[RFC] sanity check on hand-built klp_patch for version_proc_show (measurement harness ground truth)

Hi livepatch maintainers,

I am building a differential-testing harness for livepatch build
pipelines (kpatch-build vs upstream klp-build, etc.) and need a
hand-built klp_patch module as behavioral ground truth per pinned
kernel commit.

Before I lean on this in a paper, I would appreciate an informal "does
this look sane?" pass on the LP-PILOT-02 module below. This is not a
merge candidate, but just a minimal multi-relocation replacement of
`version_proc_show` on Linux v6.6.47 (`4c1a2d4cd9a5`).

Target: `version_proc_show` in `fs/proc/version.c`
API: standard `klp_func` / `klp_object` / `klp_patch` (same shape as
`samples/livepatch/livepatch-sample.c`)
Relocations in the built .ko: `R_X86_64_PLT32` to `seq_printf`,
`seq_putc`; two `R_X86_64_32S` addends into `.rodata` (marker + suffix
strings)
Validation: `insmod` succeeds; `/proc/version` shows marker
`LP-PILOT-02 patched-by-harness`; revert via sysfs `enabled=3D0`
restores banner; controlled rodata-addend swap leaves module loadable
but breaks the marker predicate (behavioral detection, not load-time
rejection)

Questions I would value eyes on:
1. Is the `klp_patch` structure correct for v6.6.x (object/func
sentinels, `MODULE_INFO(livepatch, "Y")`)?
2. Any obvious foot-guns replacing a seq_file handler that only uses
`seq_printf`/`seq_putc`?
3. Anything you'd consider invalid ground truth for equivalence
testing even if it loads?

Source (LP-PILOT-02 hand-build):

--- livepatch-version.c ---
// LP-PILOT-02 hand-built klp_patch =E2=80=94 version_proc_show
// Multi-external-reloc replacement (strlen + seq_printf + seq_putc;
dual rodata addends).
#define pr_fmt(fmt) "lp-pilot-version: " fmt

#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/livepatch.h>
#include <linux/module.h>
#include <linux/seq_file.h>
#include <linux/string.h>

static int hb_version_proc_show(struct seq_file *m, void *v)
{
static const char marker[] =3D "LP-PILOT-02";
static const char suffix[] =3D " patched-by-harness";
size_t n;

n =3D strlen(marker);
if (n =3D=3D 0)
return 0;
n +=3D strlen(suffix);
seq_printf(m, "%s%s\n", marker, suffix);
if (n > 0)
seq_putc(m, '!');
return 0;
}

static struct klp_func hb_funcs[] =3D {
{
.old_name =3D "version_proc_show",
.new_func =3D hb_version_proc_show,
},
{ }
};

static struct klp_object hb_objs[] =3D {
{
.funcs =3D hb_funcs,
},
{ }
};

static struct klp_patch hb_patch =3D {
.mod =3D THIS_MODULE,
.objs =3D hb_objs,
};

static int __init hb_version_init(void)
{
return klp_enable_patch(&hb_patch);
}

static void __exit hb_version_exit(void)
{
}

module_init(hb_version_init);
module_exit(hb_version_exit);
MODULE_LICENSE("GPL");
MODULE_INFO(livepatch, "Y");
--- end ---

Context: https://github.com/kazuru-chidumbwe/livepatch-differential-harness
(public release pending)

Thank you for any quick reactions, even "looks fine" or "fix X" is
helpful, or a correction if I am wandering the wrong direction.

Regards,

Seke