Re: [PATCH v2 net-next 3/6] net: bcmasp: Add support for ASP2.0 Ethernet controller

On Wed, Apr 26, 2023 at 11:54:29AM -0700, Justin Chen wrote:

Add support for the Broadcom ASP 2.0 Ethernet controller which is first
introduced with 72165. This controller features two distinct Ethernet
ports that can be independently operated.

This patch supports:

- Wake-on-LAN using magic packets
- basic ethtool operations (link, counters, message level)
- MAC destination address filtering (promiscuous, ALL_MULTI, etc.)

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Justin Chen <redacted>

...

...

+static int bcmasp_netfilt_get_reg_offset(struct bcmasp_priv *priv,
+					 struct bcmasp_net_filter *nfilt,
+					 enum asp_netfilt_reg_type reg_type,
+					 u32 offset)
+{
+	u32 block_index, filter_sel;
+
+	if (offset < 32) {
+		block_index = ASP_RX_FILTER_NET_L2;
+		filter_sel = nfilt->hw_index;
+	} else if (offset < 64) {
+		block_index = ASP_RX_FILTER_NET_L2;
+		filter_sel = nfilt->hw_index + 1;
+	} else if (offset < 96) {
+		block_index = ASP_RX_FILTER_NET_L3_0;
+		filter_sel = nfilt->hw_index;
+	} else if (offset < 128) {
+		block_index = ASP_RX_FILTER_NET_L3_0;
+		filter_sel = nfilt->hw_index + 1;
+	} else if (offset < 160) {
+		block_index = ASP_RX_FILTER_NET_L3_1;
+		filter_sel = nfilt->hw_index;
+	} else if (offset < 192) {
+		block_index = ASP_RX_FILTER_NET_L3_1;
+		filter_sel = nfilt->hw_index + 1;
+	} else if (offset < 224) {
+		block_index = ASP_RX_FILTER_NET_L4;
+		filter_sel = nfilt->hw_index;
+	} else if (offset < 256) {
+		block_index = ASP_RX_FILTER_NET_L4;
+		filter_sel = nfilt->hw_index + 1;
+	}

block_index and filter_sel are uninitialised if offset doesn't match any
of the conditions above. Can that happen?

+
+	switch (reg_type) {
+	case ASP_NETFILT_MATCH:
+		return ASP_RX_FILTER_NET_PAT(filter_sel, block_index,
+					     (offset % 32));
+	case ASP_NETFILT_MASK:
+		return ASP_RX_FILTER_NET_MASK(filter_sel, block_index,
+					      (offset % 32));
+	default:
+		return -EINVAL;
+	}
+}

...

+static void bcmasp_netfilt_tcpip4_wr(struct bcmasp_priv *priv,
+				     struct bcmasp_net_filter *nfilt,
+				     struct ethtool_tcpip4_spec *match,
+				     struct ethtool_tcpip4_spec *mask,
+				     u32 offset)
+{
+	__be16 val_16, mask_16;
+
+	val_16 = htons(ETH_P_IP);
+	mask_16 = 0xFFFF;

mask_17 is __be16, but 0xFFFF is host byte order.

Please make sure there are no new warnings when building with W=1 C=1.

...

+/* If no network filter found, return open filter.
+ * If no more open filters return NULL
+ */
+struct bcmasp_net_filter *bcmasp_netfilt_get_init(struct bcmasp_intf *intf,
+						  int loc, bool wake_filter,
+						  bool init)
+{
+	struct bcmasp_priv *priv = intf->parent;
+	struct bcmasp_net_filter *nfilter = NULL;
+	int i, open_index = -1;

Please use reverse xmas tree - longest line to shortest - for local
variable declarations in networking code.

You can check for this using https://github.com/ecree-solarflare/xmastree

...

+static int bcmasp_combine_set_filter(struct bcmasp_intf *intf,
+				     unsigned char *addr, unsigned char *mask,
+				     int i)
+{
+	u64 addr1, addr2, mask1, mask2, mask3;
+	struct bcmasp_priv *priv = intf->parent;
+
+	/* Switch to u64 to help with the calculations */
+	addr1 = ether_addr_to_u64(priv->mda_filters[i].addr);
+	mask1 = ether_addr_to_u64(priv->mda_filters[i].mask);
+	addr2 = ether_addr_to_u64(addr);
+	mask2 = ether_addr_to_u64(mask);
+
+	/* Check if one filter resides within the other */
+	mask3 = mask1 & mask2;
+	if (mask3 == mask1 && ((addr1 & mask1) == (addr2 & mask1))) {
+		/* Filter 2 resides within fitler 1, so everthing is good */

nit: s/fitler/filter/

Please consider running ./scripts/checkpatch.pl --codespell

...

+static void bcmasp_update_mib_counters(struct bcmasp_intf *priv)
+{
+	int i, j = 0;
+
+	for (i = 0; i < BCMASP_STATS_LEN; i++) {
+		const struct bcmasp_stats *s;
+		u16 offset = 0;
+		u32 val = 0;
+		char *p;
+
+		s = &bcmasp_gstrings_stats[i];
+		switch (s->type) {
+		case BCMASP_STAT_NETDEV:
+		case BCMASP_STAT_SOFT:
+			continue;
+		case BCMASP_STAT_RUNT:
+			offset += BCMASP_STAT_OFFSET;
+			fallthrough;
+		case BCMASP_STAT_MIB_TX:
+			offset += BCMASP_STAT_OFFSET;
+			fallthrough;
+		case BCMASP_STAT_MIB_RX:
+			val = umac_rl(priv, UMC_MIB_START + j + offset);
+			offset = 0;	/* Reset Offset */
+			break;
+		case BCMASP_STAT_RX_EDPKT:
+			val = rx_edpkt_core_rl(priv->parent, s->reg_offset);
+			break;
+		case BCMASP_STAT_RX_CTRL:
+			offset = bcmasp_stat_fixup_offset(priv, s);
+			if (offset != ASP_RX_CTRL_FB_FILT_OUT_FRAME_COUNT)
+				offset += sizeof(u32) * priv->port;
+			val = rx_ctrl_core_rl(priv->parent, offset);
+			break;
+		}
+
+		j += s->stat_sizeof;
+		p = (char *)priv + s->stat_offset;
+		*(u32 *)p = val;

Is p always 32bit aligned?

+	}
+}
+
+static void bcmasp_get_ethtool_stats(struct net_device *dev,
+				     struct ethtool_stats *stats,
+				     u64 *data)
+{
+	struct bcmasp_intf *priv = netdev_priv(dev);
+	int i, j = 0;
+
+	if (netif_running(dev))
+		bcmasp_update_mib_counters(priv);
+
+	dev->netdev_ops->ndo_get_stats(dev);
+
+	for (i = 0; i < BCMASP_STATS_LEN; i++) {
+		const struct bcmasp_stats *s;
+		char *p;
+
+		s = &bcmasp_gstrings_stats[i];
+		if (!bcmasp_stat_available(priv, s->type))
+			continue;
+		if (s->type == BCMASP_STAT_NETDEV)
+			p = (char *)&dev->stats;
+		else
+			p = (char *)priv;
+		p += s->stat_offset;
+		if (sizeof(unsigned long) != sizeof(u32) &&
+		    s->stat_sizeof == sizeof(unsigned long))
+			data[j] = *(unsigned long *)p;
+		else
+			data[j] = *(u32 *)p;

Maybe memcpy would make this a little easier to read.

+		j++;
+	}
+}

...

...

+static int bcmasp_init_rx(struct bcmasp_intf *intf)
+{
+	struct device *kdev = &intf->parent->pdev->dev;
+	struct net_device *ndev = intf->ndev;
+	void *p;
+	dma_addr_t dma;
+	struct page *buffer_pg;
+	u32 reg;
+	int ret;
+
+	intf->rx_buf_order = get_order(RING_BUFFER_SIZE);
+	buffer_pg = alloc_pages(GFP_KERNEL, intf->rx_buf_order);
+
+	dma = dma_map_page(kdev, buffer_pg, 0, RING_BUFFER_SIZE,
+			   DMA_FROM_DEVICE);
+	if (dma_mapping_error(kdev, dma)) {
+		netdev_err(ndev, "Cannot allocate RX buffer\n");

I think the core will log an error on allocation failure,
so the message above is not needed.

+		__free_pages(buffer_pg, intf->rx_buf_order);
+		return -ENOMEM;
+	}

...