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 PHC | |
Open Source and information security mailing list archives
| ||
|
Date: Tue, 1 Feb 2022 14:44:48 -0800 From: Andrew Morton <akpm@...ux-foundation.org> To: Kees Cook <keescook@...omium.org> Cc: Magnus Groß <magnus.gross@...h-aachen.de>, Alexander Viro <viro@...iv.linux.org.uk>, Eric Biederman <ebiederm@...ssion.com>, linux-fsdevel@...r.kernel.org, linux-mm@...ck.org, linux-hardening@...r.kernel.org, Alexey Dobriyan <adobriyan@...il.com> Subject: Re: [PATCH] elf: Relax assumptions about vaddr ordering On Wed, 26 Jan 2022 22:31:42 -0800 Kees Cook <keescook@...omium.org> wrote: > On Wed, Jan 26, 2022 at 08:50:15AM -0800, Kees Cook wrote: > > On Wed, Jan 26, 2022 at 05:25:20PM +0100, Magnus Groß wrote: > > > From ff4dde97e82727727bda711f2367c05663498b24 Mon Sep 17 00:00:00 2001 > > > From: =?UTF-8?q?Magnus=20Gro=C3=9F?= <magnus.gross@...h-aachen.de> > > > Date: Wed, 26 Jan 2022 16:35:07 +0100 > > > Subject: [PATCH] elf: Relax assumptions about vaddr ordering > > > MIME-Version: 1.0 > > > Content-Type: text/plain; charset=UTF-8 > > > Content-Transfer-Encoding: 8bit > > > > > > Commit 5f501d555653 ("binfmt_elf: reintroduce using > > > MAP_FIXED_NOREPLACE") introduced a regression, where the kernel now > > > assumes that PT_LOAD segments are ordered by vaddr in load_elf_binary(). > > > > > > Specifically consider an ELF binary with the following PT_LOAD segments: > > > > > > Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align > > > LOAD 0x000000 0x08000000 0x08000000 0x474585 0x474585 R E 0x1000 > > > LOAD 0x475000 0x08475000 0x08475000 0x090a4 0xc6c10 RW 0x1000 > > > LOAD 0x47f000 0x00010000 0x00010000 0x00000 0x7ff0000 0x1000 > > > > > > Note how the last segment is actually the first segment and vice versa. > > > > > > Since total_mapping_size() only computes the difference between the > > > first and the last segment in the order that they appear, it will return > > > a size of 0 in this case, thus causing load_elf_binary() to fail, which > > > did not happen before that change. > > > > > > Strictly speaking total_mapping_size() made that assumption already > > > before that patch, but the issue did not appear because the old > > > load_addr_set guards never allowed this call to total_mapping_size(). > > > > > > Instead of fixing this by reverting to the old load_addr_set logic, we > > > fix this by comparing the correct first and last segments in > > > total_mapping_size(). > > > > Ah, nice. Yeah, this is good. > > > > > Signed-off-by: Magnus Groß <magnus.gross@...h-aachen.de> > > > > Fixes: 5f501d555653 ("binfmt_elf: reintroduce using MAP_FIXED_NOREPLACE") > > Cc: stable@...r.kernel.org > > Acked-by: Kees Cook <keescook@...omium.org> > > Andrew, can you pick this up too? > It conflicts significantly with Alexey's "ELF: fix overflow in total mapping size calculation".
Powered by blists - more mailing lists