vfio/pds: Fix possible sleep while in atomic context

[ Upstream commit ae2667cd8a479bb5abd6e24c12fcc9ef5bc06d75 ]

The driver could possibly sleep while in atomic context resulting
in the following call trace while CONFIG_DEBUG_ATOMIC_SLEEP=y is
set:

BUG: sleeping function called from invalid context at kernel/locking/mutex.c:283
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 2817, name: bash
preempt_count: 1, expected: 0
RCU nest depth: 0, expected: 0
Call Trace:
 <TASK>
 dump_stack_lvl+0x36/0x50
 __might_resched+0x123/0x170
 mutex_lock+0x1e/0x50
 pds_vfio_put_lm_file+0x1e/0xa0 [pds_vfio_pci]
 pds_vfio_put_save_file+0x19/0x30 [pds_vfio_pci]
 pds_vfio_state_mutex_unlock+0x2e/0x80 [pds_vfio_pci]
 pci_reset_function+0x4b/0x70
 reset_store+0x5b/0xa0
 kernfs_fop_write_iter+0x137/0x1d0
 vfs_write+0x2de/0x410
 ksys_write+0x5d/0xd0
 do_syscall_64+0x3b/0x90
 entry_SYSCALL_64_after_hwframe+0x6e/0xd8

This can happen if pds_vfio_put_restore_file() and/or
pds_vfio_put_save_file() grab the mutex_lock(&lm_file->lock)
while the spin_lock(&pds_vfio->reset_lock) is held, which can
happen during while calling pds_vfio_state_mutex_unlock().

Fix this by changing the reset_lock to reset_mutex so there are no such
conerns. Also, make sure to destroy the reset_mutex in the driver specific
VFIO device release function.

This also fixes a spinlock bad magic BUG that was caused
by not calling spinlock_init() on the reset_lock. Since, the lock is
being changed to a mutex, make sure to call mutex_init() on it.

Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Closes: https://lore.kernel.org/kvm/1f9bc27b-3de9-4891-9687-ba2820c1b390@moroto.mountain/
Fixes: bb500dbe2ac6 ("vfio/pds: Add VFIO live migration support")
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20231122192532.25791-3-brett.creeley@amd.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

vfio/pds: Fix mutex lock->magic != lock warning

[ Upstream commit 91aeb563bd4332e2988f8c0f64f125c4ecb5bcb3 ]

The following BUG was found when running on a kernel with
CONFIG_DEBUG_MUTEXES=y set:

DEBUG_LOCKS_WARN_ON(lock->magic != lock)
RIP: 0010:mutex_trylock+0x10d/0x120
Call Trace:
 <TASK>
 ? __warn+0x85/0x140
 ? mutex_trylock+0x10d/0x120
 ? report_bug+0xfc/0x1e0
 ? handle_bug+0x3f/0x70
 ? exc_invalid_op+0x17/0x70
 ? asm_exc_invalid_op+0x1a/0x20
 ? mutex_trylock+0x10d/0x120
 ? mutex_trylock+0x10d/0x120
 pds_vfio_reset+0x3a/0x60 [pds_vfio_pci]
 pci_reset_function+0x4b/0x70
 reset_store+0x5b/0xa0
 kernfs_fop_write_iter+0x137/0x1d0
 vfs_write+0x2de/0x410
 ksys_write+0x5d/0xd0
 do_syscall_64+0x3b/0x90
 entry_SYSCALL_64_after_hwframe+0x6e/0xd8

As shown, lock->magic != lock. This is because
mutex_init(&pds_vfio->state_mutex) is called in the VFIO open path. So,
if a reset is initiated before the VFIO device is opened the mutex will
have never been initialized. Fix this by calling
mutex_init(&pds_vfio->state_mutex) in the VFIO init path.

Also, don't destroy the mutex on close because the device may
be re-opened, which would cause mutex to be uninitialized. Fix this by
implementing a driver specific vfio_device_ops.release callback that
destroys the mutex before calling vfio_pci_core_release_dev().

Fixes: bb500dbe2ac6 ("vfio/pds: Add VFIO live migration support")
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20231122192532.25791-2-brett.creeley@amd.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Fix MPCC 1DLUT programming

[ Upstream commit 6f395cebdd8927fbffdc3a55a14fcacf93634359 ]

[Why]
Wrong function is used to translate LUT values to HW format, leading to
visible artifacting in some cases.

[How]
Use the correct cm3_helper function.

Cc: stable@vger.kernel.org # 6.1+
Reviewed-by: Krunoslav Kovac <krunoslav.kovac@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Ilya Bakoulin <ilya.bakoulin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Simplify brightness initialization

[ Upstream commit d9e865826c202b262f9ee3f17a03cc4ac5d44ced ]

[Why]
Remove the brightness cache in DC. It uses a single value to represent
the brightness for both SDR and HDR mode. This leads to flash in HDR
on/off. It also unconditionally programs brightness as in HDR mode. This
may introduce garbage on SDR mode in miniLED panel.

[How]
Simplify the initialization flow by removing the DC cache and taking
what panel has as default. Expand the mechanism for PWM to DPCD Aux to
restore cached brightness value generally.

Cc: stable@vger.kernel.org # 6.1+
Reviewed-by: Krunoslav Kovac <krunoslav.kovac@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Camille Cho <camille.cho@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Reduce default backlight min from 5 nits to 1 nits

[ Upstream commit 5edb7cdff85af8f8c5fda5b88310535ab823f663 ]

[Why & How]
Currently set_default_brightness_aux function uses 5 nits as lower limit
to check for valid default_backlight setting. However some newer panels
can support even lower default settings

Reviewed-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Acked-by: Roman Li <roman.li@amd.com>
Signed-off-by: Swapnil Patel <swapnil.patel@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Stable-dep-of: d9e865826c20 ("drm/amd/display: Simplify brightness initialization")
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: refactor ILR to make it work

[ Upstream commit 6ec876472ff7edeaf2a07bf6afbff74d7f1dfa35 ]

[Why]
Current ILR toggle is on/off as a part of panel
config for new function, which breaks original
ILR logic

[How]
Refactor ILR and take panel config into account

Reviewed-by: Anthony Koo <anthony.koo@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Sherry Wang <yao.wang1@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Stable-dep-of: d9e865826c20 ("drm/amd/display: Simplify brightness initialization")
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu: Fix printk arg in of_iommu_get_resv_regions()

[ Upstream commit c2183b3dcc9dd41b768569ea88bededa58cceebb ]

The variable phys is defined as (struct resource *) which aligns with
the printk format specifier %pr. Taking the address of it results in a
value of type (struct resource **) which is incompatible with the format
specifier %pr. Therefore, remove the address of operator (&).

Fixes: a5bf3cfce8cb ("iommu: Implement of_iommu_get_resv_regions()")
Signed-off-by: Daniel Mentz <danielmentz@google.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20231108062226.928985-1-danielmentz@google.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/pm: fix a memleak in aldebaran_tables_init

[ Upstream commit 7a88f23e768491bae653b444a96091d2aaeb0818 ]

When kzalloc() for smu_table->ecc_table fails, we should free
the previously allocated resources to prevent memleak.

Fixes: edd794208555 ("drm/amd/pm: add message smu to get ecc_table v2")
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

cpufreq/amd-pstate: Only print supported EPP values for performance governor

[ Upstream commit 142c169b31beb364ef39385b4e88735bd51d37fe ]

show_energy_performance_available_preferences() to show only supported
values which is performance in performance governor policy.

-------Before--------
$ cat /sys/devices/system/cpu/cpu1/cpufreq/scaling_driver
amd-pstate-epp
$ cat /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
performance
$ cat /sys/devices/system/cpu/cpu1/cpufreq/energy_performance_preference
performance
$ cat /sys/devices/system/cpu/cpu1/cpufreq/energy_performance_available_preferences
default performance balance_performance balance_power power

-------After--------
$ cat /sys/devices/system/cpu/cpu1/cpufreq/scaling_driver
amd-pstate-epp
$ cat /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
performance
$ cat /sys/devices/system/cpu/cpu1/cpufreq/energy_performance_preference
performance
$ cat /sys/devices/system/cpu/cpu1/cpufreq/energy_performance_available_preferences
performance

Fixes: ffa5096a7c33 ("cpufreq: amd-pstate: implement Pstate EPP support for the AMD processors")
Suggested-by: Wyes Karny <wyes.karny@amd.com>
Signed-off-by: Ayush Jain <ayush.jain3@amd.com>
Reviewed-by: Wyes Karny <wyes.karny@amd.com>
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

cpufreq/amd-pstate: Fix scaling_min_freq and scaling_max_freq update

[ Upstream commit febab20caebac959fdc3d7520bc52de8b1184455 ]

When amd_pstate is running, writing to scaling_min_freq and
scaling_max_freq has no effect. These values are only passed to the
policy level, but not to the platform level. This means that the
platform does not know about the frequency limits set by the user.

To fix this, update the min_perf and max_perf values at the platform
level whenever the user changes the scaling_min_freq and scaling_max_freq
values.

Fixes: ffa5096a7c33 ("cpufreq: amd-pstate: implement Pstate EPP support for the AMD processors")
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Wyes Karny <wyes.karny@amd.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/panel: nt36523: fix return value check in nt36523_probe()

[ Upstream commit fb18fe0fdf22a2f4512a8b644bb5ea1473829cda ]

mipi_dsi_device_register_full() never returns NULL pointer, it
will return ERR_PTR() when it fails, so replace the check with
IS_ERR().

Fixes: 0993234a0045 ("drm/panel: Add driver for Novatek NT36523")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/20231129090715.856263-1-yangyingliang@huaweicloud.com
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20231129090715.856263-1-yangyingliang@huaweicloud.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/panel: starry-2081101qfh032011-53g: Fine tune the panel power sequence

[ Upstream commit fc1ccc16271a0526518f19f460fed63d575a8a42 ]

For the "starry, 2081101qfh032011-53g" panel, it is stipulated in the
panel spec that MIPI needs to keep the LP11 state before the
lcm_reset pin is pulled high.

Fixes: 6069b66cd962 ("drm/panel: support for STARRY 2081101QFH032011-53G MIPI-DSI panel")
Signed-off-by: xiazhengqiao <xiazhengqiao@huaqin.corp-partner.google.com>
Reviewed-by: Jessica Zhang <quic_jesszhan@quicinc.com>
Link: https://lore.kernel.org/r/20231129084115.7918-1-xiazhengqiao@huaqin.corp-partner.google.com
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20231129084115.7918-1-xiazhengqiao@huaqin.corp-partner.google.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/i915/gsc: Mark internal GSC engine with reserved uabi class

[ Upstream commit 503579448db93f9fbcc93cd99a1f2d5aa4b2cda6 ]

The GSC CS is not exposed to the user, so we skipped assigning a uabi
class number for it. However, the trace logs use the uabi class and
instance to identify the engine, so leaving uabi class unset makes the
GSC CS show up as the RCS in those logs.

Given that the engine is not exposed to the user, we can't add a new
case in the uabi enum, so we insted internally define a kernel
internal class as -1.

At the same time remove special handling for the name and complete
the uabi_classes array so internal class is automatically correctly
assigned.

Engine will show as 65535:0 other0 in the logs/traces which should
be unique enough.

v2:
 * Fix uabi class u8 vs u16 type confusion.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: 194babe26bdc ("drm/i915/mtl: don't expose GSC command streamer to the user")
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Alan Previn <alan.previn.teres.alexis@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231116084456.291533-1-tvrtko.ursulin@linux.intel.com
(cherry picked from commit dfed6b58d54f3a5d7e6bc1fb060e2c936330eba2)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu/vt-d: Make context clearing consistent with context mapping

[ Upstream commit 9a16ab9d640274b20813d2d17475e18d3e99d834 ]

In the iommu probe_device path, domain_context_mapping() allows setting
up the context entry for a non-PCI device. However, in the iommu
release_device path, domain_context_clear() only clears context entries
for PCI devices.

Make domain_context_clear() behave consistently with
domain_context_mapping() by clearing context entries for both PCI and
non-PCI devices.

Fixes: 579305f75d34 ("iommu/vt-d: Update to use PCI DMA aliases")
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20231114011036.70142-4-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu/vt-d: Disable PCI ATS in legacy passthrough mode

[ Upstream commit da37dddcf4caf015c400a930301d2ee27a7a15fb ]

When IOMMU hardware operates in legacy mode, the TT field of the context
entry determines the translation type, with three supported types (Section
9.3 Context Entry):

- DMA translation without device TLB support
- DMA translation with device TLB support
- Passthrough mode with translated and translation requests blocked

Device TLB support is absent when hardware is configured in passthrough
mode.

Disable the PCI ATS feature when IOMMU is configured for passthrough
translation type in legacy (non-scalable) mode.

Fixes: 0faa19a1515f ("iommu/vt-d: Decouple PASID & PRI enabling from SVA")
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20231114011036.70142-3-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu/vt-d: Omit devTLB invalidation requests when TES=0

[ Upstream commit 0f5432a9b839847dcfe9fa369d72e3d646102ddf ]

The latest VT-d spec indicates that when remapping hardware is disabled
(TES=0 in Global Status Register), upstream ATS Invalidation Completion
requests are treated as UR (Unsupported Request).

Consequently, the spec recommends in section 4.3 Handling of Device-TLB
Invalidations that software refrain from submitting any Device-TLB
invalidation requests when address remapping hardware is disabled.

Verify address remapping hardware is enabled prior to submitting Device-
TLB invalidation requests.

Fixes: 792fb43ce2c9 ("iommu/vt-d: Enable Intel IOMMU scalable mode by default")
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20231114011036.70142-2-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

cpufreq: imx6q: Don't disable 792 Mhz OPP unnecessarily

[ Upstream commit 2e4e0984c7d696cc74cf2fd7e7f62997f0e9ebe6 ]

For a 900MHz i.MX6ULL CPU the 792MHz OPP is disabled. There is no
convincing reason to disable this OPP. If a CPU can run at 900MHz,
it should also be able to cope with 792MHz. Looking at the voltage
level of 792MHz in [1] (page 24, table 10. "Operating Ranges") the
current defined OPP is above the minimum. So the voltage level
shouldn't be a problem. However in [2] (page 24, table 10.
"Operating Ranges"), it is not mentioned that 792MHz OPP isn't
allowed. Change it to only disable 792MHz OPP for i.MX6ULL types
below 792 MHz.

[1] https://www.nxp.com/docs/en/data-sheet/IMX6ULLIEC.pdf
[2] https://www.nxp.com/docs/en/data-sheet/IMX6ULLCEC.pdf

Fixes: 0aa9abd4c212 ("cpufreq: imx6q: check speed grades for i.MX6ULL")
Signed-off-by: Christoph Niedermaier <cniedermaier@dh-electronics.com>
Reviewed-by: Marek Vasut <marex@denx.de>
Reviewed-by: Fabio Estevam <festevam@denx.de>
[ Viresh: Edited subject ]
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Remove power sequencing check

[ Upstream commit b0399e22ada096435de3e3e73899aa8bc026820d ]

[Why]
Some ASICs keep backlight powered on after dpms off
command has been issued.

[How]
The check for no edp power sequencing was never going to pass.
The value is never changed from what it is set by design.

Cc: stable@vger.kernel.org # 6.1+
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2765
Reviewed-by: Swapnil Patel <swapnil.patel@amd.com>
Acked-by: Roman Li <roman.li@amd.com>
Signed-off-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Refactor edp power control

[ Upstream commit 45f98fccb1f6895f527bd5f811f23478c2f920f5 ]

[Why & How]
To organize the edp power control a bit:

1. add flag in dc_link to indicate dc to skip all implicit eDP power control.
2. add edp_set_panel_power link service for DM to call.

Reviewed-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Jun Lei <jun.lei@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Ian Chen <ian.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Stable-dep-of: b0399e22ada0 ("drm/amd/display: Remove power sequencing check")
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/cmma: fix handling of swapper_pg_dir and invalid_pg_dir

[ Upstream commit 84bb41d5df48868055d159d9247b80927f1f70f9 ]

If the cmma no-dat feature is available the kernel page tables are walked
to identify and mark all pages which are used for address translation (all
region, segment, and page tables). In a subsequent loop all other pages are
marked as "no-dat" pages with the ESSA instruction.

This information is visible to the hypervisor, so that the hypervisor can
optimize purging of guest TLB entries. All pages used for swapper_pg_dir
and invalid_pg_dir are incorrectly marked as no-dat, which in turn can
result in incorrect guest TLB flushes.

Fix this by marking those pages correctly as being used for DAT.

Cc: <stable@vger.kernel.org>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/pseries/iommu: enable_ddw incorrectly returns direct mapping for SR-IOV device

[ Upstream commit 3bf983e4e93ce8e6d69e9d63f52a66ec0856672e ]

When a device is initialized, the driver invokes dma_supported() twice -
first for streaming mappings followed by coherent mappings. For an
SR-IOV device, default window is deleted and DDW created. With vPMEM
enabled, TCE mappings are dynamically created for both vPMEM and SR-IOV
device.  There are no direct mappings.

First time when dma_supported() is called with 64 bit mask, DDW is created
and marked as dynamic window. The second time dma_supported() is called,
enable_ddw() finds existing window for the device and incorrectly returns
it as "direct mapping".

This only happens when size of DDW is big enough to map max LPAR memory.

This results in streaming TCEs to not get dynamically mapped, since code
incorrently assumes these are already pre-mapped. The adapter initially
comes up but goes down due to EEH.

Fixes: 381ceda88c4c ("powerpc/pseries/iommu: Make use of DDW for indirect mapping")
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Gaurav Batra <gbatra@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/20231003030802.47914-1-gbatra@linux.vnet.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ravb: Keep reverse order of operations in ravb_remove()

[ Upstream commit edf9bc396e05081ca281ffb0cd41e44db478ff26 ]

On RZ/G3S SMARC Carrier II board having RGMII connections b/w Ethernet
MACs and PHYs it has been discovered that doing unbind/bind for ravb
driver in a loop leads to wrong speed and duplex for Ethernet links and
broken connectivity (the connectivity cannot be restored even with
bringing interface down/up). Before doing unbind/bind the Ethernet
interfaces were configured though systemd. The sh instructions used to
do unbind/bind were:

$ cd /sys/bus/platform/drivers/ravb/
$ while :; do echo 11c30000.ethernet > unbind ; \
  echo 11c30000.ethernet > bind; done

It has been discovered that there is a race b/w IOCTLs initialized by
systemd at the response of success binding and the
"ravb_write(ndev, CCC_OPC_RESET, CCC)" call in ravb_remove() as
follows:

1/ as a result of bind success the user space open/configures the
   interfaces tough an IOCTL; the following stack trace has been
   identified on RZ/G3S:

Call trace:
dump_backtrace+0x9c/0x100
show_stack+0x20/0x38
dump_stack_lvl+0x48/0x60
dump_stack+0x18/0x28
ravb_open+0x70/0xa58
__dev_open+0xf4/0x1e8
__dev_change_flags+0x198/0x218
dev_change_flags+0x2c/0x80
devinet_ioctl+0x640/0x708
inet_ioctl+0x1e4/0x200
sock_do_ioctl+0x50/0x108
sock_ioctl+0x240/0x358
__arm64_sys_ioctl+0xb0/0x100
invoke_syscall+0x50/0x128
el0_svc_common.constprop.0+0xc8/0xf0
do_el0_svc+0x24/0x38
el0_svc+0x34/0xb8
el0t_64_sync_handler+0xc0/0xc8
el0t_64_sync+0x190/0x198

2/ this call may execute concurrently with ravb_remove() as the
   unbind/bind operation was executed in a loop
3/ if the operation mode is changed to RESET (through
   ravb_write(ndev, CCC_OPC_RESET, CCC) call in ravb_remove())
   while the above ravb_open() is in progress it may lead to MAC
   (or PHY, or MAC-PHY connection, the right point hasn't been identified
   at the moment) to be broken, thus the Ethernet connectivity fails to
   restore.

The simple fix for this is to move ravb_write(ndev, CCC_OPC_RESET, CCC))
after unregister_netdev() to avoid resetting the controller while the
netdev interface is still registered.

To avoid future issues in ravb_remove(), the patch follows the proper order
of operations in ravb_remove(): reverse order compared with ravb_probe().
This avoids described races as the IOCTLs as well as unregister_netdev()
(called now at the beginning of ravb_remove()) calls rtnl_lock() before
continuing and IOCTLs check (though devinet_ioctl()) if device is still
registered just after taking the lock:

int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
{
// ...

        rtnl_lock();

        ret = -ENODEV;
        dev = __dev_get_by_name(net, ifr->ifr_name);
        if (!dev)
                goto done;

// ...
done:
        rtnl_unlock();
out:
        return ret;
}

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ravb: Stop DMA in case of failures on ravb_open()

[ Upstream commit eac16a733427ba0de2449ffc7bd3da32ddb65cb7 ]

In case ravb_phy_start() returns with error the settings applied in
ravb_dmac_init() are not reverted (e.g. config mode). For this call
ravb_stop_dma() on failure path of ravb_open().

Fixes: a0d2f20650e8 ("Renesas Ethernet AVB PTP clock driver")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ravb: Start TX queues after HW initialization succeeded

[ Upstream commit 6f32c086602050fc11157adeafaa1c1eb393f0af ]

ravb_phy_start() may fail. If that happens, the TX queues will remain
started. Thus, move the netif_tx_start_all_queues() after PHY is
successfully initialized.

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ravb: Make write access to CXR35 first before accessing other EMAC registers

[ Upstream commit d78c0ced60d5e2f8b5a4a0468a5c400b24aeadf2 ]

Hardware manual of RZ/G3S (and RZ/G2L) specifies the following on the
description of CXR35 register (chapter "PHY interface select register
(CXR35)"): "After release reset, make write-access to this register before
making write-access to other registers (except MDIOMOD). Even if not need
to change the value of this register, make write-access to this register
at least one time. Because RGMII/MII MODE is recognized by accessing this
register".

The setup procedure for EMAC module (chapter "Setup procedure" of RZ/G3S,
RZ/G2L manuals) specifies the E-MAC.CXR35 register is the first EMAC
register that is to be configured.

Note [A] from chapter "PHY interface select register (CXR35)" specifies
the following:
[A] The case which CXR35 SEL_XMII is used for the selection of RGMII/MII
in APB Clock 100 MHz.
(1) To use RGMII interface, Set ‘H’03E8_0000’ to this register.
(2) To use MII interface, Set ‘H’03E8_0002’ to this register.

Take into account these indication.

Fixes: 1089877ada8d ("ravb: Add RZ/G2L MII interface support")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ravb: Use pm_runtime_resume_and_get()

[ Upstream commit 88b74831faaee455c2af380382d979fc38e79270 ]

pm_runtime_get_sync() may return an error. In case it returns with an error
dev->power.usage_count needs to be decremented. pm_runtime_resume_and_get()
takes care of this. Thus use it.

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ravb: Check return value of reset_control_deassert()

[ Upstream commit d8eb6ea4b302e7ff78535c205510e359ac10a0bd ]

reset_control_deassert() could return an error. Some devices cannot work
if reset signal de-assert operation fails. To avoid this check the return
code of reset_control_deassert() in ravb_probe() and take proper action.

Along with it, the free_netdev() call from the error path was moved after
reset_control_assert() on its own label (out_free_netdev) to free
netdev in case reset_control_deassert() fails.

Fixes: 0d13a1a464a0 ("ravb: Add reset support")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: Fix VF Reset paths when interface in a failed over aggregate

[ Upstream commit 9f74a3dfcf83e11aedcb98250b8040dbc6d9659a ]

There is an error when an interface has the following conditions:
- PF is in an aggregate (bond)
- PF has VFs created on it
- bond is in a state where it is failed-over to the secondary interface
- A VF reset is issued on one or more of those VFs

The issue is generated by the originating PF trying to rebuild or
reconfigure the VF resources.  Since the bond is failed over to the
secondary interface the queue contexts are in a modified state.

To fix this issue, have the originating interface reclaim its resources
prior to the tear-down and rebuild or reconfigure.  Then after the process
is complete, move the resources back to the currently active interface.

There are multiple paths that can be used depending on what triggered the
event, so create a helper function to move the queues and use paired calls
to the helper (back to origin, process, then move back to active interface)
under the same lag_mutex lock.

Fixes: 1e0f9881ef79 ("ice: Flesh out implementation of support for SRIOV on bonded interface")
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Tested-by: Sujai Buvaneswaran <sujai.buvaneswaran@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Link: https://lore.kernel.org/r/20231127212340.1137657-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf, sockmap: af_unix stream sockets need to hold ref for pair sock

[ Upstream commit 8866730aed5100f06d3d965c22f1c61f74942541 ]

AF_UNIX stream sockets are a paired socket. So sending on one of the pairs
will lookup the paired socket as part of the send operation. It is possible
however to put just one of the pairs in a BPF map. This currently increments
the refcnt on the sock in the sockmap to ensure it is not free'd by the
stack before sockmap cleans up its state and stops any skbs being sent/recv'd
to that socket.

But we missed a case. If the peer socket is closed it will be free'd by the
stack. However, the paired socket can still be referenced from BPF sockmap
side because we hold a reference there. Then if we are sending traffic through
BPF sockmap to that socket it will try to dereference the free'd pair in its
send logic creating a use after free. And following splat:

   [59.900375] BUG: KASAN: slab-use-after-free in sk_wake_async+0x31/0x1b0
   [59.901211] Read of size 8 at addr ffff88811acbf060 by task kworker/1:2/954
   [...]
   [59.905468] Call Trace:
   [59.905787]  <TASK>
   [59.906066]  dump_stack_lvl+0x130/0x1d0
   [59.908877]  print_report+0x16f/0x740
   [59.910629]  kasan_report+0x118/0x160
   [59.912576]  sk_wake_async+0x31/0x1b0
   [59.913554]  sock_def_readable+0x156/0x2a0
   [59.914060]  unix_stream_sendmsg+0x3f9/0x12a0
   [59.916398]  sock_sendmsg+0x20e/0x250
   [59.916854]  skb_send_sock+0x236/0xac0
   [59.920527]  sk_psock_backlog+0x287/0xaa0

To fix let BPF sockmap hold a refcnt on both the socket in the sockmap and its
paired socket. It wasn't obvious how to contain the fix to bpf_unix logic. The
primarily problem with keeping this logic in bpf_unix was: In the sock close()
we could handle the deref by having a close handler. But, when we are destroying
the psock through a map delete operation we wouldn't have gotten any signal
thorugh the proto struct other than it being replaced. If we do the deref from
the proto replace its too early because we need to deref the sk_pair after the
backlog worker has been stopped.

Given all this it seems best to just cache it at the end of the psock and eat 8B
for the af_unix and vsock users. Notice dgram sockets are OK because they handle
locking already.

Fixes: 94531cfcbe79 ("af_unix: Add unix_stream_proto for sockmap")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20231129012557.95371-2-john.fastabend@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ethtool: don't propagate EOPNOTSUPP from dumps

[ Upstream commit cbeb989e41f4094f54bec2cecce993f26f547bea ]

The default dump handler needs to clear ret before returning.
Otherwise if the last interface returns an inconsequential
error this error will propagate to user space.

This may confuse user space (ethtool CLI seems to ignore it,
but YNL doesn't). It will also terminate the dump early
for mutli-skb dump, because netlink core treats EOPNOTSUPP
as a real error.

Fixes: 728480f12442 ("ethtool: default handlers for GET requests")
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20231126225806.2143528-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ravb: Fix races between ravb_tx_timeout_work() and net related ops

[ Upstream commit 9870257a0a338cd8d6c1cddab74e703f490f6779 ]

Fix races between ravb_tx_timeout_work() and functions of net_device_ops
and ethtool_ops by using rtnl_trylock() and rtnl_unlock(). Note that
since ravb_close() is under the rtnl lock and calls cancel_work_sync(),
ravb_tx_timeout_work() should calls rtnl_trylock(). Otherwise, a deadlock
may happen in ravb_tx_timeout_work() like below:

CPU0 CPU1
ravb_tx_timeout()
schedule_work()
...
__dev_close_many()
// Under rtnl lock
ravb_close()
cancel_work_sync()
// Waiting
ravb_tx_timeout_work()
rtnl_lock()
// This is possible to cause a deadlock

If rtnl_trylock() fails, rescheduling the work with sleep for 1 msec.

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Link: https://lore.kernel.org/r/20231127122420.3706751-1-yoshihiro.shimoda.uh@renesas.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

r8169: prevent potential deadlock in rtl8169_close

[ Upstream commit 91d3d149978ba7b238198dd80e4b823756aa7cfa ]

ndo_stop() is RTNL-protected by net core, and the worker function takes
RTNL as well. Therefore we will deadlock when trying to execute a
pending work synchronously. To fix this execute any pending work
asynchronously. This will do no harm because netif_running() is false
in ndo_stop(), and therefore the work function is effectively a no-op.
However we have to ensure that no task is running or pending after
rtl_remove_one(), therefore add a call to cancel_work_sync().

Fixes: abe5fc42f9ce ("r8169: use RTNL to protect critical sections")
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/12395867-1d17-4cac-aa7d-c691938fcddf@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

efi/unaccepted: Fix off-by-one when checking for overlapping ranges

[ Upstream commit 01b1e3ca0e5ce47bbae8217d47376ad01b331b07 ]

When a task needs to accept memory it will scan the accepting_list
to see if any ranges already being processed by other tasks overlap
with its range. Due to an off-by-one in the range comparisons, a task
might falsely determine that an overlapping range is being accepted,
leading to an unnecessary delay before it begins processing the range.

Fix the off-by-one in the range comparison to prevent this and slightly
improve performance.

Fixes: 50e782a86c98 ("efi/unaccepted: Fix soft lockups caused by parallel memory acceptance")
Link: https://lore.kernel.org/linux-mm/20231101004523.vseyi5bezgfaht5i@amd.com/T/#me2eceb9906fcae5fe958b3fe88e41f920f8335b6
Reviewed-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

neighbour: Fix __randomize_layout crash in struct neighbour

[ Upstream commit 45b3fae4675dc1d4ee2d7aefa19d85ee4f891377 ]

Previously, one-element and zero-length arrays were treated as true
flexible arrays, even though they are actually "fake" flex arrays.
The __randomize_layout would leave them untouched at the end of the
struct, similarly to proper C99 flex-array members.

However, this approach changed with commit 1ee60356c2dc ("gcc-plugins:
randstruct: Only warn about true flexible arrays"). Now, only C99
flexible-array members will remain untouched at the end of the struct,
while one-element and zero-length arrays will be subject to randomization.

Fix a `__randomize_layout` crash in `struct neighbour` by transforming
zero-length array `primary_key` into a proper C99 flexible-array member.

Fixes: 1ee60356c2dc ("gcc-plugins: randstruct: Only warn about true flexible arrays")
Closes: https://lore.kernel.org/linux-hardening/20231124102458.GB1503258@e124191.cambridge.arm.com/
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Tested-by: Joey Gouly <joey.gouly@arm.com>
Link: https://lore.kernel.org/r/ZWJoRsJGnCPdJ3+2@work
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-pf: Restore TC ingress police rules when interface is up

[ Upstream commit fd7f98b2e12a3d96a92bde6640657ec7116f4372 ]

TC ingress policer rules depends on interface receive queue
contexts since the bandwidth profiles are attached to RQ
contexts. When an interface is brought down all the queue
contexts are freed. This in turn frees bandwidth profiles in
hardware causing ingress police rules non-functional after
the interface is brought up. Fix this by applying all the ingress
police rules config to hardware in otx2_open. Also allow
adding ingress rules only when interface is running
since no contexts exist for the interface when it is down.

Fixes: 68fbff68dbea ("octeontx2-pf: Add police action for TC flower")
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Link: https://lore.kernel.org/r/1700930217-5707-1-git-send-email-sbhatta@marvell.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-pf: Fix adding mbox work queue entry when num_vfs > 64

[ Upstream commit 51597219e0cd5157401d4d0ccb5daa4d9961676f ]

When more than 64 VFs are enabled for a PF then mbox communication
between VF and PF is not working as mbox work queueing for few VFs
are skipped due to wrong calculation of VF numbers.

Fixes: d424b6c02415 ("octeontx2-pf: Enable SRIOV and added VF mbox handling")
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Link: https://lore.kernel.org/r/1700930042-5400-1-git-send-email-sbhatta@marvell.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: stmmac: xgmac: Disable FPE MMC interrupts

[ Upstream commit e54d628a2721bfbb002c19f6e8ca6746cec7640f ]

Commit aeb18dd07692 ("net: stmmac: xgmac: Disable MMC interrupts
by default") tries to disable MMC interrupts to avoid a storm of
unhandled interrupts, but leaves the FPE(Frame Preemption) MMC
interrupts enabled, FPE MMC interrupts can cause the same problem.
Now we mask FPE TX and RX interrupts to disable all MMC interrupts.

Fixes: aeb18dd07692 ("net: stmmac: xgmac: Disable MMC interrupts by default")
Reviewed-by: Larysa Zaremba <larysa.zaremba@intel.com>
Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Serge Semin <fancer.lancer@gmail.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Link: https://lore.kernel.org/r/20231125060126.2328690-1-0x1207@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-af: Fix possible buffer overflow

[ Upstream commit ad31c629ca3c87f6d557488c1f9faaebfbcd203c ]

A loop in rvu_mbox_handler_nix_bandprof_free() contains
a break if (idx == MAX_BANDPROF_PER_PFFUNC),
but if idx may reach MAX_BANDPROF_PER_PFFUNC
buffer '(*req->prof_idx)[layer]' overflow happens before that check.

The patch moves the break to the
beginning of the loop.

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

Fixes: e8e095b3b370 ("octeontx2-af: cn10k: Bandwidth profiles config support").
Signed-off-by: Elena Salomatkina <elena.salomatkina.cmc@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Subbaraya Sundeep <sbhatta@marvell.com>
Link: https://lore.kernel.org/r/20231124210802.109763-1-elena.salomatkina.cmc@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/net: mptcp: fix uninitialized variable warnings

[ Upstream commit 00a4f8fd9c750f20d8fd4535c71c9caa7ef5ff2f ]

Same init_rng() in both tests. The function reads /dev/urandom to
initialize srand(). In case of failure, it falls back onto the
entropy in the uninitialized variable. Not sure if this is on purpose.
But failure reading urandom should be rare, so just fail hard. While
at it, convert to getrandom(). Which man 4 random suggests is simpler
and more robust.

    mptcp_inq.c:525:6:
    mptcp_connect.c:1131:6:

    error: variable 'foo' is used uninitialized
    whenever 'if' condition is false
    [-Werror,-Wsometimes-uninitialized]

Fixes: 048d19d444be ("mptcp: add basic kselftest for mptcp")
Fixes: b51880568f20 ("selftests: mptcp: add inq test case")
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: Willem de Bruijn <willemb@google.com>
----

When input is randomized because this is expected to meaningfully
explore edge cases, should we also add
1. logging the random seed to stdout and
2. adding a command line argument to replay from a specific seed
I can do this in net-next, if authors find it useful in this case.
Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Link: https://lore.kernel.org/r/20231124171645.1011043-5-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/net: unix: fix unused variable compiler warning

[ Upstream commit 59fef379d453781f0dabfa1f1a1e86e78aee919a ]

Remove an unused variable.

    diag_uid.c:151:24:
    error: unused variable 'udr'
    [-Werror,-Wunused-variable]

Fixes: ac011361bd4f ("af_unix: Add test for sock_diag and UDIAG_SHOW_UID.")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/20231124171645.1011043-4-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/net: fix a char signedness issue

[ Upstream commit 7b29828c5af6841bdeb9fafa32fdfeff7ab9c407 ]

Signedness of char is signed on x86_64, but unsigned on arm64.

Fix the warning building cmsg_sender.c on signed platforms or
forced with -fsigned-char:

    msg_sender.c:455:12:
    error: implicit conversion from 'int' to 'char'
    changes value from 128 to -128
    [-Werror,-Wconstant-conversion]
        buf[0] = ICMPV6_ECHO_REQUEST;

constant ICMPV6_ECHO_REQUEST is 128.

Link: https://lwn.net/Articles/911914
Fixes: de17e305a810 ("selftests: net: cmsg_sender: support icmp and raw sockets")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/20231124171645.1011043-3-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests/net: ipsec: fix constant out of range

[ Upstream commit 088559815477c6f623a5db5993491ddd7facbec7 ]

Fix a small compiler warning.

nr_process must be a signed long: it is assigned a signed long by
strtol() and is compared against LONG_MIN and LONG_MAX.

ipsec.c:2280:65:
    error: result of comparison of constant -9223372036854775808
    with expression of type 'unsigned int' is always false
    [-Werror,-Wtautological-constant-out-of-range-compare]

  if ((errno == ERANGE && (nr_process == LONG_MAX || nr_process == LONG_MIN))

Fixes: bc2652b7ae1e ("selftest/net/xfrm: Add test for ipsec tunnel")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmail.com>
Link: https://lore.kernel.org/r/20231124171645.1011043-2-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

uapi: propagate __struct_group() attributes to the container union

[ Upstream commit 4e86f32a13af1970d21be94f659cae56bbe487ee ]

Recently the kernel test robot has reported an ARM-specific BUILD_BUG_ON()
in an old and unmaintained wil6210 wireless driver. The problem comes from
the structure packing rules of old ARM ABI ('-mabi=apcs-gnu'). For example,
the following structure is packed to 18 bytes instead of 16:

struct poorly_packed {
        unsigned int a;
        unsigned int b;
        unsigned short c;
        union {
                struct {
                        unsigned short d;
                        unsigned int e;
                } __attribute__((packed));
                struct {
                        unsigned short d;
                        unsigned int e;
                } __attribute__((packed)) inner;
        };
} __attribute__((packed));

To fit it into 16 bytes, it's required to add packed attribute to the
container union as well:

struct poorly_packed {
        unsigned int a;
        unsigned int b;
        unsigned short c;
        union {
                struct {
                        unsigned short d;
                        unsigned int e;
                } __attribute__((packed));
                struct {
                        unsigned short d;
                        unsigned int e;
                } __attribute__((packed)) inner;
        } __attribute__((packed));
} __attribute__((packed));

Thanks to Andrew Pinski of GCC team for sorting the things out at
https://gcc.gnu.org/pipermail/gcc/2023-November/242888.html.

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202311150821.cI4yciFE-lkp@intel.com
Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>
Link: https://lore.kernel.org/r/20231120110607.98956-1-dmantipov@yandex.ru
Fixes: 50d7bd38c3aa ("stddef: Introduce struct_group() helper macro")
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Add missed allocation hint for bpf_mem_cache_alloc_flags()

[ Upstream commit 75a442581d05edaee168222ffbe00d4389785636 ]

bpf_mem_cache_alloc_flags() may call __alloc() directly when there is no
free object in free list, but it doesn't initialize the allocation hint
for the returned pointer. It may lead to bad memory dereference when
freeing the pointer, so fix it by initializing the allocation hint.

Fixes: 822fb26bdb55 ("bpf: Add a hint to allocated objects.")
Signed-off-by: Hou Tao <houtao1@huawei.com>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/r/20231111043821.2258513-1-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dpaa2-eth: recycle the RX buffer only after all processing done

[ Upstream commit beb1930f966d1517921488bd5d64147f58f79abf ]

The blamed commit added support for Rx copybreak. This meant that for
certain frame sizes, a new skb was allocated and the initial data buffer
was recycled. Instead of waiting to recycle the Rx buffer only after all
processing was done on it (like accessing the parse results or timestamp
information), the code path just went ahead and re-used the buffer right
away.

This sometimes lead to corrupted HW and SW annotation areas.
Fix this by delaying the moment when the buffer is recycled.

Fixes: 50f826999a80 ("dpaa2-eth: add rx copybreak support")
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dpaa2-eth: increase the needed headroom to account for alignment

[ Upstream commit f422abe3f23d483cf01f386819f26fb3fe0dbb2b ]

Increase the needed headroom to account for a 64 byte alignment
restriction which, with this patch, we make mandatory on the Tx path.
The case in which the amount of headroom needed is not available is
already handled by the driver which instead sends a S/G frame with the
first buffer only holding the SW and HW annotation areas.

Without this patch, we can empirically see data corruption happening
between Tx and Tx confirmation which sometimes leads to the SW
annotation area being overwritten.

Since this is an old IP where the hardware team cannot help to
understand the underlying behavior, we make the Tx alignment mandatory
for all frames to avoid the crash on Tx conf. Also, remove the comment
that suggested that this is just an optimization.

This patch also sets the needed_headroom net device field to the usual
value that the driver would need on the Tx path:
- 64 bytes for the software annotation area
- 64 bytes to account for a 64 byte aligned buffer address

Fixes: 6e2387e8f19e ("staging: fsl-dpaa2/eth: Add Freescale DPAA2 Ethernet driver")
Closes: https://lore.kernel.org/netdev/aa784d0c-85eb-4e5d-968b-c8f74fa86be6@gin.de/
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: mv88e6xxx: fix marvell 6350 probe crash

[ Upstream commit a524eabcd72d28425d9db242cf375d0389d74eba ]

As of commit b92143d4420f ("net: dsa: mv88e6xxx: add infrastructure for
phylink_pcs") probing of a Marvell 88e6350 switch causes a NULL pointer
de-reference like this example:

    ...
    mv88e6085 d0072004.mdio-mii:11: switch 0x3710 detected: Marvell 88E6350, revision 2
    8<--- cut here ---
    Unable to handle kernel NULL pointer dereference at virtual address 00000000 when read
    [00000000] *pgd=00000000
    Internal error: Oops: 5 [#1] ARM
    Modules linked in:
    CPU: 0 PID: 8 Comm: kworker/u2:0 Not tainted 6.7.0-rc2-dirty #26
    Hardware name: Marvell Armada 370/XP (Device Tree)
    Workqueue: events_unbound deferred_probe_work_func
    PC is at mv88e6xxx_port_setup+0x1c/0x44
    LR is at dsa_port_devlink_setup+0x74/0x154
    pc : [<c057ea24>]    lr : [<c0819598>]    psr: a0000013
    sp : c184fce0  ip : c542b8f4  fp : 00000000
    r10: 00000001  r9 : c542a540  r8 : c542bc00
    r7 : c542b838  r6 : c5244580  r5 : 00000005  r4 : c5244580
    r3 : 00000000  r2 : c542b840  r1 : 00000005  r0 : c1a02040
    ...

The Marvell 6350 switch has no SERDES interface and so has no
corresponding pcs_ops defined for it. But during probing a call is made
to mv88e6xxx_port_setup() which unconditionally expects pcs_ops to exist -
though the presence of the pcs_ops->pcs_init function is optional.

Modify code to check for pcs_ops first, before checking for and calling
pcs_ops->pcs_init. Modify checking and use of pcs_ops->pcs_teardown
which may potentially suffer the same problem.

Fixes: b92143d4420f ("net: dsa: mv88e6xxx: add infrastructure for phylink_pcs")
Signed-off-by: Greg Ungerer <gerg@kernel.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: mv88e6xxx: fix marvell 6350 switch probing

[ Upstream commit b3f1a164c7f742503dc7159011f7ad6b092b660e ]

As of commit de5c9bf40c45 ("net: phylink: require supported_interfaces to
be filled") Marvell 88e6350 switches fail to be probed:

    ...
    mv88e6085 d0072004.mdio-mii:11: switch 0x3710 detected: Marvell 88E6350, revision 2
    mv88e6085 d0072004.mdio-mii:11: phylink: error: empty supported_interfaces
    error creating PHYLINK: -22
    mv88e6085: probe of d0072004.mdio-mii:11 failed with error -22
    ...

The problem stems from the use of mv88e6185_phylink_get_caps() to get
the device capabilities. Create a new dedicated phylink_get_caps for the
6351 family (which the 6350 is one of) to properly support their set of
capabilities.

According to chip.h the 6351 switch family includes the 6171, 6175, 6350
and 6351 switches, so update each of these to use the correct
phylink_get_caps.

Fixes: de5c9bf40c45 ("net: phylink: require supported_interfaces to be filled")
Signed-off-by: Greg Ungerer <gerg@kernel.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wifi: mac80211: do not pass AP_VLAN vif pointer to drivers during flush

[ Upstream commit 3e3a2b645c043f7e3e488d5011478cefb69bbe8b ]

This fixes WARN_ONs when using AP_VLANs after station removal. The flush
call passed AP_VLAN vif to driver, but because these vifs are virtual and
not registered with drivers, we need to translate to the correct AP vif
first.

Closes: https://github.com/openwrt/openwrt/issues/12420
Fixes: 0b75a1b1e42e ("wifi: mac80211: flush queues on STA removal")
Fixes: d00800a289c9 ("wifi: mac80211: add flush_sta method")
Tested-by: Konstantin Demin <rockdrilla@gmail.com>
Tested-by: Koen Vandeputte <koen.vandeputte@citymesh.com>
Signed-off-by: Oldřich Jedlička <oldium.pro@gmail.com>
Link: https://lore.kernel.org/r/20231104141333.3710-1-oldium.pro@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wifi: iwlwifi: mvm: fix an error code in iwl_mvm_mld_add_sta()

[ Upstream commit 71b5e40651d89a8685bea1592dfcd2aa61559628 ]

This error path should return -EINVAL instead of success.

Fixes: 57974a55d995 ("wifi: iwlwifi: mvm: refactor iwl_mvm_mac_sta_state_common()")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Acked-by: Gregory Greenman <gregory.greenman@intel.com>
Link: https://lore.kernel.org/r/75e4ea09-db58-462f-bd4e-5ad4e5e5dcb5@moroto.mountain
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ipv4: igmp: fix refcnt uaf issue when receiving igmp query packet

[ Upstream commit e2b706c691905fe78468c361aaabc719d0a496f1 ]

When I perform the following test operations:
1.ip link add br0 type bridge
2.brctl addif br0 eth0
3.ip addr add 239.0.0.1/32 dev eth0
4.ip addr add 239.0.0.1/32 dev br0
5.ip addr add 224.0.0.1/32 dev br0
6.while ((1))
    do
        ifconfig br0 up
        ifconfig br0 down
    done
7.send IGMPv2 query packets to port eth0 continuously. For example,
./mausezahn ethX -c 0 "01 00 5e 00 00 01 00 72 19 88 aa 02 08 00 45 00 00
1c 00 01 00 00 01 02 0e 7f c0 a8 0a b7 e0 00 00 01 11 64 ee 9b 00 00 00 00"

The preceding tests may trigger the refcnt uaf issue of the mc list. The
stack is as follows:
refcount_t: addition on 0; use-after-free.
WARNING: CPU: 21 PID: 144 at lib/refcount.c:25 refcount_warn_saturate (lib/refcount.c:25)
CPU: 21 PID: 144 Comm: ksoftirqd/21 Kdump: loaded Not tainted 6.7.0-rc1-next-20231117-dirty #80
Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
RIP: 0010:refcount_warn_saturate (lib/refcount.c:25)
RSP: 0018:ffffb68f00657910 EFLAGS: 00010286
RAX: 0000000000000000 RBX: ffff8a00c3bf96c0 RCX: ffff8a07b6160908
RDX: 00000000ffffffd8 RSI: 0000000000000027 RDI: ffff8a07b6160900
RBP: ffff8a00cba36862 R08: 0000000000000000 R09: 00000000ffff7fff
R10: ffffb68f006577c0 R11: ffffffffb0fdcdc8 R12: ffff8a00c3bf9680
R13: ffff8a00c3bf96f0 R14: 0000000000000000 R15: ffff8a00d8766e00
FS:  0000000000000000(0000) GS:ffff8a07b6140000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000055f10b520b28 CR3: 000000039741a000 CR4: 00000000000006f0
Call Trace:
<TASK>
igmp_heard_query (net/ipv4/igmp.c:1068)
igmp_rcv (net/ipv4/igmp.c:1132)
ip_protocol_deliver_rcu (net/ipv4/ip_input.c:205)
ip_local_deliver_finish (net/ipv4/ip_input.c:234)
__netif_receive_skb_one_core (net/core/dev.c:5529)
netif_receive_skb_internal (net/core/dev.c:5729)
netif_receive_skb (net/core/dev.c:5788)
br_handle_frame_finish (net/bridge/br_input.c:216)
nf_hook_bridge_pre (net/bridge/br_input.c:294)
__netif_receive_skb_core (net/core/dev.c:5423)
__netif_receive_skb_list_core (net/core/dev.c:5606)
__netif_receive_skb_list (net/core/dev.c:5674)
netif_receive_skb_list_internal (net/core/dev.c:5764)
napi_gro_receive (net/core/gro.c:609)
e1000_clean_rx_irq (drivers/net/ethernet/intel/e1000/e1000_main.c:4467)
e1000_clean (drivers/net/ethernet/intel/e1000/e1000_main.c:3805)
__napi_poll (net/core/dev.c:6533)
net_rx_action (net/core/dev.c:6735)
__do_softirq (kernel/softirq.c:554)
run_ksoftirqd (kernel/softirq.c:913)
smpboot_thread_fn (kernel/smpboot.c:164)
kthread (kernel/kthread.c:388)
ret_from_fork (arch/x86/kernel/process.c:153)
ret_from_fork_asm (arch/x86/entry/entry_64.S:250)
</TASK>

The root causes are as follows:
Thread A Thread B
... netif_receive_skb
br_dev_stop ...
    br_multicast_leave_snoopers ...
        __ip_mc_dec_group ...
            __igmp_group_dropped igmp_rcv
                igmp_stop_timer     igmp_heard_query         //ref = 1
                ip_ma_put         igmp_mod_timer
                    refcount_dec_and_test             igmp_start_timer //ref = 0
...                                     refcount_inc //ref increases from 0
When the device receives an IGMPv2 Query message, it starts the timer
immediately, regardless of whether the device is running. If the device is
down and has left the multicast group, it will cause the mc list refcount
uaf issue.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: rswitch: Fix missing dev_kfree_skb_any() in error path

[ Upstream commit 782486af9b5b76493012711413c141509ac45dec ]

Before returning the rswitch_start_xmit() in the error path,
dev_kfree_skb_any() should be called. So, fix it.

Fixes: 33f5d733b589 ("net: renesas: rswitch: Improve TX timestamp accuracy")
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: rswitch: Fix return value in rswitch_start_xmit()

[ Upstream commit 1aaef8634a20b322c82e84f12a9b6aec1e2fd4fa ]

This .ndo_start_xmit() function should return netdev_tx_t value,
not -ENOMEM. So, fix it.

Fixes: 33f5d733b589 ("net: renesas: rswitch: Improve TX timestamp accuracy")
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: rswitch: Fix type of ret in rswitch_start_xmit()

[ Upstream commit 109b25d13e0054337860d44841b990d11b32d262 ]

The type of ret in rswitch_start_xmit() should be netdev_tx_t. So,
fix it.

Fixes: 3590918b5d07 ("net: ethernet: renesas: Add support for "Ethernet Switch"")
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netdevsim: Don't accept device bound programs

[ Upstream commit c0c6bde586c7dce82719b4ff32a2db6af9ee3d65 ]

Commit 2b3486bc2d23 ("bpf: Introduce device-bound XDP programs") introduced
device-bound programs by largely reusing existing offloading infrastructure.
This changed the semantics of 'prog->aux->offload' a bit. Now, it's non-NULL
for both offloaded and device-bound programs.

Instead of looking at 'prog->aux->offload' let's call bpf_prog_is_offloaded
which should be true iff the program is offloaded and not merely device-bound.

Fixes: 2b3486bc2d23 ("bpf: Introduce device-bound XDP programs")
Reported-by: syzbot+44c2416196b7c607f226@syzkaller.appspotmail.com
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Cc: Dipendra Khadka <kdipendra88@gmail.com>
Link: https://lore.kernel.org/bpf/20231114045453.1816995-2-sdf@google.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: v4l2-subdev: Fix a 64bit bug

[ Upstream commit 5d33213fac5929a2e7766c88d78779fd443b0fe8 ]

The problem is this line here from subdev_do_ioctl().

        client_cap->capabilities &= ~V4L2_SUBDEV_CLIENT_CAP_STREAMS;

The "client_cap->capabilities" variable is a u64.  The AND operation
is supposed to clear out the V4L2_SUBDEV_CLIENT_CAP_STREAMS flag.  But
because it's a 32 bit variable it accidentally clears out the high 32
bits as well.

Currently we only use the first bit and none of the upper bits so this
doesn't affect runtime behavior.

Fixes: f57fa2959244 ("media: v4l2-subdev: Add new ioctl for client capabilities")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pinctrl: stm32: fix array read out of bound

[ Upstream commit edd48fd9d45370d6c8ba0dd834fcc51ff688cc87 ]

The existing code does not verify if the "tentative" index exceeds
the size of the array, causing out of bound read.
Issue identified with kasan.

Check the index before using it.

Signed-off-by: Antonio Borneo <antonio.borneo@foss.st.com>
Fixes: 32c170ff15b0 ("pinctrl: stm32: set default gpio line names using pin names")
Link: https://lore.kernel.org/r/20231107110520.4449-1-antonio.borneo@foss.st.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pinctrl: stm32: Add check for devm_kcalloc

[ Upstream commit b0eeba527e704d6023a6cd9103f929226e326b03 ]

Add check for the return value of devm_kcalloc() and return the error
if it fails in order to avoid NULL pointer dereference.

Fixes: 32c170ff15b0 ("pinctrl: stm32: set default gpio line names using pin names")
Signed-off-by: Chen Ni <nichen@iscas.ac.cn>
Acked-by: Valentin Caron <valentin.caron@foss.st.com>
Link: https://lore.kernel.org/r/20231031080807.3600656-1-nichen@iscas.ac.cn
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

wifi: cfg80211: fix CQM for non-range use

commit 7e7efdda6adb385fbdfd6f819d76bc68c923c394 upstream.

My prior race fix here broke CQM when ranges aren't used, as
the reporting worker now requires the cqm_config to be set in
the wdev, but isn't set when there's no range configured.

Rather than continuing to special-case the range version, set
the cqm_config always and configure accordingly, also tracking
if range was used or not to be able to clear the configuration
appropriately with the same API, which was actually not right
if both were implemented by a driver for some reason, as is
the case with mac80211 (though there the implementations are
equivalent so it doesn't matter.)

Also, the original multiple-RSSI commit lost checking for the
callback, so might have potentially crashed if a driver had
neither implementation, and userspace tried to use it despite
not being advertised as supported.

Cc: stable@vger.kernel.org
Fixes: 4a4b8169501b ("cfg80211: Accept multiple RSSI thresholds for CQM")
Fixes: 37c20b2effe9 ("wifi: cfg80211: fix cqm_config access race")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring/kbuf: recycle freed mapped buffer ring entries

commit b10b73c102a2eab91e1cd62a03d6446f1dfecc64 upstream.

Right now we stash any potentially mmap'ed provided ring buffer range
for freeing at release time, regardless of when they get unregistered.
Since we're keeping track of these ranges anyway, keep track of their
registration state as well, and use that to recycle ranges when
appropriate rather than always allocate new ones.

The lookup is a basic scan of entries, checking for the best matching
free entry.

Fixes: c392cbecd8ec ("io_uring/kbuf: defer release of mapped buffer rings")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring/kbuf: defer release of mapped buffer rings

commit c392cbecd8eca4c53f2bf508731257d9d0a21c2d upstream.

If a provided buffer ring is setup with IOU_PBUF_RING_MMAP, then the
kernel allocates the memory for it and the application is expected to
mmap(2) this memory. However, io_uring uses remap_pfn_range() for this
operation, so we cannot rely on normal munmap/release on freeing them
for us.

Stash an io_buf_free entry away for each of these, if any, and provide
a helper to free them post ->release().

Cc: stable@vger.kernel.org
Fixes: c56e022c0a27 ("io_uring: add support for user mapped provided buffer ring")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring: enable io_mem_alloc/free to be used in other parts

commit edecf1689768452ba1a64b7aaf3a47a817da651a upstream.

In preparation for using these helpers, make them non-static and add
them to our internal header.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: fix 64bit compat send ioctl arguments not initializing version member

commit 5de0434bc064606d6b7467ec3e5ad22963a18c04 upstream.

When the send protocol versioning was added in 5.16 e77fbf990316
("btrfs: send: prepare for v2 protocol"), the 32/64bit compat code was
not updated (added by 2351f431f727 ("btrfs: fix send ioctl on 32bit with
64bit kernel")), missing the version struct member. The compat code is
probably rarely used, nobody reported any bugs.

Found by tool https://github.com/jirislaby/clang-struct .

Fixes: e77fbf990316 ("btrfs: send: prepare for v2 protocol")
CC: stable@vger.kernel.org # 6.1+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: free the allocated memory if btrfs_alloc_page_array() fails

commit 94dbf7c0871f7ae6349ba4b0341ce8f5f98a071d upstream.

[BUG]
If btrfs_alloc_page_array() fail to allocate all pages but part of the
slots, then the partially allocated pages would be leaked in function
btrfs_submit_compressed_read().

[CAUSE]
As explicitly stated, if btrfs_alloc_page_array() returned -ENOMEM,
caller is responsible to free the partially allocated pages.

For the existing call sites, most of them are fine:

- btrfs_raid_bio::stripe_pages
  Handled by free_raid_bio().

- extent_buffer::pages[]
  Handled btrfs_release_extent_buffer_pages().

- scrub_stripe::pages[]
  Handled by release_scrub_stripe().

But there is one exception in btrfs_submit_compressed_read(), if
btrfs_alloc_page_array() failed, we didn't cleanup the array and freed
the array pointer directly.

Initially there is still the error handling in commit dd137dd1f2d7
("btrfs: factor out allocating an array of pages"), but later in commit
544fe4a903ce ("btrfs: embed a btrfs_bio into struct compressed_bio"),
the error handling is removed, leading to the possible memory leak.

[FIX]
This patch would add back the error handling first, then to prevent such
situation from happening again, also
Make btrfs_alloc_page_array() to free the allocated pages as a extra
safety net, then we don't need to add the error handling to
btrfs_submit_compressed_read().

Fixes: 544fe4a903ce ("btrfs: embed a btrfs_bio into struct compressed_bio")
CC: stable@vger.kernel.org # 6.4+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: make error messages more clear when getting a chunk map

commit 7d410d5efe04e42a6cd959bfe6d59d559fdf8b25 upstream.

When getting a chunk map, at btrfs_get_chunk_map(), we do some sanity
checks to verify we found a chunk map and that map found covers the
logical address the caller passed in. However the messages aren't very
clear in the sense that don't mention the issue is with a chunk map and
one of them prints the 'length' argument as if it were the end offset of
the requested range (while the in the string format we use %llu-%llu
which suggests a range, and the second %llu-%llu is actually a range for
the chunk map). So improve these two details in the error messages.

CC: stable@vger.kernel.org # 5.4+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: send: ensure send_fd is writable

commit 0ac1d13a55eb37d398b63e6ff6db4a09a2c9128c upstream.

kernel_write() requires the caller to ensure that the file is writable.
Let's do that directly after looking up the ->send_fd.

We don't need a separate bailout path because the "out" path already
does fput() if ->send_filp is non-NULL.

This has no security impact for two reasons:

 - the ioctl requires CAP_SYS_ADMIN
 - __kernel_write() bails out on read-only files - but only since 5.8,
   see commit a01ac27be472 ("fs: check FMODE_WRITE in __kernel_write")

Reported-and-tested-by: syzbot+12e098239d20385264d3@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=12e098239d20385264d3
Fixes: 31db9f7c23fb ("Btrfs: introduce BTRFS_IOC_SEND for btrfs send/receive")
CC: stable@vger.kernel.org # 4.14+
Signed-off-by: Jann Horn <jannh@google.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: fix off-by-one when checking chunk map includes logical address

commit 5fba5a571858ce2d787fdaf55814e42725bfa895 upstream.

At btrfs_get_chunk_map() we get the extent map for the chunk that contains
the given logical address stored in the 'logical' argument. Then we do
sanity checks to verify the extent map contains the logical address. One
of these checks verifies if the extent map covers a range with an end
offset behind the target logical address - however this check has an
off-by-one error since it will consider an extent map whose start offset
plus its length matches the target logical address as inclusive, while
the fact is that the last byte it covers is behind the target logical
address (by 1).

So fix this condition by using '<=' rather than '<' when comparing the
extent map's "start + length" against the target logical address.

CC: stable@vger.kernel.org # 4.14+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: ref-verify: fix memory leaks in btrfs_ref_tree_mod()

commit f91192cd68591c6b037da345bc9fcd5e50540358 upstream.

In btrfs_ref_tree_mod(), when !parent 're' was allocated through
kmalloc(). In the following code, if an error occurs, the execution will
be redirected to 'out' or 'out_unlock' and the function will be exited.
However, on some of the paths, 're' are not deallocated and may lead to
memory leaks.

For example: lookup_block_entry() for 'be' returns NULL, the out label
will be invoked. During that flow ref and 'ra' are freed but not 're',
which can potentially lead to a memory leak.

CC: stable@vger.kernel.org # 5.10+
Reported-and-tested-by: syzbot+d66de4cbf532749df35f@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=d66de4cbf532749df35f
Signed-off-by: Bragatheswaran Manickavel <bragathemanick0908@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: add dmesg output for first mount and last unmount of a filesystem

commit 2db313205f8b96eea467691917138d646bb50aef upstream.

There is a feature request to add dmesg output when unmounting a btrfs.
There are several alternative methods to do the same thing, but with
their own problems:

- Use eBPF to watch btrfs_put_super()/open_ctree()
  Not end user friendly, they have to dip their head into the source
  code.

- Watch for directory /sys/fs/<uuid>/
  This is way more simple, but still requires some simple device -> uuid
  lookups.  And a script needs to use inotify to watch /sys/fs/.

Compared to all these, directly outputting the information into dmesg
would be the most simple one, with both device and UUID included.

And since we're here, also add the output when mounting a filesystem for
the first time for parity. A more fine grained monitoring of subvolume
mounts should be done by another layer, like audit.

Now mounting a btrfs with all default mkfs options would look like this:

  [81.906566] BTRFS info (device dm-8): first mount of filesystem 633b5c16-afe3-4b79-b195-138fe145e4f2
  [81.907494] BTRFS info (device dm-8): using crc32c (crc32c-intel) checksum algorithm
  [81.908258] BTRFS info (device dm-8): using free space tree
  [81.912644] BTRFS info (device dm-8): auto enabling async discard
  [81.913277] BTRFS info (device dm-8): checking UUID tree
  [91.668256] BTRFS info (device dm-8): last unmount of filesystem 633b5c16-afe3-4b79-b195-138fe145e4f2

CC: stable@vger.kernel.org # 5.4+
Link: https://github.com/kdave/btrfs-progs/issues/689
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc: Mark altinstructions read-only and 32-bit aligned

commit 33f806da2df68606f77d7b892cd1298ba3d463e8 upstream.

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

parisc: Ensure 32-bit alignment on parisc unwind section

commit c9fcb2b65c2849e8ff3be23fd8828312fb68dc19 upstream.

Make sure the .PARISC.unwind section will be 32-bit aligned.

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

parisc: Mark jump_table naturally aligned

commit 07eecff8ae78df7f28800484d31337e1f9bfca3a upstream.

The jump_table stores two 32-bit words and one 32- (on 32-bit kernel)
or one 64-bit word (on 64-bit kernel).
Ensure that the last word is always 64-bit aligned on a 64-bit kernel
by aligning the whole structure on sizeof(long).

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

parisc: Drop the HP-UX ENOSYM and EREMOTERELEASE error codes

commit e5f3e299a2b1e9c3ece24a38adfc089aef307e8a upstream.

Those return codes are only defined for the parisc architecture and
are leftovers from when we wanted to be HP-UX compatible.

They are not returned by any Linux kernel syscall but do trigger
problems with the glibc strerrorname_np() and strerror() functions as
reported in glibc issue #31080.

There is no need to keep them, so simply remove them.

Signed-off-by: Helge Deller <deller@gmx.de>
Reported-by: Bruno Haible <bruno@clisp.org>
Closes: https://sourceware.org/bugzilla/show_bug.cgi?id=31080
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc: Mark lock_aligned variables 16-byte aligned on SMP

commit b28fc0d8739c03e7b6c44914a9d00d4c6dddc0ea upstream.

On parisc we need 16-byte alignment for variables which are used for
locking. Mark the __lock_aligned attribute acordingly so that the
.data..lock_aligned section will get that alignment in the generated
object files.

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

parisc: Use natural CPU alignment for bug_table

commit fe76a1349f235969381832c83d703bc911021eb6 upstream.

Make sure that the __bug_table section gets 32- or 64-bit aligned,
depending if a 32- or 64-bit kernel is being built.
Mark it non-writeable and use .blockz instead of the .org assembler
directive to pad the struct.

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

parisc: Mark ex_table entries 32-bit aligned in uaccess.h

commit a80aeb86542a50aa8521729ea4cc731ee7174f03 upstream.

Add an align statement to tell the linker that all ex_table entries and as
such the whole ex_table section should be 32-bit aligned in vmlinux and modules.

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

parisc: Mark ex_table entries 32-bit aligned in assembly.h

commit e11d4cccd094a7cd4696c8c42e672c76c092dad5 upstream.

Add an align statement to tell the linker that all ex_table entries and as
such the whole ex_table section should be 32-bit aligned in vmlinux and modules.

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

powerpc: Don't clobber f0/vs0 during fp|altivec register save

commit 5e1d824f9a283cbf90f25241b66d1f69adb3835b upstream.

During floating point and vector save to thread data f0/vs0 are
clobbered by the FPSCR/VSCR store routine. This has been obvserved to
lead to userspace register corruption and application data corruption
with io-uring.

Fix it by restoring f0/vs0 after FPSCR/VSCR store has completed for
all the FP, altivec, VMX register save paths.

Tested under QEMU in kvm mode, running on a Talos II workstation with
dual POWER9 DD2.2 CPUs.

Additional detail (mpe):

Typically save_fpu() is called from __giveup_fpu() which saves the FP
regs and also *turns off FP* in the tasks MSR, meaning the kernel will
reload the FP regs from the thread struct before letting the task use FP
again. So in that case save_fpu() is free to clobber f0 because the FP
regs no longer hold live values for the task.

There is another case though, which is the path via:
  sys_clone()
    ...
    copy_process()
      dup_task_struct()
        arch_dup_task_struct()
          flush_all_to_thread()
            save_all()

That path saves the FP regs but leaves them live. That's meant as an
optimisation for a process that's using FP/VSX and then calls fork(),
leaving the regs live means the parent process doesn't have to take a
fault after the fork to get its FP regs back. The optimisation was added
in commit 8792468da5e1 ("powerpc: Add the ability to save FPU without
giving it up").

That path does clobber f0, but f0 is volatile across function calls,
and typically programs reach copy_process() from userspace via a syscall
wrapper function. So in normal usage f0 being clobbered across a
syscall doesn't cause visible data corruption.

But there is now a new path, because io-uring can call copy_process()
via create_io_thread() from the signal handling path. That's OK if the
signal is handled as part of syscall return, but it's not OK if the
signal is handled due to some other interrupt.

That path is:

interrupt_return_srr_user()
  interrupt_exit_user_prepare()
    interrupt_exit_user_prepare_main()
      do_notify_resume()
        get_signal()
          task_work_run()
            create_worker_cb()
              create_io_worker()
                copy_process()
                  dup_task_struct()
                    arch_dup_task_struct()
                      flush_all_to_thread()
                        save_all()
                          if (tsk->thread.regs->msr & MSR_FP)
                            save_fpu()
                            # f0 is clobbered and potentially live in userspace

Note the above discussion applies equally to save_altivec().

Fixes: 8792468da5e1 ("powerpc: Add the ability to save FPU without giving it up")
Cc: stable@vger.kernel.org # v4.6+
Closes: https://lore.kernel.org/all/480932026.45576726.1699374859845.JavaMail.zimbra@raptorengineeringinc.com/
Closes: https://lore.kernel.org/linuxppc-dev/480221078.47953493.1700206777956.JavaMail.zimbra@raptorengineeringinc.com/
Tested-by: Timothy Pearson <tpearson@raptorengineering.com>
Tested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Timothy Pearson <tpearson@raptorengineering.com>
[mpe: Reword change log to describe exact path of corruption & other minor tweaks]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/1921539696.48534988.1700407082933.JavaMail.zimbra@raptorengineeringinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: PPC: Book3S HV: Fix KVM_RUN clobbering FP/VEC user registers

commit dc158d23b33df9033bcc8e7117e8591dd2f9d125 upstream.

Before running a guest, the host process (e.g., QEMU) FP/VEC registers
are saved if they were being used, similarly to when the kernel uses FP
registers. The guest values are then loaded into regs, and the host
process registers will be restored lazily when it uses FP/VEC.

KVM HV has a bug here: the host process registers do get saved, but the
user MSR bits remain enabled, which indicates the registers are valid
for the process. After they are clobbered by running the guest, this
valid indication causes the host process to take on the FP/VEC register
values of the guest.

Fixes: 34e119c96b2b ("KVM: PPC: Book3S HV P9: Reduce mtmsrd instructions required to save host SPRs")
Cc: stable@vger.kernel.org # v5.17+
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/20231122025811.2973-1-npiggin@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

iommu/vt-d: Add MTL to quirk list to skip TE disabling

commit 85b80fdffa867d75dfb9084a839e7949e29064e8 upstream.

The VT-d spec requires (10.4.4 Global Command Register, TE field) that:

Hardware implementations supporting DMA draining must drain any in-flight
DMA read/write requests queued within the Root-Complex before switching
address translation on or off and reflecting the status of the command
through the TES field in the Global Status register.

Unfortunately, some integrated graphic devices fail to do so after some
kind of power state transition. As the result, the system might stuck in
iommu_disable_translation(), waiting for the completion of TE transition.

Add MTL to the quirk list for those devices and skips TE disabling if the
qurik hits.

Fixes: b1012ca8dc4f ("iommu/vt-d: Skip TE disabling on quirky gfx dedicated iommu")
Cc: stable@vger.kernel.org
Signed-off-by: Abdul Halim, Mohd Syazwan <mohd.syazwan.abdul.halim@intel.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20231116022324.30120-1-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext2: Fix ki_pos update for DIO buffered-io fallback case

commit 8abc712ea4867a81c860853048f24e511bbc20f2 upstream.

Commit "filemap: update ki_pos in generic_perform_write", made updating
of ki_pos into common code in generic_perform_write() function.
This also causes generic/091 to fail.
This happened due to an in-flight collision with:
fb5de4358e1a ("ext2: Move direct-io to use iomap"). I have chosen fixes tag
based on which commit got landed later to upstream kernel.

Fixes: 182c25e9c157 ("filemap: update ki_pos in generic_perform_write")
Cc: stable@vger.kernel.org
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <d595bee9f2475ed0e8a2e7fb94f7afc2c6ffc36a.1700643443.git.ritesh.list@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bcache: revert replacing IS_ERR_OR_NULL with IS_ERR

commit bb6cc253861bd5a7cf8439e2118659696df9619f upstream.

Commit 028ddcac477b ("bcache: Remove unnecessary NULL point check in
node allocations") replaced IS_ERR_OR_NULL by IS_ERR. This leads to a
NULL pointer dereference.

BUG: kernel NULL pointer dereference, address: 0000000000000080
Call Trace:
 ? __die_body.cold+0x1a/0x1f
 ? page_fault_oops+0xd2/0x2b0
 ? exc_page_fault+0x70/0x170
 ? asm_exc_page_fault+0x22/0x30
 ? btree_node_free+0xf/0x160 [bcache]
 ? up_write+0x32/0x60
 btree_gc_coalesce+0x2aa/0x890 [bcache]
 ? bch_extent_bad+0x70/0x170 [bcache]
 btree_gc_recurse+0x130/0x390 [bcache]
 ? btree_gc_mark_node+0x72/0x230 [bcache]
 bch_btree_gc+0x5da/0x600 [bcache]
 ? cpuusage_read+0x10/0x10
 ? bch_btree_gc+0x600/0x600 [bcache]
 bch_gc_thread+0x135/0x180 [bcache]

The relevant code starts with:

    new_nodes[0] = NULL;

    for (i = 0; i < nodes; i++) {
        if (__bch_keylist_realloc(&keylist, bkey_u64s(&r[i].b->key)))
            goto out_nocoalesce;
    // ...
out_nocoalesce:
    // ...
    for (i = 0; i < nodes; i++)
        if (!IS_ERR(new_nodes[i])) {  // IS_ERR_OR_NULL before
028ddcac477b
            btree_node_free(new_nodes[i]);  // new_nodes[0] is NULL
            rw_unlock(true, new_nodes[i]);
        }

This patch replaces IS_ERR() by IS_ERR_OR_NULL() to fix this.

Fixes: 028ddcac477b ("bcache: Remove unnecessary NULL point check in node allocations")
Link: https://lore.kernel.org/all/3DF4A87A-2AC1-4893-AE5F-E921478419A9@suse.de/
Cc: stable@vger.kernel.org
Cc: Zheng Wang <zyytlz.wz@163.com>
Cc: Coly Li <colyli@suse.de>
Signed-off-by: Markus Weippert <markus@gekmihesg.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

iommu: Avoid more races around device probe

commit a2e7e59a94269484a83386972ca07c22fd188854 upstream.

It turns out there are more subtle races beyond just the main part of
__iommu_probe_device() itself running in parallel - the dev_iommu_free()
on the way out of an unsuccessful probe can still manage to trip up
concurrent accesses to a device's fwspec. Thus, extend the scope of
iommu_probe_device_lock() to also serialise fwspec creation and initial
retrieval.

Reported-by: Zhenhua Huang <quic_zhenhuah@quicinc.com>
Link: https://lore.kernel.org/linux-iommu/e2e20e1c-6450-4ac5-9804-b0000acdf7de@quicinc.com/
Fixes: 01657bc14a39 ("iommu: Avoid races around device probe")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: André Draszik <andre.draszik@linaro.org>
Tested-by: André Draszik <andre.draszik@linaro.org>
Link: https://lore.kernel.org/r/16f433658661d7cadfea51e7c65da95826112a2b.1700071477.git.robin.murphy@arm.com
Cc: stable@vger.kernel.org
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring: don't guard IORING_OFF_PBUF_RING with SETUP_NO_MMAP

commit 6f007b1406637d3d73d42e41d7e8d9b245185e69 upstream.

This flag only applies to the SQ and CQ rings, it's perfectly valid
to use a mmap approach for the provided ring buffers. Move the
check into where it belongs.

Cc: stable@vger.kernel.org
Fixes: 03d89a2de25b ("io_uring: support for user allocated memory for rings/sqes")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dma-buf: fix check in dma_resv_add_fence

commit 95ba893c9f4feb836ddce627efd0bb6af6667031 upstream.

It's valid to add the same fence multiple times to a dma-resv object and
we shouldn't need one extra slot for each.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Fixes: a3f7c10a269d5 ("dma-buf/dma-resv: check if the new fence is really later")
Cc: stable@vger.kernel.org # v5.19+
Link: https://patchwork.freedesktop.org/patch/msgid/20231115093035.1889-1-christian.koenig@amd.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cpufreq/amd-pstate: Fix the return value of amd_pstate_fast_switch()

commit bb87be267b8ee9b40917fb5bf51be5ddb33c37c2 upstream.

cpufreq_driver->fast_switch() callback expects a frequency as a return
value. amd_pstate_fast_switch() was returning the return value of
amd_pstate_update_freq(), which only indicates a success or failure.

Fix this by making amd_pstate_fast_switch() return the target_freq
when the call to amd_pstate_update_freq() is successful, and return
the current frequency from policy->cur when the call to
amd_pstate_update_freq() is unsuccessful.

Fixes: 4badf2eb1e98 ("cpufreq: amd-pstate: Add ->fast_switch() callback")
Acked-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Wyes Karny <wyes.karny@amd.com>
Reviewed-by: Perry Yuan <perry.yuan@amd.com>
Cc: 6.4+ <stable@vger.kernel.org> # v6.4+
Signed-off-by: Gautham R. Shenoy <gautham.shenoy@amd.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

powercap: DTPM: Fix unneeded conversions to micro-Watts

commit b817f1488fca548fe50e2654d84a1956a16a1a8a upstream.

The power values coming from the Energy Model are already in uW.

The PowerCap and DTPM frameworks operate on uW, so all places should
just use the values from the EM.

Fix the code by removing all of the conversion to uW still present in it.

Fixes: ae6ccaa65038 (PM: EM: convert power field to micro-Watts precision and align drivers)
Cc: 5.19+ <stable@vger.kernel.org> # v5.19+
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nouveau: find the smallest page allocation to cover a buffer alloc.

commit e9ba37d9f9a6872b069dd893bd86a7d77ba8c153 upstream.

With the new uapi we don't have the comp flags on the allocation,
so we shouldn't be using the first size that works, we should be
iterating until we get the correct one.

This reduces allocations from 2MB to 64k in lots of places.

Fixes dEQP-VK.memory.allocation.basic.size_8KiB.forward.count_4000
on my ampere/gsp system.

Cc: stable@vger.kernel.org # v6.6
Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Signed-off-by: Danilo Krummrich <dakr@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230811031520.248341-1-airlied@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring: free io_buffer_list entries via RCU

commit 5cf4f52e6d8aa2d3b7728f568abbf9d42a3af252 upstream.

mmap_lock nests under uring_lock out of necessity, as we may be doing
user copies with uring_lock held. However, for mmap of provided buffer
rings, we attempt to grab uring_lock with mmap_lock already held from
do_mmap(). This makes lockdep, rightfully, complain:

WARNING: possible circular locking dependency detected
6.7.0-rc1-00009-gff3337ebaf94-dirty #4438 Not tainted
------------------------------------------------------
buf-ring.t/442 is trying to acquire lock:
ffff00020e1480a8 (&ctx->uring_lock){+.+.}-{3:3}, at: io_uring_validate_mmap_request.isra.0+0x4c/0x140

but task is already holding lock:
ffff0000dc226190 (&mm->mmap_lock){++++}-{3:3}, at: vm_mmap_pgoff+0x124/0x264

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&mm->mmap_lock){++++}-{3:3}:
       __might_fault+0x90/0xbc
       io_register_pbuf_ring+0x94/0x488
       __arm64_sys_io_uring_register+0x8dc/0x1318
       invoke_syscall+0x5c/0x17c
       el0_svc_common.constprop.0+0x108/0x130
       do_el0_svc+0x2c/0x38
       el0_svc+0x4c/0x94
       el0t_64_sync_handler+0x118/0x124
       el0t_64_sync+0x168/0x16c

-> #0 (&ctx->uring_lock){+.+.}-{3:3}:
       __lock_acquire+0x19a0/0x2d14
       lock_acquire+0x2e0/0x44c
       __mutex_lock+0x118/0x564
       mutex_lock_nested+0x20/0x28
       io_uring_validate_mmap_request.isra.0+0x4c/0x140
       io_uring_mmu_get_unmapped_area+0x3c/0x98
       get_unmapped_area+0xa4/0x158
       do_mmap+0xec/0x5b4
       vm_mmap_pgoff+0x158/0x264
       ksys_mmap_pgoff+0x1d4/0x254
       __arm64_sys_mmap+0x80/0x9c
       invoke_syscall+0x5c/0x17c
       el0_svc_common.constprop.0+0x108/0x130
       do_el0_svc+0x2c/0x38
       el0_svc+0x4c/0x94
       el0t_64_sync_handler+0x118/0x124
       el0t_64_sync+0x168/0x16c

From that mmap(2) path, we really just need to ensure that the buffer
list doesn't go away from underneath us. For the lower indexed entries,
they never go away until the ring is freed and we can always sanely
reference those as long as the caller has a file reference. For the
higher indexed ones in our xarray, we just need to ensure that the
buffer list remains valid while we return the address of it.

Free the higher indexed io_buffer_list entries via RCU. With that we can
avoid needing ->uring_lock inside mmap(2), and simply hold the RCU read
lock around the buffer list lookup and address check.

To ensure that the arrayed lookup either returns a valid fully formulated
entry via RCU lookup, add an 'is_ready' flag that we access with store
and release memory ordering. This isn't needed for the xarray lookups,
but doesn't hurt either. Since this isn't a fast path, retain it across
both types. Similarly, for the allocated array inside the ctx, ensure
we use the proper load/acquire as setup could in theory be running in
parallel with mmap.

While in there, add a few lockdep checks for documentation purposes.

Cc: stable@vger.kernel.org
Fixes: c56e022c0a27 ("io_uring: add support for user mapped provided buffer ring")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

iommu/vt-d: Fix incorrect cache invalidation for mm notification

commit e7ad6c2a4b1aa710db94060b716f53c812cef565 upstream.

Commit 6bbd42e2df8f ("mmu_notifiers: call invalidate_range() when
invalidating TLBs") moved the secondary TLB invalidations into the TLB
invalidation functions to ensure that all secondary TLB invalidations
happen at the same time as the CPU invalidation and added a flush-all
type of secondary TLB invalidation for the batched mode, where a range
of [0, -1UL) is used to indicates that the range extends to the end of
the address space.

However, using an end address of -1UL caused an overflow in the Intel
IOMMU driver, where the end address was rounded up to the next page.
As a result, both the IOTLB and device ATC were not invalidated correctly.

Add a flush all helper function and call it when the invalidation range
is from 0 to -1UL, ensuring that the entire caches are invalidated
correctly.

Fixes: 6bbd42e2df8f ("mmu_notifiers: call invalidate_range() when invalidating TLBs")
Cc: stable@vger.kernel.org
Cc: Huang Ying <ying.huang@intel.com>
Cc: Alistair Popple <apopple@nvidia.com>
Tested-by: Luo Yuzhang <yuzhang.luo@intel.com> # QAT
Tested-by: Tony Zhu <tony.zhu@intel.com> # DSA
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20231117090933.75267-1-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring: don't allow discontig pages for IORING_SETUP_NO_MMAP

commit 820d070feb668aab5bc9413c285a1dda2a70e076 upstream.

io_sqes_map() is used rather than io_mem_alloc(), if the application
passes in memory for mapping rather than have the kernel allocate it and
then mmap(2) the ranges. This then calls __io_uaddr_map() to perform the
page mapping and pinning, which checks if we end up with the same pages,
if more than one page is mapped. But this check is incorrect and only
checks if the first and last pages are the same, where it really should
be checking if the mapped pages are contigous. This allows mapping a
single normal page, or a huge page range.

Down the line we can add support for remapping pages to be virtually
contigous, which is really all that io_uring cares about.

Cc: stable@vger.kernel.org
Fixes: 03d89a2de25b ("io_uring: support for user allocated memory for rings/sqes")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ACPI: video: Use acpi_video_device for cooling-dev driver data

commit 172c48caed91a978bca078042222d09baea13717 upstream.

The acpi_video code was storing the acpi_video_device as driver_data
in the acpi_device children of the acpi_video_bus acpi_device.

But the acpi_video driver only binds to the bus acpi_device.
It uses, but does not bind to, the children. Since it is not
the driver it should not be using the driver_data of the children's
acpi_device-s.

Since commit 0d16710146a1 ("ACPI: bus: Set driver_data to NULL every
time .add() fails") the childen's driver_data ends up getting set
to NULL after a driver fails to bind to the children leading to a NULL
pointer deref in video_get_max_state when registering the cooling-dev:

[    3.148958] BUG: kernel NULL pointer dereference, address: 0000000000000090
<snip>
[    3.149015] Hardware name: Sony Corporation VPCSB2X9R/VAIO, BIOS R2087H4 06/15/2012
[    3.149021] RIP: 0010:video_get_max_state+0x17/0x30 [video]
<snip>
[    3.149105] Call Trace:
[    3.149110]  <TASK>
[    3.149114]  ? __die+0x23/0x70
[    3.149126]  ? page_fault_oops+0x171/0x4e0
[    3.149137]  ? exc_page_fault+0x7f/0x180
[    3.149147]  ? asm_exc_page_fault+0x26/0x30
[    3.149158]  ? video_get_max_state+0x17/0x30 [video 9b6f3f0d19d7b4a0e2df17a2d8b43bc19c2ed71f]
[    3.149176]  ? __pfx_video_get_max_state+0x10/0x10 [video 9b6f3f0d19d7b4a0e2df17a2d8b43bc19c2ed71f]
[    3.149192]  __thermal_cooling_device_register.part.0+0xf2/0x2f0
[    3.149205]  acpi_video_bus_register_backlight.part.0.isra.0+0x414/0x570 [video 9b6f3f0d19d7b4a0e2df17a2d8b43bc19c2ed71f]
[    3.149227]  acpi_video_register_backlight+0x57/0x80 [video 9b6f3f0d19d7b4a0e2df17a2d8b43bc19c2ed71f]
[    3.149245]  intel_acpi_video_register+0x68/0x90 [i915 1f3a758130b32ef13d301d4f8f78c7d766d57f2a]
[    3.149669]  intel_display_driver_register+0x28/0x50 [i915 1f3a758130b32ef13d301d4f8f78c7d766d57f2a]
[    3.150064]  i915_driver_probe+0x790/0xb90 [i915 1f3a758130b32ef13d301d4f8f78c7d766d57f2a]
[    3.150402]  local_pci_probe+0x45/0xa0
[    3.150412]  pci_device_probe+0xc1/0x260
<snip>

Fix this by directly using the acpi_video_device as devdata for
the cooling-device, which avoids the need to set driver-data on
the children at all.

Fixes: 0d16710146a1 ("ACPI: bus: Set driver_data to NULL every time .add() fails")
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/9718
Cc: 6.6+ <stable@vger.kernel.org> # 6.6+
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

r8169: fix deadlock on RTL8125 in jumbo mtu mode

commit 59d395ed606d8df14615712b0cdcdadb2d962175 upstream.

The original change results in a deadlock if jumbo mtu mode is used.
Reason is that the phydev lock is held when rtl_reset_work() is called
here, and rtl_jumbo_config() calls phy_start_aneg() which also tries
to acquire the phydev lock. Fix this by calling rtl_reset_work()
asynchronously.

Fixes: 621735f59064 ("r8169: fix rare issue with broken rx after link-down on RTL8125")
Reported-by: Ian Chen <free122448@hotmail.com>
Tested-by: Ian Chen <free122448@hotmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/caf6a487-ef8c-4570-88f9-f47a659faf33@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nvme: check for valid nvme_identify_ns() before using it

commit d8b90d600aff181936457f032d116dbd8534db06 upstream.

When scanning namespaces, it is possible to get valid data from the first
call to nvme_identify_ns() in nvme_alloc_ns(), but not from the second
call in nvme_update_ns_info_block().  In particular, if the NSID becomes
inactive between the two commands, a storage device may return a buffer
filled with zero as per 4.1.5.1.  In this case, we can get a kernel crash
due to a divide-by-zero in blk_stack_limits() because ns->lba_shift will
be set to zero.

PID: 326      TASK: ffff95fec3cd8000  CPU: 29   COMMAND: "kworker/u98:10"
 #0 [ffffad8f8702f9e0] machine_kexec at ffffffff91c76ec7
 #1 [ffffad8f8702fa38] __crash_kexec at ffffffff91dea4fa
 #2 [ffffad8f8702faf8] crash_kexec at ffffffff91deb788
 #3 [ffffad8f8702fb00] oops_end at ffffffff91c2e4bb
 #4 [ffffad8f8702fb20] do_trap at ffffffff91c2a4ce
 #5 [ffffad8f8702fb70] do_error_trap at ffffffff91c2a595
 #6 [ffffad8f8702fbb0] exc_divide_error at ffffffff928506e6
 #7 [ffffad8f8702fbd0] asm_exc_divide_error at ffffffff92a00926
    [exception RIP: blk_stack_limits+434]
    RIP: ffffffff92191872  RSP: ffffad8f8702fc80  RFLAGS: 00010246
    RAX: 0000000000000000  RBX: ffff95efa0c91800  RCX: 0000000000000001
    RDX: 0000000000000000  RSI: 0000000000000001  RDI: 0000000000000001
    RBP: 00000000ffffffff   R8: ffff95fec7df35a8   R9: 0000000000000000
    R10: 0000000000000000  R11: 0000000000000001  R12: 0000000000000000
    R13: 0000000000000000  R14: 0000000000000000  R15: ffff95fed33c09a8
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #8 [ffffad8f8702fce0] nvme_update_ns_info_block at ffffffffc06d3533 [nvme_core]
 #9 [ffffad8f8702fd18] nvme_scan_ns at ffffffffc06d6fa7 [nvme_core]

This happened when the check for valid data was moved out of nvme_identify_ns()
into one of the callers.  Fix this by checking in both callers.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=218186
Fixes: 0dd6fff2aad4 ("nvme: bring back auto-removal of deleted namespaces during sequential scan")
Cc: stable@vger.kernel.org
Signed-off-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm verity: don't perform FEC for failed readahead IO

commit 0193e3966ceeeef69e235975918b287ab093082b upstream.

We found an issue under Android OTA scenario that many BIOs have to do
FEC where the data under dm-verity is 100% complete and no corruption.

Android OTA has many dm-block layers, from upper to lower:
dm-verity
dm-snapshot
dm-origin & dm-cow
dm-linear
ufs

DM tables have to change 2 times during Android OTA merging process.
When doing table change, the dm-snapshot will be suspended for a while.
During this interval, many readahead IOs are submitted to dm_verity
from filesystem. Then the kverity works are busy doing FEC process
which cost too much time to finish dm-verity IO. This causes needless
delay which feels like system is hung.

After adding debugging it was found that each readahead IO needed
around 10s to finish when this situation occurred. This is due to IO
amplification:

dm-snapshot suspend
erofs_readahead     // 300+ io is submitted
dm_submit_bio (dm_verity)
dm_submit_bio (dm_snapshot)
bio return EIO
bio got nothing, it's empty
verity_end_io
verity_verify_io
forloop range(0, io->n_blocks)    // each io->nblocks ~= 20
verity_fec_decode
fec_decode_rsb
fec_read_bufs
forloop range(0, v->fec->rsn) // v->fec->rsn = 253
new_read
submit_bio (dm_snapshot)
end loop
end loop
dm-snapshot resume

Readahead BIOs get nothing while dm-snapshot is suspended, so all of
them will cause verity's FEC.
Each readahead BIO needs to verify ~20 (io->nblocks) blocks.
Each block needs to do FEC, and every block needs to do 253
(v->fec->rsn) reads.
So during the suspend interval(~200ms), 300 readahead BIOs trigger
~1518000 (300*20*253) IOs to dm-snapshot.

As readahead IO is not required by userspace, and to fix this issue,
it is best to pass readahead errors to upper layer to handle it.

Cc: stable@vger.kernel.org
Fixes: a739ff3f543a ("dm verity: add support for forward error correction")
Signed-off-by: Wu Bo <bo.wu@vivo.com>
Reviewed-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm verity: initialize fec io before freeing it

commit 7be05bdfb4efc1396f7692562c7161e2b9f595f1 upstream.

If BIO error, verity_end_io() can call verity_finish_io() before
verity_fec_init_io(). Therefore, fec_io->rs is not initialized and
may crash when doing memory freeing in verity_fec_finish_io().

Crash call stack:
 die+0x90/0x2b8
 __do_kernel_fault+0x260/0x298
 do_bad_area+0x2c/0xdc
 do_translation_fault+0x3c/0x54
 do_mem_abort+0x54/0x118
 el1_abort+0x38/0x5c
 el1h_64_sync_handler+0x50/0x90
 el1h_64_sync+0x64/0x6c
 free_rs+0x18/0xac
 fec_rs_free+0x10/0x24
 mempool_free+0x58/0x148
 verity_fec_finish_io+0x4c/0xb0
 verity_end_io+0xb8/0x150

Cc: stable@vger.kernel.org # v6.0+
Fixes: 5721d4e5a9cd ("dm verity: Add optional "try_verify_in_tasklet" feature")
Signed-off-by: Wu Bo <bo.wu@vivo.com>
Reviewed-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: force toggle rate wa for first link training for a retimer

commit eb28018943fed7639dfea1c9ec9c756ec692b99a upstream.

[WHY]
Handover from DMUB to driver does not perform link rate toggle.
It might cause link training failure for boot up.

[HOW]
Force toggle rate wa for first link train.
link->vendor_specific_lttpr_link_rate_wa should be zero then.

Cc: stable@vger.kernel.org # 6.1+
Reviewed-by: Michael Strauss <michael.strauss@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Zhongwei <zhongwei.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: fix ABM disablement

commit b9f46f0b98784e40288ee393f863f553fde062fa upstream.

On recent versions of DMUB firmware, if we want to completely disable
ABM we have to pass ABM_LEVEL_IMMEDIATE_DISABLE as the requested ABM
level to DMUB. Otherwise, LCD eDP displays are unable to reach their
maximum brightness levels. So, to fix this whenever the user requests an
ABM level of 0 pass ABM_LEVEL_IMMEDIATE_DISABLE to DMUB instead. Also,
to keep the user's experience consistent map ABM_LEVEL_IMMEDIATE_DISABLE
to 0 when a user tries to read the requested ABM level.

Cc: stable@vger.kernel.org # 6.1+
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Update min Z8 residency time to 2100 for DCN314

commit 4636a211980052ca0df90265c8a3ed2d46099091 upstream.

[Why]
Some panels with residency period of 2054 exhibit flickering with
Z8 at the end of the frame.

[How]
As a workaround, increase the limit to block these panels.

Cc: stable@vger.kernel.org # 6.1+
Reviewed-by: Syed Hassan <syed.hassan@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/display: Use DRAM speed from validation for dummy p-state

commit 9be601135ba8ac69880c01606c82140f2dde105e upstream.

[Description]
When choosing which dummy p-state latency to use, we
need to use the DRAM speed from validation. The DRAMSpeed
DML variable can change because we use different input
params to DML when populating watermarks set B.

Cc: stable@vger.kernel.org # 6.1+
Reviewed-by: Samson Tam <samson.tam@amd.com>
Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>