Skip to contentSkip to conversation

mm: process/cgroup ksm support

PATCH Superseded Stefan Roesch 3 participants 17 messages Apr 12, 2023 – Apr 13, 2023

Subsystems (12)
Memory management Documentation Kselftest MM core +8 more subsystems

v6 of 9 Superseded Latest is v9 →

lore
Share & tools
lore

Stefan Roesch

author
#0 lore raw

v6 cover mm: process/cgroup ksm support

So far KSM can only be enabled by calling madvise for memory regions. To be able to use KSM for more workloads, KSM needs to have the ability to be enabled / disabled at the process / cgroup level.

Use case 1:
The madvise call is not available in the programming language. An example for this are programs with forked workloads using a garbage collected language without pointers. In such a language madvise cannot be made available.

In addition the addresses of objects get moved around as they are garbage collected. KSM sharing needs to be enabled "from the outside" for these type of workloads.

Use case 2:
The same interpreter can also be used for workloads where KSM brings no benefit or even has overhead. We'd like to be able to enable KSM on a workload by workload basis.

Use case 3:
With the madvise call sharing opportunities are only enabled for the current
process: it is a workload-local decision. A considerable number of sharing opportunities may exist across multiple workloads or jobs (if they are part of the same security domain). Only a higler level entity like a job scheduler or container can know for certain if its running one or more instances of a job. That job scheduler however doesn't have the necessary internal workload knowledge to make targeted madvise calls.

Security concerns:
In previous discussions security concerns have been brought up. The problem is that an individual workload does not have the knowledge about what else is running on a machine. Therefore it has to be very conservative in what memory areas can be shared or not. However, if the system is dedicated to running multiple jobs within the same security domain, its the job scheduler that has the knowledge that sharing can be safely enabled and is even desirable.

Performance:
Experiments with using UKSM have shown a capacity increase of around 20%.

1. New options for prctl system command
This patch series adds two new options to the prctl system call. The first one allows to enable KSM at the process level and the second one to query the setting.

The setting will be inherited by child processes.

With the above setting, KSM can be enabled for the seed process of a cgroup and all processes in the cgroup will inherit the setting.

2. Changes to KSM processing
When KSM is enabled at the process level, the KSM code will iterate over all the VMA's and enable KSM for the eligible VMA's.

When forking a process that has KSM enabled, the setting will be inherited by the new child process.

3. Add general_profit metric
The general_profit metric of KSM is specified in the documentation, but not calculated. This adds the general profit metric to /sys/kernel/debug/mm/ksm.

4. Add more metrics to ksm_stat
This adds the process profit metric to /proc/<pid>/ksm_stat.

5. Add more tests to ksm_tests
This adds an option to specify the merge type to the ksm_tests. This allows to test madvise and prctl KSM. It also adds a new option to query if prctl KSM has been enabled. It adds a fork test to verify that the KSM process setting is inherited by client processes.

Changes:
- V6:
- Fix error condition in prctl call
- Remove ksm_merge_type function and ksm_stat output
- Some minor changes like whitespace and removing a cast.

- V5:
- When the prctl system call is invoked, mark all compatible VMA as mergeable
- Instead of checcking during scan if VMA is mergeable, mark the VMA mergeable when the VMA is created (in case the VMA is compatible)
- Remove earlier changes, they are no longer necessary
- Unset the flag MMF_VM_MERGE_ANY in gmap_mark_unmergeable().
- When unsetting the MMF_VM_MERGE_ANY flag with prctl, only unset the flag
- Remove pages_volatile function (with the simplar general_profit calculation, the function is no longer needed)
- Use simpler formula for calculation of general_profit

- V4:
- removing check in prctl for MMF_VM_MERGEABLE in PR_SET_MEMORY_MERGE handling
- Checking for VM_MERGEABLE AND MMF_VM_MERGE_ANY to avoid chaning vm_flags
- This requires also checking that the vma is compatible. The compatibility check is provided by a new helper
- processes which have set MMF_VM_MERGE_ANY, only need to call the helper and not madvise.
- removed unmerge_vmas function, this function is no longer necessary, clearing the MMF_VM_MERGE_ANY bit is sufficient

- V3:
- folded patch 1 - 6
- folded patch 7 - 14
- folded patch 15 - 19
- Expanded on the use cases in the cover letter
- Added a section on security concerns to the cover letter

- V2:
- Added use cases to the cover letter
- Removed the tracing patch from the patch series and posted it as an individual patch
- Refreshed repo

Stefan Roesch (3):
mm: add new api to enable ksm per process
mm: add new KSM process and sysfs knobs
selftests/mm: add new selftests for KSM

 Documentation/ABI/testing/sysfs-kernel-mm-ksm |   8 +
 Documentation/admin-guide/mm/ksm.rst          |   5 +-
 arch/s390/mm/gmap.c                           |   1 +
 fs/proc/base.c                                |   3 +
 include/linux/ksm.h                           |  27 +-
 include/linux/sched/coredump.h                |   1 +
 include/uapi/linux/prctl.h                    |   2 +
 kernel/fork.c                                 |   1 +
 kernel/sys.c                                  |  23 ++
 mm/ksm.c                                      | 132 +++++++--
 mm/mmap.c                                     |   7 +
 tools/include/uapi/linux/prctl.h              |   2 +
 tools/testing/selftests/mm/Makefile           |   2 +-
 tools/testing/selftests/mm/ksm_tests.c        | 254 +++++++++++++++---
 14 files changed, 400 insertions(+), 68 deletions(-)
Signature · 1 line
2.31.1
author
#1 lore raw

PATCH v6 1/3 mm: add new api to enable ksm per process

So far KSM can only be enabled by calling madvise for memory regions. To be able to use KSM for more workloads, KSM needs to have the ability to be enabled / disabled at the process / cgroup level.

1. New options for prctl system command

This patch series adds two new options to the prctl system call. The first one allows to enable KSM at the process level and the second one to query the setting.

The setting will be inherited by child processes.

With the above setting, KSM can be enabled for the seed process of a cgroup and all processes in the cgroup will inherit the setting.

2. Changes to KSM processing

When KSM is enabled at the process level, the KSM code will iterate over all the VMA's and enable KSM for the eligible VMA's.

When forking a process that has KSM enabled, the setting will be inherited by the new child process.

1) Introduce new MMF_VM_MERGE_ANY flag

This introduces the new flag MMF_VM_MERGE_ANY flag. When this flag is set, kernel samepage merging (ksm) gets enabled for all vma's of a process.

2) Setting VM_MERGEABLE on VMA creation

When a VMA is created, if the MMF_VM_MERGE_ANY flag is set, the VM_MERGEABLE flag will be set for this VMA.

3) add flag to __ksm_enter

This change adds the flag parameter to __ksm_enter. This allows to distinguish if ksm was called by prctl or madvise.

4) add flag to __ksm_exit call

This adds the flag parameter to the __ksm_exit() call. This allows to distinguish if this call is for an prctl or madvise invocation.

5) support disabling of ksm for a process

This adds the ability to disable ksm for a process if ksm has been enabled for the process.

6) add new prctl option to get and set ksm for a process

This adds two new options to the prctl system call
- enable ksm for all vmas of a process (if the vmas support it).
- query if ksm has been enabled for a process.

3. Disabling MMF_VM_MERGE_ANY for storage keys in s390

In the s390 architecture when storage keys are used, the MMF_VM_MERGE_ANY will be disabled.

Signed-off-by Stefan Roesch <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Johannes Weiner <[email protected]>
Cc Michal Hocko <[email protected]>
Cc Rik van Riel <[email protected]>
Cc Bagas Sanjaya <[email protected]>
Signed-off-by Andrew Morton <[email protected]>
---
 arch/s390/mm/gmap.c            |   1 +
 include/linux/ksm.h            |  23 +++++--
 include/linux/sched/coredump.h |   1 +
 include/uapi/linux/prctl.h     |   2 +
 kernel/fork.c                  |   1 +
 kernel/sys.c                   |  23 +++++++
 mm/ksm.c                       | 111 ++++++++++++++++++++++++++-------
 mm/mmap.c                      |   7 +++
 8 files changed, 142 insertions(+), 27 deletions(-)
arch/s390/mm/gmap.c +1 -0
index 5a716bdcba05..9d85e5589474 100644
@@ -2591,6 +2591,7 @@ int gmap_mark_unmergeable(void)
 	int ret;
 	VMA_ITERATOR(vmi, mm, 0);
 
+	clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
 	for_each_vma(vmi, vma) {
 		/* Copy vm_flags to avoid partial modifications in ksm_madvise */
 		vm_flags = vma->vm_flags;
include/linux/ksm.h +18 -5
index 7e232ba59b86..f24f9faf1561 100644
@@ -18,20 +18,29 @@
 #ifdef CONFIG_KSM
 int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
 		unsigned long end, int advice, unsigned long *vm_flags);
-int __ksm_enter(struct mm_struct *mm);
-void __ksm_exit(struct mm_struct *mm);
+
+int ksm_add_mm(struct mm_struct *mm);
+void ksm_add_vma(struct vm_area_struct *vma);
+void ksm_add_vmas(struct mm_struct *mm);
+
+int __ksm_enter(struct mm_struct *mm, int flag);
+void __ksm_exit(struct mm_struct *mm, int flag);
 
 static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
 {
+	if (test_bit(MMF_VM_MERGE_ANY, &oldmm->flags))
+		return ksm_add_mm(mm);
 	if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags))
-		return __ksm_enter(mm);
+		return __ksm_enter(mm, MMF_VM_MERGEABLE);
 	return 0;
 }
 
 static inline void ksm_exit(struct mm_struct *mm)
 {
-	if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
-		__ksm_exit(mm);
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_exit(mm, MMF_VM_MERGE_ANY);
+	else if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
+		__ksm_exit(mm, MMF_VM_MERGEABLE);
 }
 
 /*

@@ -53,6 +62,10 @@ void folio_migrate_ksm(struct folio *newfolio, struct folio *folio);
 
 #else  /* !CONFIG_KSM */
 
+static inline void ksm_add_vma(struct vm_area_struct *vma)
+{
+}
+
 static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
 {
 	return 0;
include/linux/sched/coredump.h +1 -0
index 0e17ae7fbfd3..0ee96ea7a0e9 100644
@@ -90,4 +90,5 @@ static inline int get_dumpable(struct mm_struct *mm)
 #define MMF_INIT_MASK		(MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK |\
 				 MMF_DISABLE_THP_MASK | MMF_HAS_MDWE_MASK)
 
+#define MMF_VM_MERGE_ANY	29
 #endif /* _LINUX_SCHED_COREDUMP_H */
include/uapi/linux/prctl.h +2 -0
index 1312a137f7fb..759b3f53e53f 100644
@@ -290,4 +290,6 @@ struct prctl_mm_map {
 #define PR_SET_VMA		0x53564d41
 # define PR_SET_VMA_ANON_NAME		0
 
+#define PR_SET_MEMORY_MERGE		67
+#define PR_GET_MEMORY_MERGE		68
 #endif /* _LINUX_PRCTL_H */
kernel/fork.c +1 -0
index f68954d05e89..1520697cf6c7 100644
@@ -686,6 +686,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
 		if (vma_iter_bulk_store(&vmi, tmp))
 			goto fail_nomem_vmi_store;
 
+		ksm_add_vma(tmp);
 		mm->map_count++;
 		if (!(tmp->vm_flags & VM_WIPEONFORK))
 			retval = copy_page_range(tmp, mpnt);
kernel/sys.c +23 -0
index 495cd87d9bf4..9bba163d2d04 100644
@@ -15,6 +15,7 @@
 #include <linux/highuid.h>
 #include <linux/fs.h>
 #include <linux/kmod.h>
+#include <linux/ksm.h>
 #include <linux/perf_event.h>
 #include <linux/resource.h>
 #include <linux/kernel.h>

@@ -2661,6 +2662,28 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
 	case PR_SET_VMA:
 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
 		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
+			int err = ksm_add_mm(me->mm);
+
+			if (!err)
+				ksm_add_vmas(me->mm);
+		} else {
+			clear_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		}
+		mmap_write_unlock(me->mm);
+		break;
+	case PR_GET_MEMORY_MERGE:
+		if (arg2 || arg3 || arg4 || arg5)
+			return -EINVAL;
+
+		error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		break;
+#endif
 	default:
 		error = -EINVAL;
 		break;
mm/ksm.c +89 -22

173 lines not on this page · raw message

mm/mmap.c +7 -0
index 740b54be3ed4..483e182e0b9d 100644
@@ -46,6 +46,7 @@
 #include <linux/pkeys.h>
 #include <linux/oom.h>
 #include <linux/sched/mm.h>
+#include <linux/ksm.h>
 
 #include <linux/uaccess.h>
 #include <asm/cacheflush.h>

@@ -2213,6 +2214,8 @@ int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma,
 	/* vma_complete stores the new vma */
 	vma_complete(&vp, vmi, vma->vm_mm);
 
+	ksm_add_vma(new);
+
 	/* Success. */
 	if (new_below)
 		vma_next(vmi);

@@ -2664,6 +2667,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
 	if (file && vm_flags & VM_SHARED)
 		mapping_unmap_writable(file->f_mapping);
 	file = vma->vm_file;
+	ksm_add_vma(vma);
 expanded:
 	perf_event_mmap(vma);
 

@@ -2936,6 +2940,7 @@ static int do_brk_flags(struct vma_iterator *vmi, struct vm_area_struct *vma,
 		goto mas_store_fail;
 
 	mm->map_count++;
+	ksm_add_vma(vma);
 out:
 	perf_event_mmap(vma);
 	mm->total_vm += len >> PAGE_SHIFT;

@@ -3180,6 +3185,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
 		if (vma_link(mm, new_vma))
 			goto out_vma_link;
 		*need_rmap_locks = false;
+		ksm_add_vma(new_vma);
 	}
 	validate_mm_mt(mm);
 	return new_vma;

@@ -3356,6 +3362,7 @@ static struct vm_area_struct *__install_special_mapping(
 	vm_stat_account(mm, vma->vm_flags, len >> PAGE_SHIFT);
 
 	perf_event_mmap(vma);
+	ksm_add_vma(vma);
 
 	validate_mm_mt(mm);
 	return vma;
Signature · 1 line
2.31.1
#2 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

kernel/sys.c SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, · line 2674

Quoted 13 lines · Stefan Roesch
kernel/sys.c +10 -0
 	case PR_SET_VMA:
 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
 		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
+			int err = ksm_add_mm(me->mm);
+
+			if (!err)
+				ksm_add_vmas(me->mm);

in the last version of this patch, you reported the error. Now you swallow the error. I have no idea which is correct, but you've changed the behaviour without explaining it, so I assume it's wrong.

kernel/sys.c SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, · line 2685

Quoted 11 lines
kernel/sys.c +11 -0
+		} else {
+			clear_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		}
+		mmap_write_unlock(me->mm);
+		break;
+	case PR_GET_MEMORY_MERGE:
+		if (arg2 || arg3 || arg4 || arg5)
+			return -EINVAL;
+
+		error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		break;

Why do we need a GET? Just for symmetry, or is there an actual need for it?

author
#3 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

Quoted 10 lines · Matthew Wilcox
Quoted 9 lines · Stefan Roesch
kernel/sys.c +6 -0
 	case PR_SET_VMA:
 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
 		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
Quoted 9 lines

kernel/sys.c SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, · line 2674

Quoted 4 lines
kernel/sys.c +4 -0
+			int err = ksm_add_mm(me->mm);
+
+			if (!err)
+				ksm_add_vmas(me->mm);

in the last version of this patch, you reported the error. Now you swallow the error. I have no idea which is correct, but you've changed the behaviour without explaining it, so I assume it's wrong.

I don't see how the error is swallowed in the arg2 case. If there is an error ksm_add_vmas is not executedd and at the end of the function the error is returned. Am I missing something?

Quoted 6 lines
Quoted 6 lines
kernel/sys.c +6 -0
+		} else {
+			clear_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		}
+		mmap_write_unlock(me->mm);
+		break;
+	case PR_GET_MEMORY_MERGE:
Quoted 8 lines

kernel/sys.c SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, · line 2685

Quoted 5 lines
kernel/sys.c +5 -0
+		if (arg2 || arg3 || arg4 || arg5)
+			return -EINVAL;
+
+		error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		break;

Why do we need a GET? Just for symmetry, or is there an actual need for it?

There are three reasons:
- For symmetry
- The ksm sharing is inherited by child processes. This allows the test programs to verify that this is working.
- For child processes it might be useful to have the ability to check if ksm sharing has been enabled

#4 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

Quoted 23 lines · Stefan Roesch
Quoted 10 lines · Matthew Wilcox
Quoted 9 lines · Stefan Roesch
kernel/sys.c +6 -0
 	case PR_SET_VMA:
 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
 		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
Quoted 9 lines

kernel/sys.c SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, · line 2674

Quoted 4 lines
kernel/sys.c +4 -0
+			int err = ksm_add_mm(me->mm);
+
+			if (!err)
+				ksm_add_vmas(me->mm);

in the last version of this patch, you reported the error. Now you swallow the error. I have no idea which is correct, but you've changed the behaviour without explaining it, so I assume it's wrong.

Quoted 3 lines

I don't see how the error is swallowed in the arg2 case. If there is an error ksm_add_vmas is not executedd and at the end of the function the error is returned. Am I missing something?

You said 'int err' which declares a new variable. If you want it reported, just use 'error = ksm_add_mm(me->mm);'.

#5 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

In reply to Stefan Roesch

[...]

Thanks for giving mu sugegstions a churn. I think we can further improve/simplify some things. I added some comments, but might have more regarding MMF_VM_MERGE_ANY / MMF_VM_MERGEABLE.

[I'll try reowkring your patch after I send this mail to play with some simplifications]

Quoted 15 lines
  arch/s390/mm/gmap.c            |   1 +
  include/linux/ksm.h            |  23 +++++--
  include/linux/sched/coredump.h |   1 +
  include/uapi/linux/prctl.h     |   2 +
  kernel/fork.c                  |   1 +
  kernel/sys.c                   |  23 +++++++
  mm/ksm.c                       | 111 ++++++++++++++++++++++++++-------
  mm/mmap.c                      |   7 +++
  8 files changed, 142 insertions(+), 27 deletions(-)
arch/s390/mm/gmap.c +0 -0
index 5a716bdcba05..9d85e5589474 100644

arch/s390/mm/gmap.c int gmap_mark_unmergeable(void) · line 2594

Quoted 4 lines
arch/s390/mm/gmap.c +1 -0
@@ -2591,6 +2591,7 @@ int gmap_mark_unmergeable(void)
  	int ret;
  	VMA_ITERATOR(vmi, mm, 0);
 
+	clear_bit(MMF_VM_MERGE_ANY, &mm->flags);

Okay, that should keep the existing mechanism working. (but users can still mess it up)

Might be worth a comment

/*
* Make sure to disable KSM (if enabled for the whole process or
* individual VMAs). Note that nothing currently hinders user space
* from re-enabling it.
*/

Quoted 15 lines
arch/s390/mm/gmap.c +0 -0
  	for_each_vma(vmi, vma) {
  		/* Copy vm_flags to avoid partial modifications in ksm_madvise */
  		vm_flags = vma->vm_flags;
include/linux/ksm.h +2 -2
index 7e232ba59b86..f24f9faf1561 100644
@@ -18,20 +18,29 @@
  #ifdef CONFIG_KSM
  int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
  		unsigned long end, int advice, unsigned long *vm_flags);
-int __ksm_enter(struct mm_struct *mm);
-void __ksm_exit(struct mm_struct *mm);
+
+int ksm_add_mm(struct mm_struct *mm);

include/linux/ksm.h line 32

Quoted 10 lines
include/linux/ksm.h +7 -0
+void ksm_add_vma(struct vm_area_struct *vma);
+void ksm_add_vmas(struct mm_struct *mm);
+
+int __ksm_enter(struct mm_struct *mm, int flag);
+void __ksm_exit(struct mm_struct *mm, int flag);
 
  static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
  {
+	if (test_bit(MMF_VM_MERGE_ANY, &oldmm->flags))
+		return ksm_add_mm(mm);

ksm_fork() runs before copying any VMAs. Copying the bit should be sufficient.

Would it be possible to rework to something like:

if (test_bit(MMF_VM_MERGE_ANY, &oldmm->flags))

	set_bit(MMF_VM_MERGE_ANY, &mm->flags)
if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags))
	return __ksm_enter(mm);

work? IOW, not exporting ksm_add_mm() and not passing a flag to __ksm_enter() -- it would simply set MMF_VM_MERGEABLE ?

I rememebr proposing that enabling MMF_VM_MERGE_ANY would simply enable MMF_VM_MERGEABLE.

Quoted 4 lines
include/linux/ksm.h +1 -1
  	if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags))
-		return __ksm_enter(mm);
+		return __ksm_enter(mm, MMF_VM_MERGEABLE);
  	return 0;

include/linux/ksm.h line 43

Quoted 10 lines
include/linux/ksm.h +4 -2
  }
 
  static inline void ksm_exit(struct mm_struct *mm)
  {
-	if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
-		__ksm_exit(mm);
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_exit(mm, MMF_VM_MERGE_ANY);
+	else if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
+		__ksm_exit(mm, MMF_VM_MERGEABLE);

Can we do

if (test_bit(MMF_VM_MERGEABLE, &mm->flags))

	__ksm_exit(mm);

And simply let __ksm_exit() clear both bits?

Quoted 7 lines
include/linux/ksm.h +0 -0
  }
 
  /*

@@ -53,6 +62,10 @@ void folio_migrate_ksm(struct folio *newfolio, struct folio *folio);
 
  #else  /* !CONFIG_KSM */

[...]

Quoted 17 lines
  #endif /* _LINUX_SCHED_COREDUMP_H */
include/uapi/linux/prctl.h +2 -0
index 1312a137f7fb..759b3f53e53f 100644
@@ -290,4 +290,6 @@ struct prctl_mm_map {
  #define PR_SET_VMA		0x53564d41
  # define PR_SET_VMA_ANON_NAME		0
 
+#define PR_SET_MEMORY_MERGE		67
+#define PR_GET_MEMORY_MERGE		68
  #endif /* _LINUX_PRCTL_H */
kernel/fork.c +0 -0
index f68954d05e89..1520697cf6c7 100644

kernel/fork.c static __latent_entropy int dup_mmap(struct mm_struct *mm, · line 689

Quoted 4 lines
kernel/fork.c +1 -0
@@ -686,6 +686,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
  		if (vma_iter_bulk_store(&vmi, tmp))
  			goto fail_nomem_vmi_store;
 
+		ksm_add_vma(tmp);

Is this really required? The relevant VMAs should have VM_MERGEABLE set.

Quoted 21 lines
kernel/fork.c +0 -0
  		mm->map_count++;
  		if (!(tmp->vm_flags & VM_WIPEONFORK))
  			retval = copy_page_range(tmp, mpnt);
kernel/sys.c +3 -0
index 495cd87d9bf4..9bba163d2d04 100644
@@ -15,6 +15,7 @@
  #include <linux/highuid.h>
  #include <linux/fs.h>
  #include <linux/kmod.h>
+#include <linux/ksm.h>
  #include <linux/perf_event.h>
  #include <linux/resource.h>
  #include <linux/kernel.h>

@@ -2661,6 +2662,28 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
  	case PR_SET_VMA:
  		error = prctl_set_vma(arg2, arg3, arg4, arg5);
  		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:

kernel/sys.c SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, · line 2676

Quoted 10 lines
kernel/sys.c +10 -0
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
+			int err = ksm_add_mm(me->mm);
+
+			if (!err)
+				ksm_add_vmas(me->mm);
+		} else {
+			clear_bit(MMF_VM_MERGE_ANY, &me->mm->flags);

Okay, so disabling doesn't actually unshare anything.

Quoted 38 lines
kernel/sys.c +10 -0
+		}
+		mmap_write_unlock(me->mm);
+		break;
+	case PR_GET_MEMORY_MERGE:
+		if (arg2 || arg3 || arg4 || arg5)
+			return -EINVAL;
+
+		error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		break;
+#endif
  	default:
  		error = -EINVAL;
  		break;
mm/ksm.c +17 -0
index d7bd28199f6c..ab95ae0f9def 100644
@@ -534,10 +534,33 @@ static int break_ksm(struct vm_area_struct *vma, unsigned long addr,
  	return (ret & VM_FAULT_OOM) ? -ENOMEM : 0;
  }
 
+static bool vma_ksm_compatible(struct vm_area_struct *vma)
+{
+	if (vma->vm_flags & (VM_SHARED  | VM_MAYSHARE   | VM_PFNMAP  |
+			     VM_IO      | VM_DONTEXPAND | VM_HUGETLB |
+			     VM_MIXEDMAP))
+		return false;		/* just ignore the advice */
+
+	if (vma_is_dax(vma))
+		return false;
+
+#ifdef VM_SAO
+	if (vma->vm_flags & VM_SAO)
+		return false;
+#endif
+#ifdef VM_SPARC_ADI
+	if (vma->vm_flags & VM_SPARC_ADI)
+		return false;

mm/ksm.c static int break_ksm(struct vm_area_struct *vma, unsigned long addr, · line 563

Quoted 10 lines
mm/ksm.c +6 -0
+#endif
+
+	return true;
+}
+
  static struct vm_area_struct *find_mergeable_vma(struct mm_struct *mm,
  		unsigned long addr)
  {
  	struct vm_area_struct *vma;
+

unrelated change

Quoted 28 lines
mm/ksm.c +7 -0
  	if (ksm_test_exit(mm))
  		return NULL;
  	vma = vma_lookup(mm, addr);

@@ -1065,6 +1088,7 @@ static int unmerge_and_remove_all_rmap_items(void)
 
  			mm_slot_free(mm_slot_cache, mm_slot);
  			clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+			clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
  			mmdrop(mm);
  		} else
  			spin_unlock(&ksm_mmlist_lock);

@@ -2495,6 +2519,7 @@ static struct ksm_rmap_item *scan_get_next_rmap_item(struct page **page)
 
  		mm_slot_free(mm_slot_cache, mm_slot);
  		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+		clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
  		mmap_read_unlock(mm);
  		mmdrop(mm);
  	} else {

@@ -2571,6 +2596,63 @@ static int ksm_scan_thread(void *nothing)
  	return 0;
  }
 
+static void __ksm_add_vma(struct vm_area_struct *vma)
+{
+	unsigned long vm_flags = vma->vm_flags;
+
+	if (vm_flags & VM_MERGEABLE)

mm/ksm.c static int ksm_scan_thread(void *nothing) · line 2613

Quoted 10 lines
mm/ksm.c +10 -0
+		return;
+
+	if (vma_ksm_compatible(vma)) {
+		vm_flags |= VM_MERGEABLE;
+		vm_flags_reset(vma, vm_flags);
+	}
+}
+
+/**
+ * ksm_add_vma - Mark vma as mergeable

"if compatible"

Quoted 7 lines
mm/ksm.c +7 -0
+ *
+ * @vma:  Pointer to vma
+ */
+void ksm_add_vma(struct vm_area_struct *vma)
+{
+	struct mm_struct *mm = vma->vm_mm;
+

mm/ksm.c static int ksm_scan_thread(void *nothing) · line 2630

Quoted 10 lines
mm/ksm.c +10 -0
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_add_vma(vma);
+}
+
+/**
+ * ksm_add_vmas - Mark all vma's of a process as mergeable
+ *
+ * @mm:  Pointer to mm
+ */
+void ksm_add_vmas(struct mm_struct *mm)

I'd suggest calling this

Quoted 109 lines
mm/ksm.c +32 -22
+{
+	struct vm_area_struct *vma;
+
+	VMA_ITERATOR(vmi, mm, 0);
+	for_each_vma(vmi, vma)
+		__ksm_add_vma(vma);
+}
+
+/**
+ * ksm_add_mm - Add mm to mm ksm list
+ *
+ * @mm:  Pointer to mm
+ *
+ * Returns 0 on success, otherwise error code
+ */
+int ksm_add_mm(struct mm_struct *mm)
+{
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		return -EINVAL;
+	if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
+		return -EINVAL;
+
+	return __ksm_enter(mm, MMF_VM_MERGE_ANY);
+}
+
  int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
  		unsigned long end, int advice, unsigned long *vm_flags)
  {

@@ -2579,28 +2661,13 @@ int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
 
  	switch (advice) {
  	case MADV_MERGEABLE:
-		/*
-		 * Be somewhat over-protective for now!
-		 */
-		if (*vm_flags & (VM_MERGEABLE | VM_SHARED  | VM_MAYSHARE   |
-				 VM_PFNMAP    | VM_IO      | VM_DONTEXPAND |
-				 VM_HUGETLB | VM_MIXEDMAP))
-			return 0;		/* just ignore the advice */
-
-		if (vma_is_dax(vma))
+		if (vma->vm_flags & VM_MERGEABLE)
  			return 0;
-
-#ifdef VM_SAO
-		if (*vm_flags & VM_SAO)
+		if (!vma_ksm_compatible(vma))
  			return 0;
-#endif
-#ifdef VM_SPARC_ADI
-		if (*vm_flags & VM_SPARC_ADI)
-			return 0;
-#endif
 
  		if (!test_bit(MMF_VM_MERGEABLE, &mm->flags)) {
-			err = __ksm_enter(mm);
+			err = __ksm_enter(mm, MMF_VM_MERGEABLE);
  			if (err)
  				return err;
  		}

@@ -2626,7 +2693,7 @@ int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
  }
  EXPORT_SYMBOL_GPL(ksm_madvise);
 
-int __ksm_enter(struct mm_struct *mm)
+int __ksm_enter(struct mm_struct *mm, int flag)
  {
  	struct ksm_mm_slot *mm_slot;
  	struct mm_slot *slot;

@@ -2659,7 +2726,7 @@ int __ksm_enter(struct mm_struct *mm)
  		list_add_tail(&slot->mm_node, &ksm_scan.mm_slot->slot.mm_node);
  	spin_unlock(&ksm_mmlist_lock);
 
-	set_bit(MMF_VM_MERGEABLE, &mm->flags);
+	set_bit(flag, &mm->flags);
  	mmgrab(mm);
 
  	if (needs_wakeup)

@@ -2668,7 +2735,7 @@ int __ksm_enter(struct mm_struct *mm)
  	return 0;
  }
 
-void __ksm_exit(struct mm_struct *mm)
+void __ksm_exit(struct mm_struct *mm, int flag)
  {
  	struct ksm_mm_slot *mm_slot;
  	struct mm_slot *slot;

@@ -2700,7 +2767,7 @@ void __ksm_exit(struct mm_struct *mm)
 
  	if (easy_to_free) {
  		mm_slot_free(mm_slot_cache, mm_slot);
-		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+		clear_bit(flag, &mm->flags);
  		mmdrop(mm);
  	} else if (mm_slot) {
  		mmap_write_lock(mm);
mm/mmap.c +1 -0
index 740b54be3ed4..483e182e0b9d 100644
@@ -46,6 +46,7 @@
  #include <linux/pkeys.h>
  #include <linux/oom.h>
  #include <linux/sched/mm.h>
+#include <linux/ksm.h>
 
  #include <linux/uaccess.h>
  #include <asm/cacheflush.h>

mm/mmap.c int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma, · line 2218

Quoted 5 lines
mm/mmap.c +2 -0
@@ -2213,6 +2214,8 @@ int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma,
  	/* vma_complete stores the new vma */
  	vma_complete(&vp, vmi, vma->vm_mm);
 
+	ksm_add_vma(new);
+

Splitting a VMA shouldn't modify VM_MERGEABLE, so I assume this is not required?

Quoted 20 lines
mm/mmap.c +2 -0
  	/* Success. */
  	if (new_below)
  		vma_next(vmi);

@@ -2664,6 +2667,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
  	if (file && vm_flags & VM_SHARED)
  		mapping_unmap_writable(file->f_mapping);
  	file = vma->vm_file;
+	ksm_add_vma(vma);
  expanded:
  	perf_event_mmap(vma);
 

@@ -2936,6 +2940,7 @@ static int do_brk_flags(struct vma_iterator *vmi, struct vm_area_struct *vma,
  		goto mas_store_fail;
 
  	mm->map_count++;
+	ksm_add_vma(vma);
  out:
  	perf_event_mmap(vma);
  	mm->total_vm += len >> PAGE_SHIFT;

mm/mmap.c struct vm_area_struct *copy_vma(struct vm_area_struct **vmap, · line 3188

Quoted 4 lines
mm/mmap.c +1 -0
@@ -3180,6 +3185,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
  		if (vma_link(mm, new_vma))
  			goto out_vma_link;
  		*need_rmap_locks = false;
+		ksm_add_vma(new_vma);

Copying shouldn't modify VM_MERGEABLE, so I think this is not required?

mm/mmap.c static struct vm_area_struct *__install_special_mapping( · line 3365

Quoted 8 lines
mm/mmap.c +1 -0
  	}
  	validate_mm_mt(mm);
  	return new_vma;

@@ -3356,6 +3362,7 @@ static struct vm_area_struct *__install_special_mapping(
  	vm_stat_account(mm, vma->vm_flags, len >> PAGE_SHIFT);
 
  	perf_event_mmap(vma);
+	ksm_add_vma(vma);

IIUC, special mappings will never be considered a reasonable target for KSM (especially, because at least VM_DONTEXPAND is always set).

I think you can just drop this call.

Signature · 3 lines
Thanks,

David / dhildenb
author
#6 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

Quoted 30 lines · David Hildenbrand

[...]

Thanks for giving mu sugegstions a churn. I think we can further improve/simplify some things. I added some comments, but might have more regarding MMF_VM_MERGE_ANY / MMF_VM_MERGEABLE.

[I'll try reowkring your patch after I send this mail to play with some simplifications]

Quoted 14 lines
  arch/s390/mm/gmap.c            |   1 +
  include/linux/ksm.h            |  23 +++++--
  include/linux/sched/coredump.h |   1 +
  include/uapi/linux/prctl.h     |   2 +
  kernel/fork.c                  |   1 +
  kernel/sys.c                   |  23 +++++++
  mm/ksm.c                       | 111 ++++++++++++++++++++++++++-------
  mm/mmap.c                      |   7 +++
  8 files changed, 142 insertions(+), 27 deletions(-)
arch/s390/mm/gmap.c +0 -0
index 5a716bdcba05..9d85e5589474 100644

arch/s390/mm/gmap.c int gmap_mark_unmergeable(void) · line 2593

Quoted 3 lines
arch/s390/mm/gmap.c +0 -0
@@ -2591,6 +2591,7 @@ int gmap_mark_unmergeable(void)
  	int ret;
  	VMA_ITERATOR(vmi, mm, 0);
  +	clear_bit(MMF_VM_MERGE_ANY, &mm->flags);

Okay, that should keep the existing mechanism working. (but users can still mess it up)

Quoted 8 lines

Might be worth a comment

/*
* Make sure to disable KSM (if enabled for the whole process or
* individual VMAs). Note that nothing currently hinders user space
* from re-enabling it.
*/

I'll add the comment.

Quoted 27 lines
Quoted 14 lines
arch/s390/mm/gmap.c +0 -0
  	for_each_vma(vmi, vma) {
  		/* Copy vm_flags to avoid partial modifications in ksm_madvise */
  		vm_flags = vma->vm_flags;
include/linux/ksm.h +1 -2
index 7e232ba59b86..f24f9faf1561 100644
@@ -18,20 +18,29 @@
  #ifdef CONFIG_KSM
  int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
  		unsigned long end, int advice, unsigned long *vm_flags);
-int __ksm_enter(struct mm_struct *mm);
-void __ksm_exit(struct mm_struct *mm);
+

include/linux/ksm.h line 31

Quoted 10 lines
include/linux/ksm.h +8 -0
+int ksm_add_mm(struct mm_struct *mm);
+void ksm_add_vma(struct vm_area_struct *vma);
+void ksm_add_vmas(struct mm_struct *mm);
+
+int __ksm_enter(struct mm_struct *mm, int flag);
+void __ksm_exit(struct mm_struct *mm, int flag);
    static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
  {
+	if (test_bit(MMF_VM_MERGE_ANY, &oldmm->flags))
+		return ksm_add_mm(mm);

ksm_fork() runs before copying any VMAs. Copying the bit should be sufficient.

Quoted 7 lines

Would it be possible to rework to something like:

if (test_bit(MMF_VM_MERGE_ANY, &oldmm->flags))

	set_bit(MMF_VM_MERGE_ANY, &mm->flags)
if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags))
	return __ksm_enter(mm);

That will work.

Quoted 3 lines

work? IOW, not exporting ksm_add_mm() and not passing a flag to __ksm_enter() -- it would simply set MMF_VM_MERGEABLE ?

ksm_add_mm() is also used in prctl (kernel/sys.c). Do you want to make a similar change there?

Quoted 20 lines

I rememebr proposing that enabling MMF_VM_MERGE_ANY would simply enable MMF_VM_MERGEABLE.

include/linux/ksm.h line 43

Quoted 13 lines
include/linux/ksm.h +5 -3
  	if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags))
-		return __ksm_enter(mm);
+		return __ksm_enter(mm, MMF_VM_MERGEABLE);
  	return 0;
  }
    static inline void ksm_exit(struct mm_struct *mm)
  {
-	if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
-		__ksm_exit(mm);
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_exit(mm, MMF_VM_MERGE_ANY);
+	else if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
+		__ksm_exit(mm, MMF_VM_MERGEABLE);

Can we do

Quoted 5 lines

if (test_bit(MMF_VM_MERGEABLE, &mm->flags))

	__ksm_exit(mm);

And simply let __ksm_exit() clear both bits?

Yes, I'll make the change.

Quoted 21 lines
Quoted 5 lines
include/linux/ksm.h +0 -0
  }
    /*

@@ -53,6 +62,10 @@ void folio_migrate_ksm(struct folio *newfolio, struct folio *folio);
    #else  /* !CONFIG_KSM */

[...]

Quoted 13 lines
  #endif /* _LINUX_SCHED_COREDUMP_H */
include/uapi/linux/prctl.h +1 -0
index 1312a137f7fb..759b3f53e53f 100644
@@ -290,4 +290,6 @@ struct prctl_mm_map {
  #define PR_SET_VMA		0x53564d41
  # define PR_SET_VMA_ANON_NAME		0
  +#define PR_SET_MEMORY_MERGE		67
+#define PR_GET_MEMORY_MERGE		68
  #endif /* _LINUX_PRCTL_H */
kernel/fork.c +0 -0
index f68954d05e89..1520697cf6c7 100644
Quoted 9 lines

kernel/fork.c static __latent_entropy int dup_mmap(struct mm_struct *mm, · line 688

Quoted 6 lines
kernel/fork.c +0 -0
@@ -686,6 +686,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
  		if (vma_iter_bulk_store(&vmi, tmp))
  			goto fail_nomem_vmi_store;
  +		ksm_add_vma(tmp);

Is this really required? The relevant VMAs should have VM_MERGEABLE set.

I'll fix it.

Quoted 75 lines
Quoted 21 lines
kernel/fork.c +0 -0
  		mm->map_count++;
  		if (!(tmp->vm_flags & VM_WIPEONFORK))
  			retval = copy_page_range(tmp, mpnt);
kernel/sys.c +3 -0
index 495cd87d9bf4..9bba163d2d04 100644
@@ -15,6 +15,7 @@
  #include <linux/highuid.h>
  #include <linux/fs.h>
  #include <linux/kmod.h>
+#include <linux/ksm.h>
  #include <linux/perf_event.h>
  #include <linux/resource.h>
  #include <linux/kernel.h>

@@ -2661,6 +2662,28 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
  	case PR_SET_VMA:
  		error = prctl_set_vma(arg2, arg3, arg4, arg5);
  		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:

kernel/sys.c SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, · line 2676

Quoted 10 lines
kernel/sys.c +10 -0
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
+			int err = ksm_add_mm(me->mm);
+
+			if (!err)
+				ksm_add_vmas(me->mm);
+		} else {
+			clear_bit(MMF_VM_MERGE_ANY, &me->mm->flags);

Okay, so disabling doesn't actually unshare anything.

Quoted 41 lines
kernel/sys.c +10 -0
+		}
+		mmap_write_unlock(me->mm);
+		break;
+	case PR_GET_MEMORY_MERGE:
+		if (arg2 || arg3 || arg4 || arg5)
+			return -EINVAL;
+
+		error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		break;
+#endif
  	default:
  		error = -EINVAL;
  		break;
mm/ksm.c +20 -0
index d7bd28199f6c..ab95ae0f9def 100644
@@ -534,10 +534,33 @@ static int break_ksm(struct vm_area_struct *vma, unsigned long addr,
  	return (ret & VM_FAULT_OOM) ? -ENOMEM : 0;
  }
  +static bool vma_ksm_compatible(struct vm_area_struct *vma)
+{
+	if (vma->vm_flags & (VM_SHARED  | VM_MAYSHARE   | VM_PFNMAP  |
+			     VM_IO      | VM_DONTEXPAND | VM_HUGETLB |
+			     VM_MIXEDMAP))
+		return false;		/* just ignore the advice */
+
+	if (vma_is_dax(vma))
+		return false;
+
+#ifdef VM_SAO
+	if (vma->vm_flags & VM_SAO)
+		return false;
+#endif
+#ifdef VM_SPARC_ADI
+	if (vma->vm_flags & VM_SPARC_ADI)
+		return false;
+#endif
+
+	return true;
+}
Quoted 9 lines

mm/ksm.c static int break_ksm(struct vm_area_struct *vma, unsigned long addr, · line 563

Quoted 6 lines
mm/ksm.c +2 -0
+
  static struct vm_area_struct *find_mergeable_vma(struct mm_struct *mm,
  		unsigned long addr)
  {
  	struct vm_area_struct *vma;
+

unrelated change

Removed.

Quoted 29 lines
Quoted 29 lines
mm/ksm.c +10 -0
  	if (ksm_test_exit(mm))
  		return NULL;
  	vma = vma_lookup(mm, addr);

@@ -1065,6 +1088,7 @@ static int unmerge_and_remove_all_rmap_items(void)
    			mm_slot_free(mm_slot_cache, mm_slot);
  			clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+			clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
  			mmdrop(mm);
  		} else
  			spin_unlock(&ksm_mmlist_lock);

@@ -2495,6 +2519,7 @@ static struct ksm_rmap_item *scan_get_next_rmap_item(struct page **page)
    		mm_slot_free(mm_slot_cache, mm_slot);
  		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+		clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
  		mmap_read_unlock(mm);
  		mmdrop(mm);
  	} else {

@@ -2571,6 +2596,63 @@ static int ksm_scan_thread(void *nothing)
  	return 0;
  }
  +static void __ksm_add_vma(struct vm_area_struct *vma)
+{
+	unsigned long vm_flags = vma->vm_flags;
+
+	if (vm_flags & VM_MERGEABLE)
+		return;
+
+	if (vma_ksm_compatible(vma)) {
+		vm_flags |= VM_MERGEABLE;
Quoted 9 lines

mm/ksm.c static int ksm_scan_thread(void *nothing) · line 2613

Quoted 6 lines
mm/ksm.c +6 -0
+		vm_flags_reset(vma, vm_flags);
+	}
+}
+
+/**
+ * ksm_add_vma - Mark vma as mergeable

"if compatible"

I'll added the above.

Quoted 11 lines
Quoted 11 lines
mm/ksm.c +11 -0
+ *
+ * @vma:  Pointer to vma
+ */
+void ksm_add_vma(struct vm_area_struct *vma)
+{
+	struct mm_struct *mm = vma->vm_mm;
+
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_add_vma(vma);
+}
+
Quoted 9 lines

mm/ksm.c static int ksm_scan_thread(void *nothing) · line 2630

Quoted 6 lines
mm/ksm.c +6 -0
+/**
+ * ksm_add_vmas - Mark all vma's of a process as mergeable
+ *
+ * @mm:  Pointer to mm
+ */
+void ksm_add_vmas(struct mm_struct *mm)

I'd suggest calling this

I guess you forgot your name suggestion?

Quoted 99 lines
Quoted 99 lines
mm/ksm.c +32 -19
+{
+	struct vm_area_struct *vma;
+
+	VMA_ITERATOR(vmi, mm, 0);
+	for_each_vma(vmi, vma)
+		__ksm_add_vma(vma);
+}
+
+/**
+ * ksm_add_mm - Add mm to mm ksm list
+ *
+ * @mm:  Pointer to mm
+ *
+ * Returns 0 on success, otherwise error code
+ */
+int ksm_add_mm(struct mm_struct *mm)
+{
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		return -EINVAL;
+	if (test_bit(MMF_VM_MERGEABLE, &mm->flags))
+		return -EINVAL;
+
+	return __ksm_enter(mm, MMF_VM_MERGE_ANY);
+}
+
  int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
  		unsigned long end, int advice, unsigned long *vm_flags)
  {

@@ -2579,28 +2661,13 @@ int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
    	switch (advice) {
  	case MADV_MERGEABLE:
-		/*
-		 * Be somewhat over-protective for now!
-		 */
-		if (*vm_flags & (VM_MERGEABLE | VM_SHARED  | VM_MAYSHARE   |
-				 VM_PFNMAP    | VM_IO      | VM_DONTEXPAND |
-				 VM_HUGETLB | VM_MIXEDMAP))
-			return 0;		/* just ignore the advice */
-
-		if (vma_is_dax(vma))
+		if (vma->vm_flags & VM_MERGEABLE)
  			return 0;
-
-#ifdef VM_SAO
-		if (*vm_flags & VM_SAO)
+		if (!vma_ksm_compatible(vma))
  			return 0;
-#endif
-#ifdef VM_SPARC_ADI
-		if (*vm_flags & VM_SPARC_ADI)
-			return 0;
-#endif
    		if (!test_bit(MMF_VM_MERGEABLE, &mm->flags)) {
-			err = __ksm_enter(mm);
+			err = __ksm_enter(mm, MMF_VM_MERGEABLE);
  			if (err)
  				return err;
  		}

@@ -2626,7 +2693,7 @@ int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
  }
  EXPORT_SYMBOL_GPL(ksm_madvise);
  -int __ksm_enter(struct mm_struct *mm)
+int __ksm_enter(struct mm_struct *mm, int flag)
  {
  	struct ksm_mm_slot *mm_slot;
  	struct mm_slot *slot;

@@ -2659,7 +2726,7 @@ int __ksm_enter(struct mm_struct *mm)
  		list_add_tail(&slot->mm_node, &ksm_scan.mm_slot->slot.mm_node);
  	spin_unlock(&ksm_mmlist_lock);
  -	set_bit(MMF_VM_MERGEABLE, &mm->flags);
+	set_bit(flag, &mm->flags);
  	mmgrab(mm);
    	if (needs_wakeup)

@@ -2668,7 +2735,7 @@ int __ksm_enter(struct mm_struct *mm)
  	return 0;
  }
  -void __ksm_exit(struct mm_struct *mm)
+void __ksm_exit(struct mm_struct *mm, int flag)
  {
  	struct ksm_mm_slot *mm_slot;
  	struct mm_slot *slot;

@@ -2700,7 +2767,7 @@ void __ksm_exit(struct mm_struct *mm)
    	if (easy_to_free) {
  		mm_slot_free(mm_slot_cache, mm_slot);
-		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+		clear_bit(flag, &mm->flags);
  		mmdrop(mm);
  	} else if (mm_slot) {
  		mmap_write_lock(mm);
mm/mmap.c +1 -0
index 740b54be3ed4..483e182e0b9d 100644
@@ -46,6 +46,7 @@
  #include <linux/pkeys.h>
  #include <linux/oom.h>
  #include <linux/sched/mm.h>
+#include <linux/ksm.h>
    #include <linux/uaccess.h>
Quoted 9 lines

mm/mmap.c int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma, · line 2217

Quoted 6 lines
mm/mmap.c +1 -0
  #include <asm/cacheflush.h>

@@ -2213,6 +2214,8 @@ int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma,
  	/* vma_complete stores the new vma */
  	vma_complete(&vp, vmi, vma->vm_mm);
  +	ksm_add_vma(new);
+

Splitting a VMA shouldn't modify VM_MERGEABLE, so I assume this is not required?

I'll fix it.

Quoted 17 lines
Quoted 17 lines
mm/mmap.c +1 -0
  	/* Success. */
  	if (new_below)
  		vma_next(vmi);

@@ -2664,6 +2667,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
  	if (file && vm_flags & VM_SHARED)
  		mapping_unmap_writable(file->f_mapping);
  	file = vma->vm_file;
+	ksm_add_vma(vma);
  expanded:
  	perf_event_mmap(vma);
  @@ -2936,6 +2940,7 @@ static int do_brk_flags(struct vma_iterator *vmi,

struct vm_area_struct *vma,

  		goto mas_store_fail;
    	mm->map_count++;
mm/mmap.c +1 -0
+	ksm_add_vma(vma);
  out:
  	perf_event_mmap(vma);
Quoted 9 lines

mm/mmap.c struct vm_area_struct *copy_vma(struct vm_area_struct **vmap, · line 3188

Quoted 6 lines
  	mm->total_vm += len >> PAGE_SHIFT;
mm/mmap.c +1 -0
@@ -3180,6 +3185,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
  		if (vma_link(mm, new_vma))
  			goto out_vma_link;
  		*need_rmap_locks = false;
+		ksm_add_vma(new_vma);

Copying shouldn't modify VM_MERGEABLE, so I think this is not required?

I'll fix it.

Quoted 11 lines

mm/mmap.c static struct vm_area_struct *__install_special_mapping( · line 3364

Quoted 7 lines
mm/mmap.c +1 -0
  	}
  	validate_mm_mt(mm);
  	return new_vma;

@@ -3356,6 +3362,7 @@ static struct vm_area_struct *__install_special_mapping(
  	vm_stat_account(mm, vma->vm_flags, len >> PAGE_SHIFT);
    	perf_event_mmap(vma);
+	ksm_add_vma(vma);

IIUC, special mappings will never be considered a reasonable target for KSM (especially, because at least VM_DONTEXPAND is always set).

Quoted 1 line

I think you can just drop this call.

I dropped it.

#7 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

[...]

Quoted 7 lines

That will work.

Quoted 3 lines

work? IOW, not exporting ksm_add_mm() and not passing a flag to __ksm_enter() -- it would simply set MMF_VM_MERGEABLE ?

ksm_add_mm() is also used in prctl (kernel/sys.c). Do you want to make a similar change there?

Yes.

Quoted 20 lines
Quoted 11 lines
Quoted 11 lines

+ *
+ * @vma: Pointer to vma
+ */

mm/ksm.c +8 -0
+void ksm_add_vma(struct vm_area_struct *vma)
+{
+	struct mm_struct *mm = vma->vm_mm;
+
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_add_vma(vma);
+}
+
Quoted 9 lines

mm/ksm.c static int ksm_scan_thread(void *nothing) · line 2630

Quoted 6 lines
mm/ksm.c +6 -0
+/**
+ * ksm_add_vmas - Mark all vma's of a process as mergeable
+ *
+ * @mm:  Pointer to mm
+ */
+void ksm_add_vmas(struct mm_struct *mm)

I'd suggest calling this

Quoted 1 line

I guess you forgot your name suggestion?

Yeah, I reconsidered because the first idea I had was not particularly good. Maybe

ksm_enable_for_all_vmas()

But not so sure. If you think the "add" terminology is a good fit, keep it like that.

Thanks for bearing with me :)

Signature · 3 lines
Thanks,

David / dhildenb
#8 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

Quoted 36 lines · David Hildenbrand

[...]

Quoted 7 lines

That will work.

Quoted 3 lines

work? IOW, not exporting ksm_add_mm() and not passing a flag to __ksm_enter() -- it would simply set MMF_VM_MERGEABLE ?

ksm_add_mm() is also used in prctl (kernel/sys.c). Do you want to make a similar change there?

Yes.

Quoted 20 lines
Quoted 11 lines
Quoted 11 lines

+ *
+ * @vma: Pointer to vma
+ */

(quoted diff) +8 -0
+void ksm_add_vma(struct vm_area_struct *vma)
+{
+	struct mm_struct *mm = vma->vm_mm;
+
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_add_vma(vma);
+}
+
Quoted 9 lines

Review

Quoted 6 lines
(quoted diff) +6 -0
+/**
+ * ksm_add_vmas - Mark all vma's of a process as mergeable
+ *
+ * @mm:  Pointer to mm
+ */
+void ksm_add_vmas(struct mm_struct *mm)

I'd suggest calling this

Quoted 1 line

I guess you forgot your name suggestion?

Yeah, I reconsidered because the first idea I had was not particularly good. Maybe

Quoted 8 lines

ksm_enable_for_all_vmas()

But not so sure. If you think the "add" terminology is a good fit, keep it like that.

Thanks for bearing with me :)

I briefly played with your patch to see how much it can be simplified. Always enabling ksm (setting MMF_VM_MERGEABLE) before setting MMF_VM_MERGE_ANY might simplify things. ksm_enable_merge_any() [or however it should be called] and ksm_fork() contain the interesting bits.

Feel free to incorporate what you consider valuable (uncompiled, untested).

arch/s390/mm/gmap.c +6 -0
index 5a716bdcba05..5b2eef31398e 100644
@@ -2591,6 +2591,12 @@ int gmap_mark_unmergeable(void)
 	int ret;
 	VMA_ITERATOR(vmi, mm, 0);
 
+	/*
+	 * Make sure to disable KSM (if enabled for the whole process or
+	 * individual VMAs). Note that nothing currently hinders user space
+	 * from re-enabling it.
+	 */
+	clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
 	for_each_vma(vmi, vma) {
 		/* Copy vm_flags to avoid partial modifications in ksm_madvise */
 		vm_flags = vma->vm_flags;
include/linux/ksm.h +17 -2
index 7e232ba59b86..c638b034d586 100644
@@ -18,13 +18,24 @@
 #ifdef CONFIG_KSM
 int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
 		unsigned long end, int advice, unsigned long *vm_flags);
+
+void ksm_add_vma(struct vm_area_struct *vma);
+int ksm_enable_merge_any(struct mm_struct *mm);
+
 int __ksm_enter(struct mm_struct *mm);
 void __ksm_exit(struct mm_struct *mm);
 
 static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
 {
-	if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags))
-		return __ksm_enter(mm);
+	int ret;
+
+	if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags)) {
+		ret = __ksm_enter(mm);
+		if (ret)
+			return ret;
+	}
+	if (test_bit(MMF_VM_MERGE_ANY, &oldmm->flags))
+		set_bit(MMF_VM_MERGE_ANY, &mm->flags);
 	return 0;
 }
 

@@ -53,6 +64,10 @@ void folio_migrate_ksm(struct folio *newfolio, struct folio *folio);
 
 #else  /* !CONFIG_KSM */
 
+static inline void ksm_add_vma(struct vm_area_struct *vma)
+{
+}
+
 static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
 {
 	return 0;
include/linux/sched/coredump.h +1 -0
index 0e17ae7fbfd3..0ee96ea7a0e9 100644
@@ -90,4 +90,5 @@ static inline int get_dumpable(struct mm_struct *mm)
 #define MMF_INIT_MASK		(MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK |\
 				 MMF_DISABLE_THP_MASK | MMF_HAS_MDWE_MASK)
 
+#define MMF_VM_MERGE_ANY	29
 #endif /* _LINUX_SCHED_COREDUMP_H */
include/uapi/linux/prctl.h +2 -0
index 1312a137f7fb..759b3f53e53f 100644
@@ -290,4 +290,6 @@ struct prctl_mm_map {
 #define PR_SET_VMA		0x53564d41
 # define PR_SET_VMA_ANON_NAME		0
 
+#define PR_SET_MEMORY_MERGE		67
+#define PR_GET_MEMORY_MERGE		68
 #endif /* _LINUX_PRCTL_H */
kernel/sys.c +25 -0
index 495cd87d9bf4..8c2e50edeb18 100644
@@ -15,6 +15,7 @@
 #include <linux/highuid.h>
 #include <linux/fs.h>
 #include <linux/kmod.h>
+#include <linux/ksm.h>
 #include <linux/perf_event.h>
 #include <linux/resource.h>
 #include <linux/kernel.h>

@@ -2661,6 +2662,30 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
 	case PR_SET_VMA:
 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
 		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
+			error = ksm_enable_merge_any(me->mm);
+		} else {
+			/*
+			 * TODO: we might want disable KSM on all VMAs and
+			 * trigger unsharing to completely disable KSM.
+			 */
+			clear_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+			error = 0;
+		}
+		mmap_write_unlock(me->mm);
+		break;
+	case PR_GET_MEMORY_MERGE:
+		if (arg2 || arg3 || arg4 || arg5)
+			return -EINVAL;
+
+		error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		break;
+#endif
 	default:
 		error = -EINVAL;
 		break;
mm/ksm.c +73 -17

128 lines not on this page · raw message

mm/mmap.c +3 -0
index ff68a67a2a7c..1f8619ff58ca 100644
@@ -46,6 +46,7 @@
 #include <linux/pkeys.h>
 #include <linux/oom.h>
 #include <linux/sched/mm.h>
+#include <linux/ksm.h>
 
 #include <linux/uaccess.h>
 #include <asm/cacheflush.h>

@@ -2659,6 +2660,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
 	if (file && vm_flags & VM_SHARED)
 		mapping_unmap_writable(file->f_mapping);
 	file = vma->vm_file;
+	ksm_add_vma(vma);
 expanded:
 	perf_event_mmap(vma);
 

@@ -2931,6 +2933,7 @@ static int do_brk_flags(struct vma_iterator *vmi, struct vm_area_struct *vma,
 		goto mas_store_fail;
 
 	mm->map_count++;
+	ksm_add_vma(vma);
 out:
 	perf_event_mmap(vma);
 	mm->total_vm += len >> PAGE_SHIFT;
Signature · 7 lines
2.39.2


-- 
Thanks,

David / dhildenb
author
#9 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

Quoted 40 lines · David Hildenbrand
Quoted 30 lines · David Hildenbrand

[...]

Quoted 7 lines

That will work.

Quoted 3 lines

work? IOW, not exporting ksm_add_mm() and not passing a flag to __ksm_enter() -- it would simply set MMF_VM_MERGEABLE ?

ksm_add_mm() is also used in prctl (kernel/sys.c). Do you want to make a similar change there?

Yes.

Quoted 20 lines
Quoted 20 lines
Quoted 7 lines

+ *
+ * @vma: Pointer to vma
+ */

mm/ksm.c +4 -0
+void ksm_add_vma(struct vm_area_struct *vma)
+{
+	struct mm_struct *mm = vma->vm_mm;
+

mm/ksm.c static int ksm_scan_thread(void *nothing)

Quoted 10 lines
mm/ksm.c +10 -0
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_add_vma(vma);
+}
+
+/**
+ * ksm_add_vmas - Mark all vma's of a process as mergeable
+ *
+ * @mm:  Pointer to mm
+ */
+void ksm_add_vmas(struct mm_struct *mm)

I'd suggest calling this

Quoted 8 lines
Quoted 1 line

I guess you forgot your name suggestion?

Yeah, I reconsidered because the first idea I had was not particularly good. Maybe
ksm_enable_for_all_vmas()
But not so sure. If you think the "add" terminology is a good fit, keep it like that.
Thanks for bearing with me :)

Quoted 9 lines

I briefly played with your patch to see how much it can be simplified. Always enabling ksm (setting MMF_VM_MERGEABLE) before setting MMF_VM_MERGE_ANY might simplify things. ksm_enable_merge_any() [or however it should be called] and ksm_fork() contain the interesting bits.

Feel free to incorporate what you consider valuable (uncompiled, untested).

I added most of it. The only change is that I kept ksm_add_vmas as a static function, otherwise I need to define the VMA_ITERATOR at the top of the function.

Quoted 278 lines
arch/s390/mm/gmap.c +5 -0
index 5a716bdcba05..5b2eef31398e 100644
@@ -2591,6 +2591,12 @@ int gmap_mark_unmergeable(void)
 	int ret;
 	VMA_ITERATOR(vmi, mm, 0);
 +	/*
+	 * Make sure to disable KSM (if enabled for the whole process or
+	 * individual VMAs). Note that nothing currently hinders user space
+	 * from re-enabling it.
+	 */
+	clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
 	for_each_vma(vmi, vma) {
 		/* Copy vm_flags to avoid partial modifications in ksm_madvise */
 		vm_flags = vma->vm_flags;
include/linux/ksm.h +13 -2
index 7e232ba59b86..c638b034d586 100644
@@ -18,13 +18,24 @@
 #ifdef CONFIG_KSM
 int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
 		unsigned long end, int advice, unsigned long *vm_flags);
+
+void ksm_add_vma(struct vm_area_struct *vma);
+int ksm_enable_merge_any(struct mm_struct *mm);
+
 int __ksm_enter(struct mm_struct *mm);
 void __ksm_exit(struct mm_struct *mm);
   static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
 {
-	if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags))
-		return __ksm_enter(mm);
+	int ret;
+
+	if (test_bit(MMF_VM_MERGEABLE, &oldmm->flags)) {
+		ret = __ksm_enter(mm);
+		if (ret)
+			return ret;
+	}
+	if (test_bit(MMF_VM_MERGE_ANY, &oldmm->flags))
+		set_bit(MMF_VM_MERGE_ANY, &mm->flags);
 	return 0;
 }
 @@ -53,6 +64,10 @@ void folio_migrate_ksm(struct folio *newfolio, struct folio
 *folio);

   #else  /* !CONFIG_KSM */

+static inline void ksm_add_vma(struct vm_area_struct *vma)

include/linux/ksm.h +3 -0
+{
+}
+
 static inline int ksm_fork(struct mm_struct *mm, struct mm_struct *oldmm)
 {
 	return 0;
include/linux/sched/coredump.h +0 -0
index 0e17ae7fbfd3..0ee96ea7a0e9 100644
@@ -90,4 +90,5 @@ static inline int get_dumpable(struct mm_struct *mm)
 #define MMF_INIT_MASK		(MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK |\
 				 MMF_DISABLE_THP_MASK | MMF_HAS_MDWE_MASK)
 +#define MMF_VM_MERGE_ANY	29
 #endif /* _LINUX_SCHED_COREDUMP_H */
include/uapi/linux/prctl.h +1 -0
index 1312a137f7fb..759b3f53e53f 100644
@@ -290,4 +290,6 @@ struct prctl_mm_map {
 #define PR_SET_VMA		0x53564d41
 # define PR_SET_VMA_ANON_NAME		0
 +#define PR_SET_MEMORY_MERGE		67
+#define PR_GET_MEMORY_MERGE		68
 #endif /* _LINUX_PRCTL_H */
kernel/sys.c +25 -0
index 495cd87d9bf4..8c2e50edeb18 100644
@@ -15,6 +15,7 @@
 #include <linux/highuid.h>
 #include <linux/fs.h>
 #include <linux/kmod.h>
+#include <linux/ksm.h>
 #include <linux/perf_event.h>
 #include <linux/resource.h>
 #include <linux/kernel.h>

@@ -2661,6 +2662,30 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
 	case PR_SET_VMA:
 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
 		break;
+#ifdef CONFIG_KSM
+	case PR_SET_MEMORY_MERGE:
+		if (mmap_write_lock_killable(me->mm))
+			return -EINTR;
+
+		if (arg2) {
+			error = ksm_enable_merge_any(me->mm);
+		} else {
+			/*
+			 * TODO: we might want disable KSM on all VMAs and
+			 * trigger unsharing to completely disable KSM.
+			 */
+			clear_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+			error = 0;
+		}
+		mmap_write_unlock(me->mm);
+		break;
+	case PR_GET_MEMORY_MERGE:
+		if (arg2 || arg3 || arg4 || arg5)
+			return -EINVAL;
+
+		error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
+		break;
+#endif
 	default:
 		error = -EINVAL;
 		break;
mm/ksm.c +71 -17
index 2b8d30068cbb..76ceec35395c 100644
@@ -512,6 +512,28 @@ static int break_ksm(struct vm_area_struct *vma, unsigned long addr)
 	return (ret & VM_FAULT_OOM) ? -ENOMEM : 0;
 }
 +static bool vma_ksm_compatible(struct vm_area_struct *vma)
+{
+	if (vma->vm_flags & (VM_SHARED  | VM_MAYSHARE   | VM_PFNMAP  |
+			     VM_IO      | VM_DONTEXPAND | VM_HUGETLB |
+			     VM_MIXEDMAP))
+		return false;		/* just ignore the advice */
+
+	if (vma_is_dax(vma))
+		return false;
+
+#ifdef VM_SAO
+	if (vma->vm_flags & VM_SAO)
+		return false;
+#endif
+#ifdef VM_SPARC_ADI
+	if (vma->vm_flags & VM_SPARC_ADI)
+		return false;
+#endif
+
+	return true;
+}
+
 static struct vm_area_struct *find_mergeable_vma(struct mm_struct *mm,
 		unsigned long addr)
 {

@@ -1020,6 +1042,7 @@ static int unmerge_and_remove_all_rmap_items(void)
   			mm_slot_free(mm_slot_cache, mm_slot);
 			clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+			clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
 			mmdrop(mm);
 		} else
 			spin_unlock(&ksm_mmlist_lock);

@@ -2395,6 +2418,7 @@ static struct ksm_rmap_item *scan_get_next_rmap_item(struct page **page)
   		mm_slot_free(mm_slot_cache, mm_slot);
 		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+		clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
 		mmap_read_unlock(mm);
 		mmdrop(mm);
 	} else {

@@ -2471,6 +2495,52 @@ static int ksm_scan_thread(void *nothing)
 	return 0;
 }
 +static void __ksm_add_vma(struct vm_area_struct *vma)
+{
+	unsigned long vm_flags = vma->vm_flags;
+
+	if (vm_flags & VM_MERGEABLE)
+		return;
+
+	if (vma_ksm_compatible(vma)) {
+		vm_flags |= VM_MERGEABLE;
+		vm_flags_reset(vma, vm_flags);
+	}
+}
+
+/**
+ * ksm_add_vma - Mark vma as mergeable
+ *
+ * @vma:  Pointer to vma
+ */
+void ksm_add_vma(struct vm_area_struct *vma)
+{
+	struct mm_struct *mm = vma->vm_mm;
+
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_add_vma(vma);
+}
+
+int ksm_enable_merge_any(struct mm_struct *mm)
+{
+	struct vm_area_struct *vma;
+	int ret;
+
+	if (test_bit(MMF_VM_MERGE_ANY, mm->flags))
+		return 0;
+
+	if (!test_bit(MMF_VM_MERGEABLE, mm->flags)) {
+		ret = __ksm_enter(mm);
+		if (ret)
+			return ret;
+	}
+	set_bit(MMF_VM_MERGE_ANY, &mm->flags);
+
+	VMA_ITERATOR(vmi, mm, 0);
+	for_each_vma(vmi, vma)
+		__ksm_add_vma(vma);
+}
+
 int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
 		unsigned long end, int advice, unsigned long *vm_flags)
 {

@@ -2479,25 +2549,10 @@ int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
   	switch (advice) {
 	case MADV_MERGEABLE:
-		/*
-		 * Be somewhat over-protective for now!
-		 */
-		if (*vm_flags & (VM_MERGEABLE | VM_SHARED  | VM_MAYSHARE   |
-				 VM_PFNMAP    | VM_IO      | VM_DONTEXPAND |
-				 VM_HUGETLB | VM_MIXEDMAP))
-			return 0;		/* just ignore the advice */
-
-		if (vma_is_dax(vma))
+		if (vma->vm_flags & VM_MERGEABLE)
 			return 0;
-
-#ifdef VM_SAO
-		if (*vm_flags & VM_SAO)
-			return 0;
-#endif
-#ifdef VM_SPARC_ADI
-		if (*vm_flags & VM_SPARC_ADI)
+		if (!vma_ksm_compatible(vma))
 			return 0;
-#endif
   		if (!test_bit(MMF_VM_MERGEABLE, &mm->flags)) {
 			err = __ksm_enter(mm);

@@ -2601,6 +2656,7 @@ void __ksm_exit(struct mm_struct *mm)
 	if (easy_to_free) {
 		mm_slot_free(mm_slot_cache, mm_slot);
 		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+		clear_bit(MMF_VM_MERGE_ANY, &mm->flags);
 		mmdrop(mm);
 	} else if (mm_slot) {
 		mmap_write_lock(mm);
mm/mmap.c +2 -0
index ff68a67a2a7c..1f8619ff58ca 100644
@@ -46,6 +46,7 @@
 #include <linux/pkeys.h>
 #include <linux/oom.h>
 #include <linux/sched/mm.h>
+#include <linux/ksm.h>
   #include <linux/uaccess.h>
 #include <asm/cacheflush.h>

@@ -2659,6 +2660,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
 	if (file && vm_flags & VM_SHARED)
 		mapping_unmap_writable(file->f_mapping);
 	file = vma->vm_file;
+	ksm_add_vma(vma);
 expanded:
 	perf_event_mmap(vma);
 @@ -2931,6 +2933,7 @@ static int do_brk_flags(struct vma_iterator *vmi, struct

vm_area_struct *vma,

mm/mmap.c +1 -1
 		goto mas_store_fail;
   	mm->map_count++;
+	ksm_add_vma(vma);
 out:
 	perf_event_mmap(vma);
 	mm->total_vm += len >> PAGE_SHIFT;
--

2.39.2

#10 lore raw

Re: [PATCH v6 1/3] mm: add new api to enable ksm per process

Quoted 49 lines · Stefan Roesch
Quoted 46 lines · David Hildenbrand
Quoted 30 lines · David Hildenbrand

[...]

Quoted 7 lines

That will work.

Quoted 3 lines

work? IOW, not exporting ksm_add_mm() and not passing a flag to __ksm_enter() -- it would simply set MMF_VM_MERGEABLE ?

ksm_add_mm() is also used in prctl (kernel/sys.c). Do you want to make a similar change there?

Yes.

Quoted 20 lines
Quoted 20 lines
Quoted 7 lines

+ *
+ * @vma: Pointer to vma
+ */

mm/ksm.c +4 -0
+void ksm_add_vma(struct vm_area_struct *vma)
+{
+	struct mm_struct *mm = vma->vm_mm;
+

mm/ksm.c static int ksm_scan_thread(void *nothing)

Quoted 10 lines
mm/ksm.c +10 -0
+	if (test_bit(MMF_VM_MERGE_ANY, &mm->flags))
+		__ksm_add_vma(vma);
+}
+
+/**
+ * ksm_add_vmas - Mark all vma's of a process as mergeable
+ *
+ * @mm:  Pointer to mm
+ */
+void ksm_add_vmas(struct mm_struct *mm)

I'd suggest calling this

Quoted 8 lines
Quoted 1 line

I guess you forgot your name suggestion?

Yeah, I reconsidered because the first idea I had was not particularly good. Maybe
ksm_enable_for_all_vmas()
But not so sure. If you think the "add" terminology is a good fit, keep it like that.
Thanks for bearing with me :)

I briefly played with your patch to see how much it can be simplified. Always enabling ksm (setting MMF_VM_MERGEABLE) before setting MMF_VM_MERGE_ANY might simplify things. ksm_enable_merge_any() [or however it should be called] and ksm_fork() contain the interesting bits.

Quoted 6 lines
Quoted 3 lines

Feel free to incorporate what you consider valuable (uncompiled, untested).

I added most of it. The only change is that I kept ksm_add_vmas as a static function, otherwise I need to define the VMA_ITERATOR at the top of the function.

Makes sense. I'll review patch #3 later, so we can hopefully get this into the 6.4 merge window after letting it rest at least some days in -next.

Signature · 3 lines
Thanks,

David / dhildenb
author
#11 lore raw

PATCH v6 2/3 mm: add new KSM process and sysfs knobs

In reply to Stefan Roesch

This adds the general_profit KSM sysfs knob and the process profit metric knobs to ksm_stat.

1) expose general_profit metric

The documentation mentions a general profit metric, however this metric is not calculated. In addition the formula depends on the size of internal structures, which makes it more difficult for an administrator to make the calculation. Adding the metric for a better user experience.

2) document general_profit sysfs knob

3) calculate ksm process profit metric

The ksm documentation mentions the process profit metric and how to calculate it. This adds the calculation of the metric.

4) mm: expose ksm process profit metric in ksm_stat

This exposes the ksm process profit metric in /proc/<pid>/ksm_stat. The documentation mentions the formula for the ksm process profit metric, however it does not calculate it. In addition the formula depends on the size of internal structures. So it makes sense to expose it.

5) document new procfs ksm knobs

Signed-off-by Stefan Roesch <[email protected]>
Reviewed-by Bagas Sanjaya <[email protected]>
Acked-by David Hildenbrand <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Johannes Weiner <[email protected]>
Cc Michal Hocko <[email protected]>
Cc Rik van Riel <[email protected]>
Signed-off-by Andrew Morton <[email protected]>
---
 Documentation/ABI/testing/sysfs-kernel-mm-ksm |  8 +++++++
 Documentation/admin-guide/mm/ksm.rst          |  5 ++++-
 fs/proc/base.c                                |  3 +++
 include/linux/ksm.h                           |  4 ++++
 mm/ksm.c                                      | 21 +++++++++++++++++++
 5 files changed, 40 insertions(+), 1 deletion(-)
Documentation/ABI/testing/sysfs-kernel-mm-ksm +8 -0
index d244674a9480..6041a025b65a 100644
@@ -51,3 +51,11 @@ Description:	Control merging pages across different NUMA nodes.
 
 		When it is set to 0 only pages from the same node are merged,
 		otherwise pages from all nodes can be merged together (default).
+
+What:		/sys/kernel/mm/ksm/general_profit
+Date:		April 2023
+KernelVersion:  6.4
+Contact:	Linux memory management mailing list <[email protected]>
+Description:	Measure how effective KSM is.
+		general_profit: how effective is KSM. The formula for the
+		calculation is in Documentation/admin-guide/mm/ksm.rst.
Documentation/admin-guide/mm/ksm.rst +4 -1
index 270560fef3b2..bc1dd830dd49 100644
@@ -157,6 +157,8 @@ stable_node_chains_prune_millisecs
 
 The effectiveness of KSM and MADV_MERGEABLE is shown in ``/sys/kernel/mm/ksm/``:
 
+general_profit
+        how effective is KSM. The calculation is explained below.
 pages_shared
         how many shared pages are being used
 pages_sharing

@@ -214,7 +216,8 @@ several times, which are unprofitable memory consumed.
 			  ksm_rmap_items * sizeof(rmap_item).
 
    where ksm_merging_pages is shown under the directory ``/proc/<pid>/``,
-   and ksm_rmap_items is shown in ``/proc/<pid>/ksm_stat``.
+   and ksm_rmap_items is shown in ``/proc/<pid>/ksm_stat``. The process profit
+   is also shown in ``/proc/<pid>/ksm_stat`` as ksm_process_profit.
 
 From the perspective of application, a high ratio of ``ksm_rmap_items`` to
 ``ksm_merging_pages`` means a bad madvise-applied policy, so developers or
fs/proc/base.c +3 -0
index 07463ad4a70a..cb42bb021995 100644
@@ -96,6 +96,7 @@
 #include <linux/time_namespace.h>
 #include <linux/resctrl.h>
 #include <linux/cn_proc.h>
+#include <linux/ksm.h>
 #include <trace/events/oom.h>
 #include "internal.h"
 #include "fd.h"

@@ -3208,6 +3209,8 @@ static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
 	if (mm) {
 		seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items);
 		seq_printf(m, "zero_pages_sharing %lu\n", mm->ksm_zero_pages_sharing);
+		seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages);
+		seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm));
 		mmput(mm);
 	}
include/linux/ksm.h +4 -0
index f24f9faf1561..63cb491b1740 100644
@@ -60,6 +60,10 @@ struct page *ksm_might_need_to_copy(struct page *page,
 void rmap_walk_ksm(struct folio *folio, struct rmap_walk_control *rwc);
 void folio_migrate_ksm(struct folio *newfolio, struct folio *folio);
 
+#ifdef CONFIG_PROC_FS
+long ksm_process_profit(struct mm_struct *);
+#endif /* CONFIG_PROC_FS */
+
 #else  /* !CONFIG_KSM */
 
 static inline void ksm_add_vma(struct vm_area_struct *vma)
mm/ksm.c +21 -0
index ab95ae0f9def..7982bac15d8c 100644
@@ -3042,6 +3042,14 @@ static void wait_while_offlining(void)
 }
 #endif /* CONFIG_MEMORY_HOTREMOVE */
 
+#ifdef CONFIG_PROC_FS
+long ksm_process_profit(struct mm_struct *mm)
+{
+	return mm->ksm_merging_pages * PAGE_SIZE -
+		mm->ksm_rmap_items * sizeof(struct ksm_rmap_item);
+}
+#endif /* CONFIG_PROC_FS */
+
 #ifdef CONFIG_SYSFS
 /*
  * This all compiles without CONFIG_SYSFS, but is a waste of space.

@@ -3313,6 +3321,18 @@ static ssize_t zero_pages_sharing_show(struct kobject *kobj,
 }
 KSM_ATTR_RO(zero_pages_sharing);
 
+static ssize_t general_profit_show(struct kobject *kobj,
+				   struct kobj_attribute *attr, char *buf)
+{
+	long general_profit;
+
+	general_profit = ksm_pages_sharing * PAGE_SIZE -
+				ksm_rmap_items * sizeof(struct ksm_rmap_item);
+
+	return sysfs_emit(buf, "%ld\n", general_profit);
+}
+KSM_ATTR_RO(general_profit);
+
 static ssize_t stable_node_dups_show(struct kobject *kobj,
 				     struct kobj_attribute *attr, char *buf)
 {

@@ -3378,6 +3398,7 @@ static struct attribute *ksm_attrs[] = {
 	&stable_node_dups_attr.attr,
 	&stable_node_chains_prune_millisecs_attr.attr,
 	&use_zero_pages_attr.attr,
+	&general_profit_attr.attr,
 	NULL,
 };
Signature · 1 line
2.31.1
author
#12 lore raw

PATCH v6 3/3 selftests/mm: add new selftests for KSM

In reply to Stefan Roesch

This adds three new tests to the selftests for KSM. These tests use the new prctl API's to enable and disable KSM.

1) add new prctl flags to prctl header file in tools dir

This adds the new prctl flags to the include file prct.h in the tools directory. This makes sure they are available for testing.

2) add KSM prctl merge test

This adds the -t option to the ksm_tests program. The -t flag allows to specify if it should use madvise or prctl ksm merging.

3) add KSM get merge type test

This adds the -G flag to the ksm_tests program to query the KSM status with prctl after KSM has been enabled with prctl.

4) add KSM fork test

Add fork test to verify that the MMF_VM_MERGE_ANY flag is inherited by the child process.

5) add two functions for debugging merge outcome

This adds two functions to report the metrics in /proc/self/ksm_stat and /sys/kernel/debug/mm/ksm.

The debugging can be enabled with the following command line:
make -C tools/testing/selftests TARGETS="mm" --keep-going \
EXTRA_CFLAGS=-DDEBUG=1

[[email protected]: fix Makefile]
Link https://lkml.kernel.org/r/[email protected] open here
Signed-off-by Stefan Roesch <[email protected]>
Cc Bagas Sanjaya <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Johannes Weiner <[email protected]>
Cc Michal Hocko <[email protected]>
Cc Rik van Riel <[email protected]>
Signed-off-by Andrew Morton <[email protected]>
---
 tools/include/uapi/linux/prctl.h       |   2 +
 tools/testing/selftests/mm/Makefile    |   2 +-
 tools/testing/selftests/mm/ksm_tests.c | 254 +++++++++++++++++++++----
 3 files changed, 218 insertions(+), 40 deletions(-)
tools/include/uapi/linux/prctl.h +2 -0
index a5e06dcbba13..e4c629c1f1b0 100644
@@ -284,4 +284,6 @@ struct prctl_mm_map {
 #define PR_SET_VMA		0x53564d41
 # define PR_SET_VMA_ANON_NAME		0
 
+#define PR_SET_MEMORY_MERGE		67
+#define PR_GET_MEMORY_MERGE		68
 #endif /* _LINUX_PRCTL_H */
tools/testing/selftests/mm/Makefile +1 -1
index c31d952cff68..fbf5646b1072 100644
@@ -29,7 +29,7 @@ MACHINE ?= $(shell echo $(uname_M) | sed -e 's/aarch64.*/arm64/' -e 's/ppc64.*/p
 # LDLIBS.
 MAKEFLAGS += --no-builtin-rules
 
-CFLAGS = -Wall -I $(top_srcdir) $(EXTRA_CFLAGS) $(KHDR_INCLUDES)
+CFLAGS = -Wall -I $(top_srcdir) -I $(top_srcdir)/tools/include/uapi $(EXTRA_CFLAGS) $(KHDR_INCLUDES)
 LDLIBS = -lrt -lpthread
 TEST_GEN_FILES = cow
 TEST_GEN_FILES += compaction_test
tools/testing/selftests/mm/ksm_tests.c +215 -39

466 lines not on this page · raw message

Signature · 1 line
2.31.1
#13 lore raw

Re: [PATCH v6 3/3] selftests/mm: add new selftests for KSM

Quoted 23 lines · Stefan Roesch

This adds three new tests to the selftests for KSM. These tests use the new prctl API's to enable and disable KSM.

1) add new prctl flags to prctl header file in tools dir

This adds the new prctl flags to the include file prct.h in the tools directory. This makes sure they are available for testing.

2) add KSM prctl merge test

This adds the -t option to the ksm_tests program. The -t flag allows to specify if it should use madvise or prctl ksm merging.

3) add KSM get merge type test

This adds the -G flag to the ksm_tests program to query the KSM status with prctl after KSM has been enabled with prctl.

4) add KSM fork test

Add fork test to verify that the MMF_VM_MERGE_ANY flag is inherited by the child process.

Quoted 8 lines

5) add two functions for debugging merge outcome

This adds two functions to report the metrics in /proc/self/ksm_stat and /sys/kernel/debug/mm/ksm.

The debugging can be enabled with the following command line:
make -C tools/testing/selftests TARGETS="mm" --keep-going \
EXTRA_CFLAGS=-DDEBUG=1

Would it make sense to instead have a "-D" (if still unused) runtime options to print this data? Dead code that's not compiled is a bit unfortunate as it can easily bit-rot.

This patch essentially does two things

1) Add the option to run all tests/benchmarks with the PRCTL instead of MADVISE

2) Add some functional KSM tests for the new PRCTL (fork, enabling works, disabling works).

The latter should rather go into ksm_functional_tests().

[...]

Quoted 57 lines
tools/testing/selftests/mm/ksm_tests.c +55 -1
-static int check_ksm_unmerge(int mapping, int prot, int timeout, size_t page_size)
+/* Verify that prctl ksm flag is inherited. */
+static int check_ksm_fork(void)
+{
+	int rc = KSFT_FAIL;
+	pid_t child_pid;
+
+	if (prctl(PR_SET_MEMORY_MERGE, 1)) {
+		perror("prctl");
+		return KSFT_FAIL;
+	}
+
+	child_pid = fork();
+	if (child_pid == 0) {
+		int is_on = prctl(PR_GET_MEMORY_MERGE, 0);
+
+		if (!is_on)
+			exit(KSFT_FAIL);
+
+		exit(KSFT_PASS);
+	}
+
+	if (child_pid < 0)
+		goto out;
+
+	if (waitpid(child_pid, &rc, 0) < 0)
+		rc = KSFT_FAIL;
+
+	if (prctl(PR_SET_MEMORY_MERGE, 0)) {
+		perror("prctl");
+		rc = KSFT_FAIL;
+	}
+
+out:
+	if (rc == KSFT_PASS)
+		printf("OK\n");
+	else
+		printf("Not OK\n");
+
+	return rc;
+}
+
+static int check_ksm_get_merge_type(void)
+{
+	if (prctl(PR_SET_MEMORY_MERGE, 1)) {
+		perror("prctl set");
+		return 1;
+	}
+
+	int is_on = prctl(PR_GET_MEMORY_MERGE, 0);
+
+	if (prctl(PR_SET_MEMORY_MERGE, 0)) {
+		perror("prctl set");
+		return 1;
+	}
+

tools/testing/selftests/mm/ksm_tests.c static int check_ksm_merge(int mapping, int prot, long page_count, int timeout, · line 448

Quoted 10 lines
tools/testing/selftests/mm/ksm_tests.c +10 -0
+	int is_off = prctl(PR_GET_MEMORY_MERGE, 0);
+
+	if (is_on && is_off) {
+		printf("OK\n");
+		return KSFT_PASS;
+	}
+
+	printf("Not OK\n");
+	return KSFT_FAIL;
+}

Yes, these two are better located in ksm_functional_tests() to just run them both automatically when the test is executed.

Signature · 3 lines
Thanks,

David / dhildenb
#14 lore raw

Re: [PATCH v6 3/3] selftests/mm: add new selftests for KSM

Quoted 41 lines · David Hildenbrand
Quoted 23 lines · Stefan Roesch

This adds three new tests to the selftests for KSM. These tests use the new prctl API's to enable and disable KSM.

1) add new prctl flags to prctl header file in tools dir

This adds the new prctl flags to the include file prct.h in the tools directory. This makes sure they are available for testing.

2) add KSM prctl merge test

This adds the -t option to the ksm_tests program. The -t flag allows to specify if it should use madvise or prctl ksm merging.

3) add KSM get merge type test

This adds the -G flag to the ksm_tests program to query the KSM status with prctl after KSM has been enabled with prctl.

4) add KSM fork test

Add fork test to verify that the MMF_VM_MERGE_ANY flag is inherited by the child process.

Quoted 8 lines

5) add two functions for debugging merge outcome

This adds two functions to report the metrics in /proc/self/ksm_stat and /sys/kernel/debug/mm/ksm.

The debugging can be enabled with the following command line:
make -C tools/testing/selftests TARGETS="mm" --keep-going \
EXTRA_CFLAGS=-DDEBUG=1

Would it make sense to instead have a "-D" (if still unused) runtime options to print this data? Dead code that's not compiled is a bit unfortunate as it can easily bit-rot.

This patch essentially does two things

Quoted 7 lines

1) Add the option to run all tests/benchmarks with the PRCTL instead of MADVISE

2) Add some functional KSM tests for the new PRCTL (fork, enabling works, disabling works).

The latter should rather go into ksm_functional_tests().

"tools/testing/selftests/mm/ksm_functional_tests.c" is what I wanted to say.

Signature · 3 lines
Thanks,

David / dhildenb
author
#15 lore raw

Re: [PATCH v6 3/3] selftests/mm: add new selftests for KSM

Quoted 44 lines · David Hildenbrand
Quoted 32 lines · David Hildenbrand
Quoted 23 lines · Stefan Roesch

This adds three new tests to the selftests for KSM. These tests use the new prctl API's to enable and disable KSM.

1) add new prctl flags to prctl header file in tools dir

This adds the new prctl flags to the include file prct.h in the tools directory. This makes sure they are available for testing.

2) add KSM prctl merge test

This adds the -t option to the ksm_tests program. The -t flag allows to specify if it should use madvise or prctl ksm merging.

3) add KSM get merge type test

This adds the -G flag to the ksm_tests program to query the KSM status with prctl after KSM has been enabled with prctl.

4) add KSM fork test

Add fork test to verify that the MMF_VM_MERGE_ANY flag is inherited by the child process.

Quoted 8 lines

5) add two functions for debugging merge outcome

This adds two functions to report the metrics in /proc/self/ksm_stat and /sys/kernel/debug/mm/ksm.

The debugging can be enabled with the following command line:
make -C tools/testing/selftests TARGETS="mm" --keep-going \
EXTRA_CFLAGS=-DDEBUG=1

Quoted 9 lines

Would it make sense to instead have a "-D" (if still unused) runtime options to print this data? Dead code that's not compiled is a bit unfortunate as it can easily bit-rot.
This patch essentially does two things
1) Add the option to run all tests/benchmarks with the PRCTL instead of MADVISE
2) Add some functional KSM tests for the new PRCTL (fork, enabling works, disabling works).
The latter should rather go into ksm_functional_tests().

Quoted 1 line

"tools/testing/selftests/mm/ksm_functional_tests.c" is what I wanted to say.

I understood. I'll look into moving the fork check and the disabling into the functional tests for the next version.

author
#16 lore raw

Re: [PATCH v6 3/3] selftests/mm: add new selftests for KSM

In reply to David Hildenbrand
Quoted 22 lines · David Hildenbrand
Quoted 12 lines · Stefan Roesch

This adds three new tests to the selftests for KSM. These tests use the new prctl API's to enable and disable KSM.
1) add new prctl flags to prctl header file in tools dir This adds the new prctl flags to the include file prct.h in the tools directory. This makes sure they are available for testing.
2) add KSM prctl merge test
This adds the -t option to the ksm_tests program. The -t flag allows to specify if it should use madvise or prctl ksm merging.
3) add KSM get merge type test
This adds the -G flag to the ksm_tests program to query the KSM status with prctl after KSM has been enabled with prctl.
4) add KSM fork test

Quoted 8 lines

Add fork test to verify that the MMF_VM_MERGE_ANY flag is inherited by the child process.
5) add two functions for debugging merge outcome
This adds two functions to report the metrics in /proc/self/ksm_stat and /sys/kernel/debug/mm/ksm.
The debugging can be enabled with the following command line:
make -C tools/testing/selftests TARGETS="mm" --keep-going \
EXTRA_CFLAGS=-DDEBUG=1

Quoted 5 lines

Would it make sense to instead have a "-D" (if still unused) runtime options to print this data? Dead code that's not compiled is a bit unfortunate as it can easily bit-rot.

In the next version I'll add -d option. I'll add a global debug variable for this. Otherwise we would need to pass down the debug option several levels.

Quoted 74 lines

This patch essentially does two things

1) Add the option to run all tests/benchmarks with the PRCTL instead of MADVISE

2) Add some functional KSM tests for the new PRCTL (fork, enabling works, disabling works).

The latter should rather go into ksm_functional_tests().

[...]

Quoted 62 lines

-static int check_ksm_unmerge(int mapping, int prot, int timeout, size_t
page_size)

tools/testing/selftests/mm/ksm_tests.c +60 -0
+/* Verify that prctl ksm flag is inherited. */
+static int check_ksm_fork(void)
+{
+	int rc = KSFT_FAIL;
+	pid_t child_pid;
+
+	if (prctl(PR_SET_MEMORY_MERGE, 1)) {
+		perror("prctl");
+		return KSFT_FAIL;
+	}
+
+	child_pid = fork();
+	if (child_pid == 0) {
+		int is_on = prctl(PR_GET_MEMORY_MERGE, 0);
+
+		if (!is_on)
+			exit(KSFT_FAIL);
+
+		exit(KSFT_PASS);
+	}
+
+	if (child_pid < 0)
+		goto out;
+
+	if (waitpid(child_pid, &rc, 0) < 0)
+		rc = KSFT_FAIL;
+
+	if (prctl(PR_SET_MEMORY_MERGE, 0)) {
+		perror("prctl");
+		rc = KSFT_FAIL;
+	}
+
+out:
+	if (rc == KSFT_PASS)
+		printf("OK\n");
+	else
+		printf("Not OK\n");
+
+	return rc;
+}
+
+static int check_ksm_get_merge_type(void)
+{
+	if (prctl(PR_SET_MEMORY_MERGE, 1)) {
+		perror("prctl set");
+		return 1;
+	}
+
+	int is_on = prctl(PR_GET_MEMORY_MERGE, 0);
+
+	if (prctl(PR_SET_MEMORY_MERGE, 0)) {
+		perror("prctl set");
+		return 1;
+	}
+
+	int is_off = prctl(PR_GET_MEMORY_MERGE, 0);
+
+	if (is_on && is_off) {
+		printf("OK\n");
+		return KSFT_PASS;
Quoted 8 lines

tools/testing/selftests/mm/ksm_tests.c static int check_ksm_merge(int mapping, int prot, long page_count, int timeout, · line 448

Quoted 5 lines
tools/testing/selftests/mm/ksm_tests.c +5 -0
+	}
+
+	printf("Not OK\n");
+	return KSFT_FAIL;
+}

Yes, these two are better located in ksm_functional_tests() to just run them both automatically when the test is executed.

I moved the check_ksm_get_merge_type() and check_ksm_fork() to the ksm_functional_test executable. The change will be in the next version.