drbd: switch configuration interface from connector to genetlink
authorLars Ellenberg <lars.ellenberg@linbit.com>
Mon, 7 Mar 2011 11:49:34 +0000 (12:49 +0100)
committerPhilipp Reisner <philipp.reisner@linbit.com>
Sat, 3 Nov 2012 23:16:17 +0000 (00:16 +0100)
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
drivers/block/drbd/drbd_actlog.c
drivers/block/drbd/drbd_int.h
drivers/block/drbd/drbd_main.c
drivers/block/drbd/drbd_nl.c
drivers/block/drbd/drbd_state.c
include/linux/drbd.h
include/linux/genl_magic_func.h

index 7cd7861..c1a9061 100644 (file)
@@ -702,6 +702,7 @@ static int w_update_odbm(struct drbd_work *w, int unused)
 {
        struct update_odbm_work *udw = container_of(w, struct update_odbm_work, w);
        struct drbd_conf *mdev = w->mdev;
+       struct sib_info sib = { .sib_reason = SIB_SYNC_PROGRESS, };
 
        if (!get_ldev(mdev)) {
                if (__ratelimit(&drbd_ratelimit_state))
@@ -725,7 +726,7 @@ static int w_update_odbm(struct drbd_work *w, int unused)
                        break;
                }
        }
-       drbd_bcast_sync_progress(mdev);
+       drbd_bcast_event(mdev, &sib);
 
        return 1;
 }
index e687583..429fd8d 100644 (file)
@@ -44,6 +44,7 @@
 #include <net/tcp.h>
 #include <linux/lru_cache.h>
 #include <linux/prefetch.h>
+#include <linux/drbd_genl_api.h>
 #include <linux/drbd.h>
 #include "drbd_state.h"
 
@@ -65,7 +66,6 @@
 extern unsigned int minor_count;
 extern int disable_sendpage;
 extern int allow_oos;
-extern unsigned int cn_idx;
 
 #ifdef CONFIG_DRBD_FAULT_INJECTION
 extern int enable_faults;
@@ -865,14 +865,6 @@ struct drbd_md {
         */
 };
 
-/* for sync_conf and other types... */
-#define NL_PACKET(name, number, fields) struct name { fields };
-#define NL_INTEGER(pn,pr,member) int member;
-#define NL_INT64(pn,pr,member) __u64 member;
-#define NL_BIT(pn,pr,member)   unsigned member:1;
-#define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
-#include "linux/drbd_nl.h"
-
 struct drbd_backing_dev {
        struct block_device *backing_bdev;
        struct block_device *md_bdev;
@@ -1502,7 +1494,7 @@ enum drbd_ret_code conn_new_minor(struct drbd_tconn *tconn, unsigned int minor,
 extern void drbd_free_mdev(struct drbd_conf *mdev);
 extern void drbd_delete_device(unsigned int minor);
 
-struct drbd_tconn *drbd_new_tconn(char *name);
+struct drbd_tconn *drbd_new_tconn(const char *name);
 extern void drbd_free_tconn(struct drbd_tconn *tconn);
 struct drbd_tconn *conn_by_name(const char *name);
 
@@ -1679,16 +1671,22 @@ extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
 extern void drbd_al_shrink(struct drbd_conf *mdev);
 
-
 /* drbd_nl.c */
-
-void drbd_nl_cleanup(void);
-int __init drbd_nl_init(void);
-void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
-void drbd_bcast_sync_progress(struct drbd_conf *mdev);
-void drbd_bcast_ee(struct drbd_conf *, const char *, const int, const char *,
-                  const char *, const struct drbd_peer_request *);
-
+/* state info broadcast */
+struct sib_info {
+       enum drbd_state_info_bcast_reason sib_reason;
+       union {
+               struct {
+                       char *helper_name;
+                       unsigned helper_exit_code;
+               };
+               struct {
+                       union drbd_state os;
+                       union drbd_state ns;
+               };
+       };
+};
+void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib);
 
 /*
  * inline helper functions
index 9f6db59..9697ab8 100644 (file)
@@ -86,7 +86,6 @@ MODULE_PARM_DESC(allow_oos, "DONT USE!");
 module_param(minor_count, uint, 0444);
 module_param(disable_sendpage, bool, 0644);
 module_param(allow_oos, bool, 0);
-module_param(cn_idx, uint, 0444);
 module_param(proc_details, int, 0644);
 
 #ifdef CONFIG_DRBD_FAULT_INJECTION
@@ -108,7 +107,6 @@ module_param(fault_devs, int, 0644);
 unsigned int minor_count = DRBD_MINOR_COUNT_DEF;
 int disable_sendpage;
 int allow_oos;
-unsigned int cn_idx = CN_IDX_DRBD;
 int proc_details;       /* Detail level in proc drbd*/
 
 /* Module parameter for setting the user mode helper program
@@ -2175,7 +2173,7 @@ static void drbd_cleanup(void)
        if (drbd_proc)
                remove_proc_entry("drbd", NULL);
 
-       drbd_nl_cleanup();
+       drbd_genl_unregister();
 
        idr_for_each_entry(&minors, mdev, i)
                drbd_delete_device(i);
@@ -2237,6 +2235,9 @@ struct drbd_tconn *conn_by_name(const char *name)
 {
        struct drbd_tconn *tconn;
 
+       if (!name || !name[0])
+               return NULL;
+
        write_lock_irq(&global_state_lock);
        list_for_each_entry(tconn, &drbd_tconns, all_tconn) {
                if (!strcmp(tconn->name, name))
@@ -2248,7 +2249,7 @@ found:
        return tconn;
 }
 
-struct drbd_tconn *drbd_new_tconn(char *name)
+struct drbd_tconn *drbd_new_tconn(const char *name)
 {
        struct drbd_tconn *tconn;
 
@@ -2333,6 +2334,7 @@ enum drbd_ret_code conn_new_minor(struct drbd_tconn *tconn, unsigned int minor,
 
        mdev->tconn = tconn;
        mdev->minor = minor;
+       mdev->vnr = vnr;
 
        drbd_init_set_defaults(mdev);
 
@@ -2461,10 +2463,6 @@ int __init drbd_init(void)
 #endif
        }
 
-       err = drbd_nl_init();
-       if (err)
-               return err;
-
        err = register_blkdev(DRBD_MAJOR, "drbd");
        if (err) {
                printk(KERN_ERR
@@ -2473,6 +2471,13 @@ int __init drbd_init(void)
                return err;
        }
 
+       err = drbd_genl_register();
+       if (err) {
+               printk(KERN_ERR "drbd: unable to register generic netlink family\n");
+               goto fail;
+       }
+
+
        register_reboot_notifier(&drbd_notifier);
 
        /*
@@ -2487,12 +2492,12 @@ int __init drbd_init(void)
 
        err = drbd_create_mempools();
        if (err)
-               goto Enomem;
+               goto fail;
 
        drbd_proc = proc_create_data("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops, NULL);
        if (!drbd_proc) {
                printk(KERN_ERR "drbd: unable to register proc file\n");
-               goto Enomem;
+               goto fail;
        }
 
        rwlock_init(&global_state_lock);
@@ -2507,7 +2512,7 @@ int __init drbd_init(void)
 
        return 0; /* Success! */
 
-Enomem:
+fail:
        drbd_cleanup();
        if (err == -ENOMEM)
                /* currently always the case */
index f2739fd..f9be142 100644 (file)
 #include <linux/fs.h>
 #include <linux/file.h>
 #include <linux/slab.h>
-#include <linux/connector.h>
 #include <linux/blkpg.h>
 #include <linux/cpumask.h>
 #include "drbd_int.h"
 #include "drbd_req.h"
 #include "drbd_wrappers.h"
 #include <asm/unaligned.h>
-#include <linux/drbd_tag_magic.h>
 #include <linux/drbd_limits.h>
-#include <linux/compiler.h>
 #include <linux/kthread.h>
 
-static unsigned short *tl_add_blob(unsigned short *, enum drbd_tags, const void *, int);
-static unsigned short *tl_add_str(unsigned short *, enum drbd_tags, const char *);
-static unsigned short *tl_add_int(unsigned short *, enum drbd_tags, const void *);
-
-/* see get_sb_bdev and bd_claim */
+#include <net/genetlink.h>
+
+/* .doit */
+// int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info);
+// int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info);
+
+int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info);
+
+int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info);
+
+int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_syncer(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info);
+int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info);
+/* .dumpit */
+int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb);
+
+#include <linux/drbd_genl_api.h>
+#include <linux/genl_magic_func.h>
+
+/* used blkdev_get_by_path, to claim our meta data device(s) */
 static char *drbd_m_holder = "Hands off! this is DRBD's meta data device.";
 
-/* Generate the tag_list to struct functions */
-#define NL_PACKET(name, number, fields) \
-static int name ## _from_tags( \
-       unsigned short *tags, struct name *arg) __attribute__ ((unused)); \
-static int name ## _from_tags( \
-       unsigned short *tags, struct name *arg) \
-{ \
-       int tag; \
-       int dlen; \
-       \
-       while ((tag = get_unaligned(tags++)) != TT_END) {       \
-               dlen = get_unaligned(tags++);                   \
-               switch (tag_number(tag)) { \
-               fields \
-               default: \
-                       if (tag & T_MANDATORY) { \
-                               printk(KERN_ERR "drbd: Unknown tag: %d\n", tag_number(tag)); \
-                               return 0; \
-                       } \
-               } \
-               tags = (unsigned short *)((char *)tags + dlen); \
-       } \
-       return 1; \
+/* Configuration is strictly serialized, because generic netlink message
+ * processing is strictly serialized by the genl_lock().
+ * Which means we can use one static global drbd_config_context struct.
+ */
+static struct drbd_config_context {
+       /* assigned from drbd_genlmsghdr */
+       unsigned int minor;
+       /* assigned from request attributes, if present */
+       unsigned int volume;
+#define VOLUME_UNSPECIFIED             (-1U)
+       /* pointer into the request skb,
+        * limited lifetime! */
+       char *conn_name;
+
+       /* reply buffer */
+       struct sk_buff *reply_skb;
+       /* pointer into reply buffer */
+       struct drbd_genlmsghdr *reply_dh;
+       /* resolved from attributes, if possible */
+       struct drbd_conf *mdev;
+       struct drbd_tconn *tconn;
+} adm_ctx;
+
+static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info)
+{
+       genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb))));
+       if (genlmsg_reply(skb, info))
+               printk(KERN_ERR "drbd: error sending genl reply\n");
 }
-#define NL_INTEGER(pn, pr, member) \
-       case pn: /* D_ASSERT( tag_type(tag) == TT_INTEGER ); */ \
-               arg->member = get_unaligned((int *)(tags));     \
-               break;
-#define NL_INT64(pn, pr, member) \
-       case pn: /* D_ASSERT( tag_type(tag) == TT_INT64 ); */ \
-               arg->member = get_unaligned((u64 *)(tags));     \
-               break;
-#define NL_BIT(pn, pr, member) \
-       case pn: /* D_ASSERT( tag_type(tag) == TT_BIT ); */ \
-               arg->member = *(char *)(tags) ? 1 : 0; \
-               break;
-#define NL_STRING(pn, pr, member, len) \
-       case pn: /* D_ASSERT( tag_type(tag) == TT_STRING ); */ \
-               if (dlen > len) { \
-                       printk(KERN_ERR "drbd: arg too long: %s (%u wanted, max len: %u bytes)\n", \
-                               #member, dlen, (unsigned int)len); \
-                       return 0; \
-               } \
-                arg->member ## _len = dlen; \
-                memcpy(arg->member, tags, min_t(size_t, dlen, len)); \
-                break;
-#include "linux/drbd_nl.h"
-
-/* Generate the struct to tag_list functions */
-#define NL_PACKET(name, number, fields) \
-static unsigned short* \
-name ## _to_tags( \
-       struct name *arg, unsigned short *tags) __attribute__ ((unused)); \
-static unsigned short* \
-name ## _to_tags( \
-       struct name *arg, unsigned short *tags) \
-{ \
-       fields \
-       return tags; \
+
+/* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only
+ * reason it could fail was no space in skb, and there are 4k available. */
+static int drbd_msg_put_info(const char *info)
+{
+       struct sk_buff *skb = adm_ctx.reply_skb;
+       struct nlattr *nla;
+       int err = -EMSGSIZE;
+
+       if (!info || !info[0])
+               return 0;
+
+       nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY);
+       if (!nla)
+               return err;
+
+       err = nla_put_string(skb, T_info_text, info);
+       if (err) {
+               nla_nest_cancel(skb, nla);
+               return err;
+       } else
+               nla_nest_end(skb, nla);
+       return 0;
 }
 
-#define NL_INTEGER(pn, pr, member) \
-       put_unaligned(pn | pr | TT_INTEGER, tags++);    \
-       put_unaligned(sizeof(int), tags++);             \
-       put_unaligned(arg->member, (int *)tags);        \
-       tags = (unsigned short *)((char *)tags+sizeof(int));
-#define NL_INT64(pn, pr, member) \
-       put_unaligned(pn | pr | TT_INT64, tags++);      \
-       put_unaligned(sizeof(u64), tags++);             \
-       put_unaligned(arg->member, (u64 *)tags);        \
-       tags = (unsigned short *)((char *)tags+sizeof(u64));
-#define NL_BIT(pn, pr, member) \
-       put_unaligned(pn | pr | TT_BIT, tags++);        \
-       put_unaligned(sizeof(char), tags++);            \
-       *(char *)tags = arg->member; \
-       tags = (unsigned short *)((char *)tags+sizeof(char));
-#define NL_STRING(pn, pr, member, len) \
-       put_unaligned(pn | pr | TT_STRING, tags++);     \
-       put_unaligned(arg->member ## _len, tags++);     \
-       memcpy(tags, arg->member, arg->member ## _len); \
-       tags = (unsigned short *)((char *)tags + arg->member ## _len);
-#include "linux/drbd_nl.h"
-
-void drbd_bcast_ev_helper(struct drbd_conf *mdev, char *helper_name);
-void drbd_nl_send_reply(struct cn_msg *, int);
+/* This would be a good candidate for a "pre_doit" hook,
+ * and per-family private info->pointers.
+ * But we need to stay compatible with older kernels.
+ * If it returns successfully, adm_ctx members are valid.
+ */
+#define DRBD_ADM_NEED_MINOR    1
+#define DRBD_ADM_NEED_CONN     2
+static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info,
+               unsigned flags)
+{
+       struct drbd_genlmsghdr *d_in = info->userhdr;
+       const u8 cmd = info->genlhdr->cmd;
+       int err;
+
+       memset(&adm_ctx, 0, sizeof(adm_ctx));
+
+       /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */
+       if (cmd != DRBD_ADM_GET_STATUS
+       && security_netlink_recv(skb, CAP_SYS_ADMIN))
+              return -EPERM;
+
+       adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
+       if (!adm_ctx.reply_skb)
+               goto fail;
+
+       adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb,
+                                       info, &drbd_genl_family, 0, cmd);
+       /* put of a few bytes into a fresh skb of >= 4k will always succeed.
+        * but anyways */
+       if (!adm_ctx.reply_dh)
+               goto fail;
+
+       adm_ctx.reply_dh->minor = d_in->minor;
+       adm_ctx.reply_dh->ret_code = NO_ERROR;
+
+       if (info->attrs[DRBD_NLA_CFG_CONTEXT]) {
+               struct nlattr *nla;
+               /* parse and validate only */
+               err = drbd_cfg_context_from_attrs(NULL, info->attrs);
+               if (err)
+                       goto fail;
+
+               /* It was present, and valid,
+                * copy it over to the reply skb. */
+               err = nla_put_nohdr(adm_ctx.reply_skb,
+                               info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len,
+                               info->attrs[DRBD_NLA_CFG_CONTEXT]);
+               if (err)
+                       goto fail;
+
+               /* and assign stuff to the global adm_ctx */
+               nla = nested_attr_tb[__nla_type(T_ctx_volume)];
+               adm_ctx.volume = nla ? nla_get_u32(nla) : VOLUME_UNSPECIFIED;
+               nla = nested_attr_tb[__nla_type(T_ctx_conn_name)];
+               if (nla)
+                       adm_ctx.conn_name = nla_data(nla);
+       } else
+               adm_ctx.volume = VOLUME_UNSPECIFIED;
+
+       adm_ctx.minor = d_in->minor;
+       adm_ctx.mdev = minor_to_mdev(d_in->minor);
+       adm_ctx.tconn = conn_by_name(adm_ctx.conn_name);
+
+       if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) {
+               drbd_msg_put_info("unknown minor");
+               return ERR_MINOR_INVALID;
+       }
+       if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_CONN)) {
+               drbd_msg_put_info("unknown connection");
+               return ERR_INVALID_REQUEST;
+       }
+
+       /* some more paranoia, if the request was over-determined */
+       if (adm_ctx.mdev &&
+           adm_ctx.volume != VOLUME_UNSPECIFIED &&
+           adm_ctx.volume != adm_ctx.mdev->vnr) {
+               pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n",
+                               adm_ctx.minor, adm_ctx.volume,
+                               adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->name);
+               drbd_msg_put_info("over-determined configuration context mismatch");
+               return ERR_INVALID_REQUEST;
+       }
+       if (adm_ctx.mdev && adm_ctx.tconn &&
+           adm_ctx.mdev->tconn != adm_ctx.tconn) {
+               pr_warning("request: minor=%u, conn=%s; but that minor belongs to connection %s\n",
+                               adm_ctx.minor, adm_ctx.conn_name, adm_ctx.mdev->tconn->name);
+               drbd_msg_put_info("over-determined configuration context mismatch");
+               return ERR_INVALID_REQUEST;
+       }
+       return NO_ERROR;
+
+fail:
+       nlmsg_free(adm_ctx.reply_skb);
+       adm_ctx.reply_skb = NULL;
+       return -ENOMEM;
+}
+
+static int drbd_adm_finish(struct genl_info *info, int retcode)
+{
+       struct nlattr *nla;
+       const char *conn_name = NULL;
+
+       if (!adm_ctx.reply_skb)
+               return -ENOMEM;
+
+       adm_ctx.reply_dh->ret_code = retcode;
+
+       nla = info->attrs[DRBD_NLA_CFG_CONTEXT];
+       if (nla) {
+               nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name));
+               if (nla)
+                       conn_name = nla_data(nla);
+       }
+
+       drbd_adm_send_reply(adm_ctx.reply_skb, info);
+       return 0;
+}
 
 int drbd_khelper(struct drbd_conf *mdev, char *cmd)
 {
@@ -142,9 +257,9 @@ int drbd_khelper(struct drbd_conf *mdev, char *cmd)
                        NULL, /* Will be set to address family */
                        NULL, /* Will be set to address */
                        NULL };
-
        char mb[12], af[20], ad[60], *afs;
        char *argv[] = {usermode_helper, cmd, mb, NULL };
+       struct sib_info sib;
        int ret;
 
        snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev));
@@ -177,8 +292,9 @@ int drbd_khelper(struct drbd_conf *mdev, char *cmd)
        drbd_md_sync(mdev);
 
        dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb);
-
-       drbd_bcast_ev_helper(mdev, cmd);
+       sib.sib_reason = SIB_HELPER_PRE;
+       sib.helper_name = cmd;
+       drbd_bcast_event(mdev, &sib);
        ret = call_usermodehelper(usermode_helper, argv, envp, 1);
        if (ret)
                dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
@@ -188,6 +304,9 @@ int drbd_khelper(struct drbd_conf *mdev, char *cmd)
                dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
                                usermode_helper, cmd, mb,
                                (ret >> 8) & 0xff, ret);
+       sib.sib_reason = SIB_HELPER_POST;
+       sib.helper_exit_code = ret;
+       drbd_bcast_event(mdev, &sib);
 
        if (ret < 0) /* Ignore any ERRNOs we got. */
                ret = 0;
@@ -362,7 +481,7 @@ drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force)
                }
 
                if (rv == SS_NOTHING_TO_DO)
-                       goto fail;
+                       goto out;
                if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) {
                        nps = drbd_try_outdate_peer(mdev);
 
@@ -388,13 +507,13 @@ drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force)
                        rv = _drbd_request_state(mdev, mask, val,
                                                CS_VERBOSE + CS_WAIT_COMPLETE);
                        if (rv < SS_SUCCESS)
-                               goto fail;
+                               goto out;
                }
                break;
        }
 
        if (rv < SS_SUCCESS)
-               goto fail;
+               goto out;
 
        if (forced)
                dev_warn(DEV, "Forced to consider local data as UpToDate!\n");
@@ -438,33 +557,46 @@ drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force)
        drbd_md_sync(mdev);
 
        kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
- fail:
+out:
        mutex_unlock(mdev->state_mutex);
        return rv;
 }
 
-static int drbd_nl_primary(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                          struct drbd_nl_cfg_reply *reply)
+static const char *from_attrs_err_to_txt(int err)
 {
-       struct primary primary_args;
-
-       memset(&primary_args, 0, sizeof(struct primary));
-       if (!primary_from_tags(nlp->tag_list, &primary_args)) {
-               reply->ret_code = ERR_MANDATORY_TAG;
-               return 0;
-       }
-
-       reply->ret_code =
-               drbd_set_role(mdev, R_PRIMARY, primary_args.primary_force);
-
-       return 0;
+       return  err == -ENOMSG ? "required attribute missing" :
+               err == -EOPNOTSUPP ? "unknown mandatory attribute" :
+               "invalid attribute value";
 }
 
-static int drbd_nl_secondary(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                            struct drbd_nl_cfg_reply *reply)
+int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info)
 {
-       reply->ret_code = drbd_set_role(mdev, R_SECONDARY, 0);
+       struct set_role_parms parms;
+       int err;
+       enum drbd_ret_code retcode;
 
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
+
+       memset(&parms, 0, sizeof(parms));
+       if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) {
+               err = set_role_parms_from_attrs(&parms, info->attrs);
+               if (err) {
+                       retcode = ERR_MANDATORY_TAG;
+                       drbd_msg_put_info(from_attrs_err_to_txt(err));
+                       goto out;
+               }
+       }
+
+       if (info->genlhdr->cmd == DRBD_ADM_PRIMARY)
+               retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate);
+       else
+               retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0);
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
@@ -541,6 +673,12 @@ char *ppsize(char *buf, unsigned long long size)
  *  R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET:
  *  peer may not initiate a resize.
  */
+/* Note these are not to be confused with
+ * drbd_adm_suspend_io/drbd_adm_resume_io,
+ * which are (sub) state changes triggered by admin (drbdsetup),
+ * and can be long lived.
+ * This changes an mdev->flag, is triggered by drbd internals,
+ * and should be short-lived. */
 void drbd_suspend_io(struct drbd_conf *mdev)
 {
        set_bit(SUSPEND_IO, &mdev->flags);
@@ -881,11 +1019,10 @@ static void drbd_suspend_al(struct drbd_conf *mdev)
                dev_info(DEV, "Suspended AL updates\n");
 }
 
-/* does always return 0;
- * interesting return code is in reply->ret_code */
-static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                            struct drbd_nl_cfg_reply *reply)
+int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info)
 {
+       struct drbd_conf *mdev;
+       int err;
        enum drbd_ret_code retcode;
        enum determine_dev_size dd;
        sector_t max_possible_sectors;
@@ -897,6 +1034,13 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
        enum drbd_state_rv rv;
        int cp_discovered = 0;
 
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto fail;
+
+       mdev = adm_ctx.mdev;
        conn_reconfig_start(mdev->tconn);
 
        /* if you want to reconfigure, please tear down first */
@@ -910,7 +1054,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
         * to realize a "hot spare" feature (not that I'd recommend that) */
        wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
 
-       /* allocation not in the IO path, cqueue thread context */
+       /* allocation not in the IO path, drbdsetup context */
        nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL);
        if (!nbc) {
                retcode = ERR_NOMEM;
@@ -922,12 +1066,14 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
        nbc->dc.fencing       = DRBD_FENCING_DEF;
        nbc->dc.max_bio_bvecs = DRBD_MAX_BIO_BVECS_DEF;
 
-       if (!disk_conf_from_tags(nlp->tag_list, &nbc->dc)) {
+       err = disk_conf_from_attrs(&nbc->dc, info->attrs);
+       if (err) {
                retcode = ERR_MANDATORY_TAG;
+               drbd_msg_put_info(from_attrs_err_to_txt(err));
                goto fail;
        }
 
-       if (nbc->dc.meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) {
+       if ((int)nbc->dc.meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) {
                retcode = ERR_MD_IDX_INVALID;
                goto fail;
        }
@@ -961,7 +1107,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
         */
        bdev = blkdev_get_by_path(nbc->dc.meta_dev,
                                  FMODE_READ | FMODE_WRITE | FMODE_EXCL,
-                                 (nbc->dc.meta_dev_idx < 0) ?
+                                 ((int)nbc->dc.meta_dev_idx < 0) ?
                                  (void *)mdev : (void *)drbd_m_holder);
        if (IS_ERR(bdev)) {
                dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.meta_dev,
@@ -997,7 +1143,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
                goto fail;
        }
 
-       if (nbc->dc.meta_dev_idx < 0) {
+       if ((int)nbc->dc.meta_dev_idx < 0) {
                max_possible_sectors = DRBD_MAX_SECTORS_FLEX;
                /* at least one MB, otherwise it does not make sense */
                min_md_device_sectors = (2<<10);
@@ -1028,7 +1174,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
                dev_warn(DEV, "==> truncating very big lower level device "
                        "to currently maximum possible %llu sectors <==\n",
                        (unsigned long long) max_possible_sectors);
-               if (nbc->dc.meta_dev_idx >= 0)
+               if ((int)nbc->dc.meta_dev_idx >= 0)
                        dev_warn(DEV, "==>> using internal or flexible "
                                      "meta data may help <<==\n");
        }
@@ -1242,8 +1388,8 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
 
        kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
        put_ldev(mdev);
-       reply->ret_code = retcode;
        conn_reconfig_done(mdev->tconn);
+       drbd_adm_finish(info, retcode);
        return 0;
 
  force_diskless_dec:
@@ -1251,6 +1397,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
  force_diskless:
        drbd_force_state(mdev, NS(disk, D_FAILED));
        drbd_md_sync(mdev);
+       conn_reconfig_done(mdev->tconn);
  fail:
        if (nbc) {
                if (nbc->backing_bdev)
@@ -1263,8 +1410,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
        }
        lc_destroy(resync_lru);
 
-       reply->ret_code = retcode;
-       conn_reconfig_done(mdev->tconn);
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
@@ -1273,42 +1419,54 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
  * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
  * internal references as well.
  * Only then we have finally detached. */
-static int drbd_nl_detach(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                         struct drbd_nl_cfg_reply *reply)
+int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info)
 {
+       struct drbd_conf *mdev;
        enum drbd_ret_code retcode;
-       int ret;
+
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
+
+       mdev = adm_ctx.mdev;
        drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */
-       retcode = drbd_request_state(mdev, NS(disk, D_FAILED));
-       /* D_FAILED will transition to DISKLESS. */
-       ret = wait_event_interruptible(mdev->misc_wait,
-                       mdev->state.disk != D_FAILED);
+       retcode = drbd_request_state(mdev, NS(disk, D_DISKLESS));
+       wait_event(mdev->misc_wait,
+                       mdev->state.disk != D_DISKLESS ||
+                       !atomic_read(&mdev->local_cnt));
        drbd_resume_io(mdev);
-       if ((int)retcode == (int)SS_IS_DISKLESS)
-               retcode = SS_NOTHING_TO_DO;
-       if (ret)
-               retcode = ERR_INTR;
-       reply->ret_code = retcode;
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_net_conf(struct drbd_tconn *tconn, struct drbd_nl_cfg_req *nlp,
-                           struct drbd_nl_cfg_reply *reply)
+int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info)
 {
-       int i;
-       enum drbd_ret_code retcode;
+       char hmac_name[CRYPTO_MAX_ALG_NAME];
+       struct drbd_conf *mdev;
        struct net_conf *new_conf = NULL;
        struct crypto_hash *tfm = NULL;
        struct crypto_hash *integrity_w_tfm = NULL;
        struct crypto_hash *integrity_r_tfm = NULL;
-       struct drbd_conf *mdev;
-       char hmac_name[CRYPTO_MAX_ALG_NAME];
        void *int_dig_out = NULL;
        void *int_dig_in = NULL;
        void *int_dig_vv = NULL;
        struct drbd_tconn *oconn;
+       struct drbd_tconn *tconn;
        struct sockaddr *new_my_addr, *new_peer_addr, *taken_addr;
+       enum drbd_ret_code retcode;
+       int i;
+       int err;
 
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
+
+       tconn = adm_ctx.tconn;
        conn_reconfig_start(tconn);
 
        if (tconn->cstate > C_STANDALONE) {
@@ -1343,8 +1501,10 @@ static int drbd_nl_net_conf(struct drbd_tconn *tconn, struct drbd_nl_cfg_req *nl
        new_conf->on_congestion    = DRBD_ON_CONGESTION_DEF;
        new_conf->cong_extents     = DRBD_CONG_EXTENTS_DEF;
 
-       if (!net_conf_from_tags(nlp->tag_list, new_conf)) {
+       err = net_conf_from_attrs(new_conf, info->attrs);
+       if (err) {
                retcode = ERR_MANDATORY_TAG;
+               drbd_msg_put_info(from_attrs_err_to_txt(err));
                goto fail;
        }
 
@@ -1495,8 +1655,8 @@ static int drbd_nl_net_conf(struct drbd_tconn *tconn, struct drbd_nl_cfg_req *nl
                mdev->recv_cnt = 0;
                kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
        }
-       reply->ret_code = retcode;
        conn_reconfig_done(tconn);
+       drbd_adm_finish(info, retcode);
        return 0;
 
 fail:
@@ -1508,24 +1668,37 @@ fail:
        crypto_free_hash(integrity_r_tfm);
        kfree(new_conf);
 
-       reply->ret_code = retcode;
        conn_reconfig_done(tconn);
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_disconnect(struct drbd_tconn *tconn, struct drbd_nl_cfg_req *nlp,
-                             struct drbd_nl_cfg_reply *reply)
+int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info)
 {
-       int retcode;
-       struct disconnect dc;
+       struct disconnect_parms parms;
+       struct drbd_tconn *tconn;
+       enum drbd_ret_code retcode;
+       int err;
 
-       memset(&dc, 0, sizeof(struct disconnect));
-       if (!disconnect_from_tags(nlp->tag_list, &dc)) {
-               retcode = ERR_MANDATORY_TAG;
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
                goto fail;
+
+       tconn = adm_ctx.tconn;
+       memset(&parms, 0, sizeof(parms));
+       if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) {
+               err = disconnect_parms_from_attrs(&parms, info->attrs);
+               if (err) {
+                       retcode = ERR_MANDATORY_TAG;
+                       drbd_msg_put_info(from_attrs_err_to_txt(err));
+                       goto fail;
+               }
        }
 
-       if (dc.force) {
+       if (parms.force_disconnect) {
                spin_lock_irq(&tconn->req_lock);
                if (tconn->cstate >= C_WF_CONNECTION)
                        _conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
@@ -1567,7 +1740,7 @@ static int drbd_nl_disconnect(struct drbd_tconn *tconn, struct drbd_nl_cfg_req *
  done:
        retcode = NO_ERROR;
  fail:
-       reply->ret_code = retcode;
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
@@ -1587,20 +1760,32 @@ void resync_after_online_grow(struct drbd_conf *mdev)
                _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE);
 }
 
-static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                         struct drbd_nl_cfg_reply *reply)
+int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info)
 {
-       struct resize rs;
-       int retcode = NO_ERROR;
+       struct resize_parms rs;
+       struct drbd_conf *mdev;
+       enum drbd_ret_code retcode;
        enum determine_dev_size dd;
        enum dds_flags ddsf;
+       int err;
 
-       memset(&rs, 0, sizeof(struct resize));
-       if (!resize_from_tags(nlp->tag_list, &rs)) {
-               retcode = ERR_MANDATORY_TAG;
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
                goto fail;
+
+       memset(&rs, 0, sizeof(struct resize_parms));
+       if (info->attrs[DRBD_NLA_RESIZE_PARMS]) {
+               err = resize_parms_from_attrs(&rs, info->attrs);
+               if (err) {
+                       retcode = ERR_MANDATORY_TAG;
+                       drbd_msg_put_info(from_attrs_err_to_txt(err));
+                       goto fail;
+               }
        }
 
+       mdev = adm_ctx.mdev;
        if (mdev->state.conn > C_CONNECTED) {
                retcode = ERR_RESIZE_RESYNC;
                goto fail;
@@ -1644,14 +1829,14 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
        }
 
  fail:
-       reply->ret_code = retcode;
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                              struct drbd_nl_cfg_reply *reply)
+int drbd_adm_syncer(struct sk_buff *skb, struct genl_info *info)
 {
-       int retcode = NO_ERROR;
+       struct drbd_conf *mdev;
+       enum drbd_ret_code retcode;
        int err;
        int ovr; /* online verify running */
        int rsr; /* re-sync running */
@@ -1662,12 +1847,21 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n
        int *rs_plan_s = NULL;
        int fifo_size;
 
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto fail;
+       mdev = adm_ctx.mdev;
+
        if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL)) {
                retcode = ERR_NOMEM;
+               drbd_msg_put_info("unable to allocate cpumask");
                goto fail;
        }
 
-       if (nlp->flags & DRBD_NL_SET_DEFAULTS) {
+       if (((struct drbd_genlmsghdr*)info->userhdr)->flags
+                       & DRBD_GENL_F_SET_DEFAULTS) {
                memset(&sc, 0, sizeof(struct syncer_conf));
                sc.rate       = DRBD_RATE_DEF;
                sc.after      = DRBD_AFTER_DEF;
@@ -1681,8 +1875,10 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n
        } else
                memcpy(&sc, &mdev->sync_conf, sizeof(struct syncer_conf));
 
-       if (!syncer_conf_from_tags(nlp->tag_list, &sc)) {
+       err = syncer_conf_from_attrs(&sc, info->attrs);
+       if (err) {
                retcode = ERR_MANDATORY_TAG;
+               drbd_msg_put_info(from_attrs_err_to_txt(err));
                goto fail;
        }
 
@@ -1832,14 +2028,23 @@ fail:
        free_cpumask_var(new_cpu_mask);
        crypto_free_hash(csums_tfm);
        crypto_free_hash(verify_tfm);
-       reply->ret_code = retcode;
+
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_invalidate(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                             struct drbd_nl_cfg_reply *reply)
+int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info)
 {
-       int retcode;
+       struct drbd_conf *mdev;
+       int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
+
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
+
+       mdev = adm_ctx.mdev;
 
        /* If there is still bitmap IO pending, probably because of a previous
         * resync just being finished, wait for it before requesting a new resync. */
@@ -1862,7 +2067,8 @@ static int drbd_nl_invalidate(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nl
                retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
        }
 
-       reply->ret_code = retcode;
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
@@ -1875,56 +2081,58 @@ static int drbd_bmio_set_susp_al(struct drbd_conf *mdev)
        return rv;
 }
 
-static int drbd_nl_invalidate_peer(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                                  struct drbd_nl_cfg_reply *reply)
+static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info,
+               union drbd_state mask, union drbd_state val)
 {
-       int retcode;
-
-       /* If there is still bitmap IO pending, probably because of a previous
-        * resync just being finished, wait for it before requesting a new resync. */
-       wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
+       enum drbd_ret_code retcode;
 
-       retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S), CS_ORDERED);
-
-       if (retcode < SS_SUCCESS) {
-               if (retcode == SS_NEED_CONNECTION && mdev->state.role == R_PRIMARY) {
-                       /* The peer will get a resync upon connect anyways. Just make that
-                          into a full resync. */
-                       retcode = drbd_request_state(mdev, NS(pdsk, D_INCONSISTENT));
-                       if (retcode >= SS_SUCCESS) {
-                               if (drbd_bitmap_io(mdev, &drbd_bmio_set_susp_al,
-                                       "set_n_write from invalidate_peer",
-                                       BM_LOCKED_SET_ALLOWED))
-                                       retcode = ERR_IO_MD_DISK;
-                       }
-               } else
-                       retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S));
-       }
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       reply->ret_code = retcode;
+       retcode = drbd_request_state(adm_ctx.mdev, mask, val);
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_pause_sync(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                             struct drbd_nl_cfg_reply *reply)
+int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info)
 {
-       int retcode = NO_ERROR;
+       return drbd_adm_simple_request_state(skb, info, NS(conn, C_STARTING_SYNC_S));
+}
 
-       if (drbd_request_state(mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO)
-               retcode = ERR_PAUSE_IS_SET;
+int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info)
+{
+       enum drbd_ret_code retcode;
+
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       reply->ret_code = retcode;
+       if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO)
+               retcode = ERR_PAUSE_IS_SET;
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_resume_sync(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                              struct drbd_nl_cfg_reply *reply)
+int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info)
 {
-       int retcode = NO_ERROR;
        union drbd_state s;
+       enum drbd_ret_code retcode;
+
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       if (drbd_request_state(mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) {
-               s = mdev->state;
+       if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) {
+               s = adm_ctx.mdev->state;
                if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) {
                        retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP :
                                  s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR;
@@ -1933,28 +2141,35 @@ static int drbd_nl_resume_sync(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n
                }
        }
 
-       reply->ret_code = retcode;
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_suspend_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                             struct drbd_nl_cfg_reply *reply)
+int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info)
 {
-       reply->ret_code = drbd_request_state(mdev, NS(susp, 1));
-
-       return 0;
+       return drbd_adm_simple_request_state(skb, info, NS(susp, 1));
 }
 
-static int drbd_nl_resume_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                            struct drbd_nl_cfg_reply *reply)
+int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info)
 {
+       struct drbd_conf *mdev;
+       int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
+
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
+
+       mdev = adm_ctx.mdev;
        if (test_bit(NEW_CUR_UUID, &mdev->flags)) {
                drbd_uuid_new_current(mdev);
                clear_bit(NEW_CUR_UUID, &mdev->flags);
        }
        drbd_suspend_io(mdev);
-       reply->ret_code = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0));
-       if (reply->ret_code == SS_SUCCESS) {
+       retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0));
+       if (retcode == SS_SUCCESS) {
                if (mdev->state.conn < C_CONNECTED)
                        tl_clear(mdev->tconn);
                if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED)
@@ -1962,138 +2177,259 @@ static int drbd_nl_resume_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
        }
        drbd_resume_io(mdev);
 
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_outdate(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                          struct drbd_nl_cfg_reply *reply)
+int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info)
 {
-       reply->ret_code = drbd_request_state(mdev, NS(disk, D_OUTDATED));
-       return 0;
+       return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED));
 }
 
-static int drbd_nl_get_config(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                          struct drbd_nl_cfg_reply *reply)
+int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev,
+               const struct sib_info *sib)
 {
-       unsigned short *tl;
-
-       tl = reply->tag_list;
-
-       if (get_ldev(mdev)) {
-               tl = disk_conf_to_tags(&mdev->ldev->dc, tl);
-               put_ldev(mdev);
+       struct state_info *si = NULL; /* for sizeof(si->member); */
+       struct nlattr *nla;
+       int got_ldev;
+       int got_net;
+       int err = 0;
+       int exclude_sensitive;
+
+       /* If sib != NULL, this is drbd_bcast_event, which anyone can listen
+        * to.  So we better exclude_sensitive information.
+        *
+        * If sib == NULL, this is drbd_adm_get_status, executed synchronously
+        * in the context of the requesting user process. Exclude sensitive
+        * information, unless current has superuser.
+        *
+        * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and
+        * relies on the current implementation of netlink_dump(), which
+        * executes the dump callback successively from netlink_recvmsg(),
+        * always in the context of the receiving process */
+       exclude_sensitive = sib || !capable(CAP_SYS_ADMIN);
+
+       got_ldev = get_ldev(mdev);
+       got_net = get_net_conf(mdev->tconn);
+
+       /* We need to add connection name and volume number information still.
+        * Minor number is in drbd_genlmsghdr. */
+       nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT);
+       if (!nla)
+               goto nla_put_failure;
+       NLA_PUT_U32(skb, T_ctx_volume, mdev->vnr);
+       NLA_PUT_STRING(skb, T_ctx_conn_name, mdev->tconn->name);
+       nla_nest_end(skb, nla);
+
+       if (got_ldev)
+               if (disk_conf_to_skb(skb, &mdev->ldev->dc, exclude_sensitive))
+                       goto nla_put_failure;
+       if (got_net)
+               if (net_conf_to_skb(skb, mdev->tconn->net_conf, exclude_sensitive))
+                       goto nla_put_failure;
+
+       if (syncer_conf_to_skb(skb, &mdev->sync_conf, exclude_sensitive))
+                       goto nla_put_failure;
+
+       nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO);
+       if (!nla)
+               goto nla_put_failure;
+       NLA_PUT_U32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY);
+       NLA_PUT_U32(skb, T_current_state, mdev->state.i);
+       NLA_PUT_U64(skb, T_ed_uuid, mdev->ed_uuid);
+       NLA_PUT_U64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev));
+
+       if (got_ldev) {
+               NLA_PUT_U32(skb, T_disk_flags, mdev->ldev->md.flags);
+               NLA_PUT(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid);
+               NLA_PUT_U64(skb, T_bits_total, drbd_bm_bits(mdev));
+               NLA_PUT_U64(skb, T_bits_oos, drbd_bm_total_weight(mdev));
+               if (C_SYNC_SOURCE <= mdev->state.conn &&
+                   C_PAUSED_SYNC_T >= mdev->state.conn) {
+                       NLA_PUT_U64(skb, T_bits_rs_total, mdev->rs_total);
+                       NLA_PUT_U64(skb, T_bits_rs_failed, mdev->rs_failed);
+               }
        }
 
-       if (get_net_conf(mdev->tconn)) {
-               tl = net_conf_to_tags(mdev->tconn->net_conf, tl);
-               put_net_conf(mdev->tconn);
+       if (sib) {
+               switch(sib->sib_reason) {
+               case SIB_SYNC_PROGRESS:
+               case SIB_GET_STATUS_REPLY:
+                       break;
+               case SIB_STATE_CHANGE:
+                       NLA_PUT_U32(skb, T_prev_state, sib->os.i);
+                       NLA_PUT_U32(skb, T_new_state, sib->ns.i);
+                       break;
+               case SIB_HELPER_POST:
+                       NLA_PUT_U32(skb,
+                               T_helper_exit_code, sib->helper_exit_code);
+                       /* fall through */
+               case SIB_HELPER_PRE:
+                       NLA_PUT_STRING(skb, T_helper, sib->helper_name);
+                       break;
+               }
        }
-       tl = syncer_conf_to_tags(&mdev->sync_conf, tl);
-
-       put_unaligned(TT_END, tl++); /* Close the tag list */
+       nla_nest_end(skb, nla);
 
-       return (int)((char *)tl - (char *)reply->tag_list);
+       if (0)
+nla_put_failure:
+               err = -EMSGSIZE;
+       if (got_ldev)
+               put_ldev(mdev);
+       if (got_net)
+               put_net_conf(mdev->tconn);
+       return err;
 }
 
-static int drbd_nl_get_state(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                            struct drbd_nl_cfg_reply *reply)
+int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info)
 {
-       unsigned short *tl = reply->tag_list;
-       union drbd_state s = mdev->state;
-       unsigned long rs_left;
-       unsigned int res;
+       enum drbd_ret_code retcode;
+       int err;
 
-       tl = get_state_to_tags((struct get_state *)&s, tl);
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       /* no local ref, no bitmap, no syncer progress. */
-       if (s.conn >= C_SYNC_SOURCE && s.conn <= C_PAUSED_SYNC_T) {
-               if (get_ldev(mdev)) {
-                       drbd_get_syncer_progress(mdev, &rs_left, &res);
-                       tl = tl_add_int(tl, T_sync_progress, &res);
-                       put_ldev(mdev);
-               }
+       err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL);
+       if (err) {
+               nlmsg_free(adm_ctx.reply_skb);
+               return err;
        }
-       put_unaligned(TT_END, tl++); /* Close the tag list */
-
-       return (int)((char *)tl - (char *)reply->tag_list);
+out:
+       drbd_adm_finish(info, retcode);
+       return 0;
 }
 
-static int drbd_nl_get_uuids(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                            struct drbd_nl_cfg_reply *reply)
+int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb)
 {
-       unsigned short *tl;
-
-       tl = reply->tag_list;
+       struct drbd_conf *mdev;
+       struct drbd_genlmsghdr *dh;
+       int minor = cb->args[0];
+
+       /* Open coded deferred single idr_for_each_entry iteration.
+        * This may miss entries inserted after this dump started,
+        * or entries deleted before they are reached.
+        * But we need to make sure the mdev won't disappear while
+        * we are looking at it. */
+
+       rcu_read_lock();
+       mdev = idr_get_next(&minors, &minor);
+       if (mdev) {
+               dh = genlmsg_put(skb, NETLINK_CB(cb->skb).pid,
+                               cb->nlh->nlmsg_seq, &drbd_genl_family,
+                               NLM_F_MULTI, DRBD_ADM_GET_STATUS);
+               if (!dh)
+                       goto errout;
+
+               D_ASSERT(mdev->minor == minor);
+
+               dh->minor = minor;
+               dh->ret_code = NO_ERROR;
+
+               if (nla_put_status_info(skb, mdev, NULL)) {
+                       genlmsg_cancel(skb, dh);
+                       goto errout;
+               }
+               genlmsg_end(skb, dh);
+        }
 
-       if (get_ldev(mdev)) {
-               tl = tl_add_blob(tl, T_uuids, mdev->ldev->md.uuid, UI_SIZE*sizeof(u64));
-               tl = tl_add_int(tl, T_uuids_flags, &mdev->ldev->md.flags);
-               put_ldev(mdev);
-       }
-       put_unaligned(TT_END, tl++); /* Close the tag list */
+errout:
+       rcu_read_unlock();
+       /* where to start idr_get_next with the next iteration */
+        cb->args[0] = minor+1;
 
-       return (int)((char *)tl - (char *)reply->tag_list);
+       /* No more minors found: empty skb. Which will terminate the dump. */
+        return skb->len;
 }
 
-/**
- * drbd_nl_get_timeout_flag() - Used by drbdsetup to find out which timeout value to use
- * @mdev:      DRBD device.
- * @nlp:       Netlink/connector packet from drbdsetup
- * @reply:     Reply packet for drbdsetup
- */
-static int drbd_nl_get_timeout_flag(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                                   struct drbd_nl_cfg_reply *reply)
+int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info)
 {
-       unsigned short *tl;
-       char rv;
-
-       tl = reply->tag_list;
+       enum drbd_ret_code retcode;
+       struct timeout_parms tp;
+       int err;
 
-       rv = mdev->state.pdsk == D_OUTDATED        ? UT_PEER_OUTDATED :
-         test_bit(USE_DEGR_WFC_T, &mdev->flags) ? UT_DEGRADED : UT_DEFAULT;
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       tl = tl_add_blob(tl, T_use_degraded, &rv, sizeof(rv));
-       put_unaligned(TT_END, tl++); /* Close the tag list */
+       tp.timeout_type =
+               adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED :
+               test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED :
+               UT_DEFAULT;
 
-       return (int)((char *)tl - (char *)reply->tag_list);
+       err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp);
+       if (err) {
+               nlmsg_free(adm_ctx.reply_skb);
+               return err;
+       }
+out:
+       drbd_adm_finish(info, retcode);
+       return 0;
 }
 
-static int drbd_nl_start_ov(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                                   struct drbd_nl_cfg_reply *reply)
+int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info)
 {
-       /* default to resume from last known position, if possible */
-       struct start_ov args =
-               { .start_sector = mdev->ov_start_sector };
+       struct drbd_conf *mdev;
+       enum drbd_ret_code retcode;
 
-       if (!start_ov_from_tags(nlp->tag_list, &args)) {
-               reply->ret_code = ERR_MANDATORY_TAG;
-               return 0;
-       }
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
+       mdev = adm_ctx.mdev;
+       if (info->attrs[DRBD_NLA_START_OV_PARMS]) {
+               /* resume from last known position, if possible */
+               struct start_ov_parms parms =
+                       { .ov_start_sector = mdev->ov_start_sector };
+               int err = start_ov_parms_from_attrs(&parms, info->attrs);
+               if (err) {
+                       retcode = ERR_MANDATORY_TAG;
+                       drbd_msg_put_info(from_attrs_err_to_txt(err));
+                       goto out;
+               }
+               /* w_make_ov_request expects position to be aligned */
+               mdev->ov_start_sector = parms.ov_start_sector & ~BM_SECT_PER_BIT;
+       }
        /* If there is still bitmap IO pending, e.g. previous resync or verify
         * just being finished, wait for it before requesting a new resync. */
        wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
-
-       /* w_make_ov_request expects position to be aligned */
-       mdev->ov_start_sector = args.start_sector & ~BM_SECT_PER_BIT;
-       reply->ret_code = drbd_request_state(mdev,NS(conn,C_VERIFY_S));
+       retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S));
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
 
-static int drbd_nl_new_c_uuid(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                             struct drbd_nl_cfg_reply *reply)
+int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info)
 {
-       int retcode = NO_ERROR;
+       struct drbd_conf *mdev;
+       enum drbd_ret_code retcode;
        int skip_initial_sync = 0;
        int err;
+       struct new_c_uuid_parms args;
 
-       struct new_c_uuid args;
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out_nolock;
 
-       memset(&args, 0, sizeof(struct new_c_uuid));
-       if (!new_c_uuid_from_tags(nlp->tag_list, &args)) {
-               reply->ret_code = ERR_MANDATORY_TAG;
-               return 0;
+       mdev = adm_ctx.mdev;
+       memset(&args, 0, sizeof(args));
+       if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) {
+               err = new_c_uuid_parms_from_attrs(&args, info->attrs);
+               if (err) {
+                       retcode = ERR_MANDATORY_TAG;
+                       drbd_msg_put_info(from_attrs_err_to_txt(err));
+                       goto out_nolock;
+               }
        }
 
        mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */
@@ -2139,510 +2475,164 @@ out_dec:
        put_ldev(mdev);
 out:
        mutex_unlock(mdev->state_mutex);
-
-       reply->ret_code = retcode;
-       return 0;
-}
-
-static int drbd_nl_new_conn(struct drbd_nl_cfg_req *nlp, struct drbd_nl_cfg_reply *reply)
-{
-       struct new_connection args;
-
-       if (!new_connection_from_tags(nlp->tag_list, &args)) {
-               reply->ret_code = ERR_MANDATORY_TAG;
-               return 0;
-       }
-
-       reply->ret_code = NO_ERROR;
-       if (!drbd_new_tconn(args.name))
-               reply->ret_code = ERR_NOMEM;
-
-       return 0;
-}
-
-static int drbd_nl_new_minor(struct drbd_tconn *tconn,
-                     struct drbd_nl_cfg_req *nlp, struct drbd_nl_cfg_reply *reply)
-{
-       struct new_minor args;
-
-       args.vol_nr = 0;
-       args.minor = 0;
-
-       if (!new_minor_from_tags(nlp->tag_list, &args)) {
-               reply->ret_code = ERR_MANDATORY_TAG;
-               return 0;
-       }
-
-       reply->ret_code = conn_new_minor(tconn, args.minor, args.vol_nr);
-
+out_nolock:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-static int drbd_nl_del_minor(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
-                            struct drbd_nl_cfg_reply *reply)
+static enum drbd_ret_code
+drbd_check_conn_name(const char *name)
 {
-       if (mdev->state.disk == D_DISKLESS &&
-           mdev->state.conn == C_STANDALONE &&
-           mdev->state.role == R_SECONDARY) {
-               drbd_delete_device(mdev_to_minor(mdev));
-               reply->ret_code = NO_ERROR;
-       } else {
-               reply->ret_code = ERR_MINOR_CONFIGURED;
+       if (!name || !name[0]) {
+               drbd_msg_put_info("connection name missing");
+               return ERR_MANDATORY_TAG;
        }
-       return 0;
-}
-
-static int drbd_nl_del_conn(struct drbd_tconn *tconn,
-                           struct drbd_nl_cfg_req *nlp, struct drbd_nl_cfg_reply *reply)
-{
-       if (conn_lowest_minor(tconn) < 0) {
-               drbd_free_tconn(tconn);
-               reply->ret_code = NO_ERROR;
-       } else {
-               reply->ret_code = ERR_CONN_IN_USE;
+       /* if we want to use these in sysfs/configfs/debugfs some day,
+        * we must not allow slashes */
+       if (strchr(name, '/')) {
+               drbd_msg_put_info("invalid connection name");
+               return ERR_INVALID_REQUEST;
        }
-
-       return 0;
+       return NO_ERROR;
 }
 
-enum cn_handler_type {
-       CHT_MINOR,
-       CHT_CONN,
-       CHT_CTOR,
-       /* CHT_RES, later */
-};
-struct cn_handler_struct {
-       enum cn_handler_type type;
-       union {
-               int (*minor_based)(struct drbd_conf *,
-                                  struct drbd_nl_cfg_req *,
-                                  struct drbd_nl_cfg_reply *);
-               int (*conn_based)(struct drbd_tconn *,
-                                 struct drbd_nl_cfg_req *,
-                                 struct drbd_nl_cfg_reply *);
-               int (*constructor)(struct drbd_nl_cfg_req *,
-                                  struct drbd_nl_cfg_reply *);
-       };
-       int reply_body_size;
-};
-
-static struct cn_handler_struct cnd_table[] = {
-       [ P_primary ]           = { CHT_MINOR, { &drbd_nl_primary },    0 },
-       [ P_secondary ]         = { CHT_MINOR, { &drbd_nl_secondary },  0 },
-       [ P_disk_conf ]         = { CHT_MINOR, { &drbd_nl_disk_conf },  0 },
-       [ P_detach ]            = { CHT_MINOR, { &drbd_nl_detach },     0 },
-       [ P_net_conf ]          = { CHT_CONN,  { .conn_based = &drbd_nl_net_conf },     0 },
-       [ P_disconnect ]        = { CHT_CONN,  { .conn_based = &drbd_nl_disconnect },   0 },
-       [ P_resize ]            = { CHT_MINOR, { &drbd_nl_resize },     0 },
-       [ P_syncer_conf ]       = { CHT_MINOR, { &drbd_nl_syncer_conf },0 },
-       [ P_invalidate ]        = { CHT_MINOR, { &drbd_nl_invalidate }, 0 },
-       [ P_invalidate_peer ]   = { CHT_MINOR, { &drbd_nl_invalidate_peer },0 },
-       [ P_pause_sync ]        = { CHT_MINOR, { &drbd_nl_pause_sync }, 0 },
-       [ P_resume_sync ]       = { CHT_MINOR, { &drbd_nl_resume_sync },0 },
-       [ P_suspend_io ]        = { CHT_MINOR, { &drbd_nl_suspend_io }, 0 },
-       [ P_resume_io ]         = { CHT_MINOR, { &drbd_nl_resume_io },  0 },
-       [ P_outdate ]           = { CHT_MINOR, { &drbd_nl_outdate },    0 },
-       [ P_get_config ]        = { CHT_MINOR, { &drbd_nl_get_config },
-                                   sizeof(struct syncer_conf_tag_len_struct) +
-                                   sizeof(struct disk_conf_tag_len_struct) +
-                                   sizeof(struct net_conf_tag_len_struct) },
-       [ P_get_state ]         = { CHT_MINOR, { &drbd_nl_get_state },
-                                   sizeof(struct get_state_tag_len_struct) +
-                                   sizeof(struct sync_progress_tag_len_struct) },
-       [ P_get_uuids ]         = { CHT_MINOR, { &drbd_nl_get_uuids },
-                                   sizeof(struct get_uuids_tag_len_struct) },
-       [ P_get_timeout_flag ]  = { CHT_MINOR, { &drbd_nl_get_timeout_flag },
-                                   sizeof(struct get_timeout_flag_tag_len_struct)},
-       [ P_start_ov ]          = { CHT_MINOR, { &drbd_nl_start_ov },   0 },
-       [ P_new_c_uuid ]        = { CHT_MINOR, { &drbd_nl_new_c_uuid }, 0 },
-       [ P_new_connection ]    = { CHT_CTOR,  { .constructor = &drbd_nl_new_conn }, 0 },
-       [ P_new_minor ]         = { CHT_CONN,  { .conn_based = &drbd_nl_new_minor }, 0 },
-       [ P_del_minor ]         = { CHT_MINOR, { &drbd_nl_del_minor },  0 },
-       [ P_del_connection ]    = { CHT_CONN,  { .conn_based = &drbd_nl_del_conn }, 0 },
-};
-
-static void drbd_connector_callback(struct cn_msg *req, struct netlink_skb_parms *nsp)
+int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info)
 {
-       struct drbd_nl_cfg_req *nlp = (struct drbd_nl_cfg_req *)req->data;
-       struct cn_handler_struct *cm;
-       struct cn_msg *cn_reply;
-       struct drbd_nl_cfg_reply *reply;
-       struct drbd_conf *mdev;
-       struct drbd_tconn *tconn;
-       int retcode, rr;
-       int reply_size = sizeof(struct cn_msg)
-               + sizeof(struct drbd_nl_cfg_reply)
-               + sizeof(short int);
-
-       if (!try_module_get(THIS_MODULE)) {
-               printk(KERN_ERR "drbd: try_module_get() failed!\n");
-               return;
-       }
-
-       if (!cap_raised(current_cap(), CAP_SYS_ADMIN)) {
-               retcode = ERR_PERM;
-               goto fail;
-       }
+       enum drbd_ret_code retcode;
 
-       if (nlp->packet_type >= P_nl_after_last_packet ||
-           nlp->packet_type == P_return_code_only) {
-               retcode = ERR_PACKET_NR;
-               goto fail;
-       }
+       retcode = drbd_adm_prepare(skb, info, 0);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       cm = cnd_table + nlp->packet_type;
+       retcode = drbd_check_conn_name(adm_ctx.conn_name);
+       if (retcode != NO_ERROR)
+               goto out;
 
-       /* This may happen if packet number is 0: */
-       if (cm->minor_based == NULL) {
-               retcode = ERR_PACKET_NR;
-               goto fail;
+       if (adm_ctx.tconn) {
+               retcode = ERR_INVALID_REQUEST;
+               drbd_msg_put_info("connection exists");
+               goto out;
        }
 
-       reply_size += cm->reply_body_size;
-
-       /* allocation not in the IO path, cqueue thread context */
-       cn_reply = kzalloc(reply_size, GFP_KERNEL);
-       if (!cn_reply) {
+       if (!drbd_new_tconn(adm_ctx.conn_name))
                retcode = ERR_NOMEM;
-               goto fail;
-       }
-       reply = (struct drbd_nl_cfg_reply *) cn_reply->data;
-
-       reply->packet_type =
-               cm->reply_body_size ? nlp->packet_type : P_return_code_only;
-       reply->minor = nlp->drbd_minor;
-       reply->ret_code = NO_ERROR; /* Might by modified by cm->function. */
-       /* reply->tag_list; might be modified by cm->function. */
-
-       retcode = ERR_MINOR_INVALID;
-       rr = 0;
-       switch (cm->type) {
-       case CHT_MINOR:
-               mdev = minor_to_mdev(nlp->drbd_minor);
-               if (!mdev)
-                       goto fail;
-               rr = cm->minor_based(mdev, nlp, reply);
-               break;
-       case CHT_CONN:
-               tconn = conn_by_name(nlp->obj_name);
-               if (!tconn) {
-                       retcode = ERR_CONN_NOT_KNOWN;
-                       goto fail;
-               }
-               rr = cm->conn_based(tconn, nlp, reply);
-               break;
-       case CHT_CTOR:
-               rr = cm->constructor(nlp, reply);
-               break;
-       /* case CHT_RES: */
-       }
-
-       cn_reply->id = req->id;
-       cn_reply->seq = req->seq;
-       cn_reply->ack = req->ack  + 1;
-       cn_reply->len = sizeof(struct drbd_nl_cfg_reply) + rr;
-       cn_reply->flags = 0;
-
-       rr = cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_KERNEL);
-       if (rr && rr != -ESRCH)
-               printk(KERN_INFO "drbd: cn_netlink_send()=%d\n", rr);
-
-       kfree(cn_reply);
-       module_put(THIS_MODULE);
-       return;
- fail:
-       drbd_nl_send_reply(req, retcode);
-       module_put(THIS_MODULE);
-}
-
-static atomic_t drbd_nl_seq = ATOMIC_INIT(2); /* two. */
-
-static unsigned short *
-__tl_add_blob(unsigned short *tl, enum drbd_tags tag, const void *data,
-       unsigned short len, int nul_terminated)
-{
-       unsigned short l = tag_descriptions[tag_number(tag)].max_len;
-       len = (len < l) ? len :  l;
-       put_unaligned(tag, tl++);
-       put_unaligned(len, tl++);
-       memcpy(tl, data, len);
-       tl = (unsigned short*)((char*)tl + len);
-       if (nul_terminated)
-               *((char*)tl - 1) = 0;
-       return tl;
-}
-
-static unsigned short *
-tl_add_blob(unsigned short *tl, enum drbd_tags tag, const void *data, int len)
-{
-       return __tl_add_blob(tl, tag, data, len, 0);
-}
-
-static unsigned short *
-tl_add_str(unsigned short *tl, enum drbd_tags tag, const char *str)
-{
-       return __tl_add_blob(tl, tag, str, strlen(str)+1, 0);
-}
-
-static unsigned short *
-tl_add_int(unsigned short *tl, enum drbd_tags tag, const void *val)
-{
-       put_unaligned(tag, tl++);
-       switch(tag_type(tag)) {
-       case TT_INTEGER:
-               put_unaligned(sizeof(int), tl++);
-               put_unaligned(*(int *)val, (int *)tl);
-               tl = (unsigned short*)((char*)tl+sizeof(int));
-               break;
-       case TT_INT64:
-               put_unaligned(sizeof(u64), tl++);
-               put_unaligned(*(u64 *)val, (u64 *)tl);
-               tl = (unsigned short*)((char*)tl+sizeof(u64));
-               break;
-       default:
-               /* someone did something stupid. */
-               ;
-       }
-       return tl;
-}
-
-void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state state)
-{
-       char buffer[sizeof(struct cn_msg)+
-                   sizeof(struct drbd_nl_cfg_reply)+
-                   sizeof(struct get_state_tag_len_struct)+
-                   sizeof(short int)];
-       struct cn_msg *cn_reply = (struct cn_msg *) buffer;
-       struct drbd_nl_cfg_reply *reply =
-               (struct drbd_nl_cfg_reply *)cn_reply->data;
-       unsigned short *tl = reply->tag_list;
-
-       /* dev_warn(DEV, "drbd_bcast_state() got called\n"); */
-
-       tl = get_state_to_tags((struct get_state *)&state, tl);
-
-       put_unaligned(TT_END, tl++); /* Close the tag list */
-
-       cn_reply->id.idx = CN_IDX_DRBD;
-       cn_reply->id.val = CN_VAL_DRBD;
-
-       cn_reply->seq = atomic_inc_return(&drbd_nl_seq);
-       cn_reply->ack = 0; /* not used here. */
-       cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
-               (int)((char *)tl - (char *)reply->tag_list);
-       cn_reply->flags = 0;
-
-       reply->packet_type = P_get_state;
-       reply->minor = mdev_to_minor(mdev);
-       reply->ret_code = NO_ERROR;
-
-       cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
-}
-
-void drbd_bcast_ev_helper(struct drbd_conf *mdev, char *helper_name)
-{
-       char buffer[sizeof(struct cn_msg)+
-                   sizeof(struct drbd_nl_cfg_reply)+
-                   sizeof(struct call_helper_tag_len_struct)+
-                   sizeof(short int)];
-       struct cn_msg *cn_reply = (struct cn_msg *) buffer;
-       struct drbd_nl_cfg_reply *reply =
-               (struct drbd_nl_cfg_reply *)cn_reply->data;
-       unsigned short *tl = reply->tag_list;
-
-       /* dev_warn(DEV, "drbd_bcast_state() got called\n"); */
-
-       tl = tl_add_str(tl, T_helper, helper_name);
-       put_unaligned(TT_END, tl++); /* Close the tag list */
-
-       cn_reply->id.idx = CN_IDX_DRBD;
-       cn_reply->id.val = CN_VAL_DRBD;
-
-       cn_reply->seq = atomic_inc_return(&drbd_nl_seq);
-       cn_reply->ack = 0; /* not used here. */
-       cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
-               (int)((char *)tl - (char *)reply->tag_list);
-       cn_reply->flags = 0;
-
-       reply->packet_type = P_call_helper;
-       reply->minor = mdev_to_minor(mdev);
-       reply->ret_code = NO_ERROR;
-
-       cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
+out:
+       drbd_adm_finish(info, retcode);
+       return 0;
 }
 
-void drbd_bcast_ee(struct drbd_conf *mdev, const char *reason, const int dgs,
-                  const char *seen_hash, const char *calc_hash,
-                          const struct drbd_peer_request *peer_req)
+int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info)
 {
-       struct cn_msg *cn_reply;
-       struct drbd_nl_cfg_reply *reply;
-       unsigned short *tl;
-       struct page *page;
-       unsigned len;
+       struct drbd_genlmsghdr *dh = info->userhdr;
+       enum drbd_ret_code retcode;
 
-       if (!peer_req)
-               return;
-       if (!reason || !reason[0])
-               return;
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       /* apparently we have to memcpy twice, first to prepare the data for the
-        * struct cn_msg, then within cn_netlink_send from the cn_msg to the
-        * netlink skb. */
-       /* receiver thread context, which is not in the writeout path (of this node),
-        * but may be in the writeout path of the _other_ node.
-        * GFP_NOIO to avoid potential "distributed deadlock". */
-       cn_reply = kzalloc(
-               sizeof(struct cn_msg)+
-               sizeof(struct drbd_nl_cfg_reply)+
-               sizeof(struct dump_ee_tag_len_struct)+
-               sizeof(short int),
-               GFP_NOIO);
-
-       if (!cn_reply) {
-               dev_err(DEV, "could not kmalloc buffer for drbd_bcast_ee, "
-                            "sector %llu, size %u\n",
-                       (unsigned long long)peer_req->i.sector,
-                       peer_req->i.size);
-               return;
+       /* FIXME drop minor_count parameter, limit to MINORMASK */
+       if (dh->minor >= minor_count) {
+               drbd_msg_put_info("requested minor out of range");
+               retcode = ERR_INVALID_REQUEST;
+               goto out;
        }
-
-       reply = (struct drbd_nl_cfg_reply*)cn_reply->data;
-       tl = reply->tag_list;
-
-       tl = tl_add_str(tl, T_dump_ee_reason, reason);
-       tl = tl_add_blob(tl, T_seen_digest, seen_hash, dgs);
-       tl = tl_add_blob(tl, T_calc_digest, calc_hash, dgs);
-       tl = tl_add_int(tl, T_ee_sector, &peer_req->i.sector);
-       tl = tl_add_int(tl, T_ee_block_id, &peer_req->block_id);
-
-       /* dump the first 32k */
-       len = min_t(unsigned, peer_req->i.size, 32 << 10);
-       put_unaligned(T_ee_data, tl++);
-       put_unaligned(len, tl++);
-
-       page = peer_req->pages;
-       page_chain_for_each(page) {
-               void *d = kmap_atomic(page, KM_USER0);
-               unsigned l = min_t(unsigned, len, PAGE_SIZE);
-               memcpy(tl, d, l);
-               kunmap_atomic(d, KM_USER0);
-               tl = (unsigned short*)((char*)tl + l);
-               len -= l;
-               if (len == 0)
-                       break;
+       /* FIXME we need a define here */
+       if (adm_ctx.volume >= 256) {
+               drbd_msg_put_info("requested volume id out of range");
+               retcode = ERR_INVALID_REQUEST;
+               goto out;
        }
-       put_unaligned(TT_END, tl++); /* Close the tag list */
-
-       cn_reply->id.idx = CN_IDX_DRBD;
-       cn_reply->id.val = CN_VAL_DRBD;
-
-       cn_reply->seq = atomic_inc_return(&drbd_nl_seq);
-       cn_reply->ack = 0; // not used here.
-       cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
-               (int)((char*)tl - (char*)reply->tag_list);
-       cn_reply->flags = 0;
 
-       reply->packet_type = P_dump_ee;
-       reply->minor = mdev_to_minor(mdev);
-       reply->ret_code = NO_ERROR;
-
-       cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
-       kfree(cn_reply);
+       retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume);
+out:
+       drbd_adm_finish(info, retcode);
+       return 0;
 }
 
-void drbd_bcast_sync_progress(struct drbd_conf *mdev)
+int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info)
 {
-       char buffer[sizeof(struct cn_msg)+
-                   sizeof(struct drbd_nl_cfg_reply)+
-                   sizeof(struct sync_progress_tag_len_struct)+
-                   sizeof(short int)];
-       struct cn_msg *cn_reply = (struct cn_msg *) buffer;
-       struct drbd_nl_cfg_reply *reply =
-               (struct drbd_nl_cfg_reply *)cn_reply->data;
-       unsigned short *tl = reply->tag_list;
-       unsigned long rs_left;
-       unsigned int res;
-
-       /* no local ref, no bitmap, no syncer progress, no broadcast. */
-       if (!get_ldev(mdev))
-               return;
-       drbd_get_syncer_progress(mdev, &rs_left, &res);
-       put_ldev(mdev);
-
-       tl = tl_add_int(tl, T_sync_progress, &res);
-       put_unaligned(TT_END, tl++); /* Close the tag list */
-
-       cn_reply->id.idx = CN_IDX_DRBD;
-       cn_reply->id.val = CN_VAL_DRBD;
-
-       cn_reply->seq = atomic_inc_return(&drbd_nl_seq);
-       cn_reply->ack = 0; /* not used here. */
-       cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
-               (int)((char *)tl - (char *)reply->tag_list);
-       cn_reply->flags = 0;
+       struct drbd_conf *mdev;
+       enum drbd_ret_code retcode;
 
-       reply->packet_type = P_sync_progress;
-       reply->minor = mdev_to_minor(mdev);
-       reply->ret_code = NO_ERROR;
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
 
-       cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
+       mdev = adm_ctx.mdev;
+       if (mdev->state.disk == D_DISKLESS &&
+           mdev->state.conn == C_STANDALONE &&
+           mdev->state.role == R_SECONDARY) {
+               drbd_delete_device(mdev_to_minor(mdev));
+               retcode = NO_ERROR;
+       } else
+               retcode = ERR_MINOR_CONFIGURED;
+out:
+       drbd_adm_finish(info, retcode);
+       return 0;
 }
 
-int __init drbd_nl_init(void)
+int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info)
 {
-       static struct cb_id cn_id_drbd;
-       int err, try=10;
-
-       cn_id_drbd.val = CN_VAL_DRBD;
-       do {
-               cn_id_drbd.idx = cn_idx;
-               err = cn_add_callback(&cn_id_drbd, "cn_drbd", &drbd_connector_callback);
-               if (!err)
-                       break;
-               cn_idx = (cn_idx + CN_IDX_STEP);
-       } while (try--);
+       enum drbd_ret_code retcode;
 
-       if (err) {
-               printk(KERN_ERR "drbd: cn_drbd failed to register\n");
-               return err;
+       retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
+       if (!adm_ctx.reply_skb)
+               return retcode;
+       if (retcode != NO_ERROR)
+               goto out;
+
+       if (conn_lowest_minor(adm_ctx.tconn) < 0) {
+               drbd_free_tconn(adm_ctx.tconn);
+               retcode = NO_ERROR;
+       } else {
+               retcode = ERR_CONN_IN_USE;
        }
 
+out:
+       drbd_adm_finish(info, retcode);
        return 0;
 }
 
-void drbd_nl_cleanup(void)
+void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib)
 {
-       static struct cb_id cn_id_drbd;
-
-       cn_id_drbd.idx = cn_idx;
-       cn_id_drbd.val = CN_VAL_DRBD;
+       static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */
+       struct sk_buff *msg;
+       struct drbd_genlmsghdr *d_out;
+       unsigned seq;
+       int err = -ENOMEM;
+
+       seq = atomic_inc_return(&drbd_genl_seq);
+       msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO);
+       if (!msg)
+               goto failed;
+
+       err = -EMSGSIZE;
+       d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT);
+       if (!d_out) /* cannot happen, but anyways. */
+               goto nla_put_failure;
+       d_out->minor = mdev_to_minor(mdev);
+       d_out->ret_code = 0;
+
+       if (nla_put_status_info(msg, mdev, sib))
+               goto nla_put_failure;
+       genlmsg_end(msg, d_out);
+       err = drbd_genl_multicast_events(msg, 0);
+       /* msg has been consumed or freed in netlink_broadcast() */
+       if (err && err != -ESRCH)
+               goto failed;
 
-       cn_del_callback(&cn_id_drbd);
-}
+       return;
 
-void drbd_nl_send_reply(struct cn_msg *req, int ret_code)
-{
-       char buffer[sizeof(struct cn_msg)+sizeof(struct drbd_nl_cfg_reply)];
-       struct cn_msg *cn_reply = (struct cn_msg *) buffer;
-       struct drbd_nl_cfg_reply *reply =
-               (struct drbd_nl_cfg_reply *)cn_reply->data;
-       int rr;
-
-       memset(buffer, 0, sizeof(buffer));
-       cn_reply->id = req->id;
-
-       cn_reply->seq = req->seq;
-       cn_reply->ack = req->ack  + 1;
-       cn_reply->len = sizeof(struct drbd_nl_cfg_reply);
-       cn_reply->flags = 0;
-
-       reply->packet_type = P_return_code_only;
-       reply->minor = ((struct drbd_nl_cfg_req *)req->data)->drbd_minor;
-       reply->ret_code = ret_code;
-
-       rr = cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
-       if (rr && rr != -ESRCH)
-               printk(KERN_INFO "drbd: cn_netlink_send()=%d\n", rr);
+nla_put_failure:
+       nlmsg_free(msg);
+failed:
+       dev_err(DEV, "Error %d while broadcasting event. "
+                       "Event seq:%u sib_reason:%u\n",
+                       err, seq, sib->sib_reason);
 }
-
index ffee90d..a280bc2 100644 (file)
@@ -970,6 +970,11 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
        enum drbd_fencing_p fp;
        enum drbd_req_event what = NOTHING;
        union drbd_state nsm = (union drbd_state){ .i = -1 };
+       struct sib_info sib;
+
+       sib.sib_reason = SIB_STATE_CHANGE;
+       sib.os = os;
+       sib.ns = ns;
 
        if (os.conn != C_CONNECTED && ns.conn == C_CONNECTED) {
                clear_bit(CRASHED_PRIMARY, &mdev->flags);
@@ -984,7 +989,7 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
        }
 
        /* Inform userspace about the change... */
-       drbd_bcast_state(mdev, ns);
+       drbd_bcast_event(mdev, &sib);
 
        if (!(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE) &&
            (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE))
index e192167..d28fdd8 100644 (file)
@@ -51,7 +51,6 @@
 
 #endif
 
-
 extern const char *drbd_buildtag(void);
 #define REL_VERSION "8.3.11"
 #define API_VERSION 88
@@ -159,6 +158,7 @@ enum drbd_ret_code {
        ERR_CONN_IN_USE         = 159,
        ERR_MINOR_CONFIGURED    = 160,
        ERR_MINOR_EXISTS        = 161,
+       ERR_INVALID_REQUEST     = 162,
 
        /* insert new ones above this line */
        AFTER_LAST_ERR_CODE
@@ -349,37 +349,4 @@ enum drbd_timeout_flag {
 #define DRBD_MD_INDEX_FLEX_EXT -2
 #define DRBD_MD_INDEX_FLEX_INT -3
 
-/* Start of the new netlink/connector stuff */
-
-enum drbd_ncr_flags {
-       DRBD_NL_CREATE_DEVICE = 0x01,
-       DRBD_NL_SET_DEFAULTS =  0x02,
-};
-#define DRBD_NL_OBJ_NAME_LEN 32
-
-
-/* For searching a vacant cn_idx value */
-#define CN_IDX_STEP                    6977
-
-struct drbd_nl_cfg_req {
-       int packet_type;
-       union {
-               struct {
-                       unsigned int drbd_minor;
-                       enum drbd_ncr_flags flags;
-               };
-               struct {
-                       char obj_name[DRBD_NL_OBJ_NAME_LEN];
-               };
-       };
-       unsigned short tag_list[];
-};
-
-struct drbd_nl_cfg_reply {
-       int packet_type;
-       unsigned int minor;
-       int ret_code; /* enum ret_code or set_st_err_t */
-       unsigned short tag_list[]; /* only used with get_* calls */
-};
-
 #endif
index 8a86f65..c8c6723 100644 (file)
@@ -95,7 +95,7 @@ static struct nla_policy s_name ## _nl_policy[] __read_mostly =               \
 #endif
 #endif
 
-#if 1
+#ifdef GENL_MAGIC_DEBUG
 static void dprint_field(const char *dir, int nla_type,
                const char *name, void *valp)
 {