[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250321094109.116a6915@collabora.com>
Date: Fri, 21 Mar 2025 09:41:09 +0100
From: Boris Brezillon <boris.brezillon@...labora.com>
To: Karunika Choo <karunika.choo@....com>
Cc: dri-devel@...ts.freedesktop.org, nd@....com, Steven Price
<steven.price@....com>, Liviu Dudau <liviu.dudau@....com>, Maarten
Lankhorst <maarten.lankhorst@...ux.intel.com>, Maxime Ripard
<mripard@...nel.org>, Thomas Zimmermann <tzimmermann@...e.de>, David Airlie
<airlied@...il.com>, Simona Vetter <simona@...ll.ch>,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 7/9] drm/panthor: Support GPU_CONTROL cache flush
based on feature bit
On Thu, 20 Mar 2025 11:17:39 +0000
Karunika Choo <karunika.choo@....com> wrote:
> As the FLUSH_MEM and FLUSH_PT commands are deprecated in GPUs from
> Mali-G720 onwards, this patch adds support for performing cache
> maintenance via the FLUSH_CACHES command in GPU_CONTROL, in place of
> FLUSH_MEM and FLUSH_PT based on PANTHOR_HW_FEATURE_GPU_CTRL_CACHE_FLUSH
> feature bit.
>
> Signed-off-by: Karunika Choo <karunika.choo@....com>
> ---
> drivers/gpu/drm/panthor/panthor_hw.h | 6 +++++
> drivers/gpu/drm/panthor/panthor_mmu.c | 35 +++++++++++++++++++++++++++
> 2 files changed, 41 insertions(+)
>
> diff --git a/drivers/gpu/drm/panthor/panthor_hw.h b/drivers/gpu/drm/panthor/panthor_hw.h
> index dfe0f86c5d76..4d67fdfe86f9 100644
> --- a/drivers/gpu/drm/panthor/panthor_hw.h
> +++ b/drivers/gpu/drm/panthor/panthor_hw.h
> @@ -16,6 +16,12 @@ struct panthor_device;
> * New feature flags will be added with support for newer GPU architectures.
> */
> enum panthor_hw_feature {
> + /**
> + * @PANTHOR_HW_FEATURE_GPU_CTRL_CACHE_FLUSH: Perform cache maintenance
> + * via GPU_CONTROL.
> + */
> + PANTHOR_HW_FEATURE_GPU_CTRL_CACHE_FLUSH,
> +
> /** @PANTHOR_HW_FEATURES_END: Must be last. */
> PANTHOR_HW_FEATURES_END
> };
> diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
> index a0a79f19bdea..4ac8bf36177e 100644
> --- a/drivers/gpu/drm/panthor/panthor_mmu.c
> +++ b/drivers/gpu/drm/panthor/panthor_mmu.c
> @@ -29,7 +29,9 @@
>
> #include "panthor_device.h"
> #include "panthor_gem.h"
> +#include "panthor_gpu.h"
> #include "panthor_heap.h"
> +#include "panthor_hw.h"
> #include "panthor_mmu.h"
> #include "panthor_regs.h"
> #include "panthor_sched.h"
> @@ -568,6 +570,35 @@ static void lock_region(struct panthor_device *ptdev, u32 as_nr,
> write_cmd(ptdev, as_nr, AS_COMMAND_LOCK);
> }
>
> +static int mmu_hw_do_flush_on_gpu_ctrl(struct panthor_device *ptdev, int as_nr,
> + u32 op)
> +{
> + const u32 l2_flush_op = CACHE_CLEAN | CACHE_INV;
> + u32 lsc_flush_op = 0;
> + int ret;
> +
> + if (op == AS_COMMAND_FLUSH_MEM)
> + lsc_flush_op = CACHE_CLEAN | CACHE_INV;
> +
> + ret = wait_ready(ptdev, as_nr);
> + if (ret)
> + return ret;
> +
> + ret = panthor_gpu_flush_caches(ptdev, l2_flush_op, lsc_flush_op, 0);
> + if (ret)
> + return ret;
> +
> + /*
> + * Explicitly unlock the region as the AS is not unlocked automatically
> + * at the end of the GPU_CONTROL cache flush command, unlike
> + * AS_COMMAND_FLUSH_MEM or AS_COMMAND_FLUSH_PT.
> + */
> + write_cmd(ptdev, as_nr, AS_COMMAND_UNLOCK);
> +
> + /* Wait for the unlock command to complete */
> + return wait_ready(ptdev, as_nr);
> +}
> +
> static int mmu_hw_do_operation_locked(struct panthor_device *ptdev, int as_nr,
> u64 iova, u64 size, u32 op)
> {
> @@ -585,6 +616,10 @@ static int mmu_hw_do_operation_locked(struct panthor_device *ptdev, int as_nr,
> if (op != AS_COMMAND_UNLOCK)
> lock_region(ptdev, as_nr, iova, size);
>
> + if (panthor_hw_supports(ptdev,PANTHOR_HW_FEATURE_GPU_CTRL_CACHE_FLUSH))
> + if (op == AS_COMMAND_FLUSH_MEM || op == AS_COMMAND_FLUSH_PT)
> + return mmu_hw_do_flush_on_gpu_ctrl(ptdev, as_nr, op);
Can't we use this sequence on v10 as well? The GPU flush_cache command
exists there, so I'd rather switch all CSF HW to this sequence than
diverge on v11+.
> +
> /* Run the MMU operation */
> write_cmd(ptdev, as_nr, op);
>
Powered by blists - more mailing lists