For the interested it turned out to be lack of support for some kind of bionic hack for tls. It is using address 0xffff0ff0 but it needs to be 0xffff0fe0
On Fri, May 13, 2011 at 1:41 PM, Jon Pry <jon...@gmail.com> wrote: > Anything linked with bionic will crash. Static/dynamic. Our init is > actually built using glibc so it works fine. I've got some stuff worked out > to try and debug it. It is crashing because the errno symbol is null. > Presumably because tls init failed for some reason. Thumb is fine. Can't > find anything obviously wrong with the kernel config. > > On Fri, May 13, 2011 at 1:15 PM, Tim Bird <tim.b...@am.sony.com> wrote: > >> On 05/13/2011 09:58 AM, Jon Pry wrote: >> > I had the same idea. Unfortunately strace is built with android libs, so >> > it doesn't work. I even changed it to statically linked, no go. Same >> > binaries work on all manner of kernels I have. .27, .35. no .39. >> >> My advice was geared to hedwin's situation, where he said that >> programs ran, but error'ed out without any logcat or dmesg >> info. >> >> For your problem, do ALL statically linked binaries fail, or just >> some? (Is this the same as Mario or not). >> >> If the kernel can't even start /init, you're in pretty bad >> shape, and I'd suspect a kernel configuration issue (as Mario found). >> I can't think of any arch-specific code that should affect the >> kernel/user interface, but you never know. Maybe it's a "thumb" >> setting? It would be good to know the reason the processes are >> segfaulting. >> >> If you can start in a busybox shell, then use statically-linked >> gdb/gdbserver to capture the illegal instruction, it might help >> to find the problem. If none of the bionic-based programs >> run, you can try putting a glibc-based statically linked >> busybox on the system. >> >> Sorry if you've tried all these. Just tossing out ideas... >> -- Tim >> >> >> > On Fri, May 13, 2011 at 12:26 PM, Tim Bird <tim.b...@am.sony.com >> > <mailto:tim.b...@am.sony.com>> wrote: >> > >> > On 05/13/2011 06:16 AM, hedwin wrote: >> > > Well, zygoto does start but after some time it just exits, and >> after 4 >> > > times the service manager will reboot the system. >> > > >> > > There is no info in logcat or dmesg that provides any feedback >> > about why >> > > zygoto did exit. >> > > >> > > Don't think that init.rc is the problem, booting the same android >> > system >> > > with 2.6.32 is just fine. >> > >> > For startup problems, make sure to set "loglevel 7" >> > in /init.rc >> > >> > And a really good way to delve into the system is with strace. >> > >> > This should be particularly useful for this problem, since >> > it appears it's some interaction between the user-space >> > and the kernel, which is exactly what strace instruments. >> > (well... minus the page faults) >> > >> > Use it with both 2.6.32 and whatever kernel is failing, >> > and compare the results >> > >> > To use strace to trace zygote, just use it as a wrapper >> > in /init.rc. >> > >> > Change: >> > service zygote /system/bin/app_process -Xzygote /system/bin --zygote >> > --start-system-server >> > >> > to >> > >> > service zygote /system/xbin/strace -f -tt -o >> > /cache/debug/boot.strace /system/bin/app_process -Xzygote >> > /system/bin --zygote --start-system-server >> > >> > /cache/debug/boot.strace is a writable place on my hardware >> > with lots of space - you can use any output path you want. >> > >> > Then adb pull the boot.strace file to the host, and start comparing. >> > >> > I hope this helps. >> > �-- Tim >> > >> > ============================= >> > Tim Bird >> > Architecture Group Chair, CE Workgroup of the Linux Foundation >> > Senior Staff Engineer, Sony Network Entertainment >> > ============================= >> > >> > -- >> > unsubscribe: android-kernel+unsubscr...@googlegroups.com >> > <mailto:android-kernel%2bunsubscr...@googlegroups.com> >> > website: http://groups.google.com/group/android-kernel >> > >> > >> > -- >> > unsubscribe: android-kernel+unsubscr...@googlegroups.com >> > website: http://groups.google.com/group/android-kernel >> >> >> -- >> ============================= >> Tim Bird >> Architecture Group Chair, CE Workgroup of the Linux Foundation >> Senior Staff Engineer, Sony Network Entertainment >> ============================= >> >> -- >> unsubscribe: android-kernel+unsubscr...@googlegroups.com >> website: http://groups.google.com/group/android-kernel >> > > -- unsubscribe: android-kernel+unsubscr...@googlegroups.com website: http://groups.google.com/group/android-kernel
