On Wed, Mar 07, 2007 at 11:00:31PM +0530, Srivatsa Vaddagiri wrote:
> > I'd like to see that. I suspect it will be a bit more fiddly than the
> > simple cpu_acct subsystem.
> 
> I am almost done with the conversion. And yes cpuset is a beast to
> convert over! Will test and send the patches out tomorrow.

Ok ..I am not in a state yet where I can post the patches to lkml in the
usual conventions (breaking down neatly/good documentation etc). But I
do have something which seems to work! I could mount cpuset as:

	mount -t rcfs -ocpuset none cpuset
	cd cpuset
	mkdir a
	cd a
	cat tasks	# shows nothing
	echo 7 > cpus
	echo 0 > mems
	echo 1 > cpu_exclusive
	echo some_pid > tasks
	cat tasks	# shows some_pid

top now shows some_pid running on CPU7, as expected :)
	
Instead of the usual convention of inlining patches and sending them in 
separate mails, I am sending all of them as attachments (beware, bugs around!). 
But this gives you an idea on which direction this is proceeding ..

Todo:

        - Introduce refcounting of resource objects (get/put_res_ns)
	- rmdir needs to check resource object refcount rather than
	  nsproxy's
	- Trace couple of other lockdep warnings I have hit

Patches attached.


-- 
Regards,
vatsa


---

 linux-2.6.20-vatsa/include/linux/init_task.h |   11 
 linux-2.6.20-vatsa/include/linux/nsproxy.h   |   11 
 linux-2.6.20-vatsa/init/Kconfig              |   22 
 linux-2.6.20-vatsa/init/main.c               |    3 
 linux-2.6.20-vatsa/kernel/Makefile           |    1 


---

 linux-2.6.20.1-vatsa/include/linux/init_task.h |   11 
 linux-2.6.20.1-vatsa/include/linux/nsproxy.h   |   11 
 linux-2.6.20.1-vatsa/include/linux/rcfs.h      |   76 +
 linux-2.6.20.1-vatsa/init/Kconfig              |   22 
 linux-2.6.20.1-vatsa/init/main.c               |    3 
 linux-2.6.20.1-vatsa/kernel/Makefile           |    1 
 linux-2.6.20.1-vatsa/kernel/nsproxy.c          |   65 +
 linux-2.6.20.1-vatsa/kernel/rcfs.c             | 1202 +++++++++++++++++++++++++
 8 files changed, 1391 insertions(+)

diff -puN include/linux/init_task.h~rcfs include/linux/init_task.h
--- linux-2.6.20.1/include/linux/init_task.h~rcfs	2007-03-08 21:21:33.000000000 +0530
+++ linux-2.6.20.1-vatsa/include/linux/init_task.h	2007-03-08 21:21:34.000000000 +0530
@@ -71,6 +71,16 @@
 }
 
 extern struct nsproxy init_nsproxy;
+
+#ifdef CONFIG_RCFS
+#define INIT_RCFS(nsproxy) 						\
+	.list 		= LIST_HEAD_INIT(nsproxy.list),			\
+	.ctlr_data 	= {[ 0 ... CONFIG_MAX_RC_SUBSYS-1 ] = NULL },
+#else
+#define INIT_RCFS(nsproxy)
+#endif
+
+
 #define INIT_NSPROXY(nsproxy) {						\
 	.pid_ns		= &init_pid_ns,					\
 	.count		= ATOMIC_INIT(1),				\
@@ -78,6 +88,7 @@ extern struct nsproxy init_nsproxy;
 	.uts_ns		= &init_uts_ns,					\
 	.mnt_ns		= NULL,						\
 	INIT_IPC_NS(ipc_ns)						\
+	INIT_RCFS(nsproxy)						\
 }
 
 #define INIT_SIGHAND(sighand) {						\
diff -puN include/linux/nsproxy.h~rcfs include/linux/nsproxy.h
--- linux-2.6.20.1/include/linux/nsproxy.h~rcfs	2007-03-08 21:21:33.000000000 +0530
+++ linux-2.6.20.1-vatsa/include/linux/nsproxy.h	2007-03-08 21:21:34.000000000 +0530
@@ -28,6 +28,10 @@ struct nsproxy {
 	struct ipc_namespace *ipc_ns;
 	struct mnt_namespace *mnt_ns;
 	struct pid_namespace *pid_ns;
+#ifdef CONFIG_RCFS
+	struct list_head list;
+	void *ctlr_data[CONFIG_MAX_RC_SUBSYS];
+#endif
 };
 extern struct nsproxy init_nsproxy;
 
@@ -35,6 +39,13 @@ struct nsproxy *dup_namespaces(struct ns
 int copy_namespaces(int flags, struct task_struct *tsk);
 void get_task_namespaces(struct task_struct *tsk);
 void free_nsproxy(struct nsproxy *ns);
+#ifdef CONFIG_RCFS
+struct nsproxy *find_nsproxy(struct nsproxy *ns);
+int namespaces_init(void);
+int nsproxy_task_count(void *data, int idx);
+#else
+static inline int namespaces_init(void) { return 0;}
+#endif
 
 static inline void put_nsproxy(struct nsproxy *ns)
 {
diff -puN /dev/null include/linux/rcfs.h
--- /dev/null	2007-03-08 22:46:54.325490448 +0530
+++ linux-2.6.20.1-vatsa/include/linux/rcfs.h	2007-03-08 21:21:34.000000000 +0530
@@ -0,0 +1,76 @@
+#ifndef _LINUX_RCFS_H
+#define _LINUX_RCFS_H
+
+#ifdef CONFIG_RCFS
+
+/* struct cftype:
+ *
+ * The files in the container filesystem mostly have a very simple read/write
+ * handling, some common function will take care of it. Nevertheless some cases
+ * (read tasks) are special and therefore I define this structure for every
+ * kind of file.
+ *
+ *
+ * When reading/writing to a file:
+ *	- the container to use in file->f_dentry->d_parent->d_fsdata
+ *	- the 'cftype' of the file is file->f_dentry->d_fsdata
+ */
+
+struct inode;
+#define MAX_CFTYPE_NAME 64
+struct cftype {
+	/* By convention, the name should begin with the name of the
+	 * subsystem, followed by a period */
+	char name[MAX_CFTYPE_NAME];
+	int private;
+	int (*open) (struct inode *inode, struct file *file);
+	ssize_t (*read) (struct nsproxy *ns, struct cftype *cft,
+			 struct file *file,
+			 char __user *buf, size_t nbytes, loff_t *ppos);
+	ssize_t (*write) (struct nsproxy *ns, struct cftype *cft,
+			  struct file *file,
+			  const char __user *buf, size_t nbytes, loff_t *ppos);
+	int (*release) (struct inode *inode, struct file *file);
+};
+
+/* resource control subsystem type. See Documentation/rcfs.txt for details */
+
+struct rc_subsys {
+	int (*create)(struct rc_subsys *ss, struct nsproxy *ns,
+			 struct nsproxy *parent);
+	void (*destroy)(struct rc_subsys *ss, struct nsproxy *ns);
+	int (*can_attach)(struct rc_subsys *ss, struct nsproxy *ns,
+				 struct task_struct *tsk);
+	void (*attach)(struct rc_subsys *ss, struct nsproxy *new,
+				 struct nsproxy *old, struct task_struct *tsk);
+	int (*populate)(struct rc_subsys *ss, struct dentry *d);
+	int subsys_id;
+	int active;
+
+#define MAX_CONTAINER_TYPE_NAMELEN 32
+	const char *name;
+
+	/* Protected by RCU */
+	int hierarchy;
+
+	struct list_head sibling;
+};
+
+int rc_register_subsys(struct rc_subsys *subsys);
+/* Add a new file to the given container directory. Should only be
+ * called by subsystems from within a populate() method */
+int rcfs_add_file(struct dentry *d, const struct cftype *cft);
+extern int rcfs_init(void);
+extern void rcfs_manage_lock(void);
+extern void rcfs_manage_unlock(void);
+extern int rcfs_dir_removed(struct dentry *d);
+extern int rcfs_path(struct dentry *d, char *buf, int len);
+
+#else
+
+static inline int rcfs_init(void) { return 0; }
+
+#endif
+
+
+#endif
diff -puN init/Kconfig~rcfs init/Kconfig
--- linux-2.6.20.1/init/Kconfig~rcfs	2007-03-08 21:21:33.000000000 +0530
+++ linux-2.6.20.1-vatsa/init/Kconfig	2007-03-08 22:47:50.000000000 +0530
@@ -238,6 +238,28 @@ config IKCONFIG_PROC
 	  This option enables access to the kernel configuration file
 	  through /proc/config.gz.
 
+config RCFS
+	bool "Resource control file system support"
+	default n
+	help
+	  This option will let you create and manage resource containers,
+	  which can be used to aggregate multiple processes, e.g. for
+	  the purposes of resource tracking.
+
+	  Say N if unsure
+
+config MAX_RC_SUBSYS
+       int "Number of resource control subsystems to support"
+       depends on RCFS
+       range 1 255
+       default 8
+
+config MAX_RC_HIERARCHIES
+       int "Number of rcfs hierarchies to support"
+       depends on RCFS
+       range 2 255
+       default 4
+
 config CPUSETS
 	bool "Cpuset support"
 	depends on SMP
diff -puN init/main.c~rcfs init/main.c
--- linux-2.6.20.1/init/main.c~rcfs	2007-03-08 21:21:33.000000000 +0530
+++ linux-2.6.20.1-vatsa/init/main.c	2007-03-08 21:21:34.000000000 +0530
@@ -52,6 +52,7 @@
 #include <linux/lockdep.h>
 #include <linux/pid_namespace.h>
 #include <linux/device.h>
+#include <linux/rcfs.h>
 
 #include <asm/io.h>
 #include <asm/bugs.h>
@@ -512,6 +513,7 @@ asmlinkage void __init start_kernel(void
 	setup_per_cpu_areas();
 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
 
+	namespaces_init();
 	/*
 	 * Set up the scheduler prior starting any interrupts (such as the
 	 * timer interrupt). Full topology setup happens at smp_init()
@@ -578,6 +580,7 @@ asmlinkage void __init start_kernel(void
 	}
 #endif
 	vfs_caches_init_early();
+	rcfs_init();
 	cpuset_init_early();
 	mem_init();
 	kmem_cache_init();
diff -puN kernel/Makefile~rcfs kernel/Makefile
--- linux-2.6.20.1/kernel/Makefile~rcfs	2007-03-08 21:21:34.000000000 +0530
+++ linux-2.6.20.1-vatsa/kernel/Makefile	2007-03-08 22:47:50.000000000 +0530
@@ -50,6 +50,7 @@ obj-$(CONFIG_RELAY) += relay.o
 obj-$(CONFIG_UTS_NS) += utsname.o
 obj-$(CONFIG_TASK_DELAY_ACCT) += delayacct.o
 obj-$(CONFIG_TASKSTATS) += taskstats.o tsacct.o
+obj-$(CONFIG_RCFS) += rcfs.o
 
 ifneq ($(CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER),y)
 # According to Alan Modra <alan@linuxcare.com.au>, the -fno-omit-frame-pointer is
diff -puN kernel/nsproxy.c~rcfs kernel/nsproxy.c
--- linux-2.6.20.1/kernel/nsproxy.c~rcfs	2007-03-08 21:21:34.000000000 +0530
+++ linux-2.6.20.1-vatsa/kernel/nsproxy.c	2007-03-08 22:54:04.000000000 +0530
@@ -23,6 +23,11 @@
 
 struct nsproxy init_nsproxy = INIT_NSPROXY(init_nsproxy);
 
+#ifdef CONFIG_RCFS
+static LIST_HEAD(nslisthead);
+static DEFINE_SPINLOCK(nslistlock);
+#endif
+
 static inline void get_nsproxy(struct nsproxy *ns)
 {
 	atomic_inc(&ns->count);
@@ -71,6 +76,12 @@ struct nsproxy *dup_namespaces(struct ns
 			get_pid_ns(ns->pid_ns);
 	}
 
+#ifdef CONFIG_RCFS
+	spin_lock_irq(&nslistlock);
+	list_add(&ns->list, &nslisthead);
+	spin_unlock_irq(&nslistlock);
+#endif
+
 	return ns;
 }
 
@@ -145,5 +156,59 @@ void free_nsproxy(struct nsproxy *ns)
 		put_ipc_ns(ns->ipc_ns);
 	if (ns->pid_ns)
 		put_pid_ns(ns->pid_ns);
+#ifdef CONFIG_RCFS
+	spin_lock_irq(&nslistlock);
+	list_del(&ns->list);
+	spin_unlock_irq(&nslistlock);
+#endif
 	kfree(ns);
 }
+
+#ifdef CONFIG_RCFS
+struct nsproxy *find_nsproxy(struct nsproxy *target)
+{
+	struct nsproxy *ns;
+	int i =