netdev - Re: [EXTERNAL] Re: [PATCH net-next v3 3/3] net: ti: icssg-prueth: Add XDP support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <5f716740-ac45-4881-a27d-91a93de6f8c7@stanley.mountain>
Date: Wed, 5 Mar 2025 12:31:24 +0300
From: Dan Carpenter <dan.carpenter@...aro.org>
To: "Malladi, Meghana" <m-malladi@...com>
Cc: rogerq@...nel.org, danishanwar@...com, pabeni@...hat.com,
	kuba@...nel.org, edumazet@...gle.com, davem@...emloft.net,
	andrew+netdev@...n.ch, bpf@...r.kernel.org,
	linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
	netdev@...r.kernel.org, u.kleine-koenig@...libre.com,
	matthias.schiffer@...tq-group.com, schnelle@...ux.ibm.com,
	diogo.ivo@...mens.com, glaroque@...libre.com, macro@...am.me.uk,
	john.fastabend@...il.com, hawk@...nel.org, daniel@...earbox.net,
	ast@...nel.org, srk@...com, Vignesh Raghavendra <vigneshr@...com>
Subject: Re: [EXTERNAL] Re: [PATCH net-next v3 3/3] net: ti: icssg-prueth:
 Add XDP support

On Wed, Mar 05, 2025 at 02:53:07PM +0530, Malladi, Meghana wrote:
> Hi Dan,
> 
> On 3/3/2025 7:38 PM, Dan Carpenter wrote:
> > What I mean is just compile the .o file with and without the unlikely().
> > $ md5sum drivers/net/ethernet/ti/icssg/icssg_common. o*
> > 2de875935222b9ecd8483e61848c4fc9 drivers/net/ethernet/ti/icssg/
> > icssg_common. o. annotation 2de875935222b9ecd8483e61848c4fc9
> > ZjQcmQRYFpfptBannerStart
> > This message was sent from outside of Texas Instruments.
> > Do not click links or open attachments unless you recognize the source
> > of this email and know the content is safe.
> > Report Suspicious
> > <https://us-phishalarm-ewt.proofpoint.com/EWT/v1/G3vK!
> > uldq3TevVoc7KuXEXHnDf- TXtuZ0bON9iO0jTE7PyIS1jjfs_CzpvIiMi93PVt0MVDzjHGQSK__vY_-6rO7q86rFmBMGW4SSqK5pvNE$>
> > ZjQcmQRYFpfptBannerEnd
> > 
> > What I mean is just compile the .o file with and without the unlikely().
> > 
> > $ md5sum drivers/net/ethernet/ti/icssg/icssg_common.o*
> > 2de875935222b9ecd8483e61848c4fc9  drivers/net/ethernet/ti/icssg/icssg_common.o.annotation
> > 2de875935222b9ecd8483e61848c4fc9  drivers/net/ethernet/ti/icssg/icssg_common.o.no_anotation
> > 
> > Generally the rule is that you should leave likely/unlikely() annotations
> > out unless it's going to make a difference on a benchmark.  I'm not going
> > to jump down people's throat about this, and if you want to leave it,
> > it's fine.  But it just struct me as weird so that's why I commented on
> > it.
> > 
> 
> I have done some performance tests to see if unlikely() is gonna make any
> impact and I see around ~9000 pps and 6Mbps drop without unlikely() for
> small packet sizes (60 Bytes)
> 
> You can see summary of the tests here:
> 
> packet size   with unlikely(pps)  without unlikely(pps)   regression
> 
>       60        462377                453251                 9126
> 
>       80        403020                399372                 3648
> 
>       96        402059                396881                 5178
> 
>      120        392725                391312                 4413
> 
>      140        327706                327099                 607
> 
> packet size  with unlikely(Mbps)  without unlikely(Mbps)  regression
> 
>      60         311                   305                    6
> 
>      80         335                   332                    3
> 
>      96         386                   381                    5
> 
>      120        456                   451                    5
> 
>      140        430                   429                    1
> 
> For more details on the logs, please refer:https://gist.github.com/MeghanaMalladiTI/cc6cc7709791376cb486eb1222de67be
> 

Huh.  That's very interesting.  Fine, then.

regards,
dan carpenter