Home » Mailing list archives » linux-kernel » 2008 » February » 13

Re: [PATCH] pata_cs5536 Fix secondary port configuration

Previous thread: [PATCH 4/8] sched: rt-group: make rt groups scheduling configurable by Peter Zijlstra on Monday, February 4, 2008 - 5:03 pm. (1 message)

Next thread: [drivers/misc/thinkpad_acpi.c] duplicate test if (level & TP_EC_FAN_FULLSPEED) by Roel Kluin on Monday, February 4, 2008 - 6:07 pm. (11 messages)
[view original message]
From: Gregor Radtke <mail@...>
To: <linux-kernel@...>, <mkp@...>
Cc: Tim Kunschke <beatmeistah@...>
Subject: pata_cs5536 hangs on boot with hdd attached
Date: Monday, February 4, 2008 - 5:00 pm

--=-Gf+l7GLQkygQGNdWrZqr

Content-Type: text/plain

Content-Transfer-Encoding: quoted-printable


Hi all,


there is some serious problem with pata_cs5536 in linux 2.6.24 - it does

not boot on ALIX.1B and ALIX.2C2 with hdd attached (as slave) to ide

port.


the ALIX.* are a series of single board computers, manufactured by PC

Engines in switzerland. Systems are based on AMD Geode LX, which itself

utilizes a CS5536 for IDE. ALIX.1B/1C use Award BIOS, while ALIX.2C2 and

other use tinyBIOS. See http://www.pcengines.ch/alix.htm for more

details. I have tinyBIOS v0.99 installed, using older versions makes no

difference.


I tested with various kernel releases before 2.6.24 (and -rcX), up to

2.6.23.12. Both libata and the old ATA subsystem do work, libata with

cs5535.


cs5536 _does work_, if i boot with no hdd attached... i even get a

(relatively) high  transferrate, about 11-12MB/s, compared to 6-7MB/s

using cs5535/libata/old subsys.


the system hangs when it comes to examine partitions, or so.. see


-- snip --

[   13.391699] Driver 'sd' needs updating - please use bus_type methods

[   13.411207] scsi0 : pata_cs5536

[   13.420886] scsi1 : pata_cs5536

[   13.430497] ata1: PATA max UDMA/100 cmd 0x1f0 ctl 0x3f6 bmdma 0xff00

irq 14

[   13.451386] ata2: DUMMY

[   13.636238] ata1.00: CFA: SanDisk SDCFH-1024, HDX 4.04, max MWDMA2

[   13.654793] ata1.00: 2001888 sectors, multi 0: LBA=20

[   13.669749] ata1.01: ATA-6: IC25N060ATMR04-0, MO3OAD4A, max UDMA/100

[   13.688817] ata1.01: 117210240 sectors, multi 0: LBA48=20

[   13.704524] ata1.01: limited to UDMA/33 due to 40-wire cable

[   13.722241] ata1.00: configured for MWDMA2

[   13.765912] ata1.01: configured for UDMA/33

[   13.778807] scsi 0:0:0:0: Direct-Access     ATA      SanDisk SDCFH-10

HDX  PQ: 0 ANSI: 5

[   13.803511] sd 0:0:0:0: [sda] 2001888 512-byte hardware sectors (1025

MB)

[   13.823905] sd 0:0:0:0: [sda] Write Protect is off

[   13.838368] sd 0:0:0:0: [sda] Write cache: disabled, read cache:

enabled, doesn't support DP...
[ message continues ]
[view original message]
From: Martin K. Petersen <martin.petersen@...>
To: Gregor Radtke <mail@...>
Cc: <linux-kernel@...>, Tim Kunschke <beatmeistah@...>
Subject: Re: pata_cs5536 hangs on boot with hdd attached
Date: Tuesday, February 5, 2008 - 2:16 am

>>>>> "Gregor" == Gregor Radtke <mail@g-radtke.name> writes:


Gregor,


Gregor> I tested with various kernel releases before 2.6.24 (and

Gregor> -rcX), up to 2.6.23.12. Both libata and the old ATA subsystem

Gregor> do work, libata with cs5535.


pata_cs5535 won't drive a 5536 -- they're completely different beasts.

So you haven't been testing what you think you're testing...


Try booting your pata_cs5536 kernel with libata.dma=0


I also suggest you try a kernel with pata_amd compiled in instead of

pata_cs5536.  Slight timings aside there shouldn't be any functional

difference between the way the two drivers configure things.  But let

us know how that goes.  And include the diff of the relevant portion

of the boot logs.


Finally - which drive is it?


--

Martin K. Petersen	Oracle Linux Engineering

--
[view original message]
From: Gregor Radtke <mail@...>
To: Martin K. Petersen <martin.petersen@...>
Cc: <linux-kernel@...>, Tim Kunschke <beatmeistah@...>
Subject: Re: pata_cs5536 hangs on boot with hdd attached
Date: Sunday, February 10, 2008 - 1:00 pm

hmm.. boots perfectly, of course somewhat slow.

dmesg output:


[   26.728891] Driver 'sd' needs updating - please use bus_type methods

[   26.748371] scsi0 : pata_cs5536

[   26.758052] scsi1 : pata_cs5536

[   26.767672] ata1: PATA max UDMA/100 cmd 0x1f0 ctl 0x3f6 bmdma 0xff00

irq 14

[   26.788551] ata2: DUMMY

[   26.975641] ata1.00: CFA: SanDisk SDCFH-1024, HDX 4.04, max MWDMA2

[   26.994194] ata1.00: 2001888 sectors, multi 0: LBA

[   27.011257] ata1.01: ATA-8: SAMSUNG HM160HC, LQ100-10, max UDMA/100

[   27.030069] ata1.01: 312581808 sectors, multi 0: LBA48

[   27.045774] ata1.01: limited to UDMA/33 due to 40-wire cable

[   27.063457] ata1.00: configured for PIO4

[   27.116887] ata1.01: configured for PIO4

[   27.128995] scsi 0:0:0:0: Direct-Access     ATA      SanDisk SDCFH-10

HDX  PQ: 0 ANSI: 5

[   27.153703] sd 0:0:0:0: [sda] 2001888 512-byte hardware sectors (1025

MB)

[   27.174096] sd 0:0:0:0: [sda] Write Protect is off

[   27.188485] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00

[   27.188561] sd 0:0:0:0: [sda] Write cache: disabled, read cache:

enabled, doesn't support DPO or FUA

[   27.216134] sd 0:0:0:0: [sda] 2001888 512-byte hardware sectors (1025

MB)

[   27.236528] sd 0:0:0:0: [sda] Write Protect is off

[   27.250894] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00

[   27.250968] sd 0:0:0:0: [sda] Write cache: disabled, read cache:

enabled, doesn't support DPO or FUA

[   27.278351]  sda: sda1

[   27.288400] sd 0:0:0:0: [sda] Attached SCSI removable disk

[   27.305230] scsi 0:0:1:0: Direct-Access     ATA      SAMSUNG HM160HC

LQ10 PQ: 0 ANSI: 5

[   27.329903] sd 0:0:1:0: [sdb] 312581808 512-byte hardware sectors

(160042 MB)

[   27.351336] sd 0:0:1:0: [sdb] Write Protect is off

[   27.365720] sd 0:0:1:0: [sdb] Mode Sense: 00 3a 00 00

[   27.365795] sd 0:0:1:0: [sdb] Write cache: enabled, read cache:

enabled, doesn't support DPO or FUA

[   27.393050] sd 0:0:1:0: [sdb] 312581808 512-byte hardware sectors

(160042 MB)

[   27.414491] sd 0:0:1:0: [sdb] Write Protect is off

[ ...
[ message continues ]
[view original message]
From: Martin K. Petersen <martin.petersen@...>
To: Gregor Radtke <mail@...>
Cc: Martin K. Petersen <martin.petersen@...>, <linux-kernel@...>, Tim Kunschke <beatmeistah@...>
Subject: [PATCH] pata_cs5536 Fix secondary port configuration
Date: Tuesday, February 12, 2008 - 10:37 am

Gregor> hmm.. boots perfectly, of course somewhat slow.  dmesg output:


Brown paper bag time.  My devel board has a 16MB 1st generation

compact flash on the secondary so I never noticed this was hosed.

Please try the following patch...


--

Martin K. Petersen	Oracle Linux Engineering


Fix speed negotiation for slave device.


Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>


---


diff -r 9f5ca67cc28f drivers/ata/pata_cs5536.c

--- a/drivers/ata/pata_cs5536.c	Mon Feb 11 20:52:01 2008 -0800

+++ b/drivers/ata/pata_cs5536.c	Tue Feb 12 08:52:43 2008 -0500

@@ -40,7 +40,7 @@

 #include <asm/msr.h>


 #define DRV_NAME	"pata_cs5536"

-#define DRV_VERSION	"0.0.6"

+#define DRV_VERSION	"0.0.7"


 enum {

 	CFG			= 0,

@@ -153,8 +153,8 @@ static void cs5536_set_piomode(struct at

 	struct ata_device *pair = ata_dev_pair(adev);

 	int mode = adev->pio_mode - XFER_PIO_0;

 	int cmdmode = mode;

-	int dshift = ap->port_no ? IDE_D1_SHIFT : IDE_D0_SHIFT;

-	int cshift = ap->port_no ? IDE_CAST_D1_SHIFT : IDE_CAST_D0_SHIFT;

+	int dshift = adev->devno ? IDE_D1_SHIFT : IDE_D0_SHIFT;

+	int cshift = adev->devno ? IDE_CAST_D1_SHIFT : IDE_CAST_D0_SHIFT;

 	u32 dtc, cast, etc;


 	if (pair)

@@ -201,7 +201,7 @@ static void cs5536_set_dmamode(struct at

 	struct pci_dev *pdev = to_pci_dev(ap->host->dev);

 	u32 dtc, etc;

 	int mode = adev->dma_mode;

-	int dshift = ap->port_no ? IDE_D1_SHIFT : IDE_D0_SHIFT;

+	int dshift = adev->devno ? IDE_D1_SHIFT : IDE_D0_SHIFT;


 	if (mode >= XFER_UDMA_0) {

 		cs5536_read(pdev, ETC, &etc);


--
[view original message]
From: Gregor Radtke <mail@...>
To: Martin K. Petersen <martin.petersen@...>
Cc: <linux-kernel@...>, Tim Kunschke <beatmeistah@...>
Subject: Re: [PATCH] pata_cs5536 Fix secondary port configuration
Date: Tuesday, February 12, 2008 - 7:17 pm

boots! For some reason, i had to apply it by hand (to 2.6.24.2), and i

was too stupid to examine how to produce a new patch. Looks like i have

to read the documentation at kernelnewbies.org...


Anyway, the patch fits perfectly and now UDMA/33 (due to 40wire-cable,

vendor says UDMA/100 should work either) on a HDD using pata_cs5536 just

works =3D)


--=20

Jabber: elch@ein-elch.de        PGP: 0x63FC728D

Tel: 0170/238 30 53    http://www.g-radtke.name
[view original message]
From: Martin K. Petersen <martin.petersen@...>
To: Gregor Radtke <mail@...>
Cc: Martin K. Petersen <martin.petersen@...>, <linux-kernel@...>, Tim Kunschke <beatmeistah@...>
Subject: Re: [PATCH] pata_cs5536 Fix secondary port configuration
Date: Wednesday, February 13, 2008 - 3:12 am

>>>>> "Gregor" == Gregor Radtke <mail@g-radtke.name> writes:


Gregor> boots! 


Excellent.  Thanks for testing this!  I have sent the patch upstream.


Gregor> Anyway, the patch fits perfectly and now UDMA/33 (due to

Gregor> 40wire-cable, vendor says UDMA/100 should work either) on a

Gregor> HDD using pata_cs5536 just works =)


pata_amd.c happens to ignore the 40/80 wire configuration bit on

CS5536 and a couple of other chips.  That's why it negotiates

UDMA/100 regardless of cable type.


pata_cs5536.c relies on the BIOS telling it whether an 80-wire cable

is connected or not.


I have attached a patch that dumps the 5536 IDE configuration

registers during boot. If you like you can apply that and mail me the

resulting register dump.  That'll tell us how your BIOS initialized

the chip.


--

Martin K. Petersen	Oracle Linux Engineering


diff -r b0dc16d276be drivers/ata/pata_cs5536.c

--- a/drivers/ata/pata_cs5536.c	Tue Feb 12 08:52:44 2008 -0500

+++ b/drivers/ata/pata_cs5536.c	Wed Feb 13 01:51:00 2008 -0500

@@ -73,6 +73,11 @@ enum {

 	IDE_CAST_CMD_SHIFT	= 24,


 	IDE_ETC_NODMA		= 0x03,

+

+	CFG_REG_MASK		= 0x34002,

+	DTC_REG_MASK		= 0xffff0000,

+	CAST_REG_MASK		= 0xff0000f0,

+	ETC_REG_MASK		= 0xc7c70000,

 };


 static int use_msr;

@@ -84,6 +89,23 @@ static const u8 pci_reg[4] = {

 static const u8 pci_reg[4] = {

 	PCI_IDE_CFG, PCI_IDE_DTC, PCI_IDE_CAST, PCI_IDE_ETC,

 };

+

+static void cs5536_regs(void)

+{

+        u32 reg, dummy;

+

+        rdmsr(MSR_IDE_CFG, reg, dummy);

+        printk(KERN_ERR " CFG  = 0x%08x\n", reg & CFG_REG_MASK);

+

+        rdmsr(MSR_IDE_DTC, reg, dummy);

+        printk(KERN_ERR " DTC  = 0x%08x\n", reg & DTC_REG_MASK);

+

+        rdmsr(MSR_IDE_CAST, reg, dummy);

+        printk(KERN_ERR " CAST = 0x%08x\n", reg & CAST_REG_MASK);

+

+        rdmsr(MSR_IDE_ETC, reg, dummy);

+        printk(KERN_ERR " ETC  = 0x%08x\n", reg & ETC_REG_MASK);

+}


 static inline int cs5536_read(struct pci_dev ...
[ message continues ]
Previous thread: [PATCH 4/8] sched: rt-group: make rt groups scheduling configurable by Peter Zijlstra on Monday, February 4, 2008 - 5:03 pm. (1 message)

Next thread: [drivers/misc/thinkpad_acpi.c] duplicate test if (level & TP_EC_FAN_FULLSPEED) by Roel Kluin on Monday, February 4, 2008 - 6:07 pm. (11 messages)