[Fwd: mvsas determines most often an incorrect phy addr of 0 while attaching phy to port (Was: ERROR: Unidentified device type 0)]

[Fwd: mvsas determines most often an incorrect phy addr of 0 while attaching phy to port (Was: ERROR: Unidentified device type 0)]

[Reinhard Nissl]

Hi,

sorry for cross posting but it seems to me that this list is more
appropriate for my issue.

-------- Original-Nachricht --------
Betreff: mvsas determines most often an incorrect phy addr of 0
while attaching phy to port (Was: ERROR: Unidentified device type 0)
Datum: Sun, 07 Dec 2008 00:01:25 +0100
Von: Reinhard Nissl <rnissl@gmx.de>
An: linux-kernel@vger.kernel.org
CC: kewei@marvell.com
Referenzen: <4939ABA5.4090402@gmx.de>

Hi,

I've meanwhile investigated this issue a little bit further and
changed the subject accordingly.

Reinhard Nissl schrieb:

> I've got the ASUS P6T WS Professional mainboard including Marvell
> 88SE6320 SAS controller and attached a single Seagate 10k6 SAS
> harddrive to the first SAS port.
> 
> Installation of openSUSE 11.1rc1 went well but after rebooting,
> the harddrive doesn't get found. When the initrd script loads the
> mvsas module, the following output appears on screen:
> 
> sas: phy-10:1 added to port-10:0, phy_mask:0x2 (             0)
> ERROR: Unidentified device type 0
> 
> I first tried to install openSUSE 11.0 but the installer didn't
> find the harddrive. If I recall correctly, it was for the same
> reason as it is now when trying to boot the 11.1rc1 system.

This was not completely correct. I forgot that the 11.0 installer
saw the drive at least once. And my investigations today showed
that the 11.1rc1 installer / rescue system can run into the same
error. But most often they see the drive (sometimes only after a
power cycle) while initrd didn't succeed at all (looks like a
timing issue).

> So I think, there exists already a fix for this issue, as the
> 11.1rc1 installer sees the drive, which is still missing in the
> initrd environment.
> 
> I really would like to help solving this issue, but I need a
> pointer where and how to start digging. I've started a openSUSE
> bug (https://bugzilla.novell.com/show_bug.cgi?id=450302, Bug
> 450302 - mvsas doesn't find SAS drive anymore after installation)
> almost a week ago but didn't get much help there.

Some more findings:
After the rescue system booted successfully and detected the
drive, I tried to unload mvsas. modprobe -r mvsas didn't succeed
as the module was still in use, but lsmod didn't report any user.
So I've unloaded mvsas with rmmod -f mvsas which succeeded. After
that, modprobe mvsas was loading the module successfully again
but the above error appeared. I've put up a loop with modprobe -r
/ modprobe but it never succeeded to detect the drive anymore.

My plans for the next days are to attach a SATA drive to one of
the south bridge SATA ports, install an openSUSE system there
which I will use to investigate this issue further, i. e. setup a
build environment to be able to compile the module, activate (and
possibly add additional) debug output and load the module without
having to reboot.

I really hope that someone will drop me some lines to guide me in
investigating this issue. Thanks in advance.

Bye.
-- 
Dipl.-Inform. (FH) Reinhard Nissl
mailto:rnissl@gmx.de

PATCH: mvsas: increase port type detection delay to suit Seagate's 10k6 drive ST3450856SS 0003

[Reinhard Nissl]

[-- Attachment #1: Type: text/plain, Size: 418 bytes --]

I increased the delay step by step until loading of mvsas
reliably detected the drive 200 times in sequence. A much better
approach would be to monitor the hardware for some flag which
indicates that port detection has finished, but I do not have any
hardware documentation.

BTW: patch is against
git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-2.6.27.y.git

Signed-off-by: Reinhard Nissl <rnissl@gmx.de>


[-- Attachment #2: mvsas.diff --]
[-- Type: text/x-patch, Size: 450 bytes --]

diff --git a/drivers/scsi/mvsas.c b/drivers/scsi/mvsas.c
index 1dd70d7..23e5a87 100644
--- a/drivers/scsi/mvsas.c
+++ b/drivers/scsi/mvsas.c
@@ -2959,7 +2959,7 @@ static int __devinit mvs_hw_init(struct mvs_info *mvi)
 
 	/* enable auto port detection */
 	mw32(GBL_PORT_TYPE, MODE_AUTO_DET_EN);
-	msleep(100);
+	msleep(1100);
 	/* init and reset phys */
 	for (i = 0; i < mvi->chip->n_phy; i++) {
 		u32 lo = be32_to_cpu(*(u32 *)&mvi->sas_addr[4]);

Re: PATCH: mvsas: increase port type detection delay to suit Seagate's 15k6 drive ST3450856SS 0003

[Reinhard Nissl]

[-- Attachment #1: Type: text/plain, Size: 744 bytes --]

Hi,

I wonder why no one has dropped me a line so far. I hope you
don't mind CCing you directly. I found your email addresses on
mailing list entries regarding mvsas. I attach the little patch
again for ease of use.

Reinhard Nissl schrieb:

> I increased the delay step by step until loading of mvsas
> reliably detected the drive 200 times in sequence. A much better
> approach would be to monitor the hardware for some flag which
> indicates that port detection has finished, but I do not have any
> hardware documentation.
> 
> BTW: patch is against
> git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-2.6.27.y.git
> 
> Signed-off-by: Reinhard Nissl <rnissl@gmx.de>

Bye.
-- 
Dipl.-Inform. (FH) Reinhard Nissl
mailto:rnissl@gmx.de

[-- Attachment #2: mvsas.diff --]
[-- Type: text/x-patch, Size: 450 bytes --]

diff --git a/drivers/scsi/mvsas.c b/drivers/scsi/mvsas.c
index 1dd70d7..23e5a87 100644
--- a/drivers/scsi/mvsas.c
+++ b/drivers/scsi/mvsas.c
@@ -2959,7 +2959,7 @@ static int __devinit mvs_hw_init(struct mvs_info *mvi)
 
 	/* enable auto port detection */
 	mw32(GBL_PORT_TYPE, MODE_AUTO_DET_EN);
-	msleep(100);
+	msleep(1100);
 	/* init and reset phys */
 	for (i = 0; i < mvi->chip->n_phy; i++) {
 		u32 lo = be32_to_cpu(*(u32 *)&mvi->sas_addr[4]);