Date: Tue, 17 Sep 2019 17:44:51 +0000 (UTC) From: Kirk McKusick <mckusick@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r352453 - head/sys/kern Message-ID: <201909171744.x8HHipJ4019776@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mckusick Date: Tue Sep 17 17:44:50 2019 New Revision: 352453 URL: https://svnweb.freebsd.org/changeset/base/352453 Log: The VFS-level clustering code collects together sequential blocks by issuing delayed-writes (bdwrite()) until a non-sequential block is written or the maximum cluster size is reached. At that point it collects the delayed buffers together (using bread()) to write them in a single operation. The assumption was that since we just looked at them they will still be in memory so there is no need to check for a read error from bread(). Very occationally (apparently every 10-hours or so when being pounded by Peter Holm's tests) this assumption is wrong. The fix is to check for errors from bread() and fail the cluster write thus falling back to the default individual flushing of any still dirty buffers. Reported by: Peter Holm and Chuck Silvers Reviewed by: kib MFC after: 3 days Modified: head/sys/kern/vfs_cluster.c Modified: head/sys/kern/vfs_cluster.c ============================================================================== --- head/sys/kern/vfs_cluster.c Tue Sep 17 17:28:44 2019 (r352452) +++ head/sys/kern/vfs_cluster.c Tue Sep 17 17:44:50 2019 (r352453) @@ -718,6 +718,14 @@ cluster_write(struct vnode *vp, struct buf *bp, u_quad struct cluster_save *buflist; buflist = cluster_collectbufs(vp, bp, gbflags); + if (buflist == NULL) { + /* + * Cluster build failed so just write + * it now. + */ + bawrite(bp); + return; + } endbp = &buflist->bs_children [buflist->bs_nchildren - 1]; if (VOP_REALLOCBLKS(vp, buflist)) { @@ -1056,7 +1064,7 @@ cluster_collectbufs(struct vnode *vp, struct buf *last struct cluster_save *buflist; struct buf *bp; daddr_t lbn; - int i, len; + int i, j, len, error; len = vp->v_lastw - vp->v_cstart + 1; buflist = malloc(sizeof(struct buf *) * (len + 1) + sizeof(*buflist), @@ -1064,8 +1072,18 @@ cluster_collectbufs(struct vnode *vp, struct buf *last buflist->bs_nchildren = 0; buflist->bs_children = (struct buf **) (buflist + 1); for (lbn = vp->v_cstart, i = 0; i < len; lbn++, i++) { - (void)bread_gb(vp, lbn, last_bp->b_bcount, NOCRED, + error = bread_gb(vp, lbn, last_bp->b_bcount, NOCRED, gbflags, &bp); + if (error != 0) { + /* + * If read fails, release collected buffers + * and return failure. + */ + for (j = 0; j < i; j++) + brelse(buflist->bs_children[j]); + free(buflist, M_SEGMENT); + return (NULL); + } buflist->bs_children[i] = bp; if (bp->b_blkno == bp->b_lblkno) VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno,
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201909171744.x8HHipJ4019776>