[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference turris::digital_unix

Title:DIGITAL UNIX(FORMERLY KNOWN AS DEC OSF/1)
Notice:Welcome to the Digital UNIX Conference
Moderator:SMURF::DENHAM
Created:Thu Mar 16 1995
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:10068
Total number of notes:35879

9671.0. "Kernel tuning on OSF V4.0A" by NNTPD::"[email protected]" (LudmilaMala) Wed Apr 30 1997 03:21

I have a client who is trying to tune his kernel on AlphaServer 1000A 4/266,
running Digital UNIX V4.0A.

On loading the objectstore database he gets error:

objectstore errno 22
mummap failed: c1-mmap:60f2a000 2000
(err_internal)

which apparently is a know error and is a known error, normally solved
by doubling the size of mapentries parameter ( to 10240 ) in kernel 
configuration file. 

Originally it worked for them, but with increasing sizes of their files,
it keeps comming back. They are now on 40960 and still having problems.

Can anybody give us some explanation what this mapentries parameter means,
and why they are getting are getting an unmapping error, rather than a
mapping error?

He basically wants to know what he is doing, rather than just using braille.

Thanks in advance for every bit of wisdom, pointers, etc.

Regards,

Ludmila



[Posted by WWW Notes gateway]
T.RTitleUserPersonal
Name		DateLines
9671.1ObjectStore error?GIDDAY::STRAUSStalking through my binocularsWed Apr 30 1997 18:4420
    Hi Ludmila
    
    I think this message may come from ObjectStore, not UNIX.
    errno 22 is ...
    
    	#define EINVAL          22              /* Invalid argument */
    
    The man page for munmap(2) says that EINVAL can occur if ...
    
      [EINVAL]  The addr parameter is not a multiple of the page size as
    returned
                by sysconf(_SC_PAGE_SIZE).
    
    Beyond that I don't know if we can help him much.  He should talk to
    his ObjectStore support organsiation (I hope that's not us ;-)
    
    Hope this helps
    
    	leon
9671.2unmapping errorNNTPD::"[email protected]"Larry WoodmanThu May 01 1997 10:2420
Probably what is happening here is that when a single file or anonymous
memory gets mmaped it uses one map entry and the unmap is only a subset
file that got mapped, in the center of the address space that was mapped.
This uses 2 map entries for the remaining pieces of the address space.
Can you verify whether or not this is happening???


	   after mmap()					after munmap()

	-------------------------		------------------------------
	|	mapped		|		| mapped | unmapped | mapped |
	-------------------------		------------------------------
		    ^				     ^			 ^
		    |				     |			 |
		one map entry			one map enrty	    one map entry
	


Larry Woodman
[Posted by WWW Notes gateway]
9671.3NNTPD::"[email protected]"Shashi MangalatFri May 02 1997 01:3614
This is a known problem.  In addition to increasing the mapentries limit
you will also need one of the following patches (depending on which OS release
you have installed).

  V4.0  : OSF400-200
  V4.0A : OSF405-400200
  V4.0B : OSF410-052

This fixes a regression in the munmap() system call.  ObjectStore is unmapping
regions that are not mapped (which is really a bug in their code), so munmap()
fails the request.

--shashi
[Posted by WWW Notes gateway]
9671.4SolutionNNTPD::"mala@ stl.dec.com"Ludmila MalaSun May 04 1997 22:3211
Firstly, I want to thanks everybody for their contribution.

I have learned more about mapentries that I ever wanted to know. However my
problems
was little bit more devious then that, he had the entry in the configuration
file entered
a wrong way. Getting that stright fixed the problem, now I'll just patch him
up.

Thanks for help.
[Posted by WWW Notes gateway]