Skip to contentSkip to conversation

mm: Account anon mappings as RLIMIT_DATA

RFC In review Cyrill Gorcunov 2 participants 3 messages Dec 11, 2015

Subsystems (4)
Memory management MM core Proc filesystem Memory mapping
lore
Share & tools
lore

Cyrill Gorcunov

author
#0 lore raw

RFC 1/1 mm: Account anon mappings as RLIMIT_DATA

When inspecting a vague code inside prctl(PR_SET_MM_MEM) call (which testing the RLIMIT_DATA value to figure out if we're allowed to assign new @start_brk, @brk, @start_data, @end_data from mm_struct) it's been commited that RLIMIT_DATA in a form it's implemented now doesn't do anything useful because most of user-space libraries use mmap() syscall for dynamic memory allocations.

So Linus suggested to convert RLIMIT_DATA rlimit into something suitable for anonymous memory accounting. Here we introduce new @anon_vm member into mm descriptor which is updated every vm_stat_account call. When mmap_region() is called we test if current RLIMIT_DATA limit is not exceeded.

This should give a way to control the amount of anonymous memory allocated.

p.s.: This rfc not for application, I would like to hear if I move in right direction in general. In particular I need to add hugetlb anon mappings here as well and update docs and etc.

So comments are highly appreciated.

CC Quentin Casasnovas <[email protected]>
CC Vegard Nossum <[email protected]>
CC Linus Torvalds <[email protected]>
CC Willy Tarreau <[email protected]>
CC Andy Lutomirski <[email protected]>
CC Kees Cook <[email protected]>
CC Vladimir Davydov <[email protected]>
CC Konstantin Khlebnikov <[email protected]>
CC Pavel Emelyanov <[email protected]>
CC Vladimir Davydov <[email protected]>
CC Peter Zijlstra <[email protected]>
Signed-off-by Cyrill Gorcunov <[email protected]>
---
 fs/proc/task_mmu.c       |    2 ++
 include/linux/mm_types.h |    1 +
 mm/mmap.c                |   22 ++++++++++++++++++++++
 3 files changed, 25 insertions(+)
fs/proc/task_mmu.c +2 -0
Index: linux-ml.git/fs/proc/task_mmu.c
===================================================================
@@ -53,6 +53,7 @@ void task_mem(struct seq_file *m, struct
 		"VmHWM:\t%8lu kB\n"
 		"VmRSS:\t%8lu kB\n"
 		"VmData:\t%8lu kB\n"
+		"VmAnon:\t%8lu kB\n"
 		"VmStk:\t%8lu kB\n"
 		"VmExe:\t%8lu kB\n"
 		"VmLib:\t%8lu kB\n"

@@ -66,6 +67,7 @@ void task_mem(struct seq_file *m, struct
 		hiwater_rss << (PAGE_SHIFT-10),
 		total_rss << (PAGE_SHIFT-10),
 		data << (PAGE_SHIFT-10),
+		mm->anon_vm << (PAGE_SHIFT-10),
 		mm->stack_vm << (PAGE_SHIFT-10), text, lib,
 		ptes >> 10,
 		pmds >> 10,
include/linux/mm_types.h +1 -0
Index: linux-ml.git/include/linux/mm_types.h
===================================================================
@@ -429,6 +429,7 @@ struct mm_struct {
 	unsigned long shared_vm;	/* Shared pages (files) */
 	unsigned long exec_vm;		/* VM_EXEC & ~VM_WRITE */
 	unsigned long stack_vm;		/* VM_GROWSUP/DOWN */
+	unsigned long anon_vm;		/* Anonymous pages mapped */
 	unsigned long def_flags;
 	unsigned long start_code, end_code, start_data, end_data;
 	unsigned long start_brk, brk, start_stack;
mm/mmap.c +22 -0
Index: linux-ml.git/mm/mmap.c
===================================================================
@@ -1214,6 +1214,8 @@ void vm_stat_account(struct mm_struct *m
 {
 	const unsigned long stack_flags
 		= VM_STACK_FLAGS & (VM_GROWSUP|VM_GROWSDOWN);
+	const unsigned long not_anon_acc
+		= VM_GROWSUP | VM_GROWSDOWN | VM_SHARED | VM_MAYSHARE;
 
 	mm->total_vm += pages;
 

@@ -1223,6 +1225,9 @@ void vm_stat_account(struct mm_struct *m
 			mm->exec_vm += pages;
 	} else if (flags & stack_flags)
 		mm->stack_vm += pages;
+
+	if (!file && (flags & not_anon_acc) == 0)
+		mm->anon_vm += pages;
 }
 #endif /* CONFIG_PROC_FS */
 

@@ -1534,6 +1539,13 @@ static inline int accountable_mapping(st
 	return (vm_flags & (VM_NORESERVE | VM_SHARED | VM_WRITE)) == VM_WRITE;
 }
 
+static inline int anon_accountable_mapping(struct file *file, vm_flags_t vm_flags)
+{
+	return !file &&
+		(vm_flags & (VM_GROWSDOWN | VM_GROWSUP |
+			     VM_SHARED | VM_MAYSHARE)) == 0;
+}
+
 unsigned long mmap_region(struct file *file, unsigned long addr,
 		unsigned long len, vm_flags_t vm_flags, unsigned long pgoff)
 {

@@ -1578,6 +1590,16 @@ unsigned long mmap_region(struct file *f
 	}
 
 	/*
+	 * For anon mappings make sure we don't exceed the limit.
+	 */
+	if (anon_accountable_mapping(file, vm_flags)) {
+		unsigned long lim = rlimit(RLIMIT_DATA) >> PAGE_SHIFT;
+
+		if (lim < (mm->anon_vm + (len >> PAGE_SHIFT)))
+			return -ENOMEM;
+	}
+
+	/*
 	 * Can we just expand an old mapping?
 	 */
 	vma = vma_merge(mm, prev, addr, addr + len, vm_flags,

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#1 lore raw
Quoted 3 lines · Cyrill Gorcunov

This should give a way to control the amount of anonymous memory allocated.

This looks good to me, assuming it gets testing. I think we could add the hugetlb stuff later, I think it's a separate improvement.

Small nit:

Quoted 20 lines
mm/mmap.c +5 -0
@@ -1214,6 +1214,8 @@ void vm_stat_account(struct mm_struct *m
 {
        const unsigned long stack_flags
                = VM_STACK_FLAGS & (VM_GROWSUP|VM_GROWSDOWN);
+       const unsigned long not_anon_acc
+               = VM_GROWSUP | VM_GROWSDOWN | VM_SHARED | VM_MAYSHARE;
 
        mm->total_vm += pages;
 

@@ -1223,6 +1225,9 @@ void vm_stat_account(struct mm_struct *m
                        mm->exec_vm += pages;
        } else if (flags & stack_flags)
                mm->stack_vm += pages;
+
+       if (!file && (flags & not_anon_acc) == 0)
+               mm->anon_vm += pages;
 }
 #endif /* CONFIG_PROC_FS */

mm/mmap.c static inline int accountable_mapping(st · line 1547

Quoted 9 lines
mm/mmap.c +6 -0
@@ -1534,6 +1539,13 @@ static inline int accountable_mapping(st
        return (vm_flags & (VM_NORESERVE | VM_SHARED | VM_WRITE)) == VM_WRITE;
 }
 
+static inline int anon_accountable_mapping(struct file *file, vm_flags_t vm_flags)
+{
+       return !file &&
+               (vm_flags & (VM_GROWSDOWN | VM_GROWSUP |
+                            VM_SHARED | VM_MAYSHARE)) == 0;
+}

You're duplicating that "is it an anon accountable mapping" logic. I think you should move the inline helper function up, and use it in vm_stat_account().

Other than that, I think the patch certainly looks clean and obvious enough. But I didn't actually try to *run* it, maybe it ends up not working due to something I don't see.

Linus

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

author
#2 lore raw
Quoted 12 lines · Linus Torvalds

mm/mmap.c static inline int accountable_mapping(st · line 1547

Quoted 7 lines
mm/mmap.c +6 -0
+static inline int anon_accountable_mapping(struct file *file, vm_flags_t vm_flags)
+{
+       return !file &&
+               (vm_flags & (VM_GROWSDOWN | VM_GROWSUP |
+                            VM_SHARED | VM_MAYSHARE)) == 0;
+}

You're duplicating that "is it an anon accountable mapping" logic. I think you should move the inline helper function up, and use it in vm_stat_account().

Quoted 3 lines

Other than that, I think the patch certainly looks clean and obvious enough. But I didn't actually try to *run* it, maybe it ends up not working due to something I don't see.

Thank for the note, Linus! I tried to play with it a little bit, but I need to test it more first. I sent it simply to be sure that i'm moving in right direction. Once I get reviewed all things again and test it more deeply, I'll send a new patch.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>