How to access SATA drives attached to SII3124

unix4me2002 · 04-15-2009, 06:11 PM

I have silicon image card SiI3124 attached to my pci-x, and four sata drives attached to the silicon image card. I’m running Linux kernel 2.6.28 that I build with optional modules sata_si.ok. After boot, I have successfully loaded the following modules: sata_sil.ko, ata_piix.ko and sata_sil24.ko
> cat /proc/modules
sata_sil24.ko 13188 0 - Live 0x bla bla
ata_piix.ko 21764 0 - Live ...
sata_sil.ko 0 - Live ...

I recall reading that the sata_si.ko exposes the drive as /proc/scsi/si*
however I see nothing like that.

I'm not sure if I need additional modules to access my attached sata drives or if the drives appears in the file systems elsewhere.

Any Idea?
Thanks

kentyler · 04-16-2009, 02:47 PM

What does fdisk -l return? Or dmesg say anything about the drives or card?

unix4me2002 · 04-16-2009, 08:22 PM

ken,
First thanks for showing an interest.

fdisk -l # returns nothing

dmesg -- as follows

>>>>>>>>>>>>>> dmesg begin <<<<<<<<<<<<<<<<<<<<<<<
BIOS EBDA/lowmem at: 0009f800/0009f800
Initializing cgroup subsys cpuset
Initializing cgroup subsys cpu
Linux version 2.6.28TWIDL (root@DTMCLIENT044) (gcc version 4.2.4 (Ubuntu 4.2.4-1ubuntu3)) #1 SMP Tue Apr 14 10:06:33 PDT 2009

<<<<<<<<<< i cut some text to make it readable >>>>>>>>>>>>>>>>>>>

VFS: Mounted root (ext2 filesystem) readonly.
Freeing unused kernel memory: 380k freed
Write protecting the kernel text: 2916k
Write protecting the kernel read-only data: 1340k
EXT2-fs warning: maximal mount count reached, running e2fsck is recommended
sata_sil24 0000:02:0c.0: version 1.1
sata_sil24 0000:02:0c.0: Applying completion IRQ loss on PCI-X errata fix
scsi0 : sata_sil24
scsi1 : sata_sil24
scsi2 : sata_sil24
scsi3 : sata_sil24
ata1: SATA max UDMA/100 host m128@0xec090000 port 0xec088000 irq 24
ata2: SATA max UDMA/100 host m128@0xec090000 port 0xec08a000 irq 24
ata3: SATA max UDMA/100 host m128@0xec090000 port 0xec08c000 irq 24
ata4: SATA max UDMA/100 host m128@0xec090000 port 0xec08e000 irq 24
ata1: SATA link down (SStatus 0 SControl 0)
ata2: SATA link down (SStatus 0 SControl 0)
ata3: SATA link down (SStatus 0 SControl 0)
ata4: SATA link down (SStatus 0 SControl 0)
sata_sil24 0000:02:0f.0: Applying completion IRQ loss on PCI-X errata fix
scsi4 : sata_sil24
scsi5 : sata_sil24
scsi6 : sata_sil24
scsi7 : sata_sil24
ata5: SATA max UDMA/100 host m128@0xec091000 port 0xec080000 irq 25
ata6: SATA max UDMA/100 host m128@0xec091000 port 0xec082000 irq 25
ata7: SATA max UDMA/100 host m128@0xec091000 port 0xec084000 irq 25
ata8: SATA max UDMA/100 host m128@0xec091000 port 0xec086000 irq 25
ata5: SATA link down (SStatus 0 SControl 0)
ata6: SATA link down (SStatus 0 SControl 0)
ata7: SATA link down (SStatus 0 SControl 0)
ata8: SATA link down (SStatus 0 SControl 0)
libiscsi: Unknown symbol iscsi_alloc_session
libiscsi: Unknown symbol iscsi_session_chkready
libiscsi: Unknown symbol iscsi_host_for_each_session
libiscsi: Unknown symbol iscsi_free_session
libiscsi: Unknown symbol iscsi_conn_error_event
libiscsi: Unknown symbol iscsi_destroy_conn
libiscsi: Unknown symbol iscsi_create_conn
libiscsi: Unknown symbol iscsi_recv_pdu
libiscsi: Unknown symbol iscsi_add_session
libiscsi: Unknown symbol iscsi_destroy_session
libiscsi: Unknown symbol iscsi_unblock_session
libiscsi: Unknown symbol iscsi_block_session
Loading iSCSI transport class v2.0-870.
ide_generic: Unknown symbol ide_host_add
Uniform Multi-Platform E-IDE driver
ide_generic: please use "probe_mask=0x3f" module parameter for probing all legacy ISA IDE ports
Probing IDE interface ide0...
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
Probing IDE interface ide1...
ide1 at 0x170-0x177,0x376 on irq 15

>>>>>>>>>>>>>> dmesg end <<<<<<<<<<<<<<<<<<<<<<<

unix4me2002 · 04-23-2009, 08:21 PM

I have successfully build 2.6.28 with the additional modules ata_generic and sata_sil24. loading these modules provided me with a view of my attached devices on /proc/scsi/sg/device_strs.

My next issue is the appearance of these devices as a file devices on /dev/sd?. Currently no one is listed.

Which module is responsible to map /dev/sd? or create these file devices
Any ideas?

kentyler · 04-24-2009, 10:50 AM

Did you build and load the sd module for scsi disk? or the sg module for scsi generic?

That may help.

unix4me2002 · 04-24-2009, 08:08 PM

Ken,
"Did you build and load the sd module for scsi disk? or the sg module for scsi generic?"
1) Regarding the sd I have the kernel build with CONFIG_BLK_DEV_SD=y
2) "sg module for scsi generic" I'm not sure which build option you are referring.

BTY, my system boots of a ram drive and no sysfs exist. I'm not clear what control the creation of the sysfs.

Thanks again
Ben

unix4me2002 · 04-25-2009, 11:21 AM

Locating the modules above was the first part, and then mapping the procfs to /dev was accomplished by udevd. Now I’m able to see the devices as /dev/sd?
Thanks

jimboberella · 01-21-2010, 05:35 AM

Ok so can you explain this in a little more detail because I'm in the exact same place right now!

alpha01 · 03-04-2010, 06:47 PM

Quote:

Originally Posted by unix4me2002

Locating the modules above was the first part, and then mapping the procfs to /dev was accomplished by udevd. Now I’m able to see the devices as /dev/sd?
Thanks

I'm also trying to mound four external drives using a SiI 3124 card. Did you manually had to load the kernel module?

I noticed that on my machine it was already included.

Code:

[root@backup2 kernel]# modprobe -l | grep sata_sil24
/lib/modules/2.6.18-164.el5/kernel/drivers/ata/sata_sil24.ko

I'm able to see the drives /dev/sdb [e] yet I'm not able to access the drives via fdisk.

Code:

[root@backup2 kernel]# fdisk  /dev/sdb

Unable to read /dev/sdb

[root@backup2 silicon_driver]# tail  /var/log/messages
Mar  4 15:45:25 backup2 kernel: sd 1:0:0:0: SCSI error: return code = 0x00040000
Mar  4 15:45:25 backup2 kernel: end_request: I/O error, dev sdb, sector 1953525160
Mar  4 15:45:25 backup2 kernel: sd 1:0:0:0: SCSI error: return code = 0x00040000
Mar  4 15:45:25 backup2 kernel: end_request: I/O error, dev sdb, sector 1953525160
Mar  4 15:45:25 backup2 kernel: sd 1:0:0:0: SCSI error: return code = 0x00040000
Mar  4 15:45:25 backup2 kernel: end_request: I/O error, dev sdb, sector 0
Mar  4 15:45:25 backup2 kernel: sd 1:0:0:0: SCSI error: return code = 0x00040000
Mar  4 15:45:25 backup2 kernel: end_request: I/O error, dev sdb, sector 0

wad4ever · 03-21-2010, 06:49 PM

I'm stuck here too.

My machine has an IDE boot drive, plus five SATA drives that I need to reform into an md (they are from my old file server that died). Four of them are plugged into the motherboard, and one plugged into this card: http://www.newegg.com/Product/Produc...82E16816124008 (it's a SYBA SD-SATA2-4IR PCI SATA II Controller Card, see http://www.syba.com/index.php?contro...on=Info&Id=864).

I can only see four of the SATA drives; the one on the card is missing.

Here's my linux version (Debian lenny):

Code:

> uname -a
Linux abba 2.6.26-2-686 #1 SMP Tue Mar 9 17:35:51 UTC 2010 i686 GNU/Linux
> cat /proc/version
Linux version 2.6.26-2-686 (Debian 2.6.26-21lenny4) (dannf@debian.org) (gcc version 4.1.3 20080704 (prerelease) (Debian 4.1.2-25)) #1 SMP Tue Mar 9 17:35:51 UTC 2010

I can see the controller card:

Code:

> lspci | grep 3124
03:01.0 RAID bus controller: Silicon Image, Inc. SiI 3124 PCI-X Serial ATA Controller (rev 02)

The kernel seems to have the driver (as far as I understand this):

Code:

> modprobe -l | grep sata_sil24
/lib/modules/2.6.26-2-686/kernel/drivers/ata/sata_sil24.ko

And these modules:

Code:

> cat /proc/modules | grep si
sata_sil24 11812 0 - Live 0xf8856000
libata 140448 3 ata_piix,ata_generic,sata_sil24, Live 0xf88dc000
scsi_mod 129548 2 sd_mod,libata, Live 0xf88bb000

Here are the drives I can see:

Code:

> cat /proc/partitions
major minor  #blocks  name

   8     0  488386584 sda
   8     1  488384001 sda1
   3     0   20010312 hda
   3     1   19133383 hda1
   3     2          1 hda2
   3     5     875511 hda5
   8    16  488386584 sdb
   8    17  488384001 sdb1
   8    32  488385527 sdc
   8    33  488384001 sdc1
   8    48  488386584 sdd
   8    49  488384001 sdd1

Here's the firmware version of the card, shown at boot:

Code:

6.3.18

Here's some parts of the boot log. It appears that the drive is visible to the OS.

Code:

> dmesg
.........snip...........
[    2.168286] sata_sil24 0000:03:01.0: version 1.1
[    2.168305] ACPI: PCI Interrupt 0000:03:01.0[A] -> GSI 16 (level, low) -> IRQ 16
[    2.168615] scsi0 : sata_sil24
[    2.168716] scsi1 : sata_sil24
[    2.168767] scsi2 : sata_sil24
[    2.168808] scsi3 : sata_sil24
[    2.168837] ata1: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf0000 irq 16
[    2.168841] ata2: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf2000 irq 16
[    2.168844] ata3: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf4000 irq 16
[    2.168847] ata4: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf6000 irq 16
.........snip...........
[    4.368051] ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 0)
[    4.384023] ata1.00: ATA-8: WDC WD5000AAJS-22YFA0, 12.01C02, max UDMA/133
[    4.384023] ata1.00: 976773168 sectors, multi 16: LBA48 NCQ (depth 31/32)
[    4.384023] ata1.00: configured for UDMA/100
[    6.472038] ata2: SATA link down (SStatus 0 SControl 0)
[    8.552036] ata3: SATA link down (SStatus 0 SControl 0)
[   10.632037] ata4: SATA link down (SStatus 0 SControl 0)
[   10.632167] scsi 0:0:0:0: Direct-Access     ATA      WDC WD5000AAJS-2 12.0 PQ: 0 ANSI: 5
.........snip...........

What I'd like to see is that fifth SATA drive show up, so I can re-create my md with it.

Here are a couple of options:
1. Try flashing the BIOS to a version that will allow me to just provide the connected drives to the OS, without having to use its RAID feature. Here are the BIOSes: http://www.siliconimage.com/support/...27&cat=15&os=0
2. Rebuild my md RAID with just the four devices, then use the card's firmware to wipe that fifth drive, create a RAID out of it, and expose it to the OS. Then I can add it to my RAID-5 md.
3. I have another card, it's a 2-port SATA controller (sil 3112). I know it works, was using it in my old box, the one that died. I tried plugging it into this new box, but had the same problem as I'm having now: drive doesn't show up in /proc/partitions. So I took it back out, and am trying with this 4-port card instead.

wad4ever · 03-24-2010, 12:01 AM

I flashed the BIOS on that card to be:

Code:

BASE BIOS 6.4.09

But it's still not showing me the device under /proc/partitions. Again, I can see it as ata1 under dmesg:

Code:

..... snip .......
[    2.196283] sata_sil24 0000:03:01.0: version 1.1
[    2.196302] ACPI: PCI Interrupt 0000:03:01.0[A] -> GSI 16 (level, low) -> IRQ 16
[    2.196611] scsi0 : sata_sil24
[    2.196712] scsi1 : sata_sil24
[    2.196762] scsi2 : sata_sil24
[    2.196805] scsi3 : sata_sil24
[    2.196834] ata1: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf0000 irq 16
[    2.196838] ata2: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf2000 irq 16
[    2.196842] ata3: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf4000 irq 16
[    2.196845] ata4: SATA max UDMA/100 host m128@0xfebffc00 port 0xfebf6000 irq 16
..... snip .......
[    4.396046] ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 0)
[    4.407978] ata1.00: ATA-8: WDC WD5000AAJS-22YFA0, 12.01C02, max UDMA/133
[    4.407978] ata1.00: 976773168 sectors, multi 16: LBA48 NCQ (depth 31/32)
[    4.408777] ata1.00: configured for UDMA/100
[    6.488039] ata2: SATA link down (SStatus 0 SControl 0)
[    8.568038] ata3: SATA link down (SStatus 0 SControl 0)
[   10.648038] ata4: SATA link down (SStatus 0 SControl 0)
[   10.648170] scsi 0:0:0:0: Direct-Access     ATA      WDC WD5000AAJS-2 12.0 PQ: 0 ANSI: 5
..... snip .......

Why, oh, why can't I see the drive?

wad4ever · 03-26-2010, 12:39 PM

It works! I'm not sure what exactly it was that made it work, but here is what I did:

1. It was up and running, but debian could not see anything attached to the card. There was no /dev/sde in /proc/partitions. There were four drives plugged into the motherboard, and one on the card. The OS could see the four drives just fine.
2. I powered down, and unplugged the four drives from the motherboard, then rebooted.
3. The OS showed me the drive attached to the card!
4. I powered down, and plugged one more drive into the motherboard. When I powered up the OS recognized both!
5. When I put everything into the motherboard it saw them all.

Hey, I won't argue with success... even if I don't know why.

alpha01 · 03-27-2010, 02:22 AM

Quote:

Originally Posted by wad4ever

It works! I'm not sure what exactly it was that made it work, but here is what I did:

1. It was up and running, but debian could not see anything attached to the card. There was no /dev/sde in /proc/partitions. There were four drives plugged into the motherboard, and one on the card. The OS could see the four drives just fine.
2. I powered down, and unplugged the four drives from the motherboard, then rebooted.
3. The OS showed me the drive attached to the card!
4. I powered down, and plugged one more drive into the motherboard. When I powered up the OS recognized both!
5. When I put everything into the motherboard it saw them all.

Hey, I won't argue with success... even if I don't know why.

I also got it to work, and I was able to build a software RAID 5 to an iSCSI SAN device. But the only problem that I faced afterwards was that every time I copied mass amount of files to this software RAID, it caused it automatically remount to read-only. On a production server this is catastrophic as the only way (that I know) to recover from a filesystem that auto mounted itself as read-only is by doing a reboot.

Sadly know thats to this Silicon Image card, we are now deploying a Windows Server instead of CentOS as we first planned. :-(