Skip to contentSkip to conversation

support separate namespaces for sysv

RFC PATCH NAK Dave Hansen 6 participants 29 messages Mar 6, 2006 – Mar 20, 2006

Subsystems (2)
lore
Share & tools
lore
#1 lore raw

RFC PATCH 1/6 prepare sysctls for containers

In reply to Dave Hansen

Right now, sysctls can only deal with global variables. This patch makes them a _little_ more flexible by allowing there to be an accessor function to get at the variable being changed, instead of it being global.

This allows the sysctls to be backed by variables that are, for instance, dynamically allocated and not available at compile-time.

This also provides a very simple mechanism to take things that are currently global and containerize them.

---

 work-dave/include/linux/sysctl.h |    8 ++++
 work-dave/kernel/sysctl.c        |   65 ++++++++++++++++++++++++++-------------
 2 files changed, 52 insertions(+), 21 deletions(-)
include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +8 -0
diff -puN include/linux/sysctl.h~sysctls-for-containers include/linux/sysctl.h
@@ -872,6 +872,7 @@ extern void sysctl_init(void);
 
 typedef struct ctl_table ctl_table;
 
+typedef void *ctl_data_access (void);
 typedef int ctl_handler (ctl_table *table, int __user *name, int nlen,
 			 void __user *oldval, size_t __user *oldlenp,
 			 void __user *newval, size_t newlen, 
@@ -957,6 +958,13 @@ struct ctl_table 
 	int ctl_name;			/* Binary ID */
 	const char *procname;		/* Text ID for /proc/sys, or zero */
 	void *data;
+	ctl_data_access *data_access;	/* set this to a function if you
+					 * don't have a static place to point
+					 * ->data at compile-time.  This
+					 * function will be called to dynamically
+					 * figure out a ->data pointer.  Do not
+					 * set this and ->data at once.
+					 */
 	int maxlen;
 	mode_t mode;
 	ctl_table *child;
kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +44 -21

194 lines not on this page · raw message

_

#23 lore raw

RFC PATCH 3/6 sysvmsg: containerize sysctls

In reply to Dave Hansen

Use the new sysctl data accessor function pointers to put the sysvmsg sysctl variables inside of the context structure.

Note that this will effectively remove the system-wide limits on sysv msg resources that the sysctl mechanism currently provides and move it to a per-container limit. This is currently by design, and may be later addressed when we have proper large-scale resource controls in the kernel.

---

 work-dave/ipc/msg.c       |   38 +++++++++++++++++++++++++-------------
 work-dave/ipc/util.h      |    4 ++++
 work-dave/kernel/sysctl.c |   12 ++++++------
 3 files changed, 35 insertions(+), 19 deletions(-)
ipc/msg.c renamed from ipc/msg.c~sysvmsg-containerize-sysctls +25 -13
diff -puN ipc/msg.c~sysvmsg-containerize-sysctls ipc/msg.c
@@ -32,11 +32,6 @@
 #include <asm/uaccess.h>
 #include "util.h"
 
-/* sysctl: */
-int msg_ctlmax = MSGMAX;
-int msg_ctlmnb = MSGMNB;
-int msg_ctlmni = MSGMNI;
-
 /* one msg_receiver structure for each sleeping receiver */
 struct msg_receiver {
 	struct list_head r_list;
@@ -76,7 +71,11 @@ static int sysvipc_msg_proc_show(struct 
 
 void __init msg_init (struct ipc_msg_context *context)
 {
-	ipc_init_ids(&context->ids,msg_ctlmni);
+	context->ctlmax = MSGMAX;
+	context->ctlmnb = MSGMNB;
+	context->ctlmni = MSGMNI;
+
+	ipc_init_ids(&context->ids,context->ctlmni);
 	ipc_init_proc_interface("sysvipc/msg",
 				"       key      msqid perms      cbytes       qnum lspid lrpid   uid   gid  cuid  cgid      stime      rtime      ctime\n",
 				&context->ids,
@@ -103,7 +102,7 @@ static int newque (struct ipc_msg_contex
 		return retval;
 	}
 
-	id = ipc_addid(&context->ids, &msq->q_perm, msg_ctlmni);
+	id = ipc_addid(&context->ids, &msq->q_perm, context->ctlmni);
 	if(id == -1) {
 		security_msg_queue_free(msq);
 		ipc_rcu_putref(msq);
@@ -114,7 +113,7 @@ static int newque (struct ipc_msg_contex
 	msq->q_stime = msq->q_rtime = 0;
 	msq->q_ctime = get_seconds();
 	msq->q_cbytes = msq->q_qnum = 0;
-	msq->q_qbytes = msg_ctlmnb;
+	msq->q_qbytes = context->ctlmnb;
 	msq->q_lspid = msq->q_lrpid = 0;
 	INIT_LIST_HEAD(&msq->q_messages);
 	INIT_LIST_HEAD(&msq->q_receivers);
@@ -356,9 +355,9 @@ asmlinkage long sys_msgctl (int msqid, i
 			return err;
 
 		memset(&msginfo,0,sizeof(msginfo));	
-		msginfo.msgmni = msg_ctlmni;
-		msginfo.msgmax = msg_ctlmax;
-		msginfo.msgmnb = msg_ctlmnb;
+		msginfo.msgmni = context->ctlmni;
+		msginfo.msgmax = context->ctlmax;
+		msginfo.msgmnb = context->ctlmnb;
 		msginfo.msgssz = MSGSSZ;
 		msginfo.msgseg = MSGSEG;
 		down(&context->ids.sem);
@@ -462,7 +461,7 @@ asmlinkage long sys_msgctl (int msqid, i
 	case IPC_SET:
 	{
 		err = -EPERM;
-		if (setbuf.qbytes > msg_ctlmnb && !capable(CAP_SYS_RESOURCE))
+		if (setbuf.qbytes > context->ctlmnb && !capable(CAP_SYS_RESOURCE))
 			goto out_unlock_up;
 
 		msq->q_qbytes = setbuf.qbytes;
@@ -560,7 +559,7 @@ asmlinkage long sys_msgsnd (int msqid, s
 	int err;
 	struct ipc_msg_context *context = &current->ipc_context->msg;
 
-	if (msgsz > msg_ctlmax || (long) msgsz < 0 || msqid < 0)
+	if (msgsz > context->ctlmax || (long) msgsz < 0 || msqid < 0)
 		return -EINVAL;
 	if (get_user(mtype, &msgp->mtype))
 		return -EFAULT; 
@@ -835,3 +834,16 @@ static int sysvipc_msg_proc_show(struct 
 			  msq->q_ctime);
 }
 #endif
+
+void *msg_ctlmnb_helper(void)
+{
+	return &current->ipc_context->msg.ctlmnb;
+}
+void *msg_ctlmni_helper(void)
+{
+	return &current->ipc_context->msg.ctlmni;
+}
+void *msg_ctlmax_helper(void)
+{
+	return &current->ipc_context->msg.ctlmax;
+}
ipc/util.h renamed from ipc/util.h~sysvmsg-containerize-sysctls +4 -0
diff -puN ipc/util.h~sysvmsg-containerize-sysctls ipc/util.h
@@ -34,6 +34,10 @@ struct ipc_msg_context {
 	atomic_t bytes;
 	atomic_t hdrs;
 
+	int ctlmax;
+	int ctlmnb;
+	int ctlmni;
+
 	struct ipc_ids ids;
 	struct kref count;
 };
kernel/sysctl.c renamed from kernel/sysctl.c~sysvmsg-containerize-sysctls +6 -6
diff -puN kernel/sysctl.c~sysvmsg-containerize-sysctls kernel/sysctl.c
@@ -94,10 +94,10 @@ extern int sg_big_buff;
 #ifdef CONFIG_SYSVIPC
 extern size_t shm_ctlmax;
 extern size_t shm_ctlall;
+extern void *msg_ctlmnb_helper(void);
+extern void *msg_ctlmni_helper(void);
+extern void *msg_ctlmax_helper(void);
 extern int shm_ctlmni;
-extern int msg_ctlmax;
-extern int msg_ctlmnb;
-extern int msg_ctlmni;
 extern int sem_ctls[];
 #endif
 
@@ -452,7 +452,7 @@ static ctl_table kern_table[] = {
 	{
 		.ctl_name	= KERN_MSGMAX,
 		.procname	= "msgmax",
-		.data		= &msg_ctlmax,
+		.data_access	= &msg_ctlmax_helper,
 		.maxlen		= sizeof (int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec,
@@ -460,7 +460,7 @@ static ctl_table kern_table[] = {
 	{
 		.ctl_name	= KERN_MSGMNI,
 		.procname	= "msgmni",
-		.data		= &msg_ctlmni,
+		.data_access	= &msg_ctlmni_helper,
 		.maxlen		= sizeof (int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec,
@@ -468,7 +468,7 @@ static ctl_table kern_table[] = {
 	{
 		.ctl_name	= KERN_MSGMNB,
 		.procname	=  "msgmnb",
-		.data		= &msg_ctlmnb,
+		.data_access	= &msg_ctlmnb_helper,
 		.maxlen		= sizeof (int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec,

_

#15 lore raw

RFC PATCH 2/6 sysvmsg: containerize

In reply to Dave Hansen

Take all of the global sysv msg context and put it inside of a special structure. This allows future work to give different containers completely different views into the sysv space.

I'm still using the "_context" nomenclature here. The consensus seems to be to use something like "namespace". Please speak up if you have any other suggestions. I'll fix it up before submitting upstream.

---

 work-dave/desc.txt                  |    4 +
 work-dave/include/linux/init_task.h |    1 
 work-dave/include/linux/ipc.h       |    4 +
 work-dave/include/linux/sched.h     |    1 
 work-dave/ipc/msg.c                 |  138 ++++++++++++++++++------------------
 work-dave/ipc/util.c                |    8 +-
 work-dave/ipc/util.h                |   14 +++
 7 files changed, 99 insertions(+), 71 deletions(-)
include/linux/init_task.h renamed from include/linux/init_task.h~sysvmsg-container +1 -0
diff -puN include/linux/init_task.h~sysvmsg-container include/linux/init_task.h
@@ -2,6 +2,7 @@
 #define _LINUX__INIT_TASK_H
 
 #include <linux/file.h>
+#include <linux/ipc.h>
 #include <linux/rcupdate.h>
 
 #define INIT_FDTABLE \
include/linux/ipc.h renamed from include/linux/ipc.h~sysvmsg-container +4 -0
diff -puN include/linux/ipc.h~sysvmsg-container include/linux/ipc.h
@@ -68,6 +68,10 @@ struct kern_ipc_perm
 	void		*security;
 };
 
+struct ipc_context;
+extern struct ipc_context *ipc_context;
+
+extern struct ipc_msg_context *init_ipc_context(void);
 #endif /* __KERNEL__ */
 
 #endif /* _LINUX_IPC_H */
include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +1 -0
diff -puN include/linux/sched.h~sysvmsg-container include/linux/sched.h
@@ -793,6 +793,7 @@ struct task_struct {
 	int link_count, total_link_count;
 /* ipc stuff */
 	struct sysv_sem sysvsem;
+	struct ipc_context *ipc_context;
 /* CPU-specific state of this task */
 	struct thread_struct thread;
 /* filesystem information */
ipc/msg.c renamed from ipc/msg.c~sysvmsg-container +70 -68

351 lines not on this page · raw message

ipc/util.c renamed from ipc/util.c~sysvmsg-container +6 -2
diff -puN ipc/util.c~sysvmsg-container ipc/util.c
@@ -45,11 +45,15 @@ struct ipc_proc_iface {
  *	The various system5 IPC resources (semaphores, messages and shared
  *	memory are initialised
  */
- 
+
 static int __init ipc_init(void)
 {
+	init_task.ipc_context = kzalloc(sizeof(*ipc_context), GFP_KERNEL);
+	if (!init_task.ipc_context)
+		return -ENOMEM;
+
 	sem_init();
-	msg_init();
+	msg_init(&init_task.ipc_context->msg);
 	shm_init();
 	return 0;
 }
ipc/util.h renamed from ipc/util.h~sysvmsg-container +13 -1
diff -puN ipc/util.h~sysvmsg-container ipc/util.h
@@ -12,7 +12,7 @@
 #define SEQ_MULTIPLIER	(IPCMNI)
 
 void sem_init (void);
-void msg_init (void);
+void msg_init (struct ipc_msg_context *context);
 void shm_init (void);
 
 struct ipc_id_ary {
@@ -30,6 +30,18 @@ struct ipc_ids {
 	struct ipc_id_ary* entries;
 };
 
+struct ipc_msg_context {
+	atomic_t bytes;
+	atomic_t hdrs;
+
+	struct ipc_ids ids;
+	struct kref count;
+};
+
+struct ipc_context {
+	struct ipc_msg_context msg;
+};
+
 struct seq_file;
 void __init ipc_init_ids(struct ipc_ids* ids, int size);
 #ifdef CONFIG_PROC_FS

diff -puN kernel/sysctl.c~sysvmsg-container kernel/sysctl.c

desc.txt new file +4 -0
diff -puN /dev/null desc.txt
@@ -0,0 +1,4 @@
+DESC
+support namespaces for sysv
+EDESC
+

_

#24 lore raw

RFC PATCH 4/6 sysvsem: containerize

In reply to Dave Hansen

Same thing as the msgs.

Take all of the global sysv sem context and put it inside of a special structure. This allows future work to give different containers completely different views into the sysv space.

---

 work-dave/ipc/sem.c  |  194 ++++++++++++++++++++++++++-------------------------
 work-dave/ipc/util.c |    2 
 work-dave/ipc/util.h |   15 ++-
 3 files changed, 114 insertions(+), 97 deletions(-)
ipc/sem.c renamed from ipc/sem.c~sysvsem-container +102 -92
diff -puN ipc/sem.c~sysvsem-container ipc/sem.c
@@ -79,17 +79,19 @@
 #include "util.h"
 
 
-#define sem_lock(id)	((struct sem_array*)ipc_lock(&sem_ids,id))
-#define sem_unlock(sma)	ipc_unlock(&(sma)->sem_perm)
-#define sem_rmid(id)	((struct sem_array*)ipc_rmid(&sem_ids,id))
-#define sem_checkid(sma, semid)	\
-	ipc_checkid(&sem_ids,&sma->sem_perm,semid)
-#define sem_buildid(id, seq) \
-	ipc_buildid(&sem_ids, id, seq)
-static struct ipc_ids sem_ids;
+#define sem_lock(context, id)		\
+	((struct sem_array*)ipc_lock(&context->ids,id))
+#define sem_unlock(context, sma)	\
+	ipc_unlock(&(sma)->sem_perm)
+#define sem_rmid(context, id)		\
+	((struct sem_array*)ipc_rmid(&context->ids,id))
+#define sem_checkid(context, sma, semid)	\
+	ipc_checkid(&context->ids,&sma->sem_perm,semid)
+#define sem_buildid(context, id, seq) \
+	ipc_buildid(&context->ids, id, seq)
 
-static int newary (key_t, int, int);
-static void freeary (struct sem_array *sma, int id);
+static int newary (struct ipc_sem_context *, key_t, int, int);
+static void freeary (struct ipc_sem_context *, struct sem_array *, int);
 #ifdef CONFIG_PROC_FS
 static int sysvipc_sem_proc_show(struct seq_file *s, void *it);
 #endif
@@ -112,15 +114,12 @@ int sem_ctls[4] = {SEMMSL, SEMMNS, SEMOP
 #define sc_semopm	(sem_ctls[2])
 #define sc_semmni	(sem_ctls[3])
 
-static int used_sems;
-
-void __init sem_init (void)
+void __init sem_init (struct ipc_sem_context *context)
 {
-	used_sems = 0;
-	ipc_init_ids(&sem_ids,sc_semmni);
+	ipc_init_ids(&context->ids,sc_semmni);
 	ipc_init_proc_interface("sysvipc/sem",
 				"       key      semid perms      nsems   uid   gid  cuid  cgid      otime      ctime\n",
-				&sem_ids,
+				&context->ids,
 				sysvipc_sem_proc_show);
 }
 
@@ -158,7 +157,8 @@ void __init sem_init (void)
  */
 #define IN_WAKEUP	1
 
-static int newary (key_t key, int nsems, int semflg)
+static int newary (struct ipc_sem_context *context,
+		   key_t key, int nsems, int semflg)
 {
 	int id;
 	int retval;
@@ -167,7 +167,7 @@ static int newary (key_t key, int nsems,
 
 	if (!nsems)
 		return -EINVAL;
-	if (used_sems + nsems > sc_semmns)
+	if (context->nr_used + nsems > sc_semmns)
 		return -ENOSPC;
 
 	size = sizeof (*sma) + nsems * sizeof (struct sem);
@@ -187,22 +187,22 @@ static int newary (key_t key, int nsems,
 		return retval;
 	}
 
-	id = ipc_addid(&sem_ids, &sma->sem_perm, sc_semmni);
+	id = ipc_addid(&context->ids, &sma->sem_perm, sc_semmni);
 	if(id == -1) {
 		security_sem_free(sma);
 		ipc_rcu_putref(sma);
 		return -ENOSPC;
 	}
-	used_sems += nsems;
+	context->nr_used += nsems;
 
-	sma->sem_id = sem_buildid(id, sma->sem_perm.seq);
+	sma->sem_id = sem_buildid(context, id, sma->sem_perm.seq);
 	sma->sem_base = (struct sem *) &sma[1];
 	/* sma->sem_pending = NULL; */
 	sma->sem_pending_last = &sma->sem_pending;
 	/* sma->undo = NULL; */
 	sma->sem_nsems = nsems;
 	sma->sem_ctime = get_seconds();
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 
 	return sma->sem_id;
 }
@@ -211,22 +211,23 @@ asmlinkage long sys_semget (key_t key, i
 {
 	int id, err = -EINVAL;
 	struct sem_array *sma;
+	struct ipc_sem_context *context = &current->ipc_context->sem;
 
 	if (nsems < 0 || nsems > sc_semmsl)
 		return -EINVAL;
-	down(&sem_ids.sem);
+	down(&context->ids.sem);
 	
 	if (key == IPC_PRIVATE) {
-		err = newary(key, nsems, semflg);
-	} else if ((id = ipc_findkey(&sem_ids, key)) == -1) {  /* key not used */
+		err = newary(context, key, nsems, semflg);
+	} else if ((id = ipc_findkey(&context->ids, key)) == -1) {  /* key not used */
 		if (!(semflg & IPC_CREAT))
 			err = -ENOENT;
 		else
-			err = newary(key, nsems, semflg);
+			err = newary(context, key, nsems, semflg);
 	} else if (semflg & IPC_CREAT && semflg & IPC_EXCL) {
 		err = -EEXIST;
 	} else {
-		sma = sem_lock(id);
+		sma = sem_lock(context, id);
 		if(sma==NULL)
 			BUG();
 		if (nsems > sma->sem_nsems)
@@ -234,15 +235,15 @@ asmlinkage long sys_semget (key_t key, i
 		else if (ipcperms(&sma->sem_perm, semflg))
 			err = -EACCES;
 		else {
-			int semid = sem_buildid(id, sma->sem_perm.seq);
+			int semid = sem_buildid(context, id, sma->sem_perm.seq);
 			err = security_sem_associate(sma, semflg);
 			if (!err)
 				err = semid;
 		}
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 	}
 
-	up(&sem_ids.sem);
+	up(&context->ids.sem);
 	return err;
 }
 
@@ -437,11 +438,12 @@ static int count_semzcnt (struct sem_arr
 	return semzcnt;
 }
 
-/* Free a semaphore set. freeary() is called with sem_ids.sem down and
- * the spinlock for this semaphore set hold. sem_ids.sem remains locked
+/* Free a semaphore set. freeary() is called with context->ids.sem down and
+ * the spinlock for this semaphore set hold. context->ids.sem remains locked
  * on exit.
  */
-static void freeary (struct sem_array *sma, int id)
+static void freeary (struct ipc_sem_context *context,
+		     struct sem_array *sma, int id)
 {
 	struct sem_undo *un;
 	struct sem_queue *q;
@@ -469,10 +471,10 @@ static void freeary (struct sem_array *s
 	}
 
 	/* Remove the semaphore set from the ID array*/
-	sma = sem_rmid(id);
-	sem_unlock(sma);
+	sma = sem_rmid(context, id);
+	sem_unlock(context, sma);
 
-	used_sems -= sma->sem_nsems;
+	context->nr_used -= sma->sem_nsems;
 	size = sizeof (*sma) + sma->sem_nsems * sizeof (struct sem);
 	security_sem_free(sma);
 	ipc_rcu_putref(sma);
@@ -500,7 +502,8 @@ static unsigned long copy_semid_to_user(
 	}
 }
 
-static int semctl_nolock(int semid, int semnum, int cmd, int version, union semun arg)
+static int semctl_nolock(struct ipc_sem_context *context,
+		         int semid, int semnum, int cmd, int version, union semun arg)
 {
 	int err = -EINVAL;
 	struct sem_array *sma;
@@ -525,16 +528,16 @@ static int semctl_nolock(int semid, int 
 		seminfo.semmnu = SEMMNU;
 		seminfo.semmap = SEMMAP;
 		seminfo.semume = SEMUME;
-		down(&sem_ids.sem);
+		down(&context->ids.sem);
 		if (cmd == SEM_INFO) {
-			seminfo.semusz = sem_ids.in_use;
-			seminfo.semaem = used_sems;
+			seminfo.semusz = context->ids.in_use;
+			seminfo.semaem = context->nr_used;
 		} else {
 			seminfo.semusz = SEMUSZ;
 			seminfo.semaem = SEMAEM;
 		}
-		max_id = sem_ids.max_id;
-		up(&sem_ids.sem);
+		max_id = context->ids.max_id;
+		up(&context->ids.sem);
 		if (copy_to_user (arg.__buf, &seminfo, sizeof(struct seminfo))) 
 			return -EFAULT;
 		return (max_id < 0) ? 0: max_id;
@@ -544,12 +547,12 @@ static int semctl_nolock(int semid, int 
 		struct semid64_ds tbuf;
 		int id;
 
-		if(semid >= sem_ids.entries->size)
+		if(semid >= context->ids.entries->size)
 			return -EINVAL;
 
 		memset(&tbuf,0,sizeof(tbuf));
 
-		sma = sem_lock(semid);
+		sma = sem_lock(context, semid);
 		if(sma == NULL)
 			return -EINVAL;
 
@@ -561,13 +564,13 @@ static int semctl_nolock(int semid, int 
 		if (err)
 			goto out_unlock;
 
-		id = sem_buildid(semid, sma->sem_perm.seq);
+		id = sem_buildid(context, semid, sma->sem_perm.seq);
 
 		kernel_to_ipc64_perm(&sma->sem_perm, &tbuf.sem_perm);
 		tbuf.sem_otime  = sma->sem_otime;
 		tbuf.sem_ctime  = sma->sem_ctime;
 		tbuf.sem_nsems  = sma->sem_nsems;
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		if (copy_semid_to_user (arg.buf, &tbuf, version))
 			return -EFAULT;
 		return id;
@@ -577,11 +580,13 @@ static int semctl_nolock(int semid, int 
 	}
 	return err;
 out_unlock:
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 	return err;
 }
 
-static int semctl_main(int semid, int semnum, int cmd, int version, union semun arg)
+static int semctl_main(struct ipc_sem_context *context,
+		       int semid, int semnum, int cmd,
+		       int version, union semun arg)
 {
 	struct sem_array *sma;
 	struct sem* curr;
@@ -590,14 +595,14 @@ static int semctl_main(int semid, int se
 	ushort* sem_io = fast_sem_io;
 	int nsems;
 
-	sma = sem_lock(semid);
+	sma = sem_lock(context, semid);
 	if(sma==NULL)
 		return -EINVAL;
 
 	nsems = sma->sem_nsems;
 
 	err=-EIDRM;
-	if (sem_checkid(sma,semid))
+	if (sem_checkid(context, sma,semid))
 		goto out_unlock;
 
 	err = -EACCES;
@@ -617,20 +622,20 @@ static int semctl_main(int semid, int se
 
 		if(nsems > SEMMSL_FAST) {
 			ipc_rcu_getref(sma);
-			sem_unlock(sma);			
+			sem_unlock(context, sma);
 
 			sem_io = ipc_alloc(sizeof(ushort)*nsems);
 			if(sem_io == NULL) {
 				ipc_lock_by_ptr(&sma->sem_perm);
 				ipc_rcu_putref(sma);
-				sem_unlock(sma);
+				sem_unlock(context, sma);
 				return -ENOMEM;
 			}
 
 			ipc_lock_by_ptr(&sma->sem_perm);
 			ipc_rcu_putref(sma);
 			if (sma->sem_perm.deleted) {
-				sem_unlock(sma);
+				sem_unlock(context, sma);
 				err = -EIDRM;
 				goto out_free;
 			}
@@ -638,7 +643,7 @@ static int semctl_main(int semid, int se
 
 		for (i = 0; i < sma->sem_nsems; i++)
 			sem_io[i] = sma->sem_base[i].semval;
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		err = 0;
 		if(copy_to_user(array, sem_io, nsems*sizeof(ushort)))
 			err = -EFAULT;
@@ -650,14 +655,14 @@ static int semctl_main(int semid, int se
 		struct sem_undo *un;
 
 		ipc_rcu_getref(sma);
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 
 		if(nsems > SEMMSL_FAST) {
 			sem_io = ipc_alloc(sizeof(ushort)*nsems);
 			if(sem_io == NULL) {
 				ipc_lock_by_ptr(&sma->sem_perm);
 				ipc_rcu_putref(sma);
-				sem_unlock(sma);
+				sem_unlock(context, sma);
 				return -ENOMEM;
 			}
 		}
@@ -665,7 +670,7 @@ static int semctl_main(int semid, int se
 		if (copy_from_user (sem_io, arg.array, nsems*sizeof(ushort))) {
 			ipc_lock_by_ptr(&sma->sem_perm);
 			ipc_rcu_putref(sma);
-			sem_unlock(sma);
+			sem_unlock(context, sma);
 			err = -EFAULT;
 			goto out_free;
 		}
@@ -674,7 +679,7 @@ static int semctl_main(int semid, int se
 			if (sem_io[i] > SEMVMX) {
 				ipc_lock_by_ptr(&sma->sem_perm);
 				ipc_rcu_putref(sma);
-				sem_unlock(sma);
+				sem_unlock(context, sma);
 				err = -ERANGE;
 				goto out_free;
 			}
@@ -682,7 +687,7 @@ static int semctl_main(int semid, int se
 		ipc_lock_by_ptr(&sma->sem_perm);
 		ipc_rcu_putref(sma);
 		if (sma->sem_perm.deleted) {
-			sem_unlock(sma);
+			sem_unlock(context, sma);
 			err = -EIDRM;
 			goto out_free;
 		}
@@ -706,7 +711,7 @@ static int semctl_main(int semid, int se
 		tbuf.sem_otime  = sma->sem_otime;
 		tbuf.sem_ctime  = sma->sem_ctime;
 		tbuf.sem_nsems  = sma->sem_nsems;
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		if (copy_semid_to_user (arg.buf, &tbuf, version))
 			return -EFAULT;
 		return 0;
@@ -752,7 +757,7 @@ static int semctl_main(int semid, int se
 	}
 	}
 out_unlock:
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 out_free:
 	if(sem_io != fast_sem_io)
 		ipc_free(sem_io, sizeof(ushort)*nsems);
@@ -799,7 +804,9 @@ static inline unsigned long copy_semid_f
 	}
 }
 
-static int semctl_down(int semid, int semnum, int cmd, int version, union semun arg)
+static int semctl_down(struct ipc_sem_context *context,
+		       int semid, int semnum, int cmd,
+		       int version, union semun arg)
 {
 	struct sem_array *sma;
 	int err;
@@ -812,11 +819,11 @@ static int semctl_down(int semid, int se
 		if ((err = audit_ipc_perms(0, setbuf.uid, setbuf.gid, setbuf.mode)))
 			return err;
 	}
-	sma = sem_lock(semid);
+	sma = sem_lock(context, semid);
 	if(sma==NULL)
 		return -EINVAL;
 
-	if (sem_checkid(sma,semid)) {
+	if (sem_checkid(context, sma,semid)) {
 		err=-EIDRM;
 		goto out_unlock;
 	}	
@@ -834,7 +841,7 @@ static int semctl_down(int semid, int se
 
 	switch(cmd){
 	case IPC_RMID:
-		freeary(sma, semid);
+		freeary(context, sma, semid);
 		err = 0;
 		break;
 	case IPC_SET:
@@ -843,18 +850,18 @@ static int semctl_down(int semid, int se
 		ipcp->mode = (ipcp->mode & ~S_IRWXUGO)
 				| (setbuf.mode & S_IRWXUGO);
 		sma->sem_ctime = get_seconds();
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		err = 0;
 		break;
 	default:
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		err = -EINVAL;
 		break;
 	}
 	return err;
 
 out_unlock:
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 	return err;
 }
 
@@ -862,6 +869,7 @@ asmlinkage long sys_semctl (int semid, i
 {
 	int err = -EINVAL;
 	int version;
+	struct ipc_sem_context *context = &current->ipc_context->sem;
 
 	if (semid < 0)
 		return -EINVAL;
@@ -872,7 +880,7 @@ asmlinkage long sys_semctl (int semid, i
 	case IPC_INFO:
 	case SEM_INFO:
 	case SEM_STAT:
-		err = semctl_nolock(semid,semnum,cmd,version,arg);
+		err = semctl_nolock(context, semid,semnum,cmd,version,arg);
 		return err;
 	case GETALL:
 	case GETVAL:
@@ -882,13 +890,13 @@ asmlinkage long sys_semctl (int semid, i
 	case IPC_STAT:
 	case SETVAL:
 	case SETALL:
-		err = semctl_main(semid,semnum,cmd,version,arg);
+		err = semctl_main(context, semid,semnum,cmd,version,arg);
 		return err;
 	case IPC_RMID:
 	case IPC_SET:
-		down(&sem_ids.sem);
-		err = semctl_down(semid,semnum,cmd,version,arg);
-		up(&sem_ids.sem);
+		down(&context->ids.sem);
+		err = semctl_down(context, semid,semnum,cmd,version,arg);
+		up(&context->ids.sem);
 		return err;
 	default:
 		return -EINVAL;
@@ -976,7 +984,7 @@ static struct sem_undo *lookup_undo(stru
 	return un;
 }
 
-static struct sem_undo *find_undo(int semid)
+static struct sem_undo *find_undo(struct ipc_sem_context *context, int semid)
 {
 	struct sem_array *sma;
 	struct sem_undo_list *ulp;
@@ -995,24 +1003,24 @@ static struct sem_undo *find_undo(int se
 		goto out;
 
 	/* no undo structure around - allocate one. */
-	sma = sem_lock(semid);
+	sma = sem_lock(context, semid);
 	un = ERR_PTR(-EINVAL);
 	if(sma==NULL)
 		goto out;
 	un = ERR_PTR(-EIDRM);
-	if (sem_checkid(sma,semid)) {
-		sem_unlock(sma);
+	if (sem_checkid(context, sma,semid)) {
+		sem_unlock(context, sma);
 		goto out;
 	}
 	nsems = sma->sem_nsems;
 	ipc_rcu_getref(sma);
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 
 	new = (struct sem_undo *) kmalloc(sizeof(struct sem_undo) + sizeof(short)*nsems, GFP_KERNEL);
 	if (!new) {
 		ipc_lock_by_ptr(&sma->sem_perm);
 		ipc_rcu_putref(sma);
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		return ERR_PTR(-ENOMEM);
 	}
 	memset(new, 0, sizeof(struct sem_undo) + sizeof(short)*nsems);
@@ -1026,13 +1034,13 @@ static struct sem_undo *find_undo(int se
 		kfree(new);
 		ipc_lock_by_ptr(&sma->sem_perm);
 		ipc_rcu_putref(sma);
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		goto out;
 	}
 	ipc_lock_by_ptr(&sma->sem_perm);
 	ipc_rcu_putref(sma);
 	if (sma->sem_perm.deleted) {
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		unlock_semundo();
 		kfree(new);
 		un = ERR_PTR(-EIDRM);
@@ -1042,7 +1050,7 @@ static struct sem_undo *find_undo(int se
 	ulp->proc_list = new;
 	new->id_next = sma->undo;
 	sma->undo = new;
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 	un = new;
 	unlock_semundo();
 out:
@@ -1052,6 +1060,7 @@ out:
 asmlinkage long sys_semtimedop(int semid, struct sembuf __user *tsops,
 			unsigned nsops, const struct timespec __user *timeout)
 {
+	struct ipc_sem_context *context = &current->ipc_context->sem;
 	int error = -EINVAL;
 	struct sem_array *sma;
 	struct sembuf fast_sops[SEMOPM_FAST];
@@ -1099,7 +1108,7 @@ asmlinkage long sys_semtimedop(int semid
 
 retry_undos:
 	if (undos) {
-		un = find_undo(semid);
+		un = find_undo(context, semid);
 		if (IS_ERR(un)) {
 			error = PTR_ERR(un);
 			goto out_free;
@@ -1107,12 +1116,12 @@ retry_undos:
 	} else
 		un = NULL;
 
-	sma = sem_lock(semid);
+	sma = sem_lock(context, semid);
 	error=-EINVAL;
 	if(sma==NULL)
 		goto out_free;
 	error = -EIDRM;
-	if (sem_checkid(sma,semid))
+	if (sem_checkid(context, sma,semid))
 		goto out_unlock_free;
 	/*
 	 * semid identifies are not unique - find_undo may have
@@ -1120,7 +1129,7 @@ retry_undos:
 	 * and now a new array with received the same id. Check and retry.
 	 */
 	if (un && un->semid == -1) {
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 		goto retry_undos;
 	}
 	error = -EFBIG;
@@ -1161,7 +1170,7 @@ retry_undos:
 	queue.status = -EINTR;
 	queue.sleeper = current;
 	current->state = TASK_INTERRUPTIBLE;
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 
 	if (timeout)
 		jiffies_left = schedule_timeout(jiffies_left);
@@ -1180,7 +1189,7 @@ retry_undos:
 		goto out_free;
 	}
 
-	sma = sem_lock(semid);
+	sma = sem_lock(context, semid);
 	if(sma==NULL) {
 		if(queue.prev != NULL)
 			BUG();
@@ -1205,7 +1214,7 @@ retry_undos:
 	goto out_unlock_free;
 
 out_unlock_free:
-	sem_unlock(sma);
+	sem_unlock(context, sma);
 out_free:
 	if(sops != fast_sops)
 		kfree(sops);
@@ -1258,6 +1267,7 @@ void exit_sem(struct task_struct *tsk)
 {
 	struct sem_undo_list *undo_list;
 	struct sem_undo *u, **up;
+	struct ipc_sem_context *context = &tsk->ipc_context->sem;
 
 	undo_list = tsk->sysvsem.undo_list;
 	if (!undo_list)
@@ -1279,14 +1289,14 @@ void exit_sem(struct task_struct *tsk)
 
 		if(semid == -1)
 			continue;
-		sma = sem_lock(semid);
+		sma = sem_lock(context, semid);
 		if (sma == NULL)
 			continue;
 
 		if (u->semid == -1)
 			goto next_entry;
 
-		BUG_ON(sem_checkid(sma,u->semid));
+		BUG_ON(sem_checkid(context, sma,u->semid));
 
 		/* remove u from the sma->undo list */
 		for (unp = &sma->undo; (un = *unp); unp = &un->id_next) {
@@ -1327,7 +1337,7 @@ found:
 		/* maybe some queued-up processes were waiting for this */
 		update_queue(sma);
 next_entry:
-		sem_unlock(sma);
+		sem_unlock(context, sma);
 	}
 	kfree(undo_list);
 }
ipc/util.c renamed from ipc/util.c~sysvsem-container +1 -1
diff -puN ipc/util.c~sysvsem-container ipc/util.c
@@ -52,7 +52,7 @@ static int __init ipc_init(void)
 	if (!init_task.ipc_context)
 		return -ENOMEM;
 
-	sem_init();
+	sem_init(&init_task.ipc_context->sem);
 	msg_init(&init_task.ipc_context->msg);
 	shm_init();
 	return 0;
ipc/util.h renamed from ipc/util.h~sysvsem-container +11 -4
diff -puN ipc/util.h~sysvsem-container ipc/util.h
@@ -11,10 +11,6 @@
 #define USHRT_MAX 0xffff
 #define SEQ_MULTIPLIER	(IPCMNI)
 
-void sem_init (void);
-void msg_init (struct ipc_msg_context *context);
-void shm_init (void);
-
 struct ipc_id_ary {
 	int size;
 	struct kern_ipc_perm *p[0];
@@ -42,10 +38,21 @@ struct ipc_msg_context {
 	struct kref count;
 };
 
+struct ipc_sem_context {
+	struct ipc_ids ids;
+
+	int nr_used;
+};
+
 struct ipc_context {
 	struct ipc_msg_context msg;
+	struct ipc_sem_context sem;
 };
 
+void sem_init (struct ipc_sem_context *context);
+void msg_init (struct ipc_msg_context *context);
+void shm_init (void);
+
 struct seq_file;
 void __init ipc_init_ids(struct ipc_ids* ids, int size);
 #ifdef CONFIG_PROC_FS

_

#0 lore raw

cover support separate namespaces for sysv

This is more followup on this thread:

http://marc.theaimsgroup.com/?l=linux-kernel&m=114108773827517&w=2

The following series makes it possible to isolate different sets of sysv structures (msg, sem, and shm) on one kernel.

The initial patch makes it much more feasible to access these disparite sets from the traditional sysctl interfaces.

#28 lore raw

RFC PATCH 6/6 sysvshm: containerize sysctls

Note that this will effectively remove the system-wide limits on sysv shm resources that the sysctl mechanism currently provides and move it to a per-container limit. This is currently by design, and may be later addressed when we have proper large-scale resource controls in the kernel.

---

 desc.txt                  |    0 
 work-dave/ipc/shm.c       |   33 +++++++++++++++++++++++----------
 work-dave/ipc/util.h      |    4 ++++
 work-dave/kernel/sysctl.c |   12 ++++++------
 4 files changed, 33 insertions(+), 16 deletions(-)
ipc/shm.c renamed from ipc/shm.c~sysvshm-containerize-sysctls +23 -10
diff -puN ipc/shm.c~sysvshm-containerize-sysctls ipc/shm.c
@@ -55,12 +55,12 @@ static void shm_close (struct vm_area_st
 static int sysvipc_shm_proc_show(struct seq_file *s, void *it);
 #endif
 
-size_t	shm_ctlmax = SHMMAX;
-size_t 	shm_ctlall = SHMALL;
-int 	shm_ctlmni = SHMMNI;
-
 void __init shm_init (struct ipc_shm_context *context)
 {
+	context->ctlmax = SHMMAX;
+	context->ctlall = SHMALL;
+	context->ctlmni = SHMMNI;
+
 	ipc_init_ids(&context->ids, 1);
 	ipc_init_proc_interface("sysvipc/shm",
 				"       key      shmid perms       size  cpid  lpid nattch   uid   gid  cuid  cgid      atime      dtime      ctime\n",
@@ -85,7 +85,7 @@ struct shmid_kernel *shm_rmid(struct ipc
 static inline
 int shm_addid(struct ipc_shm_context *context, struct shmid_kernel *shp)
 {
-	return ipc_addid(&context->ids, &shp->shm_perm, shm_ctlmni);
+	return ipc_addid(&context->ids, &shp->shm_perm, context->ctlmni);
 }
 
 
@@ -208,10 +208,10 @@ static int newseg (struct ipc_shm_contex
 	char name[13];
 	int id;
 
-	if (size < SHMMIN || size > shm_ctlmax)
+	if (size < SHMMIN || size > context->ctlmax)
 		return -EINVAL;
 
-	if (context->tot + numpages >= shm_ctlall)
+	if (context->tot + numpages >= context->ctlall)
 		return -ENOSPC;
 
 	shp = ipc_rcu_alloc(sizeof(*shp));
@@ -463,9 +463,9 @@ asmlinkage long sys_shmctl (int shmid, i
 			return err;
 
 		memset(&shminfo,0,sizeof(shminfo));
-		shminfo.shmmni = shminfo.shmseg = shm_ctlmni;
-		shminfo.shmmax = shm_ctlmax;
-		shminfo.shmall = shm_ctlall;
+		shminfo.shmmni = shminfo.shmseg = context->ctlmni;
+		shminfo.shmmax = context->ctlmax;
+		shminfo.shmall = context->ctlall;
 
 		shminfo.shmmin = SHMMIN;
 		if(copy_shminfo_to_user (buf, &shminfo, version))
@@ -935,3 +935,16 @@ static int sysvipc_shm_proc_show(struct 
 			  shp->shm_ctim);
 }
 #endif
+
+void *shm_ctlmax_helper(void)
+{
+	return &current->ipc_context->shm.ctlmax;
+}
+void *shm_ctlall_helper(void)
+{
+	return &current->ipc_context->shm.ctlall;
+}
+void *shm_ctlmni_helper(void)
+{
+	return &current->ipc_context->shm.ctlmni;
+}
ipc/util.h renamed from ipc/util.h~sysvshm-containerize-sysctls +4 -0
diff -puN ipc/util.h~sysvshm-containerize-sysctls ipc/util.h
@@ -47,6 +47,10 @@ struct ipc_sem_context {
 struct ipc_shm_context {
 	struct ipc_ids ids;
 
+	size_t ctlmax;
+	size_t ctlall;
+	int    ctlmni;
+
 	int tot; /* total number of shared memory pages */
 };
 
kernel/sysctl.c renamed from kernel/sysctl.c~sysvshm-containerize-sysctls +6 -6
diff -puN kernel/sysctl.c~sysvshm-containerize-sysctls kernel/sysctl.c
@@ -92,12 +92,12 @@ extern char modprobe_path[];
 extern int sg_big_buff;
 #endif
 #ifdef CONFIG_SYSVIPC
-extern size_t shm_ctlmax;
-extern size_t shm_ctlall;
+extern void *shm_ctlmax_helper(void);
+extern void *shm_ctlall_helper(void);
+extern void *shm_ctlmni_helper(void);
 extern void *msg_ctlmnb_helper(void);
 extern void *msg_ctlmni_helper(void);
 extern void *msg_ctlmax_helper(void);
-extern int shm_ctlmni;
 extern int sem_ctls[];
 #endif
 
@@ -428,7 +428,7 @@ static ctl_table kern_table[] = {
 	{
 		.ctl_name	= KERN_SHMMAX,
 		.procname	= "shmmax",
-		.data		= &shm_ctlmax,
+		.data_access	= &shm_ctlmax_helper,
 		.maxlen		= sizeof (size_t),
 		.mode		= 0644,
 		.proc_handler	= &proc_doulongvec_minmax,
@@ -436,7 +436,7 @@ static ctl_table kern_table[] = {
 	{
 		.ctl_name	= KERN_SHMALL,
 		.procname	= "shmall",
-		.data		= &shm_ctlall,
+		.data_access	= &shm_ctlall_helper,
 		.maxlen		= sizeof (size_t),
 		.mode		= 0644,
 		.proc_handler	= &proc_doulongvec_minmax,
@@ -444,7 +444,7 @@ static ctl_table kern_table[] = {
 	{
 		.ctl_name	= KERN_SHMMNI,
 		.procname	= "shmmni",
-		.data		= &shm_ctlmni,
+		.data_access	= &shm_ctlmni_helper,
 		.maxlen		= sizeof (int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec,

diff -puN desc.txt~sysvshm-containerize-sysctls desc.txt _

#27 lore raw

RFC PATCH 5/6 sysvshm: containerize

In reply to Dave Hansen

Same thing as sems and msgs.

Take all of the global sysv shm context and put it inside of a special structure. This allows future work to give different containers completely different views into the sysv space.

---

 work-dave/ipc/shm.c  |  197 +++++++++++++++++++++++++++------------------------
 work-dave/ipc/util.c |    3 
 work-dave/ipc/util.h |    9 ++
 3 files changed, 118 insertions(+), 91 deletions(-)

diff -puN include/linux/sched.h~sysvshm-container include/linux/sched.h

ipc/shm.c renamed from ipc/shm.c~sysvshm-container +108 -89
diff -puN ipc/shm.c~sysvshm-container ipc/shm.c
@@ -38,15 +38,17 @@
 static struct file_operations shm_file_operations;
 static struct vm_operations_struct shm_vm_ops;
 
-static struct ipc_ids shm_ids;
+#define shm_lock(context, id)		\
+	((struct shmid_kernel*)ipc_lock(&context->ids,id))
+#define shm_unlock(context, shp)	\
+	ipc_unlock(&(shp)->shm_perm)
+#define shm_get(context, id)\
+	((struct shmid_kernel*)ipc_get(&context->ids,id))
+#define shm_buildid(context, id, seq) \
+	ipc_buildid(&context->ids, id, seq)
 
-#define shm_lock(id)	((struct shmid_kernel*)ipc_lock(&shm_ids,id))
-#define shm_unlock(shp)	ipc_unlock(&(shp)->shm_perm)
-#define shm_get(id)	((struct shmid_kernel*)ipc_get(&shm_ids,id))
-#define shm_buildid(id, seq) \
-	ipc_buildid(&shm_ids, id, seq)
-
-static int newseg (key_t key, int shmflg, size_t size);
+static int newseg (struct ipc_shm_context *context,
+		   key_t key, int shmflg, size_t size);
 static void shm_open (struct vm_area_struct *shmd);
 static void shm_close (struct vm_area_struct *shmd);
 #ifdef CONFIG_PROC_FS
@@ -57,51 +59,53 @@ size_t	shm_ctlmax = SHMMAX;
 size_t 	shm_ctlall = SHMALL;
 int 	shm_ctlmni = SHMMNI;
 
-static int shm_tot; /* total number of shared memory pages */
-
-void __init shm_init (void)
+void __init shm_init (struct ipc_shm_context *context)
 {
-	ipc_init_ids(&shm_ids, 1);
+	ipc_init_ids(&context->ids, 1);
 	ipc_init_proc_interface("sysvipc/shm",
 				"       key      shmid perms       size  cpid  lpid nattch   uid   gid  cuid  cgid      atime      dtime      ctime\n",
-				&shm_ids,
+				&context->ids,
 				sysvipc_shm_proc_show);
 }
 
-static inline int shm_checkid(struct shmid_kernel *s, int id)
+static inline int shm_checkid(struct ipc_shm_context *context,
+			      struct shmid_kernel *s, int id)
 {
-	if (ipc_checkid(&shm_ids,&s->shm_perm,id))
+	if (ipc_checkid(&context->ids,&s->shm_perm,id))
 		return -EIDRM;
 	return 0;
 }
 
-static inline struct shmid_kernel *shm_rmid(int id)
+static inline
+struct shmid_kernel *shm_rmid(struct ipc_shm_context *context, int id)
 {
-	return (struct shmid_kernel *)ipc_rmid(&shm_ids,id);
+	return (struct shmid_kernel *)ipc_rmid(&context->ids,id);
 }
 
-static inline int shm_addid(struct shmid_kernel *shp)
+static inline
+int shm_addid(struct ipc_shm_context *context, struct shmid_kernel *shp)
 {
-	return ipc_addid(&shm_ids, &shp->shm_perm, shm_ctlmni);
+	return ipc_addid(&context->ids, &shp->shm_perm, shm_ctlmni);
 }
 
 
 
-static inline void shm_inc (int id) {
+static inline void shm_inc (struct ipc_shm_context *context, int id) {
 	struct shmid_kernel *shp;
 
-	if(!(shp = shm_lock(id)))
+	if(!(shp = shm_lock(context, id)))
 		BUG();
 	shp->shm_atim = get_seconds();
 	shp->shm_lprid = current->tgid;
 	shp->shm_nattch++;
-	shm_unlock(shp);
+	shm_unlock(context, shp);
 }
 
 /* This is called by fork, once for every shm attach. */
 static void shm_open (struct vm_area_struct *shmd)
 {
-	shm_inc (shmd->vm_file->f_dentry->d_inode->i_ino);
+	struct ipc_shm_context *context = shmd->vm_private_data;
+	shm_inc (context, shmd->vm_file->f_dentry->d_inode->i_ino);
 }
 
 /*
@@ -109,14 +113,15 @@ static void shm_open (struct vm_area_str
  *
  * @shp: struct to free
  *
- * It has to be called with shp and shm_ids.sem locked,
+ * It has to be called with shp and context->ids.sem locked,
  * but returns with shp unlocked and freed.
  */
-static void shm_destroy (struct shmid_kernel *shp)
+static void shm_destroy (struct ipc_shm_context *context,
+			 struct shmid_kernel *shp)
 {
-	shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
-	shm_rmid (shp->id);
-	shm_unlock(shp);
+	context->tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	shm_rmid (context, shp->id);
+	shm_unlock(context, shp);
 	if (!is_file_hugepages(shp->shm_file))
 		shmem_lock(shp->shm_file, 0, shp->mlock_user);
 	else
@@ -138,30 +143,39 @@ static void shm_close (struct vm_area_st
 	struct file * file = shmd->vm_file;
 	int id = file->f_dentry->d_inode->i_ino;
 	struct shmid_kernel *shp;
+	struct ipc_shm_context *context = shmd->vm_private_data;
 
-	down (&shm_ids.sem);
+	down (&context->ids.sem);
 	/* remove from the list of attaches of the shm segment */
-	if(!(shp = shm_lock(id)))
+	if(!(shp = shm_lock(context, id)))
 		BUG();
 	shp->shm_lprid = current->tgid;
 	shp->shm_dtim = get_seconds();
 	shp->shm_nattch--;
 	if(shp->shm_nattch == 0 &&
 	   shp->shm_perm.mode & SHM_DEST)
-		shm_destroy (shp);
+		shm_destroy (context, shp);
 	else
-		shm_unlock(shp);
-	up (&shm_ids.sem);
+		shm_unlock(context, shp);
+	up (&context->ids.sem);
 }
 
 static int shm_mmap(struct file * file, struct vm_area_struct * vma)
 {
 	int ret;
+	struct ipc_shm_context *context = current->ipc_shm_context;
 
 	ret = shmem_mmap(file, vma);
 	if (ret == 0) {
 		vma->vm_ops = &shm_vm_ops;
-		shm_inc(file->f_dentry->d_inode->i_ino);
+		/*
+		 * This implies that the shm context can not change
+		 * while a task is running because this mmap's
+		 * context would change underneath it.
+		 * -- You many only change shm context at exec()
+		 */
+		vma->vm_private_data = context;
+		shm_inc(context, file->f_dentry->d_inode->i_ino);
 	}
 
 	return ret;
@@ -184,7 +198,8 @@ static struct vm_operations_struct shm_v
 #endif
 };
 
-static int newseg (key_t key, int shmflg, size_t size)
+static int newseg (struct ipc_shm_context *context,
+		   key_t key, int shmflg, size_t size)
 {
 	int error;
 	struct shmid_kernel *shp;
@@ -196,7 +211,7 @@ static int newseg (key_t key, int shmflg
 	if (size < SHMMIN || size > shm_ctlmax)
 		return -EINVAL;
 
-	if (shm_tot + numpages >= shm_ctlall)
+	if (context->tot + numpages >= shm_ctlall)
 		return -ENOSPC;
 
 	shp = ipc_rcu_alloc(sizeof(*shp));
@@ -235,7 +250,7 @@ static int newseg (key_t key, int shmflg
 		goto no_file;
 
 	error = -ENOSPC;
-	id = shm_addid(shp);
+	id = shm_addid(context, shp);
 	if(id == -1) 
 		goto no_id;
 
@@ -245,7 +260,7 @@ static int newseg (key_t key, int shmflg
 	shp->shm_ctim = get_seconds();
 	shp->shm_segsz = size;
 	shp->shm_nattch = 0;
-	shp->id = shm_buildid(id,shp->shm_perm.seq);
+	shp->id = shm_buildid(context, id,shp->shm_perm.seq);
 	shp->shm_file = file;
 	file->f_dentry->d_inode->i_ino = shp->id;
 
@@ -253,8 +268,8 @@ static int newseg (key_t key, int shmflg
 	if (!(shmflg & SHM_HUGETLB))
 		file->f_op = &shm_file_operations;
 
-	shm_tot += numpages;
-	shm_unlock(shp);
+	context->tot += numpages;
+	shm_unlock(context, shp);
 	return shp->id;
 
 no_id:
@@ -269,19 +284,20 @@ asmlinkage long sys_shmget (key_t key, s
 {
 	struct shmid_kernel *shp;
 	int err, id = 0;
+	struct ipc_shm_context *context = current->ipc_shm_context;
 
-	down(&shm_ids.sem);
+	down(&context->ids.sem);
 	if (key == IPC_PRIVATE) {
-		err = newseg(key, shmflg, size);
-	} else if ((id = ipc_findkey(&shm_ids, key)) == -1) {
+		err = newseg(context, key, shmflg, size);
+	} else if ((id = ipc_findkey(&context->ids, key)) == -1) {
 		if (!(shmflg & IPC_CREAT))
 			err = -ENOENT;
 		else
-			err = newseg(key, shmflg, size);
+			err = newseg(context, key, shmflg, size);
 	} else if ((shmflg & IPC_CREAT) && (shmflg & IPC_EXCL)) {
 		err = -EEXIST;
 	} else {
-		shp = shm_lock(id);
+		shp = shm_lock(context, id);
 		if(shp==NULL)
 			BUG();
 		if (shp->shm_segsz < size)
@@ -289,14 +305,14 @@ asmlinkage long sys_shmget (key_t key, s
 		else if (ipcperms(&shp->shm_perm, shmflg))
 			err = -EACCES;
 		else {
-			int shmid = shm_buildid(id, shp->shm_perm.seq);
+			int shmid = shm_buildid(context, id, shp->shm_perm.seq);
 			err = security_shm_associate(shp, shmflg);
 			if (!err)
 				err = shmid;
 		}
-		shm_unlock(shp);
+		shm_unlock(context, shp);
 	}
-	up(&shm_ids.sem);
+	up(&context->ids.sem);
 
 	return err;
 }
@@ -392,18 +408,19 @@ static inline unsigned long copy_shminfo
 	}
 }
 
-static void shm_get_stat(unsigned long *rss, unsigned long *swp) 
+static void shm_get_stat(struct ipc_shm_context *context,
+			 unsigned long *rss, unsigned long *swp)
 {
 	int i;
 
 	*rss = 0;
 	*swp = 0;
 
-	for (i = 0; i <= shm_ids.max_id; i++) {
+	for (i = 0; i <= context->ids.max_id; i++) {
 		struct shmid_kernel *shp;
 		struct inode *inode;
 
-		shp = shm_get(i);
+		shp = shm_get(context, i);
 		if(!shp)
 			continue;
 
@@ -427,6 +444,7 @@ asmlinkage long sys_shmctl (int shmid, i
 	struct shm_setbuf setbuf;
 	struct shmid_kernel *shp;
 	int err, version;
+	struct ipc_shm_context *context = current->ipc_shm_context;
 
 	if (cmd < 0 || shmid < 0) {
 		err = -EINVAL;
@@ -453,7 +471,7 @@ asmlinkage long sys_shmctl (int shmid, i
 		if(copy_shminfo_to_user (buf, &shminfo, version))
 			return -EFAULT;
 		/* reading a integer is always atomic */
-		err= shm_ids.max_id;
+		err= context->ids.max_id;
 		if(err<0)
 			err = 0;
 		goto out;
@@ -467,14 +485,14 @@ asmlinkage long sys_shmctl (int shmid, i
 			return err;
 
 		memset(&shm_info,0,sizeof(shm_info));
-		down(&shm_ids.sem);
-		shm_info.used_ids = shm_ids.in_use;
-		shm_get_stat (&shm_info.shm_rss, &shm_info.shm_swp);
-		shm_info.shm_tot = shm_tot;
+		down(&context->ids.sem);
+		shm_info.used_ids = context->ids.in_use;
+		shm_get_stat (context, &shm_info.shm_rss, &shm_info.shm_swp);
+		shm_info.shm_tot = context->tot;
 		shm_info.swap_attempts = 0;
 		shm_info.swap_successes = 0;
-		err = shm_ids.max_id;
-		up(&shm_ids.sem);
+		err = context->ids.max_id;
+		up(&context->ids.sem);
 		if(copy_to_user (buf, &shm_info, sizeof(shm_info))) {
 			err = -EFAULT;
 			goto out;
@@ -489,17 +507,17 @@ asmlinkage long sys_shmctl (int shmid, i
 		struct shmid64_ds tbuf;
 		int result;
 		memset(&tbuf, 0, sizeof(tbuf));
-		shp = shm_lock(shmid);
+		shp = shm_lock(context, shmid);
 		if(shp==NULL) {
 			err = -EINVAL;
 			goto out;
 		} else if(cmd==SHM_STAT) {
 			err = -EINVAL;
-			if (shmid > shm_ids.max_id)
+			if (shmid > context->ids.max_id)
 				goto out_unlock;
-			result = shm_buildid(shmid, shp->shm_perm.seq);
+			result = shm_buildid(context, shmid, shp->shm_perm.seq);
 		} else {
-			err = shm_checkid(shp,shmid);
+			err = shm_checkid(context,shp,shmid);
 			if(err)
 				goto out_unlock;
 			result = 0;
@@ -521,7 +539,7 @@ asmlinkage long sys_shmctl (int shmid, i
 			tbuf.shm_nattch	= shp->shm_nattch;
 		else
 			tbuf.shm_nattch = file_count(shp->shm_file) - 1;
-		shm_unlock(shp);
+		shm_unlock(context, shp);
 		if(copy_shmid_to_user (buf, &tbuf, version))
 			err = -EFAULT;
 		else
@@ -531,12 +549,12 @@ asmlinkage long sys_shmctl (int shmid, i
 	case SHM_LOCK:
 	case SHM_UNLOCK:
 	{
-		shp = shm_lock(shmid);
+		shp = shm_lock(context, shmid);
 		if(shp==NULL) {
 			err = -EINVAL;
 			goto out;
 		}
-		err = shm_checkid(shp,shmid);
+		err = shm_checkid(context,shp,shmid);
 		if(err)
 			goto out_unlock;
 
@@ -568,7 +586,7 @@ asmlinkage long sys_shmctl (int shmid, i
 			shp->shm_perm.mode &= ~SHM_LOCKED;
 			shp->mlock_user = NULL;
 		}
-		shm_unlock(shp);
+		shm_unlock(context, shp);
 		goto out;
 	}
 	case IPC_RMID:
@@ -583,12 +601,12 @@ asmlinkage long sys_shmctl (int shmid, i
 		 *	Instead we set a destroyed flag, and then blow
 		 *	the name away when the usage hits zero.
 		 */
-		down(&shm_ids.sem);
-		shp = shm_lock(shmid);
+		down(&context->ids.sem);
+		shp = shm_lock(context, shmid);
 		err = -EINVAL;
 		if (shp == NULL) 
 			goto out_up;
-		err = shm_checkid(shp, shmid);
+		err = shm_checkid(context, shp, shmid);
 		if(err)
 			goto out_unlock_up;
 
@@ -607,10 +625,10 @@ asmlinkage long sys_shmctl (int shmid, i
 			shp->shm_perm.mode |= SHM_DEST;
 			/* Do not find it any more */
 			shp->shm_perm.key = IPC_PRIVATE;
-			shm_unlock(shp);
+			shm_unlock(context, shp);
 		} else
-			shm_destroy (shp);
-		up(&shm_ids.sem);
+			shm_destroy (context, shp);
+		up(&context->ids.sem);
 		goto out;
 	}
 
@@ -622,12 +640,12 @@ asmlinkage long sys_shmctl (int shmid, i
 		}
 		if ((err = audit_ipc_perms(0, setbuf.uid, setbuf.gid, setbuf.mode)))
 			return err;
-		down(&shm_ids.sem);
-		shp = shm_lock(shmid);
+		down(&context->ids.sem);
+		shp = shm_lock(context, shmid);
 		err=-EINVAL;
 		if(shp==NULL)
 			goto out_up;
-		err = shm_checkid(shp,shmid);
+		err = shm_checkid(context,shp,shmid);
 		if(err)
 			goto out_unlock_up;
 		err=-EPERM;
@@ -656,12 +674,12 @@ asmlinkage long sys_shmctl (int shmid, i
 
 	err = 0;
 out_unlock_up:
-	shm_unlock(shp);
+	shm_unlock(context, shp);
 out_up:
-	up(&shm_ids.sem);
+	up(&context->ids.sem);
 	goto out;
 out_unlock:
-	shm_unlock(shp);
+	shm_unlock(context, shp);
 out:
 	return err;
 }
@@ -675,6 +693,7 @@ out:
  */
 long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr)
 {
+	struct ipc_shm_context *context = current->ipc_shm_context;
 	struct shmid_kernel *shp;
 	unsigned long addr;
 	unsigned long size;
@@ -725,32 +744,32 @@ long do_shmat(int shmid, char __user *sh
 	 * We cannot rely on the fs check since SYSV IPC does have an
 	 * additional creator id...
 	 */
-	shp = shm_lock(shmid);
+	shp = shm_lock(context, shmid);
 	if(shp == NULL) {
 		err = -EINVAL;
 		goto out;
 	}
-	err = shm_checkid(shp,shmid);
+	err = shm_checkid(context,shp,shmid);
 	if (err) {
-		shm_unlock(shp);
+		shm_unlock(context, shp);
 		goto out;
 	}
 	if (ipcperms(&shp->shm_perm, acc_mode)) {
-		shm_unlock(shp);
+		shm_unlock(context, shp);
 		err = -EACCES;
 		goto out;
 	}
 
 	err = security_shm_shmat(shp, shmaddr, shmflg);
 	if (err) {
-		shm_unlock(shp);
+		shm_unlock(context, shp);
 		return err;
 	}
 		
 	file = shp->shm_file;
 	size = i_size_read(file->f_dentry->d_inode);
 	shp->shm_nattch++;
-	shm_unlock(shp);
+	shm_unlock(context, shp);
 
 	down_write(&current->mm->mmap_sem);
 	if (addr && !(shmflg & SHM_REMAP)) {
@@ -771,16 +790,16 @@ long do_shmat(int shmid, char __user *sh
 invalid:
 	up_write(&current->mm->mmap_sem);
 
-	down (&shm_ids.sem);
-	if(!(shp = shm_lock(shmid)))
+	down (&context->ids.sem);
+	if(!(shp = shm_lock(context, shmid)))
 		BUG();
 	shp->shm_nattch--;
 	if(shp->shm_nattch == 0 &&
 	   shp->shm_perm.mode & SHM_DEST)
-		shm_destroy (shp);
+		shm_destroy (context, shp);
 	else
-		shm_unlock(shp);
-	up (&shm_ids.sem);
+		shm_unlock(context, shp);
+	up (&context->ids.sem);
 
 	*raddr = (unsigned long) user_addr;
 	err = 0;
ipc/util.c renamed from ipc/util.c~sysvshm-container +2 -1
diff -puN ipc/util.c~sysvshm-container ipc/util.c
@@ -46,6 +46,7 @@ struct ipc_proc_iface {
  *	memory are initialised
  */
 
+struct ipc_shm_context *ipc_shm_context;
 static int __init ipc_init(void)
 {
 	init_task.ipc_context = kzalloc(sizeof(*ipc_context), GFP_KERNEL);
@@ -54,7 +55,7 @@ static int __init ipc_init(void)
 
 	sem_init(&init_task.ipc_context->sem);
 	msg_init(&init_task.ipc_context->msg);
-	shm_init();
+	shm_init(&init_task.ipc_context->shm);
 	return 0;
 }
 __initcall(ipc_init);
ipc/util.h renamed from ipc/util.h~sysvshm-container +8 -1
diff -puN ipc/util.h~sysvshm-container ipc/util.h
@@ -44,14 +44,21 @@ struct ipc_sem_context {
 	int nr_used;
 };
 
+struct ipc_shm_context {
+	struct ipc_ids ids;
+
+	int tot; /* total number of shared memory pages */
+};
+
 struct ipc_context {
 	struct ipc_msg_context msg;
 	struct ipc_sem_context sem;
+	struct ipc_shm_context shm;
 };
 
 void sem_init (struct ipc_sem_context *context);
 void msg_init (struct ipc_msg_context *context);
-void shm_init (void);
+void shm_init (struct ipc_shm_context *context);
 
 struct seq_file;
 void __init ipc_init_ids(struct ipc_ids* ids, int size);

_

#2 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Dave Hansen
Quoted 6 lines · Dave Hansen

Right now, sysctls can only deal with global variables. This patch makes them a _little_ more flexible by allowing there to be an accessor function to get at the variable being changed, instead of it being global.

Quoted 6 lines

This allows the sysctls to be backed by variables that are, for instance, dynamically allocated and not available at compile-time.

This also provides a very simple mechanism to take things that are currently global and containerize them.

hmm, why do you call the sysctl_table_data() over and over again? what's the purpose? what about sideeffects?

Quoted 36 lines

---

 work-dave/include/linux/sysctl.h |    8 ++++
 work-dave/kernel/sysctl.c        |   65 ++++++++++++++++++++++++++-------------
 2 files changed, 52 insertions(+), 21 deletions(-)

diff -puN include/linux/sysctl.h~sysctls-for-containers include/linux/sysctl.h

include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +8 -0
@@ -872,6 +872,7 @@ extern void sysctl_init(void);
 
 typedef struct ctl_table ctl_table;
 
+typedef void *ctl_data_access (void);
 typedef int ctl_handler (ctl_table *table, int __user *name, int nlen,
 			 void __user *oldval, size_t __user *oldlenp,
 			 void __user *newval, size_t newlen, 
@@ -957,6 +958,13 @@ struct ctl_table 
 	int ctl_name;			/* Binary ID */
 	const char *procname;		/* Text ID for /proc/sys, or zero */
 	void *data;
+	ctl_data_access *data_access;	/* set this to a function if you
+					 * don't have a static place to point
+					 * ->data at compile-time.  This
+					 * function will be called to dynamically
+					 * figure out a ->data pointer.  Do not
+					 * set this and ->data at once.
+					 */
 	int maxlen;
 	mode_t mode;
 	ctl_table *child;

diff -puN kernel/sysctl.c~sysctls-for-containers kernel/sysctl.c

kernel/sysctl.c repeat: · line 1199

Quoted 3 lines
kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +0 -0
@@ -1197,6 +1197,24 @@ repeat:
 	return -ENOTDIR;
 }
 

I'd expect that to be inline, and to vanish when containers are disabled ...

kernel/sysctl.c repeat: · line 1208

Quoted 9 lines
kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +9 -0
+void *sysctl_table_data(ctl_table *table)
+{
+	void *data;
+
+	if (table->data && table->data_access) {
+		printk(KERN_WARNING
+			"sysctl: data and accessor function set for: '%s'\n",
+			table->procname);
+		table->data = NULL;

why is ->data and ->data_access evil?

Quoted 198 lines
kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +35 -21
+	}
+
+	data = table->data;
+	if (!data && table->data_access)
+		data = table->data_access();
+
+	return data;
+}
+
 /* Perform the actual read/write of a sysctl table entry. */
 int do_sysctl_strategy (ctl_table *table, 
 			int __user *name, int nlen,
@@ -1205,6 +1223,7 @@ int do_sysctl_strategy (ctl_table *table
 {
 	int op = 0, rc;
 	size_t len;
+	void *data;
 
 	if (oldval)
 		op |= 004;
@@ -1224,14 +1243,15 @@ int do_sysctl_strategy (ctl_table *table
 
 	/* If there is no strategy routine, or if the strategy returns
 	 * zero, proceed with automatic r/w */
-	if (table->data && table->maxlen) {
+	data = sysctl_table_data(table);
+	if (data && table->maxlen) {
 		if (oldval && oldlenp) {
 			if (get_user(len, oldlenp))
 				return -EFAULT;
 			if (len) {
 				if (len > table->maxlen)
 					len = table->maxlen;
-				if(copy_to_user(oldval, table->data, len))
+				if(copy_to_user(oldval, data, len))
 					return -EFAULT;
 				if(put_user(len, oldlenp))
 					return -EFAULT;
@@ -1241,7 +1261,7 @@ int do_sysctl_strategy (ctl_table *table
 			len = newlen;
 			if (len > table->maxlen)
 				len = table->maxlen;
-			if(copy_from_user(table->data, newval, len))
+			if(copy_from_user(data, newval, len))
 				return -EFAULT;
 		}
 	}
@@ -1539,7 +1559,7 @@ int proc_dostring(ctl_table *table, int 
 	char __user *p;
 	char c;
 
-	if (!table->data || !table->maxlen || !*lenp ||
+	if (!sysctl_table_data(table) || !table->maxlen || !*lenp ||
 	    (*ppos && !write)) {
 		*lenp = 0;
 		return 0;
@@ -1557,18 +1577,18 @@ int proc_dostring(ctl_table *table, int 
 		}
 		if (len >= table->maxlen)
 			len = table->maxlen-1;
-		if(copy_from_user(table->data, buffer, len))
+		if(copy_from_user(sysctl_table_data(table), buffer, len))
 			return -EFAULT;
-		((char *) table->data)[len] = 0;
+		((char *) sysctl_table_data(table))[len] = 0;
 		*ppos += *lenp;
 	} else {
-		len = strlen(table->data);
+		len = strlen(sysctl_table_data(table));
 		if (len > table->maxlen)
 			len = table->maxlen;
 		if (len > *lenp)
 			len = *lenp;
 		if (len)
-			if(copy_to_user(buffer, table->data, len))
+			if(copy_to_user(buffer, sysctl_table_data(table), len))
 				return -EFAULT;
 		if (len < *lenp) {
 			if(put_user('\n', ((char __user *) buffer) + len))
@@ -1636,13 +1656,13 @@ static int do_proc_dointvec(ctl_table *t
 	char buf[TMPBUFLEN], *p;
 	char __user *s = buffer;
 
-	if (!table->data || !table->maxlen || !*lenp ||
+	if (!sysctl_table_data(table) || !table->maxlen || !*lenp ||
 	    (*ppos && !write)) {
 		*lenp = 0;
 		return 0;
 	}
 
-	i = (int *) table->data;
+	i = (int *) sysctl_table_data(table);
 	vleft = table->maxlen / sizeof(*i);
 	left = *lenp;
 
@@ -1878,13 +1898,13 @@ static int do_proc_doulongvec_minmax(ctl
 	char buf[TMPBUFLEN], *p;
 	char __user *s = buffer;
 
-	if (!table->data || !table->maxlen || !*lenp ||
+	if (!sysctl_table_data(table) || !table->maxlen || !*lenp ||
 	    (*ppos && !write)) {
 		*lenp = 0;
 		return 0;
 	}
 
-	i = (unsigned long *) table->data;
+	i = (unsigned long *) sysctl_table_data(table);
 	min = (unsigned long *) table->extra1;
 	max = (unsigned long *) table->extra2;
 	vleft = table->maxlen / sizeof(unsigned long);
@@ -2230,7 +2250,9 @@ int sysctl_string(ctl_table *table, int 
 		  void __user *oldval, size_t __user *oldlenp,
 		  void __user *newval, size_t newlen, void **context)
 {
-	if (!table->data || !table->maxlen) 
+	char *data = sysctl_table_data(table);
+
+	if (!data || !table->maxlen)
 		return -ENOTDIR;
 
 	if (oldval && oldlenp) {
@@ -2238,7 +2260,7 @@ int sysctl_string(ctl_table *table, int 
 		if (get_user(bufsize, oldlenp))
 			return -EFAULT;
 		if (bufsize) {
-			size_t len = strlen(table->data), copied;
+			size_t len = strlen(data), copied;
 
 			/* This shouldn't trigger for a well-formed sysctl */
 			if (len > table->maxlen)
@@ -2247,7 +2269,7 @@ int sysctl_string(ctl_table *table, int 
 			/* Copy up to a max of bufsize-1 bytes of the string */
 			copied = (len >= bufsize) ? bufsize - 1 : len;
 
-			if (copy_to_user(oldval, table->data, copied) ||
+			if (copy_to_user(oldval, data, copied) ||
 			    put_user(0, (char __user *)(oldval + copied)))
 				return -EFAULT;
 			if (put_user(len, oldlenp))
@@ -2258,11 +2280,11 @@ int sysctl_string(ctl_table *table, int 
 		size_t len = newlen;
 		if (len > table->maxlen)
 			len = table->maxlen;
-		if(copy_from_user(table->data, newval, len))
+		if(copy_from_user(data, newval, len))
 			return -EFAULT;
 		if (len == table->maxlen)
 			len--;
-		((char *) table->data)[len] = 0;
+		data[len] = 0;
 	}
 	return 1;
 }
@@ -2320,7 +2342,7 @@ int sysctl_jiffies(ctl_table *table, int
 			if (olen!=sizeof(int))
 				return -EINVAL; 
 		}
-		if (put_user(*(int *)(table->data)/HZ, (int __user *)oldval) ||
+		if (put_user(*(int *)(sysctl_table_data(table))/HZ, (int __user *)oldval) ||
 		    (oldlenp && put_user(sizeof(int),oldlenp)))
 			return -EFAULT;
 	}
@@ -2330,7 +2352,7 @@ int sysctl_jiffies(ctl_table *table, int
 			return -EINVAL; 
 		if (get_user(new, (int __user *)newval))
 			return -EFAULT;
-		*(int *)(table->data) = new*HZ; 
+		*(int *)(sysctl_table_data(table)) = new*HZ;
 	}
 	return 1;
 }
@@ -2340,6 +2362,7 @@ int sysctl_ms_jiffies(ctl_table *table, 
 		void __user *oldval, size_t __user *oldlenp,
 		void __user *newval, size_t newlen, void **context)
 {
+	int *jiffies_ptr = sysctl_table_data(table);
 	if (oldval) {
 		size_t olen;
 		if (oldlenp) { 
@@ -2348,7 +2371,7 @@ int sysctl_ms_jiffies(ctl_table *table, 
 			if (olen!=sizeof(int))
 				return -EINVAL; 
 		}
-		if (put_user(jiffies_to_msecs(*(int *)(table->data)), (int __user *)oldval) ||
+		if (put_user(jiffies_to_msecs(*jiffies_ptr), (int __user *)oldval) ||
 		    (oldlenp && put_user(sizeof(int),oldlenp)))
 			return -EFAULT;
 	}
@@ -2358,7 +2381,7 @@ int sysctl_ms_jiffies(ctl_table *table, 
 			return -EINVAL; 
 		if (get_user(new, (int __user *)newval))
 			return -EFAULT;
-		*(int *)(table->data) = msecs_to_jiffies(new);
+		*jiffies_ptr = msecs_to_jiffies(new);
 	}
 	return 1;
 }
Quoted 1 line

_

wouldn't some get/set helper make more sense? i.e. some virtualizer and devirtualizer functions?

best,
Herbert

#6 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Dave Hansen
Quoted 17 lines · * Dave Hansen ([email protected]) wrote:
kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +3 -2
 	/* If there is no strategy routine, or if the strategy returns
 	 * zero, proceed with automatic r/w */
-	if (table->data && table->maxlen) {
+	data = sysctl_table_data(table);
+	if (data && table->maxlen) {
 		if (oldval && oldlenp) {
 			if (get_user(len, oldlenp))
 				return -EFAULT;
 			if (len) {
 				if (len > table->maxlen)
 					len = table->maxlen;
-				if(copy_to_user(oldval, table->data, len))
+				if(copy_to_user(oldval, data, len))
 					return -EFAULT;
 				if(put_user(len, oldlenp))
 					return -EFAULT;

kernel/sysctl.c int do_sysctl_strategy (ctl_table *table · line 1265

Quoted 6 lines
kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +1 -1
@@ -1241,7 +1261,7 @@ int do_sysctl_strategy (ctl_table *table
 			len = newlen;
 			if (len > table->maxlen)
 				len = table->maxlen;
-			if(copy_from_user(table->data, newval, len))
+			if(copy_from_user(data, newval, len))
 				return -EFAULT;

Interesting idea. One piece that's missing is strategy for controlling creation the new context (assuming the data_access() will always evaluate into a context sensitive piece of data). Otherwise a user can get out of the limits imposed by sysadmin (since they may have placed themselves in a context which differs from admin).

thanks,
-chris

#10 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Dave Hansen
Quoted 6 lines · Dave Hansen

Right now, sysctls can only deal with global variables. This patch makes them a _little_ more flexible by allowing there to be an accessor function to get at the variable being changed, instead of it being global.

Quoted 7 lines

This allows the sysctls to be backed by variables that are, for instance, dynamically allocated and not available at compile-time.

This also provides a very simple mechanism to take things that are currently global and containerize them.

This is disgusting. Please, don't pile more and more complexity into sysctl_table - it's already choke-full of it and needs to be simplified, not to grow more crap.

NAK.

#16 lore raw

Re: [RFC][PATCH 2/6] sysvmsg: containerize

In reply to Dave Hansen

include/linux/init_task.h line 8

Quoted 10 lines · * Dave Hansen ([email protected]) wrote:
include/linux/init_task.h renamed from include/linux/init_task.h~sysvmsg-container +1 -0
@@ -2,6 +2,7 @@
 #define _LINUX__INIT_TASK_H
 
 #include <linux/file.h>
+#include <linux/ipc.h>
 #include <linux/rcupdate.h>
 
 #define INIT_FDTABLE \

missing INIT_IPC_CONTEXT type of macro?

Quoted 18 lines

diff -puN include/linux/ipc.h~sysvmsg-container include/linux/ipc.h

include/linux/ipc.h renamed from include/linux/ipc.h~sysvmsg-container +4 -0
@@ -68,6 +68,10 @@ struct kern_ipc_perm
 	void		*security;
 };
 
+struct ipc_context;
+extern struct ipc_context *ipc_context;
+
+extern struct ipc_msg_context *init_ipc_context(void);
 #endif /* __KERNEL__ */
 
 #endif /* _LINUX_IPC_H */

diff -puN include/linux/sched.h~sysvmsg-container include/linux/sched.h

include/linux/sched.h struct task_struct { · line 796

Quoted 4 lines
include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +1 -0
@@ -793,6 +793,7 @@ struct task_struct {
 	int link_count, total_link_count;
 /* ipc stuff */
 	struct sysv_sem sysvsem;
+	struct ipc_context *ipc_context;

how does this propagate on clone (presently it's just side-effect of dup_task_struct starting from init_task, with no dynamic lifetime), how is it meant to be changed?

Quoted 38 lines
include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +0 -0
 /* CPU-specific state of this task */
 	struct thread_struct thread;
 /* filesystem information */

diff -puN ipc/msg.c~sysvmsg-container ipc/msg.c

ipc/msg.c renamed from ipc/msg.c~sysvmsg-container +9 -14
@@ -60,35 +60,30 @@ struct msg_sender {
 #define SEARCH_NOTEQUAL		3
 #define SEARCH_LESSEQUAL	4
 
-static atomic_t msg_bytes = ATOMIC_INIT(0);
-static atomic_t msg_hdrs = ATOMIC_INIT(0);
+#define msg_lock(ctx, id)	((struct msg_queue*)ipc_lock(&ctx->ids,id))
+#define msg_unlock(ctx, msq)	ipc_unlock(&(msq)->q_perm)
+#define msg_rmid(ctx, id)	((struct msg_queue*)ipc_rmid(&ctx->ids,id))
+#define msg_checkid(ctx, msq, msgid)	\
+	ipc_checkid(&ctx->ids,&msq->q_perm,msgid)
+#define msg_buildid(ctx, id, seq) \
+	ipc_buildid(&ctx->ids, id, seq)
 
-static struct ipc_ids msg_ids;
-
-#define msg_lock(id)	((struct msg_queue*)ipc_lock(&msg_ids,id))
-#define msg_unlock(msq)	ipc_unlock(&(msq)->q_perm)
-#define msg_rmid(id)	((struct msg_queue*)ipc_rmid(&msg_ids,id))
-#define msg_checkid(msq, msgid)	\
-	ipc_checkid(&msg_ids,&msq->q_perm,msgid)
-#define msg_buildid(id, seq) \
-	ipc_buildid(&msg_ids, id, seq)
-
-static void freeque (struct msg_queue *msq, int id);
-static int newque (key_t key, int msgflg);
+static void freeque (struct ipc_msg_context *, struct msg_queue *msq, int id);
+static int newque (struct ipc_msg_context *context, key_t key, int id);
 #ifdef CONFIG_PROC_FS
 static int sysvipc_msg_proc_show(struct seq_file *s, void *it);
 #endif
 

ipc/msg.c struct msg_sender { · line 83

Quoted 10 lines
ipc/msg.c renamed from ipc/msg.c~sysvmsg-container +3 -3
-void __init msg_init (void)
+void __init msg_init (struct ipc_msg_context *context)
 {
-	ipc_init_ids(&msg_ids,msg_ctlmni);
+	ipc_init_ids(&context->ids,msg_ctlmni);
 	ipc_init_proc_interface("sysvipc/msg",
 				"       key      msqid perms      cbytes       qnum lspid lrpid   uid   gid  cuid  cgid      stime      rtime      ctime\n",
-				&msg_ids,
+				&context->ids,
 				sysvipc_msg_proc_show);

Does that mean /proc interface only gets init_task context? Along those lines, I think now ipcs -a is incomplete from admin perspective. Suppose that's a feature from the container/vserver POV.

Quoted 16 lines
ipc/util.h renamed from ipc/util.h~sysvmsg-container +2 -1
@@ -12,7 +12,7 @@
 #define SEQ_MULTIPLIER	(IPCMNI)
 
 void sem_init (void);
-void msg_init (void);
+void msg_init (struct ipc_msg_context *context);
 void shm_init (void);
 
 struct ipc_id_ary {
@@ -30,6 +30,18 @@ struct ipc_ids {
 	struct ipc_id_ary* entries;
 };
 
+struct ipc_msg_context {

ipc/util.h struct ipc_ids { · line 43

Quoted 10 lines
ipc/util.h renamed from ipc/util.h~sysvmsg-container +10 -0
+	atomic_t bytes;
+	atomic_t hdrs;
+
+	struct ipc_ids ids;
+	struct kref count;
+};
+
+struct ipc_context {
+	struct ipc_msg_context msg;
+};

This looks a little suspect. ref count embedded in embedded object. My first thought was, sem and shared memory would introduce same, and now we'd have 3 independent refounts for top level object. However, it's not used, and doesn't appear to be mirrored in the sem and shared mem contexts. Perhaps it is just a leftover?

thanks,
-chris

#11 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Al Viro
Quoted 3 lines · Al Viro

This is disgusting. Please, don't pile more and more complexity into sysctl_table - it's already choke-full of it and needs to be simplified, not to grow more crap.

I don't completely disagree. It certainly isn't the most elegant approach I've ever seen.

Any ideas on ways we could simplify it? I was thinking that we could get rid of the .data member and allow access only via the mechanism I just introduced. It would be pretty easy to make some macros to generate "simple" access functions for the existing global variables.

-- Dave

#12 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

Quoted 5 lines · Dave Hansen
Quoted 3 lines · Al Viro

This is disgusting. Please, don't pile more and more complexity into sysctl_table - it's already choke-full of it and needs to be simplified, not to grow more crap.

Quoted 7 lines

I don't completely disagree. It certainly isn't the most elegant approach I've ever seen.

Any ideas on ways we could simplify it? I was thinking that we could get rid of the .data member and allow access only via the mechanism I just introduced. It would be pretty easy to make some macros to generate "simple" access functions for the existing global variables.

I'll resurrect the sysctl-cleanups tree and drop it on kernel.org tonight.

#3 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Herbert Poetzl
Quoted 11 lines · Herbert Poetzl
Quoted 10 lines · Dave Hansen

Right now, sysctls can only deal with global variables. This patch makes them a _little_ more flexible by allowing there to be an accessor function to get at the variable being changed, instead of it being global.

This allows the sysctls to be backed by variables that are, for instance, dynamically allocated and not available at compile-time.

Quoted 5 lines
Quoted 2 lines

This also provides a very simple mechanism to take things that are currently global and containerize them.

hmm, why do you call the sysctl_table_data() over and over again? what's the purpose?

Letting me be lazy and code with s/// :)

For the current application, it doesn't really matter. But, I can imagine that other users could be a bit more costly.

Quoted 1 line

what about sideeffects?

Require that there aren't any. ;)

It might be necessary to have something effectively implementing put and get, but this certainly doesn't need it yet.

Quoted 34 lines
Quoted 34 lines

---

 work-dave/include/linux/sysctl.h |    8 ++++
 work-dave/kernel/sysctl.c        |   65 ++++++++++++++++++++++++++-------------
 2 files changed, 52 insertions(+), 21 deletions(-)

diff -puN include/linux/sysctl.h~sysctls-for-containers include/linux/sysctl.h

include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +8 -0
@@ -872,6 +872,7 @@ extern void sysctl_init(void);
 
 typedef struct ctl_table ctl_table;
 
+typedef void *ctl_data_access (void);
 typedef int ctl_handler (ctl_table *table, int __user *name, int nlen,
 			 void __user *oldval, size_t __user *oldlenp,
 			 void __user *newval, size_t newlen, 
@@ -957,6 +958,13 @@ struct ctl_table 
 	int ctl_name;			/* Binary ID */
 	const char *procname;		/* Text ID for /proc/sys, or zero */
 	void *data;
+	ctl_data_access *data_access;	/* set this to a function if you
+					 * don't have a static place to point
+					 * ->data at compile-time.  This
+					 * function will be called to dynamically
+					 * figure out a ->data pointer.  Do not
+					 * set this and ->data at once.
+					 */
 	int maxlen;
 	mode_t mode;
 	ctl_table *child;

diff -puN kernel/sysctl.c~sysctls-for-containers kernel/sysctl.c
--- work/kernel/sysctl.c~sysctls-for-containers 2006-03-06 15:41:55.000000000 -0800

Quoted 8 lines

include/linux/sysctl.h repeat: · line 1199

Quoted 5 lines

+++ work-dave/kernel/sysctl.c 2006-03-06 15:41:55.000000000 -0800

include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +0 -0
@@ -1197,6 +1197,24 @@ repeat:
 	return -ENOTDIR;
 }
 

I'd expect that to be inline, and to vanish when containers are disabled ...

Unless we want non-container code to be able to use it. I guess we could restrict it to containers only, though.

Quoted 11 lines

include/linux/sysctl.h repeat: · line 1208

Quoted 9 lines
include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +9 -0
+void *sysctl_table_data(ctl_table *table)
+{
+	void *data;
+
+	if (table->data && table->data_access) {
+		printk(KERN_WARNING
+			"sysctl: data and accessor function set for: '%s'\n",
+			table->procname);
+		table->data = NULL;

why is ->data and ->data_access evil?

As it stands, which one do you use? What if they aren't consistent? Do you use both? Just one? Which first? Easiest to just say that it isn't allowed.

Quoted 2 lines

wouldn't some get/set helper make more sense? i.e. some virtualizer and devirtualizer functions?

I'm not quite sure I know what you mean. Can you elaborate some more?

-- Dave

#7 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Chris Wright
Quoted 5 lines · Chris Wright

Interesting idea. One piece that's missing is strategy for controlling creation the new context (assuming the data_access() will always evaluate into a context sensitive piece of data). Otherwise a user can get out of the limits imposed by sysadmin (since they may have placed themselves in a context which differs from admin).

Yup, that is missing for now. We couldn't agree on quite which implementation we want for basic containers/vservers/vpses. So, for now, making it useful is left as an exercise to the reader. :)

BTW, the current code _is_ potentially context sensitive because "current" provides much of the context that we will ever need.

-- Dave

#17 lore raw

Re: [RFC][PATCH 2/6] sysvmsg: containerize

In reply to Chris Wright
Quoted 6 lines · Chris Wright
Quoted 5 lines · * Dave Hansen ([email protected]) wrote:
include/linux/init_task.h renamed from include/linux/init_task.h~sysvmsg-container +0 -0
@@ -2,6 +2,7 @@
 #define _LINUX__INIT_TASK_H
 
Quoted 7 lines

include/linux/init_task.h line 8

Quoted 5 lines
include/linux/init_task.h renamed from include/linux/init_task.h~sysvmsg-container +1 -0
 #include <linux/file.h>
+#include <linux/ipc.h>
 #include <linux/rcupdate.h>
 
 #define INIT_FDTABLE \

missing INIT_IPC_CONTEXT type of macro?

Yeah, I was doing something like that at first. But, I decided to dynamically allocate it, much like it will have to be done when we have _real_ namespaces for it. That means that we can't actually initialize it at init_task.h time.

Quoted 4 lines
Quoted 4 lines

diff -puN include/linux/sched.h~sysvmsg-container include/linux/sched.h

include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +0 -0
@@ -793,6 +793,7 @@ struct task_struct {
Quoted 8 lines

include/linux/sched.h struct task_struct { · line 796

Quoted 4 lines
include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +1 -0
 	int link_count, total_link_count;
 /* ipc stuff */
 	struct sysv_sem sysvsem;
+	struct ipc_context *ipc_context;

how does this propagate on clone (presently it's just side-effect of dup_task_struct starting from init_task, with no dynamic lifetime), how is it meant to be changed?

Some interface later down the road. You caught me a bit early in the cycle.

Quoted 7 lines
Quoted 7 lines
include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +2 -2
-void __init msg_init (void)
+void __init msg_init (struct ipc_msg_context *context)
 {
-	ipc_init_ids(&msg_ids,msg_ctlmni);
+	ipc_init_ids(&context->ids,msg_ctlmni);
include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +0 -0
 	ipc_init_proc_interface("sysvipc/msg",
 				"       key      msqid perms      cbytes       qnum lspid lrpid   uid   gid  cuid  cgid      stime      rtime      ctime\n",
Quoted 8 lines

include/linux/sched.h

Quoted 3 lines
include/linux/sched.h renamed from include/linux/sched.h~sysvmsg-container +1 -1
-				&msg_ids,
+				&context->ids,
 				sysvipc_msg_proc_show);

Does that mean /proc interface only gets init_task context? Along those lines, I think now ipcs -a is incomplete from admin perspective. Suppose that's a feature from the container/vserver POV.

It will get context from the current task, which means the current container. We haven't quite decided how these things will be (or if they need to be) aggregated on a a system-wide basis.

Quoted 24 lines
Quoted 24 lines
ipc/util.h renamed from ipc/util.h~sysvmsg-container +10 -1
@@ -12,7 +12,7 @@
 #define SEQ_MULTIPLIER	(IPCMNI)
 
 void sem_init (void);
-void msg_init (void);
+void msg_init (struct ipc_msg_context *context);
 void shm_init (void);
 
 struct ipc_id_ary {
@@ -30,6 +30,18 @@ struct ipc_ids {
 	struct ipc_id_ary* entries;
 };
 
+struct ipc_msg_context {
+	atomic_t bytes;
+	atomic_t hdrs;
+
+	struct ipc_ids ids;
+	struct kref count;
+};
+
+struct ipc_context {
Quoted 8 lines

ipc/util.h struct ipc_ids { · line 43

Quoted 2 lines
ipc/util.h renamed from ipc/util.h~sysvmsg-container +2 -0
+	struct ipc_msg_context msg;
+};

This looks a little suspect. ref count embedded in embedded object. My first thought was, sem and shared memory would introduce same, and now we'd have 3 independent refounts for top level object. However, it's not used, and doesn't appear to be mirrored in the sem and shared mem contexts. Perhaps it is just a leftover?

Yeah, I have the feeling that it used to be an atomic_t, and got converted over to a kref a bit needlessly. I'll investigate more when I spin the patches again.

-- Dave

#8 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Dave Hansen
Quoted 3 lines · * Dave Hansen ([email protected]) wrote:

Yup, that is missing for now. We couldn't agree on quite which implementation we want for basic containers/vservers/vpses. So, for now, making it useful is left as an exercise to the reader. :)

;-)

Quoted 2 lines

BTW, the current code _is_ potentially context sensitive because "current" provides much of the context that we will ever need.

Right. More a question of sysadmin lowering limits which aren't seen in any context but sysadmin context. But from your comments above and in other mail, I understand the requirements aren't fully baked yet.

thanks,
-chris

#18 lore raw

Re: [RFC][PATCH 2/6] sysvmsg: containerize

In reply to Dave Hansen
Quoted 18 lines · * Dave Hansen ([email protected]) wrote:
Quoted 16 lines · Chris Wright

ipc/msg.c struct msg_sender { · line 83

Quoted 10 lines · * Dave Hansen ([email protected]) wrote:
ipc/msg.c renamed from ipc/msg.c~sysvmsg-container +3 -3
-void __init msg_init (void)
+void __init msg_init (struct ipc_msg_context *context)
 {
-	ipc_init_ids(&msg_ids,msg_ctlmni);
+	ipc_init_ids(&context->ids,msg_ctlmni);
 	ipc_init_proc_interface("sysvipc/msg",
 				"       key      msqid perms      cbytes       qnum lspid lrpid   uid   gid  cuid  cgid      stime      rtime      ctime\n",
-				&msg_ids,
+				&context->ids,
 				sysvipc_msg_proc_show);

Does that mean /proc interface only gets init_task context? Along those lines, I think now ipcs -a is incomplete from admin perspective. Suppose that's a feature from the container/vserver POV.

Quoted 3 lines

It will get context from the current task, which means the current container. We haven't quite decided how these things will be (or if they need to be) aggregated on a a system-wide basis.

The /proc interface is registering with &context->ids of init_task. So, all other contexts using that interface will be looking at the wrong info, AFAICT.

As you can tell my concerns are in resource consumption. If a user can create contexts which it can hide from sysadmin, and they aren't subject to sysadmin mandated resource limits, it's effectively a leak, esp. since these resources don't die with exit(2).

thanks,
-chris

#25 lore raw

Re: [RFC][PATCH 4/6] sysvsem: containerize

In reply to Dave Hansen

ipc/sem.c int sem_ctls[4] = {SEMMSL, SEMMNS, SEMOP · line 115

Quoted 3 lines · * Dave Hansen ([email protected]) wrote:
ipc/sem.c renamed from ipc/sem.c~sysvsem-container +0 -0
@@ -112,15 +114,12 @@ int sem_ctls[4] = {SEMMSL, SEMMNS, SEMOP
 #define sc_semopm	(sem_ctls[2])
 #define sc_semmni	(sem_ctls[3])

Minor one, but didn't see an attempt to handle these sem_ctls global sysctls.

thanks,
-chris

#4 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Dave Hansen
Quoted 73 lines · Dave Hansen
Quoted 11 lines · Herbert Poetzl
Quoted 10 lines · Dave Hansen

Right now, sysctls can only deal with global variables. This patch makes them a _little_ more flexible by allowing there to be an accessor function to get at the variable being changed, instead of it being global.

This allows the sysctls to be backed by variables that are, for instance, dynamically allocated and not available at compile-time.

Quoted 5 lines
Quoted 2 lines

This also provides a very simple mechanism to take things that are currently global and containerize them.

hmm, why do you call the sysctl_table_data() over and over again? what's the purpose?

Letting me be lazy and code with s/// :)

For the current application, it doesn't really matter. But, I can imagine that other users could be a bit more costly.

Quoted 1 line

what about sideeffects?

Require that there aren't any. ;)

It might be necessary to have something effectively implementing put and get, but this certainly doesn't need it yet.

Quoted 34 lines
Quoted 34 lines

---

 work-dave/include/linux/sysctl.h |    8 ++++
 work-dave/kernel/sysctl.c        |   65 ++++++++++++++++++++++++++-------------
 2 files changed, 52 insertions(+), 21 deletions(-)

diff -puN include/linux/sysctl.h~sysctls-for-containers include/linux/sysctl.h

include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +8 -0
@@ -872,6 +872,7 @@ extern void sysctl_init(void);
 
 typedef struct ctl_table ctl_table;
 
+typedef void *ctl_data_access (void);
 typedef int ctl_handler (ctl_table *table, int __user *name, int nlen,
 			 void __user *oldval, size_t __user *oldlenp,
 			 void __user *newval, size_t newlen, 
@@ -957,6 +958,13 @@ struct ctl_table 
 	int ctl_name;			/* Binary ID */
 	const char *procname;		/* Text ID for /proc/sys, or zero */
 	void *data;
+	ctl_data_access *data_access;	/* set this to a function if you
+					 * don't have a static place to point
+					 * ->data at compile-time.  This
+					 * function will be called to dynamically
+					 * figure out a ->data pointer.  Do not
+					 * set this and ->data at once.
+					 */
 	int maxlen;
 	mode_t mode;
 	ctl_table *child;

diff -puN kernel/sysctl.c~sysctls-for-containers kernel/sysctl.c
--- work/kernel/sysctl.c~sysctls-for-containers 2006-03-06 15:41:55.000000000 -0800

Quoted 8 lines

include/linux/sysctl.h repeat: · line 1199

Quoted 5 lines

+++ work-dave/kernel/sysctl.c 2006-03-06 15:41:55.000000000 -0800

include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +0 -0
@@ -1197,6 +1197,24 @@ repeat:
 	return -ENOTDIR;
 }
 

I'd expect that to be inline, and to vanish when containers are disabled ...

Quoted 2 lines

Unless we want non-container code to be able to use it. I guess we could restrict it to containers only, though.

uhm, what about kernel compiled _without_ container support?

Quoted 12 lines
Quoted 11 lines

include/linux/sysctl.h repeat: · line 1208

Quoted 9 lines
include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +9 -0
+void *sysctl_table_data(ctl_table *table)
+{
+	void *data;
+
+	if (table->data && table->data_access) {
+		printk(KERN_WARNING
+			"sysctl: data and accessor function set for: '%s'\n",
+			table->procname);
+		table->data = NULL;

why is ->data and ->data_access evil?

Quoted 3 lines

As it stands, which one do you use? What if they aren't consistent? Do you use both? Just one? Which first? Easiest to just say that it isn't allowed.

wouldn't it make sense to have some 'default' in ->data and let ->data_access() override it when there is something to override?

Quoted 4 lines
Quoted 2 lines

wouldn't some get/set helper make more sense? i.e. some virtualizer and devirtualizer functions?

I'm not quite sure I know what you mean. Can you elaborate some more?

sure, basically we have two cases which interact with userspace: the read and the write case.

for the read case, we want something which gives us the 'virtualized' view, which might often be the same as the host gets, where for the write case it is usually not that simple, as we might not want a context to write to 'world' stuff

so, having two different functions here, or one which gets passed the direction, might be much simpler to adjust in many cases than adding more and more structures ... but YMMV

best,
Herbert

Quoted 1 line

-- Dave

#9 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Dave Hansen
Quoted 7 lines · Dave Hansen
Quoted 7 lines

Interesting idea. One piece that's missing is strategy for controlling creation the new context (assuming the data_access() will always evaluate into a context sensitive piece of data). Otherwise a user can get out of the limits imposed by sysadmin (since they may have placed themselves in a context which differs from admin).

Quoted 8 lines

Yup, that is missing for now. We couldn't agree on quite which implementation we want for basic containers/vservers/vpses. So, for now, making it useful is left as an exercise to the reader. :)

BTW, the current code _is_ potentially context sensitive because "current" provides much of the context that we will ever need.

I have an RFC quality submission ready, extracted from Herbert's work. I'll prepare and forward it to the ML now for reference.

Sam.

#26 lore raw

Re: [RFC][PATCH 4/6] sysvsem: containerize

In reply to Chris Wright
Quoted 6 lines · Chris Wright

ipc/sem.c int sem_ctls[4] = {SEMMSL, SEMMNS, SEMOP · line 115

Quoted 3 lines · * Dave Hansen ([email protected]) wrote:
ipc/sem.c renamed from ipc/sem.c~sysvsem-container +0 -0
@@ -112,15 +114,12 @@ int sem_ctls[4] = {SEMMSL, SEMMNS, SEMOP
 #define sc_semopm	(sem_ctls[2])
 #define sc_semmni	(sem_ctls[3])

Minor one, but didn't see an attempt to handle these sem_ctls global sysctls.

Nope, haven't gotten to those yet. Thanks for pointing them out.

-- Dave

#13 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Al Viro
Quoted 9 lines · Al Viro
Quoted 8 lines · Dave Hansen
Quoted 3 lines · Al Viro

This is disgusting. Please, don't pile more and more complexity into sysctl_table - it's already choke-full of it and needs to be simplified, not to grow more crap.

I don't completely disagree. It certainly isn't the most elegant approach I've ever seen.

Quoted 6 lines
Quoted 4 lines

Any ideas on ways we could simplify it? I was thinking that we could get rid of the .data member and allow access only via the mechanism I just introduced. It would be pretty easy to make some macros to generate "simple" access functions for the existing global variables.

I'll resurrect the sysctl-cleanups tree and drop it on kernel.org tonight.

Has that happened yet?

I just looked and I didn't see anything up there.

Eric

#14 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Dave Hansen
Quoted 11 lines · Dave Hansen

Right now, sysctls can only deal with global variables. This patch makes them a _little_ more flexible by allowing there to be an accessor function to get at the variable being changed, instead of it being global.

This allows the sysctls to be backed by variables that are, for instance, dynamically allocated and not available at compile-time.

This also provides a very simple mechanism to take things that are currently global and containerize them.

Sorry for taking so long to look at this I just spotted this series of patches.

The parameters that describe the variable are:
data, maxlen, extra1, and extra2.

I am concerned that this does not provide a capability for anything except data to vary at runtime.

Also so that we can do a good job and report process local resources in /proc this data_access should take a task_struct parameter.

data_access should also be passed the the ctl_table in case we can make an generic data accessor.

That would allow us to put offsetof is data and then simply add currrent->ipc_context to get the address of the variable. Which should keep the number of accessor functions down.

Which give a signature something like:

struct ctl_data_info {
       void *data;
       int maxlen;
       void *extra1;
       void *extra2;
};

void ctl_data_access(struct ctl_table *tbl, sruct task_struct *task,

			struct ctl_data_info *info);
Quoted 48 lines

---

 work-dave/include/linux/sysctl.h |    8 ++++
 work-dave/kernel/sysctl.c | 65 ++++++++++++++++++++++++++-------------
 2 files changed, 52 insertions(+), 21 deletions(-)

diff -puN include/linux/sysctl.h~sysctls-for-containers include/linux/sysctl.h --- work/include/linux/sysctl.h~sysctls-for-containers 2006-03-06 15:41:55.000000000 -0800
+++ work-dave/include/linux/sysctl.h 2006-03-06 15:41:55.000000000 -0800

include/linux/sysctl.h renamed from include/linux/sysctl.h~sysctls-for-containers +8 -0
@@ -872,6 +872,7 @@ extern void sysctl_init(void);
 
 typedef struct ctl_table ctl_table;
 
+typedef void *ctl_data_access (void);
 typedef int ctl_handler (ctl_table *table, int __user *name, int nlen,
 			 void __user *oldval, size_t __user *oldlenp,
 			 void __user *newval, size_t newlen, 
@@ -957,6 +958,13 @@ struct ctl_table 
 	int ctl_name;			/* Binary ID */
 	const char *procname;		/* Text ID for /proc/sys, or zero */
 	void *data;
+	ctl_data_access *data_access;	/* set this to a function if you
+					 * don't have a static place to point
+					 * ->data at compile-time.  This
+ * function will be called to dynamically
+					 * figure out a ->data pointer.  Do not
+					 * set this and ->data at once.
+					 */
 	int maxlen;
 	mode_t mode;
 	ctl_table *child;

diff -puN kernel/sysctl.c~sysctls-for-containers kernel/sysctl.c --- work/kernel/sysctl.c~sysctls-for-containers 2006-03-06 15:41:55.000000000 -0800
+++ work-dave/kernel/sysctl.c 2006-03-06 15:41:55.000000000 -0800

kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +7 -0
@@ -1197,6 +1197,24 @@ repeat:
 	return -ENOTDIR;
 }
 
+void *sysctl_table_data(ctl_table *table)
+{
+	void *data;
+
+	if (table->data && table->data_access) {
+		printk(KERN_WARNING
+			"sysctl: data and accessor function set for: '%s'\n",

kernel/sysctl.c repeat: · line 1216

Quoted 10 lines
kernel/sysctl.c renamed from kernel/sysctl.c~sysctls-for-containers +10 -0
+			table->procname);
+		table->data = NULL;
+	}
+
+	data = table->data;
+	if (!data && table->data_access)
+		data = table->data_access();
+
+	return data;
+}

I think we should always call data_access if it is populated.

For the rest it looks like it is getting there.

Eric

#19 lore raw

Re: [RFC][PATCH 2/6] sysvmsg: containerize

In reply to Chris Wright
Quoted 20 lines · Chris Wright
Quoted 9 lines · * Dave Hansen ([email protected]) wrote:
Quoted 8 lines · Chris Wright
Quoted 7 lines · * Dave Hansen ([email protected]) wrote:
ipc/msg.c renamed from ipc/msg.c~sysvmsg-container +2 -2
-void __init msg_init (void)
+void __init msg_init (struct ipc_msg_context *context)
 {
-	ipc_init_ids(&msg_ids,msg_ctlmni);
+	ipc_init_ids(&context->ids,msg_ctlmni);
 	ipc_init_proc_interface("sysvipc/msg",

" key msqid perms cbytes qnum lspid lrpid uid gid cuid cgid stime rtime

ctime\n",

Quoted 9 lines
Quoted 8 lines

ipc/msg.c struct msg_sender { · line 83

Quoted 3 lines
ipc/msg.c renamed from ipc/msg.c~sysvmsg-container +1 -1
-				&msg_ids,
+				&context->ids,
 				sysvipc_msg_proc_show);

Does that mean /proc interface only gets init_task context? Along those lines, I think now ipcs -a is incomplete from admin perspective. Suppose that's a feature from the container/vserver POV.

Quoted 7 lines
Quoted 3 lines

It will get context from the current task, which means the current container. We haven't quite decided how these things will be (or if they need to be) aggregated on a a system-wide basis.

The /proc interface is registering with &context->ids of init_task. So, all other contexts using that interface will be looking at the wrong info, AFAICT.

We need to make this per process in /proc to get it right. So /proc/sysvipc becomes a symlink to /proc/<pid>/sysvipc.

Quoted 4 lines

As you can tell my concerns are in resource consumption. If a user can create contexts which it can hide from sysadmin, and they aren't subject to sysadmin mandated resource limits, it's effectively a leak, esp. since these resources don't die with exit(2).

I haven't spotted it yet in Dave's series but this is something that should happen when all of the tasks using the ipc_context in this case exit.

Eric

#5 lore raw

Re: [RFC][PATCH 1/6] prepare sysctls for containers

In reply to Herbert Poetzl
Quoted 9 lines · Herbert Poetzl

sure, basically we have two cases which interact with userspace: the read and the write case.

for the read case, we want something which gives us the 'virtualized' view, which might often be the same as the host gets, where for the write case it is usually not that simple, as we might not want a context to write to 'world' stuff

Quoted 4 lines

so, having two different functions here, or one which gets passed the direction, might be much simpler to adjust in many cases than adding more and more structures ... but YMMV

I'm not convinced either way yet but my gut feel is that the use case for get/put functions is incomplete permission checks on various sysctls.

If that is the fixing the permission checks looks like the right thing to do.

Eric

#20 lore raw

Re: [RFC][PATCH 2/6] sysvmsg: containerize

In reply to Eric W. Biederman
Quoted 7 lines · * Eric W. Biederman ([email protected]) wrote:
Quoted 3 lines · Chris Wright

The /proc interface is registering with &context->ids of init_task. So, all other contexts using that interface will be looking at the wrong info, AFAICT.

We need to make this per process in /proc to get it right. So /proc/sysvipc becomes a symlink to /proc/<pid>/sysvipc.

That, and any considerations for context access to protect against reading /proc/pid/sysvipc/* (assuming you can share pid namespace, while not sharing sysvipc context).

Quoted 7 lines
Quoted 4 lines

As you can tell my concerns are in resource consumption. If a user can create contexts which it can hide from sysadmin, and they aren't subject to sysadmin mandated resource limits, it's effectively a leak, esp. since these resources don't die with exit(2).

I haven't spotted it yet in Dave's series but this is something that should happen when all of the tasks using the ipc_context in this case exit.

Making it look like an 'init 0' from the P.O.V. of that ipc_context, WFM. Seems the context needs to have context limits so any privileged process in the context is still subject to the top-level administered limits.

thanks,
-chris

#21 lore raw

Re: [RFC][PATCH 2/6] sysvmsg: containerize

Quoted 6 lines · Chris Wright
Quoted 5 lines · * Eric W. Biederman ([email protected]) wrote:
Quoted 3 lines · Chris Wright

The /proc interface is registering with &context->ids of init_task. So, all other contexts using that interface will be looking at the wrong info, AFAICT.

Quoted 6 lines
Quoted 2 lines

We need to make this per process in /proc to get it right. So /proc/sysvipc becomes a symlink to /proc/<pid>/sysvipc.

That, and any considerations for context access to protect against reading /proc/pid/sysvipc/* (assuming you can share pid namespace, while not sharing sysvipc context).

Yes. Although reading is fairly harmless activity.

Quoted 11 lines
Quoted 7 lines
Quoted 4 lines

As you can tell my concerns are in resource consumption. If a user can create contexts which it can hide from sysadmin, and they aren't subject to sysadmin mandated resource limits, it's effectively a leak, esp. since these resources don't die with exit(2).

I haven't spotted it yet in Dave's series but this is something that should happen when all of the tasks using the ipc_context in this case exit.

Making it look like an 'init 0' from the P.O.V. of that ipc_context, WFM. Seems the context needs to have context limits so any privileged process in the context is still subject to the top-level administered limits.

Agreed. Getting the limit checking correct without imposing measurable overhead is tricky. My gut feel is that the limits should remain global until we can agree on how to implement more fine grained limits.

Eric

#22 lore raw

Re: [RFC][PATCH 2/6] sysvmsg: containerize

Quoted 3 lines · * Eric W. Biederman ([email protected]) wrote:

Agreed. Getting the limit checking correct without imposing measurable overhead is tricky. My gut feel is that the limits should remain global until we can agree on how to implement more fine grained limits.

Sounds reasonable.