diff options
author | Jesper Nilsson <jespern@axis.com> | 2014-10-08 14:57:35 +0200 |
---|---|---|
committer | Jesper Nilsson <jespern@axis.com> | 2014-12-20 00:05:36 +0100 |
commit | 421d085252613c4caf9c4f2fda3c722a9f5bc554 (patch) | |
tree | fc3e9c8ddfc48a019950154d4ce6fe12f2524dbf | |
parent | ca36c1faddb48c478f7f8944b9b900fc72629310 (diff) | |
download | linux-rpi3-421d085252613c4caf9c4f2fda3c722a9f5bc554.tar.gz linux-rpi3-421d085252613c4caf9c4f2fda3c722a9f5bc554.tar.bz2 linux-rpi3-421d085252613c4caf9c4f2fda3c722a9f5bc554.zip |
CRIS: Use KALLSYMs if available in call stack dump
Also, print kernel version on oops.
Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
-rw-r--r-- | arch/cris/kernel/traps.c | 61 |
1 files changed, 34 insertions, 27 deletions
diff --git a/arch/cris/kernel/traps.c b/arch/cris/kernel/traps.c index 0ffda73734f5..da4c72401e27 100644 --- a/arch/cris/kernel/traps.c +++ b/arch/cris/kernel/traps.c @@ -14,6 +14,10 @@ #include <linux/init.h> #include <linux/module.h> +#include <linux/utsname.h> +#ifdef CONFIG_KALLSYMS +#include <linux/kallsyms.h> +#endif #include <asm/pgtable.h> #include <asm/uaccess.h> @@ -34,25 +38,24 @@ static int kstack_depth_to_print = 24; void (*nmi_handler)(struct pt_regs *); -void -show_trace(unsigned long *stack) +void show_trace(unsigned long *stack) { unsigned long addr, module_start, module_end; extern char _stext, _etext; int i; - printk("\nCall Trace: "); + pr_err("\nCall Trace: "); i = 1; module_start = VMALLOC_START; module_end = VMALLOC_END; - while (((long)stack & (THREAD_SIZE-1)) != 0) { + while (((long)stack & (THREAD_SIZE - 1)) != 0) { if (__get_user(addr, stack)) { /* This message matches "failing address" marked s390 in ksymoops, so lines containing it will not be filtered out by ksymoops. */ - printk("Failing address 0x%lx\n", (unsigned long)stack); + pr_err("Failing address 0x%lx\n", (unsigned long)stack); break; } stack++; @@ -68,10 +71,14 @@ show_trace(unsigned long *stack) if (((addr >= (unsigned long)&_stext) && (addr <= (unsigned long)&_etext)) || ((addr >= module_start) && (addr <= module_end))) { +#ifdef CONFIG_KALLSYMS + print_ip_sym(addr); +#else if (i && ((i % 8) == 0)) - printk("\n "); - printk("[<%08lx>] ", addr); + pr_err("\n "); + pr_err("[<%08lx>] ", addr); i++; +#endif } } } @@ -111,21 +118,21 @@ show_stack(struct task_struct *task, unsigned long *sp) stack = sp; - printk("\nStack from %08lx:\n ", (unsigned long)stack); + pr_err("\nStack from %08lx:\n ", (unsigned long)stack); for (i = 0; i < kstack_depth_to_print; i++) { if (((long)stack & (THREAD_SIZE-1)) == 0) break; if (i && ((i % 8) == 0)) - printk("\n "); + pr_err("\n "); if (__get_user(addr, stack)) { /* This message matches "failing address" marked s390 in ksymoops, so lines containing it will not be filtered out by ksymoops. */ - printk("Failing address 0x%lx\n", (unsigned long)stack); + pr_err("Failing address 0x%lx\n", (unsigned long)stack); break; } stack++; - printk("%08lx ", addr); + pr_err("%08lx ", addr); } show_trace(sp); } @@ -139,33 +146,32 @@ show_stack(void) unsigned long *sp = (unsigned long *)rdusp(); int i; - printk("Stack dump [0x%08lx]:\n", (unsigned long)sp); + pr_err("Stack dump [0x%08lx]:\n", (unsigned long)sp); for (i = 0; i < 16; i++) - printk("sp + %d: 0x%08lx\n", i*4, sp[i]); + pr_err("sp + %d: 0x%08lx\n", i*4, sp[i]); return 0; } #endif -void -set_nmi_handler(void (*handler)(struct pt_regs *)) +void set_nmi_handler(void (*handler)(struct pt_regs *)) { nmi_handler = handler; arch_enable_nmi(); } #ifdef CONFIG_DEBUG_NMI_OOPS -void -oops_nmi_handler(struct pt_regs *regs) +void oops_nmi_handler(struct pt_regs *regs) { stop_watchdog(); oops_in_progress = 1; - printk("NMI!\n"); + pr_err("NMI!\n"); show_registers(regs); oops_in_progress = 0; + oops_exit(); + pr_err("\n"); /* Flush mtdoops. */ } -static int __init -oops_nmi_register(void) +static int __init oops_nmi_register(void) { set_nmi_handler(oops_nmi_handler); return 0; @@ -180,8 +186,7 @@ __initcall(oops_nmi_register); * similar to an Oops dump, and if the kernel is configured to be a nice * doggy, then halt instead of reboot. */ -void -watchdog_bite_hook(struct pt_regs *regs) +void watchdog_bite_hook(struct pt_regs *regs) { #ifdef CONFIG_ETRAX_WATCHDOG_NICE_DOGGY local_irq_disable(); @@ -196,8 +201,7 @@ watchdog_bite_hook(struct pt_regs *regs) } /* This is normally the Oops function. */ -void -die_if_kernel(const char *str, struct pt_regs *regs, long err) +void die_if_kernel(const char *str, struct pt_regs *regs, long err) { if (user_mode(regs)) return; @@ -211,13 +215,17 @@ die_if_kernel(const char *str, struct pt_regs *regs, long err) stop_watchdog(); #endif + oops_enter(); handle_BUG(regs); - printk("%s: %04lx\n", str, err & 0xffff); + pr_err("Linux %s %s\n", utsname()->release, utsname()->version); + pr_err("%s: %04lx\n", str, err & 0xffff); show_registers(regs); + oops_exit(); oops_in_progress = 0; + pr_err("\n"); /* Flush mtdoops. */ #ifdef CONFIG_ETRAX_WATCHDOG_NICE_DOGGY reset_watchdog(); @@ -225,8 +233,7 @@ die_if_kernel(const char *str, struct pt_regs *regs, long err) do_exit(SIGSEGV); } -void __init -trap_init(void) +void __init trap_init(void) { /* Nothing needs to be done */ } |