Skip to contentSkip to conversation

mm/mmap: check mapping limits more strictly

PATCH Superseded [email protected] 2 participants 5 messages Mar 23, 2020 – Apr 21, 2020

Subsystems (2)
Memory management Memory mapping
lore
Share & tools
lore

[email protected]

author
#0 lore raw

cover mm/mmap: check mapping limits more strictly

From: Alexander Gordeev <[email protected]>

This series is against v5.6-rc7

Alexander Gordeev (2):
mm/mmap.c: add more sanity checks to get_unmapped_area() mm/mmap.c: do not allow mappings outside of allowed limits

 mm/mmap.c | 29 +++++++++++++++++------------
 1 file changed, 17 insertions(+), 12 deletions(-)
Signature · 1 line
2.23.0
author
#1 lore raw

PATCH 1/2 mm/mmap.c: add more sanity checks to get_unmapped_area()

From: Alexander Gordeev <[email protected]>

Generic get_unmapped_area() function does sanity checks of address and length of the area to be mapped. Yet, it lacks checking against mmap_min_addr and mmap_end limits.

At the same time the default implementation of functions arch_get_unmapped_area[_topdown]() and some architecture callbacks do mmap_min_addr and mmap_end checks on its own.

Put additional checks into the generic code and do not let architecture callbacks to get away with a possible area outside of the allowed limits.

That could also relieve arch_get_unmapped_area[_topdown]() callbacks of own address and length sanity checks.

Signed-off-by Alexander Gordeev <[email protected]>
---
 mm/mmap.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
mm/mmap.c +3 -2
index d681a20eb4ea..a0fcb5ca0e06 100644
@@ -2168,12 +2168,13 @@ get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
 	unsigned long (*get_area)(struct file *, unsigned long,
 				  unsigned long, unsigned long, unsigned long);
 
+	const unsigned long mmap_end = arch_get_mmap_end(addr);
 	unsigned long error = arch_mmap_check(addr, len, flags);
 	if (error)
 		return error;
 
 	/* Careful about overflows.. */
-	if (len > TASK_SIZE)
+	if (len > mmap_end - mmap_min_addr)
 		return -ENOMEM;
 
 	get_area = current->mm->get_unmapped_area;

@@ -2194,7 +2195,7 @@ get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
 	if (IS_ERR_VALUE(addr))
 		return addr;
 
-	if (addr > TASK_SIZE - len)
+	if ((addr < mmap_min_addr) || (addr > mmap_end - len))
 		return -ENOMEM;
 	if (offset_in_page(addr))
 		return -EINVAL;
Signature · 1 line
2.23.0
author
#2 lore raw

PATCH 2/2 mm/mmap.c: do not allow mappings outside of allowed limits

In reply to [email protected]

From: Alexander Gordeev <[email protected]>

It is possible to request a fixed mapping address below mmap_min_addr and succeed. This update adds early checks of mmap_min_addr and mmap_end boundaries and fixes the above issue.

Signed-off-by Alexander Gordeev <[email protected]>
---
 mm/mmap.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)
mm/mmap.c +14 -10
index a0fcb5ca0e06..bd673a01b1d0 100644
@@ -59,6 +59,14 @@
 #define arch_mmap_check(addr, len, flags)	(0)
 #endif
 
+#ifndef arch_get_mmap_end
+#define arch_get_mmap_end(addr)	(TASK_SIZE)
+#endif
+
+#ifndef arch_get_mmap_base
+#define arch_get_mmap_base(addr, base) (base)
+#endif
+
 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
 const int mmap_rnd_bits_min = CONFIG_ARCH_MMAP_RND_BITS_MIN;
 const int mmap_rnd_bits_max = CONFIG_ARCH_MMAP_RND_BITS_MAX;

@@ -1366,6 +1374,7 @@ unsigned long do_mmap(struct file *file, unsigned long addr,
 			unsigned long pgoff, unsigned long *populate,
 			struct list_head *uf)
 {
+	const unsigned long mmap_end = arch_get_mmap_end(addr);
 	struct mm_struct *mm = current->mm;
 	int pkey = 0;
 

@@ -1388,8 +1397,12 @@ unsigned long do_mmap(struct file *file, unsigned long addr,
 	if (flags & MAP_FIXED_NOREPLACE)
 		flags |= MAP_FIXED;
 
-	if (!(flags & MAP_FIXED))
+	if (flags & MAP_FIXED) {
+		if ((addr < mmap_min_addr) || (addr > mmap_end))
+			return -ENOMEM;
+	} else {
 		addr = round_hint_to_min(addr);
+	}
 
 	/* Careful about overflows.. */
 	len = PAGE_ALIGN(len);

@@ -2050,15 +2063,6 @@ unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info)
 	return gap_end;
 }
 
-
-#ifndef arch_get_mmap_end
-#define arch_get_mmap_end(addr)	(TASK_SIZE)
-#endif
-
-#ifndef arch_get_mmap_base
-#define arch_get_mmap_base(addr, base) (base)
-#endif
-
 /* Get an address range which is currently unmapped.
  * For shmat() with addr=0.
  *
Signature · 1 line
2.23.0
maintainer
#3 lore raw

Re: [PATCH 2/2] mm/mmap.c: do not allow mappings outside of allowed limits

Quoted 4 lines · On Mon, 23 Mar 2020 14:29:29 +0100 [email protected] wrote:

It is possible to request a fixed mapping address below mmap_min_addr and succeed. This update adds early checks of mmap_min_addr and mmap_end boundaries and fixes the above issue.

Does this solve any known problems? If not, what is the motivation for the change?

author
#4 lore raw

Re: [PATCH 2/2] mm/mmap.c: do not allow mappings outside of allowed limits

Quoted 9 lines · Andrew Morton
Quoted 4 lines · On Mon, 23 Mar 2020 14:29:29 +0100 [email protected] wrote:

It is possible to request a fixed mapping address below mmap_min_addr and succeed. This update adds early checks of mmap_min_addr and mmap_end boundaries and fixes the above issue.

Does this solve any known problems? If not, what is the motivation for the change?

One can set a lowest possible address in /proc/sys/vm/mmap_min_addr and mmap below that bound nevertheless. Apart from it is wrong I am not aware of any existing issue.

Thanks!