svn commit: r220929 - head/sys/dev/ata/chipsets

Alexander Motin mav at FreeBSD.org
Thu Apr 21 20:56:34 UTC 2011


Author: mav
Date: Thu Apr 21 20:56:34 2011
New Revision: 220929
URL: http://svn.freebsd.org/changeset/base/220929

Log:
  According to ATA specifications, when ATAPI master is the only device, it
  should respond with all zeroes to any access to slave registers. Test with
  PATA devices confirmed such behavior. Unluckily, Intel SATA controllers in
  legacy emulation mode behave differently, not making any difference between
  ATA and ATAPI devices. It causes false positive slave device detection and,
  as result, command timeouts.
  
  To workaround this problem, mask result of legacy-emulated soft-reset with
  the device presence information received from the SATA-specific registers.

Modified:
  head/sys/dev/ata/chipsets/ata-intel.c

Modified: head/sys/dev/ata/chipsets/ata-intel.c
==============================================================================
--- head/sys/dev/ata/chipsets/ata-intel.c	Thu Apr 21 19:56:06 2011	(r220928)
+++ head/sys/dev/ata/chipsets/ata-intel.c	Thu Apr 21 20:56:34 2011	(r220929)
@@ -392,8 +392,9 @@ ata_intel_reset(device_t dev)
 	device_t parent = device_get_parent(dev);
 	struct ata_pci_controller *ctlr = device_get_softc(parent);
 	struct ata_channel *ch = device_get_softc(dev);
-	int mask, pmask, timeout, devs;
+	int mask, pshift, timeout, devs;
 	u_char *smap;
+	uint16_t pcs;
 
 	/* In combined mode, skip SATA stuff for PATA channel. */
 	if ((ch->flags & ATA_SATA) == 0)
@@ -412,26 +413,35 @@ ata_intel_reset(device_t dev)
 
 	/* Wait up to 1 sec for "connect well". */
 	if (ctlr->chip->cfg1 & (INTEL_6CH | INTEL_6CH2))
-		pmask = mask << 8;
+		pshift = 8;
 	else
-		pmask = mask << 4;
+		pshift = 4;
 	for (timeout = 0; timeout < 100 ; timeout++) {
-		if (((pci_read_config(parent, 0x92, 2) & pmask) == pmask) &&
-		    (ATA_IDX_INB(ch, ATA_STATUS) != 0xff))
+		pcs = (pci_read_config(parent, 0x92, 2) >> pshift) & mask;
+		if ((pcs == mask) && (ATA_IDX_INB(ch, ATA_STATUS) != 0xff))
 			break;
 		ata_udelay(10000);
 	}
 
+	if (bootverbose)
+		device_printf(dev, "SATA reset: ports status=0x%02x\n", pcs);
 	/* If any device found, do soft-reset. */
 	if (ch->hw.pm_read != NULL) {
-		devs = ata_sata_phy_reset(dev, 0, 2);
+		devs = ata_sata_phy_reset(dev, 0, 2) ? ATA_ATA_MASTER : 0;
 		if ((ch->flags & ATA_NO_SLAVE) == 0)
-			devs += ata_sata_phy_reset(dev, 1, 2);
-	} else
-		devs = 1;
-	if (devs)
+			devs |= ata_sata_phy_reset(dev, 1, 2) ?
+			    ATA_ATA_SLAVE : 0;
+	} else {
+		devs = (pcs & (1 << smap[0])) ? ATA_ATA_MASTER : 0;
+		if ((ch->flags & ATA_NO_SLAVE) == 0)
+			devs |= (pcs & (1 << smap[1])) ?
+			    ATA_ATA_SLAVE : 0;
+	}
+	if (devs) {
 		ata_generic_reset(dev);
-	else
+		/* Reset may give fake slave when only ATAPI master present. */
+		ch->devices &= (devs | (devs * ATA_ATAPI_MASTER));
+	} else
 		ch->devices = 0;
 }
 


More information about the svn-src-head mailing list