* [PATCH 2/2] net: ti cpsw ethernet: set IFCTL_{A,B} bits for RMII mode
From: Daniel Mack @ 2012-09-26 17:24 UTC (permalink / raw)
To: netdev
Cc: devicetree-discuss, Daniel Mack, Mugunthan V N, Vaibhav Hiremath,
David S. Miller
In-Reply-To: <1348680268-8194-1-git-send-email-zonque@gmail.com>
For RMII mode operation in 100Mbps, the CPSW needs to set the
IFCTL_A / IFCTL_B bits in the MACCONTROL register.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Cc: Mugunthan V N <mugunthanvnm@ti.com>
Cc: Vaibhav Hiremath <hvaibhav@ti.com>
Cc: David S. Miller <davem@davemloft.net>
---
drivers/net/ethernet/ti/cpsw.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/drivers/net/ethernet/ti/cpsw.c b/drivers/net/ethernet/ti/cpsw.c
index 3d7594e..d88dbfa 100644
--- a/drivers/net/ethernet/ti/cpsw.c
+++ b/drivers/net/ethernet/ti/cpsw.c
@@ -386,6 +386,12 @@ static void _cpsw_adjust_link(struct cpsw_slave *slave,
mac_control |= BIT(7); /* GIGABITEN */
if (phy->duplex)
mac_control |= BIT(0); /* FULLDUPLEXEN */
+
+ /* set speed_in input in case RMII mode is used in >10Mbps */
+ if (phy->speed > 10 && slave->slave_num < 2 &&
+ phy->interface == PHY_INTERFACE_MODE_RMII)
+ mac_control |= BIT(15 + slave->slave_num);
+
*link = true;
} else {
mac_control = 0;
--
1.7.11.4
^ permalink raw reply related
* [PATCH 1/2] net: ti cpsw ethernet: allow reading phy interface mode from DT
From: Daniel Mack @ 2012-09-26 17:24 UTC (permalink / raw)
To: netdev
Cc: devicetree-discuss, Daniel Mack, Mugunthan V N, Vaibhav Hiremath,
David S. Miller
Allow users to specify the phy interface of the CPSW slaves. The new
node parameter is called "phy_if_mode" and is optional. The original
behaviour of the driver is preserved when not given.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Cc: Mugunthan V N <mugunthanvnm@ti.com>
Cc: Vaibhav Hiremath <hvaibhav@ti.com>
Cc: David S. Miller <davem@davemloft.net>
---
Documentation/devicetree/bindings/net/cpsw.txt | 3 +++
drivers/net/ethernet/ti/cpsw.c | 3 +++
2 files changed, 6 insertions(+)
diff --git a/Documentation/devicetree/bindings/net/cpsw.txt b/Documentation/devicetree/bindings/net/cpsw.txt
index dcaabe9..d87f7d2 100644
--- a/Documentation/devicetree/bindings/net/cpsw.txt
+++ b/Documentation/devicetree/bindings/net/cpsw.txt
@@ -25,6 +25,8 @@ Required properties:
- slave_reg_ofs : Specifies slave register offset
- sliver_reg_ofs : Specifies slave sliver register offset
- phy_id : Specifies slave phy id
+- phy_if_mode : Specified slave phy interface mode (optional)
+ (one of the PHY_INTERFACE_MODE_* as numerical value)
- mac-address : Specifies slave MAC address
Optional properties:
@@ -62,6 +64,7 @@ Examples:
slave_reg_ofs = <0x208>;
sliver_reg_ofs = <0xd80>;
phy_id = "davinci_mdio.16:00";
+ phy_if_mode = <6>; /* PHY_INTERFACE_MODE_RGMII */
/* Filled in by U-Boot */
mac-address = [ 00 00 00 00 00 00 ];
};
diff --git a/drivers/net/ethernet/ti/cpsw.c b/drivers/net/ethernet/ti/cpsw.c
index aa78168..3d7594e 100644
--- a/drivers/net/ethernet/ti/cpsw.c
+++ b/drivers/net/ethernet/ti/cpsw.c
@@ -850,6 +850,9 @@ static int cpsw_probe_dt(struct cpsw_platform_data *data,
}
slave_data->sliver_reg_ofs = prop;
+ if (!of_property_read_u32(slave_node, "phy_if_mode", &prop))
+ slave_data->phy_if = prop;
+
mac_addr = of_get_mac_address(slave_node);
if (mac_addr)
memcpy(slave_data->mac_addr, mac_addr, ETH_ALEN);
--
1.7.11.4
^ permalink raw reply related
* Re: [PATCH 02/11] inet_diag: pass inet_diag module to netlink_dump_start
From: Pablo Neira Ayuso @ 2012-09-26 17:26 UTC (permalink / raw)
To: Stephen Hemminger
Cc: Gao feng, davem-fT/PcQaiUtIeIZ0/mPfg9Q,
netfilter-devel-u79uwXL29TY76Z2rM5mHXA,
linux-rdma-u79uwXL29TY76Z2rM5mHXA, netdev-u79uwXL29TY76Z2rM5mHXA,
eric.dumazet-Re5JQEeQqe8AvxtiuMwx3w,
steffen.klassert-opNxpl+3fjRBDgjK7y7TUQ,
linux-crypto-u79uwXL29TY76Z2rM5mHXA, jengelh-9+2X+4sQBs8,
stephen.hemminger-ZtmgI6mnKB3QT0dZR+AlfA
In-Reply-To: <20120926090714.618a4b39-We1ePj4FEcvRI77zikRAJc56i+j3xesD0e7PPNI6Mm0@public.gmane.org>
On Wed, Sep 26, 2012 at 09:07:14AM -0700, Stephen Hemminger wrote:
> On Wed, 26 Sep 2012 12:52:11 +0800
> Gao feng <gaofeng-BthXqXjhjHXQFUHtdCDX3A@public.gmane.org> wrote:
>
> > struct netlink_dump_control c = {
> > .dump = inet_diag_dump_compat,
> > + .done = netlink_dump_done,
> > + .module = THIS_MODULE,
> > };
>
> Since these are immutable, why the netlink_dump_control structures
> on the stack (which requires code to generate at runtime), and instead
> make them global
>
> static const struct netlink_dump_control unix_diag_ctrl = {
> ...
I made that code. I can send a patch to statify those
netlink_dump_control structures.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" 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
* Re: [PATCH v5 01/10] ipc: remove forced assignment of selected message
From: Serge Hallyn @ 2012-09-26 17:37 UTC (permalink / raw)
To: Stanislav Kinsbursky
Cc: akpm, manfred, a.p.zijlstra, netdev, will.deacon, linux-kernel,
cmetcalf, jmorris, linux-driver, linux-security-module, hughd,
ron.mercer, viro, james.l.morris, catalin.marinas, casey, eparis,
sds, jitendra.kalsaria, devel
In-Reply-To: <20120919160531.11254.48182.stgit@localhost6.localdomain6>
Quoting Stanislav Kinsbursky (skinsbursky@parallels.com):
> This is a cleanup patch. The assignment is redundant.
>
> Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
> ---
> ipc/msg.c | 1 -
> 1 files changed, 0 insertions(+), 1 deletions(-)
>
> diff --git a/ipc/msg.c b/ipc/msg.c
> index 7385de2..f3bfbb8 100644
> --- a/ipc/msg.c
> +++ b/ipc/msg.c
> @@ -787,7 +787,6 @@ long do_msgrcv(int msqid, long *pmtype, void __user *mtext,
> !security_msg_queue_msgrcv(msq, walk_msg, current,
> msgtyp, mode)) {
>
> - msg = walk_msg;
> if (mode == SEARCH_LESSEQUAL &&
> walk_msg->m_type != 1) {
> msg = walk_msg;
Perhaps your tree is different from mine, but it looks to me like it would
be simpler to remove the 'msg = walk_msg' from both the 'if' and 'else',
and keep them above the if/else?
^ permalink raw reply
* Re: [PATCH] bnx2: Clean up remaining iounmap
From: Michael Chan @ 2012-09-26 18:33 UTC (permalink / raw)
To: Neil Horman; +Cc: netdev, Michael Chan, David S. Miller
In-Reply-To: <1348680122-9711-1-git-send-email-nhorman@tuxdriver.com>
On Wed, 2012-09-26 at 13:22 -0400, Neil Horman wrote:
> commit c0357e975afdbbedab5c662d19bef865f02adc17 modified bnx2 to switch from
> using ioremap/iounmap to pci_iomap/pci_iounmap. They missed a spot in the error
> path of bnx2_init_one though. This patch just cleans that up.
>
> Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
> CC: Michael Chan <mcan@broadcom.com>
> CC: "David S. Miller" <davem@davemloft.net>
Acked-by: Michael Chan <mchan@broadcom.com>
> ---
> drivers/net/ethernet/broadcom/bnx2.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/broadcom/bnx2.c b/drivers/net/ethernet/broadcom/bnx2.c
> index 79cebd8..e48312f 100644
> --- a/drivers/net/ethernet/broadcom/bnx2.c
> +++ b/drivers/net/ethernet/broadcom/bnx2.c
> @@ -8564,7 +8564,7 @@ bnx2_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
> return 0;
>
> error:
> - iounmap(bp->regview);
> + pci_iounmap(pdev, bp->regview);
> pci_release_regions(pdev);
> pci_disable_device(pdev);
> pci_set_drvdata(pdev, NULL);
^ permalink raw reply
* Re: ixgbe unstable performance at 1Gb/s
From: Charles Vejnar @ 2012-09-26 18:33 UTC (permalink / raw)
To: Tantilov, Emil S, netdev@vger.kernel.org
In-Reply-To: <87618083B2453E4A8714035B62D6799216E25153@FMSMSX102.amr.corp.intel.com>
Le 25/09/2012 19:58, Tantilov, Emil S a écrit :
>> -----Original Message-----
>> From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org] On
>> Behalf Of Charles
>> Sent: Monday, September 24, 2012 10:47 AM
>> To: netdev@vger.kernel.org
>> Subject: ixgbe unstable performance at 1Gb/s
>>
>> Hi,
>>
>> I hope I am posting on the right mailing-list. If not, sorry; please
>> redirect me
>> to the right place. Thanks.
>>
>> I have a new motherboard with integrated Intel X540 10GBase-T. For now, I
>> want
>> to use it at 1Gb/s.
>>
>> The bandwidth is only of ~300 Mbit/s (with Iperf). It's actually very
>> unstable
>> (always varies between 100 to 800 Mbit/s during the transfer).
> Do you by any chance have CONFIG_IXGBE_PTP set in your kernel config?
>
> If so, try disabling it and see if it fixes your performance.
>
> Thanks,
> Emil
>
Hi,
Thanks for your reply.
I compiled manually the ixgbe module with the default options of my
distribution (Archlinux). I had the same problem.
I then changed the CONFIG_IXGBE_PTP to no (default is yes) as you
suggested, and recompiled. The problem disappeared; normal transfer.
Could you please explain why this PTP is causing a problem? How can this
be fixed without having to recompile the module (ethtool, /sys, bios...
)? Thanks
Regards,
Charles
^ permalink raw reply
* RE: [PATCH 2/2] net: ti cpsw ethernet: set IFCTL_{A,B} bits for RMII mode
From: N, Mugunthan V @ 2012-09-26 18:50 UTC (permalink / raw)
To: Daniel Mack, netdev@vger.kernel.org
Cc: devicetree-discuss@lists.ozlabs.org, Hiremath, Vaibhav,
David S. Miller
In-Reply-To: <1348680268-8194-2-git-send-email-zonque@gmail.com>
> -----Original Message-----
> From: Daniel Mack [mailto:zonque@gmail.com]
> Sent: Wednesday, September 26, 2012 10:54 PM
> To: netdev@vger.kernel.org
> Cc: devicetree-discuss@lists.ozlabs.org; Daniel Mack; N, Mugunthan V;
> Hiremath, Vaibhav; David S. Miller
> Subject: [PATCH 2/2] net: ti cpsw ethernet: set IFCTL_{A,B} bits for
> RMII mode
>
> For RMII mode operation in 100Mbps, the CPSW needs to set the
> IFCTL_A / IFCTL_B bits in the MACCONTROL register.
>
> Signed-off-by: Daniel Mack <zonque@gmail.com>
> Cc: Mugunthan V N <mugunthanvnm@ti.com>
> Cc: Vaibhav Hiremath <hvaibhav@ti.com>
> Cc: David S. Miller <davem@davemloft.net>
> ---
> drivers/net/ethernet/ti/cpsw.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/drivers/net/ethernet/ti/cpsw.c
> b/drivers/net/ethernet/ti/cpsw.c
> index 3d7594e..d88dbfa 100644
> --- a/drivers/net/ethernet/ti/cpsw.c
> +++ b/drivers/net/ethernet/ti/cpsw.c
> @@ -386,6 +386,12 @@ static void _cpsw_adjust_link(struct cpsw_slave
> *slave,
> mac_control |= BIT(7); /* GIGABITEN */
> if (phy->duplex)
> mac_control |= BIT(0); /* FULLDUPLEXEN */
> +
> + /* set speed_in input in case RMII mode is used in >10Mbps
> */
> + if (phy->speed > 10 && slave->slave_num < 2 &&
> + phy->interface == PHY_INTERFACE_MODE_RMII)
> + mac_control |= BIT(15 + slave->slave_num);
Mac control register is separate for both the slaves and has same bit definitions,
Bit 15 has to be set for 100Mbps link for RMII and RGMII Phy interface to control
the RMII/RGMII gasket and in GMII this bit is Un-used by CPSW.
For slave 1, Bit 16 is set with the above code which is not used control the
RMII/RGMII gasket control. So it is not required to pass the Phy mode from DT.
This patch has to be reworked to set Bit 15 with any Phy mode connected.
The original driver present was tested with GMII (Beagle Bone A5) and
RGMII (AM3358 EVM) phy , but CPSW works fine without setting this bit in
RGMII phymode so this issue was not caught in testing.
Regards
Mugunthan V N
> +
> *link = true;
> } else {
> mac_control = 0;
> --
> 1.7.11.4
^ permalink raw reply
* [PATCH] rapidio/rionet: fix multicast packet transmit logic
From: Alexandre Bounine @ 2012-09-26 19:14 UTC (permalink / raw)
To: Andrew Morton, linux-kernel, linuxppc-dev
Cc: Alexandre Bounine, Matt Porter, David S. Miller, netdev
Fix multicast packet transmit logic to account for repetitive transmission
of single skb:
- correct check for available buffers (this bug may produce NULL pointer
crash dump in case of heavy traffic);
- update skb user count (incorrect user counter causes a warning dump from
net_tx_action routine during multicast transfers in systems with three or
more rionet participants).
Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
---
drivers/net/rionet.c | 20 +++++++++++++++++---
1 files changed, 17 insertions(+), 3 deletions(-)
diff --git a/drivers/net/rionet.c b/drivers/net/rionet.c
index 91d2588..1470d3e 100644
--- a/drivers/net/rionet.c
+++ b/drivers/net/rionet.c
@@ -79,6 +79,7 @@ static int rionet_capable = 1;
* on system trade-offs.
*/
static struct rio_dev **rionet_active;
+static int nact; /* total number of active rionet peers */
#define is_rionet_capable(src_ops, dst_ops) \
((src_ops & RIO_SRC_OPS_DATA_MSG) && \
@@ -175,6 +176,7 @@ static int rionet_start_xmit(struct sk_buff *skb, struct net_device *ndev)
struct ethhdr *eth = (struct ethhdr *)skb->data;
u16 destid;
unsigned long flags;
+ int add_num = 1;
local_irq_save(flags);
if (!spin_trylock(&rnet->tx_lock)) {
@@ -182,7 +184,10 @@ static int rionet_start_xmit(struct sk_buff *skb, struct net_device *ndev)
return NETDEV_TX_LOCKED;
}
- if ((rnet->tx_cnt + 1) > RIONET_TX_RING_SIZE) {
+ if (is_multicast_ether_addr(eth->h_dest))
+ add_num = nact;
+
+ if ((rnet->tx_cnt + add_num) > RIONET_TX_RING_SIZE) {
netif_stop_queue(ndev);
spin_unlock_irqrestore(&rnet->tx_lock, flags);
printk(KERN_ERR "%s: BUG! Tx Ring full when queue awake!\n",
@@ -191,11 +196,16 @@ static int rionet_start_xmit(struct sk_buff *skb, struct net_device *ndev)
}
if (is_multicast_ether_addr(eth->h_dest)) {
+ int count = 0;
for (i = 0; i < RIO_MAX_ROUTE_ENTRIES(rnet->mport->sys_size);
i++)
- if (rionet_active[i])
+ if (rionet_active[i]) {
rionet_queue_tx_msg(skb, ndev,
rionet_active[i]);
+ if (count)
+ atomic_inc(&skb->users);
+ count++;
+ }
} else if (RIONET_MAC_MATCH(eth->h_dest)) {
destid = RIONET_GET_DESTID(eth->h_dest);
if (rionet_active[destid])
@@ -220,14 +230,17 @@ static void rionet_dbell_event(struct rio_mport *mport, void *dev_id, u16 sid, u
if (info == RIONET_DOORBELL_JOIN) {
if (!rionet_active[sid]) {
list_for_each_entry(peer, &rionet_peers, node) {
- if (peer->rdev->destid == sid)
+ if (peer->rdev->destid == sid) {
rionet_active[sid] = peer->rdev;
+ nact++;
+ }
}
rio_mport_send_doorbell(mport, sid,
RIONET_DOORBELL_JOIN);
}
} else if (info == RIONET_DOORBELL_LEAVE) {
rionet_active[sid] = NULL;
+ nact--;
} else {
if (netif_msg_intr(rnet))
printk(KERN_WARNING "%s: unhandled doorbell\n",
@@ -523,6 +536,7 @@ static int rionet_probe(struct rio_dev *rdev, const struct rio_device_id *id)
rc = rionet_setup_netdev(rdev->net->hport, ndev);
rionet_check = 1;
+ nact = 0;
}
/*
--
1.7.8.4
^ permalink raw reply related
* RE: ixgbe unstable performance at 1Gb/s
From: Tantilov, Emil S @ 2012-09-26 20:22 UTC (permalink / raw)
To: Charles Vejnar, netdev@vger.kernel.org; +Cc: Kirsher, Jeffrey T
In-Reply-To: <50634A60.40609@unige.ch>
>-----Original Message-----
>From: Charles Vejnar [mailto:Charles.Vejnar@unige.ch]
>Sent: Wednesday, September 26, 2012 11:33 AM
>To: Tantilov, Emil S; netdev@vger.kernel.org
>Subject: Re: ixgbe unstable performance at 1Gb/s
>
>Le 25/09/2012 19:58, Tantilov, Emil S a écrit :
>>> -----Original Message-----
>>> From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org]
>On
>>> Behalf Of Charles
>>> Sent: Monday, September 24, 2012 10:47 AM
>>> To: netdev@vger.kernel.org
>>> Subject: ixgbe unstable performance at 1Gb/s
>>>
>>> Hi,
>>>
>>> I hope I am posting on the right mailing-list. If not, sorry; please
>>> redirect me
>>> to the right place. Thanks.
>>>
>>> I have a new motherboard with integrated Intel X540 10GBase-T. For now,
>I
>>> want
>>> to use it at 1Gb/s.
>>>
>>> The bandwidth is only of ~300 Mbit/s (with Iperf). It's actually very
>>> unstable
>>> (always varies between 100 to 800 Mbit/s during the transfer).
>> Do you by any chance have CONFIG_IXGBE_PTP set in your kernel config?
>>
>> If so, try disabling it and see if it fixes your performance.
>>
>> Thanks,
>> Emil
>>
>Hi,
>
>Thanks for your reply.
>
>I compiled manually the ixgbe module with the default options of my
>distribution (Archlinux). I had the same problem.
>
>I then changed the CONFIG_IXGBE_PTP to no (default is yes) as you
>suggested, and recompiled. The problem disappeared; normal transfer.
>
>Could you please explain why this PTP is causing a problem? How can this
>be fixed without having to recompile the module (ethtool, /sys, bios...
>)? Thanks
This is actually a bug in the driver. We should have a patch out very soon to address it.
>
>Regards,
>
>Charles
Thanks,
Emil
^ permalink raw reply
* Re: [PATCH V4 0/7] ipvs: IPv6 fragment handling for IPVS
From: Julian Anastasov @ 2012-09-26 20:50 UTC (permalink / raw)
To: Jesper Dangaard Brouer
Cc: Hans Schillstrom, Hans Schillstrom, netdev, Pablo Neira Ayuso,
lvs-devel, Patrick McHardy, Thomas Graf, Wensong Zhang,
netfilter-devel, Simon Horman
In-Reply-To: <20120926120534.24804.78415.stgit@dragon>
Hello,
On Wed, 26 Sep 2012, Jesper Dangaard Brouer wrote:
> The following patchset implement IPv6 fragment handling for IPVS.
>
> This work is based upon patches from Hans Schillstrom. I have taken
> over the patchset, in close agreement with Hans, because he don't have
> (gotten allocated) time to complete his work.
>
> I have cleaned up the patchset significantly, and split the patchset
> up into seven patches.
>
> The first 3 patches, are ready to be merged
>
> Patch01: Trivial changes, use compressed IPv6 address in output
> Patch02: IPv6 extend ICMPv6 handling for future types
> Patch03: Use config macro IS_ENABLED()
>
> The next 4 patches, is V4 of the patches I have submitted earlier.
> Where I have incorporated Julian's recent feedback.
>
> - Notice that patch04 of patchset V3, have been dropped.
>
> I have also tried to make the patches easier to review, by
> reorganizing the changes, to be more strictly split (exthdr
> vs. fragment handling).
>
> I have also removed the API changes, and moved those to patch06. This
> is done, (1) to make it easier to review the patches, and (2) to allow
> easier integration of Patricks idea and my RFC patch of caching exthdr
> info in skb->cb[]. Thus, we can get these patches applied (and later
> go back and apply the caching scheme easier).
>
> Patch04: Fix faulty IPv6 extension header handling in IPVS
> Patch05: Complete IPv6 fragment handling for IPVS
> Patch06: IPVS API change to avoid rescan of IPv6 exthdr
> Patch07: IPVS SIP fragment handling
>
> The SIP frag handling have been split into its own patch, as I have
> not been able to test this part my self.
>
> This patchset is based upon:
> Pablo's nf-next tree: git://1984.lsi.us.es/nf-next
> On top of:
> commit 2cbc78a29e76a2e92c172651204f3117491877d2
> (netfilter: combine ipt_REDIRECT and ip6t_REDIRECT)
>
> ---
>
> Jesper Dangaard Brouer (7):
> ipvs: SIP fragment handling
> ipvs: API change to avoid rescan of IPv6 exthdr
> ipvs: Complete IPv6 fragment handling for IPVS
> ipvs: Fix faulty IPv6 extension header handling in IPVS
> ipvs: Use config macro IS_ENABLED()
> ipvs: IPv6 extend ICMPv6 handling for future types
> ipvs: Trivial changes, use compressed IPv6 address in output
All 7 patches look good to me. Thanks!
Acked-by: Julian Anastasov <ja@ssi.bg>
> include/net/ip_vs.h | 194 +++++++++++----
> net/netfilter/ipvs/Kconfig | 7 -
> net/netfilter/ipvs/ip_vs_conn.c | 15 -
> net/netfilter/ipvs/ip_vs_core.c | 404 +++++++++++++++++--------------
> net/netfilter/ipvs/ip_vs_dh.c | 2
> net/netfilter/ipvs/ip_vs_lblc.c | 2
> net/netfilter/ipvs/ip_vs_lblcr.c | 2
> net/netfilter/ipvs/ip_vs_pe_sip.c | 21 +-
> net/netfilter/ipvs/ip_vs_proto.c | 6
> net/netfilter/ipvs/ip_vs_proto_ah_esp.c | 9 -
> net/netfilter/ipvs/ip_vs_proto_sctp.c | 42 +--
> net/netfilter/ipvs/ip_vs_proto_tcp.c | 40 +--
> net/netfilter/ipvs/ip_vs_proto_udp.c | 41 +--
> net/netfilter/ipvs/ip_vs_sched.c | 2
> net/netfilter/ipvs/ip_vs_sh.c | 2
> net/netfilter/ipvs/ip_vs_xmit.c | 73 +++---
> net/netfilter/xt_ipvs.c | 4
> 17 files changed, 504 insertions(+), 362 deletions(-)
>
>
> --
> Best regards,
> Jesper Dangaard Brouer
> MSc.CS, Sr. Network Kernel Developer at Red Hat
> Author of http://www.iptv-analyzer.org
> LinkedIn: http://www.linkedin.com/in/brouer
Regards
--
Julian Anastasov <ja@ssi.bg>
^ permalink raw reply
* Re: Problems with tg3 on BCM5720
From: Nithin Nayak Sujir @ 2012-09-26 21:13 UTC (permalink / raw)
To: Dirkjan Ochtman; +Cc: netdev
In-Reply-To: <CAKmKYaCBG5hURNomffgp9gRhS7g2BYsNG-LTmBrKsX-iyVzVxg@mail.gmail.com>
Hi Dirkjan,
A few questions -
1. Can you tell me the last patch that is included in the tg3 driver in 3.4.9 on your distro?
2. Can you give more info about the working setup?
3. Was there any system reset or driver reload between the working and not working setups? Or was it
just a cable switch?
4. Please give the output of
ethtool eth0
ethtool -i eth0
ethtool -k eth0
5. Can you run ethtool --test in the working setup?
6. I noticed in the syslog, the link is coming up at 100 Mbps. Is this expected?
7. Does it fail immediately on connect to the data center switch? Or is it after some traffic goes
through?
Thanks,
Nithin.
On 09/26/2012 06:48 AM, Dirkjan Ochtman wrote:
> Hi,
>
> I hope this is the right list for this; please let me know if there is
> some more appropriate venue for this kind of problem.
>
> We have bought a Dell PowerEdge server recently with Broadcom ethernet
> controllers:
>
> 01:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5720
> Gigabit Ethernet PCIe
> 01:00.1 Ethernet controller: Broadcom Corporation NetXtreme BCM5720
> Gigabit Ethernet PCIe
> 02:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5720
> Gigabit Ethernet PCIe
> 02:00.1 Ethernet controller: Broadcom Corporation NetXtreme BCM5720
> Gigabit Ethernet PCIe
>
> Using the tg3 driver as provided by the 3.4.9 kernel (it's a distro
> kernel, but I verified that there are no relevant patches on top of
> the vanilla 3.4.9), networking works fine on a small VLAN. However,
> we've run into trouble connecting one of the interfaces to switch run
> by our data center vendor (the switch side works fine when connecting
> with a laptop). The driver keeps erroring out with "transmit timed
> out, resetting" (full syslog output is at [1]). This also keeps
> happening when I use the latest tg3.{c,h} from torvalds/linux-2.6
> (though I've had to revert a few things to make it compile inside the
> 3.4.9 kernel).
>
> I've also run some tests with ethtool:
>
> djc@jansky ~ $ sudo ethtool --test eth0
> The test result is FAIL
> The test extra info:
> nvram test (online) 0
> link test (online) 1
> register test (offline) 0
> memory test (offline) 0
> mac loopback test (offline) 7
> phy loopback test (offline) 7
> ext loopback test (offline) 0
> interrupt test (offline) 1
>
> While I'd be happy to do more digging, at this point I'm out of ideas
> on how to go about that.
>
> Again, sorry if this is off-topic for this list, any help would be appreciated.
>
> Cheers,
>
> Dirkjan
>
> [1] http://dirkjan.ochtman.nl/files/tg3-fail.txt
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
^ permalink raw reply
* Re: Problems with tg3 on BCM5720
From: Michael Chan @ 2012-09-26 21:40 UTC (permalink / raw)
To: Dirkjan Ochtman; +Cc: netdev
In-Reply-To: <CAKmKYaCBG5hURNomffgp9gRhS7g2BYsNG-LTmBrKsX-iyVzVxg@mail.gmail.com>
On Wed, 2012-09-26 at 15:48 +0200, Dirkjan Ochtman wrote:
> I've also run some tests with ethtool:
>
> djc@jansky ~ $ sudo ethtool --test eth0
> The test result is FAIL
> The test extra info:
> nvram test (online) 0
> link test (online) 1
> register test (offline) 0
> memory test (offline) 0
> mac loopback test (offline) 7
> phy loopback test (offline) 7
> ext loopback test (offline) 0
> interrupt test (offline) 1
It is most likely that the device eth0 is down. The device needs to be
up in order to perform all the tests that failed. Please bring up the
device and run the test again. Thanks.
>
>
^ permalink raw reply
* [PATCH] tcp: Remove unused parameter from tcp_v4_save_options
From: Christoph Paasch @ 2012-09-26 21:59 UTC (permalink / raw)
To: David Miller; +Cc: netdev
struct sock *sk is not used inside tcp_v4_save_options. Thus it can be
removed.
Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
---
net/ipv4/tcp_ipv4.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 93406c5..385eb79 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -925,8 +925,7 @@ EXPORT_SYMBOL(tcp_syn_flood_action);
/*
* Save and compile IPv4 options into the request_sock if needed.
*/
-static struct ip_options_rcu *tcp_v4_save_options(struct sock *sk,
- struct sk_buff *skb)
+static struct ip_options_rcu *tcp_v4_save_options(struct sk_buff *skb)
{
const struct ip_options *opt = &(IPCB(skb)->opt);
struct ip_options_rcu *dopt = NULL;
@@ -1568,7 +1567,7 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
ireq->loc_addr = daddr;
ireq->rmt_addr = saddr;
ireq->no_srccheck = inet_sk(sk)->transparent;
- ireq->opt = tcp_v4_save_options(sk, skb);
+ ireq->opt = tcp_v4_save_options(skb);
if (security_inet_conn_request(sk, skb, req))
goto drop_and_free;
--
1.7.9.5
^ permalink raw reply related
* [PATCH 1/7 net-next] tg3: Introduce separate functions to allocate/free RX/TX rings.
From: Michael Chan @ 2012-09-26 22:32 UTC (permalink / raw)
To: davem; +Cc: netdev
This is preparation work to allow the number of RX and TX rings to be
configured separately.
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/tg3.c | 220 +++++++++++++++++++++++------------
drivers/net/ethernet/broadcom/tg3.h | 2 +
2 files changed, 146 insertions(+), 76 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index bf906c5..93b8120 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -7607,15 +7607,11 @@ static int tg3_init_rings(struct tg3 *tp)
return 0;
}
-/*
- * Must not be invoked with interrupt sources disabled and
- * the hardware shutdown down.
- */
-static void tg3_free_consistent(struct tg3 *tp)
+static void tg3_mem_tx_release(struct tg3 *tp)
{
int i;
- for (i = 0; i < tp->irq_cnt; i++) {
+ for (i = 0; i < tp->irq_max; i++) {
struct tg3_napi *tnapi = &tp->napi[i];
if (tnapi->tx_ring) {
@@ -7626,17 +7622,114 @@ static void tg3_free_consistent(struct tg3 *tp)
kfree(tnapi->tx_buffers);
tnapi->tx_buffers = NULL;
+ }
+}
- if (tnapi->rx_rcb) {
- dma_free_coherent(&tp->pdev->dev,
- TG3_RX_RCB_RING_BYTES(tp),
- tnapi->rx_rcb,
- tnapi->rx_rcb_mapping);
- tnapi->rx_rcb = NULL;
- }
+static int tg3_mem_tx_acquire(struct tg3 *tp)
+{
+ int i;
+ struct tg3_napi *tnapi = &tp->napi[0];
+
+ /* If multivector TSS is enabled, vector 0 does not handle
+ * tx interrupts. Don't allocate any resources for it.
+ */
+ if (tg3_flag(tp, ENABLE_TSS))
+ tnapi++;
+
+ for (i = 0; i < tp->txq_cnt; i++, tnapi++) {
+ tnapi->tx_buffers = kzalloc(sizeof(struct tg3_tx_ring_info) *
+ TG3_TX_RING_SIZE, GFP_KERNEL);
+ if (!tnapi->tx_buffers)
+ goto err_out;
+
+ tnapi->tx_ring = dma_alloc_coherent(&tp->pdev->dev,
+ TG3_TX_RING_BYTES,
+ &tnapi->tx_desc_mapping,
+ GFP_KERNEL);
+ if (!tnapi->tx_ring)
+ goto err_out;
+ }
+
+ return 0;
+
+err_out:
+ tg3_mem_tx_release(tp);
+ return -ENOMEM;
+}
+
+static void tg3_mem_rx_release(struct tg3 *tp)
+{
+ int i;
+
+ for (i = 0; i < tp->irq_max; i++) {
+ struct tg3_napi *tnapi = &tp->napi[i];
tg3_rx_prodring_fini(tp, &tnapi->prodring);
+ if (!tnapi->rx_rcb)
+ continue;
+
+ dma_free_coherent(&tp->pdev->dev,
+ TG3_RX_RCB_RING_BYTES(tp),
+ tnapi->rx_rcb,
+ tnapi->rx_rcb_mapping);
+ tnapi->rx_rcb = NULL;
+ }
+}
+
+static int tg3_mem_rx_acquire(struct tg3 *tp)
+{
+ unsigned int i, limit;
+
+ limit = tp->rxq_cnt;
+
+ /* If RSS is enabled, we need a (dummy) producer ring
+ * set on vector zero. This is the true hw prodring.
+ */
+ if (tg3_flag(tp, ENABLE_RSS))
+ limit++;
+
+ for (i = 0; i < limit; i++) {
+ struct tg3_napi *tnapi = &tp->napi[i];
+
+ if (tg3_rx_prodring_init(tp, &tnapi->prodring))
+ goto err_out;
+
+ /* If multivector RSS is enabled, vector 0
+ * does not handle rx or tx interrupts.
+ * Don't allocate any resources for it.
+ */
+ if (!i && tg3_flag(tp, ENABLE_RSS))
+ continue;
+
+ tnapi->rx_rcb = dma_alloc_coherent(&tp->pdev->dev,
+ TG3_RX_RCB_RING_BYTES(tp),
+ &tnapi->rx_rcb_mapping,
+ GFP_KERNEL);
+ if (!tnapi->rx_rcb)
+ goto err_out;
+
+ memset(tnapi->rx_rcb, 0, TG3_RX_RCB_RING_BYTES(tp));
+ }
+
+ return 0;
+
+err_out:
+ tg3_mem_rx_release(tp);
+ return -ENOMEM;
+}
+
+/*
+ * Must not be invoked with interrupt sources disabled and
+ * the hardware shutdown down.
+ */
+static void tg3_free_consistent(struct tg3 *tp)
+{
+ int i;
+
+ for (i = 0; i < tp->irq_cnt; i++) {
+ struct tg3_napi *tnapi = &tp->napi[i];
+
if (tnapi->hw_status) {
dma_free_coherent(&tp->pdev->dev, TG3_HW_STATUS_SIZE,
tnapi->hw_status,
@@ -7645,6 +7738,9 @@ static void tg3_free_consistent(struct tg3 *tp)
}
}
+ tg3_mem_rx_release(tp);
+ tg3_mem_tx_release(tp);
+
if (tp->hw_stats) {
dma_free_coherent(&tp->pdev->dev, sizeof(struct tg3_hw_stats),
tp->hw_stats, tp->stats_mapping);
@@ -7683,72 +7779,38 @@ static int tg3_alloc_consistent(struct tg3 *tp)
memset(tnapi->hw_status, 0, TG3_HW_STATUS_SIZE);
sblk = tnapi->hw_status;
- if (tg3_rx_prodring_init(tp, &tnapi->prodring))
- goto err_out;
+ if (tg3_flag(tp, ENABLE_RSS)) {
+ u16 *prodptr = 0;
- /* If multivector TSS is enabled, vector 0 does not handle
- * tx interrupts. Don't allocate any resources for it.
- */
- if ((!i && !tg3_flag(tp, ENABLE_TSS)) ||
- (i && tg3_flag(tp, ENABLE_TSS))) {
- tnapi->tx_buffers = kzalloc(
- sizeof(struct tg3_tx_ring_info) *
- TG3_TX_RING_SIZE, GFP_KERNEL);
- if (!tnapi->tx_buffers)
- goto err_out;
-
- tnapi->tx_ring = dma_alloc_coherent(&tp->pdev->dev,
- TG3_TX_RING_BYTES,
- &tnapi->tx_desc_mapping,
- GFP_KERNEL);
- if (!tnapi->tx_ring)
- goto err_out;
- }
-
- /*
- * When RSS is enabled, the status block format changes
- * slightly. The "rx_jumbo_consumer", "reserved",
- * and "rx_mini_consumer" members get mapped to the
- * other three rx return ring producer indexes.
- */
- switch (i) {
- default:
- if (tg3_flag(tp, ENABLE_RSS)) {
- tnapi->rx_rcb_prod_idx = NULL;
+ /*
+ * When RSS is enabled, the status block format changes
+ * slightly. The "rx_jumbo_consumer", "reserved",
+ * and "rx_mini_consumer" members get mapped to the
+ * other three rx return ring producer indexes.
+ */
+ switch (i) {
+ case 1:
+ prodptr = &sblk->idx[0].rx_producer;
+ break;
+ case 2:
+ prodptr = &sblk->rx_jumbo_consumer;
+ break;
+ case 3:
+ prodptr = &sblk->reserved;
+ break;
+ case 4:
+ prodptr = &sblk->rx_mini_consumer;
break;
}
- /* Fall through */
- case 1:
+ tnapi->rx_rcb_prod_idx = prodptr;
+ } else {
tnapi->rx_rcb_prod_idx = &sblk->idx[0].rx_producer;
- break;
- case 2:
- tnapi->rx_rcb_prod_idx = &sblk->rx_jumbo_consumer;
- break;
- case 3:
- tnapi->rx_rcb_prod_idx = &sblk->reserved;
- break;
- case 4:
- tnapi->rx_rcb_prod_idx = &sblk->rx_mini_consumer;
- break;
}
-
- /*
- * If multivector RSS is enabled, vector 0 does not handle
- * rx or tx interrupts. Don't allocate any resources for it.
- */
- if (!i && tg3_flag(tp, ENABLE_RSS))
- continue;
-
- tnapi->rx_rcb = dma_alloc_coherent(&tp->pdev->dev,
- TG3_RX_RCB_RING_BYTES(tp),
- &tnapi->rx_rcb_mapping,
- GFP_KERNEL);
- if (!tnapi->rx_rcb)
- goto err_out;
-
- memset(tnapi->rx_rcb, 0, TG3_RX_RCB_RING_BYTES(tp));
}
+ if (tg3_mem_tx_acquire(tp) || tg3_mem_rx_acquire(tp))
+ goto err_out;
+
return 0;
err_out:
@@ -10154,6 +10216,7 @@ static bool tg3_enable_msix(struct tg3 *tp)
* one to the number of vectors we are requesting.
*/
tp->irq_cnt = min_t(unsigned, tp->irq_cnt + 1, tp->irq_max);
+ tp->rxq_cnt = tp->irq_cnt - 1;
}
for (i = 0; i < tp->irq_max; i++) {
@@ -10170,14 +10233,13 @@ static bool tg3_enable_msix(struct tg3 *tp)
netdev_notice(tp->dev, "Requested %d MSI-X vectors, received %d\n",
tp->irq_cnt, rc);
tp->irq_cnt = rc;
+ tp->rxq_cnt = max(rc - 1, 1);
}
for (i = 0; i < tp->irq_max; i++)
tp->napi[i].irq_vec = msix_ent[i].vector;
- netif_set_real_num_tx_queues(tp->dev, 1);
- rc = tp->irq_cnt > 1 ? tp->irq_cnt - 1 : 1;
- if (netif_set_real_num_rx_queues(tp->dev, rc)) {
+ if (netif_set_real_num_rx_queues(tp->dev, tp->rxq_cnt)) {
pci_disable_msix(tp->pdev);
return false;
}
@@ -10188,7 +10250,8 @@ static bool tg3_enable_msix(struct tg3 *tp)
if (GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5719 ||
GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5720) {
tg3_flag_set(tp, ENABLE_TSS);
- netif_set_real_num_tx_queues(tp->dev, tp->irq_cnt - 1);
+ tp->txq_cnt = tp->rxq_cnt;
+ netif_set_real_num_tx_queues(tp->dev, tp->txq_cnt);
}
}
@@ -10224,6 +10287,11 @@ defcfg:
if (!tg3_flag(tp, USING_MSIX)) {
tp->irq_cnt = 1;
tp->napi[0].irq_vec = tp->pdev->irq;
+ }
+
+ if (tp->irq_cnt == 1) {
+ tp->txq_cnt = 1;
+ tp->rxq_cnt = 1;
netif_set_real_num_tx_queues(tp->dev, 1);
netif_set_real_num_rx_queues(tp->dev, 1);
}
diff --git a/drivers/net/ethernet/broadcom/tg3.h b/drivers/net/ethernet/broadcom/tg3.h
index 6d52cb2..5838dea 100644
--- a/drivers/net/ethernet/broadcom/tg3.h
+++ b/drivers/net/ethernet/broadcom/tg3.h
@@ -3037,6 +3037,7 @@ struct tg3 {
void (*write32_tx_mbox) (struct tg3 *, u32,
u32);
u32 dma_limit;
+ u32 txq_cnt;
/* begin "rx thread" cacheline section */
struct tg3_napi napi[TG3_IRQ_MAX_VECS];
@@ -3051,6 +3052,7 @@ struct tg3 {
u32 rx_std_max_post;
u32 rx_offset;
u32 rx_pkt_map_sz;
+ u32 rxq_cnt;
bool rx_refill;
--
1.7.1
^ permalink raw reply related
* [PATCH 5/7 net-next] tg3: Refactor tg3_close()
From: Michael Chan @ 2012-09-26 22:32 UTC (permalink / raw)
To: davem; +Cc: netdev
In-Reply-To: <1348698769-760-4-git-send-email-mchan@broadcom.com>
by introducing tg3_stop() that does the opposite of tg3_start(). This
function will be useful when adding the support for changing the numbe
of rx and tx rings.
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/tg3.c | 70 +++++++++++++++++++----------------
1 files changed, 38 insertions(+), 32 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index 9bd99ce..3f2197e 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -10448,6 +10448,43 @@ err_out1:
return err;
}
+static void tg3_stop(struct tg3 *tp)
+{
+ int i;
+
+ tg3_napi_disable(tp);
+ tg3_reset_task_cancel(tp);
+
+ netif_tx_disable(tp->dev);
+
+ tg3_timer_stop(tp);
+
+ tg3_hwmon_close(tp);
+
+ tg3_phy_stop(tp);
+
+ tg3_full_lock(tp, 1);
+
+ tg3_disable_ints(tp);
+
+ tg3_halt(tp, RESET_KIND_SHUTDOWN, 1);
+ tg3_free_rings(tp);
+ tg3_flag_clear(tp, INIT_COMPLETE);
+
+ tg3_full_unlock(tp);
+
+ for (i = tp->irq_cnt - 1; i >= 0; i--) {
+ struct tg3_napi *tnapi = &tp->napi[i];
+ free_irq(tnapi->irq_vec, tnapi);
+ }
+
+ tg3_ints_fini(tp);
+
+ tg3_napi_fini(tp);
+
+ tg3_free_consistent(tp);
+}
+
static int tg3_open(struct net_device *dev)
{
struct tg3 *tp = netdev_priv(dev);
@@ -10490,45 +10527,14 @@ static int tg3_open(struct net_device *dev)
static int tg3_close(struct net_device *dev)
{
- int i;
struct tg3 *tp = netdev_priv(dev);
- tg3_napi_disable(tp);
- tg3_reset_task_cancel(tp);
-
- netif_tx_stop_all_queues(dev);
-
- tg3_timer_stop(tp);
-
- tg3_hwmon_close(tp);
-
- tg3_phy_stop(tp);
-
- tg3_full_lock(tp, 1);
-
- tg3_disable_ints(tp);
-
- tg3_halt(tp, RESET_KIND_SHUTDOWN, 1);
- tg3_free_rings(tp);
- tg3_flag_clear(tp, INIT_COMPLETE);
-
- tg3_full_unlock(tp);
-
- for (i = tp->irq_cnt - 1; i >= 0; i--) {
- struct tg3_napi *tnapi = &tp->napi[i];
- free_irq(tnapi->irq_vec, tnapi);
- }
-
- tg3_ints_fini(tp);
+ tg3_stop(tp);
/* Clear stats across close / open calls */
memset(&tp->net_stats_prev, 0, sizeof(tp->net_stats_prev));
memset(&tp->estats_prev, 0, sizeof(tp->estats_prev));
- tg3_napi_fini(tp);
-
- tg3_free_consistent(tp);
-
tg3_power_down(tp);
netif_carrier_off(tp->dev);
--
1.7.1
^ permalink raw reply related
* [PATCH 3/7 net-next] tg3: Separate coalescing setup for rx and tx
From: Michael Chan @ 2012-09-26 22:32 UTC (permalink / raw)
To: davem; +Cc: netdev
In-Reply-To: <1348698769-760-2-git-send-email-mchan@broadcom.com>
since the number of rings can be different.
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/tg3.c | 74 +++++++++++++++++++++--------------
1 files changed, 45 insertions(+), 29 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index 330356b..ddf260c 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -8331,9 +8331,10 @@ static void tg3_set_bdinfo(struct tg3 *tp, u32 bdinfo_addr,
nic_addr);
}
-static void __tg3_set_coalesce(struct tg3 *tp, struct ethtool_coalesce *ec)
+
+static void tg3_coal_tx_init(struct tg3 *tp, struct ethtool_coalesce *ec)
{
- int i;
+ int i = 0;
if (!tg3_flag(tp, ENABLE_TSS)) {
tw32(HOSTCC_TXCOL_TICKS, ec->tx_coalesce_usecs);
@@ -8343,31 +8344,43 @@ static void __tg3_set_coalesce(struct tg3 *tp, struct ethtool_coalesce *ec)
tw32(HOSTCC_TXCOL_TICKS, 0);
tw32(HOSTCC_TXMAX_FRAMES, 0);
tw32(HOSTCC_TXCOAL_MAXF_INT, 0);
+
+ for (; i < tp->txq_cnt; i++) {
+ u32 reg;
+
+ reg = HOSTCC_TXCOL_TICKS_VEC1 + i * 0x18;
+ tw32(reg, ec->tx_coalesce_usecs);
+ reg = HOSTCC_TXMAX_FRAMES_VEC1 + i * 0x18;
+ tw32(reg, ec->tx_max_coalesced_frames);
+ reg = HOSTCC_TXCOAL_MAXF_INT_VEC1 + i * 0x18;
+ tw32(reg, ec->tx_max_coalesced_frames_irq);
+ }
+ }
+
+ for (; i < tp->irq_max - 1; i++) {
+ tw32(HOSTCC_TXCOL_TICKS_VEC1 + i * 0x18, 0);
+ tw32(HOSTCC_TXMAX_FRAMES_VEC1 + i * 0x18, 0);
+ tw32(HOSTCC_TXCOAL_MAXF_INT_VEC1 + i * 0x18, 0);
}
+}
+
+static void tg3_coal_rx_init(struct tg3 *tp, struct ethtool_coalesce *ec)
+{
+ int i = 0;
+ u32 limit = tp->rxq_cnt;
if (!tg3_flag(tp, ENABLE_RSS)) {
tw32(HOSTCC_RXCOL_TICKS, ec->rx_coalesce_usecs);
tw32(HOSTCC_RXMAX_FRAMES, ec->rx_max_coalesced_frames);
tw32(HOSTCC_RXCOAL_MAXF_INT, ec->rx_max_coalesced_frames_irq);
+ limit--;
} else {
tw32(HOSTCC_RXCOL_TICKS, 0);
tw32(HOSTCC_RXMAX_FRAMES, 0);
tw32(HOSTCC_RXCOAL_MAXF_INT, 0);
}
- if (!tg3_flag(tp, 5705_PLUS)) {
- u32 val = ec->stats_block_coalesce_usecs;
-
- tw32(HOSTCC_RXCOAL_TICK_INT, ec->rx_coalesce_usecs_irq);
- tw32(HOSTCC_TXCOAL_TICK_INT, ec->tx_coalesce_usecs_irq);
-
- if (!netif_carrier_ok(tp->dev))
- val = 0;
-
- tw32(HOSTCC_STAT_COAL_TICKS, val);
- }
-
- for (i = 0; i < tp->irq_cnt - 1; i++) {
+ for (; i < limit; i++) {
u32 reg;
reg = HOSTCC_RXCOL_TICKS_VEC1 + i * 0x18;
@@ -8376,27 +8389,30 @@ static void __tg3_set_coalesce(struct tg3 *tp, struct ethtool_coalesce *ec)
tw32(reg, ec->rx_max_coalesced_frames);
reg = HOSTCC_RXCOAL_MAXF_INT_VEC1 + i * 0x18;
tw32(reg, ec->rx_max_coalesced_frames_irq);
-
- if (tg3_flag(tp, ENABLE_TSS)) {
- reg = HOSTCC_TXCOL_TICKS_VEC1 + i * 0x18;
- tw32(reg, ec->tx_coalesce_usecs);
- reg = HOSTCC_TXMAX_FRAMES_VEC1 + i * 0x18;
- tw32(reg, ec->tx_max_coalesced_frames);
- reg = HOSTCC_TXCOAL_MAXF_INT_VEC1 + i * 0x18;
- tw32(reg, ec->tx_max_coalesced_frames_irq);
- }
}
for (; i < tp->irq_max - 1; i++) {
tw32(HOSTCC_RXCOL_TICKS_VEC1 + i * 0x18, 0);
tw32(HOSTCC_RXMAX_FRAMES_VEC1 + i * 0x18, 0);
tw32(HOSTCC_RXCOAL_MAXF_INT_VEC1 + i * 0x18, 0);
+ }
+}
- if (tg3_flag(tp, ENABLE_TSS)) {
- tw32(HOSTCC_TXCOL_TICKS_VEC1 + i * 0x18, 0);
- tw32(HOSTCC_TXMAX_FRAMES_VEC1 + i * 0x18, 0);
- tw32(HOSTCC_TXCOAL_MAXF_INT_VEC1 + i * 0x18, 0);
- }
+static void __tg3_set_coalesce(struct tg3 *tp, struct ethtool_coalesce *ec)
+{
+ tg3_coal_tx_init(tp, ec);
+ tg3_coal_rx_init(tp, ec);
+
+ if (!tg3_flag(tp, 5705_PLUS)) {
+ u32 val = ec->stats_block_coalesce_usecs;
+
+ tw32(HOSTCC_RXCOAL_TICK_INT, ec->rx_coalesce_usecs_irq);
+ tw32(HOSTCC_TXCOAL_TICK_INT, ec->tx_coalesce_usecs_irq);
+
+ if (!netif_carrier_ok(tp->dev))
+ val = 0;
+
+ tw32(HOSTCC_STAT_COAL_TICKS, val);
}
}
--
1.7.1
^ permalink raw reply related
* [PATCH 6/7 net-next] tg3: Add support for ethtool -L|-l to get/set the number of rings.
From: Michael Chan @ 2012-09-26 22:32 UTC (permalink / raw)
To: davem; +Cc: netdev
In-Reply-To: <1348698769-760-5-git-send-email-mchan@broadcom.com>
Default remains the same.
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/tg3.c | 64 +++++++++++++++++++++++++++++++++--
drivers/net/ethernet/broadcom/tg3.h | 2 +
2 files changed, 63 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index 3f2197e..74eea2f 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -10239,11 +10239,15 @@ static bool tg3_enable_msix(struct tg3 *tp)
int i, rc;
struct msix_entry msix_ent[tp->irq_max];
- tp->rxq_cnt = netif_get_num_default_rss_queues();
+ tp->txq_cnt = tp->txq_req;
+ tp->rxq_cnt = tp->rxq_req;
+ if (!tp->rxq_cnt)
+ tp->rxq_cnt = netif_get_num_default_rss_queues();
if (tp->rxq_cnt > tp->rxq_max)
tp->rxq_cnt = tp->rxq_max;
- if (GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5719 ||
- GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5720)
+ if ((GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5719 ||
+ GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5720) &&
+ !tp->txq_req)
tp->txq_cnt = min(tp->rxq_cnt, tp->txq_max);
tp->irq_cnt = tg3_irq_count(tp);
@@ -11384,6 +11388,58 @@ static int tg3_set_rxfh_indir(struct net_device *dev, const u32 *indir)
return 0;
}
+static void tg3_get_channels(struct net_device *dev,
+ struct ethtool_channels *channel)
+{
+ struct tg3 *tp = netdev_priv(dev);
+ u32 deflt_qs = netif_get_num_default_rss_queues();
+
+ channel->max_rx = tp->rxq_max;
+ channel->max_tx = tp->txq_max;
+
+ if (netif_running(dev)) {
+ channel->rx_count = tp->rxq_cnt;
+ channel->tx_count = tp->txq_cnt;
+ } else {
+ if (tp->rxq_req)
+ channel->rx_count = tp->rxq_req;
+ else
+ channel->rx_count = min(deflt_qs, tp->rxq_max);
+
+ if (tp->txq_req)
+ channel->tx_count = tp->txq_req;
+ else
+ channel->tx_count = min(deflt_qs, tp->txq_max);
+ }
+}
+
+static int tg3_set_channels(struct net_device *dev,
+ struct ethtool_channels *channel)
+{
+ struct tg3 *tp = netdev_priv(dev);
+
+ if (!tg3_flag(tp, SUPPORT_MSIX))
+ return -EOPNOTSUPP;
+
+ if (channel->rx_count > tp->rxq_max ||
+ channel->tx_count > tp->txq_max)
+ return -EINVAL;
+
+ tp->rxq_req = channel->rx_count;
+ tp->txq_req = channel->tx_count;
+
+ if (!netif_running(dev))
+ return 0;
+
+ tg3_stop(tp);
+
+ netif_carrier_off(dev);
+
+ tg3_start(tp, true, false);
+
+ return 0;
+}
+
static void tg3_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
{
switch (stringset) {
@@ -12632,6 +12688,8 @@ static const struct ethtool_ops tg3_ethtool_ops = {
.get_rxfh_indir_size = tg3_get_rxfh_indir_size,
.get_rxfh_indir = tg3_get_rxfh_indir,
.set_rxfh_indir = tg3_set_rxfh_indir,
+ .get_channels = tg3_get_channels,
+ .set_channels = tg3_set_channels,
.get_ts_info = ethtool_op_get_ts_info,
};
diff --git a/drivers/net/ethernet/broadcom/tg3.h b/drivers/net/ethernet/broadcom/tg3.h
index 2abe94c..d9308c3 100644
--- a/drivers/net/ethernet/broadcom/tg3.h
+++ b/drivers/net/ethernet/broadcom/tg3.h
@@ -3038,6 +3038,7 @@ struct tg3 {
void (*write32_tx_mbox) (struct tg3 *, u32,
u32);
u32 dma_limit;
+ u32 txq_req;
u32 txq_cnt;
u32 txq_max;
@@ -3054,6 +3055,7 @@ struct tg3 {
u32 rx_std_max_post;
u32 rx_offset;
u32 rx_pkt_map_sz;
+ u32 rxq_req;
u32 rxq_cnt;
u32 rxq_max;
bool rx_refill;
--
1.7.1
^ permalink raw reply related
* [PATCH 4/7 net-next] tg3: Refactor tg3_open()
From: Michael Chan @ 2012-09-26 22:32 UTC (permalink / raw)
To: davem; +Cc: netdev
In-Reply-To: <1348698769-760-3-git-send-email-mchan@broadcom.com>
by introducing tg3_start() that handles all initialization steps from
IRQ allocation. This function will be needed when adding support for
changing the number of rx and tx rings.
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/tg3.c | 78 ++++++++++++++++++++---------------
1 files changed, 45 insertions(+), 33 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index ddf260c..9bd99ce 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -10339,38 +10339,11 @@ static void tg3_ints_fini(struct tg3 *tp)
tg3_flag_clear(tp, ENABLE_TSS);
}
-static int tg3_open(struct net_device *dev)
+static int tg3_start(struct tg3 *tp, bool reset_phy, bool test_irq)
{
- struct tg3 *tp = netdev_priv(dev);
+ struct net_device *dev = tp->dev;
int i, err;
- if (tp->fw_needed) {
- err = tg3_request_firmware(tp);
- if (tp->pci_chip_rev_id == CHIPREV_ID_5701_A0) {
- if (err)
- return err;
- } else if (err) {
- netdev_warn(tp->dev, "TSO capability disabled\n");
- tg3_flag_clear(tp, TSO_CAPABLE);
- } else if (!tg3_flag(tp, TSO_CAPABLE)) {
- netdev_notice(tp->dev, "TSO capability restored\n");
- tg3_flag_set(tp, TSO_CAPABLE);
- }
- }
-
- netif_carrier_off(tp->dev);
-
- err = tg3_power_up(tp);
- if (err)
- return err;
-
- tg3_full_lock(tp, 0);
-
- tg3_disable_ints(tp);
- tg3_flag_clear(tp, INIT_COMPLETE);
-
- tg3_full_unlock(tp);
-
/*
* Setup interrupts first so we know how
* many NAPI resources to allocate
@@ -10404,7 +10377,7 @@ static int tg3_open(struct net_device *dev)
tg3_full_lock(tp, 0);
- err = tg3_init_hw(tp, 1);
+ err = tg3_init_hw(tp, reset_phy);
if (err) {
tg3_halt(tp, RESET_KIND_SHUTDOWN, 1);
tg3_free_rings(tp);
@@ -10415,7 +10388,7 @@ static int tg3_open(struct net_device *dev)
if (err)
goto err_out3;
- if (tg3_flag(tp, USING_MSI)) {
+ if (test_irq && tg3_flag(tp, USING_MSI)) {
err = tg3_test_msi(tp);
if (err) {
@@ -10471,8 +10444,47 @@ err_out2:
err_out1:
tg3_ints_fini(tp);
- tg3_frob_aux_power(tp, false);
- pci_set_power_state(tp->pdev, PCI_D3hot);
+
+ return err;
+}
+
+static int tg3_open(struct net_device *dev)
+{
+ struct tg3 *tp = netdev_priv(dev);
+ int err;
+
+ if (tp->fw_needed) {
+ err = tg3_request_firmware(tp);
+ if (tp->pci_chip_rev_id == CHIPREV_ID_5701_A0) {
+ if (err)
+ return err;
+ } else if (err) {
+ netdev_warn(tp->dev, "TSO capability disabled\n");
+ tg3_flag_clear(tp, TSO_CAPABLE);
+ } else if (!tg3_flag(tp, TSO_CAPABLE)) {
+ netdev_notice(tp->dev, "TSO capability restored\n");
+ tg3_flag_set(tp, TSO_CAPABLE);
+ }
+ }
+
+ netif_carrier_off(tp->dev);
+
+ err = tg3_power_up(tp);
+ if (err)
+ return err;
+
+ tg3_full_lock(tp, 0);
+
+ tg3_disable_ints(tp);
+ tg3_flag_clear(tp, INIT_COMPLETE);
+
+ tg3_full_unlock(tp);
+
+ err = tg3_start(tp, true, true);
+ if (err) {
+ tg3_frob_aux_power(tp, false);
+ pci_set_power_state(tp->pdev, PCI_D3hot);
+ }
return err;
}
--
1.7.1
^ permalink raw reply related
* [PATCH 7/7 net-next] tg3: Change default number of tx rings to 1.
From: Michael Chan @ 2012-09-26 22:32 UTC (permalink / raw)
To: davem; +Cc: netdev
In-Reply-To: <1348698769-760-6-git-send-email-mchan@broadcom.com>
Hardware tx scheduling can cause some starvation of a tx ring with small
packets if other tx rings have jumbo or TSO packets. The default setting
of 1 TX ring gives the best overall performance in many common traffic
scenarios. The user can change it using ethttol -L if desired.
Update version to 3.125.
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/tg3.c | 11 +++++------
1 files changed, 5 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index 74eea2f..3dff2ef 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -92,10 +92,10 @@ static inline void _tg3_flag_clear(enum TG3_FLAGS flag, unsigned long *bits)
#define DRV_MODULE_NAME "tg3"
#define TG3_MAJ_NUM 3
-#define TG3_MIN_NUM 124
+#define TG3_MIN_NUM 125
#define DRV_MODULE_VERSION \
__stringify(TG3_MAJ_NUM) "." __stringify(TG3_MIN_NUM)
-#define DRV_MODULE_RELDATE "March 21, 2012"
+#define DRV_MODULE_RELDATE "September 26, 2012"
#define RESET_KIND_SHUTDOWN 0
#define RESET_KIND_INIT 1
@@ -10245,10 +10245,9 @@ static bool tg3_enable_msix(struct tg3 *tp)
tp->rxq_cnt = netif_get_num_default_rss_queues();
if (tp->rxq_cnt > tp->rxq_max)
tp->rxq_cnt = tp->rxq_max;
- if ((GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5719 ||
- GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5720) &&
- !tp->txq_req)
- tp->txq_cnt = min(tp->rxq_cnt, tp->txq_max);
+
+ if (!tp->txq_req)
+ tp->txq_cnt = 1;
tp->irq_cnt = tg3_irq_count(tp);
--
1.7.1
^ permalink raw reply related
* [PATCH 2/7 net-next] tg3: Allow number of rx and tx rings to be set independently.
From: Michael Chan @ 2012-09-26 22:32 UTC (permalink / raw)
To: davem; +Cc: netdev
In-Reply-To: <1348698769-760-1-git-send-email-mchan@broadcom.com>
irq_cnt is no longer necessarily equal to the number rx or tx rings.
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
---
drivers/net/ethernet/broadcom/tg3.c | 74 +++++++++++++++++++++++------------
drivers/net/ethernet/broadcom/tg3.h | 5 ++-
2 files changed, 53 insertions(+), 26 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index 93b8120..330356b 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -6278,7 +6278,7 @@ static int tg3_poll_work(struct tg3_napi *tnapi, int work_done, int budget)
u32 jmb_prod_idx = dpr->rx_jmb_prod_idx;
tp->rx_refill = false;
- for (i = 1; i < tp->irq_cnt; i++)
+ for (i = 1; i <= tp->rxq_cnt; i++)
err |= tg3_rx_prodring_xfer(tp, dpr,
&tp->napi[i].prodring);
@@ -8654,13 +8654,12 @@ static void __tg3_set_rx_mode(struct net_device *dev)
}
}
-static void tg3_rss_init_dflt_indir_tbl(struct tg3 *tp)
+static void tg3_rss_init_dflt_indir_tbl(struct tg3 *tp, u32 qcnt)
{
int i;
for (i = 0; i < TG3_RSS_INDIR_TBL_SIZE; i++)
- tp->rss_ind_tbl[i] =
- ethtool_rxfh_indir_default(i, tp->irq_cnt - 1);
+ tp->rss_ind_tbl[i] = ethtool_rxfh_indir_default(i, qcnt);
}
static void tg3_rss_check_indir_tbl(struct tg3 *tp)
@@ -8682,7 +8681,7 @@ static void tg3_rss_check_indir_tbl(struct tg3 *tp)
}
if (i != TG3_RSS_INDIR_TBL_SIZE)
- tg3_rss_init_dflt_indir_tbl(tp);
+ tg3_rss_init_dflt_indir_tbl(tp, tp->rxq_cnt);
}
static void tg3_rss_write_indir_tbl(struct tg3 *tp)
@@ -10203,22 +10202,36 @@ static int tg3_request_firmware(struct tg3 *tp)
return 0;
}
-static bool tg3_enable_msix(struct tg3 *tp)
+static u32 tg3_irq_count(struct tg3 *tp)
{
- int i, rc;
- struct msix_entry msix_ent[tp->irq_max];
+ u32 irq_cnt = max(tp->rxq_cnt, tp->txq_cnt);
- tp->irq_cnt = netif_get_num_default_rss_queues();
- if (tp->irq_cnt > 1) {
+ if (irq_cnt > 1) {
/* We want as many rx rings enabled as there are cpus.
* In multiqueue MSI-X mode, the first MSI-X vector
* only deals with link interrupts, etc, so we add
* one to the number of vectors we are requesting.
*/
- tp->irq_cnt = min_t(unsigned, tp->irq_cnt + 1, tp->irq_max);
- tp->rxq_cnt = tp->irq_cnt - 1;
+ irq_cnt = min_t(unsigned, irq_cnt + 1, tp->irq_max);
}
+ return irq_cnt;
+}
+
+static bool tg3_enable_msix(struct tg3 *tp)
+{
+ int i, rc;
+ struct msix_entry msix_ent[tp->irq_max];
+
+ tp->rxq_cnt = netif_get_num_default_rss_queues();
+ if (tp->rxq_cnt > tp->rxq_max)
+ tp->rxq_cnt = tp->rxq_max;
+ if (GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5719 ||
+ GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5720)
+ tp->txq_cnt = min(tp->rxq_cnt, tp->txq_max);
+
+ tp->irq_cnt = tg3_irq_count(tp);
+
for (i = 0; i < tp->irq_max; i++) {
msix_ent[i].entry = i;
msix_ent[i].vector = 0;
@@ -10234,6 +10247,8 @@ static bool tg3_enable_msix(struct tg3 *tp)
tp->irq_cnt, rc);
tp->irq_cnt = rc;
tp->rxq_cnt = max(rc - 1, 1);
+ if (tp->txq_cnt)
+ tp->txq_cnt = min(tp->rxq_cnt, tp->txq_max);
}
for (i = 0; i < tp->irq_max; i++)
@@ -10244,16 +10259,15 @@ static bool tg3_enable_msix(struct tg3 *tp)
return false;
}
- if (tp->irq_cnt > 1) {
- tg3_flag_set(tp, ENABLE_RSS);
+ if (tp->irq_cnt == 1)
+ return true;
- if (GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5719 ||
- GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5720) {
- tg3_flag_set(tp, ENABLE_TSS);
- tp->txq_cnt = tp->rxq_cnt;
- netif_set_real_num_tx_queues(tp->dev, tp->txq_cnt);
- }
- }
+ tg3_flag_set(tp, ENABLE_RSS);
+
+ if (tp->txq_cnt > 1)
+ tg3_flag_set(tp, ENABLE_TSS);
+
+ netif_set_real_num_tx_queues(tp->dev, tp->txq_cnt);
return true;
}
@@ -11275,11 +11289,11 @@ static int tg3_get_rxnfc(struct net_device *dev, struct ethtool_rxnfc *info,
switch (info->cmd) {
case ETHTOOL_GRXRINGS:
if (netif_running(tp->dev))
- info->data = tp->irq_cnt;
+ info->data = tp->rxq_cnt;
else {
info->data = num_online_cpus();
- if (info->data > TG3_IRQ_MAX_VECS_RSS)
- info->data = TG3_IRQ_MAX_VECS_RSS;
+ if (info->data > TG3_RSS_MAX_NUM_QS)
+ info->data = TG3_RSS_MAX_NUM_QS;
}
/* The first interrupt vector only
@@ -14600,10 +14614,20 @@ static int __devinit tg3_get_invariants(struct tg3 *tp)
if (tg3_flag(tp, 57765_PLUS)) {
tg3_flag_set(tp, SUPPORT_MSIX);
tp->irq_max = TG3_IRQ_MAX_VECS;
- tg3_rss_init_dflt_indir_tbl(tp);
}
}
+ tp->txq_max = 1;
+ tp->rxq_max = 1;
+ if (tp->irq_max > 1) {
+ tp->rxq_max = TG3_RSS_MAX_NUM_QS;
+ tg3_rss_init_dflt_indir_tbl(tp, TG3_RSS_MAX_NUM_QS);
+
+ if (GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5719 ||
+ GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5720)
+ tp->txq_max = tp->irq_max - 1;
+ }
+
if (tg3_flag(tp, 5755_PLUS) ||
GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5906)
tg3_flag_set(tp, SHORT_DMA_BUG);
diff --git a/drivers/net/ethernet/broadcom/tg3.h b/drivers/net/ethernet/broadcom/tg3.h
index 5838dea..2abe94c 100644
--- a/drivers/net/ethernet/broadcom/tg3.h
+++ b/drivers/net/ethernet/broadcom/tg3.h
@@ -2860,7 +2860,8 @@ struct tg3_rx_prodring_set {
dma_addr_t rx_jmb_mapping;
};
-#define TG3_IRQ_MAX_VECS_RSS 5
+#define TG3_RSS_MAX_NUM_QS 4
+#define TG3_IRQ_MAX_VECS_RSS (TG3_RSS_MAX_NUM_QS + 1)
#define TG3_IRQ_MAX_VECS TG3_IRQ_MAX_VECS_RSS
struct tg3_napi {
@@ -3038,6 +3039,7 @@ struct tg3 {
u32);
u32 dma_limit;
u32 txq_cnt;
+ u32 txq_max;
/* begin "rx thread" cacheline section */
struct tg3_napi napi[TG3_IRQ_MAX_VECS];
@@ -3053,6 +3055,7 @@ struct tg3 {
u32 rx_offset;
u32 rx_pkt_map_sz;
u32 rxq_cnt;
+ u32 rxq_max;
bool rx_refill;
--
1.7.1
^ permalink raw reply related
* RE: [PATCH net-next] be2net: Ignore spurious UE indication from NIC
From: Khaparde, Ajit @ 2012-09-26 22:35 UTC (permalink / raw)
To: David Miller; +Cc: netdev@vger.kernel.org
In-Reply-To: <20120921.150413.2303624982785706643.davem@davemloft.net>
> From: David Miller [davem@davemloft.net]
> Sent: Friday, September 21, 2012 2:04 PM
> To: Khaparde, Ajit
> Cc: netdev@vger.kernel.org
> Subject: Re: [PATCH net-next] be2net: Ignore spurious UE indication from NIC
>
> From: Ajit Khaparde <ajit.khaparde@emulex.com>
> Date: Fri, 21 Sep 2012 11:36:20 -0500
>
>> Ignore spurious UE indication seen on some platforms.
>> Consider the error as un-recoverable only when the bits
>> stay high during second sampling.
>>
>> Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
>
> Treating uncorrectable errors as spurious seems like an invitation
> for hard to track down data corruption to me.
>
> You'll need to come up with a more sophisticated test for
> spurious other than "happens more than once" before I'm willing
> to subject the entire world to this kind of potential problem.
If the UE is real, then the hardware will stop responding to requests.
The hardware block goes offline automatically and no traffic will flow.
After this the hardware will generate a register dump, which can be
retrived using ethtool.
A spurious UE or a data corruption will not generate this dump.
The detection logic is merely to inform the user about the failure
and also avoid any further access to the NIC. This will also prevent
the driver unload from having to timeout on each access to the hardware
which could take a long time to complete.
Please let me know if this is enough to differentiate the scenarios.
Thanks
-Ajit
^ permalink raw reply
* [PATCH 0/26 v3] rdma/cm: Add support for native InfiniBand addressing
From: Hefty, Sean @ 2012-09-26 22:35 UTC (permalink / raw)
To: linux-rdma (linux-rdma@vger.kernel.org), netdev@vger.kernel.org
missed copying netdev on patch cover letter...
This patch series adds the ability to handle native Infiniband addressing
to the rdma_cm.
In addition to supporting native addresses, this support allows us to offload
name and/or address translation services to a user space daemon, providing the
user greater control over path selection. For example, a user can specify
different paths for multiple connections to the same system, or specify a
backup path for automatic path migration capabilities. The primary motivation
is to support large scale fabrics, with address and name services either cached
or bypassed completely (information is exchanged out of band by an MPI process
manager).
The intended usage model is for applications to call rdma_getaddrinfo. This
call is similar to getaddrinfo, but specific for RDMA devices.
rdma_getaddrinfo takes as input a name, IP address, GID, or LID, and returns
full connection information.
The patch set introduces af_ib and sockaddr_ib. The kernel rdma_cm is updated
accordingly, mainly to make its handling of addresses more generic. However,
since sockaddr_ib is larger than sockaddr_in6, the rdma_ucm requires changes to
its user to kernel interface. To provide backwards compatibility, the
userspace ABI is extended to support the larger address size.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
changes from v2:
updated to later kernel
^ permalink raw reply
* Re: [PATCH net-next] be2net: Ignore spurious UE indication from NIC
From: David Miller @ 2012-09-26 23:33 UTC (permalink / raw)
To: Ajit.Khaparde; +Cc: netdev
In-Reply-To: <3BB8C5C6B04A1F4CBE5B241DA98A501E67EC7B@CMEXMB1.ad.emulex.com>
From: "Khaparde, Ajit" <Ajit.Khaparde@Emulex.Com>
Date: Wed, 26 Sep 2012 22:35:14 +0000
>> From: David Miller [davem@davemloft.net]
>> Sent: Friday, September 21, 2012 2:04 PM
>> To: Khaparde, Ajit
>> Cc: netdev@vger.kernel.org
>> Subject: Re: [PATCH net-next] be2net: Ignore spurious UE indication from NIC
>>
>> From: Ajit Khaparde <ajit.khaparde@emulex.com>
>> Date: Fri, 21 Sep 2012 11:36:20 -0500
>>
>>> Ignore spurious UE indication seen on some platforms.
>>> Consider the error as un-recoverable only when the bits
>>> stay high during second sampling.
>>>
>>> Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
>>
>> Treating uncorrectable errors as spurious seems like an invitation
>> for hard to track down data corruption to me.
>>
>> You'll need to come up with a more sophisticated test for
>> spurious other than "happens more than once" before I'm willing
>> to subject the entire world to this kind of potential problem.
>
> If the UE is real, then the hardware will stop responding to requests.
> The hardware block goes offline automatically and no traffic will flow.
> After this the hardware will generate a register dump, which can be
> retrived using ethtool.
>
> A spurious UE or a data corruption will not generate this dump.
>
> The detection logic is merely to inform the user about the failure
> and also avoid any further access to the NIC.
So it sounds more like the UE test should be removed completely
since you cannot rely upon it, and if a real UE happens the hardware
will stop and you will already report the problem when the hardware
block goes offline.
Furthermore, it also implies that you can test if the hardware block
is offline to validate the UE indication.
All of which says irrefutably that your patch is still the wrong way
to do this.
^ permalink raw reply
* Re: [patch v2 04/11] nf_conntrack_netlink: pass nf_conntrack_netlink module to netlink_dump_start
From: Gao feng @ 2012-09-27 0:14 UTC (permalink / raw)
To: Pablo Neira Ayuso
Cc: davem, eric.dumazet, steffen.klassert, netfilter-devel,
linux-rdma, netdev, linux-crypto, stephen.hemminger, jengelh
In-Reply-To: <20120926150430.GA12438@1984>
于 2012年09月26日 23:04, Pablo Neira Ayuso 写道:
> On Wed, Sep 26, 2012 at 08:35:53PM +0800, Gao feng wrote:
>> 于 2012年09月26日 17:58, Pablo Neira Ayuso 写道:
>>> On Wed, Sep 26, 2012 at 05:42:31PM +0800, Gao feng wrote:
>>>> Hi Pablo:
>>>>
>>>> 于 2012年09月26日 17:26, Pablo Neira Ayuso 写道:
>>>>> On Wed, Sep 26, 2012 at 04:41:21PM +0800, Gao feng wrote:
>>>>>> use proper netlink_dump_control.done and .module to avoid panic.
>>>>>>
>>>>>> Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
>>>>>> ---
>>>>>> net/netfilter/nf_conntrack_netlink.c | 8 ++++++++
>>>>>> 1 files changed, 8 insertions(+), 0 deletions(-)
>>>>>>
>>>>>> diff --git a/net/netfilter/nf_conntrack_netlink.c b/net/netfilter/nf_conntrack_netlink.c
>>>>>> index 9807f32..509a257 100644
>>>>>> --- a/net/netfilter/nf_conntrack_netlink.c
>>>>>> +++ b/net/netfilter/nf_conntrack_netlink.c
>>>>>> @@ -706,6 +706,7 @@ static int ctnetlink_done(struct netlink_callback *cb)
>>>>>> nf_ct_put((struct nf_conn *)cb->args[1]);
>>>>>> if (cb->data)
>>>>>> kfree(cb->data);
>>>>>> + netlink_dump_done(cb);
>>>>>
>>>>> I think you can call netlink_dump_done from af_netlink.c:
>>>>>
>>>>> static int netlink_dump(struct sock *sk)
>>>>> ...
>>>>> if (cb->done) {
>>>>> cb->done(cb);
>>>>> netlink_dump_done(...);
>>>>> }
>>>>>
>>>>> Thus, you don't need to change netlink_dump_control in every netlink
>>>>> subsystem.
>>>>
>>>> because cb->done is called by netlink_sock_destruct too,it's very usefully
>>>> when userspace program only send dump request to kernel without reading
>>>> data from kernel.
>>>
>>> Then add that to netlink_sock_destruct as well. If possible, I prefer
>>> if this remains in the netlink core to avoid leaking module refcount
>>> if you forget to call netlink_dump_done.
>>
>> make sense, I will update it in next version.
>> Thanks!
>
> Great. Remove also netlink_dump_done and just use module_put instead
> after cb->done. That new function you added is so small that there is
> no way to justify its addition.
>
>>>
>>>>>
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> @@ -1022,6 +1023,7 @@ ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
>>>>>> struct netlink_dump_control c = {
>>>>>> .dump = ctnetlink_dump_table,
>>>>>> .done = ctnetlink_done,
>>>>>> + .module = THIS_MODULE,
>>>>>
>>>>> You can do something similar to:
>>>>>
>>>>> 9f00d97 netlink: hide struct module parameter in netlink_kernel_create
>>>>>
>>>>> by definiting netlink_dump_start as static inline and using
>>>>> THIS_MODULE from there.
>>>>>
>>>>> If I'm not missing anything, with those two changes, you will not need
>>>>> to modify any caller and it will result one single patch.
>>>>>
>>>>
>>>> You can see the patch [11/11], THIS_MODULE in infiniband/core/cma.c
>>>> means module rdma_cm,but we call netlink_dump_start in infiniband/core/netlink.c
>>>
>>> You can still use __netlink_dump_start for that case, which allows you
>>> to specify a custom struct module * parameter. But for most cases,
>>> netlink_dump_start (which hides THIS_MODULE) should be fine.
>>
>> I don't know how to deal with module_put in this way.
>> and I think my way is simple enough.
>
> Not sure what problem with module_put you're refering to.
>
forget this,I'm wrong.
> I think you can make this patchset way smaller with the change I'm
> proposing.
>
I know,but I choose simple and directviewing, not smaller.
I think these two function will make people confuse.
Thanks!
^ permalink raw reply
* [PATCH] netdev: pasemi: fix return value check in pasemi_mac_phy_init()
From: Wei Yongjun @ 2012-09-27 5:51 UTC (permalink / raw)
To: olof, grant.likely, rob.herring; +Cc: yongjun_wei, netdev, devicetree-discuss
From: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
In case of error, the function of_phy_connect() returns NULL
pointer not ERR_PTR(). The IS_ERR() test in the return value
check should be replaced with NULL test.
dpatch engine is used to auto generate this patch.
(https://github.com/weiyj/dpatch)
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
---
drivers/net/ethernet/pasemi/pasemi_mac.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/pasemi/pasemi_mac.c b/drivers/net/ethernet/pasemi/pasemi_mac.c
index e559dfa..6fa74d5 100644
--- a/drivers/net/ethernet/pasemi/pasemi_mac.c
+++ b/drivers/net/ethernet/pasemi/pasemi_mac.c
@@ -1101,9 +1101,9 @@ static int pasemi_mac_phy_init(struct net_device *dev)
phydev = of_phy_connect(dev, phy_dn, &pasemi_adjust_link, 0,
PHY_INTERFACE_MODE_SGMII);
- if (IS_ERR(phydev)) {
+ if (!phydev) {
printk(KERN_ERR "%s: Could not attach to phy\n", dev->name);
- return PTR_ERR(phydev);
+ return -ENODEV;
}
mac->phydev = phydev;
^ 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