freebsd-dev/sys/fs/fdescfs/fdesc_vnops.c
Robert Watson 5e3f7694b1 Replace custom file descriptor array sleep lock constructed using a mutex
and flags with an sxlock.  This leads to a significant and measurable
performance improvement as a result of access to shared locking for
frequent lookup operations, reduced general overhead, and reduced overhead
in the event of contention.  All of these are imported for threaded
applications where simultaneous access to a shared file descriptor array
occurs frequently.  Kris has reported 2x-4x transaction rate improvements
on 8-core MySQL benchmarks; smaller improvements can be expected for many
workloads as a result of reduced overhead.

- Generally eliminate the distinction between "fast" and regular
  acquisisition of the filedesc lock; the plan is that they will now all
  be fast.  Change all locking instances to either shared or exclusive
  locks.

- Correct a bug (pointed out by kib) in fdfree() where previously msleep()
  was called without the mutex held; sx_sleep() is now always called with
  the sxlock held exclusively.

- Universally hold the struct file lock over changes to struct file,
  rather than the filedesc lock or no lock.  Always update the f_ops
  field last. A further memory barrier is required here in the future
  (discussed with jhb).

- Improve locking and reference management in linux_at(), which fails to
  properly acquire vnode references before using vnode pointers.  Annotate
  improper use of vn_fullpath(), which will be replaced at a future date.

In fcntl(), we conservatively acquire an exclusive lock, even though in
some cases a shared lock may be sufficient, which should be revisited.
The dropping of the filedesc lock in fdgrowtable() is no longer required
as the sxlock can be held over the sleep operation; we should consider
removing that (pointed out by attilio).

Tested by:	kris
Discussed with:	jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00

537 lines
12 KiB
C

/*-
* Copyright (c) 1992, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software donated to Berkeley by
* Jan-Simon Pendry.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)fdesc_vnops.c 8.9 (Berkeley) 1/21/94
*
* $FreeBSD$
*/
/*
* /dev/fd Filesystem
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/conf.h>
#include <sys/dirent.h>
#include <sys/filedesc.h>
#include <sys/kernel.h> /* boottime */
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/malloc.h>
#include <sys/file.h> /* Must come after sys/malloc.h */
#include <sys/mount.h>
#include <sys/namei.h>
#include <sys/proc.h>
#include <sys/stat.h>
#include <sys/vnode.h>
#include <fs/fdescfs/fdesc.h>
#define FDL_WANT 0x01
#define FDL_LOCKED 0x02
static int fdcache_lock;
#define NFDCACHE 4
#define FD_NHASH(ix) \
(&fdhashtbl[(ix) & fdhash])
static LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl;
static u_long fdhash;
static vop_getattr_t fdesc_getattr;
static vop_inactive_t fdesc_inactive;
static vop_lookup_t fdesc_lookup;
static vop_open_t fdesc_open;
static vop_readdir_t fdesc_readdir;
static vop_reclaim_t fdesc_reclaim;
static vop_setattr_t fdesc_setattr;
static struct vop_vector fdesc_vnodeops = {
.vop_default = &default_vnodeops,
.vop_access = VOP_NULL,
.vop_getattr = fdesc_getattr,
.vop_inactive = fdesc_inactive,
.vop_lookup = fdesc_lookup,
.vop_open = fdesc_open,
.vop_pathconf = vop_stdpathconf,
.vop_readdir = fdesc_readdir,
.vop_reclaim = fdesc_reclaim,
.vop_setattr = fdesc_setattr,
};
/*
* Initialise cache headers
*/
int
fdesc_init(vfsp)
struct vfsconf *vfsp;
{
fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash);
return (0);
}
int
fdesc_allocvp(ftype, ix, mp, vpp, td)
fdntype ftype;
int ix;
struct mount *mp;
struct vnode **vpp;
struct thread *td;
{
struct fdhashhead *fc;
struct fdescnode *fd;
int error = 0;
fc = FD_NHASH(ix);
loop:
LIST_FOREACH(fd, fc, fd_hash) {
if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
if (vget(fd->fd_vnode, 0, td))
goto loop;
*vpp = fd->fd_vnode;
return (error);
}
}
/*
* otherwise lock the array while we call getnewvnode
* since that can block.
*/
if (fdcache_lock & FDL_LOCKED) {
fdcache_lock |= FDL_WANT;
(void) tsleep( &fdcache_lock, PINOD, "fdalvp", 0);
goto loop;
}
fdcache_lock |= FDL_LOCKED;
/*
* Do the MALLOC before the getnewvnode since doing so afterward
* might cause a bogus v_data pointer to get dereferenced
* elsewhere if MALLOC should block.
*/
MALLOC(fd, struct fdescnode *, sizeof(struct fdescnode), M_TEMP, M_WAITOK);
error = getnewvnode("fdesc", mp, &fdesc_vnodeops, vpp);
if (error) {
FREE(fd, M_TEMP);
goto out;
}
(*vpp)->v_data = fd;
fd->fd_vnode = *vpp;
fd->fd_type = ftype;
fd->fd_fd = -1;
fd->fd_ix = ix;
/* XXX: vnode should be locked here */
error = insmntque(*vpp, mp); /* XXX: Too early for mpsafe fs */
if (error != 0) {
free(fd, M_TEMP);
*vpp = NULLVP;
goto out;
}
LIST_INSERT_HEAD(fc, fd, fd_hash);
out:
fdcache_lock &= ~FDL_LOCKED;
if (fdcache_lock & FDL_WANT) {
fdcache_lock &= ~FDL_WANT;
wakeup( &fdcache_lock);
}
return (error);
}
/*
* vp is the current namei directory
* ndp is the name to locate in that directory...
*/
static int
fdesc_lookup(ap)
struct vop_lookup_args /* {
struct vnode * a_dvp;
struct vnode ** a_vpp;
struct componentname * a_cnp;
} */ *ap;
{
struct vnode **vpp = ap->a_vpp;
struct vnode *dvp = ap->a_dvp;
struct componentname *cnp = ap->a_cnp;
char *pname = cnp->cn_nameptr;
struct thread *td = cnp->cn_thread;
struct file *fp;
int nlen = cnp->cn_namelen;
u_int fd;
int error;
struct vnode *fvp;
if ((cnp->cn_flags & ISLASTCN) &&
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
error = EROFS;
goto bad;
}
if (cnp->cn_namelen == 1 && *pname == '.') {
*vpp = dvp;
VREF(dvp);
return (0);
}
if (VTOFDESC(dvp)->fd_type != Froot) {
error = ENOTDIR;
goto bad;
}
fd = 0;
/* the only time a leading 0 is acceptable is if it's "0" */
if (*pname == '0' && nlen != 1) {
error = ENOENT;
goto bad;
}
while (nlen--) {
if (*pname < '0' || *pname > '9') {
error = ENOENT;
goto bad;
}
fd = 10 * fd + *pname++ - '0';
}
if ((error = fget(td, fd, &fp)) != 0)
goto bad;
error = fdesc_allocvp(Fdesc, FD_DESC+fd, dvp->v_mount, &fvp, td);
fdrop(fp, td);
if (error)
goto bad;
VTOFDESC(fvp)->fd_fd = fd;
if (fvp != dvp)
vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY, td);
*vpp = fvp;
return (0);
bad:
*vpp = NULL;
return (error);
}
static int
fdesc_open(ap)
struct vop_open_args /* {
struct vnode *a_vp;
int a_mode;
struct ucred *a_cred;
struct thread *a_td;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
if (VTOFDESC(vp)->fd_type == Froot)
return (0);
/*
* XXX Kludge: set td->td_proc->p_dupfd to contain the value of the the file
* descriptor being sought for duplication. The error return ensures
* that the vnode for this device will be released by vn_open. Open
* will detect this special error and take the actions in dupfdopen.
* Other callers of vn_open or VOP_OPEN will simply report the
* error.
*/
ap->a_td->td_dupfd = VTOFDESC(vp)->fd_fd; /* XXX */
return (ENODEV);
}
static int
fdesc_getattr(ap)
struct vop_getattr_args /* {
struct vnode *a_vp;
struct vattr *a_vap;
struct ucred *a_cred;
struct thread *a_td;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
struct vattr *vap = ap->a_vap;
struct file *fp;
struct stat stb;
u_int fd;
int error = 0;
switch (VTOFDESC(vp)->fd_type) {
case Froot:
VATTR_NULL(vap);
vap->va_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
vap->va_type = VDIR;
vap->va_nlink = 2;
vap->va_size = DEV_BSIZE;
vap->va_fileid = VTOFDESC(vp)->fd_ix;
vap->va_uid = 0;
vap->va_gid = 0;
vap->va_blocksize = DEV_BSIZE;
vap->va_atime.tv_sec = boottime.tv_sec;
vap->va_atime.tv_nsec = 0;
vap->va_mtime = vap->va_atime;
vap->va_ctime = vap->va_mtime;
vap->va_gen = 0;
vap->va_flags = 0;
vap->va_rdev = 0;
vap->va_bytes = 0;
break;
case Fdesc:
fd = VTOFDESC(vp)->fd_fd;
if ((error = fget(ap->a_td, fd, &fp)) != 0)
return (error);
bzero(&stb, sizeof(stb));
error = fo_stat(fp, &stb, ap->a_td->td_ucred, ap->a_td);
fdrop(fp, ap->a_td);
if (error == 0) {
VATTR_NULL(vap);
vap->va_type = IFTOVT(stb.st_mode);
vap->va_mode = stb.st_mode;
#define FDRX (VREAD|VEXEC)
if (vap->va_type == VDIR)
vap->va_mode &= ~((FDRX)|(FDRX>>3)|(FDRX>>6));
#undef FDRX
vap->va_nlink = 1;
vap->va_flags = 0;
vap->va_bytes = stb.st_blocks * stb.st_blksize;
vap->va_fileid = VTOFDESC(vp)->fd_ix;
vap->va_size = stb.st_size;
vap->va_blocksize = stb.st_blksize;
vap->va_rdev = stb.st_rdev;
/*
* If no time data is provided, use the current time.
*/
if (stb.st_atimespec.tv_sec == 0 &&
stb.st_atimespec.tv_nsec == 0)
nanotime(&stb.st_atimespec);
if (stb.st_ctimespec.tv_sec == 0 &&
stb.st_ctimespec.tv_nsec == 0)
nanotime(&stb.st_ctimespec);
if (stb.st_mtimespec.tv_sec == 0 &&
stb.st_mtimespec.tv_nsec == 0)
nanotime(&stb.st_mtimespec);
vap->va_atime = stb.st_atimespec;
vap->va_mtime = stb.st_mtimespec;
vap->va_ctime = stb.st_ctimespec;
vap->va_uid = stb.st_uid;
vap->va_gid = stb.st_gid;
}
break;
default:
panic("fdesc_getattr");
break;
}
if (error == 0)
vp->v_type = vap->va_type;
return (error);
}
static int
fdesc_setattr(ap)
struct vop_setattr_args /* {
struct vnode *a_vp;
struct vattr *a_vap;
struct ucred *a_cred;
struct thread *a_td;
} */ *ap;
{
struct vattr *vap = ap->a_vap;
struct vnode *vp;
struct mount *mp;
struct file *fp;
unsigned fd;
int error;
/*
* Can't mess with the root vnode
*/
if (VTOFDESC(ap->a_vp)->fd_type == Froot)
return (EACCES);
fd = VTOFDESC(ap->a_vp)->fd_fd;
/*
* Allow setattr where there is an underlying vnode.
*/
error = getvnode(ap->a_td->td_proc->p_fd, fd, &fp);
if (error) {
/*
* getvnode() returns EINVAL if the file descriptor is not
* backed by a vnode. Silently drop all changes except
* chflags(2) in this case.
*/
if (error == EINVAL) {
if (vap->va_flags != VNOVAL)
error = EOPNOTSUPP;
else
error = 0;
}
return (error);
}
vp = fp->f_vnode;
if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) == 0) {
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td);
error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_td);
VOP_UNLOCK(vp, 0, ap->a_td);
vn_finished_write(mp);
}
fdrop(fp, ap->a_td);
return (error);
}
#define UIO_MX 16
static int
fdesc_readdir(ap)
struct vop_readdir_args /* {
struct vnode *a_vp;
struct uio *a_uio;
struct ucred *a_cred;
int *a_eofflag;
u_long *a_cookies;
int a_ncookies;
} */ *ap;
{
struct uio *uio = ap->a_uio;
struct filedesc *fdp;
struct dirent d;
struct dirent *dp = &d;
int error, i, off, fcnt;
/*
* We don't allow exporting fdesc mounts, and currently local
* requests do not need cookies.
*/
if (ap->a_ncookies)
panic("fdesc_readdir: not hungry");
if (VTOFDESC(ap->a_vp)->fd_type != Froot)
panic("fdesc_readdir: not dir");
off = (int)uio->uio_offset;
if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 ||
uio->uio_resid < UIO_MX)
return (EINVAL);
i = (u_int)off / UIO_MX;
fdp = uio->uio_td->td_proc->p_fd;
error = 0;
fcnt = i - 2; /* The first two nodes are `.' and `..' */
FILEDESC_SLOCK(fdp);
while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) {
switch (i) {
case 0: /* `.' */
case 1: /* `..' */
bzero((caddr_t)dp, UIO_MX);
dp->d_fileno = i + FD_ROOT;
dp->d_namlen = i + 1;
dp->d_reclen = UIO_MX;
bcopy("..", dp->d_name, dp->d_namlen);
dp->d_name[i + 1] = '\0';
dp->d_type = DT_DIR;
break;
default:
if (fdp->fd_ofiles[fcnt] == NULL) {
FILEDESC_SUNLOCK(fdp);
goto done;
}
bzero((caddr_t) dp, UIO_MX);
dp->d_namlen = sprintf(dp->d_name, "%d", fcnt);
dp->d_reclen = UIO_MX;
dp->d_type = DT_UNKNOWN;
dp->d_fileno = i + FD_DESC;
break;
}
/*
* And ship to userland
*/
FILEDESC_SUNLOCK(fdp);
error = uiomove(dp, UIO_MX, uio);
if (error)
goto done;
FILEDESC_SLOCK(fdp);
i++;
fcnt++;
}
FILEDESC_SUNLOCK(fdp);
done:
uio->uio_offset = i * UIO_MX;
return (error);
}
static int
fdesc_inactive(ap)
struct vop_inactive_args /* {
struct vnode *a_vp;
struct thread *a_td;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
/*
* Clear out the v_type field to avoid
* nasty things happening in vgone().
*/
vp->v_type = VNON;
return (0);
}
static int
fdesc_reclaim(ap)
struct vop_reclaim_args /* {
struct vnode *a_vp;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
struct fdescnode *fd = VTOFDESC(vp);
LIST_REMOVE(fd, fd_hash);
FREE(vp->v_data, M_TEMP);
vp->v_data = 0;
return (0);
}