Merge branch 'for-3.12-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata

Pull libata fixes from Tejun Heo:
 "The only interesting bit is ata_eh_qc_retry() update which fixes a
  problem where a SG_IO command may fail across suspend/resume cycle
  without the command actually being at fault.

  Other changes are low level driver specific and fairly low impact"

* 'for-3.12-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata:
  libahci: fix turning on LEDs in ahci_start_port()
  libata: make ata_eh_qc_retry() bump scmd->allowed on bogus failures
  ahci_platform: use dev_info() instead of printk()
  ahci: use dev_info() instead of printk()
  pata_isapnp: Don't use invalid I/O ports

1 files changed, 9 insertions, 1 deletions

diff --git a/drivers/ata/libahci.c b/drivers/ata/libahci.c
index acfd0f711069..aaac4fb0d564 100644
--- a/drivers/ata/libahci.c
+++ b/drivers/ata/libahci.c
@@ -778,8 +778,16 @@ static void ahci_start_port(struct ata_port *ap)
                                 rc = ap->ops->transmit_led_message(ap,
                                                                emp->led_state,
                                                                4);
+                                /*
+                                 * If busy, give a breather but do not
+                                 * release EH ownership by using msleep()
+                                 * instead of ata_msleep().  EM Transmit
+                                 * bit is busy for the whole host and
+                                 * releasing ownership will cause other
+                                 * ports to fail the same way.
+                                 */
                                 if (rc == -EBUSY)
-                                        ata_msleep(ap, 1);
+                                        msleep(1);
                                 else
                                         break;
                         }