Author: kib
Date: Tue Sep 17 18:41:39 2019
New Revision: 352457
URL: https://svnweb.freebsd.org/changeset/base/352457
Log:
Further refine r352393, only call vnode_pager_setsize() outside the
node lock when shrinking.
This is similar to r252528, applied to the above commit.
Apparently there is a race which makes necessary at least to keep the
n_size and pager size consistent when extending. Current suspect is
that iod threads perform vnode_pager_setsize() without taking the
vnode lock, which corrupts the file content.
Reported and tested by: Masachika ISHIZUKA <i...@amail.plala.or.jp>
Discussed with: rmacklem (related issues)
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Modified:
head/sys/fs/nfsclient/nfs_clport.c
Modified: head/sys/fs/nfsclient/nfs_clport.c
==============================================================================
--- head/sys/fs/nfsclient/nfs_clport.c Tue Sep 17 18:36:29 2019
(r352456)
+++ head/sys/fs/nfsclient/nfs_clport.c Tue Sep 17 18:41:39 2019
(r352457)
@@ -414,12 +414,12 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvatt
struct nfsnode *np;
struct nfsmount *nmp;
struct timespec mtime_save;
+ vm_object_t object;
u_quad_t nsize;
- int setnsize, error, force_fid_err;
+ int error, force_fid_err;
+ bool setnsize;
error = 0;
- setnsize = 0;
- nsize = 0;
/*
* If v_type == VNON it is a new node, so fill in the v_type,
@@ -511,8 +511,7 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvatt
* zero np->n_attrstamp to indicate that
* the attributes are stale.
*/
- nsize = vap->va_size = np->n_size;
- setnsize = 1;
+ vap->va_size = np->n_size;
np->n_attrstamp = 0;
KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
} else if (np->n_flag & NMODIFIED) {
@@ -526,22 +525,9 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvatt
np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
}
- nsize = np->n_size;
- setnsize = 1;
- } else if (vap->va_size < np->n_size) {
- /*
- * When shrinking the size, the call to
- * vnode_pager_setsize() cannot be done
- * with the mutex held, so delay it until
- * after the mtx_unlock call.
- */
- nsize = np->n_size = vap->va_size;
- np->n_flag |= NSIZECHANGED;
- setnsize = 1;
} else {
- nsize = np->n_size = vap->va_size;
+ np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
- setnsize = 1;
}
} else {
np->n_size = vap->va_size;
@@ -579,6 +565,23 @@ out:
if (np->n_attrstamp != 0)
KDTRACE_NFS_ATTRCACHE_LOAD_DONE(vp, vap, error);
#endif
+ nsize = vap->va_size;
+ object = vp->v_object;
+ setnsize = false;
+ if (object != NULL) {
+ if (OFF_TO_IDX(nsize + PAGE_MASK) < object->size) {
+ /*
+ * When shrinking the size, the call to
+ * vnode_pager_setsize() cannot be done with
+ * the mutex held, because we might need to
+ * wait for a busy page. Delay it until after
+ * the node is unlocked.
+ */
+ setnsize = true;
+ } else {
+ vnode_pager_setsize(vp, nsize);
+ }
+ }
NFSUNLOCKNODE(np);
if (setnsize)
vnode_pager_setsize(vp, nsize);
_______________________________________________
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
