* Size on MDS revival
*
* Author: Jinshan Xiong <jinshan.xiong@intel.com>
+ * Author: Yingjin Qian <qian@ddn.com>
*/
#define DEBUG_SUBSYSTEM S_MDS
#include "mdt_internal.h"
+/*
+ * Swab and extract SOM attributes from on-disk xattr.
+ *
+ * \param buf - is a buffer containing the on-disk LSOM extended attribute.
+ * \param rc - is the SOM xattr stored in \a buf
+ * \param ms - is the md_som structure where to extract SOM attributes.
+ */
+int lustre_buf2som(void *buf, int rc, struct md_som *ms)
+{
+ struct lustre_som_attrs *attrs = (struct lustre_som_attrs *)buf;
+ ENTRY;
+
+ if (rc == 0 || rc == -ENODATA)
+ /* no LSOM attributes */
+ RETURN(-ENODATA);
+
+ if (rc < 0)
+ /* error hit while fetching xattr */
+ RETURN(rc);
+
+ /* unpack LSOM attributes */
+ lustre_som_swab(attrs);
+
+ /* fill in-memory md_som structure */
+ ms->ms_valid = attrs->lsa_valid;
+ ms->ms_size = attrs->lsa_size;
+ ms->ms_blocks = attrs->lsa_blocks;
+
+ RETURN(0);
+}
+
int mdt_get_som(struct mdt_thread_info *info, struct mdt_object *obj,
- struct lu_attr *attr)
+ struct md_attr *ma)
{
struct lu_buf *buf = &info->mti_buf;
- struct lustre_som_attrs *som;
+ struct lu_attr *attr = &ma->ma_attr;
int rc;
- som = buf->lb_buf = info->mti_xattr_buf;
+ buf->lb_buf = info->mti_xattr_buf;
buf->lb_len = sizeof(info->mti_xattr_buf);
+ BUILD_BUG_ON(sizeof(struct lustre_som_attrs) >
+ sizeof(info->mti_xattr_buf));
rc = mo_xattr_get(info->mti_env, mdt_object_child(obj), buf,
XATTR_NAME_SOM);
- if (rc >= (int)sizeof(*som) && (som->lsa_valid & LSOM_FL_VALID)) {
- attr->la_valid |= LA_SIZE | LA_BLOCKS;
- attr->la_size = som->lsa_size;
- attr->la_blocks = som->lsa_blocks;
+ rc = lustre_buf2som(info->mti_xattr_buf, rc, &ma->ma_som);
+ if (rc == 0) {
+ struct md_som *som = &ma->ma_som;
- /* Size on MDS is valid and could be returned to client */
- info->mti_som_valid = 1;
+ ma->ma_valid |= MA_SOM;
CDEBUG(D_INODE, DFID": Reading som attrs: "
- "valid: %x, size: %lld, blocks: %lld, rc: %d.\n",
- PFID(mdt_object_fid(obj)), som->lsa_valid,
- som->lsa_size, som->lsa_blocks, rc);
+ "valid: %x, size: %lld, blocks: %lld\n",
+ PFID(mdt_object_fid(obj)), som->ms_valid,
+ som->ms_size, som->ms_blocks);
+
+ if (som->ms_valid & SOM_FL_STRICT) {
+ attr->la_valid |= LA_SIZE | LA_BLOCKS;
+
+ /*
+ * Size on MDS is valid and could be returned
+ * to client.
+ */
+ attr->la_size = som->ms_size;
+ attr->la_blocks = som->ms_blocks;
+ info->mti_som_valid = 1;
+ }
+ } else if (rc == -ENODATA) {
+ rc = 0;
}
- return (rc > 0 || rc == -ENODATA) ? 0 : rc;
+ return rc;
}
+/**
+ * Update SOM on-disk attributes.
+ */
int mdt_set_som(struct mdt_thread_info *info, struct mdt_object *obj,
- struct lu_attr *attr)
+ enum lustre_som_flags flag, __u64 size, __u64 blocks)
{
struct md_object *next = mdt_object_child(obj);
struct lu_buf *buf = &info->mti_buf;
struct lustre_som_attrs *som;
int rc;
+
ENTRY;
- buf->lb_buf = info->mti_xattr_buf;
- buf->lb_len = sizeof(info->mti_xattr_buf);
- rc = mo_xattr_get(info->mti_env, next, buf, XATTR_NAME_SOM);
- if (rc < 0 && rc != -ENODATA)
- RETURN(rc);
+ CDEBUG(D_INODE,
+ DFID": Set SOM attrs S/B/F: %lld/%lld/%x.\n",
+ PFID(mdt_object_fid(obj)), size, blocks, flag);
- som = buf->lb_buf;
+ som = (struct lustre_som_attrs *)info->mti_xattr_buf;
+ BUILD_BUG_ON(sizeof(info->mti_xattr_buf) < sizeof(*som));
- CDEBUG(D_INODE,
- DFID": Set som attrs: " "size: %lld, blocks: %lld, rc: %d\n",
- PFID(mdt_object_fid(obj)), som->lsa_size, som->lsa_blocks, rc);
-
- if (rc == -ENODATA)
- memset(som, 0, sizeof(*som));
- if (attr->la_valid & (LA_SIZE | LA_BLOCKS)) {
- som->lsa_valid |= LSOM_FL_VALID;
- som->lsa_size = attr->la_size;
- som->lsa_blocks = attr->la_blocks;
- }
+ som->lsa_valid = flag;
+ som->lsa_size = size;
+ som->lsa_blocks = blocks;
+ memset(&som->lsa_reserved, 0, sizeof(som->lsa_reserved));
+ lustre_som_swab(som);
+
+ /* update SOM attributes */
+ buf->lb_buf = som;
buf->lb_len = sizeof(*som);
rc = mo_xattr_set(info->mti_env, next, buf, XATTR_NAME_SOM, 0);
+
+ RETURN(rc);
+}
+
+/**
+ * SOM state transition from STRICT to STALE,
+ */
+int mdt_lsom_downgrade(struct mdt_thread_info *info, struct mdt_object *o)
+{
+ struct md_attr *tmp_ma;
+ int rc;
+
+ ENTRY;
+
+ mutex_lock(&o->mot_som_mutex);
+ tmp_ma = &info->mti_u.som.attr;
+ tmp_ma->ma_need = MA_SOM;
+ tmp_ma->ma_valid = 0;
+
+ rc = mdt_get_som(info, o, tmp_ma);
+ if (rc < 0)
+ GOTO(out_lock, rc);
+
+ if (tmp_ma->ma_valid & MA_SOM) {
+ struct md_som *som = &tmp_ma->ma_som;
+
+ info->mti_som_valid = 0;
+ /* The size and blocks info should be still correct. */
+ if (som->ms_valid & SOM_FL_STRICT)
+ rc = mdt_set_som(info, o, SOM_FL_STALE,
+ som->ms_size, som->ms_blocks);
+ }
+out_lock:
+ mutex_unlock(&o->mot_som_mutex);
+ RETURN(rc);
+}
+
+int mdt_lsom_update(struct mdt_thread_info *info,
+ struct mdt_object *o, bool truncate)
+{
+ struct md_attr *ma, *tmp_ma;
+ struct lu_attr *la;
+ int rc = 0;
+
+ ENTRY;
+
+ ma = &info->mti_attr;
+ la = &ma->ma_attr;
+
+ mutex_lock(&o->mot_som_mutex);
+ tmp_ma = &info->mti_u.som.attr;
+ tmp_ma->ma_need = MA_INODE | MA_SOM;
+ tmp_ma->ma_valid = 0;
+
+ rc = mdt_attr_get_complex(info, o, tmp_ma);
+ if (rc)
+ GOTO(out_lock, rc);
+
+ /**
+ * If mti_big_lmm_used is set, it indicates that mti_big_lmm
+ * should contain valid LOV EA data, and can be used directly.
+ */
+ if (!info->mti_big_lmm_used) {
+ rc = mdt_big_xattr_get(info, o, XATTR_NAME_LOV);
+ if (rc < 0 && rc != -ENODATA)
+ GOTO(out_lock, rc);
+
+ /* No LOV EA */
+ if (rc == -ENODATA)
+ GOTO(out_lock, rc = 0);
+
+ rc = 0;
+ }
+
+ /**
+ * Check if a Lazy Size-on-MDS update is needed. Skip the
+ * file with no LOV EA, unlink files or DoM-only file.
+ * MDS only updates LSOM of the file if the size or block
+ * size is being increased or the file is being truncated.
+ */
+ if (mdt_lmm_dom_entry(info->mti_big_lmm) != LMM_DOM_ONLY &&
+ !(tmp_ma->ma_valid & MA_INODE && tmp_ma->ma_attr.la_nlink == 0)) {
+ __u64 size;
+ __u64 blocks;
+ bool changed = false;
+ struct md_som *som = &tmp_ma->ma_som;
+
+ if (truncate) {
+ size = la->la_size;
+ if (size == 0) {
+ blocks = 0;
+ } else if (!(tmp_ma->ma_valid & MA_SOM) ||
+ size < som->ms_size) {
+ /* We cannot rely to blocks after
+ * truncate especially for spare file,
+ * and the truncate operation is usually
+ * followed with a close, so just set blocks
+ * to 1 here, and the following close will
+ * update it accordingly.
+ */
+ blocks = 1;
+ } else {
+ blocks = som->ms_blocks;
+ }
+ } else {
+ if (!(tmp_ma->ma_valid & MA_SOM)) {
+ /* Only set initial SOM Xattr data when both
+ * size and blocks are valid.
+ */
+ if (la->la_valid & (LA_SIZE | LA_LSIZE) &&
+ la->la_valid & (LA_BLOCKS | LA_LBLOCKS)) {
+ changed = true;
+ size = la->la_size;
+ blocks = la->la_blocks;
+ }
+ } else {
+ /* Double check whether it is already set
+ * to SOM_FL_STRICT in mdt_mfd_close.
+ * If file is in SOM_FL_STALE state, and
+ * the close indicates there is no data
+ * modified, skip to transimit to LAZY
+ * state.
+ */
+ if (som->ms_valid & SOM_FL_STRICT ||
+ (som->ms_valid & SOM_FL_STALE &&
+ !(ma->ma_attr_flags & MDS_DATA_MODIFIED)))
+ GOTO(out_lock, rc);
+
+ size = som->ms_size;
+ blocks = som->ms_blocks;
+ if (la->la_valid & (LA_SIZE | LA_LSIZE) &&
+ la->la_size > som->ms_size) {
+ changed = true;
+ size = la->la_size;
+ }
+ if (la->la_valid & (LA_BLOCKS | LA_LBLOCKS) &&
+ la->la_blocks > som->ms_blocks) {
+ changed = true;
+ blocks = la->la_blocks;
+ }
+ }
+ }
+ if (truncate || changed)
+ rc = mdt_set_som(info, o, SOM_FL_LAZY, size, blocks);
+ }
+
+out_lock:
+ mutex_unlock(&o->mot_som_mutex);
RETURN(rc);
}