<div dir="ltr"><br><div class="gmail_extra"><br><br><div class="gmail_quote">On Thu, Aug 21, 2014 at 10:11 PM, <span dir="ltr"><<a href="mailto:lst_hoe02@kwsoft.de" target="_blank">lst_hoe02@kwsoft.de</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
Zitat von Iosif Peterfi <<a href="mailto:iosif.peterfi@gmail.com" target="_blank">iosif.peterfi@gmail.com</a>>:<div><div class="h5"><br>
<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
On Thu, Aug 21, 2014 at 6:43 PM, <<a href="mailto:lst_hoe02@kwsoft.de" target="_blank">lst_hoe02@kwsoft.de</a>> wrote:<br>
<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
Zitat von Gilberto Nunes <<a href="mailto:gilberto.nunes32@gmail.com" target="_blank">gilberto.nunes32@gmail.com</a>>:<br>
<br>
<br>
<a href="http://en.wikipedia.org/wiki/X86_virtualization#I.2FO_MMU_" target="_blank">http://en.wikipedia.org/wiki/<u></u>X86_virtualization#I.2FO_MMU_</a><br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
virtualization_.28AMD-Vi_and_<u></u>VT-d.29<br>
<br>
2014-08-21 13:06 GMT-03:00 <<a href="mailto:lst_hoe02@kwsoft.de" target="_blank">lst_hoe02@kwsoft.de</a>>:<br>
<br>
Hello,<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
when inspecting the "dmesg" output on one of the PVE servers i got the<br>
follwoing:<br>
<br>
No AGP bridge found<br>
Node 0: aperture @ 0 size 32 MB<br>
Your BIOS doesn't leave a aperture memory hole<br>
Please enable the IOMMU option in the BIOS setup<br>
This costs you 64 MB of RAM<br>
Mapping aperture over 65536 KB of RAM @ 20000000<br>
PM: Registered nosave memory: 0000000020000000 - 0000000024000000<br>
AMD-Vi disabled by default: pass amd_iommu=on to enable<br>
<br>
My first action was to go to the BIOS (the machine is still test) and<br>
activated the IOMMU. The message stays the same. Next try was to set<br>
iommu=noaperture as boot parameter because it was suggested for machine<br>
having IOMMU but no AGP anymore. This lead to<br>
<br>
PCI-DMA: Using software bounce buffering for IO (SWIOTLB)<br>
Placing 64MB software IO TLB between ffff880020000000 - ffff880024000000<br>
software IO TLB at phys 0x20000000 - 0x24000000<br>
<br>
Which looks like the IOMMU emulated in software...<br>
<br>
This lead to the question what the real benefits of AMD-Vi / IOMMU are<br>
for<br>
Proxmox/KVM if they are disabled by default and can not be enabled<br>
straight<br>
forward as it looks like?<br>
<br>
The Board in question is a ASUS KCMA-D8, BIOS 3102 with two AMD Opteron<br>
4226 CPUs.<br>
<br>
<br>
</blockquote></blockquote>
Hm, yes. But this still lead to the question which boot parameter should<br>
be used if any:<br>
<br>
iommu=noaperture which looks like deaktivating GART IOMMU but also<br>
hardware IOMMU at all<br>
amd_iommu which should activate AMD-Vi which includes IOMMU<br>
no parameter which looks like using GART IOMMU but not AMD-Vi and only<br>
when IOMMU is enabled in BIOS<br>
<br>
or do i simply not need it at all if no passthrough of devices is required?<br>
<br>
me->confused<br>
<br>
Thanks<br>
<br>
<br>
Andreas<br>
<br>
<br>
______________________________<u></u>_________________<br>
pve-user mailing list<br>
<a href="mailto:pve-user@pve.proxmox.com" target="_blank">pve-user@pve.proxmox.com</a><br>
<a href="http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user" target="_blank">http://pve.proxmox.com/cgi-<u></u>bin/mailman/listinfo/pve-user</a><br>
<br>
</blockquote>
<br>
<br>
Hello,<br>
<br>
>From the original post I thought you were trying to activate AMD-Vi. What<br>
are you trying to accomplish ?<br>
<br>
Regards,<br>
Iosif<br>
</blockquote>
<br></div></div>
I try to find out if it is useful to activate the IOMMU and if yes, which kind of IOMMU and how. The default seams to use the GART (AGP?) based old style IOMMU. It would be interesting to hear if one could get a real benefit from using AMD-Vi instead which is the part which was invented for virtualization, no?<div class="">
<div class="h5"><br>
<br>
Thanks<br>
<br>
Andreas<br>
<br>
<br>
______________________________<u></u>_________________<br>
pve-user mailing list<br>
<a href="mailto:pve-user@pve.proxmox.com" target="_blank">pve-user@pve.proxmox.com</a><br>
<a href="http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user" target="_blank">http://pve.proxmox.com/cgi-<u></u>bin/mailman/listinfo/pve-user</a><br>
</div></div></blockquote></div><br><br></div><div class="gmail_extra">Hello Andreas,<br><br>Try passing amd_iommu=on to the kernel and test it yourself and tell us as well.<br><br></div><div class="gmail_extra">Regards,<br>
</div><div class="gmail_extra">Iosif<br></div></div>