* [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 @ 2002-09-25 9:02 Ingo Molnar 2002-09-25 9:32 ` [re-ANNOUNCE] " Ingo Molnar 2002-09-25 18:16 ` [ANNOUNCE] " Linus Torvalds 0 siblings, 2 replies; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 9:02 UTC (permalink / raw) To: linux-kernel Cc: Linus Torvalds, Kai Germaschewski, Rusty Russell, Arjan van de Ven the attached patch is the latest version of 'kksymoops' for the 2.5 kernel. Kksymoops is an in-kernel symbol resolver, which enables nifty things like: - in-kernel symbolic oopses, symbolic show_stack() and symbolic show_trace(). Finally correct symbolic oopses over serial consoles or netdump. - module symbols are correctly decoded as well. Ie. all the userspace oops decoding mismatches are solved, which can arise if a kernel crashes and another kernel (with different module symbols) is booted. How do you find out the symbols that a particular crashed kernel had? - list of modules are printed upon oopsing - this clearly puts every crash into perspective - exactly which modules were loaded ... a sample kksym-oops as it goes straight into the kernel log: ------------[ cut here ]------------ kernel BUG at time.c:100! invalid operand: 0000 CPU: 1 EIP: 0060:[<c011bc64>] Not tainted EFLAGS: 00010246 EIP is at sys_gettimeofday [kernel] 0x84 eax: 0000004e ebx: ceea0000 ecx: 00000000 edx: 00000068 esi: 00000000 edi: 00000000 ebp: bffffad8 esp: ceea1fa0 ds: 0068 es: 0068 ss: 0068 Process gettimeofday (pid: 547, threadinfo=ceea0000 task=cef2b0c0) Stack: 00000001 00000004 40156154 00000004 c0112a40 ceea0000 400168e4 bffffb44 c0107973 00000000 00000000 40156154 400168e4 bffffb44 bffffad8 0000004e 0000002b 0000002b 0000004e 400cecc1 00000023 00000246 bffffacc 0000002b Call Trace: [<c0112a40>] do_page_fault [kernel] 0x0 [<c0107973>] syscall_call [kernel] 0x7 Code: 0f 0b 64 00 67 b4 26 c0 eb b6 89 f6 83 ec 0c 89 5c 24 04 8b ie. all symbols resolved properly. i believe it's all for the better, much of the above featureset is also based on distributors' daily experience of how users report crashes and how it can be made sense of post-mortem. Tester feedback is often a scarce resource for distributors, so improving the quality of individual reports is of high importance. Even here on lkml the quality of oops reporting is often surprisingly low, especially taking the many years of education into account. the cost of the feature is an in-kernel copy of the symbol table - most testers will not care, and it's default-disabled in the .config. This patch has proven to be very useful in my daily kernel development activities, hopefully others will find this just as useful. I've tested the patch on x86, building and oopsing works both with kksymoops enabled and disabled. The line of credit for kksymoops goes like this: Arjan took Keith's original kallsyms work and extended it to the area of kernel oopsing and stack trace printing - this was the 2.4 kksymoops patch. Which i ported to 2.5 and added some minor fixes, which Kai improved significantly - essencially Kai rewrote much of the original patch - it's now a nice patch that fits into the 2.5 build system properly. Here's an (incomplete) list of Kai's changes: o switched from a four-stage to a two-stage process. o moved the new __kallsyms section to the end of vmlinux, so adding this section it shouldn't affect the addresses of the other symbols, which are lower. o only zero out the actual .bss o move print_modules() into module.c - it doesn't really have anything to do with kallsyms. Also, since it provides a large buffer, I made it actually use it, not truncate to 80 chars (maybe truncating makes sense, but then one should resize the buffer) o move lookup_symbol() into module.c - it's also useful when CONFIG_KALLSYMS is not set (and I killed the ifdefs from kallsyms.c, we now only compile it depending on CONFIG_KALLSYMS). When it cannot resolve symbols using kallsyms, it still tries to use the list of exported symbols (that's somewhat better than nothing, but not sure if really useful) o kallsyms.c used some rather twisted way to get the list of modules ("module_list") instead of just using the global symbol defined in kernel/module.c, so I threw that code out. kallsyms.c provides three functions, only one of which is actually used for kksymoops - I left them all in unconditionally, though, I think kdb or sth uses them. o __{start,stop}___kallsyms needs to be exported, since insmod looks for these symbols to determine if the kernel has kallsyms support (and if so, automatically adds the information to the module inserted) --- linux/arch/i386/kernel/head.S.orig Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/kernel/head.S Wed Sep 25 10:49:32 2002 @@ -121,7 +121,7 @@ */ xorl %eax,%eax movl $__bss_start,%edi - movl $_end,%ecx + movl $__bss_stop,%ecx subl %edi,%ecx rep stosb --- linux/arch/i386/kernel/process.c.orig Wed Sep 25 10:49:32 2002 +++ linux/arch/i386/kernel/process.c Wed Sep 25 10:49:32 2002 @@ -159,14 +159,12 @@ void show_regs(struct pt_regs * regs) { unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; - static char buffer[MAX_SYMBOL_SIZE]; printk("\n"); printk("Pid: %d, comm: %20s\n", current->pid, current->comm); printk("EIP: %04x:[<%08lx>] CPU: %d",0xffff & regs->xcs,regs->eip, smp_processor_id()); + print_symbol("EIP is at %s\n", regs->eip); - lookup_symbol(regs->eip, buffer, MAX_SYMBOL_SIZE); - printk("\nEIP is at %s \n", buffer); if (regs->xcs & 3) printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp); printk(" EFLAGS: %08lx %s\n",regs->eflags, print_tainted()); --- linux/arch/i386/kernel/traps.c.orig Wed Sep 25 10:49:32 2002 +++ linux/arch/i386/kernel/traps.c Wed Sep 25 10:49:32 2002 @@ -132,8 +132,6 @@ { int i; unsigned long addr; - /* static to not take up stackspace; if we race here too bad */ - static char buffer[MAX_SYMBOL_SIZE]; if (!stack) stack = (unsigned long*)&stack; @@ -143,9 +141,12 @@ while (((long) stack & (THREAD_SIZE-1)) != 0) { addr = *stack++; if (kernel_text_address(addr)) { - lookup_symbol(addr, buffer, MAX_SYMBOL_SIZE); - printk("[<%08lx>] %s \n", addr,buffer); - i++; + printk("[<%08lx>] ", addr); + if (print_symbol("%s\n", addr)) { + /* save screen space */ + if ((i++ % 6) == 0) + printk("\n "); + } } } printk("\n"); @@ -198,7 +199,6 @@ int in_kernel = 1; unsigned long esp; unsigned short ss; - static char buffer[MAX_SYMBOL_SIZE]; esp = (unsigned long) (®s->esp); ss = __KERNEL_DS; @@ -207,12 +207,11 @@ esp = regs->esp; ss = regs->xss & 0xffff; } - print_modules(); - lookup_symbol(regs->eip, buffer, MAX_SYMBOL_SIZE); printk("CPU: %d\nEIP: %04x:[<%08lx>] %s\nEFLAGS: %08lx\n", smp_processor_id(), 0xffff & regs->xcs, regs->eip, print_tainted(), regs->eflags); - printk("\nEIP is at %s \n",buffer); + + print_symbol("EIP is at %s\n", regs->eip); printk("eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n", regs->eax, regs->ebx, regs->ecx, regs->edx); printk("esi: %08lx edi: %08lx ebp: %08lx esp: %08lx\n", --- linux/arch/i386/Config.help.orig Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/Config.help Wed Sep 25 10:49:32 2002 @@ -946,6 +946,11 @@ of the BUG call as well as the EIP and oops trace. This aids debugging but costs about 70-100K of memory. +CONFIG_KALLSYMS + Say Y here to let the kernel print out symbolic crash information and + symbolic stack backtraces. This increases the size of the kernel + somewhat, as all symbols have to be loaded into the kernel image. + CONFIG_DEBUG_OBSOLETE Say Y here if you want to reduce the chances of the tree compiling, and are prepared to dig into driver internals to fix compile errors. --- linux/arch/i386/config.in.orig Wed Sep 25 10:28:11 2002 +++ linux/arch/i386/config.in Wed Sep 25 10:49:32 2002 @@ -435,6 +435,7 @@ if [ "$CONFIG_HIGHMEM" = "y" ]; then bool ' Highmem debugging' CONFIG_DEBUG_HIGHMEM fi + bool ' Load all symbols for debugging/kksymoops' CONFIG_KALLSYMS fi if [ "$CONFIG_X86_LOCAL_APIC" = "y" ]; then --- linux/arch/i386/vmlinux.lds.S.orig Fri Sep 20 17:20:19 2002 +++ linux/arch/i386/vmlinux.lds.S Wed Sep 25 10:49:32 2002 @@ -78,9 +78,13 @@ .data.cacheline_aligned : { *(.data.cacheline_aligned) } __bss_start = .; /* BSS */ - .bss : { - *(.bss) - } + .bss : { *(.bss) } + __bss_stop = .; + + __start___kallsyms = .; /* All kernel symbols */ + __kallsyms : { *(__kallsyms) } + __stop___kallsyms = .; + _end = . ; /* Sections to be discarded */ --- linux/include/linux/kallsyms.h.orig Wed Sep 25 10:49:32 2002 +++ linux/include/linux/kallsyms.h Wed Sep 25 10:49:32 2002 @@ -0,0 +1,163 @@ +/* kallsyms headers + Copyright 2000 Keith Owens <kaos@ocs.com.au> + + This file is part of the Linux modutils. It is exported to kernel + space so debuggers can access the kallsyms data. + + The kallsyms data contains all the non-stack symbols from a kernel + or a module. The kernel symbols are held between __start___kallsyms + and __stop___kallsyms. The symbols for a module are accessed via + the struct module chain which is based at module_list. + + This program is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by the + Free Software Foundation; either version 2 of the License, or (at your + option) any later version. + + This program is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software Foundation, + Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + */ + +#ident "$Id: linux-2.4.9-kallsyms.patch,v 1.8 2002/02/11 18:34:53 arjanv Exp $" + +#ifndef MODUTILS_KALLSYMS_H +#define MODUTILS_KALLSYMS_H 1 + +/* Have to (re)define these ElfW entries here because external kallsyms + * code does not have access to modutils/include/obj.h. This code is + * included from user spaces tools (modutils) and kernel, they need + * different includes. + */ + +#ifndef ELFCLASS32 +#ifdef __KERNEL__ +#include <linux/elf.h> +#else /* __KERNEL__ */ +#include <elf.h> +#endif /* __KERNEL__ */ +#endif /* ELFCLASS32 */ + +#ifndef ELFCLASSM +#define ELFCLASSM ELF_CLASS +#endif + +#ifndef ElfW +# if ELFCLASSM == ELFCLASS32 +# define ElfW(x) Elf32_ ## x +# define ELFW(x) ELF32_ ## x +# else +# define ElfW(x) Elf64_ ## x +# define ELFW(x) ELF64_ ## x +# endif +#endif + +/* Format of data in the kallsyms section. + * Most of the fields are small numbers but the total size and all + * offsets can be large so use the 32/64 bit types for these fields. + * + * Do not use sizeof() on these structures, modutils may be using extra + * fields. Instead use the size fields in the header to access the + * other bits of data. + */ + +struct kallsyms_header { + int size; /* Size of this header */ + ElfW(Word) total_size; /* Total size of kallsyms data */ + int sections; /* Number of section entries */ + ElfW(Off) section_off; /* Offset to first section entry */ + int section_size; /* Size of one section entry */ + int symbols; /* Number of symbol entries */ + ElfW(Off) symbol_off; /* Offset to first symbol entry */ + int symbol_size; /* Size of one symbol entry */ + ElfW(Off) string_off; /* Offset to first string */ + ElfW(Addr) start; /* Start address of first section */ + ElfW(Addr) end; /* End address of last section */ +}; + +struct kallsyms_section { + ElfW(Addr) start; /* Start address of section */ + ElfW(Word) size; /* Size of this section */ + ElfW(Off) name_off; /* Offset to section name */ + ElfW(Word) flags; /* Flags from section */ +}; + +struct kallsyms_symbol { + ElfW(Off) section_off; /* Offset to section that owns this symbol */ + ElfW(Addr) symbol_addr; /* Address of symbol */ + ElfW(Off) name_off; /* Offset to symbol name */ +}; + +#define KALLSYMS_SEC_NAME "__kallsyms" +#define KALLSYMS_IDX 2 /* obj_kallsyms creates kallsyms as section 2 */ + +#define kallsyms_next_sec(h,s) \ + ((s) = (struct kallsyms_section *)((char *)(s) + (h)->section_size)) +#define kallsyms_next_sym(h,s) \ + ((s) = (struct kallsyms_symbol *)((char *)(s) + (h)->symbol_size)) + +#ifdef CONFIG_KALLSYMS + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start with */ + const char **mod_name, /* Set to module name or "kernel" */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_sections(void *token, + int (*callback)(void *, /* token */ + const char *, /* module name */ + const char *, /* section name */ + ElfW(Addr), /* Section start */ + ElfW(Addr), /* Section end */ + ElfW(Word) /* Section flags */ + ) + ); + +#else + +static inline int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + return -ESRCH; +} + +#endif + +#endif /* kallsyms.h */ --- linux/include/linux/module.h.orig Wed Sep 25 10:49:32 2002 +++ linux/include/linux/module.h Wed Sep 25 10:49:32 2002 @@ -504,19 +504,26 @@ #define SET_MODULE_OWNER(some_struct) do { } while (0) #endif +extern void print_modules(void); + #if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) extern struct module *module_list; -extern int lookup_symbol(unsigned long address, char *buffer, int buflen); +/* + * print_symbols takes a format string containing one %s. + * If support for resolving symbols is compiled in, the %s will + * be replaced by the closest symbol to the address and the entire + * string is printk()ed. Otherwise, nothing is printed. + */ +extern int print_symbol(const char *fmt, unsigned long address); #else static inline int -lookup_symbol(unsigned long address, char *buffer, int buflen) +print_symbol(const char *fmt, unsigned long address) { - buffer[0] = 0; - return 0; + return -ESRCH; } #endif --- linux/include/linux/sched.h.orig Wed Sep 25 10:49:32 2002 +++ linux/include/linux/sched.h Wed Sep 25 10:49:32 2002 @@ -152,12 +152,9 @@ extern void sched_init(void); extern void init_idle(task_t *idle, int cpu); -#define MAX_SYMBOL_SIZE 512 - extern void show_state(void); extern void show_trace(unsigned long *stack); extern void show_stack(unsigned long *stack); -extern void print_modules(void); extern void show_regs(struct pt_regs *); --- linux/include/asm-i386/hardirq.h.orig Fri Sep 20 17:20:32 2002 +++ linux/include/asm-i386/hardirq.h Wed Sep 25 10:49:32 2002 @@ -97,6 +97,4 @@ extern void synchronize_irq(unsigned int irq); #endif /* CONFIG_SMP */ -extern void show_stack(unsigned long * esp); - #endif /* __ASM_HARDIRQ_H */ --- linux/include/asm-i386/ptrace.h.orig Fri Sep 20 17:20:16 2002 +++ linux/include/asm-i386/ptrace.h Wed Sep 25 10:49:32 2002 @@ -57,7 +57,6 @@ #ifdef __KERNEL__ #define user_mode(regs) ((VM_MASK & (regs)->eflags) || (3 & (regs)->xcs)) #define instruction_pointer(regs) ((regs)->eip) -extern void show_regs(struct pt_regs *); #endif #endif --- linux/kernel/Makefile.orig Fri Sep 20 17:20:19 2002 +++ linux/kernel/Makefile Wed Sep 25 10:49:32 2002 @@ -3,7 +3,7 @@ # export-objs = signal.o sys.o kmod.o context.o ksyms.o pm.o exec_domain.o \ - printk.o platform.o suspend.o dma.o + printk.o platform.o suspend.o dma.o module.o obj-y = sched.o fork.o exec_domain.o panic.o printk.o \ module.o exit.o itimer.o time.o softirq.o resource.o \ @@ -14,6 +14,7 @@ obj-$(CONFIG_SMP) += cpu.o obj-$(CONFIG_UID16) += uid16.o obj-$(CONFIG_MODULES) += ksyms.o +obj-$(CONFIG_KALLSYMS) += kallsyms.o obj-$(CONFIG_PM) += pm.o obj-$(CONFIG_BSD_PROCESS_ACCT) += acct.o obj-$(CONFIG_SOFTWARE_SUSPEND) += suspend.o --- linux/kernel/kallsyms.c.orig Wed Sep 25 10:49:32 2002 +++ linux/kernel/kallsyms.c Wed Sep 25 10:49:32 2002 @@ -0,0 +1,227 @@ +/* + * kksymoops.c: in-kernel printing of symbolic oopses and stack traces. + * + * Copyright 2000 Keith Owens <kaos@ocs.com.au> April 2000 + * Copyright 2002 Arjan van de Ven <arjanv@redhat.com> + * + This code uses the list of all kernel and module symbols to :- + + * Find any non-stack symbol in a kernel or module. Symbols do + not have to be exported for debugging. + + * Convert an address to the module (or kernel) that owns it, the + section it is in and the nearest symbol. This finds all non-stack + symbols, not just exported ones. + + */ + +#include <linux/mm.h> +#include <linux/module.h> +#include <linux/kallsyms.h> + +/* A symbol can appear in more than one module. A token is used to + * restart the scan at the next module, set the token to 0 for the + * first scan of each symbol. + */ + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start at */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec; + const struct kallsyms_symbol *ka_sym = NULL; + const char *ka_str = NULL; + const struct module *m; + int i = 0, l; + const char *p, *pt_R; + char *p2; + + /* Restart? */ + m = module_list; + if (token && *token) { + for (; m; m = m->next) + if ((unsigned long)m == *token) + break; + if (m) + m = m->next; + } + + for (; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + p = ka_str + ka_sym->name_off; + if (strcmp(p, name) == 0) + break; + /* Unversioned requests match versioned names */ + if (!(pt_R = strstr(p, "_R"))) + continue; + l = strlen(pt_R); + if (l < 10) + continue; /* Not _R.*xxxxxxxx */ + (void)simple_strtoul(pt_R+l-8, &p2, 16); + if (*p2) + continue; /* Not _R.*xxxxxxxx */ + if (strncmp(p, name, pt_R-p) == 0) + break; /* Match with version */ + } + if (i < ka_hdr->symbols) + break; + } + + if (token) + *token = (unsigned long)m; + if (!m) + return(0); /* not found */ + + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off); + *mod_name = *(m->name) ? m->name : "kernel"; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = ka_sym->name_off + ka_str; + *sym_start = ka_sym->symbol_addr; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + *sym_end = ka_symn->symbol_addr; + } + else + *sym_end = *sec_end; + return(1); +} + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const struct kallsyms_symbol *ka_sym; + const char *ka_str; + const struct module *m; + int i; + unsigned long end; + + for (m = module_list; m; m = m->next) { + + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off); + /* Is the address in any section in this module? */ + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (ka_sec->start <= address && + (ka_sec->start + ka_sec->size) > address) + break; + } + if (i < ka_hdr->sections) + break; /* Found a matching section */ + } + + if (!m) + return(0); /* not found */ + + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + *mod_name = m->name; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = *sec_name; /* In case we find no matching symbol */ + *sym_start = *sec_start; + *sym_end = *sec_end; + + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + if (ka_sym->symbol_addr > address) + continue; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + end = ka_symn->symbol_addr; + } + else + end = *sec_end; + if (end <= address) + continue; + if ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off + != (char *)ka_sec) + continue; /* wrong section */ + *sym_name = ka_str + ka_sym->name_off; + *sym_start = ka_sym->symbol_addr; + *sym_end = end; + break; + } + return(1); +} + +/* List all sections in all modules. The callback routine is invoked with + * token, module name, section name, section start, section end, section flags. + */ +int kallsyms_sections(void *token, + int (*callback)(void *, const char *, const char *, ElfW(Addr), ElfW(Addr), ElfW(Word))) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const char *ka_str; + const struct module *m; + int i; + + for (m = module_list; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) ((char *)ka_hdr + ka_hdr->section_off); + ka_str = ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (callback( + token, + *(m->name) ? m->name : "kernel", + ka_sec->name_off + ka_str, + ka_sec->start, + ka_sec->start + ka_sec->size, + ka_sec->flags)) + return(0); + } + } + return(1); +} --- linux/kernel/module.c.orig Wed Sep 25 10:49:32 2002 +++ linux/kernel/module.c Wed Sep 25 10:49:32 2002 @@ -1305,51 +1305,51 @@ #if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) -int lookup_symbol(unsigned long address, char *buffer, int buflen) +#define MAX_SYMBOL_SIZE 512 + +int print_symbol(const char *fmt, unsigned long address) { struct module *this_mod; unsigned long bestsofar; + /* static to not take up stackspace; if we race here too bad */ + static char buffer[MAX_SYMBOL_SIZE]; const char *mod_name = NULL, *sec_name = NULL, *sym_name = NULL; unsigned long mod_start, mod_end, sec_start, sec_end, sym_start, sym_end; - if (!buffer) - return -EFAULT; - - if (buflen < 256) - return -ENOMEM; - - memset(buffer, 0, buflen); + memset(buffer, 0, MAX_SYMBOL_SIZE); if (!kallsyms_address_to_symbol(address,&mod_name,&mod_start,&mod_end,&sec_name, &sec_start, &sec_end, &sym_name, &sym_start, &sym_end)) { - /* kallsyms doesn't have a clue; lets try harder */ + /* kallsyms doesn't have a clue; lets try our list + * of exported symbols */ bestsofar = 0; - snprintf(buffer, buflen-1, "[unresolved]"); + sprintf(buffer, "[unresolved]"); - this_mod = module_list; - - while (this_mod != NULL) { + for (this_mod = module_list; this_mod; this_mod = this_mod->next) { int i; /* walk the symbol list of this module. Only symbols who's address is smaller than the searched for address are relevant; and only if it's better than the best so far */ for (i = 0; i < this_mod->nsyms; i++) if ((this_mod->syms[i].value <= address) && - (bestsofar<this_mod->syms[i].value)) { - snprintf(buffer,buflen-1,"%s [%s] 0x%x", + (bestsofar < this_mod->syms[i].value)) { + snprintf(buffer, MAX_SYMBOL_SIZE - 1, + "%s [%s] 0x%x", this_mod->syms[i].name, this_mod->name, (unsigned int)(address - this_mod->syms[i].value)); bestsofar = this_mod->syms[i].value; } - this_mod = this_mod->next; } - } else /* kallsyms success */ - snprintf(buffer,buflen-1,"%s [%s] 0x%x",sym_name,mod_name,(unsigned int)(address-sym_start)); - - return strlen(buffer); + } else { /* kallsyms success */ + snprintf(buffer,MAX_SYMBOL_SIZE - 1, "%s [%s] 0x%x", + sym_name, mod_name, + (unsigned int)(address - sym_start)); + } + printk(fmt, buffer); + return 0; } #endif --- linux/Makefile.orig Wed Sep 25 10:49:32 2002 +++ linux/Makefile Wed Sep 25 10:49:32 2002 @@ -331,9 +331,13 @@ .tmp_kallsyms.o: .tmp_vmlinux $(call cmd,kallsyms) +# After generating .tmp_vmlinux just like vmlinux, decrement the version +# number again, so the final vmlinux gets the same one. +# Ignore return value of 'expr'. + define rule_.tmp_vmlinux $(rule_vmlinux) - expr 0`cat .version` - 1 > .tmp_version + if expr 0`cat .version` - 1 > .tmp_version; then true; fi mv -f .tmp_version .version endef ^ permalink raw reply [flat|nested] 23+ messages in thread
* [re-ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 9:02 [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 Ingo Molnar @ 2002-09-25 9:32 ` Ingo Molnar 2002-09-25 9:51 ` Arnaldo Carvalho de Melo 2002-09-25 18:16 ` [ANNOUNCE] " Linus Torvalds 1 sibling, 1 reply; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 9:32 UTC (permalink / raw) To: linux-kernel Cc: Linus Torvalds, Kai Germaschewski, Rusty Russell, Arjan van de Ven okay, this time the correct patch against BK-curr is included: diff -rNu linux.orig/Makefile linux/Makefile --- linux.orig/Makefile Wed Sep 25 11:30:13 2002 +++ linux/Makefile Wed Sep 25 11:26:11 2002 @@ -138,6 +138,7 @@ MAKEFILES = $(TOPDIR)/.config GENKSYMS = /sbin/genksyms DEPMOD = /sbin/depmod +KALLSYMS = /sbin/kallsyms PERL = perl MODFLAGS = -DMODULE CFLAGS_MODULE = $(MODFLAGS) @@ -291,32 +292,64 @@ vmlinux-objs := $(HEAD) $(INIT) $(CORE_FILES) $(LIBS) $(DRIVERS) $(NETWORKS) quiet_cmd_link_vmlinux = LD $@ -cmd_link_vmlinux = $(LD) $(LDFLAGS) $(LDFLAGS_$(@F)) $(HEAD) $(INIT) \ - --start-group \ - $(CORE_FILES) \ - $(LIBS) \ - $(DRIVERS) \ - $(NETWORKS) \ - --end-group \ - -o vmlinux +define cmd_link_vmlinux + $(LD) $(LDFLAGS) $(LDFLAGS_vmlinux) $(HEAD) $(INIT) \ + --start-group \ + $(CORE_FILES) \ + $(LIBS) \ + $(DRIVERS) \ + $(NETWORKS) \ + --end-group \ + $(filter $(kallsyms.o),$^) \ + -o $@ +endef # set -e makes the rule exit immediately on error -define rule_link_vmlinux +define rule_vmlinux set -e echo ' Generating build number' - . scripts/mkversion > .tmpversion - mv -f .tmpversion .version + . scripts/mkversion > .tmp_version + mv -f .tmp_version .version +$(MAKE) -C init $(call cmd,link_vmlinux) echo 'cmd_$@ := $(cmd_link_vmlinux)' > $(@D)/.$(@F).cmd - $(NM) vmlinux | grep -v '\(compiled\)\|\(\.o$$\)\|\( [aUw] \)\|\(\.\.ng$$\)\|\(LASH[RL]DI\)' | sort > System.map + $(NM) $@ | grep -v '\(compiled\)\|\(\.o$$\)\|\( [aUw] \)\|\(\.\.ng$$\)\|\(LASH[RL]DI\)' | sort > System.map endef LDFLAGS_vmlinux += -T arch/$(ARCH)/vmlinux.lds.s -vmlinux: $(vmlinux-objs) arch/$(ARCH)/vmlinux.lds.s FORCE - $(call if_changed_rule,link_vmlinux) +# Generate section listing all symbols and add it into vmlinux + +ifdef CONFIG_KALLSYMS + +kallsyms.o := .tmp_kallsyms.o + +quiet_cmd_kallsyms = KSYM $@ +cmd_kallsyms = $(KALLSYMS) $< > $@ + +.tmp_kallsyms.o: .tmp_vmlinux + $(call cmd,kallsyms) + +# After generating .tmp_vmlinux just like vmlinux, decrement the version +# number again, so the final vmlinux gets the same one. +# Ignore return value of 'expr'. + +define rule_.tmp_vmlinux + $(rule_vmlinux) + if expr 0`cat .version` - 1 > .tmp_version; then true; fi + mv -f .tmp_version .version +endef + +.tmp_vmlinux: $(vmlinux-objs) arch/$(ARCH)/vmlinux.lds.s FORCE + $(call if_changed_rule,.tmp_vmlinux) + +endif + +# Finally the vmlinux rule + +vmlinux: $(vmlinux-objs) $(kallsyms.o) arch/$(ARCH)/vmlinux.lds.s FORCE + $(call if_changed_rule,vmlinux) # The actual objects are generated when descending, # make sure no implicit rule kicks in @@ -820,7 +853,7 @@ # FIXME Should go into a make.lib or something # =========================================================================== -echo_target = $(RELDIR)/$@ +echo_target = $@ a_flags = -Wp,-MD,$(depfile) $(AFLAGS) $(NOSTDINC_FLAGS) \ $(modkern_aflags) $(EXTRA_AFLAGS) $(AFLAGS_$(*F).o) diff -rNu linux.orig/arch/i386/Config.help linux/arch/i386/Config.help --- linux.orig/arch/i386/Config.help Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/Config.help Wed Sep 25 11:26:11 2002 @@ -946,6 +946,11 @@ of the BUG call as well as the EIP and oops trace. This aids debugging but costs about 70-100K of memory. +CONFIG_KALLSYMS + Say Y here to let the kernel print out symbolic crash information and + symbolic stack backtraces. This increases the size of the kernel + somewhat, as all symbols have to be loaded into the kernel image. + CONFIG_DEBUG_OBSOLETE Say Y here if you want to reduce the chances of the tree compiling, and are prepared to dig into driver internals to fix compile errors. diff -rNu linux.orig/arch/i386/config.in linux/arch/i386/config.in --- linux.orig/arch/i386/config.in Wed Sep 25 11:30:13 2002 +++ linux/arch/i386/config.in Wed Sep 25 11:26:11 2002 @@ -435,6 +435,7 @@ if [ "$CONFIG_HIGHMEM" = "y" ]; then bool ' Highmem debugging' CONFIG_DEBUG_HIGHMEM fi + bool ' Load all symbols for debugging/kksymoops' CONFIG_KALLSYMS fi if [ "$CONFIG_X86_LOCAL_APIC" = "y" ]; then diff -rNu linux.orig/arch/i386/kernel/head.S linux/arch/i386/kernel/head.S --- linux.orig/arch/i386/kernel/head.S Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/kernel/head.S Wed Sep 25 11:26:11 2002 @@ -121,7 +121,7 @@ */ xorl %eax,%eax movl $__bss_start,%edi - movl $_end,%ecx + movl $__bss_stop,%ecx subl %edi,%ecx rep stosb diff -rNu linux.orig/arch/i386/kernel/process.c linux/arch/i386/kernel/process.c --- linux.orig/arch/i386/kernel/process.c Fri Sep 20 17:20:12 2002 +++ linux/arch/i386/kernel/process.c Wed Sep 25 11:26:11 2002 @@ -33,6 +33,7 @@ #include <linux/reboot.h> #include <linux/init.h> #include <linux/mc146818rtc.h> +#include <linux/module.h> #include <asm/uaccess.h> #include <asm/pgtable.h> @@ -155,8 +156,6 @@ __setup("idle=", idle_setup); -extern void show_trace(unsigned long* esp); - void show_regs(struct pt_regs * regs) { unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; @@ -164,6 +163,8 @@ printk("\n"); printk("Pid: %d, comm: %20s\n", current->pid, current->comm); printk("EIP: %04x:[<%08lx>] CPU: %d",0xffff & regs->xcs,regs->eip, smp_processor_id()); + print_symbol("EIP is at %s\n", regs->eip); + if (regs->xcs & 3) printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp); printk(" EFLAGS: %08lx %s\n",regs->eflags, print_tainted()); diff -rNu linux.orig/arch/i386/kernel/process.c.rej linux/arch/i386/kernel/process.c.rej --- linux.orig/arch/i386/kernel/process.c.rej Thu Jan 1 01:00:00 1970 +++ linux/arch/i386/kernel/process.c.rej Wed Sep 25 11:25:33 2002 @@ -0,0 +1,29 @@ +*************** +*** 159,172 **** + void show_regs(struct pt_regs * regs) + { + unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; +- static char buffer[MAX_SYMBOL_SIZE]; + + printk("\n"); + printk("Pid: %d, comm: %20s\n", current->pid, current->comm); + printk("EIP: %04x:[<%08lx>] CPU: %d",0xffff & regs->xcs,regs->eip, smp_processor_id()); + +- lookup_symbol(regs->eip, buffer, MAX_SYMBOL_SIZE); +- printk("\nEIP is at %s \n", buffer); + if (regs->xcs & 3) + printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp); + printk(" EFLAGS: %08lx %s\n",regs->eflags, print_tainted()); +--- 159,170 ---- + void show_regs(struct pt_regs * regs) + { + unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; + + printk("\n"); + printk("Pid: %d, comm: %20s\n", current->pid, current->comm); + printk("EIP: %04x:[<%08lx>] CPU: %d",0xffff & regs->xcs,regs->eip, smp_processor_id()); ++ print_symbol("EIP is at %s\n", regs->eip); + + if (regs->xcs & 3) + printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp); + printk(" EFLAGS: %08lx %s\n",regs->eflags, print_tainted()); diff -rNu linux.orig/arch/i386/kernel/traps.c linux/arch/i386/kernel/traps.c --- linux.orig/arch/i386/kernel/traps.c Fri Sep 20 17:20:19 2002 +++ linux/arch/i386/kernel/traps.c Wed Sep 25 11:26:11 2002 @@ -94,7 +94,6 @@ #ifdef CONFIG_MODULES -extern struct module *module_list; extern struct module kernel_module; static inline int kernel_text_address(unsigned long addr) @@ -142,10 +141,12 @@ while (((long) stack & (THREAD_SIZE-1)) != 0) { addr = *stack++; if (kernel_text_address(addr)) { - if (i && ((i % 6) == 0)) - printk("\n "); printk("[<%08lx>] ", addr); - i++; + if (print_symbol("%s\n", addr)) { + /* save screen space */ + if ((i++ % 6) == 0) + printk("\n "); + } } } printk("\n"); @@ -206,8 +207,11 @@ esp = regs->esp; ss = regs->xss & 0xffff; } + print_modules(); printk("CPU: %d\nEIP: %04x:[<%08lx>] %s\nEFLAGS: %08lx\n", smp_processor_id(), 0xffff & regs->xcs, regs->eip, print_tainted(), regs->eflags); + + print_symbol("EIP is at %s\n", regs->eip); printk("eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n", regs->eax, regs->ebx, regs->ecx, regs->edx); printk("esi: %08lx edi: %08lx ebp: %08lx esp: %08lx\n", @@ -268,6 +272,7 @@ (unsigned long)file < PAGE_OFFSET || __get_user(c, file)) file = "<bad filename>"; + printk("------------[ cut here ]------------\n"); printk("kernel BUG at %s:%d!\n", file, line); no_bug: diff -rNu linux.orig/arch/i386/vmlinux.lds.S linux/arch/i386/vmlinux.lds.S --- linux.orig/arch/i386/vmlinux.lds.S Fri Sep 20 17:20:19 2002 +++ linux/arch/i386/vmlinux.lds.S Wed Sep 25 11:26:11 2002 @@ -78,9 +78,13 @@ .data.cacheline_aligned : { *(.data.cacheline_aligned) } __bss_start = .; /* BSS */ - .bss : { - *(.bss) - } + .bss : { *(.bss) } + __bss_stop = .; + + __start___kallsyms = .; /* All kernel symbols */ + __kallsyms : { *(__kallsyms) } + __stop___kallsyms = .; + _end = . ; /* Sections to be discarded */ diff -rNu linux.orig/include/asm-i386/hardirq.h linux/include/asm-i386/hardirq.h --- linux.orig/include/asm-i386/hardirq.h Fri Sep 20 17:20:32 2002 +++ linux/include/asm-i386/hardirq.h Wed Sep 25 11:26:11 2002 @@ -97,6 +97,4 @@ extern void synchronize_irq(unsigned int irq); #endif /* CONFIG_SMP */ -extern void show_stack(unsigned long * esp); - #endif /* __ASM_HARDIRQ_H */ diff -rNu linux.orig/include/asm-i386/ptrace.h linux/include/asm-i386/ptrace.h --- linux.orig/include/asm-i386/ptrace.h Fri Sep 20 17:20:16 2002 +++ linux/include/asm-i386/ptrace.h Wed Sep 25 11:26:11 2002 @@ -57,7 +57,6 @@ #ifdef __KERNEL__ #define user_mode(regs) ((VM_MASK & (regs)->eflags) || (3 & (regs)->xcs)) #define instruction_pointer(regs) ((regs)->eip) -extern void show_regs(struct pt_regs *); #endif #endif diff -rNu linux.orig/include/linux/kallsyms.h linux/include/linux/kallsyms.h --- linux.orig/include/linux/kallsyms.h Thu Jan 1 01:00:00 1970 +++ linux/include/linux/kallsyms.h Wed Sep 25 11:26:11 2002 @@ -0,0 +1,163 @@ +/* kallsyms headers + Copyright 2000 Keith Owens <kaos@ocs.com.au> + + This file is part of the Linux modutils. It is exported to kernel + space so debuggers can access the kallsyms data. + + The kallsyms data contains all the non-stack symbols from a kernel + or a module. The kernel symbols are held between __start___kallsyms + and __stop___kallsyms. The symbols for a module are accessed via + the struct module chain which is based at module_list. + + This program is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by the + Free Software Foundation; either version 2 of the License, or (at your + option) any later version. + + This program is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software Foundation, + Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + */ + +#ident "$Id: linux-2.4.9-kallsyms.patch,v 1.8 2002/02/11 18:34:53 arjanv Exp $" + +#ifndef MODUTILS_KALLSYMS_H +#define MODUTILS_KALLSYMS_H 1 + +/* Have to (re)define these ElfW entries here because external kallsyms + * code does not have access to modutils/include/obj.h. This code is + * included from user spaces tools (modutils) and kernel, they need + * different includes. + */ + +#ifndef ELFCLASS32 +#ifdef __KERNEL__ +#include <linux/elf.h> +#else /* __KERNEL__ */ +#include <elf.h> +#endif /* __KERNEL__ */ +#endif /* ELFCLASS32 */ + +#ifndef ELFCLASSM +#define ELFCLASSM ELF_CLASS +#endif + +#ifndef ElfW +# if ELFCLASSM == ELFCLASS32 +# define ElfW(x) Elf32_ ## x +# define ELFW(x) ELF32_ ## x +# else +# define ElfW(x) Elf64_ ## x +# define ELFW(x) ELF64_ ## x +# endif +#endif + +/* Format of data in the kallsyms section. + * Most of the fields are small numbers but the total size and all + * offsets can be large so use the 32/64 bit types for these fields. + * + * Do not use sizeof() on these structures, modutils may be using extra + * fields. Instead use the size fields in the header to access the + * other bits of data. + */ + +struct kallsyms_header { + int size; /* Size of this header */ + ElfW(Word) total_size; /* Total size of kallsyms data */ + int sections; /* Number of section entries */ + ElfW(Off) section_off; /* Offset to first section entry */ + int section_size; /* Size of one section entry */ + int symbols; /* Number of symbol entries */ + ElfW(Off) symbol_off; /* Offset to first symbol entry */ + int symbol_size; /* Size of one symbol entry */ + ElfW(Off) string_off; /* Offset to first string */ + ElfW(Addr) start; /* Start address of first section */ + ElfW(Addr) end; /* End address of last section */ +}; + +struct kallsyms_section { + ElfW(Addr) start; /* Start address of section */ + ElfW(Word) size; /* Size of this section */ + ElfW(Off) name_off; /* Offset to section name */ + ElfW(Word) flags; /* Flags from section */ +}; + +struct kallsyms_symbol { + ElfW(Off) section_off; /* Offset to section that owns this symbol */ + ElfW(Addr) symbol_addr; /* Address of symbol */ + ElfW(Off) name_off; /* Offset to symbol name */ +}; + +#define KALLSYMS_SEC_NAME "__kallsyms" +#define KALLSYMS_IDX 2 /* obj_kallsyms creates kallsyms as section 2 */ + +#define kallsyms_next_sec(h,s) \ + ((s) = (struct kallsyms_section *)((char *)(s) + (h)->section_size)) +#define kallsyms_next_sym(h,s) \ + ((s) = (struct kallsyms_symbol *)((char *)(s) + (h)->symbol_size)) + +#ifdef CONFIG_KALLSYMS + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start with */ + const char **mod_name, /* Set to module name or "kernel" */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_sections(void *token, + int (*callback)(void *, /* token */ + const char *, /* module name */ + const char *, /* section name */ + ElfW(Addr), /* Section start */ + ElfW(Addr), /* Section end */ + ElfW(Word) /* Section flags */ + ) + ); + +#else + +static inline int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + return -ESRCH; +} + +#endif + +#endif /* kallsyms.h */ diff -rNu linux.orig/include/linux/module.h linux/include/linux/module.h --- linux.orig/include/linux/module.h Fri Sep 20 17:20:32 2002 +++ linux/include/linux/module.h Wed Sep 25 11:26:11 2002 @@ -316,8 +316,6 @@ #define MOD_DEC_USE_COUNT do { } while (0) #define MOD_IN_USE 1 -extern struct module *module_list; - #endif /* !__GENKSYMS__ */ #endif /* MODULE */ @@ -504,6 +502,30 @@ #define SET_MODULE_OWNER(some_struct) do { (some_struct)->owner = THIS_MODULE; } while (0) #else #define SET_MODULE_OWNER(some_struct) do { } while (0) +#endif + +extern void print_modules(void); + +#if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) + +extern struct module *module_list; + +/* + * print_symbols takes a format string containing one %s. + * If support for resolving symbols is compiled in, the %s will + * be replaced by the closest symbol to the address and the entire + * string is printk()ed. Otherwise, nothing is printed. + */ +extern int print_symbol(const char *fmt, unsigned long address); + +#else + +static inline int +print_symbol(const char *fmt, unsigned long address) +{ + return -ESRCH; +} + #endif #endif /* _LINUX_MODULE_H */ diff -rNu linux.orig/include/linux/sched.h linux/include/linux/sched.h --- linux.orig/include/linux/sched.h Wed Sep 25 11:30:14 2002 +++ linux/include/linux/sched.h Wed Sep 25 11:26:11 2002 @@ -151,7 +151,13 @@ extern void sched_init(void); extern void init_idle(task_t *idle, int cpu); + extern void show_state(void); +extern void show_trace(unsigned long *stack); +extern void show_stack(unsigned long *stack); +extern void show_regs(struct pt_regs *); + + extern void cpu_init (void); extern void trap_init(void); extern void update_process_times(int user); diff -rNu linux.orig/kernel/Makefile linux/kernel/Makefile --- linux.orig/kernel/Makefile Fri Sep 20 17:20:19 2002 +++ linux/kernel/Makefile Wed Sep 25 11:26:11 2002 @@ -3,7 +3,7 @@ # export-objs = signal.o sys.o kmod.o context.o ksyms.o pm.o exec_domain.o \ - printk.o platform.o suspend.o dma.o + printk.o platform.o suspend.o dma.o module.o obj-y = sched.o fork.o exec_domain.o panic.o printk.o \ module.o exit.o itimer.o time.o softirq.o resource.o \ @@ -14,6 +14,7 @@ obj-$(CONFIG_SMP) += cpu.o obj-$(CONFIG_UID16) += uid16.o obj-$(CONFIG_MODULES) += ksyms.o +obj-$(CONFIG_KALLSYMS) += kallsyms.o obj-$(CONFIG_PM) += pm.o obj-$(CONFIG_BSD_PROCESS_ACCT) += acct.o obj-$(CONFIG_SOFTWARE_SUSPEND) += suspend.o diff -rNu linux.orig/kernel/kallsyms.c linux/kernel/kallsyms.c --- linux.orig/kernel/kallsyms.c Thu Jan 1 01:00:00 1970 +++ linux/kernel/kallsyms.c Wed Sep 25 11:26:11 2002 @@ -0,0 +1,227 @@ +/* + * kksymoops.c: in-kernel printing of symbolic oopses and stack traces. + * + * Copyright 2000 Keith Owens <kaos@ocs.com.au> April 2000 + * Copyright 2002 Arjan van de Ven <arjanv@redhat.com> + * + This code uses the list of all kernel and module symbols to :- + + * Find any non-stack symbol in a kernel or module. Symbols do + not have to be exported for debugging. + + * Convert an address to the module (or kernel) that owns it, the + section it is in and the nearest symbol. This finds all non-stack + symbols, not just exported ones. + + */ + +#include <linux/mm.h> +#include <linux/module.h> +#include <linux/kallsyms.h> + +/* A symbol can appear in more than one module. A token is used to + * restart the scan at the next module, set the token to 0 for the + * first scan of each symbol. + */ + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start at */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec; + const struct kallsyms_symbol *ka_sym = NULL; + const char *ka_str = NULL; + const struct module *m; + int i = 0, l; + const char *p, *pt_R; + char *p2; + + /* Restart? */ + m = module_list; + if (token && *token) { + for (; m; m = m->next) + if ((unsigned long)m == *token) + break; + if (m) + m = m->next; + } + + for (; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + p = ka_str + ka_sym->name_off; + if (strcmp(p, name) == 0) + break; + /* Unversioned requests match versioned names */ + if (!(pt_R = strstr(p, "_R"))) + continue; + l = strlen(pt_R); + if (l < 10) + continue; /* Not _R.*xxxxxxxx */ + (void)simple_strtoul(pt_R+l-8, &p2, 16); + if (*p2) + continue; /* Not _R.*xxxxxxxx */ + if (strncmp(p, name, pt_R-p) == 0) + break; /* Match with version */ + } + if (i < ka_hdr->symbols) + break; + } + + if (token) + *token = (unsigned long)m; + if (!m) + return(0); /* not found */ + + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off); + *mod_name = *(m->name) ? m->name : "kernel"; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = ka_sym->name_off + ka_str; + *sym_start = ka_sym->symbol_addr; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + *sym_end = ka_symn->symbol_addr; + } + else + *sym_end = *sec_end; + return(1); +} + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const struct kallsyms_symbol *ka_sym; + const char *ka_str; + const struct module *m; + int i; + unsigned long end; + + for (m = module_list; m; m = m->next) { + + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off); + /* Is the address in any section in this module? */ + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (ka_sec->start <= address && + (ka_sec->start + ka_sec->size) > address) + break; + } + if (i < ka_hdr->sections) + break; /* Found a matching section */ + } + + if (!m) + return(0); /* not found */ + + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + *mod_name = m->name; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = *sec_name; /* In case we find no matching symbol */ + *sym_start = *sec_start; + *sym_end = *sec_end; + + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + if (ka_sym->symbol_addr > address) + continue; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + end = ka_symn->symbol_addr; + } + else + end = *sec_end; + if (end <= address) + continue; + if ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off + != (char *)ka_sec) + continue; /* wrong section */ + *sym_name = ka_str + ka_sym->name_off; + *sym_start = ka_sym->symbol_addr; + *sym_end = end; + break; + } + return(1); +} + +/* List all sections in all modules. The callback routine is invoked with + * token, module name, section name, section start, section end, section flags. + */ +int kallsyms_sections(void *token, + int (*callback)(void *, const char *, const char *, ElfW(Addr), ElfW(Addr), ElfW(Word))) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const char *ka_str; + const struct module *m; + int i; + + for (m = module_list; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) ((char *)ka_hdr + ka_hdr->section_off); + ka_str = ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (callback( + token, + *(m->name) ? m->name : "kernel", + ka_sec->name_off + ka_str, + ka_sec->start, + ka_sec->start + ka_sec->size, + ka_sec->flags)) + return(0); + } + } + return(1); +} diff -rNu linux.orig/kernel/module.c linux/kernel/module.c --- linux.orig/kernel/module.c Fri Sep 20 17:20:19 2002 +++ linux/kernel/module.c Wed Sep 25 11:26:11 2002 @@ -3,6 +3,7 @@ #include <linux/module.h> #include <asm/module.h> #include <asm/uaccess.h> +#include <linux/kallsyms.h> #include <linux/vmalloc.h> #include <linux/smp_lock.h> #include <asm/pgalloc.h> @@ -39,13 +40,19 @@ extern const struct exception_table_entry __start___ex_table[]; extern const struct exception_table_entry __stop___ex_table[]; -extern const char __start___kallsyms[] __attribute__ ((weak)); -extern const char __stop___kallsyms[] __attribute__ ((weak)); +extern const char __start___kallsyms[] __attribute__((weak)); +extern const char __stop___kallsyms[] __attribute__((weak)); + +/* modutils uses these exported symbols to figure out if + kallsyms support is present */ + +EXPORT_SYMBOL(__start___kallsyms); +EXPORT_SYMBOL(__stop___kallsyms); struct module kernel_module = { .size_of_struct = sizeof(struct module), - .name = "", + .name = "kernel", .uc = {ATOMIC_INIT(1)}, .flags = MOD_RUNNING, .syms = __start___ksymtab, @@ -1220,6 +1227,30 @@ .show = s_show }; +#define MODLIST_SIZE 4096 + +/* + * this function isn't smp safe but that's not really a problem; it's + * called from oops context only and any locking could actually prevent + * the oops from going out; the line that is generated is informational + * only and should NEVER prevent the real oops from going out. + */ +void print_modules(void) +{ + static char modlist[MODLIST_SIZE]; + struct module *this_mod; + int pos = 0; + + this_mod = module_list; + while (this_mod) { + if (this_mod->name) + pos += snprintf(modlist+pos, MODLIST_SIZE-pos-1, + "%s ", this_mod->name); + this_mod = this_mod->next; + } + printk("%s\n",modlist); +} + #else /* CONFIG_MODULES */ /* Dummy syscalls for people who don't want modules */ @@ -1265,4 +1296,60 @@ return 1; } +void print_modules(void) +{ +} + #endif /* CONFIG_MODULES */ + + +#if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) + +#define MAX_SYMBOL_SIZE 512 + +int print_symbol(const char *fmt, unsigned long address) +{ + struct module *this_mod; + unsigned long bestsofar; + /* static to not take up stackspace; if we race here too bad */ + static char buffer[MAX_SYMBOL_SIZE]; + + const char *mod_name = NULL, *sec_name = NULL, *sym_name = NULL; + unsigned long mod_start, mod_end, sec_start, sec_end, + sym_start, sym_end; + + memset(buffer, 0, MAX_SYMBOL_SIZE); + + if (!kallsyms_address_to_symbol(address,&mod_name,&mod_start,&mod_end,&sec_name, &sec_start, &sec_end, &sym_name, &sym_start, &sym_end)) { + /* kallsyms doesn't have a clue; lets try our list + * of exported symbols */ + bestsofar = 0; + sprintf(buffer, "[unresolved]"); + + for (this_mod = module_list; this_mod; this_mod = this_mod->next) { + int i; + /* walk the symbol list of this module. Only symbols + who's address is smaller than the searched for address + are relevant; and only if it's better than the best so far */ + for (i = 0; i < this_mod->nsyms; i++) + if ((this_mod->syms[i].value <= address) && + (bestsofar < this_mod->syms[i].value)) { + snprintf(buffer, MAX_SYMBOL_SIZE - 1, + "%s [%s] 0x%x", + this_mod->syms[i].name, + this_mod->name, + (unsigned int)(address - this_mod->syms[i].value)); + bestsofar = this_mod->syms[i].value; + } + } + + } else { /* kallsyms success */ + snprintf(buffer,MAX_SYMBOL_SIZE - 1, "%s [%s] 0x%x", + sym_name, mod_name, + (unsigned int)(address - sym_start)); + } + printk(fmt, buffer); + return 0; +} + +#endif ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [re-ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 9:32 ` [re-ANNOUNCE] " Ingo Molnar @ 2002-09-25 9:51 ` Arnaldo Carvalho de Melo 0 siblings, 0 replies; 23+ messages in thread From: Arnaldo Carvalho de Melo @ 2002-09-25 9:51 UTC (permalink / raw) To: Ingo Molnar Cc: linux-kernel, Linus Torvalds, Kai Germaschewski, Rusty Russell, Arjan van de Ven Em Wed, Sep 25, 2002 at 11:32:24AM +0200, Ingo Molnar escreveu: > > okay, this time the correct patch against BK-curr is included: <SNIP> > diff -rNu linux.orig/arch/i386/kernel/process.c.rej linux/arch/i386/kernel/process.c.rej > --- linux.orig/arch/i386/kernel/process.c.rej Thu Jan 1 01:00:00 1970 > +++ linux/arch/i386/kernel/process.c.rej Wed Sep 25 11:25:33 2002 ^^^^ ^^^^ ^^^^ Oops :-) > @@ -0,0 +1,29 @@ > +*************** > +*** 159,172 **** > + void show_regs(struct pt_regs * regs) > + { > + unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 9:02 [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 Ingo Molnar 2002-09-25 9:32 ` [re-ANNOUNCE] " Ingo Molnar @ 2002-09-25 18:16 ` Linus Torvalds 2002-09-25 19:23 ` Kai Germaschewski 1 sibling, 1 reply; 23+ messages in thread From: Linus Torvalds @ 2002-09-25 18:16 UTC (permalink / raw) To: Ingo Molnar Cc: linux-kernel, Kai Germaschewski, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Ingo Molnar wrote: > > EIP is at sys_gettimeofday [kernel] 0x84 > Call Trace: [<c0112a40>] do_page_fault [kernel] 0x0 > [<c0107973>] syscall_call [kernel] 0x7 At a minimum, fix this to: - not print out that stupid "kernel" thing. Of _course_ it is the kernel. Modules can put their module name to clarify, but the core kernel certainly doesn't "clarify" anything by putting "kernel" there. - print out offset/length like the user-space ksymoops does. Without the offset the thing is useless, since you still need the real address to actually look up which instruction faulted. Linus ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 18:16 ` [ANNOUNCE] " Linus Torvalds @ 2002-09-25 19:23 ` Kai Germaschewski 2002-09-25 19:42 ` Ingo Molnar 2002-09-25 19:45 ` Linus Torvalds 0 siblings, 2 replies; 23+ messages in thread From: Kai Germaschewski @ 2002-09-25 19:23 UTC (permalink / raw) To: Linus Torvalds; +Cc: Ingo Molnar, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Linus Torvalds wrote: > On Wed, 25 Sep 2002, Ingo Molnar wrote: > > > > EIP is at sys_gettimeofday [kernel] 0x84 > > Call Trace: [<c0112a40>] do_page_fault [kernel] 0x0 > > [<c0107973>] syscall_call [kernel] 0x7 > > At a minimum, fix this to: > > - not print out that stupid "kernel" thing. Of _course_ it is the kernel. > Modules can put their module name to clarify, but the core kernel > certainly doesn't "clarify" anything by putting "kernel" there. > > - print out offset/length like the user-space ksymoops does. Without the > offset the thing is useless, since you still need the real address to > actually look up which instruction faulted. Alright, I did these modifications (however, even the current version prints the offset, just not the length, which does not have too much meaning anyway) So this patch converts the output to the more familiar sys_gettimeofday+0x84/0x108 [module-name] format, where module-name is "" for the kernel. Ingo, could you test the modifications again, please? Also, I worked against my latest version, if you did additional changes after the last version I sent you, they need to be merged back in. --Kai Pull from http://linux-isdn.bkbits.net/linux-2.5.kallsyms (Merging changesets omitted for clarity) ----------------------------------------------------------------------------- ChangeSet@1.609, 2002-09-25 14:11:06-05:00, kai@tp1.ruhr-uni-bochum.de kksymoops: cosmetics Don't print "kernel" when a symbol is in the kernel. Change the output format to sys_gettimeofday+0x84/0x108 [module]. ---------------------------------------------------------------------------- module.c | 75 ++++++++++++++++++++++++++++++++++++--------------------------- 1 files changed, 44 insertions(+), 31 deletions(-) ============================================================================= unified diffs follow for reference ============================================================================= ----------------------------------------------------------------------------- ChangeSet@1.609, 2002-09-25 14:11:06-05:00, kai@tp1.ruhr-uni-bochum.de kksymoops: cosmetics Don't print "kernel" when a symbol is in the kernel. Change the output format to sys_gettimeofday+0x84/0x108 [module]. --------------------------------------------------------------------------- diff -Nru a/kernel/module.c b/kernel/module.c --- a/kernel/module.c Wed Sep 25 14:18:41 2002 +++ b/kernel/module.c Wed Sep 25 14:18:41 2002 @@ -52,7 +52,7 @@ struct module kernel_module = { .size_of_struct = sizeof(struct module), - .name = "kernel", + .name = "", .uc = {ATOMIC_INIT(1)}, .flags = MOD_RUNNING, .syms = __start___ksymtab, @@ -1307,48 +1307,61 @@ #define MAX_SYMBOL_SIZE 512 -int print_symbol(const char *fmt, unsigned long address) +static void +address_to_exported_symbol(unsigned long address, const char **mod_name, + const char **sym_name, unsigned long *sym_start, + unsigned long *sym_end) { struct module *this_mod; - unsigned long bestsofar; + int i; + + for (this_mod = module_list; this_mod; this_mod = this_mod->next) { + /* walk the symbol list of this module. Only symbols + who's address is smaller than the searched for address + are relevant; and only if it's better than the best so far */ + for (i = 0; i < this_mod->nsyms; i++) + if ((this_mod->syms[i].value <= address) && + (*sym_start < this_mod->syms[i].value)) { + *sym_start = this_mod->syms[i].value; + *sym_name = this_mod->syms[i].name; + *mod_name = this_mod->name; + if (i + 1 < this_mod->nsyms) + *sym_end = this_mod->syms[i+1].value; + else + *sym_end = (unsigned long) this_mod + this_mod->size; + } + } +} + +int +print_symbol(const char *fmt, unsigned long address) +{ /* static to not take up stackspace; if we race here too bad */ static char buffer[MAX_SYMBOL_SIZE]; const char *mod_name = NULL, *sec_name = NULL, *sym_name = NULL; unsigned long mod_start, mod_end, sec_start, sec_end, - sym_start, sym_end; + sym_start, sym_end; + char *tag = ""; memset(buffer, 0, MAX_SYMBOL_SIZE); - if (!kallsyms_address_to_symbol(address,&mod_name,&mod_start,&mod_end,&sec_name, &sec_start, &sec_end, &sym_name, &sym_start, &sym_end)) { - /* kallsyms doesn't have a clue; lets try our list - * of exported symbols */ - bestsofar = 0; - sprintf(buffer, "[unresolved]"); - - for (this_mod = module_list; this_mod; this_mod = this_mod->next) { - int i; - /* walk the symbol list of this module. Only symbols - who's address is smaller than the searched for address - are relevant; and only if it's better than the best so far */ - for (i = 0; i < this_mod->nsyms; i++) - if ((this_mod->syms[i].value <= address) && - (bestsofar < this_mod->syms[i].value)) { - snprintf(buffer, MAX_SYMBOL_SIZE - 1, - "%s [%s] 0x%x", - this_mod->syms[i].name, - this_mod->name, - (unsigned int)(address - this_mod->syms[i].value)); - bestsofar = this_mod->syms[i].value; - } - } + sym_start = 0; + if (!kallsyms_address_to_symbol(address, &mod_name, &mod_start, &mod_end, &sec_name, &sec_start, &sec_end, &sym_name, &sym_start, &sym_end)) { + tag = "E "; + address_to_exported_symbol(address, &mod_name, &sym_name, &sym_start, &sym_end); + } - } else { /* kallsyms success */ - snprintf(buffer,MAX_SYMBOL_SIZE - 1, "%s [%s] 0x%x", - sym_name, mod_name, - (unsigned int)(address - sym_start)); + if (sym_start) { + snprintf(buffer, MAX_SYMBOL_SIZE - 1, "%s%s+%#x/%#x [%s]", + tag, sym_name, + (unsigned int)(address - sym_start), + (unsigned int)(sym_end - sym_start), + mod_name); + printk(fmt, buffer); + } else { + printk(fmt, "[unresolved]"); } - printk(fmt, buffer); return 0; } ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:23 ` Kai Germaschewski @ 2002-09-25 19:42 ` Ingo Molnar 2002-09-25 19:41 ` Kai Germaschewski 2002-09-25 19:46 ` Linus Torvalds 2002-09-25 19:45 ` Linus Torvalds 1 sibling, 2 replies; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 19:42 UTC (permalink / raw) To: Kai Germaschewski Cc: Linus Torvalds, linux-kernel, Rusty Russell, Arjan van de Ven one unrelated build thing that doesnt work is arch/i386/defconfig. If i add a CONFIG_KALLSYMS=y line to its debug section, and remove the CONFIG_KALLSYMS line from the .config, and then do a 'make oldconfig', i get this: Kernel debugging (CONFIG_DEBUG_KERNEL) [Y/n/?] Debug memory allocations (CONFIG_DEBUG_SLAB) [N/y/?] Memory mapped I/O debugging (CONFIG_DEBUG_IOVIRT) [N/y/?] Magic SysRq key (CONFIG_MAGIC_SYSRQ) [Y/n/?] Spinlock debugging (CONFIG_DEBUG_SPINLOCK) [N/y/?] Load all symbols for debugging/kksymoops (CONFIG_KALLSYMS) [N/y/?] (NEW) i'd expect the 'Y' to be picked up from the defconfig - no? Ingo ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:42 ` Ingo Molnar @ 2002-09-25 19:41 ` Kai Germaschewski 2002-09-25 19:46 ` Linus Torvalds 1 sibling, 0 replies; 23+ messages in thread From: Kai Germaschewski @ 2002-09-25 19:41 UTC (permalink / raw) To: Ingo Molnar; +Cc: Linus Torvalds, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Ingo Molnar wrote: > one unrelated build thing that doesnt work is arch/i386/defconfig. If i > add a CONFIG_KALLSYMS=y line to its debug section, and remove the > CONFIG_KALLSYMS line from the .config, and then do a 'make oldconfig', i > get this: > > Kernel debugging (CONFIG_DEBUG_KERNEL) [Y/n/?] > Debug memory allocations (CONFIG_DEBUG_SLAB) [N/y/?] > Memory mapped I/O debugging (CONFIG_DEBUG_IOVIRT) [N/y/?] > Magic SysRq key (CONFIG_MAGIC_SYSRQ) [Y/n/?] > Spinlock debugging (CONFIG_DEBUG_SPINLOCK) [N/y/?] > Load all symbols for debugging/kksymoops (CONFIG_KALLSYMS) [N/y/?] (NEW) > > i'd expect the 'Y' to be picked up from the defconfig - no? No. I don't think the behavior you describe ever existed. If you do "rm .config; make oldconfig" then it'll get picked up. --Kai ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:42 ` Ingo Molnar 2002-09-25 19:41 ` Kai Germaschewski @ 2002-09-25 19:46 ` Linus Torvalds 2002-09-25 19:54 ` Ingo Molnar 1 sibling, 1 reply; 23+ messages in thread From: Linus Torvalds @ 2002-09-25 19:46 UTC (permalink / raw) To: Ingo Molnar Cc: Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Ingo Molnar wrote: > > i'd expect the 'Y' to be picked up from the defconfig - no? No. defconfig is either used 100% or not at all. Linus ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:46 ` Linus Torvalds @ 2002-09-25 19:54 ` Ingo Molnar 2002-09-25 19:52 ` Linus Torvalds 2002-09-25 19:55 ` Kai Germaschewski 0 siblings, 2 replies; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 19:54 UTC (permalink / raw) To: Linus Torvalds Cc: Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Linus Torvalds wrote: > > i'd expect the 'Y' to be picked up from the defconfig - no? > > No. defconfig is either used 100% or not at all. hm, then what is the standard way to make a new kernel option default-Y? At least for the development kernel, a default-enabled kksymoops sounds like the right way to go. Ingo ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:54 ` Ingo Molnar @ 2002-09-25 19:52 ` Linus Torvalds 2002-09-25 20:04 ` Ingo Molnar 2002-09-25 20:31 ` Jeff Garzik 2002-09-25 19:55 ` Kai Germaschewski 1 sibling, 2 replies; 23+ messages in thread From: Linus Torvalds @ 2002-09-25 19:52 UTC (permalink / raw) To: Ingo Molnar Cc: Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Ingo Molnar wrote: > > hm, then what is the standard way to make a new kernel option default-Y? There is none. In fact, there _cannot_ be any these days, since all recent kernrels have stopped using defconfig entirely, and favour using /etc/kernel-config instead (making it much easier to have per-machine default configurations). > At least for the development kernel, a default-enabled kksymoops sounds > like the right way to go. We can ask people to enable it if they can't get their oops reports together (and whether they get their oops reports in shape by using the user-space ksymoops or the kernel version really doesn't matter). Linus ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:52 ` Linus Torvalds @ 2002-09-25 20:04 ` Ingo Molnar 2002-09-25 20:31 ` Jeff Garzik 1 sibling, 0 replies; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 20:04 UTC (permalink / raw) To: Linus Torvalds Cc: Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven one config variant (no-modules, no-kksymoops) did not compile due to linux/errno.h not being included in linux/module.h, the attached patch does this properly. Ingo --- linux/arch/i386/kernel/head.S.orig Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/kernel/head.S Wed Sep 25 21:46:56 2002 @@ -121,7 +121,7 @@ */ xorl %eax,%eax movl $__bss_start,%edi - movl $_end,%ecx + movl $__bss_stop,%ecx subl %edi,%ecx rep stosb --- linux/arch/i386/kernel/process.c.orig Fri Sep 20 17:20:12 2002 +++ linux/arch/i386/kernel/process.c Wed Sep 25 21:46:56 2002 @@ -33,6 +33,7 @@ #include <linux/reboot.h> #include <linux/init.h> #include <linux/mc146818rtc.h> +#include <linux/module.h> #include <asm/uaccess.h> #include <asm/pgtable.h> @@ -155,8 +156,6 @@ __setup("idle=", idle_setup); -extern void show_trace(unsigned long* esp); - void show_regs(struct pt_regs * regs) { unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; @@ -164,6 +163,8 @@ printk("\n"); printk("Pid: %d, comm: %20s\n", current->pid, current->comm); printk("EIP: %04x:[<%08lx>] CPU: %d",0xffff & regs->xcs,regs->eip, smp_processor_id()); + print_symbol("EIP is at %s\n", regs->eip); + if (regs->xcs & 3) printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp); printk(" EFLAGS: %08lx %s\n",regs->eflags, print_tainted()); --- linux/arch/i386/kernel/traps.c.orig Fri Sep 20 17:20:19 2002 +++ linux/arch/i386/kernel/traps.c Wed Sep 25 21:46:56 2002 @@ -94,7 +94,6 @@ #ifdef CONFIG_MODULES -extern struct module *module_list; extern struct module kernel_module; static inline int kernel_text_address(unsigned long addr) @@ -142,10 +141,12 @@ while (((long) stack & (THREAD_SIZE-1)) != 0) { addr = *stack++; if (kernel_text_address(addr)) { - if (i && ((i % 6) == 0)) - printk("\n "); printk("[<%08lx>] ", addr); - i++; + if (print_symbol("%s\n", addr)) { + /* save screen space */ + if ((i++ % 6) == 0) + printk("\n "); + } } } printk("\n"); @@ -206,8 +207,11 @@ esp = regs->esp; ss = regs->xss & 0xffff; } + print_modules(); printk("CPU: %d\nEIP: %04x:[<%08lx>] %s\nEFLAGS: %08lx\n", smp_processor_id(), 0xffff & regs->xcs, regs->eip, print_tainted(), regs->eflags); + + print_symbol("EIP is at %s\n", regs->eip); printk("eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n", regs->eax, regs->ebx, regs->ecx, regs->edx); printk("esi: %08lx edi: %08lx ebp: %08lx esp: %08lx\n", @@ -268,6 +272,7 @@ (unsigned long)file < PAGE_OFFSET || __get_user(c, file)) file = "<bad filename>"; + printk("------------[ cut here ]------------\n"); printk("kernel BUG at %s:%d!\n", file, line); no_bug: --- linux/arch/i386/Config.help.orig Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/Config.help Wed Sep 25 21:46:56 2002 @@ -946,6 +946,11 @@ of the BUG call as well as the EIP and oops trace. This aids debugging but costs about 70-100K of memory. +CONFIG_KALLSYMS + Say Y here to let the kernel print out symbolic crash information and + symbolic stack backtraces. This increases the size of the kernel + somewhat, as all symbols have to be loaded into the kernel image. + CONFIG_DEBUG_OBSOLETE Say Y here if you want to reduce the chances of the tree compiling, and are prepared to dig into driver internals to fix compile errors. --- linux/arch/i386/config.in.orig Wed Sep 25 21:43:13 2002 +++ linux/arch/i386/config.in Wed Sep 25 21:46:56 2002 @@ -435,6 +435,7 @@ if [ "$CONFIG_HIGHMEM" = "y" ]; then bool ' Highmem debugging' CONFIG_DEBUG_HIGHMEM fi + bool ' Load all symbols for debugging/kksymoops' CONFIG_KALLSYMS fi if [ "$CONFIG_X86_LOCAL_APIC" = "y" ]; then --- linux/arch/i386/vmlinux.lds.S.orig Fri Sep 20 17:20:19 2002 +++ linux/arch/i386/vmlinux.lds.S Wed Sep 25 21:46:56 2002 @@ -78,9 +78,13 @@ .data.cacheline_aligned : { *(.data.cacheline_aligned) } __bss_start = .; /* BSS */ - .bss : { - *(.bss) - } + .bss : { *(.bss) } + __bss_stop = .; + + __start___kallsyms = .; /* All kernel symbols */ + __kallsyms : { *(__kallsyms) } + __stop___kallsyms = .; + _end = . ; /* Sections to be discarded */ --- linux/include/linux/kallsyms.h.orig Wed Sep 25 21:46:56 2002 +++ linux/include/linux/kallsyms.h Wed Sep 25 21:46:56 2002 @@ -0,0 +1,163 @@ +/* kallsyms headers + Copyright 2000 Keith Owens <kaos@ocs.com.au> + + This file is part of the Linux modutils. It is exported to kernel + space so debuggers can access the kallsyms data. + + The kallsyms data contains all the non-stack symbols from a kernel + or a module. The kernel symbols are held between __start___kallsyms + and __stop___kallsyms. The symbols for a module are accessed via + the struct module chain which is based at module_list. + + This program is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by the + Free Software Foundation; either version 2 of the License, or (at your + option) any later version. + + This program is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software Foundation, + Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + */ + +#ident "$Id: linux-2.4.9-kallsyms.patch,v 1.8 2002/02/11 18:34:53 arjanv Exp $" + +#ifndef MODUTILS_KALLSYMS_H +#define MODUTILS_KALLSYMS_H 1 + +/* Have to (re)define these ElfW entries here because external kallsyms + * code does not have access to modutils/include/obj.h. This code is + * included from user spaces tools (modutils) and kernel, they need + * different includes. + */ + +#ifndef ELFCLASS32 +#ifdef __KERNEL__ +#include <linux/elf.h> +#else /* __KERNEL__ */ +#include <elf.h> +#endif /* __KERNEL__ */ +#endif /* ELFCLASS32 */ + +#ifndef ELFCLASSM +#define ELFCLASSM ELF_CLASS +#endif + +#ifndef ElfW +# if ELFCLASSM == ELFCLASS32 +# define ElfW(x) Elf32_ ## x +# define ELFW(x) ELF32_ ## x +# else +# define ElfW(x) Elf64_ ## x +# define ELFW(x) ELF64_ ## x +# endif +#endif + +/* Format of data in the kallsyms section. + * Most of the fields are small numbers but the total size and all + * offsets can be large so use the 32/64 bit types for these fields. + * + * Do not use sizeof() on these structures, modutils may be using extra + * fields. Instead use the size fields in the header to access the + * other bits of data. + */ + +struct kallsyms_header { + int size; /* Size of this header */ + ElfW(Word) total_size; /* Total size of kallsyms data */ + int sections; /* Number of section entries */ + ElfW(Off) section_off; /* Offset to first section entry */ + int section_size; /* Size of one section entry */ + int symbols; /* Number of symbol entries */ + ElfW(Off) symbol_off; /* Offset to first symbol entry */ + int symbol_size; /* Size of one symbol entry */ + ElfW(Off) string_off; /* Offset to first string */ + ElfW(Addr) start; /* Start address of first section */ + ElfW(Addr) end; /* End address of last section */ +}; + +struct kallsyms_section { + ElfW(Addr) start; /* Start address of section */ + ElfW(Word) size; /* Size of this section */ + ElfW(Off) name_off; /* Offset to section name */ + ElfW(Word) flags; /* Flags from section */ +}; + +struct kallsyms_symbol { + ElfW(Off) section_off; /* Offset to section that owns this symbol */ + ElfW(Addr) symbol_addr; /* Address of symbol */ + ElfW(Off) name_off; /* Offset to symbol name */ +}; + +#define KALLSYMS_SEC_NAME "__kallsyms" +#define KALLSYMS_IDX 2 /* obj_kallsyms creates kallsyms as section 2 */ + +#define kallsyms_next_sec(h,s) \ + ((s) = (struct kallsyms_section *)((char *)(s) + (h)->section_size)) +#define kallsyms_next_sym(h,s) \ + ((s) = (struct kallsyms_symbol *)((char *)(s) + (h)->symbol_size)) + +#ifdef CONFIG_KALLSYMS + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start with */ + const char **mod_name, /* Set to module name or "kernel" */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_sections(void *token, + int (*callback)(void *, /* token */ + const char *, /* module name */ + const char *, /* section name */ + ElfW(Addr), /* Section start */ + ElfW(Addr), /* Section end */ + ElfW(Word) /* Section flags */ + ) + ); + +#else + +static inline int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + return -ESRCH; +} + +#endif + +#endif /* kallsyms.h */ --- linux/include/linux/module.h.orig Fri Sep 20 17:20:32 2002 +++ linux/include/linux/module.h Wed Sep 25 22:01:19 2002 @@ -10,6 +10,7 @@ #include <linux/config.h> #include <linux/spinlock.h> #include <linux/list.h> +#include <linux/errno.h> #include <asm/atomic.h> @@ -316,8 +317,6 @@ #define MOD_DEC_USE_COUNT do { } while (0) #define MOD_IN_USE 1 -extern struct module *module_list; - #endif /* !__GENKSYMS__ */ #endif /* MODULE */ @@ -504,6 +503,30 @@ #define SET_MODULE_OWNER(some_struct) do { (some_struct)->owner = THIS_MODULE; } while (0) #else #define SET_MODULE_OWNER(some_struct) do { } while (0) +#endif + +extern void print_modules(void); + +#if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) + +extern struct module *module_list; + +/* + * print_symbols takes a format string containing one %s. + * If support for resolving symbols is compiled in, the %s will + * be replaced by the closest symbol to the address and the entire + * string is printk()ed. Otherwise, nothing is printed. + */ +extern int print_symbol(const char *fmt, unsigned long address); + +#else + +static inline int +print_symbol(const char *fmt, unsigned long address) +{ + return -ESRCH; +} + #endif #endif /* _LINUX_MODULE_H */ --- linux/include/linux/sched.h.orig Wed Sep 25 21:43:13 2002 +++ linux/include/linux/sched.h Wed Sep 25 21:46:56 2002 @@ -151,7 +151,13 @@ extern void sched_init(void); extern void init_idle(task_t *idle, int cpu); + extern void show_state(void); +extern void show_trace(unsigned long *stack); +extern void show_stack(unsigned long *stack); +extern void show_regs(struct pt_regs *); + + extern void cpu_init (void); extern void trap_init(void); extern void update_process_times(int user); --- linux/include/asm-i386/hardirq.h.orig Fri Sep 20 17:20:32 2002 +++ linux/include/asm-i386/hardirq.h Wed Sep 25 21:46:56 2002 @@ -97,6 +97,4 @@ extern void synchronize_irq(unsigned int irq); #endif /* CONFIG_SMP */ -extern void show_stack(unsigned long * esp); - #endif /* __ASM_HARDIRQ_H */ --- linux/include/asm-i386/ptrace.h.orig Fri Sep 20 17:20:16 2002 +++ linux/include/asm-i386/ptrace.h Wed Sep 25 21:46:56 2002 @@ -57,7 +57,6 @@ #ifdef __KERNEL__ #define user_mode(regs) ((VM_MASK & (regs)->eflags) || (3 & (regs)->xcs)) #define instruction_pointer(regs) ((regs)->eip) -extern void show_regs(struct pt_regs *); #endif #endif --- linux/kernel/Makefile.orig Fri Sep 20 17:20:19 2002 +++ linux/kernel/Makefile Wed Sep 25 21:46:56 2002 @@ -3,7 +3,7 @@ # export-objs = signal.o sys.o kmod.o context.o ksyms.o pm.o exec_domain.o \ - printk.o platform.o suspend.o dma.o + printk.o platform.o suspend.o dma.o module.o obj-y = sched.o fork.o exec_domain.o panic.o printk.o \ module.o exit.o itimer.o time.o softirq.o resource.o \ @@ -14,6 +14,7 @@ obj-$(CONFIG_SMP) += cpu.o obj-$(CONFIG_UID16) += uid16.o obj-$(CONFIG_MODULES) += ksyms.o +obj-$(CONFIG_KALLSYMS) += kallsyms.o obj-$(CONFIG_PM) += pm.o obj-$(CONFIG_BSD_PROCESS_ACCT) += acct.o obj-$(CONFIG_SOFTWARE_SUSPEND) += suspend.o --- linux/kernel/kallsyms.c.orig Wed Sep 25 21:46:56 2002 +++ linux/kernel/kallsyms.c Wed Sep 25 21:51:18 2002 @@ -0,0 +1,227 @@ +/* + * kksymoops.c: in-kernel printing of symbolic oopses and stack traces. + * + * Copyright 2000 Keith Owens <kaos@ocs.com.au> April 2000 + * Copyright 2002 Arjan van de Ven <arjanv@redhat.com> + * + This code uses the list of all kernel and module symbols to :- + + * Find any non-stack symbol in a kernel or module. Symbols do + not have to be exported for debugging. + + * Convert an address to the module (or kernel) that owns it, the + section it is in and the nearest symbol. This finds all non-stack + symbols, not just exported ones. + + */ + +#include <linux/mm.h> +#include <linux/module.h> +#include <linux/kallsyms.h> + +/* A symbol can appear in more than one module. A token is used to + * restart the scan at the next module, set the token to 0 for the + * first scan of each symbol. + */ + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start at */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec; + const struct kallsyms_symbol *ka_sym = NULL; + const char *ka_str = NULL; + const struct module *m; + int i = 0, l; + const char *p, *pt_R; + char *p2; + + /* Restart? */ + m = module_list; + if (token && *token) { + for (; m; m = m->next) + if ((unsigned long)m == *token) + break; + if (m) + m = m->next; + } + + for (; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + p = ka_str + ka_sym->name_off; + if (strcmp(p, name) == 0) + break; + /* Unversioned requests match versioned names */ + if (!(pt_R = strstr(p, "_R"))) + continue; + l = strlen(pt_R); + if (l < 10) + continue; /* Not _R.*xxxxxxxx */ + (void)simple_strtoul(pt_R+l-8, &p2, 16); + if (*p2) + continue; /* Not _R.*xxxxxxxx */ + if (strncmp(p, name, pt_R-p) == 0) + break; /* Match with version */ + } + if (i < ka_hdr->symbols) + break; + } + + if (token) + *token = (unsigned long)m; + if (!m) + return(0); /* not found */ + + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off); + *mod_name = m->name; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = ka_sym->name_off + ka_str; + *sym_start = ka_sym->symbol_addr; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + *sym_end = ka_symn->symbol_addr; + } + else + *sym_end = *sec_end; + return(1); +} + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const struct kallsyms_symbol *ka_sym; + const char *ka_str; + const struct module *m; + int i; + unsigned long end; + + for (m = module_list; m; m = m->next) { + + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off); + /* Is the address in any section in this module? */ + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (ka_sec->start <= address && + (ka_sec->start + ka_sec->size) > address) + break; + } + if (i < ka_hdr->sections) + break; /* Found a matching section */ + } + + if (!m) + return(0); /* not found */ + + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + *mod_name = m->name; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = *sec_name; /* In case we find no matching symbol */ + *sym_start = *sec_start; + *sym_end = *sec_end; + + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + if (ka_sym->symbol_addr > address) + continue; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + end = ka_symn->symbol_addr; + } + else + end = *sec_end; + if (end <= address) + continue; + if ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off + != (char *)ka_sec) + continue; /* wrong section */ + *sym_name = ka_str + ka_sym->name_off; + *sym_start = ka_sym->symbol_addr; + *sym_end = end; + break; + } + return(1); +} + +/* List all sections in all modules. The callback routine is invoked with + * token, module name, section name, section start, section end, section flags. + */ +int kallsyms_sections(void *token, + int (*callback)(void *, const char *, const char *, ElfW(Addr), ElfW(Addr), ElfW(Word))) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const char *ka_str; + const struct module *m; + int i; + + for (m = module_list; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) ((char *)ka_hdr + ka_hdr->section_off); + ka_str = ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (callback( + token, + *(m->name) ? m->name : "kernel", + ka_sec->name_off + ka_str, + ka_sec->start, + ka_sec->start + ka_sec->size, + ka_sec->flags)) + return(0); + } + } + return(1); +} --- linux/kernel/module.c.orig Fri Sep 20 17:20:19 2002 +++ linux/kernel/module.c Wed Sep 25 21:52:00 2002 @@ -3,6 +3,7 @@ #include <linux/module.h> #include <asm/module.h> #include <asm/uaccess.h> +#include <linux/kallsyms.h> #include <linux/vmalloc.h> #include <linux/smp_lock.h> #include <asm/pgalloc.h> @@ -39,13 +40,19 @@ extern const struct exception_table_entry __start___ex_table[]; extern const struct exception_table_entry __stop___ex_table[]; -extern const char __start___kallsyms[] __attribute__ ((weak)); -extern const char __stop___kallsyms[] __attribute__ ((weak)); +extern const char __start___kallsyms[] __attribute__((weak)); +extern const char __stop___kallsyms[] __attribute__((weak)); + +/* modutils uses these exported symbols to figure out if + kallsyms support is present */ + +EXPORT_SYMBOL(__start___kallsyms); +EXPORT_SYMBOL(__stop___kallsyms); struct module kernel_module = { .size_of_struct = sizeof(struct module), - .name = "", + .name = NULL, .uc = {ATOMIC_INIT(1)}, .flags = MOD_RUNNING, .syms = __start___ksymtab, @@ -1220,6 +1227,30 @@ .show = s_show }; +#define MODLIST_SIZE 4096 + +/* + * this function isn't smp safe but that's not really a problem; it's + * called from oops context only and any locking could actually prevent + * the oops from going out; the line that is generated is informational + * only and should NEVER prevent the real oops from going out. + */ +void print_modules(void) +{ + static char modlist[MODLIST_SIZE]; + struct module *this_mod; + int pos = 0; + + this_mod = module_list; + while (this_mod) { + if (this_mod->name) + pos += snprintf(modlist+pos, MODLIST_SIZE-pos-1, + "%s ", this_mod->name); + this_mod = this_mod->next; + } + printk("%s\n",modlist); +} + #else /* CONFIG_MODULES */ /* Dummy syscalls for people who don't want modules */ @@ -1265,4 +1296,79 @@ return 1; } +void print_modules(void) +{ +} + #endif /* CONFIG_MODULES */ + + +#if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) + +#define MAX_SYMBOL_SIZE 512 + +static void +address_to_exported_symbol(unsigned long address, const char **mod_name, + const char **sym_name, unsigned long *sym_start, + unsigned long *sym_end) +{ + struct module *this_mod; + int i; + + for (this_mod = module_list; this_mod; this_mod = this_mod->next) { + /* walk the symbol list of this module. Only symbols + who's address is smaller than the searched for address + are relevant; and only if it's better than the best so far */ + for (i = 0; i < this_mod->nsyms; i++) + if ((this_mod->syms[i].value <= address) && + (*sym_start < this_mod->syms[i].value)) { + *sym_start = this_mod->syms[i].value; + *sym_name = this_mod->syms[i].name; + *mod_name = this_mod->name; + if (i + 1 < this_mod->nsyms) + *sym_end = this_mod->syms[i+1].value; + else + *sym_end = (unsigned long) this_mod + this_mod->size; + } + } +} + +int +print_symbol(const char *fmt, unsigned long address) +{ + /* static to not take up stackspace; if we race here too bad */ + static char buffer[MAX_SYMBOL_SIZE]; + + const char *mod_name = NULL, *sec_name = NULL, *sym_name = NULL; + unsigned long mod_start, mod_end, sec_start, sec_end, + sym_start, sym_end; + char *tag = ""; + + memset(buffer, 0, MAX_SYMBOL_SIZE); + + sym_start = 0; + if (!kallsyms_address_to_symbol(address, &mod_name, &mod_start, &mod_end, &sec_name, &sec_start, &sec_end, &sym_name, &sym_start, &sym_end)) { + tag = "E "; + address_to_exported_symbol(address, &mod_name, &sym_name, &sym_start, &sym_end); + } + + if (sym_start) { + if (mod_name) + snprintf(buffer, MAX_SYMBOL_SIZE - 1, "%s%s+%#x/%#x [%s]", + tag, sym_name, + (unsigned int)(address - sym_start), + (unsigned int)(sym_end - sym_start), + mod_name); + else + snprintf(buffer, MAX_SYMBOL_SIZE - 1, "%s%s+%#x/%#x", + tag, sym_name, + (unsigned int)(address - sym_start), + (unsigned int)(sym_end - sym_start)); + printk(fmt, buffer); + } else { + printk(fmt, "[unresolved]"); + } + return 0; +} + +#endif --- linux/Makefile.orig Wed Sep 25 21:43:13 2002 +++ linux/Makefile Wed Sep 25 21:46:56 2002 @@ -138,6 +138,7 @@ MAKEFILES = $(TOPDIR)/.config GENKSYMS = /sbin/genksyms DEPMOD = /sbin/depmod +KALLSYMS = /sbin/kallsyms PERL = perl MODFLAGS = -DMODULE CFLAGS_MODULE = $(MODFLAGS) @@ -291,32 +292,64 @@ vmlinux-objs := $(HEAD) $(INIT) $(CORE_FILES) $(LIBS) $(DRIVERS) $(NETWORKS) quiet_cmd_link_vmlinux = LD $@ -cmd_link_vmlinux = $(LD) $(LDFLAGS) $(LDFLAGS_$(@F)) $(HEAD) $(INIT) \ - --start-group \ - $(CORE_FILES) \ - $(LIBS) \ - $(DRIVERS) \ - $(NETWORKS) \ - --end-group \ - -o vmlinux +define cmd_link_vmlinux + $(LD) $(LDFLAGS) $(LDFLAGS_vmlinux) $(HEAD) $(INIT) \ + --start-group \ + $(CORE_FILES) \ + $(LIBS) \ + $(DRIVERS) \ + $(NETWORKS) \ + --end-group \ + $(filter $(kallsyms.o),$^) \ + -o $@ +endef # set -e makes the rule exit immediately on error -define rule_link_vmlinux +define rule_vmlinux set -e echo ' Generating build number' - . scripts/mkversion > .tmpversion - mv -f .tmpversion .version + . scripts/mkversion > .tmp_version + mv -f .tmp_version .version +$(MAKE) -C init $(call cmd,link_vmlinux) echo 'cmd_$@ := $(cmd_link_vmlinux)' > $(@D)/.$(@F).cmd - $(NM) vmlinux | grep -v '\(compiled\)\|\(\.o$$\)\|\( [aUw] \)\|\(\.\.ng$$\)\|\(LASH[RL]DI\)' | sort > System.map + $(NM) $@ | grep -v '\(compiled\)\|\(\.o$$\)\|\( [aUw] \)\|\(\.\.ng$$\)\|\(LASH[RL]DI\)' | sort > System.map endef LDFLAGS_vmlinux += -T arch/$(ARCH)/vmlinux.lds.s -vmlinux: $(vmlinux-objs) arch/$(ARCH)/vmlinux.lds.s FORCE - $(call if_changed_rule,link_vmlinux) +# Generate section listing all symbols and add it into vmlinux + +ifdef CONFIG_KALLSYMS + +kallsyms.o := .tmp_kallsyms.o + +quiet_cmd_kallsyms = KSYM $@ +cmd_kallsyms = $(KALLSYMS) $< > $@ + +.tmp_kallsyms.o: .tmp_vmlinux + $(call cmd,kallsyms) + +# After generating .tmp_vmlinux just like vmlinux, decrement the version +# number again, so the final vmlinux gets the same one. +# Ignore return value of 'expr'. + +define rule_.tmp_vmlinux + $(rule_vmlinux) + if expr 0`cat .version` - 1 > .tmp_version; then true; fi + mv -f .tmp_version .version +endef + +.tmp_vmlinux: $(vmlinux-objs) arch/$(ARCH)/vmlinux.lds.s FORCE + $(call if_changed_rule,.tmp_vmlinux) + +endif + +# Finally the vmlinux rule + +vmlinux: $(vmlinux-objs) $(kallsyms.o) arch/$(ARCH)/vmlinux.lds.s FORCE + $(call if_changed_rule,vmlinux) # The actual objects are generated when descending, # make sure no implicit rule kicks in @@ -820,7 +853,7 @@ # FIXME Should go into a make.lib or something # =========================================================================== -echo_target = $(RELDIR)/$@ +echo_target = $@ a_flags = -Wp,-MD,$(depfile) $(AFLAGS) $(NOSTDINC_FLAGS) \ $(modkern_aflags) $(EXTRA_AFLAGS) $(AFLAGS_$(*F).o) ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:52 ` Linus Torvalds 2002-09-25 20:04 ` Ingo Molnar @ 2002-09-25 20:31 ` Jeff Garzik 1 sibling, 0 replies; 23+ messages in thread From: Jeff Garzik @ 2002-09-25 20:31 UTC (permalink / raw) To: Linus Torvalds Cc: Ingo Molnar, Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven Linus Torvalds wrote: > In fact, there _cannot_ be any these days, since all recent kernrels have > stopped using defconfig entirely, and favour using /etc/kernel-config > instead (making it much easier to have per-machine default > configurations). wow, cool... I missed that. ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:54 ` Ingo Molnar 2002-09-25 19:52 ` Linus Torvalds @ 2002-09-25 19:55 ` Kai Germaschewski 1 sibling, 0 replies; 23+ messages in thread From: Kai Germaschewski @ 2002-09-25 19:55 UTC (permalink / raw) To: Ingo Molnar; +Cc: Linus Torvalds, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Ingo Molnar wrote: > hm, then what is the standard way to make a new kernel option default-Y? > At least for the development kernel, a default-enabled kksymoops sounds > like the right way to go. There isn't really any. You can do define_bool CONFIG_KALLSYMS y for two releases and hope to propagate it into people's .config before making it an actual choice in config.in. Or you can hack something up like if [ "$CONFIG_KALLSYMS" = "" ]; then define_bool CONFIG_KALLSYMS y fi bool 'kallsyms' CONFIG_KALLSYMS which gives the desired effect in make oldconfig, but may magically break make xconfig or so. So no, just hope for people to read the help text and enable it. --Kai ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:23 ` Kai Germaschewski 2002-09-25 19:42 ` Ingo Molnar @ 2002-09-25 19:45 ` Linus Torvalds 2002-09-25 19:53 ` Ingo Molnar 2002-09-25 19:56 ` Ingo Molnar 1 sibling, 2 replies; 23+ messages in thread From: Linus Torvalds @ 2002-09-25 19:45 UTC (permalink / raw) To: Kai Germaschewski Cc: Ingo Molnar, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Kai Germaschewski wrote: > > So this patch converts the output to the more familiar > > sys_gettimeofday+0x84/0x108 [module-name] > > format, where module-name is "" for the kernel. I want those [] gone too, I see no reason for them except to make the output ugly. Linus ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:45 ` Linus Torvalds @ 2002-09-25 19:53 ` Ingo Molnar 2002-09-25 19:04 ` Cort Dougan 2002-09-25 19:56 ` Ingo Molnar 1 sibling, 1 reply; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 19:53 UTC (permalink / raw) To: Linus Torvalds Cc: Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Linus Torvalds wrote: > I want those [] gone too, I see no reason for them except to make the > output ugly. yep, i removed them already, patch in a minute, after i'm done with testing. Ingo ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:53 ` Ingo Molnar @ 2002-09-25 19:04 ` Cort Dougan 2002-09-25 20:14 ` Ingo Molnar 0 siblings, 1 reply; 23+ messages in thread From: Cort Dougan @ 2002-09-25 19:04 UTC (permalink / raw) To: Ingo Molnar Cc: Linus Torvalds, Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven How does this change differ from the one I sent a month ago? ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:04 ` Cort Dougan @ 2002-09-25 20:14 ` Ingo Molnar 0 siblings, 0 replies; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 20:14 UTC (permalink / raw) To: Cort Dougan Cc: Linus Torvalds, Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On Wed, 25 Sep 2002, Cort Dougan wrote: > How does this change differ from the one I sent a month ago? kallsyms/kksymoops existed for quite some time already, and there are a couple of additional things it does: - it includes and uses the full symbol table, not just the module symbols. (hence the 'all' in KALLSYMS) - as far as i can see your patch did not extend to show_stack()? - kksymoops prints a module list as well. Ingo ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:45 ` Linus Torvalds 2002-09-25 19:53 ` Ingo Molnar @ 2002-09-25 19:56 ` Ingo Molnar 2002-09-25 22:04 ` J.A. Magallon 2002-09-26 17:16 ` Ruth Ivimey-Cook 1 sibling, 2 replies; 23+ messages in thread From: Ingo Molnar @ 2002-09-25 19:56 UTC (permalink / raw) To: Linus Torvalds Cc: Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven okay, here is the new oops output: ------------[ cut here ]------------ kernel BUG at time.c:99! invalid operand: 0000 CPU: 1 EIP: 0060:[<c011bd14>] Not tainted EFLAGS: 00010246 EIP is at sys_gettimeofday+0x84/0x90 eax: 0000004e ebx: cef9e000 ecx: 00000000 edx: 00000068 esi: 00000000 edi: 00000000 ebp: bffffad8 esp: cef9ffa0 ds: 0068 es: 0068 ss: 0068 Process gettimeofday (pid: 549, threadinfo=cef9e000 task=cf84d860) Stack: 4001695c bffff414 40156154 00000004 c0112a40 cef9e000 400168e4 bffffb44 c0107973 00000000 00000000 40156154 400168e4 bffffb44 bffffad8 0000004e 0000002b 0000002b 0000004e 400cecc1 00000023 00000246 bffffacc 0000002b Call Trace: [<c0112a40>] do_page_fault+0x0/0x4a2 [<c0107973>] syscall_call+0x7/0xb and the patch, against BK-curr: --- linux/arch/i386/kernel/head.S.orig Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/kernel/head.S Wed Sep 25 21:46:56 2002 @@ -121,7 +121,7 @@ */ xorl %eax,%eax movl $__bss_start,%edi - movl $_end,%ecx + movl $__bss_stop,%ecx subl %edi,%ecx rep stosb --- linux/arch/i386/kernel/process.c.orig Fri Sep 20 17:20:12 2002 +++ linux/arch/i386/kernel/process.c Wed Sep 25 21:46:56 2002 @@ -33,6 +33,7 @@ #include <linux/reboot.h> #include <linux/init.h> #include <linux/mc146818rtc.h> +#include <linux/module.h> #include <asm/uaccess.h> #include <asm/pgtable.h> @@ -155,8 +156,6 @@ __setup("idle=", idle_setup); -extern void show_trace(unsigned long* esp); - void show_regs(struct pt_regs * regs) { unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L; @@ -164,6 +163,8 @@ printk("\n"); printk("Pid: %d, comm: %20s\n", current->pid, current->comm); printk("EIP: %04x:[<%08lx>] CPU: %d",0xffff & regs->xcs,regs->eip, smp_processor_id()); + print_symbol("EIP is at %s\n", regs->eip); + if (regs->xcs & 3) printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp); printk(" EFLAGS: %08lx %s\n",regs->eflags, print_tainted()); --- linux/arch/i386/kernel/traps.c.orig Fri Sep 20 17:20:19 2002 +++ linux/arch/i386/kernel/traps.c Wed Sep 25 21:46:56 2002 @@ -94,7 +94,6 @@ #ifdef CONFIG_MODULES -extern struct module *module_list; extern struct module kernel_module; static inline int kernel_text_address(unsigned long addr) @@ -142,10 +141,12 @@ while (((long) stack & (THREAD_SIZE-1)) != 0) { addr = *stack++; if (kernel_text_address(addr)) { - if (i && ((i % 6) == 0)) - printk("\n "); printk("[<%08lx>] ", addr); - i++; + if (print_symbol("%s\n", addr)) { + /* save screen space */ + if ((i++ % 6) == 0) + printk("\n "); + } } } printk("\n"); @@ -206,8 +207,11 @@ esp = regs->esp; ss = regs->xss & 0xffff; } + print_modules(); printk("CPU: %d\nEIP: %04x:[<%08lx>] %s\nEFLAGS: %08lx\n", smp_processor_id(), 0xffff & regs->xcs, regs->eip, print_tainted(), regs->eflags); + + print_symbol("EIP is at %s\n", regs->eip); printk("eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n", regs->eax, regs->ebx, regs->ecx, regs->edx); printk("esi: %08lx edi: %08lx ebp: %08lx esp: %08lx\n", @@ -268,6 +272,7 @@ (unsigned long)file < PAGE_OFFSET || __get_user(c, file)) file = "<bad filename>"; + printk("------------[ cut here ]------------\n"); printk("kernel BUG at %s:%d!\n", file, line); no_bug: --- linux/arch/i386/Config.help.orig Fri Sep 20 17:20:16 2002 +++ linux/arch/i386/Config.help Wed Sep 25 21:46:56 2002 @@ -946,6 +946,11 @@ of the BUG call as well as the EIP and oops trace. This aids debugging but costs about 70-100K of memory. +CONFIG_KALLSYMS + Say Y here to let the kernel print out symbolic crash information and + symbolic stack backtraces. This increases the size of the kernel + somewhat, as all symbols have to be loaded into the kernel image. + CONFIG_DEBUG_OBSOLETE Say Y here if you want to reduce the chances of the tree compiling, and are prepared to dig into driver internals to fix compile errors. --- linux/arch/i386/config.in.orig Wed Sep 25 21:43:13 2002 +++ linux/arch/i386/config.in Wed Sep 25 21:46:56 2002 @@ -435,6 +435,7 @@ if [ "$CONFIG_HIGHMEM" = "y" ]; then bool ' Highmem debugging' CONFIG_DEBUG_HIGHMEM fi + bool ' Load all symbols for debugging/kksymoops' CONFIG_KALLSYMS fi if [ "$CONFIG_X86_LOCAL_APIC" = "y" ]; then --- linux/arch/i386/vmlinux.lds.S.orig Fri Sep 20 17:20:19 2002 +++ linux/arch/i386/vmlinux.lds.S Wed Sep 25 21:46:56 2002 @@ -78,9 +78,13 @@ .data.cacheline_aligned : { *(.data.cacheline_aligned) } __bss_start = .; /* BSS */ - .bss : { - *(.bss) - } + .bss : { *(.bss) } + __bss_stop = .; + + __start___kallsyms = .; /* All kernel symbols */ + __kallsyms : { *(__kallsyms) } + __stop___kallsyms = .; + _end = . ; /* Sections to be discarded */ --- linux/include/linux/kallsyms.h.orig Wed Sep 25 21:46:56 2002 +++ linux/include/linux/kallsyms.h Wed Sep 25 21:46:56 2002 @@ -0,0 +1,163 @@ +/* kallsyms headers + Copyright 2000 Keith Owens <kaos@ocs.com.au> + + This file is part of the Linux modutils. It is exported to kernel + space so debuggers can access the kallsyms data. + + The kallsyms data contains all the non-stack symbols from a kernel + or a module. The kernel symbols are held between __start___kallsyms + and __stop___kallsyms. The symbols for a module are accessed via + the struct module chain which is based at module_list. + + This program is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by the + Free Software Foundation; either version 2 of the License, or (at your + option) any later version. + + This program is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software Foundation, + Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + */ + +#ident "$Id: linux-2.4.9-kallsyms.patch,v 1.8 2002/02/11 18:34:53 arjanv Exp $" + +#ifndef MODUTILS_KALLSYMS_H +#define MODUTILS_KALLSYMS_H 1 + +/* Have to (re)define these ElfW entries here because external kallsyms + * code does not have access to modutils/include/obj.h. This code is + * included from user spaces tools (modutils) and kernel, they need + * different includes. + */ + +#ifndef ELFCLASS32 +#ifdef __KERNEL__ +#include <linux/elf.h> +#else /* __KERNEL__ */ +#include <elf.h> +#endif /* __KERNEL__ */ +#endif /* ELFCLASS32 */ + +#ifndef ELFCLASSM +#define ELFCLASSM ELF_CLASS +#endif + +#ifndef ElfW +# if ELFCLASSM == ELFCLASS32 +# define ElfW(x) Elf32_ ## x +# define ELFW(x) ELF32_ ## x +# else +# define ElfW(x) Elf64_ ## x +# define ELFW(x) ELF64_ ## x +# endif +#endif + +/* Format of data in the kallsyms section. + * Most of the fields are small numbers but the total size and all + * offsets can be large so use the 32/64 bit types for these fields. + * + * Do not use sizeof() on these structures, modutils may be using extra + * fields. Instead use the size fields in the header to access the + * other bits of data. + */ + +struct kallsyms_header { + int size; /* Size of this header */ + ElfW(Word) total_size; /* Total size of kallsyms data */ + int sections; /* Number of section entries */ + ElfW(Off) section_off; /* Offset to first section entry */ + int section_size; /* Size of one section entry */ + int symbols; /* Number of symbol entries */ + ElfW(Off) symbol_off; /* Offset to first symbol entry */ + int symbol_size; /* Size of one symbol entry */ + ElfW(Off) string_off; /* Offset to first string */ + ElfW(Addr) start; /* Start address of first section */ + ElfW(Addr) end; /* End address of last section */ +}; + +struct kallsyms_section { + ElfW(Addr) start; /* Start address of section */ + ElfW(Word) size; /* Size of this section */ + ElfW(Off) name_off; /* Offset to section name */ + ElfW(Word) flags; /* Flags from section */ +}; + +struct kallsyms_symbol { + ElfW(Off) section_off; /* Offset to section that owns this symbol */ + ElfW(Addr) symbol_addr; /* Address of symbol */ + ElfW(Off) name_off; /* Offset to symbol name */ +}; + +#define KALLSYMS_SEC_NAME "__kallsyms" +#define KALLSYMS_IDX 2 /* obj_kallsyms creates kallsyms as section 2 */ + +#define kallsyms_next_sec(h,s) \ + ((s) = (struct kallsyms_section *)((char *)(s) + (h)->section_size)) +#define kallsyms_next_sym(h,s) \ + ((s) = (struct kallsyms_symbol *)((char *)(s) + (h)->symbol_size)) + +#ifdef CONFIG_KALLSYMS + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start with */ + const char **mod_name, /* Set to module name or "kernel" */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ); + +int kallsyms_sections(void *token, + int (*callback)(void *, /* token */ + const char *, /* module name */ + const char *, /* section name */ + ElfW(Addr), /* Section start */ + ElfW(Addr), /* Section end */ + ElfW(Word) /* Section flags */ + ) + ); + +#else + +static inline int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + return -ESRCH; +} + +#endif + +#endif /* kallsyms.h */ --- linux/include/linux/module.h.orig Fri Sep 20 17:20:32 2002 +++ linux/include/linux/module.h Wed Sep 25 21:46:56 2002 @@ -316,8 +316,6 @@ #define MOD_DEC_USE_COUNT do { } while (0) #define MOD_IN_USE 1 -extern struct module *module_list; - #endif /* !__GENKSYMS__ */ #endif /* MODULE */ @@ -504,6 +502,30 @@ #define SET_MODULE_OWNER(some_struct) do { (some_struct)->owner = THIS_MODULE; } while (0) #else #define SET_MODULE_OWNER(some_struct) do { } while (0) +#endif + +extern void print_modules(void); + +#if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) + +extern struct module *module_list; + +/* + * print_symbols takes a format string containing one %s. + * If support for resolving symbols is compiled in, the %s will + * be replaced by the closest symbol to the address and the entire + * string is printk()ed. Otherwise, nothing is printed. + */ +extern int print_symbol(const char *fmt, unsigned long address); + +#else + +static inline int +print_symbol(const char *fmt, unsigned long address) +{ + return -ESRCH; +} + #endif #endif /* _LINUX_MODULE_H */ --- linux/include/linux/sched.h.orig Wed Sep 25 21:43:13 2002 +++ linux/include/linux/sched.h Wed Sep 25 21:46:56 2002 @@ -151,7 +151,13 @@ extern void sched_init(void); extern void init_idle(task_t *idle, int cpu); + extern void show_state(void); +extern void show_trace(unsigned long *stack); +extern void show_stack(unsigned long *stack); +extern void show_regs(struct pt_regs *); + + extern void cpu_init (void); extern void trap_init(void); extern void update_process_times(int user); --- linux/include/asm-i386/hardirq.h.orig Fri Sep 20 17:20:32 2002 +++ linux/include/asm-i386/hardirq.h Wed Sep 25 21:46:56 2002 @@ -97,6 +97,4 @@ extern void synchronize_irq(unsigned int irq); #endif /* CONFIG_SMP */ -extern void show_stack(unsigned long * esp); - #endif /* __ASM_HARDIRQ_H */ --- linux/include/asm-i386/ptrace.h.orig Fri Sep 20 17:20:16 2002 +++ linux/include/asm-i386/ptrace.h Wed Sep 25 21:46:56 2002 @@ -57,7 +57,6 @@ #ifdef __KERNEL__ #define user_mode(regs) ((VM_MASK & (regs)->eflags) || (3 & (regs)->xcs)) #define instruction_pointer(regs) ((regs)->eip) -extern void show_regs(struct pt_regs *); #endif #endif --- linux/kernel/Makefile.orig Fri Sep 20 17:20:19 2002 +++ linux/kernel/Makefile Wed Sep 25 21:46:56 2002 @@ -3,7 +3,7 @@ # export-objs = signal.o sys.o kmod.o context.o ksyms.o pm.o exec_domain.o \ - printk.o platform.o suspend.o dma.o + printk.o platform.o suspend.o dma.o module.o obj-y = sched.o fork.o exec_domain.o panic.o printk.o \ module.o exit.o itimer.o time.o softirq.o resource.o \ @@ -14,6 +14,7 @@ obj-$(CONFIG_SMP) += cpu.o obj-$(CONFIG_UID16) += uid16.o obj-$(CONFIG_MODULES) += ksyms.o +obj-$(CONFIG_KALLSYMS) += kallsyms.o obj-$(CONFIG_PM) += pm.o obj-$(CONFIG_BSD_PROCESS_ACCT) += acct.o obj-$(CONFIG_SOFTWARE_SUSPEND) += suspend.o --- linux/kernel/kallsyms.c.orig Wed Sep 25 21:46:56 2002 +++ linux/kernel/kallsyms.c Wed Sep 25 21:51:18 2002 @@ -0,0 +1,227 @@ +/* + * kksymoops.c: in-kernel printing of symbolic oopses and stack traces. + * + * Copyright 2000 Keith Owens <kaos@ocs.com.au> April 2000 + * Copyright 2002 Arjan van de Ven <arjanv@redhat.com> + * + This code uses the list of all kernel and module symbols to :- + + * Find any non-stack symbol in a kernel or module. Symbols do + not have to be exported for debugging. + + * Convert an address to the module (or kernel) that owns it, the + section it is in and the nearest symbol. This finds all non-stack + symbols, not just exported ones. + + */ + +#include <linux/mm.h> +#include <linux/module.h> +#include <linux/kallsyms.h> + +/* A symbol can appear in more than one module. A token is used to + * restart the scan at the next module, set the token to 0 for the + * first scan of each symbol. + */ + +int kallsyms_symbol_to_address( + const char *name, /* Name to lookup */ + unsigned long *token, /* Which module to start at */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec; + const struct kallsyms_symbol *ka_sym = NULL; + const char *ka_str = NULL; + const struct module *m; + int i = 0, l; + const char *p, *pt_R; + char *p2; + + /* Restart? */ + m = module_list; + if (token && *token) { + for (; m; m = m->next) + if ((unsigned long)m == *token) + break; + if (m) + m = m->next; + } + + for (; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + p = ka_str + ka_sym->name_off; + if (strcmp(p, name) == 0) + break; + /* Unversioned requests match versioned names */ + if (!(pt_R = strstr(p, "_R"))) + continue; + l = strlen(pt_R); + if (l < 10) + continue; /* Not _R.*xxxxxxxx */ + (void)simple_strtoul(pt_R+l-8, &p2, 16); + if (*p2) + continue; /* Not _R.*xxxxxxxx */ + if (strncmp(p, name, pt_R-p) == 0) + break; /* Match with version */ + } + if (i < ka_hdr->symbols) + break; + } + + if (token) + *token = (unsigned long)m; + if (!m) + return(0); /* not found */ + + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off); + *mod_name = m->name; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = ka_sym->name_off + ka_str; + *sym_start = ka_sym->symbol_addr; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + *sym_end = ka_symn->symbol_addr; + } + else + *sym_end = *sec_end; + return(1); +} + +int kallsyms_address_to_symbol( + unsigned long address, /* Address to lookup */ + const char **mod_name, /* Set to module name */ + unsigned long *mod_start, /* Set to start address of module */ + unsigned long *mod_end, /* Set to end address of module */ + const char **sec_name, /* Set to section name */ + unsigned long *sec_start, /* Set to start address of section */ + unsigned long *sec_end, /* Set to end address of section */ + const char **sym_name, /* Set to full symbol name */ + unsigned long *sym_start, /* Set to start address of symbol */ + unsigned long *sym_end /* Set to end address of symbol */ + ) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const struct kallsyms_symbol *ka_sym; + const char *ka_str; + const struct module *m; + int i; + unsigned long end; + + for (m = module_list; m; m = m->next) { + + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) + ((char *)ka_hdr + ka_hdr->section_off); + /* Is the address in any section in this module? */ + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (ka_sec->start <= address && + (ka_sec->start + ka_sec->size) > address) + break; + } + if (i < ka_hdr->sections) + break; /* Found a matching section */ + } + + if (!m) + return(0); /* not found */ + + ka_sym = (struct kallsyms_symbol *) + ((char *)(ka_hdr) + ka_hdr->symbol_off); + ka_str = + ((char *)(ka_hdr) + ka_hdr->string_off); + *mod_name = m->name; + *mod_start = ka_hdr->start; + *mod_end = ka_hdr->end; + *sec_name = ka_sec->name_off + ka_str; + *sec_start = ka_sec->start; + *sec_end = ka_sec->start + ka_sec->size; + *sym_name = *sec_name; /* In case we find no matching symbol */ + *sym_start = *sec_start; + *sym_end = *sec_end; + + for (i = 0; i < ka_hdr->symbols; ++i, kallsyms_next_sym(ka_hdr, ka_sym)) { + if (ka_sym->symbol_addr > address) + continue; + if (i < ka_hdr->symbols-1) { + const struct kallsyms_symbol *ka_symn = ka_sym; + kallsyms_next_sym(ka_hdr, ka_symn); + end = ka_symn->symbol_addr; + } + else + end = *sec_end; + if (end <= address) + continue; + if ((char *)ka_hdr + ka_hdr->section_off + ka_sym->section_off + != (char *)ka_sec) + continue; /* wrong section */ + *sym_name = ka_str + ka_sym->name_off; + *sym_start = ka_sym->symbol_addr; + *sym_end = end; + break; + } + return(1); +} + +/* List all sections in all modules. The callback routine is invoked with + * token, module name, section name, section start, section end, section flags. + */ +int kallsyms_sections(void *token, + int (*callback)(void *, const char *, const char *, ElfW(Addr), ElfW(Addr), ElfW(Word))) +{ + const struct kallsyms_header *ka_hdr = NULL; /* stupid gcc */ + const struct kallsyms_section *ka_sec = NULL; + const char *ka_str; + const struct module *m; + int i; + + for (m = module_list; m; m = m->next) { + if (!mod_member_present(m, kallsyms_start) || + !mod_member_present(m, kallsyms_end) || + m->kallsyms_start >= m->kallsyms_end) + continue; + ka_hdr = (struct kallsyms_header *)m->kallsyms_start; + ka_sec = (const struct kallsyms_section *) ((char *)ka_hdr + ka_hdr->section_off); + ka_str = ((char *)(ka_hdr) + ka_hdr->string_off); + for (i = 0; i < ka_hdr->sections; ++i, kallsyms_next_sec(ka_hdr, ka_sec)) { + if (callback( + token, + *(m->name) ? m->name : "kernel", + ka_sec->name_off + ka_str, + ka_sec->start, + ka_sec->start + ka_sec->size, + ka_sec->flags)) + return(0); + } + } + return(1); +} --- linux/kernel/module.c.orig Fri Sep 20 17:20:19 2002 +++ linux/kernel/module.c Wed Sep 25 21:52:00 2002 @@ -3,6 +3,7 @@ #include <linux/module.h> #include <asm/module.h> #include <asm/uaccess.h> +#include <linux/kallsyms.h> #include <linux/vmalloc.h> #include <linux/smp_lock.h> #include <asm/pgalloc.h> @@ -39,13 +40,19 @@ extern const struct exception_table_entry __start___ex_table[]; extern const struct exception_table_entry __stop___ex_table[]; -extern const char __start___kallsyms[] __attribute__ ((weak)); -extern const char __stop___kallsyms[] __attribute__ ((weak)); +extern const char __start___kallsyms[] __attribute__((weak)); +extern const char __stop___kallsyms[] __attribute__((weak)); + +/* modutils uses these exported symbols to figure out if + kallsyms support is present */ + +EXPORT_SYMBOL(__start___kallsyms); +EXPORT_SYMBOL(__stop___kallsyms); struct module kernel_module = { .size_of_struct = sizeof(struct module), - .name = "", + .name = NULL, .uc = {ATOMIC_INIT(1)}, .flags = MOD_RUNNING, .syms = __start___ksymtab, @@ -1220,6 +1227,30 @@ .show = s_show }; +#define MODLIST_SIZE 4096 + +/* + * this function isn't smp safe but that's not really a problem; it's + * called from oops context only and any locking could actually prevent + * the oops from going out; the line that is generated is informational + * only and should NEVER prevent the real oops from going out. + */ +void print_modules(void) +{ + static char modlist[MODLIST_SIZE]; + struct module *this_mod; + int pos = 0; + + this_mod = module_list; + while (this_mod) { + if (this_mod->name) + pos += snprintf(modlist+pos, MODLIST_SIZE-pos-1, + "%s ", this_mod->name); + this_mod = this_mod->next; + } + printk("%s\n",modlist); +} + #else /* CONFIG_MODULES */ /* Dummy syscalls for people who don't want modules */ @@ -1265,4 +1296,79 @@ return 1; } +void print_modules(void) +{ +} + #endif /* CONFIG_MODULES */ + + +#if defined(CONFIG_MODULES) || defined(CONFIG_KALLSYMS) + +#define MAX_SYMBOL_SIZE 512 + +static void +address_to_exported_symbol(unsigned long address, const char **mod_name, + const char **sym_name, unsigned long *sym_start, + unsigned long *sym_end) +{ + struct module *this_mod; + int i; + + for (this_mod = module_list; this_mod; this_mod = this_mod->next) { + /* walk the symbol list of this module. Only symbols + who's address is smaller than the searched for address + are relevant; and only if it's better than the best so far */ + for (i = 0; i < this_mod->nsyms; i++) + if ((this_mod->syms[i].value <= address) && + (*sym_start < this_mod->syms[i].value)) { + *sym_start = this_mod->syms[i].value; + *sym_name = this_mod->syms[i].name; + *mod_name = this_mod->name; + if (i + 1 < this_mod->nsyms) + *sym_end = this_mod->syms[i+1].value; + else + *sym_end = (unsigned long) this_mod + this_mod->size; + } + } +} + +int +print_symbol(const char *fmt, unsigned long address) +{ + /* static to not take up stackspace; if we race here too bad */ + static char buffer[MAX_SYMBOL_SIZE]; + + const char *mod_name = NULL, *sec_name = NULL, *sym_name = NULL; + unsigned long mod_start, mod_end, sec_start, sec_end, + sym_start, sym_end; + char *tag = ""; + + memset(buffer, 0, MAX_SYMBOL_SIZE); + + sym_start = 0; + if (!kallsyms_address_to_symbol(address, &mod_name, &mod_start, &mod_end, &sec_name, &sec_start, &sec_end, &sym_name, &sym_start, &sym_end)) { + tag = "E "; + address_to_exported_symbol(address, &mod_name, &sym_name, &sym_start, &sym_end); + } + + if (sym_start) { + if (mod_name) + snprintf(buffer, MAX_SYMBOL_SIZE - 1, "%s%s+%#x/%#x [%s]", + tag, sym_name, + (unsigned int)(address - sym_start), + (unsigned int)(sym_end - sym_start), + mod_name); + else + snprintf(buffer, MAX_SYMBOL_SIZE - 1, "%s%s+%#x/%#x", + tag, sym_name, + (unsigned int)(address - sym_start), + (unsigned int)(sym_end - sym_start)); + printk(fmt, buffer); + } else { + printk(fmt, "[unresolved]"); + } + return 0; +} + +#endif --- linux/Makefile.orig Wed Sep 25 21:43:13 2002 +++ linux/Makefile Wed Sep 25 21:46:56 2002 @@ -138,6 +138,7 @@ MAKEFILES = $(TOPDIR)/.config GENKSYMS = /sbin/genksyms DEPMOD = /sbin/depmod +KALLSYMS = /sbin/kallsyms PERL = perl MODFLAGS = -DMODULE CFLAGS_MODULE = $(MODFLAGS) @@ -291,32 +292,64 @@ vmlinux-objs := $(HEAD) $(INIT) $(CORE_FILES) $(LIBS) $(DRIVERS) $(NETWORKS) quiet_cmd_link_vmlinux = LD $@ -cmd_link_vmlinux = $(LD) $(LDFLAGS) $(LDFLAGS_$(@F)) $(HEAD) $(INIT) \ - --start-group \ - $(CORE_FILES) \ - $(LIBS) \ - $(DRIVERS) \ - $(NETWORKS) \ - --end-group \ - -o vmlinux +define cmd_link_vmlinux + $(LD) $(LDFLAGS) $(LDFLAGS_vmlinux) $(HEAD) $(INIT) \ + --start-group \ + $(CORE_FILES) \ + $(LIBS) \ + $(DRIVERS) \ + $(NETWORKS) \ + --end-group \ + $(filter $(kallsyms.o),$^) \ + -o $@ +endef # set -e makes the rule exit immediately on error -define rule_link_vmlinux +define rule_vmlinux set -e echo ' Generating build number' - . scripts/mkversion > .tmpversion - mv -f .tmpversion .version + . scripts/mkversion > .tmp_version + mv -f .tmp_version .version +$(MAKE) -C init $(call cmd,link_vmlinux) echo 'cmd_$@ := $(cmd_link_vmlinux)' > $(@D)/.$(@F).cmd - $(NM) vmlinux | grep -v '\(compiled\)\|\(\.o$$\)\|\( [aUw] \)\|\(\.\.ng$$\)\|\(LASH[RL]DI\)' | sort > System.map + $(NM) $@ | grep -v '\(compiled\)\|\(\.o$$\)\|\( [aUw] \)\|\(\.\.ng$$\)\|\(LASH[RL]DI\)' | sort > System.map endef LDFLAGS_vmlinux += -T arch/$(ARCH)/vmlinux.lds.s -vmlinux: $(vmlinux-objs) arch/$(ARCH)/vmlinux.lds.s FORCE - $(call if_changed_rule,link_vmlinux) +# Generate section listing all symbols and add it into vmlinux + +ifdef CONFIG_KALLSYMS + +kallsyms.o := .tmp_kallsyms.o + +quiet_cmd_kallsyms = KSYM $@ +cmd_kallsyms = $(KALLSYMS) $< > $@ + +.tmp_kallsyms.o: .tmp_vmlinux + $(call cmd,kallsyms) + +# After generating .tmp_vmlinux just like vmlinux, decrement the version +# number again, so the final vmlinux gets the same one. +# Ignore return value of 'expr'. + +define rule_.tmp_vmlinux + $(rule_vmlinux) + if expr 0`cat .version` - 1 > .tmp_version; then true; fi + mv -f .tmp_version .version +endef + +.tmp_vmlinux: $(vmlinux-objs) arch/$(ARCH)/vmlinux.lds.s FORCE + $(call if_changed_rule,.tmp_vmlinux) + +endif + +# Finally the vmlinux rule + +vmlinux: $(vmlinux-objs) $(kallsyms.o) arch/$(ARCH)/vmlinux.lds.s FORCE + $(call if_changed_rule,vmlinux) # The actual objects are generated when descending, # make sure no implicit rule kicks in @@ -820,7 +853,7 @@ # FIXME Should go into a make.lib or something # =========================================================================== -echo_target = $(RELDIR)/$@ +echo_target = $@ a_flags = -Wp,-MD,$(depfile) $(AFLAGS) $(NOSTDINC_FLAGS) \ $(modkern_aflags) $(EXTRA_AFLAGS) $(AFLAGS_$(*F).o) ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:56 ` Ingo Molnar @ 2002-09-25 22:04 ` J.A. Magallon 2002-09-27 20:48 ` Bill Davidsen 2002-09-26 17:16 ` Ruth Ivimey-Cook 1 sibling, 1 reply; 23+ messages in thread From: J.A. Magallon @ 2002-09-25 22:04 UTC (permalink / raw) To: Ingo Molnar Cc: Linus Torvalds, Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On 2002.09.25 Ingo Molnar wrote: > >okay, here is the new oops output: > Sorry to be picky, but... >------------[ cut here ]------------ >kernel BUG at time.c:99! >invalid operand: 0000 >CPU: 1 >EIP: 0060:[<c011bd14>] Not tainted >EFLAGS: 00010246 >EIP is at sys_gettimeofday+0x84/0x90 >eax: 0000004e ebx: cef9e000 ecx: 00000000 edx: 00000068 >esi: 00000000 edi: 00000000 ebp: bffffad8 esp: cef9ffa0 >ds: 0068 es: 0068 ss: 0068 >Process gettimeofday (pid: 549, threadinfo=cef9e000 task=cf84d860) >Stack: 4001695c bffff414 40156154 00000004 c0112a40 cef9e000 400168e4 bffffb44 > c0107973 00000000 00000000 40156154 400168e4 bffffb44 bffffad8 0000004e > 0000002b 0000002b 0000004e 400cecc1 00000023 00000246 bffffacc 0000002b >Call Trace: [<c0112a40>] do_page_fault+0x0/0x4a2 >[<c0107973>] syscall_call+0x7/0xb > ...wouldn't this look much nicer with some '\n' and a couple space for indent ? Stack: 4001695c bffff414 40156154 00000004 c0112a40 cef9e000 400168e4 bffffb44 c0107973 00000000 00000000 40156154 400168e4 bffffb44 bffffad8 0000004e 0000002b 0000002b 0000004e 400cecc1 00000023 00000246 bffffacc 0000002b Call Trace: [<c0112a40>] do_page_fault+0x0/0x4a2 [<c0107973>] syscall_call+0x7/0xb -- J.A. Magallon <jamagallon@able.es> \ Software is like sex: werewolf.able.es \ It's better when it's free Mandrake Linux release 9.0 (Cooker) for i586 Linux 2.4.20-pre7-jam0 (gcc 3.2 (Mandrake Linux 9.0 3.2-1mdk)) ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 22:04 ` J.A. Magallon @ 2002-09-27 20:48 ` Bill Davidsen 2002-09-27 21:13 ` Ingo Molnar 0 siblings, 1 reply; 23+ messages in thread From: Bill Davidsen @ 2002-09-27 20:48 UTC (permalink / raw) To: J.A. Magallon Cc: Ingo Molnar, Linus Torvalds, Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On Thu, 26 Sep 2002, J.A. Magallon wrote: > Sorry to be picky, but... > ...wouldn't this look much nicer with some '\n' and a couple space for indent ? > > Stack: > 4001695c bffff414 40156154 00000004 c0112a40 cef9e000 400168e4 bffffb44 > c0107973 00000000 00000000 40156154 400168e4 bffffb44 bffffad8 0000004e > 0000002b 0000002b 0000004e 400cecc1 00000023 00000246 bffffacc 0000002b > Call Trace: > [<c0112a40>] do_page_fault+0x0/0x4a2 > [<c0107973>] syscall_call+0x7/0xb You mean like making it human readable? ;-) I think that's a great idea. -- bill davidsen <davidsen@tmr.com> CTO, TMR Associates, Inc Doing interesting things with little computers since 1979. ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-27 20:48 ` Bill Davidsen @ 2002-09-27 21:13 ` Ingo Molnar 2002-09-27 23:58 ` J.A. Magallon 0 siblings, 1 reply; 23+ messages in thread From: Ingo Molnar @ 2002-09-27 21:13 UTC (permalink / raw) To: Bill Davidsen Cc: J.A. Magallon, Linus Torvalds, Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven > > Call Trace: > > [<c0112a40>] do_page_fault+0x0/0x4a2 > > [<c0107973>] syscall_call+0x7/0xb > I think that's a great idea. it's already in the kksymoops patch that Linus applied two days ago. Ingo ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-27 21:13 ` Ingo Molnar @ 2002-09-27 23:58 ` J.A. Magallon 0 siblings, 0 replies; 23+ messages in thread From: J.A. Magallon @ 2002-09-27 23:58 UTC (permalink / raw) To: Ingo Molnar Cc: Bill Davidsen, J.A. Magallon, Linus Torvalds, Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On 2002.09.27 Ingo Molnar wrote: > >> > Call Trace: >> > [<c0112a40>] do_page_fault+0x0/0x4a2 >> > [<c0107973>] syscall_call+0x7/0xb > >> I think that's a great idea. > >it's already in the kksymoops patch that Linus applied two days ago. > Any version of this for 2.4 ? I think this is going to be very, very usefull, so it is interesting to have it on 2.4... -- J.A. Magallon <jamagallon@able.es> \ Software is like sex: werewolf.able.es \ It's better when it's free Mandrake Linux release 9.0 (Cooker) for i586 Linux 2.4.20-pre8-jam0 (gcc 3.2 (Mandrake Linux 9.0 3.2-1mdk)) ^ permalink raw reply [flat|nested] 23+ messages in thread
* Re: [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 2002-09-25 19:56 ` Ingo Molnar 2002-09-25 22:04 ` J.A. Magallon @ 2002-09-26 17:16 ` Ruth Ivimey-Cook 1 sibling, 0 replies; 23+ messages in thread From: Ruth Ivimey-Cook @ 2002-09-26 17:16 UTC (permalink / raw) To: Ingo Molnar, Linus Torvalds Cc: Kai Germaschewski, linux-kernel, Rusty Russell, Arjan van de Ven On Wednesday 25 September 2002 20:56, Ingo Molnar wrote: > ------------[ cut here ]------------ > kernel BUG at time.c:99! > invalid operand: 0000 > CPU: 1 > EIP: 0060:[<c011bd14>] Not tainted > EFLAGS: 00010246 > EIP is at sys_gettimeofday+0x84/0x90 > eax: 0000004e ebx: cef9e000 ecx: 00000000 edx: 00000068 > esi: 00000000 edi: 00000000 ebp: bffffad8 esp: cef9ffa0 > ds: 0068 es: 0068 ss: 0068 > Process gettimeofday (pid: 549, threadinfo=cef9e000 task=cf84d860) > Stack: 4001695c bffff414 40156154 00000004 c0112a40 cef9e000 400168e4 Something that's been bugging me of late: the Oops output is quite expensive in lines on the terminal, which means if you get a couple of oops before the kernel stops, you're unlikely to see the one that fired first. To help with this, would it be good to use the following form for the initial part: kernel BUG at time.c:99, invalid operand: 0000 CPU 1: EIP: 0060:[<c011bd14>] Not tainted EIP is at sys_gettimeofday+0x84/0x90 eax: 0000004e ebx: cef9e000 ecx: 00000000 edx: 00000068 esi: 00000000 edi: 00000000 ebp: bffffad8 esp: cef9ffa0 ds: 0068 es: 0068 ss: 0068 eflags: 00010246 [textflags] Where [textflags] is some arch-specific representation of the flags word. Also, in the same vein I would like to be able to say (as a kernel option): if kernel-oops { copy console output to [printer|serial] port } [printer output == ascii only, of course] Regards, Ruth ^ permalink raw reply [flat|nested] 23+ messages in thread
end of thread, other threads:[~2002-09-27 23:56 UTC | newest] Thread overview: 23+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2002-09-25 9:02 [ANNOUNCE] [patch] kksymoops, in-kernel symbolic oopser, 2.5.38-B0 Ingo Molnar 2002-09-25 9:32 ` [re-ANNOUNCE] " Ingo Molnar 2002-09-25 9:51 ` Arnaldo Carvalho de Melo 2002-09-25 18:16 ` [ANNOUNCE] " Linus Torvalds 2002-09-25 19:23 ` Kai Germaschewski 2002-09-25 19:42 ` Ingo Molnar 2002-09-25 19:41 ` Kai Germaschewski 2002-09-25 19:46 ` Linus Torvalds 2002-09-25 19:54 ` Ingo Molnar 2002-09-25 19:52 ` Linus Torvalds 2002-09-25 20:04 ` Ingo Molnar 2002-09-25 20:31 ` Jeff Garzik 2002-09-25 19:55 ` Kai Germaschewski 2002-09-25 19:45 ` Linus Torvalds 2002-09-25 19:53 ` Ingo Molnar 2002-09-25 19:04 ` Cort Dougan 2002-09-25 20:14 ` Ingo Molnar 2002-09-25 19:56 ` Ingo Molnar 2002-09-25 22:04 ` J.A. Magallon 2002-09-27 20:48 ` Bill Davidsen 2002-09-27 21:13 ` Ingo Molnar 2002-09-27 23:58 ` J.A. Magallon 2002-09-26 17:16 ` Ruth Ivimey-Cook
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).