selftests: mptcp: connect: skip disconnect tests if not supported

commit 4ad39a42da2e9770c8e4c37fe632ed8898419129 upstream.

Selftests are supposed to run on any kernels, including the old ones not
supporting all MPTCP features.

One of them is the full support of disconnections from the userspace
introduced by commit b29fcfb54cd7 ("mptcp: full disconnect
implementation").

It is possible to look for "mptcp_pm_data_reset" in kallsyms because a
preparation patch added it to ease the introduction of the mentioned
feature.

Link: https://github.com/multipath-tcp/mptcp_net-next/issues/368
Fixes: 05be5e273c84 ("selftests: mptcp: add disconnect tests")
Cc: stable@vger.kernel.org
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests: mptcp: connect: skip transp tests if not supported

commit 07bf49401909264a38fa3427c3cce43e8304436a upstream.

Selftests are supposed to run on any kernels, including the old ones not
supporting all MPTCP features.

One of them is the support of IP(V6)_TRANSPARENT socket option with
MPTCP connections introduced by commit c9406a23c116 ("mptcp: sockopt:
add SOL_IP freebind & transparent options").

It is possible to look for "__ip_sock_set_tos" in kallsyms because
IP(V6)_TRANSPARENT socket option support has been added after TOS
support which came with the required infrastructure in MPTCP sockopt
code. To support TOS, the following function has been exported (T). Not
great but better than checking for a specific kernel version.

Link: https://github.com/multipath-tcp/mptcp_net-next/issues/368
Fixes: 5fb62e9cd3ad ("selftests: mptcp: add tproxy test case")
Cc: stable@vger.kernel.org
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests: mptcp: lib: skip if missing symbol

commit 673004821ab98c6645bd21af56a290854e88f533 upstream.

Selftests are supposed to run on any kernels, including the old ones not
supporting all MPTCP features.

New functions are now available to easily detect if a certain feature is
missing by looking at kallsyms.

These new helpers are going to be used in the following commits. In
order to ease the backport of such future patches, it would be good if
this patch is backported up to the introduction of MPTCP selftests,
hence the Fixes tag below: this type of check was supposed to be done
from the beginning.

Link: https://github.com/multipath-tcp/mptcp_net-next/issues/368
Fixes: 048d19d444be ("mptcp: add basic kselftest for mptcp")
Cc: stable@vger.kernel.org
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests: mptcp: join: fix ShellCheck warnings

commit 0fcd72df8847d3a62eb34a084862157ce0564a94 upstream.

Most of the code had an issue according to ShellCheck.

That's mainly due to the fact it incorrectly believes most of the code
was unreachable because it's invoked by variable name, see how the
"tests" array is used.

Once SC2317 has been ignored, three small warnings were still visible:

 - SC2155: Declare and assign separately to avoid masking return values.

 - SC2046: Quote this to prevent word splitting: can be ignored because
   "ip netns pids" can display more than one pid.

 - SC2166: Prefer [ p ] || [ q ] as [ p -o q ] is not well defined.

This probably didn't fix any actual issues but it might help spotting
new interesting warnings reported by ShellCheck as just before,
ShellCheck was reporting issues for most lines making it a bit useless.

Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests: mptcp: remove duplicated entries in usage

commit 0a85264e48b642d360720589fdb837a3643fb9b0 upstream.

mptcp_connect tool was printing some duplicated entries when showing how
to use it: -j -l -r

While at it, I also:

 - moved the very few entries that were not sorted,

 - added -R that was missing since
   commit 8a4b910d005d ("mptcp: selftests: add rcvbuf set option"),

 - removed the -u parameter that has been removed in
   commit f730b65c9d85 ("selftests: mptcp: try to set mptcp ulp mode in different sk states").

No need to backport this, it is just an internal tool used by our
selftests. The help menu is mainly useful for MPTCP kernel devs.

Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tick/common: Align tick period during sched_timer setup

commit 13bb06f8dd42071cb9a49f6e21099eea05d4b856 upstream.

The tick period is aligned very early while the first clock_event_device is
registered. At that point the system runs in periodic mode and switches
later to one-shot mode if possible.

The next wake-up event is programmed based on the aligned value
(tick_next_period) but the delta value, that is used to program the
clock_event_device, is computed based on ktime_get().

With the subtracted offset, the device fires earlier than the exact time
frame. With a large enough offset the system programs the timer for the
next wake-up and the remaining time left is too small to make any boot
progress. The system hangs.

Move the alignment later to the setup of tick_sched timer. At this point
the system switches to oneshot mode and a high resolution clocksource is
available. At this point it is safe to align tick_next_period because
ktime_get() will now return accurate (not jiffies based) time.

[bigeasy: Patch description + testing].

Fixes: e9523a0d81899 ("tick/common: Align tick period with the HZ tick.")
Reported-by: Mathias Krause <minipli@grsecurity.net>
Reported-by: "Bhatnagar, Rishabh" <risbhat@amazon.com>
Suggested-by: Mathias Krause <minipli@grsecurity.net>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Richard W.M. Jones <rjones@redhat.com>
Tested-by: Mathias Krause <minipli@grsecurity.net>
Acked-by: SeongJae Park <sj@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/5a56290d-806e-b9a5-f37c-f21958b5a8c0@grsecurity.net
Link: https://lore.kernel.org/12c6f9a3-d087-b824-0d05-0d18c9bc1bf3@amazon.com
Link: https://lore.kernel.org/r/20230615091830.RxMV2xf_@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ksmbd: validate session id and tree id in the compound request

commit 5005bcb4219156f1bf7587b185080ec1da08518e upstream.

This patch validate session id and tree id in compound request.
If first operation in the compound is SMB2 ECHO request, ksmbd bypass
session and tree validation. So work->sess and work->tcon could be NULL.
If secound request in the compound access work->sess or tcon, It cause
NULL pointer dereferecing error.

Cc: stable@vger.kernel.org
Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-21165
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ksmbd: fix out-of-bound read in smb2_write

commit 5fe7f7b78290638806211046a99f031ff26164e1 upstream.

ksmbd_smb2_check_message doesn't validate hdr->NextCommand. If
->NextCommand is bigger than Offset + Length of smb2 write, It will
allow oversized smb2 write length. It will cause OOB read in smb2_write.

Cc: stable@vger.kernel.org
Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-21164
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ksmbd: validate command payload size

commit 2b9b8f3b68edb3d67d79962f02e26dbb5ae3808d upstream.

->StructureSize2 indicates command payload size. ksmbd should validate
this size with rfc1002 length before accessing it.
This patch remove unneeded check and add the validation for this.

[    8.912583] BUG: KASAN: slab-out-of-bounds in ksmbd_smb2_check_message+0x12a/0xc50
[    8.913051] Read of size 2 at addr ffff88800ac7d92c by task kworker/0:0/7
...
[    8.914967] Call Trace:
[    8.915126]  <TASK>
[    8.915267]  dump_stack_lvl+0x33/0x50
[    8.915506]  print_report+0xcc/0x620
[    8.916558]  kasan_report+0xae/0xe0
[    8.917080]  kasan_check_range+0x35/0x1b0
[    8.917334]  ksmbd_smb2_check_message+0x12a/0xc50
[    8.917935]  ksmbd_verify_smb_message+0xae/0xd0
[    8.918223]  handle_ksmbd_work+0x192/0x820
[    8.918478]  process_one_work+0x419/0x760
[    8.918727]  worker_thread+0x2a2/0x6f0
[    8.919222]  kthread+0x187/0x1d0
[    8.919723]  ret_from_fork+0x1f/0x30
[    8.919954]  </TASK>

Cc: stable@vger.kernel.org
Reported-by: Chih-Yen Chang <cc85nod@gmail.com>
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tpm_crb: Add support for CRB devices based on Pluton

commit 4d2732882703791ea4b670df433f88fc4b40a5cb upstream.

Pluton is an integrated security processor present in some recent Ryzen
parts. If it's enabled, it presents two devices - an MSFT0101 ACPI device
that's broadly an implementation of a Command Response Buffer TPM2, and an
MSFT0200 ACPI device whose functionality I haven't examined in detail yet.
This patch only attempts to add support for the TPM device.

There's a few things that need to be handled here. The first is that the
TPM2 ACPI table uses a previously undefined start method identifier. The
table format appears to include 16 bytes of startup data, which corresponds
to one 64-bit address for a start message and one 64-bit address for a
completion response. The second is that the ACPI tables on the Thinkpad Z13
I'm testing this on don't define any memory windows in _CRS (or, more
accurately, there are two empty memory windows). This check doesn't seem
strictly necessary, so I've skipped that.

Finally, it seems like chip needs to be explicitly asked to transition into
ready status on every command. Failing to do this means that if two
commands are sent in succession without an idle/ready transition in
between, everything will appear to work fine but the response is simply the
original command. I'm working without any docs here, so I'm not sure if
this is actually the required behaviour or if I'm missing something
somewhere else, but doing this results in the chip working reliably.

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Matthew Garrett <mjg59@srcf.ucam.org>
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>
Cc: "Limonciello, Mario" <mario.limonciello@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tpm, tpm_tis: Claim locality in interrupt handler

commit 0e069265bce5a40c4eee52e2364bbbd4dabee94a upstream.

Writing the TPM_INT_STATUS register in the interrupt handler to clear the
interrupts only has effect if a locality is held. Since this is not
guaranteed at the time the interrupt is fired, claim the locality
explicitly in the handler.

Signed-off-by: Lino Sanfilippo <l.sanfilippo@kunbus.com>
Tested-by: Michael Niewöhner <linux@mniewoehner.de>
Tested-by: Jarkko Sakkinen <jarkko@kernel.org>
Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm: Fix copy_from_user_nofault().

commit d319f344561de23e810515d109c7278919bff7b0 upstream.

There are several issues with copy_from_user_nofault():

- access_ok() is designed for user context only and for that reason
it has WARN_ON_IN_IRQ() which triggers when bpf, kprobe, eprobe
and perf on ppc are calling it from irq.

- it's missing nmi_uaccess_okay() which is a nop on all architectures
except x86 where it's required.
The comment in arch/x86/mm/tlb.c explains the details why it's necessary.
Calling copy_from_user_nofault() from bpf, [ke]probe without this check is not safe.

- __copy_from_user_inatomic() under CONFIG_HARDENED_USERCOPY is calling
check_object_size()->__check_object_size()->check_heap_object()->find_vmap_area()->spin_lock()
which is not safe to do from bpf, [ke]probe and perf due to potential deadlock.

Fix all three issues. At the end the copy_from_user_nofault() becomes
equivalent to copy_from_user_nmi() from safety point of view with
a difference in the return value.

Reported-by: Hsin-Wei Hung <hsinweih@uci.edu>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Florian Lehner <dev@der-flo.net>
Tested-by: Hsin-Wei Hung <hsinweih@uci.edu>
Tested-by: Florian Lehner <dev@der-flo.net>
Link: https://lore.kernel.org/r/20230410174345.4376-2-dev@der-flo.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Cc: Javier Honduvilla Coto <javierhonduco@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ata: libata-scsi: Avoid deadlock on rescan after device resume

[ Upstream commit 6aa0365a3c8512587fffd42fe438768709ddef8e ]

When an ATA port is resumed from sleep, the port is reset and a power
management request issued to libata EH to reset the port and rescanning
the device(s) attached to the port. Device rescanning is done by
scheduling an ata_scsi_dev_rescan() work, which will execute
scsi_rescan_device().

However, scsi_rescan_device() takes the generic device lock, which is
also taken by dpm_resume() when the SCSI device is resumed as well. If
a device rescan execution starts before the completion of the SCSI
device resume, the rcu locking used to refresh the cached VPD pages of
the device, combined with the generic device locking from
scsi_rescan_device() and from dpm_resume() can cause a deadlock.

Avoid this situation by changing struct ata_port scsi_rescan_task to be
a delayed work instead of a simple work_struct. ata_scsi_dev_rescan() is
modified to check if the SCSI device associated with the ATA device that
must be rescanned is not suspended. If the SCSI device is still
suspended, ata_scsi_dev_rescan() returns early and reschedule itself for
execution after an arbitrary delay of 5ms.

Reported-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Reported-by: Joe Breuer <linux-kernel@jmbreuer.net>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217530
Fixes: a19a93e4c6a9 ("scsi: core: pm: Rely on the device driver core for async power management")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Tested-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Joe Breuer <linux-kernel@jmbreuer.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tty: serial: fsl_lpuart: reduce RX watermark to 0 on LS1028A

[ Upstream commit a82c3df955f8c1c726e4976527aa6ae924a67dd9 ]

LS1028A is using DMA with LPUART. Having RX watermark set to 1, means
DMA transactions are started only after receiving the second character.

On other platforms with newer LPUART IP, Receiver Idle Empty function
initiates the DMA request after the receiver is idling for 4 characters.
But this feature is missing on LS1028A, which is causing a 1-character
delay in the RX direction on this platform.

Set RX watermark to 0 to initiate RX DMA after each character.

Link: https://lore.kernel.org/linux-serial/20230607103459.1222426-1-robert.hodaszi@digi.com/
Fixes: 9ad9df844754 ("tty: serial: fsl_lpuart: Fix the wrong RXWATER setting for rx dma case")
Cc: stable <stable@kernel.org>
Signed-off-by: Robert Hodaszi <robert.hodaszi@digi.com>
Message-ID: <20230609121334.1878626-1-robert.hodaszi@digi.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tty: serial: fsl_lpuart: make rx_watermark configurable for different platforms

[ Upstream commit 34ebb26f12a84b744f43c5c4869516f122a2dfaa ]

Add rx_watermark parameter for struct lpuart_port to make the receive
watermark configurable for different platforms.
No function changed.

Signed-off-by: Sherry Sun <sherry.sun@nxp.com>
Link: https://lore.kernel.org/r/20230130064449.9564-2-sherry.sun@nxp.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Stable-dep-of: a82c3df955f8 ("tty: serial: fsl_lpuart: reduce RX watermark to 0 on LS1028A")
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: fix the system hang while disable PSR

[ Upstream commit ea2062dd1f0384ae1b136d333ee4ced15bedae38 ]

[Why]
When the PSR enabled. If you try to adjust the timing parameters,
it may cause system hang. Because the timing mismatch with the
DMCUB settings.

[How]
Disable the PSR before adjusting timing parameters.

Cc: Mario Limonciello <mario.limonciello@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Acked-by: Stylon Wang <stylon.wang@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Reviewed-by: Wayne Lin <Wayne.Lin@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Add wrapper to call planes and stream update

[ Upstream commit 81f743a08f3b214638aa389e252ae5e6c3592e7c ]

[Why & How]
This commit is part of a sequence of changes that replaces the commit
sequence used in the DC with a new one. As a result of this transition,
we moved some specific parts from the commit sequence and brought them
to amdgpu_dm. This commit adds a wrapper inside DM that enable our
drivers to do any necessary preparation or change before we offload the
plane/stream update to DC.

Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Stable-dep-of: ea2062dd1f03 ("drm/amd/display: fix the system hang while disable PSR")
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Use dc_update_planes_and_stream

[ Upstream commit f7511289821ffccc07579406d6ab520aa11049f5 ]

[Why & How]
The old dc_commit_updates_for_stream lacks manipulation for many corner
cases where the DC feature requires special attention; as a result, it
starts to show its limitation (e.g., the SubVP feature is not supported
by it, among other cases). To modernize and unify our internal API, this
commit replaces the old dc_commit_updates_for_stream with
dc_update_planes_and_stream, which has more features.

Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Stable-dep-of: ea2062dd1f03 ("drm/amd/display: fix the system hang while disable PSR")
Signed-off-by: Sasha Levin <sashal@kernel.org>

Linux 6.1.35

Link: https://lore.kernel.org/r/20230619102154.568541872@linuxfoundation.org
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Markus Reichelt <lkt+2023@mareichelt.com>
Tested-by: Salvatore Bonaccorso <carnil@debian.org>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Chris Paterson (CIP) <chris.paterson2@renesas.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Conor Dooley <conor.dooley@microchip.com>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Tested-by: Takeshi Ogasawara <takeshi.ogasawara@futuring-girl.com>
Tested-by: Allen Pais <apais@linux.microsoft.com>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

kbuild: Update assembler calls to use proper flags and language target

commit d5c8d6e0fa61401a729e9eb6a9c7077b2d3aebb0 upstream.

as-instr uses KBUILD_AFLAGS, but as-option uses KBUILD_CFLAGS. This can
cause as-option to fail unexpectedly when CONFIG_WERROR is set, because
clang will emit -Werror,-Wunused-command-line-argument for various -m
and -f flags in KBUILD_CFLAGS for assembler sources.

Callers of as-option and as-instr should be adding flags to
KBUILD_AFLAGS / aflags-y, not KBUILD_CFLAGS / cflags-y. Use
KBUILD_AFLAGS in all macros to clear up the initial problem.

Unfortunately, -Wunused-command-line-argument can still be triggered
with clang by the presence of warning flags or macro definitions because
'-x assembler' is used, instead of '-x assembler-with-cpp', which will
consume these flags. Switch to '-x assembler-with-cpp' in places where
'-x assembler' is used, as the compiler is always used as the driver for
out of line assembler sources in the kernel.

Finally, add -Werror to these macros so that they behave consistently
whether or not CONFIG_WERROR is set.

[nathan: Reworded and expanded on problems in commit message
         Use '-x assembler-with-cpp' in a couple more places]

Link: https://github.com/ClangBuiltLinux/linux/issues/1699
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

MIPS: Prefer cc-option for additions to cflags

commit 337ff6bb8960fdc128cabd264aaea3d42ca27a32 upstream.

A future change will switch as-option to use KBUILD_AFLAGS instead of
KBUILD_CFLAGS to allow clang to drop -Qunused-arguments, which may cause
issues if the flag being tested requires a flag previously added to
KBUILD_CFLAGS but not KBUILD_AFLAGS. Use cc-option for cflags additions
so that the flags are tested properly.

Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

MIPS: Move '-Wa,-msoft-float' check from as-option to cc-option

This patch is for linux-6.1.y and earlier, it has no direct mainline
equivalent.

In order to backport commit d5c8d6e0fa61 ("kbuild: Update assembler
calls to use proper flags and language target") to resolve a separate
issue regarding PowerPC, the problem noticed and fixed by
commit 80a20d2f8288 ("MIPS: Always use -Wa,-msoft-float and eliminate
GAS_HAS_SET_HARDFLOAT") needs to be addressed. Unfortunately, 6.1 and
earlier do not contain commit e4412739472b ("Documentation: raise
minimum supported version of binutils to 2.25"), so it cannot be assumed
that all supported versions of GNU as have support for -msoft-float.

In order to switch from KBUILD_CFLAGS to KBUILD_AFLAGS in as-option
without consequence, move the '-Wa,-msoft-float' check to cc-option,
including '$(cflags-y)' directly to avoid the issue mentioned in
commit 80a20d2f8288 ("MIPS: Always use -Wa,-msoft-float and eliminate
GAS_HAS_SET_HARDFLOAT").

Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/boot/compressed: prefer cc-option for CFLAGS additions

commit 994f5f7816ff963f49269cfc97f63cb2e4edb84f upstream.

as-option tests new options using KBUILD_CFLAGS, which causes problems
when using as-option to update KBUILD_AFLAGS because many compiler
options are not valid assembler options.

This will be fixed in a follow up patch. Before doing so, move the
assembler test for -Wa,-mrelax-relocations=no from using as-option to
cc-option.

Link: https://lore.kernel.org/llvm/CAK7LNATcHt7GcXZ=jMszyH=+M_LC9Qr6yeAGRCBbE6xriLxtUQ@mail.gmail.com/
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: target: core: Fix error path in target_setup_session()

commit 91271699228bfc66f1bc8abc0327169dc156d854 upstream.

In the error exits in target_setup_session(), if a branch is taken to
free_sess: transport_free_session() may call to target_free_cmd_counter()
and then fall through to call target_free_cmd_counter() a second time.
This can, and does, sometimes cause seg faults since the data field in
cmd_cnt->refcnt has been freed in the first call.

Fix this problem by simply returning after the call to
transport_free_session(). The second call is redundant for those cases.

Fixes: 4edba7e4a8f3 ("scsi: target: Move cmd counter allocation")
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Link: https://lore.kernel.org/r/20230613144259.12890-1-rpearsonhpe@gmail.com
Reviewed-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

neighbour: delete neigh_lookup_nodev as not used

commit 76b9bf965c98c9b53ef7420b3b11438dbd764f92 upstream.

neigh_lookup_nodev isn't used in the kernel after removal
of DECnet. So let's remove it.

Fixes: 1202cdd66531 ("Remove DECnet support from kernel")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://lore.kernel.org/r/eb5656200d7964b2d177a36b77efa3c597d6d72d.1678267343.git.leonro@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/sched: act_api: add specific EXT_WARN_MSG for tc action

commit 2f59823fe696caa844249a90bb3f9aeda69cfe5c upstream.

In my previous commit 0349b8779cc9 ("sched: add new attr TCA_EXT_WARN_MSG
to report tc extact message") I didn't notice the tc action use different
enum with filter. So we can't use TCA_EXT_WARN_MSG directly for tc action.
Let's add a TCA_ROOT_EXT_WARN_MSG for tc action specifically and put this
param before going to the TCA_ACT_TAB nest.

Fixes: 0349b8779cc9 ("sched: add new attr TCA_EXT_WARN_MSG to report tc extact message")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "net/sched: act_api: move TCA_EXT_WARN_MSG to the correct hierarchy"

commit 8de2bd02439eb839a452a853c1004c2c45ff6fef upstream.

This reverts commit 923b2e30dc9cd05931da0f64e2e23d040865c035.

This is not a correct fix as TCA_EXT_WARN_MSG is not a hierarchy to
TCA_ACT_TAB. I didn't notice the TC actions use different enum when adding
TCA_EXT_WARN_MSG. To fix the difference I will add a new WARN enum in
TCA_ROOT_MAX as Jamal suggested.

Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/sched: act_api: move TCA_EXT_WARN_MSG to the correct hierarchy

commit 923b2e30dc9cd05931da0f64e2e23d040865c035 upstream.

TCA_EXT_WARN_MSG is currently sitting outside of the expected hierarchy
for the tc actions code. It should sit within TCA_ACT_TAB.

Fixes: 0349b8779cc9 ("sched: add new attr TCA_EXT_WARN_MSG to report tc extact message")
Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: Don't set struct drm_driver.output_poll_changed

commit 0e3172bac3f43759719384403fe2d1e4c61f87e0 upstream.

Don't set struct drm_driver.output_poll_changed. It's used to restore
the fbdev console. But as amdgpu uses generic fbdev emulation, the
console is being restored by the DRM client helpers already. See the
functions drm_kms_helper_hotplug_event() and
drm_kms_helper_connector_hotplug_event() in drm_probe_helper.c.

v2:
* fix commit description (Christian)

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221103151446.2638-5-tzimmermann@suse.de
Cc: "Gagniuc, Alexandru" <alexandru.gagniuc@hp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

rcu/kvfree: Avoid freeing new kfree_rcu() memory after old grace period

commit 5da7cb193db32da783a3f3e77d8b639989321d48 upstream.

Memory passed to kvfree_rcu() that is to be freed is tracked by a
per-CPU kfree_rcu_cpu structure, which in turn contains pointers
to kvfree_rcu_bulk_data structures that contain pointers to memory
that has not yet been handed to RCU, along with an kfree_rcu_cpu_work
structure that tracks the memory that has already been handed to RCU.
These structures track three categories of memory: (1) Memory for
kfree(), (2) Memory for kvfree(), and (3) Memory for both that arrived
during an OOM episode.  The first two categories are tracked in a
cache-friendly manner involving a dynamically allocated page of pointers
(the aforementioned kvfree_rcu_bulk_data structures), while the third
uses a simple (but decidedly cache-unfriendly) linked list through the
rcu_head structures in each block of memory.

On a given CPU, these three categories are handled as a unit, with that
CPU's kfree_rcu_cpu_work structure having one pointer for each of the
three categories.  Clearly, new memory for a given category cannot be
placed in the corresponding kfree_rcu_cpu_work structure until any old
memory has had its grace period elapse and thus has been removed.  And
the kfree_rcu_monitor() function does in fact check for this.

Except that the kfree_rcu_monitor() function checks these pointers one
at a time.  This means that if the previous kfree_rcu() memory passed
to RCU had only category 1 and the current one has only category 2, the
kfree_rcu_monitor() function will send that current category-2 memory
along immediately.  This can result in memory being freed too soon,
that is, out from under unsuspecting RCU readers.

To see this, consider the following sequence of events, in which:

o Task A on CPU 0 calls rcu_read_lock(), then uses "from_cset",
then is preempted.

o CPU 1 calls kfree_rcu(cset, rcu_head) in order to free "from_cset"
after a later grace period.  Except that "from_cset" is freed
right after the previous grace period ended, so that "from_cset"
is immediately freed.  Task A resumes and references "from_cset"'s
member, after which nothing good happens.

In full detail:

CPU 0 CPU 1
---------------------- ----------------------
count_memcg_event_mm()
|rcu_read_lock()  <---
|mem_cgroup_from_task()
 |// css_set_ptr is the "from_cset" mentioned on CPU 1
 |css_set_ptr = rcu_dereference((task)->cgroups)
 |// Hard irq comes, current task is scheduled out.

cgroup_attach_task()
|cgroup_migrate()
|cgroup_migrate_execute()
|css_set_move_task(task, from_cset, to_cset, true)
|cgroup_move_task(task, to_cset)
|rcu_assign_pointer(.., to_cset)
|...
|cgroup_migrate_finish()
|put_css_set_locked(from_cset)
|from_cset->refcount return 0
|kfree_rcu(cset, rcu_head) // free from_cset after new gp
|add_ptr_to_bulk_krc_lock()
|schedule_delayed_work(&krcp->monitor_work, ..)

kfree_rcu_monitor()
|krcp->bulk_head[0]'s work attached to krwp->bulk_head_free[]
|queue_rcu_work(system_wq, &krwp->rcu_work)
|if rwork->rcu.work is not in WORK_STRUCT_PENDING_BIT state,
|call_rcu(&rwork->rcu, rcu_work_rcufn) <--- request new gp

// There is a perious call_rcu(.., rcu_work_rcufn)
// gp end, rcu_work_rcufn() is called.
rcu_work_rcufn()
|__queue_work(.., rwork->wq, &rwork->work);

|kfree_rcu_work()
|krwp->bulk_head_free[0] bulk is freed before new gp end!!!
|The "from_cset" is freed before new gp end.

// the task resumes some time later.
 |css_set_ptr->subsys[(subsys_id) <--- Caused kernel crash, because css_set_ptr is freed.

This commit therefore causes kfree_rcu_monitor() to refrain from moving
kfree_rcu() memory to the kfree_rcu_cpu_work structure until the RCU
grace period has completed for all three categories.

v2: Use helper function instead of inserted code block at kfree_rcu_monitor().

Fixes: 34c881745549 ("rcu: Support kfree_bulk() interface in kfree_rcu()")
Fixes: 5f3c8d620447 ("rcu/tree: Maintain separate array for vmalloc ptrs")
Reported-by: Mukesh Ojha <quic_mojha@quicinc.com>
Signed-off-by: Ziwei Dai <ziwei.dai@unisoc.com>
Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Tested-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc: Delete redundant register definitions in <asm/assembly.h>

commit b5b2a02bcaac7c287694aa0db4837a07bf178626 upstream.

We define sp and ipsw in <asm/asmregs.h> using ".reg", and when using
current binutils (snapshot 2.40.50.20230611) the definitions in
<asm/assembly.h> using "=" conflict with those:

arch/parisc/include/asm/assembly.h: Assembler messages:
arch/parisc/include/asm/assembly.h:93: Error: symbol `sp' is already defined
arch/parisc/include/asm/assembly.h:95: Error: symbol `ipsw' is already defined

Delete the duplicate definitions in <asm/assembly.h>.

Also delete the definition of gp, which isn't used anywhere.

Signed-off-by: Ben Hutchings <benh@debian.org>
Cc: stable@vger.kernel.org # v6.0+
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

afs: Fix vlserver probe RTT handling

[ Upstream commit ba00b190670809c1a89326d80de96d714f6004f2 ]

In the same spirit as commit ca57f02295f1 ("afs: Fix fileserver probe
RTT handling"), don't rule out using a vlserver just because there
haven't been enough packets yet to calculate a real rtt.  Always set the
server's probe rtt from the estimate provided by rxrpc_kernel_get_srtt,
which is capped at 1 second.

This could lead to EDESTADDRREQ errors when accessing a cell for the
first time, even though the vl servers are known and have responded to a
probe.

Fixes: 1d4adfaf6574 ("rxrpc: Make rxrpc_kernel_get_srtt() indicate validity")
Signed-off-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: linux-afs@lists.infradead.org
Link: http://lists.infradead.org/pipermail/linux-afs/2023-June/006746.html
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeon_ep: Add missing check for ioremap

[ Upstream commit 9a36e2d44d122fe73a2a76ba73f1d50a65cf8210 ]

Add check for ioremap() and return the error if it fails in order to
guarantee the success of ioremap().

Fixes: 862cd659a6fb ("octeon_ep: Add driver framework and device initialization")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Link: https://lore.kernel.org/r/20230615033400.2971-1-jiasheng@iscas.ac.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/ptp: Fix timestamp printf format for PTP_SYS_OFFSET

[ Upstream commit 76a4c8b82938bc5020b67663db41f451684bf327 ]

Previously, timestamps were printed using "%lld.%u" which is incorrect
for nanosecond values lower than 100,000,000 as they're fractional
digits, therefore leading zeros are meaningful.

This patch changes the format strings to "%lld.%09u" in order to add
leading zeros to the nanosecond value.

Fixes: 568ebc5985f5 ("ptp: add the PTP_SYS_OFFSET ioctl to the testptp program")
Fixes: 4ec54f95736f ("ptp: Fix compiler warnings in the testptp utility")
Fixes: 6ab0e475f1f3 ("Documentation: fix misc. warnings")
Signed-off-by: Alex Maftei <alex.maftei@amd.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Link: https://lore.kernel.org/r/20230615083404.57112-1-alex.maftei@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: tipc: resize nlattr array to correct size

[ Upstream commit 44194cb1b6045dea33ae9a0d54fb7e7cd93a2e09 ]

According to nla_parse_nested_deprecated(), the tb[] is supposed to the
destination array with maxtype+1 elements. In current
tipc_nl_media_get() and __tipc_nl_media_set(), a larger array is used
which is unnecessary. This patch resize them to a proper size.

Fixes: 1e55417d8fc6 ("tipc: add media set to new netlink api")
Fixes: 46f15c6794fb ("tipc: add media get/dump to new netlink api")
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Reviewed-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Tung Nguyen <tung.q.nguyen@dektech.com.au>
Link: https://lore.kernel.org/r/20230614120604.1196377-1-linma@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm: don't lock fs when the map is NULL during suspend or resume

[ Upstream commit 2760904d895279f87196f0fa9ec570c79fe6a2e4 ]

As described in commit 38d11da522aa ("dm: don't lock fs when the map is
NULL in process of resume"), a deadlock may be triggered between
do_resume() and do_mount().

This commit preserves the fix from commit 38d11da522aa but moves it to
where it also serves to fix a similar deadlock between do_suspend()
and do_mount().  It does so, if the active map is NULL, by clearing
DM_SUSPEND_LOCKFS_FLAG in dm_suspend() which is called by both
do_suspend() and do_resume().

Fixes: 38d11da522aa ("dm: don't lock fs when the map is NULL in process of resume")
Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sfc: fix XDP queues mode with legacy IRQ

[ Upstream commit e84a1e1e683f3558e30f437d7c99df35afb8b52c ]

In systems without MSI-X capabilities, xdp_txq_queues_mode is calculated
in efx_allocate_msix_channels, but when enabling MSI-X fails, it was not
changed to a proper default value. This was leading to the driver
thinking that it has dedicated XDP queues, when it didn't.

Fix it by setting xdp_txq_queues_mode to the correct value if the driver
fallbacks to MSI or legacy IRQ mode. The correct value is
EFX_XDP_TX_QUEUES_BORROWED because there are no XDP dedicated queues.

The issue can be easily visible if the kernel is started with pci=nomsi,
then a call trace is shown. It is not shown only with sfc's modparam
interrupt_mode=2. Call trace example:
 WARNING: CPU: 2 PID: 663 at drivers/net/ethernet/sfc/efx_channels.c:828 efx_set_xdp_channels+0x124/0x260 [sfc]
 [...skip...]
 Call Trace:
  <TASK>
  efx_set_channels+0x5c/0xc0 [sfc]
  efx_probe_nic+0x9b/0x15a [sfc]
  efx_probe_all+0x10/0x1a2 [sfc]
  efx_pci_probe_main+0x12/0x156 [sfc]
  efx_pci_probe_post_io+0x18/0x103 [sfc]
  efx_pci_probe.cold+0x154/0x257 [sfc]
  local_pci_probe+0x42/0x80

Fixes: 6215b608a8c4 ("sfc: last resort fallback for lack of xdp tx queues")
Reported-by: Yanghang Liu <yanghliu@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: macsec: fix double free of percpu stats

[ Upstream commit 0c0cf3db83f8c7c9bb141c2771a34043bcf952ef ]

Inside macsec_add_dev() we free percpu macsec->secy.tx_sc.stats and
macsec->stats on some of the memory allocation failure paths. However, the
net_device is already registered to that moment: in macsec_newlink(), just
before calling macsec_add_dev(). This means that during unregister process
its priv_destructor - macsec_free_netdev() - will be called and will free
the stats again.

Remove freeing percpu stats inside macsec_add_dev() because
macsec_free_netdev() will correctly free the already allocated ones. The
pointers to unallocated stats stay NULL, and free_percpu() treats that
correctly.

Found by Linux Verification Center (linuxtesting.org) with Syzkaller.

Fixes: 0a28bfd4971f ("net/macsec: Add MACsec skb_metadata_dst Tx Data path support")
Fixes: c09440f7dcb3 ("macsec: introduce IEEE 802.1AE driver")
Signed-off-by: Fedor Pchelkin <pchelkin@ispras.ru>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: lapbether: only support ethernet devices

[ Upstream commit 9eed321cde22fc1afd76eac563ce19d899e0d6b2 ]

It probbaly makes no sense to support arbitrary network devices
for lapbether.

syzbot reported:

skbuff: skb_under_panic: text:ffff80008934c100 len:44 put:40 head:ffff0000d18dd200 data:ffff0000d18dd1ea tail:0x16 end:0x140 dev:bond1
kernel BUG at net/core/skbuff.c:200 !
Internal error: Oops - BUG: 00000000f2000800 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 5643 Comm: dhcpcd Not tainted 6.4.0-rc5-syzkaller-g4641cff8e810 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 05/25/2023
pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : skb_panic net/core/skbuff.c:196 [inline]
pc : skb_under_panic+0x13c/0x140 net/core/skbuff.c:210
lr : skb_panic net/core/skbuff.c:196 [inline]
lr : skb_under_panic+0x13c/0x140 net/core/skbuff.c:210
sp : ffff8000973b7260
x29: ffff8000973b7270 x28: ffff8000973b7360 x27: dfff800000000000
x26: ffff0000d85d8150 x25: 0000000000000016 x24: ffff0000d18dd1ea
x23: ffff0000d18dd200 x22: 000000000000002c x21: 0000000000000140
x20: 0000000000000028 x19: ffff80008934c100 x18: ffff8000973b68a0
x17: 0000000000000000 x16: ffff80008a43bfbc x15: 0000000000000202
x14: 0000000000000000 x13: 0000000000000001 x12: 0000000000000001
x11: 0000000000000201 x10: 0000000000000000 x9 : f22f7eb937cced00
x8 : f22f7eb937cced00 x7 : 0000000000000001 x6 : 0000000000000001
x5 : ffff8000973b6b78 x4 : ffff80008df9ee80 x3 : ffff8000805974f4
x2 : 0000000000000001 x1 : 0000000100000201 x0 : 0000000000000086
Call trace:
skb_panic net/core/skbuff.c:196 [inline]
skb_under_panic+0x13c/0x140 net/core/skbuff.c:210
skb_push+0xf0/0x108 net/core/skbuff.c:2409
ip6gre_header+0xbc/0x738 net/ipv6/ip6_gre.c:1383
dev_hard_header include/linux/netdevice.h:3137 [inline]
lapbeth_data_transmit+0x1c4/0x298 drivers/net/wan/lapbether.c:257
lapb_data_transmit+0x8c/0xb0 net/lapb/lapb_iface.c:447
lapb_transmit_buffer+0x178/0x204 net/lapb/lapb_out.c:149
lapb_send_control+0x220/0x320 net/lapb/lapb_subr.c:251
lapb_establish_data_link+0x94/0xec
lapb_device_event+0x348/0x4e0
notifier_call_chain+0x1a4/0x510 kernel/notifier.c:93
raw_notifier_call_chain+0x3c/0x50 kernel/notifier.c:461
__dev_notify_flags+0x2bc/0x544
dev_change_flags+0xd0/0x15c net/core/dev.c:8643
devinet_ioctl+0x858/0x17e4 net/ipv4/devinet.c:1150
inet_ioctl+0x2ac/0x4d8 net/ipv4/af_inet.c:979
sock_do_ioctl+0x134/0x2dc net/socket.c:1201
sock_ioctl+0x4ec/0x858 net/socket.c:1318
vfs_ioctl fs/ioctl.c:51 [inline]
__do_sys_ioctl fs/ioctl.c:870 [inline]
__se_sys_ioctl fs/ioctl.c:856 [inline]
__arm64_sys_ioctl+0x14c/0x1c8 fs/ioctl.c:856
__invoke_syscall arch/arm64/kernel/syscall.c:38 [inline]
invoke_syscall+0x98/0x2c0 arch/arm64/kernel/syscall.c:52
el0_svc_common+0x138/0x244 arch/arm64/kernel/syscall.c:142
do_el0_svc+0x64/0x198 arch/arm64/kernel/syscall.c:191
el0_svc+0x4c/0x160 arch/arm64/kernel/entry-common.c:647
el0t_64_sync_handler+0x84/0xfc arch/arm64/kernel/entry-common.c:665
el0t_64_sync+0x190/0x194 arch/arm64/kernel/entry.S:591
Code: aa1803e6 aa1903e7 a90023f5 947730f5 (d4210000)

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Martin Schiller <ms@dev.tdt.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: felix: fix taprio guard band overflow at 10Mbps with jumbo frames

[ Upstream commit 6ac7a27a8b07588497ed53dfd885df9c72bc67e0 ]

The DEV_MAC_MAXLEN_CFG register contains a 16-bit value - up to 65535.
Plus 2 * VLAN_HLEN (4), that is up to 65543.

The picos_per_byte variable is the largest when "speed" is lowest -
SPEED_10 = 10. In that case it is (1000000L * 8) / 10 = 800000.

Their product - 52434400000 - exceeds 32 bits, which is a problem,
because apparently, a multiplication between two 32-bit factors is
evaluated as 32-bit before being assigned to a 64-bit variable.
In fact it's a problem for any MTU value larger than 5368.

Cast one of the factors of the multiplication to u64 to force the
multiplication to take place on 64 bits.

Issue found by Coverity.

Fixes: 55a515b1f5a9 ("net: dsa: felix: drop oversized frames with tc-taprio instead of hanging the port")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230613170907.2413559-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: cls_api: Fix lockup on flushing explicitly created chain

[ Upstream commit c9a82bec02c339cdda99b37c5e62b3b71fc4209c ]

Mingshuai Ren reports:

When a new chain is added by using tc, one soft lockup alarm will be
 generated after delete the prio 0 filter of the chain. To reproduce
 the problem, perform the following steps:
(1) tc qdisc add dev eth0 root handle 1: htb default 1
(2) tc chain add dev eth0
(3) tc filter del dev eth0 chain 0 parent 1: prio 0
(4) tc filter add dev eth0 chain 0 parent 1:

Fix the issue by accounting for additional reference to chains that are
explicitly created by RTM_NEWCHAIN message as opposed to implicitly by
RTM_NEWTFILTER message.

Fixes: 726d061286ce ("net: sched: prevent insertion of new classifiers during chain flush")
Reported-by: Mingshuai Ren <renmingshuai@huawei.com>
Closes: https://lore.kernel.org/lkml/87legswvi3.fsf@nvidia.com/T/
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Link: https://lore.kernel.org/r/20230612093426.2867183-1-vladbu@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: drop the call to ext4_error() from ext4_get_group_info()

[ Upstream commit f451fd97dd2b78f286379203a47d9d295c467255 ]

A recent patch added a call to ext4_error() which is problematic since
some callers of the ext4_get_group_info() function may be holding a
spinlock, whereas ext4_error() must never be called in atomic context.

This triggered a report from Syzbot: "BUG: sleeping function called from
invalid context in ext4_update_super" (see the link below).

Therefore, drop the call to ext4_error() from ext4_get_group_info(). In
the meantime use eight characters tabs instead of nine characters ones.

Reported-by: syzbot+4acc7d910e617b360859@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/00000000000070575805fdc6cdb2@google.com/
Fixes: 5354b2af3406 ("ext4: allow ext4_get_group_info() to fail")
Suggested-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Link: https://lore.kernel.org/r/20230614100446.14337-1-fmdefrancesco@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

cifs: fix lease break oops in xfstest generic/098

[ Upstream commit c774e6779f38bf36f0cce65e30793704bab4b0d7 ]

umount can race with lease break so need to check if
tcon->ses->server is still valid to send the lease
break response.

Reviewed-by: Bharath SM <bharathsm@microsoft.com>
Reviewed-by: Shyam Prasad N <sprasad@microsoft.com>
Fixes: 59a556aebc43 ("SMB3: drop reference to cfile before sending oplock break")
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: forwarding: hw_stats_l3: Set addrgenmode in a separate step

[ Upstream commit bef68e201e538eaa3a91f97aae8161eb2d0a8ed7 ]

Setting the IPv6 address generation mode of a net device during its
creation never worked, but after commit b0ad3c179059 ("rtnetlink: call
validate_linkmsg in rtnl_create_link") it explicitly fails [1]. The
failure is caused by the fact that validate_linkmsg() is called before
the net device is registered, when it still does not have an 'inet6_dev'.

Likewise, raising the net device before setting the address generation
mode is meaningless, because by the time the mode is set, the address
has already been generated.

Therefore, fix the test to first create the net device, then set its
IPv6 address generation mode and finally bring it up.

[1]
 # ip link add name mydev addrgenmode eui64 type dummy
 RTNETLINK answers: Address family not supported by protocol

Fixes: ba95e7930957 ("selftests: forwarding: hw_stats_l3: Add a new test")
Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Link: https://lore.kernel.org/r/f3b05d85b2bc0c3d6168fe8f7207c6c8365703db.1686580046.git.petrm@nvidia.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: qdisc_destroy() old ingress and clsact Qdiscs before grafting

[ Upstream commit 84ad0af0bccd3691cb951c2974c5cb2c10594d4a ]

mini_Qdisc_pair::p_miniq is a double pointer to mini_Qdisc, initialized
in ingress_init() to point to net_device::miniq_ingress.  ingress Qdiscs
access this per-net_device pointer in mini_qdisc_pair_swap().  Similar
for clsact Qdiscs and miniq_egress.

Unfortunately, after introducing RTNL-unlocked RTM_{NEW,DEL,GET}TFILTER
requests (thanks Hillf Danton for the hint), when replacing ingress or
clsact Qdiscs, for example, the old Qdisc ("@old") could access the same
miniq_{in,e}gress pointer(s) concurrently with the new Qdisc ("@new"),
causing race conditions [1] including a use-after-free bug in
mini_qdisc_pair_swap() reported by syzbot:

 BUG: KASAN: slab-use-after-free in mini_qdisc_pair_swap+0x1c2/0x1f0 net/sched/sch_generic.c:1573
 Write of size 8 at addr ffff888045b31308 by task syz-executor690/14901
...
 Call Trace:
  <TASK>
  __dump_stack lib/dump_stack.c:88 [inline]
  dump_stack_lvl+0xd9/0x150 lib/dump_stack.c:106
  print_address_description.constprop.0+0x2c/0x3c0 mm/kasan/report.c:319
  print_report mm/kasan/report.c:430 [inline]
  kasan_report+0x11c/0x130 mm/kasan/report.c:536
  mini_qdisc_pair_swap+0x1c2/0x1f0 net/sched/sch_generic.c:1573
  tcf_chain_head_change_item net/sched/cls_api.c:495 [inline]
  tcf_chain0_head_change.isra.0+0xb9/0x120 net/sched/cls_api.c:509
  tcf_chain_tp_insert net/sched/cls_api.c:1826 [inline]
  tcf_chain_tp_insert_unique net/sched/cls_api.c:1875 [inline]
  tc_new_tfilter+0x1de6/0x2290 net/sched/cls_api.c:2266
...

@old and @new should not affect each other.  In other words, @old should
never modify miniq_{in,e}gress after @new, and @new should not update
@old's RCU state.

Fixing without changing sch_api.c turned out to be difficult (please
refer to Closes: for discussions).  Instead, make sure @new's first call
always happen after @old's last call (in {ingress,clsact}_destroy()) has
finished:

In qdisc_graft(), return -EBUSY if @old has any ongoing filter requests,
and call qdisc_destroy() for @old before grafting @new.

Introduce qdisc_refcount_dec_if_one() as the counterpart of
qdisc_refcount_inc_nz() used for filter requests.  Introduce a
non-static version of qdisc_destroy() that does a TCQ_F_BUILTIN check,
just like qdisc_put() etc.

Depends on patch "net/sched: Refactor qdisc_graft() for ingress and
clsact Qdiscs".

[1] To illustrate, the syzkaller reproducer adds ingress Qdiscs under
TC_H_ROOT (no longer possible after commit c7cfbd115001 ("net/sched:
sch_ingress: Only create under TC_H_INGRESS")) on eth0 that has 8
transmission queues:

  Thread 1 creates ingress Qdisc A (containing mini Qdisc a1 and a2),
  then adds a flower filter X to A.

  Thread 2 creates another ingress Qdisc B (containing mini Qdisc b1 and
  b2) to replace A, then adds a flower filter Y to B.

 Thread 1               A's refcnt   Thread 2
  RTM_NEWQDISC (A, RTNL-locked)
   qdisc_create(A)               1
   qdisc_graft(A)                9

  RTM_NEWTFILTER (X, RTNL-unlocked)
   __tcf_qdisc_find(A)          10
   tcf_chain0_head_change(A)
   mini_qdisc_pair_swap(A) (1st)
            |
            |                         RTM_NEWQDISC (B, RTNL-locked)
         RCU sync                2     qdisc_graft(B)
            |                    1     notify_and_destroy(A)
            |
   tcf_block_release(A)          0    RTM_NEWTFILTER (Y, RTNL-unlocked)
   qdisc_destroy(A)                    tcf_chain0_head_change(B)
   tcf_chain0_head_change_cb_del(A)    mini_qdisc_pair_swap(B) (2nd)
   mini_qdisc_pair_swap(A) (3rd)                |
           ...                                 ...

Here, B calls mini_qdisc_pair_swap(), pointing eth0->miniq_ingress to
its mini Qdisc, b1.  Then, A calls mini_qdisc_pair_swap() again during
ingress_destroy(), setting eth0->miniq_ingress to NULL, so ingress
packets on eth0 will not find filter Y in sch_handle_ingress().

This is just one of the possible consequences of concurrently accessing
miniq_{in,e}gress pointers.

Fixes: 7a096d579e8e ("net: sched: ingress: set 'unlocked' flag for Qdisc ops")
Fixes: 87f373921c4e ("net: sched: ingress: set 'unlocked' flag for clsact Qdisc ops")
Reported-by: syzbot+b53a9c0d1ea4ad62da8b@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/r/0000000000006cf87705f79acf1a@google.com/
Cc: Hillf Danton <hdanton@sina.com>
Cc: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: Refactor qdisc_graft() for ingress and clsact Qdiscs

[ Upstream commit 2d5f6a8d7aef7852a9ecc555f88c673a1c91754f ]

Grafting ingress and clsact Qdiscs does not need a for-loop in
qdisc_graft().  Refactor it.  No functional changes intended.

Tested-by: Pedro Tammela <pctammela@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 84ad0af0bccd ("net/sched: qdisc_destroy() old ingress and clsact Qdiscs before grafting")
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched: add new attr TCA_EXT_WARN_MSG to report tc extact message

[ Upstream commit 0349b8779cc949ad9e6aced32672ee48cf79b497 ]

We will report extack message if there is an error via netlink_ack(). But
if the rule is not to be exclusively executed by the hardware, extack is not
passed along and offloading failures don't get logged.

In commit 81c7288b170a ("sched: cls: enable verbose logging") Marcelo
made cls could log verbose info for offloading failures, which helps
improving Open vSwitch debuggability when using flower offloading.

It would also be helpful if userspace monitor tools, like "tc monitor",
could log this kind of message, as it doesn't require vswitchd log level
adjusment. Let's add a new tc attributes to report the extack message so
the monitor program could receive the failures. e.g.

  # tc monitor
  added chain dev enp3s0f1np1 parent ffff: chain 0
  added filter dev enp3s0f1np1 ingress protocol all pref 49152 flower chain 0 handle 0x1
    ct_state +trk+new
    not_in_hw
          action order 1: gact action drop
           random type none pass val 0
           index 1 ref 1 bind 1

  Warning: mlx5_core: matching on ct_state +new isn't supported.

In this patch I only report the extack message on add/del operations.
It doesn't look like we need to report the extack message on get/dump
operations.

Note this message not only reporte to multicast groups, it could also
be reported unicast, which may affect the current usersapce tool's behaivor.

Suggested-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20230113034353.2766735-1-liuhangbin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 84ad0af0bccd ("net/sched: qdisc_destroy() old ingress and clsact Qdiscs before grafting")
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/tc-testing: Fix SFB db test

[ Upstream commit b39d8c41c7a8336ce85c376b5d4906089524a0ae ]

Setting very small value of db like 10ms introduces rounding errors when
converting to/from jiffies on some kernel configs. For example, on 250hz
the actual value will be set to 12ms which causes the test to fail:

 # $ sudo ./tdc.py  -d eth2 -e 3410
 #  -- ns/SubPlugin.__init__
 # Test 3410: Create SFB with db setting
 #
 # All test results:
 #
 # 1..1
 # not ok 1 3410 - Create SFB with db setting
 #         Could not match regex pattern. Verify command output:
 # qdisc sfb 1: root refcnt 2 rehash 600s db 12ms limit 1000p max 25p target 20p increment 0.000503548 decrement 4.57771e-05 penalty_rate 10pps penalty_burst 20p

Set the value to 100ms instead which currently seem to work on 100hz,
250hz, 300hz and 1000hz kernel configs.

Fixes: 6ad92dc56fca ("selftests/tc-testing: add selftests for sfb qdisc")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Pedro Tammela <pctammela@mojatatu.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/tc-testing: Fix Error: failed to find target LOG

[ Upstream commit b849c566ee9c6ed78288a522278dcaf419f8e239 ]

Add missing netfilter config dependency.

Fixes following example error when running tests via tdc.sh for all XT
tests:

 # $ sudo ./tdc.py -d eth2 -e 2029
 # Test 2029: Add xt action with log-prefix
 # exit: 255
 # exit: 0
 #  failed to find target LOG
 #
 # bad action parsing
 # parse_action: bad value (7:xt)!
 # Illegal "action"
 #
 # -----> teardown stage *** Could not execute: "$TC actions flush action xt"
 #
 # -----> teardown stage *** Error message: "Error: Cannot flush unknown TC action.
 # We have an error flushing
 # "
 # returncode 1; expected [0]
 #
 # -----> teardown stage *** Aborting test run.
 #
 # <_io.BufferedReader name=3> *** stdout ***
 #
 # <_io.BufferedReader name=5> *** stderr ***
 # "-----> teardown stage" did not complete successfully
 # Exception <class '__main__.PluginMgrTestFail'> ('teardown', ' failed to find target LOG\n\nbad action parsing\nparse_action: bad value (7:xt)!\nIllegal "action"\n', '"-----> teardown stage" did not complete successfully') (caught in test_runner, running test 2 2029 Add xt action with log-prefix stage teardown)
 # ---------------
 # traceback
 #   File "/images/src/linux/tools/testing/selftests/tc-testing/./tdc.py", line 495, in test_runner
 #     res = run_one_test(pm, args, index, tidx)
 #   File "/images/src/linux/tools/testing/selftests/tc-testing/./tdc.py", line 434, in run_one_test
 #     prepare_env(args, pm, 'teardown', '-----> teardown stage', tidx['teardown'], procout)
 #   File "/images/src/linux/tools/testing/selftests/tc-testing/./tdc.py", line 245, in prepare_env
 #     raise PluginMgrTestFail(
 # ---------------
 # accumulated output for this test:
 #  failed to find target LOG
 #
 # bad action parsing
 # parse_action: bad value (7:xt)!
 # Illegal "action"
 #
 # ---------------
 #
 # All test results:
 #
 # 1..1
 # ok 1 2029 - Add xt action with log-prefix # skipped - "-----> teardown stage" did not complete successfully

Fixes: 910d504bc187 ("selftests/tc-testings: add selftests for xt action")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Pedro Tammela <pctammela@mojatatu.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/tc-testing: Fix Error: Specified qdisc kind is unknown.

[ Upstream commit aef6e908b54200d04f2d77dab31509fcff2e60ae ]

All TEQL tests assume that sch_teql module is loaded. Load module in tdc.sh
before running qdisc tests.

Fixes following example error when running tests via tdc.sh for all TEQL
tests:

 # $ sudo ./tdc.py -d eth2 -e 84a0
 #  -- ns/SubPlugin.__init__
 # Test 84a0: Create TEQL with default setting
 # exit: 2
 # exit: 0
 # Error: Specified qdisc kind is unknown.
 #
 # -----> teardown stage *** Could not execute: "$TC qdisc del dev $DUMMY handle 1: root"
 #
 # -----> teardown stage *** Error message: "Error: Invalid handle.
 # "
 # returncode 2; expected [0]
 #
 # -----> teardown stage *** Aborting test run.
 #
 # <_io.BufferedReader name=3> *** stdout ***
 #
 # <_io.BufferedReader name=5> *** stderr ***
 # "-----> teardown stage" did not complete successfully
 # Exception <class '__main__.PluginMgrTestFail'> ('teardown', 'Error: Specified qdisc kind is unknown.\n', '"-----> teardown stage" did not complete successfully') (caught in test_runner, running test 2 84a0 Create TEQL with default setting stage teardown)
 # ---------------
 # traceback
 #   File "/images/src/linux/tools/testing/selftests/tc-testing/./tdc.py", line 495, in test_runner
 #     res = run_one_test(pm, args, index, tidx)
 #   File "/images/src/linux/tools/testing/selftests/tc-testing/./tdc.py", line 434, in run_one_test
 #     prepare_env(args, pm, 'teardown', '-----> teardown stage', tidx['teardown'], procout)
 #   File "/images/src/linux/tools/testing/selftests/tc-testing/./tdc.py", line 245, in prepare_env
 #     raise PluginMgrTestFail(
 # ---------------
 # accumulated output for this test:
 # Error: Specified qdisc kind is unknown.
 #
 # ---------------
 #
 # All test results:
 #
 # 1..1
 # ok 1 84a0 - Create TEQL with default setting # skipped - "-----> teardown stage" did not complete successfully

Fixes: cc62fbe114c9 ("selftests/tc-testing: add selftests for teql qdisc")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Victor Nogueira <victor@mojatatu.com>
Reviewed-by: Pedro Tammela <pctammela@mojatatu.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/nouveau: add nv_encoder pointer check for NULL

[ Upstream commit 55b94bb8c42464bad3d2217f6874aa1a85664eac ]

Pointer nv_encoder could be dereferenced at nouveau_connector.c
in case it's equal to NULL by jumping to goto label.
This patch adds a NULL-check to avoid it.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: 3195c5f9784a ("drm/nouveau: set encoder for lvds")
Signed-off-by: Natalia Petrova <n.petrova@fintech.ru>
Reviewed-by: Lyude Paul <lyude@redhat.com>
[Fixed patch title]
Signed-off-by: Lyude Paul <lyude@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230512103320.82234-1-n.petrova@fintech.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/nouveau/dp: check for NULL nv_connector->native_mode

[ Upstream commit 20a2ce87fbaf81e4c3dcb631d738e423959eb320 ]

Add checking for NULL before calling nouveau_connector_detect_depth() in
nouveau_connector_get_modes() function because nv_connector->native_mode
could be dereferenced there since connector pointer passed to
nouveau_connector_detect_depth() and the same value of
nv_connector->native_mode is used there.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: d4c2c99bdc83 ("drm/nouveau/dp: remove broken display depth function, use the improved one")

Signed-off-by: Natalia Petrova <n.petrova@fintech.ru>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Lyude Paul <lyude@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230512111526.82408-1-n.petrova@fintech.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/bridge: ti-sn65dsi86: Avoid possible buffer overflow

[ Upstream commit 95011f267c44a4d1f9ca1769e8a29ab2c559e004 ]

Smatch error:buffer overflow 'ti_sn_bridge_refclk_lut' 5 <= 5.

Fixes: cea86c5bb442 ("drm/bridge: ti-sn65dsi86: Implement the pwm_chip")
Signed-off-by: Su Hui <suhui@nfschina.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230608012443.839372-1-suhui@nfschina.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/nouveau: don't detect DSM for non-NVIDIA device

[ Upstream commit 11d24327c2d7ad7f24fcc44fb00e1fa91ebf6525 ]

The call site of nouveau_dsm_pci_probe() uses single set of output
variables for all invocations. So, we must not write anything to them
unless it's an NVIDIA device. Otherwise, if we are called with another
device after the NVIDIA device, we'll clober the result of the NVIDIA
device.

For example, if the other device doesn't have _PR3 resources, the
detection later would miss the presence of power resource support, and
the rest of the code will keep using Optimus DSM, breaking power
management for that machine.

Also, because we're detecting NVIDIA's DSM, it doesn't make sense to run
this detection on a non-NVIDIA device anyway. Thus, check at the
beginning of the detection code if this is an NVIDIA card, and just
return if it isn't.

This, together with commit d22915d22ded ("drm/nouveau/devinit/tu102-:
wait for GFW_BOOT_PROGRESS == COMPLETED") developed independently and
landed earlier, fixes runtime power management of the NVIDIA card in
Lenovo Legion 5-15ARH05. Without this patch, the GPU resumption code
will "timeout", sometimes hanging userspace.

As a bonus, we'll also stop preventing _PR3 usage from the bridge for
unrelated devices, which is always nice, I guess.

Fixes: ccfc2d5cdb02 ("drm/nouveau: Use generic helper to check _PR3 presence")
Signed-off-by: Ratchanan Srirattanamet <peathot@hotmail.com>
Closes: https://gitlab.freedesktop.org/drm/nouveau/-/issues/79
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/DM6PR19MB2780805D4BE1E3F9B3AC96D0BC409@DM6PR19MB2780.namprd19.prod.outlook.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: phylink: use a dedicated helper to parse usgmii control word

[ Upstream commit 923454c0368b8092e9d05c020f50abca577e7290 ]

Q-USGMII is a derivative of USGMII, that uses a specific formatting for
the control word. The layout is close to the USXGMII control word, but
doesn't support speeds over 1Gbps. Use a dedicated decoding logic for
the USGMII control word, re-using USXGMII definitions but only considering
10/100/1000Mbps speeds

Fixes: 5e61fe157a27 ("net: phy: Introduce QUSGMII PHY mode")
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: phylink: report correct max speed for QUSGMII

[ Upstream commit b9dc1046edfeb7d9dbc2272c8d9ad5a8c47f3199 ]

Q-USGMII is the quad port version of USGMII, and supports a max speed of
1Gbps on each line. Make so that phylink_interface_max_speed() reports
this information correctly.

Fixes: ae0e4bb2a0e0 ("net: phylink: Adjust link settings based on rate matching")
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

igb: fix nvm.ops.read() error handling

[ Upstream commit 48a821fd58837800750ec1b3962f0f799630a844 ]

Add error handling into igb_set_eeprom() function, in case
nvm.ops.read() fails just quit with error code asap.

Fixes: 9d5c824399de ("igb: PCI-Express 82575 Gigabit Ethernet driver")
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

igc: Fix possible system crash when loading module

[ Upstream commit c080fe262f9e73a00934b70c16b1479cf40cd2bd ]

Guarantee that when probe() is run again, PTM and PCI busmaster will be
in the same state as it was if the driver was never loaded.

Avoid an i225/i226 hardware issue that PTM requests can be made even
though PCI bus mastering is not enabled. These unexpected PTM requests
can crash some systems.

So, "force" disable PTM and busmastering before removing the driver,
so they can be re-enabled in the right order during probe(). This is
more like a workaround and should be applicable for i225 and i226, in
any platform.

Fixes: 1b5d73fb8624 ("igc: Enable PCIe PTM")
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Reviewed-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

igc: Clean the TX buffer and TX descriptor ring

[ Upstream commit e43516f5978d11d36511ce63d31d1da4db916510 ]

There could be a race condition during link down where interrupt
being generated and igc_clean_tx_irq() been called to perform the
TX completion. Properly clear the TX buffer/descriptor ring and
disable the TX Queue ring in igc_free_tx_resources() to avoid that.

Kernel trace:
[  108.237177] Hardware name: Intel Corporation Tiger Lake Client Platform/TigerLake U DDR4 SODIMM RVP, BIOS TGLIFUI1.R00.4204.A00.2105270302 05/27/2021
[  108.237178] RIP: 0010:refcount_warn_saturate+0x55/0x110
[  108.242143] RSP: 0018:ffff9e7980003db0 EFLAGS: 00010286
[  108.245555] Code: 84 bc 00 00 00 c3 cc cc cc cc 85 f6 74 46 80 3d 20 8c 4d 01 00 75 ee 48 c7 c7 88 f4 03 ab c6 05 10 8c 4d 01 01 e8 0b 10 96 ff <0f> 0b c3 cc cc cc cc 80 3d fc 8b 4d 01 00 75 cb 48 c7 c7 b0 f4 03
[  108.250434]
[  108.250434] RSP: 0018:ffff9e798125f910 EFLAGS: 00010286
[  108.254358] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
[  108.259325]
[  108.259325] RAX: 0000000000000000 RBX: ffff8ddb935b8000 RCX: 0000000000000027
[  108.261868] RDX: ffff8de250a28800 RSI: ffff8de250a1c580 RDI: ffff8de250a1c580
[  108.265538] RDX: 0000000000000027 RSI: 0000000000000002 RDI: ffff8de250a9c588
[  108.265539] RBP: ffff8ddb935b8000 R08: ffffffffab2655a0 R09: ffff9e798125f898
[  108.267914] RBP: ffff8ddb8a5b8d80 R08: 0000005648eba354 R09: 0000000000000000
[  108.270196] R10: 0000000000000001 R11: 000000002d2d2d2d R12: ffff9e798125f948
[  108.270197] R13: ffff9e798125fa1c R14: ffff8ddb8a5b8d80 R15: 7fffffffffffffff
[  108.273001] R10: 000000002d2d2d2d R11: 000000002d2d2d2d R12: ffff8ddb8a5b8ed4
[  108.276410] FS:  00007f605851b740(0000) GS:ffff8de250a80000(0000) knlGS:0000000000000000
[  108.280597] R13: 00000000000002ac R14: 00000000ffffff99 R15: ffff8ddb92561b80
[  108.282966] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  108.282967] CR2: 00007f053c039248 CR3: 0000000185850003 CR4: 0000000000f70ee0
[  108.286206] FS:  0000000000000000(0000) GS:ffff8de250a00000(0000) knlGS:0000000000000000
[  108.289701] PKRU: 55555554
[  108.289702] Call Trace:
[  108.289704]  <TASK>
[  108.293977] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  108.297562]  sock_alloc_send_pskb+0x20c/0x240
[  108.301494] CR2: 00007f053c03a168 CR3: 0000000184394002 CR4: 0000000000f70ef0
[  108.301495] PKRU: 55555554
[  108.306464]  __ip_append_data.isra.0+0x96f/0x1040
[  108.309441] Call Trace:
[  108.309443]  ? __pfx_ip_generic_getfrag+0x10/0x10
[  108.314927]  <IRQ>
[  108.314928]  sock_wfree+0x1c7/0x1d0
[  108.318078]  ? __pfx_ip_generic_getfrag+0x10/0x10
[  108.320276]  skb_release_head_state+0x32/0x90
[  108.324812]  ip_make_skb+0xf6/0x130
[  108.327188]  skb_release_all+0x16/0x40
[  108.330775]  ? udp_sendmsg+0x9f3/0xcb0
[  108.332626]  napi_consume_skb+0x48/0xf0
[  108.334134]  ? xfrm_lookup_route+0x23/0xb0
[  108.344285]  igc_poll+0x787/0x1620 [igc]
[  108.346659]  udp_sendmsg+0x9f3/0xcb0
[  108.360010]  ? ttwu_do_activate+0x40/0x220
[  108.365237]  ? __pfx_ip_generic_getfrag+0x10/0x10
[  108.366744]  ? try_to_wake_up+0x289/0x5e0
[  108.376987]  ? sock_sendmsg+0x81/0x90
[  108.395698]  ? __pfx_process_timeout+0x10/0x10
[  108.395701]  sock_sendmsg+0x81/0x90
[  108.409052]  __napi_poll+0x29/0x1c0
[  108.414279]  ____sys_sendmsg+0x284/0x310
[  108.419507]  net_rx_action+0x257/0x2d0
[  108.438216]  ___sys_sendmsg+0x7c/0xc0
[  108.439723]  __do_softirq+0xc1/0x2a8
[  108.444950]  ? finish_task_switch+0xb4/0x2f0
[  108.452077]  irq_exit_rcu+0xa9/0xd0
[  108.453584]  ? __schedule+0x372/0xd00
[  108.460713]  common_interrupt+0x84/0xa0
[  108.467840]  ? clockevents_program_event+0x95/0x100
[  108.474968]  </IRQ>
[  108.482096]  ? do_nanosleep+0x88/0x130
[  108.489224]  <TASK>
[  108.489225]  asm_common_interrupt+0x26/0x40
[  108.496353]  ? __rseq_handle_notify_resume+0xa9/0x4f0
[  108.503478] RIP: 0010:cpu_idle_poll+0x2c/0x100
[  108.510607]  __sys_sendmsg+0x5d/0xb0
[  108.518687] Code: 05 e1 d9 c8 00 65 8b 15 de 64 85 55 85 c0 7f 57 e8 b9 ef ff ff fb 65 48 8b 1c 25 00 cc 02 00 48 8b 03 a8 08 74 0b eb 1c f3 90 <48> 8b 03 a8 08 75 13 8b 05 77 63 cd 00 85 c0 75 ed e8 ce ec ff ff
[  108.525817]  do_syscall_64+0x44/0xa0
[  108.531563] RSP: 0018:ffffffffab203e70 EFLAGS: 00000202
[  108.538693]  entry_SYSCALL_64_after_hwframe+0x72/0xdc
[  108.546775]
[  108.546777] RIP: 0033:0x7f605862b7f7
[  108.549495] RAX: 0000000000000001 RBX: ffffffffab20c940 RCX: 000000000000003b
[  108.551955] Code: 0e 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b9 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 2e 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 89 54 24 1c 48 89 74 24 10
[  108.554068] RDX: 4000000000000000 RSI: 000000002da97f6a RDI: 00000000002b8ff4
[  108.559816] RSP: 002b:00007ffc99264058 EFLAGS: 00000246
[  108.564178] RBP: 0000000000000000 R08: 00000000002b8ff4 R09: ffff8ddb01554c80
[  108.571302]  ORIG_RAX: 000000000000002e
[  108.571303] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f605862b7f7
[  108.574023] R10: 000000000000015b R11: 000000000000000f R12: ffffffffab20c940
[  108.574024] R13: 0000000000000000 R14: ffff8de26fbeef40 R15: ffffffffab20c940
[  108.578727] RDX: 0000000000000000 RSI: 00007ffc992640a0 RDI: 0000000000000003
[  108.578728] RBP: 00007ffc99264110 R08: 0000000000000000 R09: 175f48ad1c3a9c00
[  108.581187]  do_idle+0x62/0x230
[  108.585890] R10: 0000000000000000 R11: 0000000000000246 R12: 00007ffc992642d8
[  108.585891] R13: 00005577814ab2ba R14: 00005577814addf0 R15: 00007f605876d000
[  108.587920]  cpu_startup_entry+0x1d/0x20
[  108.591422]  </TASK>
[  108.596127]  rest_init+0xc5/0xd0
[  108.600490] ---[ end trace 0000000000000000 ]---

Test Setup:

DUT:
- Change mac address on DUT Side. Ensure NIC not having same MAC Address
- Running udp_tai on DUT side. Let udp_tai running throughout the test

Example:
./udp_tai -i enp170s0 -P 100000 -p 90 -c 1 -t 0 -u 30004

Host:
- Perform link up/down every 5 second.

Result:
Kernel panic will happen on DUT Side.

Fixes: 13b5b7fd6a4a ("igc: Add support for Tx/Rx rings")
Signed-off-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sctp: fix an error code in sctp_sf_eat_auth()

[ Upstream commit 75e6def3b26736e7ff80639810098c9074229737 ]

The sctp_sf_eat_auth() function is supposed to enum sctp_disposition
values and returning a kernel error code will cause issues in the
caller.  Change -ENOMEM to SCTP_DISPOSITION_NOMEM.

Fixes: 65b07e5d0d09 ("[SCTP]: API updates to suport SCTP-AUTH extensions.")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Acked-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ipvlan: fix bound dev checking for IPv6 l3s mode

[ Upstream commit ce57adc222aba32431c42632b396e9213d0eb0b8 ]

The commit 59a0b022aa24 ("ipvlan: Make skb->skb_iif track skb->dev for l3s
mode") fixed ipvlan bonded dev checking by updating skb skb_iif. This fix
works for IPv4, as in raw_v4_input() the dif is from inet_iif(skb), which
is skb->skb_iif when there is no route.

But for IPv6, the fix is not enough, because in ipv6_raw_deliver() ->
raw_v6_match(), the dif is inet6_iif(skb), which is returns IP6CB(skb)->iif
instead of skb->skb_iif if it's not a l3_slave. To fix the IPv6 part
issue. Let's set IP6CB(skb)->iif to correct ifindex.

BTW, ipvlan handles NS/NA specifically. Since it works fine, I will not
reset IP6CB(skb)->iif when addr->atype is IPVL_ICMPV6.

Fixes: c675e06a98a4 ("ipvlan: decouple l3s mode dependencies from other modes")
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2196710
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: Larysa Zaremba <larysa.zaremba@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ethtool: correct MAX attribute value for stats

[ Upstream commit 52f79609c0c5b25fddb88e85f25ce08aa7e3fb42 ]

When compiling YNL generated code compiler complains about
array-initializer-out-of-bounds. Turns out the MAX value
for STATS_GRP uses the value for STATS.

This may lead to random corruptions in user space (kernel
itself doesn't use this value as it never parses stats).

Fixes: f09ea6fb1272 ("ethtool: add a new command for reading standard stats")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

IB/isert: Fix incorrect release of isert connection

[ Upstream commit 699826f4e30ab76a62c238c86fbef7e826639c8d ]

The ib_isert module is releasing the isert connection both in
isert_wait_conn() handler as well as isert_free_conn() handler.
In isert_wait_conn() handler, it is expected to wait for iSCSI
session logout operation to complete. It should free the isert
connection only in isert_free_conn() handler.

When a bunch of iSER target is cleared, this issue can lead to
use-after-free memory issue as isert conn is twice released

Fixes: b02efbfc9a05 ("iser-target: Fix implicit termination of connections")
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Saravanan Vajravel <saravanan.vajravel@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/20230606102531.162967-4-saravanan.vajravel@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

IB/isert: Fix possible list corruption in CMA handler

[ Upstream commit 7651e2d6c5b359a28c2d4c904fec6608d1021ca8 ]

When ib_isert module receives connection error event, it is
releasing the isert session and removes corresponding list
node but it doesn't take appropriate mutex lock to remove
the list node.  This can lead to linked  list corruption

Fixes: bd3792205aae ("iser-target: Fix pending connections handling in target stack shutdown sequnce")
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Saravanan Vajravel <saravanan.vajravel@broadcom.com>
Link: https://lore.kernel.org/r/20230606102531.162967-3-saravanan.vajravel@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

IB/isert: Fix dead lock in ib_isert

[ Upstream commit 691b0480933f0ce88a81ed1d1a0aff340ff6293a ]

- When a iSER session is released, ib_isert module is taking a mutex
  lock and releasing all pending connections. As part of this, ib_isert
  is destroying rdma cm_id. To destroy cm_id, rdma_cm module is sending
  CM events to CMA handler of ib_isert. This handler is taking same
  mutex lock. Hence it leads to deadlock between ib_isert & rdma_cm
  modules.

- For fix, created local list of pending connections and release the
  connection outside of mutex lock.

Calltrace:
---------
[ 1229.791410] INFO: task kworker/10:1:642 blocked for more than 120 seconds.
[ 1229.791416]       Tainted: G           OE    --------- -  - 4.18.0-372.9.1.el8.x86_64 #1
[ 1229.791418] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1229.791419] task:kworker/10:1    state:D stack:    0 pid:  642 ppid:     2 flags:0x80004000
[ 1229.791424] Workqueue: ib_cm cm_work_handler [ib_cm]
[ 1229.791436] Call Trace:
[ 1229.791438]  __schedule+0x2d1/0x830
[ 1229.791445]  ? select_idle_sibling+0x23/0x6f0
[ 1229.791449]  schedule+0x35/0xa0
[ 1229.791451]  schedule_preempt_disabled+0xa/0x10
[ 1229.791453]  __mutex_lock.isra.7+0x310/0x420
[ 1229.791456]  ? select_task_rq_fair+0x351/0x990
[ 1229.791459]  isert_cma_handler+0x224/0x330 [ib_isert]
[ 1229.791463]  ? ttwu_queue_wakelist+0x159/0x170
[ 1229.791466]  cma_cm_event_handler+0x25/0xd0 [rdma_cm]
[ 1229.791474]  cma_ib_handler+0xa7/0x2e0 [rdma_cm]
[ 1229.791478]  cm_process_work+0x22/0xf0 [ib_cm]
[ 1229.791483]  cm_work_handler+0xf4/0xf30 [ib_cm]
[ 1229.791487]  ? move_linked_works+0x6e/0xa0
[ 1229.791490]  process_one_work+0x1a7/0x360
[ 1229.791491]  ? create_worker+0x1a0/0x1a0
[ 1229.791493]  worker_thread+0x30/0x390
[ 1229.791494]  ? create_worker+0x1a0/0x1a0
[ 1229.791495]  kthread+0x10a/0x120
[ 1229.791497]  ? set_kthread_struct+0x40/0x40
[ 1229.791499]  ret_from_fork+0x1f/0x40

[ 1229.791739] INFO: task targetcli:28666 blocked for more than 120 seconds.
[ 1229.791740]       Tainted: G           OE    --------- -  - 4.18.0-372.9.1.el8.x86_64 #1
[ 1229.791741] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1229.791742] task:targetcli       state:D stack:    0 pid:28666 ppid:  5510 flags:0x00004080
[ 1229.791743] Call Trace:
[ 1229.791744]  __schedule+0x2d1/0x830
[ 1229.791746]  schedule+0x35/0xa0
[ 1229.791748]  schedule_preempt_disabled+0xa/0x10
[ 1229.791749]  __mutex_lock.isra.7+0x310/0x420
[ 1229.791751]  rdma_destroy_id+0x15/0x20 [rdma_cm]
[ 1229.791755]  isert_connect_release+0x115/0x130 [ib_isert]
[ 1229.791757]  isert_free_np+0x87/0x140 [ib_isert]
[ 1229.791761]  iscsit_del_np+0x74/0x120 [iscsi_target_mod]
[ 1229.791776]  lio_target_np_driver_store+0xe9/0x140 [iscsi_target_mod]
[ 1229.791784]  configfs_write_file+0xb2/0x110
[ 1229.791788]  vfs_write+0xa5/0x1a0
[ 1229.791792]  ksys_write+0x4f/0xb0
[ 1229.791794]  do_syscall_64+0x5b/0x1a0
[ 1229.791798]  entry_SYSCALL_64_after_hwframe+0x65/0xca

Fixes: bd3792205aae ("iser-target: Fix pending connections handling in target stack shutdown sequnce")
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Saravanan Vajravel <saravanan.vajravel@broadcom.com>
Link: https://lore.kernel.org/r/20230606102531.162967-2-saravanan.vajravel@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/mlx5: Fix affinity assignment

[ Upstream commit 617f5db1a626f18d5cbb7c7faf7bf8f9ea12be78 ]

The cited commit aimed to ensure that Virtual Functions (VFs) assign a
queue affinity to a Queue Pair (QP) to distribute traffic when
the LAG master creates a hardware LAG. If the affinity was set while
the hardware was not in LAG, the firmware would ignore the affinity value.

However, this commit unintentionally assigned an affinity to QPs on the LAG
master's VPORT even if the RDMA device was not marked as LAG-enabled.
In most cases, this was not an issue because when the hardware entered
hardware LAG configuration, the RDMA device of the LAG master would be
destroyed and a new one would be created, marked as LAG-enabled.

The problem arises when a user configures Equal-Cost Multipath (ECMP).
In ECMP mode, traffic can be directed to different physical ports based on
the queue affinity, which is intended for use by VPORTS other than the
E-Switch manager. ECMP mode is supported only if both E-Switch managers are
in switchdev mode and the appropriate route is configured via IP. In this
configuration, the RDMA device is not destroyed, and we retain the RDMA
device that is not marked as LAG-enabled.

To ensure correct behavior, Send Queues (SQs) opened by the E-Switch
manager through verbs should be assigned strict affinity. This means they
will only be able to communicate through the native physical port
associated with the E-Switch manager. This will prevent the firmware from
assigning affinity and will not allow the SQs to be remapped in case of
failover.

Fixes: 802dcc7fc5ec ("RDMA/mlx5: Support TX port affinity for VF drivers in LAG mode")
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://lore.kernel.org/r/425b05f4da840bc684b0f7e8ebf61aeb5cef09b0.1685960567.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

IB/uverbs: Fix to consider event queue closing also upon non-blocking mode

[ Upstream commit 62fab312fa1683e812e605db20d4f22de3e3fb2f ]

Fix ib_uverbs_event_read() to consider event queue closing also upon
non-blocking mode.

Once the queue is closed (e.g. hot-plug flow) all the existing events
are cleaned-up as part of ib_uverbs_free_event_queue().

An application that uses the non-blocking FD mode should get -EIO in
that case to let it knows that the device was removed already.

Otherwise, it can loose the indication that the device was removed and
won't recover.

As part of that, refactor the code to have a single flow with regards to
'is_closed' for both blocking and non-blocking modes.

Fixes: 14e23bd6d221 ("RDMA/core: Fix locking in ib_uverbs_event_read")
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Link: https://lore.kernel.org/r/97b00116a1e1e13f8dc4ec38a5ea81cf8c030210.1685960567.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/cma: Always set static rate to 0 for RoCE

[ Upstream commit 58030c76cce473b6cfd630bbecb97215def0dff8 ]

Set static rate to 0 as it should be discovered by path query and
has no meaning for RoCE.
This also avoid of using the rtnl lock and ethtool API, which is
a bottleneck when try to setup many rdma-cm connections at the same
time, especially with multiple processes.

Fixes: 3c86aa70bf67 ("RDMA/cm: Add RDMA CM support for IBoE devices")
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Link: https://lore.kernel.org/r/f72a4f8b667b803aee9fa794069f61afb5839ce4.1685960567.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/mlx5: Create an indirect flow table for steering anchor

[ Upstream commit e1f4a52ac171dd863fe89055e749ef5e0a0bc5ce ]

A misbehaved user can create a steering anchor that points to a kernel
flow table and then destroy the anchor without freeing the associated
STC. This creates a problem as the kernel can't destroy the flow
table since there is still a reference to it. As a result, this can
exhaust all available flow table resources, preventing other users from
using the RDMA device.

To prevent this problem, a solution is implemented where a special flow
table with two steering rules is created when a user creates a steering
anchor for the first time. The rules include one that drops all traffic
and another that points to the kernel flow table. If the steering anchor
is destroyed, only the rule pointing to the kernel's flow table is removed.
Any traffic reaching the special flow table after that is dropped.

Since the special flow table is not destroyed when the steering anchor is
destroyed, any issues are prevented from occurring. The remaining resources
are only destroyed when the RDMA device is destroyed, which happens after
all DEVX objects are freed, including the STCs, thus mitigating the issue.

Fixes: 0c6ab0ca9a66 ("RDMA/mlx5: Expose steering anchor to userspace")
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Link: https://lore.kernel.org/r/b4a88a871d651fa4e8f98d552553c1cfe9ba2cd6.1685960567.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/mlx5: Initiate dropless RQ for RAW Ethernet functions

[ Upstream commit ee4d269eccfea6c17b18281bef482700d898e86f ]

Delay drop data is initiated for PFs that have the capability of
rq_delay_drop and are in roce profile.

However, PFs with RAW ethernet profile do not initiate delay drop data
on function load, causing kernel panic if delay drop struct members are
accessed later on in case a dropless RQ is created.

Thus, stage the delay drop initialization as part of RAW ethernet
PF loading process.

Fixes: b5ca15ad7e61 ("IB/mlx5: Add proper representors support")
Signed-off-by: Maher Sanalla <msanalla@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Link: https://lore.kernel.org/r/2e9d386785043d48c38711826eb910315c1de141.1685960567.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-af: fix lbk link credits on cn10k

[ Upstream commit 87e12a17eef476bbf768dc3a74419ad461f36fbc ]

Fix LBK link credits on CN10K to be same as CN9K i.e
16 * MAX_LBK_DATA_RATE instead of current scheme of
calculation based on LBK buf length / FIFO size.

Fixes: 6e54e1c5399a ("octeontx2-af: cn10K: Add MTU configuration")
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-af: fixed resource availability check

[ Upstream commit 4e635f9d86165e47f5440196f2ebdb258efb8341 ]

txschq_alloc response have two different arrays to store continuous
and non-continuous schedulers of each level. Requested count should
be checked for each array separately.

Fixes: 5d9b976d4480 ("octeontx2-af: Support fixed transmit scheduler topology")
Signed-off-by: Satha Rao <skoteshwar@marvell.com>
Signed-off-by: Sunil Kovvuri Goutham <sgoutham@marvell.com>
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: remove mask from iavf_irq_enable_queues()

[ Upstream commit c37cf54c12cfaa51e7aaf88708167b0d3259e64e ]

Enable more than 32 IRQs by removing the u32 bit mask in
iavf_irq_enable_queues(). There is no need for the mask as there are no
callers that select individual IRQs through the bitmask. Also, if the PF
allocates more than 32 IRQs, this mask will prevent us from using all of
them.

Modify the comment in iavf_register.h to show that the maximum number
allowed for the IRQ index is 63 as per the iAVF standard 1.0 [1].

link: [1] https://www.intel.com/content/dam/www/public/us/en/documents/product-specifications/ethernet-adaptive-virtual-function-hardware-spec.pdf
Fixes: 5eae00c57f5e ("i40evf: main driver core")
Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20230608200226.451861-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/rxe: Fix the use-before-initialization error of resp_pkts

[ Upstream commit 2a62b6210ce876c596086ab8fd4c8a0c3d10611a ]

In the following:

  Call Trace:
   <TASK>
   __dump_stack lib/dump_stack.c:88 [inline]
   dump_stack_lvl+0xd9/0x150 lib/dump_stack.c:106
   assign_lock_key kernel/locking/lockdep.c:982 [inline]
   register_lock_class+0xdb6/0x1120 kernel/locking/lockdep.c:1295
   __lock_acquire+0x10a/0x5df0 kernel/locking/lockdep.c:4951
   lock_acquire kernel/locking/lockdep.c:5691 [inline]
   lock_acquire+0x1b1/0x520 kernel/locking/lockdep.c:5656
   __raw_spin_lock_irqsave include/linux/spinlock_api_smp.h:110 [inline]
   _raw_spin_lock_irqsave+0x3d/0x60 kernel/locking/spinlock.c:162
   skb_dequeue+0x20/0x180 net/core/skbuff.c:3639
   drain_resp_pkts drivers/infiniband/sw/rxe/rxe_comp.c:555 [inline]
   rxe_completer+0x250d/0x3cc0 drivers/infiniband/sw/rxe/rxe_comp.c:652
   rxe_qp_do_cleanup+0x1be/0x820 drivers/infiniband/sw/rxe/rxe_qp.c:761
   execute_in_process_context+0x3b/0x150 kernel/workqueue.c:3473
   __rxe_cleanup+0x21e/0x370 drivers/infiniband/sw/rxe/rxe_pool.c:233
   rxe_create_qp+0x3f6/0x5f0 drivers/infiniband/sw/rxe/rxe_verbs.c:583

This is a use-before-initialization problem.

It happens because rxe_qp_do_cleanup is called during error unwind before
the struct has been fully initialized.

Move the initialization of the skb earlier.

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Link: https://lore.kernel.org/r/20230602035408.741534-1-yanjun.zhu@intel.com
Reported-by: syzbot+eba589d8f49c73d356da@syzkaller.appspotmail.com
Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/rxe: Removed unused name from rxe_task struct

[ Upstream commit de669ae8af49ceed0eed44f5b3d51dc62affc5e4 ]

The name field in struct rxe_task is never used. This patch removes it.

Link: https://lore.kernel.org/r/20221021200118.2163-4-rpearsonhpe@gmail.com
Signed-off-by: Ian Ziemba <ian.ziemba@hpe.com>
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Stable-dep-of: 2a62b6210ce8 ("RDMA/rxe: Fix the use-before-initialization error of resp_pkts")
Signed-off-by: Sasha Levin <sashal@kernel.org>

wifi: mac80211: take lock before setting vif links

[ Upstream commit 15846f95ab01b71fdb1cef8df73680aad41edf70 ]

ieee80211_vif_set_links requires the sdata->local->mtx lock to be held.
Add the appropriate locking around the calls in both the link add and
remove handlers.

This causes a warning when e.g. ieee80211_link_release_channel is called
via ieee80211_link_stop from ieee80211_vif_update_links.

Fixes: 0d8c4a3c8688 ("wifi: mac80211: implement add/del interface link callbacks")
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20230608163202.fa0c6597fdad.I83dd70359f6cda30f86df8418d929c2064cf4995@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wifi: cfg80211: fix link del callback to call correct handler

[ Upstream commit 1ff56684fa8682bdfbbce4e12cf67ab23cb1db05 ]

The wrapper function was incorrectly calling the add handler instead of
the del handler. This had no negative side effect as the default
handlers are essentially identical.

Fixes: f2a0290b2df2 ("wifi: cfg80211: add optional link add/remove callbacks")
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20230608163202.ebd00e000459.Iaff7dc8d1cdecf77f53ea47a0e5080caa36ea02a@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wifi: mac80211: fix link activation settings order

[ Upstream commit 01605ad6c3e8608d7e147c9b75d67eb8a3d27d88 ]

In the normal MLME code we always call
ieee80211_mgd_set_link_qos_params() before
ieee80211_link_info_change_notify() and some drivers,
notably iwlwifi, rely on that as they don't do anything
(but store the data) in their conf_tx.

Fix the order here to be the same as in the normal code
paths, so this isn't broken.

Fixes: 3d9011029227 ("wifi: mac80211: implement link switching")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/20230608163202.a2a86bba2f80.Iac97e04827966d22161e63bb6e201b4061e9651b@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: cls_u32: Fix reference counter leak leading to overflow

[ Upstream commit 04c55383fa5689357bcdd2c8036725a55ed632bc ]

In the event of a failure in tcf_change_indev(), u32_set_parms() will
immediately return without decrementing the recently incremented
reference counter.  If this happens enough times, the counter will
rollover and the reference freed, leading to a double free which can be
used to do 'bad things'.

In order to prevent this, move the point of possible failure above the
point where the reference counter is incremented.  Also save any
meaningful return values to be applied to the return data at the
appropriate point in time.

This issue was caught with KASAN.

Fixes: 705c7091262d ("net: sched: cls_u32: no need to call tcf_exts_change for newly allocated struct")
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Lee Jones <lee@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-af: Fix promiscuous mode

[ Upstream commit c0e489372a294044feea650b38f38c888eff57a4 ]

CN10KB silicon introduced a new exact match feature,
which is used for DMAC filtering. The state of installed
DMAC filters in this exact match table is getting corrupted
when promiscuous mode is toggled. Fix this by not touching
Exact match related config when promiscuous mode is toggled.

Fixes: 2dba9459d2c9 ("octeontx2-af: Wrapper functions for MAC addr add/del/update/reset")
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: act_pedit: Parse L3 Header for L4 offset

[ Upstream commit 6c02568fd1ae53099b4ab86365c5be1ff15f586b ]

Instead of relying on skb->transport_header being set correctly, opt
instead to parse the L3 header length out of the L3 headers for both
IPv4/IPv6 when the Extended Layer Op for tcp/udp is used. This fixes a
bug if GRO is disabled, when GRO is disabled skb->transport_header is
set by __netif_receive_skb_core() to point to the L3 header, it's later
fixed by the upper protocol layers, but act_pedit will receive the SKB
before the fixups are completed. The existing behavior causes the
following to edit the L3 header if GRO is disabled instead of the UDP
header:

    tc filter add dev eth0 ingress protocol ip flower ip_proto udp \
 dst_ip 192.168.1.3 action pedit ex munge udp set dport 18053

Also re-introduce a rate-limited warning if we were unable to extract
the header offset when using the 'ex' interface.

Fixes: 71d0ed7079df ("net/act_pedit: Support using offset relative to
the conventional network headers")
Signed-off-by: Max Tottenham <mtottenh@akamai.com>
Reviewed-by: Josh Hunt <johunt@akamai.com>
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202305261541.N165u9TZ-lkp@intel.com/
Reviewed-by: Pedro Tammela <pctammela@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: act_pedit: remove extra check for key type

[ Upstream commit 577140180ba28d0d37bc898c7bd6702c83aa106f ]

The netlink parsing already validates the key 'htype'.
Remove the datapath check as it's redundant.

Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Stable-dep-of: 6c02568fd1ae ("net/sched: act_pedit: Parse L3 Header for L4 offset")
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: simplify tcf_pedit_act

[ Upstream commit 95b069382351826c0ae37938070aa82dbeaf288d ]

Remove the check for a negative number of keys as
this cannot ever happen

Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 6c02568fd1ae ("net/sched: act_pedit: Parse L3 Header for L4 offset")
Signed-off-by: Sasha Levin <sashal@kernel.org>

igb: Fix extts capture value format for 82580/i354/i350

[ Upstream commit 6292d7436cf2f0a2ea8800a1d2cbb155d237818a ]

82580/i354/i350 features circle-counter-like timestamp registers
that are different with newer i210. The EXTTS capture value in
AUXTSMPx should be converted from raw circle counter value to
timestamp value in resolution of 1 nanosec by the driver.

This issue can be reproduced on i350 nics, connecting an 1PPS
signal to a SDP pin, and run 'ts2phc' command to read external
1PPS timestamp value. On i210 this works fine, but on i350 the
extts is not correctly converted.

The i350/i354/82580's SYSTIM and other timestamp registers are
40bit counters, presenting time range of 2^40 ns, that means these
registers overflows every about 1099s. This causes all these regs
can't be used directly in contrast to the newer i210/i211s.

The igb driver needs to convert these raw register values to
valid time stamp format by using kernel timecounter apis for i350s
families. Here the igb_extts() just forgot to do the convert.

Fixes: 38970eac41db ("igb: support EXTTS on 82580/i354/i350")
Signed-off-by: Yuezhen Luan <eggcar.luan@gmail.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230607164116.3768175-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ping6: Fix send to link-local addresses with VRF.

[ Upstream commit 91ffd1bae1dafbb9e34b46813f5b058581d9144d ]

Ping sockets can't send packets when they're bound to a VRF master
device and the output interface is set to a slave device.

For example, when net.ipv4.ping_group_range is properly set, so that
ping6 can use ping sockets, the following kind of commands fails:
  $ ip vrf exec red ping6 fe80::854:e7ff:fe88:4bf1%eth1

What happens is that sk->sk_bound_dev_if is set to the VRF master
device, but 'oif' is set to the real output device. Since both are set
but different, ping_v6_sendmsg() sees their value as inconsistent and
fails.

Fix this by allowing 'oif' to be a slave device of ->sk_bound_dev_if.

This fixes the following kselftest failure:
  $ ./fcnal-test.sh -t ipv6_ping
  [...]
  TEST: ping out, vrf device+address bind - ns-B IPv6 LLA        [FAIL]

Reported-by: Mirsad Todorovac <mirsad.todorovac@alu.unizg.hr>
Closes: https://lore.kernel.org/netdev/b6191f90-ffca-dbca-7d06-88a9788def9c@alu.unizg.hr/
Tested-by: Mirsad Todorovac <mirsad.todorovac@alu.unizg.hr>
Fixes: 5e457896986e ("net: ipv6: Fix ping to link-local addresses.")
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/6c8b53108816a8d0d5705ae37bdc5a8322b5e3d9.1686153846.git.gnault@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: enetc: correct the indexes of highest and 2nd highest TCs

[ Upstream commit 21225873be1472b7c59ed3650396af0e40578112 ]

For ENETC hardware, the TCs are numbered from 0 to N-1, where N
is the number of TCs. Numerically higher TC has higher priority.
It's obvious that the highest priority TC index should be N-1 and
the 2nd highest priority TC index should be N-2.

However, the previous logic uses netdev_get_prio_tc_map() to get
the indexes of highest priority and 2nd highest priority TCs, it
does not make sense and is incorrect to give a "tc" argument to
netdev_get_prio_tc_map(). So the driver may get the wrong indexes
of the two highest priotiry TCs which would lead to failed to set
the CBS for the two highest priotiry TCs.

e.g.
$ tc qdisc add dev eno0 parent root handle 100: mqprio num_tc 6 \
map 0 0 1 1 2 3 4 5 queues 1@0 1@1 1@2 1@3 2@4 2@6 hw 1
$ tc qdisc replace dev eno0 parent 100:6 cbs idleslope 100000 \
sendslope -900000 hicredit 12 locredit -113 offload 1
$ Error: Specified device failed to setup cbs hardware offload.
  ^^^^^

In this example, the previous logic deems the indexes of the two
highest priotiry TCs should be 3 and 2. Actually, the indexes are
5 and 4, because the number of TCs is 6. So it would be failed to
configure the CBS for the two highest priority TCs.

Fixes: c431047c4efe ("enetc: add support Credit Based Shaper(CBS) for hardware offload")
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables: incorrect error path handling with NFT_MSG_NEWRULE

[ Upstream commit 1240eb93f0616b21c675416516ff3d74798fdc97 ]

In case of error when adding a new rule that refers to an anonymous set,
deactivate expressions via NFT_TRANS_PREPARE state, not NFT_TRANS_RELEASE.
Thus, the lookup expression marks anonymous sets as inactive in the next
generation to ensure it is not reachable in this transaction anymore and
decrement the set refcount as introduced by c1592a89942e ("netfilter:
nf_tables: deactivate anonymous set from preparation phase"). The abort
step takes care of undoing the anonymous set.

This is also consistent with rule deletion, where NFT_TRANS_PREPARE is
used. Note that this error path is exercised in the preparation step of
the commit protocol. This patch replaces nf_tables_rule_release() by the
deactivate and destroy calls, this time with NFT_TRANS_PREPARE.

Due to this incorrect error handling, it is possible to access a
dangling pointer to the anonymous set that remains in the transaction
list.

[1009.379054] BUG: KASAN: use-after-free in nft_set_lookup_global+0x147/0x1a0 [nf_tables]
[1009.379106] Read of size 8 at addr ffff88816c4c8020 by task nft-rule-add/137110
[1009.379116] CPU: 7 PID: 137110 Comm: nft-rule-add Not tainted 6.4.0-rc4+ #256
[1009.379128] Call Trace:
[1009.379132]  <TASK>
[1009.379135]  dump_stack_lvl+0x33/0x50
[1009.379146]  ? nft_set_lookup_global+0x147/0x1a0 [nf_tables]
[1009.379191]  print_address_description.constprop.0+0x27/0x300
[1009.379201]  kasan_report+0x107/0x120
[1009.379210]  ? nft_set_lookup_global+0x147/0x1a0 [nf_tables]
[1009.379255]  nft_set_lookup_global+0x147/0x1a0 [nf_tables]
[1009.379302]  nft_lookup_init+0xa5/0x270 [nf_tables]
[1009.379350]  nf_tables_newrule+0x698/0xe50 [nf_tables]
[1009.379397]  ? nf_tables_rule_release+0xe0/0xe0 [nf_tables]
[1009.379441]  ? kasan_unpoison+0x23/0x50
[1009.379450]  nfnetlink_rcv_batch+0x97c/0xd90 [nfnetlink]
[1009.379470]  ? nfnetlink_rcv_msg+0x480/0x480 [nfnetlink]
[1009.379485]  ? __alloc_skb+0xb8/0x1e0
[1009.379493]  ? __alloc_skb+0xb8/0x1e0
[1009.379502]  ? entry_SYSCALL_64_after_hwframe+0x46/0xb0
[1009.379509]  ? unwind_get_return_address+0x2a/0x40
[1009.379517]  ? write_profile+0xc0/0xc0
[1009.379524]  ? avc_lookup+0x8f/0xc0
[1009.379532]  ? __rcu_read_unlock+0x43/0x60

Fixes: 958bee14d071 ("netfilter: nf_tables: use new transaction infrastructure to handle sets")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: Fix XDP memory leak when NIC is brought up and down

[ Upstream commit 78c50d6961fc05491ebbc71c35d87324b1a4f49a ]

Fix the buffer leak that occurs while switching
the port up and down with traffic and XDP by
checking for an active XDP program and freeing all empty TX buffers.

Fixes: efc2214b6047 ("ice: Add support for XDP")
Signed-off-by: Kamil Maziarz <kamil.maziarz@intel.com>
Tested-by: Chandan Kumar Rout <chandanx.rout@intel.com> (A Contingent Worker at Intel)
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nfnetlink: skip error delivery on batch in case of ENOMEM

[ Upstream commit a1a64a151dae8ac3581c1cbde44b672045cb658b ]

If caller reports ENOMEM, then stop iterating over the batch and send a
single netlink message to userspace to report OOM.

Fixes: cbb8125eb40b ("netfilter: nfnetlink: deliver netlink errors on batch completion")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables: integrate pipapo into commit protocol

[ Upstream commit 212ed75dc5fb9d1423b3942c8f872a868cda3466 ]

The pipapo set backend follows copy-on-update approach, maintaining one
clone of the existing datastructure that is being updated. The clone
and current datastructures are swapped via rcu from the commit step.

The existing integration with the commit protocol is flawed because
there is no operation to clean up the clone if the transaction is
aborted. Moreover, the datastructure swap happens on set element
activation.

This patch adds two new operations for sets: commit and abort, these new
operations are invoked from the commit and abort steps, after the
transactions have been digested, and it updates the pipapo set backend
to use it.

This patch adds a new ->pending_update field to sets to maintain a list
of sets that require this new commit and abort operations.

Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: fsl-dspi: avoid SCK glitches with continuous transfers

[ Upstream commit c5c31fb71f16ba75bad4ade208abbae225305b65 ]

The DSPI controller has configurable timing for

(a) tCSC: the interval between the assertion of the chip select and the
    first clock edge

(b) tASC: the interval between the last clock edge and the deassertion
    of the chip select

What is a bit surprising, but is documented in the figure "Example of
continuous transfer (CPHA=1, CONT=1)" in the datasheet, is that when the
chip select stays asserted between multiple TX FIFO writes, the tCSC and
tASC times still apply. With CONT=1, chip select remains asserted, but
SCK takes a break and goes to the idle state for tASC + tCSC ns.

In other words, the default values (of 0 and 0 ns) result in SCK
glitches where the SCK transition to the idle state, as well as the SCK
transition from the idle state, will have no delay in between, and it
may appear that a SCK cycle has simply gone missing. The resulting
timing violation might cause data corruption in many peripherals, as
their chip select is asserted.

The driver has device tree bindings for tCSC ("fsl,spi-cs-sck-delay")
and tASC ("fsl,spi-sck-cs-delay"), but these are only specified to apply
when the chip select toggles in the first place, and this timing
characteristic depends on each peripheral. Many peripherals do not have
explicit timing requirements, so many device trees do not have these
properties present at all.

Nonetheless, the lack of SCK glitches is a common sense requirement, and
since the SCK stays in the idle state during transfers for tCSC+tASC ns,
and that in itself should look like half a cycle, then let's ensure that
tCSC and tASC are at least a quarter of a SCK period, such that their
sum is at least half of one.

Fixes: 95bf15f38641 ("spi: fsl-dspi: Add ~50ns delay between cs and sck")
Reported-by: Lisa Chen (陈敏捷) <minjie.chen@geekplus.com>
Debugged-by: Lisa Chen (陈敏捷) <minjie.chen@geekplus.com>
Tested-by: Lisa Chen (陈敏捷) <minjie.chen@geekplus.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20230529223402.1199503-1-vladimir.oltean@nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: cadence-quadspi: Add missing check for dma_set_mask

[ Upstream commit 947c70a213769f60e9d5aca2bc88b50a1cfaf5a6 ]

Add check for dma_set_mask() and return the error if it fails.

Fixes: 1a6f854f7daa ("spi: cadence-quadspi: Add Xilinx Versal external DMA support")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Link: https://lore.kernel.org/r/20230606093859.27818-1-jiasheng@iscas.ac.cn
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/rxe: Fix ref count error in check_rkey()

[ Upstream commit b00683422fd79dd07c9b75efdce1660e5e19150e ]

There is a reference count error in error path code and a potential race
in check_rkey() in rxe_resp.c. When looking up the rkey for a memory
window the reference to the mw from rxe_lookup_mw() is dropped before a
reference is taken on the mr referenced by the mw. If the mr is destroyed
immediately after the call to rxe_put(mw) the mr pointer is unprotected
and may end up pointing at freed memory. The rxe_get(mr) call should take
place before the rxe_put(mw) call.

All errors in check_rkey() call rxe_put(mw) if mw is not NULL but it was
already called after the above. The mw pointer should be set to NULL after
the rxe_put(mw) call to prevent this from happening.

Fixes: cdd0b85675ae ("RDMA/rxe: Implement memory access through MWs")
Link: https://lore.kernel.org/r/20230517211509.1819998-1-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/rxe: Fix packet length checks

[ Upstream commit 9a3763e87379c97a78b7c6c6f40720b1e877174f ]

In rxe_net.c a received packet, from udp or loopback, is passed to
rxe_rcv() in rxe_recv.c as a udp packet. I.e. skb->data is pointing at the
udp header. But rxe_rcv() makes length checks to verify the packet is long
enough to hold the roce headers as if it were a roce
packet. I.e. skb->data pointing at the bth header. A runt packet would
appear to have 8 more bytes than it actually does which may lead to
incorrect behavior.

This patch calls skb_pull() to adjust the skb to point at the bth header
before calling rxe_rcv() which fixes this error.

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Link: https://lore.kernel.org/r/20230517172242.1806340-1-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/rtrs: Fix rxe_dealloc_pd warning

[ Upstream commit 9c29c8c7df0688f358d2df5ddd16c97c2f7292b4 ]

In current design:
1. PD and clt_path->s.dev are shared among connections.
2. every con[n]'s cleanup phase will call destroy_con_cq_qp()
3. clt_path->s.dev will be always decreased in destroy_con_cq_qp(), and
   when clt_path->s.dev become zero, it will destroy PD.
4. when con[1] failed to create, con[1] will not take clt_path->s.dev,
   but it try to decreased clt_path->s.dev

So, in case create_cm(con[0]) succeeds but create_cm(con[1]) fails,
destroy_con_cq_qp(con[1]) will be called first which will destroy the PD
while this PD is still taken by con[0].

Here, we refactor the error path of create_cm() and init_conns(), so that
we do the cleanup in the order they are created.

The warning occurs when destroying RXE PD whose reference count is not
zero.

 rnbd_client L597: Mapping device /dev/nvme0n1 on session client, (access_mode: rw, nr_poll_queues: 0)
 ------------[ cut here ]------------
 WARNING: CPU: 0 PID: 26407 at drivers/infiniband/sw/rxe/rxe_pool.c:256 __rxe_cleanup+0x13a/0x170 [rdma_rxe]
 Modules linked in: rpcrdma rdma_ucm ib_iser rnbd_client libiscsi rtrs_client scsi_transport_iscsi rtrs_core rdma_cm iw_cm ib_cm crc32_generic rdma_rxe udp_tunnel ib_uverbs ib_core kmem device_dax nd_pmem dax_pmem nd_vme crc32c_intel fuse nvme_core nfit libnvdimm dm_multipath scsi_dh_rdac scsi_dh_emc scsi_dh_alua dm_mirror dm_region_hash dm_log dm_mod
 CPU: 0 PID: 26407 Comm: rnbd-client.sh Kdump: loaded Not tainted 6.2.0-rc6-roce-flush+ #53
 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
 RIP: 0010:__rxe_cleanup+0x13a/0x170 [rdma_rxe]
 Code: 45 84 e4 0f 84 5a ff ff ff 48 89 ef e8 5f 18 71 f9 84 c0 75 90 be c8 00 00 00 48 89 ef e8 be 89 1f fa 85 c0 0f 85 7b ff ff ff <0f> 0b 41 bc ea ff ff ff e9 71 ff ff ff e8 84 7f 1f fa e9 d0 fe ff
 RSP: 0018:ffffb09880b6f5f0 EFLAGS: 00010246
 RAX: 0000000000000000 RBX: ffff99401f15d6a8 RCX: 0000000000000000
 RDX: 0000000000000001 RSI: ffffffffbac8234b RDI: 00000000ffffffff
 RBP: ffff99401f15d6d0 R08: 0000000000000001 R09: 0000000000000001
 R10: 0000000000002d82 R11: 0000000000000000 R12: 0000000000000001
 R13: ffff994101eff208 R14: ffffb09880b6f6a0 R15: 00000000fffffe00
 FS:  00007fe113904740(0000) GS:ffff99413bc00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007ff6cde656c8 CR3: 000000001f108004 CR4: 00000000001706f0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  <TASK>
  rxe_dealloc_pd+0x16/0x20 [rdma_rxe]
  ib_dealloc_pd_user+0x4b/0x80 [ib_core]
  rtrs_ib_dev_put+0x79/0xd0 [rtrs_core]
  destroy_con_cq_qp+0x8a/0xa0 [rtrs_client]
  init_path+0x1e7/0x9a0 [rtrs_client]
  ? __pfx_autoremove_wake_function+0x10/0x10
  ? lock_is_held_type+0xd7/0x130
  ? rcu_read_lock_sched_held+0x43/0x80
  ? pcpu_alloc+0x3dd/0x7d0
  ? rtrs_clt_init_stats+0x18/0x40 [rtrs_client]
  rtrs_clt_open+0x24f/0x5a0 [rtrs_client]
  ? __pfx_rnbd_clt_link_ev+0x10/0x10 [rnbd_client]
  rnbd_clt_map_device+0x6a5/0xe10 [rnbd_client]

Fixes: 6a98d71daea1 ("RDMA/rtrs: client: main functionality")
Link: https://lore.kernel.org/r/1682384563-2-4-git-send-email-lizhijian@fujitsu.com
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Acked-by: Jack Wang <jinpu.wang@ionos.com>
Tested-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/rtrs: Fix the last iu->buf leak in err path

[ Upstream commit 3bf3a7c6985c625f64e73baefdaa36f1c2045a29 ]

The last iu->buf will leak if ib_dma_mapping_error() fails.

Fixes: c0894b3ea69d ("RDMA/rtrs: core: lib functions shared between client and server modules")
Link: https://lore.kernel.org/r/1682384563-2-3-git-send-email-lizhijian@fujitsu.com
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Acked-by: Guoqing Jiang <guoqing.jiang@linux.dev>
Acked-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

usb: dwc3: gadget: Reset num TRBs before giving back the request

commit 00f8205ffcf112dcef14f8151d78075d38d22c08 upstream.

Consider a scenario where cable disconnect happens when there is an active
usb reqest queued to the UDC. As part of the disconnect we would issue an
end transfer with no interrupt-on-completion before giving back this
request. Since we are giving back the request without skipping TRBs the
num_trbs field of dwc3_request still holds the stale value previously used.
Function drivers re-use same request for a given bind-unbind session and
hence their dwc3_request context gets preserved across cable
disconnect/connect. When such a request gets re-queued after cable connect,
we would increase the num_trbs field on top of the previous stale value
thus incorrectly representing the number of TRBs used. Fix this by
resetting num_trbs field before giving back the request.

Fixes: 09fe1f8d7e2f ("usb: dwc3: gadget: track number of TRBs per request")
Cc: stable <stable@kernel.org>
Signed-off-by: Elson Roy Serrao <quic_eserrao@quicinc.com>
Acked-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Message-ID: <1685654850-8468-1-git-send-email-quic_eserrao@quicinc.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: dwc3: fix use-after-free on core driver unbind

commit e3dbb657571509044be15184a13134fa7c1fdca1 upstream.

Some dwc3 glue drivers are currently accessing the driver data of the
child core device directly, which is clearly a bad idea as the child may
not have probed yet or may have been unbound from its driver.

As a workaround until the glue drivers have been fixed, clear the driver
data pointer before allowing the glue parent device to runtime suspend
to prevent its driver from accessing data that has been freed during
unbind.

Fixes: 6dd2565989b4 ("usb: dwc3: add imx8mp dwc3 glue layer driver")
Fixes: 6895ea55c385 ("usb: dwc3: qcom: Configure wakeup interrupts during suspend")
Cc: stable@vger.kernel.org # 5.12
Cc: Li Jun <jun.li@nxp.com>
Cc: Sandeep Maheswaram <quic_c_sanm@quicinc.com>
Cc: Krishna Kurapati <quic_kriskura@quicinc.com>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Acked-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Message-ID: <20230607100540.31045-3-johan+linaro@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: dwc3: qcom: fix NULL-deref on suspend

commit d2d69354226de0b333d4405981f3d9c41ba8430a upstream.

The Qualcomm dwc3 glue driver is currently accessing the driver data of
the child core device during suspend and on wakeup interrupts. This is
clearly a bad idea as the child may not have probed yet or could have
been unbound from its driver.

The first such layering violation was part of the initial version of the
driver, but this was later made worse when the hack that accesses the
driver data of the grand child xhci device to configure the wakeup
interrupts was added.

Fixing this properly is not that easily done, so add a sanity check to
make sure that the child driver data is non-NULL before dereferencing it
for now.

Note that this relies on subtleties like the fact that driver core is
making sure that the parent is not suspended while the child is probing.

Reported-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/all/20230325165217.31069-4-manivannan.sadhasivam@linaro.org/
Fixes: d9152161b4bf ("usb: dwc3: Add Qualcomm DWC3 glue layer driver")
Fixes: 6895ea55c385 ("usb: dwc3: qcom: Configure wakeup interrupts during suspend")
Cc: stable@vger.kernel.org # 3.18: a872ab303d5d: "usb: dwc3: qcom: fix use-after-free on runtime-PM wakeup"
Cc: Sandeep Maheswaram <quic_c_sanm@quicinc.com>
Cc: Krishna Kurapati <quic_kriskura@quicinc.com>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Acked-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Message-ID: <20230607100540.31045-2-johan+linaro@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: gadget: udc: core: Prevent soft_connect_store() race

commit 286d9975a838d0a54da049765fa1d1fb96b89682 upstream.

usb_udc_connect_control(), soft_connect_store() and
usb_gadget_deactivate() can potentially race against each other to invoke
usb_gadget_connect()/usb_gadget_disconnect(). To prevent this, guard
udc->started, gadget->allow_connect, gadget->deactivate and
gadget->connect with connect_lock so that ->pullup() is only invoked when
the gadget is bound, started and not deactivated. The routines
usb_gadget_connect_locked(), usb_gadget_disconnect_locked(),
usb_udc_connect_control_locked(), usb_gadget_udc_start_locked(),
usb_gadget_udc_stop_locked() are called with this lock held.

An earlier version of this commit was reverted due to the crash reported in
https://lore.kernel.org/all/ZF4BvgsOyoKxdPFF@francesco-nb.int.toradex.com/.
commit 16737e78d190 ("usb: gadget: udc: core: Offload usb_udc_vbus_handler processing")
addresses the crash reported.

Cc: stable@vger.kernel.org
Fixes: 628ef0d273a6 ("usb: udc: add usb_udc_vbus_handler")
Signed-off-by: Badhri Jagan Sridharan <badhri@google.com>
Reviewed-by: Alan Stern <stern@rowland.harvard.edu>
Message-ID: <20230609010227.978661-2-badhri@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>