Re: [PATCH] net/imx_fec: Updating the IMX_FEC IP to support loopback mode.

[Jason Wang]

On 2019/11/30 上午12:04, Philippe Mathieu-Daudé wrote:

On Fri, Nov 29, 2019 at 4:59 PM Wasim, Bilal <bilal_wa...@mentor.com> wrote:

Thanks for the pointers philippe.. Is the patch okay to be merged without it or 
do I need to do a re-submission with the updated username ?

If there are no review comments on your patch, I think the maintainer
taking your patch can fix this details directly, no need to resend.

-----Original Message-----
From: Philippe Mathieu-Daudé [mailto:phi...@redhat.com]
Sent: Friday, November 29, 2019 8:38 PM
To: bilalwasim...@gmail.com; qemu-devel@nongnu.org
Cc: peter.mayd...@linaro.org; aa1ron...@gmail.com; j...@tribudubois.net; 
qemu-...@nongnu.org; Wasim, Bilal <bilal_wa...@mentor.com>; li...@roeck-us.net; Jason 
Wang <jasow...@redhat.com>
Subject: Re: [PATCH] net/imx_fec: Updating the IMX_FEC IP to support loopback 
mode.

Hi Bilal,

Cc'ing Jason, the maintainer of network devices.

On 11/29/19 4:05 PM, bilalwasim...@gmail.com wrote:

From: bwasim <bilal_wa...@mentor.com>

Your git setup misses your 'user.name', you could fix it running:

    git config user.name "Bilal Wasim"

(eventually with the --global option).

The patch looks good otherwise.

Thanks!

Applied with the fix for user.name.

Thanks

Loopback mode only works when specific conditions (as dictated by the
IP guide) are met, i.e. the MII_MODE is set and the RMII_MODE is
cleared. If not, we simply send the packet on the output queue (for TX
to the host network). Tested by running a custom RTOS and TXing a ton
of packets. The same packets were received on the RX side..

Signed-off-by: Bilal Wasim <bilal_wa...@mentor.com>
---
   hw/net/imx_fec.c | 27 +++++++++++++++++++++++++--
   1 file changed, 25 insertions(+), 2 deletions(-)

diff --git a/hw/net/imx_fec.c b/hw/net/imx_fec.c index
bd99236864..c51e7f7363 100644
--- a/hw/net/imx_fec.c
+++ b/hw/net/imx_fec.c
@@ -256,6 +256,29 @@ static const VMStateDescription vmstate_imx_eth =
{

   static void imx_eth_update(IMXFECState *s);

+/*
+ * Function to check if the MAC is configured to run in loopback mode.
+ * If so, invoke the "receive" routine.
+ * Else write to the output.
+ * */
+static void send_pkt(IMXFECState *s, uint8_t *frame, int frame_size)
+{
+    NetClientState *nc = qemu_get_queue(s->nic);
+
+    /*
+     * Loopback or Normal mode ?
+     * Per the FEC Manual: If loopback is enabled, the MII_MODE
+     * should be SET and the RMII_MODE should be cleared. Loopback
+     * will only work if this criterion is met. If not met,
+     * we will send the frame on the output queue. */
+    if ((s->regs[ENET_RCR] & ENET_RCR_LOOP) && (s->regs[ENET_RCR] & 
ENET_RCR_MII_MODE)
+            && !(s->regs[ENET_RCR] & ENET_RCR_RMII_MODE)) {
+        nc->info->receive(nc, frame, frame_size);
+    } else {
+        qemu_send_packet(nc, frame, frame_size);
+    }
+}
+
   /*
    * The MII phy could raise a GPIO to the processor which in turn
    * could be handled as an interrpt by the OS.
@@ -488,7 +511,7 @@ static void imx_fec_do_tx(IMXFECState *s)
           frame_size += len;
           if (bd.flags & ENET_BD_L) {
               /* Last buffer in frame.  */
-            qemu_send_packet(qemu_get_queue(s->nic), s->frame, frame_size);
+            send_pkt(s, (uint8_t *)&s->frame, frame_size);
               ptr = s->frame;
               frame_size = 0;
               s->regs[ENET_EIR] |= ENET_INT_TXF; @@ -586,7 +609,7 @@
static void imx_enet_do_tx(IMXFECState *s, uint32_t index)
               }
               /* Last buffer in frame.  */

-            qemu_send_packet(qemu_get_queue(s->nic), s->frame, frame_size);
+            send_pkt(s, (uint8_t *)&s->frame, frame_size);
               ptr = s->frame;

               frame_size = 0;