Mostly revert a5970a529c: Make files opened with O_PATH to not block non-forced unmount

Problem is that open(O_PATH) on nullfs -o nocache is broken then,
because there is no reference on the vnode after the open syscall exits.

Reported and tested by:	ambrisko
Reviewed by:	markj
Sponsored by:	The FreeBSD Foundation
MFC after:	1 week
This commit is contained in:
Konstantin Belousov 2022-04-13 03:47:45 +03:00
parent a30eb84a5d
commit bf13db086b
3 changed files with 2 additions and 7 deletions

View file

@ -28,7 +28,7 @@
.\" @(#)open.2 8.2 (Berkeley) 11/16/93
.\" $FreeBSD$
.\"
.Dd October 9, 2021
.Dd April 22, 2022
.Dt OPEN 2
.Os
.Sh NAME
@ -351,9 +351,6 @@ But operations like
and any other that operate on file and not on file descriptor (except
.Xr fstat 2 ),
are not allowed.
File opened with the
.Dv O_PATH
flag does not prevent non-forced unmount of the volume it belongs to.
.Pp
A file descriptor created with the
.Dv O_PATH

View file

@ -5138,7 +5138,7 @@ path_close(struct file *fp, struct thread *td)
{
MPASS(fp->f_type == DTYPE_VNODE);
fp->f_ops = &badfileops;
vdrop(fp->f_vnode);
vrele(fp->f_vnode);
return (0);
}

View file

@ -1219,8 +1219,6 @@ kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
if ((flags & O_PATH) != 0) {
finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
DTYPE_VNODE, NULL, &path_fileops);
vhold(vp);
vunref(vp);
} else {
finit_vnode(fp, flags, NULL, &vnops);
}