netdev - [PATCH v2 0/3] bpf: Track MAP pointer alignment

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>] [day] [month] [year] [list]

Message-Id: <20170515.120422.1411526109824089456.davem@davemloft.net>
Date:   Mon, 15 May 2017 12:04:22 -0400 (EDT)
From:   David Miller <davem@...emloft.net>
To:     daniel@...earbox.net
CC:     ast@...com, alexei.starovoitov@...il.com, netdev@...r.kernel.org
Subject: [PATCH v2 0/3] bpf: Track MAP pointer alignment

This patch series updates the BPF verifier to properly track MAP
access alignment, just as we already do for packet pointers.

The two main elements of the implementation are putting the register
offset into a shared rather than reg type specific area of the
verifier register tracking strcutre, and also eliding packet pointer
specific checks when managing MAP pointer alignment.

More details of the implementation are in the commit message for patch
#2.

Patch #3 removes alignment restrictions from various test_verifier.c
tests which are no longer necessary with this MAP alignment
infrastructure in place.  Further improvements are possible, but
I didn't want to be overly ambitious here.

In the future we can add some MAP tests to test_align.c as well.

changes since v1:
 - Enforce offset limits just like for packet pointers
 - Handle "CONST_IMM += MAP_VALUE_ADJ"
 - Document effects on state pruning in commit message of patch #2

Signed-off-by: David S. Miller <davem@...emloft.net>