From: Andy Adamson <andros@xxxxxxxxxx>
Signed-off-by: Andy Adamson <andros@xxxxxxxxxx>
---
fs/nfs/pnfs.c | 78 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1 files changed, 78 insertions(+), 0 deletions(-)
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 924dce1..1db7416 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -934,6 +934,80 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo,
}
/*
+ * Use mdsthreshold hints set at each OPEN to determine if I/O should go
+ * to the MDS or over pNFS
+ *
+ * The nfs_inode read_io and write_io fields are cumulative counters reset
+ * before each LAYOUTGET. Note that in pnfs_update_layout iomode is set to
+ * IOMODE_READ for a READ request, and set to IOMODE_RW for a WRITE request.
+ * A return of true means use MDS I/O.
+ *
+ * From rfc 5661:
+ * If a file's length is less than the value of threshold4 read or write size,
+ * then it is RECOMMENDED that the client read from the file via the MDS and
+ * not a storage device.
+ *
+ * For I/O sizes below the threshold4 read or write iosize, it is RECOMMENDED
+ * to access data through the MDS.
+ */
+static bool pnfs_within_mdsthreshold(struct nfs_open_context *ctx,
+ struct inode *ino, int iomode)
+{
+ struct nfs4_threshold *t = ctx->mdsthreshold;
+ struct nfs_inode *nfsi = NFS_I(ino);
+ loff_t fsize = i_size_read(ino);
+ bool size = false, size_set = false, io = false, io_set = false;
+ bool ret = false;
+
+ if (t == NULL || t->bm == 0)
+ return ret;
+
+ dprintk("%s bm=0x%x rd_sz=%llu wr_sz=%llu rd_io=%llu wr_io=%llu\n",
+ __func__, t->bm, t->rd_sz, t->wr_sz, t->rd_io_sz, t->wr_io_sz);
+
+ switch (iomode) {
+ case IOMODE_READ:
+ if (t->bm & THRESHOLD_RD) {
+ dprintk("%s fsize %llu\n", __func__, fsize);
+ size_set = true;
+ if (fsize < t->rd_sz)
+ size = true;
+ }
+ if (t->bm & THRESHOLD_RD_IO) {
+ dprintk("%s nfsi->read_io %llu\n", __func__,
+ nfsi->read_io);
+ io_set = true;
+ if (nfsi->read_io < t->rd_io_sz)
+ io = true;
+ }
+ break;
+ case IOMODE_RW:
+ if (t->bm & THRESHOLD_WR) {
+ dprintk("%s fsize %llu\n", __func__, fsize);
+ size_set = true;
+ if (fsize < t->wr_sz)
+ size = true;
+ }
+ if (t->bm & THRESHOLD_WR_IO) {
+ dprintk("%s nfsi->write_io %llu\n", __func__,
+ nfsi->write_io);
+ io_set = true;
+ if (nfsi->write_io < t->wr_io_sz)
+ io = true;
+ }
+ break;
+ }
+ if (size_set && io_set) {
+ if (size && io)
+ ret = true;
+ } else if (size || io)
+ ret = true;
+
+ dprintk("<-- %s size %d io %d ret %d\n", __func__, size, io, ret);
+ return ret;
+}
+
+/*
* Layout segment is retreived from the server if not cached.
* The appropriate layout segment is referenced and returned to the caller.
*/
@@ -960,6 +1034,10 @@ pnfs_update_layout(struct inode *ino,
if (!pnfs_enabled_sb(NFS_SERVER(ino)))
return NULL;
+
+ if (pnfs_within_mdsthreshold(ctx, ino, iomode))
+ return NULL;
+
spin_lock(&ino->i_lock);
lo = pnfs_find_alloc_layout(ino, ctx, gfp_flags);
if (lo == NULL) {
--
1.7.7.6
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
