Re: [MiNT] re Martin Eric racine, Guido Leaving?

[Frank Naumann <fnaumann@prinz-atm.CS.Uni-Magdeburg.De>]

Hi!

> > Every filesystem access over the new drivers chokes the buffer cache from 
> > kernel.
> 
> Btw, choke = block, as in, when writing many files the whole kernel is
> paused between inode writing batches.  In fact, not just the kernel, but -
> I suspect - every interrupt is put on hold until the next batch of inodes
> strored in the cache is written.  This freezes the system completely and
> only corrects itself only _after_ the cache is emptied.

You don't know anything about the internal working of the buffer cache, 
the MinixFS and the protocol between these two instances.

So please don't speak about inode accesses and interrupts and so.

> But, if you that it's okay for the whole OS to halt between cache
> emptying, well....

Nobody halts. If the cache is written back MiNT must wait until finish 
these operations (missing background DMA).

The same also occurs with old MinixFS 0.60. If in 0.60 the MinixFS 
internal cache is written back the complete system must wait until these 
operations are finished.

> > It seems that cp is your problem. cp report a bus error to me if I will 
> > copy lot of files (the bus error only occurs with memory protection 
> > enabled).
> 
> Without memory protection also.

And you know that cp without memory protection can overwrite all data in 
memory? Can you run your system with MP and retest?

> No, it doesn't.  Example:  yesterday, I had about 200 files I was
> transfering from 5 floppies to Minix.  Minix crashed and took Thing
> with itself...

Which version of FreeMiNT and MinixFS are you exactly using?


Tschuess
   ...Frank

--
ATARI FALCON 040
--------------------------------------
Internet: fnaumann@cs.uni-magdeburg.de
Mausnet:  Frank Naumann @ L2