Re: [PATCH dwarves v2 0/9] pahole: Encode true signatures in kernel BTF

[Yonghong Song <yonghong.song@linux.dev>]

On 3/9/26 11:39 AM, Alan Maguire wrote:
> On 09/03/2026 15:32, Yonghong Song wrote:
>> Current vmlinux BTF encoding is based on the source level signatures.
>> But the compiler may do some optimization and changed the signature.
>> If the user tried with source level signature, their initial implementation
>> may have wrong results and then the user need to check what is the
>> problem and work around it, e.g. through kprobe since kprobe does not
>> need vmlinux BTF.
>>
>> Majority of changed signatures are due to dead argument elimination.
>> The following is a more complex one. The original source signature:
>>    typedef struct {
>>          union {
>>                  void            *kernel;
>>                  void __user     *user;
>>          };
>>          bool            is_kernel : 1;
>>    } sockptr_t;
>>    typedef sockptr_t bpfptr_t;
>>    static int map_create(union bpf_attr *attr, bpfptr_t uattr) { ... }
>> After compiler optimization, the signature becomes:
>>    static int map_create(union bpf_attr *attr, bool uattr__is_kernel) { ... }
>> In the above, uattr__is_kernel corresponds to 'is_kernel' field in sockptr_t.
>> This makes it easier for developers to understand what changed.
>>
>> The new signature needs to properly follow ABI specification based on
>> locations. Otherwise, that signature should be discarded. For example,
>>
>>      0x0242f1f7:   DW_TAG_subprogram
>>                      DW_AT_name      ("memblock_find_in_range")
>>                      DW_AT_calling_convention        (DW_CC_nocall)
>>                      DW_AT_type      (0x0242decc "phys_addr_t")
>>                      ...
>>      0x0242f22e:     DW_TAG_formal_parameter
>>                        DW_AT_location        (indexed (0x14a) loclist = 0x005595bc:
>>                           [0xffffffff87a000f9, 0xffffffff87a00178): DW_OP_reg5 RDI
>>                           [0xffffffff87a00178, 0xffffffff87a001be): DW_OP_reg14 R14
>>                           [0xffffffff87a001be, 0xffffffff87a001c7): DW_OP_entry_value(DW_OP_reg5 RDI), DW_OP_stack_value
>>                           [0xffffffff87a001c7, 0xffffffff87a00214): DW_OP_reg14 R14)
>>                        DW_AT_name    ("start")
>>                        DW_AT_type    (0x0242decc "phys_addr_t")
>>                        ...
>>      0x0242f239:     DW_TAG_formal_parameter
>>                        DW_AT_location        (indexed (0x14b) loclist = 0x005595e6:
>>                           [0xffffffff87a000f9, 0xffffffff87a00175): DW_OP_reg4 RSI
>>                           [0xffffffff87a00175, 0xffffffff87a001b8): DW_OP_reg3 RBX
>>                           [0xffffffff87a001b8, 0xffffffff87a001c7): DW_OP_entry_value(DW_OP_reg4 RSI), DW_OP_stack_value
>>                           [0xffffffff87a001c7, 0xffffffff87a00214): DW_OP_reg3 RBX)
>>                        DW_AT_name    ("end")
>>                        DW_AT_type    (0x0242decc "phys_addr_t")
>>                        ...
>>      0x0242f245:     DW_TAG_formal_parameter
>>                        DW_AT_location        (indexed (0x14c) loclist = 0x00559610:
>>                           [0xffffffff87a001e3, 0xffffffff87a001ef): DW_OP_breg4 RSI+0)
>>                        DW_AT_name    ("size")
>>                        DW_AT_type    (0x0242decc "phys_addr_t")
>>                        ...
>>      0x0242f250:     DW_TAG_formal_parameter
>>                        DW_AT_const_value     (4096)
>>                        DW_AT_name    ("align")
>>                        DW_AT_type    (0x0242decc "phys_addr_t")
>>                        ...
>>
>> The third argument should correspond to RDX for x86_64. But the location suggests that
>> the parameter value is stored in the address with 'RSI + 0'. It is not clear whether
>> the parameter value is stored in RDX or not. So we have to discard this funciton in
>> vmlinux BTF to avoid incorrect true signatures.
>>
>> For llvm, any function having
>>    DW_AT_calling_convention        (DW_CC_nocall)
>> in dwarf DW_TAG_subprogram will indicate that this function has signature changed.
>> I did experiment with latest bpf-next. For x86_64, there are 69103 kernel functions
>> and 875 kernel functions having signature changed. A series of patches are intended
>> to ensure true signatures are properly represented. Eventually, only 17 functions
>> cannot have true signatures due to locations.
>>
> hi Yonghong, one high-level question before I start digging into this further.
> Are there any minimum requirements on LLVM/clang version for this support? Thanks!

This featureDW_AT_calling_convention is introduced into llvm on June 2022: https://reviews.llvm.org/D127134
The release is llvm15. From Documentation/process/changes.rst, we have
     Clang/LLVM (optional)  15.0.0           clang --version
So we should be okay.

But if the kernel is built with -O3 or FullLTO, there will be some
additional signature changed functions and they will be checked
only available at >= llvm23 (see https://github.com/llvm/llvm-project/pull/178973).
But the number of those additional signature changed functions should not be that many.
Also typical kernel build is -O2 so we are not missing signature changed
functions in most cases.

>
> Alan
>   
>> For arm64, there are 863 kernel functions having signature changed, and
>> 79 functions cannot have true signatures due to locations. I checked those
>> functions and look like llvm arm64 backend more relaxed to compute parameter
>> values.
>>
>> For the patch set, Patch 1 introduced usage of DW_AT_calling_convention, which
>> can precisely identify which function has signature changed. This can filter
>> majority of functions where their signature won't change.
>> Patches 2 to 7 tried to find functions with true signature.
>> Patch 8 enables to btf encoder to properly generate BTF.
>> Patch 9 includes a few tests.
>>
>> Changelog:
>>    v1 -> v2:
>>      - v1: https://lore.kernel.org/bpf/20260305225455.1151066-1-yonghong.song@linux.dev/
>>      - Added producer_clang guarding in btf_encoder. Otherwise, gcc kernel build
>>        will crash pahole.
>>      - Fix an early return in parameter__reg() which didn't do pthread_mutex_unlock()
>>        which caused the deadlock for arm64.
>>      - Add a few more places to guard with producer_clang and conf->true_signature
>>        to maintain the previous behavior if not clang or conf->true_signature is false.
>>
>> Yonghong Song (9):
>>    dwarf_loader: Reduce parameter checking with clang
>>      DW_AT_calling_convention attr
>>    dwarf_loader: Handle signatures with dead arguments
>>    dwarf_loader: Refactor initial ret -1 to be macro PARM_DEFAULT_FAIL
>>    dwarf_laoder: Handle locations with DW_OP_fbreg
>>    dwarf_loader: Change exprlen checking condition in parameter__reg()
>>    dwarf_loader: Detect optimized parameters with locations having
>>      constant values
>>    dwarf_loader: Handle expression lists
>>    btf_encoder: Handle optimized parameter properly
>>    tests: Add a few clang true signature tests
>>
>>   btf_encoder.c                                 |  13 +-
>>   dwarf_loader.c                                | 397 +++++++++++++++++-
>>   dwarves.h                                     |   3 +
>>   tests/true_signatures/clang_parm_aggregate.sh |  83 ++++
>>   tests/true_signatures/clang_parm_optimized.sh |  95 +++++
>>   .../clang_parm_optimized_stack.sh             |  95 +++++
>>   .../gcc_true_signatures.sh                    |   0
>>   7 files changed, 662 insertions(+), 24 deletions(-)
>>   create mode 100755 tests/true_signatures/clang_parm_aggregate.sh
>>   create mode 100755 tests/true_signatures/clang_parm_optimized.sh
>>   create mode 100755 tests/true_signatures/clang_parm_optimized_stack.sh
>>   rename tests/{ => true_signatures}/gcc_true_signatures.sh (100%)
>>