[ Upstream commit a2e38dffcd93541914aba52b30c6a52acca35201 ]
Building with the Clang assembler shows the following warning:
arch/x86/kernel/ftrace_64.o: warning: objtool: missing symbol for insn at offset 0x16
The Clang assembler strips section symbols. That ends up giving
objtool's find_func_containing() much more test coverage than normal.
Turns out, find_func_containing() doesn't work so well for overlapping
symbols:
2: 000000000000000e 0 NOTYPE LOCAL DEFAULT 2 fgraph_trace
3: 000000000000000f 0 NOTYPE LOCAL DEFAULT 2 trace
4: 0000000000000000 165 FUNC GLOBAL DEFAULT 2 __fentry__
5: 000000000000000e 0 NOTYPE GLOBAL DEFAULT 2 ftrace_stub
The zero-length NOTYPE symbols are inside __fentry__(), confusing the
rbtree search for any __fentry__() offset coming after a NOTYPE.
Try to avoid this problem by not adding zero-length symbols to the
rbtree. They're rare and aren't needed in the rbtree anyway.
One caveat, this actually might not end up being the right fix.
Non-empty overlapping symbols, if they exist, could have the same
problem. But that would need bigger changes, let's see if we can get
away with the easy fix for now.
Reported-by: Arnd Bergmann <arnd@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 1d489151e9f9d1647110277ff77282fe4d96d09b upstream.
Thanks to a recent binutils change which doesn't generate unused
symbols, it's now possible for thunk_64.o be completely empty without
CONFIG_PREEMPTION: no text, no data, no symbols.
We could edit the Makefile to only build that file when
CONFIG_PREEMPTION is enabled, but that will likely create confusion
if/when the thunks end up getting used by some other code again.
Just ignore it and move on.
Reported-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/1254
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add the inline static call implementation for x86-64. The generated code
is identical to the out-of-line case, except we move the trampoline into
it's own section.
Objtool uses the trampoline naming convention to detect all the call
sites. It then annotates those call sites in the .static_call_sites
section.
During boot (and module init), the call sites are patched to call
directly into the destination function. The temporary trampoline is
then no longer used.
[peterz: merged trampolines, put trampoline in section]
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: https://lore.kernel.org/r/20200818135804.864271425@infradead.org
This provides infrastructure to rewrite instructions; this is
immediately useful for helping out with KCOV-vs-noinstr, but will
also come in handy for a bunch of variable sized jump-label patches
that are still on ice.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
With there being multiple ways to change the ELF data, let's more
concisely track modification.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Currently objtool only collects information about relocations with
addends. In recordmcount, which we are about to merge into objtool,
some supported architectures do not use rela relocations.
Signed-off-by: Matt Helsley <mhelsley@vmware.com>
Reviewed-by: Julien Thierry <jthierry@redhat.com>
Reviewed-by: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Before supporting additional relocation types rename the relevant
types and functions from "rela" to "reloc". This work be done with
the following regex:
sed -e 's/struct rela/struct reloc/g' \
-e 's/\([_\*]\)rela\(s\{0,1\}\)/\1reloc\2/g' \
-e 's/tmprela\(s\{0,1\}\)/tmpreloc\1/g' \
-e 's/relasec/relocsec/g' \
-e 's/rela_list/reloc_list/g' \
-e 's/rela_hash/reloc_hash/g' \
-e 's/add_rela/add_reloc/g' \
-e 's/rela->/reloc->/g' \
-e '/rela[,\.]/{ s/\([^\.>]\)rela\([\.,]\)/\1reloc\2/g ; }' \
-e 's/rela =/reloc =/g' \
-e 's/relas =/relocs =/g' \
-e 's/relas\[/relocs[/g' \
-e 's/relaname =/relocname =/g' \
-e 's/= rela\;/= reloc\;/g' \
-e 's/= relas\;/= relocs\;/g' \
-e 's/= relaname\;/= relocname\;/g' \
-e 's/, rela)/, reloc)/g' \
-e 's/\([ @]\)rela\([ "]\)/\1reloc\2/g' \
-e 's/ rela$/ reloc/g' \
-e 's/, relaname/, relocname/g' \
-e 's/sec->rela/sec->reloc/g' \
-e 's/(\(!\{0,1\}\)rela/(\1reloc/g' \
-i \
arch.h \
arch/x86/decode.c \
check.c \
check.h \
elf.c \
elf.h \
orc_gen.c \
special.c
Notable exceptions which complicate the regex include gelf_*
library calls and standard/expected section names which still use
"rela" because they encode the type of relocation expected. Also, keep
"rela" in the struct because it encodes a specific type of relocation
we currently expect.
It will eventually turn into a member of an anonymous union when a
susequent patch adds implicit addend, or "rel", relocation support.
Signed-off-by: Matt Helsley <mhelsley@vmware.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
ELF doesn't require .rela section names to match the base section. Use
the section index in sh_info to find the section instead of looking it
up by name.
LLD, for example, generates a .rela section that doesn't match the base
section name when we merge sections in a linker script for a binary
compiled with -ffunction-sections.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
If a .cold function is examined prior to it's parent, the link
to the parent/child function can be overwritten when the parent
is examined. Only update pfunc and cfunc if they were previously
nil to prevent this from happening.
This fixes an issue seen when compiling with -ffunction-sections.
Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Quoting Julien:
"And the other suggestion is my other email was that you don't even
need to add INSN_EXCEPTION_RETURN. You can keep IRET as
INSN_CONTEXT_SWITCH by default and x86 decoder lookups the symbol
conaining an iret. If it's a function symbol, it can just set the type
to INSN_OTHER so that it caries on to the next instruction after
having handled the stack_op."
Suggested-by: Julien Thierry <jthierry@redhat.com>
Signed-off-by: Miroslav Benes <mbenes@suse.cz>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Link: https://lkml.kernel.org/r/20200428191659.913283807@infradead.org
When doing kbuild tests to see if the objtool changes affected those I
found that there was a measurable regression:
pre post
real 1m13.594 1m16.488s
user 34m58.246s 35m23.947s
sys 4m0.393s 4m27.312s
Perf showed that for small files the increased hash-table sizes were a
measurable difference. Since we already have -l "vmlinux" to
distinguish between the modes, make it also use a smaller portion of
the hash-tables.
This flips it into a small win:
real 1m14.143s
user 34m49.292s
sys 3m44.746s
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Reviewed-by: Alexandre Chartre <alexandre.chartre@oracle.com>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Link: https://lkml.kernel.org/r/20200416115119.167588731@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Sometimes, WARN_FUNC() and other users of symbol_by_offset() will
associate the first instruction of a symbol with the symbol preceding
it. This is because symbol->offset + symbol->len is already outside of
the symbol's range.
Fixes: 2a362ecc3e ("objtool: Optimize find_symbol_*() and read_symbols()")
Signed-off-by: Julien Thierry <jthierry@redhat.com>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Perf shows there is significant time in find_rela_by_dest(); this is
because we have to iterate the address space per byte, looking for
relocation entries.
Optimize this by reducing the address space granularity.
This reduces objtool on vmlinux.o runtime from 4.8 to 4.4 seconds.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Link: https://lkml.kernel.org/r/20200324160924.861321325@infradead.org
Perf showed that __hash_init() is a significant portion of
read_sections(), so instead of doing a per section rela_hash, use an
elf-wide rela_hash.
Statistics show us there are about 1.1 million relas, so size it
accordingly.
This reduces the objtool on vmlinux.o runtime to a third, from 15 to 5
seconds.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Link: https://lkml.kernel.org/r/20200324160924.739153726@infradead.org
The symbol index is object wide, not per section, so it makes no sense
to have the symbol_hash be part of the section object. By moving it to
the elf object we avoid the linear sections iteration.
This reduces the runtime of objtool on vmlinux.o from over 3 hours (I
gave up) to a few minutes. The defconfig vmlinux.o has around 20k
sections.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Link: https://lkml.kernel.org/r/20200324160924.261852348@infradead.org
Based on 2 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version this program is distributed in the
hope that it will be useful but without any warranty without even
the implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details you
should have received a copy of the gnu general public license along
with this program if not see http www gnu org licenses
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version this program is distributed in the
hope that it will be useful but without any warranty without even
the implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details [based]
[from] [clk] [highbank] [c] you should have received a copy of the
gnu general public license along with this program if not see http
www gnu org licenses
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-or-later
has been chosen to replace the boilerplate/reference in 355 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: Jilayne Lovejoy <opensource@jilayne.com>
Reviewed-by: Steve Winslow <swinslow@gmail.com>
Reviewed-by: Allison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190519154041.837383322@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Function aliases result in different symbols for the same set of
instructions; track a canonical symbol so there is a unique point of
access.
This again prepares the way for function attributes. And in particular
the need for aliases comes from how KASAN uses them.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Because find_symbol_by_name() traverses the same lists as
read_symbols(), changing sym->name in place without copying it affects
the result of find_symbol_by_name(). In the case where a ".cold"
function precedes its parent in sec->symbol_list, it can result in a
function being considered a parent of itself. This leads to function
length being set to 0 and other consequent side-effects including a
segfault in add_switch_table(). The effects of this bug are only
visible when building with -ffunction-sections in KCFLAGS.
Fix by copying the search string instead of modifying it in place.
Signed-off-by: Artem Savkov <asavkov@redhat.com>
Signed-off-by: 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: 13810435b9 ("objtool: Support GCC 8's cold subfunctions")
Link: http://lkml.kernel.org/r/910abd6b5a4945130fd44f787c24e07b9e07c8da.1542736240.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Starting with GCC 8, a lot of unlikely code was moved out of line to
"cold" subfunctions in .text.unlikely.
For example, the unlikely bits of:
irq_do_set_affinity()
are moved out to the following subfunction:
irq_do_set_affinity.cold.49()
Starting with GCC 9, the numbered suffix has been removed. So in the
above example, the cold subfunction is instead:
irq_do_set_affinity.cold()
Tweak the objtool subfunction detection logic so that it detects both
GCC 8 and GCC 9 naming schemes.
Reported-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/015e9544b1f188d36a7f02fa31e9e95629aa5f50.1541040800.git.jpoimboe@redhat.com
Add support for processing switch jump tables in objects with multiple
.rodata sections, such as those created by '-ffunction-sections' and
'-fdata-sections'. Currently, objtool always looks in .rodata for jump
table information, which results in many "sibling call from callable
instruction with modified stack frame" warnings with objects compiled
using those flags.
The fix is comprised of three parts:
1. Flagging all .rodata sections when importing ELF information for
easier checking later.
2. Keeping a reference to the section each relocation is from in order
to get the list_head for the other relocations in that section.
3. Finding jump tables by following relocations to .rodata sections,
rather than always referencing a single global .rodata section.
The patch has been tested without data sections enabled and no
differences in the resulting orc unwind information were seen.
Note that as objtool adds terminators to end of each .text section the
unwind information generated between a function+data sections build and
a normal build aren't directly comparable. Manual inspection suggests
that objtool is now generating the correct information, or at least
making more of an effort to do so than it did previously.
Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
Since the following commit:
cd77849a69 ("objtool: Fix GCC 8 cold subfunction detection for aliased functions")
... if the kernel is built with EXTRA_CFLAGS='-fno-reorder-functions',
objtool can get stuck in an infinite loop.
That flag causes the new GCC 8 cold subfunctions to be placed in .text
instead of .text.unlikely. But it also has an unfortunate quirk: in the
symbol table, the subfunction (e.g., nmi_panic.cold.7) is nested inside
the parent (nmi_panic).
That function overlap confuses objtool, and causes it to get into an
infinite loop in next_insn_same_func(). Here's Allan's description of
the loop:
"Objtool iterates through the instructions in nmi_panic using
next_insn_same_func. Once it reaches the end of nmi_panic at 0x534 it
jumps to 0x528 as that's the start of nmi_panic.cold.7. However, since
the instructions starting at 0x528 are still associated with nmi_panic
objtool will get stuck in a loop, continually jumping back to 0x528
after reaching 0x534."
Fix it by shortening the length of the parent function so that the
functions no longer overlap.
Reported-and-analyzed-by: Allan Xavier <allan.x.xavier@oracle.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Allan Xavier <allan.x.xavier@oracle.com>
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: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/9e704c52bee651129b036be14feda317ae5606ae.1530136978.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
GCC 8 moves a lot of unlikely code out of line to "cold" subfunctions in
.text.unlikely. Properly detect the new subfunctions and treat them as
extensions of the original functions.
This fixes a bunch of warnings like:
kernel/cgroup/cgroup.o: warning: objtool: parse_cgroup_root_flags()+0x33: sibling call from callable instruction with modified stack frame
kernel/cgroup/cgroup.o: warning: objtool: cgroup_addrm_files()+0x290: sibling call from callable instruction with modified stack frame
kernel/cgroup/cgroup.o: warning: objtool: cgroup_apply_control_enable()+0x25b: sibling call from callable instruction with modified stack frame
kernel/cgroup/cgroup.o: warning: objtool: rebind_subsystems()+0x325: sibling call from callable instruction with modified stack frame
Reported-and-tested-by: damian <damian.tometzki@icloud.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: David Laight <David.Laight@ACULAB.COM>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/0965e7fcfc5f31a276f0c7f298ff770c19b68706.1525923412.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Arnd Bergmann reported that a randconfig build was failing with the
following link error:
built-in.o: member arch/x86/kernel/time.o in archive is not an object
It turns out the link failed because the time.o file had been corrupted
by objtool:
nm: arch/x86/kernel/time.o: File format not recognized
In certain rare cases, when a .o file's ORC table is very small, the
.data section size doesn't change because it's page aligned. Because
all the existing sections haven't changed size, libelf doesn't detect
any section header changes, and so it doesn't update the section header
table properly. Instead it writes junk in the section header entries
for the new ORC sections.
Make sure libelf properly updates the section header table by setting
the ELF_F_DIRTY flag in the top level elf struct.
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: 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: 627fce1480 ("objtool: Add ORC unwind table generation")
Link: http://lkml.kernel.org/r/e650fd0f2d8a209d1409a9785deb101fdaed55fb.1505459813.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Arnd Bergmann reported a (false positive) objtool warning:
drivers/infiniband/sw/rxe/rxe_resp.o: warning: objtool: rxe_responder()+0xfe: sibling call from callable instruction with changed frame pointer
The issue is in find_switch_table(). It tries to find a switch
statement's jump table by walking backwards from an indirect jump
instruction, looking for a relocation to the .rodata section. In this
case it stopped walking prematurely: the first .rodata relocation it
encountered was for a variable (resp_state_name) instead of a jump
table, so it just assumed there wasn't a jump table.
The fix is to ignore any .rodata relocation which refers to an ELF
object symbol. This works because the jump tables are anonymous and
have no symbols associated with them.
Reported-by: Arnd Bergmann <arnd@arndb.de>
Tested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 3732710ff6 ("objtool: Improve rare switch jump table pattern detection")
Link: http://lkml.kernel.org/r/20170302225723.3ndbsnl4hkqbne7a@treble
Signed-off-by: Ingo Molnar <mingo@kernel.org>
