tcp: Fix a data-race around sysctl_tcp_max_orphans.

[ Upstream commit 47e6ab24e8c6e3ca10ceb5835413f401f90de4bf ]

While reading sysctl_tcp_max_orphans, it can be changed concurrently.
So, we need to add READ_ONCE() to avoid a data-race.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sysctl: Fix data races in proc_dointvec_jiffies().

[ Upstream commit e877820877663fbae8cb9582ea597a7230b94df3 ]

A sysctl variable is accessed concurrently, and there is always a chance
of data-race.  So, all readers and writers need some basic protection to
avoid load/store-tearing.

This patch changes proc_dointvec_jiffies() to use READ_ONCE() and
WRITE_ONCE() internally to fix data-races on the sysctl side.  For now,
proc_dointvec_jiffies() itself is tolerant to a data-race, but we still
need to add annotations on the other subsystem's side.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sysctl: Fix data races in proc_doulongvec_minmax().

[ Upstream commit c31bcc8fb89fc2812663900589c6325ba35d9a65 ]

A sysctl variable is accessed concurrently, and there is always a chance
of data-race.  So, all readers and writers need some basic protection to
avoid load/store-tearing.

This patch changes proc_doulongvec_minmax() to use READ_ONCE() and
WRITE_ONCE() internally to fix data-races on the sysctl side.  For now,
proc_doulongvec_minmax() itself is tolerant to a data-race, but we still
need to add annotations on the other subsystem's side.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sysctl: Fix data races in proc_douintvec_minmax().

[ Upstream commit 2d3b559df3ed39258737789aae2ae7973d205bc1 ]

A sysctl variable is accessed concurrently, and there is always a chance
of data-race.  So, all readers and writers need some basic protection to
avoid load/store-tearing.

This patch changes proc_douintvec_minmax() to use READ_ONCE() and
WRITE_ONCE() internally to fix data-races on the sysctl side.  For now,
proc_douintvec_minmax() itself is tolerant to a data-race, but we still
need to add annotations on the other subsystem's side.

Fixes: 61d9b56a8920 ("sysctl: add unsigned int range support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sysctl: Fix data races in proc_dointvec_minmax().

[ Upstream commit f613d86d014b6375a4085901de39406598121e35 ]

A sysctl variable is accessed concurrently, and there is always a chance
of data-race.  So, all readers and writers need some basic protection to
avoid load/store-tearing.

This patch changes proc_dointvec_minmax() to use READ_ONCE() and
WRITE_ONCE() internally to fix data-races on the sysctl side.  For now,
proc_dointvec_minmax() itself is tolerant to a data-race, but we still
need to add annotations on the other subsystem's side.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sysctl: Fix data races in proc_douintvec().

[ Upstream commit 4762b532ec9539755aab61445d5da6e1926ccb99 ]

A sysctl variable is accessed concurrently, and there is always a chance
of data-race.  So, all readers and writers need some basic protection to
avoid load/store-tearing.

This patch changes proc_douintvec() to use READ_ONCE() and WRITE_ONCE()
internally to fix data-races on the sysctl side.  For now, proc_douintvec()
itself is tolerant to a data-race, but we still need to add annotations on
the other subsystem's side.

Fixes: e7d316a02f68 ("sysctl: handle error writing UINT_MAX to u32 fields")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sysctl: Fix data races in proc_dointvec().

[ Upstream commit 1f1be04b4d48a2475ea1aab46a99221bfc5c0968 ]

A sysctl variable is accessed concurrently, and there is always a chance
of data-race.  So, all readers and writers need some basic protection to
avoid load/store-tearing.

This patch changes proc_dointvec() to use READ_ONCE() and WRITE_ONCE()
internally to fix data-races on the sysctl side.  For now, proc_dointvec()
itself is tolerant to a data-race, but we still need to add annotations on
the other subsystem's side.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ethernet: ti: am65-cpsw: Fix devlink port register sequence

[ Upstream commit 0680e20af5fbf41df8a11b11bd9a7c25b2ca0746 ]

Renaming interfaces using udevd depends on the interface being registered
before its netdev is registered. Otherwise, udevd reads an empty
phys_port_name value, resulting in the interface not being renamed.

Fix this by registering the interface before registering its netdev
by invoking am65_cpsw_nuss_register_devlink() before invoking
register_netdev() for the interface.

Move the function call to devlink_port_type_eth_set(), invoking it after
register_netdev() is invoked, to ensure that netlink notification for the
port state change is generated after the netdev is completely initialized.

Fixes: 58356eb31d60 ("net: ti: am65-cpsw-nuss: Add devlink support")
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Link: https://lore.kernel.org/r/20220706070208.12207-1-s-vadapalli@ti.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: stmmac: dwc-qos: Disable split header for Tegra194

[ Upstream commit 029c1c2059e9c4b38f97a06204cdecd10cfbeb8a ]

There is a long-standing issue with the Synopsys DWC Ethernet driver
for Tegra194 where random system crashes have been observed [0]. The
problem occurs when the split header feature is enabled in the stmmac
driver. In the bad case, a larger than expected buffer length is
received and causes the calculation of the total buffer length to
overflow. This results in a very large buffer length that causes the
kernel to crash. Why this larger buffer length is received is not clear,
however, the feedback from the NVIDIA design team is that the split
header feature is not supported for Tegra194. Therefore, disable split
header support for Tegra194 to prevent these random crashes from
occurring.

[0] https://lore.kernel.org/linux-tegra/b0b17697-f23e-8fa5-3757-604a86f3a095@nvidia.com/

Fixes: 67afd6d1cfdf ("net: stmmac: Add Split Header support and enable it in XGMAC cores")
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Link: https://lore.kernel.org/r/20220706083913.13750-1-jonathanh@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: Intel: Skylake: Correct the handling of fmt_config flexible array

[ Upstream commit fc976f5629afb4160ee77798b14a693eac903ffd ]

The struct nhlt_format's fmt_config is a flexible array, it must not be
used as normal array.
When moving to the next nhlt_fmt_cfg we need to take into account the data
behind the ->config.caps (indicated by ->config.size).

The logic of the code also changed: it is no longer saves the _last_
fmt_cfg for all found rates.

Fixes: bc2bd45b1f7f3 ("ASoC: Intel: Skylake: Parse nhlt and register clock device")
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com>
Reviewed-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://lore.kernel.org/r/20220630065638.11183-3-peter.ujfalusi@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: Intel: Skylake: Correct the ssp rate discovery in skl_get_ssp_clks()

[ Upstream commit 219af251bd1694bce1f627d238347d2eaf13de61 ]

The present flag is only set once when one rate has been found to be saved.
This will effectively going to ignore any rate discovered at later time and
based on the code, this is not the intention.

Fixes: bc2bd45b1f7f3 ("ASoC: Intel: Skylake: Parse nhlt and register clock device")
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com>
Reviewed-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://lore.kernel.org/r/20220630065638.11183-2-peter.ujfalusi@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: tas2764: Fix amp gain register offset & default

[ Upstream commit 1c4f29ec878bbf1cc0a1eb54ae7da5ff98e19641 ]

The register default is 0x28 per the datasheet, and the amp gain field
is supposed to be shifted left by one. With the wrong default, the ALSA
controls lie about the power-up state. With the wrong shift, we get only
half the gain we expect.

Signed-off-by: Hector Martin <marcan@marcan.st>
Fixes: 827ed8a0fa50 ("ASoC: tas2764: Add the driver for the TAS2764")
Signed-off-by: Martin Povišer <povik+lin@cutebit.org>
Link: https://lore.kernel.org/r/20220630075135.2221-4-povik+lin@cutebit.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: tas2764: Correct playback volume range

[ Upstream commit 3e99e5697e1f7120b5abc755e8a560b22612d6ed ]

DVC value 0xc8 is -100dB and 0xc9 is mute; this needs to map to
-100.5dB as far as the dB scale is concerned. Fix that and enable
the mute flag, so alsamixer correctly shows the control as
<0 dB .. -100 dB, mute>.

Signed-off-by: Hector Martin <marcan@marcan.st>
Fixes: 827ed8a0fa50 ("ASoC: tas2764: Add the driver for the TAS2764")
Signed-off-by: Martin Povišer <povik+lin@cutebit.org>
Link: https://lore.kernel.org/r/20220630075135.2221-3-povik+lin@cutebit.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: tas2764: Fix and extend FSYNC polarity handling

[ Upstream commit d1a10f1b48202e2d183cce144c218a211e98d906 ]

Fix setting of FSYNC polarity in case of LEFT_J and DSP_A/B formats.
Do NOT set the SCFG field as was previously done, because that is not
correct and is also in conflict with the "ASI1 Source" control which
sets the same SCFG field!

Also add support for explicit polarity inversion.

Fixes: 827ed8a0fa50 ("ASoC: tas2764: Add the driver for the TAS2764")
Signed-off-by: Martin Povišer <povik+lin@cutebit.org>
Link: https://lore.kernel.org/r/20220630075135.2221-2-povik+lin@cutebit.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: tas2764: Add post reset delays

[ Upstream commit cd10bb89b0d57bca98eb75e0444854a1c129a14e ]

Make sure there is at least 1 ms delay from reset to first command as
is specified in the datasheet. This is a fix similar to commit
307f31452078 ("ASoC: tas2770: Insert post reset delay").

Fixes: 827ed8a0fa50 ("ASoC: tas2764: Add the driver for the TAS2764")
Signed-off-by: Martin Povišer <povik+lin@cutebit.org>
Link: https://lore.kernel.org/r/20220630075135.2221-1-povik+lin@cutebit.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: sgtl5000: Fix noise on shutdown/remove

[ Upstream commit 040e3360af3736348112d29425bf5d0be5b93115 ]

Put the SGTL5000 in a silent/safe state on shutdown/remove, this is
required since the SGTL5000 produces a constant noise on its output
after it is configured and its clock is removed. Without this change
this is happening every time the module is unbound/removed or from
reboot till the clock is enabled again.

The issue was experienced on both a Toradex Colibri/Apalis iMX6, but can
be easily reproduced everywhere just playing something on the codec and
after that removing/unbinding the driver.

Fixes: 9b34e6cc3bc2 ("ASoC: Add Freescale SGTL5000 codec support")
Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com>
Reviewed-by: Fabio Estevam <festevam@denx.de>
Link: https://lore.kernel.org/r/20220624101301.441314-1-francesco.dolcini@toradex.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ima: Fix a potential integer overflow in ima_appraise_measurement

[ Upstream commit d2ee2cfc4aa85ff6a2a3b198a3a524ec54e3d999 ]

When the ima-modsig is enabled, the rc passed to evm_verifyxattr() may be
negative, which may cause the integer overflow problem.

Fixes: 39b07096364a ("ima: Implement support for module-style appended signatures")
Signed-off-by: Huaxin Lu <luhuaxin1@huawei.com>
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/i915: fix a possible refcount leak in intel_dp_add_mst_connector()

[ Upstream commit 85144df9ff4652816448369de76897c57cbb1b93 ]

If drm_connector_init fails, intel_connector_free will be called to take
care of proper free. So it is necessary to drop the refcount of port
before intel_connector_free.

Fixes: 091a4f91942a ("drm/i915: Handle drm-layer errors in intel_dp_add_mst_connector")
Signed-off-by: Hangyu Hua <hbh25y@gmail.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220624130406.17996-1-jose.souza@intel.com
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
(cherry picked from commit cea9ed611e85d36a05db52b6457bf584b7d969e2)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Ring the TX doorbell on DMA errors

[ Upstream commit 5b759bf2f9d73db05369aef2344502095c4e5e73 ]

TX doorbells may be postponed, because sometimes the driver knows that
another packet follows (for example, when xmit_more is true, or when a
MPWQE session is closed before transmitting a packet).

However, the DMA mapping may fail for the next packet, in which case a
new WQE is not posted, the doorbell isn't updated either, and the
transmission of the previous packet will be delayed indefinitely.

This commit fixes the described rare error flow by posting a NOP and
ringing the doorbell on errors to flush all the previous packets. The
MPWQE session is closed before that. DMA mapping in the MPWQE flow is
moved to the beginning of mlx5e_sq_xmit_mpwqe, because empty sessions
are not allowed. Stop room always has enough space for a NOP, because
the actual TX WQE is not posted.

Fixes: e586b3b0baee ("net/mlx5: Ethernet Datapath files")
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Fix capability check for updating vnic env counters

[ Upstream commit 452133dd580811f184e76b1402983182ee425298 ]

The existing capability check for vnic env counters only checks for
receive steering discards, although we need the counters update for the
exposed internal queue oob counter as well. This could result in the
latter counter not being updated correctly when the receive steering
discards counter is not supported.
Fix that by checking whether any counter is supported instead of only
the steering counter capability.

Fixes: 0cfafd4b4ddf ("net/mlx5e: Add device out of buffer counter")
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Fix enabling sriov while tc nic rules are offloaded

[ Upstream commit 0c9d876545a56aebed30fa306d0460a4d28d271a ]

There is a total of four 4M entries flow tables. In sriov disabled
mode, ct, ct_nat and post_act take three of them. When adding the
first tc nic rule in this mode, it will take another 4M table
for the tc <chain,prio> table. If user then enables sriov, the legacy
flow table tries to take another 4M and fails, and so enablement fails.

To fix that, have legacy fdb take the next available maximum
size from the fs ft pool.

Fixes: 4a98544d1827 ("net/mlx5: Move chains ft pool to be used by all firmware steering")
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: kTLS, Fix build time constant test in RX

[ Upstream commit 2ec6cf9b742a5c18982861322fa5de6510f8f57e ]

Use the correct constant (TLS_DRIVER_STATE_SIZE_RX) in the comparison
against the size of the private RX TLS driver context.

Fixes: 1182f3659357 ("net/mlx5e: kTLS, Add kTLS RX HW offload support")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: kTLS, Fix build time constant test in TX

[ Upstream commit 6cc2714e85754a621219693ea8aa3077d6fca0cb ]

Use the correct constant (TLS_DRIVER_STATE_SIZE_TX) in the comparison
against the size of the private TX TLS driver context.

Fixes: df8d866770f9 ("net/mlx5e: kTLS, Use kernel API to extract private offload context")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: 9210/1: Mark the FDT_FIXED sections as shareable

[ Upstream commit 598f0a99fa8a35be44b27106b43ddc66417af3b1 ]

commit 7a1be318f579 ("ARM: 9012/1: move device tree mapping out of linear
region") use FDT_FIXED_BASE to map the whole FDT_FIXED_SIZE memory area
which contains fdt. But it only reserves the exact physical memory that
fdt occupied. Unfortunately, this mapping is non-shareable. An illegal or
speculative read access can bring the RAM content from non-fdt zone into
cache, PIPT makes it to be hit by subsequently read access through
shareable mapping(such as linear mapping), and the cache consistency
between cores is lost due to non-shareable property.

|<---------FDT_FIXED_SIZE------>|
|                               |
 -------------------------------
| <non-fdt> | <fdt> | <non-fdt> |
 -------------------------------

1. CoreA read <non-fdt> through MT_ROM mapping, the old data is loaded
   into the cache.
2. CoreB write <non-fdt> to update data through linear mapping. CoreA
   received the notification to invalid the corresponding cachelines, but
   the property non-shareable makes it to be ignored.
3. CoreA read <non-fdt> through linear mapping, cache hit, the old data
   is read.

To eliminate this risk, add a new memory type MT_MEMORY_RO. Compared to
MT_ROM, it is shareable and non-executable.

Here's an example:
  list_del corruption. prev->next should be c0ecbf74, but was c08410dc
  kernel BUG at lib/list_debug.c:53!
  ... ...
  PC is at __list_del_entry_valid+0x58/0x98
  LR is at __list_del_entry_valid+0x58/0x98
  psr: 60000093
  sp : c0ecbf30  ip : 00000000  fp : 00000001
  r10: c08410d0  r9 : 00000001  r8 : c0825e0c
  r7 : 20000013  r6 : c08410d0  r5 : c0ecbf74  r4 : c0ecbf74
  r3 : c0825d08  r2 : 00000000  r1 : df7ce6f4  r0 : 00000044
  ... ...
  Stack: (0xc0ecbf30 to 0xc0ecc000)
  bf20:                                     c0ecbf74 c0164fd0 c0ecbf70 c0165170
  bf40: c0eca000 c0840c00 c0840c00 c0824500 c0825e0c c0189bbc c088f404 60000013
  bf60: 60000013 c0e85100 000004ec 00000000 c0ebcdc0 c0ecbf74 c0ecbf74 c0825d08
  ... ...                                           <  next     prev  >
  (__list_del_entry_valid) from (__list_del_entry+0xc/0x20)
  (__list_del_entry) from (finish_swait+0x60/0x7c)
  (finish_swait) from (rcu_gp_kthread+0x560/0xa20)
  (rcu_gp_kthread) from (kthread+0x14c/0x15c)
  (kthread) from (ret_from_fork+0x14/0x24)

The faulty list node to be deleted is a local variable, its address is
c0ecbf74. The dumped stack shows that 'prev' = c0ecbf74, but its value
before lib/list_debug.c:53 is c08410dc. A large amount of printing results
in swapping out the cacheline containing the old data(MT_ROM mapping is
read only, so the cacheline cannot be dirty), and the subsequent dump
operation obtains new data from the DDR.

Fixes: 7a1be318f579 ("ARM: 9012/1: move device tree mapping out of linear region")
Suggested-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: 9209/1: Spectre-BHB: avoid pr_info() every time a CPU comes out of idle

[ Upstream commit 0609e200246bfd3b7516091c491bec4308349055 ]

Jon reports that the Spectre-BHB init code is filling up the kernel log
with spurious notifications about which mitigation has been enabled,
every time any CPU comes out of a low power state.

Given that Spectre-BHB mitigations are system wide, only a single
mitigation can be enabled, and we already print an error if two types of
CPUs coexist in a single system that require different Spectre-BHB
mitigations.

This means that the pr_info() that describes the selected mitigation
does not need to be emitted for each CPU anyway, and so we can simply
emit it only once.

In order to clarify the above in the log message, update it to describe
that the selected mitigation will be enabled on all CPUs, including ones
that are unaffected. If another CPU comes up later that is affected and
requires a different mitigation, we report an error as before.

Fixes: b9baf5c8c5c3 ("ARM: Spectre-BHB workaround")
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: amd: Limit max transfer and message size

[ Upstream commit 6ece49c56965544262523dae4a071ace3db63507 ]

Enabling the SPI CS35L41 audio codec driver for Steam Deck [1]
revealed a problem with the current AMD SPI controller driver
implementation, consisting of an unrecoverable system hang.

The issue can be prevented if we ensure the max transfer size
and the max message size do not exceed the FIFO buffer size.

According to the implementation of the downstream driver, the
AMD SPI controller is not able to handle more than 70 bytes per
transfer, which corresponds to the size of the FIFO buffer.

Hence, let's fix this by setting the SPI limits mentioned above.

[1] https://lore.kernel.org/r/20220621213819.262537-1-cristian.ciocaltea@collabora.com

Reported-by: Anastasios Vacharakis <vacharakis@o2mail.de>
Fixes: bbb336f39efc ("spi: spi-amd: Add AMD SPI controller driver support")
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Link: https://lore.kernel.org/r/20220706100626.1234731-2-cristian.ciocaltea@collabora.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: imx6qdl-ts7970: Fix ngpio typo and count

[ Upstream commit e95ea0f687e679fcb0a3a67d0755b81ee7d60db0 ]

Device-tree incorrectly used "ngpio" which caused the driver to
fallback to 32 ngpios.

This platform has 62 GPIO registers.

Fixes: 9ff8e9fccef9 ("ARM: dts: TS-7970: add basic device tree")
Signed-off-by: Kris Bahnsen <kris@embeddedTS.com>
Reviewed-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

reset: Fix devm bulk optional exclusive control getter

[ Upstream commit a57f68ddc8865d59a19783080cc52fb4a11dc209 ]

Most likely due to copy-paste mistake the device managed version of the
denoted reset control getter has been implemented with invalid semantic,
which can be immediately spotted by having "WARN_ON(shared && acquired)"
warning in the system log as soon as the method is called. Anyway let's
fix it by altering the boolean arguments passed to the
__devm_reset_control_bulk_get() method from
- shared = true, optional = false, acquired = true
to
+ shared = false, optional = true, acquired = true
That's what they were supposed to be in the first place (see the non-devm
version of the same method: reset_control_bulk_get_optional_exclusive()).

Fixes: 48d71395896d ("reset: Add reset_control_bulk API")
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Reviewed-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://lore.kernel.org/r/20220624141853.7417-2-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>

xfs: drop async cache flushes from CIL commits.

[ Upstream commit 919edbadebe17a67193533f531c2920c03e40fa4 ]

Jan Kara reported a performance regression in dbench that he
bisected down to commit bad77c375e8d ("xfs: CIL checkpoint
flushes caches unconditionally").

Whilst developing the journal flush/fua optimisations this cache was
part of, it appeared to made a significant difference to
performance. However, now that this patchset has settled and all the
correctness issues fixed, there does not appear to be any
significant performance benefit to asynchronous cache flushes.

In fact, the opposite is true on some storage types and workloads,
where additional cache flushes that can occur from fsync heavy
workloads have measurable and significant impact on overall
throughput.

Local dbench testing shows little difference on dbench runs with
sync vs async cache flushes on either fast or slow SSD storage, and
no difference in streaming concurrent async transaction workloads
like fs-mark.

Fast NVME storage.

>From `dbench -t 30`, CIL scale:

clients async sync
BW Latency BW Latency
1  935.18   0.855  915.64   0.903
8 2404.51   6.873 2341.77   6.511
16 3003.42   6.460 2931.57   6.529
32 3697.23   7.939 3596.28   7.894
128 7237.43  15.495 7217.74  11.588
512 5079.24  90.587 5167.08  95.822

fsmark, 32 threads, create w/ 64 byte xattr w/32k logbsize

create chown unlink
async   1m41s 1m16s 2m03s
sync 1m40s 1m19s 1m54s

Slower SATA SSD storage:

>From `dbench -t 30`, CIL scale:

clients async sync
BW Latency BW Latency
1   78.59  15.792   83.78  10.729
8  367.88  92.067  404.63  59.943
16  564.51  72.524  602.71  76.089
32  831.66 105.984  870.26 110.482
128 1659.76 102.969 1624.73  91.356
512 2135.91 223.054 2603.07 161.160

fsmark, 16 threads, create w/32k logbsize

create unlink
async   5m06s 4m15s
sync 5m00s 4m22s

And on Jan's test machine:

                   5.18-rc8-vanilla       5.18-rc8-patched
Amean     1        71.22 (   0.00%)       64.94 *   8.81%*
Amean     2        93.03 (   0.00%)       84.80 *   8.85%*
Amean     4       150.54 (   0.00%)      137.51 *   8.66%*
Amean     8       252.53 (   0.00%)      242.24 *   4.08%*
Amean     16      454.13 (   0.00%)      439.08 *   3.31%*
Amean     32      835.24 (   0.00%)      829.74 *   0.66%*
Amean     64     1740.59 (   0.00%)     1686.73 *   3.09%*

Performance and cache flush behaviour is restored to pre-regression
levels.

As such, we can now consider the async cache flush mechanism an
unnecessary exercise in premature optimisation and hence we can
now remove it and the infrastructure it requires completely.

Fixes: bad77c375e8d ("xfs: CIL checkpoint flushes caches unconditionally")
Reported-and-tested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfs: run callbacks before waking waiters in xlog_state_shutdown_callbacks

[ Upstream commit cd6f79d1fb324968a3bae92f82eeb7d28ca1fd22 ]

Brian reported a null pointer dereference failure during unmount in
xfs/006. He tracked the problem down to the AIL being torn down
before a log shutdown had completed and removed all the items from
the AIL. The failure occurred in this path while unmount was
proceeding in another task:

 xfs_trans_ail_delete+0x102/0x130 [xfs]
 xfs_buf_item_done+0x22/0x30 [xfs]
 xfs_buf_ioend+0x73/0x4d0 [xfs]
 xfs_trans_committed_bulk+0x17e/0x2f0 [xfs]
 xlog_cil_committed+0x2a9/0x300 [xfs]
 xlog_cil_process_committed+0x69/0x80 [xfs]
 xlog_state_shutdown_callbacks+0xce/0xf0 [xfs]
 xlog_force_shutdown+0xdf/0x150 [xfs]
 xfs_do_force_shutdown+0x5f/0x150 [xfs]
 xlog_ioend_work+0x71/0x80 [xfs]
 process_one_work+0x1c5/0x390
 worker_thread+0x30/0x350
 kthread+0xd7/0x100
 ret_from_fork+0x1f/0x30

This is processing an EIO error to a log write, and it's
triggering a force shutdown. This causes the log to be shut down,
and then it is running attached iclog callbacks from the shutdown
context. That means the fs and log has already been marked as
xfs_is_shutdown/xlog_is_shutdown and so high level code will abort
(e.g. xfs_trans_commit(), xfs_log_force(), etc) with an error
because of shutdown.

The umount would have been blocked waiting for a log force
completion inside xfs_log_cover() -> xfs_sync_sb(). The first thing
for this situation to occur is for xfs_sync_sb() to exit without
waiting for the iclog buffer to be comitted to disk. The
above trace is the completion routine for the iclog buffer, and
it is shutting down the filesystem.

xlog_state_shutdown_callbacks() does this:

{
        struct xlog_in_core     *iclog;
        LIST_HEAD(cb_list);

        spin_lock(&log->l_icloglock);
        iclog = log->l_iclog;
        do {
                if (atomic_read(&iclog->ic_refcnt)) {
                        /* Reference holder will re-run iclog callbacks. */
                        continue;
                }
                list_splice_init(&iclog->ic_callbacks, &cb_list);
>>>>>>           wake_up_all(&iclog->ic_write_wait);
>>>>>>           wake_up_all(&iclog->ic_force_wait);
        } while ((iclog = iclog->ic_next) != log->l_iclog);

        wake_up_all(&log->l_flush_wait);
        spin_unlock(&log->l_icloglock);

>>>>>>  xlog_cil_process_committed(&cb_list);
}

This wakes any thread waiting on IO completion of the iclog (in this
case the umount log force) before shutdown processes all the pending
callbacks.  That means the xfs_sync_sb() waiting on a sync
transaction in xfs_log_force() on iclog->ic_force_wait will get
woken before the callbacks attached to that iclog are run. This
results in xfs_sync_sb() returning an error, and so unmount unblocks
and continues to run whilst the log shutdown is still in progress.

Normally this is just fine because the force waiter has nothing to
do with AIL operations. But in the case of this unmount path, the
log force waiter goes on to tear down the AIL because the log is now
shut down and so nothing ever blocks it again from the wait point in
xfs_log_cover().

Hence it's a race to see who gets to the AIL first - the unmount
code or xlog_cil_process_committed() killing the superblock buffer.

To fix this, we just have to change the order of processing in
xlog_state_shutdown_callbacks() to run the callbacks before it wakes
any task waiting on completion of the iclog.

Reported-by: Brian Foster <bfoster@redhat.com>
Fixes: aad7272a9208 ("xfs: separate out log shutdown callback processing")
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfs: don't include bnobt blocks when reserving free block pool

[ Upstream commit c8c568259772751a14e969b7230990508de73d9d ]

xfs_reserve_blocks controls the size of the user-visible free space
reserve pool.  Given the difference between the current and requested
pool sizes, it will try to reserve free space from fdblocks.  However,
the amount requested from fdblocks is also constrained by the amount of
space that we think xfs_mod_fdblocks will give us.  If we forget to
subtract m_allocbt_blks before calling xfs_mod_fdblocks, it will will
return ENOSPC and we'll hang the kernel at mount due to the infinite
loop.

In commit fd43cf600cf6, we decided that xfs_mod_fdblocks should not hand
out the "free space" used by the free space btrees, because some portion
of the free space btrees hold in reserve space for future btree
expansion.  Unfortunately, xfs_reserve_blocks' estimation of the number
of blocks that it could request from xfs_mod_fdblocks was not updated to
include m_allocbt_blks, so if space is extremely low, the caller hangs.

Fix this by creating a function to estimate the number of blocks that
can be reserved from fdblocks, which needs to exclude the set-aside and
m_allocbt_blks.

Found by running xfs/306 (which formats a single-AG 20MB filesystem)
with an fstests configuration that specifies a 1k blocksize and a
specially crafted log size that will consume 7/8 of the space (17920
blocks, specifically) in that AG.

Cc: Brian Foster <bfoster@redhat.com>
Fixes: fd43cf600cf6 ("xfs: set aside allocation btree blocks from block reservation")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfs: only run COW extent recovery when there are no live extents

[ Upstream commit 7993f1a431bc5271369d359941485a9340658ac3 ]

As part of multiple customer escalations due to file data corruption
after copy on write operations, I wrote some fstests that use fsstress
to hammer on COW to shake things loose.  Regrettably, I caught some
filesystem shutdowns due to incorrect rmap operations with the following
loop:

mount <filesystem> # (0)
fsstress <run only readonly ops> & # (1)
while true; do
fsstress <run all ops>
mount -o remount,ro # (2)
fsstress <run only readonly ops>
mount -o remount,rw # (3)
done

When (2) happens, notice that (1) is still running.  xfs_remount_ro will
call xfs_blockgc_stop to walk the inode cache to free all the COW
extents, but the blockgc mechanism races with (1)'s reader threads to
take IOLOCKs and loses, which means that it doesn't clean them all out.
Call such a file (A).

When (3) happens, xfs_remount_rw calls xfs_reflink_recover_cow, which
walks the ondisk refcount btree and frees any COW extent that it finds.
This function does not check the inode cache, which means that incore
COW forks of inode (A) is now inconsistent with the ondisk metadata.  If
one of those former COW extents are allocated and mapped into another
file (B) and someone triggers a COW to the stale reservation in (A), A's
dirty data will be written into (B) and once that's done, those blocks
will be transferred to (A)'s data fork without bumping the refcount.

The results are catastrophic -- file (B) and the refcount btree are now
corrupt.  In the first patch, we fixed the race condition in (2) so that
(A) will always flush the COW fork.  In this second patch, we move the
_recover_cow call to the initial mount call in (0) for safety.

As mentioned previously, xfs_reflink_recover_cow walks the refcount
btree looking for COW staging extents, and frees them.  This was
intended to be run at mount time (when we know there are no live inodes)
to clean up any leftover staging events that may have been left behind
during an unclean shutdown.  As a time "optimization" for readonly
mounts, we deferred this to the ro->rw transition, not realizing that
any failure to clean all COW forks during a rw->ro transition would
result in catastrophic corruption.

Therefore, remove this optimization and only run the recovery routine
when we're guaranteed not to have any COW staging extents anywhere,
which means we always run this at mount time.  While we're at it, move
the callsite to xfs_log_mount_finish because any refcount btree
expansion (however unlikely given that we're removing records from the
right side of the index) must be fed by a per-AG reservation, which
doesn't exist in its current location.

Fixes: 174edb0e46e5 ("xfs: store in-progress CoW allocations in the refcount btree")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "evm: Fix memleak in init_desc"

commit 51dd64bb99e4478fc5280171acd8e1b529eadaf7 upstream.

This reverts commit ccf11dbaa07b328fa469415c362d33459c140a37.

Commit ccf11dbaa07b ("evm: Fix memleak in init_desc") said there is
memleak in init_desc. That may be incorrect, as we can see, tmp_tfm is
saved in one of the two global variables hmac_tfm or evm_tfm[hash_algo],
then if init_desc is called next time, there is no need to alloc tfm
again, so in the error path of kmalloc desc or crypto_shash_init(desc),
It is not a problem without freeing tmp_tfm.

And also that commit did not reset the global variable to NULL after
freeing tmp_tfm and this makes *tfm a dangling pointer which may cause a
UAF issue.

Reported-by: Guozihua (Scott) <guozihua@huawei.com>
Signed-off-by: Xiu Jianfeng <xiujianfeng@huawei.com>
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sh: convert nommu io{re,un}map() to static inline functions

commit d684e0a52d36f8939eda30a0f31ee235ee4ee741 upstream.

Recently, nommu iounmap() was converted from a static inline function to a
macro again, basically reverting commit 4580ba4ad2e6b8dd ("sh: Convert
iounmap() macros to inline functions").  With -Werror, this leads to build
failures like:

    drivers/iio/adc/xilinx-ams.c: In function `ams_iounmap_ps':
    drivers/iio/adc/xilinx-ams.c:1195:14: error: unused variable `ams' [-Werror=unused-variable]
     1195 |  struct ams *ams = data;
  |              ^~~

Fix this by replacing the macros for ioremap() and iounmap() by static
inline functions, based on <asm-generic/io.h>.

Link: https://lkml.kernel.org/r/8d1b1766260961799b04035e7bc39a7f59729f72.1655708312.git.geert+renesas@glider.be
Fixes: 13f1fc870dd74713 ("sh: move the ioremap implementation out of line")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nilfs2: fix incorrect masking of permission flags for symlinks

commit 5924e6ec1585445f251ea92713eb15beb732622a upstream.

The permission flags of newly created symlinks are wrongly dropped on
nilfs2 with the current umask value even though symlinks should have 777
(rwxrwxrwx) permissions:

 $ umask
 0022
 $ touch file && ln -s file symlink; ls -l file symlink
 -rw-r--r--. 1 root root 0 Jun 23 16:29 file
 lrwxr-xr-x. 1 root root 4 Jun 23 16:29 symlink -> file

This fixes the bug by inserting a missing check that excludes
symlinks.

Link: https://lkml.kernel.org/r/1655974441-5612-1-git-send-email-konishi.ryusuke@gmail.com
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Reported-by: Tommy Pettersson <ptp@lysator.liu.se>
Reported-by: Ciprian Craciun <ciprian.craciun@gmail.com>
Tested-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs/remap: constrain dedupe of EOF blocks

commit 5750676b64a561f7ec920d7c6ba130fc9c7378f3 upstream.

If dedupe of an EOF block is not constrainted to match against only
other EOF blocks with the same EOF offset into the block, it can
match against any other block that has the same matching initial
bytes in it, even if the bytes beyond EOF in the source file do
not match.

Fix this by constraining the EOF block matching to only match
against other EOF blocks that have identical EOF offsets and data.
This allows "whole file dedupe" to continue to work without allowing
eof blocks to randomly match against partial full blocks with the
same data.

Reported-by: Ansgar Lößer <ansgar.loesser@tu-darmstadt.de>
Fixes: 1383a7ed6749 ("vfs: check file ranges before cloning files")
Link: https://lore.kernel.org/linux-fsdevel/a7c93559-4ba1-df2f-7a85-55a143696405@tu-darmstadt.de/
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/panfrost: Fix shrinker list corruption by madvise IOCTL

commit 9fc33eaaa979d112d10fea729edcd2a2e21aa912 upstream.

Calling madvise IOCTL twice on BO causes memory shrinker list corruption
and crashes kernel because BO is already on the list and it's added to
the list again, while BO should be removed from the list before it's
re-added. Fix it.

Cc: stable@vger.kernel.org
Fixes: 013b65101315 ("drm/panfrost: Add madvise and shrinker support")
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220630200601.1884120-3-dmitry.osipenko@collabora.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/panfrost: Put mapping instead of shmem obj on panfrost_mmu_map_fault_addr() error

commit fb6e0637ab7ebd8e61fe24f4d663c4bae99cfa62 upstream.

When panfrost_mmu_map_fault_addr() fails, the BO's mapping should be
unreferenced and not the shmem object which backs the mapping.

Cc: stable@vger.kernel.org
Fixes: bdefca2d8dc0 ("drm/panfrost: Add the panfrost_gem_mapping concept")
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220630200601.1884120-2-dmitry.osipenko@collabora.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: return -EAGAIN for NOWAIT dio reads/writes on compressed and inline extents

commit a4527e1853f8ff6e0b7c2dadad6268bd38427a31 upstream.

When doing a direct IO read or write, we always return -ENOTBLK when we
find a compressed extent (or an inline extent) so that we fallback to
buffered IO. This however is not ideal in case we are in a NOWAIT context
(io_uring for example), because buffered IO can block and we currently
have no support for NOWAIT semantics for buffered IO, so if we need to
fallback to buffered IO we should first signal the caller that we may
need to block by returning -EAGAIN instead.

This behaviour can also result in short reads being returned to user
space, which although it's not incorrect and user space should be able
to deal with partial reads, it's somewhat surprising and even some popular
applications like QEMU (Link tag #1) and MariaDB (Link tag #2) don't
deal with short reads properly (or at all).

The short read case happens when we try to read from a range that has a
non-compressed and non-inline extent followed by a compressed extent.
After having read the first extent, when we find the compressed extent we
return -ENOTBLK from btrfs_dio_iomap_begin(), which results in iomap to
treat the request as a short read, returning 0 (success) and waiting for
previously submitted bios to complete (this happens at
fs/iomap/direct-io.c:__iomap_dio_rw()). After that, and while at
btrfs_file_read_iter(), we call filemap_read() to use buffered IO to
read the remaining data, and pass it the number of bytes we were able to
read with direct IO. Than at filemap_read() if we get a page fault error
when accessing the read buffer, we return a partial read instead of an
-EFAULT error, because the number of bytes previously read is greater
than zero.

So fix this by returning -EAGAIN for NOWAIT direct IO when we find a
compressed or an inline extent.

Reported-by: Dominique MARTINET <dominique.martinet@atmark-techno.com>
Link: https://lore.kernel.org/linux-btrfs/YrrFGO4A1jS0GI0G@atmark-techno.com/
Link: https://jira.mariadb.org/browse/MDEV-27900?focusedCommentId=216582&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-216582
Tested-by: Dominique MARTINET <dominique.martinet@atmark-techno.com>
CC: stable@vger.kernel.org # 5.10+
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cgroup: Use separate src/dst nodes when preloading css_sets for migration

commit 07fd5b6cdf3cc30bfde8fe0f644771688be04447 upstream.

Each cset (css_set) is pinned by its tasks. When we're moving tasks around
across csets for a migration, we need to hold the source and destination
csets to ensure that they don't go away while we're moving tasks about. This
is done by linking cset->mg_preload_node on either the
mgctx->preloaded_src_csets or mgctx->preloaded_dst_csets list. Using the
same cset->mg_preload_node for both the src and dst lists was deemed okay as
a cset can't be both the source and destination at the same time.

Unfortunately, this overloading becomes problematic when multiple tasks are
involved in a migration and some of them are identity noop migrations while
others are actually moving across cgroups. For example, this can happen with
the following sequence on cgroup1:

 #1> mkdir -p /sys/fs/cgroup/misc/a/b
 #2> echo $$ > /sys/fs/cgroup/misc/a/cgroup.procs
 #3> RUN_A_COMMAND_WHICH_CREATES_MULTIPLE_THREADS &
 #4> PID=$!
 #5> echo $PID > /sys/fs/cgroup/misc/a/b/tasks
 #6> echo $PID > /sys/fs/cgroup/misc/a/cgroup.procs

the process including the group leader back into a. In this final migration,
non-leader threads would be doing identity migration while the group leader
is doing an actual one.

After #3, let's say the whole process was in cset A, and that after #4, the
leader moves to cset B. Then, during #6, the following happens:

 1. cgroup_migrate_add_src() is called on B for the leader.

 2. cgroup_migrate_add_src() is called on A for the other threads.

 3. cgroup_migrate_prepare_dst() is called. It scans the src list.

 4. It notices that B wants to migrate to A, so it tries to A to the dst
    list but realizes that its ->mg_preload_node is already busy.

 5. and then it notices A wants to migrate to A as it's an identity
    migration, it culls it by list_del_init()'ing its ->mg_preload_node and
    putting references accordingly.

 6. The rest of migration takes place with B on the src list but nothing on
    the dst list.

This means that A isn't held while migration is in progress. If all tasks
leave A before the migration finishes and the incoming task pins it, the
cset will be destroyed leading to use-after-free.

This is caused by overloading cset->mg_preload_node for both src and dst
preload lists. We wanted to exclude the cset from the src list but ended up
inadvertently excluding it from the dst list too.

This patch fixes the issue by separating out cset->mg_preload_node into
->mg_src_preload_node and ->mg_dst_preload_node, so that the src and dst
preloadings don't interfere with each other.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Mukesh Ojha <quic_mojha@quicinc.com>
Reported-by: shisiyuan <shisiyuan19870131@gmail.com>
Link: http://lkml.kernel.org/r/1654187688-27411-1-git-send-email-shisiyuan@xiaomi.com
Link: https://www.spinics.net/lists/cgroups/msg33313.html
Fixes: f817de98513d ("cgroup: prepare migration path for unified hierarchy")
Cc: stable@vger.kernel.org # v3.16+
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: mac80211: fix queue selection for mesh/OCB interfaces

commit 50e2ab39291947b6c6c7025cf01707c270fcde59 upstream.

When using iTXQ, the code assumes that there is only one vif queue for
broadcast packets, using the BE queue. Allowing non-BE queue marking
violates that assumption and txq->ac == skb_queue_mapping is no longer
guaranteed. This can cause issues with queue handling in the driver and
also causes issues with the recent ATF change, resulting in an AQL
underflow warning.

Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20220702145227.39356-1-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: 9214/1: alignment: advance IT state after emulating Thumb instruction

commit e5c46fde75e43c15a29b40e5fc5641727f97ae47 upstream.

After emulating a misaligned load or store issued in Thumb mode, we have
to advance the IT state by hand, or it will get out of sync with the
actual instruction stream, which means we'll end up applying the wrong
condition code to subsequent instructions. This might corrupt the
program state rather catastrophically.

So borrow the it_advance() helper from the probing code, and use it on
CPSR if the emulated instruction is Thumb.

Cc: <stable@vger.kernel.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: 9213/1: Print message about disabled Spectre workarounds only once

commit e4ced82deb5fb17222fb82e092c3f8311955b585 upstream.

Print the message about disabled Spectre workarounds only once. The
message is printed each time CPU goes out from idling state on NVIDIA
Tegra boards, causing storm in KMSG that makes system unusable.

Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ip: fix dflt addr selection for connected nexthop

commit 747c14307214b55dbd8250e1ab44cad8305756f1 upstream.

When a nexthop is added, without a gw address, the default scope was set
to 'host'. Thus, when a source address is selected, 127.0.0.1 may be chosen
but rejected when the route is used.

When using a route without a nexthop id, the scope can be configured in the
route, thus the problem doesn't exist.

To explain more deeply: when a user creates a nexthop, it cannot specify
the scope. To create it, the function nh_create_ipv4() calls fib_check_nh()
with scope set to 0. fib_check_nh() calls fib_check_nh_nongw() wich was
setting scope to 'host'. Then, nh_create_ipv4() calls
fib_info_update_nhc_saddr() with scope set to 'host'. The src addr is
chosen before the route is inserted.

When a 'standard' route (ie without a reference to a nexthop) is added,
fib_create_info() calls fib_info_update_nhc_saddr() with the scope set by
the user. iproute2 set the scope to 'link' by default.

Here is a way to reproduce the problem:
ip netns add foo
ip -n foo link set lo up
ip netns add bar
ip -n bar link set lo up
sleep 1

ip -n foo link add name eth0 type dummy
ip -n foo link set eth0 up
ip -n foo address add 192.168.0.1/24 dev eth0

ip -n foo link add name veth0 type veth peer name veth1 netns bar
ip -n foo link set veth0 up
ip -n bar link set veth1 up

ip -n bar address add 192.168.1.1/32 dev veth1
ip -n bar route add default dev veth1

ip -n foo nexthop add id 1 dev veth0
ip -n foo route add 192.168.1.1 nhid 1

Try to get/use the route:
> $ ip -n foo route get 192.168.1.1
> RTNETLINK answers: Invalid argument
> $ ip netns exec foo ping -c1 192.168.1.1
> ping: connect: Invalid argument

Try without nexthop group (iproute2 sets scope to 'link' by dflt):
ip -n foo route del 192.168.1.1
ip -n foo route add 192.168.1.1 dev veth0

Try to get/use the route:
> $ ip -n foo route get 192.168.1.1
> 192.168.1.1 dev veth0 src 192.168.0.1 uid 0
>     cache
> $ ip netns exec foo ping -c1 192.168.1.1
> PING 192.168.1.1 (192.168.1.1) 56(84) bytes of data.
> 64 bytes from 192.168.1.1: icmp_seq=1 ttl=64 time=0.039 ms
>
> --- 192.168.1.1 ping statistics ---
> 1 packets transmitted, 1 received, 0% packet loss, time 0ms
> rtt min/avg/max/mdev = 0.039/0.039/0.039/0.000 ms

CC: stable@vger.kernel.org
Fixes: 597cfe4fc339 ("nexthop: Add support for IPv4 nexthops")
Reported-by: Edwin Brossette <edwin.brossette@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Link: https://lore.kernel.org/r/20220713114853.29406-1-nicolas.dichtel@6wind.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: sock: tracing: Fix sock_exceed_buf_limit not to dereference stale pointer

commit 820b8963adaea34a87abbecb906d1f54c0aabfb7 upstream.

The trace event sock_exceed_buf_limit saves the prot->sysctl_mem pointer
and then dereferences it in the TP_printk() portion. This is unsafe as the
TP_printk() portion is executed at the time the buffer is read. That is,
it can be seconds, minutes, days, months, even years later. If the proto
is freed, then this dereference will can also lead to a kernel crash.

Instead, save the sysctl_mem array into the ring buffer and have the
TP_printk() reference that instead. This is the proper and safe way to
read pointers in trace events.

Link: https://lore.kernel.org/all/20220706052130.16368-12-kuniyu@amazon.com/
Cc: stable@vger.kernel.org
Fixes: 3847ce32aea9f ("core: add tracepoints for queueing skb to rcvbuf")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Acked-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tracing/histograms: Fix memory leak problem

commit 7edc3945bdce9c39198a10d6129377a5c53559c2 upstream.

This reverts commit 46bbe5c671e06f070428b9be142cc4ee5cedebac.

As commit 46bbe5c671e0 ("tracing: fix double free") said, the
"double free" problem reported by clang static analyzer is:
  > In parse_var_defs() if there is a problem allocating
  > var_defs.expr, the earlier var_defs.name is freed.
  > This free is duplicated by free_var_defs() which frees
  > the rest of the list.

However, if there is a problem allocating N-th var_defs.expr:
  + in parse_var_defs(), the freed 'earlier var_defs.name' is
    actually the N-th var_defs.name;
  + then in free_var_defs(), the names from 0th to (N-1)-th are freed;

                        IF ALLOCATING PROBLEM HAPPENED HERE!!! -+
                                                                 \
                                                                  |
          0th           1th                 (N-1)-th      N-th    V
          +-------------+-------------+-----+-------------+-----------
var_defs: | name | expr | name | expr | ... | name | expr | name | ///
          +-------------+-------------+-----+-------------+-----------

These two frees don't act on same name, so there was no "double free"
problem before. Conversely, after that commit, we get a "memory leak"
problem because the above "N-th var_defs.name" is not freed.

If enable CONFIG_DEBUG_KMEMLEAK and inject a fault at where the N-th
var_defs.expr allocated, then execute on shell like:
  $ echo 'hist:key=call_site:val=$v1,$v2:v1=bytes_req,v2=bytes_alloc' > \
/sys/kernel/debug/tracing/events/kmem/kmalloc/trigger

Then kmemleak reports:
  unreferenced object 0xffff8fb100ef3518 (size 8):
    comm "bash", pid 196, jiffies 4295681690 (age 28.538s)
    hex dump (first 8 bytes):
      76 31 00 00 b1 8f ff ff                          v1......
    backtrace:
      [<0000000038fe4895>] kstrdup+0x2d/0x60
      [<00000000c99c049a>] event_hist_trigger_parse+0x206f/0x20e0
      [<00000000ae70d2cc>] trigger_process_regex+0xc0/0x110
      [<0000000066737a4c>] event_trigger_write+0x75/0xd0
      [<000000007341e40c>] vfs_write+0xbb/0x2a0
      [<0000000087fde4c2>] ksys_write+0x59/0xd0
      [<00000000581e9cdf>] do_syscall_64+0x3a/0x80
      [<00000000cf3b065c>] entry_SYSCALL_64_after_hwframe+0x46/0xb0

Link: https://lkml.kernel.org/r/20220711014731.69520-1-zhengyejian1@huawei.com
Cc: stable@vger.kernel.org
Fixes: 46bbe5c671e0 ("tracing: fix double free")
Reported-by: Hulk Robot <hulkci@huawei.com>
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Reviewed-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Zheng Yejian <zhengyejian1@huawei.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm: split huge PUD on wp_huge_pud fallback

commit 14c99d65941538aa33edd8dc7b1bbbb593c324a2 upstream.

Currently the implementation will split the PUD when a fallback is taken
inside the create_huge_pud function.  This isn't where it should be done:
the splitting should be done in wp_huge_pud, just like it's done for PMDs.
Reason being that if a callback is taken during create, there is no PUD
yet so nothing to split, whereas if a fallback is taken when encountering
a write protection fault there is something to split.

It looks like this was the original intention with the commit where the
splitting was introduced, but somehow it got moved to the wrong place
between v1 and v2 of the patch series.  Rebase mistake perhaps.

Link: https://lkml.kernel.org/r/6f48d622eb8bce1ae5dd75327b0b73894a2ec407.camel@amazon.com
Fixes: 327e9fd48972 ("mm: Split huge pages on write-notify or COW")
Signed-off-by: James Gowans <jgowans@amazon.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Jan H. Schönherr <jschoenh@amazon.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm: userfaultfd: fix UFFDIO_CONTINUE on fallocated shmem pages

commit 73f37dbcfe1763ee2294c7717a1f571e27d17fd8 upstream.

When fallocate() is used on a shmem file, the pages we allocate can end up
with !PageUptodate.

Since UFFDIO_CONTINUE tries to find the existing page the user wants to
map with SGP_READ, we would fail to find such a page, since
shmem_getpage_gfp returns with a "NULL" pagep for SGP_READ if it discovers
!PageUptodate.  As a result, UFFDIO_CONTINUE returns -EFAULT, as it would
do if the page wasn't found in the page cache at all.

This isn't the intended behavior.  UFFDIO_CONTINUE is just trying to find
if a page exists, and doesn't care whether it still needs to be cleared or
not.  So, instead of SGP_READ, pass in SGP_NOALLOC.  This is the same,
except for one critical difference: in the !PageUptodate case, SGP_NOALLOC
will clear the page and then return it.  With this change, UFFDIO_CONTINUE
works properly (succeeds) on a shmem file which has been fallocated, but
otherwise not modified.

Link: https://lkml.kernel.org/r/20220610173812.1768919-1-axelrasmussen@google.com
Fixes: 153132571f02 ("userfaultfd/shmem: support UFFDIO_CONTINUE for shmem")
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Acked-by: Peter Xu <peterx@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fix race between exit_itimers() and /proc/pid/timers

commit d5b36a4dbd06c5e8e36ca8ccc552f679069e2946 upstream.

As Chris explains, the comment above exit_itimers() is not correct,
we can race with proc_timers_seq_ops. Change exit_itimers() to clear
signal->posix_timers with ->siglock held.

Cc: <stable@vger.kernel.org>
Reported-by: chris@accessvector.net
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xen/netback: avoid entering xenvif_rx_next_skb() with an empty rx queue

commit 94e8100678889ab428e68acadf042de723f094b9 upstream.

xenvif_rx_next_skb() is expecting the rx queue not being empty, but
in case the loop in xenvif_rx_action() is doing multiple iterations,
the availability of another skb in the rx queue is not being checked.

This can lead to crashes:

[40072.537261] BUG: unable to handle kernel NULL pointer dereference at 0000000000000080
[40072.537407] IP: xenvif_rx_skb+0x23/0x590 [xen_netback]
[40072.537534] PGD 0 P4D 0
[40072.537644] Oops: 0000 [#1] SMP NOPTI
[40072.537749] CPU: 0 PID: 12505 Comm: v1-c40247-q2-gu Not tainted 4.12.14-122.121-default #1 SLE12-SP5
[40072.537867] Hardware name: HP ProLiant DL580 Gen9/ProLiant DL580 Gen9, BIOS U17 11/23/2021
[40072.537999] task: ffff880433b38100 task.stack: ffffc90043d40000
[40072.538112] RIP: e030:xenvif_rx_skb+0x23/0x590 [xen_netback]
[40072.538217] RSP: e02b:ffffc90043d43de0 EFLAGS: 00010246
[40072.538319] RAX: 0000000000000000 RBX: ffffc90043cd7cd0 RCX: 00000000000000f7
[40072.538430] RDX: 0000000000000000 RSI: 0000000000000006 RDI: ffffc90043d43df8
[40072.538531] RBP: 000000000000003f R08: 000077ff80000000 R09: 0000000000000008
[40072.538644] R10: 0000000000007ff0 R11: 00000000000008f6 R12: ffffc90043ce2708
[40072.538745] R13: 0000000000000000 R14: ffffc90043d43ed0 R15: ffff88043ea748c0
[40072.538861] FS: 0000000000000000(0000) GS:ffff880484600000(0000) knlGS:0000000000000000
[40072.538988] CS: e033 DS: 0000 ES: 0000 CR0: 0000000080050033
[40072.539088] CR2: 0000000000000080 CR3: 0000000407ac8000 CR4: 0000000000040660
[40072.539211] Call Trace:
[40072.539319] xenvif_rx_action+0x71/0x90 [xen_netback]
[40072.539429] xenvif_kthread_guest_rx+0x14a/0x29c [xen_netback]

Fix that by stopping the loop in case the rx queue becomes empty.

Cc: stable@vger.kernel.org
Fixes: 98f6d57ced73 ("xen-netback: process guest rx packets in batches")
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Link: https://lore.kernel.org/r/20220713135322.19616-1-jgross@suse.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/realtek - Enable the headset-mic on a Xiaomi's laptop

commit 9b043a8f386485c74c0f8eea2c287d5bdbdf3279 upstream.

The headset on this machine is not defined, after applying the quirk
ALC256_FIXUP_ASUS_HEADSET_MIC, the headset-mic works well

Signed-off-by: Meng Tang <tangmeng@uniontech.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220713094133.9894-1-tangmeng@uniontech.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/realtek - Fix headset mic problem for a HP machine with alc221

commit 4ba5c853d7945b3855c3dcb293f7f9f019db641e upstream.

On a HP 288 Pro G2 MT (X9W02AV), the front mic could not be detected.
In order to get it working, the pin configuration needs to be set
correctly, and the ALC221_FIXUP_HP_288PRO_MIC_NO_PRESENCE fixup needs
to be applied.

Signed-off-by: Meng Tang <tangmeng@uniontech.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220713063332.30095-1-tangmeng@uniontech.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/realtek: fix mute/micmute LEDs for HP machines

commit 61d307855eb1a2ae849da445edd5389db8a58a5c upstream.

The HP ProBook 440/450 G9 and EliteBook 640/650 G9 have multiple
motherboard design and they are using different subsystem ID of audio
codec. Add the same quirk for other MBs.

Signed-off-by: Jeremy Szu <jeremy.szu@canonical.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220713022706.22892-1-jeremy.szu@canonical.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/realtek - Fix headset mic problem for a HP machine with alc671

commit dbe75d314748e08fc6e4576d153d8a69621ee5ca upstream.

On a HP 288 Pro G6, the front mic could not be detected.In order to
get it working, the pin configuration needs to be set correctly, and
the ALC671_FIXUP_HP_HEADSET_MIC2 fixup needs to be applied.

Signed-off-by: Meng Tang <tangmeng@uniontech.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220712092222.21738-1-tangmeng@uniontech.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/realtek: Fix headset mic for Acer SF313-51

commit 5f3fe25e70559fa3b096ab17e13316c93ddb7020 upstream.

The issue on Acer SWIFT SF313-51 is that headset microphone
doesn't work. The following quirk fixed headset microphone issue.
Note that the fixup of SF314-54/55 (ALC256_FIXUP_ACER_HEADSET_MIC)
was not successful on my SF313-51.

Signed-off-by: Meng Tang <tangmeng@uniontech.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220711081527.6254-1-tangmeng@uniontech.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/conexant: Apply quirk for another HP ProDesk 600 G3 model

commit d16d69bf5a25d91c6d8f3e29711be12551bf56cd upstream.

There is another HP ProDesk 600 G3 model with the PCI SSID 103c:82b4
that requires the quirk HP_MIC_NO_PRESENCE. Add the corresponding
entry to the quirk table.

Signed-off-by: Meng Tang <tangmeng@uniontech.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220711101744.25189-1-tangmeng@uniontech.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda - Add fixup for Dell Latitidue E5430

commit 841bdf85c226803a78a9319af9b2caa9bf3e2eda upstream.

Another Dell model, another fixup entry: Latitude E5430 needs the same
fixup as other Latitude E series as workaround for noise problems.

Signed-off-by: Meng Tang <tangmeng@uniontech.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220712060005.20176-1-tangmeng@uniontech.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 5.15.55

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "mtd: rawnand: gpmi: Fix setting busy timeout setting"

This reverts commit 0af674e7a764563496480c1e30fadf0048325978 which is
commit 06781a5026350cde699d2d10c9914a25c1524f45 upstream.

It is reported to cause data loss, so revert it to prevent that from
happening for users of this driver.

Reported-by: Tomasz Moń <tomasz.mon@camlingroup.com>
Reported-by: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/all/20220701110341.3094023-1-s.hauer@pengutronix.de/
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 5.15.54

Link: https://lore.kernel.org/r/20220711090604.055883544@linuxfoundation.org
Link: https://lore.kernel.org/r/20220711145306.494277196@linuxfoundation.org
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Link: https://lore.kernel.org/r/20220712071513.420542604@linuxfoundation.org
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

selftests/net: fix section name when using xdp_dummy.o

commit d28b25a62a47a8c8aa19bd543863aab6717e68c9 upstream.

Since commit 8fffa0e3451a ("selftests/bpf: Normalize XDP section names in
selftests") the xdp_dummy.o's section name has changed to xdp. But some
tests are still using "section xdp_dummy", which make the tests failed.
Fix them by updating to the new section name.

Fixes: 8fffa0e3451a ("selftests/bpf: Normalize XDP section names in selftests")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/r/20220630062228.3453016-1-liuhangbin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: idxd: force wq context cleanup on device disable path

commit 44c4237cf3436bda2b185ff728123651ad133f69 upstream.

Testing shown that when a wq mode is setup to be dedicated and then torn
down and reconfigured to shared, the wq configured end up being dedicated
anyays. The root cause is when idxd_device_wqs_clear_state() gets called
during idxd_driver removal, idxd_wq_disable_cleanup() does not get called
vs when the wq driver is removed first. The check of wq state being
"enabled" causes the cleanup to be bypassed. However, idxd_driver->remove()
releases all wq drivers. So the wqs goes to "disabled" state and will never
be "enabled". By that point, the driver has no idea if the wq was
previously configured or clean. So force call idxd_wq_disable_cleanup() on
all wqs always to make sure everything gets cleaned up.

Reported-by: Tony Zhu <tony.zhu@intel.com>
Tested-by: Tony Zhu <tony.zhu@intel.com>
Fixes: 0dcfe41e9a4c ("dmanegine: idxd: cleanup all device related bits after disabling device")
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Co-developed-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: https://lore.kernel.org/r/20220628230056.2527816-1-fenghua.yu@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: ti: Add missing put_device in ti_dra7_xbar_route_allocate

commit 615a4bfc426e11dba05c2cf343f9ac752fb381d2 upstream.

of_find_device_by_node() takes reference, we should use put_device()
to release it when not need anymore.

Fixes: a074ae38f859 ("dmaengine: Add driver for TI DMA crossbar on DRA7x")
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Acked-by: Peter Ujfalusi <peter.ujfalusi@gmail.com>
Link: https://lore.kernel.org/r/20220605042723.17668-1-linmq006@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: qcom: bam_dma: fix runtime PM underflow

commit 0ac9c3dd0d6fe293cd5044cfad10bec27d171e4e upstream.

Commit dbad41e7bb5f ("dmaengine: qcom: bam_dma: check if the runtime pm enabled")
caused unbalanced pm_runtime_get/put() calls when the bam is
controlled remotely. This commit reverts it and just enables pm_runtime
in all cases, the clk_* functions already just nop when the clock is NULL.

Also clean up a bit by removing unnecessary bamclk null checks.

Suggested-by: Stephan Gerhold <stephan@gerhold.net>
Fixes: dbad41e7bb5f ("dmaengine: qcom: bam_dma: check if the runtime pm enabled")
Signed-off-by: Caleb Connolly <caleb.connolly@linaro.org>
Link: https://lore.kernel.org/r/20220629140559.118537-1-caleb.connolly@linaro.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: ti: Fix refcount leak in ti_dra7_xbar_route_allocate

commit c132fe78ad7b4ce8b5d49a501a15c29d08eeb23a upstream.

of_parse_phandle() returns a node pointer with refcount
incremented, we should use of_node_put() on it when not needed anymore.

Add missing of_node_put() in to fix this.

Fixes: ec9bfa1e1a79 ("dmaengine: ti-dma-crossbar: dra7: Use bitops instead of idr")
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Link: https://lore.kernel.org/r/20220605042723.17668-2-linmq006@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: at_xdma: handle errors of at_xdmac_alloc_desc() correctly

commit 3770d92bd5237d686e49da7b2fb86f53ee6ed259 upstream.

It seems that it is valid to have less than the requested number of
descriptors. But what is not valid and leads to subsequent errors is to
have zero descriptors. In that case, abort the probing.

Fixes: e1f7c9eee707 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver")
Signed-off-by: Michael Walle <michael@walle.cc>
Link: https://lore.kernel.org/r/20220526135111.1470926-1-michael@walle.cc
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: lgm: Fix an error handling path in intel_ldma_probe()

commit 1dbe67b9faea0bc340cce894018076679c16cb71 upstream.

ldma_clk_disable() calls both:
clk_disable_unprepare(d->core_clk);
reset_control_assert(d->rst);

So, should devm_reset_control_get_optional() fail, core_clk should not
be prepare_enable'd before it, otherwise it will never be
disable_unprepare'd.

Reorder the code to handle the error handling path as expected.

Fixes: 32d31c79a1a4 ("dmaengine: Add Intel LGM SoC DMA support.")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/18504549bc4d2b62a72a02cb22a2e4d8e6a58720.1653241224.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: pl330: Fix lockdep warning about non-static key

commit b64b3b2f1d81f83519582e1feee87d77f51f5f17 upstream.

The DEFINE_SPINLOCK() macro shouldn't be used for dynamically allocated
spinlocks. The lockdep warns about this and disables locking validator.
Fix the warning by making lock static.

 INFO: trying to register non-static key.
 The code is fine but needs lockdep annotation, or maybe
 you didn't initialize this object before use?
 turning off the locking correctness validator.
 Hardware name: Radxa ROCK Pi 4C (DT)
 Call trace:
  dump_backtrace.part.0+0xcc/0xe0
  show_stack+0x18/0x6c
  dump_stack_lvl+0x8c/0xb8
  dump_stack+0x18/0x34
  register_lock_class+0x4a8/0x4cc
  __lock_acquire+0x78/0x20cc
  lock_acquire.part.0+0xe0/0x230
  lock_acquire+0x68/0x84
  _raw_spin_lock_irqsave+0x84/0xc4
  add_desc+0x44/0xc0
  pl330_get_desc+0x15c/0x1d0
  pl330_prep_dma_cyclic+0x100/0x270
  snd_dmaengine_pcm_trigger+0xec/0x1c0
  dmaengine_pcm_trigger+0x18/0x24
  ...

Fixes: e588710311ee ("dmaengine: pl330: fix descriptor allocation fail")
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Link: https://lore.kernel.org/r/20220520181432.149904-1-dmitry.osipenko@collabora.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ida: don't use BUG_ON() for debugging

commit fc82bbf4dede758007763867d0282353c06d1121 upstream.

This is another old BUG_ON() that just shouldn't exist (see also commit
a382f8fee42c: "signal handling: don't use BUG_ON() for debugging").

In fact, as Matthew Wilcox points out, this condition shouldn't really
even result in a warning, since a negative id allocation result is just
a normal allocation failure:

  "I wonder if we should even warn here -- sure, the caller is trying to
   free something that wasn't allocated, but we don't warn for
   kfree(NULL)"

and goes on to point out how that current error check is only causing
people to unnecessarily do their own index range checking before freeing
it.

This was noted by Itay Iellin, because the bluetooth HCI socket cookie
code does *not* do that range checking, and ends up just freeing the
error case too, triggering the BUG_ON().

The HCI code requires CAP_NET_RAW, and seems to just result in an ugly
splat, but there really is no reason to BUG_ON() here, and we have
generally striven for allocation models where it's always ok to just do

    free(alloc());

even if the allocation were to fail for some random reason (usually
obviously that "random" reason being some resource limit).

Fixes: 88eca0207cf1 ("ida: simplified functions for id allocation")
Reported-by: Itay Iellin <ieitayie@gmail.com>
Suggested-by: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dt-bindings: dma: allwinner,sun50i-a64-dma: Fix min/max typo

commit 607a48c78e6b427b0b684d24e61c19e846ad65d6 upstream.

The conditional block for variants with a second clock should have set
minItems, not maxItems, which was already 2. Since clock-names requires
two items, this typo should not have caused any problems.

Fixes: edd14218bd66 ("dt-bindings: dmaengine: Convert Allwinner A31 and A64 DMA to a schema")
Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220702031903.21703-1-samuel@sholland.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "serial: 8250_mtk: Make sure to select the right FEATURE_SEL"

commit f0136f65285bcfb7e8f90d1013723076a35acd51 upstream.

It was found that some MediaTek SoCs are incompatible with this
change. Also, this register was mistakenly understood as it was
related to the 16550A register layout selection but, at least
on some IPs, if not all, it's related to something else unknown.

This reverts commit 6f81fdded0d024c7d4084d434764f30bca1cd6b1.

Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Fixes: 6f81fdded0d0 ("serial: 8250_mtk: Make sure to select the right FEATURE_SEL")
Reported-by: "kernelci.org bot" <bot@kernelci.org>
Link: https://lore.kernel.org/r/20220510122620.150342-1-angelogioacchino.delregno@collabora.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "mm/memory-failure.c: fix race with changing page compound again"

commit 2ba2b008a8bf5fd268a43d03ba79e0ad464d6836 upstream.

Reverts commit 888af2701db7 ("mm/memory-failure.c: fix race with changing
page compound again") because now we fetch the page refcount under
hugetlb_lock in try_memory_failure_hugetlb() so that the race check is no
longer necessary.

Link: https://lkml.kernel.org/r/20220408135323.1559401-4-naoya.horiguchi@linux.dev
Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Suggested-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx_usb: set return value in rsp_buf alloc err path

commit 2cd37c2e72449a7add6da1183d20a6247d6db111 upstream.

Set return value in rsp_buf alloc error path before going to
error handling.

drivers/misc/cardreader/rtsx_usb.c:639:6: warning: variable 'ret' is used uninitialized whenever 'if' condition is true [-Wsometimes-uninitialized]
           if (!ucr->rsp_buf)
               ^~~~~~~~~~~~~
   drivers/misc/cardreader/rtsx_usb.c:678:9: note: uninitialized use occurs here
           return ret;
                  ^~~
   drivers/misc/cardreader/rtsx_usb.c:639:2: note: remove the 'if' if its condition is always false
           if (!ucr->rsp_buf)
           ^~~~~~~~~~~~~~~~~~
   drivers/misc/cardreader/rtsx_usb.c:622:9: note: initialize the variable 'ret' to silence this warning
           int ret;
                  ^
                   = 0

Fixes: 3776c7855985 ("misc: rtsx_usb: use separate command and response buffers")
Reported-by: kernel test robot <lkp@intel.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://lore.kernel.org/r/20220701165352.15687-1-skhan@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx_usb: use separate command and response buffers

commit 3776c78559853fd151be7c41e369fd076fb679d5 upstream.

rtsx_usb uses same buffer for command and response. There could
be a potential conflict using the same buffer for both especially
if retries and timeouts are involved.

Use separate command and response buffers to avoid conflicts.

Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Cc: stable <stable@kernel.org>
Link: https://lore.kernel.org/r/07e3721804ff07aaab9ef5b39a5691d0718b9ade.1656642167.git.skhan@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx_usb: fix use of dma mapped buffer for usb bulk transfer

commit eb7f8e28420372787933eec079735c35034bda7d upstream.

rtsx_usb driver allocates coherent dma buffer for urb transfers.
This buffer is passed to usb_bulk_msg() and usb core tries to
map already mapped buffer running into a dma mapping error.

xhci_hcd 0000:01:00.0: rejecting DMA map of vmalloc memory
WARNING: CPU: 1 PID: 279 at include/linux/dma-mapping.h:326 usb_ hcd_map_urb_for_dma+0x7d6/0x820

...

xhci_map_urb_for_dma+0x291/0x4e0
usb_hcd_submit_urb+0x199/0x12b0
...
usb_submit_urb+0x3b8/0x9e0
usb_start_wait_urb+0xe3/0x2d0
usb_bulk_msg+0x115/0x240
rtsx_usb_transfer_data+0x185/0x1a8 [rtsx_usb]
rtsx_usb_send_cmd+0xbb/0x123 [rtsx_usb]
rtsx_usb_write_register+0x12c/0x143 [rtsx_usb]
rtsx_usb_probe+0x226/0x4b2 [rtsx_usb]

Fix it to use kmalloc() to get DMA-able memory region instead.

Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Cc: stable <stable@kernel.org>
Link: https://lore.kernel.org/r/667d627d502e1ba9ff4f9b94966df3299d2d3c0d.1656642167.git.skhan@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: imx-sdma: Allow imx8m for imx7 FW revs

commit a7cd3cf0b2e5aaacfe5e02c472bd28e98e640be7 upstream.

The revision of the imx-sdma IP that is in the i.MX8M series is the
same is that as that in the i.MX7 series but the imx7d MODULE_FIRMWARE
directive is wrapped in a condiditional which means it's not defined
when built for aarch64 SOC_IMX8M platforms and hence you get the
following errors when the driver loads on imx8m devices:

imx-sdma 302c0000.dma-controller: Direct firmware load for imx/sdma/sdma-imx7d.bin failed with error -2
imx-sdma 302c0000.dma-controller: external firmware not found, using ROM firmware

Add the SOC_IMX8M into the check so the firmware can load on i.MX8.

Fixes: 1474d48bd639 ("arm64: dts: imx8mq: Add SDMA nodes")
Fixes: 941acd566b18 ("dmaengine: imx-sdma: Only check ratio on parts that support 1:1")
Signed-off-by: Peter Robinson <pbrobinson@gmail.com>
Cc: stable@vger.kernel.org # v5.2+
Reviewed-by: Fabio Estevam <festevam@gmail.com>
Link: https://lore.kernel.org/r/20220606161034.3544803-1-pbrobinson@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

i2c: cadence: Unregister the clk notifier in error path

[ Upstream commit 3501f0c663063513ad604fb1b3f06af637d3396d ]

This patch ensures that the clock notifier is unregistered
when driver probe is returning error.

Fixes: df8eb5691c48 ("i2c: Add driver for Cadence I2C controller")
Signed-off-by: Satish Nagireddy <satish.nagireddy@getcruise.com>
Tested-by: Lars-Peter Clausen <lars@metafoo.de>
Reviewed-by: Michal Simek <michal.simek@amd.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

r8169: fix accessing unset transport header

[ Upstream commit faa4e04e5e140a6d02260289a8fba8fd8d7a3003 ]

66e4c8d95008 ("net: warn if transport header was not set") added
a check that triggers a warning in r8169, see [0].

The commit referenced in the Fixes tag refers to the change from
which the patch applies cleanly, there's nothing wrong with this
commit. It seems the actual issue (not bug, because the warning
is harmless here) was introduced with bdfa4ed68187
("r8169: use Giant Send").

[0] https://bugzilla.kernel.org/show_bug.cgi?id=216157

Fixes: 8d520b4de3ed ("r8169: work around RTL8125 UDP hw bug")
Reported-by: Erhard F. <erhard_f@mailbox.org>
Tested-by: Erhard F. <erhard_f@mailbox.org>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/1b2c2b29-3dc0-f7b6-5694-97ec526d51a0@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: forwarding: fix error message in learning_test

[ Upstream commit 83844aacab2015da1dba1df0cc61fc4b4c4e8076 ]

When packets are not received, they aren't received on $host1_if, so the
message talking about the second host not receiving them is incorrect.
Fix it.

Fixes: d4deb01467ec ("selftests: forwarding: Add a test for FDB learning")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: forwarding: fix learning_test when h1 supports IFF_UNICAST_FLT

[ Upstream commit 1a635d3e1c80626237fdae47a5545b6655d8d81c ]

The first host interface has by default no interest in receiving packets
MAC DA de:ad:be:ef:13:37, so it might drop them before they hit the tc
filter and this might confuse the selftest.

Enable promiscuous mode such that the filter properly counts received
packets.

Fixes: d4deb01467ec ("selftests: forwarding: Add a test for FDB learning")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: forwarding: fix flood_unicast_test when h2 supports IFF_UNICAST_FLT

[ Upstream commit b8e629b05f5d23f9649c901bef09fab8b0c2e4b9 ]

As mentioned in the blamed commit, flood_unicast_test() works by
checking the match count on a tc filter placed on the receiving
interface.

But the second host interface (host2_if) has no interest in receiving a
packet with MAC DA de:ad:be:ef:13:37, so its RX filter drops it even
before the ingress tc filter gets to be executed. So we will incorrectly
get the message "Packet was not flooded when should", when in fact, the
packet was flooded as expected but dropped due to an unrelated reason,
at some other layer on the receiving side.

Force h2 to accept this packet by temporarily placing it in promiscuous
mode. Alternatively we could either deliver to its MAC address or use
tcpdump_start, but this has the fewest complications.

This fixes the "flooding" test from bridge_vlan_aware.sh and
bridge_vlan_unaware.sh, which calls flood_test from the lib.

Fixes: 236dd50bf67a ("selftests: forwarding: Add a test for flooded traffic")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ibmvnic: Properly dispose of all skbs during a failover.

[ Upstream commit 1b18f09d31cfa7148df15a7d5c5e0e86f105f7d1 ]

During a reset, there may have been transmits in flight that are no
longer valid and cannot be fulfilled.  Resetting and clearing the
queues is insufficient; each skb also needs to be explicitly freed
so that upper levels are not left waiting for confirmation of a
transmit that will never happen.  If this happens frequently enough,
the apparent backlog will cause TCP to begin "congestion control"
unnecessarily, culminating in permanently decreased throughput.

Fixes: d7c0ef36bde03 ("ibmvnic: Free and re-allocate scrqs when tx/rx scrqs change")
Tested-by: Nick Child <nnac123@linux.ibm.com>
Reviewed-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Rick Lindsley <ricklind@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: stm32: add missing usbh clock and fix clk order on stm32mp15

[ Upstream commit 1d0c1aadf1fd9f3de95d1532b3651e8634546e71 ]

The USBH composed of EHCI and OHCI controllers needs the PHY clock to be
initialized first, before enabling (gating) them. The reverse is also
required when going to suspend.
So, add USBPHY clock as 1st entry in both controllers, so the USBPHY PLL
gets enabled 1st upon controller init. Upon suspend/resume, this also makes
the clock to be disabled/re-enabled in the correct order.
This fixes some IRQ storm conditions seen when going to low-power, due to
PHY PLL being disabled before all clocks are cleanly gated.

Fixes: 949a0c0dec85 ("ARM: dts: stm32: add USB Host (USBH) support to stm32mp157c")
Fixes: db7be2cb87ae ("ARM: dts: stm32: use usbphyc ck_usbo_48m as USBH OHCI clock on stm32mp151")
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@foss.st.com>
Signed-off-by: Alexandre Torgue <alexandre.torgue@foss.st.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: stm32: use usbphyc ck_usbo_48m as USBH OHCI clock on stm32mp151

[ Upstream commit db7be2cb87ae65e2d033a9f61f7fb94bce505177 ]

Referring to the note under USBH reset and clocks chapter of RM0436,
"In order to access USBH_OHCI registers it is necessary to activate the USB
clocks by enabling the PLL controlled by USBPHYC" (ck_usbo_48m).

The point is, when USBPHYC PLL is not enabled, OHCI register access
freezes the resume from STANDBY. It is the case when dual USBH is enabled,
instead of OTG + single USBH.
When OTG is probed, as ck_usbo_48m is USBO clock parent, then USBPHYC PLL
is enabled and OHCI register access is OK.

This patch adds ck_usbo_48m (provided by USBPHYC PLL) as clock of USBH
OHCI, thus USBPHYC PLL will be enabled and OHCI register access will be OK.

Signed-off-by: Amelie Delaunay <amelie.delaunay@foss.st.com>
Signed-off-by: Alexandre Torgue <alexandre.torgue@foss.st.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i40e: Fix VF's MAC Address change on VM

[ Upstream commit fed0d9f13266a22ce1fc9a97521ef9cdc6271a23 ]

Clear VF MAC from parent PF and remove VF filter from VSI when both
conditions are true:
-VIRTCHNL_VF_OFFLOAD_USO is not used
-VM MAC was not set from PF level

It affects older version of IAVF and it allow them to change MAC
Address on VM, newer IAVF won't change their behaviour.

Previously it wasn't possible to change VF's MAC Address on VM
because there is flag on IAVF driver that won't allow to
change MAC Address if this address is given from PF driver.

Fixes: 155f0ac2c96b ("iavf: allow permanent MAC address to change")
Signed-off-by: Norbert Zulinski <norbertx.zulinski@intel.com>
Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i40e: Fix dropped jumbo frames statistics

[ Upstream commit 1adb1563e7b7ec659379a18e607e8bc3522d8a78 ]

Dropped packets caused by too large frames were not included in
dropped RX packets statistics.
Issue was caused by not reading the GL_RXERR1 register. That register
stores count of packet which was have been dropped due to too large
size.

Fix it by reading GL_RXERR1 register for each interface.

Repro steps:
Send a packet larger than the set MTU to SUT
Observe rx statists: ethtool -S <interface> | grep rx | grep -v ": 0"

Fixes: 41a9e55c89be ("i40e: add missing VSI statistics")
Signed-off-by: Lukasz Cieplicki <lukaszx.cieplicki@intel.com>
Signed-off-by: Jedrzej Jagielski <jedrzej.jagielski@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i2c: piix4: Fix a memory leak in the EFCH MMIO support

[ Upstream commit 8ad59b397f86a4d8014966fdc0552095a0c4fb2b ]

The recently added support for EFCH MMIO regions introduced a memory
leak in that code path. The leak is caused by the fact that
release_resource() merely removes the resource from the tree but does
not free its memory. We need to call release_mem_region() instead,
which does free the memory. As a nice side effect, this brings back
some symmetry between the legacy and MMIO paths.

Signed-off-by: Jean Delvare <jdelvare@suse.de>
Reported-by: Yi Zhang <yi.zhang@redhat.com>
Tested-by: Yi Zhang <yi.zhang@redhat.com>
Reviewed-by: Terry Bowman <terry.bowman@amd.com>
Tested-by: Terry Bowman <Terry.Bowman@amd.com>
Fixes: 7c148722d074 ("i2c: piix4: Add EFCH MMIO support to region request and release")
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

xsk: Clear page contiguity bit when unmapping pool

[ Upstream commit 512d1999b8e94a5d43fba3afc73e774849674742 ]

When a XSK pool gets mapped, xp_check_dma_contiguity() adds bit 0x1
to pages' DMA addresses that go in ascending order and at 4K stride.

The problem is that the bit does not get cleared before doing unmap.
As a result, a lot of warnings from iommu_dma_unmap_page() are seen
in dmesg, which indicates that lookups by iommu_iova_to_phys() fail.

Fixes: 2b43470add8c ("xsk: Introduce AF_XDP buffer allocation API")
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/bpf/20220628091848.534803-1-ivan.malov@oktetlabs.ru
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: at91: fix soc detection for SAM9X60 SiPs

[ Upstream commit 35074df65a8d8c5328a83e2eea948f7bbc8e6e08 ]

Fix SoC detection for SAM9X60 SiPs:
SAM9X60D5M
SAM9X60D1G
SAM9X60D6K

Fixes: af3a10513cd6 ("drivers: soc: atmel: add per soc id and version match masks")
Signed-off-by: Mihai Sain <mihai.sain@microchip.com>
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20220616081344.1978664-1-claudiu.beznea@microchip.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: at91: sama5d2_icp: fix eeprom compatibles

[ Upstream commit 416ce193d73a734ded6d09fe141017b38af1c567 ]

The eeprom memories on the board are microchip 24aa025e48, which are 2 Kbits
and are compatible with at24c02 not at24c32.

Fixes: 68a95ef72cefe ("ARM: dts: at91: sama5d2-icp: add SAMA5D2-ICP")
Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Reviewed-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20220607090455.80433-2-eugen.hristev@microchip.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: at91: sam9x60ek: fix eeprom compatible and size

[ Upstream commit f2cbbc3f926316ccf8ef9363d8a60c1110afc1c7 ]

The board has a microchip 24aa025e48 eeprom, which is a 2 Kbits memory,
so it's compatible with at24c02 not at24c32.
Also the size property is wrong, it's not 128 bytes, but 256 bytes.
Thus removing and leaving it to the default (256).

Fixes: 1e5f532c27371 ("ARM: dts: at91: sam9x60: add device tree for soc and board")
Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Reviewed-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20220607090455.80433-1-eugen.hristev@microchip.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: at91: pm: use proper compatibles for sama7g5's rtc and rtt

[ Upstream commit 1c40169b35ad58906814d53a517ac92db3d20d5f ]

Use proper compatible strings for SAMA7G5's RTC and RTT IPs. These are
necessary for configuring wakeup sources for ULP1 PM mode.

Fixes: 6501330f9f5e ("ARM: at91: pm: add pm support for SAMA7G5")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20220523092421.317345-4-claudiu.beznea@microchip.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: at91: pm: use proper compatibles for sam9x60's rtc and rtt

[ Upstream commit 641522665dbb25ce117c78746df1aad8b58c80e5 ]

Use proper compatible strings for SAM9X60's RTC and RTT IPs. These are
necessary for configuring wakeup sources for ULP1 PM mode.

Fixes: eaedc0d379da ("ARM: at91: pm: add ULP1 support for SAM9X60")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20220523092421.317345-3-claudiu.beznea@microchip.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: at91: pm: use proper compatible for sama5d2's rtc

[ Upstream commit ddc980da8043779119acaca106c6d9b445c9b65b ]

Use proper compatible strings for SAMA5D2's RTC IPs. This is necessary
for configuring wakeup sources for ULP1 PM mode.

Fixes: d7484f5c6b3b ("ARM: at91: pm: configure wakeup sources for ULP1 mode")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20220523092421.317345-2-claudiu.beznea@microchip.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: qcom: msm8992-*: Fix vdd_lvs1_2-supply typo

[ Upstream commit 5fb779558f1c97e2bf2794cb59553e569c38e2f9 ]

"make dtbs_check" complains about the missing "-supply" suffix for
vdd_lvs1_2 which is clearly a typo, originally introduced in the
msm8994-smd-rpm.dtsi file and apparently later copied to
msm8992-xiaomi-libra.dts:

msm8992-lg-bullhead-rev-10/101.dtb: pm8994-regulators: 'vdd_lvs1_2'
does not match any of the regexes:
  '.*-supply$', '^((s|l|lvs|5vs)[0-9]*)|(boost-bypass)|(bob)$', 'pinctrl-[0-9]+'
From schema: regulator/qcom,smd-rpm-regulator.yaml

msm8992-xiaomi-libra.dtb: pm8994-regulators: 'vdd_lvs1_2'
does not match any of the regexes:
  '.*-supply$', '^((s|l|lvs|5vs)[0-9]*)|(boost-bypass)|(bob)$', 'pinctrl-[0-9]+'
From schema: regulator/qcom,smd-rpm-regulator.yaml

Reported-by: Rob Herring <robh@kernel.org>
Cc: Konrad Dybcio <konrad.dybcio@somainline.org>
Fixes: f3b2c99e73be ("arm64: dts: Enable onboard SDHCI on msm8992")
Fixes: 0f5cdb31e850 ("arm64: dts: qcom: Add Xiaomi Libra (Mi 4C) device tree")
Signed-off-by: Stephan Gerhold <stephan.gerhold@kernkonzept.com>
Reviewed-by: Konrad Dybcio <konrad.dybcio@somainline.org>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/20220627135938.2901871-1-stephan.gerhold@kernkonzept.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

pinctrl: sunxi: sunxi_pconf_set: use correct offset

[ Upstream commit cd4c1e65a32afd003b08ad4aafe1e4d3e4e8e61b ]

Some Allwinner SoCs have 2 pinctrls (PIO and R_PIO).
Previous implementation used absolute pin numbering and it was incorrect
for R_PIO pinctrl.
It's necessary to take into account the base pin number.

Fixes: 90be64e27621 ("pinctrl: sunxi: implement pin_config_set")
Signed-off-by: Andrei Lalaev <andrey.lalaev@gmail.com>
Reviewed-by: Samuel Holland <samuel@sholland.org>
Link: https://lore.kernel.org/r/20220525190423.410609-1-andrey.lalaev@gmail.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: imx8mp-phyboard-pollux-rdk: correct i2c2 & mmc settings

[ Upstream commit 242d8ee9111171a6e68249aaff62643c513be6ec ]

BIT3 and BIT0 are reserved bits, should not touch.

Fixes: 88f7f6bcca37 ("arm64: dts: freescale: Add support for phyBOARD-Pollux-i.MX8MP")
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: imx8mp-phyboard-pollux-rdk: correct eqos pad settings

[ Upstream commit bae4de618efe1c41d34aa2e6cef8b08e46256667 ]

BIT3 and BIT0 are reserved bits, should not touch.

Fixes: 6f96852619d5 ("arm64: dts: freescale: Add support EQOS MAC on phyBOARD-Pollux-i.MX8MP")
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: imx8mp-phyboard-pollux-rdk: correct uart pad settings

[ Upstream commit e266c155bd88e95f9b86379d6b0add6ac6e5452e ]

BIT3 and BIT0 are reserved bits, should not touch.

Fixes: 846f752866bd ("arm64: dts: imx8mp-phyboard-pollux-rdk: Change debug UART")
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: imx8mp-evk: correct I2C3 pad settings

[ Upstream commit 0836de513ebaae5f03014641eac996290d67493d ]

According to RM bit layout, BIT3 and BIT0 are reserved.
 8  7   6   5   4   3  2 1  0
PE HYS PUE ODE FSEL X  DSE  X

Although function is not broken, we should not set reserved bit.

Fixes: 5e4a67ff7f69 ("arm64: dts: imx8mp-evk: Add i2c3 support")
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>