From: Jeff Garzik <jgarzik@pobox.com>
To: Evgeny Rodichev <er@sai.msu.su>, Brett Russ <russb@emc.com>
Cc: linux-kernel@vger.kernel.org
Subject: Re: [PATCH 2.6.14-rc2] libata: Marvell SATA support (DMA mode) (resend: v0.22)
Date: Wed, 05 Oct 2005 13:04:34 -0400 [thread overview]
Message-ID: <434407A2.6010605@pobox.com> (raw)
In-Reply-To: <Pine.GSO.4.63.0510051912230.10241@ra.sai.msu.su>
[-- Attachment #1: Type: text/plain, Size: 654 bytes --]
Evgeny Rodichev wrote:
> On Mon, 3 Oct 2005, Jeff Garzik wrote:
>
>> applied
>>
>
> This patch leads to freeze with Marvell MV88SX6041 4-port SATA II PCI-X
> Controller (rev 03), as I wrote already
> (http://www.uwsg.iu.edu/hypermail/linux/kernel/0510.0/0203.html)
>
> It is impossible to reboot the system without harware reset (after
> modprobe).
I'm betting a temporary workaround should be to disable CONFIG_SMP.
The first attached patch (patch.1) fixes one of the lockups. The second
attached patch (patch.2) should fix all the lockups, but that's just a
guess, and it's a very ugly patch.
Note that patch.2 depends on patch.1.
Jeff
[-- Attachment #2: patch.1 --]
[-- Type: text/plain, Size: 545 bytes --]
diff --git a/drivers/scsi/sata_mv.c b/drivers/scsi/sata_mv.c
--- a/drivers/scsi/sata_mv.c
+++ b/drivers/scsi/sata_mv.c
@@ -409,14 +409,8 @@ static void mv_irq_clear(struct ata_port
static void mv_start_dma(void __iomem *base, struct mv_port_priv *pp,
struct ata_port *ap)
{
- unsigned long flags;
-
- spin_lock_irqsave(&ap->host_set->lock, flags);
-
writelfl(EDMA_EN, base + EDMA_CMD_OFS);
pp->pp_flags |= MV_PP_FLAG_EDMA_EN;
-
- spin_unlock_irqrestore(&ap->host_set->lock, flags);
}
static void mv_stop_dma(struct ata_port *ap)
[-- Attachment #3: patch.2 --]
[-- Type: text/plain, Size: 1815 bytes --]
diff --git a/drivers/scsi/sata_mv.c b/drivers/scsi/sata_mv.c
--- a/drivers/scsi/sata_mv.c
+++ b/drivers/scsi/sata_mv.c
@@ -413,16 +413,13 @@ static void mv_start_dma(void __iomem *b
pp->pp_flags |= MV_PP_FLAG_EDMA_EN;
}
-static void mv_stop_dma(struct ata_port *ap)
+static void __mv_stop_dma(struct ata_port *ap)
{
void __iomem *port_mmio = mv_ap_base(ap);
struct mv_port_priv *pp = ap->private_data;
- unsigned long flags;
u32 reg;
int i;
- spin_lock_irqsave(&ap->host_set->lock, flags);
-
if (!(MV_PP_FLAG_EDMA_DS_ACT & pp->pp_flags) &&
((MV_PP_FLAG_EDMA_EN & pp->pp_flags) ||
(EDMA_EN & readl(port_mmio + EDMA_CMD_OFS)))) {
@@ -433,7 +430,6 @@ static void mv_stop_dma(struct ata_port
writelfl(EDMA_DS, port_mmio + EDMA_CMD_OFS);
pp->pp_flags &= ~MV_PP_FLAG_EDMA_EN;
}
- spin_unlock_irqrestore(&ap->host_set->lock, flags);
/* now properly wait for the eDMA to stop */
for (i = 1000; i > 0; i--) {
@@ -444,15 +440,22 @@ static void mv_stop_dma(struct ata_port
udelay(100);
}
- spin_lock_irqsave(&ap->host_set->lock, flags);
pp->pp_flags &= ~MV_PP_FLAG_EDMA_DS_ACT;
- spin_unlock_irqrestore(&ap->host_set->lock, flags);
if (EDMA_EN & reg) {
printk(KERN_ERR "ata%u: Unable to stop eDMA\n", ap->id);
}
}
+static void mv_stop_dma(struct ata_port *ap)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&ap->host_set->lock, flags);
+ __mv_stop_dma(ap);
+ spin_unlock_irqrestore(&ap->host_set->lock, flags);
+}
+
static void mv_dump_mem(void __iomem *start, unsigned bytes)
{
#ifdef ATA_DEBUG
@@ -845,7 +848,7 @@ static int mv_qc_issue(struct ata_queued
* port. Turn off EDMA so there won't be problems accessing
* shadow block, etc registers.
*/
- mv_stop_dma(qc->ap);
+ __mv_stop_dma(qc->ap);
return ata_qc_issue_prot(qc);
}
next prev parent reply other threads:[~2005-10-05 17:04 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-09-30 5:36 [PATCH 2.6.14-rc2] libata: Marvell SATA support (DMA mode) (resend: v0.22) Brett Russ
2005-09-30 11:37 ` Lars Magne Ingebrigtsen
2005-10-04 2:08 ` Jeff Garzik
2005-10-05 15:17 ` Evgeny Rodichev
2005-10-05 17:04 ` Jeff Garzik [this message]
2005-10-05 17:10 ` Brett Russ
2005-10-05 19:14 ` Bogdan Costescu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=434407A2.6010605@pobox.com \
--to=jgarzik@pobox.com \
--cc=er@sai.msu.su \
--cc=linux-kernel@vger.kernel.org \
--cc=russb@emc.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).