* [git patches] libata updates
@ 2008-05-06 15:48 Jeff Garzik
2008-05-06 16:23 ` Linus Torvalds
2008-05-07 14:56 ` [git patches] libata updates - (improve post-reset device ready test) regression Markus Trippelsdorf
0 siblings, 2 replies; 22+ messages in thread
From: Jeff Garzik @ 2008-05-06 15:48 UTC (permalink / raw)
To: Andrew Morton, Linus Torvalds; +Cc: linux-ide, LKML
- sata_mv: continued fixing
- sata_inic: make it actually work
- new driver from Intel
- misc fixes from Tejun
Please pull from 'upstream-linus' branch of
master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/libata-dev.git upstream-linus
to receive the following updates:
drivers/ata/Kconfig | 13 +-
drivers/ata/Makefile | 1 +
drivers/ata/ahci.c | 4 +-
drivers/ata/ata_generic.c | 6 +
drivers/ata/ata_piix.c | 25 ++
drivers/ata/libata-core.c | 1 +
drivers/ata/libata-eh.c | 2 +-
drivers/ata/libata-sff.c | 6 +-
drivers/ata/pata_acpi.c | 6 +
drivers/ata/pata_sch.c | 206 +++++++++++++
drivers/ata/sata_inic162x.c | 646 ++++++++++++++++++++++++++--------------
drivers/ata/sata_mv.c | 688 +++++++++++++++++++++++++++++++++----------
include/linux/libata.h | 16 +
13 files changed, 1222 insertions(+), 398 deletions(-)
create mode 100644 drivers/ata/pata_sch.c
Alan Cox (1):
pata_atiixp: Don't disable
Alek Du (1):
libata: Add Intel SCH PATA driver
Mark Lord (13):
sata_mv more cosmetic changes
sata_mv pci features
sata_mv wait for empty+idle
sata_mv new mv_qc_defer method
sata_mv errata workaround for sata25 part 1
sata_mv rearrange mv_config_fbs
sata_mv NCQ and SError fixes for mv_err_intr
sata_mv fix mv_host_intr bug for hc_irq_cause
sata_mv new mv_port_intr function
libata: export ata_eh_analyze_ncq_error
sata_mv delayed eh handling
sata_mv NCQ-EH for FIS-based switching
sata_mv use hweight16() for bit counting (V2)
Tejun Heo (12):
libata: improve post-reset device ready test
ata_piix: verify SIDPR access before enabling it
sata_inic162x: misc clean ups
sata_inic162x: add / update constants
sata_inic162x: update TF read handling
sata_inic162x: use IDMA for ATA_PROT_DMA
sata_inic162x: kill now unused bmdma related stuff
sata_inic162x: use IDMA for non DMA ATA commands
sata_inic162x: use IDMA for ATAPI commands
sata_inic162x: kill now unused SFF related stuff
sata_inic162x: add cardbus support
sata_inic162x: update intro comment, up the version and drop EXPERIMENTAL
diff --git a/drivers/ata/Kconfig b/drivers/ata/Kconfig
index 1c11df9..9bf2986 100644
--- a/drivers/ata/Kconfig
+++ b/drivers/ata/Kconfig
@@ -205,8 +205,8 @@ config SATA_VITESSE
If unsure, say N.
config SATA_INIC162X
- tristate "Initio 162x SATA support (HIGHLY EXPERIMENTAL)"
- depends on PCI && EXPERIMENTAL
+ tristate "Initio 162x SATA support"
+ depends on PCI
help
This option enables support for Initio 162x Serial ATA.
@@ -697,6 +697,15 @@ config PATA_SCC
If unsure, say N.
+config PATA_SCH
+ tristate "Intel SCH PATA support"
+ depends on PCI
+ help
+ This option enables support for Intel SCH PATA on the Intel
+ SCH (US15W, US15L, UL11L) series host controllers.
+
+ If unsure, say N.
+
config PATA_BF54X
tristate "Blackfin 54x ATAPI support"
depends on BF542 || BF548 || BF549
diff --git a/drivers/ata/Makefile b/drivers/ata/Makefile
index b693d82..674965f 100644
--- a/drivers/ata/Makefile
+++ b/drivers/ata/Makefile
@@ -67,6 +67,7 @@ obj-$(CONFIG_PATA_SIS) += pata_sis.o
obj-$(CONFIG_PATA_TRIFLEX) += pata_triflex.o
obj-$(CONFIG_PATA_IXP4XX_CF) += pata_ixp4xx_cf.o
obj-$(CONFIG_PATA_SCC) += pata_scc.o
+obj-$(CONFIG_PATA_SCH) += pata_sch.o
obj-$(CONFIG_PATA_BF54X) += pata_bf54x.o
obj-$(CONFIG_PATA_PLATFORM) += pata_platform.o
obj-$(CONFIG_PATA_OF_PLATFORM) += pata_of_platform.o
diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
index 8cace9a..97f83fb 100644
--- a/drivers/ata/ahci.c
+++ b/drivers/ata/ahci.c
@@ -1267,9 +1267,7 @@ static int ahci_check_ready(struct ata_link *link)
void __iomem *port_mmio = ahci_port_base(link->ap);
u8 status = readl(port_mmio + PORT_TFDATA) & 0xFF;
- if (!(status & ATA_BUSY))
- return 1;
- return 0;
+ return ata_check_ready(status);
}
static int ahci_softreset(struct ata_link *link, unsigned int *class,
diff --git a/drivers/ata/ata_generic.c b/drivers/ata/ata_generic.c
index 47aeccd..75a406f 100644
--- a/drivers/ata/ata_generic.c
+++ b/drivers/ata/ata_generic.c
@@ -152,6 +152,12 @@ static int ata_generic_init_one(struct pci_dev *dev, const struct pci_device_id
if (dev->vendor == PCI_VENDOR_ID_AL)
ata_pci_bmdma_clear_simplex(dev);
+ if (dev->vendor == PCI_VENDOR_ID_ATI) {
+ int rc = pcim_enable_device(dev);
+ if (rc < 0)
+ return rc;
+ pcim_pin_device(dev);
+ }
return ata_pci_sff_init_one(dev, ppi, &generic_sht, NULL);
}
diff --git a/drivers/ata/ata_piix.c b/drivers/ata/ata_piix.c
index ea2c764..a9027b8 100644
--- a/drivers/ata/ata_piix.c
+++ b/drivers/ata/ata_piix.c
@@ -1348,6 +1348,8 @@ static void __devinit piix_init_sidpr(struct ata_host *host)
{
struct pci_dev *pdev = to_pci_dev(host->dev);
struct piix_host_priv *hpriv = host->private_data;
+ struct ata_device *dev0 = &host->ports[0]->link.device[0];
+ u32 scontrol;
int i;
/* check for availability */
@@ -1366,6 +1368,29 @@ static void __devinit piix_init_sidpr(struct ata_host *host)
return;
hpriv->sidpr = pcim_iomap_table(pdev)[PIIX_SIDPR_BAR];
+
+ /* SCR access via SIDPR doesn't work on some configurations.
+ * Give it a test drive by inhibiting power save modes which
+ * we'll do anyway.
+ */
+ scontrol = piix_sidpr_read(dev0, SCR_CONTROL);
+
+ /* if IPM is already 3, SCR access is probably working. Don't
+ * un-inhibit power save modes as BIOS might have inhibited
+ * them for a reason.
+ */
+ if ((scontrol & 0xf00) != 0x300) {
+ scontrol |= 0x300;
+ piix_sidpr_write(dev0, SCR_CONTROL, scontrol);
+ scontrol = piix_sidpr_read(dev0, SCR_CONTROL);
+
+ if ((scontrol & 0xf00) != 0x300) {
+ dev_printk(KERN_INFO, host->dev, "SCR access via "
+ "SIDPR is available but doesn't work\n");
+ return;
+ }
+ }
+
host->ports[0]->ops = &piix_sidpr_sata_ops;
host->ports[1]->ops = &piix_sidpr_sata_ops;
}
diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
index 3bc4885..927b692 100644
--- a/drivers/ata/libata-core.c
+++ b/drivers/ata/libata-core.c
@@ -6292,6 +6292,7 @@ EXPORT_SYMBOL_GPL(ata_eh_freeze_port);
EXPORT_SYMBOL_GPL(ata_eh_thaw_port);
EXPORT_SYMBOL_GPL(ata_eh_qc_complete);
EXPORT_SYMBOL_GPL(ata_eh_qc_retry);
+EXPORT_SYMBOL_GPL(ata_eh_analyze_ncq_error);
EXPORT_SYMBOL_GPL(ata_do_eh);
EXPORT_SYMBOL_GPL(ata_std_error_handler);
diff --git a/drivers/ata/libata-eh.c b/drivers/ata/libata-eh.c
index 61dcd00..62e0331 100644
--- a/drivers/ata/libata-eh.c
+++ b/drivers/ata/libata-eh.c
@@ -1357,7 +1357,7 @@ static void ata_eh_analyze_serror(struct ata_link *link)
* LOCKING:
* Kernel thread context (may sleep).
*/
-static void ata_eh_analyze_ncq_error(struct ata_link *link)
+void ata_eh_analyze_ncq_error(struct ata_link *link)
{
struct ata_port *ap = link->ap;
struct ata_eh_context *ehc = &link->eh_context;
diff --git a/drivers/ata/libata-sff.c b/drivers/ata/libata-sff.c
index 2ec65a8..3c2d228 100644
--- a/drivers/ata/libata-sff.c
+++ b/drivers/ata/libata-sff.c
@@ -314,11 +314,7 @@ static int ata_sff_check_ready(struct ata_link *link)
{
u8 status = link->ap->ops->sff_check_status(link->ap);
- if (!(status & ATA_BUSY))
- return 1;
- if (status == 0xff)
- return -ENODEV;
- return 0;
+ return ata_check_ready(status);
}
/**
diff --git a/drivers/ata/pata_acpi.c b/drivers/ata/pata_acpi.c
index c5f91e6..fbe6057 100644
--- a/drivers/ata/pata_acpi.c
+++ b/drivers/ata/pata_acpi.c
@@ -259,6 +259,12 @@ static int pacpi_init_one (struct pci_dev *pdev, const struct pci_device_id *id)
.port_ops = &pacpi_ops,
};
const struct ata_port_info *ppi[] = { &info, NULL };
+ if (pdev->vendor == PCI_VENDOR_ID_ATI) {
+ int rc = pcim_enable_device(pdev);
+ if (rc < 0)
+ return rc;
+ pcim_pin_device(pdev);
+ }
return ata_pci_sff_init_one(pdev, ppi, &pacpi_sht, NULL);
}
diff --git a/drivers/ata/pata_sch.c b/drivers/ata/pata_sch.c
new file mode 100644
index 0000000..c8cc027
--- /dev/null
+++ b/drivers/ata/pata_sch.c
@@ -0,0 +1,206 @@
+/*
+ * pata_sch.c - Intel SCH PATA controllers
+ *
+ * Copyright (c) 2008 Alek Du <alek.du@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License 2 as published
+ * by the Free Software Foundation.
+ *
+ * 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; see the file COPYING. If not, write to
+ * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+
+/*
+ * Supports:
+ * Intel SCH (AF82US15W, AF82US15L, AF82UL11L) chipsets -- see spec at:
+ * http://download.intel.com/design/chipsets/embedded/datashts/319537.pdf
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/init.h>
+#include <linux/blkdev.h>
+#include <linux/delay.h>
+#include <linux/device.h>
+#include <scsi/scsi_host.h>
+#include <linux/libata.h>
+#include <linux/dmi.h>
+
+#define DRV_NAME "pata_sch"
+#define DRV_VERSION "0.2"
+
+/* see SCH datasheet page 351 */
+enum {
+ D0TIM = 0x80, /* Device 0 Timing Register */
+ D1TIM = 0x84, /* Device 1 Timing Register */
+ PM = 0x07, /* PIO Mode Bit Mask */
+ MDM = (0x03 << 8), /* Multi-word DMA Mode Bit Mask */
+ UDM = (0x07 << 16), /* Ultra DMA Mode Bit Mask */
+ PPE = (1 << 30), /* Prefetch/Post Enable */
+ USD = (1 << 31), /* Use Synchronous DMA */
+};
+
+static int sch_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent);
+static void sch_set_piomode(struct ata_port *ap, struct ata_device *adev);
+static void sch_set_dmamode(struct ata_port *ap, struct ata_device *adev);
+
+static const struct pci_device_id sch_pci_tbl[] = {
+ /* Intel SCH PATA Controller */
+ { PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_SCH_IDE), 0 },
+ { } /* terminate list */
+};
+
+static struct pci_driver sch_pci_driver = {
+ .name = DRV_NAME,
+ .id_table = sch_pci_tbl,
+ .probe = sch_init_one,
+ .remove = ata_pci_remove_one,
+#ifdef CONFIG_PM
+ .suspend = ata_pci_device_suspend,
+ .resume = ata_pci_device_resume,
+#endif
+};
+
+static struct scsi_host_template sch_sht = {
+ ATA_BMDMA_SHT(DRV_NAME),
+};
+
+static struct ata_port_operations sch_pata_ops = {
+ .inherits = &ata_bmdma_port_ops,
+ .cable_detect = ata_cable_unknown,
+ .set_piomode = sch_set_piomode,
+ .set_dmamode = sch_set_dmamode,
+};
+
+static struct ata_port_info sch_port_info = {
+ .flags = 0,
+ .pio_mask = ATA_PIO4, /* pio0-4 */
+ .mwdma_mask = ATA_MWDMA2, /* mwdma0-2 */
+ .udma_mask = ATA_UDMA5, /* udma0-5 */
+ .port_ops = &sch_pata_ops,
+};
+
+MODULE_AUTHOR("Alek Du <alek.du@intel.com>");
+MODULE_DESCRIPTION("SCSI low-level driver for Intel SCH PATA controllers");
+MODULE_LICENSE("GPL");
+MODULE_DEVICE_TABLE(pci, sch_pci_tbl);
+MODULE_VERSION(DRV_VERSION);
+
+/**
+ * sch_set_piomode - Initialize host controller PATA PIO timings
+ * @ap: Port whose timings we are configuring
+ * @adev: ATA device
+ *
+ * Set PIO mode for device, in host controller PCI config space.
+ *
+ * LOCKING:
+ * None (inherited from caller).
+ */
+
+static void sch_set_piomode(struct ata_port *ap, struct ata_device *adev)
+{
+ unsigned int pio = adev->pio_mode - XFER_PIO_0;
+ struct pci_dev *dev = to_pci_dev(ap->host->dev);
+ unsigned int port = adev->devno ? D1TIM : D0TIM;
+ unsigned int data;
+
+ pci_read_config_dword(dev, port, &data);
+ /* see SCH datasheet page 351 */
+ /* set PIO mode */
+ data &= ~(PM | PPE);
+ data |= pio;
+ /* enable PPE for block device */
+ if (adev->class == ATA_DEV_ATA)
+ data |= PPE;
+ pci_write_config_dword(dev, port, data);
+}
+
+/**
+ * sch_set_dmamode - Initialize host controller PATA DMA timings
+ * @ap: Port whose timings we are configuring
+ * @adev: ATA device
+ *
+ * Set MW/UDMA mode for device, in host controller PCI config space.
+ *
+ * LOCKING:
+ * None (inherited from caller).
+ */
+
+static void sch_set_dmamode(struct ata_port *ap, struct ata_device *adev)
+{
+ unsigned int dma_mode = adev->dma_mode;
+ struct pci_dev *dev = to_pci_dev(ap->host->dev);
+ unsigned int port = adev->devno ? D1TIM : D0TIM;
+ unsigned int data;
+
+ pci_read_config_dword(dev, port, &data);
+ /* see SCH datasheet page 351 */
+ if (dma_mode >= XFER_UDMA_0) {
+ /* enable Synchronous DMA mode */
+ data |= USD;
+ data &= ~UDM;
+ data |= (dma_mode - XFER_UDMA_0) << 16;
+ } else { /* must be MWDMA mode, since we masked SWDMA already */
+ data &= ~(USD | MDM);
+ data |= (dma_mode - XFER_MW_DMA_0) << 8;
+ }
+ pci_write_config_dword(dev, port, data);
+}
+
+/**
+ * sch_init_one - Register SCH ATA PCI device with kernel services
+ * @pdev: PCI device to register
+ * @ent: Entry in sch_pci_tbl matching with @pdev
+ *
+ * LOCKING:
+ * Inherited from PCI layer (may sleep).
+ *
+ * RETURNS:
+ * Zero on success, or -ERRNO value.
+ */
+
+static int __devinit sch_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ static int printed_version;
+ const struct ata_port_info *ppi[] = { &sch_port_info, NULL };
+ struct ata_host *host;
+ int rc;
+
+ if (!printed_version++)
+ dev_printk(KERN_DEBUG, &pdev->dev,
+ "version " DRV_VERSION "\n");
+
+ /* enable device and prepare host */
+ rc = pcim_enable_device(pdev);
+ if (rc)
+ return rc;
+ rc = ata_pci_sff_prepare_host(pdev, ppi, &host);
+ if (rc)
+ return rc;
+ pci_set_master(pdev);
+ return ata_pci_sff_activate_host(host, ata_sff_interrupt, &sch_sht);
+}
+
+static int __init sch_init(void)
+{
+ return pci_register_driver(&sch_pci_driver);
+}
+
+static void __exit sch_exit(void)
+{
+ pci_unregister_driver(&sch_pci_driver);
+}
+
+module_init(sch_init);
+module_exit(sch_exit);
diff --git a/drivers/ata/sata_inic162x.c b/drivers/ata/sata_inic162x.c
index d27bb9a..3ead02f 100644
--- a/drivers/ata/sata_inic162x.c
+++ b/drivers/ata/sata_inic162x.c
@@ -10,13 +10,33 @@
* 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.
+ * This driver has interesting history. The first version was written
+ * from the documentation and a 2.4 IDE driver posted on a Taiwan
+ * company, which didn't use any IDMA features and couldn't handle
+ * LBA48. The resulting driver couldn't handle LBA48 devices either
+ * making it pretty useless.
+ *
+ * After a while, initio picked the driver up, renamed it to
+ * sata_initio162x, updated it to use IDMA for ATA DMA commands and
+ * posted it on their website. It only used ATA_PROT_DMA for IDMA and
+ * attaching both devices and issuing IDMA and !IDMA commands
+ * simultaneously broke it due to PIRQ masking interaction but it did
+ * show how to use the IDMA (ADMA + some initio specific twists)
+ * engine.
+ *
+ * Then, I picked up their changes again and here's the usable driver
+ * which uses IDMA for everything. Everything works now including
+ * LBA48, CD/DVD burning, suspend/resume and hotplug. There are some
+ * issues tho. Result Tf is not resported properly, NCQ isn't
+ * supported yet and CD/DVD writing works with DMA assisted PIO
+ * protocol (which, for native SATA devices, shouldn't cause any
+ * noticeable difference).
+ *
+ * Anyways, so, here's finally a working driver for inic162x. Enjoy!
+ *
+ * initio: If you guys wanna improve the driver regarding result TF
+ * access and other stuff, please feel free to contact me. I'll be
+ * happy to assist.
*/
#include <linux/kernel.h>
@@ -28,13 +48,19 @@
#include <scsi/scsi_device.h>
#define DRV_NAME "sata_inic162x"
-#define DRV_VERSION "0.3"
+#define DRV_VERSION "0.4"
enum {
- MMIO_BAR = 5,
+ MMIO_BAR_PCI = 5,
+ MMIO_BAR_CARDBUS = 1,
NR_PORTS = 2,
+ IDMA_CPB_TBL_SIZE = 4 * 32,
+
+ INIC_DMA_BOUNDARY = 0xffffff,
+
+ HOST_ACTRL = 0x08,
HOST_CTL = 0x7c,
HOST_STAT = 0x7e,
HOST_IRQ_STAT = 0xbc,
@@ -43,22 +69,37 @@ enum {
PORT_SIZE = 0x40,
/* registers for ATA TF operation */
- PORT_TF = 0x00,
- PORT_ALT_STAT = 0x08,
+ PORT_TF_DATA = 0x00,
+ PORT_TF_FEATURE = 0x01,
+ PORT_TF_NSECT = 0x02,
+ PORT_TF_LBAL = 0x03,
+ PORT_TF_LBAM = 0x04,
+ PORT_TF_LBAH = 0x05,
+ PORT_TF_DEVICE = 0x06,
+ PORT_TF_COMMAND = 0x07,
+ PORT_TF_ALT_STAT = 0x08,
PORT_IRQ_STAT = 0x09,
PORT_IRQ_MASK = 0x0a,
PORT_PRD_CTL = 0x0b,
PORT_PRD_ADDR = 0x0c,
PORT_PRD_XFERLEN = 0x10,
+ PORT_CPB_CPBLAR = 0x18,
+ PORT_CPB_PTQFIFO = 0x1c,
/* IDMA register */
PORT_IDMA_CTL = 0x14,
+ PORT_IDMA_STAT = 0x16,
+
+ PORT_RPQ_FIFO = 0x1e,
+ PORT_RPQ_CNT = 0x1f,
PORT_SCR = 0x20,
/* HOST_CTL bits */
HCTL_IRQOFF = (1 << 8), /* global IRQ off */
- HCTL_PWRDWN = (1 << 13), /* power down PHYs */
+ HCTL_FTHD0 = (1 << 10), /* fifo threshold 0 */
+ HCTL_FTHD1 = (1 << 11), /* fifo threshold 1*/
+ HCTL_PWRDWN = (1 << 12), /* power down PHYs */
HCTL_SOFTRST = (1 << 13), /* global reset (no phy reset) */
HCTL_RPGSEL = (1 << 15), /* register page select */
@@ -81,9 +122,7 @@ enum {
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_DEFAULT = PIRQ_REPLY | PIRQ_ATA,
PIRQ_MASK_FREEZE = 0xff,
/* PORT_PRD_CTL bits */
@@ -96,20 +135,104 @@ enum {
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 */
+
+ /* PORT_IDMA_STAT bits */
+ IDMA_STAT_PERR = (1 << 0), /* PCI ERROR MODE */
+ IDMA_STAT_CPBERR = (1 << 1), /* ADMA CPB error */
+ IDMA_STAT_LGCY = (1 << 3), /* ADMA legacy */
+ IDMA_STAT_UIRQ = (1 << 4), /* ADMA unsolicited irq */
+ IDMA_STAT_STPD = (1 << 5), /* ADMA stopped */
+ IDMA_STAT_PSD = (1 << 6), /* ADMA pause */
+ IDMA_STAT_DONE = (1 << 7), /* ADMA done */
+
+ IDMA_STAT_ERR = IDMA_STAT_PERR | IDMA_STAT_CPBERR,
+
+ /* CPB Control Flags*/
+ CPB_CTL_VALID = (1 << 0), /* CPB valid */
+ CPB_CTL_QUEUED = (1 << 1), /* queued command */
+ CPB_CTL_DATA = (1 << 2), /* data, rsvd in datasheet */
+ CPB_CTL_IEN = (1 << 3), /* PCI interrupt enable */
+ CPB_CTL_DEVDIR = (1 << 4), /* device direction control */
+
+ /* CPB Response Flags */
+ CPB_RESP_DONE = (1 << 0), /* ATA command complete */
+ CPB_RESP_REL = (1 << 1), /* ATA release */
+ CPB_RESP_IGNORED = (1 << 2), /* CPB ignored */
+ CPB_RESP_ATA_ERR = (1 << 3), /* ATA command error */
+ CPB_RESP_SPURIOUS = (1 << 4), /* ATA spurious interrupt error */
+ CPB_RESP_UNDERFLOW = (1 << 5), /* APRD deficiency length error */
+ CPB_RESP_OVERFLOW = (1 << 6), /* APRD exccess length error */
+ CPB_RESP_CPB_ERR = (1 << 7), /* CPB error flag */
+
+ /* PRD Control Flags */
+ PRD_DRAIN = (1 << 1), /* ignore data excess */
+ PRD_CDB = (1 << 2), /* atapi packet command pointer */
+ PRD_DIRECT_INTR = (1 << 3), /* direct interrupt */
+ PRD_DMA = (1 << 4), /* data transfer method */
+ PRD_WRITE = (1 << 5), /* data dir, rsvd in datasheet */
+ PRD_IOM = (1 << 6), /* io/memory transfer */
+ PRD_END = (1 << 7), /* APRD chain end */
};
+/* Comman Parameter Block */
+struct inic_cpb {
+ u8 resp_flags; /* Response Flags */
+ u8 error; /* ATA Error */
+ u8 status; /* ATA Status */
+ u8 ctl_flags; /* Control Flags */
+ __le32 len; /* Total Transfer Length */
+ __le32 prd; /* First PRD pointer */
+ u8 rsvd[4];
+ /* 16 bytes */
+ u8 feature; /* ATA Feature */
+ u8 hob_feature; /* ATA Ex. Feature */
+ u8 device; /* ATA Device/Head */
+ u8 mirctl; /* Mirror Control */
+ u8 nsect; /* ATA Sector Count */
+ u8 hob_nsect; /* ATA Ex. Sector Count */
+ u8 lbal; /* ATA Sector Number */
+ u8 hob_lbal; /* ATA Ex. Sector Number */
+ u8 lbam; /* ATA Cylinder Low */
+ u8 hob_lbam; /* ATA Ex. Cylinder Low */
+ u8 lbah; /* ATA Cylinder High */
+ u8 hob_lbah; /* ATA Ex. Cylinder High */
+ u8 command; /* ATA Command */
+ u8 ctl; /* ATA Control */
+ u8 slave_error; /* Slave ATA Error */
+ u8 slave_status; /* Slave ATA Status */
+ /* 32 bytes */
+} __packed;
+
+/* Physical Region Descriptor */
+struct inic_prd {
+ __le32 mad; /* Physical Memory Address */
+ __le16 len; /* Transfer Length */
+ u8 rsvd;
+ u8 flags; /* Control Flags */
+} __packed;
+
+struct inic_pkt {
+ struct inic_cpb cpb;
+ struct inic_prd prd[LIBATA_MAX_PRD + 1]; /* + 1 for cdb */
+ u8 cdb[ATAPI_CDB_LEN];
+} __packed;
+
struct inic_host_priv {
- u16 cached_hctl;
+ void __iomem *mmio_base;
+ u16 cached_hctl;
};
struct inic_port_priv {
- u8 dfl_prdctl;
- u8 cached_prdctl;
- u8 cached_pirq_mask;
+ struct inic_pkt *pkt;
+ dma_addr_t pkt_dma;
+ u32 *cpb_tbl;
+ dma_addr_t cpb_tbl_dma;
};
static struct scsi_host_template inic_sht = {
- ATA_BMDMA_SHT(DRV_NAME),
+ ATA_BASE_SHT(DRV_NAME),
+ .sg_tablesize = LIBATA_MAX_PRD, /* maybe it can be larger? */
+ .dma_boundary = INIC_DMA_BOUNDARY,
};
static const int scr_map[] = {
@@ -120,54 +243,34 @@ static const int scr_map[] = {
static void __iomem *inic_port_base(struct ata_port *ap)
{
- return ap->host->iomap[MMIO_BAR] + ap->port_no * PORT_SIZE;
-}
-
-static void __inic_set_pirq_mask(struct ata_port *ap, u8 mask)
-{
- void __iomem *port_base = inic_port_base(ap);
- struct inic_port_priv *pp = ap->private_data;
+ struct inic_host_priv *hpriv = ap->host->private_data;
- writeb(mask, port_base + PORT_IRQ_MASK);
- pp->cached_pirq_mask = mask;
-}
-
-static void inic_set_pirq_mask(struct ata_port *ap, u8 mask)
-{
- struct inic_port_priv *pp = ap->private_data;
-
- if (pp->cached_pirq_mask != mask)
- __inic_set_pirq_mask(ap, mask);
+ return hpriv->mmio_base + ap->port_no * PORT_SIZE;
}
static void inic_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);
+ /* stop IDMA engine */
+ readw(idma_ctl); /* flush */
+ msleep(1);
/* mask IRQ and assert reset */
- writew(ctl | IDMA_CTL_RST_IDMA | IDMA_CTL_ATA_NIEN, idma_ctl);
+ writew(IDMA_CTL_RST_IDMA, idma_ctl);
readw(idma_ctl); /* flush */
-
- /* give it some time */
msleep(1);
/* release reset */
- writew(ctl | IDMA_CTL_ATA_NIEN, idma_ctl);
+ writew(0, idma_ctl);
/* clear irq */
writeb(0xff, port_base + PORT_IRQ_STAT);
-
- /* reenable ATA IRQ, turn off IDMA mode */
- writew(ctl, idma_ctl);
}
static int inic_scr_read(struct ata_port *ap, unsigned sc_reg, u32 *val)
{
- void __iomem *scr_addr = ap->ioaddr.scr_addr;
+ void __iomem *scr_addr = inic_port_base(ap) + PORT_SCR;
void __iomem *addr;
if (unlikely(sc_reg >= ARRAY_SIZE(scr_map)))
@@ -184,120 +287,126 @@ static int inic_scr_read(struct ata_port *ap, unsigned sc_reg, u32 *val)
static int inic_scr_write(struct ata_port *ap, unsigned sc_reg, u32 val)
{
- void __iomem *scr_addr = ap->ioaddr.scr_addr;
- void __iomem *addr;
+ void __iomem *scr_addr = inic_port_base(ap) + PORT_SCR;
if (unlikely(sc_reg >= ARRAY_SIZE(scr_map)))
return -EINVAL;
- addr = scr_addr + scr_map[sc_reg] * 4;
writel(val, scr_addr + scr_map[sc_reg] * 4);
return 0;
}
-/*
- * 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)
+static void inic_stop_idma(struct ata_port *ap)
{
- struct ata_port *ap = qc->ap;
- struct inic_port_priv *pp = ap->private_data;
void __iomem *port_base = inic_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->sff_exec_command(ap, &qc->tf);
+ readb(port_base + PORT_RPQ_FIFO);
+ readb(port_base + PORT_RPQ_CNT);
+ writew(0, port_base + PORT_IDMA_CTL);
}
-static void inic_bmdma_start(struct ata_queued_cmd *qc)
+static void inic_host_err_intr(struct ata_port *ap, u8 irq_stat, u16 idma_stat)
{
- struct ata_port *ap = qc->ap;
+ struct ata_eh_info *ehi = &ap->link.eh_info;
struct inic_port_priv *pp = ap->private_data;
- void __iomem *port_base = inic_port_base(ap);
+ struct inic_cpb *cpb = &pp->pkt->cpb;
+ bool freeze = false;
- /* start host DMA transaction */
- pp->cached_prdctl |= PRD_CTL_START;
- writeb(pp->cached_prdctl, port_base + PORT_PRD_CTL);
-}
+ ata_ehi_clear_desc(ehi);
+ ata_ehi_push_desc(ehi, "irq_stat=0x%x idma_stat=0x%x",
+ irq_stat, idma_stat);
-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 = inic_port_base(ap);
+ inic_stop_idma(ap);
- /* stop DMA engine */
- writeb(pp->dfl_prdctl, port_base + PORT_PRD_CTL);
-}
+ if (irq_stat & (PIRQ_OFFLINE | PIRQ_ONLINE)) {
+ ata_ehi_push_desc(ehi, "hotplug");
+ ata_ehi_hotplugged(ehi);
+ freeze = true;
+ }
-static u8 inic_bmdma_status(struct ata_port *ap)
-{
- /* event is already verified by the interrupt handler */
- return ATA_DMA_INTR;
+ if (idma_stat & IDMA_STAT_PERR) {
+ ata_ehi_push_desc(ehi, "PCI error");
+ freeze = true;
+ }
+
+ if (idma_stat & IDMA_STAT_CPBERR) {
+ ata_ehi_push_desc(ehi, "CPB error");
+
+ if (cpb->resp_flags & CPB_RESP_IGNORED) {
+ __ata_ehi_push_desc(ehi, " ignored");
+ ehi->err_mask |= AC_ERR_INVALID;
+ freeze = true;
+ }
+
+ if (cpb->resp_flags & CPB_RESP_ATA_ERR)
+ ehi->err_mask |= AC_ERR_DEV;
+
+ if (cpb->resp_flags & CPB_RESP_SPURIOUS) {
+ __ata_ehi_push_desc(ehi, " spurious-intr");
+ ehi->err_mask |= AC_ERR_HSM;
+ freeze = true;
+ }
+
+ if (cpb->resp_flags &
+ (CPB_RESP_UNDERFLOW | CPB_RESP_OVERFLOW)) {
+ __ata_ehi_push_desc(ehi, " data-over/underflow");
+ ehi->err_mask |= AC_ERR_HSM;
+ freeze = true;
+ }
+ }
+
+ if (freeze)
+ ata_port_freeze(ap);
+ else
+ ata_port_abort(ap);
}
static void inic_host_intr(struct ata_port *ap)
{
void __iomem *port_base = inic_port_base(ap);
- struct ata_eh_info *ehi = &ap->link.eh_info;
+ struct ata_queued_cmd *qc = ata_qc_from_tag(ap, ap->link.active_tag);
u8 irq_stat;
+ u16 idma_stat;
- /* fetch and clear irq */
+ /* read and clear IRQ status */
irq_stat = readb(port_base + PORT_IRQ_STAT);
writeb(irq_stat, port_base + PORT_IRQ_STAT);
+ idma_stat = readw(port_base + PORT_IDMA_STAT);
- if (likely(!(irq_stat & PIRQ_ERR))) {
- struct ata_queued_cmd *qc =
- ata_qc_from_tag(ap, ap->link.active_tag);
+ if (unlikely((irq_stat & PIRQ_ERR) || (idma_stat & IDMA_STAT_ERR)))
+ inic_host_err_intr(ap, irq_stat, idma_stat);
- if (unlikely(!qc || (qc->tf.flags & ATA_TFLAG_POLLING))) {
- ap->ops->sff_check_status(ap); /* clear ATA interrupt */
- return;
- }
+ if (unlikely(!qc))
+ goto spurious;
- if (likely(ata_sff_host_intr(ap, qc)))
- return;
+ if (likely(idma_stat & IDMA_STAT_DONE)) {
+ inic_stop_idma(ap);
- ap->ops->sff_check_status(ap); /* clear ATA interrupt */
- ata_port_printk(ap, KERN_WARNING, "unhandled "
- "interrupt, irq_stat=%x\n", irq_stat);
+ /* Depending on circumstances, device error
+ * isn't reported by IDMA, check it explicitly.
+ */
+ if (unlikely(readb(port_base + PORT_TF_COMMAND) &
+ (ATA_DF | ATA_ERR)))
+ qc->err_mask |= AC_ERR_DEV;
+
+ ata_qc_complete(qc);
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);
+ spurious:
+ ata_port_printk(ap, KERN_WARNING, "unhandled interrupt: "
+ "cmd=0x%x irq_stat=0x%x idma_stat=0x%x\n",
+ qc ? qc->tf.command : 0xff, irq_stat, idma_stat);
}
static irqreturn_t inic_interrupt(int irq, void *dev_instance)
{
struct ata_host *host = dev_instance;
- void __iomem *mmio_base = host->iomap[MMIO_BAR];
+ struct inic_host_priv *hpriv = host->private_data;
u16 host_irq_stat;
int i, handled = 0;;
- host_irq_stat = readw(mmio_base + HOST_IRQ_STAT);
+ host_irq_stat = readw(hpriv->mmio_base + HOST_IRQ_STAT);
if (unlikely(!(host_irq_stat & HIRQ_GLOBAL)))
goto out;
@@ -327,60 +436,173 @@ static irqreturn_t inic_interrupt(int irq, void *dev_instance)
return IRQ_RETVAL(handled);
}
+static int inic_check_atapi_dma(struct ata_queued_cmd *qc)
+{
+ /* For some reason ATAPI_PROT_DMA doesn't work for some
+ * commands including writes and other misc ops. Use PIO
+ * protocol instead, which BTW is driven by the DMA engine
+ * anyway, so it shouldn't make much difference for native
+ * SATA devices.
+ */
+ if (atapi_cmd_type(qc->cdb[0]) == READ)
+ return 0;
+ return 1;
+}
+
+static void inic_fill_sg(struct inic_prd *prd, struct ata_queued_cmd *qc)
+{
+ struct scatterlist *sg;
+ unsigned int si;
+ u8 flags = 0;
+
+ if (qc->tf.flags & ATA_TFLAG_WRITE)
+ flags |= PRD_WRITE;
+
+ if (ata_is_dma(qc->tf.protocol))
+ flags |= PRD_DMA;
+
+ for_each_sg(qc->sg, sg, qc->n_elem, si) {
+ prd->mad = cpu_to_le32(sg_dma_address(sg));
+ prd->len = cpu_to_le16(sg_dma_len(sg));
+ prd->flags = flags;
+ prd++;
+ }
+
+ WARN_ON(!si);
+ prd[-1].flags |= PRD_END;
+}
+
+static void inic_qc_prep(struct ata_queued_cmd *qc)
+{
+ struct inic_port_priv *pp = qc->ap->private_data;
+ struct inic_pkt *pkt = pp->pkt;
+ struct inic_cpb *cpb = &pkt->cpb;
+ struct inic_prd *prd = pkt->prd;
+ bool is_atapi = ata_is_atapi(qc->tf.protocol);
+ bool is_data = ata_is_data(qc->tf.protocol);
+ unsigned int cdb_len = 0;
+
+ VPRINTK("ENTER\n");
+
+ if (is_atapi)
+ cdb_len = qc->dev->cdb_len;
+
+ /* prepare packet, based on initio driver */
+ memset(pkt, 0, sizeof(struct inic_pkt));
+
+ cpb->ctl_flags = CPB_CTL_VALID | CPB_CTL_IEN;
+ if (is_atapi || is_data)
+ cpb->ctl_flags |= CPB_CTL_DATA;
+
+ cpb->len = cpu_to_le32(qc->nbytes + cdb_len);
+ cpb->prd = cpu_to_le32(pp->pkt_dma + offsetof(struct inic_pkt, prd));
+
+ cpb->device = qc->tf.device;
+ cpb->feature = qc->tf.feature;
+ cpb->nsect = qc->tf.nsect;
+ cpb->lbal = qc->tf.lbal;
+ cpb->lbam = qc->tf.lbam;
+ cpb->lbah = qc->tf.lbah;
+
+ if (qc->tf.flags & ATA_TFLAG_LBA48) {
+ cpb->hob_feature = qc->tf.hob_feature;
+ cpb->hob_nsect = qc->tf.hob_nsect;
+ cpb->hob_lbal = qc->tf.hob_lbal;
+ cpb->hob_lbam = qc->tf.hob_lbam;
+ cpb->hob_lbah = qc->tf.hob_lbah;
+ }
+
+ cpb->command = qc->tf.command;
+ /* don't load ctl - dunno why. it's like that in the initio driver */
+
+ /* setup PRD for CDB */
+ if (is_atapi) {
+ memcpy(pkt->cdb, qc->cdb, ATAPI_CDB_LEN);
+ prd->mad = cpu_to_le32(pp->pkt_dma +
+ offsetof(struct inic_pkt, cdb));
+ prd->len = cpu_to_le16(cdb_len);
+ prd->flags = PRD_CDB | PRD_WRITE;
+ if (!is_data)
+ prd->flags |= PRD_END;
+ prd++;
+ }
+
+ /* setup sg table */
+ if (is_data)
+ inic_fill_sg(prd, qc);
+
+ pp->cpb_tbl[0] = pp->pkt_dma;
+}
+
static unsigned int inic_qc_issue(struct ata_queued_cmd *qc)
{
struct ata_port *ap = qc->ap;
+ void __iomem *port_base = inic_port_base(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))
- inic_set_pirq_mask(ap, PIRQ_MASK_DMA_READ);
- else
- inic_set_pirq_mask(ap, PIRQ_MASK_OTHER);
+ /* fire up the ADMA engine */
+ writew(HCTL_FTHD0, port_base + HOST_CTL);
+ writew(IDMA_CTL_GO, port_base + PORT_IDMA_CTL);
+ writeb(0, port_base + PORT_CPB_PTQFIFO);
+
+ return 0;
+}
+
+static void inic_tf_read(struct ata_port *ap, struct ata_taskfile *tf)
+{
+ void __iomem *port_base = inic_port_base(ap);
+
+ tf->feature = readb(port_base + PORT_TF_FEATURE);
+ tf->nsect = readb(port_base + PORT_TF_NSECT);
+ tf->lbal = readb(port_base + PORT_TF_LBAL);
+ tf->lbam = readb(port_base + PORT_TF_LBAM);
+ tf->lbah = readb(port_base + PORT_TF_LBAH);
+ tf->device = readb(port_base + PORT_TF_DEVICE);
+ tf->command = readb(port_base + PORT_TF_COMMAND);
+}
- /* 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.
+static bool inic_qc_fill_rtf(struct ata_queued_cmd *qc)
+{
+ struct ata_taskfile *rtf = &qc->result_tf;
+ struct ata_taskfile tf;
+
+ /* FIXME: Except for status and error, result TF access
+ * doesn't work. I tried reading from BAR0/2, CPB and BAR5.
+ * None works regardless of which command interface is used.
+ * For now return true iff status indicates device error.
+ * This means that we're reporting bogus sector for RW
+ * failures. Eeekk....
*/
- if (unlikely(qc->tf.command == ATA_CMD_ID_ATA ||
- qc->tf.command == ATA_CMD_ID_ATAPI)) {
- u8 stat = ap->ops->sff_check_status(ap);
- if (stat == 0x7f || stat == 0xff)
- return AC_ERR_HSM;
- }
+ inic_tf_read(qc->ap, &tf);
- return ata_sff_qc_issue(qc);
+ if (!(tf.command & ATA_ERR))
+ return false;
+
+ rtf->command = tf.command;
+ rtf->feature = tf.feature;
+ return true;
}
static void inic_freeze(struct ata_port *ap)
{
void __iomem *port_base = inic_port_base(ap);
- __inic_set_pirq_mask(ap, PIRQ_MASK_FREEZE);
-
- ap->ops->sff_check_status(ap);
+ writeb(PIRQ_MASK_FREEZE, port_base + PORT_IRQ_MASK);
writeb(0xff, port_base + PORT_IRQ_STAT);
-
- readb(port_base + PORT_IRQ_STAT); /* flush */
}
static void inic_thaw(struct ata_port *ap)
{
void __iomem *port_base = inic_port_base(ap);
- ap->ops->sff_check_status(ap);
writeb(0xff, port_base + PORT_IRQ_STAT);
+ writeb(PIRQ_MASK_DEFAULT, port_base + PORT_IRQ_MASK);
+}
- __inic_set_pirq_mask(ap, PIRQ_MASK_OTHER);
+static int inic_check_ready(struct ata_link *link)
+{
+ void __iomem *port_base = inic_port_base(link->ap);
- readb(port_base + PORT_IRQ_STAT); /* flush */
+ return ata_check_ready(readb(port_base + PORT_TF_COMMAND));
}
/*
@@ -394,17 +616,15 @@ static int inic_hardreset(struct ata_link *link, unsigned int *class,
void __iomem *port_base = inic_port_base(ap);
void __iomem *idma_ctl = port_base + PORT_IDMA_CTL;
const unsigned long *timing = sata_ehc_deb_timing(&link->eh_context);
- u16 val;
int rc;
/* hammer it into sane state */
inic_reset_port(port_base);
- val = readw(idma_ctl);
- writew(val | IDMA_CTL_RST_ATA, idma_ctl);
+ writew(IDMA_CTL_RST_ATA, idma_ctl);
readw(idma_ctl); /* flush */
msleep(1);
- writew(val & ~IDMA_CTL_RST_ATA, idma_ctl);
+ writew(0, idma_ctl);
rc = sata_link_resume(link, timing, deadline);
if (rc) {
@@ -418,7 +638,7 @@ static int inic_hardreset(struct ata_link *link, unsigned int *class,
struct ata_taskfile tf;
/* wait for link to become ready */
- rc = ata_sff_wait_after_reset(link, 1, deadline);
+ rc = ata_wait_after_reset(link, deadline, inic_check_ready);
/* link occupied, -ENODEV too is an error */
if (rc) {
ata_link_printk(link, KERN_WARNING, "device not ready "
@@ -426,7 +646,7 @@ static int inic_hardreset(struct ata_link *link, unsigned int *class,
return rc;
}
- ata_sff_tf_read(ap, &tf);
+ inic_tf_read(ap, &tf);
*class = ata_dev_classify(&tf);
}
@@ -436,18 +656,8 @@ static int inic_hardreset(struct ata_link *link, unsigned int *class,
static void inic_error_handler(struct ata_port *ap)
{
void __iomem *port_base = inic_port_base(ap);
- struct inic_port_priv *pp = ap->private_data;
- unsigned long flags;
- /* reset PIO HSM and stop DMA engine */
inic_reset_port(port_base);
-
- 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_std_error_handler(ap);
}
@@ -458,26 +668,18 @@ static void inic_post_internal_cmd(struct ata_queued_cmd *qc)
inic_reset_port(inic_port_base(qc->ap));
}
-static void inic_dev_config(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;
-
- if (dev->n_sectors >= 1 << 28) {
- ata_dev_printk(dev, KERN_ERR,
- "ERROR: This driver doesn't support LBA48 yet and may cause\n"
- " data corruption on such devices. Disabling.\n");
- ata_dev_disable(dev);
- }
-}
-
static void init_port(struct ata_port *ap)
{
void __iomem *port_base = inic_port_base(ap);
+ struct inic_port_priv *pp = ap->private_data;
- /* Setup PRD address */
+ /* clear packet and CPB table */
+ memset(pp->pkt, 0, sizeof(struct inic_pkt));
+ memset(pp->cpb_tbl, 0, IDMA_CPB_TBL_SIZE);
+
+ /* setup PRD and CPB lookup table addresses */
writel(ap->prd_dma, port_base + PORT_PRD_ADDR);
+ writel(pp->cpb_tbl_dma, port_base + PORT_CPB_CPBLAR);
}
static int inic_port_resume(struct ata_port *ap)
@@ -488,28 +690,30 @@ static int inic_port_resume(struct ata_port *ap)
static int inic_port_start(struct ata_port *ap)
{
- void __iomem *port_base = inic_port_base(ap);
+ struct device *dev = ap->host->dev;
struct inic_port_priv *pp;
- u8 tmp;
int rc;
/* alloc and initialize private data */
- pp = devm_kzalloc(ap->host->dev, sizeof(*pp), GFP_KERNEL);
+ pp = devm_kzalloc(dev, 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);
+ if (rc)
return rc;
- }
+
+ pp->pkt = dmam_alloc_coherent(dev, sizeof(struct inic_pkt),
+ &pp->pkt_dma, GFP_KERNEL);
+ if (!pp->pkt)
+ return -ENOMEM;
+
+ pp->cpb_tbl = dmam_alloc_coherent(dev, IDMA_CPB_TBL_SIZE,
+ &pp->cpb_tbl_dma, GFP_KERNEL);
+ if (!pp->cpb_tbl)
+ return -ENOMEM;
init_port(ap);
@@ -517,21 +721,18 @@ static int inic_port_start(struct ata_port *ap)
}
static struct ata_port_operations inic_port_ops = {
- .inherits = &ata_sff_port_ops,
+ .inherits = &sata_port_ops,
- .bmdma_setup = inic_bmdma_setup,
- .bmdma_start = inic_bmdma_start,
- .bmdma_stop = inic_bmdma_stop,
- .bmdma_status = inic_bmdma_status,
+ .check_atapi_dma = inic_check_atapi_dma,
+ .qc_prep = inic_qc_prep,
.qc_issue = inic_qc_issue,
+ .qc_fill_rtf = inic_qc_fill_rtf,
.freeze = inic_freeze,
.thaw = inic_thaw,
- .softreset = ATA_OP_NULL, /* softreset is broken */
.hardreset = inic_hardreset,
.error_handler = inic_error_handler,
.post_internal_cmd = inic_post_internal_cmd,
- .dev_config = inic_dev_config,
.scr_read = inic_scr_read,
.scr_write = inic_scr_write,
@@ -541,12 +742,6 @@ static struct ata_port_operations inic_port_ops = {
};
static struct ata_port_info inic_port_info = {
- /* For some reason, ATAPI_PROT_PIO 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 */
@@ -599,7 +794,6 @@ 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->iomap[MMIO_BAR];
int rc;
rc = ata_pci_device_do_resume(pdev);
@@ -607,7 +801,7 @@ static int inic_pci_device_resume(struct pci_dev *pdev)
return rc;
if (pdev->dev.power.power_state.event == PM_EVENT_SUSPEND) {
- rc = init_controller(mmio_base, hpriv->cached_hctl);
+ rc = init_controller(hpriv->mmio_base, hpriv->cached_hctl);
if (rc)
return rc;
}
@@ -625,6 +819,7 @@ static int inic_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
struct ata_host *host;
struct inic_host_priv *hpriv;
void __iomem * const *iomap;
+ int mmio_bar;
int i, rc;
if (!printed_version++)
@@ -638,38 +833,31 @@ static int inic_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
host->private_data = hpriv;
- /* acquire resources and fill host */
+ /* Acquire resources and fill host. Note that PCI and cardbus
+ * use different BARs.
+ */
rc = pcim_enable_device(pdev);
if (rc)
return rc;
- rc = pcim_iomap_regions(pdev, 0x3f, DRV_NAME);
+ if (pci_resource_flags(pdev, MMIO_BAR_PCI) & IORESOURCE_MEM)
+ mmio_bar = MMIO_BAR_PCI;
+ else
+ mmio_bar = MMIO_BAR_CARDBUS;
+
+ rc = pcim_iomap_regions(pdev, 1 << mmio_bar, DRV_NAME);
if (rc)
return rc;
host->iomap = iomap = pcim_iomap_table(pdev);
+ hpriv->mmio_base = iomap[mmio_bar];
+ hpriv->cached_hctl = readw(hpriv->mmio_base + HOST_CTL);
for (i = 0; i < NR_PORTS; i++) {
struct ata_port *ap = host->ports[i];
- struct ata_ioports *port = &ap->ioaddr;
- unsigned int offset = i * PORT_SIZE;
-
- port->cmd_addr = iomap[2 * i];
- port->altstatus_addr =
- port->ctl_addr = (void __iomem *)
- ((unsigned long)iomap[2 * i + 1] | ATA_PCI_CTL_OFS);
- port->scr_addr = iomap[MMIO_BAR] + offset + PORT_SCR;
-
- ata_sff_std_ports(port);
-
- ata_port_pbar_desc(ap, MMIO_BAR, -1, "mmio");
- ata_port_pbar_desc(ap, MMIO_BAR, offset, "port");
- ata_port_desc(ap, "cmd 0x%llx ctl 0x%llx",
- (unsigned long long)pci_resource_start(pdev, 2 * i),
- (unsigned long long)pci_resource_start(pdev, (2 * i + 1)) |
- ATA_PCI_CTL_OFS);
- }
- hpriv->cached_hctl = readw(iomap[MMIO_BAR] + HOST_CTL);
+ ata_port_pbar_desc(ap, mmio_bar, -1, "mmio");
+ ata_port_pbar_desc(ap, mmio_bar, i * PORT_SIZE, "port");
+ }
/* Set dma_mask. This devices doesn't support 64bit addressing. */
rc = pci_set_dma_mask(pdev, DMA_32BIT_MASK);
@@ -698,7 +886,7 @@ static int inic_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
return rc;
}
- rc = init_controller(iomap[MMIO_BAR], hpriv->cached_hctl);
+ rc = init_controller(hpriv->mmio_base, hpriv->cached_hctl);
if (rc) {
dev_printk(KERN_ERR, &pdev->dev,
"failed to initialize controller\n");
diff --git a/drivers/ata/sata_mv.c b/drivers/ata/sata_mv.c
index 842b1a1..bb73b22 100644
--- a/drivers/ata/sata_mv.c
+++ b/drivers/ata/sata_mv.c
@@ -65,6 +65,7 @@
#include <linux/platform_device.h>
#include <linux/ata_platform.h>
#include <linux/mbus.h>
+#include <linux/bitops.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_device.h>
@@ -91,9 +92,9 @@ enum {
MV_IRQ_COAL_TIME_THRESHOLD = (MV_IRQ_COAL_REG_BASE + 0xd0),
MV_SATAHC0_REG_BASE = 0x20000,
- MV_FLASH_CTL = 0x1046c,
- MV_GPIO_PORT_CTL = 0x104f0,
- MV_RESET_CFG = 0x180d8,
+ MV_FLASH_CTL_OFS = 0x1046c,
+ MV_GPIO_PORT_CTL_OFS = 0x104f0,
+ MV_RESET_CFG_OFS = 0x180d8,
MV_PCI_REG_SZ = MV_MAJOR_REG_AREA_SZ,
MV_SATAHC_REG_SZ = MV_MAJOR_REG_AREA_SZ,
@@ -147,18 +148,21 @@ enum {
/* PCI interface registers */
PCI_COMMAND_OFS = 0xc00,
+ PCI_COMMAND_MRDTRIG = (1 << 7), /* PCI Master Read Trigger */
PCI_MAIN_CMD_STS_OFS = 0xd30,
STOP_PCI_MASTER = (1 << 2),
PCI_MASTER_EMPTY = (1 << 3),
GLOB_SFT_RST = (1 << 4),
- MV_PCI_MODE = 0xd00,
+ MV_PCI_MODE_OFS = 0xd00,
+ MV_PCI_MODE_MASK = 0x30,
+
MV_PCI_EXP_ROM_BAR_CTL = 0xd2c,
MV_PCI_DISC_TIMER = 0xd04,
MV_PCI_MSI_TRIGGER = 0xc38,
MV_PCI_SERR_MASK = 0xc28,
- MV_PCI_XBAR_TMOUT = 0x1d04,
+ MV_PCI_XBAR_TMOUT_OFS = 0x1d04,
MV_PCI_ERR_LOW_ADDRESS = 0x1d40,
MV_PCI_ERR_HIGH_ADDRESS = 0x1d44,
MV_PCI_ERR_ATTRIBUTE = 0x1d48,
@@ -225,16 +229,18 @@ enum {
PHY_MODE4 = 0x314,
PHY_MODE2 = 0x330,
SATA_IFCTL_OFS = 0x344,
+ SATA_TESTCTL_OFS = 0x348,
SATA_IFSTAT_OFS = 0x34c,
VENDOR_UNIQUE_FIS_OFS = 0x35c,
- FIS_CFG_OFS = 0x360,
- FIS_CFG_SINGLE_SYNC = (1 << 16), /* SYNC on DMA activation */
+ FISCFG_OFS = 0x360,
+ FISCFG_WAIT_DEV_ERR = (1 << 8), /* wait for host on DevErr */
+ FISCFG_SINGLE_SYNC = (1 << 16), /* SYNC on DMA activation */
MV5_PHY_MODE = 0x74,
- MV5_LT_MODE = 0x30,
- MV5_PHY_CTL = 0x0C,
- SATA_INTERFACE_CFG = 0x050,
+ MV5_LTMODE_OFS = 0x30,
+ MV5_PHY_CTL_OFS = 0x0C,
+ SATA_INTERFACE_CFG_OFS = 0x050,
MV_M2_PREAMP_MASK = 0x7e0,
@@ -332,10 +338,16 @@ enum {
EDMA_CMD_OFS = 0x28, /* EDMA command register */
EDMA_EN = (1 << 0), /* enable EDMA */
EDMA_DS = (1 << 1), /* disable EDMA; self-negated */
- ATA_RST = (1 << 2), /* reset trans/link/phy */
+ EDMA_RESET = (1 << 2), /* reset eng/trans/link/phy */
+
+ EDMA_STATUS_OFS = 0x30, /* EDMA engine status */
+ EDMA_STATUS_CACHE_EMPTY = (1 << 6), /* GenIIe command cache empty */
+ EDMA_STATUS_IDLE = (1 << 7), /* GenIIe EDMA enabled/idle */
- EDMA_IORDY_TMOUT = 0x34,
- EDMA_ARB_CFG = 0x38,
+ EDMA_IORDY_TMOUT_OFS = 0x34,
+ EDMA_ARB_CFG_OFS = 0x38,
+
+ EDMA_HALTCOND_OFS = 0x60, /* GenIIe halt conditions */
GEN_II_NCQ_MAX_SECTORS = 256, /* max sects/io on Gen2 w/NCQ */
@@ -350,15 +362,19 @@ enum {
MV_HP_GEN_II = (1 << 7), /* Generation II: 60xx */
MV_HP_GEN_IIE = (1 << 8), /* Generation IIE: 6042/7042 */
MV_HP_PCIE = (1 << 9), /* PCIe bus/regs: 7042 */
+ MV_HP_CUT_THROUGH = (1 << 10), /* can use EDMA cut-through */
/* Port private flags (pp_flags) */
MV_PP_FLAG_EDMA_EN = (1 << 0), /* is EDMA engine enabled? */
MV_PP_FLAG_NCQ_EN = (1 << 1), /* is EDMA set up for NCQ? */
+ MV_PP_FLAG_FBS_EN = (1 << 2), /* is EDMA set up for FBS? */
+ MV_PP_FLAG_DELAYED_EH = (1 << 3), /* delayed dev err handling */
};
#define IS_GEN_I(hpriv) ((hpriv)->hp_flags & MV_HP_GEN_I)
#define IS_GEN_II(hpriv) ((hpriv)->hp_flags & MV_HP_GEN_II)
#define IS_GEN_IIE(hpriv) ((hpriv)->hp_flags & MV_HP_GEN_IIE)
+#define IS_PCIE(hpriv) ((hpriv)->hp_flags & MV_HP_PCIE)
#define HAS_PCI(host) (!((host)->ports[0]->flags & MV_FLAG_SOC))
#define WINDOW_CTRL(i) (0x20030 + ((i) << 4))
@@ -433,6 +449,7 @@ struct mv_port_priv {
unsigned int resp_idx;
u32 pp_flags;
+ unsigned int delayed_eh_pmp_map;
};
struct mv_port_signal {
@@ -479,6 +496,7 @@ static int mv5_scr_read(struct ata_port *ap, unsigned int sc_reg_in, u32 *val);
static int mv5_scr_write(struct ata_port *ap, unsigned int sc_reg_in, u32 val);
static int mv_port_start(struct ata_port *ap);
static void mv_port_stop(struct ata_port *ap);
+static int mv_qc_defer(struct ata_queued_cmd *qc);
static void mv_qc_prep(struct ata_queued_cmd *qc);
static void mv_qc_prep_iie(struct ata_queued_cmd *qc);
static unsigned int mv_qc_issue(struct ata_queued_cmd *qc);
@@ -527,6 +545,9 @@ static int mv_pmp_hardreset(struct ata_link *link, unsigned int *class,
unsigned long deadline);
static int mv_softreset(struct ata_link *link, unsigned int *class,
unsigned long deadline);
+static void mv_pmp_error_handler(struct ata_port *ap);
+static void mv_process_crpb_entries(struct ata_port *ap,
+ struct mv_port_priv *pp);
/* .sg_tablesize is (MV_MAX_SG_CT / 2) in the structures below
* because we have to allow room for worst case splitting of
@@ -548,6 +569,7 @@ static struct scsi_host_template mv6_sht = {
static struct ata_port_operations mv5_ops = {
.inherits = &ata_sff_port_ops,
+ .qc_defer = mv_qc_defer,
.qc_prep = mv_qc_prep,
.qc_issue = mv_qc_issue,
@@ -566,7 +588,6 @@ static struct ata_port_operations mv5_ops = {
static struct ata_port_operations mv6_ops = {
.inherits = &mv5_ops,
- .qc_defer = sata_pmp_qc_defer_cmd_switch,
.dev_config = mv6_dev_config,
.scr_read = mv_scr_read,
.scr_write = mv_scr_write,
@@ -574,12 +595,11 @@ static struct ata_port_operations mv6_ops = {
.pmp_hardreset = mv_pmp_hardreset,
.pmp_softreset = mv_softreset,
.softreset = mv_softreset,
- .error_handler = sata_pmp_error_handler,
+ .error_handler = mv_pmp_error_handler,
};
static struct ata_port_operations mv_iie_ops = {
.inherits = &mv6_ops,
- .qc_defer = ata_std_qc_defer, /* FIS-based switching */
.dev_config = ATA_OP_NULL,
.qc_prep = mv_qc_prep_iie,
};
@@ -875,6 +895,29 @@ static void mv_start_dma(struct ata_port *ap, void __iomem *port_mmio,
}
}
+static void mv_wait_for_edma_empty_idle(struct ata_port *ap)
+{
+ void __iomem *port_mmio = mv_ap_base(ap);
+ const u32 empty_idle = (EDMA_STATUS_CACHE_EMPTY | EDMA_STATUS_IDLE);
+ const int per_loop = 5, timeout = (15 * 1000 / per_loop);
+ int i;
+
+ /*
+ * Wait for the EDMA engine to finish transactions in progress.
+ * No idea what a good "timeout" value might be, but measurements
+ * indicate that it often requires hundreds of microseconds
+ * with two drives in-use. So we use the 15msec value above
+ * as a rough guess at what even more drives might require.
+ */
+ for (i = 0; i < timeout; ++i) {
+ u32 edma_stat = readl(port_mmio + EDMA_STATUS_OFS);
+ if ((edma_stat & empty_idle) == empty_idle)
+ break;
+ udelay(per_loop);
+ }
+ /* ata_port_printk(ap, KERN_INFO, "%s: %u+ usecs\n", __func__, i); */
+}
+
/**
* mv_stop_edma_engine - Disable eDMA engine
* @port_mmio: io base address
@@ -907,6 +950,7 @@ static int mv_stop_edma(struct ata_port *ap)
if (!(pp->pp_flags & MV_PP_FLAG_EDMA_EN))
return 0;
pp->pp_flags &= ~MV_PP_FLAG_EDMA_EN;
+ mv_wait_for_edma_empty_idle(ap);
if (mv_stop_edma_engine(port_mmio)) {
ata_port_printk(ap, KERN_ERR, "Unable to stop eDMA\n");
return -EIO;
@@ -1057,26 +1101,95 @@ static void mv6_dev_config(struct ata_device *adev)
}
}
-static void mv_config_fbs(void __iomem *port_mmio, int enable_fbs)
+static int mv_qc_defer(struct ata_queued_cmd *qc)
{
- u32 old_fcfg, new_fcfg, old_ltmode, new_ltmode;
+ struct ata_link *link = qc->dev->link;
+ struct ata_port *ap = link->ap;
+ struct mv_port_priv *pp = ap->private_data;
+
+ /*
+ * Don't allow new commands if we're in a delayed EH state
+ * for NCQ and/or FIS-based switching.
+ */
+ if (pp->pp_flags & MV_PP_FLAG_DELAYED_EH)
+ return ATA_DEFER_PORT;
/*
- * Various bit settings required for operation
- * in FIS-based switching (fbs) mode on GenIIe:
+ * If the port is completely idle, then allow the new qc.
*/
- old_fcfg = readl(port_mmio + FIS_CFG_OFS);
- old_ltmode = readl(port_mmio + LTMODE_OFS);
- if (enable_fbs) {
- new_fcfg = old_fcfg | FIS_CFG_SINGLE_SYNC;
- new_ltmode = old_ltmode | LTMODE_BIT8;
- } else { /* disable fbs */
- new_fcfg = old_fcfg & ~FIS_CFG_SINGLE_SYNC;
- new_ltmode = old_ltmode & ~LTMODE_BIT8;
- }
- if (new_fcfg != old_fcfg)
- writelfl(new_fcfg, port_mmio + FIS_CFG_OFS);
+ if (ap->nr_active_links == 0)
+ return 0;
+
+ if (pp->pp_flags & MV_PP_FLAG_EDMA_EN) {
+ /*
+ * The port is operating in host queuing mode (EDMA).
+ * It can accomodate a new qc if the qc protocol
+ * is compatible with the current host queue mode.
+ */
+ if (pp->pp_flags & MV_PP_FLAG_NCQ_EN) {
+ /*
+ * The host queue (EDMA) is in NCQ mode.
+ * If the new qc is also an NCQ command,
+ * then allow the new qc.
+ */
+ if (qc->tf.protocol == ATA_PROT_NCQ)
+ return 0;
+ } else {
+ /*
+ * The host queue (EDMA) is in non-NCQ, DMA mode.
+ * If the new qc is also a non-NCQ, DMA command,
+ * then allow the new qc.
+ */
+ if (qc->tf.protocol == ATA_PROT_DMA)
+ return 0;
+ }
+ }
+ return ATA_DEFER_PORT;
+}
+
+static void mv_config_fbs(void __iomem *port_mmio, int want_ncq, int want_fbs)
+{
+ u32 new_fiscfg, old_fiscfg;
+ u32 new_ltmode, old_ltmode;
+ u32 new_haltcond, old_haltcond;
+
+ old_fiscfg = readl(port_mmio + FISCFG_OFS);
+ old_ltmode = readl(port_mmio + LTMODE_OFS);
+ old_haltcond = readl(port_mmio + EDMA_HALTCOND_OFS);
+
+ new_fiscfg = old_fiscfg & ~(FISCFG_SINGLE_SYNC | FISCFG_WAIT_DEV_ERR);
+ new_ltmode = old_ltmode & ~LTMODE_BIT8;
+ new_haltcond = old_haltcond | EDMA_ERR_DEV;
+
+ if (want_fbs) {
+ new_fiscfg = old_fiscfg | FISCFG_SINGLE_SYNC;
+ new_ltmode = old_ltmode | LTMODE_BIT8;
+ if (want_ncq)
+ new_haltcond &= ~EDMA_ERR_DEV;
+ else
+ new_fiscfg |= FISCFG_WAIT_DEV_ERR;
+ }
+
+ if (new_fiscfg != old_fiscfg)
+ writelfl(new_fiscfg, port_mmio + FISCFG_OFS);
if (new_ltmode != old_ltmode)
writelfl(new_ltmode, port_mmio + LTMODE_OFS);
+ if (new_haltcond != old_haltcond)
+ writelfl(new_haltcond, port_mmio + EDMA_HALTCOND_OFS);
+}
+
+static void mv_60x1_errata_sata25(struct ata_port *ap, int want_ncq)
+{
+ struct mv_host_priv *hpriv = ap->host->private_data;
+ u32 old, new;
+
+ /* workaround for 88SX60x1 FEr SATA#25 (part 1) */
+ old = readl(hpriv->base + MV_GPIO_PORT_CTL_OFS);
+ if (want_ncq)
+ new = old | (1 << 22);
+ else
+ new = old & ~(1 << 22);
+ if (new != old)
+ writel(new, hpriv->base + MV_GPIO_PORT_CTL_OFS);
}
static void mv_edma_cfg(struct ata_port *ap, int want_ncq)
@@ -1088,25 +1201,40 @@ static void mv_edma_cfg(struct ata_port *ap, int want_ncq)
/* set up non-NCQ EDMA configuration */
cfg = EDMA_CFG_Q_DEPTH; /* always 0x1f for *all* chips */
+ pp->pp_flags &= ~MV_PP_FLAG_FBS_EN;
if (IS_GEN_I(hpriv))
cfg |= (1 << 8); /* enab config burst size mask */
- else if (IS_GEN_II(hpriv))
+ else if (IS_GEN_II(hpriv)) {
cfg |= EDMA_CFG_RD_BRST_EXT | EDMA_CFG_WR_BUFF_LEN;
+ mv_60x1_errata_sata25(ap, want_ncq);
- else if (IS_GEN_IIE(hpriv)) {
- cfg |= (1 << 23); /* do not mask PM field in rx'd FIS */
- cfg |= (1 << 22); /* enab 4-entry host queue cache */
- cfg |= (1 << 18); /* enab early completion */
- cfg |= (1 << 17); /* enab cut-through (dis stor&forwrd) */
+ } else if (IS_GEN_IIE(hpriv)) {
+ int want_fbs = sata_pmp_attached(ap);
+ /*
+ * Possible future enhancement:
+ *
+ * The chip can use FBS with non-NCQ, if we allow it,
+ * But first we need to have the error handling in place
+ * for this mode (datasheet section 7.3.15.4.2.3).
+ * So disallow non-NCQ FBS for now.
+ */
+ want_fbs &= want_ncq;
+
+ mv_config_fbs(port_mmio, want_ncq, want_fbs);
- if (want_ncq && sata_pmp_attached(ap)) {
+ if (want_fbs) {
+ pp->pp_flags |= MV_PP_FLAG_FBS_EN;
cfg |= EDMA_CFG_EDMA_FBS; /* FIS-based switching */
- mv_config_fbs(port_mmio, 1);
- } else {
- mv_config_fbs(port_mmio, 0);
}
+
+ cfg |= (1 << 23); /* do not mask PM field in rx'd FIS */
+ cfg |= (1 << 22); /* enab 4-entry host queue cache */
+ if (HAS_PCI(ap->host))
+ cfg |= (1 << 18); /* enab early completion */
+ if (hpriv->hp_flags & MV_HP_CUT_THROUGH)
+ cfg |= (1 << 17); /* enab cut-thru (dis stor&forwrd) */
}
if (want_ncq) {
@@ -1483,25 +1611,186 @@ static struct ata_queued_cmd *mv_get_active_qc(struct ata_port *ap)
return qc;
}
-static void mv_unexpected_intr(struct ata_port *ap)
+static void mv_pmp_error_handler(struct ata_port *ap)
{
+ unsigned int pmp, pmp_map;
struct mv_port_priv *pp = ap->private_data;
- struct ata_eh_info *ehi = &ap->link.eh_info;
- char *when = "";
+ if (pp->pp_flags & MV_PP_FLAG_DELAYED_EH) {
+ /*
+ * Perform NCQ error analysis on failed PMPs
+ * before we freeze the port entirely.
+ *
+ * The failed PMPs are marked earlier by mv_pmp_eh_prep().
+ */
+ pmp_map = pp->delayed_eh_pmp_map;
+ pp->pp_flags &= ~MV_PP_FLAG_DELAYED_EH;
+ for (pmp = 0; pmp_map != 0; pmp++) {
+ unsigned int this_pmp = (1 << pmp);
+ if (pmp_map & this_pmp) {
+ struct ata_link *link = &ap->pmp_link[pmp];
+ pmp_map &= ~this_pmp;
+ ata_eh_analyze_ncq_error(link);
+ }
+ }
+ ata_port_freeze(ap);
+ }
+ sata_pmp_error_handler(ap);
+}
+
+static unsigned int mv_get_err_pmp_map(struct ata_port *ap)
+{
+ void __iomem *port_mmio = mv_ap_base(ap);
+
+ return readl(port_mmio + SATA_TESTCTL_OFS) >> 16;
+}
+
+static void mv_pmp_eh_prep(struct ata_port *ap, unsigned int pmp_map)
+{
+ struct ata_eh_info *ehi;
+ unsigned int pmp;
+
+ /*
+ * Initialize EH info for PMPs which saw device errors
+ */
+ ehi = &ap->link.eh_info;
+ for (pmp = 0; pmp_map != 0; pmp++) {
+ unsigned int this_pmp = (1 << pmp);
+ if (pmp_map & this_pmp) {
+ struct ata_link *link = &ap->pmp_link[pmp];
+
+ pmp_map &= ~this_pmp;
+ ehi = &link->eh_info;
+ ata_ehi_clear_desc(ehi);
+ ata_ehi_push_desc(ehi, "dev err");
+ ehi->err_mask |= AC_ERR_DEV;
+ ehi->action |= ATA_EH_RESET;
+ ata_link_abort(link);
+ }
+ }
+}
+
+static int mv_handle_fbs_ncq_dev_err(struct ata_port *ap)
+{
+ struct mv_port_priv *pp = ap->private_data;
+ int failed_links;
+ unsigned int old_map, new_map;
+
+ /*
+ * Device error during FBS+NCQ operation:
+ *
+ * Set a port flag to prevent further I/O being enqueued.
+ * Leave the EDMA running to drain outstanding commands from this port.
+ * Perform the post-mortem/EH only when all responses are complete.
+ * Follow recovery sequence from 6042/7042 datasheet (7.3.15.4.2.2).
+ */
+ if (!(pp->pp_flags & MV_PP_FLAG_DELAYED_EH)) {
+ pp->pp_flags |= MV_PP_FLAG_DELAYED_EH;
+ pp->delayed_eh_pmp_map = 0;
+ }
+ old_map = pp->delayed_eh_pmp_map;
+ new_map = old_map | mv_get_err_pmp_map(ap);
+
+ if (old_map != new_map) {
+ pp->delayed_eh_pmp_map = new_map;
+ mv_pmp_eh_prep(ap, new_map & ~old_map);
+ }
+ failed_links = hweight16(new_map);
+
+ ata_port_printk(ap, KERN_INFO, "%s: pmp_map=%04x qc_map=%04x "
+ "failed_links=%d nr_active_links=%d\n",
+ __func__, pp->delayed_eh_pmp_map,
+ ap->qc_active, failed_links,
+ ap->nr_active_links);
+
+ if (ap->nr_active_links <= failed_links) {
+ mv_process_crpb_entries(ap, pp);
+ mv_stop_edma(ap);
+ mv_eh_freeze(ap);
+ ata_port_printk(ap, KERN_INFO, "%s: done\n", __func__);
+ return 1; /* handled */
+ }
+ ata_port_printk(ap, KERN_INFO, "%s: waiting\n", __func__);
+ return 1; /* handled */
+}
+
+static int mv_handle_fbs_non_ncq_dev_err(struct ata_port *ap)
+{
/*
- * We got a device interrupt from something that
- * was supposed to be using EDMA or polling.
+ * Possible future enhancement:
+ *
+ * FBS+non-NCQ operation is not yet implemented.
+ * See related notes in mv_edma_cfg().
+ *
+ * Device error during FBS+non-NCQ operation:
+ *
+ * We need to snapshot the shadow registers for each failed command.
+ * Follow recovery sequence from 6042/7042 datasheet (7.3.15.4.2.3).
*/
+ return 0; /* not handled */
+}
+
+static int mv_handle_dev_err(struct ata_port *ap, u32 edma_err_cause)
+{
+ struct mv_port_priv *pp = ap->private_data;
+
+ if (!(pp->pp_flags & MV_PP_FLAG_EDMA_EN))
+ return 0; /* EDMA was not active: not handled */
+ if (!(pp->pp_flags & MV_PP_FLAG_FBS_EN))
+ return 0; /* FBS was not active: not handled */
+
+ if (!(edma_err_cause & EDMA_ERR_DEV))
+ return 0; /* non DEV error: not handled */
+ edma_err_cause &= ~EDMA_ERR_IRQ_TRANSIENT;
+ if (edma_err_cause & ~(EDMA_ERR_DEV | EDMA_ERR_SELF_DIS))
+ return 0; /* other problems: not handled */
+
+ if (pp->pp_flags & MV_PP_FLAG_NCQ_EN) {
+ /*
+ * EDMA should NOT have self-disabled for this case.
+ * If it did, then something is wrong elsewhere,
+ * and we cannot handle it here.
+ */
+ if (edma_err_cause & EDMA_ERR_SELF_DIS) {
+ ata_port_printk(ap, KERN_WARNING,
+ "%s: err_cause=0x%x pp_flags=0x%x\n",
+ __func__, edma_err_cause, pp->pp_flags);
+ return 0; /* not handled */
+ }
+ return mv_handle_fbs_ncq_dev_err(ap);
+ } else {
+ /*
+ * EDMA should have self-disabled for this case.
+ * If it did not, then something is wrong elsewhere,
+ * and we cannot handle it here.
+ */
+ if (!(edma_err_cause & EDMA_ERR_SELF_DIS)) {
+ ata_port_printk(ap, KERN_WARNING,
+ "%s: err_cause=0x%x pp_flags=0x%x\n",
+ __func__, edma_err_cause, pp->pp_flags);
+ return 0; /* not handled */
+ }
+ return mv_handle_fbs_non_ncq_dev_err(ap);
+ }
+ return 0; /* not handled */
+}
+
+static void mv_unexpected_intr(struct ata_port *ap, int edma_was_enabled)
+{
+ struct ata_eh_info *ehi = &ap->link.eh_info;
+ char *when = "idle";
+
ata_ehi_clear_desc(ehi);
- if (pp->pp_flags & MV_PP_FLAG_EDMA_EN) {
- when = " while EDMA enabled";
+ if (!ap || (ap->flags & ATA_FLAG_DISABLED)) {
+ when = "disabled";
+ } else if (edma_was_enabled) {
+ when = "EDMA enabled";
} else {
struct ata_queued_cmd *qc = ata_qc_from_tag(ap, ap->link.active_tag);
if (qc && (qc->tf.flags & ATA_TFLAG_POLLING))
- when = " while polling";
+ when = "polling";
}
- ata_ehi_push_desc(ehi, "unexpected device interrupt%s", when);
+ ata_ehi_push_desc(ehi, "unexpected device interrupt while %s", when);
ehi->err_mask |= AC_ERR_OTHER;
ehi->action |= ATA_EH_RESET;
ata_port_freeze(ap);
@@ -1519,7 +1808,7 @@ static void mv_unexpected_intr(struct ata_port *ap)
* LOCKING:
* Inherited from caller.
*/
-static void mv_err_intr(struct ata_port *ap, struct ata_queued_cmd *qc)
+static void mv_err_intr(struct ata_port *ap)
{
void __iomem *port_mmio = mv_ap_base(ap);
u32 edma_err_cause, eh_freeze_mask, serr = 0;
@@ -1527,24 +1816,42 @@ static void mv_err_intr(struct ata_port *ap, struct ata_queued_cmd *qc)
struct mv_host_priv *hpriv = ap->host->private_data;
unsigned int action = 0, err_mask = 0;
struct ata_eh_info *ehi = &ap->link.eh_info;
-
- ata_ehi_clear_desc(ehi);
+ struct ata_queued_cmd *qc;
+ int abort = 0;
/*
- * Read and clear the err_cause bits. This won't actually
- * clear for some errors (eg. SError), but we will be doing
- * a hard reset in those cases regardless, which *will* clear it.
+ * Read and clear the SError and err_cause bits.
*/
+ sata_scr_read(&ap->link, SCR_ERROR, &serr);
+ sata_scr_write_flush(&ap->link, SCR_ERROR, serr);
+
edma_err_cause = readl(port_mmio + EDMA_ERR_IRQ_CAUSE_OFS);
writelfl(~edma_err_cause, port_mmio + EDMA_ERR_IRQ_CAUSE_OFS);
- ata_ehi_push_desc(ehi, "edma_err_cause=%08x", edma_err_cause);
+ ata_port_printk(ap, KERN_INFO, "%s: err_cause=%08x pp_flags=0x%x\n",
+ __func__, edma_err_cause, pp->pp_flags);
+
+ if (edma_err_cause & EDMA_ERR_DEV) {
+ /*
+ * Device errors during FIS-based switching operation
+ * require special handling.
+ */
+ if (mv_handle_dev_err(ap, edma_err_cause))
+ return;
+ }
+ qc = mv_get_active_qc(ap);
+ ata_ehi_clear_desc(ehi);
+ ata_ehi_push_desc(ehi, "edma_err_cause=%08x pp_flags=%08x",
+ edma_err_cause, pp->pp_flags);
/*
* All generations share these EDMA error cause bits:
*/
- if (edma_err_cause & EDMA_ERR_DEV)
+ if (edma_err_cause & EDMA_ERR_DEV) {
err_mask |= AC_ERR_DEV;
+ action |= ATA_EH_RESET;
+ ata_ehi_push_desc(ehi, "dev error");
+ }
if (edma_err_cause & (EDMA_ERR_D_PAR | EDMA_ERR_PRD_PAR |
EDMA_ERR_CRQB_PAR | EDMA_ERR_CRPB_PAR |
EDMA_ERR_INTRL_PAR)) {
@@ -1576,13 +1883,6 @@ static void mv_err_intr(struct ata_port *ap, struct ata_queued_cmd *qc)
ata_ehi_push_desc(ehi, "EDMA self-disable");
}
if (edma_err_cause & EDMA_ERR_SERR) {
- /*
- * Ensure that we read our own SCR, not a pmp link SCR:
- */
- ap->ops->scr_read(ap, SCR_ERROR, &serr);
- /*
- * Don't clear SError here; leave it for libata-eh:
- */
ata_ehi_push_desc(ehi, "SError=%08x", serr);
err_mask |= AC_ERR_ATA_BUS;
action |= ATA_EH_RESET;
@@ -1602,10 +1902,29 @@ static void mv_err_intr(struct ata_port *ap, struct ata_queued_cmd *qc)
else
ehi->err_mask |= err_mask;
- if (edma_err_cause & eh_freeze_mask)
+ if (err_mask == AC_ERR_DEV) {
+ /*
+ * Cannot do ata_port_freeze() here,
+ * because it would kill PIO access,
+ * which is needed for further diagnosis.
+ */
+ mv_eh_freeze(ap);
+ abort = 1;
+ } else if (edma_err_cause & eh_freeze_mask) {
+ /*
+ * Note to self: ata_port_freeze() calls ata_port_abort()
+ */
ata_port_freeze(ap);
- else
- ata_port_abort(ap);
+ } else {
+ abort = 1;
+ }
+
+ if (abort) {
+ if (qc)
+ ata_link_abort(qc->dev->link);
+ else
+ ata_port_abort(ap);
+ }
}
static void mv_process_crpb_response(struct ata_port *ap,
@@ -1632,8 +1951,9 @@ static void mv_process_crpb_response(struct ata_port *ap,
}
}
ata_status = edma_status >> CRPB_FLAG_STATUS_SHIFT;
- qc->err_mask |= ac_err_mask(ata_status);
- ata_qc_complete(qc);
+ if (!ac_err_mask(ata_status))
+ ata_qc_complete(qc);
+ /* else: leave it for mv_err_intr() */
} else {
ata_port_printk(ap, KERN_ERR, "%s: no qc for tag=%d\n",
__func__, tag);
@@ -1677,6 +1997,44 @@ static void mv_process_crpb_entries(struct ata_port *ap, struct mv_port_priv *pp
port_mmio + EDMA_RSP_Q_OUT_PTR_OFS);
}
+static void mv_port_intr(struct ata_port *ap, u32 port_cause)
+{
+ struct mv_port_priv *pp;
+ int edma_was_enabled;
+
+ if (!ap || (ap->flags & ATA_FLAG_DISABLED)) {
+ mv_unexpected_intr(ap, 0);
+ return;
+ }
+ /*
+ * Grab a snapshot of the EDMA_EN flag setting,
+ * so that we have a consistent view for this port,
+ * even if something we call of our routines changes it.
+ */
+ pp = ap->private_data;
+ edma_was_enabled = (pp->pp_flags & MV_PP_FLAG_EDMA_EN);
+ /*
+ * Process completed CRPB response(s) before other events.
+ */
+ if (edma_was_enabled && (port_cause & DONE_IRQ)) {
+ mv_process_crpb_entries(ap, pp);
+ if (pp->pp_flags & MV_PP_FLAG_DELAYED_EH)
+ mv_handle_fbs_ncq_dev_err(ap);
+ }
+ /*
+ * Handle chip-reported errors, or continue on to handle PIO.
+ */
+ if (unlikely(port_cause & ERR_IRQ)) {
+ mv_err_intr(ap);
+ } else if (!edma_was_enabled) {
+ struct ata_queued_cmd *qc = mv_get_active_qc(ap);
+ if (qc)
+ ata_sff_host_intr(ap, qc);
+ else
+ mv_unexpected_intr(ap, edma_was_enabled);
+ }
+}
+
/**
* mv_host_intr - Handle all interrupts on the given host controller
* @host: host specific structure
@@ -1688,66 +2046,58 @@ static void mv_process_crpb_entries(struct ata_port *ap, struct mv_port_priv *pp
static int mv_host_intr(struct ata_host *host, u32 main_irq_cause)
{
struct mv_host_priv *hpriv = host->private_data;
- void __iomem *mmio = hpriv->base, *hc_mmio = NULL;
- u32 hc_irq_cause = 0;
+ void __iomem *mmio = hpriv->base, *hc_mmio;
unsigned int handled = 0, port;
for (port = 0; port < hpriv->n_ports; port++) {
struct ata_port *ap = host->ports[port];
- struct mv_port_priv *pp;
- unsigned int shift, hardport, port_cause;
- /*
- * When we move to the second hc, flag our cached
- * copies of hc_mmio (and hc_irq_cause) as invalid again.
- */
- if (port == MV_PORTS_PER_HC)
- hc_mmio = NULL;
- /*
- * Do nothing if port is not interrupting or is disabled:
- */
+ unsigned int p, shift, hardport, port_cause;
+
MV_PORT_TO_SHIFT_AND_HARDPORT(port, shift, hardport);
- port_cause = (main_irq_cause >> shift) & (DONE_IRQ | ERR_IRQ);
- if (!port_cause || !ap || (ap->flags & ATA_FLAG_DISABLED))
- continue;
/*
- * Each hc within the host has its own hc_irq_cause register.
- * We defer reading it until we know we need it, right now:
- *
- * FIXME later: we don't really need to read this register
- * (some logic changes required below if we go that way),
- * because it doesn't tell us anything new. But we do need
- * to write to it, outside the top of this loop,
- * to reset the interrupt triggers for next time.
+ * Each hc within the host has its own hc_irq_cause register,
+ * where the interrupting ports bits get ack'd.
*/
- if (!hc_mmio) {
+ if (hardport == 0) { /* first port on this hc ? */
+ u32 hc_cause = (main_irq_cause >> shift) & HC0_IRQ_PEND;
+ u32 port_mask, ack_irqs;
+ /*
+ * Skip this entire hc if nothing pending for any ports
+ */
+ if (!hc_cause) {
+ port += MV_PORTS_PER_HC - 1;
+ continue;
+ }
+ /*
+ * We don't need/want to read the hc_irq_cause register,
+ * because doing so hurts performance, and
+ * main_irq_cause already gives us everything we need.
+ *
+ * But we do have to *write* to the hc_irq_cause to ack
+ * the ports that we are handling this time through.
+ *
+ * This requires that we create a bitmap for those
+ * ports which interrupted us, and use that bitmap
+ * to ack (only) those ports via hc_irq_cause.
+ */
+ ack_irqs = 0;
+ for (p = 0; p < MV_PORTS_PER_HC; ++p) {
+ if ((port + p) >= hpriv->n_ports)
+ break;
+ port_mask = (DONE_IRQ | ERR_IRQ) << (p * 2);
+ if (hc_cause & port_mask)
+ ack_irqs |= (DMA_IRQ | DEV_IRQ) << p;
+ }
hc_mmio = mv_hc_base_from_port(mmio, port);
- hc_irq_cause = readl(hc_mmio + HC_IRQ_CAUSE_OFS);
- writelfl(~hc_irq_cause, hc_mmio + HC_IRQ_CAUSE_OFS);
+ writelfl(~ack_irqs, hc_mmio + HC_IRQ_CAUSE_OFS);
handled = 1;
}
/*
- * Process completed CRPB response(s) before other events.
- */
- pp = ap->private_data;
- if (hc_irq_cause & (DMA_IRQ << hardport)) {
- if (pp->pp_flags & MV_PP_FLAG_EDMA_EN)
- mv_process_crpb_entries(ap, pp);
- }
- /*
- * Handle chip-reported errors, or continue on to handle PIO.
+ * Handle interrupts signalled for this port:
*/
- if (unlikely(port_cause & ERR_IRQ)) {
- mv_err_intr(ap, mv_get_active_qc(ap));
- } else if (hc_irq_cause & (DEV_IRQ << hardport)) {
- if (!(pp->pp_flags & MV_PP_FLAG_EDMA_EN)) {
- struct ata_queued_cmd *qc = mv_get_active_qc(ap);
- if (qc) {
- ata_sff_host_intr(ap, qc);
- continue;
- }
- }
- mv_unexpected_intr(ap);
- }
+ port_cause = (main_irq_cause >> shift) & (DONE_IRQ | ERR_IRQ);
+ if (port_cause)
+ mv_port_intr(ap, port_cause);
}
return handled;
}
@@ -1894,7 +2244,7 @@ static void mv5_reset_bus(struct ata_host *host, void __iomem *mmio)
static void mv5_reset_flash(struct mv_host_priv *hpriv, void __iomem *mmio)
{
- writel(0x0fcfffff, mmio + MV_FLASH_CTL);
+ writel(0x0fcfffff, mmio + MV_FLASH_CTL_OFS);
}
static void mv5_read_preamp(struct mv_host_priv *hpriv, int idx,
@@ -1913,7 +2263,7 @@ static void mv5_enable_leds(struct mv_host_priv *hpriv, void __iomem *mmio)
{
u32 tmp;
- writel(0, mmio + MV_GPIO_PORT_CTL);
+ writel(0, mmio + MV_GPIO_PORT_CTL_OFS);
/* FIXME: handle MV_HP_ERRATA_50XXB2 errata */
@@ -1931,14 +2281,14 @@ static void mv5_phy_errata(struct mv_host_priv *hpriv, void __iomem *mmio,
int fix_apm_sq = (hpriv->hp_flags & MV_HP_ERRATA_50XXB0);
if (fix_apm_sq) {
- tmp = readl(phy_mmio + MV5_LT_MODE);
+ tmp = readl(phy_mmio + MV5_LTMODE_OFS);
tmp |= (1 << 19);
- writel(tmp, phy_mmio + MV5_LT_MODE);
+ writel(tmp, phy_mmio + MV5_LTMODE_OFS);
- tmp = readl(phy_mmio + MV5_PHY_CTL);
+ tmp = readl(phy_mmio + MV5_PHY_CTL_OFS);
tmp &= ~0x3;
tmp |= 0x1;
- writel(tmp, phy_mmio + MV5_PHY_CTL);
+ writel(tmp, phy_mmio + MV5_PHY_CTL_OFS);
}
tmp = readl(phy_mmio + MV5_PHY_MODE);
@@ -1956,11 +2306,6 @@ static void mv5_reset_hc_port(struct mv_host_priv *hpriv, void __iomem *mmio,
{
void __iomem *port_mmio = mv_port_base(mmio, port);
- /*
- * The datasheet warns against setting ATA_RST when EDMA is active
- * (but doesn't say what the problem might be). So we first try
- * to disable the EDMA engine before doing the ATA_RST operation.
- */
mv_reset_channel(hpriv, mmio, port);
ZERO(0x028); /* command */
@@ -1975,7 +2320,7 @@ static void mv5_reset_hc_port(struct mv_host_priv *hpriv, void __iomem *mmio,
ZERO(0x024); /* respq outp */
ZERO(0x020); /* respq inp */
ZERO(0x02c); /* test control */
- writel(0xbc, port_mmio + EDMA_IORDY_TMOUT);
+ writel(0xbc, port_mmio + EDMA_IORDY_TMOUT_OFS);
}
#undef ZERO
@@ -2021,13 +2366,13 @@ static void mv_reset_pci_bus(struct ata_host *host, void __iomem *mmio)
struct mv_host_priv *hpriv = host->private_data;
u32 tmp;
- tmp = readl(mmio + MV_PCI_MODE);
+ tmp = readl(mmio + MV_PCI_MODE_OFS);
tmp &= 0xff00ffff;
- writel(tmp, mmio + MV_PCI_MODE);
+ writel(tmp, mmio + MV_PCI_MODE_OFS);
ZERO(MV_PCI_DISC_TIMER);
ZERO(MV_PCI_MSI_TRIGGER);
- writel(0x000100ff, mmio + MV_PCI_XBAR_TMOUT);
+ writel(0x000100ff, mmio + MV_PCI_XBAR_TMOUT_OFS);
ZERO(PCI_HC_MAIN_IRQ_MASK_OFS);
ZERO(MV_PCI_SERR_MASK);
ZERO(hpriv->irq_cause_ofs);
@@ -2045,10 +2390,10 @@ static void mv6_reset_flash(struct mv_host_priv *hpriv, void __iomem *mmio)
mv5_reset_flash(hpriv, mmio);
- tmp = readl(mmio + MV_GPIO_PORT_CTL);
+ tmp = readl(mmio + MV_GPIO_PORT_CTL_OFS);
tmp &= 0x3;
tmp |= (1 << 5) | (1 << 6);
- writel(tmp, mmio + MV_GPIO_PORT_CTL);
+ writel(tmp, mmio + MV_GPIO_PORT_CTL_OFS);
}
/**
@@ -2121,7 +2466,7 @@ static void mv6_read_preamp(struct mv_host_priv *hpriv, int idx,
void __iomem *port_mmio;
u32 tmp;
- tmp = readl(mmio + MV_RESET_CFG);
+ tmp = readl(mmio + MV_RESET_CFG_OFS);
if ((tmp & (1 << 0)) == 0) {
hpriv->signal[idx].amps = 0x7 << 8;
hpriv->signal[idx].pre = 0x1 << 5;
@@ -2137,7 +2482,7 @@ static void mv6_read_preamp(struct mv_host_priv *hpriv, int idx,
static void mv6_enable_leds(struct mv_host_priv *hpriv, void __iomem *mmio)
{
- writel(0x00000060, mmio + MV_GPIO_PORT_CTL);
+ writel(0x00000060, mmio + MV_GPIO_PORT_CTL_OFS);
}
static void mv6_phy_errata(struct mv_host_priv *hpriv, void __iomem *mmio,
@@ -2235,11 +2580,6 @@ static void mv_soc_reset_hc_port(struct mv_host_priv *hpriv,
{
void __iomem *port_mmio = mv_port_base(mmio, port);
- /*
- * The datasheet warns against setting ATA_RST when EDMA is active
- * (but doesn't say what the problem might be). So we first try
- * to disable the EDMA engine before doing the ATA_RST operation.
- */
mv_reset_channel(hpriv, mmio, port);
ZERO(0x028); /* command */
@@ -2254,7 +2594,7 @@ static void mv_soc_reset_hc_port(struct mv_host_priv *hpriv,
ZERO(0x024); /* respq outp */
ZERO(0x020); /* respq inp */
ZERO(0x02c); /* test control */
- writel(0xbc, port_mmio + EDMA_IORDY_TMOUT);
+ writel(0xbc, port_mmio + EDMA_IORDY_TMOUT_OFS);
}
#undef ZERO
@@ -2297,38 +2637,39 @@ static void mv_soc_reset_bus(struct ata_host *host, void __iomem *mmio)
return;
}
-static void mv_setup_ifctl(void __iomem *port_mmio, int want_gen2i)
+static void mv_setup_ifcfg(void __iomem *port_mmio, int want_gen2i)
{
- u32 ifctl = readl(port_mmio + SATA_INTERFACE_CFG);
+ u32 ifcfg = readl(port_mmio + SATA_INTERFACE_CFG_OFS);
- ifctl = (ifctl & 0xf7f) | 0x9b1000; /* from chip spec */
+ ifcfg = (ifcfg & 0xf7f) | 0x9b1000; /* from chip spec */
if (want_gen2i)
- ifctl |= (1 << 7); /* enable gen2i speed */
- writelfl(ifctl, port_mmio + SATA_INTERFACE_CFG);
+ ifcfg |= (1 << 7); /* enable gen2i speed */
+ writelfl(ifcfg, port_mmio + SATA_INTERFACE_CFG_OFS);
}
-/*
- * Caller must ensure that EDMA is not active,
- * by first doing mv_stop_edma() where needed.
- */
static void mv_reset_channel(struct mv_host_priv *hpriv, void __iomem *mmio,
unsigned int port_no)
{
void __iomem *port_mmio = mv_port_base(mmio, port_no);
+ /*
+ * The datasheet warns against setting EDMA_RESET when EDMA is active
+ * (but doesn't say what the problem might be). So we first try
+ * to disable the EDMA engine before doing the EDMA_RESET operation.
+ */
mv_stop_edma_engine(port_mmio);
- writelfl(ATA_RST, port_mmio + EDMA_CMD_OFS);
+ writelfl(EDMA_RESET, port_mmio + EDMA_CMD_OFS);
if (!IS_GEN_I(hpriv)) {
- /* Enable 3.0gb/s link speed */
- mv_setup_ifctl(port_mmio, 1);
+ /* Enable 3.0gb/s link speed: this survives EDMA_RESET */
+ mv_setup_ifcfg(port_mmio, 1);
}
/*
- * Strobing ATA_RST here causes a hard reset of the SATA transport,
+ * Strobing EDMA_RESET here causes a hard reset of the SATA transport,
* link, and physical layers. It resets all SATA interface registers
* (except for SATA_INTERFACE_CFG), and issues a COMRESET to the dev.
*/
- writelfl(ATA_RST, port_mmio + EDMA_CMD_OFS);
+ writelfl(EDMA_RESET, port_mmio + EDMA_CMD_OFS);
udelay(25); /* allow reset propagation */
writelfl(0, port_mmio + EDMA_CMD_OFS);
@@ -2392,7 +2733,7 @@ static int mv_hardreset(struct ata_link *link, unsigned int *class,
sata_scr_read(link, SCR_STATUS, &sstatus);
if (!IS_GEN_I(hpriv) && ++attempts >= 5 && sstatus == 0x121) {
/* Force 1.5gb/s link speed and try again */
- mv_setup_ifctl(mv_ap_base(ap), 0);
+ mv_setup_ifcfg(mv_ap_base(ap), 0);
if (time_after(jiffies + HZ, deadline))
extra = HZ; /* only extend it once, max */
}
@@ -2493,6 +2834,34 @@ static void mv_port_init(struct ata_ioports *port, void __iomem *port_mmio)
readl(port_mmio + EDMA_ERR_IRQ_MASK_OFS));
}
+static unsigned int mv_in_pcix_mode(struct ata_host *host)
+{
+ struct mv_host_priv *hpriv = host->private_data;
+ void __iomem *mmio = hpriv->base;
+ u32 reg;
+
+ if (!HAS_PCI(host) || !IS_PCIE(hpriv))
+ return 0; /* not PCI-X capable */
+ reg = readl(mmio + MV_PCI_MODE_OFS);
+ if ((reg & MV_PCI_MODE_MASK) == 0)
+ return 0; /* conventional PCI mode */
+ return 1; /* chip is in PCI-X mode */
+}
+
+static int mv_pci_cut_through_okay(struct ata_host *host)
+{
+ struct mv_host_priv *hpriv = host->private_data;
+ void __iomem *mmio = hpriv->base;
+ u32 reg;
+
+ if (!mv_in_pcix_mode(host)) {
+ reg = readl(mmio + PCI_COMMAND_OFS);
+ if (reg & PCI_COMMAND_MRDTRIG)
+ return 0; /* not okay */
+ }
+ return 1; /* okay */
+}
+
static int mv_chip_id(struct ata_host *host, unsigned int board_idx)
{
struct pci_dev *pdev = to_pci_dev(host->dev);
@@ -2560,7 +2929,7 @@ static int mv_chip_id(struct ata_host *host, unsigned int board_idx)
break;
case chip_7042:
- hp_flags |= MV_HP_PCIE;
+ hp_flags |= MV_HP_PCIE | MV_HP_CUT_THROUGH;
if (pdev->vendor == PCI_VENDOR_ID_TTI &&
(pdev->device == 0x2300 || pdev->device == 0x2310))
{
@@ -2590,9 +2959,12 @@ static int mv_chip_id(struct ata_host *host, unsigned int board_idx)
" and avoid the final two gigabytes on"
" all RocketRAID BIOS initialized drives.\n");
}
+ /* drop through */
case chip_6042:
hpriv->ops = &mv6xxx_ops;
hp_flags |= MV_HP_GEN_IIE;
+ if (board_idx == chip_6042 && mv_pci_cut_through_okay(host))
+ hp_flags |= MV_HP_CUT_THROUGH;
switch (pdev->revision) {
case 0x0:
diff --git a/include/linux/libata.h b/include/linux/libata.h
index d1dfe87..7e206da 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -1039,6 +1039,7 @@ extern void ata_eh_thaw_port(struct ata_port *ap);
extern void ata_eh_qc_complete(struct ata_queued_cmd *qc);
extern void ata_eh_qc_retry(struct ata_queued_cmd *qc);
+extern void ata_eh_analyze_ncq_error(struct ata_link *link);
extern void ata_do_eh(struct ata_port *ap, ata_prereset_fn_t prereset,
ata_reset_fn_t softreset, ata_reset_fn_t hardreset,
@@ -1381,6 +1382,21 @@ static inline struct ata_port *ata_shost_to_port(struct Scsi_Host *host)
return *(struct ata_port **)&host->hostdata[0];
}
+static inline int ata_check_ready(u8 status)
+{
+ /* Some controllers report 0x77 or 0x7f during intermediate
+ * not-ready stages.
+ */
+ if (status == 0x77 || status == 0x7f)
+ return 0;
+
+ /* 0xff indicates either no device or device not ready */
+ if (status == 0xff)
+ return -ENODEV;
+
+ return !(status & ATA_BUSY);
+}
+
/**************************************************************************
* PMP - drivers/ata/libata-pmp.c
^ permalink raw reply related [flat|nested] 22+ messages in thread
* Re: [git patches] libata updates
2008-05-06 15:48 [git patches] libata updates Jeff Garzik
@ 2008-05-06 16:23 ` Linus Torvalds
2008-05-07 14:56 ` [git patches] libata updates - (improve post-reset device ready test) regression Markus Trippelsdorf
1 sibling, 0 replies; 22+ messages in thread
From: Linus Torvalds @ 2008-05-06 16:23 UTC (permalink / raw)
To: Jeff Garzik; +Cc: Andrew Morton, linux-ide, LKML
On Tue, 6 May 2008, Jeff Garzik wrote:
>
> +config PATA_SCH
> + tristate "Intel SCH PATA support"
> + depends on PCI
> + help
> + This option enables support for Intel SCH PATA on the Intel
> + SCH (US15W, US15L, UL11L) series host controllers.
> +
> + If unsure, say N.
I think it would generally be good if people explained their drivers
better than this.
If you know what the hell an SCH is, you don't *need* that help message in
the first place! Agreed?
IOW, when adding drivers, please add help messages that actually explain
what the driver is from a *user* perspective, not from a chip designer
perspective. In this case, something like
The Intel SCH chipset is the new low-power, small-footprint
embedded chipset that is generally paired with the Atom CPU and is
not their normal PC chipset.
or something. So that people can actually make a somewhat informed guess
about whether it's relevant to them or not.
Linus
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-06 15:48 [git patches] libata updates Jeff Garzik
2008-05-06 16:23 ` Linus Torvalds
@ 2008-05-07 14:56 ` Markus Trippelsdorf
2008-05-07 16:35 ` Takashi Iwai
2008-05-08 15:35 ` [PATCH] " Jeff Garzik
1 sibling, 2 replies; 22+ messages in thread
From: Markus Trippelsdorf @ 2008-05-07 14:56 UTC (permalink / raw)
To: Jeff Garzik; +Cc: Andrew Morton, Linus Torvalds, linux-ide, LKML, htejun
>
> Tejun Heo (12):
> libata: improve post-reset device ready test
This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
boot delay with my onboard Promise controller. It seems like libata
probes for a nonexisting PATA drive...
ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
scsi0 : sata_promise
scsi1 : sata_promise
scsi2 : sata_promise
ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
ata1: SATA link down (SStatus 0 SControl 300)
ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
ata2.00: configured for UDMA/133
ata3: link is slow to respond, please be patient (ready=0)
ata3: device not ready (errno=-16), forcing hardreset
ata3: link is slow to respond, please be patient (ready=0)
ata3: SRST failed (errno=-16)
ata3: link is slow to respond, please be patient (ready=0)
ata3: SRST failed (errno=-16)
ata3: link is slow to respond, please be patient (ready=0)
ata3: SRST failed (errno=-16)
- Last output repeated twice -
ata3: reset failed, giving up
--
Markus
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-07 14:56 ` [git patches] libata updates - (improve post-reset device ready test) regression Markus Trippelsdorf
@ 2008-05-07 16:35 ` Takashi Iwai
2008-05-08 3:13 ` Marc Dionne
2008-05-08 9:45 ` Jan-Simon Möller
2008-05-08 15:35 ` [PATCH] " Jeff Garzik
1 sibling, 2 replies; 22+ messages in thread
From: Takashi Iwai @ 2008-05-07 16:35 UTC (permalink / raw)
To: Markus Trippelsdorf
Cc: Jeff Garzik, Andrew Morton, Linus Torvalds, linux-ide, LKML, htejun
At Wed, 7 May 2008 16:56:16 +0200,
Markus Trippelsdorf wrote:
>
> >
> > Tejun Heo (12):
> > libata: improve post-reset device ready test
>
> This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
> boot delay with my onboard Promise controller. It seems like libata
> probes for a nonexisting PATA drive...
>
> ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
> scsi0 : sata_promise
> scsi1 : sata_promise
> scsi2 : sata_promise
> ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
> ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
> ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
> ata1: SATA link down (SStatus 0 SControl 300)
> ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
> ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
> ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
> ata2.00: configured for UDMA/133
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: device not ready (errno=-16), forcing hardreset
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: SRST failed (errno=-16)
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: SRST failed (errno=-16)
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: SRST failed (errno=-16)
> - Last output repeated twice -
> ata3: reset failed, giving up
The same problem appears on my machine, too:
libata version 3.00 loaded.
ata_piix 0000:00:1f.2: version 2.12
ACPI: PCI Interrupt 0000:00:1f.2[B] -> GSI 19 (level, low) -> IRQ 19
ata_piix 0000:00:1f.2: MAP [ P0 P2 P1 P3 ]
PCI: Setting latency timer of device 0000:00:1f.2 to 64
scsi0 : ata_piix
scsi1 : ata_piix
ata1: SATA max UDMA/133 cmd 0xe400 ctl 0xe500 bmdma 0xe800 irq 19
ata2: SATA max UDMA/133 cmd 0xe600 ctl 0xe700 bmdma 0xe808 irq 19
ata1.00: ATA-7: HDS728080PLA380, PF2OA60A, max UDMA/133
ata1.00: 160836480 sectors, multi 16: LBA48 NCQ (depth 0/32)
ata1.00: configured for UDMA/133
ata2: link is slow to respond, please be patient (ready=0)
ata2: device not ready (errno=-16), forcing hardreset
ata2: link is slow to respond, please be patient (ready=0)
ata2: SRST failed (errno=-16)
ata2: link is slow to respond, please be patient (ready=0)
ata2: SRST failed (errno=-16)
ata2: link is slow to respond, please be patient (ready=0)
ata2: SRST failed (errno=-16)
ata2: SRST failed (errno=-16)
ata2: reset failed, giving up
Takashi
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-07 16:35 ` Takashi Iwai
@ 2008-05-08 3:13 ` Marc Dionne
2008-05-08 9:45 ` Jan-Simon Möller
1 sibling, 0 replies; 22+ messages in thread
From: Marc Dionne @ 2008-05-08 3:13 UTC (permalink / raw)
To: Takashi Iwai
Cc: Markus Trippelsdorf, Jeff Garzik, Andrew Morton, Linus Torvalds,
linux-ide, LKML, htejun
07/05/08 12:35 PM, Takashi Iwai:
>
> The same problem appears on my machine, too:
>
> libata version 3.00 loaded.
> ata_piix 0000:00:1f.2: version 2.12
> ACPI: PCI Interrupt 0000:00:1f.2[B] -> GSI 19 (level, low) -> IRQ 19
> ata_piix 0000:00:1f.2: MAP [ P0 P2 P1 P3 ]
> PCI: Setting latency timer of device 0000:00:1f.2 to 64
> scsi0 : ata_piix
> scsi1 : ata_piix
> ata1: SATA max UDMA/133 cmd 0xe400 ctl 0xe500 bmdma 0xe800 irq 19
> ata2: SATA max UDMA/133 cmd 0xe600 ctl 0xe700 bmdma 0xe808 irq 19
> ata1.00: ATA-7: HDS728080PLA380, PF2OA60A, max UDMA/133
> ata1.00: 160836480 sectors, multi 16: LBA48 NCQ (depth 0/32)
> ata1.00: configured for UDMA/133
> ata2: link is slow to respond, please be patient (ready=0)
> ata2: device not ready (errno=-16), forcing hardreset
> ata2: link is slow to respond, please be patient (ready=0)
> ata2: SRST failed (errno=-16)
> ata2: link is slow to respond, please be patient (ready=0)
> ata2: SRST failed (errno=-16)
> ata2: link is slow to respond, please be patient (ready=0)
> ata2: SRST failed (errno=-16)
> ata2: SRST failed (errno=-16)
> ata2: reset failed, giving up
>
>
> Takashi
Same here as well, with sata_nv.
The patch comment states:
> This patch implements ata_check_ready(), which takes TF status value
> and determines whether the port is ready or not considering the above
> and other conditions, and use it in @check_ready() functions. This is
> safe as both 0x77 and 0x7f aren't valid ready status value even though
> they have BSY bit cleared.
.. but adding a print to the ata_check_ready function shows the status in my
case is systematically 0x7f for all unconnected links. This makes
ata_*_check_ready return 0 where it previously returned 1, and triggers several
hundred retries (over several minutes) before the boot proceeds.
Marc
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-07 16:35 ` Takashi Iwai
2008-05-08 3:13 ` Marc Dionne
@ 2008-05-08 9:45 ` Jan-Simon Möller
1 sibling, 0 replies; 22+ messages in thread
From: Jan-Simon Möller @ 2008-05-08 9:45 UTC (permalink / raw)
To: Takashi Iwai
Cc: Markus Trippelsdorf, Jeff Garzik, Andrew Morton, Linus Torvalds,
linux-ide, LKML, htejun
Am Mittwoch 07 Mai 2008 18:35:48 schrieben Sie:
> At Wed, 7 May 2008 16:56:16 +0200,
> Markus Trippelsdorf wrote:
> >
> > >
> > > Tejun Heo (12):
> > > libata: improve post-reset device ready test
> >
> > This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
> > boot delay with my onboard Promise controller. It seems like libata
> > probes for a nonexisting PATA drive...
> >
Same here :
<6>ata1: SATA max UDMA/133 abar m2048@0xfc404000 port 0xfc404100 irq 4347
<6>ata2: SATA max UDMA/133 abar m2048@0xfc404000 port 0xfc404180 irq 4347
<6>ata3: SATA max UDMA/133 abar m2048@0xfc404000 port 0xfc404200 irq 4347
<6>ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
<4>ata1.00: _GTF unexpected object type 0x1
<6>ata1.00: ATA-8: WDC WD2500BEVS-22UST0, 01.01A01, max UDMA/133
<6>ata1.00: 488397168 sectors, multi 16: LBA48 NCQ (depth 31/32)
<4>ata1.00: _GTF unexpected object type 0x1
<6>ata1.00: configured for UDMA/133
<6>ata2: SATA link down (SStatus 0 SControl 300)
<6>ata3: SATA link down (SStatus 0 SControl 300)
<5>scsi 0:0:0:0: Direct-Access ATA WDC WD2500BEVS-2 01.0 PQ: 0 ANSI: 5
<7>PM: Adding info for No Bus:target0:0:0
<7>PM: Adding info for scsi:0:0:0:0
<7>PM: Adding info for No Bus:0:0:0:0
<7>ata_piix 0000:00:1f.1: version 2.12
<6>ACPI: PCI Interrupt 0000:00:1f.1[A] -> GSI 19 (level, low) -> IRQ 19
<7>PCI: Setting latency timer of device 0000:00:1f.1 to 64
<6>scsi3 : ata_piix
<7>PM: Adding info for No Bus:host3
<7>PM: Adding info for No Bus:host3
<6>scsi4 : ata_piix
<7>PM: Adding info for No Bus:host4
<7>PM: Adding info for No Bus:host4
<6>ata4: PATA max UDMA/100 cmd 0x1f0 ctl 0x3f6 bmdma 0x1810 irq 14
<6>ata5: PATA max UDMA/100 cmd 0x170 ctl 0x376 bmdma 0x1818 irq 15
<6>ata4.00: ATAPI: HL-DT-ST DVDRAM GSA-T20N, WW01, max UDMA/33
<6>ata4.00: configured for UDMA/33
<4>ata5: link is slow to respond, please be patient (ready=0)
<4>ata5: device not ready (errno=-16), forcing hardreset
<4>ata5: link is slow to respond, please be patient (ready=0)
<3>ata5: SRST failed (errno=-16)
<4>ata5: link is slow to respond, please be patient (ready=0)
<3>ata5: SRST failed (errno=-16)
<4>ata5: link is slow to respond, please be patient (ready=0)
<3>ata5: SRST failed (errno=-16)
<3>ata5: SRST failed (errno=-16)
<3>ata5: reset failed, giving up
<4>isa bounce pool size: 16 pages
<5>scsi 3:0:0:0: CD-ROM HL-DT-ST DVDRAM GSA-T20N WW01 PQ: 0 ANSI: 5
<7>PM: Adding info for No Bus:target3:0:0
<7>PM: Adding info for scsi:3:0:0:0
<7>PM: Adding info for No Bus:3:0:0:0
<6>SGI XFS with ACLs, security attributes, realtime, large block/inode numbers, no debug enabled
<6>SGI XFS Quota Management subsystem
<6>BIOS EDD facility v0.16 2004-Jun-25, 0 devices found
<6>EDD information not available.
<4>sys_init_module: 'edd'->init suspiciously returned 1, it should follow 0/-E convention
<4>sys_init_module: loading module anyway...
<4>Pid: 554, comm: modprobe Not tainted 2.6.26-rc1-linus.git-00166-gc0a1811 #1
<4>
<4>Call Trace:
<4> [sys_init_module+304/428] sys_init_module+0x130/0x1ac
<4> [system_call_after_swapgs+138/143] system_call_after_swapgs+0x8a/0x8f
<4>
<6>usbcore: registered new interface driver usbfs
<6>usbcore: registered new interface driver hub
<6>usbcore: registered new device driver usb
^ permalink raw reply [flat|nested] 22+ messages in thread
* [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-07 14:56 ` [git patches] libata updates - (improve post-reset device ready test) regression Markus Trippelsdorf
2008-05-07 16:35 ` Takashi Iwai
@ 2008-05-08 15:35 ` Jeff Garzik
2008-05-08 16:05 ` Takashi Iwai
` (3 more replies)
1 sibling, 4 replies; 22+ messages in thread
From: Jeff Garzik @ 2008-05-08 15:35 UTC (permalink / raw)
To: Markus Trippelsdorf, linux-ide
Cc: Andrew Morton, Linus Torvalds, LKML, htejun, Takashi Iwai,
marc.c.dionne, dl9pf, bug-track, sitsofe
[-- Attachment #1: Type: text/plain, Size: 1648 bytes --]
Markus Trippelsdorf wrote:
>> Tejun Heo (12):
>> libata: improve post-reset device ready test
>
> This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
> boot delay with my onboard Promise controller. It seems like libata
> probes for a nonexisting PATA drive...
>
> ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
> scsi0 : sata_promise
> scsi1 : sata_promise
> scsi2 : sata_promise
> ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
> ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
> ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
> ata1: SATA link down (SStatus 0 SControl 300)
> ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
> ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
> ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
> ata2.00: configured for UDMA/133
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: device not ready (errno=-16), forcing hardreset
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: SRST failed (errno=-16)
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: SRST failed (errno=-16)
> ata3: link is slow to respond, please be patient (ready=0)
> ata3: SRST failed (errno=-16)
> - Last output repeated twice -
> ata3: reset failed, giving up
Does the attached patch fix things?
It basically reverts the patch, while still maintaining the
consolidation. It looks like that status evaluation is not as universal
as believed.
Jeff
[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 637 bytes --]
diff --git a/include/linux/libata.h b/include/linux/libata.h
index 7e206da..0f17643 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -1384,17 +1384,14 @@ static inline struct ata_port *ata_shost_to_port(struct Scsi_Host *host)
static inline int ata_check_ready(u8 status)
{
- /* Some controllers report 0x77 or 0x7f during intermediate
- * not-ready stages.
- */
- if (status == 0x77 || status == 0x7f)
- return 0;
+ if (!(status & ATA_BUSY))
+ return 1;
/* 0xff indicates either no device or device not ready */
if (status == 0xff)
return -ENODEV;
- return !(status & ATA_BUSY);
+ return 0;
}
^ permalink raw reply related [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 15:35 ` [PATCH] " Jeff Garzik
@ 2008-05-08 16:05 ` Takashi Iwai
2008-05-08 16:18 ` Markus Trippelsdorf
2008-05-08 21:43 ` J.A. Magallón
` (2 subsequent siblings)
3 siblings, 1 reply; 22+ messages in thread
From: Takashi Iwai @ 2008-05-08 16:05 UTC (permalink / raw)
To: Jeff Garzik
Cc: Markus Trippelsdorf, linux-ide, Andrew Morton, Linus Torvalds,
LKML, htejun, marc.c.dionne, dl9pf, bug-track, sitsofe
At Thu, 08 May 2008 11:35:11 -0400,
Jeff Garzik wrote:
>
> Markus Trippelsdorf wrote:
> >> Tejun Heo (12):
> >> libata: improve post-reset device ready test
> >
> > This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
> > boot delay with my onboard Promise controller. It seems like libata
> > probes for a nonexisting PATA drive...
> >
> > ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
> > scsi0 : sata_promise
> > scsi1 : sata_promise
> > scsi2 : sata_promise
> > ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
> > ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
> > ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
> > ata1: SATA link down (SStatus 0 SControl 300)
> > ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
> > ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
> > ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
> > ata2.00: configured for UDMA/133
> > ata3: link is slow to respond, please be patient (ready=0)
> > ata3: device not ready (errno=-16), forcing hardreset
> > ata3: link is slow to respond, please be patient (ready=0)
> > ata3: SRST failed (errno=-16)
> > ata3: link is slow to respond, please be patient (ready=0)
> > ata3: SRST failed (errno=-16)
> > ata3: link is slow to respond, please be patient (ready=0)
> > ata3: SRST failed (errno=-16)
> > - Last output repeated twice -
> > ata3: reset failed, giving up
>
> Does the attached patch fix things?
>
> It basically reverts the patch, while still maintaining the
> consolidation. It looks like that status evaluation is not as universal
> as believed.
Yes, the patch fixes the problem on my machine.
Thanks,
Takashi
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 16:05 ` Takashi Iwai
@ 2008-05-08 16:18 ` Markus Trippelsdorf
2008-05-08 17:40 ` Alexey Fisher
2008-05-08 19:21 ` Hugh Dickins
0 siblings, 2 replies; 22+ messages in thread
From: Markus Trippelsdorf @ 2008-05-08 16:18 UTC (permalink / raw)
To: Takashi Iwai
Cc: Jeff Garzik, linux-ide, Andrew Morton, Linus Torvalds, LKML,
htejun, marc.c.dionne, dl9pf, bug-track, sitsofe
On Thu, May 08, 2008 at 06:05:08PM +0200, Takashi Iwai wrote:
> At Thu, 08 May 2008 11:35:11 -0400,
> Jeff Garzik wrote:
> >
> > Markus Trippelsdorf wrote:
> > >> Tejun Heo (12):
> > >> libata: improve post-reset device ready test
> > >
> > > This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
> > > boot delay with my onboard Promise controller. It seems like libata
> > > probes for a nonexisting PATA drive...
> > >
> >
> > Does the attached patch fix things?
> >
> > It basically reverts the patch, while still maintaining the
> > consolidation. It looks like that status evaluation is not as universal
> > as believed.
>
> Yes, the patch fixes the problem on my machine.
>
Your patch fixes the problem for me, too...
--
Markus
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 16:18 ` Markus Trippelsdorf
@ 2008-05-08 17:40 ` Alexey Fisher
2008-05-08 19:21 ` Hugh Dickins
1 sibling, 0 replies; 22+ messages in thread
From: Alexey Fisher @ 2008-05-08 17:40 UTC (permalink / raw)
To: Markus Trippelsdorf
Cc: Takashi Iwai, Jeff Garzik, linux-ide, Andrew Morton,
Linus Torvalds, LKML, htejun, marc.c.dionne, dl9pf, sitsofe
[-- Attachment #1: Type: text/plain, Size: 912 bytes --]
Am Donnerstag, den 08.05.2008, 18:18 +0200 schrieb Markus Trippelsdorf:
> On Thu, May 08, 2008 at 06:05:08PM +0200, Takashi Iwai wrote:
> > At Thu, 08 May 2008 11:35:11 -0400,
> > Jeff Garzik wrote:
> > >
> > > Markus Trippelsdorf wrote:
> > > >> Tejun Heo (12):
> > > >> libata: improve post-reset device ready test
> > > >
> > > > This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
> > > > boot delay with my onboard Promise controller. It seems like libata
> > > > probes for a nonexisting PATA drive...
> > > >
> > >
> > > Does the attached patch fix things?
> > >
> > > It basically reverts the patch, while still maintaining the
> > > consolidation. It looks like that status evaluation is not as universal
> > > as believed.
> >
> > Yes, the patch fixes the problem on my machine.
> >
>
> Your patch fixes the problem for me, too...
me too :)
[-- Attachment #2: Dies ist ein digital signierter Nachrichtenteil --]
[-- Type: application/pgp-signature, Size: 189 bytes --]
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 16:18 ` Markus Trippelsdorf
2008-05-08 17:40 ` Alexey Fisher
@ 2008-05-08 19:21 ` Hugh Dickins
1 sibling, 0 replies; 22+ messages in thread
From: Hugh Dickins @ 2008-05-08 19:21 UTC (permalink / raw)
To: Jeff Garzik
Cc: Markus Trippelsdorf, Takashi Iwai, Jeff Garzik, linux-ide,
Andrew Morton, Linus Torvalds, LKML, htejun, marc.c.dionne,
dl9pf, bug-track, sitsofe
On Thu, 8 May 2008, Markus Trippelsdorf wrote:
> On Thu, May 08, 2008 at 06:05:08PM +0200, Takashi Iwai wrote:
> > At Thu, 08 May 2008 11:35:11 -0400,
> > Jeff Garzik wrote:
> > >
> > > It basically reverts the patch, while still maintaining the
> > > consolidation. It looks like that status evaluation is not as universal
> > > as believed.
> >
> > Yes, the patch fixes the problem on my machine.
>
> Your patch fixes the problem for me, too...
And for me: I was preparing to report when I saw these go by.
A Fujitsu-Siemens laptop and a Fujitsu-Siemens desktop were each taking
70 seconds longer to boot than before. (The status byte was 0x7f.)
Thanks,
Hugh
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 15:35 ` [PATCH] " Jeff Garzik
2008-05-08 16:05 ` Takashi Iwai
@ 2008-05-08 21:43 ` J.A. Magallón
2008-05-09 14:54 ` Mark Lord
2008-05-08 21:51 ` Marc Dionne
2008-05-09 2:57 ` Tejun Heo
3 siblings, 1 reply; 22+ messages in thread
From: J.A. Magallón @ 2008-05-08 21:43 UTC (permalink / raw)
To: Jeff Garzik, Linux-Kernel, Linus-IDE
On Thu, 08 May 2008 11:35:11 -0400, Jeff Garzik <jeff@garzik.org> wrote:
> Markus Trippelsdorf wrote:
> >> Tejun Heo (12):
> >> libata: improve post-reset device ready test
> >
> > This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
> > boot delay with my onboard Promise controller. It seems like libata
> > probes for a nonexisting PATA drive...
> >
I also had this patch collected from LKML, that still applies to -git6.
Is it really needed ?
ref: http://marc.info/?l=linux-ide&m=120913178617926&w=2
TIA
--- upstream/drivers/ata/libata-eh.c 2008-04-30 17:35:36.000000000 -0400
+++ linux/drivers/ata/libata-eh.c 2008-04-30 17:35:45.000000000 -0400
@@ -1312,8 +1312,7 @@
err_mask |= AC_ERR_ATA_BUS;
action |= ATA_EH_RESET;
}
- if (serror &
- (SERR_DATA_RECOVERED | SERR_COMM_RECOVERED | SERR_DATA)) {
+ if (serror & (SERR_DATA_RECOVERED | SERR_DATA)) {
err_mask |= AC_ERR_ATA_BUS;
action |= ATA_EH_RESET;
}
@@ -1924,7 +1923,7 @@
}
if (ehc->i.serror)
- ata_port_printk(ap, KERN_ERR,
+ ata_link_printk(link, KERN_ERR,
"SError: { %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s}\n",
ehc->i.serror & SERR_DATA_RECOVERED ? "RecovData " : "",
ehc->i.serror & SERR_COMM_RECOVERED ? "RecovComm " : "",
--
J.A. Magallon <jamagallon()ono!com> \ Software is like sex:
\ It's better when it's free
Mandriva Linux release 2008.1 (Cooker) for i586
Linux 2.6.23-jam05 (gcc 4.2.2 20071128 (4.2.2-2mdv2008.1)) SMP PREEMPT
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 15:35 ` [PATCH] " Jeff Garzik
2008-05-08 16:05 ` Takashi Iwai
2008-05-08 21:43 ` J.A. Magallón
@ 2008-05-08 21:51 ` Marc Dionne
2008-05-09 2:57 ` Tejun Heo
3 siblings, 0 replies; 22+ messages in thread
From: Marc Dionne @ 2008-05-08 21:51 UTC (permalink / raw)
To: Jeff Garzik
Cc: Markus Trippelsdorf, linux-ide, Andrew Morton, Linus Torvalds,
LKML, htejun, Takashi Iwai, dl9pf, bug-track, sitsofe
08/05/08 11:35 AM, Jeff Garzik:
> Markus Trippelsdorf wrote:
>>> Tejun Heo (12):
>>> libata: improve post-reset device ready test
>>
>> This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
>> boot delay with my onboard Promise controller. It seems like libata
>> probes for a nonexisting PATA drive...
>>
>> ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
>> scsi0 : sata_promise
>> scsi1 : sata_promise
>> scsi2 : sata_promise
>> ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
>> ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
>> ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
>> ata1: SATA link down (SStatus 0 SControl 300)
>> ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
>> ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
>> ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
>> ata2.00: configured for
>> UDMA/133
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: device not ready (errno=-16), forcing hardreset
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: SRST failed (errno=-16)
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: SRST failed (errno=-16)
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: SRST failed (errno=-16)
>> - Last output repeated twice -
>> ata3: reset failed, giving up
>
> Does the attached patch fix things?
>
> It basically reverts the patch, while still maintaining the
> consolidation. It looks like that status evaluation is not as universal
> as believed.
>
> Jeff
Yes, with the patch I get normal booting behaviour.
Thanks,
Marc
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 15:35 ` [PATCH] " Jeff Garzik
` (2 preceding siblings ...)
2008-05-08 21:51 ` Marc Dionne
@ 2008-05-09 2:57 ` Tejun Heo
2008-05-09 3:01 ` Tejun Heo
` (4 more replies)
3 siblings, 5 replies; 22+ messages in thread
From: Tejun Heo @ 2008-05-09 2:57 UTC (permalink / raw)
To: Jeff Garzik
Cc: Markus Trippelsdorf, linux-ide, Andrew Morton, Linus Torvalds,
LKML, Takashi Iwai, marc.c.dionne, dl9pf, bug-track, sitsofe
Jeff Garzik wrote:
> Markus Trippelsdorf wrote:
>>> Tejun Heo (12):
>>> libata: improve post-reset device ready test
>>
>> This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
>> boot delay with my onboard Promise controller. It seems like libata
>> probes for a nonexisting PATA drive...
>>
>> ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
>> scsi0 : sata_promise
>> scsi1 : sata_promise
>> scsi2 : sata_promise
>> ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
>> ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
>> ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
>> ata1: SATA link down (SStatus 0 SControl 300)
>> ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
>> ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
>> ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
>> ata2.00: configured for
>> UDMA/133
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: device not ready (errno=-16), forcing hardreset
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: SRST failed (errno=-16)
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: SRST failed (errno=-16)
>> ata3: link is slow to respond, please be patient (ready=0)
>> ata3: SRST failed (errno=-16)
>> - Last output repeated twice -
>> ata3: reset failed, giving up
>
> Does the attached patch fix things?
>
> It basically reverts the patch, while still maintaining the
> consolidation. It looks like that status evaluation is not as universal
> as believed.
This means that we need to make custom readiness tests for controllers
using 0x77 or 0x7f. Eeeek... Both groups of controllers are behaving in
incorrect way. Controllers shouldn't use 0x77 or 0x7f for either busy
or ready states - it's invalid for both, yet, some use the 77/7f for
busy while others use them for ready state. Great. :-(
Jeff, can you please revert 78ab88f04f44bed566d51dce0c7cbfeff6449a06 for
now? I'll post updated patch.
Markus, Alexey, Marc, which controllers are you using?
Thanks.
--
tejun
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-09 2:57 ` Tejun Heo
@ 2008-05-09 3:01 ` Tejun Heo
2008-05-09 4:27 ` Justin Mattock
2008-05-09 3:43 ` Justin Mattock
` (3 subsequent siblings)
4 siblings, 1 reply; 22+ messages in thread
From: Tejun Heo @ 2008-05-09 3:01 UTC (permalink / raw)
To: Jeff Garzik
Cc: Markus Trippelsdorf, linux-ide, Andrew Morton, Linus Torvalds,
LKML, Takashi Iwai, marc.c.dionne, dl9pf, bug-track, sitsofe
Tejun Heo wrote:
> Markus, Alexey, Marc, which controllers are you using?
Never mind, found the rest of the thread.
Thanks.
--
tejun
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-09 2:57 ` Tejun Heo
2008-05-09 3:01 ` Tejun Heo
@ 2008-05-09 3:43 ` Justin Mattock
2008-05-09 4:37 ` Jeff Garzik
` (2 subsequent siblings)
4 siblings, 0 replies; 22+ messages in thread
From: Justin Mattock @ 2008-05-09 3:43 UTC (permalink / raw)
To: Tejun Heo
Cc: Jeff Garzik, Markus Trippelsdorf, linux-ide, Andrew Morton,
Linus Torvalds, LKML, Takashi Iwai, marc.c.dionne, dl9pf,
bug-track, sitsofe
On Fri, May 9, 2008 at 2:57 AM, Tejun Heo <htejun@gmail.com> wrote:
> Jeff Garzik wrote:
>>
>> Markus Trippelsdorf wrote:
>>>>
>>>> Tejun Heo (12):
>>>> libata: improve post-reset device ready test
>>>
>>> This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
>>> boot delay with my onboard Promise controller. It seems like libata
>>> probes for a nonexisting PATA drive...
>>>
>>> ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
>>> scsi0 : sata_promise
>>> scsi1 : sata_promise
>>> scsi2 : sata_promise
>>> ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
>>> ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
>>> ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
>>> ata1: SATA link down (SStatus 0 SControl 300)
>>> ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
>>> ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
>>> ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
>>> ata2.00: configured for UDMA/133
>>> ata3: link is slow to
>>> respond, please be patient (ready=0)
>>> ata3: device not ready (errno=-16), forcing hardreset
>>> ata3: link is slow to respond, please be patient (ready=0)
>>> ata3: SRST failed (errno=-16)
>>> ata3: link is slow to respond, please be patient (ready=0)
>>> ata3: SRST failed (errno=-16)
>>> ata3: link is slow to respond, please be patient (ready=0)
>>> ata3: SRST failed (errno=-16)
>>> - Last output repeated twice -
>>> ata3: reset failed, giving up
>>
>> Does the attached patch fix things?
>>
>> It basically reverts the patch, while still maintaining the consolidation.
>> It looks like that status evaluation is not as universal as believed.
>
> This means that we need to make custom readiness tests for controllers using
> 0x77 or 0x7f. Eeeek... Both groups of controllers are behaving in incorrect
> way. Controllers shouldn't use 0x77 or 0x7f for either busy or ready states
> - it's invalid for both, yet, some use the 77/7f for busy while others use
> them for ready state. Great. :-(
>
> Jeff, can you please revert 78ab88f04f44bed566d51dce0c7cbfeff6449a06 for
> now? I'll post updated patch.
>
> Markus, Alexey, Marc, which controllers are you using?
>
> Thanks.
>
> --
> tejun
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
Hello; Im getting that same message here as well, If you have the
patch I'll give it a try to see if it works.
regards;
--
Justin P. Mattock
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-09 3:01 ` Tejun Heo
@ 2008-05-09 4:27 ` Justin Mattock
0 siblings, 0 replies; 22+ messages in thread
From: Justin Mattock @ 2008-05-09 4:27 UTC (permalink / raw)
To: Tejun Heo
Cc: Jeff Garzik, Markus Trippelsdorf, linux-ide, Andrew Morton,
Linus Torvalds, LKML, Takashi Iwai, marc.c.dionne, dl9pf,
bug-track, sitsofe
On Fri, May 9, 2008 at 3:01 AM, Tejun Heo <htejun@gmail.com> wrote:
> Tejun Heo wrote:
>>
>> Markus, Alexey, Marc, which controllers are you using?
>
> Never mind, found the rest of the thread.
>
> Thanks.
>
> --
> tejun
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
This patch here is what fixes this issue for the macbook pro.
http://lkml.org/lkml/2008/5/8/245
regards;
--
Justin P. Mattock
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-09 2:57 ` Tejun Heo
2008-05-09 3:01 ` Tejun Heo
2008-05-09 3:43 ` Justin Mattock
@ 2008-05-09 4:37 ` Jeff Garzik
2008-05-10 6:34 ` Tejun Heo
2008-05-09 5:09 ` Alexey Fisher
2008-05-09 10:59 ` Marc Dionne
4 siblings, 1 reply; 22+ messages in thread
From: Jeff Garzik @ 2008-05-09 4:37 UTC (permalink / raw)
To: Tejun Heo
Cc: Markus Trippelsdorf, linux-ide, Andrew Morton, Linus Torvalds,
LKML, Takashi Iwai, marc.c.dionne, dl9pf, bug-track, sitsofe
Tejun Heo wrote:
> This means that we need to make custom readiness tests for controllers
> using 0x77 or 0x7f. Eeeek... Both groups of controllers are behaving in
> incorrect way. Controllers shouldn't use 0x77 or 0x7f for either busy
> or ready states - it's invalid for both, yet, some use the 77/7f for
> busy while others use them for ready state. Great. :-(
I think that's assuming too much? PATA and SATA are quite different
here... in PATA the status is mostly the value from the device directly
off the wires. in SATA, it may be from the device or from the
controller. And "smart" or firmware-based controllers may generate
their own status, too, apart from the device's status.
So that results in varied status returns, and not all the time is a
definite "ready" or "not ready" obvious.
Jeff
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-09 2:57 ` Tejun Heo
` (2 preceding siblings ...)
2008-05-09 4:37 ` Jeff Garzik
@ 2008-05-09 5:09 ` Alexey Fisher
2008-05-09 10:59 ` Marc Dionne
4 siblings, 0 replies; 22+ messages in thread
From: Alexey Fisher @ 2008-05-09 5:09 UTC (permalink / raw)
To: Tejun Heo
Cc: Jeff Garzik, Markus Trippelsdorf, linux-ide, Andrew Morton,
Linus Torvalds, LKML, Takashi Iwai, marc.c.dionne, dl9pf,
sitsofe
[-- Attachment #1: Type: text/plain, Size: 459 bytes --]
Am Freitag, den 09.05.2008, 11:57 +0900 schrieb Tejun Heo:
>
> Markus, Alexey, Marc, which controllers are you using?
>
> Thanks.
>
This is lspci of my controller.
00:1f.1 IDE interface: Intel Corporation 82801G (ICH7 Family) IDE
Controller (rev 01)
00:1f.2 IDE interface: Intel Corporation 82801GB/GR/GH (ICH7 Family)
SATA IDE Controller (rev 01)
complete lspci -vvxxn you can find here
http://bugzilla.kernel.org/show_bug.cgi?id=10619
[-- Attachment #2: Dies ist ein digital signierter Nachrichtenteil --]
[-- Type: application/pgp-signature, Size: 189 bytes --]
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-09 2:57 ` Tejun Heo
` (3 preceding siblings ...)
2008-05-09 5:09 ` Alexey Fisher
@ 2008-05-09 10:59 ` Marc Dionne
4 siblings, 0 replies; 22+ messages in thread
From: Marc Dionne @ 2008-05-09 10:59 UTC (permalink / raw)
To: Tejun Heo
Cc: Jeff Garzik, Markus Trippelsdorf, linux-ide, Andrew Morton,
Linus Torvalds, LKML, Takashi Iwai, dl9pf, bug-track, sitsofe
08/05/08 10:57 PM, Tejun Heo:
> Jeff Garzik wrote:
>> Markus Trippelsdorf wrote:
>>>> Tejun Heo (12):
>>>> libata: improve post-reset device ready test
>>>
>>> This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
>>> boot delay with my onboard Promise controller. It seems like libata
>>> probes for a nonexisting PATA drive...
>>>
>>> ACPI: PCI Interrupt 0000:00:08.0[A] -> GSI 18 (level, low) -> IRQ 18
>>> scsi0 : sata_promise
>>> scsi1 : sata_promise
>>> scsi2 : sata_promise
>>> ata1: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600200 irq 18
>>> ata2: SATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600280 irq 18
>>> ata3: PATA max UDMA/133 mmio m4096@0xfb600000 port 0xfb600300 irq 18
>>> ata1: SATA link down (SStatus 0 SControl 300)
>>> ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
>>> ata2.00: ATA-7: SAMSUNG HD753LJ, 1AA01109, max UDMA7
>>> ata2.00: 1465149168 sectors, multi 0: LBA48 NCQ (depth 0/32)
>>> ata2.00: configured for
>>> UDMA/133
>>> ata3: link is slow to respond, please be patient (ready=0)
>>> ata3: device not ready (errno=-16), forcing hardreset
>>> ata3: link is slow to respond, please be patient (ready=0)
>>> ata3: SRST failed (errno=-16)
>>> ata3: link is slow to respond, please be patient (ready=0)
>>> ata3: SRST failed (errno=-16)
>>> ata3: link is slow to respond, please be patient (ready=0)
>>> ata3: SRST failed (errno=-16)
>>> - Last output repeated twice -
>>> ata3: reset failed, giving up
>>
>> Does the attached patch fix things?
>>
>> It basically reverts the patch, while still maintaining the
>> consolidation. It looks like that status evaluation is not as
>> universal as believed.
>
> This means that we need to make custom readiness tests for controllers
> using 0x77 or 0x7f. Eeeek... Both groups of controllers are behaving in
> incorrect way. Controllers shouldn't use 0x77 or 0x7f for either busy
> or ready states - it's invalid for both, yet, some use the 77/7f for
> busy while others use them for ready state. Great. :-(
>
> Jeff, can you please revert 78ab88f04f44bed566d51dce0c7cbfeff6449a06 for
> now? I'll post updated patch.
>
> Markus, Alexey, Marc, which controllers are you using?
>
> Thanks.
>
nVidia NForce4 chipset, sata_nv driver:
00:06.0 IDE interface: nVidia Corporation CK804 IDE (rev f2)
00:08.0 IDE interface: nVidia Corporation CK804 Serial ATA Controller (rev f3)
Marc
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-08 21:43 ` J.A. Magallón
@ 2008-05-09 14:54 ` Mark Lord
0 siblings, 0 replies; 22+ messages in thread
From: Mark Lord @ 2008-05-09 14:54 UTC (permalink / raw)
To: "J.A. Magallón"; +Cc: Jeff Garzik, Linux-Kernel, Linus-IDE
J.A. Magallón wrote:
> On Thu, 08 May 2008 11:35:11 -0400, Jeff Garzik <jeff@garzik.org> wrote:
>
>> Markus Trippelsdorf wrote:
>>>> Tejun Heo (12):
>>>> libata: improve post-reset device ready test
>>> This commit (78ab88f04f44bed566d51dce0c7cbfeff6449a06) causes a long
>>> boot delay with my onboard Promise controller. It seems like libata
>>> probes for a nonexisting PATA drive...
>>>
>
> I also had this patch collected from LKML, that still applies to -git6.
> Is it really needed ?
> ref: http://marc.info/?l=linux-ide&m=120913178617926&w=2
..
> --- upstream/drivers/ata/libata-eh.c 2008-04-30 17:35:36.000000000 -0400
> +++ linux/drivers/ata/libata-eh.c 2008-04-30 17:35:45.000000000 -0400
> @@ -1312,8 +1312,7 @@
> err_mask |= AC_ERR_ATA_BUS;
> action |= ATA_EH_RESET;
> }
> - if (serror &
> - (SERR_DATA_RECOVERED | SERR_COMM_RECOVERED | SERR_DATA)) {
> + if (serror & (SERR_DATA_RECOVERED | SERR_DATA)) {
> err_mask |= AC_ERR_ATA_BUS;
> action |= ATA_EH_RESET;
> }
..
The original problem I had, which prompted this patch, eventually was
resolved in other ways. But I do think that a hw-recovered bit in SError
is not a good reason to reset everything.
Tejun suggested that we still want to know about such errors,
and eventually do something if they repeat often enough.
But doing a port/link reset every single time ?
> @@ -1924,7 +1923,7 @@
> }
>
> if (ehc->i.serror)
> - ata_port_printk(ap, KERN_ERR,
> + ata_link_printk(link, KERN_ERR,
> "SError: { %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s}\n",
> ehc->i.serror & SERR_DATA_RECOVERED ? "RecovData " : "",
> ehc->i.serror & SERR_COMM_RECOVERED ? "RecovComm " : "",
>
..
That portion of the patch is still a very good idea.
Cheers
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH] Re: [git patches] libata updates - (improve post-reset device ready test) regression
2008-05-09 4:37 ` Jeff Garzik
@ 2008-05-10 6:34 ` Tejun Heo
0 siblings, 0 replies; 22+ messages in thread
From: Tejun Heo @ 2008-05-10 6:34 UTC (permalink / raw)
To: Jeff Garzik
Cc: Markus Trippelsdorf, linux-ide, Andrew Morton, Linus Torvalds,
LKML, Takashi Iwai, marc.c.dionne, dl9pf, bug-track, sitsofe
Jeff Garzik wrote:
> Tejun Heo wrote:
>> This means that we need to make custom readiness tests for controllers
>> using 0x77 or 0x7f. Eeeek... Both groups of controllers are behaving
>> in incorrect way. Controllers shouldn't use 0x77 or 0x7f for either
>> busy or ready states - it's invalid for both, yet, some use the 77/7f
>> for busy while others use them for ready state. Great. :-(
>
> I think that's assuming too much? PATA and SATA are quite different
> here... in PATA the status is mostly the value from the device directly
> off the wires. in SATA, it may be from the device or from the
> controller. And "smart" or firmware-based controllers may generate
> their own status, too, apart from the device's status.
>
> So that results in varied status returns, and not all the time is a
> definite "ready" or "not ready" obvious.
I think it's pretty safe to say that these weird ready values are from
TF emulation on controller side. The ready/not ready distinction is
probably too simplistic but those values aren't supposed to appear
during post-reset readiness test.
Sorry about the big regression. Heh... It's amazing how all the
controllers I tested didn't show the problem and I did test a good
number of combinations.
I still think it would be better to have a unified readiness test
function. The problem is subtle (device misdetection on hotplug of
certain drives) and went unnoticed quite some time for JMB ahcis && test
coverage over those things can't be good. I'll try to think about
something better.
Thanks.
--
tejun
^ permalink raw reply [flat|nested] 22+ messages in thread
end of thread, other threads:[~2008-05-10 6:34 UTC | newest]
Thread overview: 22+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-05-06 15:48 [git patches] libata updates Jeff Garzik
2008-05-06 16:23 ` Linus Torvalds
2008-05-07 14:56 ` [git patches] libata updates - (improve post-reset device ready test) regression Markus Trippelsdorf
2008-05-07 16:35 ` Takashi Iwai
2008-05-08 3:13 ` Marc Dionne
2008-05-08 9:45 ` Jan-Simon Möller
2008-05-08 15:35 ` [PATCH] " Jeff Garzik
2008-05-08 16:05 ` Takashi Iwai
2008-05-08 16:18 ` Markus Trippelsdorf
2008-05-08 17:40 ` Alexey Fisher
2008-05-08 19:21 ` Hugh Dickins
2008-05-08 21:43 ` J.A. Magallón
2008-05-09 14:54 ` Mark Lord
2008-05-08 21:51 ` Marc Dionne
2008-05-09 2:57 ` Tejun Heo
2008-05-09 3:01 ` Tejun Heo
2008-05-09 4:27 ` Justin Mattock
2008-05-09 3:43 ` Justin Mattock
2008-05-09 4:37 ` Jeff Garzik
2008-05-10 6:34 ` Tejun Heo
2008-05-09 5:09 ` Alexey Fisher
2008-05-09 10:59 ` Marc Dionne
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).