diff options
author | Mark Haverkamp <markh@osdl.org> | 2005-10-24 10:52:22 -0700 |
---|---|---|
committer | James Bottomley <jejb@mulgrave.(none)> | 2005-10-28 11:41:53 -0500 |
commit | 8e0c5ebde82b08f6d996e11983890fc4cc085fab (patch) | |
tree | 3ba38ff8e7b9203b47d038c215d9c7d623c250ba /drivers/scsi/aacraid/rkt.c | |
parent | 38a9a621aba953ddb8051547e98c10ec3c741312 (diff) | |
download | linux-8e0c5ebde82b08f6d996e11983890fc4cc085fab.tar.bz2 |
[SCSI] aacraid: Newer adapter communication iterface support
Received from Mark Salyzyn.
This patch adds the 'new comm' interface, which modern AAC based
adapters that are less than a year old support in the name of much
improved performance. These modern adapters support both the legacy and
the 'new comm' interfaces.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Diffstat (limited to 'drivers/scsi/aacraid/rkt.c')
-rw-r--r-- | drivers/scsi/aacraid/rkt.c | 172 |
1 files changed, 110 insertions, 62 deletions
diff --git a/drivers/scsi/aacraid/rkt.c b/drivers/scsi/aacraid/rkt.c index 557287a0b80b..fc4c73c2a6a9 100644 --- a/drivers/scsi/aacraid/rkt.c +++ b/drivers/scsi/aacraid/rkt.c @@ -49,40 +49,57 @@ static irqreturn_t aac_rkt_intr(int irq, void *dev_id, struct pt_regs *regs) { struct aac_dev *dev = dev_id; - unsigned long bellbits; - u8 intstat, mask; - intstat = rkt_readb(dev, MUnit.OISR); - /* - * Read mask and invert because drawbridge is reversed. - * This allows us to only service interrupts that have - * been enabled. - */ - mask = ~(dev->OIMR); - /* Check to see if this is our interrupt. If it isn't just return */ - if (intstat & mask) - { - bellbits = rkt_readl(dev, OutboundDoorbellReg); - if (bellbits & DoorBellPrintfReady) { - aac_printf(dev, rkt_readl(dev, IndexRegs.Mailbox[5])); - rkt_writel(dev, MUnit.ODR,DoorBellPrintfReady); - rkt_writel(dev, InboundDoorbellReg,DoorBellPrintfDone); - } - else if (bellbits & DoorBellAdapterNormCmdReady) { - rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormCmdReady); - aac_command_normal(&dev->queues->queue[HostNormCmdQueue]); - } - else if (bellbits & DoorBellAdapterNormRespReady) { - aac_response_normal(&dev->queues->queue[HostNormRespQueue]); - rkt_writel(dev, MUnit.ODR,DoorBellAdapterNormRespReady); - } - else if (bellbits & DoorBellAdapterNormCmdNotFull) { - rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormCmdNotFull); + + if (dev->new_comm_interface) { + u32 Index = rkt_readl(dev, MUnit.OutboundQueue); + if (Index == 0xFFFFFFFFL) + Index = rkt_readl(dev, MUnit.OutboundQueue); + if (Index != 0xFFFFFFFFL) { + do { + if (aac_intr_normal(dev, Index)) { + rkt_writel(dev, MUnit.OutboundQueue, Index); + rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormRespReady); + } + Index = rkt_readl(dev, MUnit.OutboundQueue); + } while (Index != 0xFFFFFFFFL); + return IRQ_HANDLED; } - else if (bellbits & DoorBellAdapterNormRespNotFull) { - rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormCmdNotFull); - rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormRespNotFull); + } else { + unsigned long bellbits; + u8 intstat; + intstat = rkt_readb(dev, MUnit.OISR); + /* + * Read mask and invert because drawbridge is reversed. + * This allows us to only service interrupts that have + * been enabled. + * Check to see if this is our interrupt. If it isn't just return + */ + if (intstat & ~(dev->OIMR)) + { + bellbits = rkt_readl(dev, OutboundDoorbellReg); + if (bellbits & DoorBellPrintfReady) { + aac_printf(dev, rkt_readl (dev, IndexRegs.Mailbox[5])); + rkt_writel(dev, MUnit.ODR,DoorBellPrintfReady); + rkt_writel(dev, InboundDoorbellReg,DoorBellPrintfDone); + } + else if (bellbits & DoorBellAdapterNormCmdReady) { + rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormCmdReady); + aac_command_normal(&dev->queues->queue[HostNormCmdQueue]); +// rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormCmdReady); + } + else if (bellbits & DoorBellAdapterNormRespReady) { + rkt_writel(dev, MUnit.ODR,DoorBellAdapterNormRespReady); + aac_response_normal(&dev->queues->queue[HostNormRespQueue]); + } + else if (bellbits & DoorBellAdapterNormCmdNotFull) { + rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormCmdNotFull); + } + else if (bellbits & DoorBellAdapterNormRespNotFull) { + rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormCmdNotFull); + rkt_writel(dev, MUnit.ODR, DoorBellAdapterNormRespNotFull); + } + return IRQ_HANDLED; } - return IRQ_HANDLED; } return IRQ_NONE; } @@ -173,7 +190,10 @@ static int rkt_sync_cmd(struct aac_dev *dev, u32 command, /* * Restore interrupt mask even though we timed out */ - rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xfb); + if (dev->new_comm_interface) + rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xf7); + else + rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xfb); return -ETIMEDOUT; } /* @@ -196,7 +216,10 @@ static int rkt_sync_cmd(struct aac_dev *dev, u32 command, /* * Restore interrupt mask */ - rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xfb); + if (dev->new_comm_interface) + rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xf7); + else + rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xfb); return 0; } @@ -268,15 +291,6 @@ static void aac_rkt_start_adapter(struct aac_dev *dev) init = dev->init; init->HostElapsedSeconds = cpu_to_le32(get_seconds()); - /* - * First clear out all interrupts. Then enable the one's that we - * can handle. - */ - rkt_writeb(dev, MUnit.OIMR, 0xff); - rkt_writel(dev, MUnit.ODR, 0xffffffff); -// rkt_writeb(dev, MUnit.OIMR, ~(u8)OUTBOUND_DOORBELL_INTERRUPT_MASK); - rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xfb); - // We can only use a 32 bit address here rkt_sync_cmd(dev, INIT_STRUCT_BASE_ADDRESS, (u32)(ulong)dev->init_pa, 0, 0, 0, 0, 0, NULL, NULL, NULL, NULL, NULL); @@ -350,6 +364,39 @@ static int aac_rkt_check_health(struct aac_dev *dev) } /** + * aac_rkt_send + * @fib: fib to issue + * + * Will send a fib, returning 0 if successful. + */ +static int aac_rkt_send(struct fib * fib) +{ + u64 addr = fib->hw_fib_pa; + struct aac_dev *dev = fib->dev; + volatile void __iomem *device = dev->regs.rkt; + u32 Index; + + dprintk((KERN_DEBUG "%p->aac_rkt_send(%p->%llx)\n", dev, fib, addr)); + Index = rkt_readl(dev, MUnit.InboundQueue); + if (Index == 0xFFFFFFFFL) + Index = rkt_readl(dev, MUnit.InboundQueue); + dprintk((KERN_DEBUG "Index = 0x%x\n", Index)); + if (Index == 0xFFFFFFFFL) + return Index; + device += Index; + dprintk((KERN_DEBUG "entry = %x %x %u\n", (u32)(addr & 0xffffffff), + (u32)(addr >> 32), (u32)le16_to_cpu(fib->hw_fib->header.Size))); + writel((u32)(addr & 0xffffffff), device); + device += sizeof(u32); + writel((u32)(addr >> 32), device); + device += sizeof(u32); + writel(le16_to_cpu(fib->hw_fib->header.Size), device); + rkt_writel(dev, MUnit.InboundQueue, Index); + dprintk((KERN_DEBUG "aac_rkt_send - return 0\n")); + return 0; +} + +/** * aac_rkt_init - initialize an i960 based AAC card * @dev: device to configure * @@ -369,13 +416,8 @@ int aac_rkt_init(struct aac_dev *dev) name = dev->name; /* - * Map in the registers from the adapter. + * Check to see if the board panic'd while booting. */ - if((dev->regs.rkt = ioremap((unsigned long)dev->scsi_host_ptr->base, 8192))==NULL) - { - printk(KERN_WARNING "aacraid: unable to map i960.\n" ); - goto error_iounmap; - } /* * Check to see if the board failed any self tests. */ @@ -426,6 +468,7 @@ int aac_rkt_init(struct aac_dev *dev) dev->a_ops.adapter_notify = aac_rkt_notify_adapter; dev->a_ops.adapter_sync_cmd = rkt_sync_cmd; dev->a_ops.adapter_check_health = aac_rkt_check_health; + dev->a_ops.adapter_send = aac_rkt_send; /* * First clear out all interrupts. Then enable the one's that we @@ -437,15 +480,24 @@ int aac_rkt_init(struct aac_dev *dev) if (aac_init_adapter(dev) == NULL) goto error_irq; - /* - * Start any kernel threads needed - */ - dev->thread_pid = kernel_thread((int (*)(void *))aac_command_thread, dev, 0); - if(dev->thread_pid < 0) - { - printk(KERN_ERR "aacraid: Unable to create rkt thread.\n"); - goto error_kfree; - } + if (dev->new_comm_interface) { + /* + * FIB Setup has already been done, but we can minimize the + * damage by at least ensuring the OS never issues more + * commands than we can handle. The Rocket adapters currently + * can only handle 246 commands and 8 AIFs at the same time, + * and in fact do notify us accordingly if we negotiate the + * FIB size. The problem that causes us to add this check is + * to ensure that we do not overdo it with the adapter when a + * hard coded FIB override is being utilized. This special + * case warrants this half baked, but convenient, check here. + */ + if (dev->scsi_host_ptr->can_queue > (246 - AAC_NUM_MGT_FIB)) { + dev->init->MaxIoCommands = cpu_to_le32(246); + dev->scsi_host_ptr->can_queue = 246 - AAC_NUM_MGT_FIB; + } + rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xf7); + } /* * Tell the adapter that all is configured, and it can start * accepting requests @@ -453,15 +505,11 @@ int aac_rkt_init(struct aac_dev *dev) aac_rkt_start_adapter(dev); return 0; -error_kfree: - kfree(dev->queues); - error_irq: rkt_writeb(dev, MUnit.OIMR, dev->OIMR = 0xff); free_irq(dev->scsi_host_ptr->irq, (void *)dev); error_iounmap: - iounmap(dev->regs.rkt); return -1; } |