From 2d33cc7688caa1a8f870ad21f859903d115d1603 Mon Sep 17 00:00:00 2001 From: Alexander Motin Date: Fri, 29 Jul 2011 20:35:23 +0000 Subject: [PATCH] In some cases, at least on Marvell 88SE912x controllers, Current Command Slot field of the PxCMD register may point to an empty command slot. That breaks command timeout detection logic, making impossible to find what command actually caused timeout, and leading to infinite wait. Workaround that by checking whether pointed command slot is really used and can timeout in its time. And if not, fallback to the dumb algorithm used with FBS -- let all commands to time out and then fail all of them. Approved by: re (kib) MFC after: 1 week --- sys/dev/ahci/ahci.c | 20 ++++++++++++++------ sys/dev/ahci/ahci.h | 1 + 2 files changed, 15 insertions(+), 6 deletions(-) diff --git a/sys/dev/ahci/ahci.c b/sys/dev/ahci/ahci.c index 0bc094edccea..1d6185dce833 100644 --- a/sys/dev/ahci/ahci.c +++ b/sys/dev/ahci/ahci.c @@ -1879,12 +1879,13 @@ ahci_execute_transaction(struct ahci_slot *slot) device_printf(dev, "Poll timeout on slot %d port %d\n", slot->slot, port); device_printf(dev, "is %08x cs %08x ss %08x " - "rs %08x tfd %02x serr %08x\n", + "rs %08x tfd %02x serr %08x cmd %08x\n", ATA_INL(ch->r_mem, AHCI_P_IS), ATA_INL(ch->r_mem, AHCI_P_CI), ATA_INL(ch->r_mem, AHCI_P_SACT), ch->rslots, ATA_INL(ch->r_mem, AHCI_P_TFD), - ATA_INL(ch->r_mem, AHCI_P_SERR)); + ATA_INL(ch->r_mem, AHCI_P_SERR), + ATA_INL(ch->r_mem, AHCI_P_CMD)); et = AHCI_ERR_TIMEOUT; } @@ -1960,8 +1961,12 @@ ahci_timeout(struct ahci_slot *slot) ccs = (ATA_INL(ch->r_mem, AHCI_P_CMD) & AHCI_P_CMD_CCS_MASK) >> AHCI_P_CMD_CCS_SHIFT; if ((sstatus & (1 << slot->slot)) != 0 || ccs == slot->slot || - ch->fbs_enabled) + ch->fbs_enabled || ch->wrongccs) slot->state = AHCI_SLOT_EXECUTING; + else if ((ch->rslots & (1 << ccs)) == 0) { + ch->wrongccs = 1; + slot->state = AHCI_SLOT_EXECUTING; + } callout_reset(&slot->timeout, (int)slot->ccb->ccb_h.timeout * hz / 2000, @@ -1971,10 +1976,12 @@ ahci_timeout(struct ahci_slot *slot) device_printf(dev, "Timeout on slot %d port %d\n", slot->slot, slot->ccb->ccb_h.target_id & 0x0f); - device_printf(dev, "is %08x cs %08x ss %08x rs %08x tfd %02x serr %08x\n", + device_printf(dev, "is %08x cs %08x ss %08x rs %08x tfd %02x " + "serr %08x cmd %08x\n", ATA_INL(ch->r_mem, AHCI_P_IS), ATA_INL(ch->r_mem, AHCI_P_CI), ATA_INL(ch->r_mem, AHCI_P_SACT), ch->rslots, - ATA_INL(ch->r_mem, AHCI_P_TFD), ATA_INL(ch->r_mem, AHCI_P_SERR)); + ATA_INL(ch->r_mem, AHCI_P_TFD), ATA_INL(ch->r_mem, AHCI_P_SERR), + ATA_INL(ch->r_mem, AHCI_P_CMD)); /* Handle frozen command. */ if (ch->frozen) { @@ -1987,7 +1994,7 @@ ahci_timeout(struct ahci_slot *slot) } xpt_done(fccb); } - if (!ch->fbs_enabled) { + if (!ch->fbs_enabled && !ch->wrongccs) { /* Without FBS we know real timeout source. */ ch->fatalerr = 1; /* Handle command with timeout. */ @@ -2585,6 +2592,7 @@ ahci_reset(device_t dev) xpt_release_simq(ch->sim, TRUE); ch->eslots = 0; ch->toslots = 0; + ch->wrongccs = 0; ch->fatalerr = 0; /* Tell the XPT about the event */ xpt_async(AC_BUS_RESET, ch->path, NULL); diff --git a/sys/dev/ahci/ahci.h b/sys/dev/ahci/ahci.h index 3ace69fff925..559c67ab0e4a 100644 --- a/sys/dev/ahci/ahci.h +++ b/sys/dev/ahci/ahci.h @@ -426,6 +426,7 @@ struct ahci_channel { int resetting; /* Hard-reset in progress. */ int resetpolldiv; /* Hard-reset poll divider. */ int listening; /* SUD bit is cleared. */ + int wrongccs; /* CCS field in CMD was wrong */ union ccb *frozen; /* Frozen command */ struct callout pm_timer; /* Power management events */ struct callout reset_timer; /* Hard-reset timeout */