netdev - RE: [PATCH net-next] strparser: Return if socket does not have required number of bytes

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]

Date:   Thu, 31 Jan 2019 06:25:26 +0000
From:   Vakul Garg <vakul.garg@....com>
To:     David Miller <davem@...emloft.net>
CC:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "borisp@...lanox.com" <borisp@...lanox.com>,
        "aviadye@...lanox.com" <aviadye@...lanox.com>,
        "davejwatson@...com" <davejwatson@...com>,
        "doronrk@...com" <doronrk@...com>
Subject: RE: [PATCH net-next] strparser: Return if socket does not have
 required number of bytes



> -----Original Message-----
> From: David Miller <davem@...emloft.net>
> Sent: Thursday, January 31, 2019 11:30 AM
> To: Vakul Garg <vakul.garg@....com>
> Cc: netdev@...r.kernel.org; borisp@...lanox.com;
> aviadye@...lanox.com; davejwatson@...com; doronrk@...com
> Subject: Re: [PATCH net-next] strparser: Return if socket does not have
> required number of bytes
> 
> From: Vakul Garg <vakul.garg@....com>
> Date: Wed, 30 Jan 2019 07:31:44 +0000
> 
> > Function strp_data_ready() should peek the associated socket to check
> > whether it has the required number of bytes available before queueing
> > work or initiating socket read via strp_read_sock(). This saves cpu
> > cycles because strp_read_sock() is called only when required amount of
> > data is available.
> >
> > Signed-off-by: Vakul Garg <vakul.garg@....com>
> 
> You can't do this, I think.  It's racy and the user socket owned check is
> absolutely necessary before you do the need bytes check.

Do you mean to say that 'peek_len' operation can't be invoked if the socket is already owned by some other context?

My understanding by reading following link is that we can to peek_len operation without acquiring sock lock.

https://elixir.bootlin.com/linux/v5.0-rc4/source/include/linux/net.h#L191

Can you please kindly elaborate the problem?