Re: [PATCH v2] Bluetooth: btintel_pcie: fix PM flow for S0ix, S3 and S4

[Sergey Lebedev <lsa.uz@pm.me>]

Ravindra, Vladimir,

Ravindra — thank you, and thank you for the freeze/thaw answer. That was
the one thing I could not test here, this machine having no S4, and it
settles it: if D3_COLD preserves firmware then .thaw taking the normal D0
path is right, and the asymmetry with .restore is deliberate rather than an
oversight. Nothing to change on that account.

Ravindra has agreed to the two patches going as one series, with each of
you the author of your own half, the Fixes: tags kept separate, and his
rebased on top of Vladimir's. So unless Vladimir would rather not, this is
what I will send in a few days:

  1/2  Vladimir V. Kondratyev — fix stale cache in set_dxstate fallback
       check. v4 unchanged, with Paul's Reviewed-by restored (see below)
       and my Tested-by.

  2/2  Ravindra — fix PM flow for S0ix, S3 and S4. v2 rebased onto 1/2,
       unchanged in substance.

with a cover letter carrying the per-hunk measurements from this hardware:
2/2's hunk alone leaves the missed-interrupt case failing exactly as
unpatched, 1/2's re-read fixes it, and together they do not interfere. My
Signed-off-by goes on neither patch, only on the posting; the authorship,
the Fixes: tags and the review tags are yours and stay yours.

Vladimir, if you would rather v4 stayed a standalone patch, say so and I
will leave it exactly where it is — it is your patch and your call, and a
word is enough.

Two things that argue for doing it, both checked rather than assumed.

First, v4 lost a tag. Paul reviewed v3 at 14:36 UTC on 3 September:

  Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
  https://lore.kernel.org/all/ff049c69-1575-4a10-91e2-cd0371c98c0d@molgen.mpg.de/

v4 went out at 19:22 the same day and does not carry it. Whoever applies
v4 as it stands will apply it without Paul's review. That wants fixing
whether or not the patches are combined.

Second, neither patch is moving. Both are in patchwork at state "new" and
undelegated — v4 since 3 September, Ravindra's v2 since the 2nd. Two
independent changes to the same function, arriving separately, are exactly
what a maintainer defers until it is clear how they interact. One series
with an ordering and the measurements attached answers that before it is
asked, and neither of you loses anything by it.

Sergey