linux-kernel - Re: [PATCH v5 4/6] x86/microcode/AMD: Check microcode container data in the late loader

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] [thread-next>] [day] [month] [year] [list]

Message-ID: <756e5bd8-6cd8-5f28-6cab-c60396dc5de4@maciej.szmigiero.name>
Date:   Fri, 4 May 2018 01:26:50 +0200
From:   "Maciej S. Szmigiero" <mail@...iej.szmigiero.name>
To:     Borislav Petkov <bp@...en8.de>
Cc:     Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        "H. Peter Anvin" <hpa@...or.com>, x86@...nel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 4/6] x86/microcode/AMD: Check microcode container data
 in the late loader

On 03.05.2018 12:01, Borislav Petkov wrote:
> On Wed, May 02, 2018 at 02:47:39AM +0200, Maciej S. Szmigiero wrote:
>> On 01.05.2018 22:03, Borislav Petkov wrote:
>>> On Tue, May 01, 2018 at 06:19:56PM +0200, Maciej S. Szmigiero wrote:
>>>> -EINVAL cast to unsigned int is 4294967274 and this value is also
>>>> a valid count of bytes to skip that this function can return.
>>>
>>> And where exactly in the *old* code do we do that?
>>
>> The old code returned this value as a signed int, but then any
>> "patch_size" value (which is u32) above INT_MAX read from a section header
>> wrapped around to a negative pseudo-error code (which likely didn't match
>> any actual error number).
> 
> Lemme repeat my question: *where* *exactly* in the old code do we do that?
> 
> Feel free to paste snippets to show what you mean.
> 

>From verify_and_add_patch():
> static int verify_and_add_patch(u8 family, u8 *fw, unsigned int leftover)
> {
> 	struct microcode_header_amd *mc_hdr;
> 	struct ucode_patch *patch;
> 	unsigned int patch_size, crnt_size, ret;
> 	u32 proc_fam;
> 	u16 proc_id;
> 
> 	patch_size  = *(u32 *)(fw + 4);

Here we read a u32 (= unsigned int) value from a section header
and store it into an unsigned int variable.

> 	crnt_size   = patch_size + SECTION_HDR_SIZE;

Here we add 8 (SECTION_HDR_SIZE) to this value and once again store it
into an unsigned int variable.

> 	mc_hdr	    = (struct microcode_header_amd *)(fw + SECTION_HDR_SIZE);
> 	proc_id	    = mc_hdr->processor_rev_id;
> 
> 	proc_fam = find_cpu_family_by_equiv_cpu(proc_id);
> 	if (!proc_fam) {
> 		pr_err("No patch family for equiv ID: 0x%04x\n", proc_id);
> 		return crnt_size;

Here we return this variable, implicitly converting it into a
(signed) int.
Any value above INT_MAX will wrap around to a negative pseudo-error
code (which might not match any actual error number).

Maciej