Re: [PATCH v8 02/12] syscalls: Create unified partial table for all archs

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:
- [Medium] syscall tables: validation breaks for architectures with syscall numbers > 472
- [Medium] syscall tables: build fails under set -e if grep matches no ABIs
- [Medium] syscall tables: default execution without --abis breaks on duplicate clone3
- [Low] syscall tables: temporary file leak on script failure
--

commit 1a47a0eaef6bf764943feb1f3af3a9d960c27697
Author: André Almeida <andrealmeid@igalia.com>

syscalls: Create unified partial table for all archs

Creates a new unified syscall table and modifies generation scripts to use it.
Introduces special ABIs for clone3 variants.

> diff --git a/scripts/syscall_common.tbl b/scripts/syscall_common.tbl
> new file mode 100644
> index 0000000000000..8085030eb869e
> --- /dev/null
> +++ b/scripts/syscall_common.tbl
[ ... ]
> +434	common	pidfd_open			sys_pidfd_open
> +435	clone3	clone3				sys_clone3
> +435	__clone3	clone3			__sys_clone3

[Severity: Medium]
Will default execution without an explicit ABI filter break on these duplicate
clone3 entries?

The script permits omitting the --abis flag, which falls back to matching
all lines. Processing both of these mutually exclusive entries causes
syscalltbl.sh to erroneously fail its duplicate number validation.

[ ... ]
> diff --git a/scripts/syscallhdr.sh b/scripts/syscallhdr.sh
> index bad1ab606a0fb..8f7dfcf19a4c5 100755
> --- a/scripts/syscallhdr.sh
> +++ b/scripts/syscallhdr.sh
[ ... ]
> +max=0
> +# gen_hdr(infile)
> +gen_hdr() {
> +	input=$1
> +	tmpfile=$(mktemp)

[Severity: Low]
Does this temporary file leak on script failure? File creation occurs without
a guaranteed cleanup mechanism on error paths.

>  
> -	max=0
> -	while read nr abi name native compat ; do
> +	grep -E "^[0-9A-Fa-fXx]+[[:space:]]+$abis" "$input" > $tmpfile

[Severity: Medium]
Since this script uses set -e, will executing grep as a standalone command
uncontrollably abort the script if no matching lines are found?

> +	while read nr abi name native compat; do
>  
>  		max=$nr

[Severity: Medium]
Does this unconditionally overwrite max, discarding higher syscall numbers seen
in previously processed architecture-specific tables?

[ ... ]
> diff --git a/scripts/syscalltbl.sh b/scripts/syscalltbl.sh
> index c4b1f85c2dd63..70da0039ab22f 100755
> --- a/scripts/syscalltbl.sh
> +++ b/scripts/syscalltbl.sh
[ ... ]
>  nxt=0
>  
> -grep -E "^[0-9]+[[:space:]]+$abis" "$infile" | {
> +# gen_tbl(infile)
> +gen_tbl() {
> +	input=$1
> +	tmpfile=$(mktemp)

[Severity: Low]
Does this temporary file leak if the script aborts early? There is no
cleanup mechanism on error paths.

>  
> +	grep -E "^[0-9A-Fa-fXx]+[[:space:]]+$abis" "$input" > $tmpfile

[Severity: Medium]
If this grep finds no matches, will it exit with 1 and instantly kill the
build script under set -e?

>  	while read nr abi name native compat noreturn; do
>  
>  		if [ $nxt -gt $nr ]; then
> -			echo "error: $infile: syscall table is not sorted or duplicates the same syscall number" >&2
> +			echo "error: $input: syscall table is not sorted or duplicates the same syscall number" >&2
>  			exit 1
>  		fi

[Severity: Medium]
Does the global nxt state bleed across sequentially processed files?

If an architecture that has syscall numbers greater than the highest number in
the common table uses this script with --common-tbl, sequential processing
starts at a lower number than the end of the previous arch table, causing
this sorted order validation to fail.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260918-tonyk-syscall_table-v8-0-0e2d60cad13e@igalia.com?part=2