From 416821d3d68164909b2cbcf398e4ba0797f5f8a2 Mon Sep 17 00:00:00 2001 From: Helge Deller Date: Fri, 10 May 2013 21:24:01 +0000 Subject: parisc: implement irq stacks - part 2 (v2) This patch fixes few build issues which were introduced with the last irq stack patch, e.g. the combination of stack overflow check and irq stack. Furthermore we now do proper locking and change the irq bh handler to use the irq stack as well. In /proc/interrupts one now can monitor how huge the irq stack has grown and how often it was preferred over the kernel stack. IRQ stacks are now enabled by default just to make sure that we not overflow the kernel stack by accident. Signed-off-by: Helge Deller --- arch/parisc/Kconfig | 2 +- arch/parisc/include/asm/hardirq.h | 9 ++++ arch/parisc/include/asm/processor.h | 3 ++ arch/parisc/kernel/irq.c | 101 +++++++++++++++++++++++++++++++----- arch/parisc/mm/init.c | 4 +- 5 files changed, 102 insertions(+), 17 deletions(-) (limited to 'arch') diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig index cad060f288cf..6507dabdd5dd 100644 --- a/arch/parisc/Kconfig +++ b/arch/parisc/Kconfig @@ -245,7 +245,7 @@ config SMP config IRQSTACKS bool "Use separate kernel stacks when processing interrupts" - default n + default y help If you say Y here the kernel will use separate kernel stacks for handling hard and soft interrupts. This can help avoid diff --git a/arch/parisc/include/asm/hardirq.h b/arch/parisc/include/asm/hardirq.h index 12373c4dabab..c19f7138ba48 100644 --- a/arch/parisc/include/asm/hardirq.h +++ b/arch/parisc/include/asm/hardirq.h @@ -11,10 +11,18 @@ #include #include +#ifdef CONFIG_IRQSTACKS +#define __ARCH_HAS_DO_SOFTIRQ +#endif + typedef struct { unsigned int __softirq_pending; #ifdef CONFIG_DEBUG_STACKOVERFLOW unsigned int kernel_stack_usage; +#ifdef CONFIG_IRQSTACKS + unsigned int irq_stack_usage; + unsigned int irq_stack_counter; +#endif #endif #ifdef CONFIG_SMP unsigned int irq_resched_count; @@ -28,6 +36,7 @@ DECLARE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat); #define __ARCH_IRQ_STAT #define __IRQ_STAT(cpu, member) (irq_stat[cpu].member) #define inc_irq_stat(member) this_cpu_inc(irq_stat.member) +#define __inc_irq_stat(member) __this_cpu_inc(irq_stat.member) #define local_softirq_pending() this_cpu_read(irq_stat.__softirq_pending) #define __ARCH_SET_SOFTIRQ_PENDING diff --git a/arch/parisc/include/asm/processor.h b/arch/parisc/include/asm/processor.h index 064015547d1e..cfbc43929cf6 100644 --- a/arch/parisc/include/asm/processor.h +++ b/arch/parisc/include/asm/processor.h @@ -63,10 +63,13 @@ */ #ifdef __KERNEL__ +#include + #define IRQ_STACK_SIZE (4096 << 2) /* 16k irq stack size */ union irq_stack_union { unsigned long stack[IRQ_STACK_SIZE/sizeof(unsigned long)]; + raw_spinlock_t lock; }; DECLARE_PER_CPU(union irq_stack_union, irq_stack_union); diff --git a/arch/parisc/kernel/irq.c b/arch/parisc/kernel/irq.c index e255db0bb761..55237a70e197 100644 --- a/arch/parisc/kernel/irq.c +++ b/arch/parisc/kernel/irq.c @@ -166,22 +166,32 @@ int arch_show_interrupts(struct seq_file *p, int prec) seq_printf(p, "%*s: ", prec, "STK"); for_each_online_cpu(j) seq_printf(p, "%10u ", irq_stats(j)->kernel_stack_usage); - seq_printf(p, " Kernel stack usage\n"); + seq_puts(p, " Kernel stack usage\n"); +# ifdef CONFIG_IRQSTACKS + seq_printf(p, "%*s: ", prec, "IST"); + for_each_online_cpu(j) + seq_printf(p, "%10u ", irq_stats(j)->irq_stack_usage); + seq_puts(p, " Interrupt stack usage\n"); + seq_printf(p, "%*s: ", prec, "ISC"); + for_each_online_cpu(j) + seq_printf(p, "%10u ", irq_stats(j)->irq_stack_counter); + seq_puts(p, " Interrupt stack usage counter\n"); +# endif #endif #ifdef CONFIG_SMP seq_printf(p, "%*s: ", prec, "RES"); for_each_online_cpu(j) seq_printf(p, "%10u ", irq_stats(j)->irq_resched_count); - seq_printf(p, " Rescheduling interrupts\n"); + seq_puts(p, " Rescheduling interrupts\n"); seq_printf(p, "%*s: ", prec, "CAL"); for_each_online_cpu(j) seq_printf(p, "%10u ", irq_stats(j)->irq_call_count); - seq_printf(p, " Function call interrupts\n"); + seq_puts(p, " Function call interrupts\n"); #endif seq_printf(p, "%*s: ", prec, "TLB"); for_each_online_cpu(j) seq_printf(p, "%10u ", irq_stats(j)->irq_tlb_count); - seq_printf(p, " TLB shootdowns\n"); + seq_puts(p, " TLB shootdowns\n"); return 0; } @@ -378,6 +388,7 @@ static inline void stack_overflow_check(struct pt_regs *regs) unsigned long sp = regs->gr[30]; unsigned long stack_usage; unsigned int *last_usage; + int cpu = smp_processor_id(); /* if sr7 != 0, we interrupted a userspace process which we do not want * to check for stack overflow. We will only check the kernel stack. */ @@ -386,7 +397,31 @@ static inline void stack_overflow_check(struct pt_regs *regs) /* calculate kernel stack usage */ stack_usage = sp - stack_start; - last_usage = &per_cpu(irq_stat.kernel_stack_usage, smp_processor_id()); +#ifdef CONFIG_IRQSTACKS + if (likely(stack_usage <= THREAD_SIZE)) + goto check_kernel_stack; /* found kernel stack */ + + /* check irq stack usage */ + stack_start = (unsigned long) &per_cpu(irq_stack_union, cpu).stack; + stack_usage = sp - stack_start; + + last_usage = &per_cpu(irq_stat.irq_stack_usage, cpu); + if (unlikely(stack_usage > *last_usage)) + *last_usage = stack_usage; + + if (likely(stack_usage < (IRQ_STACK_SIZE - STACK_MARGIN))) + return; + + pr_emerg("stackcheck: %s will most likely overflow irq stack " + "(sp:%lx, stk bottom-top:%lx-%lx)\n", + current->comm, sp, stack_start, stack_start + IRQ_STACK_SIZE); + goto panic_check; + +check_kernel_stack: +#endif + + /* check kernel stack usage */ + last_usage = &per_cpu(irq_stat.kernel_stack_usage, cpu); if (unlikely(stack_usage > *last_usage)) *last_usage = stack_usage; @@ -398,31 +433,69 @@ static inline void stack_overflow_check(struct pt_regs *regs) "(sp:%lx, stk bottom-top:%lx-%lx)\n", current->comm, sp, stack_start, stack_start + THREAD_SIZE); +#ifdef CONFIG_IRQSTACKS +panic_check: +#endif if (sysctl_panic_on_stackoverflow) panic("low stack detected by irq handler - check messages\n"); #endif } #ifdef CONFIG_IRQSTACKS -DEFINE_PER_CPU(union irq_stack_union, irq_stack_union); +DEFINE_PER_CPU(union irq_stack_union, irq_stack_union) = { + .lock = __RAW_SPIN_LOCK_UNLOCKED((irq_stack_union).lock) + }; static void execute_on_irq_stack(void *func, unsigned long param1) { - unsigned long *irq_stack_start; + union irq_stack_union *union_ptr; unsigned long irq_stack; - int cpu = smp_processor_id(); + raw_spinlock_t *irq_stack_in_use; - irq_stack_start = &per_cpu(irq_stack_union, cpu).stack[0]; - irq_stack = (unsigned long) irq_stack_start; - irq_stack = ALIGN(irq_stack, 16); /* align for stack frame usage */ + union_ptr = &per_cpu(irq_stack_union, smp_processor_id()); + irq_stack = (unsigned long) &union_ptr->stack; + irq_stack = ALIGN(irq_stack + sizeof(irq_stack_union.lock), + 64); /* align for stack frame usage */ - BUG_ON(*irq_stack_start); /* report bug if we were called recursive. */ - *irq_stack_start = 1; + /* We may be called recursive. If we are already using the irq stack, + * just continue to use it. Use spinlocks to serialize + * the irq stack usage. + */ + irq_stack_in_use = &union_ptr->lock; + if (!raw_spin_trylock(irq_stack_in_use)) { + void (*direct_call)(unsigned long p1) = func; + + /* We are using the IRQ stack already. + * Do direct call on current stack. */ + direct_call(param1); + return; + } /* This is where we switch to the IRQ stack. */ call_on_stack(param1, func, irq_stack); - *irq_stack_start = 0; + __inc_irq_stat(irq_stack_counter); + + /* free up irq stack usage. */ + do_raw_spin_unlock(irq_stack_in_use); +} + +asmlinkage void do_softirq(void) +{ + __u32 pending; + unsigned long flags; + + if (in_interrupt()) + return; + + local_irq_save(flags); + + pending = local_softirq_pending(); + + if (pending) + execute_on_irq_stack(__do_softirq, 0); + + local_irq_restore(flags); } #endif /* CONFIG_IRQSTACKS */ diff --git a/arch/parisc/mm/init.c b/arch/parisc/mm/init.c index ce939ac8622b..1c965642068b 100644 --- a/arch/parisc/mm/init.c +++ b/arch/parisc/mm/init.c @@ -1069,7 +1069,7 @@ void flush_tlb_all(void) { int do_recycle; - inc_irq_stat(irq_tlb_count); + __inc_irq_stat(irq_tlb_count); do_recycle = 0; spin_lock(&sid_lock); if (dirty_space_ids > RECYCLE_THRESHOLD) { @@ -1090,7 +1090,7 @@ void flush_tlb_all(void) #else void flush_tlb_all(void) { - inc_irq_stat(irq_tlb_count); + __inc_irq_stat(irq_tlb_count); spin_lock(&sid_lock); flush_tlb_all_local(NULL); recycle_sids(); -- cgit v1.2.3-58-ga151 From f0a18819e261afc5fdbd8c5c6f9943123c5461ba Mon Sep 17 00:00:00 2001 From: John David Anglin Date: Fri, 10 May 2013 23:21:38 +0000 Subject: parisc: fix SMP races when updating PTE and TLB entries in entry.S Currently, race conditions exist in the handling of TLB interruptions in entry.S. In particular, dirty bit updates can be lost if an accessed interruption occurs just after the dirty bit interruption on a different cpu. Lost dirty bit updates result in user pages not being flushed and general system instability. This change adds lock and unlock macros to synchronize all PTE and TLB updates done in entry.S. As a result, userspace stability is significantly improved. Signed-off-by: John David Anglin Signed-off-by: Helge Deller --- arch/parisc/kernel/entry.S | 155 ++++++++++++++++++++++++--------------------- 1 file changed, 83 insertions(+), 72 deletions(-) (limited to 'arch') diff --git a/arch/parisc/kernel/entry.S b/arch/parisc/kernel/entry.S index 4bb96ad9b0b1..ae27cb6ce19a 100644 --- a/arch/parisc/kernel/entry.S +++ b/arch/parisc/kernel/entry.S @@ -452,9 +452,41 @@ L2_ptep \pgd,\pte,\index,\va,\fault .endm + /* Acquire pa_dbit_lock lock. */ + .macro dbit_lock spc,tmp,tmp1 +#ifdef CONFIG_SMP + cmpib,COND(=),n 0,\spc,2f + load32 PA(pa_dbit_lock),\tmp +1: LDCW 0(\tmp),\tmp1 + cmpib,COND(=) 0,\tmp1,1b + nop +2: +#endif + .endm + + /* Release pa_dbit_lock lock without reloading lock address. */ + .macro dbit_unlock0 spc,tmp +#ifdef CONFIG_SMP + or,COND(=) %r0,\spc,%r0 + stw \spc,0(\tmp) +#endif + .endm + + /* Release pa_dbit_lock lock. */ + .macro dbit_unlock1 spc,tmp +#ifdef CONFIG_SMP + load32 PA(pa_dbit_lock),\tmp + dbit_unlock0 \spc,\tmp +#endif + .endm + /* Set the _PAGE_ACCESSED bit of the PTE. Be clever and * don't needlessly dirty the cache line if it was already set */ - .macro update_ptep ptep,pte,tmp,tmp1 + .macro update_ptep spc,ptep,pte,tmp,tmp1 +#ifdef CONFIG_SMP + or,COND(=) %r0,\spc,%r0 + LDREG 0(\ptep),\pte +#endif ldi _PAGE_ACCESSED,\tmp1 or \tmp1,\pte,\tmp and,COND(<>) \tmp1,\pte,%r0 @@ -463,7 +495,11 @@ /* Set the dirty bit (and accessed bit). No need to be * clever, this is only used from the dirty fault */ - .macro update_dirty ptep,pte,tmp + .macro update_dirty spc,ptep,pte,tmp +#ifdef CONFIG_SMP + or,COND(=) %r0,\spc,%r0 + LDREG 0(\ptep),\pte +#endif ldi _PAGE_ACCESSED|_PAGE_DIRTY,\tmp or \tmp,\pte,\pte STREG \pte,0(\ptep) @@ -1111,11 +1147,13 @@ dtlb_miss_20w: L3_ptep ptp,pte,t0,va,dtlb_check_alias_20w - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot idtlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1135,11 +1173,13 @@ nadtlb_miss_20w: L3_ptep ptp,pte,t0,va,nadtlb_check_alias_20w - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot idtlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1161,7 +1201,8 @@ dtlb_miss_11: L2_ptep ptp,pte,t0,va,dtlb_check_alias_11 - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb_11 spc,pte,prot @@ -1172,6 +1213,7 @@ dtlb_miss_11: idtlbp prot,(%sr1,va) mtsp t0, %sr1 /* Restore sr1 */ + dbit_unlock1 spc,t0 rfir nop @@ -1192,7 +1234,8 @@ nadtlb_miss_11: L2_ptep ptp,pte,t0,va,nadtlb_check_alias_11 - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb_11 spc,pte,prot @@ -1204,6 +1247,7 @@ nadtlb_miss_11: idtlbp prot,(%sr1,va) mtsp t0, %sr1 /* Restore sr1 */ + dbit_unlock1 spc,t0 rfir nop @@ -1224,13 +1268,15 @@ dtlb_miss_20: L2_ptep ptp,pte,t0,va,dtlb_check_alias_20 - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot f_extend pte,t0 idtlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1250,13 +1296,15 @@ nadtlb_miss_20: L2_ptep ptp,pte,t0,va,nadtlb_check_alias_20 - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot f_extend pte,t0 idtlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1357,11 +1405,13 @@ itlb_miss_20w: L3_ptep ptp,pte,t0,va,itlb_fault - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot iitlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1379,11 +1429,13 @@ naitlb_miss_20w: L3_ptep ptp,pte,t0,va,naitlb_check_alias_20w - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot iitlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1405,7 +1457,8 @@ itlb_miss_11: L2_ptep ptp,pte,t0,va,itlb_fault - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb_11 spc,pte,prot @@ -1416,6 +1469,7 @@ itlb_miss_11: iitlbp prot,(%sr1,va) mtsp t0, %sr1 /* Restore sr1 */ + dbit_unlock1 spc,t0 rfir nop @@ -1427,7 +1481,8 @@ naitlb_miss_11: L2_ptep ptp,pte,t0,va,naitlb_check_alias_11 - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb_11 spc,pte,prot @@ -1438,6 +1493,7 @@ naitlb_miss_11: iitlbp prot,(%sr1,va) mtsp t0, %sr1 /* Restore sr1 */ + dbit_unlock1 spc,t0 rfir nop @@ -1459,13 +1515,15 @@ itlb_miss_20: L2_ptep ptp,pte,t0,va,itlb_fault - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot f_extend pte,t0 iitlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1477,13 +1535,15 @@ naitlb_miss_20: L2_ptep ptp,pte,t0,va,naitlb_check_alias_20 - update_ptep ptp,pte,t0,t1 + dbit_lock spc,t0,t1 + update_ptep spc,ptp,pte,t0,t1 make_insert_tlb spc,pte,prot f_extend pte,t0 iitlbt pte,prot + dbit_unlock1 spc,t0 rfir nop @@ -1507,29 +1567,13 @@ dbit_trap_20w: L3_ptep ptp,pte,t0,va,dbit_fault -#ifdef CONFIG_SMP - cmpib,COND(=),n 0,spc,dbit_nolock_20w - load32 PA(pa_dbit_lock),t0 - -dbit_spin_20w: - LDCW 0(t0),t1 - cmpib,COND(=) 0,t1,dbit_spin_20w - nop - -dbit_nolock_20w: -#endif - update_dirty ptp,pte,t1 + dbit_lock spc,t0,t1 + update_dirty spc,ptp,pte,t1 make_insert_tlb spc,pte,prot idtlbt pte,prot -#ifdef CONFIG_SMP - cmpib,COND(=),n 0,spc,dbit_nounlock_20w - ldi 1,t1 - stw t1,0(t0) - -dbit_nounlock_20w: -#endif + dbit_unlock0 spc,t0 rfir nop @@ -1543,18 +1587,8 @@ dbit_trap_11: L2_ptep ptp,pte,t0,va,dbit_fault -#ifdef CONFIG_SMP - cmpib,COND(=),n 0,spc,dbit_nolock_11 - load32 PA(pa_dbit_lock),t0 - -dbit_spin_11: - LDCW 0(t0),t1 - cmpib,= 0,t1,dbit_spin_11 - nop - -dbit_nolock_11: -#endif - update_dirty ptp,pte,t1 + dbit_lock spc,t0,t1 + update_dirty spc,ptp,pte,t1 make_insert_tlb_11 spc,pte,prot @@ -1565,13 +1599,7 @@ dbit_nolock_11: idtlbp prot,(%sr1,va) mtsp t1, %sr1 /* Restore sr1 */ -#ifdef CONFIG_SMP - cmpib,COND(=),n 0,spc,dbit_nounlock_11 - ldi 1,t1 - stw t1,0(t0) - -dbit_nounlock_11: -#endif + dbit_unlock0 spc,t0 rfir nop @@ -1583,32 +1611,15 @@ dbit_trap_20: L2_ptep ptp,pte,t0,va,dbit_fault -#ifdef CONFIG_SMP - cmpib,COND(=),n 0,spc,dbit_nolock_20 - load32 PA(pa_dbit_lock),t0 - -dbit_spin_20: - LDCW 0(t0),t1 - cmpib,= 0,t1,dbit_spin_20 - nop - -dbit_nolock_20: -#endif - update_dirty ptp,pte,t1 + dbit_lock spc,t0,t1 + update_dirty spc,ptp,pte,t1 make_insert_tlb spc,pte,prot f_extend pte,t1 idtlbt pte,prot - -#ifdef CONFIG_SMP - cmpib,COND(=),n 0,spc,dbit_nounlock_20 - ldi 1,t1 - stw t1,0(t0) - -dbit_nounlock_20: -#endif + dbit_unlock0 spc,t0 rfir nop -- cgit v1.2.3-58-ga151 From 6880b0150a7c25fd75c5ece80abc49ebf53c38c1 Mon Sep 17 00:00:00 2001 From: Helge Deller Date: Sat, 11 May 2013 19:04:09 +0000 Subject: parisc: make default cross compiler search more robust (v3) People/distros vary how they prefix the toolchain name for 64bit builds. Rather than enforce one convention over another, add a for loop which does a search for all the general prefixes. For 64bit builds, we now search for (in order): hppa64-unknown-linux-gnu hppa64-linux-gnu hppa64-linux For 32bit builds, we look for: hppa-unknown-linux-gnu hppa-linux-gnu hppa-linux hppa2.0-unknown-linux-gnu hppa2.0-linux-gnu hppa2.0-linux hppa1.1-unknown-linux-gnu hppa1.1-linux-gnu hppa1.1-linux This patch was initiated by Mike Frysinger, with feedback from Jeroen Roovers, John David Anglin and Helge Deller. Signed-off-by: Mike Frysinger Signed-off-by: Jeroen Roovers Signed-off-by: John David Anglin Signed-off-by: Helge Deller --- arch/parisc/Makefile | 21 +++++++++------------ 1 file changed, 9 insertions(+), 12 deletions(-) (limited to 'arch') diff --git a/arch/parisc/Makefile b/arch/parisc/Makefile index 2f967cc6649e..197690068f88 100644 --- a/arch/parisc/Makefile +++ b/arch/parisc/Makefile @@ -23,24 +23,21 @@ NM = sh $(srctree)/arch/parisc/nm CHECKFLAGS += -D__hppa__=1 LIBGCC = $(shell $(CC) $(KBUILD_CFLAGS) -print-libgcc-file-name) -MACHINE := $(shell uname -m) -NATIVE := $(if $(filter parisc%,$(MACHINE)),1,0) - ifdef CONFIG_64BIT UTS_MACHINE := parisc64 CHECKFLAGS += -D__LP64__=1 -m64 -WIDTH := 64 +CC_ARCHES = hppa64 else # 32-bit -WIDTH := +CC_ARCHES = hppa hppa2.0 hppa1.1 endif -# attempt to help out folks who are cross-compiling -ifeq ($(NATIVE),1) -CROSS_COMPILE := hppa$(WIDTH)-linux- -else - ifeq ($(CROSS_COMPILE),) - CROSS_COMPILE := hppa$(WIDTH)-linux-gnu- - endif +ifneq ($(SUBARCH),$(UTS_MACHINE)) + ifeq ($(CROSS_COMPILE),) + CC_SUFFIXES = linux linux-gnu unknown-linux-gnu + CROSS_COMPILE := $(call cc-cross-prefix, \ + $(foreach a,$(CC_ARCHES), \ + $(foreach s,$(CC_SUFFIXES),$(a)-$(s)-))) + endif endif OBJCOPY_FLAGS =-O binary -R .note -R .comment -S -- cgit v1.2.3-58-ga151