Re: [PATCH v2 10/12] drm/rockchip: vop2: Add support for rk3588

From: Andy Yan
Date: Tue Nov 28 2023 - 05:18:52 EST


Hi Heiko:

On 11/28/23 17:44, Heiko Stübner wrote:
Hi Andy,

Am Dienstag, 28. November 2023, 10:32:55 CET schrieb Andy Yan:
On 11/27/23 23:29, Heiko Stübner wrote:
Am Mittwoch, 22. November 2023, 13:55:44 CET schrieb Andy Yan:
From: Andy Yan <andy.yan@xxxxxxxxxxxxxx>

VOP2 on rk3588:

Four video ports:
VP0 Max 4096x2160
VP1 Max 4096x2160
VP2 Max 4096x2160
VP3 Max 2048x1080

4 4K Cluster windows with AFBC/line RGB and AFBC-only YUV support
4 4K Esmart windows with line RGB/YUV support

Signed-off-by: Andy Yan <andy.yan@xxxxxxxxxxxxxx>

---

Changes in v2:
- add rk3588_ prefix for functions which are rk3588 only
- make some calculation as fixed value and keep calculation formula as
comment
- check return value for some cru calculation functions.
- check return value for syscon_regmap_lookup_by_phandle
- add NV20/NV30 for esmart plane

drivers/gpu/drm/rockchip/rockchip_drm_vop2.c | 381 ++++++++++++++++++-
drivers/gpu/drm/rockchip/rockchip_drm_vop2.h | 66 ++++
drivers/gpu/drm/rockchip/rockchip_vop2_reg.c | 221 +++++++++++
3 files changed, 660 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
index 4bcc405bcf11..9eecbe1f71f9 100644
--- a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
+++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
@@ -271,9 +282,12 @@ static bool vop2_cluster_window(const struct vop2_win *win)
static void vop2_cfg_done(struct vop2_video_port *vp)
{
struct vop2 *vop2 = vp->vop2;
+ u32 val;
+
+ val = BIT(vp->id) | (BIT(vp->id) << 16) |
+ RK3568_REG_CFG_DONE__GLB_CFG_DONE_EN;
- regmap_set_bits(vop2->map, RK3568_REG_CFG_DONE,
- BIT(vp->id) | RK3568_REG_CFG_DONE__GLB_CFG_DONE_EN);
+ regmap_set_bits(vop2->map, RK3568_REG_CFG_DONE, val);
I don't fully understand that code:
(1) the write mask is also present on the rk3568, so should this change
be a separate patch with a fixes tag?
The write mask of VP config done on rk356x is missing, that means
you can write the corresponding mask bit, but it has no effect.

I once considered making it a separate patch, I can split it as a separate patch if
you like.
I think I'd like it to be a separate patch please.


(2) RK3568_REG_CFG_DONE__GLB_CFG_DONE_EN does not contain the part for
the write-mask

#define RK3568_REG_CFG_DONE__GLB_CFG_DONE_EN BIT(15)

why is this working then?

Actually this bit has no write-mask bit. 🙂
when doing that separate patch mentioned above, could you also add a
comment to the code stating that RK3568_REG_CFG_DONE__GLB_CFG_DONE_EN
doesn't have a write mask bit please?

Because the TRM is not clear and ideally I'd not forget this fact for
the future :-) .


Okay, will do both above.

}
static void vop2_win_disable(struct vop2_win *win)
[...]

@@ -1298,7 +1346,11 @@ static void vop2_plane_atomic_update(struct drm_plane *plane,
vop2_win_write(win, VOP2_WIN_AFBC_ENABLE, 1);
vop2_win_write(win, VOP2_WIN_AFBC_FORMAT, afbc_format);
vop2_win_write(win, VOP2_WIN_AFBC_UV_SWAP, uv_swap);
- vop2_win_write(win, VOP2_WIN_AFBC_AUTO_GATING_EN, 0);
+ if (vop2->data->soc_id == 3566 || vop2->data->soc_id == 3568)
+ vop2_win_write(win, VOP2_WIN_AFBC_AUTO_GATING_EN, 0);
+ else
+ vop2_win_write(win, VOP2_WIN_AFBC_AUTO_GATING_EN, 1);
+
I think this at least warrants a comment, what is happening here. Also,
can you already see how future vop2-users are behaving - aka are all new
socs in the "else" part of the conditional, or would a switch-case better
represent future socs?

On rk356x, this bit is auto gating enable, but this function is not work well so
we need to disable this function.
On rk3588, and the following new soc(rk3528/rk3576), this bit is gating disable,
we should write 1 to disable gating when enable a cluster window.


Maybe i add some comments in next version ?
Yep that comment would be helpful. And with your explanation the code
itself can stay as it is :-)


will do.

Thanks
Heiko


vop2_win_write(win, VOP2_WIN_AFBC_BLOCK_SPLIT_EN, 0);
transform_offset = vop2_afbc_transform_offset(pstate, half_block_en);
vop2_win_write(win, VOP2_WIN_AFBC_HDR_PTR, yrgb_mst);
@@ -1627,9 +1937,17 @@ static void vop2_crtc_atomic_enable(struct drm_crtc *crtc,
drm_for_each_encoder_mask(encoder, crtc->dev, crtc_state->encoder_mask) {
struct rockchip_encoder *rkencoder = to_rockchip_encoder(encoder);
- rk3568_set_intf_mux(vp, rkencoder->crtc_endpoint_id, polflags);
+ /*
+ * for drive a high resolution(4KP120, 8K), vop on rk3588/rk3576 need
+ * process multi(1/2/4/8) pixels per cycle, so the dclk feed by the
+ * system cru may be the 1/2 or 1/4 of mode->clock.
+ */
+ clock = vop2_set_intf_mux(vp, rkencoder->crtc_endpoint_id, polflags);
}
+ if (!clock)
+ return;
+
hmm, shouldn't the check for the validity of a mode happen before
atomic_enable is run? So this shouldn't error out in the middle of the
function?


Actually it is a check like the check of clk_prepares_enable at the beginning,

maybe one place can do this is at crtc_atomic_check ? But we really don't need to

do the calculate and enable the related interface at every frame commit.

with a grep i can find many platforms do this kind of check in crtc_atomic_enable(ade/meson/vc4/omap/malidp/tidss_crtc_atomic_enable, etc...)

so maybe just let it as it is now?



if (vcstate->output_mode == ROCKCHIP_OUT_MODE_AAAA &&
!(vp_data->feature & VOP_FEATURE_OUTPUT_10BIT))
out_mode = ROCKCHIP_OUT_MODE_P888;
Thanks
Heiko



_______________________________________________
Linux-rockchip mailing list
Linux-rockchip@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/linux-rockchip