From patchwork series 421443

https://patchwork.ozlabs.org//project/linuxppc-dev/list/?series=421443

arch/alpha/kernel/osf_sys.c

@@ -1223,6 +1223,8 @@ arch_get_unmapped_area_1(unsigned long addr, unsigned long len,
 	info.length = len;
 	info.low_limit = addr;
 	info.high_limit = limit;
+	info.hint = addr;
+	info.mmap_flags = flags;
 	return vm_unmapped_area(&info);
 }
 

arch/arc/mm/mmap.c

@@ -29,6 +29,9 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
 	struct vm_area_struct *vma;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/*
 	 * We enforce the MAP_FIXED case.
 	 */

arch/arm/mm/mmap.c

@@ -36,6 +36,9 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
 	int aliasing = cache_is_vipt_aliasing();
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/*
 	 * We only need to do colour alignment if either the I or D
 	 * caches alias.
@@ -56,6 +59,7 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
 	if (len > TASK_SIZE)
 		return -ENOMEM;
 
+
 	if (addr) {
 		if (do_align)
 			addr = COLOUR_ALIGN(addr, pgoff);
@@ -88,6 +92,9 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 	int aliasing = cache_is_vipt_aliasing();
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/*
 	 * We only need to do colour alignment if either the I or D
 	 * caches alias.

arch/csky/abiv1/mmap.c

@@ -35,6 +35,9 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
 		.align_offset = pgoff << PAGE_SHIFT
 	};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/*
 	 * We only need to do colour alignment if either the I or D
 	 * caches alias.

arch/loongarch/mm/mmap.c

@@ -27,6 +27,9 @@ static unsigned long arch_get_unmapped_area_common(struct file *filp,
 	int do_color_align;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (unlikely(len > TASK_SIZE))
 		return -ENOMEM;
 

arch/mips/mm/mmap.c

@@ -36,6 +36,9 @@ static unsigned long arch_get_unmapped_area_common(struct file *filp,
 	int do_color_align;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (unlikely(len > TASK_SIZE))
 		return -ENOMEM;
 

arch/parisc/kernel/sys_parisc.c

@@ -108,6 +108,9 @@ static unsigned long arch_get_unmapped_area_common(struct file *filp,
 		.length = len
 	};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (unlikely(len > TASK_SIZE))
 		return -ENOMEM;
 

arch/powerpc/mm/book3s64/slice.c

@@ -286,6 +286,10 @@ static unsigned long slice_find_area_bottomup(struct mm_struct *mm,
 		.length = len,
 		.align_mask = PAGE_MASK & ((1ul << pshift) - 1),
 	};
+
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/*
 	 * Check till the allow max value for this mmap request
 	 */
@@ -331,6 +335,9 @@ static unsigned long slice_find_area_topdown(struct mm_struct *mm,
 	};
 	unsigned long min_addr = max(PAGE_SIZE, mmap_min_addr);
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/*
 	 * If we are trying to allocate above DEFAULT_MAP_WINDOW
 	 * Add the different to the mmap_base.

arch/s390/mm/hugetlbpage.c

@@ -254,6 +254,8 @@ static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *file,
 	info.low_limit = current->mm->mmap_base;
 	info.high_limit = TASK_SIZE;
 	info.align_mask = PAGE_MASK & ~huge_page_mask(h);
+	info.hint = addr;
+	info.mmap_flags = flags;
 	return vm_unmapped_area(&info);
 }
 
@@ -270,6 +272,8 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 	info.low_limit = PAGE_SIZE;
 	info.high_limit = current->mm->mmap_base;
 	info.align_mask = PAGE_MASK & ~huge_page_mask(h);
+	info.hint = addr;
+	info.mmap_flags = flags;
 	addr = vm_unmapped_area(&info);
 
 	/*

arch/s390/mm/mmap.c

@@ -88,6 +88,9 @@ unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr,
 	struct vm_area_struct *vma;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (len > TASK_SIZE - mmap_min_addr)
 		return -ENOMEM;
 
@@ -123,6 +126,9 @@ unsigned long arch_get_unmapped_area_topdown(struct file *filp, unsigned long ad
 	struct mm_struct *mm = current->mm;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/* requested length too big for entire address space */
 	if (len > TASK_SIZE - mmap_min_addr)
 		return -ENOMEM;

arch/sh/mm/mmap.c

@@ -59,6 +59,9 @@ unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr,
 	int do_colour_align;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (flags & MAP_FIXED) {
 		/* We do not accept a shared mapping if it would violate
 		 * cache aliasing constraints.
@@ -107,6 +110,9 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 	int do_colour_align;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (flags & MAP_FIXED) {
 		/* We do not accept a shared mapping if it would violate
 		 * cache aliasing constraints.

arch/sparc/kernel/sys_sparc_32.c

@@ -43,6 +43,9 @@ unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr, unsi
 {
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (flags & MAP_FIXED) {
 		/* We do not accept a shared mapping if it would violate
 		 * cache aliasing constraints.

arch/sparc/kernel/sys_sparc_64.c

@@ -95,6 +95,9 @@ unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr, unsi
 	int do_color_align;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (flags & MAP_FIXED) {
 		/* We do not accept a shared mapping if it would violate
 		 * cache aliasing constraints.
@@ -155,6 +158,9 @@ arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 	int do_color_align;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/* This should only ever run for 32-bit processes.  */
 	BUG_ON(!test_thread_flag(TIF_32BIT));
 

arch/sparc/mm/hugetlbpage.c

@@ -40,6 +40,8 @@ static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *filp,
 	info.low_limit = TASK_UNMAPPED_BASE;
 	info.high_limit = min(task_size, VA_EXCLUDE_START);
 	info.align_mask = PAGE_MASK & ~huge_page_mask(h);
+	info.hint = addr;
+	info.mmap_flags = flags;
 	addr = vm_unmapped_area(&info);
 
 	if ((addr & ~PAGE_MASK) && task_size > VA_EXCLUDE_END) {
@@ -71,6 +73,8 @@ hugetlb_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 	info.low_limit = PAGE_SIZE;
 	info.high_limit = mm->mmap_base;
 	info.align_mask = PAGE_MASK & ~huge_page_mask(h);
+	info.hint = addr;
+	info.mmap_flags = flags;
 	addr = vm_unmapped_area(&info);
 
 	/*

arch/x86/kernel/sys_x86_64.c

@@ -129,6 +129,9 @@ arch_get_unmapped_area_vmflags(struct file *filp, unsigned long addr, unsigned l
 	struct vm_unmapped_area_info info = {};
 	unsigned long begin, end;
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (flags & MAP_FIXED)
 		return addr;
 
@@ -167,6 +170,9 @@ arch_get_unmapped_area_topdown_vmflags(struct file *filp, unsigned long addr0,
 	unsigned long addr = addr0;
 	struct vm_unmapped_area_info info = {};
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/* requested length too big for entire address space */
 	if (len > TASK_SIZE)
 		return -ENOMEM;

arch/x86/mm/hugetlbpage.c

@@ -29,6 +29,8 @@ static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *file,
 
 	info.length = len;
 	info.low_limit = get_mmap_base(1);
+	info.hint = addr;
+	info.mmap_flags = flags;
 
 	/*
 	 * If hint address is above DEFAULT_MAP_WINDOW, look for unmapped area
@@ -52,6 +54,8 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file,
 	info.length = len;
 	info.low_limit = PAGE_SIZE;
 	info.high_limit = get_mmap_base(0);
+	info.hint = addr;
+	info.mmap_flags = flags;
 
 	/*
 	 * If hint address is above DEFAULT_MAP_WINDOW, look for unmapped area

fs/hugetlbfs/inode.c

@@ -182,6 +182,8 @@ hugetlb_get_unmapped_area_bottomup(struct file *file, unsigned long addr,
 	info.low_limit = current->mm->mmap_base;
 	info.high_limit = arch_get_mmap_end(addr, len, flags);
 	info.align_mask = PAGE_MASK & ~huge_page_mask(h);
+	info.hint = addr;
+	info.mmap_flags = flags;
 	return vm_unmapped_area(&info);
 }
 
@@ -197,6 +199,8 @@ hugetlb_get_unmapped_area_topdown(struct file *file, unsigned long addr,
 	info.low_limit = PAGE_SIZE;
 	info.high_limit = arch_get_mmap_base(addr, current->mm->mmap_base);
 	info.align_mask = PAGE_MASK & ~huge_page_mask(h);
+	info.hint = addr;
+	info.mmap_flags = flags;
 	addr = vm_unmapped_area(&info);
 
 	/*

include/linux/mm.h

@@ -3434,6 +3434,8 @@ struct vm_unmapped_area_info {
 	unsigned long align_mask;
 	unsigned long align_offset;
 	unsigned long start_gap;
+	unsigned long hint;
+	unsigned long mmap_flags;
 };
 
 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);

include/uapi/asm-generic/mman-common.h

@@ -32,6 +32,7 @@
 
 #define MAP_UNINITIALIZED 0x4000000	/* For anonymous mmap, memory could be
 					 * uninitialized */
+#define MAP_BELOW_HINT	  0x8000000	/* give out address that is below (inclusive) hint address */
 
 /*
  * Flags for mlock

mm/mmap.c

@@ -1766,6 +1766,9 @@ unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info)
 {
 	unsigned long addr;
 
+	if (info->hint != 0 && info->mmap_flags & MAP_BELOW_HINT)
+		info->high_limit = MIN(info->high_limit, info->hint + info->length);
+
 	if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
 		addr = unmapped_area_topdown(info);
 	else
@@ -1796,6 +1799,9 @@ generic_get_unmapped_area(struct file *filp, unsigned long addr,
 	struct vm_unmapped_area_info info = {};
 	const unsigned long mmap_end = arch_get_mmap_end(addr, len, flags);
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	if (len > mmap_end - mmap_min_addr)
 		return -ENOMEM;
 
@@ -1841,6 +1847,9 @@ generic_get_unmapped_area_topdown(struct file *filp, unsigned long addr,
 	struct vm_unmapped_area_info info = {};
 	const unsigned long mmap_end = arch_get_mmap_end(addr, len, flags);
 
+	info.hint = addr;
+	info.mmap_flags = flags;
+
 	/* requested length too big for entire address space */
 	if (len > mmap_end - mmap_min_addr)
 		return -ENOMEM;

tools/include/uapi/asm-generic/mman-common.h

@@ -32,6 +32,7 @@
 
 #define MAP_UNINITIALIZED 0x4000000	/* For anonymous mmap, memory could be
 					 * uninitialized */
+#define MAP_BELOW_HINT	  0x8000000	/* give out address that is below (inclusive) hint address */
 
 /*
  * Flags for mlock

tools/testing/selftests/mm/Makefile

@@ -50,6 +50,7 @@ TEST_GEN_FILES += hugepage-shm
 TEST_GEN_FILES += hugepage-vmemmap
 TEST_GEN_FILES += khugepaged
 TEST_GEN_FILES += madv_populate
+TEST_GEN_FILES += map_below_hint
 TEST_GEN_FILES += map_fixed_noreplace
 TEST_GEN_FILES += map_hugetlb
 TEST_GEN_FILES += map_populate

tools/testing/selftests/mm/map_below_hint.c

@@ -0,0 +1,32 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Test the MAP_BELOW_HINT mmap flag.
+ */
+#include <sys/mman.h>
+#include <errno.h>
+#include "../kselftest.h"
+
+#define ADDR (1 << 20)
+#define LENGTH (ADDR / 10000)
+
+#define MAP_BELOW_HINT	  0x8000000	/* Not defined in all libc */
+
+/*
+ * Map memory with MAP_BELOW_HINT until no memory left. Ensure that all returned
+ * addresses are below the hint.
+ */
+int main(int argc, char **argv)
+{
+	void *addr;
+
+	do {
+		addr = mmap((void *)ADDR, LENGTH, PROT_READ | PROT_WRITE,
+			    MAP_PRIVATE | MAP_ANONYMOUS | MAP_BELOW_HINT, -1, 0);
+	} while (addr != MAP_FAILED && (unsigned long)addr <= ADDR);
+
+	if (errno == ENOMEM)
+		ksft_test_result_pass("MAP_BELOW_HINT works\n");
+	else
+		ksft_test_result_fail("mmap returned address above hint with MAP_BELOW_HINT with error: %s\n",
+				      strerror(errno));
+}