| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Mon, 16 Jun 2014 21:41:49 -0700
From: Florian Fainelli <f.fainelli@...il.com>
To: Andy Green <andy.green@...aro.org>
Cc: netdev <netdev@...r.kernel.org>, Joe Perches <joe@...ches.com>,
patches@...aro.org, Francois Romieu <romieu@...zoreil.com>
Subject: Re: [net-next PATCH v3] net: ethernet driver: Fujitsu OGMA
Hi Andy,
2014-06-14 21:21 GMT-07:00 Andy Green <andy.green@...aro.org>:
> This driver adds support for "ogma", a Fujitsu Semiconductor Ltd IP Gigabit
> Ethernet + PHY IP used in a variety of their ARM-based ASICs.
>
> We are preparing to upstream the main platform support for these chips
> which is currently waiting for mailbox driver in v6 now and waiting for
> a final ACK
>
> https://lkml.org/lkml/2014/5/15/49
>
> This driver was originally written by guys inside Fujitsu as an abstracted
> "you can build this for Windows as well" type code, I've removed all that,
> modernized various things, added runtime_pm, and ported it to work with
> Device Tree, using only the bindings already mentioned in
>
> ./Documentation/devicetree/bindings/net/ethernet.txt
>
> There is only one checkpatch complaint, both about documentation for the
> DT vendor missing. But we will document vendor "fujitsu" in the main mach
> support patches. Bindings documentation is added by this patch.
>
> The patch is based on net-next f9da455b93f6ba076935 from today, and the unchanged
> patch has been tested on real hardware on an integration tree based on
> pre-3.16-rc1 from a couple of days ago.
>
> Allmodconfig doesn't generate any warnings (although on current net-next build
> is broken by a staging driver).
>
> Any comments about how to further improve and align it with current best-
> practice for upstream Ethernet drivers are appreciated.
>
> Changes since v2:
>
> - Followed comments from Florian Fainelli and Joe Perches
> - Use Phylib
> - Register as mii device
> - Fill out ethtool info
> - Use ethtool coalesce control struct
> - use netdev-related log apis
> - Many other cleanups and normalizations
This looks much better now, thanks! Here are some more minor comments.
net-next is not open for new things now, so that buys you some time to
address the changes.
>
> Changes since v1:
>
> - Followed comments from Francois Romieu about style issues and eliminate
> spinlock wrappers
>
> - Remove remaining excess ()
> - Pass checkpatch --strict now
> - Use netdev_alloc_skb_ip_align as suggested
> - Set hardware endian support according to cpu endianess
> - change error handling targets from "bailX" to "errX"
>
> Signed-off-by: Andy Green <andy.green@...aro.org>
> ---
[snip]
> +
> +struct ogma_desc_ring {
> + unsigned int id;
> + bool running;
> + u32 full:1;
> + u8 len;
> + u16 head;
> + u16 tail;
> + u16 rx_num;
> + u16 tx_done_num;
> + spinlock_t spinlock_desc; /* protect descriptor access */
> + void *ring_vaddr;
> + phys_addr_t desc_phys;
> + struct ogma_frag_info *frag;
> + struct sk_buff **priv;
> +};
You might be able to better align this structure for efficiency.
[snip]
> +static int alloc_rx_pkt_buf(struct ogma_priv *priv, struct ogma_frag_info *info,
> + void **addr, phys_addr_t *pa, struct sk_buff **skb)
> +{
> + *skb = netdev_alloc_skb_ip_align(priv->net_device, info->len);
> + if (!*skb)
> + return -ENOMEM;
> +
> + ogma_mark_skb_type(*skb, OGMA_RING_RX);
> + *addr = (*skb)->data;
> + *pa = dma_map_single(priv->dev, *addr, info->len, DMA_FROM_DEVICE);
dma_map_single() returns a dma_addr_t, both types might boil down to
being the same, but that might not be true for all platforms.
> + if (dma_mapping_error(priv->dev, *pa)) {
> + dev_kfree_skb(*skb);
> + return -ENOMEM;
> + }
> +
[snip]
> +static int ogma_set_irq_coalesce_param(struct ogma_priv *priv, unsigned int id)
> +{
> + int max_frames, tmr;
> +
> + switch (id) {
You could make id an enum, such that you make sure you pass the
correct type here.
[snip
> +
> +void ogma_mac_write(struct ogma_priv *priv, u32 addr, u32 value)
> +{
> + ogma_write_reg(priv, MAC_REG_DATA, value);
> + ogma_write_reg(priv, MAC_REG_CMD, addr | OGMA_GMAC_CMD_ST_WRITE);
> + ogma_wait_while_busy(priv, MAC_REG_CMD, OGMA_GMAC_CMD_ST_BUSY);
You should propagate the error from ogma_wait_while_busy() here.
> +}
> +
> +u32 ogma_mac_read(struct ogma_priv *priv, u32 addr)
> +{
> + ogma_write_reg(priv, MAC_REG_CMD, addr | OGMA_GMAC_CMD_ST_READ);
> + ogma_wait_while_busy(priv, MAC_REG_CMD, OGMA_GMAC_CMD_ST_BUSY);
Same here.
> +
> + return ogma_read_reg(priv, MAC_REG_DATA);
> +}
> +
> +static int ogma_mac_wait_while_busy(struct ogma_priv *priv, u32 addr, u32 mask)
> +{
> + u32 timeout = TIMEOUT_SPINS_MAC;
> +
> + while (--timeout && ogma_mac_read(priv, addr) & mask)
> + ;
> + if (!timeout) {
> + netdev_WARN(priv->net_device, "%s: timeout\n", __func__);
> + return -ETIME;
-ETIMEDOUT is a better error code here.
[snip]
> + value = priv->gmac_mode.half_duplex_flag ?
> + OGMA_GMAC_MCR_REG_HALF_DUPLEX_COMMON :
> + OGMA_GMAC_MCR_REG_FULL_DUPLEX_COMMON;
This should probably be done in the PHY library adjust_link() callback.
> +
> + if (priv->gmac_mode.link_speed != SPEED_1000)
> + value |= OGMA_GMAC_MCR_PS;
> +
> + if ((priv->phy_interface != PHY_INTERFACE_MODE_GMII) &&
> + (priv->gmac_mode.link_speed == SPEED_100))
> + value |= OGMA_GMAC_MCR_REG_FES;
And this here too.
[snip]
> +
> +static int ogma_phy_write(struct mii_bus *bus, int phy_addr, int reg, u16 val)
> +{
> + struct ogma_priv *priv = bus->priv;
> +
> + BUG_ON(phy_addr >= 32 || reg >= 32);
The phy_addr and reg conditions are unreachable, the PHY library
ensures that for you.
> +
> + ogma_mac_write(priv, GMAC_REG_GDR, val);
> + ogma_mac_write(priv, GMAC_REG_GAR,
> + phy_addr << OGMA_GMAC_GAR_REG_SHIFT_PA |
> + reg << OGMA_GMAC_GAR_REG_SHIFT_GR |
> + ogma_clk_type(priv->gmac_hz) << GMAC_REG_SHIFT_CR_GAR |
> + OGMA_GMAC_GAR_REG_GW | OGMA_GMAC_GAR_REG_GB);
> +
> + return ogma_mac_wait_while_busy(priv, GMAC_REG_GAR,
> + OGMA_GMAC_GAR_REG_GB);
> +}
> +
> +static int ogma_phy_read(struct mii_bus *bus, int phy_addr, int reg_addr)
> +{
> + struct ogma_priv *priv = bus->priv;
> +
> + BUG_ON(phy_addr >= 32 || reg_addr >= 32);
Same here.
> +
> + ogma_mac_write(priv, GMAC_REG_GAR, OGMA_GMAC_GAR_REG_GB |
> + phy_addr << OGMA_GMAC_GAR_REG_SHIFT_PA |
> + reg_addr << OGMA_GMAC_GAR_REG_SHIFT_GR |
> + ogma_clk_type(priv->gmac_hz) << GMAC_REG_SHIFT_CR_GAR);
> +
> + if (ogma_mac_wait_while_busy(priv, GMAC_REG_GAR, OGMA_GMAC_GAR_REG_GB))
> + return 0;
> +
> + return ogma_mac_read(priv, GMAC_REG_GDR);
> +}
> +
> +int ogma_mii_register(struct ogma_priv *priv)
> +{
> + struct mii_bus *bus = mdiobus_alloc();
> + struct resource res;
> + int ret;
> +
> + if (!bus)
> + return -ENOMEM;
> +
> + of_address_to_resource(priv->dev->of_node, 0, &res);
> + snprintf(bus->id, MII_BUS_ID_SIZE, "%p", (void *)(long)res.start);
If you kept a pointer to the device_node you are using, you could use
np->full_name which is already unique.
[snip]
> +
> +int ogma_netdev_napi_poll(struct napi_struct *napi_p, int budget)
> +{
> + struct ogma_priv *priv = container_of(napi_p, struct ogma_priv, napi);
> + struct net_device *net_device = priv->net_device;
> + struct ogma_rx_pkt_info rx_info;
> + int ret, done = 0, rx_num = 0;
> + struct ogma_frag_info frag;
> + struct sk_buff *skb;
> + u16 len;
> +
> + ogma_ring_irq_clr(priv, OGMA_RING_TX, OGMA_IRQ_EMPTY);
> + ogma_clean_tx_desc_ring(priv);
> +
> + if (netif_queue_stopped(priv->net_device) &&
> + ogma_get_tx_avail_num(priv) >= OGMA_NETDEV_TX_PKT_SCAT_NUM_MAX)
> + netif_wake_queue(priv->net_device);
I would really move the TX processing logic to a separate function for
clarity and logically breaking down things.
> +
> + while (done < budget) {
> + if (!rx_num) {
> + rx_num = ogma_get_rx_num(priv);
> + if (!rx_num)
> + break;
> + }
> +
> + ret = ogma_get_rx_pkt_data(priv, &rx_info, &frag, &len, &skb);
> + if (unlikely(ret == -ENOMEM)) {
> + netif_err(priv, drv, priv->net_device,
> + "%s: rx fail %d\n", __func__, ret);
> + net_device->stats.rx_dropped++;
> + } else {
> + dma_unmap_single(priv->dev, frag.paddr, frag.len,
> + DMA_FROM_DEVICE);
> +
> + skb_put(skb, len);
> + skb->protocol = eth_type_trans(skb, priv->net_device);
> +
> + if (priv->rx_cksum_offload_flag &&
> + rx_info.rx_cksum_result == OGMA_RX_CKSUM_OK)
> + skb->ip_summed = CHECKSUM_UNNECESSARY;
> +
> + napi_gro_receive(napi_p, skb);
> +
> + net_device->stats.rx_packets++;
> + net_device->stats.rx_bytes += len;
> + }
> +
> + done++;
> + rx_num--;
> + }
> +
> + if (done == budget)
> + return budget;
> +
> + napi_complete(napi_p);
> + ogma_write_reg(priv, OGMA_REG_TOP_INTEN_SET,
> + OGMA_TOP_IRQ_REG_NRM_TX | OGMA_TOP_IRQ_REG_NRM_RX);
> +
> + return done;
> +}
> +
> +static int ogma_netdev_stop(struct net_device *net_device)
> +{
> + struct ogma_priv *priv = netdev_priv(net_device);
> +
> + netif_stop_queue(priv->net_device);
You should probably stop NAPI here too before you disable the hardware entirely.
> + ogma_stop_gmac(priv);
> + ogma_stop_desc_ring(priv, OGMA_RING_RX);
> + ogma_stop_desc_ring(priv, OGMA_RING_TX);
> + napi_disable(&priv->napi);
> + phy_stop(priv->phydev);
> + phy_disconnect(priv->phydev);
> + priv->phydev = NULL;
> +
> + pm_runtime_mark_last_busy(priv->dev);
> + pm_runtime_put_autosuspend(priv->dev);
> +
> + return 0;
> +}
> +
> +static netdev_tx_t ogma_netdev_start_xmit(struct sk_buff *skb,
> + struct net_device *net_device)
> +{
> + struct ogma_priv *priv = netdev_priv(net_device);
> + struct ogma_tx_pkt_ctrl tx_ctrl;
> + u16 pend_tx, tso_seg_len = 0;
> + struct ogma_frag_info *scat;
> + skb_frag_t *frag;
> + u8 scat_num;
> + int ret, i;
> +
> + memset(&tx_ctrl, 0, sizeof(struct ogma_tx_pkt_ctrl));
> +
> + ogma_ring_irq_clr(priv, OGMA_RING_TX, OGMA_IRQ_EMPTY);
> +
> + BUG_ON(skb_shinfo(skb)->nr_frags >= OGMA_NETDEV_TX_PKT_SCAT_NUM_MAX);
> + scat_num = skb_shinfo(skb)->nr_frags + 1;
> +
> + scat = kcalloc(scat_num, sizeof(*scat), GFP_NOWAIT);
> + if (!scat)
> + return NETDEV_TX_OK;
Cannot you pre-allocate an array of MAX_SKB_FRAGS + 1 scat elements
and re-use them when you transmit packets? This is a fast-path that
should avoid allocations/re-allocations as much as possible.
> +
> + if (skb->ip_summed == CHECKSUM_PARTIAL) {
> + if (skb->protocol == htons(ETH_P_IP))
> + ip_hdr(skb)->check = 0;
Not sure if that change is actually needed here.
> + tx_ctrl.cksum_offload_flag = 1;
> + }
> +
> + if (skb_is_gso(skb)) {
> + tso_seg_len = skb_shinfo(skb)->gso_size;
> +
> + BUG_ON(skb->ip_summed != CHECKSUM_PARTIAL);
> + BUG_ON(!tso_seg_len);
> + BUG_ON(tso_seg_len > (priv->param.use_jumbo_pkt_flag ?
> + OGMA_TCP_JUMBO_SEG_LEN_MAX : OGMA_TCP_SEG_LEN_MAX));
> +
> + if (tso_seg_len < OGMA_TCP_SEG_LEN_MIN) {
> + tso_seg_len = OGMA_TCP_SEG_LEN_MIN;
> +
> + if (skb->data_len < OGMA_TCP_SEG_LEN_MIN)
> + tso_seg_len = 0;
> + }
> + }
> +
> + if (tso_seg_len > 0) {
> + if (skb->protocol == htons(ETH_P_IP)) {
> + BUG_ON(!(skb_shinfo(skb)->gso_type & SKB_GSO_TCPV4));
> +
> + ip_hdr(skb)->tot_len = 0;
> + tcp_hdr(skb)->check =
> + ~tcp_v4_check(0, ip_hdr(skb)->saddr,
> + ip_hdr(skb)->daddr, 0);
> + } else {
> + BUG_ON(!(skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6));
> + ipv6_hdr(skb)->payload_len = 0;
> + tcp_hdr(skb)->check =
> + ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
> + &ipv6_hdr(skb)->daddr,
> + 0, IPPROTO_TCP, 0);
> + }
> +
> + tx_ctrl.tcp_seg_offload_flag = 1;
> + tx_ctrl.tcp_seg_len = tso_seg_len;
> + }
> +
> + scat[0].paddr = dma_map_single(priv->dev, skb->data,
> + skb_headlen(skb), DMA_TO_DEVICE);
> + if (dma_mapping_error(priv->dev, scat[0].paddr)) {
> + netif_err(priv, drv, priv->net_device,
> + "%s: DMA mapping failed\n", __func__);
> + kfree(scat);
> + return NETDEV_TX_OK;
> + }
> + scat[0].addr = skb->data;
> + scat[0].len = skb_headlen(skb);
> +
> + for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
> + frag = &skb_shinfo(skb)->frags[i];
> + scat[i + 1].paddr =
> + skb_frag_dma_map(priv->dev, frag, 0,
> + skb_frag_size(frag), DMA_TO_DEVICE);
> + scat[i + 1].addr = skb_frag_address(frag);
> + scat[i + 1].len = frag->size;
> + }
> +
> + ogma_mark_skb_type(skb, OGMA_RING_TX);
> +
> + ret = ogma_set_tx_pkt_data(priv, &tx_ctrl, scat_num, scat, skb);
> + if (ret) {
> + netif_err(priv, drv, priv->net_device,
> + "set tx pkt failed %d\n", ret);
> + for (i = 0; i < scat_num; i++)
> + dma_unmap_single(priv->dev, scat[i].paddr,
> + scat[i].len, DMA_TO_DEVICE);
> + kfree(scat);
> + net_device->stats.tx_dropped++;
> +
> + return NETDEV_TX_OK;
> + }
> +
> + kfree(scat);
> +
> + spin_lock(&priv->tx_queue_lock);
> + pend_tx = ogma_get_tx_avail_num(priv);
> +
> + if (pend_tx < OGMA_NETDEV_TX_PKT_SCAT_NUM_MAX) {
> + ogma_ring_irq_enable(priv, OGMA_RING_TX, OGMA_IRQ_EMPTY);
> + netif_stop_queue(net_device);
> + goto err;
> + }
> + if (pend_tx <= DESC_NUM - 2) {
> + ogma_ring_irq_enable(priv, OGMA_RING_TX, OGMA_IRQ_EMPTY);
> + goto err;
> + }
> + ogma_ring_irq_disable(priv, OGMA_RING_TX, OGMA_IRQ_EMPTY);
> +
> +err:
> + spin_unlock(&priv->tx_queue_lock);
> +
> + return NETDEV_TX_OK;
> +}
> +
> +static struct net_device_stats *ogma_netdev_get_stats(struct net_device
> + *net_device)
> +{
> + return &net_device->stats;
> +}
This should be the default, can be dropped.
> +
> +static int ogma_netdev_change_mtu(struct net_device *net_device, int new_mtu)
> +{
> + struct ogma_priv *priv = netdev_priv(net_device);
> +
> + if (!priv->param.use_jumbo_pkt_flag)
> + return eth_change_mtu(net_device, new_mtu);
> +
> + if ((new_mtu < 68) || (new_mtu > 9000))
> + return -EINVAL;
> +
> + net_device->mtu = new_mtu;
> +
> + return 0;
> +}
> +
> +static int ogma_netdev_set_features(struct net_device *net_device,
> + netdev_features_t features)
> +{
> + struct ogma_priv *priv = netdev_priv(net_device);
> +
> + priv->rx_cksum_offload_flag = !!(features & NETIF_F_RXCSUM);
> +
> + return 0;
> +}
> +
> +static int ogma_phy_get_link_speed(struct ogma_priv *priv, bool *half_duplex)
> +{
> + int link_speed = SPEED_10;
> + u32 u;
> +
> + *half_duplex = false;
> +
> + if ((phy_read(priv->phydev, OGMA_PHY_ADDR_MSC) & MSC_1GBIT) &&
> + (phy_read(priv->phydev, OGMA_PHY_ADDR_1000BASE_SR) & SR_1GBIT))
> + link_speed = SPEED_1000;
> + else {
> + u = phy_read(priv->phydev, OGMA_PHY_ADDR_ANA) &
> + phy_read(priv->phydev, OGMA_PHY_ADDR_ANLPA);
> +
> + if (u & OGMA_PHY_ANLPA_REG_TXF) {
> + link_speed = SPEED_100;
> + } else if (u & OGMA_PHY_ANLPA_REG_TXD) {
> + link_speed = SPEED_100;
> + *half_duplex = true;
> + }
> + }
These are already provided by the PHY library in phydev->speed,
phydev->duplex and phydev->link.
> +
> + return link_speed;
> +}
> +
> +static void ogma_phy_adjust_link(struct net_device *net_device)
> +{
> + struct ogma_priv *priv = netdev_priv(net_device);
> + bool half_duplex = false;
> + int link_speed;
> + u32 sr;
> +
> + sr = phy_read(priv->phydev, OGMA_PHY_ADDR_SR);
> +
> + if ((sr & OGMA_PHY_SR_REG_LINK) && (sr & OGMA_PHY_SR_REG_AN_C)) {
I do think this is required if the PHY device provides link status
information in MII_BMSR as it should do. If not, you would have to
write a custom PHY driver for this chip.
> + link_speed = ogma_phy_get_link_speed(priv, &half_duplex);
> + if (priv->actual_link_speed != link_speed ||
> + priv->actual_half_duplex != half_duplex) {
> + netif_info(priv, drv, priv->net_device,
> + "Autoneg: %uMbps, half-duplex:%d\n",
> + link_speed, half_duplex);
> + ogma_stop_gmac(priv);
> + priv->gmac_mode.link_speed = link_speed;
> + priv->gmac_mode.half_duplex_flag = half_duplex;
> + ogma_start_gmac(priv);
Do we really need to go through an expensive stop()/start() sequence
here? Cannot you just change the duplex/speed parameters on the fly?
> +
> + priv->actual_link_speed = link_speed;
> + priv->actual_half_duplex = half_duplex;
> + }
> + }
> +
> + if (!netif_running(priv->net_device) && (sr & OGMA_PHY_SR_REG_LINK)) {
> + netif_info(priv, drv, priv->net_device,
> + "%s: link up\n", __func__);
> + netif_carrier_on(net_device);
The PHY library calls netif_carrier_{on,off} based on the link status
it read from the PHY device.
> + netif_start_queue(net_device);
You cannot start the transmit queue like that, your TX ring might
still be congested.
> + }
> +
> + if (netif_running(priv->net_device) && !(sr & OGMA_PHY_SR_REG_LINK)) {
> + netif_info(priv, drv, priv->net_device,
> + "%s: link down\n", __func__);
> + netif_stop_queue(net_device);
> + netif_carrier_off(net_device);
> + priv->actual_link_speed = 0;
> + priv->actual_half_duplex = 0;
Same here, you should not have to do all of this.
> + }
> +}
> +
> +static int ogma_netdev_open_sub(struct ogma_priv *priv)
> +{
> + napi_enable(&priv->napi);
> +
> + if (ogma_start_desc_ring(priv, OGMA_RING_RX))
> + goto err1;
> + if (ogma_start_desc_ring(priv, OGMA_RING_TX))
> + goto err2;
> +
> + ogma_ring_irq_disable(priv, OGMA_RING_TX, OGMA_IRQ_EMPTY);
> +
> + return 0;
> +
> +err2:
> + ogma_stop_desc_ring(priv, OGMA_RING_RX);
> +err1:
> + napi_disable(&priv->napi);
> +
> + return -EINVAL;
> +}
> +
> +static int ogma_netdev_open(struct net_device *net_device)
> +{
> + struct ogma_priv *priv = netdev_priv(net_device);
> + int ret;
> +
> + pm_runtime_get_sync(priv->dev);
> +
> + ret = ogma_clean_rx_desc_ring(priv);
> + if (ret) {
> + netif_err(priv, drv, priv->net_device,
> + "%s: clean rx desc fail\n", __func__);
> + goto err;
> + }
> +
> + ret = ogma_clean_tx_desc_ring(priv);
> + if (ret) {
> + netif_err(priv, drv, priv->net_device,
> + "%s: clean tx desc fail\n", __func__);
> + goto err;
> + }
> +
> + ogma_ring_irq_clr(priv, OGMA_RING_TX, OGMA_IRQ_EMPTY);
> +
> + priv->phydev = of_phy_connect(priv->net_device, priv->phy_np,
> + &ogma_phy_adjust_link, 0,
> + priv->phy_interface);
> + if (!priv->phydev) {
> + netif_err(priv, link, priv->net_device,
> + "could not find the PHY\n");
> + goto err;
> + }
> +
> + phy_start(priv->phydev);
> +
> + ret = ogma_netdev_open_sub(priv);
> + if (ret) {
> + phy_disconnect(priv->phydev);
> + priv->phydev = NULL;
> + netif_err(priv, link, priv->net_device,
> + "ogma_netdev_open_sub() failed\n");
> + goto err;
> + }
> +
> + /* mask with MAC supported features */
> + priv->phydev->supported &= PHY_BASIC_FEATURES;
> + priv->phydev->advertising = priv->phydev->supported;
These are the defaults, even if you are using the Generic PHY driver,
setting the 'max-speed' property correctly limits the advertised and
supported capabilities.
[snip]
> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + if (!res) {
> + netif_err(priv, probe, priv->net_device,
> + "Missing base resource\n");
> + goto err1;
> + }
> +
> + priv->ioaddr = ioremap_nocache(res->start, res->end - res->start + 1);
> + if (!priv->ioaddr) {
> + netif_err(priv, probe, priv->net_device,
> + "ioremap_nocache() failed\n");
> + err = -EINVAL;
> + goto err1;
> + }
You can simplify this by calling devm_ioremap_resource().
> +
> + res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
> + if (!res) {
> + netif_err(priv, probe, priv->net_device,
> + "Missing rdlar resource\n");
> + goto err1;
> + }
> + priv->rdlar_pa = res->start;
> +
> + res = platform_get_resource(pdev, IORESOURCE_MEM, 2);
> + if (!res) {
> + netif_err(priv, probe, priv->net_device,
> + "Missing tdlar resource\n");
> + goto err1;
> + }
> + priv->tdlar_pa = res->start;
> +
> + res = platform_get_resource(pdev, IORESOURCE_IRQ, 0);
> + if (!res) {
> + netif_err(priv, probe, priv->net_device,
> + "Missing IRQ resource\n");
> + goto err2;
> + }
> + priv->net_device->irq = res->start;
> + err = request_irq(priv->net_device->irq, ogma_irq_handler,
> + IRQF_SHARED, "ogma", priv);
You should call request_irq() in the ndo_open() function such that the
interrupt remains disabled until your driver is ready to service them.
Conversely ndo_close() should call free_irq().
> + if (err) {
> + netif_err(priv, probe, priv->net_device,
> + "request_irq() failed\n");
> + goto err2;
> + }
> + disable_irq(priv->net_device->irq);
> +
> + pdev->dev.coherent_dma_mask = DMA_BIT_MASK(32);
> + pdev->dev.dma_mask = &pdev->dev.coherent_dma_mask;
> +
> + while (priv->clock_count < ARRAY_SIZE(priv->clk)) {
> + priv->clk[priv->clock_count] =
> + of_clk_get(pdev->dev.of_node, priv->clock_count);
> + if (IS_ERR(priv->clk[priv->clock_count])) {
> + if (!priv->clock_count) {
> + netif_err(priv, probe, priv->net_device,
> + "Failed to get clock\n");
> + goto err3;
> + }
> + break;
> + }
> + priv->clock_count++;
> + }
> +
> + /* disable by default */
> + priv->et_coalesce.rx_coalesce_usecs = 0;
> + priv->et_coalesce.rx_max_coalesced_frames = 1;
> + priv->et_coalesce.tx_coalesce_usecs = 0;
> + priv->et_coalesce.tx_max_coalesced_frames = 1;
> +
> + pm_runtime_set_autosuspend_delay(&pdev->dev, 2000); /* 2s delay */
> + pm_runtime_use_autosuspend(&pdev->dev);
> + pm_runtime_enable(&pdev->dev);
> +
> + /* runtime_pm coverage just for probe, enable/disable also cover it */
> + pm_runtime_get_sync(&pdev->dev);
> +
> + priv->param.use_jumbo_pkt_flag = 0;
> + p = of_get_property(pdev->dev.of_node, "max-frame-size", NULL);
> + if (p)
> + priv->param.use_jumbo_pkt_flag = !!(be32_to_cpu(*p) > 8000);
> +
> + hw_ver = ogma_read_reg(priv, OGMA_REG_F_TAIKI_VER);
> +
> + if (OGMA_F_NETSEC_VER_MAJOR_NUM(hw_ver) !=
> + OGMA_F_NETSEC_VER_MAJOR_NUM(OGMA_REG_OGMA_VER_F_TAIKI)) {
> + ret = -ENODEV;
> + goto err3b;
> + }
> +
> + if (priv->param.use_jumbo_pkt_flag)
> + priv->rx_pkt_buf_len = OGMA_RX_JUMBO_PKT_BUF_LEN;
> + else
> + priv->rx_pkt_buf_len = OGMA_RX_PKT_BUF_LEN;
> +
> + for (i = 0; i <= OGMA_RING_MAX; i++) {
> + ret = ogma_alloc_desc_ring(priv, (u8) i);
> + if (ret) {
> + netif_err(priv, probe, priv->net_device,
> + "%s: alloc ring failed\n", __func__);
> + goto err3b;
> + }
> + }
If your interface remains "down", you will be eating memory for
nothing. This is not a whole lot of memory but still, could be some
hundreds of KiB. Better do that in your ndo_open() when your interface
really gets used.
> +
> + ret = ogma_setup_rx_desc(priv, &priv->desc_ring[OGMA_RING_RX]);
> + if (ret) {
> + netif_err(priv, probe, priv->net_device,
> + "%s: fail setup ring\n", __func__);
> + goto err3b;
> + }
> +
> + netif_info(priv, probe, priv->net_device,
> + "IP version: 0x%08x\n", hw_ver);
> +
> + priv->gmac_mode.flow_start_th = OGMA_FLOW_CONTROL_START_THRESHOLD;
> + priv->gmac_mode.flow_stop_th = OGMA_FLOW_CONTROL_STOP_THRESHOLD;
> + priv->gmac_mode.pause_time = pause_time;
> + priv->gmac_hz = clk_get_rate(priv->clk[0]);
> +
> + priv->gmac_mode.half_duplex_flag = 0;
> + priv->gmac_mode.flow_ctrl_enable_flag = 0;
> +
> + scb_irq_temp = ogma_read_reg(priv, OGMA_REG_TOP_INTEN);
> + ogma_write_reg(priv, OGMA_REG_TOP_INTEN_CLR, scb_irq_temp);
> +
> + ret = ogma_hw_configure_to_normal(priv);
> + if (ret) {
> + netif_err(priv, probe, priv->net_device,
> + "%s: normal fail %d\n", __func__, ret);
> + goto err3b;
> + }
> +
> + netif_napi_add(priv->net_device, &priv->napi, ogma_netdev_napi_poll,
> + napi_weight);
> +
> + net_device->netdev_ops = &ogma_netdev_ops;
> + net_device->ethtool_ops = &ogma_ethtool_ops;
> + net_device->features = NETIF_F_SG | NETIF_F_IP_CSUM |
> + NETIF_F_IPV6_CSUM | NETIF_F_TSO |
> + NETIF_F_TSO6 | NETIF_F_GSO |
> + NETIF_F_HIGHDMA | NETIF_F_RXCSUM;
> + priv->net_device->hw_features = priv->net_device->features;
> +
> + priv->rx_cksum_offload_flag = 1;
> + spin_lock_init(&priv->tx_queue_lock);
> +
> + err = ogma_mii_register(priv);
> + if (err) {
> + netif_err(priv, probe, priv->net_device,
> + "mii bus registration failed %d\n", err);
> + goto err3c;
> + }
> +
> + err = register_netdev(priv->net_device);
> + if (err) {
> + netif_err(priv, probe, priv->net_device,
> + "register_netdev() failed\n");
> + goto err4;
> + }
> +
> + ogma_write_reg(priv, OGMA_REG_TOP_INTEN_SET,
> + OGMA_TOP_IRQ_REG_NRM_TX | OGMA_TOP_IRQ_REG_NRM_RX);
> +
> + netif_info(priv, probe, priv->net_device,
> + "%s initialized\n", priv->net_device->name);
> +
> + pm_runtime_mark_last_busy(&pdev->dev);
> + pm_runtime_put_autosuspend(&pdev->dev);
> +
> + return 0;
> +
> +err4:
> + ogma_mii_register(priv);
> +err3c:
> + free_netdev(priv->net_device);
> +err3b:
> + ogma_write_reg(priv, OGMA_REG_TOP_INTEN_SET, scb_irq_temp);
> + ogma_terminate(priv);
> +err3:
> + pm_runtime_put_sync_suspend(&pdev->dev);
> + pm_runtime_disable(&pdev->dev);
> + while (priv->clock_count > 0) {
> + priv->clock_count--;
> + clk_put(priv->clk[priv->clock_count]);
> + }
> +
> + free_irq(priv->net_device->irq, priv);
> +err2:
> + iounmap(priv->ioaddr);
> +err1:
> + kfree(priv);
> +
> + dev_err(&pdev->dev, "init failed\n");
> +
> + return ret;
> +}
> +
> +static int ogma_remove(struct platform_device *pdev)
> +{
> + struct ogma_priv *priv = platform_get_drvdata(pdev);
> + u32 timeout = 1000000;
> +
> + pm_runtime_get_sync(&pdev->dev);
> +
> + ogma_write_reg(priv, OGMA_REG_TOP_INTEN_CLR,
> + OGMA_TOP_IRQ_REG_NRM_TX | OGMA_TOP_IRQ_REG_NRM_RX);
> + BUG_ON(ogma_hw_configure_to_taiki(priv));
> +
> + phy_write(priv->phydev, 0, phy_read(priv->phydev, 0) | (1 << 15));
> + while (--timeout && (phy_read(priv->phydev, 0)) & (1 << 15))
> + ;
If you need to reset the PHY, call genphy_soft_reset() as it will make
sure it waits long enough for the PHY to be in reset.
--
Florian
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists