[Bug 10436] New: Compact Flash HD: dma and seek_error/badCRC problems

[Bug 10436] New: Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436

           Summary: Compact Flash HD: dma and seek_error/badCRC problems
           Product: IO/Storage
           Version: 2.5
     KernelVersion: 2.6.24.3 / 2.6.24.4
          Platform: All
        OS/Version: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: IDE
        AssignedTo: io_ide@kernel-bugs.osdl.org
        ReportedBy: marco@absence.it


Latest working kernel version: 2.6.22 about DMA only
Earliest failing kernel version:
Distribution: Debian testing (Lenny)
Hardware Environment: Mini-ITX VIA EPIA SN Motherboard, Sandisk SDCFX4-8192
Compact Flash (hda), Maxtor PATA HD (hdb)
Problem Description:

I'm going to report two different problems but maybe related one with the
other.
I own a VIA SN board that comes with VIA VT8251 South Bridge (that includes a
VT82C5xx IDE interface). It has a single PATA connector (allowing thus max two
PATA drives) and a CF connector soldered in the back of the board (allowing
thus one CF and one PATA drive).
The SanDisk Extreme IV CF I'm using is udma4 capable and rated for 40MB/s.

The first problem:
With a 2.6.24 kernel I see in dmesg:

  hda: drive side 80-wire cable detection failed, limiting max speed to UDMA33
  hda: UDMA/33 mode selected

and the CF drive is reported to work with udma2 by hdparm:

  DMA: mdma0 mdma1 mdma2 udma0 udma1 *udma2 udma3 udma4

Using dd for some testing, the max throughput "speed" reached is about 28MB/s,
accordingly with the configuration above.
However I discovered that using a 2.6.22 kernel, dmesg still reports that
message but the kernel configures my CF to use udma4 as it should. Now dd shows
a 38MB/s (about) "speed", reaching the maximum speed for this CF.
Well, I don't really know how the CF slot is connected to the South Bridge but
for sure there are really short traces, and I think I can safely say I had no
data corruption using a 2.6.22 kernel and udma4 mode.
Changing the disk BIOS settings has no influence at all.
It seems I'm not the only one suffering this problem:
http://crazyduke.blogspot.com/2007/12/cflinux.html
It's not a big issue, really, but I'd like to understand why this happens and
how to use my CF at full speed with udma4.

The second problem:
Adding a second HD (Maxtor PATA UDMA/133) results in a udma0 setting (for this
drive, while the CF behaves normally) and in lots of errors in dmesg:

  [...]
  hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error }
  hdb: dma_intr: error=0x84 { DriveStatusError BadCRC }
  ide: failed opcode was: unknown
  hda: DMA disabled
  hdb: UDMA/100 mode selected
  ide0: reset: success
  [...]
  hdb: dma_intr: status=0x51 { DriveReady SeekComplete Error }
  hdb: dma_intr: error=0x84 { DriveStatusError BadCRC }
  ide: failed opcode was: unknown
  hdb: no DMA mode selected
  ide0: reset: success
  ldm_validate_partition_table(): Disk read failed.
  Dev hdb: unable to read RDB block 0
   unable to read partition table

Once the system was up, only /dev/hdb had been created and I had to rewrite the
partition table to the disk to see the partitions on it and correctly access
them. However when I tried to set udma6 or udma4/2 with hdparm I obtained just
a udma0 mode (infatc the system is lagging when coping files). I don't think I
had data corruption, but in this case I'm not that sure.
In the beginning I though that the CF does not want to share the ide channel
with another device. However I had no problem when I connected a (really) old
CDROM instead (using mdma2), for the system installation to the CF.
Using a 2.6.22 or a 2.6.24 kernel makes no difference.
Here it is a similar problem but without a solution:
http://readlist.com/lists/vger.kernel.org/linux-kernel/69/348695.html
Again, not a huge problem since I'm going to buy a larger SATA drive soon, but
maybe it's usefull to understand why this happens.


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #1 from marco@absence.it  2008-04-10 06:01 -------
Created an attachment (id=15717)
 --> (http://bugzilla.kernel.org/attachment.cgi?id=15717&action=view)
dmes and test logs

Here are some datas, reporting the first problem, executed under debian with
2.6.22 and 2.6.24 plus ubuntu 7.10 livecd using a 2.6.22 kernel.
Let me know if further testing is needed.


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #2 from alan@lxorguk.ukuu.org.uk  2008-04-10 06:57 -------
No real suprises there

CF devices and adapters are sometimes good enough for UDMA at 66MHz or better
but forget to report the cable detection. For those boards we keep tables and
if you provide the output of dmidecode that can be updated to include this
board (in the windows world vendors ship drivers so they can knobble the check
in their driver or .inf file)

High UDMA modes demand good cable and signal quality and we regularly see mixed
CF and disk on the same cable as failing (the CRC message is signal corruption
on the wire). I don't therefore think this is a software bug

The dmidecode and lspci -vvxxx output would be nice so we can let the kernel
know your system has a hardwired DMA capable CF slot and automate that bit.


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #3 from marco@absence.it  2008-04-10 12:21 -------
Created an attachment (id=15724)
 --> (http://bugzilla.kernel.org/attachment.cgi?id=15724&action=view)
dmidecode output

I'm afraid dmi data isn't much of help. As you can see at
http://www.lm-sensors.org/ticket/2298 this problem already occurred with
sensors-detect, due to a non-standard address in the SuperIO chip. Maybe asking
VIA a BIOS containing a motherboard identifier string would help?
However I have two questions for you:
1) Why does it work using udma4 with the previous 2.6.22?
2) In the BIOS at "ATA(PI) 80 pin Cable Detection" I can select "Host &
Device", "Host" and "Device". Which one should I choose? Which one could/should
be better?

About the second (hardware related) issue, I simply had this drive laying
around, so I gave a try, but I'll buy a new sata II one and use AHIC.
So let me know if you find something useful in the dmidecode output, otherwise
feel free to close this bug.


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #4 from alan@lxorguk.ukuu.org.uk  2008-04-10 13:21 -------
Joyous. Is there anything useful in lspci -vvxxx ?

UDMA4 probably used to work because the old code was less careful about modes.
You can override it by hand with the libata driver and also the old driver so a
hand override exists. Not sure there is much you can do about the CRC errors
though.


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #5 from marco@absence.it  2008-04-10 13:35 -------
Created an attachment (id=15726)
 --> (http://bugzilla.kernel.org/attachment.cgi?id=15726&action=view)
lspci output

Sorry, I forgot about the lspci output.
How the override can be done? It's a specific module parameter?
I removed the second HD a couple of days ago and I'm using just the CF.

BTW, thank you for your help!


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #6 from marco@absence.it  2008-04-10 15:08 -------
Ok, I have found this:
http://www.mail-archive.com/linux-ide@vger.kernel.org/msg14722.html
Different CF adapter but same CF and same problem. Tomorrow I'll try to reboot
the computer passing the "libata.force=80c,udma4" parameter. I think it should
work fine then.


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #7 from marco@absence.it  2008-04-12 09:53 -------
It didn't worked. I tried a couple of patch and many boot parameters but it
still work with udma2. Moreover, hda is now shown as sda: what's happening with
libata?


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436





------- Comment #8 from alan@lxorguk.ukuu.org.uk  2008-04-12 10:05 -------
If your disk controller is setup to use libata it appears as /dev/sd* if it
uses the legacy IDE drivers /dev/hd*


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

[Bug 10436] Compact Flash HD: dma and seek_error/badCRC problems

[bugme-daemon]

http://bugzilla.kernel.org/show_bug.cgi?id=10436


marco@absence.it changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |CLOSED
         Resolution|                            |PATCH_ALREADY_AVAILABLE




------- Comment #9 from marco@absence.it  2008-04-12 13:51 -------
Halt! Finally I got it working! I had to manually insert the new code but then
everything went fine then. Now we can close this bug (hoping to see this this
patch in the upstream kernel).


-- 
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.