Hey,
I just started adding support for the Zynq SoC to base-hw (see [1]).
As I don't have any Zynq hardware available yet, I tried running this on QEMU 2.1.2; I compiled the run/hw_info target.
After applying some printf debugging ;), I figured out that the program dies (i.e. QEMU freezes) during kernel initialisation in init_virt_kernel(). I therefore tried attaching gdb to QEMU's internal gdb server, however, it seems to ignore my breakpoints.
Hence my question: Do you guys have any experience with low-level/kernel debugging on QEMU? Maybe I am missing something or QEMU is just weird?
[1] https://github.com/ValiValpas/genode/commit/4e375384d635d06626f0b226ac14e21f...
Cheers Johannes
Hi Johannes,
Sorry for answering that late. I had overlooked your mail.
Unfortunately the newest qemu version available for my system is 2.0.0 so there might be differences in our observations. After modifying your qemu config from "-m 1024" to "-m 256" I get serial output from the beginning of 'init_kernel_up' till 'Sctlr::write(Sctlr::init_virt_kernel());' in 'init_kernel_mp -> Cpu::init_virt_kernel'. So this seems to be a problem with enabling the MMU.
You could do single-step at this point and check the IP after the write to the SCTLR to see in which exception vector you end up (see base-hw/src/core/spec/arm_v7/mode_transition.s: _mt_kernel_entry_pic). I assume it to be either _dab_entry or _pab_entry. If this is the case, reading the ARM registers DFSR/DFAR respectively IFSR might help to track down the cause.
One known problem that arises when enabling the MMU is that this also enables ARM caches and branch predictors. You can try deactivating them for test purpose.
'Z::set(v, 0);' in 'Sctlr::init_virt_kernel' in 'base-hw/src/core/include/spec/arm_v7/cpu_support.h' and an empty 'Arm_v7::inval_branch_predicts' method should disable branch predictors.
Should 'C::set(v, 0);' and 'I::set(v, 0);' in 'Sctlr::init_common' in 'base-hw/src/core/include/spec/arm/cpu_support.h' and adapted TEX/C/B fields in the 'CACHED' case in 'Translation::_create' in 'base-hw/src/core/include/spec/arm/short_translation_table.h' should disable ARM caches.
You might also check wether your UART and your Core image are mapped correctly 1:1 by instrumenting the translation-table methods.
Cheers, Martin
On 15.12.2014 19:43, Johannes Schlatow wrote:
Hey,
I just started adding support for the Zynq SoC to base-hw (see [1]).
As I don't have any Zynq hardware available yet, I tried running this on QEMU 2.1.2; I compiled the run/hw_info target.
After applying some printf debugging ;), I figured out that the program dies (i.e. QEMU freezes) during kernel initialisation in init_virt_kernel(). I therefore tried attaching gdb to QEMU's internal gdb server, however, it seems to ignore my breakpoints.
Hence my question: Do you guys have any experience with low-level/kernel debugging on QEMU? Maybe I am missing something or QEMU is just weird?
[1] https://github.com/ValiValpas/genode/commit/4e375384d635d06626f0b226ac14e21f...
Cheers Johannes
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server from Actuate! Instantly Supercharge Your Business Reports and Dashboards with Interactivity, Sharing, Native Excel Exports, App Integration & more Get technology previously reserved for billion-dollar corporations, FREE http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.cl... _______________________________________________ genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main
Hi Martin,
thanks for the hints. I'll check this soon.
I now have a parallella board for testing as well but unfortunately no breakout board yet to which I could connect a JTAG debugger. Hence I'll stick to QEMU for the moment.
Cheers and have some nice holidays Johannes
On Fri, 19 Dec 2014 16:49:53 +0100 Martin Stein <martin.stein@...1...> wrote:
Hi Johannes,
Sorry for answering that late. I had overlooked your mail.
Unfortunately the newest qemu version available for my system is 2.0.0 so there might be differences in our observations. After modifying your qemu config from "-m 1024" to "-m 256" I get serial output from the beginning of 'init_kernel_up' till 'Sctlr::write(Sctlr::init_virt_kernel());' in 'init_kernel_mp -> Cpu::init_virt_kernel'. So this seems to be a problem with enabling the MMU.
You could do single-step at this point and check the IP after the write to the SCTLR to see in which exception vector you end up (see base-hw/src/core/spec/arm_v7/mode_transition.s: _mt_kernel_entry_pic). I assume it to be either _dab_entry or _pab_entry. If this is the case, reading the ARM registers DFSR/DFAR respectively IFSR might help to track down the cause.
One known problem that arises when enabling the MMU is that this also enables ARM caches and branch predictors. You can try deactivating them for test purpose.
'Z::set(v, 0);' in 'Sctlr::init_virt_kernel' in 'base-hw/src/core/include/spec/arm_v7/cpu_support.h' and an empty 'Arm_v7::inval_branch_predicts' method should disable branch predictors.
Should 'C::set(v, 0);' and 'I::set(v, 0);' in 'Sctlr::init_common' in 'base-hw/src/core/include/spec/arm/cpu_support.h' and adapted TEX/C/B fields in the 'CACHED' case in 'Translation::_create' in 'base-hw/src/core/include/spec/arm/short_translation_table.h' should disable ARM caches.
You might also check wether your UART and your Core image are mapped correctly 1:1 by instrumenting the translation-table methods.
Cheers, Martin
On 15.12.2014 19:43, Johannes Schlatow wrote:
Hey,
I just started adding support for the Zynq SoC to base-hw (see [1]).
As I don't have any Zynq hardware available yet, I tried running this on QEMU 2.1.2; I compiled the run/hw_info target.
After applying some printf debugging ;), I figured out that the program dies (i.e. QEMU freezes) during kernel initialisation in init_virt_kernel(). I therefore tried attaching gdb to QEMU's internal gdb server, however, it seems to ignore my breakpoints.
Hence my question: Do you guys have any experience with low-level/kernel debugging on QEMU? Maybe I am missing something or QEMU is just weird?
[1] https://github.com/ValiValpas/genode/commit/4e375384d635d06626f0b226ac14e21f...
Cheers Johannes
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server from Actuate! Instantly Supercharge Your Business Reports and Dashboards with Interactivity, Sharing, Native Excel Exports, App Integration & more Get technology previously reserved for billion-dollar corporations, FREE http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.cl... _______________________________________________ genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server from Actuate! Instantly Supercharge Your Business Reports and Dashboards with Interactivity, Sharing, Native Excel Exports, App Integration & more Get technology previously reserved for billion-dollar corporations, FREE http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.cl... _______________________________________________ genode-main mailing list genode-main@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/genode-main
-
Johannes Schlatow -
Martin Stein