oss-sec mailing list archives
Re: Intel hyper-threading security issues
From: Lukas Odzioba <lukas.odzioba () gmail com>
Date: Thu, 21 Jun 2018 14:51:51 +0200
2018-06-21 13:54 GMT+02:00 Sven Schwedas <sven.schwedas () tao at>:
Hotplug doesn't seem differentiate between HT threads and physical cores, will setting maxcpus=2 on a 2 cores+HT machine reliably disable HT, or can it disable one core and keep HT active on the other?
It depends on machine/BIOS. IIRC usually "non ht" logical cpus are enumerated in in SRAT before ht siblings (kernel goes over this table and boots cpus untill maxcpus is reached), so it will work fine, but I don't think it is standardized in any way. In /proc/cpuinfo among other things you have 1) Linux processor number, 2) apicid 3) core id and the task is to disable minimal number of cpus to make core id unique among those which are online. Of course it doesn't disable HT in any way, just removes certain logical cpus from scheduler. Thanks, Lukas
Current thread:
- Re: Intel hyper-threading security issues, (continued)
- Re: Intel hyper-threading security issues Georgi Guninski (Jun 20)
- Re: Intel hyper-threading security issues Stuart Henderson (Jun 21)
- Re: Intel hyper-threading security issues Lukas Odzioba (Jun 21)
- Re: Intel hyper-threading security issues Lukas Odzioba (Jun 21)
- Re: Intel hyper-threading security issues Sven Schwedas (Jun 21)
- Re: Intel hyper-threading security issues Solar Designer (Jun 21)
- Re: Intel hyper-threading security issues Michael Ellerman (Jun 22)
- Re: Intel hyper-threading security issues Solar Designer (Jun 22)
- Re: Intel hyper-threading security issues Seth Arnold (Jun 22)
- Re: Intel hyper-threading security issues Peter Kjellström (Jun 23)
- Re: Intel hyper-threading security issues Stuart Henderson (Jun 21)
- Re: Intel hyper-threading security issues Lukas Odzioba (Jun 21)
- Re: Intel hyper-threading security issues Georgi Guninski (Jun 20)
- Re: Intel hyper-threading security issues Gordon Tetlow (Jun 21)