netdev - Re: [PATCHv5 iproute2 net-next 2/2] lib/libnetlink: re malloc buff if size is not enough

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

Message-ID: <20171026183147.GY32305@orbyte.nwl.cc>
Date:   Thu, 26 Oct 2017 20:31:47 +0200
From:   Phil Sutter <phil@....cc>
To:     David Ahern <dsahern@...il.com>
Cc:     Stephen Hemminger <stephen@...workplumber.org>,
        Hangbin Liu <liuhangbin@...il.com>, netdev@...r.kernel.org,
        Michal Kubecek <mkubecek@...e.cz>
Subject: Re: [PATCHv5 iproute2 net-next 2/2] lib/libnetlink: re malloc buff
 if size is not enough

On Thu, Oct 26, 2017 at 09:42:46AM -0600, David Ahern wrote:
> On 10/26/17 9:33 AM, Phil Sutter wrote:
> > On Thu, Oct 26, 2017 at 09:28:00AM -0600, David Ahern wrote:
> >> On 10/26/17 4:24 AM, Stephen Hemminger wrote:
> >>>>
> >>>> The kernel needs a flag that says "give me the message of the buffer is
> >>>> large enough; if not just PEEK and tell me the length." That would avoid
> >>>> the double call in most cases.
> >>>
> >>> Actually this has little impact because old code was doing implicit zero
> >>> of whole buffer, new code does not.
> >>>
> >>
> >> The patch calls recvmsg twice; libnl does the same thing. It would be
> >> better performance wise to have a flag that allows retrieval of the
> >> message if the supplied buffer is large enough and PEEK semantics if
> >> not. It was really a comment on how we could do better with proper
> >> kernel support.
> > 
> > Doesn't MSG_TRUNC without MSG_PEEK do just that?
> 
> MSG_TRUNC returns the actual message length if it is greater than the
> buffer. The message was dequeued and what could be copied into the
> supplied buffer is copied, but that means the returned message is truncated.

Ah, so one would have to resend the request then. Stupid me. :)

Thanks, Phil