| 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
| ||
|
Message-ID: <202201281347.F36AEA5B61@keescook>
Date: Fri, 28 Jan 2022 14:30:12 -0800
From: Kees Cook <keescook@...omium.org>
To: Andrew Morton <akpm@...ux-foundation.org>,
Alexey Dobriyan <adobriyan@...il.com>
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
Subject: Re: [PATCH] elf: Relax assumptions about vaddr ordering
On Fri, Jan 28, 2022 at 09:26:09AM +0100, Magnus Groß wrote:
> On Wed, Jan 26, 2022 at 10:31:42PM -0800 Kees Cook 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?
> >
> > -Kees
> >
>
> May I also propose to include this patch in whatever mailing-list
> corresponds to the 5.16.x bugfix series?
> It turns out that almost all native Linux games published by the Virtual
> Programming company have this kind of weird PT_LOAD ordering including
> the famous Bioshock Infinite, so right now those games are all
> completely broken since Linux 5.16.
>
> P.S.: Someone should probably ask Virtual Programming, what kind of
> tooling they use to create such convoluted ELF binaries.
Oh, actually, this was independently fixed:
https://lore.kernel.org/all/YVmd7D0M6G/DcP4O@localhost.localdomain/
Alexey, you never answered by question about why we can't use a proper
type and leave the ELF_PAGESTART() macros alone:
https://lore.kernel.org/all/202110071038.B589687@keescook/
I still don't like the use of "int" in ELF_PAGESTART(), but I agree
it shouldn't cause a problem. I just really don't like mixing a signed
type with address calculations, from a robustness perspective.
Andrew, can you update elf-fix-overflow-in-total-mapping-size-calculation.patch
to include:
Fixes: 5f501d555653 ("binfmt_elf: reintroduce using MAP_FIXED_NOREPLACE")
Cc: stable@...r.kernel.org
Acked-by: Kees Cook <keescook@...omium.org>
Thanks!
-Kees
>
> > >
> > > -Kees
> > >
> > > > ---
> > > > fs/binfmt_elf.c | 18 ++++++++++++++----
> > > > 1 file changed, 14 insertions(+), 4 deletions(-)
> > > >
> > > > diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c
> > > > index f8c7f26f1fbb..0caaad9eddd1 100644
> > > > --- a/fs/binfmt_elf.c
> > > > +++ b/fs/binfmt_elf.c
> > > > @@ -402,19 +402,29 @@ static unsigned long elf_map(struct file *filep, unsigned long addr,
> > > > static unsigned long total_mapping_size(const struct elf_phdr *cmds, int nr)
> > > > {
> > > > int i, first_idx = -1, last_idx = -1;
> > > > + unsigned long min_vaddr = ULONG_MAX, max_vaddr = 0;
> > > >
> > > > for (i = 0; i < nr; i++) {
> > > > if (cmds[i].p_type == PT_LOAD) {
> > > > - last_idx = i;
> > > > - if (first_idx == -1)
> > > > + /*
> > > > + * The PT_LOAD segments are not necessarily ordered
> > > > + * by vaddr. Make sure that we get the segment with
> > > > + * minimum vaddr (maximum vaddr respectively)
> > > > + */
> > > > + if (cmds[i].p_vaddr <= min_vaddr) {
> > > > first_idx = i;
> > > > + min_vaddr = cmds[i].p_vaddr;
> > > > + }
> > > > + if (cmds[i].p_vaddr >= max_vaddr) {
> > > > + last_idx = i;
> > > > + max_vaddr = cmds[i].p_vaddr;
> > > > + }
> > > > }
> > > > }
> > > > if (first_idx == -1)
> > > > return 0;
> > > >
> > > > - return cmds[last_idx].p_vaddr + cmds[last_idx].p_memsz -
> > > > - ELF_PAGESTART(cmds[first_idx].p_vaddr);
> > > > + return max_vaddr + cmds[last_idx].p_memsz - ELF_PAGESTART(min_vaddr);
> > > > }
> > > >
> > > > static int elf_read(struct file *file, void *buf, size_t len, loff_t pos)
> > > > --
> > > > 2.34.1
> > >
> > > --
> > > Kees Cook
> >
> > --
> > Kees Cook
--
Kees Cook
Powered by blists - more mailing lists