From: Gustavo Sousa <gustavo.sousa@intel.com>
To: intel-gfx@lists.freedesktop.org, intel-xe@lists.freedesktop.org
Cc: Luca Coelho <luciano.coelho@intel.com>,
Jani Nikula <jani.nikula@intel.com>
Subject: [PATCH v3 17/18] drm/i915/dmc_wl: Do nothing until initialized
Date: Thu, 7 Nov 2024 15:27:22 -0300 [thread overview]
Message-ID: <20241107182921.102193-18-gustavo.sousa@intel.com> (raw)
In-Reply-To: <20241107182921.102193-1-gustavo.sousa@intel.com>
There is a bit of a chicken and egg situation where we depend on runtime
info to know that DMC and wakelock are supported by the hardware, and
such information is grabbed via display MMIO functions, which in turns
call intel_dmc_wl_get() and intel_dmc_wl_put() as part of their regular
flow.
Since we do not expect DC states (and consequently the wakelock
mechanism) to be enabled until DMC and DMC wakelock software structures
are initialized, a simple and safe solution to this is to turn
intel_dmc_wl_get() and intel_dmc_wl_put() into no-op until we have
properly initialized.
Let's implement that via a new field "initialized". Not that, since we
expect __intel_dmc_wl_supported() to be used for most non-static DMC
wakelock functions, let's add a drm_WARN_ONCE() there for when it is
called prior to initialization.
The only exception of functions that can be called before initialization
are intel_dmc_wl_get() and intel_dmc_wl_put(), so we bail before
calling __intel_dmc_wl_supported() if not initialized.
An alternative solution would be to revise MMIO-related stuff in the
whole driver initialization sequence, but that would possibly come with
the cost of some added ordering dependencies and complexity to the
source code.
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
---
drivers/gpu/drm/i915/display/intel_dmc_wl.c | 13 +++++++++++++
drivers/gpu/drm/i915/display/intel_dmc_wl.h | 11 +++++++++++
2 files changed, 24 insertions(+)
diff --git a/drivers/gpu/drm/i915/display/intel_dmc_wl.c b/drivers/gpu/drm/i915/display/intel_dmc_wl.c
index c164ac6e1ada..aae5ea0c72ff 100644
--- a/drivers/gpu/drm/i915/display/intel_dmc_wl.c
+++ b/drivers/gpu/drm/i915/display/intel_dmc_wl.c
@@ -264,6 +264,11 @@ static bool intel_dmc_wl_check_range(i915_reg_t reg, u32 dc_state)
static bool __intel_dmc_wl_supported(struct intel_display *display)
{
+ struct intel_dmc_wl *wl = &display->wl;
+
+ if (drm_WARN_ON(display->drm, !wl->initialized))
+ return false;
+
return display->params.enable_dmc_wl && intel_dmc_has_payload(display);
}
@@ -282,6 +287,8 @@ void intel_dmc_wl_init(struct intel_display *display)
intel_dmc_wl_sanitize_param(display);
+ wl->initialized = true;
+
if (!display->params.enable_dmc_wl)
return;
@@ -378,6 +385,9 @@ void intel_dmc_wl_get(struct intel_display *display, i915_reg_t reg)
struct intel_dmc_wl *wl = &display->wl;
unsigned long flags;
+ if (!wl->initialized)
+ return;
+
if (!__intel_dmc_wl_supported(display))
return;
@@ -410,6 +420,9 @@ void intel_dmc_wl_put(struct intel_display *display, i915_reg_t reg)
struct intel_dmc_wl *wl = &display->wl;
unsigned long flags;
+ if (!wl->initialized)
+ return;
+
if (!__intel_dmc_wl_supported(display))
return;
diff --git a/drivers/gpu/drm/i915/display/intel_dmc_wl.h b/drivers/gpu/drm/i915/display/intel_dmc_wl.h
index 147eeb4d8432..06c8b61d7e87 100644
--- a/drivers/gpu/drm/i915/display/intel_dmc_wl.h
+++ b/drivers/gpu/drm/i915/display/intel_dmc_wl.h
@@ -15,6 +15,17 @@
struct intel_display;
struct intel_dmc_wl {
+ /*
+ * There is a bit of a chicken and egg situation where we depend
+ * on runtime info to know that DMC and wakelock are supported
+ * by the hardware, and such information is grabbed via display
+ * MMIO functions, which in turns call intel_dmc_wl_get() and
+ * intel_dmc_wl_put() as part of their regular flow.
+ *
+ * So we need the initialized field to ensure that we turn the
+ * get/put routines into a no-op until we have initialized.
+ */
+ bool initialized;
spinlock_t lock; /* protects enabled, taken, dc_state and refcount */
bool enabled;
bool taken;
--
2.47.0
next prev parent reply other threads:[~2024-11-07 18:30 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-11-07 18:27 [PATCH v3 00/18] drm/i915/dmc_wl: Fixes and enablement for Xe3_LPD Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 01/18] drm/i915/dmc_wl: Use i915_mmio_reg_offset() instead of reg.reg Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 02/18] drm/xe: Mimic i915 behavior for non-sleeping MMIO wait Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 03/18] drm/i915/dmc_wl: Use non-sleeping variant of " Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 04/18] drm/i915/dmc_wl: Check for non-zero refcount in release work Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 05/18] drm/i915/dmc_wl: Get wakelock when disabling dynamic DC states Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 06/18] drm/i915/dmc_wl: Use sentinel item for range tables Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 07/18] drm/i915/dmc_wl: Extract intel_dmc_wl_reg_in_range() Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 08/18] drm/i915/dmc_wl: Rename lnl_wl_range to powered_off_ranges Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 09/18] drm/i915/dmc_wl: Track registers touched by the DMC Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 10/18] drm/i915/dmc_wl: Allow simpler syntax for single reg in range tables Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 11/18] drm/i915/dmc_wl: Deal with existing references when disabling Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 12/18] drm/i915/dmc_wl: Couple enable/disable with dynamic DC states Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 13/18] drm/i915/dmc_wl: Add and use HAS_DMC_WAKELOCK() Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 14/18] drm/i915/dmc_wl: Init only after we have runtime device info Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 15/18] drm/i915/dmc_wl: Use HAS_DMC() in HAS_DMC_WAKELOCK() Gustavo Sousa
2024-11-07 18:27 ` [PATCH v3 16/18] drm/i915/dmc_wl: Sanitize enable_dmc_wl according to hardware support Gustavo Sousa
2024-11-07 18:27 ` Gustavo Sousa [this message]
2024-11-07 19:23 ` [PATCH v3 17/18] drm/i915/dmc_wl: Do nothing until initialized Luca Coelho
2024-11-07 20:14 ` Gustavo Sousa
2024-11-07 20:22 ` Gustavo Sousa
2024-11-08 9:57 ` Luca Coelho
2024-11-08 13:10 ` Gustavo Sousa
2024-11-07 20:47 ` Gustavo Sousa
2024-11-08 10:00 ` Luca Coelho
2024-11-07 18:27 ` [PATCH v3 18/18] drm/i915/xe3lpd: Use DMC wakelock by default Gustavo Sousa
2024-11-07 19:04 ` ✗ Fi.CI.CHECKPATCH: warning for drm/i915/dmc_wl: Fixes and enablement for Xe3_LPD (rev3) Patchwork
2024-11-07 19:04 ` ✗ Fi.CI.SPARSE: " Patchwork
2024-11-07 19:29 ` ✓ Fi.CI.BAT: success " Patchwork
2024-11-07 21:32 ` ✗ Fi.CI.IGT: failure " 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=20241107182921.102193-18-gustavo.sousa@intel.com \
--to=gustavo.sousa@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
--cc=intel-xe@lists.freedesktop.org \
--cc=jani.nikula@intel.com \
--cc=luciano.coelho@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