Hi/2.

No problem. I think you'd better provide the patches implementing what
you said and the binaries with us in order to check the stability.

--- In fat32dev@yahoogroups.com, "barf1234de" <lars.erdmann@...> wrote:
>
> Hi,
>
> --- In fat32dev@yahoogroups.com, "KO Myung-Hun" <komh@> wrote:
> >
> > Hi/2.
> >
> > This is a GA version of FAT32 v0.9.13.
> >
> > And there are the following changes since v0.9.12.
> >
> > -----
> > - Fixed the problem that a program trying to READ/WRITE from/to memory
> > object
> > with OBJ_ANY attribute is crashed. But READ/WRITE performance is
> > decreased.
> > Reported by dixie.
>
> I looked at the 0.9.13 code. May I suggest the following for
> FS_READ/FS_WRITE:
> keep my code, create an additional 64 k local buffer (FS_READ/FS_WRITE
> never pass user buffer > 64k, because they cannot) then read into that
> local buffer and use memcpy to copy to final user buffer (FS_READ, and
> vice versa for FS_WRITE).
> That will at least keep the performance gain to read/write as many
> adjacent clusters as possible. Memcpy performance is considerably
> faster then calling FSH_DOVOLIO individually for each cluster.
>
> The problem is: FSH_DOVOLIO does not allow to work from/to buffer
> allocated via OBJ_ANY (for whatever reason). But memcpy does.
>
>
> Lars
>