linux-raid.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* how to disable partition search?
@ 2008-03-26 11:43 Ferenc Wagner
  2008-03-26 20:20 ` Bill Davidsen
  2008-04-06 10:37 ` Luca Berra
  0 siblings, 2 replies; 6+ messages in thread
From: Ferenc Wagner @ 2008-03-26 11:43 UTC (permalink / raw)
  To: linux-raid

Hi,

I use an FC SAN, which provides multiple pathes to multiple LUNs.
These all come up as different sd* devices, exhausting single letter
names.  I mean they are a LOT.  Using the md mulitpath driver
everything works perfectly, no problems there.  However, during boot,
the kernel tries to read the partition table from each device,
spitting out hundreds of lines of error messages: most of the devices
aren't even readable, and those which are, don't contain a valid
partition table.  They never will.  So I'd like to disable partition
detection, because these messages overflow the kernel message buffer,
depriving syslog of gathering any useful boot messages, and also
needlessly lengthening the boot process.  (Of course the noise alone
is disturbing enough when one tries to troubleshoot a boot problem.)
However, looking at the kernel sources didn't give me any hint.  Is
this possible to disable at all?

(Please Cc me, I'm not on the list.)
-- 
Thanks,
Feri.

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: how to disable partition search?
  2008-03-26 11:43 how to disable partition search? Ferenc Wagner
@ 2008-03-26 20:20 ` Bill Davidsen
  2008-03-27  0:45   ` Ferenc Wagner
  2008-04-06 10:37 ` Luca Berra
  1 sibling, 1 reply; 6+ messages in thread
From: Bill Davidsen @ 2008-03-26 20:20 UTC (permalink / raw)
  To: Ferenc Wagner; +Cc: linux-raid

Ferenc Wagner wrote:
> Hi,
>
> I use an FC SAN, which provides multiple pathes to multiple LUNs.
> These all come up as different sd* devices, exhausting single letter
> names.  I mean they are a LOT.  Using the md mulitpath driver
> everything works perfectly, no problems there.  However, during boot,
> the kernel tries to read the partition table from each device,
> spitting out hundreds of lines of error messages: most of the devices
> aren't even readable, and those which are, don't contain a valid
> partition table.  They never will.  So I'd like to disable partition
> detection, because these messages overflow the kernel message buffer,
> depriving syslog of gathering any useful boot messages, and also
> needlessly lengthening the boot process.  (Of course the noise alone
> is disturbing enough when one tries to troubleshoot a boot problem.)
> However, looking at the kernel sources didn't give me any hint.  Is
> this possible to disable at all?
>
> (Please Cc me, I'm not on the list.)
>   
After reading some code, I didn't see any boot or module load options 
you can use to prevent that, I think you may want to increase buffer 
size to hold the messages and use "grep -v" to create a sanitized 
dmesg.  If you can give an example of the error message you see I can 
look again.

Don't forget to copy the list...

-- 
Bill Davidsen <davidsen@tmr.com>
  "Woe unto the statesman who makes war without a reason that will still
  be valid when the war is over..." Otto von Bismark 



^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: how to disable partition search?
  2008-03-26 20:20 ` Bill Davidsen
@ 2008-03-27  0:45   ` Ferenc Wagner
  0 siblings, 0 replies; 6+ messages in thread
From: Ferenc Wagner @ 2008-03-27  0:45 UTC (permalink / raw)
  To: Bill Davidsen; +Cc: linux-raid

Bill Davidsen <davidsen@tmr.com> writes:

> Ferenc Wagner wrote:
>
>> I use an FC SAN, which provides multiple pathes to multiple LUNs.
>> These all come up as different sd* devices, exhausting single letter
>> names.  I mean they are a LOT.  Using the md mulitpath driver
>> everything works perfectly, no problems there.  However, during boot,
>> the kernel tries to read the partition table from each device,
>> spitting out hundreds of lines of error messages: most of the devices
>> aren't even readable, and those which are, don't contain a valid
>> partition table.  They never will.  So I'd like to disable partition
>> detection, because these messages overflow the kernel message buffer,
>> depriving syslog of gathering any useful boot messages, and also
>> needlessly lengthening the boot process.  (Of course the noise alone
>> is disturbing enough when one tries to troubleshoot a boot problem.)
>> However, looking at the kernel sources didn't give me any hint.  Is
>> this possible to disable at all?
>>
>> (Please Cc me, I'm not on the list.)
>
> After reading some code, I didn't see any boot or module load options
> you can use to prevent that, I think you may want to increase buffer
> size to hold the messages and use "grep -v" to create a sanitized
> dmesg.  If you can give an example of the error message you see I can
> look again.
>
> Don't forget to copy the list...

Thanks for looking into this.  As an illustration, a reboot looks like
this in the syslog:

Mar  3 17:20:55 mach kernel: Kernel logging (proc) stopped.
Mar  3 17:20:55 mach kernel: Kernel log daemon terminating.
Mar  3 17:20:56 mach exiting on signal 15
Mar  3 17:27:00 mach syslogd 1.4.1#18: restart.
Mar  3 17:27:00 mach kernel: klogd 1.4.1#18, log source = /proc/kmsg started.
[boot messages should have been inserted here.]
Mar  3 17:27:00 mach kernel:  Enabled
Mar  3 17:27:00 mach kernel:  sdd:<6>sd 0:0:0:3: [sdd] Device not ready: <6>: Sense Key : Not Ready [current] 
Mar  3 17:27:00 mach kernel: : Add. Sense: Logical unit not ready, manual intervention required
Mar  3 17:27:00 mach kernel: end_request: I/O error, dev sdd, sector 0
Mar  3 17:27:00 mach kernel: sd 0:0:0:3: [sdd] Device not ready: <6>: Sense Key : Not Ready [current] 
Mar  3 17:27:00 mach kernel: : Add. Sense: Logical unit not ready, manual intervention required
Mar  3 17:27:00 mach kernel: end_request: I/O error, dev sdd, sector 0
[the above 3 lines 7 more times]
Mar  3 17:27:00 mach kernel: Dev sdd: unable to read RDB block 0
[those 3 lines 2 more times]
Mar  3 17:27:00 mach kernel: sd 0:0:0:3: [sdd] Device not ready: <6>: Sense Key : Not Ready [current] 
Mar  3 17:27:00 mach kernel: : Add. Sense: Logical unit not ready, manual intervention required
Mar  3 17:27:00 mach kernel: end_request: I/O error, dev sdd, sector 24
Mar  3 17:27:00 mach kernel: printk: 29 messages suppressed.
Mar  3 17:27:00 mach kernel: sd 0:0:0:3: [sdd] Device not ready: <6>: Sense Key : Not Ready [current] 
Mar  3 17:27:00 mach kernel: : Add. Sense: Logical unit not ready, manual intervention required
Mar  3 17:27:00 mach kernel: end_request: I/O error, dev sdd, sector 24
Mar  3 17:27:00 mach kernel: sd 0:0:0:3: [sdd] Device not ready: <6>: Sense Key : Not Ready [current] 
Mar  3 17:27:00 mach kernel: : Add. Sense: Logical unit not ready, manual intervention required
Mar  3 17:27:00 mach kernel: end_request: I/O error, dev sdd, sector 0
Mar  3 17:27:00 mach kernel:  unable to read partition table
Mar  3 17:27:00 mach kernel: sd 0:0:0:3: [sdd] Attached SCSI disk
Mar  3 17:27:00 mach kernel: sd 0:0:1:0: [sde] 10485760 512-byte hardware sectors (5369 MB)
Mar  3 17:27:00 mach kernel: sd 0:0:1:0: [sde] Write Protect is off
Mar  3 17:27:00 mach kernel: sd 0:0:1:0: [sde] Cache data unavailable
Mar  3 17:27:00 mach kernel: sd 0:0:1:0: [sde] 10485760 512-byte hardware sectors (5369 MB)
Mar  3 17:27:00 mach kernel: sd 0:0:1:0: [sde] Write Protect is off
Mar  3 17:27:00 mach kernel: sd 0:0:1:0: [sde] Cache data unavailable
Mar  3 17:27:00 mach kernel:  sde: unknown partition table
Mar  3 17:27:00 mach kernel: sd 0:0:1:0: [sde] Attached SCSI disk
Mar  3 17:27:00 mach kernel: sd 0:0:1:1: [sdf] 419430400 512-byte hardware sectors (214748 MB)
Mar  3 17:27:00 mach kernel: sd 0:0:1:1: [sdf] Test WP failed, assume Write Enabled
Mar  3 17:27:00 mach kernel: sd 0:0:1:1: [sdf] 419430400 512-byte hardware sectors (214748 MB)
Mar  3 17:27:00 mach kernel: sd 0:0:1:1: [sdf] Test WP failed, assume Write Enabled
Mar  3 17:27:00 mach kernel:  sdf:<6>sd 0:0:1:1: [sdf] Device not ready: <6>: Sense Key : Not Ready [current] 
Mar  3 17:27:00 mach kernel: : Add. Sense: Logical unit not ready, manual intervention required
Mar  3 17:27:00 mach kernel: end_request: I/O error, dev sdf, sector 0
Mar  3 17:27:00 mach kernel: sd 0:0:1:1: [sdf] Device not ready: <6>: Sense Key : Not Ready [current] 
Mar  3 17:27:00 mach kernel: : Add. Sense: Logical unit not ready, manual intervention required
Mar  3 17:27:00 mach kernel: end_request: I/O error, dev sdf, sector 0
[and so on]

It would be possible to grep out the offending lines, but I'd rather
fight the cause than the symptom.  Doing my research I realized I'm
not alone with this problem.  Looks like a recompilation is necessary,
so I'm turning this into a feature request: how could selective
partition check be added to the Linux kernel?  I'm willing to give it
a shot under some expert guidance.
-- 
Thanks,
Feri.

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: how to disable partition search?
  2008-03-26 11:43 how to disable partition search? Ferenc Wagner
  2008-03-26 20:20 ` Bill Davidsen
@ 2008-04-06 10:37 ` Luca Berra
  2008-04-06 22:25   ` Ferenc Wagner
  1 sibling, 1 reply; 6+ messages in thread
From: Luca Berra @ 2008-04-06 10:37 UTC (permalink / raw)
  To: Ferenc Wagner; +Cc: linux-raid

On Wed, Mar 26, 2008 at 12:43:41PM +0100, Ferenc Wagner wrote:
>Hi,
>
>I use an FC SAN, which provides multiple pathes to multiple LUNs.
>These all come up as different sd* devices, exhausting single letter
>names.  I mean they are a LOT.  Using the md mulitpath driver
>everything works perfectly, no problems there.  However, during boot,
>the kernel tries to read the partition table from each device,
>spitting out hundreds of lines of error messages: most of the devices
>aren't even readable, and those which are, don't contain a valid
>partition table.  They never will.  So I'd like to disable partition
>detection, because these messages overflow the kernel message buffer,
>depriving syslog of gathering any useful boot messages, and also
>needlessly lengthening the boot process.  (Of course the noise alone
>is disturbing enough when one tries to troubleshoot a boot problem.)
>However, looking at the kernel sources didn't give me any hint.  Is
>this possible to disable at all?

you could try bugging lkml until it dawns on them that partition
detection code should belong in userspace by default :)

anyway you can rebuild your own kernel disabling it
just set PARTITION_ADVANCED, and disable all partition types.

you should be aware that doing this will disable partition detection on
all drives, so if you have partitioned drives (eg boot drives) you have
to run partx in initramfs or it wont be able to access them.

L.


-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: how to disable partition search?
  2008-04-06 10:37 ` Luca Berra
@ 2008-04-06 22:25   ` Ferenc Wagner
  2008-04-07 18:02     ` Ferenc Wagner
  0 siblings, 1 reply; 6+ messages in thread
From: Ferenc Wagner @ 2008-04-06 22:25 UTC (permalink / raw)
  To: linux-raid; +Cc: linux-kernel, wferi

Luca Berra <bluca@comedia.it> writes:

> On Wed, Mar 26, 2008 at 12:43:41PM +0100, Ferenc Wagner wrote:
>
>> I use an FC SAN, which provides multiple pathes to multiple LUNs.
>> These all come up as different sd* devices, exhausting single letter
>> names.  I mean they are a LOT.  Using the md mulitpath driver
>> everything works perfectly, no problems there.  However, during boot,
>> the kernel tries to read the partition table from each device,
>> spitting out hundreds of lines of error messages: most of the devices
>> aren't even readable, and those which are, don't contain a valid
>> partition table.  They never will.  So I'd like to disable partition
>> detection, because these messages overflow the kernel message buffer,
>> depriving syslog of gathering any useful boot messages, and also
>> needlessly lengthening the boot process.  (Of course the noise alone
>> is disturbing enough when one tries to troubleshoot a boot problem.)
>> However, looking at the kernel sources didn't give me any hint.  Is
>> this possible to disable at all?
>
> you could try bugging lkml until it dawns on them that partition
> detection code should belong in userspace by default :)

:) Very well put.  Thanks for strenghtening my belief.

> anyway you can rebuild your own kernel disabling it
> just set PARTITION_ADVANCED, and disable all partition types.

Sounds like a good idea, will try that!  This makes bugging lkml
unnecessary, if not for changing the default...  But that's more a
distribution issue, I think.  And now that we have udev and kpartx,
things will probably drift in this direction.

> you should be aware that doing this will disable partition detection on
> all drives, so if you have partitioned drives (eg boot drives) you have
> to run partx in initramfs or it wont be able to access them.

Not in this case.  I'm netbooting and getting / and co. from the SAN
via multipath and LVM, not a single partition on the whole cluster...
-- 
Thanks for the tip!
Feri.

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: how to disable partition search?
  2008-04-06 22:25   ` Ferenc Wagner
@ 2008-04-07 18:02     ` Ferenc Wagner
  0 siblings, 0 replies; 6+ messages in thread
From: Ferenc Wagner @ 2008-04-07 18:02 UTC (permalink / raw)
  To: linux-raid

Ferenc Wagner <wferi@niif.hu> writes:

> Luca Berra <bluca@comedia.it> writes:
>
>> anyway you can rebuild your own kernel disabling it
>> just set PARTITION_ADVANCED, and disable all partition types.
>
> Sounds like a good idea, will try that!

Thanks again, it indeed works!  My boot messages are readable again.
Interestingly, it still says "unknown partition table", but doesn't
try to read it either.  More than good enough for me.

>> you should be aware that doing this will disable partition detection on
>> all drives, so if you have partitioned drives (eg boot drives) you have
>> to run partx in initramfs or it wont be able to access them.
>
> Not in this case.  I'm netbooting and getting / and co. from the SAN
> via multipath and LVM, not a single partition on the whole cluster...

And sure everything just worked with the new kernel.  Wonderful!
-- 
Thanks,
Feri.

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2008-04-07 18:02 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-03-26 11:43 how to disable partition search? Ferenc Wagner
2008-03-26 20:20 ` Bill Davidsen
2008-03-27  0:45   ` Ferenc Wagner
2008-04-06 10:37 ` Luca Berra
2008-04-06 22:25   ` Ferenc Wagner
2008-04-07 18:02     ` Ferenc Wagner

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).