With Posted-Interrupts support in Intel CPU and IOMMU, an external
interrupt from assigned-devices could be directly delivered to a
virtual CPU in a virtual machine. Instead of hacking KVM and Intel
IOMMU drivers, we propose a platform independent interface to target
an interrupt to a specific virtual CPU in a virtual machine, or set
virtual CPU affinity for an interrupt.
By adopting this new interface and the hierarchy irqdomain, we could
easily support posted-interrupts on Intel platforms, and also provide
flexible enough interfaces for other platforms to support similar
features.
Here is the usage scenario for this interface:
Guest update MSI/MSI-X interrupt configuration
-->QEMU and KVM handle this
-->KVM call this interface (passing posted interrupts descriptor
and guest vector)
-->irq core will transfer the control to IOMMU
-->IOMMU will do the real work of updating IRTE (IRTE has new
format for VT-d Posted-Interrupts)
Signed-off-by: Jiang Liu <jiang.liu@linux.intel.com>
Signed-off-by: Feng Wu <feng.wu@intel.com>
Link: http://lkml.kernel.org/r/1432026437-16560-2-git-send-email-feng.wu@intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
So while testing kernels using tools/kvm/ (kvmtool) I noticed that it
booted super slow:
[ 0.142991] Performance Events: no PMU driver, software events only.
[ 0.149265] x86: Booting SMP configuration:
[ 0.149765] .... node #0, CPUs: #1
[ 0.148304] kvm-clock: cpu 1, msr 2:1bfe9041, secondary cpu clock
[ 10.158813] KVM setup async PF for cpu 1
[ 10.159000] #2
[ 10.159000] kvm-stealtime: cpu 1, msr 211a4d400
[ 10.158829] kvm-clock: cpu 2, msr 2:1bfe9081, secondary cpu clock
[ 20.167805] KVM setup async PF for cpu 2
[ 20.168000] #3
[ 20.168000] kvm-stealtime: cpu 2, msr 211a8d400
[ 20.167818] kvm-clock: cpu 3, msr 2:1bfe90c1, secondary cpu clock
[ 30.176902] KVM setup async PF for cpu 3
[ 30.177000] #4
[ 30.177000] kvm-stealtime: cpu 3, msr 211acd400
One CPU booted up per 10 seconds. With 120 CPUs that takes a while.
Bisection pinpointed this commit:
853b160aaa ("Revert f5d6a52f51 ("x86/smpboot: Skip delays during SMP initialization similar to Xen")")
But that commit just restores previous behavior, so it cannot cause the
problem. After some head scratching it turns out that these two commits:
1a744cb356 ("x86/smp/boot: Remove 10ms delay from cpu_up() on modern processors")
d68921f9bd ("x86/smp/boot: Add cmdline "cpu_init_udelay=N" to specify cpu_up() delay")
added the following code to smpboot.c:
- mdelay(10);
+ mdelay(init_udelay);
Note the mismatch in the units: the delay is called 'udelay' and is set
to microseconds - while the function used here is actually 'mdelay',
which counts in milliseconds ...
So the delay for legacy systems is off by a factor of 1,000, so instead
of 10 msecs we waited for 10 seconds ...
The reason bisection pointed to 853b160aaa was that 853b160aaa removed
a (broken) boot-time speedup patch, which masked the factor 1,000 bug.
Fix it by using udelay(). This fixes my bootup problems.
Cc: Len Brown <len.brown@intel.com>
Cc: Alan Cox <alan@linux.intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jan H. Schönherr <jschoenh@amazon.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Packing loops tightly (-falign-loops=1) is beneficial to code size:
text data bss dec filename
12566391 1617840 1089536 15273767 vmlinux.align.16-byte
12224951 1617840 1089536 14932327 vmlinux.align.1-byte
11976567 1617840 1089536 14683943 vmlinux.align.1-byte.funcs-1-byte
11903735 1617840 1089536 14611111 vmlinux.align.1-byte.funcs-1-byte.loops-1-byte
Which reduces the size of the kernel by another 0.6%, so the
the total combined size reduction of the alignment-packing
patches is ~5.5%.
The x86 decoder bandwidth and caching arguments laid out in:
be6cb02779 ("x86: Align jump targets to 1-byte boundaries")
apply to loop alignment as well.
Furtermore, modern CPU uarchs have a loop cache/buffer that
is a L0 cache before even any uop cache, covering a few
dozen most recently executed instructions.
This loop cache generally does not have the 16-byte alignment
restrictions of the uop cache.
Now loop alignment can still be beneficial if:
- a loop is cache-hot and its surroundings are not.
- if the loop is so cache hot that the instruction
flow becomes x86 decoder bandwidth limited
But loop alignment is harmful if:
- a loop is cache-cold
- a loop's surroundings are cache-hot as well
- two cache-hot loops are close to each other
- if the loop fits into the loop cache
- if the code flow is not decoder bandwidth limited
and I'd argue that the latter five scenarios are much
more common in the kernel, as our hottest loops are
typically:
- pointer chasing: this should fit into the loop cache
in most cases and is typically data cache and address
generation limited
- generic memory ops (memset, memcpy, etc.): these generally
fit into the loop cache as well, and are likewise data
cache limited.
So this patch packs loop addresses tightly as well.
Acked-by: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Aswin Chandramouleeswaran <aswin@hp.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/20150410123017.GB19918@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull USB fixes from Greg KH:
"Here are some USB fixes and new device ids for 4.1-rc4.
All are pretty minor, and have been in linux-next successfully"
* tag 'usb-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
usb-storage: Add NO_WP_DETECT quirk for Lacie 059f:0651 devices
Added another USB product ID for ELAN touchscreen quirks.
xhci: gracefully handle xhci_irq dead device
xhci: Solve full event ring by increasing TRBS_PER_SEGMENT to 256
xhci: fix isoc endpoint dequeue from advancing too far on transaction error
usb: chipidea: debug: avoid out of bound read
USB: visor: Match I330 phone more precisely
USB: pl2303: Remove support for Samsung I330
USB: cp210x: add ID for KCF Technologies PRN device
usb: gadget: remove incorrect __init/__exit annotations
usb: phy: isp1301: work around tps65010 dependency
usb: gadget: serial: fix re-ordering of tx data
usb: gadget: hid: Fix static variable usage
usb: gadget: configfs: Fix interfaces array NULL-termination
usb: gadget: xilinx: fix devm_ioremap_resource() check
usb: dwc3: dwc3-omap: correct the register macros
Pull tty/serial fixes from Greg KH:
"Here's some TTY and serial driver fixes for reported issues.
All of these have been in linux-next successfully"
* tag 'tty-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
pty: Fix input race when closing
tty/n_gsm.c: fix a memory leak when gsmtty is removed
Revert "serial/amba-pl011: Leave the TX IRQ alone when the UART is not open"
serial: omap: Fix error handling in probe
earlycon: Revert log warnings
Pull staging / IIO driver fixes from Greg KH:
"Here's some staging and iio driver fixes to resolve a number of
reported issues.
All of these have been in linux-next for a while"
* tag 'staging-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging: (31 commits)
iio: light: hid-sensor-prox: Fix memory leak in probe()
iio: adc: cc10001: Add delay before setting START bit
iio: adc: cc10001: Fix regulator_get_voltage() return value check
iio: adc: cc10001: Fix incorrect use of power-up/power-down register
staging: gdm724x: Correction of variable usage after applying ALIGN()
iio: adc: cc10001: Fix the channel number mapping
staging: vt6655: lock MACvWriteBSSIDAddress.
staging: vt6655: CARDbUpdateTSF bss timestamp correct tsf counter value.
staging: vt6655: vnt_tx_packet Correct TX order of OWNED_BY_NIC
staging: vt6655: Fix 80211 control and management status reporting.
staging: vt6655: implement IEEE80211_TX_STAT_NOACK_TRANSMITTED
staging: vt6655: device_free_tx_buf use only ieee80211_tx_status_irqsafe
staging: vt6656: use ieee80211_tx_info to select packet type.
staging: rtl8712: freeing an ERR_PTR
staging: sm750: remove incorrect __exit annotation
iio: kfifo: Set update_needed to false only if a buffer was allocated
iio: mcp320x: Fix occasional incorrect readings
iio: accel: mma9553: check input value for activity period
iio: accel: mma9553: add enable channel for activity
iio: accel: mma9551_core: prevent buffer overrun
...
Pull char/misc fix from Greg KH:
"Here is one fix, in the extcon subsystem, that resolves a reported
issue.
It's been in linux-next for a number of weeks now, sorry for not
getting it to you sooner"
* tag 'char-misc-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
extcon: usb-gpio: register extcon device before IRQ registration
Pull UML hostfs fix from Richard Weinberger:
"This contains a single fix for a regression introduced in 4.1-rc1"
* 'for-linus-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml:
hostfs: Use correct mask for file mode
Pull UBI bufix from Richard Weinberger:
"This contains a single bug fix for the UBI block driver"
* tag 'upstream-4.1-rc4' of git://git.infradead.org/linux-ubifs:
UBI: block: Add missing cache flushes
Pull ext4 fixes from Ted Ts'o:
"Fix a number of ext4 bugs; the most serious of which is a bug in the
lazytime mount optimization code where we could end up updating the
timestamps to the wrong inode"
* tag 'for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
ext4: fix an ext3 collapse range regression in xfstests
jbd2: fix r_count overflows leading to buffer overflow in journal recovery
ext4: check for zero length extent explicitly
ext4: fix NULL pointer dereference when journal restart fails
ext4: remove unused function prototype from ext4.h
ext4: don't save the error information if the block device is read-only
ext4: fix lazytime optimization
Pull btrfs fixes from Chris Mason:
"The first commit is a fix from Filipe for a very old extent buffer
reuse race that triggered a BUG_ON. It hasn't come up often, I looked
through old logs at FB and we hit it a handful of times over the last
year.
The rest are other corners he hit during testing"
* 'for-linus-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
Btrfs: fix race when reusing stale extent buffers that leads to BUG_ON
Btrfs: fix race between block group creation and their cache writeout
Btrfs: fix panic when starting bg cache writeout after IO error
Btrfs: fix crash after inode cache writeback failure
Pull MIPS fixes from Ralf Baechle:
"Seven small fixes. The shortlog below is a good description so no
need to elaborate.
It has sat in linux-next and survived the usual automated testing by
Imagination's test farm"
* 'master' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
MIPS: tlb-r4k: Fix PG_ELPA comment
MIPS: Fix up obsolete cpu_set usage
MIPS: IP32: Fix build errors in reset code in DS1685 platform hook.
MIPS: KVM: Fix unused variable build warning
MIPS: traps: remove extra Tainted: line from __show_regs() output
MIPS: Fix wrong CHECKFLAGS (sparse builds) with GCC 5.1
MIPS: Fix a preemption issue with thread's FPU defaults
Pull ARM SoC fixes from Arnd Bergmann:
"Nothing frightening this time, just smaller fixes in a number of
places.
The other changes contained here are:
MAINTAINERS file updates:
- The mach-gemini maintainer is back in action and has a new git tree
- Krzysztof Kozlowski has volunteered to be a new co-maintainer for
the samsung platforms
- updates to the files that belong to Marvell mvebu
Bug fixes:
- The largest changes are on omap2, but are only to avoid some
harmless warnings and to fix reset on omap4
- a small regression fix on tegra
- multiple fixes for incorrect IRQ affinity on vexpress
- the missing system controller on arm64 juno is added
- one revert of a patch that was accidentally applied twice for
mach-rockchip
- two clock related DT fixes for mvebu
- a workaround for suspend with old DT binaries on new exynos kernels
- Another fix for suspend on exynos, needs to be backported"
* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (21 commits)
MAINTAINERS: Add dts entries for some of the Marvell SoCs
MAINTAINERS: ARM: EXYNOS: Add Krzysztof Kozlowski as co-maintainer
ARM: EXYNOS: Use of_machine_is_compatible instead of soc_is_exynos4
ARM: EXYNOS: Fix failed second suspend on Exynos4
Revert "ARM: rockchip: fix undefined instruction of reset_ctrl_regs"
ARM: EXYNOS: Fix dereference of ERR_PTR returned by of_genpd_get_from_provider
ARM: EXYNOS: Don't try to initialize suspend on old DT
ARM: dts: Add keep-power-in-suspend to WiFi SDIO node for Peach Boards
ARM: gemini: fix compiler warning due wrong data type
ARM: vexpress/tc2: Add interrupt-affinity to the PMU node
ARM: vexpress/ca9: Add interrupt-affinity to the PMU node
ARM: vexpress/ca9: Add unified-cache property to l2 cache node
ARM64: juno: add sp810 support and fix sp804 clock frequency
ARM: Gemini: Maintainers update
ARM: OMAP2+: Remove bogus struct clk comparison for timer clock
ARM: dove: Add clock-names to CuBox Si5351 clk generator
ARM: AM33xx+: hwmod: re-use omap4 implementations for reset functionality
ARM: OMAP4+: PRM: add support for passing status register/bit info to reset
ARM: AM43xx: hwmod: add VPFE hwmod entries
ARM: mvebu: Fix the main PLL frequency on Armada 375, 38x and 39x SoCs
...
Pull thermal fixes from Zhang Rui:
"Specifics:
- fix an issue in intel_powerclamp driver that idle injection target
is not accurately maintained on newer Intel CPUs. Package C8 to
C10 states are introduced on these CPUs but they were not included
in the package c-state residency calculation. From Jacob Pan.
- fix a problem that package c-state idle injection was missing on
Broadwell server, by adding its id to intel_powerclamp driver.
From Jacob Pan.
- a couple of small fixes and cleanups from Joe Perches, Mathias
Krause, Dan Carpenter and Anand Moon"
* 'for-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux:
tools/thermal: tmon: fixed the 'make install' command
thermal: rockchip: fix an error code
thermal/powerclamp: fix missing newer package c-states
thermal/intel_powerclamp: add id for broadwell server
thermal/intel_powerclamp: add __init / __exit annotations
thermal: Use bool function return values of true/false not 1/0
Pull kselftest fixes from Shuah Khan:
"Urgent fix for Kselftest regression introduced in 4.1-rc1 by the new
x86 test due to its hard dependency on 32-bit build environment.
A set of 5 patches fix the make kselftest run and kselftest install"
* tag 'linux-kselftest-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
selftests, x86: Rework x86 target architecture detection
selftests, x86: Remove useless run_tests rule
selftests/x86: install tests
selftest/x86: have no dependency on all when cross building
selftest/x86: build both bitnesses
Pull parisc fixes from Helge Deller:
"One important patch which fixes crashes due to stack randomization on
architectures where the stack grows upwards (currently parisc and
metag only).
This bug went unnoticed on parisc since kernel 3.14 where the flexible
mmap memory layout support was added by commit 9dabf60dc4. The
changes in fs/exec.c are inside an #ifdef CONFIG_STACK_GROWSUP section
and will not affect other platforms.
The other two patches rename args of the kthread_arg() function and
fixes a printk output"
* 'parisc-4.1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
parisc,metag: Fix crashes due to stack randomization on stack-grows-upwards architectures
parisc: copy_thread(): rename 'arg' argument to 'kthread_arg'
parisc: %pf is only for function pointers
cpu_set was removed (along with a bunch of cpumask helpers) by
commit 2f0f267ea0 ("cpumask: remove deprecated functions.").
Fix this by replacing cpu_set with cpumask_set_cpu. Without this
fix the following error is triggered when CONFIG_MIPS_MT_FPAFF=y.
arch/mips/kernel/smp-cps.c: In function 'cps_smp_setup':
arch/mips/kernel/smp-cps.c:95:3: error: implicit declaration of function 'cpu_set' [-Werror=implicit-function-declaration]
Fixes: 90db024f14 ("MIPS: smp-cps: cpu_set FPU mask if FPU present")
Signed-off-by: Ezequiel Garcia <ezequiel.garcia@imgtec.com>
Acked-by: Niklas Cassel <niklass@axis.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9912/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Pull x86 build fix from Ingo Molnar:
"A bzImage build fix on older distros"
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/vdso: Fix 'make bzImage' on older distros
Pull scheduler fixes from Ingo Molnar:
"Two fixes: a suspend/resume related regression fix, and an RT priority
boosting fix"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/core: Fix regression in cpuset_cpu_inactive() for suspend
sched: Handle priority boosted tasks proper in setscheduler()
Pull perf fixes from Ingo Molnar:
"Mostly tooling fixes, but also a lockdep annotation fix, a PMU event
list fix and a new model addition"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
tools/liblockdep: Fix compilation error
tools/liblockdep: Fix linker error in case of cross compile
perf tools: Use getconf to determine number of online CPUs
tools: Fix tools/vm build
perf/x86/rapl: Enable Broadwell-U RAPL support
perf/x86/intel: Fix SLM cache event list
perf: Annotate inherited event ctx->mutex recursion
Pull irq fix from Ingo Molnar:
"A tegra irqchip driver memory corruption fix"
* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
irqchip: tegra: Set the proper base address in irq chip data
Pull drm fixes from Dave Airlie:
"Radeon:
one oops fix, one bug fix, one pci id addition patch
i915:
one suspend/resume regression fix.
All seems quiet enough."
* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
drm/radeon: don't do mst probing if MST isn't enabled.
drm/radeon: add new bonaire pci id
drm/radeon: fix VM_CONTEXT*_PAGE_TABLE_END_ADDR handling
drm/i915: Avoid GPU hang when coming out of s3 or s4
Merge misc fixes from Andrew Morton:
"8 fixes"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
mm, numa: really disable NUMA balancing by default on single node machines
MAINTAINERS: update Jingoo Han's email address
CMA: page_isolation: check buddy before accessing it
uidgid: make uid_valid and gid_valid work with !CONFIG_MULTIUSER
kernfs: do not account ino_ida allocations to memcg
gfp: add __GFP_NOACCOUNT
tools/vm: fix page-flags build
drivers/rtc/rtc-armada38x.c: remove unused local `flags'
Merge "Samsung 2nd fixes for v4.1" from Kukjin Kim:
- fix second S2R on exynos4412 based Trats2, Odroid U3 boards which
happened after enabling L2$ and caused by commit 13cfa6c4f7 ("ARM:
EXYNOS: Fix CPU idle clock down after CPU off")
And replace the soc_is_exynosxxx() macro with of_compatible_xxx
- fix dereference of ERR_PTR of of_genpd_get_from_provider()
- fix suspend problem on old DT machines to skip the initialization
suspend and caused by commit 8b283c0254 ("ARM: exynos4/5: convert
pmu wakeup to stacked domains")
- add keep-power-in-suspend for Peach Boards to support S2R and has
been missed in previous pull-request for fixes
* tag 'samsung-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
ARM: EXYNOS: Use of_machine_is_compatible instead of soc_is_exynos4
ARM: EXYNOS: Fix failed second suspend on Exynos4
ARM: EXYNOS: Fix dereference of ERR_PTR returned by of_genpd_get_from_provider
ARM: EXYNOS: Don't try to initialize suspend on old DT
ARM: dts: Add keep-power-in-suspend to WiFi SDIO node for Peach Boards
Merge "mvebu fixes for 4.1 (part 2)" from Gregory CLEMENT:
Fix the main PLL frequency on Armada 375, 38x and 39x SoCs
Add clock-names to CuBox Si5351 clk generator
Add dts entries in the MAINTAINERS file
* tag 'mvebu-fixes-4.1-2' of git://git.infradead.org/linux-mvebu:
MAINTAINERS: Add dts entries for some of the Marvell SoCs
ARM: dove: Add clock-names to CuBox Si5351 clk generator
ARM: mvebu: Fix the main PLL frequency on Armada 375, 38x and 39x SoCs
smp.c and irq_work.c implement the same inline helper. Move it to
apic.h and use it everywhere.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Since many releases, the modifications of the mvebu and berlin device
tree files are merged through the mvebu subsystem. This patch makes it
official in order to help the contributors using the get_maintainer.pl
to find the accurate peoples.
In the same time, updated the mvebu description which now includes the
kirkwood SoCs and new Armada SoCs.
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Acked-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Jason Cooper <jason@lakedaemon.net>
Acked-by: Andrew Lunn <andrew@lunn.ch>
The following NOP in a hot function caught my attention:
> 5a: 66 0f 1f 44 00 00 nopw 0x0(%rax,%rax,1)
That's a dead NOP that bloats the function a bit, added for the
default 16-byte alignment that GCC applies for jump targets.
I realize that x86 CPU manufacturers recommend 16-byte jump
target alignments (it's in the Intel optimization manual),
to help their relatively narrow decoder prefetch alignment
and uop cache constraints, but the cost of that is very
significant:
text data bss dec filename
12566391 1617840 1089536 15273767 vmlinux.align.16-byte
12224951 1617840 1089536 14932327 vmlinux.align.1-byte
By using 1-byte jump target alignment (i.e. no alignment at all)
we get an almost 3% reduction in kernel size (!) - and a
probably similar reduction in I$ footprint.
Now, the usual justification for jump target alignment is the
following:
- modern decoders tend to have 16-byte (effective) decoder
prefetch windows. (AMD documents it higher but measurements
suggest the effective prefetch window on curretn uarchs is
still around 16 bytes)
- on Intel there's also the uop-cache with cachelines that have
16-byte granularity and limited associativity.
- older x86 uarchs had a penalty for decoder fetches that crossed
16-byte boundaries. These limits are mostly gone from recent
uarchs.
So if a forward jump target is aligned to cacheline boundary then
prefetches will start from a new prefetch-cacheline and there's
higher chance for decoding in fewer steps and packing tightly.
But I think that argument is flawed for typical optimized kernel
code flows: forward jumps often go to 'cold' (uncommon) pieces
of code, and aligning cold code to cache lines does not bring a
lot of advantages (they are uncommon), while it causes
collateral damage:
- their alignment 'spreads out' the cache footprint, it shifts
followup hot code further out
- plus it slows down even 'cold' code that immediately follows 'hot'
code (like in the above case), which could have benefited from the
partial cacheline that comes off the end of hot code.
But even in the cache-hot case the 16 byte alignment brings
disadvantages:
- it spreads out the cache footprint, possibly making the code
fall out of the L1 I$.
- On Intel CPUs, recent microarchitectures have plenty of
uop cache (typically doubling every 3 years) - while the
size of the L1 cache grows much less aggressively. So
workloads are rarely uop cache limited.
The only situation where alignment might matter are tight
loops that could fit into a single 16 byte chunk - but those
are pretty rare in the kernel: if they exist they tend
to be pointer chasing or generic memory ops, which both tend
to be cache miss (or cache allocation) intensive and are not
decoder bandwidth limited.
So the balance of arguments strongly favors packing kernel
instructions tightly versus maximizing for decoder bandwidth:
this patch changes the jump target alignment from 16 bytes
to 1 byte (tightly packed, unaligned).
Acked-by: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Aswin Chandramouleeswaran <aswin@hp.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/20150410120846.GA17101@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
fix one gpu hang on resume.
* tag 'drm-intel-fixes-2015-05-13' of git://anongit.freedesktop.org/drm-intel:
drm/i915: Avoid GPU hang when coming out of s3 or s4
radeon minor fixes, and pci id addition.
* 'drm-fixes-4.1' of git://people.freedesktop.org/~agd5f/linux:
drm/radeon: don't do mst probing if MST isn't enabled.
drm/radeon: add new bonaire pci id
drm/radeon: fix VM_CONTEXT*_PAGE_TABLE_END_ADDR handling
The xfstests test suite assumes that an attempt to collapse range on
the range (0, 1) will return EOPNOTSUPP if the file system does not
support collapse range. Commit 280227a75b: "ext4: move check under
lock scope to close a race" broke this, and this caused xfstests to
fail when run when testing file systems that did not have the extents
feature enabled.
Reported-by: Eric Whitney <enwlinux@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Pull ACPI fixes from Rafael Wysocki:
"Two fixes here, one revert of a recent ACPICA commit that broke audio
support on one Dell machine and a fix for a long-standing issue that
may cause systems to break randomly during boot.
Specifics:
- The recent ACPICA commit that set the ACPI _REV return value to 2
(which is the value always used by Windows and now mandated by the
spec too) in order to prevent the firmware people from using it to
play tricks with us caused a serious audio regression to happen on
Dell XPS 13 (the AML on that machine uses the _REV return value to
decide how to expose audio to the OS and does that to hide the lack
of proper support for its I2S audio in Linux), so revert that
commit for now and we'll revisit the issue in the next cycle.
- Ensure that the ordering of acpi_reserve_resources() with respect
to the rest of the ACPI initialization sequence will always be the
same, or the IO or memory region occupied by the ACPI fixed
registers may be assigned to a PCI host bridge as a result of a
race and random breakage ensues going forward"
* tag 'pm+acpi-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
Revert "ACPICA: Permanently set _REV to the value '2'."
ACPI / init: Fix the ordering of acpi_reserve_resources()
Pull dmi fixes from Jean Delvare.
* 'dmi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging:
firmware: dmi_scan: Fix ordering of product_uuid
firmware: dmi_scan: Simplified displayed version
