more on 2.4.19pre... & swsusp

more on 2.4.19pre... & swsusp

[brian]

I found a .config difference between my 2.4.19-pre5-ac3 setup
and my 2.4.19-pre6 (swsusp v0.8 patched) setup.

After making both the same, both generally oops in the same place
as previously reported (oops via ksymoops previously posted).

Findings thus far:

   swsusp says it doesn't need APM. But it does. at least so far
   as menuconfig is concerned.

   With apm loaded (or built) in the kernel swsusp says it can't
   terminate/kill kapmd and gives up.

   With apm *not* loaded in the kernel swsusp oopses as previously
   reported.

Nice repeatable behavior.

Documentation/swsusp.txt which is repeatedly refered to does not
exist, either in the ac version or the v0.8 patch.

I have not been able to find the swsusp program which is also
refered to.

-- 
Brian Litzinger <brian@worldcontrol.com>

Re: more on 2.4.19pre... & swsusp

[Ed Sweetman]

On Sat, 2002-04-06 at 15:09, brian@worldcontrol.com wrote:
> I found a .config difference between my 2.4.19-pre5-ac3 setup
> and my 2.4.19-pre6 (swsusp v0.8 patched) setup.
> 
> After making both the same, both generally oops in the same place
> as previously reported (oops via ksymoops previously posted).
> 
> Findings thus far:
> 
>    swsusp says it doesn't need APM. But it does. at least so far
>    as menuconfig is concerned.
This is an error in the AC branch.   WOLK 3.2 configs fine.


>    With apm loaded (or built) in the kernel swsusp says it can't
>    terminate/kill kapmd and gives up.
> 
>    With apm *not* loaded in the kernel swsusp oopses as previously
>    reported.
> 
> Nice repeatable behavior.
> 
> Documentation/swsusp.txt which is repeatedly refered to does not
> exist, either in the ac version or the v0.8 patch.

Does not exist in the WOLK 3.2 release either. all documentation on
swsusp is horribly outdated.  I believe the website posts 2.4.9 as the
latest kernel it's been patched against. 

> I have not been able to find the swsusp program which is also
> refered to.

all programs are simply scripts people have written to fascilitate
suspending or resuming.  (turning off dma before suspending and such).  


swsusp does not need apm or acpi.  IT does need magic sysrq or acpi
however.  Which is something that's not quite documented.  Ie. You can't
do sysrq d (c in WOLK) if you dont have magic sysrq compiled in.  And
you can't echo "4" > /proc/acpi/sleep if you dont have acpi.  So one or
the other is required.  This could be an example of why they're putting
in a new build system for 2.5.  


swsusp works for me in wolk3.2 but it doesn't use rmap. Seems like the
swsusp patch was just hacked into ac so it would compile with little to
no changes from wolk3.2.   

Wolk 3.2 also uses the memeat patch which the list describes as being
necessary for many people to not oops on suspend.   This is not in the
ac branch.  


On a different note.  Why doesn't the ac branch have ftpfs yet?  Besides
the fact that it sometimes has problems with ls'ing a directory because
of a . handle error,  using mc to navigate works perfectly.  It deserves
a wider test audiance. Mounting ftp sites is a so amazingly convenient
it's hard to overestimate the coolness of ftpfs.  It patches cleanly
against the current ac as is.  

http://ftpfs.sourceforge.net/ 

Re: more on 2.4.19pre... & swsusp

[Alan Cox]

> On a different note.  Why doesn't the ac branch have ftpfs yet?  Besides
> the fact that it sometimes has problems with ls'ing a directory because

Because its perfectly doable in user space. Its for testing useful stuff not
a dumping ground

Re: more on 2.4.19pre... & swsusp

[Ed Sweetman]

On Sat, 2002-04-06 at 17:59, Alan Cox wrote:
> > On a different note.  Why doesn't the ac branch have ftpfs yet?  Besides
> > the fact that it sometimes has problems with ls'ing a directory because
> 
> Because its perfectly doable in user space. Its for testing useful stuff not
> a dumping ground
> -



Wouldn't that be true of any networked filesystem?  They should all be
able to be done in userspace.  The problem with that would be it loses
it's transparency to the user and increases latency.  Sure ftpfs can be
done in userspace, but the point of it is so i dont have to interface
with ftp's through a specific client.  I'm sure people would love it if
they had to open samba-view whenever they wanted to copy to and from
samba shares, same for nfs etc.  

There are more than a couple examples of things in the kernel that can
also be completely functional just done in userspace,  Both autofs (why
we continue to ship an older version when the newer one is reverse
compat is a mystery to me) implementations are two such examples.  

I have nothing against not including something for personal preferences
( it is your branch) or because something is too untested.. but because
it can be done in userspace just doesn't hold up when you look at some
of the things in the kernel already.   But i've wasted enough time
arguing about something that doesn't require any changes. heh 

Re: more on 2.4.19pre... & swsusp

[Pavel Machek]

Hi!

> > > On a different note.  Why doesn't the ac branch have ftpfs yet?  Besides
> > > the fact that it sometimes has problems with ls'ing a directory because
> > 
> > Because its perfectly doable in user space. Its for testing useful stuff not
> > a dumping ground
> 
> Wouldn't that be true of any networked filesystem?  They should all be
> able to be done in userspace.  The problem with that would be it loses
> it's transparency to the user and increases latency.  Sure ftpfs can be

uservfs.sf.net.
									Pavel
-- 
Philips Velo 1: 1"x4"x8", 300gram, 60, 12MB, 40bogomips, linux, mutt,
details at http://atrey.karlin.mff.cuni.cz/~pavel/velo/index.html.

Re: more on 2.4.19pre... & swsusp

[Pavel Machek]

Hi!

> swsusp works for me in wolk3.2 but it doesn't use rmap. Seems like the
> swsusp patch was just hacked into ac so it would compile with little to
> no changes from wolk3.2.   

What does it mean "doesn't use rmap"?

Oh, and it was not merged from wolk. Swsusp in both -wolk and -ac is hacked
from the same tree on my machine. I'm working on makng it work in -ac.

									Pavel
-- 
Philips Velo 1: 1"x4"x8", 300gram, 60, 12MB, 40bogomips, linux, mutt,
details at http://atrey.karlin.mff.cuni.cz/~pavel/velo/index.html.