Skip to content

L
linux
Switch branch/tag
  1. 10 Nov, 2017 1 commit
    • Michael S. Tsirkin's avatar
      locking/x86: Use LOCK ADD for smp_mb() instead of MFENCE · 450cbdd0
      Michael S. Tsirkin authored 
      MFENCE appears to be way slower than a locked instruction - let's use
LOCK ADD unconditionally, as we always did on old 32-bit.

Performance testing results:

  perf stat -r 10 -- ./virtio_ring_0_9 --sleep --host-affinity 0 --guest-affinity 0
  Before:
         0.922565990 seconds time elapsed                                          ( +-  1.15% )
  After:
         0.578667024 seconds time elapsed                                          ( +-  1.21% )

i.e. about ~60% faster.

Just poking at SP would be the most natural, but if we then read the
value from SP, we get a false dependency which will slow us down.

This was noted in this article:

  http://shipilev.net/blog/2014/on-the-fence-with-dependencies/

And is easy to reproduce by sticking a barrier in a small non-inline
function.

So let's use a negative offset - which avoids this problem since we
build with the red zone disabled.

For userspace, use an address just below the redzone.

The one difference between LOCK ADD and MFENCE is that LOCK ADD does
not affect CLFLUSH, previous patches converted all uses of CLFLUSH to
call mb(), such that changes to smp_mb() won't affect it.

Update mb/rmb/wmb() on 32-bit to use the negative offset, too, for
consistency.

As a follow-up, it might be worth considering switching users
of CLFLUSH to another API (e.g. clflush_mb()?) - we will
then be able to convert mb() to smp_mb() again.

Also arguably, GCC should switch to use LOCK ADD for __sync_synchronize().
This might be worth pursuing separately.
Suggested-by: default avatarAndy Lutomirski <luto@amacapital.net>
Signed-off-by: default avatarMichael S. Tsirkin <mst@redhat.com>
Acked-by: default avatarPeter Zijlstra <peterz@infradead.org>
Acked-by: default avatarThomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: qemu-devel@nongnu.org
Cc: virtualization@lists.linux-foundation.org
Link: http://lkml.kernel.org/r/1509118355-4890-1-git-send-email-mst@redhat.comSigned-off-by: default avatarIngo Molnar <mingo@kernel.org>
      450cbdd0
  3. 08 Nov, 2017 15 commits
  5. 07 Nov, 2017 2 commits
  7. 06 Nov, 2017 3 commits
  9. 05 Nov, 2017 8 commits
  11. 04 Nov, 2017 11 commits
    • Linus Torvalds's avatar
      Merge branch 'fixes' of git://git.armlinux.org.uk/~rmk/linux-arm · 2d634994
      Linus Torvalds authored 
      Pull ARM fixes from Russell King:

 - omit EFI memory map sorting, which was recently introduced, but
   caused problems with the decompressor due to additional sections
   being emitted.

 - avoid unaligned load fault-generating instructions in the
   decompressor by switching to a private unaligned implementation.

 - add a symbol into the decompressor to further debug non-boot
   situations (ld's documentation is extremely poor for how "." works,
   ld doesn't seem to follow its own documentation!)

 - parse endian information to sparse

* 'fixes' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: add debug ".edata_real" symbol
  ARM: 8716/1: pass endianness info to sparse
  efi/libstub: arm: omit sorting of the UEFI memory map
  ARM: 8715/1: add a private asm/unaligned.h
      2d634994
    • Linus Torvalds's avatar
      Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm · f0a32ee4
      Linus Torvalds authored 
      Pull KVM fixes from Paolo Bonzini:
 "Fixes for interrupt controller emulation in ARM/ARM64 and x86, plus a
  one-liner x86 KVM guest fix"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86: Update APICv on APIC reset
  KVM: VMX: Do not fully reset PI descriptor on vCPU reset
  kvm: Return -ENODEV from update_persistent_clock
  KVM: arm/arm64: vgic-its: Check GITS_BASER Valid bit before saving tables
  KVM: arm/arm64: vgic-its: Check CBASER/BASER validity before enabling the ITS
  KVM: arm/arm64: vgic-its: Fix vgic_its_restore_collection_table returned value
  KVM: arm/arm64: vgic-its: Fix return value for device table restore
  arm/arm64: kvm: Disable branch profiling in HYP code
  arm/arm64: kvm: Move initialization completion message
  arm/arm64: KVM: set right LR register value for 32 bit guest when inject abort
  KVM: arm64: its: Fix missing dynamic allocation check in scan_its_table
      f0a32ee4
    • Linus Torvalds's avatar
      Merge tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc · b1878b85
      Linus Torvalds authored 
      Pull ARM SoC fixes from Arnd Bergmann:
 "Only two patches came in over the last two weeks: Uniphier USB support
  needs additional clocks enabled (on both 32-bit and 64-bit ARM), and a
  Marvell MVEBU stability issue has been fixed"

* tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: dts: mvebu: pl310-cache disable double-linefill
  arm64: dts: uniphier: add STDMAC clock to EHCI nodes
  ARM: dts: uniphier: add STDMAC clock to EHCI nodes
      b1878b85
    • Linus Torvalds's avatar
      Merge tag 'mips_fixes_4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/jhogan/mips · dab30d55
      Linus Torvalds authored 
      Pull MIPS fixes from James Hogan:
 "A selection of important MIPS fixes for 4.14, and some MAINTAINERS /
  email address updates:

  Maintainership updates:
   - imgtec.com -> mips.com email addresses (this trivially updates
     comments in quite a few files, as well as MAINTAINERS)
   - Pistachio SoC maintainership update

  Fixes:
   - NI 169445 build (new platform in 4.14)
   - EVA regression (4.14)
   - SMP-CPS build & preemption regressions (4.14)
   - SMP/hotplug deadlock & race (deadlock reintroduced 4.13)
   - ebpf_jit error return (4.13)
   - SMP-CMP build regressions (4.11 and 4.14)
   - bad UASM microMIPS encoding (3.16)
   - CM definitions (3.15)"

[ I had taken the email address updates separately, because I didn't
  expect James to send a pull request, so those got applied twice.   - Linus]

* tag 'mips_fixes_4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/jhogan/mips:
  MIPS: Update email address for Marcin Nowakowski
  MIPS: smp-cmp: Fix vpe_id build error
  MAINTAINERS: Update Pistachio platform maintainers
  MIPS: smp-cmp: Use right include for task_struct
  MIPS: Update Goldfish RTC driver maintainer email address
  MIPS: Update RINT emulation maintainer email address
  MIPS: CPS: Fix use of current_cpu_data in preemptible code
  MIPS: SMP: Fix deadlock & online race
  MIPS: bpf: Fix a typo in build_one_insn()
  MIPS: microMIPS: Fix incorrect mask in insn_table_MM
  MIPS: Fix CM region target definitions
  MIPS: generic: Fix compilation error from include asm/mips-cpc.h
  MIPS: Fix exception entry when CONFIG_EVA enabled
  MIPS: generic: Fix NI 169445 its build
  Update MIPS email addresses
      dab30d55
    • Josh Poimboeuf's avatar
      objtool: Prevent GCC from merging annotate_unreachable(), take 2 · ec1e1b61
      Josh Poimboeuf authored 
      This fixes the following warning with GCC 4.6:

  mm/migrate.o: warning: objtool: migrate_misplaced_transhuge_page()+0x71: unreachable instruction

The problem is that the compiler merged identical annotate_unreachable()
inline asm blocks, resulting in a missing 'unreachable' annotation.

This problem happened before, and was partially fixed with:

  3d1e2360 ("objtool: Prevent GCC from merging annotate_unreachable()")

That commit tried to ensure that each instance of the
annotate_unreachable() inline asm statement has a unique label.  It used
the __LINE__ macro to generate the label number.  However, even the line
number isn't necessarily unique when used in an inline function with
multiple callers (in this case, __alloc_pages_node()'s use of
VM_BUG_ON).
Reported-by: default avatarkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: default avatarJosh Poimboeuf <jpoimboe@redhat.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: kbuild-all@01.org
Cc: tipbuild@zytor.com
Fixes: 3d1e2360 ("objtool: Prevent GCC from merging annotate_unreachable()")
Link: http://lkml.kernel.org/r/20171103221941.cajpwszir7ujxyc4@trebleSigned-off-by: default avatarIngo Molnar <mingo@kernel.org>
      ec1e1b61
    • Andy Lutomirski's avatar
      Revert "x86/mm: Stop calling leave_mm() in idle code" · 67535736
      Andy Lutomirski authored 
      This reverts commit 43858b4f.

The reason I removed the leave_mm() calls in question is because the
heuristic wasn't needed after that patch.  With the original version
of my PCID series, we never flushed a "lazy cpu" (i.e. a CPU running
kernel thread) due a flush on the loaded mm.

Unfortunately, that caused architectural issues, so now I've
reinstated these flushes on non-PCID systems in:

    commit b956575b ("x86/mm: Flush more aggressively in lazy TLB mode").

That, in turn, gives us a power management and occasionally
performance regression as compared to old kernels: a process that
goes into a deep idle state on a given CPU and gets its mm flushed
due to activity on a different CPU will wake the idle CPU.

Reinstate the old ugly heuristic: if a CPU goes into ACPI C3 or an
intel_idle state that is likely to cause a TLB flush gets its mm
switched to init_mm before going idle.

FWIW, this heuristic is lousy.  Whether we should change CR3 before
idle isn't a good hint except insofar as the performance hit is a bit
lower if the TLB is getting flushed by the idle code anyway.  What we
really want to know is whether we anticipate being idle long enough
that the mm is likely to be flushed before we wake up.  This is more a
matter of the expected latency than the idle state that gets chosen.
This heuristic also completely fails on systems that don't know
whether the TLB will be flushed (e.g. AMD systems?).  OTOH it may be a
bit obsolete anyway -- PCID systems don't presently benefit from this
heuristic at all.

We also shouldn't do this callback from innermost bit of the idle code
due to the RCU nastiness it causes.  All the information need is
available before rcu_idle_enter() needs to happen.
Signed-off-by: default avatarAndy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bpetkov@suse.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 43858b4f "x86/mm: Stop calling leave_mm() in idle code"
Link: http://lkml.kernel.org/r/c513bbd4e653747213e05bc7062de000bf0202a5.1509793738.git.luto@kernel.orgSigned-off-by: default avatarIngo Molnar <mingo@kernel.org>
      67535736
    • Frank Rowand's avatar
      Documentation: Add Frank Rowand to list of enforcement statement endorsers · e1b48c20
      Frank Rowand authored 
      Add my name to the list.
Signed-off-by: default avatarFrank Rowand <frank.rowand@sony.com>
Signed-off-by: default avatarGreg Kroah-Hartman <gregkh@linuxfoundation.org>
      e1b48c20
    • Willy Tarreau's avatar
      doc: add Willy Tarreau to the list of enforcement statement endorsers · 707f5a0f
      Willy Tarreau authored 
      add me to the list.
Signed-off-by: default avatarWilly Tarreau <w@1wt.eu>
Signed-off-by: default avatarGreg Kroah-Hartman <gregkh@linuxfoundation.org>
      707f5a0f
    • Ingo Molnar's avatar
      tools/headers: Synchronize kernel ABI headers · fb7df12d
      Ingo Molnar authored 
      After the SPDX license tags were added a number of tooling headers got out of
sync with their kernel variants, generating lots of build warnings.

Sync them:

 - tools/arch/x86/include/asm/disabled-features.h,
   tools/arch/x86/include/asm/required-features.h,
   tools/include/linux/hash.h:

     Remove the SPDX tag where the kernel version does not have it.

 - tools/include/asm-generic/bitops/__fls.h,
   tools/include/asm-generic/bitops/arch_hweight.h,
   tools/include/asm-generic/bitops/const_hweight.h,
   tools/include/asm-generic/bitops/fls.h,
   tools/include/asm-generic/bitops/fls64.h,
   tools/include/uapi/asm-generic/ioctls.h,
   tools/include/uapi/asm-generic/mman-common.h,
   tools/include/uapi/sound/asound.h,
   tools/include/uapi/linux/kvm.h,
   tools/include/uapi/linux/perf_event.h,
   tools/include/uapi/linux/sched.h,
   tools/include/uapi/linux/vhost.h,
   tools/include/uapi/sound/asound.h:

     Add the SPDX tag of the respective kernel header.

 - tools/include/uapi/linux/bpf_common.h,
   tools/include/uapi/linux/fcntl.h,
   tools/include/uapi/linux/hw_breakpoint.h,
   tools/include/uapi/linux/mman.h,
   tools/include/uapi/linux/stat.h,

     Change the tag to the kernel header version:

       -/* SPDX-License-Identifier: GPL-2.0 */
       +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */

Also sync other header details:

 - include/uapi/sound/asound.h:

     Fix pointless end of line whitespace noise the header grew in this cycle.

 - tools/arch/x86/lib/memcpy_64.S:

     Sync the code and add tools/include/asm/export.h with dummy wrappers
     to support building the kernel side code in a tooling header environment.

 - tools/include/uapi/asm-generic/mman.h,
   tools/include/uapi/linux/bpf.h:

     Sync other details that don't impact tooling's use of the ABIs.
Acked-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
Cc: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: default avatarIngo Molnar <mingo@kernel.org>
      fb7df12d
    • Josh Poimboeuf's avatar
      objtool: Resync objtool's instruction decoder source code copy with the kernel's latest version · da0db32b
      Josh Poimboeuf authored 
      This fixes the following warning:

  warning: objtool: x86 instruction decoder differs from kernel
Reported-by: default avatarStephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: default avatarJosh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/013315a808ccf5580abc293808827c8e2b5e1354.1509719152.git.jpoimboe@redhat.comSigned-off-by: default avatarIngo Molnar <mingo@kernel.org>
      da0db32b
    • Ingo Molnar's avatar
      Merge branch 'linus' into core/urgent, to pick up dependent commits · 649e441f
      Ingo Molnar authored 
      We want to fix an objtool build warning that got introduced in the latest upstream kernel.
Signed-off-by: default avatarIngo Molnar <mingo@kernel.org>
      649e441f
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7