Hi there. I have modified the kvmtest program in the SVN repo, so it prints out the registers. Equally, I have modified the master branch, so that it actually stores the registers on the vcpu struct upon invoking the swi handler. Of course, this can be expanded (and probably put in its own function) to also store CPSR, SPSR, etc. etc., but more about this later. If you need that soon, tell me, and I'll look into it. Regarding 42, well, I just thought the machine would feel powerfull starting with 42 in register 0 :) No, it was an old test that I forgot to remove. Taken care of now. However, I changed the program to actually use the header files for our source, so now it depends on the kernel source to be in ~/src/android/kernel. If you don't keep it there, perhaps modify the makefile or create a symlink. It will probably be easy generally later on if we all store it there. There were some issues with pulling the changes, which Andreas had made, but we fixed that now. We have to sync a little about the vcpu_arch fields, since I don't think it will be necessary to have separate registers for the banked registers 8-12. Especially not, if we are not hijacking async. interrupts - referring to our last meeting. I hope that Andreas and I will have a better view on this during the weekend and get back to you. Over the weekend, I hope we will sync better with regards to the vcpu_arch and regs struct. However, I'm down with a cold, so I can't promise to be too productive. CU, Christoffer PS: We modified the scripts/kvmtest.sh script to use the example.bin file in SVN repo if anybody is using the 4-line no-brainer script. On Fri, Apr 3, 2009 at 8:25 AM, Andreas Nilsson <apn2107 at columbia.edu>wrote: > Brian, > > Well you could just have run it from master without merging anything? The > output from the kvmtest program looks right. The address returned is the > address of the replaced interrupt handler. > > I don't know what code you are talking about that stores 42 into r0, maybe > I > got the branches wrong, I'll check it out. In example.s the tenth Fibonacci > number is calculated and stored into r2. It is later printed out using > printk in the swi_handler when running > > $ kvmtest example.bin > > on the emulator. Doing a dmesg (on the emulator) should output something > like > "the Fibonacci number calculated is ..." in the end. > > The reason why we don't print anything from example.s is that it would > require library support. Hope this clear things up somewhat. You can have a > look in arm.c and arm_interrupts.S to see what's going on. > > /Andreas > > > -----Original Message----- > From: android-virt-bounces at lists.cs.columbia.edu > [mailto:android-virt-bounces at lists.cs.columbia.edu] On Behalf Of Brian > Smith > Sent: den 2 april 2009 22:43 > To: android-virt at lists.cs.columbia.edu > Subject: Re: running flat binaries > > After a very entertaining 3 hours or so of merging your branch into > mine, I get the following when running kvmtest on android > > Created VM > Created VCPU > Got KVM registers struct > Sent back KVM registers struct > Issued run command, return value is: c5aec000 > > I didn't see anything significant in dmesg, what is supposed to show up > in there? The above doesn't look right > to me, which means either I merged something incorrectly, or something > else... Also, besides being the answer > to life, the universe and everything, what does storing 42 into guest R0 > do? > > Note I haven't tried to debug this yet, if you guys have a quick answer > to the above great, if not then I know where > my starting point is:) > > Brian > > Andreas Nilsson wrote: > > Brian (and other interested parties), > > > > we have now finished a basic kvmtest application which runs a flat > > binary. Do an svn update of android/src/kvmtest > > > > 1. run make in the kvmtest directory which will create the two binaries: > > > > kvmtest - Actual test executable > > example.bin - Flat binary which calculates the 10th Fibonacci number > > and stores it in r2 > > > > 2. copy kvmtest and example.bin to for example /data on the emulator > > > > 3. run > > $ adb shell /data/kvmtest /data/example.bin > > > > 4. run > > $ adb shell dmesg > > > > to see the fibonacci number outputted from the swi hook > > > > 5. To create more binaries simply copy example.s and add the new file > > to the Makefile. It is probably possible to do the the exact same > > thing for library-independent C source files but we didn't try it yet > > and it's also more messy to get some valuable info out of the > > execution since you don't control the registers completely. > > > > We have added two "Brian todos" in arch/arm/kvm/arm.c to indicate > > appropriate places to hook in the translation and emulation > > respectively. Both of these locations are in C files so no funky asm > > calls needed :-) > > > > We also added code to track the pid of the guest process to be able to > > determine which process caused an SWI and act accordingly. It should > > all be documented in arm.c > > > > Please shout if anything is unclear. > > > > /Andreas > > > > > > > > To create more binaries simply copy example.s and add the new file to > > the Makefile. > > > > Jason Nieh wrote: > >> I'm stuck in a meeting off campus and will be 15 min late. > >> -- > >> Jason Nieh http://www.cs.columbia.edu/~nieh<http://www.cs.columbia.edu/%7Enieh> > >> <http://www.cs.columbia.edu/%7Enieh> > >> > >> On Mar 30, 2009, at 10:26 PM, Christoffer Dall <cd2436 at columbia.edu > >> <mailto:cd2436 at columbia.edu>> wrote: > >> > >>> Hi. > >>> > >>> Sorry about this. Andreas and I just discussed this, and we couldn't > >>> find a time that was good for the both of us, so I am just gonna > >>> meet with you guys at 6:30 and represent Andreas at my best. > >>> > >>> See you tomorrow, and I will be practicing my Swedish accent in the > >>> mean time :) > >>> > >>> /Christoffer > >>> > >>> On Mon, Mar 30, 2009 at 10:20 PM, David Albert > >>> <davidbalbert at gmail.com <mailto:davidbalbert at gmail.com>> wrote: > >>> > >>> Should be fine for me. > >>> > >>> Best, > >>> -D > >>> > >>> > >>> On Mar 30, 2009, at 10:20 PM, Brian Smith wrote: > >>> > >>> Hey guys, > >>> No one has confirmed tomorrow yet, I wanted to determine if > >>> we are or are not meeting. Like I said below, I can meet > >>> anytime > >>> after 6:30 pm, is 6:30 pm good for everyone? > >>> > >>> Brian > >>> > >>> > >>> Brian Smith wrote: > >>> > >>> Hi all, > >>> Tuesdays are fine, I can definitely do 6:30 or later, 6 > >>> is a little too early for me (I can't guarantee I will be > >>> home from work by then). > >>> > >>> Regards, > >>> Brian > >>> > >>> Christoffer Dall wrote: > >>> > >>> > >>> > >>> On Fri, Mar 27, 2009 at 10:22 AM, Jason Nieh > >>> <nieh at cs.columbia.edu <mailto:nieh at cs.columbia.edu> > >>> <mailto:nieh at cs.columbia.edu > >>> <mailto:nieh at cs.columbia.edu>>> wrote: > >>> > >>> > >>> i suggest we do the following: > >>> > >>> 1. have separate development branches, one for the > >>> trap/emulate work > >>> that brian is doing, and one for the get the > >>> basic user mode stuff > >>> to work that andreas/christoffer are doing. > >>> > >>> > >>> This is already set up. If anyone needs help with > >>> practical details, please e-mail me and I shall see > >>> what I can do to help. > >>> > >>> > >>> > >>> 2. all four of you should have a weekly conference > >>> call IN ADDITION > >>> to the one that we are having on wednesdays to > >>> go over your > >>> respective parts and stay in sync. please set > >>> up a regular > >>> meeting time and let me know when - there should > >>> not be a week of > >>> "we haven't done anything and therefore want to > >>> skip the > >>> meeting". major design changes to the main > >>> branch should be > >>> discussed. monday is probably a good day for > >>> this. > >>> > >>> > >>> What times are good for you guys? Tuesdays are better > >>> for me, around 5/6pm? > >>> > >>> > >>> > >>> 3. when code is known to RUN not just compile, it > >>> should be checked > >>> into the main branch. if there are merging > >>> issues, christoffer > >>> has volunteered to take care of merging. > >>> > >>> > >>> Sure, I will take of that. We expect to check in > >>> first working stuff tomorrow evening. We are fighting > >>> with kernel panics after interrupt vector > >>> replacements. More on this later. > >>> > >>> > >>> > >>> 4. i suggest we get basically functionality > >>> working sooner as opposed > >>> to later, and check-in to the main branch > >>> smaller parts that are > >>> functional (ie RUN correctly) as opposed to > >>> waiting. > >>> > >>> 5. christoffer - please set up a mailing list for > >>> the group > >>> @lists.cs.columbia.edu > >>> <http://lists.cs.columbia.edu> > >>> <http://lists.cs.columbia.edu>. i'd > >>> like to see all email communication > >>> related to the project posted on the mailing > >>> list so everyone is > >>> aware of the technical issues. please include > >>> oren as well. > >>> > >>> > >>> in process... > >>> > >>> > >>> > >>> 6. if there are important wiki updates, please > >>> also email everyone a > >>> pointer to the new content. > >>> > >>> > >>> thanks, > >>> > >>> -- > >>> Jason Nieh http://www.cs.columbia.edu/~nieh<http://www.cs.columbia.edu/%7Enieh> > >>> <http://www.cs.columbia.edu/%7Enieh> > >>> <http://www.cs.columbia.edu/%7Enieh> > >>> > >>> > >>> > >>> > >>> > > > > > _______________________________________________ > Android-virt mailing list > Android-virt at lists.cs.columbia.edu > https://lists.cs.columbia.edu/cucslists/listinfo/android-virt > > > _______________________________________________ > Android-virt mailing list > Android-virt at lists.cs.columbia.edu > https://lists.cs.columbia.edu/cucslists/listinfo/android-virt > -------------- next part -------------- An HTML attachment was scrubbed... URL: https://lists.cs.columbia.edu/pipermail/android-virt/attachments/20090403/abfd9a60/attachment-0001.html
