Lustre의 파일 create & open 과정 분석 - 2

I am going to open [0x200000007:0x1:0x0]/(Melon->[0x20000a811:0x1:0x0]) cr_flag=0100000000102 mode=0100666 msg_flag=0x0

// lustre/mdt/mdt_open.c
int mdt_reint_open(struct mdt_thread_info *info, struct mdt_lock_handle *lhc)
{
...
        parent = mdt_object_find(info->mti_env, mdt, rr->rr_fid1);
...
        fid_zero(child_fid);
        result = -ENOENT;
...
        if (result == -ENOENT) {
...
                child = mdt_object_new(info->mti_env, mdt, child_fid);
        } else {
...
        if (result == -ENOENT) {
...
                result = mdo_create(info->mti_env, mdt_object_child(parent),
                                    &rr->rr_name, mdt_object_child(child),
                                    &info->mti_spec, &info->mti_attr);
                if (result == -ERESTART) {
...
out:
        if (result)
                lustre_msg_set_transno(req->rq_repmsg, 0);
        return result;
}

// lustre/include/lu_object.h
/**
 * Layer in the layered object.
 */
struct lu_object {
        /**
         * Header for this object.
         */
        struct lu_object_header           *lo_header;
        /**
         * Device for this layer.
         */
        struct lu_device                  *lo_dev;
        /**
         * Operations for this object.
         */
        const struct lu_object_operations *lo_ops;
        /**
         * Linkage into list of all layers.
         */
        struct list_head                   lo_linkage;
        /**
         * Link to the device, for debugging.
         */
        struct lu_ref_link                 lo_dev_ref;
};

// lustre mdd/mdd_dir.c
int mdd_create(const struct lu_env *env, struct md_object *pobj,
                      const struct lu_name *lname, struct md_object *child,
                      struct md_op_spec *spec, struct md_attr *ma)
{
...
        mdd_object_make_hint(env, mdd_pobj, son, attr, spec, hint);
...
        if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_BAD_PARENT)) {
                struct lu_fid tfid = *mdd_object_fid(mdd_pobj);

                tfid.f_oid--;
                rc = mdd_linkea_prepare(env, son, NULL, NULL,
                                        &tfid, lname, 1, 0, ldata);
        } else {
                rc = mdd_linkea_prepare(env, son, NULL, NULL,
                                        mdd_object_fid(mdd_pobj),
                                        lname, 1, 0, ldata);
        }
...
        rc = mdd_declare_create(env, mdd, mdd_pobj, son, lname, attr,
                                handle, spec, ldata, &def_acl_buf, &acl_buf,
...
        rc = mdd_create_object(env, mdd_pobj, son, attr, spec, &acl_buf,
                               &def_acl_buf, &hsm_buf, hint, handle, true);
...
}

// lustre/include/dt_object.h
/**
 * This is a general purpose dt allocation hint.
 * It now contains the parent object.
 * It can contain any allocation hint in the future.
 */
struct dt_allocation_hint {
        struct dt_object        *dah_parent;
        const void              *dah_eadata;
        int                     dah_eadata_len;
        int                     dah_acl_len;
        __u32                   dah_mode;
        int                     dah_append_stripes;
        bool                    dah_can_block;
        char                    *dah_append_pool;
};

// lustre/include/lustre_linkea.h
struct linkea_data {
        /**
         * Buffer to keep link EA body.
         */
        struct lu_buf           *ld_buf;
        /**
         * The matched header, entry and its length in the EA
         */
        struct link_ea_header   *ld_leh;
        struct link_ea_entry    *ld_lee;
        int                     ld_reclen;
};

// lustre/lod/lod_qos.c
int lod_qos_prep_create(const struct lu_env *env, struct lod_object *lo,
                        struct lu_attr *attr, struct thandle *th,
                        int comp_idx, __u64 reserve)
{
...
                if (lod_comp->llc_ostlist.op_array && // 사용자가 지정한 OST에 할당
                    lod_comp->llc_ostlist.op_count) {
                        rc = lod_alloc_ost_list(env, lo, stripe, ost_indices,
                                                th, comp_idx, reserve);
                } else if (lod_comp->llc_stripe_offset == LOV_OFFSET_DEFAULT) { // 지정하지 않으면, 먼저 QoS 정책에 의해 할당
                        /**
                         * collect OSTs and OSSs used in other mirrors whose
                         * components cross the ldo_comp_entries[comp_idx]
                         */
                        rc = lod_prepare_avoidance(env, lo);
                        if (rc)
                                GOTO(put_ldts, rc);

                        QOS_DEBUG("collecting conflict osts for comp[%d]\n",
                                  comp_idx);
                        lod_collect_avoidance(lo, lag, comp_idx);

                        rc = lod_ost_alloc_qos(env, lo, stripe, ost_indices,
                                               flag, th, comp_idx, reserve);
                        if (rc == -EAGAIN) // QoS 할당 실패 시 라운드로빈 방식으로 할당
                                rc = lod_ost_alloc_rr(env, lo, stripe,
                                                      ost_indices, flag, th,
                                                      comp_idx, reserve);
                } else { // 사용자가 지정한 offset에 맞춰 할당
                        rc = lod_ost_alloc_specific(env, lo, stripe,
                                                    ost_indices, flag, th,
                                                    comp_idx, reserve);
                }
....

// lustre/lliste/namei.c
static int ll_create_it(struct inode *dir, struct dentry *dentry,
                        struct lookup_intent *it,
                        void *secctx, __u32 secctxlen, bool encrypt,
                        void *encctx, __u32 encctxlen, unsigned int open_flags)
{
...
        inode = ll_create_node(dir, it);
        if (IS_ERR(inode))
                RETURN(PTR_ERR(inode));
...

        d_instantiate(dentry, inode);
...
}

// lustre/lov/lov_object.c
struct lov_stripe_md {
        atomic_t        lsm_refc;
        spinlock_t      lsm_lock;
        pid_t           lsm_lock_owner; /* debugging */

        union {
                /* maximum possible file size, might change as OSTs status
                 * changes, e.g. disconnected, deactivated
                 */
                loff_t          lsm_maxbytes;
                /* size of full foreign LOV */
                size_t          lsm_foreign_size;
        };
        struct ost_id   lsm_oi;
        u32             lsm_magic;
        u32             lsm_layout_gen;
        u16             lsm_flags;
        bool            lsm_is_released;
        u16             lsm_mirror_count;
        u16             lsm_entry_count;
        struct lov_stripe_md_entry *lsm_entries[];
};

lsm 00000000435c78ed, objid 0x1:43025, maxbytes 0x7fffffffffffffff, magic 0x0BD10BD0, refc: 1, entry: 1, mirror: 0, flags: 0,layout_gen 0 
[0x0, 0xffffffffffffffff): id: 0, flags: 10, magic 0x0BD10BD0, layout_gen 0, stripe count 1, sstripe size 1048576, pool: []
oinfo:00000000edee2c32: ostid: 0x0:1186 ost idx: 1 gen: 0