Staging: hv: storvsc: Move the storage driver out of the staging area
authorK. Y. Srinivasan <kys@microsoft.com>
Thu, 9 Feb 2012 22:14:48 +0000 (14:14 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 13 Feb 2012 19:35:20 +0000 (11:35 -0800)
The storage driver (storvsc_drv.c) handles all block storage devices
assigned to Linux guests hosted on Hyper-V. This driver has been in the
staging tree for a while and this patch moves it out of the staging area.

James was willing to apply this patch during the 3.3-rc phase and a decision
was taken to defer this to 3.4 since Greg had queued up a bunch of storvsc
patches for 3.4. Now that Greg has applied all of the pending storvsc patches,
I am sending this patch to move this driver out of staging. Based on James'
recommendation, this patch gets rid of the unneeded files in the staging/hv
directory.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Acked-by: James Bottomley <JBottomley@Parallels.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/scsi/Kconfig
drivers/scsi/Makefile
drivers/scsi/storvsc_drv.c [new file with mode: 0644]
drivers/staging/Kconfig
drivers/staging/Makefile
drivers/staging/hv/Kconfig [deleted file]
drivers/staging/hv/Makefile [deleted file]
drivers/staging/hv/TODO [deleted file]
drivers/staging/hv/storvsc_drv.c [deleted file]

index 16570aa84aac0e1e420a541d7224ecfd0e567bbc..d3d18e89cb5761dbca06245ff92dd7978d5690f2 100644 (file)
@@ -662,6 +662,13 @@ config VMWARE_PVSCSI
          To compile this driver as a module, choose M here: the
          module will be called vmw_pvscsi.
 
+config HYPERV_STORAGE
+       tristate "Microsoft Hyper-V virtual storage driver"
+       depends on SCSI && HYPERV
+       default HYPERV
+       help
+         Select this option to enable the Hyper-V virtual storage driver.
+
 config LIBFC
        tristate "LibFC module"
        select SCSI_FC_ATTRS
index 2b887498be50135bcd68a5e8b0a736f64482cd55..e4c1a69f8fab94abcf012a705d9c7471a21a1922 100644 (file)
@@ -142,6 +142,7 @@ obj-$(CONFIG_SCSI_BNX2_ISCSI)       += libiscsi.o bnx2i/
 obj-$(CONFIG_BE2ISCSI)         += libiscsi.o be2iscsi/
 obj-$(CONFIG_SCSI_PMCRAID)     += pmcraid.o
 obj-$(CONFIG_VMWARE_PVSCSI)    += vmw_pvscsi.o
+obj-$(CONFIG_HYPERV_STORAGE)   += hv_storvsc.o
 
 obj-$(CONFIG_ARM)              += arm/
 
@@ -170,6 +171,8 @@ scsi_mod-$(CONFIG_SCSI_PROC_FS)     += scsi_proc.o
 scsi_mod-y                     += scsi_trace.o
 scsi_mod-$(CONFIG_PM)          += scsi_pm.o
 
+hv_storvsc-y                   := storvsc_drv.o
+
 scsi_tgt-y                     += scsi_tgt_lib.o scsi_tgt_if.o
 
 sd_mod-objs    := sd.o
diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
new file mode 100644 (file)
index 0000000..695ffc3
--- /dev/null
@@ -0,0 +1,1548 @@
+/*
+ * Copyright (c) 2009, Microsoft Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Authors:
+ *   Haiyang Zhang <haiyangz@microsoft.com>
+ *   Hank Janssen  <hjanssen@microsoft.com>
+ *   K. Y. Srinivasan <kys@microsoft.com>
+ */
+
+#include <linux/kernel.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/completion.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/device.h>
+#include <linux/hyperv.h>
+#include <linux/mempool.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_tcq.h>
+#include <scsi/scsi_eh.h>
+#include <scsi/scsi_devinfo.h>
+#include <scsi/scsi_dbg.h>
+
+/*
+ * All wire protocol details (storage protocol between the guest and the host)
+ * are consolidated here.
+ *
+ * Begin protocol definitions.
+ */
+
+/*
+ * Version history:
+ * V1 Beta: 0.1
+ * V1 RC < 2008/1/31: 1.0
+ * V1 RC > 2008/1/31:  2.0
+ * Win7: 4.2
+ */
+
+#define VMSTOR_CURRENT_MAJOR  4
+#define VMSTOR_CURRENT_MINOR  2
+
+
+/*  Packet structure describing virtual storage requests. */
+enum vstor_packet_operation {
+       VSTOR_OPERATION_COMPLETE_IO             = 1,
+       VSTOR_OPERATION_REMOVE_DEVICE           = 2,
+       VSTOR_OPERATION_EXECUTE_SRB             = 3,
+       VSTOR_OPERATION_RESET_LUN               = 4,
+       VSTOR_OPERATION_RESET_ADAPTER           = 5,
+       VSTOR_OPERATION_RESET_BUS               = 6,
+       VSTOR_OPERATION_BEGIN_INITIALIZATION    = 7,
+       VSTOR_OPERATION_END_INITIALIZATION      = 8,
+       VSTOR_OPERATION_QUERY_PROTOCOL_VERSION  = 9,
+       VSTOR_OPERATION_QUERY_PROPERTIES        = 10,
+       VSTOR_OPERATION_ENUMERATE_BUS           = 11,
+       VSTOR_OPERATION_MAXIMUM                 = 11
+};
+
+/*
+ * Platform neutral description of a scsi request -
+ * this remains the same across the write regardless of 32/64 bit
+ * note: it's patterned off the SCSI_PASS_THROUGH structure
+ */
+#define STORVSC_MAX_CMD_LEN                    0x10
+#define STORVSC_SENSE_BUFFER_SIZE              0x12
+#define STORVSC_MAX_BUF_LEN_WITH_PADDING       0x14
+
+struct vmscsi_request {
+       u16 length;
+       u8 srb_status;
+       u8 scsi_status;
+
+       u8  port_number;
+       u8  path_id;
+       u8  target_id;
+       u8  lun;
+
+       u8  cdb_length;
+       u8  sense_info_length;
+       u8  data_in;
+       u8  reserved;
+
+       u32 data_transfer_length;
+
+       union {
+               u8 cdb[STORVSC_MAX_CMD_LEN];
+               u8 sense_data[STORVSC_SENSE_BUFFER_SIZE];
+               u8 reserved_array[STORVSC_MAX_BUF_LEN_WITH_PADDING];
+       };
+} __attribute((packed));
+
+
+/*
+ * This structure is sent during the intialization phase to get the different
+ * properties of the channel.
+ */
+struct vmstorage_channel_properties {
+       u16 protocol_version;
+       u8  path_id;
+       u8 target_id;
+
+       /* Note: port number is only really known on the client side */
+       u32  port_number;
+       u32  flags;
+       u32   max_transfer_bytes;
+
+       /*
+        * This id is unique for each channel and will correspond with
+        * vendor specific data in the inquiry data.
+        */
+
+       u64  unique_id;
+} __packed;
+
+/*  This structure is sent during the storage protocol negotiations. */
+struct vmstorage_protocol_version {
+       /* Major (MSW) and minor (LSW) version numbers. */
+       u16 major_minor;
+
+       /*
+        * Revision number is auto-incremented whenever this file is changed
+        * (See FILL_VMSTOR_REVISION macro above).  Mismatch does not
+        * definitely indicate incompatibility--but it does indicate mismatched
+        * builds.
+        * This is only used on the windows side. Just set it to 0.
+        */
+       u16 revision;
+} __packed;
+
+/* Channel Property Flags */
+#define STORAGE_CHANNEL_REMOVABLE_FLAG         0x1
+#define STORAGE_CHANNEL_EMULATED_IDE_FLAG      0x2
+
+struct vstor_packet {
+       /* Requested operation type */
+       enum vstor_packet_operation operation;
+
+       /*  Flags - see below for values */
+       u32 flags;
+
+       /* Status of the request returned from the server side. */
+       u32 status;
+
+       /* Data payload area */
+       union {
+               /*
+                * Structure used to forward SCSI commands from the
+                * client to the server.
+                */
+               struct vmscsi_request vm_srb;
+
+               /* Structure used to query channel properties. */
+               struct vmstorage_channel_properties storage_channel_properties;
+
+               /* Used during version negotiations. */
+               struct vmstorage_protocol_version version;
+       };
+} __packed;
+
+/*
+ * Packet Flags:
+ *
+ * This flag indicates that the server should send back a completion for this
+ * packet.
+ */
+
+#define REQUEST_COMPLETION_FLAG        0x1
+
+/* Matches Windows-end */
+enum storvsc_request_type {
+       WRITE_TYPE = 0,
+       READ_TYPE,
+       UNKNOWN_TYPE,
+};
+
+/*
+ * SRB status codes and masks; a subset of the codes used here.
+ */
+
+#define SRB_STATUS_AUTOSENSE_VALID     0x80
+#define SRB_STATUS_INVALID_LUN 0x20
+#define SRB_STATUS_SUCCESS     0x01
+#define SRB_STATUS_ERROR       0x04
+
+/*
+ * This is the end of Protocol specific defines.
+ */
+
+
+/*
+ * We setup a mempool to allocate request structures for this driver
+ * on a per-lun basis. The following define specifies the number of
+ * elements in the pool.
+ */
+
+#define STORVSC_MIN_BUF_NR                             64
+static int storvsc_ringbuffer_size = (20 * PAGE_SIZE);
+
+module_param(storvsc_ringbuffer_size, int, S_IRUGO);
+MODULE_PARM_DESC(storvsc_ringbuffer_size, "Ring buffer size (bytes)");
+
+#define STORVSC_MAX_IO_REQUESTS                                128
+
+/*
+ * In Hyper-V, each port/path/target maps to 1 scsi host adapter.  In
+ * reality, the path/target is not used (ie always set to 0) so our
+ * scsi host adapter essentially has 1 bus with 1 target that contains
+ * up to 256 luns.
+ */
+#define STORVSC_MAX_LUNS_PER_TARGET                    64
+#define STORVSC_MAX_TARGETS                            1
+#define STORVSC_MAX_CHANNELS                           1
+
+
+
+struct storvsc_cmd_request {
+       struct list_head entry;
+       struct scsi_cmnd *cmd;
+
+       unsigned int bounce_sgl_count;
+       struct scatterlist *bounce_sgl;
+
+       struct hv_device *device;
+
+       /* Synchronize the request/response if needed */
+       struct completion wait_event;
+
+       unsigned char *sense_buffer;
+       struct hv_multipage_buffer data_buffer;
+       struct vstor_packet vstor_packet;
+};
+
+
+/* A storvsc device is a device object that contains a vmbus channel */
+struct storvsc_device {
+       struct hv_device *device;
+
+       bool     destroy;
+       bool     drain_notify;
+       atomic_t num_outstanding_req;
+       struct Scsi_Host *host;
+
+       wait_queue_head_t waiting_to_drain;
+
+       /*
+        * Each unique Port/Path/Target represents 1 channel ie scsi
+        * controller. In reality, the pathid, targetid is always 0
+        * and the port is set by us
+        */
+       unsigned int port_number;
+       unsigned char path_id;
+       unsigned char target_id;
+
+       /* Used for vsc/vsp channel reset process */
+       struct storvsc_cmd_request init_request;
+       struct storvsc_cmd_request reset_request;
+};
+
+struct stor_mem_pools {
+       struct kmem_cache *request_pool;
+       mempool_t *request_mempool;
+};
+
+struct hv_host_device {
+       struct hv_device *dev;
+       unsigned int port;
+       unsigned char path;
+       unsigned char target;
+};
+
+struct storvsc_scan_work {
+       struct work_struct work;
+       struct Scsi_Host *host;
+       uint lun;
+};
+
+static void storvsc_bus_scan(struct work_struct *work)
+{
+       struct storvsc_scan_work *wrk;
+       int id, order_id;
+
+       wrk = container_of(work, struct storvsc_scan_work, work);
+       for (id = 0; id < wrk->host->max_id; ++id) {
+               if (wrk->host->reverse_ordering)
+                       order_id = wrk->host->max_id - id - 1;
+               else
+                       order_id = id;
+
+               scsi_scan_target(&wrk->host->shost_gendev, 0,
+                               order_id, SCAN_WILD_CARD, 1);
+       }
+       kfree(wrk);
+}
+
+static void storvsc_remove_lun(struct work_struct *work)
+{
+       struct storvsc_scan_work *wrk;
+       struct scsi_device *sdev;
+
+       wrk = container_of(work, struct storvsc_scan_work, work);
+       if (!scsi_host_get(wrk->host))
+               goto done;
+
+       sdev = scsi_device_lookup(wrk->host, 0, 0, wrk->lun);
+
+       if (sdev) {
+               scsi_remove_device(sdev);
+               scsi_device_put(sdev);
+       }
+       scsi_host_put(wrk->host);
+
+done:
+       kfree(wrk);
+}
+
+/*
+ * Major/minor macros.  Minor version is in LSB, meaning that earlier flat
+ * version numbers will be interpreted as "0.x" (i.e., 1 becomes 0.1).
+ */
+
+static inline u16 storvsc_get_version(u8 major, u8 minor)
+{
+       u16 version;
+
+       version = ((major << 8) | minor);
+       return version;
+}
+
+/*
+ * We can get incoming messages from the host that are not in response to
+ * messages that we have sent out. An example of this would be messages
+ * received by the guest to notify dynamic addition/removal of LUNs. To
+ * deal with potential race conditions where the driver may be in the
+ * midst of being unloaded when we might receive an unsolicited message
+ * from the host, we have implemented a mechanism to gurantee sequential
+ * consistency:
+ *
+ * 1) Once the device is marked as being destroyed, we will fail all
+ *    outgoing messages.
+ * 2) We permit incoming messages when the device is being destroyed,
+ *    only to properly account for messages already sent out.
+ */
+
+static inline struct storvsc_device *get_out_stor_device(
+                                       struct hv_device *device)
+{
+       struct storvsc_device *stor_device;
+
+       stor_device = hv_get_drvdata(device);
+
+       if (stor_device && stor_device->destroy)
+               stor_device = NULL;
+
+       return stor_device;
+}
+
+
+static inline void storvsc_wait_to_drain(struct storvsc_device *dev)
+{
+       dev->drain_notify = true;
+       wait_event(dev->waiting_to_drain,
+                  atomic_read(&dev->num_outstanding_req) == 0);
+       dev->drain_notify = false;
+}
+
+static inline struct storvsc_device *get_in_stor_device(
+                                       struct hv_device *device)
+{
+       struct storvsc_device *stor_device;
+
+       stor_device = hv_get_drvdata(device);
+
+       if (!stor_device)
+               goto get_in_err;
+
+       /*
+        * If the device is being destroyed; allow incoming
+        * traffic only to cleanup outstanding requests.
+        */
+
+       if (stor_device->destroy  &&
+               (atomic_read(&stor_device->num_outstanding_req) == 0))
+               stor_device = NULL;
+
+get_in_err:
+       return stor_device;
+
+}
+
+static void destroy_bounce_buffer(struct scatterlist *sgl,
+                                 unsigned int sg_count)
+{
+       int i;
+       struct page *page_buf;
+
+       for (i = 0; i < sg_count; i++) {
+               page_buf = sg_page((&sgl[i]));
+               if (page_buf != NULL)
+                       __free_page(page_buf);
+       }
+
+       kfree(sgl);
+}
+
+static int do_bounce_buffer(struct scatterlist *sgl, unsigned int sg_count)
+{
+       int i;
+
+       /* No need to check */
+       if (sg_count < 2)
+               return -1;
+
+       /* We have at least 2 sg entries */
+       for (i = 0; i < sg_count; i++) {
+               if (i == 0) {
+                       /* make sure 1st one does not have hole */
+                       if (sgl[i].offset + sgl[i].length != PAGE_SIZE)
+                               return i;
+               } else if (i == sg_count - 1) {
+                       /* make sure last one does not have hole */
+                       if (sgl[i].offset != 0)
+                               return i;
+               } else {
+                       /* make sure no hole in the middle */
+                       if (sgl[i].length != PAGE_SIZE || sgl[i].offset != 0)
+                               return i;
+               }
+       }
+       return -1;
+}
+
+static struct scatterlist *create_bounce_buffer(struct scatterlist *sgl,
+                                               unsigned int sg_count,
+                                               unsigned int len,
+                                               int write)
+{
+       int i;
+       int num_pages;
+       struct scatterlist *bounce_sgl;
+       struct page *page_buf;
+       unsigned int buf_len = ((write == WRITE_TYPE) ? 0 : PAGE_SIZE);
+
+       num_pages = ALIGN(len, PAGE_SIZE) >> PAGE_SHIFT;
+
+       bounce_sgl = kcalloc(num_pages, sizeof(struct scatterlist), GFP_ATOMIC);
+       if (!bounce_sgl)
+               return NULL;
+
+       for (i = 0; i < num_pages; i++) {
+               page_buf = alloc_page(GFP_ATOMIC);
+               if (!page_buf)
+                       goto cleanup;
+               sg_set_page(&bounce_sgl[i], page_buf, buf_len, 0);
+       }
+
+       return bounce_sgl;
+
+cleanup:
+       destroy_bounce_buffer(bounce_sgl, num_pages);
+       return NULL;
+}
+
+/* Assume the original sgl has enough room */
+static unsigned int copy_from_bounce_buffer(struct scatterlist *orig_sgl,
+                                           struct scatterlist *bounce_sgl,
+                                           unsigned int orig_sgl_count,
+                                           unsigned int bounce_sgl_count)
+{
+       int i;
+       int j = 0;
+       unsigned long src, dest;
+       unsigned int srclen, destlen, copylen;
+       unsigned int total_copied = 0;
+       unsigned long bounce_addr = 0;
+       unsigned long dest_addr = 0;
+       unsigned long flags;
+
+       local_irq_save(flags);
+
+       for (i = 0; i < orig_sgl_count; i++) {
+               dest_addr = (unsigned long)kmap_atomic(sg_page((&orig_sgl[i])),
+                                       KM_IRQ0) + orig_sgl[i].offset;
+               dest = dest_addr;
+               destlen = orig_sgl[i].length;
+
+               if (bounce_addr == 0)
+                       bounce_addr =
+                       (unsigned long)kmap_atomic(sg_page((&bounce_sgl[j])),
+                                                       KM_IRQ0);
+
+               while (destlen) {
+                       src = bounce_addr + bounce_sgl[j].offset;
+                       srclen = bounce_sgl[j].length - bounce_sgl[j].offset;
+
+                       copylen = min(srclen, destlen);
+                       memcpy((void *)dest, (void *)src, copylen);
+
+                       total_copied += copylen;
+                       bounce_sgl[j].offset += copylen;
+                       destlen -= copylen;
+                       dest += copylen;
+
+                       if (bounce_sgl[j].offset == bounce_sgl[j].length) {
+                               /* full */
+                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
+                               j++;
+
+                               /*
+                                * It is possible that the number of elements
+                                * in the bounce buffer may not be equal to
+                                * the number of elements in the original
+                                * scatter list. Handle this correctly.
+                                */
+
+                               if (j == bounce_sgl_count) {
+                                       /*
+                                        * We are done; cleanup and return.
+                                        */
+                                       kunmap_atomic((void *)(dest_addr -
+                                                       orig_sgl[i].offset),
+                                                       KM_IRQ0);
+                                       local_irq_restore(flags);
+                                       return total_copied;
+                               }
+
+                               /* if we need to use another bounce buffer */
+                               if (destlen || i != orig_sgl_count - 1)
+                                       bounce_addr =
+                                       (unsigned long)kmap_atomic(
+                                       sg_page((&bounce_sgl[j])), KM_IRQ0);
+                       } else if (destlen == 0 && i == orig_sgl_count - 1) {
+                               /* unmap the last bounce that is < PAGE_SIZE */
+                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
+                       }
+               }
+
+               kunmap_atomic((void *)(dest_addr - orig_sgl[i].offset),
+                             KM_IRQ0);
+       }
+
+       local_irq_restore(flags);
+
+       return total_copied;
+}
+
+/* Assume the bounce_sgl has enough room ie using the create_bounce_buffer() */
+static unsigned int copy_to_bounce_buffer(struct scatterlist *orig_sgl,
+                                         struct scatterlist *bounce_sgl,
+                                         unsigned int orig_sgl_count)
+{
+       int i;
+       int j = 0;
+       unsigned long src, dest;
+       unsigned int srclen, destlen, copylen;
+       unsigned int total_copied = 0;
+       unsigned long bounce_addr = 0;
+       unsigned long src_addr = 0;
+       unsigned long flags;
+
+       local_irq_save(flags);
+
+       for (i = 0; i < orig_sgl_count; i++) {
+               src_addr = (unsigned long)kmap_atomic(sg_page((&orig_sgl[i])),
+                               KM_IRQ0) + orig_sgl[i].offset;
+               src = src_addr;
+               srclen = orig_sgl[i].length;
+
+               if (bounce_addr == 0)
+                       bounce_addr =
+                       (unsigned long)kmap_atomic(sg_page((&bounce_sgl[j])),
+                                               KM_IRQ0);
+
+               while (srclen) {
+                       /* assume bounce offset always == 0 */
+                       dest = bounce_addr + bounce_sgl[j].length;
+                       destlen = PAGE_SIZE - bounce_sgl[j].length;
+
+                       copylen = min(srclen, destlen);
+                       memcpy((void *)dest, (void *)src, copylen);
+
+                       total_copied += copylen;
+                       bounce_sgl[j].length += copylen;
+                       srclen -= copylen;
+                       src += copylen;
+
+                       if (bounce_sgl[j].length == PAGE_SIZE) {
+                               /* full..move to next entry */
+                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
+                               j++;
+
+                               /* if we need to use another bounce buffer */
+                               if (srclen || i != orig_sgl_count - 1)
+                                       bounce_addr =
+                                       (unsigned long)kmap_atomic(
+                                       sg_page((&bounce_sgl[j])), KM_IRQ0);
+
+                       } else if (srclen == 0 && i == orig_sgl_count - 1) {
+                               /* unmap the last bounce that is < PAGE_SIZE */
+                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
+                       }
+               }
+
+               kunmap_atomic((void *)(src_addr - orig_sgl[i].offset), KM_IRQ0);
+       }
+
+       local_irq_restore(flags);
+
+       return total_copied;
+}
+
+static int storvsc_channel_init(struct hv_device *device)
+{
+       struct storvsc_device *stor_device;
+       struct storvsc_cmd_request *request;
+       struct vstor_packet *vstor_packet;
+       int ret, t;
+
+       stor_device = get_out_stor_device(device);
+       if (!stor_device)
+               return -ENODEV;
+
+       request = &stor_device->init_request;
+       vstor_packet = &request->vstor_packet;
+
+       /*
+        * Now, initiate the vsc/vsp initialization protocol on the open
+        * channel
+        */
+       memset(request, 0, sizeof(struct storvsc_cmd_request));
+       init_completion(&request->wait_event);
+       vstor_packet->operation = VSTOR_OPERATION_BEGIN_INITIALIZATION;
+       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
+
+       ret = vmbus_sendpacket(device->channel, vstor_packet,
+                              sizeof(struct vstor_packet),
+                              (unsigned long)request,
+                              VM_PKT_DATA_INBAND,
+                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
+       if (ret != 0)
+               goto cleanup;
+
+       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
+       if (t == 0) {
+               ret = -ETIMEDOUT;
+               goto cleanup;
+       }
+
+       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
+           vstor_packet->status != 0)
+               goto cleanup;
+
+
+       /* reuse the packet for version range supported */
+       memset(vstor_packet, 0, sizeof(struct vstor_packet));
+       vstor_packet->operation = VSTOR_OPERATION_QUERY_PROTOCOL_VERSION;
+       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
+
+       vstor_packet->version.major_minor =
+               storvsc_get_version(VMSTOR_CURRENT_MAJOR, VMSTOR_CURRENT_MINOR);
+
+       /*
+        * The revision number is only used in Windows; set it to 0.
+        */
+       vstor_packet->version.revision = 0;
+
+       ret = vmbus_sendpacket(device->channel, vstor_packet,
+                              sizeof(struct vstor_packet),
+                              (unsigned long)request,
+                              VM_PKT_DATA_INBAND,
+                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
+       if (ret != 0)
+               goto cleanup;
+
+       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
+       if (t == 0) {
+               ret = -ETIMEDOUT;
+               goto cleanup;
+       }
+
+       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
+           vstor_packet->status != 0)
+               goto cleanup;
+
+
+       memset(vstor_packet, 0, sizeof(struct vstor_packet));
+       vstor_packet->operation = VSTOR_OPERATION_QUERY_PROPERTIES;
+       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
+       vstor_packet->storage_channel_properties.port_number =
+                                       stor_device->port_number;
+
+       ret = vmbus_sendpacket(device->channel, vstor_packet,
+                              sizeof(struct vstor_packet),
+                              (unsigned long)request,
+                              VM_PKT_DATA_INBAND,
+                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
+
+       if (ret != 0)
+               goto cleanup;
+
+       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
+       if (t == 0) {
+               ret = -ETIMEDOUT;
+               goto cleanup;
+       }
+
+       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
+           vstor_packet->status != 0)
+               goto cleanup;
+
+       stor_device->path_id = vstor_packet->storage_channel_properties.path_id;
+       stor_device->target_id
+               = vstor_packet->storage_channel_properties.target_id;
+
+       memset(vstor_packet, 0, sizeof(struct vstor_packet));
+       vstor_packet->operation = VSTOR_OPERATION_END_INITIALIZATION;
+       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
+
+       ret = vmbus_sendpacket(device->channel, vstor_packet,
+                              sizeof(struct vstor_packet),
+                              (unsigned long)request,
+                              VM_PKT_DATA_INBAND,
+                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
+
+       if (ret != 0)
+               goto cleanup;
+
+       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
+       if (t == 0) {
+               ret = -ETIMEDOUT;
+               goto cleanup;
+       }
+
+       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
+           vstor_packet->status != 0)
+               goto cleanup;
+
+
+cleanup:
+       return ret;
+}
+
+
+static void storvsc_command_completion(struct storvsc_cmd_request *cmd_request)
+{
+       struct scsi_cmnd *scmnd = cmd_request->cmd;
+       struct hv_host_device *host_dev = shost_priv(scmnd->device->host);
+       void (*scsi_done_fn)(struct scsi_cmnd *);
+       struct scsi_sense_hdr sense_hdr;
+       struct vmscsi_request *vm_srb;
+       struct storvsc_scan_work *wrk;
+       struct stor_mem_pools *memp = scmnd->device->hostdata;
+
+       vm_srb = &cmd_request->vstor_packet.vm_srb;
+       if (cmd_request->bounce_sgl_count) {
+               if (vm_srb->data_in == READ_TYPE)
+                       copy_from_bounce_buffer(scsi_sglist(scmnd),
+                                       cmd_request->bounce_sgl,
+                                       scsi_sg_count(scmnd),
+                                       cmd_request->bounce_sgl_count);
+               destroy_bounce_buffer(cmd_request->bounce_sgl,
+                                       cmd_request->bounce_sgl_count);
+       }
+
+       /*
+        * If there is an error; offline the device since all
+        * error recovery strategies would have already been
+        * deployed on the host side.
+        */
+       if (vm_srb->srb_status == SRB_STATUS_ERROR)
+               scmnd->result = DID_TARGET_FAILURE << 16;
+       else
+               scmnd->result = vm_srb->scsi_status;
+
+       /*
+        * If the LUN is invalid; remove the device.
+        */
+       if (vm_srb->srb_status == SRB_STATUS_INVALID_LUN) {
+               struct storvsc_device *stor_dev;
+               struct hv_device *dev = host_dev->dev;
+               struct Scsi_Host *host;
+
+               stor_dev = get_in_stor_device(dev);
+               host = stor_dev->host;
+
+               wrk = kmalloc(sizeof(struct storvsc_scan_work),
+                               GFP_ATOMIC);
+               if (!wrk) {
+                       scmnd->result = DID_TARGET_FAILURE << 16;
+               } else {
+                       wrk->host = host;
+                       wrk->lun = vm_srb->lun;
+                       INIT_WORK(&wrk->work, storvsc_remove_lun);
+                       schedule_work(&wrk->work);
+               }
+       }
+
+       if (scmnd->result) {
+               if (scsi_normalize_sense(scmnd->sense_buffer,
+                               SCSI_SENSE_BUFFERSIZE, &sense_hdr))
+                       scsi_print_sense_hdr("storvsc", &sense_hdr);
+       }
+
+       scsi_set_resid(scmnd,
+               cmd_request->data_buffer.len -
+               vm_srb->data_transfer_length);
+
+       scsi_done_fn = scmnd->scsi_done;
+
+       scmnd->host_scribble = NULL;
+       scmnd->scsi_done = NULL;
+
+       scsi_done_fn(scmnd);
+
+       mempool_free(cmd_request, memp->request_mempool);
+}
+
+static void storvsc_on_io_completion(struct hv_device *device,
+                                 struct vstor_packet *vstor_packet,
+                                 struct storvsc_cmd_request *request)
+{
+       struct storvsc_device *stor_device;
+       struct vstor_packet *stor_pkt;
+
+       stor_device = hv_get_drvdata(device);
+       stor_pkt = &request->vstor_packet;
+
+       /*
+        * The current SCSI handling on the host side does
+        * not correctly handle:
+        * INQUIRY command with page code parameter set to 0x80
+        * MODE_SENSE command with cmd[2] == 0x1c
+        *
+        * Setup srb and scsi status so this won't be fatal.
+        * We do this so we can distinguish truly fatal failues
+        * (srb status == 0x4) and off-line the device in that case.
+        */
+
+       if ((stor_pkt->vm_srb.cdb[0] == INQUIRY) ||
+          (stor_pkt->vm_srb.cdb[0] == MODE_SENSE)) {
+               vstor_packet->vm_srb.scsi_status = 0;
+               vstor_packet->vm_srb.srb_status = SRB_STATUS_SUCCESS;
+       }
+
+
+       /* Copy over the status...etc */
+       stor_pkt->vm_srb.scsi_status = vstor_packet->vm_srb.scsi_status;
+       stor_pkt->vm_srb.srb_status = vstor_packet->vm_srb.srb_status;
+       stor_pkt->vm_srb.sense_info_length =
+       vstor_packet->vm_srb.sense_info_length;
+
+       if (vstor_packet->vm_srb.scsi_status != 0 ||
+               vstor_packet->vm_srb.srb_status != SRB_STATUS_SUCCESS){
+               dev_warn(&device->device,
+                        "cmd 0x%x scsi status 0x%x srb status 0x%x\n",
+                        stor_pkt->vm_srb.cdb[0],
+                        vstor_packet->vm_srb.scsi_status,
+                        vstor_packet->vm_srb.srb_status);
+       }
+
+       if ((vstor_packet->vm_srb.scsi_status & 0xFF) == 0x02) {
+               /* CHECK_CONDITION */
+               if (vstor_packet->vm_srb.srb_status &
+                       SRB_STATUS_AUTOSENSE_VALID) {
+                       /* autosense data available */
+                       dev_warn(&device->device,
+                                "stor pkt %p autosense data valid - len %d\n",
+                                request,
+                                vstor_packet->vm_srb.sense_info_length);
+
+                       memcpy(request->sense_buffer,
+                              vstor_packet->vm_srb.sense_data,
+                              vstor_packet->vm_srb.sense_info_length);
+
+               }
+       }
+
+       stor_pkt->vm_srb.data_transfer_length =
+       vstor_packet->vm_srb.data_transfer_length;
+
+       storvsc_command_completion(request);
+
+       if (atomic_dec_and_test(&stor_device->num_outstanding_req) &&
+               stor_device->drain_notify)
+               wake_up(&stor_device->waiting_to_drain);
+
+
+}
+
+static void storvsc_on_receive(struct hv_device *device,
+                            struct vstor_packet *vstor_packet,
+                            struct storvsc_cmd_request *request)
+{
+       struct storvsc_scan_work *work;
+       struct storvsc_device *stor_device;
+
+       switch (vstor_packet->operation) {
+       case VSTOR_OPERATION_COMPLETE_IO:
+               storvsc_on_io_completion(device, vstor_packet, request);
+               break;
+
+       case VSTOR_OPERATION_REMOVE_DEVICE:
+       case VSTOR_OPERATION_ENUMERATE_BUS:
+               stor_device = get_in_stor_device(device);
+               work = kmalloc(sizeof(struct storvsc_scan_work), GFP_ATOMIC);
+               if (!work)
+                       return;
+
+               INIT_WORK(&work->work, storvsc_bus_scan);
+               work->host = stor_device->host;
+               schedule_work(&work->work);
+               break;
+
+       default:
+               break;
+       }
+}
+
+static void storvsc_on_channel_callback(void *context)
+{
+       struct hv_device *device = (struct hv_device *)context;
+       struct storvsc_device *stor_device;
+       u32 bytes_recvd;
+       u64 request_id;
+       unsigned char packet[ALIGN(sizeof(struct vstor_packet), 8)];
+       struct storvsc_cmd_request *request;
+       int ret;
+
+
+       stor_device = get_in_stor_device(device);
+       if (!stor_device)
+               return;
+
+       do {
+               ret = vmbus_recvpacket(device->channel, packet,
+                                      ALIGN(sizeof(struct vstor_packet), 8),
+                                      &bytes_recvd, &request_id);
+               if (ret == 0 && bytes_recvd > 0) {
+
+                       request = (struct storvsc_cmd_request *)
+                                       (unsigned long)request_id;
+
+                       if ((request == &stor_device->init_request) ||
+                           (request == &stor_device->reset_request)) {
+
+                               memcpy(&request->vstor_packet, packet,
+                                      sizeof(struct vstor_packet));
+                               complete(&request->wait_event);
+                       } else {
+                               storvsc_on_receive(device,
+                                               (struct vstor_packet *)packet,
+                                               request);
+                       }
+               } else {
+                       break;
+               }
+       } while (1);
+
+       return;
+}
+
+static int storvsc_connect_to_vsp(struct hv_device *device, u32 ring_size)
+{
+       struct vmstorage_channel_properties props;
+       int ret;
+
+       memset(&props, 0, sizeof(struct vmstorage_channel_properties));
+
+       ret = vmbus_open(device->channel,
+                        ring_size,
+                        ring_size,
+                        (void *)&props,
+                        sizeof(struct vmstorage_channel_properties),
+                        storvsc_on_channel_callback, device);
+
+       if (ret != 0)
+               return ret;
+
+       ret = storvsc_channel_init(device);
+
+       return ret;
+}
+
+static int storvsc_dev_remove(struct hv_device *device)
+{
+       struct storvsc_device *stor_device;
+       unsigned long flags;
+
+       stor_device = hv_get_drvdata(device);
+
+       spin_lock_irqsave(&device->channel->inbound_lock, flags);
+       stor_device->destroy = true;
+       spin_unlock_irqrestore(&device->channel->inbound_lock, flags);
+
+       /*
+        * At this point, all outbound traffic should be disable. We
+        * only allow inbound traffic (responses) to proceed so that
+        * outstanding requests can be completed.
+        */
+
+       storvsc_wait_to_drain(stor_device);
+
+       /*
+        * Since we have already drained, we don't need to busy wait
+        * as was done in final_release_stor_device()
+        * Note that we cannot set the ext pointer to NULL until
+        * we have drained - to drain the outgoing packets, we need to
+        * allow incoming packets.
+        */
+       spin_lock_irqsave(&device->channel->inbound_lock, flags);
+       hv_set_drvdata(device, NULL);
+       spin_unlock_irqrestore(&device->channel->inbound_lock, flags);
+
+       /* Close the channel */
+       vmbus_close(device->channel);
+
+       kfree(stor_device);
+       return 0;
+}
+
+static int storvsc_do_io(struct hv_device *device,
+                             struct storvsc_cmd_request *request)
+{
+       struct storvsc_device *stor_device;
+       struct vstor_packet *vstor_packet;
+       int ret = 0;
+
+       vstor_packet = &request->vstor_packet;
+       stor_device = get_out_stor_device(device);
+
+       if (!stor_device)
+               return -ENODEV;
+
+
+       request->device  = device;
+
+
+       vstor_packet->flags |= REQUEST_COMPLETION_FLAG;
+
+       vstor_packet->vm_srb.length = sizeof(struct vmscsi_request);
+
+
+       vstor_packet->vm_srb.sense_info_length = STORVSC_SENSE_BUFFER_SIZE;
+
+
+       vstor_packet->vm_srb.data_transfer_length =
+       request->data_buffer.len;
+
+       vstor_packet->operation = VSTOR_OPERATION_EXECUTE_SRB;
+
+       if (request->data_buffer.len) {
+               ret = vmbus_sendpacket_multipagebuffer(device->channel,
+                               &request->data_buffer,
+                               vstor_packet,
+                               sizeof(struct vstor_packet),
+                               (unsigned long)request);
+       } else {
+               ret = vmbus_sendpacket(device->channel, vstor_packet,
+                              sizeof(struct vstor_packet),
+                              (unsigned long)request,
+                              VM_PKT_DATA_INBAND,
+                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
+       }
+
+       if (ret != 0)
+               return ret;
+
+       atomic_inc(&stor_device->num_outstanding_req);
+
+       return ret;
+}
+
+static int storvsc_device_alloc(struct scsi_device *sdevice)
+{
+       struct stor_mem_pools *memp;
+       int number = STORVSC_MIN_BUF_NR;
+
+       memp = kzalloc(sizeof(struct stor_mem_pools), GFP_KERNEL);
+       if (!memp)
+               return -ENOMEM;
+
+       memp->request_pool =
+               kmem_cache_create(dev_name(&sdevice->sdev_dev),
+                               sizeof(struct storvsc_cmd_request), 0,
+                               SLAB_HWCACHE_ALIGN, NULL);
+
+       if (!memp->request_pool)
+               goto err0;
+
+       memp->request_mempool = mempool_create(number, mempool_alloc_slab,
+                                               mempool_free_slab,
+                                               memp->request_pool);
+
+       if (!memp->request_mempool)
+               goto err1;
+
+       sdevice->hostdata = memp;
+
+       return 0;
+
+err1:
+       kmem_cache_destroy(memp->request_pool);
+
+err0:
+       kfree(memp);
+       return -ENOMEM;
+}
+
+static void storvsc_device_destroy(struct scsi_device *sdevice)
+{
+       struct stor_mem_pools *memp = sdevice->hostdata;
+
+       mempool_destroy(memp->request_mempool);
+       kmem_cache_destroy(memp->request_pool);
+       kfree(memp);
+       sdevice->hostdata = NULL;
+}
+
+static int storvsc_device_configure(struct scsi_device *sdevice)
+{
+       scsi_adjust_queue_depth(sdevice, MSG_SIMPLE_TAG,
+                               STORVSC_MAX_IO_REQUESTS);
+
+       blk_queue_max_segment_size(sdevice->request_queue, PAGE_SIZE);
+
+       blk_queue_bounce_limit(sdevice->request_queue, BLK_BOUNCE_ANY);
+
+       return 0;
+}
+
+static int storvsc_get_chs(struct scsi_device *sdev, struct block_device * bdev,
+                          sector_t capacity, int *info)
+{
+       sector_t nsect = capacity;
+       sector_t cylinders = nsect;
+       int heads, sectors_pt;
+
+       /*
+        * We are making up these values; let us keep it simple.
+        */
+       heads = 0xff;
+       sectors_pt = 0x3f;      /* Sectors per track */
+       sector_div(cylinders, heads * sectors_pt);
+       if ((sector_t)(cylinders + 1) * heads * sectors_pt < nsect)
+               cylinders = 0xffff;
+
+       info[0] = heads;
+       info[1] = sectors_pt;
+       info[2] = (int)cylinders;
+
+       return 0;
+}
+
+static int storvsc_host_reset_handler(struct scsi_cmnd *scmnd)
+{
+       struct hv_host_device *host_dev = shost_priv(scmnd->device->host);
+       struct hv_device *device = host_dev->dev;
+
+       struct storvsc_device *stor_device;
+       struct storvsc_cmd_request *request;
+       struct vstor_packet *vstor_packet;
+       int ret, t;
+
+
+       stor_device = get_out_stor_device(device);
+       if (!stor_device)
+               return FAILED;
+
+       request = &stor_device->reset_request;
+       vstor_packet = &request->vstor_packet;
+
+       init_completion(&request->wait_event);
+
+       vstor_packet->operation = VSTOR_OPERATION_RESET_BUS;
+       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
+       vstor_packet->vm_srb.path_id = stor_device->path_id;
+
+       ret = vmbus_sendpacket(device->channel, vstor_packet,
+                              sizeof(struct vstor_packet),
+                              (unsigned long)&stor_device->reset_request,
+                              VM_PKT_DATA_INBAND,
+                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
+       if (ret != 0)
+               return FAILED;
+
+       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
+       if (t == 0)
+               return TIMEOUT_ERROR;
+
+
+       /*
+        * At this point, all outstanding requests in the adapter
+        * should have been flushed out and return to us
+        */
+
+       return SUCCESS;
+}
+
+static bool storvsc_scsi_cmd_ok(struct scsi_cmnd *scmnd)
+{
+       bool allowed = true;
+       u8 scsi_op = scmnd->cmnd[0];
+
+       switch (scsi_op) {
+       /*
+        * smartd sends this command and the host does not handle
+        * this. So, don't send it.
+        */
+       case SET_WINDOW:
+               scmnd->result = ILLEGAL_REQUEST << 16;
+               allowed = false;
+               break;
+       default:
+               break;
+       }
+       return allowed;
+}
+
+static int storvsc_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scmnd)
+{
+       int ret;
+       struct hv_host_device *host_dev = shost_priv(host);
+       struct hv_device *dev = host_dev->dev;
+       struct storvsc_cmd_request *cmd_request;
+       unsigned int request_size = 0;
+       int i;
+       struct scatterlist *sgl;
+       unsigned int sg_count = 0;
+       struct vmscsi_request *vm_srb;
+       struct stor_mem_pools *memp = scmnd->device->hostdata;
+
+       if (!storvsc_scsi_cmd_ok(scmnd)) {
+               scmnd->scsi_done(scmnd);
+               return 0;
+       }
+
+       request_size = sizeof(struct storvsc_cmd_request);
+
+       cmd_request = mempool_alloc(memp->request_mempool,
+                                      GFP_ATOMIC);
+
+       /*
+        * We might be invoked in an interrupt context; hence
+        * mempool_alloc() can fail.
+        */
+       if (!cmd_request)
+               return SCSI_MLQUEUE_DEVICE_BUSY;
+
+       memset(cmd_request, 0, sizeof(struct storvsc_cmd_request));
+
+       /* Setup the cmd request */
+       cmd_request->cmd = scmnd;
+
+       scmnd->host_scribble = (unsigned char *)cmd_request;
+
+       vm_srb = &cmd_request->vstor_packet.vm_srb;
+
+
+       /* Build the SRB */
+       switch (scmnd->sc_data_direction) {
+       case DMA_TO_DEVICE:
+               vm_srb->data_in = WRITE_TYPE;
+               break;
+       case DMA_FROM_DEVICE:
+               vm_srb->data_in = READ_TYPE;
+               break;
+       default:
+               vm_srb->data_in = UNKNOWN_TYPE;
+               break;
+       }
+
+
+       vm_srb->port_number = host_dev->port;
+       vm_srb->path_id = scmnd->device->channel;
+       vm_srb->target_id = scmnd->device->id;
+       vm_srb->lun = scmnd->device->lun;
+
+       vm_srb->cdb_length = scmnd->cmd_len;
+
+       memcpy(vm_srb->cdb, scmnd->cmnd, vm_srb->cdb_length);
+
+       cmd_request->sense_buffer = scmnd->sense_buffer;
+
+
+       cmd_request->data_buffer.len = scsi_bufflen(scmnd);
+       if (scsi_sg_count(scmnd)) {
+               sgl = (struct scatterlist *)scsi_sglist(scmnd);
+               sg_count = scsi_sg_count(scmnd);
+
+               /* check if we need to bounce the sgl */
+               if (do_bounce_buffer(sgl, scsi_sg_count(scmnd)) != -1) {
+                       cmd_request->bounce_sgl =
+                               create_bounce_buffer(sgl, scsi_sg_count(scmnd),
+                                                    scsi_bufflen(scmnd),
+                                                    vm_srb->data_in);
+                       if (!cmd_request->bounce_sgl) {
+                               ret = SCSI_MLQUEUE_HOST_BUSY;
+                               goto queue_error;
+                       }
+
+                       cmd_request->bounce_sgl_count =
+                               ALIGN(scsi_bufflen(scmnd), PAGE_SIZE) >>
+                                       PAGE_SHIFT;
+
+                       if (vm_srb->data_in == WRITE_TYPE)
+                               copy_to_bounce_buffer(sgl,
+                                       cmd_request->bounce_sgl,
+                                       scsi_sg_count(scmnd));
+
+                       sgl = cmd_request->bounce_sgl;
+                       sg_count = cmd_request->bounce_sgl_count;
+               }
+
+               cmd_request->data_buffer.offset = sgl[0].offset;
+
+               for (i = 0; i < sg_count; i++)
+                       cmd_request->data_buffer.pfn_array[i] =
+                               page_to_pfn(sg_page((&sgl[i])));
+
+       } else if (scsi_sglist(scmnd)) {
+               cmd_request->data_buffer.offset =
+                       virt_to_phys(scsi_sglist(scmnd)) & (PAGE_SIZE-1);
+               cmd_request->data_buffer.pfn_array[0] =
+                       virt_to_phys(scsi_sglist(scmnd)) >> PAGE_SHIFT;
+       }
+
+       /* Invokes the vsc to start an IO */
+       ret = storvsc_do_io(dev, cmd_request);
+
+       if (ret == -EAGAIN) {
+               /* no more space */
+
+               if (cmd_request->bounce_sgl_count) {
+                       destroy_bounce_buffer(cmd_request->bounce_sgl,
+                                       cmd_request->bounce_sgl_count);
+
+                       ret = SCSI_MLQUEUE_DEVICE_BUSY;
+                       goto queue_error;
+               }
+       }
+
+       return 0;
+
+queue_error:
+       mempool_free(cmd_request, memp->request_mempool);
+       scmnd->host_scribble = NULL;
+       return ret;
+}
+
+static struct scsi_host_template scsi_driver = {
+       .module =               THIS_MODULE,
+       .name =                 "storvsc_host_t",
+       .bios_param =           storvsc_get_chs,
+       .queuecommand =         storvsc_queuecommand,
+       .eh_host_reset_handler =        storvsc_host_reset_handler,
+       .slave_alloc =          storvsc_device_alloc,
+       .slave_destroy =        storvsc_device_destroy,
+       .slave_configure =      storvsc_device_configure,
+       .cmd_per_lun =          1,
+       /* 64 max_queue * 1 target */
+       .can_queue =            STORVSC_MAX_IO_REQUESTS*STORVSC_MAX_TARGETS,
+       .this_id =              -1,
+       /* no use setting to 0 since ll_blk_rw reset it to 1 */
+       /* currently 32 */
+       .sg_tablesize =         MAX_MULTIPAGE_BUFFER_COUNT,
+       .use_clustering =       DISABLE_CLUSTERING,
+       /* Make sure we dont get a sg segment crosses a page boundary */
+       .dma_boundary =         PAGE_SIZE-1,
+};
+
+enum {
+       SCSI_GUID,
+       IDE_GUID,
+};
+
+static const struct hv_vmbus_device_id id_table[] = {
+       /* SCSI guid */
+       { VMBUS_DEVICE(0xd9, 0x63, 0x61, 0xba, 0xa1, 0x04, 0x29, 0x4d,
+                      0xb6, 0x05, 0x72, 0xe2, 0xff, 0xb1, 0xdc, 0x7f)
+         .driver_data = SCSI_GUID },
+       /* IDE guid */
+       { VMBUS_DEVICE(0x32, 0x26, 0x41, 0x32, 0xcb, 0x86, 0xa2, 0x44,
+                      0x9b, 0x5c, 0x50, 0xd1, 0x41, 0x73, 0x54, 0xf5)
+         .driver_data = IDE_GUID },
+       { },
+};
+
+MODULE_DEVICE_TABLE(vmbus, id_table);
+
+static int storvsc_probe(struct hv_device *device,
+                       const struct hv_vmbus_device_id *dev_id)
+{
+       int ret;
+       struct Scsi_Host *host;
+       struct hv_host_device *host_dev;
+       bool dev_is_ide = ((dev_id->driver_data == IDE_GUID) ? true : false);
+       int target = 0;
+       struct storvsc_device *stor_device;
+
+       host = scsi_host_alloc(&scsi_driver,
+                              sizeof(struct hv_host_device));
+       if (!host)
+               return -ENOMEM;
+
+       host_dev = shost_priv(host);
+       memset(host_dev, 0, sizeof(struct hv_host_device));
+
+       host_dev->port = host->host_no;
+       host_dev->dev = device;
+
+
+       stor_device = kzalloc(sizeof(struct storvsc_device), GFP_KERNEL);
+       if (!stor_device) {
+               ret = -ENOMEM;
+               goto err_out0;
+       }
+
+       stor_device->destroy = false;
+       init_waitqueue_head(&stor_device->waiting_to_drain);
+       stor_device->device = device;
+       stor_device->host = host;
+       hv_set_drvdata(device, stor_device);
+
+       stor_device->port_number = host->host_no;
+       ret = storvsc_connect_to_vsp(device, storvsc_ringbuffer_size);
+       if (ret)
+               goto err_out1;
+
+       host_dev->path = stor_device->path_id;
+       host_dev->target = stor_device->target_id;
+
+       /* max # of devices per target */
+       host->max_lun = STORVSC_MAX_LUNS_PER_TARGET;
+       /* max # of targets per channel */
+       host->max_id = STORVSC_MAX_TARGETS;
+       /* max # of channels */
+       host->max_channel = STORVSC_MAX_CHANNELS - 1;
+       /* max cmd length */
+       host->max_cmd_len = STORVSC_MAX_CMD_LEN;
+
+       /* Register the HBA and start the scsi bus scan */
+       ret = scsi_add_host(host, &device->device);
+       if (ret != 0)
+               goto err_out2;
+
+       if (!dev_is_ide) {
+               scsi_scan_host(host);
+       } else {
+               target = (device->dev_instance.b[5] << 8 |
+                        device->dev_instance.b[4]);
+               ret = scsi_add_device(host, 0, target, 0);
+               if (ret) {
+                       scsi_remove_host(host);
+                       goto err_out2;
+               }
+       }
+       return 0;
+
+err_out2:
+       /*
+        * Once we have connected with the host, we would need to
+        * to invoke storvsc_dev_remove() to rollback this state and
+        * this call also frees up the stor_device; hence the jump around
+        * err_out1 label.
+        */
+       storvsc_dev_remove(device);
+       goto err_out0;
+
+err_out1:
+       kfree(stor_device);
+
+err_out0:
+       scsi_host_put(host);
+       return ret;
+}
+
+static int storvsc_remove(struct hv_device *dev)
+{
+       struct storvsc_device *stor_device = hv_get_drvdata(dev);
+       struct Scsi_Host *host = stor_device->host;
+
+       scsi_remove_host(host);
+       storvsc_dev_remove(dev);
+       scsi_host_put(host);
+
+       return 0;
+}
+
+static struct hv_driver storvsc_drv = {
+       .name = KBUILD_MODNAME,
+       .id_table = id_table,
+       .probe = storvsc_probe,
+       .remove = storvsc_remove,
+};
+
+static int __init storvsc_drv_init(void)
+{
+       u32 max_outstanding_req_per_channel;
+
+       /*
+        * Divide the ring buffer data size (which is 1 page less
+        * than the ring buffer size since that page is reserved for
+        * the ring buffer indices) by the max request size (which is
+        * vmbus_channel_packet_multipage_buffer + struct vstor_packet + u64)
+        */
+       max_outstanding_req_per_channel =
+               ((storvsc_ringbuffer_size - PAGE_SIZE) /
+               ALIGN(MAX_MULTIPAGE_BUFFER_PACKET +
+               sizeof(struct vstor_packet) + sizeof(u64),
+               sizeof(u64)));
+
+       if (max_outstanding_req_per_channel <
+           STORVSC_MAX_IO_REQUESTS)
+               return -EINVAL;
+
+       return vmbus_driver_register(&storvsc_drv);
+}
+
+static void __exit storvsc_drv_exit(void)
+{
+       vmbus_driver_unregister(&storvsc_drv);
+}
+
+MODULE_LICENSE("GPL");
+MODULE_VERSION(HV_DRV_VERSION);
+MODULE_DESCRIPTION("Microsoft Hyper-V virtual storage driver");
+module_init(storvsc_drv_init);
+module_exit(storvsc_drv_exit);
index fae2e242025c60e0e42cae32452dd0460fd3fc1d..b5a311b9fe372e3d5ce78aaa39e6456f8b7471d1 100644 (file)
@@ -76,8 +76,6 @@ source "drivers/staging/vt6655/Kconfig"
 
 source "drivers/staging/vt6656/Kconfig"
 
-source "drivers/staging/hv/Kconfig"
-
 source "drivers/staging/vme/Kconfig"
 
 source "drivers/staging/sep/Kconfig"
index 84a6db25cc82f351ef335ed41e7e00a0296a8223..4908e4625e944bda180a7c9c5b00f8323320a2fd 100644 (file)
@@ -29,7 +29,6 @@ obj-$(CONFIG_USB_SERIAL_QUATECH_USB2) += quatech_usb2/
 obj-$(CONFIG_OCTEON_ETHERNET)  += octeon/
 obj-$(CONFIG_VT6655)           += vt6655/
 obj-$(CONFIG_VT6656)           += vt6656/
-obj-$(CONFIG_HYPERV)           += hv/
 obj-$(CONFIG_VME_BUS)          += vme/
 obj-$(CONFIG_DX_SEP)            += sep/
 obj-$(CONFIG_IIO)              += iio/
diff --git a/drivers/staging/hv/Kconfig b/drivers/staging/hv/Kconfig
deleted file mode 100644 (file)
index 60ac479..0000000
+++ /dev/null
@@ -1,5 +0,0 @@
-config HYPERV_STORAGE
-       tristate "Microsoft Hyper-V virtual storage driver"
-       depends on HYPERV && SCSI
-       help
-        Select this option to enable the Hyper-V virtual storage driver.
diff --git a/drivers/staging/hv/Makefile b/drivers/staging/hv/Makefile
deleted file mode 100644 (file)
index af95a6b..0000000
+++ /dev/null
@@ -1,3 +0,0 @@
-obj-$(CONFIG_HYPERV_STORAGE)   += hv_storvsc.o
-
-hv_storvsc-y := storvsc_drv.o
diff --git a/drivers/staging/hv/TODO b/drivers/staging/hv/TODO
deleted file mode 100644 (file)
index dea7d92..0000000
+++ /dev/null
@@ -1,5 +0,0 @@
-TODO:
-       - audit the scsi driver
-
-Please send patches for this code to Greg Kroah-Hartman <gregkh@suse.de>,
-Haiyang Zhang <haiyangz@microsoft.com>, and K. Y. Srinivasan <kys@microsoft.com>
diff --git a/drivers/staging/hv/storvsc_drv.c b/drivers/staging/hv/storvsc_drv.c
deleted file mode 100644 (file)
index 695ffc3..0000000
+++ /dev/null
@@ -1,1548 +0,0 @@
-/*
- * Copyright (c) 2009, Microsoft Corporation.
- *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms and conditions of the GNU General Public License,
- * version 2, as published by the Free Software Foundation.
- *
- * This program is distributed in the hope 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
- * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
- * Place - Suite 330, Boston, MA 02111-1307 USA.
- *
- * Authors:
- *   Haiyang Zhang <haiyangz@microsoft.com>
- *   Hank Janssen  <hjanssen@microsoft.com>
- *   K. Y. Srinivasan <kys@microsoft.com>
- */
-
-#include <linux/kernel.h>
-#include <linux/wait.h>
-#include <linux/sched.h>
-#include <linux/completion.h>
-#include <linux/string.h>
-#include <linux/mm.h>
-#include <linux/delay.h>
-#include <linux/init.h>
-#include <linux/slab.h>
-#include <linux/module.h>
-#include <linux/device.h>
-#include <linux/hyperv.h>
-#include <linux/mempool.h>
-#include <scsi/scsi.h>
-#include <scsi/scsi_cmnd.h>
-#include <scsi/scsi_host.h>
-#include <scsi/scsi_device.h>
-#include <scsi/scsi_tcq.h>
-#include <scsi/scsi_eh.h>
-#include <scsi/scsi_devinfo.h>
-#include <scsi/scsi_dbg.h>
-
-/*
- * All wire protocol details (storage protocol between the guest and the host)
- * are consolidated here.
- *
- * Begin protocol definitions.
- */
-
-/*
- * Version history:
- * V1 Beta: 0.1
- * V1 RC < 2008/1/31: 1.0
- * V1 RC > 2008/1/31:  2.0
- * Win7: 4.2
- */
-
-#define VMSTOR_CURRENT_MAJOR  4
-#define VMSTOR_CURRENT_MINOR  2
-
-
-/*  Packet structure describing virtual storage requests. */
-enum vstor_packet_operation {
-       VSTOR_OPERATION_COMPLETE_IO             = 1,
-       VSTOR_OPERATION_REMOVE_DEVICE           = 2,
-       VSTOR_OPERATION_EXECUTE_SRB             = 3,
-       VSTOR_OPERATION_RESET_LUN               = 4,
-       VSTOR_OPERATION_RESET_ADAPTER           = 5,
-       VSTOR_OPERATION_RESET_BUS               = 6,
-       VSTOR_OPERATION_BEGIN_INITIALIZATION    = 7,
-       VSTOR_OPERATION_END_INITIALIZATION      = 8,
-       VSTOR_OPERATION_QUERY_PROTOCOL_VERSION  = 9,
-       VSTOR_OPERATION_QUERY_PROPERTIES        = 10,
-       VSTOR_OPERATION_ENUMERATE_BUS           = 11,
-       VSTOR_OPERATION_MAXIMUM                 = 11
-};
-
-/*
- * Platform neutral description of a scsi request -
- * this remains the same across the write regardless of 32/64 bit
- * note: it's patterned off the SCSI_PASS_THROUGH structure
- */
-#define STORVSC_MAX_CMD_LEN                    0x10
-#define STORVSC_SENSE_BUFFER_SIZE              0x12
-#define STORVSC_MAX_BUF_LEN_WITH_PADDING       0x14
-
-struct vmscsi_request {
-       u16 length;
-       u8 srb_status;
-       u8 scsi_status;
-
-       u8  port_number;
-       u8  path_id;
-       u8  target_id;
-       u8  lun;
-
-       u8  cdb_length;
-       u8  sense_info_length;
-       u8  data_in;
-       u8  reserved;
-
-       u32 data_transfer_length;
-
-       union {
-               u8 cdb[STORVSC_MAX_CMD_LEN];
-               u8 sense_data[STORVSC_SENSE_BUFFER_SIZE];
-               u8 reserved_array[STORVSC_MAX_BUF_LEN_WITH_PADDING];
-       };
-} __attribute((packed));
-
-
-/*
- * This structure is sent during the intialization phase to get the different
- * properties of the channel.
- */
-struct vmstorage_channel_properties {
-       u16 protocol_version;
-       u8  path_id;
-       u8 target_id;
-
-       /* Note: port number is only really known on the client side */
-       u32  port_number;
-       u32  flags;
-       u32   max_transfer_bytes;
-
-       /*
-        * This id is unique for each channel and will correspond with
-        * vendor specific data in the inquiry data.
-        */
-
-       u64  unique_id;
-} __packed;
-
-/*  This structure is sent during the storage protocol negotiations. */
-struct vmstorage_protocol_version {
-       /* Major (MSW) and minor (LSW) version numbers. */
-       u16 major_minor;
-
-       /*
-        * Revision number is auto-incremented whenever this file is changed
-        * (See FILL_VMSTOR_REVISION macro above).  Mismatch does not
-        * definitely indicate incompatibility--but it does indicate mismatched
-        * builds.
-        * This is only used on the windows side. Just set it to 0.
-        */
-       u16 revision;
-} __packed;
-
-/* Channel Property Flags */
-#define STORAGE_CHANNEL_REMOVABLE_FLAG         0x1
-#define STORAGE_CHANNEL_EMULATED_IDE_FLAG      0x2
-
-struct vstor_packet {
-       /* Requested operation type */
-       enum vstor_packet_operation operation;
-
-       /*  Flags - see below for values */
-       u32 flags;
-
-       /* Status of the request returned from the server side. */
-       u32 status;
-
-       /* Data payload area */
-       union {
-               /*
-                * Structure used to forward SCSI commands from the
-                * client to the server.
-                */
-               struct vmscsi_request vm_srb;
-
-               /* Structure used to query channel properties. */
-               struct vmstorage_channel_properties storage_channel_properties;
-
-               /* Used during version negotiations. */
-               struct vmstorage_protocol_version version;
-       };
-} __packed;
-
-/*
- * Packet Flags:
- *
- * This flag indicates that the server should send back a completion for this
- * packet.
- */
-
-#define REQUEST_COMPLETION_FLAG        0x1
-
-/* Matches Windows-end */
-enum storvsc_request_type {
-       WRITE_TYPE = 0,
-       READ_TYPE,
-       UNKNOWN_TYPE,
-};
-
-/*
- * SRB status codes and masks; a subset of the codes used here.
- */
-
-#define SRB_STATUS_AUTOSENSE_VALID     0x80
-#define SRB_STATUS_INVALID_LUN 0x20
-#define SRB_STATUS_SUCCESS     0x01
-#define SRB_STATUS_ERROR       0x04
-
-/*
- * This is the end of Protocol specific defines.
- */
-
-
-/*
- * We setup a mempool to allocate request structures for this driver
- * on a per-lun basis. The following define specifies the number of
- * elements in the pool.
- */
-
-#define STORVSC_MIN_BUF_NR                             64
-static int storvsc_ringbuffer_size = (20 * PAGE_SIZE);
-
-module_param(storvsc_ringbuffer_size, int, S_IRUGO);
-MODULE_PARM_DESC(storvsc_ringbuffer_size, "Ring buffer size (bytes)");
-
-#define STORVSC_MAX_IO_REQUESTS                                128
-
-/*
- * In Hyper-V, each port/path/target maps to 1 scsi host adapter.  In
- * reality, the path/target is not used (ie always set to 0) so our
- * scsi host adapter essentially has 1 bus with 1 target that contains
- * up to 256 luns.
- */
-#define STORVSC_MAX_LUNS_PER_TARGET                    64
-#define STORVSC_MAX_TARGETS                            1
-#define STORVSC_MAX_CHANNELS                           1
-
-
-
-struct storvsc_cmd_request {
-       struct list_head entry;
-       struct scsi_cmnd *cmd;
-
-       unsigned int bounce_sgl_count;
-       struct scatterlist *bounce_sgl;
-
-       struct hv_device *device;
-
-       /* Synchronize the request/response if needed */
-       struct completion wait_event;
-
-       unsigned char *sense_buffer;
-       struct hv_multipage_buffer data_buffer;
-       struct vstor_packet vstor_packet;
-};
-
-
-/* A storvsc device is a device object that contains a vmbus channel */
-struct storvsc_device {
-       struct hv_device *device;
-
-       bool     destroy;
-       bool     drain_notify;
-       atomic_t num_outstanding_req;
-       struct Scsi_Host *host;
-
-       wait_queue_head_t waiting_to_drain;
-
-       /*
-        * Each unique Port/Path/Target represents 1 channel ie scsi
-        * controller. In reality, the pathid, targetid is always 0
-        * and the port is set by us
-        */
-       unsigned int port_number;
-       unsigned char path_id;
-       unsigned char target_id;
-
-       /* Used for vsc/vsp channel reset process */
-       struct storvsc_cmd_request init_request;
-       struct storvsc_cmd_request reset_request;
-};
-
-struct stor_mem_pools {
-       struct kmem_cache *request_pool;
-       mempool_t *request_mempool;
-};
-
-struct hv_host_device {
-       struct hv_device *dev;
-       unsigned int port;
-       unsigned char path;
-       unsigned char target;
-};
-
-struct storvsc_scan_work {
-       struct work_struct work;
-       struct Scsi_Host *host;
-       uint lun;
-};
-
-static void storvsc_bus_scan(struct work_struct *work)
-{
-       struct storvsc_scan_work *wrk;
-       int id, order_id;
-
-       wrk = container_of(work, struct storvsc_scan_work, work);
-       for (id = 0; id < wrk->host->max_id; ++id) {
-               if (wrk->host->reverse_ordering)
-                       order_id = wrk->host->max_id - id - 1;
-               else
-                       order_id = id;
-
-               scsi_scan_target(&wrk->host->shost_gendev, 0,
-                               order_id, SCAN_WILD_CARD, 1);
-       }
-       kfree(wrk);
-}
-
-static void storvsc_remove_lun(struct work_struct *work)
-{
-       struct storvsc_scan_work *wrk;
-       struct scsi_device *sdev;
-
-       wrk = container_of(work, struct storvsc_scan_work, work);
-       if (!scsi_host_get(wrk->host))
-               goto done;
-
-       sdev = scsi_device_lookup(wrk->host, 0, 0, wrk->lun);
-
-       if (sdev) {
-               scsi_remove_device(sdev);
-               scsi_device_put(sdev);
-       }
-       scsi_host_put(wrk->host);
-
-done:
-       kfree(wrk);
-}
-
-/*
- * Major/minor macros.  Minor version is in LSB, meaning that earlier flat
- * version numbers will be interpreted as "0.x" (i.e., 1 becomes 0.1).
- */
-
-static inline u16 storvsc_get_version(u8 major, u8 minor)
-{
-       u16 version;
-
-       version = ((major << 8) | minor);
-       return version;
-}
-
-/*
- * We can get incoming messages from the host that are not in response to
- * messages that we have sent out. An example of this would be messages
- * received by the guest to notify dynamic addition/removal of LUNs. To
- * deal with potential race conditions where the driver may be in the
- * midst of being unloaded when we might receive an unsolicited message
- * from the host, we have implemented a mechanism to gurantee sequential
- * consistency:
- *
- * 1) Once the device is marked as being destroyed, we will fail all
- *    outgoing messages.
- * 2) We permit incoming messages when the device is being destroyed,
- *    only to properly account for messages already sent out.
- */
-
-static inline struct storvsc_device *get_out_stor_device(
-                                       struct hv_device *device)
-{
-       struct storvsc_device *stor_device;
-
-       stor_device = hv_get_drvdata(device);
-
-       if (stor_device && stor_device->destroy)
-               stor_device = NULL;
-
-       return stor_device;
-}
-
-
-static inline void storvsc_wait_to_drain(struct storvsc_device *dev)
-{
-       dev->drain_notify = true;
-       wait_event(dev->waiting_to_drain,
-                  atomic_read(&dev->num_outstanding_req) == 0);
-       dev->drain_notify = false;
-}
-
-static inline struct storvsc_device *get_in_stor_device(
-                                       struct hv_device *device)
-{
-       struct storvsc_device *stor_device;
-
-       stor_device = hv_get_drvdata(device);
-
-       if (!stor_device)
-               goto get_in_err;
-
-       /*
-        * If the device is being destroyed; allow incoming
-        * traffic only to cleanup outstanding requests.
-        */
-
-       if (stor_device->destroy  &&
-               (atomic_read(&stor_device->num_outstanding_req) == 0))
-               stor_device = NULL;
-
-get_in_err:
-       return stor_device;
-
-}
-
-static void destroy_bounce_buffer(struct scatterlist *sgl,
-                                 unsigned int sg_count)
-{
-       int i;
-       struct page *page_buf;
-
-       for (i = 0; i < sg_count; i++) {
-               page_buf = sg_page((&sgl[i]));
-               if (page_buf != NULL)
-                       __free_page(page_buf);
-       }
-
-       kfree(sgl);
-}
-
-static int do_bounce_buffer(struct scatterlist *sgl, unsigned int sg_count)
-{
-       int i;
-
-       /* No need to check */
-       if (sg_count < 2)
-               return -1;
-
-       /* We have at least 2 sg entries */
-       for (i = 0; i < sg_count; i++) {
-               if (i == 0) {
-                       /* make sure 1st one does not have hole */
-                       if (sgl[i].offset + sgl[i].length != PAGE_SIZE)
-                               return i;
-               } else if (i == sg_count - 1) {
-                       /* make sure last one does not have hole */
-                       if (sgl[i].offset != 0)
-                               return i;
-               } else {
-                       /* make sure no hole in the middle */
-                       if (sgl[i].length != PAGE_SIZE || sgl[i].offset != 0)
-                               return i;
-               }
-       }
-       return -1;
-}
-
-static struct scatterlist *create_bounce_buffer(struct scatterlist *sgl,
-                                               unsigned int sg_count,
-                                               unsigned int len,
-                                               int write)
-{
-       int i;
-       int num_pages;
-       struct scatterlist *bounce_sgl;
-       struct page *page_buf;
-       unsigned int buf_len = ((write == WRITE_TYPE) ? 0 : PAGE_SIZE);
-
-       num_pages = ALIGN(len, PAGE_SIZE) >> PAGE_SHIFT;
-
-       bounce_sgl = kcalloc(num_pages, sizeof(struct scatterlist), GFP_ATOMIC);
-       if (!bounce_sgl)
-               return NULL;
-
-       for (i = 0; i < num_pages; i++) {
-               page_buf = alloc_page(GFP_ATOMIC);
-               if (!page_buf)
-                       goto cleanup;
-               sg_set_page(&bounce_sgl[i], page_buf, buf_len, 0);
-       }
-
-       return bounce_sgl;
-
-cleanup:
-       destroy_bounce_buffer(bounce_sgl, num_pages);
-       return NULL;
-}
-
-/* Assume the original sgl has enough room */
-static unsigned int copy_from_bounce_buffer(struct scatterlist *orig_sgl,
-                                           struct scatterlist *bounce_sgl,
-                                           unsigned int orig_sgl_count,
-                                           unsigned int bounce_sgl_count)
-{
-       int i;
-       int j = 0;
-       unsigned long src, dest;
-       unsigned int srclen, destlen, copylen;
-       unsigned int total_copied = 0;
-       unsigned long bounce_addr = 0;
-       unsigned long dest_addr = 0;
-       unsigned long flags;
-
-       local_irq_save(flags);
-
-       for (i = 0; i < orig_sgl_count; i++) {
-               dest_addr = (unsigned long)kmap_atomic(sg_page((&orig_sgl[i])),
-                                       KM_IRQ0) + orig_sgl[i].offset;
-               dest = dest_addr;
-               destlen = orig_sgl[i].length;
-
-               if (bounce_addr == 0)
-                       bounce_addr =
-                       (unsigned long)kmap_atomic(sg_page((&bounce_sgl[j])),
-                                                       KM_IRQ0);
-
-               while (destlen) {
-                       src = bounce_addr + bounce_sgl[j].offset;
-                       srclen = bounce_sgl[j].length - bounce_sgl[j].offset;
-
-                       copylen = min(srclen, destlen);
-                       memcpy((void *)dest, (void *)src, copylen);
-
-                       total_copied += copylen;
-                       bounce_sgl[j].offset += copylen;
-                       destlen -= copylen;
-                       dest += copylen;
-
-                       if (bounce_sgl[j].offset == bounce_sgl[j].length) {
-                               /* full */
-                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
-                               j++;
-
-                               /*
-                                * It is possible that the number of elements
-                                * in the bounce buffer may not be equal to
-                                * the number of elements in the original
-                                * scatter list. Handle this correctly.
-                                */
-
-                               if (j == bounce_sgl_count) {
-                                       /*
-                                        * We are done; cleanup and return.
-                                        */
-                                       kunmap_atomic((void *)(dest_addr -
-                                                       orig_sgl[i].offset),
-                                                       KM_IRQ0);
-                                       local_irq_restore(flags);
-                                       return total_copied;
-                               }
-
-                               /* if we need to use another bounce buffer */
-                               if (destlen || i != orig_sgl_count - 1)
-                                       bounce_addr =
-                                       (unsigned long)kmap_atomic(
-                                       sg_page((&bounce_sgl[j])), KM_IRQ0);
-                       } else if (destlen == 0 && i == orig_sgl_count - 1) {
-                               /* unmap the last bounce that is < PAGE_SIZE */
-                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
-                       }
-               }
-
-               kunmap_atomic((void *)(dest_addr - orig_sgl[i].offset),
-                             KM_IRQ0);
-       }
-
-       local_irq_restore(flags);
-
-       return total_copied;
-}
-
-/* Assume the bounce_sgl has enough room ie using the create_bounce_buffer() */
-static unsigned int copy_to_bounce_buffer(struct scatterlist *orig_sgl,
-                                         struct scatterlist *bounce_sgl,
-                                         unsigned int orig_sgl_count)
-{
-       int i;
-       int j = 0;
-       unsigned long src, dest;
-       unsigned int srclen, destlen, copylen;
-       unsigned int total_copied = 0;
-       unsigned long bounce_addr = 0;
-       unsigned long src_addr = 0;
-       unsigned long flags;
-
-       local_irq_save(flags);
-
-       for (i = 0; i < orig_sgl_count; i++) {
-               src_addr = (unsigned long)kmap_atomic(sg_page((&orig_sgl[i])),
-                               KM_IRQ0) + orig_sgl[i].offset;
-               src = src_addr;
-               srclen = orig_sgl[i].length;
-
-               if (bounce_addr == 0)
-                       bounce_addr =
-                       (unsigned long)kmap_atomic(sg_page((&bounce_sgl[j])),
-                                               KM_IRQ0);
-
-               while (srclen) {
-                       /* assume bounce offset always == 0 */
-                       dest = bounce_addr + bounce_sgl[j].length;
-                       destlen = PAGE_SIZE - bounce_sgl[j].length;
-
-                       copylen = min(srclen, destlen);
-                       memcpy((void *)dest, (void *)src, copylen);
-
-                       total_copied += copylen;
-                       bounce_sgl[j].length += copylen;
-                       srclen -= copylen;
-                       src += copylen;
-
-                       if (bounce_sgl[j].length == PAGE_SIZE) {
-                               /* full..move to next entry */
-                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
-                               j++;
-
-                               /* if we need to use another bounce buffer */
-                               if (srclen || i != orig_sgl_count - 1)
-                                       bounce_addr =
-                                       (unsigned long)kmap_atomic(
-                                       sg_page((&bounce_sgl[j])), KM_IRQ0);
-
-                       } else if (srclen == 0 && i == orig_sgl_count - 1) {
-                               /* unmap the last bounce that is < PAGE_SIZE */
-                               kunmap_atomic((void *)bounce_addr, KM_IRQ0);
-                       }
-               }
-
-               kunmap_atomic((void *)(src_addr - orig_sgl[i].offset), KM_IRQ0);
-       }
-
-       local_irq_restore(flags);
-
-       return total_copied;
-}
-
-static int storvsc_channel_init(struct hv_device *device)
-{
-       struct storvsc_device *stor_device;
-       struct storvsc_cmd_request *request;
-       struct vstor_packet *vstor_packet;
-       int ret, t;
-
-       stor_device = get_out_stor_device(device);
-       if (!stor_device)
-               return -ENODEV;
-
-       request = &stor_device->init_request;
-       vstor_packet = &request->vstor_packet;
-
-       /*
-        * Now, initiate the vsc/vsp initialization protocol on the open
-        * channel
-        */
-       memset(request, 0, sizeof(struct storvsc_cmd_request));
-       init_completion(&request->wait_event);
-       vstor_packet->operation = VSTOR_OPERATION_BEGIN_INITIALIZATION;
-       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
-
-       ret = vmbus_sendpacket(device->channel, vstor_packet,
-                              sizeof(struct vstor_packet),
-                              (unsigned long)request,
-                              VM_PKT_DATA_INBAND,
-                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
-       if (ret != 0)
-               goto cleanup;
-
-       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
-       if (t == 0) {
-               ret = -ETIMEDOUT;
-               goto cleanup;
-       }
-
-       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
-           vstor_packet->status != 0)
-               goto cleanup;
-
-
-       /* reuse the packet for version range supported */
-       memset(vstor_packet, 0, sizeof(struct vstor_packet));
-       vstor_packet->operation = VSTOR_OPERATION_QUERY_PROTOCOL_VERSION;
-       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
-
-       vstor_packet->version.major_minor =
-               storvsc_get_version(VMSTOR_CURRENT_MAJOR, VMSTOR_CURRENT_MINOR);
-
-       /*
-        * The revision number is only used in Windows; set it to 0.
-        */
-       vstor_packet->version.revision = 0;
-
-       ret = vmbus_sendpacket(device->channel, vstor_packet,
-                              sizeof(struct vstor_packet),
-                              (unsigned long)request,
-                              VM_PKT_DATA_INBAND,
-                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
-       if (ret != 0)
-               goto cleanup;
-
-       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
-       if (t == 0) {
-               ret = -ETIMEDOUT;
-               goto cleanup;
-       }
-
-       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
-           vstor_packet->status != 0)
-               goto cleanup;
-
-
-       memset(vstor_packet, 0, sizeof(struct vstor_packet));
-       vstor_packet->operation = VSTOR_OPERATION_QUERY_PROPERTIES;
-       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
-       vstor_packet->storage_channel_properties.port_number =
-                                       stor_device->port_number;
-
-       ret = vmbus_sendpacket(device->channel, vstor_packet,
-                              sizeof(struct vstor_packet),
-                              (unsigned long)request,
-                              VM_PKT_DATA_INBAND,
-                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
-
-       if (ret != 0)
-               goto cleanup;
-
-       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
-       if (t == 0) {
-               ret = -ETIMEDOUT;
-               goto cleanup;
-       }
-
-       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
-           vstor_packet->status != 0)
-               goto cleanup;
-
-       stor_device->path_id = vstor_packet->storage_channel_properties.path_id;
-       stor_device->target_id
-               = vstor_packet->storage_channel_properties.target_id;
-
-       memset(vstor_packet, 0, sizeof(struct vstor_packet));
-       vstor_packet->operation = VSTOR_OPERATION_END_INITIALIZATION;
-       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
-
-       ret = vmbus_sendpacket(device->channel, vstor_packet,
-                              sizeof(struct vstor_packet),
-                              (unsigned long)request,
-                              VM_PKT_DATA_INBAND,
-                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
-
-       if (ret != 0)
-               goto cleanup;
-
-       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
-       if (t == 0) {
-               ret = -ETIMEDOUT;
-               goto cleanup;
-       }
-
-       if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
-           vstor_packet->status != 0)
-               goto cleanup;
-
-
-cleanup:
-       return ret;
-}
-
-
-static void storvsc_command_completion(struct storvsc_cmd_request *cmd_request)
-{
-       struct scsi_cmnd *scmnd = cmd_request->cmd;
-       struct hv_host_device *host_dev = shost_priv(scmnd->device->host);
-       void (*scsi_done_fn)(struct scsi_cmnd *);
-       struct scsi_sense_hdr sense_hdr;
-       struct vmscsi_request *vm_srb;
-       struct storvsc_scan_work *wrk;
-       struct stor_mem_pools *memp = scmnd->device->hostdata;
-
-       vm_srb = &cmd_request->vstor_packet.vm_srb;
-       if (cmd_request->bounce_sgl_count) {
-               if (vm_srb->data_in == READ_TYPE)
-                       copy_from_bounce_buffer(scsi_sglist(scmnd),
-                                       cmd_request->bounce_sgl,
-                                       scsi_sg_count(scmnd),
-                                       cmd_request->bounce_sgl_count);
-               destroy_bounce_buffer(cmd_request->bounce_sgl,
-                                       cmd_request->bounce_sgl_count);
-       }
-
-       /*
-        * If there is an error; offline the device since all
-        * error recovery strategies would have already been
-        * deployed on the host side.
-        */
-       if (vm_srb->srb_status == SRB_STATUS_ERROR)
-               scmnd->result = DID_TARGET_FAILURE << 16;
-       else
-               scmnd->result = vm_srb->scsi_status;
-
-       /*
-        * If the LUN is invalid; remove the device.
-        */
-       if (vm_srb->srb_status == SRB_STATUS_INVALID_LUN) {
-               struct storvsc_device *stor_dev;
-               struct hv_device *dev = host_dev->dev;
-               struct Scsi_Host *host;
-
-               stor_dev = get_in_stor_device(dev);
-               host = stor_dev->host;
-
-               wrk = kmalloc(sizeof(struct storvsc_scan_work),
-                               GFP_ATOMIC);
-               if (!wrk) {
-                       scmnd->result = DID_TARGET_FAILURE << 16;
-               } else {
-                       wrk->host = host;
-                       wrk->lun = vm_srb->lun;
-                       INIT_WORK(&wrk->work, storvsc_remove_lun);
-                       schedule_work(&wrk->work);
-               }
-       }
-
-       if (scmnd->result) {
-               if (scsi_normalize_sense(scmnd->sense_buffer,
-                               SCSI_SENSE_BUFFERSIZE, &sense_hdr))
-                       scsi_print_sense_hdr("storvsc", &sense_hdr);
-       }
-
-       scsi_set_resid(scmnd,
-               cmd_request->data_buffer.len -
-               vm_srb->data_transfer_length);
-
-       scsi_done_fn = scmnd->scsi_done;
-
-       scmnd->host_scribble = NULL;
-       scmnd->scsi_done = NULL;
-
-       scsi_done_fn(scmnd);
-
-       mempool_free(cmd_request, memp->request_mempool);
-}
-
-static void storvsc_on_io_completion(struct hv_device *device,
-                                 struct vstor_packet *vstor_packet,
-                                 struct storvsc_cmd_request *request)
-{
-       struct storvsc_device *stor_device;
-       struct vstor_packet *stor_pkt;
-
-       stor_device = hv_get_drvdata(device);
-       stor_pkt = &request->vstor_packet;
-
-       /*
-        * The current SCSI handling on the host side does
-        * not correctly handle:
-        * INQUIRY command with page code parameter set to 0x80
-        * MODE_SENSE command with cmd[2] == 0x1c
-        *
-        * Setup srb and scsi status so this won't be fatal.
-        * We do this so we can distinguish truly fatal failues
-        * (srb status == 0x4) and off-line the device in that case.
-        */
-
-       if ((stor_pkt->vm_srb.cdb[0] == INQUIRY) ||
-          (stor_pkt->vm_srb.cdb[0] == MODE_SENSE)) {
-               vstor_packet->vm_srb.scsi_status = 0;
-               vstor_packet->vm_srb.srb_status = SRB_STATUS_SUCCESS;
-       }
-
-
-       /* Copy over the status...etc */
-       stor_pkt->vm_srb.scsi_status = vstor_packet->vm_srb.scsi_status;
-       stor_pkt->vm_srb.srb_status = vstor_packet->vm_srb.srb_status;
-       stor_pkt->vm_srb.sense_info_length =
-       vstor_packet->vm_srb.sense_info_length;
-
-       if (vstor_packet->vm_srb.scsi_status != 0 ||
-               vstor_packet->vm_srb.srb_status != SRB_STATUS_SUCCESS){
-               dev_warn(&device->device,
-                        "cmd 0x%x scsi status 0x%x srb status 0x%x\n",
-                        stor_pkt->vm_srb.cdb[0],
-                        vstor_packet->vm_srb.scsi_status,
-                        vstor_packet->vm_srb.srb_status);
-       }
-
-       if ((vstor_packet->vm_srb.scsi_status & 0xFF) == 0x02) {
-               /* CHECK_CONDITION */
-               if (vstor_packet->vm_srb.srb_status &
-                       SRB_STATUS_AUTOSENSE_VALID) {
-                       /* autosense data available */
-                       dev_warn(&device->device,
-                                "stor pkt %p autosense data valid - len %d\n",
-                                request,
-                                vstor_packet->vm_srb.sense_info_length);
-
-                       memcpy(request->sense_buffer,
-                              vstor_packet->vm_srb.sense_data,
-                              vstor_packet->vm_srb.sense_info_length);
-
-               }
-       }
-
-       stor_pkt->vm_srb.data_transfer_length =
-       vstor_packet->vm_srb.data_transfer_length;
-
-       storvsc_command_completion(request);
-
-       if (atomic_dec_and_test(&stor_device->num_outstanding_req) &&
-               stor_device->drain_notify)
-               wake_up(&stor_device->waiting_to_drain);
-
-
-}
-
-static void storvsc_on_receive(struct hv_device *device,
-                            struct vstor_packet *vstor_packet,
-                            struct storvsc_cmd_request *request)
-{
-       struct storvsc_scan_work *work;
-       struct storvsc_device *stor_device;
-
-       switch (vstor_packet->operation) {
-       case VSTOR_OPERATION_COMPLETE_IO:
-               storvsc_on_io_completion(device, vstor_packet, request);
-               break;
-
-       case VSTOR_OPERATION_REMOVE_DEVICE:
-       case VSTOR_OPERATION_ENUMERATE_BUS:
-               stor_device = get_in_stor_device(device);
-               work = kmalloc(sizeof(struct storvsc_scan_work), GFP_ATOMIC);
-               if (!work)
-                       return;
-
-               INIT_WORK(&work->work, storvsc_bus_scan);
-               work->host = stor_device->host;
-               schedule_work(&work->work);
-               break;
-
-       default:
-               break;
-       }
-}
-
-static void storvsc_on_channel_callback(void *context)
-{
-       struct hv_device *device = (struct hv_device *)context;
-       struct storvsc_device *stor_device;
-       u32 bytes_recvd;
-       u64 request_id;
-       unsigned char packet[ALIGN(sizeof(struct vstor_packet), 8)];
-       struct storvsc_cmd_request *request;
-       int ret;
-
-
-       stor_device = get_in_stor_device(device);
-       if (!stor_device)
-               return;
-
-       do {
-               ret = vmbus_recvpacket(device->channel, packet,
-                                      ALIGN(sizeof(struct vstor_packet), 8),
-                                      &bytes_recvd, &request_id);
-               if (ret == 0 && bytes_recvd > 0) {
-
-                       request = (struct storvsc_cmd_request *)
-                                       (unsigned long)request_id;
-
-                       if ((request == &stor_device->init_request) ||
-                           (request == &stor_device->reset_request)) {
-
-                               memcpy(&request->vstor_packet, packet,
-                                      sizeof(struct vstor_packet));
-                               complete(&request->wait_event);
-                       } else {
-                               storvsc_on_receive(device,
-                                               (struct vstor_packet *)packet,
-                                               request);
-                       }
-               } else {
-                       break;
-               }
-       } while (1);
-
-       return;
-}
-
-static int storvsc_connect_to_vsp(struct hv_device *device, u32 ring_size)
-{
-       struct vmstorage_channel_properties props;
-       int ret;
-
-       memset(&props, 0, sizeof(struct vmstorage_channel_properties));
-
-       ret = vmbus_open(device->channel,
-                        ring_size,
-                        ring_size,
-                        (void *)&props,
-                        sizeof(struct vmstorage_channel_properties),
-                        storvsc_on_channel_callback, device);
-
-       if (ret != 0)
-               return ret;
-
-       ret = storvsc_channel_init(device);
-
-       return ret;
-}
-
-static int storvsc_dev_remove(struct hv_device *device)
-{
-       struct storvsc_device *stor_device;
-       unsigned long flags;
-
-       stor_device = hv_get_drvdata(device);
-
-       spin_lock_irqsave(&device->channel->inbound_lock, flags);
-       stor_device->destroy = true;
-       spin_unlock_irqrestore(&device->channel->inbound_lock, flags);
-
-       /*
-        * At this point, all outbound traffic should be disable. We
-        * only allow inbound traffic (responses) to proceed so that
-        * outstanding requests can be completed.
-        */
-
-       storvsc_wait_to_drain(stor_device);
-
-       /*
-        * Since we have already drained, we don't need to busy wait
-        * as was done in final_release_stor_device()
-        * Note that we cannot set the ext pointer to NULL until
-        * we have drained - to drain the outgoing packets, we need to
-        * allow incoming packets.
-        */
-       spin_lock_irqsave(&device->channel->inbound_lock, flags);
-       hv_set_drvdata(device, NULL);
-       spin_unlock_irqrestore(&device->channel->inbound_lock, flags);
-
-       /* Close the channel */
-       vmbus_close(device->channel);
-
-       kfree(stor_device);
-       return 0;
-}
-
-static int storvsc_do_io(struct hv_device *device,
-                             struct storvsc_cmd_request *request)
-{
-       struct storvsc_device *stor_device;
-       struct vstor_packet *vstor_packet;
-       int ret = 0;
-
-       vstor_packet = &request->vstor_packet;
-       stor_device = get_out_stor_device(device);
-
-       if (!stor_device)
-               return -ENODEV;
-
-
-       request->device  = device;
-
-
-       vstor_packet->flags |= REQUEST_COMPLETION_FLAG;
-
-       vstor_packet->vm_srb.length = sizeof(struct vmscsi_request);
-
-
-       vstor_packet->vm_srb.sense_info_length = STORVSC_SENSE_BUFFER_SIZE;
-
-
-       vstor_packet->vm_srb.data_transfer_length =
-       request->data_buffer.len;
-
-       vstor_packet->operation = VSTOR_OPERATION_EXECUTE_SRB;
-
-       if (request->data_buffer.len) {
-               ret = vmbus_sendpacket_multipagebuffer(device->channel,
-                               &request->data_buffer,
-                               vstor_packet,
-                               sizeof(struct vstor_packet),
-                               (unsigned long)request);
-       } else {
-               ret = vmbus_sendpacket(device->channel, vstor_packet,
-                              sizeof(struct vstor_packet),
-                              (unsigned long)request,
-                              VM_PKT_DATA_INBAND,
-                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
-       }
-
-       if (ret != 0)
-               return ret;
-
-       atomic_inc(&stor_device->num_outstanding_req);
-
-       return ret;
-}
-
-static int storvsc_device_alloc(struct scsi_device *sdevice)
-{
-       struct stor_mem_pools *memp;
-       int number = STORVSC_MIN_BUF_NR;
-
-       memp = kzalloc(sizeof(struct stor_mem_pools), GFP_KERNEL);
-       if (!memp)
-               return -ENOMEM;
-
-       memp->request_pool =
-               kmem_cache_create(dev_name(&sdevice->sdev_dev),
-                               sizeof(struct storvsc_cmd_request), 0,
-                               SLAB_HWCACHE_ALIGN, NULL);
-
-       if (!memp->request_pool)
-               goto err0;
-
-       memp->request_mempool = mempool_create(number, mempool_alloc_slab,
-                                               mempool_free_slab,
-                                               memp->request_pool);
-
-       if (!memp->request_mempool)
-               goto err1;
-
-       sdevice->hostdata = memp;
-
-       return 0;
-
-err1:
-       kmem_cache_destroy(memp->request_pool);
-
-err0:
-       kfree(memp);
-       return -ENOMEM;
-}
-
-static void storvsc_device_destroy(struct scsi_device *sdevice)
-{
-       struct stor_mem_pools *memp = sdevice->hostdata;
-
-       mempool_destroy(memp->request_mempool);
-       kmem_cache_destroy(memp->request_pool);
-       kfree(memp);
-       sdevice->hostdata = NULL;
-}
-
-static int storvsc_device_configure(struct scsi_device *sdevice)
-{
-       scsi_adjust_queue_depth(sdevice, MSG_SIMPLE_TAG,
-                               STORVSC_MAX_IO_REQUESTS);
-
-       blk_queue_max_segment_size(sdevice->request_queue, PAGE_SIZE);
-
-       blk_queue_bounce_limit(sdevice->request_queue, BLK_BOUNCE_ANY);
-
-       return 0;
-}
-
-static int storvsc_get_chs(struct scsi_device *sdev, struct block_device * bdev,
-                          sector_t capacity, int *info)
-{
-       sector_t nsect = capacity;
-       sector_t cylinders = nsect;
-       int heads, sectors_pt;
-
-       /*
-        * We are making up these values; let us keep it simple.
-        */
-       heads = 0xff;
-       sectors_pt = 0x3f;      /* Sectors per track */
-       sector_div(cylinders, heads * sectors_pt);
-       if ((sector_t)(cylinders + 1) * heads * sectors_pt < nsect)
-               cylinders = 0xffff;
-
-       info[0] = heads;
-       info[1] = sectors_pt;
-       info[2] = (int)cylinders;
-
-       return 0;
-}
-
-static int storvsc_host_reset_handler(struct scsi_cmnd *scmnd)
-{
-       struct hv_host_device *host_dev = shost_priv(scmnd->device->host);
-       struct hv_device *device = host_dev->dev;
-
-       struct storvsc_device *stor_device;
-       struct storvsc_cmd_request *request;
-       struct vstor_packet *vstor_packet;
-       int ret, t;
-
-
-       stor_device = get_out_stor_device(device);
-       if (!stor_device)
-               return FAILED;
-
-       request = &stor_device->reset_request;
-       vstor_packet = &request->vstor_packet;
-
-       init_completion(&request->wait_event);
-
-       vstor_packet->operation = VSTOR_OPERATION_RESET_BUS;
-       vstor_packet->flags = REQUEST_COMPLETION_FLAG;
-       vstor_packet->vm_srb.path_id = stor_device->path_id;
-
-       ret = vmbus_sendpacket(device->channel, vstor_packet,
-                              sizeof(struct vstor_packet),
-                              (unsigned long)&stor_device->reset_request,
-                              VM_PKT_DATA_INBAND,
-                              VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
-       if (ret != 0)
-               return FAILED;
-
-       t = wait_for_completion_timeout(&request->wait_event, 5*HZ);
-       if (t == 0)
-               return TIMEOUT_ERROR;
-
-
-       /*
-        * At this point, all outstanding requests in the adapter
-        * should have been flushed out and return to us
-        */
-
-       return SUCCESS;
-}
-
-static bool storvsc_scsi_cmd_ok(struct scsi_cmnd *scmnd)
-{
-       bool allowed = true;
-       u8 scsi_op = scmnd->cmnd[0];
-
-       switch (scsi_op) {
-       /*
-        * smartd sends this command and the host does not handle
-        * this. So, don't send it.
-        */
-       case SET_WINDOW:
-               scmnd->result = ILLEGAL_REQUEST << 16;
-               allowed = false;
-               break;
-       default:
-               break;
-       }
-       return allowed;
-}
-
-static int storvsc_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scmnd)
-{
-       int ret;
-       struct hv_host_device *host_dev = shost_priv(host);
-       struct hv_device *dev = host_dev->dev;
-       struct storvsc_cmd_request *cmd_request;
-       unsigned int request_size = 0;
-       int i;
-       struct scatterlist *sgl;
-       unsigned int sg_count = 0;
-       struct vmscsi_request *vm_srb;
-       struct stor_mem_pools *memp = scmnd->device->hostdata;
-
-       if (!storvsc_scsi_cmd_ok(scmnd)) {
-               scmnd->scsi_done(scmnd);
-               return 0;
-       }
-
-       request_size = sizeof(struct storvsc_cmd_request);
-
-       cmd_request = mempool_alloc(memp->request_mempool,
-                                      GFP_ATOMIC);
-
-       /*
-        * We might be invoked in an interrupt context; hence
-        * mempool_alloc() can fail.
-        */
-       if (!cmd_request)
-               return SCSI_MLQUEUE_DEVICE_BUSY;
-
-       memset(cmd_request, 0, sizeof(struct storvsc_cmd_request));
-
-       /* Setup the cmd request */
-       cmd_request->cmd = scmnd;
-
-       scmnd->host_scribble = (unsigned char *)cmd_request;
-
-       vm_srb = &cmd_request->vstor_packet.vm_srb;
-
-
-       /* Build the SRB */
-       switch (scmnd->sc_data_direction) {
-       case DMA_TO_DEVICE:
-               vm_srb->data_in = WRITE_TYPE;
-               break;
-       case DMA_FROM_DEVICE:
-               vm_srb->data_in = READ_TYPE;
-               break;
-       default:
-               vm_srb->data_in = UNKNOWN_TYPE;
-               break;
-       }
-
-
-       vm_srb->port_number = host_dev->port;
-       vm_srb->path_id = scmnd->device->channel;
-       vm_srb->target_id = scmnd->device->id;
-       vm_srb->lun = scmnd->device->lun;
-
-       vm_srb->cdb_length = scmnd->cmd_len;
-
-       memcpy(vm_srb->cdb, scmnd->cmnd, vm_srb->cdb_length);
-
-       cmd_request->sense_buffer = scmnd->sense_buffer;
-
-
-       cmd_request->data_buffer.len = scsi_bufflen(scmnd);
-       if (scsi_sg_count(scmnd)) {
-               sgl = (struct scatterlist *)scsi_sglist(scmnd);
-               sg_count = scsi_sg_count(scmnd);
-
-               /* check if we need to bounce the sgl */
-               if (do_bounce_buffer(sgl, scsi_sg_count(scmnd)) != -1) {
-                       cmd_request->bounce_sgl =
-                               create_bounce_buffer(sgl, scsi_sg_count(scmnd),
-                                                    scsi_bufflen(scmnd),
-                                                    vm_srb->data_in);
-                       if (!cmd_request->bounce_sgl) {
-                               ret = SCSI_MLQUEUE_HOST_BUSY;
-                               goto queue_error;
-                       }
-
-                       cmd_request->bounce_sgl_count =
-                               ALIGN(scsi_bufflen(scmnd), PAGE_SIZE) >>
-                                       PAGE_SHIFT;
-
-                       if (vm_srb->data_in == WRITE_TYPE)
-                               copy_to_bounce_buffer(sgl,
-                                       cmd_request->bounce_sgl,
-                                       scsi_sg_count(scmnd));
-
-                       sgl = cmd_request->bounce_sgl;
-                       sg_count = cmd_request->bounce_sgl_count;
-               }
-
-               cmd_request->data_buffer.offset = sgl[0].offset;
-
-               for (i = 0; i < sg_count; i++)
-                       cmd_request->data_buffer.pfn_array[i] =
-                               page_to_pfn(sg_page((&sgl[i])));
-
-       } else if (scsi_sglist(scmnd)) {
-               cmd_request->data_buffer.offset =
-                       virt_to_phys(scsi_sglist(scmnd)) & (PAGE_SIZE-1);
-               cmd_request->data_buffer.pfn_array[0] =
-                       virt_to_phys(scsi_sglist(scmnd)) >> PAGE_SHIFT;
-       }
-
-       /* Invokes the vsc to start an IO */
-       ret = storvsc_do_io(dev, cmd_request);
-
-       if (ret == -EAGAIN) {
-               /* no more space */
-
-               if (cmd_request->bounce_sgl_count) {
-                       destroy_bounce_buffer(cmd_request->bounce_sgl,
-                                       cmd_request->bounce_sgl_count);
-
-                       ret = SCSI_MLQUEUE_DEVICE_BUSY;
-                       goto queue_error;
-               }
-       }
-
-       return 0;
-
-queue_error:
-       mempool_free(cmd_request, memp->request_mempool);
-       scmnd->host_scribble = NULL;
-       return ret;
-}
-
-static struct scsi_host_template scsi_driver = {
-       .module =               THIS_MODULE,
-       .name =                 "storvsc_host_t",
-       .bios_param =           storvsc_get_chs,
-       .queuecommand =         storvsc_queuecommand,
-       .eh_host_reset_handler =        storvsc_host_reset_handler,
-       .slave_alloc =          storvsc_device_alloc,
-       .slave_destroy =        storvsc_device_destroy,
-       .slave_configure =      storvsc_device_configure,
-       .cmd_per_lun =          1,
-       /* 64 max_queue * 1 target */
-       .can_queue =            STORVSC_MAX_IO_REQUESTS*STORVSC_MAX_TARGETS,
-       .this_id =              -1,
-       /* no use setting to 0 since ll_blk_rw reset it to 1 */
-       /* currently 32 */
-       .sg_tablesize =         MAX_MULTIPAGE_BUFFER_COUNT,
-       .use_clustering =       DISABLE_CLUSTERING,
-       /* Make sure we dont get a sg segment crosses a page boundary */
-       .dma_boundary =         PAGE_SIZE-1,
-};
-
-enum {
-       SCSI_GUID,
-       IDE_GUID,
-};
-
-static const struct hv_vmbus_device_id id_table[] = {
-       /* SCSI guid */
-       { VMBUS_DEVICE(0xd9, 0x63, 0x61, 0xba, 0xa1, 0x04, 0x29, 0x4d,
-                      0xb6, 0x05, 0x72, 0xe2, 0xff, 0xb1, 0xdc, 0x7f)
-         .driver_data = SCSI_GUID },
-       /* IDE guid */
-       { VMBUS_DEVICE(0x32, 0x26, 0x41, 0x32, 0xcb, 0x86, 0xa2, 0x44,
-                      0x9b, 0x5c, 0x50, 0xd1, 0x41, 0x73, 0x54, 0xf5)
-         .driver_data = IDE_GUID },
-       { },
-};
-
-MODULE_DEVICE_TABLE(vmbus, id_table);
-
-static int storvsc_probe(struct hv_device *device,
-                       const struct hv_vmbus_device_id *dev_id)
-{
-       int ret;
-       struct Scsi_Host *host;
-       struct hv_host_device *host_dev;
-       bool dev_is_ide = ((dev_id->driver_data == IDE_GUID) ? true : false);
-       int target = 0;
-       struct storvsc_device *stor_device;
-
-       host = scsi_host_alloc(&scsi_driver,
-                              sizeof(struct hv_host_device));
-       if (!host)
-               return -ENOMEM;
-
-       host_dev = shost_priv(host);
-       memset(host_dev, 0, sizeof(struct hv_host_device));
-
-       host_dev->port = host->host_no;
-       host_dev->dev = device;
-
-
-       stor_device = kzalloc(sizeof(struct storvsc_device), GFP_KERNEL);
-       if (!stor_device) {
-               ret = -ENOMEM;
-               goto err_out0;
-       }
-
-       stor_device->destroy = false;
-       init_waitqueue_head(&stor_device->waiting_to_drain);
-       stor_device->device = device;
-       stor_device->host = host;
-       hv_set_drvdata(device, stor_device);
-
-       stor_device->port_number = host->host_no;
-       ret = storvsc_connect_to_vsp(device, storvsc_ringbuffer_size);
-       if (ret)
-               goto err_out1;
-
-       host_dev->path = stor_device->path_id;
-       host_dev->target = stor_device->target_id;
-
-       /* max # of devices per target */
-       host->max_lun = STORVSC_MAX_LUNS_PER_TARGET;
-       /* max # of targets per channel */
-       host->max_id = STORVSC_MAX_TARGETS;
-       /* max # of channels */
-       host->max_channel = STORVSC_MAX_CHANNELS - 1;
-       /* max cmd length */
-       host->max_cmd_len = STORVSC_MAX_CMD_LEN;
-
-       /* Register the HBA and start the scsi bus scan */
-       ret = scsi_add_host(host, &device->device);
-       if (ret != 0)
-               goto err_out2;
-
-       if (!dev_is_ide) {
-               scsi_scan_host(host);
-       } else {
-               target = (device->dev_instance.b[5] << 8 |
-                        device->dev_instance.b[4]);
-               ret = scsi_add_device(host, 0, target, 0);
-               if (ret) {
-                       scsi_remove_host(host);
-                       goto err_out2;
-               }
-       }
-       return 0;
-
-err_out2:
-       /*
-        * Once we have connected with the host, we would need to
-        * to invoke storvsc_dev_remove() to rollback this state and
-        * this call also frees up the stor_device; hence the jump around
-        * err_out1 label.
-        */
-       storvsc_dev_remove(device);
-       goto err_out0;
-
-err_out1:
-       kfree(stor_device);
-
-err_out0:
-       scsi_host_put(host);
-       return ret;
-}
-
-static int storvsc_remove(struct hv_device *dev)
-{
-       struct storvsc_device *stor_device = hv_get_drvdata(dev);
-       struct Scsi_Host *host = stor_device->host;
-
-       scsi_remove_host(host);
-       storvsc_dev_remove(dev);
-       scsi_host_put(host);
-
-       return 0;
-}
-
-static struct hv_driver storvsc_drv = {
-       .name = KBUILD_MODNAME,
-       .id_table = id_table,
-       .probe = storvsc_probe,
-       .remove = storvsc_remove,
-};
-
-static int __init storvsc_drv_init(void)
-{
-       u32 max_outstanding_req_per_channel;
-
-       /*
-        * Divide the ring buffer data size (which is 1 page less
-        * than the ring buffer size since that page is reserved for
-        * the ring buffer indices) by the max request size (which is
-        * vmbus_channel_packet_multipage_buffer + struct vstor_packet + u64)
-        */
-       max_outstanding_req_per_channel =
-               ((storvsc_ringbuffer_size - PAGE_SIZE) /
-               ALIGN(MAX_MULTIPAGE_BUFFER_PACKET +
-               sizeof(struct vstor_packet) + sizeof(u64),
-               sizeof(u64)));
-
-       if (max_outstanding_req_per_channel <
-           STORVSC_MAX_IO_REQUESTS)
-               return -EINVAL;
-
-       return vmbus_driver_register(&storvsc_drv);
-}
-
-static void __exit storvsc_drv_exit(void)
-{
-       vmbus_driver_unregister(&storvsc_drv);
-}
-
-MODULE_LICENSE("GPL");
-MODULE_VERSION(HV_DRV_VERSION);
-MODULE_DESCRIPTION("Microsoft Hyper-V virtual storage driver");
-module_init(storvsc_drv_init);
-module_exit(storvsc_drv_exit);
This page took 0.05788 seconds and 5 git commands to generate.