linux-kernel - Re: [PATCH v2] x86: Clear feature bits disabled at compile-time

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: <5pzffj2kde67oqgwpvw4j3lxd3fstuhgnosmhiyf5wcdr3je6i@juy3hfn4fiw7>
Date: Wed, 23 Jul 2025 13:46:44 +0200
From: Maciej Wieczor-Retman <maciej.wieczor-retman@...el.com>
To: Greg KH <gregkh@...uxfoundation.org>
CC: Thomas Gleixner <tglx@...utronix.de>, Ingo Molnar <mingo@...hat.com>,
	Borislav Petkov <bp@...en8.de>, Dave Hansen <dave.hansen@...ux.intel.com>,
	<x86@...nel.org>, "H. Peter Anvin" <hpa@...or.com>, "Kirill A. Shutemov"
	<kas@...nel.org>, Alexander Potapenko <glider@...gle.com>, "Peter Zijlstra
 (Intel)" <peterz@...radead.org>, Xin Li <xin3.li@...el.com>, Sai Praneeth
	<sai.praneeth.prakhya@...el.com>, Jethro Beekman <jethro@...tanix.com>,
	Jarkko Sakkinen <jarkko@...nel.org>, Sean Christopherson <seanjc@...gle.com>,
	Tony Luck <tony.luck@...el.com>, Fenghua Yu <fenghua.yu@...el.com>, "Mike
 Rapoport (IBM)" <rppt@...nel.org>, Kees Cook <kees@...nel.org>, "Rick
 Edgecombe" <rick.p.edgecombe@...el.com>, Yu-cheng Yu <yu-cheng.yu@...el.com>,
	<stable@...r.kernel.org>, Borislav Petkov <bp@...e.de>,
	<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2] x86: Clear feature bits disabled at compile-time

On 2025-07-23 at 11:45:22 +0200, Greg KH wrote:
>On Wed, Jul 23, 2025 at 11:22:49AM +0200, Maciej Wieczor-Retman wrote:
>> If some config options are disabled during compile time, they still are
>> enumerated in macros that use the x86_capability bitmask - cpu_has() or
>> this_cpu_has().
>> 
>> The features are also visible in /proc/cpuinfo even though they are not
>> enabled - which is contrary to what the documentation states about the
>> file. Examples of such feature flags are lam, fred, sgx, ibrs_enhanced,
>> split_lock_detect, user_shstk, avx_vnni and enqcmd.
>> 
>> Add a DISABLED_MASK() macro that returns 32 bit chunks of the disabled
>> feature bits bitmask.
>> 
>> Initialize the cpu_caps_cleared and cpu_caps_set arrays with the
>> contents of the disabled and required bitmasks respectively. Then let
>> apply_forced_caps() clear/set these feature bits in the x86_capability.
>> 
>> Fixes: 6449dcb0cac7 ("x86: CPUID and CR3/CR4 flags for Linear Address Masking")
>> Fixes: 51c158f7aacc ("x86/cpufeatures: Add the CPU feature bit for FRED")
>> Fixes: 706d51681d63 ("x86/speculation: Support Enhanced IBRS on future CPUs")
>> Fixes: e7b6385b01d8 ("x86/cpufeatures: Add Intel SGX hardware bits")
>> Fixes: 6650cdd9a8cc ("x86/split_lock: Enable split lock detection by kernel")
>> Fixes: 701fb66d576e ("x86/cpufeatures: Add CPU feature flags for shadow stacks")
>> Fixes: ff4f82816dff ("x86/cpufeatures: Enumerate ENQCMD and ENQCMDS instructions")
>
>That is fricken insane.
>
>You are saying to people who backport stuff:
>	This fixes a commit found in the following kernel releases:
>		6.4
>		6.9
>		3.16.68 4.4.180 4.9.137 4.14.81 4.18.19 4.19
>		5.11
>		5.7
>		6.6
>		5.10
>
>You didn't even sort this in any sane order, how was it generated?
>
>What in the world is anyone supposed to do with this?
>
>If you were sent a patch with this in it, what would you think?  What
>could you do with it?
>
>Please be reasonable and consider us overworked stable maintainers and
>give us a chance to get things right.  As it is, this just makes things
>worse...
>
>greg k-h

Sorry, I certainly didn't want to add you more work.

I noted down which features are present in the x86_capability bitmask while
they're not compiled into the kernel. Then I noted down which commits added
these feature flags. So I suppose the order is from least to most significant
feature bit, which now I realize doesn't help much in backporting, again sorry.

Would a more fitting Fixes: commit be the one that changed how the feature flags
are used? At some point docs started stating to have them set only when features
are COMPILED & HARDWARE-SUPPORTED.

-- 
Kind regards
Maciej Wieczór-Retman