Serial ATA (SATA) Linux Support

coolersites · 06-08-2003, 03:43 PM

arxik,

Was running rc6 and just upgraded to rc7. As suggested, I turned off "USE multy-mode by default" and already had the tg3 driver loaded. Recompiled, and same results as before.

I was able to force the IRQ to change by blocking and doing some manul assignment in the bios, but it didn't make an differance either.

Any chance you have a .config file from your compile you could supply?

If your install works, then the only thing left that I could imagine is a hardware failure (board or drive?). I have removed all pci cards (just a bttv tuner card in slot 1) - so I don't much else could be left to conflict. However I do have ide1 with a WD ATA IDE drive and a DVD, ide 2 with a CDRW (with scsi emulation turned on)... seems pretty standard stuff.

To diagnose the hardware, I guess I am going to test the drive with windoze.. if it works there....

gabriele_101 · 06-09-2003, 11:58 PM

Delete

coolersites · 06-20-2003, 07:26 PM

Ok, well first - after years of using and caring for Redhat I decided that I was ready for an easier life (not due to the HD problems but due to problems keeping XMMS running smoothly).

So - I made a wild switch one morning to Mandrake... so far so good.

Anyhow... I came upon this thread today:

Thread at NForcersHQ.com

Basically to sum it up, I added the following line to my rc.sysinit :
rc.sysinit snippet
------------------------------------------------------------------------------------------
# Remount the root filesystem read-write.
state=`awk '/ \/ / && ($3 !~ /rootfs/) { print $4 }' /proc/mounts`
[ "$state" != "rw" ] && \
action $"Remounting root filesystem in read-write mode: " mount -n -o remount,rw /

echo max_kb_per_request:15 > /proc/ide/hde/settings
/sbin/hdparm -X69 /dev/hde
/sbin/hdparm -d1 /dev/hde
/sbin/hdparm -c1u1 /dev/hde

# LVM initialization
if [ -f /etc/lvmtab -a ! -e /proc/lvm ] ; then

And now I have success with DMA on my SATA drives. I also believe this solution would have worked equally well for Redhat.

As I suspected earlier - the timing of turning the DMA on is very important to the success....

Cheers!

Oh, BTW - here's the -tT readings:
/dev/hde:
Timing buffer-cache reads: 128 MB in 0.25 seconds =512.00 MB/sec
Timing buffered disk reads: 64 MB in 1.81 seconds = 35.36 MB/sec

I suspect these could be even higher with a newer kernel (I am on the default Mandrake 9.1 2.4.21-0.13mdkenterprise kernel) - but I notice increases when recompiling my RH kernel.

mobilagent · 06-22-2003, 02:55 AM

Worked quite nicely!

GiGabyte 8knxp i875p MB with on board SI3112 SATA, two WD Raptor SATA.

Both drives got recognized during installation, setup a software RAID0 then without any problems. I used Ctrl+Alt+F2 to switch to shell during installation and tested the HD speed, hdparm -t gives 1.8MB/s, manually enabled DMA by typing:
hdparm -X69 -d1 /dev/hde
hdparm -X69 -d1 /dev/hdg
now both drives runs at 54MB/s.

After installation, following the instructions provided by Silicon Image website and made all the settings permanent, now
hdparm -t /dev/md0 will give a stunning speed at 104MB/s!

here is the website:
http://12.24.47.40/display/2/index.a...=2&r=0.8054468

just browse into downloads->drivers->SiI3x12-> SiI3x12: Serial ATA (SATA) Linux IDE Driver

download the zip file on the right side of the page:
3x1x_x86_lin_rh90-dma-tweak.zip

You have all you need to fuel up the Raptors!

coolersites · 06-22-2003, 01:56 PM

Just a final note on this:

It looks as though for SEAGATE drives the following is necessary to add to rc.sysinit. It would appear this is a SEAGATE SATA issue, and may not be necessary with the WD Raptor drives.

echo "max_kb_per_request:15" > /proc/.ide/hdX/settings

Seems the kernel folks are just catching up on this:
(Alan Cox and Andre Hedrick). See:
http://www.cs.helsinki.fi/linux/linu...3-22/0760.html

For now, having a Seagate is just a little bit of a bummer!

-Aaron

JakeS · 06-30-2003, 03:43 PM

Hello,
I've installed RH 9.0 but
hdparm -X69 -d1 /dev/hde crashes the system instantly.

I have a Tyan 2460 dual athalon board with the SI3112 on a PCI card and the WD 36G drive. Everything installed fine, runs fine in slow-motion (hdparm -t = 2.5 MB/s). I tried booting with the SMP and non-SMP kernel got the same results.

Any ideas?

Thanks in advance.

Uniform Multi-Platform E-IDE driver Revision: 7.00beta-2.4
ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
AMD7411: IDE controller at PCI slot 00:07.1
AMD7411: chipset revision 1
AMD7411: not 100% native mode: will probe irqs later
ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda

MA, hdb

io
ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc

io, hdd

io
SiI3112 Serial ATA: IDE controller at PCI slot 00:0d.0
SiI3112 Serial ATA: chipset revision 2
SiI3112 Serial ATA: not 100% native mode: will probe irqs later
ide2: MMIO-DMA at 0xe080d800-0xe080d807, BIOS settings: hde

io, hdf

io
ide3: MMIO-DMA at 0xe080d808-0xe080d80f, BIOS settings: hdg

io, hdh

io
hda: IC35L040AVER07-0, ATA DISK drive
blk: queue c043a1e0, I/O limit 4095Mb (mask 0xffffffff)
hdc: CREATIVE CD2422E, ATAPI CD/DVD-ROM drive
hde: WDC WD360GD-00FNA0, ATA DISK drive
hdg: no response (status = 0xfe)
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
ide1 at 0x170-0x177,0x376 on irq 15
ide2 at 0xe080d880-0xe080d887,0xe080d88a on irq 11
hda: host protected area => 1
hda: 80418240 sectors (41174 MB) w/1916KiB Cache, CHS=5005/255/63, UDMA(100)
hde: host protected area => 1
hde: 72303840 sectors (37020 MB) w/8192KiB Cache, CHS=4500/255/63
ide-floppy driver 0.99.newide
Partition check:
hda: hda1 hda2
hde: hde1 hde2 hde3

JakeS · 07-01-2003, 09:29 AM

Don't know if anyone is interested in this but I solved my own problem (previous post):

Compiled new kernel - 2.4.21
Using hdparm -X69 -d1 still crashed system
Using hdparm -X69 -c3 -u1 sets it to UDMA5
then using hdparm -d1 turns on DMA
hdparm -t gets me 57.5 mb/s now so it seems to be working
For some reason, turning on DMA at the same time you use X69 seems to be fatal.

Bottom line is I got the SATA drive to play nice with a Tyan Tiger MP with the SI PCI adapter (Western Digital drive).

Hope this is useful for someone - it cost me a day and a half

Aussie · 07-01-2003, 07:28 PM

Quote:

Originally posted by coolersites

For now, having a Seagate is just a little bit of a bummer!

-Aaron

You can say that again.

I've got the following in /etc/rc.d/rc.local,

Code:

echo max_kb_per_request:15 > /proc/ide/hde/settings
hdparm -X69 -d1 -m1 -c3 -u1 /dev/hde

hdparm -tT,

Code:

/dev/hde:
 Timing buffer-cache reads:   128 MB in  0.25 seconds =512.00 MB/sec
 Timing buffered disk reads:  64 MB in  1.85 seconds = 34.59 MB/sec

My barracuda IV,

Code:

/dev/hda:
 Timing buffer-cache reads:   128 MB in  0.62 seconds =206.45 MB/sec
 Timing buffered disk reads:  64 MB in  1.68 seconds = 38.10 MB/sec

Hopefully there will be a fix in an upcoming kernel - or I'm going to be contacting seagate.

coolersites · 07-02-2003, 12:38 AM

I actually did contact Seagate. In fact I even sent them a copy of this thread since my previous correspondance 100% verified they had no clue about linux (and I might venture to say - anything beyond level 1 support).

Perhaps now they will at least be able to advise folks on how to get the SATA drives to work, but something tells me my advise will go unheeded since they did not bother to reply.

I would suggest that anyone with a Seagate SATA should begin bombarding Seagate.

Hmm, I think maybe I will send this thread to Tom's Hardware as well. It would only be fair to have a proper performance review I think.

-Aaron

dav1x · 07-03-2003, 01:22 PM

I checked with the driver supposed to be found on Silicons website but following the links listed yields the results that the driver has been moved to kernel.org.

Did anyone have the filename of the driver listed?

Was their driver a source or kernel patch or what?
I cannot seem to get rh 8.0 or rh 9 to see anything on the SATA controller, this is on a 865G intel chipset.

Thanks in advance,
dav1x

JakeS · 07-03-2003, 01:53 PM

The 2.4.21 kernel should have the code in it. I got mine to work with RH 9.0 after downloading and compiling that kernel (make sure the proper options are checked in xconfig).
The files I got from the website were rather useless - just a bunch of scripts to set hdparm variables.

Maerd · 07-03-2003, 07:07 PM

to coolersites

I had and still have a problem with SATA drives. That you have had before. But using your solution I've come, unfortunately, to nothing.

I inserted these rows

echo "max_kb_per_request:15" > /proc/ide/hde/settings
/sbin/hdparm -X69 /dev/hde
/sbin/hdparm -c1u1 /dev/hde
/sbin/hdparm -d1 /dev/hde

in my rc.sysinit right in the place showed in your example, after "# Remount the root filesystem read-write." rows. And Mandrake 9.1 hangs up during the boot right after it shows me the row

using_dma = 1 (on)

If anyone else knows how to solve the problem, help me, please. I'm very unhappy to have transfer rate 1.38 MB/sec

P.S. I have 2 HDD Seagate Barracuda 7200.7 160GB.

coolersites · 07-03-2003, 08:45 PM

After tweaking (using Mandrake 9.1) here is what I have currently:

in /etc/rc.d/rc.sysinit:

# Remount the root filesystem read-write.
state=`awk '/ \/ / && ($3 !~ /rootfs/) { print $4 }' /proc/mounts`
[ "$state" != "rw" ] && \
action "Remounting root filesystem in read-write mode: " mount -n -o remount,rw /

# ADDED to throttle DMA (some hard disks will no work above this value)
echo "max_kb_per_request:15" > /proc/ide/hde/settings

Then made file:
/etc/sysconfig/harddiskhde

USE_DMA=1
MULTIPLE_IO=16
EIDE_32BIT=1
# LOOKAHEAD=1
EXTRA_PARAMS=-X69

Maerd · 07-04-2003, 07:52 PM

Well, it works only if I set EXTRA_PARAMS=-X66
this is better than nothing but a bit frustrating...

coolersites · 07-05-2003, 12:23 PM

Not sure why you are able to get on -x66, unless it relates to the dual drives. Are you running raid? and is this the raid bios for the 3112 chip, or just the standard bios?

Understand the frustration! When these drives cost $190 and this is the best you can get-- I am wondering if this shouldn't be considered a hardware failure.. since the raptors dont seem to be influenced.

Anyhow, I have contacted everyone I can think of to try and solve and / or publicize these issues and nobody really seems to give a crap.