Author: rmacklem
Date: Thu Jul 11 01:30:24 2013
New Revision: 253177
URL: http://svnweb.freebsd.org/changeset/base/253177
Log:
MFC: r252673
A problem with the old NFS client where large writes to large files
would sometimes result in a corrupted file was reported via email.
This problem appears to have been caused by r251719 (reverting
r251719 fixed the problem). Although I have not been able to
reproduce this problem, I suspect it is caused by another thread
increasing np->n_size after the mtx_unlock(&np->n_mtx) but before
the vnode_pager_setsize() call. Since the np->n_mtx mutex serializes
updates to np->n_size, doing the vnode_pager_setsize() with the
mutex locked appears to avoid the problem.
Unfortunately, vnode_pager_setsize() where the new size is smaller,
cannot be called with a mutex held.
This patch returns the semantics to be close to pre-r251719 such that the
call to the vnode_pager_setsize() is only delayed until after the mutex is
unlocked when np->n_size is shrinking. Since the file is growing
when being written, I believe this will fix the corruption.
Tested by: remy.nonnenmacher at activnetworks.com
Modified:
stable/9/sys/nfsclient/nfs_subs.c
Directory Properties:
stable/9/sys/ (props changed)
Modified: stable/9/sys/nfsclient/nfs_subs.c
==============================================================================
--- stable/9/sys/nfsclient/nfs_subs.c Wed Jul 10 23:14:24 2013 (r253176)
+++ stable/9/sys/nfsclient/nfs_subs.c Thu Jul 11 01:30:24 2013 (r253177)
@@ -595,6 +595,7 @@ nfs_loadattrcache(struct vnode **vpp, st
vap->va_size = np->n_size;
np->n_attrstamp = 0;
KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
+ vnode_pager_setsize(vp, np->n_size);
} else if (np->n_flag & NMODIFIED) {
/*
* We've modified the file: Use the larger
@@ -606,12 +607,22 @@ nfs_loadattrcache(struct vnode **vpp, st
np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
}
+ vnode_pager_setsize(vp, np->n_size);
+ } 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 {
np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
+ vnode_pager_setsize(vp, np->n_size);
}
- setnsize = 1;
- nsize = vap->va_size;
} else {
np->n_size = vap->va_size;
}