[pups] missing space in 2.11_rp_unknown

[pups] missing space in 2.11_rp_unknown

[Christian Groessler]

Hi,

On 02/22/2002 07:31:49 PM PST "Steven M. Schultz" wrote:
>
>> From: Christian Groessler <cpg at aladdin.de>
>> I'm running above image with the p11 emulator, and the root partition
>> is almost full.
>>
>> # df
>> Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
>> root             7816     7030      786    10%    /
>>
>> but looking at files, I only see 2MB+ in use:
>>
>> # du -s /
>> 2702
>
>	I see Greg mentioned running fsck.   That sounds like an excellent
>	suggestion.

Yes, but it didn't help :-(
What can this be?

I tried something else, I copied the contents of the root fs
elsewhere, newfs'd the root partition and copied the contents back.

But now booting stops when it normally starts init,

-------------
: unix
Boot: bootdev=05010 bootcsr=0176700

2.11 BSD UNIX #1: Fri Feb 15 18:47:18 PST 2002
    chris at pdp11:/usr/src/sys/PDP11CPG

attaching qe0 csr 174440
qe0: DEC DEQNA addr 08:00:2b:07:82:6c
attaching lo0

phys mem  = 2097152
avail mem = 1647872
user mem  = 307200

-------------

... and here it hangs. Do I have to consider something else when I
newfs the root partition?

regards,
chris

[pups] missing space in 2.11_rp_unknown

[Christian Groessler]

Hi,

On 03/03/2002 01:47:34 AM PST "Steven M. Schultz" wrote:
>
>	Glad to hear you're current and are not seeing 'df' weirdness.   You
>	may want to upgrade P11 to 2.9 though - that would, I think, have
>	fixed that problem before you saw it ;)

No, wouldn't :-)
I used 2.10a all the time.

regards,
chris

[pups] missing space in 2.11_rp_unknown

[Steven M. Schultz]

Howdy -

> From: Christian Groessler <cpg at aladdin.de>
> I just last week installed v6 from tape image. I have to admit, I like
> working boot images more :-) (Since I only have an emulator and not

	An emualtor of course would use an emulated tape image ;)   That's
	how I typically install.   There are instructions and sources to
	the program to create the 'virtual tape' from the dump and tar files.

> the real thing, I don't have the need to physically transfer the stuff.)

	I keep forgetting that not everyone has a SCSI<->Qbus adaptor :)

	(it was _expensive_ at the time but gosh, after 10 years the initial
	 $ pain is long long gone and I've gotten a lot of use out of it)

> Hmm, I just did this now, but I have to admit, I only browsed the
> instructions of most of them. I followed the instructions of 412/413
> because I feared I'd forget to update init before rebooting the new kernel.

	Yes, screwing up 'init' is, to put it mildly, catastrophic.   During
	the development and testing of that set of updates I did render my
	system unbootable.   Thankfully I had a spare OS installed on the
	SCSI Zip drive - I just booted from "DU 1" and put back a working
	'init' (turns out that a 100MB Zip disk can contain a *full* 2.11BSD
	system - not a lot of space left, but it includes all sources and
	will boot).

> But otherwise I applied all patches to 442, and then rebuilt the
> kernel, rebooted, and did "make build; make installsrc". Seemed to work.

	That's fantastic to hear!  

> I noticed 2 patches, which patched /usr/src/sys/GENERIC/Makefile, but
> this is a generated file I think. At least it wasn't present, because

	Yes and No.  YES - it is generated by running './config' in /sys/conf.
	NO - it's an integral part of the OS as distributed.

> I removed /usr/src/sys/GENERIC.

	You really didn't want to do that ;)

	The Make* files for custom kernels will (100% guaranteed) diverge
	from the defaults.   That's expected.   The GENERIC kernel is a special
	case though.  When changes are made to the Make* files (overlay sizes
	change for example) the patches will not attempt to find and "fix"
	any locally created kernels - but the guarantee has always been that
	the GENERIC kernel _will_ build, thus the patches presume that the
	/sys/GENERIC directory hasn't been removed.   Indeed the kernel patches
	usually suggest rebuilding GENERIC.

	It is a Good Idea (saved my system a couple times) to keep a known
	good working _non_networking kernel (i.e. GENERIC) in /genunix.   That
	way if you are tinkering around (or a bad patch ends up in /unix) you
	have something to boot.   Many is the time (during development, testing
	of course) that I've had to rely on a /genunix to get the system
	back alive.

	Glad to hear you're current and are not seeing 'df' weirdness.   You
	may want to upgrade P11 to 2.9 though - that would, I think, have 
	fixed that problem before you saw it ;)

	Cheers,
	Steven Schultz
	sms at 2bsd.com

[pups] missing space in 2.11_rp_unknown

[Christian Groessler]

Hi,

On 02/28/2002 08:53:34 AM PST "Steven M. Schultz" wrote:
>
>> From: Christian Groessler <cpg at aladdin.de>
>> >	Look at the /VERSION file.  The first or second line will have
>> >	the patchlevel.   That file's updated by each patch.
>>
>> I have 400. I assume www.2bsd.com contains the newest patches? So 442
>> is the latest?
>
>	It _might_ be easier to create an install tape from the files in
>	the 2.11 portion of the PUPS archive - I think that was updated
>	to about patchlevel 432 or so.   There is documentation on how to
>	create a boot tape, etc from the compressed files.

I just last week installed v6 from tape image. I have to admit, I like
working boot images more :-) (Since I only have an emulator and not
the real thing, I don't have the need to physically transfer the
stuff.)

>	On the other hand it might be instructive/interesting/whatever to
>	apply the 42 updates manually - just be sure to read the instructions
>	that come with each one :)

Hmm, I just did this now, but I have to admit, I only browsed the
instructions of most of them. I followed the instructions of 412/413
because I feared I'd forget to update init before rebooting the new
kernel.
But otherwise I applied all patches to 442, and then rebuilt the
kernel, rebooted, and did "make build; make installsrc". Seemed to
work.

I noticed 2 patches, which patched /usr/src/sys/GENERIC/Makefile, but
this is a generated file I think. At least it wasn't present, because
I removed /usr/src/sys/GENERIC.

regards,
chris

[pups] missing space in 2.11_rp_unknown

[Steven M. Schultz]

Hi -

> From: Christian Groessler <cpg at aladdin.de>
> >	Look at the /VERSION file.  The first or second line will have
> >	the patchlevel.   That file's updated by each patch.
> 
> I have 400. I assume www.2bsd.com contains the newest patches? So 442
> is the latest?

	Wow, that is quite old.

	A faster link is at FTP.TO.GD-ES.COM (that's a T-1 vs the ISDL 
	link I have at home).

	It _might_ be easier to create an install tape from the files in
	the 2.11 portion of the PUPS archive - I think that was updated
	to about patchlevel 432 or so.   There is documentation on how to
	create a boot tape, etc from the compressed files.

	On the other hand it might be instructive/interesting/whatever to
	apply the 42 updates manually - just be sure to read the instructions
	that come with each one :)

	Cheers,

	Steven

[pups] missing space in 2.11_rp_unknown

[Christian Groessler]

Hi,

On 02/27/2002 03:25:22 PM PST "Steven M. Schultz" wrote:
>
>Hello again -
>
>> From: Christian Groessler <cpg at aladdin.de>
>> Regarding the patchlevels, how do I find out which patchlevel my
>> system is at?
>
>	Look at the /VERSION file.  The first or second line will have
>	the patchlevel.   That file's updated by each patch.

I have 400. I assume www.2bsd.com contains the newest patches? So 442
is the latest?

regards,
chris

[pups] missing space in 2.11_rp_unknown

[Frank Wortner]

FYI,  here's the patch that fixes the "df" problem that Christian reported.
If I recall correctly,  it took me about a day to figure out that the
problem wasn't in df,  wasn't in the 2.11 BSD C library,  but was in the p11
emulator.  The fact that the same disk image booted by Bob Supnik's SIMH
PDP/11 emulator produced correct "df" results was the big clue.  Here's the
fix for p11 (version 2.7).

$ diff -c float.h*
*** float.h     Tue Oct 10 10:36:14 2000
--- float.h.orig        Sat Mar  4 03:03:29 2000
***************
*** 442,448 ****
  # define CnvL(A,L)    (void)((A) = F_int((L)))
  # define CnvF2L(S,L)  (void)((L) = F_chop((S)))

! # define FrExp(A)     ((signed)((A).e - EOFFS))

  # define GetMant(P)   (((P)->m & ~f_msb) << 1)
  # define GetExp(P)    ((int)(P)->e - EOFFS + 1)
--- 442,448 ----
  # define CnvL(A,L)    (void)((A) = F_int((L)))
  # define CnvF2L(S,L)  (void)((L) = F_chop((S)))

! # define FrExp(A)     ((A).e - EOFFS)

  # define GetMant(P)   (((P)->m & ~f_msb) << 1)
  # define GetExp(P)    ((int)(P)->e - EOFFS + 1)

-- 
Frank

"I don't hold with all this washing.  This modern Behind-the-ears nonsense."
* Eeyore, "Winnie the Pooh"

[pups] missing space in 2.11_rp_unknown

[Steven M. Schultz]

Hello again -

> From: Christian Groessler <cpg at aladdin.de>
> >	Mmmm, I wonder if the problems you were having  were caused by
> >	/dev not being correctly populated.
> 
> Maybe. I noticed they're missing and recreated them by hand. Perhaps I
> made a mistake there.

	It would be easy enough to do - or perhaps a critical one was
	left out.    Filesystems without device nodes can be moved
	with a 'tar' pipeline but the root filesystem is special.

> It's a problem of the p11 emulator I use. I got a patch off-list which
> fixed it. It was some signed/unsigned thing.

	Ah ha!

> Regarding the patchlevels, how do I find out which patchlevel my
> system is at?

	Look at the /VERSION file.  The first or second line will have
	the patchlevel.   That file's updated by each patch.

	Cheers,
	Steven

[pups] missing space in 2.11_rp_unknown

[Christian Groessler]

Hi,

On 02/26/2002 03:29:07 PM PST "Steven M. Schultz" wrote:
>
>	Mmmm, I wonder if the problems you were having  were caused by
>	/dev not being correctly populated.

Maybe. I noticed they're missing and recreated them by hand. Perhaps I
made a mistake there.


>> $ df
>> Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
>> /dev/xp0a        7816     2658     5158    04%    /
>> /dev/xp0g      151625   117599    34026    08%    /usr
>> $
>>
>> Btw, the capacity values look a bit strange?
>
>	Yes, they do look (more than a little bit) strange.
>
>	On my system here (a P11 based emulated PDP-11 - I have a real 11/73
>	but it is only powered up when I'm actively testing):
>
>Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
>/dev/xp0a        8228     3163     5065    38%    /
>/dev/xp0h      155328    84188    71140    54%    /usr
>
>	What patchlevel did you mention the system was at?   There were a lot
>	of patches issued after the ' 2.11_rp_unknown' image was created.
>	One thing, which probably will not make any difference, to try would
>	be to recompile 'df' (and possibly 'libc') and see if the problem
>	changes.   Looks like it's a math error of some kind so either
>	the compiler/libraries are broken or P11's having a problem doing
>	arithmetic.

It's a problem of the p11 emulator I use. I got a patch off-list which
fixed it. It was some signed/unsigned thing.

Regarding the patchlevels, how do I find out which patchlevel my
system is at?

regards,
chris

[pups] missing space in 2.11_rp_unknown

[Steven M. Schultz]

Hi!

> From: Christian Groessler <cpg at aladdin.de>
> >	Did you use dump+restor?
> 
> No, tar. I tried again with dump and restor and now it works! Thanks
> for the hint! I seldomly use dump/restore.
	
	Ah ha!  For moving filesystems dump+restor or 'afio' need to be
	used.   Dump+restor also have the advantage of preserving the
	file flags (see chflags(2) and chflags(1)) - other utilities do
	not preserve that metadata.

	The other thing that dump+restor (or afio) handle correctly is
	the special files in /dev.  'tar' does not know how to archive
	files such as "/dev/rp0a".

	Mmmm, I wonder if the problems you were having  were caused by
	/dev not being correctly populated.

> Now there's enough space in /:
> 
> $ df
> Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
> /dev/xp0a        7816     2658     5158    04%    /
> /dev/xp0g      151625   117599    34026    08%    /usr
> $ 
> 
> Btw, the capacity values look a bit strange?

	Yes, they do look (more than a little bit) strange.

	On my system here (a P11 based emulated PDP-11 - I have a real 11/73
	but it is only powered up when I'm actively testing):

Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
/dev/xp0a        8228     3163     5065    38%    /
/dev/xp0h      155328    84188    71140    54%    /usr

	What patchlevel did you mention the system was at?   There were a lot
	of patches issued after the ' 2.11_rp_unknown' image was created.
	One thing, which probably will not make any difference, to try would
	be to recompile 'df' (and possibly 'libc') and see if the problem
	changes.   Looks like it's a math error of some kind so either
	the compiler/libraries are broken or P11's having a problem doing
	arithmetic.

	Cheers,
	Steven Schultz
	sms at 2bsd.com

[pups] missing space in 2.11_rp_unknown

[Christian Groessler]

Hi,

On 02/25/2002 09:16:31 AM PST "Steven M. Schultz" wrote:
>
>Hi -
>
>> >	I see Greg mentioned running fsck.   That sounds like an excellent
>> >	suggestion.
>>
>> Yes, but it didn't help :-(
>
>> What can this be?
>
>	It might be necessary to use the '-s' option .   "fsck -s" will
>	unconditionally rebuild the freelist.

This didn't work either.

>
>> I tried something else, I copied the contents of the root fs
>> elsewhere, newfs'd the root partition and copied the contents back.
>
>	Did you use dump+restor?

No, tar. I tried again with dump and restor and now it works! Thanks
for the hint! I seldomly use dump/restore.

Now there's enough space in /:

$ df
Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
/dev/xp0a        7816     2658     5158    04%    /
/dev/xp0g      151625   117599    34026    08%    /usr
$ 

Btw, the capacity values look a bit strange?

regards,
chris

[pups] missing space in 2.11_rp_unknown

[Steven M. Schultz]

Hi -

> >	I see Greg mentioned running fsck.   That sounds like an excellent
> >	suggestion.
> 
> Yes, but it didn't help :-(

> What can this be?

	It might be necessary to use the '-s' option .   "fsck -s" will
	unconditionally rebuild the freelist.

> I tried something else, I copied the contents of the root fs
> elsewhere, newfs'd the root partition and copied the contents back.

	Did you use dump+restor?

> But now booting stops when it normally starts init,
> 
	Oh no!

> -------------
> : unix
> Boot: bootdev=05010 bootcsr=0176700
> 
> 2.11 BSD UNIX #1: Fri Feb 15 18:47:18 PST 2002
>     chris at pdp11:/usr/src/sys/PDP11CPG
> 
> attaching qe0 csr 174440
> qe0: DEC DEQNA addr 08:00:2b:07:82:6c
> attaching lo0
> 
> phys mem  = 2097152
> avail mem = 1647872
> user mem  = 307200
> 
> -------------
> 
> ... and here it hangs. Do I have to consider something else when I
> newfs the root partition?

	The boot block, /boot, /unix, /netnix and /etc/init, /bin/sh are 
	intact since the system got as far as printing the memory numbers.

	After the memory stats the '/etc/autoconfig' process should be
	run ('init' runs it) and the device probes should take place.

	The only thing I can think of (and it's a wild guess) is that the
	"clock" isn't running - thru the boot process clock interrupts 
	aren't used but when 'init' goes to run 'autoconfig' the system nees
	clock interrupts in order to drive the context switching.   Either
	the clock isn't running or /etc/autoconfig got corrupted somehow
	in the copying.

	Steven Schultz
	sms at 2bsd.com

[pups] missing space in 2.11_rp_unknown

[Steven M. Schultz]

Hi!

> From: Christian Groessler <cpg at aladdin.de>
> I'm running above image with the p11 emulator, and the root partition
> is almost full.
> 
> # df
> Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
> root             7816     7030      786    10%    /
> 
> but looking at files, I only see 2MB+ in use:
> 
> # du -s /
> 2702    

> This persists over reboots, so it doesn't seem to be a large deleted
> file which is still in use.

	It might be a corrupt freelist.   If that is the case then running
	fsck will detect that fact and reclaim the space by rebuilding the
	freelist.

> Where is the missing space?

	My guess is it's "missing" - that can happen if the system's shutdown
	(or the emulator terminated) prematurely.   In that case the freelist
	metadata might not have been updated.

	I see Greg mentioned running fsck.   That sounds like an excellent
	suggestion.

	Cheers,
	Steven Schultz
	sms at 2bsd.com

[pups] missing space in 2.11_rp_unknown

[Greg Lehey]

On Friday, 22 February 2002 at 15:52:13 +0100, Christian Groessler wrote:
> Hi,
>
> I'm running above image with the p11 emulator, and the root partition
> is almost full.
> I tried to clean it up, but I cannot find where the space is used.
>
> df says:
>
> --------
> # df
> Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
> root             7816     7030      786    10%    /
> --------
>
> but looking at files, I only see 2MB+ in use:
>
> --------
> # du -s /
> 2702
> --------
>
> This persists over reboots, so it doesn't seem to be a large deleted
> file which is still in use.

Have you tried running fsck?

Greg
--
Finger grog at lemis.com for PGP public key
See complete headers for address and phone numbers

[pups] missing space in 2.11_rp_unknown

[Christian Groessler]

Hi,

I'm running above image with the p11 emulator, and the root partition
is almost full.
I tried to clean it up, but I cannot find where the space is used.

df says:

--------
# df
Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
root             7816     7030      786    10%    /
--------

but looking at files, I only see 2MB+ in use:

--------
# du -s /
2702    
--------

This persists over reboots, so it doesn't seem to be a large deleted
file which is still in use.

Where is the missing space?

regards,
chris