netdev - Re: [PATCH net-next v2 1/4] indirect call wrappers: helpers to speed-up indirect calls of builtin

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: <0ffb8bc11613ea082179142f7c90830d4e419491.camel@redhat.com>
Date:   Tue, 11 Dec 2018 23:28:32 +0100
From:   Paolo Abeni <pabeni@...hat.com>
To:     David Woodhouse <dwmw2@...radead.org>, netdev@...r.kernel.org
Cc:     "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <eric.dumazet@...il.com>,
        Paul Turner <pjt@...gle.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH net-next v2 1/4] indirect call wrappers: helpers to
 speed-up indirect calls of builtin

Hi,

I'm sorry for the long delay, I was (and still I am) diverted by some
other duty.

On Fri, 2018-12-07 at 21:46 +0000, David Woodhouse wrote:
> On Fri, 2018-12-07 at 21:46 +0100, Paolo Abeni wrote:
> > > I wonder if we can declare the common case functions as 'weak' so that
> > > the link failures don't happen when they're absent.
> > 
> > I experimented a previous version with alias. I avoided weak alias
> > usage, because I [mis?]understood not all compilers have a complete
> > support for them (e.g. clang).
> > Also, with weak ref, a coding error that is now discovered at build
> > time will result in worse performance at runtime, likely with some
> > uncommon configuration, possibly not as easily detected. I'm unsure
> > that would be better ?!?
> 
> I think everything supports weak linkage; we've been using it for
> years.

Ok, I likely was confused by some old, non first-hand info.

Anyway weak alias will turn a compile time issue in a possible run-time 
(small) regression. I think the first option would be preferable.

> > I'm sorry, I don't follow here. I think static keys can't be used for
> > the reported network case: we have different list elements each
> > contaning a different function pointer and we access/use
> > different ptr on a per packet basis.
> 
> Yes, the alternatives would be used to change the "likely" case.
> 
> We still do the "if (fn == default_fn) default_fn(); else (*fn)();"
> part; or even the variant with two (or more) common cases. 
> 
> It's just that the value of 'default_fn' can be changed at runtime
> (with patching like alternatives/static keys, since of course it has to
> be a direct call).

Thanks for clarifying. If I understood correctly, you would like some
helper for:

if (static_branch_likely(&use_default_fn_a))
    INDIRECT_CALL_1(f, default_fn_a, <args>)
else if (static_branch_likely(&use_default_fn_b))
    INDIRECT_CALL_1(f, default_fn_b, <args>)
// ...

if so, I think we can eventually add support for this kind of stuff on
top of the proposed macros. WDYT?

Thanks,

Paolo