*/
/*
- * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
* Copyright (c) 2013 Steven Hartland. All rights reserved.
* Copyright (c) 2017 Datto Inc.
* Copyright 2017 RackTop Systems.
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
+#ifdef ZFS_DEBUG
+#include <stdio.h>
+#endif
#include <errno.h>
#include <fcntl.h>
#include <pthread.h>
static pthread_mutex_t g_lock = PTHREAD_MUTEX_INITIALIZER;
static int g_refcount;
+#ifdef ZFS_DEBUG
+static zfs_ioc_t fail_ioc_cmd;
+static zfs_errno_t fail_ioc_err;
+
+static void
+libzfs_core_debug_ioc(void)
+{
+ /*
+ * To test running newer user space binaries with kernel's
+ * that don't yet support an ioctl or a new ioctl arg we
+ * provide an override to intentionally fail an ioctl.
+ *
+ * USAGE:
+ * The override variable, ZFS_IOC_TEST, is of the form "cmd:err"
+ *
+ * For example, to fail a ZFS_IOC_POOL_CHECKPOINT with a
+ * ZFS_ERR_IOC_CMD_UNAVAIL, the string would be "0x5a4d:1029"
+ *
+ * $ sudo sh -c "ZFS_IOC_TEST=0x5a4d:1029 zpool checkpoint tank"
+ * cannot checkpoint 'tank': the loaded zfs module does not support
+ * this operation. A reboot may be required to enable this operation.
+ */
+ if (fail_ioc_cmd == 0) {
+ char *ioc_test = getenv("ZFS_IOC_TEST");
+ unsigned int ioc_num = 0, ioc_err = 0;
+
+ if (ioc_test != NULL &&
+ sscanf(ioc_test, "%i:%i", &ioc_num, &ioc_err) == 2 &&
+ ioc_num < ZFS_IOC_LAST) {
+ fail_ioc_cmd = ioc_num;
+ fail_ioc_err = ioc_err;
+ }
+ }
+}
+#endif
+
int
libzfs_core_init(void)
{
}
}
g_refcount++;
+
+#ifdef ZFS_DEBUG
+ libzfs_core_debug_ioc();
+#endif
(void) pthread_mutex_unlock(&g_lock);
return (0);
}
ASSERT3S(g_refcount, >, 0);
VERIFY3S(g_fd, !=, -1);
+#ifdef ZFS_DEBUG
+ if (ioc == fail_ioc_cmd)
+ return (fail_ioc_err);
+#endif
+
if (name != NULL)
(void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name));
return (0);
}
+int
+lzc_remap(const char *fsname)
+{
+ int error;
+ nvlist_t *args = fnvlist_alloc();
+ error = lzc_ioctl(ZFS_IOC_REMAP, fsname, args, NULL);
+ nvlist_free(args);
+ return (error);
+}
+
+int
+lzc_rename(const char *source, const char *target)
+{
+ zfs_cmd_t zc = { "\0" };
+ int error;
+ ASSERT3S(g_refcount, >, 0);
+ VERIFY3S(g_fd, !=, -1);
+ (void) strlcpy(zc.zc_name, source, sizeof (zc.zc_name));
+ (void) strlcpy(zc.zc_value, target, sizeof (zc.zc_value));
+ error = ioctl(g_fd, ZFS_IOC_RENAME, &zc);
+ if (error != 0)
+ error = errno;
+ return (error);
+}
+int
+lzc_destroy(const char *fsname)
+{
+ int error;
+ nvlist_t *args = fnvlist_alloc();
+ error = lzc_ioctl(ZFS_IOC_DESTROY, fsname, args, NULL);
+ nvlist_free(args);
+ return (error);
+}
+
/*
* Creates snapshots.
*
* to contain DRR_WRITE_EMBEDDED records with drr_etype==BP_EMBEDDED_TYPE_DATA,
* which the receiving system must support (as indicated by support
* for the "embedded_data" feature).
+ *
+ * If "flags" contains LZC_SEND_FLAG_COMPRESS, the stream is generated by using
+ * compressed WRITE records for blocks which are compressed on disk and in
+ * memory. If the lz4_compress feature is active on the sending system, then
+ * the receiving system must have that feature enabled as well.
+ *
+ * If "flags" contains LZC_SEND_FLAG_RAW, the stream is generated, for encrypted
+ * datasets, by sending data exactly as it exists on disk. This allows backups
+ * to be taken even if encryption keys are not currently loaded.
*/
int
lzc_send(const char *snapname, const char *from, int fd,
*/
static int
recv_impl(const char *snapname, nvlist_t *recvdprops, nvlist_t *localprops,
- const char *origin, boolean_t force, boolean_t resumable, boolean_t raw,
- int input_fd, const dmu_replay_record_t *begin_record, int cleanup_fd,
+ uint8_t *wkeydata, uint_t wkeylen, const char *origin, boolean_t force,
+ boolean_t resumable, boolean_t raw, int input_fd,
+ const dmu_replay_record_t *begin_record, int cleanup_fd,
uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle,
nvlist_t **errors)
{
drr = *begin_record;
}
- if (resumable || raw) {
+ /*
+ * Raw receives, resumable receives, and receives that include a
+ * wrapping key all use the new interface.
+ */
+ if (resumable || raw || wkeydata != NULL) {
nvlist_t *outnvl = NULL;
nvlist_t *innvl = fnvlist_alloc();
if (localprops != NULL)
fnvlist_add_nvlist(innvl, "localprops", localprops);
+ if (wkeydata != NULL) {
+ /*
+ * wkeydata must be placed in the special
+ * ZPOOL_HIDDEN_ARGS nvlist so that it
+ * will not be printed to the zpool history.
+ */
+ nvlist_t *hidden_args = fnvlist_alloc();
+ fnvlist_add_uint8_array(hidden_args, "wkeydata",
+ wkeydata, wkeylen);
+ fnvlist_add_nvlist(innvl, ZPOOL_HIDDEN_ARGS,
+ hidden_args);
+ nvlist_free(hidden_args);
+ }
+
if (origin != NULL && strlen(origin))
fnvlist_add_string(innvl, "origin", origin);
lzc_receive(const char *snapname, nvlist_t *props, const char *origin,
boolean_t force, boolean_t raw, int fd)
{
- return (recv_impl(snapname, props, NULL, origin, force, B_FALSE, raw,
- fd, NULL, -1, NULL, NULL, NULL, NULL));
+ return (recv_impl(snapname, props, NULL, NULL, 0, origin, force,
+ B_FALSE, raw, fd, NULL, -1, NULL, NULL, NULL, NULL));
}
/*
lzc_receive_resumable(const char *snapname, nvlist_t *props, const char *origin,
boolean_t force, boolean_t raw, int fd)
{
- return (recv_impl(snapname, props, NULL, origin, force, B_TRUE, raw,
- fd, NULL, -1, NULL, NULL, NULL, NULL));
+ return (recv_impl(snapname, props, NULL, NULL, 0, origin, force,
+ B_TRUE, raw, fd, NULL, -1, NULL, NULL, NULL, NULL));
}
/*
if (begin_record == NULL)
return (EINVAL);
- return (recv_impl(snapname, props, NULL, origin, force, resumable, raw,
- fd, begin_record, -1, NULL, NULL, NULL, NULL));
+ return (recv_impl(snapname, props, NULL, NULL, 0, origin, force,
+ resumable, raw, fd, begin_record, -1, NULL, NULL, NULL, NULL));
}
/*
uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle,
nvlist_t **errors)
{
- return (recv_impl(snapname, props, NULL, origin, force, resumable,
- raw, input_fd, begin_record, cleanup_fd, read_bytes, errflags,
- action_handle, errors));
+ return (recv_impl(snapname, props, NULL, NULL, 0, origin, force,
+ resumable, raw, input_fd, begin_record, cleanup_fd, read_bytes,
+ errflags, action_handle, errors));
}
/*
* this nvlist
*/
int lzc_receive_with_cmdprops(const char *snapname, nvlist_t *props,
- nvlist_t *cmdprops, const char *origin, boolean_t force,
- boolean_t resumable, boolean_t raw, int input_fd,
+ nvlist_t *cmdprops, uint8_t *wkeydata, uint_t wkeylen, const char *origin,
+ boolean_t force, boolean_t resumable, boolean_t raw, int input_fd,
const dmu_replay_record_t *begin_record, int cleanup_fd,
uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle,
nvlist_t **errors)
{
- return (recv_impl(snapname, props, cmdprops, origin, force, resumable,
- raw, input_fd, begin_record, cleanup_fd, read_bytes, errflags,
- action_handle, errors));
+ return (recv_impl(snapname, props, cmdprops, wkeydata, wkeylen, origin,
+ force, resumable, raw, input_fd, begin_record, cleanup_fd,
+ read_bytes, errflags, action_handle, errors));
}
/*
return (error);
}
+static int
+lzc_channel_program_impl(const char *pool, const char *program, boolean_t sync,
+ uint64_t instrlimit, uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl)
+{
+ int error;
+ nvlist_t *args;
+
+ args = fnvlist_alloc();
+ fnvlist_add_string(args, ZCP_ARG_PROGRAM, program);
+ fnvlist_add_nvlist(args, ZCP_ARG_ARGLIST, argnvl);
+ fnvlist_add_boolean_value(args, ZCP_ARG_SYNC, sync);
+ fnvlist_add_uint64(args, ZCP_ARG_INSTRLIMIT, instrlimit);
+ fnvlist_add_uint64(args, ZCP_ARG_MEMLIMIT, memlimit);
+ error = lzc_ioctl(ZFS_IOC_CHANNEL_PROGRAM, pool, args, outnvl);
+ fnvlist_free(args);
+
+ return (error);
+}
+
/*
* Executes a channel program.
*
int
lzc_channel_program(const char *pool, const char *program, uint64_t instrlimit,
uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl)
+{
+ return (lzc_channel_program_impl(pool, program, B_TRUE, instrlimit,
+ memlimit, argnvl, outnvl));
+}
+
+/*
+ * Creates a checkpoint for the specified pool.
+ *
+ * If this function returns 0 the pool was successfully checkpointed.
+ *
+ * This method may also return:
+ *
+ * ZFS_ERR_CHECKPOINT_EXISTS
+ * The pool already has a checkpoint. A pools can only have one
+ * checkpoint at most, at any given time.
+ *
+ * ZFS_ERR_DISCARDING_CHECKPOINT
+ * ZFS is in the middle of discarding a checkpoint for this pool.
+ * The pool can be checkpointed again once the discard is done.
+ *
+ * ZFS_DEVRM_IN_PROGRESS
+ * A vdev is currently being removed. The pool cannot be
+ * checkpointed until the device removal is done.
+ *
+ * ZFS_VDEV_TOO_BIG
+ * One or more top-level vdevs exceed the maximum vdev size
+ * supported for this feature.
+ */
+int
+lzc_pool_checkpoint(const char *pool)
{
int error;
- nvlist_t *args;
- args = fnvlist_alloc();
- fnvlist_add_string(args, ZCP_ARG_PROGRAM, program);
- fnvlist_add_nvlist(args, ZCP_ARG_ARGLIST, argnvl);
- fnvlist_add_uint64(args, ZCP_ARG_INSTRLIMIT, instrlimit);
- fnvlist_add_uint64(args, ZCP_ARG_MEMLIMIT, memlimit);
- error = lzc_ioctl(ZFS_IOC_CHANNEL_PROGRAM, pool, args, outnvl);
+ nvlist_t *result = NULL;
+ nvlist_t *args = fnvlist_alloc();
+
+ error = lzc_ioctl(ZFS_IOC_POOL_CHECKPOINT, pool, args, &result);
+
fnvlist_free(args);
+ fnvlist_free(result);
return (error);
}
+/*
+ * Discard the checkpoint from the specified pool.
+ *
+ * If this function returns 0 the checkpoint was successfully discarded.
+ *
+ * This method may also return:
+ *
+ * ZFS_ERR_NO_CHECKPOINT
+ * The pool does not have a checkpoint.
+ *
+ * ZFS_ERR_DISCARDING_CHECKPOINT
+ * ZFS is already in the middle of discarding the checkpoint.
+ */
+int
+lzc_pool_checkpoint_discard(const char *pool)
+{
+ int error;
+
+ nvlist_t *result = NULL;
+ nvlist_t *args = fnvlist_alloc();
+
+ error = lzc_ioctl(ZFS_IOC_POOL_DISCARD_CHECKPOINT, pool, args, &result);
+
+ fnvlist_free(args);
+ fnvlist_free(result);
+
+ return (error);
+}
+
+/*
+ * Executes a read-only channel program.
+ *
+ * A read-only channel program works programmatically the same way as a
+ * normal channel program executed with lzc_channel_program(). The only
+ * difference is it runs exclusively in open-context and therefore can
+ * return faster. The downside to that, is that the program cannot change
+ * on-disk state by calling functions from the zfs.sync submodule.
+ *
+ * The return values of this function (and their meaning) are exactly the
+ * same as the ones described in lzc_channel_program().
+ */
+int
+lzc_channel_program_nosync(const char *pool, const char *program,
+ uint64_t timeout, uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl)
+{
+ return (lzc_channel_program_impl(pool, program, B_FALSE, timeout,
+ memlimit, argnvl, outnvl));
+}
+
/*
* Performs key management functions
*
- * crypto_cmd should be a value from zfs_ioc_crypto_cmd_t. If the command
- * specifies to load or change a wrapping key, the key should be specified in
- * the hidden_args nvlist so that it is not logged
+ * crypto_cmd should be a value from dcp_cmd_t. If the command specifies to
+ * load or change a wrapping key, the key should be specified in the
+ * hidden_args nvlist so that it is not logged.
*/
int
lzc_load_key(const char *fsname, boolean_t noop, uint8_t *wkeydata,
nvlist_free(args);
return (error);
}
+
+/*
+ * Changes initializing state.
+ *
+ * vdevs should be a list of (<key>, guid) where guid is a uint64 vdev GUID.
+ * The key is ignored.
+ *
+ * If there are errors related to vdev arguments, per-vdev errors are returned
+ * in an nvlist with the key "vdevs". Each error is a (guid, errno) pair where
+ * guid is stringified with PRIu64, and errno is one of the following as
+ * an int64_t:
+ * - ENODEV if the device was not found
+ * - EINVAL if the devices is not a leaf or is not concrete (e.g. missing)
+ * - EROFS if the device is not writeable
+ * - EBUSY start requested but the device is already being initialized
+ * - ESRCH cancel/suspend requested but device is not being initialized
+ *
+ * If the errlist is empty, then return value will be:
+ * - EINVAL if one or more arguments was invalid
+ * - Other spa_open failures
+ * - 0 if the operation succeeded
+ */
+int
+lzc_initialize(const char *poolname, pool_initialize_func_t cmd_type,
+ nvlist_t *vdevs, nvlist_t **errlist)
+{
+ int error;
+ nvlist_t *args = fnvlist_alloc();
+ fnvlist_add_uint64(args, ZPOOL_INITIALIZE_COMMAND, (uint64_t)cmd_type);
+ fnvlist_add_nvlist(args, ZPOOL_INITIALIZE_VDEVS, vdevs);
+
+ error = lzc_ioctl(ZFS_IOC_POOL_INITIALIZE, poolname, args, errlist);
+
+ fnvlist_free(args);
+
+ return (error);
+}