> On Jan. 6, 2015, 11:21 a.m., Steve Reinhardt wrote:
> > Can you be more specific about what doesn't work?  Do we really need to 
> > back out all of the enabled features?
> > 
> > Also, it would be nice to replace the comments with different comments, 
> > rather than just getting rid of them.  If you want to keep this as a pure 
> > backout of the previous commit and make the other changes separately, that 
> > would be fine.
> 
> Gabe Black wrote:
>     I can't point to exactly what doesn't work, but software that used to run 
> no longer runs with this CL. Another user reported similar issues, although I 
> don't know what they were running. Essentially we're telling the simulated 
> software it can use features that aren't actually implemented (or don't make 
> sense, like SVM on an intel system). If the host CPU doesn't get upset, the 
> simulated software might try actually using those features, and then 
> something bad will likely happen.
>     
>     Having comments are a good thing, but those comments weren't great. Those 
> CPUID results hold more than just feature bits, first of all. Second, it's 
> not well defined to say, effectively, that everything is enabled except such 
> and such. The set those exceptions are from will change over time, making the 
> comment invalid. It's also more relevant what features enabled than which are 
> disabled.
>     
>     My intention is to revert the original change here, and then to put up a 
> new version which does what the original was trying to do. I have that CL 
> posted over here http://reviews.gem5.org/r/2591/ which you reviewed. The 
> features that needed to be enabled to get things working for KVM in SE mode 
> (why those I have no idea) were a small subset of those enabled by this 
> change. That was a while ago, but I think I also turned on one or two other 
> things which we support but weren't telling software about.

Ah, I see... I had the order of these two patches reversed in my head, hence my 
confusion.  Makes a lot more sense now.  I'll just say that adding some 
comments to 2591 similar to the ones you're getting rid of here would be nice.


- Steve


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/2590/#review5739
-----------------------------------------------------------


On Dec. 22, 2014, 4:37 p.m., Gabe Black wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/2590/
> -----------------------------------------------------------
> 
> (Updated Dec. 22, 2014, 4:37 p.m.)
> 
> 
> Review request for Default.
> 
> 
> Repository: gem5
> 
> 
> Description
> -------
> 
> Changeset 10607:334a1f4026ca
> ---------------------------
> cpuid, x86: Revert "Enabling more features in CPUid"
> 
> That change enables CPUID bits for features that aren't implemented in gem5.
> If a simulated system tries to use those features because it was told it
> could, bad things can happen.
> 
> 
> Diffs
> -----
> 
>   src/arch/x86/cpuid.cc a0cb57e1c072965dcdd51465beff37b264b41424 
> 
> Diff: http://reviews.gem5.org/r/2590/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Gabe Black
> 
>

_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to