ARM: option for loading modules into vmalloc area

Usually modules are loaded into small area prior to the kernel
text because they are linked with the kernel using short calls.
Compile-time instrumentation like GCOV or KASAN bloats code a lot,
and as a result huge modules no longer fit into reserved area.

This patch adds option CONFIG_MODULES_USE_VMALLOC which lifts
limitation on amount of loaded modules. It links modules using
long-calls (option -mlong-calls) and loads them into vmalloc area.

In few places exported symbols are called from inline assembly.
This patch adds macro for such call sites: __asmbl and __asmbl_clobber.
Call turns into single 'bl' or sequence 'movw; movt; blx' depending on
context and state of config option.

Unfortunately this option isn't compatible with CONFIG_FUNCTION_TRACER.
Compiler emits short calls to profiling function despite of -mlong-calls.
This is a bug in GCC, but ftrace anyway needs an update to handle this.

Signed-off-by: Konstantin Khlebnikov <k.khlebnikov@samsung.com>
Patch-mainline: linux-arm-kernel @ 18/11/2014, 20:21:46
Change-Id: Iea7990a033c060c26f5782125fb63d6f96a9d218
Signed-off-by: Shubham Aggarwal <shubagga@codeaurora.org>

arch/arm/Kconfig

@@ -1726,6 +1726,26 @@ config CPU_SW_DOMAIN_PAN
 	  Their lower 1MB needs to remain accessible for the vectors, but
 	  the remainder of userspace will become appropriately inaccessible.
 
+config MODULES_USE_LONG_CALLS
+	bool
+	help
+	  Use long calls for calling exported symbols.
+
+config MODULES_USE_VMALLOC
+	bool "Put modules into vmalloc area"
+	select MODULES_USE_LONG_CALLS
+	depends on MMU && MODULES
+	depends on !XIP_KERNEL
+	depends on !FUNCTION_TRACER
+	help
+	  Usually modules are loaded into small area prior to the kernel text
+	  because they are linked with the kernel using short calls.
+
+	  This option enables long calls and moves modules into vmalloc area.
+	  This allows to load more modules but adds some perfromance penalty.
+
+	  If unsure, say n.
+
 config HW_PERF_EVENTS
 	def_bool y
 	depends on ARM_PMU

arch/arm/Makefile

@@ -136,6 +136,10 @@ CFLAGS_ISA	:=$(call cc-option,-marm,)
 AFLAGS_ISA	:=$(CFLAGS_ISA)
 endif
 
+ifeq ($(CONFIG_MODULES_USE_LONG_CALLS),y)
+CFLAGS_MODULE	+= -mlong-calls
+endif
+
 # Need -Uarm for gcc < 3.x
 KBUILD_CFLAGS	+=$(CFLAGS_ABI) $(CFLAGS_ISA) $(arch-y) $(tune-y) $(call cc-option,-mshort-load-bytes,$(call cc-option,-malignment-traps,)) -msoft-float $(call cc-option, -Uarm,)
 KBUILD_AFLAGS	+=$(CFLAGS_ABI) $(AFLAGS_ISA) $(arch-y) $(tune-y) -include asm/unified.h -msoft-float

arch/arm/include/asm/compiler.h

@@ -25,5 +25,18 @@
 	  ".endif; "				\
 	".endif\n\t"
 
+/*
+ * This is used for calling exported symbols from inline assembly code.
+ */
+#if defined(MODULE) && defined(CONFIG_MODULES_USE_LONG_CALLS)
+#define __asmbl(cond, reg, target) \
+	"movw	" reg ", #:lower16:" target "\n\t" \
+	"movt	" reg ", #:upper16:" target "\n\t" \
+	"blx" cond "	" reg "\n\t"
+#define __asmbl_clobber(reg)	,reg
+#else
+#define __asmbl(cond, reg, target) "bl" cond "	" target"\n\t"
+#define __asmbl_clobber(reg)
+#endif
 
 #endif /* __ASM_ARM_COMPILER_H */

arch/arm/include/asm/div64.h

@@ -40,7 +40,7 @@ static inline uint32_t __div64_32(uint64_t *n, uint32_t base)
 		__asmeq("%1", "r2")
 		__asmeq("%2", "r0")
 		__asmeq("%3", "r4")
-		"bl	__do_div64"
+		__asmbl("", "ip",  "__do_div64")
 		: "=r" (__rem), "=r" (__res)
 		: "r" (__n), "r" (__base)
 		: "ip", "lr", "cc");

arch/arm/include/asm/memory.h

@@ -45,6 +45,15 @@
  */
 #define TASK_SIZE_26		(UL(1) << 26)
 
+#ifdef CONFIG_MODULES_USE_VMALLOC
+/*
+ * Modules might be anywhere in the vmalloc area.
+ */
+#define MODULES_VADDR		VMALLOC_START
+#define MODULES_END		VMALLOC_END
+
+#else /* CONFIG_MODULES_USE_VMALLOC */
+
 /*
  * The module space lives between the addresses given by TASK_SIZE
  * and PAGE_OFFSET - it must be within 32MB of the kernel text.
@@ -69,6 +78,8 @@
 #define MODULES_END		(PAGE_OFFSET)
 #endif
 
+#endif /* CONFIG_MODULES_USE_VMALLOC */
+
 /*
  * The XIP kernel gets mapped at the bottom of the module vm area.
  * Since we use sections to map it, this macro replaces the physical address

arch/arm/include/asm/uaccess.h

@@ -145,21 +145,21 @@ extern int __get_user_64t_1(void *);
 extern int __get_user_64t_2(void *);
 extern int __get_user_64t_4(void *);
 
-#define __GUP_CLOBBER_1	"lr", "cc"
+#define __GUP_CLOBBER_1	"lr", "cc" __asmbl_clobber("ip")
 #ifdef CONFIG_CPU_USE_DOMAINS
 #define __GUP_CLOBBER_2	"ip", "lr", "cc"
 #else
-#define __GUP_CLOBBER_2 "lr", "cc"
+#define __GUP_CLOBBER_2 "lr", "cc" __asmbl_clobber("ip")
 #endif
-#define __GUP_CLOBBER_4	"lr", "cc"
-#define __GUP_CLOBBER_32t_8 "lr", "cc"
-#define __GUP_CLOBBER_8	"lr", "cc"
+#define __GUP_CLOBBER_4	"lr", "cc" __asmbl_clobber("ip")
+#define __GUP_CLOBBER_32t_8 "lr", "cc" __asmbl_clobber("ip")
+#define __GUP_CLOBBER_8	"lr", "cc" __asmbl_clobber("ip")
 
 #define __get_user_x(__r2, __p, __e, __l, __s)				\
 	   __asm__ __volatile__ (					\
 		__asmeq("%0", "r0") __asmeq("%1", "r2")			\
 		__asmeq("%3", "r1")					\
-		"bl	__get_user_" #__s				\
+		__asmbl("", "ip", "__get_user_" #__s)			\
 		: "=&r" (__e), "=r" (__r2)				\
 		: "0" (__p), "r" (__l)					\
 		: __GUP_CLOBBER_##__s)
@@ -181,7 +181,7 @@ extern int __get_user_64t_4(void *);
 	   __asm__ __volatile__ (					\
 		__asmeq("%0", "r0") __asmeq("%1", "r2")			\
 		__asmeq("%3", "r1")					\
-		"bl	__get_user_64t_" #__s				\
+		__asmbl("", "ip", "__get_user_64t_" #__s)		\
 		: "=&r" (__e), "=r" (__r2)				\
 		: "0" (__p), "r" (__l)					\
 		: __GUP_CLOBBER_##__s)
@@ -251,7 +251,7 @@ extern int __put_user_8(void *, unsigned long long);
 		__asm__ __volatile__ (					\
 			__asmeq("%0", "r0") __asmeq("%2", "r2")		\
 			__asmeq("%3", "r1")				\
-			"bl	__put_user_" #__s			\
+			__asmbl("", "ip", "__put_user_" #__s)		\
 			: "=&r" (__e)					\
 			: "0" (__p), "r" (__r2), "r" (__l)		\
 			: "ip", "lr", "cc");				\

arch/arm/kernel/module.c

@@ -38,6 +38,7 @@
 #endif
 
 #ifdef CONFIG_MMU
+#ifndef CONFIG_MODULES_USE_VMALLOC
 void *module_alloc(unsigned long size)
 {
 	gfp_t gfp_mask = GFP_KERNEL;
@@ -56,6 +57,7 @@ void *module_alloc(unsigned long size)
 				GFP_KERNEL, PAGE_KERNEL_EXEC, 0, NUMA_NO_NODE,
 				__builtin_return_address(0));
 }
+#endif /* CONFIG_MODULES_USE_VMALLOC */
 #endif
 
 int

arch/arm/mm/dump.c

@@ -21,6 +21,7 @@
 #include <asm/fixmap.h>
 #include <asm/memory.h>
 #include <asm/pgtable.h>
+#include <asm/highmem.h>
 
 struct addr_marker {
 	unsigned long start_address;
@@ -28,7 +29,12 @@ struct addr_marker {
 };
 
 static struct addr_marker address_markers[] = {
+#ifndef CONFIG_MODULES_USE_VMALLOC
 	{ MODULES_VADDR,	"Modules" },
+#endif
+#ifdef CONFIG_HIGHMEM
+	{ PKMAP_BASE,		"Page kmap" },
+#endif
 	{ PAGE_OFFSET,		"Kernel Mapping" },
 	{ 0,			"vmalloc() Area" },
 	{ VMALLOC_END,		"vmalloc() End" },
@@ -386,7 +392,9 @@ static int ptdump_init(void)
 			for (j = 0; j < pg_level[i].num; j++)
 				pg_level[i].mask |= pg_level[i].bits[j].mask;
 
-	address_markers[2].start_address = VMALLOC_START;
+	i = 1 + !IS_ENABLED(CONFIG_MODULES_USE_VMALLOC) +
+		!!IS_ENABLED(CONFIG_HIGHMEM);
+	address_markers[i].start_address = VMALLOC_START;
 
 	pe = debugfs_create_file("kernel_page_tables", 0400, NULL, NULL,
 				 &ptdump_fops);

arch/arm/mm/init.c

@@ -623,7 +623,9 @@ void __init mem_init(void)
 	 * be detected at build time already.
 	 */
 #ifdef CONFIG_MMU
+#ifndef CONFIG_MODULES_USE_VMALLOC
 	BUILD_BUG_ON(TASK_SIZE				> MODULES_VADDR);
+#endif
 	BUG_ON(TASK_SIZE 				> MODULES_VADDR);
 #endif
 

arch/arm/mm/mmu.c

@@ -1261,16 +1261,15 @@ void __init adjust_lowmem_bounds(void)
 
 static inline void prepare_page_table(void)
 {
-	unsigned long addr;
+	unsigned long addr = 0;
 	phys_addr_t end;
 
 	/*
 	 * Clear out all the mappings below the kernel image.
 	 */
-	for (addr = 0; addr < MODULES_VADDR; addr += PMD_SIZE)
-		pmd_clear(pmd_off_k(addr));
-
 #ifdef CONFIG_XIP_KERNEL
+	for ( ; addr < MODULES_VADDR; addr += PMD_SIZE)
+		pmd_clear(pmd_off_k(addr));
 	/* The XIP kernel is mapped in the module area -- skip over it */
 	addr = ((unsigned long)_exiprom + PMD_SIZE - 1) & PMD_MASK;
 #endif

arch/arm/mm/pgd.c

@@ -53,17 +53,18 @@ pgd_t *pgd_alloc(struct mm_struct *mm)
 	clean_dcache_area(new_pgd, PTRS_PER_PGD * sizeof(pgd_t));
 
 #ifdef CONFIG_ARM_LPAE
+#if defined(CONFIG_HIGHMEM) || !defined(CONFIG_MODULES_USE_VMALLOC)
 	/*
 	 * Allocate PMD table for modules and pkmap mappings.
 	 */
-	new_pud = pud_alloc(mm, new_pgd + pgd_index(MODULES_VADDR),
-			    MODULES_VADDR);
+	new_pud = pud_alloc(mm, new_pgd + pgd_index(PKMAP_BASE), PKMAP_BASE);
 	if (!new_pud)
 		goto no_pud;
 
 	new_pmd = pmd_alloc(mm, new_pud, 0);
 	if (!new_pmd)
 		goto no_pmd;
+#endif
 #endif
 
 	if (!vectors_high()) {