Re: [PATCH v5 1/3] mailbox: Add support for APM X-Gene platform mailbox driver

From: Duc Dang
Date: Fri Feb 12 2016 - 22:26:02 EST

Next message: Jason Gunthorpe: "[PATCH 3/3] tpm: Provide strong locking for device removal"
Previous message: Keith Packard: "Re: [RFC] A first shot at asciidoc-based formatted docs"
In reply to: Mathieu Poirier: "Re: [PATCH v5 1/3] mailbox: Add support for APM X-Gene platform mailbox driver"
Next in thread: Duc Dang: "[PATCH v6 0/3] mailbox: Add APM X-Gene platform mailbox driver"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Wed, Feb 10, 2016 at 6:41 AM, Mathieu Poirier
<mathieu.poirier@xxxxxxxxxx> wrote:
> On 9 February 2016 at 20:46, Duc Dang <dhdang@xxxxxxx> wrote:
>> On Tue, Feb 9, 2016 at 8:40 AM, Mathieu Poirier
>> <mathieu.poirier@xxxxxxxxxx> wrote:
>>> On 8 February 2016 at 15:04, Duc Dang <dhdang@xxxxxxx> wrote:
>>>> X-Gene mailbox controller provides 8 mailbox channels, with
>>>> each channel has a dedicated interrupt line.
>>>>
>>>> Signed-off-by: Feng Kan <fkan@xxxxxxx>
>>>> Signed-off-by: Duc Dang <dhdang@xxxxxxx>
>>>> ---
>>>> Changes since v4:
>>>> - Rebase over v4.5-rc1
>>>> - Fix section mistmatch warning by removing
>>>> __init in slimpro_mbox_probe declaration
>>>> - Correctly print channel number when
>>>> there is no IRQ for that channel
>>>>
>>>> Changes since v3:
>>>> - Rebase over v4.4
>>>> - Remove 'id' in slimpro_mbox_chan structure
>>>> - Remove small functions that are only called once
>>>> and fold them into the other callers
>>>> - Remove void* pointer type cast
>>>> - Relax the number of mailbox IRQs condition
>>>> - Use subsys_initcall to guarantee mailbox driver
>>>> will be registered before any other dependent driver
>>>> is loaded.
>>>>
>>>> drivers/mailbox/Kconfig | 9 ++
>>>> drivers/mailbox/Makefile | 2 +
>>>> drivers/mailbox/mailbox-xgene-slimpro.c | 264 ++++++++++++++++++++++++++++++++
>>>> 3 files changed, 275 insertions(+)
>>>> create mode 100644 drivers/mailbox/mailbox-xgene-slimpro.c
>>>>
>>>> diff --git a/drivers/mailbox/Kconfig b/drivers/mailbox/Kconfig
>>>> index 546d05f..678e434 100644
>>>> --- a/drivers/mailbox/Kconfig
>>>> +++ b/drivers/mailbox/Kconfig
>>>> @@ -85,4 +85,13 @@ config MAILBOX_TEST
>>>> Test client to help with testing new Controller driver
>>>> implementations.
>>>>
>>>> +config XGENE_SLIMPRO_MBOX
>>>> + tristate "APM SoC X-Gene SLIMpro Mailbox Controller"
>>>> + depends on ARCH_XGENE
>>>> + help
>>>> + An implementation of the APM X-Gene Interprocessor Communication
>>>> + Mailbox (IPCM) between the ARM 64-bit cores and SLIMpro controller.
>>>> + It is used to send short messages between ARM64-bit cores and
>>>> + the SLIMpro Management Engine, primarily for PM. Say Y here if you
>>>> + want to use the APM X-Gene SLIMpro IPCM support.
>>>> endif
>>>> diff --git a/drivers/mailbox/Makefile b/drivers/mailbox/Makefile
>>>> index 92435ef..b602ef8 100644
>>>> --- a/drivers/mailbox/Makefile
>>>> +++ b/drivers/mailbox/Makefile
>>>> @@ -17,3 +17,5 @@ obj-$(CONFIG_ALTERA_MBOX) += mailbox-altera.o
>>>> obj-$(CONFIG_BCM2835_MBOX) += bcm2835-mailbox.o
>>>>
>>>> obj-$(CONFIG_STI_MBOX) += mailbox-sti.o
>>>> +
>>>> +obj-$(CONFIG_XGENE_SLIMPRO_MBOX) += mailbox-xgene-slimpro.o
>>>> diff --git a/drivers/mailbox/mailbox-xgene-slimpro.c b/drivers/mailbox/mailbox-xgene-slimpro.c
>>>> new file mode 100644
>>>> index 0000000..0ea1eb8
>>>> --- /dev/null
>>>> +++ b/drivers/mailbox/mailbox-xgene-slimpro.c
>>>> @@ -0,0 +1,264 @@
>>>> +/*
>>>> + * APM X-Gene SLIMpro MailBox Driver
>>>> + *
>>>> + * Copyright (c) 2015, Applied Micro Circuits Corporation
>>>> + * Author: Feng Kan fkan@xxxxxxx
>>>> + *
>>>> + * 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; either version 2 of
>>>> + * the License, or (at your option) any later version.
>>>> + *
>>>> + * 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.
>>>> + *
>>>> + * You should have received a copy of the GNU General Public License
>>>> + * along with this program; if not, see <http://www.gnu.org/licenses/>.
>>>> + *
>>>> + */
>>>> +#include <linux/acpi.h>
>>>> +#include <linux/delay.h>
>>>> +#include <linux/interrupt.h>
>>>> +#include <linux/io.h>
>>>> +#include <linux/mailbox_controller.h>
>>>> +#include <linux/module.h>
>>>> +#include <linux/of.h>
>>>> +#include <linux/platform_device.h>
>>>> +#include <linux/spinlock.h>
>>>> +
>>>> +#define MBOX_CON_NAME "slimpro-mbox"
>>>> +#define MBOX_REG_SET_OFFSET 0x1000
>>>> +#define MBOX_CNT 8
>>>> +#define MBOX_STATUS_AVAIL_MASK BIT(16)
>>>> +#define MBOX_STATUS_ACK_MASK BIT(0)
>>>> +
>>>> +/* Configuration and Status Registers */
>>>> +#define REG_DB_IN 0x00
>>>> +#define REG_DB_DIN0 0x04
>>>> +#define REG_DB_DIN1 0x08
>>>> +#define REG_DB_OUT 0x10
>>>> +#define REG_DB_DOUT0 0x14
>>>> +#define REG_DB_DOUT1 0x18
>>>> +#define REG_DB_STAT 0x20
>>>> +#define REG_DB_STATMASK 0x24
>>>> +
>>>> +struct slimpro_mbox_chan {
>>>> + struct device *dev;
>>>> + struct mbox_chan *chan;
>>>> + void __iomem *reg;
>>>> + int irq;
>>>> + u32 rx_msg[3];
>>>> +};
>>>> +
>>>> +struct slimpro_mbox {
>>>> + struct mbox_controller mb_ctrl;
>>>> + struct slimpro_mbox_chan mc[MBOX_CNT];
>>>> + struct mbox_chan chans[MBOX_CNT];
>>>> +};
>>>
>>> The above two structure declaration could use kernel style documentation.
>>
>> Can you please clarify or give an example? I looked into several files
>> in drivers/mailbox/ and drivers/irqchip but could not find what I need
>> to tweak.
>
> http://lxr.free-electrons.com/source/drivers/mailbox/mailbox-sti.c#L42

I am modifying the structure defintion to add description and will
post a new version shortly.

Regards,
Duc Dang.
>
>>
>>>
>>>> +
>>>> +static void mb_chan_send_msg(struct slimpro_mbox_chan *mb_chan, u32 *msg)
>>>> +{
>>>> + writel(msg[1], mb_chan->reg + REG_DB_DOUT0);
>>>> + writel(msg[2], mb_chan->reg + REG_DB_DOUT1);
>>>> + writel(msg[0], mb_chan->reg + REG_DB_OUT);
>>>> +}
>>>> +
>>>> +static void mb_chan_recv_msg(struct slimpro_mbox_chan *mb_chan)
>>>> +{
>>>> + mb_chan->rx_msg[1] = readl(mb_chan->reg + REG_DB_DIN0);
>>>> + mb_chan->rx_msg[2] = readl(mb_chan->reg + REG_DB_DIN1);
>>>> + mb_chan->rx_msg[0] = readl(mb_chan->reg + REG_DB_IN);
>>>> +}
>>>
>>> Is there a requirement that operations in the above two functions be
>>> ordered? If not significant gains could be achieved by using
>>> writel/readl_relaxed(). Same comment for the rest of this driver.
>>
>> The 'writes' need to be in order as writing to REG_DB_OUT will trigger
>> interrupt, so I want the other 2 'writes' complete before it.
>>
>> After the receiver get the message, it will clear door bell available
>> interrupt request, this will also trigger a door bell acknowledge
>> interrupt. So the 'read' to get the message and the 'write' to clear
>> the door bell available interrupt must be in order as well.
>>
>> Regards,
>> Duc Dang.
>>>
>>> Thanks,
>>> Mathieu
>>>
>>>> +
>>>> +static int mb_chan_status_ack(struct slimpro_mbox_chan *mb_chan)
>>>> +{
>>>> + u32 val = readl(mb_chan->reg + REG_DB_STAT);
>>>> +
>>>> + if (val & MBOX_STATUS_ACK_MASK) {
>>>> + writel(MBOX_STATUS_ACK_MASK, mb_chan->reg + REG_DB_STAT);
>>>> + return 1;
>>>> + }
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static int mb_chan_status_avail(struct slimpro_mbox_chan *mb_chan)
>>>> +{
>>>> + u32 val = readl(mb_chan->reg + REG_DB_STAT);
>>>> +
>>>> + if (val & MBOX_STATUS_AVAIL_MASK) {
>>>> + mb_chan_recv_msg(mb_chan);
>>>> + writel(MBOX_STATUS_AVAIL_MASK, mb_chan->reg + REG_DB_STAT);
>>>> + return 1;
>>>> + }
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static irqreturn_t slimpro_mbox_irq(int irq, void *id)
>>>> +{
>>>> + struct slimpro_mbox_chan *mb_chan = id;
>>>> +
>>>> + if (mb_chan_status_ack(mb_chan))
>>>> + mbox_chan_txdone(mb_chan->chan, 0);
>>>> +
>>>> + if (mb_chan_status_avail(mb_chan))
>>>> + mbox_chan_received_data(mb_chan->chan, mb_chan->rx_msg);
>>>> +
>>>> + return IRQ_HANDLED;
>>>> +}
>>>> +
>>>> +static int slimpro_mbox_send_data(struct mbox_chan *chan, void *msg)
>>>> +{
>>>> + struct slimpro_mbox_chan *mb_chan = chan->con_priv;
>>>> +
>>>> + mb_chan_send_msg(mb_chan, msg);
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static int slimpro_mbox_startup(struct mbox_chan *chan)
>>>> +{
>>>> + struct slimpro_mbox_chan *mb_chan = chan->con_priv;
>>>> + int rc;
>>>> + u32 val;
>>>> +
>>>> + rc = devm_request_irq(mb_chan->dev, mb_chan->irq, slimpro_mbox_irq, 0,
>>>> + MBOX_CON_NAME, mb_chan);
>>>> + if (unlikely(rc)) {
>>>> + dev_err(mb_chan->dev, "failed to register mailbox interrupt %d\n",
>>>> + mb_chan->irq);
>>>> + return rc;
>>>> + }
>>>> +
>>>> + /* Enable HW interrupt */
>>>> + writel(MBOX_STATUS_ACK_MASK | MBOX_STATUS_AVAIL_MASK,
>>>> + mb_chan->reg + REG_DB_STAT);
>>>> + /* Unmask doorbell status interrupt */
>>>> + val = readl(mb_chan->reg + REG_DB_STATMASK);
>>>> + val &= ~(MBOX_STATUS_ACK_MASK | MBOX_STATUS_AVAIL_MASK);
>>>> + writel(val, mb_chan->reg + REG_DB_STATMASK);
>>>> +
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static void slimpro_mbox_shutdown(struct mbox_chan *chan)
>>>> +{
>>>> + struct slimpro_mbox_chan *mb_chan = chan->con_priv;
>>>> + u32 val;
>>>> +
>>>> + /* Mask doorbell status interrupt */
>>>> + val = readl(mb_chan->reg + REG_DB_STATMASK);
>>>> + val |= (MBOX_STATUS_ACK_MASK | MBOX_STATUS_AVAIL_MASK);
>>>> + writel(val, mb_chan->reg + REG_DB_STATMASK);
>>>> +
>>>> + devm_free_irq(mb_chan->dev, mb_chan->irq, mb_chan);
>>>> +}
>>>> +
>>>> +static struct mbox_chan_ops slimpro_mbox_ops = {
>>>> + .send_data = slimpro_mbox_send_data,
>>>> + .startup = slimpro_mbox_startup,
>>>> + .shutdown = slimpro_mbox_shutdown,
>>>> +};
>>>> +
>>>> +static int slimpro_mbox_probe(struct platform_device *pdev)
>>>> +{
>>>> + struct slimpro_mbox *ctx;
>>>> + struct resource *regs;
>>>> + void __iomem *mb_base;
>>>> + int rc;
>>>> + int i;
>>>> +
>>>> + ctx = devm_kzalloc(&pdev->dev, sizeof(struct slimpro_mbox), GFP_KERNEL);
>>>> + if (IS_ERR(ctx))
>>>> + return PTR_ERR(ctx);
>>>> +
>>>> + platform_set_drvdata(pdev, ctx);
>>>> +
>>>> + regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>>>> + mb_base = devm_ioremap(&pdev->dev, regs->start, resource_size(regs));
>>>> + if (IS_ERR(mb_base))
>>>> + return PTR_ERR(mb_base);
>>>> +
>>>> + /* Setup mailbox links */
>>>> + for (i = 0; i < MBOX_CNT; i++) {
>>>> + ctx->mc[i].irq = platform_get_irq(pdev, i);
>>>> + if (ctx->mc[i].irq < 0) {
>>>> + if (i == 0) {
>>>> + dev_err(&pdev->dev, "no available IRQ\n");
>>>> + return -EINVAL;
>>>> + }
>>>> + dev_info(&pdev->dev, "no IRQ for channel %d\n", i);
>>>> + break;
>>>> + }
>>>> +
>>>> + ctx->mc[i].dev = &pdev->dev;
>>>> + ctx->mc[i].reg = mb_base + i * MBOX_REG_SET_OFFSET;
>>>> + ctx->mc[i].chan = &ctx->chans[i];
>>>> + ctx->chans[i].con_priv = &ctx->mc[i];
>>>> + }
>>>> +
>>>> + /* Setup mailbox controller */
>>>> + ctx->mb_ctrl.dev = &pdev->dev;
>>>> + ctx->mb_ctrl.chans = ctx->chans;
>>>> + ctx->mb_ctrl.txdone_irq = true;
>>>> + ctx->mb_ctrl.ops = &slimpro_mbox_ops;
>>>> + ctx->mb_ctrl.num_chans = i;
>>>> +
>>>> + rc = mbox_controller_register(&ctx->mb_ctrl);
>>>> + if (rc) {
>>>> + dev_err(&pdev->dev,
>>>> + "APM X-Gene SLIMpro MailBox register failed:%d\n", rc);
>>>> + return rc;
>>>> + }
>>>> +
>>>> + dev_info(&pdev->dev, "APM X-Gene SLIMpro MailBox registered\n");
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static int slimpro_mbox_remove(struct platform_device *pdev)
>>>> +{
>>>> + struct slimpro_mbox *smb = platform_get_drvdata(pdev);
>>>> +
>>>> + mbox_controller_unregister(&smb->mb_ctrl);
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static const struct of_device_id slimpro_of_match[] = {
>>>> + {.compatible = "apm,xgene-slimpro-mbox" },
>>>> + { },
>>>> +};
>>>> +MODULE_DEVICE_TABLE(of, slimpro_of_match);
>>>> +
>>>> +#ifdef CONFIG_ACPI
>>>> +static const struct acpi_device_id slimpro_acpi_ids[] = {
>>>> + {"APMC0D01", 0},
>>>> + {}
>>>> +};
>>>> +MODULE_DEVICE_TABLE(acpi, slimpro_acpi_ids);
>>>> +#endif
>>>> +
>>>> +static struct platform_driver slimpro_mbox_driver = {
>>>> + .probe = slimpro_mbox_probe,
>>>> + .remove = slimpro_mbox_remove,
>>>> + .driver = {
>>>> + .name = "xgene-slimpro-mbox",
>>>> + .of_match_table = of_match_ptr(slimpro_of_match),
>>>> + .acpi_match_table = ACPI_PTR(slimpro_acpi_ids)
>>>> + },
>>>> +};
>>>> +
>>>> +static int __init slimpro_mbox_init(void)
>>>> +{
>>>> + return platform_driver_register(&slimpro_mbox_driver);
>>>> +}
>>>> +
>>>> +static void __exit slimpro_mbox_exit(void)
>>>> +{
>>>> + platform_driver_unregister(&slimpro_mbox_driver);
>>>> +}
>>>> +
>>>> +subsys_initcall(slimpro_mbox_init);
>>>> +module_exit(slimpro_mbox_exit);
>>>> +
>>>> +MODULE_DESCRIPTION("APM X-Gene SLIMpro Mailbox Driver");
>>>> +MODULE_LICENSE("GPL");
>>>> --
>>>> 1.9.1
>>>>

Next message: Jason Gunthorpe: "[PATCH 3/3] tpm: Provide strong locking for device removal"
Previous message: Keith Packard: "Re: [RFC] A first shot at asciidoc-based formatted docs"
In reply to: Mathieu Poirier: "Re: [PATCH v5 1/3] mailbox: Add support for APM X-Gene platform mailbox driver"
Next in thread: Duc Dang: "[PATCH v6 0/3] mailbox: Add APM X-Gene platform mailbox driver"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]