Re: [PATCH net-next v2 4/9] dpll: convert dpll_lock to rwsem
From: Ivan Vecera
Date: Mon Nov 16 2026 - 03:22:30 EST
Sashiko reviewed v2 only on its dashboard [1], with low severity
findings, so I am answering them here.
[1] https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261111164021.273409-1-ivecera%40redhat.com
> This isn't a bug, but zl3073x_dpll_input_pin_ref_sync_set() in
> drivers/dpll/zl3073x/dpll.c still uses the old global serialization to
> justify dropping zldpll->lock:
[...]
> Could this comment be updated in this patch, along with the dpll_lock
> reference in zl3073x_dpll_changes_check()? The later patch "dpll:
> zl3073x: update locking comments" rewrites both. The stale comment only
> exists between this patch and that one.
Right, both comments are stale between this patch and patch 7, which
rewrites them. The code is correct at every commit, because all channels
of a chip share one domain. The new comments hold from patch 3 on, so if
a respin is needed, I will move patch 7 before this one.
> Pin requests now run with dpll_lock held for read and pin->domain->lock.
> Take a fwnode pin registered on a DPLL from a different domain. Can that
> DPLL's device callbacks now run without its own domain lock?
[...]
> Would it be better to put that check in before this conversion, or in
> the same patch, so this gap never exists?
Right, between this patch and patch 6 nothing rejects such a
registration. As you note, no in-tree driver registers a pin of another
domain with .measured_freq_get, so no commit of the series can hit it
with in-tree drivers. If a respin is needed, I will move the check of
patch 6 before this patch and keep its tests in the KUnit patch.
Thanks,
Ivan