Date: Tue, 02 Jan 2018 17:11:54 -0800 From: Cy Schubert <Cy.Schubert@komquats.com> To: Michael Butler <imb@protected-networks.net>, Cy Schubert <Cy.Schubert@komquats.com>, Warner Losh <imp@bsdimp.com> Cc: FreeBSD Current <freebsd-current@freebsd.org> Subject: Re: Intel CPU design flaw - FreeBSD affected? Message-ID: <227070B7-561C-420E-BE0B-F75DD3D4D6EE@cschubert.com> In-Reply-To: <5787e63c-9a88-c807-c132-572e8454a4d0@protected-networks.net> References: <20180103001949.6AF4F2C9@spqr.komquats.com> <5787e63c-9a88-c807-c132-572e8454a4d0@protected-networks.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On January 2, 2018 4:56:48 PM PST, Michael Butler <imb@protected-networks= =2Enet> wrote: >On 01/02/18 19:20, Cy Schubert wrote: >> This Linux commit gives us a hint=2E >>=20 >> https://lkml=2Eorg/lkml/2017/12//27/2 > >Sadly, the articles I've read to date make no mention of which Intel >silicon revs are vulnerable=2E However, the use of the PCID feature, >which >is only available on more recent CPUs, does seem to afford a slightly >lesser performance hit when completely splitting the kernel and user >address space mappings :-( > > imb Yes=2E You can see if your cpu supports pcid using cpuinfo from ports=2E Then loo= k at input line 0x07 in the hexdump=2E Bit 0x0a of rbx will indicate if in= vpcid is available=2E We simply need to manipulate a copy of cr3 and reload= it=2E My amd gear in my basement isn't affected but my laptop is=2E It supports = pcid but not invpcid=2E There will be a performance hit=2E Looking at the Linux patch, looks like = the workaround is optional=2E I would suspect not through a sysctl, that wo= uld be silly=2E --- Cy Schubert <Cy=2ESchubert@cschubeet=2Ecom> or <cy@freebsd=2Eorg> -- small keyboard in use, apologies for typos and autocorrect --
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?227070B7-561C-420E-BE0B-F75DD3D4D6EE>