d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vnode read(2)/write(2): acquire rangelock regardless of do_vn_io_fault()

Browse Source 
To ensure atomicity of reads against parallel writes and truncates,
vnode lock was not enough at least since introduction of vn_io_fault().
That code only take rangelock when it was possible that vn_read() and
vn_write() could drop the vnode lock.

At least since the introduction of VOP_READ_PGCACHE() which generally
does not lock the vnode at all, rangelocks become required even
for filesystems that do not need vn_io_fault() workaround.  For
instance, tmpfs.

PR:	272678
Analyzed and reviewed by:	Andrew Gierth  <andrew@tao11.riddles.org.uk>
Sponsored by:	The FreeBSD Foundation
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D41158
This commit is contained in:
Konstantin Belousov 2023-07-23 18:55:50 +03:00
parent 4a07c77863
commit 5b353925ff
1 changed files with 14 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
sys/kern/vfs_vnops.c
View File

@ -1443,6 +1443,7 @@ vn_io_fault(struct file *fp, struct uio *uio, struct ucred *active_cred,
void *rl_cookie;
struct vn_io_fault_args args;
int error;
bool rl_locked;
doio = uio->uio_rw == UIO_READ ? vn_read : vn_write;
vp = fp->f_vnode;
@ -1465,12 +1466,7 @@ vn_io_fault(struct file *fp, struct uio *uio, struct ucred *active_cred,
}
foffset_lock_uio(fp, uio, flags);
if (do_vn_io_fault(vp, uio)) {
args.kind = VN_IO_FAULT_FOP;
args.args.fop_args.fp = fp;
args.args.fop_args.doio = doio;
args.cred = active_cred;
args.flags = flags | FOF_OFFSET;
if (vp->v_type == VREG) {
if (uio->uio_rw == UIO_READ) {
rl_cookie = vn_rangelock_rlock(vp, uio->uio_offset,
uio->uio_offset + uio->uio_resid);
@ -1482,11 +1478,22 @@ vn_io_fault(struct file *fp, struct uio *uio, struct ucred *active_cred,
rl_cookie = vn_rangelock_wlock(vp, uio->uio_offset,
uio->uio_offset + uio->uio_resid);
}
rl_locked = true;
} else {
rl_locked = false;
}
if (do_vn_io_fault(vp, uio)) {
args.kind = VN_IO_FAULT_FOP;
args.args.fop_args.fp = fp;
args.args.fop_args.doio = doio;
args.cred = active_cred;
args.flags = flags | FOF_OFFSET;
error = vn_io_fault1(vp, uio, &args, td);
vn_rangelock_unlock(vp, rl_cookie);
} else {
error = doio(fp, uio, active_cred, flags | FOF_OFFSET, td);
}
if (rl_locked)
vn_rangelock_unlock(vp, rl_cookie);
foffset_unlock_uio(fp, uio, flags);
return (error);
}