All of lore.kernel.org
 help / color / mirror / Atom feed
From: Wayne Lin <Wayne.Lin@amd.com>
To: <amd-gfx@lists.freedesktop.org>
Cc: <Harry.Wentland@amd.com>, <Sunpeng.Li@amd.com>,
	<Rodrigo.Siqueira@amd.com>, <Aurabindo.Pillai@amd.com>,
	<roman.li@amd.com>, <wayne.lin@amd.com>,
	<agustin.gutierrez@amd.com>, <chiahsuan.chung@amd.com>,
	<hersenxs.wu@amd.com>, <jerry.zuo@amd.com>,
	Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>,
	Duncan Ma <duncan.ma@amd.com>
Subject: [PATCH 35/43] drm/amd/display: Detect and disallow idle reallow during reentrancy
Date: Tue, 12 Mar 2024 17:20:28 +0800	[thread overview]
Message-ID: <20240312092036.3283319-36-Wayne.Lin@amd.com> (raw)
In-Reply-To: <20240312092036.3283319-1-Wayne.Lin@amd.com>

From: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>

[Why]
Cursor updates can be preempted by queued flips in some DMs.

The synchronization model causes this to occur within the same thread
at an intermediate level when we insert logs into the OS queue.

Since this occurs on the same thread and we're still holding the lock
(recursively) the cache is coherent.

The exit sequence will run twice since we technically haven't finished
the exit the first time, so we need a way to detect and avoid the
reallow in the middle of this call to prevent the hang on the cursor
update that was preempted.

[How]
Keep a counter that tracks the depth of the exit calls. Do not reallow
until the counter is zero.

Reviewed-by: Duncan Ma <duncan.ma@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
---
 drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c | 14 ++++++++++++--
 drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h |  1 +
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c
index f796ed061879..4878e9e50440 100644
--- a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c
+++ b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c
@@ -1437,6 +1437,8 @@ void dc_dmub_srv_apply_idle_power_optimizations(const struct dc *dc, bool allow_
 	 */
 
 	if (!allow_idle) {
+		dc_dmub_srv->idle_exit_counter += 1;
+
 		dc_dmub_srv_exit_low_power_state(dc);
 		/*
 		 * Idle is considered fully exited only after the sequence above
@@ -1448,6 +1450,12 @@ void dc_dmub_srv_apply_idle_power_optimizations(const struct dc *dc, bool allow_
 		 * dm_execute_dmub_cmd submission instead of the "wake" helpers.
 		 */
 		dc_dmub_srv->idle_allowed = false;
+
+		dc_dmub_srv->idle_exit_counter -= 1;
+		if (dc_dmub_srv->idle_exit_counter < 0) {
+			ASSERT(0);
+			dc_dmub_srv->idle_exit_counter = 0;
+		}
 	} else {
 		/* Consider idle as notified prior to the actual submission to
 		 * prevent multiple entries. */
@@ -1489,7 +1497,8 @@ bool dc_wake_and_execute_dmub_cmd_list(const struct dc_context *ctx, unsigned in
 	else
 		result = dm_execute_dmub_cmd(ctx, cmd, wait_type);
 
-	if (result && reallow_idle && !ctx->dc->debug.disable_dmub_reallow_idle)
+	if (result && reallow_idle && dc_dmub_srv->idle_exit_counter == 0 &&
+	    !ctx->dc->debug.disable_dmub_reallow_idle)
 		dc_dmub_srv_apply_idle_power_optimizations(ctx->dc, true);
 
 	return result;
@@ -1538,7 +1547,8 @@ bool dc_wake_and_execute_gpint(const struct dc_context *ctx, enum dmub_gpint_com
 
 	result = dc_dmub_execute_gpint(ctx, command_code, param, response, wait_type);
 
-	if (result && reallow_idle && !ctx->dc->debug.disable_dmub_reallow_idle)
+	if (result && reallow_idle && dc_dmub_srv->idle_exit_counter == 0 &&
+	    !ctx->dc->debug.disable_dmub_reallow_idle)
 		dc_dmub_srv_apply_idle_power_optimizations(ctx->dc, true);
 
 	return result;
diff --git a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h
index 60c93e9e3533..c0a512a12531 100644
--- a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h
+++ b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.h
@@ -51,6 +51,7 @@ struct dc_dmub_srv {
 	struct dc_context *ctx;
 	void *dm;
 
+	int32_t idle_exit_counter;
 	bool idle_allowed;
 	bool needs_idle_wake;
 };
-- 
2.37.3


  parent reply	other threads:[~2024-03-12  9:24 UTC|newest]

Thread overview: 48+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-03-12  9:19 [PATCH 00/43] DC Patches March 18, 2024 Wayne Lin
2024-03-12  9:19 ` [PATCH 01/43] drm/amd/display: Remove code duplication Wayne Lin
2024-03-12  9:19 ` [PATCH 02/43] drm/amd/display: Remove wrong signal from vrr calculation Wayne Lin
2024-03-12  9:19 ` [PATCH 03/43] drm/amd/display: Enable 2to1 ODM policy for DCN35 Wayne Lin
2024-03-12 10:20   ` Christian König
2024-03-13  6:35     ` Lin, Wayne
2024-03-12  9:19 ` [PATCH 04/43] drm/amd/display: Delete duplicated function prototypes Wayne Lin
2024-03-12  9:19 ` [PATCH 05/43] drm/amd/display: Correct indentations and spaces Wayne Lin
2024-03-12  9:19 ` [PATCH 06/43] drm/amd/display: Add the MALL size in the fallback function Wayne Lin
2024-03-12  9:20 ` [PATCH 07/43] drm/amd/display: Move define to the proper header Wayne Lin
2024-03-12  9:20 ` [PATCH 08/43] drm/amd/display: Enable fast update for DCN314 Wayne Lin
2024-03-12  9:20 ` [PATCH 09/43] drm/amd/display: Remove legacy code Wayne Lin
2024-03-12  9:20 ` [PATCH 10/43] drm/amd/display: correct hostvm flag Wayne Lin
2024-03-12  9:20 ` [PATCH 11/43] drm/amd/display: Comments adjustments Wayne Lin
2024-03-12  9:20 ` [PATCH 12/43] drm/amd/display: Add missing registers and offset Wayne Lin
2024-03-12  9:20 ` [PATCH 13/43] drm/amd/display: Fix noise issue on HDMI AV mute Wayne Lin
2024-03-12  9:20 ` [PATCH 14/43] drm/amd/display: skip forcing odm in minimal transition Wayne Lin
2024-03-12  9:20 ` [PATCH 15/43] drm/amd/display: revert Exit idle optimizations before HDCP execution Wayne Lin
2024-03-12  9:20 ` [PATCH 16/43] drm/amd/display: Add debug option for idle reg checks Wayne Lin
2024-03-12  9:20 ` [PATCH 17/43] drm/amd/display: Revert Add left edge pixel + ODM pipe split Wayne Lin
2024-03-12  9:20 ` [PATCH 18/43] drm/amd/display: To adjust dprefclk by down spread percentage Wayne Lin
2024-03-12  9:20 ` [PATCH 19/43] drm/amd/display: Enabling urgent latency adjustment for DCN35 Wayne Lin
2024-03-12  9:20 ` [PATCH 20/43] drm/amd/display: Revert "Set the power_down_on_boot function pointer to null" Wayne Lin
2024-03-12  9:20 ` [PATCH 21/43] drm/amd/display: add stream clock source to DP DTO params Wayne Lin
2024-03-12  9:20 ` [PATCH 22/43] drm/amd/display: Program pixclk according to dcn revision Wayne Lin
2024-03-12  9:20 ` [PATCH 23/43] drm/amd/display: clear mpc_tree in init_pipes Wayne Lin
2024-03-12  9:20 ` [PATCH 24/43] drm/amd/display: [FW Promotion] Release 0.0.208.0 Wayne Lin
2024-03-12  9:20 ` [PATCH 25/43] drm/amd/display: 3.2.276 Wayne Lin
2024-03-12  9:20 ` [PATCH 26/43] drm/amd/display: Workaround register access in idle race with cursor Wayne Lin
2024-03-12  9:20 ` [PATCH 27/43] drm/amd/display: Revert Remove pixle rate limit for subvp Wayne Lin
2024-03-12  9:20 ` [PATCH 28/43] drm/amd/display: fix debug key not working on dml2 Wayne Lin
2024-03-12  9:20 ` [PATCH 29/43] drm/amd/display: Power on VPG memory unconditionally if off Wayne Lin
2024-03-12  9:20 ` [PATCH 30/43] drm/amd/display: Added debug prints for zstate_support and StutterPeriod Wayne Lin
2024-03-12  9:20 ` [PATCH 31/43] drm/amd/display: change aux_init to apu version Wayne Lin
2024-03-12  9:20 ` [PATCH 32/43] drm/amd/display: Increase Z8 watermark times Wayne Lin
2024-03-12  9:20 ` [PATCH 33/43] drm/amd/display: Prevent crash on bring-up Wayne Lin
2024-03-13 14:40   ` Pillai, Aurabindo
2024-03-12  9:20 ` [PATCH 34/43] drm/amd/display: increase bb clock for DCN351 Wayne Lin
2024-03-12  9:20 ` Wayne Lin [this message]
2024-03-12  9:20 ` [PATCH 36/43] drm/amd/display: Add optional optimization for IPS handshake Wayne Lin
2024-03-12  9:20 ` [PATCH 37/43] drm/amd/display: Enable optimized handshake for DCN35 Wayne Lin
2024-03-12  9:20 ` [PATCH 38/43] drm/amd/display: Remove unnecessary hard coded DPM states Wayne Lin
2024-03-12  9:20 ` [PATCH 39/43] drm/amd/display: Enable new interface design for alternate scrambling Wayne Lin
2024-03-12  9:20 ` [PATCH 40/43] drm/amd/display: Enable reallow for idle on DCN35 Wayne Lin
2024-03-12  9:20 ` [PATCH 41/43] drm/amd/display: fix a bug to dereference already freed old current state memory Wayne Lin
2024-03-12  9:20 ` [PATCH 42/43] drm/amd/display: Add TB_BORROWED_MAX definition Wayne Lin
2024-03-12  9:20 ` [PATCH 43/43] drm/amd/display: 3.2.277 Wayne Lin
2024-03-18 13:54 ` [PATCH 00/43] DC Patches March 18, 2024 Wheeler, Daniel

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=20240312092036.3283319-36-Wayne.Lin@amd.com \
    --to=wayne.lin@amd.com \
    --cc=Aurabindo.Pillai@amd.com \
    --cc=Harry.Wentland@amd.com \
    --cc=Rodrigo.Siqueira@amd.com \
    --cc=Sunpeng.Li@amd.com \
    --cc=agustin.gutierrez@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=chiahsuan.chung@amd.com \
    --cc=duncan.ma@amd.com \
    --cc=hersenxs.wu@amd.com \
    --cc=jerry.zuo@amd.com \
    --cc=nicholas.kazlauskas@amd.com \
    --cc=roman.li@amd.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.