes in hv_set_ifconfig.sh.

These fixes are proposed after doing some internal testing at Red Hat.

CC: Shradha Gupta <shradhagupta@linux.microsoft.com>
CC: Saurabh Sengar <ssengar@linux.microsoft.com>
Fixes: 42999c904612 ("hv/hv_kvp_daemon:Support for keyfile based connection profile")
Signed-off-by: Ani Sinha <anisinha@redhat.com>
Reviewed-by: Shradha Gupta <Shradhagupta@linux.microsoft.com>
Signed-off-by: Wei Liu <wei.liu@kernel.org>
Message-ID: <20231016133122.2419537-1-anisinha@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

irqchip/gic-v3-its: Flush ITS tables correctly in non-coherent GIC designs

[ Upstream commit d3badb15613c14dd35d3495b1dde5c90fcd616dd ]

In non-coherent GIC designs, the ITS tables must be flushed before writing
to the GITS_BASER<n> registers, otherwise the ITS could read dirty tables,
which results in unpredictable behavior.

Flush the tables right at the begin of its_setup_baser() to prevent that.

[ tglx: Massage changelog ]

Fixes: a8707f553884 ("irqchip/gic-v3: Add Rockchip 3588001 erratum workaround")
Suggested-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Fang Xiang <fangxiang3@xiaomi.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20231030083256.4345-1-fangxiang3@xiaomi.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

NFSD: Fix checksum mismatches in the duplicate reply cache

[ Upstream commit bf51c52a1f3c238d72c64e14d5e7702d3a245b82 ]

nfsd_cache_csum() currently assumes that the server's RPC layer has
been advancing rq_arg.head[0].iov_base as it decodes an incoming
request, because that's the way it used to work. On entry, it
expects that buf->head[0].iov_base points to the start of the NFS
header, and excludes the already-decoded RPC header.

These days however, head[0].iov_base now points to the start of the
RPC header during all processing. It no longer points at the NFS
Call header when execution arrives at nfsd_cache_csum().

In a retransmitted RPC the XID and the NFS header are supposed to
be the same as the original message, but the contents of the
retransmitted RPC header can be different. For example, for krb5,
the GSS sequence number will be different between the two. Thus if
the RPC header is always included in the DRC checksum computation,
the checksum of the retransmitted message might not match the
checksum of the original message, even though the NFS part of these
messages is identical.

The result is that, even if a matching XID is found in the DRC,
the checksum mismatch causes the server to execute the
retransmitted RPC transaction again.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Tested-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

NFSD: Fix "start of NFS reply" pointer passed to nfsd_cache_update()

[ Upstream commit 1caf5f61dd8430ae5a0b4538afe4953ce7517cbb ]

The "statp + 1" pointer that is passed to nfsd_cache_update() is
supposed to point to the start of the egress NFS Reply header. In
fact, it does point there for AUTH_SYS and RPCSEC_GSS_KRB5 requests.

But both krb5i and krb5p add fields between the RPC header's
accept_stat field and the start of the NFS Reply header. In those
cases, "statp + 1" points at the extra fields instead of the Reply.
The result is that nfsd_cache_update() caches what looks to the
client like garbage.

A connection break can occur for a number of reasons, but the most
common reason when using krb5i/p is a GSS sequence number window
underrun. When an underrun is detected, the server is obliged to
drop the RPC and the connection to force a retransmit with a fresh
GSS sequence number. The client presents the same XID, it hits in
the server's DRC, and the server returns the garbage cache entry.

The "statp + 1" argument has been used since the oldest changeset
in the kernel history repo, so it has been in nfsd_dispatch()
literally since before history began. The problem arose only when
the server-side GSS implementation was added twenty years ago.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Tested-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 6.6.3

Link: https://lore.kernel.org/r/20231124172028.107505484@linuxfoundation.org
Tested-by: Ronald Warsow <rwarsow@gmx.de>
Tested-by: Nam Cao <namcao@linutronix.de>
Tested-by: Takeshi Ogasawara <takeshi.ogasawara@futuring-girl.com>
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Link: https://lore.kernel.org/r/20231125163158.249616313@linuxfoundation.org
Tested-by: Ronald Warsow <rwarsow@gmx.de>
Tested-by: SeongJae Park <sj@kernel.org>
Link: https://lore.kernel.org/r/20231125194417.090791215@linuxfoundation.org
Tested-by: Takeshi Ogasawara <takeshi.ogasawara@futuring-girl.com>
Tested-by: Ronald Warsow <rwarsow@gmx.de>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Link: https://lore.kernel.org/r/20231126154418.032283745@linuxfoundation.org
Tested-by: SeongJae Park <sj@kernel.org>
Tested-by: Ricardo B. Marliere <ricardo@marliere.net>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Takeshi Ogasawara <takeshi.ogasawara@futuring-girl.com>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Nam Cao <namcao@linutronix.de>
Tested-by: Allen Pais <apais@linux.microsoft.com>
Tested-by: Justin M. Forbes <jforbes@fedoraproject.org>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Tested-by: Rudi Heitbaum <rudi@heitbaum.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Change the DMCUB mailbox memory location from FB to inbox

commit 5911d02cac70d7fb52009fbd37423e63f8f6f9bc upstream.

[WHY]
Flush command sent to DMCUB spends more time for execution on
a dGPU than on an APU. This causes cursor lag when using high
refresh rate mouses.

[HOW]
1. Change the DMCUB mailbox memory location from FB to inbox.
2. Only change windows memory to inbox.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Lewis Huang <lewis.huang@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Clear dpcd_sink_ext_caps if not set

commit 923bbfe6c888812db1088d684bd30c24036226d2 upstream.

[WHY]
Some eDP panels' ext caps don't set initial values
and the value of dpcd_addr (0x317) is random.
It means that sometimes the eDP can be OLED, miniLED and etc,
and cause incorrect backlight control interface.

[HOW]
Add remove_sink_ext_caps to remove sink ext caps (HDR, OLED and etc)

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Anthony Koo <anthony.koo@amd.com>
Acked-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Paul Hsieh <paul.hsieh@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Enable fast plane updates on DCN3.2 and above

commit 435f5b369657cffee4b04db1f5805b48599f4dbe upstream.

[WHY]
When cursor moves across screen boarder, lag cursor observed,
since subvp settings need to sync up with vblank that causes
cursor updates being delayed.

[HOW]
Enable fast plane updates on DCN3.2 to fix it.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Acked-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Tianci Yin <tianci.yin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: fix a NULL pointer dereference in amdgpu_dm_i2c_xfer()

commit b71f4ade1b8900d30c661d6c27f87c35214c398c upstream.

When ddc_service_construct() is called, it explicitly checks both the
link type and whether there is something on the link which will
dictate whether the pin is marked as hw_supported.

If the pin isn't set or the link is not set (such as from
unloading/reloading amdgpu in an IGT test) then fail the
amdgpu_dm_i2c_xfer() call.

Cc: stable@vger.kernel.org
Fixes: 22676bc500c2 ("drm/amd/display: Fix dmub soft hang for PSR 1")
Link: https://github.com/fwupd/fwupd/issues/6327
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Fix DSC not Enabled on Direct MST Sink

commit a58555359a9f870543aaddef277c3396159895ce upstream.

[WHY & HOW]
For the scenario when a dsc capable MST sink device is directly
connected, it needs to use max dsc compression as the link bw constraint.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Roman Li <roman.li@amd.com>
Acked-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Fangzhi Zuo <jerry.zuo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Guard against invalid RPTR/WPTR being set

commit 1ffa8602e39b89469dc703ebab7a7e44c33da0f7 upstream.

[WHY]
HW can return invalid values on register read, guard against these being
set and causing us to access memory out of range and page fault.

[HOW]
Guard at sync_inbox1 and guard at pushing commands.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Hansen Dsouza <hansen.dsouza@amd.com>
Acked-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: Fix possible null pointer dereference

commit 256503071c2de2b5b5c20e06654aa9a44f13aa62 upstream.

mem = bo->tbo.resource may be NULL in amdgpu_vm_bo_update.

Fixes: 180253782038 ("drm/ttm: stop allocating dummy resources during BO creation")
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: lower CS errors to debug severity

commit 17daf01ab4e3e5a5929747aa05cc15eb2bad5438 upstream.

Otherwise userspace can spam the logs by using incorrect input values.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: fix error handling in amdgpu_bo_list_get()

commit 12f76050d8d4d10dab96333656b821bd4620d103 upstream.

We should not leak the pointer where we couldn't grab the reference
on to the caller because it can be that the error handling still
tries to put the reference then.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: fix error handling in amdgpu_vm_init

commit 8473bfdcb5b1a32fd05629c4535ccacd73bc5567 upstream.

When clearing the root PD fails we need to properly release it again.

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: don't use ATRM for external devices

commit 432e664e7c98c243fab4c3c95bd463bea3aeed28 upstream.

The ATRM ACPI method is for fetching the dGPU vbios rom
image on laptops and all-in-one systems.  It should not be
used for external add in cards.  If the dGPU is thunderbolt
connected, don't try ATRM.

v2: pci_is_thunderbolt_attached only works for Intel.  Use
    pdev->external_facing instead.
v3: dev_is_removable() seems to be what we want

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2925
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: add a retry for IP discovery init

commit 3938eb956e383ef88b8fc7d556492336ebee52df upstream.

AMD dGPUs have integrated FW that runs as soon as the
device gets power and initializes the board (determines
the amount of memory, provides configuration details to
the driver, etc.).  For direct PCIe attached cards this
happens as soon as power is applied and normally completes
well before the OS has even started loading.  However, with
hotpluggable ports like USB4, the driver needs to wait for
this to complete before initializing the device.

This normally takes 60-100ms, but could take longer on
some older boards periodically due to memory training.

Retry for up to a second.  In the non-hotplug case, there
should be no change in behavior and this should complete
on the first try.

v2: adjust test criteria
v3: adjust checks for the masks, only enable on removable devices
v4: skip bif_fb_en check

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2925
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: fix GRBM read timeout when do mes_self_test

commit 36e7ff5c13cb15cb7b06c76d42bb76cbf6b7ea75 upstream.

Use a proper MEID to make sure the CP_HQD_* and CP_GFX_HQD_* registers
can be touched when initialize the compute and gfx mqd in mes_self_test.
Otherwise, we expect no response from CP and an GRBM eventual timeout.

Signed-off-by: Tim Huang <Tim.Huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: don't use pci_is_thunderbolt_attached()

commit 7b1c6263eaf4fd64ffe1cafdc504a42ee4bfbb33 upstream.

It's only valid on Intel systems with the Intel VSEC.
Use dev_is_removable() instead.  This should do the right
thing regardless of the platform.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2925
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu/smu13: drop compute workload workaround

commit 23170863ea0a0965d224342c0eb2ad8303b1f267 upstream.

This was fixed in PMFW before launch and is no longer
required.

Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.1.x
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/pm: Fix error of MACO flag setting code

commit 7f3e6b840fa8b0889d776639310a5dc672c1e9e1 upstream.

MACO only works if BACO is supported

Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.1.x
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/i915: Flush WC GGTT only on required platforms

commit 7d7a328d0e8d6edefb7b0d665185d468667588d0 upstream.

gen8_ggtt_invalidate() is only needed for limited set of platforms
where GGTT is mapped as WC. This was added as way to fix WC based GGTT in
commit 0f9b91c754b7 ("drm/i915: flush system agent TLBs on SNB") and
there are no reference in HW docs that forces us to use this on non-WC
backed GGTT.

This can also cause unwanted side-effects on XE_HP platforms where
GFX_FLSH_CNTL_GEN6 is not valid anymore.

v2: Add a func to detect wc ggtt detection (Ville)
v3: Improve commit log and add reference commit (Daniel)

Fixes: d2eae8e98d59 ("drm/i915/dg2: Drop force_probe requirement")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Jonathan Cavitt <jonathan.cavitt@intel.com>
Cc: John Harrison <john.c.harrison@intel.com>
Cc: Andi Shyti <andi.shyti@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: <stable@vger.kernel.org> # v6.2+
Suggested-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231018093815.1349-1-nirmoy.das@intel.com
(cherry picked from commit 81de3e296b10a13e5c9f13172825b0d8d9495c68)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/i915: Fix potential spectre vulnerability

commit 1a8e9bad6ef563c28ab0f8619628d5511be55431 upstream.

Fix smatch warning:
drivers/gpu/drm/i915/gem/i915_gem_context.c:847 set_proto_ctx_sseu()
warn: potential spectre issue 'pc->user_engines' [r] (local cap)

Fixes: d4433c7600f7 ("drm/i915/gem: Use the proto-context to handle create parameters (v5)")
Cc: <stable@vger.kernel.org> # v5.15+
Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231103110922.430122-1-tvrtko.ursulin@linux.intel.com
(cherry picked from commit 27b086382c22efb7e0a16442f7bdc2e120108ef3)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/i915: Bump GLK CDCLK frequency when driving multiple pipes

commit 0cb89cd42fd22bbdec0b046c48f35775f5b88bdb upstream.

On GLK CDCLK frequency needs to be at least 2*96 MHz when accessing
the audio hardware. Currently we bump the CDCLK frequency up
temporarily (if not high enough already) whenever audio hardware
is being accessed, and drop it back down afterwards.

With a single active pipe this works just fine as we can switch
between all the valid CDCLK frequencies by changing the cd2x
divider, which doesn't require a full modeset. However with
multiple active pipes the cd2x divider trick no longer works,
and thus we end up blinking all displays off and back on.

To avoid this let's just bump the CDCLK frequency to >=2*96MHz
whenever multiple pipes are active. The downside is slightly
higher power consumption, but that seems like an acceptable
tradeoff. With a single active pipe we can stick to the current
more optiomal (from power comsumption POV) behaviour.

Cc: stable@vger.kernel.org
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/9599
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231031160800.18371-1-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
(cherry picked from commit 451eaa1a614c911f5a51078dcb68022874e4cb12)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/i915/mtl: Support HBR3 rate with C10 phy and eDP in MTL

commit ce4941c2d6459664761c9854701015d8e99414fb upstream.

eDP specification supports HBR3 link rate since v1.4a. Moreover,
C10 phy can support HBR3 link rate for both DP and eDP. Therefore,
do not clamp the supported rates for eDP at 6.75Gbps.

Cc: <stable@vger.kernel.org>
BSpec: 70073 74224

Signed-off-by: Chaitanya Kumar Borah <chaitanya.kumar.borah@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231018113622.2761997-1-chaitanya.kumar.borah@intel.com
(cherry picked from commit a3431650f30a94b179d419ef87c21213655c28cd)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Add Null check for DPP resource

commit e186400685d8a9287388a8535e2399bc673bfe95 upstream.

[what and why]
Check whether dpp resource pointer is null in advance and return early
if so.

Reviewed-by: Charlene Liu <charlene.liu@amd.com>
Reviewed-by: Martin Leung <martin.leung@amd.com>
Signed-off-by: Gabe Teeger <gabe.teeger@amd.com>
Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/srso: Move retbleed IBPB check into existing 'has_microcode' code block

commit 351236947a45a512c517153bbe109fe868d05e6d upstream.

Simplify the code flow a bit by moving the retbleed IBPB check into the
existing 'has_microcode' block.

Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Acked-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/0a22b86b1f6b07f9046a9ab763fc0e0d1b7a91d4.1693889988.git.jpoimboe@kernel.org
Cc: Caleb Jorden <cjorden@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm: bridge: it66121: ->get_edid callback must not return err pointers

commit 81995ee1620318b4c7bbeb02bcc372da2c078c76 upstream.

The drm stack does not expect error valued pointers for EDID anywhere.

Fixes: e66856508746 ("drm: bridge: it66121: Set DDC preamble only once before reading EDID")
Cc: Paul Cercueil <paul@crapouillou.net>
Cc: Robert Foss <robert.foss@linaro.org>
Cc: Phong LE <ple@baylibre.com>
Cc: Neil Armstrong <neil.armstrong@linaro.org>
Cc: Andrzej Hajda <andrzej.hajda@intel.com>
Cc: Robert Foss <rfoss@kernel.org>
Cc: Laurent Pinchart <Laurent.pinchart@ideasonboard.com>
Cc: Jonas Karlman <jonas@kwiboo.se>
Cc: Jernej Skrabec <jernej.skrabec@gmail.com>
Cc: <stable@vger.kernel.org> # v6.3+
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20230914131159.2472513-1-jani.nikula@intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/pm: Handle non-terminated overdrive commands.

commit 08e9ebc75b5bcfec9d226f9e16bab2ab7b25a39a upstream.

The incoming strings might not be terminated by a newline
or a 0.

(found while testing a program that just wrote the string
 itself, causing a crash)

Cc: stable@vger.kernel.org
Fixes: e3933f26b657 ("drm/amd/pp: Add edit/commit/show OD clock/voltage support in sysfs")
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: fix racy may inline data check in dio write

commit ce56d21355cd6f6937aca32f1f44ca749d1e4808 upstream.

syzbot reports that the following warning from ext4_iomap_begin()
triggers as of the commit referenced below:

        if (WARN_ON_ONCE(ext4_has_inline_data(inode)))
                return -ERANGE;

This occurs during a dio write, which is never expected to encounter
an inode with inline data. To enforce this behavior,
ext4_dio_write_iter() checks the current inline state of the inode
and clears the MAY_INLINE_DATA state flag to either fall back to
buffered writes, or enforce that any other writers in progress on
the inode are not allowed to create inline data.

The problem is that the check for existing inline data and the state
flag can span a lock cycle. For example, if the ilock is originally
locked shared and subsequently upgraded to exclusive, another writer
may have reacquired the lock and created inline data before the dio
write task acquires the lock and proceeds.

The commit referenced below loosens the lock requirements to allow
some forms of unaligned dio writes to occur under shared lock, but
AFAICT the inline data check was technically already racy for any
dio write that would have involved a lock cycle. Regardless, lift
clearing of the state bit to the same lock critical section that
checks for preexisting inline data on the inode to close the race.

Cc: stable@kernel.org
Reported-by: syzbot+307da6ca5cb0d01d581a@syzkaller.appspotmail.com
Fixes: 310ee0902b8d ("ext4: allow concurrent unaligned dio overwrites")
Signed-off-by: Brian Foster <bfoster@redhat.com>
Link: https://lore.kernel.org/r/20231002185020.531537-1-bfoster@redhat.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: properly sync file size update after O_SYNC direct IO

commit 91562895f8030cb9a0470b1db49de79346a69f91 upstream.

Gao Xiang has reported that on ext4 O_SYNC direct IO does not properly
sync file size update and thus if we crash at unfortunate moment, the
file can have smaller size although O_SYNC IO has reported successful
completion. The problem happens because update of on-disk inode size is
handled in ext4_dio_write_iter() *after* iomap_dio_rw() (and thus
dio_complete() in particular) has returned and generic_file_sync() gets
called by dio_complete(). Fix the problem by handling on-disk inode size
update directly in our ->end_io completion handler.

References: https://lore.kernel.org/all/02d18236-26ef-09b0-90ad-030c4fe3ee20@linux.alibaba.com
Reported-by: Gao Xiang <hsiangkao@linux.alibaba.com>
CC: stable@vger.kernel.org
Fixes: 378f32bab371 ("ext4: introduce direct I/O write using iomap infrastructure")
Signed-off-by: Jan Kara <jack@suse.cz>
Tested-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: "Ritesh Harjani (IBM)" <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20231013121350.26872-1-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: add missed brelse in update_backups

commit 9adac8b01f4be28acd5838aade42b8daa4f0b642 upstream.

add missed brelse in update_backups

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Link: https://lore.kernel.org/r/20230826174712.4059355-3-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: remove gdb backup copy for meta bg in setup_new_flex_group_blocks

commit 40dd7953f4d606c280074f10d23046b6812708ce upstream.

Wrong check of gdb backup in meta bg as following:
first_group is the first group of meta_bg which contains target group, so
target group is always >= first_group. We check if target group has gdb
backup by comparing first_group with [group + 1] and [group +
EXT4_DESC_PER_BLOCK(sb) - 1]. As group >= first_group, then [group + N] is
> first_group. So no copy of gdb backup in meta bg is done in
setup_new_flex_group_blocks.

No need to do gdb backup copy in meta bg from setup_new_flex_group_blocks
as we always copy updated gdb block to backups at end of
ext4_flex_group_add as following:

ext4_flex_group_add
  /* no gdb backup copy for meta bg any more */
  setup_new_flex_group_blocks

  /* update current group number */
  ext4_update_super
    sbi->s_groups_count += flex_gd->count;

  /*
   * if group in meta bg contains backup is added, the primary gdb block
   * of the meta bg will be copy to backup in new added group here.
   */
  for (; gdb_num <= gdb_num_end; gdb_num++)
    update_backups(...)

In summary, we can remove wrong gdb backup copy code in
setup_new_flex_group_blocks.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Link: https://lore.kernel.org/r/20230826174712.4059355-5-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: correct the start block of counting reserved clusters

commit 40ea98396a3659062267d1fe5f99af4f7e4f05e3 upstream.

When big allocate feature is enabled, we need to count and update
reserved clusters before removing a delayed only extent_status entry.
{init|count|get}_rsvd() have already done this, but the start block
number of this counting isn't correct in the following case.

  lblk            end
   |               |
   v               v
          -------------------------
          |                       | orig_es
          -------------------------
                   ^              ^
      len1 is 0    |     len2     |

If the start block of the orig_es entry founded is bigger than lblk, we
passed lblk as start block to count_rsvd(), but the length is correct,
finally, the range to be counted is offset. This patch fix this by
passing the start blocks to 'orig_es->lblk + len1'.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Cc: stable@kernel.org
Link: https://lore.kernel.org/r/20230824092619.1327976-2-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: correct return value of ext4_convert_meta_bg

commit 48f1551592c54f7d8e2befc72a99ff4e47f7dca0 upstream.

Avoid to ignore error in "err".

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Link: https://lore.kernel.org/r/20230826174712.4059355-4-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: mark buffer new if it is unwritten to avoid stale data exposure

commit 2cd8bdb5efc1e0d5b11a4b7ba6b922fd2736a87f upstream.

** Short Version **

In ext4 with dioread_nolock, we could have a scenario where the bh returned by
get_blocks (ext4_get_block_unwritten()) in __block_write_begin_int() has
UNWRITTEN and MAPPED flag set. Since such a bh does not have NEW flag set we
never zero out the range of bh that is not under write, causing whatever stale
data is present in the folio at that time to be written out to disk. To fix this
mark the buffer as new, in case it is unwritten, in ext4_get_block_unwritten().

** Long Version **

The issue mentioned above was resulting in two different bugs:

1. On block size < page size case in ext4, generic/269 was reliably
failing with dioread_nolock. The state of the write was as follows:

  * The write was extending i_size.
  * The last block of the file was fallocated and had an unwritten extent
  * We were near ENOSPC and hence we were switching to non-delayed alloc
    allocation.

In this case, the back trace that triggers the bug is as follows:

  ext4_da_write_begin()
    /* switch to nodelalloc due to low space */
    ext4_write_begin()
      ext4_should_dioread_nolock() // true since mount flags still have delalloc
      __block_write_begin(..., ext4_get_block_unwritten)
        __block_write_begin_int()
          for(each buffer head in page) {
            /* first iteration, this is bh1 which contains i_size */
            if (!buffer_mapped)
              get_block() /* returns bh with only UNWRITTEN and MAPPED */
            /* second iteration, bh2 */
              if (!buffer_mapped)
                get_block() /* we fail here, could be ENOSPC */
          }
          if (err)
            /*
             * this would zero out all new buffers and mark them uptodate.
             * Since bh1 was never marked new, we skip it here which causes
             * the bug later.
             */
            folio_zero_new_buffers();
      /* ext4_wrte_begin() error handling */
      ext4_truncate_failed_write()
        ext4_truncate()
          ext4_block_truncate_page()
            __ext4_block_zero_page_range()
              if(!buffer_uptodate())
                ext4_read_bh_lock()
                  ext4_read_bh() -> ... ext4_submit_bh_wbc()
                    BUG_ON(buffer_unwritten(bh)); /* !!! */

2. The second issue is stale data exposure with page size >= blocksize
with dioread_nolock. The conditions needed for it to happen are same as
the previous issue ie dioread_nolock around ENOSPC condition. The issue
is also similar where in __block_write_begin_int() when we call
ext4_get_block_unwritten() on the buffer_head and the underlying extent
is unwritten, we get an unwritten and mapped buffer head. Since it is
not new, we never zero out the partial range which is not under write,
thus writing stale data to disk. This can be easily observed with the
following reproducer:

 fallocate -l 4k testfile
 xfs_io -c "pwrite 2k 2k" testfile
 # hexdump output will have stale data in from byte 0 to 2k in testfile
 hexdump -C testfile

NOTE: To trigger this, we need dioread_nolock enabled and write happening via
ext4_write_begin(), which is usually used when we have -o nodealloc. Since
dioread_nolock is disabled with nodelalloc, the only alternate way to call
ext4_write_begin() is to ensure that delayed alloc switches to nodelalloc ie
ext4_da_write_begin() calls ext4_write_begin(). This will usually happen when
ext4 is almost full like the way generic/269 was triggering it in Issue 1 above.
This might make the issue harder to hit. Hence, for reliable replication, I used
the below patch to temporarily allow dioread_nolock with nodelalloc and then
mount the disk with -o nodealloc,dioread_nolock. With this you can hit the stale
data issue 100% of times:

@@ -508,8 +508,8 @@ static inline int ext4_should_dioread_nolock(struct inode *inode)
  if (ext4_should_journal_data(inode))
    return 0;
  /* temporary fix to prevent generic/422 test failures */
- if (!test_opt(inode->i_sb, DELALLOC))
-   return 0;
+ // if (!test_opt(inode->i_sb, DELALLOC))
+ //  return 0;
  return 1;
 }

After applying this patch to mark buffer as NEW, both the above issues are
fixed.

Signed-off-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Cc: stable@kernel.org
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: "Ritesh Harjani (IBM)" <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/d0ed09d70a9733fbb5349c5c7b125caac186ecdf.1695033645.git.ojaswin@linux.ibm.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: correct offset of gdb backup in non meta_bg group to update_backups

commit 31f13421c004a420c0e9d288859c9ea9259ea0cc upstream.

Commit 0aeaa2559d6d5 ("ext4: fix corruption when online resizing a 1K
bigalloc fs") found that primary superblock's offset in its group is
not equal to offset of backup superblock in its group when block size
is 1K and bigalloc is enabled. As group descriptor blocks are right
after superblock, we can't pass block number of gdb to update_backups
for the same reason.

The root casue of the issue above is that leading 1K padding block is
count as data block offset for primary block while backup block has no
padding block offset in its group.

Remove padding data block count to fix the issue for gdb backups.

For meta_bg case, update_backups treat blk_off as block number, do no
conversion in this case.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Link: https://lore.kernel.org/r/20230826174712.4059355-2-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: apply umask if ACL support is disabled

commit 484fd6c1de13b336806a967908a927cc0356e312 upstream.

The function ext4_init_acl() calls posix_acl_create() which is
responsible for applying the umask.  But without
CONFIG_EXT4_FS_POSIX_ACL, ext4_init_acl() is an empty inline function,
and nobody applies the umask.

This fixes a bug which causes the umask to be ignored with O_TMPFILE
on ext4:

 https://github.com/MusicPlayerDaemon/MPD/issues/558
 https://bugs.gentoo.org/show_bug.cgi?id=686142#c3
 https://bugzilla.kernel.org/show_bug.cgi?id=203625

Reviewed-by: "J. Bruce Fields" <bfields@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Max Kellermann <max.kellermann@ionos.com>
Link: https://lore.kernel.org/r/20230919081824.1096619-1-max.kellermann@ionos.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: make sure allocate pending entry not fail

commit 8e387c89e96b9543a339f84043cf9df15fed2632 upstream.

__insert_pending() allocate memory in atomic context, so the allocation
could fail, but we are not handling that failure now. It could lead
ext4_es_remove_extent() to get wrong reserved clusters, and the global
data blocks reservation count will be incorrect. The same to
extents_status entry preallocation, preallocate pending entry out of the
i_es_lock with __GFP_NOFAIL, make sure __insert_pending() and
__revise_pending() always succeeds.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Cc: stable@kernel.org
Link: https://lore.kernel.org/r/20230824092619.1327976-3-yi.zhang@huaweicloud.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: no need to generate from free list in mballoc

commit ebf6cb7c6e1241984f75f29f1bdbfa2fe7168f88 upstream.

Commit 7a2fcbf7f85 ("ext4: don't use blocks freed but not yet committed in
buddy cache init") added a code to mark as used blocks in the list of not yet
committed freed blocks during initialization of a buddy page. However
ext4_mb_free_metadata() makes sure buddy page is already loaded and takes a
reference to it so it cannot happen that ext4_mb_init_cache() is called
when efd list is non-empty. Just remove the
ext4_mb_generate_from_freelist() call.

Fixes: 7a2fcbf7f85('ext4: don't use blocks freed but not yet committed in buddy cache init')
Signed-off-by: Wang Jianjian <wangjianjian0@foxmail.com>
Link: https://lore.kernel.org/r/tencent_53CBCB1668358AE862684E453DF37B722008@qq.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: fix race between writepages and remount

commit 745f17a4166e79315e4b7f33ce89d03e75a76983 upstream.

We got a WARNING in ext4_add_complete_io:
==================================================================
 WARNING: at fs/ext4/page-io.c:231 ext4_put_io_end_defer+0x182/0x250
 CPU: 10 PID: 77 Comm: ksoftirqd/10 Tainted: 6.3.0-rc2 #85
 RIP: 0010:ext4_put_io_end_defer+0x182/0x250 [ext4]
 [...]
 Call Trace:
  <TASK>
  ext4_end_bio+0xa8/0x240 [ext4]
  bio_endio+0x195/0x310
  blk_update_request+0x184/0x770
  scsi_end_request+0x2f/0x240
  scsi_io_completion+0x75/0x450
  scsi_finish_command+0xef/0x160
  scsi_complete+0xa3/0x180
  blk_complete_reqs+0x60/0x80
  blk_done_softirq+0x25/0x40
  __do_softirq+0x119/0x4c8
  run_ksoftirqd+0x42/0x70
  smpboot_thread_fn+0x136/0x3c0
  kthread+0x140/0x1a0
  ret_from_fork+0x2c/0x50
==================================================================

Above issue may happen as follows:

            cpu1                        cpu2
----------------------------|----------------------------
mount -o dioread_lock
ext4_writepages
 ext4_do_writepages
  *if (ext4_should_dioread_nolock(inode))*
    // rsv_blocks is not assigned here
                                 mount -o remount,dioread_nolock
  ext4_journal_start_with_reserve
   __ext4_journal_start
    __ext4_journal_start_sb
     jbd2__journal_start
      *if (rsv_blocks)*
        // h_rsv_handle is not initialized here
  mpage_map_and_submit_extent
    mpage_map_one_extent
      dioread_nolock = ext4_should_dioread_nolock(inode)
      if (dioread_nolock && (map->m_flags & EXT4_MAP_UNWRITTEN))
        mpd->io_submit.io_end->handle = handle->h_rsv_handle
        ext4_set_io_unwritten_flag
          io_end->flag |= EXT4_IO_END_UNWRITTEN
      // now io_end->handle is NULL but has EXT4_IO_END_UNWRITTEN flag

scsi_finish_command
 scsi_io_completion
  scsi_io_completion_action
   scsi_end_request
    blk_update_request
     req_bio_endio
      bio_endio
       bio->bi_end_io  > ext4_end_bio
        ext4_put_io_end_defer
 ext4_add_complete_io
  // trigger WARN_ON(!io_end->handle && sbi->s_journal);

The immediate cause of this problem is that ext4_should_dioread_nolock()
function returns inconsistent values in the ext4_do_writepages() and
mpage_map_one_extent(). There are four conditions in this function that
can be changed at mount time to cause this problem. These four conditions
can be divided into two categories:

    (1) journal_data and EXT4_EXTENTS_FL, which can be changed by ioctl
    (2) DELALLOC and DIOREAD_NOLOCK, which can be changed by remount

The two in the first category have been fixed by commit c8585c6fcaf2
("ext4: fix races between changing inode journal mode and ext4_writepages")
and commit cb85f4d23f79 ("ext4: fix race between writepages and enabling
EXT4_EXTENTS_FL") respectively.

Two cases in the other category have not yet been fixed, and the above
issue is caused by this situation. We refer to the fix for the first
category, when applying options during remount, we grab s_writepages_rwsem
to avoid racing with writepages ops to trigger this problem.

Fixes: 6b523df4fb5a ("ext4: use transaction reservation for extent conversion in ext4_end_io")
Cc: stable@vger.kernel.org
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230524072538.2883391-1-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "net: r8169: Disable multicast filter for RTL8168H and RTL8107E"

commit 6a26310273c323380da21eb23fcfd50e31140913 upstream.

This reverts commit efa5f1311c4998e9e6317c52bc5ee93b3a0f36df.

I couldn't reproduce the reported issue. What I did, based on a pcap
packet log provided by the reporter:
- Used same chip version (RTL8168h)
- Set MAC address to the one used on the reporters system
- Replayed the EAPOL unicast packet that, according to the reporter,
  was filtered out by the mc filter.
The packet was properly received.

Therefore the root cause of the reported issue seems to be somewhere
else. Disabling mc filtering completely for the most common chip
version is a quite big hammer. Therefore revert the change and wait
for further analysis results from the reporter.

Cc: stable@vger.kernel.org
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "HID: logitech-dj: Add support for a new lightspeed receiver iteration"

commit 5b4ffb176d7979ac66b349addf3f7de433335e00 upstream.

This reverts commit 9d1bd9346241cd6963b58da7ffb7ed303285f684.

Multiple people reported misbehaving devices and reverting this commit fixes
the problem for them. As soon as the original commit author starts reacting
again, we can try to figure out why he hasn't seen the issues (mismatching
report descriptors?), but for the time being, fix for 6.7 by reverting.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=218172
Link: https://bugzilla.kernel.org/show_bug.cgi?id=218094
Cc: <stable@vger.kernel.org>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix csid-gen2 for test pattern generator

commit 87889f1b7ea40d2544b49c62092e6ef2792dced7 upstream.

In the current driver csid Test Pattern Generator (TPG) doesn't work.
This change:
- fixes writing frame width and height values into CSID_TPG_DT_n_CFG_0
- fixes the shift by one between test_pattern control value and the
  actual pattern.
- drops fixed VC of 0x0a which testing showed prohibited some test
  patterns in the CSID to produce output.
So that TPG starts working, but with the below limitations:
- only test_pattern=9 works as it should
- test_pattern=8 and test_pattern=7 produce black frame (all zeroes)
- the rest of test_pattern's don't work (yavta doesn't get the data)
- regardless of the CFA pattern set by 'media-ctl -V' the actual pixel
  order is always the same (RGGB for any RAW8 or RAW10P format in
  4608x2592 resolution).

Tested with:

RAW10P format, VC0:
 media-ctl -V '"msm_csid0":0[fmt:SRGGB10/4608x2592 field:none]'
 media-ctl -V '"msm_vfe0_rdi0":0[fmt:SRGGB10/4608x2592 field:none]'
 media-ctl -l '"msm_csid0":1->"msm_vfe0_rdi0":0[1]'
 v4l2-ctl -d /dev/v4l-subdev6 -c test_pattern=9
 yavta -B capture-mplane --capture=3 -n 3 -f SRGGB10P -s 4608x2592 /dev/video0

RAW10P format, VC1:
 media-ctl -V '"msm_csid0":2[fmt:SRGGB10/4608x2592 field:none]'
 media-ctl -V '"msm_vfe0_rdi1":0[fmt:SRGGB10/4608x2592 field:none]'
 media-ctl -l '"msm_csid0":2->"msm_vfe0_rdi1":0[1]'
 v4l2-ctl -d /dev/v4l-subdev6 -c test_pattern=9
 yavta -B capture-mplane --capture=3 -n 3 -f SRGGB10P -s 4608x2592 /dev/video1

RAW8 format, VC0:
 media-ctl --reset
 media-ctl -V '"msm_csid0":0[fmt:SRGGB8/4608x2592 field:none]'
 media-ctl -V '"msm_vfe0_rdi0":0[fmt:SRGGB8/4608x2592 field:none]'
 media-ctl -l '"msm_csid0":1->"msm_vfe0_rdi0":0[1]'
 yavta -B capture-mplane --capture=3 -n 3 -f SRGGB8 -s 4608x2592 /dev/video0

Fixes: eebe6d00e9bf ("media: camss: Add support for CSID hardware version Titan 170")
Cc: stable@vger.kernel.org
Signed-off-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix invalid clock enable bit disjunction

commit d8f7e1a60d01739a1d78db2b08603089c6cf7c8e upstream.

define CSIPHY_3PH_CMN_CSI_COMMON_CTRL5_CLK_ENABLE BIT(7)

disjunction for gen2 ? BIT(7) : is a nop we are setting the same bit
either way.

Fixes: 4abb21309fda ("media: camss: csiphy: Move to hardcode CSI Clock Lane number")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix set CSI2_RX_CFG1_VC_MODE when VC is greater than 3

commit e655d1ae9703286cef7fda8675cad62f649dc183 upstream.

VC_MODE = 0 implies a two bit VC address.
VC_MODE = 1 is required for VCs with a larger address than two bits.

Fixes: eebe6d00e9bf ("media: camss: Add support for CSID hardware version Titan 170")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix missing vfe_lite clocks check

commit b6e1bdca463a932c1ac02caa7d3e14bf39288e0c upstream.

check_clock doesn't account for vfe_lite which means that vfe_lite will
never get validated by this routine. Add the clock name to the expected set
to remediate.

Fixes: 7319cdf189bb ("media: camss: Add support for VFE hardware version Titan 170")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix VFE-480 vfe_disable_output()

commit 7f24d291350426d40b36dfbe6b3090617cdfd37a upstream.

vfe-480 is copied from vfe-17x and has the same racy idle timeout bug as in
17x.

Fix the vfe_disable_output() logic to no longer be racy and to conform
to the 17x way of quiescing and then resetting the VFE.

Fixes: 4edc8eae715c ("media: camss: Add initial support for VFE hardware version Titan 480")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix VFE-17x vfe_disable_output()

commit 3143ad282fc08bf995ee73e32a9e40c527bf265d upstream.

There are two problems with the current vfe_disable_output() routine.

Firstly we rightly use a spinlock to protect output->gen2.active_num
everywhere except for in the IDLE timeout path of vfe_disable_output().
Even if that is not racy "in practice" somehow it is by happenstance not
by design.

Secondly we do not get consistent behaviour from this routine. On
sc8280xp 50% of the time I get "VFE idle timeout - resetting". In this
case the subsequent capture will succeed. The other 50% of the time, we
don't hit the idle timeout, never do the VFE reset and subsequent
captures stall indefinitely.

Rewrite the vfe_disable_output() routine to

- Quiesce write masters with vfe_wm_stop()
- Set active_num = 0

remembering to hold the spinlock when we do so followed by

- Reset the VFE

Testing on sc8280xp and sdm845 shows this to be a valid fix.

Fixes: 7319cdf189bb ("media: camss: Add support for VFE hardware version Titan 170")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix vfe_get() error jump

commit 26bda3da00c3edef727a6acb00ed2eb4b22f8723 upstream.

Right now it is possible to do a vfe_get() with the internal reference
count at 1. If vfe_check_clock_rates() returns non-zero then we will
leave the reference count as-is and

run:
- pm_runtime_put_sync()
- vfe->ops->pm_domain_off()

skip:
- camss_disable_clocks()

Subsequent vfe_put() calls will when the ref-count is non-zero
unconditionally run:

- pm_runtime_put_sync()
- vfe->ops->pm_domain_off()
- camss_disable_clocks()

vfe_get() should not attempt to roll-back on error when the ref-count is
non-zero as the upper layers will still do their own vfe_put() operations.

vfe_put() will drop the reference count and do the necessary power
domain release, the cleanup jumps in vfe_get() should only be run when
the ref-count is zero.

[   50.095796] CPU: 7 PID: 3075 Comm: cam Not tainted 6.3.2+ #80
[   50.095798] Hardware name: LENOVO 21BXCTO1WW/21BXCTO1WW, BIOS N3HET82W (1.54 ) 05/26/2023
[   50.095799] pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   50.095802] pc : refcount_warn_saturate+0xf4/0x148
[   50.095804] lr : refcount_warn_saturate+0xf4/0x148
[   50.095805] sp : ffff80000c7cb8b0
[   50.095806] x29: ffff80000c7cb8b0 x28: ffff16ecc0e3fc10 x27: 0000000000000000
[   50.095810] x26: 0000000000000000 x25: 0000000000020802 x24: 0000000000000000
[   50.095813] x23: ffff16ecc7360640 x22: 00000000ffffffff x21: 0000000000000005
[   50.095815] x20: ffff16ed175f4400 x19: ffffb4d9852942a8 x18: ffffffffffffffff
[   50.095818] x17: ffffb4d9852d4a48 x16: ffffb4d983da5db8 x15: ffff80000c7cb320
[   50.095821] x14: 0000000000000001 x13: 2e656572662d7265 x12: 7466612d65737520
[   50.095823] x11: 00000000ffffefff x10: ffffb4d9850cebf0 x9 : ffffb4d9835cf954
[   50.095826] x8 : 0000000000017fe8 x7 : c0000000ffffefff x6 : 0000000000057fa8
[   50.095829] x5 : ffff16f813fe3d08 x4 : 0000000000000000 x3 : ffff621e8f4d2000
[   50.095832] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff16ed32119040
[   50.095835] Call trace:
[   50.095836]  refcount_warn_saturate+0xf4/0x148
[   50.095838]  device_link_put_kref+0x84/0xc8
[   50.095843]  device_link_del+0x38/0x58
[   50.095846]  vfe_pm_domain_off+0x3c/0x50 [qcom_camss]
[   50.095860]  vfe_put+0x114/0x140 [qcom_camss]
[   50.095869]  csid_set_power+0x2c8/0x408 [qcom_camss]
[   50.095878]  pipeline_pm_power_one+0x164/0x170 [videodev]
[   50.095896]  pipeline_pm_power+0xc4/0x110 [videodev]
[   50.095909]  v4l2_pipeline_pm_use+0x5c/0xa0 [videodev]
[   50.095923]  v4l2_pipeline_pm_get+0x1c/0x30 [videodev]
[   50.095937]  video_open+0x7c/0x100 [qcom_camss]
[   50.095945]  v4l2_open+0x84/0x130 [videodev]
[   50.095960]  chrdev_open+0xc8/0x250
[   50.095964]  do_dentry_open+0x1bc/0x498
[   50.095966]  vfs_open+0x34/0x40
[   50.095968]  path_openat+0xb44/0xf20
[   50.095971]  do_filp_open+0xa4/0x160
[   50.095974]  do_sys_openat2+0xc8/0x188
[   50.095975]  __arm64_sys_openat+0x6c/0xb8
[   50.095977]  invoke_syscall+0x50/0x128
[   50.095982]  el0_svc_common.constprop.0+0x4c/0x100
[   50.095985]  do_el0_svc+0x40/0xa8
[   50.095988]  el0_svc+0x2c/0x88
[   50.095991]  el0t_64_sync_handler+0xf4/0x120
[   50.095994]  el0t_64_sync+0x190/0x198
[   50.095996] ---[ end trace 0000000000000000 ]---

Fixes: 779096916dae ("media: camss: vfe: Fix runtime PM imbalance on error")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: qcom: camss: Fix pm_domain_on sequence in probe

commit 7405116519ad70b8c7340359bfac8db8279e7ce4 upstream.

We need to make sure camss_configure_pd() happens before
camss_register_entities() as the vfe_get() path relies on the pointer
provided by camss_configure_pd().

Fix the ordering sequence in probe to ensure the pointers vfe_get() demands
are present by the time camss_register_entities() runs.

In order to facilitate backporting to stable kernels I've moved the
configure_pd() call pretty early on the probe() function so that
irrespective of the existence of the old error handling jump labels this
patch should still apply to -next circa Aug 2023 to v5.13 inclusive.

Fixes: 2f6f8af67203 ("media: camss: Refactor VFE power domain toggling")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: sdhci-pci-gli: GL9750: Mask the replay timer timeout of AER

commit 015c9cbcf0ad709079117d27c2094a46e0eadcdb upstream.

Due to a flaw in the hardware design, the GL9750 replay timer frequently
times out when ASPM is enabled. As a result, the warning messages will
often appear in the system log when the system accesses the GL9750
PCI config. Therefore, the replay timer timeout must be masked.

Fixes: d7133797e9e1 ("mmc: sdhci-pci-gli: A workaround to allow GL9750 to enter ASPM L1.2")
Signed-off-by: Victor Shih <victor.shih@genesyslogic.com.tw>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Kai-Heng Feng <kai.heng.geng@canonical.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231107095741.8832-2-victorshihgli@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

r8169: add handling DASH when DASH is disabled

commit 0ab0c45d8aaea5192328bfa6989673aceafc767c upstream.

For devices that support DASH, even DASH is disabled, there may still
exist a default firmware that will influence device behavior.
So driver needs to handle DASH for devices that support DASH, no
matter the DASH status is.

This patch also prepares for "fix network lost after resume on DASH
systems".

Fixes: ee7a1beb9759 ("r8169:call "rtl8168_driver_start" "rtl8168_driver_stop" only when hardware dash function is enabled")
Cc: stable@vger.kernel.org
Signed-off-by: ChunHao Lin <hau@realtek.com>
Reviewed-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/20231109173400.4573-2-hau@realtek.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

r8169: fix network lost after resume on DASH systems

commit 868c3b95afef4883bfb66c9397482da6840b5baf upstream.

Device that support DASH may be reseted or powered off during suspend.
So driver needs to handle DASH during system suspend and resume. Or
DASH firmware will influence device behavior and causes network lost.

Fixes: b646d90053f8 ("r8169: magic.")
Cc: stable@vger.kernel.org
Reviewed-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: ChunHao Lin <hau@realtek.com>
Link: https://lore.kernel.org/r/20231109173400.4573-3-hau@realtek.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests: mptcp: fix fastclose with csum failure

commit 7cefbe5e1dacc7236caa77e9d072423f21422fe2 upstream.

Running the mp_join selftest manually with the following command line:

  ./mptcp_join.sh -z -C

leads to some failures:

  002 fastclose server test
  # ...
  rtx                                 [fail] got 1 MP_RST[s] TX expected 0
  # ...
  rstrx                               [fail] got 1 MP_RST[s] RX expected 0

The problem is really in the wrong expectations for the RST checks
implied by the csum validation. Note that the same check is repeated
explicitly in the same test-case, with the correct expectation and
pass successfully.

Address the issue explicitly setting the correct expectation for
the failing checks.

Reported-by: Xiumei Mu <xmu@redhat.com>
Fixes: 6bf41020b72b ("selftests: mptcp: update and extend fastclose test-cases")
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts <matttbe@kernel.org>
Link: https://lore.kernel.org/r/20231114-upstream-net-20231113-mptcp-misc-fixes-6-7-rc2-v1-5-7b9cd6a7b7f4@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mptcp: fix setsockopt(IP_TOS) subflow locking

commit 7679d34f97b7a09fd565f5729f79fd61b7c55329 upstream.

The MPTCP implementation of the IP_TOS socket option uses the lockless
variant of the TOS manipulation helper and does not hold such lock at
the helper invocation time.

Add the required locking.

Fixes: ffcacff87cd6 ("mptcp: Support for IP_TOS for MPTCP setsockopt()")
Cc: stable@vger.kernel.org
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/457
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts <matttbe@kernel.org>
Link: https://lore.kernel.org/r/20231114-upstream-net-20231113-mptcp-misc-fixes-6-7-rc2-v1-4-7b9cd6a7b7f4@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mptcp: add validity check for sending RM_ADDR

commit 8df220b29282e8b450ea57be62e1eccd4996837c upstream.

This patch adds the validity check for sending RM_ADDRs for userspace PM
in mptcp_pm_remove_addrs(), only send a RM_ADDR when the address is in the
anno_list or conn_list.

Fixes: 8b1c94da1e48 ("mptcp: only send RM_ADDR in nl_cmd_remove")
Cc: stable@vger.kernel.org
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts <matttbe@kernel.org>
Link: https://lore.kernel.org/r/20231114-upstream-net-20231113-mptcp-misc-fixes-6-7-rc2-v1-3-7b9cd6a7b7f4@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mptcp: deal with large GSO size

commit 9fce92f050f448a0d1ddd9083ef967d9930f1e52 upstream.

After the blamed commit below, the TCP sockets (and the MPTCP subflows)
can build egress packets larger than 64K. That exceeds the maximum DSS
data size, the length being misrepresent on the wire and the stream being
corrupted, as later observed on the receiver:

  WARNING: CPU: 0 PID: 9696 at net/mptcp/protocol.c:705 __mptcp_move_skbs_from_subflow+0x2604/0x26e0
  CPU: 0 PID: 9696 Comm: syz-executor.7 Not tainted 6.6.0-rc5-gcd8bdf563d46 #45
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.el7 04/01/2014
  netlink: 8 bytes leftover after parsing attributes in process `syz-executor.4'.
  RIP: 0010:__mptcp_move_skbs_from_subflow+0x2604/0x26e0 net/mptcp/protocol.c:705
  RSP: 0018:ffffc90000006e80 EFLAGS: 00010246
  RAX: ffffffff83e9f674 RBX: ffff88802f45d870 RCX: ffff888102ad0000
  netlink: 8 bytes leftover after parsing attributes in process `syz-executor.4'.
  RDX: 0000000080000303 RSI: 0000000000013908 RDI: 0000000000003908
  RBP: ffffc90000007110 R08: ffffffff83e9e078 R09: 1ffff1100e548c8a
  R10: dffffc0000000000 R11: ffffed100e548c8b R12: 0000000000013908
  R13: dffffc0000000000 R14: 0000000000003908 R15: 000000000031cf29
  FS:  00007f239c47e700(0000) GS:ffff88811b200000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00007f239c45cd78 CR3: 000000006a66c006 CR4: 0000000000770ef0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000600
  PKRU: 55555554
  Call Trace:
   <IRQ>
   mptcp_data_ready+0x263/0xac0 net/mptcp/protocol.c:819
   subflow_data_ready+0x268/0x6d0 net/mptcp/subflow.c:1409
   tcp_data_queue+0x21a1/0x7a60 net/ipv4/tcp_input.c:5151
   tcp_rcv_established+0x950/0x1d90 net/ipv4/tcp_input.c:6098
   tcp_v6_do_rcv+0x554/0x12f0 net/ipv6/tcp_ipv6.c:1483
   tcp_v6_rcv+0x2e26/0x3810 net/ipv6/tcp_ipv6.c:1749
   ip6_protocol_deliver_rcu+0xd6b/0x1ae0 net/ipv6/ip6_input.c:438
   ip6_input+0x1c5/0x470 net/ipv6/ip6_input.c:483
   ipv6_rcv+0xef/0x2c0 include/linux/netfilter.h:304
   __netif_receive_skb+0x1ea/0x6a0 net/core/dev.c:5532
   process_backlog+0x353/0x660 net/core/dev.c:5974
   __napi_poll+0xc6/0x5a0 net/core/dev.c:6536
   net_rx_action+0x6a0/0xfd0 net/core/dev.c:6603
   __do_softirq+0x184/0x524 kernel/softirq.c:553
   do_softirq+0xdd/0x130 kernel/softirq.c:454

Address the issue explicitly bounding the maximum GSO size to what MPTCP
actually allows.

Reported-by: Christoph Paasch <cpaasch@apple.com>
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/450
Fixes: 7c4e983c4f3c ("net: allow gso_max_size to exceed 65536")
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts <matttbe@kernel.org>
Link: https://lore.kernel.org/r/20231114-upstream-net-20231113-mptcp-misc-fixes-6-7-rc2-v1-1-7b9cd6a7b7f4@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>