[PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Milan Kupcevic]

From: Milan Kupcevic <milan@physics.harvard.edu>

Fix Promise SATAII 150 TX4 (PDC40518) and Promise SATA 300 TX4 
(PDC40718-GP) wrong port enumeration order that makes it (nearly) 
impossible to deal with boot problems using two or more drives.

Signed-off-by: Milan Kupcevic <milan@physics.harvard.edu>
---

The current kernel driver assumes:

port 1 - scsi3
port 2 - scsi1
port 3 - scsi0
port 4 - scsi2

Having 4 hard drives connected to the controller grub recognizes the 
port 1 connected drive as "(hd0)" but kernel recognizes the port 3 
connected drive as scsi0:0:0:0 (/dev/sda). There is no clean way to make 
it boot correctly.

 sata_promise.c |   16 ++++++++++------
 1 files changed, 10 insertions(+), 6 deletions(-)

diff -uprN ./drivers/scsi/sata_promise.c 
../linux/drivers/scsi/sata_promise.c
--- ./drivers/scsi/sata_promise.c       2006-02-17 17:23:45.000000000 -0500
+++ ../linux/drivers/scsi/sata_promise.c        2006-02-24 
19:35:16.000000000 -0500
@@ -707,14 +707,18 @@ static int pdc_ata_init_one (struct pci_
 
        /* notice 4-port boards */
        switch (board_idx) {
-       case board_20319:
+       case board_20319: /* tx4  */
                        probe_ent->n_ports = 4;
 
-               pdc_ata_setup_port(&probe_ent->port[2], base + 0x300);
-               pdc_ata_setup_port(&probe_ent->port[3], base + 0x380);
-
-               probe_ent->port[2].scr_addr = base + 0x600;
-               probe_ent->port[3].scr_addr = base + 0x700;
+               pdc_ata_setup_port(&probe_ent->port[0], base + 0x380);
+               pdc_ata_setup_port(&probe_ent->port[1], base + 0x280);
+               pdc_ata_setup_port(&probe_ent->port[2], base + 0x200);
+               pdc_ata_setup_port(&probe_ent->port[3], base + 0x300);
+
+               probe_ent->port[0].scr_addr = base + 0x700;
+               probe_ent->port[1].scr_addr = base + 0x500;
+               probe_ent->port[2].scr_addr = base + 0x400;
+               probe_ent->port[3].scr_addr = base + 0x600;
                break;
        case board_2037x:
                probe_ent->n_ports = 2;

-- 
Milan Kupcevic
System Administrator
Harvard University
Department of Physics

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Jeff Garzik]

Milan Kupcevic wrote:
> From: Milan Kupcevic <milan@physics.harvard.edu>
> 
> Fix Promise SATAII 150 TX4 (PDC40518) and Promise SATA 300 TX4 
> (PDC40718-GP) wrong port enumeration order that makes it (nearly) 
> impossible to deal with boot problems using two or more drives.
> 
> Signed-off-by: Milan Kupcevic <milan@physics.harvard.edu>
> ---
> 
> The current kernel driver assumes:
> 
> port 1 - scsi3
> port 2 - scsi1
> port 3 - scsi0
> port 4 - scsi2

The current kernel driver assumes nothing, but simply exports what the 
hardware gives us.

It sounds like you are trying to patch the kernel because you received 
an incorrectly-wired board.  NAK.

	Jeff

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Milan Kupcevic]

Jeff Garzik wrote:

> Milan Kupcevic wrote:
>
>> From: Milan Kupcevic <milan@physics.harvard.edu>
>>
>> Fix Promise SATAII 150 TX4 (PDC40518) and Promise SATA 300 TX4 
>> (PDC40718-GP) wrong port enumeration order that makes it (nearly) 
>> impossible to deal with boot problems using two or more drives.
>>
>> Signed-off-by: Milan Kupcevic <milan@physics.harvard.edu>
>> ---
>>
>> The current kernel driver assumes:
>>
>> port 1 - scsi3
>> port 2 - scsi1
>> port 3 - scsi0
>> port 4 - scsi2
>
>
> The current kernel driver assumes nothing, but simply exports what the 
> hardware gives us.
>
> It sounds like you are trying to patch the kernel because you received 
> an incorrectly-wired board.  NAK.
>
>     Jeff
>

I have tested two SATAII150TX4 (chip PDC40518 id: 105a:3d18 (rev 02)) 
adapters and one SATA300TX4 (chip PDC40718-GP id: 105a:3d17 (rev 02)) 
adapter on three different boards in last several days with exactly the 
same results.

The problem disappears when using the driver form the www.promise.com site.

BIOS, Grub and promise.com driver agree the printed and documented port 
order is correct but the current kernel driver exports (wrong) 3-2-4-1 
port order.


Thanks,

Milan

-- 
Milan Kupcevic
System Administrator
Harvard University
Department of Physics

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Eyal Lebedinsky]

Milan Kupcevic wrote:
> From: Milan Kupcevic <milan@physics.harvard.edu>
> 
> Fix Promise SATAII 150 TX4 (PDC40518) and Promise SATA 300 TX4
> (PDC40718-GP) wrong port enumeration order that makes it (nearly)
> impossible to deal with boot problems using two or more drives.
> 
> Signed-off-by: Milan Kupcevic <milan@physics.harvard.edu>
> ---
> 
> The current kernel driver assumes:
> 
> port 1 - scsi3
> port 2 - scsi1
> port 3 - scsi0
> port 4 - scsi2

I totally agree with the fact that the Linux driver gets the ports wrong
when compared to the BIOS, Windows and surely contradicts the port
numbers printed on the board. I doubt we all got samples on the one
bad batch...

It *is* a real problem and if the solution is correct then I support it.

Maybe we need a quick feedback from current users: do you guys find
that the ports are detected as they are labelled (white silk screen)
on the board or do they show up out of order (as listed above by
Milan)?

-- 
Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.org/eyal/>
	attach .zip as .dat

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Christiaan den Besten]

Hi !

We have several of these boards in use [SATA 300 TX4] (bought over time .. not in one batch). All of them have the ordering as 
described below. So another vote for "Please fix!" :)

bye,
Chris

----- Original Message ----- 
From: "Eyal Lebedinsky" <eyal@eyal.emu.id.au>
To: "Milan Kupcevic" <milan@physics.harvard.edu>
Cc: "Jeff Garzik" <jgarzik@pobox.com>; <linux-ide@vger.kernel.org>; <linux-scsi@vger.kernel.org>; <linux-kernel@vger.kernel.org>; 
<trivial@rustcorp.com.au>; <torvalds@osdl.org>
Sent: Saturday, February 25, 2006 7:59 AM
Subject: Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4


> Milan Kupcevic wrote:
>> From: Milan Kupcevic <milan@physics.harvard.edu>
>>
>> Fix Promise SATAII 150 TX4 (PDC40518) and Promise SATA 300 TX4
>> (PDC40718-GP) wrong port enumeration order that makes it (nearly)
>> impossible to deal with boot problems using two or more drives.
>>
>> Signed-off-by: Milan Kupcevic <milan@physics.harvard.edu>
>> ---
>>
>> The current kernel driver assumes:
>>
>> port 1 - scsi3
>> port 2 - scsi1
>> port 3 - scsi0
>> port 4 - scsi2
>
> I totally agree with the fact that the Linux driver gets the ports wrong
> when compared to the BIOS, Windows and surely contradicts the port
> numbers printed on the board. I doubt we all got samples on the one
> bad batch...
>
> It *is* a real problem and if the solution is correct then I support it.
>
> Maybe we need a quick feedback from current users: do you guys find
> that the ports are detected as they are labelled (white silk screen)
> on the board or do they show up out of order (as listed above by
> Milan)?
>
> -- 
> Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.org/eyal/>
> attach .zip as .dat
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
>
> 

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Milan Kupcevic]

Fabian Knittel wrote:

>In other words: The boards appear to be wired correctly (or maybe just
>uniformly the wrong way) and the bios, the windows driver and the closed
>source promise driver (reportedly) know how to handle it.
>
>We have a SATA 150 TX4 board with the same behaviour and would love to
>see this annoying little bug fixed in linux. :)
>
>  Fabian
>
>Christiaan den Besten wrote:
>  
>
>>We have several of these boards in use [SATA 300 TX4] (bought over time
>>.. not in one batch). All of them have the ordering as described below.
>>So another vote for "Please fix!" :)
>>
>>    
>>
>>>Milan Kupcevic wrote:
>>>
>>>      
>>>
>>>>From: Milan Kupcevic <milan@physics.harvard.edu>
>>>>
>>>>Fix Promise SATAII 150 TX4 (PDC40518) and Promise SATA 300 TX4
>>>>(PDC40718-GP) wrong port enumeration order that makes it (nearly)
>>>>impossible to deal with boot problems using two or more drives.
>>>>
>>>>Signed-off-by: Milan Kupcevic <milan@physics.harvard.edu>
>>>>---
>>>>
>>>>The current kernel driver assumes:
>>>>
>>>>port 1 - scsi3
>>>>port 2 - scsi1
>>>>port 3 - scsi0
>>>>port 4 - scsi2
>>>>        
>>>>
>>>I totally agree with the fact that the Linux driver gets the ports wrong
>>>when compared to the BIOS, Windows and surely contradicts the port
>>>numbers printed on the board. I doubt we all got samples on the one
>>>bad batch...
>>>
>>>It *is* a real problem and if the solution is correct then I support it.
>>>
>>>Maybe we need a quick feedback from current users: do you guys find
>>>that the ports are detected as they are labelled (white silk screen)
>>>on the board or do they show up out of order (as listed above by
>>>Milan)?
>>>      
>>>

This is a list of Promise SATA TX4 and FastTrak TX4xxx controllers, I 
have in my lab, affected with the "wiring" bug:

Retail name: SATAII150 TX4
Chip label: PDC40518  SATAII150
Vendor-Device number: 105a:3d18 (rev 02)
Wiring: NEW

Retail name: FastTrak TX4200
Chip label: PDC40519  RAID  SATAII150
Vendor-Device number: 105a:3519
Wiring: NEW

Retail name: SATA300 TX4
Chip label: PDC40718-GP  SATAII300
Vendor-Device number: 105a:3d17 (rev 02)
Wiring: NEW


This is the only one Promise TX4 controller, I have in my lab, that is 
working properly regarding the "wiring" bug with the current kernel driver:

Retail name: FastTrak S150 TX4
Chip label: PDC20319 RAID SATA 150
Vendor-Device number: 105a:3319 (rev 02)
Wiring: OLD


This is a list of Promise SATA TX2 and FastTrak TX2xxx controllers, I 
have in my lab, that are working correctly regarding the "wiring" bug 
with the current kernel driver:

Retail name: FastTrak S150 TX2plus
Chip label: PDC20371  SATA 150
Vendor-Device number: 105a:3371 (rev 02)

Retail name: SATA150 TX2plus
Chip label: PDC20375  SATA 150
Vendor-Device number: 105a:3375 (rev 02)

Retail name: FastTrak TX2200
Chip label: PDC20571  SATAII150
Vendor-Device number: 105a:3571 (rev 02)


It seems the problem exists on all newer Promise SATA TX4 and FastTrak 
TX4xxx controllers, so I refer to them as the "new wiring" Promise SATA 
controllers.  All the Promise SATA TX2 and FastTrak TX2xxx I have in my 
lab are working correctly with the current kernel driver, so it seems 
this "wiring" problem does not affect the TX2(xxx) controllers; only 
SATA TX4 and FastTrak TX4xxx are affected.

For driver to be able to distinguish the "new wiring" and the "old 
wiring" Promise TX4(xxx) controllers we need a feedback from the users 
that are aware of this problem.

Q. How to know if a controller has the "new wiring"? 
A. You need to be able to boot your testing machine using a hard drive 
not attached to the controller you are going to test.  Connect 4 
different size/brand/model SATA hard drives to the testing controller so 
you can see the particular order they are recognized by the BIOS and by 
the kernel (not patched for the "new wiring"). 
   Boot the machine and look for the BIOS recognized hard drive order.  
The BIOS recognized hard drive order always matches the order the hard 
drives are connected to, with respect to port number labels.  If you are 
testing FastTrak TX series controller you may need to press Ctrl-F (or 
Ctrl-A) to get into controllers' BIOS and then press "2" to see the BIOS 
recognized order.  Plain SATA models do not have controller specific 
BIOS and they will report the BIOS recognized order without user 
intervention.  You may want to press the "Pause" key on your keyboard to 
have enough time to read the text on the screen.  If you are arguing 
with your machine using a serial terminal, there is a Hold Screen button 
somewhere on the terminal keyboard. 
   When the machine boots up, type "cat /proc/scsi/scsi", it will show 
up the order hard drives are recognized by the kernel.  Make sure the 
kernel you are using is NOT patched for the "new wiring" bug.  If you 
have the "new wiring" case, the order will be 3-2-4-1; that means, the 
drive connected to the "port 3" and recognized as the third drive 
connected to the controller by the BIOS, will be seen by the kernel as 
first hard drive connected to this controller.  The second drive is 
always at second place, the fourth one goes at third place and the first 
one goes at fourth place.


Please respond with this data:

- Your Promise SATA controller retail name
- Chip label (PDCxxxxx)
- PCI vendor and device code as you can get with "lspci -n"
- Say if the controler has the new or the old wiring
 
Your feedback will be appreciated.


NOTE: the patch I have submitted ( 
http://marc.theaimsgroup.com/?l=linux-ide&m=114082978311290&w=2 ) is a 
solution that doesn't know about the older Promise SATA controllers, 
which are not affected with the "new wiring" problem, so the older 
controllers will appear screwed if you use it.

Hopefully we will collect enough info about all the SATA Promise 
controllers to distinguish the new and the old wiring controllers, then 
produce a new patch that will be a correct solution to the "new wiring" 
problem.


The best to all,

Milan

-- 
Milan Kupcevic
System Administrator
Harvard University
Department of Physics

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Fabian Knittel]

Milan Kupcevic wrote:
> Please respond with this data:
> 
> - Your Promise SATA controller retail name
> - Chip label (PDCxxxxx)
> - PCI vendor and device code as you can get with "lspci -n"
> - Say if the controler has the new or the old wiring

We only have one card and it matches one of yours:

  Retail name: SATAII150 TX4
  Chip label: PDC40518  SATAII150
  Vendor-Device number: 105a:3d18 (rev 02)
  Wiring: NEW

Greetings, Fabian
-- 
avona media GbR      | Neureuter Str. 5-7   | D-76185 Karlsruhe
fon +49 721 53169901 | fax +49 721 53169904 | www.avona.com

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Evan Harris]

I have a card that mirrors this one from your list:

Retail name: SATA300 TX4
Chip label: PDC40718-GP  SATAII300
Vendor-Device number: 105a:3d17 (rev 02)

Through testing, I've found linux 2.6.16 and 2.6.17 find the ports in this 
order (the list is ordered by linux detection):

1. silkscreen port 3
2. silkscreen port 2
3. silkscreen port 4
4. silkscreen port 1

Evan

On Tue, 30 May 2006, Milan Kupcevic wrote:

>
> Fabian Knittel wrote:
>
>> In other words: The boards appear to be wired correctly (or maybe just
>> uniformly the wrong way) and the bios, the windows driver and the closed
>> source promise driver (reportedly) know how to handle it.
>> 
>> We have a SATA 150 TX4 board with the same behaviour and would love to
>> see this annoying little bug fixed in linux. :)
>> 
>>  Fabian
>> 
>> Christiaan den Besten wrote:
>> 
>>> We have several of these boards in use [SATA 300 TX4] (bought over time
>>> .. not in one batch). All of them have the ordering as described below.
>>> So another vote for "Please fix!" :)
>>> 
>>> 
>>>> Milan Kupcevic wrote:
>>>> 
>>>> 
>>>>> From: Milan Kupcevic <milan@physics.harvard.edu>
>>>>> 
>>>>> Fix Promise SATAII 150 TX4 (PDC40518) and Promise SATA 300 TX4
>>>>> (PDC40718-GP) wrong port enumeration order that makes it (nearly)
>>>>> impossible to deal with boot problems using two or more drives.
>>>>> 
>>>>> Signed-off-by: Milan Kupcevic <milan@physics.harvard.edu>
>>>>> ---
>>>>> 
>>>>> The current kernel driver assumes:
>>>>> 
>>>>> port 1 - scsi3
>>>>> port 2 - scsi1
>>>>> port 3 - scsi0
>>>>> port 4 - scsi2
>>>>> 
>>>> I totally agree with the fact that the Linux driver gets the ports wrong
>>>> when compared to the BIOS, Windows and surely contradicts the port
>>>> numbers printed on the board. I doubt we all got samples on the one
>>>> bad batch...
>>>> 
>>>> It *is* a real problem and if the solution is correct then I support it.
>>>> 
>>>> Maybe we need a quick feedback from current users: do you guys find
>>>> that the ports are detected as they are labelled (white silk screen)
>>>> on the board or do they show up out of order (as listed above by
>>>> Milan)?
>>>> 
>
> This is a list of Promise SATA TX4 and FastTrak TX4xxx controllers, I have in 
> my lab, affected with the "wiring" bug:
>
> Retail name: SATAII150 TX4
> Chip label: PDC40518  SATAII150
> Vendor-Device number: 105a:3d18 (rev 02)
> Wiring: NEW
>
> Retail name: FastTrak TX4200
> Chip label: PDC40519  RAID  SATAII150
> Vendor-Device number: 105a:3519
> Wiring: NEW
>
> Retail name: SATA300 TX4
> Chip label: PDC40718-GP  SATAII300
> Vendor-Device number: 105a:3d17 (rev 02)
> Wiring: NEW
>
>
> This is the only one Promise TX4 controller, I have in my lab, that is 
> working properly regarding the "wiring" bug with the current kernel driver:
>
> Retail name: FastTrak S150 TX4
> Chip label: PDC20319 RAID SATA 150
> Vendor-Device number: 105a:3319 (rev 02)
> Wiring: OLD
>
>
> This is a list of Promise SATA TX2 and FastTrak TX2xxx controllers, I have in 
> my lab, that are working correctly regarding the "wiring" bug with the 
> current kernel driver:
>
> Retail name: FastTrak S150 TX2plus
> Chip label: PDC20371  SATA 150
> Vendor-Device number: 105a:3371 (rev 02)
>
> Retail name: SATA150 TX2plus
> Chip label: PDC20375  SATA 150
> Vendor-Device number: 105a:3375 (rev 02)
>
> Retail name: FastTrak TX2200
> Chip label: PDC20571  SATAII150
> Vendor-Device number: 105a:3571 (rev 02)
>
>
> It seems the problem exists on all newer Promise SATA TX4 and FastTrak TX4xxx 
> controllers, so I refer to them as the "new wiring" Promise SATA controllers. 
> All the Promise SATA TX2 and FastTrak TX2xxx I have in my lab are working 
> correctly with the current kernel driver, so it seems this "wiring" problem 
> does not affect the TX2(xxx) controllers; only SATA TX4 and FastTrak TX4xxx 
> are affected.
>
> For driver to be able to distinguish the "new wiring" and the "old wiring" 
> Promise TX4(xxx) controllers we need a feedback from the users that are aware 
> of this problem.
>
> Q. How to know if a controller has the "new wiring"? A. You need to be able 
> to boot your testing machine using a hard drive not attached to the 
> controller you are going to test.  Connect 4 different size/brand/model SATA 
> hard drives to the testing controller so you can see the particular order 
> they are recognized by the BIOS and by the kernel (not patched for the "new 
> wiring").   Boot the machine and look for the BIOS recognized hard drive 
> order.  The BIOS recognized hard drive order always matches the order the 
> hard drives are connected to, with respect to port number labels.  If you are 
> testing FastTrak TX series controller you may need to press Ctrl-F (or 
> Ctrl-A) to get into controllers' BIOS and then press "2" to see the BIOS 
> recognized order.  Plain SATA models do not have controller specific BIOS and 
> they will report the BIOS recognized order without user intervention.  You 
> may want to press the "Pause" key on your keyboard to have enough time to 
> read the text on the screen.  If you are arguing with your machine using a 
> serial terminal, there is a Hold Screen button somewhere on the terminal 
> keyboard.   When the machine boots up, type "cat /proc/scsi/scsi", it will 
> show up the order hard drives are recognized by the kernel.  Make sure the 
> kernel you are using is NOT patched for the "new wiring" bug.  If you have 
> the "new wiring" case, the order will be 3-2-4-1; that means, the drive 
> connected to the "port 3" and recognized as the third drive connected to the 
> controller by the BIOS, will be seen by the kernel as first hard drive 
> connected to this controller.  The second drive is always at second place, 
> the fourth one goes at third place and the first one goes at fourth place.
>
>
> Please respond with this data:
>
> - Your Promise SATA controller retail name
> - Chip label (PDCxxxxx)
> - PCI vendor and device code as you can get with "lspci -n"
> - Say if the controler has the new or the old wiring
>
> Your feedback will be appreciated.
>
>
> NOTE: the patch I have submitted ( 
> http://marc.theaimsgroup.com/?l=linux-ide&m=114082978311290&w=2 ) is a 
> solution that doesn't know about the older Promise SATA controllers, which 
> are not affected with the "new wiring" problem, so the older controllers will 
> appear screwed if you use it.
>
> Hopefully we will collect enough info about all the SATA Promise controllers 
> to distinguish the new and the old wiring controllers, then produce a new 
> patch that will be a correct solution to the "new wiring" problem.
>
>
> The best to all,
>
> Milan
>
> -- 
> Milan Kupcevic
> System Administrator
> Harvard University
> Department of Physics
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

Re: [PATCH] sata_promise: Port enumeration order - SATA 150 TX4, SATA 300 TX4

[Jeff Garzik]

Evan Harris wrote:
> 
> I have a card that mirrors this one from your list:
> 
> Retail name: SATA300 TX4
> Chip label: PDC40718-GP  SATAII300
> Vendor-Device number: 105a:3d17 (rev 02)
> 
> Through testing, I've found linux 2.6.16 and 2.6.17 find the ports in 
> this order (the list is ordered by linux detection):
> 
> 1. silkscreen port 3
> 2. silkscreen port 2
> 3. silkscreen port 4
> 4. silkscreen port 1
>> NOTE: the patch I have submitted ( 
>> http://marc.theaimsgroup.com/?l=linux-ide&m=114082978311290&w=2 ) is a 
>> solution that doesn't know about the older Promise SATA controllers, 
>> which are not affected with the "new wiring" problem, so the older 
>> controllers will appear screwed if you use it.
>>
>> Hopefully we will collect enough info about all the SATA Promise 
>> controllers to distinguish the new and the old wiring controllers, 
>> then produce a new patch that will be a correct solution to the "new 
>> wiring" problem.

Mikael Pettersson has been doing some excellent work recently on 
sata_promise.  If enough data has been collected on this sata_promise 
port enumeration problem, maybe the data could be collated and proposed 
via Mikael as a patch?

	Jeff