<div dir="ltr"><div dir="ltr"><div>Yeah,I tried to join the IRC channel several times,but after some hours I left the channel because no one gave support. It seemed to me that the channel was full of bot and not humans. Anyway,I posted the question on different channels,such as ML and Reddit,but in all these cases,I didn't get some support,I should be honest. There is an update for this problem. I've got qemu from here :<code> fetch <a href="https://leaf.dragonflybsd.org/~aly/nvmm/qemu-6.0.0_1.txz" target="_blank">https://leaf.dragonflybsd.org/~aly/nvmm/qemu-6.0.0_1.txz</a></code></div><div><br></div><div>and I tried again :</div><div><br></div><div>/home/marietto/Desktop/Files/DFLY/qemu-6.0.0_1/usr/local/bin/./qemu-system-x86_64 \</div><span class="gmail-im">-machine type=q35,accel=nvmm \<br>-smp cpus=4 -m 8G \<br>-drive if=pflash,format=raw,readonly=on,file=/usr/local/share/uefi-edk2-qemu/QEMU_UEFI_CODE-x86_64.fd \<br>-drive if=pflash,format=raw,file=/usr/local/share/uefi-edk2-qemu/QEMU_UEFI_VARS-x86_64.fd \<br>-drive file=/mnt/dragonfly-ufs/bhyve/impish-cuda-11-4-nvidia-470.img,if=none,id=disk0 \<br>-device virtio-blk-pci,drive=disk0 \<br>-netdev user,id=net0,hostfwd=tcp:127.0.0.1:6022-:22 \<br>-device virtio-net-pci,netdev=net0 \<br>-object rng-random,id=rng0,filename=/dev/urandom \<br>-device virtio-rng-pci,rng=rng0 \<br><div>-display curses \</div><div><br></div></span><div>but I've got this output :</div><div><br></div><div>qemu-system-x86_64: NVMM: Mem Assist Failed [gpa=0xfffffff0]<br>qemu-system-x86_64: NVMM: Failed to execute a VCPU.<br>Abort trap (core dumped)<br>VGA Blank mode<br></div></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">Il giorno sab 1 gen 2022 alle ore 00:54 Aaron LI <<a href="mailto:aly@aaronly.me">aly@aaronly.me</a>> ha scritto:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="auto"><div dir="ltr">Hi Mario,</div><div dir="ltr"><br></div><div dir="ltr"><blockquote type="cite">On Dec 31, 2021, at 23:51, Mario Marietto <<a href="mailto:marietto2008@gmail.com" target="_blank">marietto2008@gmail.com</a>> wrote:<br></blockquote></div><blockquote type="cite"><div dir="ltr"><div dir="ltr"><p>Hello.</p>

<p>For sure I don't want to bother you,but after having posted this question in several places and no one replied (like IRC,the ML,or Reddit or the UnitedBSD forum),</p></div></div></blockquote>I didn’t seem to see this question on the #dragonflybsd IRC and this mailing list.<br><blockquote type="cite"><div dir="ltr"><div dir="ltr"><p>I've thought of trying this <span lang="en">unorthodox method,just because...I need support to understand the reasons for this error. I also think that by helping me,you are also helping the other DFLY lovers. Thanks for the</span><span lang="en"> understanding. Taking in consideration that NVMM has been ported from NetBSD and on the NetBSD this same error has been fixed,as you can see below :<br></span></p><p><span lang="en"><a href="http://mail-index.netbsd.org/pkgsrc-users/2021/09/05/msg034346.html" target="_blank">http://mail-index.netbsd.org/pkgsrc-users/2021/09/05/msg034346.html</a></span></p><p><span lang="en"> </span></p><p>it seems that the fix has not been ported to DFLY,so for this reason we can't use NVMM.<br></p></div></div></blockquote>Sure. I noticed such a fix to Qemu upstream some time ago.<div><br></div><div>But well, we can’t directly use the official unmodified version. The DPorts version should be used and there are several necessary patches (ideally we’d better push to upstream, but that takes time and efforts).</div><div><br></div><div>It looks to me you were using the Qemu build from upstream rather than installed via pkg(8). If yes, then that’s the issue, and you just switch to our version and should just work.</div><div><br></div><div>Check:</div><div><a href="https://www.dragonflybsd.org/docs/docs/howtos/nvmm/" target="_blank">https://www.dragonflybsd.org/docs/docs/howtos/nvmm/</a></div><div><br><blockquote type="cite"><div dir="ltr"><div dir="ltr"><p>So,this is what I did :<br></p><p>I'm trying to test qemu and nvmm on :</p>


<p><code>DragonFly marietto 6.1-DEVELOPMENT DragonFly 
v6.1.0.573.gfca8e8-DEVELOPMENT #0: Wed Dec 22 09:11:32 CET 2021 
marietto@marietto:/usr/obj/usr/src/sys/X86_64_GENERIC x86_64</code></p>

<p>first of all I added these users on the nvmm group :</p>

<pre><code>root<span>@marietto</span><span>:/home/marietto</span> <span># pw groupmod nvmm -m marietto</span>
root<span>@marietto</span><span>:/home/marietto</span> <span># pw groupmod nvmm -m root</span></code></pre>

<p>then,I've launched this vm :</p>


<pre><code>qemu-system-x86_64 \
-machine <span>type</span>=q35,accel=nvmm \
-smp cpus=<span>4</span> -m <span>8</span>G \
-drive <span>if</span>=pflash,<span>format</span>=raw,readonly=on,file=/usr/<span>local</span>/share/uefi-edk2-qemu/QEMU_UEFI_CODE-x86_64.fd \
-drive <span>if</span>=pflash,<span>format</span>=raw,file=/usr/<span>local</span>/share/uefi-edk2-qemu/QEMU_UEFI_VARS-x86_64.fd \
-drive file=/mnt/dragonfly-ufs/bhyve/impish-cuda<span>-11</span><span>-4</span>-nvidia<span>-470.</span>img,<span>if</span>=none,id=disk0 \
-device virtio-blk-pci,drive=disk0 \
-netdev user,id=net0,hostfwd=tcp:<span>127.0</span><span>.0</span><span>.1</span>:<span>6022</span>-:<span>22</span> \
-device virtio-net-pci,netdev=net0 \
-object rng-<span>random</span>,id=rng0,filename=/dev/urandom \
-device virtio-rng-pci,rng=rng0 \
-display curses \

WARNING: Image <span>format</span> was <span>not</span> specified <span>for</span> <span>'/mnt/dragonfly-ufs/bhyve/impish-cuda-11-4-nvidia-470.img'</span> <span>and</span> probing guessed raw. Automatically detecting the <span>format</span> is dangerous <span>for</span> raw images, <span>write</span> operations on block <span>0</span> will be restricted. Specify the <span>'raw'</span> <span>format</span> explicitly to <span>remove</span> the restrictions.</code></pre>
<p>unfortunately I get this error,that it seems a bug to me :</p>

<p><code>qemu-system-x86_64: invalid accelerator nvmm</code></p>


<p>just because nvmm works great :</p>


<pre><code><span>root@marietto:/home/marietto/Desktop</span> <span># nvmmctl identify</span>

<span>nvmm:</span> <span>Kernel</span> <span>API</span> <span>version</span> <span>3</span>
<span>nvmm:</span> <span>State</span> <span>size</span> <span>1008</span>
<span>nvmm:</span> <span>Comm</span> <span>size</span> <span>4096</span>
<span>nvmm:</span> <span>Max</span> <span>machines</span> <span>128</span>
<span>nvmm:</span> <span>Max</span> <span>VCPUs</span> <span>per</span> <span>machine</span> <span>128</span>
<span>nvmm:</span> <span>Max</span> <span>RAM</span> <span>per</span> <span>machine</span> <span>127T</span>
<span>nvmm:</span> <span>Arch</span> <span>Mach</span> <span>conf</span> <span>0</span>
<span>nvmm:</span> <span>Arch</span> <span>VCPU</span> <span>conf</span> <span>0x3</span><span><CPUID,TPR></span>
<span>nvmm:</span> <span>Guest</span> <span>FPU</span> <span>states</span> <span>0x3</span><span><x87,SSE></span></code></pre></div></div></blockquote><div>The NVMM part looks good.</div><div><br></div></div><div>Cheers,</div><div>Aaron </div></div></blockquote></div><br clear="all"><br>-- <br><div dir="ltr" class="gmail_signature">Mario.<br></div>