Hi!
19-Июл-2004 01:23 [EMAIL PROTECTED] (Eric Auer) wrote to
[EMAIL PROTECTED]:
[...]
EA> The DosGetFree function (remember my bug report...) calls
EA> flush_buffers if called from FatGetDrvData (int 21.1c?), maybe this is
EA> where it crashed? Otherwise it also is int 21.36 ...
EA> Plus, how do we know that this does not get a NULL pointer? -->
EA> if (*nc != 0xffff)
EA> *navc = (COUNT) dos_free(dpbp);
In theory, *nc will never 0xffff (because it points to DX, and DL on
input is a drive number). Thus, you right, there is bug (which garbages
value of INT0 vector).
EA> That fake for 64k cluster size looks evil, too.
EA> while (ntotal > FAT_MAGIC16 && cluster_size < 0x8000)
EA> { SHL cluster size and spc, SHR ntotal and nfree }
EA> ... seems to be a cluster size determination loop which should probably
No. As I understand, this loop is used to imitate FAT1x over FAT32. For
example - _known_ size of cluster is increased up to 32k to decrease ntotal
("total clusters on drive"). Example: we have cluster 4k and total clusters
0x10000 (long value). After increasing cluster size to 32k we get (16-bit!)
value 0x2000 for "total clusters" (and return 0x2000 instead 0xffff).
[...]
EA> BOOL dos_setfsize(COUNT fd, LONG size)
EA> should be ULONG, as it already is for dos_getfsize!
[...]
I wrote all these small fixes in my todos. When I will change these
files, I will fix also this.
EA> void far *DynAlloc(char *what, unsigned num, unsigned size)
EA> ...
EA> if ((ULONG) total + Dynp->Allocated > 0xffff)
EA> ... should be possible to do that without having to use ULONG.
:) Yes. "if (total > ~Dynp->allocated)".
EA> STATIC int LBA_Get_Drive_Parameters(int drive, struct DriveParamS
EA> *driveParam)
-->> bug:
EA> if (lba_bios_parameters.heads > 0xffff ||
EA> lba_bios_parameters.sectors > 0xffff ||
EA> lba_bios_parameters.totalSectHigh != 0)
EA> {
EA> printf("Drive is too large to handle, using only 1st 8 GB\n"
EA> ...
EA> Explanation: totalSectHigh nonzero means either buggy BIOS (happens!) or
EA> disk size above 2 TeraBytes! This does definitely NOT mean that you should:
EA> ...
EA> goto StandardBios;
This is not bug, this is, probably, misfeature.
EA> ... not at all! Just use LBA and clip the useable disk size to 0xFFFFFFFFUL
EA> sectors.
Nice idea, but Bart and Brian also should give their opinion.
EA> Can somebody explain - in rwblock - what this does and why it uses UCOUNT?:
EA> UCOUNT sectors_to_xfer, sectors_wanted;
What wonder you?
EA> sectors_wanted /= secsize;
EA> ... can this be optimized to use a shift instead? 16bit>>value -> 16bit.
DPB doesn't contains value "bits per sector_size value" (similar to
dpb_shftcnt).
EA> I saw the compile time option WRITEZEROS, could that be extended to
EA> be supported by shrink_file (or maybe in rwblock itself?) ...?
?
EA> Reading LBA_Transfer now, very interesting... play_dj (TODO: should call
EA> the int 2f interface which allows to suppress the message!!),
?
EA> int 1e stuff
EA> and disk reset, track wrap avoidance for CHS, DMA wrap avoidance -->
EA> ** You should be allowed to disable DMA_max_transfer check for non-floppy!?
Why? DMA aligned access is useful in any case.
EA> ** You should be allowed to disable track wrap protection for non-floppy!?
?
EA> Wow. That was huge.
Yes, this letter _very_ huge. In future, please, split such long things
over more letters.
-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idG21&alloc_id040&op=click
_______________________________________________
Freedos-kernel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/freedos-kernel
