Ian Rogers [Sat, 22 Jan 2022 04:58:10 +0000 (20:58 -0800)]
perf cpumap: Migrate to libperf cpumap api
Switch from directly accessing the perf_cpu_map to using the appropriate
libperf API when possible. Using the API simplifies the job of
refactoring use of perf_cpu_map.
Perf script was failed to print the phys_addr for SPE profiling.
One 'dummy' event is added by SPE profiling but it doesn't have PHYS_ADDR
attribute set, perf script then exits with error.
Now referring to 'addr', use evsel__do_check_stype() to check the type.
Before:
# perf record -e arm_spe_0/branch_filter=0,ts_enable=1,pa_enable=1,load_filter=1,jitter=0,\
store_filter=0,min_latency=0,event_filter=2/ -p 4064384 -- sleep 3
# perf script -F pid,tid,addr,phys_addr
Samples for 'dummy:u' event do not have PHYS_ADDR attribute set. Cannot print 'phys_addr' field.
Masahiro Yamada [Thu, 20 Jan 2022 19:22:05 +0000 (04:22 +0900)]
certs: Fix build error when CONFIG_MODULE_SIG_KEY is empty
Since b8c96a6b466c ("certs: simplify $(srctree)/ handling and remove
config_filename macro"), when CONFIG_MODULE_SIG_KEY is empty,
signing_key.x509 fails to build:
Dmitry V. Levin [Mon, 3 Jan 2022 01:24:02 +0000 (04:24 +0300)]
usr/include/Makefile: add linux/nfc.h to the compile-test coverage
As linux/nfc.h userspace compilation was finally fixed by commits 79b69a83705e ("nfc: uapi: use kernel size_t to fix user-space builds")
and 7175f02c4e5f ("uapi: fix linux/nfc.h userspace compilation errors"),
there is no need to keep the compile-test exception for it in
usr/include/Makefile.
Linus Torvalds [Sat, 22 Jan 2022 09:28:23 +0000 (11:28 +0200)]
Merge branch 'akpm' (patches from Andrew)
Merge yet more updates from Andrew Morton:
"This is the post-linux-next queue. Material which was based on or
dependent upon material which was in -next.
69 patches.
Subsystems affected by this patch series: mm (migration and zsmalloc),
sysctl, proc, and lib"
* emailed patches from Andrew Morton <[email protected]>: (69 commits)
mm: hide the FRONTSWAP Kconfig symbol
frontswap: remove support for multiple ops
mm: mark swap_lock and swap_active_head static
frontswap: simplify frontswap_register_ops
frontswap: remove frontswap_test
mm: simplify try_to_unuse
frontswap: remove the frontswap exports
frontswap: simplify frontswap_init
frontswap: remove frontswap_curr_pages
frontswap: remove frontswap_shrink
frontswap: remove frontswap_tmem_exclusive_gets
frontswap: remove frontswap_writethrough
mm: remove cleancache
lib/stackdepot: always do filter_irq_stacks() in stack_depot_save()
lib/stackdepot: allow optional init and stack_table allocation by kvmalloc()
proc: remove PDE_DATA() completely
fs: proc: store PDE()->data into inode->i_private
zsmalloc: replace get_cpu_var with local_lock
zsmalloc: replace per zpage lock with pool->migrate_lock
locking/rwlocks: introduce write_lock_nested
...
Linus Torvalds [Sat, 22 Jan 2022 09:12:26 +0000 (11:12 +0200)]
Merge tag '5.17-rc-part2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Pull cifs fixes from Steve French:
- multichannel fixes, addressing additional reconnect and DFS scenarios
- reenabling fscache support (indexing rewrite, metadata caching e.g.)
- send additional version information during NTLMSSP negotiate to
improve debugging
- fix for a mount race
- DFS fixes
- fix for a memory leak for stable
* tag '5.17-rc-part2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
cifs: update internal module number
smb3: send NTLMSSP version information
cifs: Support fscache indexing rewrite
cifs: cifs_ses_mark_for_reconnect should also update reconnect bits
cifs: update tcpStatus during negotiate and sess setup
cifs: make status checks in version independent callers
cifs: remove repeated state change in dfs tree connect
cifs: fix the cifs_reconnect path for DFS
cifs: remove unused variable ses_selected
cifs: protect all accesses to chan_* with chan_lock
cifs: fix the connection state transitions with multichannel
cifs: check reconnects for channels of active tcons too
smb3: add new defines from protocol specification
cifs: serialize all mount attempts
cifs: quirk for STATUS_OBJECT_NAME_INVALID returned for non-ASCII dfs refs
cifs: alloc_path_with_tree_prefix: do not append sep. if the path is empty
cifs: clean up an inconsistent indenting
cifs: free ntlmsspblob allocated in negotiate
Linus Torvalds [Sat, 22 Jan 2022 09:04:27 +0000 (11:04 +0200)]
Merge tag 'xfs-5.17-merge-7' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull xfs fixes from Darrick Wong:
"One of the patches removes some dead code from xfs_ioctl32.h and the
other fixes broken workqueue flushing in the inode garbage collector.
- Minor cleanup of ioctl32 cruft
- Clean up open coded inodegc workqueue function calls"
* tag 'xfs-5.17-merge-7' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
xfs: flush inodegc workqueue tasks before cancel
xfs: remove unused xfs_ioctl32.h declarations
Linus Torvalds [Sat, 22 Jan 2022 08:59:32 +0000 (10:59 +0200)]
Merge tag 'fscache-fixes-20220121' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs
Pull more fscache updates from David Howells:
"A set of fixes and minor updates for the fscache rewrite:
- Fix mishandling of volume collisions (the wait condition is
inverted and so it was only waiting if the volume collision was
already resolved).
- Fix miscalculation of whether there's space available in
cachefiles.
- Make sure a default cache name is set on a cache if the user hasn't
set one by the time they bind the cache.
- Adjust the way the backing inode is presented in tracepoints, add a
tracepoint for mkdir and trace directory lookup.
- Add a tracepoint for failure to set the active file mark.
- Add an explanation of the checks made on the backing filesystem.
- Check that the backing filesystem supports tmpfile.
- Document how the page-release cancellation of the read-skip
optimisation works.
And I've included a change for netfslib:
- Make ops->init_rreq() optional"
* tag 'fscache-fixes-20220121' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs:
netfs: Make ops->init_rreq() optional
fscache: Add a comment explaining how page-release optimisation works
cachefiles: Check that the backing filesystem supports tmpfiles
cachefiles: Explain checks in a comment
cachefiles: Trace active-mark failure
cachefiles: Make some tracepoint adjustments
cachefiles: set default tag name if it's unspecified
cachefiles: Calculate the blockshift in terms of bytes, not pages
fscache: Fix the volume collision wait condition
Linus Torvalds [Sat, 22 Jan 2022 08:43:07 +0000 (10:43 +0200)]
Merge tag 'folio-5.17a' of git://git.infradead.org/users/willy/pagecache
Pull more folio updates from Matthew Wilcox:
"Three small folio patches.
One bug fix, one patch pulled forward from the patches destined for
5.18 and then a patch to make use of that functionality"
* tag 'folio-5.17a' of git://git.infradead.org/users/willy/pagecache:
filemap: Use folio_put_refs() in filemap_free_folio()
mm: Add folio_put_refs()
pagevec: Initialise folio_batch->percpu_pvec_drained
Linus Torvalds [Sat, 22 Jan 2022 08:24:02 +0000 (10:24 +0200)]
Merge tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull more SCSI updates from James Bottomley:
"This series is all the stragglers that didn't quite make the first
merge window pull. It's mostly minor updates and bug fixes of merge
window code"
* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: nsp_cs: Check of ioremap return value
scsi: ufs: ufs-mediatek: Fix error checking in ufs_mtk_init_va09_pwr_ctrl()
scsi: ufs: Modify Tactive time setting conditions
scsi: efct: Remove useless DMA-32 fallback configuration
scsi: message: fusion: mptctl: Use dma_alloc_coherent()
scsi: message: fusion: mptsas: Use dma_alloc_coherent()
scsi: message: fusion: Use dma_alloc_coherent() in mptsas_exp_repmanufacture_info()
scsi: message: fusion: mptbase: Use dma_alloc_coherent()
scsi: message: fusion: Use dma_alloc_coherent() in mpt_alloc_fw_memory()
scsi: message: fusion: Remove usage of the deprecated "pci-dma-compat.h" API
scsi: megaraid: Avoid mismatched storage type sizes
scsi: hisi_sas: Remove unused variable and check in hisi_sas_send_ata_reset_each_phy()
scsi: aic79xx: Remove redundant error variable
scsi: pm80xx: Port reset timeout error handling correction
scsi: mpi3mr: Fix formatting problems in some kernel-doc comments
scsi: mpi3mr: Fix some spelling mistakes
scsi: mpt3sas: Update persistent trigger pages from sysfs interface
scsi: core: Fix scsi_mode_select() interface
scsi: aacraid: Fix spelling of "its"
scsi: qedf: Fix potential dereference of NULL pointer
Linus Torvalds [Sat, 22 Jan 2022 08:15:41 +0000 (10:15 +0200)]
Merge tag 'thermal-5.17-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull more thermal control updates from Rafael Wysocki:
"Add device IDs for Raptor Lake to the int340x thermal control driver
(Srinivas Pandruvada)"
* tag 'thermal-5.17-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
thermal: int340x: Add Raptor Lake PCI device id
thermal: int340x: Support Raptor Lake
Linus Torvalds [Sat, 22 Jan 2022 08:09:51 +0000 (10:09 +0200)]
Merge tag 'acpi-5.17-rc1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull extra ACPI updates from Rafael Wysocki:
"These fix and clean up the ACPI CPPC driver on top of the recent
changes in it merged previously and add some new device IDs to the
ACPI DPTF driver.
Specifics:
- Fix a recently introduced endianness-related issue in the ACPI CPPC
library and clean it up on top of that (Rafael Wysocki)
- Add new device IDs for the Raptor Lake SoC to the ACPI DPTF driver
(Srinivas Pandruvada)"
* tag 'acpi-5.17-rc1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
ACPI: DPTF: Support Raptor Lake
ACPI: CPPC: Drop redundant local variable from cpc_read()
ACPI: CPPC: Fix up I/O port access in cpc_read()
Linus Torvalds [Sat, 22 Jan 2022 07:52:17 +0000 (09:52 +0200)]
Merge tag 'devicetree-fixes-for-5.17-1' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux
Pull devicetree fixes and cleanups from Rob Herring:
- Fix a regression when probing a child device reusing the parent
device's DT node pointer
- Refactor of_parse_phandle*() variants to static inlines
- Drop Enric Balletbo i Serra as a maintainer
- Fix DT schemas with arrays incorrectly encoded as a matrix
- Drop unneeded pinctrl properties from schemas
- Add SPI peripheral schema to SPI based displays
- Clean-up several schema examples
- Clean-up trivial-devices.yaml comments
- Add missing, in use vendor prefixes: Wingtech, Thundercomm, Huawei,
F(x)tec, 8devices
* tag 'devicetree-fixes-for-5.17-1' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
dt-bindings: google,cros-ec: drop Enric Balletbo i Serra from maintainers
dt-bindings: display: bridge: drop Enric Balletbo i Serra from maintainers
of: Check 'of_node_reused' flag on of_match_device()
of: property: define of_property_read_u{8,16,32,64}_array() unconditionally
of: base: make small of_parse_phandle() variants static inline
dt-bindings: mfd: cirrus,madera: Fix 'interrupts' in example
dt-bindings: Fix array schemas encoded as matrices
dt-bindings: Drop unnecessary pinctrl properties
dt-bindings: rtc: st,stm32-rtc: Make each example a separate entry
dt-bindings: mmc: arm,pl18x: Make each example a separate entry
dt-bindings: display: Add SPI peripheral schema to SPI based displays
scripts/dtc: dtx_diff: remove broken example from help text
dt-bindings: trivial-devices: fix double spaces in comments
dt-bindings: trivial-devices: fix swapped comments
dt-bindings: vendor-prefixes: add Wingtech
dt-bindings: vendor-prefixes: add Thundercomm
dt-bindings: vendor-prefixes: add Huawei
dt-bindings: vendor-prefixes: add F(x)tec
dt-bindings: vendor-prefixes: add 8devices
dt-bindings: power: reset: gpio-restart: Correct default priority
Linus Torvalds [Sat, 22 Jan 2022 07:40:01 +0000 (09:40 +0200)]
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull more kvm updates from Paolo Bonzini:
"Generic:
- selftest compilation fix for non-x86
- KVM: avoid warning on s390 in mark_page_dirty
x86:
- fix page write-protection bug and improve comments
- use binary search to lookup the PMU event filter, add test
- enable_pmu module parameter support for Intel CPUs
- switch blocked_vcpu_on_cpu_lock to raw spinlock
- cleanups of blocked vCPU logic
- partially allow KVM_SET_CPUID{,2} after KVM_RUN (5.16 regression)
- various small fixes"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (46 commits)
docs: kvm: fix WARNINGs from api.rst
selftests: kvm/x86: Fix the warning in lib/x86_64/processor.c
selftests: kvm/x86: Fix the warning in pmu_event_filter_test.c
kvm: selftests: Do not indent with spaces
kvm: selftests: sync uapi/linux/kvm.h with Linux header
selftests: kvm: add amx_test to .gitignore
KVM: SVM: Nullify vcpu_(un)blocking() hooks if AVIC is disabled
KVM: SVM: Move svm_hardware_setup() and its helpers below svm_x86_ops
KVM: SVM: Drop AVIC's intermediate avic_set_running() helper
KVM: VMX: Don't do full kick when handling posted interrupt wakeup
KVM: VMX: Fold fallback path into triggering posted IRQ helper
KVM: VMX: Pass desired vector instead of bool for triggering posted IRQ
KVM: VMX: Don't do full kick when triggering posted interrupt "fails"
KVM: SVM: Skip AVIC and IRTE updates when loading blocking vCPU
KVM: SVM: Use kvm_vcpu_is_blocking() in AVIC load to handle preemption
KVM: SVM: Remove unnecessary APICv/AVIC update in vCPU unblocking path
KVM: SVM: Don't bother checking for "running" AVIC when kicking for IPIs
KVM: SVM: Signal AVIC doorbell iff vCPU is in guest mode
KVM: x86: Remove defunct pre_block/post_block kvm_x86_ops hooks
KVM: x86: Unexport LAPIC's switch_to_{hv,sw}_timer() helpers
...
Linus Torvalds [Sat, 22 Jan 2022 07:37:31 +0000 (09:37 +0200)]
Merge tag 'for-5.17/parisc-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux
Pull more parisc architecture updates from Helge Deller:
"Fixes and enhancements:
- a memory leak fix in an error path in pdc_stable (Miaoqian Lin)
- two compiler warning fixes in the TOC code
- added autodetection for currently used console type (serial or
graphics) which inserts console=<type> if it's missing"
* tag 'for-5.17/parisc-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
parisc: pdc_stable: Fix memory leak in pdcs_register_pathentries
parisc: Fix missing prototype for 'toc_intr' warning in toc.c
parisc: Autodetect default output device and set console= kernel parameter
parisc: Use safer strscpy() in setup_cmdline()
parisc: Add visible flag to toc_stack variable
Linus Torvalds [Sat, 22 Jan 2022 07:34:49 +0000 (09:34 +0200)]
Merge tag 'riscv-for-linus-5.17-mw1' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux
Pull more RISC-V updates from Palmer Dabbelt:
- Support for sv48 paging
- Hart ID mappings are now sparse, which enables more CPUs to come up
on systems with sparse hart IDs
- A handful of cleanups and fixes
* tag 'riscv-for-linus-5.17-mw1' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux: (27 commits)
RISC-V: nommu_virt: Drop unused SLAB_MERGE_DEFAULT
RISC-V: Remove redundant err variable
riscv: dts: sifive unmatched: Add gpio poweroff
riscv: canaan: remove useless select of non-existing config SYSCON
RISC-V: Do not use cpumask data structure for hartid bitmap
RISC-V: Move spinwait booting method to its own config
RISC-V: Move the entire hart selection via lottery to SMP
RISC-V: Use __cpu_up_stack/task_pointer only for spinwait method
RISC-V: Do not print the SBI version during HSM extension boot print
RISC-V: Avoid using per cpu array for ordered booting
riscv: default to CONFIG_RISCV_SBI_V01=n
riscv: fix boolconv.cocci warnings
riscv: Explicit comment about user virtual address space size
riscv: Use pgtable_l4_enabled to output mmu_type in cpuinfo
riscv: Implement sv48 support
asm-generic: Prepare for riscv use of pud_alloc_one and pud_free
riscv: Allow to dynamically define VA_BITS
riscv: Introduce functions to switch pt_ops
riscv: Split early kasan mapping to prepare sv48 introduction
riscv: Move KASAN mapping next to the kernel mapping
...
Linus Torvalds [Sat, 22 Jan 2022 07:02:57 +0000 (09:02 +0200)]
Merge tag 'docs-5.17-2' of git://git.lwn.net/linux
Pull documentation fixes from Jonathan Corbet:
"Three small documentation fixes"
* tag 'docs-5.17-2' of git://git.lwn.net/linux:
Documentation: fix firewire.rst ABI file path error
docs: ftrace: fix ambiguous sentence
docs: staging/tee.rst: fix two typos found while reading
Given that frontswap_register_ops must be called from built-in code,
there is no need to handle the case of swapfiles coming online before or
during it, so delete the code that deals with that case.
Since the removal of the Xen tmem driver in 2019, the cleancache hooks
are entirely unused, as are large parts of frontswap. This series
against linux-next (with the folio changes included) removes
cleancaches, and cuts down frontswap to the bits actually used by zswap.
This patch (of 13):
The cleancache subsystem is unused since the removal of Xen tmem driver
in commit 814bbf49dcd0 ("xen: remove tmem driver").
Marco Elver [Sat, 22 Jan 2022 06:14:31 +0000 (22:14 -0800)]
lib/stackdepot: always do filter_irq_stacks() in stack_depot_save()
The non-interrupt portion of interrupt stack traces before interrupt
entry is usually arbitrary. Therefore, saving stack traces of
interrupts (that include entries before interrupt entry) to stack depot
leads to unbounded stackdepot growth.
As such, use of filter_irq_stacks() is a requirement to ensure
stackdepot can efficiently deduplicate interrupt stacks.
Looking through all current users of stack_depot_save(), none (except
KASAN) pass the stack trace through filter_irq_stacks() before passing
it on to stack_depot_save().
Rather than adding filter_irq_stacks() to all current users of
stack_depot_save(), it became clear that stack_depot_save() should
simply do filter_irq_stacks().
Vlastimil Babka [Sat, 22 Jan 2022 06:14:27 +0000 (22:14 -0800)]
lib/stackdepot: allow optional init and stack_table allocation by kvmalloc()
Currently, enabling CONFIG_STACKDEPOT means its stack_table will be
allocated from memblock, even if stack depot ends up not actually used.
The default size of stack_table is 4MB on 32-bit, 8MB on 64-bit.
This is fine for use-cases such as KASAN which is also a config option
and has overhead on its own. But it's an issue for functionality that
has to be actually enabled on boot (page_owner) or depends on hardware
(GPU drivers) and thus the memory might be wasted. This was raised as
an issue [1] when attempting to add stackdepot support for SLUB's debug
object tracking functionality. It's common to build kernels with
CONFIG_SLUB_DEBUG and enable slub_debug on boot only when needed, or
create only specific kmem caches with debugging for testing purposes.
It would thus be more efficient if stackdepot's table was allocated only
when actually going to be used. This patch thus makes the allocation
(and whole stack_depot_init() call) optional:
- Add a CONFIG_STACKDEPOT_ALWAYS_INIT flag to keep using the current
well-defined point of allocation as part of mem_init(). Make
CONFIG_KASAN select this flag.
- Other users have to call stack_depot_init() as part of their own init
when it's determined that stack depot will actually be used. This may
depend on both config and runtime conditions. Convert current users
which are page_owner and several in the DRM subsystem. Same will be
done for SLUB later.
- Because the init might now be called after the boot-time memblock
allocation has given all memory to the buddy allocator, change
stack_depot_init() to allocate stack_table with kvmalloc() when
memblock is no longer available. Also handle allocation failure by
disabling stackdepot (could have theoretically happened even with
memblock allocation previously), and don't unnecessarily align the
memblock allocation to its own size anymore.
On FLATMEM, we call page_ext_init_flatmem_late() just before
kmem_cache_init() which means stack_depot_init() (called by page owner
init) will not recognize properly it should use kvmalloc() and not
memblock_alloc(). memblock_alloc() will also not issue a warning and
return a block memory that can be invalid and cause kernel page fault when
saving stacks, as reported by the kernel test robot [1].
Fix this by moving page_ext_init_flatmem_late() below kmem_cache_init() so
that slab_is_available() is true during stack_depot_init(). SPARSEMEM
doesn't have this issue, as it doesn't do page_ext_init_flatmem_late(),
but a different page_ext_init() even later in the boot process.
Thanks to Mike Rapoport for pointing out the FLATMEM init ordering issue.
While at it, also actually resolve a checkpatch warning in stack_depot_init()
from DRM CI, which was supposed to be in the original patch already.
Due to cd06ab2fd48f ("drm/locking: add backtrace for locking contended
locks without backoff") landing recently to -next adding a new stack depot
user in drivers/gpu/drm/drm_modeset_lock.c we need to add an appropriate
call to stack_depot_init() there as well.
Due to 4e66934eaadc ("lib: add reference counting tracking
infrastructure") landing recently to net-next adding a new stack depot
user in lib/ref_tracker.c we need to add an appropriate call to
stack_depot_init() there as well.
Muchun Song [Sat, 22 Jan 2022 06:14:20 +0000 (22:14 -0800)]
fs: proc: store PDE()->data into inode->i_private
PDE_DATA(inode) is introduced to get user private data and hide the
layout of struct proc_dir_entry. The inode->i_private is used to do the
same thing as well. Save a copy of user private data to inode->
i_private when proc inode is allocated. This means the user also can
get their private data by inode->i_private.
Introduce pde_data() to wrap inode->i_private so that we can remove
PDE_DATA() from fs/proc/generic.c and make PTE_DATE() as a wrapper of
pde_data(). It will be easier if we decide to remove PDE_DATE() in the
future.
Mike Galbraith [Sat, 22 Jan 2022 06:14:17 +0000 (22:14 -0800)]
zsmalloc: replace get_cpu_var with local_lock
The usage of get_cpu_var() in zs_map_object() is problematic because it
disables preemption and makes it impossible to acquire any sleeping lock
on PREEMPT_RT such as a spinlock_t.
Replace the get_cpu_var() usage with a local_lock_t which is embedded
struct mapping_area. It ensures that the access the struct is
synchronized against all users on the same CPU.
[minchan: remove the bit_spin_lock part and change the title]
Minchan Kim [Sat, 22 Jan 2022 06:14:13 +0000 (22:14 -0800)]
zsmalloc: replace per zpage lock with pool->migrate_lock
The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage
object alive during several operations. However, it causes the problem
for PREEMPT_RT as well as introducing too complicated.
This patch replaces the bit spin_lock with pool->migrate_lock rwlock.
It could make the code simple as well as zsmalloc work under PREEMPT_RT.
The drawback is the pool->migrate_lock is bigger granuarity than per
zpage lock so the contention would be higher than old when both
IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and
compaction(page/zpage migration) are going in parallel(*, the
migrate_lock is rwlock and IO related functions are all read side lock
so there is no contention). However, the write-side is fast
enough(dominant overhead is just page copy) so it wouldn't affect much.
If the lock granurity becomes more problem later, we could introduce
table locks based on handle as a hash value.
Minchan Kim [Sat, 22 Jan 2022 06:14:10 +0000 (22:14 -0800)]
locking/rwlocks: introduce write_lock_nested
In preparation for converting bit_spin_lock to rwlock in zsmalloc so
that multiple writers of zspages can run at the same time but those
zspages are supposed to be different zspage instance. Thus, it's not
deadlock. This patch adds write_lock_nested to support the case for
LOCKDEP.
Minchan Kim [Sat, 22 Jan 2022 06:14:07 +0000 (22:14 -0800)]
zsmalloc: remove zspage isolation for migration
zspage isolation for migration introduced additional exceptions to be
dealt with since the zspage was isolated from class list. The reason
why I isolated zspage from class list was to prevent race between
obj_malloc and page migration via allocating zpage from the zspage
further. However, it couldn't prevent object freeing from zspage so it
needed corner case handling.
This patch removes the whole mess. Now, we are fine since class->lock
and zspage->lock can prevent the race.
Minchan Kim [Sat, 22 Jan 2022 06:13:57 +0000 (22:13 -0800)]
zsmalloc: decouple class actions from zspage works
This patch moves class stat update out of obj_malloc since it's not
related to zspage operation. This is a preparation to introduce new
lock scheme in next patch.
Minchan Kim [Sat, 22 Jan 2022 06:13:51 +0000 (22:13 -0800)]
zsmalloc: introduce some helper functions
Patch series "zsmalloc: remove bit_spin_lock", v2.
zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage
granularity lock. However, it causes zsmalloc non-working under
PREEMPT_RT as well as adding too much complication.
This patchset tries to replace the bit_spin_lock with per-pool rwlock.
It also removes unnecessary zspage isolation logic from class, which was
the other part too much complication added into zsmalloc.
Last patch changes the get_cpu_var to local_lock to make it work in
PREEMPT_RT.
This patch (of 9):
get_zspage_mapping returns fullness as well as class_idx. However, the
fullness is usually not used since it could be stale in some contexts.
It causes misleading as well as unnecessary instructions so this patch
introduces zspage_class.
obj_to_location also produces page and index but we don't need always
the index, either so this patch introduces obj_to_page.
Xiaoming Ni [Sat, 22 Jan 2022 06:13:41 +0000 (22:13 -0800)]
kprobe: move sysctl_kprobes_optimization to kprobes.c
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
Move sysctl_kprobes_optimization from kernel/sysctl.c to
kernel/kprobes.c. Use register_sysctl() to register the sysctl
interface.
[[email protected]: fix compile issue when CONFIG_OPTPROBES is disabled]
Xiaoming Ni [Sat, 22 Jan 2022 06:13:34 +0000 (22:13 -0800)]
printk: fix build warning when CONFIG_PRINTK=n
build warning when CONFIG_PRINTK=n
kernel/printk/printk.c:175:5: warning: no previous prototype for
'devkmsg_sysctl_set_loglvl' [-Wmissing-prototypes]
devkmsg_sysctl_set_loglvl() is only used in sysctl.c when
CONFIG_PRINTK=y, but it participates in the build when CONFIG_PRINTK=n.
So add compile dependency CONFIG_PRINTK=y && CONFIG_SYSCTL=y to fix the
build warning.
Luis Chamberlain [Sat, 22 Jan 2022 06:13:27 +0000 (22:13 -0800)]
fs: move namespace sysctls and declare fs base directory
This moves the namespace sysctls to its own file as part of the
kernel/sysctl.c spring cleaning
Since we have now removed all sysctls for "fs", we now have to declare
it on the filesystem code, we do that using the new helper, which
reduces boiler plate code.
We rename init_fs_shared_sysctls() to init_fs_sysctls() to reflect that
now fs/sysctls.c is taking on the burden of being the first to register
the base directory as well.
Lastly, since init code will load in the order in which we link it we
have to move the sysctl code to be linked in early, so that its early
init routine runs prior to other fs code. This way, other filesystem
code can register their own sysctls using the helpers after this:
Luis Chamberlain [Sat, 22 Jan 2022 06:13:24 +0000 (22:13 -0800)]
sysctl: add and use base directory declarer and registration helper
Patch series "sysctl: add and use base directory declarer and
registration helper".
In this patch series we start addressing base directories, and so we
start with the "fs" sysctls. The end goal is we end up completely
moving all "fs" sysctl knobs out from kernel/sysctl.
This patch (of 6):
Add a set of helpers which can be used to declare and register base
directory sysctls on their own. We do this so we can later move each of
the base sysctl directories like "fs", "kernel", etc, to their own
respective files instead of shoving the declarations and registrations
all on kernel/sysctl.c. The lazy approach has caught up and with this,
we just end up extending the list of base directories / sysctls on one
file and this makes maintenance difficult due to merge conflicts from
many developers.
The declarations are used first by kernel/sysctl.c for registration its
own base which over time we'll try to clean up. It will be used in the
next patch to demonstrate how to cleanly deal with base sysctl
directories.
Luis Chamberlain [Sat, 22 Jan 2022 06:13:20 +0000 (22:13 -0800)]
fs: move pipe sysctls to is own file
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
Luis Chamberlain [Sat, 22 Jan 2022 06:13:17 +0000 (22:13 -0800)]
fs: move fs/exec.c sysctls into its own file
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the fs/exec.c respective sysctls to its own file.
Since checkpatch complains about style issues with the old code, this
move also fixes a few of those minor style issues:
* Use pr_warn() instead of prink(WARNING
* New empty lines are wanted at the beginning of routines
Luis Chamberlain [Sat, 22 Jan 2022 06:13:13 +0000 (22:13 -0800)]
fs: move namei sysctls to its own file
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move namei's own sysctl knobs to its own file.
Other than the move we also avoid initializing two static variables to 0
as this is not needed:
Luis Chamberlain [Sat, 22 Jan 2022 06:13:10 +0000 (22:13 -0800)]
fs: move locking sysctls where they are used
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
The locking fs sysctls are only used on fs/locks.c, so move them there.
Luis Chamberlain [Sat, 22 Jan 2022 06:13:06 +0000 (22:13 -0800)]
fs: move shared sysctls to fs/sysctls.c
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move sysctls which are shared between filesystems into a common file
outside of kernel/sysctl.c.
Luis Chamberlain [Sat, 22 Jan 2022 06:12:59 +0000 (22:12 -0800)]
fs: move dcache sysctls to its own file
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the dcache sysctl clutter out of kernel/sysctl.c. This is a
small one-off entry, perhaps later we can simplify this representation,
but for now we use the helpers we have. We won't know how we can
simplify this further untl we're fully done with the cleanup.
Luis Chamberlain [Sat, 22 Jan 2022 06:12:56 +0000 (22:12 -0800)]
fs: move fs stat sysctls to file_table.c
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
We can create the sysctl dynamically on early init for fs stat to help
with this clutter. This dusts off the fs stat syctls knobs and puts
them into where they are declared.
Luis Chamberlain [Sat, 22 Jan 2022 06:12:52 +0000 (22:12 -0800)]
fs: move inode sysctls to its own file
Patch series "sysctl: 4th set of kernel/sysctl cleanups".
This is slimming down the fs uses of kernel/sysctl.c to the point that
the next step is to just get rid of the fs base directory for it and
move that elsehwere, so that next patch series starts dealing with that
to demo how we can end up cleaning up a full base directory from
kernel/sysctl.c, one at a time.
This patch (of 9):
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the inode sysctls to its own file. Since we are no longer using
this outside of fs/ remove the extern declaration of its respective proc
helper.
We use early_initcall() as it is the earliest we can use.
Xiaoming Ni [Sat, 22 Jan 2022 06:12:43 +0000 (22:12 -0800)]
stackleak: move stack_erasing sysctl to stackleak.c
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the stack_erasing sysctl from kernel/sysctl.c to
kernel/stackleak.c and use register_sysctl() to register the sysctl
interface.
Xiaoming Ni [Sat, 22 Jan 2022 06:12:38 +0000 (22:12 -0800)]
scsi/sg: move sg-big-buff sysctl to scsi/sg.c
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the sg-big-buff sysctl from kernel/sysctl.c to drivers/scsi/sg.c
and use register_sysctl() to register the sysctl interface.
Xiaoming Ni [Sat, 22 Jan 2022 06:12:33 +0000 (22:12 -0800)]
printk: move printk sysctl to printk/sysctl.c
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move printk sysctl from kernel/sysctl.c to kernel/printk/sysctl.c.
Use register_sysctl() to register the sysctl interface.
[[email protected]: fixed compile issues when PRINTK is not set, commit log update]
Luis Chamberlain [Sat, 22 Jan 2022 06:12:28 +0000 (22:12 -0800)]
fs: move binfmt_misc sysctl to its own file
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
This moves the binfmt_misc sysctl to its own file to help remove clutter
from kernel/sysctl.c.
Luis Chamberlain [Sat, 22 Jan 2022 06:12:23 +0000 (22:12 -0800)]
sysctl: add helper to register a sysctl mount point
The way to create a subdirectory on top of sysctl_mount_point is a bit
obscure, and *why* we do that even so more. Provide a helper which
makes it clear why we do this.
[[email protected]: export register_sysctl_mount_point() to
modules]
Xiaoming Ni [Sat, 22 Jan 2022 06:12:18 +0000 (22:12 -0800)]
random: move the random sysctl declarations to its own file
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the random sysctls to their own file and use
register_sysctl_init().
Xiaoming Ni [Sat, 22 Jan 2022 06:12:13 +0000 (22:12 -0800)]
firmware_loader: move firmware sysctl to its own files
Patch series "sysctl: 3rd set of kernel/sysctl cleanups", v2.
This is the third set of patches to help address cleaning the kitchen
seink in kernel/sysctl.c and to move sysctls away to where they are
actually implemented / used.
This patch (of 8):
kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the firmware configuration sysctl table to the only place where
it is used, and make it clear that if sysctls are disabled this is not
used.
Xiaoming Ni [Sat, 22 Jan 2022 06:12:09 +0000 (22:12 -0800)]
eventpoll: simplify sysctl declaration with register_sysctl()
The kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move the epoll_table sysctl to fs/eventpoll.c and use
register_sysctl().
Luis Chamberlain [Sat, 22 Jan 2022 06:12:04 +0000 (22:12 -0800)]
cdrom: simplify subdirectory registration with register_sysctl()
There is no need to user boiler plate code to specify a set of base
directories we're going to stuff sysctls under. Simplify this by using
register_sysctl() and specifying the directory path directly.
Xiaoming Ni [Sat, 22 Jan 2022 06:11:59 +0000 (22:11 -0800)]
inotify: simplify subdirectory registration with register_sysctl()
There is no need to user boiler plate code to specify a set of base
directories we're going to stuff sysctls under. Simplify this by using
register_sysctl() and specifying the directory path directly.
Move inotify_user sysctl to inotify_user.c while at it to remove clutter
from kernel/sysctl.c.
Luis Chamberlain [Sat, 22 Jan 2022 06:11:54 +0000 (22:11 -0800)]
test_sysctl: simplify subdirectory registration with register_sysctl()
There is no need to user boiler plate code to specify a set of base
directories we're going to stuff sysctls under. Simplify this by using
register_sysctl() and specifying the directory path directly.
Luis Chamberlain [Sat, 22 Jan 2022 06:11:49 +0000 (22:11 -0800)]
ocfs2: simplify subdirectory registration with register_sysctl()
There is no need to user boiler plate code to specify a set of base
directories we're going to stuff sysctls under. Simplify this by using
register_sysctl() and specifying the directory path directly.
Luis Chamberlain [Sat, 22 Jan 2022 06:11:44 +0000 (22:11 -0800)]
macintosh/mac_hid.c: simplify subdirectory registration with register_sysctl()
There is no need to user boiler plate code to specify a set of base
directories we're going to stuff sysctls under. Simplify this by using
register_sysctl() and specifying the directory path directly.
Luis Chamberlain [Sat, 22 Jan 2022 06:11:39 +0000 (22:11 -0800)]
i915: simplify subdirectory registration with register_sysctl()
There is no need to user boiler plate code to specify a set of base
directories we're going to stuff sysctls under. Simplify this by using
register_sysctl() and specifying the directory path directly.
Luis Chamberlain [Sat, 22 Jan 2022 06:11:34 +0000 (22:11 -0800)]
hpet: simplify subdirectory registration with register_sysctl()
Patch series "sysctl: second set of kernel/sysctl cleanups", v2.
This is the 2nd set of kernel/sysctl.c cleanups. The diff stat should
reflect how this is a much better way to deal with theses. Fortunately
coccinelle can be used to ensure correctness for most of these and/or
future merge conflicts.
Note that since this is part of a larger effort to cleanup
kernel/sysctl.c I think we have no other option but to go with merging
these patches in either Andrew's tree or keep them staged in a separate
tree and send a merge request later. Otherwise kernel/sysctl.c will end
up becoming a sore spot for the next merge window.
This patch (of 8):
There is no need to user boiler plate code to specify a set of base
directories we're going to stuff sysctls under. Simplify this by using
register_sysctl() and specifying the directory path directly.
Xiaoming Ni [Sat, 22 Jan 2022 06:11:29 +0000 (22:11 -0800)]
dnotify: move dnotify sysctl to dnotify.c
The kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move dnotify sysctls to dnotify.c and use the new
register_sysctl_init() to register the sysctl interface.
Xiaoming Ni [Sat, 22 Jan 2022 06:11:24 +0000 (22:11 -0800)]
aio: move aio sysctl to aio.c
The kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
Move aio sysctl to aio.c and use the new register_sysctl_init() to
register the sysctl interface for aio.
Xiaoming Ni [Sat, 22 Jan 2022 06:11:19 +0000 (22:11 -0800)]
sysctl: use SYSCTL_ZERO to replace some static int zero uses
Use the variable SYSCTL_ZERO to replace some static int boundary
variables with a value of 0 (minolduid, min_extfrag_threshold,
min_wakeup_granularity_ns).
Xiaoming Ni [Sat, 22 Jan 2022 06:11:14 +0000 (22:11 -0800)]
sysctl: use const for typically used max/min proc sysctls
When proc_dointvec_minmax() or proc_doulongvec_minmax() are used we are
using the extra1 and extra2 parameters on the sysctl table only for a
min and max boundary, these extra1 and extra2 arguments are then used
for read-only operations. So make them const to reflect this.
Xiaoming Ni [Sat, 22 Jan 2022 06:11:05 +0000 (22:11 -0800)]
watchdog: move watchdog sysctl interface to watchdog.c
The kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic of proc sysctl.
So, move the watchdog syscl interface to watchdog.c. Use
register_sysctl() to register the sysctl interface to avoid merge
conflicts when different features modify sysctl.c at the same time.
Xiaoming Ni [Sat, 22 Jan 2022 06:11:00 +0000 (22:11 -0800)]
hung_task: move hung_task sysctl interface to hung_task.c
The kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
So move hung_task sysctl interface to hung_task.c and use
register_sysctl() to register the sysctl interface.
Xiaoming Ni [Sat, 22 Jan 2022 06:10:55 +0000 (22:10 -0800)]
sysctl: move some boundary constants from sysctl.c to sysctl_vals
sysctl has helpers which let us specify boundary values for a min or max
int value. Since these are used for a boundary check only they don't
change, so move these variables to sysctl_vals to avoid adding duplicate
variables. This will help with our cleanup of kernel/sysctl.c.
[[email protected]: update it for "mm/pagealloc: sysctl: change watermark_scale_factor max limit to 30%"]
[[email protected]: major rebase]
Xiaoming Ni [Sat, 22 Jan 2022 06:10:50 +0000 (22:10 -0800)]
sysctl: add a new register_sysctl_init() interface
Patch series "sysctl: first set of kernel/sysctl cleanups", v2.
Finally had time to respin the series of the work we had started last
year on cleaning up the kernel/sysct.c kitchen sink. People keeps
stuffing their sysctls in that file and this creates a maintenance
burden. So this effort is aimed at placing sysctls where they actually
belong.
I'm going to split patches up into series as there is quite a bit of
work.
This first set adds register_sysctl_init() for uses of registerting a
sysctl on the init path, adds const where missing to a few places,
generalizes common values so to be more easy to share, and starts the
move of a few kernel/sysctl.c out where they belong.
The majority of rework on v2 in this first patch set is 0-day fixes.
Eric Biederman's feedback is later addressed in subsequent patch sets.
I'll only post the first two patch sets for now. We can address the
rest once the first two patch sets get completely reviewed / Acked.
This patch (of 9):
The kernel/sysctl.c is a kitchen sink where everyone leaves their dirty
dishes, this makes it very difficult to maintain.
To help with this maintenance let's start by moving sysctls to places
where they actually belong. The proc sysctl maintainers do not want to
know what sysctl knobs you wish to add for your own piece of code, we
just care about the core logic.
Today though folks heavily rely on tables on kernel/sysctl.c so they can
easily just extend this table with their needed sysctls. In order to
help users move their sysctls out we need to provide a helper which can
be used during code initialization.
We special-case the initialization use of register_sysctl() since it
*is* safe to fail, given all that sysctls do is provide a dynamic
interface to query or modify at runtime an existing variable. So the
use case of register_sysctl() on init should *not* stop if the sysctls
don't end up getting registered. It would be counter productive to stop
boot if a simple sysctl registration failed.
Provide a helper for init then, and document the recommended init levels
to use for callers of this routine. We will later use this in
subsequent patches to start slimming down kernel/sysctl.c tables and
moving sysctl registration to the code which actually needs these
sysctls.
[[email protected]: major commit log and documentation rephrasing also moved to fs/proc/proc_sysctl.c ]
Alistair Popple [Sat, 22 Jan 2022 06:10:46 +0000 (22:10 -0800)]
mm/migrate.c: rework migration_entry_wait() to not take a pageref
This fixes the FIXME in migrate_vma_check_page().
Before migrating a page migration code will take a reference and check
there are no unexpected page references, failing the migration if there
are. When a thread faults on a migration entry it will take a temporary
reference to the page to wait for the page to become unlocked signifying
the migration entry has been removed.
This reference is dropped just prior to waiting on the page lock,
however the extra reference can cause migration failures so it is
desirable to avoid taking it.
As migration code already has a reference to the migrating page an extra
reference to wait on PG_locked is unnecessary so long as the reference
can't be dropped whilst setting up the wait.
When faulting on a migration entry the ptl is taken to check the
migration entry. Removing a migration entry also requires the ptl, and
migration code won't drop its page reference until after the migration
entry has been removed. Therefore retaining the ptl of a migration
entry is sufficient to ensure the page has a reference. Reworking
migration_entry_wait() to hold the ptl until the wait setup is complete
means the extra page reference is no longer needed.
Jeffle Xu [Tue, 28 Dec 2021 12:44:19 +0000 (20:44 +0800)]
netfs: Make ops->init_rreq() optional
Make the ops->init_rreq() callback optional. This isn't required for the
erofs changes I'm implementing to do on-demand read through fscache[1].
Further, ceph has an empty init_rreq method that can then be removed and
it's marked optional in the documentation.
David Howells [Fri, 7 Jan 2022 09:28:41 +0000 (09:28 +0000)]
fscache: Add a comment explaining how page-release optimisation works
Add a comment into fscache_note_page_release() to explain how the
page-release optimisation logic works[1]. It's not entirely obvious as it
has nothing to do with whether or not the netfs file contains data.
FSCACHE_COOKIE_NO_DATA_TO_READ is set if we have no data in the cache yet
(ie. the backing file lookup was negative, the file is 0 length or the
cookie got invalidated). It means that we have no data in the cache, not
that the file is necessarily empty on the server.
FSCACHE_COOKIE_HAVE_DATA is set once we've stored data in the backing file.
From that point on, we have data we *could* read - however, it's covered by
pages in the netfs pagecache until at such time one of those covering pages
is released.
So if we've written data to the cache (HAVE_DATA) and there wasn't any data
in the cache when we started (NO_DATA_TO_READ), it may no longer be true
that we can skip reading from the cache.
Read skipping is done by cachefiles_prepare_read().
Note that tracking is not done on a per-page basis, but only on a per-file
basis.
projects / linux.git / log