tbench regression in 2.6.25-rc1

tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

Comparing with kernel 2.6.24, tbench result has regression with
2.6.25-rc1.

1) On 2 quad-core processor stoakley: 4%.
2) On 4 quad-core processor tigerton: more than 30%.

bisect located below patch.

b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit
commit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b
Author: Herbert Xu <herbert@gondor.apana.org.au>
Date:   Tue Nov 13 21:33:32 2007 -0800

    [IPV6]: Move nfheader_len into rt6_info
    
    The dst member nfheader_len is only used by IPv6.  It's also currently
    creating a rather ugly alignment hole in struct dst.  Therefore this patch
    moves it from there into struct rt6_info.


As tbench uses ipv4, so the patch's real impact on ipv4 is it deletes
nfheader_len in dst_entry. It might change cache line alignment.

To verify my finding, I just added nfheader_len back to dst_entry in 2.6.25-rc1
and reran tbench on the 2 machines. Performance could be recovered completely.

I started cpu_number*2 tbench processes. On my 16-core tigerton:
#./tbench_srv &
#./tbench 32 127.0.0.1

-yanmin

Re: tbench regression in 2.6.25-rc1

[Eric Dumazet]

Zhang, Yanmin a écrit :
> Comparing with kernel 2.6.24, tbench result has regression with
> 2.6.25-rc1.
> 
> 1) On 2 quad-core processor stoakley: 4%.
> 2) On 4 quad-core processor tigerton: more than 30%.
> 
> bisect located below patch.
> 
> b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit
> commit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b
> Author: Herbert Xu <herbert@gondor.apana.org.au>
> Date:   Tue Nov 13 21:33:32 2007 -0800
> 
>     [IPV6]: Move nfheader_len into rt6_info
>     
>     The dst member nfheader_len is only used by IPv6.  It's also currently
>     creating a rather ugly alignment hole in struct dst.  Therefore this patch
>     moves it from there into struct rt6_info.
> 
> 
> As tbench uses ipv4, so the patch's real impact on ipv4 is it deletes
> nfheader_len in dst_entry. It might change cache line alignment.
> 
> To verify my finding, I just added nfheader_len back to dst_entry in 2.6.25-rc1
> and reran tbench on the 2 machines. Performance could be recovered completely.
> 
> I started cpu_number*2 tbench processes. On my 16-core tigerton:
> #./tbench_srv &
> #./tbench 32 127.0.0.1
> 
> -yanmin

Yup. struct dst is sensitive to alignements, especially for benches.

In the real world, we need to make sure that next pointer start at a cache 
line bondary (or a litle bit after), so that RT cache lookups use one cache 
line per entry instead of two. This permits better behavior in DDOS attacks.

(check commit 1e19e02ca0c5e33ea73a25127dbe6c3b8fcaac4b for reference)

Are you using a 64 or a 32 bit kernel ?

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=utf-8, Size: 1809 bytes --]

On Fri, 2008-02-15 at 07:05 +0100, Eric Dumazet wrote:
> Zhang, Yanmin a écrit :
> > Comparing with kernel 2.6.24, tbench result has regression with
> > 2.6.25-rc1.
> > 
> > 1) On 2 quad-core processor stoakley: 4%.
> > 2) On 4 quad-core processor tigerton: more than 30%.
> > 
> > bisect located below patch.
> > 
> > b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit
> > commit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b
> > Author: Herbert Xu <herbert@gondor.apana.org.au>
> > Date:   Tue Nov 13 21:33:32 2007 -0800
> > 
> >     [IPV6]: Move nfheader_len into rt6_info
> >     
> >     The dst member nfheader_len is only used by IPv6.  It's also currently
> >     creating a rather ugly alignment hole in struct dst.  Therefore this patch
> >     moves it from there into struct rt6_info.
> > 
> > 
> > As tbench uses ipv4, so the patch's real impact on ipv4 is it deletes
> > nfheader_len in dst_entry. It might change cache line alignment.
> > 
> > To verify my finding, I just added nfheader_len back to dst_entry in 2.6.25-rc1
> > and reran tbench on the 2 machines. Performance could be recovered completely.
> > 
> > I started cpu_number*2 tbench processes. On my 16-core tigerton:
> > #./tbench_srv &
> > #./tbench 32 127.0.0.1
> > 
> > -yanmin
> 
> Yup. struct dst is sensitive to alignements, especially for benches.
> 
> In the real world, we need to make sure that next pointer start at a cache 
> line bondary (or a litle bit after), so that RT cache lookups use one cache 
> line per entry instead of two. This permits better behavior in DDOS attacks.
> 
> (check commit 1e19e02ca0c5e33ea73a25127dbe6c3b8fcaac4b for reference)
> 
> Are you using a 64 or a 32 bit kernel ?
64bit x86-64 machine. On another 4-way Madison Itanium machine, tbench has the
similiar regression.

-yanmin

Re: tbench regression in 2.6.25-rc1

[Eric Dumazet]

Zhang, Yanmin a écrit :
> On Fri, 2008-02-15 at 07:05 +0100, Eric Dumazet wrote:
>   
>> Zhang, Yanmin a �crit :
>>     
>>> Comparing with kernel 2.6.24, tbench result has regression with
>>> 2.6.25-rc1.
>>>
>>> 1) On 2 quad-core processor stoakley: 4%.
>>> 2) On 4 quad-core processor tigerton: more than 30%.
>>>
>>> bisect located below patch.
>>>
>>> b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit
>>> commit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b
>>> Author: Herbert Xu <herbert@gondor.apana.org.au>
>>> Date:   Tue Nov 13 21:33:32 2007 -0800
>>>
>>>     [IPV6]: Move nfheader_len into rt6_info
>>>     
>>>     The dst member nfheader_len is only used by IPv6.  It's also currently
>>>     creating a rather ugly alignment hole in struct dst.  Therefore this patch
>>>     moves it from there into struct rt6_info.
>>>
>>>
>>> As tbench uses ipv4, so the patch's real impact on ipv4 is it deletes
>>> nfheader_len in dst_entry. It might change cache line alignment.
>>>
>>> To verify my finding, I just added nfheader_len back to dst_entry in 2.6.25-rc1
>>> and reran tbench on the 2 machines. Performance could be recovered completely.
>>>
>>> I started cpu_number*2 tbench processes. On my 16-core tigerton:
>>> #./tbench_srv &
>>> #./tbench 32 127.0.0.1
>>>
>>> -yanmin
>>>       
>> Yup. struct dst is sensitive to alignements, especially for benches.
>>
>> In the real world, we need to make sure that next pointer start at a cache 
>> line bondary (or a litle bit after), so that RT cache lookups use one cache 
>> line per entry instead of two. This permits better behavior in DDOS attacks.
>>
>> (check commit 1e19e02ca0c5e33ea73a25127dbe6c3b8fcaac4b for reference)
>>
>> Are you using a 64 or a 32 bit kernel ?
>>     
> 64bit x86-64 machine. On another 4-way Madison Itanium machine, tbench has the
> similiar regression.
>
>   

On linux-2.6.25-rc1 x86_64 :

offsetof(struct dst_entry, lastuse)=0xb0
offsetof(struct dst_entry, __refcnt)=0xb8
offsetof(struct dst_entry, __use)=0xbc
offsetof(struct dst_entry, next)=0xc0

So it should be optimal... I dont know why tbench prefers __refcnt being 
on 0xc0, since in this case lastuse will be on a different cache line...

Each incoming IP packet will need to change lastuse, __refcnt and __use, 
so keeping them in the same cache line is a win.

I suspect then that even this patch could help tbench, since it avoids 
writing lastuse...

diff --git a/include/net/dst.h b/include/net/dst.h
index e3ac7d0..24d3c4e 100644
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -147,7 +147,8 @@ static inline void dst_use(struct dst_entry *dst, 
unsigned long time)
 {
        dst_hold(dst);
        dst->__use++;
-       dst->lastuse = time;
+       if (time != dst->lastuse)
+               dst->lastuse = time;
 }

Re: tbench regression in 2.6.25-rc1

[David Miller]

From: Eric Dumazet <dada1@cosmosbay.com>
Date: Fri, 15 Feb 2008 15:21:48 +0100

> On linux-2.6.25-rc1 x86_64 :
> 
> offsetof(struct dst_entry, lastuse)=0xb0
> offsetof(struct dst_entry, __refcnt)=0xb8
> offsetof(struct dst_entry, __use)=0xbc
> offsetof(struct dst_entry, next)=0xc0
> 
> So it should be optimal... I dont know why tbench prefers __refcnt being 
> on 0xc0, since in this case lastuse will be on a different cache line...
> 
> Each incoming IP packet will need to change lastuse, __refcnt and __use, 
> so keeping them in the same cache line is a win.
> 
> I suspect then that even this patch could help tbench, since it avoids 
> writing lastuse...

I think your suspicions are right, and even moreso
it helps to keep __refcnt out of the same cache line
as input/output/ops which are read-almost-entirely :-)

I haven't done an exhaustive analysis, but it seems that
the write traffic to lastuse and __refcnt are about the
same.  However if we find that __refcnt gets hit more
than lastuse in this workload, it explains the regression.

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

On Fri, 2008-02-15 at 15:21 +0100, Eric Dumazet wrote:
> Zhang, Yanmin a écrit :
> > On Fri, 2008-02-15 at 07:05 +0100, Eric Dumazet wrote:
> >   
> >> Zhang, Yanmin a �crit :
> >>     
> >>> Comparing with kernel 2.6.24, tbench result has regression with
> >>> 2.6.25-rc1.
> >>>
> >>> 1) On 2 quad-core processor stoakley: 4%.
> >>> 2) On 4 quad-core processor tigerton: more than 30%.
> >>>
> >>> bisect located below patch.
> >>>
> >>> b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit
> >>> commit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b
> >>> Author: Herbert Xu <herbert@gondor.apana.org.au>
> >>> Date:   Tue Nov 13 21:33:32 2007 -0800
> >>>
> >>>     [IPV6]: Move nfheader_len into rt6_info
> >>>     
> >>>     The dst member nfheader_len is only used by IPv6.  It's also currently
> >>>     creating a rather ugly alignment hole in struct dst.  Therefore this patch
> >>>     moves it from there into struct rt6_info.
> >>>
> >>>
> >>> As tbench uses ipv4, so the patch's real impact on ipv4 is it deletes
> >>> nfheader_len in dst_entry. It might change cache line alignment.
> >>>
> >>> To verify my finding, I just added nfheader_len back to dst_entry in 2.6.25-rc1
> >>> and reran tbench on the 2 machines. Performance could be recovered completely.
> >>>
> >>> I started cpu_number*2 tbench processes. On my 16-core tigerton:
> >>> #./tbench_srv &
> >>> #./tbench 32 127.0.0.1
> >>>
> >>> -yanmin
> >>>       
> >> Yup. struct dst is sensitive to alignements, especially for benches.
> >>
> >> In the real world, we need to make sure that next pointer start at a cache 
> >> line bondary (or a litle bit after), so that RT cache lookups use one cache 
> >> line per entry instead of two. This permits better behavior in DDOS attacks.
> >>
> >> (check commit 1e19e02ca0c5e33ea73a25127dbe6c3b8fcaac4b for reference)
> >>
> >> Are you using a 64 or a 32 bit kernel ?
> >>     
> > 64bit x86-64 machine. On another 4-way Madison Itanium machine, tbench has the
> > similiar regression.
> >
> >   
> 
> On linux-2.6.25-rc1 x86_64 :
> 
> offsetof(struct dst_entry, lastuse)=0xb0
> offsetof(struct dst_entry, __refcnt)=0xb8
> offsetof(struct dst_entry, __use)=0xbc
> offsetof(struct dst_entry, next)=0xc0
> 
> So it should be optimal... I dont know why tbench prefers __refcnt being 
> on 0xc0, since in this case lastuse will be on a different cache line...
> 
> Each incoming IP packet will need to change lastuse, __refcnt and __use, 
> so keeping them in the same cache line is a win.
> 
> I suspect then that even this patch could help tbench, since it avoids 
> writing lastuse...
> 
> diff --git a/include/net/dst.h b/include/net/dst.h
> index e3ac7d0..24d3c4e 100644
> --- a/include/net/dst.h
> +++ b/include/net/dst.h
> @@ -147,7 +147,8 @@ static inline void dst_use(struct dst_entry *dst, 
> unsigned long time)
>  {
>         dst_hold(dst);
>         dst->__use++;
> -       dst->lastuse = time;
> +       if (time != dst->lastuse)
> +               dst->lastuse = time;
>  }
I did a quick test and this patch doesn't help tbench.

Thanks,
-yanmin

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

On Fri, 2008-02-15 at 15:22 -0800, David Miller wrote:
> From: Eric Dumazet <dada1@cosmosbay.com>
> Date: Fri, 15 Feb 2008 15:21:48 +0100
> 
> > On linux-2.6.25-rc1 x86_64 :
> > 
> > offsetof(struct dst_entry, lastuse)=0xb0
> > offsetof(struct dst_entry, __refcnt)=0xb8
> > offsetof(struct dst_entry, __use)=0xbc
> > offsetof(struct dst_entry, next)=0xc0
> > 
> > So it should be optimal... I dont know why tbench prefers __refcnt being 
> > on 0xc0, since in this case lastuse will be on a different cache line...
> > 
> > Each incoming IP packet will need to change lastuse, __refcnt and __use, 
> > so keeping them in the same cache line is a win.
> > 
> > I suspect then that even this patch could help tbench, since it avoids 
> > writing lastuse...
> 
> I think your suspicions are right, and even moreso
> it helps to keep __refcnt out of the same cache line
> as input/output/ops which are read-almost-entirely :-
I think you are right. The issue is these three variables sharing the same cache line
with input/output/ops.

> )
> 
> I haven't done an exhaustive analysis, but it seems that
> the write traffic to lastuse and __refcnt are about the
> same.  However if we find that __refcnt gets hit more
> than lastuse in this workload, it explains the regression.
I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
pading before lastuse, so the 3 members are moved to next cache line. The performance is
recovered.

How about below patch? Almost all performance is recovered with the new patch.

Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>

---

--- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
+++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-21 14:36:22.000000000 +0800
@@ -52,11 +52,10 @@ struct dst_entry
 	unsigned short		header_len;	/* more space at head required */
 	unsigned short		trailer_len;	/* space to reserve at tail */
 
-	u32			metrics[RTAX_MAX];
-	struct dst_entry	*path;
-
-	unsigned long		rate_last;	/* rate limiting for ICMP */
 	unsigned int		rate_tokens;
+	unsigned long		rate_last;	/* rate limiting for ICMP */
+
+	struct dst_entry	*path;
 
 #ifdef CONFIG_NET_CLS_ROUTE
 	__u32			tclassid;
@@ -70,10 +69,12 @@ struct dst_entry
 	int			(*output)(struct sk_buff*);
 
 	struct  dst_ops	        *ops;
-		
-	unsigned long		lastuse;
+
+	u32			metrics[RTAX_MAX];
+
 	atomic_t		__refcnt;	/* client references	*/
 	int			__use;
+	unsigned long		lastuse;
 	union {
 		struct dst_entry *next;
 		struct rtable    *rt_next;

Re: tbench regression in 2.6.25-rc1

[Eric Dumazet]

On Mon, 18 Feb 2008 16:12:38 +0800
"Zhang, Yanmin" <yanmin_zhang@linux.intel.com> wrote:

> On Fri, 2008-02-15 at 15:22 -0800, David Miller wrote:
> > From: Eric Dumazet <dada1@cosmosbay.com>
> > Date: Fri, 15 Feb 2008 15:21:48 +0100
> > 
> > > On linux-2.6.25-rc1 x86_64 :
> > > 
> > > offsetof(struct dst_entry, lastuse)=0xb0
> > > offsetof(struct dst_entry, __refcnt)=0xb8
> > > offsetof(struct dst_entry, __use)=0xbc
> > > offsetof(struct dst_entry, next)=0xc0
> > > 
> > > So it should be optimal... I dont know why tbench prefers __refcnt being 
> > > on 0xc0, since in this case lastuse will be on a different cache line...
> > > 
> > > Each incoming IP packet will need to change lastuse, __refcnt and __use, 
> > > so keeping them in the same cache line is a win.
> > > 
> > > I suspect then that even this patch could help tbench, since it avoids 
> > > writing lastuse...
> > 
> > I think your suspicions are right, and even moreso
> > it helps to keep __refcnt out of the same cache line
> > as input/output/ops which are read-almost-entirely :-
> I think you are right. The issue is these three variables sharing the same cache line
> with input/output/ops.
> 
> > )
> > 
> > I haven't done an exhaustive analysis, but it seems that
> > the write traffic to lastuse and __refcnt are about the
> > same.  However if we find that __refcnt gets hit more
> > than lastuse in this workload, it explains the regression.
> I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
> pading before lastuse, so the 3 members are moved to next cache line. The performance is
> recovered.
> 
> How about below patch? Almost all performance is recovered with the new patch.
> 
> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
> 
> ---
> 
> --- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
> +++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-21 14:36:22.000000000 +0800
> @@ -52,11 +52,10 @@ struct dst_entry
>  	unsigned short		header_len;	/* more space at head required */
>  	unsigned short		trailer_len;	/* space to reserve at tail */
>  
> -	u32			metrics[RTAX_MAX];
> -	struct dst_entry	*path;
> -
> -	unsigned long		rate_last;	/* rate limiting for ICMP */
>  	unsigned int		rate_tokens;
> +	unsigned long		rate_last;	/* rate limiting for ICMP */
> +
> +	struct dst_entry	*path;
>  
>  #ifdef CONFIG_NET_CLS_ROUTE
>  	__u32			tclassid;
> @@ -70,10 +69,12 @@ struct dst_entry
>  	int			(*output)(struct sk_buff*);
>  
>  	struct  dst_ops	        *ops;
> -		
> -	unsigned long		lastuse;
> +
> +	u32			metrics[RTAX_MAX];
> +
>  	atomic_t		__refcnt;	/* client references	*/
>  	int			__use;
> +	unsigned long		lastuse;
>  	union {
>  		struct dst_entry *next;
>  		struct rtable    *rt_next;
> 
> 

Well, after this patch, we grow dst_entry by 8 bytes :

sizeof(struct dst_entry)=0xd0
offsetof(struct dst_entry, input)=0x68
offsetof(struct dst_entry, output)=0x70
offsetof(struct dst_entry, __refcnt)=0xb4
offsetof(struct dst_entry, lastuse)=0xc0
offsetof(struct dst_entry, __use)=0xb8
sizeof(struct rtable)=0x140


So we dirty two cache lines instead of one, unless your cpu have 128 bytes cache lines ?

I am quite suprised that my patch to not change lastuse if already set to jiffies changes nothing...

If you have some time, could you also test this (unrelated) patch ?

We can avoid dirty all the time a cache line of loopback device.

diff --git a/drivers/net/loopback.c b/drivers/net/loopback.c
index f2a6e71..0a4186a 100644
--- a/drivers/net/loopback.c
+++ b/drivers/net/loopback.c
@@ -150,7 +150,10 @@ static int loopback_xmit(struct sk_buff *skb, struct net_device *dev)
                return 0;
        }
 #endif
-       dev->last_rx = jiffies;
+#ifdef CONFIG_SMP
+       if (dev->last_rx != jiffies)
+#endif
+               dev->last_rx = jiffies;
 
        /* it's OK to use per_cpu_ptr() because BHs are off */
        pcpu_lstats = netdev_priv(dev);

Re: tbench regression in 2.6.25-rc1

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 610 bytes --]

On Mon, 18 Feb 2008 16:12:38 +0800, "Zhang, Yanmin" said:

> I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
> pading before lastuse, so the 3 members are moved to next cache line. The performance is
> recovered.
> 
> How about below patch? Almost all performance is recovered with the new patch.
> 
> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>

Could you add a comment someplace that says "refcnt wants to be on a different
cache line from input/output/ops or performance tanks badly", to warn some
future kernel hacker who starts adding new fields to the structure?

[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

On Mon, 2008-02-18 at 11:11 +0100, Eric Dumazet wrote:
> On Mon, 18 Feb 2008 16:12:38 +0800
> "Zhang, Yanmin" <yanmin_zhang@linux.intel.com> wrote:
> 
> > On Fri, 2008-02-15 at 15:22 -0800, David Miller wrote:
> > > From: Eric Dumazet <dada1@cosmosbay.com>
> > > Date: Fri, 15 Feb 2008 15:21:48 +0100
> > > 
> > > > On linux-2.6.25-rc1 x86_64 :
> > > > 
> > > > offsetof(struct dst_entry, lastuse)=0xb0
> > > > offsetof(struct dst_entry, __refcnt)=0xb8
> > > > offsetof(struct dst_entry, __use)=0xbc
> > > > offsetof(struct dst_entry, next)=0xc0
> > > > 
> > > > So it should be optimal... I dont know why tbench prefers __refcnt being 
> > > > on 0xc0, since in this case lastuse will be on a different cache line...
> > > > 
> > > > Each incoming IP packet will need to change lastuse, __refcnt and __use, 
> > > > so keeping them in the same cache line is a win.
> > > > 
> > > > I suspect then that even this patch could help tbench, since it avoids 
> > > > writing lastuse...
> > > 
> > > I think your suspicions are right, and even moreso
> > > it helps to keep __refcnt out of the same cache line
> > > as input/output/ops which are read-almost-entirely :-
> > I think you are right. The issue is these three variables sharing the same cache line
> > with input/output/ops.
> > 
> > > )
> > > 
> > > I haven't done an exhaustive analysis, but it seems that
> > > the write traffic to lastuse and __refcnt are about the
> > > same.  However if we find that __refcnt gets hit more
> > > than lastuse in this workload, it explains the regression.
> > I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
> > pading before lastuse, so the 3 members are moved to next cache line. The performance is
> > recovered.
> > 
> > How about below patch? Almost all performance is recovered with the new patch.
> > 
> > Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
> > 
> > ---
> > 
> > --- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
> > +++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-21 14:36:22.000000000 +0800
> > @@ -52,11 +52,10 @@ struct dst_entry
> >  	unsigned short		header_len;	/* more space at head required */
> >  	unsigned short		trailer_len;	/* space to reserve at tail */
> >  
> > -	u32			metrics[RTAX_MAX];
> > -	struct dst_entry	*path;
> > -
> > -	unsigned long		rate_last;	/* rate limiting for ICMP */
> >  	unsigned int		rate_tokens;
> > +	unsigned long		rate_last;	/* rate limiting for ICMP */
> > +
> > +	struct dst_entry	*path;
> >  
> >  #ifdef CONFIG_NET_CLS_ROUTE
> >  	__u32			tclassid;
> > @@ -70,10 +69,12 @@ struct dst_entry
> >  	int			(*output)(struct sk_buff*);
> >  
> >  	struct  dst_ops	        *ops;
> > -		
> > -	unsigned long		lastuse;
> > +
> > +	u32			metrics[RTAX_MAX];
> > +
> >  	atomic_t		__refcnt;	/* client references	*/
> >  	int			__use;
> > +	unsigned long		lastuse;
> >  	union {
> >  		struct dst_entry *next;
> >  		struct rtable    *rt_next;
> > 
> > 
> 
> Well, after this patch, we grow dst_entry by 8 bytes :
With my .config, it doesn't grow. Perhaps because of CONFIG_NET_CLS_ROUTE, I don't
enable it. I will move tclassid under ops.

> 
> sizeof(struct dst_entry)=0xd0
> offsetof(struct dst_entry, input)=0x68
> offsetof(struct dst_entry, output)=0x70
> offsetof(struct dst_entry, __refcnt)=0xb4
> offsetof(struct dst_entry, lastuse)=0xc0
> offsetof(struct dst_entry, __use)=0xb8
> sizeof(struct rtable)=0x140
> 
> 
> So we dirty two cache lines instead of one, unless your cpu have 128 bytes cache lines ?
> 
> I am quite suprised that my patch to not change lastuse if already set to jiffies changes nothing...
> 
> If you have some time, could you also test this (unrelated) patch ?
> 
> We can avoid dirty all the time a cache line of loopback device.
> 
> diff --git a/drivers/net/loopback.c b/drivers/net/loopback.c
> index f2a6e71..0a4186a 100644
> --- a/drivers/net/loopback.c
> +++ b/drivers/net/loopback.c
> @@ -150,7 +150,10 @@ static int loopback_xmit(struct sk_buff *skb, struct net_device *dev)
>                 return 0;
>         }
>  #endif
> -       dev->last_rx = jiffies;
> +#ifdef CONFIG_SMP
> +       if (dev->last_rx != jiffies)
> +#endif
> +               dev->last_rx = jiffies;
>  
>         /* it's OK to use per_cpu_ptr() because BHs are off */
>         pcpu_lstats = netdev_priv(dev);
> 
Although I didn't test it, I don't think it's ok. The key is __refcnt shares the same
cache line with ops/input/output.

-yanmin

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

On Mon, 2008-02-18 at 12:33 -0500, Valdis.Kletnieks@vt.edu wrote: 
> On Mon, 18 Feb 2008 16:12:38 +0800, "Zhang, Yanmin" said:
> 
> > I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
> > pading before lastuse, so the 3 members are moved to next cache line. The performance is
> > recovered.
> > 
> > How about below patch? Almost all performance is recovered with the new patch.
> > 
> > Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
> 
> Could you add a comment someplace that says "refcnt wants to be on a different
> cache line from input/output/ops or performance tanks badly", to warn some
> future kernel hacker who starts adding new fields to the structure?
Ok. Below is the new patch.

1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE=y. So sizeof(dst_entry)=200
no matter if CONFIG_NET_CLS_ROUTE=y/n. I tested many patches on my 16-core tigerton by
moving tclassid to different place. It looks like tclassid could also have impact on
performance.
If moving tclassid before metrics, or just don't move tclassid, the performance isn't
good. So I move it behind metrics.

2) Add comments before __refcnt.

If CONFIG_NET_CLS_ROUTE=y, the result with below patch is about 18% better than
the one without the patch.

If CONFIG_NET_CLS_ROUTE=n, the result with below patch is about 30% better than
the one without the patch.

Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>

---

--- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
+++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-22 12:52:19.000000000 +0800
@@ -52,15 +52,10 @@ struct dst_entry
 	unsigned short		header_len;	/* more space at head required */
 	unsigned short		trailer_len;	/* space to reserve at tail */
 
-	u32			metrics[RTAX_MAX];
-	struct dst_entry	*path;
-
-	unsigned long		rate_last;	/* rate limiting for ICMP */
 	unsigned int		rate_tokens;
+	unsigned long		rate_last;	/* rate limiting for ICMP */
 
-#ifdef CONFIG_NET_CLS_ROUTE
-	__u32			tclassid;
-#endif
+	struct dst_entry	*path;
 
 	struct neighbour	*neighbour;
 	struct hh_cache		*hh;
@@ -70,10 +65,20 @@ struct dst_entry
 	int			(*output)(struct sk_buff*);
 
 	struct  dst_ops	        *ops;
-		
-	unsigned long		lastuse;
+
+	u32			metrics[RTAX_MAX];
+
+#ifdef CONFIG_NET_CLS_ROUTE
+	__u32			tclassid;
+#endif
+
+	/*
+	 * __refcnt wants to be on a different cache line from
+	 * input/output/ops or performance tanks badly
+	 */
 	atomic_t		__refcnt;	/* client references	*/
 	int			__use;
+	unsigned long		lastuse;
 	union {
 		struct dst_entry *next;
 		struct rtable    *rt_next;

Re: tbench regression in 2.6.25-rc1

[Eric Dumazet]

Zhang, Yanmin a écrit :
> On Mon, 2008-02-18 at 11:11 +0100, Eric Dumazet wrote:
>> On Mon, 18 Feb 2008 16:12:38 +0800
>> "Zhang, Yanmin" <yanmin_zhang@linux.intel.com> wrote:
>>
>>> On Fri, 2008-02-15 at 15:22 -0800, David Miller wrote:
>>>> From: Eric Dumazet <dada1@cosmosbay.com>
>>>> Date: Fri, 15 Feb 2008 15:21:48 +0100
>>>>
>>>>> On linux-2.6.25-rc1 x86_64 :
>>>>>
>>>>> offsetof(struct dst_entry, lastuse)=0xb0
>>>>> offsetof(struct dst_entry, __refcnt)=0xb8
>>>>> offsetof(struct dst_entry, __use)=0xbc
>>>>> offsetof(struct dst_entry, next)=0xc0
>>>>>
>>>>> So it should be optimal... I dont know why tbench prefers __refcnt being 
>>>>> on 0xc0, since in this case lastuse will be on a different cache line...
>>>>>
>>>>> Each incoming IP packet will need to change lastuse, __refcnt and __use, 
>>>>> so keeping them in the same cache line is a win.
>>>>>
>>>>> I suspect then that even this patch could help tbench, since it avoids 
>>>>> writing lastuse...
>>>> I think your suspicions are right, and even moreso
>>>> it helps to keep __refcnt out of the same cache line
>>>> as input/output/ops which are read-almost-entirely :-
>>> I think you are right. The issue is these three variables sharing the same cache line
>>> with input/output/ops.
>>>
>>>> )
>>>>
>>>> I haven't done an exhaustive analysis, but it seems that
>>>> the write traffic to lastuse and __refcnt are about the
>>>> same.  However if we find that __refcnt gets hit more
>>>> than lastuse in this workload, it explains the regression.
>>> I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
>>> pading before lastuse, so the 3 members are moved to next cache line. The performance is
>>> recovered.
>>>
>>> How about below patch? Almost all performance is recovered with the new patch.
>>>
>>> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
>>>
>>> ---
>>>
>>> --- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
>>> +++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-21 14:36:22.000000000 +0800
>>> @@ -52,11 +52,10 @@ struct dst_entry
>>>  	unsigned short		header_len;	/* more space at head required */
>>>  	unsigned short		trailer_len;	/* space to reserve at tail */
>>>  
>>> -	u32			metrics[RTAX_MAX];
>>> -	struct dst_entry	*path;
>>> -
>>> -	unsigned long		rate_last;	/* rate limiting for ICMP */
>>>  	unsigned int		rate_tokens;
>>> +	unsigned long		rate_last;	/* rate limiting for ICMP */
>>> +
>>> +	struct dst_entry	*path;
>>>  
>>>  #ifdef CONFIG_NET_CLS_ROUTE
>>>  	__u32			tclassid;
>>> @@ -70,10 +69,12 @@ struct dst_entry
>>>  	int			(*output)(struct sk_buff*);
>>>  
>>>  	struct  dst_ops	        *ops;
>>> -		
>>> -	unsigned long		lastuse;
>>> +
>>> +	u32			metrics[RTAX_MAX];
>>> +
>>>  	atomic_t		__refcnt;	/* client references	*/
>>>  	int			__use;
>>> +	unsigned long		lastuse;
>>>  	union {
>>>  		struct dst_entry *next;
>>>  		struct rtable    *rt_next;
>>>
>>>
>> Well, after this patch, we grow dst_entry by 8 bytes :
> With my .config, it doesn't grow. Perhaps because of CONFIG_NET_CLS_ROUTE, I don't
> enable it. I will move tclassid under ops.
> 
>> sizeof(struct dst_entry)=0xd0
>> offsetof(struct dst_entry, input)=0x68
>> offsetof(struct dst_entry, output)=0x70
>> offsetof(struct dst_entry, __refcnt)=0xb4
>> offsetof(struct dst_entry, lastuse)=0xc0
>> offsetof(struct dst_entry, __use)=0xb8
>> sizeof(struct rtable)=0x140
>>
>>
>> So we dirty two cache lines instead of one, unless your cpu have 128 bytes cache lines ?
>>
>> I am quite suprised that my patch to not change lastuse if already set to jiffies changes nothing...
>>
>> If you have some time, could you also test this (unrelated) patch ?
>>
>> We can avoid dirty all the time a cache line of loopback device.
>>
>> diff --git a/drivers/net/loopback.c b/drivers/net/loopback.c
>> index f2a6e71..0a4186a 100644
>> --- a/drivers/net/loopback.c
>> +++ b/drivers/net/loopback.c
>> @@ -150,7 +150,10 @@ static int loopback_xmit(struct sk_buff *skb, struct net_device *dev)
>>                 return 0;
>>         }
>>  #endif
>> -       dev->last_rx = jiffies;
>> +#ifdef CONFIG_SMP
>> +       if (dev->last_rx != jiffies)
>> +#endif
>> +               dev->last_rx = jiffies;
>>  
>>         /* it's OK to use per_cpu_ptr() because BHs are off */
>>         pcpu_lstats = netdev_priv(dev);
>>
> Although I didn't test it, I don't think it's ok. The key is __refcnt shares the same
> cache line with ops/input/output.
> 

Note it was unrelated to struct dst, but dirtying of one cache line of 
'loopback netdevice'

I tested it, and tbench result was better with this patch : 890 MB/s instead 
of 870 MB/s on a bi dual core machine.


I was curious of the potential gain on your 16 cores (4x4) machine.

Re: tbench regression in 2.6.25-rc1

[Eric Dumazet]

Zhang, Yanmin a écrit :
> On Mon, 2008-02-18 at 12:33 -0500, Valdis.Kletnieks@vt.edu wrote: 
>> On Mon, 18 Feb 2008 16:12:38 +0800, "Zhang, Yanmin" said:
>>
>>> I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
>>> pading before lastuse, so the 3 members are moved to next cache line. The performance is
>>> recovered.
>>>
>>> How about below patch? Almost all performance is recovered with the new patch.
>>>
>>> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
>> Could you add a comment someplace that says "refcnt wants to be on a different
>> cache line from input/output/ops or performance tanks badly", to warn some
>> future kernel hacker who starts adding new fields to the structure?
> Ok. Below is the new patch.
> 
> 1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE=y. So sizeof(dst_entry)=200
> no matter if CONFIG_NET_CLS_ROUTE=y/n. I tested many patches on my 16-core tigerton by
> moving tclassid to different place. It looks like tclassid could also have impact on
> performance.
> If moving tclassid before metrics, or just don't move tclassid, the performance isn't
> good. So I move it behind metrics.
> 
> 2) Add comments before __refcnt.
> 
> If CONFIG_NET_CLS_ROUTE=y, the result with below patch is about 18% better than
> the one without the patch.
> 
> If CONFIG_NET_CLS_ROUTE=n, the result with below patch is about 30% better than
> the one without the patch.
> 
> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
> 
> ---
> 
> --- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
> +++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-22 12:52:19.000000000 +0800
> @@ -52,15 +52,10 @@ struct dst_entry
>  	unsigned short		header_len;	/* more space at head required */
>  	unsigned short		trailer_len;	/* space to reserve at tail */
>  
> -	u32			metrics[RTAX_MAX];
> -	struct dst_entry	*path;
> -
> -	unsigned long		rate_last;	/* rate limiting for ICMP */
>  	unsigned int		rate_tokens;
> +	unsigned long		rate_last;	/* rate limiting for ICMP */
>  
> -#ifdef CONFIG_NET_CLS_ROUTE
> -	__u32			tclassid;
> -#endif
> +	struct dst_entry	*path;
>  
>  	struct neighbour	*neighbour;
>  	struct hh_cache		*hh;
> @@ -70,10 +65,20 @@ struct dst_entry
>  	int			(*output)(struct sk_buff*);
>  
>  	struct  dst_ops	        *ops;
> -		
> -	unsigned long		lastuse;
> +
> +	u32			metrics[RTAX_MAX];
> +
> +#ifdef CONFIG_NET_CLS_ROUTE
> +	__u32			tclassid;
> +#endif
> +
> +	/*
> +	 * __refcnt wants to be on a different cache line from
> +	 * input/output/ops or performance tanks badly
> +	 */
>  	atomic_t		__refcnt;	/* client references	*/
>  	int			__use;
> +	unsigned long		lastuse;
>  	union {
>  		struct dst_entry *next;
>  		struct rtable    *rt_next;
> 
> 
> 

I prefer this patch, but unfortunatly your perf numbers are for 64 bits kernels.

Could you please test now with 32 bits one ?

Thank you

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=utf-8, Size: 5246 bytes --]

On Tue, 2008-02-19 at 08:35 +0100, Eric Dumazet wrote:
> Zhang, Yanmin a écrit :
> > On Mon, 2008-02-18 at 11:11 +0100, Eric Dumazet wrote:
> >> On Mon, 18 Feb 2008 16:12:38 +0800
> >> "Zhang, Yanmin" <yanmin_zhang@linux.intel.com> wrote:
> >>
> >>> On Fri, 2008-02-15 at 15:22 -0800, David Miller wrote:
> >>>> From: Eric Dumazet <dada1@cosmosbay.com>
> >>>> Date: Fri, 15 Feb 2008 15:21:48 +0100
> >>>>
> >>>>> On linux-2.6.25-rc1 x86_64 :
> >>>>>
> >>>>> offsetof(struct dst_entry, lastuse)=0xb0
> >>>>> offsetof(struct dst_entry, __refcnt)=0xb8
> >>>>> offsetof(struct dst_entry, __use)=0xbc
> >>>>> offsetof(struct dst_entry, next)=0xc0
> >>>>>
> >>>>> So it should be optimal... I dont know why tbench prefers __refcnt being 
> >>>>> on 0xc0, since in this case lastuse will be on a different cache line...
> >>>>>
> >>>>> Each incoming IP packet will need to change lastuse, __refcnt and __use, 
> >>>>> so keeping them in the same cache line is a win.
> >>>>>
> >>>>> I suspect then that even this patch could help tbench, since it avoids 
> >>>>> writing lastuse...
> >>>> I think your suspicions are right, and even moreso
> >>>> it helps to keep __refcnt out of the same cache line
> >>>> as input/output/ops which are read-almost-entirely :-
> >>> I think you are right. The issue is these three variables sharing the same cache line
> >>> with input/output/ops.
> >>>
> >>>> )
> >>>>
> >>>> I haven't done an exhaustive analysis, but it seems that
> >>>> the write traffic to lastuse and __refcnt are about the
> >>>> same.  However if we find that __refcnt gets hit more
> >>>> than lastuse in this workload, it explains the regression.
> >>> I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
> >>> pading before lastuse, so the 3 members are moved to next cache line. The performance is
> >>> recovered.
> >>>
> >>> How about below patch? Almost all performance is recovered with the new patch.
> >>>
> >>> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
> >>>
> >>> ---
> >>>
> >>> --- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
> >>> +++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-21 14:36:22.000000000 +0800
> >>> @@ -52,11 +52,10 @@ struct dst_entry
> >>>  	unsigned short		header_len;	/* more space at head required */
> >>>  	unsigned short		trailer_len;	/* space to reserve at tail */
> >>>  
> >>> -	u32			metrics[RTAX_MAX];
> >>> -	struct dst_entry	*path;
> >>> -
> >>> -	unsigned long		rate_last;	/* rate limiting for ICMP */
> >>>  	unsigned int		rate_tokens;
> >>> +	unsigned long		rate_last;	/* rate limiting for ICMP */
> >>> +
> >>> +	struct dst_entry	*path;
> >>>  
> >>>  #ifdef CONFIG_NET_CLS_ROUTE
> >>>  	__u32			tclassid;
> >>> @@ -70,10 +69,12 @@ struct dst_entry
> >>>  	int			(*output)(struct sk_buff*);
> >>>  
> >>>  	struct  dst_ops	        *ops;
> >>> -		
> >>> -	unsigned long		lastuse;
> >>> +
> >>> +	u32			metrics[RTAX_MAX];
> >>> +
> >>>  	atomic_t		__refcnt;	/* client references	*/
> >>>  	int			__use;
> >>> +	unsigned long		lastuse;
> >>>  	union {
> >>>  		struct dst_entry *next;
> >>>  		struct rtable    *rt_next;
> >>>
> >>>
> >> Well, after this patch, we grow dst_entry by 8 bytes :
> > With my .config, it doesn't grow. Perhaps because of CONFIG_NET_CLS_ROUTE, I don't
> > enable it. I will move tclassid under ops.
> > 
> >> sizeof(struct dst_entry)=0xd0
> >> offsetof(struct dst_entry, input)=0x68
> >> offsetof(struct dst_entry, output)=0x70
> >> offsetof(struct dst_entry, __refcnt)=0xb4
> >> offsetof(struct dst_entry, lastuse)=0xc0
> >> offsetof(struct dst_entry, __use)=0xb8
> >> sizeof(struct rtable)=0x140
> >>
> >>
> >> So we dirty two cache lines instead of one, unless your cpu have 128 bytes cache lines ?
> >>
> >> I am quite suprised that my patch to not change lastuse if already set to jiffies changes nothing...
> >>
> >> If you have some time, could you also test this (unrelated) patch ?
> >>
> >> We can avoid dirty all the time a cache line of loopback device.
> >>
> >> diff --git a/drivers/net/loopback.c b/drivers/net/loopback.c
> >> index f2a6e71..0a4186a 100644
> >> --- a/drivers/net/loopback.c
> >> +++ b/drivers/net/loopback.c
> >> @@ -150,7 +150,10 @@ static int loopback_xmit(struct sk_buff *skb, struct net_device *dev)
> >>                 return 0;
> >>         }
> >>  #endif
> >> -       dev->last_rx = jiffies;
> >> +#ifdef CONFIG_SMP
> >> +       if (dev->last_rx != jiffies)
> >> +#endif
> >> +               dev->last_rx = jiffies;
> >>  
> >>         /* it's OK to use per_cpu_ptr() because BHs are off */
> >>         pcpu_lstats = netdev_priv(dev);
> >>
> > Although I didn't test it, I don't think it's ok. The key is __refcnt shares the same
> > cache line with ops/input/output.
> > 
> 
> Note it was unrelated to struct dst, but dirtying of one cache line of 
> 'loopback netdevice'
> 
> I tested it, and tbench result was better with this patch : 890 MB/s instead 
> of 870 MB/s on a bi dual core machine.
I tested your new patch and it doesn't help tbench.

On my 8-core stoakley machine, the regression is only 5%, but it's 30% on 16-core tigerton.
It looks like the scalability is poor.

> 
> 
> I was curious of the potential gain on your 16 cores (4x4) machine.

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=utf-8, Size: 3960 bytes --]

On Tue, 2008-02-19 at 08:40 +0100, Eric Dumazet wrote:
> Zhang, Yanmin a écrit :
> > On Mon, 2008-02-18 at 12:33 -0500, Valdis.Kletnieks@vt.edu wrote: 
> >> On Mon, 18 Feb 2008 16:12:38 +0800, "Zhang, Yanmin" said:
> >>
> >>> I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
> >>> pading before lastuse, so the 3 members are moved to next cache line. The performance is
> >>> recovered.
> >>>
> >>> How about below patch? Almost all performance is recovered with the new patch.
> >>>
> >>> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
> >> Could you add a comment someplace that says "refcnt wants to be on a different
> >> cache line from input/output/ops or performance tanks badly", to warn some
> >> future kernel hacker who starts adding new fields to the structure?
> > Ok. Below is the new patch.
> > 
> > 1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE=y. So sizeof(dst_entry)=200
> > no matter if CONFIG_NET_CLS_ROUTE=y/n. I tested many patches on my 16-core tigerton by
> > moving tclassid to different place. It looks like tclassid could also have impact on
> > performance.
> > If moving tclassid before metrics, or just don't move tclassid, the performance isn't
> > good. So I move it behind metrics.
> > 
> > 2) Add comments before __refcnt.
> > 
> > If CONFIG_NET_CLS_ROUTE=y, the result with below patch is about 18% better than
> > the one without the patch.
> > 
> > If CONFIG_NET_CLS_ROUTE=n, the result with below patch is about 30% better than
> > the one without the patch.
> > 
> > Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
> > 
> > ---
> > 
> > --- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
> > +++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-22 12:52:19.000000000 +0800
> > @@ -52,15 +52,10 @@ struct dst_entry
> >  	unsigned short		header_len;	/* more space at head required */
> >  	unsigned short		trailer_len;	/* space to reserve at tail */
> >  
> > -	u32			metrics[RTAX_MAX];
> > -	struct dst_entry	*path;
> > -
> > -	unsigned long		rate_last;	/* rate limiting for ICMP */
> >  	unsigned int		rate_tokens;
> > +	unsigned long		rate_last;	/* rate limiting for ICMP */
> >  
> > -#ifdef CONFIG_NET_CLS_ROUTE
> > -	__u32			tclassid;
> > -#endif
> > +	struct dst_entry	*path;
> >  
> >  	struct neighbour	*neighbour;
> >  	struct hh_cache		*hh;
> > @@ -70,10 +65,20 @@ struct dst_entry
> >  	int			(*output)(struct sk_buff*);
> >  
> >  	struct  dst_ops	        *ops;
> > -		
> > -	unsigned long		lastuse;
> > +
> > +	u32			metrics[RTAX_MAX];
> > +
> > +#ifdef CONFIG_NET_CLS_ROUTE
> > +	__u32			tclassid;
> > +#endif
> > +
> > +	/*
> > +	 * __refcnt wants to be on a different cache line from
> > +	 * input/output/ops or performance tanks badly
> > +	 */
> >  	atomic_t		__refcnt;	/* client references	*/
> >  	int			__use;
> > +	unsigned long		lastuse;
> >  	union {
> >  		struct dst_entry *next;
> >  		struct rtable    *rt_next;
> > 
> > 
> > 
> 
> I prefer this patch, but unfortunatly your perf numbers are for 64 bits kernels.
> 
> Could you please test now with 32 bits one ?
I tested it with 32bit 2.6.25-rc1 on 8-core stoakley. The result almost has no difference
between pure kernel and patched kernel.

New update: On 8-core stoakley, the regression becomes 2~3% with kernel 2.6.25-rc2. On
tigerton, the regression is still 30% with 2.6.25-rc2. On Tulsa( 8 cores+hyperthreading),
the regression is still 4% with 2.6.25-rc2.

With my patch, on tigerton, almost all regression disappears. On tulsa, only about 2%
regression disappears.

So this issue is triggerred with multiple-cpu. Perhaps process scheduler is another
factor causing the issue to happen, but it's very hard to change scheduler.


Eric,

I tested your new patch in function loopback_xmit. It has no improvement, while it doesn't
introduce new issues. As you tested it on dual-core machine and got improvement, how about
merging your patch with mine?

-yanmin

Re: tbench regression in 2.6.25-rc1

[Eric Dumazet]

Zhang, Yanmin a écrit :
> On Tue, 2008-02-19 at 08:40 +0100, Eric Dumazet wrote:
>> Zhang, Yanmin a �crit :
>>> On Mon, 2008-02-18 at 12:33 -0500, Valdis.Kletnieks@vt.edu wrote: 
>>>> On Mon, 18 Feb 2008 16:12:38 +0800, "Zhang, Yanmin" said:
>>>>
>>>>> I also think __refcnt is the key. I did a new testing by adding 2 unsigned long
>>>>> pading before lastuse, so the 3 members are moved to next cache line. The performance is
>>>>> recovered.
>>>>>
>>>>> How about below patch? Almost all performance is recovered with the new patch.
>>>>>
>>>>> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
>>>> Could you add a comment someplace that says "refcnt wants to be on a different
>>>> cache line from input/output/ops or performance tanks badly", to warn some
>>>> future kernel hacker who starts adding new fields to the structure?
>>> Ok. Below is the new patch.
>>>
>>> 1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE=y. So sizeof(dst_entry)=200
>>> no matter if CONFIG_NET_CLS_ROUTE=y/n. I tested many patches on my 16-core tigerton by
>>> moving tclassid to different place. It looks like tclassid could also have impact on
>>> performance.
>>> If moving tclassid before metrics, or just don't move tclassid, the performance isn't
>>> good. So I move it behind metrics.
>>>
>>> 2) Add comments before __refcnt.
>>>
>>> If CONFIG_NET_CLS_ROUTE=y, the result with below patch is about 18% better than
>>> the one without the patch.
>>>
>>> If CONFIG_NET_CLS_ROUTE=n, the result with below patch is about 30% better than
>>> the one without the patch.
>>>
>>> Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
>>>
>>> ---
>>>
>>> --- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
>>> +++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-22 12:52:19.000000000 +0800
>>> @@ -52,15 +52,10 @@ struct dst_entry
>>>  	unsigned short		header_len;	/* more space at head required */
>>>  	unsigned short		trailer_len;	/* space to reserve at tail */
>>>  
>>> -	u32			metrics[RTAX_MAX];
>>> -	struct dst_entry	*path;
>>> -
>>> -	unsigned long		rate_last;	/* rate limiting for ICMP */
>>>  	unsigned int		rate_tokens;
>>> +	unsigned long		rate_last;	/* rate limiting for ICMP */
>>>  
>>> -#ifdef CONFIG_NET_CLS_ROUTE
>>> -	__u32			tclassid;
>>> -#endif
>>> +	struct dst_entry	*path;
>>>  
>>>  	struct neighbour	*neighbour;
>>>  	struct hh_cache		*hh;
>>> @@ -70,10 +65,20 @@ struct dst_entry
>>>  	int			(*output)(struct sk_buff*);
>>>  
>>>  	struct  dst_ops	        *ops;
>>> -		
>>> -	unsigned long		lastuse;
>>> +
>>> +	u32			metrics[RTAX_MAX];
>>> +
>>> +#ifdef CONFIG_NET_CLS_ROUTE
>>> +	__u32			tclassid;
>>> +#endif
>>> +
>>> +	/*
>>> +	 * __refcnt wants to be on a different cache line from
>>> +	 * input/output/ops or performance tanks badly
>>> +	 */
>>>  	atomic_t		__refcnt;	/* client references	*/
>>>  	int			__use;
>>> +	unsigned long		lastuse;
>>>  	union {
>>>  		struct dst_entry *next;
>>>  		struct rtable    *rt_next;
>>>
>>>
>>>
>> I prefer this patch, but unfortunatly your perf numbers are for 64 bits kernels.
>>
>> Could you please test now with 32 bits one ?
> I tested it with 32bit 2.6.25-rc1 on 8-core stoakley. The result almost has no difference
> between pure kernel and patched kernel.
> 
> New update: On 8-core stoakley, the regression becomes 2~3% with kernel 2.6.25-rc2. On
> tigerton, the regression is still 30% with 2.6.25-rc2. On Tulsa( 8 cores+hyperthreading),
> the regression is still 4% with 2.6.25-rc2.
> 
> With my patch, on tigerton, almost all regression disappears. On tulsa, only about 2%
> regression disappears.
> 
> So this issue is triggerred with multiple-cpu. Perhaps process scheduler is another
> factor causing the issue to happen, but it's very hard to change scheduler.
> 

Thanks very much Yanmin, I think we can apply your patch as is, if no 
regression was found for 32bits.

> 
> Eric,
> 
> I tested your new patch in function loopback_xmit. It has no improvement, while it doesn't
> introduce new issues. As you tested it on dual-core machine and got improvement, how about
> merging your patch with mine?

No, thank you, that was an experiment and is not related to your findings on 
dst_entry.

I am currently working on a 'distributed refcount' infrastructure, to be able 
to spread on several nodes (for NUMA machines) or several cache lines (normal 
SMP machines)  the high pressure we currently have on some refcnt (struct 
dst_entry, struct net_device, and many more refcnts ...)

Instead of NR_CPUS allocations, goal is to be able to restrict to a small 
value like 4, 8 or 16 the number of 32bits entities used to store one refcnt, 
even if NR_CPUS=4096 or so.

atomic_inc(&p->refcnt) ->  distref_inc(&p->refcnt)

distref_inc(struct distref *p)
{
	atomic_inc(myptr[p->offset]);
}

Re: tbench regression in 2.6.25-rc1

[David Miller]

From: Eric Dumazet <dada1@cosmosbay.com>
Date: Wed, 20 Feb 2008 08:38:17 +0100

> Thanks very much Yanmin, I think we can apply your patch as is, if no 
> regression was found for 32bits.

Great.  Can I get a resubmission of the patch with a cleaned up
changelog entry that describes in the regression along with the
changelog bits I saw in the most recent version of the patch?

An explicit "Acked-by:" from Eric would be nice too :-)

Thanks!

Re: tbench regression in 2.6.25-rc1

[Zhang, Yanmin]

Comparing with kernel 2.6.24, tbench result has regression with
2.6.25-rc1.
1) On 2 quad-core processor stoakley: 4%.
2) On 4 quad-core processor tigerton: more than 30%.

bisect located below patch.

b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit
commit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b
Author: Herbert Xu <herbert@gondor.apana.org.au>
Date:   Tue Nov 13 21:33:32 2007 -0800

    [IPV6]: Move nfheader_len into rt6_info
    
    The dst member nfheader_len is only used by IPv6.  It's also currently
    creating a rather ugly alignment hole in struct dst.  Therefore this patch
    moves it from there into struct rt6_info.

Above patch changes the cache line alignment, especially member __refcnt. I did a 
testing by adding 2 unsigned long pading before lastuse, so the 3 members,
lastuse/__refcnt/__use, are moved to next cache line. The performance is recovered.

I created a patch to rearrange the members in struct dst_entry.

With Eric and Valdis Kletnieks's suggestion, I made finer arrangement.
1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE=y. So sizeof(dst_entry)=200
no matter if CONFIG_NET_CLS_ROUTE=y/n. I tested many patches on my 16-core tigerton by
moving tclassid to different place. It looks like tclassid could also have impact on
performance.
If moving tclassid before metrics, or just don't move tclassid, the performance isn't
good. So I move it behind metrics.
2) Add comments before __refcnt.

On 16-core tigerton:
If CONFIG_NET_CLS_ROUTE=y, the result with below patch is about 18% better than
the one without the patch;
If CONFIG_NET_CLS_ROUTE=n, the result with below patch is about 30% better than
the one without the patch.

With 32bit 2.6.25-rc1 on 8-core stoakley, the new patch doesn't introduce regression.

Thank Eric, Valdis, and David!

Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Acked-by: Eric Dumazet <dada1@cosmosbay.com>

---

--- linux-2.6.25-rc1/include/net/dst.h	2008-02-21 14:33:43.000000000 +0800
+++ linux-2.6.25-rc1_work/include/net/dst.h	2008-02-22 12:52:19.000000000 +0800
@@ -52,15 +52,10 @@ struct dst_entry
 	unsigned short		header_len;	/* more space at head required */
 	unsigned short		trailer_len;	/* space to reserve at tail */
 
-	u32			metrics[RTAX_MAX];
-	struct dst_entry	*path;
-
-	unsigned long		rate_last;	/* rate limiting for ICMP */
 	unsigned int		rate_tokens;
+	unsigned long		rate_last;	/* rate limiting for ICMP */
 
-#ifdef CONFIG_NET_CLS_ROUTE
-	__u32			tclassid;
-#endif
+	struct dst_entry	*path;
 
 	struct neighbour	*neighbour;
 	struct hh_cache		*hh;
@@ -70,10 +65,20 @@ struct dst_entry
 	int			(*output)(struct sk_buff*);
 
 	struct  dst_ops	        *ops;
-		
-	unsigned long		lastuse;
+
+	u32			metrics[RTAX_MAX];
+
+#ifdef CONFIG_NET_CLS_ROUTE
+	__u32			tclassid;
+#endif
+
+	/*
+	 * __refcnt wants to be on a different cache line from
+	 * input/output/ops or performance tanks badly
+	 */
 	atomic_t		__refcnt;	/* client references	*/
 	int			__use;
+	unsigned long		lastuse;
 	union {
 		struct dst_entry *next;
 		struct rtable    *rt_next;