Message-Id: <200805111050.55662.saschasommer@freenet.de>
Date:	Sun, 11 May 2008 10:50:55 +0200
From:	Sascha Sommer <saschasommer@...enet.de>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	linux-kernel@...r.kernel.org,
	sdricohcs-devel@...ts.sourceforge.net,
	Pierre Ossman <drzeus-list@...eus.cx>
Subject: Re: [PATCH] MMC/SD host driver for Ricoh Bay1Controllers

Hi,

Thanks for the review. I attached a new patch.

On Freitag, 2. Mai 2008, Andrew Morton wrote:
> On Sun, 27 Apr 2008 16:17:53 +0200
>
> Sascha Sommer <saschasommer@...enet.de> wrote:
> > Hi,
> >
> > attached you can find a driver for MMC/SD card controllers that have been
> > integrated into some Notebooks with Ricoh Co Ltd RL5c476 II Cardbus
> > bridges.
>
> Thanks.
>
> Please do copy Pierre on mmc patches.
>
> > This includes the Samsung P35 notebook, the Dell X300 and some notebooks
> > from the Asus M6N series.
> > Whenever a MMC or SD card is insterted in the cardslot of one of these
> > notebooks the cardbus bridge will announce a
> > virtual Ricoh Bay1Controller PCMCIA card.
> >
> > Since I submitted the first version of this driver to lkml last year many
> > changes have been made. The driver can now read and write to MMC and SD
> > cards and the reading speed for SD cards should be close to the windows
> > driver now. Many thanks to all the people that tested the driver and
> > helped me to fix some of the remaining bugs. I think it is now time to
> > find out what needs to be fixed before the driver can go into kernel.
> >
> > While the driver is working fine for me there are still some not so nice
> > parts in it.
> > - the registers are on the cardbus bridge that is already claimed by
> > yenta socket therefore the driver has to use pci_get_device
> > - there is a lot of register polling because the reader does not support
> > irqs - some mmc commands require extra hacks. One of those commands is
> > the SD_APP_SEND_SCR command (a user reported that it is required for his
> > SDHC card). Like the read and write commands this command
> >   requires a block read to read the scr but this does not seem to work.
> >   I do not have any documentation for this device and the windows driver
> > does not support such commands so I cannot find the
> >   needed info in its logfiles.
> > ...
> >
> > I'm thankfull for any kind of feedback.
> >
> > Signed-off-by: Sascha Sommer <saschasommer@...enet.de>
>
> The patch has a large number of trivial layout problems, most of which we
> wold ordinarily prefer be fixed.  Please pass the diff through
> scripts/checkpatch.pl and consider the output?
>

Done.

> > ...
> >
> > +#define DRIVER_NAME "sdricoh_cs"
> > +#define DRIVER_VERSION "0.1.4"
>
> Please consider removing the driver version.  It becomes useless once the
> code is merge into the kernel - it cannot be used to determine exactly
> which version of the driver your users are running.  Distros may patch the
> driver, other kernel develoeprs may patch it and forget to increment the
> version number, etc.
>
> To regenerate a user's driver source the only reliable approach is to find
> out their kernel version and then go find the source to that kernel.

Removed.

>
> > +static unsigned int debug = 0;
> > +static unsigned int switchlocked = 0;
> > +
> > +/* #define DEBUG */
> > +
> > +/* debug macros */
> > +
> > +#ifdef DEBUG
> > +#define REGDBG(fmt, arg...) do {\
> > +        if (debug > 1) \
> > +                printk(KERN_INFO "sdricoh_cs: "fmt, \
>
> Use DRIVER_NAME here
>
> > +                         ##arg); } while (0)
> > +#else
> > +#define REGDBG(fmt, arg...)
> > +#endif
> > +
> > +#define DBG(fmt, arg...) do {\
> > +        if (debug > 0) \
> > +                printk(KERN_INFO DRIVER_NAME ": "fmt, \
> > +                         ##arg); } while (0)
> > +
> > +#define ERR(fmt, arg...) do {\
> > +                printk(KERN_INFO DRIVER_NAME ": "fmt, \
> > +                         ##arg); } while (0)
> > +
> > +#define INFO(fmt, arg...) do {\
> > +                printk(KERN_INFO DRIVER_NAME ": "fmt, \
> > +                         ##arg); } while (0)
>
> It'd be nice to use common debug macros rather than home-made ones.
> include/linux/kernel.h has one.  But don't bust a gut over it - thousands
> of drivers do this :(
>

Changed.

> > ...
> >
> > +static int sdricoh_query_status(struct sdricoh_host *host,unsigned int
> > wanted, +				unsigned int timeout){
> > +	unsigned int loop;
> > +	unsigned int status = 0;
> > +	for (loop = 0; loop < timeout; loop++) {
> > +		status = sdricoh_readl(host, R21C_STATUS);
> > +		sdricoh_writel(host, R2E4_STATUS_RESP, status);
> > +		if (status & wanted)
> > +			break;
> > +	}
> > +
> > +        if (loop == timeout) {
> > +                ERR("query_status: timeout waiting for data\n");
> > +                return -ETIMEDOUT;
> > +        }
>
> Something went wrong with the indenting there, although I expect checkpatch
> will notice it.
>

Fixed.

> > +	/* do not do this check in the loop as some commands fail otherwise */
> > +	if(status & 0x7F0000){
> > +		ERR("waiting for status bit %x failed\n",wanted);
> > +		return -EINVAL;
> > +	}
> > +	return 0;
> > +
> > +}
> > +
> > +
> > +
> > +
> > +
>
> ?
>

Fixed.

> > +static int sdricoh_mmc_cmd(struct sdricoh_host *host, unsigned char
> > opcode, +			   unsigned int arg)
> > +{
> > +	unsigned int status;
> > +	int result = 0;
> > +	unsigned int loop = 0;
> > +	/* reset status reg? */
> > +	sdricoh_writel(host, R21C_STATUS, 0x18);
> > +	/* fill parameters */
> > +	sdricoh_writel(host, R204_CMD_ARG, arg);
> > +	sdricoh_writel(host, R200_CMD, (0x10000 << 8) | opcode);
> > +	/* wait for command completion */
> > +	if (opcode) {
> > +		for (loop = 0; loop < CMD_TIMEOUT; loop++) {
> > +			status = sdricoh_readl(host, R21C_STATUS);
> > +			sdricoh_writel(host, R2E4_STATUS_RESP, status);
> > +			if (status  & STATUS_CMD_FINISHED)
> > +				break;
> > +		}
> > +		if (loop == CMD_TIMEOUT || status & STATUS_CMD_TIMEOUT)
>
> The test of STATUS_CMD_TIMEOUT here is redundant.
>
> Perhaps you meant to break out of the loop if STATUS_CMD_TIMEOUT becomes
> set.
>

It is a bit strange. The bits sometimes do not get reset so doing the check 
inside the loop leads to wrong negatives. I added a comment.


> > +			result = -ETIMEDOUT;
> > +
> > +	}
> > +	DBG("mmc_cmd opcode=%i arg=0x%x => %i (queries=%i)\n",
> > +	    opcode, arg, result, loop);
> > +
> > +	if(result == 0){
> > +		/* EXT_CSD are filtered so this should be save */
> > +                if(opcode == SD_SEND_IF_COND){
> > +                        if(host->mode != MODE_SDHC){
> > +                                INFO("switching to SDHC mode\n");
> > +                                host->mode = MODE_SDHC;
> > +                        }
> > +                }
> > +
> > +		/* switch to SD mode if APP_CMDs are supported */
> > +		if(opcode == MMC_APP_CMD){
> > +			if(host->mode == MODE_MMC){
> > +				INFO("switching to SD mode\n");
> > +				host->mode = MODE_SD;
> > +			}
> > +		}
> > +	}
> > +
> > +	return result;
> > +
> > +}
> >
> > ...
> >
> > +static int sdricoh_blockio(struct sdricoh_host *host, int read,
> > +				unsigned int* buf)
> > +{
> > +	int i;
> > +	/* wait until the data is available */
> > +	if(read){
> > +		if(sdricoh_query_status(host,STATUS_READY_TO_READ,
> > +						TRANSFER_TIMEOUT))
> > +			return 0;
> > +		sdricoh_writel(host, R21C_STATUS, 0x18);
> > +		/* read data */
> > +		for (i = 0; i < 512 / 4; i++) {
> > +			buf[i] = sdricoh_readl(host, R230_DATA);
> > +		}
>
> Will this code work correctly on big-endian machines?
>

Most likely not and there might be other big-endian related problems in the 
driver. I don't know of a case where this hardware can be found on big 
endian.

> > +	}else{
> > +		if(sdricoh_query_status(host,STATUS_READY_TO_WRITE,
> > +						TRANSFER_TIMEOUT))
> > +			return 0;
> > +		sdricoh_writel(host, R21C_STATUS, 0x18);
> > +		/* write data */
> > +		for (i = 0; i < 512 / 4; i++) {
> > +			sdricoh_writel(host, R230_DATA, buf[i]);
> > +		}
> > +	}
> > +
> > +	return 512;
> > +}
> > +
> > +static int sdricoh_busy(struct sdricoh_host* host){
>
> the brace goes on the next line, please.
>

Fixed.

> > +	unsigned int status;
> > +	int i;
> > +	/* wait until the tranfer is finished */
> > +	for (i = 0; i < BUSY_TIMEOUT; i++) {
> > +		status = sdricoh_readl(host, R21C_STATUS);
> > +		sdricoh_writel(host, R2E4_STATUS_RESP, status);
> > +		if (!(status & STATUS_BUSY))
> > +		break;
>
> missing tab
>

Fixed.

> > +	}
> > +	if(status & 0x7F0000)
> > +		return -EINVAL;
> > +	if(i == BUSY_TIMEOUT)
> > +		return -ETIMEDOUT;
> > +	return 0;
> > +}
> > +
> > +
> > +static void sdricoh_request(struct mmc_host *mmc, struct mmc_request
> > *mrq) +{
> > +	struct sdricoh_host *host = mmc_priv(mmc);
> > +	struct mmc_command *cmd = mrq->cmd;
> > +	struct mmc_data *data = cmd->data;
> > +	int i;
> > +
> > +	DBG("=============================\n");
> > +	DBG("sdricoh_request opcode=%i\n", cmd->opcode);
> > +
> > +	sdricoh_writel(host, R21C_STATUS, 0x18);
> > +
> > +	/* we cannot handle all commands that require a block transfer
> > +	   therefore do some ugly special handling here
> > +	*/
> > +	if(cmd->data){
> > +		switch(cmd->opcode){
> > +			/* working commands */
> > +			case MMC_READ_SINGLE_BLOCK:
> > +			case MMC_READ_MULTIPLE_BLOCK:
> > +			case MMC_WRITE_BLOCK:
> > +				break;
> > +	               	case SD_APP_SEND_SCR: /* required for SDHC */
> > +				cmd->error = sdricoh_mmc_cmd(host,
> > +						       cmd->opcode,cmd->arg);
> > +                       		mmc_request_done(mmc, mrq);
> > +                       		return;
> > +			default:
> > +				DBG("unsupported command %i\n",cmd->opcode);
> > +				cmd->error = -EINVAL;
> > +				mmc_request_done(mmc, mrq);
> > +				return;
> > +		}
>
> we normally will indent the body of the switch statement one tabstop less
> than this.
>

Fixed.

> > +	}
> > +
> > +
> > +	/* read/write commands seem to require this */
> > +	if (data) {
> > +		if((cmd->error = sdricoh_busy(host)))
> > +			ERR("sdricoh_request: unable to prepare transfer %x\n",
> > +				cmd->error);
> > +		sdricoh_writel(host, R208_DATAIO, 0);
> > +	}
> > +
> > +
> > +	cmd->error = sdricoh_mmc_cmd(host, cmd->opcode, cmd->arg);
> > +
> > +	/* read response buffer */
> > +	if (cmd->flags & MMC_RSP_PRESENT) {
> > +		if (cmd->flags & MMC_RSP_136) {
> > +			/* CRC is stripped so we need to do some shifting. */
> > +			for (i = 0; i < 4; i++) {
> > +				cmd->resp[i] =
> > +				    sdricoh_readl(host,
> > +						  R20C_RESP + (3 - i) * 4) << 8;
> > +				if (i != 3)
> > +					cmd->resp[i] |=
> > +					    sdricoh_readb(host, R20C_RESP +
> > +							  (3 - i) *4 - 1);
> > +			}
> > +			DBG("resp[0]=0x%x\n", cmd->resp[0]);
> > +			DBG("resp[1]=0x%x\n", cmd->resp[1]);
> > +			DBG("resp[2]=0x%x\n", cmd->resp[2]);
> > +			DBG("resp[2]=0x%x\n", cmd->resp[3]);
> > +		} else {
> > +			cmd->resp[0] = sdricoh_readl(host, R20C_RESP);
> > +			DBG("resp[0]=0x%x\n", cmd->resp[0]);
> > +		}
> > +	}
> > +
> > +	/* yet another workaround */
> > +	/* without the extra command SD cards do not work at all */
> > +	if (cmd->opcode == MMC_SELECT_CARD) {
> > +		if(host->mode != MODE_MMC){
> > +			sdricoh_mmc_cmd(host, MMC_APP_CMD, cmd->arg);
> > +			sdricoh_mmc_cmd(host, 0x46, 0x02);
> > +		}else{
> > +			sdricoh_writel(host, R228_POWER, 0xc0e0);
> > +			sdricoh_writel(host, R224_MODE, 0x2000301);
> > +		}
> > +	}
> > +
> > +	/* transfer data */
> > +	if (data && cmd->error == 0) {
> > +		DBG("transfer: blksz %i blocks %i sg_len %i sg length %i\n",
> > +                    data->blksz, data->blocks, data->sg_len,
> > data->sg->length); +
> > +		/* enter data reading mode */
> > +		sdricoh_writel(host, R21C_STATUS, 0x837f031e);
> > +		for (i = 0; i < data->blocks; i++) {
> > +			unsigned int *buf;
> > +			struct page* page;
> > +			size_t xfered;
> > +			page = sg_page(data->sg);
> > +
> > +        		buf = kmap(page) + data->sg->offset + (512 * i);
> > +			xfered =
> > +			  sdricoh_blockio(host,data->flags & MMC_DATA_READ,buf);
> > +			kunmap(page);
>
> hm.  I don't really know how the kernel gets down into here, but I wonder
> if this driver (and, I bet, lots of similar ones) should be doing
> flush_dcache_page() after modifying the page.  If this page can be file
> pagecache or user memory then "yes".  Unless it's done elsewhere for us.
>

Some other mmc host drivers seem to do it so I added it, too.

> > +			if(!xfered){
> > +				ERR("sdricoh_request: block transfer failed\n");
> > +				cmd->error = -EINVAL;
> > +				break;
> > +			}else
> > +				data->bytes_xfered += xfered;
> > +		}
> > +
> > +		sdricoh_writel(host, R208_DATAIO, 1);
> > +
> > +		if(sdricoh_query_status(host,STATUS_TRANSFER_FINISHED,
> > +					TRANSFER_TIMEOUT)){
> > +			ERR("sdricoh_request: transfer end error\n");
> > +			cmd->error = -EINVAL;
> > +		}
> > +
> > +		if(!cmd->error && (cmd->error = sdricoh_busy(host)))
> > +			ERR("sdricoh_request: transfer not finished %x\n",
> > +				cmd->error);
> > +
> > +	}
> > +
> > +	mmc_request_done(mmc, mrq);
> > +	DBG("=============================\n");
> > +}
> >
> > ...
> >
> > +/* initialize the control and register it to the mmc framework */
> > +static int sdricoh_init_mmc(struct pci_dev *pci_dev,
> > +			    struct pcmcia_device *pcmcia_dev)
> > +{
> > +	int result = 0;
> > +	void __iomem *iobase = NULL;
> > +	struct mmc_host *mmc = NULL;
> > +	struct sdricoh_host *host = NULL;
> > +	/* map iomem */
> > +	if (pci_resource_len(pci_dev, SDRICOH_PCI_REGION) !=
> > +	    SDRICOH_PCI_REGION_SIZE) {
> > +		DBG("unexpected pci resource len\n");
> > +		return -ENODEV;
> > +	}
> > +	iobase =
> > +	    pci_iomap(pci_dev, SDRICOH_PCI_REGION, SDRICOH_PCI_REGION_SIZE);
> > +	if (!iobase) {
> > +		ERR("unable to map iobase\n");
> > +		return -ENODEV;
> > +	}
> > +	/* check version? */
> > +	if (readl(iobase + R104_VERSION) != 0x4000) {
> > +		DBG("no supported mmc controller found\n");
> > +		result = -ENODEV;
> > +		goto err;
> > +	}
> > +	/* allocate privdata */
> > +	mmc = pcmcia_dev->priv =
> > +	    mmc_alloc_host(sizeof(struct sdricoh_host), &pcmcia_dev->dev);
> > +	if (!mmc) {
> > +		ERR("mmc_alloc_host failed\n");
> > +		result = -ENOMEM;
> > +		goto err;
> > +	}
> > +	host = mmc_priv(mmc);
> > +
> > +	host->iobase = iobase;
> > +
> > +	mmc->ops = &sdricoh_ops;
> > +
> > +	/* FIXME: frequency and voltage handling is done by the controller
> > +	 */
> > +	mmc->f_min = 450000;
> > +	mmc->f_max = 24000000;
> > +	mmc->ocr_avail = MMC_VDD_32_33 | MMC_VDD_33_34;
> > +
> > +	mmc->max_seg_size = 1024 * 512;
> > +
> > +	/* reset the controler */
> > +	if (sdricoh_reset(host)) {
> > +		DBG("could not reset\n");
> > +		result = -EIO;
> > +		goto err;
>
> forgot to free pcmcia_dev->priv here?
>

No. It is the same as mmc and that gets freed later.

> > +	}
> > +
> > +	result = mmc_add_host(mmc);
> > +
> > +	if (!result) {
> > +		DBG("mmc host registered\n");
> > +		return 0;
> > +	}
> > +
> > +      err:
> > +	if (iobase)
> > +		iounmap(iobase);
> > +	if (mmc)
> > +		mmc_free_host(mmc);
> > +
> > +	return result;
> > +}
> >
> > ...
> >
> > +static void sdricoh_pcmcia_detach(struct pcmcia_device *link)
> > +{
> > +	struct mmc_host *mmc = link->priv;
> > +
> > +	DBG("detach\n");
> > +
> > +	flush_scheduled_work();
>
> what work are we flushing here?
>

Removed.

> > +	/* remove mmc host */
> > +	if (mmc) {
> > +		struct sdricoh_host *host = mmc_priv(mmc);
> > +		mmc_remove_host(mmc);
> > +		pci_iounmap(host->pci_dev, host->iobase);
> > +		pci_dev_put(host->pci_dev);
> > +		mmc_free_host(mmc);
> > +	}
> > +	pcmcia_disable_device(link);
> > +
> > +}
> >
> > ...
> >
> > --- drivers/mmc/host.org/Kconfig	2008-04-27 14:35:31.000000000 +0200
> > +++ drivers/mmc/host/Kconfig	2008-04-27 15:15:29.000000000 +0200
> > @@ -130,3 +130,13 @@
> >
> >  	  If unsure, or if your system has no SPI master driver, say N.
> >
> > +config MMC_SDRICOH_CS
> > +	tristate "MMC/SD driver for Ricoh Bay1Controllers (EXPERIMENTAL)"
> > +	depends on EXPERIMENTAL && MMC && PCI && PCMCIA && YENTA
>
> Is the dependency on YENTA correct?
>

Yes. At least noone who wants to enable this driver will forget the yenta 
driver then.

Regards

Sascha

View attachment "sdricoh_cs_try2.patch" of type "text/x-diff" (18365 bytes)

Powered by blists - more mailing lists