* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Alexander Duyck @ 2012-07-26 18:26 UTC (permalink / raw)
To: Eric Dumazet; +Cc: David Miller, netdev
In-Reply-To: <1343324896.2626.11808.camel@edumazet-glaptop>
On Thu, Jul 26, 2012 at 10:48 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> On Thu, 2012-07-26 at 19:43 +0200, Eric Dumazet wrote:
>> On Thu, 2012-07-26 at 19:36 +0200, Eric Dumazet wrote:
>> > On Thu, 2012-07-26 at 19:31 +0200, Eric Dumazet wrote:
>> > > On Thu, 2012-07-26 at 10:18 -0700, Alexander Duyck wrote:
>> > >
>> > > > I tested this patch and it looks like it runs, but still has the same
>> > > > performance issue. I did some digging into the annotation for
>> > > > ip_route_intput_noref and it seems like the issue is that I am hitting
>> > > > the dst_hold call in __mkroute_input.
>> > >
>> > > David suggested a percpu cache.
>> > >
>> > > nh_rth_input would be allocated by alloc_percpu(struct dst *)
>> > >
>> > > I can work on this.
>> >
>> > Wait a minute, on input we should use the noref trick too.
>> >
>>
>> Something like : (on top of latest David patch)
>
> Sorry updated patch : (missing skb_dst_set() before 'out' label)
The previous results were with a slight modifications to your earlier
patch. With this patch applied I am seeing 10.4Mpps with 8 queues,
reaching a maximum of 11.6Mpps with 9 queues.
When we have a final version for this patch let me know and I will
give it a quick run and throw in my tested by.
Thanks,
Alex
^ permalink raw reply
* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Alexander Duyck @ 2012-07-26 18:06 UTC (permalink / raw)
To: Eric Dumazet; +Cc: David Miller, netdev
In-Reply-To: <1343324633.2626.11801.camel@edumazet-glaptop>
On Thu, Jul 26, 2012 at 10:43 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> On Thu, 2012-07-26 at 19:36 +0200, Eric Dumazet wrote:
>> On Thu, 2012-07-26 at 19:31 +0200, Eric Dumazet wrote:
>> > On Thu, 2012-07-26 at 10:18 -0700, Alexander Duyck wrote:
>> >
>> > > I tested this patch and it looks like it runs, but still has the same
>> > > performance issue. I did some digging into the annotation for
>> > > ip_route_intput_noref and it seems like the issue is that I am hitting
>> > > the dst_hold call in __mkroute_input.
>> >
>> > David suggested a percpu cache.
>> >
>> > nh_rth_input would be allocated by alloc_percpu(struct dst *)
>> >
>> > I can work on this.
>>
>> Wait a minute, on input we should use the noref trick too.
>>
>
> Something like : (on top of latest David patch)
>
> diff --git a/net/ipv4/route.c b/net/ipv4/route.c
> index 7a591aa..d5d2ad1 100644
> --- a/net/ipv4/route.c
> +++ b/net/ipv4/route.c
> @@ -1371,8 +1371,7 @@ static void ip_handle_martian_source(struct net_device *dev,
> static int __mkroute_input(struct sk_buff *skb,
> const struct fib_result *res,
> struct in_device *in_dev,
> - __be32 daddr, __be32 saddr, u32 tos,
> - struct rtable **result)
> + __be32 daddr, __be32 saddr, u32 tos)
> {
> struct rtable *rth;
> int err;
> @@ -1423,7 +1422,7 @@ static int __mkroute_input(struct sk_buff *skb,
> if (!itag) {
> rth = FIB_RES_NH(*res).nh_rth_input;
> if (rt_cache_valid(rth)) {
> - dst_hold(&rth->dst);
> + skb_dst_set_noref(skb, &rth->dst);
> goto out;
> }
> do_cache = true;
> @@ -1451,7 +1450,6 @@ static int __mkroute_input(struct sk_buff *skb,
>
> rt_set_nexthop(rth, daddr, res, NULL, res->fi, res->type, itag);
> out:
> - *result = rth;
> err = 0;
> cleanup:
> return err;
> @@ -1463,21 +1461,13 @@ static int ip_mkroute_input(struct sk_buff *skb,
> struct in_device *in_dev,
> __be32 daddr, __be32 saddr, u32 tos)
> {
> - struct rtable *rth = NULL;
> - int err;
> -
> #ifdef CONFIG_IP_ROUTE_MULTIPATH
> if (res->fi && res->fi->fib_nhs > 1)
> fib_select_multipath(res);
> #endif
>
> /* create a routing cache entry */
> - err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth);
> - if (err)
> - return err;
> -
> - skb_dst_set(skb, &rth->dst);
> - return 0;
> + return __mkroute_input(skb, res, in_dev, daddr, saddr, tos);
> }
>
> /*
>
With your changes in place I see an increase from 7.5Mpps to 9.9Mpps
for 8 queues, and increasing the queues to 9 gets me up to 11Mpps even
if the 9th queue is on another node. This is a HUGE improvement over
what we had before.
The only remaining overhead that has been introduced with the recent
changes appears to be the fib_table_lookup which doesn't have any hot
spots that jump out at me. The performance is in-line with what I was
seeing when I was randomly generating source IPs from a fairly large
set so I suspect this is just the expected behaviour without a routing
cache in place.
Thanks,
Alex
^ permalink raw reply
* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Eric Dumazet @ 2012-07-26 17:48 UTC (permalink / raw)
To: Alexander Duyck; +Cc: David Miller, netdev
In-Reply-To: <1343324633.2626.11801.camel@edumazet-glaptop>
On Thu, 2012-07-26 at 19:43 +0200, Eric Dumazet wrote:
> On Thu, 2012-07-26 at 19:36 +0200, Eric Dumazet wrote:
> > On Thu, 2012-07-26 at 19:31 +0200, Eric Dumazet wrote:
> > > On Thu, 2012-07-26 at 10:18 -0700, Alexander Duyck wrote:
> > >
> > > > I tested this patch and it looks like it runs, but still has the same
> > > > performance issue. I did some digging into the annotation for
> > > > ip_route_intput_noref and it seems like the issue is that I am hitting
> > > > the dst_hold call in __mkroute_input.
> > >
> > > David suggested a percpu cache.
> > >
> > > nh_rth_input would be allocated by alloc_percpu(struct dst *)
> > >
> > > I can work on this.
> >
> > Wait a minute, on input we should use the noref trick too.
> >
>
> Something like : (on top of latest David patch)
Sorry updated patch : (missing skb_dst_set() before 'out' label)
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 7a591aa..fc1a81c 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -1371,8 +1371,7 @@ static void ip_handle_martian_source(struct net_device *dev,
static int __mkroute_input(struct sk_buff *skb,
const struct fib_result *res,
struct in_device *in_dev,
- __be32 daddr, __be32 saddr, u32 tos,
- struct rtable **result)
+ __be32 daddr, __be32 saddr, u32 tos)
{
struct rtable *rth;
int err;
@@ -1423,7 +1422,7 @@ static int __mkroute_input(struct sk_buff *skb,
if (!itag) {
rth = FIB_RES_NH(*res).nh_rth_input;
if (rt_cache_valid(rth)) {
- dst_hold(&rth->dst);
+ skb_dst_set_noref(skb, &rth->dst);
goto out;
}
do_cache = true;
@@ -1450,8 +1449,8 @@ static int __mkroute_input(struct sk_buff *skb,
rth->dst.output = ip_output;
rt_set_nexthop(rth, daddr, res, NULL, res->fi, res->type, itag);
+ skb_dst_set(skb, &rth->dst);
out:
- *result = rth;
err = 0;
cleanup:
return err;
@@ -1463,21 +1462,13 @@ static int ip_mkroute_input(struct sk_buff *skb,
struct in_device *in_dev,
__be32 daddr, __be32 saddr, u32 tos)
{
- struct rtable *rth = NULL;
- int err;
-
#ifdef CONFIG_IP_ROUTE_MULTIPATH
if (res->fi && res->fi->fib_nhs > 1)
fib_select_multipath(res);
#endif
/* create a routing cache entry */
- err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth);
- if (err)
- return err;
-
- skb_dst_set(skb, &rth->dst);
- return 0;
+ return __mkroute_input(skb, res, in_dev, daddr, saddr, tos);
}
/*
^ permalink raw reply related
* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Eric Dumazet @ 2012-07-26 17:43 UTC (permalink / raw)
To: Alexander Duyck; +Cc: David Miller, netdev
In-Reply-To: <1343324175.2626.11790.camel@edumazet-glaptop>
On Thu, 2012-07-26 at 19:36 +0200, Eric Dumazet wrote:
> On Thu, 2012-07-26 at 19:31 +0200, Eric Dumazet wrote:
> > On Thu, 2012-07-26 at 10:18 -0700, Alexander Duyck wrote:
> >
> > > I tested this patch and it looks like it runs, but still has the same
> > > performance issue. I did some digging into the annotation for
> > > ip_route_intput_noref and it seems like the issue is that I am hitting
> > > the dst_hold call in __mkroute_input.
> >
> > David suggested a percpu cache.
> >
> > nh_rth_input would be allocated by alloc_percpu(struct dst *)
> >
> > I can work on this.
>
> Wait a minute, on input we should use the noref trick too.
>
Something like : (on top of latest David patch)
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 7a591aa..d5d2ad1 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -1371,8 +1371,7 @@ static void ip_handle_martian_source(struct net_device *dev,
static int __mkroute_input(struct sk_buff *skb,
const struct fib_result *res,
struct in_device *in_dev,
- __be32 daddr, __be32 saddr, u32 tos,
- struct rtable **result)
+ __be32 daddr, __be32 saddr, u32 tos)
{
struct rtable *rth;
int err;
@@ -1423,7 +1422,7 @@ static int __mkroute_input(struct sk_buff *skb,
if (!itag) {
rth = FIB_RES_NH(*res).nh_rth_input;
if (rt_cache_valid(rth)) {
- dst_hold(&rth->dst);
+ skb_dst_set_noref(skb, &rth->dst);
goto out;
}
do_cache = true;
@@ -1451,7 +1450,6 @@ static int __mkroute_input(struct sk_buff *skb,
rt_set_nexthop(rth, daddr, res, NULL, res->fi, res->type, itag);
out:
- *result = rth;
err = 0;
cleanup:
return err;
@@ -1463,21 +1461,13 @@ static int ip_mkroute_input(struct sk_buff *skb,
struct in_device *in_dev,
__be32 daddr, __be32 saddr, u32 tos)
{
- struct rtable *rth = NULL;
- int err;
-
#ifdef CONFIG_IP_ROUTE_MULTIPATH
if (res->fi && res->fi->fib_nhs > 1)
fib_select_multipath(res);
#endif
/* create a routing cache entry */
- err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth);
- if (err)
- return err;
-
- skb_dst_set(skb, &rth->dst);
- return 0;
+ return __mkroute_input(skb, res, in_dev, daddr, saddr, tos);
}
/*
^ permalink raw reply related
* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Eric Dumazet @ 2012-07-26 17:36 UTC (permalink / raw)
To: Alexander Duyck; +Cc: David Miller, netdev
In-Reply-To: <1343323858.2626.11783.camel@edumazet-glaptop>
On Thu, 2012-07-26 at 19:31 +0200, Eric Dumazet wrote:
> On Thu, 2012-07-26 at 10:18 -0700, Alexander Duyck wrote:
>
> > I tested this patch and it looks like it runs, but still has the same
> > performance issue. I did some digging into the annotation for
> > ip_route_intput_noref and it seems like the issue is that I am hitting
> > the dst_hold call in __mkroute_input.
>
> David suggested a percpu cache.
>
> nh_rth_input would be allocated by alloc_percpu(struct dst *)
>
> I can work on this.
Wait a minute, on input we should use the noref trick too.
^ permalink raw reply
* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Eric Dumazet @ 2012-07-26 17:30 UTC (permalink / raw)
To: Alexander Duyck; +Cc: David Miller, netdev
In-Reply-To: <CAKgT0UfYEBYQOAoZ77rtwctbwau=2kx4rYdTyeaPaipeMUP9aw@mail.gmail.com>
On Thu, 2012-07-26 at 10:18 -0700, Alexander Duyck wrote:
> I tested this patch and it looks like it runs, but still has the same
> performance issue. I did some digging into the annotation for
> ip_route_intput_noref and it seems like the issue is that I am hitting
> the dst_hold call in __mkroute_input.
David suggested a percpu cache.
nh_rth_input would be allocated by alloc_percpu(struct dst *)
I can work on this.
^ permalink raw reply
* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Alexander Duyck @ 2012-07-26 17:18 UTC (permalink / raw)
To: David Miller; +Cc: eric.dumazet, netdev
In-Reply-To: <20120726.014736.1066206957795563053.davem@davemloft.net>
On Thu, Jul 26, 2012 at 1:47 AM, David Miller <davem@davemloft.net> wrote:
> From: Eric Dumazet <eric.dumazet@gmail.com>
> Date: Thu, 26 Jul 2012 10:27:58 +0200
>
>> But isnt DST_NOCACHE intent reverted then ?
>>
>> like you meant :
>>
>> + } else {
>> + /* Routes we intend to cache in the FIB nexthop have
>> + * the DST_NOCACHE bit unset. However, if we are
>> + * unsuccessful at storing this route into the cache
>> + * we really need to set that bit.
>> + */
>> + rt->dst.flags |= DST_NOCACHE;
>> }
>
> Indeed, thanks for catching this bug.
>
> Here's a new version of the patch, as I found another error. In
> fib_semantics.c, we have to change dst_release() to dst_free() for the
> liberation the nexthop cached routes.
I tested this patch and it looks like it runs, but still has the same
performance issue. I did some digging into the annotation for
ip_route_intput_noref and it seems like the issue is that I am hitting
the dst_hold call in __mkroute_input.
You were correct about the ARP entry. It looks like I had messed up
my test script and it wasn't putting the static ARP entry back in.
I'm amazed it was even working since my SmartBits system doesn't
normally even reply to ARPs.
Thanks,
Alex
^ permalink raw reply
* Re: open sockets preventing unregister_netdevice from completing in linux-next (next-20120724)
From: Eric Dumazet @ 2012-07-26 17:15 UTC (permalink / raw)
To: David Miller; +Cc: bjorn, netdev
In-Reply-To: <20120725.151737.1807149851147342859.davem@davemloft.net>
On Wed, 2012-07-25 at 15:17 -0700, David Miller wrote:
> From: Eric Dumazet <eric.dumazet@gmail.com>
> Date: Wed, 25 Jul 2012 16:38:48 +0200
>
> > Yes, we miss what was done with rt_cache_flush() : find all cached
> > routes and release all dev references...
>
> We can fix this with a two-pronged approach:
>
> 1) Walk the FIB info nexthops and invalidate.
>
> 2) Entries not cached in the FIB info nexthops go into a
> per-netns list which is scanned as well.
>
> I'll try to work on this if nobody beats me to it.
With your latest patch, I can "rmmod tg3" while sockets are active.
Not sure we need all this now ?
(the trick is probably in fib_semantics.c, when you changed
dst_release() to dst_free())
^ permalink raw reply
* Re: [PATCH 00/16] Remove the ipv4 routing cache
From: Eric Dumazet @ 2012-07-26 17:02 UTC (permalink / raw)
To: Paweł Staszewski; +Cc: David Miller, netdev
In-Reply-To: <500D8FEE.7050802@itcare.pl>
On Mon, 2012-07-23 at 19:54 +0200, Paweł Staszewski wrote:
> I make some real life tests with 10Gbit traffic and make some rdos.
> With kernel 3.4.5 when I reach 6M routing cache entries - traffic is not
> forwarded - on my hardware with 64G RAM
>
> With kernel - 3.5.0-rc7-next-20120720 + David M. patches - route cache
> removal + fix
> Traffic is forwarder all the time at speed 8.2Mpps RX from traffic
> generator and 8.2Mpps TX to the sink
>
> Also I see performance improvement.
>
> With kernel 3.4.5 I can reach maximum od 7.5Mpps with my hardware
> And with kernel 3.5.0-rc7-next-20120720 i can reach 8.2Mpps - forwarding
> UDP performance - can be more - but my TX pktgen that use 6 cores can't
> do more.
>
> Really thanks for this :)
In real life, with TCP traffic, make sure you
set /proc/sys/net/ipv4/ip_early_demux to 0
^ permalink raw reply
* Re: [RFC ETHTOOL PATCH 2/2] ethtool: allow setting MDI-X state
From: Ben Hutchings @ 2012-07-26 16:35 UTC (permalink / raw)
To: Jesse Brandeburg; +Cc: netdev
In-Reply-To: <20120726163033.GA20122@jbrandeb-snb.jf.intel.com>
On Thu, 2012-07-26 at 09:30 -0700, Jesse Brandeburg wrote:
> On Wed, Jul 25, 2012 at 11:59:02PM +0100, Ben Hutchings wrote:
> > On Wed, 2012-07-25 at 10:53 -0700, Jesse Brandeburg wrote:
> > How about when you have a forced mode but the driver reports eth_tp_mdix
> > = ETH_TP_MDI_INVALID because the link is down? This is going to result
> > in:
> >
> > MDI-X: Unknown (forced)
> >
> > which makes no sense at all. So I think that we have to do something
> > like:
> >
> > if (ep->eth_tp_mdix_ctrl == ETH_TP_MDI) {
> > fprintf(stdout, "off (forced)\n");
> > } else if (ep->eth_tp_mdix_ctrl == ETH_TP_MDI_X) {
> > fprintf(stdout, "on (forced)\n");
> > } else {
> > switch (ep->eth_tp_mdix) {
> > ...
> > }
> > if (ep->eth_tp_mdix_ctrl == ETH_TP_MDI_AUTO)
> > fprintf(stdout, " (auto)");
> > fprintf(stdout, "\n");
> > }
> >
> > Or else we require that when the mode is forced then drivers report that
> > as the current status even if the link is down.
>
> I tried that and it works swimmingly. How does this look?
[...]
That looks good, thanks.
Ben.
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply
* Re: [PATCH] iproute: Add route showdump command
From: Stephen Hemminger @ 2012-07-26 16:03 UTC (permalink / raw)
To: Pavel Emelyanov; +Cc: Dan Smith, Linux Netdev List
In-Reply-To: <50110593.1070502@parallels.com>
On Thu, 26 Jul 2012 12:53:39 +0400
Pavel Emelyanov <xemul@parallels.com> wrote:
> Some time ago the save+restore commands were added to ip route (git
> id f4ff11e3, Add ip route save/restore). These two save the raw rtnl
> stream into a file and restore one (reading it from stdin).
>
> The problem is that there's no way to get the contents of the dump
> file in a human readable form. How about adding a command that reads
> the rtnl stream from stdin and prints the data in a way the usual
> "ip route list" does?
>
> Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Being able to decode a dump is great idea.
Are the user's smart enough not to try it out at the command prompt
and get totally lost? Maybe another isatty() check is needed.
Another solution would be to put a small header on the save file with
a magic number that could be checked. This would mean changing save/restore/showdump
and ideally updating the magic file in distributions.
^ permalink raw reply
* [PATCH 3/3] pch_gbe: vlan skb len fix
From: Andy Cress @ 2012-07-26 16:01 UTC (permalink / raw)
To: netdev
pch_gbe_xmit_frame skb->len verification was incorrect in vlan case
causing bogus transfer length errors. One correction could be:
offset = skb->protocol == htons(ETH_P_8021Q) ? 0 : 4;
if (unlikely(skb->len > (adapter->hw.mac.max_frame_size - offset)))
However, this verification is not necessary, so remove it.
Signed-off-by: Andy Cress <andy.cress@us.kontron.com>
diff --git a/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c b/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
index 000ef68..817614c 100644
--- a/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
+++ b/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
@@ -2157,13 +2157,6 @@ static int pch_gbe_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
struct pch_gbe_tx_ring *tx_ring = adapter->tx_ring;
unsigned long flags;
- if (unlikely(skb->len > (adapter->hw.mac.max_frame_size - 4))) {
- pr_err("Transfer length Error: skb len: %d > max: %d\n",
- skb->len, adapter->hw.mac.max_frame_size);
- dev_kfree_skb_any(skb);
- adapter->stats.tx_length_errors++;
- return NETDEV_TX_OK;
- }
if (!spin_trylock_irqsave(&tx_ring->tx_lock, flags)) {
/* Collision - tell upper layer to requeue */
return NETDEV_TX_LOCKED;
^ permalink raw reply related
* [PATCH 2/3] pch_gbe: add extra clean tx
From: Andy Cress @ 2012-07-26 16:00 UTC (permalink / raw)
To: netdev
This adds extra cleaning to the pch_gbe_clean_tx routine to avoid
transmit timeouts on some BCM PHYs that have different timing.
Also update the DRV_VERSION to 1.01, and show it.
Signed-off-by: Andy Cress <andy.cress@us.kontron.com>
diff --git a/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c b/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
index 8694704..000ef68 100644
--- a/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
+++ b/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
@@ -26,7 +26,7 @@
#include <linux/ptp_classify.h>
#endif
-#define DRV_VERSION "1.00"
+#define DRV_VERSION "1.01"
const char pch_driver_version[] = DRV_VERSION;
#define PCI_DEVICE_ID_INTEL_IOH1_GBE 0x8802 /* Pci device ID */
@@ -1579,7 +1579,8 @@ pch_gbe_clean_tx(struct pch_gbe_adapter *adapter,
struct sk_buff *skb;
unsigned int i;
unsigned int cleaned_count = 0;
- bool cleaned = true;
+ bool cleaned = false;
+ int unused, thresh;
pr_debug("next_to_clean : %d\n", tx_ring->next_to_clean);
@@ -1588,10 +1589,36 @@ pch_gbe_clean_tx(struct pch_gbe_adapter *adapter,
pr_debug("gbec_status:0x%04x dma_status:0x%04x\n",
tx_desc->gbec_status, tx_desc->dma_status);
+ unused = PCH_GBE_DESC_UNUSED(tx_ring);
+ thresh = tx_ring->count - PCH_GBE_TX_WEIGHT;
+ if ((tx_desc->gbec_status == DSC_INIT16) && (unused < thresh))
+ { /* current marked clean, tx queue filling up, do extra clean */
+ int j, k;
+ if (unused < 8) { /* tx queue nearly full */
+ pr_debug("clean_tx: transmit queue warning (%x,%x) unused=%d\n",
+ tx_ring->next_to_clean,tx_ring->next_to_use,unused);
+ }
+
+ /* current marked clean, scan for more that need cleaning. */
+ k = i;
+ for (j = 0; j < PCH_GBE_TX_WEIGHT; j++)
+ {
+ tx_desc = PCH_GBE_TX_DESC(*tx_ring, k);
+ if (tx_desc->gbec_status != DSC_INIT16) break; /*found*/
+ if (++k >= tx_ring->count) k = 0; /*increment, wrap*/
+ }
+ if (j < PCH_GBE_TX_WEIGHT) {
+ pr_debug("clean_tx: unused=%d loops=%d found tx_desc[%x,%x:%x].gbec_status=%04x\n",
+ unused,j, i,k, tx_ring->next_to_use, tx_desc->gbec_status);
+ i = k; /*found one to clean, usu gbec_status==2000.*/
+ }
+ }
+
while ((tx_desc->gbec_status & DSC_INIT16) == 0x0000) {
pr_debug("gbec_status:0x%04x\n", tx_desc->gbec_status);
buffer_info = &tx_ring->buffer_info[i];
skb = buffer_info->skb;
+ cleaned = true;
if ((tx_desc->gbec_status & PCH_GBE_TXD_GMAC_STAT_ABT)) {
adapter->stats.tx_aborted_errors++;
@@ -1639,18 +1666,21 @@ pch_gbe_clean_tx(struct pch_gbe_adapter *adapter,
}
pr_debug("called pch_gbe_unmap_and_free_tx_resource() %d count\n",
cleaned_count);
- /* Recover from running out of Tx resources in xmit_frame */
- spin_lock(&tx_ring->tx_lock);
- if (unlikely(cleaned && (netif_queue_stopped(adapter->netdev)))) {
- netif_wake_queue(adapter->netdev);
- adapter->stats.tx_restart_count++;
- pr_debug("Tx wake queue\n");
- }
+ if (cleaned_count > 0) { /*skip this if nothing cleaned*/
+ /* Recover from running out of Tx resources in xmit_frame */
+ spin_lock(&tx_ring->tx_lock);
+ if (unlikely(cleaned && (netif_queue_stopped(adapter->netdev))))
+ {
+ netif_wake_queue(adapter->netdev);
+ adapter->stats.tx_restart_count++;
+ pr_debug("Tx wake queue\n");
+ }
- tx_ring->next_to_clean = i;
+ tx_ring->next_to_clean = i;
- pr_debug("next_to_clean : %d\n", tx_ring->next_to_clean);
- spin_unlock(&tx_ring->tx_lock);
+ pr_debug("next_to_clean : %d\n", tx_ring->next_to_clean);
+ spin_unlock(&tx_ring->tx_lock);
+ }
return cleaned;
}
@@ -2387,7 +2417,7 @@ static int pch_gbe_napi_poll(struct napi_struct *napi, int budget)
pch_gbe_clean_rx(adapter, adapter->rx_ring, &work_done, budget);
cleaned = pch_gbe_clean_tx(adapter, adapter->tx_ring);
- if (!cleaned)
+ if (cleaned)
work_done = budget;
/* If no Tx and not enough Rx work done,
* exit the polling mode
@@ -2793,6 +2823,7 @@ static int __init pch_gbe_init_module(void)
{
int ret;
+ pr_info("EG20T PCH Gigabit Ethernet Driver - version %s\n",DRV_VERSION);
ret = pci_register_driver(&pch_gbe_driver);
if (copybreak != PCH_GBE_COPYBREAK_DEFAULT) {
if (copybreak == 0) {
^ permalink raw reply related
* [PATCH 1/3] pch_gbe: fix transmit watchdog timeout
From: Andy Cress @ 2012-07-26 15:59 UTC (permalink / raw)
To: netdev
An extended ping test with 6 vlans resulted in a driver oops with a
netdev transmit timeout.
Fix WATCHDOG_TIMEOUT to be more like e1000e at 5 * HZ, to avoid
unnecessary transmit timeouts.
Signed-off-by: Andy Cress <andy.cress@us.kontron.com>
diff --git a/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c b/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
index 4c04843..a746064 100644
--- a/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
+++ b/drivers/net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c
@@ -35,7 +35,7 @@ const char pch_driver_version[] = DRV_VERSION;
#define DSC_INIT16 0xC000
#define PCH_GBE_DMA_ALIGN 0
#define PCH_GBE_DMA_PADDING 2
-#define PCH_GBE_WATCHDOG_PERIOD (1 * HZ) /* watchdog time */
+#define PCH_GBE_WATCHDOG_PERIOD (5 * HZ) /* watchdog time */
#define PCH_GBE_COPYBREAK_DEFAULT 256
#define PCH_GBE_PCI_BAR 1
#define PCH_GBE_RESERVE_MEMORY 0x200000 /* 2MB */
^ permalink raw reply related
* [PATCH 0/4] pch_gbe: avoiding transmit timeouts (rev3)
From: Andy Cress @ 2012-07-26 15:57 UTC (permalink / raw)
To: netdev
When the interface is stressed with 6 VLANs, some transmit timeout stats were
observed, which is a potential precursor to the more severe netdev watchdog
timeout oops. Also we saw more than the expected number of transmit restarts, which impacted performance. The following patches were applied and resolved
the symptom of the transmit timeout stats, and reduced the number of
transmit restarts.
This patch set includes the following patches:
0001-pch_gbe-fix-transmit-watchdog-timeout.patch
0002-pch_gbe-add-extra-clean-tx.patch (includes bumping the version to 1.01)
0003-pch_gbe-vlan-skb-len-fix.patch
This rev3 has the following changes:
* Removed Fix-the-checksum-fill-to-the-error-location.patch for now so Hongbo can evaluate/revise/test based on feedback.
The resulting pch_gbe 1.01 driver has been tested on Kontron Tunnel Creek
EG20T modules and Intel Crown Bay EG20T modules, so I believe that these are
appropriate for consideration in the upstream pch_gbe driver.
Please review and comment.
Thanks,
Andy
^ permalink raw reply
* Premature timeout for MLDv1 Host compatibility mode?
From: Dragos Ilie @ 2012-07-26 14:57 UTC (permalink / raw)
To: netdev
Hi!
I suspect that the MLDv1 Host compatibility mode is ended prematurely.
In net/ipv6/mcast.c the "Older Version Querier Present" timeout is
computed as
max_delay = (ntohs(mld->mld_maxdelay)*HZ)/1000;
switchback = (idev->mc_qrv + 1) * max_delay;
idev->mc_v1_seen = jiffies + switchback;
RFC 3810 says that the timeout should be computed as
Robustness_Variable * Query_Interval + Query_Response_Interval. This
suggests that the line where switchback is computed should be changed
to something like
switchback = (idev->mc_qrv * 125 * HZ) + max_delay;
where 125 is the default Query_Interval in seconds
I would appreciate it if somebody can confirm if my suspicion is correct.
Regards,
Dragos
^ permalink raw reply
* Re: [PATCH]net:appletalk:ddp:fixed coding style issue in net/appletalk/ddp.c
From: Joe Perches @ 2012-07-26 14:11 UTC (permalink / raw)
To: Jeffrin Jose; +Cc: acme, davem, netdev, linux-kernel
In-Reply-To: <1343311142-4539-1-git-send-email-ahiliation@yahoo.co.in>
On Thu, 2012-07-26 at 19:29 +0530, Jeffrin Jose wrote:
> Fixed coding style issue relating to switch and case
> statement in file net/appletalk/ddp.c
Don't just move the case labels, move the code blocks
including comments.
> diff --git a/net/appletalk/ddp.c b/net/appletalk/ddp.c
[]
> @@ -1798,7 +1798,7 @@ static int atalk_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
>
> switch (cmd) {
> /* Protocol layer */
> - case TIOCOUTQ: {
> + case TIOCOUTQ: {
> long amount = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
long amount = etc...
Post the patch with a commit message that says that
that a git diff -w is empty and the compiled objects
are the same too.
^ permalink raw reply
* [PATCH]net:appletalk:ddp:fixed coding style issue in net/appletalk/ddp.c
From: Jeffrin Jose @ 2012-07-26 13:59 UTC (permalink / raw)
To: acme, davem; +Cc: netdev, linux-kernel, ahiliation
Fixed coding style issue relating to switch and case
statement in file net/appletalk/ddp.c
Signed-off-by: Jeffrin Jose <ahiliation@yahoo.co.in>
---
net/appletalk/ddp.c | 26 +++++++++++++-------------
1 file changed, 13 insertions(+), 13 deletions(-)
diff --git a/net/appletalk/ddp.c b/net/appletalk/ddp.c
index 0301b32..c8bbcd2 100644
--- a/net/appletalk/ddp.c
+++ b/net/appletalk/ddp.c
@@ -1798,7 +1798,7 @@ static int atalk_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
switch (cmd) {
/* Protocol layer */
- case TIOCOUTQ: {
+ case TIOCOUTQ: {
long amount = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
if (amount < 0)
@@ -1806,7 +1806,7 @@ static int atalk_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
rc = put_user(amount, (int __user *)argp);
break;
}
- case TIOCINQ: {
+ case TIOCINQ: {
/*
* These two are safe on a single CPU system as only
* user tasks fiddle here
@@ -1819,27 +1819,27 @@ static int atalk_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
rc = put_user(amount, (int __user *)argp);
break;
}
- case SIOCGSTAMP:
+ case SIOCGSTAMP:
rc = sock_get_timestamp(sk, argp);
break;
- case SIOCGSTAMPNS:
+ case SIOCGSTAMPNS:
rc = sock_get_timestampns(sk, argp);
break;
/* Routing */
- case SIOCADDRT:
- case SIOCDELRT:
+ case SIOCADDRT:
+ case SIOCDELRT:
rc = -EPERM;
if (capable(CAP_NET_ADMIN))
rc = atrtr_ioctl(cmd, argp);
break;
/* Interface */
- case SIOCGIFADDR:
- case SIOCSIFADDR:
- case SIOCGIFBRDADDR:
- case SIOCATALKDIFADDR:
- case SIOCDIFADDR:
- case SIOCSARP: /* proxy AARP */
- case SIOCDARP: /* proxy AARP */
+ case SIOCGIFADDR:
+ case SIOCSIFADDR:
+ case SIOCGIFBRDADDR:
+ case SIOCATALKDIFADDR:
+ case SIOCDIFADDR:
+ case SIOCSARP: /* proxy AARP */
+ case SIOCDARP: /* proxy AARP */
rtnl_lock();
rc = atif_ioctl(cmd, argp);
rtnl_unlock();
--
1.7.10
^ permalink raw reply related
* Re: [PATCH RESEND 0/9] OLPC: create a generic OLPC EC driver
From: Thomas Gleixner @ 2012-07-26 13:51 UTC (permalink / raw)
To: Andres Salomon
Cc: Andrew Morton, Paul Fox, Daniel Drake, Richard A. Smith,
linux-kernel, libertas-dev, linux-wireless, netdev,
platform-driver-x86, devel, Ingo Molnar, H. Peter Anvin, x86,
Dan Williams, John W. Linville, Matthew Garrett, Anton Vorontsov,
David Woodhouse, Chris Ball, Jon Nettleton, Greg Kroah-Hartman
In-Reply-To: <20120718213713.232e4161@dev.queued.net>
On Wed, 18 Jul 2012, Andres Salomon wrote:
> The OLPC EC (Embedded Controller) code that is currently upstream is
> x86-only, originally written for the XO-1. Since then, we've had the
> XO-1.5 (also x86), and XO-1.75 (arm-based) enter mass production. The
> 1.75 uses a vastly different EC protocol, and future hardware revisions
> are likely to change it even further.
>
> However, the drivers do share quite a bit of code, so it makes sense to
> have a platform-agnostic driver that calls into platform-specific hooks
> for each XO's EC driver. This is the first stab and creating such a
> beast (with further patches pending). Aside from the lack of code
> duplication, this is helpful for fixing bugs in one place (for example,
> we fixed an EC suspend/resume bug in 1.75 that I've just seen happen on
> 1.5 without these patches. With these patches, the problem goes away).
>
> These patches are against Linus's current HEAD; let me know if they
> don't apply somewhere, and I'll happily redo them against the -next
> tree. I'm assuming that these changes (which touch places like x86,
> wireless, and staging) should go through either the x86 tree, or
> through akpm's tree.
>
> Alternatively, if the reviews are positive and I can get SOBs from the
> relevant maintainers, I can set up a platform-olpc tree somewhere and
> request a pull from Linus.
Either via akpm or a separate tree are fine with me.
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
^ permalink raw reply
* Re: orphan process is causing deadlock in __kernel_vsyscall ()
From: Sri Ram Vemulpali @ 2012-07-26 12:54 UTC (permalink / raw)
To: linux-kernel-mail, linux-newbie, linux-netdev, kernelnewbies,
Kernel-newbies
In-Reply-To: <CALyraeOQZ+FPTqaOk6XjHsZwwCfNuo2XLo6bhw44maSEkDf7tA@mail.gmail.com>
> Hello All,
>
> I am debugging a problem which I do not know how to solve. I know what
> is happening, but cannot find solution to it. If anyone has any
> suggestions or solution that would be really grateful and I appreciate
> it.
>
> Here is the description.
>
> I have a thread which spawns a child process using fork() and in child
> context it execv to shell and parent thread waits on waitpid(). This
> child process is nothing but a shell /bin/sh, so that from program I
> can create shell and run some commands.
> In this shell instance, I run tshark for capturing network traffic. I
> have really high traffic going, so system is really busy. All is good
> so far, but when I send SIGKILL to child process (shell) from another
> thread from same application,
> child process (shell) gets terminated and tshark is assigned to
> init(1) process as orphan. The parent thread which spawned the child
> process unblocks from waitpid() after cleaning child instance and
> continues it execution.
>
> Here is the problem:
> Once the parent thread resumes from waitpid(), it hits fprintf to
> print message "exiting from shell" to console. In here the main thread
> gets blocked in __kernel_vsyscall () entered from fprintf. Following
> is my stack trace from gdb.
> But after some more debugging and found that, if I kill from another
> shell "tshark" process which is assigned to init(1) as child, the
> parent thread unblocks from __kernel_vsyscall ().
>
> Can anyone tell what is going on, is this related to futex spinning
> problem where lock is been held by tshark. How would I approach this
> problem?
> Or is there a way to kill all child process of a process without
> making them orphan. Because when a child process is running at the
> time parent is exiting, that child process is assigned to init(1) as
> orphan process.
> Can this behavior be modified. Please let me know any suggestions and
> thoughts. Thanks in advance.
>
> --
> Regards,
> Sri.
>
> (gdb) backtrace
> #0 0xffffe424 in __kernel_vsyscall ()
> #1 0xb7e6396b in write () from /lib/libc.so.6
> #2 0xb7e0c2ef in _IO_new_file_write () from /lib/libc.so.6
> #3 0xb7e0bf93 in new_do_write () from /lib/libc.so.6
> #4 0xb7e0c296 in _IO_new_do_write () from /lib/libc.so.6
> #5 0xb7e0ce20 in _IO_new_file_overflow () from /lib/libc.so.6
> #6 0xb7e0c0cd in _IO_new_file_xsputn () from /lib/libc.so.6
> #7 0xb7de3634 in vfprintf () from /lib/libc.so.6
> #8 0x09ad62a2 in acme_fprintf (stream=0xb7ee2560,
> format=0x9f17694 "shell exited. returning to acli\n")
> at /home/svemulpali/cc/svemulpali_SCOTTY_main/linux/private/common/src/coutOverride.cpp:42
> #9 0x094a9345 in vxShellCmdStartWait (syscmd=0x0)
> at /home/svemulpali/cc/svemulpali_SCOTTY_main/linux/private/lib/stub/vxShellLib.cpp:210
--
Regards,
Sri.
--
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs
^ permalink raw reply
* r8169, 3.5.0, does't work at all
From: Denys Fedoryshchenko @ 2012-07-26 11:58 UTC (permalink / raw)
To: netdev, romieu
Just pushed new kernel to test cluster and got this:
[ 1.604470] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[ 1.604709] r8169 0000:02:00.0: irq 43 for MSI/MSI-X
[ 1.605211] r8169 0000:02:00.0: eth0: RTL8168b/8111b at 0xf8604000,
00:08:54:57:7a:54, XID 18000000 IRQ 43
[ 1.605364] r8169 0000:02:00.0: eth0: jumbo features [frames: 4080
bytes, tx checksumming: ko]
[ 1.605549] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[ 1.605780] r8169 0000:03:00.0: irq 44 for MSI/MSI-X
[ 1.606212] r8169 0000:03:00.0: eth1: RTL8168b/8111b at 0xf8606000,
00:08:54:57:7a:33, XID 18000000 IRQ 44
[ 1.606366] r8169 0000:03:00.0: eth1: jumbo features [frames: 4080
bytes, tx checksumming: ko]
....
[ 20.406090] input device check on
[ 20.406177] Actions configured
[ 21.526836] r8169 0000:02:00.0: eth0: link up
[ 28.038386] ------------[ cut here ]------------
[ 28.038438] WARNING: at net/sched/sch_generic.c:255
dev_watchdog+0xd6/0x12a()
[ 28.038489] Hardware name:
[ 28.038534] NETDEV WATCHDOG: eth0 (r8169): transmit queue 0 timed
out
[ 28.038584] Modules linked in: cls_flow cls_u32 em_meta cls_basic
xt_dscp ts_bm xt_string xt_hl ifb cls_fw sch_tbf sch_htb act_ipt
act_mirred ipt_REDIRECT ipt_REJECT xt_TCPMSS xt_DSCP xt_mark
iptable_mangle iptable_nat nf_nat nf_conntrack_ipv4 nf_conntrack
nf_defrag_ipv4 iptable_filter 8021q garp stp llc loop usb_storage mei
lpc_ich mfd_core ahci libahci libata sr_mod cdrom tulip r8169 sky2
via_velocity via_rhine sis900 ne2k_pci 8390 skge tg3 libphy 8139too
e1000 e100 usbhid ohci_hcd uhci_hcd ehci_hcd usbcore usb_common [last
unloaded: usb_storage]
[ 28.038616] Pid: 4, comm: kworker/0:0 Not tainted 3.5.0-build-0063
#4
[ 28.038617] Call Trace:
[ 28.038620] [<c0127962>] warn_slowpath_common+0x77/0x8c
[ 28.038623] [<c02f507e>] ? dev_watchdog+0xd6/0x12a
[ 28.038625] [<c02f507e>] ? dev_watchdog+0xd6/0x12a
[ 28.038628] [<c01279f3>] warn_slowpath_fmt+0x2e/0x30
[ 28.038630] [<c02f507e>] dev_watchdog+0xd6/0x12a
[ 28.038633] [<c0131c45>] run_timer_softirq+0x1ab/0x254
[ 28.038635] [<c0131bc9>] ? run_timer_softirq+0x12f/0x254
[ 28.038638] [<c015d181>] ? trace_hardirqs_on_caller+0xf5/0x13f
[ 28.038640] [<c02f4fa8>] ? netif_tx_unlock+0x41/0x41
[ 28.038643] [<c012d38e>] __do_softirq+0x7b/0x118
[ 28.038644] [<c012d313>] ? local_bh_enable+0xd/0xd
[ 28.038645] <IRQ> [<c012d5a9>] ? irq_exit+0x41/0x91
[ 28.038650] [<c0118ac0>] ? smp_apic_timer_interrupt+0x64/0x71
[ 28.038652] [<c023054c>] ? trace_hardirqs_off_thunk+0xc/0x10
[ 28.038655] [<c0353977>] ? apic_timer_interrupt+0x2f/0x34
[ 28.038661] [<f85f7e9f>] ? rtl_slow_event_work+0x14/0x1a2 [r8169]
[ 28.038664] [<c0351898>] ? mutex_lock_nested+0x20/0x22
[ 28.038667] [<f85f49d1>] ? rtl_task+0x49/0x5f [r8169]
[ 28.038670] [<c013992a>] ? process_one_work+0x1a1/0x2ab
[ 28.038672] [<c01398b9>] ? process_one_work+0x130/0x2ab
[ 28.038675] [<f85f4988>] ? rtl8169_set_features+0x39/0x39 [r8169]
[ 28.038678] [<c013a3be>] ? worker_thread+0xd1/0x14c
[ 28.038680] [<c013a2ed>] ? manage_workers.clone.20+0x14e/0x14e
[ 28.038682] [<c013d3ec>] ? kthread+0x67/0x6c
[ 28.038685] [<c0350000>] ? migration_call+0x10/0x189
[ 28.038687] [<c013d385>] ? __init_kthread_worker+0x47/0x47
[ 28.038689] [<c03542fa>] ? kernel_thread_helper+0x6/0xd
[ 28.038690] ---[ end trace 9701f577e9a74181 ]---
[ 28.043532] r8169 0000:02:00.0: eth0: link up
[ 34.043357] r8169 0000:02:00.0: eth0: link up
[ 40.043202] r8169 0000:02:00.0: eth0: link up
[ 46.043052] r8169 0000:02:00.0: eth0: link up
[ 52.042893] r8169 0000:02:00.0: eth0: link up
[ 58.042739] r8169 0000:02:00.0: eth0: link up
[ 64.042584] r8169 0000:02:00.0: eth0: link up
[ 70.042430] r8169 0000:02:00.0: eth0: link up
[ 76.042276] r8169 0000:02:00.0: eth0: link up
[ 82.042123] r8169 0000:02:00.0: eth0: link up
[ 88.041968] r8169 0000:02:00.0: eth0: link up
[ 94.041813] r8169 0000:02:00.0: eth0: link up
---
Denys Fedoryshchenko, Network Engineer, Virtual ISP S.A.L.
^ permalink raw reply
* [net] ixgbe: fix panic while dumping packets on Tx hang with IOMMU
From: Jeff Kirsher @ 2012-07-26 11:21 UTC (permalink / raw)
To: davem; +Cc: Emil Tantilov, netdev, gospo, sassmann, Jeff Kirsher
From: Emil Tantilov <emil.s.tantilov@intel.com>
This patch resolves a "BUG: unable to handle kernel paging request at ..."
oops while dumping packet data. The issue occurs with IOMMU enabled due to
the address provided by phys_to_virt().
This patch makes use of skb->data on Tx and the virtual address of the pages
allocated for Rx.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 3b6784c..c709eae 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -396,11 +396,10 @@ static void ixgbe_dump(struct ixgbe_adapter *adapter)
pr_cont("\n");
if (netif_msg_pktdata(adapter) &&
- dma_unmap_len(tx_buffer, len) != 0)
+ tx_buffer->skb)
print_hex_dump(KERN_INFO, "",
DUMP_PREFIX_ADDRESS, 16, 1,
- phys_to_virt(dma_unmap_addr(tx_buffer,
- dma)),
+ tx_buffer->skb->data,
dma_unmap_len(tx_buffer, len),
true);
}
@@ -474,10 +473,12 @@ rx_ring_summary:
(u64)rx_buffer_info->dma,
rx_buffer_info->skb);
- if (netif_msg_pktdata(adapter)) {
+ if (netif_msg_pktdata(adapter) &&
+ rx_buffer_info->dma) {
print_hex_dump(KERN_INFO, "",
DUMP_PREFIX_ADDRESS, 16, 1,
- phys_to_virt(rx_buffer_info->dma),
+ page_address(rx_buffer_info->page) +
+ rx_buffer_info->page_offset,
ixgbe_rx_bufsz(rx_ring), true);
}
}
--
1.7.11.2
^ permalink raw reply related
* [PATCH] ipv6: Early TCP socket demux
From: Eric Dumazet @ 2012-07-26 10:34 UTC (permalink / raw)
To: David Miller; +Cc: netdev
From: Eric Dumazet <edumazet@google.com>
This is the IPv6 missing bits for infratructure added in commit
41063e9dd1195 (ipv4: Early TCP socket demux.)
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
include/net/protocol.h | 2 ++
net/ipv6/ip6_input.c | 13 +++++++++++--
net/ipv6/tcp_ipv6.c | 38 ++++++++++++++++++++++++++++++++++++++
3 files changed, 51 insertions(+), 2 deletions(-)
diff --git a/include/net/protocol.h b/include/net/protocol.h
index 057f2d3..929528c 100644
--- a/include/net/protocol.h
+++ b/include/net/protocol.h
@@ -52,6 +52,8 @@ struct net_protocol {
#if IS_ENABLED(CONFIG_IPV6)
struct inet6_protocol {
+ void (*early_demux)(struct sk_buff *skb);
+
int (*handler)(struct sk_buff *skb);
void (*err_handler)(struct sk_buff *skb,
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index 5ab923e..47975e3 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -47,9 +47,18 @@
-inline int ip6_rcv_finish( struct sk_buff *skb)
+int ip6_rcv_finish(struct sk_buff *skb)
{
- if (skb_dst(skb) == NULL)
+ if (sysctl_ip_early_demux && !skb_dst(skb)) {
+ const struct inet6_protocol *ipprot;
+
+ rcu_read_lock();
+ ipprot = rcu_dereference(inet6_protos[ipv6_hdr(skb)->nexthdr]);
+ if (ipprot && ipprot->early_demux)
+ ipprot->early_demux(skb);
+ rcu_read_unlock();
+ }
+ if (!skb_dst(skb))
ip6_route_input(skb);
return dst_input(skb);
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index f49476e..221224e 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1674,6 +1674,43 @@ do_time_wait:
goto discard_it;
}
+static void tcp_v6_early_demux(struct sk_buff *skb)
+{
+ const struct ipv6hdr *hdr;
+ const struct tcphdr *th;
+ struct sock *sk;
+
+ if (skb->pkt_type != PACKET_HOST)
+ return;
+
+ if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
+ return;
+
+ hdr = ipv6_hdr(skb);
+ th = tcp_hdr(skb);
+
+ if (th->doff < sizeof(struct tcphdr) / 4)
+ return;
+
+ sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
+ &hdr->saddr, th->source,
+ &hdr->daddr, ntohs(th->dest),
+ inet6_iif(skb));
+ if (sk) {
+ skb->sk = sk;
+ skb->destructor = sock_edemux;
+ if (sk->sk_state != TCP_TIME_WAIT) {
+ struct dst_entry *dst = sk->sk_rx_dst;
+ struct inet_sock *icsk = inet_sk(sk);
+ if (dst)
+ dst = dst_check(dst, 0);
+ if (dst &&
+ icsk->rx_dst_ifindex == inet6_iif(skb))
+ skb_dst_set_noref(skb, dst);
+ }
+ }
+}
+
static struct timewait_sock_ops tcp6_timewait_sock_ops = {
.twsk_obj_size = sizeof(struct tcp6_timewait_sock),
.twsk_unique = tcp_twsk_unique,
@@ -1984,6 +2021,7 @@ struct proto tcpv6_prot = {
};
static const struct inet6_protocol tcpv6_protocol = {
+ .early_demux = tcp_v6_early_demux,
.handler = tcp_v6_rcv,
.err_handler = tcp_v6_err,
.gso_send_check = tcp_v6_gso_send_check,
^ permalink raw reply related
* Re: [PATCH] bcma: fix regression in pmu workaround reg masks
From: David Miller @ 2012-07-26 10:06 UTC (permalink / raw)
To: hauke
Cc: torvalds, linville, brcm80211-dev-list, linux-wireless, netdev,
linux-kernel, seth.forshee, pieterpg, brudley, arend, zajec5
In-Reply-To: <1343294151-5691-1-git-send-email-hauke@hauke-m.de>
From: Hauke Mehrtens <hauke@hauke-m.de>
Date: Thu, 26 Jul 2012 11:15:51 +0200
> This fixes a regression introduced in:
> commit b9562545ef0b13c0440ccd8d6dd4111fb77cb17a
> Author: Hauke Mehrtens <hauke@hauke-m.de>
> Date: Sat Jun 30 01:44:41 2012 +0200
>
> bcma: complete workaround for BCMA43224 and BCM4313
>
> The regression broke reading of the sprom from the chip and seamed to
> cause a change of the pci id on a BCMA43224 device so that it is
> rejected by brcmsmac. If this problem occurred, after applying this
> patch, a cold boot is needed to make the device work again.
>
> In the original patch the workaround set the intended bits in the chip
> common core chip control registers, but the patch also unset all the
> other bits in that register, because of a wrong mask. The original
> patch was based on code from brcmsmac and there was an additional
> inversing of the mask in the code setting these regs which was missing
> here. Now the regs are set like brcmsmac did it before.
>
> Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
> Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
> Tested-by: Seth Forshee <seth.forshee@canonical.com>
Acked-by: David S. Miller <davem@davemloft.net>
^ permalink raw reply
* [PATCH iproute2] ss: report SK_MEMINFO_BACKLOG
From: Eric Dumazet @ 2012-07-26 9:20 UTC (permalink / raw)
To: Stephen Hemminger; +Cc: netdev, Vijay Subramanian
From: Eric Dumazet <edumazet@google.com>
linux-3.6-rc1 supports SK_MEMINFO_BACKLOG with commit d594e987c6f54
(sock_diag: add SK_MEMINFO_BACKLOG)
ss command can display it if provided by the kernel.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
---
include/linux/sock_diag.h | 1 +
misc/ss.c | 13 ++++++++++---
2 files changed, 11 insertions(+), 3 deletions(-)
diff --git a/include/linux/sock_diag.h b/include/linux/sock_diag.h
index 39e4b1c..ac9db19 100644
--- a/include/linux/sock_diag.h
+++ b/include/linux/sock_diag.h
@@ -18,6 +18,7 @@ enum {
SK_MEMINFO_FWD_ALLOC,
SK_MEMINFO_WMEM_QUEUED,
SK_MEMINFO_OPTMEM,
+ SK_MEMINFO_BACKLOG,
SK_MEMINFO_VARS,
};
diff --git a/misc/ss.c b/misc/ss.c
index cf529ef..d8429cc 100644
--- a/misc/ss.c
+++ b/misc/ss.c
@@ -1337,8 +1337,9 @@ static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r)
nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*r)));
if (tb[INET_DIAG_SKMEMINFO]) {
- const __u32 *skmeminfo = RTA_DATA(tb[INET_DIAG_SKMEMINFO]);
- printf(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u)",
+ const __u32 *skmeminfo = RTA_DATA(tb[INET_DIAG_SKMEMINFO]);
+
+ printf(" skmem:(r%u,rb%u,t%u,tb%u,f%u,w%u,o%u",
skmeminfo[SK_MEMINFO_RMEM_ALLOC],
skmeminfo[SK_MEMINFO_RCVBUF],
skmeminfo[SK_MEMINFO_WMEM_ALLOC],
@@ -1346,7 +1347,13 @@ static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r)
skmeminfo[SK_MEMINFO_FWD_ALLOC],
skmeminfo[SK_MEMINFO_WMEM_QUEUED],
skmeminfo[SK_MEMINFO_OPTMEM]);
- }else if (tb[INET_DIAG_MEMINFO]) {
+
+ if (RTA_PAYLOAD(tb[INET_DIAG_SKMEMINFO]) >=
+ (SK_MEMINFO_BACKLOG + 1) * sizeof(__u32))
+ printf(",bl%u", skmeminfo[SK_MEMINFO_BACKLOG]);
+
+ printf(")");
+ } else if (tb[INET_DIAG_MEMINFO]) {
const struct inet_diag_meminfo *minfo
= RTA_DATA(tb[INET_DIAG_MEMINFO]);
printf(" mem:(r%u,w%u,f%u,t%u)",
^ 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