ARC: SMP support

ARC common code to enable a SMP system + ISS provided SMP extensions.

ARC700 natively lacks SMP support, hence some of the core features are
are only enabled if SoCs have the necessary h/w pixie-dust. This
includes:
-Inter Processor Interrupts (IPI)
-Cache coherency
-load-locked/store-conditional
...

The low level exception handling would be completely broken in SMP
because we don't have hardware assisted stack switching. Thus a fair bit
of this code is repurposing the MMU_SCRATCH reg for event handler
prologues to keep them re-entrant.

Many thanks to Rajeshwar Ranga for his initial "major" contributions to
SMP Port (back in 2008), and to Noam Camus and Gilad Ben-Yossef for help
with resurrecting that in 3.2 kernel (2012).

Note that this platform code is again singleton design pattern - so
multiple SMP platforms won't build at the moment - this deficiency is
addressed in subsequent patches within this series.

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Rajeshwar Ranga <rajeshwar.ranga@gmail.com>
Cc: Noam Camus <noamc@ezchip.com>
Cc: Gilad Ben-Yossef <gilad@benyossef.com>

6 files changed, 316 insertions, 1 deletions

diff --git a/arch/arc/plat-arcfpga/Kconfig b/arch/arc/plat-arcfpga/Kconfig
index 7af3a4e7f7d..38752bfb91e 100644
--- a/arch/arc/plat-arcfpga/Kconfig
+++ b/arch/arc/plat-arcfpga/Kconfig
@@ -13,6 +13,7 @@ choice
 
 config ARC_BOARD_ANGEL4
 	bool "ARC Angel4"
+	select ISS_SMP_EXTN if SMP
 	help
 	  ARC Angel4 FPGA Ref Platform (Xilinx Virtex Based)
 
@@ -21,6 +22,19 @@ config ARC_BOARD_ML509
 	help
 	  ARC ML509 FPGA Ref Platform (Xilinx Virtex-5 Based)
 
+config ISS_SMP_EXTN
+	bool "ARC SMP Extensions (ISS Models only)"
+	default n
+	depends on SMP
+	select ARC_HAS_COH_RTSC
+	help
+	  SMP Extensions to ARC700, in a "simulation only" Model, supported in
+	  ARC ISS (Instruction Set Simulator).
+	  The SMP extensions include:
+	  -IDU (Interrupt Distribution Unit)
+	  -XTL (To enable CPU start/stop/set-PC for another CPU)
+	  It doesn't provide coherent Caches and/or Atomic Ops (LLOCK/SCOND)
+
 endchoice
 
 config ARC_SERIAL_BAUD
diff --git a/arch/arc/plat-arcfpga/Makefile b/arch/arc/plat-arcfpga/Makefile
index 385eb9d83b6..2a828bec821 100644
--- a/arch/arc/plat-arcfpga/Makefile
+++ b/arch/arc/plat-arcfpga/Makefile
@@ -7,3 +7,4 @@
 #
 
 obj-y := platform.o irq.o
+obj-$(CONFIG_SMP)		+= smp.o
diff --git a/arch/arc/plat-arcfpga/include/plat/irq.h b/arch/arc/plat-arcfpga/include/plat/irq.h
index b34e08734c6..255b90e973e 100644
--- a/arch/arc/plat-arcfpga/include/plat/irq.h
+++ b/arch/arc/plat-arcfpga/include/plat/irq.h
@@ -12,7 +12,11 @@
 #ifndef __PLAT_IRQ_H
 #define __PLAT_IRQ_H
 
-#define NR_IRQS		16
+#ifdef CONFIG_SMP
+#define NR_IRQS 32
+#else
+#define NR_IRQS 16
+#endif
 
 #define UART0_IRQ	5
 #define UART1_IRQ	10
@@ -24,4 +28,8 @@
 #define PCI_IRQ		14
 #define PS2_IRQ		15
 
+#ifdef CONFIG_SMP
+#define IDU_INTERRUPT_0 16
+#endif
+
 #endif
diff --git a/arch/arc/plat-arcfpga/include/plat/smp.h b/arch/arc/plat-arcfpga/include/plat/smp.h
new file mode 100644
index 00000000000..8c5e46c01b1
--- /dev/null
+++ b/arch/arc/plat-arcfpga/include/plat/smp.h
@@ -0,0 +1,115 @@
+/*
+ * Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com)
+ *
+ * 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.
+ *
+ *  Rajeshwar Ranga: Interrupt Distribution Unit API's
+ */
+
+#ifndef __PLAT_ARCFPGA_SMP_H
+#define __PLAT_ARCFPGA_SMP_H
+
+#ifdef CONFIG_SMP
+
+#include <linux/types.h>
+#include <asm/arcregs.h>
+
+#define ARC_AUX_IDU_REG_CMD		0x2000
+#define ARC_AUX_IDU_REG_PARAM		0x2001
+
+#define ARC_AUX_XTL_REG_CMD		0x2002
+#define ARC_AUX_XTL_REG_PARAM		0x2003
+
+#define ARC_REG_MP_BCR			0x2021
+
+#define ARC_XTL_CMD_WRITE_PC		0x04
+#define ARC_XTL_CMD_CLEAR_HALT		0x02
+
+/*
+ * Build Configuration Register which identifies the sub-components
+ */
+struct bcr_mp {
+#ifdef CONFIG_CPU_BIG_ENDIAN
+	unsigned int mp_arch:16, pad:5, sdu:1, idu:1, scu:1, ver:8;
+#else
+	unsigned int ver:8, scu:1, idu:1, sdu:1, pad:5, mp_arch:16;
+#endif
+};
+
+/* IDU supports 256 common interrupts */
+#define NR_IDU_IRQS			256
+
+/*
+ * The Aux Regs layout is same bit-by-bit in both BE/LE modes.
+ * However when casted as a bitfield encoded "C" struct, gcc treats it as
+ * memory, generating different code for BE/LE, requiring strcture adj (see
+ * include/asm/arcregs.h)
+ *
+ * However when manually "carving" the value for a Aux, no special handling
+ * of BE is needed because of the property discribed above
+ */
+#define IDU_SET_COMMAND(irq, cmd)			\
+do {							\
+	uint32_t __val;					\
+	__val = (((irq & 0xFF) << 8) | (cmd & 0xFF));	\
+	write_aux_reg(ARC_AUX_IDU_REG_CMD, __val);	\
+} while (0)
+
+#define IDU_SET_PARAM(par)  write_aux_reg(ARC_AUX_IDU_REG_PARAM, par)
+#define IDU_GET_PARAM()     read_aux_reg(ARC_AUX_IDU_REG_PARAM)
+
+/* IDU Commands */
+#define IDU_DISABLE			0x00
+#define IDU_ENABLE			0x01
+#define IDU_IRQ_CLEAR			0x02
+#define IDU_IRQ_ASSERT			0x03
+#define IDU_IRQ_WMODE			0x04
+#define IDU_IRQ_STATUS			0x05
+#define IDU_IRQ_ACK			0x06
+#define IDU_IRQ_PEND			0x07
+#define IDU_IRQ_RMODE			0x08
+#define IDU_IRQ_WBITMASK		0x09
+#define IDU_IRQ_RBITMASK		0x0A
+
+#define idu_enable()		IDU_SET_COMMAND(0, IDU_ENABLE)
+#define idu_disable()		IDU_SET_COMMAND(0, IDU_DISABLE)
+
+#define idu_irq_assert(irq)	IDU_SET_COMMAND((irq), IDU_IRQ_ASSERT)
+#define idu_irq_clear(irq)	IDU_SET_COMMAND((irq), IDU_IRQ_CLEAR)
+
+/* IDU Interrupt Mode - Destination Encoding */
+#define IDU_IRQ_MOD_DISABLE		0x00
+#define IDU_IRQ_MOD_ROUND_RECP		0x01
+#define IDU_IRQ_MOD_TCPU_FIRSTRECP	0x02
+#define IDU_IRQ_MOD_TCPU_ALLRECP	0x03
+
+/* IDU Interrupt Mode  - Triggering Mode */
+#define IDU_IRQ_MODE_LEVEL_TRIG		0x00
+#define IDU_IRQ_MODE_PULSE_TRIG		0x01
+
+#define IDU_IRQ_MODE_PARAM(dest_mode, trig_mode)   \
+	(((trig_mode & 0x01) << 15) | (dest_mode & 0xFF))
+
+struct idu_irq_config {
+	uint8_t irq;
+	uint8_t dest_mode;
+	uint8_t trig_mode;
+};
+
+struct idu_irq_status {
+	uint8_t irq;
+	bool enabled;
+	bool status;
+	bool ack;
+	bool pend;
+	uint8_t next_rr;
+};
+
+extern void idu_irq_set_tgtcpu(uint8_t irq, uint32_t mask);
+extern void idu_irq_set_mode(uint8_t irq, uint8_t dest_mode, uint8_t trig_mode);
+
+#endif	/* CONFIG_SMP */
+
+#endif
diff --git a/arch/arc/plat-arcfpga/irq.c b/arch/arc/plat-arcfpga/irq.c
index ed726360b9f..590edd174c4 100644
--- a/arch/arc/plat-arcfpga/irq.c
+++ b/arch/arc/plat-arcfpga/irq.c
@@ -9,7 +9,17 @@
  */
 
 #include <linux/interrupt.h>
+#include <asm/irq.h>
 
 void __init plat_init_IRQ(void)
 {
+	/*
+	 * SMP Hack because UART IRQ hardwired to cpu0 (boot-cpu) but if the
+	 * request_irq() comes from any other CPU, the low level IRQ unamsking
+	 * essential for getting Interrupts won't be enabled on cpu0, locking
+	 * up the UART state machine.
+	 */
+#ifdef CONFIG_SMP
+	arch_unmask_irq(UART0_IRQ);
+#endif
 }
diff --git a/arch/arc/plat-arcfpga/smp.c b/arch/arc/plat-arcfpga/smp.c
new file mode 100644
index 00000000000..a95fcdb2903
--- /dev/null
+++ b/arch/arc/plat-arcfpga/smp.c
@@ -0,0 +1,167 @@
+/*
+ * ARC700 Simulation-only Extensions for SMP
+ *
+ * Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com)
+ *
+ * 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.
+ *
+ *  Vineet Gupta    - 2012 : split off arch common and plat specific SMP
+ *  Rajeshwar Ranga - 2007 : Interrupt Distribution Unit API's
+ */
+
+#include <linux/smp.h>
+#include <asm/irq.h>
+#include <plat/smp.h>
+
+static char smp_cpuinfo_buf[128];
+
+/*
+ *-------------------------------------------------------------------
+ * Platform specific callbacks expected by arch SMP code
+ *-------------------------------------------------------------------
+ */
+
+const char *arc_platform_smp_cpuinfo(void)
+{
+#define IS_AVAIL1(var, str)    ((var) ? str : "")
+
+	struct bcr_mp mp;
+
+	READ_BCR(ARC_REG_MP_BCR, mp);
+
+	sprintf(smp_cpuinfo_buf, "Extn [700-SMP]: v%d, arch(%d) %s %s %s\n",
+		mp.ver, mp.mp_arch, IS_AVAIL1(mp.scu, "SCU"),
+		IS_AVAIL1(mp.idu, "IDU"), IS_AVAIL1(mp.sdu, "SDU"));
+
+	return smp_cpuinfo_buf;
+}
+
+/*
+ * Master kick starting another CPU
+ */
+void arc_platform_smp_wakeup_cpu(int cpu, unsigned long pc)
+{
+	/* setup the start PC */
+	write_aux_reg(ARC_AUX_XTL_REG_PARAM, pc);
+
+	/* Trigger WRITE_PC cmd for this cpu */
+	write_aux_reg(ARC_AUX_XTL_REG_CMD,
+			(ARC_XTL_CMD_WRITE_PC | (cpu << 8)));
+
+	/* Take the cpu out of Halt */
+	write_aux_reg(ARC_AUX_XTL_REG_CMD,
+			(ARC_XTL_CMD_CLEAR_HALT | (cpu << 8)));
+
+}
+
+/*
+ * Any SMP specific init any CPU does when it comes up.
+ * Here we setup the CPU to enable Inter-Processor-Interrupts
+ * Called for each CPU
+ * -Master      : init_IRQ()
+ * -Other(s)    : start_kernel_secondary()
+ */
+void arc_platform_smp_init_cpu(void)
+{
+	int cpu = smp_processor_id();
+
+	/* Check if CPU is configured for more than 16 interrupts */
+	if (NR_IRQS <= 16 || get_hw_config_num_irq() <= 16)
+		panic("[arcfpga] IRQ system can't support IDU IPI\n");
+
+	idu_disable();
+
+	/****************************************************************
+	 * IDU provides a set of Common IRQs, each of which can be dynamically
+	 * attached to (1|many|all) CPUs.
+	 * The Common IRQs [0-15] are mapped as CPU pvt [16-31]
+	 *
+	 * Here we use a simple 1:1 mapping:
+	 * A CPU 'x' is wired to Common IRQ 'x'.
+	 * So an IDU ASSERT on IRQ 'x' will trigger Interupt on CPU 'x', which
+	 * makes up for our simple IPI plumbing.
+	 *
+	 * TBD: Have a dedicated multicast IRQ for sending IPIs to all CPUs
+	 *      w/o having to do one-at-a-time
+	 ******************************************************************/
+
+	/*
+	 * Claim an IRQ which would trigger IPI on this CPU.
+	 * In IDU parlance it involves setting up a cpu bitmask for the IRQ
+	 * The bitmap here contains only 1 CPU (self).
+	 */
+	idu_irq_set_tgtcpu(cpu, 0x1 << cpu);
+
+	/* Set the IRQ destination to use the bitmask above */
+	idu_irq_set_mode(cpu, 7, /* XXX: IDU_IRQ_MOD_TCPU_ALLRECP: ISS bug */
+			 IDU_IRQ_MODE_PULSE_TRIG);
+
+	idu_enable();
+
+	/* Attach the arch-common IPI ISR to our IDU IRQ */
+	smp_ipi_irq_setup(cpu, IDU_INTERRUPT_0 + cpu);
+}
+
+void arc_platform_ipi_send(const struct cpumask *callmap)
+{
+	unsigned int cpu;
+
+	for_each_cpu(cpu, callmap)
+		idu_irq_assert(cpu);
+}
+
+void arc_platform_ipi_clear(int cpu, int irq)
+{
+	idu_irq_clear(IDU_INTERRUPT_0 + cpu);
+}
+
+/*
+ *-------------------------------------------------------------------
+ * Low level Platform IPI Providers
+ *-------------------------------------------------------------------
+ */
+
+/* Set the Mode for the Common IRQ */
+void idu_irq_set_mode(uint8_t irq, uint8_t dest_mode, uint8_t trig_mode)
+{
+	uint32_t par = IDU_IRQ_MODE_PARAM(dest_mode, trig_mode);
+
+	IDU_SET_PARAM(par);
+	IDU_SET_COMMAND(irq, IDU_IRQ_WMODE);
+}
+
+/* Set the target cpu Bitmask for Common IRQ */
+void idu_irq_set_tgtcpu(uint8_t irq, uint32_t mask)
+{
+	IDU_SET_PARAM(mask);
+	IDU_SET_COMMAND(irq, IDU_IRQ_WBITMASK);
+}
+
+/* Get the Interrupt Acknowledged status for IRQ (as CPU Bitmask) */
+bool idu_irq_get_ack(uint8_t irq)
+{
+	uint32_t val;
+
+	IDU_SET_COMMAND(irq, IDU_IRQ_ACK);
+	val = IDU_GET_PARAM();
+
+	return val & (1 << irq);
+}
+
+/*
+ * Get the Interrupt Pending status for IRQ (as CPU Bitmask)
+ * -Pending means CPU has not yet noticed the IRQ (e.g. disabled)
+ * -After Interrupt has been taken, the IPI expcitily needs to be
+ *  cleared, to be acknowledged.
+ */
+bool idu_irq_get_pend(uint8_t irq)
+{
+	uint32_t val;
+
+	IDU_SET_COMMAND(irq, IDU_IRQ_PEND);
+	val = IDU_GET_PARAM();
+
+	return val & (1 << irq);
+}