From: Jiri Pirko <jiri@resnulli.us>
To: Scott Feldman <sfeldma@gmail.com>
Cc: Netdev <netdev@vger.kernel.org>,
"David S. Miller" <davem@davemloft.net>,
idosch@mellanox.com, eladr@mellanox.com,
"ogerlitz@mellanox.com" <ogerlitz@mellanox.com>,
Roopa Prabhu <roopa@cumulusnetworks.com>,
Florian Fainelli <f.fainelli@gmail.com>,
Thomas Graf <tgraf@suug.ch>,
Alexei Starovoitov <ast@plumgrid.com>,
Jamal Hadi Salim <jhs@mojatatu.com>,
Daniel Borkmann <daniel@iogearbox.net>,
john fastabend <john.fastabend@gmail.com>,
"simon.horman@netronome.com" <simon.horman@netronome.com>,
John Linville <linville@tuxdriver.com>,
Andy Gospodarek <andy@greyhouse.net>,
Shrijeet Mukherjee <shm@cumulusnetworks.com>,
"nhorman@tuxdriver.com" <nhorman@tuxdriver.com>,
Jiri Pirko <jiri@mellanox.com>
Subject: Re: [patch net-next 4/4] mlxsw: Introduce Mellanox SwitchX-2 ASIC support
Date: Sun, 26 Jul 2015 09:10:10 +0200 [thread overview]
Message-ID: <20150726071010.GA2150@nanopsycho.fritz.box> (raw)
In-Reply-To: <CAE4R7bAk_nUCuqcJxuAAr6pOqNX7zXWv1zFZDVGwBi7_nk8=kQ@mail.gmail.com>
Sun, Jul 26, 2015 at 04:45:07AM CEST, sfeldma@gmail.com wrote:
>On Thu, Jul 23, 2015 at 8:43 AM, Jiri Pirko <jiri@resnulli.us> wrote:
>> From: Jiri Pirko <jiri@mellanox.com>
>>
>> Benefit from the previously introduced Mellanox Switch infrastructure and
>> add driver for SwitchX-2 ASIC. Note that this driver is very simple now.
>> It implements bare minimum for getting device to work on slow-path.
>> Fast-path offload functionality is going to be added soon.
>>
>> Signed-off-by: Jiri Pirko <jiri@mellanox.com>
>> Signed-off-by: Ido Schimmel <idosch@mellanox.com>
>> Signed-off-by: Elad Raz <eladr@mellanox.com>
>
>[cut]
>
>> +static netdev_tx_t mlxsw_sx_port_xmit(struct sk_buff *skb,
>> + struct net_device *dev)
>> +{
>> + struct mlxsw_sx_port *mlxsw_sx_port = netdev_priv(dev);
>> + struct mlxsw_sx *mlxsw_sx = mlxsw_sx_port->mlxsw_sx;
>> + struct mlxsw_sx_port_pcpu_stats *pcpu_stats;
>> + const struct mlxsw_tx_info tx_info = {
>> + .local_port = mlxsw_sx_port->local_port,
>> + .is_emad = false,
>> + };
>> + int err;
>> +
>> + if (unlikely(skb_headroom(skb) < MLXSW_TXHDR_LEN)) {
>
>Does this happen at all since dev->hard_header_len was set in probe to
>add MLXSW_TXHDR_LEN?
This needs to be done for example for bridge forwarding case and other
forwarding cases.
>
>> + struct sk_buff *skb_new;
>> +
>> + skb_new = skb_realloc_headroom(skb, MLXSW_TXHDR_LEN);
>> + dev_kfree_skb_any(skb);
>> + if (!skb_new) {
>> + this_cpu_inc(mlxsw_sx_port->pcpu_stats->tx_dropped);
>> + return NETDEV_TX_OK;
>> + }
>> + skb = skb_new;
>> + }
>> + mlxsw_sx_txhdr_construct(skb, &tx_info);
>> + err = mlxsw_core_skb_transmit(mlxsw_sx, skb, &tx_info);
>> + if (err == -EAGAIN)
>> + return NETDEV_TX_BUSY;
>
>I think there is a problem here when returning NETDEV_TX_BUSY when
>original skb might have been freed above in the headroom check. (ref
>Documentation/networking/driver.txt).
I have to check this out a bit more. Thanks for pointing that out.
>
>[cut]
>
>> +static int mlxsw_sx_port_dev_addr_get(struct mlxsw_sx_port *mlxsw_sx_port)
>> +{
>> + struct mlxsw_sx *mlxsw_sx = mlxsw_sx_port->mlxsw_sx;
>> + struct net_device *dev = mlxsw_sx_port->dev;
>> + char ppad_pl[MLXSW_REG_PPAD_LEN];
>> + int err;
>> +
>> + mlxsw_reg_ppad_pack(ppad_pl, false, 0);
>> + err = mlxsw_reg_query(mlxsw_sx->core, MLXSW_REG(ppad), ppad_pl);
>> + if (err)
>> + return err;
>> + mlxsw_reg_ppad_mac_memcpy_from(ppad_pl, dev->dev_addr);
>> + /* The last byte in base mac address is always 0 */
>> + dev->dev_addr[ETH_ALEN - 1] += mlxsw_sx_port->local_port;
>
>If MLXSW_PORT_MAX_PORTS > 256, you'll wrap this. Is dev_addr[ETH_ALEN
>- 2] available to carry into?
That will never happen. MLXSW_PORT_MAX_PORTS is 0x40 and address got
from ppad register always ends with 0
>
>> + return 0;
>> +}
>> +
>
>[cut]
>
>> +static int mlxsw_sx_port_create(struct mlxsw_sx *mlxsw_sx, u8 local_port)
>> +{
>> + struct mlxsw_sx_port *mlxsw_sx_port;
>> + struct net_device *dev;
>> + bool usable;
>> + int err;
>> +
>> + dev = alloc_etherdev(sizeof(struct mlxsw_sx_port));
>> + if (!dev)
>> + return -ENOMEM;
>> + mlxsw_sx_port = netdev_priv(dev);
>> + mlxsw_sx_port->dev = dev;
>> + mlxsw_sx_port->mlxsw_sx = mlxsw_sx;
>> + mlxsw_sx_port->local_port = local_port;
>> +
>> + mlxsw_sx_port->pcpu_stats =
>> + netdev_alloc_pcpu_stats(struct mlxsw_sx_port_pcpu_stats);
>> + if (!mlxsw_sx_port->pcpu_stats) {
>> + err = -ENOMEM;
>> + goto err_alloc_stats;
>> + }
>> +
>> + dev->netdev_ops = &mlxsw_sx_port_netdev_ops;
>> + dev->ethtool_ops = &mlxsw_sx_port_ethtool_ops;
>> + dev->switchdev_ops = &mlxsw_sx_port_switchdev_ops;
>> +
>> + err = mlxsw_sx_port_dev_addr_get(mlxsw_sx_port);
>> + if (err) {
>> + dev_err(mlxsw_sx->bus_info->dev, "Port %d: Unable to get port mac address\n",
>> + mlxsw_sx_port->local_port);
>> + goto err_dev_addr_get;
>> + }
>> +
>> + netif_carrier_off(dev);
>> +
>> + dev->features |= NETIF_F_NETNS_LOCAL | NETIF_F_LLTX |
>
>Not supposed to use LLTX in new drivers, according to
>include/linux/netdev_features.h.
In our case, wee need to use this. Since multiple port netdevs may use
the same send dataqueue, we need to do locking ourselves.
Thanks for review!
next prev parent reply other threads:[~2015-07-26 7:10 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-07-23 15:43 [patch net-next 0/4] Introduce Mellanox Technologies Switch ASICs switchdev drivers Jiri Pirko
2015-07-23 15:43 ` [patch net-next 1/4] mlxsw: Introduce Mellanox switch driver core Jiri Pirko
2015-07-23 15:43 ` [patch net-next 2/4] mlxsw: Add PCI bus implementation Jiri Pirko
2015-07-24 4:52 ` Scott Feldman
2015-07-24 5:30 ` Jiri Pirko
2015-07-26 5:15 ` Scott Feldman
2015-07-26 6:49 ` Jiri Pirko
2015-07-23 15:43 ` [patch net-next 3/4] mlxsw: Add interface to access registers and process events Jiri Pirko
2015-07-23 21:12 ` Andy Gospodarek
2015-07-24 5:24 ` Jiri Pirko
2015-07-24 12:09 ` Andy Gospodarek
2015-07-24 5:13 ` Scott Feldman
2015-07-24 5:18 ` Jiri Pirko
2015-07-24 6:48 ` Elad Raz
2015-07-23 15:43 ` [patch net-next 4/4] mlxsw: Introduce Mellanox SwitchX-2 ASIC support Jiri Pirko
2015-07-23 17:19 ` Alexander Duyck
2015-07-23 19:42 ` Jiri Pirko
2015-07-26 2:45 ` Scott Feldman
2015-07-26 7:10 ` Jiri Pirko [this message]
2015-07-26 17:14 ` Jiri Pirko
2015-07-27 6:17 ` Rosen, Rami
2015-07-28 18:06 ` Jiri Pirko
2015-07-24 0:03 ` [patch net-next 0/4] Introduce Mellanox Technologies Switch ASICs switchdev drivers Scott Feldman
2015-07-24 0:38 ` Rustad, Mark D
2015-07-24 5:38 ` Jiri Pirko
2015-07-27 6:51 ` David Miller
2015-07-27 20:21 ` Scott Feldman
2015-07-27 20:27 ` Jiri Pirko
2015-07-27 21:21 ` Florian Fainelli
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20150726071010.GA2150@nanopsycho.fritz.box \
--to=jiri@resnulli.us \
--cc=andy@greyhouse.net \
--cc=ast@plumgrid.com \
--cc=daniel@iogearbox.net \
--cc=davem@davemloft.net \
--cc=eladr@mellanox.com \
--cc=f.fainelli@gmail.com \
--cc=idosch@mellanox.com \
--cc=jhs@mojatatu.com \
--cc=jiri@mellanox.com \
--cc=john.fastabend@gmail.com \
--cc=linville@tuxdriver.com \
--cc=netdev@vger.kernel.org \
--cc=nhorman@tuxdriver.com \
--cc=ogerlitz@mellanox.com \
--cc=roopa@cumulusnetworks.com \
--cc=sfeldma@gmail.com \
--cc=shm@cumulusnetworks.com \
--cc=simon.horman@netronome.com \
--cc=tgraf@suug.ch \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).