Commit 4e982311 authored by Kirill Korotaev's avatar Kirill Korotaev Committed by Linus Torvalds
Browse files

[PATCH] IPC namespace - shm



IPC namespace support for IPC shm code.
Signed-off-by: default avatarPavel Emelianiov <xemul@openvz.org>
Signed-off-by: default avatarKirill Korotaev <dev@openvz.org>
Signed-off-by: default avatarAndrew Morton <akpm@osdl.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@osdl.org>
parent e3893534
...@@ -15,6 +15,10 @@ ...@@ -15,6 +15,10 @@
* *
* support for audit of ipc object properties and permission changes * support for audit of ipc object properties and permission changes
* Dustin Kirkland <dustin.kirkland@us.ibm.com> * Dustin Kirkland <dustin.kirkland@us.ibm.com>
*
* namespaces support
* OpenVZ, SWsoft Inc.
* Pavel Emelianov <xemul@openvz.org>
*/ */
#include <linux/slab.h> #include <linux/slab.h>
...@@ -32,6 +36,7 @@ ...@@ -32,6 +36,7 @@
#include <linux/ptrace.h> #include <linux/ptrace.h>
#include <linux/seq_file.h> #include <linux/seq_file.h>
#include <linux/mutex.h> #include <linux/mutex.h>
#include <linux/nsproxy.h>
#include <asm/uaccess.h> #include <asm/uaccess.h>
...@@ -40,59 +45,115 @@ ...@@ -40,59 +45,115 @@
static struct file_operations shm_file_operations; static struct file_operations shm_file_operations;
static struct vm_operations_struct shm_vm_ops; static struct vm_operations_struct shm_vm_ops;
static struct ipc_ids shm_ids; static struct ipc_ids init_shm_ids;
#define shm_ids(ns) (*((ns)->ids[IPC_SHM_IDS]))
#define shm_lock(id) ((struct shmid_kernel*)ipc_lock(&shm_ids,id)) #define shm_lock(ns, id) \
#define shm_unlock(shp) ipc_unlock(&(shp)->shm_perm) ((struct shmid_kernel*)ipc_lock(&shm_ids(ns),id))
#define shm_get(id) ((struct shmid_kernel*)ipc_get(&shm_ids,id)) #define shm_unlock(shp) \
#define shm_buildid(id, seq) \ ipc_unlock(&(shp)->shm_perm)
ipc_buildid(&shm_ids, id, seq) #define shm_get(ns, id) \
((struct shmid_kernel*)ipc_get(&shm_ids(ns),id))
#define shm_buildid(ns, id, seq) \
ipc_buildid(&shm_ids(ns), id, seq)
static int newseg (key_t key, int shmflg, size_t size); static int newseg (struct ipc_namespace *ns, key_t key,
int shmflg, size_t size);
static void shm_open (struct vm_area_struct *shmd); static void shm_open (struct vm_area_struct *shmd);
static void shm_close (struct vm_area_struct *shmd); static void shm_close (struct vm_area_struct *shmd);
static void shm_destroy (struct ipc_namespace *ns, struct shmid_kernel *shp);
#ifdef CONFIG_PROC_FS #ifdef CONFIG_PROC_FS
static int sysvipc_shm_proc_show(struct seq_file *s, void *it); static int sysvipc_shm_proc_show(struct seq_file *s, void *it);
#endif #endif
size_t shm_ctlmax = SHMMAX; static void __ipc_init __shm_init_ns(struct ipc_namespace *ns, struct ipc_ids *ids)
size_t shm_ctlall = SHMALL; {
int shm_ctlmni = SHMMNI; ns->ids[IPC_SHM_IDS] = ids;
ns->shm_ctlmax = SHMMAX;
ns->shm_ctlall = SHMALL;
ns->shm_ctlmni = SHMMNI;
ns->shm_tot = 0;
ipc_init_ids(ids, 1);
}
static void do_shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *shp)
{
if (shp->shm_nattch){
shp->shm_perm.mode |= SHM_DEST;
/* Do not find it any more */
shp->shm_perm.key = IPC_PRIVATE;
shm_unlock(shp);
} else
shm_destroy(ns, shp);
}
#ifdef CONFIG_IPC_NS
int shm_init_ns(struct ipc_namespace *ns)
{
struct ipc_ids *ids;
ids = kmalloc(sizeof(struct ipc_ids), GFP_KERNEL);
if (ids == NULL)
return -ENOMEM;
static int shm_tot; /* total number of shared memory pages */ __shm_init_ns(ns, ids);
return 0;
}
void shm_exit_ns(struct ipc_namespace *ns)
{
int i;
struct shmid_kernel *shp;
mutex_lock(&shm_ids(ns).mutex);
for (i = 0; i <= shm_ids(ns).max_id; i++) {
shp = shm_lock(ns, i);
if (shp == NULL)
continue;
do_shm_rmid(ns, shp);
}
mutex_unlock(&shm_ids(ns).mutex);
kfree(ns->ids[IPC_SHM_IDS]);
ns->ids[IPC_SHM_IDS] = NULL;
}
#endif
void __init shm_init (void) void __init shm_init (void)
{ {
ipc_init_ids(&shm_ids, 1); __shm_init_ns(&init_ipc_ns, &init_shm_ids);
ipc_init_proc_interface("sysvipc/shm", ipc_init_proc_interface("sysvipc/shm",
" key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime\n", " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime\n",
&shm_ids, IPC_SHM_IDS, sysvipc_shm_proc_show);
sysvipc_shm_proc_show);
} }
static inline int shm_checkid(struct shmid_kernel *s, int id) static inline int shm_checkid(struct ipc_namespace *ns,
struct shmid_kernel *s, int id)
{ {
if (ipc_checkid(&shm_ids,&s->shm_perm,id)) if (ipc_checkid(&shm_ids(ns), &s->shm_perm, id))
return -EIDRM; return -EIDRM;
return 0; return 0;
} }
static inline struct shmid_kernel *shm_rmid(int id) static inline struct shmid_kernel *shm_rmid(struct ipc_namespace *ns, int id)
{ {
return (struct shmid_kernel *)ipc_rmid(&shm_ids,id); return (struct shmid_kernel *)ipc_rmid(&shm_ids(ns), id);
} }
static inline int shm_addid(struct shmid_kernel *shp) static inline int shm_addid(struct ipc_namespace *ns, struct shmid_kernel *shp)
{ {
return ipc_addid(&shm_ids, &shp->shm_perm, shm_ctlmni); return ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
} }
static inline void shm_inc (int id) { static inline void shm_inc(struct ipc_namespace *ns, int id)
{
struct shmid_kernel *shp; struct shmid_kernel *shp;
shp = shm_lock(id); shp = shm_lock(ns, id);
BUG_ON(!shp); BUG_ON(!shp);
shp->shm_atim = get_seconds(); shp->shm_atim = get_seconds();
shp->shm_lprid = current->tgid; shp->shm_lprid = current->tgid;
...@@ -100,10 +161,13 @@ static inline void shm_inc (int id) { ...@@ -100,10 +161,13 @@ static inline void shm_inc (int id) {
shm_unlock(shp); shm_unlock(shp);
} }
#define shm_file_ns(file) (*((struct ipc_namespace **)&(file)->private_data))
/* This is called by fork, once for every shm attach. */ /* This is called by fork, once for every shm attach. */
static void shm_open (struct vm_area_struct *shmd) static void shm_open(struct vm_area_struct *shmd)
{ {
shm_inc (shmd->vm_file->f_dentry->d_inode->i_ino); shm_inc(shm_file_ns(shmd->vm_file),
shmd->vm_file->f_dentry->d_inode->i_ino);
} }
/* /*
...@@ -114,10 +178,10 @@ static void shm_open (struct vm_area_struct *shmd) ...@@ -114,10 +178,10 @@ static void shm_open (struct vm_area_struct *shmd)
* It has to be called with shp and shm_ids.mutex locked, * It has to be called with shp and shm_ids.mutex locked,
* but returns with shp unlocked and freed. * but returns with shp unlocked and freed.
*/ */
static void shm_destroy (struct shmid_kernel *shp) static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
{ {
shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
shm_rmid (shp->id); shm_rmid(ns, shp->id);
shm_unlock(shp); shm_unlock(shp);
if (!is_file_hugepages(shp->shm_file)) if (!is_file_hugepages(shp->shm_file))
shmem_lock(shp->shm_file, 0, shp->mlock_user); shmem_lock(shp->shm_file, 0, shp->mlock_user);
...@@ -140,20 +204,23 @@ static void shm_close (struct vm_area_struct *shmd) ...@@ -140,20 +204,23 @@ static void shm_close (struct vm_area_struct *shmd)
struct file * file = shmd->vm_file; struct file * file = shmd->vm_file;
int id = file->f_dentry->d_inode->i_ino; int id = file->f_dentry->d_inode->i_ino;
struct shmid_kernel *shp; struct shmid_kernel *shp;
struct ipc_namespace *ns;
mutex_lock(&shm_ids.mutex); ns = shm_file_ns(file);
mutex_lock(&shm_ids(ns).mutex);
/* remove from the list of attaches of the shm segment */ /* remove from the list of attaches of the shm segment */
shp = shm_lock(id); shp = shm_lock(ns, id);
BUG_ON(!shp); BUG_ON(!shp);
shp->shm_lprid = current->tgid; shp->shm_lprid = current->tgid;
shp->shm_dtim = get_seconds(); shp->shm_dtim = get_seconds();
shp->shm_nattch--; shp->shm_nattch--;
if(shp->shm_nattch == 0 && if(shp->shm_nattch == 0 &&
shp->shm_perm.mode & SHM_DEST) shp->shm_perm.mode & SHM_DEST)
shm_destroy (shp); shm_destroy(ns, shp);
else else
shm_unlock(shp); shm_unlock(shp);
mutex_unlock(&shm_ids.mutex); mutex_unlock(&shm_ids(ns).mutex);
} }
static int shm_mmap(struct file * file, struct vm_area_struct * vma) static int shm_mmap(struct file * file, struct vm_area_struct * vma)
...@@ -165,14 +232,25 @@ static int shm_mmap(struct file * file, struct vm_area_struct * vma) ...@@ -165,14 +232,25 @@ static int shm_mmap(struct file * file, struct vm_area_struct * vma)
vma->vm_ops = &shm_vm_ops; vma->vm_ops = &shm_vm_ops;
if (!(vma->vm_flags & VM_WRITE)) if (!(vma->vm_flags & VM_WRITE))
vma->vm_flags &= ~VM_MAYWRITE; vma->vm_flags &= ~VM_MAYWRITE;
shm_inc(file->f_dentry->d_inode->i_ino); shm_inc(shm_file_ns(file), file->f_dentry->d_inode->i_ino);
} }
return ret; return ret;
} }
static int shm_release(struct inode *ino, struct file *file)
{
struct ipc_namespace *ns;
ns = shm_file_ns(file);
put_ipc_ns(ns);
shm_file_ns(file) = NULL;
return 0;
}
static struct file_operations shm_file_operations = { static struct file_operations shm_file_operations = {
.mmap = shm_mmap, .mmap = shm_mmap,
.release = shm_release,
#ifndef CONFIG_MMU #ifndef CONFIG_MMU
.get_unmapped_area = shmem_get_unmapped_area, .get_unmapped_area = shmem_get_unmapped_area,
#endif #endif
...@@ -188,7 +266,7 @@ static struct vm_operations_struct shm_vm_ops = { ...@@ -188,7 +266,7 @@ static struct vm_operations_struct shm_vm_ops = {
#endif #endif
}; };
static int newseg (key_t key, int shmflg, size_t size) static int newseg (struct ipc_namespace *ns, key_t key, int shmflg, size_t size)
{ {
int error; int error;
struct shmid_kernel *shp; struct shmid_kernel *shp;
...@@ -197,10 +275,10 @@ static int newseg (key_t key, int shmflg, size_t size) ...@@ -197,10 +275,10 @@ static int newseg (key_t key, int shmflg, size_t size)
char name[13]; char name[13];
int id; int id;
if (size < SHMMIN || size > shm_ctlmax) if (size < SHMMIN || size > ns->shm_ctlmax)
return -EINVAL; return -EINVAL;
if (shm_tot + numpages >= shm_ctlall) if (ns->shm_tot + numpages >= ns->shm_ctlall)
return -ENOSPC; return -ENOSPC;
shp = ipc_rcu_alloc(sizeof(*shp)); shp = ipc_rcu_alloc(sizeof(*shp));
...@@ -239,7 +317,7 @@ static int newseg (key_t key, int shmflg, size_t size) ...@@ -239,7 +317,7 @@ static int newseg (key_t key, int shmflg, size_t size)
goto no_file; goto no_file;
error = -ENOSPC; error = -ENOSPC;
id = shm_addid(shp); id = shm_addid(ns, shp);
if(id == -1) if(id == -1)
goto no_id; goto no_id;
...@@ -249,15 +327,17 @@ static int newseg (key_t key, int shmflg, size_t size) ...@@ -249,15 +327,17 @@ static int newseg (key_t key, int shmflg, size_t size)
shp->shm_ctim = get_seconds(); shp->shm_ctim = get_seconds();
shp->shm_segsz = size; shp->shm_segsz = size;
shp->shm_nattch = 0; shp->shm_nattch = 0;
shp->id = shm_buildid(id,shp->shm_perm.seq); shp->id = shm_buildid(ns, id, shp->shm_perm.seq);
shp->shm_file = file; shp->shm_file = file;
file->f_dentry->d_inode->i_ino = shp->id; file->f_dentry->d_inode->i_ino = shp->id;
shm_file_ns(file) = get_ipc_ns(ns);
/* Hugetlb ops would have already been assigned. */ /* Hugetlb ops would have already been assigned. */
if (!(shmflg & SHM_HUGETLB)) if (!(shmflg & SHM_HUGETLB))
file->f_op = &shm_file_operations; file->f_op = &shm_file_operations;
shm_tot += numpages; ns->shm_tot += numpages;
shm_unlock(shp); shm_unlock(shp);
return shp->id; return shp->id;
...@@ -273,33 +353,36 @@ asmlinkage long sys_shmget (key_t key, size_t size, int shmflg) ...@@ -273,33 +353,36 @@ asmlinkage long sys_shmget (key_t key, size_t size, int shmflg)
{ {
struct shmid_kernel *shp; struct shmid_kernel *shp;
int err, id = 0; int err, id = 0;
struct ipc_namespace *ns;
ns = current->nsproxy->ipc_ns;
mutex_lock(&shm_ids.mutex); mutex_lock(&shm_ids(ns).mutex);
if (key == IPC_PRIVATE) { if (key == IPC_PRIVATE) {
err = newseg(key, shmflg, size); err = newseg(ns, key, shmflg, size);
} else if ((id = ipc_findkey(&shm_ids, key)) == -1) { } else if ((id = ipc_findkey(&shm_ids(ns), key)) == -1) {
if (!(shmflg & IPC_CREAT)) if (!(shmflg & IPC_CREAT))
err = -ENOENT; err = -ENOENT;
else else
err = newseg(key, shmflg, size); err = newseg(ns, key, shmflg, size);
} else if ((shmflg & IPC_CREAT) && (shmflg & IPC_EXCL)) { } else if ((shmflg & IPC_CREAT) && (shmflg & IPC_EXCL)) {
err = -EEXIST; err = -EEXIST;
} else { } else {
shp = shm_lock(id); shp = shm_lock(ns, id);
BUG_ON(shp==NULL); BUG_ON(shp==NULL);
if (shp->shm_segsz < size) if (shp->shm_segsz < size)
err = -EINVAL; err = -EINVAL;
else if (ipcperms(&shp->shm_perm, shmflg)) else if (ipcperms(&shp->shm_perm, shmflg))
err = -EACCES; err = -EACCES;
else { else {
int shmid = shm_buildid(id, shp->shm_perm.seq); int shmid = shm_buildid(ns, id, shp->shm_perm.seq);
err = security_shm_associate(shp, shmflg); err = security_shm_associate(shp, shmflg);
if (!err) if (!err)
err = shmid; err = shmid;
} }
shm_unlock(shp); shm_unlock(shp);
} }
mutex_unlock(&shm_ids.mutex); mutex_unlock(&shm_ids(ns).mutex);
return err; return err;
} }
...@@ -395,18 +478,19 @@ static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminf ...@@ -395,18 +478,19 @@ static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminf
} }
} }
static void shm_get_stat(unsigned long *rss, unsigned long *swp) static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss,
unsigned long *swp)
{ {
int i; int i;
*rss = 0; *rss = 0;
*swp = 0; *swp = 0;
for (i = 0; i <= shm_ids.max_id; i++) { for (i = 0; i <= shm_ids(ns).max_id; i++) {
struct shmid_kernel *shp; struct shmid_kernel *shp;
struct inode *inode; struct inode *inode;
shp = shm_get(i); shp = shm_get(ns, i);
if(!shp) if(!shp)
continue; continue;
...@@ -430,6 +514,7 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) ...@@ -430,6 +514,7 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
struct shm_setbuf setbuf; struct shm_setbuf setbuf;
struct shmid_kernel *shp; struct shmid_kernel *shp;
int err, version; int err, version;
struct ipc_namespace *ns;
if (cmd < 0 || shmid < 0) { if (cmd < 0 || shmid < 0) {
err = -EINVAL; err = -EINVAL;
...@@ -437,6 +522,7 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) ...@@ -437,6 +522,7 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
} }
version = ipc_parse_version(&cmd); version = ipc_parse_version(&cmd);
ns = current->nsproxy->ipc_ns;
switch (cmd) { /* replace with proc interface ? */ switch (cmd) { /* replace with proc interface ? */
case IPC_INFO: case IPC_INFO:
...@@ -448,15 +534,15 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) ...@@ -448,15 +534,15 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
return err; return err;
memset(&shminfo,0,sizeof(shminfo)); memset(&shminfo,0,sizeof(shminfo));
shminfo.shmmni = shminfo.shmseg = shm_ctlmni; shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni;
shminfo.shmmax = shm_ctlmax; shminfo.shmmax = ns->shm_ctlmax;
shminfo.shmall = shm_ctlall; shminfo.shmall = ns->shm_ctlall;
shminfo.shmmin = SHMMIN; shminfo.shmmin = SHMMIN;
if(copy_shminfo_to_user (buf, &shminfo, version)) if(copy_shminfo_to_user (buf, &shminfo, version))
return -EFAULT; return -EFAULT;
/* reading a integer is always atomic */ /* reading a integer is always atomic */
err= shm_ids.max_id; err= shm_ids(ns).max_id;
if(err<0) if(err<0)
err = 0; err = 0;
goto out; goto out;
...@@ -470,14 +556,14 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) ...@@ -470,14 +556,14 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
return err; return err;
memset(&shm_info,0,sizeof(shm_info)); memset(&shm_info,0,sizeof(shm_info));
mutex_lock(&shm_ids.mutex); mutex_lock(&shm_ids(ns).mutex);
shm_info.used_ids = shm_ids.in_use; shm_info.used_ids = shm_ids(ns).in_use;
shm_get_stat (&shm_info.shm_rss, &shm_info.shm_swp); shm_get_stat (ns, &shm_info.shm_rss, &shm_info.shm_swp);
shm_info.shm_tot = shm_tot; shm_info.shm_tot = ns->shm_tot;
shm_info.swap_attempts = 0; shm_info.swap_attempts = 0;
shm_info.swap_successes = 0; shm_info.swap_successes = 0;
err = shm_ids.max_id; err = shm_ids(ns).max_id;
mutex_unlock(&shm_ids.mutex); mutex_unlock(&shm_ids(ns).mutex);
if(copy_to_user (buf, &shm_info, sizeof(shm_info))) { if(copy_to_user (buf, &shm_info, sizeof(shm_info))) {
err = -EFAULT; err = -EFAULT;
goto out; goto out;
...@@ -492,17 +578,17 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) ...@@ -492,17 +578,17 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
struct shmid64_ds tbuf; struct shmid64_ds tbuf;
int result; int result;
memset(&tbuf, 0, sizeof(tbuf)); memset(&tbuf, 0, sizeof(tbuf));
shp = shm_lock(shmid); shp = shm_lock(ns, shmid);
if(shp==NULL) { if(shp==NULL) {
err = -EINVAL; err = -EINVAL;
goto out; goto out;
} else if(cmd==SHM_STAT) { } else if(cmd==SHM_STAT) {
err = -EINVAL; err = -EINVAL;
if (shmid > shm_ids.max_id) if (shmid > shm_ids(ns).max_id)
goto out_unlock; goto out_unlock;
result = shm_buildid(shmid, shp->shm_perm.seq); result = shm_buildid(ns, shmid, shp->shm_perm.seq);
} else { } else {
err = shm_checkid(shp,shmid); err = shm_checkid(ns, shp,shmid);
if(err) if(err)
goto out_unlock; goto out_unlock;
result = 0; result = 0;
...@@ -534,12 +620,12 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) ...@@ -534,12 +620,12 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
case SHM_LOCK: case SHM_LOCK:
case SHM_UNLOCK: case SHM_UNLOCK:
{ {
shp = shm_lock(shmid); shp = shm_lock(ns, shmid);
if(shp==NULL) { if(shp==NULL) {
err = -EINVAL; err = -EINVAL;
goto out; goto out;
} }
err = shm_checkid(shp,shmid); err = shm_checkid(ns, shp,shmid);
if(err) if(err)
goto out_unlock; goto out_unlock;
...@@ -590,12 +676,12 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf) ...@@ -590,12 +676,12 @@ asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
* Instead we set a destroyed flag, and then blow * Instead we set a destroyed flag, and then blow
* the name away when the usage hits zero. * the name away when the usage hits zero.
*/ */
mutex_lock(&shm_ids.mutex); mutex_lock(&shm_ids(ns).mutex);
shp = shm_lock(shmid); shp = shm_lock(ns, shmid);
err = -EINVAL; err = -EINVAL;