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

Do not upgrade the vnode lock to call getinoquota().

Browse Source 
Doing so can deadlock when the thread already owns another vnode lock,
e.g. during a rename, as was demonstrated by the reporter.  In fact,
there seems to be no need to force the call to getinoquota() always,
because vn_open() locks vnode exclusively, and this is the most
important case.  To add to the point, directories where the dirent is
added or removed, are locked exclusively as well.

Reported by:	bwidawsk
Tested by:	bwidawsk, pho (as part of the larger patch)
Sponsored by:	The FreeBSD Foundation
Approved by:	re (gjb)
MFC after:	1 week
This commit is contained in:
Konstantin Belousov 2018-09-17 19:38:43 +00:00
parent 77e4a39103
commit a408841593
1 changed files with 7 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
sys/ufs/ufs/ufs_vnops.c
View File

@ -325,9 +325,6 @@ ufs_accessx(ap)
struct inode *ip = VTOI(vp);
accmode_t accmode = ap->a_accmode;
int error;
#ifdef QUOTA
int relocked;
#endif
#ifdef UFS_ACL
struct acl *acl;
acl_type_t type;
@ -350,32 +347,14 @@ ufs_accessx(ap)
* Inode is accounted in the quotas only if struct
* dquot is attached to it. VOP_ACCESS() is called
* from vn_open_cred() and provides a convenient
* point to call getinoquota().
* point to call getinoquota(). The lock mode is
* exclusive when the file is opening for write.
*/
if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
/*
* Upgrade vnode lock, since getinoquota()
* requires exclusive lock to modify inode.
*/
relocked = 1;
vhold(vp);
vn_lock(vp, LK_UPGRADE | LK_RETRY);
VI_LOCK(vp);
if (vp->v_iflag & VI_DOOMED) {
vdropl(vp);
error = ENOENT;
goto relock;
}
vdropl(vp);
} else
relocked = 0;
error = getinoquota(ip);
relock:
if (relocked)
vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
if (error != 0)
return (error);
if (VOP_ISLOCKED(vp) == LK_EXCLUSIVE) {
error = getinoquota(ip);
if (error != 0)
return (error);
}
#endif
break;
default: