whiterose

linux unikernel
Log | Files | Refs | README | LICENSE | git clone https://git.ne02ptzero.me/git/whiterose

commit e8a71a38668919c53e6ca9dd1bfa977e5690523f
parent 2b9c272cf5cd81708e51b4ce3e432ce9566cfa47
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Fri, 15 Mar 2019 14:32:59 -0700

Merge tag 'ntb-5.1' of git://github.com/jonmason/ntb

Pull NTB updates from Jon Mason:

 - fixes for switchtec debugability and mapping table entries

 - NTB transport improvements

 - a reworking of the peer_db_addr for better abstraction

* tag 'ntb-5.1' of git://github.com/jonmason/ntb:
  NTB: add new parameter to peer_db_addr() db_bit and db_data
  NTB: ntb_transport: Ensure the destination buffer is mapped for TX DMA
  NTB: ntb_transport: Free MWs in ntb_transport_link_cleanup()
  ntb_hw_switchtec: Added support of >=4G memory windows
  ntb_hw_switchtec: NT req id mapping table register entry number should be 512
  ntb_hw_switchtec: debug print 64bit aligned crosslink BAR Numbers

Diffstat:
Mdrivers/ntb/hw/intel/ntb_hw_gen1.c | 25+++++++++++++++++++------
Mdrivers/ntb/hw/intel/ntb_hw_gen1.h | 5+++--
Mdrivers/ntb/hw/intel/ntb_hw_gen3.c | 33++++++++++++++++++++++++++++++++-
Mdrivers/ntb/hw/mscc/ntb_hw_switchtec.c | 20++++++++++++++++----
Mdrivers/ntb/ntb_transport.c | 31+++++++++++++++++++++++++++++--
Minclude/linux/ntb.h | 10+++++++---
Minclude/linux/switchtec.h | 10+++++++---
7 files changed, 113 insertions(+), 21 deletions(-)

diff --git a/drivers/ntb/hw/intel/ntb_hw_gen1.c b/drivers/ntb/hw/intel/ntb_hw_gen1.c @@ -180,7 +180,7 @@ int ndev_mw_to_bar(struct intel_ntb_dev *ndev, int idx) return ndev->reg->mw_bar[idx]; } -static inline int ndev_db_addr(struct intel_ntb_dev *ndev, +void ndev_db_addr(struct intel_ntb_dev *ndev, phys_addr_t *db_addr, resource_size_t *db_size, phys_addr_t reg_addr, unsigned long reg) { @@ -196,8 +196,6 @@ static inline int ndev_db_addr(struct intel_ntb_dev *ndev, *db_size = ndev->reg->db_size; dev_dbg(&ndev->ntb.pdev->dev, "Peer db size %llx\n", *db_size); } - - return 0; } u64 ndev_db_read(struct intel_ntb_dev *ndev, @@ -1111,13 +1109,28 @@ int intel_ntb_db_clear_mask(struct ntb_dev *ntb, u64 db_bits) ndev->self_reg->db_mask); } -int intel_ntb_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr, - resource_size_t *db_size) +static int intel_ntb_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr, + resource_size_t *db_size, u64 *db_data, int db_bit) { + u64 db_bits; struct intel_ntb_dev *ndev = ntb_ndev(ntb); - return ndev_db_addr(ndev, db_addr, db_size, ndev->peer_addr, + if (unlikely(db_bit >= BITS_PER_LONG_LONG)) + return -EINVAL; + + db_bits = BIT_ULL(db_bit); + + if (unlikely(db_bits & ~ntb_ndev(ntb)->db_valid_mask)) + return -EINVAL; + + ndev_db_addr(ndev, db_addr, db_size, ndev->peer_addr, ndev->peer_reg->db_bell); + + if (db_data) + *db_data = db_bits; + + + return 0; } static int intel_ntb_peer_db_set(struct ntb_dev *ntb, u64 db_bits) diff --git a/drivers/ntb/hw/intel/ntb_hw_gen1.h b/drivers/ntb/hw/intel/ntb_hw_gen1.h @@ -147,6 +147,9 @@ extern struct intel_b2b_addr xeon_b2b_dsd_addr; int ndev_init_isr(struct intel_ntb_dev *ndev, int msix_min, int msix_max, int msix_shift, int total_shift); enum ntb_topo xeon_ppd_topo(struct intel_ntb_dev *ndev, u8 ppd); +void ndev_db_addr(struct intel_ntb_dev *ndev, + phys_addr_t *db_addr, resource_size_t *db_size, + phys_addr_t reg_addr, unsigned long reg); u64 ndev_db_read(struct intel_ntb_dev *ndev, void __iomem *mmio); int ndev_db_write(struct intel_ntb_dev *ndev, u64 db_bits, void __iomem *mmio); @@ -166,8 +169,6 @@ int intel_ntb_db_vector_count(struct ntb_dev *ntb); u64 intel_ntb_db_vector_mask(struct ntb_dev *ntb, int db_vector); int intel_ntb_db_set_mask(struct ntb_dev *ntb, u64 db_bits); int intel_ntb_db_clear_mask(struct ntb_dev *ntb, u64 db_bits); -int intel_ntb_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr, - resource_size_t *db_size); int intel_ntb_spad_is_unsafe(struct ntb_dev *ntb); int intel_ntb_spad_count(struct ntb_dev *ntb); u32 intel_ntb_spad_read(struct ntb_dev *ntb, int idx); diff --git a/drivers/ntb/hw/intel/ntb_hw_gen3.c b/drivers/ntb/hw/intel/ntb_hw_gen3.c @@ -532,6 +532,37 @@ static int intel_ntb3_mw_set_trans(struct ntb_dev *ntb, int pidx, int idx, return 0; } +int intel_ntb3_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr, + resource_size_t *db_size, + u64 *db_data, int db_bit) +{ + phys_addr_t db_addr_base; + struct intel_ntb_dev *ndev = ntb_ndev(ntb); + + if (unlikely(db_bit >= BITS_PER_LONG_LONG)) + return -EINVAL; + + if (unlikely(BIT_ULL(db_bit) & ~ntb_ndev(ntb)->db_valid_mask)) + return -EINVAL; + + ndev_db_addr(ndev, &db_addr_base, db_size, ndev->peer_addr, + ndev->peer_reg->db_bell); + + if (db_addr) { + *db_addr = db_addr_base + (db_bit * 4); + dev_dbg(&ndev->ntb.pdev->dev, "Peer db addr %llx db bit %d\n", + *db_addr, db_bit); + } + + if (db_data) { + *db_data = 1; + dev_dbg(&ndev->ntb.pdev->dev, "Peer db data %llx db bit %d\n", + *db_data, db_bit); + } + + return 0; +} + static int intel_ntb3_peer_db_set(struct ntb_dev *ntb, u64 db_bits) { struct intel_ntb_dev *ndev = ntb_ndev(ntb); @@ -584,7 +615,7 @@ const struct ntb_dev_ops intel_ntb3_ops = { .db_clear = intel_ntb3_db_clear, .db_set_mask = intel_ntb_db_set_mask, .db_clear_mask = intel_ntb_db_clear_mask, - .peer_db_addr = intel_ntb_peer_db_addr, + .peer_db_addr = intel_ntb3_peer_db_addr, .peer_db_set = intel_ntb3_peer_db_set, .spad_is_unsafe = intel_ntb_spad_is_unsafe, .spad_count = intel_ntb_spad_count, diff --git a/drivers/ntb/hw/mscc/ntb_hw_switchtec.c b/drivers/ntb/hw/mscc/ntb_hw_switchtec.c @@ -236,6 +236,7 @@ static void switchtec_ntb_mw_clr_direct(struct switchtec_ntb *sndev, int idx) ctl_val &= ~NTB_CTRL_BAR_DIR_WIN_EN; iowrite32(ctl_val, &ctl->bar_entry[bar].ctl); iowrite32(0, &ctl->bar_entry[bar].win_size); + iowrite32(0, &ctl->bar_ext_entry[bar].win_size); iowrite64(sndev->self_partition, &ctl->bar_entry[bar].xlate_addr); } @@ -258,7 +259,9 @@ static void switchtec_ntb_mw_set_direct(struct switchtec_ntb *sndev, int idx, ctl_val |= NTB_CTRL_BAR_DIR_WIN_EN; iowrite32(ctl_val, &ctl->bar_entry[bar].ctl); - iowrite32(xlate_pos | size, &ctl->bar_entry[bar].win_size); + iowrite32(xlate_pos | (lower_32_bits(size) & 0xFFFFF000), + &ctl->bar_entry[bar].win_size); + iowrite32(upper_32_bits(size), &ctl->bar_ext_entry[bar].win_size); iowrite64(sndev->self_partition | addr, &ctl->bar_entry[bar].xlate_addr); } @@ -679,11 +682,16 @@ static u64 switchtec_ntb_db_read_mask(struct ntb_dev *ntb) static int switchtec_ntb_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr, - resource_size_t *db_size) + resource_size_t *db_size, + u64 *db_data, + int db_bit) { struct switchtec_ntb *sndev = ntb_sndev(ntb); unsigned long offset; + if (unlikely(db_bit >= BITS_PER_LONG_LONG)) + return -EINVAL; + offset = (unsigned long)sndev->mmio_peer_dbmsg->odb - (unsigned long)sndev->stdev->mmio; @@ -693,6 +701,8 @@ static int switchtec_ntb_peer_db_addr(struct ntb_dev *ntb, *db_addr = pci_resource_start(ntb->pdev, 0) + offset; if (db_size) *db_size = sizeof(u32); + if (db_data) + *db_data = BIT_ULL(db_bit) << sndev->db_peer_shift; return 0; } @@ -1025,7 +1035,9 @@ static int crosslink_setup_mws(struct switchtec_ntb *sndev, int ntb_lut_idx, ctl_val |= NTB_CTRL_BAR_DIR_WIN_EN; iowrite32(ctl_val, &ctl->bar_entry[bar].ctl); - iowrite32(xlate_pos | size, &ctl->bar_entry[bar].win_size); + iowrite32(xlate_pos | (lower_32_bits(size) & 0xFFFFF000), + &ctl->bar_entry[bar].win_size); + iowrite32(upper_32_bits(size), &ctl->bar_ext_entry[bar].win_size); iowrite64(sndev->peer_partition | addr, &ctl->bar_entry[bar].xlate_addr); } @@ -1092,7 +1104,7 @@ static int crosslink_enum_partition(struct switchtec_ntb *sndev, dev_dbg(&sndev->stdev->dev, "Crosslink BAR%d addr: %llx\n", - i, bar_addr); + i*2, bar_addr); if (bar_addr != bar_space * i) continue; diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c @@ -144,7 +144,9 @@ struct ntb_transport_qp { struct list_head tx_free_q; spinlock_t ntb_tx_free_q_lock; void __iomem *tx_mw; - dma_addr_t tx_mw_phys; + phys_addr_t tx_mw_phys; + size_t tx_mw_size; + dma_addr_t tx_mw_dma_addr; unsigned int tx_index; unsigned int tx_max_entry; unsigned int tx_max_frame; @@ -862,6 +864,9 @@ static void ntb_transport_link_cleanup(struct ntb_transport_ctx *nt) if (!nt->link_is_up) cancel_delayed_work_sync(&nt->link_work); + for (i = 0; i < nt->mw_count; i++) + ntb_free_mw(nt, i); + /* The scratchpad registers keep the values if the remote side * goes down, blast them now to give them a sane value the next * time they are accessed @@ -1049,6 +1054,7 @@ static int ntb_transport_init_queue(struct ntb_transport_ctx *nt, tx_size = (unsigned int)mw_size / num_qps_mw; qp_offset = tx_size * (qp_num / mw_count); + qp->tx_mw_size = tx_size; qp->tx_mw = nt->mw_vec[mw_num].vbase + qp_offset; if (!qp->tx_mw) return -EINVAL; @@ -1644,7 +1650,7 @@ static int ntb_async_tx_submit(struct ntb_transport_qp *qp, dma_cookie_t cookie; device = chan->device; - dest = qp->tx_mw_phys + qp->tx_max_frame * entry->tx_index; + dest = qp->tx_mw_dma_addr + qp->tx_max_frame * entry->tx_index; buff_off = (size_t)buf & ~PAGE_MASK; dest_off = (size_t)dest & ~PAGE_MASK; @@ -1863,6 +1869,18 @@ ntb_transport_create_queue(void *data, struct device *client_dev, qp->rx_dma_chan = NULL; } + if (qp->tx_dma_chan) { + qp->tx_mw_dma_addr = + dma_map_resource(qp->tx_dma_chan->device->dev, + qp->tx_mw_phys, qp->tx_mw_size, + DMA_FROM_DEVICE, 0); + if (dma_mapping_error(qp->tx_dma_chan->device->dev, + qp->tx_mw_dma_addr)) { + qp->tx_mw_dma_addr = 0; + goto err1; + } + } + dev_dbg(&pdev->dev, "Using %s memcpy for TX\n", qp->tx_dma_chan ? "DMA" : "CPU"); @@ -1904,6 +1922,10 @@ err1: qp->rx_alloc_entry = 0; while ((entry = ntb_list_rm(&qp->ntb_rx_q_lock, &qp->rx_free_q))) kfree(entry); + if (qp->tx_mw_dma_addr) + dma_unmap_resource(qp->tx_dma_chan->device->dev, + qp->tx_mw_dma_addr, qp->tx_mw_size, + DMA_FROM_DEVICE, 0); if (qp->tx_dma_chan) dma_release_channel(qp->tx_dma_chan); if (qp->rx_dma_chan) @@ -1945,6 +1967,11 @@ void ntb_transport_free_queue(struct ntb_transport_qp *qp) */ dma_sync_wait(chan, qp->last_cookie); dmaengine_terminate_all(chan); + + dma_unmap_resource(chan->device->dev, + qp->tx_mw_dma_addr, qp->tx_mw_size, + DMA_FROM_DEVICE, 0); + dma_release_channel(chan); } diff --git a/include/linux/ntb.h b/include/linux/ntb.h @@ -296,7 +296,8 @@ struct ntb_dev_ops { int (*db_clear_mask)(struct ntb_dev *ntb, u64 db_bits); int (*peer_db_addr)(struct ntb_dev *ntb, - phys_addr_t *db_addr, resource_size_t *db_size); + phys_addr_t *db_addr, resource_size_t *db_size, + u64 *db_data, int db_bit); u64 (*peer_db_read)(struct ntb_dev *ntb); int (*peer_db_set)(struct ntb_dev *ntb, u64 db_bits); int (*peer_db_clear)(struct ntb_dev *ntb, u64 db_bits); @@ -1078,6 +1079,8 @@ static inline int ntb_db_clear_mask(struct ntb_dev *ntb, u64 db_bits) * @ntb: NTB device context. * @db_addr: OUT - The address of the peer doorbell register. * @db_size: OUT - The number of bytes to write the peer doorbell register. + * @db_data: OUT - The data of peer doorbell register + * @db_bit: door bell bit number * * Return the address of the peer doorbell register. This may be used, for * example, by drivers that offload memory copy operations to a dma engine. @@ -1091,12 +1094,13 @@ static inline int ntb_db_clear_mask(struct ntb_dev *ntb, u64 db_bits) */ static inline int ntb_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr, - resource_size_t *db_size) + resource_size_t *db_size, + u64 *db_data, int db_bit) { if (!ntb->ops->peer_db_addr) return -EINVAL; - return ntb->ops->peer_db_addr(ntb, db_addr, db_size); + return ntb->ops->peer_db_addr(ntb, db_addr, db_size, db_data, db_bit); } /** diff --git a/include/linux/switchtec.h b/include/linux/switchtec.h @@ -248,9 +248,13 @@ struct ntb_ctrl_regs { u32 win_size; u64 xlate_addr; } bar_entry[6]; - u32 reserved2[216]; - u32 req_id_table[256]; - u32 reserved3[512]; + struct { + u32 win_size; + u32 reserved[3]; + } bar_ext_entry[6]; + u32 reserved2[192]; + u32 req_id_table[512]; + u32 reserved3[256]; u64 lut_entry[512]; } __packed;