bin/122925: sftp(1) duplicates filename when get listing directory on CDROM

[Stas Timokhin]

The following reply was made to PR bin/122925; it has been noted by GNATS.

From: Stas Timokhin <stast at bsdportal.ru>
To: Jaakko Heinonen <jh at saunalahti.fi>
Cc: Bruce Cran <bruce at cran.org.uk>,
 bug-followup at freebsd.org
Subject: Re: bin/122925: sftp(1) duplicates filename when get listing directory on CDROM
Date: Tue, 20 May 2008 15:44:15 +0700

 =F7 =D3=CF=CF=C2=DD=C5=CE=C9=C9 =CF=D4 Tuesday 20 May 2008 14:48:33 Jaakko =
 Heinonen =CE=C1=D0=C9=D3=C1=CC(=C1):
 > On 2008-05-18, Bruce Cran wrote:
 > >  This is occurring because sftp-server expects readdir(3) to return NULL
 > >  for a given DIR* twice in a row after all the files have been retrieve=
 d.
 > >    It seems that under certain conditions that isn't true.
 >
 > Thanks for the test case.
 >
 > The problem seems to lie in cd9660_readdir()
 > (src/sys/fs/cd9660/cd9660_vnops.c). The problem is that if we have
 > reached end of file (directory) and enter again to cd9660_readdir()
 > the idp->uio_off variable is not initialized correctly. In the end of
 > the function the file offset is set to idp->uio_off. So this basically
 > means that the file offset changes to a random value. This causes
 > effects such readdir(3) starting again at some position or g_vfs_done()
 > errors when data is attempted to read from bogus offset.
 >
 > I believe that this patch fixes the problem:
 >
 > Index: cd9660_vnops.c
 > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
 =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
 =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
 > RCS file: /home/ncvs/src/sys/fs/cd9660/cd9660_vnops.c,v
 > retrieving revision 1.113
 > diff -p -u -r1.113 cd9660_vnops.c
 > --- cd9660_vnops.c	15 Feb 2007 22:08:34 -0000	1.113
 > +++ cd9660_vnops.c	20 May 2008 06:45:20 -0000
 > @@ -495,6 +495,7 @@ cd9660_readdir(ap)
 >  	}
 >  	idp->eofflag =3D 1;
 >  	idp->curroff =3D uio->uio_offset;
 > +	idp->uio_off =3D uio->uio_offset;
 >
 >  	if ((entryoffsetinblock =3D idp->curroff & bmask) &&
 >  	    (error =3D cd9660_blkatoff(vdp, (off_t)idp->curroff, NULL, &bp))) {
 >
 > Could you please test the patch?
 Yes, i  done this test. Everything looks good, thanks !