Bug 2633 - fsck -p in /etc/rc fails with cannot alloc nnnn bytes for lncntp
Summary: fsck -p in /etc/rc fails with cannot alloc nnnn bytes for lncntp
Status: Closed FIXED
Alias: None
Product: Base System
Classification: Unclassified
Component: bin (show other bugs)
Version: Unspecified
Hardware: Any Any
: Normal Affects Only Me
Assignee: freebsd-bugs (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 1997-02-01 02:00 UTC by shimon
Modified: 1998-06-03 02:04 UTC (History)
0 users

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description shimon 1997-02-01 02:00:01 UTC
	When booting, fsck -p (in /etc/rc) fails with ``cannot alloc nnnn bytes for
	lncntp''  error, thus makes it impossible to boot into init level 2.

	This happens consistently for file systems which are about 4GB large.
	Does not seem to happen on smaller ones.
	Reports for more sever behavior under SMP kernels.

Fix: 

Workaround:  Comment out the fsck -p in /etc/rc and pray.
How-To-Repeat: 
	Create a file system $GB or more in size
	Boot the system and enjoy.
Comment 1 Joerg Wunsch 1997-02-01 13:13:09 UTC
As shimon@i-Connect.Net wrote:

> 	When booting, fsck -p (in /etc/rc) fails with ``cannot alloc
> nnnn bytes for lncntp'' error, thus makes it impossible to boot into
> init level 2.  This happens consistently for file systems which are
> about 4GB large.

Does the error go away if you enable swapping before?

-- 
cheers, J"org

joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/ -- NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)
Comment 2 Bill Fenner freebsd_committer freebsd_triage 1997-02-02 18:08:36 UTC
State Changed
From-To: open->open

It didn't.  darn that edit-pr anyway. 

Comment 3 shimon 1997-02-03 19:37:31 UTC
Hi J Wunsch;  On 01-Feb-97 you wrote: 
> As shimon@i-Connect.Net wrote:
> 
> >     When booting, fsck -p (in /etc/rc) fails with ``cannot alloc
> > nnnn bytes for lncntp'' error, thus makes it impossible to boot into
> > init level 2.  This happens consistently for file systems which are
> > about 4GB large.
> 
> Does the error go away if you enable swapping before?

I did not try.  Being that this is a standard /etc/rc, and the machine has
64MB of RAM, it should not have a problem.  I think.

Simon

P.S.  I will try.

Simon
Comment 4 shimon 1997-02-03 19:40:49 UTC
Hi J Wunsch;  On 01-Feb-97 you wrote: 
> As shimon@i-Connect.Net wrote:
> 
> >     When booting, fsck -p (in /etc/rc) fails with ``cannot alloc
> > nnnn bytes for lncntp'' error, thus makes it impossible to boot into
> > init level 2.  This happens consistently for file systems which are
> > about 4GB large.
> 
> Does the error go away if you enable swapping before?

If you look in /etc/rc, you will see that swap -a runs before the fsck.
I do use ccdconfig and have /etc/ccd.conf file.  These may be related.

Simon
Comment 5 Joerg Wunsch 1997-02-03 21:15:02 UTC
As Simon Shapiro wrote:

> > Does the error go away if you enable swapping before?
> 
> If you look in /etc/rc, you will see that swap -a runs before the fsck.

You'd like to say, i should *really* have looked before? :-)

On another chance, does fsck simply run out of VM limits?  See
ulimit...

-- 
cheers, J"org

joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/ -- NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)
Comment 6 shimon 1997-02-04 07:32:26 UTC
Hi J Wunsch;  On 03-Feb-97 you wrote: 
> As Simon Shapiro wrote:
> 
> > > Does the error go away if you enable swapping before?
> > 
> > If you look in /etc/rc, you will see that swap -a runs before the fsck.
> 
> You'd like to say, i should *really* have looked before? :-)
> 
> On another chance, does fsck simply run out of VM limits?  See
> ulimit...

Ulimit is ``unlimited'', unless it does something different (and silly?)
at boot time.

As this one is a basic, boot time, plain vanilla issue, I am sort of
``playing dumb'' here.  As I am still new here, am I expected to fix that
(since I spoke first :-) or there is an ``official maintainer'' for this
piece?

Simon
Comment 7 asami 1997-02-04 16:29:25 UTC
 *  > >     When booting, fsck -p (in /etc/rc) fails with ``cannot alloc
 *  > > nnnn bytes for lncntp'' error, thus makes it impossible to boot into
 *  > > init level 2.  This happens consistently for file systems which are
 *  > > about 4GB large.

I've seen this before, but not for 4GB disks.  More like 40GB.  I had
to raise the kernel data size limit (MAXDSIZ).

 *  > Does the error go away if you enable swapping before?
 *  
 *  If you look in /etc/rc, you will see that swap -a runs before the fsck.
 *  I do use ccdconfig and have /etc/ccd.conf file.  These may be related.

You are most likely running out of swap space, although it's hard to
believe with 64MB of RAM.  How much swap do you have?

Satoshi
Comment 8 shimon 1997-02-04 22:02:04 UTC
Hi Satoshi Asami;  On 04-Feb-97 you wrote: 
>  *  > >     When booting, fsck -p (in /etc/rc) fails with ``cannot alloc
>  *  > > nnnn bytes for lncntp'' error, thus makes it impossible to boot
into
>  *  > > init level 2.  This happens consistently for file systems which
are
>  *  > > about 4GB large.
> 
> I've seen this before, but not for 4GB disks.  More like 40GB.  I had
> to raise the kernel data size limit (MAXDSIZ).
> 
>  *  > Does the error go away if you enable swapping before?
>  *  
>  *  If you look in /etc/rc, you will see that swap -a runs before the
fsck.
>  *  I do use ccdconfig and have /etc/ccd.conf file.  These may be related.
> 
> You are most likely running out of swap space, although it's hard to
> believe with 64MB of RAM.  How much swap do you have?
> 
> Satoshi

128MB.  All this at boot time, from /etc/rc.  I doubt I am running out of
swap space at that point.  especially when NOTHING else runs out of swap
space later.  If the filesystem is anything less than 4GB, it works fine.
Have you considered looking at the fact that fsck may be doing lseek or
some other computation, based on the SIZE of the partition in bytes,
encountering an integer overflow, or sign extention problem and trying to
malloc (for example) a negative number, or a very small one, etc.

Do we have llseek(2) in FreeBSD?

Simon
Comment 9 shimon 1997-02-04 22:37:15 UTC
Hi Satoshi Asami;  On 04-Feb-97 you wrote: 

...

> I've seen this before, but not for 4GB disks.  More like 40GB.  I had
> to raise the kernel data size limit (MAXDSIZ).

This is the most dangerous parameter in the system!  Just grepping for
it has froze the machine solid :-)  I need a serial console...

It is set to 256MB.  Should be enough...

Simon
Comment 10 Steve Price freebsd_committer freebsd_triage 1998-06-03 02:02:07 UTC
State Changed
From-To: open->closed

The originator says that after tweaking some parameters in login.conf 
the problem seems to have gone away.