linux.git/scripts/faddr2line, branch v6.18.21

scripts/faddr2line: Fix "Argument list too long" error

2026-01-02T11:56:38+00:00

[ Upstream commit ff5c0466486ba8d07ab2700380e8fd6d5344b4e9 ]

The run_readelf() function reads the entire output of readelf into a
single shell variable. For large object files with extensive debug
information, the size of this variable can exceed the system's
command-line argument length limit.

When this variable is subsequently passed to sed via `echo "${out}"`, it
triggers an "Argument list too long" error, causing the script to fail.

Fix this by redirecting the output of readelf to a temporary file
instead of a variable. The sed commands are then modified to read from
this file, avoiding the argument length limitation entirely.

Signed-off-by: Pankaj Raghav 
Signed-off-by: Josh Poimboeuf 
Signed-off-by: Sasha Levin

Revert "scripts/faddr2line: Check only two symbols when calculating symbol size"

2024-10-17T22:16:04+00:00

This reverts commit c02904f05ff805d6c0631634d5751ebd338f75ec.

Such commit assumed that only two symbols are relevant for the symbol
size calculation. However, this can lead to an incorrect symbol size
calculation when there are mapping symbols emitted by readelf.

For instance, when feeding 'update_irq_load_avg+0x1c/0x1c4', faddr2line
might need to process the following readelf lines:

 784284: ffffffc0081cca30   428 FUNC    GLOBAL DEFAULT     2 update_irq_load_avg
  87319: ffffffc0081ccb0c     0 NOTYPE  LOCAL  DEFAULT     2 $x.62522
  87321: ffffffc0081ccbdc     0 NOTYPE  LOCAL  DEFAULT     2 $x.62524
  87323: ffffffc0081ccbe0     0 NOTYPE  LOCAL  DEFAULT     2 $x.62526
  87325: ffffffc0081ccbe4     0 NOTYPE  LOCAL  DEFAULT     2 $x.62528
  87327: ffffffc0081ccbe8     0 NOTYPE  LOCAL  DEFAULT     2 $x.62530
  87329: ffffffc0081ccbec     0 NOTYPE  LOCAL  DEFAULT     2 $x.62532
  87331: ffffffc0081ccbf0     0 NOTYPE  LOCAL  DEFAULT     2 $x.62534
  87332: ffffffc0081ccbf4     0 NOTYPE  LOCAL  DEFAULT     2 $x.62535
 783403: ffffffc0081ccbf4   424 FUNC    GLOBAL DEFAULT     2 sched_pelt_multiplier

The symbol size of 'update_irq_load_avg' should be calculated with the
address of 'sched_pelt_multiplier', after skipping the mapping symbols
seen in between. However, the offending commit cuts the list short and
faddr2line incorrectly assumes 'update_irq_load_avg' is the last symbol
in the section, resulting in:

  $ scripts/faddr2line vmlinux update_irq_load_avg+0x1c/0x1c4
  skipping update_irq_load_avg address at 0xffffffc0081cca4c due to size mismatch (0x1c4 != 0x3ff9a59988)
  no match for update_irq_load_avg+0x1c/0x1c4

After reverting the commit the issue is resolved:

  $ scripts/faddr2line vmlinux update_irq_load_avg+0x1c/0x1c4
  update_irq_load_avg+0x1c/0x1c4:
  cpu_of at kernel/sched/sched.h:1109
  (inlined by) update_irq_load_avg at kernel/sched/pelt.c:481

Fixes: c02904f05ff8 ("scripts/faddr2line: Check only two symbols when calculating symbol size")
Signed-off-by: Carlos Llamas 
Acked-by: Will Deacon 
Acked-by: Brian Johannesmeyer 
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Check only two symbols when calculating symbol size

2024-07-03T06:38:37+00:00

Rather than looping through each symbol in a particular section to
calculate a symbol's size, grep for the symbol and its immediate
successor, and only use those two symbols.

Signed-off-by: Brian Johannesmeyer 
Link: https://lore.kernel.org/r/20240415145538.1938745-8-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Remove call to addr2line from find_dir_prefix()

2024-07-03T06:38:37+00:00

Use the single long-running faddr2line process from find_dir_prefix().

Signed-off-by: Brian Johannesmeyer 
Link: https://lore.kernel.org/r/20240415145538.1938745-7-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Invoke addr2line as a single long-running process

2024-07-03T06:38:36+00:00

Rather than invoking a separate addr2line process for each address, invoke
a single addr2line coprocess, and pass each address to its stdin. Previous
work [0] applied a similar change to perf, leading to a ~60x speed-up [1].

If using an object file that is _not_ vmlinux, faddr2line passes a section
name argument to addr2line. Because we do not know until runtime which
section names will be passed to addr2line, we cannot apply this change to
non-vmlinux object files. Hence, it only applies to vmlinux.

[0] commit be8ecc57f180 ("perf srcline: Use long-running addr2line per
DSO")
[1] Link:
https://eighty-twenty.org/2021/09/09/perf-addr2line-speed-improvement

Signed-off-by: Brian Johannesmeyer 
Link: https://lore.kernel.org/r/20240415145538.1938745-6-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Pass --addresses argument to addr2line

2024-07-03T06:38:36+00:00

In preparation for identifying an addr2line sentinel. See previous work
[0], which applies a similar change to perf.

[0] commit 8dc26b6f718a ("perf srcline: Make sentinel reading for binutils
addr2line more robust")

Signed-off-by: Brian Johannesmeyer 
Link: https://lore.kernel.org/r/20240415145538.1938745-5-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Check vmlinux only once

2024-07-03T06:38:36+00:00

Rather than checking whether the object file is vmlinux for each invocation
of __faddr2line, check it only once beforehand.

Signed-off-by: Brian Johannesmeyer 
Link: https://lore.kernel.org/r/20240415145538.1938745-4-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Combine three readelf calls into one

2024-07-03T06:38:36+00:00

Rather than calling readelf three separate times to collect three different
types of info, call it only once, and parse out the different types of info
from its output.

Signed-off-by: Brian Johannesmeyer 
Link: https://lore.kernel.org/r/20240415145538.1938745-3-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Reduce number of readelf calls to three

2024-07-03T06:38:36+00:00

Rather than calling readelf several times for each invocation of
__faddr2line, call readelf only three times at the beginning, and save its
result for future use.

Signed-off-by: Brian Johannesmeyer 
Link: https://lore.kernel.org/r/20240415145538.1938745-2-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf

scripts/faddr2line: Skip over mapping symbols in output from readelf

2023-10-23T15:36:46+00:00

Mapping symbols emitted in the readelf output can confuse the
'faddr2line' symbol size calculation, resulting in the erroneous
rejection of valid offsets. This is especially prevalent when building
an arm64 kernel with CONFIG_CFI_CLANG=y, where most functions are
prefixed with a 32-bit data value in a '$d.n' section. For example:

447538: ffff800080014b80   548 FUNC    GLOBAL DEFAULT    2 do_one_initcall
   104: ffff800080014c74     0 NOTYPE  LOCAL  DEFAULT    2 $x.73
   106: ffff800080014d30     0 NOTYPE  LOCAL  DEFAULT    2 $x.75
   111: ffff800080014da4     0 NOTYPE  LOCAL  DEFAULT    2 $d.78
   112: ffff800080014da8     0 NOTYPE  LOCAL  DEFAULT    2 $x.79
    36: ffff800080014de0   200 FUNC    LOCAL  DEFAULT    2 run_init_process

Adding a warning to do_one_initcall() results in:

  | WARNING: CPU: 0 PID: 1 at init/main.c:1236 do_one_initcall+0xf4/0x260

Which 'faddr2line' refuses to accept:

$ ./scripts/faddr2line vmlinux do_one_initcall+0xf4/0x260
skipping do_one_initcall address at 0xffff800080014c74 due to size mismatch (0x260 != 0x224)
no match for do_one_initcall+0xf4/0x260

Filter out these entries from readelf using a shell reimplementation of
is_mapping_symbol(), so that the size of a symbol is calculated as a
delta to the next symbol present in ksymtab.

Suggested-by: Masahiro Yamada 
Signed-off-by: Will Deacon 
Reviewed-by: Nick Desaulniers 
Link: https://lore.kernel.org/r/20231002165750.1661-4-will@kernel.org
Signed-off-by: Josh Poimboeuf