linux-kernel - Re: [BUG] list corruption while enabling multi call uprobes via perf

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]

Message-Id: <20200110103647.f2182be4584886b9a62d6161@kernel.org>
Date:   Fri, 10 Jan 2020 10:36:47 +0900
From:   Masami Hiramatsu <mhiramat@...nel.org>
To:     Arnaldo Carvalho de Melo <acme@...nel.org>
Cc:     Jiri Olsa <jolsa@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...nel.org>,
        "Naveen N. Rao" <naveen.n.rao@...ux.ibm.com>,
        Anil S Keshavamurthy <anil.s.keshavamurthy@...el.com>,
        "David S. Miller" <davem@...emloft.net>,
        Namhyung Kim <namhyung@...nel.org>,
        Toke Høiland-Jørgensen 
        <thoiland@...hat.com>, Jean-Tsung Hsiao <jhsiao@...hat.com>,
        Jesper Dangaard Brouer <brouer@...hat.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [BUG] list corruption while enabling multi call uprobes via
 perf

On Thu, 9 Jan 2020 10:38:46 -0300
Arnaldo Carvalho de Melo <acme@...nel.org> wrote:

> Em Thu, Jan 09, 2020 at 06:33:56PM +0900, Masami Hiramatsu escreveu:
> > Hi,
> > 
> > On Thu, 9 Jan 2020 11:10:56 +0900
> > Masami Hiramatsu <mhiramat@...nel.org> wrote:
> > 
> > > Hmm, this seems that the event->hw.tp_list is not initialized when removing
> > > from the list in uprobe_perf_close().
> > 
> > Oops, that's wrong. Of course my patch can ease (avoid kernel panic) the
> > issue, but not fixing the root cause.
> > The root cause is that the uprobe event tries to open multiple probes with
> > one perf_event. So the perf_event is reused on different probes.
> > 
> > In the reported case, if we remove the multiple probe event before perf-stat,
> > no problem happens.
> > 
> > I'll try to fix it.
> 
> Ok!
> 
> For reference, I rebooted it with a fedora kernel, 5.3ish and it seems
> to work:

Yes, since it was my commit 60d53e2c3b75 ("tracing/probe: Split trace_event
related data from trace_probe") broke list operation...
I'll send a fix soon.

Thank you,

-- 
Masami Hiramatsu <mhiramat@...nel.org>