Re: [PATCH v7 03/13] scripts/syscall.tbl: Use the common table

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] perf: perf trace silently drops generic syscalls >= 403
- [Low] docs: adding-syscalls.rst directs developers to break the build
--

commit b66e7fc900570caaf1efbae8e5955a536079dad9
Author: André Almeida <andrealmeid@igalia.com>

scripts/syscall.tbl: Use the common table

Remove some of the duplicated code by using the common syscall number
table.

> diff --git a/scripts/Makefile.asm-headers b/scripts/Makefile.asm-headers
> index f1c3d287c14b0..b09dbc8992c32 100644
> --- a/scripts/Makefile.asm-headers
> +++ b/scripts/Makefile.asm-headers
[ ... ]
> @@ -15,9 +15,10 @@ all:
>  
>  src := $(srctree)/$(subst /generated,,$(obj))
>  
> -syscall_abis_32  += common,32
> -syscall_abis_64  += common,64
> +syscall_abis_32  += clone3,common,32
> +syscall_abis_64  += clone3,common,64
>  syscalltbl := $(srctree)/scripts/syscall.tbl
> +syscalltbl_common := $(srctree)/scripts/syscall_common.tbl

[Severity: Medium]
Does this change cause perf trace to silently drop generic syscalls >= 403?

The tools/perf/trace/beauty/syscalltbl.sh script is hardcoded to only
parse scripts/syscall.tbl for generic architectures like CSKY and RISC-V:

tools/perf/trace/beauty/syscalltbl.sh:build_tables() {
    build_tables "$tools_dir/scripts/syscall.tbl" "$outfile" \
        common,32,csky,time32,stat64,rlimit EM_CSKY
}

Because the newly introduced syscall_common.tbl is not parsed by perf trace,
local execution of perf trace on affected architectures will fail to resolve
the names of generic syscalls >= 403 (e.g., clone3, openat2, io_uring_setup),
displaying raw numbers instead.

> diff --git a/scripts/syscall.tbl b/scripts/syscall.tbl
> index 0ab5316051202..ee828d03094ed 100644
> --- a/scripts/syscall.tbl
> +++ b/scripts/syscall.tbl
> @@ -344,72 +344,3 @@
>  293	common	rseq				sys_rseq
>  294	common	kexec_file_load			sys_kexec_file_load
>  # 295 through 402 are unassigned to sync up with generic numbers don't use
> -403	32	clock_gettime64			sys_clock_gettime
> -404	32	clock_settime64			sys_clock_settime
[ ... ]
> -471	common	rseq_slice_yield		sys_rseq_slice_yield
> -472	common	fchroot				sys_fchroot

[Severity: Low]
With these entries moved, does Documentation/process/adding-syscalls.rst now
direct developers to break the build?

The documentation actively instructs developers to place new entries in the
now-deprecated primary table:

Documentation/process/adding-syscalls.rst:
    As scripts/syscall.tbl serves as a common syscall table across multiple
    architectures, a new entry is required in this table::

        468   common   xyzzy     sys_xyzzy

Because syscallhdr.sh and syscalltbl.sh append syscall_common.tbl (which
starts at 403) after processing syscall.tbl, a developer following these
instructions and adding a new generic syscall (e.g., 468) to syscall.tbl will
cause the internal nxt counter to exceed the first entry of the common table
(469 > 403). This triggers the strict ascending-order check and results in a
fatal "syscall table is not sorted" build error.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260917-tonyk-syscall_table-v7-0-f4cd8f8e3265@igalia.com?part=3