Pull more iomap updates from Darrick Wong:
"Remove the now unused 'io_private' field from struct iomap_ioend, for
a modest savings in memory allocation"
* tag 'iomap-5.13-merge-3' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
iomap: remove unused private field from ioend
Pull input updates from Dmitry Torokhov:
- three new touchscreen drivers: Hycon HY46XX, ILITEK Lego Series,
and MStar MSG2638
- a new driver for Azoteq IQS626A proximity and touch controller
- addition of Amazon Game Controller to the list of devices handled
by the xpad driver
- Elan touchscreen driver will avoid binding to devices described as
I2CHID compatible in ACPI tables
- various driver fixes
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input: (56 commits)
Input: xpad - add support for Amazon Game Controller
Input: ili210x - add missing negation for touch indication on ili210x
MAINTAINERS: repair reference in HYCON HY46XX TOUCHSCREEN SUPPORT
Input: add driver for the Hycon HY46XX touchpanel series
dt-bindings: touchscreen: Add HY46XX bindings
dt-bindings: Add Hycon Technology vendor prefix
Input: cyttsp - flag the device properly
Input: cyttsp - set abs params for ABS_MT_TOUCH_MAJOR
Input: cyttsp - drop the phys path
Input: cyttsp - reduce reset pulse timings
Input: cyttsp - error message on boot mode exit error
Input: apbps2 - remove useless variable
Input: mms114 - support MMS136
Input: mms114 - convert bindings to YAML and extend
Input: Add support for ILITEK Lego Series
dt-bindings: input: touchscreen: ilitek_ts_i2c: Add bindings
Input: add MStar MSG2638 touchscreen driver
dt-bindings: input/touchscreen: add bindings for msg2638
Input: silead - add workaround for x86 BIOS-es which bring the chip up in a stuck state
Input: elants_i2c - do not bind to i2c-hid compatible ACPI instantiated devices
...
The bitmap.h header is used in a lot of code around the kernel. Besides
that it includes kernel.h which sometimes makes a loop.
The problem here is many unneeded loops that make header hell
dependencies. For example, how may you move bitmap_zalloc() from C-file
to the header? Currently it's impossible. And bitmap.h here is only the
tip of an iceberg.
kerne.h is a dump of everything that even has nothing in common at all.
We may still have it, but in my new code I prefer to include only the
headers that I want to use, without the bulk of unneeded kernel code.
Break the loop by introducing align.h, including it in kernel.h and
bitmap.h followed by replacing kernel.h with limits.h.
Link: https://lkml.kernel.org/r/20210326170347.37441-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Yury Norov <yury.norov@gmail.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull more VFIO updates from Alex Williamson:
"A second small set of commits for this merge window, primarily to
unbreak some deletions from our uAPI header.
- Additional mdev sample driver cleanup (Dan Carpenter)
- Doc fix (Alyssa Ross)
- Unbreak uAPI from NVLink2 support removal (Alex Williamson)"
* tag 'vfio-v5.13-rc1pt2' of git://github.com/awilliam/linux-vfio:
docs: vfio: fix typo
vfio/pci: Revert nvlink removal uAPI breakage
vfio/mdev: remove unnecessary NULL check in mbochs_create()
Pull RISC-V updates from Palmer Dabbelt:
- Support for the memtest= kernel command-line argument.
- Support for building the kernel with FORTIFY_SOURCE.
- Support for generic clockevent broadcasts.
- Support for the buildtar build target.
- Some build system cleanups to pass more LLVM-friendly arguments.
- Support for kprobes.
- A rearranged kernel memory map, the first part of supporting sv48
systems.
- Improvements to kexec, along with support for kdump and crash
kernels.
- An alternatives-based errata framework, along with support for
handling a pair of errata that manifest on some SiFive designs
(including the HiFive Unmatched).
- Support for XIP.
- A device tree for the Microchip PolarFire ICICLE SoC and associated
dev board.
... along with a bunch of cleanups. There are already a handful of fixes
on the list so there will likely be a part 2.
* tag 'riscv-for-linus-5.13-mw0' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux: (45 commits)
RISC-V: Always define XIP_FIXUP
riscv: Remove 32b kernel mapping from page table dump
riscv: Fix 32b kernel build with CONFIG_DEBUG_VIRTUAL=y
RISC-V: Fix error code returned by riscv_hartid_to_cpuid()
RISC-V: Enable Microchip PolarFire ICICLE SoC
RISC-V: Initial DTS for Microchip ICICLE board
dt-bindings: riscv: microchip: Add YAML documentation for the PolarFire SoC
RISC-V: Add Microchip PolarFire SoC kconfig option
RISC-V: enable XIP
RISC-V: Add crash kernel support
RISC-V: Add kdump support
RISC-V: Improve init_resources()
RISC-V: Add kexec support
RISC-V: Add EM_RISCV to kexec UAPI header
riscv: vdso: fix and clean-up Makefile
riscv/mm: Use BUG_ON instead of if condition followed by BUG.
riscv/kprobe: fix kernel panic when invoking sys_read traced by kprobe
riscv: Set ARCH_HAS_STRICT_MODULE_RWX if MMU
riscv: module: Create module allocations without exec permissions
riscv: bpf: Avoid breaking W^X
...
As of commit 966a967116 ("smp: Avoid using two cache lines for struct
call_single_data"), the smp code prefers 32-byte aligned call_single_data
objects for performance reasons, but the block layer includes an instance
of this structure in the main 'struct request' that is more senstive
to size than to performance here, see 4ccafe0320 ("block: unalign
call_single_data in struct request").
The result is a violation of the calling conventions that clang correctly
points out:
block/blk-mq.c:630:39: warning: passing 8-byte aligned argument to 32-byte aligned parameter 2 of 'smp_call_function_single_async' may result in an unaligned pointer access [-Walign-mismatch]
smp_call_function_single_async(cpu, &rq->csd);
It does seem that the usage of the call_single_data without cache line
alignment should still be allowed by the smp code, so just change the
function prototype so it accepts both, but leave the default alignment
unchanged for the other users. This seems better to me than adding
a local hack to shut up an otherwise correct warning in the caller.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Jens Axboe <axboe@kernel.dk>
Link: https://lkml.kernel.org/r/20210505211300.3174456-1-arnd@kernel.org
Provide separate helpers for guest enter vtime accounting (in addition to
the existing guest exit helpers), and move all vtime accounting helpers
to vtime.h where the existing #ifdef infrastructure can be leveraged to
better delineate the different types of accounting. This will also allow
future cleanups via deduplication of context tracking code.
Opportunstically delete the vtime_account_kernel() stub now that all
callers are wrapped with CONFIG_VIRT_CPU_ACCOUNTING_NATIVE=y.
No functional change intended.
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210505002735.1684165-6-seanjc@google.com
Move the blob of external declarations (and their stubs) above the set of
inline definitions (and their stubs) for vtime accounting. This will
allow a future patch to bring in more inline definitions without also
having to shuffle large chunks of code.
No functional change intended.
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Link: https://lore.kernel.org/r/20210505002735.1684165-5-seanjc@google.com
Merge more updates from Andrew Morton:
"The remainder of the main mm/ queue.
143 patches.
Subsystems affected by this patch series (all mm): pagecache, hugetlb,
userfaultfd, vmscan, compaction, migration, cma, ksm, vmstat, mmap,
kconfig, util, memory-hotplug, zswap, zsmalloc, highmem, cleanups, and
kfence"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (143 commits)
kfence: use power-efficient work queue to run delayed work
kfence: maximize allocation wait timeout duration
kfence: await for allocation using wait_event
kfence: zero guard page after out-of-bounds access
mm/process_vm_access.c: remove duplicate include
mm/mempool: minor coding style tweaks
mm/highmem.c: fix coding style issue
btrfs: use memzero_page() instead of open coded kmap pattern
iov_iter: lift memzero_page() to highmem.h
mm/zsmalloc: use BUG_ON instead of if condition followed by BUG.
mm/zswap.c: switch from strlcpy to strscpy
arm64/Kconfig: introduce ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
x86/Kconfig: introduce ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
mm,memory_hotplug: add kernel boot option to enable memmap_on_memory
acpi,memhotplug: enable MHP_MEMMAP_ON_MEMORY when supported
mm,memory_hotplug: allocate memmap from the added memory range
mm,memory_hotplug: factor out adjusting present pages into adjust_present_page_count()
mm,memory_hotplug: relax fully spanned sections check
drivers/base/memory: introduce memory_block_{online,offline}
mm/memory_hotplug: remove broken locking of zone PCP structures during hot remove
...
Pull virtio updates from Michael Tsirkin:
"A bunch of new drivers including vdpa support for block and
virtio-vdpa.
Beginning of vq kick (aka doorbell) mapping support.
Misc fixes"
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost: (40 commits)
virtio_pci_modern: correct sparse tags for notify
virtio_pci_modern: __force cast the notify mapping
vDPA/ifcvf: get_config_size should return dev specific config size
vDPA/ifcvf: enable Intel C5000X-PL virtio-block for vDPA
vDPA/ifcvf: deduce VIRTIO device ID when probe
vdpa_sim_blk: add support for vdpa management tool
vdpa_sim_blk: handle VIRTIO_BLK_T_GET_ID
vdpa_sim_blk: implement ramdisk behaviour
vdpa: add vdpa simulator for block device
vhost/vdpa: Remove the restriction that only supports virtio-net devices
vhost/vdpa: use get_config_size callback in vhost_vdpa_config_validate()
vdpa: add get_config_size callback in vdpa_config_ops
vdpa_sim: cleanup kiovs in vdpasim_free()
vringh: add vringh_kiov_length() helper
vringh: implement vringh_kiov_advance()
vringh: explain more about cleaning riov and wiov
vringh: reset kiov 'consumed' field in __vringh_iov()
vringh: add 'iotlb_lock' to synchronize iotlb accesses
vdpa_sim: use iova module to allocate IOVA addresses
vDPA/ifcvf: deduce VIRTIO device ID from pdev ids
...
Pull pwm updates from Thierry Reding:
"This adds support for the PWM controller found on Toshiba Visconti
SoCs and converts a couple of drivers to the atomic API.
There's also a bunch of cleanups and minor fixes across the board"
* tag 'pwm/for-5.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/thierry.reding/linux-pwm: (35 commits)
pwm: Reword docs about pwm_apply_state()
pwm: atmel: Improve duty cycle calculation in .apply()
pwm: atmel: Fix duty cycle calculation in .get_state()
pwm: visconti: Add Toshiba Visconti SoC PWM support
dt-bindings: pwm: Add bindings for Toshiba Visconti PWM Controller
arm64: dts: rockchip: Remove clock-names from PWM nodes
ARM: dts: rockchip: Remove clock-names from PWM nodes
dt-bindings: pwm: rockchip: Add more compatible strings
dt-bindings: pwm: Convert pwm-rockchip.txt to YAML
pwm: mediatek: Remove unused function
pwm: pca9685: Improve runtime PM behavior
pwm: pca9685: Support hardware readout
pwm: pca9685: Switch to atomic API
pwm: lpss: Don't modify HW state in .remove callback
pwm: sti: Free resources only after pwmchip_remove()
pwm: sti: Don't modify HW state in .remove callback
pwm: lpc3200: Don't modify HW state in .remove callback
pwm: lpc18xx-sct: Free resources only after pwmchip_remove()
pwm: bcm-kona: Don't modify HW state in .remove callback
pwm: bcm2835: Free resources only after pwmchip_remove()
...
Pull thermal updates from Daniel Lezcano:
- Remove duplicate error message for the amlogic driver (Tang Bin)
- Fix spellos in comments for the tegra and sun8i (Bhaskar Chowdhury)
- Add the missing fifth node on the rcar_gen3 sensor (Niklas Söderlund)
- Remove duplicate include in ti-bandgap (Zhang Yunkai)
- Assign error code in the error path in the function
thermal_of_populate_bind_params() (Jia-Ju Bai)
- Fix spelling mistake in a comment 'disabed' -> 'disabled' (Colin Ian
King)
- Use the device name instead of auto-numbering for a better
identification of the cooling device (Daniel Lezcano)
- Improve a bit the division accuracy in the power allocator governor
(Jeson Gao)
- Enable the missing third sensor on msm8976 (Konrad Dybcio)
- Add QCom tsens driver co-maintainer (Thara Gopinath)
- Fix memory leak and use after free errors in the core code (Daniel
Lezcano)
- Add the MDM9607 compatible bindings (Konrad Dybcio)
- Fix trivial spello in the copyright name for Hisilicon (Hao Fang)
- Fix negative index array access when converting the frequency to
power in the energy model (Brian-sy Yang)
- Add support for Gen2 new PMIC support for Qcom SPMI (David Collins)
- Update maintainer file for CPU cooling device section (Lukasz Luba)
- Fix missing put_device on error in the Qcom tsens driver (Guangqing
Zhu)
- Add compatible DT binding for sm8350 (Robert Foss)
- Add support for the MDM9607's tsens driver (Konrad Dybcio)
- Remove duplicate error messages in thermal_mmio and the bcm2835
driver (Ruiqi Gong)
- Add the Thermal Temperature Cooling driver (Zhang Rui)
- Remove duplicate error messages in the Hisilicon sensor driver (Ye
Bin)
- Use the devm_platform_ioremap_resource_byname() function instead of a
couple of corresponding calls (dingsenjie)
- Sort the headers alphabetically in the ti-bandgap driver (Zhen Lei)
- Add missing property in the DT thermal sensor binding (Rafał Miłecki)
- Remove dead code in the ti-bandgap sensor driver (Lin Ruizhe)
- Convert the BRCM DT bindings to the yaml schema (Rafał Miłecki)
- Replace the thermal_notify_framework() call by a call to the
thermal_zone_device_update() function. Remove the function as well as
the corresponding documentation (Thara Gopinath)
- Add support for the ipq8064-tsens sensor along with a set of cleanups
and code preparation (Ansuel Smith)
- Add a lockless __thermal_cdev_update() function to improve the
locking scheme in the core code and governors (Lukasz Luba)
- Fix multiple cooling device notification changes (Lukasz Luba)
- Remove unneeded variable initialization (Colin Ian King)
* tag 'thermal-v5.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/thermal/linux: (55 commits)
thermal/drivers/mtk_thermal: Remove redundant initializations of several variables
thermal/core/power allocator: Use the lockless __thermal_cdev_update() function
thermal/core/fair share: Use the lockless __thermal_cdev_update() function
thermal/core/fair share: Lock the thermal zone while looping over instances
thermal/core/power_allocator: Update once cooling devices when temp is low
thermal/core/power_allocator: Maintain the device statistics from going stale
thermal/core: Create a helper __thermal_cdev_update() without a lock
dt-bindings: thermal: tsens: Document ipq8064 bindings
thermal/drivers/tsens: Add support for ipq8064-tsens
thermal/drivers/tsens: Drop unused define for msm8960
thermal/drivers/tsens: Replace custom 8960 apis with generic apis
thermal/drivers/tsens: Fix bug in sensor enable for msm8960
thermal/drivers/tsens: Use init_common for msm8960
thermal/drivers/tsens: Add VER_0 tsens version
thermal/drivers/tsens: Convert msm8960 to reg_field
thermal/drivers/tsens: Don't hardcode sensor slope
Documentation: driver-api: thermal: Remove thermal_notify_framework from documentation
thermal/core: Remove thermal_notify_framework
iwlwifi: mvm: tt: Replace thermal_notify_framework
dt-bindings: thermal: brcm,ns-thermal: Convert to the json-schema
...
Pull gpio updates from Bartosz Golaszewski:
- new driver for the Realtek Otto GPIO controller
- ACPI support for gpio-mpc8xxx
- edge event support for gpio-sch (+ Kconfig fixes)
- Kconfig improvements in gpio-ich
- fixes to older issues in gpio-mockup
- ACPI quirk for ignoring EC wakeups on Dell Venue 10 Pro 5055
- improve the GPIO aggregator code by using more generic interfaces
instead of reimplementing them in the driver
- convert the DT bindings for gpio-74x164 to yaml
- documentation improvements
- a slew of other minor fixes and improvements to GPIO drivers
* tag 'gpio-updates-for-v5.13-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux: (34 commits)
dt-bindings: gpio: add YAML description for rockchip,gpio-bank
gpio: mxs: remove useless function
dt-bindings: gpio: fairchild,74hc595: Convert to json-schema
gpio: it87: remove unused code
gpio: 104-dio-48e: Fix coding style issues
gpio: mpc8xxx: Add ACPI support
gpio: ich: Switch to be dependent on LPC_ICH
gpio: sch: Drop MFD_CORE selection
gpio: sch: depends on LPC_SCH
gpiolib: acpi: Add quirk to ignore EC wakeups on Dell Venue 10 Pro 5055
gpio: sch: Hook into ACPI GPE handler to catch GPIO edge events
gpio: sch: Add edge event support
gpio: aggregator: Replace custom get_arg() with a generic next_arg()
lib/cmdline: Export next_arg() for being used in modules
gpio: omap: Use device_get_match_data() helper
gpio: Add Realtek Otto GPIO support
dt-bindings: gpio: Binding for Realtek Otto GPIO
docs: kernel-parameters: Add gpio_mockup_named_lines
docs: kernel-parameters: Move gpio-mockup for alphabetic order
lib: bitmap: provide devm_bitmap_alloc() and devm_bitmap_zalloc()
...
Physical memory hotadd has to allocate a memmap (struct page array) for
the newly added memory section. Currently, alloc_pages_node() is used
for those allocations.
This has some disadvantages:
a) an existing memory is consumed for that purpose
(eg: ~2MB per 128MB memory section on x86_64)
This can even lead to extreme cases where system goes OOM because
the physically hotplugged memory depletes the available memory before
it is onlined.
b) if the whole node is movable then we have off-node struct pages
which has performance drawbacks.
c) It might be there are no PMD_ALIGNED chunks so memmap array gets
populated with base pages.
This can be improved when CONFIG_SPARSEMEM_VMEMMAP is enabled.
Vmemap page tables can map arbitrary memory. That means that we can
reserve a part of the physically hotadded memory to back vmemmap page
tables. This implementation uses the beginning of the hotplugged memory
for that purpose.
There are some non-obviously things to consider though.
Vmemmap pages are allocated/freed during the memory hotplug events
(add_memory_resource(), try_remove_memory()) when the memory is
added/removed. This means that the reserved physical range is not
online although it is used. The most obvious side effect is that
pfn_to_online_page() returns NULL for those pfns. The current design
expects that this should be OK as the hotplugged memory is considered a
garbage until it is onlined. For example hibernation wouldn't save the
content of those vmmemmaps into the image so it wouldn't be restored on
resume but this should be OK as there no real content to recover anyway
while metadata is reachable from other data structures (e.g. vmemmap
page tables).
The reserved space is therefore (de)initialized during the {on,off}line
events (mhp_{de}init_memmap_on_memory). That is done by extracting page
allocator independent initialization from the regular onlining path.
The primary reason to handle the reserved space outside of
{on,off}line_pages is to make each initialization specific to the
purpose rather than special case them in a single function.
As per above, the functions that are introduced are:
- mhp_init_memmap_on_memory:
Initializes vmemmap pages by calling move_pfn_range_to_zone(), calls
kasan_add_zero_shadow(), and onlines as many sections as vmemmap pages
fully span.
- mhp_deinit_memmap_on_memory:
Offlines as many sections as vmemmap pages fully span, removes the
range from zhe zone by remove_pfn_range_from_zone(), and calls
kasan_remove_zero_shadow() for the range.
The new function memory_block_online() calls mhp_init_memmap_on_memory()
before doing the actual online_pages(). Should online_pages() fail, we
clean up by calling mhp_deinit_memmap_on_memory(). Adjusting of
present_pages is done at the end once we know that online_pages()
succedeed.
On offline, memory_block_offline() needs to unaccount vmemmap pages from
present_pages() before calling offline_pages(). This is necessary because
offline_pages() tears down some structures based on the fact whether the
node or the zone become empty. If offline_pages() fails, we account back
vmemmap pages. If it succeeds, we call mhp_deinit_memmap_on_memory().
Hot-remove:
We need to be careful when removing memory, as adding and
removing memory needs to be done with the same granularity.
To check that this assumption is not violated, we check the
memory range we want to remove and if a) any memory block has
vmemmap pages and b) the range spans more than a single memory
block, we scream out loud and refuse to proceed.
If all is good and the range was using memmap on memory (aka vmemmap pages),
we construct an altmap structure so free_hugepage_table does the right
thing and calls vmem_altmap_free instead of free_pagetable.
Link: https://lkml.kernel.org/r/20210421102701.25051-5-osalvador@suse.de
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.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>
To help with debugging the sluggishness caused by TLB miss/reload, we
introduce monotonic hugepage [direct mapped] split event counts since
system state: SYSTEM_RUNNING to be displayed as part of /proc/vmstat in
x86 servers
The lifetime split event information will be displayed at the bottom of
/proc/vmstat
....
swap_ra 0
swap_ra_hit 0
direct_map_level2_splits 94
direct_map_level3_splits 4
nr_unstable 0
....
One of the many lasting sources of direct hugepage splits is kernel
tracing (kprobes, tracepoints).
Note that the kernel's code segment [512 MB] points to the same physical
addresses that have been already mapped in the kernel's direct mapping
range.
Source : Documentation/x86/x86_64/mm.rst
When we enable kernel tracing, the kernel has to modify
attributes/permissions of the text segment hugepages that are direct
mapped causing them to split.
Kernel's direct mapped hugepages do not coalesce back after split and
remain in place for the remainder of the lifetime.
An instance of direct page splits when we turn on dynamic kernel tracing
....
cat /proc/vmstat | grep -i direct_map_level
direct_map_level2_splits 784
direct_map_level3_splits 12
bpftrace -e 'tracepoint:raw_syscalls:sys_enter { @ [pid, comm] =
count(); }'
cat /proc/vmstat | grep -i
direct_map_level
direct_map_level2_splits 789
direct_map_level3_splits 12
....
Link: https://lkml.kernel.org/r/20210218235744.1040634-1-saravanand@fb.com
Signed-off-by: Saravanan D <saravanand@fb.com>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pages containing buffer_heads that are in one of the per-CPU buffer_head
LRU caches will be pinned and thus cannot be migrated. This can prevent
CMA allocations from succeeding, which are often used on platforms with
co-processors (such as a DSP) that can only use physically contiguous
memory. It can also prevent memory hot-unplugging from succeeding,
which involves migrating at least MIN_MEMORY_BLOCK_SIZE bytes of memory,
which ranges from 8 MiB to 1 GiB based on the architecture in use.
Correspondingly, invalidate the BH LRU caches before a migration starts
and stop any buffer_head from being cached in the LRU caches, until
migration has finished.
Link: https://lkml.kernel.org/r/20210319175127.886124-3-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reported-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Reported-by: Laura Abbott <labbott@kernel.org>
Tested-by: Oliver Sang <oliver.sang@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: John Dias <joaodias@google.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Suren Baghdasaryan <surenb@google.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>
LRU pagevec holds refcount of pages until the pagevec are drained. It
could prevent migration since the refcount of the page is greater than
the expection in migration logic. To mitigate the issue, callers of
migrate_pages drains LRU pagevec via migrate_prep or lru_add_drain_all
before migrate_pages call.
However, it's not enough because pages coming into pagevec after the
draining call still could stay at the pagevec so it could keep
preventing page migration. Since some callers of migrate_pages have
retrial logic with LRU draining, the page would migrate at next trail
but it is still fragile in that it doesn't close the fundamental race
between upcoming LRU pages into pagvec and migration so the migration
failure could cause contiguous memory allocation failure in the end.
To close the race, this patch disables lru caches(i.e, pagevec) during
ongoing migration until migrate is done.
Since it's really hard to reproduce, I measured how many times
migrate_pages retried with force mode(it is about a fallback to a sync
migration) with below debug code.
int migrate_pages(struct list_head *from, new_page_t get_new_page,
..
..
if (rc && reason == MR_CONTIG_RANGE && pass > 2) {
printk(KERN_ERR, "pfn 0x%lx reason %d", page_to_pfn(page), rc);
dump_page(page, "fail to migrate");
}
The test was repeating android apps launching with cma allocation in
background every five seconds. Total cma allocation count was about 500
during the testing. With this patch, the dump_page count was reduced
from 400 to 30.
The new interface is also useful for memory hotplug which currently
drains lru pcp caches after each migration failure. This is rather
suboptimal as it has to disrupt others running during the operation.
With the new interface the operation happens only once. This is also in
line with pcp allocator cache which are disabled for the offlining as
well.
Link: https://lkml.kernel.org/r/20210319175127.886124-1-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: John Dias <joaodias@google.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Oliver Sang <oliver.sang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently the number of deferred objects are per shrinker, but some
slabs, for example, vfs inode/dentry cache are per memcg, this would
result in poor isolation among memcgs.
The deferred objects typically are generated by __GFP_NOFS allocations,
one memcg with excessive __GFP_NOFS allocations may blow up deferred
objects, then other innocent memcgs may suffer from over shrink,
excessive reclaim latency, etc.
For example, two workloads run in memcgA and memcgB respectively,
workload in B is vfs heavy workload. Workload in A generates excessive
deferred objects, then B's vfs cache might be hit heavily (drop half of
caches) by B's limit reclaim or global reclaim.
We observed this hit in our production environment which was running vfs
heavy workload shown as the below tracing log:
<...>-409454 [016] .... 28286961.747146: mm_shrink_slab_start: super_cache_scan+0x0/0x1a0 ffff9a83046f3458:
nid: 1 objects to shrink 3641681686040 gfp_flags GFP_HIGHUSER_MOVABLE|__GFP_ZERO pgs_scanned 1 lru_pgs 15721
cache items 246404277 delta 31345 total_scan 123202138
<...>-409454 [022] .... 28287105.928018: mm_shrink_slab_end: super_cache_scan+0x0/0x1a0 ffff9a83046f3458:
nid: 1 unused scan count 3641681686040 new scan count 3641798379189 total_scan 602
last shrinker return val 123186855
The vfs cache and page cache ratio was 10:1 on this machine, and half of
caches were dropped. This also resulted in significant amount of page
caches were dropped due to inodes eviction.
Make nr_deferred per memcg for memcg aware shrinkers would solve the
unfairness and bring better isolation.
The following patch will add nr_deferred to parent memcg when memcg
offline. To preserve nr_deferred when reparenting memcgs to root, root
memcg needs shrinker_info allocated too.
When memcg is not enabled (!CONFIG_MEMCG or memcg disabled), the
shrinker's nr_deferred would be used. And non memcg aware shrinkers use
shrinker's nr_deferred all the time.
Link: https://lkml.kernel.org/r/20210311190845.9708-10-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Acked-by: Roman Gushchin <guro@fb.com>
Acked-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
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>
