Subject: The yamt-x86pmap merge breaks running under Parallels.
To: None <current-users@netbsd.org>
From: Andrew Gillham <gillham@gmail.com>
List: current-users
Date: 10/24/2007 11:51:13
------=_Part_8113_19662456.1193251873347
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

Kernels from -current built after the merge hang like this with Intel VT-x
active:
...
cpu0 at mainbus0: (uniprocessor)
cpu0: Intel Celeron (Mendocino) (686-class), 2660.28 MHz, id 0x665
cpu0: "Intel(R) Xeon(R) CPU               5150  @ 2.66GHz"
<hang>

A working boot of the same kernel (with Intel VT-x disabled) looks like
this:
...
mainbus0 (root)
cpu0 at mainbus0: (uniprocessor)
cpu0: Intel Celeron (Mendocino) (686-class), 2662.25 MHz, id 0x665
cpu0: features f80b9b9<FPU,PSE,TSC,MSR,MCE,CX8,SEP,MTRR>
cpu0: features f80b9b9<PGE,CMOV,MMX>
cpu0: features f80b9b9<FXSR,SSE,SSE2,SS>
cpu0: features2 1<SSE3>
cpu0: "Intel(R) Xeon(R) CPU            5150  @ 2.66GHz"
cpu0: I-cache 32 KB 64B/line 8-way, D-cache 32 KB 64B/line 8-way
cpu0: L2 cache 4 MB 64B/line 16-way
cpu0: ITLB 128 4 KB entries 4-way
cpu0: 64 page colors
pci0 at mainbus0 bus 0: configuration mode 1
...

Disabling largepages seems to fix this issue and allow it to boot and run
properly.
I've tried the different OS options under Parallels but they seem to just
change what CPU is identified and have no effect on the hang.  I don't know
if there is an easy way to recognize Parallels' virtual CPU, but we need
that or a way to disable largepages.

This is on a Mac Pro with Parallels 2.5.  Can anyone else reproduce this?

-Andrew

------=_Part_8113_19662456.1193251873347
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

Kernels from -current built after the merge hang like this with Intel VT-x active:<br>...<br>cpu0 at mainbus0: (uniprocessor)<br>cpu0: Intel Celeron (Mendocino) (686-class), 2660.28 MHz, id 0x665<br>cpu0: &quot;Intel(R) Xeon(R) CPU&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 5150&nbsp; @ 
2.66GHz&quot;<br>&lt;hang&gt;<br><br>A working boot of the same kernel (with Intel VT-x disabled) looks like this:<br>...<br>mainbus0 (root)<br>cpu0 at mainbus0: (uniprocessor)<br>cpu0: Intel Celeron (Mendocino) (686-class), 
2662.25 MHz, id 0x665<br>cpu0: features f80b9b9&lt;FPU,PSE,TSC,MSR,MCE,CX8,SEP,MTRR&gt;<br>cpu0: features f80b9b9&lt;PGE,CMOV,MMX&gt;<br>cpu0: features f80b9b9&lt;FXSR,SSE,SSE2,SS&gt;<br>cpu0: features2 1&lt;SSE3&gt;<br>
cpu0: &quot;Intel(R) Xeon(R) CPU&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 5150&nbsp; @ 
2.66GHz&quot;<br>cpu0: I-cache 32 KB 64B/line 8-way, D-cache 32 KB 64B/line 8-way<br>cpu0: L2 cache 4 MB 64B/line 16-way<br>cpu0: ITLB 128 4 KB entries 4-way<br>cpu0: 64 page colors<br>pci0 at mainbus0 bus 0: configuration mode 1
<br>...<br><br>Disabling largepages seems to fix this issue and allow it to boot and run properly.<br>I&#39;ve tried the different OS options under Parallels but they seem to just change what CPU is identified and have no effect on the hang.&nbsp; I don&#39;t know if there is an easy way to recognize Parallels&#39; virtual CPU, but we need that or a way to disable largepages.
<br><br>This is on a Mac Pro with Parallels 2.5.&nbsp; Can anyone else reproduce this?<br><br>-Andrew

------=_Part_8113_19662456.1193251873347--