[PATCH 3/3] staging: qlge: add comment explaining memory barrier

From: drake
Date: Mon Oct 31 2022 - 10:33:10 EST


From: Drake Talley <drake@xxxxxxxxxxxxxxx>

codestyle change that fixes the following report from checkpatch:

> WARNING: memory barrier without comment
> #2101: FILE: drivers/staging/qlge/qlge_main.c:2101:

The added comment identifies the next item from the circular
buffer (rx_ring->curr_entry) and its handling/unmapping as the two
operations that must not be reordered. Based on the kernel
documentation for memory barriers in circular buffers
(https://www.kernel.org/doc/Documentation/circular-buffers.txt) and
the presence of atomic operations in the current context I'm assuming
this usage of the memory barrier is akin to what is explained in the
linked doc.

There are a couple of other uncommented usages of memory barriers in
the current file. If this comment is adequate I can add similar
comments to the others.

Signed-off-by: Drake Talley <drake@xxxxxxxxxxxxxxx>
---
drivers/staging/qlge/qlge_main.c | 6 ++++++
1 file changed, 6 insertions(+)

diff --git a/drivers/staging/qlge/qlge_main.c b/drivers/staging/qlge/qlge_main.c
index c8403dbb5bad..f70390bce6d8 100644
--- a/drivers/staging/qlge/qlge_main.c
+++ b/drivers/staging/qlge/qlge_main.c
@@ -2098,6 +2098,12 @@ static int qlge_clean_outbound_rx_ring(struct rx_ring *rx_ring)
rx_ring->cq_id, prod, rx_ring->cnsmr_idx);

net_rsp = (struct qlge_ob_mac_iocb_rsp *)rx_ring->curr_entry;
+ /*
+ * Ensure that the next item from the ring buffer is loaded
+ * before being processed.
+ * Adding rmb() prevents the compiler from reordering the read
+ * and subsequent handling of the outbound completion pointer.
+ */
rmb();
switch (net_rsp->opcode) {
case OPCODE_OB_MAC_TSO_IOCB:
--
2.34.1