Message-Id: <20220517092701.1662641-1-xu.xin16@zte.com.cn>
Date:   Tue, 17 May 2022 09:27:01 +0000
From:   cgel.zte@...il.com
To:     akpm@...ux-foundation.org
Cc:     ammarfaizi2@...weeb.org, oleksandr@...alenko.name,
        willy@...radead.org, linux-mm@...ck.org, corbet@....net,
        linux-kernel@...r.kernel.org, xu xin <xu.xin16@....com.cn>,
        Yang Yang <yang.yang29@....com.cn>,
        Ran Xiaokai <ran.xiaokai@....com.cn>,
        wangyong <wang.yong12@....com.cn>,
        Yunkai Zhang <zhang.yunkai@....com.cn>,
        Jiang Xuexin <jiang.xuexin@....com.cn>,
        CGEL <cgel.zte@...il.com>
Subject: [PATCH] mm/ksm: introduce ksm_enabled for each process

From: xu xin <xu.xin16@....com.cn>

For now, if we want to use KSM to merge pages of some apps, we have to
explicitly call madvise() in application code, which means installed
apps on OS needs to be uninstall and source code needs to be modified.
It is very inconvenient because sometimes users or app developers are not
willing to modify their app source codes for any reasons.

So to use KSM more flexibly, we provide a new proc file "ksm_enabled" under
/proc/<pid>/. We can pass parameter into this file with one of three values
as follows:

	always:
		force all anonymous and eligible VMAs of this process to be
		scanned by ksmd.
	madvise:
		the default state, unless user code call madvise, ksmd
		doesn't scan this process.
	never:
		this process will never be scanned by ksmd and no merged
		pages occurred in this process.

With this patch, we can control KSM with ``/proc/<pid>/ksm_enabled``
based on every process. KSM for each process can be entirely disabled
(mostly for debugging purposes) or only enabled inside MADV_MERGEABLE
regions (to avoid the risk of consuming more cpu resources to scan for
ksmd) or enabled entirely for a process.

Signed-off-by: xu xin <xu.xin16@....com.cn>
Reviewed-by: Yang Yang <yang.yang29@....com.cn>
Reviewed-by: Ran Xiaokai <ran.xiaokai@....com.cn>
Reviewed-by: wangyong <wang.yong12@....com.cn>
Reviewed-by: Yunkai Zhang <zhang.yunkai@....com.cn>
Reviewed-by: Jiang Xuexin <jiang.xuexin@....com.cn>
Signed-off-by: CGEL <cgel.zte@...il.com>
---
 Documentation/admin-guide/mm/ksm.rst |  24 ++++++-
 Documentation/filesystems/proc.rst   |  14 ++++
 fs/proc/base.c                       | 102 ++++++++++++++++++++++++++-
 include/linux/ksm.h                  |   5 ++
 include/linux/mm_types.h             |  10 +++
 mm/ksm.c                             |  35 ++++++++-
 6 files changed, 185 insertions(+), 5 deletions(-)

diff --git a/Documentation/admin-guide/mm/ksm.rst b/Documentation/admin-guide/mm/ksm.rst
index b244f0202a03..91326198e37f 100644
--- a/Documentation/admin-guide/mm/ksm.rst
+++ b/Documentation/admin-guide/mm/ksm.rst
@@ -32,7 +32,7 @@ are swapped back in: ksmd must rediscover their identity and merge again).
 Controlling KSM with madvise
 ============================
 
-KSM only operates on those areas of address space which an application
+KSM can operate on those areas of address space which an application
 has advised to be likely candidates for merging, by using the madvise(2)
 system call::
 
@@ -70,6 +70,28 @@ Applications should be considerate in their use of MADV_MERGEABLE,
 restricting its use to areas likely to benefit.  KSM's scans may use a lot
 of processing power: some installations will disable KSM for that reason.
 
+Controlling KSM with procfs
+===========================
+We can also control KSM with ``/proc/<pid>/ksm_enabled`` based on every
+process. KSM for each process can be entirely disabled (mostly for
+debugging purposes) or only enabled inside MADV_MERGEABLE regions (to avoid
+the risk of consuming more cpu resources to scan for ksmd) or enabled entirely
+for a process. This can be achieved with one of::
+
+    echo always > /proc/<pid>/ksm_enabled
+    echo madvise > /proc/<pid>/ksm_enabled
+    echo never > /proc/<pid>/ksm_enabled
+
+always:
+        force all anonymous and eligible VMAs of this process to be scanned
+        by ksmd.
+madvise:
+        the default state, unless user code call madvise, ksmd doesn't scan
+        this process.
+never:
+        this process will never be scanned by ksmd and no merged pages
+	occurred in this process.
+
 .. _ksm_sysfs:
 
 KSM daemon sysfs interface
diff --git a/Documentation/filesystems/proc.rst b/Documentation/filesystems/proc.rst
index 1bc91fb8c321..ea7e08a1c143 100644
--- a/Documentation/filesystems/proc.rst
+++ b/Documentation/filesystems/proc.rst
@@ -47,6 +47,7 @@ fixes/update part 1.1  Stefani Seibold <stefani@...bold.net>    June 9 2009
   3.10  /proc/<pid>/timerslack_ns - Task timerslack value
   3.11	/proc/<pid>/patch_state - Livepatch patch operation state
   3.12	/proc/<pid>/arch_status - Task architecture specific information
+  3.13  /proc/<pid>/ksm_enabled - Controlling KSM based on process
 
   4	Configuring procfs
   4.1	Mount options
@@ -2140,6 +2141,19 @@ AVX512_elapsed_ms
   the task is unlikely an AVX512 user, but depends on the workload and the
   scheduling scenario, it also could be a false negative mentioned above.
 
+3.13 /proc/<pid>/ksm_enabled - Controlling KSM based on process
+---------------------------------------------------------------
+When CONFIG_KSM is enabled, this file can be used to specify how this
+process's anonymous memory gets involved in KSM scanning.
+
+If writing "always" to this file, it will force all anonymous and eligible
+VMAs of this process to be scanned by ksmd.
+
+If writing "madvise" to this file, turn to the default state, unless user
+code call madvise, ksmd doesn't scan this process.
+
+If writing "never" to this file, this process will never be scanned by ksmd.
+
 Chapter 4: Configuring procfs
 =============================
 
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 617816168748..760ceeab4aa1 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -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"
@@ -3171,7 +3172,104 @@ static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *
 
 	return 0;
 }
-#endif /* CONFIG_KSM */
+
+static int ksm_enabled_show(struct seq_file *m, void *v)
+{
+	struct inode *inode = m->private;
+	struct mm_struct *mm;
+	struct task_struct *task = get_proc_task(inode);
+
+	if (!task)
+		return -ESRCH;
+
+	mm = get_task_mm(task);
+	if (mm) {
+		if (mm->ksm_enabled == KSM_PROC_ALWAYS)
+			seq_puts(m, "[always] madvise never\n");
+		else if (mm->ksm_enabled == KSM_PROC_MADVISE)
+			seq_puts(m, "always [madvise] never\n");
+		else
+			seq_puts(m, "always madvise [never]\n");
+		mmput(mm);
+	}
+
+	put_task_struct(task);
+	return 0;
+}
+
+static ssize_t ksm_enabled_write(struct file *file, const char __user *buf,
+								size_t count, loff_t *ppos)
+{
+	struct task_struct *task;
+	struct mm_struct *mm;
+	char buffer[PROC_NUMBUF];
+	int value;
+	int err = 0;
+	long str_len;
+
+	if (count > sizeof(buffer) - 1)
+		count = sizeof(buffer) - 1;
+	str_len = strncpy_from_user(buffer, buf, count);
+	if (str_len < 0)
+		return -EFAULT;
+	buffer[str_len - 1] = '\0';
+
+	if (!strcmp("always", buffer))
+		value = KSM_PROC_ALWAYS;
+	else if (!strcmp("madvise", buffer))
+		value = KSM_PROC_MADVISE;
+	else if (!strcmp("never", buffer))
+		value = KSM_PROC_NEVER;
+	else
+		return -EINVAL;
+
+	task = get_proc_task(file_inode(file));
+	if (!task)
+		return -ESRCH;
+	mm = get_task_mm(task);
+	if (!mm)
+		goto out_put_task;
+
+	if (mm->ksm_enabled != value) {
+		if (mmap_write_lock_killable(mm)) {
+			err = -EINTR;
+			goto out_mmput;
+		}
+		if (value == KSM_PROC_NEVER)
+			mm->ksm_enabled = value;
+		else {
+			/*
+			 * No matter whether it's KSM_PROC_ALWAYS or KSM_PROC_MADVISE, we need
+			 * to recheck mm->flags to guarantee that this mm is in ksm_scan.
+			 */
+			if (!test_bit(MMF_VM_MERGEABLE, &mm->flags))
+				err = __ksm_enter(mm);
+			if (!err)
+				mm->ksm_enabled = value;
+		}
+		mmap_write_unlock(mm);
+	}
+
+out_mmput:
+	mmput(mm);
+out_put_task:
+	put_task_struct(task);
+	return err < 0 ? err : count;
+}
+
+static int ksm_enabled_open(struct inode *inode, struct file *filp)
+{
+	return single_open(filp, ksm_enabled_show, inode);
+}
+
+static const struct file_operations proc_pid_ksm_enabled_operations = {
+	.open       = ksm_enabled_open,
+	.read		= seq_read,
+	.write		= ksm_enabled_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+#endif /*CONFIG_KSM */
 
 #ifdef CONFIG_STACKLEAK_METRICS
 static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
@@ -3306,6 +3404,7 @@ static const struct pid_entry tgid_base_stuff[] = {
 #endif
 #ifdef CONFIG_KSM
 	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
+	REG("ksm_enabled", S_IRUGO|S_IWUSR, proc_pid_ksm_enabled_operations),
 #endif
 };
 
@@ -3642,6 +3741,7 @@ static const struct pid_entry tid_base_stuff[] = {
 #endif
 #ifdef CONFIG_KSM
 	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
+	REG("ksm_enabled", S_IRUGO|S_IWUSR, proc_pid_ksm_enabled_operations),
 #endif
 };
 
diff --git a/include/linux/ksm.h b/include/linux/ksm.h
index 0b4f17418f64..29d23d208b54 100644
--- a/include/linux/ksm.h
+++ b/include/linux/ksm.h
@@ -19,6 +19,11 @@ struct stable_node;
 struct mem_cgroup;
 
 #ifdef CONFIG_KSM
+
+#define KSM_PROC_MADVISE	0
+#define KSM_PROC_ALWAYS		1
+#define KSM_PROC_NEVER		2
+
 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);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 417ef1519475..29fd4c84d08c 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -649,6 +649,16 @@ struct mm_struct {
 		 * merging.
 		 */
 		unsigned long ksm_merging_pages;
+
+		/*
+		 * Represent the state of this mm involing in KSM, with 3 states:
+		 * 1) KSM_PROC_ALWAYS:  force all anonymous VMAs of this process to
+		 *						be scanned.
+		 * 2) KSM_PROC_MADVISE: the default state, unless user code call
+		 *						madvise, don't scan this process.
+		 * 3) KSM_PROC_NEVER: never be involed in KSM.
+		 */
+		int ksm_enabled;
 #endif
 	} __randomize_layout;
 
diff --git a/mm/ksm.c b/mm/ksm.c
index 26da7f813f23..90cc8eda8bca 100644
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -334,6 +334,35 @@ static void __init ksm_slab_free(void)
 	mm_slot_cache = NULL;
 }
 
+static bool vma_scannable(struct vm_area_struct *vma)
+{
+	unsigned long vm_flags = vma->vm_flags;
+	struct mm_struct *mm = vma->vm_mm;
+
+	if (mm->ksm_enabled == KSM_PROC_NEVER ||
+			(!(vma->vm_flags & VM_MERGEABLE) &&
+			mm->ksm_enabled != KSM_PROC_ALWAYS))
+		return false;
+
+	if (vm_flags & (VM_SHARED	| VM_MAYSHARE	|
+			VM_PFNMAP	| VM_IO | VM_DONTEXPAND |
+			VM_HUGETLB	| VM_MIXEDMAP))
+		return false;       /* just ignore this vma*/
+
+	if (vma_is_dax(vma))
+		return false;
+#ifdef VM_SAO
+	if (vm_flags & VM_SAO)
+		return false;
+#endif
+#ifdef VM_SPARC_ADI
+	if (vm_flags & VM_SPARC_ADI)
+		return false;
+#endif
+
+	return true;
+}
+
 static __always_inline bool is_stable_node_chain(struct stable_node *chain)
 {
 	return chain->rmap_hlist_len == STABLE_NODE_CHAIN;
@@ -523,7 +552,7 @@ static struct vm_area_struct *find_mergeable_vma(struct mm_struct *mm,
 	if (ksm_test_exit(mm))
 		return NULL;
 	vma = vma_lookup(mm, addr);
-	if (!vma || !(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
+	if (!vma || !vma_scannable(vma) || !vma->anon_vma)
 		return NULL;
 	return vma;
 }
@@ -990,7 +1019,7 @@ static int unmerge_and_remove_all_rmap_items(void)
 		for_each_vma(vmi, vma) {
 			if (ksm_test_exit(mm))
 				break;
-			if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
+			if (!vma_scannable(vma) || !vma->anon_vma)
 				continue;
 			err = unmerge_ksm_pages(vma,
 						vma->vm_start, vma->vm_end);
@@ -2300,7 +2329,7 @@ static struct rmap_item *scan_get_next_rmap_item(struct page **page)
 		goto no_vmas;
 
 	for_each_vma(vmi, vma) {
-		if (!(vma->vm_flags & VM_MERGEABLE))
+		if (!vma_scannable(vma))
 			continue;
 		if (ksm_scan.address < vma->vm_start)
 			ksm_scan.address = vma->vm_start;
-- 
2.25.1

Powered by blists - more mailing lists