Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20261119140102.GZ5716@brightrain.aerifal.cx>
Date: Thu, 19 Nov 2026 09:01:02 -0500
From: Rich Felker <dalias@...c.org>
To: zhaotq@....edu.cn
Cc: musl@...ts.openwall.com
Subject: Re: musl libc 1.2.5: PT_TLS p_align=0 causes TLS layout crash

On Thu, Nov 19, 2026 at 01:24:43PM +0800, zhaotq@....edu.cn wrote:
> Hello,
> 
> We found an availability bug in musl libc's static TLS initialization when an
> ELF executable has a PT_TLS program header with p_align equal to zero.
> 
> The ELF gABI defines p_align values 0 and 1 as equivalent "no alignment
> requirement" encodings. This rule comes from ELF gABI; POSIX does not directly
> specify this program-header encoding.
> 
> The PoC compiles a normal static musl program, changes the PT_TLS p_align
> field to zero, and executes it again. The unmodified binary
> runs normally; the modified binary crashes before main().

It is outside the scope of the dynamic linker to validate anything
about the programs it's running. If you are running a program/loading
a library, it already has full access to execute code in its execution
context.

If there is actually tooling that emits PT_TLS with p_align==0, then
this is an incompatibility with binaries produced with such tooling.
But I don't think there is any such tooling.

> We would appreciate your assessment of the security impact and whether
> a CVE is warranted. 

Absolutely not. "We can edit a binary and make it misbehave when run"
is not a security issue whatsoever. You can also just put system("rm
-rf /") in the binary. That is not a vuln.

Rich

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.