diff options
Diffstat (limited to 'arch/avr32/kernel')
26 files changed, 2131 insertions, 1845 deletions
diff --git a/arch/avr32/kernel/.gitignore b/arch/avr32/kernel/.gitignore new file mode 100644 index 00000000000..c5f676c3c22 --- /dev/null +++ b/arch/avr32/kernel/.gitignore @@ -0,0 +1 @@ +vmlinux.lds diff --git a/arch/avr32/kernel/Makefile b/arch/avr32/kernel/Makefile index 90e5afff54a..119a2e41def 100644 --- a/arch/avr32/kernel/Makefile +++ b/arch/avr32/kernel/Makefile @@ -6,13 +6,10 @@ extra-y := head.o vmlinux.lds obj-$(CONFIG_SUBARCH_AVR32B) += entry-avr32b.o obj-y += syscall_table.o syscall-stubs.o irq.o -obj-y += setup.o traps.o semaphore.o ptrace.o -obj-y += signal.o sys_avr32.o process.o time.o -obj-y += init_task.o switch_to.o cpu.o +obj-y += setup.o traps.o ocd.o ptrace.o +obj-y += signal.o process.o time.o +obj-y += switch_to.o cpu.o obj-$(CONFIG_MODULES) += module.o avr32_ksyms.o obj-$(CONFIG_KPROBES) += kprobes.o - -USE_STANDARD_AS_RULE := true - -%.lds: %.lds.c FORCE - $(call if_changed_dep,cpp_lds_S) +obj-$(CONFIG_STACKTRACE) += stacktrace.o +obj-$(CONFIG_NMI_DEBUGGING) += nmi_debug.o diff --git a/arch/avr32/kernel/asm-offsets.c b/arch/avr32/kernel/asm-offsets.c index 97d86586566..d6a8193a1d2 100644 --- a/arch/avr32/kernel/asm-offsets.c +++ b/arch/avr32/kernel/asm-offsets.c @@ -4,15 +4,10 @@ * to extract and format the required data. */ +#include <linux/mm.h> +#include <linux/sched.h> #include <linux/thread_info.h> - -#define DEFINE(sym, val) \ - asm volatile("\n->" #sym " %0 " #val : : "i" (val)) - -#define BLANK() asm volatile("\n->" : : ) - -#define OFFSET(sym, str, mem) \ - DEFINE(sym, offsetof(struct str, mem)); +#include <linux/kbuild.h> void foo(void) { @@ -21,5 +16,11 @@ void foo(void) OFFSET(TI_flags, thread_info, flags); OFFSET(TI_cpu, thread_info, cpu); OFFSET(TI_preempt_count, thread_info, preempt_count); + OFFSET(TI_rar_saved, thread_info, rar_saved); + OFFSET(TI_rsr_saved, thread_info, rsr_saved); OFFSET(TI_restart_block, thread_info, restart_block); + BLANK(); + OFFSET(TSK_active_mm, task_struct, active_mm); + BLANK(); + OFFSET(MM_pgd, mm_struct, pgd); } diff --git a/arch/avr32/kernel/avr32_ksyms.c b/arch/avr32/kernel/avr32_ksyms.c index 04f767a272b..d93ead02dae 100644 --- a/arch/avr32/kernel/avr32_ksyms.c +++ b/arch/avr32/kernel/avr32_ksyms.c @@ -7,11 +7,12 @@ * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. */ +#include <linux/delay.h> +#include <linux/io.h> #include <linux/module.h> #include <asm/checksum.h> #include <asm/uaccess.h> -#include <asm/delay.h> /* * GCC functions @@ -29,6 +30,9 @@ EXPORT_SYMBOL(__avr32_asr64); EXPORT_SYMBOL(memset); EXPORT_SYMBOL(memcpy); +EXPORT_SYMBOL(clear_page); +EXPORT_SYMBOL(copy_page); + /* * Userspace access stuff. */ @@ -39,6 +43,8 @@ EXPORT_SYMBOL(strncpy_from_user); EXPORT_SYMBOL(__strncpy_from_user); EXPORT_SYMBOL(clear_user); EXPORT_SYMBOL(__clear_user); +EXPORT_SYMBOL(strnlen_user); + EXPORT_SYMBOL(csum_partial); EXPORT_SYMBOL(csum_partial_copy_generic); @@ -52,4 +58,13 @@ EXPORT_SYMBOL(find_first_zero_bit); EXPORT_SYMBOL(find_next_zero_bit); EXPORT_SYMBOL(find_first_bit); EXPORT_SYMBOL(find_next_bit); -EXPORT_SYMBOL(generic_find_next_zero_le_bit); +EXPORT_SYMBOL(find_next_bit_le); +EXPORT_SYMBOL(find_next_zero_bit_le); + +/* I/O primitives (lib/io-*.S) */ +EXPORT_SYMBOL(__raw_readsb); +EXPORT_SYMBOL(__raw_readsw); +EXPORT_SYMBOL(__raw_readsl); +EXPORT_SYMBOL(__raw_writesb); +EXPORT_SYMBOL(__raw_writesw); +EXPORT_SYMBOL(__raw_writesl); diff --git a/arch/avr32/kernel/cpu.c b/arch/avr32/kernel/cpu.c index 342452ba204..0341ae27c9e 100644 --- a/arch/avr32/kernel/cpu.c +++ b/arch/avr32/kernel/cpu.c @@ -6,12 +6,14 @@ * published by the Free Software Foundation. */ #include <linux/init.h> -#include <linux/sysdev.h> +#include <linux/device.h> #include <linux/seq_file.h> #include <linux/cpu.h> +#include <linux/module.h> #include <linux/percpu.h> #include <linux/param.h> #include <linux/errno.h> +#include <linux/clk.h> #include <asm/setup.h> #include <asm/sysreg.h> @@ -24,125 +26,141 @@ static DEFINE_PER_CPU(struct cpu, cpu_devices); * XXX: If/when a SMP-capable implementation of AVR32 will ever be * made, we must make sure that the code executes on the correct CPU. */ -static ssize_t show_pc0event(struct sys_device *dev, char *buf) +static ssize_t show_pc0event(struct device *dev, + struct device_attribute *attr, char *buf) { unsigned long pccr; pccr = sysreg_read(PCCR); return sprintf(buf, "0x%lx\n", (pccr >> 12) & 0x3f); } -static ssize_t store_pc0event(struct sys_device *dev, const char *buf, +static ssize_t store_pc0event(struct device *dev, + struct device_attribute *attr, const char *buf, size_t count) { unsigned long val; - char *endp; + int ret; - val = simple_strtoul(buf, &endp, 0); - if (endp == buf || val > 0x3f) + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; + if (val > 0x3f) return -EINVAL; val = (val << 12) | (sysreg_read(PCCR) & 0xfffc0fff); sysreg_write(PCCR, val); return count; } -static ssize_t show_pc0count(struct sys_device *dev, char *buf) +static ssize_t show_pc0count(struct device *dev, + struct device_attribute *attr, char *buf) { unsigned long pcnt0; pcnt0 = sysreg_read(PCNT0); return sprintf(buf, "%lu\n", pcnt0); } -static ssize_t store_pc0count(struct sys_device *dev, const char *buf, - size_t count) +static ssize_t store_pc0count(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) { unsigned long val; - char *endp; + int ret; - val = simple_strtoul(buf, &endp, 0); - if (endp == buf) - return -EINVAL; + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; sysreg_write(PCNT0, val); return count; } -static ssize_t show_pc1event(struct sys_device *dev, char *buf) +static ssize_t show_pc1event(struct device *dev, + struct device_attribute *attr, char *buf) { unsigned long pccr; pccr = sysreg_read(PCCR); return sprintf(buf, "0x%lx\n", (pccr >> 18) & 0x3f); } -static ssize_t store_pc1event(struct sys_device *dev, const char *buf, +static ssize_t store_pc1event(struct device *dev, + struct device_attribute *attr, const char *buf, size_t count) { unsigned long val; - char *endp; + int ret; - val = simple_strtoul(buf, &endp, 0); - if (endp == buf || val > 0x3f) + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; + if (val > 0x3f) return -EINVAL; val = (val << 18) | (sysreg_read(PCCR) & 0xff03ffff); sysreg_write(PCCR, val); return count; } -static ssize_t show_pc1count(struct sys_device *dev, char *buf) +static ssize_t show_pc1count(struct device *dev, + struct device_attribute *attr, char *buf) { unsigned long pcnt1; pcnt1 = sysreg_read(PCNT1); return sprintf(buf, "%lu\n", pcnt1); } -static ssize_t store_pc1count(struct sys_device *dev, const char *buf, +static ssize_t store_pc1count(struct device *dev, + struct device_attribute *attr, const char *buf, size_t count) { unsigned long val; - char *endp; + int ret; - val = simple_strtoul(buf, &endp, 0); - if (endp == buf) - return -EINVAL; + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; sysreg_write(PCNT1, val); return count; } -static ssize_t show_pccycles(struct sys_device *dev, char *buf) +static ssize_t show_pccycles(struct device *dev, + struct device_attribute *attr, char *buf) { unsigned long pccnt; pccnt = sysreg_read(PCCNT); return sprintf(buf, "%lu\n", pccnt); } -static ssize_t store_pccycles(struct sys_device *dev, const char *buf, +static ssize_t store_pccycles(struct device *dev, + struct device_attribute *attr, const char *buf, size_t count) { unsigned long val; - char *endp; + int ret; - val = simple_strtoul(buf, &endp, 0); - if (endp == buf) - return -EINVAL; + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; sysreg_write(PCCNT, val); return count; } -static ssize_t show_pcenable(struct sys_device *dev, char *buf) +static ssize_t show_pcenable(struct device *dev, + struct device_attribute *attr, char *buf) { unsigned long pccr; pccr = sysreg_read(PCCR); return sprintf(buf, "%c\n", (pccr & 1)?'1':'0'); } -static ssize_t store_pcenable(struct sys_device *dev, const char *buf, +static ssize_t store_pcenable(struct device *dev, + struct device_attribute *attr, const char *buf, size_t count) { unsigned long pccr, val; - char *endp; + int ret; - val = simple_strtoul(buf, &endp, 0); - if (endp == buf) - return -EINVAL; + ret = kstrtoul(buf, 0, &val); + if (ret) + return ret; if (val) val = 1; @@ -153,12 +171,12 @@ static ssize_t store_pcenable(struct sys_device *dev, const char *buf, return count; } -static SYSDEV_ATTR(pc0event, 0600, show_pc0event, store_pc0event); -static SYSDEV_ATTR(pc0count, 0600, show_pc0count, store_pc0count); -static SYSDEV_ATTR(pc1event, 0600, show_pc1event, store_pc1event); -static SYSDEV_ATTR(pc1count, 0600, show_pc1count, store_pc1count); -static SYSDEV_ATTR(pccycles, 0600, show_pccycles, store_pccycles); -static SYSDEV_ATTR(pcenable, 0600, show_pcenable, store_pcenable); +static DEVICE_ATTR(pc0event, 0600, show_pc0event, store_pc0event); +static DEVICE_ATTR(pc0count, 0600, show_pc0count, store_pc0count); +static DEVICE_ATTR(pc1event, 0600, show_pc1event, store_pc1event); +static DEVICE_ATTR(pc1count, 0600, show_pc1count, store_pc1count); +static DEVICE_ATTR(pccycles, 0600, show_pccycles, store_pccycles); +static DEVICE_ATTR(pcenable, 0600, show_pcenable, store_pcenable); #endif /* CONFIG_PERFORMANCE_COUNTERS */ @@ -172,12 +190,12 @@ static int __init topology_init(void) register_cpu(c, cpu); #ifdef CONFIG_PERFORMANCE_COUNTERS - sysdev_create_file(&c->sysdev, &attr_pc0event); - sysdev_create_file(&c->sysdev, &attr_pc0count); - sysdev_create_file(&c->sysdev, &attr_pc1event); - sysdev_create_file(&c->sysdev, &attr_pc1count); - sysdev_create_file(&c->sysdev, &attr_pccycles); - sysdev_create_file(&c->sysdev, &attr_pcenable); + device_create_file(&c->dev, &dev_attr_pc0event); + device_create_file(&c->dev, &dev_attr_pc0count); + device_create_file(&c->dev, &dev_attr_pc1event); + device_create_file(&c->dev, &dev_attr_pc1count); + device_create_file(&c->dev, &dev_attr_pccycles); + device_create_file(&c->dev, &dev_attr_pcenable); #endif } @@ -186,9 +204,20 @@ static int __init topology_init(void) subsys_initcall(topology_init); +struct chip_id_map { + u16 mid; + u16 pn; + const char *name; +}; + +static const struct chip_id_map chip_names[] = { + { .mid = 0x1f, .pn = 0x1e82, .name = "AT32AP700x" }, +}; +#define NR_CHIP_NAMES ARRAY_SIZE(chip_names) + static const char *cpu_names[] = { "Morgan", - "AP7000", + "AP7", }; #define NR_CPU_NAMES ARRAY_SIZE(cpu_names) @@ -205,36 +234,60 @@ static const char *mmu_types[] = { "MPU" }; +static const char *cpu_feature_flags[] = { + "rmw", "dsp", "simd", "ocd", "perfctr", "java", "fpu", +}; + +static const char *get_chip_name(struct avr32_cpuinfo *cpu) +{ + unsigned int i; + unsigned int mid = avr32_get_manufacturer_id(cpu); + unsigned int pn = avr32_get_product_number(cpu); + + for (i = 0; i < NR_CHIP_NAMES; i++) { + if (chip_names[i].mid == mid && chip_names[i].pn == pn) + return chip_names[i].name; + } + + return "(unknown)"; +} + void __init setup_processor(void) { unsigned long config0, config1; + unsigned long features; unsigned cpu_id, cpu_rev, arch_id, arch_rev, mmu_type; + unsigned device_id; unsigned tmp; + unsigned i; + + config0 = sysreg_read(CONFIG0); + config1 = sysreg_read(CONFIG1); + cpu_id = SYSREG_BFEXT(PROCESSORID, config0); + cpu_rev = SYSREG_BFEXT(PROCESSORREVISION, config0); + arch_id = SYSREG_BFEXT(AT, config0); + arch_rev = SYSREG_BFEXT(AR, config0); + mmu_type = SYSREG_BFEXT(MMUT, config0); - config0 = sysreg_read(CONFIG0); /* 0x0000013e; */ - config1 = sysreg_read(CONFIG1); /* 0x01f689a2; */ - cpu_id = config0 >> 24; - cpu_rev = (config0 >> 16) & 0xff; - arch_id = (config0 >> 13) & 0x07; - arch_rev = (config0 >> 10) & 0x07; - mmu_type = (config0 >> 7) & 0x03; + device_id = ocd_read(DID); boot_cpu_data.arch_type = arch_id; boot_cpu_data.cpu_type = cpu_id; boot_cpu_data.arch_revision = arch_rev; boot_cpu_data.cpu_revision = cpu_rev; boot_cpu_data.tlb_config = mmu_type; + boot_cpu_data.device_id = device_id; - tmp = (config1 >> 13) & 0x07; + tmp = SYSREG_BFEXT(ILSZ, config1); if (tmp) { - boot_cpu_data.icache.ways = 1 << ((config1 >> 10) & 0x07); - boot_cpu_data.icache.sets = 1 << ((config1 >> 16) & 0x0f); + boot_cpu_data.icache.ways = 1 << SYSREG_BFEXT(IASS, config1); + boot_cpu_data.icache.sets = 1 << SYSREG_BFEXT(ISET, config1); boot_cpu_data.icache.linesz = 1 << (tmp + 1); } - tmp = (config1 >> 3) & 0x07; + tmp = SYSREG_BFEXT(DLSZ, config1); if (tmp) { - boot_cpu_data.dcache.ways = 1 << (config1 & 0x07); - boot_cpu_data.dcache.sets = 1 << ((config1 >> 6) & 0x0f); + boot_cpu_data.dcache.ways = 1 << SYSREG_BFEXT(DASS, config1); + boot_cpu_data.dcache.sets = 1 << SYSREG_BFEXT(DSET, config1); boot_cpu_data.dcache.linesz = 1 << (tmp + 1); } @@ -245,20 +298,36 @@ void __init setup_processor(void) return; } - printk ("CPU: %s [%02x] revision %d (%s revision %d)\n", + printk ("CPU: %s chip revision %c\n", get_chip_name(&boot_cpu_data), + avr32_get_chip_revision(&boot_cpu_data) + 'A'); + printk ("CPU: %s [%02x] core revision %d (%s arch revision %d)\n", cpu_names[cpu_id], cpu_id, cpu_rev, arch_names[arch_id], arch_rev); printk ("CPU: MMU configuration: %s\n", mmu_types[mmu_type]); + printk ("CPU: features:"); - if (config0 & (1 << 6)) - printk(" fpu"); - if (config0 & (1 << 5)) - printk(" java"); - if (config0 & (1 << 4)) - printk(" perfctr"); - if (config0 & (1 << 3)) - printk(" ocd"); + features = 0; + if (config0 & SYSREG_BIT(CONFIG0_R)) + features |= AVR32_FEATURE_RMW; + if (config0 & SYSREG_BIT(CONFIG0_D)) + features |= AVR32_FEATURE_DSP; + if (config0 & SYSREG_BIT(CONFIG0_S)) + features |= AVR32_FEATURE_SIMD; + if (config0 & SYSREG_BIT(CONFIG0_O)) + features |= AVR32_FEATURE_OCD; + if (config0 & SYSREG_BIT(CONFIG0_P)) + features |= AVR32_FEATURE_PCTR; + if (config0 & SYSREG_BIT(CONFIG0_J)) + features |= AVR32_FEATURE_JAVA; + if (config0 & SYSREG_BIT(CONFIG0_F)) + features |= AVR32_FEATURE_FPU; + + for (i = 0; i < ARRAY_SIZE(cpu_feature_flags); i++) + if (features & (1 << i)) + printk(" %s", cpu_feature_flags[i]); + printk("\n"); + boot_cpu_data.features = features; } #ifdef CONFIG_PROC_FS @@ -266,6 +335,8 @@ static int c_show(struct seq_file *m, void *v) { unsigned int icache_size, dcache_size; unsigned int cpu = smp_processor_id(); + unsigned int freq; + unsigned int i; icache_size = boot_cpu_data.icache.ways * boot_cpu_data.icache.sets * @@ -276,15 +347,21 @@ static int c_show(struct seq_file *m, void *v) seq_printf(m, "processor\t: %d\n", cpu); + seq_printf(m, "chip type\t: %s revision %c\n", + get_chip_name(&boot_cpu_data), + avr32_get_chip_revision(&boot_cpu_data) + 'A'); if (boot_cpu_data.arch_type < NR_ARCH_NAMES) - seq_printf(m, "cpu family\t: %s revision %d\n", + seq_printf(m, "cpu arch\t: %s revision %d\n", arch_names[boot_cpu_data.arch_type], boot_cpu_data.arch_revision); if (boot_cpu_data.cpu_type < NR_CPU_NAMES) - seq_printf(m, "cpu type\t: %s revision %d\n", + seq_printf(m, "cpu core\t: %s revision %d\n", cpu_names[boot_cpu_data.cpu_type], boot_cpu_data.cpu_revision); + freq = (clk_get_rate(boot_cpu_data.clk) + 500) / 1000; + seq_printf(m, "cpu MHz\t\t: %u.%03u\n", freq / 1000, freq % 1000); + seq_printf(m, "i-cache\t\t: %dK (%u ways x %u sets x %u)\n", icache_size >> 10, boot_cpu_data.icache.ways, @@ -295,7 +372,13 @@ static int c_show(struct seq_file *m, void *v) boot_cpu_data.dcache.ways, boot_cpu_data.dcache.sets, boot_cpu_data.dcache.linesz); - seq_printf(m, "bogomips\t: %lu.%02lu\n", + + seq_printf(m, "features\t:"); + for (i = 0; i < ARRAY_SIZE(cpu_feature_flags); i++) + if (boot_cpu_data.features & (1 << i)) + seq_printf(m, " %s", cpu_feature_flags[i]); + + seq_printf(m, "\nbogomips\t: %lu.%02lu\n", boot_cpu_data.loops_per_jiffy / (500000/HZ), (boot_cpu_data.loops_per_jiffy / (5000/HZ)) % 100); @@ -318,7 +401,7 @@ static void c_stop(struct seq_file *m, void *v) } -struct seq_operations cpuinfo_op = { +const struct seq_operations cpuinfo_op = { .start = c_start, .next = c_next, .stop = c_stop, diff --git a/arch/avr32/kernel/entry-avr32b.S b/arch/avr32/kernel/entry-avr32b.S index eeb66792bc3..7301f4806bb 100644 --- a/arch/avr32/kernel/entry-avr32b.S +++ b/arch/avr32/kernel/entry-avr32b.S @@ -74,102 +74,133 @@ exception_vectors: .align 2 bral do_dtlb_modified - /* - * r0 : PGD/PT/PTE - * r1 : Offending address - * r2 : Scratch register - * r3 : Cause (5, 12 or 13) - */ #define tlbmiss_save pushm r0-r3 #define tlbmiss_restore popm r0-r3 - .section .tlbx.ex.text,"ax",@progbits + .org 0x50 .global itlb_miss itlb_miss: tlbmiss_save rjmp tlb_miss_common - .section .tlbr.ex.text,"ax",@progbits + .org 0x60 dtlb_miss_read: tlbmiss_save rjmp tlb_miss_common - .section .tlbw.ex.text,"ax",@progbits + .org 0x70 dtlb_miss_write: tlbmiss_save .global tlb_miss_common + .align 2 tlb_miss_common: - mfsr r0, SYSREG_PTBR - mfsr r1, SYSREG_TLBEAR - - /* Is it the vmalloc space? */ - bld r1, 31 - brcs handle_vmalloc_miss + mfsr r0, SYSREG_TLBEAR + mfsr r1, SYSREG_PTBR - /* First level lookup */ + /* + * First level lookup: The PGD contains virtual pointers to + * the second-level page tables, but they may be NULL if not + * present. + */ pgtbl_lookup: - lsr r2, r1, PGDIR_SHIFT - ld.w r0, r0[r2 << 2] - bld r0, _PAGE_BIT_PRESENT - brcc page_table_not_present - - /* TODO: Check access rights on page table if necessary */ - - /* Translate to virtual address in P1. */ - andl r0, 0xf000 - sbr r0, 31 + lsr r2, r0, PGDIR_SHIFT + ld.w r3, r1[r2 << 2] + bfextu r1, r0, PAGE_SHIFT, PGDIR_SHIFT - PAGE_SHIFT + cp.w r3, 0 + breq page_table_not_present /* Second level lookup */ - lsl r1, (32 - PGDIR_SHIFT) - lsr r1, (32 - PGDIR_SHIFT) + PAGE_SHIFT - add r2, r0, r1 << 2 - ld.w r1, r2[0] - bld r1, _PAGE_BIT_PRESENT + ld.w r2, r3[r1 << 2] + mfsr r0, SYSREG_TLBARLO + bld r2, _PAGE_BIT_PRESENT brcc page_not_present /* Mark the page as accessed */ - sbr r1, _PAGE_BIT_ACCESSED - st.w r2[0], r1 + sbr r2, _PAGE_BIT_ACCESSED + st.w r3[r1 << 2], r2 /* Drop software flags */ - andl r1, _PAGE_FLAGS_HARDWARE_MASK & 0xffff - mtsr SYSREG_TLBELO, r1 + andl r2, _PAGE_FLAGS_HARDWARE_MASK & 0xffff + mtsr SYSREG_TLBELO, r2 /* Figure out which entry we want to replace */ - mfsr r0, SYSREG_TLBARLO + mfsr r1, SYSREG_MMUCR clz r2, r0 brcc 1f - mov r1, -1 /* All entries have been accessed, */ - mtsr SYSREG_TLBARLO, r1 /* so reset TLBAR */ - mov r2, 0 /* and start at 0 */ -1: mfsr r1, SYSREG_MMUCR - lsl r2, 14 - andl r1, 0x3fff, COH - or r1, r2 - mtsr SYSREG_MMUCR, r1 + mov r3, -1 /* All entries have been accessed, */ + mov r2, 0 /* so start at 0 */ + mtsr SYSREG_TLBARLO, r3 /* and reset TLBAR */ +1: bfins r1, r2, SYSREG_DRP_OFFSET, SYSREG_DRP_SIZE + mtsr SYSREG_MMUCR, r1 tlbw tlbmiss_restore rete -handle_vmalloc_miss: - /* Simply do the lookup in init's page table */ - mov r0, lo(swapper_pg_dir) - orh r0, hi(swapper_pg_dir) + /* The slow path of the TLB miss handler */ + .align 2 +page_table_not_present: + /* Do we need to synchronize with swapper_pg_dir? */ + bld r0, 31 + brcs sync_with_swapper_pg_dir + +page_not_present: + tlbmiss_restore + sub sp, 4 + stmts --sp, r0-lr + call save_full_context_ex + mfsr r12, SYSREG_ECR + mov r11, sp + call do_page_fault + rjmp ret_from_exception + + .align 2 +sync_with_swapper_pg_dir: + /* + * If swapper_pg_dir contains a non-NULL second-level page + * table pointer, copy it into the current PGD. If not, we + * must handle it as a full-blown page fault. + * + * Jumping back to pgtbl_lookup causes an unnecessary lookup, + * but it is guaranteed to be a cache hit, it won't happen + * very often, and we absolutely do not want to sacrifice any + * performance in the fast path in order to improve this. + */ + mov r1, lo(swapper_pg_dir) + orh r1, hi(swapper_pg_dir) + ld.w r3, r1[r2 << 2] + cp.w r3, 0 + breq page_not_present + mfsr r1, SYSREG_PTBR + st.w r1[r2 << 2], r3 rjmp pgtbl_lookup + /* + * We currently have two bytes left at this point until we + * crash into the system call handler... + * + * Don't worry, the assembler will let us know. + */ + /* --- System Call --- */ - .section .scall.text,"ax",@progbits + .org 0x100 system_call: +#ifdef CONFIG_PREEMPT + mask_interrupts +#endif pushm r12 /* r12_orig */ stmts --sp, r0-lr - zero_fp + mfsr r0, SYSREG_RAR_SUP mfsr r1, SYSREG_RSR_SUP +#ifdef CONFIG_PREEMPT + unmask_interrupts +#endif + zero_fp stm --sp, r0-r1 /* check for syscall tracing */ @@ -219,18 +250,20 @@ syscall_badsys: .global ret_from_fork ret_from_fork: - rcall schedule_tail + call schedule_tail + mov r12, 0 + rjmp syscall_return - /* check for syscall tracing */ - get_thread_info r0 - ld.w r1, r0[TI_flags] - andl r1, _TIF_ALLWORK_MASK, COH - brne syscall_exit_work - rjmp syscall_exit_cont + .global ret_from_kernel_thread +ret_from_kernel_thread: + call schedule_tail + mov r12, r0 + mov lr, r2 /* syscall_return */ + mov pc, r1 syscall_trace_enter: pushm r8-r12 - rcall syscall_trace + call syscall_trace popm r8-r12 rjmp syscall_trace_cont @@ -238,60 +271,44 @@ syscall_exit_work: bld r1, TIF_SYSCALL_TRACE brcc 1f unmask_interrupts - rcall syscall_trace + call syscall_trace mask_interrupts ld.w r1, r0[TI_flags] 1: bld r1, TIF_NEED_RESCHED brcc 2f unmask_interrupts - rcall schedule + call schedule mask_interrupts ld.w r1, r0[TI_flags] rjmp 1b -2: mov r2, _TIF_SIGPENDING | _TIF_RESTORE_SIGMASK +2: mov r2, _TIF_SIGPENDING | _TIF_NOTIFY_RESUME tst r1, r2 breq 3f unmask_interrupts mov r12, sp mov r11, r0 - rcall do_notify_resume + call do_notify_resume mask_interrupts ld.w r1, r0[TI_flags] rjmp 1b 3: bld r1, TIF_BREAKPOINT brcc syscall_exit_cont - mfsr r3, SYSREG_TLBEHI - lddsp r2, sp[REG_PC] - andl r3, 0xff, COH - lsl r3, 1 - sbr r3, 30 - sbr r3, 0 - mtdr DBGREG_BWA2A, r2 - mtdr DBGREG_BWC2A, r3 - rjmp syscall_exit_cont - - - /* The slow path of the TLB miss handler */ -page_table_not_present: -page_not_present: - tlbmiss_restore - sub sp, 4 - stmts --sp, r0-lr - rcall save_full_context_ex - mfsr r12, SYSREG_ECR - mov r11, sp - rcall do_page_fault - rjmp ret_from_exception + rjmp enter_monitor_mode /* This function expects to find offending PC in SYSREG_RAR_EX */ + .type save_full_context_ex, @function + .align 2 save_full_context_ex: + mfsr r11, SYSREG_RAR_EX + sub r9, pc, . - debug_trampoline mfsr r8, SYSREG_RSR_EX + cp.w r9, r11 + breq 3f mov r12, r8 andh r8, (MODE_MASK >> 16), COH - mfsr r11, SYSREG_RAR_EX brne 2f 1: pushm r11, r12 /* PC and SR */ @@ -302,77 +319,170 @@ save_full_context_ex: stdsp sp[4], r10 /* replace saved SP */ rjmp 1b + /* + * The debug handler set up a trampoline to make us + * automatically enter monitor mode upon return, but since + * we're saving the full context, we must assume that the + * exception handler might want to alter the return address + * and/or status register. So we need to restore the original + * context and enter monitor mode manually after the exception + * has been handled. + */ +3: get_thread_info r8 + ld.w r11, r8[TI_rar_saved] + ld.w r12, r8[TI_rsr_saved] + rjmp 1b + .size save_full_context_ex, . - save_full_context_ex + /* Low-level exception handlers */ handle_critical: - pushm r12 + /* + * AT32AP700x errata: + * + * After a Java stack overflow or underflow trap, any CPU + * memory access may cause erratic behavior. This will happen + * when the four least significant bits of the JOSP system + * register contains any value between 9 and 15 (inclusive). + * + * Possible workarounds: + * - Don't use the Java Extension Module + * - Ensure that the stack overflow and underflow trap + * handlers do not do any memory access or trigger any + * exceptions before the overflow/underflow condition is + * cleared (by incrementing or decrementing the JOSP) + * - Make sure that JOSP does not contain any problematic + * value before doing any exception or interrupt + * processing. + * - Set up a critical exception handler which writes a + * known-to-be-safe value, e.g. 4, to JOSP before doing + * any further processing. + * + * We'll use the last workaround for now since we cannot + * guarantee that user space processes don't use Java mode. + * Non-well-behaving userland will be terminated with extreme + * prejudice. + */ +#ifdef CONFIG_CPU_AT32AP700X + /* + * There's a chance we can't touch memory, so temporarily + * borrow PTBR to save the stack pointer while we fix things + * up... + */ + mtsr SYSREG_PTBR, sp + mov sp, 4 + mtsr SYSREG_JOSP, sp + mfsr sp, SYSREG_PTBR + sub pc, -2 + + /* Push most of pt_regs on stack. We'll do the rest later */ + sub sp, 4 pushm r0-r12 - rcall save_full_context_ex + + /* PTBR mirrors current_thread_info()->task->active_mm->pgd */ + get_thread_info r0 + ld.w r1, r0[TI_task] + ld.w r2, r1[TSK_active_mm] + ld.w r3, r2[MM_pgd] + mtsr SYSREG_PTBR, r3 +#else + sub sp, 4 + pushm r0-r12 +#endif + sub r0, sp, -(14 * 4) + mov r1, lr + mfsr r2, SYSREG_RAR_EX + mfsr r3, SYSREG_RSR_EX + pushm r0-r3 + mfsr r12, SYSREG_ECR mov r11, sp - rcall do_critical_exception + call do_critical_exception /* We should never get here... */ bad_return: sub r12, pc, (. - 1f) - bral panic + lddpc pc, 2f .align 2 1: .asciz "Return from critical exception!" +2: .long panic .align 1 do_bus_error_write: sub sp, 4 stmts --sp, r0-lr - rcall save_full_context_ex + call save_full_context_ex mov r11, 1 rjmp 1f do_bus_error_read: sub sp, 4 stmts --sp, r0-lr - rcall save_full_context_ex + call save_full_context_ex mov r11, 0 1: mfsr r12, SYSREG_BEAR mov r10, sp - rcall do_bus_error + call do_bus_error rjmp ret_from_exception .align 1 do_nmi_ll: sub sp, 4 stmts --sp, r0-lr - /* FIXME: Make sure RAR_NMI and RSR_NMI are pushed instead of *_EX */ - rcall save_full_context_ex + mfsr r9, SYSREG_RSR_NMI + mfsr r8, SYSREG_RAR_NMI + bfextu r0, r9, MODE_SHIFT, 3 + brne 2f + +1: pushm r8, r9 /* PC and SR */ mfsr r12, SYSREG_ECR mov r11, sp - rcall do_nmi - rjmp bad_return + call do_nmi + popm r8-r9 + mtsr SYSREG_RAR_NMI, r8 + tst r0, r0 + mtsr SYSREG_RSR_NMI, r9 + brne 3f + + ldmts sp++, r0-lr + sub sp, -4 /* skip r12_orig */ + rete + +2: sub r10, sp, -(FRAME_SIZE_FULL - REG_LR) + stdsp sp[4], r10 /* replace saved SP */ + rjmp 1b + +3: popm lr + sub sp, -4 /* skip sp */ + popm r0-r12 + sub sp, -4 /* skip r12_orig */ + rete handle_address_fault: sub sp, 4 stmts --sp, r0-lr - rcall save_full_context_ex + call save_full_context_ex mfsr r12, SYSREG_ECR mov r11, sp - rcall do_address_exception + call do_address_exception rjmp ret_from_exception handle_protection_fault: sub sp, 4 stmts --sp, r0-lr - rcall save_full_context_ex + call save_full_context_ex mfsr r12, SYSREG_ECR mov r11, sp - rcall do_page_fault + call do_page_fault rjmp ret_from_exception .align 1 do_illegal_opcode_ll: sub sp, 4 stmts --sp, r0-lr - rcall save_full_context_ex + call save_full_context_ex mfsr r12, SYSREG_ECR mov r11, sp - rcall do_illegal_opcode + call do_illegal_opcode rjmp ret_from_exception do_dtlb_modified: @@ -406,16 +516,17 @@ do_dtlb_modified: do_fpe_ll: sub sp, 4 stmts --sp, r0-lr - rcall save_full_context_ex + call save_full_context_ex unmask_interrupts mov r12, 26 mov r11, sp - rcall do_fpe + call do_fpe rjmp ret_from_exception ret_from_exception: mask_interrupts lddsp r4, sp[REG_SR] + andh r4, (MODE_MASK >> 16), COH brne fault_resume_kernel @@ -445,7 +556,7 @@ fault_resume_kernel: lddsp r4, sp[REG_SR] bld r4, SYSREG_GM_OFFSET brcs 1f - rcall preempt_schedule_irq + call preempt_schedule_irq 1: #endif @@ -474,137 +585,142 @@ fault_exit_work: bld r1, TIF_NEED_RESCHED brcc 1f unmask_interrupts - rcall schedule + call schedule mask_interrupts ld.w r1, r0[TI_flags] rjmp fault_exit_work -1: mov r2, _TIF_SIGPENDING | _TIF_RESTORE_SIGMASK +1: mov r2, _TIF_SIGPENDING | _TIF_NOTIFY_RESUME tst r1, r2 breq 2f unmask_interrupts mov r12, sp mov r11, r0 - rcall do_notify_resume + call do_notify_resume mask_interrupts ld.w r1, r0[TI_flags] rjmp fault_exit_work 2: bld r1, TIF_BREAKPOINT brcc fault_resume_user - mfsr r3, SYSREG_TLBEHI - lddsp r2, sp[REG_PC] - andl r3, 0xff, COH - lsl r3, 1 - sbr r3, 30 - sbr r3, 0 - mtdr DBGREG_BWA2A, r2 - mtdr DBGREG_BWC2A, r3 - rjmp fault_resume_user - - /* If we get a debug trap from privileged context we end up here */ -handle_debug_priv: - /* Fix up LR and SP in regs. r11 contains the mode we came from */ + rjmp enter_monitor_mode + + .section .kprobes.text, "ax", @progbits + .type handle_debug, @function +handle_debug: + sub sp, 4 /* r12_orig */ + stmts --sp, r0-lr + mfsr r8, SYSREG_RAR_DBG + mfsr r9, SYSREG_RSR_DBG + unmask_exceptions + pushm r8-r9 + bfextu r9, r9, SYSREG_MODE_OFFSET, SYSREG_MODE_SIZE + brne debug_fixup_regs + +.Ldebug_fixup_cont: +#ifdef CONFIG_TRACE_IRQFLAGS + call trace_hardirqs_off +#endif + mov r12, sp + call do_debug + mov sp, r12 + + lddsp r2, sp[REG_SR] + bfextu r3, r2, SYSREG_MODE_OFFSET, SYSREG_MODE_SIZE + brne debug_resume_kernel + + get_thread_info r0 + ld.w r1, r0[TI_flags] + mov r2, _TIF_DBGWORK_MASK + tst r1, r2 + brne debug_exit_work + + bld r1, TIF_SINGLE_STEP + brcc 1f + mfdr r4, OCD_DC + sbr r4, OCD_DC_SS_BIT + mtdr OCD_DC, r4 + +1: popm r10,r11 + mask_exceptions + mtsr SYSREG_RSR_DBG, r11 + mtsr SYSREG_RAR_DBG, r10 +#ifdef CONFIG_TRACE_IRQFLAGS + call trace_hardirqs_on +1: +#endif + ldmts sp++, r0-lr + sub sp, -4 + retd + .size handle_debug, . - handle_debug + + /* Mode of the trapped context is in r9 */ + .type debug_fixup_regs, @function +debug_fixup_regs: mfsr r8, SYSREG_SR - mov r9, r8 - andh r8, hi(~MODE_MASK) - or r8, r11 + mov r10, r8 + bfins r8, r9, SYSREG_MODE_OFFSET, SYSREG_MODE_SIZE mtsr SYSREG_SR, r8 sub pc, -2 stdsp sp[REG_LR], lr - mtsr SYSREG_SR, r9 + mtsr SYSREG_SR, r10 sub pc, -2 - sub r10, sp, -FRAME_SIZE_FULL - stdsp sp[REG_SP], r10 - mov r12, sp - rcall do_debug_priv + sub r8, sp, -FRAME_SIZE_FULL + stdsp sp[REG_SP], r8 + rjmp .Ldebug_fixup_cont + .size debug_fixup_regs, . - debug_fixup_regs - /* Now, put everything back */ - ssrf SR_EM_BIT + .type debug_resume_kernel, @function +debug_resume_kernel: + mask_exceptions popm r10, r11 mtsr SYSREG_RAR_DBG, r10 mtsr SYSREG_RSR_DBG, r11 - mfsr r8, SYSREG_SR - mov r9, r8 - andh r8, hi(~MODE_MASK) - andh r11, hi(MODE_MASK) - or r8, r11 - mtsr SYSREG_SR, r8 +#ifdef CONFIG_TRACE_IRQFLAGS + bld r11, SYSREG_GM_OFFSET + brcc 1f + call trace_hardirqs_on +1: +#endif + mfsr r2, SYSREG_SR + mov r1, r2 + bfins r2, r3, SYSREG_MODE_OFFSET, SYSREG_MODE_SIZE + mtsr SYSREG_SR, r2 sub pc, -2 popm lr - mtsr SYSREG_SR, r9 + mtsr SYSREG_SR, r1 sub pc, -2 sub sp, -4 /* skip SP */ popm r0-r12 sub sp, -4 retd + .size debug_resume_kernel, . - debug_resume_kernel + .type debug_exit_work, @function +debug_exit_work: /* - * At this point, everything is masked, that is, interrupts, - * exceptions and debugging traps. We might get called from - * interrupt or exception context in some rare cases, but this - * will be taken care of by do_debug(), so we're not going to - * do a 100% correct context save here. + * We must return from Monitor Mode using a retd, and we must + * not schedule since that involves the D bit in SR getting + * cleared by something other than the debug hardware. This + * may cause undefined behaviour according to the Architecture + * manual. + * + * So we fix up the return address and status and return to a + * stub below in Exception mode. From there, we can follow the + * normal exception return path. + * + * The real return address and status registers are stored on + * the stack in the way the exception return path understands, + * so no need to fix anything up there. */ -handle_debug: - sub sp, 4 /* r12_orig */ - stmts --sp, r0-lr - mfsr r10, SYSREG_RAR_DBG - mfsr r11, SYSREG_RSR_DBG - unmask_exceptions - pushm r10,r11 - andh r11, (MODE_MASK >> 16), COH - brne handle_debug_priv - - mov r12, sp - rcall do_debug - - lddsp r10, sp[REG_SR] - andh r10, (MODE_MASK >> 16), COH - breq debug_resume_user - -debug_restore_all: - popm r10,r11 - mask_exceptions - mtsr SYSREG_RSR_DBG, r11 - mtsr SYSREG_RAR_DBG, r10 - ldmts sp++, r0-lr - sub sp, -4 + sub r8, pc, . - fault_exit_work + mtsr SYSREG_RAR_DBG, r8 + mov r9, 0 + orh r9, hi(SR_EM | SR_GM | MODE_EXCEPTION) + mtsr SYSREG_RSR_DBG, r9 + sub pc, -2 retd - -debug_resume_user: - get_thread_info r0 - mask_interrupts - - ld.w r1, r0[TI_flags] - andl r1, _TIF_DBGWORK_MASK, COH - breq debug_restore_all - -1: bld r1, TIF_NEED_RESCHED - brcc 2f - unmask_interrupts - rcall schedule - mask_interrupts - ld.w r1, r0[TI_flags] - rjmp 1b - -2: mov r2, _TIF_SIGPENDING | _TIF_RESTORE_SIGMASK - tst r1, r2 - breq 3f - unmask_interrupts - mov r12, sp - mov r11, r0 - rcall do_notify_resume - mask_interrupts - ld.w r1, r0[TI_flags] - rjmp 1b - -3: bld r1, TIF_SINGLE_STEP - brcc debug_restore_all - mfdr r2, DBGREG_DC - sbr r2, DC_SS_BIT - mtdr DBGREG_DC, r2 - rjmp debug_restore_all + .size debug_exit_work, . - debug_exit_work .set rsr_int0, SYSREG_RSR_INT0 .set rsr_int1, SYSREG_RSR_INT1 @@ -622,17 +738,27 @@ irq_level\level: stmts --sp,r0-lr mfsr r8, rar_int\level mfsr r9, rsr_int\level + +#ifdef CONFIG_PREEMPT + sub r11, pc, (. - system_call) + cp.w r11, r8 + breq 4f +#endif + pushm r8-r9 mov r11, sp mov r12, \level - rcall do_IRQ + call do_IRQ lddsp r4, sp[REG_SR] - andh r4, (MODE_MASK >> 16), COH + bfextu r4, r4, SYSREG_M0_OFFSET, 3 + cp.w r4, MODE_SUPERVISOR >> SYSREG_M0_OFFSET + breq 2f + cp.w r4, MODE_USER >> SYSREG_M0_OFFSET #ifdef CONFIG_PREEMPT - brne 2f + brne 3f #else brne 1f #endif @@ -642,7 +768,11 @@ irq_level\level: andl r1, _TIF_WORK_MASK, COH brne irq_exit_work -1: popm r8-r9 +1: +#ifdef CONFIG_TRACE_IRQFLAGS + call trace_hardirqs_on +#endif + popm r8-r9 mtsr rar_int\level, r8 mtsr rsr_int\level, r9 ldmts sp++,r0-lr @@ -650,8 +780,27 @@ irq_level\level: rete #ifdef CONFIG_PREEMPT -2: - get_thread_info r0 +4: mask_interrupts + mfsr r8, rsr_int\level + sbr r8, 16 + mtsr rsr_int\level, r8 + ldmts sp++, r0-lr + sub sp, -4 /* ignore r12_orig */ + rete +#endif + +2: get_thread_info r0 + ld.w r1, r0[TI_flags] + bld r1, TIF_CPU_GOING_TO_SLEEP +#ifdef CONFIG_PREEMPT + brcc 3f +#else + brcc 1b +#endif + sub r1, pc, . - cpu_idle_skip_sleep + stdsp sp[REG_PC], r1 +#ifdef CONFIG_PREEMPT +3: get_thread_info r0 ld.w r2, r0[TI_preempt_count] cp.w r2, 0 brne 1b @@ -661,9 +810,9 @@ irq_level\level: lddsp r4, sp[REG_SR] bld r4, SYSREG_GM_OFFSET brcs 1b - rcall preempt_schedule_irq - rjmp 1b + call preempt_schedule_irq #endif + rjmp 1b .endm .section .irq.text,"ax",@progbits @@ -676,3 +825,53 @@ irq_level\level: IRQ_LEVEL 1 IRQ_LEVEL 2 IRQ_LEVEL 3 + + .section .kprobes.text, "ax", @progbits + .type enter_monitor_mode, @function +enter_monitor_mode: + /* + * We need to enter monitor mode to do a single step. The + * monitor code will alter the return address so that we + * return directly to the user instead of returning here. + */ + breakpoint + rjmp breakpoint_failed + + .size enter_monitor_mode, . - enter_monitor_mode + + .type debug_trampoline, @function + .global debug_trampoline +debug_trampoline: + /* + * Save the registers on the stack so that the monitor code + * can find them easily. + */ + sub sp, 4 /* r12_orig */ + stmts --sp, r0-lr + get_thread_info r0 + ld.w r8, r0[TI_rar_saved] + ld.w r9, r0[TI_rsr_saved] + pushm r8-r9 + + /* + * The monitor code will alter the return address so we don't + * return here. + */ + breakpoint + rjmp breakpoint_failed + .size debug_trampoline, . - debug_trampoline + + .type breakpoint_failed, @function +breakpoint_failed: + /* + * Something went wrong. Perhaps the debug hardware isn't + * enabled? + */ + lda.w r12, msg_breakpoint_failed + mov r11, sp + mov r10, 9 /* SIGKILL */ + call die +1: rjmp 1b + +msg_breakpoint_failed: + .asciz "Failed to enter Debug Mode" diff --git a/arch/avr32/kernel/head.S b/arch/avr32/kernel/head.S index 773b7ad87be..59eae6dfbed 100644 --- a/arch/avr32/kernel/head.S +++ b/arch/avr32/kernel/head.S @@ -10,36 +10,13 @@ #include <linux/linkage.h> #include <asm/page.h> -#include <asm/thread_info.h> -#include <asm/sysreg.h> .section .init.text,"ax" .global kernel_entry kernel_entry: - /* Initialize status register */ - lddpc r0, init_sr - mtsr SYSREG_SR, r0 - - /* Set initial stack pointer */ - lddpc sp, stack_addr - sub sp, -THREAD_SIZE - -#ifdef CONFIG_FRAME_POINTER - /* Mark last stack frame */ - mov lr, 0 - mov r7, 0 -#endif - - /* Set up the PIO, SDRAM controller, early printk, etc. */ - rcall board_early_init - /* Start the show */ lddpc pc, kernel_start_addr .align 2 -init_sr: - .long 0x007f0000 /* Supervisor mode, everything masked */ -stack_addr: - .long init_thread_union kernel_start_addr: .long start_kernel diff --git a/arch/avr32/kernel/init_task.c b/arch/avr32/kernel/init_task.c deleted file mode 100644 index effcacf9d1a..00000000000 --- a/arch/avr32/kernel/init_task.c +++ /dev/null @@ -1,38 +0,0 @@ -/* - * Copyright (C) 2004-2006 Atmel Corporation - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License version 2 as - * published by the Free Software Foundation. - */ -#include <linux/module.h> -#include <linux/fs.h> -#include <linux/sched.h> -#include <linux/init_task.h> -#include <linux/mqueue.h> - -#include <asm/pgtable.h> - -static struct fs_struct init_fs = INIT_FS; -static struct files_struct init_files = INIT_FILES; -static struct signal_struct init_signals = INIT_SIGNALS(init_signals); -static struct sighand_struct init_sighand = INIT_SIGHAND(init_sighand); -struct mm_struct init_mm = INIT_MM(init_mm); - -EXPORT_SYMBOL(init_mm); - -/* - * Initial thread structure. Must be aligned on an 8192-byte boundary. - */ -union thread_union init_thread_union - __attribute__((__section__(".data.init_task"))) = - { INIT_THREAD_INFO(init_task) }; - -/* - * Initial task structure. - * - * All other task structs will be allocated on slabs in fork.c - */ -struct task_struct init_task = INIT_TASK(init_task); - -EXPORT_SYMBOL(init_task); diff --git a/arch/avr32/kernel/irq.c b/arch/avr32/kernel/irq.c index 856f3548e66..900e49b2258 100644 --- a/arch/avr32/kernel/irq.c +++ b/arch/avr32/kernel/irq.c @@ -7,15 +7,6 @@ * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. - * - * This file contains the code used by various IRQ handling routines: - * asking for different IRQ's should be done through these routines - * instead of just grabbing them. Thus setups with different IRQ numbers - * shouldn't result in any weird surprises, and installing new handlers - * should be easier. - * - * IRQ's are in fact implemented a bit like signal handlers for the kernel. - * Naturally it's not a 1:1 relation, but there are similarities. */ #include <linux/interrupt.h> @@ -23,49 +14,15 @@ #include <linux/kernel_stat.h> #include <linux/proc_fs.h> #include <linux/seq_file.h> -#include <linux/sysdev.h> +#include <linux/device.h> -/* - * 'what should we do if we get a hw irq event on an illegal vector'. - * each architecture has to answer this themselves. - */ -void ack_bad_irq(unsigned int irq) +/* May be overridden by platform code */ +int __weak nmi_enable(void) { - printk("unexpected IRQ %u\n", irq); + return -ENOSYS; } -#ifdef CONFIG_PROC_FS -int show_interrupts(struct seq_file *p, void *v) +void __weak nmi_disable(void) { - int i = *(loff_t *)v, cpu; - struct irqaction *action; - unsigned long flags; - - if (i == 0) { - seq_puts(p, " "); - for_each_online_cpu(cpu) - seq_printf(p, "CPU%d ", cpu); - seq_putc(p, '\n'); - } - - if (i < NR_IRQS) { - spin_lock_irqsave(&irq_desc[i].lock, flags); - action = irq_desc[i].action; - if (!action) - goto unlock; - - seq_printf(p, "%3d: ", i); - for_each_online_cpu(cpu) - seq_printf(p, "%10u ", kstat_cpu(cpu).irqs[i]); - seq_printf(p, " %s", action->name); - for (action = action->next; action; action = action->next) - seq_printf(p, ", %s", action->name); - - seq_putc(p, '\n'); - unlock: - spin_unlock_irqrestore(&irq_desc[i].lock, flags); - } - return 0; } -#endif diff --git a/arch/avr32/kernel/kprobes.c b/arch/avr32/kernel/kprobes.c index 6caf9e8d808..f820e9f2552 100644 --- a/arch/avr32/kernel/kprobes.c +++ b/arch/avr32/kernel/kprobes.c @@ -15,13 +15,15 @@ #include <linux/ptrace.h> #include <asm/cacheflush.h> -#include <asm/kdebug.h> +#include <linux/kdebug.h> #include <asm/ocd.h> DEFINE_PER_CPU(struct kprobe *, current_kprobe); static unsigned long kprobe_status; static struct pt_regs jprobe_saved_regs; +struct kretprobe_blackpoint kretprobe_blacklist[] = {{NULL, NULL}}; + int __kprobes arch_prepare_kprobe(struct kprobe *p) { int ret = 0; @@ -46,6 +48,7 @@ int __kprobes arch_prepare_kprobe(struct kprobe *p) void __kprobes arch_arm_kprobe(struct kprobe *p) { pr_debug("arming kprobe at %p\n", p->addr); + ocd_enable(NULL); *p->addr = BREAKPOINT_INSTRUCTION; flush_icache_range((unsigned long)p->addr, (unsigned long)p->addr + sizeof(kprobe_opcode_t)); @@ -54,6 +57,7 @@ void __kprobes arch_arm_kprobe(struct kprobe *p) void __kprobes arch_disarm_kprobe(struct kprobe *p) { pr_debug("disarming kprobe at %p\n", p->addr); + ocd_disable(NULL); *p->addr = p->opcode; flush_icache_range((unsigned long)p->addr, (unsigned long)p->addr + sizeof(kprobe_opcode_t)); @@ -68,9 +72,9 @@ static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs) BUG_ON(!(sysreg_read(SR) & SYSREG_BIT(SR_D))); - dc = __mfdr(DBGREG_DC); - dc |= DC_SS; - __mtdr(DBGREG_DC, dc); + dc = ocd_read(DC); + dc |= 1 << OCD_DC_SS_BIT; + ocd_write(DC, dc); /* * We must run the instruction from its original location @@ -89,9 +93,9 @@ static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs) pr_debug("resuming execution at PC=%08lx\n", regs->pc); - dc = __mfdr(DBGREG_DC); - dc &= ~DC_SS; - __mtdr(DBGREG_DC, dc); + dc = ocd_read(DC); + dc &= ~(1 << OCD_DC_SS_BIT); + ocd_write(DC, dc); *p->addr = BREAKPOINT_INSTRUCTION; flush_icache_range((unsigned long)p->addr, @@ -109,7 +113,7 @@ static int __kprobes kprobe_handler(struct pt_regs *regs) void *addr = (void *)regs->pc; int ret = 0; - pr_debug("kprobe_handler: kprobe_running=%d\n", + pr_debug("kprobe_handler: kprobe_running=%p\n", kprobe_running()); /* @@ -154,6 +158,7 @@ ss_probe: return 1; no_kprobe: + preempt_enable_no_resched(); return ret; } @@ -178,7 +183,7 @@ static int __kprobes post_kprobe_handler(struct pt_regs *regs) return 1; } -static int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr) +int __kprobes kprobe_fault_handler(struct pt_regs *regs, int trapnr) { struct kprobe *cur = kprobe_running(); @@ -215,11 +220,6 @@ int __kprobes kprobe_exceptions_notify(struct notifier_block *self, if (post_kprobe_handler(args->regs)) ret = NOTIFY_STOP; break; - case DIE_FAULT: - if (kprobe_running() - && kprobe_fault_handler(args->regs, args->trapnr)) - ret = NOTIFY_STOP; - break; default: break; } @@ -262,9 +262,6 @@ int __kprobes longjmp_break_handler(struct kprobe *p, struct pt_regs *regs) int __init arch_init_kprobes(void) { - printk("KPROBES: Enabling monitor mode (MM|DBE)...\n"); - __mtdr(DBGREG_DC, DC_MM | DC_DBE); - /* TODO: Register kretprobe trampoline */ return 0; } diff --git a/arch/avr32/kernel/module.c b/arch/avr32/kernel/module.c index dfc32f2817b..2c941290802 100644 --- a/arch/avr32/kernel/module.c +++ b/arch/avr32/kernel/module.c @@ -12,27 +12,19 @@ * published by the Free Software Foundation. */ -#include <linux/moduleloader.h> -#include <linux/module.h> -#include <linux/kernel.h> +#include <linux/bug.h> #include <linux/elf.h> +#include <linux/kernel.h> +#include <linux/module.h> +#include <linux/moduleloader.h> #include <linux/vmalloc.h> -void *module_alloc(unsigned long size) -{ - if (size == 0) - return NULL; - return vmalloc(size); -} - void module_free(struct module *mod, void *module_region) { vfree(mod->arch.syminfo); mod->arch.syminfo = NULL; vfree(module_region); - /* FIXME: if module_region == mod->init_region, trim exception - * table entries. */ } static inline int check_rela(Elf32_Rela *rela, struct module *module, @@ -263,7 +255,7 @@ int apply_relocate_add(Elf32_Shdr *sechdrs, const char *strtab, * value of PC. Just subtract the value of * GOT, and we're done. */ - pr_debug("GOTPC: PC=0x%lx, got_offset=0x%lx, core=0x%p\n", + pr_debug("GOTPC: PC=0x%x, got_offset=0x%lx, core=0x%p\n", relocation, module->arch.got_offset, module->module_core); relocation -= ((unsigned long)module->module_core @@ -272,7 +264,7 @@ int apply_relocate_add(Elf32_Shdr *sechdrs, const char *strtab, break; case R_AVR32_GOT18SW: if ((relocation & 0xfffe0003) != 0 - && (relocation & 0xfffc0003) != 0xffff0000) + && (relocation & 0xfffc0000) != 0xfffc0000) return reloc_overflow(module, "R_AVR32_GOT18SW", relocation); relocation >>= 2; @@ -282,7 +274,7 @@ int apply_relocate_add(Elf32_Shdr *sechdrs, const char *strtab, && (relocation & 0xffff0000) != 0xffff0000) return reloc_overflow(module, "R_AVR32_GOT16S", relocation); - pr_debug("GOT reloc @ 0x%lx -> %lu\n", + pr_debug("GOT reloc @ 0x%x -> %u\n", rel->r_offset, relocation); value = *location; value = ((value & 0xffff0000) @@ -300,15 +292,6 @@ int apply_relocate_add(Elf32_Shdr *sechdrs, const char *strtab, return ret; } -int apply_relocate(Elf32_Shdr *sechdrs, const char *strtab, - unsigned int symindex, unsigned int relindex, - struct module *module) -{ - printk(KERN_ERR "module %s: REL relocations are not supported\n", - module->name); - return -ENOEXEC; -} - int module_finalize(const Elf_Ehdr *hdr, const Elf_Shdr *sechdrs, struct module *module) { @@ -317,8 +300,3 @@ int module_finalize(const Elf_Ehdr *hdr, const Elf_Shdr *sechdrs, return 0; } - -void module_arch_cleanup(struct module *module) -{ - -} diff --git a/arch/avr32/kernel/nmi_debug.c b/arch/avr32/kernel/nmi_debug.c new file mode 100644 index 00000000000..3414b8566c2 --- /dev/null +++ b/arch/avr32/kernel/nmi_debug.c @@ -0,0 +1,82 @@ +/* + * Copyright (C) 2007 Atmel Corporation + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ +#include <linux/delay.h> +#include <linux/kdebug.h> +#include <linux/notifier.h> +#include <linux/sched.h> + +#include <asm/irq.h> + +enum nmi_action { + NMI_SHOW_STATE = 1 << 0, + NMI_SHOW_REGS = 1 << 1, + NMI_DIE = 1 << 2, + NMI_DEBOUNCE = 1 << 3, +}; + +static unsigned long nmi_actions; + +static int nmi_debug_notify(struct notifier_block *self, + unsigned long val, void *data) +{ + struct die_args *args = data; + + if (likely(val != DIE_NMI)) + return NOTIFY_DONE; + + if (nmi_actions & NMI_SHOW_STATE) + show_state(); + if (nmi_actions & NMI_SHOW_REGS) + show_regs(args->regs); + if (nmi_actions & NMI_DEBOUNCE) + mdelay(10); + if (nmi_actions & NMI_DIE) + return NOTIFY_BAD; + + return NOTIFY_OK; +} + +static struct notifier_block nmi_debug_nb = { + .notifier_call = nmi_debug_notify, +}; + +static int __init nmi_debug_setup(char *str) +{ + char *p, *sep; + + register_die_notifier(&nmi_debug_nb); + if (nmi_enable()) { + printk(KERN_WARNING "Unable to enable NMI.\n"); + return 0; + } + + if (*str != '=') + return 0; + + for (p = str + 1; *p; p = sep + 1) { + sep = strchr(p, ','); + if (sep) + *sep = 0; + if (strcmp(p, "state") == 0) + nmi_actions |= NMI_SHOW_STATE; + else if (strcmp(p, "regs") == 0) + nmi_actions |= NMI_SHOW_REGS; + else if (strcmp(p, "debounce") == 0) + nmi_actions |= NMI_DEBOUNCE; + else if (strcmp(p, "die") == 0) + nmi_actions |= NMI_DIE; + else + printk(KERN_WARNING "NMI: Unrecognized action `%s'\n", + p); + if (!sep) + break; + } + + return 0; +} +__setup("nmi_debug", nmi_debug_setup); diff --git a/arch/avr32/kernel/ocd.c b/arch/avr32/kernel/ocd.c new file mode 100644 index 00000000000..1b0245d4e0c --- /dev/null +++ b/arch/avr32/kernel/ocd.c @@ -0,0 +1,167 @@ +/* + * Copyright (C) 2007 Atmel Corporation + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ +#include <linux/init.h> +#include <linux/sched.h> +#include <linux/spinlock.h> + +#include <asm/ocd.h> + +static long ocd_count; +static spinlock_t ocd_lock; + +/** + * ocd_enable - enable on-chip debugging + * @child: task to be debugged + * + * If @child is non-NULL, ocd_enable() first checks if debugging has + * already been enabled for @child, and if it has, does nothing. + * + * If @child is NULL (e.g. when debugging the kernel), or debugging + * has not already been enabled for it, ocd_enable() increments the + * reference count and enables the debugging hardware. + */ +void ocd_enable(struct task_struct *child) +{ + u32 dc; + + if (child) + pr_debug("ocd_enable: child=%s [%u]\n", + child->comm, child->pid); + else + pr_debug("ocd_enable (no child)\n"); + + if (!child || !test_and_set_tsk_thread_flag(child, TIF_DEBUG)) { + spin_lock(&ocd_lock); + ocd_count++; + dc = ocd_read(DC); + dc |= (1 << OCD_DC_MM_BIT) | (1 << OCD_DC_DBE_BIT); + ocd_write(DC, dc); + spin_unlock(&ocd_lock); + } +} + +/** + * ocd_disable - disable on-chip debugging + * @child: task that was being debugged, but isn't anymore + * + * If @child is non-NULL, ocd_disable() checks if debugging is enabled + * for @child, and if it isn't, does nothing. + * + * If @child is NULL (e.g. when debugging the kernel), or debugging is + * enabled, ocd_disable() decrements the reference count, and if it + * reaches zero, disables the debugging hardware. + */ +void ocd_disable(struct task_struct *child) +{ + u32 dc; + + if (!child) + pr_debug("ocd_disable (no child)\n"); + else if (test_tsk_thread_flag(child, TIF_DEBUG)) + pr_debug("ocd_disable: child=%s [%u]\n", + child->comm, child->pid); + + if (!child || test_and_clear_tsk_thread_flag(child, TIF_DEBUG)) { + spin_lock(&ocd_lock); + ocd_count--; + + WARN_ON(ocd_count < 0); + + if (ocd_count <= 0) { + dc = ocd_read(DC); + dc &= ~((1 << OCD_DC_MM_BIT) | (1 << OCD_DC_DBE_BIT)); + ocd_write(DC, dc); + } + spin_unlock(&ocd_lock); + } +} + +#ifdef CONFIG_DEBUG_FS +#include <linux/debugfs.h> +#include <linux/module.h> + +static struct dentry *ocd_debugfs_root; +static struct dentry *ocd_debugfs_DC; +static struct dentry *ocd_debugfs_DS; +static struct dentry *ocd_debugfs_count; + +static int ocd_DC_get(void *data, u64 *val) +{ + *val = ocd_read(DC); + return 0; +} +static int ocd_DC_set(void *data, u64 val) +{ + ocd_write(DC, val); + return 0; +} +DEFINE_SIMPLE_ATTRIBUTE(fops_DC, ocd_DC_get, ocd_DC_set, "0x%08llx\n"); + +static int ocd_DS_get(void *data, u64 *val) +{ + *val = ocd_read(DS); + return 0; +} +DEFINE_SIMPLE_ATTRIBUTE(fops_DS, ocd_DS_get, NULL, "0x%08llx\n"); + +static int ocd_count_get(void *data, u64 *val) +{ + *val = ocd_count; + return 0; +} +DEFINE_SIMPLE_ATTRIBUTE(fops_count, ocd_count_get, NULL, "%lld\n"); + +static void ocd_debugfs_init(void) +{ + struct dentry *root; + + root = debugfs_create_dir("ocd", NULL); + if (IS_ERR(root) || !root) + goto err_root; + ocd_debugfs_root = root; + + ocd_debugfs_DC = debugfs_create_file("DC", S_IRUSR | S_IWUSR, + root, NULL, &fops_DC); + if (!ocd_debugfs_DC) + goto err_DC; + + ocd_debugfs_DS = debugfs_create_file("DS", S_IRUSR, root, + NULL, &fops_DS); + if (!ocd_debugfs_DS) + goto err_DS; + + ocd_debugfs_count = debugfs_create_file("count", S_IRUSR, root, + NULL, &fops_count); + if (!ocd_debugfs_count) + goto err_count; + + return; + +err_count: + debugfs_remove(ocd_debugfs_DS); +err_DS: + debugfs_remove(ocd_debugfs_DC); +err_DC: + debugfs_remove(ocd_debugfs_root); +err_root: + printk(KERN_WARNING "OCD: Failed to create debugfs entries\n"); +} +#else +static inline void ocd_debugfs_init(void) +{ + +} +#endif + +static int __init ocd_init(void) +{ + spin_lock_init(&ocd_lock); + ocd_debugfs_init(); + return 0; +} +arch_initcall(ocd_init); diff --git a/arch/avr32/kernel/process.c b/arch/avr32/kernel/process.c index 317dc50945f..42a53e740a7 100644 --- a/arch/avr32/kernel/process.c +++ b/arch/avr32/kernel/process.c @@ -9,92 +9,62 @@ #include <linux/module.h> #include <linux/kallsyms.h> #include <linux/fs.h> +#include <linux/pm.h> #include <linux/ptrace.h> +#include <linux/slab.h> #include <linux/reboot.h> +#include <linux/tick.h> +#include <linux/uaccess.h> #include <linux/unistd.h> #include <asm/sysreg.h> #include <asm/ocd.h> +#include <asm/syscalls.h> -void (*pm_power_off)(void) = NULL; +#include <mach/pm.h> + +void (*pm_power_off)(void); EXPORT_SYMBOL(pm_power_off); /* * This file handles the architecture-dependent parts of process handling.. */ -void cpu_idle(void) +void arch_cpu_idle(void) { - /* endless idle loop with no priority at all */ - while (1) { - /* TODO: Enter sleep mode */ - while (!need_resched()) - cpu_relax(); - preempt_enable_no_resched(); - schedule(); - preempt_disable(); - } + cpu_enter_idle(); } void machine_halt(void) { + /* + * Enter Stop mode. The 32 kHz oscillator will keep running so + * the RTC will keep the time properly and the system will + * boot quickly. + */ + asm volatile("sleep 3\n\t" + "sub pc, -2"); } void machine_power_off(void) { + if (pm_power_off) + pm_power_off(); } void machine_restart(char *cmd) { - __mtdr(DBGREG_DC, DC_DBE); - __mtdr(DBGREG_DC, DC_RES); + ocd_write(DC, (1 << OCD_DC_DBE_BIT)); + ocd_write(DC, (1 << OCD_DC_RES_BIT)); while (1) ; } /* - * PC is actually discarded when returning from a system call -- the - * return address must be stored in LR. This function will make sure - * LR points to do_exit before starting the thread. - * - * Also, when returning from fork(), r12 is 0, so we must copy the - * argument as well. - * - * r0 : The argument to the main thread function - * r1 : The address of do_exit - * r2 : The address of the main thread function - */ -asmlinkage extern void kernel_thread_helper(void); -__asm__(" .type kernel_thread_helper, @function\n" - "kernel_thread_helper:\n" - " mov r12, r0\n" - " mov lr, r2\n" - " mov pc, r1\n" - " .size kernel_thread_helper, . - kernel_thread_helper"); - -int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) -{ - struct pt_regs regs; - - memset(®s, 0, sizeof(regs)); - - regs.r0 = (unsigned long)arg; - regs.r1 = (unsigned long)fn; - regs.r2 = (unsigned long)do_exit; - regs.lr = (unsigned long)kernel_thread_helper; - regs.pc = (unsigned long)kernel_thread_helper; - regs.sr = MODE_SUPERVISOR; - - return do_fork(flags | CLONE_VM | CLONE_UNTRACED, - 0, ®s, 0, NULL, NULL); -} -EXPORT_SYMBOL(kernel_thread); - -/* * Free current thread data structures etc */ void exit_thread(void) { - /* nothing to do */ + ocd_disable(current); } void flush_thread(void) @@ -107,51 +77,194 @@ void release_thread(struct task_struct *dead_task) /* do nothing */ } +static void dump_mem(const char *str, const char *log_lvl, + unsigned long bottom, unsigned long top) +{ + unsigned long p; + int i; + + printk("%s%s(0x%08lx to 0x%08lx)\n", log_lvl, str, bottom, top); + + for (p = bottom & ~31; p < top; ) { + printk("%s%04lx: ", log_lvl, p & 0xffff); + + for (i = 0; i < 8; i++, p += 4) { + unsigned int val; + + if (p < bottom || p >= top) + printk(" "); + else { + if (__get_user(val, (unsigned int __user *)p)) { + printk("\n"); + goto out; + } + printk("%08x ", val); + } + } + printk("\n"); + } + +out: + return; +} + +static inline int valid_stack_ptr(struct thread_info *tinfo, unsigned long p) +{ + return (p > (unsigned long)tinfo) + && (p < (unsigned long)tinfo + THREAD_SIZE - 3); +} + +#ifdef CONFIG_FRAME_POINTER +static void show_trace_log_lvl(struct task_struct *tsk, unsigned long *sp, + struct pt_regs *regs, const char *log_lvl) +{ + unsigned long lr, fp; + struct thread_info *tinfo; + + if (regs) + fp = regs->r7; + else if (tsk == current) + asm("mov %0, r7" : "=r"(fp)); + else + fp = tsk->thread.cpu_context.r7; + + /* + * Walk the stack as long as the frame pointer (a) is within + * the kernel stack of the task, and (b) it doesn't move + * downwards. + */ + tinfo = task_thread_info(tsk); + printk("%sCall trace:\n", log_lvl); + while (valid_stack_ptr(tinfo, fp)) { + unsigned long new_fp; + + lr = *(unsigned long *)fp; +#ifdef CONFIG_KALLSYMS + printk("%s [<%08lx>] ", log_lvl, lr); +#else + printk(" [<%08lx>] ", lr); +#endif + print_symbol("%s\n", lr); + + new_fp = *(unsigned long *)(fp + 4); + if (new_fp <= fp) + break; + fp = new_fp; + } + printk("\n"); +} +#else +static void show_trace_log_lvl(struct task_struct *tsk, unsigned long *sp, + struct pt_regs *regs, const char *log_lvl) +{ + unsigned long addr; + + printk("%sCall trace:\n", log_lvl); + + while (!kstack_end(sp)) { + addr = *sp++; + if (kernel_text_address(addr)) { +#ifdef CONFIG_KALLSYMS + printk("%s [<%08lx>] ", log_lvl, addr); +#else + printk(" [<%08lx>] ", addr); +#endif + print_symbol("%s\n", addr); + } + } + printk("\n"); +} +#endif + +void show_stack_log_lvl(struct task_struct *tsk, unsigned long sp, + struct pt_regs *regs, const char *log_lvl) +{ + struct thread_info *tinfo; + + if (sp == 0) { + if (tsk) + sp = tsk->thread.cpu_context.ksp; + else + sp = (unsigned long)&tinfo; + } + if (!tsk) + tsk = current; + + tinfo = task_thread_info(tsk); + + if (valid_stack_ptr(tinfo, sp)) { + dump_mem("Stack: ", log_lvl, sp, + THREAD_SIZE + (unsigned long)tinfo); + show_trace_log_lvl(tsk, (unsigned long *)sp, regs, log_lvl); + } +} + +void show_stack(struct task_struct *tsk, unsigned long *stack) +{ + show_stack_log_lvl(tsk, (unsigned long)stack, NULL, ""); +} + static const char *cpu_modes[] = { "Application", "Supervisor", "Interrupt level 0", "Interrupt level 1", "Interrupt level 2", "Interrupt level 3", "Exception", "NMI" }; -void show_regs(struct pt_regs *regs) +void show_regs_log_lvl(struct pt_regs *regs, const char *log_lvl) { unsigned long sp = regs->sp; unsigned long lr = regs->lr; unsigned long mode = (regs->sr & MODE_MASK) >> MODE_SHIFT; - if (!user_mode(regs)) + show_regs_print_info(log_lvl); + + if (!user_mode(regs)) { sp = (unsigned long)regs + FRAME_SIZE_FULL; - print_symbol("PC is at %s\n", instruction_pointer(regs)); - print_symbol("LR is at %s\n", lr); - printk("pc : [<%08lx>] lr : [<%08lx>] %s\n" - "sp : %08lx r12: %08lx r11: %08lx\n", - instruction_pointer(regs), - lr, print_tainted(), sp, regs->r12, regs->r11); - printk("r10: %08lx r9 : %08lx r8 : %08lx\n", - regs->r10, regs->r9, regs->r8); - printk("r7 : %08lx r6 : %08lx r5 : %08lx r4 : %08lx\n", - regs->r7, regs->r6, regs->r5, regs->r4); - printk("r3 : %08lx r2 : %08lx r1 : %08lx r0 : %08lx\n", - regs->r3, regs->r2, regs->r1, regs->r0); - printk("Flags: %c%c%c%c%c\n", + printk("%s", log_lvl); + print_symbol("PC is at %s\n", instruction_pointer(regs)); + printk("%s", log_lvl); + print_symbol("LR is at %s\n", lr); + } + + printk("%spc : [<%08lx>] lr : [<%08lx>] %s\n" + "%ssp : %08lx r12: %08lx r11: %08lx\n", + log_lvl, instruction_pointer(regs), lr, print_tainted(), + log_lvl, sp, regs->r12, regs->r11); + printk("%sr10: %08lx r9 : %08lx r8 : %08lx\n", + log_lvl, regs->r10, regs->r9, regs->r8); + printk("%sr7 : %08lx r6 : %08lx r5 : %08lx r4 : %08lx\n", + log_lvl, regs->r7, regs->r6, regs->r5, regs->r4); + printk("%sr3 : %08lx r2 : %08lx r1 : %08lx r0 : %08lx\n", + log_lvl, regs->r3, regs->r2, regs->r1, regs->r0); + printk("%sFlags: %c%c%c%c%c\n", log_lvl, regs->sr & SR_Q ? 'Q' : 'q', regs->sr & SR_V ? 'V' : 'v', regs->sr & SR_N ? 'N' : 'n', regs->sr & SR_Z ? 'Z' : 'z', regs->sr & SR_C ? 'C' : 'c'); - printk("Mode bits: %c%c%c%c%c%c%c%c%c\n", + printk("%sMode bits: %c%c%c%c%c%c%c%c%c%c\n", log_lvl, regs->sr & SR_H ? 'H' : 'h', - regs->sr & SR_R ? 'R' : 'r', regs->sr & SR_J ? 'J' : 'j', + regs->sr & SR_DM ? 'M' : 'm', + regs->sr & SR_D ? 'D' : 'd', regs->sr & SR_EM ? 'E' : 'e', regs->sr & SR_I3M ? '3' : '.', regs->sr & SR_I2M ? '2' : '.', regs->sr & SR_I1M ? '1' : '.', regs->sr & SR_I0M ? '0' : '.', regs->sr & SR_GM ? 'G' : 'g'); - printk("CPU Mode: %s\n", cpu_modes[mode]); + printk("%sCPU Mode: %s\n", log_lvl, cpu_modes[mode]); +} + +void show_regs(struct pt_regs *regs) +{ + unsigned long sp = regs->sp; + + if (!user_mode(regs)) + sp = (unsigned long)regs + FRAME_SIZE_FULL; - show_trace(NULL, (unsigned long *)sp, regs); + show_regs_log_lvl(regs, ""); + show_trace_log_lvl(current, (unsigned long *)sp, regs, ""); } EXPORT_SYMBOL(show_regs); @@ -163,74 +276,40 @@ int dump_fpu(struct pt_regs *regs, elf_fpregset_t *fpu) } asmlinkage void ret_from_fork(void); +asmlinkage void ret_from_kernel_thread(void); +asmlinkage void syscall_return(void); -int copy_thread(int nr, unsigned long clone_flags, unsigned long usp, - unsigned long unused, - struct task_struct *p, struct pt_regs *regs) +int copy_thread(unsigned long clone_flags, unsigned long usp, + unsigned long arg, + struct task_struct *p) { - struct pt_regs *childregs; - - childregs = ((struct pt_regs *)(THREAD_SIZE + (unsigned long)p->thread_info)) - 1; - *childregs = *regs; - - if (user_mode(regs)) - childregs->sp = usp; - else - childregs->sp = (unsigned long)p->thread_info + THREAD_SIZE; - - childregs->r12 = 0; /* Set return value for child */ + struct pt_regs *childregs = task_pt_regs(p); + + if (unlikely(p->flags & PF_KTHREAD)) { + memset(childregs, 0, sizeof(struct pt_regs)); + p->thread.cpu_context.r0 = arg; + p->thread.cpu_context.r1 = usp; /* fn */ + p->thread.cpu_context.r2 = (unsigned long)syscall_return; + p->thread.cpu_context.pc = (unsigned long)ret_from_kernel_thread; + childregs->sr = MODE_SUPERVISOR; + } else { + *childregs = *current_pt_regs(); + if (usp) + childregs->sp = usp; + childregs->r12 = 0; /* Set return value for child */ + p->thread.cpu_context.pc = (unsigned long)ret_from_fork; + } p->thread.cpu_context.sr = MODE_SUPERVISOR | SR_GM; p->thread.cpu_context.ksp = (unsigned long)childregs; - p->thread.cpu_context.pc = (unsigned long)ret_from_fork; - - return 0; -} -/* r12-r8 are dummy parameters to force the compiler to use the stack */ -asmlinkage int sys_fork(struct pt_regs *regs) -{ - return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL); -} + clear_tsk_thread_flag(p, TIF_DEBUG); + if ((clone_flags & CLONE_PTRACE) && test_thread_flag(TIF_DEBUG)) + ocd_enable(p); -asmlinkage int sys_clone(unsigned long clone_flags, unsigned long newsp, - unsigned long parent_tidptr, - unsigned long child_tidptr, struct pt_regs *regs) -{ - if (!newsp) - newsp = regs->sp; - return do_fork(clone_flags, newsp, regs, 0, - (int __user *)parent_tidptr, - (int __user *)child_tidptr); -} - -asmlinkage int sys_vfork(struct pt_regs *regs) -{ - return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp, regs, - 0, NULL, NULL); -} - -asmlinkage int sys_execve(char __user *ufilename, char __user *__user *uargv, - char __user *__user *uenvp, struct pt_regs *regs) -{ - int error; - char *filename; - - filename = getname(ufilename); - error = PTR_ERR(filename); - if (IS_ERR(filename)) - goto out; - - error = do_execve(filename, uargv, uenvp, regs); - if (error == 0) - current->ptrace &= ~PT_DTRACE; - putname(filename); - -out: - return error; + return 0; } - /* * This function is supposed to answer the question "who called * schedule()?" @@ -243,7 +322,7 @@ unsigned long get_wchan(struct task_struct *p) if (!p || p == current || p->state == TASK_RUNNING) return 0; - stack_page = (unsigned long)p->thread_info; + stack_page = (unsigned long)task_stack_page(p); BUG_ON(!stack_page); /* @@ -262,7 +341,7 @@ unsigned long get_wchan(struct task_struct *p) * is actually quite ugly. It might be possible to * determine the frame size automatically at build * time by doing this: - * - compile sched.c + * - compile sched/core.c * - disassemble the resulting sched.o * - look for 'sub sp,??' shortly after '<schedule>:' */ diff --git a/arch/avr32/kernel/ptrace.c b/arch/avr32/kernel/ptrace.c index 3c89e59029a..4aedcab7cd4 100644 --- a/arch/avr32/kernel/ptrace.c +++ b/arch/avr32/kernel/ptrace.c @@ -9,7 +9,6 @@ #include <linux/kernel.h> #include <linux/sched.h> #include <linux/mm.h> -#include <linux/smp_lock.h> #include <linux/ptrace.h> #include <linux/errno.h> #include <linux/user.h> @@ -21,33 +20,40 @@ #include <asm/uaccess.h> #include <asm/ocd.h> #include <asm/mmu_context.h> -#include <asm/kdebug.h> +#include <linux/kdebug.h> static struct pt_regs *get_user_regs(struct task_struct *tsk) { - return (struct pt_regs *)((unsigned long) tsk->thread_info + + return (struct pt_regs *)((unsigned long)task_stack_page(tsk) + THREAD_SIZE - sizeof(struct pt_regs)); } -static void ptrace_single_step(struct task_struct *tsk) +void user_enable_single_step(struct task_struct *tsk) { - pr_debug("ptrace_single_step: pid=%u, SR=0x%08lx\n", - tsk->pid, tsk->thread.cpu_context.sr); - if (!(tsk->thread.cpu_context.sr & SR_D)) { - /* - * Set a breakpoint at the current pc to force the - * process into debug mode. The syscall/exception - * exit code will set a breakpoint at the return - * address when this flag is set. - */ - pr_debug("ptrace_single_step: Setting TIF_BREAKPOINT\n"); - set_tsk_thread_flag(tsk, TIF_BREAKPOINT); - } + pr_debug("user_enable_single_step: pid=%u, PC=0x%08lx, SR=0x%08lx\n", + tsk->pid, task_pt_regs(tsk)->pc, task_pt_regs(tsk)->sr); - /* The monitor code will do the actual step for us */ + /* + * We can't schedule in Debug mode, so when TIF_BREAKPOINT is + * set, the system call or exception handler will do a + * breakpoint to enter monitor mode before returning to + * userspace. + * + * The monitor code will then notice that TIF_SINGLE_STEP is + * set and return to userspace with single stepping enabled. + * The CPU will then enter monitor mode again after exactly + * one instruction has been executed, and the monitor code + * will then send a SIGTRAP to the process. + */ + set_tsk_thread_flag(tsk, TIF_BREAKPOINT); set_tsk_thread_flag(tsk, TIF_SINGLE_STEP); } +void user_disable_single_step(struct task_struct *child) +{ + /* XXX(hch): a no-op here seems wrong.. */ +} + /* * Called by kernel/ptrace.c when detaching * @@ -56,23 +62,8 @@ static void ptrace_single_step(struct task_struct *tsk) void ptrace_disable(struct task_struct *child) { clear_tsk_thread_flag(child, TIF_SINGLE_STEP); -} - -/* - * Handle hitting a breakpoint - */ -static void ptrace_break(struct task_struct *tsk, struct pt_regs *regs) -{ - siginfo_t info; - - info.si_signo = SIGTRAP; - info.si_errno = 0; - info.si_code = TRAP_BRKPT; - info.si_addr = (void __user *)instruction_pointer(regs); - - pr_debug("ptrace_break: Sending SIGTRAP to PID %u (pc = 0x%p)\n", - tsk->pid, info.si_addr); - force_sig_info(SIGTRAP, &info, tsk); + clear_tsk_thread_flag(child, TIF_BREAKPOINT); + ocd_disable(child); } /* @@ -85,9 +76,6 @@ static int ptrace_read_user(struct task_struct *tsk, unsigned long offset, unsigned long *regs; unsigned long value; - pr_debug("ptrace_read_user(%p, %#lx, %p)\n", - tsk, offset, data); - if (offset & 3 || offset >= sizeof(struct user)) { printk("ptrace_read_user: invalid offset 0x%08lx\n", offset); return -EIO; @@ -99,6 +87,9 @@ static int ptrace_read_user(struct task_struct *tsk, unsigned long offset, if (offset < sizeof(struct pt_regs)) value = regs[offset / sizeof(regs[0])]; + pr_debug("ptrace_read_user(%s[%u], %#lx, %p) -> %#lx\n", + tsk->comm, tsk->pid, offset, data, value); + return put_user(value, data); } @@ -112,8 +103,11 @@ static int ptrace_write_user(struct task_struct *tsk, unsigned long offset, { unsigned long *regs; + pr_debug("ptrace_write_user(%s[%u], %#lx, %#lx)\n", + tsk->comm, tsk->pid, offset, value); + if (offset & 3 || offset >= sizeof(struct user)) { - printk("ptrace_write_user: invalid offset 0x%08lx\n", offset); + pr_debug(" invalid offset 0x%08lx\n", offset); return -EIO; } @@ -152,102 +146,39 @@ static int ptrace_setregs(struct task_struct *tsk, const void __user *uregs) return ret; } -long arch_ptrace(struct task_struct *child, long request, long addr, long data) +long arch_ptrace(struct task_struct *child, long request, + unsigned long addr, unsigned long data) { - unsigned long tmp; int ret; - - pr_debug("arch_ptrace(%ld, %ld, %#lx, %#lx)\n", - request, child->pid, addr, data); - - pr_debug("ptrace: Enabling monitor mode...\n"); - __mtdr(DBGREG_DC, __mfdr(DBGREG_DC) | DC_MM | DC_DBE); + void __user *datap = (void __user *) data; switch (request) { /* Read the word at location addr in the child process */ case PTRACE_PEEKTEXT: case PTRACE_PEEKDATA: - ret = access_process_vm(child, addr, &tmp, sizeof(tmp), 0); - if (ret == sizeof(tmp)) - ret = put_user(tmp, (unsigned long __user *)data); - else - ret = -EIO; + ret = generic_ptrace_peekdata(child, addr, data); break; case PTRACE_PEEKUSR: - ret = ptrace_read_user(child, addr, - (unsigned long __user *)data); + ret = ptrace_read_user(child, addr, datap); break; /* Write the word in data at location addr */ case PTRACE_POKETEXT: case PTRACE_POKEDATA: - ret = access_process_vm(child, addr, &data, sizeof(data), 1); - if (ret == sizeof(data)) - ret = 0; - else - ret = -EIO; + ret = generic_ptrace_pokedata(child, addr, data); break; case PTRACE_POKEUSR: ret = ptrace_write_user(child, addr, data); break; - /* continue and stop at next (return from) syscall */ - case PTRACE_SYSCALL: - /* restart after signal */ - case PTRACE_CONT: - ret = -EIO; - if (!valid_signal(data)) - break; - if (request == PTRACE_SYSCALL) - set_tsk_thread_flag(child, TIF_SYSCALL_TRACE); - else - clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE); - child->exit_code = data; - /* XXX: Are we sure no breakpoints are active here? */ - wake_up_process(child); - ret = 0; - break; - - /* - * Make the child exit. Best I can do is send it a - * SIGKILL. Perhaps it should be put in the status that it - * wants to exit. - */ - case PTRACE_KILL: - ret = 0; - if (child->exit_state == EXIT_ZOMBIE) - break; - child->exit_code = SIGKILL; - wake_up_process(child); - break; - - /* - * execute single instruction. - */ - case PTRACE_SINGLESTEP: - ret = -EIO; - if (!valid_signal(data)) - break; - clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE); - ptrace_single_step(child); - child->exit_code = data; - wake_up_process(child); - ret = 0; - break; - - /* Detach a process that was attached */ - case PTRACE_DETACH: - ret = ptrace_detach(child, data); - break; - case PTRACE_GETREGS: - ret = ptrace_getregs(child, (void __user *)data); + ret = ptrace_getregs(child, datap); break; case PTRACE_SETREGS: - ret = ptrace_setregs(child, (const void __user *)data); + ret = ptrace_setregs(child, datap); break; default: @@ -255,19 +186,16 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data) break; } - pr_debug("sys_ptrace returning %d (DC = 0x%08lx)\n", ret, __mfdr(DBGREG_DC)); return ret; } asmlinkage void syscall_trace(void) { - pr_debug("syscall_trace called\n"); if (!test_thread_flag(TIF_SYSCALL_TRACE)) return; if (!(current->ptrace & PT_PTRACED)) return; - pr_debug("syscall_trace: notifying parent\n"); /* The 0x80 provides a way for the tracing parent to * distinguish between a syscall stop and SIGTRAP delivery */ ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD) @@ -286,86 +214,143 @@ asmlinkage void syscall_trace(void) } } -asmlinkage void do_debug_priv(struct pt_regs *regs) -{ - unsigned long dc, ds; - unsigned long die_val; - - ds = __mfdr(DBGREG_DS); - - pr_debug("do_debug_priv: pc = %08lx, ds = %08lx\n", regs->pc, ds); - - if (ds & DS_SSS) - die_val = DIE_SSTEP; - else - die_val = DIE_BREAKPOINT; - - if (notify_die(die_val, regs, 0, SIGTRAP) == NOTIFY_STOP) - return; - - if (likely(ds & DS_SSS)) { - extern void itlb_miss(void); - extern void tlb_miss_common(void); - struct thread_info *ti; - - dc = __mfdr(DBGREG_DC); - dc &= ~DC_SS; - __mtdr(DBGREG_DC, dc); - - ti = current_thread_info(); - ti->flags |= _TIF_BREAKPOINT; - - /* The TLB miss handlers don't check thread flags */ - if ((regs->pc >= (unsigned long)&itlb_miss) - && (regs->pc <= (unsigned long)&tlb_miss_common)) { - __mtdr(DBGREG_BWA2A, sysreg_read(RAR_EX)); - __mtdr(DBGREG_BWC2A, 0x40000001 | (get_asid() << 1)); - } - - /* - * If we're running in supervisor mode, the breakpoint - * will take us where we want directly, no need to - * single step. - */ - if ((regs->sr & MODE_MASK) != MODE_SUPERVISOR) - ti->flags |= TIF_SINGLE_STEP; - } else { - panic("Unable to handle debug trap at pc = %08lx\n", - regs->pc); - } -} - /* - * Handle breakpoints, single steps and other debuggy things. To keep - * things simple initially, we run with interrupts and exceptions - * disabled all the time. + * debug_trampoline() is an assembly stub which will store all user + * registers on the stack and execute a breakpoint instruction. + * + * If we single-step into an exception handler which runs with + * interrupts disabled the whole time so it doesn't have to check for + * pending work, its return address will be modified so that it ends + * up returning to debug_trampoline. + * + * If the exception handler decides to store the user context and + * enable interrupts after all, it will restore the original return + * address and status register value. Before it returns, it will + * notice that TIF_BREAKPOINT is set and execute a breakpoint + * instruction. */ -asmlinkage void do_debug(struct pt_regs *regs) -{ - unsigned long dc, ds; +extern void debug_trampoline(void); - ds = __mfdr(DBGREG_DS); - pr_debug("do_debug: pc = %08lx, ds = %08lx\n", regs->pc, ds); +asmlinkage struct pt_regs *do_debug(struct pt_regs *regs) +{ + struct thread_info *ti; + unsigned long trampoline_addr; + u32 status; + u32 ctrl; + int code; + + status = ocd_read(DS); + ti = current_thread_info(); + code = TRAP_BRKPT; + + pr_debug("do_debug: status=0x%08x PC=0x%08lx SR=0x%08lx tif=0x%08lx\n", + status, regs->pc, regs->sr, ti->flags); + + if (!user_mode(regs)) { + unsigned long die_val = DIE_BREAKPOINT; + + if (status & (1 << OCD_DS_SSS_BIT)) + die_val = DIE_SSTEP; + + if (notify_die(die_val, "ptrace", regs, 0, 0, SIGTRAP) + == NOTIFY_STOP) + return regs; + + if ((status & (1 << OCD_DS_SWB_BIT)) + && test_and_clear_ti_thread_flag( + ti, TIF_BREAKPOINT)) { + /* + * Explicit breakpoint from trampoline or + * exception/syscall/interrupt handler. + * + * The real saved regs are on the stack right + * after the ones we saved on entry. + */ + regs++; + pr_debug(" -> TIF_BREAKPOINT done, adjusted regs:" + "PC=0x%08lx SR=0x%08lx\n", + regs->pc, regs->sr); + BUG_ON(!user_mode(regs)); + + if (test_thread_flag(TIF_SINGLE_STEP)) { + pr_debug("Going to do single step...\n"); + return regs; + } + + /* + * No TIF_SINGLE_STEP means we're done + * stepping over a syscall. Do the trap now. + */ + code = TRAP_TRACE; + } else if ((status & (1 << OCD_DS_SSS_BIT)) + && test_ti_thread_flag(ti, TIF_SINGLE_STEP)) { + + pr_debug("Stepped into something, " + "setting TIF_BREAKPOINT...\n"); + set_ti_thread_flag(ti, TIF_BREAKPOINT); + + /* + * We stepped into an exception, interrupt or + * syscall handler. Some exception handlers + * don't check for pending work, so we need to + * set up a trampoline just in case. + * + * The exception entry code will undo the + * trampoline stuff if it does a full context + * save (which also means that it'll check for + * pending work later.) + */ + if ((regs->sr & MODE_MASK) == MODE_EXCEPTION) { + trampoline_addr + = (unsigned long)&debug_trampoline; + + pr_debug("Setting up trampoline...\n"); + ti->rar_saved = sysreg_read(RAR_EX); + ti->rsr_saved = sysreg_read(RSR_EX); + sysreg_write(RAR_EX, trampoline_addr); + sysreg_write(RSR_EX, (MODE_EXCEPTION + | SR_EM | SR_GM)); + BUG_ON(ti->rsr_saved & MODE_MASK); + } + + /* + * If we stepped into a system call, we + * shouldn't do a single step after we return + * since the return address is right after the + * "scall" instruction we were told to step + * over. + */ + if ((regs->sr & MODE_MASK) == MODE_SUPERVISOR) { + pr_debug("Supervisor; no single step\n"); + clear_ti_thread_flag(ti, TIF_SINGLE_STEP); + } + + ctrl = ocd_read(DC); + ctrl &= ~(1 << OCD_DC_SS_BIT); + ocd_write(DC, ctrl); + + return regs; + } else { + printk(KERN_ERR "Unexpected OCD_DS value: 0x%08x\n", + status); + printk(KERN_ERR "Thread flags: 0x%08lx\n", ti->flags); + die("Unhandled debug trap in kernel mode", + regs, SIGTRAP); + } + } else if (status & (1 << OCD_DS_SSS_BIT)) { + /* Single step in user mode */ + code = TRAP_TRACE; - if (test_thread_flag(TIF_BREAKPOINT)) { - pr_debug("TIF_BREAKPOINT set\n"); - /* We're taking care of it */ - clear_thread_flag(TIF_BREAKPOINT); - __mtdr(DBGREG_BWC2A, 0); + ctrl = ocd_read(DC); + ctrl &= ~(1 << OCD_DC_SS_BIT); + ocd_write(DC, ctrl); } - if (test_thread_flag(TIF_SINGLE_STEP)) { - pr_debug("TIF_SINGLE_STEP set, ds = 0x%08lx\n", ds); - if (ds & DS_SSS) { - dc = __mfdr(DBGREG_DC); - dc &= ~DC_SS; - __mtdr(DBGREG_DC, dc); + pr_debug("Sending SIGTRAP: code=%d PC=0x%08lx SR=0x%08lx\n", + code, regs->pc, regs->sr); - clear_thread_flag(TIF_SINGLE_STEP); - ptrace_break(current, regs); - } - } else { - /* regular breakpoint */ - ptrace_break(current, regs); - } + clear_thread_flag(TIF_SINGLE_STEP); + _exception(SIGTRAP, regs, code, instruction_pointer(regs)); + + return regs; } diff --git a/arch/avr32/kernel/semaphore.c b/arch/avr32/kernel/semaphore.c deleted file mode 100644 index 1e2705a0501..00000000000 --- a/arch/avr32/kernel/semaphore.c +++ /dev/null @@ -1,148 +0,0 @@ -/* - * AVR32 sempahore implementation. - * - * Copyright (C) 2004-2006 Atmel Corporation - * - * Based on linux/arch/i386/kernel/semaphore.c - * Copyright (C) 1999 Linus Torvalds - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License version 2 as - * published by the Free Software Foundation. - */ - -#include <linux/sched.h> -#include <linux/errno.h> -#include <linux/module.h> - -#include <asm/semaphore.h> -#include <asm/atomic.h> - -/* - * Semaphores are implemented using a two-way counter: - * The "count" variable is decremented for each process - * that tries to acquire the semaphore, while the "sleeping" - * variable is a count of such acquires. - * - * Notably, the inline "up()" and "down()" functions can - * efficiently test if they need to do any extra work (up - * needs to do something only if count was negative before - * the increment operation. - * - * "sleeping" and the contention routine ordering is protected - * by the spinlock in the semaphore's waitqueue head. - * - * Note that these functions are only called when there is - * contention on the lock, and as such all this is the - * "non-critical" part of the whole semaphore business. The - * critical part is the inline stuff in <asm/semaphore.h> - * where we want to avoid any extra jumps and calls. - */ - -/* - * Logic: - * - only on a boundary condition do we need to care. When we go - * from a negative count to a non-negative, we wake people up. - * - when we go from a non-negative count to a negative do we - * (a) synchronize with the "sleeper" count and (b) make sure - * that we're on the wakeup list before we synchronize so that - * we cannot lose wakeup events. - */ - -void __up(struct semaphore *sem) -{ - wake_up(&sem->wait); -} -EXPORT_SYMBOL(__up); - -void __sched __down(struct semaphore *sem) -{ - struct task_struct *tsk = current; - DECLARE_WAITQUEUE(wait, tsk); - unsigned long flags; - - tsk->state = TASK_UNINTERRUPTIBLE; - spin_lock_irqsave(&sem->wait.lock, flags); - add_wait_queue_exclusive_locked(&sem->wait, &wait); - - sem->sleepers++; - for (;;) { - int sleepers = sem->sleepers; - - /* - * Add "everybody else" into it. They aren't - * playing, because we own the spinlock in - * the wait_queue_head. - */ - if (atomic_add_return(sleepers - 1, &sem->count) >= 0) { - sem->sleepers = 0; - break; - } - sem->sleepers = 1; /* us - see -1 above */ - spin_unlock_irqrestore(&sem->wait.lock, flags); - - schedule(); - - spin_lock_irqsave(&sem->wait.lock, flags); - tsk->state = TASK_UNINTERRUPTIBLE; - } - remove_wait_queue_locked(&sem->wait, &wait); - wake_up_locked(&sem->wait); - spin_unlock_irqrestore(&sem->wait.lock, flags); - tsk->state = TASK_RUNNING; -} -EXPORT_SYMBOL(__down); - -int __sched __down_interruptible(struct semaphore *sem) -{ - int retval = 0; - struct task_struct *tsk = current; - DECLARE_WAITQUEUE(wait, tsk); - unsigned long flags; - - tsk->state = TASK_INTERRUPTIBLE; - spin_lock_irqsave(&sem->wait.lock, flags); - add_wait_queue_exclusive_locked(&sem->wait, &wait); - - sem->sleepers++; - for (;;) { - int sleepers = sem->sleepers; - - /* - * With signals pending, this turns into the trylock - * failure case - we won't be sleeping, and we can't - * get the lock as it has contention. Just correct the - * count and exit. - */ - if (signal_pending(current)) { - retval = -EINTR; - sem->sleepers = 0; - atomic_add(sleepers, &sem->count); - break; - } - - /* - * Add "everybody else" into it. They aren't - * playing, because we own the spinlock in - * the wait_queue_head. - */ - if (atomic_add_return(sleepers - 1, &sem->count) >= 0) { - sem->sleepers = 0; - break; - } - sem->sleepers = 1; /* us - see -1 above */ - spin_unlock_irqrestore(&sem->wait.lock, flags); - - schedule(); - - spin_lock_irqsave(&sem->wait.lock, flags); - tsk->state = TASK_INTERRUPTIBLE; - } - remove_wait_queue_locked(&sem->wait, &wait); - wake_up_locked(&sem->wait); - spin_unlock_irqrestore(&sem->wait.lock, flags); - - tsk->state = TASK_RUNNING; - return retval; -} -EXPORT_SYMBOL(__down_interruptible); diff --git a/arch/avr32/kernel/setup.c b/arch/avr32/kernel/setup.c index 5d68f3c6990..209ae5ad349 100644 --- a/arch/avr32/kernel/setup.c +++ b/arch/avr32/kernel/setup.c @@ -8,14 +8,17 @@ #include <linux/clk.h> #include <linux/init.h> +#include <linux/initrd.h> #include <linux/sched.h> #include <linux/console.h> #include <linux/ioport.h> #include <linux/bootmem.h> #include <linux/fs.h> #include <linux/module.h> +#include <linux/pfn.h> #include <linux/root_dev.h> #include <linux/cpu.h> +#include <linux/kernel.h> #include <asm/sections.h> #include <asm/processor.h> @@ -23,19 +26,12 @@ #include <asm/setup.h> #include <asm/sysreg.h> -#include <asm/arch/board.h> -#include <asm/arch/init.h> +#include <mach/board.h> +#include <mach/init.h> extern int root_mountflags; /* - * Bootloader-provided information about physical memory - */ -struct tag_mem_range *mem_phys; -struct tag_mem_range *mem_reserved; -struct tag_mem_range *mem_ramdisk; - -/* * Initialize loops_per_jiffy as 5000000 (500MIPS). * Better make it too large than too small... */ @@ -44,51 +40,197 @@ struct avr32_cpuinfo boot_cpu_data = { }; EXPORT_SYMBOL(boot_cpu_data); -static char command_line[COMMAND_LINE_SIZE]; +static char __initdata command_line[COMMAND_LINE_SIZE]; /* - * Should be more than enough, but if you have a _really_ complex - * setup, you might need to increase the size of this... + * Standard memory resources */ -static struct tag_mem_range __initdata mem_range_cache[32]; -static unsigned mem_range_next_free; +static struct resource __initdata kernel_data = { + .name = "Kernel data", + .start = 0, + .end = 0, + .flags = IORESOURCE_MEM, +}; +static struct resource __initdata kernel_code = { + .name = "Kernel code", + .start = 0, + .end = 0, + .flags = IORESOURCE_MEM, + .sibling = &kernel_data, +}; /* - * Standard memory resources + * Available system RAM and reserved regions as singly linked + * lists. These lists are traversed using the sibling pointer in + * struct resource and are kept sorted at all times. */ -static struct resource mem_res[] = { - { - .name = "Kernel code", - .start = 0, - .end = 0, - .flags = IORESOURCE_MEM - }, - { - .name = "Kernel data", - .start = 0, - .end = 0, - .flags = IORESOURCE_MEM, - }, -}; +static struct resource *__initdata system_ram; +static struct resource *__initdata reserved = &kernel_code; + +/* + * We need to allocate these before the bootmem allocator is up and + * running, so we need this "cache". 32 entries are probably enough + * for all but the most insanely complex systems. + */ +static struct resource __initdata res_cache[32]; +static unsigned int __initdata res_cache_next_free; + +static void __init resource_init(void) +{ + struct resource *mem, *res; + struct resource *new; + + kernel_code.start = __pa(init_mm.start_code); + + for (mem = system_ram; mem; mem = mem->sibling) { + new = alloc_bootmem_low(sizeof(struct resource)); + memcpy(new, mem, sizeof(struct resource)); + + new->sibling = NULL; + if (request_resource(&iomem_resource, new)) + printk(KERN_WARNING "Bad RAM resource %08x-%08x\n", + mem->start, mem->end); + } + + for (res = reserved; res; res = res->sibling) { + new = alloc_bootmem_low(sizeof(struct resource)); + memcpy(new, res, sizeof(struct resource)); + + new->sibling = NULL; + if (insert_resource(&iomem_resource, new)) + printk(KERN_WARNING + "Bad reserved resource %s (%08x-%08x)\n", + res->name, res->start, res->end); + } +} + +static void __init +add_physical_memory(resource_size_t start, resource_size_t end) +{ + struct resource *new, *next, **pprev; + + for (pprev = &system_ram, next = system_ram; next; + pprev = &next->sibling, next = next->sibling) { + if (end < next->start) + break; + if (start <= next->end) { + printk(KERN_WARNING + "Warning: Physical memory map is broken\n"); + printk(KERN_WARNING + "Warning: %08x-%08x overlaps %08x-%08x\n", + start, end, next->start, next->end); + return; + } + } + + if (res_cache_next_free >= ARRAY_SIZE(res_cache)) { + printk(KERN_WARNING + "Warning: Failed to add physical memory %08x-%08x\n", + start, end); + return; + } + + new = &res_cache[res_cache_next_free++]; + new->start = start; + new->end = end; + new->name = "System RAM"; + new->flags = IORESOURCE_MEM; + + *pprev = new; +} + +static int __init +add_reserved_region(resource_size_t start, resource_size_t end, + const char *name) +{ + struct resource *new, *next, **pprev; + + if (end < start) + return -EINVAL; -#define kernel_code mem_res[0] -#define kernel_data mem_res[1] + if (res_cache_next_free >= ARRAY_SIZE(res_cache)) + return -ENOMEM; + + for (pprev = &reserved, next = reserved; next; + pprev = &next->sibling, next = next->sibling) { + if (end < next->start) + break; + if (start <= next->end) + return -EBUSY; + } + + new = &res_cache[res_cache_next_free++]; + new->start = start; + new->end = end; + new->name = name; + new->sibling = next; + new->flags = IORESOURCE_MEM; + + *pprev = new; + + return 0; +} + +static unsigned long __init +find_free_region(const struct resource *mem, resource_size_t size, + resource_size_t align) +{ + struct resource *res; + unsigned long target; + + target = ALIGN(mem->start, align); + for (res = reserved; res; res = res->sibling) { + if ((target + size) <= res->start) + break; + if (target <= res->end) + target = ALIGN(res->end + 1, align); + } + + if ((target + size) > (mem->end + 1)) + return mem->end + 1; + + return target; +} + +static int __init +alloc_reserved_region(resource_size_t *start, resource_size_t size, + resource_size_t align, const char *name) +{ + struct resource *mem; + resource_size_t target; + int ret; + + for (mem = system_ram; mem; mem = mem->sibling) { + target = find_free_region(mem, size, align); + if (target <= mem->end) { + ret = add_reserved_region(target, target + size - 1, + name); + if (!ret) + *start = target; + return ret; + } + } + + return -ENOMEM; +} /* * Early framebuffer allocation. Works as follows: * - If fbmem_size is zero, nothing will be allocated or reserved. * - If fbmem_start is zero when setup_bootmem() is called, - * fbmem_size bytes will be allocated from the bootmem allocator. + * a block of fbmem_size bytes will be reserved before bootmem + * initialization. It will be aligned to the largest page size + * that fbmem_size is a multiple of. * - If fbmem_start is nonzero, an area of size fbmem_size will be - * reserved at the physical address fbmem_start if necessary. If - * the area isn't in a memory region known to the kernel, it will - * be left alone. + * reserved at the physical address fbmem_start if possible. If + * it collides with other reserved memory, a different block of + * same size will be allocated, just as if fbmem_start was zero. * * Board-specific code may use these variables to set up platform data * for the framebuffer driver if fbmem_size is nonzero. */ -static unsigned long __initdata fbmem_start; -static unsigned long __initdata fbmem_size; +resource_size_t __initdata fbmem_start; +resource_size_t __initdata fbmem_size; /* * "fbmem=xxx[kKmM]" allocates the specified amount of boot memory for @@ -102,48 +244,63 @@ static unsigned long __initdata fbmem_size; */ static int __init early_parse_fbmem(char *p) { + int ret; + unsigned long align; + fbmem_size = memparse(p, &p); - if (*p == '@') - fbmem_start = memparse(p, &p); + if (*p == '@') { + fbmem_start = memparse(p + 1, &p); + ret = add_reserved_region(fbmem_start, + fbmem_start + fbmem_size - 1, + "Framebuffer"); + if (ret) { + printk(KERN_WARNING + "Failed to reserve framebuffer memory\n"); + fbmem_start = 0; + } + } + + if (!fbmem_start) { + if ((fbmem_size & 0x000fffffUL) == 0) + align = 0x100000; /* 1 MiB */ + else if ((fbmem_size & 0x0000ffffUL) == 0) + align = 0x10000; /* 64 KiB */ + else + align = 0x1000; /* 4 KiB */ + + ret = alloc_reserved_region(&fbmem_start, fbmem_size, + align, "Framebuffer"); + if (ret) { + printk(KERN_WARNING + "Failed to allocate framebuffer memory\n"); + fbmem_size = 0; + } else { + memset(__va(fbmem_start), 0, fbmem_size); + } + } + return 0; } early_param("fbmem", early_parse_fbmem); -static inline void __init resource_init(void) +/* + * Pick out the memory size. We look for mem=size@start, + * where start and size are "size[KkMmGg]" + */ +static int __init early_mem(char *p) { - struct tag_mem_range *region; - - kernel_code.start = __pa(init_mm.start_code); - kernel_code.end = __pa(init_mm.end_code - 1); - kernel_data.start = __pa(init_mm.end_code); - kernel_data.end = __pa(init_mm.brk - 1); - - for (region = mem_phys; region; region = region->next) { - struct resource *res; - unsigned long phys_start, phys_end; - - if (region->size == 0) - continue; - - phys_start = region->addr; - phys_end = phys_start + region->size - 1; - - res = alloc_bootmem_low(sizeof(*res)); - res->name = "System RAM"; - res->start = phys_start; - res->end = phys_end; - res->flags = IORESOURCE_MEM | IORESOURCE_BUSY; + resource_size_t size, start; - request_resource (&iomem_resource, res); + start = system_ram->start; + size = memparse(p, &p); + if (*p == '@') + start = memparse(p + 1, &p); - if (kernel_code.start >= res->start && - kernel_code.end <= res->end) - request_resource (res, &kernel_code); - if (kernel_data.start >= res->start && - kernel_data.end <= res->end) - request_resource (res, &kernel_data); - } + system_ram->start = start; + system_ram->end = system_ram->start + size - 1; + return 0; } +early_param("mem", early_mem); static int __init parse_tag_core(struct tag *tag) { @@ -156,11 +313,9 @@ static int __init parse_tag_core(struct tag *tag) } __tagtable(ATAG_CORE, parse_tag_core); -static int __init parse_tag_mem_range(struct tag *tag, - struct tag_mem_range **root) +static int __init parse_tag_mem(struct tag *tag) { - struct tag_mem_range *cur, **pprev; - struct tag_mem_range *new; + unsigned long start, end; /* * Ignore zero-sized entries. If we're running standalone, the @@ -170,49 +325,61 @@ static int __init parse_tag_mem_range(struct tag *tag, if (tag->u.mem_range.size == 0) return 0; - /* - * Copy the data so the bootmem init code doesn't need to care - * about it. - */ - if (mem_range_next_free >= - (sizeof(mem_range_cache) / sizeof(mem_range_cache[0]))) - panic("Physical memory map too complex!\n"); - - new = &mem_range_cache[mem_range_next_free++]; - *new = tag->u.mem_range; - - pprev = root; - cur = *root; - while (cur) { - pprev = &cur->next; - cur = cur->next; + start = tag->u.mem_range.addr; + end = tag->u.mem_range.addr + tag->u.mem_range.size - 1; + + add_physical_memory(start, end); + return 0; +} +__tagtable(ATAG_MEM, parse_tag_mem); + +static int __init parse_tag_rdimg(struct tag *tag) +{ +#ifdef CONFIG_BLK_DEV_INITRD + struct tag_mem_range *mem = &tag->u.mem_range; + int ret; + + if (initrd_start) { + printk(KERN_WARNING + "Warning: Only the first initrd image will be used\n"); + return 0; } - *pprev = new; - new->next = NULL; + ret = add_reserved_region(mem->addr, mem->addr + mem->size - 1, + "initrd"); + if (ret) { + printk(KERN_WARNING + "Warning: Failed to reserve initrd memory\n"); + return ret; + } + + initrd_start = (unsigned long)__va(mem->addr); + initrd_end = initrd_start + mem->size; +#else + printk(KERN_WARNING "RAM disk image present, but " + "no initrd support in kernel, ignoring\n"); +#endif return 0; } +__tagtable(ATAG_RDIMG, parse_tag_rdimg); -static int __init parse_tag_mem(struct tag *tag) +static int __init parse_tag_rsvd_mem(struct tag *tag) { - return parse_tag_mem_range(tag, &mem_phys); + struct tag_mem_range *mem = &tag->u.mem_range; + + return add_reserved_region(mem->addr, mem->addr + mem->size - 1, + "Reserved"); } -__tagtable(ATAG_MEM, parse_tag_mem); +__tagtable(ATAG_RSVD_MEM, parse_tag_rsvd_mem); static int __init parse_tag_cmdline(struct tag *tag) { - strlcpy(saved_command_line, tag->u.cmdline.cmdline, COMMAND_LINE_SIZE); + strlcpy(boot_command_line, tag->u.cmdline.cmdline, COMMAND_LINE_SIZE); return 0; } __tagtable(ATAG_CMDLINE, parse_tag_cmdline); -static int __init parse_tag_rdimg(struct tag *tag) -{ - return parse_tag_mem_range(tag, &mem_ramdisk); -} -__tagtable(ATAG_RDIMG, parse_tag_rdimg); - static int __init parse_tag_clock(struct tag *tag) { /* @@ -223,35 +390,20 @@ static int __init parse_tag_clock(struct tag *tag) } __tagtable(ATAG_CLOCK, parse_tag_clock); -static int __init parse_tag_rsvd_mem(struct tag *tag) -{ - return parse_tag_mem_range(tag, &mem_reserved); -} -__tagtable(ATAG_RSVD_MEM, parse_tag_rsvd_mem); +/* + * The board_number correspond to the bd->bi_board_number in U-Boot. This + * parameter is only available during initialisation and can be used in some + * kind of board identification. + */ +u32 __initdata board_number; -static int __init parse_tag_ethernet(struct tag *tag) +static int __init parse_tag_boardinfo(struct tag *tag) { -#if 0 - const struct platform_device *pdev; + board_number = tag->u.boardinfo.board_number; - /* - * We really need a bus type that supports "classes"...this - * will do for now (until we must handle other kinds of - * ethernet controllers) - */ - pdev = platform_get_device("macb", tag->u.ethernet.mac_index); - if (pdev && pdev->dev.platform_data) { - struct eth_platform_data *data = pdev->dev.platform_data; - - data->valid = 1; - data->mii_phy_addr = tag->u.ethernet.mii_phy_addr; - memcpy(data->hw_addr, tag->u.ethernet.hw_address, - sizeof(data->hw_addr)); - } -#endif return 0; } -__tagtable(ATAG_ETHERNET, parse_tag_ethernet); +__tagtable(ATAG_BOARDINFO, parse_tag_boardinfo); /* * Scan the tag table for this tag, and call its parse function. The @@ -284,14 +436,144 @@ static void __init parse_tags(struct tag *t) t->hdr.tag); } +/* + * Find a free memory region large enough for storing the + * bootmem bitmap. + */ +static unsigned long __init +find_bootmap_pfn(const struct resource *mem) +{ + unsigned long bootmap_pages, bootmap_len; + unsigned long node_pages = PFN_UP(resource_size(mem)); + unsigned long bootmap_start; + + bootmap_pages = bootmem_bootmap_pages(node_pages); + bootmap_len = bootmap_pages << PAGE_SHIFT; + + /* + * Find a large enough region without reserved pages for + * storing the bootmem bitmap. We can take advantage of the + * fact that all lists have been sorted. + * + * We have to check that we don't collide with any reserved + * regions, which includes the kernel image and any RAMDISK + * images. + */ + bootmap_start = find_free_region(mem, bootmap_len, PAGE_SIZE); + + return bootmap_start >> PAGE_SHIFT; +} + +#define MAX_LOWMEM HIGHMEM_START +#define MAX_LOWMEM_PFN PFN_DOWN(MAX_LOWMEM) + +static void __init setup_bootmem(void) +{ + unsigned bootmap_size; + unsigned long first_pfn, bootmap_pfn, pages; + unsigned long max_pfn, max_low_pfn; + unsigned node = 0; + struct resource *res; + + printk(KERN_INFO "Physical memory:\n"); + for (res = system_ram; res; res = res->sibling) + printk(" %08x-%08x\n", res->start, res->end); + printk(KERN_INFO "Reserved memory:\n"); + for (res = reserved; res; res = res->sibling) + printk(" %08x-%08x: %s\n", + res->start, res->end, res->name); + + nodes_clear(node_online_map); + + if (system_ram->sibling) + printk(KERN_WARNING "Only using first memory bank\n"); + + for (res = system_ram; res; res = NULL) { + first_pfn = PFN_UP(res->start); + max_low_pfn = max_pfn = PFN_DOWN(res->end + 1); + bootmap_pfn = find_bootmap_pfn(res); + if (bootmap_pfn > max_pfn) + panic("No space for bootmem bitmap!\n"); + + if (max_low_pfn > MAX_LOWMEM_PFN) { + max_low_pfn = MAX_LOWMEM_PFN; +#ifndef CONFIG_HIGHMEM + /* + * Lowmem is memory that can be addressed + * directly through P1/P2 + */ + printk(KERN_WARNING + "Node %u: Only %ld MiB of memory will be used.\n", + node, MAX_LOWMEM >> 20); + printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); +#else +#error HIGHMEM is not supported by AVR32 yet +#endif + } + + /* Initialize the boot-time allocator with low memory only. */ + bootmap_size = init_bootmem_node(NODE_DATA(node), bootmap_pfn, + first_pfn, max_low_pfn); + + /* + * Register fully available RAM pages with the bootmem + * allocator. + */ + pages = max_low_pfn - first_pfn; + free_bootmem_node (NODE_DATA(node), PFN_PHYS(first_pfn), + PFN_PHYS(pages)); + + /* Reserve space for the bootmem bitmap... */ + reserve_bootmem_node(NODE_DATA(node), + PFN_PHYS(bootmap_pfn), + bootmap_size, + BOOTMEM_DEFAULT); + + /* ...and any other reserved regions. */ + for (res = reserved; res; res = res->sibling) { + if (res->start > PFN_PHYS(max_pfn)) + break; + + /* + * resource_init will complain about partial + * overlaps, so we'll just ignore such + * resources for now. + */ + if (res->start >= PFN_PHYS(first_pfn) + && res->end < PFN_PHYS(max_pfn)) + reserve_bootmem_node(NODE_DATA(node), + res->start, + resource_size(res), + BOOTMEM_DEFAULT); + } + + node_set_online(node); + } +} + void __init setup_arch (char **cmdline_p) { struct clk *cpu_clk; + init_mm.start_code = (unsigned long)_stext; + init_mm.end_code = (unsigned long)_etext; + init_mm.end_data = (unsigned long)_edata; + init_mm.brk = (unsigned long)_end; + + /* + * Include .init section to make allocations easier. It will + * be removed before the resource is actually requested. + */ + kernel_code.start = __pa(__init_begin); + kernel_code.end = __pa(init_mm.end_code - 1); + kernel_data.start = __pa(init_mm.end_code); + kernel_data.end = __pa(init_mm.brk - 1); + parse_tags(bootloader_tags); setup_processor(); setup_platform(); + setup_board(); cpu_clk = clk_get(NULL, "cpu"); if (IS_ERR(cpu_clk)) { @@ -312,24 +594,16 @@ void __init setup_arch (char **cmdline_p) ((cpu_hz + 500) / 1000) % 1000); } - init_mm.start_code = (unsigned long) &_text; - init_mm.end_code = (unsigned long) &_etext; - init_mm.end_data = (unsigned long) &_edata; - init_mm.brk = (unsigned long) &_end; - - strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE); + strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); *cmdline_p = command_line; parse_early_param(); setup_bootmem(); - board_setup_fbmem(fbmem_start, fbmem_size); - #ifdef CONFIG_VT conswitchp = &dummy_con; #endif paging_init(); - resource_init(); } diff --git a/arch/avr32/kernel/signal.c b/arch/avr32/kernel/signal.c index 33096651c24..b80c0b3d2ba 100644 --- a/arch/avr32/kernel/signal.c +++ b/arch/avr32/kernel/signal.c @@ -15,18 +15,11 @@ #include <linux/errno.h> #include <linux/ptrace.h> #include <linux/unistd.h> -#include <linux/suspend.h> +#include <linux/tracehook.h> #include <asm/uaccess.h> #include <asm/ucontext.h> - -#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP))) - -asmlinkage int sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss, - struct pt_regs *regs) -{ - return do_sigaltstack(uss, uoss, regs->sp); -} +#include <asm/syscalls.h> struct rt_sigframe { @@ -75,6 +68,9 @@ asmlinkage int sys_rt_sigreturn(struct pt_regs *regs) struct rt_sigframe __user *frame; sigset_t set; + /* Always make any pending restarted system calls return -EINTR */ + current_thread_info()->restart_block.fn = do_no_restart_syscall; + frame = (struct rt_sigframe __user *)regs->sp; pr_debug("SIG return: frame = %p\n", frame); @@ -84,15 +80,14 @@ asmlinkage int sys_rt_sigreturn(struct pt_regs *regs) if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set))) goto badframe; - sigdelsetmask(&set, ~_BLOCKABLE); - spin_lock_irq(¤t->sighand->siglock); - current->blocked = set; - recalc_sigpending(); - spin_unlock_irq(¤t->sighand->siglock); + set_current_blocked(&set); if (restore_sigcontext(regs, &frame->uc.uc_mcontext)) goto badframe; + if (restore_altstack(&frame->uc.uc_stack)) + goto badframe; + pr_debug("Context restored: pc = %08lx, lr = %08lx, sp = %08lx\n", regs->pc, regs->lr, regs->sp); @@ -174,12 +169,7 @@ setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info, /* Set up the ucontext */ err |= __put_user(0, &frame->uc.uc_flags); err |= __put_user(NULL, &frame->uc.uc_link); - err |= __put_user((void __user *)current->sas_ss_sp, - &frame->uc.uc_stack.ss_sp); - err |= __put_user(sas_ss_flags(regs->sp), - &frame->uc.uc_stack.ss_flags); - err |= __put_user(current->sas_ss_size, - &frame->uc.uc_stack.ss_size); + err |= __save_altstack(&frame->uc.uc_stack, regs->sp); err |= setup_sigcontext(&frame->uc.uc_mcontext, regs); err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set)); @@ -208,7 +198,7 @@ out: return err; } -static inline void restart_syscall(struct pt_regs *regs) +static inline void setup_syscall_restart(struct pt_regs *regs) { if (regs->r12 == -ERESTART_RESTARTBLOCK) regs->r8 = __NR_restart_syscall; @@ -219,14 +209,14 @@ static inline void restart_syscall(struct pt_regs *regs) static inline void handle_signal(unsigned long sig, struct k_sigaction *ka, siginfo_t *info, - sigset_t *oldset, struct pt_regs *regs, int syscall) + struct pt_regs *regs, int syscall) { int ret; /* * Set up the stack frame */ - ret = setup_rt_frame(sig, ka, info, oldset, regs); + ret = setup_rt_frame(sig, ka, info, sigmask_to_save(), regs); /* * Check that the resulting registers are sane @@ -234,21 +224,12 @@ handle_signal(unsigned long sig, struct k_sigaction *ka, siginfo_t *info, ret |= !valid_user_regs(regs); /* - * Block the signal if we were unsuccessful. + * Block the signal if we were successful. */ - if (ret != 0 || !(ka->sa.sa_flags & SA_NODEFER)) { - spin_lock_irq(¤t->sighand->siglock); - sigorsets(¤t->blocked, ¤t->blocked, - &ka->sa.sa_mask); - sigaddset(¤t->blocked, sig); - recalc_sigpending(); - spin_unlock_irq(¤t->sighand->siglock); - } - - if (ret == 0) - return; - - force_sigsegv(sig, current); + if (ret != 0) + force_sigsegv(sig, current); + else + signal_delivered(sig, info, ka, regs, 0); } /* @@ -256,7 +237,7 @@ handle_signal(unsigned long sig, struct k_sigaction *ka, siginfo_t *info, * doesn't want to handle. Thus you cannot kill init even with a * SIGKILL even by mistake. */ -int do_signal(struct pt_regs *regs, sigset_t *oldset, int syscall) +static void do_signal(struct pt_regs *regs, int syscall) { siginfo_t info; int signr; @@ -268,21 +249,9 @@ int do_signal(struct pt_regs *regs, sigset_t *oldset, int syscall) * without doing anything if so. */ if (!user_mode(regs)) - return 0; - - if (try_to_freeze()) { - signr = 0; - if (!signal_pending(current)) - goto no_signal; - } - - if (test_thread_flag(TIF_RESTORE_SIGMASK)) - oldset = ¤t->saved_sigmask; - else if (!oldset) - oldset = ¤t->blocked; + return; signr = get_signal_to_deliver(&info, &ka, regs, NULL); -no_signal: if (syscall) { switch (regs->r12) { case -ERESTART_RESTARTBLOCK: @@ -299,21 +268,17 @@ no_signal: } /* fall through */ case -ERESTARTNOINTR: - restart_syscall(regs); + setup_syscall_restart(regs); } } if (signr == 0) { /* No signal to deliver -- put the saved sigmask back */ - if (test_thread_flag(TIF_RESTORE_SIGMASK)) { - clear_thread_flag(TIF_RESTORE_SIGMASK); - sigprocmask(SIG_SETMASK, ¤t->saved_sigmask, NULL); - } - return 0; + restore_saved_sigmask(); + return; } - handle_signal(signr, &ka, &info, oldset, regs, syscall); - return 1; + handle_signal(signr, &ka, &info, regs, syscall); } asmlinkage void do_notify_resume(struct pt_regs *regs, struct thread_info *ti) @@ -323,6 +288,11 @@ asmlinkage void do_notify_resume(struct pt_regs *regs, struct thread_info *ti) if ((sysreg_read(SR) & MODE_MASK) == MODE_SUPERVISOR) syscall = 1; - if (ti->flags & (_TIF_SIGPENDING | _TIF_RESTORE_SIGMASK)) - do_signal(regs, ¤t->blocked, syscall); + if (ti->flags & _TIF_SIGPENDING) + do_signal(regs, syscall); + + if (ti->flags & _TIF_NOTIFY_RESUME) { + clear_thread_flag(TIF_NOTIFY_RESUME); + tracehook_notify_resume(regs); + } } diff --git a/arch/avr32/kernel/stacktrace.c b/arch/avr32/kernel/stacktrace.c new file mode 100644 index 00000000000..c09f0d8dd67 --- /dev/null +++ b/arch/avr32/kernel/stacktrace.c @@ -0,0 +1,55 @@ +/* + * Stack trace management functions + * + * Copyright (C) 2007 Atmel Corporation + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ +#include <linux/sched.h> +#include <linux/stacktrace.h> +#include <linux/thread_info.h> +#include <linux/module.h> + +register unsigned long current_frame_pointer asm("r7"); + +struct stackframe { + unsigned long lr; + unsigned long fp; +}; + +/* + * Save stack-backtrace addresses into a stack_trace buffer. + */ +void save_stack_trace(struct stack_trace *trace) +{ + unsigned long low, high; + unsigned long fp; + struct stackframe *frame; + int skip = trace->skip; + + low = (unsigned long)task_stack_page(current); + high = low + THREAD_SIZE; + fp = current_frame_pointer; + + while (fp >= low && fp <= (high - 8)) { + frame = (struct stackframe *)fp; + + if (skip) { + skip--; + } else { + trace->entries[trace->nr_entries++] = frame->lr; + if (trace->nr_entries >= trace->max_entries) + break; + } + + /* + * The next frame must be at a higher address than the + * current frame. + */ + low = fp + 8; + fp = frame->fp; + } +} +EXPORT_SYMBOL_GPL(save_stack_trace); diff --git a/arch/avr32/kernel/sys_avr32.c b/arch/avr32/kernel/sys_avr32.c deleted file mode 100644 index 6ec5693da44..00000000000 --- a/arch/avr32/kernel/sys_avr32.c +++ /dev/null @@ -1,51 +0,0 @@ -/* - * Copyright (C) 2004-2006 Atmel Corporation - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License version 2 as - * published by the Free Software Foundation. - */ -#include <linux/errno.h> -#include <linux/fs.h> -#include <linux/file.h> -#include <linux/mm.h> -#include <linux/unistd.h> - -#include <asm/mman.h> -#include <asm/uaccess.h> - -asmlinkage int sys_pipe(unsigned long __user *filedes) -{ - int fd[2]; - int error; - - error = do_pipe(fd); - if (!error) { - if (copy_to_user(filedes, fd, sizeof(fd))) - error = -EFAULT; - } - return error; -} - -asmlinkage long sys_mmap2(unsigned long addr, unsigned long len, - unsigned long prot, unsigned long flags, - unsigned long fd, off_t offset) -{ - int error = -EBADF; - struct file *file = NULL; - - flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE); - if (!(flags & MAP_ANONYMOUS)) { - file = fget(fd); - if (!file) - return error; - } - - down_write(¤t->mm->mmap_sem); - error = do_mmap_pgoff(file, addr, len, prot, flags, offset); - up_write(¤t->mm->mmap_sem); - - if (file) - fput(file); - return error; -} diff --git a/arch/avr32/kernel/syscall-stubs.S b/arch/avr32/kernel/syscall-stubs.S index 7589a9b426c..b5fc927cd39 100644 --- a/arch/avr32/kernel/syscall-stubs.S +++ b/arch/avr32/kernel/syscall-stubs.S @@ -20,48 +20,18 @@ __sys_rt_sigsuspend: mov r10, sp rjmp sys_rt_sigsuspend - .global __sys_sigaltstack - .type __sys_sigaltstack,@function -__sys_sigaltstack: - mov r10, sp - rjmp sys_sigaltstack - .global __sys_rt_sigreturn .type __sys_rt_sigreturn,@function __sys_rt_sigreturn: mov r12, sp rjmp sys_rt_sigreturn - .global __sys_fork - .type __sys_fork,@function -__sys_fork: - mov r12, sp - rjmp sys_fork - - .global __sys_clone - .type __sys_clone,@function -__sys_clone: - mov r8, sp - rjmp sys_clone - - .global __sys_vfork - .type __sys_vfork,@function -__sys_vfork: - mov r12, sp - rjmp sys_vfork - - .global __sys_execve - .type __sys_execve,@function -__sys_execve: - mov r9, sp - rjmp sys_execve - .global __sys_mmap2 .type __sys_mmap2,@function __sys_mmap2: pushm lr st.w --sp, ARG6 - rcall sys_mmap2 + call sys_mmap_pgoff sub sp, -4 popm pc @@ -70,7 +40,7 @@ __sys_mmap2: __sys_sendto: pushm lr st.w --sp, ARG6 - rcall sys_sendto + call sys_sendto sub sp, -4 popm pc @@ -79,7 +49,7 @@ __sys_sendto: __sys_recvfrom: pushm lr st.w --sp, ARG6 - rcall sys_recvfrom + call sys_recvfrom sub sp, -4 popm pc @@ -88,7 +58,7 @@ __sys_recvfrom: __sys_pselect6: pushm lr st.w --sp, ARG6 - rcall sys_pselect6 + call sys_pselect6 sub sp, -4 popm pc @@ -97,6 +67,24 @@ __sys_pselect6: __sys_splice: pushm lr st.w --sp, ARG6 - rcall sys_splice + call sys_splice + sub sp, -4 + popm pc + + .global __sys_epoll_pwait + .type __sys_epoll_pwait,@function +__sys_epoll_pwait: + pushm lr + st.w --sp, ARG6 + call sys_epoll_pwait + sub sp, -4 + popm pc + + .global __sys_sync_file_range + .type __sys_sync_file_range,@function +__sys_sync_file_range: + pushm lr + st.w --sp, ARG6 + call sys_sync_file_range sub sp, -4 popm pc diff --git a/arch/avr32/kernel/syscall_table.S b/arch/avr32/kernel/syscall_table.S index 63b206965d0..017a904180c 100644 --- a/arch/avr32/kernel/syscall_table.S +++ b/arch/avr32/kernel/syscall_table.S @@ -8,14 +8,6 @@ * published by the Free Software Foundation. */ -#if !defined(CONFIG_NFSD) && !defined(CONFIG_NFSD_MODULE) -#define sys_nfsservctl sys_ni_syscall -#endif - -#if !defined(CONFIG_SYSV_IPC) -# define sys_ipc sys_ni_syscall -#endif - .section .rodata,"a",@progbits .type sys_call_table,@object .global sys_call_table @@ -23,7 +15,7 @@ sys_call_table: .long sys_restart_syscall .long sys_exit - .long __sys_fork + .long sys_fork .long sys_read .long sys_write .long sys_open /* 5 */ @@ -32,7 +24,7 @@ sys_call_table: .long sys_creat .long sys_link .long sys_unlink /* 10 */ - .long __sys_execve + .long sys_execve .long sys_chdir .long sys_time .long sys_mknod @@ -65,7 +57,7 @@ sys_call_table: .long sys_dup .long sys_pipe .long sys_times - .long __sys_clone + .long sys_clone .long sys_brk /* 45 */ .long sys_setgid .long sys_getgid @@ -123,19 +115,19 @@ sys_call_table: .long sys_statfs .long sys_fstatfs /* 100 */ .long sys_vhangup - .long __sys_sigaltstack + .long sys_sigaltstack .long sys_syslog .long sys_setitimer .long sys_getitimer /* 105 */ .long sys_swapoff .long sys_sysinfo - .long sys_ipc + .long sys_ni_syscall /* was sys_ipc briefly */ .long sys_sendfile .long sys_setdomainname /* 110 */ .long sys_newuname .long sys_adjtimex .long sys_mprotect - .long __sys_vfork + .long sys_vfork .long sys_init_module /* 115 */ .long sys_delete_module .long sys_quotactl @@ -166,7 +158,7 @@ sys_call_table: .long sys_sched_rr_get_interval .long sys_nanosleep .long sys_poll - .long sys_nfsservctl /* 145 */ + .long sys_ni_syscall /* 145 was nfsservctl */ .long sys_setresgid .long sys_getresgid .long sys_prctl @@ -283,7 +275,26 @@ sys_call_table: .long sys_set_robust_list .long sys_get_robust_list /* 260 */ .long __sys_splice - .long sys_sync_file_range + .long __sys_sync_file_range .long sys_tee .long sys_vmsplice + .long __sys_epoll_pwait /* 265 */ + .long sys_msgget + .long sys_msgsnd + .long sys_msgrcv + .long sys_msgctl + .long sys_semget /* 270 */ + .long sys_semop + .long sys_semctl + .long sys_semtimedop + .long sys_shmat + .long sys_shmget /* 275 */ + .long sys_shmdt + .long sys_shmctl + .long sys_utimensat + .long sys_signalfd + .long sys_ni_syscall /* 280, was sys_timerfd */ + .long sys_eventfd + .long sys_recvmmsg + .long sys_setns .long sys_ni_syscall /* r8 is saturated at nr_syscalls */ diff --git a/arch/avr32/kernel/time.c b/arch/avr32/kernel/time.c index b0e6b5855a3..d0f771be9e9 100644 --- a/arch/avr32/kernel/time.c +++ b/arch/avr32/kernel/time.c @@ -1,238 +1,160 @@ /* - * Copyright (C) 2004-2006 Atmel Corporation - * - * Based on MIPS implementation arch/mips/kernel/time.c - * Copyright 2001 MontaVista Software Inc. + * Copyright (C) 2004-2007 Atmel Corporation * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. */ - #include <linux/clk.h> -#include <linux/clocksource.h> -#include <linux/time.h> -#include <linux/module.h> +#include <linux/clockchips.h> +#include <linux/init.h> #include <linux/interrupt.h> #include <linux/irq.h> -#include <linux/kernel_stat.h> -#include <linux/errno.h> -#include <linux/init.h> -#include <linux/profile.h> -#include <linux/sysdev.h> +#include <linux/kernel.h> +#include <linux/time.h> +#include <linux/cpu.h> -#include <asm/div64.h> #include <asm/sysreg.h> -#include <asm/io.h> -#include <asm/sections.h> - -static cycle_t read_cycle_count(void) -{ - return (cycle_t)sysreg_read(COUNT); -} - -static struct clocksource clocksource_avr32 = { - .name = "avr32", - .rating = 350, - .read = read_cycle_count, - .mask = CLOCKSOURCE_MASK(32), - .shift = 16, - .is_continuous = 1, -}; - -/* - * By default we provide the null RTC ops - */ -static unsigned long null_rtc_get_time(void) -{ - return mktime(2004, 1, 1, 0, 0, 0); -} -static int null_rtc_set_time(unsigned long sec) -{ - return 0; -} - -static unsigned long (*rtc_get_time)(void) = null_rtc_get_time; -static int (*rtc_set_time)(unsigned long) = null_rtc_set_time; - -/* how many counter cycles in a jiffy? */ -static unsigned long cycles_per_jiffy; - -/* cycle counter value at the previous timer interrupt */ -static unsigned int timerhi, timerlo; - -/* the count value for the next timer interrupt */ -static unsigned int expirelo; - -static void avr32_timer_ack(void) -{ - unsigned int count; - - /* Ack this timer interrupt and set the next one */ - expirelo += cycles_per_jiffy; - if (expirelo == 0) { - printk(KERN_DEBUG "expirelo == 0\n"); - sysreg_write(COMPARE, expirelo + 1); - } else { - sysreg_write(COMPARE, expirelo); - } +#include <mach/pm.h> - /* Check to see if we have missed any timer interrupts */ - count = sysreg_read(COUNT); - if ((count - expirelo) < 0x7fffffff) { - expirelo = count + cycles_per_jiffy; - sysreg_write(COMPARE, expirelo); - } -} - -static unsigned int avr32_hpt_read(void) -{ - return sysreg_read(COUNT); -} - -/* - * Taken from MIPS c0_hpt_timer_init(). - * - * Why is it so complicated, and what is "count"? My assumption is - * that `count' specifies the "reference cycle", i.e. the cycle since - * reset that should mean "zero". The reason COUNT is written twice is - * probably to make sure we don't get any timer interrupts while we - * are messing with the counter. - */ -static void avr32_hpt_init(unsigned int count) -{ - count = sysreg_read(COUNT) - count; - expirelo = (count / cycles_per_jiffy + 1) * cycles_per_jiffy; - sysreg_write(COUNT, expirelo - cycles_per_jiffy); - sysreg_write(COMPARE, expirelo); - sysreg_write(COUNT, count); -} -/* - * Scheduler clock - returns current time in nanosec units. - */ -unsigned long long sched_clock(void) +static cycle_t read_cycle_count(struct clocksource *cs) { - /* There must be better ways...? */ - return (unsigned long long)jiffies * (1000000000 / HZ); + return (cycle_t)sysreg_read(COUNT); } /* - * local_timer_interrupt() does profiling and process accounting on a - * per-CPU basis. - * - * In UP mode, it is invoked from the (global) timer_interrupt. + * The architectural cycle count registers are a fine clocksource unless + * the system idle loop use sleep states like "idle": the CPU cycles + * measured by COUNT (and COMPARE) don't happen during sleep states. + * Their duration also changes if cpufreq changes the CPU clock rate. + * So we rate the clocksource using COUNT as very low quality. */ -static void local_timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) -{ - if (current->pid) - profile_tick(CPU_PROFILING, regs); - update_process_times(user_mode(regs)); -} +static struct clocksource counter = { + .name = "avr32_counter", + .rating = 50, + .read = read_cycle_count, + .mask = CLOCKSOURCE_MASK(32), + .flags = CLOCK_SOURCE_IS_CONTINUOUS, +}; -static irqreturn_t -timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) +static irqreturn_t timer_interrupt(int irq, void *dev_id) { - unsigned int count; + struct clock_event_device *evdev = dev_id; - /* ack timer interrupt and try to set next interrupt */ - count = avr32_hpt_read(); - avr32_timer_ack(); - - /* Update timerhi/timerlo for intra-jiffy calibration */ - timerhi += count < timerlo; /* Wrap around */ - timerlo = count; + if (unlikely(!(intc_get_pending(0) & 1))) + return IRQ_NONE; /* - * Call the generic timer interrupt handler + * Disable the interrupt until the clockevent subsystem + * reprograms it. */ - write_seqlock(&xtime_lock); - do_timer(regs); - write_sequnlock(&xtime_lock); - - /* - * In UP mode, we call local_timer_interrupt() to do profiling - * and process accounting. - * - * SMP is not supported yet. - */ - local_timer_interrupt(irq, dev_id, regs); + sysreg_write(COMPARE, 0); + evdev->event_handler(evdev); return IRQ_HANDLED; } static struct irqaction timer_irqaction = { .handler = timer_interrupt, - .flags = IRQF_DISABLED, - .name = "timer", + /* Oprofile uses the same irq as the timer, so allow it to be shared */ + .flags = IRQF_TIMER | IRQF_SHARED, + .name = "avr32_comparator", }; -void __init time_init(void) +static int comparator_next_event(unsigned long delta, + struct clock_event_device *evdev) { - unsigned long mult, shift, count_hz; - int ret; + unsigned long flags; - xtime.tv_sec = rtc_get_time(); - xtime.tv_nsec = 0; + raw_local_irq_save(flags); - set_normalized_timespec(&wall_to_monotonic, - -xtime.tv_sec, -xtime.tv_nsec); + /* The time to read COUNT then update COMPARE must be less + * than the min_delta_ns value for this clockevent source. + */ + sysreg_write(COMPARE, (sysreg_read(COUNT) + delta) ? : 1); - printk("Before time_init: count=%08lx, compare=%08lx\n", - (unsigned long)sysreg_read(COUNT), - (unsigned long)sysreg_read(COMPARE)); + raw_local_irq_restore(flags); - count_hz = clk_get_rate(boot_cpu_data.clk); - shift = clocksource_avr32.shift; - mult = clocksource_hz2mult(count_hz, shift); - clocksource_avr32.mult = mult; + return 0; +} - printk("Cycle counter: mult=%lu, shift=%lu\n", mult, shift); +static void comparator_mode(enum clock_event_mode mode, + struct clock_event_device *evdev) +{ + switch (mode) { + case CLOCK_EVT_MODE_ONESHOT: + pr_debug("%s: start\n", evdev->name); + /* FALLTHROUGH */ + case CLOCK_EVT_MODE_RESUME: + /* + * If we're using the COUNT and COMPARE registers we + * need to force idle poll. + */ + cpu_idle_poll_ctrl(true); + break; + case CLOCK_EVT_MODE_UNUSED: + case CLOCK_EVT_MODE_SHUTDOWN: + sysreg_write(COMPARE, 0); + pr_debug("%s: stop\n", evdev->name); + if (evdev->mode == CLOCK_EVT_MODE_ONESHOT || + evdev->mode == CLOCK_EVT_MODE_RESUME) { + /* + * Only disable idle poll if we have forced that + * in a previous call. + */ + cpu_idle_poll_ctrl(false); + } + break; + default: + BUG(); + } +} - { - u64 tmp; +static struct clock_event_device comparator = { + .name = "avr32_comparator", + .features = CLOCK_EVT_FEAT_ONESHOT, + .shift = 16, + .rating = 50, + .set_next_event = comparator_next_event, + .set_mode = comparator_mode, +}; - tmp = TICK_NSEC; - tmp <<= shift; - tmp += mult / 2; - do_div(tmp, mult); +void read_persistent_clock(struct timespec *ts) +{ + ts->tv_sec = mktime(2007, 1, 1, 0, 0, 0); + ts->tv_nsec = 0; +} - cycles_per_jiffy = tmp; - } +void __init time_init(void) +{ + unsigned long counter_hz; + int ret; - /* This sets up the high precision timer for the first interrupt. */ - avr32_hpt_init(avr32_hpt_read()); + /* figure rate for counter */ + counter_hz = clk_get_rate(boot_cpu_data.clk); + ret = clocksource_register_hz(&counter, counter_hz); + if (ret) + pr_debug("timer: could not register clocksource: %d\n", ret); - printk("After time_init: count=%08lx, compare=%08lx\n", - (unsigned long)sysreg_read(COUNT), - (unsigned long)sysreg_read(COMPARE)); + /* setup COMPARE clockevent */ + comparator.mult = div_sc(counter_hz, NSEC_PER_SEC, comparator.shift); + comparator.max_delta_ns = clockevent_delta2ns((u32)~0, &comparator); + comparator.min_delta_ns = clockevent_delta2ns(50, &comparator) + 1; + comparator.cpumask = cpumask_of(0); - ret = clocksource_register(&clocksource_avr32); - if (ret) - printk(KERN_ERR - "timer: could not register clocksource: %d\n", ret); + sysreg_write(COMPARE, 0); + timer_irqaction.dev_id = &comparator; ret = setup_irq(0, &timer_irqaction); if (ret) - printk("timer: could not request IRQ 0: %d\n", ret); -} - -static struct sysdev_class timer_class = { - set_kset_name("timer"), -}; + pr_debug("timer: could not request IRQ 0: %d\n", ret); + else { + clockevents_register_device(&comparator); -static struct sys_device timer_device = { - .id = 0, - .cls = &timer_class, -}; - -static int __init init_timer_sysfs(void) -{ - int err = sysdev_class_register(&timer_class); - if (!err) - err = sysdev_register(&timer_device); - return err; + pr_info("%s: irq 0, %lu.%03lu MHz\n", comparator.name, + ((counter_hz + 500) / 1000) / 1000, + ((counter_hz + 500) / 1000) % 1000); + } } - -device_initcall(init_timer_sysfs); diff --git a/arch/avr32/kernel/traps.c b/arch/avr32/kernel/traps.c index 7e803f4d7a1..682b2478691 100644 --- a/arch/avr32/kernel/traps.c +++ b/arch/avr32/kernel/traps.c @@ -5,256 +5,132 @@ * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. */ -#undef DEBUG -#include <linux/sched.h> + +#include <linux/bug.h> +#include <linux/hardirq.h> #include <linux/init.h> -#include <linux/module.h> #include <linux/kallsyms.h> +#include <linux/kdebug.h> +#include <linux/module.h> #include <linux/notifier.h> +#include <linux/sched.h> +#include <linux/uaccess.h> -#include <asm/traps.h> -#include <asm/sysreg.h> #include <asm/addrspace.h> -#include <asm/ocd.h> #include <asm/mmu_context.h> -#include <asm/uaccess.h> - -static void dump_mem(const char *str, unsigned long bottom, unsigned long top) -{ - unsigned long p; - int i; +#include <asm/ocd.h> +#include <asm/sysreg.h> +#include <asm/traps.h> - printk("%s(0x%08lx to 0x%08lx)\n", str, bottom, top); +static DEFINE_SPINLOCK(die_lock); - for (p = bottom & ~31; p < top; ) { - printk("%04lx: ", p & 0xffff); +void die(const char *str, struct pt_regs *regs, long err) +{ + static int die_counter; - for (i = 0; i < 8; i++, p += 4) { - unsigned int val; + console_verbose(); + spin_lock_irq(&die_lock); + bust_spinlocks(1); - if (p < bottom || p >= top) - printk(" "); - else { - if (__get_user(val, (unsigned int __user *)p)) { - printk("\n"); - goto out; - } - printk("%08x ", val); - } - } - printk("\n"); - } + printk(KERN_ALERT "Oops: %s, sig: %ld [#%d]\n", + str, err, ++die_counter); -out: - return; -} + printk(KERN_EMERG); +#ifdef CONFIG_PREEMPT + printk(KERN_CONT "PREEMPT "); +#endif #ifdef CONFIG_FRAME_POINTER -static inline void __show_trace(struct task_struct *tsk, unsigned long *sp, - struct pt_regs *regs) -{ - unsigned long __user *fp; - unsigned long __user *last_fp = NULL; - - if (regs) { - fp = (unsigned long __user *)regs->r7; - } else if (tsk == current) { - register unsigned long __user *real_fp __asm__("r7"); - fp = real_fp; + printk(KERN_CONT "FRAME_POINTER "); +#endif + if (current_cpu_data.features & AVR32_FEATURE_OCD) { + unsigned long did = ocd_read(DID); + printk(KERN_CONT "chip: 0x%03lx:0x%04lx rev %lu\n", + (did >> 1) & 0x7ff, + (did >> 12) & 0x7fff, + (did >> 28) & 0xf); } else { - fp = (unsigned long __user *)tsk->thread.cpu_context.r7; + printk(KERN_CONT "cpu: arch %u r%u / core %u r%u\n", + current_cpu_data.arch_type, + current_cpu_data.arch_revision, + current_cpu_data.cpu_type, + current_cpu_data.cpu_revision); } - /* - * Walk the stack until (a) we get an exception, (b) the frame - * pointer becomes zero, or (c) the frame pointer gets stuck - * at the same value. - */ - while (fp && fp != last_fp) { - unsigned long lr, new_fp = 0; + print_modules(); + show_regs_log_lvl(regs, KERN_EMERG); + show_stack_log_lvl(current, regs->sp, regs, KERN_EMERG); + bust_spinlocks(0); + add_taint(TAINT_DIE, LOCKDEP_NOW_UNRELIABLE); + spin_unlock_irq(&die_lock); - last_fp = fp; - if (__get_user(lr, fp)) - break; - if (fp && __get_user(new_fp, fp + 1)) - break; - fp = (unsigned long __user *)new_fp; + if (in_interrupt()) + panic("Fatal exception in interrupt"); - printk(" [<%08lx>] ", lr); - print_symbol("%s\n", lr); - } - printk("\n"); -} -#else -static inline void __show_trace(struct task_struct *tsk, unsigned long *sp, - struct pt_regs *regs) -{ - unsigned long addr; + if (panic_on_oops) + panic("Fatal exception"); - while (!kstack_end(sp)) { - addr = *sp++; - if (kernel_text_address(addr)) { - printk(" [<%08lx>] ", addr); - print_symbol("%s\n", addr); - } - } + do_exit(err); } -#endif -void show_trace(struct task_struct *tsk, unsigned long *sp, - struct pt_regs *regs) +void _exception(long signr, struct pt_regs *regs, int code, + unsigned long addr) { - if (regs && - (((regs->sr & MODE_MASK) == MODE_EXCEPTION) || - ((regs->sr & MODE_MASK) == MODE_USER))) - return; - - printk ("Call trace:"); -#ifdef CONFIG_KALLSYMS - printk("\n"); -#endif + siginfo_t info; - __show_trace(tsk, sp, regs); - printk("\n"); -} + if (!user_mode(regs)) { + const struct exception_table_entry *fixup; -void show_stack(struct task_struct *tsk, unsigned long *sp) -{ - unsigned long stack; - - if (!tsk) - tsk = current; - if (sp == 0) { - if (tsk == current) { - register unsigned long *real_sp __asm__("sp"); - sp = real_sp; - } else { - sp = (unsigned long *)tsk->thread.cpu_context.ksp; + /* Are we prepared to handle this kernel fault? */ + fixup = search_exception_tables(regs->pc); + if (fixup) { + regs->pc = fixup->fixup; + return; } + die("Unhandled exception in kernel mode", regs, signr); } - stack = (unsigned long)sp; - dump_mem("Stack: ", stack, - THREAD_SIZE + (unsigned long)tsk->thread_info); - show_trace(tsk, sp, NULL); -} - -void dump_stack(void) -{ - show_stack(NULL, NULL); + memset(&info, 0, sizeof(info)); + info.si_signo = signr; + info.si_code = code; + info.si_addr = (void __user *)addr; + force_sig_info(signr, &info, current); } -EXPORT_SYMBOL(dump_stack); - -ATOMIC_NOTIFIER_HEAD(avr32_die_chain); -int register_die_notifier(struct notifier_block *nb) +asmlinkage void do_nmi(unsigned long ecr, struct pt_regs *regs) { - pr_debug("register_die_notifier: %p\n", nb); - - return atomic_notifier_chain_register(&avr32_die_chain, nb); -} -EXPORT_SYMBOL(register_die_notifier); + int ret; -int unregister_die_notifier(struct notifier_block *nb) -{ - return atomic_notifier_chain_unregister(&avr32_die_chain, nb); -} -EXPORT_SYMBOL(unregister_die_notifier); + nmi_enter(); -static DEFINE_SPINLOCK(die_lock); - -void __die(const char *str, struct pt_regs *regs, unsigned long err, - const char *file, const char *func, unsigned long line) -{ - struct task_struct *tsk = current; - static int die_counter; - - console_verbose(); - spin_lock_irq(&die_lock); - bust_spinlocks(1); - - printk(KERN_ALERT "%s", str); - if (file && func) - printk(" in %s:%s, line %ld", file, func, line); - printk("[#%d]:\n", ++die_counter); - print_modules(); - show_regs(regs); - printk("Process %s (pid: %d, stack limit = 0x%p)\n", - tsk->comm, tsk->pid, tsk->thread_info + 1); - - if (!user_mode(regs) || in_interrupt()) { - dump_mem("Stack: ", regs->sp, - THREAD_SIZE + (unsigned long)tsk->thread_info); + ret = notify_die(DIE_NMI, "NMI", regs, 0, ecr, SIGINT); + switch (ret) { + case NOTIFY_OK: + case NOTIFY_STOP: + break; + case NOTIFY_BAD: + die("Fatal Non-Maskable Interrupt", regs, SIGINT); + default: + printk(KERN_ALERT "Got NMI, but nobody cared. Disabling...\n"); + nmi_disable(); + break; } - - bust_spinlocks(0); - spin_unlock_irq(&die_lock); - do_exit(SIGSEGV); -} - -void __die_if_kernel(const char *str, struct pt_regs *regs, unsigned long err, - const char *file, const char *func, unsigned long line) -{ - if (!user_mode(regs)) - __die(str, regs, err, file, func, line); -} - -asmlinkage void do_nmi(unsigned long ecr, struct pt_regs *regs) -{ -#ifdef CONFIG_SUBARCH_AVR32B - /* - * The exception entry always saves RSR_EX. For NMI, this is - * wrong; it should be RSR_NMI - */ - regs->sr = sysreg_read(RSR_NMI); -#endif - - printk("NMI taken!!!!\n"); - die("NMI", regs, ecr); - BUG(); + nmi_exit(); } asmlinkage void do_critical_exception(unsigned long ecr, struct pt_regs *regs) { - printk("Unable to handle critical exception %lu at pc = %08lx!\n", - ecr, regs->pc); - die("Oops", regs, ecr); - BUG(); + die("Critical exception", regs, SIGKILL); } asmlinkage void do_address_exception(unsigned long ecr, struct pt_regs *regs) { - siginfo_t info; - - die_if_kernel("Oops: Address exception in kernel mode", regs, ecr); - -#ifdef DEBUG - if (ecr == ECR_ADDR_ALIGN_X) - pr_debug("Instruction Address Exception at pc = %08lx\n", - regs->pc); - else if (ecr == ECR_ADDR_ALIGN_R) - pr_debug("Data Address Exception (Read) at pc = %08lx\n", - regs->pc); - else if (ecr == ECR_ADDR_ALIGN_W) - pr_debug("Data Address Exception (Write) at pc = %08lx\n", - regs->pc); - else - BUG(); - - show_regs(regs); -#endif - - info.si_signo = SIGBUS; - info.si_errno = 0; - info.si_code = BUS_ADRALN; - info.si_addr = (void __user *)regs->pc; - - force_sig_info(SIGBUS, &info, current); + _exception(SIGBUS, regs, BUS_ADRALN, regs->pc); } /* This way of handling undefined instructions is stolen from ARM */ static LIST_HEAD(undef_hook); -static spinlock_t undef_lock = SPIN_LOCK_UNLOCKED; +static DEFINE_SPINLOCK(undef_lock); void register_undef_hook(struct undef_hook *hook) { @@ -274,7 +150,8 @@ static int do_cop_absent(u32 insn) { int cop_nr; u32 cpucr; - if ( (insn & 0xfdf00000) == 0xf1900000 ) + + if ((insn & 0xfdf00000) == 0xf1900000) /* LDC0 */ cop_nr = 0; else @@ -286,136 +163,95 @@ static int do_cop_absent(u32 insn) sysreg_write(CPUCR, cpucr); cpucr = sysreg_read(CPUCR); - if ( !(cpucr & (1 << (24 + cop_nr))) ){ - printk("Coprocessor #%i not found!\n", cop_nr); - return -1; - } + if (!(cpucr & (1 << (24 + cop_nr)))) + return -ENODEV; return 0; } #ifdef CONFIG_BUG -#ifdef CONFIG_DEBUG_BUGVERBOSE -static inline void do_bug_verbose(struct pt_regs *regs, u32 insn) -{ - char *file; - u16 line; - char c; - - if (__get_user(line, (u16 __user *)(regs->pc + 2))) - return; - if (__get_user(file, (char * __user *)(regs->pc + 4)) - || (unsigned long)file < PAGE_OFFSET - || __get_user(c, file)) - file = "<bad filename>"; - - printk(KERN_ALERT "kernel BUG at %s:%d!\n", file, line); -} -#else -static inline void do_bug_verbose(struct pt_regs *regs, u32 insn) +int is_valid_bugaddr(unsigned long pc) { + unsigned short opcode; + + if (pc < PAGE_OFFSET) + return 0; + if (probe_kernel_address((u16 *)pc, opcode)) + return 0; + return opcode == AVR32_BUG_OPCODE; } #endif -#endif asmlinkage void do_illegal_opcode(unsigned long ecr, struct pt_regs *regs) { u32 insn; struct undef_hook *hook; - siginfo_t info; void __user *pc; + long code; - if (!user_mode(regs)) - goto kernel_trap; +#ifdef CONFIG_BUG + if (!user_mode(regs) && (ecr == ECR_ILLEGAL_OPCODE)) { + enum bug_trap_type type; + + type = report_bug(regs->pc, regs); + switch (type) { + case BUG_TRAP_TYPE_NONE: + break; + case BUG_TRAP_TYPE_WARN: + regs->pc += 2; + return; + case BUG_TRAP_TYPE_BUG: + die("Kernel BUG", regs, SIGKILL); + } + } +#endif local_irq_enable(); - pc = (void __user *)instruction_pointer(regs); - if (__get_user(insn, (u32 __user *)pc)) - goto invalid_area; + if (user_mode(regs)) { + pc = (void __user *)instruction_pointer(regs); + if (get_user(insn, (u32 __user *)pc)) + goto invalid_area; - if (ecr == ECR_COPROC_ABSENT) { - if (do_cop_absent(insn) == 0) + if (ecr == ECR_COPROC_ABSENT && !do_cop_absent(insn)) return; - } - spin_lock_irq(&undef_lock); - list_for_each_entry(hook, &undef_hook, node) { - if ((insn & hook->insn_mask) == hook->insn_val) { - if (hook->fn(regs, insn) == 0) { - spin_unlock_irq(&undef_lock); - return; + spin_lock_irq(&undef_lock); + list_for_each_entry(hook, &undef_hook, node) { + if ((insn & hook->insn_mask) == hook->insn_val) { + if (hook->fn(regs, insn) == 0) { + spin_unlock_irq(&undef_lock); + return; + } } } + spin_unlock_irq(&undef_lock); } - spin_unlock_irq(&undef_lock); - -invalid_area: - -#ifdef DEBUG - printk("Illegal instruction at pc = %08lx\n", regs->pc); - if (regs->pc < TASK_SIZE) { - unsigned long ptbr, pgd, pte, *p; - - ptbr = sysreg_read(PTBR); - p = (unsigned long *)ptbr; - pgd = p[regs->pc >> 22]; - p = (unsigned long *)((pgd & 0x1ffff000) | 0x80000000); - pte = p[(regs->pc >> 12) & 0x3ff]; - printk("page table: 0x%08lx -> 0x%08lx -> 0x%08lx\n", ptbr, pgd, pte); - } -#endif - info.si_signo = SIGILL; - info.si_errno = 0; - info.si_addr = (void __user *)regs->pc; switch (ecr) { - case ECR_ILLEGAL_OPCODE: - case ECR_UNIMPL_INSTRUCTION: - info.si_code = ILL_ILLOPC; - break; case ECR_PRIVILEGE_VIOLATION: - info.si_code = ILL_PRVOPC; + code = ILL_PRVOPC; break; case ECR_COPROC_ABSENT: - info.si_code = ILL_COPROC; + code = ILL_COPROC; break; default: - BUG(); + code = ILL_ILLOPC; + break; } - force_sig_info(SIGILL, &info, current); + _exception(SIGILL, regs, code, regs->pc); return; -kernel_trap: -#ifdef CONFIG_BUG - if (__kernel_text_address(instruction_pointer(regs))) { - insn = *(u16 *)instruction_pointer(regs); - if (insn == AVR32_BUG_OPCODE) { - do_bug_verbose(regs, insn); - die("Kernel BUG", regs, 0); - return; - } - } -#endif - - die("Oops: Illegal instruction in kernel code", regs, ecr); +invalid_area: + _exception(SIGSEGV, regs, SEGV_MAPERR, regs->pc); } asmlinkage void do_fpe(unsigned long ecr, struct pt_regs *regs) { - siginfo_t info; - - printk("Floating-point exception at pc = %08lx\n", regs->pc); - - /* We have no FPU... */ - info.si_signo = SIGILL; - info.si_errno = 0; - info.si_addr = (void __user *)regs->pc; - info.si_code = ILL_COPROC; - - force_sig_info(SIGILL, &info, current); + /* We have no FPU yet */ + _exception(SIGILL, regs, ILL_COPROC, regs->pc); } diff --git a/arch/avr32/kernel/vmlinux.lds.S b/arch/avr32/kernel/vmlinux.lds.S new file mode 100644 index 00000000000..a4589176bed --- /dev/null +++ b/arch/avr32/kernel/vmlinux.lds.S @@ -0,0 +1,88 @@ +/* + * AVR32 linker script for the Linux kernel + * + * Copyright (C) 2004-2006 Atmel Corporation + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ +#define LOAD_OFFSET 0x00000000 +#include <asm-generic/vmlinux.lds.h> +#include <asm/cache.h> +#include <asm/thread_info.h> + +OUTPUT_FORMAT("elf32-avr32", "elf32-avr32", "elf32-avr32") +OUTPUT_ARCH(avr32) +ENTRY(_start) + +/* Big endian */ +jiffies = jiffies_64 + 4; + +SECTIONS +{ + . = CONFIG_ENTRY_ADDRESS; + .init : AT(ADDR(.init) - LOAD_OFFSET) { + _text = .; + __init_begin = .; + _sinittext = .; + *(.text.reset) + INIT_TEXT + /* + * .exit.text is discarded at runtime, not + * link time, to deal with references from + * __bug_table + */ + EXIT_TEXT + _einittext = .; + . = ALIGN(4); + __tagtable_begin = .; + *(.taglist.init) + __tagtable_end = .; + } + INIT_DATA_SECTION(16) + . = ALIGN(PAGE_SIZE); + __init_end = .; + + .text : AT(ADDR(.text) - LOAD_OFFSET) { + _evba = .; + _stext = .; + *(.ex.text) + *(.irq.text) + KPROBES_TEXT + TEXT_TEXT + SCHED_TEXT + LOCK_TEXT + *(.fixup) + *(.gnu.warning) + _etext = .; + } = 0xd703d703 + + EXCEPTION_TABLE(4) + RODATA + + .data : AT(ADDR(.data) - LOAD_OFFSET) { + _data = .; + _sdata = .; + + INIT_TASK_DATA(THREAD_SIZE) + PAGE_ALIGNED_DATA(PAGE_SIZE); + CACHELINE_ALIGNED_DATA(L1_CACHE_BYTES) + *(.data.rel*) + DATA_DATA + CONSTRUCTORS + + _edata = .; + } + + BSS_SECTION(0, 8, 8) + _end = .; + + DWARF_DEBUG + + /* When something in the kernel is NOT compiled as a module, the module + * cleanup code and data are put into these segments. Both can then be + * thrown away, as cleanup code is never called unless it's a module. + */ + DISCARDS +} diff --git a/arch/avr32/kernel/vmlinux.lds.c b/arch/avr32/kernel/vmlinux.lds.c deleted file mode 100644 index cdd627c6b7d..00000000000 --- a/arch/avr32/kernel/vmlinux.lds.c +++ /dev/null @@ -1,139 +0,0 @@ -/* - * AVR32 linker script for the Linux kernel - * - * Copyright (C) 2004-2006 Atmel Corporation - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License version 2 as - * published by the Free Software Foundation. - */ -#define LOAD_OFFSET 0x00000000 -#include <asm-generic/vmlinux.lds.h> - -OUTPUT_FORMAT("elf32-avr32", "elf32-avr32", "elf32-avr32") -OUTPUT_ARCH(avr32) -ENTRY(_start) - -/* Big endian */ -jiffies = jiffies_64 + 4; - -SECTIONS -{ - . = CONFIG_ENTRY_ADDRESS; - .init : AT(ADDR(.init) - LOAD_OFFSET) { - _stext = .; - __init_begin = .; - _sinittext = .; - *(.text.reset) - *(.init.text) - _einittext = .; - . = ALIGN(4); - __tagtable_begin = .; - *(.taglist) - __tagtable_end = .; - *(.init.data) - . = ALIGN(16); - __setup_start = .; - *(.init.setup) - __setup_end = .; - . = ALIGN(4); - __initcall_start = .; - *(.initcall1.init) - *(.initcall2.init) - *(.initcall3.init) - *(.initcall4.init) - *(.initcall5.init) - *(.initcall6.init) - *(.initcall7.init) - __initcall_end = .; - __con_initcall_start = .; - *(.con_initcall.init) - __con_initcall_end = .; - __security_initcall_start = .; - *(.security_initcall.init) - __security_initcall_end = .; - . = ALIGN(32); - __initramfs_start = .; - *(.init.ramfs) - __initramfs_end = .; - . = ALIGN(4096); - __init_end = .; - } - - . = ALIGN(8192); - .text : AT(ADDR(.text) - LOAD_OFFSET) { - _evba = .; - _text = .; - *(.ex.text) - . = 0x50; - *(.tlbx.ex.text) - . = 0x60; - *(.tlbr.ex.text) - . = 0x70; - *(.tlbw.ex.text) - . = 0x100; - *(.scall.text) - *(.irq.text) - *(.text) - SCHED_TEXT - LOCK_TEXT - KPROBES_TEXT - *(.fixup) - *(.gnu.warning) - _etext = .; - } = 0xd703d703 - - . = ALIGN(4); - __ex_table : AT(ADDR(__ex_table) - LOAD_OFFSET) { - __start___ex_table = .; - *(__ex_table) - __stop___ex_table = .; - } - - RODATA - - . = ALIGN(8192); - - .data : AT(ADDR(.data) - LOAD_OFFSET) { - _data = .; - _sdata = .; - /* - * First, the init task union, aligned to an 8K boundary. - */ - *(.data.init_task) - - /* Then, the cacheline aligned data */ - . = ALIGN(32); - *(.data.cacheline_aligned) - - /* And the rest... */ - *(.data.rel*) - *(.data) - CONSTRUCTORS - - _edata = .; - } - - - . = ALIGN(8); - .bss : AT(ADDR(.bss) - LOAD_OFFSET) { - __bss_start = .; - *(.bss) - *(COMMON) - . = ALIGN(8); - __bss_stop = .; - _end = .; - } - - /* When something in the kernel is NOT compiled as a module, the module - * cleanup code and data are put into these segments. Both can then be - * thrown away, as cleanup code is never called unless it's a module. - */ - /DISCARD/ : { - *(.exit.text) - *(.exit.data) - *(.exitcall.exit) - } - - DWARF_DEBUG -} |