[Devel] [PATCH RFC] c/r mounts ns. Except not really.
Serge E. Hallyn
serue at us.ibm.com
Mon Dec 14 18:39:45 PST 2009
We only c/r a mounts ns with objref 0, meaning inherit the existing
mounts ns. We do intend to implement c/r of mounts and mounts
namespaces in the kernel. It shouldn't be ugly or complicate locking
to do so. Just haven't gotten around to it.
Why did I bother with this? Because we can't re-create private
mounts yet, and while I don't expect trouble doing so, I think
it's more than we want to take on now for v19. But I'd like
as much as possible for everything which we don't support, to not
be checkpointable, since not doing so has in the past invited
slanderous accusations of being a toy implementation :)
Comments?
Signed-off-by: Serge E. Hallyn <serue at us.ibm.com>
---
checkpoint/objhash.c | 26 ++++++++++++++++++++++++++
fs/namespace.c | 35 +++++++++++++++++++++++++++++++++++
include/linux/checkpoint.h | 2 +-
include/linux/checkpoint_hdr.h | 10 ++++++++++
include/linux/mnt_namespace.h | 8 ++++++++
kernel/nsproxy.c | 33 +++++++++++++++++++++++++++++++--
6 files changed, 111 insertions(+), 3 deletions(-)
diff --git a/checkpoint/objhash.c b/checkpoint/objhash.c
index 20fd3e9..b379809 100644
--- a/checkpoint/objhash.c
+++ b/checkpoint/objhash.c
@@ -19,6 +19,7 @@
#include <linux/kref.h>
#include <linux/ipc_namespace.h>
#include <linux/user_namespace.h>
+#include <linux/mnt_namespace.h>
#include <linux/checkpoint.h>
#include <linux/checkpoint_hdr.h>
#include <net/sock.h>
@@ -236,6 +237,22 @@ static void obj_groupinfo_drop(void *ptr, int lastref)
put_group_info((struct group_info *) ptr);
}
+static int obj_mntns_grab(void *ptr)
+{
+ get_mnt_ns((struct mnt_namespace *) ptr);
+ return 0;
+}
+
+static void obj_mntns_drop(void *ptr, int lastref)
+{
+ put_mnt_ns((struct mnt_namespace *) ptr);
+}
+
+static int obj_mntns_users(void *ptr)
+{
+ return atomic_read(&((struct mnt_namespace *) ptr)->count);
+}
+
static int obj_sock_grab(void *ptr)
{
sock_hold((struct sock *) ptr);
@@ -500,6 +517,15 @@ static struct ckpt_obj_ops ckpt_obj_ops[] = {
.checkpoint = checkpoint_lsm_string,
.restore = restore_lsm_string_wrap,
},
+ {
+ .obj_name = "MOUNTS NS",
+ .obj_type = CKPT_OBJ_MNT_NS,
+ .ref_grab = obj_mntns_grab,
+ .ref_drop = obj_mntns_drop,
+ .ref_users = obj_mntns_users,
+ .checkpoint = checkpoint_mnt_ns,
+ .restore = restore_mnt_ns,
+ },
};
diff --git a/fs/namespace.c b/fs/namespace.c
index bdc3cb4..3a61197 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -29,6 +29,7 @@
#include <linux/log2.h>
#include <linux/idr.h>
#include <linux/fs_struct.h>
+#include <linux/checkpoint.h>
#include <asm/uaccess.h>
#include <asm/unistd.h>
#include "pnode.h"
@@ -2323,3 +2324,37 @@ void put_mnt_ns(struct mnt_namespace *ns)
kfree(ns);
}
EXPORT_SYMBOL(put_mnt_ns);
+
+#ifdef CONFIG_CHECKPOINT
+int checkpoint_mounts_ns(struct ckpt_ctx *ctx, struct mnt_namespace *mnt_ns)
+{
+ if (mnt_ns == ctx->root_nsproxy->mnt_ns)
+ return 0;
+ return checkpoint_obj(ctx, mnt_ns, CKPT_OBJ_MNT_NS);
+}
+/*
+ * c/r with private mounts namespaces is unsupported. We'll need
+ * to also support c/r of mounts to meaningfully do so.
+ */
+static int do_checkpoint_mnt_ns(struct ckpt_ctx *ctx,
+ struct mnt_namespace *mnt_ns)
+{
+ return -EINVAL;
+}
+
+int checkpoint_mnt_ns(struct ckpt_ctx *ctx, void *ptr)
+{
+ return do_checkpoint_mnt_ns(ctx, (struct mnt_namespace *) ptr);
+}
+
+static struct mnt_namespace *do_restore_mnt_ns(struct ckpt_ctx *ctx)
+{
+ ckpt_err(ctx, -EINVAL, "private mounts namespace unsupported\n");
+ return ERR_PTR(-EINVAL);
+}
+
+void *restore_mnt_ns(struct ckpt_ctx *ctx)
+{
+ return (void *) do_restore_mnt_ns(ctx);
+}
+#endif
diff --git a/include/linux/checkpoint.h b/include/linux/checkpoint.h
index 92d47fa..93bdbdd 100644
--- a/include/linux/checkpoint.h
+++ b/include/linux/checkpoint.h
@@ -10,7 +10,7 @@
* distribution for more details.
*/
-#define CHECKPOINT_VERSION 5
+#define CHECKPOINT_VERSION 6
/* checkpoint user flags */
#define CHECKPOINT_SUBTREE 0x1
diff --git a/include/linux/checkpoint_hdr.h b/include/linux/checkpoint_hdr.h
index f07e527..cdc6531 100644
--- a/include/linux/checkpoint_hdr.h
+++ b/include/linux/checkpoint_hdr.h
@@ -117,6 +117,8 @@ enum {
#define CKPT_HDR_GROUPINFO CKPT_HDR_GROUPINFO
CKPT_HDR_TASK_CREDS,
#define CKPT_HDR_TASK_CREDS CKPT_HDR_TASK_CREDS
+ CKPT_HDR_MNT_NS,
+#define CKPT_HDR_MNT_NS CKPT_HDR_MNT_NS
/* 201-299: reserved for arch-dependent */
@@ -246,6 +248,8 @@ enum obj_type {
#define CKPT_OBJ_SECURITY_PTR CKPT_OBJ_SECURITY_PTR
CKPT_OBJ_SECURITY,
#define CKPT_OBJ_SECURITY CKPT_OBJ_SECURITY
+ CKPT_OBJ_MNT_NS,
+#define CKPT_OBJ_MNT_NS CKPT_OBJ_MNT_NS
CKPT_OBJ_MAX
#define CKPT_OBJ_MAX CKPT_OBJ_MAX
};
@@ -428,10 +432,16 @@ struct ckpt_hdr_task_ns {
__s32 ns_objref;
} __attribute__((aligned(8)));
+/*
+ * mnt_objref is 0 if shared with ctx->root_task->parent's mntns
+ * Come up with a better name?
+ */
+#define CKPT_MNT_NS_INHERIT 0
struct ckpt_hdr_ns {
struct ckpt_hdr h;
__s32 uts_objref;
__s32 ipc_objref;
+ __s32 mnt_objref;
} __attribute__((aligned(8)));
/* cannot include <linux/tty.h> from userspace, so define: */
diff --git a/include/linux/mnt_namespace.h b/include/linux/mnt_namespace.h
index d74785c..ad242bb 100644
--- a/include/linux/mnt_namespace.h
+++ b/include/linux/mnt_namespace.h
@@ -32,6 +32,14 @@ static inline void get_mnt_ns(struct mnt_namespace *ns)
atomic_inc(&ns->count);
}
+#ifdef CONFIG_CHECKPOINT
+struct ckpt_ctx;
+extern void *restore_mnt_ns(struct ckpt_ctx *ctx);
+extern int checkpoint_mnt_ns(struct ckpt_ctx *ctx, void *ptr);
+extern int checkpoint_mounts_ns(struct ckpt_ctx *ctx,
+ struct mnt_namespace *mnt_ns);
+#endif
+
extern const struct seq_operations mounts_op;
extern const struct seq_operations mountinfo_op;
extern const struct seq_operations mountstats_op;
diff --git a/kernel/nsproxy.c b/kernel/nsproxy.c
index e7aaa00..c91b725 100644
--- a/kernel/nsproxy.c
+++ b/kernel/nsproxy.c
@@ -282,6 +282,13 @@ static int do_checkpoint_ns(struct ckpt_ctx *ctx, struct nsproxy *nsproxy)
goto out;
h->ipc_objref = ret;
+ ret = checkpoint_mounts_ns(ctx, nsproxy->mnt_ns);
+ if (ret < 0) {
+ ckpt_err(ctx, ret, "%(T)process mounts ns\n");
+ goto out;
+ }
+ h->mnt_objref = ret;
+
/* TODO: Write other namespaces here */
ret = ckpt_write_obj(ctx, &h->h);
@@ -302,6 +309,7 @@ static struct nsproxy *do_restore_ns(struct ckpt_ctx *ctx)
struct nsproxy *nsproxy = NULL;
struct uts_namespace *uts_ns;
struct ipc_namespace *ipc_ns;
+ struct mnt_namespace *mnt_ns = NULL;
int ret;
h = ckpt_read_obj_type(ctx, sizeof(*h), CKPT_HDR_NS);
@@ -323,6 +331,25 @@ static struct nsproxy *do_restore_ns(struct ckpt_ctx *ctx)
ret = PTR_ERR(ipc_ns);
goto out;
}
+ if (h->mnt_objref == CKPT_MNT_NS_INHERIT)
+ /* inherit from root task.
+ * XXX And right here we come upon a dilemma. Note that by
+ * doing that, we wouldn't allow userspace to recreate mounts
+ * namespace and mounts between being cloned() and running
+ * sys_restart().
+ *
+ * SO, we can claim that if userspace does nothing fancy, then
+ * current->nsproxy->mnt_ns() will be root task mnt_ns. If
+ * userspace does, then it clearly wants to set up mounts, so
+ * we let it.
+ */
+ mnt_ns = current->nsproxy->mnt_ns;
+ else
+ mnt_ns = ckpt_obj_fetch(ctx, h->mnt_objref, CKPT_OBJ_MNT_NS);
+ if (IS_ERR(mnt_ns)) {
+ ret = PTR_ERR(mnt_ns);
+ goto out;
+ }
#if defined(COFNIG_UTS_NS) || defined(CONFIG_IPC_NS)
ret = -ENOMEM;
@@ -335,10 +362,9 @@ static struct nsproxy *do_restore_ns(struct ckpt_ctx *ctx)
get_ipc_ns(ipc_ns);
nsproxy->ipc_ns = ipc_ns;
+ /* unsupported namespaces: */
get_pid_ns(current->nsproxy->pid_ns);
nsproxy->pid_ns = current->nsproxy->pid_ns;
- get_mnt_ns(current->nsproxy->mnt_ns);
- nsproxy->mnt_ns = current->nsproxy->mnt_ns;
get_net(current->nsproxy->net_ns);
nsproxy->net_ns = current->nsproxy->net_ns;
#else
@@ -349,6 +375,9 @@ static struct nsproxy *do_restore_ns(struct ckpt_ctx *ctx)
BUG_ON(nsproxy->ipc_ns != ipc_ns);
#endif
+ get_mnt_ns(mnt_ns);
+ nsproxy->mnt_ns = mnt_ns;
+
/* TODO: add more namespaces here */
ret = 0;
out:
--
1.6.0.4
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers
More information about the Devel
mailing list