Hi!
29-Мар-2004 21:05 [EMAIL PROTECTED] (Bernd Blaauw) wrote to
[EMAIL PROTECTED]:
>> Wrong.
BB> it was a proposed algorythm for a patch.
From your letter I suggest that you try explain how you understand how
it works.
BB> since YOUR patch assumes 6 buffers minimum,
No, my patch _allows_ to allocate less than 20 buffers (but not less
than 6, as it is/was be).
BB> I thought I could use that as a minimum amount
BB> (instead of allowing "buffers=1" )
Currently, there is allowed to allocate only one buffer by buffers=-1.
Of course, this is bug, and without my patch kernel tries to deallocate all
buffers for INT 2F/4A02. (I already enumerate all related bugs, which fixes
my patch).
BB> that would explain my algorythm:
BB> *HMA must have space to store the minimum amount of 6 YOU mentioned.
This limit was selected before me.
BB> *requested number of buffers must be at least this minimum. (6<=N)
BB> *request number of buffers must fit in free HMA (N<=M)
With my patch, if requested buffers not fit into HMA, they allocated in
lower memory. And I don't see the reason, why we should cut value, requested
by user - if by some reason he wants 80 buffers, kernel shouldn't limit
buffers by 27 (current value of buffers in HMA).
BB> *number of requested buffers must be below 100, as 99 is the maximum.
if{M>>=6}
BB> {
BB> if(6<=N<=M<100) --> mark N buffers as used
BB> after marking, fill HMA with N=min(M,99) buffers //
BB> fill up until full or
BB> 99 reached.
BB> }
Let me quote my patch:
______________O\_/_________________________________\_/O______________
unsigned buffers = 0;
/* fill HMA with buffers if BUFFERS count >=0 and DOS in HMA */
if (wantedbuffers < 0)
wantedbuffers = -wantedbuffers;
else if (HMAState == HMA_DONE)
buffers = (0xFFF0 - HMAFree) / sizeof (struct buffer);
if (wantedbuffers < 6) /* min 6 buffers */
wantedbuffers = 6;
if (wantedbuffers > 99) /* max 99 buffers */
{
printf("BUFFERS=%u not supported, reducing to 99\n", wantedbuffers);
wantedbuffers = 99;
}
if (buffers < wantedbuffers) /* not less than requested */
buffers = wantedbuffers;
LoL->nbuffers = buffers;
_____________________________________________________________________
O/~\ /~\O
BB> if any of above conditions fail, buffers are stored in conventional memory.
BB> and that explains my use of "-" (negative value)
In current case (also as with my patch) negative value used only to
prevent buffer count expansion (to fill rest of HMA).
BB> Tom said the requested number of buffers are marked as "used" in HMA,
They are _not_ marked as "used" - there is pointer into HMA, which
points to Nth buffer in buffers array. Memory after this pointer may be
later reused for INT 2F/4A01 - for this all buffers, which intersects with
requested ared will be removed from chain. My patch changes (fixes) the
pointer behavior: now it always points to HMA (if present), even if buffers
are in lower memory; this allows to reuse rest of HMA notwithstanding
buffers placement.
BB> and that the ones used for filling up could be used for a (non-existing..)
BB> HMA-using program instead.
I repeat again: memory, allocated for buffers over requested count,
later may be removed to make memory available for INT 2F/4A02.
Because there are no function to release memory (by application), this
memory should be allocated for permanent needings (ie., for TSR). Currently
I know only one TSR, which uses HMA for its needings (so, tom is wrong when
says that such TSR are non-existing).
-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id70&alloc_id638&op=click
_______________________________________________
Freedos-kernel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/freedos-kernel
