From: Jeff Garzik <jgarzik@pobox.com>
To: Tejun Heo <htejun@gmail.com>
Cc: linux-ide@vger.kernel.org, bob@evoria.net,
nabble@theanthonys.net, alan@lxorguk.ukuu.org.uk,
matthias@urlichs.de, romieu@fr.zoreil.com,
carlosmarcelomartinez@gmail.com
Subject: Re: [PATCH 2/2] sata_inic162x: driver for initio 162x SATA controllers, take 2
Date: Wed, 20 Dec 2006 14:56:16 -0500 [thread overview]
Message-ID: <45899560.8090401@pobox.com> (raw)
In-Reply-To: <20061220062546.GA4823@htj.dyndns.org>
Tejun Heo wrote:
> Driver for Initio 162x SATA controllers. ATA r/w, ATAPI r, hotplug
> and suspend/resume work. ATAPI w (recording, that is) broken. Feel
> free to fix it, but be warned, this controller is weird.
>
> Signed-off-by: Tejun Heo <htejun@gmail.com>
Minor issues only... See inline comments.
> drivers/ata/Kconfig | 6
> drivers/ata/Makefile | 1
> drivers/ata/sata_inic162x.c | 802 ++++++++++++++++++++++++++++++++++++++++++++
> 3 files changed, 809 insertions(+)
>
> Index: work/drivers/ata/Kconfig
> ===================================================================
> --- work.orig/drivers/ata/Kconfig
> +++ work/drivers/ata/Kconfig
> @@ -143,6 +143,12 @@ config SATA_VITESSE
>
> If unsure, say N.
>
> +config SATA_INIC162X
> + tristate "Initio 162x SATA support"
> + depends on PCI
> + help
> + This option enables support for Initio 162x Serial ATA.
> +
> config SATA_INTEL_COMBINED
> bool
> depends on IDE=y && !BLK_DEV_IDE_SATA && (SATA_AHCI || ATA_PIIX)
> Index: work/drivers/ata/Makefile
> ===================================================================
> --- work.orig/drivers/ata/Makefile
> +++ work/drivers/ata/Makefile
> @@ -15,6 +15,7 @@ obj-$(CONFIG_SATA_SX4) += sata_sx4.o
> obj-$(CONFIG_SATA_NV) += sata_nv.o
> obj-$(CONFIG_SATA_ULI) += sata_uli.o
> obj-$(CONFIG_SATA_MV) += sata_mv.o
> +obj-$(CONFIG_SATA_INIC162X) += sata_inic162x.o
> obj-$(CONFIG_PDC_ADMA) += pdc_adma.o
>
> obj-$(CONFIG_PATA_ALI) += pata_ali.o
> Index: work/drivers/ata/sata_inic162x.c
> ===================================================================
> --- /dev/null
> +++ work/drivers/ata/sata_inic162x.c
> @@ -0,0 +1,802 @@
> +/*
> + * sata_inic162x.c - Driver for Initio 162x SATA controllers
> + *
> + * Copyright 2006 SUSE Linux Products GmbH
> + * Copyright 2006 Tejun Heo <teheo@novell.com>
> + *
> + * This file is released under GPL v2.
> + *
> + * This controller is eccentric and easily locks up if something isn't
> + * right. Documentation is available at initio's website but it only
> + * documents registers (not programming model).
> + *
> + * - ATA disks work.
> + * - Hotplug works.
> + * - ATAPI read works but burning doesn't. This thing is really
> + * peculiar about ATAPI and I couldn't figure out how ATAPI PIO and
> + * ATAPI DMA WRITE should be programmed. If you've got a clue, be
> + * my guest.
> + * - Both STR and STD work.
Do everyday users get a sane error, or something bad like a lockup, when
trying to ATAPI write?
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/pci.h>
> +#include <scsi/scsi_host.h>
> +#include <linux/libata.h>
> +#include <linux/blkdev.h>
> +#include <scsi/scsi_device.h>
> +
> +#define DRV_NAME "sata_inic162x"
> +#define DRV_VERSION "0.1"
> +
> +enum {
> + MMIO_BAR = 5,
> +
> + NR_PORTS = 2,
> +
> + HOST_CTL = 0x7c,
> + HOST_STAT = 0x7e,
> + HOST_IRQ_STAT = 0xbc,
> + HOST_IRQ_MASK = 0xbe,
> +
> + PORT_SIZE = 0x40,
> +
> + /* registers for ATA TF operation */
> + PORT_TF = 0x00,
> + PORT_ALT_STAT = 0x08,
> + PORT_IRQ_STAT = 0x09,
> + PORT_IRQ_MASK = 0x0a,
> + PORT_PRD_CTL = 0x0b,
> + PORT_PRD_ADDR = 0x0c,
> + PORT_PRD_XFERLEN = 0x10,
> +
> + /* IDMA register */
> + PORT_IDMA_CTL = 0x14,
> +
> + PORT_SCR = 0x20,
> +
> + /* HOST_CTL bits */
> + HCTL_IRQOFF = (1 << 8), /* global IRQ off */
> + HCTL_PWRDWN = (1 << 13), /* power down PHYs */
> + HCTL_SOFTRST = (1 << 13), /* global reset (no phy reset) */
> + HCTL_RPGSEL = (1 << 15), /* register page select */
> +
> + HCTL_KNOWN_BITS = HCTL_IRQOFF | HCTL_PWRDWN | HCTL_SOFTRST |
> + HCTL_RPGSEL,
> +
> + /* HOST_IRQ_(STAT|MASK) bits */
> + HIRQ_PORT0 = (1 << 0),
> + HIRQ_PORT1 = (1 << 1),
> + HIRQ_SOFT = (1 << 14),
> + HIRQ_GLOBAL = (1 << 15), /* STAT only */
> +
> + /* PORT_IRQ_(STAT|MASK) bits */
> + PIRQ_OFFLINE = (1 << 0), /* device unplugged */
> + PIRQ_ONLINE = (1 << 1), /* device plugged */
> + PIRQ_COMPLETE = (1 << 2), /* completion interrupt */
> + PIRQ_FATAL = (1 << 3), /* fatal error */
> + PIRQ_ATA = (1 << 4), /* ATA interrupt */
> + PIRQ_REPLY = (1 << 5), /* reply FIFO not empty */
> + PIRQ_PENDING = (1 << 7), /* port IRQ pending (STAT only) */
> +
> + PIRQ_ERR = PIRQ_OFFLINE | PIRQ_ONLINE | PIRQ_FATAL,
> +
> + PIRQ_MASK_DMA_READ = PIRQ_REPLY | PIRQ_ATA,
> + PIRQ_MASK_OTHER = PIRQ_REPLY | PIRQ_COMPLETE,
> + PIRQ_MASK_FREEZE = 0xff,
> +
> + /* PORT_PRD_CTL bits */
> + PRD_CTL_START = (1 << 0),
> + PRD_CTL_WR = (1 << 3),
> + PRD_CTL_DMAEN = (1 << 7), /* DMA enable */
> +
> + /* PORT_IDMA_CTL bits */
> + IDMA_CTL_RST_ATA = (1 << 2), /* hardreset ATA bus */
> + IDMA_CTL_RST_IDMA = (1 << 5), /* reset IDMA machinary */
> + IDMA_CTL_GO = (1 << 7), /* IDMA mode go */
> + IDMA_CTL_ATA_NIEN = (1 << 8), /* ATA IRQ disable */
> +};
> +
> +struct inic_host_priv {
> + u16 cached_hctl;
> +};
> +
> +struct inic_port_priv {
> + u8 dfl_prdctl, cached_prdctl;
> + u8 cached_pirq_mask;
apply standard struct style:
1) one struct member per line
2) indent between type and member name
> +static int inic_slave_config(struct scsi_device *sdev)
> +{
> + /* This controller is braindamaged. dma_boundary is 0xffff
> + * like others but it will lock up the whole machine HARD if
> + * 65536 byte PRD entry is fed. Reduce maximum segment size.
> + */
> + blk_queue_max_segment_size(sdev->request_queue, 65536 - 512);
> +
> + return ata_scsi_slave_config(sdev);
> +}
> +
> +static struct scsi_host_template inic_sht = {
> + .module = THIS_MODULE,
> + .name = DRV_NAME,
> + .ioctl = ata_scsi_ioctl,
> + .queuecommand = ata_scsi_queuecmd,
> + .can_queue = ATA_DEF_QUEUE,
> + .this_id = ATA_SHT_THIS_ID,
> + .sg_tablesize = LIBATA_MAX_PRD,
> + .cmd_per_lun = ATA_SHT_CMD_PER_LUN,
> + .emulated = ATA_SHT_EMULATED,
> + .use_clustering = ATA_SHT_USE_CLUSTERING,
> + .proc_name = DRV_NAME,
> + .dma_boundary = ATA_DMA_BOUNDARY,
> + .slave_configure = inic_slave_config,
> + .slave_destroy = ata_scsi_slave_destroy,
> + .bios_param = ata_std_bios_param,
> + .suspend = ata_scsi_device_suspend,
> + .resume = ata_scsi_device_resume,
> +};
> +
> +static const int scr_map[] = {
> + [SCR_STATUS] = 0,
> + [SCR_ERROR] = 1,
> + [SCR_CONTROL] = 2,
> +};
> +
> +static void __iomem * get_port_base(struct ata_port *ap)
> +{
> + return ap->host->mmio_base + ap->port_no * PORT_SIZE;
> +}
> +
> +static void __set_pirq_mask(struct ata_port *ap, u8 mask)
> +{
> + void __iomem *port_base = get_port_base(ap);
> + struct inic_port_priv *pp = ap->private_data;
> +
> + writeb(mask, port_base + PORT_IRQ_MASK);
> + pp->cached_pirq_mask = mask;
> +}
> +
> +static void set_pirq_mask(struct ata_port *ap, u8 mask)
> +{
> + struct inic_port_priv *pp = ap->private_data;
> +
> + if (pp->cached_pirq_mask != mask)
> + __set_pirq_mask(ap, mask);
> +}
You should either flush here, or in the one case you need it, ->thaw
> +static void reset_port(void __iomem *port_base)
> +{
> + void __iomem *idma_ctl = port_base + PORT_IDMA_CTL;
> + u16 ctl;
> +
> + ctl = readw(idma_ctl);
> + ctl &= ~(IDMA_CTL_RST_IDMA | IDMA_CTL_ATA_NIEN | IDMA_CTL_GO);
> +
> + /* mask IRQ and assert reset */
> + writew(ctl | IDMA_CTL_RST_IDMA | IDMA_CTL_ATA_NIEN, idma_ctl);
> + readw(idma_ctl); /* flush */
> +
> + /* give it some time */
> + msleep(1);
> +
> + /* release reset */
> + writew(ctl | IDMA_CTL_ATA_NIEN, idma_ctl);
> +
> + /* clear irq */
> + writeb(0xff, port_base + PORT_IRQ_STAT);
> +
> + /* reenable ATA IRQ, turn off IDMA mode */
> + writew(ctl, idma_ctl);
> +}
> +
> +static u32 inic_scr_read(struct ata_port *ap, unsigned sc_reg)
> +{
> + void __iomem *scr_addr = (void __iomem *)ap->ioaddr.scr_addr;
> + if (sc_reg < ARRAY_SIZE(scr_map)) {
> + void __iomem *addr;
> + u32 val;
> +
> + addr = scr_addr + scr_map[sc_reg] * 4;
> + val = readl(scr_addr + scr_map[sc_reg] * 4);
> +
> + /* this controller has stuck DIAG.N, ignore it */
> + if (sc_reg == SCR_ERROR)
> + val &= ~SERR_PHYRDY_CHG;
> + return val;
> + }
> + return 0xffffffffU;
> +}
style: the main body of code should not be indented.
more appropriate:
if (unlikely(sc_reg >= ARRAY_SIZE(scr_map)))
return 0xffffffffU;
...
Or perhaps audit the code and ensure that the core never calls with an
SCR index greater than 2 (SCR_CONTROL), and then add
BUG_ON(sc_ref > SCR_CONTROL);
> +static void inic_scr_write(struct ata_port *ap, unsigned sc_reg, u32 val)
> +{
> + void __iomem *scr_addr = (void __iomem *)ap->ioaddr.scr_addr;
> + if (sc_reg < ARRAY_SIZE(scr_map)) {
> + void __iomem *addr;
> + addr = scr_addr + scr_map[sc_reg] * 4;
> + writel(val, scr_addr + scr_map[sc_reg] * 4);
> + }
> +}
ditto
> +/*
> + * In TF mode, inic162x is very similar to SFF device. TF registers
> + * function the same. DMA engine behaves similary using the same PRD
> + * format as BMDMA but different command register, interrupt and event
> + * notification methods are used. The following inic_bmdma_*()
> + * functions do the impedance matching.
> + */
> +static void inic_bmdma_setup(struct ata_queued_cmd *qc)
> +{
> + struct ata_port *ap = qc->ap;
> + struct inic_port_priv *pp = ap->private_data;
> + void __iomem *port_base = get_port_base(ap);
> + int rw = qc->tf.flags & ATA_TFLAG_WRITE;
> +
> + /* make sure device sees PRD table writes */
> + wmb();
> +
> + /* load transfer length */
> + writel(qc->nbytes, port_base + PORT_PRD_XFERLEN);
> +
> + /* turn on DMA and specify data direction */
> + pp->cached_prdctl = pp->dfl_prdctl | PRD_CTL_DMAEN;
> + if (!rw)
> + pp->cached_prdctl |= PRD_CTL_WR;
> + writeb(pp->cached_prdctl, port_base + PORT_PRD_CTL);
> +
> + /* issue r/w command */
> + ap->ops->exec_command(ap, &qc->tf);
> +}
> +
> +static void inic_bmdma_start(struct ata_queued_cmd *qc)
> +{
> + struct ata_port *ap = qc->ap;
> + struct inic_port_priv *pp = ap->private_data;
> + void __iomem *port_base = get_port_base(ap);
> +
> + /* start host DMA transaction */
> + pp->cached_prdctl |= PRD_CTL_START;
> + writeb(pp->cached_prdctl, port_base + PORT_PRD_CTL);
> +}
> +
> +static void inic_bmdma_stop(struct ata_queued_cmd *qc)
> +{
> + struct ata_port *ap = qc->ap;
> + struct inic_port_priv *pp = ap->private_data;
> + void __iomem *port_base = get_port_base(ap);
> +
> + /* stop DMA engine */
> + writeb(pp->dfl_prdctl, port_base + PORT_PRD_CTL);
> +}
> +
> +static u8 inic_bmdma_status(struct ata_port *ap)
> +{
> + /* event is already verified by the interrupt handler */
> + return ATA_DMA_INTR;
> +}
> +
> +static void inic_irq_clear(struct ata_port *ap)
> +{
> + /* noop */
> +}
> +
> +static void inic_host_intr(struct ata_port *ap)
> +{
> + void __iomem *port_base = get_port_base(ap);
> + struct ata_eh_info *ehi = &ap->eh_info;
> + u8 irq_stat;
> +
> + /* fetch and clear irq */
> + irq_stat = readb(port_base + PORT_IRQ_STAT);
> + writeb(irq_stat, port_base + PORT_IRQ_STAT);
> +
> + if (likely(!(irq_stat & PIRQ_ERR))) {
> + struct ata_queued_cmd *qc = ata_qc_from_tag(ap, ap->active_tag);
> + u8 stat;
> +
> + if (unlikely(!qc || (qc->tf.flags & ATA_TFLAG_POLLING))) {
> + ata_chk_status(ap); /* clear ATA interrupt */
> + return;
> + }
> +
> + if (likely(ata_host_intr(ap, qc)))
> + return;
> +
> + stat = ata_chk_status(ap); /* clear ATA interrupt */
> + ata_port_printk(ap, KERN_WARNING, "unhandled interrupt, "
> + "stat=0x%x irq_stat=0x%x\n", stat, irq_stat);
> + return;
> + }
> +
> + /* error */
> + ata_ehi_push_desc(ehi, "irq_stat=0x%x", irq_stat);
> +
> + if (irq_stat & (PIRQ_OFFLINE | PIRQ_ONLINE)) {
> + ata_ehi_hotplugged(ehi);
> + ata_port_freeze(ap);
> + } else
> + ata_port_abort(ap);
> +}
> +
> +static irqreturn_t inic_interrupt(int irq, void *dev_instance)
> +{
> + struct ata_host *host = dev_instance;
> + void __iomem *mmio_base = host->mmio_base;
> + u16 host_irq_stat;
> + int i, handled = 0;;
> +
> + host_irq_stat = readw(mmio_base + HOST_IRQ_STAT);
> +
> + if (unlikely(!(host_irq_stat & HIRQ_GLOBAL)))
> + goto out;
> +
> + spin_lock(&host->lock);
> +
> + for (i = 0; i < NR_PORTS; i++) {
> + struct ata_port *ap = host->ports[i];
> +
> + if (!(host_irq_stat & (HIRQ_PORT0 << i)))
> + continue;
> +
> + if (likely(ap && !(ap->flags & ATA_FLAG_DISABLED))) {
> + inic_host_intr(ap);
> + handled++;
> + } else {
> + if (ata_ratelimit())
> + dev_printk(KERN_ERR, host->dev, "interrupt "
> + "from disabled port %d (0x%x)\n",
> + i, host_irq_stat);
> + }
> + }
> +
> + spin_unlock(&host->lock);
> +
> + out:
> + return IRQ_RETVAL(handled);
> +}
> +
> +static unsigned int inic_qc_issue(struct ata_queued_cmd *qc)
> +{
> + struct ata_port *ap = qc->ap;
> +
> + /* ATA IRQ doesn't wait for DMA transfer completion and vice
> + * versa. Mask IRQ selectively to detect command completion.
> + * Without it, ATA DMA read command can cause data corruption.
> + *
> + * Something similar might be needed for ATAPI writes. I
> + * tried a lot of combinations but couldn't find the solution.
> + */
> + if (qc->tf.protocol == ATA_PROT_DMA &&
> + !(qc->tf.flags & ATA_TFLAG_WRITE))
> + set_pirq_mask(ap, PIRQ_MASK_DMA_READ);
> + else
> + set_pirq_mask(ap, PIRQ_MASK_OTHER);
> +
> + /* Issuing a command to yet uninitialized port locks up the
> + * controller. Most of the time, this happens for the first
> + * command after reset which are ATA and ATAPI IDENTIFYs.
> + * Fast fail if stat is 0x7f or 0xff for those commands.
> + */
> + if (unlikely(qc->tf.command == ATA_CMD_ID_ATA ||
> + qc->tf.command == ATA_CMD_ID_ATAPI)) {
> + u8 stat = ata_chk_status(ap);
> + if (stat == 0x7f || stat == 0xff)
> + return AC_ERR_HSM;
> + }
> +
> + return ata_qc_issue_prot(qc);
> +}
> +
> +static void inic_freeze(struct ata_port *ap)
> +{
> + void __iomem *port_base = get_port_base(ap);
> +
> + __set_pirq_mask(ap, PIRQ_MASK_FREEZE);
> +
> + ata_chk_status(ap);
> + writeb(0xff, port_base + PORT_IRQ_STAT);
> +}
> +
> +static void inic_thaw(struct ata_port *ap)
> +{
> + void __iomem *port_base = get_port_base(ap);
> +
> + ata_chk_status(ap);
> + writeb(0xff, port_base + PORT_IRQ_STAT);
> +
> + __set_pirq_mask(ap, PIRQ_MASK_OTHER);
> +}
> +
> +/*
> + * SRST and SControl hardreset don't give valid signature on this
> + * controller. Only controller specific hardreset mechanism works.
> + */
> +static int inic_hardreset(struct ata_port *ap, unsigned int *class)
> +{
> + void __iomem *port_base = get_port_base(ap);
> + void __iomem *idma_ctl = port_base + PORT_IDMA_CTL;
> + const unsigned long *timing = sata_ehc_deb_timing(&ap->eh_context);
> + u16 val;
> + int rc;
> +
> + /* hammer it into sane state */
> + reset_port(port_base);
> +
> + if (ata_port_offline(ap)) {
> + *class = ATA_DEV_NONE;
> + return 0;
> + }
> +
> + val = readw(idma_ctl);
> + writew(val | IDMA_CTL_RST_ATA, idma_ctl);
> + readw(idma_ctl); /* flush */
> + msleep(1);
> + writew(val & ~IDMA_CTL_RST_ATA, idma_ctl);
> +
> + rc = sata_phy_resume(ap, timing);
> + if (rc) {
> + ata_port_printk(ap, KERN_WARNING, "failed to resume "
> + "link for reset (errno=%d)\n", rc);
> + return rc;
> + }
> +
> + msleep(150);
> +
> + *class = ATA_DEV_NONE;
> + if (ata_port_online(ap)) {
> + struct ata_taskfile tf;
> +
> + if (ata_busy_sleep(ap, ATA_TMOUT_BOOT_QUICK, ATA_TMOUT_BOOT)) {
> + ata_port_printk(ap, KERN_WARNING,
> + "device busy after hardreset\n");
> + return -EIO;
> + }
> +
> + ata_tf_read(ap, &tf);
> + *class = ata_dev_classify(&tf);
> + if (*class == ATA_DEV_UNKNOWN)
> + *class = ATA_DEV_NONE;
> + }
> +
> + return 0;
> +}
> +
> +static void inic_error_handler(struct ata_port *ap)
> +{
> + void __iomem *port_base = get_port_base(ap);
> + struct inic_port_priv *pp = ap->private_data;
> + unsigned long flags;
> +
> + /* reset PIO HSM and stop DMA engine */
> + reset_port(port_base);
This function name is a bit too generic, and more difficult to narrow
down to the single driver when viewed in an oops stack trace
> + spin_lock_irqsave(ap->lock, flags);
> + ap->hsm_task_state = HSM_ST_IDLE;
> + writeb(pp->dfl_prdctl, port_base + PORT_PRD_CTL);
> + spin_unlock_irqrestore(ap->lock, flags);
> +
> + /* PIO and DMA engines have been stopped, perform recovery */
> + ata_do_eh(ap, ata_std_prereset, NULL, inic_hardreset,
> + ata_std_postreset);
> +}
> +
> +static void inic_post_internal_cmd(struct ata_queued_cmd *qc)
> +{
> + /* make DMA engine forget about the failed command */
> + if (qc->err_mask)
> + reset_port(get_port_base(qc->ap));
> +}
> +
> +static void inic_dev_config(struct ata_port *ap, struct ata_device *dev)
> +{
> + /* inic can only handle upto LBA28 max sectors */
> + if (dev->max_sectors > ATA_MAX_SECTORS)
> + dev->max_sectors = ATA_MAX_SECTORS;
> +}
why is this needed? scsi host template should take care of this, right?
> +static void init_port(struct ata_port *ap)
> +{
> + void __iomem *port_base = get_port_base(ap);
> +
> + /* Setup PRD address */
> + writel(ap->prd_dma, port_base + PORT_PRD_ADDR);
> +}
> +
> +static int inic_port_resume(struct ata_port *ap)
> +{
> + init_port(ap);
likewise, name is too generic for the global namespace (even though its
a 'static' function)
> + return 0;
> +}
> +
> +static int inic_port_start(struct ata_port *ap)
> +{
> + void __iomem *port_base = get_port_base(ap);
> + struct inic_port_priv *pp;
> + u8 tmp;
> + int rc;
> +
> + /* alloc and initialize private data */
> + pp = kzalloc(sizeof(*pp), GFP_KERNEL);
> + if (!pp)
> + return -ENOMEM;
> + ap->private_data = pp;
> +
> + /* default PRD_CTL value, DMAEN, WR and START off */
> + tmp = readb(port_base + PORT_PRD_CTL);
> + tmp &= ~(PRD_CTL_DMAEN | PRD_CTL_WR | PRD_CTL_START);
> + pp->dfl_prdctl = tmp;
> +
> + /* Alloc resources */
> + rc = ata_port_start(ap);
> + if (rc) {
> + kfree(pp);
> + return rc;
> + }
> +
> + init_port(ap);
> +
> + return 0;
> +}
> +
> +static void inic_port_stop(struct ata_port *ap)
> +{
> + ata_port_stop(ap);
> + kfree(ap->private_data);
> +}
> +
> +static struct ata_port_operations inic_port_ops = {
> + .port_disable = ata_port_disable,
> + .tf_load = ata_tf_load,
> + .tf_read = ata_tf_read,
> + .check_status = ata_check_status,
> + .exec_command = ata_exec_command,
> + .dev_select = ata_std_dev_select,
> +
> + .scr_read = inic_scr_read,
> + .scr_write = inic_scr_write,
> +
> + .bmdma_setup = inic_bmdma_setup,
> + .bmdma_start = inic_bmdma_start,
> + .bmdma_stop = inic_bmdma_stop,
> + .bmdma_status = inic_bmdma_status,
> +
> + .irq_handler = inic_interrupt,
> + .irq_clear = inic_irq_clear,
> +
> + .qc_prep = ata_qc_prep,
> + .qc_issue = inic_qc_issue,
> + .data_xfer = ata_pio_data_xfer,
> +
> + .freeze = inic_freeze,
> + .thaw = inic_thaw,
> + .error_handler = inic_error_handler,
> + .post_internal_cmd = inic_post_internal_cmd,
> + .dev_config = inic_dev_config,
> +
> + .port_resume = inic_port_resume,
> +
> + .port_start = inic_port_start,
> + .port_stop = inic_port_stop,
> + .host_stop = ata_pci_host_stop
> +};
> +
> +static struct ata_port_info inic_port_info = {
> + .sht = &inic_sht,
> + /* For some reason, ATA_PROT_ATAPI is broken on this
> + * controller, and no, PIO_POLLING does't fix it. It somehow
> + * manages to report the wrong ireason and ignoring ireason
> + * results in machine lock up. Tell libata to always prefer
> + * DMA.
> + */
> + .flags = ATA_FLAG_SATA | ATA_FLAG_PIO_DMA,
> + .pio_mask = 0x1f, /* pio0-4 */
> + .mwdma_mask = 0x07, /* mwdma0-2 */
> + .udma_mask = 0x7f, /* udma0-6 */
> + .port_ops = &inic_port_ops
> +};
> +
> +static int init_controller(void __iomem *mmio_base, u16 hctl)
> +{
> + int i;
> + u16 val;
> +
> + hctl &= ~HCTL_KNOWN_BITS;
> +
> + /* Soft reset whole controller. Spec says reset duration is 3
> + * PCI clocks, be generous and give it 10ms.
> + */
> + writew(hctl | HCTL_SOFTRST, mmio_base + HOST_CTL);
> + readw(mmio_base + HOST_CTL); /* flush */
> +
> + for (i = 0; i < 10; i++) {
> + msleep(1);
> + val = readw(mmio_base + HOST_CTL);
> + if (!(val & HCTL_SOFTRST))
> + break;
> + }
> +
> + if (val & HCTL_SOFTRST)
> + return -EIO;
> +
> + /* mask all interrupts and reset ports */
> + for (i = 0; i < NR_PORTS; i++) {
> + void __iomem *port_base = mmio_base + i * PORT_SIZE;
> +
> + writeb(0xff, port_base + PORT_IRQ_MASK);
> + reset_port(port_base);
> + }
> +
> + /* port IRQ is masked now, unmask global IRQ */
> + writew(hctl & ~HCTL_IRQOFF, mmio_base + HOST_CTL);
> + val = readw(mmio_base + HOST_IRQ_MASK);
> + val &= ~(HIRQ_PORT0 | HIRQ_PORT1);
> + writew(val, mmio_base + HOST_IRQ_MASK);
> +
> + return 0;
> +}
> +
> +static int inic_pci_device_resume(struct pci_dev *pdev)
> +{
> + struct ata_host *host = dev_get_drvdata(&pdev->dev);
> + struct inic_host_priv *hpriv = host->private_data;
> + void __iomem *mmio_base = host->mmio_base;
> + int rc;
> +
> + ata_pci_device_do_resume(pdev);
> +
> + if (pdev->dev.power.power_state.event == PM_EVENT_SUSPEND) {
> + rc = init_controller(mmio_base, hpriv->cached_hctl);
> + if (rc)
> + return rc;
> + }
> +
> + ata_host_resume(host);
> +
> + return 0;
> +}
> +
> +static int inic_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
> +{
> + static int printed_version;
> + struct ata_port_info *pinfo = &inic_port_info;
> + struct ata_probe_ent *probe_ent;
> + struct inic_host_priv *hpriv;
> + void __iomem *mmio_base;
> + int i, rc;
> +
> + if (!printed_version++)
> + dev_printk(KERN_DEBUG, &pdev->dev, "version " DRV_VERSION "\n");
> +
> + rc = pci_enable_device(pdev);
> + if (rc)
> + return rc;
> +
> + rc = pci_request_regions(pdev, DRV_NAME);
> + if (rc)
> + goto err_out;
> +
> + rc = -ENOMEM;
> + mmio_base = pci_iomap(pdev, MMIO_BAR, 0);
> + if (!mmio_base)
> + goto err_out_regions;
> +
> + /* Set dma_mask. This devices doesn't support 64bit addressing. */
> + rc = pci_set_dma_mask(pdev, DMA_32BIT_MASK);
> + if (rc) {
> + dev_printk(KERN_ERR, &pdev->dev,
> + "32-bit DMA enable failed\n");
> + goto err_out_map;
> + }
> +
> + rc = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK);
> + if (rc) {
> + dev_printk(KERN_ERR, &pdev->dev,
> + "32-bit consistent DMA enable failed\n");
> + goto err_out_map;
> + }
> +
> + rc = -ENOMEM;
> + probe_ent = kzalloc(sizeof(*probe_ent), GFP_KERNEL);
> + if (!probe_ent)
> + goto err_out_map;
> +
> + hpriv = kzalloc(sizeof(*hpriv), GFP_KERNEL);
> + if (!hpriv)
> + goto err_out_ent;
> +
> + probe_ent->dev = &pdev->dev;
> + INIT_LIST_HEAD(&probe_ent->node);
> +
> + probe_ent->sht = pinfo->sht;
> + probe_ent->port_flags = pinfo->flags;
> + probe_ent->pio_mask = pinfo->pio_mask;
> + probe_ent->mwdma_mask = pinfo->mwdma_mask;
> + probe_ent->udma_mask = pinfo->udma_mask;
> + probe_ent->port_ops = pinfo->port_ops;
> + probe_ent->n_ports = NR_PORTS;
> +
> + probe_ent->irq = pdev->irq;
> + probe_ent->irq_flags = SA_SHIRQ;
> +
> + probe_ent->mmio_base = mmio_base;
> +
> + for (i = 0; i < NR_PORTS; i++) {
> + struct ata_ioports *port = &probe_ent->port[i];
> + unsigned long port_base =
> + (unsigned long)mmio_base + i * PORT_SIZE;
> +
> + port->cmd_addr = pci_resource_start(pdev, 2 * i);
> + port->altstatus_addr =
> + port->ctl_addr =
> + pci_resource_start(pdev, 2 * i + 1) | ATA_PCI_CTL_OFS;
> + port->scr_addr = port_base + PORT_SCR;
> +
> + ata_std_ports(port);
> + }
> +
> + probe_ent->private_data = hpriv;
> + hpriv->cached_hctl = readw(mmio_base + HOST_CTL);
> +
> + rc = init_controller(mmio_base, hpriv->cached_hctl);
> + if (rc) {
> + dev_printk(KERN_ERR, &pdev->dev,
> + "failed to initialize controller\n");
> + goto err_out_hpriv;
> + }
> +
> + pci_set_master(pdev);
> +
> + rc = -ENODEV;
> + if (!ata_device_add(probe_ent))
> + goto err_out_hpriv;
> +
> + kfree(probe_ent);
> +
> + return 0;
> +
> + err_out_hpriv:
> + kfree(hpriv);
> + err_out_ent:
> + kfree(probe_ent);
> + err_out_map:
> + pci_iounmap(pdev, mmio_base);
> + err_out_regions:
> + pci_release_regions(pdev);
> + err_out:
> + pci_disable_device(pdev);
> + return rc;
> +}
> +
> +static const struct pci_device_id inic_pci_tbl[] = {
> + { PCI_VDEVICE(INIT, 0x1622), },
> + { },
> +};
> +
> +static struct pci_driver inic_pci_driver = {
> + .name = DRV_NAME,
> + .id_table = inic_pci_tbl,
> + .suspend = ata_pci_device_suspend,
> + .resume = inic_pci_device_resume,
> + .probe = inic_init_one,
> + .remove = ata_pci_remove_one,
> +};
> +
> +static int __init inic_init(void)
> +{
> + return pci_register_driver(&inic_pci_driver);
> +}
> +
> +static void __exit inic_exit(void)
> +{
> + pci_unregister_driver(&inic_pci_driver);
> +}
> +
> +MODULE_AUTHOR("Tejun Heo");
> +MODULE_DESCRIPTION("low-level driver for Initio 162x SATA");
> +MODULE_LICENSE("GPL v2");
> +MODULE_DEVICE_TABLE(pci, inic_pci_tbl);
> +MODULE_VERSION(DRV_VERSION);
> +
> +module_init(inic_init);
> +module_exit(inic_exit);
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ide" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
next prev parent reply other threads:[~2006-12-20 19:56 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-12-17 1:48 [PATCH 1/2] libata: kill qc->nsect and cursect Tejun Heo
2006-12-17 1:50 ` [PATCH 2/2] sata_inic162x: finally, driver for initio 162x SATA controllers Tejun Heo
2006-12-17 12:24 ` Alan
2006-12-18 0:04 ` Tejun Heo
2006-12-18 0:25 ` Bob Stewart
2006-12-20 6:21 ` sata_inic162x driver for 2.6.19 Tejun Heo
2006-12-21 0:06 ` Bob Stewart
2006-12-21 1:48 ` Bob Stewart
2006-12-21 4:16 ` Bob Stewart
2007-01-10 23:34 ` Richard Purdie
2007-01-11 0:04 ` Alan
2007-01-11 2:00 ` Bob Stewart
2006-12-20 6:25 ` [PATCH 2/2] sata_inic162x: driver for initio 162x SATA controllers, take 2 Tejun Heo
2006-12-20 19:56 ` Jeff Garzik [this message]
2007-01-03 8:18 ` Tejun Heo
2007-01-03 8:30 ` [PATCH 1/2] libata: kill qc->nsect and cursect Tejun Heo
2007-01-03 8:32 ` [PATCH 2/2] sata_inic162x: finally, driver for initio 162x SATA controllers, take #2 Tejun Heo
2007-01-20 0:04 ` Jeff Garzik
2007-01-20 0:04 ` [PATCH 1/2] libata: kill qc->nsect and cursect Jeff Garzik
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=45899560.8090401@pobox.com \
--to=jgarzik@pobox.com \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=bob@evoria.net \
--cc=carlosmarcelomartinez@gmail.com \
--cc=htejun@gmail.com \
--cc=linux-ide@vger.kernel.org \
--cc=matthias@urlichs.de \
--cc=nabble@theanthonys.net \
--cc=romieu@fr.zoreil.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).