From b32728ffef7f233dbdabb3f11814bdf692aaf501 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Mon, 23 Jan 2017 08:12:19 -0700 Subject: [PATCH 01/10] xen-blkfront: feature flags handling adjustments Don't truncate the "feature-persistent" value read from xenstore: Any non-zero value is supposed to enable the feature, just like is already being done for feature_secdiscard. Just like the other feature_* fields, feature_flush and feature_fua are boolean flags, and hence fit well into a single bit. Keep all bit fields together to limit gaps. Signed-off-by: Jan Beulich Signed-off-by: Konrad Rzeszutek Wilk --- drivers/block/xen-blkfront.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index b2bdfa81f929..386b17208605 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -197,13 +197,13 @@ struct blkfront_info /* Number of pages per ring buffer. */ unsigned int nr_ring_pages; struct request_queue *rq; - unsigned int feature_flush; - unsigned int feature_fua; + unsigned int feature_flush:1; + unsigned int feature_fua:1; unsigned int feature_discard:1; unsigned int feature_secdiscard:1; + unsigned int feature_persistent:1; unsigned int discard_granularity; unsigned int discard_alignment; - unsigned int feature_persistent:1; /* Number of 4KB segments handled */ unsigned int max_indirect_segments; int is_ready; @@ -2323,8 +2323,8 @@ static void blkfront_gather_backend_features(struct blkfront_info *info) blkfront_setup_discard(info); info->feature_persistent = - xenbus_read_unsigned(info->xbdev->otherend, - "feature-persistent", 0); + !!xenbus_read_unsigned(info->xbdev->otherend, + "feature-persistent", 0); indirect_segments = xenbus_read_unsigned(info->xbdev->otherend, "feature-max-indirect-segments", 0); From 3b4f18843e511193e7eb616710e838f5852e661d Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Mon, 23 Jan 2017 08:11:37 -0700 Subject: [PATCH 02/10] xen-blkfront: correct maximum segment accounting Making use of "max_indirect_segments=" has issues: - blkfront_setup_indirect() may end up with zero psegs when PAGE_SIZE is sufficiently much larger than XEN_PAGE_SIZE - the variable driven by the command line option (xen_blkif_max_segments) has a somewhat different purpose, and hence should namely never end up being zero - as long as the specified value is lower than the legacy default, we better don't use indirect segments at all (or we'd in fact lower throughput) Signed-off-by: Jan Beulich Signed-off-by: Konrad Rzeszutek Wilk --- drivers/block/xen-blkfront.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index 386b17208605..265f1a7072e9 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -2223,7 +2223,7 @@ static int blkfront_setup_indirect(struct blkfront_ring_info *rinfo) } else grants = info->max_indirect_segments; - psegs = grants / GRANTS_PER_PSEG; + psegs = DIV_ROUND_UP(grants, GRANTS_PER_PSEG); err = fill_grant_buffer(rinfo, (grants + INDIRECT_GREFS(grants)) * BLK_RING_SIZE(info)); @@ -2328,8 +2328,11 @@ static void blkfront_gather_backend_features(struct blkfront_info *info) indirect_segments = xenbus_read_unsigned(info->xbdev->otherend, "feature-max-indirect-segments", 0); - info->max_indirect_segments = min(indirect_segments, - xen_blkif_max_segments); + if (indirect_segments > xen_blkif_max_segments) + indirect_segments = xen_blkif_max_segments; + if (indirect_segments <= BLKIF_MAX_SEGMENTS_PER_REQUEST) + indirect_segments = 0; + info->max_indirect_segments = indirect_segments; } /* @@ -2652,6 +2655,9 @@ static int __init xlblk_init(void) if (!xen_domain()) return -ENODEV; + if (xen_blkif_max_segments < BLKIF_MAX_SEGMENTS_PER_REQUEST) + xen_blkif_max_segments = BLKIF_MAX_SEGMENTS_PER_REQUEST; + if (xen_blkif_max_ring_order > XENBUS_MAX_RING_GRANT_ORDER) { pr_info("Invalid max_ring_order (%d), will use default max: %d.\n", xen_blkif_max_ring_order, XENBUS_MAX_RING_GRANT_ORDER); From 690e5325b8c7d5db05fc569c0f7b888bb4248272 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 24 Jan 2017 14:50:19 +0100 Subject: [PATCH 03/10] block: fix use after free in __blkdev_direct_IO MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We can't dereference the dio structure after submitting the last bio for this request, as I/O completion might have happened before the code is run. Introduce a local is_sync variable instead. Fixes: 542ff7bf ("block: new direct I/O implementation") Signed-off-by: Christoph Hellwig Reported-by: Matias Bjørling Tested-by: Matias Bjørling Signed-off-by: Jens Axboe --- fs/block_dev.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/block_dev.c b/fs/block_dev.c index 5db5d1340d69..3c47614a4b32 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -331,7 +331,7 @@ __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter, int nr_pages) struct blk_plug plug; struct blkdev_dio *dio; struct bio *bio; - bool is_read = (iov_iter_rw(iter) == READ); + bool is_read = (iov_iter_rw(iter) == READ), is_sync; loff_t pos = iocb->ki_pos; blk_qc_t qc = BLK_QC_T_NONE; int ret; @@ -344,7 +344,7 @@ __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter, int nr_pages) bio_get(bio); /* extra ref for the completion handler */ dio = container_of(bio, struct blkdev_dio, bio); - dio->is_sync = is_sync_kiocb(iocb); + dio->is_sync = is_sync = is_sync_kiocb(iocb); if (dio->is_sync) dio->waiter = current; else @@ -398,7 +398,7 @@ __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter, int nr_pages) } blk_finish_plug(&plug); - if (!dio->is_sync) + if (!is_sync) return -EIOCBQUEUED; for (;;) { From c81e55e057b6458aac6d96a6429ef021b7f6f62c Mon Sep 17 00:00:00 2001 From: James Smart Date: Sat, 24 Dec 2016 09:46:43 -0800 Subject: [PATCH 04/10] nvmet_fc: correct logic in disconnect queue LS handling Correct logic in disconnect queue LS handling. Rework so that queue searching and error reporting is above the section to send back a ls rjt Signed-off-by: James Smart Signed-off-by: Sagi Grimberg --- drivers/nvme/target/fc.c | 36 ++++++++++++++++++++++-------------- 1 file changed, 22 insertions(+), 14 deletions(-) diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c index 173e842f19c9..ba57f9852bde 100644 --- a/drivers/nvme/target/fc.c +++ b/drivers/nvme/target/fc.c @@ -1314,7 +1314,7 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport, (struct fcnvme_ls_disconnect_rqst *)iod->rqstbuf; struct fcnvme_ls_disconnect_acc *acc = (struct fcnvme_ls_disconnect_acc *)iod->rspbuf; - struct nvmet_fc_tgt_queue *queue; + struct nvmet_fc_tgt_queue *queue = NULL; struct nvmet_fc_tgt_assoc *assoc; int ret = 0; bool del_assoc = false; @@ -1348,7 +1348,18 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport, assoc = nvmet_fc_find_target_assoc(tgtport, be64_to_cpu(rqst->associd.association_id)); iod->assoc = assoc; - if (!assoc) + if (assoc) { + if (rqst->discon_cmd.scope == + FCNVME_DISCONN_CONNECTION) { + queue = nvmet_fc_find_target_queue(tgtport, + be64_to_cpu( + rqst->discon_cmd.id)); + if (!queue) { + nvmet_fc_tgt_a_put(assoc); + ret = VERR_NO_CONN; + } + } + } else ret = VERR_NO_ASSOC; } @@ -1373,21 +1384,18 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport, FCNVME_LS_DISCONNECT); - if (rqst->discon_cmd.scope == FCNVME_DISCONN_CONNECTION) { - queue = nvmet_fc_find_target_queue(tgtport, - be64_to_cpu(rqst->discon_cmd.id)); - if (queue) { - int qid = queue->qid; + /* are we to delete a Connection ID (queue) */ + if (queue) { + int qid = queue->qid; - nvmet_fc_delete_target_queue(queue); + nvmet_fc_delete_target_queue(queue); - /* release the get taken by find_target_queue */ - nvmet_fc_tgt_q_put(queue); + /* release the get taken by find_target_queue */ + nvmet_fc_tgt_q_put(queue); - /* tear association down if io queue terminated */ - if (!qid) - del_assoc = true; - } + /* tear association down if io queue terminated */ + if (!qid) + del_assoc = true; } /* release get taken in nvmet_fc_find_target_assoc */ From 344770b07b7ae70639ebf110010eb6156a6e55e9 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Sun, 27 Nov 2016 22:29:17 +0200 Subject: [PATCH 05/10] nvmet: delete controllers deletion upon subsystem release No reason for them to be kept around if we are deleting the subsystem, so instead of passively wait for the host to disconnect, actively delete the controllers. Signed-off-by: Sagi Grimberg Reviewed-by: Johannes Thumshirn Reviewed-by: Max Gurtovoy Reviewed-by: Christoph Hellwig --- drivers/nvme/target/configfs.c | 1 + drivers/nvme/target/core.c | 10 ++++++++++ drivers/nvme/target/nvmet.h | 1 + 3 files changed, 12 insertions(+) diff --git a/drivers/nvme/target/configfs.c b/drivers/nvme/target/configfs.c index 6f5074153dcd..be8c800078e2 100644 --- a/drivers/nvme/target/configfs.c +++ b/drivers/nvme/target/configfs.c @@ -631,6 +631,7 @@ static void nvmet_subsys_release(struct config_item *item) { struct nvmet_subsys *subsys = to_subsys(item); + nvmet_subsys_del_ctrls(subsys); nvmet_subsys_put(subsys); } diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c index b1d66ed655c9..4a367549eb93 100644 --- a/drivers/nvme/target/core.c +++ b/drivers/nvme/target/core.c @@ -935,6 +935,16 @@ static void nvmet_subsys_free(struct kref *ref) kfree(subsys); } +void nvmet_subsys_del_ctrls(struct nvmet_subsys *subsys) +{ + struct nvmet_ctrl *ctrl; + + mutex_lock(&subsys->lock); + list_for_each_entry(ctrl, &subsys->ctrls, subsys_entry) + ctrl->ops->delete_ctrl(ctrl); + mutex_unlock(&subsys->lock); +} + void nvmet_subsys_put(struct nvmet_subsys *subsys) { kref_put(&subsys->ref, nvmet_subsys_free); diff --git a/drivers/nvme/target/nvmet.h b/drivers/nvme/target/nvmet.h index 23d5eb1c944f..cc7ad06b43a7 100644 --- a/drivers/nvme/target/nvmet.h +++ b/drivers/nvme/target/nvmet.h @@ -282,6 +282,7 @@ void nvmet_ctrl_put(struct nvmet_ctrl *ctrl); struct nvmet_subsys *nvmet_subsys_alloc(const char *subsysnqn, enum nvme_subsys_type type); void nvmet_subsys_put(struct nvmet_subsys *subsys); +void nvmet_subsys_del_ctrls(struct nvmet_subsys *subsys); struct nvmet_ns *nvmet_find_namespace(struct nvmet_ctrl *ctrl, __le32 nsid); void nvmet_put_namespace(struct nvmet_ns *ns); From 06406d81a2d7cfb8abcc4fa6cdfeb8e5897007c5 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Sun, 1 Jan 2017 13:41:56 +0200 Subject: [PATCH 06/10] nvmet: cancel fatal error and flush async work before free controller Make sure they are not running and we can free the controller safely. Signed-off-by: Roy Shterman Signed-off-by: Sagi Grimberg Reviewed-by: Christoph Hellwig --- drivers/nvme/target/core.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c index 4a367549eb93..a327a43f8166 100644 --- a/drivers/nvme/target/core.c +++ b/drivers/nvme/target/core.c @@ -816,6 +816,9 @@ static void nvmet_ctrl_free(struct kref *ref) list_del(&ctrl->subsys_entry); mutex_unlock(&subsys->lock); + flush_work(&ctrl->async_event_work); + cancel_work_sync(&ctrl->fatal_err_work); + ida_simple_remove(&subsys->cntlid_ida, ctrl->cntlid); nvmet_subsys_put(subsys); From 23a8ed4a624324dc696c328f09bd502c4a3816f0 Mon Sep 17 00:00:00 2001 From: Sagi Grimberg Date: Sun, 1 Jan 2017 13:18:26 +0200 Subject: [PATCH 07/10] nvmet: Call fatal_error from keep-alive timout expiration We only need to call delete_ctrl once, so given that both keep-alive timeout and any other fatal error can trigger it, just make sure we only call delete_ctrl once. Signed-off-by: Sagi Grimberg Reviewed-by: Christoph Hellwig --- drivers/nvme/target/core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c index a327a43f8166..fc5ba2f9e15f 100644 --- a/drivers/nvme/target/core.c +++ b/drivers/nvme/target/core.c @@ -200,7 +200,7 @@ static void nvmet_keep_alive_timer(struct work_struct *work) pr_err("ctrl %d keep-alive timer (%d seconds) expired!\n", ctrl->cntlid, ctrl->kato); - ctrl->ops->delete_ctrl(ctrl); + nvmet_ctrl_fatal_error(ctrl); } static void nvmet_start_keep_alive_timer(struct nvmet_ctrl *ctrl) From 748ff8408f8e208f279ba221e5c12612fbb4dddb Mon Sep 17 00:00:00 2001 From: Parav Pandit Date: Thu, 19 Jan 2017 09:55:08 -0600 Subject: [PATCH 08/10] nvmet-rdma: Fix missing dma sync to nvme data structures This patch performs dma sync operations on nvme_command and nvme_completion. nvme_command is synced (a) on receiving of the recv queue completion for cpu access. (b) before posting recv wqe back to rdma adapter for device access. nvme_completion is synced (a) on receiving of the recv queue completion of associated nvme_command for cpu access. (b) before posting send wqe to rdma adapter for device access. This patch is generated for git://git.infradead.org/nvme-fabrics.git Branch: nvmf-4.10 Signed-off-by: Parav Pandit Reviewed-by: Max Gurtovoy --- drivers/nvme/target/rdma.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c index 8c3760a78ac0..60990220bd83 100644 --- a/drivers/nvme/target/rdma.c +++ b/drivers/nvme/target/rdma.c @@ -438,6 +438,10 @@ static int nvmet_rdma_post_recv(struct nvmet_rdma_device *ndev, { struct ib_recv_wr *bad_wr; + ib_dma_sync_single_for_device(ndev->device, + cmd->sge[0].addr, cmd->sge[0].length, + DMA_FROM_DEVICE); + if (ndev->srq) return ib_post_srq_recv(ndev->srq, &cmd->wr, &bad_wr); return ib_post_recv(cmd->queue->cm_id->qp, &cmd->wr, &bad_wr); @@ -538,6 +542,11 @@ static void nvmet_rdma_queue_response(struct nvmet_req *req) first_wr = &rsp->send_wr; nvmet_rdma_post_recv(rsp->queue->dev, rsp->cmd); + + ib_dma_sync_single_for_device(rsp->queue->dev->device, + rsp->send_sge.addr, rsp->send_sge.length, + DMA_TO_DEVICE); + if (ib_post_send(cm_id->qp, first_wr, &bad_wr)) { pr_err("sending cmd response failed\n"); nvmet_rdma_release_rsp(rsp); @@ -698,6 +707,14 @@ static void nvmet_rdma_handle_command(struct nvmet_rdma_queue *queue, cmd->n_rdma = 0; cmd->req.port = queue->port; + + ib_dma_sync_single_for_cpu(queue->dev->device, + cmd->cmd->sge[0].addr, cmd->cmd->sge[0].length, + DMA_FROM_DEVICE); + ib_dma_sync_single_for_cpu(queue->dev->device, + cmd->send_sge.addr, cmd->send_sge.length, + DMA_TO_DEVICE); + if (!nvmet_req_init(&cmd->req, &queue->nvme_cq, &queue->nvme_sq, &nvmet_rdma_ops)) return; From 19e420bb4076ace670addc55300e3b8c4a02dfc6 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 19 Jan 2017 16:55:57 +0100 Subject: [PATCH 09/10] nvme-fc: use blk_rq_nr_phys_segments Without this deallocate won't work properly due to the mismatch of the bio/request size and the actual payload size. Signed-off-by: Christoph Hellwig Reviewed-by: James Smart Reviewed-by: Johannes Thumshirn Signed-off-by: Sagi Grimberg --- drivers/nvme/host/fc.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c index fcc9dcfdf675..e65041c640cb 100644 --- a/drivers/nvme/host/fc.c +++ b/drivers/nvme/host/fc.c @@ -1663,13 +1663,13 @@ nvme_fc_map_data(struct nvme_fc_ctrl *ctrl, struct request *rq, return 0; freq->sg_table.sgl = freq->first_sgl; - ret = sg_alloc_table_chained(&freq->sg_table, rq->nr_phys_segments, - freq->sg_table.sgl); + ret = sg_alloc_table_chained(&freq->sg_table, + blk_rq_nr_phys_segments(rq), freq->sg_table.sgl); if (ret) return -ENOMEM; op->nents = blk_rq_map_sg(rq->q, rq, freq->sg_table.sgl); - WARN_ON(op->nents > rq->nr_phys_segments); + WARN_ON(op->nents > blk_rq_nr_phys_segments(rq)); dir = (rq_data_dir(rq) == WRITE) ? DMA_TO_DEVICE : DMA_FROM_DEVICE; freq->sg_cnt = fc_dma_map_sg(ctrl->lport->dev, freq->sg_table.sgl, op->nents, dir); From 08965c2eba135bdfb6e86cf25308e01421c7e0ce Mon Sep 17 00:00:00 2001 From: Bart Van Assche Date: Wed, 25 Jan 2017 13:43:56 -0800 Subject: [PATCH 10/10] Revert "sd: remove __data_len hack for WRITE SAME" This patch reverts commit f80de881d8df and avoids that sending a WRITE SAME command to the iSCSI initiator triggers the following: BUG: unable to handle kernel NULL pointer dereference at 0000000000000014 TARGET_CORE[iSCSI]: Expected Transfer Length: 260096 does not match SCSI CDB Length: 512 for SAM Opcode: 0x41 IP: iscsi_tcp_segment_done+0x20b/0x310 [libiscsi_tcp] Oops: 0000 [#1] SMP Modules linked in: target_core_user uio target_core_iblock target_core_file iscsi_target_mod target_core_mod netconsole configfs crct10dif_pclmul crc32_pclmul ghash_clmulni_intel aesni_intel aes_x86_64 crypto_simd cryptd glue_helper virtio_console virtio_rng virtio_balloon serio_raw i2c_piix4 acpi_cpufreq button iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi ext4 jbd2 mbcache virtio_blk virtio_net psmouse floppy drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops ttm drm virtio_pci CPU: 2 PID: 5 Comm: kworker/u8:0 Not tainted 4.10.0-rc5-debug+ #3 Workqueue: iscsi_q_0 iscsi_xmitworker [libiscsi] RIP: 0010:iscsi_tcp_segment_done+0x20b/0x310 [libiscsi_tcp] Call Trace: iscsi_sw_tcp_xmit_segment+0x84/0x120 [iscsi_tcp] iscsi_sw_tcp_pdu_xmit+0x51/0x180 [iscsi_tcp] iscsi_tcp_task_xmit+0xb3/0x290 [libiscsi_tcp] iscsi_xmit_task+0x4e/0xc0 [libiscsi] iscsi_xmitworker+0x243/0x330 [libiscsi] process_one_work+0x1d8/0x4b0 worker_thread+0x49/0x4a0 kthread+0x102/0x140 Fixes: f80de881d8df ("sd: remove __data_len hack for WRITE SAME") Signed-off-by: Bart Van Assche Cc: Hannes Reinecke Cc: Sagi Grimberg Cc: Jens Axboe Cc: Lee Duncan Cc: Chris Leech Acked-by: Christoph Hellwig Acked-by: Martin K. Petersen Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 0b09638fa39b..1f5d92a25a49 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -836,6 +836,7 @@ static int sd_setup_write_same_cmnd(struct scsi_cmnd *cmd) struct bio *bio = rq->bio; sector_t sector = blk_rq_pos(rq); unsigned int nr_sectors = blk_rq_sectors(rq); + unsigned int nr_bytes = blk_rq_bytes(rq); int ret; if (sdkp->device->no_write_same) @@ -868,7 +869,21 @@ static int sd_setup_write_same_cmnd(struct scsi_cmnd *cmd) cmd->transfersize = sdp->sector_size; cmd->allowed = SD_MAX_RETRIES; - return scsi_init_io(cmd); + + /* + * For WRITE SAME the data transferred via the DATA OUT buffer is + * different from the amount of data actually written to the target. + * + * We set up __data_len to the amount of data transferred via the + * DATA OUT buffer so that blk_rq_map_sg sets up the proper S/G list + * to transfer a single sector of data first, but then reset it to + * the amount of data to be written right after so that the I/O path + * knows how much to actually write. + */ + rq->__data_len = sdp->sector_size; + ret = scsi_init_io(cmd); + rq->__data_len = nr_bytes; + return ret; } static int sd_setup_flush_cmnd(struct scsi_cmnd *cmd)