* Lustre files while hiding details of the internal data structures
* from the user.
*
- * Copyright (c) 2016, Intel Corporation.
+ * Copyright (c) 2016, 2017, Intel Corporation.
*
* Author: Ned Bass <bass6@llnl.gov>
*/
#include <unistd.h>
#include <errno.h>
#include <limits.h>
+#include <assert.h>
#include <sys/xattr.h>
+#include <sys/param.h>
#include <libcfs/util/list.h>
#include <lustre/lustreapi.h>
struct lu_extent llc_extent; /* [start, end) of component */
uint32_t llc_id; /* unique ID of component */
uint32_t llc_flags; /* LCME_FL_* flags */
+ uint64_t llc_timestamp; /* snapshot timestamp */
struct list_head llc_list; /* linked to the llapi_layout
components list */
};
uint32_t llot_gen;
uint32_t llot_flags;
bool llot_is_composite;
+ uint16_t llot_mirror_count;
/* Cursor pointing to one of the components in llot_comp_list */
struct llapi_layout_comp *llot_cur_comp;
struct list_head llot_comp_list;
ent = &comp_v1->lcm_entries[i];
__swab32s(&ent->lcme_id);
__swab32s(&ent->lcme_flags);
+ __swab64s(&ent->lcme_timestamp);
__swab64s(&ent->lcme_extent.e_start);
__swab64s(&ent->lcme_extent.e_end);
__swab32s(&ent->lcme_offset);
layout->llot_gen = 0;
layout->llot_flags = 0;
layout->llot_is_composite = false;
+ layout->llot_mirror_count = 1;
layout->llot_cur_comp = NULL;
INIT_LIST_HEAD(&layout->llot_comp_list);
* Convert the data from a lov_user_md to a newly allocated llapi_layout.
* The caller is responsible for freeing the returned pointer.
*
- * \param[in] lum LOV user metadata structure to copy data from
- * \param[in] lum_size size the the lum passed in
+ * \param[in] lov_xattr LOV user metadata xattr to copy data from
+ * \param[in] lov_xattr_size size the lov_xattr_size passed in
*
* \retval valid llapi_layout pointer on success
* \retval NULL if memory allocation fails
*/
-static struct llapi_layout *
-llapi_layout_from_lum(const struct lov_user_md *lum, int lum_size)
+struct llapi_layout *llapi_layout_get_by_xattr(const void *lov_xattr,
+ ssize_t lov_xattr_size)
{
+ const struct lov_user_md *lum = lov_xattr;
struct lov_comp_md_v1 *comp_v1 = NULL;
struct lov_comp_md_entry_v1 *ent;
struct lov_user_md *v1;
if (lum->lmm_magic == LOV_MAGIC_COMP_V1) {
comp_v1 = (struct lov_comp_md_v1 *)lum;
ent_count = comp_v1->lcm_entry_count;
+ layout->llot_gen = comp_v1->lcm_layout_gen;
layout->llot_is_composite = true;
+ layout->llot_mirror_count = comp_v1->lcm_mirror_count + 1;
layout->llot_gen = comp_v1->lcm_layout_gen;
layout->llot_flags = comp_v1->lcm_flags;
} else if (lum->lmm_magic == LOV_MAGIC_V1 ||
lum->lmm_magic == LOV_MAGIC_V3) {
ent_count = 1;
layout->llot_is_composite = false;
+
+ if (lov_xattr_size <= 0) {
+ errno = EINVAL;
+ goto error;
+ }
+ } else {
+ errno = EOPNOTSUPP;
+ goto error;
}
if (ent_count == 0) {
ent = &comp_v1->lcm_entries[i];
v1 = (struct lov_user_md *)((char *)comp_v1 +
ent->lcme_offset);
- lum_size = ent->lcme_size;
+ lov_xattr_size = ent->lcme_size;
} else {
ent = NULL;
}
- obj_count = llapi_layout_objects_in_lum(v1, lum_size);
+ obj_count = llapi_layout_objects_in_lum(v1, lov_xattr_size);
comp = __llapi_comp_alloc(obj_count);
if (comp == NULL)
goto error;
comp->llc_extent.e_end = ent->lcme_extent.e_end;
comp->llc_id = ent->lcme_id;
comp->llc_flags = ent->lcme_flags;
+ if (comp->llc_flags & LCME_FL_NOSYNC)
+ comp->llc_timestamp = ent->lcme_timestamp;
} else {
comp->llc_extent.e_start = 0;
comp->llc_extent.e_end = LUSTRE_EOF;
comp_cnt++;
lum_size = sizeof(*comp_v1) + comp_cnt * sizeof(*ent);
- lum = malloc(lum_size);
+ lum = calloc(lum_size, 1);
if (lum == NULL) {
errno = ENOMEM;
return NULL;
comp_v1->lcm_magic = LOV_USER_MAGIC_COMP_V1;
comp_v1->lcm_size = lum_size;
comp_v1->lcm_layout_gen = 0;
- comp_v1->lcm_flags = 0;
+ comp_v1->lcm_flags = layout->llot_flags;
comp_v1->lcm_entry_count = comp_cnt;
+ comp_v1->lcm_mirror_count = layout->llot_mirror_count - 1;
offset += lum_size;
}
blob->lmm_magic = magic;
if (pattern == LLAPI_LAYOUT_DEFAULT)
- blob->lmm_pattern = 0;
- else if (pattern == LLAPI_LAYOUT_RAID0)
blob->lmm_pattern = LOV_PATTERN_RAID0;
else if (pattern == LLAPI_LAYOUT_MDT)
blob->lmm_pattern = LOV_PATTERN_MDT;
ent = &comp_v1->lcm_entries[ent_idx];
ent->lcme_id = comp->llc_id;
ent->lcme_flags = comp->llc_flags;
+ if (ent->lcme_flags & LCME_FL_NOSYNC)
+ ent->lcme_timestamp = comp->llc_timestamp;
ent->lcme_extent.e_start = comp->llc_extent.e_start;
ent->lcme_extent.e_end = comp->llc_extent.e_end;
ent->lcme_size = blob_size;
if (comp == NULL)
return false;
- if (layout->llot_is_composite)
+ if (layout->llot_is_composite || layout->llot_mirror_count != 1)
return true;
return comp->llc_pattern != LLAPI_LAYOUT_DEFAULT ||
goto out;
}
- layout = llapi_layout_from_lum(lum, bytes_read);
+ layout = llapi_layout_get_by_xattr(lum, bytes_read);
out:
free(lum);
return layout;
return -1;
if (pattern != LLAPI_LAYOUT_DEFAULT &&
- pattern != LLAPI_LAYOUT_RAID0 &&
- pattern != LLAPI_LAYOUT_MDT) {
+ pattern != LLAPI_LAYOUT_RAID0 && pattern != LLAPI_LAYOUT_MDT) {
errno = EOPNOTSUPP;
return -1;
}
*
* \param[in] path name of the file to open
* \param[in] open_flags open() flags
- * \param[in] mode permissions to create new file with
+ * \param[in] mode permissions to create file, filtered by umask
* \param[in] layout layout to create new file with
*
* \retval non-negative file descriptor on successful open
layout);
}
+int llapi_layout_flags_get(struct llapi_layout *layout, uint32_t *flags)
+{
+ if (layout->llot_magic != LLAPI_LAYOUT_MAGIC) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ *flags = layout->llot_flags;
+ return 0;
+}
+
+/**
+ * Set flags to the header of a component layout.
+ */
+int llapi_layout_flags_set(struct llapi_layout *layout, uint32_t flags)
+{
+ if (layout->llot_magic != LLAPI_LAYOUT_MAGIC) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ layout->llot_flags = flags;
+ return 0;
+}
+
+const char *llapi_layout_flags_string(uint32_t flags)
+{
+ switch (flags & LCM_FL_FLR_MASK) {
+ case LCM_FL_RDONLY:
+ return "ro";
+ case LCM_FL_WRITE_PENDING:
+ return "wp";
+ case LCM_FL_SYNC_PENDING:
+ return "sp";
+ }
+
+ return "0";
+}
+
+const __u16 llapi_layout_string_flags(char *string)
+{
+ if (strncmp(string, "ro", strlen(string)) == 0)
+ return LCM_FL_RDONLY;
+ if (strncmp(string, "wp", strlen(string)) == 0)
+ return LCM_FL_WRITE_PENDING;
+ if (strncmp(string, "sp", strlen(string)) == 0)
+ return LCM_FL_SYNC_PENDING;
+
+ return 0;
+}
+
+/**
+ * llapi_layout_mirror_count_is_valid() - Check the validity of mirror count.
+ * @count: Mirror count value to be checked.
+ *
+ * This function checks the validity of mirror count.
+ *
+ * Return: true on success or false on failure.
+ */
+static bool llapi_layout_mirror_count_is_valid(uint16_t count)
+{
+ return count >= 0 && count <= LUSTRE_MIRROR_COUNT_MAX;
+}
+
+/**
+ * llapi_layout_mirror_count_get() - Get mirror count from the header of
+ * a layout.
+ * @layout: Layout to get mirror count from.
+ * @count: Returned mirror count value.
+ *
+ * This function gets mirror count from the header of a layout.
+ *
+ * Return: 0 on success or -1 on failure.
+ */
+int llapi_layout_mirror_count_get(struct llapi_layout *layout,
+ uint16_t *count)
+{
+ if (layout->llot_magic != LLAPI_LAYOUT_MAGIC) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ *count = layout->llot_mirror_count;
+ return 0;
+}
+
+/**
+ * llapi_layout_mirror_count_set() - Set mirror count to the header of a layout.
+ * @layout: Layout to set mirror count in.
+ * @count: Mirror count value to be set.
+ *
+ * This function sets mirror count to the header of a layout.
+ *
+ * Return: 0 on success or -1 on failure.
+ */
+int llapi_layout_mirror_count_set(struct llapi_layout *layout,
+ uint16_t count)
+{
+ if (layout->llot_magic != LLAPI_LAYOUT_MAGIC) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ if (!llapi_layout_mirror_count_is_valid(count)) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ layout->llot_mirror_count = count;
+ return 0;
+}
+
/**
* Fetch the start and end offset of the current layout component.
*
if (comp->llc_list.prev != &layout->llot_comp_list) {
prev = list_entry(comp->llc_list.prev, typeof(*prev),
llc_list);
- if (start != prev->llc_extent.e_end) {
+ if (start != 0 && start != prev->llc_extent.e_end) {
errno = EINVAL;
return -1;
}
if (comp->llc_list.next != &layout->llot_comp_list) {
next = list_entry(comp->llc_list.next, typeof(*next),
llc_list);
- if (end != next->llc_extent.e_start) {
+ if (next->llc_extent.e_start != 0 &&
+ end != next->llc_extent.e_start) {
errno = EINVAL;
return -1;
}
}
/**
+ * Return the mirror id of the current layout component.
+ *
+ * \param[in] layout the layout component
+ * \param[out] id stored the returned mirror ID
+ *
+ * \retval 0 on success
+ * \retval <0 if error occurs
+ */
+int llapi_layout_mirror_id_get(const struct llapi_layout *layout, uint32_t *id)
+{
+ struct llapi_layout_comp *comp;
+
+ comp = __llapi_layout_cur_comp(layout);
+ if (comp == NULL)
+ return -1;
+
+ if (id == NULL) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ *id = mirror_id_of(comp->llc_id);
+
+ return 0;
+}
+
+/**
* Adds a component to \a layout, the new component will be added to
* the tail of components list and it'll inherit attributes of existing
* ones. The \a layout will change it's current component pointer to
last = list_entry(layout->llot_comp_list.prev, typeof(*last),
llc_list);
- /* Inherit some attributes from existing component */
- new->llc_stripe_size = comp->llc_stripe_size;
- new->llc_stripe_count = comp->llc_stripe_count;
- if (comp->llc_pool_name[0] != '\0')
- strncpy(new->llc_pool_name, comp->llc_pool_name,
- sizeof(comp->llc_pool_name));
if (new->llc_extent.e_end <= last->llc_extent.e_end) {
__llapi_comp_free(new);
errno = EINVAL;
return 0;
}
+/**
+ * Adds a first component of a mirror to \a layout.
+ * The \a layout will change it's current component pointer to
+ * the newly added component, and it'll be turned into a composite
+ * layout if it was not before the adding.
+ *
+ * \param[in] layout existing composite or plain layout
+ *
+ * \retval 0 on success
+ * \retval <0 if error occurs
+ */
+int llapi_layout_add_first_comp(struct llapi_layout *layout)
+{
+ struct llapi_layout_comp *comp, *new;
+
+ comp = __llapi_layout_cur_comp(layout);
+ if (comp == NULL)
+ return -1;
+
+ new = __llapi_comp_alloc(0);
+ if (new == NULL)
+ return -1;
+
+ new->llc_extent.e_start = 0;
+
+ list_add_tail(&new->llc_list, &layout->llot_comp_list);
+ layout->llot_cur_comp = new;
+ layout->llot_is_composite = true;
+
+ return 0;
+}
/**
* Deletes current component from the composite layout. The component
* comp->lcme_id value, which must be an unique component ID. The new
* attributes are passed in by @comp and @valid is used to specify which
* attributes in the component are going to be changed.
+ *
+ * \param[in] path path name of the file
+ * \param[in] ids An array of component IDs
+ * \param[in] flags flags: LCME_FL_* or;
+ * negative flags: (LCME_FL_NEG|LCME_FL_*)
+ * \param[in] count Number of elements in ids and flags array
*/
-int llapi_layout_file_comp_set(const char *path,
- const struct llapi_layout *comp,
- uint32_t valid)
+int llapi_layout_file_comp_set(const char *path, uint32_t *ids, uint32_t *flags,
+ size_t count)
{
- errno = EOPNOTSUPP;
- return -1;
+ int rc = -1, fd = -1, i;
+ size_t lum_size;
+ struct llapi_layout *layout;
+ struct llapi_layout_comp *comp;
+ struct lov_user_md *lum = NULL;
+
+ if (path == NULL) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ if (!count)
+ return 0;
+
+ for (i = 0; i < count; i++) {
+ if (!ids[i] || !flags[i]) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ if (ids[i] > LCME_ID_MAX || (flags[i] & ~LCME_KNOWN_FLAGS)) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ /* do not allow to set or clear INIT flag */
+ if (flags[i] & LCME_FL_INIT) {
+ errno = EINVAL;
+ return -1;
+ }
+ }
+
+ layout = __llapi_layout_alloc();
+ if (layout == NULL)
+ return -1;
+
+ layout->llot_is_composite = true;
+ for (i = 0; i < count; i++) {
+ comp = __llapi_comp_alloc(0);
+ if (comp == NULL)
+ goto out;
+
+ comp->llc_id = ids[i];
+ comp->llc_flags = flags[i];
+
+ list_add_tail(&comp->llc_list, &layout->llot_comp_list);
+ layout->llot_cur_comp = comp;
+ }
+
+ lum = llapi_layout_to_lum(layout);
+ if (lum == NULL)
+ goto out;
+
+ lum_size = ((struct lov_comp_md_v1 *)lum)->lcm_size;
+
+ fd = open(path, O_RDWR);
+ if (fd < 0)
+ goto out;
+
+ /* flush cached pages from clients */
+ rc = llapi_file_flush(fd);
+ if (rc) {
+ errno = -rc;
+ rc = -1;
+ goto out_close;
+ }
+
+ rc = fsetxattr(fd, XATTR_LUSTRE_LOV".set.flags", lum, lum_size, 0);
+ if (rc < 0)
+ goto out_close;
+
+ rc = 0;
+
+out_close:
+ if (fd >= 0) {
+ int tmp_errno = errno;
+ close(fd);
+ errno = tmp_errno;
+ }
+out:
+ if (lum)
+ free(lum);
+ llapi_layout_free(layout);
+ return rc;
}
/**
{
return layout->llot_is_composite;
}
+
+/**
+ * Iterate every components in the @layout and call callback function @cb.
+ *
+ * \param[in] layout component layout list.
+ * \param[in] cb callback for each component
+ * \param[in] cbdata callback data
+ *
+ * \retval < 0 error happens during the iteration
+ * \retval LLAPI_LAYOUT_ITER_CONT finished the iteration w/o error
+ * \retval LLAPI_LAYOUT_ITER_STOP got something, stop the iteration
+ */
+int llapi_layout_comp_iterate(struct llapi_layout *layout,
+ llapi_layout_iter_cb cb, void *cbdata)
+{
+ int rc;
+
+ rc = llapi_layout_comp_use(layout, LLAPI_LAYOUT_COMP_USE_FIRST);
+ if (rc < 0)
+ return rc;
+
+ /**
+ * make sure on success llapi_layout_comp_use() API returns 0 with
+ * USE_FIRST.
+ */
+ assert(rc == 0);
+
+ while (1) {
+ rc = cb(layout, cbdata);
+ if (rc != LLAPI_LAYOUT_ITER_CONT)
+ break;
+
+ rc = llapi_layout_comp_use(layout, LLAPI_LAYOUT_COMP_USE_NEXT);
+ if (rc < 0)
+ return rc;
+ else if (rc == 1) /* reached the last comp */
+ return LLAPI_LAYOUT_ITER_CONT;
+ }
+
+ return rc;
+}
+
+/**
+ * llapi_layout_merge() - Merge a composite layout into another one.
+ * @dst_layout: Destination composite layout.
+ * @src_layout: Source composite layout.
+ *
+ * This function copies all of the components from @src_layout and
+ * appends them to @dst_layout.
+ *
+ * Return: 0 on success or -1 on failure.
+ */
+int llapi_layout_merge(struct llapi_layout **dst_layout,
+ const struct llapi_layout *src_layout)
+{
+ struct llapi_layout *new_layout = *dst_layout;
+ struct llapi_layout_comp *new = NULL;
+ struct llapi_layout_comp *comp = NULL;
+ int i = 0;
+
+ if (src_layout == NULL ||
+ list_empty((struct list_head *)&src_layout->llot_comp_list))
+ return 0;
+
+ if (new_layout == NULL) {
+ new_layout = __llapi_layout_alloc();
+ if (new_layout == NULL) {
+ errno = ENOMEM;
+ return -1;
+ }
+ }
+
+ list_for_each_entry(comp, &src_layout->llot_comp_list, llc_list) {
+ new = __llapi_comp_alloc(0);
+ if (new == NULL) {
+ errno = ENOMEM;
+ goto error;
+ }
+
+ new->llc_pattern = comp->llc_pattern;
+ new->llc_stripe_size = comp->llc_stripe_size;
+ new->llc_stripe_count = comp->llc_stripe_count;
+ new->llc_stripe_offset = comp->llc_stripe_offset;
+
+ if (comp->llc_pool_name[0] != '\0')
+ strncpy(new->llc_pool_name, comp->llc_pool_name,
+ sizeof(new->llc_pool_name));
+
+ for (i = 0; i < comp->llc_objects_count; i++) {
+ if (__llapi_comp_objects_realloc(new,
+ stripe_number_roundup(i)) < 0) {
+ errno = EINVAL;
+ __llapi_comp_free(new);
+ goto error;
+ }
+ new->llc_objects[i].l_ost_idx = \
+ comp->llc_objects[i].l_ost_idx;
+ }
+
+ new->llc_objects_count = comp->llc_objects_count;
+ new->llc_extent.e_start = comp->llc_extent.e_start;
+ new->llc_extent.e_end = comp->llc_extent.e_end;
+ new->llc_id = comp->llc_id;
+ new->llc_flags = comp->llc_flags;
+
+ list_add_tail(&new->llc_list, &new_layout->llot_comp_list);
+ new_layout->llot_cur_comp = new;
+ }
+ new_layout->llot_is_composite = true;
+
+ *dst_layout = new_layout;
+ return 0;
+error:
+ llapi_layout_free(new_layout);
+ return -1;
+}
+
+/**
+ * Find all stale components.
+ *
+ * \param[in] layout component layout list.
+ * \param[out] comp array of stale component info.
+ * \param[in] comp_size array size of @comp.
+ * \param[in] mirror_ids array of mirror id that only components
+ * belonging to these mirror will be collected.
+ * \param[in] ids_nr number of mirror ids array.
+ *
+ * \retval number of component info collected on sucess or
+ * an error code on failure.
+ */
+int llapi_mirror_find_stale(struct llapi_layout *layout,
+ struct llapi_resync_comp *comp, size_t comp_size,
+ __u16 *mirror_ids, int ids_nr)
+{
+ int idx = 0;
+ int rc;
+
+ rc = llapi_layout_comp_use(layout, LLAPI_LAYOUT_COMP_USE_FIRST);
+ if (rc < 0)
+ goto error;
+
+ while (rc == 0) {
+ uint32_t id;
+ uint32_t mirror_id;
+ uint32_t flags;
+ uint64_t start, end;
+
+ rc = llapi_layout_comp_flags_get(layout, &flags);
+ if (rc < 0)
+ goto error;
+
+ if (!(flags & LCME_FL_STALE))
+ goto next;
+
+ rc = llapi_layout_mirror_id_get(layout, &mirror_id);
+ if (rc < 0)
+ goto error;
+
+ /* the caller only wants stale components from specific
+ * mirrors */
+ if (ids_nr > 0) {
+ int j;
+
+ for (j = 0; j < ids_nr; j++) {
+ if (mirror_ids[j] == mirror_id)
+ break;
+ }
+
+ /* not in the specified mirror */
+ if (j == ids_nr)
+ goto next;
+ } else if (flags & LCME_FL_NOSYNC) {
+ /* if not specified mirrors, do not resync "nosync"
+ * mirrors */
+ goto next;
+ }
+
+ rc = llapi_layout_comp_id_get(layout, &id);
+ if (rc < 0)
+ goto error;
+
+ rc = llapi_layout_comp_extent_get(layout, &start, &end);
+ if (rc < 0)
+ goto error;
+
+ /* pack this component into @comp array */
+ comp[idx].lrc_id = id;
+ comp[idx].lrc_mirror_id = mirror_id;
+ comp[idx].lrc_start = start;
+ comp[idx].lrc_end = end;
+ idx++;
+
+ if (idx >= comp_size) {
+ rc = -EINVAL;
+ goto error;
+ }
+
+ next:
+ rc = llapi_layout_comp_use(layout, LLAPI_LAYOUT_COMP_USE_NEXT);
+ if (rc < 0) {
+ rc = -EINVAL;
+ goto error;
+ }
+ }
+error:
+ return rc < 0 ? rc : idx;
+}
+
+/* locate @layout to a valid component covering file [file_start, file_end) */
+uint32_t llapi_mirror_find(struct llapi_layout *layout,
+ uint64_t file_start, uint64_t file_end,
+ uint64_t *endp)
+{
+ uint32_t mirror_id = 0;
+ int rc;
+
+ rc = llapi_layout_comp_use(layout, LLAPI_LAYOUT_COMP_USE_FIRST);
+ if (rc < 0)
+ return rc;
+
+ *endp = 0;
+ while (rc == 0) {
+ uint64_t start, end;
+ uint32_t flags, id, rid;
+
+ rc = llapi_layout_comp_flags_get(layout, &flags);
+ if (rc < 0)
+ return rc;
+
+ if (flags & LCME_FL_STALE)
+ goto next;
+
+ rc = llapi_layout_mirror_id_get(layout, &rid);
+ if (rc < 0)
+ return rc;
+
+ rc = llapi_layout_comp_id_get(layout, &id);
+ if (rc < 0)
+ return rc;
+
+ rc = llapi_layout_comp_extent_get(layout, &start, &end);
+ if (rc < 0)
+ return rc;
+
+ if (file_start >= start && file_start < end) {
+ if (!mirror_id)
+ mirror_id = rid;
+ else if (mirror_id != rid || *endp != start)
+ break;
+
+ file_start = *endp = end;
+ if (end >= file_end)
+ break;
+ }
+
+ next:
+ rc = llapi_layout_comp_use(layout, LLAPI_LAYOUT_COMP_USE_NEXT);
+ if (rc < 0)
+ return rc;
+ }
+
+ return mirror_id;
+}
+
+int llapi_mirror_resync_many(int fd, struct llapi_layout *layout,
+ struct llapi_resync_comp *comp_array,
+ int comp_size, uint64_t start, uint64_t end)
+{
+ uint64_t count;
+ size_t page_size = sysconf(_SC_PAGESIZE);
+ const size_t buflen = 4 << 20; /* 4M */
+ void *buf;
+ uint64_t pos = start;
+ int i;
+ int rc;
+
+ rc = posix_memalign(&buf, page_size, buflen);
+ if (rc)
+ return -rc;
+
+ if (end == OBD_OBJECT_EOF)
+ count = OBD_OBJECT_EOF;
+ else
+ count = end - start;
+
+ while (count > 0) {
+ uint32_t src;
+ uint64_t mirror_end = 0;
+ uint64_t bytes_left;
+ ssize_t bytes_read;
+ size_t to_read;
+ size_t to_write;
+
+ src = llapi_mirror_find(layout, pos, end, &mirror_end);
+ if (src == 0)
+ return -ENOENT;
+
+ if (mirror_end == OBD_OBJECT_EOF) {
+ bytes_left = count;
+ } else {
+ bytes_left = MIN(count, mirror_end - pos);
+ bytes_left = ((bytes_left - 1) | (page_size - 1)) + 1;
+ }
+ to_read = MIN(buflen, bytes_left);
+
+ bytes_read = llapi_mirror_read(fd, src, buf, to_read, pos);
+ if (bytes_read == 0) {
+ /* end of file */
+ break;
+ }
+ if (bytes_read < 0) {
+ rc = bytes_read;
+ break;
+ }
+
+ /* round up to page align to make direct IO happy. */
+ to_write = ((bytes_read - 1) | (page_size - 1)) + 1;
+
+ for (i = 0; i < comp_size; i++) {
+ ssize_t written;
+ off_t pos2 = pos;
+ size_t to_write2 = to_write;
+
+ /* skip non-overlapped component */
+ if (pos >= comp_array[i].lrc_end ||
+ pos + to_write <= comp_array[i].lrc_start)
+ continue;
+
+ if (pos < comp_array[i].lrc_start)
+ pos2 = comp_array[i].lrc_start;
+
+ to_write2 -= pos2 - pos;
+
+ if ((pos + to_write) > comp_array[i].lrc_end)
+ to_write2 -= pos + to_write -
+ comp_array[i].lrc_end;
+
+ written = llapi_mirror_write(fd,
+ comp_array[i].lrc_mirror_id,
+ buf + pos2 - pos,
+ to_write2, pos2);
+ if (written < 0) {
+ /**
+ * this component is not written successfully,
+ * mark it using its lrc_synced, it is supposed
+ * to be false before getting here.
+ *
+ * And before this function returns, all
+ * elements of comp_array will reverse their
+ * lrc_synced flag to reflect their true
+ * meanings.
+ */
+ comp_array[i].lrc_synced = true;
+ continue;
+ }
+ assert(written == to_write2);
+ }
+
+ pos += bytes_read;
+ count -= bytes_read;
+ }
+
+ free(buf);
+
+ if (rc < 0) {
+ for (i = 0; i < comp_size; i++)
+ comp_array[i].lrc_synced = false;
+ return rc;
+ }
+
+ for (i = 0; i < comp_size; i++) {
+ comp_array[i].lrc_synced = !comp_array[i].lrc_synced;
+ if (comp_array[i].lrc_synced && pos & (page_size - 1)) {
+ rc = llapi_mirror_truncate(fd,
+ comp_array[i].lrc_mirror_id, pos);
+ if (rc < 0)
+ comp_array[i].lrc_synced = false;
+ }
+ }
+
+ /* partially successful is successful */
+ return 0;
+}