Emilio, Thank you for sharing this. I think this will be very helpful for me. I am going to look it over and see if I can get 3.2.24 up and running, and I'll plan to let you know what happens. (It might take me awhile, as I have other things going on.)
Thanks again and Regards, Patrick On Fri, Mar 6, 2015 at 1:44 PM, Castillo Villar, Emilio < emilio.casti...@unican.es> wrote: > Hello all, > > I have been able to boot newer and unmodified x86 kernels after doing a > bit of research and debugging, I managed to get a 3.2.24 kernel to work. > > One of the main issues was the SMP initialization with the "trampoline" > code. > > This is the very first routine that any slave processor executes when it > boots. > > Before 2.6.38 series, the trampoline code was always fixed in a very low > memory position, as the cpu boots in real mode it has to be within the > first MB. > > Newer kernels changed this allocation to be dynamic. They read the e820 > bios memory tables at the beginning and try to allocate memory starting > from the high positions. As a result, the trampoline code will be in the > higher part of this first MB. As Gem5 lacks support for doing large jumps > while in real mode. When the slave processor booted, the simulator will > fail due to an unimplemented "long jump" instruction. > > In order to work around this, I modified the e820 memory maps to make > some of the lower positions allocatable. > > In file configs/common/FSConfig.py > > change the e820 map for the following, this will force the trampoline > code to load on the first 64kB segment. > > entries = \ > [ > X86E820Entry(addr = 0x0, size = '64kB', range_type = 1), > X86E820Entry(addr = 0x10000, size = '960kB', range_type = 2), > # Mark the rest of physical memory as available > X86E820Entry(addr = 0x100000, > size = '%dB' % (self.mem_ranges[0].size() - 0x100000), > range_type = 1), > ] > > Now you need one of the .config files provided in the wiki: > wget http://www.m5sim.org/dist/current/x86/config-x86.tar.bz2 > http://www.m5sim.org/dist/current/x86/config-x86.tar.bz2 > > THIS IS MANDATORY as this config file sets some options needed for the > kernel to work correctly, > copy the linux-2.6.28.4 file to the kernel 3.2 root and rename it to > ".config", then execute make oldconfig. Just press enter till is over. > > There is an option that must be changed in the kernel > CONFIG_X86_RESERVE_LOW, this is set to 64 (KB) by default but for the > kernel to boot in smp change it to 4, > if using "make menuconfig": > Processor types and features -> Amount of low memory, in kilobytes, to > reserve for the bios > > This should be enough, I have managed to boot up to 16 cores with this > kernel. > > Thanks for your time and hope this info may be of any use. > > ------------------------------ > *De:* gem5-users [gem5-users-boun...@gem5.org] en nombre de Joel Hestness > via gem5-users [gem5-users@gem5.org] > *Enviado:* viernes, 06 de marzo de 2015 20:50 > *Para:* Patrick; gem5 users mailing list > *Asunto:* Re: [gem5-users] Linux kernel > > > To get a working x86 Linux kernel, you shouldn't need to patch the >> source. The patches were developed to speed up the kernel timer >> configuration for ALPHA, but x86 uses a different mechanism, so those are >> unnecessary. >> > > Apologies... to be clearer: The major thing for getting a working kernel > is configuring the build correctly to keep gem5 from trying to use devices > and CPU features that are not implemented. The build config files > <http://www.m5sim.org/dist/current/x86/config-x86.tar.bz2> available on > the gem5 site should be a good starting point. > > Joel > > > > On Fri, Mar 6, 2015 at 12:01 PM, Patrick via gem5-users < >> gem5-users@gem5.org> wrote: >> >>> Andreas, >>> >>> Thanks. ARM would be useful for me, but I would like at least to try >>> to get x86 working, since that would be more useful for my purposes. I came >>> across this page (link <http://www.m5sim.org/Compiling_a_Linux_Kernel>) >>> explaining the patching that has to be done to the kernel to make it work. >>> I will try to work through that. >>> >>> Thanks, >>> Patrick >>> >>> On Fri, Mar 6, 2015 at 1:13 AM, Andreas Hansson <andreas.hans...@arm.com >>> > wrote: >>> >>>> Hi Patrick, >>>> >>>> It might not be what you are looking for, but just in case… if you >>>> want the latest linux kernel you should be just fine with ARM (aarch64). >>>> There are lots of people using ARM full system so there should also be >>>> plenty help available if needed. >>>> >>>> Andreas >>>> >>>> From: Patrick via gem5-users <gem5-users@gem5.org> >>>> Reply-To: Patrick <plafr...@gmail.com>, gem5 users mailing list < >>>> gem5-users@gem5.org> >>>> Date: Thursday, 5 March 2015 22:28 >>>> To: gem5 users mailing list <gem5-users@gem5.org> >>>> Subject: Re: [gem5-users] Linux kernel >>>> >>>> Just to offer an update, I was able to build kernel version 2.6.28.4 >>>> by rolling back to gcc 3.4. When I try to run gem5 in FS mode with this >>>> kernel, though, I get this error: >>>> >>>> Kernel panic - not syncing: Attempted to kill the idle task! >>>> >>>> >>>> It looks like this error has come up in past discussions on the gem5 >>>> boards, so I will look over those. (Sorry for sidetracking the >>>> conversation. I will eventually like to look into newer versions of the >>>> kernel, but I figured starting with an older, known working version of the >>>> kernel was a good starting point.) >>>> >>>> -Patrick >>>> >>>> On Thu, Mar 5, 2015 at 10:53 AM, Patrick <plafr...@gmail.com> wrote: >>>> >>>>> Konstantinos, >>>>> >>>>> I tried running the latest stable version (3.19). It built >>>>> succesfully from source, but when I tried to run it, I got the following >>>>> error: >>>>> >>>>> panic: Resetting mouse wrap mode unimplemented. >>>>> @ tick 928089856500 >>>>> [processData:build/X86/dev/x86/i8042.cc, line 258] >>>>> Memory Usage: 1084128 KBytes >>>>> Program aborted at cycle 928089856500 >>>>> Aborted (core dumped) >>>>> >>>>> I tried building two of the versions of the kernel for which >>>>> there are configuration files provided for x86 on the gem5 wiki (here >>>>> <http://www.m5sim.org/Linux_kernel>). This included versions 2.6.22.9 >>>>> and 2.6.28.4. When I tried building 2.6.22.9, I got this error: >>>>> >>>>> >>>>> In file included from arch/x86_64/kernel/asm-offsets.c:7:0: >>>>> include/linux/crypto.h:20:24: fatal error: asm/atomic.h: No such >>>>> file or directory >>>>> >>>>> >>>>> When I tried to build 2.6.28.4, I got this error: >>>>> >>>>> gcc: error: unrecognized command line option ‘-m’ >>>>> gcc: error: elf_x86_64: No such file or directory >>>>> >>>>> >>>>> From reading discussion boards about building the kernel, I think >>>>> these errors might mean I need to roll back my toolchain to build these >>>>> previous versions of the kernel. However, I have little experience with >>>>> building kernels, so I need to do more reading about this. >>>>> >>>>> Just thought I'd share my experience, as it sounds like we are >>>>> working on similar problems. >>>>> >>>>> I had had a brief discussion with another individual on this message >>>>> board a long time ago on the related topic of creating a disk image for >>>>> gem5 (link >>>>> <http://www.mail-archive.com/gem5-users%40gem5.org/msg10731.html>). >>>>> This discussion had lead me to the script gem5/gem5img.py, but I haven't >>>>> yet gotten around to revisiting that to learn more about it. >>>>> >>>>> I'd be interested in any updates you might have on this. >>>>> >>>>> Thanks, >>>>> Patrick >>>>> >>>>> On Sat, Feb 14, 2015 at 2:40 AM, Konstadinos PARASYRIS via gem5-users >>>>> <gem5-users@gem5.org> wrote: >>>>> >>>>>> >>>>>> Hi, >>>>>> >>>>>> Has anyone successfully executed a kernel 3.2 or newer under X86 full >>>>>> system simulation? >>>>>> >>>>>> >>>>>> Konstantinos Parasyris >>>>>> >>>>>> _______________________________________________ >>>>>> gem5-users mailing list >>>>>> gem5-users@gem5.org >>>>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >>>>>> >>>>> >>>>> >>>> >>>> -- IMPORTANT NOTICE: The contents of this email and any attachments are >>>> confidential and may also be privileged. If you are not the intended >>>> recipient, please notify the sender immediately and do not disclose the >>>> contents to any other person, use it for any purpose, or store or copy the >>>> information in any medium. Thank you. >>>> >>>> ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, >>>> Registered in England & Wales, Company No: 2557590 >>>> ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 >>>> 9NJ, Registered in England & Wales, Company No: 2548782 >>>> >>> >>> >>> _______________________________________________ >>> gem5-users mailing list >>> gem5-users@gem5.org >>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >>> >> >> >> >> -- >> Joel Hestness >> PhD Candidate, Computer Architecture >> Dept. of Computer Science, University of Wisconsin - Madison >> http://pages.cs.wisc.edu/~hestness/ >> > > > > -- > Joel Hestness > PhD Candidate, Computer Architecture > Dept. of Computer Science, University of Wisconsin - Madison > http://pages.cs.wisc.edu/~hestness/ >
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
