Blackfin arch: SMP supporting patchset: Blackfin kernel and memory ma…

…nagement code

Blackfin dual core BF561 processor can support SMP like features.
https://docs.blackfin.uclinux.org/doku.php?id=linux-kernel:smp-like

In this patch, we provide SMP extend to Blackfin kernel and memory management code

Singed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>

arch/blackfin/kernel/asm-offsets.c

@@ -56,6 +56,9 @@ int main(void)
 	/* offsets into the thread struct */
 	DEFINE(THREAD_KSP, offsetof(struct thread_struct, ksp));
 	DEFINE(THREAD_USP, offsetof(struct thread_struct, usp));
+	DEFINE(THREAD_SR, offsetof(struct thread_struct, seqstat));
+	DEFINE(PT_SR, offsetof(struct thread_struct, seqstat));
+	DEFINE(THREAD_ESP0, offsetof(struct thread_struct, esp0));
 	DEFINE(THREAD_PC, offsetof(struct thread_struct, pc));
 	DEFINE(KERNEL_STACK_SIZE, THREAD_SIZE);
 
@@ -128,5 +131,31 @@ int main(void)
 	DEFINE(SIGSEGV, SIGSEGV);
 	DEFINE(SIGTRAP, SIGTRAP);
 
+	/* PDA management (in L1 scratchpad) */
+	DEFINE(PDA_SYSCFG, offsetof(struct blackfin_pda, syscfg));
+#ifdef CONFIG_SMP
+	DEFINE(PDA_IRQFLAGS, offsetof(struct blackfin_pda, imask));
+#endif
+	DEFINE(PDA_IPDT, offsetof(struct blackfin_pda, ipdt));
+	DEFINE(PDA_IPDT_SWAPCOUNT, offsetof(struct blackfin_pda, ipdt_swapcount));
+	DEFINE(PDA_DPDT, offsetof(struct blackfin_pda, dpdt));
+	DEFINE(PDA_DPDT_SWAPCOUNT, offsetof(struct blackfin_pda, dpdt_swapcount));
+	DEFINE(PDA_EXIPTR, offsetof(struct blackfin_pda, ex_iptr));
+	DEFINE(PDA_EXOPTR, offsetof(struct blackfin_pda, ex_optr));
+	DEFINE(PDA_EXBUF, offsetof(struct blackfin_pda, ex_buf));
+	DEFINE(PDA_EXIMASK, offsetof(struct blackfin_pda, ex_imask));
+	DEFINE(PDA_EXSTACK, offsetof(struct blackfin_pda, ex_stack));
+#ifdef ANOMALY_05000261
+	DEFINE(PDA_LFRETX, offsetof(struct blackfin_pda, last_cplb_fault_retx));
+#endif
+	DEFINE(PDA_DCPLB, offsetof(struct blackfin_pda, dcplb_fault_addr));
+	DEFINE(PDA_ICPLB, offsetof(struct blackfin_pda, icplb_fault_addr));
+	DEFINE(PDA_RETX, offsetof(struct blackfin_pda, retx));
+	DEFINE(PDA_SEQSTAT, offsetof(struct blackfin_pda, seqstat));
+#ifdef CONFIG_SMP
+	/* Inter-core lock (in L2 SRAM) */
+	DEFINE(SIZEOF_CORELOCK, sizeof(struct corelock_slot));
+#endif
+
 	return 0;
 }

arch/blackfin/kernel/bfin_ksyms.c

@@ -68,3 +68,37 @@ EXPORT_SYMBOL(insw_8);
 EXPORT_SYMBOL(outsl);
 EXPORT_SYMBOL(insl);
 EXPORT_SYMBOL(insl_16);
+
+#ifdef CONFIG_SMP
+EXPORT_SYMBOL(__raw_atomic_update_asm);
+EXPORT_SYMBOL(__raw_atomic_clear_asm);
+EXPORT_SYMBOL(__raw_atomic_set_asm);
+EXPORT_SYMBOL(__raw_atomic_xor_asm);
+EXPORT_SYMBOL(__raw_atomic_test_asm);
+EXPORT_SYMBOL(__raw_xchg_1_asm);
+EXPORT_SYMBOL(__raw_xchg_2_asm);
+EXPORT_SYMBOL(__raw_xchg_4_asm);
+EXPORT_SYMBOL(__raw_cmpxchg_1_asm);
+EXPORT_SYMBOL(__raw_cmpxchg_2_asm);
+EXPORT_SYMBOL(__raw_cmpxchg_4_asm);
+EXPORT_SYMBOL(__raw_spin_is_locked_asm);
+EXPORT_SYMBOL(__raw_spin_lock_asm);
+EXPORT_SYMBOL(__raw_spin_trylock_asm);
+EXPORT_SYMBOL(__raw_spin_unlock_asm);
+EXPORT_SYMBOL(__raw_read_lock_asm);
+EXPORT_SYMBOL(__raw_read_trylock_asm);
+EXPORT_SYMBOL(__raw_read_unlock_asm);
+EXPORT_SYMBOL(__raw_write_lock_asm);
+EXPORT_SYMBOL(__raw_write_trylock_asm);
+EXPORT_SYMBOL(__raw_write_unlock_asm);
+EXPORT_SYMBOL(__raw_bit_set_asm);
+EXPORT_SYMBOL(__raw_bit_clear_asm);
+EXPORT_SYMBOL(__raw_bit_toggle_asm);
+EXPORT_SYMBOL(__raw_bit_test_asm);
+EXPORT_SYMBOL(__raw_bit_test_set_asm);
+EXPORT_SYMBOL(__raw_bit_test_clear_asm);
+EXPORT_SYMBOL(__raw_bit_test_toggle_asm);
+EXPORT_SYMBOL(__raw_uncached_fetch_asm);
+EXPORT_SYMBOL(__raw_smp_mark_barrier_asm);
+EXPORT_SYMBOL(__raw_smp_check_barrier_asm);
+#endif

arch/blackfin/kernel/entry.S

@@ -30,6 +30,7 @@
 #include <linux/linkage.h>
 #include <asm/thread_info.h>
 #include <asm/errno.h>
+#include <asm/blackfin.h>
 #include <asm/asm-offsets.h>
 
 #include <asm/context.S>

arch/blackfin/kernel/irqchip.c

@@ -36,7 +36,7 @@
 #include <linux/irq.h>
 #include <asm/trace.h>
 
-static unsigned long irq_err_count;
+static atomic_t irq_err_count;
 static spinlock_t irq_controller_lock;
 
 /*
@@ -48,7 +48,7 @@ void dummy_mask_unmask_irq(unsigned int irq)
 
 void ack_bad_irq(unsigned int irq)
 {
-	irq_err_count += 1;
+	atomic_inc(&irq_err_count);
 	printk(KERN_ERR "IRQ: spurious interrupt %d\n", irq);
 }
 EXPORT_SYMBOL(ack_bad_irq);
@@ -72,27 +72,28 @@ static struct irq_desc bad_irq_desc = {
 
 int show_interrupts(struct seq_file *p, void *v)
 {
-	int i = *(loff_t *) v;
+	int i = *(loff_t *) v, j;
 	struct irqaction *action;
 	unsigned long flags;
 
 	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: %10u ", i, kstat_irqs(i));
+			goto skip;
+		seq_printf(p, "%3d: ", i);
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
+		seq_printf(p, " %8s", irq_desc[i].chip->name);
 		seq_printf(p, "  %s", action->name);
 		for (action = action->next; action; action = action->next)
-			seq_printf(p, ", %s", action->name);
+			seq_printf(p, "  %s", action->name);
 
 		seq_putc(p, '\n');
- unlock:
+ skip:
 		spin_unlock_irqrestore(&irq_desc[i].lock, flags);
-	} else if (i == NR_IRQS) {
-		seq_printf(p, "Err: %10lu\n", irq_err_count);
-	}
+	} else if (i == NR_IRQS)
+		seq_printf(p, "Err: %10u\n",  atomic_read(&irq_err_count));
 	return 0;
 }
 
@@ -101,7 +102,6 @@ int show_interrupts(struct seq_file *p, void *v)
  * come via this function.  Instead, they should provide their
  * own 'handler'
  */
-
 #ifdef CONFIG_DO_IRQ_L1
 __attribute__((l1_text))
 #endif

arch/blackfin/kernel/kgdb.c

@@ -363,12 +363,12 @@ void kgdb_passive_cpu_callback(void *info)
 
 void kgdb_roundup_cpus(unsigned long flags)
 {
-	smp_call_function(kgdb_passive_cpu_callback, NULL, 0, 0);
+	smp_call_function(kgdb_passive_cpu_callback, NULL, 0);
 }
 
 void kgdb_roundup_cpu(int cpu, unsigned long flags)
 {
-	smp_call_function_single(cpu, kgdb_passive_cpu_callback, NULL, 0, 0);
+	smp_call_function_single(cpu, kgdb_passive_cpu_callback, NULL, 0);
 }
 #endif
 

arch/blackfin/kernel/module.c

@@ -343,7 +343,13 @@ apply_relocate_add(Elf_Shdr * sechdrs, const char *strtab,
 		pr_debug("location is %x, value is %x type is %d \n",
 			 (unsigned int) location32, value,
 			 ELF32_R_TYPE(rel[i].r_info));
-
+#ifdef CONFIG_SMP
+		if ((unsigned long)location16 >= COREB_L1_DATA_A_START) {
+			printk(KERN_ERR "module %s: cannot relocate in L1: %u (SMP kernel)",
+				       mod->name, ELF32_R_TYPE(rel[i].r_info));
+			return -ENOEXEC;
+		}
+#endif
 		switch (ELF32_R_TYPE(rel[i].r_info)) {
 
 		case R_pcrel24:
@@ -436,6 +442,7 @@ module_finalize(const Elf_Ehdr * hdr,
 {
 	unsigned int i, strindex = 0, symindex = 0;
 	char *secstrings;
+	long err = 0;
 
 	secstrings = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset;
 
@@ -460,8 +467,10 @@ module_finalize(const Elf_Ehdr * hdr,
 		    (strcmp(".rela.l1.text", secstrings + sechdrs[i].sh_name) == 0) ||
 		    ((strcmp(".rela.text", secstrings + sechdrs[i].sh_name) == 0) &&
 			(hdr->e_flags & (EF_BFIN_CODE_IN_L1|EF_BFIN_CODE_IN_L2))))) {
-			apply_relocate_add((Elf_Shdr *) sechdrs, strtab,
+			err = apply_relocate_add((Elf_Shdr *) sechdrs, strtab,
 					   symindex, i, mod);
+			if (err < 0)
+				return -ENOEXEC;
 		}
 	}
 	return 0;

arch/blackfin/kernel/process.c

@@ -171,6 +171,13 @@ asmlinkage int bfin_clone(struct pt_regs *regs)
 	unsigned long clone_flags;
 	unsigned long newsp;
 
+#ifdef __ARCH_SYNC_CORE_DCACHE
+	if (current->rt.nr_cpus_allowed == num_possible_cpus()) {
+		current->cpus_allowed = cpumask_of_cpu(smp_processor_id());
+		current->rt.nr_cpus_allowed = 1;
+	}
+#endif
+
 	/* syscall2 puts clone_flags in r0 and usp in r1 */
 	clone_flags = regs->r0;
 	newsp = regs->r1;
@@ -338,22 +345,22 @@ int _access_ok(unsigned long addr, unsigned long size)
 	if (addr >= (unsigned long)__init_begin &&
 	    addr + size <= (unsigned long)__init_end)
 		return 1;
-	if (addr >= L1_SCRATCH_START
-	    && addr + size <= L1_SCRATCH_START + L1_SCRATCH_LENGTH)
+	if (addr >= get_l1_scratch_start()
+	    && addr + size <= get_l1_scratch_start() + L1_SCRATCH_LENGTH)
 		return 1;
 #if L1_CODE_LENGTH != 0
-	if (addr >= L1_CODE_START + (_etext_l1 - _stext_l1)
-	    && addr + size <= L1_CODE_START + L1_CODE_LENGTH)
+	if (addr >= get_l1_code_start() + (_etext_l1 - _stext_l1)
+	    && addr + size <= get_l1_code_start() + L1_CODE_LENGTH)
 		return 1;
 #endif
 #if L1_DATA_A_LENGTH != 0
-	if (addr >= L1_DATA_A_START + (_ebss_l1 - _sdata_l1)
-	    && addr + size <= L1_DATA_A_START + L1_DATA_A_LENGTH)
+	if (addr >= get_l1_data_a_start() + (_ebss_l1 - _sdata_l1)
+	    && addr + size <= get_l1_data_a_start() + L1_DATA_A_LENGTH)
 		return 1;
 #endif
 #if L1_DATA_B_LENGTH != 0
-	if (addr >= L1_DATA_B_START + (_ebss_b_l1 - _sdata_b_l1)
-	    && addr + size <= L1_DATA_B_START + L1_DATA_B_LENGTH)
+	if (addr >= get_l1_data_b_start() + (_ebss_b_l1 - _sdata_b_l1)
+	    && addr + size <= get_l1_data_b_start() + L1_DATA_B_LENGTH)
 		return 1;
 #endif
 #if L2_LENGTH != 0

arch/blackfin/kernel/ptrace.c

@@ -220,8 +220,8 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 				break;
 			pr_debug("ptrace: user address is valid\n");
 
-			if (L1_CODE_LENGTH != 0 && addr >= L1_CODE_START
-			    && addr + sizeof(tmp) <= L1_CODE_START + L1_CODE_LENGTH) {
+			if (L1_CODE_LENGTH != 0 && addr >= get_l1_code_start()
+			    && addr + sizeof(tmp) <= get_l1_code_start() + L1_CODE_LENGTH) {
 				safe_dma_memcpy (&tmp, (const void *)(addr), sizeof(tmp));
 				copied = sizeof(tmp);
 
@@ -300,8 +300,8 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 				break;
 			pr_debug("ptrace: user address is valid\n");
 
-			if (L1_CODE_LENGTH != 0 && addr >= L1_CODE_START
-			    && addr + sizeof(data) <= L1_CODE_START + L1_CODE_LENGTH) {
+			if (L1_CODE_LENGTH != 0 && addr >= get_l1_code_start()
+			    && addr + sizeof(data) <= get_l1_code_start() + L1_CODE_LENGTH) {
 				safe_dma_memcpy ((void *)(addr), &data, sizeof(data));
 				copied = sizeof(data);
 

arch/blackfin/kernel/reboot.c

@@ -21,7 +21,7 @@
  * the core reset.
  */
 __attribute__((l1_text))
-static void bfin_reset(void)
+static void _bfin_reset(void)
 {
 	/* Wait for completion of "system" events such as cache line
 	 * line fills so that we avoid infinite stalls later on as
@@ -66,6 +66,18 @@ static void bfin_reset(void)
 	}
 }
 
+static void bfin_reset(void)
+{
+	if (ANOMALY_05000353 || ANOMALY_05000386)
+		_bfin_reset();
+	else
+		/* the bootrom checks to see how it was reset and will
+		 * automatically perform a software reset for us when
+		 * it starts executing boot
+		 */
+		asm("raise 1;");
+}
+
 __attribute__((weak))
 void native_machine_restart(char *cmd)
 {
@@ -75,14 +87,10 @@ void machine_restart(char *cmd)
 {
 	native_machine_restart(cmd);
 	local_irq_disable();
-	if (ANOMALY_05000353 || ANOMALY_05000386)
-		bfin_reset();
+	if (smp_processor_id())
+		smp_call_function((void *)bfin_reset, 0, 1);
 	else
-		/* the bootrom checks to see how it was reset and will
-		 * automatically perform a software reset for us when
-		 * it starts executing boot
-		 */
-		asm("raise 1;");
+		bfin_reset();
 }
 
 __attribute__((weak))