[PATCH] x86: fix EFI memory map parsing

The memory descriptors that comprise the EFI memory map are not fixed in stone such that the size could change in the future. This uses the memory descriptor size obtained from EFI to iterate over the memory map entries during boot. This enables the removal of an x86 specific pad (and ifdef) in the EFI header. I also couldn't stomach the broken up nature of the function to put EFI runtime calls into virtual mode any longer so I fixed that up a bit as well. For reference, this patch only impacts x86. Signed-off-by: Matt Tolentino <matthew.e.tolentino@intel.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86: fix EFI memory map parsing
The memory descriptors that comprise the EFI memory map are not fixed in stone such that the size could change in the future. This uses the memory descriptor size obtained from EFI to iterate over the memory map entries during boot. This enables the removal of an x86 specific pad (and ifdef) in the EFI header. I also couldn't stomach the broken up nature of the function to put EFI runtime calls into virtual mode any longer so I fixed that up a bit as well. For reference, this patch only impacts x86. Signed-off-by: Matt Tolentino <matthew.e.tolentino@intel.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Matt Tolentino · Linus Torvalds
1 parent 4116c527ea
Showing 5 changed files with 67 additions and 69 deletions Side-by-side Diff
arch/i386/kernel/efi.c
arch/i386/kernel/setup.c
arch/i386/mm/init.c
include/asm-i386/setup.h
include/linux/efi.h
@@ -233,22 +233,23 @@
 {
 	memmap.map = NULL;
  
-	memmap.map = (efi_memory_desc_t *)
-		bt_ioremap((unsigned long) memmap.phys_map,
-			(memmap.nr_map * sizeof(efi_memory_desc_t)));
-
+	memmap.map = bt_ioremap((unsigned long) memmap.phys_map,
+			(memmap.nr_map * memmap.desc_size));
 	if (memmap.map == NULL)
 		printk(KERN_ERR PFX "Could not remap the EFI memmap!\n");
+
+	memmap.map_end = memmap.map + (memmap.nr_map * memmap.desc_size);
 }
  
 #if EFI_DEBUG
 static void __init print_efi_memmap(void)
 {
 	efi_memory_desc_t *md;
+	void *p;
 	int i;
  
-	for (i = 0; i < memmap.nr_map; i++) {
-		md = &memmap.map[i];
+	for (p = memmap.map, i = 0; p < memmap.map_end; p += memmap.desc_size, i++) {
+		md = p;
 		printk(KERN_INFO "mem%02u: type=%u, attr=0x%llx, "
 			"range=[0x%016llx-0x%016llx) (%lluMB)\n",
 			i, md->type, md->attribute, md->phys_addr,
  
@@ -271,10 +272,10 @@
 	} prev, curr;
 	efi_memory_desc_t *md;
 	unsigned long start, end;
-	int i;
+	void *p;
  
-	for (i = 0; i < memmap.nr_map; i++) {
-		md = &memmap.map[i];
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
  
 		if ((md->num_pages == 0) || (!is_available_memory(md)))
 			continue;
@@ -325,6 +326,7 @@
 	memmap.phys_map = EFI_MEMMAP;
 	memmap.nr_map = EFI_MEMMAP_SIZE/EFI_MEMDESC_SIZE;
 	memmap.desc_version = EFI_MEMDESC_VERSION;
+	memmap.desc_size = EFI_MEMDESC_SIZE;
  
 	efi.systab = (efi_system_table_t *)
 		boot_ioremap((unsigned long) efi_phys.systab,
  
  
@@ -428,22 +430,30 @@
 		printk(KERN_ERR PFX "Could not map the runtime service table!\n");
  
 	/* Map the EFI memory map for use until paging_init() */
-
-	memmap.map = (efi_memory_desc_t *)
-		boot_ioremap((unsigned long) EFI_MEMMAP, EFI_MEMMAP_SIZE);
-
+	memmap.map = boot_ioremap((unsigned long) EFI_MEMMAP, EFI_MEMMAP_SIZE);
 	if (memmap.map == NULL)
 		printk(KERN_ERR PFX "Could not map the EFI memory map!\n");
  
-	if (EFI_MEMDESC_SIZE != sizeof(efi_memory_desc_t)) {
-		printk(KERN_WARNING PFX "Warning! Kernel-defined memdesc doesn't "
-			   "match the one from EFI!\n");
-	}
+	memmap.map_end = memmap.map + (memmap.nr_map * memmap.desc_size);
+
 #if EFI_DEBUG
 	print_efi_memmap();
 #endif
 }
  
+static inline void __init check_range_for_systab(efi_memory_desc_t *md)
+{
+	if (((unsigned long)md->phys_addr <= (unsigned long)efi_phys.systab) &&
+		((unsigned long)efi_phys.systab < md->phys_addr +
+		((unsigned long)md->num_pages << EFI_PAGE_SHIFT))) {
+		unsigned long addr;
+
+		addr = md->virt_addr - md->phys_addr +
+			(unsigned long)efi_phys.systab;
+		efi.systab = (efi_system_table_t *)addr;
+	}
+}
+
 /*
  * This function will switch the EFI runtime services to virtual mode.
  * Essentially, look through the EFI memmap and map every region that
  
  
  
  
  
@@ -457,43 +467,32 @@
 {
 	efi_memory_desc_t *md;
 	efi_status_t status;
-	int i;
+	void *p;
  
 	efi.systab = NULL;
  
-	for (i = 0; i < memmap.nr_map; i++) {
-		md = &memmap.map[i];
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
  
-		if (md->attribute & EFI_MEMORY_RUNTIME) {
-			md->virt_addr =
-				(unsigned long)ioremap(md->phys_addr,
-					md->num_pages << EFI_PAGE_SHIFT);
-			if (!(unsigned long)md->virt_addr) {
-				printk(KERN_ERR PFX "ioremap of 0x%lX failed\n",
-					(unsigned long)md->phys_addr);
-			}
+		if (!(md->attribute & EFI_MEMORY_RUNTIME))
+			continue;
  
-			if (((unsigned long)md->phys_addr <=
-					(unsigned long)efi_phys.systab) &&
-				((unsigned long)efi_phys.systab <
-					md->phys_addr +
-					((unsigned long)md->num_pages <<
-						EFI_PAGE_SHIFT))) {
-				unsigned long addr;
-
-				addr = md->virt_addr - md->phys_addr +
-						(unsigned long)efi_phys.systab;
-				efi.systab = (efi_system_table_t *)addr;
-			}
+		md->virt_addr = (unsigned long)ioremap(md->phys_addr,
+			md->num_pages << EFI_PAGE_SHIFT);
+		if (!(unsigned long)md->virt_addr) {
+			printk(KERN_ERR PFX "ioremap of 0x%lX failed\n",
+				(unsigned long)md->phys_addr);
 		}
+		/* update the virtual address of the EFI system table */
+		check_range_for_systab(md);
 	}
  
 	if (!efi.systab)
 		BUG();
  
 	status = phys_efi_set_virtual_address_map(
-			sizeof(efi_memory_desc_t) * memmap.nr_map,
-			sizeof(efi_memory_desc_t),
+			memmap.desc_size * memmap.nr_map,
+			memmap.desc_size,
 			memmap.desc_version,
 		       	memmap.phys_map);
  
  
@@ -533,10 +532,10 @@
 {
 	struct resource *res;
 	efi_memory_desc_t *md;
-	int i;
+	void *p;
  
-	for (i = 0; i < memmap.nr_map; i++) {
-		md = &memmap.map[i];
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
  
 		if ((md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT)) >
 		    0x100000000ULL)
  
@@ -613,10 +612,10 @@
 u32 efi_mem_type(unsigned long phys_addr)
 {
 	efi_memory_desc_t *md;
-	int i;
+	void *p;
  
-	for (i = 0; i < memmap.nr_map; i++) {
-		md = &memmap.map[i];
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
 		if ((md->phys_addr <= phys_addr) && (phys_addr <
 			(md->phys_addr + (md-> num_pages << EFI_PAGE_SHIFT)) ))
 			return md->type;
  
@@ -627,10 +626,10 @@
 u64 efi_mem_attributes(unsigned long phys_addr)
 {
 	efi_memory_desc_t *md;
-	int i;
+	void *p;
  
-	for (i = 0; i < memmap.nr_map; i++) {
-		md = &memmap.map[i];
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
 		if ((md->phys_addr <= phys_addr) && (phys_addr <
 			(md->phys_addr + (md-> num_pages << EFI_PAGE_SHIFT)) ))
 			return md->attribute;
@@ -370,12 +370,16 @@
 	int i;
  
 	if (efi_enabled) {
-		for (i = 0; i < memmap.nr_map; i++) {
-			current_addr = memmap.map[i].phys_addr +
-				       (memmap.map[i].num_pages << 12);
-			if (memmap.map[i].type == EFI_CONVENTIONAL_MEMORY) {
+		efi_memory_desc_t *md;
+		void *p;
+
+		for (p = memmap.map, i = 0; p < memmap.map_end;
+			p += memmap.desc_size, i++) {
+			md = p;
+			current_addr = md->phys_addr + (md->num_pages << 12);
+			if (md->type == EFI_CONVENTIONAL_MEMORY) {
 				if (current_addr >= size) {
-					memmap.map[i].num_pages -=
+					md->num_pages -=
 						(((current_addr-size) + PAGE_SIZE-1) >> PAGE_SHIFT);
 					memmap.nr_map = i + 1;
 					return;
@@ -198,9 +198,10 @@
  
 	if (efi_enabled) {
 		efi_memory_desc_t *md;
+		void *p;
  
-		for (i = 0; i < memmap.nr_map; i++) {
-			md = &memmap.map[i];
+		for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+			md = p;
 			if (!is_available_memory(md))
 				continue;
 			addr = (md->phys_addr+PAGE_SIZE-1) >> PAGE_SHIFT;
@@ -44,7 +44,7 @@
 #define EFI_SYSTAB ((efi_system_table_t *) *((unsigned long *)(PARAM+0x1c4)))
 #define EFI_MEMDESC_SIZE (*((unsigned long *) (PARAM+0x1c8)))
 #define EFI_MEMDESC_VERSION (*((unsigned long *) (PARAM+0x1cc)))
-#define EFI_MEMMAP ((efi_memory_desc_t *) *((unsigned long *)(PARAM+0x1d0)))
+#define EFI_MEMMAP ((void *) *((unsigned long *)(PARAM+0x1d0)))
 #define EFI_MEMMAP_SIZE (*((unsigned long *) (PARAM+0x1d4)))
 #define MOUNT_ROOT_RDONLY (*(unsigned short *) (PARAM+0x1F2))
 #define RAMDISK_FLAGS (*(unsigned short *) (PARAM+0x1F8))
@@ -91,11 +91,6 @@
  
 #define EFI_PAGE_SHIFT		12
  
-/*
- * For current x86 implementations of EFI, there is
- * additional padding in the mem descriptors.  This is not
- * the case in ia64.  Need to have this fixed in the f/w.
- */
 typedef struct {
 	u32 type;
 	u32 pad;
@@ -103,9 +98,6 @@
 	u64 virt_addr;
 	u64 num_pages;
 	u64 attribute;
-#if defined (__i386__)
-	u64 pad1;
-#endif
 } efi_memory_desc_t;
  
 typedef int (*efi_freemem_callback_t) (unsigned long start, unsigned long end, void *arg);
  
@@ -240,10 +232,12 @@
 } efi_system_table_t;
  
 struct efi_memory_map {
-	efi_memory_desc_t *phys_map;
-	efi_memory_desc_t *map;
+	void *phys_map;
+	void *map;
+	void *map_end;
 	int nr_map;
 	unsigned long desc_version;
+	unsigned long desc_size;
 };
  
 /*
...	...	@@ -233,22 +233,23 @@
233	233	{
234	234	memmap.map = NULL;
235	235
236		- memmap.map = (efi_memory_desc_t *)
237		- bt_ioremap((unsigned long) memmap.phys_map,
238		- (memmap.nr_map * sizeof(efi_memory_desc_t)));
239		-
	236	+ memmap.map = bt_ioremap((unsigned long) memmap.phys_map,
	237	+ (memmap.nr_map * memmap.desc_size));
240	238	if (memmap.map == NULL)
241	239	printk(KERN_ERR PFX "Could not remap the EFI memmap!\n");
	240	+
	241	+ memmap.map_end = memmap.map + (memmap.nr_map * memmap.desc_size);
242	242	}
243	243
244	244	#if EFI_DEBUG
245	245	static void __init print_efi_memmap(void)
246	246	{
247	247	efi_memory_desc_t *md;
	248	+ void *p;
248	249	int i;
249	250
250		- for (i = 0; i < memmap.nr_map; i++) {
251		- md = &memmap.map[i];
	251	+ for (p = memmap.map, i = 0; p < memmap.map_end; p += memmap.desc_size, i++) {
	252	+ md = p;
252	253	printk(KERN_INFO "mem%02u: type=%u, attr=0x%llx, "
253	254	"range=[0x%016llx-0x%016llx) (%lluMB)\n",
254	255	i, md->type, md->attribute, md->phys_addr,
255	256
...	...	@@ -271,10 +272,10 @@
271	272	} prev, curr;
272	273	efi_memory_desc_t *md;
273	274	unsigned long start, end;
274		- int i;
	275	+ void *p;
275	276
276		- for (i = 0; i < memmap.nr_map; i++) {
277		- md = &memmap.map[i];
	277	+ for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
	278	+ md = p;
278	279
279	280	if ((md->num_pages == 0) \|\| (!is_available_memory(md)))
280	281	continue;
...	...	@@ -325,6 +326,7 @@
325	326	memmap.phys_map = EFI_MEMMAP;
326	327	memmap.nr_map = EFI_MEMMAP_SIZE/EFI_MEMDESC_SIZE;
327	328	memmap.desc_version = EFI_MEMDESC_VERSION;
	329	+ memmap.desc_size = EFI_MEMDESC_SIZE;
328	330
329	331	efi.systab = (efi_system_table_t *)
330	332	boot_ioremap((unsigned long) efi_phys.systab,
331	333
332	334
...	...	@@ -428,22 +430,30 @@
428	430	printk(KERN_ERR PFX "Could not map the runtime service table!\n");
429	431
430	432	/* Map the EFI memory map for use until paging_init() */
431		-
432		- memmap.map = (efi_memory_desc_t *)
433		- boot_ioremap((unsigned long) EFI_MEMMAP, EFI_MEMMAP_SIZE);
434		-
	433	+ memmap.map = boot_ioremap((unsigned long) EFI_MEMMAP, EFI_MEMMAP_SIZE);
435	434	if (memmap.map == NULL)
436	435	printk(KERN_ERR PFX "Could not map the EFI memory map!\n");
437	436
438		- if (EFI_MEMDESC_SIZE != sizeof(efi_memory_desc_t)) {
439		- printk(KERN_WARNING PFX "Warning! Kernel-defined memdesc doesn't "
440		- "match the one from EFI!\n");
441		- }
	437	+ memmap.map_end = memmap.map + (memmap.nr_map * memmap.desc_size);
	438	+
442	439	#if EFI_DEBUG
443	440	print_efi_memmap();
444	441	#endif
445	442	}
446	443
	444	+static inline void __init check_range_for_systab(efi_memory_desc_t *md)
	445	+{
	446	+ if (((unsigned long)md->phys_addr <= (unsigned long)efi_phys.systab) &&
	447	+ ((unsigned long)efi_phys.systab < md->phys_addr +
	448	+ ((unsigned long)md->num_pages << EFI_PAGE_SHIFT))) {
	449	+ unsigned long addr;
	450	+
	451	+ addr = md->virt_addr - md->phys_addr +
	452	+ (unsigned long)efi_phys.systab;
	453	+ efi.systab = (efi_system_table_t *)addr;
	454	+ }
	455	+}
	456	+
447	457	/*
448	458	* This function will switch the EFI runtime services to virtual mode.
449	459	* Essentially, look through the EFI memmap and map every region that
450	460
451	461
452	462
453	463
454	464
...	...	@@ -457,43 +467,32 @@
457	467	{
458	468	efi_memory_desc_t *md;
459	469	efi_status_t status;
460		- int i;
	470	+ void *p;
461	471
462	472	efi.systab = NULL;
463	473
464		- for (i = 0; i < memmap.nr_map; i++) {
465		- md = &memmap.map[i];
	474	+ for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
	475	+ md = p;
466	476
467		- if (md->attribute & EFI_MEMORY_RUNTIME) {
468		- md->virt_addr =
469		- (unsigned long)ioremap(md->phys_addr,
470		- md->num_pages << EFI_PAGE_SHIFT);
471		- if (!(unsigned long)md->virt_addr) {
472		- printk(KERN_ERR PFX "ioremap of 0x%lX failed\n",
473		- (unsigned long)md->phys_addr);
474		- }
	477	+ if (!(md->attribute & EFI_MEMORY_RUNTIME))
	478	+ continue;
475	479
476		- if (((unsigned long)md->phys_addr <=
477		- (unsigned long)efi_phys.systab) &&
478		- ((unsigned long)efi_phys.systab <
479		- md->phys_addr +
480		- ((unsigned long)md->num_pages <<
481		- EFI_PAGE_SHIFT))) {
482		- unsigned long addr;
483		-
484		- addr = md->virt_addr - md->phys_addr +
485		- (unsigned long)efi_phys.systab;
486		- efi.systab = (efi_system_table_t *)addr;
487		- }
	480	+ md->virt_addr = (unsigned long)ioremap(md->phys_addr,
	481	+ md->num_pages << EFI_PAGE_SHIFT);
	482	+ if (!(unsigned long)md->virt_addr) {
	483	+ printk(KERN_ERR PFX "ioremap of 0x%lX failed\n",
	484	+ (unsigned long)md->phys_addr);
488	485	}
	486	+ /* update the virtual address of the EFI system table */
	487	+ check_range_for_systab(md);
489	488	}
490	489
491	490	if (!efi.systab)
492	491	BUG();
493	492
494	493	status = phys_efi_set_virtual_address_map(
495		- sizeof(efi_memory_desc_t) * memmap.nr_map,
496		- sizeof(efi_memory_desc_t),
	494	+ memmap.desc_size * memmap.nr_map,
	495	+ memmap.desc_size,
497	496	memmap.desc_version,
498	497	memmap.phys_map);
499	498
500	499
...	...	@@ -533,10 +532,10 @@
533	532	{
534	533	struct resource *res;
535	534	efi_memory_desc_t *md;
536		- int i;
	535	+ void *p;
537	536
538		- for (i = 0; i < memmap.nr_map; i++) {
539		- md = &memmap.map[i];
	537	+ for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
	538	+ md = p;
540	539
541	540	if ((md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT)) >
542	541	0x100000000ULL)
543	542
...	...	@@ -613,10 +612,10 @@
613	612	u32 efi_mem_type(unsigned long phys_addr)
614	613	{
615	614	efi_memory_desc_t *md;
616		- int i;
	615	+ void *p;
617	616
618		- for (i = 0; i < memmap.nr_map; i++) {
619		- md = &memmap.map[i];
	617	+ for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
	618	+ md = p;
620	619	if ((md->phys_addr <= phys_addr) && (phys_addr <
621	620	(md->phys_addr + (md-> num_pages << EFI_PAGE_SHIFT)) ))
622	621	return md->type;
623	622
...	...	@@ -627,10 +626,10 @@
627	626	u64 efi_mem_attributes(unsigned long phys_addr)
628	627	{
629	628	efi_memory_desc_t *md;
630		- int i;
	629	+ void *p;
631	630
632		- for (i = 0; i < memmap.nr_map; i++) {
633		- md = &memmap.map[i];
	631	+ for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
	632	+ md = p;
634	633	if ((md->phys_addr <= phys_addr) && (phys_addr <
635	634	(md->phys_addr + (md-> num_pages << EFI_PAGE_SHIFT)) ))
636	635	return md->attribute;
...	...	@@ -370,12 +370,16 @@
370	370	int i;
371	371
372	372	if (efi_enabled) {
373		- for (i = 0; i < memmap.nr_map; i++) {
374		- current_addr = memmap.map[i].phys_addr +
375		- (memmap.map[i].num_pages << 12);
376		- if (memmap.map[i].type == EFI_CONVENTIONAL_MEMORY) {
	373	+ efi_memory_desc_t *md;
	374	+ void *p;
	375	+
	376	+ for (p = memmap.map, i = 0; p < memmap.map_end;
	377	+ p += memmap.desc_size, i++) {
	378	+ md = p;
	379	+ current_addr = md->phys_addr + (md->num_pages << 12);
	380	+ if (md->type == EFI_CONVENTIONAL_MEMORY) {
377	381	if (current_addr >= size) {
378		- memmap.map[i].num_pages -=
	382	+ md->num_pages -=
379	383	(((current_addr-size) + PAGE_SIZE-1) >> PAGE_SHIFT);
380	384	memmap.nr_map = i + 1;
381	385	return;
...	...	@@ -198,9 +198,10 @@
198	198
199	199	if (efi_enabled) {
200	200	efi_memory_desc_t *md;
	201	+ void *p;
201	202
202		- for (i = 0; i < memmap.nr_map; i++) {
203		- md = &memmap.map[i];
	203	+ for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
	204	+ md = p;
204	205	if (!is_available_memory(md))
205	206	continue;
206	207	addr = (md->phys_addr+PAGE_SIZE-1) >> PAGE_SHIFT;
...	...	@@ -44,7 +44,7 @@
44	44	#define EFI_SYSTAB ((efi_system_table_t ) ((unsigned long *)(PARAM+0x1c4)))
45	45	#define EFI_MEMDESC_SIZE (((unsigned long ) (PARAM+0x1c8)))
46	46	#define EFI_MEMDESC_VERSION (((unsigned long ) (PARAM+0x1cc)))
47		-#define EFI_MEMMAP ((efi_memory_desc_t ) ((unsigned long *)(PARAM+0x1d0)))
	47	+#define EFI_MEMMAP ((void ) ((unsigned long *)(PARAM+0x1d0)))
48	48	#define EFI_MEMMAP_SIZE (((unsigned long ) (PARAM+0x1d4)))
49	49	#define MOUNT_ROOT_RDONLY ((unsigned short ) (PARAM+0x1F2))
50	50	#define RAMDISK_FLAGS ((unsigned short ) (PARAM+0x1F8))
...	...	@@ -91,11 +91,6 @@
91	91
92	92	#define EFI_PAGE_SHIFT 12
93	93
94		-/*
95		- * For current x86 implementations of EFI, there is
96		- * additional padding in the mem descriptors. This is not
97		- * the case in ia64. Need to have this fixed in the f/w.
98		- */
99	94	typedef struct {
100	95	u32 type;
101	96	u32 pad;
...	...	@@ -103,9 +98,6 @@
103	98	u64 virt_addr;
104	99	u64 num_pages;
105	100	u64 attribute;
106		-#if defined (__i386__)
107		- u64 pad1;
108		-#endif
109	101	} efi_memory_desc_t;
110	102
111	103	typedef int (efi_freemem_callback_t) (unsigned long start, unsigned long end, void arg);
112	104
...	...	@@ -240,10 +232,12 @@
240	232	} efi_system_table_t;
241	233
242	234	struct efi_memory_map {
243		- efi_memory_desc_t *phys_map;
244		- efi_memory_desc_t *map;
	235	+ void *phys_map;
	236	+ void *map;
	237	+ void *map_end;
245	238	int nr_map;
246	239	unsigned long desc_version;
	240	+ unsigned long desc_size;
247	241	};
248	242
249	243	/*