Regression? 2.6.23.16 -> 2.6.32.9: ide/libata HD recognition

Regression? 2.6.23.16 -> 2.6.32.9: ide/libata HD recognition

[CaT]

Hi,

Just had a strange situation with the above upgrade. It was done on an
old server that used the serverworks ide driver. The IDE hw config under
2.6.23.16 was:

hda:	HD
hdb:	CDROM
hdc:	empty
hdd:	HD

After upgrade, using either IDE or libata the config changed to:

hda:	HD
hdb:	CDROM
hdc:	empty
hdd:	empty

The BIOS recognised the disk. I went through 3 disks and it wasn't until
I made the second HD a master that it was recognised. So now I have a
2.6.32.9 server, using the IDE subsystem with the following config:

hda:	HD
hdb:	CDROM
hdc:	HD
hdd:	empty

Otherwise it does not work.

I had an outage window that I already blew so I was not able to do
any more diagnosis than what it took to get the system up.

If anyone wants the full config I'll send. Otherwise:

IDE subsystem (no libata subsystem configured):

CONFIG_HAVE_IDE=y
CONFIG_IDE=y
CONFIG_IDE_XFER_MODE=y
CONFIG_IDE_ATAPI=y
CONFIG_IDE_GD=y
CONFIG_IDE_GD_ATA=y
CONFIG_BLK_DEV_IDECD=y
CONFIG_BLK_DEV_IDECD_VERBOSE_ERRORS=y
CONFIG_BLK_DEV_IDEACPI=y
CONFIG_IDE_PROC_FS=y
CONFIG_BLK_DEV_IDEDMA_SFF=y
CONFIG_BLK_DEV_IDEPCI=y
CONFIG_IDEPCI_PCIBUS_ORDER=y
CONFIG_BLK_DEV_IDEDMA_PCI=y
CONFIG_BLK_DEV_IDEDMA=y
CONFIG_LEDS_TRIGGER_IDE_DISK=y

libata kernel (no IDE subsystem configured):

CONFIG_ATA=y
CONFIG_ATA_VERBOSE_ERROR=y
CONFIG_ATA_ACPI=y
CONFIG_SATA_PMP=y
CONFIG_ATA_SFF=y
CONFIG_PATA_SERVERWORKS=y

-- 
  "A search of his car uncovered pornography, a homemade sex aid, women's 
  stockings and a Jack Russell terrier."
    - http://www.news.com.au/story/0%2C27574%2C24675808-421%2C00.html

Re: Regression? 2.6.23.16 -> 2.6.32.9: ide/libata HD recognition

[Bill Davidsen]

CaT wrote:
> Hi,
> 
> Just had a strange situation with the above upgrade. It was done on an
> old server that used the serverworks ide driver. The IDE hw config under
> 2.6.23.16 was:
> 
> hda:	HD
> hdb:	CDROM
> hdc:	empty
> hdd:	HD
> 
> After upgrade, using either IDE or libata the config changed to:
> 
> hda:	HD
> hdb:	CDROM
> hdc:	empty
> hdd:	empty
> 
> The BIOS recognised the disk. I went through 3 disks and it wasn't until
> I made the second HD a master that it was recognised. So now I have a
> 2.6.32.9 server, using the IDE subsystem with the following config:
> 
> hda:	HD
> hdb:	CDROM
> hdc:	HD
> hdd:	empty
> 
> Otherwise it does not work.
> 
> I had an outage window that I already blew so I was not able to do
> any more diagnosis than what it took to get the system up.
> 
> If anyone wants the full config I'll send. Otherwise:
> 
Could you "ls /sys/block" and post? And /var/log/messages from the start to the 
last device found? Thanks.

-- 
Bill Davidsen <davidsen@tmr.com>
   "We have more to fear from the bungling of the incompetent than from
the machinations of the wicked."  - from Slashdot

Re: Regression? 2.6.23.16 -> 2.6.32.9: ide/libata HD recognition

[CaT]

[-- Attachment #1: Type: text/plain, Size: 1094 bytes --]

On Wed, Mar 10, 2010 at 01:47:49PM -0500, Bill Davidsen wrote:
>> I had an outage window that I already blew so I was not able to do
>> any more diagnosis than what it took to get the system up.
>>
>> If anyone wants the full config I'll send. Otherwise:
>
> Could you "ls /sys/block" and post? And /var/log/messages from the start 

this is for the working setup:

$ ls /sys/block/
fd0  hdb  loop0  loop2	loop4  loop6  md0  md2	md4
hda  hdc  loop1  loop3	loop5  loop7  md1  md3	md5

I don't have one for the notworking side of things as that'd require
taking the server down, opening it, cursing heavily as I tried to
manipulate a tiny jumper in the worst location possible (thank god
for sata), etc.

> to the last device found? Thanks.

>From /var/log/messages I'm assuming you want the dmesg output. That's
attached. I'll provide both the libata and ide subsystem ones as well
as the 2.6.23.16 one.

-- 
  "A search of his car uncovered pornography, a homemade sex aid, women's 
  stockings and a Jack Russell terrier."
    - http://www.news.com.au/story/0%2C27574%2C24675808-421%2C00.html

[-- Attachment #2: dmesg.2.6.23.16-ide.bz2 --]
[-- Type: application/octet-stream, Size: 5358 bytes --]

[-- Attachment #3: dmesg.2.6.32.9-ide.bz2 --]
[-- Type: application/octet-stream, Size: 7532 bytes --]

[-- Attachment #4: dmesg.2.6.32.9-libata.bz2 --]
[-- Type: application/octet-stream, Size: 7736 bytes --]

Re: Regression? 2.6.23.16 -> 2.6.32.9: ide/libata HD recognition

[Robert Hancock]

On 03/09/2010 10:16 PM, CaT wrote:
> Hi,
>
> Just had a strange situation with the above upgrade. It was done on an
> old server that used the serverworks ide driver. The IDE hw config under
> 2.6.23.16 was:
>
> hda:	HD
> hdb:	CDROM
> hdc:	empty
> hdd:	HD

That sounds like you had a slave with no master on the second IDE 
channel. That really isn't a wise configuration and is prone to break 
(some devices don't work in this setup, some BIOSes choke, etc.) since 
the standard doesn't really allow this (the spec says that support for 
this configuration is "host specific" and not mandatory).

There have been some reports about this breaking when moving to libata 
(not sure about newer versions of IDE). I'm not sure exactly what has 
changed to cause this, but it might not be possible to get around it 
without breaking other things (could be the 80-wire cable detection 
protocol is what changed things, but IIRC the reason it changed was the 
original code didn't follow the spec and didn't always work properly 
with some hardware).

Essentially, fixing the hardware config to set the drive for master if 
it's the only device on the channel is likely the best solution overall.

>
> After upgrade, using either IDE or libata the config changed to:
>
> hda:	HD
> hdb:	CDROM
> hdc:	empty
> hdd:	empty
>
> The BIOS recognised the disk. I went through 3 disks and it wasn't until
> I made the second HD a master that it was recognised. So now I have a
> 2.6.32.9 server, using the IDE subsystem with the following config:
>
> hda:	HD
> hdb:	CDROM
> hdc:	HD
> hdd:	empty
>
> Otherwise it does not work.
>
> I had an outage window that I already blew so I was not able to do
> any more diagnosis than what it took to get the system up.
>
> If anyone wants the full config I'll send. Otherwise:
>
> IDE subsystem (no libata subsystem configured):
>
> CONFIG_HAVE_IDE=y
> CONFIG_IDE=y
> CONFIG_IDE_XFER_MODE=y
> CONFIG_IDE_ATAPI=y
> CONFIG_IDE_GD=y
> CONFIG_IDE_GD_ATA=y
> CONFIG_BLK_DEV_IDECD=y
> CONFIG_BLK_DEV_IDECD_VERBOSE_ERRORS=y
> CONFIG_BLK_DEV_IDEACPI=y
> CONFIG_IDE_PROC_FS=y
> CONFIG_BLK_DEV_IDEDMA_SFF=y
> CONFIG_BLK_DEV_IDEPCI=y
> CONFIG_IDEPCI_PCIBUS_ORDER=y
> CONFIG_BLK_DEV_IDEDMA_PCI=y
> CONFIG_BLK_DEV_IDEDMA=y
> CONFIG_LEDS_TRIGGER_IDE_DISK=y
>
> libata kernel (no IDE subsystem configured):
>
> CONFIG_ATA=y
> CONFIG_ATA_VERBOSE_ERROR=y
> CONFIG_ATA_ACPI=y
> CONFIG_SATA_PMP=y
> CONFIG_ATA_SFF=y
> CONFIG_PATA_SERVERWORKS=y
>

Re: Regression? 2.6.23.16 -> 2.6.32.9: ide/libata HD recognition

[CaT]

On Thu, Mar 11, 2010 at 06:04:38PM -0600, Robert Hancock wrote:
> That sounds like you had a slave with no master on the second IDE  

Indeed.

> channel. That really isn't a wise configuration and is prone to break  
> (some devices don't work in this setup, some BIOSes choke, etc.) since  
> the standard doesn't really allow this (the spec says that support for  
> this configuration is "host specific" and not mandatory).

I've never had it cause issues. BIOS has always recognised the HDs and
linux, until now, has always been more than happy to work with them.
This is across many different mbs, ide chips and hds.

On this server a lone hdd has been the case for 4 years and has worked
fine.

> Essentially, fixing the hardware config to set the drive for master if  
> it's the only device on the channel is likely the best solution overall.

It is what worked, after a lot of futzing around. Still, this is a
regression (from the pov of someone who just had to futz). Before
whatever change was made things were fine and futz-free.

-- 
  "A search of his car uncovered pornography, a homemade sex aid, women's 
  stockings and a Jack Russell terrier."
    - http://www.news.com.au/story/0%2C27574%2C24675808-421%2C00.html