SAMV7 Ethernet: Fix some errors in circular queue handling

This commit is contained in:
Gregory Nutt 2015-03-27 13:04:43 -06:00
parent dd33bb14ac
commit 750fad37f0

View File

@ -1010,7 +1010,7 @@ static uint16_t sam_txfree(struct sam_emac_s *priv, int qid)
* the configured size minus 1.
*/
return (priv->xfrq[qid].ntxbuffers-1) - sam_txinuse(priv, qid);
return (priv->xfrq[qid].ntxbuffers - 1) - sam_txinuse(priv, qid);
}
/****************************************************************************
@ -1321,7 +1321,7 @@ static int sam_transmit(struct sam_emac_s *priv, int qid)
/* Update TX descriptor status. */
status = dev->d_len | EMACTXD_STA_LAST;
if (txhead == priv->xfrq[qid].ntxbuffers-1)
if (txhead == priv->xfrq[qid].ntxbuffers - 1)
{
status |= EMACTXD_STA_WRAP;
}
@ -1556,7 +1556,9 @@ static int sam_recvframe(struct sam_emac_s *priv, int qid)
if ((rxdesc->status & EMACRXD_STA_SOF) != 0)
{
/* Skip previous fragments */
/* Skip previous fragments. Loop incrementing the index to the
* start fragment until it is equal to the index with the SOF mark
*/
while (rxndx != priv->xfrq[qid].rxndx)
{
@ -1571,7 +1573,7 @@ static int sam_recvframe(struct sam_emac_s *priv, int qid)
(uintptr_t)rxdesc +
sizeof(struct emac_rxdesc_s));
/* Increment the RX index */
/* Increment the RX index to the start fragment */
if (++priv->xfrq[qid].rxndx >= priv->xfrq[qid].nrxbuffers)
{
@ -1589,7 +1591,10 @@ static int sam_recvframe(struct sam_emac_s *priv, int qid)
isframe = true;
}
/* Increment the working index */
/* Increment the working index. rxndx points to the next fragment
* in this frame (or, if the LAST bit was set, to the first fragment
* of the next frame).
*/
if (++rxndx >= priv->xfrq[qid].nrxbuffers)
{
@ -1664,7 +1669,8 @@ static int sam_recvframe(struct sam_emac_s *priv, int qid)
priv->xfrq[qid].rxndx, rxndx, dev->d_len);
/* All data have been copied in the application frame buffer,
* release the RX descriptor
* release the RX descriptor(s). Loop until all descriptors
* have been released up to the start of the next frame.
*/
while (priv->xfrq[qid].rxndx != rxndx)
@ -1680,7 +1686,9 @@ static int sam_recvframe(struct sam_emac_s *priv, int qid)
(uintptr_t)rxdesc +
sizeof(struct emac_rxdesc_s));
/* Increment the RX index */
/* Increment the RX index of the descriptor that was just
* released.
*/
if (++priv->xfrq[qid].rxndx >= priv->xfrq[qid].nrxbuffers)
{
@ -1705,8 +1713,9 @@ static int sam_recvframe(struct sam_emac_s *priv, int qid)
}
}
/* We have not encountered the SOF yet... discard this fragment and
* keep looking
/* We found a descriptor that we own, but we have not encountered the
* SOF yet... discard this fragment and keep looking starting at the
* next fragment.
*/
else
@ -1720,10 +1729,17 @@ static int sam_recvframe(struct sam_emac_s *priv, int qid)
arch_clean_dcache((uintptr_t)rxdesc,
(uintptr_t)rxdesc +
sizeof(struct emac_rxdesc_s));
/* rxndx already points to the next fragment to be examined.
* Use it to update the candidate Start-of-Frame.
*/
priv->xfrq[qid].rxndx = rxndx;
}
/* Process the next buffer */
/* Set-up rocess the next fragment. Get the RX descriptor
* associated with the next fragment.
*/
rxdesc = &priv->xfrq[qid].rxdesc[rxndx];
@ -1932,7 +1948,7 @@ static void sam_txdone(struct sam_emac_s *priv, int qid)
txdesc = &priv->xfrq[qid].txdesc[tail];
/* REVISIT: If the rxdesc is not aligned with the cacheline boundary
/* REVISIT: If the txdesc is not aligned with the cacheline boundary
* then won't this also invalidate some surrounding memory?
*/
@ -1959,7 +1975,7 @@ static void sam_txdone(struct sam_emac_s *priv, int qid)
{
/* Increment the tail index */
if (++tail > priv->xfrq[qid].ntxbuffers)
if (++tail >= priv->xfrq[qid].ntxbuffers)
{
tail = 0;
}
@ -1974,7 +1990,7 @@ static void sam_txdone(struct sam_emac_s *priv, int qid)
/* Go to first buffer of the next frame */
if (tail != priv->xfrq[qid].txhead &&
++tail > priv->xfrq[qid].ntxbuffers)
++tail >= priv->xfrq[qid].ntxbuffers)
{
tail = 0;
}
@ -2070,7 +2086,7 @@ static void sam_txerr_interrupt(FAR struct sam_emac_s *priv, int qid)
{
/* Increment the tail index */
if (++tail > priv->xfrq[qid].ntxbuffers)
if (++tail >= priv->xfrq[qid].ntxbuffers)
{
tail = 0;
}
@ -2085,7 +2101,7 @@ static void sam_txerr_interrupt(FAR struct sam_emac_s *priv, int qid)
/* Go to first buffer of the next frame */
if (tail != priv->xfrq[qid].txhead &&
++tail > priv->xfrq[qid].ntxbuffers)
++tail >= priv->xfrq[qid].ntxbuffers)
{
tail = 0;
}
@ -5073,8 +5089,8 @@ errout:
* netdev ioctl. The application level code have gotten the MAC
* address from some configuration parameter or by accessing some
* non-volatile storage containing the address. This is the
* "cannonically correct" way to set the MAC address.
* 2) Alterntively, the board logic may support some other less obvious
* "canonically correct" way to set the MAC address.
* 2) Alternatively, the board logic may support some other less obvious
* non-volatile storage and the board-level boot-up code may access
* this and use this interface to set the Ethernet MAC address more
* directly. This is mostly a kludge for the case where you just don't