Re: [seL4] arm_data_abort_exception

Hi Alex,

A trick that I use to overcome this problem is to introduce a global variable that reflects the status of the PD. The UART driver then switches between the physical and virtual address of the UART as required.

that's it - seems before kernel activates the PD, Kernel might run into assertion which wants to print. I've changed the UART driver like --- a/src/plat/keystone/machine/io.c +++ b/src/plat/keystone/machine/io.c @@ -18,14 +18,19 @@ #define ULSR 0x14 /* UART Line Status Register */ #define ULSR_THRE BIT(5) /* Transmit Holding Register Empty */ -#define UART_REG(x) ((volatile uint32_t *)(UART0_PPTR + (x))) +#define UART_REG(x,uart_addr) ((volatile uint32_t *)(uart_addr + (x))) #if defined(CONFIG_PRINTING) || defined(CONFIG_DEBUG_BUILD) void putDebugChar(unsigned char c) { - while ((*UART_REG(ULSR) & ULSR_THRE) == 0); - *UART_REG(UTHR) = c; + if (status_activate_global_pd) { + while ((*UART_REG(ULSR, UART0_PPTR) & ULSR_THRE) == 0); + *UART_REG(UTHR,UART0_PPTR) = c; + } else { + while ((*UART_REG(ULSR, UART0_PADDR) & ULSR_THRE) == 0); + *UART_REG(UTHR,UART0_PADDR) = c; + } } #endif When running the machine now, I am at least able to see the assertion .. ELF-loader started on CPU: ARM Ltd. Cortex-A15 r2p4 paddr=[90000000..9048841f] ELF-loading image 'kernel' paddr=[80000000..8002ffff] vaddr=[e0000000..e002ffff] virt_entry=e0000000 ELF-loading image 'sel4test-driver' paddr=[80030000..80411fff] vaddr=[10000..3f1fff] virt_entry=1e920 Enabling MMU and paging Jumping to kernel-image entry point... seL4 failed assertion '(address & ~0xfffff000ul) == ((0 && (address & (1ul << 31))) ? 0x0 : 0)' at ./arch/object/structures_gen.h:2495 in function pte_pte_small_new halting... Thanks! - Wladi 2017-01-25 1:10 GMT+01:00 :

Hi Wladislav,

If the fault occurs before the activation of the kernel PD, this likely means that you are trying to printf before the UART is mapped.

If you are using printf for debugging before the PD is activated, you will need to change your kernel UART driver to use the physical address of the peripheral (the elfloader sets up unity mappings for 0x00000000-0xe0000000).

A trick that I use to overcome this problem is to introduce a global variable that reflects the status of the PD. The UART driver then switches between the physical and virtual address of the UART as required.

- Alex

On Tue, 2017-01-24 at 23:55 +0100, Wladislav Wiebe wrote:

...

Hi Alex,

thanks for pointing that out - I will double check the device mapping.

- Wladislav Wiebe

2017-01-24 23:24 GMT+01:00 :

...

A "Synchronous Data abort" occurs when a virtual memory translation fails on load/store operations. A "Prefetch abort" occurs when a virtual memory translation fails when loading instructions into the CPU pipeline. An "Asynchronous Data abort" occurs when a physical memory translation fails (ie, no RAM or peripheral is mapped to the physical address)

The address of the fault corresponds to a kernel device mapping. My guess is that you are not mapping kernel devices correctly at boot time: https://github.com/seL4/seL4/blob/master/include/plat/exynos5/plat/machine/d... https://github.com/seL4/seL4/blob/master/include/plat/exynos5/plat/machine/h... https://github.com/seL4/seL4/blob/master/src/arch/arm/32/machine/hardware.c#...

- Alex

On Tue, 2017-01-24 at 17:17 +0100, Wladislav Wiebe wrote:

...

could it be an TLB fault? For my understandig, based on http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0438g/BABFFDF... it is the DFSR = 0x7 -> b00111 Translation fault, 2nd level.

Thanks! - Wladislav Wiebe

2017-01-24 15:42 GMT+01:00 Wladislav Wiebe :

...

Hello,

has somebody experience with running into arm_data_abort_exception after enabling the MMU/paging and jumping to the kernel image: .. ELF-loader started on CPU: ARM Ltd. Cortex-A15 r2p4 paddr=[90000000..904b041f] ELF-loading image 'kernel' paddr=[80000000..80033fff] vaddr=[f0000000..f0033fff] virt_entry=f0000000 ELF-loading image 'sel4test-driver' paddr=[80034000..8042dfff] vaddr=[10000..409fff] virt_entry=1e924 Enabling MMU and paging Jumping to kernel-image entry point...

check_data_abort_exception() DFAR = 0xfff01014 DFSR = 0x7 ADFSR = 0x0 abort() called.

I've dumped the DFAR/DFSR register --

Relevant config parts about the setup: CONFIG_ARCH_ARM_V7A=y CONFIG_ARCH_ARM=y CONFIG_ARCH_AARCH32=y CONFIG_ARM_CORTEX_A15=y CONFIG_PLAT_KEYSTONE=y

Thanks in advance! -- WBR, Wladislav WIebe

-- WBR, Wladislav WIebe