headers
itojun | 2 Mar 2004 00:36

1.6.2 -> current: /usr/include/kerberosIV/kafs.h

	when upgrading 1.6.2 to current, the existence of
	/usr/include/kerberosIV/kafs.h will break the build.  shouldn't we
	have "remove obsolete include files" process during the build process
	(before or after "make includes" process)?

itojun
Permalink | Reply | 7 comments |
headers
Luke Mewburn | 2 Mar 2004 01:36
Picon

Re: 1.6.2 -> current: /usr/include/kerberosIV/kafs.h

On Tue, Mar 02, 2004 at 08:36:22AM +0900, itojun <at> iijlab.net wrote:
  | 	when upgrading 1.6.2 to current, the existence of
  | 	/usr/include/kerberosIV/kafs.h will break the build.  shouldn't we
  | 	have "remove obsolete include files" process during the build process
  | 	(before or after "make includes" process)?

This type of "chicken & egg" problem with the host includes being used
as the build includes is one of the reasons we have build.sh and we
strongly recommened building into a DESTDIR ...

Cheers,
Luke.
Permalink | Reply | 6 comments |
headers
nmahajan | 2 Mar 2004 07:11

file creation decreases memory !


Hi,

we are facing a problem in our system runinng netbsd.
we are trying to create a file on our file system (compact flash mounted
ffs and memory based file system for system image and NO SWAP mounted !) ..

we are observing that the file copied decreases the top shown free memory ?
and increases the File memory shown by top i.e
Memory: 50M Act, 72M Inact, 220K Wired, 1944K Exec, 110M File, 96M Free

The File and Free are proportinately increased and decreased by same amount
respectively.

This goes on even if we issue sync command (wasnt it supposed to write back
these dirty pages back to disk ?)
io_flush kernel thread has also been seen running but doesnt changes these
top output. ..
the file in question being  say 38 MB .. and command is " cp file1 file 2"

Can some body explain why it is happening ..

My questions /:
what happens further is that if say a program asks for memory it goes down
saying out of swap ??
so why isnt the file system being written back to ??
Could declaration of no swap affecting the vnode handling by netbsd ??
we dont have swap declared / mounted is all I can see as a difference
between a server runinng netbsd where also repetitive file copies does
decreases memory free as reported by top but after a point the error shown
(Continue reading)

Permalink | Reply | 1 comment |
headers
David Young | 2 Mar 2004 09:11
Picon
Favicon

Re: PATCH: disklabel & fdisk regular files

On Tue, Feb 17, 2004 at 07:39:06PM +1100, Luke Mewburn wrote:
> On Tue, Feb 17, 2004 at 01:46:38AM -0600, David Young wrote:
>   | It's useful sometimes for a privilegeless user to disklabel or fdisk
>   | a file containing a disk image---a file was made by makefs, say. Here
>   | are patches that make it possible. The disklabel patch is pretty simple:
>   | detect that the target is a file and avoid certain steps.  Because fdisk
>   | needs a disklabel to get started, my patch lets you give it a disktab
>   | and disktype with -t and -T options.
> 
> I'd prefer that disklabel(8) took a specific option to use a regular
> file to enable this behaviour; other disk related tools use '-F' for this.

Done. Patches attached.

> As for fdisk; is there a way we can implement '-F' to enable the
> "don't bother disklabeling" stuff, rather than the -t/-T stuff ?

There is a way.  I can provide -F by "faking up" the disk geometry.
Patches attached.

I had used -t/-T because it was expedient, but I would like for it
to stick around. It lets one avoid "faking up" C/H/S for the file,
if one knows the geometry of the device where one will dd(1) the disk
image later.

Dave

--

-- 
David Young             OJC Technologies
dyoung <at> ojctech.com      Urbana, IL * (217) 278-3933
(Continue reading)

Permalink | Reply | 1 comment |
headers
nmahajan | 2 Mar 2004 11:12

swap space for a memory based filesystem


Hi,

We have a system which boots using memory based file system,
i.e we have a ramdisk having some file system mounted on it and parameters
for it passed via kernel options file ..
but I couldnt find any option / way to define swap partition and space for
same ..
could some one please tell me how to declare and allocate space for swap
for it ?
I couldnt find any  relavant option in swapctl and swapon ..  ?

Regards,
Nitin
Permalink | Reply | 0 comments |
headers
Jaromir Dolecek | 2 Mar 2004 11:57
Picon

Re: swap space for a memory based filesystem

nmahajan <at> hss.hns.com wrote:
> We have a system which boots using memory based file system,
> i.e we have a ramdisk having some file system mounted on it and parameters
> for it passed via kernel options file ..
> but I couldnt find any option / way to define swap partition and space for
> same ..
> could some one please tell me how to declare and allocate space for swap
> for it ?
> I couldnt find any  relavant option in swapctl and swapon ..  ?

It's probably not very useful to have swap on MFS - it just wastes
additional memory for swap structures. You can safely run without
any swap configured.

Jaromir
--

-- 
Jaromir Dolecek <jdolecek <at> NetBSD.org>            http://www.NetBSD.cz/
-=- We should be mindful of the potential goal, but as the Buddhist -=-
-=- masters say, ``You may notice during meditation that you        -=-
-=- sometimes levitate or glow.   Do not let this distract you.''   -=-
Permalink | Reply | 0 comments |
headers
nmahajan | 2 Mar 2004 13:57

Re: swap space for a memory based filesystem


Hi,

actually what you are saying made sense to me also but we are finding a
peculiar problem:
(from my earlier post)
we are facing a problem in our system runinng netbsd.
we are trying to create a file on our file system (compact flash mounted
ffs and memory based file system for system image and NO SWAP mounted !) ..

we are observing that the file copied decreases the top shown free memory ?
and increases the File memory shown by top i.e
Memory: 50M Act, 72M Inact, 220K Wired, 1944K Exec, 110M File, 96M Free

The File and Free are proportinately increased and decreased by same amount
respectively.

This goes on even if we issue sync command (wasnt it supposed to write back
these dirty pages back to disk ?)
io_flush kernel thread has also been seen running but doesnt changes these
top output. ..
the file in question being  say 38 MB .. and command is " cp file1 file 2"

Can some body explain why it is happening ..

My questions /:
what happens further is that if say a program asks for memory it goes down
saying out of swap ??
so why isnt the file system being written back to ??
Could declaration of no swap affecting the vnode handling by netbsd ??
(Continue reading)

Permalink | Reply | 0 comments |
headers
cube | 2 Mar 2004 14:28

Re: swap space for a memory based filesystem

On Tue, Mar 02, 2004 at 06:27:11PM +0530, nmahajan <at> hss.hns.com wrote:
[...]
> actually what you are saying made sense to me also but we are finding a
> peculiar problem:
> (from my earlier post)
> we are facing a problem in our system runinng netbsd.
> we are trying to create a file on our file system (compact flash mounted
> ffs and memory based file system for system image and NO SWAP mounted !) ..
> 
> we are observing that the file copied decreases the top shown free memory ?
> and increases the File memory shown by top i.e
> Memory: 50M Act, 72M Inact, 220K Wired, 1944K Exec, 110M File, 96M Free
> 
> The File and Free are proportinately increased and decreased by same amount
> respectively.
> 
> This goes on even if we issue sync command (wasnt it supposed to write back
> these dirty pages back to disk ?)

Who says everything classified as 'File' is dirty?  What would be the point
of caching the data if the pages were to be freed immediately after being
written?

> io_flush kernel thread has also been seen running but doesnt changes these
> top output. ..
> the file in question being  say 38 MB .. and command is " cp file1 file 2"
> 
> Can some body explain why it is happening ..

The kernel keeps the cached data.  That sounds rather reasonable to me.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Greg Troxel | 2 Mar 2004 16:29
Picon

Re: 1.6.2 -> current: /usr/include/kerberosIV/kafs.h

For what it's worth, I just took a 1.6.2ish (not quite) system to
-current (sources from mid february).  The build (with build.sh) went
fine, and I tried to do the install after placing the current kernel
in / but before booting it.  The base set installed, and then I got
bad syscall errors on the rest.  So I rebooted into the 1.6ZI kernel
(which I know the instructions say; I'm used to 1.6->1.6 upgrades
where you can get away with this) and did the install again, and all
seems ok.

--

-- 
        Greg Troxel <gdt <at> ir.bbn.com>
Permalink | Reply | 5 comments |
headers
Manuel Bouyer | 2 Mar 2004 20:55

Re: file creation decreases memory !

On Tue, Mar 02, 2004 at 11:41:48AM +0530, nmahajan <at> hss.hns.com wrote:
> 
> Hi,
> 
> 
> [...]
> 
> My questions /:
> what happens further is that if say a program asks for memory it goes down
> saying out of swap ??
> so why isnt the file system being written back to ??

I've seen this happen on NetBSD 1.6 (you don't say which release you're
running) and I think this a consequence of what Jason exposed on tech-kern a
few days ago ("Page reactivation path in pdaemon"). I don't think there
is a solution for this yet. A workaround may be to tune the vm.* sysctls
to reduce vm.filemax to something like 20, and maybe bump the exec and anon
parameters.

--

-- 
Manuel Bouyer <bouyer <at> antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--
Permalink | Reply | 0 comments |

Gmane