linux-kernel - Re: [RFC PATCH v4 00/18] pkeys-based page table hardening

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: <5bad616f-8a62-46f4-9c0c-929f4ab1dde0@arm.com>
Date: Fri, 11 Apr 2025 14:37:28 +0200
From: Kevin Brodsky <kevin.brodsky@....com>
To: Ingo Molnar <mingo@...nel.org>
Cc: linux-mm@...ck.org, linux-kernel@...r.kernel.org,
 Andrew Morton <akpm@...ux-foundation.org>, Mark Brown <broonie@...nel.org>,
 Catalin Marinas <catalin.marinas@....com>,
 Dave Hansen <dave.hansen@...ux.intel.com>,
 David Hildenbrand <david@...hat.com>, Ira Weiny <ira.weiny@...el.com>,
 Jann Horn <jannh@...gle.com>, Jeff Xu <jeffxu@...omium.org>,
 Joey Gouly <joey.gouly@....com>, Kees Cook <kees@...nel.org>,
 Linus Walleij <linus.walleij@...aro.org>, Andy Lutomirski <luto@...nel.org>,
 Marc Zyngier <maz@...nel.org>, Peter Zijlstra <peterz@...radead.org>,
 Pierre Langlois <pierre.langlois@....com>,
 Quentin Perret <qperret@...gle.com>,
 Rick Edgecombe <rick.p.edgecombe@...el.com>,
 "Mike Rapoport (IBM)" <rppt@...nel.org>, Ryan Roberts
 <ryan.roberts@....com>, Thomas Gleixner <tglx@...utronix.de>,
 Will Deacon <will@...nel.org>, Matthew Wilcox <willy@...radead.org>,
 Qi Zheng <zhengqi.arch@...edance.com>, linux-arm-kernel@...ts.infradead.org,
 x86@...nel.org
Subject: Re: [RFC PATCH v4 00/18] pkeys-based page table hardening

On 11/04/2025 11:21, Ingo Molnar wrote:
> * Kevin Brodsky <kevin.brodsky@....com> wrote:
>
>> Performance
>> ===========
>>
>> Caveat: these numbers should be seen as a lower bound for the overhead
>> of a real POE-based protection. The hardware checks added by POE are
>> however not expected to incur significant extra overhead.
>>
>> +-------------------+----------------------------------+------------------+---------------+
>> | Benchmark         | Result Class                     | Without batching | With batching |
>> +===================+==================================+==================+===============+
>> | mmtests/kernbench | elsp-64                          |            0.20% |         0.20% |
>> |                   | syst-64                          |            1.62% |         0.63% |
>> |                   | user-64                          |           -0.04% |         0.05% |
>> +-------------------+----------------------------------+------------------+---------------+
>> | micromm/fork      | fork: p:1                        |      (R) 225.56% |        -0.07% |
>> |                   | fork: p:512                      |      (R) 254.32% |         0.73% |
>> +-------------------+----------------------------------+------------------+---------------+
>> | micromm/munmap    | munmap: p:1                      |       (R) 24.49% |         4.29% |
>> |                   | munmap: p:512                    |      (R) 161.47% |     (R) 6.06% |
>> +-------------------+----------------------------------+------------------+---------------+
>> | micromm/vmalloc   | fix_size_alloc_test: p:1, h:0    |       (R) 14.80% |    (R) 11.85% |
>> |                   | fix_size_alloc_test: p:4, h:0    |       (R) 38.42% |    (R) 10.47% |
>> |                   | fix_size_alloc_test: p:16, h:0   |       (R) 64.74% |     (R) 6.41% |
>> |                   | fix_size_alloc_test: p:64, h:0   |       (R) 79.98% |     (R) 3.24% |
>> |                   | fix_size_alloc_test: p:256, h:0  |       (R) 85.46% |     (R) 2.77% |
>> |                   | fix_size_alloc_test: p:16, h:1   |       (R) 47.89% |         3.10% |
>> |                   | fix_size_alloc_test: p:64, h:1   |       (R) 62.43% |         3.36% |
>> |                   | fix_size_alloc_test: p:256, h:1  |       (R) 64.30% |     (R) 2.68% |
>> |                   | random_size_alloc_test: p:1, h:0 |       (R) 74.94% |     (R) 3.13% |
>> |                   | vm_map_ram_test: p:1, h:0        |       (R) 30.53% |    (R) 26.20% |
>> +-------------------+----------------------------------+------------------+---------------+
> So I had to look 3 times to figure out what the numbers mean: they are 
> the extra overhead from this hardening feature, measured in system time 
> percentage, right?

These are relative increases compared to the baseline for this series
(described earlier on: 6.15-rc1 + 2 additional series). Real time is
measured, except for kernbench where all 3 measurements are provided.

> So "4.29%" means there's a 4.29% slowdown on that particular workload 
> when the feature is enabled. Maybe add an explanation to the next iteration? :-)

Yes that's right. I thought it was clear from the description above but
evidently I was wrong :) I'll add a "plain text" reading like this one
in the next version. I should also have mentioned which config was used,
namely: defconfig + CONFIG_KPKEYS_HARDENED_PGTABLES=y

- Kevin