j...@wxcvbn.org (Jeremie Courreges-Anglas) writes: > j...@wxcvbn.org (Jeremie Courreges-Anglas) writes: > >> "Todd C. Miller" <todd.mil...@courtesan.com> writes: >> >>> From source inspection, Net and Free appear to allow read(2) of >>> dirs to succeed. However, since Linux, Mac OS X and Solaris have >>> the EISDIR behavior I think it is probably safe from a portability >>> standpoint. >>> >>> We're long past the days when opendir(3)/readdir(3) used read(2)... >>> >>> HP-UX and AIX still allow reads of directories but no one cares >>> about them ;-) >> >> So I think that we agree that EISDIR is more useful, and seems safe from >> a portability POV. I've built base and x sets on i386, and ajacoutot >> ran the ports bulk builds. The two offenders in the ports tree were due >> to an unrelated glitch in base libtool which has since been fixed. > > I've hold this for a few days until the release cycle comes back to an > almost normal state. But, as Theo points out, we have to think about the > drawbacks. Some configurations that do work right now are very likely > to break. > > $ cat /; echo $? > cat: /: Is a directory > 1 > > is only a trivial example. > > I still think that it is better to error out instead of, for example, > silently ignoring invalid/mistyped CLI arguments and config parameters. > But I don't want to mindlessly break people's setups behind their backs, > so I'd like to hear more opinions and... *test reports*.
I haven't received a single test report, which is far from sufficient for such a change. Even though I'm convinced that such a change would be a benefit, I won't push this further. > Diff below for convenience. > > > Index: lib/libc/sys/read.2 > =================================================================== > RCS file: /cvs/src/lib/libc/sys/read.2,v > retrieving revision 1.35 > diff -u -p -r1.35 read.2 > --- lib/libc/sys/read.2 5 Feb 2015 02:33:09 -0000 1.35 > +++ lib/libc/sys/read.2 9 Jul 2016 17:20:39 -0000 > @@ -152,13 +152,15 @@ is not a valid file or socket descriptor > Part of > .Fa buf > points outside the process's allocated address space. > -.It Bq Er EIO > -An I/O error occurred while reading from the file system. > .It Bq Er EINTR > A read from a slow device > (i.e. one that might block for an arbitrary amount of time) > was interrupted by the delivery of a signal > before any data arrived. > +.It Bq Er EIO > +An I/O error occurred while reading from the file system. > +.It Bq Er EISDIR > +The underlying file is a directory. > .El > .Pp > In addition, > Index: sys/kern/vfs_vnops.c > =================================================================== > RCS file: /cvs/src/sys/kern/vfs_vnops.c,v > retrieving revision 1.85 > diff -u -p -r1.85 vfs_vnops.c > --- sys/kern/vfs_vnops.c 19 Jun 2016 11:54:33 -0000 1.85 > +++ sys/kern/vfs_vnops.c 9 Jul 2016 17:20:39 -0000 > @@ -336,11 +336,13 @@ vn_read(struct file *fp, off_t *poff, st > if (vp->v_type != VCHR && count > LLONG_MAX - *poff) > return (EINVAL); > > + if (vp->v_type == VDIR) > + return (EISDIR); > + > vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); > uio->uio_offset = *poff; > - if (vp->v_type != VDIR) > - error = VOP_READ(vp, uio, > - (fp->f_flag & FNONBLOCK) ? IO_NDELAY : 0, cred); > + error = VOP_READ(vp, uio, (fp->f_flag & FNONBLOCK) ? IO_NDELAY : 0, > + cred); > *poff += count - uio->uio_resid; > VOP_UNLOCK(vp, p); > return (error); -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
