From: Edward Cree <ecree@solarflare.com>
To: Alexei Starovoitov <ast@fb.com>, <davem@davemloft.net>,
Alexei Starovoitov <alexei.starovoitov@gmail.com>,
Daniel Borkmann <daniel@iogearbox.net>
Cc: <netdev@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
iovisor-dev <iovisor-dev@lists.iovisor.org>
Subject: Re: [PATCH v3 net-next] bpf/verifier: track liveness for pruning
Date: Mon, 21 Aug 2017 19:36:44 +0100 [thread overview]
Message-ID: <b39b4a6f-907e-c684-b9e4-285ff9c7ca9a@solarflare.com> (raw)
In-Reply-To: <5d4e12aa-6861-a176-a8cf-a766bbca0a7a@fb.com>
On 19/08/17 00:37, Alexei Starovoitov wrote:
> that '14: safe' above is not correct.
>
> Disabling liveness as:
> @@ -3282,7 +3288,7 @@ static bool regsafe(struct bpf_reg_state *rold,
> struct bpf_reg_state *rcur,
> bool varlen_map_access, struct idpair *idmap)
> {
> - if (!(rold->live & REG_LIVE_READ))
> + if (0 && !(rold->live & REG_LIVE_READ))
>
> makes the test work properly and proper verifier output is:
> from 9 to 11: R0=map_value(id=0,off=0,ks=8,vs=48,imm=0) R1=inv(id=0,smax_value=10) R2=inv11 R10=fp0
> 11: (64) (u32) r1 <<= (u32) 2
> 12: (0f) r0 += r1
> 13: (05) goto pc+0
> 14: (7a) *(u64 *)(r0 +0) = 4
>
> R0=map_value(id=0,off=0,ks=8,vs=48,umax_value=17179869180,var_off=(0x0; 0x3fffffffc)) R1=inv(id=0,umax_value=17179869180,var_off=(0x0; 0x3fffffffc)) R2=inv11 R10=fp0
> R0 unbounded memory access, make sure to bounds check any array access into a map
>
> I don't yet understand the underlying reason. R0 should have been
> marked as LIVE_READ by ST_MEM...
> Please help debug it further.
>
Having added a bunch of debugging, I found out that indeed R0 _had_ been
marked as LIVE_READ. The problem was that env->varlen_map_value_access
wasn't set, because the access was at a constant offset (imm=0), but then
when we compare register states we just say "oh yeah, it's a map_value,
we don't need to look at the var_off".
This probably results from my unifying PTR_TO_MAP_VALUE with
PTR_TO_MAP_VALUE_ADJ; before that the old and new R0 would have different
reg->type so wouldn't match.
I'm tempted to just rip out env->varlen_map_value_access and always check
the whole thing, because honestly I don't know what it was meant to do
originally or how it can ever do any useful pruning. While drastic, it
does cause your test case to pass.
I'm not quite sure why your test passed when you disabled liveness, though;
that I can't explain.
-Ed
next prev parent reply other threads:[~2017-08-21 18:36 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-08-15 19:34 Edward Cree
2017-08-15 22:12 ` Daniel Borkmann
2017-08-15 23:32 ` David Miller
2017-08-18 3:21 ` Alexei Starovoitov
2017-08-18 14:16 ` Edward Cree
2017-08-18 23:37 ` Alexei Starovoitov
2017-08-21 18:36 ` Edward Cree [this message]
2017-08-21 20:27 ` Daniel Borkmann
2017-08-21 20:44 ` Edward Cree
2017-08-21 21:00 ` Daniel Borkmann
2017-08-21 21:23 ` Alexei Starovoitov
2017-08-21 20:24 ` Edward Cree
2017-08-21 21:18 ` Alexei Starovoitov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=b39b4a6f-907e-c684-b9e4-285ff9c7ca9a@solarflare.com \
--to=ecree@solarflare.com \
--cc=alexei.starovoitov@gmail.com \
--cc=ast@fb.com \
--cc=daniel@iogearbox.net \
--cc=davem@davemloft.net \
--cc=iovisor-dev@lists.iovisor.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®