Re: Buildroot on 6446EVM
I've used OpenEmbedded to build images for the EVM before. I built glibc images, but it should be possible to create uclibc images also. Unfortunately, I do not have an EVM to test against so I do not keep a set of current images available for download. Philip Brian Niebuhr wrote: Has anyone got buildroot to work on the 6446EVM? I am currently running the 2.6.23-davinci1 git kernel with a Montavista-generated cramfs root filesystem and that is working fine. However, adding just a couple of packages to create a minimal filesystem makes the filesystem huge. So I wanted to try buildroot to generate a uClibc-based toolchain and filesystem. I was able to get buildroot to create a cramfs root filesystem and I used it to replace my Montavista filesystem. However when the kernel gets to the point where it should run init, it just apparently hangs (no output whatsoever). I've tried various init= kernel parameters, including just init=/bin/sh, and I always get the same results. It's as if busybox can't start running. I assume that I just messed something up in the buildroot configuration, but I'm not sure what. If anyone has this working, could I get a copy of your buildroot .config and/or any tips on how to make it work? Thanks! Brian _ Brian Niebuhr Senior Design Engineer E.F. Johnson ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source smime.p7s Description: S/MIME Cryptographic Signature ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
Re: Buildroot on 6446EVM
Brian Niebuhr wrote: Buildroot is good at building all together, but not good at cleaning. Since your console is working and even init=/bin/sh is not working I suspected EABI/OABI stuff. Moreover, I read some warnings some time ago on the buildroot mailing list about this issue. Caglar - You were correct. After getting a brand new copy of buildroot and configuring it for OABI, /sbin/getty now runs. Unfortunately it hasn't gotten me much further because a shell still won't run. Once I log in, a shell never starts and getty just respawns. Using init=/bin/sh or init=/bin/ash just hangs like it did before with no output. I compiled a hello world program with the uclibc toolchain, and using that for init works just fine. So there's something about the busybox shell that isn't working right. One independent thing to try: Check make menuconfig of your kernel if any floating point emulation is enabled: Floating point emulation --- [*] NWFPE math emulation Regards Dirk ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
RE: Buildroot on 6446EVM
One independent thing to try: Check make menuconfig of your kernel if any floating point emulation is enabled: Floating point emulation --- [*] NWFPE math emulation Dirk - Thanks for this suggestion - I would never have figured that one out. I had a different floating point emulation implementation selected, and when I switched to NWFPE things started working. Can I ask why you suspected floating point emulation? Brian ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
Re: Buildroot on 6446EVM
Brian Niebuhr wrote: One independent thing to try: Check make menuconfig of your kernel if any floating point emulation is enabled: Floating point emulation --- [*] NWFPE math emulation Dirk - Thanks for this suggestion - I would never have figured that one out. I had a different floating point emulation implementation selected, and when I switched to NWFPE things started working. :) Can I ask why you suspected floating point emulation? Had a similiar issue some weeks ago ;) The switch from kernel (no floating point at all) to user space (heavily using floating point, depending on toolchain library) is always a tricky thing cause you get *no* debugging output. Things simply do not work :( I think to remember that default kernel config of git kernel has no floating point emulation enabled at all. That an other FPE don't work is new to me, never tried others ;) Maybe you like to ask on arm-linux list regarding this. Would be an interesting question. Regards Dirk ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
Buildroot on 6446EVM
Has anyone got buildroot to work on the 6446EVM? I am currently running the 2.6.23-davinci1 git kernel with a Montavista-generated cramfs root filesystem and that is working fine. However, adding just a couple of packages to create a minimal filesystem makes the filesystem huge. So I wanted to try buildroot to generate a uClibc-based toolchain and filesystem. I was able to get buildroot to create a cramfs root filesystem and I used it to replace my Montavista filesystem. However when the kernel gets to the point where it should run init, it just apparently hangs (no output whatsoever). I've tried various init= kernel parameters, including just init=/bin/sh, and I always get the same results. It's as if busybox can't start running. I assume that I just messed something up in the buildroot configuration, but I'm not sure what. If anyone has this working, could I get a copy of your buildroot .config and/or any tips on how to make it work? Thanks! Brian _ Brian Niebuhr Senior Design Engineer E.F. Johnson ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
RE: Buildroot on 6446EVM
Brian Niebuhr wrote: Has anyone got buildroot to work on the 6446EVM? I am currently running the 2.6.23-davinci1 git kernel with a Montavista-generated cramfs root filesystem and that is working fine. However, adding just a couple of packages to create a minimal filesystem makes the filesystem huge. So I wanted to try buildroot to generate a uClibc-based toolchain and filesystem. I was able to get buildroot to create a cramfs root filesystem and I used it to replace my Montavista filesystem. However when the kernel gets to the point where it should run init, it just apparently hangs (no output whatsoever). I've tried various init= kernel parameters, including just init=/bin/sh, and I always get the same results. It's as if busybox can't start running. I assume that I just messed something up in the buildroot configuration, but I'm not sure what. If anyone has this working, could I get a copy of your buildroot .config and/or any tips on how to make it work? I guess I have faced this problem before. I'm really not sure, but this can be related to EABI/OABI incompatibility. Just a tought. That was my original thought too. In fact, I had configured my kernel without EABI and I configured buildroot to use EABI. Once I noticed this, I did a make clean in buildroot, changed the buildroot ABI setting to OABI, and rebuilt buildroot. But unfortunately I got the same results. I am going to try getting a pristine copy of buildroot and configuring it correctly in the first place, but I'm guessing that's not going to help. Brian ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
Re: Buildroot on 6446EVM
Brian Niebuhr wrote: That was my original thought too. In fact, I had configured my kernel without EABI and I configured buildroot to use EABI. Once I noticed this, I did a make clean in buildroot, changed the buildroot ABI setting to OABI, and rebuilt buildroot. But unfortunately I got the same results. I am going to try getting a pristine copy of buildroot and configuring it correctly in the first place, but I'm guessing that's not going to help. Oops! First mail was not replied to all. Sorry for the inconvenience. Buildroot is good at building all together, but not good at cleaning. Since your console is working and even init=/bin/sh is not working I suspected EABI/OABI stuff. Moreover, I read some warnings some time ago on the buildroot mailing list about this issue. Regards, Caglar ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
RE: Buildroot on 6446EVM
Buildroot is good at building all together, but not good at cleaning. Since your console is working and even init=/bin/sh is not working I suspected EABI/OABI stuff. Moreover, I read some warnings some time ago on the buildroot mailing list about this issue. Caglar - You were correct. After getting a brand new copy of buildroot and configuring it for OABI, /sbin/getty now runs. Unfortunately it hasn't gotten me much further because a shell still won't run. Once I log in, a shell never starts and getty just respawns. Using init=/bin/sh or init=/bin/ash just hangs like it did before with no output. I compiled a hello world program with the uclibc toolchain, and using that for init works just fine. So there's something about the busybox shell that isn't working right. In any case, I got a little further. Thanks for the help. Brian ___ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source