Skip to contentSkip to conversation

FEC: Rework RX buffer init

PATCH New Marek Vasut 2 participants 2 messages Oct 9, 2012 – Oct 12, 2012 Network drivers

lore
Share & tools
lore

Marek Vasut

author
#0 lore raw

PATCH 1/1 FEC: Rework RX buffer init

Rework the RX buffer init so that it's easier to understand. Firstly, allocate the whole RX buffer as one large continuous piece of memory. Also, drop all these writel() accessors used on the FEC RX DMA descriptor, since it's all flat bogus.

Finally, this makes recoverable stalls of the FEC on i.MX28 disappear completely.

Also, it removes whole 80 bytes from the size of u-boot! (Tested with Debian GCC 4.7.1-3)

Signed-off-by Marek Vasut <[email protected]>
Cc Fabio Estevam <[email protected]>
Cc Joe Hershberger <[email protected]>
Cc Otavio Salvador <[email protected]>
Cc Stefano Babic <[email protected]>
---
 drivers/net/fec_mxc.c |   47 ++++++++++++++++++++---------------------------
 1 file changed, 20 insertions(+), 27 deletions(-)
drivers/net/fec_mxc.c +20 -27
index 3e232c7..1897397 100644
@@ -289,43 +289,36 @@ static int fec_tx_task_disable(struct fec_priv *fec)
  */
 static int fec_rbd_init(struct fec_priv *fec, int count, int dsize)
 {
+	struct fec_bd *rbd = fec->rbd_base;
 	uint32_t size;
+	void *mem;
 	int i;
 
-	/*
-	 * Allocate memory for the buffers. This allocation respects the
-	 * alignment
-	 */
-	size = roundup(dsize, ARCH_DMA_MINALIGN);
+	if (!rbd->data_pointer) {
+		/*
+		 * Allocate memory for the buffers. This allocation
+		 * respects the alignment.
+		 */
+		size = roundup(dsize, ARCH_DMA_MINALIGN);
+		mem = memalign(ARCH_DMA_MINALIGN, size * count);
+		if (!mem)
+			return -ENOMEM;
+
+		for (i = 0; i < count; i++)
+			rbd[i].data_pointer = (uint32_t)(mem + size * i);
+	}
+
 	for (i = 0; i < count; i++) {
-		uint32_t data_ptr = readl(&fec->rbd_base[i].data_pointer);
-		if (data_ptr == 0) {
-			uint8_t *data = memalign(ARCH_DMA_MINALIGN,
-						 size);
-			if (!data) {
-				printf("%s: error allocating rxbuf %d\n",
-				       __func__, i);
-				goto err;
-			}
-			writel((uint32_t)data, &fec->rbd_base[i].data_pointer);
-		} /* needs allocation */
-		writew(FEC_RBD_EMPTY, &fec->rbd_base[i].status);
-		writew(0, &fec->rbd_base[i].data_length);
+		fec->rbd_base[i].status = FEC_RBD_EMPTY;
+		fec->rbd_base[i].data_length = 0;
 	}
 
 	/* Mark the last RBD to close the ring. */
-	writew(FEC_RBD_WRAP | FEC_RBD_EMPTY, &fec->rbd_base[i - 1].status);
+	fec->rbd_base[count - 1].status |= FEC_RBD_WRAP;
+
 	fec->rbd_index = 0;
 
 	return 0;
-
-err:
-	for (; i >= 0; i--) {
-		uint32_t data_ptr = readl(&fec->rbd_base[i].data_pointer);
-		free((void *)data_ptr);
-	}
-
-	return -ENOMEM;
 }
 
 /**
Signature · 1 line
1.7.10.4
#1 lore raw

PATCH 1/1 FEC: Rework RX buffer init

Hi Marek,

Quoted 4 lines · Marek Vasut

Rework the RX buffer init so that it's easier to understand. Firstly, allocate the whole RX buffer as one large continuous piece of memory. Also, drop all these writel() accessors used on the FEC RX DMA descriptor, since it's all flat bogus.

I'd like this commit message to be more usefully descriptive than "all flat bogus". What exactly is bogus in it?

Quoted 2 lines

Finally, this makes recoverable stalls of the FEC on i.MX28 disappear completely.

Are both the single allocation change and the writel() change needed for this?

Amicalement,

Signature · 1 line
Albert.