Message-ID: <1530663042.21991.64.camel@mhfsdcap03>
Date:   Wed, 4 Jul 2018 08:10:42 +0800
From:   houlong wei <houlong.wei@...iatek.com>
To:     CK Hu <ck.hu@...iatek.com>
CC:     Jassi Brar <jassisinghbrar@...il.com>,
        Matthias Brugger <matthias.bgg@...il.com>,
        Rob Herring <robh+dt@...nel.org>,
        Daniel Kurtz <djkurtz@...omium.org>,
        Sascha Hauer <s.hauer@...gutronix.de>,
        "devicetree@...r.kernel.org" <devicetree@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        "linux-mediatek@...ts.infradead.org" 
        <linux-mediatek@...ts.infradead.org>,
        srv_heupstream <srv_heupstream@...iatek.com>,
        Sascha Hauer <kernel@...gutronix.de>,
        Philipp Zabel <p.zabel@...gutronix.de>,
        "Nicolas Boichat" <drinkcat@...omium.org>,
        Bibby Hsieh (謝濟遠) 
        <Bibby.Hsieh@...iatek.com>,
        YT Shen (沈岳霆) 
        <Yt.Shen@...iatek.com>,
        Daoyuan Huang (黃道原) 
        <Daoyuan.Huang@...iatek.com>,
        Jiaguang Zhang (张加广) 
        <Jiaguang.Zhang@...iatek.com>,
        Dennis-YC Hsieh (謝宇哲) 
        <Dennis-YC.Hsieh@...iatek.com>,
        Monica Wang (王孟婷) 
        <monica.wang@...iatek.com>,
        Hs Liao (廖宏祥) 
        <Hs.Liao@...iatek.com>,
        Ginny Chen (陳治傑) 
        <ginny.chen@...iatek.com>,
        Enzhu Wang (王恩柱) 
        <enzhu.wang@...iatek.com>, <houlong.wei@...iatek.com>
Subject: Re: [PATCH v22 2/4] mailbox: mediatek: Add Mediatek CMDQ driver

On Fri, 2018-06-29 at 15:08 +0800, CK Hu wrote:
> Hi, Houlong:
> 
> Some inline comment.
> 
> On Wed, 2018-06-27 at 19:16 +0800, Houlong Wei wrote:
> > This patch is first version of Mediatek Command Queue(CMDQ) driver. The
> > CMDQ is used to help write registers with critical time limitation,
> > such as updating display configuration during the vblank. It controls
> > Global Command Engine (GCE) hardware to achieve this requirement.
> > Currently, CMDQ only supports display related hardwares, but we expect
> > it can be extended to other hardwares for future requirements.
> > 
> > Signed-off-by: Houlong Wei <houlong.wei@...iatek.com>
> > Signed-off-by: HS Liao <hs.liao@...iatek.com>
> > Signed-off-by: CK Hu <ck.hu@...iatek.com>
> > ---
> >  drivers/mailbox/Kconfig                  |   10 +
> >  drivers/mailbox/Makefile                 |    2 +
> >  drivers/mailbox/mtk-cmdq-mailbox.c       |  634 ++++++++++++++++++++++++++++++
> >  include/linux/mailbox/mtk-cmdq-mailbox.h |   70 ++++
> >  4 files changed, 716 insertions(+)
> >  create mode 100644 drivers/mailbox/mtk-cmdq-mailbox.c
> >  create mode 100644 include/linux/mailbox/mtk-cmdq-mailbox.h
> > 
> 
> [...]
> 
> > +
> > +static int cmdq_thread_reset(struct cmdq *cmdq, struct cmdq_thread *thread)
> > +{
> > +	u32 warm_reset;
> > +
> > +	writel(CMDQ_THR_DO_WARM_RESET, thread->base + CMDQ_THR_WARM_RESET);
> > +	if (readl_poll_timeout_atomic(thread->base + CMDQ_THR_WARM_RESET,
> > +			warm_reset, !(warm_reset & CMDQ_THR_DO_WARM_RESET),
> > +			0, 10)) {
> > +		dev_err(cmdq->mbox.dev, "reset GCE thread 0x%x failed\n",
> > +			(u32)(thread->base - cmdq->base));
> > +		return -EFAULT;
> > +	}
> > +	writel(CMDQ_THR_ACTIVE_SLOT_CYCLES, cmdq->base + CMDQ_THR_SLOT_CYCLES);
> 
> The CMDQ_THR_SLOT_CYCLES looks like not relevant to reset. Maybe you
> just need to set this value when startup.

Will move configuration of CMDQ_THR_SLOT_CYCLES to cmdq_xlate() where is
startup of a GCE thread.

> 
> > +
> > +	return 0;
> > +}
> > +
> 
> [...]
> 
> > +
> > +static void cmdq_task_exec(struct cmdq_pkt *pkt, struct cmdq_thread *thread)
> > +{
> > +	struct cmdq *cmdq;
> > +	struct cmdq_task *task;
> > +	unsigned long curr_pa, end_pa;
> > +
> > +	cmdq = dev_get_drvdata(thread->chan->mbox->dev);
> > +
> > +	/* Client should not flush new tasks if suspended. */
> > +	WARN_ON(cmdq->suspended);
> > +
> > +	task = kzalloc(sizeof(*task), GFP_ATOMIC);
> > +	task->cmdq = cmdq;
> > +	INIT_LIST_HEAD(&task->list_entry);
> > +	task->pa_base = pkt->pa_base;
> > +	task->thread = thread;
> > +	task->pkt = pkt;
> > +
> > +	if (list_empty(&thread->task_busy_list)) {
> > +		WARN_ON(clk_enable(cmdq->clock) < 0);
> > +		WARN_ON(cmdq_thread_reset(cmdq, thread) < 0);
> > +
> > +		writel(task->pa_base, thread->base + CMDQ_THR_CURR_ADDR);
> > +		writel(task->pa_base + pkt->cmd_buf_size,
> > +		       thread->base + CMDQ_THR_END_ADDR);
> > +		writel(thread->priority, thread->base + CMDQ_THR_PRIORITY);
> > +		writel(CMDQ_THR_IRQ_EN, thread->base + CMDQ_THR_IRQ_ENABLE);
> > +		writel(CMDQ_THR_ENABLED, thread->base + CMDQ_THR_ENABLE_TASK);
> > +
> > +		if (thread->timeout_ms != CMDQ_NO_TIMEOUT)
> > +			mod_timer(&thread->timeout,
> > +				jiffies + msecs_to_jiffies(thread->timeout_ms));
> 
> I think the timeout processing should be done by client driver. The
> total time to execute a command buffer does not depend on GCE HW speed
> because the WFE (wait for event) command would wait for client HW event,
> so the total time depend on how long a client HW send this event to GCE
> and the timeout processing should be client driver's job. Each client
> may have different timeout processing mechanism, for example, if display
> could dynamic change panel frame rate between 120Hz and 60Hz, and the
> timeout time is 2 frame, so it may dynamically change timeout time
> between 17ms and 33ms. Another reason is that display have interrupt
> every vblank, and it could check timeout in that interrupt, so the timer
> in cmdq driver looks redundant. Because each client would define its own
> timeout processing mechanism, so it's not wise to put timeout processing
> in cmdq driver.

The client drivers' owners strongly hope to keep the current timeout
mechanism, the reasons are below.
1. If remove, all clients should add timeout mechanism and the code will
be redundant.
2. If timeout happens, only GCE driver can do reset and continue to
execute next packet.

> 
> > +	} else {
> > +		WARN_ON(cmdq_thread_suspend(cmdq, thread) < 0);
> > +		curr_pa = readl(thread->base + CMDQ_THR_CURR_ADDR);
> > +		end_pa = readl(thread->base + CMDQ_THR_END_ADDR);
> > +
> > +		/*
> > +		 * Atomic execution should remove the following wfe, i.e. only
> > +		 * wait event at first task, and prevent to pause when running.
> > +		 */
> > +		if (thread->atomic_exec) {
> > +			/* GCE is executing if command is not WFE */
> > +			if (!cmdq_thread_is_in_wfe(thread)) {
> > +				cmdq_thread_resume(thread);
> > +				cmdq_thread_wait_end(thread, end_pa);
> > +				WARN_ON(cmdq_thread_suspend(cmdq, thread) < 0);
> > +				/* set to this task directly */
> > +				writel(task->pa_base,
> > +				       thread->base + CMDQ_THR_CURR_ADDR);
> > +			} else {
> > +				cmdq_task_insert_into_thread(task);
> > +				cmdq_task_remove_wfe(task);
> > +				smp_mb(); /* modify jump before enable thread */
> > +			}
> > +		} else {
> > +			/* check boundary */
> > +			if (curr_pa == end_pa - CMDQ_INST_SIZE ||
> > +			    curr_pa == end_pa) {
> > +				/* set to this task directly */
> > +				writel(task->pa_base,
> > +				       thread->base + CMDQ_THR_CURR_ADDR);
> > +			} else {
> > +				cmdq_task_insert_into_thread(task);
> > +				smp_mb(); /* modify jump before enable thread */
> > +			}
> > +		}
> > +		writel(task->pa_base + pkt->cmd_buf_size,
> > +		       thread->base + CMDQ_THR_END_ADDR);
> > +		cmdq_thread_resume(thread);
> > +	}
> > +	list_move_tail(&task->list_entry, &thread->task_busy_list);
> > +}
> > +
> > +static void cmdq_task_exec_done(struct cmdq_task *task, bool err)
> > +{
> > +	struct device *dev = task->cmdq->mbox.dev;
> > +	struct cmdq_cb_data cmdq_cb_data;
> > +
> > +	dma_unmap_single(dev, task->pa_base, task->pkt->cmd_buf_size,
> > +			 DMA_TO_DEVICE);
> 
> Move this to client driver.

map/unmap are common code for clients driver, could we move it to cmdq
helper?

> 
> > +	if (task->pkt->cb.cb) {
> > +		cmdq_cb_data.err = err;
> > +		cmdq_cb_data.data = task->pkt->cb.data;
> > +		task->pkt->cb.cb(cmdq_cb_data);
> > +	}
> > +	list_del(&task->list_entry);
> > +}
> > +
> 
> [...]
> 
> > +
> > +static bool cmdq_mbox_last_tx_done(struct mbox_chan *chan)
> > +{
> > +	return true;
> > +}
> > +
> > +static const struct mbox_chan_ops cmdq_mbox_chan_ops = {
> > +	.send_data = cmdq_mbox_send_data,
> > +	.startup = cmdq_mbox_startup,
> > +	.shutdown = cmdq_mbox_shutdown,
> > +	.last_tx_done = cmdq_mbox_last_tx_done,
> 
> Because mbox->txdone_poll is false, so you need not to implement
> last_tx_done.
> 
> Regards,
> CK

Will remove cmdq_mbox_last_tx_done().

> 
> > +};
> > +
> > +static struct mbox_chan *cmdq_xlate(struct mbox_controller *mbox,
> > +		const struct of_phandle_args *sp)
> > +{
> > +	int ind = sp->args[0];
> > +	struct cmdq_thread *thread;
> > +
> > +	if (ind >= mbox->num_chans)
> > +		return ERR_PTR(-EINVAL);
> > +
> > +	thread = mbox->chans[ind].con_priv;
> > +	thread->timeout_ms = sp->args[1];
> > +	thread->priority = sp->args[2];
> > +	thread->atomic_exec = (sp->args[3] != 0);
> > +	thread->chan = &mbox->chans[ind];
> > +
> > +	return &mbox->chans[ind];
> > +}
> > +
> [...]
> 
> 

Powered by blists - more mailing lists