diff options
Diffstat (limited to 'fs/nfs/pnfs_nfs.c')
-rw-r--r-- | fs/nfs/pnfs_nfs.c | 514 |
1 files changed, 375 insertions, 139 deletions
diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c index 8b37e7f8e789..25f135572fc8 100644 --- a/fs/nfs/pnfs_nfs.c +++ b/fs/nfs/pnfs_nfs.c @@ -59,6 +59,17 @@ void pnfs_generic_commit_release(void *calldata) } EXPORT_SYMBOL_GPL(pnfs_generic_commit_release); +static struct pnfs_layout_segment * +pnfs_free_bucket_lseg(struct pnfs_commit_bucket *bucket) +{ + if (list_empty(&bucket->committing) && list_empty(&bucket->written)) { + struct pnfs_layout_segment *freeme = bucket->lseg; + bucket->lseg = NULL; + return freeme; + } + return NULL; +} + /* The generic layer is about to remove the req from the commit list. * If this will make the bucket empty, it will need to put the lseg reference. * Note this must be called holding nfsi->commit_mutex @@ -78,8 +89,7 @@ pnfs_generic_clear_request_commit(struct nfs_page *req, bucket = list_first_entry(&req->wb_list, struct pnfs_commit_bucket, written); - freeme = bucket->wlseg; - bucket->wlseg = NULL; + freeme = pnfs_free_bucket_lseg(bucket); } out: nfs_request_remove_commit_list(req, cinfo); @@ -87,10 +97,154 @@ out: } EXPORT_SYMBOL_GPL(pnfs_generic_clear_request_commit); +struct pnfs_commit_array * +pnfs_alloc_commit_array(size_t n, gfp_t gfp_flags) +{ + struct pnfs_commit_array *p; + struct pnfs_commit_bucket *b; + + p = kmalloc(struct_size(p, buckets, n), gfp_flags); + if (!p) + return NULL; + p->nbuckets = n; + INIT_LIST_HEAD(&p->cinfo_list); + INIT_LIST_HEAD(&p->lseg_list); + p->lseg = NULL; + for (b = &p->buckets[0]; n != 0; b++, n--) { + INIT_LIST_HEAD(&b->written); + INIT_LIST_HEAD(&b->committing); + b->lseg = NULL; + b->direct_verf.committed = NFS_INVALID_STABLE_HOW; + } + return p; +} +EXPORT_SYMBOL_GPL(pnfs_alloc_commit_array); + +void +pnfs_free_commit_array(struct pnfs_commit_array *p) +{ + kfree_rcu(p, rcu); +} +EXPORT_SYMBOL_GPL(pnfs_free_commit_array); + +static struct pnfs_commit_array * +pnfs_find_commit_array_by_lseg(struct pnfs_ds_commit_info *fl_cinfo, + struct pnfs_layout_segment *lseg) +{ + struct pnfs_commit_array *array; + + list_for_each_entry_rcu(array, &fl_cinfo->commits, cinfo_list) { + if (array->lseg == lseg) + return array; + } + return NULL; +} + +struct pnfs_commit_array * +pnfs_add_commit_array(struct pnfs_ds_commit_info *fl_cinfo, + struct pnfs_commit_array *new, + struct pnfs_layout_segment *lseg) +{ + struct pnfs_commit_array *array; + + array = pnfs_find_commit_array_by_lseg(fl_cinfo, lseg); + if (array) + return array; + new->lseg = lseg; + refcount_set(&new->refcount, 1); + list_add_rcu(&new->cinfo_list, &fl_cinfo->commits); + list_add(&new->lseg_list, &lseg->pls_commits); + return new; +} +EXPORT_SYMBOL_GPL(pnfs_add_commit_array); + +static struct pnfs_commit_array * +pnfs_lookup_commit_array(struct pnfs_ds_commit_info *fl_cinfo, + struct pnfs_layout_segment *lseg) +{ + struct pnfs_commit_array *array; + + rcu_read_lock(); + array = pnfs_find_commit_array_by_lseg(fl_cinfo, lseg); + if (!array) { + rcu_read_unlock(); + fl_cinfo->ops->setup_ds_info(fl_cinfo, lseg); + rcu_read_lock(); + array = pnfs_find_commit_array_by_lseg(fl_cinfo, lseg); + } + rcu_read_unlock(); + return array; +} + +static void +pnfs_release_commit_array_locked(struct pnfs_commit_array *array) +{ + list_del_rcu(&array->cinfo_list); + list_del(&array->lseg_list); + pnfs_free_commit_array(array); +} + +static void +pnfs_put_commit_array_locked(struct pnfs_commit_array *array) +{ + if (refcount_dec_and_test(&array->refcount)) + pnfs_release_commit_array_locked(array); +} + +static void +pnfs_put_commit_array(struct pnfs_commit_array *array, struct inode *inode) +{ + if (refcount_dec_and_lock(&array->refcount, &inode->i_lock)) { + pnfs_release_commit_array_locked(array); + spin_unlock(&inode->i_lock); + } +} + +static struct pnfs_commit_array * +pnfs_get_commit_array(struct pnfs_commit_array *array) +{ + if (refcount_inc_not_zero(&array->refcount)) + return array; + return NULL; +} + +static void +pnfs_remove_and_free_commit_array(struct pnfs_commit_array *array) +{ + array->lseg = NULL; + list_del_init(&array->lseg_list); + pnfs_put_commit_array_locked(array); +} + +void +pnfs_generic_ds_cinfo_release_lseg(struct pnfs_ds_commit_info *fl_cinfo, + struct pnfs_layout_segment *lseg) +{ + struct pnfs_commit_array *array, *tmp; + + list_for_each_entry_safe(array, tmp, &lseg->pls_commits, lseg_list) + pnfs_remove_and_free_commit_array(array); +} +EXPORT_SYMBOL_GPL(pnfs_generic_ds_cinfo_release_lseg); + +void +pnfs_generic_ds_cinfo_destroy(struct pnfs_ds_commit_info *fl_cinfo) +{ + struct pnfs_commit_array *array, *tmp; + + list_for_each_entry_safe(array, tmp, &fl_cinfo->commits, cinfo_list) + pnfs_remove_and_free_commit_array(array); +} +EXPORT_SYMBOL_GPL(pnfs_generic_ds_cinfo_destroy); + +/* + * Locks the nfs_page requests for commit and moves them to + * @bucket->committing. + */ static int -pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, - struct nfs_commit_info *cinfo, - int max) +pnfs_bucket_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, + struct nfs_commit_info *cinfo, + int max) { struct list_head *src = &bucket->written; struct list_head *dst = &bucket->committing; @@ -101,158 +255,253 @@ pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, if (ret) { cinfo->ds->nwritten -= ret; cinfo->ds->ncommitting += ret; - if (bucket->clseg == NULL) - bucket->clseg = pnfs_get_lseg(bucket->wlseg); - if (list_empty(src)) { - pnfs_put_lseg(bucket->wlseg); - bucket->wlseg = NULL; - } } return ret; } +static int pnfs_bucket_scan_array(struct nfs_commit_info *cinfo, + struct pnfs_commit_bucket *buckets, + unsigned int nbuckets, + int max) +{ + unsigned int i; + int rv = 0, cnt; + + for (i = 0; i < nbuckets && max != 0; i++) { + cnt = pnfs_bucket_scan_ds_commit_list(&buckets[i], cinfo, max); + rv += cnt; + max -= cnt; + } + return rv; +} + /* Move reqs from written to committing lists, returning count * of number moved. */ -int pnfs_generic_scan_commit_lists(struct nfs_commit_info *cinfo, - int max) +int pnfs_generic_scan_commit_lists(struct nfs_commit_info *cinfo, int max) { - int i, rv = 0, cnt; + struct pnfs_ds_commit_info *fl_cinfo = cinfo->ds; + struct pnfs_commit_array *array; + int rv = 0, cnt; - lockdep_assert_held(&NFS_I(cinfo->inode)->commit_mutex); - for (i = 0; i < cinfo->ds->nbuckets && max != 0; i++) { - cnt = pnfs_generic_scan_ds_commit_list(&cinfo->ds->buckets[i], - cinfo, max); - max -= cnt; + rcu_read_lock(); + list_for_each_entry_rcu(array, &fl_cinfo->commits, cinfo_list) { + if (!array->lseg || !pnfs_get_commit_array(array)) + continue; + rcu_read_unlock(); + cnt = pnfs_bucket_scan_array(cinfo, array->buckets, + array->nbuckets, max); + rcu_read_lock(); + pnfs_put_commit_array(array, cinfo->inode); rv += cnt; + max -= cnt; + if (!max) + break; } + rcu_read_unlock(); return rv; } EXPORT_SYMBOL_GPL(pnfs_generic_scan_commit_lists); -/* Pull everything off the committing lists and dump into @dst. */ -void pnfs_generic_recover_commit_reqs(struct list_head *dst, - struct nfs_commit_info *cinfo) +static unsigned int +pnfs_bucket_recover_commit_reqs(struct list_head *dst, + struct pnfs_commit_bucket *buckets, + unsigned int nbuckets, + struct nfs_commit_info *cinfo) { struct pnfs_commit_bucket *b; struct pnfs_layout_segment *freeme; - int nwritten; - int i; + unsigned int nwritten, ret = 0; + unsigned int i; - lockdep_assert_held(&NFS_I(cinfo->inode)->commit_mutex); restart: - for (i = 0, b = cinfo->ds->buckets; i < cinfo->ds->nbuckets; i++, b++) { + for (i = 0, b = buckets; i < nbuckets; i++, b++) { nwritten = nfs_scan_commit_list(&b->written, dst, cinfo, 0); if (!nwritten) continue; - cinfo->ds->nwritten -= nwritten; - if (list_empty(&b->written)) { - freeme = b->wlseg; - b->wlseg = NULL; + ret += nwritten; + freeme = pnfs_free_bucket_lseg(b); + if (freeme) { pnfs_put_lseg(freeme); goto restart; } } + return ret; +} + +/* Pull everything off the committing lists and dump into @dst. */ +void pnfs_generic_recover_commit_reqs(struct list_head *dst, + struct nfs_commit_info *cinfo) +{ + struct pnfs_ds_commit_info *fl_cinfo = cinfo->ds; + struct pnfs_commit_array *array; + unsigned int nwritten; + + lockdep_assert_held(&NFS_I(cinfo->inode)->commit_mutex); + rcu_read_lock(); + list_for_each_entry_rcu(array, &fl_cinfo->commits, cinfo_list) { + if (!array->lseg || !pnfs_get_commit_array(array)) + continue; + rcu_read_unlock(); + nwritten = pnfs_bucket_recover_commit_reqs(dst, + array->buckets, + array->nbuckets, + cinfo); + rcu_read_lock(); + pnfs_put_commit_array(array, cinfo->inode); + fl_cinfo->nwritten -= nwritten; + } + rcu_read_unlock(); } EXPORT_SYMBOL_GPL(pnfs_generic_recover_commit_reqs); -static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx) +static struct nfs_page * +pnfs_bucket_search_commit_reqs(struct pnfs_commit_bucket *buckets, + unsigned int nbuckets, struct page *page) +{ + struct nfs_page *req; + struct pnfs_commit_bucket *b; + unsigned int i; + + /* Linearly search the commit lists for each bucket until a matching + * request is found */ + for (i = 0, b = buckets; i < nbuckets; i++, b++) { + list_for_each_entry(req, &b->written, wb_list) { + if (req->wb_page == page) + return req->wb_head; + } + list_for_each_entry(req, &b->committing, wb_list) { + if (req->wb_page == page) + return req->wb_head; + } + } + return NULL; +} + +/* pnfs_generic_search_commit_reqs - Search lists in @cinfo for the head reqest + * for @page + * @cinfo - commit info for current inode + * @page - page to search for matching head request + * + * Returns a the head request if one is found, otherwise returns NULL. + */ +struct nfs_page * +pnfs_generic_search_commit_reqs(struct nfs_commit_info *cinfo, struct page *page) { struct pnfs_ds_commit_info *fl_cinfo = cinfo->ds; + struct pnfs_commit_array *array; + struct nfs_page *req; + + list_for_each_entry(array, &fl_cinfo->commits, cinfo_list) { + req = pnfs_bucket_search_commit_reqs(array->buckets, + array->nbuckets, page); + if (req) + return req; + } + return NULL; +} +EXPORT_SYMBOL_GPL(pnfs_generic_search_commit_reqs); + +static struct pnfs_layout_segment * +pnfs_bucket_get_committing(struct list_head *head, + struct pnfs_commit_bucket *bucket, + struct nfs_commit_info *cinfo) +{ + struct list_head *pos; + + list_for_each(pos, &bucket->committing) + cinfo->ds->ncommitting--; + list_splice_init(&bucket->committing, head); + return pnfs_free_bucket_lseg(bucket); +} + +static struct nfs_commit_data * +pnfs_bucket_fetch_commitdata(struct pnfs_commit_bucket *bucket, + struct nfs_commit_info *cinfo) +{ + struct nfs_commit_data *data = nfs_commitdata_alloc(false); + + if (!data) + return NULL; + data->lseg = pnfs_bucket_get_committing(&data->pages, bucket, cinfo); + if (!data->lseg) + data->lseg = pnfs_get_lseg(bucket->lseg); + return data; +} + +static void pnfs_generic_retry_commit(struct pnfs_commit_bucket *buckets, + unsigned int nbuckets, + struct nfs_commit_info *cinfo, + unsigned int idx) +{ struct pnfs_commit_bucket *bucket; struct pnfs_layout_segment *freeme; - struct list_head *pos; LIST_HEAD(pages); - int i; - mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); - for (i = idx; i < fl_cinfo->nbuckets; i++) { - bucket = &fl_cinfo->buckets[i]; + for (bucket = buckets; idx < nbuckets; bucket++, idx++) { if (list_empty(&bucket->committing)) continue; - freeme = bucket->clseg; - bucket->clseg = NULL; - list_for_each(pos, &bucket->committing) - cinfo->ds->ncommitting--; - list_splice_init(&bucket->committing, &pages); + mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); + freeme = pnfs_bucket_get_committing(&pages, bucket, cinfo); mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); - nfs_retry_commit(&pages, freeme, cinfo, i); + nfs_retry_commit(&pages, freeme, cinfo, idx); pnfs_put_lseg(freeme); - mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); } - mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); } static unsigned int -pnfs_generic_alloc_ds_commits(struct nfs_commit_info *cinfo, - struct list_head *list) +pnfs_bucket_alloc_ds_commits(struct list_head *list, + struct pnfs_commit_bucket *buckets, + unsigned int nbuckets, + struct nfs_commit_info *cinfo) { - struct pnfs_ds_commit_info *fl_cinfo; struct pnfs_commit_bucket *bucket; struct nfs_commit_data *data; - int i; + unsigned int i; unsigned int nreq = 0; - fl_cinfo = cinfo->ds; - bucket = fl_cinfo->buckets; - for (i = 0; i < fl_cinfo->nbuckets; i++, bucket++) { + for (i = 0, bucket = buckets; i < nbuckets; i++, bucket++) { if (list_empty(&bucket->committing)) continue; - data = nfs_commitdata_alloc(false); - if (!data) - break; - data->ds_commit_index = i; - list_add(&data->pages, list); - nreq++; + mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); + if (!list_empty(&bucket->committing)) { + data = pnfs_bucket_fetch_commitdata(bucket, cinfo); + if (!data) + goto out_error; + data->ds_commit_index = i; + list_add_tail(&data->list, list); + atomic_inc(&cinfo->mds->rpcs_out); + nreq++; + } + mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); } - + return nreq; +out_error: + mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); /* Clean up on error */ - pnfs_generic_retry_commit(cinfo, i); + pnfs_generic_retry_commit(buckets, nbuckets, cinfo, i); return nreq; } -static inline -void pnfs_fetch_commit_bucket_list(struct list_head *pages, - struct nfs_commit_data *data, - struct nfs_commit_info *cinfo) +static unsigned int +pnfs_alloc_ds_commits_list(struct list_head *list, + struct pnfs_ds_commit_info *fl_cinfo, + struct nfs_commit_info *cinfo) { - struct pnfs_commit_bucket *bucket; - struct list_head *pos; - - bucket = &cinfo->ds->buckets[data->ds_commit_index]; - mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); - list_for_each(pos, &bucket->committing) - cinfo->ds->ncommitting--; - list_splice_init(&bucket->committing, pages); - data->lseg = bucket->clseg; - bucket->clseg = NULL; - mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); - -} + struct pnfs_commit_array *array; + unsigned int ret = 0; -/* Helper function for pnfs_generic_commit_pagelist to catch an empty - * page list. This can happen when two commits race. - * - * This must be called instead of nfs_init_commit - call one or the other, but - * not both! - */ -static bool -pnfs_generic_commit_cancel_empty_pagelist(struct list_head *pages, - struct nfs_commit_data *data, - struct nfs_commit_info *cinfo) -{ - if (list_empty(pages)) { - if (atomic_dec_and_test(&cinfo->mds->rpcs_out)) - wake_up_var(&cinfo->mds->rpcs_out); - /* don't call nfs_commitdata_release - it tries to put - * the open_context which is not acquired until nfs_init_commit - * which has not been called on @data */ - WARN_ON_ONCE(data->context); - nfs_commit_free(data); - return true; + rcu_read_lock(); + list_for_each_entry_rcu(array, &fl_cinfo->commits, cinfo_list) { + if (!array->lseg || !pnfs_get_commit_array(array)) + continue; + rcu_read_unlock(); + ret += pnfs_bucket_alloc_ds_commits(list, array->buckets, + array->nbuckets, cinfo); + rcu_read_lock(); + pnfs_put_commit_array(array, cinfo->inode); } - - return false; + return ret; } /* This follows nfs_commit_list pretty closely */ @@ -262,6 +511,7 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, int (*initiate_commit)(struct nfs_commit_data *data, int how)) { + struct pnfs_ds_commit_info *fl_cinfo = cinfo->ds; struct nfs_commit_data *data, *tmp; LIST_HEAD(list); unsigned int nreq = 0; @@ -269,40 +519,25 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, if (!list_empty(mds_pages)) { data = nfs_commitdata_alloc(true); data->ds_commit_index = -1; - list_add(&data->pages, &list); + list_splice_init(mds_pages, &data->pages); + list_add_tail(&data->list, &list); + atomic_inc(&cinfo->mds->rpcs_out); nreq++; } - nreq += pnfs_generic_alloc_ds_commits(cinfo, &list); - + nreq += pnfs_alloc_ds_commits_list(&list, fl_cinfo, cinfo); if (nreq == 0) goto out; - atomic_add(nreq, &cinfo->mds->rpcs_out); - - list_for_each_entry_safe(data, tmp, &list, pages) { - list_del_init(&data->pages); + list_for_each_entry_safe(data, tmp, &list, list) { + list_del(&data->list); if (data->ds_commit_index < 0) { - /* another commit raced with us */ - if (pnfs_generic_commit_cancel_empty_pagelist(mds_pages, - data, cinfo)) - continue; - - nfs_init_commit(data, mds_pages, NULL, cinfo); + nfs_init_commit(data, NULL, NULL, cinfo); nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(data->inode), data->mds_ops, how, 0); } else { - LIST_HEAD(pages); - - pnfs_fetch_commit_bucket_list(&pages, data, cinfo); - - /* another commit raced with us */ - if (pnfs_generic_commit_cancel_empty_pagelist(&pages, - data, cinfo)) - continue; - - nfs_init_commit(data, &pages, data->lseg, cinfo); + nfs_init_commit(data, NULL, data->lseg, cinfo); initiate_commit(data, how); } } @@ -930,32 +1165,33 @@ pnfs_layout_mark_request_commit(struct nfs_page *req, u32 ds_commit_idx) { struct list_head *list; - struct pnfs_commit_bucket *buckets; + struct pnfs_commit_array *array; + struct pnfs_commit_bucket *bucket; mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); - buckets = cinfo->ds->buckets; - list = &buckets[ds_commit_idx].written; - if (list_empty(list)) { - if (!pnfs_is_valid_lseg(lseg)) { - mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); - cinfo->completion_ops->resched_write(cinfo, req); - return; - } - /* Non-empty buckets hold a reference on the lseg. That ref - * is normally transferred to the COMMIT call and released - * there. It could also be released if the last req is pulled - * off due to a rewrite, in which case it will be done in - * pnfs_common_clear_request_commit - */ - WARN_ON_ONCE(buckets[ds_commit_idx].wlseg != NULL); - buckets[ds_commit_idx].wlseg = pnfs_get_lseg(lseg); - } + array = pnfs_lookup_commit_array(cinfo->ds, lseg); + if (!array || !pnfs_is_valid_lseg(lseg)) + goto out_resched; + bucket = &array->buckets[ds_commit_idx]; + list = &bucket->written; + /* Non-empty buckets hold a reference on the lseg. That ref + * is normally transferred to the COMMIT call and released + * there. It could also be released if the last req is pulled + * off due to a rewrite, in which case it will be done in + * pnfs_common_clear_request_commit + */ + if (!bucket->lseg) + bucket->lseg = pnfs_get_lseg(lseg); set_bit(PG_COMMIT_TO_DS, &req->wb_flags); cinfo->ds->nwritten++; nfs_request_add_commit_list_locked(req, list, cinfo); mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); nfs_mark_page_unstable(req->wb_page, cinfo); + return; +out_resched: + mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); + cinfo->completion_ops->resched_write(cinfo, req); } EXPORT_SYMBOL_GPL(pnfs_layout_mark_request_commit); |