linux-kernel - Re: [V9fs-developer] [PATCH 2/2] net/9p: add a per-client fcall kmem

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]

Date:   Tue, 31 Jul 2018 03:35:56 +0200
From:   Dominique Martinet <asmadeus@...ewreck.org>
To:     piaojun <piaojun@...wei.com>
Cc:     v9fs-developer@...ts.sourceforge.net,
        linux-fsdevel@...r.kernel.org, Greg Kurz <groug@...d.org>,
        Matthew Wilcox <willy@...radead.org>,
        linux-kernel@...r.kernel.org
Subject: Re: [V9fs-developer] [PATCH 2/2] net/9p: add a per-client fcall
 kmem_cache

piaojun wrote on Tue, Jul 31, 2018:
> Could you help paste some test result before-and-after the patch applied?

The only performance tests I did were sent to the list a couple of mails
earlier, you can find it here:
http://lkml.kernel.org/r/20180730093101.GA7894@nautica

In particular, the results for benchmark on small writes just before and
after this patch, without KASAN (these are the same numbers as in the
link, hardware/setup is described there):
 - no alloc (4.18-rc7 request cache): 65.4k req/s
 - non-power of two alloc, no patch: 61.6k req/s
 - power of two alloc, no patch: 62.2k req/s
 - non-power of two alloc, with patch: 64.7k req/s
 - power of two alloc, with patch: 65.1k req/s

I'm rather happy with the result, I didn't expect using a dedicated
cache would bring this much back but it's certainly worth it.

> > @@ -1011,6 +1034,7 @@ void p9_client_destroy(struct p9_client *clnt)
> >  
> >  	p9_tag_cleanup(clnt);
> >  
> > +	kmem_cache_destroy(clnt->fcall_cache);
> 
> We could set NULL for fcall_cache in case of use-after-free.
> 
> >  	kfree(clnt);

Hmm, I understand where this comes from, but I'm not sure I agree.
If someone tries to access the client while/after it is freed things are
going to break anyway, I'd rather let things break as obviously as
possible than try to cover it up.

-- 
Dominique