Message-ID: <59FAE221.3080600@huawei.com>
Date:   Thu, 2 Nov 2017 17:15:13 +0800
From:   Zhong Kaihua <zhongkaihua@...wei.com>
To:     Julien Thierry <julien.thierry@....com>
CC:     <robh+dt@...nel.org>, <mark.rutland@....com>,
        <xuwei5@...ilicon.com>, <catalin.marinas@....com>,
        <will.deacon@....com>, <jassisinghbrar@...il.com>,
        <xuezhiliang@...ilicon.com>, <devicetree@...r.kernel.org>,
        <guodong.xu@...aro.org>, <suzhuangluan@...ilicon.com>,
        <linux-kernel@...r.kernel.org>, <haojian.zhuang@...aro.org>,
        <kevin.wangtao@...ilicon.com>,
        <linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH v2 2/3] mailbox: Add support for Hi3660 mailbox

Hi, Julien,

On 2017/10/27 18:40, Julien Thierry wrote:
> Hi Kaihua,
> 
> On 27/10/17 07:15, Kaihua Zhong wrote:
>> Hi3660 mailbox controller is used to send message within multiple
>> processors, MCU, HIFI, etc.  It supports 32 mailbox channels and every
>> channel can only be used for single transferring direction.  Once the
>> channel is enabled, it needs to specify the destination interrupt and
>> acknowledge interrupt, these two interrupt vectors are used to create
>> the connection between the mailbox and interrupt controllers.
>>
>> The application processor (or from point of view of kernel) is not the
>> only one master which can launch the data transferring, other
>> processors or MCU/DSP also can kick off the data transferring.  So this
>> driver implements a locking mechanism to support exclusive accessing.
>>
>> The data transferring supports two modes, one is named as "automatic
>> acknowledge" mode so after send message the kernel doesn't need to wait
>> for acknowledge from remote and directly return; there have another mode
>> is to rely on handling interrupt for acknowledge.
>>
>> This commit is for initial version driver, which only supports
>> "automatic acknowledge" mode to support CPU clock, which is the only
>> one consumer to use mailbox and has been verified.  Later may enhance
>> this driver for interrupt mode (e.g. for supporting HIFI).
>>
>> Cc: John Stultz <john.stultz@...aro.org>
>> Cc: Guodong Xu <guodong.xu@...aro.org>
>> Cc: Haojian Zhuang <haojian.zhuang@...aro.org>
>> Cc: Niranjan Yadla <nyadla@...ence.com>
>> Cc: Raj Pawate <pawateb@...ence.com>
>> Signed-off-by: Leo Yan <leo.yan@...aro.org>
>> Signed-off-by: Ruyi Wang <wangruyi@...wei.com>
>> Signed-off-by: Kaihua Zhong <zhongkaihua@...wei.com>
>> Signed-off-by: Kevin Wang <kevin.wangtao@...ilicon.com>
>> ---
>>   drivers/mailbox/Kconfig          |   8 +
>>   drivers/mailbox/Makefile         |   2 +
>>   drivers/mailbox/hi3660-mailbox.c | 331 +++++++++++++++++++++++++++++++++++++++
>>   3 files changed, 341 insertions(+)
>>   create mode 100644 drivers/mailbox/hi3660-mailbox.c
>>
>> diff --git a/drivers/mailbox/Kconfig b/drivers/mailbox/Kconfig
>> index c5731e5..4b5d6e9 100644
>> --- a/drivers/mailbox/Kconfig
>> +++ b/drivers/mailbox/Kconfig
>> @@ -108,6 +108,14 @@ config TI_MESSAGE_MANAGER
>>         multiple processors within the SoC. Select this driver if your
>>         platform has support for the hardware block.
>>   +config HI3660_MBOX
>> +    tristate "Hi3660 Mailbox"
>> +    depends on ARCH_HISI && OF
>> +    help
>> +      An implementation of the hi3660 mailbox. It is used to send message
>> +      between application processors and other processors/MCU/DSP. Select
>> +      Y here if you want to use Hi3660 mailbox controller.
>> +
>>   config HI6220_MBOX
>>       tristate "Hi6220 Mailbox"
>>       depends on ARCH_HISI
>> diff --git a/drivers/mailbox/Makefile b/drivers/mailbox/Makefile
>> index d54e412..7d1bd51 100644
>> --- a/drivers/mailbox/Makefile
>> +++ b/drivers/mailbox/Makefile
>> @@ -26,6 +26,8 @@ obj-$(CONFIG_TI_MESSAGE_MANAGER) += ti-msgmgr.o
>>     obj-$(CONFIG_XGENE_SLIMPRO_MBOX) += mailbox-xgene-slimpro.o
>>   +obj-$(CONFIG_HI3660_MBOX)    += hi3660-mailbox.o
>> +
>>   obj-$(CONFIG_HI6220_MBOX)    += hi6220-mailbox.o
>>     obj-$(CONFIG_BCM_PDC_MBOX)    += bcm-pdc-mailbox.o
>> diff --git a/drivers/mailbox/hi3660-mailbox.c b/drivers/mailbox/hi3660-mailbox.c
>> new file mode 100644
>> index 0000000..67df8f8
>> --- /dev/null
>> +++ b/drivers/mailbox/hi3660-mailbox.c
>> @@ -0,0 +1,331 @@
>> +/*
>> + * Hisilicon's Hi3660 mailbox controller driver
>> + *
>> + * Copyright (c) 2017 Hisilicon Limited.
>> + * Copyright (c) 2017 Linaro Limited.
>> + *
>> + * Author: Leo Yan <leo.yan@...aro.org>
>> + *
>> + * This program is free software: you can redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation, version 2 of the License.
>> + *
>> + * This program is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> + * GNU General Public License for more details.
>> + *
>> + */
>> +
>> +#include <linux/bitops.h>
>> +#include <linux/delay.h>
>> +#include <linux/device.h>
>> +#include <linux/err.h>
>> +#include <linux/interrupt.h>
>> +#include <linux/io.h>
>> +#include <linux/iopoll.h>
>> +#include <linux/mailbox_controller.h>
>> +#include <linux/module.h>
>> +#include <linux/platform_device.h>
>> +#include <linux/slab.h>
>> +
>> +#include "mailbox.h"
>> +
>> +#define MBOX_CHAN_MAX            32
>> +
>> +#define MBOX_RX                (0x0)
>> +#define MBOX_TX                (0x1)
>> +
>> +#define MBOX_BASE(mbox, ch)        ((mbox)->base + ((ch) * 0x40))
>> +#define MBOX_SRC_REG            (0x00)
>> +#define MBOX_DST_REG            (0x04)
>> +#define MBOX_DCLR_REG            (0x08)
>> +#define MBOX_DSTAT_REG            (0x0c)
>> +#define MBOX_MODE_REG            (0x10)
>> +#define MBOX_IMASK_REG            (0x14)
>> +#define MBOX_ICLR_REG            (0x18)
>> +#define MBOX_SEND_REG            (0x1c)
>> +#define MBOX_DATA_REG            (0x20)
>> +
>> +#define MBOX_IPC_LOCK_REG        (0xa00)
>> +#define MBOX_IPC_UNLOCK            (0x1acce551)
>> +
>> +#define MBOX_AUTOMATIC_ACK        (1)
>> +
>> +#define MBOX_STATE_IDLE            BIT(4)
>> +#define MBOX_STATE_ACK            BIT(7)
>> +
>> +#define MBOX_MSG_LEN            8
>> +
>> +/**
>> + * Hi3660 mailbox channel device data
>> + *
>> + * A channel can be used for TX or RX, it can trigger remote
>> + * processor interrupt to notify remote processor and can receive
>> + * interrupt if has incoming message.
>> + *
>> + * @dst_irq:    Interrupt vector for remote processor
>> + * @ack_irq:    Interrupt vector for local processor
>> + */
>> +struct hi3660_mbox_dev {
>> +    unsigned int dst_irq;
>> +    unsigned int ack_irq;
>> +};
>> +
>> +/**
>> + * Hi3660 mailbox controller data
>> + *
>> + * Mailbox controller includes 32 channels and can allocate
>> + * channel for message transferring.
>> + *
>> + * @dev:    Device to which it is attached
>> + * @base:    Base address of the register mapping region
>> + * @chan:    Representation of channels in mailbox controller
>> + * @mdev:    Representation of channel device data
>> + * @controller:    Representation of a communication channel controller
>> + */
>> +struct hi3660_mbox {
>> +    struct device *dev;
>> +    void __iomem *base;
>> +    struct mbox_chan chan[MBOX_CHAN_MAX];
>> +    struct hi3660_mbox_dev mdev[MBOX_CHAN_MAX];
>> +    struct mbox_controller controller;
>> +};
>> +
>> +static inline struct hi3660_mbox *to_hi3660_mbox(struct mbox_chan *chan)
>> +{
>> +    return container_of(chan->mbox, struct hi3660_mbox, controller);
>> +}
>> +
>> +static int hi3660_mbox_check_state(struct mbox_chan *chan)
>> +{
>> +    unsigned long ch = (unsigned long)chan->con_priv;
>> +    struct hi3660_mbox *mbox = to_hi3660_mbox(chan);
>> +    struct hi3660_mbox_dev *mdev = &mbox->mdev[ch];
>> +    void __iomem *base = MBOX_BASE(mbox, ch);
>> +    unsigned long val;
>> +    unsigned int state, ret;
>> +
>> +    /* Mailbox is idle so directly bail out */
>> +    state = readl_relaxed(base + MBOX_MODE_REG);
>> +    if (state & MBOX_STATE_IDLE)
>> +        return 0;
>> +
>> +    /* Wait for acknowledge from remote */
>> +    ret = readx_poll_timeout_atomic(readl_relaxed, base + MBOX_MODE_REG,
>> +            val, (val & MBOX_STATE_ACK), 1000, 300000);
>> +    if (ret) {
>> +        dev_err(mbox->dev, "%s: timeout for receiving ack\n", __func__);
>> +        return ret;
>> +    }
>> +
>> +    /* Ensure channel is released */
>> +    writel_relaxed(0xffffffff, base + MBOX_IMASK_REG);
>> +    writel_relaxed(BIT(mdev->ack_irq), base + MBOX_SRC_REG);
>> +    __asm__ volatile ("sev");
> 
> There is an existing sev() macro for this.
> 
Macro will be used in next patch，if this function is not redesigned.
Mark and Leo is discussing memory barrier protection issues.

>> +    return 0;
>> +}
>> +
>> +static int hi3660_mbox_unlock(struct mbox_chan *chan)
>> +{
>> +    struct hi3660_mbox *mbox = to_hi3660_mbox(chan);
>> +    unsigned int val, retry = 3;
>> +
>> +    do {
>> +        writel_relaxed(MBOX_IPC_UNLOCK, mbox->base + MBOX_IPC_LOCK_REG);
>> +
>> +        val = readl_relaxed(mbox->base + MBOX_IPC_LOCK_REG);
>> +        if (!val)
>> +            break;
>> +
>> +        udelay(10);
>> +    } while (retry--);
>> +
>> +    return (!val) ? 0 : -ETIMEDOUT;
>> +}
>> +
>> +static int hi3660_mbox_acquire_channel(struct mbox_chan *chan)
>> +{
>> +    unsigned long ch = (unsigned long)chan->con_priv;
>> +    struct hi3660_mbox *mbox = to_hi3660_mbox(chan);
>> +    struct hi3660_mbox_dev *mdev = &mbox->mdev[ch];
>> +    void __iomem *base = MBOX_BASE(mbox, ch);
>> +    unsigned int val = 0, retry = 10;
>> +
>> +    /*
>> +     * Hardware locking for exclusive accessing within CPUs
>> +     * without exclusive monitor mechanism.
>> +     */
>> +    do {
>> +        val = readl_relaxed(base + MBOX_MODE_REG);
>> +        if (!(val & MBOX_STATE_IDLE)) {
>> +            __asm__ volatile ("wfe");
>> +            continue;
> 
> So this is going to "wfe" when retry == 0 and the continue statement will take us out of the loop?
> 
> Also, there is a wfe() macro for arm and arm64 which could be used here.
> 
>> +        }
>> +
>> +        /* Check if channel has be acquired */
>> +        writel_relaxed(BIT(mdev->ack_irq), base + MBOX_SRC_REG);
>> +        val = readl_relaxed(base + MBOX_SRC_REG) & BIT(mdev->ack_irq);
>> +        if (val)
>> +            break;
>> +
>> +    } while (retry--);
>> +
>> +    return (val) ? 0 : -ETIMEDOUT;
> 
> If timeout occurs while waiting for the MBOX to be idle, val will hold the last value from MBOX_MODE_REG, and if this can be different than 0, this will hide the fact it timed out.
> 
> Maybe the following would be more appropriate:
> 
> return (retry >= 0) ? 0 : -ETIMEDOUT;
> 
> Also, your do {} while loops for the timeouts falsely give the impression we do "retry" attempts when we actually do "retry + 1" attempts (but I guess it doesn't make a big difference from the functional point of view).
> 

Agreed. The current logic of code is confusing as val is used twice.

>> +}
>> +
>> +static int hi3660_mbox_send(struct mbox_chan *chan, u32 *msg)
>> +{
>> +    unsigned long ch = (unsigned long)chan->con_priv;
>> +    struct hi3660_mbox *mbox = to_hi3660_mbox(chan);
>> +    struct hi3660_mbox_dev *mdev = &mbox->mdev[ch];
>> +    void __iomem *base = MBOX_BASE(mbox, ch);
>> +    unsigned int i;
>> +
>> +    /* Clear mask for destination interrupt */
>> +    writel_relaxed(~BIT(mdev->dst_irq), base + MBOX_IMASK_REG);
>> +
>> +    /* Config destination for interrupt vector */
>> +    writel_relaxed(BIT(mdev->dst_irq), base + MBOX_DST_REG);
>> +
>> +    /* Automatic acknowledge mode */
>> +    writel_relaxed(MBOX_AUTOMATIC_ACK, base + MBOX_MODE_REG);
>> +
>> +    /* Fill message data */
>> +    for (i = 0; i < MBOX_MSG_LEN; i++)
>> +        writel_relaxed(msg[i], base + MBOX_DATA_REG + i * 4);
>> +
>> +    /* Trigger data transferring */
>> +    writel_relaxed(BIT(mdev->ack_irq), base + MBOX_SEND_REG);
>> +    return 0;
>> +}
>> +
>> +static int hi3660_mbox_send_data(struct mbox_chan *chan, void *msg)
>> +{
>> +    struct hi3660_mbox *mbox = to_hi3660_mbox(chan);
>> +    int err;
>> +
>> +    err = hi3660_mbox_check_state(chan);
>> +    if (err) {
>> +        dev_err(mbox->dev, "checking state failed\n");
>> +        return err;
>> +    }
>> +
>> +    err = hi3660_mbox_unlock(chan);
>> +    if (err) {
>> +        dev_err(mbox->dev, "unlocking mailbox failed\n");
>> +        return err;
>> +    }
>> +
>> +    err = hi3660_mbox_acquire_channel(chan);
>> +    if (err) {
>> +        dev_err(mbox->dev, "acquiring channel failed\n");
>> +        return err;
>> +    }
>> +
>> +    return hi3660_mbox_send(chan, msg);
>> +}
>> +
>> +static struct mbox_chan_ops hi3660_mbox_ops = {
>> +    .send_data    = hi3660_mbox_send_data,
>> +};
>> +
>> +static struct mbox_chan *hi3660_mbox_xlate(struct mbox_controller *controller,
>> +                       const struct of_phandle_args *spec)
>> +{
>> +    struct hi3660_mbox *mbox = dev_get_drvdata(controller->dev);
> 
> nit:
> In to_hi3660_mbox, you use the fact that the mbox_controller structure is part of hi3660_mbox and use container of to retrieve hi3660_mbox.
> 
> I think it would be nice to be consistent about this, either use container_of or dev drvdata, but not both.
> 
> Cheers,
> 
I will discuss with Leo.

Thanks for your comments.

Best Regards,

Powered by blists - more mailing lists