GitHub/LineageOS/android_kernel_motorola_exynos9610.git
7 years agomd/raid5-cache: in r5l_do_submit_io(), submit io->split_bio first
Song Liu [Tue, 9 May 2017 00:39:24 +0000 (17:39 -0700)]
md/raid5-cache: in r5l_do_submit_io(), submit io->split_bio first

In r5l_do_submit_io(), it is necessary to check io->split_bio before
submit io->current_bio. This is because, endio of current_bio may
free the whole IO unit, and thus change io->split_bio.

Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Shaohua Li <shli@fb.com>
7 years agomd/md0: optimize raid0 discard handling
Shaohua Li [Mon, 8 May 2017 00:36:24 +0000 (17:36 -0700)]
md/md0: optimize raid0 discard handling

There are complaints that raid0 discard handling is slow. Currently we
divide discard request into chunks and dispatch to underlayer disks. The
block layer will do merge to form big requests. This causes a lot of
request split/merge and uses significant CPU time.

A simple idea is to calculate the range for each raid disk for an IO
request and send a discard request to raid disks, which will avoid the
split/merge completely. Previously Coly tried the approach, but the
implementation was too complex because of raid0 zones. This patch always
split bio in zone boundary and handle bio within one zone. It simplifies
the implementation a lot.

Reviewed-by: NeilBrown <neilb@suse.com>
Acked-by: Coly Li <colyli@suse.de>
Signed-off-by: Shaohua Li <shli@fb.com>
7 years agomd: don't return -EAGAIN in md_allow_write for external metadata arrays
Artur Paszkiewicz [Mon, 8 May 2017 09:56:55 +0000 (11:56 +0200)]
md: don't return -EAGAIN in md_allow_write for external metadata arrays

This essentially reverts commit b5470dc5fc18 ("md: resolve external
metadata handling deadlock in md_allow_write") with some adjustments.

Since commit 6791875e2e53 ("md: make reconfig_mutex optional for writes
to md sysfs files.") changing array_state to 'active' does not use
mddev_lock() and will not cause a deadlock with md_allow_write(). This
revert simplifies userspace tools that write to sysfs attributes like
"stripe_cache_size" or "consistency_policy" because it removes the need
for special handling for external metadata arrays, checking for EAGAIN
and retrying the write.

Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
7 years agomd/raid5: make use of spin_lock_irq over local_irq_disable + spin_lock
Julia Cartwright [Fri, 28 Apr 2017 17:41:02 +0000 (12:41 -0500)]
md/raid5: make use of spin_lock_irq over local_irq_disable + spin_lock

On mainline, there is no functional difference, just less code, and
symmetric lock/unlock paths.

On PREEMPT_RT builds, this fixes the following warning, seen by
Alexander GQ Gerasiov, due to the sleeping nature of spinlocks.

   BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:993
   in_atomic(): 0, irqs_disabled(): 1, pid: 58, name: kworker/u12:1
   CPU: 5 PID: 58 Comm: kworker/u12:1 Tainted: G        W       4.9.20-rt16-stand6-686 #1
   Hardware name: Supermicro SYS-5027R-WRF/X9SRW-F, BIOS 3.2a 10/28/2015
   Workqueue: writeback wb_workfn (flush-253:0)
   Call Trace:
    dump_stack+0x47/0x68
    ? migrate_enable+0x4a/0xf0
    ___might_sleep+0x101/0x180
    rt_spin_lock+0x17/0x40
    add_stripe_bio+0x4e3/0x6c0 [raid456]
    ? preempt_count_add+0x42/0xb0
    raid5_make_request+0x737/0xdd0 [raid456]

Reported-by: Alexander GQ Gerasiov <gq@redlab-i.ru>
Tested-by: Alexander GQ Gerasiov <gq@redlab-i.ru>
Signed-off-by: Julia Cartwright <julia@ni.com>
Signed-off-by: Shaohua Li <shli@fb.com>
7 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Thu, 4 May 2017 19:26:43 +0000 (12:26 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) The wireless rate info fix from Johannes Berg.

 2) When a RAW socket is in hdrincl mode, we need to make sure that the
    user provided at least a minimally sized ipv4/ipv6 header. Fix from
    Alexander Potapenko.

 3) We must emit IFLA_PHYS_PORT_NAME netlink attributes using
    nla_put_string() so that it is NULL terminated.

 4) Fix a bug in TCP fastopen handling, wherein child sockets
    erroneously inherit the fastopen_req from the parent, and later can
    end up derefencing freed memory or doing a double free. From Eric
    Dumazet.

 5) Don't clear out netdev stats at close time in tg3 driver, from
    YueHaibing.

 6) Fix refcount leak in xt_CT, from Gao Feng.

 7) In nft_set_bitmap() don't leak dummy elements, from Liping Zhang.

 8) Fix deadlock due to taking the expectation lock twice, also from
    Liping Zhang.

 9) Make xt_socket work again with ipv6, from Peter Tirsek.

10) Don't allow IPV6 to be used with IPVS if ipv6.disable=1, from Paolo
    Abeni.

11) Make the BPF loader more flexible wrt. changes to the bpf MAP entry
    layout. From Jesper Dangaard Brouer.

12) Fix ethtool reported device name in aquantia driver, from Pavel
    Belous.

13) Fix build failures due to the compile time size test not working in
    netfilter conntrack. From Geert Uytterhoeven.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (52 commits)
  cfg80211: make RATE_INFO_BW_20 the default
  ipv6: initialize route null entry in addrconf_init()
  qede: Fix possible misconfiguration of advertised autoneg value.
  qed: Fix overriding of supported autoneg value.
  qed*: Fix possible overflow for status block id field.
  rtnetlink: NUL-terminate IFLA_PHYS_PORT_NAME string
  netvsc: make sure napi enabled before vmbus_open
  aquantia: Fix driver name reported by ethtool
  ipv4, ipv6: ensure raw socket message is big enough to hold an IP header
  net/sched: remove redundant null check on head
  tcp: do not inherit fastopen_req from parent
  forcedeth: remove unnecessary carrier status check
  ibmvnic: Move queue restarting in ibmvnic_tx_complete
  ibmvnic: Record SKB RX queue during poll
  ibmvnic: Continue skb processing after skb completion error
  ibmvnic: Check for driver reset first in ibmvnic_xmit
  ibmvnic: Wait for any pending scrqs entries at driver close
  ibmvnic: Clean up tx pools when closing
  ibmvnic: Whitespace correction in release_rx_pools
  ibmvnic: Delete napi's when releasing driver resources
  ...

7 years agoMerge tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Thu, 4 May 2017 19:19:44 +0000 (12:19 -0700)]
Merge tag 'scsi-misc' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI updates from James Bottomley:
 "This update includes the usual round of major driver updates
  (hisi_sas, ufs, fnic, cxlflash, be2iscsi, ipr, stex). There's also the
  usual amount of cosmetic and spelling stuff"

* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (155 commits)
  scsi: qla4xxx: fix spelling mistake: "Tempalate" -> "Template"
  scsi: stex: make S6flag static
  scsi: mac_esp: fix to pass correct device identity to free_irq()
  scsi: aacraid: pci_alloc_consistent() failures on ARM64
  scsi: ufs: make ufshcd_get_lists_status() register operation obvious
  scsi: ufs: use MASK_EE_STATUS
  scsi: mac_esp: Replace bogus memory barrier with spinlock
  scsi: fcoe: make fcoe_e_d_tov and fcoe_r_a_tov static
  scsi: sd_zbc: Do not write lock zones for reset
  scsi: sd_zbc: Remove superfluous assignments
  scsi: sd: sd_zbc: Rename sd_zbc_setup_write_cmnd
  scsi: Improve scsi_get_sense_info_fld
  scsi: sd: Cleanup sd_done sense data handling
  scsi: sd: Improve sd_completed_bytes
  scsi: sd: Fix function descriptions
  scsi: mpt3sas: remove redundant wmb
  scsi: mpt: Move scsi_remove_host() out of mptscsih_remove_host()
  scsi: sg: reset 'res_in_use' after unlinking reserved array
  scsi: mvumi: remove code handling zero scsi_sg_count(scmd) case
  scsi: fusion: fix spelling mistake: "Persistancy" -> "Persistency"
  ...

7 years agoMerge tag 'gpio-v4.12-1' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Thu, 4 May 2017 19:05:32 +0000 (12:05 -0700)]
Merge tag 'gpio-v4.12-1' of git://git./linux/kernel/git/linusw/linux-gpio

Pull GPIO updates from Linus Walleij:
 "This is the bulk of GPIO changes for the v4.12 kernel cycle.

  Core changes:

   - Return NULL from gpiod_get_optional() when GPIOLIB is disabled.
     This was a much discussed change. It affects use cases where people
     write drivers that might or might not be using GPIO resources. I
     have decided that this is the lesser evil right now.

   - Make gpiod_count() behave consistently across different hardware
     descriptions.

   - Fix the syntax around open drain/open source to not infer active
     high/low semantics.

  New drivers:

   - A new single-register fixed-direction framework driver for hardware
     that have lines controlled by a single register that just work in
     one direction (out or in), including IRQ support.

   - Support the Fintek F71889A GPIO SuperIO controller.

   - Support the National NI 169445 MMIO GPIO.

   - Support for the X-Gene derivative of the DWC GPIO controller

   - Support for the Rohm BD9571MWV-M PMIC GPIO controller.

   - Refactor the Gemini GPIO driver to a generic Faraday FTGPIO driver
     and replace both the Gemini and the Moxa ART custom drivers with
     this driver.

  Driver improvements:

   - A whole slew of drivers have their spinlocks chaned to raw
     spinlocks as they provide irqchips, and thus we are progressing on
     realtime compliance.

   - Use devm_irq_alloc_descs() in a slew of drivers, getting managed
     resources.

   - Support for the embedded PWM controller inside the MVEBU driver.

   - Debounce, open source and open drain support for the Aspeed driver.

   - Misc smaller fixes like spelling and syntax and whatnot"

* tag 'gpio-v4.12-1' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio: (77 commits)
  gpio: f7188x: Add a missing break
  gpio: omap: return error if requested debounce time is not possible
  gpio: Add ROHM BD9571MWV-M PMIC GPIO driver
  gpio: gpio-wcove: fix GPIO IRQ status mask
  gpio: DT bindings, move tca9554 from pcf857x to pca953x
  gpio: move tca9554 from pcf857x to pca953x
  gpio: arizona: Correct check whether the pin is an input
  gpio: Add XRA1403 DTS binding documentation
  dt-bindings: add exar to vendor prefixes list
  gpio: gpio-wcove: fix irq pending status bit width
  gpio: dwapb: use dwapb_read instead of readl_relaxed
  gpio: aspeed: Add open-source and open-drain support
  gpio: aspeed: Add debounce support
  gpio: aspeed: dt: Add optional clocks property
  gpio: aspeed: dt: Fix description alignment in bindings document
  gpio: mvebu: Add limited PWM support
  gpio: Use unsigned int for interrupt numbers
  gpio: f7188x: Add F71889A GPIO support.
  gpio: core: Decouple open drain/source flag with active low/high
  gpio: arizona: Correct handling for reading input GPIOs
  ...

7 years agoMerge tag 'platform-drivers-x86-v4.12-1' of git://git.infradead.org/linux-platform...
Linus Torvalds [Thu, 4 May 2017 18:56:59 +0000 (11:56 -0700)]
Merge tag 'platform-drivers-x86-v4.12-1' of git://git.infradead.org/linux-platform-drivers-x86

Pull x86 platform-drivers update from Darren Hart:
 "This represents a significantly larger and more complex set of changes
  than those of prior merge windows.

  In particular, we had several changes with dependencies on other
  subsystems which we felt were best managed through merges of immutable
  branches, including one each from input, i2c, and leds. Two patches
  for the watchdog subsystem are included after discussion with Wim and
  Guenter following a collision in linux-next (this should be resolved
  and you should only see these two appear in this pull request). These
  are called out in the "External" section below.

  Summary of changes:
   - significant further cleanup of fujitsu-laptop and hp-wmi
   - new model support for ideapad, asus, silead, and xiaomi
   - new hotkeys for thinkpad and models using intel-vbtn
   - dell keyboard backlight improvements
   - build and dependency improvements
   - intel * ipc fixes, cleanups, and api updates
   - single isolated fixes noted below

  External:
   - watchdog: iTCO_wdt: Add PMC specific noreboot update api
   - watchdog: iTCO_wdt: cleanup set/unset no_reboot_bit functions
   - Merge branch 'ib/4.10-sparse-keymap-managed'
   - Merge branch 'i2c/for-INT33FE'
   - Merge branch 'linux-leds/dell-laptop-changes-for-4.12'

  platform/x86:
   - Add Intel Cherry Trail ACPI INT33FE device driver
   - remove sparse_keymap_free() calls
   - Make SILEAD_DMI depend on TOUCHSCREEN_SILEAD

  asus-wmi:
   - try to set als by default
   - fix cpufv sysfs file permission

  acer-wmi:
   - setup accelerometer when ACPI device was found

  ideapad-laptop:
   - Add IdeaPad V310-15ISK to no_hw_rfkill
   - Add IdeaPad 310-15IKB to no_hw_rfkill

  intel_pmc_ipc:
   - use gcr mem base for S0ix counter read
   - Fix iTCO_wdt GCS memory mapping failure
   - Add pmc gcr read/write/update api's
   - fix gcr offset

  dell-laptop:
   - Add keyboard backlight timeout AC settings
   - Handle return error form dell_get_intensity.
   - Protect kbd_state against races
   - Refactor kbd_led_triggers_store()

  hp-wireless:
   - reuse module_acpi_driver
   - add Xiaomi's hardware id to the supported list

  intel-vbtn:
   - add volume up and down

  INT33FE:
   - add i2c dependency

  hp-wmi:
   - Cleanup exit paths
   - Do not shadow errors in sysfs show functions
   - Use DEVICE_ATTR_(RO|RW) helper macros
   - Refactor dock and tablet state fetchers
   - Cleanup wireless get_(hw|sw)state functions
   - Refactor redundant HPWMI_READ functions
   - Standardize enum usage for constants
   - Cleanup local variable declarations
   - Do not shadow error values
   - Fix detection for dock and tablet mode
   - Fix error value for hp_wmi_tablet_state

  fujitsu-laptop:
   - simplify error handling in acpi_fujitsu_laptop_add()
   - do not log LED registration failures
   - switch to managed LED class devices
   - reorganize LED-related code
   - refactor LED registration
   - select LEDS_CLASS
   - remove redundant fields from struct fujitsu_bl
   - account for backlight power when determining brightness
   - do not log set_lcd_level() failures in bl_update_status()
   - ignore errors when setting backlight power
   - make disable_brightness_adjust a boolean
   - clean up use_alt_lcd_levels handling
   - sync brightness in set_lcd_level()
   - simplify set_lcd_level()
   - merge set_lcd_level_alt() into set_lcd_level()
   - switch to a managed backlight device
   - only handle backlight when appropriate
   - update debug message logged by call_fext_func()
   - rename call_fext_func() arguments
   - simplify call_fext_func()
   - clean up local variables in call_fext_func()
   - remove keycode fields from struct fujitsu_bl
   - model-dependent sparse keymap overrides
   - use a sparse keymap for hotkey event generation
   - switch to a managed hotkey input device
   - refactor hotkey input device setup
   - use a sparse keymap for brightness key events
   - switch to a managed backlight input device
   - refactor backlight input device setup
   - remove pf_device field from struct fujitsu_bl
   - only register platform device if FUJ02E3 is present
   - add and remove platform device in separate functions
   - simplify platform device attribute definitions
   - remove backlight-related attributes from the platform device
   - cleanup error labels in fujitsu_init()
   - only register backlight device if FUJ02B1 is present
   - sync backlight power status in acpi_fujitsu_laptop_add()
   - register backlight device in a separate function
   - simplify brightness key event generation logic
   - decrease indentation in acpi_fujitsu_bl_notify()

  intel-hid:
   - Add missing ->thaw callback
   - do not set parents of input devices explicitly
   - remove redundant set_bit() call
   - use devm_input_allocate_device() for HID events input device
   - make intel_hid_set_enable() take a boolean argument
   - simplify enabling/disabling HID events

  silead_dmi:
   - Add touchscreen info for Surftab Wintron 7.0
   - Abort early if DMI does not match
   - Do not treat all devices as i2c_clients
   - Add entry for Insyde 7W tablets
   - Constify properties arrays

  intel_scu_ipc:
   - Introduce intel_scu_ipc_raw_command()
   - Introduce SCU_DEVICE() macro
   - Remove redundant subarch check
   - Rearrange init sequence
   - Platform data is mandatory

  asus-nb-wmi:
   - Add wapf4 quirk for the X302UA

  dell-*:
   - Call new led hw_changed API on kbd brightness change
   - Add a generic dell-laptop notifier chain

  eeepc-laptop:
   - Skip unknown key messages 0x50 0x51

  thinkpad_acpi:
   - add mapping for new hotkeys
   - guard generic hotkey case"

* tag 'platform-drivers-x86-v4.12-1' of git://git.infradead.org/linux-platform-drivers-x86: (108 commits)
  platform/x86: Make SILEAD_DMI depend on TOUCHSCREEN_SILEAD
  platform/x86: asus-wmi: try to set als by default
  platform/x86: asus-wmi: fix cpufv sysfs file permission
  platform/x86: acer-wmi: setup accelerometer when ACPI device was found
  platform/x86: ideapad-laptop: Add IdeaPad V310-15ISK to no_hw_rfkill
  platform/x86: intel_pmc_ipc: use gcr mem base for S0ix counter read
  platform/x86: intel_pmc_ipc: Fix iTCO_wdt GCS memory mapping failure
  watchdog: iTCO_wdt: Add PMC specific noreboot update api
  watchdog: iTCO_wdt: cleanup set/unset no_reboot_bit functions
  platform/x86: intel_pmc_ipc: Add pmc gcr read/write/update api's
  platform/x86: intel_pmc_ipc: fix gcr offset
  platform/x86: dell-laptop: Add keyboard backlight timeout AC settings
  platform/x86: dell-laptop: Handle return error form dell_get_intensity.
  platform/x86: hp-wireless: reuse module_acpi_driver
  platform/x86: intel-vbtn: add volume up and down
  platform/x86: INT33FE: add i2c dependency
  platform/x86: hp-wmi: Cleanup exit paths
  platform/x86: hp-wmi: Do not shadow errors in sysfs show functions
  platform/x86: hp-wmi: Use DEVICE_ATTR_(RO|RW) helper macros
  platform/x86: hp-wmi: Refactor dock and tablet state fetchers
  ...

7 years agoMerge tag 'vfio-v4.12-rc1' of git://github.com/awilliam/linux-vfio
Linus Torvalds [Thu, 4 May 2017 18:53:24 +0000 (11:53 -0700)]
Merge tag 'vfio-v4.12-rc1' of git://github.com/awilliam/linux-vfio

Pull VFIO updates from Alex Williamson:

 - Updates for SPAPR IOMMU backend including compatibility test and
   memory allocation check (Alexey Kardashevskiy)

 - Updates for type1 IOMMU backend to remove asynchronous locked page
   accounting and remove redundancy (Alex Williamson)

* tag 'vfio-v4.12-rc1' of git://github.com/awilliam/linux-vfio:
  vfio/type1: Reduce repetitive calls in vfio_pin_pages_remote()
  vfio/type1: Prune vfio_pin_page_external()
  vfio/type1: Remove locked page accounting workqueue
  vfio/spapr_tce: Check kzalloc() return when preregistering memory
  vfio/powerpc/spapr_tce: Enforce IOMMU type compatibility check

7 years agoMerge tag 'for-linus-4.12b-rc0b-tag' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 4 May 2017 18:37:09 +0000 (11:37 -0700)]
Merge tag 'for-linus-4.12b-rc0b-tag' of git://git./linux/kernel/git/xen/tip

Pull xen updates from Juergen Gross:
 "Xen fixes and featrues for 4.12. The main changes are:

   - enable building the kernel with Xen support but without enabling
     paravirtualized mode (Vitaly Kuznetsov)

   - add a new 9pfs xen frontend driver (Stefano Stabellini)

   - simplify Xen's cpuid handling by making use of cpu capabilities
     (Juergen Gross)

   - add/modify some headers for new Xen paravirtualized devices
     (Oleksandr Andrushchenko)

   - EFI reset_system support under Xen (Julien Grall)

   - and the usual cleanups and corrections"

* tag 'for-linus-4.12b-rc0b-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip: (57 commits)
  xen: Move xen_have_vector_callback definition to enlighten.c
  xen: Implement EFI reset_system callback
  arm/xen: Consolidate calls to shutdown hypercall in a single helper
  xen: Export xen_reboot
  xen/x86: Call xen_smp_intr_init_pv() on BSP
  xen: Revert commits da72ff5bfcb0 and 72a9b186292d
  xen/pvh: Do not fill kernel's e820 map in init_pvh_bootparams()
  xen/scsifront: use offset_in_page() macro
  xen/arm,arm64: rename __generic_dma_ops to xen_get_dma_ops
  xen/arm,arm64: fix xen_dma_ops after 815dd18 "Consolidate get_dma_ops..."
  xen/9pfs: select CONFIG_XEN_XENBUS_FRONTEND
  x86/cpu: remove hypervisor specific set_cpu_features
  vmware: set cpu capabilities during platform initialization
  x86/xen: use capabilities instead of fake cpuid values for xsave
  x86/xen: use capabilities instead of fake cpuid values for x2apic
  x86/xen: use capabilities instead of fake cpuid values for mwait
  x86/xen: use capabilities instead of fake cpuid values for acpi
  x86/xen: use capabilities instead of fake cpuid values for acc
  x86/xen: use capabilities instead of fake cpuid values for mtrr
  x86/xen: use capabilities instead of fake cpuid values for aperf
  ...

7 years agocfg80211: make RATE_INFO_BW_20 the default
Johannes Berg [Thu, 4 May 2017 06:42:30 +0000 (08:42 +0200)]
cfg80211: make RATE_INFO_BW_20 the default

Due to the way I did the RX bitrate conversions in mac80211 with
spatch, going setting flags to setting the value, many drivers now
don't set the bandwidth value for 20 MHz, since with the flags it
wasn't necessary to (there was no 20 MHz flag, only the others.)

Rather than go through and try to fix up all the drivers, instead
renumber the enum so that 20 MHz, which is the typical bandwidth,
actually has the value 0, making those drivers all work again.

If VHT was hit used with a driver not reporting it, e.g. iwlmvm,
this manifested in hitting the bandwidth warning in
cfg80211_calculate_bitrate_vht().

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Tested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoipv6: initialize route null entry in addrconf_init()
WANG Cong [Thu, 4 May 2017 05:07:31 +0000 (22:07 -0700)]
ipv6: initialize route null entry in addrconf_init()

Andrey reported a crash on init_net.ipv6.ip6_null_entry->rt6i_idev
since it is always NULL.

This is clearly wrong, we have code to initialize it to loopback_dev,
unfortunately the order is still not correct.

loopback_dev is registered very early during boot, we lose a chance
to re-initialize it in notifier. addrconf_init() is called after
ip6_route_init(), which means we have no chance to correct it.

Fix it by moving this initialization explicitly after
ipv6_add_dev(init_net.loopback_dev) in addrconf_init().

Reported-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Tested-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'qed-fixes'
David S. Miller [Thu, 4 May 2017 16:31:03 +0000 (12:31 -0400)]
Merge branch 'qed-fixes'

Sudarsana Reddy Kalluru says:

====================
qed*: Bug fix series.

The series contains minor bug fixes for qed/qede drivers.

Please consider applying it to 'net' branch.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqede: Fix possible misconfiguration of advertised autoneg value.
sudarsana.kalluru@cavium.com [Thu, 4 May 2017 15:15:05 +0000 (08:15 -0700)]
qede: Fix possible misconfiguration of advertised autoneg value.

Fail the configuration of advertised speed-autoneg value if the config
update is not supported.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed: Fix overriding of supported autoneg value.
sudarsana.kalluru@cavium.com [Thu, 4 May 2017 15:15:04 +0000 (08:15 -0700)]
qed: Fix overriding of supported autoneg value.

Driver currently uses advertised-autoneg value to populate the
supported-autoneg field. When advertised field is updated, user gets
the same value for supported field. Supported-autoneg value need to be
populated from the link capabilities value returned by the MFW.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed*: Fix possible overflow for status block id field.
sudarsana.kalluru@cavium.com [Thu, 4 May 2017 15:15:03 +0000 (08:15 -0700)]
qed*: Fix possible overflow for status block id field.

Value for status block id could be more than 256 in 100G mode, need to
update its data type from u8 to u16.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agortnetlink: NUL-terminate IFLA_PHYS_PORT_NAME string
Michal Schmidt [Thu, 4 May 2017 14:48:58 +0000 (16:48 +0200)]
rtnetlink: NUL-terminate IFLA_PHYS_PORT_NAME string

IFLA_PHYS_PORT_NAME is a string attribute, so terminate it with \0.
Otherwise libnl3 fails to validate netlink messages with this attribute.
"ip -detail a" assumes too that the attribute is NUL-terminated when
printing it. It often was, due to padding.

I noticed this as libvirtd failing to start on a system with sfc driver
after upgrading it to Linux 4.11, i.e. when sfc added support for
phys_port_name.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonetvsc: make sure napi enabled before vmbus_open
stephen hemminger [Wed, 3 May 2017 23:59:21 +0000 (16:59 -0700)]
netvsc: make sure napi enabled before vmbus_open

This fixes a race where vmbus callback for new packet arriving
could occur before NAPI is initialized.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoaquantia: Fix driver name reported by ethtool
Pavel Belous [Wed, 3 May 2017 18:17:44 +0000 (21:17 +0300)]
aquantia: Fix driver name reported by ethtool

V2: using "aquantia" subsystem tag.

The command "ethtool -i ethX" should display driver name (driver: atlantic)
instead vendor name (driver: aquantia).

Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoipv4, ipv6: ensure raw socket message is big enough to hold an IP header
Alexander Potapenko [Wed, 3 May 2017 15:06:58 +0000 (17:06 +0200)]
ipv4, ipv6: ensure raw socket message is big enough to hold an IP header

raw_send_hdrinc() and rawv6_send_hdrinc() expect that the buffer copied
from the userspace contains the IPv4/IPv6 header, so if too few bytes are
copied, parts of the header may remain uninitialized.

This bug has been detected with KMSAN.

For the record, the KMSAN report:

==================================================================
BUG: KMSAN: use of unitialized memory in nf_ct_frag6_gather+0xf5a/0x44a0
inter: 0
CPU: 0 PID: 1036 Comm: probe Not tainted 4.11.0-rc5+ #2455
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:16
 dump_stack+0x143/0x1b0 lib/dump_stack.c:52
 kmsan_report+0x16b/0x1e0 mm/kmsan/kmsan.c:1078
 __kmsan_warning_32+0x5c/0xa0 mm/kmsan/kmsan_instr.c:510
 nf_ct_frag6_gather+0xf5a/0x44a0 net/ipv6/netfilter/nf_conntrack_reasm.c:577
 ipv6_defrag+0x1d9/0x280 net/ipv6/netfilter/nf_defrag_ipv6_hooks.c:68
 nf_hook_entry_hookfn ./include/linux/netfilter.h:102
 nf_hook_slow+0x13f/0x3c0 net/netfilter/core.c:310
 nf_hook ./include/linux/netfilter.h:212
 NF_HOOK ./include/linux/netfilter.h:255
 rawv6_send_hdrinc net/ipv6/raw.c:673
 rawv6_sendmsg+0x2fcb/0x41a0 net/ipv6/raw.c:919
 inet_sendmsg+0x3f8/0x6d0 net/ipv4/af_inet.c:762
 sock_sendmsg_nosec net/socket.c:633
 sock_sendmsg net/socket.c:643
 SYSC_sendto+0x6a5/0x7c0 net/socket.c:1696
 SyS_sendto+0xbc/0xe0 net/socket.c:1664
 do_syscall_64+0x72/0xa0 arch/x86/entry/common.c:285
 entry_SYSCALL64_slow_path+0x25/0x25 arch/x86/entry/entry_64.S:246
RIP: 0033:0x436e03
RSP: 002b:00007ffce48baf38 EFLAGS: 00000246 ORIG_RAX: 000000000000002c
RAX: ffffffffffffffda RBX: 00000000004002b0 RCX: 0000000000436e03
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000003
RBP: 00007ffce48baf90 R08: 00007ffce48baf50 R09: 000000000000001c
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000401790 R14: 0000000000401820 R15: 0000000000000000
origin: 00000000d9400053
 save_stack_trace+0x16/0x20 arch/x86/kernel/stacktrace.c:59
 kmsan_save_stack_with_flags mm/kmsan/kmsan.c:362
 kmsan_internal_poison_shadow+0xb1/0x1a0 mm/kmsan/kmsan.c:257
 kmsan_poison_shadow+0x6d/0xc0 mm/kmsan/kmsan.c:270
 slab_alloc_node mm/slub.c:2735
 __kmalloc_node_track_caller+0x1f4/0x390 mm/slub.c:4341
 __kmalloc_reserve net/core/skbuff.c:138
 __alloc_skb+0x2cd/0x740 net/core/skbuff.c:231
 alloc_skb ./include/linux/skbuff.h:933
 alloc_skb_with_frags+0x209/0xbc0 net/core/skbuff.c:4678
 sock_alloc_send_pskb+0x9ff/0xe00 net/core/sock.c:1903
 sock_alloc_send_skb+0xe4/0x100 net/core/sock.c:1920
 rawv6_send_hdrinc net/ipv6/raw.c:638
 rawv6_sendmsg+0x2918/0x41a0 net/ipv6/raw.c:919
 inet_sendmsg+0x3f8/0x6d0 net/ipv4/af_inet.c:762
 sock_sendmsg_nosec net/socket.c:633
 sock_sendmsg net/socket.c:643
 SYSC_sendto+0x6a5/0x7c0 net/socket.c:1696
 SyS_sendto+0xbc/0xe0 net/socket.c:1664
 do_syscall_64+0x72/0xa0 arch/x86/entry/common.c:285
 return_from_SYSCALL_64+0x0/0x6a arch/x86/entry/entry_64.S:246
==================================================================

, triggered by the following syscalls:
  socket(PF_INET6, SOCK_RAW, IPPROTO_RAW) = 3
  sendto(3, NULL, 0, 0, {sa_family=AF_INET6, sin6_port=htons(0), inet_pton(AF_INET6, "ff00::", &sin6_addr), sin6_flowinfo=0, sin6_scope_id=0}, 28) = -1 EPERM

A similar report is triggered in net/ipv4/raw.c if we use a PF_INET socket
instead of a PF_INET6 one.

Signed-off-by: Alexander Potapenko <glider@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/sched: remove redundant null check on head
Colin Ian King [Wed, 3 May 2017 13:50:40 +0000 (14:50 +0100)]
net/sched: remove redundant null check on head

head is previously null checked and so the 2nd null check on head
is redundant and therefore can be removed.

Detected by CoverityScan, CID#1399505 ("Logically dead code")

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotcp: do not inherit fastopen_req from parent
Eric Dumazet [Wed, 3 May 2017 13:39:31 +0000 (06:39 -0700)]
tcp: do not inherit fastopen_req from parent

Under fuzzer stress, it is possible that a child gets a non NULL
fastopen_req pointer from its parent at accept() time, when/if parent
morphs from listener to active session.

We need to make sure this can not happen, by clearing the field after
socket cloning.

BUG: Double free or freeing an invalid pointer
Unexpected shadow byte: 0xFB
CPU: 3 PID: 20933 Comm: syz-executor3 Not tainted 4.11.0+ #306
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs
01/01/2011
Call Trace:
 <IRQ>
 __dump_stack lib/dump_stack.c:16 [inline]
 dump_stack+0x292/0x395 lib/dump_stack.c:52
 kasan_object_err+0x1c/0x70 mm/kasan/report.c:164
 kasan_report_double_free+0x5c/0x70 mm/kasan/report.c:185
 kasan_slab_free+0x9d/0xc0 mm/kasan/kasan.c:580
 slab_free_hook mm/slub.c:1357 [inline]
 slab_free_freelist_hook mm/slub.c:1379 [inline]
 slab_free mm/slub.c:2961 [inline]
 kfree+0xe8/0x2b0 mm/slub.c:3882
 tcp_free_fastopen_req net/ipv4/tcp.c:1077 [inline]
 tcp_disconnect+0xc15/0x13e0 net/ipv4/tcp.c:2328
 inet_child_forget+0xb8/0x600 net/ipv4/inet_connection_sock.c:898
 inet_csk_reqsk_queue_add+0x1e7/0x250
net/ipv4/inet_connection_sock.c:928
 tcp_get_cookie_sock+0x21a/0x510 net/ipv4/syncookies.c:217
 cookie_v4_check+0x1a19/0x28b0 net/ipv4/syncookies.c:384
 tcp_v4_cookie_check net/ipv4/tcp_ipv4.c:1384 [inline]
 tcp_v4_do_rcv+0x731/0x940 net/ipv4/tcp_ipv4.c:1421
 tcp_v4_rcv+0x2dc0/0x31c0 net/ipv4/tcp_ipv4.c:1715
 ip_local_deliver_finish+0x4cc/0xc20 net/ipv4/ip_input.c:216
 NF_HOOK include/linux/netfilter.h:257 [inline]
 ip_local_deliver+0x1ce/0x700 net/ipv4/ip_input.c:257
 dst_input include/net/dst.h:492 [inline]
 ip_rcv_finish+0xb1d/0x20b0 net/ipv4/ip_input.c:396
 NF_HOOK include/linux/netfilter.h:257 [inline]
 ip_rcv+0xd8c/0x19c0 net/ipv4/ip_input.c:487
 __netif_receive_skb_core+0x1ad1/0x3400 net/core/dev.c:4210
 __netif_receive_skb+0x2a/0x1a0 net/core/dev.c:4248
 process_backlog+0xe5/0x6c0 net/core/dev.c:4868
 napi_poll net/core/dev.c:5270 [inline]
 net_rx_action+0xe70/0x18e0 net/core/dev.c:5335
 __do_softirq+0x2fb/0xb99 kernel/softirq.c:284
 do_softirq_own_stack+0x1c/0x30 arch/x86/entry/entry_64.S:899
 </IRQ>
 do_softirq.part.17+0x1e8/0x230 kernel/softirq.c:328
 do_softirq kernel/softirq.c:176 [inline]
 __local_bh_enable_ip+0x1cf/0x1e0 kernel/softirq.c:181
 local_bh_enable include/linux/bottom_half.h:31 [inline]
 rcu_read_unlock_bh include/linux/rcupdate.h:931 [inline]
 ip_finish_output2+0x9ab/0x15e0 net/ipv4/ip_output.c:230
 ip_finish_output+0xa35/0xdf0 net/ipv4/ip_output.c:316
 NF_HOOK_COND include/linux/netfilter.h:246 [inline]
 ip_output+0x1f6/0x7b0 net/ipv4/ip_output.c:404
 dst_output include/net/dst.h:486 [inline]
 ip_local_out+0x95/0x160 net/ipv4/ip_output.c:124
 ip_queue_xmit+0x9a8/0x1a10 net/ipv4/ip_output.c:503
 tcp_transmit_skb+0x1ade/0x3470 net/ipv4/tcp_output.c:1057
 tcp_write_xmit+0x79e/0x55b0 net/ipv4/tcp_output.c:2265
 __tcp_push_pending_frames+0xfa/0x3a0 net/ipv4/tcp_output.c:2450
 tcp_push+0x4ee/0x780 net/ipv4/tcp.c:683
 tcp_sendmsg+0x128d/0x39b0 net/ipv4/tcp.c:1342
 inet_sendmsg+0x164/0x5b0 net/ipv4/af_inet.c:762
 sock_sendmsg_nosec net/socket.c:633 [inline]
 sock_sendmsg+0xca/0x110 net/socket.c:643
 SYSC_sendto+0x660/0x810 net/socket.c:1696
 SyS_sendto+0x40/0x50 net/socket.c:1664
 entry_SYSCALL_64_fastpath+0x1f/0xbe
RIP: 0033:0x446059
RSP: 002b:00007faa6761fb58 EFLAGS: 00000282 ORIG_RAX: 000000000000002c
RAX: ffffffffffffffda RBX: 0000000000000017 RCX: 0000000000446059
RDX: 0000000000000001 RSI: 0000000020ba3fcd RDI: 0000000000000017
RBP: 00000000006e40a0 R08: 0000000020ba4ff0 R09: 0000000000000010
R10: 0000000020000000 R11: 0000000000000282 R12: 0000000000708150
R13: 0000000000000000 R14: 00007faa676209c0 R15: 00007faa67620700
Object at ffff88003b5bbcb8, in cache kmalloc-64 size: 64
Allocated:
PID = 20909
 save_stack_trace+0x16/0x20 arch/x86/kernel/stacktrace.c:59
 save_stack+0x43/0xd0 mm/kasan/kasan.c:513
 set_track mm/kasan/kasan.c:525 [inline]
 kasan_kmalloc+0xad/0xe0 mm/kasan/kasan.c:616
 kmem_cache_alloc_trace+0x82/0x270 mm/slub.c:2745
 kmalloc include/linux/slab.h:490 [inline]
 kzalloc include/linux/slab.h:663 [inline]
 tcp_sendmsg_fastopen net/ipv4/tcp.c:1094 [inline]
 tcp_sendmsg+0x221a/0x39b0 net/ipv4/tcp.c:1139
 inet_sendmsg+0x164/0x5b0 net/ipv4/af_inet.c:762
 sock_sendmsg_nosec net/socket.c:633 [inline]
 sock_sendmsg+0xca/0x110 net/socket.c:643
 SYSC_sendto+0x660/0x810 net/socket.c:1696
 SyS_sendto+0x40/0x50 net/socket.c:1664
 entry_SYSCALL_64_fastpath+0x1f/0xbe
Freed:
PID = 20909
 save_stack_trace+0x16/0x20 arch/x86/kernel/stacktrace.c:59
 save_stack+0x43/0xd0 mm/kasan/kasan.c:513
 set_track mm/kasan/kasan.c:525 [inline]
 kasan_slab_free+0x73/0xc0 mm/kasan/kasan.c:589
 slab_free_hook mm/slub.c:1357 [inline]
 slab_free_freelist_hook mm/slub.c:1379 [inline]
 slab_free mm/slub.c:2961 [inline]
 kfree+0xe8/0x2b0 mm/slub.c:3882
 tcp_free_fastopen_req net/ipv4/tcp.c:1077 [inline]
 tcp_disconnect+0xc15/0x13e0 net/ipv4/tcp.c:2328
 __inet_stream_connect+0x20c/0xf90 net/ipv4/af_inet.c:593
 tcp_sendmsg_fastopen net/ipv4/tcp.c:1111 [inline]
 tcp_sendmsg+0x23a8/0x39b0 net/ipv4/tcp.c:1139
 inet_sendmsg+0x164/0x5b0 net/ipv4/af_inet.c:762
 sock_sendmsg_nosec net/socket.c:633 [inline]
 sock_sendmsg+0xca/0x110 net/socket.c:643
 SYSC_sendto+0x660/0x810 net/socket.c:1696
 SyS_sendto+0x40/0x50 net/socket.c:1664
 entry_SYSCALL_64_fastpath+0x1f/0xbe

Fixes: e994b2f0fb92 ("tcp: do not lock listener to process SYN packets")
Fixes: 7db92362d2fe ("tcp: fix potential double free issue for fastopen_req")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Wei Wang <weiwan@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoforcedeth: remove unnecessary carrier status check
Zhu Yanjun [Wed, 3 May 2017 04:43:42 +0000 (00:43 -0400)]
forcedeth: remove unnecessary carrier status check

Since netif_carrier_on() will do nothing if device's
carrier is already on, so it's unnecessary to do
carrier status check.

It's the same for netif_carrier_off().

Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge tag 'modules-for-v4.12' of git://git.kernel.org/pub/scm/linux/kernel/git/jeyu...
Linus Torvalds [Thu, 4 May 2017 02:12:27 +0000 (19:12 -0700)]
Merge tag 'modules-for-v4.12' of git://git./linux/kernel/git/jeyu/linux

Pull modules updates from Jessica Yu:

 - Minor code cleanups

 - Fix section alignment for .init_array

* tag 'modules-for-v4.12' of git://git.kernel.org/pub/scm/linux/kernel/git/jeyu/linux:
  kallsyms: Use bounded strnchr() when parsing string
  module: Unify the return value type of try_module_get
  module: set .init_array alignment to 8

7 years agoMerge tag 'trace-v4.12' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Thu, 4 May 2017 01:41:21 +0000 (18:41 -0700)]
Merge tag 'trace-v4.12' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing updates from Steven Rostedt:
 "New features for this release:

   - Pretty much a full rewrite of the processing of function plugins.
     i.e. echo do_IRQ:stacktrace > set_ftrace_filter

   - The rewrite was needed to add plugins to be unique to tracing
     instances. i.e. mkdir instance/foo; cd instances/foo; echo
     do_IRQ:stacktrace > set_ftrace_filter The old way was written very
     hacky. This removes a lot of those hacks.

   - New "function-fork" tracing option. When set, pids in the
     set_ftrace_pid will have their children added when the processes
     with their pids listed in the set_ftrace_pid file forks.

   - Exposure of "maxactive" for kretprobe in kprobe_events

   - Allow for builtin init functions to be traced by the function
     tracer (via the kernel command line). Module init function tracing
     will come in the next release.

   - Added more selftests, and have selftests also test in an instance"

* tag 'trace-v4.12' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace: (60 commits)
  ring-buffer: Return reader page back into existing ring buffer
  selftests: ftrace: Allow some event trigger tests to run in an instance
  selftests: ftrace: Have some basic tests run in a tracing instance too
  selftests: ftrace: Have event tests also run in an tracing instance
  selftests: ftrace: Make func_event_triggers and func_traceonoff_triggers tests do instances
  selftests: ftrace: Allow some tests to be run in a tracing instance
  tracing/ftrace: Allow for instances to trigger their own stacktrace probes
  tracing/ftrace: Allow for the traceonoff probe be unique to instances
  tracing/ftrace: Enable snapshot function trigger to work with instances
  tracing/ftrace: Allow instances to have their own function probes
  tracing/ftrace: Add a better way to pass data via the probe functions
  ftrace: Dynamically create the probe ftrace_ops for the trace_array
  tracing: Pass the trace_array into ftrace_probe_ops functions
  tracing: Have the trace_array hold the list of registered func probes
  ftrace: If the hash for a probe fails to update then free what was initialized
  ftrace: Have the function probes call their own function
  ftrace: Have each function probe use its own ftrace_ops
  ftrace: Have unregister_ftrace_function_probe_func() return a value
  ftrace: Add helper function ftrace_hash_move_and_update_ops()
  ftrace: Remove data field from ftrace_func_probe structure
  ...

7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/pmladek...
Linus Torvalds [Thu, 4 May 2017 01:29:28 +0000 (18:29 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/pmladek/printk

Pull printk updates from Petr Mladek:

 - There is a situation when early console is not deregistered because
   the preferred one matches a wrong entry. It caused messages to appear
   twice.

   This is the 2nd attempt to fix it. The first one was wrong, see the
   commit c6c7d83b9c9e ('Revert "console: don't prefer first registered
   if DT specifies stdout-path"').

   The fix is coupled with some small code clean up. Well, the console
   registration code would deserve a big one. We need to think about it.

 - Do not lose information about the preemtive context when the console
   semaphore is re-taken.

 - Do not block CPU hotplug when someone else is already pushing
   messages to the console.

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/pmladek/printk:
  printk: fix double printing with earlycon
  printk: rename selected_console -> preferred_console
  printk: fix name/type/scope of preferred_console var
  printk: Correctly handle preemption in console_unlock()
  printk: use console_trylock() in console_cpu_notify()

7 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Thu, 4 May 2017 00:55:59 +0000 (17:55 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc updates from Andrew Morton:

 - a few misc things

 - most of MM

 - KASAN updates

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (102 commits)
  kasan: separate report parts by empty lines
  kasan: improve double-free report format
  kasan: print page description after stacks
  kasan: improve slab object description
  kasan: change report header
  kasan: simplify address description logic
  kasan: change allocation and freeing stack traces headers
  kasan: unify report headers
  kasan: introduce helper functions for determining bug type
  mm: hwpoison: call shake_page() after try_to_unmap() for mlocked page
  mm: hwpoison: call shake_page() unconditionally
  mm/swapfile.c: fix swap space leak in error path of swap_free_entries()
  mm/gup.c: fix access_ok() argument type
  mm/truncate: avoid pointless cleancache_invalidate_inode() calls.
  mm/truncate: bail out early from invalidate_inode_pages2_range() if mapping is empty
  fs/block_dev: always invalidate cleancache in invalidate_bdev()
  fs: fix data invalidation in the cleancache during direct IO
  zram: reduce load operation in page_same_filled
  zram: use zram_free_page instead of open-coded
  zram: introduce zram data accessor
  ...

7 years agokasan: separate report parts by empty lines
Andrey Konovalov [Wed, 3 May 2017 21:56:50 +0000 (14:56 -0700)]
kasan: separate report parts by empty lines

Makes the report easier to read.

Link: http://lkml.kernel.org/r/20170302134851.101218-10-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: improve double-free report format
Andrey Konovalov [Wed, 3 May 2017 21:56:47 +0000 (14:56 -0700)]
kasan: improve double-free report format

Changes double-free report header from

  BUG: Double free or freeing an invalid pointer
  Unexpected shadow byte: 0xFB

to

  BUG: KASAN: double-free or invalid-free in kmalloc_oob_left+0xe5/0xef

This makes a bug uniquely identifiable by the first report line.  To
account for removing of the unexpected shadow value, print shadow bytes
at the end of the report as in reports for other kinds of bugs.

Link: http://lkml.kernel.org/r/20170302134851.101218-9-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: print page description after stacks
Andrey Konovalov [Wed, 3 May 2017 21:56:44 +0000 (14:56 -0700)]
kasan: print page description after stacks

Moves page description after the stacks since it's less important.

Link: http://lkml.kernel.org/r/20170302134851.101218-8-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: improve slab object description
Andrey Konovalov [Wed, 3 May 2017 21:56:41 +0000 (14:56 -0700)]
kasan: improve slab object description

Changes slab object description from:

  Object at ffff880068388540, in cache kmalloc-128 size: 128

to:

  The buggy address belongs to the object at ffff880068388540
   which belongs to the cache kmalloc-128 of size 128
  The buggy address is located 123 bytes inside of
   128-byte region [ffff880068388540ffff8800683885c0)

Makes it more explanatory and adds information about relative offset of
the accessed address to the start of the object.

Link: http://lkml.kernel.org/r/20170302134851.101218-7-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: change report header
Andrey Konovalov [Wed, 3 May 2017 21:56:38 +0000 (14:56 -0700)]
kasan: change report header

Change report header format from:

  BUG: KASAN: use-after-free in unwind_get_return_address+0x28a/0x2c0 at addr ffff880069437950
  Read of size 8 by task insmod/3925

to:

  BUG: KASAN: use-after-free in unwind_get_return_address+0x28a/0x2c0
  Read of size 8 at addr ffff880069437950 by task insmod/3925

The exact access address is not usually important, so move it to the
second line.  This also makes the header look visually balanced.

Link: http://lkml.kernel.org/r/20170302134851.101218-6-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: simplify address description logic
Andrey Konovalov [Wed, 3 May 2017 21:56:34 +0000 (14:56 -0700)]
kasan: simplify address description logic

Simplify logic for describing a memory address.  Add addr_to_page()
helper function.

Makes the code easier to follow.

Link: http://lkml.kernel.org/r/20170302134851.101218-5-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: change allocation and freeing stack traces headers
Andrey Konovalov [Wed, 3 May 2017 21:56:31 +0000 (14:56 -0700)]
kasan: change allocation and freeing stack traces headers

Change stack traces headers from:

  Allocated:
  PID = 42

to:

  Allocated by task 42:

Makes the report one line shorter and look better.

Link: http://lkml.kernel.org/r/20170302134851.101218-4-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: unify report headers
Andrey Konovalov [Wed, 3 May 2017 21:56:28 +0000 (14:56 -0700)]
kasan: unify report headers

Unify KASAN report header format for different kinds of bad memory
accesses.  Makes the code simpler.

Link: http://lkml.kernel.org/r/20170302134851.101218-3-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokasan: introduce helper functions for determining bug type
Andrey Konovalov [Wed, 3 May 2017 21:56:25 +0000 (14:56 -0700)]
kasan: introduce helper functions for determining bug type

Patch series "kasan: improve error reports", v2.

This patchset improves KASAN reports by making them easier to read and a
little more detailed.  Also improves mm/kasan/report.c readability.

Effectively changes a use-after-free report to:

  ==================================================================
  BUG: KASAN: use-after-free in kmalloc_uaf+0xaa/0xb6 [test_kasan]
  Write of size 1 at addr ffff88006aa59da8 by task insmod/3951

  CPU: 1 PID: 3951 Comm: insmod Tainted: G    B           4.10.0+ #84
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
  Call Trace:
   dump_stack+0x292/0x398
   print_address_description+0x73/0x280
   kasan_report.part.2+0x207/0x2f0
   __asan_report_store1_noabort+0x2c/0x30
   kmalloc_uaf+0xaa/0xb6 [test_kasan]
   kmalloc_tests_init+0x4f/0xa48 [test_kasan]
   do_one_initcall+0xf3/0x390
   do_init_module+0x215/0x5d0
   load_module+0x54de/0x82b0
   SYSC_init_module+0x3be/0x430
   SyS_init_module+0x9/0x10
   entry_SYSCALL_64_fastpath+0x1f/0xc2
  RIP: 0033:0x7f22cfd0b9da
  RSP: 002b:00007ffe69118a78 EFLAGS: 00000206 ORIG_RAX: 00000000000000af
  RAX: ffffffffffffffda RBX: 0000555671242090 RCX: 00007f22cfd0b9da
  RDX: 00007f22cffcaf88 RSI: 000000000004df7e RDI: 00007f22d0399000
  RBP: 00007f22cffcaf88 R08: 0000000000000003 R09: 0000000000000000
  R10: 00007f22cfd07d0a R11: 0000000000000206 R12: 0000555671243190
  R13: 000000000001fe81 R14: 0000000000000000 R15: 0000000000000004

  Allocated by task 3951:
   save_stack_trace+0x16/0x20
   save_stack+0x43/0xd0
   kasan_kmalloc+0xad/0xe0
   kmem_cache_alloc_trace+0x82/0x270
   kmalloc_uaf+0x56/0xb6 [test_kasan]
   kmalloc_tests_init+0x4f/0xa48 [test_kasan]
   do_one_initcall+0xf3/0x390
   do_init_module+0x215/0x5d0
   load_module+0x54de/0x82b0
   SYSC_init_module+0x3be/0x430
   SyS_init_module+0x9/0x10
   entry_SYSCALL_64_fastpath+0x1f/0xc2

  Freed by task 3951:
   save_stack_trace+0x16/0x20
   save_stack+0x43/0xd0
   kasan_slab_free+0x72/0xc0
   kfree+0xe8/0x2b0
   kmalloc_uaf+0x85/0xb6 [test_kasan]
   kmalloc_tests_init+0x4f/0xa48 [test_kasan]
   do_one_initcall+0xf3/0x390
   do_init_module+0x215/0x5d0
   load_module+0x54de/0x82b0
   SYSC_init_module+0x3be/0x430
   SyS_init_module+0x9/0x10
   entry_SYSCALL_64_fastpath+0x1f/0xc

  The buggy address belongs to the object at ffff88006aa59da0
   which belongs to the cache kmalloc-16 of size 16
  The buggy address is located 8 bytes inside of
   16-byte region [ffff88006aa59da0ffff88006aa59db0)
  The buggy address belongs to the page:
  page:ffffea0001aa9640 count:1 mapcount:0 mapping:          (null) index:0x0
  flags: 0x100000000000100(slab)
  raw: 0100000000000100 0000000000000000 0000000000000000 0000000180800080
  raw: ffffea0001abe380 0000000700000007 ffff88006c401b40 0000000000000000
  page dumped because: kasan: bad access detected

  Memory state around the buggy address:
   ffff88006aa59c80: 00 00 fc fc 00 00 fc fc 00 00 fc fc 00 00 fc fc
   ffff88006aa59d00: 00 00 fc fc 00 00 fc fc 00 00 fc fc 00 00 fc fc
  >ffff88006aa59d80: fb fb fc fc fb fb fc fc fb fb fc fc fb fb fc fc
                                    ^
   ffff88006aa59e00: fb fb fc fc fb fb fc fc fb fb fc fc fb fb fc fc
   ffff88006aa59e80: fb fb fc fc 00 00 fc fc 00 00 fc fc 00 00 fc fc
  ==================================================================

from:

  ==================================================================
  BUG: KASAN: use-after-free in kmalloc_uaf+0xaa/0xb6 [test_kasan] at addr ffff88006c4dcb28
  Write of size 1 by task insmod/3984
  CPU: 1 PID: 3984 Comm: insmod Tainted: G    B           4.10.0+ #83
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
  Call Trace:
   dump_stack+0x292/0x398
   kasan_object_err+0x1c/0x70
   kasan_report.part.1+0x20e/0x4e0
   __asan_report_store1_noabort+0x2c/0x30
   kmalloc_uaf+0xaa/0xb6 [test_kasan]
   kmalloc_tests_init+0x4f/0xa48 [test_kasan]
   do_one_initcall+0xf3/0x390
   do_init_module+0x215/0x5d0
   load_module+0x54de/0x82b0
   SYSC_init_module+0x3be/0x430
   SyS_init_module+0x9/0x10
   entry_SYSCALL_64_fastpath+0x1f/0xc2
  RIP: 0033:0x7feca0f779da
  RSP: 002b:00007ffdfeae5218 EFLAGS: 00000206 ORIG_RAX: 00000000000000af
  RAX: ffffffffffffffda RBX: 000055a064c13090 RCX: 00007feca0f779da
  RDX: 00007feca1236f88 RSI: 000000000004df7e RDI: 00007feca1605000
  RBP: 00007feca1236f88 R08: 0000000000000003 R09: 0000000000000000
  R10: 00007feca0f73d0a R11: 0000000000000206 R12: 000055a064c14190
  R13: 000000000001fe81 R14: 0000000000000000 R15: 0000000000000004
  Object at ffff88006c4dcb20, in cache kmalloc-16 size: 16
  Allocated:
  PID = 3984
   save_stack_trace+0x16/0x20
   save_stack+0x43/0xd0
   kasan_kmalloc+0xad/0xe0
   kmem_cache_alloc_trace+0x82/0x270
   kmalloc_uaf+0x56/0xb6 [test_kasan]
   kmalloc_tests_init+0x4f/0xa48 [test_kasan]
   do_one_initcall+0xf3/0x390
   do_init_module+0x215/0x5d0
   load_module+0x54de/0x82b0
   SYSC_init_module+0x3be/0x430
   SyS_init_module+0x9/0x10
   entry_SYSCALL_64_fastpath+0x1f/0xc2
  Freed:
  PID = 3984
   save_stack_trace+0x16/0x20
   save_stack+0x43/0xd0
   kasan_slab_free+0x73/0xc0
   kfree+0xe8/0x2b0
   kmalloc_uaf+0x85/0xb6 [test_kasan]
   kmalloc_tests_init+0x4f/0xa48 [test_kasan]
   do_one_initcall+0xf3/0x390
   do_init_module+0x215/0x5d0
   load_module+0x54de/0x82b0
   SYSC_init_module+0x3be/0x430
   SyS_init_module+0x9/0x10
   entry_SYSCALL_64_fastpath+0x1f/0xc2
  Memory state around the buggy address:
   ffff88006c4dca00: fb fb fc fc fb fb fc fc fb fb fc fc fb fb fc fc
   ffff88006c4dca80: fb fb fc fc fb fb fc fc fb fb fc fc fb fb fc fc
  >ffff88006c4dcb00: fb fb fc fc fb fb fc fc fb fb fc fc fb fb fc fc
                                    ^
   ffff88006c4dcb80: fb fb fc fc 00 00 fc fc fb fb fc fc fb fb fc fc
   ffff88006c4dcc00: fb fb fc fc fb fb fc fc fb fb fc fc fb fb fc fc
  ==================================================================

This patch (of 9):

Introduce get_shadow_bug_type() function, which determines bug type
based on the shadow value for a particular kernel address.  Introduce
get_wild_bug_type() function, which determines bug type for addresses
which don't have a corresponding shadow value.

Link: http://lkml.kernel.org/r/20170302134851.101218-2-andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: hwpoison: call shake_page() after try_to_unmap() for mlocked page
Naoya Horiguchi [Wed, 3 May 2017 21:56:22 +0000 (14:56 -0700)]
mm: hwpoison: call shake_page() after try_to_unmap() for mlocked page

Memory error handler calls try_to_unmap() for error pages in various
states.  If the error page is a mlocked page, error handling could fail
with "still referenced by 1 users" message.  This is because the page is
linked to and stays in lru cache after the following call chain.

  try_to_unmap_one
    page_remove_rmap
      clear_page_mlock
        putback_lru_page
          lru_cache_add

memory_failure() calls shake_page() to hanlde the similar issue, but
current code doesn't cover because shake_page() is called only before
try_to_unmap().  So this patches adds shake_page().

Fixes: 23a003bfd23ea9ea0b7756b920e51f64b284b468 ("mm/madvise: pass return code of memory_failure() to userspace")
Link: http://lkml.kernel.org/r/20170417055948.GM31394@yexl-desktop
Link: http://lkml.kernel.org/r/1493197841-23986-3-git-send-email-n-horiguchi@ah.jp.nec.com
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Xiaolong Ye <xiaolong.ye@intel.com>
Cc: Chen Gong <gong.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: hwpoison: call shake_page() unconditionally
Naoya Horiguchi [Wed, 3 May 2017 21:56:19 +0000 (14:56 -0700)]
mm: hwpoison: call shake_page() unconditionally

shake_page() is called before going into core error handling code in
order to ensure that the error page is flushed from lru_cache lists
where pages stay during transferring among LRU lists.

But currently it's not fully functional because when the page is linked
to lru_cache by calling activate_page(), its PageLRU flag is set and
shake_page() is skipped.  The result is to fail error handling with
"still referenced by 1 users" message.

When the page is linked to lru_cache by isolate_lru_page(), its PageLRU
is clear, so that's fine.

This patch makes shake_page() unconditionally called to avoild the
failure.

Fixes: 23a003bfd23ea9ea0b7756b920e51f64b284b468 ("mm/madvise: pass return code of memory_failure() to userspace")
Link: http://lkml.kernel.org/r/20170417055948.GM31394@yexl-desktop
Link: http://lkml.kernel.org/r/1493197841-23986-2-git-send-email-n-horiguchi@ah.jp.nec.com
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Xiaolong Ye <xiaolong.ye@intel.com>
Cc: Chen Gong <gong.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/swapfile.c: fix swap space leak in error path of swap_free_entries()
Huang Ying [Wed, 3 May 2017 21:56:16 +0000 (14:56 -0700)]
mm/swapfile.c: fix swap space leak in error path of swap_free_entries()

In swapcache_free_entries(), if swap_info_get_cont() returns NULL,
something wrong occurs for the swap entry.  But we should still continue
to free the following swap entries in the array instead of skip them to
avoid swap space leak.  This is just problem in error path, where system
may be in an inconsistent state, but it is still good to fix it.

Link: http://lkml.kernel.org/r/20170421124739.24534-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shaohua Li <shli@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/gup.c: fix access_ok() argument type
Arnd Bergmann [Wed, 3 May 2017 21:56:12 +0000 (14:56 -0700)]
mm/gup.c: fix access_ok() argument type

MIPS just got changed to only accept a pointer argument for access_ok(),
causing one warning in drivers/scsi/pmcraid.c.  I tried changing x86 the
same way and found the same warning in __get_user_pages_fast() and
nowhere else in the kernel during randconfig testing:

  mm/gup.c: In function '__get_user_pages_fast':
  mm/gup.c:1578:6: error: passing argument 1 of '__chk_range_not_ok' makes pointer from integer without a cast [-Werror=int-conversion]

It would probably be a good idea to enforce type-safety in general, so
let's change this file to not cause a warning if we do that.

I don't know why the warning did not appear on MIPS.

Fixes: 2667f50e8b81 ("mm: introduce a general RCU get_user_pages_fast()")
Link: http://lkml.kernel.org/r/20170421162659.3314521-1-arnd@arndb.de
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Lorenzo Stoakes <lstoakes@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/truncate: avoid pointless cleancache_invalidate_inode() calls.
Andrey Ryabinin [Wed, 3 May 2017 21:56:09 +0000 (14:56 -0700)]
mm/truncate: avoid pointless cleancache_invalidate_inode() calls.

cleancache_invalidate_inode() called truncate_inode_pages_range() and
invalidate_inode_pages2_range() twice - on entry and on exit.  It's
stupid and waste of time.  It's enough to call it once at exit.

Link: http://lkml.kernel.org/r/20170424164135.22350-5-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Alexey Kuznetsov <kuznet@virtuozzo.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/truncate: bail out early from invalidate_inode_pages2_range() if mapping is empty
Andrey Ryabinin [Wed, 3 May 2017 21:56:06 +0000 (14:56 -0700)]
mm/truncate: bail out early from invalidate_inode_pages2_range() if mapping is empty

If mapping is empty (both ->nrpages and ->nrexceptional is zero) we can
avoid pointless lookups in empty radix tree and bail out immediately
after cleancache invalidation.

Link: http://lkml.kernel.org/r/20170424164135.22350-4-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Alexey Kuznetsov <kuznet@virtuozzo.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agofs/block_dev: always invalidate cleancache in invalidate_bdev()
Andrey Ryabinin [Wed, 3 May 2017 21:56:02 +0000 (14:56 -0700)]
fs/block_dev: always invalidate cleancache in invalidate_bdev()

invalidate_bdev() calls cleancache_invalidate_inode() iff ->nrpages != 0
which doen't make any sense.

Make sure that invalidate_bdev() always calls cleancache_invalidate_inode()
regardless of mapping->nrpages value.

Fixes: c515e1fd361c ("mm/fs: add hooks to support cleancache")
Link: http://lkml.kernel.org/r/20170424164135.22350-3-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Alexey Kuznetsov <kuznet@virtuozzo.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agofs: fix data invalidation in the cleancache during direct IO
Andrey Ryabinin [Wed, 3 May 2017 21:55:59 +0000 (14:55 -0700)]
fs: fix data invalidation in the cleancache during direct IO

Patch series "Properly invalidate data in the cleancache", v2.

We've noticed that after direct IO write, buffered read sometimes gets
stale data which is coming from the cleancache.  The reason for this is
that some direct write hooks call call invalidate_inode_pages2[_range]()
conditionally iff mapping->nrpages is not zero, so we may not invalidate
data in the cleancache.

Another odd thing is that we check only for ->nrpages and don't check
for ->nrexceptional, but invalidate_inode_pages2[_range] also
invalidates exceptional entries as well.  So we invalidate exceptional
entries only if ->nrpages != 0? This doesn't feel right.

 - Patch 1 fixes direct IO writes by removing ->nrpages check.
 - Patch 2 fixes similar case in invalidate_bdev().
     Note: I only fixed conditional cleancache_invalidate_inode() here.
       Do we also need to add ->nrexceptional check in into invalidate_bdev()?

 - Patches 3-4: some optimizations.

This patch (of 4):

Some direct IO write fs hooks call invalidate_inode_pages2[_range]()
conditionally iff mapping->nrpages is not zero.  This can't be right,
because invalidate_inode_pages2[_range]() also invalidate data in the
cleancache via cleancache_invalidate_inode() call.  So if page cache is
empty but there is some data in the cleancache, buffered read after
direct IO write would get stale data from the cleancache.

Also it doesn't feel right to check only for ->nrpages because
invalidate_inode_pages2[_range] invalidates exceptional entries as well.

Fix this by calling invalidate_inode_pages2[_range]() regardless of
nrpages state.

Note: nfs,cifs,9p doesn't need similar fix because the never call
cleancache_get_page() (nor directly, nor via mpage_readpage[s]()), so
they are not affected by this bug.

Fixes: c515e1fd361c ("mm/fs: add hooks to support cleancache")
Link: http://lkml.kernel.org/r/20170424164135.22350-2-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Alexey Kuznetsov <kuznet@virtuozzo.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Nikolay Borisov <n.borisov.lkml@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agozram: reduce load operation in page_same_filled
Sangwoo Park [Wed, 3 May 2017 21:55:56 +0000 (14:55 -0700)]
zram: reduce load operation in page_same_filled

In page_same_filled function, all elements in the page is compared with
next index value.  The current comparison routine compares the (i)th and
(i+1)th values of the page.

In this case, two load operaions occur for each comparison.  But if we
store first value of the page stores at 'val' variable and using it to
compare with others, the load opearation is reduced.  It reduce load
operation per page by up to 64times.

Link: http://lkml.kernel.org/r/1488428104-7257-1-git-send-email-sangwoo2.park@lge.com
Signed-off-by: Sangwoo Park <sangwoo2.park@lge.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agozram: use zram_free_page instead of open-coded
Minchan Kim [Wed, 3 May 2017 21:55:53 +0000 (14:55 -0700)]
zram: use zram_free_page instead of open-coded

The zram_free_page already handles NULL handle case and same page so use
it to reduce error probability.  (Acutaully, I made a mistake when I
handled same page feature)

Link: http://lkml.kernel.org/r/1492052365-16169-7-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agozram: introduce zram data accessor
Minchan Kim [Wed, 3 May 2017 21:55:50 +0000 (14:55 -0700)]
zram: introduce zram data accessor

With element, sometime I got confused handle and element access.  It
might be my bad but I think it's time to introduce accessor to prevent
future idiot like me.  This patch is just clean-up patch so it shouldn't
change any behavior.

Link: http://lkml.kernel.org/r/1492052365-16169-6-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agozram: remove zram_meta structure
Minchan Kim [Wed, 3 May 2017 21:55:47 +0000 (14:55 -0700)]
zram: remove zram_meta structure

It's redundant now.  Instead, remove it and use zram structure directly.

Link: http://lkml.kernel.org/r/1492052365-16169-5-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agozram: use zram_slot_lock instead of raw bit_spin_lock op
Minchan Kim [Wed, 3 May 2017 21:55:44 +0000 (14:55 -0700)]
zram: use zram_slot_lock instead of raw bit_spin_lock op

With this clean-up phase, I want to use zram's wrapper function to lock
table access which is more consistent with other zram's functions.

Link: http://lkml.kernel.org/r/1492052365-16169-4-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agozram: partial IO refactoring
Minchan Kim [Wed, 3 May 2017 21:55:41 +0000 (14:55 -0700)]
zram: partial IO refactoring

For architecture(PAGE_SIZE > 4K), zram have supported partial IO.
However, the mixed code for handling normal/partial IO is too mess,
error-prone to modify IO handler functions with upcoming feature so this
patch aims for cleaning up zram's IO handling functions.

Link: http://lkml.kernel.org/r/1492052365-16169-3-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agozram: handle multiple pages attached bio's bvec
Minchan Kim [Wed, 3 May 2017 21:55:38 +0000 (14:55 -0700)]
zram: handle multiple pages attached bio's bvec

Patch series "zram clean up", v2.

This patchset aims to clean up zram .

[1] clean up multiple pages's bvec handling.
[2] clean up partial IO handling
[3-6] clean up zram via using accessor and removing pointless structure.

With [2-6] applied, we can get a few hundred bytes as well as huge
readibility enhance.

x86: 708 byte save

    add/remove: 1/1 grow/shrink: 0/11 up/down: 478/-1186 (-708)
    function                                     old     new   delta
    zram_special_page_read                         -     478    +478
    zram_reset_device                            317     314      -3
    mem_used_max_store                           131     128      -3
    compact_store                                 96      93      -3
    mm_stat_show                                 203     197      -6
    zram_add                                     719     712      -7
    zram_slot_free_notify                        229     214     -15
    zram_make_request                            819     803     -16
    zram_meta_free                               128     111     -17
    zram_free_page                               180     151     -29
    disksize_store                               432     361     -71
    zram_decompress_page.isra                    504       -    -504
    zram_bvec_rw                                2592    2080    -512
    Total: Before=25350773, After=25350065, chg -0.00%

ppc64: 231 byte save

    add/remove: 2/0 grow/shrink: 1/9 up/down: 681/-912 (-231)
    function                                     old     new   delta
    zram_special_page_read                         -     480    +480
    zram_slot_lock                                 -     200    +200
    vermagic                                      39      40      +1
    mm_stat_show                                 256     248      -8
    zram_meta_free                               200     184     -16
    zram_add                                     944     912     -32
    zram_free_page                               348     308     -40
    disksize_store                               572     492     -80
    zram_decompress_page                         664     564    -100
    zram_slot_free_notify                        292     160    -132
    zram_make_request                           1132    1000    -132
    zram_bvec_rw                                2768    2396    -372
    Total: Before=17565825, After=17565594, chg -0.00%

This patch (of 6):

Johannes Thumshirn reported system goes the panic when using NVMe over
Fabrics loopback target with zram.

The reason is zram expects each bvec in bio contains a single page
but nvme can attach a huge bulk of pages attached to the bio's bvec
so that zram's index arithmetic could be wrong so that out-of-bound
access makes system panic.

[1] in mainline solved solved the problem by limiting max_sectors with
SECTORS_PER_PAGE but it makes zram slow because bio should split with
each pages so this patch makes zram aware of multiple pages in a bvec
so it could solve without any regression(ie, bio split).

[1] 0bc315381fe9, zram: set physical queue limits to avoid array out of
    bounds accesses

Link: http://lkml.kernel.org/r/20170413134057.GA27499@bbox
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reported-by: Johannes Thumshirn <jthumshirn@suse.de>
Tested-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Hannes Reinecke <hare@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, page_alloc: remove debug_guardpage_minorder() test in warn_alloc()
Tetsuo Handa [Wed, 3 May 2017 21:55:34 +0000 (14:55 -0700)]
mm, page_alloc: remove debug_guardpage_minorder() test in warn_alloc()

Commit c0a32fc5a2e4 ("mm: more intensive memory corruption debugging")
changed to check debug_guardpage_minorder() > 0 when reporting
allocation failures.  The reasoning was

  When we use guard page to debug memory corruption, it shrinks
  available pages to 1/2, 1/4, 1/8 and so on, depending on parameter
  value. In such case memory allocation failures can be common and
  printing errors can flood dmesg. If somebody debug corruption,
  allocation failures are not the things he/she is interested about.

but this is misguided.

Allocation requests with __GFP_NOWARN flag by definition do not cause
flooding of allocation failure messages.  Allocation requests with
__GFP_NORETRY flag likely also have __GFP_NOWARN flag.  Costly
allocation requests likely also have __GFP_NOWARN flag.

Allocation requests without __GFP_DIRECT_RECLAIM flag likely also have
__GFP_NOWARN flag or __GFP_HIGH flag.  Non-costly allocation requests
with __GFP_DIRECT_RECLAIM flag basically retry forever due to the "too
small to fail" memory-allocation rule.

Therefore, as a whole, shrinking available pages by
debug_guardpage_minorder= kernel boot parameter might cause flooding of
OOM killer messages but unlikely causes flooding of allocation failure
messages.  Let's remove debug_guardpage_minorder() > 0 check which would
likely be pointless.

Link: http://lkml.kernel.org/r/1491910035-4231-1-git-send-email-penguin-kernel@I-love.SAKURA.ne.jp
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Stanislaw Gruszka <sgruszka@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Rafael J . Wysocki" <rafael.j.wysocki@intel.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/memory-failure.c: add page flag description in error paths
Anshuman Khandual [Wed, 3 May 2017 21:55:31 +0000 (14:55 -0700)]
mm/memory-failure.c: add page flag description in error paths

It helps to provide page flag description along with the raw value in
error paths during soft offline process.  From sample experiments

Before the patch:

  soft offline: 0x6100: migration failed 1, type 3ffff800008018
  soft offline: 0x7400: migration failed 1, type 3ffff800008018

After the patch:

  soft offline: 0x5900: migration failed 1, type 3ffff800008018 (uptodate|dirty|head)
  soft offline: 0x6c00: migration failed 1, type 3ffff800008018 (uptodate|dirty|head)

Link: http://lkml.kernel.org/r/20170409023829.10788-1-khandual@linux.vnet.ibm.com
Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Acked-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/madvise: move up the behavior parameter validation
Anshuman Khandual [Wed, 3 May 2017 21:55:28 +0000 (14:55 -0700)]
mm/madvise: move up the behavior parameter validation

madvise_behavior_valid() should be called before acting upon the
behavior parameter.  Hence move up the function.  This also includes
MADV_SOFT_OFFLINE and MADV_HWPOISON options as valid behavior parameter
for the system call madvise().

Link: http://lkml.kernel.org/r/20170418052844.24891-1-khandual@linux.vnet.ibm.com
Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/madvise.c: clean up MADV_SOFT_OFFLINE and MADV_HWPOISON
Anshuman Khandual [Wed, 3 May 2017 21:55:25 +0000 (14:55 -0700)]
mm/madvise.c: clean up MADV_SOFT_OFFLINE and MADV_HWPOISON

This cleans up handling MADV_SOFT_OFFLINE and MADV_HWPOISON called
through madvise() system call.

* madvise_memory_failure() was misleading to accommodate handling of
  both memory_failure() as well as soft_offline_page() functions.
  Basically it handles memory error injection from user space which
  can go either way as memory failure or soft offline. Renamed as
  madvise_inject_error() instead.

* Renamed struct page pointer 'p' to 'page'.

* pr_info() was essentially printing PFN value but it said 'page'
  which was misleading. Made the process virtual address explicit.

Before the patch:

Soft offlining page 0x15e3e at 0x3fff8c230000
Soft offlining page 0x1f3 at 0x3fffa0da0000
Soft offlining page 0x744 at 0x3fff7d200000
Soft offlining page 0x1634d at 0x3fff95e20000
Soft offlining page 0x16349 at 0x3fff95e30000
Soft offlining page 0x1d6 at 0x3fff9e8b0000
Soft offlining page 0x5f3 at 0x3fff91bd0000

Injecting memory failure for page 0x15c8b at 0x3fff83280000
Injecting memory failure for page 0x16190 at 0x3fff83290000
Injecting memory failure for page 0x740 at 0x3fff9a2e0000
Injecting memory failure for page 0x741 at 0x3fff9a2f0000

After the patch:

Soft offlining pfn 0x1484e at process virtual address 0x3fff883c0000
Soft offlining pfn 0x1484f at process virtual address 0x3fff883d0000
Soft offlining pfn 0x14850 at process virtual address 0x3fff883e0000
Soft offlining pfn 0x14851 at process virtual address 0x3fff883f0000
Soft offlining pfn 0x14852 at process virtual address 0x3fff88400000
Soft offlining pfn 0x14853 at process virtual address 0x3fff88410000
Soft offlining pfn 0x14854 at process virtual address 0x3fff88420000
Soft offlining pfn 0x1521c at process virtual address 0x3fff6bc70000

Injecting memory failure for pfn 0x10fcf at process virtual address 0x3fff86310000
Injecting memory failure for pfn 0x10fd0 at process virtual address 0x3fff86320000
Injecting memory failure for pfn 0x10fd1 at process virtual address 0x3fff86330000
Injecting memory failure for pfn 0x10fd2 at process virtual address 0x3fff86340000
Injecting memory failure for pfn 0x10fd3 at process virtual address 0x3fff86350000
Injecting memory failure for pfn 0x10fd4 at process virtual address 0x3fff86360000
Injecting memory failure for pfn 0x10fd5 at process virtual address 0x3fff86370000

Link: http://lkml.kernel.org/r/20170410084701.11248-1-khandual@linux.vnet.ibm.com
Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoDocumentation: vm, add hugetlbfs reservation overview
Mike Kravetz [Wed, 3 May 2017 21:55:22 +0000 (14:55 -0700)]
Documentation: vm, add hugetlbfs reservation overview

Adding a brief overview of hugetlbfs reservation design and
implementation as an aid to those making code modifications in this
area.

Link: http://lkml.kernel.org/r/1491586995-13085-1-git-send-email-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, swap: remove unused function prototype
Huang Ying [Wed, 3 May 2017 21:55:19 +0000 (14:55 -0700)]
mm, swap: remove unused function prototype

This is a code cleanup patch, no functionality changes.  There are 2
unused function prototype in swap.h, they are removed.

Link: http://lkml.kernel.org/r/20170405071017.23677-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: memcontrol: use node page state naming scheme for memcg
Johannes Weiner [Wed, 3 May 2017 21:55:16 +0000 (14:55 -0700)]
mm: memcontrol: use node page state naming scheme for memcg

The memory controllers stat function names are awkwardly long and
arbitrarily different from the zone and node stat functions.

The current interface is named:

  mem_cgroup_read_stat()
  mem_cgroup_update_stat()
  mem_cgroup_inc_stat()
  mem_cgroup_dec_stat()
  mem_cgroup_update_page_stat()
  mem_cgroup_inc_page_stat()
  mem_cgroup_dec_page_stat()

This patch renames it to match the corresponding node stat functions:

  memcg_page_state() [node_page_state()]
  mod_memcg_state() [mod_node_state()]
  inc_memcg_state() [inc_node_state()]
  dec_memcg_state() [dec_node_state()]
  mod_memcg_page_state() [mod_node_page_state()]
  inc_memcg_page_state() [inc_node_page_state()]
  dec_memcg_page_state() [dec_node_page_state()]

Link: http://lkml.kernel.org/r/20170404220148.28338-4-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: memcontrol: re-use node VM page state enum
Johannes Weiner [Wed, 3 May 2017 21:55:13 +0000 (14:55 -0700)]
mm: memcontrol: re-use node VM page state enum

The current duplication is a high-maintenance mess, and it's painful to
add new items or query memcg state from the rest of the VM.

This increases the size of the stat array marginally, but we should aim
to track all these stats on a per-cgroup level anyway.

Link: http://lkml.kernel.org/r/20170404220148.28338-3-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: memcontrol: re-use global VM event enum
Johannes Weiner [Wed, 3 May 2017 21:55:10 +0000 (14:55 -0700)]
mm: memcontrol: re-use global VM event enum

The current duplication is a high-maintenance mess, and it's painful to
add new items.

This increases the size of the event array, but we'll eventually want
most of the VM events tracked on a per-cgroup basis anyway.

Link: http://lkml.kernel.org/r/20170404220148.28338-2-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: memcontrol: clean up memory.events counting function
Johannes Weiner [Wed, 3 May 2017 21:55:07 +0000 (14:55 -0700)]
mm: memcontrol: clean up memory.events counting function

We only ever count single events, drop the @nr parameter.  Rename the
function accordingly.  Remove low-information kerneldoc.

Link: http://lkml.kernel.org/r/20170404220148.28338-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: vmscan: fix IO/refault regression in cache workingset transition
Johannes Weiner [Wed, 3 May 2017 21:55:03 +0000 (14:55 -0700)]
mm: vmscan: fix IO/refault regression in cache workingset transition

Since commit 59dc76b0d4df ("mm: vmscan: reduce size of inactive file
list") we noticed bigger IO spikes during changes in cache access
patterns.

The patch in question shrunk the inactive list size to leave more room
for the current workingset in the presence of streaming IO.  However,
workingset transitions that previously happened on the inactive list are
now pushed out of memory and incur more refaults to complete.

This patch disables active list protection when refaults are being
observed.  This accelerates workingset transitions, and allows more of
the new set to establish itself from memory, without eating into the
ability to protect the established workingset during stable periods.

The workloads that were measurably affected for us were hit pretty bad
by it, with refault/majfault rates doubling and tripling during cache
transitions, and the machines sustaining half-hour periods of 100% IO
utilization, where they'd previously have sub-minute peaks at 60-90%.

Stateful services that handle user data tend to be more conservative
with kernel upgrades.  As a result we hit most page cache issues with
some delay, as was the case here.

The severity seemed to warrant a stable tag.

Fixes: 59dc76b0d4df ("mm: vmscan: reduce size of inactive file list")
Link: http://lkml.kernel.org/r/20170404220052.27593-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: <stable@vger.kernel.org> [4.7+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/mmap: replace SHM_HUGE_MASK with MAP_HUGE_MASK inside mmap_pgoff
Anshuman Khandual [Wed, 3 May 2017 21:55:00 +0000 (14:55 -0700)]
mm/mmap: replace SHM_HUGE_MASK with MAP_HUGE_MASK inside mmap_pgoff

Commit 091d0d55b286 ("shm: fix null pointer deref when userspace
specifies invalid hugepage size") had replaced MAP_HUGE_MASK with
SHM_HUGE_MASK.  Though both of them contain the same numeric value of
0x3f, MAP_HUGE_MASK flag sounds more appropriate than the other one in
the context.  Hence change it back.

Link: http://lkml.kernel.org/r/20170404045635.616-1-khandual@linux.vnet.ibm.com
Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Balbir Singh <bsingharora@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agooom: improve oom disable handling
Michal Hocko [Wed, 3 May 2017 21:54:57 +0000 (14:54 -0700)]
oom: improve oom disable handling

Tetsuo has reported that sysrq triggered OOM killer will print a
misleading information when no tasks are selected:

  sysrq: SysRq : Manual OOM execution
  Out of memory: Kill process 4468 ((agetty)) score 0 or sacrifice child
  Killed process 4468 ((agetty)) total-vm:43704kB, anon-rss:1760kB, file-rss:0kB, shmem-rss:0kB
  sysrq: SysRq : Manual OOM execution
  Out of memory: Kill process 4469 (systemd-cgroups) score 0 or sacrifice child
  Killed process 4469 (systemd-cgroups) total-vm:10704kB, anon-rss:120kB, file-rss:0kB, shmem-rss:0kB
  sysrq: SysRq : Manual OOM execution
  sysrq: OOM request ignored because killer is disabled
  sysrq: SysRq : Manual OOM execution
  sysrq: OOM request ignored because killer is disabled
  sysrq: SysRq : Manual OOM execution
  sysrq: OOM request ignored because killer is disabled

The real reason is that there are no eligible tasks for the OOM killer
to select but since commit 7c5f64f84483 ("mm: oom: deduplicate victim
selection code for memcg and global oom") the semantic of out_of_memory
has changed without updating moom_callback.

This patch updates moom_callback to tell that no task was eligible which
is the case for both oom killer disabled and no eligible tasks.  In
order to help distinguish first case from the second add printk to both
oom_killer_{enable,disable}.  This information is useful on its own
because it might help debugging potential memory allocation failures.

Fixes: 7c5f64f84483 ("mm: oom: deduplicate victim selection code for memcg and global oom")
Link: http://lkml.kernel.org/r/20170404134705.6361-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agouserfaultfd: selftest: combine all cases into a single executable
Mike Rapoport [Wed, 3 May 2017 21:54:54 +0000 (14:54 -0700)]
userfaultfd: selftest: combine all cases into a single executable

Currently, selftest for userfaultfd is compiled three times: for
anonymous, shared and hugetlb memory.  Let's combine all the cases into
a single executable which will have a command line option for selection
of the test type.

Link: http://lkml.kernel.org/r/1490869741-5913-1-git-send-email-rppt@linux.vnet.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.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>
7 years agomm: fix spelling error
Hao Lee [Wed, 3 May 2017 21:54:51 +0000 (14:54 -0700)]
mm: fix spelling error

Fix variable name error in comments. No code changes.

Link: http://lkml.kernel.org/r/20170403161655.5081-1-haolee.swjtu@gmail.com
Signed-off-by: Hao Lee <haolee.swjtu@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/swap_slots.c: add warning if swap slots cache failed to initialize
Tim Chen [Wed, 3 May 2017 21:54:48 +0000 (14:54 -0700)]
mm/swap_slots.c: add warning if swap slots cache failed to initialize

Add a warning diagnostics to user if we failed to allocate swap slots
cache and use it.

[akpm@linux-foundation.org: use WARN_ONCE return value, fix grammar in message]
Link: http://lkml.kernel.org/r/20170328234827.GA10107@linux.intel.com
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoinclude/linux/migrate.h: add arg names to prototype
Pushkar Jambhlekar [Wed, 3 May 2017 21:54:45 +0000 (14:54 -0700)]
include/linux/migrate.h: add arg names to prototype

It is preferred, and the rest of migrate.h gets it right.

Link: http://lkml.kernel.org/r/1490336009-8024-1-git-send-email-pushkar.iit@gmail.com
Signed-off-by: Pushkar Jambhlekar <pushkar.iit@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: enable page poisoning early at boot
Vinayak Menon [Wed, 3 May 2017 21:54:42 +0000 (14:54 -0700)]
mm: enable page poisoning early at boot

On SPARSEMEM systems page poisoning is enabled after buddy is up,
because of the dependency on page extension init.  This causes the pages
released by free_all_bootmem not to be poisoned.  This either delays or
misses the identification of some issues because the pages have to
undergo another cycle of alloc-free-alloc for any corruption to be
detected.

Enable page poisoning early by getting rid of the PAGE_EXT_DEBUG_POISON
flag.  Since all the free pages will now be poisoned, the flag need not
be verified before checking the poison during an alloc.

[vinmenon@codeaurora.org: fix Kconfig]
Link: http://lkml.kernel.org/r/1490878002-14423-1-git-send-email-vinmenon@codeaurora.org
Link: http://lkml.kernel.org/r/1490358246-11001-1-git-send-email-vinmenon@codeaurora.org
Signed-off-by: Vinayak Menon <vinmenon@codeaurora.org>
Acked-by: Laura Abbott <labbott@redhat.com>
Tested-by: Laura Abbott <labbott@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, swap: avoid lock swap_avail_lock when held cluster lock
Huang Ying [Wed, 3 May 2017 21:54:39 +0000 (14:54 -0700)]
mm, swap: avoid lock swap_avail_lock when held cluster lock

Cluster lock is used to protect the swap_cluster_info and corresponding
elements in swap_info_struct->swap_map[].  But it is found that now in
scan_swap_map_slots(), swap_avail_lock may be acquired when cluster lock
is held.  This does no good except making the locking more complex and
improving the potential locking contention, because the
swap_info_struct->lock is used to protect the data structure operated in
the code already.  Fix this via moving the corresponding operations in
scan_swap_map_slots() out of cluster lock.

Link: http://lkml.kernel.org/r/20170317064635.12792-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Tim Chen <tim.c.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, swap: improve readability via make spin_lock/unlock balanced
Huang Ying [Wed, 3 May 2017 21:54:36 +0000 (14:54 -0700)]
mm, swap: improve readability via make spin_lock/unlock balanced

This is just a cleanup patch, no functionality change.

In cluster_list_add_tail(), spin_lock_nested() is used to lock the
cluster, while unlock_cluster() is used to unlock the cluster.  To
improve the code readability, use spin_unlock() directly to unlock the
cluster.

Link: http://lkml.kernel.org/r/20170317064635.12792-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Tim Chen <tim.c.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, swap: fix comment in __read_swap_cache_async
Huang Ying [Wed, 3 May 2017 21:54:33 +0000 (14:54 -0700)]
mm, swap: fix comment in __read_swap_cache_async

Commit cbab0e4eec29 ("swap: avoid read_swap_cache_async() race to
deadlock while waiting on discard I/O completion") fixed a deadlock in
read_swap_cache_async().  Because at that time, in swap allocation path,
a swap entry may be set as SWAP_HAS_CACHE, then wait for discarding to
complete before the page for the swap entry is added to the swap cache.

But in commit 815c2c543d3a ("swap: make swap discard async"), the
discarding for swap become asynchronous, waiting for discarding to
complete will be done before the swap entry is set as SWAP_HAS_CACHE.
So the comments in code is incorrect now.  This patch fixes the
comments.

The cond_resched() added in the commit cbab0e4eec29 is not necessary now
too.  But if we added some sleep in swap allocation path in the future,
there may be some hard to debug/reproduce deadlock bug.  So it is kept.

Link: http://lkml.kernel.org/r/20170317064635.12792-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Rafael Aquini <aquini@redhat.com>
Cc: Shaohua Li <shli@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove SWAP_[SUCCESS|AGAIN|FAIL]
Minchan Kim [Wed, 3 May 2017 21:54:30 +0000 (14:54 -0700)]
mm: remove SWAP_[SUCCESS|AGAIN|FAIL]

There is no user for it.  Remove it.

[minchan@kernel.org: use false instead of SWAP_FAIL]
Link: http://lkml.kernel.org/r/20170316053313.GA19241@bbox
Link: http://lkml.kernel.org/r/1489555493-14659-11-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: make rmap_one boolean function
Minchan Kim [Wed, 3 May 2017 21:54:27 +0000 (14:54 -0700)]
mm: make rmap_one boolean function

rmap_one's return value controls whether rmap_work should contine to
scan other ptes or not so it's target for changing to boolean.  Return
true if the scan should be continued.  Otherwise, return false to stop
the scanning.

This patch makes rmap_one's return value to boolean.

Link: http://lkml.kernel.org/r/1489555493-14659-10-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: make rmap_walk() return void
Minchan Kim [Wed, 3 May 2017 21:54:23 +0000 (14:54 -0700)]
mm: make rmap_walk() return void

There is no user of the return value from rmap_walk() and friends so
this patch makes them void-returning functions.

Link: http://lkml.kernel.org/r/1489555493-14659-9-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: make ttu's return boolean
Minchan Kim [Wed, 3 May 2017 21:54:20 +0000 (14:54 -0700)]
mm: make ttu's return boolean

try_to_unmap() returns SWAP_SUCCESS or SWAP_FAIL so it's suitable for
boolean return.  This patch changes it.

Link: http://lkml.kernel.org/r/1489555493-14659-8-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove SWAP_AGAIN in ttu
Minchan Kim [Wed, 3 May 2017 21:54:17 +0000 (14:54 -0700)]
mm: remove SWAP_AGAIN in ttu

In 2002, [1] introduced SWAP_AGAIN.  At that time, try_to_unmap_one used
spin_trylock(&mm->page_table_lock) so it's really easy to contend and
fail to hold a lock so SWAP_AGAIN to keep LRU status makes sense.

However, now we changed it to mutex-based lock and be able to block
without skip pte so there is few of small window to return SWAP_AGAIN so
remove SWAP_AGAIN and just return SWAP_FAIL.

[1] c48c43e, minimal rmap

Link: http://lkml.kernel.org/r/1489555493-14659-7-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove SWAP_MLOCK in ttu
Minchan Kim [Wed, 3 May 2017 21:54:13 +0000 (14:54 -0700)]
mm: remove SWAP_MLOCK in ttu

ttu doesn't need to return SWAP_MLOCK.  Instead, just return SWAP_FAIL
because it means the page is not-swappable so it should move to another
LRU list(active or unevictable).  putback friends will move it to right
list depending on the page's LRU flag.

Link: http://lkml.kernel.org/r/1489555493-14659-6-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: make try_to_munlock() return void
Minchan Kim [Wed, 3 May 2017 21:54:10 +0000 (14:54 -0700)]
mm: make try_to_munlock() return void

try_to_munlock returns SWAP_MLOCK if the one of VMAs mapped the page has
VM_LOCKED flag.  In that time, VM set PG_mlocked to the page if the page
is not pte-mapped THP which cannot be mlocked, either.

With that, __munlock_isolated_page can use PageMlocked to check whether
try_to_munlock is successful or not without relying on try_to_munlock's
retval.  It helps to make try_to_unmap/try_to_unmap_one simple with
upcoming patches.

[minchan@kernel.org: remove PG_Mlocked VM_BUG_ON check]
Link: http://lkml.kernel.org/r/20170411025615.GA6545@bbox
Link: http://lkml.kernel.org/r/1489555493-14659-5-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove SWAP_MLOCK check for SWAP_SUCCESS in ttu
Minchan Kim [Wed, 3 May 2017 21:54:07 +0000 (14:54 -0700)]
mm: remove SWAP_MLOCK check for SWAP_SUCCESS in ttu

If the page is mapped and rescue in try_to_unmap_one, the
page_mapcount() of a page cannot be zero, so the page_mapcount check in
try_to_unmap is enough to return SWAP_SUCCESS.  IOW, SWAP_MLOCK check is
redundant so remove it.

Link: http://lkml.kernel.org/r/1489555493-14659-4-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove SWAP_DIRTY in ttu
Minchan Kim [Wed, 3 May 2017 21:54:04 +0000 (14:54 -0700)]
mm: remove SWAP_DIRTY in ttu

If we found lazyfree page is dirty, try_to_unmap_one can just
SetPageSwapBakced in there like PG_mlocked page and just return with
SWAP_FAIL which is very natural because the page is not swappable right
now so that vmscan can activate it.  There is no point to introduce new
return value SWAP_DIRTY in try_to_unmap at the moment.

Link: http://lkml.kernel.org/r/1489555493-14659-3-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove unncessary ret in page_referenced
Minchan Kim [Wed, 3 May 2017 21:54:00 +0000 (14:54 -0700)]
mm: remove unncessary ret in page_referenced

Nobody uses ret variable. Remove it.

Link: http://lkml.kernel.org/r/1489555493-14659-2-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/vmscan: more restrictive condition for retry in do_try_to_free_pages
Yisheng Xie [Wed, 3 May 2017 21:53:57 +0000 (14:53 -0700)]
mm/vmscan: more restrictive condition for retry in do_try_to_free_pages

By reviewing code, I find that when enter do_try_to_free_pages, the
may_thrash is always clear, and it will retry shrink zones to tap
cgroup's reserves memory by setting may_thrash when the former
shrink_zones reclaim nothing.

However, when memcg is disabled or on legacy hierarchy, or there do not
have any memcg protected by low limit, it should not do this useless
retry at all, for we do not have any cgroup's reserves memory to tap,
and we have already done hard work but made no progress, which as Michal
pointed out in former version, we are trying hard to control the retry
logical of page alloctor, and the current additional round of reclaim is
just lame.

Therefore, to avoid this unneeded retrying and make code more readable,
we remove the may_thrash field in scan_control, instead, introduce
memcg_low_reclaim and memcg_low_skipped, and only retry when
memcg_low_skipped, by setting memcg_low_reclaim.

[xieyisheng1@huawei.com: remove may_thrash field, introduce mem_cgroup_reclaim]
Link: http://lkml.kernel.org/r/1490191893-5923-1-git-send-email-ysxie@foxmail.com
Link: http://lkml.kernel.org/r/1490191893-5923-1-git-send-email-ysxie@foxmail.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Suggested-by: Michal Hocko <mhocko@kernel.org>
Suggested-by: Shakeel Butt <shakeelb@google.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/compaction: ignore block suitable after check large free page
Yisheng Xie [Wed, 3 May 2017 21:53:54 +0000 (14:53 -0700)]
mm/compaction: ignore block suitable after check large free page

By reviewing code, I find that if the migrate target is a large free
page and we ignore suitable, it may splite large target free page into
smaller block which is not good for defrag.  So move the ignore block
suitable after check large free page.

As Vlastimil pointed out in RFC version that this patch is just based on
logical analyses which might be better for future-proofing the function
and it is most likely won't have any visible effect right now, for
direct compaction shouldn't have to be called if there's a
>=pageblock_order page already available.

Link: http://lkml.kernel.org/r/1489490743-5364-1-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/sparse: refine usemap_size() a little
Wei Yang [Wed, 3 May 2017 21:53:51 +0000 (14:53 -0700)]
mm/sparse: refine usemap_size() a little

The current implementation calculates usemap_size in two steps:
    * calculate number of bytes to cover these bits
    * calculate number of "unsigned long" to cover these bytes

It would be more clear by:
    * calculate number of "unsigned long" to cover these bits
    * multiple it with sizeof(unsigned long)

This patch refine usemap_size() a little to make it more easy to
understand.

Link: http://lkml.kernel.org/r/20170310043713.96871-1-richard.weiyang@gmail.com
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: page_alloc: __GFP_NOWARN shouldn't suppress stall warnings
Johannes Weiner [Wed, 3 May 2017 21:53:48 +0000 (14:53 -0700)]
mm: page_alloc: __GFP_NOWARN shouldn't suppress stall warnings

__GFP_NOWARN, which is usually added to avoid warnings from callsites
that expect to fail and have fallbacks, currently also suppresses
allocation stall warnings.  These trigger when an allocation is stuck
inside the allocator for 10 seconds or longer.

But there is no class of allocations that can get legitimately stuck in
the allocator for this long.  This always indicates a problem.

Always emit stall warnings.  Restrict __GFP_NOWARN to alloc failures.

Link: http://lkml.kernel.org/r/20170125181150.GA16398@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, vmscan: prevent kswapd sleeping prematurely due to mismatched classzone_idx
Mel Gorman [Wed, 3 May 2017 21:53:45 +0000 (14:53 -0700)]
mm, vmscan: prevent kswapd sleeping prematurely due to mismatched classzone_idx

kswapd is woken to reclaim a node based on a failed allocation request
from any eligible zone.  Once reclaiming in balance_pgdat(), it will
continue reclaiming until there is an eligible zone available for the
zone it was woken for.  kswapd tracks what zone it was recently woken
for in pgdat->kswapd_classzone_idx.  If it has not been woken recently,
this zone will be 0.

However, the decision on whether to sleep is made on
kswapd_classzone_idx which is 0 without a recent wakeup request and that
classzone does not account for lowmem reserves.  This allows kswapd to
sleep when a low small zone such as ZONE_DMA is balanced for a GFP_DMA
request even if a stream of allocations cannot use that zone.  While
kswapd may be woken again shortly in the near future there are two
consequences -- the pgdat bits that control congestion are cleared
prematurely and direct reclaim is more likely as kswapd slept
prematurely.

This patch flips kswapd_classzone_idx to default to MAX_NR_ZONES (an
invalid index) when there has been no recent wakeups.  If there are no
wakeups, it'll decide whether to sleep based on the highest possible
zone available (MAX_NR_ZONES - 1).  It then becomes critical that the
"pgdat balanced" decisions during reclaim and when deciding to sleep are
the same.  If there is a mismatch, kswapd can stay awake continually
trying to balance tiny zones.

simoop was used to evaluate it again.  Two of the preparation patches
regressed the workload so they are included as the second set of
results.  Otherwise this patch looks artifically excellent

                                         4.11.0-rc1            4.11.0-rc1            4.11.0-rc1
                                            vanilla              clear-v2          keepawake-v2
Amean    p50-Read             21670074.18 (  0.00%) 19786774.76 (  8.69%) 22668332.52 ( -4.61%)
Amean    p95-Read             25456267.64 (  0.00%) 24101956.27 (  5.32%) 26738688.00 ( -5.04%)
Amean    p99-Read             29369064.73 (  0.00%) 27691872.71 (  5.71%) 30991404.52 ( -5.52%)
Amean    p50-Write                1390.30 (  0.00%)     1011.91 ( 27.22%)      924.91 ( 33.47%)
Amean    p95-Write              412901.57 (  0.00%)    34874.98 ( 91.55%)     1362.62 ( 99.67%)
Amean    p99-Write             6668722.09 (  0.00%)   575449.60 ( 91.37%)    16854.04 ( 99.75%)
Amean    p50-Allocation          78714.31 (  0.00%)    84246.26 ( -7.03%)    74729.74 (  5.06%)
Amean    p95-Allocation         175533.51 (  0.00%)   400058.43 (-127.91%)   101609.74 ( 42.11%)
Amean    p99-Allocation         247003.02 (  0.00%) 10905600.00 (-4315.17%)   125765.57 ( 49.08%)

With this patch on top, write and allocation latencies are massively
improved.  The read latencies are slightly impaired but it's worth
noting that this is mostly due to the IO scheduler and not directly
related to reclaim.  The vmstats are a bit of a mix but the relevant
ones are as follows;

                            4.10.0-rc7  4.10.0-rc7  4.10.0-rc7
                          mmots-20170209 clear-v1r25keepawake-v1r25
Swap Ins                             0           0           0
Swap Outs                            0         608           0
Direct pages scanned           6910672     3132699     6357298
Kswapd pages scanned          57036946    82488665    56986286
Kswapd pages reclaimed        55993488    63474329    55939113
Direct pages reclaimed         6905990     2964843     6352115
Kswapd efficiency                  98%         76%         98%
Kswapd velocity              12494.375   17597.507   12488.065
Direct efficiency                  99%         94%         99%
Direct velocity               1513.835     668.306    1393.148
Page writes by reclaim           0.000 4410243.000       0.000
Page writes file                     0     4409635           0
Page writes anon                     0         608           0
Page reclaim immediate         1036792    14175203     1042571

                            4.11.0-rc1  4.11.0-rc1  4.11.0-rc1
                               vanilla  clear-v2  keepawake-v2
Swap Ins                             0          12           0
Swap Outs                            0         838           0
Direct pages scanned           6579706     3237270     6256811
Kswapd pages scanned          61853702    79961486    54837791
Kswapd pages reclaimed        60768764    60755788    53849586
Direct pages reclaimed         6579055     2987453     6256151
Kswapd efficiency                  98%         75%         98%
Page writes by reclaim           0.000 4389496.000       0.000
Page writes file                     0     4388658           0
Page writes anon                     0         838           0
Page reclaim immediate         1073573    14473009      982507

Swap-outs are equivalent to baseline.

Direct reclaim is reduced but not eliminated.  It's worth noting that
there are two periods of direct reclaim for this workload.  The first is
when it switches from preparing the files for the actual test itself.
It's a lot of file IO followed by a lot of allocs that reclaims heavily
for a brief window.  While direct reclaim is lower with clear-v2, it is
due to kswapd scanning aggressively and trying to reclaim the world
which is not the right thing to do.  With the patches applied, there is
still direct reclaim but the phase change from "creating work files" to
starting multiple threads that allocate a lot of anonymous memory faster
than kswapd can reclaim.

Scanning/reclaim efficiency is restored by this patch.

Page writes from reclaim context are back at 0 which is ideal.

Pages immediately reclaimed after IO completes is slightly improved but
it is expected this will vary slightly.

On UMA, there is almost no change so this is not expected to be a
universal win.

[mgorman@suse.de: fix ->kswapd_classzone_idx initialization]
Link: http://lkml.kernel.org/r/20170406174538.5msrznj6nt6qpbx5@suse.de
Link: http://lkml.kernel.org/r/20170309075657.25121-4-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Shantanu Goel <sgoel01@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, vmscan: only clear pgdat congested/dirty/writeback state when balanced
Mel Gorman [Wed, 3 May 2017 21:53:41 +0000 (14:53 -0700)]
mm, vmscan: only clear pgdat congested/dirty/writeback state when balanced

A pgdat tracks if recent reclaim encountered too many dirty, writeback
or congested pages.  The flags control whether kswapd writes pages back
from reclaim context, tags pages for immediate reclaim when IO
completes, whether processes block on wait_iff_congested and whether
kswapd blocks when too many pages marked for immediate reclaim are
encountered.

The state is cleared in a check function with side-effects.  With the
patch "mm, vmscan: fix zone balance check in prepare_kswapd_sleep", the
timing of when the bits get cleared changed.  Due to the way the check
works, it'll clear the bits if ZONE_DMA is balanced for a GFP_DMA
allocation because it does not account for lowmem reserves properly.

For the simoop workload, kswapd is not stalling when it should due to
the premature clearing, writing pages from reclaim context like crazy
and generally being unhelpful.

This patch resets the pgdat bits related to page reclaim only when
kswapd is going to sleep.  The comparison with simoop is then

                                         4.11.0-rc1            4.11.0-rc1            4.11.0-rc1
                                            vanilla           fixcheck-v2              clear-v2
Amean    p50-Read             21670074.18 (  0.00%) 20464344.18 (  5.56%) 19786774.76 (  8.69%)
Amean    p95-Read             25456267.64 (  0.00%) 25721423.64 ( -1.04%) 24101956.27 (  5.32%)
Amean    p99-Read             29369064.73 (  0.00%) 30174230.76 ( -2.74%) 27691872.71 (  5.71%)
Amean    p50-Write                1390.30 (  0.00%)     1395.28 ( -0.36%)     1011.91 ( 27.22%)
Amean    p95-Write              412901.57 (  0.00%)    37737.74 ( 90.86%)    34874.98 ( 91.55%)
Amean    p99-Write             6668722.09 (  0.00%)   666489.04 ( 90.01%)   575449.60 ( 91.37%)
Amean    p50-Allocation          78714.31 (  0.00%)    86286.22 ( -9.62%)    84246.26 ( -7.03%)
Amean    p95-Allocation         175533.51 (  0.00%)   351812.27 (-100.42%)   400058.43 (-127.91%)
Amean    p99-Allocation         247003.02 (  0.00%)  6291171.56 (-2447.00%) 10905600.00 (-4315.17%)

Read latency is improved, write latency is mostly improved but
allocation latency is regressed.  kswapd is still reclaiming
inefficiently, pages are being written back from writeback context and a
host of other issues.  However, given the change, it needed to be
spelled out why the side-effect was moved.

Link: http://lkml.kernel.org/r/20170309075657.25121-3-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Shantanu Goel <sgoel01@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, vmscan: fix zone balance check in prepare_kswapd_sleep
Shantanu Goel [Wed, 3 May 2017 21:53:38 +0000 (14:53 -0700)]
mm, vmscan: fix zone balance check in prepare_kswapd_sleep

Patch series "Reduce amount of time kswapd sleeps prematurely", v2.

The series is unusual in that the first patch fixes one problem and
introduces other issues that are noted in the changelog.  Patch 2 makes
a minor modification that is worth considering on its own but leaves the
kernel in a state where it behaves badly.  It's not until patch 3 that
there is an improvement against baseline.

This was mostly motivated by examining Chris Mason's "simoop" benchmark
which puts the VM under similar pressure to HADOOP.  It has been
reported that the benchmark has regressed severely during the last
number of releases.  While I cannot reproduce all the same problems
Chris experienced due to hardware limitations, there was a number of
problems on a 2-socket machine with a single disk.

simoop latencies
                                         4.11.0-rc1            4.11.0-rc1
                                            vanilla          keepawake-v2
Amean    p50-Read             21670074.18 (  0.00%) 22668332.52 ( -4.61%)
Amean    p95-Read             25456267.64 (  0.00%) 26738688.00 ( -5.04%)
Amean    p99-Read             29369064.73 (  0.00%) 30991404.52 ( -5.52%)
Amean    p50-Write                1390.30 (  0.00%)      924.91 ( 33.47%)
Amean    p95-Write              412901.57 (  0.00%)     1362.62 ( 99.67%)
Amean    p99-Write             6668722.09 (  0.00%)    16854.04 ( 99.75%)
Amean    p50-Allocation          78714.31 (  0.00%)    74729.74 (  5.06%)
Amean    p95-Allocation         175533.51 (  0.00%)   101609.74 ( 42.11%)
Amean    p99-Allocation         247003.02 (  0.00%)   125765.57 ( 49.08%)

These are latencies.  Read/write are threads reading fixed-size random
blocks from a simulated database.  The allocation latency is mmaping and
faulting regions of memory.  The p50, 95 and p99 reports the worst
latencies for 50% of the samples, 95% and 99% respectively.

For example, the report indicates that while the test was running 99% of
writes completed 99.75% faster.  It's worth noting that on a UMA machine
that no difference in performance with simoop was observed so milage
will vary.

It's noted that there is a slight impact to read latencies but it's
mostly due to IO scheduler decisions and offset by the large reduction
in other latencies.

This patch (of 3):

The check in prepare_kswapd_sleep needs to match the one in
balance_pgdat since the latter will return as soon as any one of the
zones in the classzone is above the watermark.  This is specially
important for higher order allocations since balance_pgdat will
typically reset the order to zero relying on compaction to create the
higher order pages.  Without this patch, prepare_kswapd_sleep fails to
wake up kcompactd since the zone balance check fails.

It was first reported against 4.9.7 that kswapd is failing to wake up
kcompactd due to a mismatch in the zone balance check between
balance_pgdat() and prepare_kswapd_sleep().

balance_pgdat() returns as soon as a single zone satisfies the
allocation but prepare_kswapd_sleep() requires all zones to do +the
same.  This causes prepare_kswapd_sleep() to never succeed except in the
order == 0 case and consequently, wakeup_kcompactd() is never called.
For the machine that originally motivated this patch, the state of
compaction from /proc/vmstat looked this way after a day and a half +of
uptime:

compact_migrate_scanned 240496
compact_free_scanned 76238632
compact_isolated 123472
compact_stall 1791
compact_fail 29
compact_success 1762
compact_daemon_wake 0

After applying the patch and about 10 hours of uptime the state looks
like this:

compact_migrate_scanned 59927299
compact_free_scanned 2021075136
compact_isolated 640926
compact_stall 4
compact_fail 2
compact_success 2
compact_daemon_wake 5160

Further notes from Mel that motivated him to pick this patch up and
resend it;

It was observed for the simoop workload (pressures the VM similar to
HADOOP) that kswapd was failing to keep ahead of direct reclaim.  The
investigation noted that there was a need to rationalise kswapd
decisions to reclaim with kswapd decisions to sleep.  With this patch on
a 2-socket box, there was a 49% reduction in direct reclaim scanning.

However, the impact otherwise is extremely negative.  Kswapd reclaim
efficiency dropped from 98% to 76%.  simoop has three latency-related
metrics for read, write and allocation (an anonymous mmap and fault).

                                         4.11.0-rc1            4.11.0-rc1
                                            vanilla           fixcheck-v2
Amean    p50-Read             21670074.18 (  0.00%) 20464344.18 (  5.56%)
Amean    p95-Read             25456267.64 (  0.00%) 25721423.64 ( -1.04%)
Amean    p99-Read             29369064.73 (  0.00%) 30174230.76 ( -2.74%)
Amean    p50-Write                1390.30 (  0.00%)     1395.28 ( -0.36%)
Amean    p95-Write              412901.57 (  0.00%)    37737.74 ( 90.86%)
Amean    p99-Write             6668722.09 (  0.00%)   666489.04 ( 90.01%)
Amean    p50-Allocation          78714.31 (  0.00%)    86286.22 ( -9.62%)
Amean    p95-Allocation         175533.51 (  0.00%)   351812.27 (-100.42%)
Amean    p99-Allocation         247003.02 (  0.00%)  6291171.56 (-2447.00%)

Of greater concern is that the patch causes swapping and page writes
from kswapd context rose from 0 pages to 4189753 pages during the hour
the workload ran for.  By and large, the patch has very bad behaviour
but easily missed as the impact on a UMA machine is negligible.

This patch is included with the data in case a bisection leads to this
area.  This patch is also a pre-requisite for the rest of the series.

Link: http://lkml.kernel.org/r/20170309075657.25121-2-mgorman@techsingularity.net
Signed-off-by: Shantanu Goel <sgoel01@yahoo.com>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: do not use double negation for testing page flags
Minchan Kim [Wed, 3 May 2017 21:53:35 +0000 (14:53 -0700)]
mm: do not use double negation for testing page flags

With the discussion[1], I found it seems there are every PageFlags
functions return bool at this moment so we don't need double negation
any more.  Although it's not a problem to keep it, it makes future users
confused to use double negation for them, too.

Remove such possibility.

[1] https://marc.info/?l=linux-kernel&m=148881578820434

Frankly sepaking, I like every PageFlags to return bool instead of int.
It will make it clear.  AFAIR, Chen Gang had tried it but don't know why
it was not merged at that time.

http://lkml.kernel.org/r/1469336184-1904-1-git-send-email-chengang@emindsoft.com.cn

Link: http://lkml.kernel.org/r/1488868597-32222-1-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Chen Gang <gang.chen.5i5j@gmail.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove rodata_test_data export, add pr_fmt
Kees Cook [Wed, 3 May 2017 21:53:32 +0000 (14:53 -0700)]
mm: remove rodata_test_data export, add pr_fmt

Since commit 3ad38ceb2769 ("x86/mm: Remove CONFIG_DEBUG_NX_TEST"),
nothing is using the exported rodata_test_data variable, so drop the
export.

This additionally updates the pr_fmt to avoid redundant strings and
adjusts some whitespace.

Link: http://lkml.kernel.org/r/20170307005313.GA85809@beast
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Jinbum Park <jinb.park7@gmail.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: tighten up the fault path a little
Matthew Wilcox [Wed, 3 May 2017 21:53:29 +0000 (14:53 -0700)]
mm: tighten up the fault path a little

The round_up() macro generates a couple of unnecessary instructions
in this usage:

    48cd:       49 8b 47 50             mov    0x50(%r15),%rax
    48d1:       48 83 e8 01             sub    $0x1,%rax
    48d5:       48 0d ff 0f 00 00       or     $0xfff,%rax
    48db:       48 83 c0 01             add    $0x1,%rax
    48df:       48 c1 f8 0c             sar    $0xc,%rax
    48e3:       48 39 c3                cmp    %rax,%rbx
    48e6:       72 2e                   jb     4916 <filemap_fault+0x96>

If we change round_up() to ((x) + __round_mask(x, y)) & ~__round_mask(x, y)
then GCC can see through it and remove the mask (because that would be
dead code given the subsequent shift):

    48cd:       49 8b 47 50             mov    0x50(%r15),%rax
    48d1:       48 05 ff 0f 00 00       add    $0xfff,%rax
    48d7:       48 c1 e8 0c             shr    $0xc,%rax
    48db:       48 39 c3                cmp    %rax,%rbx
    48de:       72 2e                   jb     490e <filemap_fault+0x8e>

But that's problematic because we'd evaluate 'y' twice.  Converting
round_up into an inline function prevents it from being used in other
definitions.  The easiest thing to do is just change these three usages
of round_up to use DIV_ROUND_UP.  Also add an unlikely() because GCC's
heuristic is wrong in this case.

Link: http://lkml.kernel.org/r/20170207192812.5281-1-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agojbd2: make the whole kjournald2 kthread NOFS safe
Michal Hocko [Wed, 3 May 2017 21:53:26 +0000 (14:53 -0700)]
jbd2: make the whole kjournald2 kthread NOFS safe

kjournald2 is central to the transaction commit processing.  As such any
potential allocation from this kernel thread has to be GFP_NOFS.  Make
sure to mark the whole kernel thread GFP_NOFS by the memalloc_nofs_save.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20170306131408.9828-8-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Suggested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agojbd2: mark the transaction context with the scope GFP_NOFS context
Michal Hocko [Wed, 3 May 2017 21:53:22 +0000 (14:53 -0700)]
jbd2: mark the transaction context with the scope GFP_NOFS context

now that we have memalloc_nofs_{save,restore} api we can mark the whole
transaction context as implicitly GFP_NOFS.  All allocations will
automatically inherit GFP_NOFS this way.  This means that we do not have
to mark any of those requests with GFP_NOFS and moreover all the
ext4_kv[mz]alloc(GFP_NOFS) are also safe now because even the hardcoded
GFP_KERNEL allocations deep inside the vmalloc will be NOFS now.

[akpm@linux-foundation.org: tweak comments]
Link: http://lkml.kernel.org/r/20170306131408.9828-7-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoxfs: use memalloc_nofs_{save,restore} instead of memalloc_noio*
Michal Hocko [Wed, 3 May 2017 21:53:19 +0000 (14:53 -0700)]
xfs: use memalloc_nofs_{save,restore} instead of memalloc_noio*

kmem_zalloc_large and _xfs_buf_map_pages use memalloc_noio_{save,restore}
API to prevent from reclaim recursion into the fs because vmalloc can
invoke unconditional GFP_KERNEL allocations and these functions might be
called from the NOFS contexts.  The memalloc_noio_save will enforce
GFP_NOIO context which is even weaker than GFP_NOFS and that seems to be
unnecessary.  Let's use memalloc_nofs_{save,restore} instead as it
should provide exactly what we need here - implicit GFP_NOFS context.

Link: http://lkml.kernel.org/r/20170306131408.9828-6-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: introduce memalloc_nofs_{save,restore} API
Michal Hocko [Wed, 3 May 2017 21:53:15 +0000 (14:53 -0700)]
mm: introduce memalloc_nofs_{save,restore} API

GFP_NOFS context is used for the following 5 reasons currently:

 - to prevent from deadlocks when the lock held by the allocation
   context would be needed during the memory reclaim

 - to prevent from stack overflows during the reclaim because the
   allocation is performed from a deep context already

 - to prevent lockups when the allocation context depends on other
   reclaimers to make a forward progress indirectly

 - just in case because this would be safe from the fs POV

 - silence lockdep false positives

Unfortunately overuse of this allocation context brings some problems to
the MM.  Memory reclaim is much weaker (especially during heavy FS
metadata workloads), OOM killer cannot be invoked because the MM layer
doesn't have enough information about how much memory is freeable by the
FS layer.

In many cases it is far from clear why the weaker context is even used
and so it might be used unnecessarily.  We would like to get rid of
those as much as possible.  One way to do that is to use the flag in
scopes rather than isolated cases.  Such a scope is declared when really
necessary, tracked per task and all the allocation requests from within
the context will simply inherit the GFP_NOFS semantic.

Not only this is easier to understand and maintain because there are
much less problematic contexts than specific allocation requests, this
also helps code paths where FS layer interacts with other layers (e.g.
crypto, security modules, MM etc...) and there is no easy way to convey
the allocation context between the layers.

Introduce memalloc_nofs_{save,restore} API to control the scope of
GFP_NOFS allocation context.  This is basically copying
memalloc_noio_{save,restore} API we have for other restricted allocation
context GFP_NOIO.  The PF_MEMALLOC_NOFS flag already exists and it is
just an alias for PF_FSTRANS which has been xfs specific until recently.
There are no more PF_FSTRANS users anymore so let's just drop it.

PF_MEMALLOC_NOFS is now checked in the MM layer and drops __GFP_FS
implicitly same as PF_MEMALLOC_NOIO drops __GFP_IO.  memalloc_noio_flags
is renamed to current_gfp_context because it now cares about both
PF_MEMALLOC_NOFS and PF_MEMALLOC_NOIO contexts.  Xfs code paths preserve
their semantic.  kmem_flags_convert() doesn't need to evaluate the flag
anymore.

This patch shouldn't introduce any functional changes.

Let's hope that filesystems will drop direct GFP_NOFS (resp.  ~__GFP_FS)
usage as much as possible and only use a properly documented
memalloc_nofs_{save,restore} checkpoints where they are appropriate.

[akpm@linux-foundation.org: fix comment typo, reflow comment]
Link: http://lkml.kernel.org/r/20170306131408.9828-5-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoxfs: abstract PF_FSTRANS to PF_MEMALLOC_NOFS
Michal Hocko [Wed, 3 May 2017 21:53:12 +0000 (14:53 -0700)]
xfs: abstract PF_FSTRANS to PF_MEMALLOC_NOFS

xfs has defined PF_FSTRANS to declare a scope GFP_NOFS semantic quite
some time ago.  We would like to make this concept more generic and use
it for other filesystems as well.  Let's start by giving the flag a more
generic name PF_MEMALLOC_NOFS which is in line with an exiting
PF_MEMALLOC_NOIO already used for the same purpose for GFP_NOIO
contexts.  Replace all PF_FSTRANS usage from the xfs code in the first
step before we introduce a full API for it as xfs uses the flag directly
anyway.

This patch doesn't introduce any functional change.

Link: http://lkml.kernel.org/r/20170306131408.9828-4-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agolockdep: allow to disable reclaim lockup detection
Michal Hocko [Wed, 3 May 2017 21:53:09 +0000 (14:53 -0700)]
lockdep: allow to disable reclaim lockup detection

The current implementation of the reclaim lockup detection can lead to
false positives and those even happen and usually lead to tweak the code
to silence the lockdep by using GFP_NOFS even though the context can use
__GFP_FS just fine.

See

  http://lkml.kernel.org/r/20160512080321.GA18496@dastard

as an example.

  =================================
  [ INFO: inconsistent lock state ]
  4.5.0-rc2+ #4 Tainted: G           O
  ---------------------------------
  inconsistent {RECLAIM_FS-ON-R} -> {IN-RECLAIM_FS-W} usage.
  kswapd0/543 [HC0[0]:SC0[0]:HE1:SE1] takes:

  (&xfs_nondir_ilock_class){++++-+}, at: xfs_ilock+0x177/0x200 [xfs]

  {RECLAIM_FS-ON-R} state was registered at:
    mark_held_locks+0x79/0xa0
    lockdep_trace_alloc+0xb3/0x100
    kmem_cache_alloc+0x33/0x230
    kmem_zone_alloc+0x81/0x120 [xfs]
    xfs_refcountbt_init_cursor+0x3e/0xa0 [xfs]
    __xfs_refcount_find_shared+0x75/0x580 [xfs]
    xfs_refcount_find_shared+0x84/0xb0 [xfs]
    xfs_getbmap+0x608/0x8c0 [xfs]
    xfs_vn_fiemap+0xab/0xc0 [xfs]
    do_vfs_ioctl+0x498/0x670
    SyS_ioctl+0x79/0x90
    entry_SYSCALL_64_fastpath+0x12/0x6f

         CPU0
         ----
    lock(&xfs_nondir_ilock_class);
    <Interrupt>
      lock(&xfs_nondir_ilock_class);

   *** DEADLOCK ***

  3 locks held by kswapd0/543:

  stack backtrace:
  CPU: 0 PID: 543 Comm: kswapd0 Tainted: G           O    4.5.0-rc2+ #4
  Call Trace:
   lock_acquire+0xd8/0x1e0
   down_write_nested+0x5e/0xc0
   xfs_ilock+0x177/0x200 [xfs]
   xfs_reflink_cancel_cow_range+0x150/0x300 [xfs]
   xfs_fs_evict_inode+0xdc/0x1e0 [xfs]
   evict+0xc5/0x190
   dispose_list+0x39/0x60
   prune_icache_sb+0x4b/0x60
   super_cache_scan+0x14f/0x1a0
   shrink_slab.part.63.constprop.79+0x1e9/0x4e0
   shrink_zone+0x15e/0x170
   kswapd+0x4f1/0xa80
   kthread+0xf2/0x110
   ret_from_fork+0x3f/0x70

To quote Dave:
 "Ignoring whether reflink should be doing anything or not, that's a
  "xfs_refcountbt_init_cursor() gets called both outside and inside
  transactions" lockdep false positive case. The problem here is lockdep
  has seen this allocation from within a transaction, hence a GFP_NOFS
  allocation, and now it's seeing it in a GFP_KERNEL context. Also note
  that we have an active reference to this inode.

  So, because the reclaim annotations overload the interrupt level
  detections and it's seen the inode ilock been taken in reclaim
  ("interrupt") context, this triggers a reclaim context warning where
  it thinks it is unsafe to do this allocation in GFP_KERNEL context
  holding the inode ilock..."

This sounds like a fundamental problem of the reclaim lock detection.
It is really impossible to annotate such a special usecase IMHO unless
the reclaim lockup detection is reworked completely.  Until then it is
much better to provide a way to add "I know what I am doing flag" and
mark problematic places.  This would prevent from abusing GFP_NOFS flag
which has a runtime effect even on configurations which have lockdep
disabled.

Introduce __GFP_NOLOCKDEP flag which tells the lockdep gfp tracking to
skip the current allocation request.

While we are at it also make sure that the radix tree doesn't
accidentaly override tags stored in the upper part of the gfp_mask.

Link: http://lkml.kernel.org/r/20170306131408.9828-3-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agolockdep: teach lockdep about memalloc_noio_save
Nikolay Borisov [Wed, 3 May 2017 21:53:05 +0000 (14:53 -0700)]
lockdep: teach lockdep about memalloc_noio_save

Patch series "scope GFP_NOFS api", v5.

This patch (of 7):

Commit 21caf2fc1931 ("mm: teach mm by current context info to not do I/O
during memory allocation") added the memalloc_noio_(save|restore)
functions to enable people to modify the MM behavior by disabling I/O
during memory allocation.

This was further extended in commit 934f3072c17c ("mm: clear __GFP_FS
when PF_MEMALLOC_NOIO is set").

memalloc_noio_* functions prevent allocation paths recursing back into
the filesystem without explicitly changing the flags for every
allocation site.

However, lockdep hasn't been keeping up with the changes and it entirely
misses handling the memalloc_noio adjustments.  Instead, it is left to
the callers of __lockdep_trace_alloc to call the function after they
have shaven the respective GFP flags which can lead to false positives:

  =================================
   [ INFO: inconsistent lock state ]
   4.10.0-nbor #134 Not tainted
   ---------------------------------
   inconsistent {IN-RECLAIM_FS-W} -> {RECLAIM_FS-ON-W} usage.
   fsstress/3365 [HC0[0]:SC0[0]:HE1:SE1] takes:
    (&xfs_nondir_ilock_class){++++?.}, at: xfs_ilock+0x141/0x230
   {IN-RECLAIM_FS-W} state was registered at:
     __lock_acquire+0x62a/0x17c0
     lock_acquire+0xc5/0x220
     down_write_nested+0x4f/0x90
     xfs_ilock+0x141/0x230
     xfs_reclaim_inode+0x12a/0x320
     xfs_reclaim_inodes_ag+0x2c8/0x4e0
     xfs_reclaim_inodes_nr+0x33/0x40
     xfs_fs_free_cached_objects+0x19/0x20
     super_cache_scan+0x191/0x1a0
     shrink_slab+0x26f/0x5f0
     shrink_node+0xf9/0x2f0
     kswapd+0x356/0x920
     kthread+0x10c/0x140
     ret_from_fork+0x31/0x40
   irq event stamp: 173777
   hardirqs last  enabled at (173777): __local_bh_enable_ip+0x70/0xc0
   hardirqs last disabled at (173775): __local_bh_enable_ip+0x37/0xc0
   softirqs last  enabled at (173776): _xfs_buf_find+0x67a/0xb70
   softirqs last disabled at (173774): _xfs_buf_find+0x5db/0xb70

   other info that might help us debug this:
    Possible unsafe locking scenario:

          CPU0
          ----
     lock(&xfs_nondir_ilock_class);
     <Interrupt>
       lock(&xfs_nondir_ilock_class);

    *** DEADLOCK ***

   4 locks held by fsstress/3365:
    #0:  (sb_writers#10){++++++}, at: mnt_want_write+0x24/0x50
    #1:  (&sb->s_type->i_mutex_key#12){++++++}, at: vfs_setxattr+0x6f/0xb0
    #2:  (sb_internal#2){++++++}, at: xfs_trans_alloc+0xfc/0x140
    #3:  (&xfs_nondir_ilock_class){++++?.}, at: xfs_ilock+0x141/0x230

   stack backtrace:
   CPU: 0 PID: 3365 Comm: fsstress Not tainted 4.10.0-nbor #134
   Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
   Call Trace:
    kmem_cache_alloc_node_trace+0x3a/0x2c0
    vm_map_ram+0x2a1/0x510
    _xfs_buf_map_pages+0x77/0x140
    xfs_buf_get_map+0x185/0x2a0
    xfs_attr_rmtval_set+0x233/0x430
    xfs_attr_leaf_addname+0x2d2/0x500
    xfs_attr_set+0x214/0x420
    xfs_xattr_set+0x59/0xb0
    __vfs_setxattr+0x76/0xa0
    __vfs_setxattr_noperm+0x5e/0xf0
    vfs_setxattr+0xae/0xb0
    setxattr+0x15e/0x1a0
    path_setxattr+0x8f/0xc0
    SyS_lsetxattr+0x11/0x20
    entry_SYSCALL_64_fastpath+0x23/0xc6

Let's fix this by making lockdep explicitly do the shaving of respective
GFP flags.

Fixes: 934f3072c17c ("mm: clear __GFP_FS when PF_MEMALLOC_NOIO is set")
Link: http://lkml.kernel.org/r/20170306131408.9828-2-mhocko@kernel.org
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Chris Mason <clm@fb.com>
Cc: David Sterba <dsterba@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, vmstat: suppress pcp stats for unpopulated zones in zoneinfo
David Rientjes [Wed, 3 May 2017 21:53:02 +0000 (14:53 -0700)]
mm, vmstat: suppress pcp stats for unpopulated zones in zoneinfo

After "mm, vmstat: print non-populated zones in zoneinfo",
/proc/zoneinfo will show unpopulated zones.

The per-cpu pageset statistics are not relevant for unpopulated zones
and can be potentially lengthy, so supress them when they are not
interesting.

Also moves lowmem reserve protection information above pcp stats since
it is relevant for all zones per vm.lowmem_reserve_ratio.

Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1703061400500.46428@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>