Problem with SiS Ide/Pata/Sata on 2.6.22-rc7-git4

Problem with SiS Ide/Pata/Sata on 2.6.22-rc7-git4

[Thomas Backlund]

Hi,
(Please CC me in any responses)
There are some weirdness going on with SiS IDE/Pata/Sata support...

After moving from a 2.6.17 based kernel with ide dupport built in to a 
2.6.22-rc7-git4 based kernel with fully modular ide/pata/sata we see the 
following problem on a system with ide disks:


If you modprobe sata_sis:
(it also pulls pata_sis, is that really intended)
  - both disks are detected ok

If you modprobe sata_sis/pata_sis and then sis5513:
(it also pulls pata_sis, is that really intended)
  - both disks are detected ok

If you modprobe only sis5513
  OR
If you modprobe only sis5513 and then sata_sis/pata_sis:
  - you get this:
    (the hdb drive does not get detected/usable)


<6>Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2
<6>ide: Assuming 33MHz system bus speed for PIO modes; override with 
idebus=xx
<5>SCSI subsystem initialized
<6>SIS5513: IDE controller at PCI slot 0000:00:02.5
<6>ACPI: PCI Interrupt 0000:00:02.5[A] -> GSI 16 (level, low) -> IRQ 21
<6>SIS5513: chipset revision 1
<6>SIS5513: not 100% native mode: will probe irqs later
<6>SIS5513: SiS 962/963 MuTIOL IDE UDMA133 controller
<6>    ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio
<6>    ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:pio, hdd:pio
<7>Probing IDE interface ide0...
<4>hdb: QUANTUM FIREBALLlct15 07, ATA DISK drive
<4>hdb: host side 80-wire cable detection failed, limiting max speed to 
UDMA33
<7>hdb: selected mode 0x42
<4>ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
<7>Probing IDE interface ide1...
<4>hdc: ATAPI-CD ROM-DRIVE-50MAX, ATAPI CD/DVD-ROM drive
<7>hdc: selected mode 0x42
<4>ide1 at 0x170-0x177,0x376 on irq 15
<6>hdc: ATAPI 50X CD-ROM drive, 128kB Cache, UDMA(33)
<6>Uniform CD-ROM driver Revision: 3.20
<7>libata version 2.21 loaded.
<6>sata_sis 0000:00:05.0: version 0.8
<6>ACPI: PCI Interrupt 0000:00:05.0[A] -> GSI 17 (level, low) -> IRQ 22
<6>sata_sis 0000:00:05.0: Detected SiS 180/181/964 chipset in SATA mode
<6>scsi0 : sata_sis
<6>scsi1 : sata_sis
<6>ata1: SATA max UDMA/133 cmd 0x0001ac00 ctl 0x0001b002 bmdma 
0x0001bc00 irq 0
<6>ata2: SATA max UDMA/133 cmd 0x0001b400 ctl 0x0001b802 bmdma 
0x0001bc08 irq 0
<6>ata1: SATA link down (SStatus 0 SControl 300)
<6>ata2: SATA link down (SStatus 0 SControl 300)



kernel config is here:
http://tmb.kkc.fi/Kernels/Cooker/BUG/SiS/config

and Mandriva Installer bug report here:
http://tmb.kkc.fi/Kernels/Cooker/BUG/SiS/report.bug

Note! this report is from a 2.6.22-tmb-desktop586-0.rc7.1mdv
Mandriva Cooker Contrib kernel wich is 2.6.22-rc7-git1 wich has several 
patches including CFS, HRT, and several other patches (but none of them 
touching ide/pata/sata)...

But we got the exact same behaviour with a clean 2.6.22-rc7-git4 + 
Squashfs-3.2-r2 (needed for the installer)

http://tmb.kkc.fi/Kernels/Cooker/i586/kernel-linus-2.6.22-0.rc7.3mdv-1-1mdv2007.1.i586.rpm

Any pointers of what/how to fix ?

--
Thomas

Re: Problem with SiS Ide/Pata/Sata on 2.6.22-rc7-git4

[Alan Cox]

> If you modprobe only sis5513
>   OR
> If you modprobe only sis5513 and then sata_sis/pata_sis:
>   - you get this:
>     (the hdb drive does not get detected/usable)

Don't mix old and new IDE like that. We allow both so you can use old SIS
and new Jmicron etc for example but there are some cases where trying to
load both old and new IDE support for the same chip will do strange
things.

Alan

Re: Problem with SiS Ide/Pata/Sata on 2.6.22-rc7-git4

[Thomas Backlund]

Alan Cox skrev:
>> If you modprobe only sis5513
>>   OR
>> If you modprobe only sis5513 and then sata_sis/pata_sis:
>>   - you get this:
>>     (the hdb drive does not get detected/usable)
> 
> Don't mix old and new IDE like that. We allow both so you can use old SIS
> and new Jmicron etc for example but there are some cases where trying to
> load both old and new IDE support for the same chip will do strange
> things.
> 
> Alan
> 
> 

So this means that if you want to use pata & sata disks connected to sis 
chipset you cant use sis5513 at all, if I understand this correct?

If that's the case I guess we simply have to blacklist (or stop 
building) the sis5513 and force all sis ide users to pata_sis...

Now, do you consider pata_sis as stable as the sis5513 ?

--
Thomas

Re: Problem with SiS Ide/Pata/Sata on 2.6.22-rc7-git4

[Thomas Backlund]

Alan Cox skrev:
>> If you modprobe only sis5513
>>   OR
>> If you modprobe only sis5513 and then sata_sis/pata_sis:
>>   - you get this:
>>     (the hdb drive does not get detected/usable)
> 
> Don't mix old and new IDE like that. We allow both so you can use old SIS
> and new Jmicron etc for example but there are some cases where trying to
> load both old and new IDE support for the same chip will do strange
> things.
> 

But there is still a regression here compared to 2.6.17...

If you load only sis5513 on a 2.6.22 you still get:
--- quote ---
<6>SIS5513: chipset revision 1
<6>SIS5513: not 100% native mode: will probe irqs later
<6>SIS5513: SiS 962/963 MuTIOL IDE UDMA133 controller
<6>    ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio
<6>    ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:pio, hdd:pio
<7>Probing IDE interface ide0...
<4>hdb: QUANTUM FIREBALLlct15 07, ATA DISK drive
<4>hdb: host side 80-wire cable detection failed, limiting max speed to 
UDMA33
<7>hdb: selected mode 0x42
--- quote ---

wich does not happend on 2.6.17...

The disk in question is ATA66 capable

--
Thomas

Re: Problem with SiS Ide/Pata/Sata on 2.6.22-rc7-git4

[Bartlomiej Zolnierkiewicz]

Hi,

On Tuesday 10 July 2007, Thomas Backlund wrote:
> Alan Cox skrev:
> >> If you modprobe only sis5513
> >>   OR
> >> If you modprobe only sis5513 and then sata_sis/pata_sis:
> >>   - you get this:
> >>     (the hdb drive does not get detected/usable)
> > 
> > Don't mix old and new IDE like that. We allow both so you can use old SIS
> > and new Jmicron etc for example but there are some cases where trying to
> > load both old and new IDE support for the same chip will do strange
> > things.
> > 
> 
> But there is still a regression here compared to 2.6.17...
> 
> If you load only sis5513 on a 2.6.22 you still get:
> --- quote ---
> <6>SIS5513: chipset revision 1
> <6>SIS5513: not 100% native mode: will probe irqs later
> <6>SIS5513: SiS 962/963 MuTIOL IDE UDMA133 controller
> <6>    ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio
> <6>    ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:pio, hdd:pio
> <7>Probing IDE interface ide0...
> <4>hdb: QUANTUM FIREBALLlct15 07, ATA DISK drive
> <4>hdb: host side 80-wire cable detection failed, limiting max speed to 
> UDMA33
> <7>hdb: selected mode 0x42
> --- quote ---
> 
> wich does not happend on 2.6.17...

I'm a bit confused here.  Do you mean that the warning message
is not printed or that disk is not limited to UDMA33?

Could you send "hdb" part of 2.6.17 dmesg?

> The disk in question is ATA66 capable

Please send output of "hdparm --Istdout /dev/hdb" command.

This is probably one of the early ATA66 disks with the buggy device side
cable detection.  CONFIG_IDEDMA_IVB=y should help but I'm currently working
on adding automatic detection of such disks.

Thanks,
Bart