Re: [PATCH v3 25/27] ASoC: codecs: rt712: Use guard() for mutex locks

[Bui Duc Phuc]

Hi all,

While looking into removing the goto label in rt712_sdca_dev_resume(),
I dug a bit deeper into regcache_sync() and how it manages the
cache_dirty flag. I'd like to share the analysis and propose a small patch
for review.

1. cache_dirty state before the out: label in regcache_sync()
Before the out: block, cache_dirty can only be in one of two states:
+ true: if writing the patch or the sync itself failed partway through.
+ false: either it was already false from the start, or — after the patch
   has been applied successfully (i.e. the patch loop didn't jump to goto
   out) — the subsequent sync() call also returns success (ret == 0).
→ The logic here is sound and consistent.

2. cache_dirty state after the out: label in regcache_sync()
If _regmap_write() fails when re-pushing the paging register here,
none of its three error paths (-EIO, regcache_write() failure,
or map->reg_write() failure) touch cache_dirty.
Only the cache_only success path touches cache_dirty.
So cache_dirty simply keeps whatever value it already had before out:
(as described above), even though this write genuinely failed.

3. Proposed patch
-------------
ret = _regmap_write(map, this->selector_reg, i);
if (ret != 0) {
        dev_err(map->dev, "Failed to write %x = %x: %d\n",
                      this->selector_reg, i, ret);
+      map->cache_dirty = true;
        break;
}
--------------
This ensures cache_dirty correctly reflects a real cache/hardware
mismatch whenever this write fails.

Feedback welcome before I submit the formal patch.

Best regards,
Phuc