Received-SPF: pass (google.com: domain of linux-kernel+bounces-206179-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Precedence: bulk
MIME-Version: 1.0
References: <CABXGCsNptxsQO=5=qi-JYiFX=rX8Ok5inK80Gn0qrUFWbtBGng@mail.gmail.com>
In-Reply-To: <CABXGCsNptxsQO=5=qi-JYiFX=rX8Ok5inK80Gn0qrUFWbtBGng@mail.gmail.com>
From: Alex Deucher <alexdeucher@gmail.com>
Date: Fri, 7 Jun 2024 09:39:35 -0400
Message-ID: <CADnq5_PDxJ8O1JUQ9RBYRFB9G1WZJos05ZAM4jUKuPBwPxjNkA@mail.gmail.com>
Subject: Re: 6.10/bisected/regression - commits bc87d666c05 and 6d4279cb99ac
 cause appearing green flashing bar on top of screen on Radeon 6900XT and 120Hz
To: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
Cc: Hamza Mahfooz <hamza.mahfooz@amd.com>, Rodrigo.Siqueira@amd.com, 
	"Deucher, Alexander" <alexander.deucher@amd.com>, amd-gfx list <amd-gfx@lists.freedesktop.org>, 
	dri-devel <dri-devel@lists.freedesktop.org>, 
	Linux List Kernel Mailing <linux-kernel@vger.kernel.org>, 
	Linux regressions mailing list <regressions@lists.linux.dev>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Sun, May 26, 2024 at 10:12=E2=80=AFAM Mikhail Gavrilov
<mikhail.v.gavrilov@gmail.com> wrote:
>
> Hi,
> Day before yesterday I replaced 7900XTX to 6900XT for got clear in
> which kernel first time appeared warning message "DMA-API: amdgpu
> 0000:0f:00.0: cacheline tracking EEXIST, overlapping mappings aren't
> supported".
> The kernel 6.3 and older won't boot on a computer with Radeon 7900XTX.
> When I booted the system with 6900XT I saw a green flashing bar on top
> of the screen when I typed commands in the gnome terminal which was
> maximized on full screen.
> Demonstration: https://youtu.be/tTvwQ_5pRkk
> For reproduction you need Radeon 6900XT GPU connected to 120Hz OLED TV by=
 HDMI.
>
> I bisected the issue and the first commit which I found was 6d4279cb99ac.
> commit 6d4279cb99ac4f51d10409501d29969f687ac8dc (HEAD)
> Author: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
> Date:   Tue Mar 26 10:42:05 2024 -0600
>
>     drm/amd/display: Drop legacy code
>
>     This commit removes code that are not used by display anymore.
>
>     Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
>     Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
>     Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
>
>  drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h         |  4 ----
>  drivers/gpu/drm/amd/display/dc/inc/resource.h                  |  7 ----=
---
>  drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c         | 10 ----=
------
>  drivers/gpu/drm/amd/display/dc/resource/dcn21/dcn21_resource.c | 33
> +--------------------------------
>  4 files changed, 1 insertion(+), 53 deletions(-)
>
> Every time after bisecting I usually make sure that I found the right
> commit and build the kernel with revert of the bad commit.
> But this time I again observed an issue after running a kernel builded
> without commit 6d4279cb99ac.
> And I decided to find a second bad commit.
> The second bad commit has been bc87d666c05.
> commit bc87d666c05a13e6d4ae1ddce41fc43d2567b9a2 (HEAD)
> Author: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
> Date:   Tue Mar 26 11:55:19 2024 -0600
>
>     drm/amd/display: Add fallback configuration for set DRR in DCN10
>
>     Set OTG/OPTC parameters to 0 if something goes wrong on DCN10.
>
>     Acked-by: Hamza Mahfooz <hamza.mahfooz@amd.com>
>     Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
>     Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
>
>  drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c | 15 ++++++++++++=
---
>  1 file changed, 12 insertions(+), 3 deletions(-)
>
> After reverting both these commits on top of 54f71b0369c9 the issue is go=
ne.
>
> I also attach the build config.
>
> My hardware specs: https://linux-hardware.org/?probe=3Df25a873c5e
>
> Rodrigo or anyone else from the AMD team can you look please.

@Siqueira, Rodrigo can you take a look?  The two patches change the
programming of OTG_V_TOTAL_CONTROL.  The first patch removes this
code:

diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c
b/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c
index 58bdbd859bf9..d6f095b4555d 100644
--- a/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c
+++ b/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c
@@ -462,16 +462,6 @@ void optc2_setup_manual_trigger(struct
timing_generator *optc)
 {
        struct optc *optc1 =3D DCN10TG_FROM_TG(optc);

-       /* Set the min/max selectors unconditionally so that
-        * DMCUB fw may change OTG timings when necessary
-        * TODO: Remove the w/a after fixing the issue in DMCUB firmware
-        */
-       REG_UPDATE_4(OTG_V_TOTAL_CONTROL,
-                                OTG_V_TOTAL_MIN_SEL, 1,
-                                OTG_V_TOTAL_MAX_SEL, 1,
-                                OTG_FORCE_LOCK_ON_EVENT, 0,
-                                OTG_SET_V_TOTAL_MIN_MASK, (1 << 1));
/* TRIGA */
-
        REG_SET_8(OTG_TRIGA_CNTL, 0,
                        OTG_TRIGA_SOURCE_SELECT, 21,
                        OTG_TRIGA_SOURCE_PIPE_SELECT, optc->inst,

and the second patch adds this hunk:

diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
index f109a101d84f..5574bc628053 100644
--- a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
+++ b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
@@ -945,10 +945,19 @@ void optc1_set_drr(
                                OTG_FORCE_LOCK_ON_EVENT, 0,
                                OTG_SET_V_TOTAL_MIN_MASK_EN, 0,
                                OTG_SET_V_TOTAL_MIN_MASK, 0);
-       }

-       // Setup manual flow control for EOF via TRIG_A
-       optc->funcs->setup_manual_trigger(optc);
+               // Setup manual flow control for EOF via TRIG_A
+               optc->funcs->setup_manual_trigger(optc);
+
+       } else {
+               REG_UPDATE_4(OTG_V_TOTAL_CONTROL,
+                               OTG_SET_V_TOTAL_MIN_MASK, 0,
+                               OTG_V_TOTAL_MIN_SEL, 0,
+                               OTG_V_TOTAL_MAX_SEL, 0,
+                               OTG_FORCE_LOCK_ON_EVENT, 0);
+
+               optc->funcs->set_vtotal_min_max(optc, 0, 0);
+       }
 }


Looks like both the if and the else side paths end up programming
OTG_V_TOTAL_CONTROL differently after the change.  Perhaps
OTG_SET_V_TOTAL_MIN_MASK needs to be set differently depending on the
DMCUB firmware version? @Mikhail Gavrilov does this patch fix it?

diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
index 336488c0574e..933c7a342936 100644
--- a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
+++ b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.c
@@ -944,7 +944,7 @@ void optc1_set_drr(
                                OTG_V_TOTAL_MAX_SEL, 1,
                                OTG_FORCE_LOCK_ON_EVENT, 0,
                                OTG_SET_V_TOTAL_MIN_MASK_EN, 0,
-                               OTG_SET_V_TOTAL_MIN_MASK, 0);
+                               OTG_SET_V_TOTAL_MIN_MASK, (1 << 1)); /* TRI=
GA */

                // Setup manual flow control for EOF via TRIG_A
                optc->funcs->setup_manual_trigger(optc);


Thanks,

Alex

>
> --
> Best Regards,
> Mike Gavrilov.