[vbox-dev] Null pointer access in VMEmt.cpp by vmR3EmulationThreadWithId

Knut St. Osmundsen knut.osmundsen at oracle.com
Fri Mar 17 15:00:11 UTC 2017

Hi Alexander,

thanks for identifying a race.  Looking at the code, I can only imagine
this happening if the other EMTs (not 0) are late into the loop due to
busy host / scheduling fun.  The only alternative I can see would be
some wait error in vmR3BootstrapWait and causing it to return.

I've committed a different fix than what you suggest in your patch,
VMEmt.cpp should check that both pVM and pVCpu are sane before selecting
the non-bootstrap code path.  The fix should appear in the public SVN soon.


More information about the vbox-dev mailing list