linux-kernel - RE: x86/mce: Potential Information Leak in __mce_read

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>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <SJ1PR11MB608343DA951D303A36B2A29EFCE92@SJ1PR11MB6083.namprd11.prod.outlook.com>
Date: Thu, 30 Jan 2025 17:13:28 +0000
From: "Luck, Tony" <tony.luck@...el.com>
To: Haoyu Li <lihaoyu499@...il.com>, Borislav Petkov <bp@...en8.de>, "Thomas
 Gleixner" <tglx@...utronix.de>, Ingo Molnar <mingo@...hat.com>, Dave Hansen
	<dave.hansen@...ux.intel.com>, "H. Peter Anvin" <hpa@...or.com>,
	"linux-edac@...r.kernel.org" <linux-edac@...r.kernel.org>
CC: "x86@...nel.org" <x86@...nel.org>, "linux-kernel@...r.kernel.org"
	<linux-kernel@...r.kernel.org>, "chenyuan0y@...il.com" <chenyuan0y@...il.com>
Subject: RE: x86/mce: Potential Information Leak in __mce_read_apei Due to
 Uninitialized Bytes

> I would like to bring to your attention a potential issue in the
> "__mce_read_apei" function. Specifically, the struct mce m is
> initialized via the "apei_read_mce" function before being copied to
> user space. However, upon inspecting the call trace (apei_read_mce ->
> erst_read_record -> erst_read), we found that not all fields of struct
> mce are properly initialized.

Thanks for the report.

I'm not seeing it.(in upstream v6.13 kernel).

__mce_read_apei() does declare "struct mce m" as local, so there is a concern about leakage.

Now apei_read_mce() does:

	memcpy(m, &rcd.mce, sizeof(*m));

so all bytes are overwritte. Digging down "rcd.mce" is also a stack local, so need to
look into erst_read_record() for how it is filled.

That happens with

	erst_read(record_id, record, buflen):

which will overwrite the whole of the "struct cper_mce_record rcd"
declared by apei_read_mce().

All of the error paths look to avoid the copy_to_user() in __mce_read_apei()


Did I miss something that you saw? Are you looking at an older version?

-Tony