diff --git a/include/asm-sparc/agp.h b/include/asm-sparc/agp.h
new file mode 100644
index 0000000..c245687
--- /dev/null
+++ b/include/asm-sparc/agp.h
@@ -0,0 +1,20 @@
+#ifndef AGP_H
+#define AGP_H 1
+
+/* dummy for now */
+
+#define map_page_into_agp(page)
+#define unmap_page_from_agp(page)
+#define flush_agp_cache() mb()
+
+/* Convert a physical address to an address suitable for the GART. */
+#define phys_to_gart(x) (x)
+#define gart_to_phys(x) (x)
+
+/* GATT allocation. Returns/accepts GATT kernel virtual address. */
+#define alloc_gatt_pages(order)		\
+	((char *)__get_free_pages(GFP_KERNEL, (order)))
+#define free_gatt_pages(table, order)	\
+	free_pages((unsigned long)(table), (order))
+
+#endif
diff --git a/include/asm-sparc/apb.h b/include/asm-sparc/apb.h
new file mode 100644
index 0000000..8f3b57d
--- /dev/null
+++ b/include/asm-sparc/apb.h
@@ -0,0 +1,36 @@
+/*
+ * apb.h: Advanced PCI Bridge Configuration Registers and Bits
+ *
+ * Copyright (C) 1998  Eddie C. Dost  (ecd@skynet.be)
+ */
+
+#ifndef _SPARC64_APB_H
+#define _SPARC64_APB_H
+
+#define APB_TICK_REGISTER			0xb0
+#define APB_INT_ACK				0xb8
+#define APB_PRIMARY_MASTER_RETRY_LIMIT		0xc0
+#define APB_DMA_ASFR				0xc8
+#define APB_DMA_AFAR				0xd0
+#define APB_PIO_TARGET_RETRY_LIMIT		0xd8
+#define APB_PIO_TARGET_LATENCY_TIMER		0xd9
+#define APB_DMA_TARGET_RETRY_LIMIT		0xda
+#define APB_DMA_TARGET_LATENCY_TIMER		0xdb
+#define APB_SECONDARY_MASTER_RETRY_LIMIT	0xdc
+#define APB_SECONDARY_CONTROL			0xdd
+#define APB_IO_ADDRESS_MAP			0xde
+#define APB_MEM_ADDRESS_MAP			0xdf
+
+#define APB_PCI_CONTROL_LOW			0xe0
+#  define APB_PCI_CTL_LOW_ARB_PARK			(1 << 21)
+#  define APB_PCI_CTL_LOW_ERRINT_EN			(1 << 8)
+
+#define APB_PCI_CONTROL_HIGH			0xe4
+#  define APB_PCI_CTL_HIGH_SERR				(1 << 2)
+#  define APB_PCI_CTL_HIGH_ARBITER_EN			(1 << 0)
+
+#define APB_PIO_ASFR				0xe8
+#define APB_PIO_AFAR				0xf0
+#define APB_DIAG_REGISTER			0xf8
+
+#endif /* !(_SPARC64_APB_H) */
diff --git a/include/asm-sparc/backoff.h b/include/asm-sparc/backoff.h
new file mode 100644
index 0000000..fa1fdf6
--- /dev/null
+++ b/include/asm-sparc/backoff.h
@@ -0,0 +1,31 @@
+#ifndef _SPARC64_BACKOFF_H
+#define _SPARC64_BACKOFF_H
+
+#define BACKOFF_LIMIT	(4 * 1024)
+
+#ifdef CONFIG_SMP
+
+#define BACKOFF_SETUP(reg)	\
+	mov	1, reg
+
+#define BACKOFF_SPIN(reg, tmp, label)	\
+	mov	reg, tmp; \
+88:	brnz,pt	tmp, 88b; \
+	 sub	tmp, 1, tmp; \
+	set	BACKOFF_LIMIT, tmp; \
+	cmp	reg, tmp; \
+	bg,pn	%xcc, label; \
+	 nop; \
+	ba,pt	%xcc, label; \
+	 sllx	reg, 1, reg;
+
+#else
+
+#define BACKOFF_SETUP(reg)
+#define BACKOFF_SPIN(reg, tmp, label) \
+	ba,pt	%xcc, label; \
+	 nop;
+
+#endif
+
+#endif /* _SPARC64_BACKOFF_H */
diff --git a/include/asm-sparc/bbc.h b/include/asm-sparc/bbc.h
new file mode 100644
index 0000000..423a858
--- /dev/null
+++ b/include/asm-sparc/bbc.h
@@ -0,0 +1,225 @@
+/*
+ * bbc.h: Defines for BootBus Controller found on UltraSPARC-III
+ *        systems.
+ *
+ * Copyright (C) 2000 David S. Miller (davem@redhat.com)
+ */
+
+#ifndef _SPARC64_BBC_H
+#define _SPARC64_BBC_H
+
+/* Register sizes are indicated by "B" (Byte, 1-byte),
+ * "H" (Half-word, 2 bytes), "W" (Word, 4 bytes) or
+ * "Q" (Quad, 8 bytes) inside brackets.
+ */
+
+#define BBC_AID		0x00	/* [B] Agent ID			*/
+#define BBC_DEVP	0x01	/* [B] Device Present		*/
+#define BBC_ARB		0x02	/* [B] Arbitration		*/
+#define BBC_QUIESCE	0x03	/* [B] Quiesce			*/
+#define BBC_WDACTION	0x04	/* [B] Watchdog Action		*/
+#define BBC_SPG		0x06	/* [B] Soft POR Gen		*/
+#define BBC_SXG		0x07	/* [B] Soft XIR Gen		*/
+#define BBC_PSRC	0x08	/* [W] POR Source		*/
+#define BBC_XSRC	0x0c	/* [B] XIR Source		*/
+#define BBC_CSC		0x0d	/* [B] Clock Synthesizers Control*/
+#define BBC_ES_CTRL	0x0e	/* [H] Energy Star Control	*/
+#define BBC_ES_ACT	0x10	/* [W] E* Assert Change Time	*/
+#define BBC_ES_DACT	0x14	/* [B] E* De-Assert Change Time	*/
+#define BBC_ES_DABT	0x15	/* [B] E* De-Assert Bypass Time	*/
+#define BBC_ES_ABT	0x16	/* [H] E* Assert Bypass Time	*/
+#define BBC_ES_PST	0x18	/* [W] E* PLL Settle Time	*/
+#define BBC_ES_FSL	0x1c	/* [W] E* Frequency Switch Latency*/
+#define BBC_EBUST	0x20	/* [Q] EBUS Timing		*/
+#define BBC_JTAG_CMD	0x28	/* [W] JTAG+ Command		*/
+#define BBC_JTAG_CTRL	0x2c	/* [B] JTAG+ Control		*/
+#define BBC_I2C_SEL	0x2d	/* [B] I2C Selection		*/
+#define BBC_I2C_0_S1	0x2e	/* [B] I2C ctrlr-0 reg S1	*/
+#define BBC_I2C_0_S0	0x2f	/* [B] I2C ctrlr-0 regs S0,S0',S2,S3*/
+#define BBC_I2C_1_S1	0x30	/* [B] I2C ctrlr-1 reg S1	*/
+#define BBC_I2C_1_S0	0x31	/* [B] I2C ctrlr-1 regs S0,S0',S2,S3*/
+#define BBC_KBD_BEEP	0x32	/* [B] Keyboard Beep		*/
+#define BBC_KBD_BCNT	0x34	/* [W] Keyboard Beep Counter	*/
+
+#define BBC_REGS_SIZE	0x40
+
+/* There is a 2K scratch ram area at offset 0x80000 but I doubt
+ * we will use it for anything.
+ */
+
+/* Agent ID register.  This register shows the Safari Agent ID
+ * for the processors.  The value returned depends upon which
+ * cpu is reading the register.
+ */
+#define BBC_AID_ID	0x07	/* Safari ID		*/
+#define BBC_AID_RESV	0xf8	/* Reserved		*/
+
+/* Device Present register.  One can determine which cpus are actually
+ * present in the machine by interrogating this register.
+ */
+#define BBC_DEVP_CPU0	0x01	/* Processor 0 present	*/
+#define BBC_DEVP_CPU1	0x02	/* Processor 1 present	*/
+#define BBC_DEVP_CPU2	0x04	/* Processor 2 present	*/
+#define BBC_DEVP_CPU3	0x08	/* Processor 3 present	*/
+#define BBC_DEVP_RESV	0xf0	/* Reserved		*/
+
+/* Arbitration register.  This register is used to block access to
+ * the BBC from a particular cpu.
+ */
+#define BBC_ARB_CPU0	0x01	/* Enable cpu 0 BBC arbitratrion */
+#define BBC_ARB_CPU1	0x02	/* Enable cpu 1 BBC arbitratrion */
+#define BBC_ARB_CPU2	0x04	/* Enable cpu 2 BBC arbitratrion */
+#define BBC_ARB_CPU3	0x08	/* Enable cpu 3 BBC arbitratrion */
+#define BBC_ARB_RESV	0xf0	/* Reserved			 */
+
+/* Quiesce register.  Bus and BBC segments for cpus can be disabled
+ * with this register, ie. for hot plugging.
+ */
+#define BBC_QUIESCE_S02	0x01	/* Quiesce Safari segment for cpu 0 and 2 */
+#define BBC_QUIESCE_S13	0x02	/* Quiesce Safari segment for cpu 1 and 3 */
+#define BBC_QUIESCE_B02	0x04	/* Quiesce BBC segment for cpu 0 and 2    */
+#define BBC_QUIESCE_B13	0x08	/* Quiesce BBC segment for cpu 1 and 3    */
+#define BBC_QUIESCE_FD0 0x10	/* Disable Fatal_Error[0] reporting	  */
+#define BBC_QUIESCE_FD1 0x20	/* Disable Fatal_Error[1] reporting	  */
+#define BBC_QUIESCE_FD2 0x40	/* Disable Fatal_Error[2] reporting	  */
+#define BBC_QUIESCE_FD3 0x80	/* Disable Fatal_Error[3] reporting	  */
+
+/* Watchdog Action register.  When the watchdog device timer expires
+ * a line is enabled to the BBC.  The action BBC takes when this line
+ * is asserted can be controlled by this regiser.
+ */
+#define BBC_WDACTION_RST  0x01	/* When set, watchdog causes system reset.
+				 * When clear, BBC ignores watchdog signal.
+				 */
+#define BBC_WDACTION_RESV 0xfe	/* Reserved */
+
+/* Soft_POR_GEN register.  The POR (Power On Reset) signal may be asserted
+ * for specific processors or all processors via this register.
+ */
+#define BBC_SPG_CPU0	0x01 /* Assert POR for processor 0	*/
+#define BBC_SPG_CPU1	0x02 /* Assert POR for processor 1	*/
+#define BBC_SPG_CPU2	0x04 /* Assert POR for processor 2	*/
+#define BBC_SPG_CPU3	0x08 /* Assert POR for processor 3	*/
+#define BBC_SPG_CPUALL	0x10 /* Reset all processors and reset
+			      * the entire system.
+			      */
+#define BBC_SPG_RESV	0xe0 /* Reserved			*/
+
+/* Soft_XIR_GEN register.  The XIR (eXternally Initiated Reset) signal
+ * may be asserted to specific processors via this register.
+ */
+#define BBC_SXG_CPU0	0x01 /* Assert XIR for processor 0	*/
+#define BBC_SXG_CPU1	0x02 /* Assert XIR for processor 1	*/
+#define BBC_SXG_CPU2	0x04 /* Assert XIR for processor 2	*/
+#define BBC_SXG_CPU3	0x08 /* Assert XIR for processor 3	*/
+#define BBC_SXG_RESV	0xf0 /* Reserved			*/
+
+/* POR Source register.  One may identify the cause of the most recent
+ * reset by reading this register.
+ */
+#define BBC_PSRC_SPG0	0x0001 /* CPU 0 reset via BBC_SPG register	*/
+#define BBC_PSRC_SPG1	0x0002 /* CPU 1 reset via BBC_SPG register	*/
+#define BBC_PSRC_SPG2	0x0004 /* CPU 2 reset via BBC_SPG register	*/
+#define BBC_PSRC_SPG3	0x0008 /* CPU 3 reset via BBC_SPG register	*/
+#define BBC_PSRC_SPGSYS	0x0010 /* System reset via BBC_SPG register	*/
+#define BBC_PSRC_JTAG	0x0020 /* System reset via JTAG+		*/
+#define BBC_PSRC_BUTTON	0x0040 /* System reset via push-button dongle	*/
+#define BBC_PSRC_PWRUP	0x0080 /* System reset via power-up		*/
+#define BBC_PSRC_FE0	0x0100 /* CPU 0 reported Fatal_Error		*/
+#define BBC_PSRC_FE1	0x0200 /* CPU 1 reported Fatal_Error		*/
+#define BBC_PSRC_FE2	0x0400 /* CPU 2 reported Fatal_Error		*/
+#define BBC_PSRC_FE3	0x0800 /* CPU 3 reported Fatal_Error		*/
+#define BBC_PSRC_FE4	0x1000 /* Schizo reported Fatal_Error		*/
+#define BBC_PSRC_FE5	0x2000 /* Safari device 5 reported Fatal_Error	*/
+#define BBC_PSRC_FE6	0x4000 /* CPMS reported Fatal_Error		*/
+#define BBC_PSRC_SYNTH	0x8000 /* System reset when on-board clock synthesizers
+				* were updated.
+				*/
+#define BBC_PSRC_WDT   0x10000 /* System reset via Super I/O watchdog	*/
+#define BBC_PSRC_RSC   0x20000 /* System reset via RSC remote monitoring
+				* device
+				*/
+
+/* XIR Source register.  The source of an XIR event sent to a processor may
+ * be determined via this register.
+ */
+#define BBC_XSRC_SXG0	0x01	/* CPU 0 received XIR via Soft_XIR_GEN reg */
+#define BBC_XSRC_SXG1	0x02	/* CPU 1 received XIR via Soft_XIR_GEN reg */
+#define BBC_XSRC_SXG2	0x04	/* CPU 2 received XIR via Soft_XIR_GEN reg */
+#define BBC_XSRC_SXG3	0x08	/* CPU 3 received XIR via Soft_XIR_GEN reg */
+#define BBC_XSRC_JTAG	0x10	/* All CPUs received XIR via JTAG+         */
+#define BBC_XSRC_W_OR_B	0x20	/* All CPUs received XIR either because:
+				 * a) Super I/O watchdog fired, or
+				 * b) XIR push button was activated
+				 */
+#define BBC_XSRC_RESV	0xc0	/* Reserved				   */
+
+/* Clock Synthesizers Control register.  This register provides the big-bang
+ * programming interface to the two clock synthesizers of the machine.
+ */
+#define BBC_CSC_SLOAD	0x01	/* Directly connected to S_LOAD pins	*/
+#define BBC_CSC_SDATA	0x02	/* Directly connected to S_DATA pins	*/
+#define BBC_CSC_SCLOCK	0x04	/* Directly connected to S_CLOCK pins	*/
+#define BBC_CSC_RESV	0x78	/* Reserved				*/
+#define BBC_CSC_RST	0x80	/* Generate system reset when S_LOAD==1	*/
+
+/* Energy Star Control register.  This register is used to generate the
+ * clock frequency change trigger to the main system devices (Schizo and
+ * the processors).  The transition occurs when bits in this register
+ * go from 0 to 1, only one bit must be set at once else no action
+ * occurs.  Basically the sequence of events is:
+ * a) Choose new frequency: full, 1/2 or 1/32
+ * b) Program this desired frequency into the cpus and Schizo.
+ * c) Set the same value in this register.
+ * d) 16 system clocks later, clear this register.
+ */
+#define BBC_ES_CTRL_1_1		0x01	/* Full frequency	*/
+#define BBC_ES_CTRL_1_2		0x02	/* 1/2 frequency	*/
+#define BBC_ES_CTRL_1_32	0x20	/* 1/32 frequency	*/
+#define BBC_ES_RESV		0xdc	/* Reserved		*/
+
+/* Energy Star Assert Change Time register.  This determines the number
+ * of BBC clock cycles (which is half the system frequency) between
+ * the detection of FREEZE_ACK being asserted and the assertion of
+ * the CLK_CHANGE_L[2:0] signals.
+ */
+#define BBC_ES_ACT_VAL	0xff
+
+/* Energy Star Assert Bypass Time register.  This determines the number
+ * of BBC clock cycles (which is half the system frequency) between
+ * the assertion of the CLK_CHANGE_L[2:0] signals and the assertion of
+ * the ESTAR_PLL_BYPASS signal.
+ */
+#define BBC_ES_ABT_VAL	0xffff
+
+/* Energy Star PLL Settle Time register.  This determines the number of
+ * BBC clock cycles (which is half the system frequency) between the
+ * de-assertion of CLK_CHANGE_L[2:0] and the de-assertion of the FREEZE_L
+ * signal.
+ */
+#define BBC_ES_PST_VAL	0xffffffff
+
+/* Energy Star Frequency Switch Latency register.  This is the number of
+ * BBC clocks between the de-assertion of CLK_CHANGE_L[2:0] and the first
+ * edge of the Safari clock at the new frequency.
+ */
+#define BBC_ES_FSL_VAL	0xffffffff
+
+/* Keyboard Beep control register.  This is a simple enabler for the audio
+ * beep sound.
+ */
+#define BBC_KBD_BEEP_ENABLE	0x01 /* Enable beep	*/
+#define BBC_KBD_BEEP_RESV	0xfe /* Reserved	*/
+
+/* Keyboard Beep Counter register.  There is a free-running counter inside
+ * the BBC which runs at half the system clock.  The bit set in this register
+ * determines when the audio sound is generated.  So for example if bit
+ * 10 is set, the audio beep will oscillate at 1/(2**12).  The keyboard beep
+ * generator automatically selects a different bit to use if the system clock
+ * is changed via Energy Star.
+ */
+#define BBC_KBD_BCNT_BITS	0x0007fc00
+#define BBC_KBC_BCNT_RESV	0xfff803ff
+
+#endif /* _SPARC64_BBC_H */
+
diff --git a/include/asm-sparc/chafsr.h b/include/asm-sparc/chafsr.h
new file mode 100644
index 0000000..85c69b3
--- /dev/null
+++ b/include/asm-sparc/chafsr.h
@@ -0,0 +1,241 @@
+#ifndef _SPARC64_CHAFSR_H
+#define _SPARC64_CHAFSR_H
+
+/* Cheetah Asynchronous Fault Status register, ASI=0x4C VA<63:0>=0x0 */
+
+/* Comments indicate which processor variants on which the bit definition
+ * is valid.  Codes are:
+ * ch	-->	cheetah
+ * ch+	-->	cheetah plus
+ * jp	-->	jalapeno
+ */
+
+/* All bits of this register except M_SYNDROME and E_SYNDROME are
+ * read, write 1 to clear.  M_SYNDROME and E_SYNDROME are read-only.
+ */
+
+/* Software bit set by linux trap handlers to indicate that the trap was
+ * signalled at %tl >= 1.
+ */
+#define CHAFSR_TL1		(1UL << 63UL) /* n/a */
+
+/* Unmapped error from system bus for prefetch queue or
+ * store queue read operation
+ */
+#define CHPAFSR_DTO		(1UL << 59UL) /* ch+ */
+
+/* Bus error from system bus for prefetch queue or store queue
+ * read operation
+ */
+#define CHPAFSR_DBERR		(1UL << 58UL) /* ch+ */
+
+/* Hardware corrected E-cache Tag ECC error */
+#define CHPAFSR_THCE		(1UL << 57UL) /* ch+ */
+/* System interface protocol error, hw timeout caused */
+#define JPAFSR_JETO		(1UL << 57UL) /* jp */
+
+/* SW handled correctable E-cache Tag ECC error */
+#define CHPAFSR_TSCE		(1UL << 56UL) /* ch+ */
+/* Parity error on system snoop results */
+#define JPAFSR_SCE		(1UL << 56UL) /* jp */
+
+/* Uncorrectable E-cache Tag ECC error */
+#define CHPAFSR_TUE		(1UL << 55UL) /* ch+ */
+/* System interface protocol error, illegal command detected */
+#define JPAFSR_JEIC		(1UL << 55UL) /* jp */
+
+/* Uncorrectable system bus data ECC error due to prefetch
+ * or store fill request
+ */
+#define CHPAFSR_DUE		(1UL << 54UL) /* ch+ */
+/* System interface protocol error, illegal ADTYPE detected */
+#define JPAFSR_JEIT		(1UL << 54UL) /* jp */
+
+/* Multiple errors of the same type have occurred.  This bit is set when
+ * an uncorrectable error or a SW correctable error occurs and the status
+ * bit to report that error is already set.  When multiple errors of
+ * different types are indicated by setting multiple status bits.
+ *
+ * This bit is not set if multiple HW corrected errors with the same
+ * status bit occur, only uncorrectable and SW correctable ones have
+ * this behavior.
+ *
+ * This bit is not set when multiple ECC errors happen within a single
+ * 64-byte system bus transaction.  Only the first ECC error in a 16-byte
+ * subunit will be logged.  All errors in subsequent 16-byte subunits
+ * from the same 64-byte transaction are ignored.
+ */
+#define CHAFSR_ME		(1UL << 53UL) /* ch,ch+,jp */
+
+/* Privileged state error has occurred.  This is a capture of PSTATE.PRIV
+ * at the time the error is detected.
+ */
+#define CHAFSR_PRIV		(1UL << 52UL) /* ch,ch+,jp */
+
+/* The following bits 51 (CHAFSR_PERR) to 33 (CHAFSR_CE) are sticky error
+ * bits and record the most recently detected errors.  Bits accumulate
+ * errors that have been detected since the last write to clear the bit.
+ */
+
+/* System interface protocol error.  The processor asserts its' ERROR
+ * pin when this event occurs and it also logs a specific cause code
+ * into a JTAG scannable flop.
+ */
+#define CHAFSR_PERR		(1UL << 51UL) /* ch,ch+,jp */
+
+/* Internal processor error.  The processor asserts its' ERROR
+ * pin when this event occurs and it also logs a specific cause code
+ * into a JTAG scannable flop.
+ */
+#define CHAFSR_IERR		(1UL << 50UL) /* ch,ch+,jp */
+
+/* System request parity error on incoming address */
+#define CHAFSR_ISAP		(1UL << 49UL) /* ch,ch+,jp */
+
+/* HW Corrected system bus MTAG ECC error */
+#define CHAFSR_EMC		(1UL << 48UL) /* ch,ch+ */
+/* Parity error on L2 cache tag SRAM */
+#define JPAFSR_ETP		(1UL << 48UL) /* jp */
+
+/* Uncorrectable system bus MTAG ECC error */
+#define CHAFSR_EMU		(1UL << 47UL) /* ch,ch+ */
+/* Out of range memory error has occurred */
+#define JPAFSR_OM		(1UL << 47UL) /* jp */
+
+/* HW Corrected system bus data ECC error for read of interrupt vector */
+#define CHAFSR_IVC		(1UL << 46UL) /* ch,ch+ */
+/* Error due to unsupported store */
+#define JPAFSR_UMS		(1UL << 46UL) /* jp */
+
+/* Uncorrectable system bus data ECC error for read of interrupt vector */
+#define CHAFSR_IVU		(1UL << 45UL) /* ch,ch+,jp */
+
+/* Unmapped error from system bus */
+#define CHAFSR_TO		(1UL << 44UL) /* ch,ch+,jp */
+
+/* Bus error response from system bus */
+#define CHAFSR_BERR		(1UL << 43UL) /* ch,ch+,jp */
+
+/* SW Correctable E-cache ECC error for instruction fetch or data access
+ * other than block load.
+ */
+#define CHAFSR_UCC		(1UL << 42UL) /* ch,ch+,jp */
+
+/* Uncorrectable E-cache ECC error for instruction fetch or data access
+ * other than block load.
+ */
+#define CHAFSR_UCU		(1UL << 41UL) /* ch,ch+,jp */
+
+/* Copyout HW Corrected ECC error */
+#define CHAFSR_CPC		(1UL << 40UL) /* ch,ch+,jp */
+
+/* Copyout Uncorrectable ECC error */
+#define CHAFSR_CPU		(1UL << 39UL) /* ch,ch+,jp */
+
+/* HW Corrected ECC error from E-cache for writeback */
+#define CHAFSR_WDC		(1UL << 38UL) /* ch,ch+,jp */
+
+/* Uncorrectable ECC error from E-cache for writeback */
+#define CHAFSR_WDU		(1UL << 37UL) /* ch,ch+,jp */
+
+/* HW Corrected ECC error from E-cache for store merge or block load */
+#define CHAFSR_EDC		(1UL << 36UL) /* ch,ch+,jp */
+
+/* Uncorrectable ECC error from E-cache for store merge or block load */
+#define CHAFSR_EDU		(1UL << 35UL) /* ch,ch+,jp */
+
+/* Uncorrectable system bus data ECC error for read of memory or I/O */
+#define CHAFSR_UE		(1UL << 34UL) /* ch,ch+,jp */
+
+/* HW Corrected system bus data ECC error for read of memory or I/O */
+#define CHAFSR_CE		(1UL << 33UL) /* ch,ch+,jp */
+
+/* Uncorrectable ECC error from remote cache/memory */
+#define JPAFSR_RUE		(1UL << 32UL) /* jp */
+
+/* Correctable ECC error from remote cache/memory */
+#define JPAFSR_RCE		(1UL << 31UL) /* jp */
+
+/* JBUS parity error on returned read data */
+#define JPAFSR_BP		(1UL << 30UL) /* jp */
+
+/* JBUS parity error on data for writeback or block store */
+#define JPAFSR_WBP		(1UL << 29UL) /* jp */
+
+/* Foreign read to DRAM incurring correctable ECC error */
+#define JPAFSR_FRC		(1UL << 28UL) /* jp */
+
+/* Foreign read to DRAM incurring uncorrectable ECC error */
+#define JPAFSR_FRU		(1UL << 27UL) /* jp */
+
+#define CHAFSR_ERRORS		(CHAFSR_PERR | CHAFSR_IERR | CHAFSR_ISAP | CHAFSR_EMC | \
+				 CHAFSR_EMU | CHAFSR_IVC | CHAFSR_IVU | CHAFSR_TO | \
+				 CHAFSR_BERR | CHAFSR_UCC | CHAFSR_UCU | CHAFSR_CPC | \
+				 CHAFSR_CPU | CHAFSR_WDC | CHAFSR_WDU | CHAFSR_EDC | \
+				 CHAFSR_EDU | CHAFSR_UE | CHAFSR_CE)
+#define CHPAFSR_ERRORS		(CHPAFSR_DTO | CHPAFSR_DBERR | CHPAFSR_THCE | \
+				 CHPAFSR_TSCE | CHPAFSR_TUE | CHPAFSR_DUE | \
+				 CHAFSR_PERR | CHAFSR_IERR | CHAFSR_ISAP | CHAFSR_EMC | \
+				 CHAFSR_EMU | CHAFSR_IVC | CHAFSR_IVU | CHAFSR_TO | \
+				 CHAFSR_BERR | CHAFSR_UCC | CHAFSR_UCU | CHAFSR_CPC | \
+				 CHAFSR_CPU | CHAFSR_WDC | CHAFSR_WDU | CHAFSR_EDC | \
+				 CHAFSR_EDU | CHAFSR_UE | CHAFSR_CE)
+#define JPAFSR_ERRORS		(JPAFSR_JETO | JPAFSR_SCE | JPAFSR_JEIC | \
+				 JPAFSR_JEIT | CHAFSR_PERR | CHAFSR_IERR | \
+				 CHAFSR_ISAP | JPAFSR_ETP | JPAFSR_OM | \
+				 JPAFSR_UMS | CHAFSR_IVU | CHAFSR_TO | \
+				 CHAFSR_BERR | CHAFSR_UCC | CHAFSR_UCU | \
+				 CHAFSR_CPC | CHAFSR_CPU | CHAFSR_WDC | \
+				 CHAFSR_WDU | CHAFSR_EDC | CHAFSR_EDU | \
+				 CHAFSR_UE | CHAFSR_CE | JPAFSR_RUE | \
+				 JPAFSR_RCE | JPAFSR_BP | JPAFSR_WBP | \
+				 JPAFSR_FRC | JPAFSR_FRU)
+
+/* Active JBUS request signal when error occurred */
+#define JPAFSR_JBREQ		(0x7UL << 24UL) /* jp */
+#define JPAFSR_JBREQ_SHIFT	24UL
+
+/* L2 cache way information */
+#define JPAFSR_ETW		(0x3UL << 22UL) /* jp */
+#define JPAFSR_ETW_SHIFT	22UL
+
+/* System bus MTAG ECC syndrome.  This field captures the status of the
+ * first occurrence of the highest-priority error according to the M_SYND
+ * overwrite policy.  After the AFSR sticky bit, corresponding to the error
+ * for which the M_SYND is reported, is cleared, the contents of the M_SYND
+ * field will be unchanged by will be unfrozen for further error capture.
+ */
+#define CHAFSR_M_SYNDROME	(0xfUL << 16UL) /* ch,ch+,jp */
+#define CHAFSR_M_SYNDROME_SHIFT	16UL
+
+/* Agenid Id of the foreign device causing the UE/CE errors */
+#define JPAFSR_AID		(0x1fUL << 9UL) /* jp */
+#define JPAFSR_AID_SHIFT	9UL
+
+/* System bus or E-cache data ECC syndrome.  This field captures the status
+ * of the first occurrence of the highest-priority error according to the
+ * E_SYND overwrite policy.  After the AFSR sticky bit, corresponding to the
+ * error for which the E_SYND is reported, is cleare, the contents of the E_SYND
+ * field will be unchanged but will be unfrozen for further error capture.
+ */
+#define CHAFSR_E_SYNDROME	(0x1ffUL << 0UL) /* ch,ch+,jp */
+#define CHAFSR_E_SYNDROME_SHIFT	0UL
+
+/* The AFSR must be explicitly cleared by software, it is not cleared automatically
+ * by a read.  Writes to bits <51:33> with bits set will clear the corresponding
+ * bits in the AFSR.  Bits associated with disrupting traps must be cleared before
+ * interrupts are re-enabled to prevent multiple traps for the same error.  I.e.
+ * PSTATE.IE and AFSR bits control delivery of disrupting traps.
+ *
+ * Since there is only one AFAR, when multiple events have been logged by the
+ * bits in the AFSR, at most one of these events will have its status captured
+ * in the AFAR.  The highest priority of those event bits will get AFAR logging.
+ * The AFAR will be unlocked and available to capture the address of another event
+ * as soon as the one bit in AFSR that corresponds to the event logged in AFAR is
+ * cleared.  For example, if AFSR.CE is detected, then AFSR.UE (which overwrites
+ * the AFAR), and AFSR.UE is cleared by not AFSR.CE, then the AFAR will be unlocked
+ * and ready for another event, even though AFSR.CE is still set.  The same rules
+ * also apply to the M_SYNDROME and E_SYNDROME fields of the AFSR.
+ */
+
+#endif /* _SPARC64_CHAFSR_H */
diff --git a/include/asm-sparc/chmctrl.h b/include/asm-sparc/chmctrl.h
new file mode 100644
index 0000000..859b4a4
--- /dev/null
+++ b/include/asm-sparc/chmctrl.h
@@ -0,0 +1,183 @@
+#ifndef _SPARC64_CHMCTRL_H
+#define _SPARC64_CHMCTRL_H
+
+/* Cheetah memory controller programmable registers. */
+#define CHMCTRL_TCTRL1		0x00 /* Memory Timing Control I		*/
+#define CHMCTRL_TCTRL2		0x08 /* Memory Timing Control II	*/
+#define CHMCTRL_TCTRL3		0x38 /* Memory Timing Control III	*/
+#define CHMCTRL_TCTRL4		0x40 /* Memory Timing Control IV	*/
+#define CHMCTRL_DECODE1		0x10 /* Memory Address Decode I		*/
+#define CHMCTRL_DECODE2		0x18 /* Memory Address Decode II	*/
+#define CHMCTRL_DECODE3		0x20 /* Memory Address Decode III	*/
+#define CHMCTRL_DECODE4		0x28 /* Memory Address Decode IV	*/
+#define CHMCTRL_MACTRL		0x30 /* Memory Address Control		*/
+
+/* Memory Timing Control I */
+#define TCTRL1_SDRAMCTL_DLY	0xf000000000000000UL
+#define TCTRL1_SDRAMCTL_DLY_SHIFT     60
+#define TCTRL1_SDRAMCLK_DLY	0x0e00000000000000UL
+#define TCTRL1_SDRAMCLK_DLY_SHIFT     57
+#define TCTRL1_R		0x0100000000000000UL
+#define TCTRL1_R_SHIFT 		      56
+#define TCTRL1_AUTORFR_CYCLE	0x00fe000000000000UL
+#define TCTRL1_AUTORFR_CYCLE_SHIFT    49
+#define TCTRL1_RD_WAIT		0x0001f00000000000UL
+#define TCTRL1_RD_WAIT_SHIFT	      44
+#define TCTRL1_PC_CYCLE		0x00000fc000000000UL
+#define TCTRL1_PC_CYCLE_SHIFT	      38
+#define TCTRL1_WR_MORE_RAS_PW	0x0000003f00000000UL
+#define TCTRL1_WR_MORE_RAS_PW_SHIFT   32
+#define TCTRL1_RD_MORE_RAW_PW	0x00000000fc000000UL
+#define TCTRL1_RD_MORE_RAS_PW_SHIFT   26
+#define TCTRL1_ACT_WR_DLY	0x0000000003f00000UL
+#define TCTRL1_ACT_WR_DLY_SHIFT	      20
+#define TCTRL1_ACT_RD_DLY	0x00000000000fc000UL
+#define TCTRL1_ACT_RD_DLY_SHIFT	      14
+#define TCTRL1_BANK_PRESENT	0x0000000000003000UL
+#define TCTRL1_BANK_PRESENT_SHIFT     12
+#define TCTRL1_RFR_INT		0x0000000000000ff8UL
+#define TCTRL1_RFR_INT_SHIFT	      3
+#define TCTRL1_SET_MODE_REG	0x0000000000000004UL
+#define TCTRL1_SET_MODE_REG_SHIFT     2
+#define TCTRL1_RFR_ENABLE	0x0000000000000002UL
+#define TCTRL1_RFR_ENABLE_SHIFT	      1
+#define TCTRL1_PRECHG_ALL	0x0000000000000001UL
+#define TCTRL1_PRECHG_ALL_SHIFT	      0
+
+/* Memory Timing Control II */
+#define TCTRL2_WR_MSEL_DLY	0xfc00000000000000UL
+#define TCTRL2_WR_MSEL_DLY_SHIFT      58
+#define TCTRL2_RD_MSEL_DLY	0x03f0000000000000UL
+#define TCTRL2_RD_MSEL_DLY_SHIFT      52
+#define TCTRL2_WRDATA_THLD	0x000c000000000000UL
+#define TCTRL2_WRDATA_THLD_SHIFT      50
+#define TCTRL2_RDWR_RD_TI_DLY	0x0003f00000000000UL
+#define TCTRL2_RDWR_RD_TI_DLY_SHIFT   44
+#define TCTRL2_AUTOPRECHG_ENBL	0x0000080000000000UL
+#define TCTRL2_AUTOPRECHG_ENBL_SHIFT  43
+#define TCTRL2_RDWR_PI_MORE_DLY	0x000007c000000000UL
+#define TCTRL2_RDWR_PI_MORE_DLY_SHIFT 38
+#define TCTRL2_RDWR_1_DLY	0x0000003f00000000UL
+#define TCTRL2_RDWR_1_DLY_SHIFT       32
+#define TCTRL2_WRWR_PI_MORE_DLY	0x00000000f8000000UL
+#define TCTRL2_WRWR_PI_MORE_DLY_SHIFT 27
+#define TCTRL2_WRWR_1_DLY	0x0000000007e00000UL
+#define TCTRL2_WRWR_1_DLY_SHIFT       21
+#define TCTRL2_RDWR_RD_PI_MORE_DLY 0x00000000001f0000UL
+#define TCTRL2_RDWR_RD_PI_MORE_DLY_SHIFT 16
+#define TCTRL2_R		0x0000000000008000UL
+#define TCTRL2_R_SHIFT		      15
+#define TCTRL2_SDRAM_MODE_REG_DATA 0x0000000000007fffUL
+#define TCTRL2_SDRAM_MODE_REG_DATA_SHIFT 0
+
+/* Memory Timing Control III */
+#define TCTRL3_SDRAM_CTL_DLY	0xf000000000000000UL
+#define TCTRL3_SDRAM_CTL_DLY_SHIFT    60
+#define TCTRL3_SDRAM_CLK_DLY	0x0e00000000000000UL
+#define TCTRL3_SDRAM_CLK_DLY_SHIFT    57
+#define TCTRL3_R		0x0100000000000000UL
+#define TCTRL3_R_SHIFT		      56
+#define TCTRL3_AUTO_RFR_CYCLE	0x00fe000000000000UL
+#define TCTRL3_AUTO_RFR_CYCLE_SHIFT   49
+#define TCTRL3_RD_WAIT		0x0001f00000000000UL
+#define TCTRL3_RD_WAIT_SHIFT	      44
+#define TCTRL3_PC_CYCLE		0x00000fc000000000UL
+#define TCTRL3_PC_CYCLE_SHIFT	      38
+#define TCTRL3_WR_MORE_RAW_PW	0x0000003f00000000UL
+#define TCTRL3_WR_MORE_RAW_PW_SHIFT   32
+#define TCTRL3_RD_MORE_RAW_PW	0x00000000fc000000UL
+#define TCTRL3_RD_MORE_RAW_PW_SHIFT   26
+#define TCTRL3_ACT_WR_DLY	0x0000000003f00000UL
+#define TCTRL3_ACT_WR_DLY_SHIFT       20
+#define TCTRL3_ACT_RD_DLY	0x00000000000fc000UL
+#define TCTRL3_ACT_RD_DLY_SHIFT       14
+#define TCTRL3_BANK_PRESENT	0x0000000000003000UL
+#define TCTRL3_BANK_PRESENT_SHIFT     12
+#define TCTRL3_RFR_INT		0x0000000000000ff8UL
+#define TCTRL3_RFR_INT_SHIFT	      3
+#define TCTRL3_SET_MODE_REG	0x0000000000000004UL
+#define TCTRL3_SET_MODE_REG_SHIFT     2
+#define TCTRL3_RFR_ENABLE	0x0000000000000002UL
+#define TCTRL3_RFR_ENABLE_SHIFT       1
+#define TCTRL3_PRECHG_ALL	0x0000000000000001UL
+#define TCTRL3_PRECHG_ALL_SHIFT	      0
+
+/* Memory Timing Control IV */
+#define TCTRL4_WR_MSEL_DLY	0xfc00000000000000UL
+#define TCTRL4_WR_MSEL_DLY_SHIFT      58
+#define TCTRL4_RD_MSEL_DLY	0x03f0000000000000UL
+#define TCTRL4_RD_MSEL_DLY_SHIFT      52
+#define TCTRL4_WRDATA_THLD	0x000c000000000000UL
+#define TCTRL4_WRDATA_THLD_SHIFT      50
+#define TCTRL4_RDWR_RD_RI_DLY	0x0003f00000000000UL
+#define TCTRL4_RDWR_RD_RI_DLY_SHIFT   44
+#define TCTRL4_AUTO_PRECHG_ENBL	0x0000080000000000UL
+#define TCTRL4_AUTO_PRECHG_ENBL_SHIFT 43
+#define TCTRL4_RD_WR_PI_MORE_DLY 0x000007c000000000UL
+#define TCTRL4_RD_WR_PI_MORE_DLY_SHIFT 38
+#define TCTRL4_RD_WR_TI_DLY	0x0000003f00000000UL
+#define TCTRL4_RD_WR_TI_DLY_SHIFT     32
+#define TCTRL4_WR_WR_PI_MORE_DLY 0x00000000f8000000UL
+#define TCTRL4_WR_WR_PI_MORE_DLY_SHIFT 27
+#define TCTRL4_WR_WR_TI_DLY	0x0000000007e00000UL
+#define TCTRL4_WR_WR_TI_DLY_SHIFT     21
+#define TCTRL4_RDWR_RD_PI_MORE_DLY 0x00000000001f000UL0
+#define TCTRL4_RDWR_RD_PI_MORE_DLY_SHIFT 16
+#define TCTRL4_R		0x0000000000008000UL
+#define TCTRL4_R_SHIFT		      15
+#define TCTRL4_SDRAM_MODE_REG_DATA 0x0000000000007fffUL
+#define TCTRL4_SDRAM_MODE_REG_DATA_SHIFT 0
+
+/* All 4 memory address decoding registers have the
+ * same layout.
+ */
+#define MEM_DECODE_VALID	0x8000000000000000UL /* Valid */
+#define MEM_DECODE_VALID_SHIFT	      63
+#define MEM_DECODE_UK		0x001ffe0000000000UL /* Upper mask */
+#define MEM_DECODE_UK_SHIFT	      41
+#define MEM_DECODE_UM		0x0000001ffff00000UL /* Upper match */
+#define MEM_DECODE_UM_SHIFT	      20
+#define MEM_DECODE_LK		0x000000000003c000UL /* Lower mask */
+#define MEM_DECODE_LK_SHIFT	      14
+#define MEM_DECODE_LM		0x0000000000000f00UL /* Lower match */
+#define MEM_DECODE_LM_SHIFT           8
+
+#define PA_UPPER_BITS		0x000007fffc000000UL
+#define PA_UPPER_BITS_SHIFT	26
+#define PA_LOWER_BITS		0x00000000000003c0UL
+#define PA_LOWER_BITS_SHIFT	6
+
+#define MACTRL_R0		         0x8000000000000000UL
+#define MACTRL_R0_SHIFT		         63
+#define MACTRL_ADDR_LE_PW                0x7000000000000000UL
+#define MACTRL_ADDR_LE_PW_SHIFT		 60
+#define MACTRL_CMD_PW                    0x0f00000000000000UL
+#define MACTRL_CMD_PW_SHIFT		 56
+#define MACTRL_HALF_MODE_WR_MSEL_DLY     0x00fc000000000000UL
+#define MACTRL_HALF_MODE_WR_MSEL_DLY_SHIFT 50
+#define MACTRL_HALF_MODE_RD_MSEL_DLY     0x0003f00000000000UL
+#define MACTRL_HALF_MODE_RD_MSEL_DLY_SHIFT 44
+#define MACTRL_HALF_MODE_SDRAM_CTL_DLY   0x00000f0000000000UL
+#define MACTRL_HALF_MODE_SDRAM_CTL_DLY_SHIFT 40
+#define MACTRL_HALF_MODE_SDRAM_CLK_DLY   0x000000e000000000UL
+#define MACTRL_HALF_MODE_SDRAM_CLK_DLY_SHIFT 37
+#define MACTRL_R1                        0x0000001000000000UL
+#define MACTRL_R1_SHIFT                      36
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B3 0x0000000f00000000UL
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B3_SHIFT 32
+#define MACTRL_ENC_INTLV_B3              0x00000000f8000000UL
+#define MACTRL_ENC_INTLV_B3_SHIFT              27
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B2 0x0000000007800000UL
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B2_SHIFT 23
+#define MACTRL_ENC_INTLV_B2              0x00000000007c0000UL
+#define MACTRL_ENC_INTLV_B2_SHIFT              18
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B1 0x000000000003c000UL
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B1_SHIFT 14
+#define MACTRL_ENC_INTLV_B1              0x0000000000003e00UL
+#define MACTRL_ENC_INTLV_B1_SHIFT               9
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B0 0x00000000000001e0UL
+#define MACTRL_BANKSEL_N_ROWADDR_SIZE_B0_SHIFT  5
+#define MACTRL_ENC_INTLV_B0              0x000000000000001fUL
+#define MACTRL_ENC_INTLV_B0_SHIFT               0
+
+#endif /* _SPARC64_CHMCTRL_H */
diff --git a/include/asm-sparc/cmt.h b/include/asm-sparc/cmt.h
new file mode 100644
index 0000000..870db59
--- /dev/null
+++ b/include/asm-sparc/cmt.h
@@ -0,0 +1,59 @@
+#ifndef _SPARC64_CMT_H
+#define _SPARC64_CMT_H
+
+/* cmt.h: Chip Multi-Threading register definitions
+ *
+ * Copyright (C) 2004 David S. Miller (davem@redhat.com)
+ */
+
+/* ASI_CORE_ID - private */
+#define LP_ID		0x0000000000000010UL
+#define  LP_ID_MAX	0x00000000003f0000UL
+#define  LP_ID_ID	0x000000000000003fUL
+
+/* ASI_INTR_ID - private */
+#define LP_INTR_ID	0x0000000000000000UL
+#define  LP_INTR_ID_ID	0x00000000000003ffUL
+
+/* ASI_CESR_ID - private */
+#define CESR_ID		0x0000000000000040UL
+#define  CESR_ID_ID	0x00000000000000ffUL
+
+/* ASI_CORE_AVAILABLE - shared */
+#define LP_AVAIL	0x0000000000000000UL
+#define  LP_AVAIL_1	0x0000000000000002UL
+#define  LP_AVAIL_0	0x0000000000000001UL
+
+/* ASI_CORE_ENABLE_STATUS - shared */
+#define LP_ENAB_STAT	0x0000000000000010UL
+#define  LP_ENAB_STAT_1	0x0000000000000002UL
+#define  LP_ENAB_STAT_0	0x0000000000000001UL
+
+/* ASI_CORE_ENABLE - shared */
+#define LP_ENAB		0x0000000000000020UL
+#define  LP_ENAB_1	0x0000000000000002UL
+#define  LP_ENAB_0	0x0000000000000001UL
+
+/* ASI_CORE_RUNNING - shared */
+#define LP_RUNNING_RW	0x0000000000000050UL
+#define LP_RUNNING_W1S	0x0000000000000060UL
+#define LP_RUNNING_W1C	0x0000000000000068UL
+#define  LP_RUNNING_1	0x0000000000000002UL
+#define  LP_RUNNING_0	0x0000000000000001UL
+
+/* ASI_CORE_RUNNING_STAT - shared */
+#define LP_RUN_STAT	0x0000000000000058UL
+#define  LP_RUN_STAT_1	0x0000000000000002UL
+#define  LP_RUN_STAT_0	0x0000000000000001UL
+
+/* ASI_XIR_STEERING - shared */
+#define LP_XIR_STEER	0x0000000000000030UL
+#define  LP_XIR_STEER_1	0x0000000000000002UL
+#define  LP_XIR_STEER_0	0x0000000000000001UL
+
+/* ASI_CMT_ERROR_STEERING - shared */
+#define CMT_ER_STEER	0x0000000000000040UL
+#define  CMT_ER_STEER_1	0x0000000000000002UL
+#define  CMT_ER_STEER_0	0x0000000000000001UL
+
+#endif /* _SPARC64_CMT_H */
diff --git a/include/asm-sparc/compat.h b/include/asm-sparc/compat.h
new file mode 100644
index 0000000..f260b58
--- /dev/null
+++ b/include/asm-sparc/compat.h
@@ -0,0 +1,243 @@
+#ifndef _ASM_SPARC64_COMPAT_H
+#define _ASM_SPARC64_COMPAT_H
+/*
+ * Architecture specific compatibility types
+ */
+#include <linux/types.h>
+
+#define COMPAT_USER_HZ	100
+
+typedef u32		compat_size_t;
+typedef s32		compat_ssize_t;
+typedef s32		compat_time_t;
+typedef s32		compat_clock_t;
+typedef s32		compat_pid_t;
+typedef u16		__compat_uid_t;
+typedef u16		__compat_gid_t;
+typedef u32		__compat_uid32_t;
+typedef u32		__compat_gid32_t;
+typedef u16		compat_mode_t;
+typedef u32		compat_ino_t;
+typedef u16		compat_dev_t;
+typedef s32		compat_off_t;
+typedef s64		compat_loff_t;
+typedef s16		compat_nlink_t;
+typedef u16		compat_ipc_pid_t;
+typedef s32		compat_daddr_t;
+typedef u32		compat_caddr_t;
+typedef __kernel_fsid_t	compat_fsid_t;
+typedef s32		compat_key_t;
+typedef s32		compat_timer_t;
+
+typedef s32		compat_int_t;
+typedef s32		compat_long_t;
+typedef s64		compat_s64;
+typedef u32		compat_uint_t;
+typedef u32		compat_ulong_t;
+typedef u64		compat_u64;
+
+struct compat_timespec {
+	compat_time_t	tv_sec;
+	s32		tv_nsec;
+};
+
+struct compat_timeval {
+	compat_time_t	tv_sec;
+	s32		tv_usec;
+};
+
+struct compat_stat {
+	compat_dev_t	st_dev;
+	compat_ino_t	st_ino;
+	compat_mode_t	st_mode;
+	compat_nlink_t	st_nlink;
+	__compat_uid_t	st_uid;
+	__compat_gid_t	st_gid;
+	compat_dev_t	st_rdev;
+	compat_off_t	st_size;
+	compat_time_t	st_atime;
+	compat_ulong_t	st_atime_nsec;
+	compat_time_t	st_mtime;
+	compat_ulong_t	st_mtime_nsec;
+	compat_time_t	st_ctime;
+	compat_ulong_t	st_ctime_nsec;
+	compat_off_t	st_blksize;
+	compat_off_t	st_blocks;
+	u32		__unused4[2];
+};
+
+struct compat_stat64 {
+	unsigned long long	st_dev;
+
+	unsigned long long	st_ino;
+
+	unsigned int	st_mode;
+	unsigned int	st_nlink;
+
+	unsigned int	st_uid;
+	unsigned int	st_gid;
+
+	unsigned long long	st_rdev;
+
+	unsigned char	__pad3[8];
+
+	long long	st_size;
+	unsigned int	st_blksize;
+
+	unsigned char	__pad4[8];
+	unsigned int	st_blocks;
+
+	unsigned int	st_atime;
+	unsigned int	st_atime_nsec;
+
+	unsigned int	st_mtime;
+	unsigned int	st_mtime_nsec;
+
+	unsigned int	st_ctime;
+	unsigned int	st_ctime_nsec;
+
+	unsigned int	__unused4;
+	unsigned int	__unused5;
+};
+
+struct compat_flock {
+	short		l_type;
+	short		l_whence;
+	compat_off_t	l_start;
+	compat_off_t	l_len;
+	compat_pid_t	l_pid;
+	short		__unused;
+};
+
+#define F_GETLK64	12
+#define F_SETLK64	13
+#define F_SETLKW64	14
+
+struct compat_flock64 {
+	short		l_type;
+	short		l_whence;
+	compat_loff_t	l_start;
+	compat_loff_t	l_len;
+	compat_pid_t	l_pid;
+	short		__unused;
+};
+
+struct compat_statfs {
+	int		f_type;
+	int		f_bsize;
+	int		f_blocks;
+	int		f_bfree;
+	int		f_bavail;
+	int		f_files;
+	int		f_ffree;
+	compat_fsid_t	f_fsid;
+	int		f_namelen;	/* SunOS ignores this field. */
+	int		f_frsize;
+	int		f_spare[5];
+};
+
+#define COMPAT_RLIM_INFINITY 0x7fffffff
+
+typedef u32		compat_old_sigset_t;
+
+#define _COMPAT_NSIG		64
+#define _COMPAT_NSIG_BPW	32
+
+typedef u32		compat_sigset_word;
+
+#define COMPAT_OFF_T_MAX	0x7fffffff
+#define COMPAT_LOFF_T_MAX	0x7fffffffffffffffL
+
+/*
+ * A pointer passed in from user mode. This should not
+ * be used for syscall parameters, just declare them
+ * as pointers because the syscall entry code will have
+ * appropriately converted them already.
+ */
+typedef	u32		compat_uptr_t;
+
+static inline void __user *compat_ptr(compat_uptr_t uptr)
+{
+	return (void __user *)(unsigned long)uptr;
+}
+
+static inline compat_uptr_t ptr_to_compat(void __user *uptr)
+{
+	return (u32)(unsigned long)uptr;
+}
+
+static inline void __user *compat_alloc_user_space(long len)
+{
+	struct pt_regs *regs = current_thread_info()->kregs;
+	unsigned long usp = regs->u_regs[UREG_I6];
+
+	if (!(test_thread_flag(TIF_32BIT)))
+		usp += STACK_BIAS;
+	else
+		usp &= 0xffffffffUL;
+
+	usp -= len;
+	usp &= ~0x7UL;
+
+	return (void __user *) usp;
+}
+
+struct compat_ipc64_perm {
+	compat_key_t key;
+	__compat_uid32_t uid;
+	__compat_gid32_t gid;
+	__compat_uid32_t cuid;
+	__compat_gid32_t cgid;
+	unsigned short __pad1;
+	compat_mode_t mode;
+	unsigned short __pad2;
+	unsigned short seq;
+	unsigned long __unused1;	/* yes they really are 64bit pads */
+	unsigned long __unused2;
+};
+
+struct compat_semid64_ds {
+	struct compat_ipc64_perm sem_perm;
+	unsigned int	__pad1;
+	compat_time_t	sem_otime;
+	unsigned int	__pad2;
+	compat_time_t	sem_ctime;
+	u32		sem_nsems;
+	u32		__unused1;
+	u32		__unused2;
+};
+
+struct compat_msqid64_ds {
+	struct compat_ipc64_perm msg_perm;
+	unsigned int	__pad1;
+	compat_time_t	msg_stime;
+	unsigned int	__pad2;
+	compat_time_t	msg_rtime;
+	unsigned int	__pad3;
+	compat_time_t	msg_ctime;
+	unsigned int	msg_cbytes;
+	unsigned int	msg_qnum;
+	unsigned int	msg_qbytes;
+	compat_pid_t	msg_lspid;
+	compat_pid_t	msg_lrpid;
+	unsigned int	__unused1;
+	unsigned int	__unused2;
+};
+
+struct compat_shmid64_ds {
+	struct compat_ipc64_perm shm_perm;
+	unsigned int	__pad1;
+	compat_time_t	shm_atime;
+	unsigned int	__pad2;
+	compat_time_t	shm_dtime;
+	unsigned int	__pad3;
+	compat_time_t	shm_ctime;
+	compat_size_t	shm_segsz;
+	compat_pid_t	shm_cpid;
+	compat_pid_t	shm_lpid;
+	unsigned int	shm_nattch;
+	unsigned int	__unused1;
+	unsigned int	__unused2;
+};
+
+#endif /* _ASM_SPARC64_COMPAT_H */
diff --git a/include/asm-sparc/compat_signal.h b/include/asm-sparc/compat_signal.h
new file mode 100644
index 0000000..b759eab
--- /dev/null
+++ b/include/asm-sparc/compat_signal.h
@@ -0,0 +1,29 @@
+#ifndef _COMPAT_SIGNAL_H
+#define _COMPAT_SIGNAL_H
+
+#include <linux/compat.h>
+#include <asm/signal.h>
+
+#ifdef CONFIG_COMPAT
+struct __new_sigaction32 {
+	unsigned		sa_handler;
+	unsigned int    	sa_flags;
+	unsigned		sa_restorer;     /* not used by Linux/SPARC yet */
+	compat_sigset_t 	sa_mask;
+};
+
+struct __old_sigaction32 {
+	unsigned		sa_handler;
+	compat_old_sigset_t  	sa_mask;
+	unsigned int    	sa_flags;
+	unsigned		sa_restorer;     /* not used by Linux/SPARC yet */
+};
+
+typedef struct sigaltstack32 {
+	u32			ss_sp;
+	int			ss_flags;
+	compat_size_t		ss_size;
+} stack_t32;
+#endif
+
+#endif /* !(_COMPAT_SIGNAL_H) */
diff --git a/include/asm-sparc/dcr.h b/include/asm-sparc/dcr.h
new file mode 100644
index 0000000..620c9ba
--- /dev/null
+++ b/include/asm-sparc/dcr.h
@@ -0,0 +1,14 @@
+#ifndef _SPARC64_DCR_H
+#define _SPARC64_DCR_H
+
+/* UltraSparc-III/III+ Dispatch Control Register, ASR 0x12 */
+#define DCR_DPE		0x0000000000001000 /* III+: D$ Parity Error Enable	*/
+#define DCR_OBS		0x0000000000000fc0 /* Observability Bus Controls	*/
+#define DCR_BPE		0x0000000000000020 /* Branch Predict Enable		*/
+#define DCR_RPE		0x0000000000000010 /* Return Address Prediction Enable	*/
+#define DCR_SI		0x0000000000000008 /* Single Instruction Disable	*/
+#define DCR_IPE		0x0000000000000004 /* III+: I$ Parity Error Enable	*/
+#define DCR_IFPOE	0x0000000000000002 /* IRQ FP Operation Enable		*/
+#define DCR_MS		0x0000000000000001 /* Multi-Scalar dispatch		*/
+
+#endif /* _SPARC64_DCR_H */
diff --git a/include/asm-sparc/dcu.h b/include/asm-sparc/dcu.h
new file mode 100644
index 0000000..0f704e10
--- /dev/null
+++ b/include/asm-sparc/dcu.h
@@ -0,0 +1,27 @@
+#ifndef _SPARC64_DCU_H
+#define _SPARC64_DCU_H
+
+#include <linux/const.h>
+
+/* UltraSparc-III Data Cache Unit Control Register */
+#define DCU_CP	_AC(0x0002000000000000,UL) /* Phys Cache Enable w/o mmu	*/
+#define DCU_CV	_AC(0x0001000000000000,UL) /* Virt Cache Enable w/o mmu	*/
+#define DCU_ME	_AC(0x0000800000000000,UL) /* NC-store Merging Enable	*/
+#define DCU_RE	_AC(0x0000400000000000,UL) /* RAW bypass Enable		*/
+#define DCU_PE	_AC(0x0000200000000000,UL) /* PCache Enable		*/
+#define DCU_HPE	_AC(0x0000100000000000,UL) /* HW prefetch Enable	*/
+#define DCU_SPE	_AC(0x0000080000000000,UL) /* SW prefetch Enable	*/
+#define DCU_SL	_AC(0x0000040000000000,UL) /* Secondary ld-steering Enab*/
+#define DCU_WE	_AC(0x0000020000000000,UL) /* WCache enable		*/
+#define DCU_PM	_AC(0x000001fe00000000,UL) /* PA Watchpoint Byte Mask	*/
+#define DCU_VM	_AC(0x00000001fe000000,UL) /* VA Watchpoint Byte Mask	*/
+#define DCU_PR	_AC(0x0000000001000000,UL) /* PA Watchpoint Read Enable	*/
+#define DCU_PW	_AC(0x0000000000800000,UL) /* PA Watchpoint Write Enable*/
+#define DCU_VR	_AC(0x0000000000400000,UL) /* VA Watchpoint Read Enable	*/
+#define DCU_VW	_AC(0x0000000000200000,UL) /* VA Watchpoint Write Enable*/
+#define DCU_DM	_AC(0x0000000000000008,UL) /* DMMU Enable		*/
+#define DCU_IM	_AC(0x0000000000000004,UL) /* IMMU Enable		*/
+#define DCU_DC	_AC(0x0000000000000002,UL) /* Data Cache Enable		*/
+#define DCU_IC	_AC(0x0000000000000001,UL) /* Instruction Cache Enable	*/
+
+#endif /* _SPARC64_DCU_H */
diff --git a/include/asm-sparc/estate.h b/include/asm-sparc/estate.h
new file mode 100644
index 0000000..520c085
--- /dev/null
+++ b/include/asm-sparc/estate.h
@@ -0,0 +1,49 @@
+#ifndef _SPARC64_ESTATE_H
+#define _SPARC64_ESTATE_H
+
+/* UltraSPARC-III E-cache Error Enable */
+#define ESTATE_ERROR_FMT	0x0000000000040000 /* Force MTAG ECC		*/
+#define ESTATE_ERROR_FMESS	0x000000000003c000 /* Forced MTAG ECC val	*/
+#define ESTATE_ERROR_FMD	0x0000000000002000 /* Force DATA ECC		*/
+#define ESTATE_ERROR_FDECC	0x0000000000001ff0 /* Forced DATA ECC val	*/
+#define ESTATE_ERROR_UCEEN	0x0000000000000008 /* See below			*/
+#define ESTATE_ERROR_NCEEN	0x0000000000000002 /* See below			*/
+#define ESTATE_ERROR_CEEN	0x0000000000000001 /* See below			*/
+
+/* UCEEN enables the fast_ECC_error trap for: 1) software correctable E-cache
+ * errors 2) uncorrectable E-cache errors.  Such events only occur on reads
+ * of the E-cache by the local processor for: 1) data loads 2) instruction
+ * fetches 3) atomic operations.  Such events _cannot_ occur for: 1) merge
+ * 2) writeback 2) copyout.  The AFSR bits associated with these traps are
+ * UCC and UCU.
+ */
+
+/* NCEEN enables instruction_access_error, data_access_error, and ECC_error traps
+ * for uncorrectable ECC errors and system errors.
+ *
+ * Uncorrectable system bus data error or MTAG ECC error, system bus TimeOUT,
+ * or system bus BusERR:
+ * 1) As the result of an instruction fetch, will generate instruction_access_error
+ * 2) As the result of a load etc. will generate data_access_error.
+ * 3) As the result of store merge completion, writeback, or copyout will
+ *    generate a disrupting ECC_error trap.
+ * 4) As the result of such errors on instruction vector fetch can generate any
+ *    of the 3 trap types.
+ *
+ * The AFSR bits associated with these traps are EMU, EDU, WDU, CPU, IVU, UE,
+ * BERR, and TO.
+ */
+
+/* CEEN enables the ECC_error trap for hardware corrected ECC errors.  System bus
+ * reads resulting in a hardware corrected data or MTAG ECC error will generate an
+ * ECC_error disrupting trap with this bit enabled.
+ *
+ * This same trap will also be generated when a hardware corrected ECC error results
+ * during store merge, writeback, and copyout operations.
+ */
+
+/* In general, if the trap enable bits above are disabled the AFSR bits will still
+ * log the events even though the trap will not be generated by the processor.
+ */
+
+#endif /* _SPARC64_ESTATE_H */
diff --git a/include/asm-sparc/fhc.h b/include/asm-sparc/fhc.h
new file mode 100644
index 0000000..788cbc4
--- /dev/null
+++ b/include/asm-sparc/fhc.h
@@ -0,0 +1,121 @@
+/*
+ * fhc.h: Structures for central/fhc pseudo driver on Sunfire/Starfire/Wildfire.
+ *
+ * Copyright (C) 1997, 1999 David S. Miller (davem@redhat.com)
+ */
+
+#ifndef _SPARC64_FHC_H
+#define _SPARC64_FHC_H
+
+#include <linux/timer.h>
+
+#include <asm/oplib.h>
+#include <asm/prom.h>
+#include <asm/upa.h>
+
+struct linux_fhc;
+
+/* Clock board register offsets. */
+#define CLOCK_CTRL	0x00UL	/* Main control */
+#define CLOCK_STAT1	0x10UL	/* Status one */
+#define CLOCK_STAT2	0x20UL	/* Status two */
+#define CLOCK_PWRSTAT	0x30UL	/* Power status */
+#define CLOCK_PWRPRES	0x40UL	/* Power presence */
+#define CLOCK_TEMP	0x50UL	/* Temperature */
+#define CLOCK_IRQDIAG	0x60UL	/* IRQ diagnostics */
+#define CLOCK_PWRSTAT2	0x70UL	/* Power status two */
+
+#define CLOCK_CTRL_LLED		0x04	/* Left LED, 0 == on */
+#define CLOCK_CTRL_MLED		0x02	/* Mid LED, 1 == on */
+#define CLOCK_CTRL_RLED		0x01	/* RIght LED, 1 == on */
+
+struct linux_central {
+	struct linux_fhc		*child;
+	unsigned long			cfreg;
+	unsigned long			clkregs;
+	unsigned long			clkver;
+	int				slots;
+	struct device_node		*prom_node;
+
+	struct linux_prom_ranges	central_ranges[PROMREG_MAX];
+	int				num_central_ranges;
+};
+
+/* Firehose controller register offsets */
+struct fhc_regs {
+	unsigned long			pregs;	/* FHC internal regs */
+#define FHC_PREGS_ID	0x00UL	/* FHC ID */
+#define  FHC_ID_VERS		0xf0000000 /* Version of this FHC		*/
+#define  FHC_ID_PARTID		0x0ffff000 /* Part ID code (0x0f9f == FHC)	*/
+#define  FHC_ID_MANUF		0x0000007e /* Manufacturer (0x3e == SUN's JEDEC)*/
+#define  FHC_ID_RESV		0x00000001 /* Read as one			*/
+#define FHC_PREGS_RCS	0x10UL	/* FHC Reset Control/Status Register */
+#define  FHC_RCS_POR		0x80000000 /* Last reset was a power cycle	*/
+#define  FHC_RCS_SPOR		0x40000000 /* Last reset was sw power on reset	*/
+#define  FHC_RCS_SXIR		0x20000000 /* Last reset was sw XIR reset	*/
+#define  FHC_RCS_BPOR		0x10000000 /* Last reset was due to POR button	*/
+#define  FHC_RCS_BXIR		0x08000000 /* Last reset was due to XIR button	*/
+#define  FHC_RCS_WEVENT		0x04000000 /* CPU reset was due to wakeup event	*/
+#define  FHC_RCS_CFATAL		0x02000000 /* Centerplane Fatal Error signalled	*/
+#define  FHC_RCS_FENAB		0x01000000 /* Fatal errors elicit system reset	*/
+#define FHC_PREGS_CTRL	0x20UL	/* FHC Control Register */
+#define  FHC_CONTROL_ICS	0x00100000 /* Ignore Centerplane Signals	*/
+#define  FHC_CONTROL_FRST	0x00080000 /* Fatal Error Reset Enable		*/
+#define  FHC_CONTROL_LFAT	0x00040000 /* AC/DC signalled a local error	*/
+#define  FHC_CONTROL_SLINE	0x00010000 /* Firmware Synchronization Line	*/
+#define  FHC_CONTROL_DCD	0x00008000 /* DC-->DC Converter Disable		*/
+#define  FHC_CONTROL_POFF	0x00004000 /* AC/DC Controller PLL Disable	*/
+#define  FHC_CONTROL_FOFF	0x00002000 /* FHC Controller PLL Disable	*/
+#define  FHC_CONTROL_AOFF	0x00001000 /* CPU A SRAM/SBD Low Power Mode	*/
+#define  FHC_CONTROL_BOFF	0x00000800 /* CPU B SRAM/SBD Low Power Mode	*/
+#define  FHC_CONTROL_PSOFF	0x00000400 /* Turns off this FHC's power supply	*/
+#define  FHC_CONTROL_IXIST	0x00000200 /* 0=FHC tells clock board it exists	*/
+#define  FHC_CONTROL_XMSTR	0x00000100 /* 1=Causes this FHC to be XIR master*/
+#define  FHC_CONTROL_LLED	0x00000040 /* 0=Left LED ON			*/
+#define  FHC_CONTROL_MLED	0x00000020 /* 1=Middle LED ON			*/
+#define  FHC_CONTROL_RLED	0x00000010 /* 1=Right LED			*/
+#define  FHC_CONTROL_BPINS	0x00000003 /* Spare Bidirectional Pins		*/
+#define FHC_PREGS_BSR	0x30UL	/* FHC Board Status Register */
+#define  FHC_BSR_DA64		0x00040000 /* Port A: 0=128bit 1=64bit data path */
+#define  FHC_BSR_DB64		0x00020000 /* Port B: 0=128bit 1=64bit data path */
+#define  FHC_BSR_BID		0x0001e000 /* Board ID                           */
+#define  FHC_BSR_SA		0x00001c00 /* Port A UPA Speed (from the pins)   */
+#define  FHC_BSR_SB		0x00000380 /* Port B UPA Speed (from the pins)   */
+#define  FHC_BSR_NDIAG		0x00000040 /* Not in Diag Mode                   */
+#define  FHC_BSR_NTBED		0x00000020 /* Not in TestBED Mode                */
+#define  FHC_BSR_NIA		0x0000001c /* Jumper, bit 18 in PROM space       */
+#define  FHC_BSR_SI		0x00000001 /* Spare input pin value              */
+#define FHC_PREGS_ECC	0x40UL	/* FHC ECC Control Register (16 bits) */
+#define FHC_PREGS_JCTRL	0xf0UL	/* FHC JTAG Control Register */
+#define  FHC_JTAG_CTRL_MENAB	0x80000000 /* Indicates this is JTAG Master	 */
+#define  FHC_JTAG_CTRL_MNONE	0x40000000 /* Indicates no JTAG Master present	 */
+#define FHC_PREGS_JCMD	0x100UL	/* FHC JTAG Command Register */
+	unsigned long			ireg;	/* FHC IGN reg */
+#define FHC_IREG_IGN	0x00UL	/* This FHC's IGN */
+	unsigned long			ffregs;	/* FHC fanfail regs */
+#define FHC_FFREGS_IMAP	0x00UL	/* FHC Fanfail IMAP */
+#define FHC_FFREGS_ICLR	0x10UL	/* FHC Fanfail ICLR */
+	unsigned long			sregs;	/* FHC system regs */
+#define FHC_SREGS_IMAP	0x00UL	/* FHC System IMAP */
+#define FHC_SREGS_ICLR	0x10UL	/* FHC System ICLR */
+	unsigned long			uregs;	/* FHC uart regs */
+#define FHC_UREGS_IMAP	0x00UL	/* FHC Uart IMAP */
+#define FHC_UREGS_ICLR	0x10UL	/* FHC Uart ICLR */
+	unsigned long			tregs;	/* FHC TOD regs */
+#define FHC_TREGS_IMAP	0x00UL	/* FHC TOD IMAP */
+#define FHC_TREGS_ICLR	0x10UL	/* FHC TOD ICLR */
+};
+
+struct linux_fhc {
+	struct linux_fhc		*next;
+	struct linux_central		*parent;	/* NULL if not central FHC */
+	struct fhc_regs			fhc_regs;
+	int				board;
+	int				jtag_master;
+	struct device_node		*prom_node;
+
+	struct linux_prom_ranges	fhc_ranges[PROMREG_MAX];
+	int				num_fhc_ranges;
+};
+
+#endif /* !(_SPARC64_FHC_H) */
diff --git a/include/asm-sparc/fpumacro.h b/include/asm-sparc/fpumacro.h
new file mode 100644
index 0000000..cc463fe
--- /dev/null
+++ b/include/asm-sparc/fpumacro.h
@@ -0,0 +1,33 @@
+/* fpumacro.h: FPU related macros.
+ *
+ * Copyright (C) 1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
+ * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
+ */
+
+#ifndef _SPARC64_FPUMACRO_H
+#define _SPARC64_FPUMACRO_H
+
+#include <asm/asi.h>
+#include <asm/visasm.h>
+
+struct fpustate {
+	u32	regs[64];
+};
+
+#define FPUSTATE (struct fpustate *)(current_thread_info()->fpregs)
+
+static inline unsigned long fprs_read(void)
+{
+	unsigned long retval;
+
+	__asm__ __volatile__("rd %%fprs, %0" : "=r" (retval));
+
+	return retval;
+}
+
+static inline void fprs_write(unsigned long val)
+{
+	__asm__ __volatile__("wr %0, 0x0, %%fprs" : : "r" (val));
+}
+
+#endif /* !(_SPARC64_FPUMACRO_H) */
diff --git a/include/asm-sparc/hugetlb.h b/include/asm-sparc/hugetlb.h
new file mode 100644
index 0000000..412af58
--- /dev/null
+++ b/include/asm-sparc/hugetlb.h
@@ -0,0 +1,84 @@
+#ifndef _ASM_SPARC64_HUGETLB_H
+#define _ASM_SPARC64_HUGETLB_H
+
+#include <asm/page.h>
+
+
+void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
+		     pte_t *ptep, pte_t pte);
+
+pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
+			      pte_t *ptep);
+
+void hugetlb_prefault_arch_hook(struct mm_struct *mm);
+
+static inline int is_hugepage_only_range(struct mm_struct *mm,
+					 unsigned long addr,
+					 unsigned long len) {
+	return 0;
+}
+
+/*
+ * If the arch doesn't supply something else, assume that hugepage
+ * size aligned regions are ok without further preparation.
+ */
+static inline int prepare_hugepage_range(unsigned long addr, unsigned long len)
+{
+	if (len & ~HPAGE_MASK)
+		return -EINVAL;
+	if (addr & ~HPAGE_MASK)
+		return -EINVAL;
+	return 0;
+}
+
+static inline void hugetlb_free_pgd_range(struct mmu_gather **tlb,
+					  unsigned long addr, unsigned long end,
+					  unsigned long floor,
+					  unsigned long ceiling)
+{
+	free_pgd_range(tlb, addr, end, floor, ceiling);
+}
+
+static inline void huge_ptep_clear_flush(struct vm_area_struct *vma,
+					 unsigned long addr, pte_t *ptep)
+{
+}
+
+static inline int huge_pte_none(pte_t pte)
+{
+	return pte_none(pte);
+}
+
+static inline pte_t huge_pte_wrprotect(pte_t pte)
+{
+	return pte_wrprotect(pte);
+}
+
+static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
+					   unsigned long addr, pte_t *ptep)
+{
+	ptep_set_wrprotect(mm, addr, ptep);
+}
+
+static inline int huge_ptep_set_access_flags(struct vm_area_struct *vma,
+					     unsigned long addr, pte_t *ptep,
+					     pte_t pte, int dirty)
+{
+	return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
+}
+
+static inline pte_t huge_ptep_get(pte_t *ptep)
+{
+	return *ptep;
+}
+
+static inline int arch_prepare_hugepage(struct page *page)
+{
+	return 0;
+}
+
+static inline void arch_release_hugepage(struct page *page)
+{
+}
+
+#endif /* _ASM_SPARC64_HUGETLB_H */
diff --git a/include/asm-sparc/hvtramp.h b/include/asm-sparc/hvtramp.h
new file mode 100644
index 0000000..b2b9b94
--- /dev/null
+++ b/include/asm-sparc/hvtramp.h
@@ -0,0 +1,37 @@
+#ifndef _SPARC64_HVTRAP_H
+#define _SPARC64_HVTRAP_H
+
+#ifndef __ASSEMBLY__
+
+#include <linux/types.h>
+
+struct hvtramp_mapping {
+	__u64		vaddr;
+	__u64		tte;
+};
+
+struct hvtramp_descr {
+	__u32			cpu;
+	__u32			num_mappings;
+	__u64			fault_info_va;
+	__u64			fault_info_pa;
+	__u64			thread_reg;
+	struct hvtramp_mapping	maps[1];
+};
+
+extern void hv_cpu_startup(unsigned long hvdescr_pa);
+
+#endif
+
+#define HVTRAMP_DESCR_CPU		0x00
+#define HVTRAMP_DESCR_NUM_MAPPINGS	0x04
+#define HVTRAMP_DESCR_FAULT_INFO_VA	0x08
+#define HVTRAMP_DESCR_FAULT_INFO_PA	0x10
+#define HVTRAMP_DESCR_THREAD_REG	0x18
+#define HVTRAMP_DESCR_MAPS		0x20
+
+#define HVTRAMP_MAPPING_VADDR		0x00
+#define HVTRAMP_MAPPING_TTE		0x08
+#define HVTRAMP_MAPPING_SIZE		0x10
+
+#endif /* _SPARC64_HVTRAP_H */
diff --git a/include/asm-sparc/hypervisor.h b/include/asm-sparc/hypervisor.h
new file mode 100644
index 0000000..0c0f41c
--- /dev/null
+++ b/include/asm-sparc/hypervisor.h
@@ -0,0 +1,2945 @@
+#ifndef _SPARC64_HYPERVISOR_H
+#define _SPARC64_HYPERVISOR_H
+
+/* Sun4v hypervisor interfaces and defines.
+ *
+ * Hypervisor calls are made via traps to software traps number 0x80
+ * and above.  Registers %o0 to %o5 serve as argument, status, and
+ * return value registers.
+ *
+ * There are two kinds of these traps.  First there are the normal
+ * "fast traps" which use software trap 0x80 and encode the function
+ * to invoke by number in register %o5.  Argument and return value
+ * handling is as follows:
+ *
+ * -----------------------------------------------
+ * |  %o5  | function number |     undefined     |
+ * |  %o0  |   argument 0    |   return status   |
+ * |  %o1  |   argument 1    |   return value 1  |
+ * |  %o2  |   argument 2    |   return value 2  |
+ * |  %o3  |   argument 3    |   return value 3  |
+ * |  %o4  |   argument 4    |   return value 4  |
+ * -----------------------------------------------
+ *
+ * The second type are "hyper-fast traps" which encode the function
+ * number in the software trap number itself.  So these use trap
+ * numbers > 0x80.  The register usage for hyper-fast traps is as
+ * follows:
+ *
+ * -----------------------------------------------
+ * |  %o0  |   argument 0    |   return status   |
+ * |  %o1  |   argument 1    |   return value 1  |
+ * |  %o2  |   argument 2    |   return value 2  |
+ * |  %o3  |   argument 3    |   return value 3  |
+ * |  %o4  |   argument 4    |   return value 4  |
+ * -----------------------------------------------
+ *
+ * Registers providing explicit arguments to the hypervisor calls
+ * are volatile across the call.  Upon return their values are
+ * undefined unless explicitly specified as containing a particular
+ * return value by the specific call.  The return status is always
+ * returned in register %o0, zero indicates a successful execution of
+ * the hypervisor call and other values indicate an error status as
+ * defined below.  So, for example, if a hyper-fast trap takes
+ * arguments 0, 1, and 2, then %o0, %o1, and %o2 are volatile across
+ * the call and %o3, %o4, and %o5 would be preserved.
+ *
+ * If the hypervisor trap is invalid, or the fast trap function number
+ * is invalid, HV_EBADTRAP will be returned in %o0.  Also, all 64-bits
+ * of the argument and return values are significant.
+ */
+
+/* Trap numbers.  */
+#define HV_FAST_TRAP		0x80
+#define HV_MMU_MAP_ADDR_TRAP	0x83
+#define HV_MMU_UNMAP_ADDR_TRAP	0x84
+#define HV_TTRACE_ADDENTRY_TRAP	0x85
+#define HV_CORE_TRAP		0xff
+
+/* Error codes.  */
+#define HV_EOK				0  /* Successful return            */
+#define HV_ENOCPU			1  /* Invalid CPU id               */
+#define HV_ENORADDR			2  /* Invalid real address         */
+#define HV_ENOINTR			3  /* Invalid interrupt id         */
+#define HV_EBADPGSZ			4  /* Invalid pagesize encoding    */
+#define HV_EBADTSB			5  /* Invalid TSB description      */
+#define HV_EINVAL			6  /* Invalid argument             */
+#define HV_EBADTRAP			7  /* Invalid function number      */
+#define HV_EBADALIGN			8  /* Invalid address alignment    */
+#define HV_EWOULDBLOCK			9  /* Cannot complete w/o blocking */
+#define HV_ENOACCESS			10 /* No access to resource        */
+#define HV_EIO				11 /* I/O error                    */
+#define HV_ECPUERROR			12 /* CPU in error state           */
+#define HV_ENOTSUPPORTED		13 /* Function not supported       */
+#define HV_ENOMAP			14 /* No mapping found             */
+#define HV_ETOOMANY			15 /* Too many items specified     */
+#define HV_ECHANNEL			16 /* Invalid LDC channel          */
+#define HV_EBUSY			17 /* Resource busy                */
+
+/* mach_exit()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_EXIT
+ * ARG0:	exit code
+ * ERRORS:	This service does not return.
+ *
+ * Stop all CPUs in the virtual domain and place them into the stopped
+ * state.  The 64-bit exit code may be passed to a service entity as
+ * the domain's exit status.  On systems without a service entity, the
+ * domain will undergo a reset, and the boot firmware will be
+ * reloaded.
+ *
+ * This function will never return to the guest that invokes it.
+ *
+ * Note: By convention an exit code of zero denotes a successful exit by
+ *       the guest code.  A non-zero exit code denotes a guest specific
+ *       error indication.
+ *
+ */
+#define HV_FAST_MACH_EXIT		0x00
+
+#ifndef __ASSEMBLY__
+extern void sun4v_mach_exit(unsigned long exit_code);
+#endif
+
+/* Domain services.  */
+
+/* mach_desc()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_DESC
+ * ARG0:	buffer
+ * ARG1:	length
+ * RET0:	status
+ * RET1:	length
+ * ERRORS:	HV_EBADALIGN	Buffer is badly aligned
+ *		HV_ENORADDR	Buffer is to an illegal real address.
+ *		HV_EINVAL	Buffer length is too small for complete
+ *				machine description.
+ *
+ * Copy the most current machine description into the buffer indicated
+ * by the real address in ARG0.  The buffer provided must be 16 byte
+ * aligned.  Upon success or HV_EINVAL, this service returns the
+ * actual size of the machine description in the RET1 return value.
+ *
+ * Note: A method of determining the appropriate buffer size for the
+ *       machine description is to first call this service with a buffer
+ *       length of 0 bytes.
+ */
+#define HV_FAST_MACH_DESC		0x01
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_mach_desc(unsigned long buffer_pa,
+				     unsigned long buf_len,
+				     unsigned long *real_buf_len);
+#endif
+
+/* mach_sir()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_SIR
+ * ERRORS:	This service does not return.
+ *
+ * Perform a software initiated reset of the virtual machine domain.
+ * All CPUs are captured as soon as possible, all hardware devices are
+ * returned to the entry default state, and the domain is restarted at
+ * the SIR (trap type 0x04) real trap table (RTBA) entry point on one
+ * of the CPUs.  The single CPU restarted is selected as determined by
+ * platform specific policy.  Memory is preserved across this
+ * operation.
+ */
+#define HV_FAST_MACH_SIR		0x02
+
+#ifndef __ASSEMBLY__
+extern void sun4v_mach_sir(void);
+#endif
+
+/* mach_set_watchdog()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_SET_WATCHDOG
+ * ARG0:	timeout in milliseconds
+ * RET0:	status
+ * RET1:	time remaining in milliseconds
+ *
+ * A guest uses this API to set a watchdog timer.  Once the gues has set
+ * the timer, it must call the timer service again either to disable or
+ * postpone the expiration.  If the timer expires before being reset or
+ * disabled, then the hypervisor take a platform specific action leading
+ * to guest termination within a bounded time period.  The platform action
+ * may include recovery actions such as reporting the expiration to a
+ * Service Processor, and/or automatically restarting the gues.
+ *
+ * The 'timeout' parameter is specified in milliseconds, however the
+ * implementated granularity is given by the 'watchdog-resolution'
+ * property in the 'platform' node of the guest's machine description.
+ * The largest allowed timeout value is specified by the
+ * 'watchdog-max-timeout' property of the 'platform' node.
+ *
+ * If the 'timeout' argument is not zero, the watchdog timer is set to
+ * expire after a minimum of 'timeout' milliseconds.
+ *
+ * If the 'timeout' argument is zero, the watchdog timer is disabled.
+ *
+ * If the 'timeout' value exceeds the value of the 'max-watchdog-timeout'
+ * property, the hypervisor leaves the watchdog timer state unchanged,
+ * and returns a status of EINVAL.
+ *
+ * The 'time remaining' return value is valid regardless of whether the
+ * return status is EOK or EINVAL.  A non-zero return value indicates the
+ * number of milliseconds that were remaining until the timer was to expire.
+ * If less than one millisecond remains, the return value is '1'.  If the
+ * watchdog timer was disabled at the time of the call, the return value is
+ * zero.
+ *
+ * If the hypervisor cannot support the exact timeout value requested, but
+ * can support a larger timeout value, the hypervisor may round the actual
+ * timeout to a value larger than the requested timeout, consequently the
+ * 'time remaining' return value may be larger than the previously requested
+ * timeout value.
+ *
+ * Any guest OS debugger should be aware that the watchdog service may be in
+ * use.  Consequently, it is recommended that the watchdog service is
+ * disabled upon debugger entry (e.g. reaching a breakpoint), and then
+ * re-enabled upon returning to normal execution.  The API has been designed
+ * with this in mind, and the 'time remaining' result of the disable call may
+ * be used directly as the timeout argument of the re-enable call.
+ */
+#define HV_FAST_MACH_SET_WATCHDOG	0x05
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_mach_set_watchdog(unsigned long timeout,
+					     unsigned long *orig_timeout);
+#endif
+
+/* CPU services.
+ *
+ * CPUs represent devices that can execute software threads.  A single
+ * chip that contains multiple cores or strands is represented as
+ * multiple CPUs with unique CPU identifiers.  CPUs are exported to
+ * OBP via the machine description (and to the OS via the OBP device
+ * tree).  CPUs are always in one of three states: stopped, running,
+ * or error.
+ *
+ * A CPU ID is a pre-assigned 16-bit value that uniquely identifies a
+ * CPU within a logical domain.  Operations that are to be performed
+ * on multiple CPUs specify them via a CPU list.  A CPU list is an
+ * array in real memory, of which each 16-bit word is a CPU ID.  CPU
+ * lists are passed through the API as two arguments.  The first is
+ * the number of entries (16-bit words) in the CPU list, and the
+ * second is the (real address) pointer to the CPU ID list.
+ */
+
+/* cpu_start()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_START
+ * ARG0:	CPU ID
+ * ARG1:	PC
+ * ARG2:	RTBA
+ * ARG3:	target ARG0
+ * RET0:	status
+ * ERRORS:	ENOCPU		Invalid CPU ID
+ *		EINVAL		Target CPU ID is not in the stopped state
+ *		ENORADDR	Invalid PC or RTBA real address
+ *		EBADALIGN	Unaligned PC or unaligned RTBA
+ *		EWOULDBLOCK	Starting resources are not available
+ *
+ * Start CPU with given CPU ID with PC in %pc and with a real trap
+ * base address value of RTBA.  The indicated CPU must be in the
+ * stopped state.  The supplied RTBA must be aligned on a 256 byte
+ * boundary.  On successful completion, the specified CPU will be in
+ * the running state and will be supplied with "target ARG0" in %o0
+ * and RTBA in %tba.
+ */
+#define HV_FAST_CPU_START		0x10
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_cpu_start(unsigned long cpuid,
+				     unsigned long pc,
+				     unsigned long rtba,
+				     unsigned long arg0);
+#endif
+
+/* cpu_stop()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_STOP
+ * ARG0:	CPU ID
+ * RET0:	status
+ * ERRORS:	ENOCPU		Invalid CPU ID
+ *		EINVAL		Target CPU ID is the current cpu
+ *		EINVAL		Target CPU ID is not in the running state
+ *		EWOULDBLOCK	Stopping resources are not available
+ *		ENOTSUPPORTED	Not supported on this platform
+ *
+ * The specified CPU is stopped.  The indicated CPU must be in the
+ * running state.  On completion, it will be in the stopped state.  It
+ * is not legal to stop the current CPU.
+ *
+ * Note: As this service cannot be used to stop the current cpu, this service
+ *       may not be used to stop the last running CPU in a domain.  To stop
+ *       and exit a running domain, a guest must use the mach_exit() service.
+ */
+#define HV_FAST_CPU_STOP		0x11
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_cpu_stop(unsigned long cpuid);
+#endif
+
+/* cpu_yield()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_YIELD
+ * RET0:	status
+ * ERRORS:	No possible error.
+ *
+ * Suspend execution on the current CPU.  Execution will resume when
+ * an interrupt (device, %stick_compare, or cross-call) is targeted to
+ * the CPU.  On some CPUs, this API may be used by the hypervisor to
+ * save power by disabling hardware strands.
+ */
+#define HV_FAST_CPU_YIELD		0x12
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_cpu_yield(void);
+#endif
+
+/* cpu_qconf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_QCONF
+ * ARG0:	queue
+ * ARG1:	base real address
+ * ARG2:	number of entries
+ * RET0:	status
+ * ERRORS:	ENORADDR	Invalid base real address
+ *		EINVAL		Invalid queue or number of entries is less
+ *				than 2 or too large.
+ *		EBADALIGN	Base real address is not correctly aligned
+ *				for size.
+ *
+ * Configure the given queue to be placed at the given base real
+ * address, with the given number of entries.  The number of entries
+ * must be a power of 2.  The base real address must be aligned
+ * exactly to match the queue size.  Each queue entry is 64 bytes
+ * long, so for example a 32 entry queue must be aligned on a 2048
+ * byte real address boundary.
+ *
+ * The specified queue is unconfigured if the number of entries is given
+ * as zero.
+ *
+ * For the current version of this API service, the argument queue is defined
+ * as follows:
+ *
+ *	queue		description
+ *	-----		-------------------------
+ *	0x3c		cpu mondo queue
+ *	0x3d		device mondo queue
+ *	0x3e		resumable error queue
+ *	0x3f		non-resumable error queue
+ *
+ * Note: The maximum number of entries for each queue for a specific cpu may
+ *       be determined from the machine description.
+ */
+#define HV_FAST_CPU_QCONF		0x14
+#define  HV_CPU_QUEUE_CPU_MONDO		 0x3c
+#define  HV_CPU_QUEUE_DEVICE_MONDO	 0x3d
+#define  HV_CPU_QUEUE_RES_ERROR		 0x3e
+#define  HV_CPU_QUEUE_NONRES_ERROR	 0x3f
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_cpu_qconf(unsigned long type,
+				     unsigned long queue_paddr,
+				     unsigned long num_queue_entries);
+#endif
+
+/* cpu_qinfo()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_QINFO
+ * ARG0:	queue
+ * RET0:	status
+ * RET1:	base real address
+ * RET1:	number of entries
+ * ERRORS:	EINVAL		Invalid queue
+ *
+ * Return the configuration info for the given queue.  The base real
+ * address and number of entries of the defined queue are returned.
+ * The queue argument values are the same as for cpu_qconf() above.
+ *
+ * If the specified queue is a valid queue number, but no queue has
+ * been defined, the number of entries will be set to zero and the
+ * base real address returned is undefined.
+ */
+#define HV_FAST_CPU_QINFO		0x15
+
+/* cpu_mondo_send()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_MONDO_SEND
+ * ARG0-1:	CPU list
+ * ARG2:	data real address
+ * RET0:	status
+ * ERRORS:	EBADALIGN	Mondo data is not 64-byte aligned or CPU list
+ *				is not 2-byte aligned.
+ *		ENORADDR	Invalid data mondo address, or invalid cpu list
+ *				address.
+ *		ENOCPU		Invalid cpu in CPU list
+ *		EWOULDBLOCK	Some or all of the listed CPUs did not receive
+ *				the mondo
+ *		ECPUERROR	One or more of the listed CPUs are in error
+ *				state, use HV_FAST_CPU_STATE to see which ones
+ *		EINVAL		CPU list includes caller's CPU ID
+ *
+ * Send a mondo interrupt to the CPUs in the given CPU list with the
+ * 64-bytes at the given data real address.  The data must be 64-byte
+ * aligned.  The mondo data will be delivered to the cpu_mondo queues
+ * of the recipient CPUs.
+ *
+ * In all cases, error or not, the CPUs in the CPU list to which the
+ * mondo has been successfully delivered will be indicated by having
+ * their entry in CPU list updated with the value 0xffff.
+ */
+#define HV_FAST_CPU_MONDO_SEND		0x42
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_cpu_mondo_send(unsigned long cpu_count, unsigned long cpu_list_pa, unsigned long mondo_block_pa);
+#endif
+
+/* cpu_myid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_MYID
+ * RET0:	status
+ * RET1:	CPU ID
+ * ERRORS:	No errors defined.
+ *
+ * Return the hypervisor ID handle for the current CPU.  Use by a
+ * virtual CPU to discover it's own identity.
+ */
+#define HV_FAST_CPU_MYID		0x16
+
+/* cpu_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_STATE
+ * ARG0:	CPU ID
+ * RET0:	status
+ * RET1:	state
+ * ERRORS:	ENOCPU		Invalid CPU ID
+ *
+ * Retrieve the current state of the CPU with the given CPU ID.
+ */
+#define HV_FAST_CPU_STATE		0x17
+#define  HV_CPU_STATE_STOPPED		 0x01
+#define  HV_CPU_STATE_RUNNING		 0x02
+#define  HV_CPU_STATE_ERROR		 0x03
+
+#ifndef __ASSEMBLY__
+extern long sun4v_cpu_state(unsigned long cpuid);
+#endif
+
+/* cpu_set_rtba()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_SET_RTBA
+ * ARG0:	RTBA
+ * RET0:	status
+ * RET1:	previous RTBA
+ * ERRORS:	ENORADDR	Invalid RTBA real address
+ *		EBADALIGN	RTBA is incorrectly aligned for a trap table
+ *
+ * Set the real trap base address of the local cpu to the given RTBA.
+ * The supplied RTBA must be aligned on a 256 byte boundary.  Upon
+ * success the previous value of the RTBA is returned in RET1.
+ *
+ * Note: This service does not affect %tba
+ */
+#define HV_FAST_CPU_SET_RTBA		0x18
+
+/* cpu_set_rtba()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CPU_GET_RTBA
+ * RET0:	status
+ * RET1:	previous RTBA
+ * ERRORS:	No possible error.
+ *
+ * Returns the current value of RTBA in RET1.
+ */
+#define HV_FAST_CPU_GET_RTBA		0x19
+
+/* MMU services.
+ *
+ * Layout of a TSB description for mmu_tsb_ctx{,non}0() calls.
+ */
+#ifndef __ASSEMBLY__
+struct hv_tsb_descr {
+	unsigned short		pgsz_idx;
+	unsigned short		assoc;
+	unsigned int		num_ttes;	/* in TTEs */
+	unsigned int		ctx_idx;
+	unsigned int		pgsz_mask;
+	unsigned long		tsb_base;
+	unsigned long		resv;
+};
+#endif
+#define HV_TSB_DESCR_PGSZ_IDX_OFFSET	0x00
+#define HV_TSB_DESCR_ASSOC_OFFSET	0x02
+#define HV_TSB_DESCR_NUM_TTES_OFFSET	0x04
+#define HV_TSB_DESCR_CTX_IDX_OFFSET	0x08
+#define HV_TSB_DESCR_PGSZ_MASK_OFFSET	0x0c
+#define HV_TSB_DESCR_TSB_BASE_OFFSET	0x10
+#define HV_TSB_DESCR_RESV_OFFSET	0x18
+
+/* Page size bitmask.  */
+#define HV_PGSZ_MASK_8K			(1 << 0)
+#define HV_PGSZ_MASK_64K		(1 << 1)
+#define HV_PGSZ_MASK_512K		(1 << 2)
+#define HV_PGSZ_MASK_4MB		(1 << 3)
+#define HV_PGSZ_MASK_32MB		(1 << 4)
+#define HV_PGSZ_MASK_256MB		(1 << 5)
+#define HV_PGSZ_MASK_2GB		(1 << 6)
+#define HV_PGSZ_MASK_16GB		(1 << 7)
+
+/* Page size index.  The value given in the TSB descriptor must correspond
+ * to the smallest page size specified in the pgsz_mask page size bitmask.
+ */
+#define HV_PGSZ_IDX_8K			0
+#define HV_PGSZ_IDX_64K			1
+#define HV_PGSZ_IDX_512K		2
+#define HV_PGSZ_IDX_4MB			3
+#define HV_PGSZ_IDX_32MB		4
+#define HV_PGSZ_IDX_256MB		5
+#define HV_PGSZ_IDX_2GB			6
+#define HV_PGSZ_IDX_16GB		7
+
+/* MMU fault status area.
+ *
+ * MMU related faults have their status and fault address information
+ * placed into a memory region made available by privileged code.  Each
+ * virtual processor must make a mmu_fault_area_conf() call to tell the
+ * hypervisor where that processor's fault status should be stored.
+ *
+ * The fault status block is a multiple of 64-bytes and must be aligned
+ * on a 64-byte boundary.
+ */
+#ifndef __ASSEMBLY__
+struct hv_fault_status {
+	unsigned long		i_fault_type;
+	unsigned long		i_fault_addr;
+	unsigned long		i_fault_ctx;
+	unsigned long		i_reserved[5];
+	unsigned long		d_fault_type;
+	unsigned long		d_fault_addr;
+	unsigned long		d_fault_ctx;
+	unsigned long		d_reserved[5];
+};
+#endif
+#define HV_FAULT_I_TYPE_OFFSET	0x00
+#define HV_FAULT_I_ADDR_OFFSET	0x08
+#define HV_FAULT_I_CTX_OFFSET	0x10
+#define HV_FAULT_D_TYPE_OFFSET	0x40
+#define HV_FAULT_D_ADDR_OFFSET	0x48
+#define HV_FAULT_D_CTX_OFFSET	0x50
+
+#define HV_FAULT_TYPE_FAST_MISS	1
+#define HV_FAULT_TYPE_FAST_PROT	2
+#define HV_FAULT_TYPE_MMU_MISS	3
+#define HV_FAULT_TYPE_INV_RA	4
+#define HV_FAULT_TYPE_PRIV_VIOL	5
+#define HV_FAULT_TYPE_PROT_VIOL	6
+#define HV_FAULT_TYPE_NFO	7
+#define HV_FAULT_TYPE_NFO_SEFF	8
+#define HV_FAULT_TYPE_INV_VA	9
+#define HV_FAULT_TYPE_INV_ASI	10
+#define HV_FAULT_TYPE_NC_ATOMIC	11
+#define HV_FAULT_TYPE_PRIV_ACT	12
+#define HV_FAULT_TYPE_RESV1	13
+#define HV_FAULT_TYPE_UNALIGNED	14
+#define HV_FAULT_TYPE_INV_PGSZ	15
+/* Values 16 --> -2 are reserved.  */
+#define HV_FAULT_TYPE_MULTIPLE	-1
+
+/* Flags argument for mmu_{map,unmap}_addr(), mmu_demap_{page,context,all}(),
+ * and mmu_{map,unmap}_perm_addr().
+ */
+#define HV_MMU_DMMU			0x01
+#define HV_MMU_IMMU			0x02
+#define HV_MMU_ALL			(HV_MMU_DMMU | HV_MMU_IMMU)
+
+/* mmu_map_addr()
+ * TRAP:	HV_MMU_MAP_ADDR_TRAP
+ * ARG0:	virtual address
+ * ARG1:	mmu context
+ * ARG2:	TTE
+ * ARG3:	flags (HV_MMU_{IMMU,DMMU})
+ * ERRORS:	EINVAL		Invalid virtual address, mmu context, or flags
+ *		EBADPGSZ	Invalid page size value
+ *		ENORADDR	Invalid real address in TTE
+ *
+ * Create a non-permanent mapping using the given TTE, virtual
+ * address, and mmu context.  The flags argument determines which
+ * (data, or instruction, or both) TLB the mapping gets loaded into.
+ *
+ * The behavior is undefined if the valid bit is clear in the TTE.
+ *
+ * Note: This API call is for privileged code to specify temporary translation
+ *       mappings without the need to create and manage a TSB.
+ */
+
+/* mmu_unmap_addr()
+ * TRAP:	HV_MMU_UNMAP_ADDR_TRAP
+ * ARG0:	virtual address
+ * ARG1:	mmu context
+ * ARG2:	flags (HV_MMU_{IMMU,DMMU})
+ * ERRORS:	EINVAL		Invalid virtual address, mmu context, or flags
+ *
+ * Demaps the given virtual address in the given mmu context on this
+ * CPU.  This function is intended to be used to demap pages mapped
+ * with mmu_map_addr.  This service is equivalent to invoking
+ * mmu_demap_page() with only the current CPU in the CPU list. The
+ * flags argument determines which (data, or instruction, or both) TLB
+ * the mapping gets unmapped from.
+ *
+ * Attempting to perform an unmap operation for a previously defined
+ * permanent mapping will have undefined results.
+ */
+
+/* mmu_tsb_ctx0()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_TSB_CTX0
+ * ARG0:	number of TSB descriptions
+ * ARG1:	TSB descriptions pointer
+ * RET0:	status
+ * ERRORS:	ENORADDR		Invalid TSB descriptions pointer or
+ *					TSB base within a descriptor
+ *		EBADALIGN		TSB descriptions pointer is not aligned
+ *					to an 8-byte boundary, or TSB base
+ *					within a descriptor is not aligned for
+ *					the given TSB size
+ *		EBADPGSZ		Invalid page size in a TSB descriptor
+ *		EBADTSB			Invalid associativity or size in a TSB
+ *					descriptor
+ *		EINVAL			Invalid number of TSB descriptions, or
+ *					invalid context index in a TSB
+ *					descriptor, or index page size not
+ *					equal to smallest page size in page
+ *					size bitmask field.
+ *
+ * Configures the TSBs for the current CPU for virtual addresses with
+ * context zero.  The TSB descriptions pointer is a pointer to an
+ * array of the given number of TSB descriptions.
+ *
+ * Note: The maximum number of TSBs available to a virtual CPU is given by the
+ *       mmu-max-#tsbs property of the cpu's corresponding "cpu" node in the
+ *       machine description.
+ */
+#define HV_FAST_MMU_TSB_CTX0		0x20
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_mmu_tsb_ctx0(unsigned long num_descriptions,
+					unsigned long tsb_desc_ra);
+#endif
+
+/* mmu_tsb_ctxnon0()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_TSB_CTXNON0
+ * ARG0:	number of TSB descriptions
+ * ARG1:	TSB descriptions pointer
+ * RET0:	status
+ * ERRORS:	Same as for mmu_tsb_ctx0() above.
+ *
+ * Configures the TSBs for the current CPU for virtual addresses with
+ * non-zero contexts.  The TSB descriptions pointer is a pointer to an
+ * array of the given number of TSB descriptions.
+ *
+ * Note: A maximum of 16 TSBs may be specified in the TSB description list.
+ */
+#define HV_FAST_MMU_TSB_CTXNON0		0x21
+
+/* mmu_demap_page()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_DEMAP_PAGE
+ * ARG0:	reserved, must be zero
+ * ARG1:	reserved, must be zero
+ * ARG2:	virtual address
+ * ARG3:	mmu context
+ * ARG4:	flags (HV_MMU_{IMMU,DMMU})
+ * RET0:	status
+ * ERRORS:	EINVAL			Invalid virutal address, context, or
+ *					flags value
+ *		ENOTSUPPORTED		ARG0 or ARG1 is non-zero
+ *
+ * Demaps any page mapping of the given virtual address in the given
+ * mmu context for the current virtual CPU.  Any virtually tagged
+ * caches are guaranteed to be kept consistent.  The flags argument
+ * determines which TLB (instruction, or data, or both) participate in
+ * the operation.
+ *
+ * ARG0 and ARG1 are both reserved and must be set to zero.
+ */
+#define HV_FAST_MMU_DEMAP_PAGE		0x22
+
+/* mmu_demap_ctx()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_DEMAP_CTX
+ * ARG0:	reserved, must be zero
+ * ARG1:	reserved, must be zero
+ * ARG2:	mmu context
+ * ARG3:	flags (HV_MMU_{IMMU,DMMU})
+ * RET0:	status
+ * ERRORS:	EINVAL			Invalid context or flags value
+ *		ENOTSUPPORTED		ARG0 or ARG1 is non-zero
+ *
+ * Demaps all non-permanent virtual page mappings previously specified
+ * for the given context for the current virtual CPU.  Any virtual
+ * tagged caches are guaranteed to be kept consistent.  The flags
+ * argument determines which TLB (instruction, or data, or both)
+ * participate in the operation.
+ *
+ * ARG0 and ARG1 are both reserved and must be set to zero.
+ */
+#define HV_FAST_MMU_DEMAP_CTX		0x23
+
+/* mmu_demap_all()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_DEMAP_ALL
+ * ARG0:	reserved, must be zero
+ * ARG1:	reserved, must be zero
+ * ARG2:	flags (HV_MMU_{IMMU,DMMU})
+ * RET0:	status
+ * ERRORS:	EINVAL			Invalid flags value
+ *		ENOTSUPPORTED		ARG0 or ARG1 is non-zero
+ *
+ * Demaps all non-permanent virtual page mappings previously specified
+ * for the current virtual CPU.  Any virtual tagged caches are
+ * guaranteed to be kept consistent.  The flags argument determines
+ * which TLB (instruction, or data, or both) participate in the
+ * operation.
+ *
+ * ARG0 and ARG1 are both reserved and must be set to zero.
+ */
+#define HV_FAST_MMU_DEMAP_ALL		0x24
+
+#ifndef __ASSEMBLY__
+extern void sun4v_mmu_demap_all(void);
+#endif
+
+/* mmu_map_perm_addr()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_MAP_PERM_ADDR
+ * ARG0:	virtual address
+ * ARG1:	reserved, must be zero
+ * ARG2:	TTE
+ * ARG3:	flags (HV_MMU_{IMMU,DMMU})
+ * RET0:	status
+ * ERRORS:	EINVAL			Invalid virutal address or flags value
+ *		EBADPGSZ		Invalid page size value
+ *		ENORADDR		Invalid real address in TTE
+ *		ETOOMANY		Too many mappings (max of 8 reached)
+ *
+ * Create a permanent mapping using the given TTE and virtual address
+ * for context 0 on the calling virtual CPU.  A maximum of 8 such
+ * permanent mappings may be specified by privileged code.  Mappings
+ * may be removed with mmu_unmap_perm_addr().
+ *
+ * The behavior is undefined if a TTE with the valid bit clear is given.
+ *
+ * Note: This call is used to specify address space mappings for which
+ *       privileged code does not expect to receive misses.  For example,
+ *       this mechanism can be used to map kernel nucleus code and data.
+ */
+#define HV_FAST_MMU_MAP_PERM_ADDR	0x25
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_mmu_map_perm_addr(unsigned long vaddr,
+					     unsigned long set_to_zero,
+					     unsigned long tte,
+					     unsigned long flags);
+#endif
+
+/* mmu_fault_area_conf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_FAULT_AREA_CONF
+ * ARG0:	real address
+ * RET0:	status
+ * RET1:	previous mmu fault area real address
+ * ERRORS:	ENORADDR		Invalid real address
+ *		EBADALIGN		Invalid alignment for fault area
+ *
+ * Configure the MMU fault status area for the calling CPU.  A 64-byte
+ * aligned real address specifies where MMU fault status information
+ * is placed.  The return value is the previously specified area, or 0
+ * for the first invocation.  Specifying a fault area at real address
+ * 0 is not allowed.
+ */
+#define HV_FAST_MMU_FAULT_AREA_CONF	0x26
+
+/* mmu_enable()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_ENABLE
+ * ARG0:	enable flag
+ * ARG1:	return target address
+ * RET0:	status
+ * ERRORS:	ENORADDR		Invalid real address when disabling
+ *					translation.
+ *		EBADALIGN		The return target address is not
+ *					aligned to an instruction.
+ *		EINVAL			The enable flag request the current
+ *					operating mode (e.g. disable if already
+ *					disabled)
+ *
+ * Enable or disable virtual address translation for the calling CPU
+ * within the virtual machine domain.  If the enable flag is zero,
+ * translation is disabled, any non-zero value will enable
+ * translation.
+ *
+ * When this function returns, the newly selected translation mode
+ * will be active.  If the mmu is being enabled, then the return
+ * target address is a virtual address else it is a real address.
+ *
+ * Upon successful completion, control will be returned to the given
+ * return target address (ie. the cpu will jump to that address).  On
+ * failure, the previous mmu mode remains and the trap simply returns
+ * as normal with the appropriate error code in RET0.
+ */
+#define HV_FAST_MMU_ENABLE		0x27
+
+/* mmu_unmap_perm_addr()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_UNMAP_PERM_ADDR
+ * ARG0:	virtual address
+ * ARG1:	reserved, must be zero
+ * ARG2:	flags (HV_MMU_{IMMU,DMMU})
+ * RET0:	status
+ * ERRORS:	EINVAL			Invalid virutal address or flags value
+ *		ENOMAP			Specified mapping was not found
+ *
+ * Demaps any permanent page mapping (established via
+ * mmu_map_perm_addr()) at the given virtual address for context 0 on
+ * the current virtual CPU.  Any virtual tagged caches are guaranteed
+ * to be kept consistent.
+ */
+#define HV_FAST_MMU_UNMAP_PERM_ADDR	0x28
+
+/* mmu_tsb_ctx0_info()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_TSB_CTX0_INFO
+ * ARG0:	max TSBs
+ * ARG1:	buffer pointer
+ * RET0:	status
+ * RET1:	number of TSBs
+ * ERRORS:	EINVAL			Supplied buffer is too small
+ *		EBADALIGN		The buffer pointer is badly aligned
+ *		ENORADDR		Invalid real address for buffer pointer
+ *
+ * Return the TSB configuration as previous defined by mmu_tsb_ctx0()
+ * into the provided buffer.  The size of the buffer is given in ARG1
+ * in terms of the number of TSB description entries.
+ *
+ * Upon return, RET1 always contains the number of TSB descriptions
+ * previously configured.  If zero TSBs were configured, EOK is
+ * returned with RET1 containing 0.
+ */
+#define HV_FAST_MMU_TSB_CTX0_INFO	0x29
+
+/* mmu_tsb_ctxnon0_info()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_TSB_CTXNON0_INFO
+ * ARG0:	max TSBs
+ * ARG1:	buffer pointer
+ * RET0:	status
+ * RET1:	number of TSBs
+ * ERRORS:	EINVAL			Supplied buffer is too small
+ *		EBADALIGN		The buffer pointer is badly aligned
+ *		ENORADDR		Invalid real address for buffer pointer
+ *
+ * Return the TSB configuration as previous defined by
+ * mmu_tsb_ctxnon0() into the provided buffer.  The size of the buffer
+ * is given in ARG1 in terms of the number of TSB description entries.
+ *
+ * Upon return, RET1 always contains the number of TSB descriptions
+ * previously configured.  If zero TSBs were configured, EOK is
+ * returned with RET1 containing 0.
+ */
+#define HV_FAST_MMU_TSB_CTXNON0_INFO	0x2a
+
+/* mmu_fault_area_info()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMU_FAULT_AREA_INFO
+ * RET0:	status
+ * RET1:	fault area real address
+ * ERRORS:	No errors defined.
+ *
+ * Return the currently defined MMU fault status area for the current
+ * CPU.  The real address of the fault status area is returned in
+ * RET1, or 0 is returned in RET1 if no fault status area is defined.
+ *
+ * Note: mmu_fault_area_conf() may be called with the return value (RET1)
+ *       from this service if there is a need to save and restore the fault
+ *	 area for a cpu.
+ */
+#define HV_FAST_MMU_FAULT_AREA_INFO	0x2b
+
+/* Cache and Memory services. */
+
+/* mem_scrub()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MEM_SCRUB
+ * ARG0:	real address
+ * ARG1:	length
+ * RET0:	status
+ * RET1:	length scrubbed
+ * ERRORS:	ENORADDR	Invalid real address
+ *		EBADALIGN	Start address or length are not correctly
+ *				aligned
+ *		EINVAL		Length is zero
+ *
+ * Zero the memory contents in the range real address to real address
+ * plus length minus 1.  Also, valid ECC will be generated for that
+ * memory address range.  Scrubbing is started at the given real
+ * address, but may not scrub the entire given length.  The actual
+ * length scrubbed will be returned in RET1.
+ *
+ * The real address and length must be aligned on an 8K boundary, or
+ * contain the start address and length from a sun4v error report.
+ *
+ * Note: There are two uses for this function.  The first use is to block clear
+ *       and initialize memory and the second is to scrub an u ncorrectable
+ *       error reported via a resumable or non-resumable trap.  The second
+ *       use requires the arguments to be equal to the real address and length
+ *       provided in a sun4v memory error report.
+ */
+#define HV_FAST_MEM_SCRUB		0x31
+
+/* mem_sync()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MEM_SYNC
+ * ARG0:	real address
+ * ARG1:	length
+ * RET0:	status
+ * RET1:	length synced
+ * ERRORS:	ENORADDR	Invalid real address
+ *		EBADALIGN	Start address or length are not correctly
+ *				aligned
+ *		EINVAL		Length is zero
+ *
+ * Force the next access within the real address to real address plus
+ * length minus 1 to be fetches from main system memory.  Less than
+ * the given length may be synced, the actual amount synced is
+ * returned in RET1.  The real address and length must be aligned on
+ * an 8K boundary.
+ */
+#define HV_FAST_MEM_SYNC		0x32
+
+/* Time of day services.
+ *
+ * The hypervisor maintains the time of day on a per-domain basis.
+ * Changing the time of day in one domain does not affect the time of
+ * day on any other domain.
+ *
+ * Time is described by a single unsigned 64-bit word which is the
+ * number of seconds since the UNIX Epoch (00:00:00 UTC, January 1,
+ * 1970).
+ */
+
+/* tod_get()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_TOD_GET
+ * RET0:	status
+ * RET1:	TOD
+ * ERRORS:	EWOULDBLOCK	TOD resource is temporarily unavailable
+ *		ENOTSUPPORTED	If TOD not supported on this platform
+ *
+ * Return the current time of day.  May block if TOD access is
+ * temporarily not possible.
+ */
+#define HV_FAST_TOD_GET			0x50
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_tod_get(unsigned long *time);
+#endif
+
+/* tod_set()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_TOD_SET
+ * ARG0:	TOD
+ * RET0:	status
+ * ERRORS:	EWOULDBLOCK	TOD resource is temporarily unavailable
+ *		ENOTSUPPORTED	If TOD not supported on this platform
+ *
+ * The current time of day is set to the value specified in ARG0.  May
+ * block if TOD access is temporarily not possible.
+ */
+#define HV_FAST_TOD_SET			0x51
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_tod_set(unsigned long time);
+#endif
+
+/* Console services */
+
+/* con_getchar()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CONS_GETCHAR
+ * RET0:	status
+ * RET1:	character
+ * ERRORS:	EWOULDBLOCK	No character available.
+ *
+ * Returns a character from the console device.  If no character is
+ * available then an EWOULDBLOCK error is returned.  If a character is
+ * available, then the returned status is EOK and the character value
+ * is in RET1.
+ *
+ * A virtual BREAK is represented by the 64-bit value -1.
+ *
+ * A virtual HUP signal is represented by the 64-bit value -2.
+ */
+#define HV_FAST_CONS_GETCHAR		0x60
+
+/* con_putchar()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CONS_PUTCHAR
+ * ARG0:	character
+ * RET0:	status
+ * ERRORS:	EINVAL		Illegal character
+ *		EWOULDBLOCK	Output buffer currently full, would block
+ *
+ * Send a character to the console device.  Only character values
+ * between 0 and 255 may be used.  Values outside this range are
+ * invalid except for the 64-bit value -1 which is used to send a
+ * virtual BREAK.
+ */
+#define HV_FAST_CONS_PUTCHAR		0x61
+
+/* con_read()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CONS_READ
+ * ARG0:	buffer real address
+ * ARG1:	buffer size in bytes
+ * RET0:	status
+ * RET1:	bytes read or BREAK or HUP
+ * ERRORS:	EWOULDBLOCK	No character available.
+ *
+ * Reads characters into a buffer from the console device.  If no
+ * character is available then an EWOULDBLOCK error is returned.
+ * If a character is available, then the returned status is EOK
+ * and the number of bytes read into the given buffer is provided
+ * in RET1.
+ *
+ * A virtual BREAK is represented by the 64-bit RET1 value -1.
+ *
+ * A virtual HUP signal is represented by the 64-bit RET1 value -2.
+ *
+ * If BREAK or HUP are indicated, no bytes were read into buffer.
+ */
+#define HV_FAST_CONS_READ		0x62
+
+/* con_write()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_CONS_WRITE
+ * ARG0:	buffer real address
+ * ARG1:	buffer size in bytes
+ * RET0:	status
+ * RET1:	bytes written
+ * ERRORS:	EWOULDBLOCK	Output buffer currently full, would block
+ *
+ * Send a characters in buffer to the console device.  Breaks must be
+ * sent using con_putchar().
+ */
+#define HV_FAST_CONS_WRITE		0x63
+
+#ifndef __ASSEMBLY__
+extern long sun4v_con_getchar(long *status);
+extern long sun4v_con_putchar(long c);
+extern long sun4v_con_read(unsigned long buffer,
+			   unsigned long size,
+			   unsigned long *bytes_read);
+extern unsigned long sun4v_con_write(unsigned long buffer,
+				     unsigned long size,
+				     unsigned long *bytes_written);
+#endif
+
+/* mach_set_soft_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_SET_SOFT_STATE
+ * ARG0:	software state
+ * ARG1:	software state description pointer
+ * RET0:	status
+ * ERRORS:	EINVAL		software state not valid or software state
+ *				description is not NULL terminated
+ *		ENORADDR	software state description pointer is not a
+ *				valid real address
+ *		EBADALIGNED	software state description is not correctly
+ *				aligned
+ *
+ * This allows the guest to report it's soft state to the hypervisor.  There
+ * are two primary components to this state.  The first part states whether
+ * the guest software is running or not.  The second containts optional
+ * details specific to the software.
+ *
+ * The software state argument is defined below in HV_SOFT_STATE_*, and
+ * indicates whether the guest is operating normally or in a transitional
+ * state.
+ *
+ * The software state description argument is a real address of a data buffer
+ * of size 32-bytes aligned on a 32-byte boundary.  It is treated as a NULL
+ * terminated 7-bit ASCII string of up to 31 characters not including the
+ * NULL termination.
+ */
+#define HV_FAST_MACH_SET_SOFT_STATE	0x70
+#define  HV_SOFT_STATE_NORMAL		 0x01
+#define  HV_SOFT_STATE_TRANSITION	 0x02
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_mach_set_soft_state(unsigned long soft_state,
+					       unsigned long msg_string_ra);
+#endif
+
+/* mach_get_soft_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_GET_SOFT_STATE
+ * ARG0:	software state description pointer
+ * RET0:	status
+ * RET1:	software state
+ * ERRORS:	ENORADDR	software state description pointer is not a
+ *				valid real address
+ *		EBADALIGNED	software state description is not correctly
+ *				aligned
+ *
+ * Retrieve the current value of the guest's software state.  The rules
+ * for the software state pointer are the same as for mach_set_soft_state()
+ * above.
+ */
+#define HV_FAST_MACH_GET_SOFT_STATE	0x71
+
+/* svc_send()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_SVC_SEND
+ * ARG0:	service ID
+ * ARG1:	buffer real address
+ * ARG2:	buffer size
+ * RET0:	STATUS
+ * RET1:	sent_bytes
+ *
+ * Be careful, all output registers are clobbered by this operation,
+ * so for example it is not possible to save away a value in %o4
+ * across the trap.
+ */
+#define HV_FAST_SVC_SEND		0x80
+
+/* svc_recv()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_SVC_RECV
+ * ARG0:	service ID
+ * ARG1:	buffer real address
+ * ARG2:	buffer size
+ * RET0:	STATUS
+ * RET1:	recv_bytes
+ *
+ * Be careful, all output registers are clobbered by this operation,
+ * so for example it is not possible to save away a value in %o4
+ * across the trap.
+ */
+#define HV_FAST_SVC_RECV		0x81
+
+/* svc_getstatus()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_SVC_GETSTATUS
+ * ARG0:	service ID
+ * RET0:	STATUS
+ * RET1:	status bits
+ */
+#define HV_FAST_SVC_GETSTATUS		0x82
+
+/* svc_setstatus()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_SVC_SETSTATUS
+ * ARG0:	service ID
+ * ARG1:	bits to set
+ * RET0:	STATUS
+ */
+#define HV_FAST_SVC_SETSTATUS		0x83
+
+/* svc_clrstatus()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_SVC_CLRSTATUS
+ * ARG0:	service ID
+ * ARG1:	bits to clear
+ * RET0:	STATUS
+ */
+#define HV_FAST_SVC_CLRSTATUS		0x84
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_svc_send(unsigned long svc_id,
+				    unsigned long buffer,
+				    unsigned long buffer_size,
+				    unsigned long *sent_bytes);
+extern unsigned long sun4v_svc_recv(unsigned long svc_id,
+				    unsigned long buffer,
+				    unsigned long buffer_size,
+				    unsigned long *recv_bytes);
+extern unsigned long sun4v_svc_getstatus(unsigned long svc_id,
+					 unsigned long *status_bits);
+extern unsigned long sun4v_svc_setstatus(unsigned long svc_id,
+					 unsigned long status_bits);
+extern unsigned long sun4v_svc_clrstatus(unsigned long svc_id,
+					 unsigned long status_bits);
+#endif
+
+/* Trap trace services.
+ *
+ * The hypervisor provides a trap tracing capability for privileged
+ * code running on each virtual CPU.  Privileged code provides a
+ * round-robin trap trace queue within which the hypervisor writes
+ * 64-byte entries detailing hyperprivileged traps taken n behalf of
+ * privileged code.  This is provided as a debugging capability for
+ * privileged code.
+ *
+ * The trap trace control structure is 64-bytes long and placed at the
+ * start (offset 0) of the trap trace buffer, and is described as
+ * follows:
+ */
+#ifndef __ASSEMBLY__
+struct hv_trap_trace_control {
+	unsigned long		head_offset;
+	unsigned long		tail_offset;
+	unsigned long		__reserved[0x30 / sizeof(unsigned long)];
+};
+#endif
+#define HV_TRAP_TRACE_CTRL_HEAD_OFFSET	0x00
+#define HV_TRAP_TRACE_CTRL_TAIL_OFFSET	0x08
+
+/* The head offset is the offset of the most recently completed entry
+ * in the trap-trace buffer.  The tail offset is the offset of the
+ * next entry to be written.  The control structure is owned and
+ * modified by the hypervisor.  A guest may not modify the control
+ * structure contents.  Attempts to do so will result in undefined
+ * behavior for the guest.
+ *
+ * Each trap trace buffer entry is layed out as follows:
+ */
+#ifndef __ASSEMBLY__
+struct hv_trap_trace_entry {
+	unsigned char	type;		/* Hypervisor or guest entry?	*/
+	unsigned char	hpstate;	/* Hyper-privileged state	*/
+	unsigned char	tl;		/* Trap level			*/
+	unsigned char	gl;		/* Global register level	*/
+	unsigned short	tt;		/* Trap type			*/
+	unsigned short	tag;		/* Extended trap identifier	*/
+	unsigned long	tstate;		/* Trap state			*/
+	unsigned long	tick;		/* Tick				*/
+	unsigned long	tpc;		/* Trap PC			*/
+	unsigned long	f1;		/* Entry specific		*/
+	unsigned long	f2;		/* Entry specific		*/
+	unsigned long	f3;		/* Entry specific		*/
+	unsigned long	f4;		/* Entry specific		*/
+};
+#endif
+#define HV_TRAP_TRACE_ENTRY_TYPE	0x00
+#define HV_TRAP_TRACE_ENTRY_HPSTATE	0x01
+#define HV_TRAP_TRACE_ENTRY_TL		0x02
+#define HV_TRAP_TRACE_ENTRY_GL		0x03
+#define HV_TRAP_TRACE_ENTRY_TT		0x04
+#define HV_TRAP_TRACE_ENTRY_TAG		0x06
+#define HV_TRAP_TRACE_ENTRY_TSTATE	0x08
+#define HV_TRAP_TRACE_ENTRY_TICK	0x10
+#define HV_TRAP_TRACE_ENTRY_TPC		0x18
+#define HV_TRAP_TRACE_ENTRY_F1		0x20
+#define HV_TRAP_TRACE_ENTRY_F2		0x28
+#define HV_TRAP_TRACE_ENTRY_F3		0x30
+#define HV_TRAP_TRACE_ENTRY_F4		0x38
+
+/* The type field is encoded as follows.  */
+#define HV_TRAP_TYPE_UNDEF		0x00 /* Entry content undefined     */
+#define HV_TRAP_TYPE_HV			0x01 /* Hypervisor trap entry       */
+#define HV_TRAP_TYPE_GUEST		0xff /* Added via ttrace_addentry() */
+
+/* ttrace_buf_conf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_TTRACE_BUF_CONF
+ * ARG0:	real address
+ * ARG1:	number of entries
+ * RET0:	status
+ * RET1:	number of entries
+ * ERRORS:	ENORADDR	Invalid real address
+ *		EINVAL		Size is too small
+ *		EBADALIGN	Real address not aligned on 64-byte boundary
+ *
+ * Requests hypervisor trap tracing and declares a virtual CPU's trap
+ * trace buffer to the hypervisor.  The real address supplies the real
+ * base address of the trap trace queue and must be 64-byte aligned.
+ * Specifying a value of 0 for the number of entries disables trap
+ * tracing for the calling virtual CPU.  The buffer allocated must be
+ * sized for a power of two number of 64-byte trap trace entries plus
+ * an initial 64-byte control structure.
+ *
+ * This may be invoked any number of times so that a virtual CPU may
+ * relocate a trap trace buffer or create "snapshots" of information.
+ *
+ * If the real address is illegal or badly aligned, then trap tracing
+ * is disabled and an error is returned.
+ *
+ * Upon failure with EINVAL, this service call returns in RET1 the
+ * minimum number of buffer entries required.  Upon other failures
+ * RET1 is undefined.
+ */
+#define HV_FAST_TTRACE_BUF_CONF		0x90
+
+/* ttrace_buf_info()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_TTRACE_BUF_INFO
+ * RET0:	status
+ * RET1:	real address
+ * RET2:	size
+ * ERRORS:	None defined.
+ *
+ * Returns the size and location of the previously declared trap-trace
+ * buffer.  In the event that no buffer was previously defined, or the
+ * buffer is disabled, this call will return a size of zero bytes.
+ */
+#define HV_FAST_TTRACE_BUF_INFO		0x91
+
+/* ttrace_enable()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_TTRACE_ENABLE
+ * ARG0:	enable
+ * RET0:	status
+ * RET1:	previous enable state
+ * ERRORS:	EINVAL		No trap trace buffer currently defined
+ *
+ * Enable or disable trap tracing, and return the previous enabled
+ * state in RET1.  Future systems may define various flags for the
+ * enable argument (ARG0), for the moment a guest should pass
+ * "(uint64_t) -1" to enable, and "(uint64_t) 0" to disable all
+ * tracing - which will ensure future compatability.
+ */
+#define HV_FAST_TTRACE_ENABLE		0x92
+
+/* ttrace_freeze()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_TTRACE_FREEZE
+ * ARG0:	freeze
+ * RET0:	status
+ * RET1:	previous freeze state
+ * ERRORS:	EINVAL		No trap trace buffer currently defined
+ *
+ * Freeze or unfreeze trap tracing, returning the previous freeze
+ * state in RET1.  A guest should pass a non-zero value to freeze and
+ * a zero value to unfreeze all tracing.  The returned previous state
+ * is 0 for not frozen and 1 for frozen.
+ */
+#define HV_FAST_TTRACE_FREEZE		0x93
+
+/* ttrace_addentry()
+ * TRAP:	HV_TTRACE_ADDENTRY_TRAP
+ * ARG0:	tag (16-bits)
+ * ARG1:	data word 0
+ * ARG2:	data word 1
+ * ARG3:	data word 2
+ * ARG4:	data word 3
+ * RET0:	status
+ * ERRORS:	EINVAL		No trap trace buffer currently defined
+ *
+ * Add an entry to the trap trace buffer.  Upon return only ARG0/RET0
+ * is modified - none of the other registers holding arguments are
+ * volatile across this hypervisor service.
+ */
+
+/* Core dump services.
+ *
+ * Since the hypervisor viraulizes and thus obscures a lot of the
+ * physical machine layout and state, traditional OS crash dumps can
+ * be difficult to diagnose especially when the problem is a
+ * configuration error of some sort.
+ *
+ * The dump services provide an opaque buffer into which the
+ * hypervisor can place it's internal state in order to assist in
+ * debugging such situations.  The contents are opaque and extremely
+ * platform and hypervisor implementation specific.  The guest, during
+ * a core dump, requests that the hypervisor update any information in
+ * the dump buffer in preparation to being dumped as part of the
+ * domain's memory image.
+ */
+
+/* dump_buf_update()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_DUMP_BUF_UPDATE
+ * ARG0:	real address
+ * ARG1:	size
+ * RET0:	status
+ * RET1:	required size of dump buffer
+ * ERRORS:	ENORADDR	Invalid real address
+ *		EBADALIGN	Real address is not aligned on a 64-byte
+ *				boundary
+ *		EINVAL		Size is non-zero but less than minimum size
+ *				required
+ *		ENOTSUPPORTED	Operation not supported on current logical
+ *				domain
+ *
+ * Declare a domain dump buffer to the hypervisor.  The real address
+ * provided for the domain dump buffer must be 64-byte aligned.  The
+ * size specifies the size of the dump buffer and may be larger than
+ * the minimum size specified in the machine description.  The
+ * hypervisor will fill the dump buffer with opaque data.
+ *
+ * Note: A guest may elect to include dump buffer contents as part of a crash
+ *       dump to assist with debugging.  This function may be called any number
+ *       of times so that a guest may relocate a dump buffer, or create
+ *       "snapshots" of any dump-buffer information.  Each call to
+ *       dump_buf_update() atomically declares the new dump buffer to the
+ *       hypervisor.
+ *
+ * A specified size of 0 unconfigures the dump buffer.  If the real
+ * address is illegal or badly aligned, then any currently active dump
+ * buffer is disabled and an error is returned.
+ *
+ * In the event that the call fails with EINVAL, RET1 contains the
+ * minimum size requires by the hypervisor for a valid dump buffer.
+ */
+#define HV_FAST_DUMP_BUF_UPDATE		0x94
+
+/* dump_buf_info()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_DUMP_BUF_INFO
+ * RET0:	status
+ * RET1:	real address of current dump buffer
+ * RET2:	size of current dump buffer
+ * ERRORS:	No errors defined.
+ *
+ * Return the currently configures dump buffer description.  A
+ * returned size of 0 bytes indicates an undefined dump buffer.  In
+ * this case the return address in RET1 is undefined.
+ */
+#define HV_FAST_DUMP_BUF_INFO		0x95
+
+/* Device interrupt services.
+ *
+ * Device interrupts are allocated to system bus bridges by the hypervisor,
+ * and described to OBP in the machine description.  OBP then describes
+ * these interrupts to the OS via properties in the device tree.
+ *
+ * Terminology:
+ *
+ *	cpuid		Unique opaque value which represents a target cpu.
+ *
+ *	devhandle	Device handle.  It uniquely identifies a device, and
+ *			consistes of the lower 28-bits of the hi-cell of the
+ *			first entry of the device's "reg" property in the
+ *			OBP device tree.
+ *
+ *	devino		Device interrupt number.  Specifies the relative
+ *			interrupt number within the device.  The unique
+ *			combination of devhandle and devino are used to
+ *			identify a specific device interrupt.
+ *
+ *			Note: The devino value is the same as the values in the
+ *			      "interrupts" property or "interrupt-map" property
+ *			      in the OBP device tree for that device.
+ *
+ *	sysino		System interrupt number.  A 64-bit unsigned interger
+ *			representing a unique interrupt within a virtual
+ *			machine.
+ *
+ *	intr_state	A flag representing the interrupt state for a given
+ *			sysino.  The state values are defined below.
+ *
+ *	intr_enabled	A flag representing the 'enabled' state for a given
+ *			sysino.  The enable values are defined below.
+ */
+
+#define HV_INTR_STATE_IDLE		0 /* Nothing pending */
+#define HV_INTR_STATE_RECEIVED		1 /* Interrupt received by hardware */
+#define HV_INTR_STATE_DELIVERED		2 /* Interrupt delivered to queue */
+
+#define HV_INTR_DISABLED		0 /* sysino not enabled */
+#define HV_INTR_ENABLED			1 /* sysino enabled */
+
+/* intr_devino_to_sysino()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_INTR_DEVINO2SYSINO
+ * ARG0:	devhandle
+ * ARG1:	devino
+ * RET0:	status
+ * RET1:	sysino
+ * ERRORS:	EINVAL		Invalid devhandle/devino
+ *
+ * Converts a device specific interrupt number of the given
+ * devhandle/devino into a system specific ino (sysino).
+ */
+#define HV_FAST_INTR_DEVINO2SYSINO	0xa0
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_devino_to_sysino(unsigned long devhandle,
+					    unsigned long devino);
+#endif
+
+/* intr_getenabled()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_INTR_GETENABLED
+ * ARG0:	sysino
+ * RET0:	status
+ * RET1:	intr_enabled (HV_INTR_{DISABLED,ENABLED})
+ * ERRORS:	EINVAL		Invalid sysino
+ *
+ * Returns interrupt enabled state in RET1 for the interrupt defined
+ * by the given sysino.
+ */
+#define HV_FAST_INTR_GETENABLED		0xa1
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_intr_getenabled(unsigned long sysino);
+#endif
+
+/* intr_setenabled()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_INTR_SETENABLED
+ * ARG0:	sysino
+ * ARG1:	intr_enabled (HV_INTR_{DISABLED,ENABLED})
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid sysino or intr_enabled value
+ *
+ * Set the 'enabled' state of the interrupt sysino.
+ */
+#define HV_FAST_INTR_SETENABLED		0xa2
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_intr_setenabled(unsigned long sysino, unsigned long intr_enabled);
+#endif
+
+/* intr_getstate()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_INTR_GETSTATE
+ * ARG0:	sysino
+ * RET0:	status
+ * RET1:	intr_state (HV_INTR_STATE_*)
+ * ERRORS:	EINVAL		Invalid sysino
+ *
+ * Returns current state of the interrupt defined by the given sysino.
+ */
+#define HV_FAST_INTR_GETSTATE		0xa3
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_intr_getstate(unsigned long sysino);
+#endif
+
+/* intr_setstate()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_INTR_SETSTATE
+ * ARG0:	sysino
+ * ARG1:	intr_state (HV_INTR_STATE_*)
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid sysino or intr_state value
+ *
+ * Sets the current state of the interrupt described by the given sysino
+ * value.
+ *
+ * Note: Setting the state to HV_INTR_STATE_IDLE clears any pending
+ *       interrupt for sysino.
+ */
+#define HV_FAST_INTR_SETSTATE		0xa4
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_intr_setstate(unsigned long sysino, unsigned long intr_state);
+#endif
+
+/* intr_gettarget()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_INTR_GETTARGET
+ * ARG0:	sysino
+ * RET0:	status
+ * RET1:	cpuid
+ * ERRORS:	EINVAL		Invalid sysino
+ *
+ * Returns CPU that is the current target of the interrupt defined by
+ * the given sysino.  The CPU value returned is undefined if the target
+ * has not been set via intr_settarget().
+ */
+#define HV_FAST_INTR_GETTARGET		0xa5
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_intr_gettarget(unsigned long sysino);
+#endif
+
+/* intr_settarget()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_INTR_SETTARGET
+ * ARG0:	sysino
+ * ARG1:	cpuid
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid sysino
+ *		ENOCPU		Invalid cpuid
+ *
+ * Set the target CPU for the interrupt defined by the given sysino.
+ */
+#define HV_FAST_INTR_SETTARGET		0xa6
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_intr_settarget(unsigned long sysino, unsigned long cpuid);
+#endif
+
+/* vintr_get_cookie()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_GET_COOKIE
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * RET0:	status
+ * RET1:	cookie
+ */
+#define HV_FAST_VINTR_GET_COOKIE	0xa7
+
+/* vintr_set_cookie()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_SET_COOKIE
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * ARG2:	cookie
+ * RET0:	status
+ */
+#define HV_FAST_VINTR_SET_COOKIE	0xa8
+
+/* vintr_get_valid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_GET_VALID
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * RET0:	status
+ * RET1:	valid state
+ */
+#define HV_FAST_VINTR_GET_VALID		0xa9
+
+/* vintr_set_valid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_SET_VALID
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * ARG2:	valid state
+ * RET0:	status
+ */
+#define HV_FAST_VINTR_SET_VALID		0xaa
+
+/* vintr_get_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_GET_STATE
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * RET0:	status
+ * RET1:	state
+ */
+#define HV_FAST_VINTR_GET_STATE		0xab
+
+/* vintr_set_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_SET_STATE
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * ARG2:	state
+ * RET0:	status
+ */
+#define HV_FAST_VINTR_SET_STATE		0xac
+
+/* vintr_get_target()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_GET_TARGET
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * RET0:	status
+ * RET1:	cpuid
+ */
+#define HV_FAST_VINTR_GET_TARGET	0xad
+
+/* vintr_set_target()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_VINTR_SET_TARGET
+ * ARG0:	device handle
+ * ARG1:	device ino
+ * ARG2:	cpuid
+ * RET0:	status
+ */
+#define HV_FAST_VINTR_SET_TARGET	0xae
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_vintr_get_cookie(unsigned long dev_handle,
+					    unsigned long dev_ino,
+					    unsigned long *cookie);
+extern unsigned long sun4v_vintr_set_cookie(unsigned long dev_handle,
+					    unsigned long dev_ino,
+					    unsigned long cookie);
+extern unsigned long sun4v_vintr_get_valid(unsigned long dev_handle,
+					   unsigned long dev_ino,
+					   unsigned long *valid);
+extern unsigned long sun4v_vintr_set_valid(unsigned long dev_handle,
+					   unsigned long dev_ino,
+					   unsigned long valid);
+extern unsigned long sun4v_vintr_get_state(unsigned long dev_handle,
+					   unsigned long dev_ino,
+					   unsigned long *state);
+extern unsigned long sun4v_vintr_set_state(unsigned long dev_handle,
+					   unsigned long dev_ino,
+					   unsigned long state);
+extern unsigned long sun4v_vintr_get_target(unsigned long dev_handle,
+					    unsigned long dev_ino,
+					    unsigned long *cpuid);
+extern unsigned long sun4v_vintr_set_target(unsigned long dev_handle,
+					    unsigned long dev_ino,
+					    unsigned long cpuid);
+#endif
+
+/* PCI IO services.
+ *
+ * See the terminology descriptions in the device interrupt services
+ * section above as those apply here too.  Here are terminology
+ * definitions specific to these PCI IO services:
+ *
+ *	tsbnum		TSB number.  Indentifies which io-tsb is used.
+ *			For this version of the specification, tsbnum
+ *			must be zero.
+ *
+ *	tsbindex	TSB index.  Identifies which entry in the TSB
+ *			is used.  The first entry is zero.
+ *
+ *	tsbid		A 64-bit aligned data structure which contains
+ *			a tsbnum and a tsbindex.  Bits 63:32 contain the
+ *			tsbnum and bits 31:00 contain the tsbindex.
+ *
+ *			Use the HV_PCI_TSBID() macro to construct such
+ * 			values.
+ *
+ *	io_attributes	IO attributes for IOMMU mappings.  One of more
+ *			of the attritbute bits are stores in a 64-bit
+ *			value.  The values are defined below.
+ *
+ *	r_addr		64-bit real address
+ *
+ *	pci_device	PCI device address.  A PCI device address identifies
+ *			a specific device on a specific PCI bus segment.
+ *			A PCI device address ia a 32-bit unsigned integer
+ *			with the following format:
+ *
+ *				00000000.bbbbbbbb.dddddfff.00000000
+ *
+ *			Use the HV_PCI_DEVICE_BUILD() macro to construct
+ *			such values.
+ *
+ *	pci_config_offset
+ *			PCI configureation space offset.  For conventional
+ *			PCI a value between 0 and 255.  For extended
+ *			configuration space, a value between 0 and 4095.
+ *
+ *			Note: For PCI configuration space accesses, the offset
+ *			      must be aligned to the access size.
+ *
+ *	error_flag	A return value which specifies if the action succeeded
+ *			or failed.  0 means no error, non-0 means some error
+ *			occurred while performing the service.
+ *
+ *	io_sync_direction
+ *			Direction definition for pci_dma_sync(), defined
+ *			below in HV_PCI_SYNC_*.
+ *
+ *	io_page_list	A list of io_page_addresses, an io_page_address is
+ *			a real address.
+ *
+ *	io_page_list_p	A pointer to an io_page_list.
+ *
+ *	"size based byte swap" - Some functions do size based byte swapping
+ *				 which allows sw to access pointers and
+ *				 counters in native form when the processor
+ *				 operates in a different endianness than the
+ *				 IO bus.  Size-based byte swapping converts a
+ *				 multi-byte field between big-endian and
+ *				 little-endian format.
+ */
+
+#define HV_PCI_MAP_ATTR_READ		0x01
+#define HV_PCI_MAP_ATTR_WRITE		0x02
+
+#define HV_PCI_DEVICE_BUILD(b,d,f)	\
+	((((b) & 0xff) << 16) | \
+	 (((d) & 0x1f) << 11) | \
+	 (((f) & 0x07) <<  8))
+
+#define HV_PCI_TSBID(__tsb_num, __tsb_index) \
+	((((u64)(__tsb_num)) << 32UL) | ((u64)(__tsb_index)))
+
+#define HV_PCI_SYNC_FOR_DEVICE		0x01
+#define HV_PCI_SYNC_FOR_CPU		0x02
+
+/* pci_iommu_map()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_IOMMU_MAP
+ * ARG0:	devhandle
+ * ARG1:	tsbid
+ * ARG2:	#ttes
+ * ARG3:	io_attributes
+ * ARG4:	io_page_list_p
+ * RET0:	status
+ * RET1:	#ttes mapped
+ * ERRORS:	EINVAL		Invalid devhandle/tsbnum/tsbindex/io_attributes
+ *		EBADALIGN	Improperly aligned real address
+ *		ENORADDR	Invalid real address
+ *
+ * Create IOMMU mappings in the sun4v device defined by the given
+ * devhandle.  The mappings are created in the TSB defined by the
+ * tsbnum component of the given tsbid.  The first mapping is created
+ * in the TSB i ndex defined by the tsbindex component of the given tsbid.
+ * The call creates up to #ttes mappings, the first one at tsbnum, tsbindex,
+ * the second at tsbnum, tsbindex + 1, etc.
+ *
+ * All mappings are created with the attributes defined by the io_attributes
+ * argument.  The page mapping addresses are described in the io_page_list
+ * defined by the given io_page_list_p, which is a pointer to the io_page_list.
+ * The first entry in the io_page_list is the address for the first iotte, the
+ * 2nd for the 2nd iotte, and so on.
+ *
+ * Each io_page_address in the io_page_list must be appropriately aligned.
+ * #ttes must be greater than zero.  For this version of the spec, the tsbnum
+ * component of the given tsbid must be zero.
+ *
+ * Returns the actual number of mappings creates, which may be less than
+ * or equal to the argument #ttes.  If the function returns a value which
+ * is less than the #ttes, the caller may continus to call the function with
+ * an updated tsbid, #ttes, io_page_list_p arguments until all pages are
+ * mapped.
+ *
+ * Note: This function does not imply an iotte cache flush.  The guest must
+ *       demap an entry before re-mapping it.
+ */
+#define HV_FAST_PCI_IOMMU_MAP		0xb0
+
+/* pci_iommu_demap()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_IOMMU_DEMAP
+ * ARG0:	devhandle
+ * ARG1:	tsbid
+ * ARG2:	#ttes
+ * RET0:	status
+ * RET1:	#ttes demapped
+ * ERRORS:	EINVAL		Invalid devhandle/tsbnum/tsbindex
+ *
+ * Demap and flush IOMMU mappings in the device defined by the given
+ * devhandle.  Demaps up to #ttes entries in the TSB defined by the tsbnum
+ * component of the given tsbid, starting at the TSB index defined by the
+ * tsbindex component of the given tsbid.
+ *
+ * For this version of the spec, the tsbnum of the given tsbid must be zero.
+ * #ttes must be greater than zero.
+ *
+ * Returns the actual number of ttes demapped, which may be less than or equal
+ * to the argument #ttes.  If #ttes demapped is less than #ttes, the caller
+ * may continue to call this function with updated tsbid and #ttes arguments
+ * until all pages are demapped.
+ *
+ * Note: Entries do not have to be mapped to be demapped.  A demap of an
+ *       unmapped page will flush the entry from the tte cache.
+ */
+#define HV_FAST_PCI_IOMMU_DEMAP		0xb1
+
+/* pci_iommu_getmap()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_IOMMU_GETMAP
+ * ARG0:	devhandle
+ * ARG1:	tsbid
+ * RET0:	status
+ * RET1:	io_attributes
+ * RET2:	real address
+ * ERRORS:	EINVAL		Invalid devhandle/tsbnum/tsbindex
+ *		ENOMAP		Mapping is not valid, no translation exists
+ *
+ * Read and return the mapping in the device described by the given devhandle
+ * and tsbid.  If successful, the io_attributes shall be returned in RET1
+ * and the page address of the mapping shall be returned in RET2.
+ *
+ * For this version of the spec, the tsbnum component of the given tsbid
+ * must be zero.
+ */
+#define HV_FAST_PCI_IOMMU_GETMAP	0xb2
+
+/* pci_iommu_getbypass()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_IOMMU_GETBYPASS
+ * ARG0:	devhandle
+ * ARG1:	real address
+ * ARG2:	io_attributes
+ * RET0:	status
+ * RET1:	io_addr
+ * ERRORS:	EINVAL		Invalid devhandle/io_attributes
+ *		ENORADDR	Invalid real address
+ *		ENOTSUPPORTED	Function not supported in this implementation.
+ *
+ * Create a "special" mapping in the device described by the given devhandle,
+ * for the given real address and attributes.  Return the IO address in RET1
+ * if successful.
+ */
+#define HV_FAST_PCI_IOMMU_GETBYPASS	0xb3
+
+/* pci_config_get()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_CONFIG_GET
+ * ARG0:	devhandle
+ * ARG1:	pci_device
+ * ARG2:	pci_config_offset
+ * ARG3:	size
+ * RET0:	status
+ * RET1:	error_flag
+ * RET2:	data
+ * ERRORS:	EINVAL		Invalid devhandle/pci_device/offset/size
+ *		EBADALIGN	pci_config_offset not size aligned
+ *		ENOACCESS	Access to this offset is not permitted
+ *
+ * Read PCI configuration space for the adapter described by the given
+ * devhandle.  Read size (1, 2, or 4) bytes of data from the given
+ * pci_device, at pci_config_offset from the beginning of the device's
+ * configuration space.  If there was no error, RET1 is set to zero and
+ * RET2 is set to the data read.  Insignificant bits in RET2 are not
+ * guarenteed to have any specific value and therefore must be ignored.
+ *
+ * The data returned in RET2 is size based byte swapped.
+ *
+ * If an error occurs during the read, set RET1 to a non-zero value.  The
+ * given pci_config_offset must be 'size' aligned.
+ */
+#define HV_FAST_PCI_CONFIG_GET		0xb4
+
+/* pci_config_put()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_CONFIG_PUT
+ * ARG0:	devhandle
+ * ARG1:	pci_device
+ * ARG2:	pci_config_offset
+ * ARG3:	size
+ * ARG4:	data
+ * RET0:	status
+ * RET1:	error_flag
+ * ERRORS:	EINVAL		Invalid devhandle/pci_device/offset/size
+ *		EBADALIGN	pci_config_offset not size aligned
+ *		ENOACCESS	Access to this offset is not permitted
+ *
+ * Write PCI configuration space for the adapter described by the given
+ * devhandle.  Write size (1, 2, or 4) bytes of data in a single operation,
+ * at pci_config_offset from the beginning of the device's configuration
+ * space.  The data argument contains the data to be written to configuration
+ * space.  Prior to writing, the data is size based byte swapped.
+ *
+ * If an error occurs during the write access, do not generate an error
+ * report, do set RET1 to a non-zero value.  Otherwise RET1 is zero.
+ * The given pci_config_offset must be 'size' aligned.
+ *
+ * This function is permitted to read from offset zero in the configuration
+ * space described by the given pci_device if necessary to ensure that the
+ * write access to config space completes.
+ */
+#define HV_FAST_PCI_CONFIG_PUT		0xb5
+
+/* pci_peek()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_PEEK
+ * ARG0:	devhandle
+ * ARG1:	real address
+ * ARG2:	size
+ * RET0:	status
+ * RET1:	error_flag
+ * RET2:	data
+ * ERRORS:	EINVAL		Invalid devhandle or size
+ *		EBADALIGN	Improperly aligned real address
+ *		ENORADDR	Bad real address
+ *		ENOACCESS	Guest access prohibited
+ *
+ * Attempt to read the IO address given by the given devhandle, real address,
+ * and size.  Size must be 1, 2, 4, or 8.  The read is performed as a single
+ * access operation using the given size.  If an error occurs when reading
+ * from the given location, do not generate an error report, but return a
+ * non-zero value in RET1.  If the read was successful, return zero in RET1
+ * and return the actual data read in RET2.  The data returned is size based
+ * byte swapped.
+ *
+ * Non-significant bits in RET2 are not guarenteed to have any specific value
+ * and therefore must be ignored.  If RET1 is returned as non-zero, the data
+ * value is not guarenteed to have any specific value and should be ignored.
+ *
+ * The caller must have permission to read from the given devhandle, real
+ * address, which must be an IO address.  The argument real address must be a
+ * size aligned address.
+ *
+ * The hypervisor implementation of this function must block access to any
+ * IO address that the guest does not have explicit permission to access.
+ */
+#define HV_FAST_PCI_PEEK		0xb6
+
+/* pci_poke()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_POKE
+ * ARG0:	devhandle
+ * ARG1:	real address
+ * ARG2:	size
+ * ARG3:	data
+ * ARG4:	pci_device
+ * RET0:	status
+ * RET1:	error_flag
+ * ERRORS:	EINVAL		Invalid devhandle, size, or pci_device
+ *		EBADALIGN	Improperly aligned real address
+ *		ENORADDR	Bad real address
+ *		ENOACCESS	Guest access prohibited
+ *		ENOTSUPPORTED	Function is not supported by implementation
+ *
+ * Attempt to write data to the IO address given by the given devhandle,
+ * real address, and size.  Size must be 1, 2, 4, or 8.  The write is
+ * performed as a single access operation using the given size. Prior to
+ * writing the data is size based swapped.
+ *
+ * If an error occurs when writing to the given location, do not generate an
+ * error report, but return a non-zero value in RET1.  If the write was
+ * successful, return zero in RET1.
+ *
+ * pci_device describes the configuration address of the device being
+ * written to.  The implementation may safely read from offset 0 with
+ * the configuration space of the device described by devhandle and
+ * pci_device in order to guarantee that the write portion of the operation
+ * completes
+ *
+ * Any error that occurs due to the read shall be reported using the normal
+ * error reporting mechanisms .. the read error is not suppressed.
+ *
+ * The caller must have permission to write to the given devhandle, real
+ * address, which must be an IO address.  The argument real address must be a
+ * size aligned address.  The caller must have permission to read from
+ * the given devhandle, pci_device cofiguration space offset 0.
+ *
+ * The hypervisor implementation of this function must block access to any
+ * IO address that the guest does not have explicit permission to access.
+ */
+#define HV_FAST_PCI_POKE		0xb7
+
+/* pci_dma_sync()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_DMA_SYNC
+ * ARG0:	devhandle
+ * ARG1:	real address
+ * ARG2:	size
+ * ARG3:	io_sync_direction
+ * RET0:	status
+ * RET1:	#synced
+ * ERRORS:	EINVAL		Invalid devhandle or io_sync_direction
+ *		ENORADDR	Bad real address
+ *
+ * Synchronize a memory region described by the given real address and size,
+ * for the device defined by the given devhandle using the direction(s)
+ * defined by the given io_sync_direction.  The argument size is the size of
+ * the memory region in bytes.
+ *
+ * Return the actual number of bytes synchronized in the return value #synced,
+ * which may be less than or equal to the argument size.  If the return
+ * value #synced is less than size, the caller must continue to call this
+ * function with updated real address and size arguments until the entire
+ * memory region is synchronized.
+ */
+#define HV_FAST_PCI_DMA_SYNC		0xb8
+
+/* PCI MSI services.  */
+
+#define HV_MSITYPE_MSI32		0x00
+#define HV_MSITYPE_MSI64		0x01
+
+#define HV_MSIQSTATE_IDLE		0x00
+#define HV_MSIQSTATE_ERROR		0x01
+
+#define HV_MSIQ_INVALID			0x00
+#define HV_MSIQ_VALID			0x01
+
+#define HV_MSISTATE_IDLE		0x00
+#define HV_MSISTATE_DELIVERED		0x01
+
+#define HV_MSIVALID_INVALID		0x00
+#define HV_MSIVALID_VALID		0x01
+
+#define HV_PCIE_MSGTYPE_PME_MSG		0x18
+#define HV_PCIE_MSGTYPE_PME_ACK_MSG	0x1b
+#define HV_PCIE_MSGTYPE_CORR_MSG	0x30
+#define HV_PCIE_MSGTYPE_NONFATAL_MSG	0x31
+#define HV_PCIE_MSGTYPE_FATAL_MSG	0x33
+
+#define HV_MSG_INVALID			0x00
+#define HV_MSG_VALID			0x01
+
+/* pci_msiq_conf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_CONF
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * ARG2:	real address
+ * ARG3:	number of entries
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle, msiqid or nentries
+ *		EBADALIGN	Improperly aligned real address
+ *		ENORADDR	Bad real address
+ *
+ * Configure the MSI queue given by the devhandle and msiqid arguments,
+ * and to be placed at the given real address and be of the given
+ * number of entries.  The real address must be aligned exactly to match
+ * the queue size.  Each queue entry is 64-bytes long, so f.e. a 32 entry
+ * queue must be aligned on a 2048 byte real address boundary.  The MSI-EQ
+ * Head and Tail are initialized so that the MSI-EQ is 'empty'.
+ *
+ * Implementation Note: Certain implementations have fixed sized queues.  In
+ *                      that case, number of entries must contain the correct
+ *                      value.
+ */
+#define HV_FAST_PCI_MSIQ_CONF		0xc0
+
+/* pci_msiq_info()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_INFO
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * RET0:	status
+ * RET1:	real address
+ * RET2:	number of entries
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid
+ *
+ * Return the configuration information for the MSI queue described
+ * by the given devhandle and msiqid.  The base address of the queue
+ * is returned in ARG1 and the number of entries is returned in ARG2.
+ * If the queue is unconfigured, the real address is undefined and the
+ * number of entries will be returned as zero.
+ */
+#define HV_FAST_PCI_MSIQ_INFO		0xc1
+
+/* pci_msiq_getvalid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_GETVALID
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * RET0:	status
+ * RET1:	msiqvalid	(HV_MSIQ_VALID or HV_MSIQ_INVALID)
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid
+ *
+ * Get the valid state of the MSI-EQ described by the given devhandle and
+ * msiqid.
+ */
+#define HV_FAST_PCI_MSIQ_GETVALID	0xc2
+
+/* pci_msiq_setvalid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_SETVALID
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * ARG2:	msiqvalid	(HV_MSIQ_VALID or HV_MSIQ_INVALID)
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid or msiqvalid
+ *				value or MSI EQ is uninitialized
+ *
+ * Set the valid state of the MSI-EQ described by the given devhandle and
+ * msiqid to the given msiqvalid.
+ */
+#define HV_FAST_PCI_MSIQ_SETVALID	0xc3
+
+/* pci_msiq_getstate()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_GETSTATE
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * RET0:	status
+ * RET1:	msiqstate	(HV_MSIQSTATE_IDLE or HV_MSIQSTATE_ERROR)
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid
+ *
+ * Get the state of the MSI-EQ described by the given devhandle and
+ * msiqid.
+ */
+#define HV_FAST_PCI_MSIQ_GETSTATE	0xc4
+
+/* pci_msiq_getvalid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_GETVALID
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * ARG2:	msiqstate	(HV_MSIQSTATE_IDLE or HV_MSIQSTATE_ERROR)
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid or msiqstate
+ *				value or MSI EQ is uninitialized
+ *
+ * Set the state of the MSI-EQ described by the given devhandle and
+ * msiqid to the given msiqvalid.
+ */
+#define HV_FAST_PCI_MSIQ_SETSTATE	0xc5
+
+/* pci_msiq_gethead()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_GETHEAD
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * RET0:	status
+ * RET1:	msiqhead
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid
+ *
+ * Get the current MSI EQ queue head for the MSI-EQ described by the
+ * given devhandle and msiqid.
+ */
+#define HV_FAST_PCI_MSIQ_GETHEAD	0xc6
+
+/* pci_msiq_sethead()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_SETHEAD
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * ARG2:	msiqhead
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid or msiqhead,
+ *				or MSI EQ is uninitialized
+ *
+ * Set the current MSI EQ queue head for the MSI-EQ described by the
+ * given devhandle and msiqid.
+ */
+#define HV_FAST_PCI_MSIQ_SETHEAD	0xc7
+
+/* pci_msiq_gettail()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSIQ_GETTAIL
+ * ARG0:	devhandle
+ * ARG1:	msiqid
+ * RET0:	status
+ * RET1:	msiqtail
+ * ERRORS:	EINVAL		Invalid devhandle or msiqid
+ *
+ * Get the current MSI EQ queue tail for the MSI-EQ described by the
+ * given devhandle and msiqid.
+ */
+#define HV_FAST_PCI_MSIQ_GETTAIL	0xc8
+
+/* pci_msi_getvalid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSI_GETVALID
+ * ARG0:	devhandle
+ * ARG1:	msinum
+ * RET0:	status
+ * RET1:	msivalidstate
+ * ERRORS:	EINVAL		Invalid devhandle or msinum
+ *
+ * Get the current valid/enabled state for the MSI defined by the
+ * given devhandle and msinum.
+ */
+#define HV_FAST_PCI_MSI_GETVALID	0xc9
+
+/* pci_msi_setvalid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSI_SETVALID
+ * ARG0:	devhandle
+ * ARG1:	msinum
+ * ARG2:	msivalidstate
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle or msinum or msivalidstate
+ *
+ * Set the current valid/enabled state for the MSI defined by the
+ * given devhandle and msinum.
+ */
+#define HV_FAST_PCI_MSI_SETVALID	0xca
+
+/* pci_msi_getmsiq()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSI_GETMSIQ
+ * ARG0:	devhandle
+ * ARG1:	msinum
+ * RET0:	status
+ * RET1:	msiqid
+ * ERRORS:	EINVAL		Invalid devhandle or msinum or MSI is unbound
+ *
+ * Get the MSI EQ that the MSI defined by the given devhandle and
+ * msinum is bound to.
+ */
+#define HV_FAST_PCI_MSI_GETMSIQ		0xcb
+
+/* pci_msi_setmsiq()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSI_SETMSIQ
+ * ARG0:	devhandle
+ * ARG1:	msinum
+ * ARG2:	msitype
+ * ARG3:	msiqid
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle or msinum or msiqid
+ *
+ * Set the MSI EQ that the MSI defined by the given devhandle and
+ * msinum is bound to.
+ */
+#define HV_FAST_PCI_MSI_SETMSIQ		0xcc
+
+/* pci_msi_getstate()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSI_GETSTATE
+ * ARG0:	devhandle
+ * ARG1:	msinum
+ * RET0:	status
+ * RET1:	msistate
+ * ERRORS:	EINVAL		Invalid devhandle or msinum
+ *
+ * Get the state of the MSI defined by the given devhandle and msinum.
+ * If not initialized, return HV_MSISTATE_IDLE.
+ */
+#define HV_FAST_PCI_MSI_GETSTATE	0xcd
+
+/* pci_msi_setstate()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSI_SETSTATE
+ * ARG0:	devhandle
+ * ARG1:	msinum
+ * ARG2:	msistate
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle or msinum or msistate
+ *
+ * Set the state of the MSI defined by the given devhandle and msinum.
+ */
+#define HV_FAST_PCI_MSI_SETSTATE	0xce
+
+/* pci_msg_getmsiq()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSG_GETMSIQ
+ * ARG0:	devhandle
+ * ARG1:	msgtype
+ * RET0:	status
+ * RET1:	msiqid
+ * ERRORS:	EINVAL		Invalid devhandle or msgtype
+ *
+ * Get the MSI EQ of the MSG defined by the given devhandle and msgtype.
+ */
+#define HV_FAST_PCI_MSG_GETMSIQ		0xd0
+
+/* pci_msg_setmsiq()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSG_SETMSIQ
+ * ARG0:	devhandle
+ * ARG1:	msgtype
+ * ARG2:	msiqid
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle, msgtype, or msiqid
+ *
+ * Set the MSI EQ of the MSG defined by the given devhandle and msgtype.
+ */
+#define HV_FAST_PCI_MSG_SETMSIQ		0xd1
+
+/* pci_msg_getvalid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSG_GETVALID
+ * ARG0:	devhandle
+ * ARG1:	msgtype
+ * RET0:	status
+ * RET1:	msgvalidstate
+ * ERRORS:	EINVAL		Invalid devhandle or msgtype
+ *
+ * Get the valid/enabled state of the MSG defined by the given
+ * devhandle and msgtype.
+ */
+#define HV_FAST_PCI_MSG_GETVALID	0xd2
+
+/* pci_msg_setvalid()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_PCI_MSG_SETVALID
+ * ARG0:	devhandle
+ * ARG1:	msgtype
+ * ARG2:	msgvalidstate
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid devhandle or msgtype or msgvalidstate
+ *
+ * Set the valid/enabled state of the MSG defined by the given
+ * devhandle and msgtype.
+ */
+#define HV_FAST_PCI_MSG_SETVALID	0xd3
+
+/* Logical Domain Channel services.  */
+
+#define LDC_CHANNEL_DOWN		0
+#define LDC_CHANNEL_UP			1
+#define LDC_CHANNEL_RESETTING		2
+
+/* ldc_tx_qconf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_TX_QCONF
+ * ARG0:	channel ID
+ * ARG1:	real address base of queue
+ * ARG2:	num entries in queue
+ * RET0:	status
+ *
+ * Configure transmit queue for the LDC endpoint specified by the
+ * given channel ID, to be placed at the given real address, and
+ * be of the given num entries.  Num entries must be a power of two.
+ * The real address base of the queue must be aligned on the queue
+ * size.  Each queue entry is 64-bytes, so for example, a 32 entry
+ * queue must be aligned on a 2048 byte real address boundary.
+ *
+ * Upon configuration of a valid transmit queue the head and tail
+ * pointers are set to a hypervisor specific identical value indicating
+ * that the queue initially is empty.
+ *
+ * The endpoint's transmit queue is un-configured if num entries is zero.
+ *
+ * The maximum number of entries for each queue for a specific cpu may be
+ * determined from the machine description.  A transmit queue may be
+ * specified even in the event that the LDC is down (peer endpoint has no
+ * receive queue specified).  Transmission will begin as soon as the peer
+ * endpoint defines a receive queue.
+ *
+ * It is recommended that a guest wait for a transmit queue to empty prior
+ * to reconfiguring it, or un-configuring it.  Re or un-configuring of a
+ * non-empty transmit queue behaves exactly as defined above, however it
+ * is undefined as to how many of the pending entries in the original queue
+ * will be delivered prior to the re-configuration taking effect.
+ * Furthermore, as the queue configuration causes a reset of the head and
+ * tail pointers there is no way for a guest to determine how many entries
+ * have been sent after the configuration operation.
+ */
+#define HV_FAST_LDC_TX_QCONF		0xe0
+
+/* ldc_tx_qinfo()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_TX_QINFO
+ * ARG0:	channel ID
+ * RET0:	status
+ * RET1:	real address base of queue
+ * RET2:	num entries in queue
+ *
+ * Return the configuration info for the transmit queue of LDC endpoint
+ * defined by the given channel ID.  The real address is the currently
+ * defined real address base of the defined queue, and num entries is the
+ * size of the queue in terms of number of entries.
+ *
+ * If the specified channel ID is a valid endpoint number, but no transmit
+ * queue has been defined this service will return success, but with num
+ * entries set to zero and the real address will have an undefined value.
+ */
+#define HV_FAST_LDC_TX_QINFO		0xe1
+
+/* ldc_tx_get_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_TX_GET_STATE
+ * ARG0:	channel ID
+ * RET0:	status
+ * RET1:	head offset
+ * RET2:	tail offset
+ * RET3:	channel state
+ *
+ * Return the transmit state, and the head and tail queue pointers, for
+ * the transmit queue of the LDC endpoint defined by the given channel ID.
+ * The head and tail values are the byte offset of the head and tail
+ * positions of the transmit queue for the specified endpoint.
+ */
+#define HV_FAST_LDC_TX_GET_STATE	0xe2
+
+/* ldc_tx_set_qtail()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_TX_SET_QTAIL
+ * ARG0:	channel ID
+ * ARG1:	tail offset
+ * RET0:	status
+ *
+ * Update the tail pointer for the transmit queue associated with the LDC
+ * endpoint defined by the given channel ID.  The tail offset specified
+ * must be aligned on a 64 byte boundary, and calculated so as to increase
+ * the number of pending entries on the transmit queue.  Any attempt to
+ * decrease the number of pending transmit queue entires is considered
+ * an invalid tail offset and will result in an EINVAL error.
+ *
+ * Since the tail of the transmit queue may not be moved backwards, the
+ * transmit queue may be flushed by configuring a new transmit queue,
+ * whereupon the hypervisor will configure the initial transmit head and
+ * tail pointers to be equal.
+ */
+#define HV_FAST_LDC_TX_SET_QTAIL	0xe3
+
+/* ldc_rx_qconf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_RX_QCONF
+ * ARG0:	channel ID
+ * ARG1:	real address base of queue
+ * ARG2:	num entries in queue
+ * RET0:	status
+ *
+ * Configure receive queue for the LDC endpoint specified by the
+ * given channel ID, to be placed at the given real address, and
+ * be of the given num entries.  Num entries must be a power of two.
+ * The real address base of the queue must be aligned on the queue
+ * size.  Each queue entry is 64-bytes, so for example, a 32 entry
+ * queue must be aligned on a 2048 byte real address boundary.
+ *
+ * The endpoint's transmit queue is un-configured if num entries is zero.
+ *
+ * If a valid receive queue is specified for a local endpoint the LDC is
+ * in the up state for the purpose of transmission to this endpoint.
+ *
+ * The maximum number of entries for each queue for a specific cpu may be
+ * determined from the machine description.
+ *
+ * As receive queue configuration causes a reset of the queue's head and
+ * tail pointers there is no way for a gues to determine how many entries
+ * have been received between a preceeding ldc_get_rx_state() API call
+ * and the completion of the configuration operation.  It should be noted
+ * that datagram delivery is not guarenteed via domain channels anyway,
+ * and therefore any higher protocol should be resilient to datagram
+ * loss if necessary.  However, to overcome this specific race potential
+ * it is recommended, for example, that a higher level protocol be employed
+ * to ensure either retransmission, or ensure that no datagrams are pending
+ * on the peer endpoint's transmit queue prior to the configuration process.
+ */
+#define HV_FAST_LDC_RX_QCONF		0xe4
+
+/* ldc_rx_qinfo()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_RX_QINFO
+ * ARG0:	channel ID
+ * RET0:	status
+ * RET1:	real address base of queue
+ * RET2:	num entries in queue
+ *
+ * Return the configuration info for the receive queue of LDC endpoint
+ * defined by the given channel ID.  The real address is the currently
+ * defined real address base of the defined queue, and num entries is the
+ * size of the queue in terms of number of entries.
+ *
+ * If the specified channel ID is a valid endpoint number, but no receive
+ * queue has been defined this service will return success, but with num
+ * entries set to zero and the real address will have an undefined value.
+ */
+#define HV_FAST_LDC_RX_QINFO		0xe5
+
+/* ldc_rx_get_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_RX_GET_STATE
+ * ARG0:	channel ID
+ * RET0:	status
+ * RET1:	head offset
+ * RET2:	tail offset
+ * RET3:	channel state
+ *
+ * Return the receive state, and the head and tail queue pointers, for
+ * the receive queue of the LDC endpoint defined by the given channel ID.
+ * The head and tail values are the byte offset of the head and tail
+ * positions of the receive queue for the specified endpoint.
+ */
+#define HV_FAST_LDC_RX_GET_STATE	0xe6
+
+/* ldc_rx_set_qhead()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_RX_SET_QHEAD
+ * ARG0:	channel ID
+ * ARG1:	head offset
+ * RET0:	status
+ *
+ * Update the head pointer for the receive queue associated with the LDC
+ * endpoint defined by the given channel ID.  The head offset specified
+ * must be aligned on a 64 byte boundary, and calculated so as to decrease
+ * the number of pending entries on the receive queue.  Any attempt to
+ * increase the number of pending receive queue entires is considered
+ * an invalid head offset and will result in an EINVAL error.
+ *
+ * The receive queue may be flushed by setting the head offset equal
+ * to the current tail offset.
+ */
+#define HV_FAST_LDC_RX_SET_QHEAD	0xe7
+
+/* LDC Map Table Entry.  Each slot is defined by a translation table
+ * entry, as specified by the LDC_MTE_* bits below, and a 64-bit
+ * hypervisor invalidation cookie.
+ */
+#define LDC_MTE_PADDR	0x0fffffffffffe000 /* pa[55:13]          */
+#define LDC_MTE_COPY_W	0x0000000000000400 /* copy write access  */
+#define LDC_MTE_COPY_R	0x0000000000000200 /* copy read access   */
+#define LDC_MTE_IOMMU_W	0x0000000000000100 /* IOMMU write access */
+#define LDC_MTE_IOMMU_R	0x0000000000000080 /* IOMMU read access  */
+#define LDC_MTE_EXEC	0x0000000000000040 /* execute            */
+#define LDC_MTE_WRITE	0x0000000000000020 /* read               */
+#define LDC_MTE_READ	0x0000000000000010 /* write              */
+#define LDC_MTE_SZALL	0x000000000000000f /* page size bits     */
+#define LDC_MTE_SZ16GB	0x0000000000000007 /* 16GB page          */
+#define LDC_MTE_SZ2GB	0x0000000000000006 /* 2GB page           */
+#define LDC_MTE_SZ256MB	0x0000000000000005 /* 256MB page         */
+#define LDC_MTE_SZ32MB	0x0000000000000004 /* 32MB page          */
+#define LDC_MTE_SZ4MB	0x0000000000000003 /* 4MB page           */
+#define LDC_MTE_SZ512K	0x0000000000000002 /* 512K page          */
+#define LDC_MTE_SZ64K	0x0000000000000001 /* 64K page           */
+#define LDC_MTE_SZ8K	0x0000000000000000 /* 8K page            */
+
+#ifndef __ASSEMBLY__
+struct ldc_mtable_entry {
+	unsigned long	mte;
+	unsigned long	cookie;
+};
+#endif
+
+/* ldc_set_map_table()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_SET_MAP_TABLE
+ * ARG0:	channel ID
+ * ARG1:	table real address
+ * ARG2:	num entries
+ * RET0:	status
+ *
+ * Register the MTE table at the given table real address, with the
+ * specified num entries, for the LDC indicated by the given channel
+ * ID.
+ */
+#define HV_FAST_LDC_SET_MAP_TABLE	0xea
+
+/* ldc_get_map_table()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_GET_MAP_TABLE
+ * ARG0:	channel ID
+ * RET0:	status
+ * RET1:	table real address
+ * RET2:	num entries
+ *
+ * Return the configuration of the current mapping table registered
+ * for the given channel ID.
+ */
+#define HV_FAST_LDC_GET_MAP_TABLE	0xeb
+
+#define LDC_COPY_IN	0
+#define LDC_COPY_OUT	1
+
+/* ldc_copy()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_COPY
+ * ARG0:	channel ID
+ * ARG1:	LDC_COPY_* direction code
+ * ARG2:	target real address
+ * ARG3:	local real address
+ * ARG4:	length in bytes
+ * RET0:	status
+ * RET1:	actual length in bytes
+ */
+#define HV_FAST_LDC_COPY		0xec
+
+#define LDC_MEM_READ	1
+#define LDC_MEM_WRITE	2
+#define LDC_MEM_EXEC	4
+
+/* ldc_mapin()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_MAPIN
+ * ARG0:	channel ID
+ * ARG1:	cookie
+ * RET0:	status
+ * RET1:	real address
+ * RET2:	LDC_MEM_* permissions
+ */
+#define HV_FAST_LDC_MAPIN		0xed
+
+/* ldc_unmap()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_UNMAP
+ * ARG0:	real address
+ * RET0:	status
+ */
+#define HV_FAST_LDC_UNMAP		0xee
+
+/* ldc_revoke()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_LDC_REVOKE
+ * ARG0:	channel ID
+ * ARG1:	cookie
+ * ARG2:	ldc_mtable_entry cookie
+ * RET0:	status
+ */
+#define HV_FAST_LDC_REVOKE		0xef
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_ldc_tx_qconf(unsigned long channel,
+					unsigned long ra,
+					unsigned long num_entries);
+extern unsigned long sun4v_ldc_tx_qinfo(unsigned long channel,
+					unsigned long *ra,
+					unsigned long *num_entries);
+extern unsigned long sun4v_ldc_tx_get_state(unsigned long channel,
+					    unsigned long *head_off,
+					    unsigned long *tail_off,
+					    unsigned long *chan_state);
+extern unsigned long sun4v_ldc_tx_set_qtail(unsigned long channel,
+					    unsigned long tail_off);
+extern unsigned long sun4v_ldc_rx_qconf(unsigned long channel,
+					unsigned long ra,
+					unsigned long num_entries);
+extern unsigned long sun4v_ldc_rx_qinfo(unsigned long channel,
+					unsigned long *ra,
+					unsigned long *num_entries);
+extern unsigned long sun4v_ldc_rx_get_state(unsigned long channel,
+					    unsigned long *head_off,
+					    unsigned long *tail_off,
+					    unsigned long *chan_state);
+extern unsigned long sun4v_ldc_rx_set_qhead(unsigned long channel,
+					    unsigned long head_off);
+extern unsigned long sun4v_ldc_set_map_table(unsigned long channel,
+					     unsigned long ra,
+					     unsigned long num_entries);
+extern unsigned long sun4v_ldc_get_map_table(unsigned long channel,
+					     unsigned long *ra,
+					     unsigned long *num_entries);
+extern unsigned long sun4v_ldc_copy(unsigned long channel,
+				    unsigned long dir_code,
+				    unsigned long tgt_raddr,
+				    unsigned long lcl_raddr,
+				    unsigned long len,
+				    unsigned long *actual_len);
+extern unsigned long sun4v_ldc_mapin(unsigned long channel,
+				     unsigned long cookie,
+				     unsigned long *ra,
+				     unsigned long *perm);
+extern unsigned long sun4v_ldc_unmap(unsigned long ra);
+extern unsigned long sun4v_ldc_revoke(unsigned long channel,
+				      unsigned long cookie,
+				      unsigned long mte_cookie);
+#endif
+
+/* Performance counter services.  */
+
+#define HV_PERF_JBUS_PERF_CTRL_REG	0x00
+#define HV_PERF_JBUS_PERF_CNT_REG	0x01
+#define HV_PERF_DRAM_PERF_CTRL_REG_0	0x02
+#define HV_PERF_DRAM_PERF_CNT_REG_0	0x03
+#define HV_PERF_DRAM_PERF_CTRL_REG_1	0x04
+#define HV_PERF_DRAM_PERF_CNT_REG_1	0x05
+#define HV_PERF_DRAM_PERF_CTRL_REG_2	0x06
+#define HV_PERF_DRAM_PERF_CNT_REG_2	0x07
+#define HV_PERF_DRAM_PERF_CTRL_REG_3	0x08
+#define HV_PERF_DRAM_PERF_CNT_REG_3	0x09
+
+/* get_perfreg()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_GET_PERFREG
+ * ARG0:	performance reg number
+ * RET0:	status
+ * RET1:	performance reg value
+ * ERRORS:	EINVAL		Invalid performance register number
+ *		ENOACCESS	No access allowed to performance counters
+ *
+ * Read the value of the given DRAM/JBUS performance counter/control register.
+ */
+#define HV_FAST_GET_PERFREG		0x100
+
+/* set_perfreg()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_SET_PERFREG
+ * ARG0:	performance reg number
+ * ARG1:	performance reg value
+ * RET0:	status
+ * ERRORS:	EINVAL		Invalid performance register number
+ *		ENOACCESS	No access allowed to performance counters
+ *
+ * Write the given performance reg value to the given DRAM/JBUS
+ * performance counter/control register.
+ */
+#define HV_FAST_SET_PERFREG		0x101
+
+/* MMU statistics services.
+ *
+ * The hypervisor maintains MMU statistics and privileged code provides
+ * a buffer where these statistics can be collected.  It is continually
+ * updated once configured.  The layout is as follows:
+ */
+#ifndef __ASSEMBLY__
+struct hv_mmu_statistics {
+	unsigned long immu_tsb_hits_ctx0_8k_tte;
+	unsigned long immu_tsb_ticks_ctx0_8k_tte;
+	unsigned long immu_tsb_hits_ctx0_64k_tte;
+	unsigned long immu_tsb_ticks_ctx0_64k_tte;
+	unsigned long __reserved1[2];
+	unsigned long immu_tsb_hits_ctx0_4mb_tte;
+	unsigned long immu_tsb_ticks_ctx0_4mb_tte;
+	unsigned long __reserved2[2];
+	unsigned long immu_tsb_hits_ctx0_256mb_tte;
+	unsigned long immu_tsb_ticks_ctx0_256mb_tte;
+	unsigned long __reserved3[4];
+	unsigned long immu_tsb_hits_ctxnon0_8k_tte;
+	unsigned long immu_tsb_ticks_ctxnon0_8k_tte;
+	unsigned long immu_tsb_hits_ctxnon0_64k_tte;
+	unsigned long immu_tsb_ticks_ctxnon0_64k_tte;
+	unsigned long __reserved4[2];
+	unsigned long immu_tsb_hits_ctxnon0_4mb_tte;
+	unsigned long immu_tsb_ticks_ctxnon0_4mb_tte;
+	unsigned long __reserved5[2];
+	unsigned long immu_tsb_hits_ctxnon0_256mb_tte;
+	unsigned long immu_tsb_ticks_ctxnon0_256mb_tte;
+	unsigned long __reserved6[4];
+	unsigned long dmmu_tsb_hits_ctx0_8k_tte;
+	unsigned long dmmu_tsb_ticks_ctx0_8k_tte;
+	unsigned long dmmu_tsb_hits_ctx0_64k_tte;
+	unsigned long dmmu_tsb_ticks_ctx0_64k_tte;
+	unsigned long __reserved7[2];
+	unsigned long dmmu_tsb_hits_ctx0_4mb_tte;
+	unsigned long dmmu_tsb_ticks_ctx0_4mb_tte;
+	unsigned long __reserved8[2];
+	unsigned long dmmu_tsb_hits_ctx0_256mb_tte;
+	unsigned long dmmu_tsb_ticks_ctx0_256mb_tte;
+	unsigned long __reserved9[4];
+	unsigned long dmmu_tsb_hits_ctxnon0_8k_tte;
+	unsigned long dmmu_tsb_ticks_ctxnon0_8k_tte;
+	unsigned long dmmu_tsb_hits_ctxnon0_64k_tte;
+	unsigned long dmmu_tsb_ticks_ctxnon0_64k_tte;
+	unsigned long __reserved10[2];
+	unsigned long dmmu_tsb_hits_ctxnon0_4mb_tte;
+	unsigned long dmmu_tsb_ticks_ctxnon0_4mb_tte;
+	unsigned long __reserved11[2];
+	unsigned long dmmu_tsb_hits_ctxnon0_256mb_tte;
+	unsigned long dmmu_tsb_ticks_ctxnon0_256mb_tte;
+	unsigned long __reserved12[4];
+};
+#endif
+
+/* mmustat_conf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMUSTAT_CONF
+ * ARG0:	real address
+ * RET0:	status
+ * RET1:	real address
+ * ERRORS:	ENORADDR	Invalid real address
+ *		EBADALIGN	Real address not aligned on 64-byte boundary
+ *		EBADTRAP	API not supported on this processor
+ *
+ * Enable MMU statistic gathering using the buffer at the given real
+ * address on the current virtual CPU.  The new buffer real address
+ * is given in ARG1, and the previously specified buffer real address
+ * is returned in RET1, or is returned as zero for the first invocation.
+ *
+ * If the passed in real address argument is zero, this will disable
+ * MMU statistic collection on the current virtual CPU.  If an error is
+ * returned then no statistics are collected.
+ *
+ * The buffer contents should be initialized to all zeros before being
+ * given to the hypervisor or else the statistics will be meaningless.
+ */
+#define HV_FAST_MMUSTAT_CONF		0x102
+
+/* mmustat_info()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MMUSTAT_INFO
+ * RET0:	status
+ * RET1:	real address
+ * ERRORS:	EBADTRAP	API not supported on this processor
+ *
+ * Return the current state and real address of the currently configured
+ * MMU statistics buffer on the current virtual CPU.
+ */
+#define HV_FAST_MMUSTAT_INFO		0x103
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_mmustat_conf(unsigned long ra, unsigned long *orig_ra);
+extern unsigned long sun4v_mmustat_info(unsigned long *ra);
+#endif
+
+/* NCS crypto services  */
+
+/* ncs_request() sub-function numbers */
+#define HV_NCS_QCONF			0x01
+#define HV_NCS_QTAIL_UPDATE		0x02
+
+#ifndef __ASSEMBLY__
+struct hv_ncs_queue_entry {
+	/* MAU Control Register */
+	unsigned long	mau_control;
+#define MAU_CONTROL_INV_PARITY	0x0000000000002000
+#define MAU_CONTROL_STRAND	0x0000000000001800
+#define MAU_CONTROL_BUSY	0x0000000000000400
+#define MAU_CONTROL_INT		0x0000000000000200
+#define MAU_CONTROL_OP		0x00000000000001c0
+#define MAU_CONTROL_OP_SHIFT	6
+#define MAU_OP_LOAD_MA_MEMORY	0x0
+#define MAU_OP_STORE_MA_MEMORY	0x1
+#define MAU_OP_MODULAR_MULT	0x2
+#define MAU_OP_MODULAR_REDUCE	0x3
+#define MAU_OP_MODULAR_EXP_LOOP	0x4
+#define MAU_CONTROL_LEN		0x000000000000003f
+#define MAU_CONTROL_LEN_SHIFT	0
+
+	/* Real address of bytes to load or store bytes
+	 * into/out-of the MAU.
+	 */
+	unsigned long	mau_mpa;
+
+	/* Modular Arithmetic MA Offset Register.  */
+	unsigned long	mau_ma;
+
+	/* Modular Arithmetic N Prime Register.  */
+	unsigned long	mau_np;
+};
+
+struct hv_ncs_qconf_arg {
+	unsigned long	mid;      /* MAU ID, 1 per core on Niagara */
+	unsigned long	base;     /* Real address base of queue */
+	unsigned long	end;	  /* Real address end of queue */
+	unsigned long	num_ents; /* Number of entries in queue */
+};
+
+struct hv_ncs_qtail_update_arg {
+	unsigned long	mid;      /* MAU ID, 1 per core on Niagara */
+	unsigned long	tail;     /* New tail index to use */
+	unsigned long	syncflag; /* only SYNCFLAG_SYNC is implemented */
+#define HV_NCS_SYNCFLAG_SYNC	0x00
+#define HV_NCS_SYNCFLAG_ASYNC	0x01
+};
+#endif
+
+/* ncs_request()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_REQUEST
+ * ARG0:	NCS sub-function
+ * ARG1:	sub-function argument real address
+ * ARG2:	size in bytes of sub-function argument
+ * RET0:	status
+ *
+ * The MAU chip of the Niagara processor is not directly accessible
+ * to privileged code, instead it is programmed indirectly via this
+ * hypervisor API.
+ *
+ * The interfaces defines a queue of MAU operations to perform.
+ * Privileged code registers a queue with the hypervisor by invoking
+ * this HVAPI with the HV_NCS_QCONF sub-function, which defines the
+ * base, end, and number of entries of the queue.  Each queue entry
+ * contains a MAU register struct block.
+ *
+ * The privileged code then proceeds to add entries to the queue and
+ * then invoke the HV_NCS_QTAIL_UPDATE sub-function.  Since only
+ * synchronous operations are supported by the current hypervisor,
+ * HV_NCS_QTAIL_UPDATE will run all the pending queue entries to
+ * completion and return HV_EOK, or return an error code.
+ *
+ * The real address of the sub-function argument must be aligned on at
+ * least an 8-byte boundary.
+ *
+ * The tail argument of HV_NCS_QTAIL_UPDATE is an index, not a byte
+ * offset, into the queue and must be less than or equal the 'num_ents'
+ * argument given in the HV_NCS_QCONF call.
+ */
+#define HV_FAST_NCS_REQUEST		0x110
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_ncs_request(unsigned long request,
+				       unsigned long arg_ra,
+				       unsigned long arg_size);
+#endif
+
+#define HV_FAST_FIRE_GET_PERFREG	0x120
+#define HV_FAST_FIRE_SET_PERFREG	0x121
+
+/* Function numbers for HV_CORE_TRAP.  */
+#define HV_CORE_SET_VER			0x00
+#define HV_CORE_PUTCHAR			0x01
+#define HV_CORE_EXIT			0x02
+#define HV_CORE_GET_VER			0x03
+
+/* Hypervisor API groups for use with HV_CORE_SET_VER and
+ * HV_CORE_GET_VER.
+ */
+#define HV_GRP_SUN4V			0x0000
+#define HV_GRP_CORE			0x0001
+#define HV_GRP_INTR			0x0002
+#define HV_GRP_SOFT_STATE		0x0003
+#define HV_GRP_PCI			0x0100
+#define HV_GRP_LDOM			0x0101
+#define HV_GRP_SVC_CHAN			0x0102
+#define HV_GRP_NCS			0x0103
+#define HV_GRP_NIAG_PERF		0x0200
+#define HV_GRP_FIRE_PERF		0x0201
+#define HV_GRP_DIAG			0x0300
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_get_version(unsigned long group,
+				       unsigned long *major,
+				       unsigned long *minor);
+extern unsigned long sun4v_set_version(unsigned long group,
+				       unsigned long major,
+				       unsigned long minor,
+				       unsigned long *actual_minor);
+
+extern int sun4v_hvapi_register(unsigned long group, unsigned long major,
+				unsigned long *minor);
+extern void sun4v_hvapi_unregister(unsigned long group);
+extern int sun4v_hvapi_get(unsigned long group,
+			   unsigned long *major,
+			   unsigned long *minor);
+extern void sun4v_hvapi_init(void);
+#endif
+
+#endif /* !(_SPARC64_HYPERVISOR_H) */
diff --git a/include/asm-sparc/intr_queue.h b/include/asm-sparc/intr_queue.h
new file mode 100644
index 0000000..206077d
--- /dev/null
+++ b/include/asm-sparc/intr_queue.h
@@ -0,0 +1,15 @@
+#ifndef _SPARC64_INTR_QUEUE_H
+#define _SPARC64_INTR_QUEUE_H
+
+/* Sun4v interrupt queue registers, accessed via ASI_QUEUE.  */
+
+#define INTRQ_CPU_MONDO_HEAD	  0x3c0 /* CPU mondo head	          */
+#define INTRQ_CPU_MONDO_TAIL	  0x3c8 /* CPU mondo tail	          */
+#define INTRQ_DEVICE_MONDO_HEAD	  0x3d0 /* Device mondo head	          */
+#define INTRQ_DEVICE_MONDO_TAIL	  0x3d8 /* Device mondo tail	          */
+#define INTRQ_RESUM_MONDO_HEAD	  0x3e0 /* Resumable error mondo head     */
+#define INTRQ_RESUM_MONDO_TAIL	  0x3e8 /* Resumable error mondo tail     */
+#define INTRQ_NONRESUM_MONDO_HEAD 0x3f0 /* Non-resumable error mondo head */
+#define INTRQ_NONRESUM_MONDO_TAIL 0x3f8 /* Non-resumable error mondo head */
+
+#endif /* !(_SPARC64_INTR_QUEUE_H) */
diff --git a/include/asm-sparc/kprobes.h b/include/asm-sparc/kprobes.h
new file mode 100644
index 0000000..5879d71
--- /dev/null
+++ b/include/asm-sparc/kprobes.h
@@ -0,0 +1,49 @@
+#ifndef _SPARC64_KPROBES_H
+#define _SPARC64_KPROBES_H
+
+#include <linux/types.h>
+#include <linux/percpu.h>
+
+typedef u32 kprobe_opcode_t;
+
+#define BREAKPOINT_INSTRUCTION   0x91d02070 /* ta 0x70 */
+#define BREAKPOINT_INSTRUCTION_2 0x91d02071 /* ta 0x71 */
+#define MAX_INSN_SIZE 2
+
+#define kretprobe_blacklist_size 0
+
+#define arch_remove_kprobe(p)	do {} while (0)
+
+#define flush_insn_slot(p)		\
+do { 	flushi(&(p)->ainsn.insn[0]);	\
+	flushi(&(p)->ainsn.insn[1]);	\
+} while (0)
+
+void kretprobe_trampoline(void);
+
+/* Architecture specific copy of original instruction*/
+struct arch_specific_insn {
+	/* copy of the original instruction */
+	kprobe_opcode_t insn[MAX_INSN_SIZE];
+};
+
+struct prev_kprobe {
+	struct kprobe *kp;
+	unsigned long status;
+	unsigned long orig_tnpc;
+	unsigned long orig_tstate_pil;
+};
+
+/* per-cpu kprobe control block */
+struct kprobe_ctlblk {
+	unsigned long kprobe_status;
+	unsigned long kprobe_orig_tnpc;
+	unsigned long kprobe_orig_tstate_pil;
+	struct pt_regs jprobe_saved_regs;
+	struct prev_kprobe prev_kprobe;
+};
+
+extern int kprobe_exceptions_notify(struct notifier_block *self,
+				    unsigned long val, void *data);
+extern int kprobe_fault_handler(struct pt_regs *regs, int trapnr);
+#endif /* _SPARC64_KPROBES_H */
diff --git a/include/asm-sparc/ldc.h b/include/asm-sparc/ldc.h
new file mode 100644
index 0000000..bdb524a
--- /dev/null
+++ b/include/asm-sparc/ldc.h
@@ -0,0 +1,138 @@
+#ifndef _SPARC64_LDC_H
+#define _SPARC64_LDC_H
+
+#include <asm/hypervisor.h>
+
+extern int ldom_domaining_enabled;
+extern void ldom_set_var(const char *var, const char *value);
+extern void ldom_reboot(const char *boot_command);
+extern void ldom_power_off(void);
+
+/* The event handler will be evoked when link state changes
+ * or data becomes available on the receive side.
+ *
+ * For non-RAW links, if the LDC_EVENT_RESET event arrives the
+ * driver should reset all of it's internal state and reinvoke
+ * ldc_connect() to try and bring the link up again.
+ *
+ * For RAW links, ldc_connect() is not used.  Instead the driver
+ * just waits for the LDC_EVENT_UP event.
+ */
+struct ldc_channel_config {
+	void (*event)(void *arg, int event);
+
+	u32			mtu;
+	unsigned int		rx_irq;
+	unsigned int		tx_irq;
+	u8			mode;
+#define LDC_MODE_RAW		0x00
+#define LDC_MODE_UNRELIABLE	0x01
+#define LDC_MODE_RESERVED	0x02
+#define LDC_MODE_STREAM		0x03
+
+	u8			debug;
+#define LDC_DEBUG_HS		0x01
+#define LDC_DEBUG_STATE		0x02
+#define LDC_DEBUG_RX		0x04
+#define LDC_DEBUG_TX		0x08
+#define LDC_DEBUG_DATA		0x10
+};
+
+#define LDC_EVENT_RESET		0x01
+#define LDC_EVENT_UP		0x02
+#define LDC_EVENT_DATA_READY	0x04
+
+#define LDC_STATE_INVALID	0x00
+#define LDC_STATE_INIT		0x01
+#define LDC_STATE_BOUND		0x02
+#define LDC_STATE_READY		0x03
+#define LDC_STATE_CONNECTED	0x04
+
+struct ldc_channel;
+
+/* Allocate state for a channel.  */
+extern struct ldc_channel *ldc_alloc(unsigned long id,
+				     const struct ldc_channel_config *cfgp,
+				     void *event_arg);
+
+/* Shut down and free state for a channel.  */
+extern void ldc_free(struct ldc_channel *lp);
+
+/* Register TX and RX queues of the link with the hypervisor.  */
+extern int ldc_bind(struct ldc_channel *lp, const char *name);
+
+/* For non-RAW protocols we need to complete a handshake before
+ * communication can proceed.  ldc_connect() does that, if the
+ * handshake completes successfully, an LDC_EVENT_UP event will
+ * be sent up to the driver.
+ */
+extern int ldc_connect(struct ldc_channel *lp);
+extern int ldc_disconnect(struct ldc_channel *lp);
+
+extern int ldc_state(struct ldc_channel *lp);
+
+/* Read and write operations.  Only valid when the link is up.  */
+extern int ldc_write(struct ldc_channel *lp, const void *buf,
+		     unsigned int size);
+extern int ldc_read(struct ldc_channel *lp, void *buf, unsigned int size);
+
+#define LDC_MAP_SHADOW	0x01
+#define LDC_MAP_DIRECT	0x02
+#define LDC_MAP_IO	0x04
+#define LDC_MAP_R	0x08
+#define LDC_MAP_W	0x10
+#define LDC_MAP_X	0x20
+#define LDC_MAP_RW	(LDC_MAP_R | LDC_MAP_W)
+#define LDC_MAP_RWX	(LDC_MAP_R | LDC_MAP_W | LDC_MAP_X)
+#define LDC_MAP_ALL	0x03f
+
+struct ldc_trans_cookie {
+	u64			cookie_addr;
+	u64			cookie_size;
+};
+
+struct scatterlist;
+extern int ldc_map_sg(struct ldc_channel *lp,
+		      struct scatterlist *sg, int num_sg,
+		      struct ldc_trans_cookie *cookies, int ncookies,
+		      unsigned int map_perm);
+
+extern int ldc_map_single(struct ldc_channel *lp,
+			  void *buf, unsigned int len,
+			  struct ldc_trans_cookie *cookies, int ncookies,
+			  unsigned int map_perm);
+
+extern void ldc_unmap(struct ldc_channel *lp, struct ldc_trans_cookie *cookies,
+		      int ncookies);
+
+extern int ldc_copy(struct ldc_channel *lp, int copy_dir,
+		    void *buf, unsigned int len, unsigned long offset,
+		    struct ldc_trans_cookie *cookies, int ncookies);
+
+static inline int ldc_get_dring_entry(struct ldc_channel *lp,
+				      void *buf, unsigned int len,
+				      unsigned long offset,
+				      struct ldc_trans_cookie *cookies,
+				      int ncookies)
+{
+	return ldc_copy(lp, LDC_COPY_IN, buf, len, offset, cookies, ncookies);
+}
+
+static inline int ldc_put_dring_entry(struct ldc_channel *lp,
+				      void *buf, unsigned int len,
+				      unsigned long offset,
+				      struct ldc_trans_cookie *cookies,
+				      int ncookies)
+{
+	return ldc_copy(lp, LDC_COPY_OUT, buf, len, offset, cookies, ncookies);
+}
+
+extern void *ldc_alloc_exp_dring(struct ldc_channel *lp, unsigned int len,
+				 struct ldc_trans_cookie *cookies,
+				 int *ncookies, unsigned int map_perm);
+
+extern void ldc_free_exp_dring(struct ldc_channel *lp, void *buf,
+			       unsigned int len,
+			       struct ldc_trans_cookie *cookies, int ncookies);
+
+#endif /* _SPARC64_LDC_H */
diff --git a/include/asm-sparc/lmb.h b/include/asm-sparc/lmb.h
new file mode 100644
index 0000000..6a352cb
--- /dev/null
+++ b/include/asm-sparc/lmb.h
@@ -0,0 +1,10 @@
+#ifndef _SPARC64_LMB_H
+#define _SPARC64_LMB_H
+
+#include <asm/oplib.h>
+
+#define LMB_DBG(fmt...) prom_printf(fmt)
+
+#define LMB_REAL_LIMIT	0
+
+#endif /* !(_SPARC64_LMB_H) */
diff --git a/include/asm-sparc/lsu.h b/include/asm-sparc/lsu.h
new file mode 100644
index 0000000..7190f8d
--- /dev/null
+++ b/include/asm-sparc/lsu.h
@@ -0,0 +1,19 @@
+#ifndef _SPARC64_LSU_H
+#define _SPARC64_LSU_H
+
+#include <linux/const.h>
+
+/* LSU Control Register */
+#define LSU_CONTROL_PM _AC(0x000001fe00000000,UL) /* Phys-watchpoint byte mask*/
+#define LSU_CONTROL_VM _AC(0x00000001fe000000,UL) /* Virt-watchpoint byte mask*/
+#define LSU_CONTROL_PR _AC(0x0000000001000000,UL) /* Phys-rd watchpoint enable*/
+#define LSU_CONTROL_PW _AC(0x0000000000800000,UL) /* Phys-wr watchpoint enable*/
+#define LSU_CONTROL_VR _AC(0x0000000000400000,UL) /* Virt-rd watchpoint enable*/
+#define LSU_CONTROL_VW _AC(0x0000000000200000,UL) /* Virt-wr watchpoint enable*/
+#define LSU_CONTROL_FM _AC(0x00000000000ffff0,UL) /* Parity mask enables.     */
+#define LSU_CONTROL_DM _AC(0x0000000000000008,UL) /* Data MMU enable.         */
+#define LSU_CONTROL_IM _AC(0x0000000000000004,UL) /* Instruction MMU enable.  */
+#define LSU_CONTROL_DC _AC(0x0000000000000002,UL) /* Data cache enable.       */
+#define LSU_CONTROL_IC _AC(0x0000000000000001,UL) /* Instruction cache enable.*/
+
+#endif /* !(_SPARC64_LSU_H) */
diff --git a/include/asm-sparc/mdesc.h b/include/asm-sparc/mdesc.h
new file mode 100644
index 0000000..1acc727
--- /dev/null
+++ b/include/asm-sparc/mdesc.h
@@ -0,0 +1,78 @@
+#ifndef _SPARC64_MDESC_H
+#define _SPARC64_MDESC_H
+
+#include <linux/types.h>
+#include <linux/cpumask.h>
+#include <asm/prom.h>
+
+struct mdesc_handle;
+
+/* Machine description operations are to be surrounded by grab and
+ * release calls.  The mdesc_handle returned from the grab is
+ * the first argument to all of the operational calls that work
+ * on mdescs.
+ */
+extern struct mdesc_handle *mdesc_grab(void);
+extern void mdesc_release(struct mdesc_handle *);
+
+#define MDESC_NODE_NULL		(~(u64)0)
+
+extern u64 mdesc_node_by_name(struct mdesc_handle *handle,
+			      u64 from_node, const char *name);
+#define mdesc_for_each_node_by_name(__hdl, __node, __name) \
+	for (__node = mdesc_node_by_name(__hdl, MDESC_NODE_NULL, __name); \
+	     (__node) != MDESC_NODE_NULL; \
+	     __node = mdesc_node_by_name(__hdl, __node, __name))
+
+/* Access to property values returned from mdesc_get_property() are
+ * only valid inside of a mdesc_grab()/mdesc_release() sequence.
+ * Once mdesc_release() is called, the memory backed up by these
+ * pointers may reference freed up memory.
+ *
+ * Therefore callers must make copies of any property values
+ * they need.
+ *
+ * These same rules apply to mdesc_node_name().
+ */
+extern const void *mdesc_get_property(struct mdesc_handle *handle,
+				      u64 node, const char *name, int *lenp);
+extern const char *mdesc_node_name(struct mdesc_handle *hp, u64 node);
+
+/* MD arc iteration, the standard sequence is:
+ *
+ *	unsigned long arc;
+ *	mdesc_for_each_arc(arc, handle, node, MDESC_ARC_TYPE_{FWD,BACK}) {
+ *		unsigned long target = mdesc_arc_target(handle, arc);
+ *		...
+ *	}
+ */
+
+#define MDESC_ARC_TYPE_FWD	"fwd"
+#define MDESC_ARC_TYPE_BACK	"back"
+
+extern u64 mdesc_next_arc(struct mdesc_handle *handle, u64 from,
+			  const char *arc_type);
+#define mdesc_for_each_arc(__arc, __hdl, __node, __type) \
+	for (__arc = mdesc_next_arc(__hdl, __node, __type); \
+	     (__arc) != MDESC_NODE_NULL; \
+	     __arc = mdesc_next_arc(__hdl, __arc, __type))
+
+extern u64 mdesc_arc_target(struct mdesc_handle *hp, u64 arc);
+
+extern void mdesc_update(void);
+
+struct mdesc_notifier_client {
+	void (*add)(struct mdesc_handle *handle, u64 node);
+	void (*remove)(struct mdesc_handle *handle, u64 node);
+
+	const char			*node_name;
+	struct mdesc_notifier_client	*next;
+};
+
+extern void mdesc_register_notifier(struct mdesc_notifier_client *client);
+
+extern void mdesc_fill_in_cpu_data(cpumask_t mask);
+
+extern void sun4v_mdesc_init(void);
+
+#endif
diff --git a/include/asm-sparc/mmzone.h b/include/asm-sparc/mmzone.h
new file mode 100644
index 0000000..ebf5986
--- /dev/null
+++ b/include/asm-sparc/mmzone.h
@@ -0,0 +1,17 @@
+#ifndef _SPARC64_MMZONE_H
+#define _SPARC64_MMZONE_H
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+
+extern struct pglist_data *node_data[];
+
+#define NODE_DATA(nid)		(node_data[nid])
+#define node_start_pfn(nid)	(NODE_DATA(nid)->node_start_pfn)
+#define node_end_pfn(nid)	(NODE_DATA(nid)->node_end_pfn)
+
+extern int numa_cpu_lookup_table[];
+extern cpumask_t numa_cpumask_lookup_table[];
+
+#endif /* CONFIG_NEED_MULTIPLE_NODES */
+
+#endif /* _SPARC64_MMZONE_H */
diff --git a/include/asm-sparc/ns87303.h b/include/asm-sparc/ns87303.h
new file mode 100644
index 0000000..686defe
--- /dev/null
+++ b/include/asm-sparc/ns87303.h
@@ -0,0 +1,118 @@
+/* ns87303.h: Configuration Register Description for the
+ *            National Semiconductor PC87303 (SuperIO).
+ *
+ * Copyright (C) 1997  Eddie C. Dost  (ecd@skynet.be)
+ */
+
+#ifndef _SPARC_NS87303_H
+#define _SPARC_NS87303_H 1
+
+/*
+ * Control Register Index Values
+ */
+#define FER	0x00
+#define FAR	0x01
+#define PTR	0x02
+#define FCR	0x03
+#define PCR	0x04
+#define KRR	0x05
+#define PMC	0x06
+#define TUP	0x07
+#define SID	0x08
+#define ASC	0x09
+#define CS0CF0	0x0a
+#define CS0CF1	0x0b
+#define CS1CF0	0x0c
+#define CS1CF1	0x0d
+
+/* Function Enable Register (FER) bits */
+#define FER_EDM		0x10	/* Encoded Drive and Motor pin information   */
+
+/* Function Address Register (FAR) bits */
+#define FAR_LPT_MASK	0x03
+#define FAR_LPTB	0x00
+#define FAR_LPTA	0x01
+#define FAR_LPTC	0x02
+
+/* Power and Test Register (PTR) bits */
+#define PTR_LPTB_IRQ7	0x08
+#define PTR_LEVEL_IRQ	0x80	/* When not ECP/EPP: Use level IRQ           */
+#define PTR_LPT_REG_DIR	0x80	/* When ECP/EPP: LPT CTR controlls direction */
+				/*               of the parallel port	     */
+
+/* Function Control Register (FCR) bits */
+#define FCR_LDE		0x10	/* Logical Drive Exchange                    */
+#define FCR_ZWS_ENA	0x20	/* Enable short host read/write in ECP/EPP   */
+
+/* Printer Control Register (PCR) bits */
+#define PCR_EPP_ENABLE	0x01
+#define PCR_EPP_IEEE	0x02	/* Enable EPP Version 1.9 (IEEE 1284)        */
+#define PCR_ECP_ENABLE	0x04
+#define PCR_ECP_CLK_ENA	0x08	/* If 0 ECP Clock is stopped on Power down   */
+#define PCR_IRQ_POLAR	0x20	/* If 0 IRQ is level high or negative pulse, */
+				/* if 1 polarity is inverted                 */
+#define PCR_IRQ_ODRAIN	0x40	/* If 1, IRQ is open drain                   */
+
+/* Tape UARTs and Parallel Port Config Register (TUP) bits */
+#define TUP_EPP_TIMO	0x02	/* Enable EPP timeout IRQ                    */
+
+/* Advanced SuperIO Config Register (ASC) bits */
+#define ASC_LPT_IRQ7	0x01	/* Always use IRQ7 for LPT                  */
+#define ASC_DRV2_SEL	0x02	/* Logical Drive Exchange controlled by TDR  */
+
+#define FER_RESERVED	0x00
+#define FAR_RESERVED	0x00
+#define PTR_RESERVED	0x73
+#define FCR_RESERVED	0xc4
+#define PCR_RESERVED	0x10
+#define KRR_RESERVED	0x00
+#define PMC_RESERVED	0x98
+#define TUP_RESERVED	0xfb
+#define SIP_RESERVED	0x00
+#define ASC_RESERVED	0x18
+#define CS0CF0_RESERVED	0x00
+#define CS0CF1_RESERVED	0x08
+#define CS1CF0_RESERVED	0x00
+#define CS1CF1_RESERVED	0x08
+
+#ifdef __KERNEL__
+
+#include <linux/spinlock.h>
+
+#include <asm/system.h>
+#include <asm/io.h>
+
+extern spinlock_t ns87303_lock;
+
+static inline int ns87303_modify(unsigned long port, unsigned int index,
+				     unsigned char clr, unsigned char set)
+{
+	static unsigned char reserved[] = {
+		FER_RESERVED, FAR_RESERVED, PTR_RESERVED, FCR_RESERVED,
+		PCR_RESERVED, KRR_RESERVED, PMC_RESERVED, TUP_RESERVED,
+		SIP_RESERVED, ASC_RESERVED, CS0CF0_RESERVED, CS0CF1_RESERVED,
+		CS1CF0_RESERVED, CS1CF1_RESERVED
+	};
+	unsigned long flags;
+	unsigned char value;
+
+	if (index > 0x0d)
+		return -EINVAL;
+
+	spin_lock_irqsave(&ns87303_lock, flags);
+
+	outb(index, port);
+	value = inb(port + 1);
+	value &= ~(reserved[index] | clr);
+	value |= set;
+	outb(value, port + 1);
+	outb(value, port + 1);
+
+	spin_unlock_irqrestore(&ns87303_lock, flags);
+
+	return 0;
+}
+
+#endif /* __KERNEL__ */
+
+#endif /* !(_SPARC_NS87303_H) */
diff --git a/include/asm-sparc/parport.h b/include/asm-sparc/parport.h
new file mode 100644
index 0000000..7818b25
--- /dev/null
+++ b/include/asm-sparc/parport.h
@@ -0,0 +1,246 @@
+/* parport.h: sparc64 specific parport initialization and dma.
+ *
+ * Copyright (C) 1999  Eddie C. Dost  (ecd@skynet.be)
+ */
+
+#ifndef _ASM_SPARC64_PARPORT_H
+#define _ASM_SPARC64_PARPORT_H 1
+
+#include <asm/ebus.h>
+#include <asm/ns87303.h>
+#include <asm/of_device.h>
+#include <asm/prom.h>
+
+#define PARPORT_PC_MAX_PORTS	PARPORT_MAX
+
+/*
+ * While sparc64 doesn't have an ISA DMA API, we provide something that looks
+ * close enough to make parport_pc happy
+ */
+#define HAS_DMA
+
+static DEFINE_SPINLOCK(dma_spin_lock);
+
+#define claim_dma_lock() \
+({	unsigned long flags; \
+	spin_lock_irqsave(&dma_spin_lock, flags); \
+	flags; \
+})
+
+#define release_dma_lock(__flags) \
+	spin_unlock_irqrestore(&dma_spin_lock, __flags);
+
+static struct sparc_ebus_info {
+	struct ebus_dma_info info;
+	unsigned int addr;
+	unsigned int count;
+	int lock;
+
+	struct parport *port;
+} sparc_ebus_dmas[PARPORT_PC_MAX_PORTS];
+
+static DECLARE_BITMAP(dma_slot_map, PARPORT_PC_MAX_PORTS);
+
+static inline int request_dma(unsigned int dmanr, const char *device_id)
+{
+	if (dmanr >= PARPORT_PC_MAX_PORTS)
+		return -EINVAL;
+	if (xchg(&sparc_ebus_dmas[dmanr].lock, 1) != 0)
+		return -EBUSY;
+	return 0;
+}
+
+static inline void free_dma(unsigned int dmanr)
+{
+	if (dmanr >= PARPORT_PC_MAX_PORTS) {
+		printk(KERN_WARNING "Trying to free DMA%d\n", dmanr);
+		return;
+	}
+	if (xchg(&sparc_ebus_dmas[dmanr].lock, 0) == 0) {
+		printk(KERN_WARNING "Trying to free free DMA%d\n", dmanr);
+		return;
+	}
+}
+
+static inline void enable_dma(unsigned int dmanr)
+{
+	ebus_dma_enable(&sparc_ebus_dmas[dmanr].info, 1);
+
+	if (ebus_dma_request(&sparc_ebus_dmas[dmanr].info,
+			     sparc_ebus_dmas[dmanr].addr,
+			     sparc_ebus_dmas[dmanr].count))
+		BUG();
+}
+
+static inline void disable_dma(unsigned int dmanr)
+{
+	ebus_dma_enable(&sparc_ebus_dmas[dmanr].info, 0);
+}
+
+static inline void clear_dma_ff(unsigned int dmanr)
+{
+	/* nothing */
+}
+
+static inline void set_dma_mode(unsigned int dmanr, char mode)
+{
+	ebus_dma_prepare(&sparc_ebus_dmas[dmanr].info, (mode != DMA_MODE_WRITE));
+}
+
+static inline void set_dma_addr(unsigned int dmanr, unsigned int addr)
+{
+	sparc_ebus_dmas[dmanr].addr = addr;
+}
+
+static inline void set_dma_count(unsigned int dmanr, unsigned int count)
+{
+	sparc_ebus_dmas[dmanr].count = count;
+}
+
+static inline unsigned int get_dma_residue(unsigned int dmanr)
+{
+	return ebus_dma_residue(&sparc_ebus_dmas[dmanr].info);
+}
+
+static int __devinit ecpp_probe(struct of_device *op, const struct of_device_id *match)
+{
+	unsigned long base = op->resource[0].start;
+	unsigned long config = op->resource[1].start;
+	unsigned long d_base = op->resource[2].start;
+	unsigned long d_len;
+	struct device_node *parent;
+	struct parport *p;
+	int slot, err;
+
+	parent = op->node->parent;
+	if (!strcmp(parent->name, "dma")) {
+		p = parport_pc_probe_port(base, base + 0x400,
+					  op->irqs[0], PARPORT_DMA_NOFIFO,
+					  op->dev.parent->parent);
+		if (!p)
+			return -ENOMEM;
+		dev_set_drvdata(&op->dev, p);
+		return 0;
+	}
+
+	for (slot = 0; slot < PARPORT_PC_MAX_PORTS; slot++) {
+		if (!test_and_set_bit(slot, dma_slot_map))
+			break;
+	}
+	err = -ENODEV;
+	if (slot >= PARPORT_PC_MAX_PORTS)
+		goto out_err;
+
+	spin_lock_init(&sparc_ebus_dmas[slot].info.lock);
+
+	d_len = (op->resource[2].end - d_base) + 1UL;
+	sparc_ebus_dmas[slot].info.regs =
+		of_ioremap(&op->resource[2], 0, d_len, "ECPP DMA");
+
+	if (!sparc_ebus_dmas[slot].info.regs)
+		goto out_clear_map;
+
+	sparc_ebus_dmas[slot].info.flags = 0;
+	sparc_ebus_dmas[slot].info.callback = NULL;
+	sparc_ebus_dmas[slot].info.client_cookie = NULL;
+	sparc_ebus_dmas[slot].info.irq = 0xdeadbeef;
+	strcpy(sparc_ebus_dmas[slot].info.name, "parport");
+	if (ebus_dma_register(&sparc_ebus_dmas[slot].info))
+		goto out_unmap_regs;
+
+	ebus_dma_irq_enable(&sparc_ebus_dmas[slot].info, 1);
+
+	/* Configure IRQ to Push Pull, Level Low */
+	/* Enable ECP, set bit 2 of the CTR first */
+	outb(0x04, base + 0x02);
+	ns87303_modify(config, PCR,
+		       PCR_EPP_ENABLE |
+		       PCR_IRQ_ODRAIN,
+		       PCR_ECP_ENABLE |
+		       PCR_ECP_CLK_ENA |
+		       PCR_IRQ_POLAR);
+
+	/* CTR bit 5 controls direction of port */
+	ns87303_modify(config, PTR,
+		       0, PTR_LPT_REG_DIR);
+
+	p = parport_pc_probe_port(base, base + 0x400,
+				  op->irqs[0],
+				  slot,
+				  op->dev.parent);
+	err = -ENOMEM;
+	if (!p)
+		goto out_disable_irq;
+
+	dev_set_drvdata(&op->dev, p);
+
+	return 0;
+
+out_disable_irq:
+	ebus_dma_irq_enable(&sparc_ebus_dmas[slot].info, 0);
+	ebus_dma_unregister(&sparc_ebus_dmas[slot].info);
+
+out_unmap_regs:
+	of_iounmap(&op->resource[2], sparc_ebus_dmas[slot].info.regs, d_len);
+
+out_clear_map:
+	clear_bit(slot, dma_slot_map);
+
+out_err:
+	return err;
+}
+
+static int __devexit ecpp_remove(struct of_device *op)
+{
+	struct parport *p = dev_get_drvdata(&op->dev);
+	int slot = p->dma;
+
+	parport_pc_unregister_port(p);
+
+	if (slot != PARPORT_DMA_NOFIFO) {
+		unsigned long d_base = op->resource[2].start;
+		unsigned long d_len;
+
+		d_len = (op->resource[2].end - d_base) + 1UL;
+
+		ebus_dma_irq_enable(&sparc_ebus_dmas[slot].info, 0);
+		ebus_dma_unregister(&sparc_ebus_dmas[slot].info);
+		of_iounmap(&op->resource[2],
+			   sparc_ebus_dmas[slot].info.regs,
+			   d_len);
+		clear_bit(slot, dma_slot_map);
+	}
+
+	return 0;
+}
+
+static struct of_device_id ecpp_match[] = {
+	{
+		.name = "ecpp",
+	},
+	{
+		.name = "parallel",
+		.compatible = "ecpp",
+	},
+	{
+		.name = "parallel",
+		.compatible = "ns87317-ecpp",
+	},
+	{},
+};
+
+static struct of_platform_driver ecpp_driver = {
+	.name			= "ecpp",
+	.match_table		= ecpp_match,
+	.probe			= ecpp_probe,
+	.remove			= __devexit_p(ecpp_remove),
+};
+
+static int parport_pc_find_nonpci_ports(int autoirq, int autodma)
+{
+	of_register_driver(&ecpp_driver, &of_bus_type);
+
+	return 0;
+}
+
+#endif /* !(_ASM_SPARC64_PARPORT_H */
diff --git a/include/asm-sparc/pil.h b/include/asm-sparc/pil.h
new file mode 100644
index 0000000..eaac842
--- /dev/null
+++ b/include/asm-sparc/pil.h
@@ -0,0 +1,21 @@
+#ifndef _SPARC64_PIL_H
+#define _SPARC64_PIL_H
+
+/* To avoid some locking problems, we hard allocate certain PILs
+ * for SMP cross call messages that must do a etrap/rtrap.
+ *
+ * A local_irq_disable() does not block the cross call delivery, so
+ * when SMP locking is an issue we reschedule the event into a PIL
+ * interrupt which is blocked by local_irq_disable().
+ *
+ * In fact any XCALL which has to etrap/rtrap has a problem because
+ * it is difficult to prevent rtrap from running BH's, and that would
+ * need to be done if the XCALL arrived while %pil==15.
+ */
+#define PIL_SMP_CALL_FUNC	1
+#define PIL_SMP_RECEIVE_SIGNAL	2
+#define PIL_SMP_CAPTURE		3
+#define PIL_SMP_CTX_NEW_VERSION	4
+#define PIL_DEVICE_IRQ		5
+
+#endif /* !(_SPARC64_PIL_H) */
diff --git a/include/asm-sparc/reboot.h b/include/asm-sparc/reboot.h
new file mode 100644
index 0000000..3f3f43f
--- /dev/null
+++ b/include/asm-sparc/reboot.h
@@ -0,0 +1,6 @@
+#ifndef _SPARC64_REBOOT_H
+#define _SPARC64_REBOOT_H
+
+extern void machine_alt_power_off(void);
+
+#endif /* _SPARC64_REBOOT_H */
diff --git a/include/asm-sparc/rwsem-const.h b/include/asm-sparc/rwsem-const.h
new file mode 100644
index 0000000..a303c9d
--- /dev/null
+++ b/include/asm-sparc/rwsem-const.h
@@ -0,0 +1,12 @@
+/* rwsem-const.h: RW semaphore counter constants.  */
+#ifndef _SPARC64_RWSEM_CONST_H
+#define _SPARC64_RWSEM_CONST_H
+
+#define RWSEM_UNLOCKED_VALUE		0x00000000
+#define RWSEM_ACTIVE_BIAS		0x00000001
+#define RWSEM_ACTIVE_MASK		0x0000ffff
+#define RWSEM_WAITING_BIAS		0xffff0000
+#define RWSEM_ACTIVE_READ_BIAS		RWSEM_ACTIVE_BIAS
+#define RWSEM_ACTIVE_WRITE_BIAS		(RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
+
+#endif /* _SPARC64_RWSEM_CONST_H */
diff --git a/include/asm-sparc/rwsem.h b/include/asm-sparc/rwsem.h
new file mode 100644
index 0000000..1dc129a
--- /dev/null
+++ b/include/asm-sparc/rwsem.h
@@ -0,0 +1,84 @@
+/*
+ * rwsem.h: R/W semaphores implemented using CAS
+ *
+ * Written by David S. Miller (davem@redhat.com), 2001.
+ * Derived from asm-i386/rwsem.h
+ */
+#ifndef _SPARC64_RWSEM_H
+#define _SPARC64_RWSEM_H
+
+#ifndef _LINUX_RWSEM_H
+#error "please don't include asm/rwsem.h directly, use linux/rwsem.h instead"
+#endif
+
+#ifdef __KERNEL__
+
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <asm/rwsem-const.h>
+
+struct rwsem_waiter;
+
+struct rw_semaphore {
+	signed int count;
+	spinlock_t		wait_lock;
+	struct list_head	wait_list;
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	struct lockdep_map	dep_map;
+#endif
+};
+
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+# define __RWSEM_DEP_MAP_INIT(lockname) , .dep_map = { .name = #lockname }
+#else
+# define __RWSEM_DEP_MAP_INIT(lockname)
+#endif
+
+#define __RWSEM_INITIALIZER(name) \
+{ RWSEM_UNLOCKED_VALUE, SPIN_LOCK_UNLOCKED, LIST_HEAD_INIT((name).wait_list) \
+  __RWSEM_DEP_MAP_INIT(name) }
+
+#define DECLARE_RWSEM(name) \
+	struct rw_semaphore name = __RWSEM_INITIALIZER(name)
+
+extern void __init_rwsem(struct rw_semaphore *sem, const char *name,
+			 struct lock_class_key *key);
+
+#define init_rwsem(sem)						\
+do {								\
+	static struct lock_class_key __key;			\
+								\
+	__init_rwsem((sem), #sem, &__key);			\
+} while (0)
+
+extern void __down_read(struct rw_semaphore *sem);
+extern int __down_read_trylock(struct rw_semaphore *sem);
+extern void __down_write(struct rw_semaphore *sem);
+extern int __down_write_trylock(struct rw_semaphore *sem);
+extern void __up_read(struct rw_semaphore *sem);
+extern void __up_write(struct rw_semaphore *sem);
+extern void __downgrade_write(struct rw_semaphore *sem);
+
+static inline void __down_write_nested(struct rw_semaphore *sem, int subclass)
+{
+	__down_write(sem);
+}
+
+static inline int rwsem_atomic_update(int delta, struct rw_semaphore *sem)
+{
+	return atomic_add_return(delta, (atomic_t *)(&sem->count));
+}
+
+static inline void rwsem_atomic_add(int delta, struct rw_semaphore *sem)
+{
+	atomic_add(delta, (atomic_t *)(&sem->count));
+}
+
+static inline int rwsem_is_locked(struct rw_semaphore *sem)
+{
+	return (sem->count != 0);
+}
+
+#endif /* __KERNEL__ */
+
+#endif /* _SPARC64_RWSEM_H */
diff --git a/include/asm-sparc/scratchpad.h b/include/asm-sparc/scratchpad.h
new file mode 100644
index 0000000..5e8b01f
--- /dev/null
+++ b/include/asm-sparc/scratchpad.h
@@ -0,0 +1,14 @@
+#ifndef _SPARC64_SCRATCHPAD_H
+#define _SPARC64_SCRATCHPAD_H
+
+/* Sun4v scratchpad registers, accessed via ASI_SCRATCHPAD.  */
+
+#define SCRATCHPAD_MMU_MISS	0x00 /* Shared with OBP - set by OBP	    */
+#define SCRATCHPAD_CPUID	0x08 /* Shared with OBP - set by hypervisor */
+#define SCRATCHPAD_UTSBREG1	0x10
+#define SCRATCHPAD_UTSBREG2	0x18
+	/* 0x20 and 0x28, hypervisor only... */
+#define SCRATCHPAD_UNUSED1	0x30
+#define SCRATCHPAD_UNUSED2	0x38 /* Reserved for OBP		    */
+
+#endif /* !(_SPARC64_SCRATCHPAD_H) */
diff --git a/include/asm-sparc/seccomp.h b/include/asm-sparc/seccomp.h
new file mode 100644
index 0000000..7fcd996
--- /dev/null
+++ b/include/asm-sparc/seccomp.h
@@ -0,0 +1,21 @@
+#ifndef _ASM_SECCOMP_H
+
+#include <linux/thread_info.h> /* already defines TIF_32BIT */
+
+#ifndef TIF_32BIT
+#error "unexpected TIF_32BIT on sparc64"
+#endif
+
+#include <linux/unistd.h>
+
+#define __NR_seccomp_read __NR_read
+#define __NR_seccomp_write __NR_write
+#define __NR_seccomp_exit __NR_exit
+#define __NR_seccomp_sigreturn __NR_rt_sigreturn
+
+#define __NR_seccomp_read_32 __NR_read
+#define __NR_seccomp_write_32 __NR_write
+#define __NR_seccomp_exit_32 __NR_exit
+#define __NR_seccomp_sigreturn_32 __NR_sigreturn
+
+#endif /* _ASM_SECCOMP_H */
diff --git a/include/asm-sparc/sfafsr.h b/include/asm-sparc/sfafsr.h
new file mode 100644
index 0000000..e96137b
--- /dev/null
+++ b/include/asm-sparc/sfafsr.h
@@ -0,0 +1,82 @@
+#ifndef _SPARC64_SFAFSR_H
+#define _SPARC64_SFAFSR_H
+
+#include <linux/const.h>
+
+/* Spitfire Asynchronous Fault Status register, ASI=0x4C VA<63:0>=0x0 */
+
+#define SFAFSR_ME		(_AC(1,UL) << SFAFSR_ME_SHIFT)
+#define SFAFSR_ME_SHIFT		32
+#define SFAFSR_PRIV		(_AC(1,UL) << SFAFSR_PRIV_SHIFT)
+#define SFAFSR_PRIV_SHIFT	31
+#define SFAFSR_ISAP		(_AC(1,UL) << SFAFSR_ISAP_SHIFT)
+#define SFAFSR_ISAP_SHIFT	30
+#define SFAFSR_ETP		(_AC(1,UL) << SFAFSR_ETP_SHIFT)
+#define SFAFSR_ETP_SHIFT	29
+#define SFAFSR_IVUE		(_AC(1,UL) << SFAFSR_IVUE_SHIFT)
+#define SFAFSR_IVUE_SHIFT	28
+#define SFAFSR_TO		(_AC(1,UL) << SFAFSR_TO_SHIFT)
+#define SFAFSR_TO_SHIFT		27
+#define SFAFSR_BERR		(_AC(1,UL) << SFAFSR_BERR_SHIFT)
+#define SFAFSR_BERR_SHIFT	26
+#define SFAFSR_LDP		(_AC(1,UL) << SFAFSR_LDP_SHIFT)
+#define SFAFSR_LDP_SHIFT	25
+#define SFAFSR_CP		(_AC(1,UL) << SFAFSR_CP_SHIFT)
+#define SFAFSR_CP_SHIFT		24
+#define SFAFSR_WP		(_AC(1,UL) << SFAFSR_WP_SHIFT)
+#define SFAFSR_WP_SHIFT		23
+#define SFAFSR_EDP		(_AC(1,UL) << SFAFSR_EDP_SHIFT)
+#define SFAFSR_EDP_SHIFT	22
+#define SFAFSR_UE		(_AC(1,UL) << SFAFSR_UE_SHIFT)
+#define SFAFSR_UE_SHIFT		21
+#define SFAFSR_CE		(_AC(1,UL) << SFAFSR_CE_SHIFT)
+#define SFAFSR_CE_SHIFT		20
+#define SFAFSR_ETS		(_AC(0xf,UL) << SFAFSR_ETS_SHIFT)
+#define SFAFSR_ETS_SHIFT	16
+#define SFAFSR_PSYND		(_AC(0xffff,UL) << SFAFSR_PSYND_SHIFT)
+#define SFAFSR_PSYND_SHIFT	0
+
+/* UDB Error Register, ASI=0x7f VA<63:0>=0x0(High),0x18(Low) for read
+ *                     ASI=0x77 VA<63:0>=0x0(High),0x18(Low) for write
+ */
+
+#define UDBE_UE			(_AC(1,UL) << 9)
+#define UDBE_CE			(_AC(1,UL) << 8)
+#define UDBE_E_SYNDR		(_AC(0xff,UL) << 0)
+
+/* The trap handlers for asynchronous errors encode the AFSR and
+ * other pieces of information into a 64-bit argument for C code
+ * encoded as follows:
+ *
+ * -----------------------------------------------
+ * |  UDB_H  |  UDB_L  | TL>1  |  TT  |   AFSR   |
+ * -----------------------------------------------
+ *  63     54 53     44    42   41  33 32       0
+ *
+ * The AFAR is passed in unchanged.
+ */
+#define SFSTAT_UDBH_MASK	(_AC(0x3ff,UL) << SFSTAT_UDBH_SHIFT)
+#define SFSTAT_UDBH_SHIFT	54
+#define SFSTAT_UDBL_MASK	(_AC(0x3ff,UL) << SFSTAT_UDBH_SHIFT)
+#define SFSTAT_UDBL_SHIFT	44
+#define SFSTAT_TL_GT_ONE	(_AC(1,UL) << SFSTAT_TL_GT_ONE_SHIFT)
+#define SFSTAT_TL_GT_ONE_SHIFT	42
+#define SFSTAT_TRAP_TYPE	(_AC(0x1FF,UL) << SFSTAT_TRAP_TYPE_SHIFT)
+#define SFSTAT_TRAP_TYPE_SHIFT	33
+#define SFSTAT_AFSR_MASK	(_AC(0x1ffffffff,UL) << SFSTAT_AFSR_SHIFT)
+#define SFSTAT_AFSR_SHIFT	0
+
+/* ESTATE Error Enable Register, ASI=0x4b VA<63:0>=0x0 */
+#define ESTATE_ERR_CE		0x1 /* Correctable errors                    */
+#define ESTATE_ERR_NCE		0x2 /* TO, BERR, LDP, ETP, EDP, WP, UE, IVUE */
+#define ESTATE_ERR_ISAP		0x4 /* System address parity error           */
+#define ESTATE_ERR_ALL		(ESTATE_ERR_CE | \
+				 ESTATE_ERR_NCE | \
+				 ESTATE_ERR_ISAP)
+
+/* The various trap types that report using the above state. */
+#define TRAP_TYPE_IAE		0x09 /* Instruction Access Error             */
+#define TRAP_TYPE_DAE		0x32 /* Data Access Error                    */
+#define TRAP_TYPE_CEE		0x63 /* Correctable ECC Error                */
+
+#endif /* _SPARC64_SFAFSR_H */
diff --git a/include/asm-sparc/sparsemem.h b/include/asm-sparc/sparsemem.h
new file mode 100644
index 0000000..b99d4e4
--- /dev/null
+++ b/include/asm-sparc/sparsemem.h
@@ -0,0 +1,12 @@
+#ifndef _SPARC64_SPARSEMEM_H
+#define _SPARC64_SPARSEMEM_H
+
+#ifdef __KERNEL__
+
+#define SECTION_SIZE_BITS       30
+#define MAX_PHYSADDR_BITS       42
+#define MAX_PHYSMEM_BITS        42
+
+#endif /* !(__KERNEL__) */
+
+#endif /* !(_SPARC64_SPARSEMEM_H) */
diff --git a/include/asm-sparc/spitfire.h b/include/asm-sparc/spitfire.h
new file mode 100644
index 0000000..985ea7e
--- /dev/null
+++ b/include/asm-sparc/spitfire.h
@@ -0,0 +1,342 @@
+/* spitfire.h: SpitFire/BlackBird/Cheetah inline MMU operations.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@davemloft.net)
+ */
+
+#ifndef _SPARC64_SPITFIRE_H
+#define _SPARC64_SPITFIRE_H
+
+#include <asm/asi.h>
+
+/* The following register addresses are accessible via ASI_DMMU
+ * and ASI_IMMU, that is there is a distinct and unique copy of
+ * each these registers for each TLB.
+ */
+#define TSB_TAG_TARGET		0x0000000000000000 /* All chips				*/
+#define TLB_SFSR		0x0000000000000018 /* All chips				*/
+#define TSB_REG			0x0000000000000028 /* All chips				*/
+#define TLB_TAG_ACCESS		0x0000000000000030 /* All chips				*/
+#define VIRT_WATCHPOINT		0x0000000000000038 /* All chips				*/
+#define PHYS_WATCHPOINT		0x0000000000000040 /* All chips				*/
+#define TSB_EXTENSION_P		0x0000000000000048 /* Ultra-III and later		*/
+#define TSB_EXTENSION_S		0x0000000000000050 /* Ultra-III and later, D-TLB only	*/
+#define TSB_EXTENSION_N		0x0000000000000058 /* Ultra-III and later		*/
+#define TLB_TAG_ACCESS_EXT	0x0000000000000060 /* Ultra-III+ and later		*/
+
+/* These registers only exist as one entity, and are accessed
+ * via ASI_DMMU only.
+ */
+#define PRIMARY_CONTEXT		0x0000000000000008
+#define SECONDARY_CONTEXT	0x0000000000000010
+#define DMMU_SFAR		0x0000000000000020
+#define VIRT_WATCHPOINT		0x0000000000000038
+#define PHYS_WATCHPOINT		0x0000000000000040
+
+#define SPITFIRE_HIGHEST_LOCKED_TLBENT	(64 - 1)
+#define CHEETAH_HIGHEST_LOCKED_TLBENT	(16 - 1)
+
+#define L1DCACHE_SIZE		0x4000
+
+#define SUN4V_CHIP_INVALID	0x00
+#define SUN4V_CHIP_NIAGARA1	0x01
+#define SUN4V_CHIP_NIAGARA2	0x02
+#define SUN4V_CHIP_UNKNOWN	0xff
+
+#ifndef __ASSEMBLY__
+
+enum ultra_tlb_layout {
+	spitfire = 0,
+	cheetah = 1,
+	cheetah_plus = 2,
+	hypervisor = 3,
+};
+
+extern enum ultra_tlb_layout tlb_type;
+
+extern int sun4v_chip_type;
+
+extern int cheetah_pcache_forced_on;
+extern void cheetah_enable_pcache(void);
+
+#define sparc64_highest_locked_tlbent()	\
+	(tlb_type == spitfire ? \
+	 SPITFIRE_HIGHEST_LOCKED_TLBENT : \
+	 CHEETAH_HIGHEST_LOCKED_TLBENT)
+
+extern int num_kernel_image_mappings;
+
+/* The data cache is write through, so this just invalidates the
+ * specified line.
+ */
+static inline void spitfire_put_dcache_tag(unsigned long addr, unsigned long tag)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (tag), "r" (addr), "i" (ASI_DCACHE_TAG));
+}
+
+/* The instruction cache lines are flushed with this, but note that
+ * this does not flush the pipeline.  It is possible for a line to
+ * get flushed but stale instructions to still be in the pipeline,
+ * a flush instruction (to any address) is sufficient to handle
+ * this issue after the line is invalidated.
+ */
+static inline void spitfire_put_icache_tag(unsigned long addr, unsigned long tag)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (tag), "r" (addr), "i" (ASI_IC_TAG));
+}
+
+static inline unsigned long spitfire_get_dtlb_data(int entry)
+{
+	unsigned long data;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (data)
+			     : "r" (entry << 3), "i" (ASI_DTLB_DATA_ACCESS));
+
+	/* Clear TTE diag bits. */
+	data &= ~0x0003fe0000000000UL;
+
+	return data;
+}
+
+static inline unsigned long spitfire_get_dtlb_tag(int entry)
+{
+	unsigned long tag;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (tag)
+			     : "r" (entry << 3), "i" (ASI_DTLB_TAG_READ));
+	return tag;
+}
+
+static inline void spitfire_put_dtlb_data(int entry, unsigned long data)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (data), "r" (entry << 3),
+			       "i" (ASI_DTLB_DATA_ACCESS));
+}
+
+static inline unsigned long spitfire_get_itlb_data(int entry)
+{
+	unsigned long data;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (data)
+			     : "r" (entry << 3), "i" (ASI_ITLB_DATA_ACCESS));
+
+	/* Clear TTE diag bits. */
+	data &= ~0x0003fe0000000000UL;
+
+	return data;
+}
+
+static inline unsigned long spitfire_get_itlb_tag(int entry)
+{
+	unsigned long tag;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (tag)
+			     : "r" (entry << 3), "i" (ASI_ITLB_TAG_READ));
+	return tag;
+}
+
+static inline void spitfire_put_itlb_data(int entry, unsigned long data)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (data), "r" (entry << 3),
+			       "i" (ASI_ITLB_DATA_ACCESS));
+}
+
+static inline void spitfire_flush_dtlb_nucleus_page(unsigned long page)
+{
+	__asm__ __volatile__("stxa	%%g0, [%0] %1\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (page | 0x20), "i" (ASI_DMMU_DEMAP));
+}
+
+static inline void spitfire_flush_itlb_nucleus_page(unsigned long page)
+{
+	__asm__ __volatile__("stxa	%%g0, [%0] %1\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (page | 0x20), "i" (ASI_IMMU_DEMAP));
+}
+
+/* Cheetah has "all non-locked" tlb flushes. */
+static inline void cheetah_flush_dtlb_all(void)
+{
+	__asm__ __volatile__("stxa	%%g0, [%0] %1\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (0x80), "i" (ASI_DMMU_DEMAP));
+}
+
+static inline void cheetah_flush_itlb_all(void)
+{
+	__asm__ __volatile__("stxa	%%g0, [%0] %1\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (0x80), "i" (ASI_IMMU_DEMAP));
+}
+
+/* Cheetah has a 4-tlb layout so direct access is a bit different.
+ * The first two TLBs are fully assosciative, hold 16 entries, and are
+ * used only for locked and >8K sized translations.  One exists for
+ * data accesses and one for instruction accesses.
+ *
+ * The third TLB is for data accesses to 8K non-locked translations, is
+ * 2 way assosciative, and holds 512 entries.  The fourth TLB is for
+ * instruction accesses to 8K non-locked translations, is 2 way
+ * assosciative, and holds 128 entries.
+ *
+ * Cheetah has some bug where bogus data can be returned from
+ * ASI_{D,I}TLB_DATA_ACCESS loads, doing the load twice fixes
+ * the problem for me. -DaveM
+ */
+static inline unsigned long cheetah_get_ldtlb_data(int entry)
+{
+	unsigned long data;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %%g0\n\t"
+			     "ldxa	[%1] %2, %0"
+			     : "=r" (data)
+			     : "r" ((0 << 16) | (entry << 3)),
+			     "i" (ASI_DTLB_DATA_ACCESS));
+
+	return data;
+}
+
+static inline unsigned long cheetah_get_litlb_data(int entry)
+{
+	unsigned long data;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %%g0\n\t"
+			     "ldxa	[%1] %2, %0"
+			     : "=r" (data)
+			     : "r" ((0 << 16) | (entry << 3)),
+			     "i" (ASI_ITLB_DATA_ACCESS));
+
+	return data;
+}
+
+static inline unsigned long cheetah_get_ldtlb_tag(int entry)
+{
+	unsigned long tag;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (tag)
+			     : "r" ((0 << 16) | (entry << 3)),
+			     "i" (ASI_DTLB_TAG_READ));
+
+	return tag;
+}
+
+static inline unsigned long cheetah_get_litlb_tag(int entry)
+{
+	unsigned long tag;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (tag)
+			     : "r" ((0 << 16) | (entry << 3)),
+			     "i" (ASI_ITLB_TAG_READ));
+
+	return tag;
+}
+
+static inline void cheetah_put_ldtlb_data(int entry, unsigned long data)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (data),
+			       "r" ((0 << 16) | (entry << 3)),
+			       "i" (ASI_DTLB_DATA_ACCESS));
+}
+
+static inline void cheetah_put_litlb_data(int entry, unsigned long data)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (data),
+			       "r" ((0 << 16) | (entry << 3)),
+			       "i" (ASI_ITLB_DATA_ACCESS));
+}
+
+static inline unsigned long cheetah_get_dtlb_data(int entry, int tlb)
+{
+	unsigned long data;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %%g0\n\t"
+			     "ldxa	[%1] %2, %0"
+			     : "=r" (data)
+			     : "r" ((tlb << 16) | (entry << 3)), "i" (ASI_DTLB_DATA_ACCESS));
+
+	return data;
+}
+
+static inline unsigned long cheetah_get_dtlb_tag(int entry, int tlb)
+{
+	unsigned long tag;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (tag)
+			     : "r" ((tlb << 16) | (entry << 3)), "i" (ASI_DTLB_TAG_READ));
+	return tag;
+}
+
+static inline void cheetah_put_dtlb_data(int entry, unsigned long data, int tlb)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (data),
+			       "r" ((tlb << 16) | (entry << 3)),
+			       "i" (ASI_DTLB_DATA_ACCESS));
+}
+
+static inline unsigned long cheetah_get_itlb_data(int entry)
+{
+	unsigned long data;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %%g0\n\t"
+			     "ldxa	[%1] %2, %0"
+			     : "=r" (data)
+			     : "r" ((2 << 16) | (entry << 3)),
+                               "i" (ASI_ITLB_DATA_ACCESS));
+
+	return data;
+}
+
+static inline unsigned long cheetah_get_itlb_tag(int entry)
+{
+	unsigned long tag;
+
+	__asm__ __volatile__("ldxa	[%1] %2, %0"
+			     : "=r" (tag)
+			     : "r" ((2 << 16) | (entry << 3)), "i" (ASI_ITLB_TAG_READ));
+	return tag;
+}
+
+static inline void cheetah_put_itlb_data(int entry, unsigned long data)
+{
+	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
+			     "membar	#Sync"
+			     : /* No outputs */
+			     : "r" (data), "r" ((2 << 16) | (entry << 3)),
+			       "i" (ASI_ITLB_DATA_ACCESS));
+}
+
+#endif /* !(__ASSEMBLY__) */
+
+#endif /* !(_SPARC64_SPITFIRE_H) */
diff --git a/include/asm-sparc/sstate.h b/include/asm-sparc/sstate.h
new file mode 100644
index 0000000..a7c35db
--- /dev/null
+++ b/include/asm-sparc/sstate.h
@@ -0,0 +1,13 @@
+#ifndef _SPARC64_SSTATE_H
+#define _SPARC64_SSTATE_H
+
+extern void sstate_booting(void);
+extern void sstate_running(void);
+extern void sstate_halt(void);
+extern void sstate_poweroff(void);
+extern void sstate_panic(void);
+extern void sstate_reboot(void);
+
+extern void sun4v_sstate_init(void);
+
+#endif /* _SPARC64_SSTATE_H */
diff --git a/include/asm-sparc/stacktrace.h b/include/asm-sparc/stacktrace.h
new file mode 100644
index 0000000..6cee39a
--- /dev/null
+++ b/include/asm-sparc/stacktrace.h
@@ -0,0 +1,6 @@
+#ifndef _SPARC64_STACKTRACE_H
+#define _SPARC64_STACKTRACE_H
+
+extern void stack_trace_flush(void);
+
+#endif /* _SPARC64_STACKTRACE_H */
diff --git a/include/asm-sparc/starfire.h b/include/asm-sparc/starfire.h
new file mode 100644
index 0000000..07bafd3
--- /dev/null
+++ b/include/asm-sparc/starfire.h
@@ -0,0 +1,21 @@
+/*
+ * starfire.h: Group all starfire specific code together.
+ *
+ * Copyright (C) 2000 Anton Blanchard (anton@samba.org)
+ */
+
+#ifndef _SPARC64_STARFIRE_H
+#define _SPARC64_STARFIRE_H
+
+#ifndef __ASSEMBLY__
+
+extern int this_is_starfire;
+
+extern void check_if_starfire(void);
+extern void starfire_cpu_setup(void);
+extern int starfire_hard_smp_processor_id(void);
+extern void starfire_hookup(int);
+extern unsigned int starfire_translate(unsigned long imap, unsigned int upaid);
+
+#endif
+#endif
diff --git a/include/asm-sparc/syscalls.h b/include/asm-sparc/syscalls.h
new file mode 100644
index 0000000..45a43f6
--- /dev/null
+++ b/include/asm-sparc/syscalls.h
@@ -0,0 +1,13 @@
+#ifndef _SPARC64_SYSCALLS_H
+#define _SPARC64_SYSCALLS_H
+
+struct pt_regs;
+
+extern asmlinkage long sparc_do_fork(unsigned long clone_flags,
+				     unsigned long stack_start,
+				     struct pt_regs *regs,
+				     unsigned long stack_size);
+
+extern asmlinkage int sparc_execve(struct pt_regs *regs);
+
+#endif /* _SPARC64_SYSCALLS_H */
diff --git a/include/asm-sparc/tsb.h b/include/asm-sparc/tsb.h
new file mode 100644
index 0000000..76e4299
--- /dev/null
+++ b/include/asm-sparc/tsb.h
@@ -0,0 +1,283 @@
+#ifndef _SPARC64_TSB_H
+#define _SPARC64_TSB_H
+
+/* The sparc64 TSB is similar to the powerpc hashtables.  It's a
+ * power-of-2 sized table of TAG/PTE pairs.  The cpu precomputes
+ * pointers into this table for 8K and 64K page sizes, and also a
+ * comparison TAG based upon the virtual address and context which
+ * faults.
+ *
+ * TLB miss trap handler software does the actual lookup via something
+ * of the form:
+ *
+ * 	ldxa		[%g0] ASI_{D,I}MMU_TSB_8KB_PTR, %g1
+ * 	ldxa		[%g0] ASI_{D,I}MMU, %g6
+ *	sllx		%g6, 22, %g6
+ *	srlx		%g6, 22, %g6
+ * 	ldda		[%g1] ASI_NUCLEUS_QUAD_LDD, %g4
+ * 	cmp		%g4, %g6
+ * 	bne,pn	%xcc, tsb_miss_{d,i}tlb
+ * 	 mov		FAULT_CODE_{D,I}TLB, %g3
+ * 	stxa		%g5, [%g0] ASI_{D,I}TLB_DATA_IN
+ * 	retry
+ *
+ *
+ * Each 16-byte slot of the TSB is the 8-byte tag and then the 8-byte
+ * PTE.  The TAG is of the same layout as the TLB TAG TARGET mmu
+ * register which is:
+ *
+ * -------------------------------------------------
+ * |  -  |  CONTEXT |  -  |    VADDR bits 63:22    |
+ * -------------------------------------------------
+ *  63 61 60      48 47 42 41                     0
+ *
+ * But actually, since we use per-mm TSB's, we zero out the CONTEXT
+ * field.
+ *
+ * Like the powerpc hashtables we need to use locking in order to
+ * synchronize while we update the entries.  PTE updates need locking
+ * as well.
+ *
+ * We need to carefully choose a lock bits for the TSB entry.  We
+ * choose to use bit 47 in the tag.  Also, since we never map anything
+ * at page zero in context zero, we use zero as an invalid tag entry.
+ * When the lock bit is set, this forces a tag comparison failure.
+ */
+
+#define TSB_TAG_LOCK_BIT	47
+#define TSB_TAG_LOCK_HIGH	(1 << (TSB_TAG_LOCK_BIT - 32))
+
+#define TSB_TAG_INVALID_BIT	46
+#define TSB_TAG_INVALID_HIGH	(1 << (TSB_TAG_INVALID_BIT - 32))
+
+#define TSB_MEMBAR	membar	#StoreStore
+
+/* Some cpus support physical address quad loads.  We want to use
+ * those if possible so we don't need to hard-lock the TSB mapping
+ * into the TLB.  We encode some instruction patching in order to
+ * support this.
+ *
+ * The kernel TSB is locked into the TLB by virtue of being in the
+ * kernel image, so we don't play these games for swapper_tsb access.
+ */
+#ifndef __ASSEMBLY__
+struct tsb_ldquad_phys_patch_entry {
+	unsigned int	addr;
+	unsigned int	sun4u_insn;
+	unsigned int	sun4v_insn;
+};
+extern struct tsb_ldquad_phys_patch_entry __tsb_ldquad_phys_patch,
+	__tsb_ldquad_phys_patch_end;
+
+struct tsb_phys_patch_entry {
+	unsigned int	addr;
+	unsigned int	insn;
+};
+extern struct tsb_phys_patch_entry __tsb_phys_patch, __tsb_phys_patch_end;
+#endif
+#define TSB_LOAD_QUAD(TSB, REG)	\
+661:	ldda		[TSB] ASI_NUCLEUS_QUAD_LDD, REG; \
+	.section	.tsb_ldquad_phys_patch, "ax"; \
+	.word		661b; \
+	ldda		[TSB] ASI_QUAD_LDD_PHYS, REG; \
+	ldda		[TSB] ASI_QUAD_LDD_PHYS_4V, REG; \
+	.previous
+
+#define TSB_LOAD_TAG_HIGH(TSB, REG) \
+661:	lduwa		[TSB] ASI_N, REG; \
+	.section	.tsb_phys_patch, "ax"; \
+	.word		661b; \
+	lduwa		[TSB] ASI_PHYS_USE_EC, REG; \
+	.previous
+
+#define TSB_LOAD_TAG(TSB, REG) \
+661:	ldxa		[TSB] ASI_N, REG; \
+	.section	.tsb_phys_patch, "ax"; \
+	.word		661b; \
+	ldxa		[TSB] ASI_PHYS_USE_EC, REG; \
+	.previous
+
+#define TSB_CAS_TAG_HIGH(TSB, REG1, REG2) \
+661:	casa		[TSB] ASI_N, REG1, REG2; \
+	.section	.tsb_phys_patch, "ax"; \
+	.word		661b; \
+	casa		[TSB] ASI_PHYS_USE_EC, REG1, REG2; \
+	.previous
+
+#define TSB_CAS_TAG(TSB, REG1, REG2) \
+661:	casxa		[TSB] ASI_N, REG1, REG2; \
+	.section	.tsb_phys_patch, "ax"; \
+	.word		661b; \
+	casxa		[TSB] ASI_PHYS_USE_EC, REG1, REG2; \
+	.previous
+
+#define TSB_STORE(ADDR, VAL) \
+661:	stxa		VAL, [ADDR] ASI_N; \
+	.section	.tsb_phys_patch, "ax"; \
+	.word		661b; \
+	stxa		VAL, [ADDR] ASI_PHYS_USE_EC; \
+	.previous
+
+#define TSB_LOCK_TAG(TSB, REG1, REG2)	\
+99:	TSB_LOAD_TAG_HIGH(TSB, REG1);	\
+	sethi	%hi(TSB_TAG_LOCK_HIGH), REG2;\
+	andcc	REG1, REG2, %g0;	\
+	bne,pn	%icc, 99b;		\
+	 nop;				\
+	TSB_CAS_TAG_HIGH(TSB, REG1, REG2);	\
+	cmp	REG1, REG2;		\
+	bne,pn	%icc, 99b;		\
+	 nop;				\
+	TSB_MEMBAR
+
+#define TSB_WRITE(TSB, TTE, TAG) \
+	add	TSB, 0x8, TSB;   \
+	TSB_STORE(TSB, TTE);     \
+	sub	TSB, 0x8, TSB;   \
+	TSB_MEMBAR;              \
+	TSB_STORE(TSB, TAG);
+
+#define KTSB_LOAD_QUAD(TSB, REG) \
+	ldda		[TSB] ASI_NUCLEUS_QUAD_LDD, REG;
+
+#define KTSB_STORE(ADDR, VAL) \
+	stxa		VAL, [ADDR] ASI_N;
+
+#define KTSB_LOCK_TAG(TSB, REG1, REG2)	\
+99:	lduwa	[TSB] ASI_N, REG1;	\
+	sethi	%hi(TSB_TAG_LOCK_HIGH), REG2;\
+	andcc	REG1, REG2, %g0;	\
+	bne,pn	%icc, 99b;		\
+	 nop;				\
+	casa	[TSB] ASI_N, REG1, REG2;\
+	cmp	REG1, REG2;		\
+	bne,pn	%icc, 99b;		\
+	 nop;				\
+	TSB_MEMBAR
+
+#define KTSB_WRITE(TSB, TTE, TAG) \
+	add	TSB, 0x8, TSB;   \
+	stxa	TTE, [TSB] ASI_N;     \
+	sub	TSB, 0x8, TSB;   \
+	TSB_MEMBAR;              \
+	stxa	TAG, [TSB] ASI_N;
+
+	/* Do a kernel page table walk.  Leaves physical PTE pointer in
+	 * REG1.  Jumps to FAIL_LABEL on early page table walk termination.
+	 * VADDR will not be clobbered, but REG2 will.
+	 */
+#define KERN_PGTABLE_WALK(VADDR, REG1, REG2, FAIL_LABEL)	\
+	sethi		%hi(swapper_pg_dir), REG1; \
+	or		REG1, %lo(swapper_pg_dir), REG1; \
+	sllx		VADDR, 64 - (PGDIR_SHIFT + PGDIR_BITS), REG2; \
+	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
+	andn		REG2, 0x3, REG2; \
+	lduw		[REG1 + REG2], REG1; \
+	brz,pn		REG1, FAIL_LABEL; \
+	 sllx		VADDR, 64 - (PMD_SHIFT + PMD_BITS), REG2; \
+	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
+	sllx		REG1, 11, REG1; \
+	andn		REG2, 0x3, REG2; \
+	lduwa		[REG1 + REG2] ASI_PHYS_USE_EC, REG1; \
+	brz,pn		REG1, FAIL_LABEL; \
+	 sllx		VADDR, 64 - PMD_SHIFT, REG2; \
+	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
+	sllx		REG1, 11, REG1; \
+	andn		REG2, 0x7, REG2; \
+	add		REG1, REG2, REG1;
+
+	/* Do a user page table walk in MMU globals.  Leaves physical PTE
+	 * pointer in REG1.  Jumps to FAIL_LABEL on early page table walk
+	 * termination.  Physical base of page tables is in PHYS_PGD which
+	 * will not be modified.
+	 *
+	 * VADDR will not be clobbered, but REG1 and REG2 will.
+	 */
+#define USER_PGTABLE_WALK_TL1(VADDR, PHYS_PGD, REG1, REG2, FAIL_LABEL)	\
+	sllx		VADDR, 64 - (PGDIR_SHIFT + PGDIR_BITS), REG2; \
+	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
+	andn		REG2, 0x3, REG2; \
+	lduwa		[PHYS_PGD + REG2] ASI_PHYS_USE_EC, REG1; \
+	brz,pn		REG1, FAIL_LABEL; \
+	 sllx		VADDR, 64 - (PMD_SHIFT + PMD_BITS), REG2; \
+	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
+	sllx		REG1, 11, REG1; \
+	andn		REG2, 0x3, REG2; \
+	lduwa		[REG1 + REG2] ASI_PHYS_USE_EC, REG1; \
+	brz,pn		REG1, FAIL_LABEL; \
+	 sllx		VADDR, 64 - PMD_SHIFT, REG2; \
+	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
+	sllx		REG1, 11, REG1; \
+	andn		REG2, 0x7, REG2; \
+	add		REG1, REG2, REG1;
+
+/* Lookup a OBP mapping on VADDR in the prom_trans[] table at TL>0.
+ * If no entry is found, FAIL_LABEL will be branched to.  On success
+ * the resulting PTE value will be left in REG1.  VADDR is preserved
+ * by this routine.
+ */
+#define OBP_TRANS_LOOKUP(VADDR, REG1, REG2, REG3, FAIL_LABEL) \
+	sethi		%hi(prom_trans), REG1; \
+	or		REG1, %lo(prom_trans), REG1; \
+97:	ldx		[REG1 + 0x00], REG2; \
+	brz,pn		REG2, FAIL_LABEL; \
+	 nop; \
+	ldx		[REG1 + 0x08], REG3; \
+	add		REG2, REG3, REG3; \
+	cmp		REG2, VADDR; \
+	bgu,pt		%xcc, 98f; \
+	 cmp		VADDR, REG3; \
+	bgeu,pt		%xcc, 98f; \
+	 ldx		[REG1 + 0x10], REG3; \
+	sub		VADDR, REG2, REG2; \
+	ba,pt		%xcc, 99f; \
+	 add		REG3, REG2, REG1; \
+98:	ba,pt		%xcc, 97b; \
+	 add		REG1, (3 * 8), REG1; \
+99:
+
+	/* We use a 32K TSB for the whole kernel, this allows to
+	 * handle about 16MB of modules and vmalloc mappings without
+	 * incurring many hash conflicts.
+	 */
+#define KERNEL_TSB_SIZE_BYTES	(32 * 1024)
+#define KERNEL_TSB_NENTRIES	\
+	(KERNEL_TSB_SIZE_BYTES / 16)
+#define KERNEL_TSB4M_NENTRIES	4096
+
+	/* Do a kernel TSB lookup at tl>0 on VADDR+TAG, branch to OK_LABEL
+	 * on TSB hit.  REG1, REG2, REG3, and REG4 are used as temporaries
+	 * and the found TTE will be left in REG1.  REG3 and REG4 must
+	 * be an even/odd pair of registers.
+	 *
+	 * VADDR and TAG will be preserved and not clobbered by this macro.
+	 */
+#define KERN_TSB_LOOKUP_TL1(VADDR, TAG, REG1, REG2, REG3, REG4, OK_LABEL) \
+	sethi		%hi(swapper_tsb), REG1; \
+	or		REG1, %lo(swapper_tsb), REG1; \
+	srlx		VADDR, PAGE_SHIFT, REG2; \
+	and		REG2, (KERNEL_TSB_NENTRIES - 1), REG2; \
+	sllx		REG2, 4, REG2; \
+	add		REG1, REG2, REG2; \
+	KTSB_LOAD_QUAD(REG2, REG3); \
+	cmp		REG3, TAG; \
+	be,a,pt		%xcc, OK_LABEL; \
+	 mov		REG4, REG1;
+
+#ifndef CONFIG_DEBUG_PAGEALLOC
+	/* This version uses a trick, the TAG is already (VADDR >> 22) so
+	 * we can make use of that for the index computation.
+	 */
+#define KERN_TSB4M_LOOKUP_TL1(TAG, REG1, REG2, REG3, REG4, OK_LABEL) \
+	sethi		%hi(swapper_4m_tsb), REG1; \
+	or		REG1, %lo(swapper_4m_tsb), REG1; \
+	and		TAG, (KERNEL_TSB4M_NENTRIES - 1), REG2; \
+	sllx		REG2, 4, REG2; \
+	add		REG1, REG2, REG2; \
+	KTSB_LOAD_QUAD(REG2, REG3); \
+	cmp		REG3, TAG; \
+	be,a,pt		%xcc, OK_LABEL; \
+	 mov		REG4, REG1;
+#endif
+
+#endif /* !(_SPARC64_TSB_H) */
diff --git a/include/asm-sparc/ttable.h b/include/asm-sparc/ttable.h
new file mode 100644
index 0000000..5708ba2
--- /dev/null
+++ b/include/asm-sparc/ttable.h
@@ -0,0 +1,658 @@
+#ifndef _SPARC64_TTABLE_H
+#define _SPARC64_TTABLE_H
+
+#include <asm/utrap.h>
+
+#ifdef __ASSEMBLY__
+#include <asm/thread_info.h>
+#endif
+
+#define BOOT_KERNEL b sparc64_boot; nop; nop; nop; nop; nop; nop; nop;
+
+/* We need a "cleaned" instruction... */
+#define CLEAN_WINDOW							\
+	rdpr	%cleanwin, %l0;		add	%l0, 1, %l0;		\
+	wrpr	%l0, 0x0, %cleanwin;					\
+	clr	%o0;	clr	%o1;	clr	%o2;	clr	%o3;	\
+	clr	%o4;	clr	%o5;	clr	%o6;	clr	%o7;	\
+	clr	%l0;	clr	%l1;	clr	%l2;	clr	%l3;	\
+	clr	%l4;	clr	%l5;	clr	%l6;	clr	%l7;	\
+	retry;								\
+	nop;nop;nop;nop;nop;nop;nop;nop;nop;nop;nop;nop;
+
+#define TRAP(routine)					\
+	sethi	%hi(109f), %g7;				\
+	ba,pt	%xcc, etrap;				\
+109:	 or	%g7, %lo(109b), %g7;			\
+	call	routine;				\
+	 add	%sp, PTREGS_OFF, %o0;			\
+	ba,pt	%xcc, rtrap;				\
+	 nop;						\
+	nop;
+
+#define TRAP_7INSNS(routine)				\
+	sethi	%hi(109f), %g7;				\
+	ba,pt	%xcc, etrap;				\
+109:	 or	%g7, %lo(109b), %g7;			\
+	call	routine;				\
+	 add	%sp, PTREGS_OFF, %o0;			\
+	ba,pt	%xcc, rtrap;				\
+	 nop;
+
+#define TRAP_SAVEFPU(routine)				\
+	sethi	%hi(109f), %g7;				\
+	ba,pt	%xcc, do_fptrap;			\
+109:	 or	%g7, %lo(109b), %g7;			\
+	call	routine;				\
+	 add	%sp, PTREGS_OFF, %o0;			\
+	ba,pt	%xcc, rtrap;				\
+	 nop;						\
+	nop;
+
+#define TRAP_NOSAVE(routine)				\
+	ba,pt	%xcc, routine;				\
+	 nop;						\
+	nop; nop; nop; nop; nop; nop;
+
+#define TRAP_NOSAVE_7INSNS(routine)			\
+	ba,pt	%xcc, routine;				\
+	 nop;						\
+	nop; nop; nop; nop; nop;
+
+#define TRAPTL1(routine)				\
+	sethi	%hi(109f), %g7;				\
+	ba,pt	%xcc, etraptl1;				\
+109:	 or	%g7, %lo(109b), %g7;			\
+	call	routine;				\
+	 add	%sp, PTREGS_OFF, %o0;			\
+	ba,pt	%xcc, rtrap;				\
+	 nop;						\
+	nop;
+
+#define TRAP_ARG(routine, arg)				\
+	sethi	%hi(109f), %g7;				\
+	ba,pt	%xcc, etrap;				\
+109:	 or	%g7, %lo(109b), %g7;			\
+	add	%sp, PTREGS_OFF, %o0;			\
+	call	routine;				\
+	 mov	arg, %o1;				\
+	ba,pt	%xcc, rtrap;				\
+	 nop;
+
+#define TRAPTL1_ARG(routine, arg)			\
+	sethi	%hi(109f), %g7;				\
+	ba,pt	%xcc, etraptl1;				\
+109:	 or	%g7, %lo(109b), %g7;			\
+	add	%sp, PTREGS_OFF, %o0;			\
+	call	routine;				\
+	 mov	arg, %o1;				\
+	ba,pt	%xcc, rtrap;				\
+	 nop;
+
+#define SYSCALL_TRAP(routine, systbl)			\
+	rdpr	%pil, %g2;				\
+	mov	TSTATE_SYSCALL, %g3;			\
+	sethi	%hi(109f), %g7;				\
+	ba,pt	%xcc, etrap_syscall;			\
+109:	 or	%g7, %lo(109b), %g7;			\
+	sethi	%hi(systbl), %l7;			\
+	ba,pt	%xcc, routine;				\
+	 or	%l7, %lo(systbl), %l7;
+
+#define TRAP_UTRAP(handler,lvl)				\
+	mov	handler, %g3;				\
+	ba,pt	%xcc, utrap_trap;			\
+	 mov	lvl, %g4;				\
+	nop;						\
+	nop;						\
+	nop;						\
+	nop;						\
+	nop;
+
+#ifdef CONFIG_COMPAT
+#define	LINUX_32BIT_SYSCALL_TRAP SYSCALL_TRAP(linux_sparc_syscall32, sys_call_table32)
+#else
+#define	LINUX_32BIT_SYSCALL_TRAP BTRAP(0x110)
+#endif
+#define LINUX_64BIT_SYSCALL_TRAP SYSCALL_TRAP(linux_sparc_syscall, sys_call_table64)
+#define GETCC_TRAP TRAP(getcc)
+#define SETCC_TRAP TRAP(setcc)
+#define BREAKPOINT_TRAP TRAP(breakpoint_trap)
+
+#ifdef CONFIG_TRACE_IRQFLAGS
+
+#define TRAP_IRQ(routine, level)			\
+	rdpr	%pil, %g2;				\
+	wrpr	%g0, 15, %pil;				\
+	sethi	%hi(1f-4), %g7;				\
+	ba,pt	%xcc, etrap_irq;			\
+	 or	%g7, %lo(1f-4), %g7;			\
+	nop;						\
+	nop;						\
+	nop;						\
+	.subsection	2;				\
+1:	call	trace_hardirqs_off;			\
+	 nop;						\
+	mov	level, %o0;				\
+	call	routine;				\
+	 add	%sp, PTREGS_OFF, %o1;			\
+	ba,a,pt	%xcc, rtrap_irq;			\
+	.previous;
+
+#else
+
+#define TRAP_IRQ(routine, level)			\
+	rdpr	%pil, %g2;				\
+	wrpr	%g0, 15, %pil;				\
+	ba,pt	%xcc, etrap_irq;			\
+	 rd	%pc, %g7;				\
+	mov	level, %o0;				\
+	call	routine;				\
+	 add	%sp, PTREGS_OFF, %o1;			\
+	ba,a,pt	%xcc, rtrap_irq;
+
+#endif
+
+#define TRAP_IVEC TRAP_NOSAVE(do_ivec)
+
+#define BTRAP(lvl) TRAP_ARG(bad_trap, lvl)
+
+#define BTRAPTL1(lvl) TRAPTL1_ARG(bad_trap_tl1, lvl)
+
+#define FLUSH_WINDOW_TRAP						\
+	ba,pt	%xcc, etrap;						\
+	 rd	%pc, %g7;						\
+	flushw;								\
+	ldx	[%sp + PTREGS_OFF + PT_V9_TNPC], %l1;			\
+	add	%l1, 4, %l2;						\
+	stx	%l1, [%sp + PTREGS_OFF + PT_V9_TPC];			\
+	ba,pt	%xcc, rtrap;						\
+	 stx	%l2, [%sp + PTREGS_OFF + PT_V9_TNPC];
+
+#ifdef CONFIG_KPROBES
+#define KPROBES_TRAP(lvl) TRAP_IRQ(kprobe_trap, lvl)
+#else
+#define KPROBES_TRAP(lvl) TRAP_ARG(bad_trap, lvl)
+#endif
+
+#ifdef CONFIG_KGDB
+#define KGDB_TRAP(lvl) TRAP_IRQ(kgdb_trap, lvl)
+#else
+#define KGDB_TRAP(lvl) TRAP_ARG(bad_trap, lvl)
+#endif
+
+#define SUN4V_ITSB_MISS					\
+	ldxa	[%g0] ASI_SCRATCHPAD, %g2;		\
+	ldx	[%g2 + HV_FAULT_I_ADDR_OFFSET], %g4;	\
+	ldx	[%g2 + HV_FAULT_I_CTX_OFFSET], %g5;	\
+	srlx	%g4, 22, %g6;				\
+	ba,pt	%xcc, sun4v_itsb_miss;			\
+	 nop;						\
+	nop;						\
+	nop;
+
+#define SUN4V_DTSB_MISS					\
+	ldxa	[%g0] ASI_SCRATCHPAD, %g2;		\
+	ldx	[%g2 + HV_FAULT_D_ADDR_OFFSET], %g4;	\
+	ldx	[%g2 + HV_FAULT_D_CTX_OFFSET], %g5;	\
+	srlx	%g4, 22, %g6;				\
+	ba,pt	%xcc, sun4v_dtsb_miss;			\
+	 nop;						\
+	nop;						\
+	nop;
+
+/* Before touching these macros, you owe it to yourself to go and
+ * see how arch/sparc64/kernel/winfixup.S works... -DaveM
+ *
+ * For the user cases we used to use the %asi register, but
+ * it turns out that the "wr xxx, %asi" costs ~5 cycles, so
+ * now we use immediate ASI loads and stores instead.  Kudos
+ * to Greg Onufer for pointing out this performance anomaly.
+ *
+ * Further note that we cannot use the g2, g4, g5, and g7 alternate
+ * globals in the spill routines, check out the save instruction in
+ * arch/sparc64/kernel/etrap.S to see what I mean about g2, and
+ * g4/g5 are the globals which are preserved by etrap processing
+ * for the caller of it.  The g7 register is the return pc for
+ * etrap.  Finally, g6 is the current thread register so we cannot
+ * us it in the spill handlers either.  Most of these rules do not
+ * apply to fill processing, only g6 is not usable.
+ */
+
+/* Normal kernel spill */
+#define SPILL_0_NORMAL					\
+	stx	%l0, [%sp + STACK_BIAS + 0x00];		\
+	stx	%l1, [%sp + STACK_BIAS + 0x08];		\
+	stx	%l2, [%sp + STACK_BIAS + 0x10];		\
+	stx	%l3, [%sp + STACK_BIAS + 0x18];		\
+	stx	%l4, [%sp + STACK_BIAS + 0x20];		\
+	stx	%l5, [%sp + STACK_BIAS + 0x28];		\
+	stx	%l6, [%sp + STACK_BIAS + 0x30];		\
+	stx	%l7, [%sp + STACK_BIAS + 0x38];		\
+	stx	%i0, [%sp + STACK_BIAS + 0x40];		\
+	stx	%i1, [%sp + STACK_BIAS + 0x48];		\
+	stx	%i2, [%sp + STACK_BIAS + 0x50];		\
+	stx	%i3, [%sp + STACK_BIAS + 0x58];		\
+	stx	%i4, [%sp + STACK_BIAS + 0x60];		\
+	stx	%i5, [%sp + STACK_BIAS + 0x68];		\
+	stx	%i6, [%sp + STACK_BIAS + 0x70];		\
+	stx	%i7, [%sp + STACK_BIAS + 0x78];		\
+	saved; retry; nop; nop; nop; nop; nop; nop;	\
+	nop; nop; nop; nop; nop; nop; nop; nop;
+
+#define SPILL_0_NORMAL_ETRAP				\
+etrap_kernel_spill:					\
+	stx	%l0, [%sp + STACK_BIAS + 0x00];		\
+	stx	%l1, [%sp + STACK_BIAS + 0x08];		\
+	stx	%l2, [%sp + STACK_BIAS + 0x10];		\
+	stx	%l3, [%sp + STACK_BIAS + 0x18];		\
+	stx	%l4, [%sp + STACK_BIAS + 0x20];		\
+	stx	%l5, [%sp + STACK_BIAS + 0x28];		\
+	stx	%l6, [%sp + STACK_BIAS + 0x30];		\
+	stx	%l7, [%sp + STACK_BIAS + 0x38];		\
+	stx	%i0, [%sp + STACK_BIAS + 0x40];		\
+	stx	%i1, [%sp + STACK_BIAS + 0x48];		\
+	stx	%i2, [%sp + STACK_BIAS + 0x50];		\
+	stx	%i3, [%sp + STACK_BIAS + 0x58];		\
+	stx	%i4, [%sp + STACK_BIAS + 0x60];		\
+	stx	%i5, [%sp + STACK_BIAS + 0x68];		\
+	stx	%i6, [%sp + STACK_BIAS + 0x70];		\
+	stx	%i7, [%sp + STACK_BIAS + 0x78];		\
+	saved;						\
+	sub	%g1, 2, %g1;				\
+	ba,pt	%xcc, etrap_save;			\
+	wrpr	%g1, %cwp;				\
+	nop; nop; nop; nop; nop; nop; nop; nop;		\
+	nop; nop; nop; nop;
+
+/* Normal 64bit spill */
+#define SPILL_1_GENERIC(ASI)				\
+	add	%sp, STACK_BIAS + 0x00, %g1;		\
+	stxa	%l0, [%g1 + %g0] ASI;			\
+	mov	0x08, %g3;				\
+	stxa	%l1, [%g1 + %g3] ASI;			\
+	add	%g1, 0x10, %g1;				\
+	stxa	%l2, [%g1 + %g0] ASI;			\
+	stxa	%l3, [%g1 + %g3] ASI;			\
+	add	%g1, 0x10, %g1;				\
+	stxa	%l4, [%g1 + %g0] ASI;			\
+	stxa	%l5, [%g1 + %g3] ASI;			\
+	add	%g1, 0x10, %g1;				\
+	stxa	%l6, [%g1 + %g0] ASI;			\
+	stxa	%l7, [%g1 + %g3] ASI;			\
+	add	%g1, 0x10, %g1;				\
+	stxa	%i0, [%g1 + %g0] ASI;			\
+	stxa	%i1, [%g1 + %g3] ASI;			\
+	add	%g1, 0x10, %g1;				\
+	stxa	%i2, [%g1 + %g0] ASI;			\
+	stxa	%i3, [%g1 + %g3] ASI;			\
+	add	%g1, 0x10, %g1;				\
+	stxa	%i4, [%g1 + %g0] ASI;			\
+	stxa	%i5, [%g1 + %g3] ASI;			\
+	add	%g1, 0x10, %g1;				\
+	stxa	%i6, [%g1 + %g0] ASI;			\
+	stxa	%i7, [%g1 + %g3] ASI;			\
+	saved;						\
+	retry; nop; nop;				\
+	b,a,pt	%xcc, spill_fixup_dax;			\
+	b,a,pt	%xcc, spill_fixup_mna;			\
+	b,a,pt	%xcc, spill_fixup;
+
+#define SPILL_1_GENERIC_ETRAP				\
+etrap_user_spill_64bit:					\
+	stxa	%l0, [%sp + STACK_BIAS + 0x00] %asi;	\
+	stxa	%l1, [%sp + STACK_BIAS + 0x08] %asi;	\
+	stxa	%l2, [%sp + STACK_BIAS + 0x10] %asi;	\
+	stxa	%l3, [%sp + STACK_BIAS + 0x18] %asi;	\
+	stxa	%l4, [%sp + STACK_BIAS + 0x20] %asi;	\
+	stxa	%l5, [%sp + STACK_BIAS + 0x28] %asi;	\
+	stxa	%l6, [%sp + STACK_BIAS + 0x30] %asi;	\
+	stxa	%l7, [%sp + STACK_BIAS + 0x38] %asi;	\
+	stxa	%i0, [%sp + STACK_BIAS + 0x40] %asi;	\
+	stxa	%i1, [%sp + STACK_BIAS + 0x48] %asi;	\
+	stxa	%i2, [%sp + STACK_BIAS + 0x50] %asi;	\
+	stxa	%i3, [%sp + STACK_BIAS + 0x58] %asi;	\
+	stxa	%i4, [%sp + STACK_BIAS + 0x60] %asi;	\
+	stxa	%i5, [%sp + STACK_BIAS + 0x68] %asi;	\
+	stxa	%i6, [%sp + STACK_BIAS + 0x70] %asi;	\
+	stxa	%i7, [%sp + STACK_BIAS + 0x78] %asi;	\
+	saved;						\
+	sub	%g1, 2, %g1;				\
+	ba,pt	%xcc, etrap_save;			\
+	 wrpr	%g1, %cwp;				\
+	nop; nop; nop; nop; nop;			\
+	nop; nop; nop; nop;				\
+	ba,a,pt	%xcc, etrap_spill_fixup_64bit;		\
+	ba,a,pt	%xcc, etrap_spill_fixup_64bit;		\
+	ba,a,pt	%xcc, etrap_spill_fixup_64bit;
+
+#define SPILL_1_GENERIC_ETRAP_FIXUP			\
+etrap_spill_fixup_64bit:				\
+	ldub	[%g6 + TI_WSAVED], %g1;			\
+	sll	%g1, 3, %g3;				\
+	add	%g6, %g3, %g3;				\
+	stx	%sp, [%g3 + TI_RWIN_SPTRS];		\
+	sll	%g1, 7, %g3;				\
+	add	%g6, %g3, %g3;				\
+	stx	%l0, [%g3 + TI_REG_WINDOW + 0x00];	\
+	stx	%l1, [%g3 + TI_REG_WINDOW + 0x08];	\
+	stx	%l2, [%g3 + TI_REG_WINDOW + 0x10];	\
+	stx	%l3, [%g3 + TI_REG_WINDOW + 0x18];	\
+	stx	%l4, [%g3 + TI_REG_WINDOW + 0x20];	\
+	stx	%l5, [%g3 + TI_REG_WINDOW + 0x28];	\
+	stx	%l6, [%g3 + TI_REG_WINDOW + 0x30];	\
+	stx	%l7, [%g3 + TI_REG_WINDOW + 0x38];	\
+	stx	%i0, [%g3 + TI_REG_WINDOW + 0x40];	\
+	stx	%i1, [%g3 + TI_REG_WINDOW + 0x48];	\
+	stx	%i2, [%g3 + TI_REG_WINDOW + 0x50];	\
+	stx	%i3, [%g3 + TI_REG_WINDOW + 0x58];	\
+	stx	%i4, [%g3 + TI_REG_WINDOW + 0x60];	\
+	stx	%i5, [%g3 + TI_REG_WINDOW + 0x68];	\
+	stx	%i6, [%g3 + TI_REG_WINDOW + 0x70];	\
+	stx	%i7, [%g3 + TI_REG_WINDOW + 0x78];	\
+	add	%g1, 1, %g1;				\
+	stb	%g1, [%g6 + TI_WSAVED];			\
+	saved;						\
+	rdpr	%cwp, %g1;				\
+	sub	%g1, 2, %g1;				\
+	ba,pt	%xcc, etrap_save;			\
+	 wrpr	%g1, %cwp;				\
+	nop; nop; nop
+
+/* Normal 32bit spill */
+#define SPILL_2_GENERIC(ASI)				\
+	srl	%sp, 0, %sp;				\
+	stwa	%l0, [%sp + %g0] ASI;			\
+	mov	0x04, %g3;				\
+	stwa	%l1, [%sp + %g3] ASI;			\
+	add	%sp, 0x08, %g1;				\
+	stwa	%l2, [%g1 + %g0] ASI;			\
+	stwa	%l3, [%g1 + %g3] ASI;			\
+	add	%g1, 0x08, %g1;				\
+	stwa	%l4, [%g1 + %g0] ASI;			\
+	stwa	%l5, [%g1 + %g3] ASI;			\
+	add	%g1, 0x08, %g1;				\
+	stwa	%l6, [%g1 + %g0] ASI;			\
+	stwa	%l7, [%g1 + %g3] ASI;			\
+	add	%g1, 0x08, %g1;				\
+	stwa	%i0, [%g1 + %g0] ASI;			\
+	stwa	%i1, [%g1 + %g3] ASI;			\
+	add	%g1, 0x08, %g1;				\
+	stwa	%i2, [%g1 + %g0] ASI;			\
+	stwa	%i3, [%g1 + %g3] ASI;			\
+	add	%g1, 0x08, %g1;				\
+	stwa	%i4, [%g1 + %g0] ASI;			\
+	stwa	%i5, [%g1 + %g3] ASI;			\
+	add	%g1, 0x08, %g1;				\
+	stwa	%i6, [%g1 + %g0] ASI;			\
+	stwa	%i7, [%g1 + %g3] ASI;			\
+	saved;						\
+        retry; nop; nop;				\
+	b,a,pt	%xcc, spill_fixup_dax;			\
+	b,a,pt	%xcc, spill_fixup_mna;			\
+	b,a,pt	%xcc, spill_fixup;
+
+#define SPILL_2_GENERIC_ETRAP		\
+etrap_user_spill_32bit:			\
+	srl	%sp, 0, %sp;		\
+	stwa	%l0, [%sp + 0x00] %asi;	\
+	stwa	%l1, [%sp + 0x04] %asi;	\
+	stwa	%l2, [%sp + 0x08] %asi;	\
+	stwa	%l3, [%sp + 0x0c] %asi;	\
+	stwa	%l4, [%sp + 0x10] %asi;	\
+	stwa	%l5, [%sp + 0x14] %asi;	\
+	stwa	%l6, [%sp + 0x18] %asi;	\
+	stwa	%l7, [%sp + 0x1c] %asi;	\
+	stwa	%i0, [%sp + 0x20] %asi;	\
+	stwa	%i1, [%sp + 0x24] %asi;	\
+	stwa	%i2, [%sp + 0x28] %asi;	\
+	stwa	%i3, [%sp + 0x2c] %asi;	\
+	stwa	%i4, [%sp + 0x30] %asi;	\
+	stwa	%i5, [%sp + 0x34] %asi;	\
+	stwa	%i6, [%sp + 0x38] %asi;	\
+	stwa	%i7, [%sp + 0x3c] %asi;	\
+	saved;				\
+	sub	%g1, 2, %g1;		\
+	ba,pt	%xcc, etrap_save;	\
+	 wrpr	%g1, %cwp;		\
+	nop; nop; nop; nop;		\
+	nop; nop; nop; nop;		\
+	ba,a,pt	%xcc, etrap_spill_fixup_32bit; \
+	ba,a,pt	%xcc, etrap_spill_fixup_32bit; \
+	ba,a,pt	%xcc, etrap_spill_fixup_32bit;
+
+#define SPILL_2_GENERIC_ETRAP_FIXUP			\
+etrap_spill_fixup_32bit:				\
+	ldub	[%g6 + TI_WSAVED], %g1;			\
+	sll	%g1, 3, %g3;				\
+	add	%g6, %g3, %g3;				\
+	stx	%sp, [%g3 + TI_RWIN_SPTRS];		\
+	sll	%g1, 7, %g3;				\
+	add	%g6, %g3, %g3;				\
+	stw	%l0, [%g3 + TI_REG_WINDOW + 0x00];	\
+	stw	%l1, [%g3 + TI_REG_WINDOW + 0x04];	\
+	stw	%l2, [%g3 + TI_REG_WINDOW + 0x08];	\
+	stw	%l3, [%g3 + TI_REG_WINDOW + 0x0c];	\
+	stw	%l4, [%g3 + TI_REG_WINDOW + 0x10];	\
+	stw	%l5, [%g3 + TI_REG_WINDOW + 0x14];	\
+	stw	%l6, [%g3 + TI_REG_WINDOW + 0x18];	\
+	stw	%l7, [%g3 + TI_REG_WINDOW + 0x1c];	\
+	stw	%i0, [%g3 + TI_REG_WINDOW + 0x20];	\
+	stw	%i1, [%g3 + TI_REG_WINDOW + 0x24];	\
+	stw	%i2, [%g3 + TI_REG_WINDOW + 0x28];	\
+	stw	%i3, [%g3 + TI_REG_WINDOW + 0x2c];	\
+	stw	%i4, [%g3 + TI_REG_WINDOW + 0x30];	\
+	stw	%i5, [%g3 + TI_REG_WINDOW + 0x34];	\
+	stw	%i6, [%g3 + TI_REG_WINDOW + 0x38];	\
+	stw	%i7, [%g3 + TI_REG_WINDOW + 0x3c];	\
+	add	%g1, 1, %g1;				\
+	stb	%g1, [%g6 + TI_WSAVED];			\
+	saved;						\
+	rdpr	%cwp, %g1;				\
+	sub	%g1, 2, %g1;				\
+	ba,pt	%xcc, etrap_save;			\
+	 wrpr	%g1, %cwp;				\
+	nop; nop; nop
+
+#define SPILL_1_NORMAL SPILL_1_GENERIC(ASI_AIUP)
+#define SPILL_2_NORMAL SPILL_2_GENERIC(ASI_AIUP)
+#define SPILL_3_NORMAL SPILL_0_NORMAL
+#define SPILL_4_NORMAL SPILL_0_NORMAL
+#define SPILL_5_NORMAL SPILL_0_NORMAL
+#define SPILL_6_NORMAL SPILL_0_NORMAL
+#define SPILL_7_NORMAL SPILL_0_NORMAL
+
+#define SPILL_0_OTHER SPILL_0_NORMAL
+#define SPILL_1_OTHER SPILL_1_GENERIC(ASI_AIUS)
+#define SPILL_2_OTHER SPILL_2_GENERIC(ASI_AIUS)
+#define SPILL_3_OTHER SPILL_3_NORMAL
+#define SPILL_4_OTHER SPILL_4_NORMAL
+#define SPILL_5_OTHER SPILL_5_NORMAL
+#define SPILL_6_OTHER SPILL_6_NORMAL
+#define SPILL_7_OTHER SPILL_7_NORMAL
+
+/* Normal kernel fill */
+#define FILL_0_NORMAL					\
+	ldx	[%sp + STACK_BIAS + 0x00], %l0;		\
+	ldx	[%sp + STACK_BIAS + 0x08], %l1;		\
+	ldx	[%sp + STACK_BIAS + 0x10], %l2;		\
+	ldx	[%sp + STACK_BIAS + 0x18], %l3;		\
+	ldx	[%sp + STACK_BIAS + 0x20], %l4;		\
+	ldx	[%sp + STACK_BIAS + 0x28], %l5;		\
+	ldx	[%sp + STACK_BIAS + 0x30], %l6;		\
+	ldx	[%sp + STACK_BIAS + 0x38], %l7;		\
+	ldx	[%sp + STACK_BIAS + 0x40], %i0;		\
+	ldx	[%sp + STACK_BIAS + 0x48], %i1;		\
+	ldx	[%sp + STACK_BIAS + 0x50], %i2;		\
+	ldx	[%sp + STACK_BIAS + 0x58], %i3;		\
+	ldx	[%sp + STACK_BIAS + 0x60], %i4;		\
+	ldx	[%sp + STACK_BIAS + 0x68], %i5;		\
+	ldx	[%sp + STACK_BIAS + 0x70], %i6;		\
+	ldx	[%sp + STACK_BIAS + 0x78], %i7;		\
+	restored; retry; nop; nop; nop; nop; nop; nop;	\
+	nop; nop; nop; nop; nop; nop; nop; nop;
+
+#define FILL_0_NORMAL_RTRAP				\
+kern_rtt_fill:						\
+	rdpr	%cwp, %g1;				\
+	sub	%g1, 1, %g1;				\
+	wrpr	%g1, %cwp;				\
+	ldx	[%sp + STACK_BIAS + 0x00], %l0;		\
+	ldx	[%sp + STACK_BIAS + 0x08], %l1;		\
+	ldx	[%sp + STACK_BIAS + 0x10], %l2;		\
+	ldx	[%sp + STACK_BIAS + 0x18], %l3;		\
+	ldx	[%sp + STACK_BIAS + 0x20], %l4;		\
+	ldx	[%sp + STACK_BIAS + 0x28], %l5;		\
+	ldx	[%sp + STACK_BIAS + 0x30], %l6;		\
+	ldx	[%sp + STACK_BIAS + 0x38], %l7;		\
+	ldx	[%sp + STACK_BIAS + 0x40], %i0;		\
+	ldx	[%sp + STACK_BIAS + 0x48], %i1;		\
+	ldx	[%sp + STACK_BIAS + 0x50], %i2;		\
+	ldx	[%sp + STACK_BIAS + 0x58], %i3;		\
+	ldx	[%sp + STACK_BIAS + 0x60], %i4;		\
+	ldx	[%sp + STACK_BIAS + 0x68], %i5;		\
+	ldx	[%sp + STACK_BIAS + 0x70], %i6;		\
+	ldx	[%sp + STACK_BIAS + 0x78], %i7;		\
+	restored;					\
+	add	%g1, 1, %g1;				\
+	ba,pt	%xcc, kern_rtt_restore;			\
+	 wrpr	%g1, %cwp;				\
+	nop; nop; nop; nop; nop;			\
+	nop; nop; nop; nop;
+
+
+/* Normal 64bit fill */
+#define FILL_1_GENERIC(ASI)				\
+	add	%sp, STACK_BIAS + 0x00, %g1;		\
+	ldxa	[%g1 + %g0] ASI, %l0;			\
+	mov	0x08, %g2;				\
+	mov	0x10, %g3;				\
+	ldxa	[%g1 + %g2] ASI, %l1;			\
+	mov	0x18, %g5;				\
+	ldxa	[%g1 + %g3] ASI, %l2;			\
+	ldxa	[%g1 + %g5] ASI, %l3;			\
+	add	%g1, 0x20, %g1;				\
+	ldxa	[%g1 + %g0] ASI, %l4;			\
+	ldxa	[%g1 + %g2] ASI, %l5;			\
+	ldxa	[%g1 + %g3] ASI, %l6;			\
+	ldxa	[%g1 + %g5] ASI, %l7;			\
+	add	%g1, 0x20, %g1;				\
+	ldxa	[%g1 + %g0] ASI, %i0;			\
+	ldxa	[%g1 + %g2] ASI, %i1;			\
+	ldxa	[%g1 + %g3] ASI, %i2;			\
+	ldxa	[%g1 + %g5] ASI, %i3;			\
+	add	%g1, 0x20, %g1;				\
+	ldxa	[%g1 + %g0] ASI, %i4;			\
+	ldxa	[%g1 + %g2] ASI, %i5;			\
+	ldxa	[%g1 + %g3] ASI, %i6;			\
+	ldxa	[%g1 + %g5] ASI, %i7;			\
+	restored;					\
+	retry; nop; nop; nop; nop;			\
+	b,a,pt	%xcc, fill_fixup_dax;			\
+	b,a,pt	%xcc, fill_fixup_mna;			\
+	b,a,pt	%xcc, fill_fixup;
+
+#define FILL_1_GENERIC_RTRAP				\
+user_rtt_fill_64bit:					\
+	ldxa	[%sp + STACK_BIAS + 0x00] %asi, %l0;	\
+	ldxa	[%sp + STACK_BIAS + 0x08] %asi, %l1;	\
+	ldxa	[%sp + STACK_BIAS + 0x10] %asi, %l2;	\
+	ldxa	[%sp + STACK_BIAS + 0x18] %asi, %l3;	\
+	ldxa	[%sp + STACK_BIAS + 0x20] %asi, %l4;	\
+	ldxa	[%sp + STACK_BIAS + 0x28] %asi, %l5;	\
+	ldxa	[%sp + STACK_BIAS + 0x30] %asi, %l6;	\
+	ldxa	[%sp + STACK_BIAS + 0x38] %asi, %l7;	\
+	ldxa	[%sp + STACK_BIAS + 0x40] %asi, %i0;	\
+	ldxa	[%sp + STACK_BIAS + 0x48] %asi, %i1;	\
+	ldxa	[%sp + STACK_BIAS + 0x50] %asi, %i2;	\
+	ldxa	[%sp + STACK_BIAS + 0x58] %asi, %i3;	\
+	ldxa	[%sp + STACK_BIAS + 0x60] %asi, %i4;	\
+	ldxa	[%sp + STACK_BIAS + 0x68] %asi, %i5;	\
+	ldxa	[%sp + STACK_BIAS + 0x70] %asi, %i6;	\
+	ldxa	[%sp + STACK_BIAS + 0x78] %asi, %i7;	\
+	ba,pt	%xcc, user_rtt_pre_restore;		\
+	 restored;					\
+	nop; nop; nop; nop; nop; nop;			\
+	nop; nop; nop; nop; nop;			\
+	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
+	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
+	ba,a,pt	%xcc, user_rtt_fill_fixup;
+
+
+/* Normal 32bit fill */
+#define FILL_2_GENERIC(ASI)				\
+	srl	%sp, 0, %sp;				\
+	lduwa	[%sp + %g0] ASI, %l0;			\
+	mov	0x04, %g2;				\
+	mov	0x08, %g3;				\
+	lduwa	[%sp + %g2] ASI, %l1;			\
+	mov	0x0c, %g5;				\
+	lduwa	[%sp + %g3] ASI, %l2;			\
+	lduwa	[%sp + %g5] ASI, %l3;			\
+	add	%sp, 0x10, %g1;				\
+	lduwa	[%g1 + %g0] ASI, %l4;			\
+	lduwa	[%g1 + %g2] ASI, %l5;			\
+	lduwa	[%g1 + %g3] ASI, %l6;			\
+	lduwa	[%g1 + %g5] ASI, %l7;			\
+	add	%g1, 0x10, %g1;				\
+	lduwa	[%g1 + %g0] ASI, %i0;			\
+	lduwa	[%g1 + %g2] ASI, %i1;			\
+	lduwa	[%g1 + %g3] ASI, %i2;			\
+	lduwa	[%g1 + %g5] ASI, %i3;			\
+	add	%g1, 0x10, %g1;				\
+	lduwa	[%g1 + %g0] ASI, %i4;			\
+	lduwa	[%g1 + %g2] ASI, %i5;			\
+	lduwa	[%g1 + %g3] ASI, %i6;			\
+	lduwa	[%g1 + %g5] ASI, %i7;			\
+	restored;					\
+	retry; nop; nop; nop; nop;			\
+	b,a,pt	%xcc, fill_fixup_dax;			\
+	b,a,pt	%xcc, fill_fixup_mna;			\
+	b,a,pt	%xcc, fill_fixup;
+
+#define FILL_2_GENERIC_RTRAP				\
+user_rtt_fill_32bit:					\
+	srl	%sp, 0, %sp;				\
+	lduwa	[%sp + 0x00] %asi, %l0;			\
+	lduwa	[%sp + 0x04] %asi, %l1;			\
+	lduwa	[%sp + 0x08] %asi, %l2;			\
+	lduwa	[%sp + 0x0c] %asi, %l3;			\
+	lduwa	[%sp + 0x10] %asi, %l4;			\
+	lduwa	[%sp + 0x14] %asi, %l5;			\
+	lduwa	[%sp + 0x18] %asi, %l6;			\
+	lduwa	[%sp + 0x1c] %asi, %l7;			\
+	lduwa	[%sp + 0x20] %asi, %i0;			\
+	lduwa	[%sp + 0x24] %asi, %i1;			\
+	lduwa	[%sp + 0x28] %asi, %i2;			\
+	lduwa	[%sp + 0x2c] %asi, %i3;			\
+	lduwa	[%sp + 0x30] %asi, %i4;			\
+	lduwa	[%sp + 0x34] %asi, %i5;			\
+	lduwa	[%sp + 0x38] %asi, %i6;			\
+	lduwa	[%sp + 0x3c] %asi, %i7;			\
+	ba,pt	%xcc, user_rtt_pre_restore;		\
+	 restored;					\
+	nop; nop; nop; nop; nop;			\
+	nop; nop; nop; nop; nop;			\
+	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
+	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
+	ba,a,pt	%xcc, user_rtt_fill_fixup;
+
+
+#define FILL_1_NORMAL FILL_1_GENERIC(ASI_AIUP)
+#define FILL_2_NORMAL FILL_2_GENERIC(ASI_AIUP)
+#define FILL_3_NORMAL FILL_0_NORMAL
+#define FILL_4_NORMAL FILL_0_NORMAL
+#define FILL_5_NORMAL FILL_0_NORMAL
+#define FILL_6_NORMAL FILL_0_NORMAL
+#define FILL_7_NORMAL FILL_0_NORMAL
+
+#define FILL_0_OTHER FILL_0_NORMAL
+#define FILL_1_OTHER FILL_1_GENERIC(ASI_AIUS)
+#define FILL_2_OTHER FILL_2_GENERIC(ASI_AIUS)
+#define FILL_3_OTHER FILL_3_NORMAL
+#define FILL_4_OTHER FILL_4_NORMAL
+#define FILL_5_OTHER FILL_5_NORMAL
+#define FILL_6_OTHER FILL_6_NORMAL
+#define FILL_7_OTHER FILL_7_NORMAL
+
+#endif /* !(_SPARC64_TTABLE_H) */
diff --git a/include/asm-sparc/upa.h b/include/asm-sparc/upa.h
new file mode 100644
index 0000000..5b16332
--- /dev/null
+++ b/include/asm-sparc/upa.h
@@ -0,0 +1,109 @@
+#ifndef _SPARC64_UPA_H
+#define _SPARC64_UPA_H
+
+#include <asm/asi.h>
+
+/* UPA level registers and defines. */
+
+/* UPA Config Register */
+#define UPA_CONFIG_RESV		0xffffffffc0000000 /* Reserved.                    */
+#define UPA_CONFIG_PCON		0x000000003fc00000 /* Depth of various sys queues. */
+#define UPA_CONFIG_MID		0x00000000003e0000 /* Module ID.                   */
+#define UPA_CONFIG_PCAP		0x000000000001ffff /* Port Capabilities.           */
+
+/* UPA Port ID Register */
+#define UPA_PORTID_FNP		0xff00000000000000 /* Hardcoded to 0xfc on ultra.  */
+#define UPA_PORTID_RESV		0x00fffff800000000 /* Reserved.                    */
+#define UPA_PORTID_ECCVALID     0x0000000400000000 /* Zero if mod can generate ECC */
+#define UPA_PORTID_ONEREAD      0x0000000200000000 /* Set if mod generates P_RASB  */
+#define UPA_PORTID_PINTRDQ      0x0000000180000000 /* # outstanding P_INT_REQ's    */
+#define UPA_PORTID_PREQDQ       0x000000007e000000 /* slave-wr's to mod supported  */
+#define UPA_PORTID_PREQRD       0x0000000001e00000 /* # incoming P_REQ's supported */
+#define UPA_PORTID_UPACAP       0x00000000001f0000 /* UPA capabilities of mod      */
+#define UPA_PORTID_ID           0x000000000000ffff /* Module Identification bits  */
+
+/* UPA I/O space accessors */
+#if defined(__KERNEL__) && !defined(__ASSEMBLY__)
+static inline unsigned char _upa_readb(unsigned long addr)
+{
+	unsigned char ret;
+
+	__asm__ __volatile__("lduba\t[%1] %2, %0\t/* upa_readb */"
+			     : "=r" (ret)
+			     : "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+
+	return ret;
+}
+
+static inline unsigned short _upa_readw(unsigned long addr)
+{
+	unsigned short ret;
+
+	__asm__ __volatile__("lduha\t[%1] %2, %0\t/* upa_readw */"
+			     : "=r" (ret)
+			     : "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+
+	return ret;
+}
+
+static inline unsigned int _upa_readl(unsigned long addr)
+{
+	unsigned int ret;
+
+	__asm__ __volatile__("lduwa\t[%1] %2, %0\t/* upa_readl */"
+			     : "=r" (ret)
+			     : "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+
+	return ret;
+}
+
+static inline unsigned long _upa_readq(unsigned long addr)
+{
+	unsigned long ret;
+
+	__asm__ __volatile__("ldxa\t[%1] %2, %0\t/* upa_readq */"
+			     : "=r" (ret)
+			     : "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+
+	return ret;
+}
+
+static inline void _upa_writeb(unsigned char b, unsigned long addr)
+{
+	__asm__ __volatile__("stba\t%0, [%1] %2\t/* upa_writeb */"
+			     : /* no outputs */
+			     : "r" (b), "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+}
+
+static inline void _upa_writew(unsigned short w, unsigned long addr)
+{
+	__asm__ __volatile__("stha\t%0, [%1] %2\t/* upa_writew */"
+			     : /* no outputs */
+			     : "r" (w), "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+}
+
+static inline void _upa_writel(unsigned int l, unsigned long addr)
+{
+	__asm__ __volatile__("stwa\t%0, [%1] %2\t/* upa_writel */"
+			     : /* no outputs */
+			     : "r" (l), "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+}
+
+static inline void _upa_writeq(unsigned long q, unsigned long addr)
+{
+	__asm__ __volatile__("stxa\t%0, [%1] %2\t/* upa_writeq */"
+			     : /* no outputs */
+			     : "r" (q), "r" (addr), "i" (ASI_PHYS_BYPASS_EC_E));
+}
+
+#define upa_readb(__addr)		(_upa_readb((unsigned long)(__addr)))
+#define upa_readw(__addr)		(_upa_readw((unsigned long)(__addr)))
+#define upa_readl(__addr)		(_upa_readl((unsigned long)(__addr)))
+#define upa_readq(__addr)		(_upa_readq((unsigned long)(__addr)))
+#define upa_writeb(__b, __addr)		(_upa_writeb((__b), (unsigned long)(__addr)))
+#define upa_writew(__w, __addr)		(_upa_writew((__w), (unsigned long)(__addr)))
+#define upa_writel(__l, __addr)		(_upa_writel((__l), (unsigned long)(__addr)))
+#define upa_writeq(__q, __addr)		(_upa_writeq((__q), (unsigned long)(__addr)))
+#endif /* __KERNEL__ && !__ASSEMBLY__ */
+
+#endif /* !(_SPARC64_UPA_H) */
diff --git a/include/asm-sparc/vio.h b/include/asm-sparc/vio.h
new file mode 100644
index 0000000..d4de32f
--- /dev/null
+++ b/include/asm-sparc/vio.h
@@ -0,0 +1,406 @@
+#ifndef _SPARC64_VIO_H
+#define _SPARC64_VIO_H
+
+#include <linux/kernel.h>
+#include <linux/device.h>
+#include <linux/mod_devicetable.h>
+#include <linux/timer.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/list.h>
+#include <linux/log2.h>
+
+#include <asm/ldc.h>
+#include <asm/mdesc.h>
+
+struct vio_msg_tag {
+	u8			type;
+#define VIO_TYPE_CTRL		0x01
+#define VIO_TYPE_DATA		0x02
+#define VIO_TYPE_ERR		0x04
+
+	u8			stype;
+#define VIO_SUBTYPE_INFO	0x01
+#define VIO_SUBTYPE_ACK		0x02
+#define VIO_SUBTYPE_NACK	0x04
+
+	u16			stype_env;
+#define VIO_VER_INFO		0x0001
+#define VIO_ATTR_INFO		0x0002
+#define VIO_DRING_REG		0x0003
+#define VIO_DRING_UNREG		0x0004
+#define VIO_RDX			0x0005
+#define VIO_PKT_DATA		0x0040
+#define VIO_DESC_DATA		0x0041
+#define VIO_DRING_DATA		0x0042
+#define VNET_MCAST_INFO		0x0101
+
+	u32		sid;
+};
+
+struct vio_rdx {
+	struct vio_msg_tag	tag;
+	u64			resv[6];
+};
+
+struct vio_ver_info {
+	struct vio_msg_tag	tag;
+	u16			major;
+	u16			minor;
+	u8			dev_class;
+#define VDEV_NETWORK		0x01
+#define VDEV_NETWORK_SWITCH	0x02
+#define VDEV_DISK		0x03
+#define VDEV_DISK_SERVER	0x04
+
+	u8			resv1[3];
+	u64			resv2[5];
+};
+
+struct vio_dring_register {
+	struct vio_msg_tag	tag;
+	u64			dring_ident;
+	u32			num_descr;
+	u32			descr_size;
+	u16			options;
+#define VIO_TX_DRING		0x0001
+#define VIO_RX_DRING		0x0002
+	u16			resv;
+	u32			num_cookies;
+	struct ldc_trans_cookie	cookies[0];
+};
+
+struct vio_dring_unregister {
+	struct vio_msg_tag	tag;
+	u64			dring_ident;
+	u64			resv[5];
+};
+
+/* Data transfer modes */
+#define VIO_PKT_MODE		0x01 /* Packet based transfer	*/
+#define VIO_DESC_MODE		0x02 /* In-band descriptors	*/
+#define VIO_DRING_MODE		0x03 /* Descriptor rings	*/
+
+struct vio_dring_data {
+	struct vio_msg_tag	tag;
+	u64			seq;
+	u64			dring_ident;
+	u32			start_idx;
+	u32			end_idx;
+	u8			state;
+#define VIO_DRING_ACTIVE	0x01
+#define VIO_DRING_STOPPED	0x02
+
+	u8			__pad1;
+	u16			__pad2;
+	u32			__pad3;
+	u64			__par4[2];
+};
+
+struct vio_dring_hdr {
+	u8			state;
+#define VIO_DESC_FREE		0x01
+#define VIO_DESC_READY		0x02
+#define VIO_DESC_ACCEPTED	0x03
+#define VIO_DESC_DONE		0x04
+	u8			ack;
+#define VIO_ACK_ENABLE		0x01
+#define VIO_ACK_DISABLE		0x00
+
+	u16			__pad1;
+	u32			__pad2;
+};
+
+/* VIO disk specific structures and defines */
+struct vio_disk_attr_info {
+	struct vio_msg_tag	tag;
+	u8			xfer_mode;
+	u8			vdisk_type;
+#define VD_DISK_TYPE_SLICE	0x01 /* Slice in block device	*/
+#define VD_DISK_TYPE_DISK	0x02 /* Entire block device	*/
+	u16			resv1;
+	u32			vdisk_block_size;
+	u64			operations;
+	u64			vdisk_size;
+	u64			max_xfer_size;
+	u64			resv2[2];
+};
+
+struct vio_disk_desc {
+	struct vio_dring_hdr	hdr;
+	u64			req_id;
+	u8			operation;
+#define VD_OP_BREAD		0x01 /* Block read			*/
+#define VD_OP_BWRITE		0x02 /* Block write			*/
+#define VD_OP_FLUSH		0x03 /* Flush disk contents		*/
+#define VD_OP_GET_WCE		0x04 /* Get write-cache status		*/
+#define VD_OP_SET_WCE		0x05 /* Enable/disable write-cache	*/
+#define VD_OP_GET_VTOC		0x06 /* Get VTOC			*/
+#define VD_OP_SET_VTOC		0x07 /* Set VTOC			*/
+#define VD_OP_GET_DISKGEOM	0x08 /* Get disk geometry		*/
+#define VD_OP_SET_DISKGEOM	0x09 /* Set disk geometry		*/
+#define VD_OP_SCSICMD		0x0a /* SCSI control command		*/
+#define VD_OP_GET_DEVID		0x0b /* Get device ID			*/
+#define VD_OP_GET_EFI		0x0c /* Get EFI				*/
+#define VD_OP_SET_EFI		0x0d /* Set EFI				*/
+	u8			slice;
+	u16			resv1;
+	u32			status;
+	u64			offset;
+	u64			size;
+	u32			ncookies;
+	u32			resv2;
+	struct ldc_trans_cookie	cookies[0];
+};
+
+#define VIO_DISK_VNAME_LEN	8
+#define VIO_DISK_ALABEL_LEN	128
+#define VIO_DISK_NUM_PART	8
+
+struct vio_disk_vtoc {
+	u8			volume_name[VIO_DISK_VNAME_LEN];
+	u16			sector_size;
+	u16			num_partitions;
+	u8			ascii_label[VIO_DISK_ALABEL_LEN];
+	struct {
+		u16		id;
+		u16		perm_flags;
+		u32		resv;
+		u64		start_block;
+		u64		num_blocks;
+	} partitions[VIO_DISK_NUM_PART];
+};
+
+struct vio_disk_geom {
+	u16			num_cyl; /* Num data cylinders		*/
+	u16			alt_cyl; /* Num alternate cylinders	*/
+	u16			beg_cyl; /* Cyl off of fixed head area	*/
+	u16			num_hd;  /* Num heads			*/
+	u16			num_sec; /* Num sectors			*/
+	u16			ifact;   /* Interleave factor		*/
+	u16			apc;     /* Alts per cylinder (SCSI)	*/
+	u16			rpm;	 /* Revolutions per minute	*/
+	u16			phy_cyl; /* Num physical cylinders	*/
+	u16			wr_skip; /* Num sects to skip, writes	*/
+	u16			rd_skip; /* Num sects to skip, writes	*/
+};
+
+struct vio_disk_devid {
+	u16			resv;
+	u16			type;
+	u32			len;
+	char			id[0];
+};
+
+struct vio_disk_efi {
+	u64			lba;
+	u64			len;
+	char			data[0];
+};
+
+/* VIO net specific structures and defines */
+struct vio_net_attr_info {
+	struct vio_msg_tag	tag;
+	u8			xfer_mode;
+	u8			addr_type;
+#define VNET_ADDR_ETHERMAC	0x01
+	u16			ack_freq;
+	u32			resv1;
+	u64			addr;
+	u64			mtu;
+	u64			resv2[3];
+};
+
+#define VNET_NUM_MCAST		7
+
+struct vio_net_mcast_info {
+	struct vio_msg_tag	tag;
+	u8			set;
+	u8			count;
+	u8			mcast_addr[VNET_NUM_MCAST * 6];
+	u32			resv;
+};
+
+struct vio_net_desc {
+	struct vio_dring_hdr	hdr;
+	u32			size;
+	u32			ncookies;
+	struct ldc_trans_cookie	cookies[0];
+};
+
+#define VIO_MAX_RING_COOKIES	24
+
+struct vio_dring_state {
+	u64			ident;
+	void			*base;
+	u64			snd_nxt;
+	u64			rcv_nxt;
+	u32			entry_size;
+	u32			num_entries;
+	u32			prod;
+	u32			cons;
+	u32			pending;
+	int			ncookies;
+	struct ldc_trans_cookie	cookies[VIO_MAX_RING_COOKIES];
+};
+
+static inline void *vio_dring_cur(struct vio_dring_state *dr)
+{
+	return dr->base + (dr->entry_size * dr->prod);
+}
+
+static inline void *vio_dring_entry(struct vio_dring_state *dr,
+				    unsigned int index)
+{
+	return dr->base + (dr->entry_size * index);
+}
+
+static inline u32 vio_dring_avail(struct vio_dring_state *dr,
+				  unsigned int ring_size)
+{
+	BUILD_BUG_ON(!is_power_of_2(ring_size));
+
+	return (dr->pending -
+		((dr->prod - dr->cons) & (ring_size - 1)));
+}
+
+#define VIO_MAX_TYPE_LEN	32
+#define VIO_MAX_COMPAT_LEN	64
+
+struct vio_dev {
+	u64			mp;
+	struct device_node	*dp;
+
+	char			type[VIO_MAX_TYPE_LEN];
+	char			compat[VIO_MAX_COMPAT_LEN];
+	int			compat_len;
+
+	u64			dev_no;
+
+	unsigned long		channel_id;
+
+	unsigned int		tx_irq;
+	unsigned int		rx_irq;
+
+	struct device		dev;
+};
+
+struct vio_driver {
+	struct list_head		node;
+	const struct vio_device_id	*id_table;
+	int (*probe)(struct vio_dev *dev, const struct vio_device_id *id);
+	int (*remove)(struct vio_dev *dev);
+	void (*shutdown)(struct vio_dev *dev);
+	unsigned long			driver_data;
+	struct device_driver		driver;
+};
+
+struct vio_version {
+	u16		major;
+	u16		minor;
+};
+
+struct vio_driver_state;
+struct vio_driver_ops {
+	int	(*send_attr)(struct vio_driver_state *vio);
+	int	(*handle_attr)(struct vio_driver_state *vio, void *pkt);
+	void	(*handshake_complete)(struct vio_driver_state *vio);
+};
+
+struct vio_completion {
+	struct completion	com;
+	int			err;
+	int			waiting_for;
+};
+
+struct vio_driver_state {
+	/* Protects VIO handshake and, optionally, driver private state.  */
+	spinlock_t		lock;
+
+	struct ldc_channel	*lp;
+
+	u32			_peer_sid;
+	u32			_local_sid;
+	struct vio_dring_state	drings[2];
+#define VIO_DRIVER_TX_RING	0
+#define VIO_DRIVER_RX_RING	1
+
+	u8			hs_state;
+#define VIO_HS_INVALID		0x00
+#define VIO_HS_GOTVERS		0x01
+#define VIO_HS_GOT_ATTR		0x04
+#define VIO_HS_SENT_DREG	0x08
+#define VIO_HS_SENT_RDX		0x10
+#define VIO_HS_GOT_RDX_ACK	0x20
+#define VIO_HS_GOT_RDX		0x40
+#define VIO_HS_SENT_RDX_ACK	0x80
+#define VIO_HS_COMPLETE		(VIO_HS_GOT_RDX_ACK | VIO_HS_SENT_RDX_ACK)
+
+	u8			dev_class;
+
+	u8			dr_state;
+#define VIO_DR_STATE_TXREG	0x01
+#define VIO_DR_STATE_RXREG	0x02
+#define VIO_DR_STATE_TXREQ	0x10
+#define VIO_DR_STATE_RXREQ	0x20
+
+	u8			debug;
+#define VIO_DEBUG_HS		0x01
+#define VIO_DEBUG_DATA		0x02
+
+	void			*desc_buf;
+	unsigned int		desc_buf_len;
+
+	struct vio_completion	*cmp;
+
+	struct vio_dev		*vdev;
+
+	struct timer_list	timer;
+
+	struct vio_version	ver;
+
+	struct vio_version	*ver_table;
+	int			ver_table_entries;
+
+	char			*name;
+
+	struct vio_driver_ops	*ops;
+};
+
+#define viodbg(TYPE, f, a...) \
+do {	if (vio->debug & VIO_DEBUG_##TYPE) \
+		printk(KERN_INFO "vio: ID[%lu] " f, \
+		       vio->vdev->channel_id, ## a); \
+} while (0)
+
+extern int vio_register_driver(struct vio_driver *drv);
+extern void vio_unregister_driver(struct vio_driver *drv);
+
+static inline struct vio_driver *to_vio_driver(struct device_driver *drv)
+{
+	return container_of(drv, struct vio_driver, driver);
+}
+
+static inline struct vio_dev *to_vio_dev(struct device *dev)
+{
+	return container_of(dev, struct vio_dev, dev);
+}
+
+extern int vio_ldc_send(struct vio_driver_state *vio, void *data, int len);
+extern void vio_link_state_change(struct vio_driver_state *vio, int event);
+extern void vio_conn_reset(struct vio_driver_state *vio);
+extern int vio_control_pkt_engine(struct vio_driver_state *vio, void *pkt);
+extern int vio_validate_sid(struct vio_driver_state *vio,
+			    struct vio_msg_tag *tp);
+extern u32 vio_send_sid(struct vio_driver_state *vio);
+extern int vio_ldc_alloc(struct vio_driver_state *vio,
+			 struct ldc_channel_config *base_cfg, void *event_arg);
+extern void vio_ldc_free(struct vio_driver_state *vio);
+extern int vio_driver_init(struct vio_driver_state *vio, struct vio_dev *vdev,
+			   u8 dev_class, struct vio_version *ver_table,
+			   int ver_table_size, struct vio_driver_ops *ops,
+			   char *name);
+
+extern void vio_port_up(struct vio_driver_state *vio);
+
+#endif /* _SPARC64_VIO_H */
diff --git a/include/asm-sparc/visasm.h b/include/asm-sparc/visasm.h
new file mode 100644
index 0000000..de797b9
--- /dev/null
+++ b/include/asm-sparc/visasm.h
@@ -0,0 +1,62 @@
+#ifndef _SPARC64_VISASM_H
+#define _SPARC64_VISASM_H
+
+/* visasm.h:  FPU saving macros for VIS routines
+ *
+ * Copyright (C) 1998 Jakub Jelinek (jj@ultra.linux.cz)
+ */
+
+#include <asm/pstate.h>
+#include <asm/ptrace.h>
+
+/* Clobbers %o5, %g1, %g2, %g3, %g7, %icc, %xcc */
+
+#define VISEntry					\
+	rd		%fprs, %o5;			\
+	andcc		%o5, (FPRS_FEF|FPRS_DU), %g0;	\
+	be,pt		%icc, 297f;			\
+	 sethi		%hi(297f), %g7;			\
+	sethi		%hi(VISenter), %g1;		\
+	jmpl		%g1 + %lo(VISenter), %g0;	\
+	 or		%g7, %lo(297f), %g7;		\
+297:	wr		%g0, FPRS_FEF, %fprs;		\
+
+#define VISExit						\
+	wr		%g0, 0, %fprs;
+
+/* Clobbers %o5, %g1, %g2, %g3, %g7, %icc, %xcc.
+ * Must preserve %o5 between VISEntryHalf and VISExitHalf */
+
+#define VISEntryHalf					\
+	rd		%fprs, %o5;			\
+	andcc		%o5, FPRS_FEF, %g0;		\
+	be,pt		%icc, 297f;			\
+	 sethi		%hi(298f), %g7;			\
+	sethi		%hi(VISenterhalf), %g1;		\
+	jmpl		%g1 + %lo(VISenterhalf), %g0;	\
+	 or		%g7, %lo(298f), %g7;		\
+	clr		%o5;				\
+297:	wr		%o5, FPRS_FEF, %fprs;		\
+298:
+
+#define VISExitHalf					\
+	wr		%o5, 0, %fprs;
+
+#ifndef __ASSEMBLY__
+static inline void save_and_clear_fpu(void) {
+	__asm__ __volatile__ (
+"		rd %%fprs, %%o5\n"
+"		andcc %%o5, %0, %%g0\n"
+"		be,pt %%icc, 299f\n"
+"		 sethi %%hi(298f), %%g7\n"
+"		sethi %%hi(VISenter), %%g1\n"
+"		jmpl %%g1 + %%lo(VISenter), %%g0\n"
+"		 or %%g7, %%lo(298f), %%g7\n"
+"	298:	wr %%g0, 0, %%fprs\n"
+"	299:\n"
+"		" : : "i" (FPRS_FEF|FPRS_DU) :
+		"o5", "g1", "g2", "g3", "g7", "cc");
+}
+#endif
+
+#endif /* _SPARC64_ASI_H */