The V4 filesystem format contains known weaknesses in the on-disk format
that make metadata verification diffiult. In addition, the format does
not support dates past 2038 and will not be upgraded to do so. We
should start the process of retiring the old format to close off attack
surfaces and to encourage users to migrate onto V5.
Therefore, make XFS V4 support a configurable option. For the first
period it will be default Y in case some distributors want to withdraw
support early; for the second period it will be default N so that anyone
who wishes to continue support can do so; and after that, support will
be removed from the kernel. Dates for these events have been added to
the upstream kernel.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Add support for extending the newest data block. For this, introduce
a new finalization state (desc_finalized) denoting a committed
descriptor that cannot be extended.
Until a record is finalized, a writer can reopen that record to
append new data. Reopening a record means transitioning from the
desc_committed state back to the desc_reserved state.
A writer can explicitly finalize a record if there is no intention
of extending it. Also, records are automatically finalized when a
new record is reserved. This relieves writers of needing to
explicitly finalize while also making such records available to
readers sooner. (Readers can only traverse finalized records.)
Four new memory barrier pairs are introduced. Two of them are
insignificant additions (data_realloc:A/desc_read:D and
data_realloc:A/data_push_tail:B) because they are alternate path
memory barriers that exactly match the purpose, pairing, and
context of the two existing memory barrier pairs they provide an
alternate path for. The other two new memory barrier pairs are
significant additions:
desc_reopen_last:A / _prb_commit:B - When reopening a descriptor,
ensure the state transitions back to desc_reserved before
fully trusting the descriptor data.
_prb_commit:B / desc_reserve:D - When committing a descriptor,
ensure the state transitions to desc_committed before checking
the head ID to see if the descriptor needs to be finalized.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20200914123354.832-6-john.ogness@linutronix.de
Rather than deriving the state by evaluating bits within the flags
area of the state variable, assign the states explicit values and
set those values in the flags area. Introduce macros to make it
simple to read and write state values for the state variable.
Although the functionality is preserved, the binary representation
for the states is changed.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20200914123354.832-5-john.ogness@linutronix.de
Pull char / misc driver fixes from Greg KH:
"Here are a number of small driver fixes for 5.9-rc5
Included in here are:
- habanalabs driver fixes
- interconnect driver fixes
- soundwire driver fixes
- dyndbg fixes for reported issues, and then reverts to fix it all up
to a sane state.
- phy driver fixes
All of these have been in linux-next for a while with no reported
issues"
* tag 'char-misc-5.9-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
Revert "dyndbg: accept query terms like file=bar and module=foo"
Revert "dyndbg: fix problem parsing format="foo bar""
scripts/tags.sh: exclude tools directory from tags generation
video: fbdev: fix OOB read in vga_8planes_imageblit()
dyndbg: fix problem parsing format="foo bar"
dyndbg: refine export, rename to dynamic_debug_exec_queries()
dyndbg: give %3u width in pr-format, cosmetic only
interconnect: qcom: Fix small BW votes being truncated to zero
soundwire: fix double free of dangling pointer
interconnect: Show bandwidth for disabled paths as zero in debugfs
habanalabs: fix report of RAZWI initiator coordinates
habanalabs: prevent user buff overflow
phy: omap-usb2-phy: disable PHY charger detect
phy: qcom-qmp: Use correct values for ipq8074 PCIe Gen2 PHY init
soundwire: bus: fix typo in comment on INTSTAT registers
phy: qualcomm: fix return value check in qcom_ipq806x_usb_phy_probe()
phy: qualcomm: fix platform_no_drv_owner.cocci warnings
Commit
0c2a3913d6 ("x86/fpu: Parse clearcpuid= as early XSAVE argument")
changed clearcpuid parsing from __setup() to cmdline_find_option().
While the __setup() function would have been called for each clearcpuid=
parameter on the command line, cmdline_find_option() will only return
the last one, so the change effectively made it impossible to disable
more than one bit.
Allow a comma-separated list of bit numbers as the argument for
clearcpuid to allow multiple bits to be disabled again. Log the bits
being disabled for informational purposes.
Also fix the check on the return value of cmdline_find_option(). It
returns -1 when the option is not found, so testing as a boolean is
incorrect.
Fixes: 0c2a3913d6 ("x86/fpu: Parse clearcpuid= as early XSAVE argument")
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20200907213919.2423441-1-nivedita@alum.mit.edu
To quote the TODO of this driver:
--------------------------------------------------------------
The driver is deprecated and scheduled for removal by the end
of 2020.
In order to prevent removal the following actions would have to
be taken:
- clean up the code
- convert to the vb2 framework
- fix the disconnect and free-on-last-user handling (i.e., add
a release callback for struct v4l2_device and rework the code
to use that correctly).
--------------------------------------------------------------
Nobody picked this up, so it's time to retire this driver.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Linux 5.9-rc4
* tag 'v5.9-rc4': (1001 commits)
Linux 5.9-rc4
io_uring: fix linked deferred ->files cancellation
io_uring: fix cancel of deferred reqs with ->files
include/linux/log2.h: add missing () around n in roundup_pow_of_two()
mm/khugepaged.c: fix khugepaged's request size in collapse_file
mm/hugetlb: fix a race between hugetlb sysctl handlers
mm/hugetlb: try preferred node first when alloc gigantic page from cma
mm/migrate: preserve soft dirty in remove_migration_pte()
mm/migrate: remove unnecessary is_zone_device_page() check
mm/rmap: fixup copying of soft dirty and uffd ptes
mm/migrate: fixup setting UFFD_WP flag
mm: madvise: fix vma user-after-free
checkpatch: fix the usage of capture group ( ... )
fork: adjust sysctl_max_threads definition to match prototype
ipc: adjust proc_ipc_sem_dointvec definition to match prototype
mm: track page table modifications in __apply_to_page_range()
MAINTAINERS: IA64: mark Status as Odd Fixes only
MAINTAINERS: add LLVM maintainers
MAINTAINERS: update Cavium/Marvell entries
mm: slub: fix conversion of freelist_corrupted()
...
We got slightly different patches removing a double word
in a comment in net/ipv4/raw.c - picked the version from net.
Simple conflict in drivers/net/ethernet/ibm/ibmvnic.c. Use cached
values instead of VNIC login response buffer (following what
commit 507ebe6444 ("ibmvnic: Fix use-after-free of VNIC login
response buffer") did).
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull power management fixes from Rafael Wysocki:
"These fix reference counting in the operating performance points (OPP)
framework and address a few intel_pstate driver issues, mostly related
to switching driver operation modes and similar with hardware-managed
P-states (HWP) enabled.
Specifics:
- Fix reference counting of operating performance points (OPP) tables
(Viresh Kumar).
- Address intel_pstate driver interface issues, mostly related to
switching operation modes and handling CPU offline and online and
system-wide suspend/resume with hardware-managed P-states (HWP)
enabled (Rafael Wysocki).
- Fix the maximum frequency computation in the intel_pstate driver
with turbo P-states disabled by the platform firmware and HWP
enabled (Francisco Jerez)"
* tag 'pm-5.9-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
cpufreq: intel_pstate: Fix intel_pstate_get_hwp_max() for turbo disabled
cpufreq: intel_pstate: Free memory only when turning off
cpufreq: intel_pstate: Add ->offline and ->online callbacks
cpufreq: intel_pstate: Tweak the EPP sysfs interface
cpufreq: intel_pstate: Update cached EPP in the active mode
cpufreq: intel_pstate: Refuse to turn off with HWP enabled
opp: Don't drop reference for an OPP table that was never parsed
Modify the EPP sysfs interface to reject attempts to change the EPP
to values different from 0 ("performance") in the active mode with
the "performance" policy (ie. scaling_governor set to "performance"),
to avoid situations in which the kernel appears to discard data
passed to it via the EPP sysfs attribute.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Right now, drivers like ARM SMMU are using dma_alloc_coherent() to get
coherent DMA buffers to save their command queues and page tables. As
there is only one default CMA in the whole system, SMMUs on nodes other
than node0 will get remote memory. This leads to significant latency.
This patch provides per-numa CMA so that drivers like SMMU can get local
memory. Tests show localizing CMA can decrease dma_unmap latency much.
For instance, before this patch, SMMU on node2 has to wait for more than
560ns for the completion of CMD_SYNC in an empty command queue; with this
patch, it needs 240ns only.
A positive side effect of this patch would be improving performance even
further for those users who are worried about performance more than DMA
security and use iommu.passthrough=1 to skip IOMMU. With local CMA, all
drivers can get local coherent DMA buffers.
Also, this patch changes the default CONFIG_CMA_AREAS to 19 in NUMA. As
1+CONFIG_CMA_AREAS should be quite enough for most servers on the market
even they enable both hugetlb_cma and pernuma_cma.
2 numa nodes: 2(hugetlb) + 2(pernuma) + 1(default global cma) = 5
4 numa nodes: 4(hugetlb) + 4(pernuma) + 1(default global cma) = 9
8 numa nodes: 8(hugetlb) + 8(pernuma) + 1(default global cma) = 17
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
* pm-cpufreq:
cpufreq: Use WARN_ON_ONCE() for invalid relation
cpufreq: No need to verify cpufreq_driver in show_scaling_cur_freq()
Documentation: fix pm/intel_pstate build warning and wording
cpufreq: replace cpu_logical_map() with read_cpuid_mpir()
The sysctl that was added earlier by commit 79134e6ce2 ("net: do
not create fallback tunnels for non-default namespaces") to create
fall-back only in root-ns. This patch enhances that behavior to provide
option not to create fallback tunnels in root-ns as well. Since modules
that create fallback tunnels could be built-in and setting the sysctl
value after booting is pointless, so added a kernel cmdline options to
change this default. The default setting is preseved for backward
compatibility. The kernel command line option of fb_tunnels=initns will
set the sysctl value to 1 and will create fallback tunnels only in initns
while kernel cmdline fb_tunnels=none will set the sysctl value to 2 and
fallback tunnels are skipped in every netns.
Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Maciej Zenczykowski <maze@google.com>
Cc: Jian Yang <jianyang@google.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit adds an rcutorture.leakpointer module parameter that
intentionally leaks an RCU-protected pointer out of the RCU read-side
critical section and checks to see if the corresponding grace period
has elapsed, emitting a WARN_ON_ONCE() if so. This module parameter can
be used to test facilities like CONFIG_RCU_STRICT_GRACE_PERIOD that end
grace periods quickly.
While in the area, also document rcutorture.irqreader, which was
previously left out.
Reported-by Jann Horn <jannh@google.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The goal of this series is to increase the probability of tools like
KASAN detecting that an RCU-protected pointer was used outside of its
RCU read-side critical section. Thus far, the approach has been to make
grace periods and callback processing happen faster. Another approach
is to delay the pointer leaker. This commit therefore allows a delay
to be applied to exit from RCU read-side critical sections.
This slowdown is specified by a new rcutree.rcu_unlock_delay kernel boot
parameter that specifies this delay in microseconds, defaulting to zero.
Reported-by Jann Horn <jannh@google.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit further avoids conflation of rcuperf with the kernel's perf
feature by renaming kernel/rcu/rcuperf.c to kernel/rcu/rcuscale.c, and
also by similarly renaming the functions and variables inside this file.
This has the side effect of changing the names of the kernel boot
parameters, so kernel-parameters.txt and ver_functions.sh are also
updated. The rcutorture --torture type was also updated from rcuperf
to rcuscale.
[ paulmck: Fix bugs located by Stephen Rothwell. ]
Reported-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
This commit adds an smp_call_function() torture test that repeatedly
invokes this function and complains if things go badly awry.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Currently, the nmi_cpu_backtrace() declines to produce backtraces for
idle CPUs. This is a good choice in the common case in which problems are
caused only by non-idle CPUs. However, there are occasionally situations
in which idle CPUs are helping to cause problems. This commit therefore
adds an nmi_backtrace.backtrace_idle kernel boot parameter that causes
nmi_cpu_backtrace() to dump stacks even of idle CPUs.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: <linux-doc@vger.kernel.org>
Pull EFI fixes from Thomas Gleixner:
- Enforce NX on RO data in mixed EFI mode
- Destroy workqueue in an error handling path to prevent UAF
- Stop argument parser at '--' which is the delimiter for init
- Treat a NULL command line pointer as empty instead of dereferncing it
unconditionally.
- Handle an unterminated command line correctly
- Cleanup the 32bit code leftovers and remove obsolete documentation
* tag 'efi-urgent-2020-08-23' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
Documentation: efi: remove description of efi=old_map
efi/x86: Move 32-bit code into efi_32.c
efi/libstub: Handle unterminated cmdline
efi/libstub: Handle NULL cmdline
efi/libstub: Stop parsing arguments at "--"
efi: add missed destroy_workqueue when efisubsys_init fails
efi/x86: Mark kernel rodata non-executable for mixed mode
Pull ext4 updates from Ted Ts'o:
"Improvements to ext4's block allocator performance for very large file
systems, especially when the file system or files which are highly
fragmented. There is a new mount option, prefetch_block_bitmaps which
will pull in the block bitmaps and set up the in-memory buddy bitmaps
when the file system is initially mounted.
Beyond that, a lot of bug fixes and cleanups. In particular, a number
of changes to make ext4 more robust in the face of write errors or
file system corruptions"
* tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (46 commits)
ext4: limit the length of per-inode prealloc list
ext4: reorganize if statement of ext4_mb_release_context()
ext4: add mb_debug logging when there are lost chunks
ext4: Fix comment typo "the the".
jbd2: clean up checksum verification in do_one_pass()
ext4: change to use fallthrough macro
ext4: remove unused parameter of ext4_generic_delete_entry function
mballoc: replace seq_printf with seq_puts
ext4: optimize the implementation of ext4_mb_good_group()
ext4: delete invalid comments near ext4_mb_check_limits()
ext4: fix typos in ext4_mb_regular_allocator() comment
ext4: fix checking of directory entry validity for inline directories
fs: prevent BUG_ON in submit_bh_wbc()
ext4: correctly restore system zone info when remount fails
ext4: handle add_system_zone() failure in ext4_setup_system_zone()
ext4: fold ext4_data_block_valid_rcu() into the caller
ext4: check journal inode extents more carefully
ext4: don't allow overlapping system zones
ext4: handle error of ext4_setup_system_zone() on remount
ext4: delete the invalid BUGON in ext4_mb_load_buddy_gfp()
...
The old EFI runtime region mapping logic that was kept around for some
time has finally been removed entirely, along with the SGI UV1 support
code that was its last remaining user. So remove any mention of the
efi=old_map command line parameter from the docs.
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-doc@vger.kernel.org
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
In the scenario of writing sparse files, the per-inode prealloc list may
be very long, resulting in high overhead for ext4_mb_use_preallocated().
To circumvent this problem, we limit the maximum length of per-inode
prealloc list to 512 and allow users to modify it.
After patching, we observed that the sys ratio of cpu has dropped, and
the system throughput has increased significantly. We created a process
to write the sparse file, and the running time of the process on the
fixed kernel was significantly reduced, as follows:
Running time on unfixed kernel：
[root@TENCENT64 ~]# time taskset 0x01 ./sparse /data1/sparce.dat
real 0m2.051s
user 0m0.008s
sys 0m2.026s
Running time on fixed kernel：
[root@TENCENT64 ~]# time taskset 0x01 ./sparse /data1/sparce.dat
real 0m0.471s
user 0m0.004s
sys 0m0.395s
Signed-off-by: Chunguang Xu <brookxu@tencent.com>
Link: https://lore.kernel.org/r/d7a98178-056b-6db5-6bce-4ead23f4a257@gmail.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Pull x86 fixes from Ingo Molnar:
"Misc fixes and small updates all around the place:
- Fix mitigation state sysfs output
- Fix an FPU xstate/sxave code assumption bug triggered by
Architectural LBR support
- Fix Lightning Mountain SoC TSC frequency enumeration bug
- Fix kexec debug output
- Fix kexec memory range assumption bug
- Fix a boundary condition in the crash kernel code
- Optimize porgatory.ro generation a bit
- Enable ACRN guests to use X2APIC mode
- Reduce a __text_poke() IRQs-off critical section for the benefit of
PREEMPT_RT"
* tag 'x86-urgent-2020-08-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/alternatives: Acquire pte lock with interrupts enabled
x86/bugs/multihit: Fix mitigation reporting when VMX is not in use
x86/fpu/xstate: Fix an xstate size check warning with architectural LBRs
x86/purgatory: Don't generate debug info for purgatory.ro
x86/tsr: Fix tsc frequency enumeration bug on Lightning Mountain SoC
kexec_file: Correctly output debugging information for the PT_LOAD ELF header
kexec: Improve & fix crash_exclude_mem_range() to handle overlapping ranges
x86/crash: Correct the address boundary of function parameters
x86/acrn: Remove redundant chars from ACRN signature
x86/acrn: Allow ACRN guest to use X2APIC mode
Pull one more power management update from Rafael Wysocki:
"Modify the intel_pstate driver to allow it to work in the passive mode
with hardware-managed P-states (HWP) enabled"
* tag 'pm-5.9-rc1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
cpufreq: intel_pstate: Implement passive mode with HWP enabled
Pull documentation fixes from Jonathan Corbet:
"A handful of obvious fixes that wandered in during the merge window"
* tag 'docs-5.9-2' of git://git.lwn.net/linux:
Documentation/locking/locktypes: fix the typo
doc/zh_CN: resolve undefined label warning in admin-guide index
doc/zh_CN: fix title heading markup in admin-guide cpu-load
docs: remove the 2.6 "Upgrading I2C Drivers" guide
docs: Correct the release date of 5.2 stable
mailmap: Update comments for with format and more detalis
docs: cdrom: Fix a typo and rst markup
Doc: admin-guide: use correct legends in kernel-parameters.txt
Documentation/features: refresh RISC-V arch support files
documentation: coccinelle: Improve command example for make C={1,2}
Core-api: Documentation: Replace deprecated :c:func: Usage
Dev-tools: Documentation: Replace deprecated :c:func: Usage
Filesystems: Documentation: Replace deprecated :c:func: Usage
docs: trace: fix a typo
The document reads "%e" should be "executable filename" while actually it
could be changed by things like pr_ctl PR_SET_NAME. People who uses "%e"
in core_pattern get surprised when they find out they get thread name
instead of executable filename.
This is either a bug of document or a bug of code. Since the behavior of
"%e" is there for long time, it could bring another surprise for users if
we "fix" the code.
So we just "fix" the document. And more, for users who really need the
"executable filename" in core_pattern, we introduce a new "%f" for the
real executable filename. We already have "%E" for executable path in
kernel, so just reuse most of its code for the new added "%f" format.
Signed-off-by: Lepton Wu <ytht.net@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/20200701031432.2978761-1-ytht.net@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For some applications, we need to allocate almost all memory as hugepages.
However, on a running system, higher-order allocations can fail if the
memory is fragmented. Linux kernel currently does on-demand compaction as
we request more hugepages, but this style of compaction incurs very high
latency. Experiments with one-time full memory compaction (followed by
hugepage allocations) show that kernel is able to restore a highly
fragmented memory state to a fairly compacted memory state within <1 sec
for a 32G system. Such data suggests that a more proactive compaction can
help us allocate a large fraction of memory as hugepages keeping
allocation latencies low.
For a more proactive compaction, the approach taken here is to define a
new sysctl called 'vm.compaction_proactiveness' which dictates bounds for
external fragmentation which kcompactd tries to maintain.
The tunable takes a value in range [0, 100], with a default of 20.
Note that a previous version of this patch [1] was found to introduce too
many tunables (per-order extfrag{low, high}), but this one reduces them to
just one sysctl. Also, the new tunable is an opaque value instead of
asking for specific bounds of "external fragmentation", which would have
been difficult to estimate. The internal interpretation of this opaque
value allows for future fine-tuning.
Currently, we use a simple translation from this tunable to [low, high]
"fragmentation score" thresholds (low=100-proactiveness, high=low+10%).
The score for a node is defined as weighted mean of per-zone external
fragmentation. A zone's present_pages determines its weight.
To periodically check per-node score, we reuse per-node kcompactd threads,
which are woken up every 500 milliseconds to check the same. If a node's
score exceeds its high threshold (as derived from user-provided
proactiveness value), proactive compaction is started until its score
reaches its low threshold value. By default, proactiveness is set to 20,
which implies threshold values of low=80 and high=90.
This patch is largely based on ideas from Michal Hocko [2]. See also the
LWN article [3].
Performance data
================
System: x64_64, 1T RAM, 80 CPU threads.
Kernel: 5.6.0-rc3 + this patch
echo madvise | sudo tee /sys/kernel/mm/transparent_hugepage/enabled
echo madvise | sudo tee /sys/kernel/mm/transparent_hugepage/defrag
Before starting the driver, the system was fragmented from a userspace
program that allocates all memory and then for each 2M aligned section,
frees 3/4 of base pages using munmap. The workload is mainly anonymous
userspace pages, which are easy to move around. I intentionally avoided
unmovable pages in this test to see how much latency we incur when
hugepage allocations hit direct compaction.
1. Kernel hugepage allocation latencies
With the system in such a fragmented state, a kernel driver then allocates
as many hugepages as possible and measures allocation latency:
(all latency values are in microseconds)
- With vanilla 5.6.0-rc3
percentile latency
–––––––––– –––––––
5 7894
10 9496
25 12561
30 15295
40 18244
50 21229
60 27556
75 30147
80 31047
90 32859
95 33799
Total 2M hugepages allocated = 383859 (749G worth of hugepages out of 762G
total free => 98% of free memory could be allocated as hugepages)
- With 5.6.0-rc3 + this patch, with proactiveness=20
sysctl -w vm.compaction_proactiveness=20
percentile latency
–––––––––– –––––––
5 2
10 2
25 3
30 3
40 3
50 4
60 4
75 4
80 4
90 5
95 429
Total 2M hugepages allocated = 384105 (750G worth of hugepages out of 762G
total free => 98% of free memory could be allocated as hugepages)
2. JAVA heap allocation
In this test, we first fragment memory using the same method as for (1).
Then, we start a Java process with a heap size set to 700G and request the
heap to be allocated with THP hugepages. We also set THP to madvise to
allow hugepage backing of this heap.
/usr/bin/time
java -Xms700G -Xmx700G -XX:+UseTransparentHugePages -XX:+AlwaysPreTouch
The above command allocates 700G of Java heap using hugepages.
- With vanilla 5.6.0-rc3
17.39user 1666.48system 27:37.89elapsed
- With 5.6.0-rc3 + this patch, with proactiveness=20
8.35user 194.58system 3:19.62elapsed
Elapsed time remains around 3:15, as proactiveness is further increased.
Note that proactive compaction happens throughout the runtime of these
workloads. The situation of one-time compaction, sufficient to supply
hugepages for following allocation stream, can probably happen for more
extreme proactiveness values, like 80 or 90.
In the above Java workload, proactiveness is set to 20. The test starts
with a node's score of 80 or higher, depending on the delay between the
fragmentation step and starting the benchmark, which gives more-or-less
time for the initial round of compaction. As t he benchmark consumes
hugepages, node's score quickly rises above the high threshold (90) and
proactive compaction starts again, which brings down the score to the low
threshold level (80). Repeat.
bpftrace also confirms proactive compaction running 20+ times during the
runtime of this Java benchmark. kcompactd threads consume 100% of one of
the CPUs while it tries to bring a node's score within thresholds.
Backoff behavior
================
Above workloads produce a memory state which is easy to compact. However,
if memory is filled with unmovable pages, proactive compaction should
essentially back off. To test this aspect:
- Created a kernel driver that allocates almost all memory as hugepages
followed by freeing first 3/4 of each hugepage.
- Set proactiveness=40
- Note that proactive_compact_node() is deferred maximum number of times
with HPAGE_FRAG_CHECK_INTERVAL_MSEC of wait between each check
(=> ~30 seconds between retries).
[1] https://patchwork.kernel.org/patch/11098289/
[2] https://lore.kernel.org/linux-mm/20161230131412.GI13301@dhcp22.suse.cz/
[3] https://lwn.net/Articles/817905/
Signed-off-by: Nitin Gupta <nigupta@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Oleksandr Natalenko <oleksandr@redhat.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Khalid Aziz <khalid.aziz@oracle.com>
Reviewed-by: Oleksandr Natalenko <oleksandr@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Khalid Aziz <khalid.aziz@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Nitin Gupta <ngupta@nitingupta.dev>
Cc: Oleksandr Natalenko <oleksandr@redhat.com>
Link: http://lkml.kernel.org/r/20200616204527.19185-1-nigupta@nvidia.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Allow intel_pstate to work in the passive mode with HWP enabled and
make it set the HWP minimum performance limit (HWP floor) to the
P-state value given by the target frequency supplied by the cpufreq
governor, so as to prevent the HWP algorithm and the CPU scheduler
from working against each other, at least when the schedutil governor
is in use, and update the intel_pstate documentation accordingly.
Among other things, this allows utilization clamps to be taken
into account, at least to a certain extent, when intel_pstate is
in use and makes it more likely that sufficient capacity for
deadline tasks will be provided.
After this change, the resulting behavior of an HWP system with
intel_pstate in the passive mode should be close to the behavior
of the analogous non-HWP system with intel_pstate in the passive
mode, except that the HWP algorithm is generally allowed to make the
CPU run at a frequency above the floor P-state set by intel_pstate in
the entire available range of P-states, while without HWP a CPU can
run in a P-state above the requested one if the latter falls into the
range of turbo P-states (referred to as the turbo range) or if the
P-states of all CPUs in one package are coordinated with each other
at the hardware level.
[Note that in principle the HWP floor may not be taken into account
by the processor if it falls into the turbo range, in which case the
processor has a license to choose any P-state, either below or above
the HWP floor, just like a non-HWP processor in the case when the
target P-state falls into the turbo range.]
With this change applied, intel_pstate in the passive mode assumes
complete control over the HWP request MSR and concurrent changes of
that MSR (eg. via the direct MSR access interface) are overridden by
it.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Pull tracing updates from Steven Rostedt:
- The biggest news in that the tracing ring buffer can now time events
that interrupted other ring buffer events.
Before this change, if an interrupt came in while recording another
event, and that interrupt also had an event, those events would all
have the same time stamp as the event it interrupted.
Now, with the new design, those events will have a unique time stamp
and rightfully display the time for those events that were recorded
while interrupting another event.
- Bootconfig how has an "override" operator that lets the users have a
default config, but then add options to override the default.
- A fix was made to properly filter function graph tracing to the
ftrace PIDs. This came in at the end of the -rc cycle, and needs to
be backported.
- Several clean ups, performance updates, and minor fixes as well.
* tag 'trace-v5.9' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace: (39 commits)
tracing: Add trace_array_init_printk() to initialize instance trace_printk() buffers
kprobes: Fix compiler warning for !CONFIG_KPROBES_ON_FTRACE
tracing: Use trace_sched_process_free() instead of exit() for pid tracing
bootconfig: Fix to find the initargs correctly
Documentation: bootconfig: Add bootconfig override operator
tools/bootconfig: Add testcases for value override operator
lib/bootconfig: Add override operator support
kprobes: Remove show_registers() function prototype
tracing/uprobe: Remove dead code in trace_uprobe_register()
kprobes: Fix NULL pointer dereference at kprobe_ftrace_handler
ftrace: Fix ftrace_trace_task return value
tracepoint: Use __used attribute definitions from compiler_attributes.h
tracepoint: Mark __tracepoint_string's __used
trace : Have tracing buffer info use kvzalloc instead of kzalloc
tracing: Remove outdated comment in stack handling
ftrace: Do not let direct or IPMODIFY ftrace_ops be added to module and set trampolines
ftrace: Setup correct FTRACE_FL_REGS flags for module
tracing/hwlat: Honor the tracing_cpumask
tracing/hwlat: Drop the duplicate assignment in start_kthread()
tracing: Save one trace_event->type by using __TRACE_LAST_TYPE
...
Pull device mapper updates from Mike Snitzer:
- DM multipath locking fixes around m->flags tests and improvements to
bio-based code so that it follows patterns established by
request-based code.
- Request-based DM core improvement to eliminate unnecessary call to
blk_mq_queue_stopped().
- Add "panic_on_corruption" error handling mode to DM verity target.
- DM bufio fix to to perform buffer cleanup from a workqueue rather
than wait for IO in reclaim context from shrinker.
- DM crypt improvement to optionally avoid async processing via
workqueues for reads and/or writes -- via "no_read_workqueue" and
"no_write_workqueue" features. This more direct IO processing
improves latency and throughput with faster storage. Avoiding
workqueue IO submission for writes (DM_CRYPT_NO_WRITE_WORKQUEUE) is a
requirement for adding zoned block device support to DM crypt.
- Add zoned block device support to DM crypt. Makes use of
DM_CRYPT_NO_WRITE_WORKQUEUE and a new optional feature
(DM_CRYPT_WRITE_INLINE) that allows write completion to wait for
encryption to complete. This allows write ordering to be preserved,
which is needed for zoned block devices.
- Fix DM ebs target's check for REQ_OP_FLUSH.
- Fix DM core's report zones support to not report more zones than were
requested.
- A few small compiler warning fixes.
- DM dust improvements to return output directly to the user rather
than require they scrape the system log for output.
* tag 'for-5.9/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
dm: don't call report zones for more than the user requested
dm ebs: Fix incorrect checking for REQ_OP_FLUSH
dm init: Set file local variable static
dm ioctl: Fix compilation warning
dm raid: Remove empty if statement
dm verity: Fix compilation warning
dm crypt: Enable zoned block device support
dm crypt: add flags to optionally bypass kcryptd workqueues
dm bufio: do buffer cleanup from a workqueue
dm rq: don't call blk_mq_queue_stopped() in dm_stop_queue()
dm dust: add interface to list all badblocks
dm dust: report some message results directly back to user
dm verity: add "panic_on_corruption" error handling mode
dm mpath: use double checked locking in fast path
dm mpath: rename current_pgpath to pgpath in multipath_prepare_ioctl
dm mpath: rework __map_bio()
dm mpath: factor out multipath_queue_bio
dm mpath: push locking down to must_push_back_rq()
dm mpath: take m->lock spinlock when testing QUEUE_IF_NO_PATH
dm mpath: changes from initial m->flags locking audit
Pull media updates from Mauro Carvalho Chehab:
- Legacy soc_camera driver was removed from staging
- New I2C sensor related drivers: dw9768, ch7322, max9271, rdacm20
- TI vpe driver code was re-organized and had new features added
- Added Xilinx MIPI CSI-2 Rx Subsystem driver
- Added support for Infrared Toy and IR Droid devices
- Lots of random driver fixes, new features and cleanups
* tag 'media/v5.9-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media: (318 commits)
media: camss: fix memory leaks on error handling paths in probe
media: davinci: vpif_capture: fix potential double free
media: radio: remove redundant assignment to variable retval
media: allegro: fix potential null dereference on header
media: mtk-mdp: Fix a refcounting bug on error in init
media: allegro: fix an error pointer vs NULL check
media: meye: fix missing pm_mchip_mode field
media: cafe-driver: use generic power management
media: saa7164: use generic power management
media: v4l2-dev/ioctl: Fix document for VIDIOC_QUERYCAP
media: v4l2: Correct kernel-doc inconsistency
media: v4l2: Correct kernel-doc inconsistency
media: dvbdev.h: keep * together with the type
media: v4l2-subdev.h: keep * together with the type
media: videobuf2: Print videobuf2 buffer state by name
media: colorspaces-details.rst: fix V4L2_COLORSPACE_JPEG description
media: tw68: use generic power management
media: meye: use generic power management
media: cx88: use generic power management
media: cx25821: use generic power management
...
Merge misc updates from Andrew Morton:
- a few MM hotfixes
- kthread, tools, scripts, ntfs and ocfs2
- some of MM
Subsystems affected by this patch series: kthread, tools, scripts, ntfs,
ocfs2 and mm (hofixes, pagealloc, slab-generic, slab, slub, kcsan,
debug, pagecache, gup, swap, shmem, memcg, pagemap, mremap, mincore,
sparsemem, vmalloc, kasan, pagealloc, hugetlb and vmscan).
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (162 commits)
mm: vmscan: consistent update to pgrefill
mm/vmscan.c: fix typo
khugepaged: khugepaged_test_exit() check mmget_still_valid()
khugepaged: retract_page_tables() remember to test exit
khugepaged: collapse_pte_mapped_thp() protect the pmd lock
khugepaged: collapse_pte_mapped_thp() flush the right range
mm/hugetlb: fix calculation of adjust_range_if_pmd_sharing_possible
mm: thp: replace HTTP links with HTTPS ones
mm/page_alloc: fix memalloc_nocma_{save/restore} APIs
mm/page_alloc.c: skip setting nodemask when we are in interrupt
mm/page_alloc: fallbacks at most has 3 elements
mm/page_alloc: silence a KASAN false positive
mm/page_alloc.c: remove unnecessary end_bitidx for [set|get]_pfnblock_flags_mask()
mm/page_alloc.c: simplify pageblock bitmap access
mm/page_alloc.c: extract the common part in pfn_to_bitidx()
mm/page_alloc.c: replace the definition of NR_MIGRATETYPE_BITS with PB_migratetype_bits
mm/shuffle: remove dynamic reconfiguration
mm/memory_hotplug: document why shuffle_zone() is relevant
mm/page_alloc: remove nr_free_pagecache_pages()
mm: remove vm_total_pages
...
