Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/…

…git/x86/linux-2.6-tip * 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-tip: x86: Fix double enable_IR_x2apic() call on SMP kernel on !SMP boards x86: Increase CONFIG_NODES_SHIFT max to 10 ibft, x86: Change reserve_ibft_region() to find_ibft_region() x86, hpet: Fix bug in RTC emulation x86, hpet: Erratum workaround for read after write of HPET comparator bootmem, x86: Fix 32bit numa system without RAM on node 0 nobootmem, x86: Fix 32bit numa system without RAM on node 0 x86: Handle overlapping mptables x86: Make e820_remove_range to handle all covered case x86-32, resume: do a global tlb flush in S4 resume

Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/…
…git/x86/linux-2.6-tip * 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-tip: x86: Fix double enable_IR_x2apic() call on SMP kernel on !SMP boards x86: Increase CONFIG_NODES_SHIFT max to 10 ibft, x86: Change reserve_ibft_region() to find_ibft_region() x86, hpet: Fix bug in RTC emulation x86, hpet: Erratum workaround for read after write of HPET comparator bootmem, x86: Fix 32bit numa system without RAM on node 0 nobootmem, x86: Fix 32bit numa system without RAM on node 0 x86: Handle overlapping mptables x86: Make e820_remove_range to handle all covered case x86-32, resume: do a global tlb flush in S4 resume
Linus Torvalds
2 parents addb2d6c13 472a474c66
Showing 10 changed files Side-by-side Diff
arch/x86/Kconfig
arch/x86/kernel/apic/apic.c
arch/x86/kernel/e820.c
arch/x86/kernel/hpet.c
arch/x86/kernel/mpparse.c
arch/x86/kernel/setup.c
arch/x86/power/hibernate_asm_32.S
drivers/firmware/iscsi_ibft_find.c
include/linux/iscsi_ibft.h
mm/bootmem.c
@@ -1216,8 +1216,8 @@
  
 config NODES_SHIFT
 	int "Maximum NUMA Nodes (as a power of 2)" if !MAXSMP
-	range 1 9
-	default "9" if MAXSMP
+	range 1 10
+	default "10" if MAXSMP
 	default "6" if X86_64
 	default "4" if X86_NUMAQ
 	default "3"
@@ -1640,8 +1640,10 @@
 	}
 #endif
  
+#ifndef CONFIG_SMP
 	enable_IR_x2apic();
 	default_setup_apic_routing();
+#endif
  
 	verify_local_APIC();
 	connect_bsp_APIC();
@@ -519,29 +519,45 @@
 	printk(KERN_DEBUG "e820 remove range: %016Lx - %016Lx ",
 		       (unsigned long long) start,
 		       (unsigned long long) end);
-	e820_print_type(old_type);
+	if (checktype)
+		e820_print_type(old_type);
 	printk(KERN_CONT "\n");
  
 	for (i = 0; i < e820.nr_map; i++) {
 		struct e820entry *ei = &e820.map[i];
 		u64 final_start, final_end;
+		u64 ei_end;
  
 		if (checktype && ei->type != old_type)
 			continue;
+
+		ei_end = ei->addr + ei->size;
 		/* totally covered? */
-		if (ei->addr >= start &&
-		    (ei->addr + ei->size) <= (start + size)) {
+		if (ei->addr >= start && ei_end <= end) {
 			real_removed_size += ei->size;
 			memset(ei, 0, sizeof(struct e820entry));
 			continue;
 		}
+
+		/* new range is totally covered? */
+		if (ei->addr < start && ei_end > end) {
+			e820_add_region(end, ei_end - end, ei->type);
+			ei->size = start - ei->addr;
+			real_removed_size += size;
+			continue;
+		}
+
 		/* partially covered */
 		final_start = max(start, ei->addr);
-		final_end = min(start + size, ei->addr + ei->size);
+		final_end = min(end, ei_end);
 		if (final_start >= final_end)
 			continue;
 		real_removed_size += final_end - final_start;
  
+		/*
+		 * left range could be head or tail, so need to update
+		 * size at first.
+		 */
 		ei->size -= final_end - final_start;
 		if (ei->addr < final_start)
 			continue;
@@ -400,9 +400,15 @@
 	 * then we might have a real hardware problem. We can not do
 	 * much about it here, but at least alert the user/admin with
 	 * a prominent warning.
+	 * An erratum on some chipsets (ICH9,..), results in comparator read
+	 * immediately following a write returning old value. Workaround
+	 * for this is to read this value second time, when first
+	 * read returns old value.
 	 */
-	WARN_ONCE(hpet_readl(HPET_Tn_CMP(timer)) != cnt,
+	if (unlikely((u32)hpet_readl(HPET_Tn_CMP(timer)) != cnt)) {
+		WARN_ONCE(hpet_readl(HPET_Tn_CMP(timer)) != cnt,
 		  KERN_WARNING "hpet: compare register read back failed.\n");
+	}
  
 	return (s32)(hpet_readl(HPET_COUNTER) - cnt) >= 0 ? -ETIME : 0;
 }
@@ -1144,6 +1150,7 @@
 		do_div(clc, freq);
 		clc >>= hpet_clockevent.shift;
 		hpet_pie_delta = clc;
+		hpet_pie_limit = 0;
 	}
 	return 1;
 }
@@ -664,7 +664,7 @@
 {
 	unsigned long size = get_mpc_size(mpf->physptr);
  
-	reserve_early(mpf->physptr, mpf->physptr+size, "MP-table mpc");
+	reserve_early_overlap_ok(mpf->physptr, mpf->physptr+size, "MP-table mpc");
 }
  
 static int __init smp_scan_config(unsigned long base, unsigned long length)
@@ -693,7 +693,7 @@
 			       mpf, (u64)virt_to_phys(mpf));
  
 			mem = virt_to_phys(mpf);
-			reserve_early(mem, mem + sizeof(*mpf), "MP-table mpf");
+			reserve_early_overlap_ok(mem, mem + sizeof(*mpf), "MP-table mpf");
 			if (mpf->physptr)
 				smp_reserve_memory(mpf);
  
@@ -607,6 +607,16 @@
 early_param("elfcorehdr", setup_elfcorehdr);
 #endif
  
+static __init void reserve_ibft_region(void)
+{
+	unsigned long addr, size = 0;
+
+	addr = find_ibft_region(&size);
+
+	if (size)
+		reserve_early_overlap_ok(addr, addr + size, "ibft");
+}
+
 #ifdef CONFIG_X86_RESERVE_LOW_64K
 static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
 {
@@ -909,6 +919,8 @@
 	 */
 	find_smp_config();
  
+	reserve_ibft_region();
+
 	reserve_trampoline_memory();
  
 #ifdef CONFIG_ACPI_SLEEP
@@ -975,8 +987,6 @@
 #endif
  
 	dma32_reserve_bootmem();
-
-	reserve_ibft_region();
  
 #ifdef CONFIG_KVM_CLOCK
 	kvmclock_init();
@@ -27,10 +27,17 @@
 	ret
  
 ENTRY(restore_image)
+	movl	mmu_cr4_features, %ecx
 	movl	resume_pg_dir, %eax
 	subl	$__PAGE_OFFSET, %eax
 	movl	%eax, %cr3
  
+	jecxz	1f	# cr4 Pentium and higher, skip if zero
+	andl	$~(X86_CR4_PGE), %ecx
+	movl	%ecx, %cr4;  # turn off PGE
+	movl	%cr3, %eax;  # flush TLB
+	movl	%eax, %cr3
+1:
 	movl	restore_pblist, %edx
 	.p2align 4,,7
  
  
@@ -54,15 +61,7 @@
 	movl	$swapper_pg_dir, %eax
 	subl	$__PAGE_OFFSET, %eax
 	movl	%eax, %cr3
-	/* Flush TLB, including "global" things (vmalloc) */
 	movl	mmu_cr4_features, %ecx
-	jecxz	1f	# cr4 Pentium and higher, skip if zero
-	movl	%ecx, %edx
-	andl	$~(X86_CR4_PGE), %edx
-	movl	%edx, %cr4;  # turn off PGE
-1:
-	movl	%cr3, %eax;  # flush TLB
-	movl	%eax, %cr3
 	jecxz	1f	# cr4 Pentium and higher, skip if zero
 	movl	%ecx, %cr4;  # turn PGE back on
 1:
@@ -51,7 +51,7 @@
  * Routine used to find the iSCSI Boot Format Table. The logical
  * kernel address is set in the ibft_addr global variable.
  */
-void __init reserve_ibft_region(void)
+unsigned long __init find_ibft_region(unsigned long *sizep)
 {
 	unsigned long pos;
 	unsigned int len = 0;
@@ -77,7 +77,12 @@
 			}
 		}
 	}
-	if (ibft_addr)
-		reserve_bootmem(pos, PAGE_ALIGN(len), BOOTMEM_DEFAULT);
+	if (ibft_addr) {
+		*sizep = PAGE_ALIGN(len);
+		return pos;
+	}
+
+	*sizep = 0;
+	return 0;
 }
@@ -42,9 +42,13 @@
  * mapped address is set in the ibft_addr variable.
  */
 #ifdef CONFIG_ISCSI_IBFT_FIND
-extern void __init reserve_ibft_region(void);
+unsigned long find_ibft_region(unsigned long *sizep);
 #else
-static inline void reserve_ibft_region(void) { }
+static inline unsigned long find_ibft_region(unsigned long *sizep)
+{
+	*sizep = 0;
+	return 0;
+}
 #endif
  
 #endif /* ISCSI_IBFT_H */
@@ -304,9 +304,22 @@
 unsigned long __init free_all_bootmem(void)
 {
 #ifdef CONFIG_NO_BOOTMEM
-	return free_all_memory_core_early(NODE_DATA(0)->node_id);
+	/*
+	 * We need to use MAX_NUMNODES instead of NODE_DATA(0)->node_id
+	 *  because in some case like Node0 doesnt have RAM installed
+	 *  low ram will be on Node1
+	 * Use MAX_NUMNODES will make sure all ranges in early_node_map[]
+	 *  will be used instead of only Node0 related
+	 */
+	return free_all_memory_core_early(MAX_NUMNODES);
 #else
-	return free_all_bootmem_core(NODE_DATA(0)->bdata);
+	unsigned long total_pages = 0;
+	bootmem_data_t *bdata;
+
+	list_for_each_entry(bdata, &bdata_list, list)
+		total_pages += free_all_bootmem_core(bdata);
+
+	return total_pages;
 #endif
 }
...	...	@@ -1216,8 +1216,8 @@
1216	1216
1217	1217	config NODES_SHIFT
1218	1218	int "Maximum NUMA Nodes (as a power of 2)" if !MAXSMP
1219		- range 1 9
1220		- default "9" if MAXSMP
	1219	+ range 1 10
	1220	+ default "10" if MAXSMP
1221	1221	default "6" if X86_64
1222	1222	default "4" if X86_NUMAQ
1223	1223	default "3"
...	...	@@ -1640,8 +1640,10 @@
1640	1640	}
1641	1641	#endif
1642	1642
	1643	+#ifndef CONFIG_SMP
1643	1644	enable_IR_x2apic();
1644	1645	default_setup_apic_routing();
	1646	+#endif
1645	1647
1646	1648	verify_local_APIC();
1647	1649	connect_bsp_APIC();
...	...	@@ -519,29 +519,45 @@
519	519	printk(KERN_DEBUG "e820 remove range: %016Lx - %016Lx ",
520	520	(unsigned long long) start,
521	521	(unsigned long long) end);
522		- e820_print_type(old_type);
	522	+ if (checktype)
	523	+ e820_print_type(old_type);
523	524	printk(KERN_CONT "\n");
524	525
525	526	for (i = 0; i < e820.nr_map; i++) {
526	527	struct e820entry *ei = &e820.map[i];
527	528	u64 final_start, final_end;
	529	+ u64 ei_end;
528	530
529	531	if (checktype && ei->type != old_type)
530	532	continue;
	533	+
	534	+ ei_end = ei->addr + ei->size;
531	535	/* totally covered? */
532		- if (ei->addr >= start &&
533		- (ei->addr + ei->size) <= (start + size)) {
	536	+ if (ei->addr >= start && ei_end <= end) {
534	537	real_removed_size += ei->size;
535	538	memset(ei, 0, sizeof(struct e820entry));
536	539	continue;
537	540	}
	541	+
	542	+ /* new range is totally covered? */
	543	+ if (ei->addr < start && ei_end > end) {
	544	+ e820_add_region(end, ei_end - end, ei->type);
	545	+ ei->size = start - ei->addr;
	546	+ real_removed_size += size;
	547	+ continue;
	548	+ }
	549	+
538	550	/* partially covered */
539	551	final_start = max(start, ei->addr);
540		- final_end = min(start + size, ei->addr + ei->size);
	552	+ final_end = min(end, ei_end);
541	553	if (final_start >= final_end)
542	554	continue;
543	555	real_removed_size += final_end - final_start;
544	556
	557	+ /*
	558	+ * left range could be head or tail, so need to update
	559	+ * size at first.
	560	+ */
545	561	ei->size -= final_end - final_start;
546	562	if (ei->addr < final_start)
547	563	continue;
...	...	@@ -400,9 +400,15 @@
400	400	* then we might have a real hardware problem. We can not do
401	401	* much about it here, but at least alert the user/admin with
402	402	* a prominent warning.
	403	+ * An erratum on some chipsets (ICH9,..), results in comparator read
	404	+ * immediately following a write returning old value. Workaround
	405	+ * for this is to read this value second time, when first
	406	+ * read returns old value.
403	407	*/
404		- WARN_ONCE(hpet_readl(HPET_Tn_CMP(timer)) != cnt,
	408	+ if (unlikely((u32)hpet_readl(HPET_Tn_CMP(timer)) != cnt)) {
	409	+ WARN_ONCE(hpet_readl(HPET_Tn_CMP(timer)) != cnt,
405	410	KERN_WARNING "hpet: compare register read back failed.\n");
	411	+ }
406	412
407	413	return (s32)(hpet_readl(HPET_COUNTER) - cnt) >= 0 ? -ETIME : 0;
408	414	}
...	...	@@ -1144,6 +1150,7 @@
1144	1150	do_div(clc, freq);
1145	1151	clc >>= hpet_clockevent.shift;
1146	1152	hpet_pie_delta = clc;
	1153	+ hpet_pie_limit = 0;
1147	1154	}
1148	1155	return 1;
1149	1156	}
...	...	@@ -664,7 +664,7 @@
664	664	{
665	665	unsigned long size = get_mpc_size(mpf->physptr);
666	666
667		- reserve_early(mpf->physptr, mpf->physptr+size, "MP-table mpc");
	667	+ reserve_early_overlap_ok(mpf->physptr, mpf->physptr+size, "MP-table mpc");
668	668	}
669	669
670	670	static int __init smp_scan_config(unsigned long base, unsigned long length)
...	...	@@ -693,7 +693,7 @@
693	693	mpf, (u64)virt_to_phys(mpf));
694	694
695	695	mem = virt_to_phys(mpf);
696		- reserve_early(mem, mem + sizeof(*mpf), "MP-table mpf");
	696	+ reserve_early_overlap_ok(mem, mem + sizeof(*mpf), "MP-table mpf");
697	697	if (mpf->physptr)
698	698	smp_reserve_memory(mpf);
699	699
...	...	@@ -607,6 +607,16 @@
607	607	early_param("elfcorehdr", setup_elfcorehdr);
608	608	#endif
609	609
	610	+static __init void reserve_ibft_region(void)
	611	+{
	612	+ unsigned long addr, size = 0;
	613	+
	614	+ addr = find_ibft_region(&size);
	615	+
	616	+ if (size)
	617	+ reserve_early_overlap_ok(addr, addr + size, "ibft");
	618	+}
	619	+
610	620	#ifdef CONFIG_X86_RESERVE_LOW_64K
611	621	static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
612	622	{
...	...	@@ -909,6 +919,8 @@
909	919	*/
910	920	find_smp_config();
911	921
	922	+ reserve_ibft_region();
	923	+
912	924	reserve_trampoline_memory();
913	925
914	926	#ifdef CONFIG_ACPI_SLEEP
...	...	@@ -975,8 +987,6 @@
975	987	#endif
976	988
977	989	dma32_reserve_bootmem();
978		-
979		- reserve_ibft_region();
980	990
981	991	#ifdef CONFIG_KVM_CLOCK
982	992	kvmclock_init();
...	...	@@ -27,10 +27,17 @@
27	27	ret
28	28
29	29	ENTRY(restore_image)
	30	+ movl mmu_cr4_features, %ecx
30	31	movl resume_pg_dir, %eax
31	32	subl $__PAGE_OFFSET, %eax
32	33	movl %eax, %cr3
33	34
	35	+ jecxz 1f # cr4 Pentium and higher, skip if zero
	36	+ andl $~(X86_CR4_PGE), %ecx
	37	+ movl %ecx, %cr4; # turn off PGE
	38	+ movl %cr3, %eax; # flush TLB
	39	+ movl %eax, %cr3
	40	+1:
34	41	movl restore_pblist, %edx
35	42	.p2align 4,,7
36	43
37	44
...	...	@@ -54,15 +61,7 @@
54	61	movl $swapper_pg_dir, %eax
55	62	subl $__PAGE_OFFSET, %eax
56	63	movl %eax, %cr3
57		- /* Flush TLB, including "global" things (vmalloc) */
58	64	movl mmu_cr4_features, %ecx
59		- jecxz 1f # cr4 Pentium and higher, skip if zero
60		- movl %ecx, %edx
61		- andl $~(X86_CR4_PGE), %edx
62		- movl %edx, %cr4; # turn off PGE
63		-1:
64		- movl %cr3, %eax; # flush TLB
65		- movl %eax, %cr3
66	65	jecxz 1f # cr4 Pentium and higher, skip if zero
67	66	movl %ecx, %cr4; # turn PGE back on
68	67	1:
...	...	@@ -51,7 +51,7 @@
51	51	* Routine used to find the iSCSI Boot Format Table. The logical
52	52	* kernel address is set in the ibft_addr global variable.
53	53	*/
54		-void __init reserve_ibft_region(void)
	54	+unsigned long __init find_ibft_region(unsigned long *sizep)
55	55	{
56	56	unsigned long pos;
57	57	unsigned int len = 0;
...	...	@@ -77,7 +77,12 @@
77	77	}
78	78	}
79	79	}
80		- if (ibft_addr)
81		- reserve_bootmem(pos, PAGE_ALIGN(len), BOOTMEM_DEFAULT);
	80	+ if (ibft_addr) {
	81	+ *sizep = PAGE_ALIGN(len);
	82	+ return pos;
	83	+ }
	84	+
	85	+ *sizep = 0;
	86	+ return 0;
82	87	}
...	...	@@ -42,9 +42,13 @@
42	42	* mapped address is set in the ibft_addr variable.
43	43	*/
44	44	#ifdef CONFIG_ISCSI_IBFT_FIND
45		-extern void __init reserve_ibft_region(void);
	45	+unsigned long find_ibft_region(unsigned long *sizep);
46	46	#else
47		-static inline void reserve_ibft_region(void) { }
	47	+static inline unsigned long find_ibft_region(unsigned long *sizep)
	48	+{
	49	+ *sizep = 0;
	50	+ return 0;
	51	+}
48	52	#endif
49	53
50	54	#endif /* ISCSI_IBFT_H */
...	...	@@ -304,9 +304,22 @@
304	304	unsigned long __init free_all_bootmem(void)
305	305	{
306	306	#ifdef CONFIG_NO_BOOTMEM
307		- return free_all_memory_core_early(NODE_DATA(0)->node_id);
	307	+ /*
	308	+ * We need to use MAX_NUMNODES instead of NODE_DATA(0)->node_id
	309	+ * because in some case like Node0 doesnt have RAM installed
	310	+ * low ram will be on Node1
	311	+ * Use MAX_NUMNODES will make sure all ranges in early_node_map[]
	312	+ * will be used instead of only Node0 related
	313	+ */
	314	+ return free_all_memory_core_early(MAX_NUMNODES);
308	315	#else
309		- return free_all_bootmem_core(NODE_DATA(0)->bdata);
	316	+ unsigned long total_pages = 0;
	317	+ bootmem_data_t *bdata;
	318	+
	319	+ list_for_each_entry(bdata, &bdata_list, list)
	320	+ total_pages += free_all_bootmem_core(bdata);
	321	+
	322	+ return total_pages;
310	323	#endif
311	324	}
312	325