* [PATCH net 0/2] mlx4_core fixes
From: Or Gerlitz @ 2013-08-01 16:54 UTC (permalink / raw)
To: davem; +Cc: netdev, amirv, Or Gerlitz
Hi Dave,
Patch #1 complements commit ef96f7d46ad "net/mlx4_en: Handle unassigned VF MAC
address correctly" which took care of the VM side of things. It follows the
approach taken by the ixgbevf (e1941a7 "ixgbevf: Adjust to handle unassigned
MAC address from PF") and bnx2x (e09b74d01 "bnx2x: Zero VFs starting MACs")
drivers too, so users will experience unified approach across vendors.
Would be happy to see both patches land in -stable for 3.10.y - do you
want to handle that through the peridioc batches you send to Greg?
The patches were done against the net tree, commit e44892c1b "igb:
fix vlan filtering in promisc mode when not in VT mode"
Or.
Jack Morgenstein (1):
net/mlx4_core: VFs must ignore the enable_64b_cqe_eqe module param
Or Gerlitz (1):
net/mlx4_core: Don't give VFs MAC addresses which are derived from the PF MAC
drivers/net/ethernet/mellanox/mlx4/fw.c | 11 +----------
drivers/net/ethernet/mellanox/mlx4/main.c | 2 +-
2 files changed, 2 insertions(+), 11 deletions(-)
^ permalink raw reply
* [PATCH net 1/2] net/mlx4_core: Don't give VFs MAC addresses which are derived from the PF MAC
From: Or Gerlitz @ 2013-08-01 16:55 UTC (permalink / raw)
To: davem; +Cc: netdev, amirv, Or Gerlitz
In-Reply-To: <1375376101-20435-1-git-send-email-ogerlitz@mellanox.com>
If the user has not assigned a MAC address to a VM, then don't give it MAC which
is based on the PF one. The current derivation scheme is wrong and leads to VM
MAC collisions when the number of cards/hypervisors becomes big enough.
Instead, just give it zeros and let them figure out what to do with that.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx4/fw.c | 11 +----------
1 files changed, 1 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.c b/drivers/net/ethernet/mellanox/mlx4/fw.c
index 8873d68..6fc6dab 100644
--- a/drivers/net/ethernet/mellanox/mlx4/fw.c
+++ b/drivers/net/ethernet/mellanox/mlx4/fw.c
@@ -845,16 +845,7 @@ int mlx4_QUERY_PORT_wrapper(struct mlx4_dev *dev, int slave,
MLX4_CMD_NATIVE);
if (!err && dev->caps.function != slave) {
- /* if config MAC in DB use it */
- if (priv->mfunc.master.vf_oper[slave].vport[vhcr->in_modifier].state.mac)
- def_mac = priv->mfunc.master.vf_oper[slave].vport[vhcr->in_modifier].state.mac;
- else {
- /* set slave default_mac address */
- MLX4_GET(def_mac, outbox->buf, QUERY_PORT_MAC_OFFSET);
- def_mac += slave << 8;
- priv->mfunc.master.vf_admin[slave].vport[vhcr->in_modifier].mac = def_mac;
- }
-
+ def_mac = priv->mfunc.master.vf_oper[slave].vport[vhcr->in_modifier].state.mac;
MLX4_PUT(outbox->buf, def_mac, QUERY_PORT_MAC_OFFSET);
/* get port type - currently only eth is enabled */
--
1.7.1
^ permalink raw reply related
* Re: [PATCH 0/8] include/net: next set of extern removals
From: Eric Dumazet @ 2013-08-01 16:44 UTC (permalink / raw)
To: Joe Perches
Cc: David Howells, netdev, linux-hams, linux-wireless, linux-kernel,
linux-afs
In-Reply-To: <1375374560.2034.52.camel@joe-AO722>
On Thu, 2013-08-01 at 09:29 -0700, Joe Perches wrote:
> On Thu, 2013-08-01 at 13:04 +0100, David Howells wrote:
> > Joe Perches <joe@perches.com> wrote:
> > > Standardize on no extern use on function prototypes
> > Can we please standardise on _having_ externs on function prototypes?
>
> Why?
>
> What value is there in using extern for function prototypes?
>
> Your argument for "picking out at a glance"
> https://lkml.org/lkml/2013/8/1/237
> really doesn't make sense to me.
>
> Basically, anything with parentheses that's not a #define
> is an extern.
>
> Exceptions exist for extern function pointers, but those
> are fairly unusual anyway. Outside of netfilter,
> extern function pointers are only used about a dozen times
> total in the kernel tree.
>
> So, please provide some examples supporting your view.
My main concern about these changes is they are a huge pain for us
doing bug tracking, rebases and backports. "git blame" and friends will
show lot of noise.
'extern' in include files are an easy way to have a grep friendly
marker, and otherwise are harmless.
_You_ believe they are useless, other people think otherwise.
I really don't see any value doing all these changes on existing code,
apart from adding noise to netdev and adding more work for us.
Using rules for new code is fine, but changing 10 years old code is
really not worth the pain.
I learned C 30 years ago, and using 'extern' is quite natural for me.
It's crazy the time we have to spend on these issues, while we have so
many bugs to fix in the kernel.
^ permalink raw reply
* Re: [PATCH 0/8] include/net: next set of extern removals
From: Joe Perches @ 2013-08-01 16:29 UTC (permalink / raw)
To: David Howells; +Cc: netdev, linux-hams, linux-wireless, linux-kernel, linux-afs
In-Reply-To: <24958.1375358650@warthog.procyon.org.uk>
On Thu, 2013-08-01 at 13:04 +0100, David Howells wrote:
> Joe Perches <joe@perches.com> wrote:
> > Standardize on no extern use on function prototypes
> Can we please standardise on _having_ externs on function prototypes?
Why?
What value is there in using extern for function prototypes?
Your argument for "picking out at a glance"
https://lkml.org/lkml/2013/8/1/237
really doesn't make sense to me.
Basically, anything with parentheses that's not a #define
is an extern.
Exceptions exist for extern function pointers, but those
are fairly unusual anyway. Outside of netfilter,
extern function pointers are only used about a dozen times
total in the kernel tree.
So, please provide some examples supporting your view.
^ permalink raw reply
* Re: [PATCH 4/4] USBNET: ax88179_178a: enable tso if host supports sg dma
From: Grant Grundler @ 2013-08-01 16:12 UTC (permalink / raw)
To: Eric Dumazet
Cc: Ming Lei, David S. Miller, Greg Kroah-Hartman, Oliver Neukum,
Freddy Xin, Ben Hutchings, netdev,
linux-usb-u79uwXL29TY76Z2rM5mHXA
In-Reply-To: <1375371400.10515.155.camel@edumazet-glaptop>
On Thu, Aug 1, 2013 at 8:36 AM, Eric Dumazet <eric.dumazet-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
...
>> IIRC, cpu_to_leXX() macros return the endian "corrected" value.
>> In other words, they need to be assigned to something, no?
>
> Nope, this in in-place byte swapping (for Big Endian only)
Ah ok - thanks for clarifying. I couldn't find the implementation and
this documentation also helped:
http://kernelnewbies.org/EndianIssues
I don't remember implementing in-place endian correction for
arch/parisc (which is BE). I was expecting to find a generic
definition that did a load/swap/store and didn't.
thanks,
grant
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* [PATCH v2] macvlan: validate flags
From: Michael S. Tsirkin @ 2013-08-01 16:09 UTC (permalink / raw)
To: linux-kernel; +Cc: David S. Miller, John Fastabend, Patrick McHardy, netdev
commit df8ef8f3aaa6692970a436204c4429210addb23a
macvlan: add FDB bridge ops and macvlan flags
added a flags field to macvlan, which can be
controlled from userspace.
The idea is to make the interface future-proof
so we can add flags and not new fields.
However, flags value isn't validated, as a result,
userspace can't detect which flags are supported.
Cc: "David S. Miller" <davem@davemloft.net>
Cc: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
Changes from v1:
tweaked commit message
no code changes
Please consider this patch for -stable.
The idea is by the time we add more flags,
everyone has updated to a kernel that
detects errors, so userspace will be able
to detect supported flags cleanly.
drivers/net/macvlan.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/net/macvlan.c b/drivers/net/macvlan.c
index 18373b6..8445a94 100644
--- a/drivers/net/macvlan.c
+++ b/drivers/net/macvlan.c
@@ -736,6 +736,10 @@ static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
return -EADDRNOTAVAIL;
}
+ if (data && data[IFLA_MACVLAN_FLAGS] &&
+ nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
+ return -EINVAL;
+
if (data && data[IFLA_MACVLAN_MODE]) {
switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
case MACVLAN_MODE_PRIVATE:
@@ -809,6 +813,9 @@ int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
if (data && data[IFLA_MACVLAN_FLAGS])
vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
+ if (vlan->flags & ~MACVLAN_FLAG_NOPROMISC)
+ return -EINVAL;
+
if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
if (port->count)
return -EINVAL;
--
MST
^ permalink raw reply related
* Re: [PATCH] fib_rules: add route suppression based on ifgroup
From: Eric Dumazet @ 2013-08-01 16:06 UTC (permalink / raw)
To: Stefan Tomanek; +Cc: netdev
In-Reply-To: <20130801151809.GB21970@zirkel.wertarbyte.de>
On Thu, 2013-08-01 at 17:18 +0200, Stefan Tomanek wrote:
> This change adds the ability to suppress a routing decision based upon the
> interface group the selected interface belongs to.
>
> Signed-off-by: Stefan Tomanek <stefan.tomanek@wertarbyte.de>
> ---
> include/net/fib_rules.h | 2 ++
> include/uapi/linux/fib_rules.h | 2 +-
> net/core/fib_rules.c | 10 ++++++++++
> net/ipv4/fib_rules.c | 23 +++++++++++++++++------
> net/ipv6/fib6_rules.c | 16 +++++++++++++---
> 5 files changed, 43 insertions(+), 10 deletions(-)
>
> diff --git a/include/net/fib_rules.h b/include/net/fib_rules.h
> index 2f286dc..4d32ecc 100644
> --- a/include/net/fib_rules.h
> +++ b/include/net/fib_rules.h
> @@ -19,6 +19,7 @@ struct fib_rule {
> u32 flags;
> u32 table;
> u8 table_prefixlen_min;
> + int suppress_ifgroup;
> u8 action;
> u32 target;
> struct fib_rule __rcu *ctarget;
Why adding holes in this structure ?
^ permalink raw reply
* RE: [PATCH v2] sis900: Fix the tx queue timeout issue
From: Duan Fugang-B38611 @ 2013-08-01 15:50 UTC (permalink / raw)
To: Denis Kirjanov, davem@davemloft.net, venza@brownhat.org
Cc: netdev@vger.kernel.org
In-Reply-To: <1375340470-2126-1-git-send-email-kda@linux-powerpc.org>
On Thursday, August 01, 2013 at 3:01 PM, Denis Kirjanov wrote:
> @@ -1613,9 +1613,12 @@ sis900_start_xmit(struct sk_buff *skb, struct net_device *net_dev)
> unsigned int count_dirty_tx;
>
> /* Don't transmit data before the complete of auto-negotiation */
> - if(!sis_priv->autong_complete){
> - netif_stop_queue(net_dev);
> - return NETDEV_TX_BUSY;
> + if (unlikely(!sis_priv->autong_complete)) {
> + if (netif_msg_tx_err(sis_priv) && net_ratelimit())
> + netdev_dbg(net_dev, "Auto negotiation in progress\n");
> + net_dev->stats.tx_dropped++;
> + dev_kfree_skb(skb);
> + return NETDEV_TX_OK;
> }
I always insist on that tx queue timeout is caused by un-correctly phy link handler flow.
The link up handler flow: Auto-negotiation done-> netif_carrier_on()
The issue cannot be processed in xmit callback, which must be handled in link adjust process.
Thanks,
Andy
^ permalink raw reply
* [PATCH v2 2/2] ip rule: suppress routes based on ifgroup
From: Stefan Tomanek @ 2013-08-01 15:47 UTC (permalink / raw)
To: netdev
Signed-off-by: Stefan Tomanek <stefan.tomanek@wertarbyte.de>
---
include/linux/fib_rules.h | 2 +-
ip/iprule.c | 13 +++++++++++++
2 files changed, 14 insertions(+), 1 deletion(-)
diff --git a/include/linux/fib_rules.h b/include/linux/fib_rules.h
index 59cd31b..63e3116 100644
--- a/include/linux/fib_rules.h
+++ b/include/linux/fib_rules.h
@@ -44,7 +44,7 @@ enum {
FRA_FWMARK, /* mark */
FRA_FLOW, /* flow/class id */
FRA_UNUSED6,
- FRA_UNUSED7,
+ FRA_SUPPRESS_IFGROUP,
FRA_TABLE_PREFIXLEN_MIN,
FRA_TABLE, /* Extended table id */
FRA_FWMASK, /* mask for netfilter mark */
diff --git a/ip/iprule.c b/ip/iprule.c
index a789863..de9f292 100644
--- a/ip/iprule.c
+++ b/ip/iprule.c
@@ -41,6 +41,7 @@ static void usage(void)
fprintf(stderr, " [ goto NUMBER ]\n");
fprintf(stderr, " SUPPRESSOR\n");
fprintf(stderr, "SUPPRESSOR := [ prefixlength NUMBER ]\n");
+ fprintf(stderr, " [ suppress_group DEVGROUP ]\n");
fprintf(stderr, "TABLE_ID := [ local | main | default | NUMBER ]\n");
exit(-1);
}
@@ -163,6 +164,12 @@ int print_rule(const struct sockaddr_nl *who, struct nlmsghdr *n, void *arg)
fprintf(fp, "prefixlength %u ", pl);
}
}
+ if (tb[FRA_SUPPRESS_IFGROUP]) {
+ int group = rta_getattr_u32(tb[FRA_SUPPRESS_IFGROUP]);
+ if (group) {
+ fprintf(fp, "suppress_ifgroup %d ", group);
+ }
+ }
}
if (tb[FRA_FLOW]) {
__u32 to = rta_getattr_u32(tb[FRA_FLOW]);
@@ -325,6 +332,12 @@ static int iprule_modify(int cmd, int argc, char **argv)
if (get_u8(&pl, *argv, 0))
invarg("prefixlength value is invalid\n", *argv);
addattr8(&req.n, sizeof(req), FRA_TABLE_PREFIXLEN_MIN, pl);
+ } else if (matches(*argv, "suppress_group") == 0) {
+ NEXT_ARG();
+ int group;
+ if (rtnl_group_a2n(&group, *argv))
+ invarg("Invalid \"suppress_group\" value\n", *argv);
+ addattr32(&req.n, sizeof(req), FRA_SUPPRESS_IFGROUP, group);
} else if (strcmp(*argv, "dev") == 0 ||
strcmp(*argv, "iif") == 0) {
NEXT_ARG();
--
1.7.10.4
^ permalink raw reply related
* [PATCH v2 1/2] add prefixlength option to "ip rule"
From: Stefan Tomanek @ 2013-08-01 15:46 UTC (permalink / raw)
To: netdev
When configuring a system with multiple network uplinks and default routes, it
is often convenient to reference the main routing table multiple times - but
omitting the default route. Using this modified "ip" utility and the
corresponding kernel patch, this can be achieved by using the following command
sequence:
$ ip route add table secuplink default via 10.42.23.1
$ ip rule add pref 100 table main prefixlength 1
$ ip rule add pref 150 fwmark 0xA table secuplink
With this setup, packets marked 0xA will be processed by the additional routing
table "secuplink", but only if no suitable route in the main routing table can
be found. By using a minimum prefixlength of 1, the default route (/0) of the
table "main" is hidden to packets processed by rule 100; packets traveling to
destinations via more specific routes are processed as usual.
Signed-off-by: Stefan Tomanek <stefan.tomanek@wertarbyte.de>
---
include/linux/fib_rules.h | 2 +-
ip/iprule.c | 19 +++++++++++++++++--
man/man8/ip-rule.8 | 12 ++++++++++++
3 files changed, 30 insertions(+), 3 deletions(-)
diff --git a/include/linux/fib_rules.h b/include/linux/fib_rules.h
index 51da65b..59cd31b 100644
--- a/include/linux/fib_rules.h
+++ b/include/linux/fib_rules.h
@@ -45,7 +45,7 @@ enum {
FRA_FLOW, /* flow/class id */
FRA_UNUSED6,
FRA_UNUSED7,
- FRA_UNUSED8,
+ FRA_TABLE_PREFIXLEN_MIN,
FRA_TABLE, /* Extended table id */
FRA_FWMASK, /* mask for netfilter mark */
FRA_OIFNAME,
diff --git a/ip/iprule.c b/ip/iprule.c
index a5fcd43..a789863 100644
--- a/ip/iprule.c
+++ b/ip/iprule.c
@@ -39,6 +39,8 @@ static void usage(void)
fprintf(stderr, " [ prohibit | reject | unreachable ]\n");
fprintf(stderr, " [ realms [SRCREALM/]DSTREALM ]\n");
fprintf(stderr, " [ goto NUMBER ]\n");
+ fprintf(stderr, " SUPPRESSOR\n");
+ fprintf(stderr, "SUPPRESSOR := [ prefixlength NUMBER ]\n");
fprintf(stderr, "TABLE_ID := [ local | main | default | NUMBER ]\n");
exit(-1);
}
@@ -153,9 +155,15 @@ int print_rule(const struct sockaddr_nl *who, struct nlmsghdr *n, void *arg)
}
table = rtm_get_table(r, tb);
- if (table)
+ if (table) {
fprintf(fp, "lookup %s ", rtnl_rttable_n2a(table, b1, sizeof(b1)));
-
+ if (tb[FRA_TABLE_PREFIXLEN_MIN]) {
+ __u8 pl = rta_getattr_u8(tb[FRA_TABLE_PREFIXLEN_MIN]);
+ if (pl) {
+ fprintf(fp, "prefixlength %u ", pl);
+ }
+ }
+ }
if (tb[FRA_FLOW]) {
__u32 to = rta_getattr_u32(tb[FRA_FLOW]);
__u32 from = to>>16;
@@ -310,6 +318,13 @@ static int iprule_modify(int cmd, int argc, char **argv)
addattr32(&req.n, sizeof(req), FRA_TABLE, tid);
}
table_ok = 1;
+ } else if (matches(*argv, "prefixlength") == 0 ||
+ strcmp(*argv, "pl") == 0) {
+ __u8 pl;
+ NEXT_ARG();
+ if (get_u8(&pl, *argv, 0))
+ invarg("prefixlength value is invalid\n", *argv);
+ addattr8(&req.n, sizeof(req), FRA_TABLE_PREFIXLEN_MIN, pl);
} else if (strcmp(*argv, "dev") == 0 ||
strcmp(*argv, "iif") == 0) {
NEXT_ARG();
diff --git a/man/man8/ip-rule.8 b/man/man8/ip-rule.8
index 36e46f1..ad48f6e 100644
--- a/man/man8/ip-rule.8
+++ b/man/man8/ip-rule.8
@@ -43,6 +43,12 @@ ip-rule \- routing policy database management
.IR ADDRESS " ] [ "
.BR prohibit " | " reject " | " unreachable " ] [ " realms
.RI "[" SRCREALM "/]" DSTREALM " ]"
+.I SUPPRESSOR
+
+.ti -8
+.IR SUPPRESSOR " := [ "
+.B prefixlength
+.IR NUMBER " ]"
.ti -8
.IR TABLE_ID " := [ "
@@ -217,6 +223,12 @@ the routing table identifier to lookup if the rule selector matches.
It is also possible to use lookup instead of table.
.TP
+.BI prefixlength " NUMBER"
+only consider routes that have a minimum prefix length of NUMBER.
+Entries inside the referenced table with a lower prefix length
+will be ignored.
+
+.TP
.BI realms " FROM/TO"
Realms to select if the rule matched and the routing table lookup
succeeded. Realm
--
1.7.10.4
^ permalink raw reply related
* Re: [PATCH 4/4] USBNET: ax88179_178a: enable tso if host supports sg dma
From: Eric Dumazet @ 2013-08-01 15:36 UTC (permalink / raw)
To: Grant Grundler
Cc: Ming Lei, David S. Miller, Greg Kroah-Hartman, Oliver Neukum,
Freddy Xin, Ben Hutchings, netdev, linux-usb
In-Reply-To: <CANEJEGuY3WzxnN4NEKx0RxAJNPO_hy+yt2AqhuTqHQavmawTcw@mail.gmail.com>
On Thu, 2013-08-01 at 08:30 -0700, Grant Grundler wrote:
> http://lxr.free-electrons.com/source/include/linux/byteorder/generic.h#L111
> http://lxr.free-electrons.com/ident?i=__cpu_to_le32s
>
> IIRC, cpu_to_leXX() macros return the endian "corrected" value.
> In other words, they need to be assigned to something, no?
>
Nope, this in in-place byte swapping (for Big Endian only)
^ permalink raw reply
* [PATCH v3 03/13] tile: avoid bug in tilepro net driver built with old hypervisor
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Building against headers from an older Tilera hypervisor can cause
the frags[] array to be overrun. Don't enable TSO in that case.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilepro.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilepro.c b/drivers/net/ethernet/tile/tilepro.c
index 327ff7b..2f4b7b9 100644
--- a/drivers/net/ethernet/tile/tilepro.c
+++ b/drivers/net/ethernet/tile/tilepro.c
@@ -1929,7 +1929,7 @@ static int tile_net_tx(struct sk_buff *skb, struct net_device *dev)
unsigned int csum_start = skb_checksum_start_offset(skb);
- lepp_frag_t frags[LEPP_MAX_FRAGS];
+ lepp_frag_t frags[1 + MAX_SKB_FRAGS];
unsigned int num_frags;
@@ -1944,7 +1944,7 @@ static int tile_net_tx(struct sk_buff *skb, struct net_device *dev)
unsigned int cmd_head, cmd_tail, cmd_next;
unsigned int comp_tail;
- lepp_cmd_t cmds[LEPP_MAX_FRAGS];
+ lepp_cmd_t cmds[1 + MAX_SKB_FRAGS];
/*
@@ -2332,7 +2332,10 @@ static void tile_net_setup(struct net_device *dev)
features |= NETIF_F_LLTX;
features |= NETIF_F_HW_CSUM;
features |= NETIF_F_SG;
- features |= NETIF_F_TSO;
+
+ /* We support TSO iff the HV supports sufficient frags. */
+ if (LEPP_MAX_FRAGS >= 1 + MAX_SKB_FRAGS)
+ features |= NETIF_F_TSO;
/* We can't support HIGHDMA without hash_default, since we need
* to be able to finv() with a VA if we don't have hash_default.
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 10/13] tile: support TSO for IPv6 in tilegx network driver
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilegx.c | 44 +++++++++++++++++++++++++-------------
1 file changed, 29 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index b80a91f..6d94d58 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -36,6 +36,7 @@
#include <linux/io.h>
#include <linux/ctype.h>
#include <linux/ip.h>
+#include <linux/ipv6.h>
#include <linux/tcp.h>
#include <asm/checksum.h>
@@ -1512,20 +1513,20 @@ static int tso_count_edescs(struct sk_buff *skb)
return num_edescs;
}
-/* Prepare modified copies of the skbuff headers.
- * FIXME: add support for IPv6.
- */
+/* Prepare modified copies of the skbuff headers. */
static void tso_headers_prepare(struct sk_buff *skb, unsigned char *headers,
s64 slot)
{
struct skb_shared_info *sh = skb_shinfo(skb);
struct iphdr *ih;
+ struct ipv6hdr *ih6;
struct tcphdr *th;
unsigned int sh_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
unsigned int data_len = skb->len - sh_len;
unsigned char *data = skb->data;
unsigned int ih_off, th_off, p_len;
unsigned int isum_seed, tsum_seed, id, seq;
+ int is_ipv6;
long f_id = -1; /* id of the current fragment */
long f_size = skb_headlen(skb) - sh_len; /* current fragment size */
long f_used = 0; /* bytes used from the current fragment */
@@ -1533,18 +1534,24 @@ static void tso_headers_prepare(struct sk_buff *skb, unsigned char *headers,
int segment;
/* Locate original headers and compute various lengths. */
- ih = ip_hdr(skb);
+ is_ipv6 = skb_is_gso_v6(skb);
+ if (is_ipv6) {
+ ih6 = ipv6_hdr(skb);
+ ih_off = skb_network_offset(skb);
+ } else {
+ ih = ip_hdr(skb);
+ ih_off = skb_network_offset(skb);
+ isum_seed = ((0xFFFF - ih->check) +
+ (0xFFFF - ih->tot_len) +
+ (0xFFFF - ih->id));
+ id = ntohs(ih->id);
+ }
+
th = tcp_hdr(skb);
- ih_off = skb_network_offset(skb);
th_off = skb_transport_offset(skb);
p_len = sh->gso_size;
- /* Set up seed values for IP and TCP csum and initialize id and seq. */
- isum_seed = ((0xFFFF - ih->check) +
- (0xFFFF - ih->tot_len) +
- (0xFFFF - ih->id));
tsum_seed = th->check + (0xFFFF ^ htons(skb->len));
- id = ntohs(ih->id);
seq = ntohl(th->seq);
/* Prepare all the headers. */
@@ -1558,11 +1565,17 @@ static void tso_headers_prepare(struct sk_buff *skb, unsigned char *headers,
memcpy(buf, data, sh_len);
/* Update copied ip header. */
- ih = (struct iphdr *)(buf + ih_off);
- ih->tot_len = htons(sh_len + p_len - ih_off);
- ih->id = htons(id);
- ih->check = csum_long(isum_seed + ih->tot_len +
- ih->id) ^ 0xffff;
+ if (is_ipv6) {
+ ih6 = (struct ipv6hdr *)(buf + ih_off);
+ ih6->payload_len = htons(sh_len + p_len - ih_off -
+ sizeof(*ih6));
+ } else {
+ ih = (struct iphdr *)(buf + ih_off);
+ ih->tot_len = htons(sh_len + p_len - ih_off);
+ ih->id = htons(id);
+ ih->check = csum_long(isum_seed + ih->tot_len +
+ ih->id) ^ 0xffff;
+ }
/* Update copied tcp header. */
th = (struct tcphdr *)(buf + th_off);
@@ -1954,6 +1967,7 @@ static void tile_net_setup(struct net_device *dev)
features |= NETIF_F_HW_CSUM;
features |= NETIF_F_SG;
features |= NETIF_F_TSO;
+ features |= NETIF_F_TSO6;
dev->hw_features |= features;
dev->vlan_features |= features;
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 09/13] tile: support multiple mPIPE shims in tilegx network driver
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
The initial driver support was for a single mPIPE shim on the chip
(as is the case for the Gx36 hardware). The Gx72 chip has two mPIPE
shims, so we extend the driver to handle that case.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
arch/tile/gxio/iorpc_mpipe_info.c | 18 +
arch/tile/gxio/mpipe.c | 25 +-
arch/tile/include/gxio/iorpc_mpipe_info.h | 4 +
arch/tile/include/gxio/mpipe.h | 28 ++
arch/tile/include/hv/drv_mpipe_intf.h | 3 +
drivers/net/ethernet/tile/tilegx.c | 551 ++++++++++++++++++------------
6 files changed, 417 insertions(+), 212 deletions(-)
diff --git a/arch/tile/gxio/iorpc_mpipe_info.c b/arch/tile/gxio/iorpc_mpipe_info.c
index d0254aa..64883aa 100644
--- a/arch/tile/gxio/iorpc_mpipe_info.c
+++ b/arch/tile/gxio/iorpc_mpipe_info.c
@@ -16,6 +16,24 @@
#include "gxio/iorpc_mpipe_info.h"
+struct instance_aux_param {
+ _gxio_mpipe_link_name_t name;
+};
+
+int gxio_mpipe_info_instance_aux(gxio_mpipe_info_context_t * context,
+ _gxio_mpipe_link_name_t name)
+{
+ struct instance_aux_param temp;
+ struct instance_aux_param *params = &temp;
+
+ params->name = name;
+
+ return hv_dev_pwrite(context->fd, 0, (HV_VirtAddr) params,
+ sizeof(*params), GXIO_MPIPE_INFO_OP_INSTANCE_AUX);
+}
+
+EXPORT_SYMBOL(gxio_mpipe_info_instance_aux);
+
struct enumerate_aux_param {
_gxio_mpipe_link_name_t name;
_gxio_mpipe_link_mac_t mac;
diff --git a/arch/tile/gxio/mpipe.c b/arch/tile/gxio/mpipe.c
index 0567cf0..5301a9f 100644
--- a/arch/tile/gxio/mpipe.c
+++ b/arch/tile/gxio/mpipe.c
@@ -36,8 +36,14 @@ int gxio_mpipe_init(gxio_mpipe_context_t *context, unsigned int mpipe_index)
int fd;
int i;
+ if (mpipe_index >= GXIO_MPIPE_INSTANCE_MAX)
+ return -EINVAL;
+
snprintf(file, sizeof(file), "mpipe/%d/iorpc", mpipe_index);
fd = hv_dev_open((HV_VirtAddr) file, 0);
+
+ context->fd = fd;
+
if (fd < 0) {
if (fd >= GXIO_ERR_MIN && fd <= GXIO_ERR_MAX)
return fd;
@@ -45,8 +51,6 @@ int gxio_mpipe_init(gxio_mpipe_context_t *context, unsigned int mpipe_index)
return -ENODEV;
}
- context->fd = fd;
-
/* Map in the MMIO space. */
context->mmio_cfg_base = (void __force *)
iorpc_ioremap(fd, HV_MPIPE_CONFIG_MMIO_OFFSET,
@@ -64,12 +68,15 @@ int gxio_mpipe_init(gxio_mpipe_context_t *context, unsigned int mpipe_index)
for (i = 0; i < 8; i++)
context->__stacks.stacks[i] = 255;
+ context->instance = mpipe_index;
+
return 0;
fast_failed:
iounmap((void __force __iomem *)(context->mmio_cfg_base));
cfg_failed:
hv_dev_close(context->fd);
+ context->fd = -1;
return -ENODEV;
}
@@ -496,6 +503,20 @@ static gxio_mpipe_context_t *_gxio_get_link_context(void)
return contextp;
}
+int gxio_mpipe_link_instance(const char *link_name)
+{
+ _gxio_mpipe_link_name_t name;
+ gxio_mpipe_context_t *context = _gxio_get_link_context();
+
+ if (!context)
+ return GXIO_ERR_NO_DEVICE;
+
+ strncpy(name.name, link_name, sizeof(name.name));
+ name.name[GXIO_MPIPE_LINK_NAME_LEN - 1] = '\0';
+
+ return gxio_mpipe_info_instance_aux(context, name);
+}
+
int gxio_mpipe_link_enumerate_mac(int idx, char *link_name, uint8_t *link_mac)
{
int rv;
diff --git a/arch/tile/include/gxio/iorpc_mpipe_info.h b/arch/tile/include/gxio/iorpc_mpipe_info.h
index 0bcf3f7..476c5e5 100644
--- a/arch/tile/include/gxio/iorpc_mpipe_info.h
+++ b/arch/tile/include/gxio/iorpc_mpipe_info.h
@@ -27,11 +27,15 @@
#include <asm/pgtable.h>
+#define GXIO_MPIPE_INFO_OP_INSTANCE_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1250)
#define GXIO_MPIPE_INFO_OP_ENUMERATE_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1251)
#define GXIO_MPIPE_INFO_OP_GET_MMIO_BASE IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x8000)
#define GXIO_MPIPE_INFO_OP_CHECK_MMIO_OFFSET IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x8001)
+int gxio_mpipe_info_instance_aux(gxio_mpipe_info_context_t * context,
+ _gxio_mpipe_link_name_t name);
+
int gxio_mpipe_info_enumerate_aux(gxio_mpipe_info_context_t * context,
unsigned int idx,
_gxio_mpipe_link_name_t * name,
diff --git a/arch/tile/include/gxio/mpipe.h b/arch/tile/include/gxio/mpipe.h
index ed742e3..eb7fee4 100644
--- a/arch/tile/include/gxio/mpipe.h
+++ b/arch/tile/include/gxio/mpipe.h
@@ -220,6 +220,13 @@ typedef MPIPE_PDESC_t gxio_mpipe_idesc_t;
*/
typedef MPIPE_EDMA_DESC_t gxio_mpipe_edesc_t;
+/*
+ * Max # of mpipe instances. 2 currently.
+ */
+#define GXIO_MPIPE_INSTANCE_MAX HV_MPIPE_INSTANCE_MAX
+
+#define NR_MPIPE_MAX GXIO_MPIPE_INSTANCE_MAX
+
/* Get the "va" field from an "idesc".
*
* This is the address at which the ingress hardware copied the first
@@ -311,6 +318,9 @@ typedef struct {
/* File descriptor for calling up to Linux (and thus the HV). */
int fd;
+ /* Corresponding mpipe instance #. */
+ int instance;
+
/* The VA at which configuration registers are mapped. */
char *mmio_cfg_base;
@@ -1716,6 +1726,24 @@ typedef struct {
uint8_t mac;
} gxio_mpipe_link_t;
+/* Translate a link name to the instance number of the mPIPE shim which is
+ * connected to that link. This call does not verify whether the link is
+ * currently available, and does not reserve any link resources;
+ * gxio_mpipe_link_open() must be called to perform those functions.
+ *
+ * Typically applications will call this function to translate a link name
+ * to an mPIPE instance number; call gxio_mpipe_init(), passing it that
+ * instance number, to initialize the mPIPE shim; and then call
+ * gxio_mpipe_link_open(), passing it the same link name plus the mPIPE
+ * context, to configure the link.
+ *
+ * @param link_name Name of the link; see @ref gxio_mpipe_link_names.
+ * @return The mPIPE instance number which is associated with the named
+ * link, or a negative error code (::GXIO_ERR_NO_DEVICE) if the link does
+ * not exist.
+ */
+extern int gxio_mpipe_link_instance(const char *link_name);
+
/* Retrieve one of this system's legal link names, and its MAC address.
*
* @param index Link name index. If a system supports N legal link names,
diff --git a/arch/tile/include/hv/drv_mpipe_intf.h b/arch/tile/include/hv/drv_mpipe_intf.h
index 6cdae3b..c97e416 100644
--- a/arch/tile/include/hv/drv_mpipe_intf.h
+++ b/arch/tile/include/hv/drv_mpipe_intf.h
@@ -23,6 +23,9 @@
#include <arch/mpipe_constants.h>
+/** Number of mPIPE instances supported */
+#define HV_MPIPE_INSTANCE_MAX (2)
+
/** Number of buffer stacks (32). */
#define HV_MPIPE_NUM_BUFFER_STACKS \
(MPIPE_MMIO_INIT_DAT_GX36_1__BUFFER_STACK_MASK_WIDTH)
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index 2b1c31f..b80a91f 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -133,27 +133,31 @@ struct tile_net_tx_wake {
/* Info for a specific cpu. */
struct tile_net_info {
- /* The NAPI struct. */
- struct napi_struct napi;
- /* Packet queue. */
- gxio_mpipe_iqueue_t iqueue;
/* Our cpu. */
int my_cpu;
- /* True if iqueue is valid. */
- bool has_iqueue;
- /* NAPI flags. */
- bool napi_added;
- bool napi_enabled;
- /* Number of buffers (by kind) which must still be provided. */
- unsigned int num_needed_buffers[MAX_KINDS];
/* A timer for handling egress completions. */
struct hrtimer egress_timer;
/* True if "egress_timer" is scheduled. */
bool egress_timer_scheduled;
- /* Comps for each egress channel. */
- struct tile_net_comps *comps_for_echannel[TILE_NET_CHANNELS];
- /* Transmit wake timer for each egress channel. */
- struct tile_net_tx_wake tx_wake[TILE_NET_CHANNELS];
+ struct info_mpipe {
+ /* Packet queue. */
+ gxio_mpipe_iqueue_t iqueue;
+ /* The NAPI struct. */
+ struct napi_struct napi;
+ /* Number of buffers (by kind) which must still be provided. */
+ unsigned int num_needed_buffers[MAX_KINDS];
+ /* instance id. */
+ int instance;
+ /* True if iqueue is valid. */
+ bool has_iqueue;
+ /* NAPI flags. */
+ bool napi_added;
+ bool napi_enabled;
+ /* Comps for each egress channel. */
+ struct tile_net_comps *comps_for_echannel[TILE_NET_CHANNELS];
+ /* Transmit wake timer for each egress channel. */
+ struct tile_net_tx_wake tx_wake[TILE_NET_CHANNELS];
+ } mpipe[NR_MPIPE_MAX];
};
/* Info for egress on a particular egress channel. */
@@ -178,17 +182,54 @@ struct tile_net_priv {
int loopify_channel;
/* The egress channel (channel or loopify_channel). */
int echannel;
+ /* mPIPE instance, 0 or 1. */
+ int instance;
};
-/* Egress info, indexed by "priv->echannel" (lazily created as needed). */
-static struct tile_net_egress egress_for_echannel[TILE_NET_CHANNELS];
+static struct mpipe_data {
+ /* The ingress irq. */
+ int ingress_irq;
-/* Devices currently associated with each channel.
- * NOTE: The array entry can become NULL after ifconfig down, but
- * we do not free the underlying net_device structures, so it is
- * safe to use a pointer after reading it from this array.
- */
-static struct net_device *tile_net_devs_for_channel[TILE_NET_CHANNELS];
+ /* The "context" for all devices. */
+ gxio_mpipe_context_t context;
+
+ /* Egress info, indexed by "priv->echannel"
+ * (lazily created as needed).
+ */
+ struct tile_net_egress
+ egress_for_echannel[TILE_NET_CHANNELS];
+
+ /* Devices currently associated with each channel.
+ * NOTE: The array entry can become NULL after ifconfig down, but
+ * we do not free the underlying net_device structures, so it is
+ * safe to use a pointer after reading it from this array.
+ */
+ struct net_device
+ *tile_net_devs_for_channel[TILE_NET_CHANNELS];
+
+ /* The actual memory allocated for the buffer stacks. */
+ void *buffer_stack_vas[MAX_KINDS];
+
+ /* The amount of memory allocated for each buffer stack. */
+ size_t buffer_stack_bytes[MAX_KINDS];
+
+ /* The first buffer stack index
+ * (small = +0, large = +1, jumbo = +2).
+ */
+ int first_buffer_stack;
+
+ /* The buckets. */
+ int first_bucket;
+ int num_buckets;
+
+} mpipe_data[NR_MPIPE_MAX] = {
+ [0 ... (NR_MPIPE_MAX - 1)] {
+ .ingress_irq = -1,
+ .first_buffer_stack = -1,
+ .first_bucket = -1,
+ .num_buckets = 1
+ }
+};
/* A mutex for "tile_net_devs_for_channel". */
static DEFINE_MUTEX(tile_net_devs_for_channel_mutex);
@@ -196,8 +237,6 @@ static DEFINE_MUTEX(tile_net_devs_for_channel_mutex);
/* The per-cpu info. */
static DEFINE_PER_CPU(struct tile_net_info, per_cpu_info);
-/* The "context" for all devices. */
-static gxio_mpipe_context_t context;
/* The buffer size enums for each buffer stack.
* See arch/tile/include/gxio/mpipe.h for the set of possible values.
@@ -210,22 +249,6 @@ static gxio_mpipe_buffer_size_enum_t buffer_size_enums[MAX_KINDS] = {
GXIO_MPIPE_BUFFER_SIZE_16384
};
-/* The actual memory allocated for the buffer stacks. */
-static void *buffer_stack_vas[MAX_KINDS];
-
-/* The amount of memory allocated for each buffer stack. */
-static size_t buffer_stack_bytes[MAX_KINDS];
-
-/* The first buffer stack index (small = +0, large = +1, jumbo = +2). */
-static int first_buffer_stack = -1;
-
-/* The buckets. */
-static int first_bucket = -1;
-static int num_buckets = 1;
-
-/* The ingress irq. */
-static int ingress_irq = -1;
-
/* Text value of tile_net.cpus if passed as a module parameter. */
static char *network_cpus_string;
@@ -241,6 +264,13 @@ static char *custom_str;
/* If "tile_net.jumbo=NUM" was specified, this is "NUM". */
static uint jumbo_num;
+/* Obtain mpipe instance from struct tile_net_priv given struct net_device. */
+static inline int mpipe_instance(struct net_device *dev)
+{
+ struct tile_net_priv *priv = netdev_priv(dev);
+ return priv->instance;
+}
+
/* The "tile_net.cpus" argument specifies the cpus that are dedicated
* to handle ingress packets.
*
@@ -314,8 +344,9 @@ static void tile_net_stats_add(unsigned long value, unsigned long *field)
}
/* Allocate and push a buffer. */
-static bool tile_net_provide_buffer(int kind)
+static bool tile_net_provide_buffer(int instance, int kind)
{
+ struct mpipe_data *md = &mpipe_data[instance];
gxio_mpipe_buffer_size_enum_t bse = buffer_size_enums[kind];
size_t bs = gxio_mpipe_buffer_size_enum_to_buffer_size(bse);
const unsigned long buffer_alignment = 128;
@@ -337,7 +368,7 @@ static bool tile_net_provide_buffer(int kind)
/* Make sure "skb" and the back-pointer have been flushed. */
wmb();
- gxio_mpipe_push_buffer(&context, first_buffer_stack + kind,
+ gxio_mpipe_push_buffer(&md->context, md->first_buffer_stack + kind,
(void *)va_to_tile_io_addr(skb->data));
return true;
@@ -363,11 +394,14 @@ static struct sk_buff *mpipe_buf_to_skb(void *va)
return skb;
}
-static void tile_net_pop_all_buffers(int stack)
+static void tile_net_pop_all_buffers(int instance, int stack)
{
+ struct mpipe_data *md = &mpipe_data[instance];
+
for (;;) {
tile_io_addr_t addr =
- (tile_io_addr_t)gxio_mpipe_pop_buffer(&context, stack);
+ (tile_io_addr_t)gxio_mpipe_pop_buffer(&md->context,
+ stack);
if (addr == 0)
break;
dev_kfree_skb_irq(mpipe_buf_to_skb(tile_io_addr_to_va(addr)));
@@ -378,17 +412,21 @@ static void tile_net_pop_all_buffers(int stack)
static void tile_net_provide_needed_buffers(void)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
- int kind;
-
- for (kind = 0; kind < MAX_KINDS; kind++) {
- while (info->num_needed_buffers[kind] != 0) {
- if (!tile_net_provide_buffer(kind)) {
- /* Add info to the allocation failure dump. */
- pr_notice("Tile %d still needs some buffers\n",
- info->my_cpu);
- return;
+ int instance, kind;
+ for (instance = 0; instance < NR_MPIPE_MAX &&
+ info->mpipe[instance].has_iqueue; instance++) {
+ for (kind = 0; kind < MAX_KINDS; kind++) {
+ while (info->mpipe[instance].num_needed_buffers[kind]
+ != 0) {
+ if (!tile_net_provide_buffer(instance, kind)) {
+ pr_notice("Tile %d still needs"
+ " some buffers\n",
+ info->my_cpu);
+ return;
+ }
+ info->mpipe[instance].
+ num_needed_buffers[kind]--;
}
- info->num_needed_buffers[kind]--;
}
}
}
@@ -412,6 +450,7 @@ static void tile_net_receive_skb(struct net_device *dev, struct sk_buff *skb,
gxio_mpipe_idesc_t *idesc, unsigned long len)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
+ int instance = mpipe_instance(dev);
/* Encode the actual packet length. */
skb_put(skb, len);
@@ -422,7 +461,7 @@ static void tile_net_receive_skb(struct net_device *dev, struct sk_buff *skb,
if (idesc->cs && idesc->csum_seed_val == 0xFFFF)
skb->ip_summed = CHECKSUM_UNNECESSARY;
- napi_gro_receive(&info->napi, skb);
+ napi_gro_receive(&info->mpipe[instance].napi, skb);
/* Update stats. */
tile_net_stats_add(1, &dev->stats.rx_packets);
@@ -430,18 +469,19 @@ static void tile_net_receive_skb(struct net_device *dev, struct sk_buff *skb,
/* Need a new buffer. */
if (idesc->size == buffer_size_enums[0])
- info->num_needed_buffers[0]++;
+ info->mpipe[instance].num_needed_buffers[0]++;
else if (idesc->size == buffer_size_enums[1])
- info->num_needed_buffers[1]++;
+ info->mpipe[instance].num_needed_buffers[1]++;
else
- info->num_needed_buffers[2]++;
+ info->mpipe[instance].num_needed_buffers[2]++;
}
/* Handle a packet. Return true if "processed", false if "filtered". */
-static bool tile_net_handle_packet(gxio_mpipe_idesc_t *idesc)
+static bool tile_net_handle_packet(int instance, gxio_mpipe_idesc_t *idesc)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
- struct net_device *dev = tile_net_devs_for_channel[idesc->channel];
+ struct mpipe_data *md = &mpipe_data[instance];
+ struct net_device *dev = md->tile_net_devs_for_channel[idesc->channel];
uint8_t l2_offset;
void *va;
void *buf;
@@ -477,7 +517,7 @@ static bool tile_net_handle_packet(gxio_mpipe_idesc_t *idesc)
if (dev)
tile_net_stats_add(1, &dev->stats.rx_dropped);
drop:
- gxio_mpipe_iqueue_drop(&info->iqueue, idesc);
+ gxio_mpipe_iqueue_drop(&info->mpipe[instance].iqueue, idesc);
} else {
struct sk_buff *skb = mpipe_buf_to_skb(va);
@@ -487,7 +527,7 @@ drop:
tile_net_receive_skb(dev, skb, idesc, len);
}
- gxio_mpipe_iqueue_consume(&info->iqueue, idesc);
+ gxio_mpipe_iqueue_consume(&info->mpipe[instance].iqueue, idesc);
return !filter;
}
@@ -508,14 +548,20 @@ static int tile_net_poll(struct napi_struct *napi, int budget)
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
unsigned int work = 0;
gxio_mpipe_idesc_t *idesc;
- int i, n;
-
- /* Process packets. */
- while ((n = gxio_mpipe_iqueue_try_peek(&info->iqueue, &idesc)) > 0) {
+ int instance, i, n;
+ struct mpipe_data *md;
+ struct info_mpipe *info_mpipe =
+ container_of(napi, struct info_mpipe, napi);
+
+ instance = info_mpipe->instance;
+ while ((n = gxio_mpipe_iqueue_try_peek(
+ &info_mpipe->iqueue,
+ &idesc)) > 0) {
for (i = 0; i < n; i++) {
if (i == TILE_NET_BATCH)
goto done;
- if (tile_net_handle_packet(idesc + i)) {
+ if (tile_net_handle_packet(instance,
+ idesc + i)) {
if (++work >= budget)
goto done;
}
@@ -523,14 +569,16 @@ static int tile_net_poll(struct napi_struct *napi, int budget)
}
/* There are no packets left. */
- napi_complete(&info->napi);
+ napi_complete(&info_mpipe->napi);
+ md = &mpipe_data[instance];
/* Re-enable hypervisor interrupts. */
- gxio_mpipe_enable_notif_ring_interrupt(&context, info->iqueue.ring);
+ gxio_mpipe_enable_notif_ring_interrupt(
+ &md->context, info->mpipe[instance].iqueue.ring);
/* HACK: Avoid the "rotting packet" problem. */
- if (gxio_mpipe_iqueue_try_peek(&info->iqueue, &idesc) > 0)
- napi_schedule(&info->napi);
+ if (gxio_mpipe_iqueue_try_peek(&info_mpipe->iqueue, &idesc) > 0)
+ napi_schedule(&info_mpipe->napi);
/* ISSUE: Handle completions? */
@@ -540,11 +588,11 @@ done:
return work;
}
-/* Handle an ingress interrupt on the current cpu. */
-static irqreturn_t tile_net_handle_ingress_irq(int irq, void *unused)
+/* Handle an ingress interrupt from an instance on the current cpu. */
+static irqreturn_t tile_net_handle_ingress_irq(int irq, void *id)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
- napi_schedule(&info->napi);
+ napi_schedule(&info->mpipe[(uint64_t)id].napi);
return IRQ_HANDLED;
}
@@ -586,7 +634,9 @@ static void tile_net_schedule_tx_wake_timer(struct net_device *dev,
{
struct tile_net_info *info = &per_cpu(per_cpu_info, tx_queue_idx);
struct tile_net_priv *priv = netdev_priv(dev);
- struct tile_net_tx_wake *tx_wake = &info->tx_wake[priv->echannel];
+ int instance = priv->instance;
+ struct tile_net_tx_wake *tx_wake =
+ &info->mpipe[instance].tx_wake[priv->echannel];
hrtimer_start(&tx_wake->timer,
ktime_set(0, TX_TIMER_DELAY_USEC * 1000UL),
@@ -624,7 +674,7 @@ static enum hrtimer_restart tile_net_handle_egress_timer(struct hrtimer *t)
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
unsigned long irqflags;
bool pending = false;
- int i;
+ int i, instance;
local_irq_save(irqflags);
@@ -632,13 +682,19 @@ static enum hrtimer_restart tile_net_handle_egress_timer(struct hrtimer *t)
info->egress_timer_scheduled = false;
/* Free all possible comps for this tile. */
- for (i = 0; i < TILE_NET_CHANNELS; i++) {
- struct tile_net_egress *egress = &egress_for_echannel[i];
- struct tile_net_comps *comps = info->comps_for_echannel[i];
- if (comps->comp_last >= comps->comp_next)
- continue;
- tile_net_free_comps(egress->equeue, comps, -1, true);
- pending = pending || (comps->comp_last < comps->comp_next);
+ for (instance = 0; instance < NR_MPIPE_MAX &&
+ info->mpipe[instance].has_iqueue; instance++) {
+ for (i = 0; i < TILE_NET_CHANNELS; i++) {
+ struct tile_net_egress *egress =
+ &mpipe_data[instance].egress_for_echannel[i];
+ struct tile_net_comps *comps =
+ info->mpipe[instance].comps_for_echannel[i];
+ if (!egress || comps->comp_last >= comps->comp_next)
+ continue;
+ tile_net_free_comps(egress->equeue, comps, -1, true);
+ pending = pending ||
+ (comps->comp_last < comps->comp_next);
+ }
}
/* Reschedule timer if needed. */
@@ -650,13 +706,15 @@ static enum hrtimer_restart tile_net_handle_egress_timer(struct hrtimer *t)
return HRTIMER_NORESTART;
}
-/* Helper function for "tile_net_update()". */
-static void manage_ingress_irq(void *enable)
+/* Helper functions for "tile_net_update()". */
+static void enable_ingress_irq(void *irq)
{
- if (enable)
- enable_percpu_irq(ingress_irq, 0);
- else
- disable_percpu_irq(ingress_irq);
+ enable_percpu_irq((long)irq, 0);
+}
+
+static void disable_ingress_irq(void *irq)
+{
+ disable_percpu_irq((long)irq);
}
/* Helper function for tile_net_open() and tile_net_stop().
@@ -666,19 +724,22 @@ static int tile_net_update(struct net_device *dev)
{
static gxio_mpipe_rules_t rules; /* too big to fit on the stack */
bool saw_channel = false;
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
int channel;
int rc;
int cpu;
- gxio_mpipe_rules_init(&rules, &context);
+ saw_channel = false;
+ gxio_mpipe_rules_init(&rules, &md->context);
for (channel = 0; channel < TILE_NET_CHANNELS; channel++) {
- if (tile_net_devs_for_channel[channel] == NULL)
+ if (md->tile_net_devs_for_channel[channel] == NULL)
continue;
if (!saw_channel) {
saw_channel = true;
- gxio_mpipe_rules_begin(&rules, first_bucket,
- num_buckets, NULL);
+ gxio_mpipe_rules_begin(&rules, md->first_bucket,
+ md->num_buckets, NULL);
gxio_mpipe_rules_set_headroom(&rules, NET_IP_ALIGN);
}
gxio_mpipe_rules_add_channel(&rules, channel);
@@ -689,7 +750,8 @@ static int tile_net_update(struct net_device *dev)
*/
rc = gxio_mpipe_rules_commit(&rules);
if (rc != 0) {
- netdev_warn(dev, "gxio_mpipe_rules_commit failed: %d\n", rc);
+ netdev_warn(dev, "gxio_mpipe_rules_commit: mpipe[%d] %d\n",
+ instance, rc);
return -EIO;
}
@@ -697,35 +759,38 @@ static int tile_net_update(struct net_device *dev)
* We use on_each_cpu to handle the IPI mask or unmask.
*/
if (!saw_channel)
- on_each_cpu(manage_ingress_irq, (void *)0, 1);
+ on_each_cpu(disable_ingress_irq,
+ (void *)(long)(md->ingress_irq), 1);
for_each_online_cpu(cpu) {
struct tile_net_info *info = &per_cpu(per_cpu_info, cpu);
- if (!info->has_iqueue)
+
+ if (!info->mpipe[instance].has_iqueue)
continue;
if (saw_channel) {
- if (!info->napi_added) {
- netif_napi_add(dev, &info->napi,
+ if (!info->mpipe[instance].napi_added) {
+ netif_napi_add(dev, &info->mpipe[instance].napi,
tile_net_poll, TILE_NET_WEIGHT);
- info->napi_added = true;
+ info->mpipe[instance].napi_added = true;
}
- if (!info->napi_enabled) {
- napi_enable(&info->napi);
- info->napi_enabled = true;
+ if (!info->mpipe[instance].napi_enabled) {
+ napi_enable(&info->mpipe[instance].napi);
+ info->mpipe[instance].napi_enabled = true;
}
} else {
- if (info->napi_enabled) {
- napi_disable(&info->napi);
- info->napi_enabled = false;
+ if (info->mpipe[instance].napi_enabled) {
+ napi_disable(&info->mpipe[instance].napi);
+ info->mpipe[instance].napi_enabled = false;
}
/* FIXME: Drain the iqueue. */
}
}
if (saw_channel)
- on_each_cpu(manage_ingress_irq, (void *)1, 1);
+ on_each_cpu(enable_ingress_irq,
+ (void *)(long)(md->ingress_irq), 1);
/* HACK: Allow packets to flow in the simulator. */
if (saw_channel)
- sim_enable_mpipe_links(0, -1);
+ sim_enable_mpipe_links(instance, -1);
return 0;
}
@@ -735,46 +800,52 @@ static int create_buffer_stack(struct net_device *dev,
int kind, size_t num_buffers)
{
pte_t hash_pte = pte_set_home((pte_t) { 0 }, PAGE_HOME_HASH);
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
size_t needed = gxio_mpipe_calc_buffer_stack_bytes(num_buffers);
- int stack_idx = first_buffer_stack + kind;
+ int stack_idx = md->first_buffer_stack + kind;
void *va;
int i, rc;
/* Round up to 64KB and then use alloc_pages() so we get the
* required 64KB alignment.
*/
- buffer_stack_bytes[kind] = ALIGN(needed, 64 * 1024);
+ md->buffer_stack_bytes[kind] =
+ ALIGN(needed, 64 * 1024);
- va = alloc_pages_exact(buffer_stack_bytes[kind], GFP_KERNEL);
+ va = alloc_pages_exact(md->buffer_stack_bytes[kind], GFP_KERNEL);
if (va == NULL) {
netdev_err(dev,
"Could not alloc %zd bytes for buffer stack %d\n",
- buffer_stack_bytes[kind], kind);
+ md->buffer_stack_bytes[kind], kind);
return -ENOMEM;
}
/* Initialize the buffer stack. */
- rc = gxio_mpipe_init_buffer_stack(&context, stack_idx,
- buffer_size_enums[kind],
- va, buffer_stack_bytes[kind], 0);
+ rc = gxio_mpipe_init_buffer_stack(&md->context, stack_idx,
+ buffer_size_enums[kind], va,
+ md->buffer_stack_bytes[kind], 0);
if (rc != 0) {
- netdev_err(dev, "gxio_mpipe_init_buffer_stack: %d\n", rc);
- free_pages_exact(va, buffer_stack_bytes[kind]);
+ netdev_err(dev, "gxio_mpipe_init_buffer_stack: mpipe[%d] %d\n",
+ instance, rc);
+ free_pages_exact(va, md->buffer_stack_bytes[kind]);
return rc;
}
- buffer_stack_vas[kind] = va;
+ md->buffer_stack_vas[kind] = va;
- rc = gxio_mpipe_register_client_memory(&context, stack_idx,
+ rc = gxio_mpipe_register_client_memory(&md->context, stack_idx,
hash_pte, 0);
if (rc != 0) {
- netdev_err(dev, "gxio_mpipe_register_client_memory: %d\n", rc);
+ netdev_err(dev,
+ "gxio_mpipe_register_client_memory: mpipe[%d] %d\n",
+ instance, rc);
return rc;
}
/* Provide initial buffers. */
for (i = 0; i < num_buffers; i++) {
- if (!tile_net_provide_buffer(kind)) {
+ if (!tile_net_provide_buffer(instance, kind)) {
netdev_err(dev, "Cannot allocate initial sk_bufs!\n");
return -ENOMEM;
}
@@ -793,14 +864,18 @@ static int init_buffer_stacks(struct net_device *dev,
int num_kinds = MAX_KINDS - (jumbo_num == 0);
size_t num_buffers;
int rc;
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
/* Allocate the buffer stacks. */
- rc = gxio_mpipe_alloc_buffer_stacks(&context, num_kinds, 0, 0);
+ rc = gxio_mpipe_alloc_buffer_stacks(&md->context, num_kinds, 0, 0);
if (rc < 0) {
- netdev_err(dev, "gxio_mpipe_alloc_buffer_stacks: %d\n", rc);
+ netdev_err(dev,
+ "gxio_mpipe_alloc_buffer_stacks: mpipe[%d] %d\n",
+ instance, rc);
return rc;
}
- first_buffer_stack = rc;
+ md->first_buffer_stack = rc;
/* Enough small/large buffers to (normally) avoid buffer errors. */
num_buffers =
@@ -829,6 +904,8 @@ static int alloc_percpu_mpipe_resources(struct net_device *dev,
{
struct tile_net_info *info = &per_cpu(per_cpu_info, cpu);
int order, i, rc;
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
struct page *page;
void *addr;
@@ -843,7 +920,7 @@ static int alloc_percpu_mpipe_resources(struct net_device *dev,
addr = pfn_to_kaddr(page_to_pfn(page));
memset(addr, 0, COMPS_SIZE);
for (i = 0; i < TILE_NET_CHANNELS; i++)
- info->comps_for_echannel[i] =
+ info->mpipe[instance].comps_for_echannel[i] =
addr + i * sizeof(struct tile_net_comps);
/* If this is a network cpu, create an iqueue. */
@@ -857,14 +934,15 @@ static int alloc_percpu_mpipe_resources(struct net_device *dev,
return -ENOMEM;
}
addr = pfn_to_kaddr(page_to_pfn(page));
- rc = gxio_mpipe_iqueue_init(&info->iqueue, &context, ring++,
- addr, NOTIF_RING_SIZE, 0);
+ rc = gxio_mpipe_iqueue_init(&info->mpipe[instance].iqueue,
+ &md->context, ring++, addr,
+ NOTIF_RING_SIZE, 0);
if (rc < 0) {
netdev_err(dev,
"gxio_mpipe_iqueue_init failed: %d\n", rc);
return rc;
}
- info->has_iqueue = true;
+ info->mpipe[instance].has_iqueue = true;
}
return ring;
@@ -877,40 +955,41 @@ static int init_notif_group_and_buckets(struct net_device *dev,
int ring, int network_cpus_count)
{
int group, rc;
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
/* Allocate one NotifGroup. */
- rc = gxio_mpipe_alloc_notif_groups(&context, 1, 0, 0);
+ rc = gxio_mpipe_alloc_notif_groups(&md->context, 1, 0, 0);
if (rc < 0) {
- netdev_err(dev, "gxio_mpipe_alloc_notif_groups failed: %d\n",
- rc);
+ netdev_err(dev, "gxio_mpipe_alloc_notif_groups: mpipe[%d] %d\n",
+ instance, rc);
return rc;
}
group = rc;
/* Initialize global num_buckets value. */
if (network_cpus_count > 4)
- num_buckets = 256;
+ md->num_buckets = 256;
else if (network_cpus_count > 1)
- num_buckets = 16;
+ md->num_buckets = 16;
/* Allocate some buckets, and set global first_bucket value. */
- rc = gxio_mpipe_alloc_buckets(&context, num_buckets, 0, 0);
+ rc = gxio_mpipe_alloc_buckets(&md->context, md->num_buckets, 0, 0);
if (rc < 0) {
- netdev_err(dev, "gxio_mpipe_alloc_buckets failed: %d\n", rc);
+ netdev_err(dev, "gxio_mpipe_alloc_buckets: mpipe[%d] %d\n",
+ instance, rc);
return rc;
}
- first_bucket = rc;
+ md->first_bucket = rc;
/* Init group and buckets. */
rc = gxio_mpipe_init_notif_group_and_buckets(
- &context, group, ring, network_cpus_count,
- first_bucket, num_buckets,
+ &md->context, group, ring, network_cpus_count,
+ md->first_bucket, md->num_buckets,
GXIO_MPIPE_BUCKET_STICKY_FLOW_LOCALITY);
if (rc != 0) {
- netdev_err(
- dev,
- "gxio_mpipe_init_notif_group_and_buckets failed: %d\n",
- rc);
+ netdev_err(dev, "gxio_mpipe_init_notif_group_and_buckets: "
+ "mpipe[%d] %d\n", instance, rc);
return rc;
}
@@ -924,30 +1003,39 @@ static int init_notif_group_and_buckets(struct net_device *dev,
*/
static int tile_net_setup_interrupts(struct net_device *dev)
{
- int cpu, rc;
+ int cpu, rc, irq;
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
+
+ irq = md->ingress_irq;
+ if (irq < 0) {
+ irq = create_irq();
+ if (irq < 0) {
+ netdev_err(dev,
+ "create_irq failed: mpipe[%d] %d\n",
+ instance, irq);
+ return irq;
+ }
+ tile_irq_activate(irq, TILE_IRQ_PERCPU);
- rc = create_irq();
- if (rc < 0) {
- netdev_err(dev, "create_irq failed: %d\n", rc);
- return rc;
- }
- ingress_irq = rc;
- tile_irq_activate(ingress_irq, TILE_IRQ_PERCPU);
- rc = request_irq(ingress_irq, tile_net_handle_ingress_irq,
- 0, "tile_net", NULL);
- if (rc != 0) {
- netdev_err(dev, "request_irq failed: %d\n", rc);
- destroy_irq(ingress_irq);
- ingress_irq = -1;
- return rc;
+ rc = request_irq(irq, tile_net_handle_ingress_irq,
+ 0, "tile_net", (void *)((uint64_t)instance));
+
+ if (rc != 0) {
+ netdev_err(dev, "request_irq failed: mpipe[%d] %d\n",
+ instance, rc);
+ destroy_irq(irq);
+ return rc;
+ }
+ md->ingress_irq = irq;
}
for_each_online_cpu(cpu) {
struct tile_net_info *info = &per_cpu(per_cpu_info, cpu);
- if (info->has_iqueue) {
- gxio_mpipe_request_notif_ring_interrupt(
- &context, cpu_x(cpu), cpu_y(cpu),
- KERNEL_PL, ingress_irq, info->iqueue.ring);
+ if (info->mpipe[instance].has_iqueue) {
+ gxio_mpipe_request_notif_ring_interrupt(&md->context,
+ cpu_x(cpu), cpu_y(cpu), KERNEL_PL, irq,
+ info->mpipe[instance].iqueue.ring);
}
}
@@ -955,40 +1043,45 @@ static int tile_net_setup_interrupts(struct net_device *dev)
}
/* Undo any state set up partially by a failed call to tile_net_init_mpipe. */
-static void tile_net_init_mpipe_fail(void)
+static void tile_net_init_mpipe_fail(int instance)
{
int kind, cpu;
+ struct mpipe_data *md = &mpipe_data[instance];
/* Do cleanups that require the mpipe context first. */
for (kind = 0; kind < MAX_KINDS; kind++) {
- if (buffer_stack_vas[kind] != NULL) {
- tile_net_pop_all_buffers(first_buffer_stack + kind);
+ if (md->buffer_stack_vas[kind] != NULL) {
+ tile_net_pop_all_buffers(instance,
+ md->first_buffer_stack +
+ kind);
}
}
/* Destroy mpipe context so the hardware no longer owns any memory. */
- gxio_mpipe_destroy(&context);
+ gxio_mpipe_destroy(&md->context);
for_each_online_cpu(cpu) {
struct tile_net_info *info = &per_cpu(per_cpu_info, cpu);
- free_pages((unsigned long)(info->comps_for_echannel[0]),
- get_order(COMPS_SIZE));
- info->comps_for_echannel[0] = NULL;
- free_pages((unsigned long)(info->iqueue.idescs),
+ free_pages(
+ (unsigned long)(
+ info->mpipe[instance].comps_for_echannel[0]),
+ get_order(COMPS_SIZE));
+ info->mpipe[instance].comps_for_echannel[0] = NULL;
+ free_pages((unsigned long)(info->mpipe[instance].iqueue.idescs),
get_order(NOTIF_RING_SIZE));
- info->iqueue.idescs = NULL;
+ info->mpipe[instance].iqueue.idescs = NULL;
}
for (kind = 0; kind < MAX_KINDS; kind++) {
- if (buffer_stack_vas[kind] != NULL) {
- free_pages_exact(buffer_stack_vas[kind],
- buffer_stack_bytes[kind]);
- buffer_stack_vas[kind] = NULL;
+ if (md->buffer_stack_vas[kind] != NULL) {
+ free_pages_exact(md->buffer_stack_vas[kind],
+ md->buffer_stack_bytes[kind]);
+ md->buffer_stack_vas[kind] = NULL;
}
}
- first_buffer_stack = -1;
- first_bucket = -1;
+ md->first_buffer_stack = -1;
+ md->first_bucket = -1;
}
/* The first time any tilegx network device is opened, we initialize
@@ -1005,6 +1098,8 @@ static int tile_net_init_mpipe(struct net_device *dev)
int rc;
int cpu;
int first_ring, ring;
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
int network_cpus_count = cpus_weight(network_cpus_map);
if (!hash_default) {
@@ -1012,9 +1107,10 @@ static int tile_net_init_mpipe(struct net_device *dev)
return -EIO;
}
- rc = gxio_mpipe_init(&context, 0);
+ rc = gxio_mpipe_init(&md->context, instance);
if (rc != 0) {
- netdev_err(dev, "gxio_mpipe_init failed: %d\n", rc);
+ netdev_err(dev, "gxio_mpipe_init: mpipe[%d] %d\n",
+ instance, rc);
return -EIO;
}
@@ -1024,7 +1120,8 @@ static int tile_net_init_mpipe(struct net_device *dev)
goto fail;
/* Allocate one NotifRing for each network cpu. */
- rc = gxio_mpipe_alloc_notif_rings(&context, network_cpus_count, 0, 0);
+ rc = gxio_mpipe_alloc_notif_rings(&md->context,
+ network_cpus_count, 0, 0);
if (rc < 0) {
netdev_err(dev, "gxio_mpipe_alloc_notif_rings failed %d\n",
rc);
@@ -1054,7 +1151,7 @@ static int tile_net_init_mpipe(struct net_device *dev)
return 0;
fail:
- tile_net_init_mpipe_fail();
+ tile_net_init_mpipe_fail(instance);
return rc;
}
@@ -1072,9 +1169,11 @@ static int tile_net_init_egress(struct net_device *dev, int echannel)
int headers_order, edescs_order, equeue_order;
size_t edescs_size;
int rc = -ENOMEM;
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
/* Only initialize once. */
- if (egress_for_echannel[echannel].equeue != NULL)
+ if (md->egress_for_echannel[echannel].equeue != NULL)
return 0;
/* Allocate memory for the "headers". */
@@ -1113,20 +1212,21 @@ static int tile_net_init_egress(struct net_device *dev, int echannel)
/* Allocate an edma ring (using a one entry "free list"). */
if (ering < 0) {
- rc = gxio_mpipe_alloc_edma_rings(&context, 1, 0, 0);
+ rc = gxio_mpipe_alloc_edma_rings(&md->context, 1, 0, 0);
if (rc < 0) {
- netdev_warn(dev, "gxio_mpipe_alloc_edma_rings: %d\n",
- rc);
+ netdev_warn(dev, "gxio_mpipe_alloc_edma_rings: "
+ "mpipe[%d] %d\n", instance, rc);
goto fail_equeue;
}
ering = rc;
}
/* Initialize the equeue. */
- rc = gxio_mpipe_equeue_init(equeue, &context, ering, echannel,
+ rc = gxio_mpipe_equeue_init(equeue, &md->context, ering, echannel,
edescs, edescs_size, 0);
if (rc != 0) {
- netdev_err(dev, "gxio_mpipe_equeue_init failed: %d\n", rc);
+ netdev_err(dev, "gxio_mpipe_equeue_init: mpipe[%d] %d\n",
+ instance, rc);
goto fail_equeue;
}
@@ -1143,8 +1243,8 @@ static int tile_net_init_egress(struct net_device *dev, int echannel)
}
/* Done. */
- egress_for_echannel[echannel].equeue = equeue;
- egress_for_echannel[echannel].headers = headers;
+ md->egress_for_echannel[echannel].equeue = equeue;
+ md->egress_for_echannel[echannel].headers = headers;
return 0;
fail_equeue:
@@ -1164,9 +1264,12 @@ fail:
static int tile_net_link_open(struct net_device *dev, gxio_mpipe_link_t *link,
const char *link_name)
{
- int rc = gxio_mpipe_link_open(link, &context, link_name, 0);
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
+ int rc = gxio_mpipe_link_open(link, &md->context, link_name, 0);
if (rc < 0) {
- netdev_err(dev, "Failed to open '%s'\n", link_name);
+ netdev_err(dev, "Failed to open '%s', mpipe[%d], %d\n",
+ link_name, instance, rc);
return rc;
}
if (jumbo_num != 0) {
@@ -1193,12 +1296,21 @@ static int tile_net_link_open(struct net_device *dev, gxio_mpipe_link_t *link,
static int tile_net_open(struct net_device *dev)
{
struct tile_net_priv *priv = netdev_priv(dev);
- int cpu, rc;
+ int cpu, rc, instance;
mutex_lock(&tile_net_devs_for_channel_mutex);
- /* Do one-time initialization the first time any device is opened. */
- if (ingress_irq < 0) {
+ /* Get the instance info. */
+ rc = gxio_mpipe_link_instance(dev->name);
+ if (rc < 0 || rc >= NR_MPIPE_MAX)
+ return -EIO;
+
+ priv->instance = rc;
+ instance = rc;
+ if (!mpipe_data[rc].context.mmio_fast_base) {
+ /* Do one-time initialization per instance the first time
+ * any device is opened.
+ */
rc = tile_net_init_mpipe(dev);
if (rc != 0)
goto fail;
@@ -1229,7 +1341,7 @@ static int tile_net_open(struct net_device *dev)
if (rc != 0)
goto fail;
- tile_net_devs_for_channel[priv->channel] = dev;
+ mpipe_data[instance].tile_net_devs_for_channel[priv->channel] = dev;
rc = tile_net_update(dev);
if (rc != 0)
@@ -1241,7 +1353,7 @@ static int tile_net_open(struct net_device *dev)
for_each_online_cpu(cpu) {
struct tile_net_info *info = &per_cpu(per_cpu_info, cpu);
struct tile_net_tx_wake *tx_wake =
- &info->tx_wake[priv->echannel];
+ &info->mpipe[instance].tx_wake[priv->echannel];
hrtimer_init(&tx_wake->timer, CLOCK_MONOTONIC,
HRTIMER_MODE_REL);
@@ -1267,7 +1379,7 @@ fail:
priv->channel = -1;
}
priv->echannel = -1;
- tile_net_devs_for_channel[priv->channel] = NULL;
+ mpipe_data[instance].tile_net_devs_for_channel[priv->channel] = NULL;
mutex_unlock(&tile_net_devs_for_channel_mutex);
/* Don't return raw gxio error codes to generic Linux. */
@@ -1279,18 +1391,20 @@ static int tile_net_stop(struct net_device *dev)
{
struct tile_net_priv *priv = netdev_priv(dev);
int cpu;
+ int instance = priv->instance;
+ struct mpipe_data *md = &mpipe_data[instance];
for_each_online_cpu(cpu) {
struct tile_net_info *info = &per_cpu(per_cpu_info, cpu);
struct tile_net_tx_wake *tx_wake =
- &info->tx_wake[priv->echannel];
+ &info->mpipe[instance].tx_wake[priv->echannel];
hrtimer_cancel(&tx_wake->timer);
netif_stop_subqueue(dev, cpu);
}
mutex_lock(&tile_net_devs_for_channel_mutex);
- tile_net_devs_for_channel[priv->channel] = NULL;
+ md->tile_net_devs_for_channel[priv->channel] = NULL;
(void)tile_net_update(dev);
if (priv->loopify_channel >= 0) {
if (gxio_mpipe_link_close(&priv->loopify_link) != 0)
@@ -1500,6 +1614,8 @@ static void tso_egress(struct net_device *dev, gxio_mpipe_equeue_t *equeue,
struct sk_buff *skb, unsigned char *headers, s64 slot)
{
struct skb_shared_info *sh = skb_shinfo(skb);
+ int instance = mpipe_instance(dev);
+ struct mpipe_data *md = &mpipe_data[instance];
unsigned int sh_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
unsigned int data_len = skb->len - sh_len;
unsigned int p_len = sh->gso_size;
@@ -1522,8 +1638,8 @@ static void tso_egress(struct net_device *dev, gxio_mpipe_equeue_t *equeue,
edesc_head.xfer_size = sh_len;
/* This is only used to specify the TLB. */
- edesc_head.stack_idx = first_buffer_stack;
- edesc_body.stack_idx = first_buffer_stack;
+ edesc_head.stack_idx = md->first_buffer_stack;
+ edesc_body.stack_idx = md->first_buffer_stack;
/* Egress all the edescs. */
for (segment = 0; segment < sh->gso_segs; segment++) {
@@ -1598,8 +1714,11 @@ static int tile_net_tx_tso(struct sk_buff *skb, struct net_device *dev)
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
struct tile_net_priv *priv = netdev_priv(dev);
int channel = priv->echannel;
- struct tile_net_egress *egress = &egress_for_echannel[channel];
- struct tile_net_comps *comps = info->comps_for_echannel[channel];
+ int instance = priv->instance;
+ struct mpipe_data *md = &mpipe_data[instance];
+ struct tile_net_egress *egress = &md->egress_for_echannel[channel];
+ struct tile_net_comps *comps =
+ info->mpipe[instance].comps_for_echannel[channel];
gxio_mpipe_equeue_t *equeue = egress->equeue;
unsigned long irqflags;
int num_edescs;
@@ -1663,10 +1782,13 @@ static int tile_net_tx(struct sk_buff *skb, struct net_device *dev)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
struct tile_net_priv *priv = netdev_priv(dev);
- struct tile_net_egress *egress = &egress_for_echannel[priv->echannel];
+ int instance = priv->instance;
+ struct mpipe_data *md = &mpipe_data[instance];
+ struct tile_net_egress *egress =
+ &md->egress_for_echannel[priv->echannel];
gxio_mpipe_equeue_t *equeue = egress->equeue;
struct tile_net_comps *comps =
- info->comps_for_echannel[priv->echannel];
+ info->mpipe[instance].comps_for_echannel[priv->echannel];
unsigned int len = skb->len;
unsigned char *data = skb->data;
unsigned int num_edescs;
@@ -1683,7 +1805,7 @@ static int tile_net_tx(struct sk_buff *skb, struct net_device *dev)
num_edescs = tile_net_tx_frags(frags, skb, data, skb_headlen(skb));
/* This is only used to specify the TLB. */
- edesc.stack_idx = first_buffer_stack;
+ edesc.stack_idx = md->first_buffer_stack;
/* Prepare the edescs. */
for (i = 0; i < num_edescs; i++) {
@@ -1790,9 +1912,13 @@ static int tile_net_set_mac_address(struct net_device *dev, void *p)
*/
static void tile_net_netpoll(struct net_device *dev)
{
- disable_percpu_irq(ingress_irq);
- tile_net_handle_ingress_irq(ingress_irq, NULL);
- enable_percpu_irq(ingress_irq, 0);
+ int instance = mpipe_instance(dev);
+ struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
+ struct mpipe_data *md = &mpipe_data[instance];
+
+ disable_percpu_irq(md->ingress_irq);
+ napi_schedule(&info->mpipe[instance].napi);
+ enable_percpu_irq(md->ingress_irq, 0);
}
#endif
@@ -1895,9 +2021,12 @@ static void tile_net_init_module_percpu(void *unused)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
int my_cpu = smp_processor_id();
+ int instance;
- info->has_iqueue = false;
-
+ for (instance = 0; instance < NR_MPIPE_MAX; instance++) {
+ info->mpipe[instance].has_iqueue = false;
+ info->mpipe[instance].instance = instance;
+ }
info->my_cpu = my_cpu;
/* Initialize the egress timer. */
@@ -1914,6 +2043,8 @@ static int __init tile_net_init_module(void)
pr_info("Tilera Network Driver\n");
+ BUILD_BUG_ON(NR_MPIPE_MAX != 2);
+
mutex_init(&tile_net_devs_for_channel_mutex);
/* Initialize each CPU. */
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 12/13] tile: remove deprecated NETIF_F_LLTX flag from tile drivers
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilegx.c | 1 -
drivers/net/ethernet/tile/tilepro.c | 1 -
2 files changed, 2 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index 7302e84..0beed7e 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -1963,7 +1963,6 @@ static void tile_net_setup(struct net_device *dev)
dev->watchdog_timeo = TILE_NET_TIMEOUT;
dev->mtu = 1500;
- features |= NETIF_F_LLTX;
features |= NETIF_F_HW_CSUM;
features |= NETIF_F_SG;
features |= NETIF_F_TSO;
diff --git a/drivers/net/ethernet/tile/tilepro.c b/drivers/net/ethernet/tile/tilepro.c
index fd17af5..106be47 100644
--- a/drivers/net/ethernet/tile/tilepro.c
+++ b/drivers/net/ethernet/tile/tilepro.c
@@ -2253,7 +2253,6 @@ static void tile_net_setup(struct net_device *dev)
dev->tx_queue_len = TILE_NET_TX_QUEUE_LEN;
dev->mtu = TILE_NET_MTU;
- features |= NETIF_F_LLTX;
features |= NETIF_F_HW_CSUM;
features |= NETIF_F_SG;
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 13/13] tile: support PTP using the tilegx mPIPE (IEEE 1588)
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
---
arch/tile/gxio/iorpc_mpipe.c | 19 +++
arch/tile/include/gxio/iorpc_mpipe.h | 10 +-
arch/tile/include/gxio/mpipe.h | 14 +++
drivers/net/ethernet/tile/Kconfig | 11 ++
drivers/net/ethernet/tile/tilegx.c | 216 ++++++++++++++++++++++++++++++++++-
5 files changed, 266 insertions(+), 4 deletions(-)
diff --git a/arch/tile/gxio/iorpc_mpipe.c b/arch/tile/gxio/iorpc_mpipe.c
index c2fb1516..4f8f3d6 100644
--- a/arch/tile/gxio/iorpc_mpipe.c
+++ b/arch/tile/gxio/iorpc_mpipe.c
@@ -475,6 +475,25 @@ int gxio_mpipe_adjust_timestamp_aux(gxio_mpipe_context_t * context,
EXPORT_SYMBOL(gxio_mpipe_adjust_timestamp_aux);
+struct adjust_timestamp_freq_param {
+ int32_t ppb;
+};
+
+int gxio_mpipe_adjust_timestamp_freq(gxio_mpipe_context_t * context,
+ int32_t ppb)
+{
+ struct adjust_timestamp_freq_param temp;
+ struct adjust_timestamp_freq_param *params = &temp;
+
+ params->ppb = ppb;
+
+ return hv_dev_pwrite(context->fd, 0, (HV_VirtAddr) params,
+ sizeof(*params),
+ GXIO_MPIPE_OP_ADJUST_TIMESTAMP_FREQ);
+}
+
+EXPORT_SYMBOL(gxio_mpipe_adjust_timestamp_freq);
+
struct config_edma_ring_blks_param {
unsigned int ering;
unsigned int max_blks;
diff --git a/arch/tile/include/gxio/iorpc_mpipe.h b/arch/tile/include/gxio/iorpc_mpipe.h
index eef60fd..fdd07f8 100644
--- a/arch/tile/include/gxio/iorpc_mpipe.h
+++ b/arch/tile/include/gxio/iorpc_mpipe.h
@@ -46,10 +46,11 @@
#define GXIO_MPIPE_OP_LINK_CLOSE_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1212)
#define GXIO_MPIPE_OP_LINK_SET_ATTR_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1213)
-#define GXIO_MPIPE_OP_GET_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x121e)
-#define GXIO_MPIPE_OP_SET_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x121f)
-#define GXIO_MPIPE_OP_ADJUST_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x1220)
+#define GXIO_MPIPE_OP_GET_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x121e)
+#define GXIO_MPIPE_OP_SET_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x121f)
+#define GXIO_MPIPE_OP_ADJUST_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1220)
#define GXIO_MPIPE_OP_CONFIG_EDMA_RING_BLKS IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1221)
+#define GXIO_MPIPE_OP_ADJUST_TIMESTAMP_FREQ IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1222)
#define GXIO_MPIPE_OP_ARM_POLLFD IORPC_OPCODE(IORPC_FORMAT_KERNEL_POLLFD, 0x9000)
#define GXIO_MPIPE_OP_CLOSE_POLLFD IORPC_OPCODE(IORPC_FORMAT_KERNEL_POLLFD, 0x9001)
#define GXIO_MPIPE_OP_GET_MMIO_BASE IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x8000)
@@ -128,6 +129,9 @@ int gxio_mpipe_set_timestamp_aux(gxio_mpipe_context_t * context, uint64_t sec,
int gxio_mpipe_adjust_timestamp_aux(gxio_mpipe_context_t * context,
int64_t nsec);
+int gxio_mpipe_adjust_timestamp_freq(gxio_mpipe_context_t * context,
+ int32_t ppb);
+
int gxio_mpipe_arm_pollfd(gxio_mpipe_context_t * context, int pollfd_cookie);
int gxio_mpipe_close_pollfd(gxio_mpipe_context_t * context, int pollfd_cookie);
diff --git a/arch/tile/include/gxio/mpipe.h b/arch/tile/include/gxio/mpipe.h
index eb7fee4..e37cf4f 100644
--- a/arch/tile/include/gxio/mpipe.h
+++ b/arch/tile/include/gxio/mpipe.h
@@ -1854,4 +1854,18 @@ extern int gxio_mpipe_set_timestamp(gxio_mpipe_context_t *context,
extern int gxio_mpipe_adjust_timestamp(gxio_mpipe_context_t *context,
int64_t delta);
+/** Adjust the mPIPE timestamp clock frequency.
+ *
+ * @param context An initialized mPIPE context.
+ * @param ppb A 32-bit signed PPB (Parts Per Billion) value to adjust.
+ * The absolute value of ppb must be less than or equal to 1000000000.
+ * Values less than about 30000 will generally cause a GXIO_ERR_INVAL
+ * return due to the granularity of the hardware that converts reference
+ * clock cycles into seconds and nanoseconds.
+ * @return If the call was successful, zero; otherwise, a negative error
+ * code.
+ */
+extern int gxio_mpipe_adjust_timestamp_freq(gxio_mpipe_context_t* context,
+ int32_t ppb);
+
#endif /* !_GXIO_MPIPE_H_ */
diff --git a/drivers/net/ethernet/tile/Kconfig b/drivers/net/ethernet/tile/Kconfig
index 098b1c4..4083ba8 100644
--- a/drivers/net/ethernet/tile/Kconfig
+++ b/drivers/net/ethernet/tile/Kconfig
@@ -15,3 +15,14 @@ config TILE_NET
To compile this driver as a module, choose M here: the module
will be called tile_net.
+
+config PTP_1588_CLOCK_TILEGX
+ tristate "Tilera TILE-Gx mPIPE as PTP clock"
+ select PTP_1588_CLOCK
+ depends on TILE_NET
+ depends on TILEGX
+ ---help---
+ This driver adds support for using the mPIPE as a PTP
+ clock. This clock is only useful if your PTP programs are
+ getting hardware time stamps on the PTP Ethernet packets
+ using the SO_TIMESTAMPING API.
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index 0beed7e..907b577 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -38,6 +38,8 @@
#include <linux/ip.h>
#include <linux/ipv6.h>
#include <linux/tcp.h>
+#include <linux/net_tstamp.h>
+#include <linux/ptp_clock_kernel.h>
#include <asm/checksum.h>
#include <asm/homecache.h>
@@ -185,6 +187,10 @@ struct tile_net_priv {
int echannel;
/* mPIPE instance, 0 or 1. */
int instance;
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+ /* The timestamp config. */
+ struct hwtstamp_config stamp_cfg;
+#endif
};
static struct mpipe_data {
@@ -223,6 +229,15 @@ static struct mpipe_data {
int first_bucket;
int num_buckets;
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+ /* PTP-specific data. */
+ struct ptp_clock *ptp_clock;
+ struct ptp_clock_info caps;
+
+ /* Lock for ptp accessors. */
+ struct mutex ptp_lock;
+#endif
+
} mpipe_data[NR_MPIPE_MAX] = {
[0 ... (NR_MPIPE_MAX - 1)] {
.ingress_irq = -1,
@@ -432,6 +447,94 @@ static void tile_net_provide_needed_buffers(void)
}
}
+/* Get RX timestamp, and store it in the skb. */
+static void tile_rx_timestamp(struct tile_net_priv *priv, struct sk_buff *skb,
+ gxio_mpipe_idesc_t *idesc)
+{
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+ if (unlikely(priv->stamp_cfg.rx_filter != HWTSTAMP_FILTER_NONE)) {
+ struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
+ memset(shhwtstamps, 0, sizeof(*shhwtstamps));
+ shhwtstamps->hwtstamp = ktime_set(idesc->time_stamp_sec,
+ idesc->time_stamp_ns);
+ }
+#endif
+}
+
+/* Get TX timestamp, and store it in the skb. */
+static void tile_tx_timestamp(struct sk_buff *skb, int instance)
+{
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+ struct skb_shared_info *shtx = skb_shinfo(skb);
+ if (unlikely((shtx->tx_flags & SKBTX_HW_TSTAMP) != 0)) {
+ struct mpipe_data *md = &mpipe_data[instance];
+ struct skb_shared_hwtstamps shhwtstamps;
+ struct timespec ts;
+
+ shtx->tx_flags |= SKBTX_IN_PROGRESS;
+ gxio_mpipe_get_timestamp(&md->context, &ts);
+ memset(&shhwtstamps, 0, sizeof(shhwtstamps));
+ shhwtstamps.hwtstamp = ktime_set(ts.tv_sec, ts.tv_nsec);
+ skb_tstamp_tx(skb, &shhwtstamps);
+ }
+#endif
+}
+
+/* Use ioctl() to enable or disable TX or RX timestamping. */
+static int tile_hwtstamp_ioctl(struct net_device *dev, struct ifreq *rq,
+ int cmd)
+{
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+ struct hwtstamp_config config;
+ struct tile_net_priv *priv = netdev_priv(dev);
+
+ if (copy_from_user(&config, rq->ifr_data, sizeof(config)))
+ return -EFAULT;
+
+ if (config.flags) /* reserved for future extensions */
+ return -EINVAL;
+
+ switch (config.tx_type) {
+ case HWTSTAMP_TX_OFF:
+ case HWTSTAMP_TX_ON:
+ break;
+ default:
+ return -ERANGE;
+ }
+
+ switch (config.rx_filter) {
+ case HWTSTAMP_FILTER_NONE:
+ break;
+ case HWTSTAMP_FILTER_ALL:
+ case HWTSTAMP_FILTER_SOME:
+ case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
+ config.rx_filter = HWTSTAMP_FILTER_ALL;
+ break;
+ default:
+ return -ERANGE;
+ }
+
+ if (copy_to_user(rq->ifr_data, &config, sizeof(config)))
+ return -EFAULT;
+
+ priv->stamp_cfg = config;
+ return 0;
+#else
+ return -EOPNOTSUPP;
+#endif
+}
+
static inline bool filter_packet(struct net_device *dev, void *buf)
{
/* Filter packets received before we're up. */
@@ -451,7 +554,8 @@ static void tile_net_receive_skb(struct net_device *dev, struct sk_buff *skb,
gxio_mpipe_idesc_t *idesc, unsigned long len)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
- int instance = mpipe_instance(dev);
+ struct tile_net_priv *priv = netdev_priv(dev);
+ int instance = priv->instance;
/* Encode the actual packet length. */
skb_put(skb, len);
@@ -462,6 +566,9 @@ static void tile_net_receive_skb(struct net_device *dev, struct sk_buff *skb,
if (idesc->cs && idesc->csum_seed_val == 0xFFFF)
skb->ip_summed = CHECKSUM_UNNECESSARY;
+ /* Get RX timestamp from idesc. */
+ tile_rx_timestamp(priv, skb, idesc);
+
napi_gro_receive(&info->mpipe[instance].napi, skb);
/* Update stats. */
@@ -707,6 +814,103 @@ static enum hrtimer_restart tile_net_handle_egress_timer(struct hrtimer *t)
return HRTIMER_NORESTART;
}
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+
+/* PTP clock operations. */
+
+static int ptp_mpipe_adjfreq(struct ptp_clock_info *ptp, s32 ppb)
+{
+ int ret = 0;
+ struct mpipe_data *md = container_of(ptp, struct mpipe_data, caps);
+ mutex_lock(&md->ptp_lock);
+ if (gxio_mpipe_adjust_timestamp_freq(&md->context, ppb))
+ ret = -EINVAL;
+ mutex_unlock(&md->ptp_lock);
+ return ret;
+}
+
+static int ptp_mpipe_adjtime(struct ptp_clock_info *ptp, s64 delta)
+{
+ int ret = 0;
+ struct mpipe_data *md = container_of(ptp, struct mpipe_data, caps);
+ mutex_lock(&md->ptp_lock);
+ if (gxio_mpipe_adjust_timestamp(&md->context, delta))
+ ret = -EBUSY;
+ mutex_unlock(&md->ptp_lock);
+ return ret;
+}
+
+static int ptp_mpipe_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+{
+ int ret = 0;
+ struct mpipe_data *md = container_of(ptp, struct mpipe_data, caps);
+ mutex_lock(&md->ptp_lock);
+ if (gxio_mpipe_get_timestamp(&md->context, ts))
+ ret = -EBUSY;
+ mutex_unlock(&md->ptp_lock);
+ return ret;
+}
+
+static int ptp_mpipe_settime(struct ptp_clock_info *ptp,
+ const struct timespec *ts)
+{
+ int ret = 0;
+ struct mpipe_data *md = container_of(ptp, struct mpipe_data, caps);
+ mutex_lock(&md->ptp_lock);
+ if (gxio_mpipe_set_timestamp(&md->context, ts))
+ ret = -EBUSY;
+ mutex_unlock(&md->ptp_lock);
+ return ret;
+}
+
+static int ptp_mpipe_enable(struct ptp_clock_info *ptp,
+ struct ptp_clock_request *request, int on)
+{
+ return -EOPNOTSUPP;
+}
+
+static struct ptp_clock_info ptp_mpipe_caps = {
+ .owner = THIS_MODULE,
+ .name = "mPIPE clock",
+ .max_adj = 999999999,
+ .n_ext_ts = 0,
+ .pps = 0,
+ .adjfreq = ptp_mpipe_adjfreq,
+ .adjtime = ptp_mpipe_adjtime,
+ .gettime = ptp_mpipe_gettime,
+ .settime = ptp_mpipe_settime,
+ .enable = ptp_mpipe_enable,
+};
+
+#endif /* CONFIG_PTP_1588_CLOCK_TILEGX */
+
+/* Sync mPIPE's timestamp up with Linux system time and register PTP clock. */
+static void register_ptp_clock(struct net_device *dev, struct mpipe_data *md)
+{
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+ struct timespec ts;
+
+ getnstimeofday(&ts);
+ gxio_mpipe_set_timestamp(&md->context, &ts);
+
+ mutex_init(&md->ptp_lock);
+ md->caps = ptp_mpipe_caps;
+ md->ptp_clock = ptp_clock_register(&md->caps, NULL);
+ if (IS_ERR(md->ptp_clock))
+ netdev_err(dev, "ptp_clock_register failed %ld\n",
+ PTR_ERR(md->ptp_clock));
+#endif
+}
+
+/* Initialize PTP fields in a new device. */
+static void init_ptp_dev(struct tile_net_priv *priv)
+{
+#ifdef CONFIG_PTP_1588_CLOCK_TILEGX
+ priv->stamp_cfg.rx_filter = HWTSTAMP_FILTER_NONE;
+ priv->stamp_cfg.tx_type = HWTSTAMP_TX_OFF;
+#endif
+}
+
/* Helper functions for "tile_net_update()". */
static void enable_ingress_irq(void *irq)
{
@@ -1149,6 +1353,9 @@ static int tile_net_init_mpipe(struct net_device *dev)
if (rc != 0)
goto fail;
+ /* Register PTP clock and set mPIPE timestamp, if configured. */
+ register_ptp_clock(dev, md);
+
return 0;
fail:
@@ -1851,6 +2058,9 @@ static int tile_net_tx(struct sk_buff *skb, struct net_device *dev)
for (i = 0; i < num_edescs; i++)
gxio_mpipe_equeue_put_at(equeue, edescs[i], slot++);
+ /* Store TX timestamp if needed. */
+ tile_tx_timestamp(skb, instance);
+
/* Add a completion record. */
add_comp(equeue, comps, slot - 1, skb);
@@ -1885,6 +2095,9 @@ static void tile_net_tx_timeout(struct net_device *dev)
/* Ioctl commands. */
static int tile_net_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
{
+ if (cmd == SIOCSHWTSTAMP)
+ return tile_hwtstamp_ioctl(dev, rq, cmd);
+
return -EOPNOTSUPP;
}
@@ -2005,6 +2218,7 @@ static void tile_net_dev_init(const char *name, const uint8_t *mac)
priv->channel = -1;
priv->loopify_channel = -1;
priv->echannel = -1;
+ init_ptp_dev(priv);
/* Get the MAC address and set it in the device struct; this must
* be done before the device is opened. If the MAC is all zeroes,
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 08/13] tile: enable GRO in the tilegx network driver
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilegx.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index 17bcf33..2b1c31f 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -422,7 +422,7 @@ static void tile_net_receive_skb(struct net_device *dev, struct sk_buff *skb,
if (idesc->cs && idesc->csum_seed_val == 0xFFFF)
skb->ip_summed = CHECKSUM_UNNECESSARY;
- netif_receive_skb(skb);
+ napi_gro_receive(&info->napi, skb);
/* Update stats. */
tile_net_stats_add(1, &dev->stats.rx_packets);
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 07/13] tile: fix panic bug in napi support for tilegx network driver
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
The code used to call napi_disable() in an interrupt handler
(from smp_call_function), which in turn could call msleep().
Unfortunately you can't sleep in an interrupt context.
Luckily it turns out all the NAPI support functions are
just operating on data structures and not on any deeply
per-cpu data, so we can arrange to set up and tear down all
the NAPI state on the core driving the process, and just
do the IRQ enable/disable as a smp_call_function thing.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilegx.c | 65 +++++++++++++++++++-------------------
1 file changed, 33 insertions(+), 32 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index e9eba2b..17bcf33 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -650,37 +650,13 @@ static enum hrtimer_restart tile_net_handle_egress_timer(struct hrtimer *t)
return HRTIMER_NORESTART;
}
-/* Helper function for "tile_net_update()".
- * "dev" (i.e. arg) is the device being brought up or down,
- * or NULL if all devices are now down.
- */
-static void tile_net_update_cpu(void *arg)
+/* Helper function for "tile_net_update()". */
+static void manage_ingress_irq(void *enable)
{
- struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
- struct net_device *dev = arg;
-
- if (!info->has_iqueue)
- return;
-
- if (dev != NULL) {
- if (!info->napi_added) {
- netif_napi_add(dev, &info->napi,
- tile_net_poll, TILE_NET_WEIGHT);
- info->napi_added = true;
- }
- if (!info->napi_enabled) {
- napi_enable(&info->napi);
- info->napi_enabled = true;
- }
+ if (enable)
enable_percpu_irq(ingress_irq, 0);
- } else {
+ else
disable_percpu_irq(ingress_irq);
- if (info->napi_enabled) {
- napi_disable(&info->napi);
- info->napi_enabled = false;
- }
- /* FIXME: Drain the iqueue. */
- }
}
/* Helper function for tile_net_open() and tile_net_stop().
@@ -717,10 +693,35 @@ static int tile_net_update(struct net_device *dev)
return -EIO;
}
- /* Update all cpus, sequentially (to protect "netif_napi_add()"). */
- for_each_online_cpu(cpu)
- smp_call_function_single(cpu, tile_net_update_cpu,
- (saw_channel ? dev : NULL), 1);
+ /* Update all cpus, sequentially (to protect "netif_napi_add()").
+ * We use on_each_cpu to handle the IPI mask or unmask.
+ */
+ if (!saw_channel)
+ on_each_cpu(manage_ingress_irq, (void *)0, 1);
+ for_each_online_cpu(cpu) {
+ struct tile_net_info *info = &per_cpu(per_cpu_info, cpu);
+ if (!info->has_iqueue)
+ continue;
+ if (saw_channel) {
+ if (!info->napi_added) {
+ netif_napi_add(dev, &info->napi,
+ tile_net_poll, TILE_NET_WEIGHT);
+ info->napi_added = true;
+ }
+ if (!info->napi_enabled) {
+ napi_enable(&info->napi);
+ info->napi_enabled = true;
+ }
+ } else {
+ if (info->napi_enabled) {
+ napi_disable(&info->napi);
+ info->napi_enabled = false;
+ }
+ /* FIXME: Drain the iqueue. */
+ }
+ }
+ if (saw_channel)
+ on_each_cpu(manage_ingress_irq, (void *)1, 1);
/* HACK: Allow packets to flow in the simulator. */
if (saw_channel)
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 04/13] tile: remove dead is_dup_ack() function from tilepro net driver
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilepro.c | 76 -------------------------------------
1 file changed, 76 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilepro.c b/drivers/net/ethernet/tile/tilepro.c
index 2f4b7b9..fd17af5 100644
--- a/drivers/net/ethernet/tile/tilepro.c
+++ b/drivers/net/ethernet/tile/tilepro.c
@@ -89,9 +89,6 @@
/* ISSUE: This has not been thoroughly tested (except at 1500). */
#define TILE_NET_MTU 1500
-/* Define this to collapse "duplicate" acks. */
-/* #define IGNORE_DUP_ACKS */
-
/* HACK: Define this to verify incoming packets. */
/* #define TILE_NET_VERIFY_INGRESS */
@@ -625,79 +622,6 @@ static void tile_net_handle_egress_timer(unsigned long arg)
}
-#ifdef IGNORE_DUP_ACKS
-
-/*
- * Help detect "duplicate" ACKs. These are sequential packets (for a
- * given flow) which are exactly 66 bytes long, sharing everything but
- * ID=2@0x12, Hsum=2@0x18, Ack=4@0x2a, WinSize=2@0x30, Csum=2@0x32,
- * Tstamps=10@0x38. The ID's are +1, the Hsum's are -1, the Ack's are
- * +N, and the Tstamps are usually identical.
- *
- * NOTE: Apparently truly duplicate acks (with identical "ack" values),
- * should not be collapsed, as they are used for some kind of flow control.
- */
-static bool is_dup_ack(char *s1, char *s2, unsigned int len)
-{
- int i;
-
- unsigned long long ignorable = 0;
-
- /* Identification. */
- ignorable |= (1ULL << 0x12);
- ignorable |= (1ULL << 0x13);
-
- /* Header checksum. */
- ignorable |= (1ULL << 0x18);
- ignorable |= (1ULL << 0x19);
-
- /* ACK. */
- ignorable |= (1ULL << 0x2a);
- ignorable |= (1ULL << 0x2b);
- ignorable |= (1ULL << 0x2c);
- ignorable |= (1ULL << 0x2d);
-
- /* WinSize. */
- ignorable |= (1ULL << 0x30);
- ignorable |= (1ULL << 0x31);
-
- /* Checksum. */
- ignorable |= (1ULL << 0x32);
- ignorable |= (1ULL << 0x33);
-
- for (i = 0; i < len; i++, ignorable >>= 1) {
-
- if ((ignorable & 1) || (s1[i] == s2[i]))
- continue;
-
-#ifdef TILE_NET_DEBUG
- /* HACK: Mention non-timestamp diffs. */
- if (i < 0x38 && i != 0x2f &&
- net_ratelimit())
- pr_info("Diff at 0x%x\n", i);
-#endif
-
- return false;
- }
-
-#ifdef TILE_NET_NO_SUPPRESS_DUP_ACKS
- /* HACK: Do not suppress truly duplicate ACKs. */
- /* ISSUE: Is this actually necessary or helpful? */
- if (s1[0x2a] == s2[0x2a] &&
- s1[0x2b] == s2[0x2b] &&
- s1[0x2c] == s2[0x2c] &&
- s1[0x2d] == s2[0x2d]) {
- return false;
- }
-#endif
-
- return true;
-}
-
-#endif
-
-
-
static void tile_net_discard_aux(struct tile_net_cpu *info, int index)
{
struct tile_netio_queue *queue = &info->queue;
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 01/13] tile: set hw_features and vlan_features in setup
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
This change allows the user to configure various features of the tile
networking drivers on and off. There is no change to the default
initialization state of either the tilegx or tilepro drivers.
Neither driver needs the ndo_fix_features or ndo_set_features callbacks,
since the generic code already handles the dependencies for
fix_features, and there is no hardware state to tweak in set_features.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilegx.c | 15 +++++++++----
drivers/net/ethernet/tile/tilepro.c | 43 ++++++++++++-------------------------
2 files changed, 25 insertions(+), 33 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index f3c2d03..6085571 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -1800,14 +1800,21 @@ static const struct net_device_ops tile_net_ops = {
*/
static void tile_net_setup(struct net_device *dev)
{
+ netdev_features_t features = 0;
+
ether_setup(dev);
dev->netdev_ops = &tile_net_ops;
dev->watchdog_timeo = TILE_NET_TIMEOUT;
- dev->features |= NETIF_F_LLTX;
- dev->features |= NETIF_F_HW_CSUM;
- dev->features |= NETIF_F_SG;
- dev->features |= NETIF_F_TSO;
dev->mtu = 1500;
+
+ features |= NETIF_F_LLTX;
+ features |= NETIF_F_HW_CSUM;
+ features |= NETIF_F_SG;
+ features |= NETIF_F_TSO;
+
+ dev->hw_features |= features;
+ dev->vlan_features |= features;
+ dev->features |= features;
}
/* Allocate the device structure, register the device, and obtain the
diff --git a/drivers/net/ethernet/tile/tilepro.c b/drivers/net/ethernet/tile/tilepro.c
index f66ac20..34b43b4 100644
--- a/drivers/net/ethernet/tile/tilepro.c
+++ b/drivers/net/ethernet/tile/tilepro.c
@@ -89,10 +89,6 @@
/* ISSUE: This has not been thoroughly tested (except at 1500). */
#define TILE_NET_MTU 1500
-/* HACK: Define to support GSO. */
-/* ISSUE: This may actually hurt performance of the TCP blaster. */
-/* #define TILE_NET_GSO */
-
/* Define this to collapse "duplicate" acks. */
/* #define IGNORE_DUP_ACKS */
@@ -2336,39 +2332,28 @@ static const struct net_device_ops tile_net_ops = {
*/
static void tile_net_setup(struct net_device *dev)
{
- PDEBUG("tile_net_setup()\n");
+ netdev_features_t features = 0;
ether_setup(dev);
-
dev->netdev_ops = &tile_net_ops;
-
dev->watchdog_timeo = TILE_NET_TIMEOUT;
+ dev->tx_queue_len = TILE_NET_TX_QUEUE_LEN;
+ dev->mtu = TILE_NET_MTU;
- /* We want lockless xmit. */
- dev->features |= NETIF_F_LLTX;
-
- /* We support hardware tx checksums. */
- dev->features |= NETIF_F_HW_CSUM;
-
- /* We support scatter/gather. */
- dev->features |= NETIF_F_SG;
-
- /* We support TSO. */
- dev->features |= NETIF_F_TSO;
-
-#ifdef TILE_NET_GSO
- /* We support GSO. */
- dev->features |= NETIF_F_GSO;
-#endif
+ features |= NETIF_F_LLTX;
+ features |= NETIF_F_HW_CSUM;
+ features |= NETIF_F_SG;
+ features |= NETIF_F_TSO;
+ /* We can't support HIGHDMA without hash_default, since we need
+ * to be able to finv() with a VA if we don't have hash_default.
+ */
if (hash_default)
- dev->features |= NETIF_F_HIGHDMA;
-
- /* ISSUE: We should support NETIF_F_UFO. */
+ features |= NETIF_F_HIGHDMA;
- dev->tx_queue_len = TILE_NET_TX_QUEUE_LEN;
-
- dev->mtu = TILE_NET_MTU;
+ dev->hw_features |= features;
+ dev->vlan_features |= features;
+ dev->features |= features;
}
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 05/13] tile: support jumbo frames in the tilegx network driver
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
arch/tile/gxio/iorpc_mpipe.c | 47 +++++
arch/tile/gxio/mpipe.c | 18 +-
arch/tile/include/gxio/iorpc_mpipe.h | 4 +
arch/tile/include/gxio/mpipe.h | 101 +++++++++-
drivers/net/ethernet/tile/tilegx.c | 347 +++++++++++++++++++----------------
5 files changed, 351 insertions(+), 166 deletions(-)
diff --git a/arch/tile/gxio/iorpc_mpipe.c b/arch/tile/gxio/iorpc_mpipe.c
index 31b87bf..c2fb1516 100644
--- a/arch/tile/gxio/iorpc_mpipe.c
+++ b/arch/tile/gxio/iorpc_mpipe.c
@@ -387,6 +387,27 @@ int gxio_mpipe_link_close_aux(gxio_mpipe_context_t * context, int mac)
EXPORT_SYMBOL(gxio_mpipe_link_close_aux);
+struct link_set_attr_aux_param {
+ int mac;
+ uint32_t attr;
+ int64_t val;
+};
+
+int gxio_mpipe_link_set_attr_aux(gxio_mpipe_context_t * context, int mac,
+ uint32_t attr, int64_t val)
+{
+ struct link_set_attr_aux_param temp;
+ struct link_set_attr_aux_param *params = &temp;
+
+ params->mac = mac;
+ params->attr = attr;
+ params->val = val;
+
+ return hv_dev_pwrite(context->fd, 0, (HV_VirtAddr) params,
+ sizeof(*params), GXIO_MPIPE_OP_LINK_SET_ATTR_AUX);
+}
+
+EXPORT_SYMBOL(gxio_mpipe_link_set_attr_aux);
struct get_timestamp_aux_param {
uint64_t sec;
@@ -454,6 +475,32 @@ int gxio_mpipe_adjust_timestamp_aux(gxio_mpipe_context_t * context,
EXPORT_SYMBOL(gxio_mpipe_adjust_timestamp_aux);
+struct config_edma_ring_blks_param {
+ unsigned int ering;
+ unsigned int max_blks;
+ unsigned int min_snf_blks;
+ unsigned int db;
+};
+
+int gxio_mpipe_config_edma_ring_blks(gxio_mpipe_context_t * context,
+ unsigned int ering, unsigned int max_blks,
+ unsigned int min_snf_blks, unsigned int db)
+{
+ struct config_edma_ring_blks_param temp;
+ struct config_edma_ring_blks_param *params = &temp;
+
+ params->ering = ering;
+ params->max_blks = max_blks;
+ params->min_snf_blks = min_snf_blks;
+ params->db = db;
+
+ return hv_dev_pwrite(context->fd, 0, (HV_VirtAddr) params,
+ sizeof(*params),
+ GXIO_MPIPE_OP_CONFIG_EDMA_RING_BLKS);
+}
+
+EXPORT_SYMBOL(gxio_mpipe_config_edma_ring_blks);
+
struct arm_pollfd_param {
union iorpc_pollfd pollfd;
};
diff --git a/arch/tile/gxio/mpipe.c b/arch/tile/gxio/mpipe.c
index e71c633..0567cf0 100644
--- a/arch/tile/gxio/mpipe.c
+++ b/arch/tile/gxio/mpipe.c
@@ -383,7 +383,7 @@ EXPORT_SYMBOL_GPL(gxio_mpipe_iqueue_init);
int gxio_mpipe_equeue_init(gxio_mpipe_equeue_t *equeue,
gxio_mpipe_context_t *context,
- unsigned int edma_ring_id,
+ unsigned int ering,
unsigned int channel,
void *mem, unsigned int mem_size,
unsigned int mem_flags)
@@ -394,7 +394,7 @@ int gxio_mpipe_equeue_init(gxio_mpipe_equeue_t *equeue,
/* Offset used to read number of completed commands. */
MPIPE_EDMA_POST_REGION_ADDR_t offset;
- int result = gxio_mpipe_init_edma_ring(context, edma_ring_id, channel,
+ int result = gxio_mpipe_init_edma_ring(context, ering, channel,
mem, mem_size, mem_flags);
if (result < 0)
return result;
@@ -405,7 +405,7 @@ int gxio_mpipe_equeue_init(gxio_mpipe_equeue_t *equeue,
offset.region =
MPIPE_MMIO_ADDR__REGION_VAL_EDMA -
MPIPE_MMIO_ADDR__REGION_VAL_IDMA;
- offset.ring = edma_ring_id;
+ offset.ring = ering;
__gxio_dma_queue_init(&equeue->dma_queue,
context->mmio_fast_base + offset.word,
@@ -413,6 +413,9 @@ int gxio_mpipe_equeue_init(gxio_mpipe_equeue_t *equeue,
equeue->edescs = mem;
equeue->mask_num_entries = num_entries - 1;
equeue->log2_num_entries = __builtin_ctz(num_entries);
+ equeue->context = context;
+ equeue->ering = ering;
+ equeue->channel = channel;
return 0;
}
@@ -543,3 +546,12 @@ int gxio_mpipe_link_close(gxio_mpipe_link_t *link)
}
EXPORT_SYMBOL_GPL(gxio_mpipe_link_close);
+
+int gxio_mpipe_link_set_attr(gxio_mpipe_link_t *link, uint32_t attr,
+ int64_t val)
+{
+ return gxio_mpipe_link_set_attr_aux(link->context, link->mac, attr,
+ val);
+}
+
+EXPORT_SYMBOL_GPL(gxio_mpipe_link_set_attr);
diff --git a/arch/tile/include/gxio/iorpc_mpipe.h b/arch/tile/include/gxio/iorpc_mpipe.h
index 9d50fce..eef60fd 100644
--- a/arch/tile/include/gxio/iorpc_mpipe.h
+++ b/arch/tile/include/gxio/iorpc_mpipe.h
@@ -44,10 +44,12 @@
#define GXIO_MPIPE_OP_REGISTER_CLIENT_MEMORY IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x1210)
#define GXIO_MPIPE_OP_LINK_OPEN_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1211)
#define GXIO_MPIPE_OP_LINK_CLOSE_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1212)
+#define GXIO_MPIPE_OP_LINK_SET_ATTR_AUX IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1213)
#define GXIO_MPIPE_OP_GET_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x121e)
#define GXIO_MPIPE_OP_SET_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x121f)
#define GXIO_MPIPE_OP_ADJUST_TIMESTAMP_AUX IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x1220)
+#define GXIO_MPIPE_OP_CONFIG_EDMA_RING_BLKS IORPC_OPCODE(IORPC_FORMAT_NONE, 0x1221)
#define GXIO_MPIPE_OP_ARM_POLLFD IORPC_OPCODE(IORPC_FORMAT_KERNEL_POLLFD, 0x9000)
#define GXIO_MPIPE_OP_CLOSE_POLLFD IORPC_OPCODE(IORPC_FORMAT_KERNEL_POLLFD, 0x9001)
#define GXIO_MPIPE_OP_GET_MMIO_BASE IORPC_OPCODE(IORPC_FORMAT_NONE_NOUSER, 0x8000)
@@ -114,6 +116,8 @@ int gxio_mpipe_link_open_aux(gxio_mpipe_context_t * context,
int gxio_mpipe_link_close_aux(gxio_mpipe_context_t * context, int mac);
+int gxio_mpipe_link_set_attr_aux(gxio_mpipe_context_t * context, int mac,
+ uint32_t attr, int64_t val);
int gxio_mpipe_get_timestamp_aux(gxio_mpipe_context_t * context, uint64_t * sec,
uint64_t * nsec, uint64_t * cycles);
diff --git a/arch/tile/include/gxio/mpipe.h b/arch/tile/include/gxio/mpipe.h
index b74f470..ed742e3 100644
--- a/arch/tile/include/gxio/mpipe.h
+++ b/arch/tile/include/gxio/mpipe.h
@@ -810,7 +810,7 @@ extern int gxio_mpipe_alloc_edma_rings(gxio_mpipe_context_t *context,
/* Initialize an eDMA ring, using the given memory and size.
*
* @param context An initialized mPIPE context.
- * @param ring The eDMA ring index.
+ * @param ering The eDMA ring index.
* @param channel The channel to use. This must be one of the channels
* associated with the context's set of open links.
* @param mem A physically contiguous region of memory to be filled
@@ -823,10 +823,37 @@ extern int gxio_mpipe_alloc_edma_rings(gxio_mpipe_context_t *context,
* ::GXIO_ERR_INVAL_MEMORY_SIZE on failure.
*/
extern int gxio_mpipe_init_edma_ring(gxio_mpipe_context_t *context,
- unsigned int ring, unsigned int channel,
+ unsigned int ering, unsigned int channel,
void *mem, size_t mem_size,
unsigned int mem_flags);
+/* Set the "max_blks", "min_snf_blks", and "db" fields of
+ * ::MPIPE_EDMA_RG_INIT_DAT_THRESH_t for a given edma ring.
+ *
+ * The global pool of dynamic blocks will be automatically adjusted.
+ *
+ * This function should not be called after any egress has been done
+ * on the edma ring.
+ *
+ * Most applications should just use gxio_mpipe_equeue_set_snf_size().
+ *
+ * @param context An initialized mPIPE context.
+ * @param ering The eDMA ring index.
+ * @param max_blks The number of blocks to dedicate to the ring
+ * (normally min_snf_blks + 1). Must be greater than min_snf_blocks.
+ * @param min_snf_blks The number of blocks which must be stored
+ * prior to starting to send the packet (normally 12).
+ * @param db Whether to allow use of dynamic blocks by the ring
+ * (normally 1).
+ *
+ * @return 0 on success, negative on error.
+ */
+extern int gxio_mpipe_config_edma_ring_blks(gxio_mpipe_context_t *context,
+ unsigned int ering,
+ unsigned int max_blks,
+ unsigned int min_snf_blks,
+ unsigned int db);
+
/*****************************************************************
* Classifier Program *
******************************************************************/
@@ -1288,15 +1315,39 @@ typedef struct {
/* The log2() of the number of entries. */
unsigned long log2_num_entries;
+ /* The context. */
+ gxio_mpipe_context_t *context;
+
+ /* The ering. */
+ unsigned int ering;
+
+ /* The channel. */
+ unsigned int channel;
+
} gxio_mpipe_equeue_t;
/* Initialize an "equeue".
*
- * Takes the equeue plus the same args as gxio_mpipe_init_edma_ring().
+ * This function uses gxio_mpipe_init_edma_ring() to initialize the
+ * underlying edma_ring using the provided arguments.
+ *
+ * @param equeue An egress queue to be initialized.
+ * @param context An initialized mPIPE context.
+ * @param ering The eDMA ring index.
+ * @param channel The channel to use. This must be one of the channels
+ * associated with the context's set of open links.
+ * @param mem A physically contiguous region of memory to be filled
+ * with a ring of ::gxio_mpipe_edesc_t structures.
+ * @param mem_size Number of bytes in the ring. Must be 512, 2048,
+ * 8192 or 65536, times 16 (i.e. sizeof(gxio_mpipe_edesc_t)).
+ * @param mem_flags ::gxio_mpipe_mem_flags_e memory flags.
+ *
+ * @return 0 on success, ::GXIO_MPIPE_ERR_BAD_EDMA_RING or
+ * ::GXIO_ERR_INVAL_MEMORY_SIZE on failure.
*/
extern int gxio_mpipe_equeue_init(gxio_mpipe_equeue_t *equeue,
gxio_mpipe_context_t *context,
- unsigned int edma_ring_id,
+ unsigned int ering,
unsigned int channel,
void *mem, unsigned int mem_size,
unsigned int mem_flags);
@@ -1494,6 +1545,37 @@ static inline int gxio_mpipe_equeue_is_complete(gxio_mpipe_equeue_t *equeue,
completion_slot, update);
}
+/* Set the snf (store and forward) size for an equeue.
+ *
+ * The snf size for an equeue defaults to 1536, and encodes the size
+ * of the largest packet for which egress is guaranteed to avoid
+ * transmission underruns and/or corrupt checksums under heavy load.
+ *
+ * The snf size affects a global resource pool which cannot support,
+ * for example, all 24 equeues each requesting an snf size of 8K.
+ *
+ * To ensure that jumbo packets can be egressed properly, the snf size
+ * should be set to the size of the largest possible packet, which
+ * will usually be limited by the size of the app's largest buffer.
+ *
+ * This is a convenience wrapper around
+ * gxio_mpipe_config_edma_ring_blks().
+ *
+ * This function should not be called after any egress has been done
+ * on the equeue.
+ *
+ * @param equeue An egress queue initialized via gxio_mpipe_equeue_init().
+ * @param size The snf size, in bytes.
+ * @return Zero on success, negative error otherwise.
+ */
+static inline int gxio_mpipe_equeue_set_snf_size(gxio_mpipe_equeue_t *equeue,
+ size_t size)
+{
+ int blks = (size + 127) / 128;
+ return gxio_mpipe_config_edma_ring_blks(equeue->context, equeue->ering,
+ blks + 1, blks, 1);
+}
+
/*****************************************************************
* Link Management *
******************************************************************/
@@ -1697,6 +1779,17 @@ static inline int gxio_mpipe_link_channel(gxio_mpipe_link_t *link)
return link->channel;
}
+/* Set a link attribute.
+ *
+ * @param link A properly initialized link state object.
+ * @param attr An attribute from the set of @ref gxio_mpipe_link_attrs.
+ * @param val New value of the attribute.
+ * @return 0 if the attribute was successfully set, or a negative error
+ * code.
+ */
+extern int gxio_mpipe_link_set_attr(gxio_mpipe_link_t *link, uint32_t attr,
+ int64_t val);
+
///////////////////////////////////////////////////////////////////
// Timestamp //
///////////////////////////////////////////////////////////////////
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index 6085571..39c1e9e 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -76,6 +76,9 @@
#define MAX_FRAGS (MAX_SKB_FRAGS + 1)
+/* The "kinds" of buffer stacks (small/large/jumbo). */
+#define MAX_KINDS 3
+
/* Size of completions data to allocate.
* ISSUE: Probably more than needed since we don't use all the channels.
*/
@@ -141,10 +144,8 @@ struct tile_net_info {
/* NAPI flags. */
bool napi_added;
bool napi_enabled;
- /* Number of small sk_buffs which must still be provided. */
- unsigned int num_needed_small_buffers;
- /* Number of large sk_buffs which must still be provided. */
- unsigned int num_needed_large_buffers;
+ /* Number of buffers (by kind) which must still be provided. */
+ unsigned int num_needed_buffers[MAX_KINDS];
/* A timer for handling egress completions. */
struct hrtimer egress_timer;
/* True if "egress_timer" is scheduled. */
@@ -200,24 +201,25 @@ static DEFINE_PER_CPU(struct tile_net_info, per_cpu_info);
/* The "context" for all devices. */
static gxio_mpipe_context_t context;
-/* Buffer sizes and mpipe enum codes for buffer stacks.
+/* The buffer size enums for each buffer stack.
* See arch/tile/include/gxio/mpipe.h for the set of possible values.
+ * We avoid the "10384" size because it can induce "false chaining"
+ * on "cut-through" jumbo packets.
*/
-#define BUFFER_SIZE_SMALL_ENUM GXIO_MPIPE_BUFFER_SIZE_128
-#define BUFFER_SIZE_SMALL 128
-#define BUFFER_SIZE_LARGE_ENUM GXIO_MPIPE_BUFFER_SIZE_1664
-#define BUFFER_SIZE_LARGE 1664
+static gxio_mpipe_buffer_size_enum_t buffer_size_enums[MAX_KINDS] = {
+ GXIO_MPIPE_BUFFER_SIZE_128,
+ GXIO_MPIPE_BUFFER_SIZE_1664,
+ GXIO_MPIPE_BUFFER_SIZE_16384
+};
-/* The small/large "buffer stacks". */
-static int small_buffer_stack = -1;
-static int large_buffer_stack = -1;
+/* The actual memory allocated for the buffer stacks. */
+static void *buffer_stack_vas[MAX_KINDS];
-/* Amount of memory allocated for each buffer stack. */
-static size_t buffer_stack_size;
+/* The amount of memory allocated for each buffer stack. */
+static size_t buffer_stack_bytes[MAX_KINDS];
-/* The actual memory allocated for the buffer stacks. */
-static void *small_buffer_stack_va;
-static void *large_buffer_stack_va;
+/* The first buffer stack index (small = +0, large = +1, jumbo = +2). */
+static int first_buffer_stack = -1;
/* The buckets. */
static int first_bucket = -1;
@@ -238,6 +240,9 @@ static char *loopify_link_name;
/* If "tile_net.custom" was specified, this is non-NULL. */
static char *custom_str;
+/* If "tile_net.jumbo=NUM" was specified, this is "NUM". */
+static uint jumbo_num;
+
/* The "tile_net.cpus" argument specifies the cpus that are dedicated
* to handle ingress packets.
*
@@ -292,6 +297,12 @@ MODULE_PARM_DESC(loopify, "name the device to use loop0/1 for ingress/egress");
module_param_named(custom, custom_str, charp, 0444);
MODULE_PARM_DESC(custom, "indicates a (heavily) customized classifier");
+/* The "tile_net.jumbo" argument causes us to support "jumbo" packets,
+ * and to allocate the given number of "jumbo" buffers.
+ */
+module_param_named(jumbo, jumbo_num, uint, 0444);
+MODULE_PARM_DESC(jumbo, "the number of buffers to support jumbo packets");
+
/* Atomically update a statistics field.
* Note that on TILE-Gx, this operation is fire-and-forget on the
* issuing core (single-cycle dispatch) and takes only a few cycles
@@ -305,15 +316,15 @@ static void tile_net_stats_add(unsigned long value, unsigned long *field)
}
/* Allocate and push a buffer. */
-static bool tile_net_provide_buffer(bool small)
+static bool tile_net_provide_buffer(int kind)
{
- int stack = small ? small_buffer_stack : large_buffer_stack;
+ gxio_mpipe_buffer_size_enum_t bse = buffer_size_enums[kind];
+ size_t bs = gxio_mpipe_buffer_size_enum_to_buffer_size(bse);
const unsigned long buffer_alignment = 128;
struct sk_buff *skb;
int len;
- len = sizeof(struct sk_buff **) + buffer_alignment;
- len += (small ? BUFFER_SIZE_SMALL : BUFFER_SIZE_LARGE);
+ len = sizeof(struct sk_buff **) + buffer_alignment + bs;
skb = dev_alloc_skb(len);
if (skb == NULL)
return false;
@@ -328,7 +339,7 @@ static bool tile_net_provide_buffer(bool small)
/* Make sure "skb" and the back-pointer have been flushed. */
wmb();
- gxio_mpipe_push_buffer(&context, stack,
+ gxio_mpipe_push_buffer(&context, first_buffer_stack + kind,
(void *)va_to_tile_io_addr(skb->data));
return true;
@@ -369,24 +380,19 @@ static void tile_net_pop_all_buffers(int stack)
static void tile_net_provide_needed_buffers(void)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
-
- while (info->num_needed_small_buffers != 0) {
- if (!tile_net_provide_buffer(true))
- goto oops;
- info->num_needed_small_buffers--;
- }
-
- while (info->num_needed_large_buffers != 0) {
- if (!tile_net_provide_buffer(false))
- goto oops;
- info->num_needed_large_buffers--;
+ int kind;
+
+ for (kind = 0; kind < MAX_KINDS; kind++) {
+ while (info->num_needed_buffers[kind] != 0) {
+ if (!tile_net_provide_buffer(kind)) {
+ /* Add info to the allocation failure dump. */
+ pr_notice("Tile %d still needs some buffers\n",
+ info->my_cpu);
+ return;
+ }
+ info->num_needed_buffers[kind]--;
+ }
}
-
- return;
-
-oops:
- /* Add a description to the page allocation failure dump. */
- pr_notice("Tile %d still needs some buffers\n", info->my_cpu);
}
static inline bool filter_packet(struct net_device *dev, void *buf)
@@ -426,10 +432,12 @@ static void tile_net_receive_skb(struct net_device *dev, struct sk_buff *skb,
tile_net_stats_add(len, &priv->stats.rx_bytes);
/* Need a new buffer. */
- if (idesc->size == BUFFER_SIZE_SMALL_ENUM)
- info->num_needed_small_buffers++;
+ if (idesc->size == buffer_size_enums[0])
+ info->num_needed_buffers[0]++;
+ else if (idesc->size == buffer_size_enums[1])
+ info->num_needed_buffers[1]++;
else
- info->num_needed_large_buffers++;
+ info->num_needed_buffers[2]++;
}
/* Handle a packet. Return true if "processed", false if "filtered". */
@@ -437,29 +445,29 @@ static bool tile_net_handle_packet(gxio_mpipe_idesc_t *idesc)
{
struct tile_net_info *info = &__get_cpu_var(per_cpu_info);
struct net_device *dev = tile_net_devs_for_channel[idesc->channel];
+ struct tile_net_priv *priv = netdev_priv(dev);
uint8_t l2_offset;
void *va;
void *buf;
unsigned long len;
bool filter;
- /* Drop packets for which no buffer was available.
- * NOTE: This happens under heavy load.
+ /* Drop packets for which no buffer was available (which can
+ * happen under heavy load), or for which the me/tr/ce flags
+ * are set (which can happen for jumbo cut-through packets,
+ * or with a customized classifier).
*/
- if (idesc->be) {
- struct tile_net_priv *priv = netdev_priv(dev);
- tile_net_stats_add(1, &priv->stats.rx_dropped);
- gxio_mpipe_iqueue_consume(&info->iqueue, idesc);
- if (net_ratelimit())
- pr_info("Dropping packet (insufficient buffers).\n");
- return false;
+ if (idesc->be || idesc->me || idesc->tr || idesc->ce) {
+ if (dev)
+ tile_net_stats_add(1, &priv->stats.rx_errors);
+ goto drop;
}
/* Get the "l2_offset", if allowed. */
l2_offset = custom_str ? 0 : gxio_mpipe_idesc_get_l2_offset(idesc);
- /* Get the raw buffer VA (includes "headroom"). */
- va = tile_io_addr_to_va((unsigned long)(long)idesc->va);
+ /* Get the VA (including NET_IP_ALIGN bytes of "headroom"). */
+ va = tile_io_addr_to_va((unsigned long)idesc->va);
/* Get the actual packet start/length. */
buf = va + l2_offset;
@@ -470,6 +478,9 @@ static bool tile_net_handle_packet(gxio_mpipe_idesc_t *idesc)
filter = filter_packet(dev, buf);
if (filter) {
+ if (dev)
+ tile_net_stats_add(1, &priv->stats.rx_dropped);
+drop:
gxio_mpipe_iqueue_drop(&info->iqueue, idesc);
} else {
struct sk_buff *skb = mpipe_buf_to_skb(va);
@@ -722,86 +733,95 @@ static int tile_net_update(struct net_device *dev)
return 0;
}
-/* Allocate and initialize mpipe buffer stacks, and register them in
- * the mPIPE TLBs, for both small and large packet sizes.
- * This routine supports tile_net_init_mpipe(), below.
- */
-static int init_buffer_stacks(struct net_device *dev, int num_buffers)
+/* Initialize a buffer stack. */
+static int create_buffer_stack(struct net_device *dev,
+ int kind, size_t num_buffers)
{
pte_t hash_pte = pte_set_home((pte_t) { 0 }, PAGE_HOME_HASH);
- int rc;
+ size_t needed = gxio_mpipe_calc_buffer_stack_bytes(num_buffers);
+ int stack_idx = first_buffer_stack + kind;
+ void *va;
+ int i, rc;
- /* Compute stack bytes; we round up to 64KB and then use
- * alloc_pages() so we get the required 64KB alignment as well.
+ /* Round up to 64KB and then use alloc_pages() so we get the
+ * required 64KB alignment.
*/
- buffer_stack_size =
- ALIGN(gxio_mpipe_calc_buffer_stack_bytes(num_buffers),
- 64 * 1024);
-
- /* Allocate two buffer stack indices. */
- rc = gxio_mpipe_alloc_buffer_stacks(&context, 2, 0, 0);
- if (rc < 0) {
- netdev_err(dev, "gxio_mpipe_alloc_buffer_stacks failed: %d\n",
- rc);
- return rc;
- }
- small_buffer_stack = rc;
- large_buffer_stack = rc + 1;
+ buffer_stack_bytes[kind] = ALIGN(needed, 64 * 1024);
- /* Allocate the small memory stack. */
- small_buffer_stack_va =
- alloc_pages_exact(buffer_stack_size, GFP_KERNEL);
- if (small_buffer_stack_va == NULL) {
+ va = alloc_pages_exact(buffer_stack_bytes[kind], GFP_KERNEL);
+ if (va == NULL) {
netdev_err(dev,
- "Could not alloc %zd bytes for buffer stacks\n",
- buffer_stack_size);
+ "Could not alloc %zd bytes for buffer stack %d\n",
+ buffer_stack_bytes[kind], kind);
return -ENOMEM;
}
- rc = gxio_mpipe_init_buffer_stack(&context, small_buffer_stack,
- BUFFER_SIZE_SMALL_ENUM,
- small_buffer_stack_va,
- buffer_stack_size, 0);
+
+ /* Initialize the buffer stack. */
+ rc = gxio_mpipe_init_buffer_stack(&context, stack_idx,
+ buffer_size_enums[kind],
+ va, buffer_stack_bytes[kind], 0);
if (rc != 0) {
netdev_err(dev, "gxio_mpipe_init_buffer_stack: %d\n", rc);
+ free_pages_exact(va, buffer_stack_bytes[kind]);
return rc;
}
- rc = gxio_mpipe_register_client_memory(&context, small_buffer_stack,
+
+ buffer_stack_vas[kind] = va;
+
+ rc = gxio_mpipe_register_client_memory(&context, stack_idx,
hash_pte, 0);
if (rc != 0) {
- netdev_err(dev,
- "gxio_mpipe_register_buffer_memory failed: %d\n",
- rc);
+ netdev_err(dev, "gxio_mpipe_register_client_memory: %d\n", rc);
return rc;
}
- /* Allocate the large buffer stack. */
- large_buffer_stack_va =
- alloc_pages_exact(buffer_stack_size, GFP_KERNEL);
- if (large_buffer_stack_va == NULL) {
- netdev_err(dev,
- "Could not alloc %zd bytes for buffer stacks\n",
- buffer_stack_size);
- return -ENOMEM;
- }
- rc = gxio_mpipe_init_buffer_stack(&context, large_buffer_stack,
- BUFFER_SIZE_LARGE_ENUM,
- large_buffer_stack_va,
- buffer_stack_size, 0);
- if (rc != 0) {
- netdev_err(dev, "gxio_mpipe_init_buffer_stack failed: %d\n",
- rc);
- return rc;
+ /* Provide initial buffers. */
+ for (i = 0; i < num_buffers; i++) {
+ if (!tile_net_provide_buffer(kind)) {
+ netdev_err(dev, "Cannot allocate initial sk_bufs!\n");
+ return -ENOMEM;
+ }
}
- rc = gxio_mpipe_register_client_memory(&context, large_buffer_stack,
- hash_pte, 0);
- if (rc != 0) {
- netdev_err(dev,
- "gxio_mpipe_register_buffer_memory failed: %d\n",
- rc);
+
+ return 0;
+}
+
+/* Allocate and initialize mpipe buffer stacks, and register them in
+ * the mPIPE TLBs, for small, large, and (possibly) jumbo packet sizes.
+ * This routine supports tile_net_init_mpipe(), below.
+ */
+static int init_buffer_stacks(struct net_device *dev,
+ int network_cpus_count)
+{
+ int num_kinds = MAX_KINDS - (jumbo_num == 0);
+ size_t num_buffers;
+ int rc;
+
+ /* Allocate the buffer stacks. */
+ rc = gxio_mpipe_alloc_buffer_stacks(&context, num_kinds, 0, 0);
+ if (rc < 0) {
+ netdev_err(dev, "gxio_mpipe_alloc_buffer_stacks: %d\n", rc);
return rc;
}
+ first_buffer_stack = rc;
- return 0;
+ /* Enough small/large buffers to (normally) avoid buffer errors. */
+ num_buffers =
+ network_cpus_count * (IQUEUE_ENTRIES + TILE_NET_BATCH);
+
+ /* Allocate the small memory stack. */
+ if (rc >= 0)
+ rc = create_buffer_stack(dev, 0, num_buffers);
+
+ /* Allocate the large buffer stack. */
+ if (rc >= 0)
+ rc = create_buffer_stack(dev, 1, num_buffers);
+
+ /* Allocate the jumbo buffer stack if needed. */
+ if (rc >= 0 && jumbo_num != 0)
+ rc = create_buffer_stack(dev, 2, jumbo_num);
+
+ return rc;
}
/* Allocate per-cpu resources (memory for completions and idescs).
@@ -940,13 +960,14 @@ static int tile_net_setup_interrupts(struct net_device *dev)
/* Undo any state set up partially by a failed call to tile_net_init_mpipe. */
static void tile_net_init_mpipe_fail(void)
{
- int cpu;
+ int kind, cpu;
/* Do cleanups that require the mpipe context first. */
- if (small_buffer_stack >= 0)
- tile_net_pop_all_buffers(small_buffer_stack);
- if (large_buffer_stack >= 0)
- tile_net_pop_all_buffers(large_buffer_stack);
+ for (kind = 0; kind < MAX_KINDS; kind++) {
+ if (buffer_stack_vas[kind] != NULL) {
+ tile_net_pop_all_buffers(first_buffer_stack + kind);
+ }
+ }
/* Destroy mpipe context so the hardware no longer owns any memory. */
gxio_mpipe_destroy(&context);
@@ -961,15 +982,15 @@ static void tile_net_init_mpipe_fail(void)
info->iqueue.idescs = NULL;
}
- if (small_buffer_stack_va)
- free_pages_exact(small_buffer_stack_va, buffer_stack_size);
- if (large_buffer_stack_va)
- free_pages_exact(large_buffer_stack_va, buffer_stack_size);
+ for (kind = 0; kind < MAX_KINDS; kind++) {
+ if (buffer_stack_vas[kind] != NULL) {
+ free_pages_exact(buffer_stack_vas[kind],
+ buffer_stack_bytes[kind]);
+ buffer_stack_vas[kind] = NULL;
+ }
+ }
- small_buffer_stack_va = NULL;
- large_buffer_stack_va = NULL;
- large_buffer_stack = -1;
- small_buffer_stack = -1;
+ first_buffer_stack = -1;
first_bucket = -1;
}
@@ -984,7 +1005,7 @@ static void tile_net_init_mpipe_fail(void)
*/
static int tile_net_init_mpipe(struct net_device *dev)
{
- int i, num_buffers, rc;
+ int rc;
int cpu;
int first_ring, ring;
int network_cpus_count = cpus_weight(network_cpus_map);
@@ -1001,27 +1022,10 @@ static int tile_net_init_mpipe(struct net_device *dev)
}
/* Set up the buffer stacks. */
- num_buffers =
- network_cpus_count * (IQUEUE_ENTRIES + TILE_NET_BATCH);
- rc = init_buffer_stacks(dev, num_buffers);
+ rc = init_buffer_stacks(dev, network_cpus_count);
if (rc != 0)
goto fail;
- /* Provide initial buffers. */
- rc = -ENOMEM;
- for (i = 0; i < num_buffers; i++) {
- if (!tile_net_provide_buffer(true)) {
- netdev_err(dev, "Cannot allocate initial sk_bufs!\n");
- goto fail;
- }
- }
- for (i = 0; i < num_buffers; i++) {
- if (!tile_net_provide_buffer(false)) {
- netdev_err(dev, "Cannot allocate initial sk_bufs!\n");
- goto fail;
- }
- }
-
/* Allocate one NotifRing for each network cpu. */
rc = gxio_mpipe_alloc_notif_rings(&context, network_cpus_count, 0, 0);
if (rc < 0) {
@@ -1063,13 +1067,13 @@ fail:
*/
static int tile_net_init_egress(struct net_device *dev, int echannel)
{
+ static int ering = -1;
struct page *headers_page, *edescs_page, *equeue_page;
gxio_mpipe_edesc_t *edescs;
gxio_mpipe_equeue_t *equeue;
unsigned char *headers;
int headers_order, edescs_order, equeue_order;
size_t edescs_size;
- int edma;
int rc = -ENOMEM;
/* Only initialize once. */
@@ -1110,25 +1114,37 @@ static int tile_net_init_egress(struct net_device *dev, int echannel)
}
equeue = pfn_to_kaddr(page_to_pfn(equeue_page));
- /* Allocate an edma ring. Note that in practice this can't
- * fail, which is good, because we will leak an edma ring if so.
- */
- rc = gxio_mpipe_alloc_edma_rings(&context, 1, 0, 0);
- if (rc < 0) {
- netdev_warn(dev, "gxio_mpipe_alloc_edma_rings failed: %d\n",
- rc);
- goto fail_equeue;
+ /* Allocate an edma ring (using a one entry "free list"). */
+ if (ering < 0) {
+ rc = gxio_mpipe_alloc_edma_rings(&context, 1, 0, 0);
+ if (rc < 0) {
+ netdev_warn(dev, "gxio_mpipe_alloc_edma_rings: %d\n",
+ rc);
+ goto fail_equeue;
+ }
+ ering = rc;
}
- edma = rc;
/* Initialize the equeue. */
- rc = gxio_mpipe_equeue_init(equeue, &context, edma, echannel,
+ rc = gxio_mpipe_equeue_init(equeue, &context, ering, echannel,
edescs, edescs_size, 0);
if (rc != 0) {
netdev_err(dev, "gxio_mpipe_equeue_init failed: %d\n", rc);
goto fail_equeue;
}
+ /* Don't reuse the ering later. */
+ ering = -1;
+
+ if (jumbo_num != 0) {
+ /* Make sure "jumbo" packets can be egressed safely. */
+ if (gxio_mpipe_equeue_set_snf_size(equeue, 10368) < 0) {
+ /* ISSUE: There is no "gxio_mpipe_equeue_destroy()". */
+ netdev_warn(dev, "Jumbo packets may not be egressed"
+ " properly on channel %d\n", echannel);
+ }
+ }
+
/* Done. */
egress_for_echannel[echannel].equeue = equeue;
egress_for_echannel[echannel].headers = headers;
@@ -1156,6 +1172,17 @@ static int tile_net_link_open(struct net_device *dev, gxio_mpipe_link_t *link,
netdev_err(dev, "Failed to open '%s'\n", link_name);
return rc;
}
+ if (jumbo_num != 0) {
+ u32 attr = GXIO_MPIPE_LINK_RECEIVE_JUMBO;
+ rc = gxio_mpipe_link_set_attr(link, attr, 1);
+ if (rc != 0) {
+ netdev_err(dev,
+ "Cannot receive jumbo packets on '%s'\n",
+ link_name);
+ gxio_mpipe_link_close(link);
+ return rc;
+ }
+ }
rc = gxio_mpipe_link_channel(link);
if (rc < 0 || rc >= TILE_NET_CHANNELS) {
netdev_err(dev, "gxio_mpipe_link_channel bad value: %d\n", rc);
@@ -1499,8 +1526,8 @@ static void tso_egress(struct net_device *dev, gxio_mpipe_equeue_t *equeue,
edesc_head.xfer_size = sh_len;
/* This is only used to specify the TLB. */
- edesc_head.stack_idx = large_buffer_stack;
- edesc_body.stack_idx = large_buffer_stack;
+ edesc_head.stack_idx = first_buffer_stack;
+ edesc_body.stack_idx = first_buffer_stack;
/* Egress all the edescs. */
for (segment = 0; segment < sh->gso_segs; segment++) {
@@ -1660,7 +1687,7 @@ static int tile_net_tx(struct sk_buff *skb, struct net_device *dev)
num_edescs = tile_net_tx_frags(frags, skb, data, skb_headlen(skb));
/* This is only used to specify the TLB. */
- edesc.stack_idx = large_buffer_stack;
+ edesc.stack_idx = first_buffer_stack;
/* Prepare the edescs. */
for (i = 0; i < num_edescs; i++) {
@@ -1740,7 +1767,9 @@ static struct net_device_stats *tile_net_get_stats(struct net_device *dev)
/* Change the MTU. */
static int tile_net_change_mtu(struct net_device *dev, int new_mtu)
{
- if ((new_mtu < 68) || (new_mtu > 1500))
+ if (new_mtu < 68)
+ return -EINVAL;
+ if (new_mtu > ((jumbo_num != 0) ? 9000 : 1500))
return -EINVAL;
dev->mtu = new_mtu;
return 0;
--
1.8.3.1
^ permalink raw reply related
* [PATCH v3 02/13] tile: support rx_dropped/rx_errors in tilepro net driver
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: linux-kernel, netdev
In-Reply-To: <cover.1375385119.git.cmetcalf@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
drivers/net/ethernet/tile/tilepro.c | 41 ++++++++++++++-----------------------
1 file changed, 15 insertions(+), 26 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilepro.c b/drivers/net/ethernet/tile/tilepro.c
index 34b43b4..327ff7b 100644
--- a/drivers/net/ethernet/tile/tilepro.c
+++ b/drivers/net/ethernet/tile/tilepro.c
@@ -772,6 +772,7 @@ static bool tile_net_poll_aux(struct tile_net_cpu *info, int index)
netio_pkt_t *pkt = (netio_pkt_t *)((unsigned long) &qsp[1] + index);
netio_pkt_metadata_t *metadata = NETIO_PKT_METADATA(pkt);
+ netio_pkt_status_t pkt_status = NETIO_PKT_STATUS_M(metadata, pkt);
/* Extract the packet size. FIXME: Shouldn't the second line */
/* get subtracted? Mostly moot, since it should be "zero". */
@@ -804,40 +805,25 @@ static bool tile_net_poll_aux(struct tile_net_cpu *info, int index)
#endif /* TILE_NET_DUMP_PACKETS */
#ifdef TILE_NET_VERIFY_INGRESS
- if (!NETIO_PKT_L4_CSUM_CORRECT_M(metadata, pkt) &&
- NETIO_PKT_L4_CSUM_CALCULATED_M(metadata, pkt)) {
- /* Bug 6624: Includes UDP packets with a "zero" checksum. */
- pr_warning("Bad L4 checksum on %d byte packet.\n", len);
- }
- if (!NETIO_PKT_L3_CSUM_CORRECT_M(metadata, pkt) &&
- NETIO_PKT_L3_CSUM_CALCULATED_M(metadata, pkt)) {
+ if (pkt_status == NETIO_PKT_STATUS_OVERSIZE && len >= 64) {
dump_packet(buf, len, "rx");
- panic("Bad L3 checksum.");
- }
- switch (NETIO_PKT_STATUS_M(metadata, pkt)) {
- case NETIO_PKT_STATUS_OVERSIZE:
- if (len >= 64) {
- dump_packet(buf, len, "rx");
- panic("Unexpected OVERSIZE.");
- }
- break;
- case NETIO_PKT_STATUS_BAD:
- pr_warning("Unexpected BAD %ld byte packet.\n", len);
+ panic("Unexpected OVERSIZE.");
}
#endif
filter = 0;
- /* ISSUE: Filter TCP packets with "bad" checksums? */
-
- if (!(dev->flags & IFF_UP)) {
+ if (pkt_status == NETIO_PKT_STATUS_BAD) {
+ /* Handle CRC error and hardware truncation. */
+ filter = 2;
+ } else if (!(dev->flags & IFF_UP)) {
/* Filter packets received before we're up. */
filter = 1;
- } else if (NETIO_PKT_STATUS_M(metadata, pkt) == NETIO_PKT_STATUS_BAD) {
+ } else if (NETIO_PKT_ETHERTYPE_RECOGNIZED_M(metadata, pkt) &&
+ pkt_status == NETIO_PKT_STATUS_UNDERSIZE) {
/* Filter "truncated" packets. */
- filter = 1;
+ filter = 2;
} else if (!(dev->flags & IFF_PROMISC)) {
- /* FIXME: Implement HW multicast filter. */
if (!is_multicast_ether_addr(buf)) {
/* Filter packets not for our address. */
const u8 *mine = dev->dev_addr;
@@ -847,9 +833,12 @@ static bool tile_net_poll_aux(struct tile_net_cpu *info, int index)
u64_stats_update_begin(&stats->syncp);
- if (filter) {
+ if (filter != 0) {
- /* ISSUE: Update "drop" statistics? */
+ if (filter == 1)
+ stats->rx_dropped++;
+ else
+ stats->rx_errors++;
tile_net_provide_linux_buffer(info, va, small);
--
1.8.3.1
^ permalink raw reply related
* [PATCH v2 13/12] tile: set hw_features and vlan_features in setup
From: Chris Metcalf @ 2013-08-01 15:36 UTC (permalink / raw)
To: David Miller; +Cc: linux-kernel, netdev
In-Reply-To: <20130731.122534.774117327230299927.davem@redhat.com>
This change allows the user to configure various features of the tile
networking drivers on and off. There is no change to the default
initialization state of either the tilegx or tilepro drivers.
Neither driver needs the ndo_fix_features or ndo_set_features callbacks,
since the generic code already handles the dependencies for
fix_features, and there is no hardware state to tweak in set_features.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
---
David, I've fixed the drivers to use hw_features as a follow-on "13 of
12" patch in the v2 series, rather than rebasing the previous patches,
since they weren't actually buggy as-is, and this was simpler. However,
I am happy to provide a v3 of the patch series with the hw_features
change rebased to the front if that feels better to you.
drivers/net/ethernet/tile/tilegx.c | 15 ++++++++++----
drivers/net/ethernet/tile/tilepro.c | 39 +++++++++++++++----------------------
2 files changed, 27 insertions(+), 27 deletions(-)
diff --git a/drivers/net/ethernet/tile/tilegx.c b/drivers/net/ethernet/tile/tilegx.c
index cd9bfc7..3b1dc66 100644
--- a/drivers/net/ethernet/tile/tilegx.c
+++ b/drivers/net/ethernet/tile/tilegx.c
@@ -2169,14 +2169,21 @@ static const struct net_device_ops tile_net_ops = {
*/
static void tile_net_setup(struct net_device *dev)
{
+ netdev_features_t features = 0;
+
ether_setup(dev);
dev->netdev_ops = &tile_net_ops;
dev->watchdog_timeo = TILE_NET_TIMEOUT;
- dev->features |= NETIF_F_HW_CSUM;
- dev->features |= NETIF_F_SG;
- dev->features |= NETIF_F_TSO;
- dev->features |= NETIF_F_TSO6;
dev->mtu = 1500;
+
+ features |= NETIF_F_HW_CSUM;
+ features |= NETIF_F_SG;
+ features |= NETIF_F_TSO;
+ features |= NETIF_F_TSO6;
+
+ dev->hw_features |= features;
+ dev->vlan_features |= features;
+ dev->features |= features;
}
/* Allocate the device structure, register the device, and obtain the
diff --git a/drivers/net/ethernet/tile/tilepro.c b/drivers/net/ethernet/tile/tilepro.c
index 874c7eb..2360b38 100644
--- a/drivers/net/ethernet/tile/tilepro.c
+++ b/drivers/net/ethernet/tile/tilepro.c
@@ -89,9 +89,9 @@
/* ISSUE: This has not been thoroughly tested (except at 1500). */
#define TILE_NET_MTU 1500
-/* HACK: Define to support GSO. */
-/* ISSUE: This may actually hurt performance of the TCP blaster. */
-/* #define TILE_NET_GSO */
+/* Define as 0 to enable all features by default. */
+/* ISSUE: GSO appears to actually hurt performance of the TCP blaster. */
+#define TILE_NETIF_F_DEFAULT_DISABLE NETIF_F_GSO
/* HACK: Define this to verify incoming packets. */
/* #define TILE_NET_VERIFY_INGRESS */
@@ -2249,37 +2249,30 @@ static const struct net_device_ops tile_net_ops = {
*/
static void tile_net_setup(struct net_device *dev)
{
- PDEBUG("tile_net_setup()\n");
+ netdev_features_t features = 0;
ether_setup(dev);
-
dev->netdev_ops = &tile_net_ops;
-
dev->watchdog_timeo = TILE_NET_TIMEOUT;
+ dev->tx_queue_len = TILE_NET_TX_QUEUE_LEN;
+ dev->mtu = TILE_NET_MTU;
- /* We support hardware tx checksums. */
- dev->features |= NETIF_F_HW_CSUM;
-
- /* We support scatter/gather. */
- dev->features |= NETIF_F_SG;
+ features |= NETIF_F_HW_CSUM;
+ features |= NETIF_F_SG;
/* We support TSO iff the HV supports sufficient frags. */
if (LEPP_MAX_FRAGS >= 1 + MAX_SKB_FRAGS)
- dev->features |= NETIF_F_TSO;
-
-#ifdef TILE_NET_GSO
- /* We support GSO. */
- dev->features |= NETIF_F_GSO;
-#endif
+ features |= NETIF_F_TSO;
+ /* We can't support HIGHDMA without hash_default, since we need
+ * to be able to finv() with a VA if we don't have hash_default.
+ */
if (hash_default)
- dev->features |= NETIF_F_HIGHDMA;
-
- /* ISSUE: We should support NETIF_F_UFO. */
+ features |= NETIF_F_HIGHDMA;
- dev->tx_queue_len = TILE_NET_TX_QUEUE_LEN;
-
- dev->mtu = TILE_NET_MTU;
+ dev->hw_features |= features;
+ dev->vlan_features |= features;
+ dev->features |= features & ~TILE_NETIF_F_DEFAULT_DISABLE;
}
--
1.8.3.1
^ permalink raw reply related
* Re: [PATCH 4/4] USBNET: ax88179_178a: enable tso if host supports sg dma
From: Grant Grundler @ 2013-08-01 15:30 UTC (permalink / raw)
To: Ming Lei
Cc: David S. Miller, Greg Kroah-Hartman, Oliver Neukum, Freddy Xin,
Eric Dumazet, Ben Hutchings, netdev, linux-usb
In-Reply-To: <1375267909-30373-5-git-send-email-ming.lei@canonical.com>
On Wed, Jul 31, 2013 at 3:51 AM, Ming Lei <ming.lei@canonical.com> wrote:
> This patch enables 'can_dma_sg' flag for ax88179_178a device
> if the attached host controller supports building packet from
> discontinuous buffers(DMA SG is possible), so both frame header
> and skb data buffers can be passed to usb stack via urb->sg,
> then skb data copy can be saved.
>
> With the patch, CPU utilization decreased much in iperf test at
> client mode.
>
> Signed-off-by: Ming Lei <ming.lei@canonical.com>
> ---
> drivers/net/usb/ax88179_178a.c | 30 ++++++++++++++++++++++++++++++
> 1 file changed, 30 insertions(+)
>
> diff --git a/drivers/net/usb/ax88179_178a.c b/drivers/net/usb/ax88179_178a.c
> index 5a468f3..c75bded 100644
> --- a/drivers/net/usb/ax88179_178a.c
> +++ b/drivers/net/usb/ax88179_178a.c
> @@ -1031,12 +1031,20 @@ static int ax88179_bind(struct usbnet *dev, struct usb_interface *intf)
> dev->mii.phy_id = 0x03;
> dev->mii.supports_gmii = 1;
>
> + if (dev->udev->bus->no_sg_limit)
> + dev->can_dma_sg = 1;
> +
> dev->net->features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
> NETIF_F_RXCSUM;
>
> dev->net->hw_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
> NETIF_F_RXCSUM;
>
> + if (dev->can_dma_sg) {
> + dev->net->features |= NETIF_F_SG | NETIF_F_TSO;
> + dev->net->hw_features |= NETIF_F_SG | NETIF_F_TSO;
> + }
> +
> /* Enable checksum offload */
> *tmp = AX_RXCOE_IP | AX_RXCOE_TCP | AX_RXCOE_UDP |
> AX_RXCOE_TCPV6 | AX_RXCOE_UDPV6;
> @@ -1170,12 +1178,30 @@ ax88179_tx_fixup(struct usbnet *dev, struct sk_buff *skb, gfp_t flags)
> int mss = skb_shinfo(skb)->gso_size;
> int headroom;
> int tailroom;
> + struct skb_data *entry = (struct skb_data *)skb->cb;
>
> tx_hdr1 = skb->len;
> tx_hdr2 = mss;
> if (((skb->len + 8) % frame_size) == 0)
> tx_hdr2 |= 0x80008000; /* Enable padding */
>
> + if (dev->can_dma_sg) {
> + if (!(entry->header = kmalloc(8, GFP_ATOMIC)))
> + goto no_sg;
> +
> + entry->length = 8;
> + cpu_to_le32s(&tx_hdr1);
> + cpu_to_le32s(&tx_hdr2);
http://lxr.free-electrons.com/source/include/linux/byteorder/generic.h#L111
http://lxr.free-electrons.com/ident?i=__cpu_to_le32s
IIRC, cpu_to_leXX() macros return the endian "corrected" value.
In other words, they need to be assigned to something, no?
thanks,
grant
> + memcpy(entry->header, &tx_hdr1, 4);
> + memcpy(entry->header + 4, &tx_hdr2, 4);
> +
> + return skb;
> + } else {
> +no_sg:
> + entry->header = NULL;
> + entry->length = 0;
> + }
> +
> headroom = skb_headroom(skb);
> tailroom = skb_tailroom(skb);
>
> @@ -1323,6 +1349,10 @@ static int ax88179_reset(struct usbnet *dev)
>
> dev->net->hw_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
> NETIF_F_RXCSUM;
> + if (dev->can_dma_sg) {
> + dev->net->features |= NETIF_F_SG | NETIF_F_TSO;
> + dev->net->hw_features |= NETIF_F_SG | NETIF_F_TSO;
> + }
>
> /* Enable checksum offload */
> *tmp = AX_RXCOE_IP | AX_RXCOE_TCP | AX_RXCOE_UDP |
> --
> 1.7.9.5
>
^ permalink raw reply
* [PATCH net-next] net: add SNMP counters tracking incoming ECN bits
From: Eric Dumazet @ 2013-08-01 15:26 UTC (permalink / raw)
To: David Miller; +Cc: netdev
From: Eric Dumazet <edumazet@google.com>
With GRO/LRO processing, there is a problem because Ip[6]InReceives SNMP
counters do not count the number of frames, but number of aggregated
segments.
Its probably too late to change this now.
This patch adds four new counters, tracking number of frames, regardless
of LRO/GRO, and on a per ECN status basis, for IPv4 and IPv6.
Ip[6]NoECTPkts : Number of packets received with NOECT
Ip[6]ECT1Pkts : Number of packets received with ECT(1)
Ip[6]ECT0Pkts : Number of packets received with ECT(0)
Ip[6]CEPkts : Number of packets received with Congestion Experimented
lph37:~# nstat | egrep "Pkts|InReceive"
IpInReceives 1634137 0.0
Ip6InReceives 3714107 0.0
Ip6InNoECTPkts 19205 0.0
Ip6InECT0Pkts 52651828 0.0
IpExtInNoECTPkts 33630 0.0
IpExtInECT0Pkts 15581379 0.0
IpExtInCEPkts 6 0.0
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
include/uapi/linux/snmp.h | 4 ++++
net/ipv4/ip_input.c | 8 ++++++++
net/ipv4/proc.c | 7 ++++++-
net/ipv6/ip6_input.c | 6 +++++-
net/ipv6/proc.c | 4 ++++
5 files changed, 27 insertions(+), 2 deletions(-)
diff --git a/include/uapi/linux/snmp.h b/include/uapi/linux/snmp.h
index af0a674..60601d2 100644
--- a/include/uapi/linux/snmp.h
+++ b/include/uapi/linux/snmp.h
@@ -51,6 +51,10 @@ enum
IPSTATS_MIB_INBCASTOCTETS, /* InBcastOctets */
IPSTATS_MIB_OUTBCASTOCTETS, /* OutBcastOctets */
IPSTATS_MIB_CSUMERRORS, /* InCsumErrors */
+ IPSTATS_MIB_NOECTPKTS, /* InNoECTPkts */
+ IPSTATS_MIB_ECT1PKTS, /* InECT1Pkts */
+ IPSTATS_MIB_ECT0PKTS, /* InECT0Pkts */
+ IPSTATS_MIB_CEPKTS, /* InCEPkts */
__IPSTATS_MIB_MAX
};
diff --git a/net/ipv4/ip_input.c b/net/ipv4/ip_input.c
index 15e3e68..054a3e9 100644
--- a/net/ipv4/ip_input.c
+++ b/net/ipv4/ip_input.c
@@ -141,6 +141,7 @@
#include <net/icmp.h>
#include <net/raw.h>
#include <net/checksum.h>
+#include <net/inet_ecn.h>
#include <linux/netfilter_ipv4.h>
#include <net/xfrm.h>
#include <linux/mroute.h>
@@ -410,6 +411,13 @@ int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
if (iph->ihl < 5 || iph->version != 4)
goto inhdr_error;
+ BUILD_BUG_ON(IPSTATS_MIB_ECT1PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_1);
+ BUILD_BUG_ON(IPSTATS_MIB_ECT0PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_0);
+ BUILD_BUG_ON(IPSTATS_MIB_CEPKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_CE);
+ IP_ADD_STATS_BH(dev_net(dev),
+ IPSTATS_MIB_NOECTPKTS + (iph->tos & INET_ECN_MASK),
+ max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
+
if (!pskb_may_pull(skb, iph->ihl*4))
goto inhdr_error;
diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c
index 6577a11..5f5fa61 100644
--- a/net/ipv4/proc.c
+++ b/net/ipv4/proc.c
@@ -111,7 +111,7 @@ static const struct snmp_mib snmp4_ipstats_list[] = {
SNMP_MIB_SENTINEL
};
-/* Following RFC4293 items are displayed in /proc/net/netstat */
+/* Following items are displayed in /proc/net/netstat */
static const struct snmp_mib snmp4_ipextstats_list[] = {
SNMP_MIB_ITEM("InNoRoutes", IPSTATS_MIB_INNOROUTES),
SNMP_MIB_ITEM("InTruncatedPkts", IPSTATS_MIB_INTRUNCATEDPKTS),
@@ -125,7 +125,12 @@ static const struct snmp_mib snmp4_ipextstats_list[] = {
SNMP_MIB_ITEM("OutMcastOctets", IPSTATS_MIB_OUTMCASTOCTETS),
SNMP_MIB_ITEM("InBcastOctets", IPSTATS_MIB_INBCASTOCTETS),
SNMP_MIB_ITEM("OutBcastOctets", IPSTATS_MIB_OUTBCASTOCTETS),
+ /* Non RFC4293 fields */
SNMP_MIB_ITEM("InCsumErrors", IPSTATS_MIB_CSUMERRORS),
+ SNMP_MIB_ITEM("InNoECTPkts", IPSTATS_MIB_NOECTPKTS),
+ SNMP_MIB_ITEM("InECT1Pkts", IPSTATS_MIB_ECT1PKTS),
+ SNMP_MIB_ITEM("InECT0Pkts", IPSTATS_MIB_ECT0PKTS),
+ SNMP_MIB_ITEM("InCEPkts", IPSTATS_MIB_CEPKTS),
SNMP_MIB_SENTINEL
};
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index 2bab2aa..302d6fb 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -44,7 +44,7 @@
#include <net/ip6_route.h>
#include <net/addrconf.h>
#include <net/xfrm.h>
-
+#include <net/inet_ecn.h>
int ip6_rcv_finish(struct sk_buff *skb)
@@ -109,6 +109,10 @@ int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt
if (hdr->version != 6)
goto err;
+ IP6_ADD_STATS_BH(dev_net(dev), idev,
+ IPSTATS_MIB_NOECTPKTS +
+ (ipv6_get_dsfield(hdr) & INET_ECN_MASK),
+ max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
/*
* RFC4291 2.5.3
* A packet received on an interface with a destination address
diff --git a/net/ipv6/proc.c b/net/ipv6/proc.c
index 51c3285..091d066 100644
--- a/net/ipv6/proc.c
+++ b/net/ipv6/proc.c
@@ -91,6 +91,10 @@ static const struct snmp_mib snmp6_ipstats_list[] = {
SNMP_MIB_ITEM("Ip6InBcastOctets", IPSTATS_MIB_INBCASTOCTETS),
SNMP_MIB_ITEM("Ip6OutBcastOctets", IPSTATS_MIB_OUTBCASTOCTETS),
/* IPSTATS_MIB_CSUMERRORS is not relevant in IPv6 (no checksum) */
+ SNMP_MIB_ITEM("Ip6InNoECTPkts", IPSTATS_MIB_NOECTPKTS),
+ SNMP_MIB_ITEM("Ip6InECT1Pkts", IPSTATS_MIB_ECT1PKTS),
+ SNMP_MIB_ITEM("Ip6InECT0Pkts", IPSTATS_MIB_ECT0PKTS),
+ SNMP_MIB_ITEM("Ip6InCEPkts", IPSTATS_MIB_CEPKTS),
SNMP_MIB_SENTINEL
};
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox