linux-stable.git/tools/objtool/builtin-check.c, branch linux-4.8.y

objtool: Support new GCC 6 switch jump table pattern

2016-07-29T14:56:21+00:00

This fixes some false positive objtool warnings seen with gcc 6.1.1:

  kernel/trace/ring_buffer.o: warning: objtool: ring_buffer_read_page()+0x36c: sibling call from callable instruction with changed frame pointer
  arch/x86/kernel/reboot.o: warning: objtool: native_machine_emergency_restart()+0x139: sibling call from callable instruction with changed frame pointer
  lib/xz/xz_dec_stream.o: warning: objtool: xz_dec_run()+0xc2: sibling call from callable instruction with changed frame pointer

With GCC 6, a new code pattern is sometimes used to access a switch
statement jump table in .rodata, which objtool doesn't yet recognize:

  mov [rodata addr],%reg1
  ... some instructions ...
  jmpq *(%reg1,%reg2,8)

Add support for detecting that pattern.  The detection code is rather
crude, but it's still effective at weeding out false positives and
catching real warnings.  It can be refined later once objtool starts
reading DWARF CFI.

Reported-by: Linus Torvalds 
Signed-off-by: Josh Poimboeuf 
Cc: Andy Lutomirski 
Cc: Borislav Petkov 
Cc: Brian Gerst 
Cc: Denys Vlasenko 
Cc: H. Peter Anvin 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/b8c9503b4ad8c8a827cc5400db4c1b40a3ea07bc.1469751119.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar

Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2016-07-25T20:20:41+00:00

Pull perf updates from Ingo Molnar:
 "With over 300 commits it's been a busy cycle - with most of the work
  concentrated on the tooling side (as it should).

  The main kernel side enhancements were:

   - Add per event callchain limit: Recently we introduced a sysctl to
     tune the max-stack for all events for which callchains were
     requested:

       $ sysctl kernel.perf_event_max_stack
       kernel.perf_event_max_stack = 127

     Now this patch introduces a way to configure this per event, i.e.
     this becomes possible:

       $ perf record -e sched:*/max-stack=2/ -e block:*/max-stack=10/ -a

     allowing finer tuning of how much buffer space callchains use.

     This uses an u16 from the reserved space at the end, leaving
     another u16 for future use.

     There has been interest in even finer tuning, namely to control the
     max stack for kernel and userspace callchains separately.  Further
     discussion is needed, we may for instance use the remaining u16 for
     that and when it is present, assume that the sample_max_stack
     introduced in this patch applies for the kernel, and the u16 left
     is used for limiting the userspace callchain (Arnaldo Carvalho de
     Melo)

   - Optimize AUX event (hardware assisted side-band event) delivery
     (Kan Liang)

   - Rework Intel family name macro usage (this is partially x86 arch
     work) (Dave Hansen)

   - Refine and fix Intel LBR support (David Carrillo-Cisneros)

   - Add support for Intel 'TopDown' events (Andi Kleen)

   - Intel uncore PMU driver fixes and enhancements (Kan Liang)

   - ... other misc changes.

  Here's an incomplete list of the tooling enhancements (but there's
  much more, see the shortlog and the git log for details):

   - Support cross unwinding, i.e.  collecting '--call-graph dwarf'
     perf.data files in one machine and then doing analysis in another
     machine of a different hardware architecture.  This enables, for
     instance, to do:

       $ perf record -a --call-graph dwarf

     on a x86-32 or aarch64 system and then do 'perf report' on it on a
     x86_64 workstation (He Kuang)

   - Allow reading from a backward ring buffer (one setup via
     sys_perf_event_open() with perf_event_attr.write_backward = 1)
     (Wang Nan)

   - Finish merging initial SDT (Statically Defined Traces) support, see
     cset comments for details about how it all works (Masami Hiramatsu)

   - Support attaching eBPF programs to tracepoints (Wang Nan)

   - Add demangling of symbols in programs written in the Rust language
     (David Tolnay)

   - Add support for tracepoints in the python binding, including an
     example, that sets up and parses sched:sched_switch events,
     tools/perf/python/tracepoint.py (Jiri Olsa)

   - Introduce --stdio-color to set up the color output mode selection
     in 'annotate' and 'report', allowing emit color escape sequences
     when redirecting the output of these tools (Arnaldo Carvalho de
     Melo)

   - Add 'callindent' option to 'perf script -F', to indent the Intel PT
     call stack, making this output more ftrace-like (Adrian Hunter,
     Andi Kleen)

   - Allow dumping the object files generated by llvm when processing
     eBPF scriptlet events (Wang Nan)

   - Add stackcollapse.py script to help generating flame graphs (Paolo
     Bonzini)

   - Add --ldlat option to 'perf mem' to specify load latency for loads
     event (e.g. cpu/mem-loads/ ) (Jiri Olsa)

   - Tooling support for Intel TopDown counters, recently added to the
     kernel (Andi Kleen)"

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (303 commits)
  perf tests: Add is_printable_array test
  perf tools: Make is_printable_array global
  perf script python: Fix string vs byte array resolving
  perf probe: Warn unmatched function filter correctly
  perf cpu_map: Add more helpers
  perf stat: Balance opening and reading events
  tools: Copy linux/{hash,poison}.h and check for drift
  perf tools: Remove include/linux/list.h from perf's MANIFEST
  tools: Copy the bitops files accessed from the kernel and check for drift
  Remove: kernel unistd*h files from perf's MANIFEST, not used
  perf tools: Remove tools/perf/util/include/linux/const.h
  perf tools: Remove tools/perf/util/include/asm/byteorder.h
  perf tools: Add missing linux/compiler.h include to perf-sys.h
  perf jit: Remove some no-op error handling
  perf jit: Add missing curly braces
  objtool: Initialize variable to silence old compiler
  objtool: Add -I$(srctree)/tools/arch/$(ARCH)/include/uapi
  perf record: Add --tail-synthesize option
  perf session: Don't warn about out of order event if write_backward is used
  perf tools: Enable overwrite settings
  ...

objtool: Initialize variable to silence old compiler

2016-07-15T20:32:52+00:00

gcc version 4.1.2 20080704 (Red Hat 4.1.2-55) barfs with:

    CC       /tmp/build/objtool/builtin-check.o
  cc1: warnings being treated as errors
  builtin-check.c: In function 'cmd_check':
  builtin-check.c:667: warning: 'prev_rela' may be used uninitialized in this function
  mv: cannot stat `/tmp/build/objtool/.builtin-check.o.tmp': No such file or directory
  make[1]: *** [/tmp/build/objtool/builtin-check.o] Error 1

Init it to NULL to silence it.

Cc: Adrian Hunter 
Cc: David Ahern 
Cc: Jiri Olsa 
Cc: Josh Poimboeuf 
Cc: Namhyung Kim 
Cc: Wang Nan 
Link: http://lkml.kernel.org/n/tip-qolo31rl2ojlwj1lj9dhemyz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo

perf tools: Uninline scnprintf() and vscnprint()

2016-07-12T18:20:24+00:00

They were in tools/include/linux/kernel.h, requiring that it in turn
included stdio.h, which is way too heavy.

Cc: Adrian Hunter 
Cc: David Ahern 
Cc: Jiri Olsa 
Cc: Josh Poimboeuf 
Cc: Namhyung Kim 
Cc: Wang Nan 
Link: http://lkml.kernel.org/n/tip-855h8olnkot9v0dajuee1lo3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo

objtool: Fix STACK_FRAME_NON_STANDARD macro checking for function symbols

2016-07-10T15:15:58+00:00

Mathieu Desnoyers reported that the STACK_FRAME_NON_STANDARD macro
wasn't working with the lttng_filter_interpret_bytecode() function in
the lttng-modules code.

Usually the relocation created by STACK_FRAME_NON_STANDARD creates a
reference to a section symbol like this:

  Offset              Type            Value               Addend Name
  000000000000000000  X86_64_64       000000000000000000   +3136 .text

But in this case it created a reference to a function symbol:

  Offset              Type            Value               Addend Name
  000000000000000000  X86_64_64       0x00000000000003a0      +0 lttng_filter_interpret_bytecode

To be honest I have no idea what causes gcc to decide to do one over the
other.  But both are valid ELF, so add support for the function symbol.

Reported-by: Mathieu Desnoyers 
Signed-off-by: Josh Poimboeuf 
Cc: Linus Torvalds 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: lttng-dev@lists.lttng.org
Link: http://lkml.kernel.org/r/9cee42843bc6d94e990a152e4e0319cfdf6756ef.1466023450.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar

objtool: Detect falling through to the next function

2016-04-16T09:14:17+00:00

There are several cases in compiled C code where a function may not
return at the end, and may instead fall through to the next function.

That may indicate a bug in the code, or a gcc bug, or even an objtool
bug.  But in each case, objtool reports an unhelpful warning, something
like:

  drivers/scsi/qla2xxx/qla_attr.o: warning: objtool: qla2x00_get_fc_host_stats()+0x0: duplicate frame pointer save
  drivers/scsi/qla2xxx/qla_attr.o: warning: objtool: qla2x00_get_fc_host_stats()+0x0: frame pointer state mismatch

Detect this situation and print a more useful error message:

  drivers/scsi/qla2xxx/qla_attr.o: warning: objtool: qla2x00_get_host_fabric_name() falls through to next function qla2x00_get_starget_node_name()

Also add some information about this warning and its potential causes to
the documentation.

Reported-by: kbuild test robot 
Signed-off-by: Josh Poimboeuf 
Cc: Alexander Shishkin 
Cc: Andy Lutomirski 
Cc: Arnaldo Carvalho de Melo 
Cc: Borislav Petkov 
Cc: Brian Gerst 
Cc: Denys Vlasenko 
Cc: H. Peter Anvin 
Cc: Jiri Olsa 
Cc: Linus Torvalds 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/caa4ec6c687931db805e692d4e4bf06cd87d33e6.1460729697.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar

objtool: Add workaround for GCC switch jump table bug

2016-04-15T09:42:13+00:00

GCC has a rare quirk, currently only seen in three driver functions in
the kernel, and only with certain obscure non-distro configs, which can
cause objtool to produce "unreachable instruction" false positive
warnings.

As part of an optimization, GCC makes a copy of an existing switch jump
table, modifies it, and then hard-codes the jump (albeit with an
indirect jump) to use a single entry in the table.  The rest of the jump
table and some of its jump targets remain as dead code.

In such a case we can just crudely ignore all unreachable instruction
warnings for the entire object file.  Ideally we would just ignore them
for the function, but that would require redesigning the code quite a
bit.  And honestly that's just not worth doing: unreachable instruction
warnings are of questionable value anyway, and this is a very rare
issue.

kbuild reports:

  https://lkml.kernel.org/r/201603231906.LWcVUpxm%25fengguang.wu@intel.com
  https://lkml.kernel.org/r/201603271114.K9i45biy%25fengguang.wu@intel.com
  https://lkml.kernel.org/r/201603291058.zuJ6ben1%25fengguang.wu@intel.com

GCC bug:

  https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70604

Reported-by: kbuild test robot 
Signed-off-by: Josh Poimboeuf 
Cc: Linus Torvalds 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/700fa029bbb0feff34f03ffc69d666a3c3b57a61.1460663532.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar

objtool: Only print one warning per function

2016-03-09T09:48:10+00:00

When objtool discovers an issue, it's very common for it to flood the
terminal with a lot of duplicate warnings.  For example:

  warning: objtool: rtlwifi_rate_mapping()+0x2e7: frame pointer state mismatch
  warning: objtool: rtlwifi_rate_mapping()+0x2f3: frame pointer state mismatch
  warning: objtool: rtlwifi_rate_mapping()+0x2ff: frame pointer state mismatch
  warning: objtool: rtlwifi_rate_mapping()+0x30b: frame pointer state mismatch
  ...

The first warning is usually all you need.  Change it to only warn once
per function.

Suggested-by: Ingo Molnar 
Signed-off-by: Josh Poimboeuf 
Cc: Andrew Morton 
Cc: Andy Lutomirski 
Cc: Arnaldo Carvalho de Melo 
Cc: Arnaldo Carvalho de Melo 
Cc: Bernd Petrovitsch 
Cc: Borislav Petkov 
Cc: Chris J Arges 
Cc: Jiri Slaby 
Cc: Linus Torvalds 
Cc: Michal Marek 
Cc: Namhyung Kim 
Cc: Pedro Alves 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: live-patching@vger.kernel.org
Link: http://lkml.kernel.org/r/c47f3ca38aa01e2a9b6601f9e38efd414c3f3c18.1457502970.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar

objtool: Add several performance improvements

2016-03-09T09:48:10+00:00

Use hash tables for instruction and rela lookups (and keep the linked
lists around for sequential access).

Also cache the section struct for the "__func_stack_frame_non_standard"
section.

With this change, "objtool check net/wireless/nl80211.o" goes from:

  real	0m1.168s
  user	0m1.163s
  sys	0m0.005s

to:

  real	0m0.059s
  user	0m0.042s
  sys	0m0.017s

for a 20x speedup.

With the same object, it should be noted that the memory heap usage grew
from 8MB to 62MB.  Reducing the memory usage is on the TODO list.

Reported-by: Ingo Molnar 
Signed-off-by: Josh Poimboeuf 
Cc: Andrew Morton 
Cc: Andy Lutomirski 
Cc: Arnaldo Carvalho de Melo 
Cc: Arnaldo Carvalho de Melo 
Cc: Bernd Petrovitsch 
Cc: Borislav Petkov 
Cc: Chris J Arges 
Cc: Jiri Slaby 
Cc: Linus Torvalds 
Cc: Michal Marek 
Cc: Namhyung Kim 
Cc: Pedro Alves 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: live-patching@vger.kernel.org
Link: http://lkml.kernel.org/r/dd0d8e1449506cfa7701b4e7ba73577077c44253.1457502970.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar

objtool: Fix false positive warnings for functions with multiple switch statements

2016-03-09T09:48:09+00:00

Ingo reported [1] some false positive objtool warnings:

  drivers/net/wireless/realtek/rtlwifi/base.o: warning: objtool: rtlwifi_rate_mapping()+0x2e7: frame pointer state mismatch
  drivers/net/wireless/realtek/rtlwifi/base.o: warning: objtool: rtlwifi_rate_mapping()+0x2f3: frame pointer state mismatch
  ...

And so did the 0-day bot [2]:

  drivers/gpu/drm/radeon/cik.o: warning: objtool: cik_tiling_mode_table_init()+0x6ce: call without frame pointer save/setup
  drivers/gpu/drm/radeon/cik.o: warning: objtool: cik_tiling_mode_table_init()+0x72b: call without frame pointer save/setup
  ...

Both sets of warnings involve functions which have multiple switch
statements.  When there's more than one switch statement in a function,
objtool interprets all the switch jump tables as a single table.  If the
targets of one jump table assume a stack frame and the targets of
another one don't, it prints false positive warnings.

Fix the bug by detecting the size of each switch jump table.  For
multiple tables, each one ends where the next one begins.

[1] https://lkml.kernel.org/r/20160308103716.GA9618@gmail.com
[2] https://lists.01.org/pipermail/kbuild-all/2016-March/018124.html

Reported-by: Ingo Molnar 
Reported-by: kbuild test robot 
Signed-off-by: Josh Poimboeuf 
Cc: Andrew Morton 
Cc: Andy Lutomirski 
Cc: Arnaldo Carvalho de Melo 
Cc: Arnaldo Carvalho de Melo 
Cc: Bernd Petrovitsch 
Cc: Borislav Petkov 
Cc: Chris J Arges 
Cc: Jiri Slaby 
Cc: Linus Torvalds 
Cc: Michal Marek 
Cc: Namhyung Kim 
Cc: Pedro Alves 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: live-patching@vger.kernel.org
Link: http://lkml.kernel.org/r/2d7eecc6bc52d301f494b80f5fd62c2b6c895658.1457502970.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar