Merge tag 'parisc-for-6.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux

Pull parisc architecture updates from Helge Deller:
 "PA-RISC now has a native eBPF JIT compiler for 32- and 64-bit kernels,
  the LED driver was rewritten to use the Linux LED framework and most
  of the parisc bootup code was switched to use *_initcall() functions.

  Summary:

   - add eBPF JIT compiler for 32- and 64-bit kernel

   - LCD/LED driver rewrite to utilize Linux LED subsystem

   - switch to generic mmap top-down layout and brk randomization

   - kernel startup cleanup by loading most drivers via arch_initcall()"

* tag 'parisc-for-6.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux: (31 commits)
  parisc: ccio-dma: Create private runway procfs root entry
  parisc: chassis: Do not overwrite string on LCD display
  parisc: led: Rewrite LED/LCD driver to utilizize Linux LED subsystem
  parisc: led: Fix LAN receive and transmit LEDs
  parisc: lasi: Initialize LASI driver via arch_initcall()
  parisc: asp: Initialize asp driver via arch_initcall()
  parisc: wax: Initialize wax driver via arch_initcall()
  parisc: iosapic: Convert I/O Sapic driver to use arch_initcall()
  parisc: sba_iommu: Convert SBA IOMMU driver to use arch_initcall()
  parisc: led: Move register_led_regions() to late_initcall()
  parisc: lba: Convert LBA PCI bus driver to use arch_initcall()
  parisc: gsc: Convert GSC bus driver to use arch_initcall()
  parisc: ccio: Convert CCIO driver to use arch_initcall()
  parisc: eisa: Convert HP EISA bus driver to use arch_initcall()
  parisc: hppb: Convert HP PB bus driver to use arch_initcall()
  parisc: dino: Convert dino PCI bus driver to use arch_initcall()
  parisc: Makefile: Adjust order in which drivers should be loaded
  parisc: led: Reduce CPU overhead for disk & lan LED computation
  parisc: Avoid ioremap() for same addresss in iosapic_register()
  parisc: unaligned: Simplify 32-bit assembly in emulate_std()
  ...

23 files changed, 3572 insertions, 238 deletions

diff --git a/arch/parisc/Kbuild b/arch/parisc/Kbuild
index a6d3b280ba0c..749b195f2894 100644
--- a/arch/parisc/Kbuild
+++ b/arch/parisc/Kbuild
@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0-only
-obj-y	+= mm/ kernel/ math-emu/
+obj-y	+= mm/ kernel/ math-emu/ net/
 
 # for cleaning
 subdir- += boot
diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
index 4cb46d5c64a2..dad281808ab5 100644
--- a/arch/parisc/Kconfig
+++ b/arch/parisc/Kconfig
@@ -49,6 +49,9 @@ config PARISC
 	select TTY # Needed for pdc_cons.c
 	select HAS_IOPORT if PCI || EISA
 	select HAVE_DEBUG_STACKOVERFLOW
+	select ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT
+	select HAVE_ARCH_MMAP_RND_COMPAT_BITS if COMPAT
+	select HAVE_ARCH_MMAP_RND_BITS
 	select HAVE_ARCH_AUDITSYSCALL
 	select HAVE_ARCH_HASH
 	select HAVE_ARCH_JUMP_LABEL
@@ -56,6 +59,8 @@ config PARISC
 	select HAVE_ARCH_KFENCE
 	select HAVE_ARCH_SECCOMP_FILTER
 	select HAVE_ARCH_TRACEHOOK
+	select HAVE_EBPF_JIT
+	select ARCH_WANT_DEFAULT_BPF_JIT
 	select HAVE_REGS_AND_STACK_ACCESS_API
 	select HOTPLUG_CORE_SYNC_DEAD if HOTPLUG_CPU
 	select GENERIC_SCHED_CLOCK
@@ -124,6 +129,20 @@ config TIME_LOW_RES
 	depends on SMP
 	default y
 
+config ARCH_MMAP_RND_BITS_MIN
+	default 18 if 64BIT
+	default 8
+
+config ARCH_MMAP_RND_COMPAT_BITS_MIN
+	default 8
+
+config ARCH_MMAP_RND_BITS_MAX
+	default 24 if 64BIT
+	default 17
+
+config ARCH_MMAP_RND_COMPAT_BITS_MAX
+	default 17
+
 # unless you want to implement ACPI on PA-RISC ... ;-)
 config PM
 	bool
diff --git a/arch/parisc/Kconfig.debug b/arch/parisc/Kconfig.debug
index bf2b21b96f0b..f4f164eb12df 100644
--- a/arch/parisc/Kconfig.debug
+++ b/arch/parisc/Kconfig.debug
@@ -13,7 +13,7 @@ config LIGHTWEIGHT_SPINLOCK_CHECK
 
 config TLB_PTLOCK
 	bool "Use page table locks in TLB fault handler"
-	depends on SMP
+	depends on DEBUG_KERNEL && SMP
 	default n
 	help
 	  Select this option to enable page table locking in the TLB
diff --git a/arch/parisc/include/asm/elf.h b/arch/parisc/include/asm/elf.h
index cc426d365892..140eaa97bf21 100644
--- a/arch/parisc/include/asm/elf.h
+++ b/arch/parisc/include/asm/elf.h
@@ -163,8 +163,7 @@ typedef struct elf32_fdesc {
 
 /* Format for the Elf64 Function descriptor */
 typedef struct elf64_fdesc {
-	__u64	dummy[2]; /* FIXME: nothing uses these, why waste
-			   * the space */
+	__u64	dummy[2]; /* used by 64-bit eBPF and tracing functions */
 	__u64	addr;
 	__u64	gp;
 } Elf64_Fdesc;
diff --git a/arch/parisc/include/asm/led.h b/arch/parisc/include/asm/led.h
index 6de13d08a388..0aea47eff48d 100644
--- a/arch/parisc/include/asm/led.h
+++ b/arch/parisc/include/asm/led.h
@@ -11,8 +11,8 @@
 #define	LED1		0x02
 #define	LED0		0x01		/* bottom (or furthest left) LED */
 
-#define	LED_LAN_TX	LED0		/* for LAN transmit activity */
-#define	LED_LAN_RCV	LED1		/* for LAN receive activity */
+#define	LED_LAN_RCV	LED0		/* for LAN receive activity */
+#define	LED_LAN_TX	LED1		/* for LAN transmit activity */
 #define	LED_DISK_IO	LED2		/* for disk activity */
 #define	LED_HEARTBEAT	LED3		/* heartbeat */
 
@@ -25,19 +25,13 @@
 #define LED_CMD_REG_NONE 0		/* NULL == no addr for the cmd register */
 
 /* register_led_driver() */
-int __init register_led_driver(int model, unsigned long cmd_reg, unsigned long data_reg);
-
-/* registers the LED regions for procfs */
-void __init register_led_regions(void);
+int register_led_driver(int model, unsigned long cmd_reg, unsigned long data_reg);
 
 #ifdef CONFIG_CHASSIS_LCD_LED
 /* writes a string to the LCD display (if possible on this h/w) */
-int lcd_print(const char *str);
+void lcd_print(const char *str);
 #else
-#define lcd_print(str)
+#define lcd_print(str) do { } while (0)
 #endif
 
-/* main LED initialization function (uses PDC) */ 
-int __init led_init(void);
-
 #endif /* LED_H */
diff --git a/arch/parisc/include/asm/machdep.h b/arch/parisc/include/asm/machdep.h
deleted file mode 100644
index 215d2c43989d..000000000000
--- a/arch/parisc/include/asm/machdep.h
+++ /dev/null
@@ -1,17 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-#ifndef _PARISC_MACHDEP_H
-#define _PARISC_MACHDEP_H
-
-#include <linux/notifier.h>
-
-#define	MACH_RESTART	1
-#define	MACH_HALT	2
-#define MACH_POWER_ON	3
-#define	MACH_POWER_OFF	4
-
-extern struct notifier_block *mach_notifier;
-extern void pa7300lc_init(void);
-
-extern void (*cpu_lpmc)(int, struct pt_regs *);
-
-#endif
diff --git a/arch/parisc/include/asm/processor.h b/arch/parisc/include/asm/processor.h
index e132b2819fc9..d77c43d32974 100644
--- a/arch/parisc/include/asm/processor.h
+++ b/arch/parisc/include/asm/processor.h
@@ -313,15 +313,7 @@ extern void collect_boot_cpu_data(void);
 extern int show_cpuinfo (struct seq_file *m, void *v);
 
 /* driver code in driver/parisc */
-extern void gsc_init(void);
 extern void processor_init(void);
-extern void ccio_init(void);
-extern void hppb_init(void);
-extern void dino_init(void);
-extern void iosapic_init(void);
-extern void lba_init(void);
-extern void sba_init(void);
-extern void parisc_eisa_init(void);
 struct parisc_device;
 struct resource;
 extern void sba_distributed_lmmio(struct parisc_device *, struct resource *);
diff --git a/arch/parisc/include/asm/ropes.h b/arch/parisc/include/asm/ropes.h
index 8e51c775c80a..fd96706c7234 100644
--- a/arch/parisc/include/asm/ropes.h
+++ b/arch/parisc/include/asm/ropes.h
@@ -252,7 +252,7 @@ static inline int agp_mode_mercury(void __iomem *hpa) {
 ** fixup_irq is to initialize PCI IRQ line support and
 ** virtualize pcidev->irq value. To be called by pci_fixup_bus().
 */
-extern void *iosapic_register(unsigned long hpa);
+extern void *iosapic_register(unsigned long hpa, void __iomem *vaddr);
 extern int iosapic_fixup_irq(void *obj, struct pci_dev *pcidev);
 
 #define LBA_FUNC_ID	0x0000	/* function id */
diff --git a/arch/parisc/include/asm/runway.h b/arch/parisc/include/asm/runway.h
index 5cf061376ddb..2837f0223d6d 100644
--- a/arch/parisc/include/asm/runway.h
+++ b/arch/parisc/include/asm/runway.h
@@ -2,9 +2,6 @@
 #ifndef ASM_PARISC_RUNWAY_H
 #define ASM_PARISC_RUNWAY_H
 
-/* declared in arch/parisc/kernel/setup.c */
-extern struct proc_dir_entry * proc_runway_root;
-
 #define RUNWAY_STATUS	0x10
 #define RUNWAY_DEBUG	0x40
 
diff --git a/arch/parisc/kernel/Makefile b/arch/parisc/kernel/Makefile
index 2d1478fc4aa5..5ab0467be70a 100644
--- a/arch/parisc/kernel/Makefile
+++ b/arch/parisc/kernel/Makefile
@@ -6,7 +6,7 @@
 extra-y		:= vmlinux.lds
 
 obj-y		:= head.o cache.o pacache.o setup.o pdt.o traps.o time.o irq.o \
-		   pa7300lc.o syscall.o entry.o sys_parisc.o firmware.o \
+		   syscall.o entry.o sys_parisc.o firmware.o \
 		   ptrace.o hardware.o inventory.o drivers.o alternative.o \
 		   signal.o hpmc.o real2.o parisc_ksyms.o unaligned.o \
 		   process.o processor.o pdc_cons.o pdc_chassis.o unwind.o \
diff --git a/arch/parisc/kernel/pa7300lc.c b/arch/parisc/kernel/pa7300lc.c
deleted file mode 100644
index 0d770ac83f70..000000000000
--- a/arch/parisc/kernel/pa7300lc.c
+++ /dev/null
@@ -1,51 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0
-/*
- *   linux/arch/parisc/kernel/pa7300lc.c
- *	- PA7300LC-specific functions	
- *
- *   Copyright (C) 2000 Philipp Rumpf */
-
-#include <linux/sched.h>
-#include <linux/sched/debug.h>
-#include <linux/smp.h>
-#include <linux/kernel.h>
-#include <asm/io.h>
-#include <asm/ptrace.h>
-#include <asm/machdep.h>
-
-/* CPU register indices */
-
-#define MIOC_STATUS	0xf040
-#define MIOC_CONTROL	0xf080
-#define MDERRADD	0xf0e0
-#define DMAERR		0xf0e8
-#define DIOERR		0xf0ec
-#define HIDMAMEM	0xf0f4
-
-/* this returns the HPA of the CPU it was called on */
-static u32 cpu_hpa(void)
-{
-	return 0xfffb0000;
-}
-
-static void pa7300lc_lpmc(int code, struct pt_regs *regs)
-{
-	u32 hpa;
-	printk(KERN_WARNING "LPMC on CPU %d\n", smp_processor_id());
-
-	show_regs(regs);
-
-	hpa = cpu_hpa();
-	printk(KERN_WARNING
-		"MIOC_CONTROL %08x\n" "MIOC_STATUS  %08x\n"
-		"MDERRADD     %08x\n" "DMAERR       %08x\n"
-		"DIOERR       %08x\n" "HIDMAMEM     %08x\n",
-		gsc_readl(hpa+MIOC_CONTROL), gsc_readl(hpa+MIOC_STATUS),
-		gsc_readl(hpa+MDERRADD), gsc_readl(hpa+DMAERR),
-		gsc_readl(hpa+DIOERR), gsc_readl(hpa+HIDMAMEM));
-}
-
-void pa7300lc_init(void)
-{
-	cpu_lpmc = pa7300lc_lpmc;
-}
diff --git a/arch/parisc/kernel/pdc_chassis.c b/arch/parisc/kernel/pdc_chassis.c
index 0a9d7008ef2a..d477d0177c2f 100644
--- a/arch/parisc/kernel/pdc_chassis.c
+++ b/arch/parisc/kernel/pdc_chassis.c
@@ -31,6 +31,7 @@
 #include <asm/processor.h>
 #include <asm/pdc.h>
 #include <asm/pdcpat.h>
+#include <asm/led.h>
 
 #define PDC_CHASSIS_VER	"0.05"
 
@@ -234,6 +235,11 @@ int pdc_chassis_send_status(int message)
 		} else retval = -1;
 #endif /* CONFIG_64BIT */
 	}	/* if (pdc_chassis_enabled) */
+
+	/* if system has LCD display, update current string */
+	if (retval != -1 && IS_ENABLED(CONFIG_CHASSIS_LCD_LED))
+		lcd_print(NULL);
+
 #endif /* CONFIG_PDC_CHASSIS */
 	return retval;
 }
diff --git a/arch/parisc/kernel/process.c b/arch/parisc/kernel/process.c
index abdbf038d643..ed93bd8c1545 100644
--- a/arch/parisc/kernel/process.c
+++ b/arch/parisc/kernel/process.c
@@ -97,18 +97,12 @@ void machine_restart(char *cmd)
 
 }
 
-void (*chassis_power_off)(void);
-
 /*
  * This routine is called from sys_reboot to actually turn off the
  * machine 
  */
 void machine_power_off(void)
 {
-	/* If there is a registered power off handler, call it. */
-	if (chassis_power_off)
-		chassis_power_off();
-
 	/* Put the soft power button back under hardware control.
 	 * If the user had already pressed the power button, the
 	 * following call will immediately power off. */
@@ -284,17 +278,3 @@ __get_wchan(struct task_struct *p)
 	} while (count++ < MAX_UNWIND_ENTRIES);
 	return 0;
 }
-
-static inline unsigned long brk_rnd(void)
-{
-	return (get_random_u32() & BRK_RND_MASK) << PAGE_SHIFT;
-}
-
-unsigned long arch_randomize_brk(struct mm_struct *mm)
-{
-	unsigned long ret = PAGE_ALIGN(mm->brk + brk_rnd());
-
-	if (ret < mm->brk)
-		return mm->brk;
-	return ret;
-}
diff --git a/arch/parisc/kernel/processor.c b/arch/parisc/kernel/processor.c
index 762289b9984e..a0e2d37c5b3b 100644
--- a/arch/parisc/kernel/processor.c
+++ b/arch/parisc/kernel/processor.c
@@ -378,10 +378,18 @@ int
 show_cpuinfo (struct seq_file *m, void *v)
 {
 	unsigned long cpu;
+	char cpu_name[60], *p;
+
+	/* strip PA path from CPU name to not confuse lscpu */
+	strlcpy(cpu_name, per_cpu(cpu_data, 0).dev->name, sizeof(cpu_name));
+	p = strrchr(cpu_name, '[');
+	if (p)
+		*(--p) = 0;
 
 	for_each_online_cpu(cpu) {
-		const struct cpuinfo_parisc *cpuinfo = &per_cpu(cpu_data, cpu);
 #ifdef CONFIG_SMP
+		const struct cpuinfo_parisc *cpuinfo = &per_cpu(cpu_data, cpu);
+
 		if (0 == cpuinfo->hpa)
 			continue;
 #endif
@@ -426,8 +434,7 @@ show_cpuinfo (struct seq_file *m, void *v)
 
 		seq_printf(m, "model\t\t: %s - %s\n",
 				 boot_cpu_data.pdc.sys_model_name,
-				 cpuinfo->dev ?
-				 cpuinfo->dev->name : "Unknown");
+				 cpu_name);
 
 		seq_printf(m, "hversion\t: 0x%08x\n"
 			        "sversion\t: 0x%08x\n",
diff --git a/arch/parisc/kernel/setup.c b/arch/parisc/kernel/setup.c
index 211a4afdd282..2f434f2da185 100644
--- a/arch/parisc/kernel/setup.c
+++ b/arch/parisc/kernel/setup.c
@@ -31,7 +31,6 @@
 #include <asm/sections.h>
 #include <asm/pdc.h>
 #include <asm/led.h>
-#include <asm/machdep.h>	/* for pa7300lc_init() proto */
 #include <asm/pdc_chassis.h>
 #include <asm/io.h>
 #include <asm/setup.h>
@@ -93,8 +92,6 @@ static void __init dma_ops_init(void)
 			"the PA-RISC 1.1 or 2.0 architecture specification.\n");
 
 	case pcxl2:
-		pa7300lc_init();
-		break;
 	default:
 		break;
 	}
@@ -146,11 +143,6 @@ void __init setup_arch(char **cmdline_p)
 	parisc_cache_init();
 	paging_init();
 
-#ifdef CONFIG_CHASSIS_LCD_LED
-	/* initialize the LCD/LED after boot_cpu_data is available ! */
-	led_init();		/* LCD/LED initialization */
-#endif
-
 #ifdef CONFIG_PA11
 	dma_ops_init();
 #endif
@@ -281,47 +273,6 @@ static int __init parisc_init(void)
 
 	apply_alternatives_all();
 	parisc_setup_cache_timing();
-
-	/* These are in a non-obvious order, will fix when we have an iotree */
-#if defined(CONFIG_IOSAPIC)
-	iosapic_init();
-#endif
-#if defined(CONFIG_IOMMU_SBA)
-	sba_init();
-#endif
-#if defined(CONFIG_PCI_LBA)
-	lba_init();
-#endif
-
-	/* CCIO before any potential subdevices */
-#if defined(CONFIG_IOMMU_CCIO)
-	ccio_init();
-#endif
-
-	/*
-	 * Need to register Asp & Wax before the EISA adapters for the IRQ
-	 * regions.  EISA must come before PCI to be sure it gets IRQ region
-	 * 0.
-	 */
-#if defined(CONFIG_GSC_LASI) || defined(CONFIG_GSC_WAX)
-	gsc_init();
-#endif
-#ifdef CONFIG_EISA
-	parisc_eisa_init();
-#endif
-
-#if defined(CONFIG_HPPB)
-	hppb_init();
-#endif
-
-#if defined(CONFIG_GSC_DINO)
-	dino_init();
-#endif
-
-#ifdef CONFIG_CHASSIS_LCD_LED
-	register_led_regions();	/* register LED port info in procfs */
-#endif
-
 	return 0;
 }
 arch_initcall(parisc_init);
diff --git a/arch/parisc/kernel/sys_parisc.c b/arch/parisc/kernel/sys_parisc.c
index 9915062d5243..ab896eff7a1d 100644
--- a/arch/parisc/kernel/sys_parisc.c
+++ b/arch/parisc/kernel/sys_parisc.c
@@ -161,7 +161,7 @@ static unsigned long arch_get_unmapped_area_common(struct file *filp,
 	}
 
 	info.flags = 0;
-	info.low_limit = mm->mmap_legacy_base;
+	info.low_limit = mm->mmap_base;
 	info.high_limit = mmap_upper_limit(NULL);
 	return vm_unmapped_area(&info);
 }
@@ -181,58 +181,6 @@ unsigned long arch_get_unmapped_area_topdown(struct file *filp,
 			addr, len, pgoff, flags, DOWN);
 }
 
-static int mmap_is_legacy(void)
-{
-	if (current->personality & ADDR_COMPAT_LAYOUT)
-		return 1;
-
-	/* parisc stack always grows up - so a unlimited stack should
-	 * not be an indicator to use the legacy memory layout.
-	 * if (rlimit(RLIMIT_STACK) == RLIM_INFINITY)
-	 *	return 1;
-	 */
-
-	return sysctl_legacy_va_layout;
-}
-
-static unsigned long mmap_rnd(void)
-{
-	unsigned long rnd = 0;
-
-	if (current->flags & PF_RANDOMIZE)
-		rnd = get_random_u32() & MMAP_RND_MASK;
-
-	return rnd << PAGE_SHIFT;
-}
-
-unsigned long arch_mmap_rnd(void)
-{
-	return (get_random_u32() & MMAP_RND_MASK) << PAGE_SHIFT;
-}
-
-static unsigned long mmap_legacy_base(void)
-{
-	return TASK_UNMAPPED_BASE + mmap_rnd();
-}
-
-/*
- * This function, called very early during the creation of a new
- * process VM image, sets up which VM layout function to use:
- */
-void arch_pick_mmap_layout(struct mm_struct *mm, struct rlimit *rlim_stack)
-{
-	mm->mmap_legacy_base = mmap_legacy_base();
-	mm->mmap_base = mmap_upper_limit(rlim_stack);
-
-	if (mmap_is_legacy()) {
-		mm->mmap_base = mm->mmap_legacy_base;
-		mm->get_unmapped_area = arch_get_unmapped_area;
-	} else {
-		mm->get_unmapped_area = arch_get_unmapped_area_topdown;
-	}
-}
-
-
 asmlinkage unsigned long sys_mmap2(unsigned long addr, unsigned long len,
 	unsigned long prot, unsigned long flags, unsigned long fd,
 	unsigned long pgoff)
diff --git a/arch/parisc/kernel/traps.c b/arch/parisc/kernel/traps.c
index 3b97944c7291..1107ca819ac8 100644
--- a/arch/parisc/kernel/traps.c
+++ b/arch/parisc/kernel/traps.c
@@ -335,9 +335,6 @@ static void default_trap(int code, struct pt_regs *regs)
 	show_regs(regs);
 }
 
-void (*cpu_lpmc) (int code, struct pt_regs *regs) __read_mostly = default_trap;
-
-
 static void transfer_pim_to_trap_frame(struct pt_regs *regs)
 {
     register int i;
@@ -557,7 +554,7 @@ void notrace handle_interruption(int code, struct pt_regs *regs)
 		
 		flush_cache_all();
 		flush_tlb_all();
-		cpu_lpmc(5, regs);
+		default_trap(code, regs);
 		return;
 
 	case  PARISC_ITLB_TRAP:
diff --git a/arch/parisc/kernel/unaligned.c b/arch/parisc/kernel/unaligned.c
index 170d0dda4213..ce25acfe4889 100644
--- a/arch/parisc/kernel/unaligned.c
+++ b/arch/parisc/kernel/unaligned.c
@@ -338,25 +338,24 @@ static int emulate_std(struct pt_regs *regs, int frreg, int flop)
 	: "r19", "r20", "r21", "r22", "r1" );
 #else
     {
-	unsigned long valh = (val >> 32), vall = (val & 0xffffffffl);
 	__asm__ __volatile__ (
-"	mtsp	%4, %%sr1\n"
-"	zdep	%2, 29, 2, %%r19\n"
-"	dep	%%r0, 31, 2, %3\n"
+"	mtsp	%3, %%sr1\n"
+"	zdep	%R1, 29, 2, %%r19\n"
+"	dep	%%r0, 31, 2, %2\n"
 "	mtsar	%%r19\n"
 "	zvdepi	-2, 32, %%r19\n"
-"1:	ldw	0(%%sr1,%3),%%r20\n"
-"2:	ldw	8(%%sr1,%3),%%r21\n"
-"	vshd	%1, %2, %%r1\n"
+"1:	ldw	0(%%sr1,%2),%%r20\n"
+"2:	ldw	8(%%sr1,%2),%%r21\n"
+"	vshd	%1, %R1, %%r1\n"
 "	vshd	%%r0, %1, %1\n"
-"	vshd	%2, %%r0, %2\n"
+"	vshd	%R1, %%r0, %R1\n"
 "	and	%%r20, %%r19, %%r20\n"
 "	andcm	%%r21, %%r19, %%r21\n"
 "	or	%1, %%r20, %1\n"
-"	or	%2, %%r21, %2\n"
-"3:	stw	%1,0(%%sr1,%3)\n"
-"4:	stw	%%r1,4(%%sr1,%3)\n"
-"5:	stw	%2,8(%%sr1,%3)\n"
+"	or	%R1, %%r21, %R1\n"
+"3:	stw	%1,0(%%sr1,%2)\n"
+"4:	stw	%%r1,4(%%sr1,%2)\n"
+"5:	stw	%R1,8(%%sr1,%2)\n"
 "6:	\n"
 	ASM_EXCEPTIONTABLE_ENTRY_EFAULT(1b, 6b)
 	ASM_EXCEPTIONTABLE_ENTRY_EFAULT(2b, 6b)
@@ -364,7 +363,7 @@ static int emulate_std(struct pt_regs *regs, int frreg, int flop)
 	ASM_EXCEPTIONTABLE_ENTRY_EFAULT(4b, 6b)
 	ASM_EXCEPTIONTABLE_ENTRY_EFAULT(5b, 6b)
 	: "+r" (ret)
-	: "r" (valh), "r" (vall), "r" (regs->ior), "r" (regs->isr)
+	: "r" (val), "r" (regs->ior), "r" (regs->isr)
 	: "r19", "r20", "r21", "r1" );
     }
 #endif
diff --git a/arch/parisc/net/Makefile b/arch/parisc/net/Makefile
new file mode 100644
index 000000000000..22b12024d4c3
--- /dev/null
+++ b/arch/parisc/net/Makefile
@@ -0,0 +1,9 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+obj-$(CONFIG_BPF_JIT) += bpf_jit_core.o
+
+ifeq ($(CONFIG_64BIT),y)
+	obj-$(CONFIG_BPF_JIT) += bpf_jit_comp64.o
+else
+	obj-$(CONFIG_BPF_JIT) += bpf_jit_comp32.o
+endif
diff --git a/arch/parisc/net/bpf_jit.h b/arch/parisc/net/bpf_jit.h
new file mode 100644
index 000000000000..8b8896959f04
--- /dev/null
+++ b/arch/parisc/net/bpf_jit.h
@@ -0,0 +1,479 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Common functionality for PARISC32 and PARISC64 BPF JIT compilers
+ *
+ * Copyright (c) 2023 Helge Deller <deller@gmx.de>
+ *
+ */
+
+#ifndef _BPF_JIT_H
+#define _BPF_JIT_H
+
+#include <linux/bpf.h>
+#include <linux/filter.h>
+#include <asm/cacheflush.h>
+
+#define HPPA_JIT_DEBUG	0
+#define HPPA_JIT_REBOOT	0
+#define HPPA_JIT_DUMP	0
+
+#define OPTIMIZE_HPPA	1	/* enable some asm optimizations */
+// echo 1 > /proc/sys/net/core/bpf_jit_enable
+
+#define HPPA_R(nr)	nr	/* use HPPA register #nr */
+
+enum {
+	HPPA_REG_ZERO =	0,	/* The constant value 0 */
+	HPPA_REG_R1 =	1,	/* used for addil */
+	HPPA_REG_RP =	2,	/* Return address */
+
+	HPPA_REG_ARG7 =	19,	/* ARG4-7 used in 64-bit ABI */
+	HPPA_REG_ARG6 =	20,
+	HPPA_REG_ARG5 =	21,
+	HPPA_REG_ARG4 =	22,
+
+	HPPA_REG_ARG3 =	23,	/* ARG0-3 in 32- and 64-bit ABI */
+	HPPA_REG_ARG2 =	24,
+	HPPA_REG_ARG1 =	25,
+	HPPA_REG_ARG0 =	26,
+	HPPA_REG_GP =	27,	/* Global pointer */
+	HPPA_REG_RET0 =	28,	/* Return value, HI in 32-bit */
+	HPPA_REG_RET1 =	29,	/* Return value, LOW in 32-bit */
+	HPPA_REG_SP =	30,	/* Stack pointer */
+	HPPA_REG_R31 =	31,
+
+#ifdef CONFIG_64BIT
+	HPPA_REG_TCC	     = 3,
+	HPPA_REG_TCC_SAVED   = 4,
+	HPPA_REG_TCC_IN_INIT = HPPA_REG_R31,
+#else
+	HPPA_REG_TCC	     = 18,
+	HPPA_REG_TCC_SAVED   = 17,
+	HPPA_REG_TCC_IN_INIT = HPPA_REG_R31,
+#endif
+
+	HPPA_REG_T0 =	HPPA_REG_R1,	/* Temporaries */
+	HPPA_REG_T1 =	HPPA_REG_R31,
+	HPPA_REG_T2 =	HPPA_REG_ARG4,
+#ifndef CONFIG_64BIT
+	HPPA_REG_T3 =	HPPA_REG_ARG5,	/* not used in 64-bit */
+	HPPA_REG_T4 =	HPPA_REG_ARG6,
+	HPPA_REG_T5 =	HPPA_REG_ARG7,
+#endif
+};
+
+struct hppa_jit_context {
+	struct bpf_prog *prog;
+	u32 *insns;		/* HPPA insns */
+	int ninsns;
+	int reg_seen_collect;
+	int reg_seen;
+	int body_len;
+	int epilogue_offset;
+	int prologue_len;
+	int *offset;		/* BPF to HPPA */
+};
+
+#define REG_SET_SEEN(ctx, nr)	{ if (ctx->reg_seen_collect) ctx->reg_seen |= BIT(nr); }
+#define REG_SET_SEEN_ALL(ctx)	{ if (ctx->reg_seen_collect) ctx->reg_seen = -1; }
+#define REG_FORCE_SEEN(ctx, nr)	{ ctx->reg_seen |= BIT(nr); }
+#define REG_WAS_SEEN(ctx, nr)	(ctx->reg_seen & BIT(nr))
+#define REG_ALL_SEEN(ctx)	(ctx->reg_seen == -1)
+
+#define HPPA_INSN_SIZE		4	/* bytes per HPPA asm instruction */
+#define REG_SIZE		REG_SZ	/* bytes per native "long" word */
+
+/* subtract hppa displacement on branches which is .+8 */
+#define HPPA_BRANCH_DISPLACEMENT  2	/* instructions */
+
+/* asm statement indicator to execute delay slot */
+#define EXEC_NEXT_INSTR	0
+#define NOP_NEXT_INSTR	1
+
+#define im11(val)	(((u32)(val)) & 0x07ff)
+
+#define hppa_ldil(addr, reg) \
+	hppa_t5_insn(0x08, reg, ((u32)(addr)) >> 11)		/* ldil im21,reg */
+#define hppa_addil(addr, reg) \
+	hppa_t5_insn(0x0a, reg, ((u32)(addr)) >> 11)		/* addil im21,reg -> result in gr1 */
+#define hppa_ldo(im14, reg, target) \
+	hppa_t1_insn(0x0d, reg, target, im14)			/* ldo val14(reg),target */
+#define hppa_ldi(im14, reg) \
+	hppa_ldo(im14, HPPA_REG_ZERO, reg)			/* ldi val14,reg */
+#define hppa_or(reg1, reg2, target) \
+	hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x09, target)	/* or reg1,reg2,target */
+#define hppa_or_cond(reg1, reg2, cond, f, target) \
+	hppa_t6_insn(0x02, reg2, reg1, cond, f, 0x09, target)
+#define hppa_and(reg1, reg2, target) \
+	hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x08, target)	/* and reg1,reg2,target */
+#define hppa_and_cond(reg1, reg2, cond, f, target) \
+	hppa_t6_insn(0x02, reg2, reg1, cond, f, 0x08, target)
+#define hppa_xor(reg1, reg2, target) \
+	hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x0a, target)	/* xor reg1,reg2,target */
+#define hppa_add(reg1, reg2, target) \
+	hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x18, target)	/* add reg1,reg2,target */
+#define hppa_addc(reg1, reg2, target) \
+	hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x1c, target)	/* add,c reg1,reg2,target */
+#define hppa_sub(reg1, reg2, target) \
+	hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x10, target)	/* sub reg1,reg2,target */
+#define hppa_subb(reg1, reg2, target) \
+	hppa_t6_insn(0x02, reg2, reg1, 0, 0, 0x14, target)	/* sub,b reg1,reg2,target */
+#define hppa_nop() \
+	hppa_or(0,0,0)						/* nop: or 0,0,0 */
+#define hppa_addi(val11, reg, target) \
+	hppa_t7_insn(0x2d, reg, target, val11)			/* addi im11,reg,target */
+#define hppa_subi(val11, reg, target) \
+	hppa_t7_insn(0x25, reg, target, val11)			/* subi im11,reg,target */
+#define hppa_copy(reg, target) \
+	hppa_or(reg, HPPA_REG_ZERO, target)			/* copy reg,target */
+#define hppa_ldw(val14, reg, target) \
+	hppa_t1_insn(0x12, reg, target, val14)			/* ldw im14(reg),target */
+#define hppa_ldb(val14, reg, target) \
+	hppa_t1_insn(0x10, reg, target, val14)			/* ldb im14(reg),target */
+#define hppa_ldh(val14, reg, target) \
+	hppa_t1_insn(0x11, reg, target, val14)			/* ldh im14(reg),target */
+#define hppa_stw(reg, val14, base) \
+	hppa_t1_insn(0x1a, base, reg, val14)			/* stw reg,im14(base) */
+#define hppa_stb(reg, val14, base) \
+	hppa_t1_insn(0x18, base, reg, val14)			/* stb reg,im14(base) */
+#define hppa_sth(reg, val14, base) \
+	hppa_t1_insn(0x19, base, reg, val14)			/* sth reg,im14(base) */
+#define hppa_stwma(reg, val14, base) \
+	hppa_t1_insn(0x1b, base, reg, val14)			/* stw,ma reg,im14(base) */
+#define hppa_bv(reg, base, nop) \
+	hppa_t11_insn(0x3a, base, reg, 0x06, 0, nop)		/* bv(,n) reg(base) */
+#define hppa_be(offset, base) \
+	hppa_t12_insn(0x38, base, offset, 0x00, 1)		/* be,n offset(0,base) */
+#define hppa_be_l(offset, base, nop) \
+	hppa_t12_insn(0x39, base, offset, 0x00, nop)		/* ble(,nop) offset(0,base) */
+#define hppa_mtctl(reg, cr) \
+	hppa_t21_insn(0x00, cr, reg, 0xc2, 0)			/* mtctl reg,cr */
+#define hppa_mtsar(reg) \
+	hppa_mtctl(reg, 11)					/* mtsar reg */
+#define hppa_zdep(r, p, len, target) \
+	hppa_t10_insn(0x35, target, r, 0, 2, p, len)		/* zdep r,a,b,t */
+#define hppa_shl(r, len, target) \
+	hppa_zdep(r, len, len, lo(rd))
+#define hppa_depwz(r, p, len, target) \
+	hppa_t10_insn(0x35, target, r, 0, 3, 31-(p), 32-(len))	/* depw,z r,p,len,ret1 */
+#define hppa_depwz_sar(reg, target) \
+	hppa_t1_insn(0x35, target, reg, 0)			/* depw,z reg,sar,32,target */
+#define hppa_shrpw_sar(reg, target) \
+	hppa_t10_insn(0x34, reg, 0, 0, 0, 0, target)		/* shrpw r0,reg,sar,target */
+#define hppa_shrpw(r1, r2, p, target) \
+	hppa_t10_insn(0x34, r2, r1, 0, 2, 31-(p), target)	/* shrpw r1,r2,p,target */
+#define hppa_shd(r1, r2, p, target) \
+	hppa_t10_insn(0x34, r2, r1, 0, 2, 31-(p), target)	/* shrpw r1,r2,p,tarfer */
+#define hppa_extrws_sar(reg, target) \
+	hppa_t10_insn(0x34, reg, target, 0, 5, 0, 0)		/* extrw,s reg,sar,32,ret0 */
+#define hppa_extrws(reg, p, len, target) \
+	hppa_t10_insn(0x34, reg, target, 0, 7, p, len)		/* extrw,s reg,p,len,target */
+#define hppa_extru(r, p, len, target) \
+	hppa_t10_insn(0x34, r, target, 0, 6, p, 32-(len))
+#define hppa_shr(r, len, target) \
+	hppa_extru(r, 31-(len), 32-(len), target)
+#define hppa_bl(imm17, rp) \
+	hppa_t12_insn(0x3a, rp, imm17, 0x00, 1)			/* bl,n target_addr,rp */
+#define hppa_sh2add(r1, r2, target) \
+	hppa_t6_insn(0x02, r2, r1, 0, 0, 0x1a, target)		/* sh2add r1,r2,target */
+
+#define hppa_combt(r1, r2, target_addr, condition, nop) \
+	hppa_t11_insn(IS_ENABLED(CONFIG_64BIT) ? 0x27 : 0x20, \
+		r2, r1, condition, target_addr, nop)		/* combt,cond,n r1,r2,addr */
+#define hppa_beq(r1, r2, target_addr) \
+	hppa_combt(r1, r2, target_addr, 1, NOP_NEXT_INSTR)
+#define hppa_blt(r1, r2, target_addr) \
+	hppa_combt(r1, r2, target_addr, 2, NOP_NEXT_INSTR)
+#define hppa_ble(r1, r2, target_addr) \
+	hppa_combt(r1, r2, target_addr, 3, NOP_NEXT_INSTR)
+#define hppa_bltu(r1, r2, target_addr) \
+	hppa_combt(r1, r2, target_addr, 4, NOP_NEXT_INSTR)
+#define hppa_bleu(r1, r2, target_addr) \
+	hppa_combt(r1, r2, target_addr, 5, NOP_NEXT_INSTR)
+
+#define hppa_combf(r1, r2, target_addr, condition, nop) \
+	hppa_t11_insn(IS_ENABLED(CONFIG_64BIT) ? 0x2f : 0x22, \
+		r2, r1, condition, target_addr, nop)		/* combf,cond,n r1,r2,addr */
+#define hppa_bne(r1, r2, target_addr) \
+	hppa_combf(r1, r2, target_addr, 1, NOP_NEXT_INSTR)
+#define hppa_bge(r1, r2, target_addr) \
+	hppa_combf(r1, r2, target_addr, 2, NOP_NEXT_INSTR)
+#define hppa_bgt(r1, r2, target_addr) \
+	hppa_combf(r1, r2, target_addr, 3, NOP_NEXT_INSTR)
+#define hppa_bgeu(r1, r2, target_addr) \
+	hppa_combf(r1, r2, target_addr, 4, NOP_NEXT_INSTR)
+#define hppa_bgtu(r1, r2, target_addr) \
+	hppa_combf(r1, r2, target_addr, 5, NOP_NEXT_INSTR)
+
+/* 64-bit instructions */
+#ifdef CONFIG_64BIT
+#define hppa64_ldd_reg(reg, b, target) \
+	hppa_t10_insn(0x03, b, reg, 0, 0, 3<<1, target)
+#define hppa64_ldd_im5(im5, b, target) \
+	hppa_t10_insn(0x03, b, low_sign_unext(im5,5), 0, 1<<2, 3<<1, target)
+#define hppa64_ldd_im16(im16, b, target) \
+	hppa_t10_insn(0x14, b, target, 0, 0, 0, 0) | re_assemble_16(im16)
+#define hppa64_std_im5(src, im5, b) \
+	hppa_t10_insn(0x03, b, src, 0, 1<<2, 0xB<<1, low_sign_unext(im5,5))
+#define hppa64_std_im16(src, im16, b) \
+	hppa_t10_insn(0x1c, b, src, 0, 0, 0, 0) | re_assemble_16(im16)
+#define hppa64_bl_long(offs22) \
+	hppa_t12_L_insn(0x3a, offs22, 1)
+#define hppa64_mtsarcm(reg) \
+	hppa_t21_insn(0x00, 11, reg, 0xc6, 0)
+#define hppa64_shrpd_sar(reg, target) \
+	hppa_t10_insn(0x34, reg, 0, 0, 0, 1<<4, target)
+#define hppa64_shladd(r1, sa, r2, target) \
+	hppa_t6_insn(0x02, r2, r1, 0, 0, 1<<4|1<<3|sa, target)
+#define hppa64_depdz_sar(reg, target) \
+	hppa_t21_insn(0x35, target, reg, 3<<3, 0)
+#define hppa_extrd_sar(reg, target, se) \
+	hppa_t10_insn(0x34, reg, target, 0, 0, 0, 0) | 2<<11 | (se&1)<<10 | 1<<9 | 1<<8
+#define hppa64_bve_l_rp(base) \
+	(0x3a << 26) | (base << 21) | 0xf000
+#define hppa64_permh_3210(r, target) \
+	(0x3e << 26) | (r << 21) | (r << 16) | (target) | 0x00006900
+#define hppa64_hshl(r, sa, target) \
+	(0x3e << 26) | (0 << 21) | (r << 16) | (sa << 6) | (target) | 0x00008800
+#define hppa64_hshr_u(r, sa, target) \
+	(0x3e << 26) | (r << 21) | (0 << 16) | (sa << 6) | (target) | 0x0000c800
+#endif
+
+struct hppa_jit_data {
+	struct bpf_binary_header *header;
+	u8 *image;
+	struct hppa_jit_context ctx;
+};
+
+static inline void bpf_fill_ill_insns(void *area, unsigned int size)
+{
+	memset(area, 0, size);
+}
+
+static inline void bpf_flush_icache(void *start, void *end)
+{
+	flush_icache_range((unsigned long)start, (unsigned long)end);
+}
+
+/* Emit a 4-byte HPPA instruction. */
+static inline void emit(const u32 insn, struct hppa_jit_context *ctx)
+{
+	if (ctx->insns) {
+		ctx->insns[ctx->ninsns] = insn;
+	}
+
+	ctx->ninsns++;
+}
+
+static inline int epilogue_offset(struct hppa_jit_context *ctx)
+{
+	int to = ctx->epilogue_offset, from = ctx->ninsns;
+
+	return (to - from);
+}
+
+/* Return -1 or inverted cond. */
+static inline int invert_bpf_cond(u8 cond)
+{
+	switch (cond) {
+	case BPF_JEQ:
+		return BPF_JNE;
+	case BPF_JGT:
+		return BPF_JLE;
+	case BPF_JLT:
+		return BPF_JGE;
+	case BPF_JGE:
+		return BPF_JLT;
+	case BPF_JLE:
+		return BPF_JGT;
+	case BPF_JNE:
+		return BPF_JEQ;
+	case BPF_JSGT:
+		return BPF_JSLE;
+	case BPF_JSLT:
+		return BPF_JSGE;
+	case BPF_JSGE:
+		return BPF_JSLT;
+	case BPF_JSLE:
+		return BPF_JSGT;
+	}
+	return -1;
+}
+
+
+static inline signed long hppa_offset(int insn, int off, struct hppa_jit_context *ctx)
+{
+	signed long from, to;
+
+	off++; /* BPF branch is from PC+1 */
+	from = (insn > 0) ? ctx->offset[insn - 1] : 0;
+	to = (insn + off > 0) ? ctx->offset[insn + off - 1] : 0;
+	return (to - from);
+}
+
+/* does the signed value fits into a given number of bits ? */
+static inline int check_bits_int(signed long val, int bits)
+{
+	return	((val >= 0) && ((val >> bits) == 0)) ||
+		 ((val < 0) && (((~((u32)val)) >> (bits-1)) == 0));
+}
+
+/* can the signed value be used in relative code ? */
+static inline int relative_bits_ok(signed long val, int bits)
+{
+	return	((val >= 0) && (val < (1UL << (bits-1)))) || /* XXX */
+		 ((val < 0) && (((~((unsigned long)val)) >> (bits-1)) == 0)
+			    && (val & (1UL << (bits-1))));
+}
+
+/* can the signed value be used in relative branches ? */
+static inline int relative_branch_ok(signed long val, int bits)
+{
+	return	((val >= 0) && (val < (1UL << (bits-2)))) || /* XXX */
+		 ((val < 0) && (((~((unsigned long)val)) < (1UL << (bits-2))))
+			    && (val & (1UL << (bits-1))));
+}
+
+
+#define is_5b_int(val)		check_bits_int(val, 5)
+
+static inline unsigned sign_unext(unsigned x, unsigned len)
+{
+	unsigned len_ones;
+
+	len_ones = (1 << len) - 1;
+	return x & len_ones;
+}
+
+static inline unsigned low_sign_unext(unsigned x, unsigned len)
+{
+	unsigned temp;
+	unsigned sign;
+
+	sign = (x >> (len-1)) & 1;
+	temp = sign_unext (x, len-1);
+	return (temp << 1) | sign;
+}
+
+static inline unsigned re_assemble_12(unsigned as12)
+{
+	return ((  (as12 & 0x800) >> 11)
+		| ((as12 & 0x400) >> (10 - 2))
+		| ((as12 & 0x3ff) << (1 + 2)));
+}
+
+static inline unsigned re_assemble_14(unsigned as14)
+{
+	return ((  (as14 & 0x1fff) << 1)
+		| ((as14 & 0x2000) >> 13));
+}
+
+#ifdef CONFIG_64BIT
+static inline unsigned re_assemble_16(unsigned as16)
+{
+	unsigned s, t;
+
+	/* Unusual 16-bit encoding, for wide mode only.  */
+	t = (as16 << 1) & 0xffff;
+	s = (as16 & 0x8000);
+	return (t ^ s ^ (s >> 1)) | (s >> 15);
+}
+#endif
+
+static inline unsigned re_assemble_17(unsigned as17)
+{
+	return ((  (as17 & 0x10000) >> 16)
+		| ((as17 & 0x0f800) << (16 - 11))
+		| ((as17 & 0x00400) >> (10 - 2))
+		| ((as17 & 0x003ff) << (1 + 2)));
+}
+
+static inline unsigned re_assemble_21(unsigned as21)
+{
+	return ((  (as21 & 0x100000) >> 20)
+		| ((as21 & 0x0ffe00) >> 8)
+		| ((as21 & 0x000180) << 7)
+		| ((as21 & 0x00007c) << 14)
+		| ((as21 & 0x000003) << 12));
+}
+
+static inline unsigned re_assemble_22(unsigned as22)
+{
+	return ((  (as22 & 0x200000) >> 21)
+		| ((as22 & 0x1f0000) << (21 - 16))
+		| ((as22 & 0x00f800) << (16 - 11))
+		| ((as22 & 0x000400) >> (10 - 2))
+		| ((as22 & 0x0003ff) << (1 + 2)));
+}
+
+/* Various HPPA instruction formats. */
+/* see https://parisc.wiki.kernel.org/images-parisc/6/68/Pa11_acd.pdf, appendix C */
+
+static inline u32 hppa_t1_insn(u8 opcode, u8 b, u8 r, s16 im14)
+{
+	return ((opcode << 26) | (b << 21) | (r << 16) | re_assemble_14(im14));
+}
+
+static inline u32 hppa_t5_insn(u8 opcode, u8 tr, u32 val21)
+{
+	return ((opcode << 26) | (tr << 21) | re_assemble_21(val21));
+}
+
+static inline u32 hppa_t6_insn(u8 opcode, u8 r2, u8 r1, u8 c, u8 f, u8 ext6, u16 t)
+{
+	return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (c << 13) | (f << 12) |
+		(ext6 << 6) | t);
+}
+
+/* 7. Arithmetic immediate */
+static inline u32 hppa_t7_insn(u8 opcode, u8 r, u8 t, u32 im11)
+{
+	return ((opcode << 26) | (r << 21) | (t << 16) | low_sign_unext(im11, 11));
+}
+
+/* 10. Shift instructions */
+static inline u32 hppa_t10_insn(u8 opcode, u8 r2, u8 r1, u8 c, u8 ext3, u8 cp, u8 t)
+{
+	return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (c << 13) |
+		(ext3 << 10) | (cp << 5) | t);
+}
+
+/* 11. Conditional branch instructions */
+static inline u32 hppa_t11_insn(u8 opcode, u8 r2, u8 r1, u8 c, u32 w, u8 nop)
+{
+	u32 ra = re_assemble_12(w);
+	// ra = low_sign_unext(w,11) | (w & (1<<10)
+	return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (c << 13) | (nop << 1) | ra);
+}
+
+/* 12. Branch instructions */
+static inline u32 hppa_t12_insn(u8 opcode, u8 rp, u32 w, u8 ext3, u8 nop)
+{
+	return ((opcode << 26) | (rp << 21) | (ext3 << 13) | (nop << 1) | re_assemble_17(w));
+}
+
+static inline u32 hppa_t12_L_insn(u8 opcode, u32 w, u8 nop)
+{
+	return ((opcode << 26) | (0x05 << 13) | (nop << 1) | re_assemble_22(w));
+}
+
+/* 21. Move to control register */
+static inline u32 hppa_t21_insn(u8 opcode, u8 r2, u8 r1, u8 ext8, u8 t)
+{
+	return ((opcode << 26) | (r2 << 21) | (r1 << 16) | (ext8 << 5) | t);
+}
+
+/* Helper functions called by jit code on HPPA32 and HPPA64. */
+
+u64 hppa_div64(u64 div, u64 divisor);
+u64 hppa_div64_rem(u64 div, u64 divisor);
+
+/* Helper functions that emit HPPA instructions when possible. */
+
+void bpf_jit_build_prologue(struct hppa_jit_context *ctx);
+void bpf_jit_build_epilogue(struct hppa_jit_context *ctx);
+
+int bpf_jit_emit_insn(const struct bpf_insn *insn, struct hppa_jit_context *ctx,
+		      bool extra_pass);
+
+#endif /* _BPF_JIT_H */
diff --git a/arch/parisc/net/bpf_jit_comp32.c b/arch/parisc/net/bpf_jit_comp32.c
new file mode 100644
index 000000000000..5ff0cf925fe9
--- /dev/null
+++ b/arch/parisc/net/bpf_jit_comp32.c
@@ -0,0 +1,1615 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * BPF JIT compiler for PA-RISC (32-bit)
+ *
+ * Copyright (c) 2023 Helge Deller <deller@gmx.de>
+ *
+ * The code is based on the BPF JIT compiler for RV64 by Björn Töpel and
+ * the BPF JIT compiler for 32-bit ARM by Shubham Bansal and Mircea Gherzan.
+ */
+
+#include <linux/bpf.h>
+#include <linux/filter.h>
+#include <linux/libgcc.h>
+#include "bpf_jit.h"
+
+/*
+ * Stack layout during BPF program execution (note: stack grows up):
+ *
+ *                     high
+ *   HPPA32 sp =>  +----------+ <= HPPA32 fp
+ *                 | saved sp |
+ *                 | saved rp |
+ *                 |   ...    | HPPA32 callee-saved registers
+ *                 | curr args|
+ *                 | local var|
+ *                 +----------+ <= (sp - 4 * NR_SAVED_REGISTERS)
+ *                 |  lo(R9)  |
+ *                 |  hi(R9)  |
+ *                 |  lo(FP)  | JIT scratch space for BPF registers
+ *                 |  hi(FP)  |
+ *                 |   ...    |
+ *                 +----------+ <= (sp - 4 * NR_SAVED_REGISTERS
+ *                 |          |        - 4 * BPF_JIT_SCRATCH_REGS)
+ *                 |          |
+ *                 |   ...    | BPF program stack
+ *                 |          |
+ *                 |   ...    | Function call stack
+ *                 |          |
+ *                 +----------+
+ *                     low
+ */
+
+enum {
+	/* Stack layout - these are offsets from top of JIT scratch space. */
+	BPF_R8_HI,
+	BPF_R8_LO,
+	BPF_R9_HI,
+	BPF_R9_LO,
+	BPF_FP_HI,
+	BPF_FP_LO,
+	BPF_AX_HI,
+	BPF_AX_LO,
+	BPF_R0_TEMP_HI,
+	BPF_R0_TEMP_LO,
+	BPF_JIT_SCRATCH_REGS,
+};
+
+/* Number of callee-saved registers stored to stack: rp, r3-r18. */
+#define NR_SAVED_REGISTERS	(18 - 3 + 1 + 8)
+
+/* Offset from fp for BPF registers stored on stack. */
+#define STACK_OFFSET(k)	(- (NR_SAVED_REGISTERS + k + 1))
+#define STACK_ALIGN	FRAME_SIZE
+
+#define EXIT_PTR_LOAD(reg)	hppa_ldw(-0x08, HPPA_REG_SP, reg)
+#define EXIT_PTR_STORE(reg)	hppa_stw(reg, -0x08, HPPA_REG_SP)
+#define EXIT_PTR_JUMP(reg, nop)	hppa_bv(HPPA_REG_ZERO, reg, nop)
+
+#define TMP_REG_1	(MAX_BPF_JIT_REG + 0)
+#define TMP_REG_2	(MAX_BPF_JIT_REG + 1)
+#define TMP_REG_R0	(MAX_BPF_JIT_REG + 2)
+
+static const s8 regmap[][2] = {
+	/* Return value from in-kernel function, and exit value from eBPF. */
+	[BPF_REG_0] = {HPPA_REG_RET0, HPPA_REG_RET1},		/* HI/LOW */
+
+	/* Arguments from eBPF program to in-kernel function. */
+	[BPF_REG_1] = {HPPA_R(3), HPPA_R(4)},
+	[BPF_REG_2] = {HPPA_R(5), HPPA_R(6)},
+	[BPF_REG_3] = {HPPA_R(7), HPPA_R(8)},
+	[BPF_REG_4] = {HPPA_R(9), HPPA_R(10)},
+	[BPF_REG_5] = {HPPA_R(11), HPPA_R(12)},
+
+	[BPF_REG_6] = {HPPA_R(13), HPPA_R(14)},
+	[BPF_REG_7] = {HPPA_R(15), HPPA_R(16)},
+	/*
+	 * Callee-saved registers that in-kernel function will preserve.
+	 * Stored on the stack.
+	 */
+	[BPF_REG_8] = {STACK_OFFSET(BPF_R8_HI), STACK_OFFSET(BPF_R8_LO)},
+	[BPF_REG_9] = {STACK_OFFSET(BPF_R9_HI), STACK_OFFSET(BPF_R9_LO)},
+
+	/* Read-only frame pointer to access BPF stack. Not needed. */
+	[BPF_REG_FP] = {STACK_OFFSET(BPF_FP_HI), STACK_OFFSET(BPF_FP_LO)},
+
+	/* Temporary register for blinding constants. Stored on the stack. */
+	[BPF_REG_AX] = {STACK_OFFSET(BPF_AX_HI), STACK_OFFSET(BPF_AX_LO)},
+	/*
+	 * Temporary registers used by the JIT to operate on registers stored
+	 * on the stack. Save t0 and t1 to be used as temporaries in generated
+	 * code.
+	 */
+	[TMP_REG_1] = {HPPA_REG_T3, HPPA_REG_T2},
+	[TMP_REG_2] = {HPPA_REG_T5, HPPA_REG_T4},
+
+	/* temporary space for BPF_R0 during libgcc and millicode calls */
+	[TMP_REG_R0] = {STACK_OFFSET(BPF_R0_TEMP_HI), STACK_OFFSET(BPF_R0_TEMP_LO)},
+};
+
+static s8 hi(const s8 *r)
+{
+	return r[0];
+}
+
+static s8 lo(const s8 *r)
+{
+	return r[1];
+}
+
+static void emit_hppa_copy(const s8 rs, const s8 rd, struct hppa_jit_context *ctx)
+{
+	REG_SET_SEEN(ctx, rd);
+	if (OPTIMIZE_HPPA && (rs == rd))
+		return;
+	REG_SET_SEEN(ctx, rs);
+	emit(hppa_copy(rs, rd), ctx);
+}
+
+static void emit_hppa_xor(const s8 r1, const s8 r2, const s8 r3, struct hppa_jit_context *ctx)
+{
+	REG_SET_SEEN(ctx, r1);
+	REG_SET_SEEN(ctx, r2);
+	REG_SET_SEEN(ctx, r3);
+	if (OPTIMIZE_HPPA && (r1 == r2)) {
+		emit(hppa_copy(HPPA_REG_ZERO, r3), ctx);
+	} else {
+		emit(hppa_xor(r1, r2, r3), ctx);
+	}
+}
+
+static void emit_imm(const s8 rd, s32 imm, struct hppa_jit_context *ctx)
+{
+	u32 lower = im11(imm);
+
+	REG_SET_SEEN(ctx, rd);
+	if (OPTIMIZE_HPPA && relative_bits_ok(imm, 14)) {
+		emit(hppa_ldi(imm, rd), ctx);
+		return;
+	}
+	emit(hppa_ldil(imm, rd), ctx);
+	if (OPTIMIZE_HPPA && (lower == 0))
+		return;
+	emit(hppa_ldo(lower, rd, rd), ctx);
+}
+
+static void emit_imm32(const s8 *rd, s32 imm, struct hppa_jit_context *ctx)
+{
+	/* Emit immediate into lower bits. */
+	REG_SET_SEEN(ctx, lo(rd));
+	emit_imm(lo(rd), imm, ctx);
+
+	/* Sign-extend into upper bits. */
+	REG_SET_SEEN(ctx, hi(rd));
+	if (imm >= 0)
+		emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+	else
+		emit(hppa_ldi(-1, hi(rd)), ctx);
+}
+
+static void emit_imm64(const s8 *rd, s32 imm_hi, s32 imm_lo,
+		       struct hppa_jit_context *ctx)
+{
+	emit_imm(hi(rd), imm_hi, ctx);
+	emit_imm(lo(rd), imm_lo, ctx);
+}
+
+static void __build_epilogue(bool is_tail_call, struct hppa_jit_context *ctx)
+{
+	const s8 *r0 = regmap[BPF_REG_0];
+	int i;
+
+	if (is_tail_call) {
+		/*
+		 * goto *(t0 + 4);
+		 * Skips first instruction of prologue which initializes tail
+		 * call counter. Assumes t0 contains address of target program,
+		 * see emit_bpf_tail_call.
+		 */
+		emit(hppa_ldo(1 * HPPA_INSN_SIZE, HPPA_REG_T0, HPPA_REG_T0), ctx);
+		emit(hppa_bv(HPPA_REG_ZERO, HPPA_REG_T0, EXEC_NEXT_INSTR), ctx);
+		/* in delay slot: */
+		emit(hppa_copy(HPPA_REG_TCC, HPPA_REG_TCC_IN_INIT), ctx);
+
+		return;
+	}
+
+	/* load epilogue function pointer and jump to it. */
+	/* exit point is either directly below, or the outest TCC exit function */
+	emit(EXIT_PTR_LOAD(HPPA_REG_RP), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/* NOTE: we are 32-bit and big-endian, so return lower 32-bit value */
+	emit_hppa_copy(lo(r0), HPPA_REG_RET0, ctx);
+
+	/* Restore callee-saved registers. */
+	for (i = 3; i <= 18; i++) {
+		if (OPTIMIZE_HPPA && !REG_WAS_SEEN(ctx, HPPA_R(i)))
+			continue;
+		emit(hppa_ldw(-REG_SIZE * (8 + (i-3)), HPPA_REG_SP, HPPA_R(i)), ctx);
+	}
+
+	/* load original return pointer (stored by outest TCC function) */
+	emit(hppa_ldw(-0x14, HPPA_REG_SP, HPPA_REG_RP), ctx);
+	emit(hppa_bv(HPPA_REG_ZERO, HPPA_REG_RP, EXEC_NEXT_INSTR), ctx);
+	/* in delay slot: */
+	emit(hppa_ldw(-0x04, HPPA_REG_SP, HPPA_REG_SP), ctx);
+}
+
+static bool is_stacked(s8 reg)
+{
+	return reg < 0;
+}
+
+static const s8 *bpf_get_reg64_offset(const s8 *reg, const s8 *tmp,
+		u16 offset_sp, struct hppa_jit_context *ctx)
+{
+	if (is_stacked(hi(reg))) {
+		emit(hppa_ldw(REG_SIZE * hi(reg) - offset_sp, HPPA_REG_SP, hi(tmp)), ctx);
+		emit(hppa_ldw(REG_SIZE * lo(reg) - offset_sp, HPPA_REG_SP, lo(tmp)), ctx);
+		reg = tmp;
+	}
+	REG_SET_SEEN(ctx, hi(reg));
+	REG_SET_SEEN(ctx, lo(reg));
+	return reg;
+}
+
+static const s8 *bpf_get_reg64(const s8 *reg, const s8 *tmp,
+			       struct hppa_jit_context *ctx)
+{
+	return bpf_get_reg64_offset(reg, tmp, 0, ctx);
+}
+
+static const s8 *bpf_get_reg64_ref(const s8 *reg, const s8 *tmp,
+		bool must_load, struct hppa_jit_context *ctx)
+{
+	if (!OPTIMIZE_HPPA)
+		return bpf_get_reg64(reg, tmp, ctx);
+
+	if (is_stacked(hi(reg))) {
+		if (must_load)
+			emit(hppa_ldw(REG_SIZE * hi(reg), HPPA_REG_SP, hi(tmp)), ctx);
+		reg = tmp;
+	}
+	REG_SET_SEEN(ctx, hi(reg));
+	REG_SET_SEEN(ctx, lo(reg));
+	return reg;
+}
+
+
+static void bpf_put_reg64(const s8 *reg, const s8 *src,
+			  struct hppa_jit_context *ctx)
+{
+	if (is_stacked(hi(reg))) {
+		emit(hppa_stw(hi(src), REG_SIZE * hi(reg), HPPA_REG_SP), ctx);
+		emit(hppa_stw(lo(src), REG_SIZE * lo(reg), HPPA_REG_SP), ctx);
+	}
+}
+
+static void bpf_save_R0(struct hppa_jit_context *ctx)
+{
+	bpf_put_reg64(regmap[TMP_REG_R0], regmap[BPF_REG_0], ctx);
+}
+
+static void bpf_restore_R0(struct hppa_jit_context *ctx)
+{
+	bpf_get_reg64(regmap[TMP_REG_R0], regmap[BPF_REG_0], ctx);
+}
+
+
+static const s8 *bpf_get_reg32(const s8 *reg, const s8 *tmp,
+			       struct hppa_jit_context *ctx)
+{
+	if (is_stacked(lo(reg))) {
+		emit(hppa_ldw(REG_SIZE * lo(reg), HPPA_REG_SP, lo(tmp)), ctx);
+		reg = tmp;
+	}
+	REG_SET_SEEN(ctx, lo(reg));
+	return reg;
+}
+
+static const s8 *bpf_get_reg32_ref(const s8 *reg, const s8 *tmp,
+		struct hppa_jit_context *ctx)
+{
+	if (!OPTIMIZE_HPPA)
+		return bpf_get_reg32(reg, tmp, ctx);
+
+	if (is_stacked(hi(reg))) {
+		reg = tmp;
+	}
+	REG_SET_SEEN(ctx, lo(reg));
+	return reg;
+}
+
+static void bpf_put_reg32(const s8 *reg, const s8 *src,
+			  struct hppa_jit_context *ctx)
+{
+	if (is_stacked(lo(reg))) {
+		REG_SET_SEEN(ctx, lo(src));
+		emit(hppa_stw(lo(src), REG_SIZE * lo(reg), HPPA_REG_SP), ctx);
+		if (1 && !ctx->prog->aux->verifier_zext) {
+			REG_SET_SEEN(ctx, hi(reg));
+			emit(hppa_stw(HPPA_REG_ZERO, REG_SIZE * hi(reg), HPPA_REG_SP), ctx);
+		}
+	} else if (1 && !ctx->prog->aux->verifier_zext) {
+		REG_SET_SEEN(ctx, hi(reg));
+		emit_hppa_copy(HPPA_REG_ZERO, hi(reg), ctx);
+	}
+}
+
+/* extern hppa millicode functions */
+extern void $$mulI(void);
+extern void $$divU(void);
+extern void $$remU(void);
+
+static void emit_call_millicode(void *func, const s8 arg0,
+		const s8 arg1, u8 opcode, struct hppa_jit_context *ctx)
+{
+	u32 func_addr;
+
+	emit_hppa_copy(arg0, HPPA_REG_ARG0, ctx);
+	emit_hppa_copy(arg1, HPPA_REG_ARG1, ctx);
+
+	/* libcgcc overwrites HPPA_REG_RET0/1, save temp. in dest. */
+	if (arg0 != HPPA_REG_RET1)
+		bpf_save_R0(ctx);
+
+	func_addr = (uintptr_t) dereference_function_descriptor(func);
+	emit(hppa_ldil(func_addr, HPPA_REG_R31), ctx);
+	/* skip the following be_l instruction if divisor is zero. */
+	if (BPF_OP(opcode) == BPF_DIV || BPF_OP(opcode) == BPF_MOD) {
+		if (BPF_OP(opcode) == BPF_DIV)
+			emit_hppa_copy(HPPA_REG_ZERO, HPPA_REG_RET1, ctx);
+		else
+			emit_hppa_copy(HPPA_REG_ARG0, HPPA_REG_RET1, ctx);
+		emit(hppa_or_cond(HPPA_REG_ARG1, HPPA_REG_ZERO, 1, 0, HPPA_REG_ZERO), ctx);
+	}
+	/* Note: millicode functions use r31 as return pointer instead of rp */
+	emit(hppa_be_l(im11(func_addr) >> 2, HPPA_REG_R31, NOP_NEXT_INSTR), ctx);
+	emit(hppa_nop(), ctx); /* this nop is needed here for delay slot */
+
+	/* Note: millicode functions return result in RET1, not RET0 */
+	emit_hppa_copy(HPPA_REG_RET1, arg0, ctx);
+
+	/* restore HPPA_REG_RET0/1, temp. save in dest. */
+	if (arg0 != HPPA_REG_RET1)
+		bpf_restore_R0(ctx);
+}
+
+static void emit_call_libgcc_ll(void *func, const s8 *arg0,
+		const s8 *arg1, u8 opcode, struct hppa_jit_context *ctx)
+{
+	u32 func_addr;
+
+	emit_hppa_copy(lo(arg0), HPPA_REG_ARG0, ctx);
+	emit_hppa_copy(hi(arg0), HPPA_REG_ARG1, ctx);
+	emit_hppa_copy(lo(arg1), HPPA_REG_ARG2, ctx);
+	emit_hppa_copy(hi(arg1), HPPA_REG_ARG3, ctx);
+
+	/* libcgcc overwrites HPPA_REG_RET0/_RET1, so keep copy of R0 on stack */
+	if (hi(arg0) != HPPA_REG_RET0)
+		bpf_save_R0(ctx);
+
+	/* prepare stack */
+	emit(hppa_ldo(2 * FRAME_SIZE, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	func_addr = (uintptr_t) dereference_function_descriptor(func);
+	emit(hppa_ldil(func_addr, HPPA_REG_R31), ctx);
+        /* zero out the following be_l instruction if divisor is 0 (and set default values) */
+	if (BPF_OP(opcode) == BPF_DIV || BPF_OP(opcode) == BPF_MOD) {
+		emit_hppa_copy(HPPA_REG_ZERO, HPPA_REG_RET0, ctx);
+		if (BPF_OP(opcode) == BPF_DIV)
+			emit_hppa_copy(HPPA_REG_ZERO, HPPA_REG_RET1, ctx);
+		else
+			emit_hppa_copy(HPPA_REG_ARG0, HPPA_REG_RET1, ctx);
+		emit(hppa_or_cond(HPPA_REG_ARG2, HPPA_REG_ARG3, 1, 0, HPPA_REG_ZERO), ctx);
+	}
+	emit(hppa_be_l(im11(func_addr) >> 2, HPPA_REG_R31, EXEC_NEXT_INSTR), ctx);
+	emit_hppa_copy(HPPA_REG_R31, HPPA_REG_RP, ctx);
+
+	/* restore stack */
+	emit(hppa_ldo(-2 * FRAME_SIZE, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	emit_hppa_copy(HPPA_REG_RET0, hi(arg0), ctx);
+	emit_hppa_copy(HPPA_REG_RET1, lo(arg0), ctx);
+
+	/* restore HPPA_REG_RET0/_RET1 */
+	if (hi(arg0) != HPPA_REG_RET0)
+		bpf_restore_R0(ctx);
+}
+
+static void emit_jump(s32 paoff, bool force_far,
+			       struct hppa_jit_context *ctx)
+{
+	unsigned long pc, addr;
+
+	/* Note: allocate 2 instructions for jumps if force_far is set. */
+	if (relative_bits_ok(paoff - HPPA_BRANCH_DISPLACEMENT, 17)) {
+		/* use BL,short branch followed by nop() */
+		emit(hppa_bl(paoff - HPPA_BRANCH_DISPLACEMENT, HPPA_REG_ZERO), ctx);
+		if (force_far)
+			emit(hppa_nop(), ctx);
+		return;
+	}
+
+	pc = (uintptr_t) &ctx->insns[ctx->ninsns];
+	addr = pc + (paoff * HPPA_INSN_SIZE);
+	emit(hppa_ldil(addr, HPPA_REG_R31), ctx);
+	emit(hppa_be_l(im11(addr) >> 2, HPPA_REG_R31, NOP_NEXT_INSTR), ctx); // be,l,n addr(sr4,r31), %sr0, %r31
+}
+
+static void emit_alu_i64(const s8 *dst, s32 imm,
+			 struct hppa_jit_context *ctx, const u8 op)
+{
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *rd;
+
+	if (0 && op == BPF_MOV)
+		rd = bpf_get_reg64_ref(dst, tmp1, false, ctx);
+	else
+		rd = bpf_get_reg64(dst, tmp1, ctx);
+
+	/* dst = dst OP imm */
+	switch (op) {
+	case BPF_MOV:
+		emit_imm32(rd, imm, ctx);
+		break;
+	case BPF_AND:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit(hppa_and(lo(rd), HPPA_REG_T0, lo(rd)), ctx);
+		if (imm >= 0)
+			emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+		break;
+	case BPF_OR:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit(hppa_or(lo(rd), HPPA_REG_T0, lo(rd)), ctx);
+		if (imm < 0)
+			emit_imm(hi(rd), -1, ctx);
+		break;
+	case BPF_XOR:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit_hppa_xor(lo(rd), HPPA_REG_T0, lo(rd), ctx);
+		if (imm < 0) {
+			emit_imm(HPPA_REG_T0, -1, ctx);
+			emit_hppa_xor(hi(rd), HPPA_REG_T0, hi(rd), ctx);
+		}
+		break;
+	case BPF_LSH:
+		if (imm == 0)
+			break;
+		if (imm > 32) {
+			imm -= 32;
+			emit(hppa_zdep(lo(rd), imm, imm, hi(rd)), ctx);
+			emit_hppa_copy(HPPA_REG_ZERO, lo(rd), ctx);
+		} else if (imm == 32) {
+			emit_hppa_copy(lo(rd), hi(rd), ctx);
+			emit_hppa_copy(HPPA_REG_ZERO, lo(rd), ctx);
+		} else {
+			emit(hppa_shd(hi(rd), lo(rd), 32 - imm, hi(rd)), ctx);
+			emit(hppa_zdep(lo(rd), imm, imm, lo(rd)), ctx);
+		}
+		break;
+	case BPF_RSH:
+		if (imm == 0)
+			break;
+		if (imm > 32) {
+			imm -= 32;
+			emit(hppa_shr(hi(rd), imm, lo(rd)), ctx);
+			emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+		} else if (imm == 32) {
+			emit_hppa_copy(hi(rd), lo(rd), ctx);
+			emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+		} else {
+			emit(hppa_shrpw(hi(rd), lo(rd), imm, lo(rd)), ctx);
+			emit(hppa_shr(hi(rd), imm, hi(rd)), ctx);
+		}
+		break;
+	case BPF_ARSH:
+		if (imm == 0)
+			break;
+		if (imm > 32) {
+			imm -= 32;
+			emit(hppa_extrws(hi(rd), 31 - imm, imm, lo(rd)), ctx);
+			emit(hppa_extrws(hi(rd), 0, 31, hi(rd)), ctx);
+		} else if (imm == 32) {
+			emit_hppa_copy(hi(rd), lo(rd), ctx);
+			emit(hppa_extrws(hi(rd), 0, 31, hi(rd)), ctx);
+		} else {
+			emit(hppa_shrpw(hi(rd), lo(rd), imm, lo(rd)), ctx);
+			emit(hppa_extrws(hi(rd), 31 - imm, imm, hi(rd)), ctx);
+		}
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	bpf_put_reg64(dst, rd, ctx);
+}
+
+static void emit_alu_i32(const s8 *dst, s32 imm,
+			 struct hppa_jit_context *ctx, const u8 op)
+{
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *rd = bpf_get_reg32(dst, tmp1, ctx);
+
+	if (op == BPF_MOV)
+		rd = bpf_get_reg32_ref(dst, tmp1, ctx);
+	else
+		rd = bpf_get_reg32(dst, tmp1, ctx);
+
+	/* dst = dst OP imm */
+	switch (op) {
+	case BPF_MOV:
+		emit_imm(lo(rd), imm, ctx);
+		break;
+	case BPF_ADD:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit(hppa_add(lo(rd), HPPA_REG_T0, lo(rd)), ctx);
+		break;
+	case BPF_SUB:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit(hppa_sub(lo(rd), HPPA_REG_T0, lo(rd)), ctx);
+		break;
+	case BPF_AND:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit(hppa_and(lo(rd), HPPA_REG_T0, lo(rd)), ctx);
+		break;
+	case BPF_OR:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit(hppa_or(lo(rd), HPPA_REG_T0, lo(rd)), ctx);
+		break;
+	case BPF_XOR:
+		emit_imm(HPPA_REG_T0, imm, ctx);
+		emit_hppa_xor(lo(rd), HPPA_REG_T0, lo(rd), ctx);
+		break;
+	case BPF_LSH:
+		if (imm != 0)
+			emit(hppa_zdep(lo(rd), imm, imm, lo(rd)), ctx);
+		break;
+	case BPF_RSH:
+		if (imm != 0)
+			emit(hppa_shr(lo(rd), imm, lo(rd)), ctx);
+		break;
+	case BPF_ARSH:
+		if (imm != 0)
+			emit(hppa_extrws(lo(rd), 31 - imm, imm, lo(rd)), ctx);
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	bpf_put_reg32(dst, rd, ctx);
+}
+
+static void emit_alu_r64(const s8 *dst, const s8 *src,
+			 struct hppa_jit_context *ctx, const u8 op)
+{
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *tmp2 = regmap[TMP_REG_2];
+	const s8 *rd;
+	const s8 *rs = bpf_get_reg64(src, tmp2, ctx);
+
+	if (op == BPF_MOV)
+		rd = bpf_get_reg64_ref(dst, tmp1, false, ctx);
+	else
+		rd = bpf_get_reg64(dst, tmp1, ctx);
+
+	/* dst = dst OP src */
+	switch (op) {
+	case BPF_MOV:
+		emit_hppa_copy(lo(rs), lo(rd), ctx);
+		emit_hppa_copy(hi(rs), hi(rd), ctx);
+		break;
+	case BPF_ADD:
+		emit(hppa_add(lo(rd), lo(rs), lo(rd)), ctx);
+		emit(hppa_addc(hi(rd), hi(rs), hi(rd)), ctx);
+		break;
+	case BPF_SUB:
+		emit(hppa_sub(lo(rd), lo(rs), lo(rd)), ctx);
+		emit(hppa_subb(hi(rd), hi(rs), hi(rd)), ctx);
+		break;
+	case BPF_AND:
+		emit(hppa_and(lo(rd), lo(rs), lo(rd)), ctx);
+		emit(hppa_and(hi(rd), hi(rs), hi(rd)), ctx);
+		break;
+	case BPF_OR:
+		emit(hppa_or(lo(rd), lo(rs), lo(rd)), ctx);
+		emit(hppa_or(hi(rd), hi(rs), hi(rd)), ctx);
+		break;
+	case BPF_XOR:
+		emit_hppa_xor(lo(rd), lo(rs), lo(rd), ctx);
+		emit_hppa_xor(hi(rd), hi(rs), hi(rd), ctx);
+		break;
+	case BPF_MUL:
+		emit_call_libgcc_ll(__muldi3, rd, rs, op, ctx);
+		break;
+	case BPF_DIV:
+		emit_call_libgcc_ll(&hppa_div64, rd, rs, op, ctx);
+		break;
+	case BPF_MOD:
+		emit_call_libgcc_ll(&hppa_div64_rem, rd, rs, op, ctx);
+		break;
+	case BPF_LSH:
+		emit_call_libgcc_ll(__ashldi3, rd, rs, op, ctx);
+		break;
+	case BPF_RSH:
+		emit_call_libgcc_ll(__lshrdi3, rd, rs, op, ctx);
+		break;
+	case BPF_ARSH:
+		emit_call_libgcc_ll(__ashrdi3, rd, rs, op, ctx);
+		break;
+	case BPF_NEG:
+		emit(hppa_sub(HPPA_REG_ZERO, lo(rd), lo(rd)), ctx);
+		emit(hppa_subb(HPPA_REG_ZERO, hi(rd), hi(rd)), ctx);
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	bpf_put_reg64(dst, rd, ctx);
+}
+
+static void emit_alu_r32(const s8 *dst, const s8 *src,
+			 struct hppa_jit_context *ctx, const u8 op)
+{
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *tmp2 = regmap[TMP_REG_2];
+	const s8 *rd;
+	const s8 *rs = bpf_get_reg32(src, tmp2, ctx);
+
+	if (op == BPF_MOV)
+		rd = bpf_get_reg32_ref(dst, tmp1, ctx);
+	else
+		rd = bpf_get_reg32(dst, tmp1, ctx);
+
+	/* dst = dst OP src */
+	switch (op) {
+	case BPF_MOV:
+		emit_hppa_copy(lo(rs), lo(rd), ctx);
+		break;
+	case BPF_ADD:
+		emit(hppa_add(lo(rd), lo(rs), lo(rd)), ctx);
+		break;
+	case BPF_SUB:
+		emit(hppa_sub(lo(rd), lo(rs), lo(rd)), ctx);
+		break;
+	case BPF_AND:
+		emit(hppa_and(lo(rd), lo(rs), lo(rd)), ctx);
+		break;
+	case BPF_OR:
+		emit(hppa_or(lo(rd), lo(rs), lo(rd)), ctx);
+		break;
+	case BPF_XOR:
+		emit_hppa_xor(lo(rd), lo(rs), lo(rd), ctx);
+		break;
+	case BPF_MUL:
+		emit_call_millicode($$mulI, lo(rd), lo(rs), op, ctx);
+		break;
+	case BPF_DIV:
+		emit_call_millicode($$divU, lo(rd), lo(rs), op, ctx);
+		break;
+	case BPF_MOD:
+		emit_call_millicode($$remU, lo(rd), lo(rs), op, ctx);
+		break;
+	case BPF_LSH:
+		emit(hppa_subi(0x1f, lo(rs), HPPA_REG_T0), ctx);
+		emit(hppa_mtsar(HPPA_REG_T0), ctx);
+		emit(hppa_depwz_sar(lo(rd), lo(rd)), ctx);
+		break;
+	case BPF_RSH:
+		emit(hppa_mtsar(lo(rs)), ctx);
+		emit(hppa_shrpw_sar(lo(rd), lo(rd)), ctx);
+		break;
+	case BPF_ARSH: /* sign extending arithmetic shift right */
+		// emit(hppa_beq(lo(rs), HPPA_REG_ZERO, 2), ctx);
+		emit(hppa_subi(0x1f, lo(rs), HPPA_REG_T0), ctx);
+		emit(hppa_mtsar(HPPA_REG_T0), ctx);
+		emit(hppa_extrws_sar(lo(rd), lo(rd)), ctx);
+		break;
+	case BPF_NEG:
+		emit(hppa_sub(HPPA_REG_ZERO, lo(rd), lo(rd)), ctx);  // sub r0,rd,rd
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	bpf_put_reg32(dst, rd, ctx);
+}
+
+static int emit_branch_r64(const s8 *src1, const s8 *src2, s32 paoff,
+			   struct hppa_jit_context *ctx, const u8 op)
+{
+	int e, s = ctx->ninsns;
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *tmp2 = regmap[TMP_REG_2];
+
+	const s8 *rs1 = bpf_get_reg64(src1, tmp1, ctx);
+	const s8 *rs2 = bpf_get_reg64(src2, tmp2, ctx);
+
+	/*
+	 * NO_JUMP skips over the rest of the instructions and the
+	 * emit_jump, meaning the BPF branch is not taken.
+	 * JUMP skips directly to the emit_jump, meaning
+	 * the BPF branch is taken.
+	 *
+	 * The fallthrough case results in the BPF branch being taken.
+	 */
+#define NO_JUMP(idx)	(2 + (idx) - 1)
+#define JUMP(idx)	(0 + (idx) - 1)
+
+	switch (op) {
+	case BPF_JEQ:
+		emit(hppa_bne(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bne(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JGT:
+		emit(hppa_bgtu(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_bltu(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bleu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JLT:
+		emit(hppa_bltu(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_bgtu(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bgeu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JGE:
+		emit(hppa_bgtu(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_bltu(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bltu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JLE:
+		emit(hppa_bltu(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_bgtu(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bgtu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JNE:
+		emit(hppa_bne(hi(rs1), hi(rs2), JUMP(1)), ctx);
+		emit(hppa_beq(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JSGT:
+		emit(hppa_bgt(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_blt(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bleu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JSLT:
+		emit(hppa_blt(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_bgt(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bgeu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JSGE:
+		emit(hppa_bgt(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_blt(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bltu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JSLE:
+		emit(hppa_blt(hi(rs1), hi(rs2), JUMP(2)), ctx);
+		emit(hppa_bgt(hi(rs1), hi(rs2), NO_JUMP(1)), ctx);
+		emit(hppa_bgtu(lo(rs1), lo(rs2), NO_JUMP(0)), ctx);
+		break;
+	case BPF_JSET:
+		emit(hppa_and(hi(rs1), hi(rs2), HPPA_REG_T0), ctx);
+		emit(hppa_and(lo(rs1), lo(rs2), HPPA_REG_T1), ctx);
+		emit(hppa_bne(HPPA_REG_T0, HPPA_REG_ZERO, JUMP(1)), ctx);
+		emit(hppa_beq(HPPA_REG_T1, HPPA_REG_ZERO, NO_JUMP(0)), ctx);
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+#undef NO_JUMP
+#undef JUMP
+
+	e = ctx->ninsns;
+	/* Adjust for extra insns. */
+	paoff -= (e - s);
+	emit_jump(paoff, true, ctx);
+	return 0;
+}
+
+static int emit_bcc(u8 op, u8 rd, u8 rs, int paoff, struct hppa_jit_context *ctx)
+{
+	int e, s;
+	bool far = false;
+	int off;
+
+	if (op == BPF_JSET) {
+		/*
+		 * BPF_JSET is a special case: it has no inverse so we always
+		 * treat it as a far branch.
+		 */
+		emit(hppa_and(rd, rs, HPPA_REG_T0), ctx);
+		paoff -= 1; /* reduce offset due to hppa_and() above */
+		rd = HPPA_REG_T0;
+		rs = HPPA_REG_ZERO;
+		op = BPF_JNE;
+	}
+
+	s = ctx->ninsns;
+
+	if (!relative_bits_ok(paoff - HPPA_BRANCH_DISPLACEMENT, 12)) {
+		op = invert_bpf_cond(op);
+		far = true;
+	}
+
+	/*
+	 * For a far branch, the condition is negated and we jump over the
+	 * branch itself, and the three instructions from emit_jump.
+	 * For a near branch, just use paoff.
+	 */
+	off = far ? (HPPA_BRANCH_DISPLACEMENT - 1) : paoff - HPPA_BRANCH_DISPLACEMENT;
+
+	switch (op) {
+	/* IF (dst COND src) JUMP off */
+	case BPF_JEQ:
+		emit(hppa_beq(rd, rs, off), ctx);
+		break;
+	case BPF_JGT:
+		emit(hppa_bgtu(rd, rs, off), ctx);
+		break;
+	case BPF_JLT:
+		emit(hppa_bltu(rd, rs, off), ctx);
+		break;
+	case BPF_JGE:
+		emit(hppa_bgeu(rd, rs, off), ctx);
+		break;
+	case BPF_JLE:
+		emit(hppa_bleu(rd, rs, off), ctx);
+		break;
+	case BPF_JNE:
+		emit(hppa_bne(rd, rs, off), ctx);
+		break;
+	case BPF_JSGT:
+		emit(hppa_bgt(rd, rs, off), ctx);
+		break;
+	case BPF_JSLT:
+		emit(hppa_blt(rd, rs, off), ctx);
+		break;
+	case BPF_JSGE:
+		emit(hppa_bge(rd, rs, off), ctx);
+		break;
+	case BPF_JSLE:
+		emit(hppa_ble(rd, rs, off), ctx);
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	if (far) {
+		e = ctx->ninsns;
+		/* Adjust for extra insns. */
+		paoff -= (e - s);
+		emit_jump(paoff, true, ctx);
+	}
+	return 0;
+}
+
+static int emit_branch_r32(const s8 *src1, const s8 *src2, s32 paoff,
+			   struct hppa_jit_context *ctx, const u8 op)
+{
+	int e, s = ctx->ninsns;
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *tmp2 = regmap[TMP_REG_2];
+
+	const s8 *rs1 = bpf_get_reg32(src1, tmp1, ctx);
+	const s8 *rs2 = bpf_get_reg32(src2, tmp2, ctx);
+
+	e = ctx->ninsns;
+	/* Adjust for extra insns. */
+	paoff -= (e - s);
+
+	if (emit_bcc(op, lo(rs1), lo(rs2), paoff, ctx))
+		return -1;
+
+	return 0;
+}
+
+static void emit_call(bool fixed, u64 addr, struct hppa_jit_context *ctx)
+{
+	const s8 *tmp = regmap[TMP_REG_1];
+	const s8 *r0 = regmap[BPF_REG_0];
+	const s8 *reg;
+	const int offset_sp = 2 * STACK_ALIGN;
+
+	/* prepare stack */
+	emit(hppa_ldo(offset_sp, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	/* load R1 & R2 in registers, R3-R5 to stack. */
+	reg = bpf_get_reg64_offset(regmap[BPF_REG_5], tmp, offset_sp, ctx);
+	emit(hppa_stw(hi(reg), -0x48, HPPA_REG_SP), ctx);
+	emit(hppa_stw(lo(reg), -0x44, HPPA_REG_SP), ctx);
+
+	reg = bpf_get_reg64_offset(regmap[BPF_REG_4], tmp, offset_sp, ctx);
+	emit(hppa_stw(hi(reg), -0x40, HPPA_REG_SP), ctx);
+	emit(hppa_stw(lo(reg), -0x3c, HPPA_REG_SP), ctx);
+
+	reg = bpf_get_reg64_offset(regmap[BPF_REG_3], tmp, offset_sp, ctx);
+	emit(hppa_stw(hi(reg), -0x38, HPPA_REG_SP), ctx);
+	emit(hppa_stw(lo(reg), -0x34, HPPA_REG_SP), ctx);
+
+	reg = bpf_get_reg64_offset(regmap[BPF_REG_2], tmp, offset_sp, ctx);
+	emit_hppa_copy(hi(reg), HPPA_REG_ARG3, ctx);
+	emit_hppa_copy(lo(reg), HPPA_REG_ARG2, ctx);
+
+	reg = bpf_get_reg64_offset(regmap[BPF_REG_1], tmp, offset_sp, ctx);
+	emit_hppa_copy(hi(reg), HPPA_REG_ARG1, ctx);
+	emit_hppa_copy(lo(reg), HPPA_REG_ARG0, ctx);
+
+	/* backup TCC */
+	if (REG_WAS_SEEN(ctx, HPPA_REG_TCC))
+		emit(hppa_copy(HPPA_REG_TCC, HPPA_REG_TCC_SAVED), ctx);
+
+	/*
+	 * Use ldil() to load absolute address. Don't use emit_imm as the
+	 * number of emitted instructions should not depend on the value of
+	 * addr.
+	 */
+	emit(hppa_ldil(addr, HPPA_REG_R31), ctx);
+	emit(hppa_be_l(im11(addr) >> 2, HPPA_REG_R31, EXEC_NEXT_INSTR), ctx);
+	/* set return address in delay slot */
+	emit_hppa_copy(HPPA_REG_R31, HPPA_REG_RP, ctx);
+
+	/* restore TCC */
+	if (REG_WAS_SEEN(ctx, HPPA_REG_TCC))
+		emit(hppa_copy(HPPA_REG_TCC_SAVED, HPPA_REG_TCC), ctx);
+
+	/* restore stack */
+	emit(hppa_ldo(-offset_sp, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	/* set return value. */
+	emit_hppa_copy(HPPA_REG_RET0, hi(r0), ctx);
+	emit_hppa_copy(HPPA_REG_RET1, lo(r0), ctx);
+}
+
+static int emit_bpf_tail_call(int insn, struct hppa_jit_context *ctx)
+{
+	/*
+	 * R1 -> &ctx
+	 * R2 -> &array
+	 * R3 -> index
+	 */
+	int off;
+	const s8 *arr_reg = regmap[BPF_REG_2];
+	const s8 *idx_reg = regmap[BPF_REG_3];
+	struct bpf_array bpfa;
+	struct bpf_prog bpfp;
+
+	/* get address of TCC main exit function for error case into rp */
+	emit(EXIT_PTR_LOAD(HPPA_REG_RP), ctx);
+
+	/* max_entries = array->map.max_entries; */
+	off = offsetof(struct bpf_array, map.max_entries);
+	BUILD_BUG_ON(sizeof(bpfa.map.max_entries) != 4);
+	emit(hppa_ldw(off, lo(arr_reg), HPPA_REG_T1), ctx);
+
+	/*
+	 * if (index >= max_entries)
+	 *   goto out;
+	 */
+	emit(hppa_bltu(lo(idx_reg), HPPA_REG_T1, 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/*
+	 * if (--tcc < 0)
+	 *   goto out;
+	 */
+	REG_FORCE_SEEN(ctx, HPPA_REG_TCC);
+	emit(hppa_ldo(-1, HPPA_REG_TCC, HPPA_REG_TCC), ctx);
+	emit(hppa_bge(HPPA_REG_TCC, HPPA_REG_ZERO, 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/*
+	 * prog = array->ptrs[index];
+	 * if (!prog)
+	 *   goto out;
+	 */
+	BUILD_BUG_ON(sizeof(bpfa.ptrs[0]) != 4);
+	emit(hppa_sh2add(lo(idx_reg), lo(arr_reg), HPPA_REG_T0), ctx);
+	off = offsetof(struct bpf_array, ptrs);
+	BUILD_BUG_ON(!relative_bits_ok(off, 11));
+	emit(hppa_ldw(off, HPPA_REG_T0, HPPA_REG_T0), ctx);
+	emit(hppa_bne(HPPA_REG_T0, HPPA_REG_ZERO, 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/*
+	 * tcc = temp_tcc;
+	 * goto *(prog->bpf_func + 4);
+	 */
+	off = offsetof(struct bpf_prog, bpf_func);
+	BUILD_BUG_ON(!relative_bits_ok(off, 11));
+	BUILD_BUG_ON(sizeof(bpfp.bpf_func) != 4);
+	emit(hppa_ldw(off, HPPA_REG_T0, HPPA_REG_T0), ctx);
+	/* Epilogue jumps to *(t0 + 4). */
+	__build_epilogue(true, ctx);
+	return 0;
+}
+
+static int emit_load_r64(const s8 *dst, const s8 *src, s16 off,
+			 struct hppa_jit_context *ctx, const u8 size)
+{
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *tmp2 = regmap[TMP_REG_2];
+	const s8 *rd = bpf_get_reg64_ref(dst, tmp1, ctx->prog->aux->verifier_zext, ctx);
+	const s8 *rs = bpf_get_reg64(src, tmp2, ctx);
+	s8 srcreg;
+
+	/* need to calculate address since offset does not fit in 14 bits? */
+	if (relative_bits_ok(off, 14))
+		srcreg = lo(rs);
+	else {
+		/* need to use R1 here, since addil puts result into R1 */
+		srcreg = HPPA_REG_R1;
+		emit(hppa_addil(off, lo(rs)), ctx);
+		off = im11(off);
+	}
+
+	/* LDX: dst = *(size *)(src + off) */
+	switch (size) {
+	case BPF_B:
+		emit(hppa_ldb(off + 0, srcreg, lo(rd)), ctx);
+		if (!ctx->prog->aux->verifier_zext)
+			emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+		break;
+	case BPF_H:
+		emit(hppa_ldh(off + 0, srcreg, lo(rd)), ctx);
+		if (!ctx->prog->aux->verifier_zext)
+			emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+		break;
+	case BPF_W:
+		emit(hppa_ldw(off + 0, srcreg, lo(rd)), ctx);
+		if (!ctx->prog->aux->verifier_zext)
+			emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+		break;
+	case BPF_DW:
+		emit(hppa_ldw(off + 0, srcreg, hi(rd)), ctx);
+		emit(hppa_ldw(off + 4, srcreg, lo(rd)), ctx);
+		break;
+	}
+
+	bpf_put_reg64(dst, rd, ctx);
+	return 0;
+}
+
+static int emit_store_r64(const s8 *dst, const s8 *src, s16 off,
+			  struct hppa_jit_context *ctx, const u8 size,
+			  const u8 mode)
+{
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *tmp2 = regmap[TMP_REG_2];
+	const s8 *rd = bpf_get_reg64(dst, tmp1, ctx);
+	const s8 *rs = bpf_get_reg64(src, tmp2, ctx);
+	s8 dstreg;
+
+	/* need to calculate address since offset does not fit in 14 bits? */
+	if (relative_bits_ok(off, 14))
+		dstreg = lo(rd);
+	else {
+		/* need to use R1 here, since addil puts result into R1 */
+		dstreg = HPPA_REG_R1;
+		emit(hppa_addil(off, lo(rd)), ctx);
+		off = im11(off);
+	}
+
+	/* ST: *(size *)(dst + off) = imm */
+	switch (size) {
+	case BPF_B:
+		emit(hppa_stb(lo(rs), off + 0, dstreg), ctx);
+		break;
+	case BPF_H:
+		emit(hppa_sth(lo(rs), off + 0, dstreg), ctx);
+		break;
+	case BPF_W:
+		emit(hppa_stw(lo(rs), off + 0, dstreg), ctx);
+		break;
+	case BPF_DW:
+		emit(hppa_stw(hi(rs), off + 0, dstreg), ctx);
+		emit(hppa_stw(lo(rs), off + 4, dstreg), ctx);
+		break;
+	}
+
+	return 0;
+}
+
+static void emit_rev16(const s8 rd, struct hppa_jit_context *ctx)
+{
+	emit(hppa_extru(rd, 23, 8, HPPA_REG_T1), ctx);
+	emit(hppa_depwz(rd, 23, 8, HPPA_REG_T1), ctx);
+	emit(hppa_extru(HPPA_REG_T1, 31, 16, rd), ctx);
+}
+
+static void emit_rev32(const s8 rs, const s8 rd, struct hppa_jit_context *ctx)
+{
+	emit(hppa_shrpw(rs, rs, 16, HPPA_REG_T1), ctx);
+	emit(hppa_depwz(HPPA_REG_T1, 15, 8, HPPA_REG_T1), ctx);
+	emit(hppa_shrpw(rs, HPPA_REG_T1, 8, rd), ctx);
+}
+
+static void emit_zext64(const s8 *dst, struct hppa_jit_context *ctx)
+{
+	const s8 *rd;
+	const s8 *tmp1 = regmap[TMP_REG_1];
+
+	rd = bpf_get_reg64(dst, tmp1, ctx);
+	emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+	bpf_put_reg64(dst, rd, ctx);
+}
+
+int bpf_jit_emit_insn(const struct bpf_insn *insn, struct hppa_jit_context *ctx,
+		      bool extra_pass)
+{
+	bool is64 = BPF_CLASS(insn->code) == BPF_ALU64 ||
+		BPF_CLASS(insn->code) == BPF_JMP;
+	int s, e, paoff, i = insn - ctx->prog->insnsi;
+	u8 code = insn->code;
+	s16 off = insn->off;
+	s32 imm = insn->imm;
+
+	const s8 *dst = regmap[insn->dst_reg];
+	const s8 *src = regmap[insn->src_reg];
+	const s8 *tmp1 = regmap[TMP_REG_1];
+	const s8 *tmp2 = regmap[TMP_REG_2];
+
+	if (0) printk("CLASS %03d  CODE %#02x ALU64:%d BPF_SIZE %#02x  "
+		"BPF_CODE %#02x  src_reg %d  dst_reg %d\n",
+		BPF_CLASS(code), code, (code & BPF_ALU64) ? 1:0, BPF_SIZE(code),
+		BPF_OP(code), insn->src_reg, insn->dst_reg);
+
+	switch (code) {
+	/* dst = src */
+	case BPF_ALU64 | BPF_MOV | BPF_X:
+
+	case BPF_ALU64 | BPF_ADD | BPF_X:
+	case BPF_ALU64 | BPF_ADD | BPF_K:
+
+	case BPF_ALU64 | BPF_SUB | BPF_X:
+	case BPF_ALU64 | BPF_SUB | BPF_K:
+
+	case BPF_ALU64 | BPF_AND | BPF_X:
+	case BPF_ALU64 | BPF_OR | BPF_X:
+	case BPF_ALU64 | BPF_XOR | BPF_X:
+
+	case BPF_ALU64 | BPF_MUL | BPF_X:
+	case BPF_ALU64 | BPF_MUL | BPF_K:
+
+	case BPF_ALU64 | BPF_DIV | BPF_X:
+	case BPF_ALU64 | BPF_DIV | BPF_K:
+
+	case BPF_ALU64 | BPF_MOD | BPF_X:
+	case BPF_ALU64 | BPF_MOD | BPF_K:
+
+	case BPF_ALU64 | BPF_LSH | BPF_X:
+	case BPF_ALU64 | BPF_RSH | BPF_X:
+	case BPF_ALU64 | BPF_ARSH | BPF_X:
+		if (BPF_SRC(code) == BPF_K) {
+			emit_imm32(tmp2, imm, ctx);
+			src = tmp2;
+		}
+		emit_alu_r64(dst, src, ctx, BPF_OP(code));
+		break;
+
+	/* dst = -dst */
+	case BPF_ALU64 | BPF_NEG:
+		emit_alu_r64(dst, tmp2, ctx, BPF_OP(code));
+		break;
+
+	case BPF_ALU64 | BPF_MOV | BPF_K:
+	case BPF_ALU64 | BPF_AND | BPF_K:
+	case BPF_ALU64 | BPF_OR | BPF_K:
+	case BPF_ALU64 | BPF_XOR | BPF_K:
+	case BPF_ALU64 | BPF_LSH | BPF_K:
+	case BPF_ALU64 | BPF_RSH | BPF_K:
+	case BPF_ALU64 | BPF_ARSH | BPF_K:
+		emit_alu_i64(dst, imm, ctx, BPF_OP(code));
+		break;
+
+	case BPF_ALU | BPF_MOV | BPF_X:
+		if (imm == 1) {
+			/* Special mov32 for zext. */
+			emit_zext64(dst, ctx);
+			break;
+		}
+		fallthrough;
+	/* dst = dst OP src */
+	case BPF_ALU | BPF_ADD | BPF_X:
+	case BPF_ALU | BPF_SUB | BPF_X:
+	case BPF_ALU | BPF_AND | BPF_X:
+	case BPF_ALU | BPF_OR | BPF_X:
+	case BPF_ALU | BPF_XOR | BPF_X:
+
+	case BPF_ALU | BPF_MUL | BPF_X:
+	case BPF_ALU | BPF_MUL | BPF_K:
+
+	case BPF_ALU | BPF_DIV | BPF_X:
+	case BPF_ALU | BPF_DIV | BPF_K:
+
+	case BPF_ALU | BPF_MOD | BPF_X:
+	case BPF_ALU | BPF_MOD | BPF_K:
+
+	case BPF_ALU | BPF_LSH | BPF_X:
+	case BPF_ALU | BPF_RSH | BPF_X:
+	case BPF_ALU | BPF_ARSH | BPF_X:
+		if (BPF_SRC(code) == BPF_K) {
+			emit_imm32(tmp2, imm, ctx);
+			src = tmp2;
+		}
+		emit_alu_r32(dst, src, ctx, BPF_OP(code));
+		break;
+
+	/* dst = dst OP imm */
+	case BPF_ALU | BPF_MOV | BPF_K:
+	case BPF_ALU | BPF_ADD | BPF_K:
+	case BPF_ALU | BPF_SUB | BPF_K:
+	case BPF_ALU | BPF_AND | BPF_K:
+	case BPF_ALU | BPF_OR | BPF_K:
+	case BPF_ALU | BPF_XOR | BPF_K:
+	case BPF_ALU | BPF_LSH | BPF_K:
+	case BPF_ALU | BPF_RSH | BPF_K:
+	case BPF_ALU | BPF_ARSH | BPF_K:
+		/*
+		 * mul,div,mod are handled in the BPF_X case.
+		 */
+		emit_alu_i32(dst, imm, ctx, BPF_OP(code));
+		break;
+
+	/* dst = -dst */
+	case BPF_ALU | BPF_NEG:
+		/*
+		 * src is ignored---choose tmp2 as a dummy register since it
+		 * is not on the stack.
+		 */
+		emit_alu_r32(dst, tmp2, ctx, BPF_OP(code));
+		break;
+
+	/* dst = BSWAP##imm(dst) */
+	case BPF_ALU | BPF_END | BPF_FROM_BE:
+	{
+		const s8 *rd = bpf_get_reg64(dst, tmp1, ctx);
+
+		switch (imm) {
+		case 16:
+			/* zero-extend 16 bits into 64 bits */
+			emit(hppa_extru(lo(rd), 31, 16, lo(rd)), ctx);
+			fallthrough;
+		case 32:
+			/* zero-extend 32 bits into 64 bits */
+			if (!ctx->prog->aux->verifier_zext)
+				emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+			break;
+		case 64:
+			/* Do nothing. */
+			break;
+		default:
+			pr_err("bpf-jit: BPF_END imm %d invalid\n", imm);
+			return -1;
+		}
+
+		bpf_put_reg64(dst, rd, ctx);
+		break;
+	}
+
+	case BPF_ALU | BPF_END | BPF_FROM_LE:
+	{
+		const s8 *rd = bpf_get_reg64(dst, tmp1, ctx);
+
+		switch (imm) {
+		case 16:
+			emit_rev16(lo(rd), ctx);
+			if (!ctx->prog->aux->verifier_zext)
+				emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+			break;
+		case 32:
+			emit_rev32(lo(rd), lo(rd), ctx);
+			if (!ctx->prog->aux->verifier_zext)
+				emit_hppa_copy(HPPA_REG_ZERO, hi(rd), ctx);
+			break;
+		case 64:
+			/* Swap upper and lower halves, then each half. */
+			emit_hppa_copy(hi(rd), HPPA_REG_T0, ctx);
+			emit_rev32(lo(rd), hi(rd), ctx);
+			emit_rev32(HPPA_REG_T0, lo(rd), ctx);
+			break;
+		default:
+			pr_err("bpf-jit: BPF_END imm %d invalid\n", imm);
+			return -1;
+		}
+
+		bpf_put_reg64(dst, rd, ctx);
+		break;
+	}
+	/* JUMP off */
+	case BPF_JMP | BPF_JA:
+		paoff = hppa_offset(i, off, ctx);
+		emit_jump(paoff, false, ctx);
+		break;
+	/* function call */
+	case BPF_JMP | BPF_CALL:
+	{
+		bool fixed;
+		int ret;
+		u64 addr;
+
+		ret = bpf_jit_get_func_addr(ctx->prog, insn, extra_pass, &addr,
+					    &fixed);
+		if (ret < 0)
+			return ret;
+		emit_call(fixed, addr, ctx);
+		break;
+	}
+	/* tail call */
+	case BPF_JMP | BPF_TAIL_CALL:
+		REG_SET_SEEN_ALL(ctx);
+		if (emit_bpf_tail_call(i, ctx))
+			return -1;
+		break;
+	/* IF (dst COND imm) JUMP off */
+	case BPF_JMP | BPF_JEQ | BPF_X:
+	case BPF_JMP | BPF_JEQ | BPF_K:
+	case BPF_JMP32 | BPF_JEQ | BPF_X:
+	case BPF_JMP32 | BPF_JEQ | BPF_K:
+
+	case BPF_JMP | BPF_JNE | BPF_X:
+	case BPF_JMP | BPF_JNE | BPF_K:
+	case BPF_JMP32 | BPF_JNE | BPF_X:
+	case BPF_JMP32 | BPF_JNE | BPF_K:
+
+	case BPF_JMP | BPF_JLE | BPF_X:
+	case BPF_JMP | BPF_JLE | BPF_K:
+	case BPF_JMP32 | BPF_JLE | BPF_X:
+	case BPF_JMP32 | BPF_JLE | BPF_K:
+
+	case BPF_JMP | BPF_JLT | BPF_X:
+	case BPF_JMP | BPF_JLT | BPF_K:
+	case BPF_JMP32 | BPF_JLT | BPF_X:
+	case BPF_JMP32 | BPF_JLT | BPF_K:
+
+	case BPF_JMP | BPF_JGE | BPF_X:
+	case BPF_JMP | BPF_JGE | BPF_K:
+	case BPF_JMP32 | BPF_JGE | BPF_X:
+	case BPF_JMP32 | BPF_JGE | BPF_K:
+
+	case BPF_JMP | BPF_JGT | BPF_X:
+	case BPF_JMP | BPF_JGT | BPF_K:
+	case BPF_JMP32 | BPF_JGT | BPF_X:
+	case BPF_JMP32 | BPF_JGT | BPF_K:
+
+	case BPF_JMP | BPF_JSLE | BPF_X:
+	case BPF_JMP | BPF_JSLE | BPF_K:
+	case BPF_JMP32 | BPF_JSLE | BPF_X:
+	case BPF_JMP32 | BPF_JSLE | BPF_K:
+
+	case BPF_JMP | BPF_JSLT | BPF_X:
+	case BPF_JMP | BPF_JSLT | BPF_K:
+	case BPF_JMP32 | BPF_JSLT | BPF_X:
+	case BPF_JMP32 | BPF_JSLT | BPF_K:
+
+	case BPF_JMP | BPF_JSGE | BPF_X:
+	case BPF_JMP | BPF_JSGE | BPF_K:
+	case BPF_JMP32 | BPF_JSGE | BPF_X:
+	case BPF_JMP32 | BPF_JSGE | BPF_K:
+
+	case BPF_JMP | BPF_JSGT | BPF_X:
+	case BPF_JMP | BPF_JSGT | BPF_K:
+	case BPF_JMP32 | BPF_JSGT | BPF_X:
+	case BPF_JMP32 | BPF_JSGT | BPF_K:
+
+	case BPF_JMP | BPF_JSET | BPF_X:
+	case BPF_JMP | BPF_JSET | BPF_K:
+	case BPF_JMP32 | BPF_JSET | BPF_X:
+	case BPF_JMP32 | BPF_JSET | BPF_K:
+		paoff = hppa_offset(i, off, ctx);
+		if (BPF_SRC(code) == BPF_K) {
+			s = ctx->ninsns;
+			emit_imm32(tmp2, imm, ctx);
+			src = tmp2;
+			e = ctx->ninsns;
+			paoff -= (e - s);
+		}
+		if (is64)
+			emit_branch_r64(dst, src, paoff, ctx, BPF_OP(code));
+		else
+			emit_branch_r32(dst, src, paoff, ctx, BPF_OP(code));
+		break;
+	/* function return */
+	case BPF_JMP | BPF_EXIT:
+		if (i == ctx->prog->len - 1)
+			break;
+		/* load epilogue function pointer and jump to it. */
+		emit(EXIT_PTR_LOAD(HPPA_REG_RP), ctx);
+		emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+		break;
+
+	/* dst = imm64 */
+	case BPF_LD | BPF_IMM | BPF_DW:
+	{
+		struct bpf_insn insn1 = insn[1];
+		u32 upper = insn1.imm;
+		u32 lower = imm;
+		const s8 *rd = bpf_get_reg64_ref(dst, tmp1, false, ctx);
+
+		if (0 && bpf_pseudo_func(insn)) {
+			WARN_ON(upper); /* we are 32-bit! */
+			upper = 0;
+			lower = (uintptr_t) dereference_function_descriptor(lower);
+		}
+
+		emit_imm64(rd, upper, lower, ctx);
+		bpf_put_reg64(dst, rd, ctx);
+		return 1;
+	}
+
+	/* LDX: dst = *(size *)(src + off) */
+	case BPF_LDX | BPF_MEM | BPF_B:
+	case BPF_LDX | BPF_MEM | BPF_H:
+	case BPF_LDX | BPF_MEM | BPF_W:
+	case BPF_LDX | BPF_MEM | BPF_DW:
+		if (emit_load_r64(dst, src, off, ctx, BPF_SIZE(code)))
+			return -1;
+		break;
+
+	/* speculation barrier */
+	case BPF_ST | BPF_NOSPEC:
+		break;
+
+	/* ST: *(size *)(dst + off) = imm */
+	case BPF_ST | BPF_MEM | BPF_B:
+	case BPF_ST | BPF_MEM | BPF_H:
+	case BPF_ST | BPF_MEM | BPF_W:
+	case BPF_ST | BPF_MEM | BPF_DW:
+
+	case BPF_STX | BPF_MEM | BPF_B:
+	case BPF_STX | BPF_MEM | BPF_H:
+	case BPF_STX | BPF_MEM | BPF_W:
+	case BPF_STX | BPF_MEM | BPF_DW:
+		if (BPF_CLASS(code) == BPF_ST) {
+			emit_imm32(tmp2, imm, ctx);
+			src = tmp2;
+		}
+
+		if (emit_store_r64(dst, src, off, ctx, BPF_SIZE(code),
+				   BPF_MODE(code)))
+			return -1;
+		break;
+
+	case BPF_STX | BPF_ATOMIC | BPF_W:
+	case BPF_STX | BPF_ATOMIC | BPF_DW:
+		pr_info_once(
+			"bpf-jit: not supported: atomic operation %02x ***\n",
+			insn->imm);
+		return -EFAULT;
+
+	default:
+		pr_err("bpf-jit: unknown opcode %02x\n", code);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+void bpf_jit_build_prologue(struct hppa_jit_context *ctx)
+{
+	const s8 *tmp = regmap[TMP_REG_1];
+	const s8 *dst, *reg;
+	int stack_adjust = 0;
+	int i;
+	unsigned long addr;
+	int bpf_stack_adjust;
+
+	/*
+	 * stack on hppa grows up, so if tail calls are used we need to
+	 * allocate the maximum stack size
+	 */
+	if (REG_ALL_SEEN(ctx))
+		bpf_stack_adjust = MAX_BPF_STACK;
+	else
+		bpf_stack_adjust = ctx->prog->aux->stack_depth;
+	bpf_stack_adjust = round_up(bpf_stack_adjust, STACK_ALIGN);
+
+	/* make space for callee-saved registers. */
+	stack_adjust += NR_SAVED_REGISTERS * REG_SIZE;
+	/* make space for BPF registers on stack. */
+	stack_adjust += BPF_JIT_SCRATCH_REGS * REG_SIZE;
+	/* make space for BPF stack. */
+	stack_adjust += bpf_stack_adjust;
+	/* round up for stack alignment. */
+	stack_adjust = round_up(stack_adjust, STACK_ALIGN);
+
+	/*
+	 * The first instruction sets the tail-call-counter (TCC) register.
+	 * This instruction is skipped by tail calls.
+	 * Use a temporary register instead of a caller-saved register initially.
+	 */
+	emit(hppa_ldi(MAX_TAIL_CALL_CNT, HPPA_REG_TCC_IN_INIT), ctx);
+
+	/*
+	 * skip all initializations when called as BPF TAIL call.
+	 */
+	emit(hppa_ldi(MAX_TAIL_CALL_CNT, HPPA_REG_R1), ctx);
+	emit(hppa_bne(HPPA_REG_TCC_IN_INIT, HPPA_REG_R1, ctx->prologue_len - 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+
+	/* set up hppa stack frame. */
+	emit_hppa_copy(HPPA_REG_SP, HPPA_REG_R1, ctx);			// copy sp,r1 (=prev_sp)
+	emit(hppa_ldo(stack_adjust, HPPA_REG_SP, HPPA_REG_SP), ctx);	// ldo stack_adjust(sp),sp (increase stack)
+	emit(hppa_stw(HPPA_REG_R1, -REG_SIZE, HPPA_REG_SP), ctx);	// stw prev_sp,-0x04(sp)
+	emit(hppa_stw(HPPA_REG_RP, -0x14, HPPA_REG_SP), ctx);		// stw rp,-0x14(sp)
+
+	REG_FORCE_SEEN(ctx, HPPA_REG_T0);
+	REG_FORCE_SEEN(ctx, HPPA_REG_T1);
+	REG_FORCE_SEEN(ctx, HPPA_REG_T2);
+	REG_FORCE_SEEN(ctx, HPPA_REG_T3);
+	REG_FORCE_SEEN(ctx, HPPA_REG_T4);
+	REG_FORCE_SEEN(ctx, HPPA_REG_T5);
+
+	/* save callee-save registers. */
+	for (i = 3; i <= 18; i++) {
+		if (OPTIMIZE_HPPA && !REG_WAS_SEEN(ctx, HPPA_R(i)))
+			continue;
+		emit(hppa_stw(HPPA_R(i), -REG_SIZE * (8 + (i-3)), HPPA_REG_SP), ctx);	// stw ri,-save_area(sp)
+	}
+
+	/*
+	 * now really set the tail call counter (TCC) register.
+	 */
+	if (REG_WAS_SEEN(ctx, HPPA_REG_TCC))
+		emit(hppa_ldi(MAX_TAIL_CALL_CNT, HPPA_REG_TCC), ctx);
+
+	/*
+	 * save epilogue function pointer for outer TCC call chain.
+	 * The main TCC call stores the final RP on stack.
+	 */
+	addr = (uintptr_t) &ctx->insns[ctx->epilogue_offset];
+	/* skip first two instructions of exit function, which jump to exit */
+	addr += 2 * HPPA_INSN_SIZE;
+	emit(hppa_ldil(addr, HPPA_REG_T2), ctx);
+	emit(hppa_ldo(im11(addr), HPPA_REG_T2, HPPA_REG_T2), ctx);
+	emit(EXIT_PTR_STORE(HPPA_REG_T2), ctx);
+
+	/* load R1 & R2 from registers, R3-R5 from stack. */
+	/* use HPPA_REG_R1 which holds the old stack value */
+	dst = regmap[BPF_REG_5];
+	reg = bpf_get_reg64_ref(dst, tmp, false, ctx);
+	if (REG_WAS_SEEN(ctx, lo(reg)) | REG_WAS_SEEN(ctx, hi(reg))) {
+		if (REG_WAS_SEEN(ctx, hi(reg)))
+			emit(hppa_ldw(-0x48, HPPA_REG_R1, hi(reg)), ctx);
+		if (REG_WAS_SEEN(ctx, lo(reg)))
+			emit(hppa_ldw(-0x44, HPPA_REG_R1, lo(reg)), ctx);
+		bpf_put_reg64(dst, tmp, ctx);
+	}
+
+	dst = regmap[BPF_REG_4];
+	reg = bpf_get_reg64_ref(dst, tmp, false, ctx);
+	if (REG_WAS_SEEN(ctx, lo(reg)) | REG_WAS_SEEN(ctx, hi(reg))) {
+		if (REG_WAS_SEEN(ctx, hi(reg)))
+			emit(hppa_ldw(-0x40, HPPA_REG_R1, hi(reg)), ctx);
+		if (REG_WAS_SEEN(ctx, lo(reg)))
+			emit(hppa_ldw(-0x3c, HPPA_REG_R1, lo(reg)), ctx);
+		bpf_put_reg64(dst, tmp, ctx);
+	}
+
+	dst = regmap[BPF_REG_3];
+	reg = bpf_get_reg64_ref(dst, tmp, false, ctx);
+	if (REG_WAS_SEEN(ctx, lo(reg)) | REG_WAS_SEEN(ctx, hi(reg))) {
+		if (REG_WAS_SEEN(ctx, hi(reg)))
+			emit(hppa_ldw(-0x38, HPPA_REG_R1, hi(reg)), ctx);
+		if (REG_WAS_SEEN(ctx, lo(reg)))
+			emit(hppa_ldw(-0x34, HPPA_REG_R1, lo(reg)), ctx);
+		bpf_put_reg64(dst, tmp, ctx);
+	}
+
+	dst = regmap[BPF_REG_2];
+	reg = bpf_get_reg64_ref(dst, tmp, false, ctx);
+	if (REG_WAS_SEEN(ctx, lo(reg)) | REG_WAS_SEEN(ctx, hi(reg))) {
+		if (REG_WAS_SEEN(ctx, hi(reg)))
+			emit_hppa_copy(HPPA_REG_ARG3, hi(reg), ctx);
+		if (REG_WAS_SEEN(ctx, lo(reg)))
+			emit_hppa_copy(HPPA_REG_ARG2, lo(reg), ctx);
+		bpf_put_reg64(dst, tmp, ctx);
+	}
+
+	dst = regmap[BPF_REG_1];
+	reg = bpf_get_reg64_ref(dst, tmp, false, ctx);
+	if (REG_WAS_SEEN(ctx, lo(reg)) | REG_WAS_SEEN(ctx, hi(reg))) {
+		if (REG_WAS_SEEN(ctx, hi(reg)))
+			emit_hppa_copy(HPPA_REG_ARG1, hi(reg), ctx);
+		if (REG_WAS_SEEN(ctx, lo(reg)))
+			emit_hppa_copy(HPPA_REG_ARG0, lo(reg), ctx);
+		bpf_put_reg64(dst, tmp, ctx);
+	}
+
+	/* Set up BPF frame pointer. */
+	dst = regmap[BPF_REG_FP];
+	reg = bpf_get_reg64_ref(dst, tmp, false, ctx);
+	if (REG_WAS_SEEN(ctx, lo(reg)) | REG_WAS_SEEN(ctx, hi(reg))) {
+		if (REG_WAS_SEEN(ctx, lo(reg)))
+			emit(hppa_ldo(-REG_SIZE * (NR_SAVED_REGISTERS + BPF_JIT_SCRATCH_REGS),
+				HPPA_REG_SP, lo(reg)), ctx);
+		if (REG_WAS_SEEN(ctx, hi(reg)))
+			emit_hppa_copy(HPPA_REG_ZERO, hi(reg), ctx);
+		bpf_put_reg64(dst, tmp, ctx);
+	}
+
+	emit(hppa_nop(), ctx);
+}
+
+void bpf_jit_build_epilogue(struct hppa_jit_context *ctx)
+{
+	__build_epilogue(false, ctx);
+}
diff --git a/arch/parisc/net/bpf_jit_comp64.c b/arch/parisc/net/bpf_jit_comp64.c
new file mode 100644
index 000000000000..54b0d5e25e02
--- /dev/null
+++ b/arch/parisc/net/bpf_jit_comp64.c
@@ -0,0 +1,1209 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * BPF JIT compiler for PA-RISC (64-bit)
+ *
+ * Copyright(c) 2023 Helge Deller <deller@gmx.de>
+ *
+ * The code is based on the BPF JIT compiler for RV64 by Björn Töpel.
+ *
+ * TODO:
+ * - check if bpf_jit_needs_zext() is needed (currently enabled)
+ * - implement arch_prepare_bpf_trampoline(), poke(), ...
+ */
+
+#include <linux/bitfield.h>
+#include <linux/bpf.h>
+#include <linux/filter.h>
+#include <linux/libgcc.h>
+#include "bpf_jit.h"
+
+static const int regmap[] = {
+	[BPF_REG_0] =	HPPA_REG_RET0,
+	[BPF_REG_1] =	HPPA_R(5),
+	[BPF_REG_2] =	HPPA_R(6),
+	[BPF_REG_3] =	HPPA_R(7),
+	[BPF_REG_4] =	HPPA_R(8),
+	[BPF_REG_5] =	HPPA_R(9),
+	[BPF_REG_6] =	HPPA_R(10),
+	[BPF_REG_7] =	HPPA_R(11),
+	[BPF_REG_8] =	HPPA_R(12),
+	[BPF_REG_9] =	HPPA_R(13),
+	[BPF_REG_FP] =	HPPA_R(14),
+	[BPF_REG_AX] =	HPPA_R(15),
+};
+
+/*
+ * Stack layout during BPF program execution (note: stack grows up):
+ *
+ *                     high
+ *   HPPA64 sp =>  +----------+ <= HPPA64 fp
+ *                 | saved sp |
+ *                 | saved rp |
+ *                 |   ...    | HPPA64 callee-saved registers
+ *                 | curr args|
+ *                 | local var|
+ *                 +----------+ <= (BPF FP)
+ *                 |          |
+ *                 |   ...    | BPF program stack
+ *                 |          |
+ *                 |   ...    | Function call stack
+ *                 |          |
+ *                 +----------+
+ *                     low
+ */
+
+/* Offset from fp for BPF registers stored on stack. */
+#define STACK_ALIGN	FRAME_SIZE
+
+#define EXIT_PTR_LOAD(reg)	hppa64_ldd_im16(-FRAME_SIZE, HPPA_REG_SP, reg)
+#define EXIT_PTR_STORE(reg)	hppa64_std_im16(reg, -FRAME_SIZE, HPPA_REG_SP)
+#define EXIT_PTR_JUMP(reg, nop)	hppa_bv(HPPA_REG_ZERO, reg, nop)
+
+static u8 bpf_to_hppa_reg(int bpf_reg, struct hppa_jit_context *ctx)
+{
+	u8 reg = regmap[bpf_reg];
+
+	REG_SET_SEEN(ctx, reg);
+	return reg;
+};
+
+static void emit_hppa_copy(const s8 rs, const s8 rd, struct hppa_jit_context *ctx)
+{
+	REG_SET_SEEN(ctx, rd);
+	if (OPTIMIZE_HPPA && (rs == rd))
+		return;
+	REG_SET_SEEN(ctx, rs);
+	emit(hppa_copy(rs, rd), ctx);
+}
+
+static void emit_hppa64_depd(u8 src, u8 pos, u8 len, u8 target, bool no_zero, struct hppa_jit_context *ctx)
+{
+	int c;
+
+	pos &= (BITS_PER_LONG - 1);
+	pos = 63 - pos;
+	len = 64 - len;
+	c =  (len < 32)  ? 0x4 : 0;
+	c |= (pos >= 32) ? 0x2 : 0;
+	c |= (no_zero)   ? 0x1 : 0;
+	emit(hppa_t10_insn(0x3c, target, src, 0, c, pos & 0x1f, len & 0x1f), ctx);
+}
+
+static void emit_hppa64_shld(u8 src, int num, u8 target, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_depd(src, 63-num, 64-num, target, 0, ctx);
+}
+
+static void emit_hppa64_extrd(u8 src, u8 pos, u8 len, u8 target, bool signed_op, struct hppa_jit_context *ctx)
+{
+	int c;
+
+	pos &= (BITS_PER_LONG - 1);
+	len = 64 - len;
+	c =  (len <  32) ? 0x4 : 0;
+	c |= (pos >= 32) ? 0x2 : 0;
+	c |= signed_op   ? 0x1 : 0;
+	emit(hppa_t10_insn(0x36, src, target, 0, c, pos & 0x1f, len & 0x1f), ctx);
+}
+
+static void emit_hppa64_extrw(u8 src, u8 pos, u8 len, u8 target, bool signed_op, struct hppa_jit_context *ctx)
+{
+	int c;
+
+	pos &= (32 - 1);
+	len = 32 - len;
+	c = 0x06 | (signed_op ? 1 : 0);
+	emit(hppa_t10_insn(0x34, src, target, 0, c, pos, len), ctx);
+}
+
+#define emit_hppa64_zext32(r, target, ctx) \
+	emit_hppa64_extrd(r, 63, 32, target, false, ctx)
+#define emit_hppa64_sext32(r, target, ctx) \
+	emit_hppa64_extrd(r, 63, 32, target, true, ctx)
+
+static void emit_hppa64_shrd(u8 src, int num, u8 target, bool signed_op, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_extrd(src, 63-num, 64-num, target, signed_op, ctx);
+}
+
+static void emit_hppa64_shrw(u8 src, int num, u8 target, bool signed_op, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_extrw(src, 31-num, 32-num, target, signed_op, ctx);
+}
+
+/* Emit variable-length instructions for 32-bit imm */
+static void emit_imm32(u8 rd, s32 imm, struct hppa_jit_context *ctx)
+{
+	u32 lower = im11(imm);
+
+	REG_SET_SEEN(ctx, rd);
+	if (OPTIMIZE_HPPA && relative_bits_ok(imm, 14)) {
+		emit(hppa_ldi(imm, rd), ctx);
+		return;
+	}
+	if (OPTIMIZE_HPPA && lower == imm) {
+		emit(hppa_ldo(lower, HPPA_REG_ZERO, rd), ctx);
+		return;
+	}
+	emit(hppa_ldil(imm, rd), ctx);
+	if (OPTIMIZE_HPPA && (lower == 0))
+		return;
+	emit(hppa_ldo(lower, rd, rd), ctx);
+}
+
+static bool is_32b_int(s64 val)
+{
+	return val == (s32) val;
+}
+
+/* Emit variable-length instructions for 64-bit imm */
+static void emit_imm(u8 rd, s64 imm, u8 tmpreg, struct hppa_jit_context *ctx)
+{
+	u32 upper32;
+
+	/* get lower 32-bits into rd, sign extended */
+	emit_imm32(rd, imm, ctx);
+
+	/* do we have upper 32-bits too ? */
+	if (OPTIMIZE_HPPA && is_32b_int(imm))
+		return;
+
+	/* load upper 32-bits into lower tmpreg and deposit into rd */
+	upper32 = imm >> 32;
+	if (upper32 || !OPTIMIZE_HPPA) {
+		emit_imm32(tmpreg, upper32, ctx);
+		emit_hppa64_depd(tmpreg, 31, 32, rd, 1, ctx);
+	} else
+		emit_hppa64_depd(HPPA_REG_ZERO, 31, 32, rd, 1, ctx);
+
+}
+
+static int emit_jump(signed long paoff, bool force_far,
+			       struct hppa_jit_context *ctx)
+{
+	unsigned long pc, addr;
+
+	/* Note: Use 2 instructions for jumps if force_far is set. */
+	if (relative_bits_ok(paoff - HPPA_BRANCH_DISPLACEMENT, 22)) {
+		/* use BL,long branch followed by nop() */
+		emit(hppa64_bl_long(paoff - HPPA_BRANCH_DISPLACEMENT), ctx);
+		if (force_far)
+			emit(hppa_nop(), ctx);
+		return 0;
+	}
+
+	pc = (uintptr_t) &ctx->insns[ctx->ninsns];
+	addr = pc + (paoff * HPPA_INSN_SIZE);
+	/* even the 64-bit kernel runs in memory below 4GB */
+	if (WARN_ON_ONCE(addr >> 32))
+		return -E2BIG;
+	emit(hppa_ldil(addr, HPPA_REG_R31), ctx);
+	emit(hppa_be_l(im11(addr) >> 2, HPPA_REG_R31, NOP_NEXT_INSTR), ctx);
+	return 0;
+}
+
+static void __build_epilogue(bool is_tail_call, struct hppa_jit_context *ctx)
+{
+	int i;
+
+	if (is_tail_call) {
+		/*
+		 * goto *(t0 + 4);
+		 * Skips first instruction of prologue which initializes tail
+		 * call counter. Assumes t0 contains address of target program,
+		 * see emit_bpf_tail_call.
+		 */
+		emit(hppa_ldo(1 * HPPA_INSN_SIZE, HPPA_REG_T0, HPPA_REG_T0), ctx);
+		emit(hppa_bv(HPPA_REG_ZERO, HPPA_REG_T0, EXEC_NEXT_INSTR), ctx);
+		/* in delay slot: */
+		emit(hppa_copy(HPPA_REG_TCC, HPPA_REG_TCC_IN_INIT), ctx);
+
+		return;
+	}
+
+	/* load epilogue function pointer and jump to it. */
+	/* exit point is either at next instruction, or the outest TCC exit function */
+	emit(EXIT_PTR_LOAD(HPPA_REG_RP), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/* NOTE: we are 64-bit and big-endian, so return lower sign-extended 32-bit value */
+	emit_hppa64_sext32(regmap[BPF_REG_0], HPPA_REG_RET0, ctx);
+
+	/* Restore callee-saved registers. */
+	for (i = 3; i <= 15; i++) {
+		if (OPTIMIZE_HPPA && !REG_WAS_SEEN(ctx, HPPA_R(i)))
+			continue;
+		emit(hppa64_ldd_im16(-REG_SIZE * i, HPPA_REG_SP, HPPA_R(i)), ctx);
+	}
+
+	/* load original return pointer (stored by outest TCC function) */
+	emit(hppa64_ldd_im16(-2*REG_SIZE, HPPA_REG_SP, HPPA_REG_RP), ctx);
+	emit(hppa_bv(HPPA_REG_ZERO, HPPA_REG_RP, EXEC_NEXT_INSTR), ctx);
+	/* in delay slot: */
+	emit(hppa64_ldd_im5(-REG_SIZE, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	emit(hppa_nop(), ctx); // XXX WARUM einer zu wenig ??
+}
+
+static int emit_branch(u8 op, u8 rd, u8 rs, signed long paoff,
+			struct hppa_jit_context *ctx)
+{
+	int e, s;
+	bool far = false;
+	int off;
+
+	if (op == BPF_JSET) {
+		/*
+		 * BPF_JSET is a special case: it has no inverse so translate
+		 * to and() function and compare against zero
+		 */
+		emit(hppa_and(rd, rs, HPPA_REG_T0), ctx);
+		paoff -= 1; /* reduce offset due to hppa_and() above */
+		rd = HPPA_REG_T0;
+		rs = HPPA_REG_ZERO;
+		op = BPF_JNE;
+	}
+
+	/* set start after BPF_JSET */
+	s = ctx->ninsns;
+
+	if (!relative_branch_ok(paoff - HPPA_BRANCH_DISPLACEMENT + 1, 12)) {
+		op = invert_bpf_cond(op);
+		far = true;
+	}
+
+	/*
+	 * For a far branch, the condition is negated and we jump over the
+	 * branch itself, and the two instructions from emit_jump.
+	 * For a near branch, just use paoff.
+	 */
+	off = far ? (2 - HPPA_BRANCH_DISPLACEMENT) : paoff - HPPA_BRANCH_DISPLACEMENT;
+
+	switch (op) {
+	/* IF (dst COND src) JUMP off */
+	case BPF_JEQ:
+		emit(hppa_beq(rd, rs, off), ctx);
+		break;
+	case BPF_JGT:
+		emit(hppa_bgtu(rd, rs, off), ctx);
+		break;
+	case BPF_JLT:
+		emit(hppa_bltu(rd, rs, off), ctx);
+		break;
+	case BPF_JGE:
+		emit(hppa_bgeu(rd, rs, off), ctx);
+		break;
+	case BPF_JLE:
+		emit(hppa_bleu(rd, rs, off), ctx);
+		break;
+	case BPF_JNE:
+		emit(hppa_bne(rd, rs, off), ctx);
+		break;
+	case BPF_JSGT:
+		emit(hppa_bgt(rd, rs, off), ctx);
+		break;
+	case BPF_JSLT:
+		emit(hppa_blt(rd, rs, off), ctx);
+		break;
+	case BPF_JSGE:
+		emit(hppa_bge(rd, rs, off), ctx);
+		break;
+	case BPF_JSLE:
+		emit(hppa_ble(rd, rs, off), ctx);
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	if (far) {
+		int ret;
+		e = ctx->ninsns;
+		/* Adjust for extra insns. */
+		paoff -= (e - s);
+		ret = emit_jump(paoff, true, ctx);
+		if (ret)
+			return ret;
+	} else {
+		/*
+		 * always allocate 2 nops instead of the far branch to
+		 * reduce translation loops
+		 */
+		emit(hppa_nop(), ctx);
+		emit(hppa_nop(), ctx);
+	}
+	return 0;
+}
+
+static void emit_zext_32(u8 reg, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_zext32(reg, reg, ctx);
+}
+
+static void emit_bpf_tail_call(int insn, struct hppa_jit_context *ctx)
+{
+	/*
+	 * R1 -> &ctx
+	 * R2 -> &array
+	 * R3 -> index
+	 */
+	int off;
+	const s8 arr_reg = regmap[BPF_REG_2];
+	const s8 idx_reg = regmap[BPF_REG_3];
+	struct bpf_array bpfa;
+	struct bpf_prog bpfp;
+
+	/* if there is any tail call, we need to save & restore all registers */
+	REG_SET_SEEN_ALL(ctx);
+
+	/* get address of TCC main exit function for error case into rp */
+	emit(EXIT_PTR_LOAD(HPPA_REG_RP), ctx);
+
+	/* max_entries = array->map.max_entries; */
+	off = offsetof(struct bpf_array, map.max_entries);
+	BUILD_BUG_ON(sizeof(bpfa.map.max_entries) != 4);
+	emit(hppa_ldw(off, arr_reg, HPPA_REG_T1), ctx);
+
+	/*
+	 * if (index >= max_entries)
+	 *   goto out;
+	 */
+	emit(hppa_bltu(idx_reg, HPPA_REG_T1, 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/*
+	 * if (--tcc < 0)
+	 *   goto out;
+	 */
+	REG_FORCE_SEEN(ctx, HPPA_REG_TCC);
+	emit(hppa_ldo(-1, HPPA_REG_TCC, HPPA_REG_TCC), ctx);
+	emit(hppa_bge(HPPA_REG_TCC, HPPA_REG_ZERO, 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/*
+	 * prog = array->ptrs[index];
+	 * if (!prog)
+	 *   goto out;
+	 */
+	BUILD_BUG_ON(sizeof(bpfa.ptrs[0]) != 8);
+	emit(hppa64_shladd(idx_reg, 3, arr_reg, HPPA_REG_T0), ctx);
+	off = offsetof(struct bpf_array, ptrs);
+	BUILD_BUG_ON(off < 16);
+	emit(hppa64_ldd_im16(off, HPPA_REG_T0, HPPA_REG_T0), ctx);
+	emit(hppa_bne(HPPA_REG_T0, HPPA_REG_ZERO, 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	emit(EXIT_PTR_JUMP(HPPA_REG_RP, NOP_NEXT_INSTR), ctx);
+
+	/*
+	 * tcc = temp_tcc;
+	 * goto *(prog->bpf_func + 4);
+	 */
+	off = offsetof(struct bpf_prog, bpf_func);
+	BUILD_BUG_ON(off < 16);
+	BUILD_BUG_ON(sizeof(bpfp.bpf_func) != 8);
+	emit(hppa64_ldd_im16(off, HPPA_REG_T0, HPPA_REG_T0), ctx);
+	/* Epilogue jumps to *(t0 + 4). */
+	__build_epilogue(true, ctx);
+}
+
+static void init_regs(u8 *rd, u8 *rs, const struct bpf_insn *insn,
+		      struct hppa_jit_context *ctx)
+{
+	u8 code = insn->code;
+
+	switch (code) {
+	case BPF_JMP | BPF_JA:
+	case BPF_JMP | BPF_CALL:
+	case BPF_JMP | BPF_EXIT:
+	case BPF_JMP | BPF_TAIL_CALL:
+		break;
+	default:
+		*rd = bpf_to_hppa_reg(insn->dst_reg, ctx);
+	}
+
+	if (code & (BPF_ALU | BPF_X) || code & (BPF_ALU64 | BPF_X) ||
+	    code & (BPF_JMP | BPF_X) || code & (BPF_JMP32 | BPF_X) ||
+	    code & BPF_LDX || code & BPF_STX)
+		*rs = bpf_to_hppa_reg(insn->src_reg, ctx);
+}
+
+static void emit_zext_32_rd_rs(u8 *rd, u8 *rs, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_zext32(*rd, HPPA_REG_T2, ctx);
+	*rd = HPPA_REG_T2;
+	emit_hppa64_zext32(*rs, HPPA_REG_T1, ctx);
+	*rs = HPPA_REG_T1;
+}
+
+static void emit_sext_32_rd_rs(u8 *rd, u8 *rs, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_sext32(*rd, HPPA_REG_T2, ctx);
+	*rd = HPPA_REG_T2;
+	emit_hppa64_sext32(*rs, HPPA_REG_T1, ctx);
+	*rs = HPPA_REG_T1;
+}
+
+static void emit_zext_32_rd_t1(u8 *rd, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_zext32(*rd, HPPA_REG_T2, ctx);
+	*rd = HPPA_REG_T2;
+	emit_zext_32(HPPA_REG_T1, ctx);
+}
+
+static void emit_sext_32_rd(u8 *rd, struct hppa_jit_context *ctx)
+{
+	emit_hppa64_sext32(*rd, HPPA_REG_T2, ctx);
+	*rd = HPPA_REG_T2;
+}
+
+static bool is_signed_bpf_cond(u8 cond)
+{
+	return cond == BPF_JSGT || cond == BPF_JSLT ||
+		cond == BPF_JSGE || cond == BPF_JSLE;
+}
+
+static void emit_call(u64 addr, bool fixed, struct hppa_jit_context *ctx)
+{
+	const int offset_sp = 2*FRAME_SIZE;
+
+	emit(hppa_ldo(offset_sp, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	emit_hppa_copy(regmap[BPF_REG_1], HPPA_REG_ARG0, ctx);
+	emit_hppa_copy(regmap[BPF_REG_2], HPPA_REG_ARG1, ctx);
+	emit_hppa_copy(regmap[BPF_REG_3], HPPA_REG_ARG2, ctx);
+	emit_hppa_copy(regmap[BPF_REG_4], HPPA_REG_ARG3, ctx);
+	emit_hppa_copy(regmap[BPF_REG_5], HPPA_REG_ARG4, ctx);
+
+	/* Backup TCC. */
+	REG_FORCE_SEEN(ctx, HPPA_REG_TCC_SAVED);
+	if (REG_WAS_SEEN(ctx, HPPA_REG_TCC))
+		emit(hppa_copy(HPPA_REG_TCC, HPPA_REG_TCC_SAVED), ctx);
+
+	/*
+	 * Use ldil() to load absolute address. Don't use emit_imm as the
+	 * number of emitted instructions should not depend on the value of
+	 * addr.
+	 */
+	WARN_ON(addr >> 32);
+	/* load function address and gp from Elf64_Fdesc descriptor */
+	emit(hppa_ldil(addr, HPPA_REG_R31), ctx);
+	emit(hppa_ldo(im11(addr), HPPA_REG_R31, HPPA_REG_R31), ctx);
+	emit(hppa64_ldd_im16(offsetof(struct elf64_fdesc, addr),
+			     HPPA_REG_R31, HPPA_REG_RP), ctx);
+	emit(hppa64_bve_l_rp(HPPA_REG_RP), ctx);
+	emit(hppa64_ldd_im16(offsetof(struct elf64_fdesc, gp),
+			     HPPA_REG_R31, HPPA_REG_GP), ctx);
+
+	/* Restore TCC. */
+	if (REG_WAS_SEEN(ctx, HPPA_REG_TCC))
+		emit(hppa_copy(HPPA_REG_TCC_SAVED, HPPA_REG_TCC), ctx);
+
+	emit(hppa_ldo(-offset_sp, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	/* Set return value. */
+	emit_hppa_copy(HPPA_REG_RET0, regmap[BPF_REG_0], ctx);
+}
+
+static void emit_call_libgcc_ll(void *func, const s8 arg0,
+		const s8 arg1, u8 opcode, struct hppa_jit_context *ctx)
+{
+	u64 func_addr;
+
+	if (BPF_CLASS(opcode) == BPF_ALU) {
+		emit_hppa64_zext32(arg0, HPPA_REG_ARG0, ctx);
+		emit_hppa64_zext32(arg1, HPPA_REG_ARG1, ctx);
+	} else {
+		emit_hppa_copy(arg0, HPPA_REG_ARG0, ctx);
+		emit_hppa_copy(arg1, HPPA_REG_ARG1, ctx);
+	}
+
+	/* libcgcc overwrites HPPA_REG_RET0, so keep copy in HPPA_REG_TCC_SAVED */
+	if (arg0 != HPPA_REG_RET0) {
+		REG_SET_SEEN(ctx, HPPA_REG_TCC_SAVED);
+		emit(hppa_copy(HPPA_REG_RET0, HPPA_REG_TCC_SAVED), ctx);
+	}
+
+	/* set up stack */
+	emit(hppa_ldo(FRAME_SIZE, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	func_addr = (uintptr_t) func;
+	/* load function func_address and gp from Elf64_Fdesc descriptor */
+	emit_imm(HPPA_REG_R31, func_addr, arg0, ctx);
+	emit(hppa64_ldd_im16(offsetof(struct elf64_fdesc, addr),
+			     HPPA_REG_R31, HPPA_REG_RP), ctx);
+        /* skip the following bve_l instruction if divisor is 0. */
+        if (BPF_OP(opcode) == BPF_DIV || BPF_OP(opcode) == BPF_MOD) {
+		if (BPF_OP(opcode) == BPF_DIV)
+			emit_hppa_copy(HPPA_REG_ZERO, HPPA_REG_RET0, ctx);
+		else {
+			emit_hppa_copy(HPPA_REG_ARG0, HPPA_REG_RET0, ctx);
+		}
+		emit(hppa_beq(HPPA_REG_ARG1, HPPA_REG_ZERO, 2 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	}
+	emit(hppa64_bve_l_rp(HPPA_REG_RP), ctx);
+	emit(hppa64_ldd_im16(offsetof(struct elf64_fdesc, gp),
+			     HPPA_REG_R31, HPPA_REG_GP), ctx);
+
+	emit(hppa_ldo(-FRAME_SIZE, HPPA_REG_SP, HPPA_REG_SP), ctx);
+
+	emit_hppa_copy(HPPA_REG_RET0, arg0, ctx);
+
+	/* restore HPPA_REG_RET0 */
+	if (arg0 != HPPA_REG_RET0)
+		emit(hppa_copy(HPPA_REG_TCC_SAVED, HPPA_REG_RET0), ctx);
+}
+
+static void emit_store(const s8 rd, const s8 rs, s16 off,
+			  struct hppa_jit_context *ctx, const u8 size,
+			  const u8 mode)
+{
+	s8 dstreg;
+
+	/* need to calculate address since offset does not fit in 14 bits? */
+	if (relative_bits_ok(off, 14))
+		dstreg = rd;
+	else {
+		/* need to use R1 here, since addil puts result into R1 */
+		dstreg = HPPA_REG_R1;
+		emit(hppa_addil(off, rd), ctx);
+		off = im11(off);
+	}
+
+	switch (size) {
+	case BPF_B:
+		emit(hppa_stb(rs, off, dstreg), ctx);
+		break;
+	case BPF_H:
+		emit(hppa_sth(rs, off, dstreg), ctx);
+		break;
+	case BPF_W:
+		emit(hppa_stw(rs, off, dstreg), ctx);
+		break;
+	case BPF_DW:
+		if (off & 7) {
+			emit(hppa_ldo(off, dstreg, HPPA_REG_R1), ctx);
+			emit(hppa64_std_im5(rs, 0, HPPA_REG_R1), ctx);
+		} else if (off >= -16 && off <= 15)
+			emit(hppa64_std_im5(rs, off, dstreg), ctx);
+		else
+			emit(hppa64_std_im16(rs, off, dstreg), ctx);
+		break;
+	}
+}
+
+int bpf_jit_emit_insn(const struct bpf_insn *insn, struct hppa_jit_context *ctx,
+		      bool extra_pass)
+{
+	bool is64 = BPF_CLASS(insn->code) == BPF_ALU64 ||
+		    BPF_CLASS(insn->code) == BPF_JMP;
+	int s, e, ret, i = insn - ctx->prog->insnsi;
+	s64 paoff;
+	struct bpf_prog_aux *aux = ctx->prog->aux;
+	u8 rd = -1, rs = -1, code = insn->code;
+	s16 off = insn->off;
+	s32 imm = insn->imm;
+
+	init_regs(&rd, &rs, insn, ctx);
+
+	switch (code) {
+	/* dst = src */
+	case BPF_ALU | BPF_MOV | BPF_X:
+	case BPF_ALU64 | BPF_MOV | BPF_X:
+		if (imm == 1) {
+			/* Special mov32 for zext */
+			emit_zext_32(rd, ctx);
+			break;
+		}
+		if (!is64 && !aux->verifier_zext)
+			emit_hppa64_zext32(rs, rd, ctx);
+		else
+			emit_hppa_copy(rs, rd, ctx);
+		break;
+
+	/* dst = dst OP src */
+	case BPF_ALU | BPF_ADD | BPF_X:
+	case BPF_ALU64 | BPF_ADD | BPF_X:
+                emit(hppa_add(rd, rs, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_SUB | BPF_X:
+	case BPF_ALU64 | BPF_SUB | BPF_X:
+                emit(hppa_sub(rd, rs, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_AND | BPF_X:
+	case BPF_ALU64 | BPF_AND | BPF_X:
+                emit(hppa_and(rd, rs, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_OR | BPF_X:
+	case BPF_ALU64 | BPF_OR | BPF_X:
+                emit(hppa_or(rd, rs, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_XOR | BPF_X:
+	case BPF_ALU64 | BPF_XOR | BPF_X:
+                emit(hppa_xor(rd, rs, rd), ctx);
+		if (!is64 && !aux->verifier_zext && rs != rd)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_MUL | BPF_K:
+	case BPF_ALU64 | BPF_MUL | BPF_K:
+		emit_imm(HPPA_REG_T1, is64 ? (s64)(s32)imm : (u32)imm, HPPA_REG_T2, ctx);
+		rs = HPPA_REG_T1;
+		fallthrough;
+	case BPF_ALU | BPF_MUL | BPF_X:
+	case BPF_ALU64 | BPF_MUL | BPF_X:
+		emit_call_libgcc_ll(__muldi3, rd, rs, code, ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_DIV | BPF_K:
+	case BPF_ALU64 | BPF_DIV | BPF_K:
+		emit_imm(HPPA_REG_T1, is64 ? (s64)(s32)imm : (u32)imm, HPPA_REG_T2, ctx);
+		rs = HPPA_REG_T1;
+		fallthrough;
+	case BPF_ALU | BPF_DIV | BPF_X:
+	case BPF_ALU64 | BPF_DIV | BPF_X:
+		emit_call_libgcc_ll(&hppa_div64, rd, rs, code, ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_MOD | BPF_K:
+	case BPF_ALU64 | BPF_MOD | BPF_K:
+		emit_imm(HPPA_REG_T1, is64 ? (s64)(s32)imm : (u32)imm, HPPA_REG_T2, ctx);
+		rs = HPPA_REG_T1;
+		fallthrough;
+	case BPF_ALU | BPF_MOD | BPF_X:
+	case BPF_ALU64 | BPF_MOD | BPF_X:
+		emit_call_libgcc_ll(&hppa_div64_rem, rd, rs, code, ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+
+	case BPF_ALU | BPF_LSH | BPF_X:
+	case BPF_ALU64 | BPF_LSH | BPF_X:
+		emit_hppa64_sext32(rs, HPPA_REG_T0, ctx);
+		emit(hppa64_mtsarcm(HPPA_REG_T0), ctx);
+		if (is64)
+			emit(hppa64_depdz_sar(rd, rd), ctx);
+		else
+			emit(hppa_depwz_sar(rd, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_RSH | BPF_X:
+	case BPF_ALU64 | BPF_RSH | BPF_X:
+		emit(hppa_mtsar(rs), ctx);
+		if (is64)
+			emit(hppa64_shrpd_sar(rd, rd), ctx);
+		else
+			emit(hppa_shrpw_sar(rd, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_ARSH | BPF_X:
+	case BPF_ALU64 | BPF_ARSH | BPF_X:
+		emit_hppa64_sext32(rs, HPPA_REG_T0, ctx);
+                emit(hppa64_mtsarcm(HPPA_REG_T0), ctx);
+		if (is64)
+			emit(hppa_extrd_sar(rd, rd, 1), ctx);
+		else
+			emit(hppa_extrws_sar(rd, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+
+	/* dst = -dst */
+	case BPF_ALU | BPF_NEG:
+	case BPF_ALU64 | BPF_NEG:
+		emit(hppa_sub(HPPA_REG_ZERO, rd, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+
+	/* dst = BSWAP##imm(dst) */
+	case BPF_ALU | BPF_END | BPF_FROM_BE:
+		switch (imm) {
+		case 16:
+			/* zero-extend 16 bits into 64 bits */
+			emit_hppa64_depd(HPPA_REG_ZERO, 63-16, 64-16, rd, 1, ctx);
+			break;
+		case 32:
+			if (!aux->verifier_zext)
+				emit_zext_32(rd, ctx);
+			break;
+		case 64:
+			/* Do nothing */
+			break;
+		}
+		break;
+
+	case BPF_ALU | BPF_END | BPF_FROM_LE:
+		switch (imm) {
+		case 16:
+			emit(hppa_extru(rd, 31 - 8, 8, HPPA_REG_T1), ctx);
+			emit(hppa_depwz(rd, 23, 8, HPPA_REG_T1), ctx);
+			emit(hppa_extru(HPPA_REG_T1, 31, 16, rd), ctx);
+			emit_hppa64_extrd(HPPA_REG_T1, 63, 16, rd, 0, ctx);
+			break;
+		case 32:
+			emit(hppa_shrpw(rd, rd, 16, HPPA_REG_T1), ctx);
+			emit_hppa64_depd(HPPA_REG_T1, 63-16, 8, HPPA_REG_T1, 1, ctx);
+			emit(hppa_shrpw(rd, HPPA_REG_T1, 8, HPPA_REG_T1), ctx);
+			emit_hppa64_extrd(HPPA_REG_T1, 63, 32, rd, 0, ctx);
+			break;
+		case 64:
+			emit(hppa64_permh_3210(rd, HPPA_REG_T1), ctx);
+			emit(hppa64_hshl(HPPA_REG_T1, 8, HPPA_REG_T2), ctx);
+			emit(hppa64_hshr_u(HPPA_REG_T1, 8, HPPA_REG_T1), ctx);
+			emit(hppa_or(HPPA_REG_T2, HPPA_REG_T1, rd), ctx);
+			break;
+		default:
+			pr_err("bpf-jit: BPF_END imm %d invalid\n", imm);
+			return -1;
+		}
+		break;
+
+	/* dst = imm */
+	case BPF_ALU | BPF_MOV | BPF_K:
+	case BPF_ALU64 | BPF_MOV | BPF_K:
+		emit_imm(rd, imm, HPPA_REG_T2, ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+
+	/* dst = dst OP imm */
+	case BPF_ALU | BPF_ADD | BPF_K:
+	case BPF_ALU64 | BPF_ADD | BPF_K:
+		if (relative_bits_ok(imm, 14)) {
+			emit(hppa_ldo(imm, rd, rd), ctx);
+		} else {
+			emit_imm(HPPA_REG_T1, imm, HPPA_REG_T2, ctx);
+			emit(hppa_add(rd, HPPA_REG_T1, rd), ctx);
+		}
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_SUB | BPF_K:
+	case BPF_ALU64 | BPF_SUB | BPF_K:
+		if (relative_bits_ok(-imm, 14)) {
+			emit(hppa_ldo(-imm, rd, rd), ctx);
+		} else {
+			emit_imm(HPPA_REG_T1, imm, HPPA_REG_T2, ctx);
+			emit(hppa_sub(rd, HPPA_REG_T1, rd), ctx);
+		}
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_AND | BPF_K:
+	case BPF_ALU64 | BPF_AND | BPF_K:
+		emit_imm(HPPA_REG_T1, imm, HPPA_REG_T2, ctx);
+                emit(hppa_and(rd, HPPA_REG_T1, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_OR | BPF_K:
+	case BPF_ALU64 | BPF_OR | BPF_K:
+		emit_imm(HPPA_REG_T1, imm, HPPA_REG_T2, ctx);
+                emit(hppa_or(rd, HPPA_REG_T1, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_XOR | BPF_K:
+	case BPF_ALU64 | BPF_XOR | BPF_K:
+		emit_imm(HPPA_REG_T1, imm, HPPA_REG_T2, ctx);
+                emit(hppa_xor(rd, HPPA_REG_T1, rd), ctx);
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_LSH | BPF_K:
+	case BPF_ALU64 | BPF_LSH | BPF_K:
+		if (imm != 0) {
+			emit_hppa64_shld(rd, imm, rd, ctx);
+		}
+
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_RSH | BPF_K:
+	case BPF_ALU64 | BPF_RSH | BPF_K:
+		if (imm != 0) {
+			if (is64)
+				emit_hppa64_shrd(rd, imm, rd, false, ctx);
+			else
+				emit_hppa64_shrw(rd, imm, rd, false, ctx);
+		}
+
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+	case BPF_ALU | BPF_ARSH | BPF_K:
+	case BPF_ALU64 | BPF_ARSH | BPF_K:
+		if (imm != 0) {
+			if (is64)
+				emit_hppa64_shrd(rd, imm, rd, true, ctx);
+			else
+				emit_hppa64_shrw(rd, imm, rd, true, ctx);
+		}
+
+		if (!is64 && !aux->verifier_zext)
+			emit_zext_32(rd, ctx);
+		break;
+
+	/* JUMP off */
+	case BPF_JMP | BPF_JA:
+		paoff = hppa_offset(i, off, ctx);
+		ret = emit_jump(paoff, false, ctx);
+		if (ret)
+			return ret;
+		break;
+
+	/* IF (dst COND src) JUMP off */
+	case BPF_JMP | BPF_JEQ | BPF_X:
+	case BPF_JMP32 | BPF_JEQ | BPF_X:
+	case BPF_JMP | BPF_JGT | BPF_X:
+	case BPF_JMP32 | BPF_JGT | BPF_X:
+	case BPF_JMP | BPF_JLT | BPF_X:
+	case BPF_JMP32 | BPF_JLT | BPF_X:
+	case BPF_JMP | BPF_JGE | BPF_X:
+	case BPF_JMP32 | BPF_JGE | BPF_X:
+	case BPF_JMP | BPF_JLE | BPF_X:
+	case BPF_JMP32 | BPF_JLE | BPF_X:
+	case BPF_JMP | BPF_JNE | BPF_X:
+	case BPF_JMP32 | BPF_JNE | BPF_X:
+	case BPF_JMP | BPF_JSGT | BPF_X:
+	case BPF_JMP32 | BPF_JSGT | BPF_X:
+	case BPF_JMP | BPF_JSLT | BPF_X:
+	case BPF_JMP32 | BPF_JSLT | BPF_X:
+	case BPF_JMP | BPF_JSGE | BPF_X:
+	case BPF_JMP32 | BPF_JSGE | BPF_X:
+	case BPF_JMP | BPF_JSLE | BPF_X:
+	case BPF_JMP32 | BPF_JSLE | BPF_X:
+	case BPF_JMP | BPF_JSET | BPF_X:
+	case BPF_JMP32 | BPF_JSET | BPF_X:
+		paoff = hppa_offset(i, off, ctx);
+		if (!is64) {
+			s = ctx->ninsns;
+			if (is_signed_bpf_cond(BPF_OP(code)))
+				emit_sext_32_rd_rs(&rd, &rs, ctx);
+			else
+				emit_zext_32_rd_rs(&rd, &rs, ctx);
+			e = ctx->ninsns;
+
+			/* Adjust for extra insns */
+			paoff -= (e - s);
+		}
+		if (BPF_OP(code) == BPF_JSET) {
+			/* Adjust for and */
+			paoff -= 1;
+			emit(hppa_and(rs, rd, HPPA_REG_T1), ctx);
+			emit_branch(BPF_JNE, HPPA_REG_T1, HPPA_REG_ZERO, paoff,
+				    ctx);
+		} else {
+			emit_branch(BPF_OP(code), rd, rs, paoff, ctx);
+		}
+		break;
+
+	/* IF (dst COND imm) JUMP off */
+	case BPF_JMP | BPF_JEQ | BPF_K:
+	case BPF_JMP32 | BPF_JEQ | BPF_K:
+	case BPF_JMP | BPF_JGT | BPF_K:
+	case BPF_JMP32 | BPF_JGT | BPF_K:
+	case BPF_JMP | BPF_JLT | BPF_K:
+	case BPF_JMP32 | BPF_JLT | BPF_K:
+	case BPF_JMP | BPF_JGE | BPF_K:
+	case BPF_JMP32 | BPF_JGE | BPF_K:
+	case BPF_JMP | BPF_JLE | BPF_K:
+	case BPF_JMP32 | BPF_JLE | BPF_K:
+	case BPF_JMP | BPF_JNE | BPF_K:
+	case BPF_JMP32 | BPF_JNE | BPF_K:
+	case BPF_JMP | BPF_JSGT | BPF_K:
+	case BPF_JMP32 | BPF_JSGT | BPF_K:
+	case BPF_JMP | BPF_JSLT | BPF_K:
+	case BPF_JMP32 | BPF_JSLT | BPF_K:
+	case BPF_JMP | BPF_JSGE | BPF_K:
+	case BPF_JMP32 | BPF_JSGE | BPF_K:
+	case BPF_JMP | BPF_JSLE | BPF_K:
+	case BPF_JMP32 | BPF_JSLE | BPF_K:
+		paoff = hppa_offset(i, off, ctx);
+		s = ctx->ninsns;
+		if (imm) {
+			emit_imm(HPPA_REG_T1, imm, HPPA_REG_T2, ctx);
+			rs = HPPA_REG_T1;
+		} else {
+			rs = HPPA_REG_ZERO;
+		}
+		if (!is64) {
+			if (is_signed_bpf_cond(BPF_OP(code)))
+				emit_sext_32_rd(&rd, ctx);
+			else
+				emit_zext_32_rd_t1(&rd, ctx);
+		}
+		e = ctx->ninsns;
+
+		/* Adjust for extra insns */
+		paoff -= (e - s);
+		emit_branch(BPF_OP(code), rd, rs, paoff, ctx);
+		break;
+	case BPF_JMP | BPF_JSET | BPF_K:
+	case BPF_JMP32 | BPF_JSET | BPF_K:
+		paoff = hppa_offset(i, off, ctx);
+		s = ctx->ninsns;
+		emit_imm(HPPA_REG_T1, imm, HPPA_REG_T2, ctx);
+		emit(hppa_and(HPPA_REG_T1, rd, HPPA_REG_T1), ctx);
+		/* For jset32, we should clear the upper 32 bits of t1, but
+		 * sign-extension is sufficient here and saves one instruction,
+		 * as t1 is used only in comparison against zero.
+		 */
+		if (!is64 && imm < 0)
+			emit_hppa64_sext32(HPPA_REG_T1, HPPA_REG_T1, ctx);
+		e = ctx->ninsns;
+		paoff -= (e - s);
+		emit_branch(BPF_JNE, HPPA_REG_T1, HPPA_REG_ZERO, paoff, ctx);
+		break;
+	/* function call */
+	case BPF_JMP | BPF_CALL:
+	{
+		bool fixed_addr;
+		u64 addr;
+
+		ret = bpf_jit_get_func_addr(ctx->prog, insn, extra_pass,
+					    &addr, &fixed_addr);
+		if (ret < 0)
+			return ret;
+
+		REG_SET_SEEN_ALL(ctx);
+		emit_call(addr, fixed_addr, ctx);
+		break;
+	}
+	/* tail call */
+	case BPF_JMP | BPF_TAIL_CALL:
+		emit_bpf_tail_call(i, ctx);
+		break;
+
+	/* function return */
+	case BPF_JMP | BPF_EXIT:
+		if (i == ctx->prog->len - 1)
+			break;
+
+		paoff = epilogue_offset(ctx);
+		ret = emit_jump(paoff, false, ctx);
+		if (ret)
+			return ret;
+		break;
+
+	/* dst = imm64 */
+	case BPF_LD | BPF_IMM | BPF_DW:
+	{
+		struct bpf_insn insn1 = insn[1];
+		u64 imm64 = (u64)insn1.imm << 32 | (u32)imm;
+		if (bpf_pseudo_func(insn))
+			imm64 = (uintptr_t)dereference_function_descriptor((void*)imm64);
+		emit_imm(rd, imm64, HPPA_REG_T2, ctx);
+
+		return 1;
+	}
+
+	/* LDX: dst = *(size *)(src + off) */
+	case BPF_LDX | BPF_MEM | BPF_B:
+	case BPF_LDX | BPF_MEM | BPF_H:
+	case BPF_LDX | BPF_MEM | BPF_W:
+	case BPF_LDX | BPF_MEM | BPF_DW:
+	case BPF_LDX | BPF_PROBE_MEM | BPF_B:
+	case BPF_LDX | BPF_PROBE_MEM | BPF_H:
+	case BPF_LDX | BPF_PROBE_MEM | BPF_W:
+	case BPF_LDX | BPF_PROBE_MEM | BPF_DW:
+	{
+		u8 srcreg;
+
+		/* need to calculate address since offset does not fit in 14 bits? */
+		if (relative_bits_ok(off, 14))
+			srcreg = rs;
+		else {
+			/* need to use R1 here, since addil puts result into R1 */
+			srcreg = HPPA_REG_R1;
+			BUG_ON(rs == HPPA_REG_R1);
+			BUG_ON(rd == HPPA_REG_R1);
+			emit(hppa_addil(off, rs), ctx);
+			off = im11(off);
+		}
+
+		switch (BPF_SIZE(code)) {
+		case BPF_B:
+			emit(hppa_ldb(off, srcreg, rd), ctx);
+			if (insn_is_zext(&insn[1]))
+				return 1;
+			break;
+		case BPF_H:
+			emit(hppa_ldh(off, srcreg, rd), ctx);
+			if (insn_is_zext(&insn[1]))
+				return 1;
+			break;
+		case BPF_W:
+			emit(hppa_ldw(off, srcreg, rd), ctx);
+			if (insn_is_zext(&insn[1]))
+				return 1;
+			break;
+		case BPF_DW:
+			if (off & 7) {
+				emit(hppa_ldo(off, srcreg, HPPA_REG_R1), ctx);
+				emit(hppa64_ldd_reg(HPPA_REG_ZERO, HPPA_REG_R1, rd), ctx);
+			} else if (off >= -16 && off <= 15)
+				emit(hppa64_ldd_im5(off, srcreg, rd), ctx);
+			else
+				emit(hppa64_ldd_im16(off, srcreg, rd), ctx);
+			break;
+		}
+		break;
+	}
+	/* speculation barrier */
+	case BPF_ST | BPF_NOSPEC:
+		break;
+
+	/* ST: *(size *)(dst + off) = imm */
+	/* STX: *(size *)(dst + off) = src */
+	case BPF_ST | BPF_MEM | BPF_B:
+	case BPF_ST | BPF_MEM | BPF_H:
+	case BPF_ST | BPF_MEM | BPF_W:
+	case BPF_ST | BPF_MEM | BPF_DW:
+
+	case BPF_STX | BPF_MEM | BPF_B:
+	case BPF_STX | BPF_MEM | BPF_H:
+	case BPF_STX | BPF_MEM | BPF_W:
+	case BPF_STX | BPF_MEM | BPF_DW:
+		if (BPF_CLASS(code) == BPF_ST) {
+			emit_imm(HPPA_REG_T2, imm, HPPA_REG_T1, ctx);
+			rs = HPPA_REG_T2;
+		}
+
+		emit_store(rd, rs, off, ctx, BPF_SIZE(code), BPF_MODE(code));
+		break;
+
+	case BPF_STX | BPF_ATOMIC | BPF_W:
+	case BPF_STX | BPF_ATOMIC | BPF_DW:
+		pr_info_once(
+			"bpf-jit: not supported: atomic operation %02x ***\n",
+			insn->imm);
+		return -EFAULT;
+
+	default:
+		pr_err("bpf-jit: unknown opcode %02x\n", code);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+void bpf_jit_build_prologue(struct hppa_jit_context *ctx)
+{
+	int bpf_stack_adjust, stack_adjust, i;
+	unsigned long addr;
+	s8 reg;
+
+	/*
+	 * stack on hppa grows up, so if tail calls are used we need to
+	 * allocate the maximum stack size
+	 */
+	if (REG_ALL_SEEN(ctx))
+		bpf_stack_adjust = MAX_BPF_STACK;
+	else
+		bpf_stack_adjust = ctx->prog->aux->stack_depth;
+	bpf_stack_adjust = round_up(bpf_stack_adjust, STACK_ALIGN);
+
+	stack_adjust = FRAME_SIZE + bpf_stack_adjust;
+	stack_adjust = round_up(stack_adjust, STACK_ALIGN);
+
+	/*
+	 * NOTE: We construct an Elf64_Fdesc descriptor here.
+	 * The first 4 words initialize the TCC and compares them.
+	 * Then follows the virtual address of the eBPF function,
+	 * and the gp for this function.
+	 *
+	 * The first instruction sets the tail-call-counter (TCC) register.
+	 * This instruction is skipped by tail calls.
+	 * Use a temporary register instead of a caller-saved register initially.
+	 */
+	REG_FORCE_SEEN(ctx, HPPA_REG_TCC_IN_INIT);
+	emit(hppa_ldi(MAX_TAIL_CALL_CNT, HPPA_REG_TCC_IN_INIT), ctx);
+
+	/*
+	 * Skip all initializations when called as BPF TAIL call.
+	 */
+	emit(hppa_ldi(MAX_TAIL_CALL_CNT, HPPA_REG_R1), ctx);
+	emit(hppa_beq(HPPA_REG_TCC_IN_INIT, HPPA_REG_R1, 6 - HPPA_BRANCH_DISPLACEMENT), ctx);
+	emit(hppa64_bl_long(ctx->prologue_len - 3 - HPPA_BRANCH_DISPLACEMENT), ctx);
+
+	/* store entry address of this eBPF function */
+	addr = (uintptr_t) &ctx->insns[0];
+	emit(addr >> 32, ctx);
+	emit(addr & 0xffffffff, ctx);
+
+	/* store gp of this eBPF function */
+	asm("copy %%r27,%0" : "=r" (addr) );
+	emit(addr >> 32, ctx);
+	emit(addr & 0xffffffff, ctx);
+
+	/* Set up hppa stack frame. */
+	emit_hppa_copy(HPPA_REG_SP, HPPA_REG_R1, ctx);
+	emit(hppa_ldo(stack_adjust, HPPA_REG_SP, HPPA_REG_SP), ctx);
+	emit(hppa64_std_im5 (HPPA_REG_R1, -REG_SIZE, HPPA_REG_SP), ctx);
+	emit(hppa64_std_im16(HPPA_REG_RP, -2*REG_SIZE, HPPA_REG_SP), ctx);
+
+	/* Save callee-save registers. */
+	for (i = 3; i <= 15; i++) {
+		if (OPTIMIZE_HPPA && !REG_WAS_SEEN(ctx, HPPA_R(i)))
+			continue;
+		emit(hppa64_std_im16(HPPA_R(i), -REG_SIZE * i, HPPA_REG_SP), ctx);
+	}
+
+	/* load function parameters; load all if we use tail functions */
+	#define LOAD_PARAM(arg, dst) \
+		if (REG_WAS_SEEN(ctx, regmap[dst]) ||	\
+		    REG_WAS_SEEN(ctx, HPPA_REG_TCC))	\
+			emit_hppa_copy(arg, regmap[dst], ctx)
+	LOAD_PARAM(HPPA_REG_ARG0, BPF_REG_1);
+	LOAD_PARAM(HPPA_REG_ARG1, BPF_REG_2);
+	LOAD_PARAM(HPPA_REG_ARG2, BPF_REG_3);
+	LOAD_PARAM(HPPA_REG_ARG3, BPF_REG_4);
+	LOAD_PARAM(HPPA_REG_ARG4, BPF_REG_5);
+	#undef LOAD_PARAM
+
+	REG_FORCE_SEEN(ctx, HPPA_REG_T0);
+	REG_FORCE_SEEN(ctx, HPPA_REG_T1);
+	REG_FORCE_SEEN(ctx, HPPA_REG_T2);
+
+	/*
+	 * Now really set the tail call counter (TCC) register.
+	 */
+	if (REG_WAS_SEEN(ctx, HPPA_REG_TCC))
+		emit(hppa_ldi(MAX_TAIL_CALL_CNT, HPPA_REG_TCC), ctx);
+
+	/*
+	 * Save epilogue function pointer for outer TCC call chain.
+	 * The main TCC call stores the final RP on stack.
+	 */
+	addr = (uintptr_t) &ctx->insns[ctx->epilogue_offset];
+	/* skip first two instructions which jump to exit */
+	addr += 2 * HPPA_INSN_SIZE;
+	emit_imm(HPPA_REG_T2, addr, HPPA_REG_T1, ctx);
+	emit(EXIT_PTR_STORE(HPPA_REG_T2), ctx);
+
+	/* Set up BPF frame pointer. */
+	reg = regmap[BPF_REG_FP];	/* -> HPPA_REG_FP */
+	if (REG_WAS_SEEN(ctx, reg)) {
+		emit(hppa_ldo(-FRAME_SIZE, HPPA_REG_SP, reg), ctx);
+	}
+}
+
+void bpf_jit_build_epilogue(struct hppa_jit_context *ctx)
+{
+	__build_epilogue(false, ctx);
+}
+
+bool bpf_jit_supports_kfunc_call(void)
+{
+	return true;
+}
diff --git a/arch/parisc/net/bpf_jit_core.c b/arch/parisc/net/bpf_jit_core.c
new file mode 100644
index 000000000000..d6ee2fd45550
--- /dev/null
+++ b/arch/parisc/net/bpf_jit_core.c
@@ -0,0 +1,201 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Common functionality for HPPA32 and HPPA64 BPF JIT compilers
+ *
+ * Copyright (c) 2023 Helge Deller <deller@gmx.de>
+ *
+ */
+
+#include <linux/bpf.h>
+#include <linux/filter.h>
+#include "bpf_jit.h"
+
+/* Number of iterations to try until offsets converge. */
+#define NR_JIT_ITERATIONS	35
+
+static int build_body(struct hppa_jit_context *ctx, bool extra_pass, int *offset)
+{
+	const struct bpf_prog *prog = ctx->prog;
+	int i;
+
+	ctx->reg_seen_collect = true;
+	for (i = 0; i < prog->len; i++) {
+		const struct bpf_insn *insn = &prog->insnsi[i];
+		int ret;
+
+		ret = bpf_jit_emit_insn(insn, ctx, extra_pass);
+		/* BPF_LD | BPF_IMM | BPF_DW: skip the next instruction. */
+		if (ret > 0)
+			i++;
+		if (offset)
+			offset[i] = ctx->ninsns;
+		if (ret < 0)
+			return ret;
+	}
+	ctx->reg_seen_collect = false;
+	return 0;
+}
+
+bool bpf_jit_needs_zext(void)
+{
+	return true;
+}
+
+struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
+{
+	unsigned int prog_size = 0, extable_size = 0;
+	bool tmp_blinded = false, extra_pass = false;
+	struct bpf_prog *tmp, *orig_prog = prog;
+	int pass = 0, prev_ninsns = 0, prologue_len, i;
+	struct hppa_jit_data *jit_data;
+	struct hppa_jit_context *ctx;
+
+	if (!prog->jit_requested)
+		return orig_prog;
+
+	tmp = bpf_jit_blind_constants(prog);
+	if (IS_ERR(tmp))
+		return orig_prog;
+	if (tmp != prog) {
+		tmp_blinded = true;
+		prog = tmp;
+	}
+
+	jit_data = prog->aux->jit_data;
+	if (!jit_data) {
+		jit_data = kzalloc(sizeof(*jit_data), GFP_KERNEL);
+		if (!jit_data) {
+			prog = orig_prog;
+			goto out;
+		}
+		prog->aux->jit_data = jit_data;
+	}
+
+	ctx = &jit_data->ctx;
+
+	if (ctx->offset) {
+		extra_pass = true;
+		prog_size = sizeof(*ctx->insns) * ctx->ninsns;
+		goto skip_init_ctx;
+	}
+
+	ctx->prog = prog;
+	ctx->offset = kcalloc(prog->len, sizeof(int), GFP_KERNEL);
+	if (!ctx->offset) {
+		prog = orig_prog;
+		goto out_offset;
+	}
+	for (i = 0; i < prog->len; i++) {
+		prev_ninsns += 20;
+		ctx->offset[i] = prev_ninsns;
+	}
+
+	for (i = 0; i < NR_JIT_ITERATIONS; i++) {
+		pass++;
+		ctx->ninsns = 0;
+		if (build_body(ctx, extra_pass, ctx->offset)) {
+			prog = orig_prog;
+			goto out_offset;
+		}
+		ctx->body_len = ctx->ninsns;
+		bpf_jit_build_prologue(ctx);
+		ctx->prologue_len = ctx->ninsns - ctx->body_len;
+		ctx->epilogue_offset = ctx->ninsns;
+		bpf_jit_build_epilogue(ctx);
+
+		if (ctx->ninsns == prev_ninsns) {
+			if (jit_data->header)
+				break;
+			/* obtain the actual image size */
+			extable_size = prog->aux->num_exentries *
+				sizeof(struct exception_table_entry);
+			prog_size = sizeof(*ctx->insns) * ctx->ninsns;
+
+			jit_data->header =
+				bpf_jit_binary_alloc(prog_size + extable_size,
+						     &jit_data->image,
+						     sizeof(u32),
+						     bpf_fill_ill_insns);
+			if (!jit_data->header) {
+				prog = orig_prog;
+				goto out_offset;
+			}
+
+			ctx->insns = (u32 *)jit_data->image;
+			/*
+			 * Now, when the image is allocated, the image can
+			 * potentially shrink more (auipc/jalr -> jal).
+			 */
+		}
+		prev_ninsns = ctx->ninsns;
+	}
+
+	if (i == NR_JIT_ITERATIONS) {
+		pr_err("bpf-jit: image did not converge in <%d passes!\n", i);
+		if (jit_data->header)
+			bpf_jit_binary_free(jit_data->header);
+		prog = orig_prog;
+		goto out_offset;
+	}
+
+	if (extable_size)
+		prog->aux->extable = (void *)ctx->insns + prog_size;
+
+skip_init_ctx:
+	pass++;
+	ctx->ninsns = 0;
+
+	bpf_jit_build_prologue(ctx);
+	if (build_body(ctx, extra_pass, NULL)) {
+		bpf_jit_binary_free(jit_data->header);
+		prog = orig_prog;
+		goto out_offset;
+	}
+	bpf_jit_build_epilogue(ctx);
+
+	if (HPPA_JIT_DEBUG || bpf_jit_enable > 1) {
+		if (HPPA_JIT_DUMP)
+			bpf_jit_dump(prog->len, prog_size, pass, ctx->insns);
+		if (HPPA_JIT_REBOOT)
+			{ extern int machine_restart(char *); machine_restart(""); }
+	}
+
+	prog->bpf_func = (void *)ctx->insns;
+	prog->jited = 1;
+	prog->jited_len = prog_size;
+
+	bpf_flush_icache(jit_data->header, ctx->insns + ctx->ninsns);
+
+	if (!prog->is_func || extra_pass) {
+		bpf_jit_binary_lock_ro(jit_data->header);
+		prologue_len = ctx->epilogue_offset - ctx->body_len;
+		for (i = 0; i < prog->len; i++)
+			ctx->offset[i] += prologue_len;
+		bpf_prog_fill_jited_linfo(prog, ctx->offset);
+out_offset:
+		kfree(ctx->offset);
+		kfree(jit_data);
+		prog->aux->jit_data = NULL;
+	}
+out:
+	if (HPPA_JIT_REBOOT)
+		{ extern int machine_restart(char *); machine_restart(""); }
+
+	if (tmp_blinded)
+		bpf_jit_prog_release_other(prog, prog == orig_prog ?
+					   tmp : orig_prog);
+	return prog;
+}
+
+u64 hppa_div64(u64 div, u64 divisor)
+{
+	div = div64_u64(div, divisor);
+	return div;
+}
+
+u64 hppa_div64_rem(u64 div, u64 divisor)
+{
+	u64 rem;
+	div64_u64_rem(div, divisor, &rem);
+	return rem;
+}