Page MenuHomeFreeBSD

Allow aesni(4) module to be loaded on all VMs
AbandonedPublic

Authored by eri on Jul 4 2015, 9:41 AM.
Tags
None
Referenced Files
F82268874: D2990.diff
Sat, Apr 27, 3:46 AM
Unknown Object (File)
Thu, Apr 25, 2:55 PM
Unknown Object (File)
Wed, Apr 17, 7:55 PM
Unknown Object (File)
Dec 22 2023, 9:11 PM
Unknown Object (File)
Nov 10 2023, 5:54 AM
Unknown Object (File)
Nov 6 2023, 5:46 PM
Unknown Object (File)
Oct 31 2023, 12:38 PM
Unknown Object (File)
Oct 9 2023, 4:50 AM
Subscribers

Details

Reviewers
jmg
gnn
Summary

Some VMs advertise the support of SSE instruction in a different set of the cpuid flags.
For more read https://forum.pfsense.org/index.php?topic=87314.0

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Lint Skipped
Unit
Tests Skipped

Event Timeline

eri retitled this revision from to Allow aesni(4) module to be loaded on all VMs.
eri updated this object.
eri edited the test plan for this revision. (Show Details)
eri added reviewers: jmg, gnn.
eri set the repository for this revision to rS FreeBSD src repository - subversion.
jmg requested changes to this revision.Jul 5 2015, 7:18 AM
jmg edited edge metadata.
jmg added inline comments.
sys/crypto/aesni/aesni.c
89

you're checking for CPUID_CLFSH not CPUID2_SSE41 on cpu_feature, if you're doing that spell it correctly.

According to my reading, CLFSH just denotes the presence of SSE2, not the required SSE4 instructions that are now required by GCM... If the need for supporting older, non-SSE4 is desired, then an audit needs to be done, and only enable those algorithms that do not require SSE4.

I did an audit of what machines are out there, and the second thread linked, all AES-NI machines have SSE4 support, if this is a bug in ESX/Xen/etc, then we need to special case it only for the respective cases.

Please provide a dmesg output showing the features bits requiring this change. I currently don't see a need for this change.

This revision now requires changes to proceed.Jul 5 2015, 7:18 AM