Problems with DMA - udma mode switches

Susu · 2008-02-10 13:10:02

I was checking speed with hdparm by interest and came to a very strange thing: After running hdparm -tT /dev/sda, it switches from udma6 to udma2. Testing same thing with sdb it still shows udma5. What the heck is wrong? Okay, here's my setting, output of hdparm etc. If something is missing, please let me know.

My disks:

sda: Samsung SP1604N (ide, 160 GB, 7200 rpm) - root, home, video
sdb: Samsung SV1604N (ide, 160 GB, 5400 rpm) - mucke
sdc: Siemens Storagebird XL-E (usb sata, 500 GB, 7200 rpm) - usbdisk

My fstab

#
# /etc/fstab: static file system information
#
# <file system>        <dir>         <type>    <options>          <dump> <pass>
none                   /dev/pts      devpts    defaults            0      0
none                   /dev/shm      tmpfs     defaults            0      0


#/dev/cdrom /mnt/cdrom   iso9660   ro,user,noauto,unhide   0      0
#/dev/cdrom1 /mnt/cdrom1   iso9660   ro,user,noauto,unhide   0      0
#/dev/dvd /mnt/dvd   udf   ro,user,noauto,unhide   0      0
#/dev/dvd1 /mnt/dvd1   udf   ro,user,noauto,unhide   0      0

## Floppy
/dev/fd0 /mnt/fd0   vfat   user,noauto   0 0
## swap
/dev/disk/by-label/swap swap swap defaults 0 0
## root
/dev/disk/by-label/root / ext3 defaults 0 1
## home
/dev/disk/by-label/home /home ext3 defaults 0 0
## DVB/Torrents
/dev/disk/by-label/video /mnt/video ext3 defaults 0 0
## Musik
/dev/disk/by-label/mucke /mnt/mucke ext3 defaults 0 0
## Filme
/dev/disk/by-label/usbdisk /mnt/usbdisk ext3 defaults 0 0

## iRiver H300
/dev/iriver /mnt/iriver vfat noauto,rw,users,uid=1000,gid=100,shortname=win95,iocharset=utf8 0 0

## currently not in use (20 GB)
#/dev/disk/by-label/backup /mnt/backup ext3 defaults 0 0
## currently not in use (10 GB)
#/dev/disk/by-label/backup2 /mnt/backup2 ext3 defaults 0 0

dmesg | grep -i ata.*dma
ata1: PATA max UDMA/133 cmd 0x000101f0 ctl 0x000103f6 bmdma 0x0001fb00 irq 14
ata2: PATA max UDMA/133 cmd 0x00010170 ctl 0x00010376 bmdma 0x0001fb08 irq 15
ata1.00: ATA-7: SAMSUNG SP1604N, TM100-23, max UDMA/133
ata1.01: ATA-7: SAMSUNG SV1604N, TR100-24, max UDMA/100
ata1.00: configured for UDMA/133
ata1.01: configured for UDMA/100
ata2.00: ATAPI: HL-DT-STDVD-ROM GDR8162B, 0017, max UDMA/33
ata2.01: ATAPI: HL-DT-ST DVDRAM GSA-H10N, JL10, max UDMA/33
ata2.00: configured for UDMA/33
ata2.01: configured for UDMA/33
sata_nv 0000:00:07.0: Using ADMA mode
ata3: SATA max UDMA/133 cmd 0xf8820480 ctl 0xf88204a0 bmdma 0x0001f600 irq 16
ata4: SATA max UDMA/133 cmd 0xf8820580 ctl 0xf88205a0 bmdma 0x0001f608 irq 16
sata_nv 0000:00:08.0: Using ADMA mode
ata5: SATA max UDMA/133 cmd 0xf8822480 ctl 0xf88224a0 bmdma 0x0001f100 irq 17
ata6: SATA max UDMA/133 cmd 0xf8822580 ctl 0xf88225a0 bmdma 0x0001f108 irq 17

From mkinitcpio.conf

MODULES="pata_amd ata_generic sata_nv"
HOOKS="base udev autodetect pata scsi sata usb keymap filesystems"

hdparm -i /dev/sda

/dev/sda:
 Model=SAMSUNG SP1604N                         , FwRev=TM100-23, SerialNo=S013J10WA20104
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
 RawCHS=16383/16/63, TrkSize=34902, SectSize=554, ECCbytes=4
 BuffType=DualPortCache, BuffSize=2048kB, MaxMultSect=16, MultSect=?16?
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=268435455
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4
 DMA modes:  mdma0 mdma1 mdma2
 UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6
 AdvancedPM=no WriteCache=enabled
 Drive conforms to: ATA/ATAPI-7 T13 1532D revision 0:  ATA/ATAPI-1,2,3,4,5,6,7

hdparm -i /dev/sdb

/dev/sdb:
 Model=SAMSUNG SV1604N                         , FwRev=TR100-24, SerialNo=0810J1FWC08804
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
 RawCHS=16383/16/63, TrkSize=34902, SectSize=554, ECCbytes=4
 BuffType=DualPortCache, BuffSize=2048kB, MaxMultSect=16, MultSect=?16?
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=268435455
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4
 DMA modes:  mdma0 mdma1 mdma2
 UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5
 AdvancedPM=no WriteCache=enabled
 Drive conforms to: ATA/ATAPI-7 T13 1532D revision 0:  ATA/ATAPI-1,2,3,4,5,6,7

Now it's getting interesting!

hdparm -tT /dev/sda

/dev/sda:
 Timing cached reads:   1134 MB in  2.00 seconds = 567.26 MB/sec
 Timing buffered disk reads:  read(2097152) returned 1286144 bytes

After that hdparm -i refers to udma2

hdparm -tT /dev/sdb

/dev/sdb:
 Timing cached reads:   1150 MB in  2.00 seconds = 574.67 MB/sec
 Timing buffered disk reads:  138 MB in  3.03 seconds =  45.47 MB/sec

while sdb is still in udma5.

Anyone any ideas with that?

Last edited by Susu (2008-02-10 17:52:03)

lilsirecho · 2008-02-10 18:24:32

UDMA 2 is 33 speed.

Wonder if it runs slow...this shows up in many posts where the DVD slows down to a crawl, ending up with dma25!

Does yours run slow in sda udma 2? If so, the hdparm reading is correct. Otherwise, hdparm is erroneous for that drive.

You may wish to exchange the two and see if it is drive specific.....

Susu · 2008-02-10 19:44:00

Okay, I tested it. Coping a file from sda to sdc, before running hdparm -tT /dev/sda went smooth as expected. Copying the same file after running hdparm -tT /dev/sda went so slow it hurts. The mouse pointer is jerky and everything responds very slow. So I might say the hdparm reading is correct.

To test this a little further, I would like to change from libata to "old ide" but I'm not sure how I should change mkinitcpio.conf. Is it okay when it looks like this or do I need other modules in the MODULE array?

MODULES="ata_generic sata_nv"
HOOKS="base udev autodetect ide scsi sata usb keymap filesystems"

lilsirecho · 2008-02-10 21:25:35

One such change was by Ashren and he included the modules referenced in the comment above MODULES "".

See my bug report 9386...

lilsirecho · 2008-02-10 21:31:28

Am I to understand that following the use of hdparm -tT on a drive that it lowered the speed of that drive on subsequent operations?

lilsirecho · 2008-02-10 23:56:09

The following command may be of help in correcting the difficulty.

After receiving the slow down, try the following;

.....hdparm -tTX69 /dev/sda....

This will set the device to udma 5 and read the speed.

It may not stay udma5 but this method should reinstate it when it is lost. That is, if hdparm works in ata as it does in ide.

You can use ...hdparm -X69 /dev/sda...as well should the method work, without reading the speeds. It seems that reading the speeds is what upsets the apple cart......

Susu · 2008-02-11 05:32:18

hdparm -X69 /dev/sda

/dev/sda:
 setting xfermode to 69 (UltraDMA mode5)
SG_IO: bad/missing ATA_16 sense data::  70 00 05 00 00 00 00 0a 00 00 00 00 24 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 HDIO_DRIVE_CMD(setxfermode) failed: Input/output error

Susu · 2008-02-11 14:58:54

I googled for the problem and found a lot of links. Seems like a kernel/libata problem.
https://bugs.launchpad.net/ubuntu/+sour … bug/110636

lilsirecho · 2008-02-11 15:54:48

I received a reply on the bug I submitted which required me to use kernel 24 for reporting a bug.

My system is a LIVE system and I cannot change the kernel.

Perhaps your system is kernel 24 and a bug submitted with reference to the googled links will aid in getting the problem solved.

It gives various system problems mostly poor speed and very slow performance. The response you received is identical to those I receive.

I have attempted many methods of averting the problem with no success including boot prompt and mkinitcpio hooks. Ashren was successful in correcting his DVD problem with a return to IDE.

From your description, it seems your drive was recognized as udma 5 before you tried hdparm -tT on it. Then it failed to run at udma 5 following that .

If that is the case, then hdparm use initiates a change in the udma params for some drives but not others.

I note your drives are different spindle speeds, one is 5400 and the other is 7200. This may be a factor in the failure mode.

Your experience with the problem sheds some light on real reason for the failed HDD udma since one drive was affected but not the other (unless I misinterpret the results).

My application involves attempts to set udma 5 on a compact flash/IDE adapter combo which fails with the same error you posted. A hdd with udma 6 installed in the same IDE slot gave udma 6 in hdparm with 62MB/sec speed.

This test also confirms that different devices create the failure mode. The test results you posted reveal an ata request for 16 bit sense data from the device, thus signifying the lack of response from the device in udma mode).

Perhaps your system will allow you to interchange the two drives and test the performance again to verify the problem is device related.

There may be a need for a delay in performing the ata sense data detection within the libata interface for some devices.

Just tinkering with the possibilities.....

lilsirecho · 2008-02-11 18:04:07

Googling reveals a possible solution:

instead of ....hdparm -X69..... reset hdparm with the following:

hdparm -d1 -X69 -i -v -t /dev/sdx

This will read out the result as well as set the udma, maybe!

Susu · 2008-02-11 18:48:35

lilsirecho wrote:

Googling reveals a possible solution:
instead of ....hdparm -X69..... reset hdparm with the following:
hdparm -d1 -X69 -i -v -t /dev/sdx
This will read out the result as well as set the udma, maybe!

Nope, same problem... I will interchange both drives and see how far it gets me.

Zer0 · 2008-02-11 20:27:28

Why is it that hdparm -tT /dev/sda returns a value like the following?

Timing buffered disk reads:  read(2097152) returned 1286144 bytes

When the expected output should look like..

Timing buffered disk reads:  [x number of] MB in 3.00 Seconds = [x number]

curious..

Last edited by Zer0 (2008-02-11 20:28:05)

lilsirecho · 2008-02-11 20:31:11

We had hoped you wouldn't ask!!

The problem is the expectation for any reasonable response is thwarted.

UDMA has varied responses from all of Linux and devices used in Linux and it wouldn't be a surprise to me to see even more weird things!!!

lilsirecho · 2008-02-12 18:07:58

Susu;

The latest kernel has an "acpi in libata" change that might solve your problem. It is kernel26.24.1-2 hot on the mirrors!

Ashren · 2008-02-13 11:58:30

pata_acpi: ACPI driver support

On a cable there may be
eighty wires or perhaps forty
and we learn about its type
In the world of ACPI

So we call the GTM
And we find the the timing rate
And we look through it to see
If eighty wire it must be

Timing lives in ACPI routines
ACPI routines, ACPI routines
Timing lives in ACPI routines
ACPI routes ACPI routines

And the drivers last you see
Picking up unknown pci ids
and the code begins to work

Timing lives in ACPI routines
ACPI routines, ACPI routines
Timing lives in ACPI routines
ACPI routes ACPI routines

[Full speed ahead, Mr Hacker, full speed ahead]
Full speed over here sir!
Checking Cable, checking cable
Aye aye, 80 wire,
Heaven heaven]

If we use ACPI (ACPI)
Every box (every box) has all we need (has all we need)
Cable type (cable type) and mode timing (mode timing)
In our ATA (in our ATA) subroutines (subroutines, ha ha)

Timing lives in ACPI routines
ACPI routines, ACPI routines
Timing lives in ACPI routines
ACPI routes ACPI routines
Timing lives in ACPI routines
ACPI routines, ACPI routines
Timing lives in ACPI routines
ACPI routes ACPI routines

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Source: http://git.kernel.org/?p=linux/kernel/g … 0f3e8ea675

lilsirecho · 2008-02-13 14:36:25

Bravo!! I dub thee Sir Ashren, king of the ACPI....long live the KING!:lol::D:cool:

Susu · 2008-02-13 15:33:07

Lol, what a lovely song! But I'm waiting for kernel26 update. I'm just to lazy to build myself...

Susu · 2008-02-13 17:06:20

Just a stupid question (maybe): Where to put libata.dma=3? Kernel command line? If so, it's not working...

dmesg | grep libata
Kernel command line: root=/dev/disk/by-label/root libata.dma=3 ro vga=795
Unknown boot option `libata.dma=3': ignoring

lilsirecho · 2008-02-14 02:49:17

I don't have a stupid answer for your stupid question....but I do have a light at the end of the tunnel,,,

My problem has been the lack of recognition of a compact flash device with udma 5 capability (Lexar 2GB udma 300x compact flash when connected into an IDE interface adapter. I always was left at udma 2.

This occured with two separate kernels as provided by Faunos Live install. The dmesg gave a limit due to 40-wire cable.

This made the adapter suspect since it could be incorrectly wired or could introduce crosstalk.

Today, with the latest kernel 2.6.24.1-2 in arch , I ran hdparm -tT /dev/hdc and received buffered disk reads of 43MB/sec which is udma 5 level. Thus the adapter is not the cause of the dmesg but the kernel is the culprit. I did not boot into faunos with this device, however, only verified it would provide udma 5 in the latest kernel for archlinux.

The kernel in my Faunos may be 2.6.24 but am not sure of the trailing dots and dashes. It does not produce the udma 5 when booted into faunos but limits it to udma2.

There is soon to be another kernel released...2.6.24.2-1 which will be included in Faunos rc3 and may have the corrections needed to ID udma on the compact flash device.

At udma 2 it exhibits 16-18MB/sec so the udma 5 should be very impressive.

I am planning to examine the Lexar performance in another machine having booted into the Faunos with another device for each of two kernel boots.

Back to the salt mines!!!!

lilsirecho · 2008-02-14 15:43:48

Observations with Lexarflash device in gigabyte 81865gme-775-rh computer mobo pentium4 2GB 400MHZ.

The Lexar compact flash 2GB 300x installed with IDE adapter always takes over boot regardless of the boot device setting in BIOS.

It is recognized as a removeable drive when so installed. This will prevent the assignment of udma 5 as I understand the params.

When this device is installed in my asus a7v333 at 1600mhz cpu with 3GB ram, it is recognized as a fixed IDE device and reaches udma 5 in hdparm with 43mb/sec test result. This reading is obtained on the device while connected in IDE socket but not booting from it.

At present, the params in xorg prevent its use as boot device, but changes in that area are in work.

I suspect it has little chance of passing muster with the existing kernel but because it does respond to hdparm, the interface is shown to be viable. The kernel in the asus is 2.6.24.1-2 but the kernel in the Lexar is not necessarily that kernel since it is called 2.6.24-faunos.

At present it does not sync with my VGA lcd and the asus has an agp driver installed.

Stay tuned...............................

Susu · 2008-02-16 13:45:56

By now I think that my Samsung SP1604N might be broken. I will probably get me a new disk...

lilsirecho · 2008-02-18 16:51:58

My problems with UDMA boot continue to get worse!

In implementing the latest kernel by Faunos, I have a number of new problems preventing boot-up with either of my computers.

It seems as the complexity of the kernel increases the ability to administer the consequences is inversely proportional to the coding effort required or even recognized by the coder. Hardware nuances become achilles heels.

I hope your proposed new hardware solves the difficulty you encountered. Be sure to try one you can return if it doesn't perform.

Best of luck!!

Susu · 2008-02-18 17:00:20

Now up and running again with a new disk... have a look!

hdparm -i /dev/sda

dev/sda:

 Model=ExcelStor Technology J8160              , FwRev=P22OA85A, SerialNo=      PV4301Q495PV3C
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=51
 BuffType=DualPortCache, BuffSize=7384kB, MaxMultSect=16, MultSect=?16?
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=268435455
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4
 DMA modes:  mdma0 mdma1 mdma2
 UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6
 AdvancedPM=yes: disabled (255) WriteCache=enabled
 Drive conforms to: ATA/ATAPI-7 T13 1532D revision 1:  ATA/ATAPI-2,3,4,5,6,7

And....tadaaaaaaaa

hdparm -tT /dev/sda

/dev/sda:
 Timing cached reads:   1166 MB in  2.00 seconds = 582.90 MB/sec
 Timing buffered disk reads:  222 MB in  3.01 seconds =  73.64 MB/sec

I'm very glad :-)

lilsirecho · 2008-02-18 17:58:50

W00t!

lilsirecho · 2008-02-19 19:04:43

My problem with UDMA remains but I did get Faunos rc3 running with the Lexar 2GB flash drive but at udma2 instead of udma 5.

The scsi,pata and sata don't respond to the 80-wire test AFAICT.

Guess the name needs to be changed to Unresponsive Direct Memory Access.....!

Best to you....

EDIT: My problem might be solved with the patch described in this URL;

http://pixca.net/2008/02/03/limited-to- … ire-cable/

woot!

Last edited by lilsirecho (2008-02-19 20:07:14)

Arch Linux

#1 2008-02-10 13:10:02

Problems with DMA - udma mode switches

#2 2008-02-10 18:24:32

Re: Problems with DMA - udma mode switches

#3 2008-02-10 19:44:00

Re: Problems with DMA - udma mode switches

#4 2008-02-10 21:25:35

Re: Problems with DMA - udma mode switches

#5 2008-02-10 21:31:28

Re: Problems with DMA - udma mode switches

#6 2008-02-10 23:56:09

Re: Problems with DMA - udma mode switches

#7 2008-02-11 05:32:18

Re: Problems with DMA - udma mode switches

#8 2008-02-11 14:58:54

Re: Problems with DMA - udma mode switches

#9 2008-02-11 15:54:48

Re: Problems with DMA - udma mode switches

#10 2008-02-11 18:04:07

Re: Problems with DMA - udma mode switches

#11 2008-02-11 18:48:35

Re: Problems with DMA - udma mode switches

#12 2008-02-11 20:27:28

Re: Problems with DMA - udma mode switches

#13 2008-02-11 20:31:11

Re: Problems with DMA - udma mode switches

#14 2008-02-12 18:07:58

Re: Problems with DMA - udma mode switches

#15 2008-02-13 11:58:30

Re: Problems with DMA - udma mode switches

#16 2008-02-13 14:36:25

Re: Problems with DMA - udma mode switches

#17 2008-02-13 15:33:07

Re: Problems with DMA - udma mode switches

#18 2008-02-13 17:06:20

Re: Problems with DMA - udma mode switches

#19 2008-02-14 02:49:17

Re: Problems with DMA - udma mode switches

#20 2008-02-14 15:43:48

Re: Problems with DMA - udma mode switches

#21 2008-02-16 13:45:56

Re: Problems with DMA - udma mode switches

#22 2008-02-18 16:51:58

Re: Problems with DMA - udma mode switches

#23 2008-02-18 17:00:20

Re: Problems with DMA - udma mode switches

#24 2008-02-18 17:58:50

Re: Problems with DMA - udma mode switches

#25 2008-02-19 19:04:43

Re: Problems with DMA - udma mode switches

Board footer