From c8f7b073e0e81499474a84ee2a90f77f7805c7f8 Mon Sep 17 00:00:00 2001 From: Mark Haverkamp Date: Thu, 3 Aug 2006 08:02:24 -0700 Subject: [SCSI] aacraid: interruptible ioctl Received from Mark Salyzyn This patch allows the FSACTL_SEND_LARGE_FIB, FSACTL_SENDFIB and FSACTL_SEND_RAW_SRB ioctl calls into the aacraid driver to be interruptible. Only necessary if the adapter and/or the management software has gone into some sort of misbehavior and the system is being rebooted, thus permitting the user management software applications to be killed relatively cleanly. The FIB queue resource is held out of the free queue until the adapter finally, if ever, completes the command. Signed-off-by: Mark Haverkamp Signed-off-by: James Bottomley --- drivers/scsi/aacraid/commctrl.c | 23 ++++++++++++++++++----- drivers/scsi/aacraid/commsup.c | 15 ++++++++++++--- drivers/scsi/aacraid/dpcsup.c | 10 ++++++++-- 3 files changed, 38 insertions(+), 10 deletions(-) (limited to 'drivers/scsi/aacraid') diff --git a/drivers/scsi/aacraid/commctrl.c b/drivers/scsi/aacraid/commctrl.c index 255421de9d1..14d7aa9b7df 100644 --- a/drivers/scsi/aacraid/commctrl.c +++ b/drivers/scsi/aacraid/commctrl.c @@ -38,7 +38,7 @@ #include #include #include -#include +#include /* ssleep prototype */ #include #include #include @@ -140,7 +140,8 @@ cleanup: fibptr->hw_fib_pa = hw_fib_pa; fibptr->hw_fib = hw_fib; } - aac_fib_free(fibptr); + if (retval != -EINTR) + aac_fib_free(fibptr); return retval; } @@ -621,7 +622,13 @@ static int aac_send_raw_srb(struct aac_dev* dev, void __user * arg) actual_fibsize = sizeof (struct aac_srb) + (((user_srbcmd->sg.count & 0xff) - 1) * sizeof (struct sgentry)); if(actual_fibsize != fibsize){ // User made a mistake - should not continue - dprintk((KERN_DEBUG"aacraid: Bad Size specified in Raw SRB command\n")); + dprintk((KERN_DEBUG"aacraid: Bad Size specified in " + "Raw SRB command calculated fibsize=%d " + "user_srbcmd->sg.count=%d aac_srb=%d sgentry=%d " + "issued fibsize=%d\n", + actual_fibsize, user_srbcmd->sg.count, + sizeof(struct aac_srb), sizeof(struct sgentry), + fibsize)); rcode = -EINVAL; goto cleanup; } @@ -663,6 +670,10 @@ static int aac_send_raw_srb(struct aac_dev* dev, void __user * arg) psg->count = cpu_to_le32(sg_indx+1); status = aac_fib_send(ScsiPortCommand, srbfib, actual_fibsize, FsaNormal, 1, 1, NULL, NULL); } + if (status == -EINTR) { + rcode = -EINTR; + goto cleanup; + } if (status != 0){ dprintk((KERN_DEBUG"aacraid: Could not send raw srb fib to hba\n")); @@ -696,8 +707,10 @@ cleanup: for(i=0; i <= sg_indx; i++){ kfree(sg_list[i]); } - aac_fib_complete(srbfib); - aac_fib_free(srbfib); + if (rcode != -EINTR) { + aac_fib_complete(srbfib); + aac_fib_free(srbfib); + } return rcode; } diff --git a/drivers/scsi/aacraid/commsup.c b/drivers/scsi/aacraid/commsup.c index 3f27419c66a..c67da132113 100644 --- a/drivers/scsi/aacraid/commsup.c +++ b/drivers/scsi/aacraid/commsup.c @@ -464,6 +464,8 @@ int aac_fib_send(u16 command, struct fib *fibptr, unsigned long size, dprintk((KERN_DEBUG " hw_fib pa being sent=%lx\n",(ulong)fibptr->hw_fib_pa)); dprintk((KERN_DEBUG " fib being sent=%p\n",fibptr)); + if (!dev->queues) + return -ENODEV; q = &dev->queues->queue[AdapNormCmdQueue]; if(wait) @@ -527,8 +529,15 @@ int aac_fib_send(u16 command, struct fib *fibptr, unsigned long size, } udelay(5); } - } else - down(&fibptr->event_wait); + } else if (down_interruptible(&fibptr->event_wait)) { + spin_lock_irqsave(&fibptr->event_lock, flags); + if (fibptr->done == 0) { + fibptr->done = 2; /* Tell interrupt we aborted */ + spin_unlock_irqrestore(&fibptr->event_lock, flags); + return -EINTR; + } + spin_unlock_irqrestore(&fibptr->event_lock, flags); + } BUG_ON(fibptr->done == 0); if((fibptr->flags & FIB_CONTEXT_FLAG_TIMED_OUT)){ @@ -795,7 +804,7 @@ static void aac_handle_aif(struct aac_dev * dev, struct fib * fibptr) /* Sniff for container changes */ - if (!dev) + if (!dev || !dev->fsa_dev) return; container = (u32)-1; diff --git a/drivers/scsi/aacraid/dpcsup.c b/drivers/scsi/aacraid/dpcsup.c index b2a5c7262f3..8335f07b772 100644 --- a/drivers/scsi/aacraid/dpcsup.c +++ b/drivers/scsi/aacraid/dpcsup.c @@ -124,10 +124,15 @@ unsigned int aac_response_normal(struct aac_queue * q) } else { unsigned long flagv; spin_lock_irqsave(&fib->event_lock, flagv); - fib->done = 1; + if (!fib->done) + fib->done = 1; up(&fib->event_wait); spin_unlock_irqrestore(&fib->event_lock, flagv); FIB_COUNTER_INCREMENT(aac_config.NormalRecved); + if (fib->done == 2) { + aac_fib_complete(fib); + aac_fib_free(fib); + } } consumed++; spin_lock_irqsave(q->lock, flags); @@ -316,7 +321,8 @@ unsigned int aac_intr_normal(struct aac_dev * dev, u32 Index) unsigned long flagv; dprintk((KERN_INFO "event_wait up\n")); spin_lock_irqsave(&fib->event_lock, flagv); - fib->done = 1; + if (!fib->done) + fib->done = 1; up(&fib->event_wait); spin_unlock_irqrestore(&fib->event_lock, flagv); FIB_COUNTER_INCREMENT(aac_config.NormalRecved); -- cgit v1.2.3-70-g09d2 From 8c23cd7457151fc8ace79ec700a8aeaa9fc5b3d9 Mon Sep 17 00:00:00 2001 From: Mark Haverkamp Date: Tue, 8 Aug 2006 08:52:14 -0700 Subject: [SCSI] aacraid: Restart adapter on firmware assert (Update 2) Received from Mark Salyzyn If the adapter should be in a blinkled (Firmware Assert) state when the driver loads, we will perform a warm restart of the Adapter Firmware to see if we can rescue the adapter. Possible causes of a blinkled can occur on some early release motherboard BIOSes, transitory PCI bus problems on embedded systems or non-x86 based architectures, transitory startup failures of early release drives or transitory hardware failures; some of which can bite the adapter later at runtime. Future enhancements will include recovery during runtime. Fixed extra whitespace space issue. Signed-off-by: Mark Haverkamp Signed-off-by: James Bottomley --- drivers/scsi/aacraid/aacraid.h | 1 + drivers/scsi/aacraid/rkt.c | 29 ++++++++++++++++++++++------- drivers/scsi/aacraid/rx.c | 29 ++++++++++++++++++++++------- 3 files changed, 45 insertions(+), 14 deletions(-) (limited to 'drivers/scsi/aacraid') diff --git a/drivers/scsi/aacraid/aacraid.h b/drivers/scsi/aacraid/aacraid.h index d0eecd4bec8..05f80982efa 100644 --- a/drivers/scsi/aacraid/aacraid.h +++ b/drivers/scsi/aacraid/aacraid.h @@ -1670,6 +1670,7 @@ extern struct aac_common aac_config; #define RCV_TEMP_READINGS 0x00000025 #define GET_COMM_PREFERRED_SETTINGS 0x00000026 #define IOP_RESET 0x00001000 +#define IOP_RESET_ALWAYS 0x00001001 #define RE_INIT_ADAPTER 0x000000ee /* diff --git a/drivers/scsi/aacraid/rkt.c b/drivers/scsi/aacraid/rkt.c index 458ea897fd7..f850c3a7cce 100644 --- a/drivers/scsi/aacraid/rkt.c +++ b/drivers/scsi/aacraid/rkt.c @@ -395,6 +395,25 @@ static int aac_rkt_send(struct fib * fib) return 0; } +static int aac_rkt_restart_adapter(struct aac_dev *dev) +{ + u32 var; + + printk(KERN_ERR "%s%d: adapter kernel panic'd.\n", + dev->name, dev->id); + + if (aac_rkt_check_health(dev) <= 0) + return 1; + if (rkt_sync_cmd(dev, IOP_RESET, 0, 0, 0, 0, 0, 0, + &var, NULL, NULL, NULL, NULL)) + return 1; + if (var != 0x00000001) + return 1; + if (rkt_readl(dev, MUnit.OMRx[0]) & KERNEL_PANIC) + return 1; + return 0; +} + /** * aac_rkt_init - initialize an i960 based AAC card * @dev: device to configure @@ -417,6 +436,9 @@ int aac_rkt_init(struct aac_dev *dev) /* * Check to see if the board panic'd while booting. */ + if (rkt_readl(dev, MUnit.OMRx[0]) & KERNEL_PANIC) + if (aac_rkt_restart_adapter(dev)) + goto error_iounmap; /* * Check to see if the board failed any self tests. */ @@ -431,13 +453,6 @@ int aac_rkt_init(struct aac_dev *dev) printk(KERN_ERR "%s%d: adapter monitor panic.\n", dev->name, instance); goto error_iounmap; } - /* - * Check to see if the board panic'd while booting. - */ - if (rkt_readl(dev, MUnit.OMRx[0]) & KERNEL_PANIC) { - printk(KERN_ERR "%s%d: adapter kernel panic'd.\n", dev->name, instance); - goto error_iounmap; - } start = jiffies; /* * Wait for the adapter to be up and running. Wait up to 3 minutes diff --git a/drivers/scsi/aacraid/rx.c b/drivers/scsi/aacraid/rx.c index 035018db69b..c715c4b2442 100644 --- a/drivers/scsi/aacraid/rx.c +++ b/drivers/scsi/aacraid/rx.c @@ -394,6 +394,25 @@ static int aac_rx_send(struct fib * fib) return 0; } +static int aac_rx_restart_adapter(struct aac_dev *dev) +{ + u32 var; + + printk(KERN_ERR "%s%d: adapter kernel panic'd.\n", + dev->name, dev->id); + + if (aac_rx_check_health(dev) <= 0) + return 1; + if (rx_sync_cmd(dev, IOP_RESET, 0, 0, 0, 0, 0, 0, + &var, NULL, NULL, NULL, NULL)) + return 1; + if (var != 0x00000001) + return 1; + if (rx_readl(dev, MUnit.OMRx[0]) & KERNEL_PANIC) + return 1; + return 0; +} + /** * aac_rx_init - initialize an i960 based AAC card * @dev: device to configure @@ -416,6 +435,9 @@ int aac_rx_init(struct aac_dev *dev) /* * Check to see if the board panic'd while booting. */ + if (rx_readl(dev, MUnit.OMRx[0]) & KERNEL_PANIC) + if (aac_rx_restart_adapter(dev)) + goto error_iounmap; /* * Check to see if the board failed any self tests. */ @@ -423,13 +445,6 @@ int aac_rx_init(struct aac_dev *dev) printk(KERN_ERR "%s%d: adapter self-test failed.\n", dev->name, instance); goto error_iounmap; } - /* - * Check to see if the board panic'd while booting. - */ - if (rx_readl(dev, MUnit.OMRx[0]) & KERNEL_PANIC) { - printk(KERN_ERR "%s%d: adapter kernel panic.\n", dev->name, instance); - goto error_iounmap; - } /* * Check to see if the monitor panic'd while booting. */ -- cgit v1.2.3-70-g09d2 From 90ee346651524eb275405d410f5d3bb6765a2d93 Mon Sep 17 00:00:00 2001 From: Mark Haverkamp Date: Thu, 3 Aug 2006 08:03:07 -0700 Subject: [SCSI] aacraid: Check for unlikely errors Received from Mark Salyzyn The enclosed patch cleans up some code fragments, adds some paranoia (unproven causes of potential driver failures). Signed-off-by: Mark Haverkamp Signed-off-by: James Bottomley --- drivers/scsi/aacraid/aachba.c | 15 +++++++++++++-- drivers/scsi/aacraid/comminit.c | 2 +- drivers/scsi/aacraid/linit.c | 4 ++++ 3 files changed, 18 insertions(+), 3 deletions(-) (limited to 'drivers/scsi/aacraid') diff --git a/drivers/scsi/aacraid/aachba.c b/drivers/scsi/aacraid/aachba.c index 83b5c7d085f..699351c15cc 100644 --- a/drivers/scsi/aacraid/aachba.c +++ b/drivers/scsi/aacraid/aachba.c @@ -489,6 +489,8 @@ int aac_probe_container(struct aac_dev *dev, int cid) unsigned instance; fsa_dev_ptr = dev->fsa_dev; + if (!fsa_dev_ptr) + return -ENOMEM; instance = dev->scsi_host_ptr->unique_id; if (!(fibptr = aac_fib_alloc(dev))) @@ -1392,6 +1394,7 @@ static int aac_synchronize(struct scsi_cmnd *scsicmd, int cid) struct scsi_cmnd *cmd; struct scsi_device *sdev = scsicmd->device; int active = 0; + struct aac_dev *aac; unsigned long flags; /* @@ -1413,11 +1416,11 @@ static int aac_synchronize(struct scsi_cmnd *scsicmd, int cid) if (active) return SCSI_MLQUEUE_DEVICE_BUSY; + aac = (struct aac_dev *)scsicmd->device->host->hostdata; /* * Allocate and initialize a Fib */ - if (!(cmd_fibcontext = - aac_fib_alloc((struct aac_dev *)scsicmd->device->host->hostdata))) + if (!(cmd_fibcontext = aac_fib_alloc(aac))) return SCSI_MLQUEUE_HOST_BUSY; aac_fib_init(cmd_fibcontext); @@ -1470,6 +1473,8 @@ int aac_scsi_cmd(struct scsi_cmnd * scsicmd) struct aac_dev *dev = (struct aac_dev *)host->hostdata; struct fsa_dev_info *fsa_dev_ptr = dev->fsa_dev; + if (fsa_dev_ptr == NULL) + return -1; /* * If the bus, id or lun is out of range, return fail * Test does not apply to ID 16, the pseudo id for the controller @@ -1782,6 +1787,8 @@ static int query_disk(struct aac_dev *dev, void __user *arg) struct fsa_dev_info *fsa_dev_ptr; fsa_dev_ptr = dev->fsa_dev; + if (!fsa_dev_ptr) + return -ENODEV; if (copy_from_user(&qd, arg, sizeof (struct aac_query_disk))) return -EFAULT; if (qd.cnum == -1) @@ -1843,6 +1850,10 @@ static int delete_disk(struct aac_dev *dev, void __user *arg) struct fsa_dev_info *fsa_dev_ptr; fsa_dev_ptr = dev->fsa_dev; + if (!fsa_dev_ptr) + return -ENODEV; + if (!fsa_dev_ptr) + return -ENODEV; if (copy_from_user(&dd, arg, sizeof (struct aac_delete_disk))) return -EFAULT; diff --git a/drivers/scsi/aacraid/comminit.c b/drivers/scsi/aacraid/comminit.c index 1cd3584ba7f..87a95509676 100644 --- a/drivers/scsi/aacraid/comminit.c +++ b/drivers/scsi/aacraid/comminit.c @@ -180,7 +180,7 @@ int aac_send_shutdown(struct aac_dev * dev) -2 /* Timeout silently */, 1, NULL, NULL); - if (status == 0) + if (status >= 0) aac_fib_complete(fibctx); aac_fib_free(fibctx); return status; diff --git a/drivers/scsi/aacraid/linit.c b/drivers/scsi/aacraid/linit.c index e42a479ce64..9d8b550a91c 100644 --- a/drivers/scsi/aacraid/linit.c +++ b/drivers/scsi/aacraid/linit.c @@ -1013,6 +1013,10 @@ static void __devexit aac_remove_one(struct pci_dev *pdev) list_del(&aac->entry); scsi_host_put(shost); pci_disable_device(pdev); + if (list_empty(&aac_devices)) { + unregister_chrdev(aac_cfg_major, "aac"); + aac_cfg_major = -1; + } } static struct pci_driver aac_pci_driver = { -- cgit v1.2.3-70-g09d2 From 8c867b257d159ca04602d7087fa29f846785f9ea Mon Sep 17 00:00:00 2001 From: Mark Haverkamp Date: Thu, 3 Aug 2006 08:03:30 -0700 Subject: [SCSI] aacraid: Reset adapter in recovery timeout Received from Mark Salyzyn If the adapter is in blinkled (Firmware Assert) when error recovery timeout actions have been triggered, perform an adapter warm reset and restart the initialization. Signed-off-by: Mark Haverkamp Signed-off-by: James Bottomley --- drivers/scsi/aacraid/aachba.c | 39 ++++-- drivers/scsi/aacraid/aacraid.h | 4 +- drivers/scsi/aacraid/commctrl.c | 2 +- drivers/scsi/aacraid/commsup.c | 258 ++++++++++++++++++++++++++++++++++++++++ drivers/scsi/aacraid/linit.c | 14 +-- 5 files changed, 296 insertions(+), 21 deletions(-) (limited to 'drivers/scsi/aacraid') diff --git a/drivers/scsi/aacraid/aachba.c b/drivers/scsi/aacraid/aachba.c index 699351c15cc..37c55ddce21 100644 --- a/drivers/scsi/aacraid/aachba.c +++ b/drivers/scsi/aacraid/aachba.c @@ -175,7 +175,7 @@ MODULE_PARM_DESC(acbsize, "Request a specific adapter control block (FIB) size. * * Query config status, and commit the configuration if needed. */ -int aac_get_config_status(struct aac_dev *dev) +int aac_get_config_status(struct aac_dev *dev, int commit_flag) { int status = 0; struct fib * fibptr; @@ -219,7 +219,7 @@ int aac_get_config_status(struct aac_dev *dev) aac_fib_complete(fibptr); /* Send a CT_COMMIT_CONFIG to enable discovery of devices */ if (status >= 0) { - if (commit == 1) { + if ((commit == 1) || commit_flag) { struct aac_commit_config * dinfo; aac_fib_init(fibptr); dinfo = (struct aac_commit_config *) fib_data(fibptr); @@ -784,8 +784,9 @@ int aac_get_adapter_info(struct aac_dev* dev) dev->maximum_num_channels = le32_to_cpu(bus_info->BusCount); } - tmp = le32_to_cpu(dev->adapter_info.kernelrev); - printk(KERN_INFO "%s%d: kernel %d.%d-%d[%d] %.*s\n", + if (!dev->in_reset) { + tmp = le32_to_cpu(dev->adapter_info.kernelrev); + printk(KERN_INFO "%s%d: kernel %d.%d-%d[%d] %.*s\n", dev->name, dev->id, tmp>>24, @@ -794,20 +795,21 @@ int aac_get_adapter_info(struct aac_dev* dev) le32_to_cpu(dev->adapter_info.kernelbuild), (int)sizeof(dev->supplement_adapter_info.BuildDate), dev->supplement_adapter_info.BuildDate); - tmp = le32_to_cpu(dev->adapter_info.monitorrev); - printk(KERN_INFO "%s%d: monitor %d.%d-%d[%d]\n", + tmp = le32_to_cpu(dev->adapter_info.monitorrev); + printk(KERN_INFO "%s%d: monitor %d.%d-%d[%d]\n", dev->name, dev->id, tmp>>24,(tmp>>16)&0xff,tmp&0xff, le32_to_cpu(dev->adapter_info.monitorbuild)); - tmp = le32_to_cpu(dev->adapter_info.biosrev); - printk(KERN_INFO "%s%d: bios %d.%d-%d[%d]\n", + tmp = le32_to_cpu(dev->adapter_info.biosrev); + printk(KERN_INFO "%s%d: bios %d.%d-%d[%d]\n", dev->name, dev->id, tmp>>24,(tmp>>16)&0xff,tmp&0xff, le32_to_cpu(dev->adapter_info.biosbuild)); - if (le32_to_cpu(dev->adapter_info.serial[0]) != 0xBAD0) - printk(KERN_INFO "%s%d: serial %x\n", - dev->name, dev->id, - le32_to_cpu(dev->adapter_info.serial[0])); + if (le32_to_cpu(dev->adapter_info.serial[0]) != 0xBAD0) + printk(KERN_INFO "%s%d: serial %x\n", + dev->name, dev->id, + le32_to_cpu(dev->adapter_info.serial[0])); + } dev->nondasd_support = 0; dev->raid_scsi_mode = 0; @@ -1417,6 +1419,9 @@ static int aac_synchronize(struct scsi_cmnd *scsicmd, int cid) return SCSI_MLQUEUE_DEVICE_BUSY; aac = (struct aac_dev *)scsicmd->device->host->hostdata; + if (aac->in_reset) + return SCSI_MLQUEUE_HOST_BUSY; + /* * Allocate and initialize a Fib */ @@ -1504,6 +1509,8 @@ int aac_scsi_cmd(struct scsi_cmnd * scsicmd) case INQUIRY: case READ_CAPACITY: case TEST_UNIT_READY: + if (dev->in_reset) + return -1; spin_unlock_irq(host->host_lock); aac_probe_container(dev, cid); if ((fsa_dev_ptr[cid].valid & 1) == 0) @@ -1529,6 +1536,8 @@ int aac_scsi_cmd(struct scsi_cmnd * scsicmd) } } else { /* check for physical non-dasd devices */ if(dev->nondasd_support == 1){ + if (dev->in_reset) + return -1; return aac_send_srb_fib(scsicmd); } else { scsicmd->result = DID_NO_CONNECT << 16; @@ -1584,6 +1593,8 @@ int aac_scsi_cmd(struct scsi_cmnd * scsicmd) scsicmd->scsi_done(scsicmd); return 0; } + if (dev->in_reset) + return -1; setinqstr(dev, (void *) (inq_data.inqd_vid), fsa_dev_ptr[cid].type); inq_data.inqd_pdt = INQD_PDT_DA; /* Direct/random access device */ aac_internal_transfer(scsicmd, &inq_data, 0, sizeof(inq_data)); @@ -1739,6 +1750,8 @@ int aac_scsi_cmd(struct scsi_cmnd * scsicmd) case READ_10: case READ_12: case READ_16: + if (dev->in_reset) + return -1; /* * Hack to keep track of ordinal number of the device that * corresponds to a container. Needed to convert @@ -1757,6 +1770,8 @@ int aac_scsi_cmd(struct scsi_cmnd * scsicmd) case WRITE_10: case WRITE_12: case WRITE_16: + if (dev->in_reset) + return -1; return aac_write(scsicmd, cid); case SYNCHRONIZE_CACHE: diff --git a/drivers/scsi/aacraid/aacraid.h b/drivers/scsi/aacraid/aacraid.h index 05f80982efa..8924c183d9c 100644 --- a/drivers/scsi/aacraid/aacraid.h +++ b/drivers/scsi/aacraid/aacraid.h @@ -1029,6 +1029,7 @@ struct aac_dev init->InitStructRevision==cpu_to_le32(ADAPTER_INIT_STRUCT_REVISION_4) u8 raw_io_64; u8 printf_enabled; + u8 in_reset; }; #define aac_adapter_interrupt(dev) \ @@ -1789,7 +1790,7 @@ void aac_consumer_free(struct aac_dev * dev, struct aac_queue * q, u32 qnum); int aac_fib_complete(struct fib * context); #define fib_data(fibctx) ((void *)(fibctx)->hw_fib->data) struct aac_dev *aac_init_adapter(struct aac_dev *dev); -int aac_get_config_status(struct aac_dev *dev); +int aac_get_config_status(struct aac_dev *dev, int commit_flag); int aac_get_containers(struct aac_dev *dev); int aac_scsi_cmd(struct scsi_cmnd *cmd); int aac_dev_ioctl(struct aac_dev *dev, int cmd, void __user *arg); @@ -1800,6 +1801,7 @@ int aac_sa_init(struct aac_dev *dev); unsigned int aac_response_normal(struct aac_queue * q); unsigned int aac_command_normal(struct aac_queue * q); unsigned int aac_intr_normal(struct aac_dev * dev, u32 Index); +int aac_check_health(struct aac_dev * dev); int aac_command_thread(void *data); int aac_close_fib_context(struct aac_dev * dev, struct aac_fib_context *fibctx); int aac_fib_adapter_complete(struct fib * fibptr, unsigned short size); diff --git a/drivers/scsi/aacraid/commctrl.c b/drivers/scsi/aacraid/commctrl.c index 14d7aa9b7df..da1d3a9212f 100644 --- a/drivers/scsi/aacraid/commctrl.c +++ b/drivers/scsi/aacraid/commctrl.c @@ -298,7 +298,7 @@ return_fib: spin_unlock_irqrestore(&dev->fib_lock, flags); /* If someone killed the AIF aacraid thread, restart it */ status = !dev->aif_thread; - if (status && dev->queues && dev->fsa_dev) { + if (status && !dev->in_reset && dev->queues && dev->fsa_dev) { /* Be paranoid, be very paranoid! */ kthread_stop(dev->thread); ssleep(1); diff --git a/drivers/scsi/aacraid/commsup.c b/drivers/scsi/aacraid/commsup.c index c67da132113..53add53be0b 100644 --- a/drivers/scsi/aacraid/commsup.c +++ b/drivers/scsi/aacraid/commsup.c @@ -40,8 +40,10 @@ #include #include #include +#include #include #include +#include #include #include "aacraid.h" @@ -1054,6 +1056,262 @@ static void aac_handle_aif(struct aac_dev * dev, struct fib * fibptr) } +static int _aac_reset_adapter(struct aac_dev *aac) +{ + int index, quirks; + u32 ret; + int retval; + struct Scsi_Host *host; + struct scsi_device *dev; + struct scsi_cmnd *command; + struct scsi_cmnd *command_list; + + /* + * Assumptions: + * - host is locked. + * - in_reset is asserted, so no new i/o is getting to the + * card. + * - The card is dead. + */ + host = aac->scsi_host_ptr; + scsi_block_requests(host); + aac_adapter_disable_int(aac); + spin_unlock_irq(host->host_lock); + kthread_stop(aac->thread); + + /* + * If a positive health, means in a known DEAD PANIC + * state and the adapter could be reset to `try again'. + */ + retval = aac_adapter_check_health(aac); + if (retval == 0) + retval = aac_adapter_sync_cmd(aac, IOP_RESET_ALWAYS, + 0, 0, 0, 0, 0, 0, &ret, NULL, NULL, NULL, NULL); + if (retval) + retval = aac_adapter_sync_cmd(aac, IOP_RESET, + 0, 0, 0, 0, 0, 0, &ret, NULL, NULL, NULL, NULL); + + if (retval) + goto out; + if (ret != 0x00000001) { + retval = -ENODEV; + goto out; + } + + index = aac->cardtype; + + /* + * Re-initialize the adapter, first free resources, then carefully + * apply the initialization sequence to come back again. Only risk + * is a change in Firmware dropping cache, it is assumed the caller + * will ensure that i/o is queisced and the card is flushed in that + * case. + */ + aac_fib_map_free(aac); + aac->hw_fib_va = NULL; + aac->hw_fib_pa = 0; + pci_free_consistent(aac->pdev, aac->comm_size, aac->comm_addr, aac->comm_phys); + aac->comm_addr = NULL; + aac->comm_phys = 0; + kfree(aac->queues); + aac->queues = NULL; + free_irq(aac->pdev->irq, aac); + kfree(aac->fsa_dev); + aac->fsa_dev = NULL; + if (aac_get_driver_ident(index)->quirks & AAC_QUIRK_31BIT) { + if (((retval = pci_set_dma_mask(aac->pdev, DMA_32BIT_MASK))) || + ((retval = pci_set_consistent_dma_mask(aac->pdev, DMA_32BIT_MASK)))) + goto out; + } else { + if (((retval = pci_set_dma_mask(aac->pdev, 0x7FFFFFFFULL))) || + ((retval = pci_set_consistent_dma_mask(aac->pdev, 0x7FFFFFFFULL)))) + goto out; + } + if ((retval = (*(aac_get_driver_ident(index)->init))(aac))) + goto out; + if (aac_get_driver_ident(index)->quirks & AAC_QUIRK_31BIT) + if ((retval = pci_set_dma_mask(aac->pdev, DMA_32BIT_MASK))) + goto out; + aac->thread = kthread_run(aac_command_thread, aac, aac->name); + if (IS_ERR(aac->thread)) { + retval = PTR_ERR(aac->thread); + goto out; + } + (void)aac_get_adapter_info(aac); + quirks = aac_get_driver_ident(index)->quirks; + if ((quirks & AAC_QUIRK_34SG) && (host->sg_tablesize > 34)) { + host->sg_tablesize = 34; + host->max_sectors = (host->sg_tablesize * 8) + 112; + } + if ((quirks & AAC_QUIRK_17SG) && (host->sg_tablesize > 17)) { + host->sg_tablesize = 17; + host->max_sectors = (host->sg_tablesize * 8) + 112; + } + aac_get_config_status(aac, 1); + aac_get_containers(aac); + /* + * This is where the assumption that the Adapter is quiesced + * is important. + */ + command_list = NULL; + __shost_for_each_device(dev, host) { + unsigned long flags; + spin_lock_irqsave(&dev->list_lock, flags); + list_for_each_entry(command, &dev->cmd_list, list) + if (command->SCp.phase == AAC_OWNER_FIRMWARE) { + command->SCp.buffer = (struct scatterlist *)command_list; + command_list = command; + } + spin_unlock_irqrestore(&dev->list_lock, flags); + } + while ((command = command_list)) { + command_list = (struct scsi_cmnd *)command->SCp.buffer; + command->SCp.buffer = NULL; + command->result = DID_OK << 16 + | COMMAND_COMPLETE << 8 + | SAM_STAT_TASK_SET_FULL; + command->SCp.phase = AAC_OWNER_ERROR_HANDLER; + command->scsi_done(command); + } + retval = 0; + +out: + aac->in_reset = 0; + scsi_unblock_requests(host); + spin_lock_irq(host->host_lock); + return retval; +} + +int aac_check_health(struct aac_dev * aac) +{ + int BlinkLED; + unsigned long time_now, flagv = 0; + struct list_head * entry; + struct Scsi_Host * host; + + /* Extending the scope of fib_lock slightly to protect aac->in_reset */ + if (spin_trylock_irqsave(&aac->fib_lock, flagv) == 0) + return 0; + + if (aac->in_reset || !(BlinkLED = aac_adapter_check_health(aac))) { + spin_unlock_irqrestore(&aac->fib_lock, flagv); + return 0; /* OK */ + } + + aac->in_reset = 1; + + /* Fake up an AIF: + * aac_aifcmd.command = AifCmdEventNotify = 1 + * aac_aifcmd.seqnum = 0xFFFFFFFF + * aac_aifcmd.data[0] = AifEnExpEvent = 23 + * aac_aifcmd.data[1] = AifExeFirmwarePanic = 3 + * aac.aifcmd.data[2] = AifHighPriority = 3 + * aac.aifcmd.data[3] = BlinkLED + */ + + time_now = jiffies/HZ; + entry = aac->fib_list.next; + + /* + * For each Context that is on the + * fibctxList, make a copy of the + * fib, and then set the event to wake up the + * thread that is waiting for it. + */ + while (entry != &aac->fib_list) { + /* + * Extract the fibctx + */ + struct aac_fib_context *fibctx = list_entry(entry, struct aac_fib_context, next); + struct hw_fib * hw_fib; + struct fib * fib; + /* + * Check if the queue is getting + * backlogged + */ + if (fibctx->count > 20) { + /* + * It's *not* jiffies folks, + * but jiffies / HZ, so do not + * panic ... + */ + u32 time_last = fibctx->jiffies; + /* + * Has it been > 2 minutes + * since the last read off + * the queue? + */ + if ((time_now - time_last) > aif_timeout) { + entry = entry->next; + aac_close_fib_context(aac, fibctx); + continue; + } + } + /* + * Warning: no sleep allowed while + * holding spinlock + */ + hw_fib = kmalloc(sizeof(struct hw_fib), GFP_ATOMIC); + fib = kmalloc(sizeof(struct fib), GFP_ATOMIC); + if (fib && hw_fib) { + struct aac_aifcmd * aif; + + memset(hw_fib, 0, sizeof(struct hw_fib)); + memset(fib, 0, sizeof(struct fib)); + fib->hw_fib = hw_fib; + fib->dev = aac; + aac_fib_init(fib); + fib->type = FSAFS_NTC_FIB_CONTEXT; + fib->size = sizeof (struct fib); + fib->data = hw_fib->data; + aif = (struct aac_aifcmd *)hw_fib->data; + aif->command = cpu_to_le32(AifCmdEventNotify); + aif->seqnum = cpu_to_le32(0xFFFFFFFF); + aif->data[0] = cpu_to_le32(AifEnExpEvent); + aif->data[1] = cpu_to_le32(AifExeFirmwarePanic); + aif->data[2] = cpu_to_le32(AifHighPriority); + aif->data[3] = cpu_to_le32(BlinkLED); + + /* + * Put the FIB onto the + * fibctx's fibs + */ + list_add_tail(&fib->fiblink, &fibctx->fib_list); + fibctx->count++; + /* + * Set the event to wake up the + * thread that will waiting. + */ + up(&fibctx->wait_sem); + } else { + printk(KERN_WARNING "aifd: didn't allocate NewFib.\n"); + kfree(fib); + kfree(hw_fib); + } + entry = entry->next; + } + + spin_unlock_irqrestore(&aac->fib_lock, flagv); + + if (BlinkLED < 0) { + printk(KERN_ERR "%s: Host adapter dead %d\n", aac->name, BlinkLED); + goto out; + } + + printk(KERN_ERR "%s: Host adapter BLINK LED 0x%x\n", aac->name, BlinkLED); + + host = aac->scsi_host_ptr; + spin_lock_irqsave(host->host_lock, flagv); + BlinkLED = _aac_reset_adapter(aac); + spin_unlock_irqrestore(host->host_lock, flagv); + return BlinkLED; + +out: + aac->in_reset = 0; + return BlinkLED; +} + + /** * aac_command_thread - command processing thread * @dev: Adapter to monitor diff --git a/drivers/scsi/aacraid/linit.c b/drivers/scsi/aacraid/linit.c index 9d8b550a91c..d67058f8081 100644 --- a/drivers/scsi/aacraid/linit.c +++ b/drivers/scsi/aacraid/linit.c @@ -454,17 +454,17 @@ static int aac_eh_reset(struct scsi_cmnd* cmd) printk(KERN_ERR "%s: Host adapter reset request. SCSI hang ?\n", AAC_DRIVERNAME); aac = (struct aac_dev *)host->hostdata; - if (aac_adapter_check_health(aac)) { - printk(KERN_ERR "%s: Host adapter appears dead\n", - AAC_DRIVERNAME); - return -ENODEV; - } + + if ((count = aac_check_health(aac))) + return count; /* * Wait for all commands to complete to this specific * target (block maximum 60 seconds). */ for (count = 60; count; --count) { - int active = 0; + int active = aac->in_reset; + + if (active == 0) __shost_for_each_device(dev, host) { spin_lock_irqsave(&dev->list_lock, flags); list_for_each_entry(command, &dev->cmd_list, list) { @@ -933,7 +933,7 @@ static int __devinit aac_probe_one(struct pci_dev *pdev, else shost->max_channel = 0; - aac_get_config_status(aac); + aac_get_config_status(aac, 0); aac_get_containers(aac); list_add(&aac->entry, insert); -- cgit v1.2.3-70-g09d2