Help to submit a fix or patch

Help
VBWizard
2013-02-18
2013-02-19
  • VBWizard
    VBWizard
    2013-02-18

    Hi, long time user, never posted before. I have been working lately with the SVM code, and ran into a problem. The interrupt number was not in VMCB EXITINFO1 on an SVM EXIT triggered by an INTn (software interrupt). I found the root of the problem and would like to submit a fix request, or for that matter I wouldn't mind writing a patch, I've just never done so before. Please let me know what channel to use request a fix, or what tool(s) to use to create the patch. (yes, I am a coder so you don't have to dumb it down)

    I am working on another SVM issue related to stepping through VMRUN vs running with no breaking, but more about that later. (i.e. I have hacked a fix but haven't found the "real" fix yet)

    Please let me know the best route to proceed. Thanks!

     
  • Hello,

    The patches should created using UNIX diff utility (diff -u folder1 folder2) and posed into SF patches tracker or just directly into Bochs-Developers mailing list as attachment to message.

    All patches are welcome. But especially in the case you explain -
    Acoording to AMD's docs in SVM IntN instruction should not report vector number in EXITINFO1. It reports the vector only if decode assist is supported by the processor. The first processor supporting the decode assists (AFAIK) was Zambezi - Bulldozer and you at least will have to enable it in the configuration to get it in. But for now it won't be enough as well - the decode assist is not fully implemented in Bochs yet and therefore disabled even in Zambezi configuration.

    Stanislav

     
  • VBWizard
    VBWizard
    2013-02-19

    Stanislav,

    Thank you for taking the time to respond. You know, when I was reading the AMD docs I didn't even notice that ExitInfo1 for IntN Intercepts was a sub-topic of Decode Assists. I am very obviously not the expert here, and I apologize for assuming it was a miss/bug.

    So unless I want to add some code to tie my reporting of the INT # in EXITINFO1 to a specific CPU configuration setting (i.e. 'decode assists supported') , I guess I should remove my "patch" and simply interrogate the instruction code bytes when an interrupt occurs, instead of counting on EXITINFO1.

    Also, is this discussion board/forum the appropriate channel for discussing these types of issues? There is another more complicated situation that I have seen, and although I am now doubtful that it is due to a bug, confirmation from people such as yourself would be a great help and possibly save me days of toiling over something that isn't even a bug, but something I'm doing wrong.

    If this isn't the correct channel, would the mailing list be? Don't worry, before I use any channel to discuss this other situation, I will go back and look at it again from a "maybe its me doing something wrong" perspective.

    Thanks again and let me know if there is something I can do to pay back the time you are spending on me.

    Chris

     
    Last edit: VBWizard 2013-02-19
    • ERROR! The markdown supplied could not be parsed correctly. Did you forget to surround a code snippet with "~~~~"?

      Hi,
      
      I really glad that somebody finally came to play with the code I've done
      and report if there are issues or problems.
      The SVM code in Bochs is not very much validated (I only booted Linux in
      the SVM enabled VirtualBox under Bochs, nothing more).
      
      The better channel to discuss it would be the Bochs-Developers mailing list.
      
      Thanks,
      Stanislav
      
      
      On Tue, Feb 19, 2013 at 2:55 PM, VBWizard <vbwizard1969@users.sf.net> wrote:
      
      > Stanislav,
      >
      > Thank you for taking the time to respond.  You know, when I was reading
      > the AMD docs I didn't even notice that ExitInfo1 for IntN Intercepts was a
      > sub-topic of Decode Assists.  I am very obviously not the expert here, and
      > I apologize for assuming it was a miss/bug.
      >
      > So unless I want to add some code to tie my reporting of the INT # in
      > EXITINFO1 to a specific CPU configuration setting (i.e. 'decode assists
      > supported') , I guess I should remove my "patch" and simply interrogate AH
      > when an interrupt occurs, instead of counting on EXITINFO1.
      >
      > Also, is this discussion board/forum the appropriate channel for
      > discussing these types of issues? There is another more complicated
      > situation that I have seen, and although I am now doubtful that it is due
      > to a bug, confirmation from people such as yourself would be a great help
      > and possibly save me days of toiling over something that isn't even a bug,
      > but something I'm doing wrong.
      >
      > If this isn't the correct channel, would the mailing list be?  Don't
      > worry, before I use any channel to discuss this other situation, I will go
      > back and look at it again from a "maybe its me doing something wrong"
      > perspective.
      >
      > Thanks again and let me know if there is something I can do to pay back
      > the time you are spending on me.
      >
      > Chris
      > ---
      > [Help to submit a fix or patch](
      > https://sourceforge.net/p/bochs/discussion/39592/thread/0e103a69/?limit=25#3bbb
      > )
      >
      > ---
      >
      > Sent from sourceforge.net because you indicated interest in <
      > https://sourceforge.net/p/bochs/discussion/39592/>
      >
      > To unsubscribe from further messages, please visit <
      > https://sourceforge.net/auth/prefs/>
      >
      
       
  • Hi,

    I really glad that somebody finally came to play with the code I've done and report if there are issues or problems.
    The SVM code in Bochs is not very much validated (I only booted Linux in the SVM enabled VirtualBox under Bochs, nothing more).

    The better channel to discuss it would be the Bochs-Developers mailing list.

    Thanks,
    Stanislav

     
  • VBWizard
    VBWizard
    2013-02-19

    Stanislav,

    I am working on an idea I had to create something along the lines of ESX. I am very interested in the idea of running a hypervisor on bare metal instead of on top of an OS. Before I undertake such a big idea though, I thought I should learn everything I can about both SVM and VMX. I've always found machine level programming to be the most fun kind. In fact the first computer language I ever learned was x86 assembly language.

    I find BOCHS to be an awesome system to use for learning because it gives you full control over the machine. I am glad to test out your functionality and will discuss any findings on the developers mailing list.

    I've gone back to check everything I can about the issue I am seeing before I report it. If I don't find any errors/omissions in my code, I'll be sending my first mailing list email soon.

    Thanks again,

    Chris