diff options
| author | David Gwynne <dlg@cvs.openbsd.org> | 2010-05-24 04:40:15 +0000 |
|---|---|---|
| committer | David Gwynne <dlg@cvs.openbsd.org> | 2010-05-24 04:40:15 +0000 |
| commit | af5ac829713346309d8784d737e7979488f5f348 (patch) | |
| tree | 3150759b466faf14c80d6f2ef5d4eab8529544de | |
| parent | 8290987282000ff434f8c0fd97c12b256d1127cd (diff) | |
atascsi may (will) reuse ata_xfers, meaning the xa state that
ahci_ata_cmd is called with might be the state of the xfer that
ahci set previously. this stops ahci_ata_cmd from checking the
state of ata_xfers its given, and instead trusts that atascsi is
doing the right thing.
reported and fix tested by matthew dempsky, who also did an awesome
job of tracing the problem.
| -rw-r--r-- | sys/dev/pci/ahci.c | 6 |
1 files changed, 1 insertions, 5 deletions
diff --git a/sys/dev/pci/ahci.c b/sys/dev/pci/ahci.c index 98630fcb7c0..4f0b095fcb3 100644 --- a/sys/dev/pci/ahci.c +++ b/sys/dev/pci/ahci.c @@ -1,4 +1,4 @@ -/* $OpenBSD: ahci.c,v 1.164 2010/05/19 15:27:35 oga Exp $ */ +/* $OpenBSD: ahci.c,v 1.165 2010/05/24 04:40:14 dlg Exp $ */ /* * Copyright (c) 2006 David Gwynne <dlg@openbsd.org> @@ -1564,8 +1564,6 @@ ahci_start(struct ahci_ccb *ccb) struct ahci_port *ap = ccb->ccb_port; struct ahci_softc *sc = ap->ap_sc; - KASSERT(ccb->ccb_xa.state == ATA_S_PENDING); - /* Zero transferred byte count before transfer */ ccb->ccb_cmd_hdr->prdbc = 0; @@ -2349,8 +2347,6 @@ ahci_ata_cmd(struct ata_xfer *xa) struct ahci_cmd_hdr *cmd_slot; int s; - KASSERT(xa->state == ATA_S_SETUP); - if (ccb->ccb_port->ap_state == AP_S_FATAL_ERROR) goto failcmd; |