Re: [PATCH 1/5] HDQ Driver for OMAP2430/3430

From: Andrew Morton
Date: Fri Oct 10 2008 - 16:57:37 EST


On Wed, 8 Oct 2008 12:46:25 +0530
"Gadiyar, Anand" <gadiyar@xxxxxx> wrote:

> From: Madhusudhan Chikkature <madhu.cr@xxxxxx>
>
> The HDQ/1-Wire module of TI OMAP2430/3430 platforms implement the hardware
> protocol of the master functions of the Benchmark HDQ and the Dallas
> Semiconductor 1-Wire protocols. These protocols use a single wire for
> communication between the master (HDQ/1-Wire controller) and the slave
> (HDQ/1-Wire external compliant device).
>
> This patch provides the HDQ driver to suppport TI OMAP2430/3430 platforms.

Every tab character in all five patches was converted to eight-spaces by
your email client. Please fix the mailer and resend everything.

> +++ linux-2.6/drivers/w1/masters/omap_hdq.c 2008-09-26 14:28:36.000000000 +0530
> @@ -0,0 +1,730 @@
> +/*
> + * drivers/w1/masters/omap_hdq.c
> + *
> + * Copyright (C) 2007 Texas Instruments, Inc.
> + *
> + * This file is licensed under the terms of the GNU General Public License
> + * version 2. This program is licensed "as is" without any warranty of any
> + * kind, whether express or implied.
> + *
> + */
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/platform_device.h>
> +#include <linux/interrupt.h>
> +#include <linux/err.h>
> +#include <linux/clk.h>
> +#include <linux/io.h>
> +#include <asm/irq.h>
> +#include <mach/hardware.h>

We conventionally put a blank line between the linux/ includes and the
asm/ includes.

> +static int omap_hdq_get(struct hdq_data *hdq_data);
> +static int omap_hdq_put(struct hdq_data *hdq_data);
> +static int omap_hdq_break(struct hdq_data *hdq_data);

These three aren't strictly needed, because these functions are defined
before first use. I think it's best to not declare them.

> +static int hdq_wait_for_flag(struct hdq_data *hdq_data, u32 offset,
> + u8 flag, u8 flag_set, u8 *status)
> +{
> + int ret = 0;
> + unsigned long timeout = jiffies + OMAP_HDQ_TIMEOUT;
> +
> + if (flag_set == OMAP_HDQ_FLAG_CLEAR) {
> + /* wait for the flag clear */
> + while (((*status = hdq_reg_in(hdq_data, offset)) & flag)
> + && time_before(jiffies, timeout)) {
> + set_current_state(TASK_UNINTERRUPTIBLE);
> + schedule_timeout(1);

Use schedule_timeout_uninterruptible(1)

> + }
> + if (*status & flag)
> + ret = -ETIMEDOUT;
> + } else if (flag_set == OMAP_HDQ_FLAG_SET) {
> + /* wait for the flag set */
> + while (!((*status = hdq_reg_in(hdq_data, offset)) & flag)
> + && time_before(jiffies, timeout)) {
> + set_current_state(TASK_UNINTERRUPTIBLE);
> + schedule_timeout(1);

elsewhere..

> + }
> + if (!(*status & flag))
> + ret = -ETIMEDOUT;
> + } else
> + return -EINVAL;
> +
> + return ret;
> +}
> +
> +/* write out a byte and fill *status with HDQ_INT_STATUS */
> +static int hdq_write_byte(struct hdq_data *hdq_data, u8 val, u8 *status)
> +{
> + int ret;
> + u8 tmp_status;
> + unsigned long irqflags;
> +
> + *status = 0;
> +
> + spin_lock_irqsave(&hdq_data->hdq_spinlock, irqflags);
> + /* clear interrupt flags via a dummy read */
> + hdq_reg_in(hdq_data, OMAP_HDQ_INT_STATUS);
> + /* ISR loads it with new INT_STATUS */
> + hdq_data->hdq_irqstatus = 0;
> + spin_unlock_irqrestore(&hdq_data->hdq_spinlock, irqflags);
> +
> + hdq_reg_out(hdq_data, OMAP_HDQ_TX_DATA, val);
> +
> + /* set the GO bit */
> + hdq_reg_merge(hdq_data, OMAP_HDQ_CTRL_STATUS, OMAP_HDQ_CTRL_STATUS_GO,
> + OMAP_HDQ_CTRL_STATUS_DIR | OMAP_HDQ_CTRL_STATUS_GO);
> + /* wait for the TXCOMPLETE bit */
> + ret = wait_event_interruptible_timeout(hdq_wait_queue,
> + hdq_data->hdq_irqstatus, OMAP_HDQ_TIMEOUT);
> + if (ret < 0) {
> + dev_dbg(hdq_data->dev, "wait interrupted");
> + return -EINTR;
> + }

Is this desirable? The user hits ^C and the driver bails out?

I assume so, but was this tested?

> + spin_lock_irqsave(&hdq_data->hdq_spinlock, irqflags);
> + *status = hdq_data->hdq_irqstatus;
> + spin_unlock_irqrestore(&hdq_data->hdq_spinlock, irqflags);

It's unusual to put a lock around a single atomic move instruction.

> + /* check irqstatus */
> + if (!(*status & OMAP_HDQ_INT_STATUS_TXCOMPLETE)) {
> + dev_dbg(hdq_data->dev, "timeout waiting for"
> + "TXCOMPLETE/RXCOMPLETE, %x", *status);
> + return -ETIMEDOUT;
> + }
> +
> + /* wait for the GO bit return to zero */
> + ret = hdq_wait_for_flag(hdq_data, OMAP_HDQ_CTRL_STATUS,
> + OMAP_HDQ_CTRL_STATUS_GO,
> + OMAP_HDQ_FLAG_CLEAR, &tmp_status);
> + if (ret) {
> + dev_dbg(hdq_data->dev, "timeout waiting GO bit"
> + "return to zero, %x", tmp_status);
> + return ret;
> + }
> +
> + return ret;
> +}
>
> ...
>
> +/* Issue break pulse to the device */
> +static int omap_hdq_break(struct hdq_data *hdq_data)
> +{
> + int ret;
> + u8 tmp_status;
> + unsigned long irqflags;
> +
> + ret = mutex_lock_interruptible(&hdq_data->hdq_mutex);
> + if (ret < 0)
> + return -EINTR;
> +
> + if (!hdq_data->hdq_usecount) {
> + mutex_unlock(&hdq_data->hdq_mutex);
> + return -EINVAL;
> + }
> +
> + spin_lock_irqsave(&hdq_data->hdq_spinlock, irqflags);
> + /* clear interrupt flags via a dummy read */
> + hdq_reg_in(hdq_data, OMAP_HDQ_INT_STATUS);
> + /* ISR loads it with new INT_STATUS */
> + hdq_data->hdq_irqstatus = 0;
> + spin_unlock_irqrestore(&hdq_data->hdq_spinlock, irqflags);
> +
> + /* set the INIT and GO bit */
> + hdq_reg_merge(hdq_data, OMAP_HDQ_CTRL_STATUS,
> + OMAP_HDQ_CTRL_STATUS_INITIALIZATION | OMAP_HDQ_CTRL_STATUS_GO,
> + OMAP_HDQ_CTRL_STATUS_DIR | OMAP_HDQ_CTRL_STATUS_INITIALIZATION |
> + OMAP_HDQ_CTRL_STATUS_GO);
> +
> + /* wait for the TIMEOUT bit */
> + ret = wait_event_interruptible_timeout(hdq_wait_queue,
> + hdq_data->hdq_irqstatus, OMAP_HDQ_TIMEOUT);
> + if (ret < 0) {
> + dev_dbg(hdq_data->dev, "wait interrupted");
> + mutex_unlock(&hdq_data->hdq_mutex);
> + return -EINTR;

Multiple-return-statements-per-function are a common source of
maintenance problems: locking errors, resource leaks. This is why
kernel code usually does the `goto out' way of avoiding this.

So.. please consider. In this case

ret = -EINTR;
goto out;

would fit nicely.

> + }
> +
> + spin_lock_irqsave(&hdq_data->hdq_spinlock, irqflags);
> + tmp_status = hdq_data->hdq_irqstatus;
> + spin_unlock_irqrestore(&hdq_data->hdq_spinlock, irqflags);
> + /* check irqstatus */
> + if (!(tmp_status & OMAP_HDQ_INT_STATUS_TIMEOUT)) {
> + dev_dbg(hdq_data->dev, "timeout waiting for TIMEOUT, %x",
> + tmp_status);
> + mutex_unlock(&hdq_data->hdq_mutex);
> + return -ETIMEDOUT;

Here too.

> + }
> + /*
> + * wait for both INIT and GO bits rerurn to zero.
> + * zero wait time expected for interrupt mode.
> + */
> + ret = hdq_wait_for_flag(hdq_data, OMAP_HDQ_CTRL_STATUS,
> + OMAP_HDQ_CTRL_STATUS_INITIALIZATION |
> + OMAP_HDQ_CTRL_STATUS_GO, OMAP_HDQ_FLAG_CLEAR,
> + &tmp_status);
> + if (ret)
> + dev_dbg(hdq_data->dev, "timeout waiting INIT&GO bits"
> + "return to zero, %x", tmp_status);
> +
> + mutex_unlock(&hdq_data->hdq_mutex);
> +
> + return ret;
> +}
> +
> +static int hdq_read_byte(struct hdq_data *hdq_data, u8 *val)
> +{
> + int ret;
> + u8 status;
> + unsigned long irqflags;
> + unsigned long timeout = jiffies + OMAP_HDQ_TIMEOUT;
> +
> + ret = mutex_lock_interruptible(&hdq_data->hdq_mutex);
> + if (ret < 0)
> + return -EINTR;
> +
> + if (!hdq_data->hdq_usecount) {
> + mutex_unlock(&hdq_data->hdq_mutex);
> + return -EINVAL;
> + }
> +
> + if (!(hdq_data->hdq_irqstatus & OMAP_HDQ_INT_STATUS_RXCOMPLETE)) {
> + hdq_reg_merge(hdq_data, OMAP_HDQ_CTRL_STATUS,
> + OMAP_HDQ_CTRL_STATUS_DIR | OMAP_HDQ_CTRL_STATUS_GO,
> + OMAP_HDQ_CTRL_STATUS_DIR | OMAP_HDQ_CTRL_STATUS_GO);
> + /*
> + * The RX comes immediately after TX. It
> + * triggers another interrupt before we
> + * sleep. So we have to wait for RXCOMPLETE bit.
> + */
> + while (!(hdq_data->hdq_irqstatus
> + & OMAP_HDQ_INT_STATUS_RXCOMPLETE)
> + && time_before(jiffies, timeout)) {
> + set_current_state(TASK_UNINTERRUPTIBLE);
> + schedule_timeout(1);

schedule_timeout_uninterruptible(1)

If we were to implement the presently-missing
wait_event_uninterruptible_timeout() (like
wait_event_interruptible_timeout), could we use it here?

> + }
> + hdq_reg_merge(hdq_data, OMAP_HDQ_CTRL_STATUS, 0,
> + OMAP_HDQ_CTRL_STATUS_DIR);
> + spin_lock_irqsave(&hdq_data->hdq_spinlock, irqflags);
> + status = hdq_data->hdq_irqstatus;
> + spin_unlock_irqrestore(&hdq_data->hdq_spinlock, irqflags);
> + /* check irqstatus */
> + if (!(status & OMAP_HDQ_INT_STATUS_RXCOMPLETE)) {
> + dev_dbg(hdq_data->dev, "timeout waiting for"
> + "RXCOMPLETE, %x", status);
> + mutex_unlock(&hdq_data->hdq_mutex);
> + return -ETIMEDOUT;
> + }
> + }
> + /* the data is ready. Read it in! */
> + *val = hdq_reg_in(hdq_data, OMAP_HDQ_RX_DATA);
> + mutex_unlock(&hdq_data->hdq_mutex);
> +
> + return 0;
> +
> +}
> +
>
> ...
>
> +static int __init omap_hdq_probe(struct platform_device *pdev)
> +{
> + struct hdq_data *hdq_data;
> + struct resource *res;
> + int ret, irq;
> + u8 rev;
> +
> + if (!pdev)
> + return -ENODEV;

Can this happen?

> + hdq_data = kmalloc(sizeof(*hdq_data), GFP_KERNEL);
> + if (!hdq_data) {
> + dev_dbg(&pdev->dev, "unable to allocate memory\n");
> + ret = -ENODEV;

-ENOMEM

> + goto err_kmalloc;
> + }
> +
> + hdq_data->dev = &pdev->dev;
> + platform_set_drvdata(pdev, hdq_data);
> +
> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + if (!res) {
> + dev_dbg(&pdev->dev, "unable to get resource\n");
> + ret = ENXIO;

bzzt, that should have been -ENXIO.

> + goto err_resource;
> + }
> +
> + hdq_data->hdq_base = ioremap(res->start, SZ_4K);
> + if (!hdq_data->hdq_base) {
> + dev_dbg(&pdev->dev, "ioremap failed\n");
> + ret = -EINVAL;
> + goto err_ioremap;
> + }
> +
> + /* get interface & functional clock objects */
> + hdq_data->hdq_ick = clk_get(&pdev->dev, "hdq_ick");
> + hdq_data->hdq_fck = clk_get(&pdev->dev, "hdq_fck");
> +
> + if (IS_ERR(hdq_data->hdq_ick) || IS_ERR(hdq_data->hdq_fck)) {
> + dev_dbg(&pdev->dev, "Can't get HDQ clock objects\n");
> + if (IS_ERR(hdq_data->hdq_ick)) {
> + ret = PTR_ERR(hdq_data->hdq_ick);
> + goto err_clk;
> + }
> + if (IS_ERR(hdq_data->hdq_fck)) {
> + ret = PTR_ERR(hdq_data->hdq_fck);
> + clk_put(hdq_data->hdq_ick);
> + goto err_clk;
> + }
> + }
> +
> + hdq_data->hdq_usecount = 0;
> + mutex_init(&hdq_data->hdq_mutex);
> +
> + if (clk_enable(hdq_data->hdq_ick)) {
> + dev_dbg(&pdev->dev, "Can not enable ick\n");
> + ret = -ENODEV;
> + goto err_ick;
> + }
> +
> + if (clk_enable(hdq_data->hdq_fck)) {
> + dev_dbg(&pdev->dev, "Can not enable fck\n");
> + ret = -ENODEV;
> + goto err_fck;
> + }

ooh, I like err_ick and err_fck a lot. They sound like akpm review
comments at the end of a long day.

> + rev = hdq_reg_in(hdq_data, OMAP_HDQ_REVISION);
> + dev_info(&pdev->dev, "OMAP HDQ Hardware Rev %c.%c. Driver in %s mode\n",
> + (rev >> 4) + '0', (rev & 0x0f) + '0', "Interrupt");
> +
> + spin_lock_init(&hdq_data->hdq_spinlock);
> + omap_hdq_break(hdq_data);
> +
> + irq = platform_get_irq(pdev, 0);
> + if (irq < 0) {
> + ret = -ENXIO;
> + goto err_irq;
> + }
> +
> + ret = request_irq(irq, hdq_isr, IRQF_DISABLED, "omap_hdq", hdq_data);
> + if (ret < 0) {
> + dev_dbg(&pdev->dev, "could not request irq\n");
> + goto err_irq;
> + }
> +
> + /* don't clock the HDQ until it is needed */
> + clk_disable(hdq_data->hdq_ick);
> + clk_disable(hdq_data->hdq_fck);
> +
> + omap_w1_master.data = hdq_data;
> +
> + ret = w1_add_master_device(&omap_w1_master);
> + if (ret) {
> + dev_dbg(&pdev->dev, "Failure in registering w1 master\n");
> + goto err_w1;
> + }
> +
> + return 0;
> +
> +err_w1:
> +err_irq:
> + clk_disable(hdq_data->hdq_fck);
> +
> +err_fck:
> + clk_disable(hdq_data->hdq_ick);
> +
> +err_ick:
> + clk_put(hdq_data->hdq_ick);
> + clk_put(hdq_data->hdq_fck);
> +
> +err_clk:
> + iounmap(hdq_data->hdq_base);
> +
> +err_ioremap:
> +err_resource:
> + platform_set_drvdata(pdev, NULL);
> + kfree(hdq_data);
> +
> +err_kmalloc:
> + return ret;
> +
> +}
> +
> +static int omap_hdq_remove(struct platform_device *pdev)
> +{
> + struct hdq_data *hdq_data = platform_get_drvdata(pdev);
> +
> + mutex_lock(&hdq_data->hdq_mutex);
> +
> + if (0 != hdq_data->hdq_usecount) {

Well. Lots of people dislike that trick. It's used to catch
accidental use of = where == was intended, but the compiler will warn
anyway. There's less point in using it for !=.

> + dev_dbg(&pdev->dev, "removed when use count is not zero\n");
> + return -EBUSY;
> + }
> +
> + mutex_unlock(&hdq_data->hdq_mutex);
> +
> + /* remove module dependency */
> + clk_put(hdq_data->hdq_ick);
> + clk_put(hdq_data->hdq_fck);
> + free_irq(INT_24XX_HDQ_IRQ, hdq_data);
> + platform_set_drvdata(pdev, NULL);
> + iounmap(hdq_data->hdq_base);
> + kfree(hdq_data);
> +
> + return 0;
> +}
> +
> +static int __init
> +omap_hdq_init(void)
> +{
> + return platform_driver_register(&omap_hdq_driver);
> +}
> +module_init(omap_hdq_init);
> +
> +static void __exit
> +omap_hdq_exit(void)
> +{
> + platform_driver_unregister(&omap_hdq_driver);
> +}
> +module_exit(omap_hdq_exit);
> +
> +module_param(w1_id, int, S_IRUSR);
> +MODULE_PARM_DESC(w1_id, "1-wire id for the slave detection");
> +
> +MODULE_AUTHOR("Texas Instruments");
> +MODULE_DESCRIPTION("HDQ driver Library");
> +MODULE_LICENSE("GPL");

The code looks pretty good.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/