Merge http://lia64.bkbits.net/to-linus-2.5

into home.osdl.org:/home/torvalds/v2.5/linux

arch/ia64/hp/common/sba_iommu.c

@@ -48,9 +48,11 @@
 ** This option allows cards capable of 64bit DMA to bypass the IOMMU.  If
 ** not defined, all DMA will be 32bit and go through the TLB.
 ** There's potentially a conflict in the bio merge code with us
-** advertising an iommu, but then bypassing it.  Disabled for now.
+** advertising an iommu, but then bypassing it.  Since I/O MMU bypassing
+** appears to give more performance than bio-level virtual merging, we'll
+** do the former for now.
 */
-#undef ALLOW_IOV_BYPASS
+#define ALLOW_IOV_BYPASS
 
 /*
 ** If a device prefetches beyond the end of a valid pdir entry, it will cause

arch/ia64/hp/sim/simeth.c

@@ -223,7 +223,7 @@ simeth_probe1(void)
 	dev->set_multicast_list = set_multicast_list; /* no yet used */
 
 	err = register_netdev(dev);
-	if (dev) {
+	if (err) {
 		kfree(dev);
 		return err;
 	}

arch/ia64/kernel/head.S

@@ -863,7 +863,7 @@ GLOBAL_ENTRY(ia64_spinlock_contention_pre3_4)
 .wait:
 	// exponential backoff, kdb, lockmeter etc. go in here
 	hint @pause
-	ld4.bias r30=[r31]
+	ld4 r30=[r31]		// don't use ld4.bias; if it's contended, we won't write the word
 	nop 0
 	;;
 	cmp4.eq p14,p0=r30,r0
@@ -880,7 +880,7 @@ GLOBAL_ENTRY(ia64_spinlock_contention)
 .wait:
 	// exponential backoff, kdb, lockmeter etc. go in here
 	hint @pause
-	ld4.bias r30=[r31]
+	ld4 r30=[r31]		// don't use ld4.bias; if it's contended, we won't write the word
 	;;
 	cmp4.ne p14,p0=r30,r0
 	mov r30 = 1

arch/ia64/kernel/patch.c

@@ -9,6 +9,7 @@
 
 #include <asm/patch.h>
 #include <asm/processor.h>
+#include <asm/sections.h>
 #include <asm/system.h>
 #include <asm/unistd.h>
 
@@ -176,16 +177,8 @@ patch_brl_fsys_bubble_down (unsigned long start, unsigned long end)
 void
 ia64_patch_gate (void)
 {
-	extern char __start_gate_mckinley_e9_patchlist;
-	extern char __end_gate_mckinley_e9_patchlist;
-	extern char __start_gate_vtop_patchlist;
-	extern char __end_gate_vtop_patchlist;
-	extern char __start_gate_fsyscall_patchlist;
-	extern char __end_gate_fsyscall_patchlist;
-	extern char __start_gate_brl_fsys_bubble_down_patchlist;
-	extern char __end_gate_brl_fsys_bubble_down_patchlist;
-#	define START(name)	((unsigned long) &__start_gate_##name##_patchlist)
-#	define END(name)	((unsigned long)&__end_gate_##name##_patchlist)
+#	define START(name)	((unsigned long) __start_gate_##name##_patchlist)
+#	define END(name)	((unsigned long)__end_gate_##name##_patchlist)
 
 	patch_fsyscall_table(START(fsyscall), END(fsyscall));
 	patch_brl_fsys_bubble_down(START(brl_fsys_bubble_down), END(brl_fsys_bubble_down));

arch/ia64/kernel/setup.c

@@ -41,6 +41,7 @@
 #include <asm/pgtable.h>
 #include <asm/processor.h>
 #include <asm/sal.h>
+#include <asm/sections.h>
 #include <asm/smp.h>
 #include <asm/system.h>
 #include <asm/unistd.h>
@@ -49,8 +50,6 @@
 # error "struct cpuinfo_ia64 too big!"
 #endif
 
-extern char _end;
-
 #ifdef CONFIG_SMP
 unsigned long __per_cpu_offset[NR_CPUS];
 #endif
@@ -279,7 +278,6 @@ sort_regions (struct rsvd_region *rsvd_region, int max)
 static void
 find_memory (void)
 {
-#	define KERNEL_END	(&_end)
 	unsigned long bootmap_size;
 	int n = 0;
 
@@ -300,7 +298,7 @@ find_memory (void)
 	n++;
 
 	rsvd_region[n].start = (unsigned long) ia64_imva((void *)KERNEL_START);
-	rsvd_region[n].end   = (unsigned long) ia64_imva(KERNEL_END);
+	rsvd_region[n].end   = (unsigned long) ia64_imva(_end);
 	n++;
 
 #ifdef CONFIG_BLK_DEV_INITRD
@@ -363,13 +361,12 @@ find_memory (void)
 void __init
 setup_arch (char **cmdline_p)
 {
-	extern unsigned long *__start___vtop_patchlist[], *__end____vtop_patchlist[];
 	extern unsigned long ia64_iobase;
 	unsigned long phys_iobase;
 
 	unw_init();
 
-	ia64_patch_vtop((u64) __start___vtop_patchlist, (u64) __end____vtop_patchlist);
+	ia64_patch_vtop((u64) __start___vtop_patchlist, (u64) __end___vtop_patchlist);
 
 	*cmdline_p = __va(ia64_boot_param->command_line);
 	strlcpy(saved_command_line, *cmdline_p, sizeof(saved_command_line));
@@ -390,19 +387,6 @@ setup_arch (char **cmdline_p)
 
 	find_memory();
 
-#if 0
-	/* XXX fix me */
-	init_mm.start_code = (unsigned long) &_stext;
-	init_mm.end_code = (unsigned long) &_etext;
-	init_mm.end_data = (unsigned long) &_edata;
-	init_mm.brk = (unsigned long) &_end;
-
-	code_resource.start = virt_to_bus(&_text);
-	code_resource.end = virt_to_bus(&_etext) - 1;
-	data_resource.start = virt_to_bus(&_etext);
-	data_resource.end = virt_to_bus(&_edata) - 1;
-#endif
-
 	/* process SAL system table: */
 	ia64_sal_init(efi.sal_systab);
 
@@ -687,7 +671,6 @@ get_max_cacheline_size (void)
 void
 cpu_init (void)
 {
-	extern char __per_cpu_start[], __phys_per_cpu_start[];
 	extern void __init ia64_mmu_init (void *);
 	unsigned long num_phys_stacked;
 	pal_vm_info_2_u_t vmi;
@@ -696,7 +679,6 @@ cpu_init (void)
 	void *cpu_data;
 
 #ifdef CONFIG_SMP
-	extern char __per_cpu_end[];
 	int cpu;
 
 	/*
@@ -812,9 +794,6 @@ cpu_init (void)
 void
 check_bugs (void)
 {
-	extern char __start___mckinley_e9_bundles[];
-	extern char __end___mckinley_e9_bundles[];
-
 	ia64_patch_mckinley_e9((unsigned long) __start___mckinley_e9_bundles,
 			       (unsigned long) __end___mckinley_e9_bundles);
 }

arch/ia64/kernel/time.c

@@ -23,6 +23,7 @@
 #include <asm/hw_irq.h>
 #include <asm/ptrace.h>
 #include <asm/sal.h>
+#include <asm/sections.h>
 #include <asm/system.h>
 
 extern unsigned long wall_jiffies;
@@ -41,7 +42,6 @@ static void
 do_profile (unsigned long ip)
 {
 	extern unsigned long prof_cpu_mask;
-	extern char _stext;
 
 	if (!prof_buffer)
 		return;
@@ -49,7 +49,7 @@ do_profile (unsigned long ip)
 	if (!((1UL << smp_processor_id()) & prof_cpu_mask))
 		return;
 
-	ip -= (unsigned long) &_stext;
+	ip -= (unsigned long) _stext;
 	ip >>= prof_shift;
 	/*
 	 * Don't ignore out-of-bounds IP values silently, put them into the last

arch/ia64/kernel/unwind.c

@@ -39,6 +39,7 @@
 #include <asm/ptrace.h>
 #include <asm/ptrace_offsets.h>
 #include <asm/rse.h>
+#include <asm/sections.h>
 #include <asm/system.h>
 #include <asm/uaccess.h>
 
@@ -2178,7 +2179,7 @@ __initcall(create_gate_table);
 void __init
 unw_init (void)
 {
-	extern int ia64_unw_start, ia64_unw_end, __gp;
+	extern char __gp[];
 	extern void unw_hash_index_t_is_too_narrow (void);
 	long i, off;
 
@@ -2211,8 +2212,8 @@ unw_init (void)
 	unw.lru_head = UNW_CACHE_SIZE - 1;
 	unw.lru_tail = 0;
 
-	init_unwind_table(&unw.kernel_table, "kernel", KERNEL_START, (unsigned long) &__gp,
-			  &ia64_unw_start, &ia64_unw_end);
+	init_unwind_table(&unw.kernel_table, "kernel", KERNEL_START, (unsigned long) __gp,
+			  __start_unwind, __end_unwind);
 }
 
 /*

arch/ia64/mm/init.c

@@ -27,6 +27,7 @@
 #include <asm/patch.h>
 #include <asm/pgalloc.h>
 #include <asm/sal.h>
+#include <asm/sections.h>
 #include <asm/system.h>
 #include <asm/tlb.h>
 #include <asm/uaccess.h>
@@ -34,9 +35,6 @@
 
 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 
-/* References to section boundaries: */
-extern char _stext, _etext, _edata, __init_begin, __init_end, _end;
-
 extern void ia64_tlb_init (void);
 
 unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
@@ -151,8 +149,8 @@ free_initmem (void)
 {
 	unsigned long addr, eaddr;
 
-	addr = (unsigned long) ia64_imva(&__init_begin);
-	eaddr = (unsigned long) ia64_imva(&__init_end);
+	addr = (unsigned long) ia64_imva(__init_begin);
+	eaddr = (unsigned long) ia64_imva(__init_end);
 	while (addr < eaddr) {
 		ClearPageReserved(virt_to_page(addr));
 		set_page_count(virt_to_page(addr), 1);
@@ -161,7 +159,7 @@ free_initmem (void)
 		addr += PAGE_SIZE;
 	}
 	printk(KERN_INFO "Freeing unused kernel memory: %ldkB freed\n",
-	       (&__init_end - &__init_begin) >> 10);
+	       (__init_end - __init_begin) >> 10);
 }
 
 void
@@ -308,7 +306,6 @@ static void
 setup_gate (void)
 {
 	struct page *page;
-	extern char __start_gate_section[];
 
 	/*
 	 * Map the gate page twice: once read-only to export the ELF headers etc. and once
@@ -671,7 +668,7 @@ mem_init (void)
 
 	kclist_add(&kcore_mem, __va(0), max_low_pfn * PAGE_SIZE);
 	kclist_add(&kcore_vmem, (void *)VMALLOC_START, VMALLOC_END-VMALLOC_START);
-	kclist_add(&kcore_kernel, &_stext, &_end - &_stext);
+	kclist_add(&kcore_kernel, _stext, _end - _stext);
 
 	for_each_pgdat(pgdat)
 		totalram_pages += free_all_bootmem_node(pgdat);
@@ -679,9 +676,9 @@ mem_init (void)
 	reserved_pages = 0;
 	efi_memmap_walk(count_reserved_pages, &reserved_pages);
 
-	codesize =  (unsigned long) &_etext - (unsigned long) &_stext;
-	datasize =  (unsigned long) &_edata - (unsigned long) &_etext;
-	initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
+	codesize =  (unsigned long) _etext - (unsigned long) _stext;
+	datasize =  (unsigned long) _edata - (unsigned long) _etext;
+	initsize =  (unsigned long) __init_end - (unsigned long) __init_begin;
 
 	printk(KERN_INFO "Memory: %luk/%luk available (%luk code, %luk reserved, "
 	       "%luk data, %luk init)\n", (unsigned long) nr_free_pages() << (PAGE_SHIFT - 10),

arch/ia64/scripts/toolchain-flags

@@ -20,7 +20,7 @@ warning: your linker cannot handle cross-segment segment-relative relocations.
 EOF
 fi
 
-if ! $CC -c $dir/check-model.c -o $out | grep -q 'attribute directive ignored'
+if ! $CC -c $dir/check-model.c -o $out 2>&1 | grep -q 'attribute directive ignored'
 then
     CPPFLAGS="$CPPFLAGS -DHAVE_MODEL_SMALL_ATTRIBUTE"
 fi

arch/ia64/vmlinux.lds.S

@@ -59,7 +59,7 @@ SECTIONS
 	{
 	  __start___vtop_patchlist = .;
 	  *(.data.patch.vtop)
-	  __end____vtop_patchlist = .;
+	  __end___vtop_patchlist = .;
 	}
 
   .data.patch.mckinley_e9 : AT(ADDR(.data.patch.mckinley_e9) - LOAD_OFFSET)
@@ -89,9 +89,9 @@ SECTIONS
 	{ *(.IA_64.unwind_info*) }
   .IA_64.unwind : AT(ADDR(.IA_64.unwind) - LOAD_OFFSET)
 	{
-	  ia64_unw_start = .;
+	  __start_unwind = .;
 	  *(.IA_64.unwind*)
-	  ia64_unw_end = .;
+	  __end_unwind = .;
 	}
 
   RODATA

include/asm-ia64/atomic.h

@@ -47,7 +47,7 @@ ia64_atomic_add (int i, atomic_t *v)
 }
 
 static __inline__ int
-ia64_atomic64_add (int i, atomic64_t *v)
+ia64_atomic64_add (__s64 i, atomic64_t *v)
 {
 	__s64 old, new;
 	CMPXCHG_BUGCHECK_DECL
@@ -75,7 +75,7 @@ ia64_atomic_sub (int i, atomic_t *v)
 }
 
 static __inline__ int
-ia64_atomic64_sub (int i, atomic64_t *v)
+ia64_atomic64_sub (__s64 i, atomic64_t *v)
 {
 	__s64 old, new;
 	CMPXCHG_BUGCHECK_DECL
@@ -123,7 +123,7 @@ atomic_add_negative (int i, atomic_t *v)
 }
 
 static __inline__ int
-atomic64_add_negative (int i, atomic64_t *v)
+atomic64_add_negative (__s64 i, atomic64_t *v)
 {
 	return atomic64_add_return(i, v) < 0;
 }

include/asm-ia64/io.h

@@ -413,6 +413,16 @@ extern void __ia64_memset_c_io (unsigned long, unsigned long, long);
 
 # endif /* __KERNEL__ */
 
+/*
+ * Enabling BIO_VMERGE_BOUNDARY forces us to turn off I/O MMU bypassing.  It is said that
+ * BIO-level virtual merging can give up to 4% performance boost (not verified for ia64).
+ * On the other hand, we know that I/O MMU bypassing gives ~8% performance improvement on
+ * SPECweb-like workloads on zx1-based machines.  Thus, for now we favor I/O MMU bypassing
+ * over BIO-level virtual merging.
+ */
+#if 1
+#define BIO_VMERGE_BOUNDARY	0
+#else
 /*
  * It makes no sense at all to have this BIO_VMERGE_BOUNDARY macro here.  Should be
  * replaced by dma_merge_mask() or something of that sort.  Note: the only way
@@ -425,5 +435,6 @@ extern void __ia64_memset_c_io (unsigned long, unsigned long, long);
  */
 extern unsigned long ia64_max_iommu_merge_mask;
 #define BIO_VMERGE_BOUNDARY	(ia64_max_iommu_merge_mask + 1)
+#endif
 
 #endif /* _ASM_IA64_IO_H */

include/asm-ia64/spinlock.h

@@ -22,9 +22,6 @@ typedef struct {
 #define SPIN_LOCK_UNLOCKED			(spinlock_t) { 0 }
 #define spin_lock_init(x)			((x)->lock = 0)
 
-#define NEW_LOCK
-#ifdef NEW_LOCK
-
 /*
  * Try to get the lock.  If we fail to get the lock, make a non-standard call to
  * ia64_spinlock_contention().  We do not use a normal call because that would force all
@@ -87,31 +84,6 @@ _raw_spin_lock (spinlock_t *lock)
 #endif
 }
 
-#else /* !NEW_LOCK */
-
-/*
- * Streamlined test_and_set_bit(0, (x)).  We use test-and-test-and-set
- * rather than a simple xchg to avoid writing the cache-line when
- * there is contention.
- */
-#define _raw_spin_lock(x) __asm__ __volatile__ (		\
-	"mov ar.ccv = r0\n"					\
-	"mov r29 = 1\n"						\
-	";;\n"							\
-	"1:\n"							\
-	"ld4.bias r2 = [%0]\n"					\
-	";;\n"							\
-	"cmp4.eq p0,p7 = r0,r2\n"				\
-	"(p7) br.cond.spnt.few 1b \n"				\
-	"cmpxchg4.acq r2 = [%0], r29, ar.ccv\n"			\
-	";;\n"							\
-	"cmp4.eq p0,p7 = r0, r2\n"				\
-	"(p7) br.cond.spnt.few 1b\n"				\
-	";;\n"							\
-	:: "r"(&(x)->lock) : "ar.ccv", "p7", "r2", "r29", "memory")
-
-#endif /* !NEW_LOCK */
-
 #define spin_is_locked(x)	((x)->lock != 0)
 #define _raw_spin_unlock(x)	do { barrier(); ((spinlock_t *) x)->lock = 0; } while (0)
 #define _raw_spin_trylock(x)	(cmpxchg_acq(&(x)->lock, 0, 1) == 0)