Skip to content

L
linux
Switch branch/tag
  1. 08 Jul, 2020 2 commits
    • Steve MacLean's avatar
      perf inject jit: Remove //anon mmap events · c8f6ae1f
      Steve MacLean authored 
      **perf-<pid>.map and jit-<pid>.dump designs:

When a JIT generates code to be executed, it must allocate memory and
mark it executable using an mmap call.

*** perf-<pid>.map design

The perf-<pid>.map assumes that any sample recorded in an anonymous
memory page is JIT code. It then tries to resolve the symbol name by
looking at the process' perf-<pid>.map.

*** jit-<pid>.dump design

The jit-<pid>.dump mechanism takes a different approach. It requires a
JIT to write a `<path>/jit-<pid>.dump` file. This file must also be
mmapped so that perf inject -jit can find the file. The JIT must also
add JIT_CODE_LOAD records for any functions it generates. The records
are timestamped using a clock which can be correlated to the perf record
clock.

After perf record,  the `perf inject -jit` pass parses the recording
looking for a `<path>/jit-<pid>.dump` file. When it finds the file, it
parses it and for each JIT_CODE_LOAD record:
* creates an elf file `<path>/jitted-<pid>-<code_index>.so
* injects a new mmap record mapping the new elf file into the process.

*** Coexistence design

The kernel and perf support both of these mechanisms. We need to make
sure perf works on an app supporting either or both of these mechanisms.
Both designs rely on mmap records to determine how to resolve an ip
address.

The mmap records of both techniques by definition overlap. When the JIT
compiles a method, it must:

* allocate memory (mmap)
* add execution privilege (mprotect or mmap. either will
generate an mmap event form the kernel to perf)
* compile code into memory
* add a function record to perf-<pid>.map and/or jit-<pid>.dump

Because the jit-<pid>.dump mechanism supports greater capabilities, perf
prefers the symbols from jit-<pid>.dump. It implements this based on
timestamp ordering of events. There is an implicit ASSUMPTION that the
JIT_CODE_LOAD record timestamp will be after the // anon mmap event that
was generated during memory allocation or adding the execution privilege setting.

*** Problems with the ASSUMPTION

The ASSUMPTION made in the Coexistence design section above is violated
in the following scenario.

*** Scenario

While a JIT is jitting code it will eventually need to commit more
pages and change these pages to executable permissions. Typically the
JIT will want these collocated to minimize branch displacements.

The kernel will coalesce these anonymous mapping with identical
permissions before sending an MMAP event for the new pages. The address
range of the new mmap will not be just the most recently mmap pages.
It will include the entire coalesced mmap region.

See mm/mmap.c

unsigned long mmap_region(struct file *file, unsigned long addr,
                unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
                struct list_head *uf)
{
...
        /*
         * Can we just expand an old mapping?
         */
...
        perf_event_mmap(vma);
...
}

*** Symptoms

The coalesced // anon mmap event will be timestamped after the
JIT_CODE_LOAD records. This means it will be used as the most recent
mapping for that entire address range. For remaining events it will look
at the inferior perf-<pid>.map for symbols.

If both mechanisms are supported, the symbol will appear twice with
different module names. This causes weird behavior in reporting.

If only jit-<pid>.dump is supported, the symbol will no longer be resolved.

** Implemented solution

This patch solves the issue by removing // anon mmap events for any
process which has a valid jit-<pid>.dump file.

It tracks on a per process basis to handle the case where some running
apps support jit-<pid>.dump, but some only support perf-<pid>.map.

It adds new assumptions:
* // anon mmap events are only required for perf-<pid>.map support.
* An app that uses jit-<pid>.dump, no longer needs
perf-<pid>.map support. It assumes that any perf-<pid>.map info is
inferior.

*** Details

Use thread->priv to store whether a jitdump file has been processed

During "perf inject --jit", discard "//anon*" mmap events for any pid which
has sucessfully processed a jitdump file.

** Testing:

// jitdump case

  perf record <app with jitdump>
  perf inject --jit --input perf.data --output perfjit.data

// verify mmap "//anon" events present initially

  perf script --input perf.data --show-mmap-events | grep '//anon'

// verify mmap "//anon" events removed

  perf script --input perfjit.data --show-mmap-events | grep '//anon'

// no jitdump case

  perf record <app without jitdump>
  perf inject --jit --input perf.data --output perfjit.data

// verify mmap "//anon" events present initially

  perf script --input perf.data --show-mmap-events | grep '//anon'

// verify mmap "//anon" events not removed

  perf script --input perfjit.data --show-mmap-events | grep '//anon'

** Repro:

This issue was discovered while testing the initial CoreCLR jitdump
implementation. https://github.com/dotnet/coreclr/pull/26897.

** Alternate solutions considered

These were also briefly considered:

* Change kernel to not coalesce mmap regions.

* Change kernel reporting of coalesced mmap regions to perf. Only
include newly mapped memory.

* Only strip parts of // anon mmap events overlapping existing
jitted-<pid>-<code_index>.so mmap events.
Signed-off-by: default avatarSteve MacLean <Steve.MacLean@Microsoft.com>
Acked-by: default avatarIan Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/1590544271-125795-1-git-send-email-steve.maclean@linux.microsoft.comSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      c8f6ae1f
    • Arnaldo Carvalho de Melo's avatar
      Merge remote-tracking branch 'torvalds/master' into perf/core · facbf0b9
      Arnaldo Carvalho de Melo authored 
      To pick up fixes and move perf/core forward, minor conflict as
perf_evlist__add_dummy() lost its 'perf_' prefix as it operates on a
'struct evlist', not on a 'struct perf_evlist', i.e. its tools/perf/
specific, it is not in libperf.
Signed-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      facbf0b9
  3. 07 Jul, 2020 8 commits
  5. 06 Jul, 2020 24 commits
  7. 05 Jul, 2020 6 commits
    • Linus Torvalds's avatar
      Linux 5.8-rc4 · dcb7fd82
      Linus Torvalds authored
      dcb7fd82
    • Linus Torvalds's avatar
      x86/ldt: use "pr_info_once()" instead of open-coding it badly · bb5a93aa
      Linus Torvalds authored 
      Using a mutex for "print this warning only once" is so overdesigned as
to be actively offensive to my sensitive stomach.

Just use "pr_info_once()" that already does this, although in a
(harmlessly) racy manner that can in theory cause the message to be
printed twice if more than one CPU races on that "is this the first
time" test.

[ If somebody really cares about that harmless data race (which sounds
  very unlikely indeed), that person can trivially fix printk_once() by
  using a simple atomic access, preferably with an optimistic non-atomic
  test first before even bothering to treat the pointless "make sure it
  is _really_ just once" case.

  A mutex is most definitely never the right primitive to use for
  something like this. ]

Yes, this is a small and meaningless detail in a code path that hardly
matters.  But let's keep some code quality standards here, and not
accept outrageously bad code.

Link: https://lore.kernel.org/lkml/CAHk-=wgV9toS7GU3KmNpj8hCS9SeF+A0voHS8F275_mgLhL4Lw@mail.gmail.com/
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
      bb5a93aa
    • Linus Torvalds's avatar
      Merge tag 'x86-urgent-2020-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip · 72674d48
      Linus Torvalds authored 
      Pull x86 fixes from Thomas Gleixner:
 "A series of fixes for x86:

   - Reset MXCSR in kernel_fpu_begin() to prevent using a stale user
     space value.

   - Prevent writing MSR_TEST_CTRL on CPUs which are not explicitly
     whitelisted for split lock detection. Some CPUs which do not
     support it crash even when the MSR is written to 0 which is the
     default value.

   - Fix the XEN PV fallout of the entry code rework

   - Fix the 32bit fallout of the entry code rework

   - Add more selftests to ensure that these entry problems don't come
     back.

   - Disable 16 bit segments on XEN PV. It's not supported because XEN
     PV does not implement ESPFIX64"

* tag 'x86-urgent-2020-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/ldt: Disable 16-bit segments on Xen PV
  x86/entry/32: Fix #MC and #DB wiring on x86_32
  x86/entry/xen: Route #DB correctly on Xen PV
  x86/entry, selftests: Further improve user entry sanity checks
  x86/entry/compat: Clear RAX high bits on Xen PV SYSENTER
  selftests/x86: Consolidate and fix get/set_eflags() helpers
  selftests/x86/syscall_nt: Clear weird flags after each test
  selftests/x86/syscall_nt: Add more flag combinations
  x86/entry/64/compat: Fix Xen PV SYSENTER frame setup
  x86/entry: Move SYSENTER's regs->sp and regs->flags fixups into C
  x86/entry: Assert that syscalls are on the right stack
  x86/split_lock: Don't write MSR_TEST_CTRL on CPUs that aren't whitelisted
  x86/fpu: Reset MXCSR to default in kernel_fpu_begin()
      72674d48
    • Linus Torvalds's avatar
      Merge tag 'irq-urgent-2020-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip · f23dbe18
      Linus Torvalds authored 
      Pull irq fixes from Thomas Gleixner:
 "A set of interrupt chip driver fixes:

   - Ensure the atomicity of affinity updates in the GIC driver

   - Don't try to sleep in atomic context when waiting for the GICv4.1
     to respond. Use polling instead.

   - Typo fixes in Kconfig and warnings"

* tag 'irq-urgent-2020-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/gic: Atomically update affinity
  irqchip/riscv-intc: Fix a typo in a pr_warn()
  irqchip/gic-v4.1: Use readx_poll_timeout_atomic() to fix sleep in atomic
  irqchip/loongson-pci-msi: Fix a typo in Kconfig
      f23dbe18
    • Linus Torvalds's avatar
      Merge tag 'core-urgent-2020-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip · 5465a324
      Linus Torvalds authored 
      Pull rcu fixlet from Thomas Gleixner:
 "A single fix for a printk format warning in RCU"

* tag 'core-urgent-2020-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  rcuperf: Fix printk format warning
      5465a324
    • Linus Torvalds's avatar
      Merge tag 'kbuild-fixes-v5.8-2' of... · 4bc92736
      Linus Torvalds authored 
      Merge tag 'kbuild-fixes-v5.8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes frin Masahiro Yamada:

 - fix various bugs in xconfig

 - fix some issues in cross-compilation using Clang

 - fix documentation

* tag 'kbuild-fixes-v5.8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  .gitignore: Do not track `defconfig` from `make savedefconfig`
  kbuild: make Clang build userprogs for target architecture
  kbuild: fix CONFIG_CC_CAN_LINK(_STATIC) for cross-compilation with Clang
  kconfig: qconf: parse newer types at debug info
  kconfig: qconf: navigate menus on hyperlinks
  kconfig: qconf: don't show goback button on splitMode
  kconfig: qconf: simplify the goBack() logic
  kconfig: qconf: re-implement setSelected()
  kconfig: qconf: make debug links work again
  kconfig: qconf: make search fully work again on split mode
  kconfig: qconf: cleanup includes
  docs: kbuild: fix ReST formatting
  gcc-plugins: fix gcc-plugins directory path in documentation
      4bc92736
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7