question about SATA and IDE DVD/CD drives.

question about SATA and IDE DVD/CD drives.

[Tigran Aivazian]

Hi Jeff,

I was a bit confused with the "SATA vs IDE" thing recently, namely the 
fact that if I compile _both_ IDE and SATA support into the kernel 
then IDE takes over and SATA gets a "busy" error, and as a result I get 
3M/sec performance instead of 56M/sec.

Now, realizing this error I have reconfigured the kernel to NOT include 
ide and ONLY include SATA plus chipset-specific ata_piix.c (I have ICH6 
device 2652). However, now the new problem is that I can't see the IDE 
DVD-ROM drive anymore. Here is dmesg:

ata1: SATA max UDMA/133 cmd 0x1F0 ctl 0x3F6 bmdma 0xFFA0 irq 14
ata1: dev 0 cfg 49:2f00 82:346b 83:7d01 84:4003 85:3469 86:3c01 87:4003 88:207f
ata1: dev 0 ATA, max UDMA/133, 234441648 sectors: lba48
ata1: dev 0 configured for UDMA/133
scsi0 : ata_piix
  Vendor: ATA       Model: ST3120026AS       Rev: 3.05
  Type:   Direct-Access                      ANSI SCSI revision: 05
ata2: SATA max UDMA/133 cmd 0x170 ctl 0x376 bmdma 0xFFA8 irq 15
ata2: dev 0 cfg 49:0b00 82:0210 83:1000 84:0000 85:0000 86:0000 87:0000 88:0407
ata2: dev 0 ATAPI, max UDMA/33
ata2: dev 0 configured for UDMA/33
scsi1 : ata_piix
SCSI device sda: 234441648 512-byte hdwr sectors (120034 MB)
SCSI device sda: drive cache: write back
 sda: sda1 sda2 sda3
Attached scsi disk sda at scsi0, channel 0, id 0, lun 0
Attached scsi generic sg0 at scsi0, channel 0, id 0, lun 0,  type 0

As you see, sda is there, but where is my DVD drive gone?

# mount -t iso9660 /dev/scd0 /mnt/cdrom
mount: /dev/scd0 is not a valid block device
# uname -a
Linux tank 2.6.7 #3 SMP Fri Jul 2 13:40:07 BST 2004 x86_64 x86_64 x86_64 GNU/Linux
# grep " sr" /proc/devices 
 11 sr

Kind regards
Tigran

Re: question about SATA and IDE DVD/CD drives.

[Jeff Garzik]

Tigran Aivazian wrote:
> Hi Jeff,
> 
> I was a bit confused with the "SATA vs IDE" thing recently, namely the 
> fact that if I compile _both_ IDE and SATA support into the kernel 
> then IDE takes over and SATA gets a "busy" error, and as a result I get 
> 3M/sec performance instead of 56M/sec.
> 
> Now, realizing this error I have reconfigured the kernel to NOT include 
> ide and ONLY include SATA plus chipset-specific ata_piix.c (I have ICH6 
> device 2652). However, now the new problem is that I can't see the IDE 
> DVD-ROM drive anymore. Here is dmesg:
[...]
> As you see, sda is there, but where is my DVD drive gone?

I am guessing that your DVD drive is connected to a PATA bus, but you 
disabled the PATA driver (CONFIG_IDE) completely.

Enable CONFIG_IDE, and disable CONFIG_BLK_DEV_IDE_SATA, and that will 
fix things I bet.

	Jeff

Re: question about SATA and IDE DVD/CD drives.

[Tigran Aivazian]

if I enable CONFIG_IDE then IDE subsystem takes over 0170-0177
and 01f0-01f7 io ranges and SATA gets "busy" error and doesn't handle the 
hard disk (i.e. hard disk remains 3M/sec slow).

As for CONFIG_BLK_DEV_IDE_SATA, there is no such thing:

# find -name Kconfig | xargs fgrep CONFIG_BLK_DEV_IDE_SATA
# 

Maybe you meant CONFIG_SCSI_SATA? But if I disable it then the hard disk 
will no longer be 56M/sec but will go back to generic IDE 3M/sec 
performance.

My guess is that the solution to my problem is to compile both IDE and 
SATA but somehow pass some "ide?=???" boot options which would tell IDE to 
take over ide1 channel and let SATA handle the rest. Is this incorrect?

Kind regards
Tigran

On Fri, 2 Jul 2004, Jeff Garzik wrote:

> Tigran Aivazian wrote:
> > Hi Jeff,
> > 
> > I was a bit confused with the "SATA vs IDE" thing recently, namely the 
> > fact that if I compile _both_ IDE and SATA support into the kernel 
> > then IDE takes over and SATA gets a "busy" error, and as a result I get 
> > 3M/sec performance instead of 56M/sec.
> > 
> > Now, realizing this error I have reconfigured the kernel to NOT include 
> > ide and ONLY include SATA plus chipset-specific ata_piix.c (I have ICH6 
> > device 2652). However, now the new problem is that I can't see the IDE 
> > DVD-ROM drive anymore. Here is dmesg:
> [...]
> > As you see, sda is there, but where is my DVD drive gone?
> 
> I am guessing that your DVD drive is connected to a PATA bus, but you 
> disabled the PATA driver (CONFIG_IDE) completely.
> 
> Enable CONFIG_IDE, and disable CONFIG_BLK_DEV_IDE_SATA, and that will 
> fix things I bet.
> 
> 	Jeff
> 

Re: question about SATA and IDE DVD/CD drives.

[James Courtier-Dutton]

Jeff Garzik wrote:
> Tigran Aivazian wrote:
> 
>> Hi Jeff,
>>
>> I was a bit confused with the "SATA vs IDE" thing recently, namely the 
>> fact that if I compile _both_ IDE and SATA support into the kernel 
>> then IDE takes over and SATA gets a "busy" error, and as a result I 
>> get 3M/sec performance instead of 56M/sec.
>>
>> Now, realizing this error I have reconfigured the kernel to NOT 
>> include ide and ONLY include SATA plus chipset-specific ata_piix.c (I 
>> have ICH6 device 2652). However, now the new problem is that I can't 
>> see the IDE DVD-ROM drive anymore. Here is dmesg:
> 
> [...]
> 
>> As you see, sda is there, but where is my DVD drive gone?
> 
> 
> I am guessing that your DVD drive is connected to a PATA bus, but you 
> disabled the PATA driver (CONFIG_IDE) completely.
> 
> Enable CONFIG_IDE, and disable CONFIG_BLK_DEV_IDE_SATA, and that will 
> fix things I bet.
> 
>     Jeff
> 
> 

For some strange reason, you need the PATA HD driver compiled into the 
kernel, together with the SATA driver compiled into the kernel.
Then SATA takes the HDs, and the PATA HD driver is used to identify the 
DVD drive, and thus you can then load the PATA CD module, or the 
IDE-SCSI module to access the DVD drives.

I have a ICH5 chipset, so I had exactly the same problems as you.
The strange bit is needing the PATA HD driver compiled into the kernel, 
even though you only need PATA CD/DVD drivers.

Cheers
James

Re: question about SATA and IDE DVD/CD drives.

[Tigran Aivazian]

Ok, thanks, chaps, I'll try again now and see if the solution you propose
works. I'll let you know in either case.

Kind regards
Tigran

On Fri, 2 Jul 2004, James Courtier-Dutton wrote:

> Jeff Garzik wrote:
> > Tigran Aivazian wrote:
> > 
> >> Hi Jeff,
> >>
> >> I was a bit confused with the "SATA vs IDE" thing recently, namely the 
> >> fact that if I compile _both_ IDE and SATA support into the kernel 
> >> then IDE takes over and SATA gets a "busy" error, and as a result I 
> >> get 3M/sec performance instead of 56M/sec.
> >>
> >> Now, realizing this error I have reconfigured the kernel to NOT 
> >> include ide and ONLY include SATA plus chipset-specific ata_piix.c (I 
> >> have ICH6 device 2652). However, now the new problem is that I can't 
> >> see the IDE DVD-ROM drive anymore. Here is dmesg:
> > 
> > [...]
> > 
> >> As you see, sda is there, but where is my DVD drive gone?
> > 
> > 
> > I am guessing that your DVD drive is connected to a PATA bus, but you 
> > disabled the PATA driver (CONFIG_IDE) completely.
> > 
> > Enable CONFIG_IDE, and disable CONFIG_BLK_DEV_IDE_SATA, and that will 
> > fix things I bet.
> > 
> >     Jeff
> > 
> > 
> 
> For some strange reason, you need the PATA HD driver compiled into the 
> kernel, together with the SATA driver compiled into the kernel.
> Then SATA takes the HDs, and the PATA HD driver is used to identify the 
> DVD drive, and thus you can then load the PATA CD module, or the 
> IDE-SCSI module to access the DVD drives.
> 
> I have a ICH5 chipset, so I had exactly the same problems as you.
> The strange bit is needing the PATA HD driver compiled into the kernel, 
> even though you only need PATA CD/DVD drivers.
> 
> Cheers
> James
> 

Re: question about SATA and IDE DVD/CD drives.

[Jeff Garzik]

Tigran Aivazian wrote:
> if I enable CONFIG_IDE then IDE subsystem takes over 0170-0177
> and 01f0-01f7 io ranges and SATA gets "busy" error and doesn't handle the 
> hard disk (i.e. hard disk remains 3M/sec slow).

The quirk in drivers/pci/quirks.c should reserve one range for SATA, 
presuming that CONFIG_SCSI_SATA is set.


> As for CONFIG_BLK_DEV_IDE_SATA, there is no such thing:

> # find -name Kconfig | xargs fgrep CONFIG_BLK_DEV_IDE_SATA

You need to update your kernel, it looks like, that was recently added.


> My guess is that the solution to my problem is to compile both IDE and 
> SATA but somehow pass some "ide?=???" boot options which would tell IDE to 
> take over ide1 channel and let SATA handle the rest. Is this incorrect?

That's one way to do it.  ide?=noprobe probably.  But disabling 
CONFIG_BLK_DEV_IDE_SATA and enabling CONFIG_SCSI_SATA should get things 
going.

	Jeff

Re: question about SATA and IDE DVD/CD drives.

[Tigran Aivazian]

> > For some strange reason, you need the PATA HD driver compiled into the 
> > kernel, together with the SATA driver compiled into the kernel.
> > Then SATA takes the HDs, and the PATA HD driver is used to identify the 
> > DVD drive, and thus you can then load the PATA CD module, or the 
> > IDE-SCSI module to access the DVD drives.
> > 
> > I have a ICH5 chipset, so I had exactly the same problems as you.
> > The strange bit is needing the PATA HD driver compiled into the kernel, 
> > even though you only need PATA CD/DVD drivers.

No, still doesn't work. SATA disk is seen but DVD is not seen.

Here is config:

CONFIG_IDE=y
CONFIG_BLK_DEV_IDE=y
CONFIG_BLK_DEV_IDEDISK=y
CONFIG_BLK_DEV_IDECD=y

but worse, now I get an oops by doing cat /proc/iomem consistently:

Unable to handle kernel paging request at ffffffff00000000 RIP: 
<ffffffff801ce50c>{strnlen+13}
PML4 103027 PGD 0 
Oops: 0000 [1] SMP 
CPU 0 
Modules linked in: udf iptable_filter ip_tables
Pid: 1593, comm: cat Not tainted 2.6.7
RIP: 0010:[<ffffffff801ce50c>] <ffffffff801ce50c>{strnlen+13}
RSP: 0018:000001002ae3fd60  EFLAGS: 00010297
RAX: ffffffff00000000 RBX: 000001003d178014 RCX: 000000000000000a
RDX: 000001002ae3feb8 RSI: fffffffffffffffe RDI: ffffffff00000000
RBP: 0000000000000000 R08: 00000000fffffffe R09: 0000000000000004
R10: 00000000ffffffff R11: 0000000000000000 R12: 000001003d178fff
R13: ffffffff00000000 R14: 00000000ffffffff R15: 000001002ae3fdc8
FS:  0000002a958624c0(0000) GS:ffffffff8040a800(0000) 
knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: ffffffff00000000 CR3: 0000000000101000 CR4: 00000000000006e0
Process cat (pid: 1593, threadinfo 000001002ae3e000, task 
000001003c9ea6b0)
Stack: ffffffff801cedc6 0000010000000001 ffffffff80195bfd 0000000000001000 
       000001003d178000 ffffffff802c90bd 0000010001e39468 0000010000002000 
       0000000000002000 0000000000000400 
Call Trace:<ffffffff801cedc6>{vsnprintf+752} 
<ffffffff80195bfd>{proc_lookup+223} 
       <ffffffff80183909>{seq_printf+165} 
<ffffffff801740c2>{do_lookup+105} 
       <ffffffff8017497b>{link_path_walk+2157} 
<ffffffff801702fc>{cp_new_stat+233} 
       <ffffffff80135d4e>{r_show+113} <ffffffff801833fa>{seq_read+273} 
       <ffffffff801678af>{vfs_read+201} <ffffffff80167b00>{sys_read+73} 
       <ffffffff8010e976>{system_call+126} 

Code: 80 3f 00 74 13 48 83 ee 01 48 83 c0 01 48 83 fe ff 74 05 80 
RIP <ffffffff801ce50c>{strnlen+13} RSP <000001002ae3fd60>
CR2: ffffffff00000000

Kind regards
Tigran

Re: question about SATA and IDE DVD/CD drives.

[Tigran Aivazian]

On Fri, 2 Jul 2004, Jeff Garzik wrote:
> Enable CONFIG_IDE, and disable CONFIG_BLK_DEV_IDE_SATA, and that will 
> fix things I bet.

Tried this as well on the latest snapshot (2.6.7-bk9) and it failed as 
well. Namely, SATA disk works fine but IDE subsystem doesn't see the DVD 
drive.

Are you sure that I only need to enable CONFIG_IDE and not some of the
other IDE options (disk, cdrom, chipset-specific etc)?

Kind regards
Tigran

Re: question about SATA and IDE DVD/CD drives.

[Jeff Garzik]

Tigran Aivazian wrote:
> On Fri, 2 Jul 2004, Jeff Garzik wrote:
> 
>>Enable CONFIG_IDE, and disable CONFIG_BLK_DEV_IDE_SATA, and that will 
>>fix things I bet.
> 
> 
> Tried this as well on the latest snapshot (2.6.7-bk9) and it failed as 
> well. Namely, SATA disk works fine but IDE subsystem doesn't see the DVD 
> drive.
> 
> Are you sure that I only need to enable CONFIG_IDE and not some of the
> other IDE options (disk, cdrom, chipset-specific etc)?

Sorry, I was summarizing...  you definitely need a "personality" driver 
such as the ide-cdrom driver in order to make use of your DVD drive.

Really, though, I need to get off my butt and finish ATAPI in libata. 
Then, libata can drive PATA devices and we can get rid of all these 
headaches with combined mode being split between two drivers.

I'm going to be doing some libata hacking this weekend, looking 
particularly at a regression not caused by ACPI (hi Fabio).  I'll see if 
I can finish up ATAPI at that time.

	Jeff

Re: question about SATA and IDE DVD/CD drives.

[Tigran Aivazian]

On Fri, 2 Jul 2004, Jeff Garzik wrote:
> Tigran Aivazian wrote:
> > On Fri, 2 Jul 2004, Jeff Garzik wrote:
> > 
> >>Enable CONFIG_IDE, and disable CONFIG_BLK_DEV_IDE_SATA, and that will 
> >>fix things I bet.
> > 
> > 
> > Tried this as well on the latest snapshot (2.6.7-bk9) and it failed as 
> > well. Namely, SATA disk works fine but IDE subsystem doesn't see the DVD 
> > drive.
> > 
> > Are you sure that I only need to enable CONFIG_IDE and not some of the
> > other IDE options (disk, cdrom, chipset-specific etc)?
> 
> Sorry, I was summarizing...  you definitely need a "personality" driver 
> such as the ide-cdrom driver in order to make use of your DVD drive.

Tried it as well, namely:

CONFIG_IDE=y
CONFIG_BLK_DEV_IDE=y
# CONFIG_BLK_DEV_IDE_SATA is not set
CONFIG_BLK_DEV_IDECD=y
CONFIG_SCSI_SATA=y
CONFIG_SCSI_ATA_PIIX=y

and still the same result, i.e. SATA takes the disk but IDE only says 
this:

Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2
ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx

and DVD drive is still not seen.

I also tried enabling CONFIG_IDE_GENERIC but then it (IDE) reserves both 
io ranges and so SATA doesn't work (i.e. ata_piix probe fails with error 
-16).

I also tried not enabling CONFIG_IDE_GENERIC but instead enable PCI:

CONFIG_BLK_DEV_IDEPCI=y
CONFIG_BLK_DEV_IDEDMA_PCI=y
CONFIG_BLK_DEV_ADMA=y
CONFIG_BLK_DEV_PIIX=y
CONFIG_BLK_DEV_IDEDMA=y

Same result as the first one, i.e. SATA disk is OK but IDE DVD is not.

(btw there is a bug whereby one cannot disable CONFIG_SCSI_QLA2XXX for 
some reason)

I also tried passing ide0=noprobe to the IDE driver to make it leave ide0 
alone. Didn't work either, namely IDE took over both ide0 and ide1 and 
SATA failed to register both io regions.

So I still have a dilemma of either having a slow hdd or not being able to 
access DVD. 

Kind regards
Tigran

PS. Btw, I still get oops on cat /proc/iomem:

Unable to handle kernel paging request at ffffffff00000000 RIP: 
<ffffffff801cf084>{strnlen+13}
PML4 103027 PGD 0 
Oops: 0000 [1] SMP 
CPU 0 
Modules linked in: iptable_filter ip_tables
Pid: 2190, comm: cat Not tainted 2.6.7-bk9
RIP: 0010:[<ffffffff801cf084>] <ffffffff801cf084>{strnlen+13}
RSP: 0018:0000010028cbbd60  EFLAGS: 00010297
RAX: ffffffff00000000 RBX: 0000010023a00014 RCX: 000000000000000a
RDX: 0000010028cbbeb8 RSI: fffffffffffffffe RDI: ffffffff00000000
RBP: 0000000000000000 R08: 00000000fffffffe R09: 0000000000000004
R10: 00000000ffffffff R11: 0000000000000000 R12: 0000010023a00fff
R13: ffffffff00000000 R14: 00000000ffffffff R15: 0000010028cbbdc8
FS:  0000002a958624c0(0000) GS:ffffffff803fc600(0000) 
knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: ffffffff00000000 CR3: 0000000000101000 CR4: 00000000000006e0
Process cat (pid: 2190, threadinfo 0000010028cba000, task 
000001003ec95450)
Stack: ffffffff801cf9c1 0000010000000001 fffffffffffffff4 0000000000001000 
       0000010023a00000 ffffffff802c018d 0000010002395400 0000010000002000 
       0000000000002000 0000000000000400 
Call Trace:<ffffffff801cf9c1>{vsnprintf+752} 
<ffffffff80183aa1>{seq_printf+165} 
       <ffffffff80174944>{link_path_walk+2374} 
<ffffffff80173b04>{permission+38} 
       <ffffffff80170198>{cp_new_stat+233} <ffffffff80135a66>{r_show+113} 
       <ffffffff80183592>{seq_read+273} <ffffffff80167673>{vfs_read+201} 
       <ffffffff801678c4>{sys_read+73} <ffffffff8010e9d2>{system_call+126} 
       

Code: 80 3f 00 74 13 48 83 ee 01 48 83 c0 01 48 83 fe ff 74 05 80 
RIP <ffffffff801cf084>{strnlen+13} RSP <0000010028cbbd60>
CR2: ffffffff00000000

Re: question about SATA and IDE DVD/CD drives.

[Tigran Aivazian]

Hi Jeff,

After I gave up fiddling with "ideX=???" options, an alternative
possibility just occurred to me, looking at this dmesg output from libata:

ata_piix version 1.02
PCI: Setting latency timer of device 0000:00:1f.2 to 64
ata1: SATA max UDMA/133 cmd 0x1F0 ctl 0x3F6 bmdma 0xFFA0 irq 14
ata1: dev 0 cfg 49:2f00 82:346b 83:7d01 84:4003 85:3469 86:3c01 87:4003 88:207f
ata1: dev 0 ATA, max UDMA/133, 234441648 sectors: lba48
ata1: dev 0 configured for UDMA/133
scsi0 : ata_piix
  Vendor: ATA       Model: ST3120026AS       Rev: 3.05
  Type:   Direct-Access                      ANSI SCSI revision: 05
ata2: SATA max UDMA/133 cmd 0x170 ctl 0x376 bmdma 0xFFA8 irq 15
ata2: dev 0 cfg 49:0b00 82:0210 83:1000 84:0000 85:0000 86:0000 87:0000 88:0407
ata2: dev 0 ATAPI, max UDMA/33
ata2: dev 0 configured for UDMA/33
scsi1 : ata_piix
SCSI device sda: 234441648 512-byte hdwr sectors (120034 MB)
SCSI device sda: drive cache: write back
 sda: sda1 sda2 sda3
Attached scsi disk sda at scsi0, channel 0, id 0, lun 0

wouldn't it be nice if I could specify somehow to skip the "ata2" instance
and leave the 0x170/0x376 resource to be allocated by IDE, which would be
compiled as a module? Or, once probed and allocated (but not in use) I
could tell the driver to "release" a particular instance and free all
resources? I think this is not possible with the current infrastructure
and, seeing you want to get rid of this conflict altogether, it is
unlikely you would be interested in adding such support, right?

Kind regards
Tigran