On Fri, May 11, 2007 at 10:35:22PM +0300, Antti Kantee wrote:
> On Fri May 11 2007 at 21:08:44 +0200, Nicolas Joly wrote:
> > njoly@lanfeust [embl/blastwu]> ls -l embl.*
> > -rw-r--r--  1 997  997  12882733250 May  9 00:17 embl.xnd
> > -rw-r--r--  1 997  997  42205936685 May  9 00:17 embl.xns
> > -rw-r--r--  1 997  997    914661870 May  9 00:17 embl.xnt
> > 
> > njoly@lanfeust [embl/blastwu]> ~/mmap ./embl.xnt 
> > njoly@lanfeust [embl/blastwu]> ~/mmap ./embl.xnd
> > mmap: munmap failed: Invalid argument
> > njoly@lanfeust [embl/blastwu]> ~/mmap ./embl.xns
> > mmap: munmap failed: Invalid argument
> > 
> > Here follow the ktrace output for the last command:
> > 
> >   8950      1 mmap     CALL  open(0x7f7fffffed19,0,0x7f7fffffeba0)
> >   8950      1 mmap     NAMI  "./embl.xns"
> >   8950      1 mmap     RET   open 3
> >   8950      1 mmap     CALL  __fstat30(3,0x7f7fffffeaa0)
> >   8950      1 mmap     RET   __fstat30 0
> >   8950      1 mmap     CALL  mmap(0,0x9d3ab7c2d,1,1,3,0,0)
> >   8950      1 mmap     RET   mmap 140145487806464/0x7f762a048000
> >   8950      1 mmap     CALL  munmap(0x7f762a048000,0x9d3ab7c2d)
> >   8950      1 mmap     RET   munmap -1 errno 22 Invalid argument
> 
> Does the following patch help?  FWIW, I have no idea why the check is
> in there.

Yes, thanks.

> -	if ((int)size < 0)
> -		return (EINVAL);

In the mean time, i had a look to sys_mmap() and noticed a similar
check, except that the cast was not `int' but `ssize_t' ... and this
solved the problem too.

-- 
Nicolas Joly

Biological Software and Databanks.
Institut Pasteur, Paris.