]> Git Repo - linux.git/log
linux.git
7 months agoMerge tag 'riscv-for-linus-6.11-mw2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 Jul 2024 17:14:34 +0000 (10:14 -0700)]
Merge tag 'riscv-for-linus-6.11-mw2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux

Pull more RISC-V updates from Palmer Dabbelt:

 - Support for NUMA (via SRAT and SLIT), console output (via SPCR), and
   cache info (via PPTT) on ACPI-based systems.

 - The trap entry/exit code no longer breaks the return address stack
   predictor on many systems, which results in an improvement to trap
   latency.

 - Support for HAVE_ARCH_STACKLEAK.

 - The sv39 linear map has been extended to support 128GiB mappings.

 - The frequency of the mtime CSR is now visible via hwprobe.

* tag 'riscv-for-linus-6.11-mw2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux: (21 commits)
  RISC-V: Provide the frequency of time CSR via hwprobe
  riscv: Extend sv39 linear mapping max size to 128G
  riscv: enable HAVE_ARCH_STACKLEAK
  riscv: signal: Remove unlikely() from WARN_ON() condition
  riscv: Improve exception and system call latency
  RISC-V: Select ACPI PPTT drivers
  riscv: cacheinfo: initialize cacheinfo's level and type from ACPI PPTT
  riscv: cacheinfo: remove the useless input parameter (node) of ci_leaf_init()
  RISC-V: ACPI: Enable SPCR table for console output on RISC-V
  riscv: boot: remove duplicated targets line
  trace: riscv: Remove deprecated kprobe on ftrace support
  riscv: cpufeature: Extract common elements from extension checking
  riscv: Introduce vendor variants of extension helpers
  riscv: Add vendor extensions to /proc/cpuinfo
  riscv: Extend cpufeature.c to detect vendor extensions
  RISC-V: run savedefconfig for defconfig
  RISC-V: hwprobe: sort EXT_KEY()s in hwprobe_isa_ext0() alphabetically
  ACPI: NUMA: replace pr_info with pr_debug in arch_acpi_numa_init
  ACPI: NUMA: change the ACPI_NUMA to a hidden option
  ACPI: NUMA: Add handler for SRAT RINTC affinity structure
  ...

7 months agoMerge tag 'for-linus-6.11-rc1a-tag' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 Jul 2024 16:58:24 +0000 (09:58 -0700)]
Merge tag 'for-linus-6.11-rc1a-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:
 "Two fixes for issues introduced in this merge window:

   - fix enhanced debugging in the Xen multicall handling

   - two patches fixing a boot failure when running as dom0 in PVH mode"

* tag 'for-linus-6.11-rc1a-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  x86/xen: fix memblock_reserve() usage on PVH
  x86/xen: move xen_reserve_extra_memory()
  xen: fix multicall debug data referencing

7 months agominmax: avoid overly complicated constant expressions in VM code
Linus Torvalds [Fri, 26 Jul 2024 22:32:27 +0000 (15:32 -0700)]
minmax: avoid overly complicated constant expressions in VM code

The minmax infrastructure is overkill for simple constants, and can
cause huge expansions because those simple constants are then used by
other things.

For example, 'pageblock_order' is a core VM constant, but because it was
implemented using 'min_t()' and all the type-checking that involves, it
actually expanded to something like 2.5kB of preprocessor noise.

And when that simple constant was then used inside other expansions:

  #define pageblock_nr_pages      (1UL << pageblock_order)
  #define pageblock_start_pfn(pfn)  ALIGN_DOWN((pfn), pageblock_nr_pages)

and we then use that inside a 'max()' macro:

case ISOLATE_SUCCESS:
update_cached = false;
last_migrated_pfn = max(cc->zone->zone_start_pfn,
pageblock_start_pfn(cc->migrate_pfn - 1));

the end result was that one statement expanding to 253kB in size.

There are probably other cases of this, but this one case certainly
stood out.

I've added 'MIN_T()' and 'MAX_T()' macros for this kind of "core simple
constant with specific type" use.  These macros skip the type checking,
and as such need to be very sparingly used only for obvious cases that
have active issues like this.

Reported-by: Lorenzo Stoakes <[email protected]>
Link: https://lore.kernel.org/all/[email protected]/
Cc: David Laight <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
7 months agominmax: avoid overly complex min()/max() macro arguments in xen
Linus Torvalds [Fri, 26 Jul 2024 22:09:07 +0000 (15:09 -0700)]
minmax: avoid overly complex min()/max() macro arguments in xen

We have some very fancy min/max macros that have tons of sanity checking
to warn about mixed signedness etc.

This is all things that a sane compiler should warn about, but there are
no sane compiler interfaces for this, and '-Wsign-compare' is broken [1]
and not useful.

So then we compensate (some would say over-compensate) by doing the
checks manually with some truly horrid macro games.

And no, we can't just use __builtin_types_compatible_p(), because the
whole question of "does it make sense to compare these two values" is a
lot more complicated than that.

For example, it makes a ton of sense to compare unsigned values with
simple constants like "5", even if that is indeed a signed type.  So we
have these very strange macros to try to make sensible type checking
decisions on the arguments to 'min()' and 'max()'.

But that can cause enormous code expansion if the min()/max() macros are
used with complicated expressions, and particularly if you nest these
things so that you get the first big expansion then expanded again.

The xen setup.c file ended up ballooning to over 50MB of preprocessed
noise that takes 15s to compile (obviously depending on the build host),
largely due to one single line.

So let's split that one single line to just be simpler.  I think it ends
up being more legible to humans too at the same time.  Now that single
file compiles in under a second.

Reported-and-reviewed-by: Lorenzo Stoakes <[email protected]>
Link: https://lore.kernel.org/all/[email protected]/
Link: https://staticthinking.wordpress.com/2023/07/25/wsign-compare-is-garbage/
Cc: David Laight <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
7 months agoMerge tag 'auxdisplay-for-v6.11-tag1' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 26 Jul 2024 18:04:28 +0000 (11:04 -0700)]
Merge tag 'auxdisplay-for-v6.11-tag1' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k

Pull auxdisplay updates from Geert Uytterhoeven:

  - add support for configuring the boot message on line displays

  - miscellaneous fixes and improvements

* tag 'auxdisplay-for-v6.11-tag1' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k:
  auxdisplay: ht16k33: Drop reference after LED registration
  auxdisplay: Use sizeof(*pointer) instead of sizeof(type)
  auxdisplay: hd44780: add missing MODULE_DESCRIPTION() macro
  auxdisplay: linedisp: add missing MODULE_DESCRIPTION() macro
  auxdisplay: linedisp: Support configuring the boot message
  auxdisplay: charlcd: Provide a forward declaration

7 months agoMerge tag 'sound-fix-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 26 Jul 2024 18:01:31 +0000 (11:01 -0700)]
Merge tag 'sound-fix-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A collection of fixes gathered since the previous pull.

  We see a bit large LOCs at a HD-audio quirk, but that's only bulk COEF
  data, hence it's safe to take. In addition to that, there were two
  minor fixes for MIDI 2.0 handling for ALSA core, and the rest are all
  rather random small and device-specific fixes"

* tag 'sound-fix-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ASoC: fsl-asoc-card: Dynamically allocate memory for snd_soc_dai_link_components
  ASoC: amd: yc: Support mic on Lenovo Thinkpad E16 Gen 2
  ALSA: hda/realtek: Implement sound init sequence for Samsung Galaxy Book3 Pro 360
  ALSA: hda/realtek: cs35l41: Fixup remaining asus strix models
  ASoC: SOF: ipc4-topology: Preserve the DMA Link ID for ChainDMA on unprepare
  ASoC: SOF: ipc4-topology: Only handle dai_config with HW_PARAMS for ChainDMA
  ALSA: ump: Force 1 Group for MIDI1 FBs
  ALSA: ump: Don't update FB name for static blocks
  ALSA: usb-audio: Add a quirk for Sonix HD USB Camera
  ASoC: TAS2781: Fix tasdev_load_calibrated_data()
  ASoC: tegra: select CONFIG_SND_SIMPLE_CARD_UTILS
  ASoC: Intel: use soc_intel_is_byt_cr() only when IOSF_MBI is reachable
  ALSA: usb-audio: Move HD Webcam quirk to the right place
  ALSA: hda: tas2781: mark const variables as __maybe_unused
  ALSA: usb-audio: Fix microphone sound on HD webcam.
  ASoC: sof: amd: fix for firmware reload failure in Vangogh platform
  ASoC: Intel: Fix RT5650 SSP lookup
  ASOC: SOF: Intel: hda-loader: only wait for HDaudio IOC for IPC4 devices
  ASoC: SOF: imx8m: Fix DSP control regmap retrieval

7 months agoMerge tag 'drm-next-2024-07-26' of https://gitlab.freedesktop.org/drm/kernel
Linus Torvalds [Fri, 26 Jul 2024 17:57:07 +0000 (10:57 -0700)]
Merge tag 'drm-next-2024-07-26' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "Fixes for rc1, mostly amdgpu, i915 and xe, with some other misc ones,
  doesn't seem to be anything too serious.

  amdgpu:
   - Bump driver version for GFX12 DCC
   - DC documention warning fixes
   - VCN unified queue power fix
   - SMU fix
   - RAS fix
   - Display corruption fix
   - SDMA 5.2 workaround
   - GFX12 fixes
   - Uninitialized variable fix
   - VCN/JPEG 4.0.3 fixes
   - Misc display fixes
   - RAS fixes
   - VCN4/5 harvest fix
   - GPU reset fix

  i915:
   - Reset intel_dp->link_trained before retraining the link
   - Don't switch the LTTPR mode on an active link
   - Do not consider preemption during execlists_dequeue for gen8
   - Allow NULL memory region

  xe:
   - xe_exec ioctl minor fix on sync entry cleanup upon error
   - SRIOV: limit VF LMEM provisioning
   - Wedge mode fixes

  v3d:
   - fix indirect dispatch on newer v3d revs

  panel:
   - fix panel backlight bindings"

* tag 'drm-next-2024-07-26' of https://gitlab.freedesktop.org/drm/kernel: (39 commits)
  drm/amdgpu: reset vm state machine after gpu reset(vram lost)
  drm/amdgpu: add missed harvest check for VCN IP v4/v5
  drm/amdgpu: Fix eeprom max record count
  drm/amdgpu: fix ras UE error injection failure issue
  drm/amd/display: Remove ASSERT if significance is zero in math_ceil2
  drm/amd/display: Check for NULL pointer
  drm/amdgpu/vcn: Use offsets local to VCN/JPEG in VF
  drm/amdgpu: Add empty HDP flush function to VCN v4.0.3
  drm/amdgpu: Add empty HDP flush function to JPEG v4.0.3
  drm/amd/amdgpu: Fix uninitialized variable warnings
  drm/amdgpu: Fix atomics on GFX12
  drm/amdgpu/sdma5.2: Update wptr registers as well as doorbell
  drm/i915: Allow NULL memory region
  drm/i915/gt: Do not consider preemption during execlists_dequeue for gen8
  dt-bindings: display: panel: samsung,atna33xc20: Document ATNA45AF01
  drm/xe: Don't suspend device upon wedge
  drm/xe: Wedge the entire device
  drm/xe/pf: Limit fair VF LMEM provisioning
  drm/xe/exec: Fix minor bug related to xe_sync_entry_cleanup
  drm/amd/display: fix corruption with high refresh rates on DCN 3.0
  ...

7 months agoMerge tag 's390-6.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Fri, 26 Jul 2024 17:47:53 +0000 (10:47 -0700)]
Merge tag 's390-6.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux

Pull more s390 updates from Vasily Gorbik:

 - Fix KMSAN build breakage caused by the conflict between s390 and
   mm-stable trees

 - Add KMSAN page markers for ptdump

 - Add runtime constant support

 - Fix __pa/__va for modules under non-GPL licenses by exporting
   necessary vm_layout struct with EXPORT_SYMBOL to prevent linkage
   problems

 - Fix an endless loop in the CF_DIAG event stop in the CPU Measurement
   Counter Facility code when the counter set size is zero

 - Remove the PROTECTED_VIRTUALIZATION_GUEST config option and enable
   its functionality by default

 - Support allocation of multiple MSI interrupts per device and improve
   logging of architecture-specific limitations

 - Add support for lowcore relocation as a debugging feature to catch
   all null ptr dereferences in the kernel address space, improving
   detection beyond the current implementation's limited write access
   protection

 - Clean up and rework CPU alternatives to allow for callbacks and early
   patching for the lowcore relocation

* tag 's390-6.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: (39 commits)
  s390: Remove protvirt and kvm config guards for uv code
  s390/boot: Add cmdline option to relocate lowcore
  s390/kdump: Make kdump ready for lowcore relocation
  s390/entry: Make system_call() ready for lowcore relocation
  s390/entry: Make ret_from_fork() ready for lowcore relocation
  s390/entry: Make __switch_to() ready for lowcore relocation
  s390/entry: Make restart_int_handler() ready for lowcore relocation
  s390/entry: Make mchk_int_handler() ready for lowcore relocation
  s390/entry: Make int handlers ready for lowcore relocation
  s390/entry: Make pgm_check_handler() ready for lowcore relocation
  s390/entry: Add base register to CHECK_VMAP_STACK/CHECK_STACK macro
  s390/entry: Add base register to SIEEXIT macro
  s390/entry: Add base register to MBEAR macro
  s390/entry: Make __sie64a() ready for lowcore relocation
  s390/head64: Make startup code ready for lowcore relocation
  s390: Add infrastructure to patch lowcore accesses
  s390/atomic_ops: Disable flag outputs constraint for GCC versions below 14.2.0
  s390/entry: Move SIE indicator flag to thread info
  s390/nmi: Simplify ptregs setup
  s390/alternatives: Remove alternative facility list
  ...

7 months agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 26 Jul 2024 17:39:10 +0000 (10:39 -0700)]
Merge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:
 "The usual summary below, but the main fix is for the fast GUP lockless
  page-table walk when we have a combination of compile-time and
  run-time folding of the p4d and the pud respectively.

   - Remove some redundant Kconfig conditionals

   - Fix string output in ptrace selftest

   - Fix fast GUP crashes in some page-table configurations

   - Remove obsolete linker option when building the vDSO

   - Fix some sysreg field definitions for the GIC"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: mm: Fix lockless walks with static and dynamic page-table folding
  arm64/sysreg: Correct the values for GICv4.1
  arm64/vdso: Remove --hash-style=sysv
  kselftest: missing arg in ptrace.c
  arm64/Kconfig: Remove redundant 'if HAVE_FUNCTION_GRAPH_TRACER'
  arm64: remove redundant 'if HAVE_ARCH_KASAN' in Kconfig

7 months agoMerge tag 'ceph-for-6.11-rc1' of https://github.com/ceph/ceph-client
Linus Torvalds [Fri, 26 Jul 2024 17:34:42 +0000 (10:34 -0700)]
Merge tag 'ceph-for-6.11-rc1' of https://github.com/ceph/ceph-client

Pull ceph updates from Ilya Dryomov:
 "A small patchset to address bogus I/O errors and ultimately an
  assertion failure in the face of watch errors with -o exclusive
  mappings in RBD marked for stable and some assorted CephFS fixes"

* tag 'ceph-for-6.11-rc1' of https://github.com/ceph/ceph-client:
  rbd: don't assume rbd_is_lock_owner() for exclusive mappings
  rbd: don't assume RBD_LOCK_STATE_LOCKED for exclusive mappings
  rbd: rename RBD_LOCK_STATE_RELEASING and releasing_wait
  ceph: fix incorrect kmalloc size of pagevec mempool
  ceph: periodically flush the cap releases
  ceph: convert comma to semicolon in __ceph_dentry_dir_lease_touch()
  ceph: use cap_wait_list only if debugfs is enabled

7 months agoMerge tag 'erofs-for-6.11-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 26 Jul 2024 17:31:03 +0000 (10:31 -0700)]
Merge tag 'erofs-for-6.11-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs

Pull more erofs updates from Gao Xiang:

 - Support STATX_DIOALIGN and FS_IOC_GETFSSYSFSPATH

 - Fix a race of LZ4 decompression due to recent refactoring

 - Another multi-page folio adaption in erofs_bread()

* tag 'erofs-for-6.11-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: convert comma to semicolon
  erofs: support multi-page folios for erofs_bread()
  erofs: add support for FS_IOC_GETFSSYSFSPATH
  erofs: fix race in z_erofs_get_gbuf()
  erofs: support STATX_DIOALIGN

7 months agoMerge tag 'pull-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Fri, 26 Jul 2024 17:26:33 +0000 (10:26 -0700)]
Merge tag 'pull-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull struct file leak fixes from Al Viro:
 "a couple of leaks on failure exits missing fdput()"

* tag 'pull-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  lirc: rc_dev_get_from_fd(): fix file leak
  powerpc: fix a file leak in kvm_vcpu_ioctl_enable_cap()

7 months agoarm64: allow installing compressed image by default
Linus Torvalds [Tue, 23 Jul 2024 20:21:02 +0000 (13:21 -0700)]
arm64: allow installing compressed image by default

On arm64 we build compressed images, but "make install" by default will
install the old non-compressed one.  To actually get the compressed
image install, you need to use "make zinstall", which is not the usual
way to install a kernel.

Which may not sound like much of an issue, but when you deal with
multiple architectures (and years of your fingers knowing the regular
"make install" incantation), this inconsistency is pretty annoying.

But as Will Deacon says:
 "Sadly, bootloaders being as top quality as you might expect, I don't
  think we're in a position to rely on decompressor support across the
  board. Our Image.gz is literally just that -- we don't have a built-in
  decompressor (nor do I think we want to rush into that again after the
  fun we had on arm32) and the recent EFI zboot support solves that
  problem for platforms using EFI.

  Changing the default 'install' target terrifies me. There are bound to
  be folks with embedded boards who've scripted this and we could really
  ruin their day if we quietly give them a compressed kernel that their
  bootloader doesn't know how to handle :/"

So make this conditional on a new "COMPRESSED_INSTALL" option.

Cc: Catalin Marinas <[email protected]>
Acked-by: Will Deacon <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
7 months agoMerge tag 'bitmap-6.11-rc1' of https://github.com:/norov/linux
Linus Torvalds [Fri, 26 Jul 2024 16:50:36 +0000 (09:50 -0700)]
Merge tag 'bitmap-6.11-rc1' of https://github.com:/norov/linux

Pull bitmap updates from Yury Norov:
 "Random fixes"

* tag 'bitmap-6.11-rc1' of https://github.com:/norov/linux:
  riscv: Remove unnecessary int cast in variable_fls()
  radix tree test suite: put definition of bitmap_clear() into lib/bitmap.c
  bitops: Add a comment explaining the double underscore macros
  lib: bitmap: add missing MODULE_DESCRIPTION() macros
  cpumask: introduce assign_cpu() macro

7 months agoRISC-V: Provide the frequency of time CSR via hwprobe
Palmer Dabbelt [Tue, 2 Jul 2024 03:37:31 +0000 (11:37 +0800)]
RISC-V: Provide the frequency of time CSR via hwprobe

The RISC-V architecture makes a real time counter CSR (via RDTIME
instruction) available for applications in U-mode but there is no
architected mechanism for an application to discover the frequency
the counter is running at. Some applications (e.g., DPDK) use the
time counter for basic performance analysis as well as fine grained
time-keeping.

Add support to the hwprobe system call to export the time CSR
frequency to code running in U-mode.

Signed-off-by: Yunhui Cui <[email protected]>
Reviewed-by: Evan Green <[email protected]>
Reviewed-by: Anup Patel <[email protected]>
Acked-by: Punit Agrawal <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoriscv: Extend sv39 linear mapping max size to 128G
Stuart Menefy [Sun, 30 Jun 2024 11:05:49 +0000 (12:05 +0100)]
riscv: Extend sv39 linear mapping max size to 128G

This harmonizes all virtual addressing modes which can now all map
(PGDIR_SIZE * PTRS_PER_PGD) / 4 of physical memory.

The RISCV implementation of KASAN requires that the boundary between
shallow mappings are aligned on an 8G boundary. In this case we need
VMALLOC_START to be 8G aligned. So although we only need to move the
start of the linear mapping down by 4GiB to allow 128GiB to be mapped,
we actually move it down by 8GiB (creating a 4GiB hole between the
linear mapping and KASAN shadow space) to maintain the alignment
requirement.

Signed-off-by: Stuart Menefy <[email protected]>
Reviewed-by: Alexandre Ghiti <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoMerge patch series "RISC-V: Select ACPI PPTT drivers"
Palmer Dabbelt [Wed, 24 Jul 2024 14:39:39 +0000 (07:39 -0700)]
Merge patch series "RISC-V: Select ACPI PPTT drivers"

This series adds support for ACPI PPTT via cacheinfo.

* b4-shazam-merge:
  RISC-V: Select ACPI PPTT drivers
  riscv: cacheinfo: initialize cacheinfo's level and type from ACPI PPTT
  riscv: cacheinfo: remove the useless input parameter (node) of ci_leaf_init()

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoMerge patch "Enable SPCR table for console output on RISC-V"
Palmer Dabbelt [Wed, 24 Jul 2024 14:33:41 +0000 (07:33 -0700)]
Merge patch "Enable SPCR table for console output on RISC-V"

Sia Jee Heng <[email protected]> says:

The ACPI SPCR code has been used to enable console output for ARM64 and
X86. The same code can be reused for RISC-V. Furthermore, SPCR table is
mandated for headless system as outlined in the RISC-V BRS
Specification, chapter 6.

* b4-shazam-merge:
  RISC-V: ACPI: Enable SPCR table for console output on RISC-V

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoriscv: enable HAVE_ARCH_STACKLEAK
Jisheng Zhang [Sun, 23 Jun 2024 23:53:16 +0000 (07:53 +0800)]
riscv: enable HAVE_ARCH_STACKLEAK

Add support for the stackleak feature. Whenever the kernel returns to user
space the kernel stack is filled with a poison value.

At the same time, disables the plugin in EFI stub code because EFI stub
is out of scope for the protection.

Tested on qemu and milkv duo:
/ # echo STACKLEAK_ERASING > /sys/kernel/debug/provoke-crash/DIRECT
[   38.675575] lkdtm: Performing direct entry STACKLEAK_ERASING
[   38.678448] lkdtm: stackleak stack usage:
[   38.678448]   high offset: 288 bytes
[   38.678448]   current:     496 bytes
[   38.678448]   lowest:      1328 bytes
[   38.678448]   tracked:     1328 bytes
[   38.678448]   untracked:   448 bytes
[   38.678448]   poisoned:    14312 bytes
[   38.678448]   low offset:  8 bytes
[   38.689887] lkdtm: OK: the rest of the thread stack is properly erased

Signed-off-by: Jisheng Zhang <[email protected]>
Reviewed-by: Charlie Jenkins <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoriscv: signal: Remove unlikely() from WARN_ON() condition
Zhongqiu Han [Thu, 20 Jun 2024 03:34:34 +0000 (11:34 +0800)]
riscv: signal: Remove unlikely() from WARN_ON() condition

"WARN_ON(unlikely(x))" is excessive. WARN_ON() already uses unlikely()
internally.

Signed-off-by: Zhongqiu Han <[email protected]>
Reviewed-by: Bjorn Andersson <[email protected]>
Reviewed-by: Andy Chiu <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoriscv: Improve exception and system call latency
Anton Blanchard [Fri, 7 Jun 2024 06:13:35 +0000 (23:13 -0700)]
riscv: Improve exception and system call latency

Many CPUs implement return address branch prediction as a stack. The
RISCV architecture refers to this as a return address stack (RAS). If
this gets corrupted then the CPU will mispredict at least one but
potentally many function returns.

There are two issues with the current RISCV exception code:

- We are using the alternate link stack (x5/t0) for the indirect branch
  which makes the hardware think this is a function return. This will
  corrupt the RAS.

- We modify the return address of handle_exception to point to
  ret_from_exception. This will also corrupt the RAS.

Testing the null system call latency before and after the patch:

Visionfive2 (StarFive JH7110 / U74)
baseline: 189.87 ns
patched:  176.76 ns

Lichee pi 4a (T-Head TH1520 / C910)
baseline: 666.58 ns
patched:  636.90 ns

Just over 7% on the U74 and just over 4% on the C910.

Signed-off-by: Anton Blanchard <[email protected]>
Signed-off-by: Cyril Bur <[email protected]>
Tested-by: Jisheng Zhang <[email protected]>
Reviewed-by: Jisheng Zhang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoerofs: convert comma to semicolon
Chen Ni [Wed, 24 Jul 2024 02:07:21 +0000 (10:07 +0800)]
erofs: convert comma to semicolon

Replace a comma between expression statements by a semicolon.

Signed-off-by: Chen Ni <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Chao Yu <[email protected]>
Signed-off-by: Gao Xiang <[email protected]>
7 months agoerofs: support multi-page folios for erofs_bread()
Gao Xiang [Tue, 23 Jul 2024 07:30:24 +0000 (15:30 +0800)]
erofs: support multi-page folios for erofs_bread()

If the requested page is part of the previous multi-page folio, there
is no need to call read_mapping_folio() again.

Also, get rid of the remaining one of page->index [1] in our codebase.

[1] https://lore.kernel.org/r/[email protected]

Cc: Matthew Wilcox <[email protected]>
Reviewed-by: Chao Yu <[email protected]>
Signed-off-by: Gao Xiang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
7 months agoerofs: add support for FS_IOC_GETFSSYSFSPATH
Huang Xiaojia [Sat, 20 Jul 2024 08:23:35 +0000 (16:23 +0800)]
erofs: add support for FS_IOC_GETFSSYSFSPATH

FS_IOC_GETFSSYSFSPATH ioctl exposes /sys/fs path of a given filesystem,
potentially standarizing sysfs reporting. This patch add support for
FS_IOC_GETFSSYSFSPATH for erofs, "erofs/<dev>" will be outputted for bdev
cases, "erofs/[domain_id,]<fs_id>" will be outputted for fscache cases.

Signed-off-by: Huang Xiaojia <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Chao Yu <[email protected]>
Signed-off-by: Gao Xiang <[email protected]>
7 months agoerofs: fix race in z_erofs_get_gbuf()
Gao Xiang [Mon, 22 Jul 2024 03:51:10 +0000 (11:51 +0800)]
erofs: fix race in z_erofs_get_gbuf()

In z_erofs_get_gbuf(), the current task may be migrated to another
CPU between `z_erofs_gbuf_id()` and `spin_lock(&gbuf->lock)`.

Therefore, z_erofs_put_gbuf() will trigger the following issue
which was found by stress test:

<2>[772156.434168] kernel BUG at fs/erofs/zutil.c:58!
..
<4>[772156.435007]
<4>[772156.439237] CPU: 0 PID: 3078 Comm: stress Kdump: loaded Tainted: G            E      6.10.0-rc7+ #2
<4>[772156.439239] Hardware name: Alibaba Cloud Alibaba Cloud ECS, BIOS 1.0.0 01/01/2017
<4>[772156.439241] pstate: 83400005 (Nzcv daif +PAN -UAO +TCO +DIT -SSBS BTYPE=--)
<4>[772156.439243] pc : z_erofs_put_gbuf+0x64/0x70 [erofs]
<4>[772156.439252] lr : z_erofs_lz4_decompress+0x600/0x6a0 [erofs]
..
<6>[772156.445958] stress (3127): drop_caches: 1
<4>[772156.446120] Call trace:
<4>[772156.446121]  z_erofs_put_gbuf+0x64/0x70 [erofs]
<4>[772156.446761]  z_erofs_lz4_decompress+0x600/0x6a0 [erofs]
<4>[772156.446897]  z_erofs_decompress_queue+0x740/0xa10 [erofs]
<4>[772156.447036]  z_erofs_runqueue+0x428/0x8c0 [erofs]
<4>[772156.447160]  z_erofs_readahead+0x224/0x390 [erofs]
..

Fixes: f36f3010f676 ("erofs: rename per-CPU buffers to global buffer pool and make it configurable")
Cc: <[email protected]> # 6.10+
Reviewed-by: Chunhai Guo <[email protected]>
Reviewed-by: Sandeep Dhavale <[email protected]>
Reviewed-by: Chao Yu <[email protected]>
Signed-off-by: Gao Xiang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
7 months agoerofs: support STATX_DIOALIGN
Hongbo Li [Thu, 18 Jul 2024 08:32:43 +0000 (16:32 +0800)]
erofs: support STATX_DIOALIGN

Add support for STATX_DIOALIGN to EROFS, so that direct I/O
alignment restrictions are exposed to userspace in a generic
way.

[Before]
```
./statx_test /mnt/erofs/testfile
statx(/mnt/erofs/testfile) = 0
dio mem align:0
dio offset align:0
```

[After]
```
./statx_test /mnt/erofs/testfile
statx(/mnt/erofs/testfile) = 0
dio mem align:512
dio offset align:512
```

Signed-off-by: Hongbo Li <[email protected]>
Reviewed-by: Chao Yu <[email protected]>
Signed-off-by: Gao Xiang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
7 months agoMerge tag 'amd-drm-fixes-6.11-2024-07-25' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Thu, 25 Jul 2024 23:51:59 +0000 (09:51 +1000)]
Merge tag 'amd-drm-fixes-6.11-2024-07-25' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-fixes-6.11-2024-07-25:

amdgpu:
- SDMA 5.2 workaround
- GFX12 fixes
- Uninitialized variable fix
- VCN/JPEG 4.0.3 fixes
- Misc display fixes
- RAS fixes
- VCN4/5 harvest fix
- GPU reset fix

Signed-off-by: Dave Airlie <[email protected]>
From: Alex Deucher <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
7 months agoMerge tag 'drm-misc-next-fixes-2024-07-25' of https://gitlab.freedesktop.org/drm...
Dave Airlie [Thu, 25 Jul 2024 23:51:10 +0000 (09:51 +1000)]
Merge tag 'drm-misc-next-fixes-2024-07-25' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next

A single fix for a panel compatible

Signed-off-by: Dave Airlie <[email protected]>
From: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/20240725-frisky-wren-of-tact-f5f504@houat
7 months agoMerge tag 'drm-intel-next-fixes-2024-07-25' of https://gitlab.freedesktop.org/drm...
Dave Airlie [Thu, 25 Jul 2024 20:41:02 +0000 (06:41 +1000)]
Merge tag 'drm-intel-next-fixes-2024-07-25' of https://gitlab.freedesktop.org/drm/i915/kernel into drm-next

- Do not consider preemption during execlists_dequeue for gen8 [gt] (Nitin Gote)
- Allow NULL memory region (Jonathan Cavitt)

Signed-off-by: Dave Airlie <[email protected]>
From: Tvrtko Ursulin <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/ZqICQzyzm/6hDWy4@linux
7 months agoMerge tag 'net-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 25 Jul 2024 20:32:25 +0000 (13:32 -0700)]
Merge tag 'net-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from bpf and netfilter.

  A lot of networking people were at a conference last week, busy
  catching COVID, so relatively short PR.

  Current release - regressions:

   - tcp: process the 3rd ACK with sk_socket for TFO and MPTCP

  Current release - new code bugs:

   - l2tp: protect session IDR and tunnel session list with one lock,
     make sure the state is coherent to avoid a warning

   - eth: bnxt_en: update xdp_rxq_info in queue restart logic

   - eth: airoha: fix location of the MBI_RX_AGE_SEL_MASK field

  Previous releases - regressions:

   - xsk: require XDP_UMEM_TX_METADATA_LEN to actuate tx_metadata_len,
     the field reuses previously un-validated pad

  Previous releases - always broken:

   - tap/tun: drop short frames to prevent crashes later in the stack

   - eth: ice: add a per-VF limit on number of FDIR filters

   - af_unix: disable MSG_OOB handling for sockets in sockmap/sockhash"

* tag 'net-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (34 commits)
  tun: add missing verification for short frame
  tap: add missing verification for short frame
  mISDN: Fix a use after free in hfcmulti_tx()
  gve: Fix an edge case for TSO skb validity check
  bnxt_en: update xdp_rxq_info in queue restart logic
  tcp: process the 3rd ACK with sk_socket for TFO/MPTCP
  selftests/bpf: Add XDP_UMEM_TX_METADATA_LEN to XSK TX metadata test
  xsk: Require XDP_UMEM_TX_METADATA_LEN to actuate tx_metadata_len
  bpf: Fix a segment issue when downgrading gso_size
  net: mediatek: Fix potential NULL pointer dereference in dummy net_device handling
  MAINTAINERS: make Breno the netconsole maintainer
  MAINTAINERS: Update bonding entry
  net: nexthop: Initialize all fields in dumped nexthops
  net: stmmac: Correct byte order of perfect_match
  selftests: forwarding: skip if kernel not support setting bridge fdb learning limit
  tipc: Return non-zero value from tipc_udp_addr2str() on error
  netfilter: nft_set_pipapo_avx2: disable softinterrupts
  ice: Fix recipe read procedure
  ice: Add a per-VF limit on number of FDIR filters
  net: bonding: correctly annotate RCU in bond_should_notify_peers()
  ...

7 months agoMerge tag 'printk-for-6.11-trivial' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 25 Jul 2024 20:18:41 +0000 (13:18 -0700)]
Merge tag 'printk-for-6.11-trivial' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux

Pull printk updates from Petr Mladek:

 - trivial printk changes

The bigger "real" printk work is still being discussed.

* tag 'printk-for-6.11-trivial' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux:
  vsprintf: add missing MODULE_DESCRIPTION() macro
  printk: Rename console_replay_all() and update context

7 months agoMerge tag 'constfy-sysctl-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 25 Jul 2024 19:58:36 +0000 (12:58 -0700)]
Merge tag 'constfy-sysctl-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/sysctl/sysctl

Pull sysctl constification from Joel Granados:
 "Treewide constification of the ctl_table argument of proc_handlers
  using a coccinelle script and some manual code formatting fixups.

  This is a prerequisite to moving the static ctl_table structs into
  read-only data section which will ensure that proc_handler function
  pointers cannot be modified"

* tag 'constfy-sysctl-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/sysctl/sysctl:
  sysctl: treewide: constify the ctl_table argument of proc_handlers

7 months agoMerge tag 'efi-fixes-for-v6.11-1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 25 Jul 2024 19:55:21 +0000 (12:55 -0700)]
Merge tag 'efi-fixes-for-v6.11-1' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi

Pull EFI fixes from Ard Biesheuvel:

 - Wipe screen_info after allocating it from the heap - used by arm32
   and EFI zboot, other EFI architectures allocate it statically

 - Revert to allocating boot_params from the heap on x86 when entering
   via the native PE entrypoint, to work around a regression on older
   Dell hardware

* tag 'efi-fixes-for-v6.11-1' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi:
  x86/efistub: Revert to heap allocated boot_params for PE entrypoint
  efi/libstub: Zero initialize heap allocated struct screen_info

7 months agoMerge tag 'kgdb-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt...
Linus Torvalds [Thu, 25 Jul 2024 19:48:42 +0000 (12:48 -0700)]
Merge tag 'kgdb-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt/linux

Pull kgdb updates from Daniel Thompson:
 "Three small changes this cycle:

   - Clean up an architecture abstraction that is no longer needed
     because all the architectures have converged.

   - Actually use the prompt argument to kdb_position_cursor() instead
     of ignoring it (functionally this fix is a nop but that was due to
     luck rather than good judgement)

   - Fix a -Wformat-security warning"

* tag 'kgdb-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt/linux:
  kdb: Get rid of redundant kdb_curr_task()
  kdb: Use the passed prompt in kdb_position_cursor()
  kdb: address -Wformat-security warnings

7 months agoMerge tag 'mips_6.11_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux
Linus Torvalds [Thu, 25 Jul 2024 19:41:53 +0000 (12:41 -0700)]
Merge tag 'mips_6.11_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux

Pull MIPS updates from Thomas Bogendoerfer:

 - Use improved timer sync for Loongson64

 - Fix address of GCR_ACCESS register

 - Add missing MODULE_DESCRIPTION

* tag 'mips_6.11_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  mips: sibyte: add missing MODULE_DESCRIPTION() macro
  MIPS: SMP-CPS: Fix address for GCR_ACCESS register for CM3 and later
  MIPS: Loongson64: Switch to SYNC_R4K

7 months agoMerge tag 'parisc-for-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 25 Jul 2024 19:37:42 +0000 (12:37 -0700)]
Merge tag 'parisc-for-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux

Pull parisc updates from Helge Deller:
 "The gettimeofday() and clock_gettime() syscalls are now available as
  vDSO functions, and Dave added a patch which allows to use NVMe cards
  in the PCI slots as fast and easy alternative to SCSI discs.

  Summary:

   - add gettimeofday() and clock_gettime() vDSO functions

   - enable PCI_MSI_ARCH_FALLBACKS to allow PCI to PCIe bridge adaptor
     with PCIe NVME card to function in parisc machines

   - allow users to reduce kernel unaligned runtime warnings

   - minor code cleanups"

* tag 'parisc-for-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Add support for CONFIG_SYSCTL_ARCH_UNALIGN_NO_WARN
  parisc: Use max() to calculate parisc_tlb_flush_threshold
  parisc: Fix warning at drivers/pci/msi/msi.h:121
  parisc: Add 64-bit gettimeofday() and clock_gettime() vDSO functions
  parisc: Add 32-bit gettimeofday() and clock_gettime() vDSO functions
  parisc: Clean up unistd.h file

7 months agoMerge tag 'uml-for-linus-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 25 Jul 2024 19:33:08 +0000 (12:33 -0700)]
Merge tag 'uml-for-linus-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/uml/linux

Pull UML updates from Richard Weinberger:

 - Support for preemption

 - i386 Rust support

 - Huge cleanup by Benjamin Berg

 - UBSAN support

 - Removal of dead code

* tag 'uml-for-linus-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/uml/linux: (41 commits)
  um: vector: always reset vp->opened
  um: vector: remove vp->lock
  um: register power-off handler
  um: line: always fill *error_out in setup_one_line()
  um: remove pcap driver from documentation
  um: Enable preemption in UML
  um: refactor TLB update handling
  um: simplify and consolidate TLB updates
  um: remove force_flush_all from fork_handler
  um: Do not flush MM in flush_thread
  um: Delay flushing syscalls until the thread is restarted
  um: remove copy_context_skas0
  um: remove LDT support
  um: compress memory related stub syscalls while adding them
  um: Rework syscall handling
  um: Add generic stub_syscall6 function
  um: Create signal stack memory assignment in stub_data
  um: Remove stub-data.h include from common-offsets.h
  um: time-travel: fix signal blocking race/hang
  um: time-travel: remove time_exit()
  ...

7 months agoMerge tag 'driver-core-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 25 Jul 2024 17:42:22 +0000 (10:42 -0700)]
Merge tag 'driver-core-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core

Pull driver core updates from Greg KH:
 "Here is the big set of driver core changes for 6.11-rc1.

  Lots of stuff in here, with not a huge diffstat, but apis are evolving
  which required lots of files to be touched. Highlights of the changes
  in here are:

   - platform remove callback api final fixups (Uwe took many releases
     to get here, finally!)

   - Rust bindings for basic firmware apis and initial driver-core
     interactions.

     It's not all that useful for a "write a whole driver in rust" type
     of thing, but the firmware bindings do help out the phy rust
     drivers, and the driver core bindings give a solid base on which
     others can start their work.

     There is still a long way to go here before we have a multitude of
     rust drivers being added, but it's a great first step.

   - driver core const api changes.

     This reached across all bus types, and there are some fix-ups for
     some not-common bus types that linux-next and 0-day testing shook
     out.

     This work is being done to help make the rust bindings more safe,
     as well as the C code, moving toward the end-goal of allowing us to
     put driver structures into read-only memory. We aren't there yet,
     but are getting closer.

   - minor devres cleanups and fixes found by code inspection

   - arch_topology minor changes

   - other minor driver core cleanups

  All of these have been in linux-next for a very long time with no
  reported problems"

* tag 'driver-core-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core: (55 commits)
  ARM: sa1100: make match function take a const pointer
  sysfs/cpu: Make crash_hotplug attribute world-readable
  dio: Have dio_bus_match() callback take a const *
  zorro: make match function take a const pointer
  driver core: module: make module_[add|remove]_driver take a const *
  driver core: make driver_find_device() take a const *
  driver core: make driver_[create|remove]_file take a const *
  firmware_loader: fix soundness issue in `request_internal`
  firmware_loader: annotate doctests as `no_run`
  devres: Correct code style for functions that return a pointer type
  devres: Initialize an uninitialized struct member
  devres: Fix memory leakage caused by driver API devm_free_percpu()
  devres: Fix devm_krealloc() wasting memory
  driver core: platform: Switch to use kmemdup_array()
  driver core: have match() callback in struct bus_type take a const *
  MAINTAINERS: add Rust device abstractions to DRIVER CORE
  device: rust: improve safety comments
  MAINTAINERS: add Danilo as FIRMWARE LOADER maintainer
  MAINTAINERS: add Rust FW abstractions to FIRMWARE LOADER
  firmware: rust: improve safety comments
  ...

7 months agoMerge tag 'linux-watchdog-6.11-rc1' of git://www.linux-watchdog.org/linux-watchdog
Linus Torvalds [Thu, 25 Jul 2024 17:18:35 +0000 (10:18 -0700)]
Merge tag 'linux-watchdog-6.11-rc1' of git://www.linux-watchdog.org/linux-watchdog

Pull watchdog updates from Wim Van Sebroeck:

 - make watchdog_class const

 - rework of the rzg2l_wdt driver

 - other small fixes and improvements

* tag 'linux-watchdog-6.11-rc1' of git://www.linux-watchdog.org/linux-watchdog:
  dt-bindings: watchdog: dlg,da9062-watchdog: Drop blank space
  watchdog: rzn1: Convert comma to semicolon
  watchdog: lenovo_se10_wdt: Convert comma to semicolon
  dt-bindings: watchdog: renesas,wdt: Document RZ/G3S support
  watchdog: rzg2l_wdt: Add suspend/resume support
  watchdog: rzg2l_wdt: Rely on the reset driver for doing proper reset
  watchdog: rzg2l_wdt: Remove comparison with zero
  watchdog: rzg2l_wdt: Remove reset de-assert from probe
  watchdog: rzg2l_wdt: Check return status of pm_runtime_put()
  watchdog: rzg2l_wdt: Use pm_runtime_resume_and_get()
  watchdog: rzg2l_wdt: Make the driver depend on PM
  watchdog: rzg2l_wdt: Restrict the driver to ARCH_RZG2L and ARCH_R9A09G011
  watchdog: imx7ulp_wdt: keep already running watchdog enabled
  watchdog: starfive: Add missing clk_disable_unprepare()
  watchdog: Make watchdog_class const

7 months agoMerge tag 'dma-mapping-6.11-2024-07-24' of git://git.infradead.org/users/hch/dma...
Linus Torvalds [Thu, 25 Jul 2024 17:10:34 +0000 (10:10 -0700)]
Merge tag 'dma-mapping-6.11-2024-07-24' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fix from Christoph Hellwig:

 - fix the order of actions in dmam_free_coherent (Lance Richardson)

* tag 'dma-mapping-6.11-2024-07-24' of git://git.infradead.org/users/hch/dma-mapping:
  dma: fix call order in dmam_free_coherent

7 months agoMerge tag 'asoc-fix-v6.11-merge-window' of https://git.kernel.org/pub/scm/linux/kerne...
Takashi Iwai [Thu, 25 Jul 2024 16:04:55 +0000 (18:04 +0200)]
Merge tag 'asoc-fix-v6.11-merge-window' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound into for-linus

ASoC: Fixes for v6.11

A selection of routine fixes and quirks that came in since the merge
window.  The fsl-asoc-card change is a fix for systems with multiple
cards where updating templates in place leaks data from one card to
another.

7 months agoMerge branch 'tap-tun-harden-by-dropping-short-frame'
Jakub Kicinski [Thu, 25 Jul 2024 15:07:06 +0000 (08:07 -0700)]
Merge branch 'tap-tun-harden-by-dropping-short-frame'

Dongli Zhang says:

====================
tap/tun: harden by dropping short frame

This is to harden all of tap/tun to avoid any short frame smaller than the
Ethernet header (ETH_HLEN).

While the xen-netback already rejects short frame smaller than ETH_HLEN ...

 914 static void xenvif_tx_build_gops(struct xenvif_queue *queue,
 915                                      int budget,
 916                                      unsigned *copy_ops,
 917                                      unsigned *map_ops)
 918 {
... ...
1007                 if (unlikely(txreq.size < ETH_HLEN)) {
1008                         netdev_dbg(queue->vif->dev,
1009                                    "Bad packet size: %d\n", txreq.size);
1010                         xenvif_tx_err(queue, &txreq, extra_count, idx);
1011                         break;
1012                 }

... the short frame may not be dropped by vhost-net/tap/tun.

This fixes CVE-2024-41090 and CVE-2024-41091.
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
7 months agotun: add missing verification for short frame
Dongli Zhang [Wed, 24 Jul 2024 17:04:52 +0000 (10:04 -0700)]
tun: add missing verification for short frame

The cited commit missed to check against the validity of the frame length
in the tun_xdp_one() path, which could cause a corrupted skb to be sent
downstack. Even before the skb is transmitted, the
tun_xdp_one-->eth_type_trans() may access the Ethernet header although it
can be less than ETH_HLEN. Once transmitted, this could either cause
out-of-bound access beyond the actual length, or confuse the underlayer
with incorrect or inconsistent header length in the skb metadata.

In the alternative path, tun_get_user() already prohibits short frame which
has the length less than Ethernet header size from being transmitted for
IFF_TAP.

This is to drop any frame shorter than the Ethernet header size just like
how tun_get_user() does.

CVE: CVE-2024-41091
Inspired-by: https://lore.kernel.org/netdev/[email protected]/
Fixes: 043d222f93ab ("tuntap: accept an array of XDP buffs through sendmsg()")
Cc: [email protected]
Signed-off-by: Dongli Zhang <[email protected]>
Reviewed-by: Si-Wei Liu <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Reviewed-by: Paolo Abeni <[email protected]>
Reviewed-by: Jason Wang <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
7 months agotap: add missing verification for short frame
Si-Wei Liu [Wed, 24 Jul 2024 17:04:51 +0000 (10:04 -0700)]
tap: add missing verification for short frame

The cited commit missed to check against the validity of the frame length
in the tap_get_user_xdp() path, which could cause a corrupted skb to be
sent downstack. Even before the skb is transmitted, the
tap_get_user_xdp()-->skb_set_network_header() may assume the size is more
than ETH_HLEN. Once transmitted, this could either cause out-of-bound
access beyond the actual length, or confuse the underlayer with incorrect
or inconsistent header length in the skb metadata.

In the alternative path, tap_get_user() already prohibits short frame which
has the length less than Ethernet header size from being transmitted.

This is to drop any frame shorter than the Ethernet header size just like
how tap_get_user() does.

CVE: CVE-2024-41090
Link: https://lore.kernel.org/netdev/[email protected]/
Fixes: 0efac27791ee ("tap: accept an array of XDP buffs through sendmsg()")
Cc: [email protected]
Signed-off-by: Si-Wei Liu <[email protected]>
Signed-off-by: Dongli Zhang <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Reviewed-by: Paolo Abeni <[email protected]>
Reviewed-by: Jason Wang <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
7 months agomISDN: Fix a use after free in hfcmulti_tx()
Dan Carpenter [Wed, 24 Jul 2024 16:08:18 +0000 (11:08 -0500)]
mISDN: Fix a use after free in hfcmulti_tx()

Don't dereference *sp after calling dev_kfree_skb(*sp).

Fixes: af69fb3a8ffa ("Add mISDN HFC multiport driver")
Signed-off-by: Dan Carpenter <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
7 months agogve: Fix an edge case for TSO skb validity check
Bailey Forrest [Wed, 24 Jul 2024 14:34:31 +0000 (07:34 -0700)]
gve: Fix an edge case for TSO skb validity check

The NIC requires each TSO segment to not span more than 10
descriptors. NIC further requires each descriptor to not exceed
16KB - 1 (GVE_TX_MAX_BUF_SIZE_DQO).

The descriptors for an skb are generated by
gve_tx_add_skb_no_copy_dqo() for DQO RDA queue format.
gve_tx_add_skb_no_copy_dqo() loops through each skb frag and
generates a descriptor for the entire frag if the frag size is
not greater than GVE_TX_MAX_BUF_SIZE_DQO. If the frag size is
greater than GVE_TX_MAX_BUF_SIZE_DQO, it is split into descriptor(s)
of size GVE_TX_MAX_BUF_SIZE_DQO and a descriptor is generated for
the remainder (frag size % GVE_TX_MAX_BUF_SIZE_DQO).

gve_can_send_tso() checks if the descriptors thus generated for an
skb would meet the requirement that each TSO-segment not span more
than 10 descriptors. However, the current code misses an edge case
when a TSO segment spans multiple descriptors within a large frag.
This change fixes the edge case.

gve_can_send_tso() relies on the assumption that max gso size (9728)
is less than GVE_TX_MAX_BUF_SIZE_DQO and therefore within an skb
fragment a TSO segment can never span more than 2 descriptors.

Fixes: a57e5de476be ("gve: DQO: Add TX path")
Signed-off-by: Praveen Kaligineedi <[email protected]>
Signed-off-by: Bailey Forrest <[email protected]>
Reviewed-by: Jeroen de Borst <[email protected]>
Cc: [email protected]
Reviewed-by: Willem de Bruijn <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
7 months agobnxt_en: update xdp_rxq_info in queue restart logic
Taehee Yoo [Sun, 21 Jul 2024 05:35:54 +0000 (05:35 +0000)]
bnxt_en: update xdp_rxq_info in queue restart logic

When the netdev_rx_queue_restart() restarts queues, the bnxt_en driver
updates(creates and deletes) a page_pool.
But it doesn't update xdp_rxq_info, so the xdp_rxq_info is still
connected to an old page_pool.
So, bnxt_rx_ring_info->page_pool indicates a new page_pool, but
bnxt_rx_ring_info->xdp_rxq is still connected to an old page_pool.

An old page_pool is no longer used so it is supposed to be
deleted by page_pool_destroy() but it isn't.
Because the xdp_rxq_info is holding the reference count for it and the
xdp_rxq_info is not updated, an old page_pool will not be deleted in
the queue restart logic.

Before restarting 1 queue:
./tools/net/ynl/samples/page-pool
enp10s0f1np1[6] page pools: 4 (zombies: 0)
refs: 8192 bytes: 33554432 (refs: 0 bytes: 0)
recycling: 0.0% (alloc: 128:8048 recycle: 0:0)

After restarting 1 queue:
./tools/net/ynl/samples/page-pool
enp10s0f1np1[6] page pools: 5 (zombies: 0)
refs: 10240 bytes: 41943040 (refs: 0 bytes: 0)
recycling: 20.0% (alloc: 160:10080 recycle: 1920:128)

Before restarting queues, an interface has 4 page_pools.
After restarting one queue, an interface has 5 page_pools, but it
should be 4, not 5.
The reason is that queue restarting logic creates a new page_pool and
an old page_pool is not deleted due to the absence of an update of
xdp_rxq_info logic.

Fixes: 2d694c27d32e ("bnxt_en: implement netdev_queue_mgmt_ops")
Signed-off-by: Taehee Yoo <[email protected]>
Reviewed-by: David Wei <[email protected]>
Reviewed-by: Somnath Kotur <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
7 months agoMerge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Jakub Kicinski [Thu, 25 Jul 2024 14:40:24 +0000 (07:40 -0700)]
Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2024-07-25

We've added 14 non-merge commits during the last 8 day(s) which contain
a total of 19 files changed, 177 insertions(+), 70 deletions(-).

The main changes are:

1) Fix af_unix to disable MSG_OOB handling for sockets in BPF sockmap and
   BPF sockhash. Also add test coverage for this case, from Michal Luczaj.

2) Fix a segmentation issue when downgrading gso_size in the BPF helper
   bpf_skb_adjust_room(), from Fred Li.

3) Fix a compiler warning in resolve_btfids due to a missing type cast,
   from Liwei Song.

4) Fix stack allocation for arm64 to align the stack pointer at a 16 byte
   boundary in the fexit_sleep BPF selftest, from Puranjay Mohan.

5) Fix a xsk regression to require a flag when actuating tx_metadata_len,
   from Stanislav Fomichev.

6) Fix function prototype BTF dumping in libbpf for prototypes that have
   no input arguments, from Andrii Nakryiko.

7) Fix stacktrace symbol resolution in perf script for BPF programs
   containing subprograms, from Hou Tao.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
  selftests/bpf: Add XDP_UMEM_TX_METADATA_LEN to XSK TX metadata test
  xsk: Require XDP_UMEM_TX_METADATA_LEN to actuate tx_metadata_len
  bpf: Fix a segment issue when downgrading gso_size
  tools/resolve_btfids: Fix comparison of distinct pointer types warning in resolve_btfids
  bpf, events: Use prog to emit ksymbol event for main program
  selftests/bpf: Test sockmap redirect for AF_UNIX MSG_OOB
  selftests/bpf: Parametrize AF_UNIX redir functions to accept send() flags
  selftests/bpf: Support SOCK_STREAM in unix_inet_redir_to_connected()
  af_unix: Disable MSG_OOB handling for sockets in sockmap/sockhash
  bpftool: Fix typo in usage help
  libbpf: Fix no-args func prototype BTF dumping syntax
  MAINTAINERS: Update powerpc BPF JIT maintainers
  MAINTAINERS: Update email address of Naveen
  selftests/bpf: fexit_sleep: Fix stack allocation for arm64
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
7 months agoASoC: fsl-asoc-card: Dynamically allocate memory for snd_soc_dai_link_components
Shengjiu Wang [Thu, 25 Jul 2024 03:22:53 +0000 (11:22 +0800)]
ASoC: fsl-asoc-card: Dynamically allocate memory for snd_soc_dai_link_components

The static snd_soc_dai_link_components cause conflict for multiple
instances of this generic driver. For example, when there is
wm8962 and SPDIF case enabled together, the contaminated
snd_soc_dai_link_components will cause another device probe fail.

Fixes: 6d174cc4f224 ("ASoC: fsl-asoc-card: merge spdif support from imx-spdif.c")
Signed-off-by: Shengjiu Wang <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Mark Brown <[email protected]>
7 months agoarm64: mm: Fix lockless walks with static and dynamic page-table folding
Will Deacon [Thu, 25 Jul 2024 09:03:45 +0000 (10:03 +0100)]
arm64: mm: Fix lockless walks with static and dynamic page-table folding

Lina reports random oopsen originating from the fast GUP code when
16K pages are used with 4-level page-tables, the fourth level being
folded at runtime due to lack of LPA2.

In this configuration, the generic implementation of
p4d_offset_lockless() will return a 'p4d_t *' corresponding to the
'pgd_t' allocated on the stack of the caller, gup_fast_pgd_range().
This is normally fine, but when the fourth level of page-table is folded
at runtime, pud_offset_lockless() will offset from the address of the
'p4d_t' to calculate the address of the PUD in the same page-table page.
This results in a stray stack read when the 'p4d_t' has been allocated
on the stack and can send the walker into the weeds.

Fix the problem by providing our own definition of p4d_offset_lockless()
when CONFIG_PGTABLE_LEVELS <= 4 which returns the real page-table
pointer rather than the address of the local stack variable.

Cc: Catalin Marinas <[email protected]>
Cc: Ard Biesheuvel <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Fixes: 0dd4f60a2c76 ("arm64: mm: Add support for folding PUDs at runtime")
Reported-by: Asahi Lina <[email protected]>
Reviewed-by: Ard Biesheuvel <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Will Deacon <[email protected]>
7 months agoASoC: amd: yc: Support mic on Lenovo Thinkpad E16 Gen 2
Takashi Iwai [Thu, 25 Jul 2024 06:54:28 +0000 (08:54 +0200)]
ASoC: amd: yc: Support mic on Lenovo Thinkpad E16 Gen 2

Lenovo Thinkpad E16 Gen 2 AMD model (model 21M5) needs a corresponding
quirk entry for making the internal mic working.

Link: https://bugzilla.suse.com/show_bug.cgi?id=1228269
Cc: [email protected]
Signed-off-by: Takashi Iwai <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Mark Brown <[email protected]>
7 months agox86/xen: fix memblock_reserve() usage on PVH
Roger Pau Monne [Thu, 25 Jul 2024 07:31:15 +0000 (09:31 +0200)]
x86/xen: fix memblock_reserve() usage on PVH

The current usage of memblock_reserve() in init_pvh_bootparams() is done before
the .bss is zeroed, and that used to be fine when
memblock_reserved_init_regions implicitly ended up in the .meminit.data
section.  However after commit 73db3abdca58c memblock_reserved_init_regions
ends up in the .bss section, thus breaking it's usage before the .bss is
cleared.

Move and rename the call to xen_reserve_extra_memory() so it's done in the
x86_init.oem.arch_setup hook, which gets executed after the .bss has been
zeroed, but before calling e820__memory_setup().

Fixes: 73db3abdca58c ("init/modpost: conditionally check section mismatch to __meminit*")
Signed-off-by: Roger Pau Monné <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Message-ID: <20240725073116[email protected]>
Signed-off-by: Juergen Gross <[email protected]>
7 months agox86/xen: move xen_reserve_extra_memory()
Roger Pau Monne [Thu, 25 Jul 2024 07:31:14 +0000 (09:31 +0200)]
x86/xen: move xen_reserve_extra_memory()

In preparation for making the function static.

No functional change.

Signed-off-by: Roger Pau Monné <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Message-ID: <20240725073116[email protected]>
Signed-off-by: Juergen Gross <[email protected]>
7 months agotcp: process the 3rd ACK with sk_socket for TFO/MPTCP
Matthieu Baerts (NGI0) [Wed, 24 Jul 2024 10:25:16 +0000 (12:25 +0200)]
tcp: process the 3rd ACK with sk_socket for TFO/MPTCP

The 'Fixes' commit recently changed the behaviour of TCP by skipping the
processing of the 3rd ACK when a sk->sk_socket is set. The goal was to
skip tcp_ack_snd_check() in tcp_rcv_state_process() not to send an
unnecessary ACK in case of simultaneous connect(). Unfortunately, that
had an impact on TFO and MPTCP.

I started to look at the impact on MPTCP, because the MPTCP CI found
some issues with the MPTCP Packetdrill tests [1]. Then Paolo Abeni
suggested me to look at the impact on TFO with "plain" TCP.

For MPTCP, when receiving the 3rd ACK of a request adding a new path
(MP_JOIN), sk->sk_socket will be set, and point to the MPTCP sock that
has been created when the MPTCP connection got established before with
the first path. The newly added 'goto' will then skip the processing of
the segment text (step 7) and not go through tcp_data_queue() where the
MPTCP options are validated, and some actions are triggered, e.g.
sending the MPJ 4th ACK [2] as demonstrated by the new errors when
running a packetdrill test [3] establishing a second subflow.

This doesn't fully break MPTCP, mainly the 4th MPJ ACK that will be
delayed. Still, we don't want to have this behaviour as it delays the
switch to the fully established mode, and invalid MPTCP options in this
3rd ACK will not be caught any more. This modification also affects the
MPTCP + TFO feature as well, and being the reason why the selftests
started to be unstable the last few days [4].

For TFO, the existing 'basic-cookie-not-reqd' test [5] was no longer
passing: if the 3rd ACK contains data, and the connection is accept()ed
before receiving them, these data would no longer be processed, and thus
not ACKed.

One last thing about MPTCP, in case of simultaneous connect(), a
fallback to TCP will be done, which seems fine:

  `../common/defaults.sh`

   0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_MPTCP) = 3
  +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)

  +0 > S  0:0(0)                 <mss 1460, sackOK, TS val 100 ecr 0,   nop, wscale 8, mpcapable v1 flags[flag_h] nokey>
  +0 < S  0:0(0) win 1000        <mss 1460, sackOK, TS val 407 ecr 0,   nop, wscale 8, mpcapable v1 flags[flag_h] nokey>
  +0 > S. 0:0(0) ack 1           <mss 1460, sackOK, TS val 330 ecr 0,   nop, wscale 8, mpcapable v1 flags[flag_h] nokey>
  +0 < S. 0:0(0) ack 1 win 65535 <mss 1460, sackOK, TS val 700 ecr 100, nop, wscale 8, mpcapable v1 flags[flag_h] key[skey=2]>
  +0 >  . 1:1(0) ack 1           <nop, nop, TS val 845707014 ecr 700, nop, nop, sack 0:1>

Simultaneous SYN-data crossing is also not supported by TFO, see [6].

Kuniyuki Iwashima suggested to restrict the processing to SYN+ACK only:
that's a more generic solution than the one initially proposed, and
also enough to fix the issues described above.

Later on, Eric Dumazet mentioned that an ACK should still be sent in
reaction to the second SYN+ACK that is received: not sending a DUPACK
here seems wrong and could hurt:

   0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
  +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)

  +0 > S  0:0(0)                <mss 1460, sackOK, TS val 1000 ecr 0,nop,wscale 8>
  +0 < S  0:0(0)       win 1000 <mss 1000, sackOK, nop, nop>
  +0 > S. 0:0(0) ack 1          <mss 1460, sackOK, TS val 3308134035 ecr 0,nop,wscale 8>
  +0 < S. 0:0(0) ack 1 win 1000 <mss 1000, sackOK, nop, nop>
  +0 >  . 1:1(0) ack 1          <nop, nop, sack 0:1>  // <== Here

So in this version, the 'goto consume' is dropped, to always send an ACK
when switching from TCP_SYN_RECV to TCP_ESTABLISHED. This ACK will be
seen as a DUPACK -- with DSACK if SACK has been negotiated -- in case of
simultaneous SYN crossing: that's what is expected here.

Link: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/9936227696
Link: https://datatracker.ietf.org/doc/html/rfc8684#fig_tokens
Link: https://github.com/multipath-tcp/packetdrill/blob/mptcp-net-next/gtests/net/mptcp/syscalls/accept.pkt#L28
Link: https://netdev.bots.linux.dev/contest.html?executor=vmksft-mptcp-dbg&test=mptcp-connect-sh
Link: https://github.com/google/packetdrill/blob/master/gtests/net/tcp/fastopen/server/basic-cookie-not-reqd.pkt#L21
Link: https://github.com/google/packetdrill/blob/master/gtests/net/tcp/fastopen/client/simultaneous-fast-open.pkt
Fixes: 23e89e8ee7be ("tcp: Don't drop SYN+ACK for simultaneous connect().")
Suggested-by: Paolo Abeni <[email protected]>
Suggested-by: Kuniyuki Iwashima <[email protected]>
Suggested-by: Eric Dumazet <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Link: https://patch.msgid.link/20240724-upstream-net-next-20240716-tcp-3rd-ack-consume-sk_socket-v3-1-d48339764ce9@kernel.org
Signed-off-by: Paolo Abeni <[email protected]>
7 months agorbd: don't assume rbd_is_lock_owner() for exclusive mappings
Ilya Dryomov [Tue, 23 Jul 2024 16:08:08 +0000 (18:08 +0200)]
rbd: don't assume rbd_is_lock_owner() for exclusive mappings

Expanding on the previous commit, assuming that rbd_is_lock_owner()
always returns true (i.e. that we are either in RBD_LOCK_STATE_LOCKED
or RBD_LOCK_STATE_QUIESCING) if the mapping is exclusive is wrong too.
In case ceph_cls_set_cookie() fails, the lock would be temporarily
released even if the mapping is exclusive, meaning that we can end up
even in RBD_LOCK_STATE_UNLOCKED.

IOW, exclusive mappings are really "just" about disabling automatic
lock transitions (as documented in the man page), not about grabbing
the lock and holding on to it whatever it takes.

Cc: [email protected]
Fixes: 637cd060537d ("rbd: new exclusive lock wait/wake code")
Signed-off-by: Ilya Dryomov <[email protected]>
Reviewed-by: Dongsheng Yang <[email protected]>
7 months agorbd: don't assume RBD_LOCK_STATE_LOCKED for exclusive mappings
Ilya Dryomov [Tue, 23 Jul 2024 16:07:59 +0000 (18:07 +0200)]
rbd: don't assume RBD_LOCK_STATE_LOCKED for exclusive mappings

Every time a watch is reestablished after getting lost, we need to
update the cookie which involves quiescing exclusive lock.  For this,
we transition from RBD_LOCK_STATE_LOCKED to RBD_LOCK_STATE_QUIESCING
roughly for the duration of rbd_reacquire_lock() call.  If the mapping
is exclusive and I/O happens to arrive in this time window, it's failed
with EROFS (later translated to EIO) based on the wrong assumption in
rbd_img_exclusive_lock() -- "lock got released?" check there stopped
making sense with commit a2b1da09793d ("rbd: lock should be quiesced on
reacquire").

To make it worse, any such I/O is added to the acquiring list before
EROFS is returned and this sets up for violating rbd_lock_del_request()
precondition that the request is either on the running list or not on
any list at all -- see commit ded080c86b3f ("rbd: don't move requests
to the running list on errors").  rbd_lock_del_request() ends up
processing these requests as if they were on the running list which
screws up quiescing_wait completion counter and ultimately leads to

    rbd_assert(!completion_done(&rbd_dev->quiescing_wait));

being triggered on the next watch error.

Cc: [email protected] # 06ef84c4e9c4: rbd: rename RBD_LOCK_STATE_RELEASING and releasing_wait
Cc: [email protected]
Fixes: 637cd060537d ("rbd: new exclusive lock wait/wake code")
Signed-off-by: Ilya Dryomov <[email protected]>
Reviewed-by: Dongsheng Yang <[email protected]>
7 months agorbd: rename RBD_LOCK_STATE_RELEASING and releasing_wait
Ilya Dryomov [Tue, 23 Jul 2024 15:54:39 +0000 (17:54 +0200)]
rbd: rename RBD_LOCK_STATE_RELEASING and releasing_wait

... to RBD_LOCK_STATE_QUIESCING and quiescing_wait to recognize that
this state and the associated completion are backing rbd_quiesce_lock(),
which isn't specific to releasing the lock.

While exclusive lock does get quiesced before it's released, it also
gets quiesced before an attempt to update the cookie is made and there
the lock is not released as long as ceph_cls_set_cookie() succeeds.

Signed-off-by: Ilya Dryomov <[email protected]>
Reviewed-by: Dongsheng Yang <[email protected]>
7 months agoselftests/bpf: Add XDP_UMEM_TX_METADATA_LEN to XSK TX metadata test
Stanislav Fomichev [Sat, 13 Jul 2024 01:52:52 +0000 (18:52 -0700)]
selftests/bpf: Add XDP_UMEM_TX_METADATA_LEN to XSK TX metadata test

This flag is now required to use tx_metadata_len.

Fixes: 40808a237d9c ("selftests/bpf: Add TX side to xdp_metadata")
Reported-by: Julian Schindel <[email protected]>
Signed-off-by: Stanislav Fomichev <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Reviewed-by: Maciej Fijalkowski <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
7 months agoxsk: Require XDP_UMEM_TX_METADATA_LEN to actuate tx_metadata_len
Stanislav Fomichev [Sat, 13 Jul 2024 01:52:51 +0000 (18:52 -0700)]
xsk: Require XDP_UMEM_TX_METADATA_LEN to actuate tx_metadata_len

Julian reports that commit 341ac980eab9 ("xsk: Support tx_metadata_len")
can break existing use cases which don't zero-initialize xdp_umem_reg
padding. Introduce new XDP_UMEM_TX_METADATA_LEN to make sure we
interpret the padding as tx_metadata_len only when being explicitly
asked.

Fixes: 341ac980eab9 ("xsk: Support tx_metadata_len")
Reported-by: Julian Schindel <[email protected]>
Signed-off-by: Stanislav Fomichev <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Reviewed-by: Maciej Fijalkowski <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
7 months agobpf: Fix a segment issue when downgrading gso_size
Fred Li [Fri, 19 Jul 2024 02:46:53 +0000 (10:46 +0800)]
bpf: Fix a segment issue when downgrading gso_size

Linearize the skb when downgrading gso_size because it may trigger a
BUG_ON() later when the skb is segmented as described in [1,2].

Fixes: 2be7e212d5419 ("bpf: add bpf_skb_adjust_room helper")
Signed-off-by: Fred Li <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Acked-by: Daniel Borkmann <[email protected]>
Link: https://lore.kernel.org/all/[email protected]
Link: https://lore.kernel.org/all/[email protected]
Link: https://lore.kernel.org/bpf/[email protected]
7 months agonet: mediatek: Fix potential NULL pointer dereference in dummy net_device handling
Breno Leitao [Wed, 24 Jul 2024 08:05:23 +0000 (01:05 -0700)]
net: mediatek: Fix potential NULL pointer dereference in dummy net_device handling

Move the freeing of the dummy net_device from mtk_free_dev() to
mtk_remove().

Previously, if alloc_netdev_dummy() failed in mtk_probe(),
eth->dummy_dev would be NULL. The error path would then call
mtk_free_dev(), which in turn called free_netdev() assuming dummy_dev
was allocated (but it was not), potentially causing a NULL pointer
dereference.

By moving free_netdev() to mtk_remove(), we ensure it's only called when
mtk_probe() has succeeded and dummy_dev is fully allocated. This
addresses a potential NULL pointer dereference detected by Smatch[1].

Fixes: b209bd6d0bff ("net: mediatek: mtk_eth_sock: allocate dummy net_device dynamically")
Reported-by: Dan Carpenter <[email protected]>
Closes: https://lore.kernel.org/all/[email protected]/ [1]
Suggested-by: Dan Carpenter <[email protected]>
Reviewed-by: Dan Carpenter <[email protected]>
Signed-off-by: Breno Leitao <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>
7 months agoMerge tag 'nf-24-07-24' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Paolo Abeni [Thu, 25 Jul 2024 09:17:21 +0000 (11:17 +0200)]
Merge tag 'nf-24-07-24' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

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

The following patchset contains a Netfilter fix for net:

Patch #1 if FPU is busy, then pipapo set backend falls back to standard
         set element lookup. Moreover, disable bh while at this.
 From Florian Westphal.

netfilter pull request 24-07-24

* tag 'nf-24-07-24' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nft_set_pipapo_avx2: disable softinterrupts
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>
7 months agoALSA: hda/realtek: Implement sound init sequence for Samsung Galaxy Book3 Pro 360
Nick Weihs [Thu, 25 Jul 2024 05:47:22 +0000 (22:47 -0700)]
ALSA: hda/realtek: Implement sound init sequence for Samsung Galaxy Book3 Pro 360

Samsung Galaxy Book3 Pro 360 sends a large amount of data to the codec
through hda processing coefficients.  This data was captured using a
modified version of QEMU, but the actual content of the data remains
opaque to me.  Elliding any part of the data seems to cause sound to
not work.

Signed-off-by: Nick Weihs <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>
7 months agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net...
Paolo Abeni [Thu, 25 Jul 2024 08:10:05 +0000 (10:10 +0200)]
Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
This series contains updates to ice driver only.

Ahmed enforces the iavf per VF filter limit on ice (PF) driver to prevent
possible resource exhaustion.

Wojciech corrects assignment of l2 flags read from firmware.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  ice: Fix recipe read procedure
  ice: Add a per-VF limit on number of FDIR filters
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>
7 months agodrm/amdgpu: reset vm state machine after gpu reset(vram lost)
ZhenGuo Yin [Fri, 19 Jul 2024 08:10:40 +0000 (16:10 +0800)]
drm/amdgpu: reset vm state machine after gpu reset(vram lost)

[Why]
Page table of compute VM in the VRAM will lost after gpu reset.
VRAM won't be restored since compute VM has no shadows.

[How]
Use higher 32-bit of vm->generation to record a vram_lost_counter.
Reset the VM state machine when vm->genertaion is not equal to
the new generation token.

v2: Check vm->generation instead of calling drm_sched_entity_error
in amdgpu_vm_validate.
v3: Use new generation token instead of vram_lost_counter for check.

Signed-off-by: ZhenGuo Yin <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
(cherry picked from commit 47c0388b0589cb481c294dcb857d25a214c46eb3)

7 months agodrm/amdgpu: add missed harvest check for VCN IP v4/v5
Tim Huang [Tue, 23 Jul 2024 08:54:34 +0000 (16:54 +0800)]
drm/amdgpu: add missed harvest check for VCN IP v4/v5

To prevent below probe failure, add a check for models with VCN
IP v4.0.6 where VCN1 may be harvested.

v2:
Apply the same check to VCN IP v4.0 and v5.0.

[   54.070117] RIP: 0010:vcn_v4_0_5_start_dpg_mode+0x9be/0x36b0 [amdgpu]
[   54.071055] Code: 80 fb ff 8d 82 00 80 fe ff 81 fe 00 06 00 00 0f 43
c2 49 69 d5 38 0d 00 00 48 8d 71 04 c1 e8 02 4c 01 f2 48 89 b2 50 f6 02
00 <89> 01 48 8b 82 50 f6 02 00 48 8d 48 04 48 89 8a 50 f6 02 00 c7 00
[   54.072408] RSP: 0018:ffffb17985f736f8 EFLAGS: 00010286
[   54.072793] RAX: 00000000000000d6 RBX: ffff99a82f680000 RCX:
0000000000000000
[   54.073315] RDX: ffff99a82f680000 RSI: 0000000000000004 RDI:
ffff99a82f680000
[   54.073835] RBP: ffffb17985f73730 R08: 0000000000000001 R09:
0000000000000000
[   54.074353] R10: 0000000000000008 R11: ffffb17983c05000 R12:
0000000000000000
[   54.074879] R13: 0000000000000000 R14: ffff99a82f680000 R15:
0000000000000001
[   54.075400] FS:  00007f8d9c79a000(0000) GS:ffff99ab2f140000(0000)
knlGS:0000000000000000
[   54.075988] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   54.076408] CR2: 0000000000000000 CR3: 0000000140c3a000 CR4:
0000000000750ef0
[   54.076927] PKRU: 55555554
[   54.077132] Call Trace:
[   54.077319]  <TASK>
[   54.077484]  ? show_regs+0x69/0x80
[   54.077747]  ? __die+0x28/0x70
[   54.077979]  ? page_fault_oops+0x180/0x4b0
[   54.078286]  ? do_user_addr_fault+0x2d2/0x680
[   54.078610]  ? exc_page_fault+0x84/0x190
[   54.078910]  ? asm_exc_page_fault+0x2b/0x30
[   54.079224]  ? vcn_v4_0_5_start_dpg_mode+0x9be/0x36b0 [amdgpu]
[   54.079941]  ? vcn_v4_0_5_start_dpg_mode+0xe6/0x36b0 [amdgpu]
[   54.080617]  vcn_v4_0_5_set_powergating_state+0x82/0x19b0 [amdgpu]
[   54.081316]  amdgpu_device_ip_set_powergating_state+0x64/0xc0
[amdgpu]
[   54.082057]  amdgpu_vcn_ring_begin_use+0x6f/0x1d0 [amdgpu]
[   54.082727]  amdgpu_ring_alloc+0x44/0x70 [amdgpu]
[   54.083351]  amdgpu_vcn_dec_sw_ring_test_ring+0x40/0x110 [amdgpu]
[   54.084054]  amdgpu_ring_test_helper+0x22/0x90 [amdgpu]
[   54.084698]  vcn_v4_0_5_hw_init+0x87/0xc0 [amdgpu]
[   54.085307]  amdgpu_device_init+0x1f96/0x2780 [amdgpu]
[   54.085951]  amdgpu_driver_load_kms+0x1e/0xc0 [amdgpu]
[   54.086591]  amdgpu_pci_probe+0x19f/0x550 [amdgpu]
[   54.087215]  local_pci_probe+0x48/0xa0
[   54.087509]  pci_device_probe+0xc9/0x250
[   54.087812]  really_probe+0x1a4/0x3f0
[   54.088101]  __driver_probe_device+0x7d/0x170
[   54.088443]  driver_probe_device+0x24/0xa0
[   54.088765]  __driver_attach+0xdd/0x1d0
[   54.089068]  ? __pfx___driver_attach+0x10/0x10
[   54.089417]  bus_for_each_dev+0x8e/0xe0
[   54.089718]  driver_attach+0x22/0x30
[   54.090000]  bus_add_driver+0x120/0x220
[   54.090303]  driver_register+0x62/0x120
[   54.090606]  ? __pfx_amdgpu_init+0x10/0x10 [amdgpu]
[   54.091255]  __pci_register_driver+0x62/0x70
[   54.091593]  amdgpu_init+0x67/0xff0 [amdgpu]
[   54.092190]  do_one_initcall+0x5f/0x330
[   54.092495]  do_init_module+0x68/0x240
[   54.092794]  load_module+0x201c/0x2110
[   54.093093]  init_module_from_file+0x97/0xd0
[   54.093428]  ? init_module_from_file+0x97/0xd0
[   54.093777]  idempotent_init_module+0x11c/0x2a0
[   54.094134]  __x64_sys_finit_module+0x64/0xc0
[   54.094476]  do_syscall_64+0x58/0x120
[   54.094767]  entry_SYSCALL_64_after_hwframe+0x6e/0x76

Signed-off-by: Tim Huang <[email protected]>
Reviewed-by: Saleemkhan Jamadar <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
(cherry picked from commit 0b071245ddd98539d4f7493bdd188417fcf2d629)

7 months agodrm/amdgpu: Fix eeprom max record count
Stanley.Yang [Thu, 18 Jul 2024 02:58:04 +0000 (10:58 +0800)]
drm/amdgpu: Fix eeprom max record count

The eeprom table is empty before initializing,
set eeprom table version first before initializing.

Changed from V1:
Reuse amdgpu_ras_set_eeprom_table_version function

Signed-off-by: Stanley.Yang <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 015b8a2fdf39a4c288ff24e7b715b8d9198e56dc)

7 months agodrm/amdgpu: fix ras UE error injection failure issue
YiPeng Chai [Fri, 19 Jul 2024 12:43:04 +0000 (20:43 +0800)]
drm/amdgpu: fix ras UE error injection failure issue

The ras command shared memory is allocated from
VRAM and the response status of the command
buffer will not be zero due to gpu being in
fatal error state after ras UE error injection.

Signed-off-by: YiPeng Chai <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 8284951a6e79c6806c675e5f68a4cd425dd56bc4)

7 months agodrm/amd/display: Remove ASSERT if significance is zero in math_ceil2
Rodrigo Siqueira [Thu, 4 Jul 2024 17:54:34 +0000 (11:54 -0600)]
drm/amd/display: Remove ASSERT if significance is zero in math_ceil2

In the DML math_ceil2 function, there is one ASSERT if the significance
is equal to zero. However, significance might be equal to zero
sometimes, and this is not an issue for a ceil function, but the current
ASSERT will trigger warnings in those cases. This commit removes the
ASSERT if the significance is equal to zero to avoid unnecessary noise.

Cc: Mario Limonciello <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: [email protected]
Reviewed-by: Chaitanya Dhere <[email protected]>
Signed-off-by: Rodrigo Siqueira <[email protected]>
Signed-off-by: Aurabindo Pillai <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 332315885d3ccc6d8fe99700f3c2e4c24aa65ab7)

7 months agodrm/amd/display: Check for NULL pointer
Sung Joon Kim [Mon, 8 Jul 2024 23:29:49 +0000 (19:29 -0400)]
drm/amd/display: Check for NULL pointer

[why & how]
Need to make sure plane_state is initialized
before accessing its members.

Cc: Mario Limonciello <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: [email protected]
Reviewed-by: Xi (Alex) Liu <[email protected]>
Signed-off-by: Sung Joon Kim <[email protected]>
Signed-off-by: Aurabindo Pillai <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 295d91cbc700651782a60572f83c24861607b648)

7 months agodrm/amdgpu/vcn: Use offsets local to VCN/JPEG in VF
Jane Jian [Mon, 15 Jul 2024 10:48:31 +0000 (18:48 +0800)]
drm/amdgpu/vcn: Use offsets local to VCN/JPEG in VF

For VCN/JPEG 4.0.3, use only the local addressing scheme.

- Mask bit higher than AID0 range

v2
remain the case for mmhub use master XCC

Signed-off-by: Jane Jian <[email protected]>
Reviewed-by: Lijo Lazar <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit caaf576292f8ccef5cdc0ac16e77b87dbf6e17ab)

7 months agodrm/amdgpu: Add empty HDP flush function to VCN v4.0.3
Lijo Lazar [Mon, 11 Dec 2023 05:48:42 +0000 (11:18 +0530)]
drm/amdgpu: Add empty HDP flush function to VCN v4.0.3

VCN 4.0.3 does not HDP flush with RRMT enabled. Instead, mmsch
will do the HDP flush.

This change is necessary for VCN v4.0.3, no need for backward compatibility

Signed-off-by: Lijo Lazar <[email protected]>
Signed-off-by: Jane Jian <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 49cfaebe48e97500a68d5322a8194736b0a2c3cf)

7 months agodrm/amdgpu: Add empty HDP flush function to JPEG v4.0.3
Lijo Lazar [Mon, 11 Dec 2023 05:15:38 +0000 (10:45 +0530)]
drm/amdgpu: Add empty HDP flush function to JPEG v4.0.3

JPEG v4.0.3 doesn't support HDP flush when RRMT is enabled. Instead,
mmsch fw will do the flush.

This change is necessary for JPEG v4.0.3, no need for backward compatibility

Signed-off-by: Lijo Lazar <[email protected]>
Signed-off-by: Jane Jian <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 585e3fdb36f59c5cfed0ae06c852dc1df22b1d60)

7 months agodrm/amd/amdgpu: Fix uninitialized variable warnings
Ma Ke [Thu, 18 Jul 2024 14:17:35 +0000 (22:17 +0800)]
drm/amd/amdgpu: Fix uninitialized variable warnings

Return 0 to avoid returning an uninitialized variable r.

Cc: [email protected]
Fixes: 230dd6bb6117 ("drm/amd/amdgpu: implement mode2 reset on smu_v13_0_10")
Signed-off-by: Ma Ke <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 6472de66c0aa18d50a4b5ca85f8272e88a737676)

7 months agodrm/amdgpu: Fix atomics on GFX12
David Belanger [Mon, 10 Jun 2024 20:38:55 +0000 (16:38 -0400)]
drm/amdgpu: Fix atomics on GFX12

If PCIe supports atomics, configure register to prevent DF from
breaking atomics in separate load/store operations.

Signed-off-by: David Belanger <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
(cherry picked from commit 666f14cab21b17ccc1bdfe1e82458aa429b3b7e0)

7 months agodrm/amdgpu/sdma5.2: Update wptr registers as well as doorbell
Alex Deucher [Tue, 9 Jul 2024 21:54:11 +0000 (17:54 -0400)]
drm/amdgpu/sdma5.2: Update wptr registers as well as doorbell

We seem to have a case where SDMA will sometimes miss a doorbell
if GFX is entering the powergating state when the doorbell comes in.
To workaround this, we can update the wptr via MMIO, however,
this is only safe because we disallow gfxoff in begin_ring() for
SDMA 5.2 and then allow it again in end_ring().

Enable this workaround while we are root causing the issue with
the HW team.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/3440
Tested-by: Friedrich Vock <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
(cherry picked from commit f2ac52634963fc38e4935e11077b6f7854e5d700)

7 months agoMerge tag 'phy-for-6.11' of git://git.kernel.org/pub/scm/linux/kernel/git/phy/linux-phy
Linus Torvalds [Wed, 24 Jul 2024 20:11:28 +0000 (13:11 -0700)]
Merge tag 'phy-for-6.11' of git://git.kernel.org/pub/scm/linux/kernel/git/phy/linux-phy

Pull phy updates from Vinod Koul:
 "New Support
   - Samsung Exynos gs101 drd combo phy
   - Qualcomm SC8180x USB uniphy, IPQ9574 QMP PCIe phy
   - Airoha EN7581 PCIe phy
   - Freescale i.MX8Q HSIO SerDes phy
   - Starfive jh7110 dphy tx

  Updates:
   - Resume support for j721e-wiz driver
   - Updates to Exynos usbdrd driver
   - Support for optional power domains in g12a usb2-phy driver
   - Debugfs support and updates to zynqmp driver"

* tag 'phy-for-6.11' of git://git.kernel.org/pub/scm/linux/kernel/git/phy/linux-phy: (56 commits)
  phy: airoha: Add dtime and Rx AEQ IO registers
  dt-bindings: phy: airoha: Add dtime and Rx AEQ IO registers
  dt-bindings: phy: rockchip-emmc-phy: Convert to dtschema
  dt-bindings: phy: qcom,qmp-usb: fix spelling error
  phy: exynos5-usbdrd: support Exynos USBDRD 3.1 combo phy (HS & SS)
  phy: exynos5-usbdrd: convert Vbus supplies to regulator_bulk
  phy: exynos5-usbdrd: convert (phy) register access clock to clk_bulk
  phy: exynos5-usbdrd: convert core clocks to clk_bulk
  phy: exynos5-usbdrd: support isolating HS and SS ports independently
  dt-bindings: phy: samsung,usb3-drd-phy: add gs101 compatible
  phy: core: Fix documentation of of_phy_get
  phy: starfive: Correct the dphy configure process
  phy: zynqmp: Add debugfs support
  phy: zynqmp: Take the phy mutex in xlate
  phy: zynqmp: Only wait for PLL lock "primary" instances
  phy: zynqmp: Store instance instead of type
  phy: zynqmp: Enable reference clock correctly
  phy: cadence-torrent: Check return value on register read
  phy: Fix the cacography in phy-exynos5250-usb2.c
  phy: phy-rockchip-samsung-hdptx: Select CONFIG_MFD_SYSCON
  ...

7 months agoMerge tag 'soundwire-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul...
Linus Torvalds [Wed, 24 Jul 2024 20:04:43 +0000 (13:04 -0700)]
Merge tag 'soundwire-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/soundwire

Pull soundwire updates from Vinod Koul:

 - Simplification across subsystem using cleanup.h

 - Support for debugfs to read/write commands

 - Few Intel and Qualcomm driver updates

* tag 'soundwire-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/soundwire:
  soundwire: debugfs: simplify with cleanup.h
  soundwire: cadence: simplify with cleanup.h
  soundwire: intel_ace2x: simplify with cleanup.h
  soundwire: intel_ace2x: simplify return path in hw_params
  soundwire: intel: simplify with cleanup.h
  soundwire: intel: simplify return path in hw_params
  soundwire: amd_init: simplify with cleanup.h
  soundwire: amd: simplify with cleanup.h
  soundwire: amd: simplify return path in hw_params
  soundwire: intel_auxdevice: start the bus at default frequency
  soundwire: intel_auxdevice: add cs42l43 codec to wake_capable_list
  drivers:soundwire: qcom: cleanup port maask calculations
  soundwire: bus: simplify by using local slave->prop
  soundwire: generic_bandwidth_allocation: change port_bo parameter to pointer
  soundwire: Intel: clarify Copyright information
  soundwire: intel_ace2.x: add AC timing extensions for PantherLake
  soundwire: bus: add stream refcount
  soundwire: debugfs: add interface to read/write commands

7 months agoMerge tag 'dmaengine-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul...
Linus Torvalds [Wed, 24 Jul 2024 19:34:47 +0000 (12:34 -0700)]
Merge tag 'dmaengine-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine

Pull dmaengine updates from Vinod Koul:
 "New support:

   - New dmaengine_prep_peripheral_dma_vec() to support transfers using
     dma vectors and documentation and user in AXI dma

   - STMicro STM32 DMA3 support and new capabilities of cyclic dma

  Updates:

   - Yaml conversion for Freescale imx dma and qdma bindings,
     sprd sc9860 dma binding

   - Altera msgdma updates for descriptor management"

* tag 'dmaengine-6.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine: (35 commits)
  dt-bindings: fsl-qdma: fix interrupts 'if' check logic
  dt-bindings: dma: sprd,sc9860-dma: convert to YAML
  dmaengine: fsl-dpaa2-qdma: add missing MODULE_DESCRIPTION() macro
  dmaengine: ti: add missing MODULE_DESCRIPTION() macros
  dmaengine: ti: cppi41: add missing MODULE_DESCRIPTION() macro
  dmaengine: virt-dma: add missing MODULE_DESCRIPTION() macro
  dmaengine: ti: k3-udma: Fix BCHAN count with UHC and HC channels
  dmaengine: sh: rz-dmac: Fix lockdep assert warning
  dmaengine: qcom: gpi: clean up the IRQ disable/enable in gpi_reset_chan()
  dmaengine: fsl-edma: change the memory access from local into remote mode in i.MX 8QM
  dmaengine: qcom: gpi: remove unused struct 'reg_info'
  dmaengine: moxart-dma: remove unused struct 'moxart_filter_data'
  dt-bindings: fsl-qdma: Convert to yaml format
  dmaengine: fsl-edma: remove redundant "idle" field from fsl_chan
  dmaengine: fsl-edma: request per-channel IRQ only when channel is allocated
  dmaengine: stm32-dma3: defer channel registration to specify channel name
  dmaengine: add channel device name to channel registration
  dmaengine: stm32-dma3: improve residue granularity
  dmaengine: stm32-dma3: add device_pause and device_resume ops
  dmaengine: stm32-dma3: add DMA_MEMCPY capability
  ...

7 months agosysctl: treewide: constify the ctl_table argument of proc_handlers
Joel Granados [Wed, 24 Jul 2024 18:59:29 +0000 (20:59 +0200)]
sysctl: treewide: constify the ctl_table argument of proc_handlers

const qualify the struct ctl_table argument in the proc_handler function
signatures. This is a prerequisite to moving the static ctl_table
structs into .rodata data which will ensure that proc_handler function
pointers cannot be modified.

This patch has been generated by the following coccinelle script:

```
  virtual patch

  @r1@
  identifier ctl, write, buffer, lenp, ppos;
  identifier func !~ "appldata_(timer|interval)_handler|sched_(rt|rr)_handler|rds_tcp_skbuf_handler|proc_sctp_do_(hmac_alg|rto_min|rto_max|udp_port|alpha_beta|auth|probe_interval)";
  @@

  int func(
  - struct ctl_table *ctl
  + const struct ctl_table *ctl
    ,int write, void *buffer, size_t *lenp, loff_t *ppos);

  @r2@
  identifier func, ctl, write, buffer, lenp, ppos;
  @@

  int func(
  - struct ctl_table *ctl
  + const struct ctl_table *ctl
    ,int write, void *buffer, size_t *lenp, loff_t *ppos)
  { ... }

  @r3@
  identifier func;
  @@

  int func(
  - struct ctl_table *
  + const struct ctl_table *
    ,int , void *, size_t *, loff_t *);

  @r4@
  identifier func, ctl;
  @@

  int func(
  - struct ctl_table *ctl
  + const struct ctl_table *ctl
    ,int , void *, size_t *, loff_t *);

  @r5@
  identifier func, write, buffer, lenp, ppos;
  @@

  int func(
  - struct ctl_table *
  + const struct ctl_table *
    ,int write, void *buffer, size_t *lenp, loff_t *ppos);

```

* Code formatting was adjusted in xfs_sysctl.c to comply with code
  conventions. The xfs_stats_clear_proc_handler,
  xfs_panic_mask_proc_handler and xfs_deprecated_dointvec_minmax where
  adjusted.

* The ctl_table argument in proc_watchdog_common was const qualified.
  This is called from a proc_handler itself and is calling back into
  another proc_handler, making it necessary to change it as part of the
  proc_handler migration.

Co-developed-by: Thomas Weißschuh <[email protected]>
Signed-off-by: Thomas Weißschuh <[email protected]>
Co-developed-by: Joel Granados <[email protected]>
Signed-off-by: Joel Granados <[email protected]>
7 months agoMerge tag 'random-6.11-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 24 Jul 2024 17:29:50 +0000 (10:29 -0700)]
Merge tag 'random-6.11-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random

Pull random number generator updates from Jason Donenfeld:
 "This adds getrandom() support to the vDSO.

  First, it adds a new kind of mapping to mmap(2), MAP_DROPPABLE, which
  lets the kernel zero out pages anytime under memory pressure, which
  enables allocating memory that never gets swapped to disk but also
  doesn't count as being mlocked.

  Then, the vDSO implementation of getrandom() is introduced in a
  generic manner and hooked into random.c.

  Next, this is implemented on x86. (Also, though it's not ready for
  this pull, somebody has begun an arm64 implementation already)

  Finally, two vDSO selftests are added.

  There are also two housekeeping cleanup commits"

* tag 'random-6.11-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random:
  MAINTAINERS: add random.h headers to RNG subsection
  random: note that RNDGETPOOL was removed in 2.6.9-rc2
  selftests/vDSO: add tests for vgetrandom
  x86: vdso: Wire up getrandom() vDSO implementation
  random: introduce generic vDSO getrandom() implementation
  mm: add MAP_DROPPABLE for designating always lazily freeable mappings

7 months agoMerge tag 'vfs-6.11-rc1.fixes.2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 24 Jul 2024 16:42:51 +0000 (09:42 -0700)]
Merge tag 'vfs-6.11-rc1.fixes.2' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

Pull vfs fixes from Christian Brauner:
 "VFS:

   - The new 64bit mount ids start after the old mount id, i.e., at the
     first non-32 bit value. However, we started counting one id too
     late and thus lost 4294967296 as the first valid id. Fix that.

   - Update a few comments on some vfs_*() creation helpers.

   - Move copying of the xattr name out from the locks required to start
     a filesystem write.

   - Extend the filelock lock UAF fix to the compat code as well.

   - Now that we added the ability to look up an inode under RCU it's
     possible that lockless hash lookup can find and lock an inode after
     it gets I_FREEING set. It then waits until inode teardown in
     evict() is finished.

     The flag however is still set after evict() has woken up all
     waiters. If the inode lock is taken late enough on the waiting side
     after hash removal and wakeup happened the waiting thread will
     never be woken.

     Before RCU based lookup this was synchronized via the
     inode_hash_lock. But since unhashing requires the inode lock as
     well we can check whether the inode is unhashed while holding inode
     lock even without holding inode_hash_lock.

  pidfd:

   - The nsproxy structure contains nearly all of the namespaces
     associated with a task. When a namespace type isn't supported
     nsproxy might contain a NULL pointer or always point to the initial
     namespace type. The logic isn't consistent. So when deriving
     namespace fds we need to ensure that the namespace type is
     supported.

     First, so that we don't risk dereferncing NULL pointers. The
     correct bigger fix would be to change all namespaces to always set
     a valid namespace pointer in struct nsproxy independent of whether
     or not it is compiled in. But that requires quite a few changes.

     Second, so that we don't allow deriving namespace fds when the
     namespace type doesn't exist and thus when they couldn't also be
     derived via /proc/self/ns/.

   - Add missing selftests for the new pidfd ioctls to derive namespace
     fds. This simply extends the already existing testsuite.

  netfs:

   - Fix debug logging and fix kconfig variable name so it actually
     works.

   - Fix writeback that goes both to the server and cache. The streams
     are only activated once a subreq is added. When a server write
     happens the subreq doesn't need to have finished by the time the
     cache write is started. If the server write has already finished by
     the time the cache write is about to start the cache write will
     operate on a folio that might already have been reused. Fix this by
     preactivating the cache write.

   - Limit cachefiles subreq size for cache writes to MAX_RW_COUNT"

* tag 'vfs-6.11-rc1.fixes.2' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
  inode: clarify what's locked
  vfs: Fix potential circular locking through setxattr() and removexattr()
  filelock: Fix fcntl/close race recovery compat path
  fs: use all available ids
  cachefiles: Set the max subreq size for cache writes to MAX_RW_COUNT
  netfs: Fix writeback that needs to go to both server and cache
  pidfs: add selftests for new namespace ioctls
  pidfs: handle kernels without namespaces cleanly
  pidfs: when time ns disabled add check for ioctl
  vfs: correct the comments of vfs_*() helpers
  vfs: handle __wait_on_freeing_inode() and evict() race
  netfs: Rename CONFIG_FSCACHE_DEBUG to CONFIG_NETFS_DEBUG
  netfs: Revert "netfs: Switch debug logging to pr_debug()"

7 months agohostfs: fix folio conversion
Linus Torvalds [Wed, 24 Jul 2024 01:14:54 +0000 (18:14 -0700)]
hostfs: fix folio conversion

Commit e3ec0fe944d2 ("hostfs: Convert hostfs_read_folio() to use a
folio") simplified hostfs_read_folio(), but in the process of converting
to using folios natively also mis-used the folio_zero_tail() function
due to the very confusing API of that function.

Very arguably it's folio_zero_tail() API itself that is buggy, since it
would make more sense (and the documentation kind of implies) that the
third argument would be the pointer to the beginning of the folio
buffer.

But no, the third argument to folio_zero_tail() is where we should start
zeroing the tail (even if we already also pass in the offset separately
as the second argument).

So fix the hostfs caller, and we can leave any folio_zero_tail() sanity
cleanup for later.

Reported-and-tested-by: Maciej Żenczykowski <[email protected]>
Fixes: e3ec0fe944d2 ("hostfs: Convert hostfs_read_folio() to use a folio")
Link: https://lore.kernel.org/all/CANP3RGceNzwdb7w=vPf5=7BCid5HVQDmz1K5kC9JG42+HVAh_g@mail.gmail.com/
Cc: Matthew Wilcox <[email protected]>
Cc: Christian Brauner <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
7 months agoALSA: hda/realtek: cs35l41: Fixup remaining asus strix models
Luke D. Jones [Tue, 23 Jul 2024 01:12:24 +0000 (13:12 +1200)]
ALSA: hda/realtek: cs35l41: Fixup remaining asus strix models

Adjust quirks for 0x3a20, 0x3a30, 0x3a50 to match the 0x3a60. This
set has now been confirmed to work with this patch.

Signed-off-by: Luke D. Jones <[email protected]>
Fixes: 811dd426a9b1 ("ALSA: hda/realtek: Add quirks for Asus ROG 2024 laptops using CS35L41")
Cc: <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>
7 months agoRISC-V: Select ACPI PPTT drivers
Yunhui Cui [Mon, 17 Jun 2024 13:14:25 +0000 (21:14 +0800)]
RISC-V: Select ACPI PPTT drivers

After adding ACPI support to populate_cache_leaves(), RISC-V can build
cacheinfo through the ACPI PPTT table, thus enabling the ACPI_PPTT
configuration.

Signed-off-by: Yunhui Cui <[email protected]>
Reviewed-by: Jeremy Linton <[email protected]>
Reviewed-by: Sudeep Holla <[email protected]>
Reviewed-by: Sunil V L <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoriscv: cacheinfo: initialize cacheinfo's level and type from ACPI PPTT
Yunhui Cui [Mon, 17 Jun 2024 13:14:24 +0000 (21:14 +0800)]
riscv: cacheinfo: initialize cacheinfo's level and type from ACPI PPTT

Before cacheinfo can be built correctly, we need to initialize level
and type. Since RISC-V currently does not have a register group that
describes cache-related attributes like ARM64, we cannot obtain them
directly, so now we obtain cache leaves from the ACPI PPTT table
(acpi_get_cache_info()) and set the cache type through split_levels.

Suggested-by: Jeremy Linton <[email protected]>
Suggested-by: Sudeep Holla <[email protected]>
Reviewed-by: Conor Dooley <[email protected]>
Reviewed-by: Sunil V L <[email protected]>
Reviewed-by: Jeremy Linton <[email protected]>
Reviewed-by: Sudeep Holla <[email protected]>
Signed-off-by: Yunhui Cui <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoriscv: cacheinfo: remove the useless input parameter (node) of ci_leaf_init()
Yunhui Cui [Mon, 17 Jun 2024 13:14:23 +0000 (21:14 +0800)]
riscv: cacheinfo: remove the useless input parameter (node) of ci_leaf_init()

ci_leaf_init() is a declared static function. The implementation of the
function body and the caller do not use the parameter (struct device_node
*node) input parameter, so remove it.

Fixes: 6a24915145c9 ("Revert "riscv: Set more data to cacheinfo"")
Signed-off-by: Yunhui Cui <[email protected]>
Reviewed-by: Jeremy Linton <[email protected]>
Reviewed-by: Sudeep Holla <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoRISC-V: ACPI: Enable SPCR table for console output on RISC-V
Sia Jee Heng [Thu, 2 May 2024 07:37:51 +0000 (00:37 -0700)]
RISC-V: ACPI: Enable SPCR table for console output on RISC-V

The ACPI SPCR code has been used to enable console output for ARM64 and
X86. The same code can be reused for RISC-V. Furthermore, SPCR table is
mandated for headless system as outlined in the RISC-V BRS
Specification, chapter 6.

Signed-off-by: Sia Jee Heng <[email protected]>
Reviewed-by: Sunil V L <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoMAINTAINERS: make Breno the netconsole maintainer
Jakub Kicinski [Tue, 23 Jul 2024 22:34:05 +0000 (15:34 -0700)]
MAINTAINERS: make Breno the netconsole maintainer

netconsole has no maintainer, and Breno has been working on
improving it consistently for some time. So I think we found
the maintainer :)

Acked-by: Paolo Abeni <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>
Acked-by: Breno Leitao <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
7 months agoMAINTAINERS: Update bonding entry
Jay Vosburgh [Tue, 23 Jul 2024 22:22:36 +0000 (15:22 -0700)]
MAINTAINERS: Update bonding entry

Update my email address, clarify support status, and delete the
web site that hasn't been used in a long time.

Signed-off-by: Jay Vosburgh <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
7 months agonet: nexthop: Initialize all fields in dumped nexthops
Petr Machata [Tue, 23 Jul 2024 16:04:16 +0000 (18:04 +0200)]
net: nexthop: Initialize all fields in dumped nexthops

struct nexthop_grp contains two reserved fields that are not initialized by
nla_put_nh_group(), and carry garbage. This can be observed e.g. with
strace (edited for clarity):

    # ip nexthop add id 1 dev lo
    # ip nexthop add id 101 group 1
    # strace -e recvmsg ip nexthop get id 101
    ...
    recvmsg(... [{nla_len=12, nla_type=NHA_GROUP},
                 [{id=1, weight=0, resvd1=0x69, resvd2=0x67}]] ...) = 52

The fields are reserved and therefore not currently used. But as they are, they
leak kernel memory, and the fact they are not just zero complicates repurposing
of the fields for new ends. Initialize the full structure.

Fixes: 430a049190de ("nexthop: Add support for nexthop groups")
Signed-off-by: Petr Machata <[email protected]>
Reviewed-by: Ido Schimmel <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
7 months agonet: stmmac: Correct byte order of perfect_match
Simon Horman [Tue, 23 Jul 2024 13:29:27 +0000 (14:29 +0100)]
net: stmmac: Correct byte order of perfect_match

The perfect_match parameter of the update_vlan_hash operation is __le16,
and is correctly converted from host byte-order in the lone caller,
stmmac_vlan_update().

However, the implementations of this caller, dwxgmac2_update_vlan_hash()
and dwxgmac2_update_vlan_hash(), both treat this parameter as host byte
order, using the following pattern:

u32 value = ...
...
writel(value | perfect_match, ...);

This is not correct because both:
1) value is host byte order; and
2) writel expects a host byte order value as it's first argument

I believe that this will break on big endian systems. And I expect it
has gone unnoticed by only being exercised on little endian systems.

The approach taken by this patch is to update the callback, and it's
caller to simply use a host byte order value.

Flagged by Sparse.
Compile tested only.

Fixes: c7ab0b8088d7 ("net: stmmac: Fallback to VLAN Perfect filtering if HASH is not available")
Signed-off-by: Simon Horman <[email protected]>
Reviewed-by: Maxime Chevallier <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
7 months agoriscv: boot: remove duplicated targets line
Jisheng Zhang [Thu, 13 Jun 2024 15:30:53 +0000 (23:30 +0800)]
riscv: boot: remove duplicated targets line

The "targets:" is duplicated in another line, remove the one with less
targets.

Signed-off-by: Jisheng Zhang <[email protected]>
Reviewed-by: Emil Renner Berthing <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agotrace: riscv: Remove deprecated kprobe on ftrace support
Jinjie Ruan [Thu, 13 Jun 2024 11:13:47 +0000 (19:13 +0800)]
trace: riscv: Remove deprecated kprobe on ftrace support

Since commit 7caa9765465f60 ("ftrace: riscv: move from REGS to ARGS"),
kprobe on ftrace is not supported by riscv, because riscv's support for
FTRACE_WITH_REGS has been replaced with support for FTRACE_WITH_ARGS, and
KPROBES_ON_FTRACE will be supplanted by FPROBES. So remove the deprecated
kprobe on ftrace support, which is misunderstood.

Signed-off-by: Jinjie Ruan <[email protected]>
Acked-by: Masami Hiramatsu (Google) <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
7 months agoselftests: forwarding: skip if kernel not support setting bridge fdb learning limit
Hangbin Liu [Tue, 23 Jul 2024 08:22:52 +0000 (16:22 +0800)]
selftests: forwarding: skip if kernel not support setting bridge fdb learning limit

If the testing kernel doesn't support setting fdb_max_learned or show
fdb_n_learned, just skip it. Or we will get errors like

./bridge_fdb_learning_limit.sh: line 218: [: null: integer expression expected
./bridge_fdb_learning_limit.sh: line 225: [: null: integer expression expected

Fixes: 6f84090333bb ("selftests: forwarding: bridge_fdb_learning_limit: Add a new selftest")
Signed-off-by: Hangbin Liu <[email protected]>
Acked-by: Nikolay Aleksandrov <[email protected]>
Reviewed-by: Johannes Nixdorf <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
7 months agotipc: Return non-zero value from tipc_udp_addr2str() on error
Shigeru Yoshida [Tue, 16 Jul 2024 02:09:05 +0000 (11:09 +0900)]
tipc: Return non-zero value from tipc_udp_addr2str() on error

tipc_udp_addr2str() should return non-zero value if the UDP media
address is invalid. Otherwise, a buffer overflow access can occur in
tipc_media_addr_printf(). Fix this by returning 1 on an invalid UDP
media address.

Fixes: d0f91938bede ("tipc: add ip/udp media type")
Signed-off-by: Shigeru Yoshida <[email protected]>
Reviewed-by: Tung Nguyen <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
7 months agoASoC: SOF: ipc4-topology: Preserve the DMA Link ID for ChainDMA on unprepare
Peter Ujfalusi [Wed, 24 Jul 2024 08:19:32 +0000 (11:19 +0300)]
ASoC: SOF: ipc4-topology: Preserve the DMA Link ID for ChainDMA on unprepare

The DMA Link ID is set to the IPC message's primary during dai_config,
which is only during hw_params.
During xrun handling the hw_params is not called and the DMA Link ID
information will be lost.

All other fields in the message expected to be 0 for re-configuration, only
the DMA Link ID needs to be preserved and the in case of repeated
dai_config, it is correctly updated (masked and then set).

Cc: [email protected]
Fixes: ca5ce0caa67f ("ASoC: SOF: ipc4/intel: Add support for chained DMA")
Link: https://github.com/thesofproject/linux/issues/5116
Signed-off-by: Peter Ujfalusi <[email protected]>
Reviewed-by: Bard Liao <[email protected]>
Reviewed-by: Ranjani Sridharan <[email protected]>
Reviewed-by: Pierre-Louis Bossart <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Mark Brown <[email protected]>
7 months agoASoC: SOF: ipc4-topology: Only handle dai_config with HW_PARAMS for ChainDMA
Peter Ujfalusi [Wed, 24 Jul 2024 08:19:31 +0000 (11:19 +0300)]
ASoC: SOF: ipc4-topology: Only handle dai_config with HW_PARAMS for ChainDMA

The DMA Link ID is only valid in snd_sof_dai_config_data when the
dai_config is called with HW_PARAMS.

The commit that this patch fixes is actually moved a code section without
changing it, the same bug exists in the original code, needing different
patch to kernel prior to 6.9 kernels.

Cc: [email protected]
Fixes: 3858464de57b ("ASoC: SOF: ipc4-topology: change chain_dma handling in dai_config")
Link: https://github.com/thesofproject/linux/issues/5116
Signed-off-by: Peter Ujfalusi <[email protected]>
Reviewed-by: Bard Liao <[email protected]>
Reviewed-by: Ranjani Sridharan <[email protected]>
Reviewed-by: Pierre-Louis Bossart <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Mark Brown <[email protected]>
7 months agoinode: clarify what's locked
Christian Brauner [Wed, 24 Jul 2024 08:50:33 +0000 (10:50 +0200)]
inode: clarify what's locked

In __wait_on_freeing_inode() we warn in case the inode_hash_lock is held
but the inode is unhashed. We then release the inode_lock. So using
"locked" as parameter name is confusing. Use is_inode_hash_locked as
parameter name instead.

Signed-off-by: Christian Brauner <[email protected]>
7 months agovfs: Fix potential circular locking through setxattr() and removexattr()
David Howells [Tue, 23 Jul 2024 08:59:54 +0000 (09:59 +0100)]
vfs: Fix potential circular locking through setxattr() and removexattr()

When using cachefiles, lockdep may emit something similar to the circular
locking dependency notice below.  The problem appears to stem from the
following:

 (1) Cachefiles manipulates xattrs on the files in its cache when called
     from ->writepages().

 (2) The setxattr() and removexattr() system call handlers get the name
     (and value) from userspace after taking the sb_writers lock, putting
     accesses of the vma->vm_lock and mm->mmap_lock inside of that.

 (3) The afs filesystem uses a per-inode lock to prevent multiple
     revalidation RPCs and in writeback vs truncate to prevent parallel
     operations from deadlocking against the server on one side and local
     page locks on the other.

Fix this by moving the getting of the name and value in {get,remove}xattr()
outside of the sb_writers lock.  This also has the minor benefits that we
don't need to reget these in the event of a retry and we never try to take
the sb_writers lock in the event we can't pull the name and value into the
kernel.

Alternative approaches that might fix this include moving the dispatch of a
write to the cache off to a workqueue or trying to do without the
validation lock in afs.  Note that this might also affect other filesystems
that use netfslib and/or cachefiles.

 ======================================================
 WARNING: possible circular locking dependency detected
 6.10.0-build2+ #956 Not tainted
 ------------------------------------------------------
 fsstress/6050 is trying to acquire lock:
 ffff888138fd82f0 (mapping.invalidate_lock#3){++++}-{3:3}, at: filemap_fault+0x26e/0x8b0

 but task is already holding lock:
 ffff888113f26d18 (&vma->vm_lock->lock){++++}-{3:3}, at: lock_vma_under_rcu+0x165/0x250

 which lock already depends on the new lock.

 the existing dependency chain (in reverse order) is:

 -> #4 (&vma->vm_lock->lock){++++}-{3:3}:
        __lock_acquire+0xaf0/0xd80
        lock_acquire.part.0+0x103/0x280
        down_write+0x3b/0x50
        vma_start_write+0x6b/0xa0
        vma_link+0xcc/0x140
        insert_vm_struct+0xb7/0xf0
        alloc_bprm+0x2c1/0x390
        kernel_execve+0x65/0x1a0
        call_usermodehelper_exec_async+0x14d/0x190
        ret_from_fork+0x24/0x40
        ret_from_fork_asm+0x1a/0x30

 -> #3 (&mm->mmap_lock){++++}-{3:3}:
        __lock_acquire+0xaf0/0xd80
        lock_acquire.part.0+0x103/0x280
        __might_fault+0x7c/0xb0
        strncpy_from_user+0x25/0x160
        removexattr+0x7f/0x100
        __do_sys_fremovexattr+0x7e/0xb0
        do_syscall_64+0x9f/0x100
        entry_SYSCALL_64_after_hwframe+0x76/0x7e

 -> #2 (sb_writers#14){.+.+}-{0:0}:
        __lock_acquire+0xaf0/0xd80
        lock_acquire.part.0+0x103/0x280
        percpu_down_read+0x3c/0x90
        vfs_iocb_iter_write+0xe9/0x1d0
        __cachefiles_write+0x367/0x430
        cachefiles_issue_write+0x299/0x2f0
        netfs_advance_write+0x117/0x140
        netfs_write_folio.isra.0+0x5ca/0x6e0
        netfs_writepages+0x230/0x2f0
        afs_writepages+0x4d/0x70
        do_writepages+0x1e8/0x3e0
        filemap_fdatawrite_wbc+0x84/0xa0
        __filemap_fdatawrite_range+0xa8/0xf0
        file_write_and_wait_range+0x59/0x90
        afs_release+0x10f/0x270
        __fput+0x25f/0x3d0
        __do_sys_close+0x43/0x70
        do_syscall_64+0x9f/0x100
        entry_SYSCALL_64_after_hwframe+0x76/0x7e

 -> #1 (&vnode->validate_lock){++++}-{3:3}:
        __lock_acquire+0xaf0/0xd80
        lock_acquire.part.0+0x103/0x280
        down_read+0x95/0x200
        afs_writepages+0x37/0x70
        do_writepages+0x1e8/0x3e0
        filemap_fdatawrite_wbc+0x84/0xa0
        filemap_invalidate_inode+0x167/0x1e0
        netfs_unbuffered_write_iter+0x1bd/0x2d0
        vfs_write+0x22e/0x320
        ksys_write+0xbc/0x130
        do_syscall_64+0x9f/0x100
        entry_SYSCALL_64_after_hwframe+0x76/0x7e

 -> #0 (mapping.invalidate_lock#3){++++}-{3:3}:
        check_noncircular+0x119/0x160
        check_prev_add+0x195/0x430
        __lock_acquire+0xaf0/0xd80
        lock_acquire.part.0+0x103/0x280
        down_read+0x95/0x200
        filemap_fault+0x26e/0x8b0
        __do_fault+0x57/0xd0
        do_pte_missing+0x23b/0x320
        __handle_mm_fault+0x2d4/0x320
        handle_mm_fault+0x14f/0x260
        do_user_addr_fault+0x2a2/0x500
        exc_page_fault+0x71/0x90
        asm_exc_page_fault+0x22/0x30

 other info that might help us debug this:

 Chain exists of:
   mapping.invalidate_lock#3 --> &mm->mmap_lock --> &vma->vm_lock->lock

  Possible unsafe locking scenario:

        CPU0                    CPU1
        ----                    ----
   rlock(&vma->vm_lock->lock);
                                lock(&mm->mmap_lock);
                                lock(&vma->vm_lock->lock);
   rlock(mapping.invalidate_lock#3);

  *** DEADLOCK ***

 1 lock held by fsstress/6050:
  #0: ffff888113f26d18 (&vma->vm_lock->lock){++++}-{3:3}, at: lock_vma_under_rcu+0x165/0x250

 stack backtrace:
 CPU: 0 PID: 6050 Comm: fsstress Not tainted 6.10.0-build2+ #956
 Hardware name: ASUS All Series/H97-PLUS, BIOS 2306 10/09/2014
 Call Trace:
  <TASK>
  dump_stack_lvl+0x57/0x80
  check_noncircular+0x119/0x160
  ? queued_spin_lock_slowpath+0x4be/0x510
  ? __pfx_check_noncircular+0x10/0x10
  ? __pfx_queued_spin_lock_slowpath+0x10/0x10
  ? mark_lock+0x47/0x160
  ? init_chain_block+0x9c/0xc0
  ? add_chain_block+0x84/0xf0
  check_prev_add+0x195/0x430
  __lock_acquire+0xaf0/0xd80
  ? __pfx___lock_acquire+0x10/0x10
  ? __lock_release.isra.0+0x13b/0x230
  lock_acquire.part.0+0x103/0x280
  ? filemap_fault+0x26e/0x8b0
  ? __pfx_lock_acquire.part.0+0x10/0x10
  ? rcu_is_watching+0x34/0x60
  ? lock_acquire+0xd7/0x120
  down_read+0x95/0x200
  ? filemap_fault+0x26e/0x8b0
  ? __pfx_down_read+0x10/0x10
  ? __filemap_get_folio+0x25/0x1a0
  filemap_fault+0x26e/0x8b0
  ? __pfx_filemap_fault+0x10/0x10
  ? find_held_lock+0x7c/0x90
  ? __pfx___lock_release.isra.0+0x10/0x10
  ? __pte_offset_map+0x99/0x110
  __do_fault+0x57/0xd0
  do_pte_missing+0x23b/0x320
  __handle_mm_fault+0x2d4/0x320
  ? __pfx___handle_mm_fault+0x10/0x10
  handle_mm_fault+0x14f/0x260
  do_user_addr_fault+0x2a2/0x500
  exc_page_fault+0x71/0x90
  asm_exc_page_fault+0x22/0x30

Signed-off-by: David Howells <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
cc: Alexander Viro <[email protected]>
cc: Christian Brauner <[email protected]>
cc: Jan Kara <[email protected]>
cc: Jeff Layton <[email protected]>
cc: Gao Xiang <[email protected]>
cc: Matthew Wilcox <[email protected]>
cc: [email protected]
cc: [email protected]
cc: [email protected]
[brauner: fix minor issues]
Signed-off-by: Christian Brauner <[email protected]>
This page took 0.150495 seconds and 4 git commands to generate.