Adding ‘--enable-openbsd-malloc’ configure-flag to Tor.

Adding ‘--enable-openbsd-malloc’ configure-flag to Tor.

[Tobias Geerinckx-Rice]

Guix,

On 30/01/17 23:44, Ludovic Courtès wrote:
> Re: [PATCH] gnu: tor: Add hardening configure-flags.
> 
> Applied, thanks.

Ah, just too late :-) Although this isn't hardening-related.

I'd like to also add the ‘--enable-openbsd-malloc’ configure flag to our 
Tor <https://www.torproject.org/docs/faq.html.en#RelayMemory>.

Since adding it, I've been able to serve more traffic more reliably on a 
memory-constrained relay — without any noticable increase in load. This 
might be different on a SoC, although I'd suspect RAM would be the 
limiting factor there too.

To anyone willing to give it a try & give their opinion on the matter: 
I'd be very grateful.

Kind regards,

T G-R

PS: Sorry, no patch, because of

-- 
Sent from a Web browser. Excuse/enjoy my brevity.

Re: Adding ‘--enable-openbsd-malloc’ configure-flag to Tor.

[ng0]

Tobias Geerinckx-Rice <me@tobias.gr> writes:

> Guix,
>
> On 30/01/17 23:44, Ludovic Courtès wrote:
>> Re: [PATCH] gnu: tor: Add hardening configure-flags.
>> 
>> Applied, thanks.
>
> Ah, just too late :-) Although this isn't hardening-related.
>
> I'd like to also add the ‘--enable-openbsd-malloc’ configure flag to our 
> Tor <https://www.torproject.org/docs/faq.html.en#RelayMemory>.
>
> Since adding it, I've been able to serve more traffic more reliably on a 
> memory-constrained relay — without any noticable increase in load. This 
> might be different on a SoC, although I'd suspect RAM would be the 
> limiting factor there too.
>
> To anyone willing to give it a try & give their opinion on the matter: 
> I'd be very grateful.
>
> Kind regards,
>
> T G-R
>
> PS: Sorry, no patch, because of
>
> -- 
> Sent from a Web browser. Excuse/enjoy my brevity.
>

The tradeoff is more CPU load. For tor-relays I know how annoying
this can be, but I'm not sure if less RAM, more CPU is a good
idea (or any improvement) for the client-only side.

I need to read more about this and comment again once I know
more.
-- 
ng0 -- https://www.inventati.org/patternsinthechaos/

Re: Adding ‘--enable-openbsd-malloc’ configure-flag to Tor.

[ng0]

ng0 <contact.ng0@cryptolab.net> writes:

> Tobias Geerinckx-Rice <me@tobias.gr> writes:
>
>> Guix,
>>
>> On 30/01/17 23:44, Ludovic Courtès wrote:
>>> Re: [PATCH] gnu: tor: Add hardening configure-flags.
>>> 
>>> Applied, thanks.
>>
>> Ah, just too late :-) Although this isn't hardening-related.
>>
>> I'd like to also add the ‘--enable-openbsd-malloc’ configure flag to our 
>> Tor <https://www.torproject.org/docs/faq.html.en#RelayMemory>.
>>
>> Since adding it, I've been able to serve more traffic more reliably on a 
>> memory-constrained relay — without any noticable increase in load. This 
>> might be different on a SoC, although I'd suspect RAM would be the 
>> limiting factor there too.
>>
>> To anyone willing to give it a try & give their opinion on the matter: 
>> I'd be very grateful.
>>
>> Kind regards,
>>
>> T G-R
>>
>> PS: Sorry, no patch, because of
>>
>> -- 
>> Sent from a Web browser. Excuse/enjoy my brevity.
>>
>
> The tradeoff is more CPU load. For tor-relays I know how annoying
> this can be, but I'm not sure if less RAM, more CPU is a good
> idea (or any improvement) for the client-only side.
>
> I need to read more about this and comment again once I know
> more.
> -- 
> ng0 -- https://www.inventati.org/patternsinthechaos/

I try to take the www documentation of torproject with a grain of
salt, they have seen some improvement but usually it's better to
read the sourcecode and commits.

In commit 6a17067d38679224e8f59dd486132400e6869dc6 Andrew made
--enable-openbsd-mallow no longer the default, you now have to opt-in
to it if I understand the short commit right. This was in June
2008. Some commit before that, and in 2012 was  the last openbsd
malloc related commit.
It also seems to depend on glibc, at least according to some
commit in 2008, tor 0.2.0 which removed the option entirely for a
while (https://marc.info/?l=tor-cvs&m=121946290600768&w=2).

commit 6a17067d38679224e8f59dd486132400e6869dc6
Author: Andrew Lewman <andrew@torproject.org>
Date:   Sat Jun 21 13:44:03 2008 +0000

    with all of these fancy memory fixes, openbsd malloc appears
    to not be
        needed any more.


    svn:r15390
-- 
ng0 -- https://www.inventati.org/patternsinthechaos/