svcrdma: handle rdma read with a non-zero initial page offset

commit c91aed9 upstream.

The server rdma_read_chunk_lcl() and rdma_read_chunk_frmr() functions
were not taking into account the initial page_offset when determining
the rdma read length.  This resulted in a read who's starting address
and length exceeded the base/bounds of the frmr.

The server gets an async error from the rdma device and kills the
connection, and the client then reconnects and resends.  This repeats
indefinitely, and the application hangs.

Most work loads don't tickle this bug apparently, but one test hit it
every time: building the linux kernel on a 16 core node with 'make -j
16 O=/mnt/0' where /mnt/0 is a ramdisk mounted via NFSRDMA.

This bug seems to only be tripped with devices having small fastreg page
list depths.  I didn't see it with mlx4, for instance.

Fixes: 0bf4828 ('svcrdma: refactor marshalling logic')
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/sunrpc/xprtrdma/svc_rdma_recvfrom.c

@@ -146,7 +146,8 @@ int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt,
 	ctxt->read_hdr = head;
 	pages_needed =
 		min_t(int, pages_needed, rdma_read_max_sge(xprt, pages_needed));
-	read = min_t(int, pages_needed << PAGE_SHIFT, rs_length);
+	read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset,
+		     rs_length);
 
 	for (pno = 0; pno < pages_needed; pno++) {
 		int len = min_t(int, rs_length, PAGE_SIZE - pg_off);
@@ -245,7 +246,8 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt,
 	ctxt->direction = DMA_FROM_DEVICE;
 	ctxt->frmr = frmr;
 	pages_needed = min_t(int, pages_needed, xprt->sc_frmr_pg_list_len);
-	read = min_t(int, pages_needed << PAGE_SHIFT, rs_length);
+	read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset,
+		     rs_length);
 
 	frmr->kva = page_address(rqstp->rq_arg.pages[pg_no]);
 	frmr->direction = DMA_FROM_DEVICE;