> On 12. Jan 2020, at 18:49, Andrew Doran <ad%netbsd.org@localhost> wrote:
>
> Module Name: src
> Committed By: ad
> Date: Sun Jan 12 17:49:17 UTC 2020
>
> Modified Files:
> src/sys/kern: vfs_vnode.c
>
> Log Message:
> vput(): don't drop the vnode lock, carry the hold over into vrelel() which
> might need it anyway.
>
>
> To generate a diff of this commit:
> cvs rdiff -u -r1.106 -r1.107 src/sys/kern/vfs_vnode.c
>
> Please note that diffs are not public domain; they are subject to the
> copyright notices on the relevant files.
>
vput(vnode_t *vp)
{
+ int lktype;
- VOP_UNLOCK(vp);
- vrele(vp);
+ if ((vp->v_vflag & VV_LOCKSWORK) == 0) {
+ lktype = LK_EXCLUSIVE;
+ } else {
+ lktype = VOP_ISLOCKED(vp);
+ KASSERT(lktype != LK_NONE);
+ }
+ mutex_enter(vp->v_interlock);
+ vrelel(vp, 0, lktype);
}
This is quite wrong, from the manual:
VOP_ISLOCKED(vp)
Test if the vnode vp is locked. Possible return values are
LK_EXCLUSIVE, LK_SHARED or 0 for lock held exclusively by the
calling thread, shared lock held by anyone or unlocked,
respectively.
This function must never be used to make locking decisions at
run time: it is provided only for diagnostic purposes.
I suppose you cannot determine if the current thread holds
a shared lock.
--
J. Hannken-Illjes - hannken%eis.cs.tu-bs.de@localhost - TU Braunschweig
Attachment:
signature.asc
Description: Message signed with OpenPGP