| Age | Commit message (Collapse) | Author |
|---|
|
Clang incorrectly inlines functions with differing stack protector
attributes, which breaks __restore_processor_state() that relies on
stack protector being disabled. This change disables LTO for cpu.c
to work aroung the bug.
Link: https://bugs.llvm.org/show_bug.cgi?id=47479
Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
|
|
Disable LTO for the vDSO. Note that while we could use Clang's LTO
for the 64-bit vDSO, it won't add noticeable benefit for the small
amount of C code.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
|
|
With LTO, LLVM bitcode won't be compiled into native code until
modpost_link, or modfinal for modules. This change postpones calls
to objtool until after these steps, and moves objtool_args to
Makefile.lib, so the arguments can be reused in Makefile.modfinal.
As we didn't have objects to process earlier, we use --duplicate
when processing vmlinux.o. This change also disables unreachable
instruction warnings with LTO to avoid warnings about the int3
padding between functions.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
|
|
This change adds a --noinstr flag to objtool to allow us to specify
that we're processing vmlinux.o without also enabling noinstr
validation. This is needed to avoid false positives with LTO when we
run objtool on vmlinux.o without CONFIG_DEBUG_ENTRY.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
|
|
Select HAVE_OBJTOOL_MCOUNT if STACK_VALIDATION is selected to use
objtool to generate __mcount_loc sections for dynamic ftrace with
Clang and gcc <5 (later versions of gcc use -mrecord-mcount).
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
|
|
This change adds build support for using objtool to generate
__mcount_loc sections.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
|
|
With LTO, we run objtool on vmlinux.o, but don't want noinstr
validation. This change requires --vmlinux to be passed to objtool
explicitly.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
|
|
When objtool generates relocations for the __mcount_loc section, it
tries to reference __fentry__ calls by their section symbol offset.
However, this fails with Clang's integrated assembler as it may not
generate section symbols for every section. This patch looks up a
function symbol instead if the section symbol is missing, similarly
to commit e81e07244325 ("objtool: Support Clang non-section symbols
in ORC generation").
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
|
|
Add the --mcount option for generating __mcount_loc sections
needed for dynamic ftrace. Using this pass requires the kernel to
be compiled with -mfentry and CC_USING_NOP_MCOUNT to be defined
in Makefile.
Link: https://lore.kernel.org/lkml/20200625200235.GQ4781@hirez.programming.kicks-ass.net/
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
[Sami: rebased, dropped config changes, fixed to actually use --mcount,
and wrote a commit message.]
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/dennis/percpu
Pull percpu updates from Dennis Zhou:
"Percpu had a cleanup come in that makes use of the cpu bitmask helpers
instead of the current iterative approach.
This clean up then had an adverse interaction when clang's inlining
sensitivity is changed such that not all sites are inlined resulting
in modpost being upset with section mismatch due to percpu setup being
marked __init.
That was fixed by introducing __flatten to compiler_attributes.h"
* 'for-5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/dennis/percpu:
percpu: fix clang modpost section mismatch
percpu: reduce the number of cpu distance comparisons
|
|
The removal of EXPORT_UNUSED_SYMBOL() in commit 367948220fce looks like
(and was sold as) a no-op, but it actually had a rather serious and
subtle side effect: the UNUSED_SYMBOLS option not only enabled the
removed (unused) functionality, it also _disabled_ the TRIM_UNUSED_KSYMS
functionality.
And it turns out that TRIM_UNUSED_KSYMS is a huge time waste, and takes
up a third of the kernel build time for me. For no actual upside, since
no distro is likely to ever be able to enable it (because they all
support external kernel modules).
Rather than re-enable EXPORT_UNUSED_SYMBOL, this just disables the
TRIM_UNUSED_KSYMS option by marking it broken. I'm tempted to just
remove the support entirely, but maybe somebody has a use-case and can
fix the behavior of it.
I could have just disabled it for COMPILE_TEST, but it really smells
like the TRIM_UNUSED_KSYMS option is badly done and not really useful,
so this takes the more direct approach - let's see if anybody ever
actually notices or complains.
Cc: Miroslav Benes <mbenes@suse.cz>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Jessica Yu <jeyu@kernel.org>
Fixes: 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*")
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
* acpica:
ACPICA: Remove some code duplication from acpi_ev_address_space_dispatch
ACPICA: Fix race in generic_serial_bus (I2C) and GPIO op_region parameter handling
|
|
* pm-cpufreq:
cpufreq: Fix typo in kerneldoc comment
cpufreq: schedutil: Remove update_lock comment from struct sugov_policy definition
cpufreq: schedutil: Remove needless sg_policy parameter from ignore_dl_rate_limit()
cpufreq: ACPI: Set cpuinfo.max_freq directly if max boost is known
cpufreq: qcom-hw: drop devm_xxx() calls from init/exit hooks
* pm-opp:
opp: Don't skip freq update for different frequency
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/jeyu/linux
Pull module updates from Jessica Yu:
- Retire EXPORT_UNUSED_SYMBOL() and EXPORT_SYMBOL_GPL_FUTURE(). These
export types were introduced between 2006 - 2008. All the of the
unused symbols have been long removed and gpl future symbols were
converted to gpl quite a long time ago, and I don't believe these
export types have been used ever since. So, I think it should be safe
to retire those export types now (Christoph Hellwig)
- Refactor and clean up some aged code cruft in the module loader
(Christoph Hellwig)
- Build {,module_}kallsyms_on_each_symbol only when livepatching is
enabled, as it is the only caller (Christoph Hellwig)
- Unexport find_module() and module_mutex and fix the last module
callers to not rely on these anymore. Make module_mutex internal to
the module loader (Christoph Hellwig)
- Harden ELF checks on module load and validate ELF structures before
checking the module signature (Frank van der Linden)
- Fix undefined symbol warning for clang (Fangrui Song)
- Fix smatch warning (Dan Carpenter)
* tag 'modules-for-v5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/jeyu/linux:
module: potential uninitialized return in module_kallsyms_on_each_symbol()
module: remove EXPORT_UNUSED_SYMBOL*
module: remove EXPORT_SYMBOL_GPL_FUTURE
module: move struct symsearch to module.c
module: pass struct find_symbol_args to find_symbol
module: merge each_symbol_section into find_symbol
module: remove each_symbol_in_section
module: mark module_mutex static
kallsyms: only build {,module_}kallsyms_on_each_symbol when required
kallsyms: refactor {,module_}kallsyms_on_each_symbol
module: use RCU to synchronize find_module
module: unexport find_module and module_mutex
drm: remove drm_fb_helper_modinit
powerpc/powernv: remove get_cxl_module
module: harden ELF info handling
module: Ignore _GLOBAL_OFFSET_TABLE_ when warning for undefined symbols
|
|
Pull microblaze updates from Michal Simek:
- Fix DTB alignment
- Remove code for very old GCC versions
- Remove TRACING_SUPPORT selection
* tag 'microblaze-v5.12' of git://git.monstr.eu/linux-2.6-microblaze:
microblaze: Fix built-in DTB alignment to be 8-byte aligned
microblaze: Remove support for gcc < 4
microblaze: do not select TRACING_SUPPORT directly
|
|
In gfs2_ail1_flush, we're using I/O plugging to give the block layer a
better chance of merging I/O requests. If we're too aggressive here, we
can end up waiting on I/O to complete while still plugged. Fix that in
a way similar to writeback_sb_inodes, except that we can't use
blk_flush_plug because blk_flush_plug_list is not exported.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull objtool updates from Thomas Gleixner:
- Make objtool work for big-endian cross compiles
- Make stack tracking via stack pointer memory operations match
push/pop semantics to prepare for architectures w/o PUSH/POP
instructions.
- Add support for analyzing alternatives
- Improve retpoline detection and handling
- Improve assembly code coverage on x86
- Provide support for inlined stack switching
* tag 'objtool-core-2021-02-23' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (33 commits)
objtool: Support stack-swizzle
objtool,x86: Additionally decode: mov %rsp, (%reg)
x86/unwind/orc: Change REG_SP_INDIRECT
x86/power: Support objtool validation in hibernate_asm_64.S
x86/power: Move restore_registers() to top of the file
x86/power: Annotate indirect branches as safe
x86/acpi: Support objtool validation in wakeup_64.S
x86/acpi: Annotate indirect branch as safe
x86/ftrace: Support objtool vmlinux.o validation in ftrace_64.S
x86/xen/pvh: Annotate indirect branch as safe
x86/xen: Support objtool vmlinux.o validation in xen-head.S
x86/xen: Support objtool validation in xen-asm.S
objtool: Add xen_start_kernel() to noreturn list
objtool: Combine UNWIND_HINT_RET_OFFSET and UNWIND_HINT_FUNC
objtool: Add asm version of STACK_FRAME_NON_STANDARD
objtool: Assume only ELF functions do sibling calls
x86/ftrace: Add UNWIND_HINT_FUNC annotation for ftrace_stub
objtool: Support retpoline jump detection for vmlinux.o
objtool: Fix ".cold" section suffix check for newer versions of GCC
objtool: Fix retpoline detection in asm code
...
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git
Merge the resource group glock sharing feature and the revoke accounting rework.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
|
|
NORMAL (0x0) and IDLE (0x4) are really two different states. Hence you
cannot check for both using a bitmask, as that checks for IDLE only,
breaking operation for devices that are in NORMAL state.
Fix the wait function to report either state as ready.
Fixes: 6524d8eac258452e ("Input: st1232 - add IDLE state as ready condition")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Michael Tretter <m.tretter@pengutronix.de>
Link: https://lore.kernel.org/r/20210223090201.1430542-1-geert+renesas@glider.be
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux
Pull clang LTO updates from Kees Cook:
"Clang Link Time Optimization.
This is built on the work done preparing for LTO by arm64 folks,
tracing folks, etc. This includes the core changes as well as the
remaining pieces for arm64 (LTO has been the default build method on
Android for about 3 years now, as it is the prerequisite for the
Control Flow Integrity protections).
While x86 LTO enablement is done, it depends on some pending objtool
clean-ups. It's possible that I'll send a "part 2" pull request for
LTO that includes x86 support.
For merge log posterity, and as detailed in commit dc5723b02e52
("kbuild: add support for Clang LTO"), here is the lt;dr to do an LTO
build:
make LLVM=1 LLVM_IAS=1 defconfig
scripts/config -e LTO_CLANG_THIN
make LLVM=1 LLVM_IAS=1
(To do a cross-compile of arm64, add "CROSS_COMPILE=aarch64-linux-gnu-"
and "ARCH=arm64" to the "make" command lines.)
Summary:
- Clang LTO build infrastructure and arm64-specific enablement (Sami
Tolvanen)
- Recursive build CC_FLAGS_LTO fix (Alexander Lobakin)"
* tag 'clang-lto-v5.12-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
kbuild: prevent CC_FLAGS_LTO self-bloating on recursive rebuilds
arm64: allow LTO to be selected
arm64: disable recordmcount with DYNAMIC_FTRACE_WITH_REGS
arm64: vdso: disable LTO
drivers/misc/lkdtm: disable LTO for rodata.o
efi/libstub: disable LTO
scripts/mod: disable LTO for empty.c
modpost: lto: strip .lto from module names
PCI: Fix PREL32 relocations for LTO
init: lto: fix PREL32 relocations
init: lto: ensure initcall ordering
kbuild: lto: add a default list of used symbols
kbuild: lto: merge module sections
kbuild: lto: limit inlining
kbuild: lto: fix module versioning
kbuild: add support for Clang LTO
tracing: move function tracer options to Kconfig
|
|
- Trust Panora support improvements from Cristian Klein
|
|
- tilt support on Sony Guitar Hero devices from Sanjay Govind
|
|
- Playstation DualSense support from Roderick Colenbrander
|
|
- hid-multitouch devices should be put into high-latency mode when
suspended in order to be in line with Windows Precision Touchpad
guidelines. From Blaž Hrastnik.
|
|
- support for "Unified Battery (1004) feature" from Filipe Laíns
|
|
- power management improvements for intel-ish driver from Zhang Lixu
|
|
- ACPI and OF support made more generic / decoupled. From Douglas Anderson
- support for Goodix devices from Douglas Anderson
|
|
- User experience improvements for hid-google from Nicolas Boichat
|
|
- HID documentation fixes from Randy Dunlap
|
|
- KEY_RFKILL support for Chicony devices from Jian-Hong Pan
|
|
- improved handling of generic HID keyboard (no more splitting system
and consumer controls away), in order to make it easier for userspace
to figure out the details of the device easier. From Dmitry Torokhov.
- report data sanitization fixes from Will McVicker and Randy Dunlap
|
|
Prepare input updates for 5.12 merge window.
|
|
Fix the following coccicheck warning:
./drivers/input/tablet/aiptek.c:1629:8-16: WARNING: use scnprintf or
sprintf.
Reported-by: Abaci Robot<abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Link: https://lore.kernel.org/r/1612419191-1078-1-git-send-email-jiapeng.chong@linux.alibaba.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
s/postive/positive/
Signed-off-by: Bhaskar Chowdhury <unixbhaskar@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Reviewed-by: Pali Rohár <pali@kernel.org>
Link: https://lore.kernel.org/r/20210222075439.32201-1-unixbhaskar@gmail.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
The common cros-ec keymap has been defined as macros. This patch uses
the macros to simply linux,keymap in cros-ec-keyboard.dtsi file.
This patch also creates an alias for keyboard-controller to make it
easier to override the keymap in board-specific dts later.
Signed-off-by: Philip Chen <philipchen@chromium.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Link: https://lore.kernel.org/r/20210115143555.v6.2.I9ec9c3c61eded22a5f7fbff838d23fc95ec7cfe0@changeid
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
Decouple LOCK from F13 and directly map the LOCK key (KSI3/KSO9) to
KEY_SLEEP action key code.
Signed-off-by: Philip Chen <philipchen@chromium.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Acked-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20210115143555.v6.3.I96134907488f41f358d03f3c1b08194f9547e670@changeid
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
In Chrome OS, the keyboard matrix can be split to two groups:
The keymap for the top row keys can be customized based on OEM
preference, while the keymap for the other keys is generic/fixed
across boards.
This patch creates marcos for the keymaps of these two groups, making
it easier to reuse the generic portion of keymap when we override the
keymap in the board-specific dts for custom top row design.
Signed-off-by: Philip Chen <philipchen@chromium.org>
Acked-by: Rob Herring <robh@kernel.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Link: https://lore.kernel.org/r/20210115143555.v6.1.Iaa8a60cf2ed4b7ad5e2fbb4ad76a1c600ee36113@changeid
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
The top-row keys in a keyboard usually have dual functionalities.
E.g. A function key "F1" is also an action key "Browser back".
Therefore, when an application receives an action key code from
a top-row key press, the application needs to know how to correlate
the action key code with the function key code and do the conversion
whenever necessary.
Since the userpace already knows the key scanlines (row/column)
associated with a received key code. Essentially, the userspace only
needs a mapping between the key row/column and the matching physical
location in the top row.
So, enhance the cros-ec-keyb driver to create such a mapping
and expose it to userspace in the form of a function_row_physmap
attribute. The attribute would be a space separated ordered list of
row/column codes for the keys in the function row, in a left-to-right
order.
The attribute will only be present when the device has a custom design
for the top-row keys.
Signed-off-by: Philip Chen <philipchen@chromium.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Link: https://lore.kernel.org/r/20210115122412.v7.2.I6542d7d9d0b246e7079bb16b41e697b2ac4b4e39@changeid
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
Add a new property `function-row-physmap` to the
device tree for the custom keyboard top row design.
The property describes the rows/columns of the top row keys
from left to right.
Signed-off-by: Philip Chen <philipchen@chromium.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20210115122412.v7.1.I025fb861cd5fa0ef5286b7dce514728e9df7ae74@changeid
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
|
|
Merge with mainline to get latest APIs and device tree bindings.
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl
Pull pin control updates from Linus Walleij:
"This is the bulk of pin control changes for the v5.12 kernel.
This time a calm set with no core changes.
New drivers/subdrivers:
- Renesas R8A7790A0 pin controller.
- Allwinner H616 and H616-R pin controllers.
- Qualcomm SM8350 and SC8180x pin controllers.
Improvements:
- Redo the DT bindings for Ralink RT2880.
- A common Qualcomm TLMM DT binding in YAML.
- Delete the unused drivers for U300, COH901, Sirf Atlas, and ZTE ZX"
* tag 'pinctrl-v5.12-1' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl: (71 commits)
pinctrl: mediatek: Fix trigger type setting follow for unexpected interrupt
dt-bindings: pinctrl: Group tuples in pin control properties
pinctrl: nuvoton: npcm7xx: Fix alignment of table header comment
pinctrl: at91-pio4: fix "Prefer 'unsigned int' to bare use of 'unsigned'"
pinctrl: at91-pio4: add support for slew-rate
dt-bindings: pinctrl: at91-pio4: add slew-rate
pinctrl: actions: Add depends on || COMPILE_TEST
pinctrl: single: set function name when adding function
pinctrl: qcom: Add sc8180x TLMM driver
dt-bindings: pinctrl: qcom: Add sc8180x binding
dt-bindings: pinctrl: qcom: Define common TLMM binding
pinctrl: qcom: Add SM8350 pinctrl driver
dt-bindings: pinctrl: qcom: Add SM8350 pinctrl bindings
pinctrl: samsung: use raw_spinlock for s3c64xx
dt-bindings: mediatek: mt8192: Fix dt_binding_check warning
pinctrl: qcom: spmi-mpp: Add PM8019 compatible
pinctrl: pinmux: add function selector to pinmux-functions
pinctrl: samsung: use raw_spinlock for locking
pinctrl: clarify #pinctrl-cells for pinctrl-single,pins
pinctrl: actions: Add the platform dependency to drivers
...
|
|
Pull IPMI update from Corey Minyard:
"Only one change, but it's required for other things, so it needs to go
in"
* tag 'for-linus-5.12-1' of git://github.com/cminyard/linux-ipmi:
ipmi: remove open coded version of SMBus block write
|
|
git://anongit.freedesktop.org/drm/drm
Pull follow_pfn() updates from Daniel Vetter:
"Fixes around VM_FPNMAP and follow_pfn:
- replace mm/frame_vector.c by get_user_pages in misc/habana and
drm/exynos drivers, then move that into media as it's sole user
- close race in generic_access_phys
- s390 pci ioctl fix of this series landed in 5.11 already
- properly revoke iomem mappings (/dev/mem, pci files)"
* tag 'topic/iomem-mmap-vs-gup-2021-02-22' of git://anongit.freedesktop.org/drm/drm:
PCI: Revoke mappings like devmem
PCI: Also set up legacy files only after sysfs init
sysfs: Support zapping of binary attr mmaps
resource: Move devmem revoke code to resource framework
/dev/mem: Only set filp->f_mapping
PCI: Obey iomem restrictions for procfs mmap
mm: Close race in generic_access_phys
media: videobuf2: Move frame_vector into media subsystem
mm/frame-vector: Use FOLL_LONGTERM
misc/habana: Use FOLL_LONGTERM for userptr
misc/habana: Stop using frame_vector helpers
drm/exynos: Use FOLL_LONGTERM for g2d cmdlists
drm/exynos: Stop using frame_vector helpers
|
|
git://anongit.freedesktop.org/drm/drm
Pull kcmp kconfig update from Daniel Vetter:
"Make the kcmp syscall available independently of checkpoint/restore.
drm userspaces uses this, systemd uses this, so makes sense to pull it
out from the checkpoint-restore bundle.
Kees reviewed this from security pov and is happy with the final
version"
Link: https://lwn.net/Articles/845448/
* tag 'topic/kcmp-kconfig-2021-02-22' of git://anongit.freedesktop.org/drm/drm:
kcmp: Support selection of SYS_kcmp without CHECKPOINT_RESTORE
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace
Pull user namespace update from Eric Biederman:
"There are several pieces of active development, but only a single
change made it through the gauntlet to be ready for v5.12. That change
is tightening up the semantics of the v3 capabilities xattr. It is
just short of being a bug-fix/security issue as no user space is known
to even generate the problem case"
* 'userns-for-v5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
capabilities: Don't allow writing ambiguous v3 file capabilities
|
|
Pull qorkqueue updates from Tejun Heo:
"Tracepoint and comment updates only"
* 'for-5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: Use %s instead of function name
workqueue: tracing the name of the workqueue instead of it's address
workqueue: fix annotation for WQ_SYSFS
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Pull cgroup updates from Tejun Heo:
"Nothing interesting. Just two minor patches"
* 'for-5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cpuset: fix typos in comments
cgroup: cgroup.{procs,threads} factor out common parts
|
|
Pinning one 4K page at a time is inefficient, so do it in batches of 512
instead. This is just an optimization with no functional change
intended, and in particular the driver still calls iommu_map() with the
largest physically contiguous range possible.
Add two fields in vfio_batch to remember where to start between calls to
vfio_pin_pages_remote(), and use vfio_batch_unpin() to handle remaining
pages in the batch in case of error.
qemu pins pages for guests around 8% faster on my test system, a
two-node Broadwell server with 128G memory per node. The qemu process
was bound to one node with its allocations constrained there as well.
base test
guest ---------------- ----------------
mem (GB) speedup avg sec (std) avg sec (std)
1 7.4% 0.61 (0.00) 0.56 (0.00)
2 8.3% 0.93 (0.00) 0.85 (0.00)
4 8.4% 1.46 (0.00) 1.34 (0.00)
8 8.6% 2.54 (0.01) 2.32 (0.00)
16 8.3% 4.66 (0.00) 4.27 (0.01)
32 8.3% 8.94 (0.01) 8.20 (0.01)
64 8.2% 17.47 (0.01) 16.04 (0.03)
120 8.5% 32.45 (0.13) 29.69 (0.01)
perf diff confirms less time spent in pup. Here are the top ten
functions:
Baseline Delta Abs Symbol
78.63% +6.64% clear_page_erms
1.50% -1.50% __gup_longterm_locked
1.27% -0.78% __get_user_pages
+0.76% kvm_zap_rmapp.constprop.0
0.54% -0.53% vmacache_find
0.55% -0.51% get_pfnblock_flags_mask
0.48% -0.48% __get_user_pages_remote
+0.39% slot_rmap_walk_next
+0.32% vfio_pin_map_dma
+0.26% kvm_handle_hva_range
...
Suggested-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
|
|
Get ready to pin more pages at once with struct vfio_batch, which
represents a batch of pinned pages.
The struct has a fallback page pointer to avoid two unlikely scenarios:
pointlessly allocating a page if disable_hugepages is enabled or failing
the whole pinning operation if the kernel can't allocate memory.
vaddr_get_pfn() becomes vaddr_get_pfns() to prepare for handling
multiple pages, though for now only one page is stored in the pages
array.
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
|
|
vaddr_get_pfn() simply returns 0 on success. Have it report the number
of pfns successfully gotten instead, whether from page pinning or
follow_fault_pfn(), which will be used later when batching pinning.
Change the last check in vfio_pin_pages_remote() for consistency with
the other two.
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
|
