1) Fix a deadlock in the VFS:

- threadA runs vfs_rel(mp1)
- threadB does unmount the mp1 fs, sets MNTK_UNMOUNT and drop MNT_ILOCK()
- threadA runs vfs_busy(mp1) and, as long as, MNTK_UNMOUNT is set, sleeps
  waiting for threadB to complete the unmount
- threadB, in vfs_mount_destroy(), finds mnt_lock > 0 and sleeps waiting
  for the refcount to expire.

Fix the deadlock by adding a flag called MNTK_REFEXPIRE which signals the
unmounter is waiting for mnt_ref to expire.
The vfs_busy contenders got awake, fails, and if they retry the
MNTK_REFEXPIRE won't allow them to sleep again.

2) Simplify significantly the code of vfs_mount_destroy() trimming
   unnecessary codes:
   - as long as any reference exited, it is no-more possible to have
     write-op (primarty and secondary) in progress.
   - it is no needed to drop and reacquire the mount lock.
   - filling the structures with dummy values is unuseful as long as
     it is going to be freed.

Tested by:	pho, Andrea Barberio <insomniac at slackware dot it>
Discussed with:	kib
This commit is contained in:
Attilio Rao 2008-12-16 23:16:10 +00:00
parent fd14c50bbb
commit 4a0f807602
3 changed files with 16 additions and 32 deletions

View File

@ -507,29 +507,20 @@ vfs_mount_destroy(struct mount *mp)
{
MNT_ILOCK(mp);
mp->mnt_kern_flag |= MNTK_REFEXPIRE;
if (mp->mnt_kern_flag & MNTK_MWAIT) {
mp->mnt_kern_flag &= ~MNTK_MWAIT;
wakeup(mp);
}
while (mp->mnt_ref)
msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
if (mp->mnt_writeopcount > 0) {
printf("Waiting for mount point write ops\n");
while (mp->mnt_writeopcount > 0) {
mp->mnt_kern_flag |= MNTK_SUSPEND;
msleep(&mp->mnt_writeopcount,
MNT_MTX(mp),
PZERO, "mntdestroy2", 0);
}
printf("mount point write ops completed\n");
}
if (mp->mnt_secondary_writes > 0) {
printf("Waiting for mount point secondary write ops\n");
while (mp->mnt_secondary_writes > 0) {
mp->mnt_kern_flag |= MNTK_SUSPEND;
msleep(&mp->mnt_secondary_writes,
MNT_MTX(mp),
PZERO, "mntdestroy3", 0);
}
printf("mount point secondary write ops completed\n");
}
MNT_IUNLOCK(mp);
KASSERT(mp->mnt_ref == 0,
("%s: invalid refcount in the drain path @ %s:%d", __func__,
__FILE__, __LINE__));
if (mp->mnt_writeopcount != 0)
panic("vfs_mount_destroy: nonzero writeopcount");
if (mp->mnt_secondary_writes != 0)
panic("vfs_mount_destroy: nonzero secondary_writes");
mp->mnt_vfc->vfc_refcount--;
if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
struct vnode *vp;
@ -538,18 +529,10 @@ vfs_mount_destroy(struct mount *mp)
vprint("", vp);
panic("unmount: dangling vnode");
}
MNT_ILOCK(mp);
if (mp->mnt_kern_flag & MNTK_MWAIT)
wakeup(mp);
if (mp->mnt_writeopcount != 0)
panic("vfs_mount_destroy: nonzero writeopcount");
if (mp->mnt_secondary_writes != 0)
panic("vfs_mount_destroy: nonzero secondary_writes");
if (mp->mnt_nvnodelistsize != 0)
panic("vfs_mount_destroy: nonzero nvnodelistsize");
mp->mnt_writeopcount = -1000;
mp->mnt_nvnodelistsize = -1000;
mp->mnt_secondary_writes = -1000;
if (mp->mnt_lockref != 0)
panic("vfs_mount_destroy: nonzero lock refcount");
MNT_IUNLOCK(mp);
#ifdef MAC
mac_mount_destroy(mp);

View File

@ -345,7 +345,7 @@ vfs_busy(struct mount *mp, int flags)
MNT_ILOCK(mp);
MNT_REF(mp);
if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
if (flags & MBF_NOWAIT) {
if (flags & MBF_NOWAIT || mp->mnt_kern_flag & MNTK_REFEXPIRE) {
MNT_REL(mp);
MNT_IUNLOCK(mp);
return (ENOENT);

View File

@ -316,6 +316,7 @@ void __mnt_vnode_markerfree(struct vnode **mvp, struct mount *mp);
#define MNTK_SOFTDEP 0x00000004 /* async disabled by softdep */
#define MNTK_NOINSMNTQ 0x00000008 /* insmntque is not allowed */
#define MNTK_DRAINING 0x00000010 /* lock draining is happening */
#define MNTK_REFEXPIRE 0x00000020 /* refcount expiring is happening */
#define MNTK_UNMOUNT 0x01000000 /* unmount in progress */
#define MNTK_MWAIT 0x02000000 /* waiting for unmount to finish */
#define MNTK_SUSPEND 0x08000000 /* request write suspension */