Merge branch 'parisc-5.2-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux

Pull more parisc updates from Helge Deller:
 "Two small enhancements, which I didn't included in the last pull
  request because I wanted to keep them a few more days in for-next
  before sending upstream:

   - Replace the ldcw barrier instruction by a nop instruction in the
     CAS code on uniprocessor machines.

   - Map variables read-only after init (enable ro_after_init feature)"

* 'parisc-5.2-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Use __ro_after_init in init.c
  parisc: Use __ro_after_init in unwind.c
  parisc: Use __ro_after_init in time.c
  parisc: Use __ro_after_init in processor.c
  parisc: Use __ro_after_init in process.c
  parisc: Use __ro_after_init in perf_images.h
  parisc: Use __ro_after_init in pci.c
  parisc: Use __ro_after_init in inventory.c
  parisc: Use __ro_after_init in head.S
  parisc: Use __ro_after_init in firmware.c
  parisc: Use __ro_after_init in drivers.c
  parisc: Use __ro_after_init in cache.c
  parisc: Enable the ro_after_init feature
  parisc: Drop LDCW barrier in CAS code when running UP

arch/parisc/include/asm/cache.h

@@ -24,9 +24,6 @@
 
 #define __read_mostly __attribute__((__section__(".data..read_mostly")))
 
-/* Read-only memory is marked before mark_rodata_ro() is called. */
-#define __ro_after_init	__read_mostly
-
 void parisc_cache_init(void);	/* initializes cache-flushing */
 void disable_sr_hashing_asm(int); /* low level support for above */
 void disable_sr_hashing(void);   /* turns off space register hashing */

arch/parisc/kernel/cache.c

@@ -29,9 +29,9 @@
 #include <asm/sections.h>
 #include <asm/shmparam.h>
 
-int split_tlb __read_mostly;
-int dcache_stride __read_mostly;
-int icache_stride __read_mostly;
+int split_tlb __ro_after_init;
+int dcache_stride __ro_after_init;
+int icache_stride __ro_after_init;
 EXPORT_SYMBOL(dcache_stride);
 
 void flush_dcache_page_asm(unsigned long phys_addr, unsigned long vaddr);
@@ -51,12 +51,12 @@ DEFINE_SPINLOCK(pa_tlb_flush_lock);
 DEFINE_SPINLOCK(pa_swapper_pg_lock);
 
 #if defined(CONFIG_64BIT) && defined(CONFIG_SMP)
-int pa_serialize_tlb_flushes __read_mostly;
+int pa_serialize_tlb_flushes __ro_after_init;
 #endif
 
-struct pdc_cache_info cache_info __read_mostly;
+struct pdc_cache_info cache_info __ro_after_init;
 #ifndef CONFIG_PA20
-static struct pdc_btlb_info btlb_info __read_mostly;
+static struct pdc_btlb_info btlb_info __ro_after_init;
 #endif
 
 #ifdef CONFIG_SMP
@@ -381,10 +381,10 @@ EXPORT_SYMBOL(flush_data_cache_local);
 EXPORT_SYMBOL(flush_kernel_icache_range_asm);
 
 #define FLUSH_THRESHOLD 0x80000 /* 0.5MB */
-static unsigned long parisc_cache_flush_threshold __read_mostly = FLUSH_THRESHOLD;
+static unsigned long parisc_cache_flush_threshold __ro_after_init = FLUSH_THRESHOLD;
 
 #define FLUSH_TLB_THRESHOLD (16*1024) /* 16 KiB minimum TLB threshold */
-static unsigned long parisc_tlb_flush_threshold __read_mostly = FLUSH_TLB_THRESHOLD;
+static unsigned long parisc_tlb_flush_threshold __ro_after_init = FLUSH_TLB_THRESHOLD;
 
 void __init parisc_setup_cache_timing(void)
 {

arch/parisc/kernel/drivers.c

@@ -41,7 +41,7 @@
 #include <asm/ropes.h>
 
 /* See comments in include/asm-parisc/pci.h */
-const struct dma_map_ops *hppa_dma_ops __read_mostly;
+const struct dma_map_ops *hppa_dma_ops __ro_after_init;
 EXPORT_SYMBOL(hppa_dma_ops);
 
 static struct device root = {

arch/parisc/kernel/firmware.c

@@ -87,7 +87,7 @@ extern unsigned long pdc_result2[NUM_PDC_RESULT];
 
 /* Firmware needs to be initially set to narrow to determine the 
  * actual firmware width. */
-int parisc_narrow_firmware __read_mostly = 1;
+int parisc_narrow_firmware __ro_after_init = 1;
 #endif
 
 /* On most currently-supported platforms, IODC I/O calls are 32-bit calls

arch/parisc/kernel/head.S

@@ -376,7 +376,7 @@ smp_slave_stext:
 ENDPROC(parisc_kernel_start)
 
 #ifndef CONFIG_64BIT
-	.section .data..read_mostly
+	.section .data..ro_after_init
 
 	.align	4
 	.export	$global$,data

arch/parisc/kernel/inventory.c

@@ -39,12 +39,12 @@
 */
 #undef DEBUG_PAT
 
-int pdc_type __read_mostly = PDC_TYPE_ILLEGAL;
+int pdc_type __ro_after_init = PDC_TYPE_ILLEGAL;
 
 /* cell number and location (PAT firmware only) */
-unsigned long parisc_cell_num __read_mostly;
-unsigned long parisc_cell_loc __read_mostly;
-unsigned long parisc_pat_pdc_cap __read_mostly;
+unsigned long parisc_cell_num __ro_after_init;
+unsigned long parisc_cell_loc __ro_after_init;
+unsigned long parisc_pat_pdc_cap __ro_after_init;
 
 
 void __init setup_pdc(void)

arch/parisc/kernel/pci.c

@@ -45,14 +45,14 @@
  * #define pci_post_reset_delay 50
  */
 
-struct pci_port_ops *pci_port __read_mostly;
-struct pci_bios_ops *pci_bios __read_mostly;
+struct pci_port_ops *pci_port __ro_after_init;
+struct pci_bios_ops *pci_bios __ro_after_init;
 
-static int pci_hba_count __read_mostly;
+static int pci_hba_count __ro_after_init;
 
 /* parisc_pci_hba used by pci_port->in/out() ops to lookup bus data.  */
 #define PCI_HBA_MAX 32
-static struct pci_hba_data *parisc_pci_hba[PCI_HBA_MAX] __read_mostly;
+static struct pci_hba_data *parisc_pci_hba[PCI_HBA_MAX] __ro_after_init;
 
 
 /********************************************************************

arch/parisc/kernel/perf_images.h

@@ -25,7 +25,7 @@
 
 #define PCXU_IMAGE_SIZE 584
 
-static uint32_t onyx_images[][PCXU_IMAGE_SIZE/sizeof(uint32_t)] __read_mostly = {
+static uint32_t onyx_images[][PCXU_IMAGE_SIZE/sizeof(uint32_t)] __ro_after_init = {
 /*
  * CPI:
  *
@@ -2093,7 +2093,7 @@ static uint32_t onyx_images[][PCXU_IMAGE_SIZE/sizeof(uint32_t)] __read_mostly =
 };
 #define PCXW_IMAGE_SIZE 576
 
-static uint32_t cuda_images[][PCXW_IMAGE_SIZE/sizeof(uint32_t)] __read_mostly = {
+static uint32_t cuda_images[][PCXW_IMAGE_SIZE/sizeof(uint32_t)] __ro_after_init = {
 /*
  * CPI:     FROM CPI.IDF (Image 0)
  *

arch/parisc/kernel/process.c

@@ -192,7 +192,7 @@ int dump_task_fpu (struct task_struct *tsk, elf_fpregset_t *r)
  * QEMU idle the host too.
  */
 
-int running_on_qemu __read_mostly;
+int running_on_qemu __ro_after_init;
 EXPORT_SYMBOL(running_on_qemu);
 
 void __cpuidle arch_cpu_idle_dead(void)

arch/parisc/kernel/processor.c

@@ -43,10 +43,10 @@
 #include <asm/irq.h>		/* for struct irq_region */
 #include <asm/parisc-device.h>
 
-struct system_cpuinfo_parisc boot_cpu_data __read_mostly;
+struct system_cpuinfo_parisc boot_cpu_data __ro_after_init;
 EXPORT_SYMBOL(boot_cpu_data);
 #ifdef CONFIG_PA8X00
-int _parisc_requires_coherency __read_mostly;
+int _parisc_requires_coherency __ro_after_init;
 EXPORT_SYMBOL(_parisc_requires_coherency);
 #endif
 

arch/parisc/kernel/syscall.S

@@ -641,7 +641,8 @@ cas_action:
 2:	stw	%r24, 0(%r26)
 	/* Free lock */
 #ifdef CONFIG_SMP
-	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+98:	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+99:	ALTERNATIVE(98b, 99b, ALT_COND_NO_SMP, INSN_NOP)
 #endif
 	stw	%r20, 0(%sr2,%r20)
 #if ENABLE_LWS_DEBUG
@@ -658,7 +659,8 @@ cas_action:
 	/* Error occurred on load or store */
 	/* Free lock */
 #ifdef CONFIG_SMP
-	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+98:	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+99:	ALTERNATIVE(98b, 99b, ALT_COND_NO_SMP, INSN_NOP)
 #endif
 	stw	%r20, 0(%sr2,%r20)
 #if ENABLE_LWS_DEBUG
@@ -862,7 +864,8 @@ cas2_action:
 cas2_end:
 	/* Free lock */
 #ifdef CONFIG_SMP
-	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+98:	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+99:	ALTERNATIVE(98b, 99b, ALT_COND_NO_SMP, INSN_NOP)
 #endif
 	stw	%r20, 0(%sr2,%r20)
 	/* Enable interrupts */
@@ -875,7 +878,8 @@ cas2_end:
 	/* Error occurred on load or store */
 	/* Free lock */
 #ifdef CONFIG_SMP
-	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+98:	LDCW	0(%sr2,%r20), %r1			/* Barrier */
+99:	ALTERNATIVE(98b, 99b, ALT_COND_NO_SMP, INSN_NOP)
 #endif
 	stw	%r20, 0(%sr2,%r20)
 	ssm	PSW_SM_I, %r0

arch/parisc/kernel/time.c

@@ -40,7 +40,7 @@
 
 #include <linux/timex.h>
 
-static unsigned long clocktick __read_mostly;	/* timer cycles per tick */
+static unsigned long clocktick __ro_after_init;	/* timer cycles per tick */
 
 /*
  * We keep time on PA-RISC Linux by using the Interval Timer which is

arch/parisc/kernel/unwind.c

@@ -40,7 +40,7 @@ static DEFINE_SPINLOCK(unwind_lock);
  * we can call unwind_init as early in the bootup process as 
  * possible (before the slab allocator is initialized)
  */
-static struct unwind_table kernel_unwind_table __read_mostly;
+static struct unwind_table kernel_unwind_table __ro_after_init;
 static LIST_HEAD(unwind_tables);
 
 static inline const struct unwind_table_entry *

arch/parisc/kernel/vmlinux.lds.S

@@ -18,9 +18,6 @@
 				*(.data..vm0.pgd) \
 				*(.data..vm0.pte)
 
-/* No __ro_after_init data in the .rodata section - which will always be ro */
-#define RO_AFTER_INIT_DATA
-
 #include <asm-generic/vmlinux.lds.h>
 
 /* needed for the processor specific cache alignment size */	

arch/parisc/mm/init.c

@@ -66,7 +66,7 @@ static struct resource pdcdata_resource = {
 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM,
 };
 
-static struct resource sysram_resources[MAX_PHYSMEM_RANGES] __read_mostly;
+static struct resource sysram_resources[MAX_PHYSMEM_RANGES] __ro_after_init;
 
 /* The following array is initialized from the firmware specific
  * information retrieved in kernel/inventory.c.
@@ -345,16 +345,7 @@ static void __init setup_bootmem(void)
 	memblock_dump_all();
 }
 
-static int __init parisc_text_address(unsigned long vaddr)
-{
-	static unsigned long head_ptr __initdata;
-
-	if (!head_ptr)
-		head_ptr = PAGE_MASK & (unsigned long)
-			dereference_function_descriptor(&parisc_kernel_start);
-
-	return core_kernel_text(vaddr) || vaddr == head_ptr;
-}
+static bool kernel_set_to_readonly;
 
 static void __init map_pages(unsigned long start_vaddr,
 			     unsigned long start_paddr, unsigned long size,
@@ -372,10 +363,11 @@ static void __init map_pages(unsigned long start_vaddr,
 	unsigned long vaddr;
 	unsigned long ro_start;
 	unsigned long ro_end;
-	unsigned long kernel_end;
+	unsigned long kernel_start, kernel_end;
 
 	ro_start = __pa((unsigned long)_text);
 	ro_end   = __pa((unsigned long)&data_start);
+	kernel_start = __pa((unsigned long)&__init_begin);
 	kernel_end  = __pa((unsigned long)&_end);
 
 	end_paddr = start_paddr + size;
@@ -438,26 +430,30 @@ static void __init map_pages(unsigned long start_vaddr,
 			pg_table = (pte_t *) __va(pg_table) + start_pte;
 			for (tmp2 = start_pte; tmp2 < PTRS_PER_PTE; tmp2++, pg_table++) {
 				pte_t pte;
-
-				if (force)
-					pte =  __mk_pte(address, pgprot);
-				else if (parisc_text_address(vaddr)) {
-					pte = __mk_pte(address, PAGE_KERNEL_EXEC);
-					if (address >= ro_start && address < kernel_end)
-						pte = pte_mkhuge(pte);
+				pgprot_t prot;
+				bool huge = false;
+
+				if (force) {
+					prot = pgprot;
+				} else if (address < kernel_start || address >= kernel_end) {
+					/* outside kernel memory */
+					prot = PAGE_KERNEL;
+				} else if (!kernel_set_to_readonly) {
+					/* still initializing, allow writing to RO memory */
+					prot = PAGE_KERNEL_RWX;
+					huge = true;
+				} else if (address >= ro_start) {
+					/* Code (ro) and Data areas */
+					prot = (address < ro_end) ?
+						PAGE_KERNEL_EXEC : PAGE_KERNEL;
+					huge = true;
+				} else {
+					prot = PAGE_KERNEL;
 				}
-				else
-#if defined(CONFIG_PARISC_PAGE_SIZE_4KB)
-				if (address >= ro_start && address < ro_end) {
-					pte = __mk_pte(address, PAGE_KERNEL_EXEC);
+
+				pte = __mk_pte(address, prot);
+				if (huge)
 					pte = pte_mkhuge(pte);
-				} else
-#endif
-				{
-					pte = __mk_pte(address, pgprot);
-					if (address >= ro_start && address < kernel_end)
-						pte = pte_mkhuge(pte);
-				}
 
 				if (address >= end_paddr)
 					break;
@@ -493,6 +489,12 @@ void __ref free_initmem(void)
 {
 	unsigned long init_begin = (unsigned long)__init_begin;
 	unsigned long init_end = (unsigned long)__init_end;
+	unsigned long kernel_end  = (unsigned long)&_end;
+
+	/* Remap kernel text and data, but do not touch init section yet. */
+	kernel_set_to_readonly = true;
+	map_pages(init_end, __pa(init_end), kernel_end - init_end,
+		  PAGE_KERNEL, 0);
 
 	/* The init text pages are marked R-X.  We have to
 	 * flush the icache and mark them RW-
@@ -509,7 +511,7 @@ void __ref free_initmem(void)
 		  PAGE_KERNEL, 1);
 
 	/* force the kernel to see the new TLB entries */
-	__flush_tlb_range(0, init_begin, init_end);
+	__flush_tlb_range(0, init_begin, kernel_end);
 
 	/* finally dump all the instructions which were cached, since the
 	 * pages are no-longer executable */
@@ -527,8 +529,9 @@ void mark_rodata_ro(void)
 {
 	/* rodata memory was already mapped with KERNEL_RO access rights by
            pagetable_init() and map_pages(). No need to do additional stuff here */
-	printk (KERN_INFO "Write protecting the kernel read-only data: %luk\n",
-		(unsigned long)(__end_rodata - __start_rodata) >> 10);
+	unsigned long roai_size = __end_ro_after_init - __start_ro_after_init;
+
+	pr_info("Write protected read-only-after-init data: %luk\n", roai_size >> 10);
 }
 #endif
 
@@ -554,11 +557,11 @@ void mark_rodata_ro(void)
 #define SET_MAP_OFFSET(x) ((void *)(((unsigned long)(x) + VM_MAP_OFFSET) \
 				     & ~(VM_MAP_OFFSET-1)))
 
-void *parisc_vmalloc_start __read_mostly;
+void *parisc_vmalloc_start __ro_after_init;
 EXPORT_SYMBOL(parisc_vmalloc_start);
 
 #ifdef CONFIG_PA11
-unsigned long pcxl_dma_start __read_mostly;
+unsigned long pcxl_dma_start __ro_after_init;
 #endif
 
 void __init mem_init(void)
@@ -632,7 +635,7 @@ void __init mem_init(void)
 #endif
 }
 
-unsigned long *empty_zero_page __read_mostly;
+unsigned long *empty_zero_page __ro_after_init;
 EXPORT_SYMBOL(empty_zero_page);
 
 /*