X-Git-Url: https://gerrit.opnfv.org/gerrit/gitweb?a=blobdiff_plain;f=kernel%2Fdrivers%2Fscsi%2Faic94xx%2Faic94xx_hwi.c;fp=kernel%2Fdrivers%2Fscsi%2Faic94xx%2Faic94xx_hwi.c;h=9f636a34d5957c82e155fc63002301a4fbee8c3c;hb=9ca8dbcc65cfc63d6f5ef3312a33184e1d726e00;hp=0000000000000000000000000000000000000000;hpb=98260f3884f4a202f9ca5eabed40b1354c489b29;p=kvmfornfv.git diff --git a/kernel/drivers/scsi/aic94xx/aic94xx_hwi.c b/kernel/drivers/scsi/aic94xx/aic94xx_hwi.c new file mode 100644 index 000000000..9f636a34d --- /dev/null +++ b/kernel/drivers/scsi/aic94xx/aic94xx_hwi.c @@ -0,0 +1,1390 @@ +/* + * Aic94xx SAS/SATA driver hardware interface. + * + * Copyright (C) 2005 Adaptec, Inc. All rights reserved. + * Copyright (C) 2005 Luben Tuikov + * + * This file is licensed under GPLv2. + * + * This file is part of the aic94xx driver. + * + * The aic94xx driver is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License as + * published by the Free Software Foundation; version 2 of the + * License. + * + * The aic94xx driver is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with the aic94xx driver; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + * + */ + +#include +#include +#include +#include +#include + +#include "aic94xx.h" +#include "aic94xx_reg.h" +#include "aic94xx_hwi.h" +#include "aic94xx_seq.h" +#include "aic94xx_dump.h" + +u32 MBAR0_SWB_SIZE; + +/* ---------- Initialization ---------- */ + +static int asd_get_user_sas_addr(struct asd_ha_struct *asd_ha) +{ + /* adapter came with a sas address */ + if (asd_ha->hw_prof.sas_addr[0]) + return 0; + + return sas_request_addr(asd_ha->sas_ha.core.shost, + asd_ha->hw_prof.sas_addr); +} + +static void asd_propagate_sas_addr(struct asd_ha_struct *asd_ha) +{ + int i; + + for (i = 0; i < ASD_MAX_PHYS; i++) { + if (asd_ha->hw_prof.phy_desc[i].sas_addr[0] == 0) + continue; + /* Set a phy's address only if it has none. + */ + ASD_DPRINTK("setting phy%d addr to %llx\n", i, + SAS_ADDR(asd_ha->hw_prof.sas_addr)); + memcpy(asd_ha->hw_prof.phy_desc[i].sas_addr, + asd_ha->hw_prof.sas_addr, SAS_ADDR_SIZE); + } +} + +/* ---------- PHY initialization ---------- */ + +static void asd_init_phy_identify(struct asd_phy *phy) +{ + phy->identify_frame = phy->id_frm_tok->vaddr; + + memset(phy->identify_frame, 0, sizeof(*phy->identify_frame)); + + phy->identify_frame->dev_type = SAS_END_DEVICE; + if (phy->sas_phy.role & PHY_ROLE_INITIATOR) + phy->identify_frame->initiator_bits = phy->sas_phy.iproto; + if (phy->sas_phy.role & PHY_ROLE_TARGET) + phy->identify_frame->target_bits = phy->sas_phy.tproto; + memcpy(phy->identify_frame->sas_addr, phy->phy_desc->sas_addr, + SAS_ADDR_SIZE); + phy->identify_frame->phy_id = phy->sas_phy.id; +} + +static int asd_init_phy(struct asd_phy *phy) +{ + struct asd_ha_struct *asd_ha = phy->sas_phy.ha->lldd_ha; + struct asd_sas_phy *sas_phy = &phy->sas_phy; + + sas_phy->enabled = 1; + sas_phy->class = SAS; + sas_phy->iproto = SAS_PROTOCOL_ALL; + sas_phy->tproto = 0; + sas_phy->type = PHY_TYPE_PHYSICAL; + sas_phy->role = PHY_ROLE_INITIATOR; + sas_phy->oob_mode = OOB_NOT_CONNECTED; + sas_phy->linkrate = SAS_LINK_RATE_UNKNOWN; + + phy->id_frm_tok = asd_alloc_coherent(asd_ha, + sizeof(*phy->identify_frame), + GFP_KERNEL); + if (!phy->id_frm_tok) { + asd_printk("no mem for IDENTIFY for phy%d\n", sas_phy->id); + return -ENOMEM; + } else + asd_init_phy_identify(phy); + + memset(phy->frame_rcvd, 0, sizeof(phy->frame_rcvd)); + + return 0; +} + +static void asd_init_ports(struct asd_ha_struct *asd_ha) +{ + int i; + + spin_lock_init(&asd_ha->asd_ports_lock); + for (i = 0; i < ASD_MAX_PHYS; i++) { + struct asd_port *asd_port = &asd_ha->asd_ports[i]; + + memset(asd_port->sas_addr, 0, SAS_ADDR_SIZE); + memset(asd_port->attached_sas_addr, 0, SAS_ADDR_SIZE); + asd_port->phy_mask = 0; + asd_port->num_phys = 0; + } +} + +static int asd_init_phys(struct asd_ha_struct *asd_ha) +{ + u8 i; + u8 phy_mask = asd_ha->hw_prof.enabled_phys; + + for (i = 0; i < ASD_MAX_PHYS; i++) { + struct asd_phy *phy = &asd_ha->phys[i]; + + phy->phy_desc = &asd_ha->hw_prof.phy_desc[i]; + phy->asd_port = NULL; + + phy->sas_phy.enabled = 0; + phy->sas_phy.id = i; + phy->sas_phy.sas_addr = &phy->phy_desc->sas_addr[0]; + phy->sas_phy.frame_rcvd = &phy->frame_rcvd[0]; + phy->sas_phy.ha = &asd_ha->sas_ha; + phy->sas_phy.lldd_phy = phy; + } + + /* Now enable and initialize only the enabled phys. */ + for_each_phy(phy_mask, phy_mask, i) { + int err = asd_init_phy(&asd_ha->phys[i]); + if (err) + return err; + } + + return 0; +} + +/* ---------- Sliding windows ---------- */ + +static int asd_init_sw(struct asd_ha_struct *asd_ha) +{ + struct pci_dev *pcidev = asd_ha->pcidev; + int err; + u32 v; + + /* Unlock MBARs */ + err = pci_read_config_dword(pcidev, PCI_CONF_MBAR_KEY, &v); + if (err) { + asd_printk("couldn't access conf. space of %s\n", + pci_name(pcidev)); + goto Err; + } + if (v) + err = pci_write_config_dword(pcidev, PCI_CONF_MBAR_KEY, v); + if (err) { + asd_printk("couldn't write to MBAR_KEY of %s\n", + pci_name(pcidev)); + goto Err; + } + + /* Set sliding windows A, B and C to point to proper internal + * memory regions. + */ + pci_write_config_dword(pcidev, PCI_CONF_MBAR0_SWA, REG_BASE_ADDR); + pci_write_config_dword(pcidev, PCI_CONF_MBAR0_SWB, + REG_BASE_ADDR_CSEQCIO); + pci_write_config_dword(pcidev, PCI_CONF_MBAR0_SWC, REG_BASE_ADDR_EXSI); + asd_ha->io_handle[0].swa_base = REG_BASE_ADDR; + asd_ha->io_handle[0].swb_base = REG_BASE_ADDR_CSEQCIO; + asd_ha->io_handle[0].swc_base = REG_BASE_ADDR_EXSI; + MBAR0_SWB_SIZE = asd_ha->io_handle[0].len - 0x80; + if (!asd_ha->iospace) { + /* MBAR1 will point to OCM (On Chip Memory) */ + pci_write_config_dword(pcidev, PCI_CONF_MBAR1, OCM_BASE_ADDR); + asd_ha->io_handle[1].swa_base = OCM_BASE_ADDR; + } + spin_lock_init(&asd_ha->iolock); +Err: + return err; +} + +/* ---------- SCB initialization ---------- */ + +/** + * asd_init_scbs - manually allocate the first SCB. + * @asd_ha: pointer to host adapter structure + * + * This allocates the very first SCB which would be sent to the + * sequencer for execution. Its bus address is written to + * CSEQ_Q_NEW_POINTER, mode page 2, mode 8. Since the bus address of + * the _next_ scb to be DMA-ed to the host adapter is read from the last + * SCB DMA-ed to the host adapter, we have to always stay one step + * ahead of the sequencer and keep one SCB already allocated. + */ +static int asd_init_scbs(struct asd_ha_struct *asd_ha) +{ + struct asd_seq_data *seq = &asd_ha->seq; + int bitmap_bytes; + + /* allocate the index array and bitmap */ + asd_ha->seq.tc_index_bitmap_bits = asd_ha->hw_prof.max_scbs; + asd_ha->seq.tc_index_array = kzalloc(asd_ha->seq.tc_index_bitmap_bits* + sizeof(void *), GFP_KERNEL); + if (!asd_ha->seq.tc_index_array) + return -ENOMEM; + + bitmap_bytes = (asd_ha->seq.tc_index_bitmap_bits+7)/8; + bitmap_bytes = BITS_TO_LONGS(bitmap_bytes*8)*sizeof(unsigned long); + asd_ha->seq.tc_index_bitmap = kzalloc(bitmap_bytes, GFP_KERNEL); + if (!asd_ha->seq.tc_index_bitmap) + return -ENOMEM; + + spin_lock_init(&seq->tc_index_lock); + + seq->next_scb.size = sizeof(struct scb); + seq->next_scb.vaddr = dma_pool_alloc(asd_ha->scb_pool, GFP_KERNEL, + &seq->next_scb.dma_handle); + if (!seq->next_scb.vaddr) { + kfree(asd_ha->seq.tc_index_bitmap); + kfree(asd_ha->seq.tc_index_array); + asd_ha->seq.tc_index_bitmap = NULL; + asd_ha->seq.tc_index_array = NULL; + return -ENOMEM; + } + + seq->pending = 0; + spin_lock_init(&seq->pend_q_lock); + INIT_LIST_HEAD(&seq->pend_q); + + return 0; +} + +static void asd_get_max_scb_ddb(struct asd_ha_struct *asd_ha) +{ + asd_ha->hw_prof.max_scbs = asd_get_cmdctx_size(asd_ha)/ASD_SCB_SIZE; + asd_ha->hw_prof.max_ddbs = asd_get_devctx_size(asd_ha)/ASD_DDB_SIZE; + ASD_DPRINTK("max_scbs:%d, max_ddbs:%d\n", + asd_ha->hw_prof.max_scbs, + asd_ha->hw_prof.max_ddbs); +} + +/* ---------- Done List initialization ---------- */ + +static void asd_dl_tasklet_handler(unsigned long); + +static int asd_init_dl(struct asd_ha_struct *asd_ha) +{ + asd_ha->seq.actual_dl + = asd_alloc_coherent(asd_ha, + ASD_DL_SIZE * sizeof(struct done_list_struct), + GFP_KERNEL); + if (!asd_ha->seq.actual_dl) + return -ENOMEM; + asd_ha->seq.dl = asd_ha->seq.actual_dl->vaddr; + asd_ha->seq.dl_toggle = ASD_DEF_DL_TOGGLE; + asd_ha->seq.dl_next = 0; + tasklet_init(&asd_ha->seq.dl_tasklet, asd_dl_tasklet_handler, + (unsigned long) asd_ha); + + return 0; +} + +/* ---------- EDB and ESCB init ---------- */ + +static int asd_alloc_edbs(struct asd_ha_struct *asd_ha, gfp_t gfp_flags) +{ + struct asd_seq_data *seq = &asd_ha->seq; + int i; + + seq->edb_arr = kmalloc(seq->num_edbs*sizeof(*seq->edb_arr), gfp_flags); + if (!seq->edb_arr) + return -ENOMEM; + + for (i = 0; i < seq->num_edbs; i++) { + seq->edb_arr[i] = asd_alloc_coherent(asd_ha, ASD_EDB_SIZE, + gfp_flags); + if (!seq->edb_arr[i]) + goto Err_unroll; + memset(seq->edb_arr[i]->vaddr, 0, ASD_EDB_SIZE); + } + + ASD_DPRINTK("num_edbs:%d\n", seq->num_edbs); + + return 0; + +Err_unroll: + for (i-- ; i >= 0; i--) + asd_free_coherent(asd_ha, seq->edb_arr[i]); + kfree(seq->edb_arr); + seq->edb_arr = NULL; + + return -ENOMEM; +} + +static int asd_alloc_escbs(struct asd_ha_struct *asd_ha, + gfp_t gfp_flags) +{ + struct asd_seq_data *seq = &asd_ha->seq; + struct asd_ascb *escb; + int i, escbs; + + seq->escb_arr = kmalloc(seq->num_escbs*sizeof(*seq->escb_arr), + gfp_flags); + if (!seq->escb_arr) + return -ENOMEM; + + escbs = seq->num_escbs; + escb = asd_ascb_alloc_list(asd_ha, &escbs, gfp_flags); + if (!escb) { + asd_printk("couldn't allocate list of escbs\n"); + goto Err; + } + seq->num_escbs -= escbs; /* subtract what was not allocated */ + ASD_DPRINTK("num_escbs:%d\n", seq->num_escbs); + + for (i = 0; i < seq->num_escbs; i++, escb = list_entry(escb->list.next, + struct asd_ascb, + list)) { + seq->escb_arr[i] = escb; + escb->scb->header.opcode = EMPTY_SCB; + } + + return 0; +Err: + kfree(seq->escb_arr); + seq->escb_arr = NULL; + return -ENOMEM; + +} + +static void asd_assign_edbs2escbs(struct asd_ha_struct *asd_ha) +{ + struct asd_seq_data *seq = &asd_ha->seq; + int i, k, z = 0; + + for (i = 0; i < seq->num_escbs; i++) { + struct asd_ascb *ascb = seq->escb_arr[i]; + struct empty_scb *escb = &ascb->scb->escb; + + ascb->edb_index = z; + + escb->num_valid = ASD_EDBS_PER_SCB; + + for (k = 0; k < ASD_EDBS_PER_SCB; k++) { + struct sg_el *eb = &escb->eb[k]; + struct asd_dma_tok *edb = seq->edb_arr[z++]; + + memset(eb, 0, sizeof(*eb)); + eb->bus_addr = cpu_to_le64(((u64) edb->dma_handle)); + eb->size = cpu_to_le32(((u32) edb->size)); + } + } +} + +/** + * asd_init_escbs -- allocate and initialize empty scbs + * @asd_ha: pointer to host adapter structure + * + * An empty SCB has sg_elements of ASD_EDBS_PER_SCB (7) buffers. + * They transport sense data, etc. + */ +static int asd_init_escbs(struct asd_ha_struct *asd_ha) +{ + struct asd_seq_data *seq = &asd_ha->seq; + int err = 0; + + /* Allocate two empty data buffers (edb) per sequencer. */ + int edbs = 2*(1+asd_ha->hw_prof.num_phys); + + seq->num_escbs = (edbs+ASD_EDBS_PER_SCB-1)/ASD_EDBS_PER_SCB; + seq->num_edbs = seq->num_escbs * ASD_EDBS_PER_SCB; + + err = asd_alloc_edbs(asd_ha, GFP_KERNEL); + if (err) { + asd_printk("couldn't allocate edbs\n"); + return err; + } + + err = asd_alloc_escbs(asd_ha, GFP_KERNEL); + if (err) { + asd_printk("couldn't allocate escbs\n"); + return err; + } + + asd_assign_edbs2escbs(asd_ha); + /* In order to insure that normal SCBs do not overfill sequencer + * memory and leave no space for escbs (halting condition), + * we increment pending here by the number of escbs. However, + * escbs are never pending. + */ + seq->pending = seq->num_escbs; + seq->can_queue = 1 + (asd_ha->hw_prof.max_scbs - seq->pending)/2; + + return 0; +} + +/* ---------- HW initialization ---------- */ + +/** + * asd_chip_hardrst -- hard reset the chip + * @asd_ha: pointer to host adapter structure + * + * This takes 16 cycles and is synchronous to CFCLK, which runs + * at 200 MHz, so this should take at most 80 nanoseconds. + */ +int asd_chip_hardrst(struct asd_ha_struct *asd_ha) +{ + int i; + int count = 100; + u32 reg; + + for (i = 0 ; i < 4 ; i++) { + asd_write_reg_dword(asd_ha, COMBIST, HARDRST); + } + + do { + udelay(1); + reg = asd_read_reg_dword(asd_ha, CHIMINT); + if (reg & HARDRSTDET) { + asd_write_reg_dword(asd_ha, CHIMINT, + HARDRSTDET|PORRSTDET); + return 0; + } + } while (--count > 0); + + return -ENODEV; +} + +/** + * asd_init_chip -- initialize the chip + * @asd_ha: pointer to host adapter structure + * + * Hard resets the chip, disables HA interrupts, downloads the sequnecer + * microcode and starts the sequencers. The caller has to explicitly + * enable HA interrupts with asd_enable_ints(asd_ha). + */ +static int asd_init_chip(struct asd_ha_struct *asd_ha) +{ + int err; + + err = asd_chip_hardrst(asd_ha); + if (err) { + asd_printk("couldn't hard reset %s\n", + pci_name(asd_ha->pcidev)); + goto out; + } + + asd_disable_ints(asd_ha); + + err = asd_init_seqs(asd_ha); + if (err) { + asd_printk("couldn't init seqs for %s\n", + pci_name(asd_ha->pcidev)); + goto out; + } + + err = asd_start_seqs(asd_ha); + if (err) { + asd_printk("coudln't start seqs for %s\n", + pci_name(asd_ha->pcidev)); + goto out; + } +out: + return err; +} + +#define MAX_DEVS ((OCM_MAX_SIZE) / (ASD_DDB_SIZE)) + +static int max_devs = 0; +module_param_named(max_devs, max_devs, int, S_IRUGO); +MODULE_PARM_DESC(max_devs, "\n" + "\tMaximum number of SAS devices to support (not LUs).\n" + "\tDefault: 2176, Maximum: 65663.\n"); + +static int max_cmnds = 0; +module_param_named(max_cmnds, max_cmnds, int, S_IRUGO); +MODULE_PARM_DESC(max_cmnds, "\n" + "\tMaximum number of commands queuable.\n" + "\tDefault: 512, Maximum: 66047.\n"); + +static void asd_extend_devctx_ocm(struct asd_ha_struct *asd_ha) +{ + unsigned long dma_addr = OCM_BASE_ADDR; + u32 d; + + dma_addr -= asd_ha->hw_prof.max_ddbs * ASD_DDB_SIZE; + asd_write_reg_addr(asd_ha, DEVCTXBASE, (dma_addr_t) dma_addr); + d = asd_read_reg_dword(asd_ha, CTXDOMAIN); + d |= 4; + asd_write_reg_dword(asd_ha, CTXDOMAIN, d); + asd_ha->hw_prof.max_ddbs += MAX_DEVS; +} + +static int asd_extend_devctx(struct asd_ha_struct *asd_ha) +{ + dma_addr_t dma_handle; + unsigned long dma_addr; + u32 d; + int size; + + asd_extend_devctx_ocm(asd_ha); + + asd_ha->hw_prof.ddb_ext = NULL; + if (max_devs <= asd_ha->hw_prof.max_ddbs || max_devs > 0xFFFF) { + max_devs = asd_ha->hw_prof.max_ddbs; + return 0; + } + + size = (max_devs - asd_ha->hw_prof.max_ddbs + 1) * ASD_DDB_SIZE; + + asd_ha->hw_prof.ddb_ext = asd_alloc_coherent(asd_ha, size, GFP_KERNEL); + if (!asd_ha->hw_prof.ddb_ext) { + asd_printk("couldn't allocate memory for %d devices\n", + max_devs); + max_devs = asd_ha->hw_prof.max_ddbs; + return -ENOMEM; + } + dma_handle = asd_ha->hw_prof.ddb_ext->dma_handle; + dma_addr = ALIGN((unsigned long) dma_handle, ASD_DDB_SIZE); + dma_addr -= asd_ha->hw_prof.max_ddbs * ASD_DDB_SIZE; + dma_handle = (dma_addr_t) dma_addr; + asd_write_reg_addr(asd_ha, DEVCTXBASE, dma_handle); + d = asd_read_reg_dword(asd_ha, CTXDOMAIN); + d &= ~4; + asd_write_reg_dword(asd_ha, CTXDOMAIN, d); + + asd_ha->hw_prof.max_ddbs = max_devs; + + return 0; +} + +static int asd_extend_cmdctx(struct asd_ha_struct *asd_ha) +{ + dma_addr_t dma_handle; + unsigned long dma_addr; + u32 d; + int size; + + asd_ha->hw_prof.scb_ext = NULL; + if (max_cmnds <= asd_ha->hw_prof.max_scbs || max_cmnds > 0xFFFF) { + max_cmnds = asd_ha->hw_prof.max_scbs; + return 0; + } + + size = (max_cmnds - asd_ha->hw_prof.max_scbs + 1) * ASD_SCB_SIZE; + + asd_ha->hw_prof.scb_ext = asd_alloc_coherent(asd_ha, size, GFP_KERNEL); + if (!asd_ha->hw_prof.scb_ext) { + asd_printk("couldn't allocate memory for %d commands\n", + max_cmnds); + max_cmnds = asd_ha->hw_prof.max_scbs; + return -ENOMEM; + } + dma_handle = asd_ha->hw_prof.scb_ext->dma_handle; + dma_addr = ALIGN((unsigned long) dma_handle, ASD_SCB_SIZE); + dma_addr -= asd_ha->hw_prof.max_scbs * ASD_SCB_SIZE; + dma_handle = (dma_addr_t) dma_addr; + asd_write_reg_addr(asd_ha, CMDCTXBASE, dma_handle); + d = asd_read_reg_dword(asd_ha, CTXDOMAIN); + d &= ~1; + asd_write_reg_dword(asd_ha, CTXDOMAIN, d); + + asd_ha->hw_prof.max_scbs = max_cmnds; + + return 0; +} + +/** + * asd_init_ctxmem -- initialize context memory + * asd_ha: pointer to host adapter structure + * + * This function sets the maximum number of SCBs and + * DDBs which can be used by the sequencer. This is normally + * 512 and 128 respectively. If support for more SCBs or more DDBs + * is required then CMDCTXBASE, DEVCTXBASE and CTXDOMAIN are + * initialized here to extend context memory to point to host memory, + * thus allowing unlimited support for SCBs and DDBs -- only limited + * by host memory. + */ +static int asd_init_ctxmem(struct asd_ha_struct *asd_ha) +{ + int bitmap_bytes; + + asd_get_max_scb_ddb(asd_ha); + asd_extend_devctx(asd_ha); + asd_extend_cmdctx(asd_ha); + + /* The kernel wants bitmaps to be unsigned long sized. */ + bitmap_bytes = (asd_ha->hw_prof.max_ddbs+7)/8; + bitmap_bytes = BITS_TO_LONGS(bitmap_bytes*8)*sizeof(unsigned long); + asd_ha->hw_prof.ddb_bitmap = kzalloc(bitmap_bytes, GFP_KERNEL); + if (!asd_ha->hw_prof.ddb_bitmap) + return -ENOMEM; + spin_lock_init(&asd_ha->hw_prof.ddb_lock); + + return 0; +} + +int asd_init_hw(struct asd_ha_struct *asd_ha) +{ + int err; + u32 v; + + err = asd_init_sw(asd_ha); + if (err) + return err; + + err = pci_read_config_dword(asd_ha->pcidev, PCIC_HSTPCIX_CNTRL, &v); + if (err) { + asd_printk("couldn't read PCIC_HSTPCIX_CNTRL of %s\n", + pci_name(asd_ha->pcidev)); + return err; + } + pci_write_config_dword(asd_ha->pcidev, PCIC_HSTPCIX_CNTRL, + v | SC_TMR_DIS); + if (err) { + asd_printk("couldn't disable split completion timer of %s\n", + pci_name(asd_ha->pcidev)); + return err; + } + + err = asd_read_ocm(asd_ha); + if (err) { + asd_printk("couldn't read ocm(%d)\n", err); + /* While suspicios, it is not an error that we + * couldn't read the OCM. */ + } + + err = asd_read_flash(asd_ha); + if (err) { + asd_printk("couldn't read flash(%d)\n", err); + /* While suspicios, it is not an error that we + * couldn't read FLASH memory. + */ + } + + asd_init_ctxmem(asd_ha); + + if (asd_get_user_sas_addr(asd_ha)) { + asd_printk("No SAS Address provided for %s\n", + pci_name(asd_ha->pcidev)); + err = -ENODEV; + goto Out; + } + + asd_propagate_sas_addr(asd_ha); + + err = asd_init_phys(asd_ha); + if (err) { + asd_printk("couldn't initialize phys for %s\n", + pci_name(asd_ha->pcidev)); + goto Out; + } + + asd_init_ports(asd_ha); + + err = asd_init_scbs(asd_ha); + if (err) { + asd_printk("couldn't initialize scbs for %s\n", + pci_name(asd_ha->pcidev)); + goto Out; + } + + err = asd_init_dl(asd_ha); + if (err) { + asd_printk("couldn't initialize the done list:%d\n", + err); + goto Out; + } + + err = asd_init_escbs(asd_ha); + if (err) { + asd_printk("couldn't initialize escbs\n"); + goto Out; + } + + err = asd_init_chip(asd_ha); + if (err) { + asd_printk("couldn't init the chip\n"); + goto Out; + } +Out: + return err; +} + +/* ---------- Chip reset ---------- */ + +/** + * asd_chip_reset -- reset the host adapter, etc + * @asd_ha: pointer to host adapter structure of interest + * + * Called from the ISR. Hard reset the chip. Let everything + * timeout. This should be no different than hot-unplugging the + * host adapter. Once everything times out we'll init the chip with + * a call to asd_init_chip() and enable interrupts with asd_enable_ints(). + * XXX finish. + */ +static void asd_chip_reset(struct asd_ha_struct *asd_ha) +{ + struct sas_ha_struct *sas_ha = &asd_ha->sas_ha; + + ASD_DPRINTK("chip reset for %s\n", pci_name(asd_ha->pcidev)); + asd_chip_hardrst(asd_ha); + sas_ha->notify_ha_event(sas_ha, HAE_RESET); +} + +/* ---------- Done List Routines ---------- */ + +static void asd_dl_tasklet_handler(unsigned long data) +{ + struct asd_ha_struct *asd_ha = (struct asd_ha_struct *) data; + struct asd_seq_data *seq = &asd_ha->seq; + unsigned long flags; + + while (1) { + struct done_list_struct *dl = &seq->dl[seq->dl_next]; + struct asd_ascb *ascb; + + if ((dl->toggle & DL_TOGGLE_MASK) != seq->dl_toggle) + break; + + /* find the aSCB */ + spin_lock_irqsave(&seq->tc_index_lock, flags); + ascb = asd_tc_index_find(seq, (int)le16_to_cpu(dl->index)); + spin_unlock_irqrestore(&seq->tc_index_lock, flags); + if (unlikely(!ascb)) { + ASD_DPRINTK("BUG:sequencer:dl:no ascb?!\n"); + goto next_1; + } else if (ascb->scb->header.opcode == EMPTY_SCB) { + goto out; + } else if (!ascb->uldd_timer && !del_timer(&ascb->timer)) { + goto next_1; + } + spin_lock_irqsave(&seq->pend_q_lock, flags); + list_del_init(&ascb->list); + seq->pending--; + spin_unlock_irqrestore(&seq->pend_q_lock, flags); + out: + ascb->tasklet_complete(ascb, dl); + + next_1: + seq->dl_next = (seq->dl_next + 1) & (ASD_DL_SIZE-1); + if (!seq->dl_next) + seq->dl_toggle ^= DL_TOGGLE_MASK; + } +} + +/* ---------- Interrupt Service Routines ---------- */ + +/** + * asd_process_donelist_isr -- schedule processing of done list entries + * @asd_ha: pointer to host adapter structure + */ +static void asd_process_donelist_isr(struct asd_ha_struct *asd_ha) +{ + tasklet_schedule(&asd_ha->seq.dl_tasklet); +} + +/** + * asd_com_sas_isr -- process device communication interrupt (COMINT) + * @asd_ha: pointer to host adapter structure + */ +static void asd_com_sas_isr(struct asd_ha_struct *asd_ha) +{ + u32 comstat = asd_read_reg_dword(asd_ha, COMSTAT); + + /* clear COMSTAT int */ + asd_write_reg_dword(asd_ha, COMSTAT, 0xFFFFFFFF); + + if (comstat & CSBUFPERR) { + asd_printk("%s: command/status buffer dma parity error\n", + pci_name(asd_ha->pcidev)); + } else if (comstat & CSERR) { + int i; + u32 dmaerr = asd_read_reg_dword(asd_ha, DMAERR); + dmaerr &= 0xFF; + asd_printk("%s: command/status dma error, DMAERR: 0x%02x, " + "CSDMAADR: 0x%04x, CSDMAADR+4: 0x%04x\n", + pci_name(asd_ha->pcidev), + dmaerr, + asd_read_reg_dword(asd_ha, CSDMAADR), + asd_read_reg_dword(asd_ha, CSDMAADR+4)); + asd_printk("CSBUFFER:\n"); + for (i = 0; i < 8; i++) { + asd_printk("%08x %08x %08x %08x\n", + asd_read_reg_dword(asd_ha, CSBUFFER), + asd_read_reg_dword(asd_ha, CSBUFFER+4), + asd_read_reg_dword(asd_ha, CSBUFFER+8), + asd_read_reg_dword(asd_ha, CSBUFFER+12)); + } + asd_dump_seq_state(asd_ha, 0); + } else if (comstat & OVLYERR) { + u32 dmaerr = asd_read_reg_dword(asd_ha, DMAERR); + dmaerr = (dmaerr >> 8) & 0xFF; + asd_printk("%s: overlay dma error:0x%x\n", + pci_name(asd_ha->pcidev), + dmaerr); + } + asd_chip_reset(asd_ha); +} + +static void asd_arp2_err(struct asd_ha_struct *asd_ha, u32 dchstatus) +{ + static const char *halt_code[256] = { + "UNEXPECTED_INTERRUPT0", + "UNEXPECTED_INTERRUPT1", + "UNEXPECTED_INTERRUPT2", + "UNEXPECTED_INTERRUPT3", + "UNEXPECTED_INTERRUPT4", + "UNEXPECTED_INTERRUPT5", + "UNEXPECTED_INTERRUPT6", + "UNEXPECTED_INTERRUPT7", + "UNEXPECTED_INTERRUPT8", + "UNEXPECTED_INTERRUPT9", + "UNEXPECTED_INTERRUPT10", + [11 ... 19] = "unknown[11,19]", + "NO_FREE_SCB_AVAILABLE", + "INVALID_SCB_OPCODE", + "INVALID_MBX_OPCODE", + "INVALID_ATA_STATE", + "ATA_QUEUE_FULL", + "ATA_TAG_TABLE_FAULT", + "ATA_TAG_MASK_FAULT", + "BAD_LINK_QUEUE_STATE", + "DMA2CHIM_QUEUE_ERROR", + "EMPTY_SCB_LIST_FULL", + "unknown[30]", + "IN_USE_SCB_ON_FREE_LIST", + "BAD_OPEN_WAIT_STATE", + "INVALID_STP_AFFILIATION", + "unknown[34]", + "EXEC_QUEUE_ERROR", + "TOO_MANY_EMPTIES_NEEDED", + "EMPTY_REQ_QUEUE_ERROR", + "Q_MONIRTT_MGMT_ERROR", + "TARGET_MODE_FLOW_ERROR", + "DEVICE_QUEUE_NOT_FOUND", + "START_IRTT_TIMER_ERROR", + "ABORT_TASK_ILLEGAL_REQ", + [43 ... 255] = "unknown[43,255]" + }; + + if (dchstatus & CSEQINT) { + u32 arp2int = asd_read_reg_dword(asd_ha, CARP2INT); + + if (arp2int & (ARP2WAITTO|ARP2ILLOPC|ARP2PERR|ARP2CIOPERR)) { + asd_printk("%s: CSEQ arp2int:0x%x\n", + pci_name(asd_ha->pcidev), + arp2int); + } else if (arp2int & ARP2HALTC) + asd_printk("%s: CSEQ halted: %s\n", + pci_name(asd_ha->pcidev), + halt_code[(arp2int>>16)&0xFF]); + else + asd_printk("%s: CARP2INT:0x%x\n", + pci_name(asd_ha->pcidev), + arp2int); + } + if (dchstatus & LSEQINT_MASK) { + int lseq; + u8 lseq_mask = dchstatus & LSEQINT_MASK; + + for_each_sequencer(lseq_mask, lseq_mask, lseq) { + u32 arp2int = asd_read_reg_dword(asd_ha, + LmARP2INT(lseq)); + if (arp2int & (ARP2WAITTO | ARP2ILLOPC | ARP2PERR + | ARP2CIOPERR)) { + asd_printk("%s: LSEQ%d arp2int:0x%x\n", + pci_name(asd_ha->pcidev), + lseq, arp2int); + /* XXX we should only do lseq reset */ + } else if (arp2int & ARP2HALTC) + asd_printk("%s: LSEQ%d halted: %s\n", + pci_name(asd_ha->pcidev), + lseq,halt_code[(arp2int>>16)&0xFF]); + else + asd_printk("%s: LSEQ%d ARP2INT:0x%x\n", + pci_name(asd_ha->pcidev), lseq, + arp2int); + } + } + asd_chip_reset(asd_ha); +} + +/** + * asd_dch_sas_isr -- process device channel interrupt (DEVINT) + * @asd_ha: pointer to host adapter structure + */ +static void asd_dch_sas_isr(struct asd_ha_struct *asd_ha) +{ + u32 dchstatus = asd_read_reg_dword(asd_ha, DCHSTATUS); + + if (dchstatus & CFIFTOERR) { + asd_printk("%s: CFIFTOERR\n", pci_name(asd_ha->pcidev)); + asd_chip_reset(asd_ha); + } else + asd_arp2_err(asd_ha, dchstatus); +} + +/** + * ads_rbi_exsi_isr -- process external system interface interrupt (INITERR) + * @asd_ha: pointer to host adapter structure + */ +static void asd_rbi_exsi_isr(struct asd_ha_struct *asd_ha) +{ + u32 stat0r = asd_read_reg_dword(asd_ha, ASISTAT0R); + + if (!(stat0r & ASIERR)) { + asd_printk("hmm, EXSI interrupted but no error?\n"); + return; + } + + if (stat0r & ASIFMTERR) { + asd_printk("ASI SEEPROM format error for %s\n", + pci_name(asd_ha->pcidev)); + } else if (stat0r & ASISEECHKERR) { + u32 stat1r = asd_read_reg_dword(asd_ha, ASISTAT1R); + asd_printk("ASI SEEPROM checksum 0x%x error for %s\n", + stat1r & CHECKSUM_MASK, + pci_name(asd_ha->pcidev)); + } else { + u32 statr = asd_read_reg_dword(asd_ha, ASIERRSTATR); + + if (!(statr & CPI2ASIMSTERR_MASK)) { + ASD_DPRINTK("hmm, ASIERR?\n"); + return; + } else { + u32 addr = asd_read_reg_dword(asd_ha, ASIERRADDR); + u32 data = asd_read_reg_dword(asd_ha, ASIERRDATAR); + + asd_printk("%s: CPI2 xfer err: addr: 0x%x, wdata: 0x%x, " + "count: 0x%x, byteen: 0x%x, targerr: 0x%x " + "master id: 0x%x, master err: 0x%x\n", + pci_name(asd_ha->pcidev), + addr, data, + (statr & CPI2ASIBYTECNT_MASK) >> 16, + (statr & CPI2ASIBYTEEN_MASK) >> 12, + (statr & CPI2ASITARGERR_MASK) >> 8, + (statr & CPI2ASITARGMID_MASK) >> 4, + (statr & CPI2ASIMSTERR_MASK)); + } + } + asd_chip_reset(asd_ha); +} + +/** + * asd_hst_pcix_isr -- process host interface interrupts + * @asd_ha: pointer to host adapter structure + * + * Asserted on PCIX errors: target abort, etc. + */ +static void asd_hst_pcix_isr(struct asd_ha_struct *asd_ha) +{ + u16 status; + u32 pcix_status; + u32 ecc_status; + + pci_read_config_word(asd_ha->pcidev, PCI_STATUS, &status); + pci_read_config_dword(asd_ha->pcidev, PCIX_STATUS, &pcix_status); + pci_read_config_dword(asd_ha->pcidev, ECC_CTRL_STAT, &ecc_status); + + if (status & PCI_STATUS_DETECTED_PARITY) + asd_printk("parity error for %s\n", pci_name(asd_ha->pcidev)); + else if (status & PCI_STATUS_REC_MASTER_ABORT) + asd_printk("master abort for %s\n", pci_name(asd_ha->pcidev)); + else if (status & PCI_STATUS_REC_TARGET_ABORT) + asd_printk("target abort for %s\n", pci_name(asd_ha->pcidev)); + else if (status & PCI_STATUS_PARITY) + asd_printk("data parity for %s\n", pci_name(asd_ha->pcidev)); + else if (pcix_status & RCV_SCE) { + asd_printk("received split completion error for %s\n", + pci_name(asd_ha->pcidev)); + pci_write_config_dword(asd_ha->pcidev,PCIX_STATUS,pcix_status); + /* XXX: Abort task? */ + return; + } else if (pcix_status & UNEXP_SC) { + asd_printk("unexpected split completion for %s\n", + pci_name(asd_ha->pcidev)); + pci_write_config_dword(asd_ha->pcidev,PCIX_STATUS,pcix_status); + /* ignore */ + return; + } else if (pcix_status & SC_DISCARD) + asd_printk("split completion discarded for %s\n", + pci_name(asd_ha->pcidev)); + else if (ecc_status & UNCOR_ECCERR) + asd_printk("uncorrectable ECC error for %s\n", + pci_name(asd_ha->pcidev)); + asd_chip_reset(asd_ha); +} + +/** + * asd_hw_isr -- host adapter interrupt service routine + * @irq: ignored + * @dev_id: pointer to host adapter structure + * + * The ISR processes done list entries and level 3 error handling. + */ +irqreturn_t asd_hw_isr(int irq, void *dev_id) +{ + struct asd_ha_struct *asd_ha = dev_id; + u32 chimint = asd_read_reg_dword(asd_ha, CHIMINT); + + if (!chimint) + return IRQ_NONE; + + asd_write_reg_dword(asd_ha, CHIMINT, chimint); + (void) asd_read_reg_dword(asd_ha, CHIMINT); + + if (chimint & DLAVAIL) + asd_process_donelist_isr(asd_ha); + if (chimint & COMINT) + asd_com_sas_isr(asd_ha); + if (chimint & DEVINT) + asd_dch_sas_isr(asd_ha); + if (chimint & INITERR) + asd_rbi_exsi_isr(asd_ha); + if (chimint & HOSTERR) + asd_hst_pcix_isr(asd_ha); + + return IRQ_HANDLED; +} + +/* ---------- SCB handling ---------- */ + +static struct asd_ascb *asd_ascb_alloc(struct asd_ha_struct *asd_ha, + gfp_t gfp_flags) +{ + extern struct kmem_cache *asd_ascb_cache; + struct asd_seq_data *seq = &asd_ha->seq; + struct asd_ascb *ascb; + unsigned long flags; + + ascb = kmem_cache_zalloc(asd_ascb_cache, gfp_flags); + + if (ascb) { + ascb->dma_scb.size = sizeof(struct scb); + ascb->dma_scb.vaddr = dma_pool_alloc(asd_ha->scb_pool, + gfp_flags, + &ascb->dma_scb.dma_handle); + if (!ascb->dma_scb.vaddr) { + kmem_cache_free(asd_ascb_cache, ascb); + return NULL; + } + memset(ascb->dma_scb.vaddr, 0, sizeof(struct scb)); + asd_init_ascb(asd_ha, ascb); + + spin_lock_irqsave(&seq->tc_index_lock, flags); + ascb->tc_index = asd_tc_index_get(seq, ascb); + spin_unlock_irqrestore(&seq->tc_index_lock, flags); + if (ascb->tc_index == -1) + goto undo; + + ascb->scb->header.index = cpu_to_le16((u16)ascb->tc_index); + } + + return ascb; +undo: + dma_pool_free(asd_ha->scb_pool, ascb->dma_scb.vaddr, + ascb->dma_scb.dma_handle); + kmem_cache_free(asd_ascb_cache, ascb); + ASD_DPRINTK("no index for ascb\n"); + return NULL; +} + +/** + * asd_ascb_alloc_list -- allocate a list of aSCBs + * @asd_ha: pointer to host adapter structure + * @num: pointer to integer number of aSCBs + * @gfp_flags: GFP_ flags. + * + * This is the only function which is used to allocate aSCBs. + * It can allocate one or many. If more than one, then they form + * a linked list in two ways: by their list field of the ascb struct + * and by the next_scb field of the scb_header. + * + * Returns NULL if no memory was available, else pointer to a list + * of ascbs. When this function returns, @num would be the number + * of SCBs which were not able to be allocated, 0 if all requested + * were able to be allocated. + */ +struct asd_ascb *asd_ascb_alloc_list(struct asd_ha_struct + *asd_ha, int *num, + gfp_t gfp_flags) +{ + struct asd_ascb *first = NULL; + + for ( ; *num > 0; --*num) { + struct asd_ascb *ascb = asd_ascb_alloc(asd_ha, gfp_flags); + + if (!ascb) + break; + else if (!first) + first = ascb; + else { + struct asd_ascb *last = list_entry(first->list.prev, + struct asd_ascb, + list); + list_add_tail(&ascb->list, &first->list); + last->scb->header.next_scb = + cpu_to_le64(((u64)ascb->dma_scb.dma_handle)); + } + } + + return first; +} + +/** + * asd_swap_head_scb -- swap the head scb + * @asd_ha: pointer to host adapter structure + * @ascb: pointer to the head of an ascb list + * + * The sequencer knows the DMA address of the next SCB to be DMAed to + * the host adapter, from initialization or from the last list DMAed. + * seq->next_scb keeps the address of this SCB. The sequencer will + * DMA to the host adapter this list of SCBs. But the head (first + * element) of this list is not known to the sequencer. Here we swap + * the head of the list with the known SCB (memcpy()). + * Only one memcpy() is required per list so it is in our interest + * to keep the list of SCB as long as possible so that the ratio + * of number of memcpy calls to the number of SCB DMA-ed is as small + * as possible. + * + * LOCKING: called with the pending list lock held. + */ +static void asd_swap_head_scb(struct asd_ha_struct *asd_ha, + struct asd_ascb *ascb) +{ + struct asd_seq_data *seq = &asd_ha->seq; + struct asd_ascb *last = list_entry(ascb->list.prev, + struct asd_ascb, + list); + struct asd_dma_tok t = ascb->dma_scb; + + memcpy(seq->next_scb.vaddr, ascb->scb, sizeof(*ascb->scb)); + ascb->dma_scb = seq->next_scb; + ascb->scb = ascb->dma_scb.vaddr; + seq->next_scb = t; + last->scb->header.next_scb = + cpu_to_le64(((u64)seq->next_scb.dma_handle)); +} + +/** + * asd_start_timers -- (add and) start timers of SCBs + * @list: pointer to struct list_head of the scbs + * @to: timeout in jiffies + * + * If an SCB in the @list has no timer function, assign the default + * one, then start the timer of the SCB. This function is + * intended to be called from asd_post_ascb_list(), just prior to + * posting the SCBs to the sequencer. + */ +static void asd_start_scb_timers(struct list_head *list) +{ + struct asd_ascb *ascb; + list_for_each_entry(ascb, list, list) { + if (!ascb->uldd_timer) { + ascb->timer.data = (unsigned long) ascb; + ascb->timer.function = asd_ascb_timedout; + ascb->timer.expires = jiffies + AIC94XX_SCB_TIMEOUT; + add_timer(&ascb->timer); + } + } +} + +/** + * asd_post_ascb_list -- post a list of 1 or more aSCBs to the host adapter + * @asd_ha: pointer to a host adapter structure + * @ascb: pointer to the first aSCB in the list + * @num: number of aSCBs in the list (to be posted) + * + * See queueing comment in asd_post_escb_list(). + * + * Additional note on queuing: In order to minimize the ratio of memcpy() + * to the number of ascbs sent, we try to batch-send as many ascbs as possible + * in one go. + * Two cases are possible: + * A) can_queue >= num, + * B) can_queue < num. + * Case A: we can send the whole batch at once. Increment "pending" + * in the beginning of this function, when it is checked, in order to + * eliminate races when this function is called by multiple processes. + * Case B: should never happen. + */ +int asd_post_ascb_list(struct asd_ha_struct *asd_ha, struct asd_ascb *ascb, + int num) +{ + unsigned long flags; + LIST_HEAD(list); + int can_queue; + + spin_lock_irqsave(&asd_ha->seq.pend_q_lock, flags); + can_queue = asd_ha->hw_prof.max_scbs - asd_ha->seq.pending; + if (can_queue >= num) + asd_ha->seq.pending += num; + else + can_queue = 0; + + if (!can_queue) { + spin_unlock_irqrestore(&asd_ha->seq.pend_q_lock, flags); + asd_printk("%s: scb queue full\n", pci_name(asd_ha->pcidev)); + return -SAS_QUEUE_FULL; + } + + asd_swap_head_scb(asd_ha, ascb); + + __list_add(&list, ascb->list.prev, &ascb->list); + + asd_start_scb_timers(&list); + + asd_ha->seq.scbpro += num; + list_splice_init(&list, asd_ha->seq.pend_q.prev); + asd_write_reg_dword(asd_ha, SCBPRO, (u32)asd_ha->seq.scbpro); + spin_unlock_irqrestore(&asd_ha->seq.pend_q_lock, flags); + + return 0; +} + +/** + * asd_post_escb_list -- post a list of 1 or more empty scb + * @asd_ha: pointer to a host adapter structure + * @ascb: pointer to the first empty SCB in the list + * @num: number of aSCBs in the list (to be posted) + * + * This is essentially the same as asd_post_ascb_list, but we do not + * increment pending, add those to the pending list or get indexes. + * See asd_init_escbs() and asd_init_post_escbs(). + * + * Since sending a list of ascbs is a superset of sending a single + * ascb, this function exists to generalize this. More specifically, + * when sending a list of those, we want to do only a _single_ + * memcpy() at swap head, as opposed to for each ascb sent (in the + * case of sending them one by one). That is, we want to minimize the + * ratio of memcpy() operations to the number of ascbs sent. The same + * logic applies to asd_post_ascb_list(). + */ +int asd_post_escb_list(struct asd_ha_struct *asd_ha, struct asd_ascb *ascb, + int num) +{ + unsigned long flags; + + spin_lock_irqsave(&asd_ha->seq.pend_q_lock, flags); + asd_swap_head_scb(asd_ha, ascb); + asd_ha->seq.scbpro += num; + asd_write_reg_dword(asd_ha, SCBPRO, (u32)asd_ha->seq.scbpro); + spin_unlock_irqrestore(&asd_ha->seq.pend_q_lock, flags); + + return 0; +} + +/* ---------- LED ---------- */ + +/** + * asd_turn_led -- turn on/off an LED + * @asd_ha: pointer to host adapter structure + * @phy_id: the PHY id whose LED we want to manupulate + * @op: 1 to turn on, 0 to turn off + */ +void asd_turn_led(struct asd_ha_struct *asd_ha, int phy_id, int op) +{ + if (phy_id < ASD_MAX_PHYS) { + u32 v = asd_read_reg_dword(asd_ha, LmCONTROL(phy_id)); + if (op) + v |= LEDPOL; + else + v &= ~LEDPOL; + asd_write_reg_dword(asd_ha, LmCONTROL(phy_id), v); + } +} + +/** + * asd_control_led -- enable/disable an LED on the board + * @asd_ha: pointer to host adapter structure + * @phy_id: integer, the phy id + * @op: integer, 1 to enable, 0 to disable the LED + * + * First we output enable the LED, then we set the source + * to be an external module. + */ +void asd_control_led(struct asd_ha_struct *asd_ha, int phy_id, int op) +{ + if (phy_id < ASD_MAX_PHYS) { + u32 v; + + v = asd_read_reg_dword(asd_ha, GPIOOER); + if (op) + v |= (1 << phy_id); + else + v &= ~(1 << phy_id); + asd_write_reg_dword(asd_ha, GPIOOER, v); + + v = asd_read_reg_dword(asd_ha, GPIOCNFGR); + if (op) + v |= (1 << phy_id); + else + v &= ~(1 << phy_id); + asd_write_reg_dword(asd_ha, GPIOCNFGR, v); + } +} + +/* ---------- PHY enable ---------- */ + +static int asd_enable_phy(struct asd_ha_struct *asd_ha, int phy_id) +{ + struct asd_phy *phy = &asd_ha->phys[phy_id]; + + asd_write_reg_byte(asd_ha, LmSEQ_OOB_REG(phy_id, INT_ENABLE_2), 0); + asd_write_reg_byte(asd_ha, LmSEQ_OOB_REG(phy_id, HOT_PLUG_DELAY), + HOTPLUG_DELAY_TIMEOUT); + + /* Get defaults from manuf. sector */ + /* XXX we need defaults for those in case MS is broken. */ + asd_write_reg_byte(asd_ha, LmSEQ_OOB_REG(phy_id, PHY_CONTROL_0), + phy->phy_desc->phy_control_0); + asd_write_reg_byte(asd_ha, LmSEQ_OOB_REG(phy_id, PHY_CONTROL_1), + phy->phy_desc->phy_control_1); + asd_write_reg_byte(asd_ha, LmSEQ_OOB_REG(phy_id, PHY_CONTROL_2), + phy->phy_desc->phy_control_2); + asd_write_reg_byte(asd_ha, LmSEQ_OOB_REG(phy_id, PHY_CONTROL_3), + phy->phy_desc->phy_control_3); + + asd_write_reg_dword(asd_ha, LmSEQ_TEN_MS_COMINIT_TIMEOUT(phy_id), + ASD_COMINIT_TIMEOUT); + + asd_write_reg_addr(asd_ha, LmSEQ_TX_ID_ADDR_FRAME(phy_id), + phy->id_frm_tok->dma_handle); + + asd_control_led(asd_ha, phy_id, 1); + + return 0; +} + +int asd_enable_phys(struct asd_ha_struct *asd_ha, const u8 phy_mask) +{ + u8 phy_m; + u8 i; + int num = 0, k; + struct asd_ascb *ascb; + struct asd_ascb *ascb_list; + + if (!phy_mask) { + asd_printk("%s called with phy_mask of 0!?\n", __func__); + return 0; + } + + for_each_phy(phy_mask, phy_m, i) { + num++; + asd_enable_phy(asd_ha, i); + } + + k = num; + ascb_list = asd_ascb_alloc_list(asd_ha, &k, GFP_KERNEL); + if (!ascb_list) { + asd_printk("no memory for control phy ascb list\n"); + return -ENOMEM; + } + num -= k; + + ascb = ascb_list; + for_each_phy(phy_mask, phy_m, i) { + asd_build_control_phy(ascb, i, ENABLE_PHY); + ascb = list_entry(ascb->list.next, struct asd_ascb, list); + } + ASD_DPRINTK("posting %d control phy scbs\n", num); + k = asd_post_ascb_list(asd_ha, ascb_list, num); + if (k) + asd_ascb_free_list(ascb_list); + + return k; +}