linux-toradex.git/tools/objtool, branch v6.19-rc2

objtool: Add more robust signal error handling, detect and warn about stack overflows

2025-12-03T18:42:37+00:00

When the kernel build fails due to an objtool segfault, the error
message is a bit obtuse and confusing:

  make[5]: *** [scripts/Makefile.build:503: drivers/scsi/qla2xxx/qla2xxx.o] Error 139
                                                                            ^^^^^^^^^
  make[5]: *** Deleting file 'drivers/scsi/qla2xxx/qla2xxx.o'
  make[4]: *** [scripts/Makefile.build:556: drivers/scsi/qla2xxx] Error 2
  make[3]: *** [scripts/Makefile.build:556: drivers/scsi] Error 2
  make[2]: *** [scripts/Makefile.build:556: drivers] Error 2
  make[1]: *** [/home/jpoimboe/git/linux/Makefile:2013: .] Error 2
  make: *** [Makefile:248: __sub-make] Error 2

Add a signal handler to objtool which prints an error message like if
the local stack has overflown (for which there's a chance as objtool
makes heavy use of recursion):

  drivers/scsi/qla2xxx/qla2xxx.o: error: SIGSEGV: objtool stack overflow!

or:

  drivers/scsi/qla2xxx/qla2xxx.o: error: SIGSEGV: objtool crash!

Also, re-raise the signal so the core dump still gets triggered.

[ mingo: Applied a build fix, added more comments and prettified the code. ]

Suggested-by: Ingo Molnar 
Signed-off-by: Josh Poimboeuf 
Signed-off-by: Ingo Molnar 
Cc: Alexandre Chartre 
Cc: David Laight 
Cc: Linus Torvalds 
Cc: Nathan Chancellor 
Cc: Peter Zijlstra 
Link: https://patch.msgid.link/mi4tihk4dbncn7belrhp6ooudhpw4vdggerktu5333w3gqf3uf@vqlhc3y667mg

objtool: Fix stack overflow in validate_branch()

2025-12-02T16:40:35+00:00

On an allmodconfig kernel compiled with Clang, objtool is segfaulting in
drivers/scsi/qla2xxx/qla2xxx.o due to a stack overflow in
validate_branch().

Due in part to KASAN being enabled, the qla2xxx code has a large number
of conditional jumps, causing objtool to go quite deep in its recursion.

By far the biggest offender of stack usage is the recently added
'prev_state' stack variable in validate_insn(), coming in at 328 bytes.

Move that variable (and its tracing usage) to handle_insn_ops() and make
handle_insn_ops() noinline to keep its stack frame outside the recursive
call chain.

Reported-by: Nathan Chancellor 
Fixes: fcb268b47a2f ("objtool: Trace instruction state changes during function validation")
Signed-off-by: Josh Poimboeuf 
Signed-off-by: Ingo Molnar 
Link: https://patch.msgid.link/21bb161c23ca0d8c942a960505c0d327ca2dc7dc.1764691895.git.jpoimboe@kernel.org
Closes: https://lore.kernel.org/20251201202329.GA3225984@ax162

objtool: Fix segfault on unknown alternatives

2025-12-01T09:42:27+00:00

So 'objtool --link -d vmlinux.o' gets surprised by this endbr64+endbr64 pattern
in ___bpf_prog_run():

	___bpf_prog_run:
	1e7680:  ___bpf_prog_run+0x0                                                     push   %r12
	1e7682:  ___bpf_prog_run+0x2                                                     mov    %rdi,%r12
	1e7685:  ___bpf_prog_run+0x5                                                     push   %rbp
	1e7686:  ___bpf_prog_run+0x6                                                     xor    %ebp,%ebp
	1e7688:  ___bpf_prog_run+0x8                                                     push   %rbx
	1e7689:  ___bpf_prog_run+0x9                                                     mov    %rsi,%rbx
	1e768c:  ___bpf_prog_run+0xc                                                     movzbl (%rbx),%esi
	1e768f:  ___bpf_prog_run+0xf                                                     movzbl %sil,%edx
	1e7693:  ___bpf_prog_run+0x13                                                    mov    %esi,%eax
	1e7695:  ___bpf_prog_run+0x15                                                    mov    0x0(,%rdx,8),%rdx
	1e769d:  ___bpf_prog_run+0x1d                                                    jmp    0x1e76a2 <__x86_indirect_thunk_rdx>
	1e76a2:  ___bpf_prog_run+0x22                                                    endbr64
	1e76a6:  ___bpf_prog_run+0x26                                                    endbr64
	1e76aa:  ___bpf_prog_run+0x2a                                                    mov    0x4(%rbx),%edx

And crashes due to blindly dereferencing alt->insn->alt_group.

Bail out on NULL ->alt_group, which produces this warning and continues
with the disassembly, instead of a segfault:

  .git/O/vmlinux.o: warning: objtool: : failed to disassemble alternative

Cc: Alexandre Chartre 
Cc: Peter Zijlstra (Intel) 
Cc: Josh Poimboeuf 
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar

objtool: Build with disassembly can fail when including bdf.h

2025-11-27T08:32:46+00:00

Building objtool with disassembly support can fail when including
the bdf.h file:

  In file included from tools/objtool/include/objtool/arch.h:108,
                   from check.c:14:
  /usr/include/bfd.h:35:2: error: #error config.h must be included before this header
     35 | #error config.h must be included before this header
        |  ^~~~~

This check is present in the bfd.h file generated from the binutils
source code, but it is not necessarily present in the bfd.h file
provided in a binutil package (for example, it is not present in
the binutil RPM).

The solution to this issue is to define the PACKAGE macro before
including bfd.h. This is the solution suggested by the binutil
developer in bug 14243, and it is used by other kernel tools
which also use bfd.h (perf and bpf).

Fixes: 59953303827ec ("objtool: Disassemble code with libopcodes instead of running objdump")
Closes: https://lore.kernel.org/all/3fa261fd-3b46-4cbe-b48d-7503aabc96cb@oracle.com/
Reported-by: Nathan Chancellor 
Suggested-by: Nathan Chancellor 
Signed-off-by: Alexandre Chartre 
Signed-off-by: Peter Zijlstra (Intel) 
Link: https://sourceware.org/bugzilla/show_bug.cgi?id=14243
Link: https://patch.msgid.link/20251126134519.1760889-1-alexandre.chartre@oracle.com

objtool: Trim trailing NOPs in alternative

2025-11-24T19:40:48+00:00

When disassembling alternatives replace trailing NOPs with a single
indication of the number of bytes covered with NOPs.

Signed-off-by: Alexandre Chartre 
Signed-off-by: Peter Zijlstra (Intel) 
Acked-by: Josh Poimboeuf 
Link: https://patch.msgid.link/20251121095340.464045-31-alexandre.chartre@oracle.com

objtool: Add wide output for disassembly

2025-11-24T19:40:48+00:00

Add the --wide option to provide a wide output when disassembling.
With this option, the disassembly of alternatives is displayed
side-by-side instead of one above the other.

Signed-off-by: Alexandre Chartre 
Signed-off-by: Peter Zijlstra (Intel) 
Acked-by: Josh Poimboeuf 
Link: https://patch.msgid.link/20251121095340.464045-30-alexandre.chartre@oracle.com

objtool: Compact output for alternatives with one instruction

2025-11-24T19:40:48+00:00

When disassembling, if an instruction has alternatives which are all
made of a single instruction then print each alternative on a single
line (instruction + description) so that the output is more compact.

Signed-off-by: Alexandre Chartre 
Signed-off-by: Peter Zijlstra (Intel) 
Acked-by: Josh Poimboeuf 
Link: https://patch.msgid.link/20251121095340.464045-29-alexandre.chartre@oracle.com

objtool: Improve naming of group alternatives

2025-11-24T19:40:48+00:00

Improve the naming of group alternatives by showing the feature name and
flags used by the alternative.

Signed-off-by: Alexandre Chartre 
Signed-off-by: Peter Zijlstra (Intel) 
Acked-by: Josh Poimboeuf 
Link: https://patch.msgid.link/20251121095340.464045-28-alexandre.chartre@oracle.com

objtool: Add Function to get the name of a CPU feature

2025-11-24T19:39:47+00:00

Add a function to get the name of a CPU feature. The function is
architecture dependent and currently only implemented for x86. The
feature names are automatically generated from the cpufeatures.h
include file.

Signed-off-by: Alexandre Chartre 
Signed-off-by: Peter Zijlstra (Intel) 
Acked-by: Josh Poimboeuf 
Link: https://patch.msgid.link/20251121095340.464045-27-alexandre.chartre@oracle.com

objtool: Provide access to feature and flags of group alternatives

2025-11-21T14:30:14+00:00

Each alternative of a group alternative depends on a specific
feature and flags. Provide access to the feature/flags for each
alternative as an attribute (feature) in struct alt_group.

Signed-off-by: Alexandre Chartre 
Signed-off-by: Peter Zijlstra (Intel) 
Acked-by: Josh Poimboeuf 
Link: https://patch.msgid.link/20251121095340.464045-26-alexandre.chartre@oracle.com