linux-2.6-microblaze.git
2 years agoMerge tag 'gpio-updates-for-v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 8 Nov 2021 19:55:21 +0000 (11:55 -0800)]
Merge tag 'gpio-updates-for-v5.16' of git://git./linux/kernel/git/brgl/linux

Pull gpio updates from Bartosz Golaszewski:
 "We have a single new driver, new features in others and some cleanups
  all over the place.

  Nothing really stands out and it is all relatively small.

   - new driver: gpio-modepin (plus relevant change in zynqmp firmware)

   - add interrupt support to gpio-virtio

   - enable the 'gpio-line-names' property in the DT bindings for
     gpio-rockchip

   - use the subsystem helpers where applicable in gpio-uniphier instead
     of accessing IRQ structures directly

   - code shrink in gpio-xilinx

   - add interrupt to gpio-mlxbf2 (and include the removal of custom
     interrupt code from the mellanox ethernet driver)

   - support multiple interrupts per bank in gpio-tegra186 (and force
     one interrupt per bank in older models)

   - fix GPIO line IRQ offset calculation in gpio-realtek-otto

   - drop unneeded MODULE_ALIAS expansions in multiple drivers

   - code cleanup in gpio-aggregator

   - minor improvements in gpio-max730x and gpio-mc33880

   - Kconfig cleanups"

* tag 'gpio-updates-for-v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux:
  virtio_gpio: drop packed attribute
  gpio: virtio: Add IRQ support
  gpio: realtek-otto: fix GPIO line IRQ offset
  gpio: clean up Kconfig file
  net: mellanox: mlxbf_gige: Replace non-standard interrupt handling
  gpio: mlxbf2: Introduce IRQ support
  gpio: mc33880: Drop if with an always false condition
  gpio: max730x: Make __max730x_remove() return void
  gpio: aggregator: Wrap access to gpiochip_fwd.tmp[]
  gpio: modepin: Add driver support for modepin GPIO controller
  dt-bindings: gpio: zynqmp: Add binding documentation for modepin
  firmware: zynqmp: Add MMIO read and write support for PS_MODE pin
  gpio: tps65218: drop unneeded MODULE_ALIAS
  gpio: max77620: drop unneeded MODULE_ALIAS
  gpio: xilinx: simplify getting .driver_data
  gpio: tegra186: Support multiple interrupts per bank
  gpio: tegra186: Force one interrupt per bank
  gpio: uniphier: Use helper functions to get private data from IRQ data
  gpio: uniphier: Use helper function to get IRQ hardware number
  dt-bindings: gpio: add gpio-line-names to rockchip,gpio-bank.yaml

2 years agoMerge tag 'cxl-for-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl
Linus Torvalds [Mon, 8 Nov 2021 19:49:48 +0000 (11:49 -0800)]
Merge tag 'cxl-for-5.16' of git://git./linux/kernel/git/cxl/cxl

Pull cxl updates from Dan Williams:
 "More preparation and plumbing work in the CXL subsystem.

  From an end user perspective the highlight here is lighting up the CXL
  Persistent Memory related commands (label read / write) with the
  generic ioctl() front-end in LIBNVDIMM.

  Otherwise, the ability to instantiate new persistent and volatile
  memory regions is still on track for v5.17.

  Summary:

   - Fix support for platforms that do not enumerate every ACPI0016 (CXL
     Host Bridge) in the CHBS (ACPI Host Bridge Structure).

   - Introduce a common pci_find_dvsec_capability() helper, clean up
     open coded implementations in various drivers.

   - Add 'cxl_test' for regression testing CXL subsystem ABIs.
     'cxl_test' is a module built from tools/testing/cxl/ that mocks up
     a CXL topology to augment the nascent support for emulation of CXL
     devices in QEMU.

   - Convert libnvdimm to use the uuid API.

   - Complete the definition of CXL namespace labels in libnvdimm.

   - Tunnel libnvdimm label operations from nd_ioctl() back to the CXL
     mailbox driver. Enable 'ndctl {read,write}-labels' for CXL.

   - Continue to sort and refactor functionality into distinct driver
     and core-infrastructure buckets. For example, mailbox handling is
     now a generic core capability consumed by the PCI and cxl_test
     drivers"

* tag 'cxl-for-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl: (34 commits)
  ocxl: Use pci core's DVSEC functionality
  cxl/pci: Use pci core's DVSEC functionality
  PCI: Add pci_find_dvsec_capability to find designated VSEC
  cxl/pci: Split cxl_pci_setup_regs()
  cxl/pci: Add @base to cxl_register_map
  cxl/pci: Make more use of cxl_register_map
  cxl/pci: Remove pci request/release regions
  cxl/pci: Fix NULL vs ERR_PTR confusion
  cxl/pci: Remove dev_dbg for unknown register blocks
  cxl/pci: Convert register block identifiers to an enum
  cxl/acpi: Do not fail cxl_acpi_probe() based on a missing CHBS
  cxl/pci: Disambiguate cxl_pci further from cxl_mem
  Documentation/cxl: Add bus internal docs
  cxl/core: Split decoder setup into alloc + add
  tools/testing/cxl: Introduce a mock memory device + driver
  cxl/mbox: Move command definitions to common location
  cxl/bus: Populate the target list at decoder create
  tools/testing/cxl: Introduce a mocked-up CXL port hierarchy
  cxl/pmem: Add support for multiple nvdimm-bridge objects
  cxl/pmem: Translate NVDIMM label commands to CXL label commands
  ...

2 years agoMerge branch 'i2c/for-mergewindow' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 8 Nov 2021 19:46:10 +0000 (11:46 -0800)]
Merge branch 'i2c/for-mergewindow' of git://git./linux/kernel/git/wsa/linux

Pull i2c updates from Wolfram Sang:

 - big refactoring of the PASEMI driver to support the Apple M1

 - huge improvements to the XIIC in terms of locking and SMP safety

 - refactoring and clean ups for the i801 driver

... and the usual bunch of small driver updates

* 'i2c/for-mergewindow' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux: (43 commits)
  i2c: amd-mp2-plat: ACPI: Use ACPI_COMPANION() directly
  i2c: i801: Add support for Intel Ice Lake PCH-N
  i2c: virtio: update the maintainer to Conghui
  i2c: xlr: Fix a resource leak in the error handling path of 'xlr_i2c_probe()'
  i2c: qup: move to use request_irq by IRQF_NO_AUTOEN flag
  i2c: qup: fix a trivial typo
  i2c: tegra: Ensure that device is suspended before driver is removed
  i2c: i801: Fix incorrect and needless software PEC disabling
  i2c: mediatek: Dump i2c/dma register when a timeout occurs
  i2c: mediatek: Reset the handshake signal between i2c and dma
  i2c: mlxcpld: Allow flexible polling time setting for I2C transactions
  i2c: pasemi: Set enable bit for Apple variant
  i2c: pasemi: Add Apple platform driver
  i2c: pasemi: Refactor _probe to use devm_*
  i2c: pasemi: Allow to configure bus frequency
  i2c: pasemi: Move common reset code to own function
  i2c: pasemi: Split pci driver to its own file
  i2c: pasemi: Split off common probing code
  i2c: pasemi: Remove usage of pci_dev
  i2c: pasemi: Use dev_name instead of port number
  ...

2 years agoMerge tag 'mtd/for-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux
Linus Torvalds [Mon, 8 Nov 2021 19:37:39 +0000 (11:37 -0800)]
Merge tag 'mtd/for-5.16' of git://git./linux/kernel/git/mtd/linux

Pull mtd updates from Miquel Raynal:
 "Core:
   - Remove obsolete macros only used by the old nand_ecclayout struct
   - Don't remove debugfs directory if device is in use
   - MAINTAINERS:
      - Add entry for Qualcomm NAND controller driver
      - Update the devicetree documentation path of hyperbus

  MTD devices:
   - block2mtd:
      - Add support for an optional custom MTD label
      - Minor refactor to avoid hard coded constant
   - mtdswap: Remove redundant assignment of pointer eb

  CFI:
   - Fixup CFI on ixp4xx

  Raw NAND controller drivers:
   - Arasan:
      - Prevent an unsupported configuration
   - Xway, Socrates: plat_nand, Pasemi, Orion, mpc5121, GPIO, Au1550nd,
     AMS-Delta:
      - Keep the driver compatible with on-die ECC engines
   - cs553x, lpc32xx_slc, ndfc, sharpsl, tmio, txx9ndfmc:
      - Revert the commits: "Fix external use of SW Hamming ECC helper"
      - And let callers use the bare Hamming helpers
   - Fsmc: Fix use of SM ORDER
   - Intel:
      - Fix potential buffer overflow in probe
   - xway, vf610, txx9ndfm, tegra, stm32, plat_nand, oxnas, omap, mtk,
     hisi504, gpmi, gpio, denali, bcm6368, atmel:
      - Make use of the helper function devm_platform_ioremap_resource{,byname}()

  Onenand drivers:
   - Samsung: Drop Exynos4 and describe driver in KConfig

  Raw NAND chip drivers:
   - Hynix: Add support for H27UCG8T2ETR-BC MLC NAND

  SPI NOR core:
   - Add spi-nor device tree binding under SPI NOR maintainers

  SPI NOR manufacturer drivers:
   - Enable locking for n25q128a13

  SPI NOR controller drivers:
   - Use devm_platform_ioremap_resource_byname()"

* tag 'mtd/for-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux: (50 commits)
  mtd: core: don't remove debugfs directory if device is in use
  MAINTAINERS: Update the devicetree documentation path of hyperbus
  mtd: block2mtd: add support for an optional custom MTD label
  mtd: block2mtd: minor refactor to avoid hard coded constant
  mtd: fixup CFI on ixp4xx
  mtd: rawnand: arasan: Prevent an unsupported configuration
  MAINTAINERS: Add entry for Qualcomm NAND controller driver
  mtd: rawnand: hynix: Add support for H27UCG8T2ETR-BC MLC NAND
  mtd: rawnand: xway: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: socrates: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: plat_nand: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: pasemi: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: orion: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: mpc5121: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: gpio: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: au1550nd: Keep the driver compatible with on-die ECC engines
  mtd: rawnand: ams-delta: Keep the driver compatible with on-die ECC engines
  Revert "mtd: rawnand: cs553x: Fix external use of SW Hamming ECC helper"
  Revert "mtd: rawnand: lpc32xx_slc: Fix external use of SW Hamming ECC helper"
  Revert "mtd: rawnand: ndfc: Fix external use of SW Hamming ECC helper"
  ...

2 years agoAdd 'tools/perf/libbpf/' to ignored files
Linus Torvalds [Mon, 8 Nov 2021 19:33:35 +0000 (11:33 -0800)]
Add 'tools/perf/libbpf/' to ignored files

Commit 6b491a86b77c ("perf build: Install libbpf headers locally when
building") installed copies of the libbpf headers into the build tree,
causing unnecessary noise from 'git status' after a perf tools build.

Add the 'libbpf/' subdirectory to the .gitignore file to silence it all
again.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoMerge tag 'kgdb-5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt...
Linus Torvalds [Mon, 8 Nov 2021 17:35:43 +0000 (09:35 -0800)]
Merge tag 'kgdb-5.16-rc1' of git://git./linux/kernel/git/danielt/linux

Pull kgdb update from Daniel Thompson:
 "A single patch this cycle.

  We replace some open-coded routines to classify task states with the
  scheduler's own function to do this. Alongside the obvious benefits of
  removing funky code and aligning more exactly with the scheduler's
  task classification, this also fixes a long standing compiler warning
  by removing the open-coded routines that generated the warning"

* tag 'kgdb-5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/danielt/linux:
  kdb: Adopt scheduler's task classification

2 years agoMerge tag 'for-linus' of git://github.com/openrisc/linux
Linus Torvalds [Mon, 8 Nov 2021 17:31:25 +0000 (09:31 -0800)]
Merge tag 'for-linus' of git://github.com/openrisc/linux

Pull OpenRISC updates from Stafford Horne:
 "This includes two minor cleanups, plus a bug fix for OpenRISC TLB
  flush code that allows the the SMP kernel to boot again"

* tag 'for-linus' of git://github.com/openrisc/linux:
  openrisc: fix SMP tlb flush NULL pointer dereference
  openrisc: signal: remove unused DEBUG_SIG macro
  openrisc: time: don't mark comment as kernel-doc

2 years agoMerge tag 'perf-tools-for-v5.16-2021-11-07-without-bpftool-fix' of git://git.kernel...
Linus Torvalds [Mon, 8 Nov 2021 17:25:26 +0000 (09:25 -0800)]
Merge tag 'perf-tools-for-v5.16-2021-11-07-without-bpftool-fix' of git://git./linux/kernel/git/acme/linux

Pull perf tools updates from Arnaldo Carvalho de Melo:
 "perf annotate:
   - Add riscv64 support.
   - Add fusion logic for AMD microarchs.

  perf record:
   - Add an option to control the synthesizing behavior:
       --synth <no|all|task|mmap|cgroup>

  core:
   - Allow controlling synthesizing PERF_RECORD_ metadata events during
     record.
   - perf.data reader prep work for multithreaded processing.
   - Fix missing exclude_{host,guest} setting in PMUs that don't support
     it and that were causing the feature detection code to disable it
     for all events, even the ones in PMUs that support it.
   - Fix the default use of precise events on AMD, that were always
     falling back to non-precise because perf_event_attr.exclude_guest=1
     was set and IBS does not have filtering capability, refusing
     precise + exclude_guest.
   - Add bitfield_swap() to handle branch_stack endian issue.

  perf script:
   - Show binary offsets for userspace addresses in callchains.
   - Support instruction latency via new "ins_lat" selectable field.
   - Add dlfilter-show-cycles

  perf inject:
   - Add vmlinux and ignore-vmlinux arguments, similar to other tools.

  perf list:
   - Display PMU prefix for partially supported hybrid cache events.
   - Display hybrid PMU events with cpu type.

  perf stat:
   - Improve metrics documentation of data structures.
   - Fix memory leaks in the metric code.
   - Use NAN for missing event IDs.
   - Don't compute unused events.
   - Fix memory leak on error path.
   - Encode and use metric-id as a metric qualifier.
   - Allow metrics with no events.
   - Avoid events for an 'if' constant result.
   - Only add a referenced metric once.
   - Simplify metric_refs calculation.
   - Allow modifiers on metrics.

  perf test:
   - Add workload test of metric and metric groups.
   - Workload test of all PMUs.
   - vmlinux-kallsyms: Ignore hidden symbols.
   - Add pmu-event test for event described as "config=".
   - Verify more event members in pmu-events test.
   - Add endian test for struct branch_flags on the sample-parsing test.
   - Improve temp file cleanup in several tests.

  perf daemon:
   - Address MSAN warnings on send_cmd().

  perf kmem:
   - Improve man page for record options

  perf srcline:
   - Use long-running addr2line per DSO, greatly speeding up the
     'srcline' sort order.

  perf symbols:
   - Ignore $a/$d symbols for ARM modules.
   - Fix /proc/kcore access on 32 bit systems.

  Kernel UAPI copies:
   - Update copy of linux/socket.h with the kernel sources, no change in
     tooling output.

  libbpf:
   - Pull in bpf_program__get_prog_info_linear() from libbpf, too much
     specific to perf.
   - Deprecate bpf_map__resize() in favor of bpf_map_set_max_entries()
   - Install libbpf headers locally when building.
   - Bump minimum LLVM C++ std to GNU++14.

  libperf:
   - Use binary search in perf_cpu_map__idx() as array are sorted.

  libtracefs:
   - Enable libtracefs dynamic linking.

  libtraceevent:
   - Increase logging when verbose.

  Arch specific:

   * PowerPC:
      - Add support to expose instruction and data address registers as
        part of extended regs.

  Vendor events:

   * JSON parser:
      - Support ConfigCode to set the config= in PMUs
      - Make the JSON parser more conformant when in strict mode.

   * All JSON files:
      - Fix all remaining invalid JSON files.

   * ARM:
      - Syntax corrections in Neoverse N1 json.
      - Categorise the Neoverse V1 counters.
      - Add new armv8 PMU events.
      - Revise hip08 uncore events.

  Hardware tracing:

   * auxtrace:
      - Add missing Z option to ITRACE_HELP.
      - Add itrace A option to approximate IPC.
      - Add itrace d+o option to direct debug log to stdout.

   * Intel PT:
      - Add support for PERF_RECORD_AUX_OUTPUT_HW_ID
      - Support itrace A option to approximate IPC
      - Support itrace d+o option to direct debug log to stdout"

* tag 'perf-tools-for-v5.16-2021-11-07-without-bpftool-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (120 commits)
  perf build: Install libbpf headers locally when building
  perf MANIFEST: Add bpftool files to allow building with BUILD_BPF_SKEL=1
  perf metric: Fix memory leaks
  perf parse-event: Add init and exit to parse_event_error
  perf parse-events: Rename parse_events_error functions
  perf stat: Fix memory leak on error path
  perf tools: Use __BYTE_ORDER__
  perf inject: Add vmlinux and ignore-vmlinux arguments
  perf tools: Check vmlinux/kallsyms arguments in all tools
  perf tools: Refactor out kernel symbol argument sanity checking
  perf symbols: Ignore $a/$d symbols for ARM modules
  perf evsel: Don't set exclude_guest by default
  perf evsel: Fix missing exclude_{host,guest} setting
  perf bpf: Add missing free to bpf_event__print_bpf_prog_info()
  perf beauty: Update copy of linux/socket.h with the kernel sources
  perf clang: Fixes for more recent LLVM/clang
  tools: Bump minimum LLVM C++ std to GNU++14
  perf bpf: Pull in bpf_program__get_prog_info_linear()
  Revert "perf bench futex: Add support for 32-bit systems with 64-bit time_t"
  perf test sample-parsing: Add endian test for struct branch_flags
  ...

2 years agoMerge tag 'kbuild-v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy...
Linus Torvalds [Mon, 8 Nov 2021 17:15:45 +0000 (09:15 -0800)]
Merge tag 'kbuild-v5.16' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild updates from Masahiro Yamada:

 - Remove the global -isystem compiler flag, which was made possible by
   the introduction of <linux/stdarg.h>

 - Improve the Kconfig help to print the location in the top menu level

 - Fix "FORCE prerequisite is missing" build warning for sparc

 - Add new build targets, tarzst-pkg and perf-tarzst-src-pkg, which
   generate a zstd-compressed tarball

 - Prevent gen_init_cpio tool from generating a corrupted cpio when
   KBUILD_BUILD_TIMESTAMP is set to 2106-02-07 or later

 - Misc cleanups

* tag 'kbuild-v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild: (28 commits)
  kbuild: use more subdir- for visiting subdirectories while cleaning
  sh: remove meaningless archclean line
  initramfs: Check timestamp to prevent broken cpio archive
  kbuild: split DEBUG_CFLAGS out to scripts/Makefile.debug
  gen_init_cpio: add static const qualifiers
  kbuild: Add make tarzst-pkg build option
  scripts: update the comments of kallsyms support
  sparc: Add missing "FORCE" target when using if_changed
  kconfig: refactor conf_touch_dep()
  kconfig: refactor conf_write_dep()
  kconfig: refactor conf_write_autoconf()
  kconfig: add conf_get_autoheader_name()
  kconfig: move sym_escape_string_value() to confdata.c
  kconfig: refactor listnewconfig code
  kconfig: refactor conf_write_symbol()
  kconfig: refactor conf_write_heading()
  kconfig: remove 'const' from the return type of sym_escape_string_value()
  kconfig: rename a variable in the lexer to a clearer name
  kconfig: narrow the scope of variables in the lexer
  kconfig: Create links to main menu items in search
  ...

2 years agoMerge tag 'modules-5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof...
Linus Torvalds [Mon, 8 Nov 2021 17:04:59 +0000 (09:04 -0800)]
Merge tag 'modules-5.16-rc1' of git://git./linux/kernel/git/mcgrof/linux

Pull module updates from Luis Chamberlain:
 "As requested by Jessica I'm stepping in to help with modules
  maintenance. This is my first pull request to you.

  I've collected only two patches for modules for the 5.16-rc1 merge
  window. These patches are from Shuah Khan as she debugged some corner
  case error with modules. The error messages are improved for
  elf_validity_check(). While doing this work a corner case fix was
  spotted on validate_section_offset() due to a possible overflow bug on
  64-bit. The impact of this fix is low given this just limits module
  section headers placed within the 32-bit boundary, and we obviously
  don't have insane module sizes. Even if a specially crafted module is
  constructed later checks would invalidate the module right away.

  I've let this sit through 0-day testing since October 15th with no
  issues found"

* tag 'modules-5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux:
  module: change to print useful messages from elf_validity_check()
  module: fix validate_section_offset() overflow bug on 64-bit

2 years agoAdd gitignore file for samples/fanotify/ subdirectory
Linus Torvalds [Sun, 7 Nov 2021 19:19:24 +0000 (11:19 -0800)]
Add gitignore file for samples/fanotify/ subdirectory

Commit 5451093081db ("samples: Add fs error monitoring example") added a
new sample program, but didn't teach git to ignore the new generated
files, causing unnecessary noise from 'git status' after a full build.

Add the 'fs-monitor' sample executable to the .gitignore for this
subdirectory to silence it all again.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sun, 7 Nov 2021 18:52:04 +0000 (10:52 -0800)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fix from Wolfram Sang:
 "Hot-fix for I2C"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: xgene-slimpro: Fix wrong pointer passed to PTR_ERR()

2 years agoMerge tag 'auxdisplay-for-linus-v5.16' of git://github.com/ojeda/linux
Linus Torvalds [Sun, 7 Nov 2021 18:47:27 +0000 (10:47 -0800)]
Merge tag 'auxdisplay-for-linus-v5.16' of git://github.com/ojeda/linux

Pull auxdisplay updates from Miguel Ojeda:

 - 4-digit 7-segment and quad alphanumeric display support for the
   ht16k33 driver, allowing the user to display and scroll text
   messages, from Geert Uytterhoeven.

 - An assortment of fixes and cleanups from Geert Uytterhoeven.

 - Header cleanups from Mianhan Liu.

 - Whitespace cleanup from Huiquan Deng.

* tag 'auxdisplay-for-linus-v5.16' of git://github.com/ojeda/linux: (26 commits)
  MAINTAINERS: Add DT Bindings for Auxiliary Display Drivers
  auxdisplay: cfag12864bfb: code indent should use tabs where possible
  auxdisplay: ht16k33: remove superfluous header files
  auxdisplay: ks0108: remove superfluous header files
  auxdisplay: cfag12864bfb: remove superfluous header files
  auxdisplay: ht16k33: Make use of device properties
  auxdisplay: ht16k33: Add LED support
  dt-bindings: auxdisplay: ht16k33: Document LED subnode
  auxdisplay: ht16k33: Add support for segment displays
  auxdisplay: ht16k33: Extract frame buffer probing
  auxdisplay: ht16k33: Extract ht16k33_brightness_set()
  auxdisplay: ht16k33: Move delayed work
  auxdisplay: ht16k33: Add helper variable dev
  auxdisplay: ht16k33: Convert to simple i2c probe function
  auxdisplay: ht16k33: Remove unneeded error check in keypad probe()
  auxdisplay: ht16k33: Use HT16K33_FB_SIZE in ht16k33_initialize()
  auxdisplay: ht16k33: Fix frame buffer device blanking
  auxdisplay: ht16k33: Connect backlight to fbdev
  auxdisplay: linedisp: Add support for changing scroll rate
  auxdisplay: linedisp: Use kmemdup_nul() helper
  ...

2 years agoperf build: Install libbpf headers locally when building
Quentin Monnet [Sun, 7 Nov 2021 00:24:45 +0000 (00:24 +0000)]
perf build: Install libbpf headers locally when building

API headers from libbpf should not be accessed directly from the
library's source directory. Instead, they should be exported with "make
install_headers". Let's adjust perf's Makefile to install those headers
locally when building libbpf.

v2:

- Fix $(LIBBPF_OUTPUT) when $(OUTPUT) is null.
- Make sure the recipe for $(LIBBPF_OUTPUT) is not under a "ifdef".

Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: bpf@vger.kernel.org
Cc: netdev@vger.kernel.org
Link: http://lore.kernel.org/lkml/20211107002445.4790-1-quentin@isovalent.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf MANIFEST: Add bpftool files to allow building with BUILD_BPF_SKEL=1
Arnaldo Carvalho de Melo [Sun, 7 Nov 2021 15:43:20 +0000 (12:43 -0300)]
perf MANIFEST: Add bpftool files to allow building with BUILD_BPF_SKEL=1

We need bpftool and required kernel/bpf/disasm.[ch] to bootstrap the
cgroups, bperf and other BPF skels used by perf.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf metric: Fix memory leaks
Ian Rogers [Sun, 7 Nov 2021 09:00:02 +0000 (01:00 -0800)]
perf metric: Fix memory leaks

Certain error paths may leak memory as caught by address sanitizer.
Ensure this is cleaned up to make sure address/leak sanitizer is happy.

Fixes: 5ecd5a0c7d1cca79 ("perf metrics: Modify setup and deduplication")
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20211107090002.3784612-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf parse-event: Add init and exit to parse_event_error
Ian Rogers [Sun, 7 Nov 2021 09:00:01 +0000 (01:00 -0800)]
perf parse-event: Add init and exit to parse_event_error

parse_events() may succeed but leave string memory allocations reachable
in the error.

Add an init/exit that must be called to initialize and clean up the
error. This fixes a leak in metricgroup parse_ids.

Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20211107090002.3784612-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf parse-events: Rename parse_events_error functions
Ian Rogers [Sun, 7 Nov 2021 09:00:00 +0000 (01:00 -0800)]
perf parse-events: Rename parse_events_error functions

Group error functions and name after the data type they manipulate.

Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20211107090002.3784612-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoMerge tag 'compiler-attributes-for-linus-v5.16' of git://github.com/ojeda/linux
Linus Torvalds [Sun, 7 Nov 2021 18:38:17 +0000 (10:38 -0800)]
Merge tag 'compiler-attributes-for-linus-v5.16' of git://github.com/ojeda/linux

Pull compiler attributes update from Miguel Ojeda:
 "An improvement for `__compiletime_assert` and a trivial cleanup"

* tag 'compiler-attributes-for-linus-v5.16' of git://github.com/ojeda/linux:
  compiler_types: mark __compiletime_assert failure as __noreturn
  Compiler Attributes: remove GCC 5.1 mention

2 years agoMerge tag 'spi-nor/for-5.16' into mtd/next
Miquel Raynal [Sun, 7 Nov 2021 16:38:36 +0000 (17:38 +0100)]
Merge tag 'spi-nor/for-5.16' into mtd/next

SPI NOR core changes:
- Add spi-nor device tree binding under SPI NOR maintainers

SPI NOR manufacturer drivers changes:
- Enable locking for n25q128a13

SPI NOR controller drivers changes:
- Use devm_platform_ioremap_resource_byname()

2 years agoMerge tag 'nand/for-5.16' into mtd/next
Miquel Raynal [Sun, 7 Nov 2021 16:38:05 +0000 (17:38 +0100)]
Merge tag 'nand/for-5.16' into mtd/next

Core:
* Remove obsolete macros only used by the old nand_ecclayout struct
* MAINTAINERS: Add entry for Qualcomm NAND controller driver

Raw NAND controller drivers:
* Arasan:
  - Prevent an unsupported configuration
* Xway, Socrates: plat_nand, Pasemi, Orion, mpc5121, GPIO, Au1550nd, AMS-Delta:
  - Keep the driver compatible with on-die ECC engines
* cs553x, lpc32xx_slc, ndfc, sharpsl, tmio, txx9ndfmc:
  - Revert the commits: "Fix external use of SW Hamming ECC helper"
  - And let callers use the bare Hamming helpers
* Fsmc: Fix use of SM ORDER
* Intel:
  - Fix potential buffer overflow in probe
* xway, vf610, txx9ndfm, tegra, stm32, plat_nand, oxnas, omap, mtk, hisi504,
  gpmi, gpio, denali, bcm6368, atmel:
  - Make use of the helper function devm_platform_ioremap_resource{,byname}()

Onenand driver:
* Samsung: Drop Exynos4 and describe driver in KConfig

Raw NAND chip drivers:
* Hynix: Add support for H27UCG8T2ETR-BC MLC NAND

2 years agoperf stat: Fix memory leak on error path
Ian Rogers [Sun, 7 Nov 2021 08:54:44 +0000 (01:54 -0700)]
perf stat: Fix memory leak on error path

strdup() is used to deduplicate, ensure it isn't leaking an already
created string by freeing first.

Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Garry <john.garry@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20211107085444.3781604-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf tools: Use __BYTE_ORDER__
Ilya Leoshkevich [Thu, 4 Nov 2021 13:23:11 +0000 (14:23 +0100)]
perf tools: Use __BYTE_ORDER__

Switch from the libc-defined __BYTE_ORDER to the compiler-defined
__BYTE_ORDER__ in order to make endianness detection more robust, like
it was done for libbpf.

Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Suggested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Link: https://lore.kernel.org/r/20211104132311.984703-1-iii@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf inject: Add vmlinux and ignore-vmlinux arguments
James Clark [Mon, 18 Oct 2021 13:48:43 +0000 (14:48 +0100)]
perf inject: Add vmlinux and ignore-vmlinux arguments

Other perf tools allow specifying the path to vmlinux. 'perf inject'
didn't have this argument which made some auxtrace workflows difficult.

Also add --ignore-vmlinux for consistency with other tools.

Suggested-by: Denis Nikitin <denik@chromium.org>
Signed-off-by: James Clark <james.clark@arm.com>
Tested-by: Denis Nikitin <denik@chromium.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20211018134844.2627174-4-james.clark@arm.com
[ Added the perf-inject man page entries for these options, as noted by Denis ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf tools: Check vmlinux/kallsyms arguments in all tools
James Clark [Mon, 18 Oct 2021 13:48:42 +0000 (14:48 +0100)]
perf tools: Check vmlinux/kallsyms arguments in all tools

Only perf report checked the validity of these arguments so apply the
same check to all tools that read them for consistency.

Signed-off-by: James Clark <james.clark@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Denis Nikitin <denik@chromium.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20211018134844.2627174-3-james.clark@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf tools: Refactor out kernel symbol argument sanity checking
James Clark [Mon, 18 Oct 2021 13:48:41 +0000 (14:48 +0100)]
perf tools: Refactor out kernel symbol argument sanity checking

User supplied values for vmlinux and kallsyms are checked before
continuing. Refactor this into a function so that it can be used
elsewhere.

Reviewed-by: Denis Nikitin <denik@chromium.org>
Signed-off-by: James Clark <james.clark@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20211018134844.2627174-2-james.clark@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf symbols: Ignore $a/$d symbols for ARM modules
Lexi Shao [Fri, 29 Oct 2021 06:50:37 +0000 (14:50 +0800)]
perf symbols: Ignore $a/$d symbols for ARM modules

On anARM machine, kernel symbols from modules can be resolved to $a
instead of printing the actual symbol name. Ignore symbols starting with
"$" when building kallsyms rbtree.

A sample stacktrace is shown as follows:

  c0f2e39c schedule_hrtimeout+0x14 ([kernel.kallsyms])
  bf4a66d8 $a+0x78 ([test_module])
  c0a4f5f4 kthread+0x15c ([kernel.kallsyms])
  c0a001f8 ret_from_fork+0x14 ([kernel.kallsyms])

On an ARM machine, $a/$d symbols are used by the compiler to mark the
beginning of code/data part in code section. These symbols are filtered
out when linking vmlinux(see scripts/kallsyms.c ignored_prefixes), but
are left on modules. So there are $a symbols in /proc/kallsyms which
share the same addresses with the actual module symbols and confuses
perf when resolving symbols.

After this patch, the module symbol name is printed:

  c0f2e39c schedule_hrtimeout+0x14 ([kernel.kallsyms])
  bf4a66d8 test_func+0x78 ([test_module])
  c0a4f5f4 kthread+0x15c ([kernel.kallsyms])
  c0a001f8 ret_from_fork+0x14 ([kernel.kallsyms])

Reviewed-by: James Clark <james.clark@arm.com>
Signed-off-by: Lexi Shao <shaolexi@huawei.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jessica Yu <jeyu@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: KP Singh <kpsingh@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nathan Chancellor <natechancellor@gmail.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: QiuXi <qiuxi1@huawei.com>
Cc: Song Liu <songliubraving@fb.com>
Cc: Wangbing <wangbing6@huawei.com>
Cc: Xiaoming Ni <nixiaoming@huawei.com>
Cc: Yonghong Song <yhs@fb.com>
Cc: bpf@vger.kernel.org
Cc: clang-built-linux@googlegroups.com
Link: https://lore.kernel.org/r/20211029065038.39449-2-shaolexi@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf evsel: Don't set exclude_guest by default
Ravi Bangoria [Tue, 2 Nov 2021 05:31:12 +0000 (11:01 +0530)]
perf evsel: Don't set exclude_guest by default

Perf tool sets exclude_guest by default while calling perf_event_open().
Because IBS does not have filtering capability, it always gets rejected
by IBS PMU driver and thus perf falls back to non-precise sampling. Fix
it by not setting exclude_guest by default on AMD.

Before:
  $ sudo ./perf record -C 0 -vvv true |& grep precise
    precise_ip                       3
  decreasing precise_ip by one (2)
    precise_ip                       2
  decreasing precise_ip by one (1)
    precise_ip                       1
  decreasing precise_ip by one (0)

After:
  $ sudo ./perf record -C 0 -vvv true |& grep precise
    precise_ip                       3
  decreasing precise_ip by one (2)
    precise_ip                       2

Committer notes:

Fixup init to zero for perf_env in older compilers:

  arch/x86/util/evsel.c:15:26: error: missing field 'os_release' initializer [-Werror,-Wmissing-field-initializers]
          struct perf_env env = {0};
                                  ^

Committer notes:

Namhyung remarked:

  It'd be nice if it can cover explicit "-e cycles:pp" as well.

Ravi clarified:

  For explicit :pp modifier, evsel->precise_max does not get set and thus perf
  does not try with different attr->precise_ip values while exclude_guest set.
  So no issue with explicit :pp:

    $ sudo ./perf record -C 0 -e cycles:pp -vvv |& grep "precise_ip\|exclude_guest"
      precise_ip                       2
      exclude_guest                    1
      precise_ip                       2
      exclude_guest                    1
    switching off exclude_guest, exclude_host
      precise_ip                       2
    ^C

  Also, with :P modifier, evsel->precise_max gets set but exclude_guest does
  not and thus :P also works fine:

    $ sudo ./perf record -C 0 -e cycles:P -vvv |& grep "precise_ip\|exclude_guest"
      precise_ip                       3
    decreasing precise_ip by one (2)
      precise_ip                       2
    ^C

Reported-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20211103072112.32312-1-ravi.bangoria@amd.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoi2c: amd-mp2-plat: ACPI: Use ACPI_COMPANION() directly
Rafael J. Wysocki [Wed, 13 Oct 2021 16:09:39 +0000 (18:09 +0200)]
i2c: amd-mp2-plat: ACPI: Use ACPI_COMPANION() directly

The ACPI_HANDLE() macro is a wrapper arond the ACPI_COMPANION()
macro and the ACPI handle produced by the former comes from the
ACPI device object produced by the latter, so it is way more
straightforward to evaluate the latter directly instead of passing
the handle produced by the former to acpi_bus_get_device().

Modify i2c_amd_probe() accordingly (no intentional functional impact).

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
2 years agoi2c: i801: Add support for Intel Ice Lake PCH-N
Andy Shevchenko [Fri, 1 Oct 2021 17:21:54 +0000 (20:21 +0300)]
i2c: i801: Add support for Intel Ice Lake PCH-N

Add PCI ID of SMBus controller on Intel Ice Lake PCH-N.

The device can be found on MacBookPro16,2 [1].

[1]: https://linux-hardware.org/?probe=f1c5cf0c43

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Jean Delvare <jdelvare@suse.de>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
2 years agoMerge tag '5.16-rc-part1-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 6 Nov 2021 23:47:53 +0000 (16:47 -0700)]
Merge tag '5.16-rc-part1-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs updates from Steve French:

 - reconnect fix for stable

 - minor mount option fix

 - debugging improvement for (TCP) connection issues

 - refactoring of common code to help ksmbd

* tag '5.16-rc-part1-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  smb3: add dynamic trace points for socket connection
  cifs: Move SMB2_Create definitions to the shared area
  cifs: Move more definitions into the shared area
  cifs: move NEGOTIATE_PROTOCOL definitions out into the common area
  cifs: Create a new shared file holding smb2 pdu definitions
  cifs: add mount parameter tcpnodelay
  cifs: To match file servers, make sure the server hostname matches

2 years agoMerge tag 'fsnotify_for_v5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 6 Nov 2021 23:43:20 +0000 (16:43 -0700)]
Merge tag 'fsnotify_for_v5.16-rc1' of git://git./linux/kernel/git/jack/linux-fs

Pull fsnotify updates from Jan Kara:
 "Support for reporting filesystem errors through fanotify so that
  system health monitoring daemons can watch for these and act instead
  of scraping system logs"

* tag 'fsnotify_for_v5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs: (34 commits)
  samples: remove duplicate include in fs-monitor.c
  samples: Fix warning in fsnotify sample
  docs: Fix formatting of literal sections in fanotify docs
  samples: Make fs-monitor depend on libc and headers
  docs: Document the FAN_FS_ERROR event
  samples: Add fs error monitoring example
  ext4: Send notifications on error
  fanotify: Allow users to request FAN_FS_ERROR events
  fanotify: Emit generic error info for error event
  fanotify: Report fid info for file related file system errors
  fanotify: WARN_ON against too large file handles
  fanotify: Add helpers to decide whether to report FID/DFID
  fanotify: Wrap object_fh inline space in a creator macro
  fanotify: Support merging of error events
  fanotify: Support enqueueing of error events
  fanotify: Pre-allocate pool of error events
  fanotify: Reserve UAPI bits for FAN_FS_ERROR
  fsnotify: Support FS_ERROR event type
  fanotify: Require fid_mode for any non-fd event
  fanotify: Encode empty file handle when no inode is provided
  ...

2 years agoMerge tag 'fs_for_v5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jack...
Linus Torvalds [Sat, 6 Nov 2021 23:40:48 +0000 (16:40 -0700)]
Merge tag 'fs_for_v5.16-rc1' of git://git./linux/kernel/git/jack/linux-fs

Pull quota, isofs, and reiserfs updates from Jan Kara:
 "Fixes for handling of corrupted quota files, fix for handling of
  corrupted isofs filesystem, and a small cleanup for reiserfs"

* tag 'fs_for_v5.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  fs: reiserfs: remove useless new_opts in reiserfs_remount
  isofs: Fix out of bound access for corrupted isofs image
  quota: correct error number in free_dqentry()
  quota: check block number when reading the block in quota file

2 years agoMerge tag 'xtensa-20211105' of git://github.com/jcmvbkbc/linux-xtensa
Linus Torvalds [Sat, 6 Nov 2021 21:54:03 +0000 (14:54 -0700)]
Merge tag 'xtensa-20211105' of git://github.com/jcmvbkbc/linux-xtensa

Pull xtensa updates from Max Filippov:

 - add support for xtensa cores without windowed registers option

* tag 'xtensa-20211105' of git://github.com/jcmvbkbc/linux-xtensa:
  xtensa: move section symbols to asm/sections.h
  xtensa: remove unused variable wmask
  xtensa: only build windowed register support code when needed
  xtensa: use register window specific opcodes only when present
  xtensa: implement call0 ABI support in assembly
  xtensa: definitions for call0 ABI
  xtensa: don't use a12 in __xtensa_copy_user in call0 ABI
  xtensa: don't use a12 in strncpy_user
  xtensa: use a14 instead of a15 in inline assembly
  xtensa: move _SimulateUserKernelVectorException out of WindowVectors

2 years agoMerge tag 's390-5.16-1' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Sat, 6 Nov 2021 21:48:06 +0000 (14:48 -0700)]
Merge tag 's390-5.16-1' of git://git./linux/kernel/git/s390/linux

Pull s390 updates from Vasily Gorbik:

 - Add support for ftrace with direct call and ftrace direct call
   samples.

 - Add support for kernel command lines longer than current 896 bytes
   and make its length configurable.

 - Add support for BEAR enhancement facility to improve last breaking
   event instruction tracking.

 - Add kprobes sanity checks and testcases to prevent kprobe in the mid
   of an instruction.

 - Allow concurrent access to /dev/hwc for the CPUMF users.

 - Various ftrace / jump label improvements.

 - Convert unwinder tests to KUnit.

 - Add s390_iommu_aperture kernel parameter to tweak the limits on
   concurrently usable DMA mappings.

 - Add ap.useirq AP module option which can be used to disable interrupt
   use.

 - Add add_disk() error handling support to block device drivers.

 - Drop arch specific and use generic implementation of strlcpy and
   strrchr.

 - Several __pa/__va usages fixes.

 - Various cio, crypto, pci, kernel doc and other small fixes and
   improvements all over the code.

[ Merge fixup as per https://lore.kernel.org/all/YXAqZ%2FEszRisunQw@osiris/ ]

* tag 's390-5.16-1' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: (63 commits)
  s390: make command line configurable
  s390: support command lines longer than 896 bytes
  s390/kexec_file: move kernel image size check
  s390/pci: add s390_iommu_aperture kernel parameter
  s390/spinlock: remove incorrect kernel doc indicator
  s390/string: use generic strlcpy
  s390/string: use generic strrchr
  s390/ap: function rework based on compiler warning
  s390/cio: make ccw_device_dma_* more robust
  s390/vfio-ap: s390/crypto: fix all kernel-doc warnings
  s390/hmcdrv: fix kernel doc comments
  s390/ap: new module option ap.useirq
  s390/cpumf: Allow multiple processes to access /dev/hwc
  s390/bitops: return true/false (not 1/0) from bool functions
  s390: add support for BEAR enhancement facility
  s390: introduce nospec_uses_trampoline()
  s390: rename last_break to pgm_last_break
  s390/ptrace: add last_break member to pt_regs
  s390/sclp: sort out physical vs virtual pointers usage
  s390/setup: convert start and end initrd pointers to virtual
  ...

2 years agoMerge tag 'pci-v5.16-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Sat, 6 Nov 2021 21:36:12 +0000 (14:36 -0700)]
Merge tag 'pci-v5.16-changes' of git://git./linux/kernel/git/helgaas/pci

Pull pci updates from Bjorn Helgaas:
 "Enumeration:
   - Conserve IRQs by setting up portdrv IRQs only when there are users
     (Jan Kiszka)
   - Rework and simplify _OSC negotiation for control of PCIe features
     (Joerg Roedel)
   - Remove struct pci_dev.driver pointer since it's redundant with the
     struct device.driver pointer (Uwe Kleine-König)

  Resource management:
   - Coalesce contiguous host bridge apertures from _CRS to accommodate
     BARs that cover more than one aperture (Kai-Heng Feng)

  Sysfs:
   - Check CAP_SYS_ADMIN before parsing user input (Krzysztof
     WilczyÅ„ski)
   - Return -EINVAL consistently from "store" functions (Krzysztof
     WilczyÅ„ski)
   - Use sysfs_emit() in endpoint "show" functions to avoid buffer
     overruns (Kunihiko Hayashi)

  PCIe native device hotplug:
   - Ignore Link Down/Up caused by resets during error recovery so
     endpoint drivers can remain bound to the device (Lukas Wunner)

  Virtualization:
   - Avoid bus resets on Atheros QCA6174, where they hang the device
     (Ingmar Klein)
   - Work around Pericom PI7C9X2G switch packet drop erratum by using
     store and forward mode instead of cut-through (Nathan Rossi)
   - Avoid trying to enable AtomicOps on VFs; the PF setting applies to
     all VFs (Selvin Xavier)

  MSI:
   - Document that /sys/bus/pci/devices/.../irq contains the legacy INTx
     interrupt or the IRQ of the first MSI (not MSI-X) vector (Barry
     Song)

  VPD:
   - Add pci_read_vpd_any() and pci_write_vpd_any() to access anywhere
     in the possible VPD space; use these to simplify the cxgb3 driver
     (Heiner Kallweit)

  Peer-to-peer DMA:
   - Add (not subtract) the bus offset when calculating DMA address
     (Wang Lu)

  ASPM:
   - Re-enable LTR at Downstream Ports so they don't report Unsupported
     Requests when reset or hot-added devices send LTR messages
     (Mingchuang Qiao)

  Apple PCIe controller driver:
   - Add driver for Apple M1 PCIe controller (Alyssa Rosenzweig, Marc
     Zyngier)

  Cadence PCIe controller driver:
   - Return success when probe succeeds instead of falling into error
     path (Li Chen)

  HiSilicon Kirin PCIe controller driver:
   - Reorganize PHY logic and add support for external PHY drivers
     (Mauro Carvalho Chehab)
   - Support PERST# GPIOs for HiKey970 external PEX 8606 bridge (Mauro
     Carvalho Chehab)
   - Add Kirin 970 support (Mauro Carvalho Chehab)
   - Make driver removable (Mauro Carvalho Chehab)

  Intel VMD host bridge driver:
   - If IOMMU supports interrupt remapping, leave VMD MSI-X remapping
     enabled (Adrian Huang)
   - Number each controller so we can tell them apart in
     /proc/interrupts (Chunguang Xu)
   - Avoid building on UML because VMD depends on x86 bare metal APIs
     (Johannes Berg)

  Marvell Aardvark PCIe controller driver:
   - Define macros for PCI_EXP_DEVCTL_PAYLOAD_* (Pali Rohár)
   - Set Max Payload Size to 512 bytes per Marvell spec (Pali Rohár)
   - Downgrade PIO Response Status messages to debug level (Marek Behún)
   - Preserve CRS SV (Config Request Retry Software Visibility) bit in
     emulated Root Control register (Pali Rohár)
   - Fix issue in configuring reference clock (Pali Rohár)
   - Don't clear status bits for masked interrupts (Pali Rohár)
   - Don't mask unused interrupts (Pali Rohár)
   - Avoid code repetition in advk_pcie_rd_conf() (Marek Behún)
   - Retry config accesses on CRS response (Pali Rohár)
   - Simplify emulated Root Capabilities initialization (Pali Rohár)
   - Fix several link training issues (Pali Rohár)
   - Fix link-up checking via LTSSM (Pali Rohár)
   - Fix reporting of Data Link Layer Link Active (Pali Rohár)
   - Fix emulation of W1C bits (Marek Behún)
   - Fix MSI domain .alloc() method to return zero on success (Marek
     Behún)
   - Read entire 16-bit MSI vector in MSI handler, not just low 8 bits
     (Marek Behún)
   - Clear Root Port I/O Space, Memory Space, and Bus Master Enable bits
     at startup; PCI core will set those as necessary (Pali Rohár)
   - When operating as a Root Port, set class code to "PCI Bridge"
     instead of the default "Mass Storage Controller" (Pali Rohár)
   - Add emulation for PCI_BRIDGE_CTL_BUS_RESET since aardvark doesn't
     implement this per spec (Pali Rohár)
   - Add emulation of option ROM BAR since aardvark doesn't implement
     this per spec (Pali Rohár)

  MediaTek MT7621 PCIe controller driver:
   - Add MediaTek MT7621 PCIe host controller driver and DT binding
     (Sergio Paracuellos)

  Qualcomm PCIe controller driver:
   - Add SC8180x compatible string (Bjorn Andersson)
   - Add endpoint controller driver and DT binding (Manivannan
     Sadhasivam)
   - Restructure to use of_device_get_match_data() (Prasad Malisetty)
   - Add SC7280-specific pcie_1_pipe_clk_src handling (Prasad Malisetty)

  Renesas R-Car PCIe controller driver:
   - Remove unnecessary includes (Geert Uytterhoeven)

  Rockchip DesignWare PCIe controller driver:
   - Add DT binding (Simon Xue)

  Socionext UniPhier Pro5 controller driver:
   - Serialize INTx masking/unmasking (Kunihiko Hayashi)

  Synopsys DesignWare PCIe controller driver:
   - Run dwc .host_init() method before registering MSI interrupt
     handler so we can deal with pending interrupts left by bootloader
     (Bjorn Andersson)
   - Clean up Kconfig dependencies (Andy Shevchenko)
   - Export symbols to allow more modular drivers (Luca Ceresoli)

  TI DRA7xx PCIe controller driver:
   - Allow host and endpoint drivers to be modules (Luca Ceresoli)
   - Enable external clock if present (Luca Ceresoli)

  TI J721E PCIe driver:
   - Disable PHY when probe fails after initializing it (Christophe
     JAILLET)

  MicroSemi Switchtec management driver:
   - Return error to application when command execution fails because an
     out-of-band reset has cleared the device BARs, Memory Space Enable,
     etc (Kelvin Cao)
   - Fix MRPC error status handling issue (Kelvin Cao)
   - Mask out other bits when reading of management VEP instance ID
     (Kelvin Cao)
   - Return EOPNOTSUPP instead of ENOTSUPP from sysfs show functions
     (Kelvin Cao)
   - Add check of event support (Logan Gunthorpe)

  Miscellaneous:
   - Remove unused pci_pool wrappers, which have been replaced by
     dma_pool (Cai Huoqing)
   - Use 'unsigned int' instead of bare 'unsigned' (Krzysztof
     WilczyÅ„ski)
   - Use kstrtobool() directly, sans strtobool() wrapper (Krzysztof
     WilczyÅ„ski)
   - Fix some sscanf(), sprintf() format mismatches (Krzysztof
     WilczyÅ„ski)
   - Update PCI subsystem information in MAINTAINERS (Krzysztof
     WilczyÅ„ski)
   - Correct some misspellings (Krzysztof WilczyÅ„ski)"

* tag 'pci-v5.16-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci: (137 commits)
  PCI: Add ACS quirk for Pericom PI7C9X2G switches
  PCI: apple: Configure RID to SID mapper on device addition
  iommu/dart: Exclude MSI doorbell from PCIe device IOVA range
  PCI: apple: Implement MSI support
  PCI: apple: Add INTx and per-port interrupt support
  PCI: kirin: Allow removing the driver
  PCI: kirin: De-init the dwc driver
  PCI: kirin: Disable clkreq during poweroff sequence
  PCI: kirin: Move the power-off code to a common routine
  PCI: kirin: Add power_off support for Kirin 960 PHY
  PCI: kirin: Allow building it as a module
  PCI: kirin: Add MODULE_* macros
  PCI: kirin: Add Kirin 970 compatible
  PCI: kirin: Support PERST# GPIOs for HiKey970 external PEX 8606 bridge
  PCI: apple: Set up reference clocks when probing
  PCI: apple: Add initial hardware bring-up
  PCI: of: Allow matching of an interrupt-map local to a PCI device
  of/irq: Allow matching of an interrupt-map local to an interrupt controller
  irqdomain: Make of_phandle_args_to_fwspec() generally available
  PCI: Do not enable AtomicOps on VFs
  ...

2 years agoi2c: xgene-slimpro: Fix wrong pointer passed to PTR_ERR()
Wei Yongjun [Mon, 1 Nov 2021 14:02:35 +0000 (14:02 +0000)]
i2c: xgene-slimpro: Fix wrong pointer passed to PTR_ERR()

PTR_ERR should access the value just tested by IS_ERR, otherwise
the wrong error code will be returned.

Fixes: 7b6da7fe7bba ("mailbox: pcc: Use PCC mailbox channel pointer instead of standard")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
2 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sat, 6 Nov 2021 21:08:17 +0000 (14:08 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc updates from Andrew Morton:
 "257 patches.

  Subsystems affected by this patch series: scripts, ocfs2, vfs, and
  mm (slab-generic, slab, slub, kconfig, dax, kasan, debug, pagecache,
  gup, swap, memcg, pagemap, mprotect, mremap, iomap, tracing, vmalloc,
  pagealloc, memory-failure, hugetlb, userfaultfd, vmscan, tools,
  memblock, oom-kill, hugetlbfs, migration, thp, readahead, nommu, ksm,
  vmstat, madvise, memory-hotplug, rmap, zsmalloc, highmem, zram,
  cleanups, kfence, and damon)"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (257 commits)
  mm/damon: remove return value from before_terminate callback
  mm/damon: fix a few spelling mistakes in comments and a pr_debug message
  mm/damon: simplify stop mechanism
  Docs/admin-guide/mm/pagemap: wordsmith page flags descriptions
  Docs/admin-guide/mm/damon/start: simplify the content
  Docs/admin-guide/mm/damon/start: fix a wrong link
  Docs/admin-guide/mm/damon/start: fix wrong example commands
  mm/damon/dbgfs: add adaptive_targets list check before enable monitor_on
  mm/damon: remove unnecessary variable initialization
  Documentation/admin-guide/mm/damon: add a document for DAMON_RECLAIM
  mm/damon: introduce DAMON-based Reclamation (DAMON_RECLAIM)
  selftests/damon: support watermarks
  mm/damon/dbgfs: support watermarks
  mm/damon/schemes: activate schemes based on a watermarks mechanism
  tools/selftests/damon: update for regions prioritization of schemes
  mm/damon/dbgfs: support prioritization weights
  mm/damon/vaddr,paddr: support pageout prioritization
  mm/damon/schemes: prioritize regions within the quotas
  mm/damon/selftests: support schemes quotas
  mm/damon/dbgfs: support quotas of schemes
  ...

2 years agoperf evsel: Fix missing exclude_{host,guest} setting
Namhyung Kim [Fri, 5 Nov 2021 20:58:47 +0000 (13:58 -0700)]
perf evsel: Fix missing exclude_{host,guest} setting

The current logic for the perf missing feature has a bug that it can
wrongly clear some modifiers like G or H.  Actually some PMUs don't
support any filtering or exclusion while others do.  But we check it as
a global feature.

For example, the cycles event can have 'G' modifier to enable it only in
the guest mode on x86.  When you don't run any VMs it'll return 0.

  # perf stat -a -e cycles:G sleep 1

    Performance counter stats for 'system wide':

                    0      cycles:G

          1.000721670 seconds time elapsed

But when it's used with other pmu events that don't support G modifier,
it'll be reset and return non-zero values.

  # perf stat -a -e cycles:G,msr/tsc/ sleep 1

    Performance counter stats for 'system wide':

          538,029,960      cycles:G
       16,924,010,738      msr/tsc/

          1.001815327 seconds time elapsed

This is because of the missing feature detection logic being global.
Add a hashmap to set pmu-specific exclude_host/guest features.

Committer notes:

Fix 'perf test python' by adding a stub for evsel__find_pmu() in
tools/perf/util/python.c, document that it is used so far only for the
above reasons so that if anybody needs this in the python binding
usecases, we can revisit this.

Reported-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Link: http://lore.kernel.org/lkml/20211105205847.120950-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf bpf: Add missing free to bpf_event__print_bpf_prog_info()
Ian Rogers [Sat, 6 Nov 2021 05:37:33 +0000 (22:37 -0700)]
perf bpf: Add missing free to bpf_event__print_bpf_prog_info()

If btf__new() is called then there needs to be a corresponding btf__free().

Fixes: f8dfeae009effc0b ("perf bpf: Show more BPF program info in print_bpf_prog_info()")
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: KP Singh <kpsingh@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tiezhu Yang <yangtiezhu@loongson.cn>
Cc: Yonghong Song <yhs@fb.com>
Cc: bpf@vger.kernel.org
Cc: netdev@vger.kernel.org
Link: http://lore.kernel.org/lkml/20211106053733.3580931-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf beauty: Update copy of linux/socket.h with the kernel sources
Arnaldo Carvalho de Melo [Sat, 19 Jun 2021 13:09:08 +0000 (10:09 -0300)]
perf beauty: Update copy of linux/socket.h with the kernel sources

To pick the changes in:

  99ce45d5e7dbde39 ("mctp: Implement extended addressing")
  55c42fa7fa331f98 ("mptcp: add MPTCP_INFO getsockopt")

That don't result in any changes in the tables generated from that
header.

A table generator for setsockopt is needed, probably will be done in the
5.16 cycle.

This silences this perf build warning:

  Warning: Kernel ABI header at 'tools/perf/trace/beauty/include/linux/socket.h' differs from latest version at 'include/linux/socket.h'
  diff -u tools/perf/trace/beauty/include/linux/socket.h include/linux/socket.h

Cc: David S. Miller <davem@davemloft.net>
Cc: Florian Westphal <fw@strlen.de>
Cc: Jeremy Kerr <jk@codeconstruct.com.au>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agomm/damon: remove return value from before_terminate callback
Changbin Du [Fri, 5 Nov 2021 20:48:27 +0000 (13:48 -0700)]
mm/damon: remove return value from before_terminate callback

Since the return value of 'before_terminate' callback is never used, we
make it have no return value.

Link: https://lkml.kernel.org/r/20211029005023.8895-1-changbin.du@gmail.com
Signed-off-by: Changbin Du <changbin.du@gmail.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: fix a few spelling mistakes in comments and a pr_debug message
Colin Ian King [Fri, 5 Nov 2021 20:48:24 +0000 (13:48 -0700)]
mm/damon: fix a few spelling mistakes in comments and a pr_debug message

There are a few spelling mistakes in the code.  Fix these.

Link: https://lkml.kernel.org/r/20211028184157.614544-1-colin.i.king@gmail.com
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: simplify stop mechanism
Changbin Du [Fri, 5 Nov 2021 20:48:22 +0000 (13:48 -0700)]
mm/damon: simplify stop mechanism

A kernel thread can exit gracefully with kthread_stop().  So we don't
need a new flag 'kdamond_stop'.  And to make sure the task struct is not
freed when accessing it, get reference to it before termination.

Link: https://lkml.kernel.org/r/20211027130517.4404-1-changbin.du@gmail.com
Signed-off-by: Changbin Du <changbin.du@gmail.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs/admin-guide/mm/pagemap: wordsmith page flags descriptions
SeongJae Park [Fri, 5 Nov 2021 20:48:19 +0000 (13:48 -0700)]
Docs/admin-guide/mm/pagemap: wordsmith page flags descriptions

Some descriptions of page flags in 'pagemap.rst' are written in
assumption of none-rst, which respects every new line, as below:

    7 - SLAB
       page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator
       When compound page is used, SLUB/SLQB will only set this flag on the head

Because rst ignores the new line between the first sentence and second
sentence, resulting html looks a little bit weird, as below.

    7 - SLAB
    page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator When
                                                                       ^
    compound page is used, SLUB/SLQB will only set this flag on the head
    page; SLOB will not flag it at all.

This change makes it more natural and consistent with other parts in the
rendered version.

Link: https://lkml.kernel.org/r/20211022090311.3856-5-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs/admin-guide/mm/damon/start: simplify the content
SeongJae Park [Fri, 5 Nov 2021 20:48:16 +0000 (13:48 -0700)]
Docs/admin-guide/mm/damon/start: simplify the content

Information in 'TL; DR' section of 'Getting Started' is duplicated in
other parts of the doc.  It is also asking readers to visit the access
pattern visualizations gallery web site to show the results of example
visualization commands, while the users of the commands can use terminal
output.

To make the doc simple, this removes the duplicated 'TL; DR' section and
replaces the visualization example commands with versions using terminal
outputs.

Link: https://lkml.kernel.org/r/20211022090311.3856-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs/admin-guide/mm/damon/start: fix a wrong link
SeongJae Park [Fri, 5 Nov 2021 20:48:13 +0000 (13:48 -0700)]
Docs/admin-guide/mm/damon/start: fix a wrong link

The 'Getting Started' of DAMON is providing a link to DAMON's user
interface document while saying about its user space tool's detailed
usages.  This fixes the link.

Link: https://lkml.kernel.org/r/20211022090311.3856-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs/admin-guide/mm/damon/start: fix wrong example commands
SeongJae Park [Fri, 5 Nov 2021 20:48:10 +0000 (13:48 -0700)]
Docs/admin-guide/mm/damon/start: fix wrong example commands

Patch series "Fix trivial nits in Documentation/admin-guide/mm".

This patchset fixes trivial nits in admin guide documents for DAMON and
pagemap.

This patch (of 4):

Some of the example commands in DAMON getting started guide are
outdated, missing sudo, or just wrong.  This fixes those.

Link: https://lkml.kernel.org/r/20211022090311.3856-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: add adaptive_targets list check before enable monitor_on
Xin Hao [Fri, 5 Nov 2021 20:48:07 +0000 (13:48 -0700)]
mm/damon/dbgfs: add adaptive_targets list check before enable monitor_on

When the ctx->adaptive_targets list is empty, I did some test on
monitor_on interface like this.

    # cat /sys/kernel/debug/damon/target_ids
    #
    # echo on > /sys/kernel/debug/damon/monitor_on
    # damon: kdamond (5390) starts

Though the ctx->adaptive_targets list is empty, but the kthread_run
still be called, and the kdamond.x thread still be created, this is
meaningless.

So there adds a judgment in 'dbgfs_monitor_on_write', if the
ctx->adaptive_targets list is empty, return -EINVAL.

Link: https://lkml.kernel.org/r/0a60a6e8ec9d71989e0848a4dc3311996ca3b5d4.1634720326.git.xhao@linux.alibaba.com
Signed-off-by: Xin Hao <xhao@linux.alibaba.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: remove unnecessary variable initialization
Xin Hao [Fri, 5 Nov 2021 20:48:04 +0000 (13:48 -0700)]
mm/damon: remove unnecessary variable initialization

Patch series "mm/damon: Fix some small bugs", v4.

This patch (of 2):

In 'damon_va_apply_three_regions' there is no need to set variable 'i'
to zero.

Link: https://lkml.kernel.org/r/b7df8d3dad0943a37e01f60c441b1968b2b20354.1634720326.git.xhao@linux.alibaba.com
Link: https://lkml.kernel.org/r/cover.1634720326.git.xhao@linux.alibaba.com
Signed-off-by: Xin Hao <xhao@linux.alibaba.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocumentation/admin-guide/mm/damon: add a document for DAMON_RECLAIM
SeongJae Park [Fri, 5 Nov 2021 20:48:01 +0000 (13:48 -0700)]
Documentation/admin-guide/mm/damon: add a document for DAMON_RECLAIM

This adds an admin-guide document for DAMON-based Reclamation.

Link: https://lkml.kernel.org/r/20211019150731.16699-16-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: introduce DAMON-based Reclamation (DAMON_RECLAIM)
SeongJae Park [Fri, 5 Nov 2021 20:47:57 +0000 (13:47 -0700)]
mm/damon: introduce DAMON-based Reclamation (DAMON_RECLAIM)

This implements a new kernel subsystem that finds cold memory regions
using DAMON and reclaims those immediately.  It is intended to be used
as proactive lightweigh reclamation logic for light memory pressure.
For heavy memory pressure, it could be inactivated and fall back to the
traditional page-scanning based reclamation.

It's implemented on top of DAMON framework to use the DAMON-based
Operation Schemes (DAMOS) feature.  It utilizes all the DAMOS features
including speed limit, prioritization, and watermarks.

It could be enabled and tuned in boot time via the kernel boot
parameter, and in run time via its module parameters
('/sys/module/damon_reclaim/parameters/') interface.

[yangyingliang@huawei.com: fix error return code in damon_reclaim_turn()]
Link: https://lkml.kernel.org/r/20211025124500.2758060-1-yangyingliang@huawei.com
Link: https://lkml.kernel.org/r/20211019150731.16699-15-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoselftests/damon: support watermarks
SeongJae Park [Fri, 5 Nov 2021 20:47:54 +0000 (13:47 -0700)]
selftests/damon: support watermarks

This updates DAMON selftests for 'schemes' debugfs file to reflect the
changes in the format.

Link: https://lkml.kernel.org/r/20211019150731.16699-14-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: support watermarks
SeongJae Park [Fri, 5 Nov 2021 20:47:50 +0000 (13:47 -0700)]
mm/damon/dbgfs: support watermarks

This updates DAMON debugfs interface to support the watermarks based
schemes activation.  For this, now 'schemes' file receives five more
values.

Link: https://lkml.kernel.org/r/20211019150731.16699-13-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/schemes: activate schemes based on a watermarks mechanism
SeongJae Park [Fri, 5 Nov 2021 20:47:47 +0000 (13:47 -0700)]
mm/damon/schemes: activate schemes based on a watermarks mechanism

DAMON-based operation schemes need to be manually turned on and off.  In
some use cases, however, the condition for turning a scheme on and off
would depend on the system's situation.  For example, schemes for
proactive pages reclamation would need to be turned on when some memory
pressure is detected, and turned off when the system has enough free
memory.

For easier control of schemes activation based on the system situation,
this introduces a watermarks-based mechanism.  The client can describe
the watermark metric (e.g., amount of free memory in the system),
watermark check interval, and three watermarks, namely high, mid, and
low.  If the scheme is deactivated, it only gets the metric and compare
that to the three watermarks for every check interval.  If the metric is
higher than the high watermark, the scheme is deactivated.  If the
metric is between the mid watermark and the low watermark, the scheme is
activated.  If the metric is lower than the low watermark, the scheme is
deactivated again.  This is to allow users fall back to traditional
page-granularity mechanisms.

Link: https://lkml.kernel.org/r/20211019150731.16699-12-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agotools/selftests/damon: update for regions prioritization of schemes
SeongJae Park [Fri, 5 Nov 2021 20:47:44 +0000 (13:47 -0700)]
tools/selftests/damon: update for regions prioritization of schemes

This updates the DAMON selftests for 'schemes' debugfs file, as the file
format is updated.

Link: https://lkml.kernel.org/r/20211019150731.16699-11-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: support prioritization weights
SeongJae Park [Fri, 5 Nov 2021 20:47:40 +0000 (13:47 -0700)]
mm/damon/dbgfs: support prioritization weights

This allows DAMON debugfs interface users set the prioritization weights
by putting three more numbers to the 'schemes' file.

Link: https://lkml.kernel.org/r/20211019150731.16699-10-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/vaddr,paddr: support pageout prioritization
SeongJae Park [Fri, 5 Nov 2021 20:47:37 +0000 (13:47 -0700)]
mm/damon/vaddr,paddr: support pageout prioritization

This makes the default monitoring primitives for virtual address spaces
and the physical address sapce to support memory regions prioritization
for 'PAGEOUT' DAMOS action.  It calculates hotness of each region as
weighted sum of 'nr_accesses' and 'age' of the region and get the
priority score as reverse of the hotness, so that cold regions can be
paged out first.

Link: https://lkml.kernel.org/r/20211019150731.16699-9-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/schemes: prioritize regions within the quotas
SeongJae Park [Fri, 5 Nov 2021 20:47:33 +0000 (13:47 -0700)]
mm/damon/schemes: prioritize regions within the quotas

This makes DAMON apply schemes to regions having higher priority first,
if it cannot apply schemes to all regions due to the quotas.

The prioritization function should be implemented in the monitoring
primitives.  Those would commonly calculate the priority of the region
using attributes of regions, namely 'size', 'nr_accesses', and 'age'.
For example, some primitive would calculate the priority of each region
using a weighted sum of 'nr_accesses' and 'age' of the region.

The optimal weights would depend on give environments, so this makes
those customizable.  Nevertheless, the score calculation functions are
only encouraged to respect the weights, not mandated.

Link: https://lkml.kernel.org/r/20211019150731.16699-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/selftests: support schemes quotas
SeongJae Park [Fri, 5 Nov 2021 20:47:30 +0000 (13:47 -0700)]
mm/damon/selftests: support schemes quotas

This updates DAMON selftests to support updated schemes debugfs file
format for the quotas.

Link: https://lkml.kernel.org/r/20211019150731.16699-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: support quotas of schemes
SeongJae Park [Fri, 5 Nov 2021 20:47:27 +0000 (13:47 -0700)]
mm/damon/dbgfs: support quotas of schemes

This makes the debugfs interface of DAMON support the scheme quotas by
chaning the format of the input for the schemes file.

Link: https://lkml.kernel.org/r/20211019150731.16699-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/schemes: implement time quota
SeongJae Park [Fri, 5 Nov 2021 20:47:23 +0000 (13:47 -0700)]
mm/damon/schemes: implement time quota

The size quota feature of DAMOS is useful for IO resource-critical
systems, but not so intuitive for CPU time-critical systems.  Systems
using zram or zswap-like swap device would be examples.

To provide another intuitive ways for such systems, this implements
time-based quota for DAMON-based Operation Schemes.  If the quota is
set, DAMOS tries to use only up to the user-defined quota of CPU time
within a given time window.

Link: https://lkml.kernel.org/r/20211019150731.16699-5-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/schemes: skip already charged targets and regions
SeongJae Park [Fri, 5 Nov 2021 20:47:20 +0000 (13:47 -0700)]
mm/damon/schemes: skip already charged targets and regions

If DAMOS has stopped applying action in the middle of a group of memory
regions due to its size quota, it starts the work again from the
beginning of the address space in the next charge window.  If there is a
huge memory region at the beginning of the address space and it fulfills
the scheme's target data access pattern always, the action will applied
to only the region.

This mitigates the case by skipping memory regions that charged in
current charge window at the beginning of next charge window.

Link: https://lkml.kernel.org/r/20211019150731.16699-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/schemes: implement size quota for schemes application speed control
SeongJae Park [Fri, 5 Nov 2021 20:47:16 +0000 (13:47 -0700)]
mm/damon/schemes: implement size quota for schemes application speed control

There could be arbitrarily large memory regions fulfilling the target
data access pattern of a DAMON-based operation scheme.  In the case,
applying the action of the scheme could incur too high overhead.  To
provide an intuitive way for avoiding it, this implements a feature
called size quota.  If the quota is set, DAMON tries to apply the action
only up to the given amount of memory regions within a given time
window.

Link: https://lkml.kernel.org/r/20211019150731.16699-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/paddr: support the pageout scheme
SeongJae Park [Fri, 5 Nov 2021 20:47:13 +0000 (13:47 -0700)]
mm/damon/paddr: support the pageout scheme

Introduction
============

This patchset 1) makes the engine for general data access
pattern-oriented memory management (DAMOS) be more useful for production
environments, and 2) implements a static kernel module for lightweight
proactive reclamation using the engine.

Proactive Reclamation
---------------------

On general memory over-committed systems, proactively reclaiming cold
pages helps saving memory and reducing latency spikes that incurred by
the direct reclaim or the CPU consumption of kswapd, while incurring
only minimal performance degradation[2].

A Free Pages Reporting[8] based memory over-commit virtualization system
would be one more specific use case.  In the system, the guest VMs
reports their free memory to host, and the host reallocates the reported
memory to other guests.  As a result, the system's memory utilization
can be maximized.  However, the guests could be not so memory-frugal,
because some kernel subsystems and user-space applications are designed
to use as much memory as available.  Then, guests would report only
small amount of free memory to host, results in poor memory utilization.
Running the proactive reclamation in such guests could help mitigating
this problem.

Google has also implemented this idea and using it in their data center.
They further proposed upstreaming it in LSFMM'19, and "the general
consensus was that, while this sort of proactive reclaim would be useful
for a number of users, the cost of this particular solution was too high
to consider merging it upstream"[3].  The cost mainly comes from the
coldness tracking.  Roughly speaking, the implementation periodically
scans the 'Accessed' bit of each page.  For the reason, the overhead
linearly increases as the size of the memory and the scanning frequency
grows.  As a result, Google is known to dedicating one CPU for the work.
That's a reasonable option to someone like Google, but it wouldn't be so
to some others.

DAMON and DAMOS: An engine for data access pattern-oriented memory management
-----------------------------------------------------------------------------

DAMON[4] is a framework for general data access monitoring.  Its
adaptive monitoring overhead control feature minimizes its monitoring
overhead.  It also let the upper-bound of the overhead be configurable
by clients, regardless of the size of the monitoring target memory.
While monitoring 70 GiB memory of a production system every 5
milliseconds, it consumes less than 1% single CPU time.  For this, it
could sacrify some of the quality of the monitoring results.
Nevertheless, the lower-bound of the quality is configurable, and it
uses a best-effort algorithm for better quality.  Our test results[5]
show the quality is practical enough.  From the production system
monitoring, we were able to find a 4 KiB region in the 70 GiB memory
that shows highest access frequency.

We normally don't monitor the data access pattern just for fun but to
improve something like memory management.  Proactive reclamation is one
such usage.  For such general cases, DAMON provides a feature called
DAMon-based Operation Schemes (DAMOS)[6].  It makes DAMON an engine for
general data access pattern oriented memory management.  Using this,
clients can ask DAMON to find memory regions of specific data access
pattern and apply some memory management action (e.g., page out, move to
head of the LRU list, use huge page, ...).  We call the request
'scheme'.

Proactive Reclamation on top of DAMON/DAMOS
-------------------------------------------

Therefore, by using DAMON for the cold pages detection, the proactive
reclamation's monitoring overhead issue can be solved.  Actually, we
previously implemented a version of proactive reclamation using DAMOS
and achieved noticeable improvements with our evaluation setup[5].
Nevertheless, it more for a proof-of-concept, rather than production
uses.  It supports only virtual address spaces of processes, and require
additional tuning efforts for given workloads and the hardware.  For the
tuning, we introduced a simple auto-tuning user space tool[8].  Google
is also known to using a ML-based similar approach for their fleets[2].
But, making it just works with intuitive knobs in the kernel would be
helpful for general users.

To this end, this patchset improves DAMOS to be ready for such
production usages, and implements another version of the proactive
reclamation, namely DAMON_RECLAIM, on top of it.

DAMOS Improvements: Aggressiveness Control, Prioritization, and Watermarks
--------------------------------------------------------------------------

First of all, the current version of DAMOS supports only virtual address
spaces.  This patchset makes it supports the physical address space for
the page out action.

Next major problem of the current version of DAMOS is the lack of the
aggressiveness control, which can results in arbitrary overhead.  For
example, if huge memory regions having the data access pattern of
interest are found, applying the requested action to all of the regions
could incur significant overhead.  It can be controlled by tuning the
target data access pattern with manual or automated approaches[2,7].
But, some people would prefer the kernel to just work with only
intuitive tuning or default values.

For such cases, this patchset implements a safeguard, namely time/size
quota.  Using this, the clients can specify up to how much time can be
used for applying the action, and/or up to how much memory regions the
action can be applied within a user-specified time duration.  A followup
question is, to which memory regions should the action applied within
the limits? We implement a simple regions prioritization mechanism for
each action and make DAMOS to apply the action to high priority regions
first.  It also allows clients tune the prioritization mechanism to use
different weights for size, access frequency, and age of memory regions.
This means we could use not only LRU but also LFU or some fancy
algorithms like CAR[9] with lightweight overhead.

Though DAMON is lightweight, someone would want to remove even the cold
pages monitoring overhead when it is unnecessary.  Currently, it should
manually turned on and off by clients, but some clients would simply
want to turn it on and off based on some metrics like free memory ratio
or memory fragmentation.  For such cases, this patchset implements a
watermarks-based automatic activation feature.  It allows the clients
configure the metric of their interest, and three watermarks of the
metric.  If the metric is higher than the high watermark or lower than
the low watermark, the scheme is deactivated.  If the metric is lower
than the mid watermark but higher than the low watermark, the scheme is
activated.

DAMON-based Reclaim
-------------------

Using the improved version of DAMOS, this patchset implements a static
kernel module called 'damon_reclaim'.  It finds memory regions that
didn't accessed for specific time duration and page out.  Consuming too
much CPU for the paging out operations, or doing pageout too frequently
can be critical for systems configuring their swap devices with
software-defined in-memory block devices like zram/zswap or total number
of writes limited devices like SSDs, respectively.  To avoid the
problems, the time/size quotas can be configured.  Under the quotas, it
pages out memory regions that didn't accessed longer first.  Also, to
remove the monitoring overhead under peaceful situation, and to fall
back to the LRU-list based page granularity reclamation when it doesn't
make progress, the three watermarks based activation mechanism is used,
with the free memory ratio as the watermark metric.

For convenient configurations, it provides several module parameters.
Using these, sysadmins can enable/disable it, and tune its parameters
including the coldness identification time threshold, the time/size
quotas and the three watermarks.

Evaluation
==========

In short, DAMON_RECLAIM with 50ms/s time quota and regions
prioritization on v5.15-rc5 Linux kernel with ZRAM swap device achieves
38.58% memory saving with only 1.94% runtime overhead.  For this,
DAMON_RECLAIM consumes only 4.97% of single CPU time.

Setup
-----

We evaluate DAMON_RECLAIM to show how each of the DAMOS improvements
make effect.  For this, we measure DAMON_RECLAIM's CPU consumption,
entire system memory footprint, total number of major page faults, and
runtime of 24 realistic workloads in PARSEC3 and SPLASH-2X benchmark
suites on my QEMU/KVM based virtual machine.  The virtual machine runs
on an i3.metal AWS instance, has 130GiB memory, and runs a linux kernel
built on latest -mm tree[1] plus this patchset.  It also utilizes a 4
GiB ZRAM swap device.  We repeats the measurement 5 times and use
averages.

[1] https://github.com/hnaz/linux-mm/tree/v5.15-rc5-mmots-2021-10-13-19-55

Detailed Results
----------------

The results are summarized in the below table.

With coldness identification threshold of 5 seconds, DAMON_RECLAIM
without the time quota-based speed limit achieves 47.21% memory saving,
but incur 4.59% runtime slowdown to the workloads on average.  For this,
DAMON_RECLAIM consumes about 11.28% single CPU time.

Applying time quotas of 200ms/s, 50ms/s, and 10ms/s without the regions
prioritization reduces the slowdown to 4.89%, 2.65%, and 1.5%,
respectively.  Time quota of 200ms/s (20%) makes no real change compared
to the quota unapplied version, because the quota unapplied version
consumes only 11.28% CPU time.  DAMON_RECLAIM's CPU utilization also
similarly reduced: 11.24%, 5.51%, and 2.01% of single CPU time.  That
is, the overhead is proportional to the speed limit.  Nevertheless, it
also reduces the memory saving because it becomes less aggressive.  In
detail, the three variants show 48.76%, 37.83%, and 7.85% memory saving,
respectively.

Applying the regions prioritization (page out regions that not accessed
longer first within the time quota) further reduces the performance
degradation.  Runtime slowdowns and total number of major page faults
increase has been 4.89%/218,690% -> 4.39%/166,136% (200ms/s),
2.65%/111,886% -> 1.94%/59,053% (50ms/s), and 1.5%/34,973.40% ->
2.08%/8,781.75% (10ms/s).  The runtime under 10ms/s time quota has
increased with prioritization, but apparently that's under the margin of
error.

    time quota   prioritization  memory_saving  cpu_util  slowdown  pgmajfaults overhead
    N            N               47.21%         11.28%    4.59%     194,802%
    200ms/s      N               48.76%         11.24%    4.89%     218,690%
    50ms/s       N               37.83%         5.51%     2.65%     111,886%
    10ms/s       N               7.85%          2.01%     1.5%      34,793.40%
    200ms/s      Y               50.08%         10.38%    4.39%     166,136%
    50ms/s       Y               38.58%         4.97%     1.94%     59,053%
    10ms/s       Y               3.63%          1.73%     2.08%     8,781.75%

Baseline and Complete Git Trees
===============================

The patches are based on the latest -mm tree
(v5.15-rc5-mmots-2021-10-13-19-55).  You can also clone the complete git tree
from:

    $ git clone git://github.com/sjp38/linux -b damon_reclaim/patches/v1

The web is also available:
https://git.kernel.org/pub/scm/linux/kernel/git/sj/linux.git/tag/?h=damon_reclaim/patches/v1

Sequence Of Patches
===================

The first patch makes DAMOS support the physical address space for the
page out action.  Following five patches (patches 2-6) implement the
time/size quotas.  Next four patches (patches 7-10) implement the memory
regions prioritization within the limit.  Then, three following patches
(patches 11-13) implement the watermarks-based schemes activation.

Finally, the last two patches (patches 14-15) implement and document the
DAMON-based reclamation using the advanced DAMOS.

[1] https://www.kernel.org/doc/html/v5.15-rc1/vm/damon/index.html
[2] https://research.google/pubs/pub48551/
[3] https://lwn.net/Articles/787611/
[4] https://damonitor.github.io
[5] https://damonitor.github.io/doc/html/latest/vm/damon/eval.html
[6] https://lore.kernel.org/linux-mm/20211001125604.29660-1-sj@kernel.org/
[7] https://github.com/awslabs/damoos
[8] https://www.kernel.org/doc/html/latest/vm/free_page_reporting.html
[9] https://www.usenix.org/conference/fast-04/car-clock-adaptive-replacement

This patch (of 15):

This makes the DAMON primitives for physical address space support the
pageout action for DAMON-based Operation Schemes.  With this commit,
hence, users can easily implement system-level data access-aware
reclamations using DAMOS.

[sj@kernel.org: fix missing-prototype build warning]
Link: https://lkml.kernel.org/r/20211025064220.13904-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20211019150731.16699-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20211019150731.16699-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Marco Elver <elver@google.com>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Greg Thelen <gthelen@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: remove unnecessary variables
Rongwei Wang [Fri, 5 Nov 2021 20:47:09 +0000 (13:47 -0700)]
mm/damon/dbgfs: remove unnecessary variables

In some functions, it's unnecessary to declare 'err' and 'ret' variables
at the same time.  This patch mainly to simplify the issue of such
declarations by reusing one variable.

Link: https://lkml.kernel.org/r/20211014073014.35754-1-sj@kernel.org
Signed-off-by: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/vaddr: constify static mm_walk_ops
Rikard Falkeborn [Fri, 5 Nov 2021 20:47:07 +0000 (13:47 -0700)]
mm/damon/vaddr: constify static mm_walk_ops

The only usage of these structs is to pass their addresses to
walk_page_range(), which takes a pointer to const mm_walk_ops as
argument.  Make them const to allow the compiler to put them in
read-only memory.

Link: https://lkml.kernel.org/r/20211014075042.17174-2-rikard.falkeborn@gmail.com
Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs/DAMON: document physical memory monitoring support
SeongJae Park [Fri, 5 Nov 2021 20:47:03 +0000 (13:47 -0700)]
Docs/DAMON: document physical memory monitoring support

This updates the DAMON documents for the physical memory address space
monitoring support.

Link: https://lkml.kernel.org/r/20211012205711.29216-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: support physical memory monitoring
SeongJae Park [Fri, 5 Nov 2021 20:47:00 +0000 (13:47 -0700)]
mm/damon/dbgfs: support physical memory monitoring

This makes the 'damon-dbgfs' to support the physical memory monitoring,
in addition to the virtual memory monitoring.

Users can do the physical memory monitoring by writing a special
keyword, 'paddr' to the 'target_ids' debugfs file.  Then, DAMON will
check the special keyword and configure the monitoring context to run
with the primitives for the physical address space.

Unlike the virtual memory monitoring, the monitoring target region will
not be automatically set.  Therefore, users should also set the
monitoring target address region using the 'init_regions' debugfs file.

Also, note that the physical memory monitoring will not automatically
terminated.  The user should explicitly turn off the monitoring by
writing 'off' to the 'monitor_on' debugfs file.

Link: https://lkml.kernel.org/r/20211012205711.29216-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: implement primitives for physical address space monitoring
SeongJae Park [Fri, 5 Nov 2021 20:46:56 +0000 (13:46 -0700)]
mm/damon: implement primitives for physical address space monitoring

This implements the monitoring primitives for the physical memory
address space.  Internally, it uses the PTE Accessed bit, similar to
that of the virtual address spaces monitoring primitives.  It supports
only user memory pages, as idle pages tracking does.  If the monitoring
target physical memory address range contains non-user memory pages,
access check of the pages will do nothing but simply treat the pages as
not accessed.

Link: https://lkml.kernel.org/r/20211012205711.29216-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/vaddr: separate commonly usable functions
SeongJae Park [Fri, 5 Nov 2021 20:46:53 +0000 (13:46 -0700)]
mm/damon/vaddr: separate commonly usable functions

This moves functions in the default virtual address spaces monitoring
primitives that commonly usable from other address spaces like physical
address space into a header file.  Those will be reused by the physical
address space monitoring primitives which will be implemented by the
following commit.

[sj@kernel.org: include 'highmem.h' to fix a build failure]
Link: https://lkml.kernel.org/r/20211014110848.5204-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20211012205711.29216-5-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs/admin-guide/mm/damon: document 'init_regions' feature
SeongJae Park [Fri, 5 Nov 2021 20:46:49 +0000 (13:46 -0700)]
Docs/admin-guide/mm/damon: document 'init_regions' feature

This adds description of the 'init_regions' feature in the DAMON usage
document.

Link: https://lkml.kernel.org/r/20211012205711.29216-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs-test: add a unit test case for 'init_regions'
SeongJae Park [Fri, 5 Nov 2021 20:46:46 +0000 (13:46 -0700)]
mm/damon/dbgfs-test: add a unit test case for 'init_regions'

This adds another test case for the new feature, 'init_regions'.

Link: https://lkml.kernel.org/r/20211012205711.29216-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: allow users to set initial monitoring target regions
SeongJae Park [Fri, 5 Nov 2021 20:46:42 +0000 (13:46 -0700)]
mm/damon/dbgfs: allow users to set initial monitoring target regions

Patch series "DAMON: Support Physical Memory Address Space Monitoring:.

DAMON currently supports only virtual address spaces monitoring.  It can
be easily extended for various use cases and address spaces by
configuring its monitoring primitives layer to use appropriate
primitives implementations, though.  This patchset implements monitoring
primitives for the physical address space monitoring using the
structure.

The first 3 patches allow the user space users manually set the
monitoring regions.  The 1st patch implements the feature in the
'damon-dbgfs'.  Then, patches for adding a unit tests (the 2nd patch)
and updating the documentation (the 3rd patch) follow.

Following 4 patches implement the physical address space monitoring
primitives.  The 4th patch makes some primitive functions for the
virtual address spaces primitives reusable.  The 5th patch implements
the physical address space monitoring primitives.  The 6th patch links
the primitives to the 'damon-dbgfs'.  Finally, 7th patch documents this
new features.

This patch (of 7):

Some 'damon-dbgfs' users would want to monitor only a part of the entire
virtual memory address space.  The program interface users in the kernel
space could use '->before_start()' callback or set the regions inside
the context struct as they want, but 'damon-dbgfs' users cannot.

For that reason, this introduces a new debugfs file called
'init_region'.  'damon-dbgfs' users can specify which initial monitoring
target address regions they want by writing special input to the file.
The input should describe each region in each line in the below form:

    <pid> <start address> <end address>

Note that the regions will be updated to cover entire memory mapped
regions after a 'regions update interval' is passed.  If you want the
regions to not be updated after the initial setting, you could set the
interval as a very long time, say, a few decades.

Link: https://lkml.kernel.org/r/20211012205711.29216-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20211012205711.29216-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Marco Elver <elver@google.com>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Greg Thelen <gthelen@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: David Rienjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs/admin-guide/mm/damon: document DAMON-based Operation Schemes
SeongJae Park [Fri, 5 Nov 2021 20:46:39 +0000 (13:46 -0700)]
Docs/admin-guide/mm/damon: document DAMON-based Operation Schemes

This adds the description of DAMON-based operation schemes in the DAMON
documents.

Link: https://lkml.kernel.org/r/20211001125604.29660-8-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoselftests/damon: add 'schemes' debugfs tests
SeongJae Park [Fri, 5 Nov 2021 20:46:36 +0000 (13:46 -0700)]
selftests/damon: add 'schemes' debugfs tests

This adds simple selftets for 'schemes' debugfs file of DAMON.

Link: https://lkml.kernel.org/r/20211001125604.29660-7-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/schemes: implement statistics feature
SeongJae Park [Fri, 5 Nov 2021 20:46:32 +0000 (13:46 -0700)]
mm/damon/schemes: implement statistics feature

To tune the DAMON-based operation schemes, knowing how many and how
large regions are affected by each of the schemes will be helful.  Those
stats could be used for not only the tuning, but also monitoring of the
working set size and the number of regions, if the scheme does not
change the program behavior too much.

For the reason, this implements the statistics for the schemes.  The
total number and size of the regions that each scheme is applied are
exported to users via '->stat_count' and '->stat_sz' of 'struct damos'.
Admins can also check the number by reading 'schemes' debugfs file.  The
last two integers now represents the stats.  To allow collecting the
stats without changing the program behavior, this also adds new scheme
action, 'DAMOS_STAT'.  Note that 'DAMOS_STAT' is not only making no
memory operation actions, but also does not reset the age of regions.

Link: https://lkml.kernel.org/r/20211001125604.29660-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/dbgfs: support DAMON-based Operation Schemes
SeongJae Park [Fri, 5 Nov 2021 20:46:29 +0000 (13:46 -0700)]
mm/damon/dbgfs: support DAMON-based Operation Schemes

This makes 'damon-dbgfs' to support the data access monitoring oriented
memory management schemes.  Users can read and update the schemes using
``<debugfs>/damon/schemes`` file.  The format is::

    <min/max size> <min/max access frequency> <min/max age> <action>

Link: https://lkml.kernel.org/r/20211001125604.29660-5-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/vaddr: support DAMON-based Operation Schemes
SeongJae Park [Fri, 5 Nov 2021 20:46:25 +0000 (13:46 -0700)]
mm/damon/vaddr: support DAMON-based Operation Schemes

This makes DAMON's default primitives for virtual address spaces to
support DAMON-based Operation Schemes (DAMOS) by implementing actions
application functions and registering it to the monitoring context.  The
implementation simply links 'madvise()' for related DAMOS actions.  That
is, 'madvise(MADV_WILLNEED)' is called for 'WILLNEED' DAMOS action and
similar for other actions ('COLD', 'PAGEOUT', 'HUGEPAGE', 'NOHUGEPAGE').

So, the kernel space DAMON users can now use the DAMON-based
optimizations with only small amount of code.

Link: https://lkml.kernel.org/r/20211001125604.29660-4-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/core: implement DAMON-based Operation Schemes (DAMOS)
SeongJae Park [Fri, 5 Nov 2021 20:46:22 +0000 (13:46 -0700)]
mm/damon/core: implement DAMON-based Operation Schemes (DAMOS)

In many cases, users might use DAMON for simple data access aware memory
management optimizations such as applying an operation scheme to a
memory region of a specific size having a specific access frequency for
a specific time.  For example, "page out a memory region larger than 100
MiB but having a low access frequency more than 10 minutes", or "Use THP
for a memory region larger than 2 MiB having a high access frequency for
more than 2 seconds".

Most simple form of the solution would be doing offline data access
pattern profiling using DAMON and modifying the application source code
or system configuration based on the profiling results.  Or, developing
a daemon constructed with two modules (one for access monitoring and the
other for applying memory management actions via mlock(), madvise(),
sysctl, etc) is imaginable.

To avoid users spending their time for implementation of such simple
data access monitoring-based operation schemes, this makes DAMON to
handle such schemes directly.  With this change, users can simply
specify their desired schemes to DAMON.  Then, DAMON will automatically
apply the schemes to the user-specified target processes.

Each of the schemes is composed with conditions for filtering of the
target memory regions and desired memory management action for the
target.  Specifically, the format is::

    <min/max size> <min/max access frequency> <min/max age> <action>

The filtering conditions are size of memory region, number of accesses
to the region monitored by DAMON, and the age of the region.  The age of
region is incremented periodically but reset when its addresses or
access frequency has significantly changed or the action of a scheme was
applied.  For the action, current implementation supports a few of
madvise()-like hints, ``WILLNEED``, ``COLD``, ``PAGEOUT``, ``HUGEPAGE``,
and ``NOHUGEPAGE``.

Because DAMON supports various address spaces and application of the
actions to a monitoring target region is dependent to the type of the
target address space, the application code should be implemented by each
primitives and registered to the framework.  Note that this only
implements the framework part.  Following commit will implement the
action applications for virtual address spaces primitives.

Link: https://lkml.kernel.org/r/20211001125604.29660-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rienjes <rientjes@google.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Marco Elver <elver@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/core: account age of target regions
SeongJae Park [Fri, 5 Nov 2021 20:46:18 +0000 (13:46 -0700)]
mm/damon/core: account age of target regions

Patch series "Implement Data Access Monitoring-based Memory Operation Schemes".

Introduction
============

DAMON[1] can be used as a primitive for data access aware memory
management optimizations.  For that, users who want such optimizations
should run DAMON, read the monitoring results, analyze it, plan a new
memory management scheme, and apply the new scheme by themselves.  Such
efforts will be inevitable for some complicated optimizations.

However, in many other cases, the users would simply want the system to
apply a memory management action to a memory region of a specific size
having a specific access frequency for a specific time.  For example,
"page out a memory region larger than 100 MiB keeping only rare accesses
more than 2 minutes", or "Do not use THP for a memory region larger than
2 MiB rarely accessed for more than 1 seconds".

To make the works easier and non-redundant, this patchset implements a
new feature of DAMON, which is called Data Access Monitoring-based
Operation Schemes (DAMOS).  Using the feature, users can describe the
normal schemes in a simple way and ask DAMON to execute those on its
own.

[1] https://damonitor.github.io

Evaluations
===========

DAMOS is accurate and useful for memory management optimizations.  An
experimental DAMON-based operation scheme for THP, 'ethp', removes
76.15% of THP memory overheads while preserving 51.25% of THP speedup.
Another experimental DAMON-based 'proactive reclamation' implementation,
'prcl', reduces 93.38% of residential sets and 23.63% of system memory
footprint while incurring only 1.22% runtime overhead in the best case
(parsec3/freqmine).

NOTE that the experimental THP optimization and proactive reclamation
are not for production but only for proof of concepts.

Please refer to the showcase web site's evaluation document[1] for
detailed evaluation setup and results.

[1] https://damonitor.github.io/doc/html/v34/vm/damon/eval.html

Long-term Support Trees
-----------------------

For people who want to test DAMON but using LTS kernels, there are
another couple of trees based on two latest LTS kernels respectively and
containing the 'damon/master' backports.

- For v5.4.y: https://git.kernel.org/sj/h/damon/for-v5.4.y
- For v5.10.y: https://git.kernel.org/sj/h/damon/for-v5.10.y

Sequence Of Patches
===================

The 1st patch accounts age of each region.  The 2nd patch implements the
core of the DAMON-based operation schemes feature.  The 3rd patch makes
the default monitoring primitives for virtual address spaces to support
the schemes.  From this point, the kernel space users can use DAMOS.
The 4th patch exports the feature to the user space via the debugfs
interface.  The 5th patch implements schemes statistics feature for
easier tuning of the schemes and runtime access pattern analysis, and
the 6th patch adds selftests for these changes.  Finally, the 7th patch
documents this new feature.

This patch (of 7):

DAMON can be used for data access pattern aware memory management
optimizations.  For that, users should run DAMON, read the monitoring
results, analyze it, plan a new memory management scheme, and apply the
new scheme by themselves.  It would not be too hard, but still require
some level of effort.  For complicated cases, this effort is inevitable.

That said, in many cases, users would simply want to apply an actions to
a memory region of a specific size having a specific access frequency
for a specific time.  For example, "page out a memory region larger than
100 MiB but having a low access frequency more than 10 minutes", or "Use
THP for a memory region larger than 2 MiB having a high access frequency
for more than 2 seconds".

For such optimizations, users will need to first account the age of each
region themselves.  To reduce such efforts, this implements a simple age
account of each region in DAMON.  For each aggregation step, DAMON
compares the access frequency with that from last aggregation and reset
the age of the region if the change is significant.  Else, the age is
incremented.  Also, in case of the merge of regions, the region
size-weighted average of the ages is set as the age of merged new
region.

Link: https://lkml.kernel.org/r/20211001125604.29660-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20211001125604.29660-2-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Amit Shah <amit@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Woodhouse <dwmw@amazon.com>
Cc: Marco Elver <elver@google.com>
Cc: Leonard Foerster <foersleo@amazon.de>
Cc: Greg Thelen <gthelen@google.com>
Cc: Markus Boehme <markubo@amazon.de>
Cc: David Rienjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/core: nullify pointer ctx->kdamond with a NULL
Colin Ian King [Fri, 5 Nov 2021 20:46:15 +0000 (13:46 -0700)]
mm/damon/core: nullify pointer ctx->kdamond with a NULL

Currently a plain integer is being used to nullify the pointer
ctx->kdamond.  Use NULL instead.  Cleans up sparse warning:

  mm/damon/core.c:317:40: warning: Using plain integer as NULL pointer

Link: https://lkml.kernel.org/r/20210925215908.181226-1-colin.king@canonical.com
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: needn't hold kdamond_lock to print pid of kdamond
Changbin Du [Fri, 5 Nov 2021 20:46:12 +0000 (13:46 -0700)]
mm/damon: needn't hold kdamond_lock to print pid of kdamond

Just get the pid by 'current->pid'.  Meanwhile, to be symmetrical make
the 'starts' and 'finishes' logs both use debug level.

Link: https://lkml.kernel.org/r/20210927232432.17750-1-changbin.du@gmail.com
Signed-off-by: Changbin Du <changbin.du@gmail.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: remove unnecessary do_exit() from kdamond
Changbin Du [Fri, 5 Nov 2021 20:46:09 +0000 (13:46 -0700)]
mm/damon: remove unnecessary do_exit() from kdamond

Just return from the kthread function.

Link: https://lkml.kernel.org/r/20210927232421.17694-1-changbin.du@gmail.com
Signed-off-by: Changbin Du <changbin.du@gmail.com>
Cc: SeongJae Park <sjpark@amazon.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon/core: print kdamond start log in debug mode only
SeongJae Park [Fri, 5 Nov 2021 20:46:06 +0000 (13:46 -0700)]
mm/damon/core: print kdamond start log in debug mode only

Logging of kdamond startup is using 'pr_info()' unnecessarily.  This
makes it to use 'pr_debug()' instead.

Link: https://lkml.kernel.org/r/20210917123958.3819-6-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: SeongJae Park <sjpark@amazon.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoinclude/linux/damon.h: fix kernel-doc comments for 'damon_callback'
SeongJae Park [Fri, 5 Nov 2021 20:46:04 +0000 (13:46 -0700)]
include/linux/damon.h: fix kernel-doc comments for 'damon_callback'

A few Kernel-doc comments in 'damon.h' are broken.  This fixes them.

Link: https://lkml.kernel.org/r/20210917123958.3819-5-sj@kernel.org
Signed-off-by: SeongJae Park <sjpark@amazon.de>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agodocs/vm/damon: remove broken reference
SeongJae Park [Fri, 5 Nov 2021 20:46:01 +0000 (13:46 -0700)]
docs/vm/damon: remove broken reference

Building DAMON documents warns for a reference to nonexisting doc, as
below:

    $ time make htmldocs
    [...]
    Documentation/vm/damon/index.rst:24: WARNING: toctree contains reference to nonexisting document 'vm/damon/plans'

This fixes the warning by removing the wrong reference.

Link: https://lkml.kernel.org/r/20210917123958.3819-4-sj@kernel.org
Signed-off-by: SeongJae Park <sjpark@amazon.de>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoMAINTAINERS: update SeongJae's email address
SeongJae Park [Fri, 5 Nov 2021 20:45:58 +0000 (13:45 -0700)]
MAINTAINERS: update SeongJae's email address

This updates SeongJae's email address in MAINTAINERS file to his
preferred one.

Link: https://lkml.kernel.org/r/20210917123958.3819-3-sj@kernel.org
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: SeongJae Park <sjpark@amazon.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocumentation/vm: move user guides to admin-guide/mm/
SeongJae Park [Fri, 5 Nov 2021 20:45:55 +0000 (13:45 -0700)]
Documentation/vm: move user guides to admin-guide/mm/

Most memory management user guide documents are in 'admin-guide/mm/',
but two of those are in 'vm/'.  This moves the two docs into
'admin-guide/mm' for easier documents finding.

Link: https://lkml.kernel.org/r/20210917123958.3819-2-sj@kernel.org
Signed-off-by: SeongJae Park <sjpark@amazon.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm/damon: grammar s/works/work/
Geert Uytterhoeven [Fri, 5 Nov 2021 20:45:52 +0000 (13:45 -0700)]
mm/damon: grammar s/works/work/

Correct a singular versus plural grammar mistake in the help text for
the DAMON_VADDR config symbol.

Link: https://lkml.kernel.org/r/20210914073451.3883834-1-geert@linux-m68k.org
Fixes: 3f49584b262cf8f4 ("mm/damon: implement primitives for the virtual memory address spaces")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: SeongJae Park <sjpark@amazon.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: default to dynamic branch instead of static keys mode
Marco Elver [Fri, 5 Nov 2021 20:45:49 +0000 (13:45 -0700)]
kfence: default to dynamic branch instead of static keys mode

We have observed that on very large machines with newer CPUs, the static
key/branch switching delay is on the order of milliseconds.  This is due
to the required broadcast IPIs, which simply does not scale well to
hundreds of CPUs (cores).  If done too frequently, this can adversely
affect tail latencies of various workloads.

One workaround is to increase the sample interval to several seconds,
while decreasing sampled allocation coverage, but the problem still
exists and could still increase tail latencies.

As already noted in the Kconfig help text, there are trade-offs: at
lower sample intervals the dynamic branch results in better performance;
however, at very large sample intervals, the static keys mode can result
in better performance -- careful benchmarking is recommended.

Our initial benchmarking showed that with large enough sample intervals
and workloads stressing the allocator, the static keys mode was slightly
better.  Evaluating and observing the possible system-wide side-effects
of the static-key-switching induced broadcast IPIs, however, was a blind
spot (in particular on large machines with 100s of cores).

Therefore, a major downside of the static keys mode is, unfortunately,
that it is hard to predict performance on new system architectures and
topologies, but also making conclusions about performance of new
workloads based on a limited set of benchmarks.

Most distributions will simply select the defaults, while targeting a
large variety of different workloads and system architectures.  As such,
the better default is CONFIG_KFENCE_STATIC_KEYS=n, and re-enabling it is
only recommended after careful evaluation.

For reference, on x86-64 the condition in kfence_alloc() generates
exactly
2 instructions in the kmem_cache_alloc() fast-path:

 | ...
 | cmpl   $0x0,0x1a8021c(%rip)  # ffffffff82d560d0 <kfence_allocation_gate>
 | je     ffffffff812d6003      <kmem_cache_alloc+0x243>
 | ...

which, given kfence_allocation_gate is infrequently modified, should be
well predicted by most CPUs.

Link: https://lkml.kernel.org/r/20211019102524.2807208-2-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jann Horn <jannh@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: always use static branches to guard kfence_alloc()
Marco Elver [Fri, 5 Nov 2021 20:45:46 +0000 (13:45 -0700)]
kfence: always use static branches to guard kfence_alloc()

Regardless of KFENCE mode (CONFIG_KFENCE_STATIC_KEYS: either using
static keys to gate allocations, or using a simple dynamic branch),
always use a static branch to avoid the dynamic branch in kfence_alloc()
if KFENCE was disabled at boot.

For CONFIG_KFENCE_STATIC_KEYS=n, this now avoids the dynamic branch if
KFENCE was disabled at boot.

To simplify, also unifies the location where kfence_allocation_gate is
read-checked to just be inline in kfence_alloc().

Link: https://lkml.kernel.org/r/20211019102524.2807208-1-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jann Horn <jannh@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: shorten critical sections of alloc/free
Marco Elver [Fri, 5 Nov 2021 20:45:43 +0000 (13:45 -0700)]
kfence: shorten critical sections of alloc/free

Initializing memory and setting/checking the canary bytes is relatively
expensive, and doing so in the meta->lock critical sections extends the
duration with preemption and interrupts disabled unnecessarily.

Any reads to meta->addr and meta->size in kfence_guarded_alloc() and
kfence_guarded_free() don't require locking meta->lock as long as the
object is removed from the freelist: only kfence_guarded_alloc() sets
meta->addr and meta->size after removing it from the freelist, which
requires a preceding kfence_guarded_free() returning it to the list or
the initial state.

Therefore move reads to meta->addr and meta->size, including expensive
memory initialization using them, out of meta->lock critical sections.

Link: https://lkml.kernel.org/r/20210930153706.2105471-1-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jann Horn <jannh@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: test: use kunit_skip() to skip tests
Marco Elver [Fri, 5 Nov 2021 20:45:40 +0000 (13:45 -0700)]
kfence: test: use kunit_skip() to skip tests

Use the new kunit_skip() to skip tests if requirements were not met.  It
makes it easier to see in KUnit's summary if there were skipped tests.

Link: https://lkml.kernel.org/r/20210922182541.1372400-1-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Taras Madan <tarasmadan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: add note to documentation about skipping covered allocations
Marco Elver [Fri, 5 Nov 2021 20:45:37 +0000 (13:45 -0700)]
kfence: add note to documentation about skipping covered allocations

Add a note briefly mentioning the new policy about "skipping currently
covered allocations if pool close to full." Since this has a notable
impact on KFENCE's bug-detection ability on systems with large uptimes,
it is worth pointing out the feature.

Link: https://lkml.kernel.org/r/20210923104803.2620285-5-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Taras Madan <tarasmadan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: limit currently covered allocations when pool nearly full
Marco Elver [Fri, 5 Nov 2021 20:45:34 +0000 (13:45 -0700)]
kfence: limit currently covered allocations when pool nearly full

One of KFENCE's main design principles is that with increasing uptime,
allocation coverage increases sufficiently to detect previously
undetected bugs.

We have observed that frequent long-lived allocations of the same source
(e.g.  pagecache) tend to permanently fill up the KFENCE pool with
increasing system uptime, thus breaking the above requirement.  The
workaround thus far had been increasing the sample interval and/or
increasing the KFENCE pool size, but is no reliable solution.

To ensure diverse coverage of allocations, limit currently covered
allocations of the same source once pool utilization reaches 75%
(configurable via `kfence.skip_covered_thresh`) or above.  The effect is
retaining reasonable allocation coverage when the pool is close to full.

A side-effect is that this also limits frequent long-lived allocations
of the same source filling up the pool permanently.

Uniqueness of an allocation for coverage purposes is based on its
(partial) allocation stack trace (the source).  A Counting Bloom filter
is used to check if an allocation is covered; if the allocation is
currently covered, the allocation is skipped by KFENCE.

Testing was done using:

(a) a synthetic workload that performs frequent long-lived
    allocations (default config values; sample_interval=1;
    num_objects=63), and

(b) normal desktop workloads on an otherwise idle machine where
    the problem was first reported after a few days of uptime
    (default config values).

In both test cases the sampled allocation rate no longer drops to zero
at any point.  In the case of (b) we observe (after 2 days uptime) 15%
unique allocations in the pool, 77% pool utilization, with 20% "skipped
allocations (covered)".

[elver@google.com: simplify and just use hash_32(), use more random stack_hash_seed]
Link: https://lkml.kernel.org/r/YU3MRGaCaJiYht5g@elver.google.com
[elver@google.com: fix 32 bit]

Link: https://lkml.kernel.org/r/20210923104803.2620285-4-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Taras Madan <tarasmadan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: move saving stack trace of allocations into __kfence_alloc()
Marco Elver [Fri, 5 Nov 2021 20:45:31 +0000 (13:45 -0700)]
kfence: move saving stack trace of allocations into __kfence_alloc()

Move the saving of the stack trace of allocations into __kfence_alloc(),
so that the stack entries array can be used outside of
kfence_guarded_alloc() and we avoid potentially unwinding the stack
multiple times.

Link: https://lkml.kernel.org/r/20210923104803.2620285-3-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Taras Madan <tarasmadan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agokfence: count unexpectedly skipped allocations
Marco Elver [Fri, 5 Nov 2021 20:45:28 +0000 (13:45 -0700)]
kfence: count unexpectedly skipped allocations

Maintain a counter to count allocations that are skipped due to being
incompatible (oversized, incompatible gfp flags) or no capacity.

This is to compute the fraction of allocations that could not be
serviced by KFENCE, which we expect to be rare.

Link: https://lkml.kernel.org/r/20210923104803.2620285-2-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Acked-by: Alexander Potapenko <glider@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Taras Madan <tarasmadan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agostacktrace: move filter_irq_stacks() to kernel/stacktrace.c
Marco Elver [Fri, 5 Nov 2021 20:45:25 +0000 (13:45 -0700)]
stacktrace: move filter_irq_stacks() to kernel/stacktrace.c

filter_irq_stacks() has little to do with the stackdepot implementation,
except that it is usually used by users (such as KASAN) of stackdepot to
reduce the stack trace.

However, filter_irq_stacks() itself is not useful without a stack trace
as obtained by stack_trace_save() and friends.

Therefore, move filter_irq_stacks() to kernel/stacktrace.c, so that new
users of filter_irq_stacks() do not have to start depending on
STACKDEPOT only for filter_irq_stacks().

Link: https://lkml.kernel.org/r/20210923104803.2620285-1-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Taras Madan <tarasmadan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoinclude/linux/mm.h: move nr_free_buffer_pages from swap.h to mm.h
Mianhan Liu [Fri, 5 Nov 2021 20:45:21 +0000 (13:45 -0700)]
include/linux/mm.h: move nr_free_buffer_pages from swap.h to mm.h

nr_free_buffer_pages could be exposed through mm.h instead of swap.h.
The advantage of this change is that it can reduce the obsolete
includes.  For example, net/ipv4/tcp.c wouldn't need swap.h any more
since it has already included mm.h.  Similarly, after checking all the
other files, it comes that tcp.c, udp.c meter.c ,...  follow the same
rule, so these files can have swap.h removed too.

Moreover, after preprocessing all the files that use
nr_free_buffer_pages, it turns out that those files have already
included mm.h.Thus, we can move nr_free_buffer_pages from swap.h to mm.h
safely.  This change will not affect the compilation of other files.

Link: https://lkml.kernel.org/r/20210912133640.1624-1-liumh1@shanghaitech.edu.cn
Signed-off-by: Mianhan Liu <liumh1@shanghaitech.edu.cn>
Cc: Jakub Kicinski <kuba@kernel.org>
CC: Ulf Hansson <ulf.hansson@linaro.org>
Cc: "David S . Miller" <davem@davemloft.net>
Cc: Simon Horman <horms@verge.net.au>
Cc: Pravin B Shelar <pshelar@ovn.org>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>