From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: Jake Steinman <j@metarealtyinc.ca>
Cc: jouni.hogander@intel.com, intel-xe@lists.freedesktop.org,
intel-gfx@lists.freedesktop.org, ankit.k.nautiyal@intel.com,
rodrigo.vivi@intel.com, animesh.manna@intel.com,
uma.shankar@intel.com, jani.nikula@linux.intel.com
Subject: Re: [RFC PATCH] drm/i915/psr: Send the AS SDP in PR Active while vblank is enabled
Date: Fri, 20 Nov 2026 19:23:37 +0200 [thread overview]
Message-ID: <awCCGeCxHToY8tsI@intel.com> (raw)
In-Reply-To: <20261120163416.66175-1-j@metarealtyinc.ca>
On Fri, Nov 20, 2026 at 11:34:16AM -0500, Jake Steinman wrote:
> Hi Ville,
>
> On Fri, Nov 20, 2026 at 05:03:08PM +0200, Ville Syrjälä wrote:
> > Is this some weird hardware quirk of ours? I don't see what vblank
> > interrupts (purely an internal thing to the display engine) have to
> > do with the external panel receiving SDPs.
>
> The panel doesn't care about vblank interrupts. What it needs is the AS
> SDP while frames are being updated. With the AS SDP suspended in PR
> Active it judders, and with it sent all the time it's smooth but
> DC5/DC6 is blocked.
>
> Vblank enable is only used as the "frames are being updated" signal.
> Userspace holds a vblank reference while it flips, and Panel Replay
> already holds DC_OFF for that window, so sending the AS SDP then costs
> no DC residency. It is suspended again before DC_OFF is dropped.
Vblanks being on during flips all completely accidental. Userspace
asking for the vblank counter ahead of a flip is not mandatory.
The temporary vblank reference that query grabs also gets dropped
immediately, and if the flip isn't actually committed during the
same frame as the query the vblank interrupt will also have been
turned off by the time the flip hits the hardware.
With the DSB/flipq paths we don't hold any in-driver reference
to the vblank, and the only reason the mmio path does so is
because we're using the vblank interrupt for the completion. If
we had a better interrupt for the mmio path we wouldn't need the
vblank reference there either.
We should also be looking for a way to have userspace not do the
vblank query, or make the vblank query correct without actually
enabling vblank interrupts. Turning on vblank interrupts requires
us to turn the hardware on possibly earlier than necessary, which
will waste power. Without the vblank interrupt getting enabled
we could let the hw sleep until it actually needs to wake up for
the actual flip. Though I don't really have an answer how to get
an accurate vblank count for the query without turning some
hardware on at least.
>
> Jouni can say whether the internal confirmation explains why the panel
> needs it. From my side I only know it's this sink (LG, OUI 00:22:b9)
> on two Dell models.
>
> Thanks,
> Jake
--
Ville Syrjälä
Intel
next prev parent reply other threads:[~2026-11-20 17:23 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-23 4:13 [PATCH] drm/i915/dp: Don't promise AS SDPs in PR Active when they are disabled Jake Steinman
2026-09-23 14:27 ` sashiko-bot
2026-09-23 15:25 ` Jake Steinman
2026-09-23 15:10 ` ✗ LGCI.VerificationFailed: failure for " Patchwork
2026-09-28 12:23 ` [PATCH] " Jake Steinman
2026-10-06 22:05 ` [RFC PATCH] drm/i915/psr: Send the AS SDP in PR Active while vblank is enabled Jake Steinman
2026-10-07 14:20 ` sashiko-bot
2026-11-20 9:42 ` Hogander, Jouni
2026-11-20 15:03 ` Ville Syrjälä
[not found] ` <20261120163416.66175-1-j@metarealtyinc.ca>
2026-11-20 17:23 ` Ville Syrjälä [this message]
2026-10-07 6:41 ` [PATCH] drm/i915/dp: Don't promise AS SDPs in PR Active when they are disabled Hogander, Jouni
2026-10-07 11:49 ` Jake Steinman
2026-10-07 15:21 ` ✗ LGCI.VerificationFailed: failure for drm/i915/dp: Don't promise AS SDPs in PR Active when they are disabled (rev2) Patchwork
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=awCCGeCxHToY8tsI@intel.com \
--to=ville.syrjala@linux.intel.com \
--cc=animesh.manna@intel.com \
--cc=ankit.k.nautiyal@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
--cc=intel-xe@lists.freedesktop.org \
--cc=j@metarealtyinc.ca \
--cc=jani.nikula@linux.intel.com \
--cc=jouni.hogander@intel.com \
--cc=rodrigo.vivi@intel.com \
--cc=uma.shankar@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox