Re: [ndctl PATCH v3 2/2] daxctl, util/sysfs: skip module probe-insert when driver is builtin or live

On Thu, 18 Jun 2026 17:06:53 +0800
Chen Pei <[email protected]> wrote:

> kmod_module_probe_insert_module() is supposed to return 0 for builtin
> modules, but only when libkmod can locate the modules.builtin index. If
> the index is missing (e.g. a kernel built with the driver as builtin
> but installed without running modules_install), libkmod falls through
> to the real init_module() syscall and returns an error such as -ENOENT,
> producing a spurious "insert failure" even though the driver is already
> part of the running kernel.
> 
> Add a helper util_kmod_skip_probe_insert() that returns true when the
> module state is KMOD_MODULE_BUILTIN or KMOD_MODULE_LIVE. As an
> additional heuristic, treat KMOD_MODULE_COMING as builtin when
> /sys/module/<name>/ exists but the initstate file does not - this is
> the exact pattern libkmod's sysfs fallback emits for builtin drivers
> when the modules.builtin index is unavailable. The pattern mirrors the
> KMOD_MODULE_LIVE / KMOD_MODULE_BUILTIN check already used by ndctl's
> own test/core.c (see test/core.c:218-236).
> 
> The helper also returns the observed libkmod state via an out parameter
> so daxctl_insert_kmod_for_mode() can distinguish LIVE (retain the kmod
> reference in dev->module) from BUILTIN (drop it, since builtin drivers
> cannot be unloaded) without re-reading /sys/module/<name>/initstate.
> __util_bind() passes NULL since it does not need the state.
> 
> Reported-by: Jonathan Cameron <[email protected]>
> Suggested-by: Alison Schofield <[email protected]>
> Reviewed-by: Dave Jiang <[email protected]>
> Signed-off-by: Chen Pei <[email protected]>

I'm not set up to test this right now but fix looks good to me.
So a tentative
Reviewed-by: Jonathan Cameron <[email protected]>