Return-Path: Received: from daytona.panasas.com ([67.152.220.89]:57188 "EHLO daytona.panasas.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751366Ab1DTR24 (ORCPT ); Wed, 20 Apr 2011 13:28:56 -0400 From: Benny Halevy To: linux-nfs@vger.kernel.org Subject: [RFC 21/27] pnfs-obj: objio_osd real IO implementation Date: Wed, 20 Apr 2011 20:28:53 +0300 Message-Id: <1303320533-21750-1-git-send-email-bhalevy@panasas.com> In-Reply-To: <4DAF0DE1.6020609@panasas.com> References: <4DAF0DE1.6020609@panasas.com> Sender: linux-nfs-owner@vger.kernel.org List-ID: Content-Type: text/plain MIME-Version: 1.0 From: Boaz Harrosh With the use of the in-kernel osd library. Implement read/write of data from/to osd-objects according to information specified in the objects-layout. TODO: Only a limited Mirror arrangement is implemented. stripping/raid will come in at later patches. [pnfs-obj: objio: cleanup un-indent _read_mirrors] Signed-off-by: Boaz Harrosh [added FIXME comment] [use REQ flags rather than BIO flags] As they were converged starting 2.6.36 [fix REQ flags usage] fixes "pnfs-obj use REQ flags rather than BIO flags" Cc: Boaz Harrosh Signed-off-by: Benny Halevy --- fs/nfs/objlayout/objio_osd.c | 418 +++++++++++++++++++++++++++++++++++++++++- 1 files changed, 415 insertions(+), 3 deletions(-) diff --git a/fs/nfs/objlayout/objio_osd.c b/fs/nfs/objlayout/objio_osd.c index a24bf82..75d4ebb 100644 --- a/fs/nfs/objlayout/objio_osd.c +++ b/fs/nfs/objlayout/objio_osd.c @@ -142,11 +142,26 @@ struct objio_segment { struct osd_dev *ods[1]; }; +struct objio_state; +typedef ssize_t (*objio_done_fn)(struct objio_state *ios); + struct objio_state { /* Generic layer */ struct objlayout_io_state ol_state; struct objio_segment *objio_seg; + + struct kref kref; + objio_done_fn done; + void *private; + + unsigned long length; + unsigned numdevs; /* Actually used devs in this IO */ + /* A per-device variable array of size numdevs */ + struct _objio_per_comp { + struct bio *bio; + struct osd_request *or; + } per_dev[]; }; /* Send and wait for a get_device_info of devices in the layout, @@ -232,6 +247,38 @@ out: return err; } +static int _verify_data_map(struct pnfs_osd_layout *layout) +{ + struct pnfs_osd_data_map *data_map = &layout->olo_map; + +/* FIXME: Only Mirror arangment for now. if not so, do not mount */ + if (data_map->odm_group_width || data_map->odm_group_depth) { + printk(KERN_ERR "Group width/depth not supported\n"); + return -ENOTSUPP; + } + if (data_map->odm_num_comps != layout->olo_num_comps) { + printk(KERN_ERR "odm_num_comps(%u) != olo_num_comps(%u)\n", + data_map->odm_num_comps, layout->olo_num_comps); + return -ENOTSUPP; + } + if (data_map->odm_raid_algorithm != PNFS_OSD_RAID_0) { + printk(KERN_ERR "Only RAID_0 for now\n"); + return -ENOTSUPP; + } + if (data_map->odm_num_comps != data_map->odm_mirror_cnt + 1) { + printk(KERN_ERR "Mirror only!, num_comps=%u mirrors=%u\n", + data_map->odm_num_comps, data_map->odm_mirror_cnt); + return -ENOTSUPP; + } + + if (data_map->odm_stripe_unit != PAGE_SIZE) { + printk(KERN_ERR "Stripe Unit != PAGE_SIZE not supported\n"); + return -ENOTSUPP; + } + + return 0; +} + int objio_alloc_lseg(void **outp, struct pnfs_layout_hdr *pnfslay, struct pnfs_layout_segment *lseg, @@ -240,6 +287,10 @@ int objio_alloc_lseg(void **outp, struct objio_segment *objio_seg; int err; + err = _verify_data_map(layout); + if (unlikely(err)) + return err; + objio_seg = kzalloc(sizeof(*objio_seg) + (layout->olo_num_comps - 1) * sizeof(objio_seg->ods[0]), GFP_KERNEL); @@ -272,8 +323,11 @@ int objio_alloc_io_state(void *seg, struct objlayout_io_state **outp) { struct objio_segment *objio_seg = seg; struct objio_state *ios; + const unsigned first_size = sizeof(*ios) + + objio_seg->num_comps * sizeof(ios->per_dev[0]); - ios = kzalloc(sizeof(*ios), GFP_KERNEL); + dprintk("%s: num_comps=%d\n", __func__, objio_seg->num_comps); + ios = kzalloc(first_size, GFP_KERNEL); if (unlikely(!ios)) return -ENOMEM; @@ -291,20 +345,378 @@ void objio_free_io_state(struct objlayout_io_state *ol_state) kfree(ios); } +enum pnfs_osd_errno osd_pri_2_pnfs_err(enum osd_err_priority oep) +{ + switch (oep) { + case OSD_ERR_PRI_NO_ERROR: + return (enum pnfs_osd_errno)0; + + case OSD_ERR_PRI_CLEAR_PAGES: + BUG_ON(1); + return 0; + + case OSD_ERR_PRI_RESOURCE: + return PNFS_OSD_ERR_RESOURCE; + case OSD_ERR_PRI_BAD_CRED: + return PNFS_OSD_ERR_BAD_CRED; + case OSD_ERR_PRI_NO_ACCESS: + return PNFS_OSD_ERR_NO_ACCESS; + case OSD_ERR_PRI_UNREACHABLE: + return PNFS_OSD_ERR_UNREACHABLE; + case OSD_ERR_PRI_NOT_FOUND: + return PNFS_OSD_ERR_NOT_FOUND; + case OSD_ERR_PRI_NO_SPACE: + return PNFS_OSD_ERR_NO_SPACE; + default: + WARN_ON(1); + /* fallthrough */ + case OSD_ERR_PRI_EIO: + return PNFS_OSD_ERR_EIO; + } +} + +static void _clear_bio(struct bio *bio) +{ + struct bio_vec *bv; + unsigned i; + + __bio_for_each_segment(bv, bio, i, 0) { + unsigned this_count = bv->bv_len; + + if (likely(PAGE_SIZE == this_count)) + clear_highpage(bv->bv_page); + else + zero_user(bv->bv_page, bv->bv_offset, this_count); + } +} + +static int _io_check(struct objio_state *ios, bool is_write) +{ + enum osd_err_priority oep = OSD_ERR_PRI_NO_ERROR; + int lin_ret = 0; + int i; + + for (i = 0; i < ios->numdevs; i++) { + struct osd_sense_info osi; + struct osd_request *or = ios->per_dev[i].or; + int ret; + + if (!or) + continue; + + ret = osd_req_decode_sense(or, &osi); + if (likely(!ret)) + continue; + + if (OSD_ERR_PRI_CLEAR_PAGES == osi.osd_err_pri) { + /* start read offset passed endof file */ + BUG_ON(is_write); + _clear_bio(ios->per_dev[i].bio); + dprintk("%s: start read offset passed end of file " + "offset=0x%llx, length=0x%lx\n", __func__, + _LLU(ios->ol_state.offset), ios->length); + + continue; /* we recovered */ + } + + if (osi.osd_err_pri >= oep) { + oep = osi.osd_err_pri; + lin_ret = ret; + } + } + + return lin_ret; +} + +/* + * Common IO state helpers. + */ +static void _io_free(struct objio_state *ios) +{ + unsigned i; + + for (i = 0; i < ios->numdevs; i++) { + struct _objio_per_comp *per_dev = &ios->per_dev[i]; + + if (per_dev->or) { + osd_end_request(per_dev->or); + per_dev->or = NULL; + } + + if (per_dev->bio) { + bio_put(per_dev->bio); + per_dev->bio = NULL; + } + } +} + +static int _io_rw_pagelist(struct objio_state *ios) +{ + u64 length = ios->ol_state.count; + unsigned pgbase = ios->ol_state.pgbase; + unsigned nr_pages = ios->ol_state.nr_pages; + struct page **pages = ios->ol_state.pages; + struct bio *master_bio; + unsigned bio_size = min_t(unsigned, nr_pages, BIO_MAX_PAGES_KMALLOC); + + master_bio = bio_kmalloc(GFP_KERNEL, bio_size); + if (unlikely(!master_bio)) { + dprintk("%s: Faild to alloc bio pages=%d\n", + __func__, bio_size); + return -ENOMEM; + } + + ios->per_dev[0].bio = master_bio; + + while (length) { + unsigned cur_len, added_len; + + cur_len = min_t(u64, length, PAGE_SIZE - pgbase); + + added_len = bio_add_pc_page( + osd_request_queue(ios->objio_seg->ods[0]), + master_bio, *pages, cur_len, pgbase); + if (unlikely(cur_len != added_len)) + break; + + pgbase = 0; + ++pages; + length -= cur_len; + ios->length += cur_len; + } + + /* this should never happen */ + WARN_ON(!ios->length); + + return 0; +} + +static ssize_t _sync_done(struct objio_state *ios) +{ + struct completion *waiting = ios->private; + + complete(waiting); + return 0; +} + +static void _last_io(struct kref *kref) +{ + struct objio_state *ios = container_of(kref, struct objio_state, kref); + + ios->done(ios); +} + +static void _done_io(struct osd_request *or, void *p) +{ + struct objio_state *ios = p; + + kref_put(&ios->kref, _last_io); +} + +static ssize_t _io_exec(struct objio_state *ios) +{ + DECLARE_COMPLETION_ONSTACK(wait); + ssize_t status = 0; /* sync status */ + unsigned i; + objio_done_fn saved_done_fn = ios->done; + bool sync = ios->ol_state.sync; + + if (sync) { + ios->done = _sync_done; + ios->private = &wait; + } + + kref_init(&ios->kref); + + for (i = 0; i < ios->numdevs; i++) { + struct osd_request *or = ios->per_dev[i].or; + + if (!or) + continue; + + kref_get(&ios->kref); + osd_execute_request_async(or, _done_io, ios); + } + + kref_put(&ios->kref, _last_io); + + if (sync) { + wait_for_completion(&wait); + status = saved_done_fn(ios); + } + + return status; +} + /* * read */ +static ssize_t _read_done(struct objio_state *ios) +{ + ssize_t status; + int ret = _io_check(ios, false); + + _io_free(ios); + + if (likely(!ret)) + status = ios->length; + else + status = ret; + + objlayout_read_done(&ios->ol_state, status, ios->ol_state.sync); + return status; +} + +static ssize_t _read_exec(struct objio_state *ios) +{ + struct osd_request *or = NULL; + struct _objio_per_comp *per_dev = &ios->per_dev[0]; + unsigned dev = 0; + struct pnfs_osd_object_cred *cred = + &ios->objio_seg->layout->olo_comps[dev]; + struct osd_obj_id obj = { + .partition = cred->oc_object_id.oid_partition_id, + .id = cred->oc_object_id.oid_object_id, + }; + int ret; + + or = osd_start_request(ios->objio_seg->ods[dev], GFP_KERNEL); + if (unlikely(!or)) { + ret = -ENOMEM; + goto err; + } + per_dev->or = or; + ios->numdevs++; + + osd_req_read(or, &obj, ios->ol_state.offset, per_dev->bio, ios->length); + + ret = osd_finalize_request(or, 0, cred->oc_cap.cred, NULL); + if (ret) { + dprintk("%s: Faild to osd_finalize_request() => %d\n", + __func__, ret); + goto err; + } + + dprintk("%s: obj=0x%llx start=0x%llx length=0x%lx\n", + __func__, obj.id, _LLU(ios->ol_state.offset), ios->length); + ios->done = _read_done; + return _io_exec(ios); /* In sync mode exec returns the io status */ + +err: + _io_free(ios); + return ret; +} + ssize_t objio_read_pagelist(struct objlayout_io_state *ol_state) { - return -EIO; + struct objio_state *ios = container_of(ol_state, struct objio_state, + ol_state); + int ret; + + ret = _io_rw_pagelist(ios); + if (unlikely(ret)) + return ret; + + return _read_exec(ios); } /* * write */ +static ssize_t _write_done(struct objio_state *ios) +{ + ssize_t status; + int ret = _io_check(ios, true); + + _io_free(ios); + + if (likely(!ret)) { + /* FIXME: should be based on the OSD's persistence model + * See OSD2r05 Section 4.13 Data persistence model */ + ios->ol_state.committed = NFS_FILE_SYNC; + status = ios->length; + } else { + status = ret; + } + + objlayout_write_done(&ios->ol_state, status, ios->ol_state.sync); + return status; +} + +static int _write_exec(struct objio_state *ios) +{ + int i, ret; + struct bio *master_bio = ios->per_dev[0].bio; + + for (i = 0; i < ios->objio_seg->num_comps; i++) { + struct osd_request *or = NULL; + struct pnfs_osd_object_cred *cred = + &ios->objio_seg->layout->olo_comps[i]; + struct osd_obj_id obj = {cred->oc_object_id.oid_partition_id, + cred->oc_object_id.oid_object_id}; + struct _objio_per_comp *per_dev = &ios->per_dev[i]; + struct bio *bio; + + or = osd_start_request(ios->objio_seg->ods[i], GFP_KERNEL); + if (unlikely(!or)) { + ret = -ENOMEM; + goto err; + } + per_dev->or = or; + ios->numdevs++; + + if (i != 0) { + bio = bio_kmalloc(GFP_KERNEL, master_bio->bi_max_vecs); + if (unlikely(!bio)) { + dprintk("Faild to allocate BIO size=%u\n", + master_bio->bi_max_vecs); + ret = -ENOMEM; + goto err; + } + + __bio_clone(bio, master_bio); + bio->bi_bdev = NULL; + bio->bi_next = NULL; + per_dev->bio = bio; + } else { + bio = master_bio; + bio->bi_rw |= REQ_WRITE; + } + + osd_req_write(or, &obj, ios->ol_state.offset, bio, ios->length); + + ret = osd_finalize_request(or, 0, cred->oc_cap.cred, NULL); + if (ret) { + dprintk("%s: Faild to osd_finalize_request() => %d\n", + __func__, ret); + goto err; + } + + dprintk("%s: [%d] obj=0x%llx start=0x%llx length=0x%lx\n", + __func__, i, obj.id, _LLU(ios->ol_state.offset), + ios->length); + } + + ios->done = _write_done; + return _io_exec(ios); /* In sync mode exec returns the io->status */ + +err: + _io_free(ios); + return ret; +} + ssize_t objio_write_pagelist(struct objlayout_io_state *ol_state, bool stable) { - return -EIO; + struct objio_state *ios = container_of(ol_state, struct objio_state, + ol_state); + int ret; + + /* TODO: ios->stable = stable; */ + ret = _io_rw_pagelist(ios); + if (unlikely(ret)) + return ret; + + return _write_exec(ios); } /* -- 1.7.3.4