Re: [RESEND, PATCH 4/5] soc: mediatek: mtk-cmdq: Add cmdq_pkt_acquire_event() function

From: "CK Hu (胡俊光)" <ck.hu@mediatek.com>
To: "jassisinghbrar@gmail.com" <jassisinghbrar@gmail.com>,
	"matthias.bgg@gmail.com" <matthias.bgg@gmail.com>,
	"Jason-JH Lin (林睿祥)" <Jason-JH.Lin@mediatek.com>,
	"chunkuang.hu@kernel.org" <chunkuang.hu@kernel.org>
Cc: "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"Singo Chang (張興國)" <Singo.Chang@mediatek.com>,
	"linux-mediatek@lists.infradead.org"
	<linux-mediatek@lists.infradead.org>,
	"Jason-ch Chen (陳建豪)" <Jason-ch.Chen@mediatek.com>,
	"Shawn Sung (宋孝謙)" <Shawn.Sung@mediatek.com>,
	"Nancy Lin (林欣螢)" <Nancy.Lin@mediatek.com>,
	Project_Global_Chrome_Upstream_Group
	<Project_Global_Chrome_Upstream_Group@mediatek.com>,
	"linux-arm-kernel@lists.infradead.org"
	<linux-arm-kernel@lists.infradead.org>,
	"angelogioacchino.delregno@collabora.com"
	<angelogioacchino.delregno@collabora.com>
Subject: Re: [RESEND, PATCH 4/5] soc: mediatek: mtk-cmdq: Add cmdq_pkt_acquire_event() function
Date: Mon, 4 Mar 2024 02:11:18 +0000	[thread overview]
Message-ID: <ade36b07c4f60a755a3b7700250e56ec61711d53.camel@mediatek.com> (raw)
In-Reply-To: <20240301144403.2977-5-jason-jh.lin@mediatek.com>

Hi, Jason:

On Fri, 2024-03-01 at 22:44 +0800, Jason-JH.Lin wrote:
> Add cmdq_pkt_acquire_event() function to support CMDQ user making
> an instruction for acquiring event.
> 
> CMDQ users can use cmdq_pkt_acquire_event() and
> cmdq_pkt_clear_event()
> to acquire GCE event and release GCE event and achieve the MUTEX_LOCK
> protection between GCE threads.

I'm not clear what acquire do in detail. This is what I guess:

cmdq_pkt_acquire_event() would wait for event to be cleared. After
event is cleared, cmdq_pkt_acquire_event() would set event and keep
executing next command. So the mutex would work like this

cmdq_pkt_acquire_event() /* mutex lock */

/* critical secton */

cmdq_pkt_clear_event() /* mutex unlock */

If it's so, describe as detail as this. If not, describe how it do.

As I know, GCE is single core, so multiple thread is served by single
GCE, why need mutex lock?

Regards,
CK

> 
> Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> ---
>  drivers/soc/mediatek/mtk-cmdq-helper.c | 15 +++++++++++++++
>  include/linux/soc/mediatek/mtk-cmdq.h  |  9 +++++++++
>  2 files changed, 24 insertions(+)
> 
> diff --git a/drivers/soc/mediatek/mtk-cmdq-helper.c
> b/drivers/soc/mediatek/mtk-cmdq-helper.c
> index 2e9fc9bb1183..0183b40a0eff 100644
> --- a/drivers/soc/mediatek/mtk-cmdq-helper.c
> +++ b/drivers/soc/mediatek/mtk-cmdq-helper.c
> @@ -342,6 +342,21 @@ int cmdq_pkt_wfe(struct cmdq_pkt *pkt, u16
> event, bool clear)
>  }
>  EXPORT_SYMBOL(cmdq_pkt_wfe);
>  
> +int cmdq_pkt_acquire_event(struct cmdq_pkt *pkt, u16 event)
> +{
> +	struct cmdq_instruction inst = {};
> +
> +	if (event >= CMDQ_MAX_EVENT)
> +		return -EINVAL;
> +
> +	inst.op = CMDQ_CODE_WFE;
> +	inst.value = CMDQ_WFE_UPDATE | CMDQ_WFE_UPDATE_VALUE |
> CMDQ_WFE_WAIT;
> +	inst.event = event;
> +
> +	return cmdq_pkt_append_command(pkt, inst);
> +}
> +EXPORT_SYMBOL(cmdq_pkt_acquire_event);
> +
>  int cmdq_pkt_clear_event(struct cmdq_pkt *pkt, u16 event)
>  {
>  	struct cmdq_instruction inst = { {0} };
> diff --git a/include/linux/soc/mediatek/mtk-cmdq.h
> b/include/linux/soc/mediatek/mtk-cmdq.h
> index 2fe9be240fbc..de93c0a8e8a9 100644
> --- a/include/linux/soc/mediatek/mtk-cmdq.h
> +++ b/include/linux/soc/mediatek/mtk-cmdq.h
> @@ -202,6 +202,15 @@ int cmdq_pkt_mem_move(struct cmdq_pkt *pkt,
> dma_addr_t src_addr, dma_addr_t dst_
>   */
>  int cmdq_pkt_wfe(struct cmdq_pkt *pkt, u16 event, bool clear);
>  
> +/**
> + * cmdq_pkt_acquire_event() - append acquire event command to the
> CMDQ packet
> + * @pkt:	the CMDQ packet
> + * @event:	the desired event to be acquired
> + *
> + * Return: 0 for success; else the error code is returned
> + */
> +int cmdq_pkt_acquire_event(struct cmdq_pkt *pkt, u16 event);
> +
>  /**
>   * cmdq_pkt_clear_event() - append clear event command to the CMDQ
> packet
>   * @pkt:	the CMDQ packet