[PATCH 0/] OSD changes for 2.6.31

[PATCH 0/] OSD changes for 2.6.31

[Boaz Harrosh]

These are the accumulated osd patches for the next merge window,
Linux 2.6.31. 

Most of these patches are a repost rebased on latest scsi-misc

List of patches:
[PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
[PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
[PATCH 03/10 resend] libosd: Better printout of OSD target system information
[PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
[PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
[PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter

   These are a resend. The last 2 patches are important for supporting Jeff's
   osdblk driver. (The first 4 are old patches that thier time has come)

[PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
   Simple wrapper

[PATCH 08/10] osduld: use filp_open() when looking up an osd-device

    This patch was inspired by Al Viro. It uses the in-kernel open-file
    mechanism to look up the osd char-device for Kernel users like exofs
    and later pNFS-Objlayout.
    It was heavily tested and works well.
    It fixes the problem we had before, of need to open+close a file handle
    on the osd-char-device before mounting an exofs.
    Please review

[PATCH 09/10] libosd: Use REQ_QUIET requests.
    Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
    because of the way we routinely receive a recovered scsi sense. Recently, for
    reason I could not find, this condition started to print messages at:
       scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
    Use of a REQ_QUIET flag is needed to silence these prints.
    OSD has it's own prints for scsi-sense returns, when needed.

[PATCH 10/10] osd: Remove out-of-tree left overs

log stat
 drivers/scsi/osd/Kbuild          |   25 -----------
 drivers/scsi/osd/Makefile        |   37 -----------------
 drivers/scsi/osd/osd_initiator.c |   83 +++++++++++++++++++++++++++-----------
 drivers/scsi/osd/osd_uld.c       |   66 ++++++++++++++----------------
 fs/exofs/common.h                |    6 ---
 fs/exofs/inode.c                 |    8 ++--
 fs/exofs/osd.c                   |   26 ------------
 include/scsi/osd_attributes.h    |   74 +++++++++++++++++++++++++++++++++-
 include/scsi/osd_initiator.h     |   14 +++++-
 include/scsi/osd_protocol.h      |    8 ++++
 10 files changed, 185 insertions(+), 162 deletions(-)
 delete mode 100755 drivers/scsi/osd/Makefile

These patches are available in the git repository at:

  git://git.open-osd.org/linux-open-osd.git osd-devel

Thanks
Boaz

[PATCH 01/10 resend] libosd: OSD2r05: Additional command enums

[Boaz Harrosh]

Add all constant definitions of new OSD commands added in
revision 4 & 5. Mainly for creating snapshots and clones.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 include/scsi/osd_protocol.h |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)

diff --git a/include/scsi/osd_protocol.h b/include/scsi/osd_protocol.h
index 62b2ab8..2cc8e8b 100644
--- a/include/scsi/osd_protocol.h
+++ b/include/scsi/osd_protocol.h
@@ -303,7 +303,15 @@ enum osd_service_actions {
 	OSD_ACT_V2(REMOVE_MEMBER_OBJECTS,	0x21)
 	OSD_ACT_V2(GET_MEMBER_ATTRIBUTES,	0x22)
 	OSD_ACT_V2(SET_MEMBER_ATTRIBUTES,	0x23)
+
+	OSD_ACT_V2(CREATE_CLONE,		0x28)
+	OSD_ACT_V2(CREATE_SNAPSHOT,		0x29)
+	OSD_ACT_V2(DETACH_CLONE,		0x2A)
+	OSD_ACT_V2(REFRESH_SNAPSHOT_CLONE,	0x2B)
+	OSD_ACT_V2(RESTORE_PARTITION_FROM_SNAPSHOT, 0x2C)
+
 	OSD_ACT_V2(READ_MAP,			0x31)
+	OSD_ACT_V2(READ_MAPS_COMPARE,		0x32)
 
 	OSD_ACT_V1_V2(PERFORM_SCSI_COMMAND,	0x8F7E, 0x8F7C)
 	OSD_ACT_V1_V2(SCSI_TASK_MANAGEMENT,	0x8F7F, 0x8F7D)
-- 
1.6.2.1

[PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions

[Boaz Harrosh]

Some New revision 5 Attribute definitions.
Some missing definitions of Attributes and pages

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 include/scsi/osd_attributes.h |   74 +++++++++++++++++++++++++++++++++++++++-
 1 files changed, 72 insertions(+), 2 deletions(-)

diff --git a/include/scsi/osd_attributes.h b/include/scsi/osd_attributes.h
index f888a6f..56e920a 100644
--- a/include/scsi/osd_attributes.h
+++ b/include/scsi/osd_attributes.h
@@ -29,6 +29,7 @@ enum {
 	OSD_APAGE_PARTITION_INFORMATION = OSD_APAGE_PARTITION_FIRST + 1,
 	OSD_APAGE_PARTITION_QUOTAS	= OSD_APAGE_PARTITION_FIRST + 2,
 	OSD_APAGE_PARTITION_TIMESTAMP	= OSD_APAGE_PARTITION_FIRST + 3,
+	OSD_APAGE_PARTITION_ATTR_ACCESS = OSD_APAGE_PARTITION_FIRST + 4,
 	OSD_APAGE_PARTITION_SECURITY	= OSD_APAGE_PARTITION_FIRST + 5,
 	OSD_APAGE_PARTITION_LAST	= 0x5FFFFFFF,
 
@@ -51,7 +52,9 @@ enum {
 	OSD_APAGE_RESERVED_TYPE_LAST	= 0xEFFFFFFF,
 
 	OSD_APAGE_COMMON_FIRST		= 0xF0000000,
-	OSD_APAGE_COMMON_LAST		= 0xFFFFFFFE,
+	OSD_APAGE_COMMON_LAST		= 0xFFFFFFFD,
+
+	OSD_APAGE_CURRENT_COMMAND	= 0xFFFFFFFE,
 
 	OSD_APAGE_REQUEST_ALL		= 0xFFFFFFFF,
 };
@@ -106,10 +109,30 @@ enum {
 	OSD_ATTR_RI_PRODUCT_REVISION_LEVEL   = 0x7,   /* 4        */
 	OSD_ATTR_RI_PRODUCT_SERIAL_NUMBER    = 0x8,   /* variable */
 	OSD_ATTR_RI_OSD_NAME                 = 0x9,   /* variable */
+	OSD_ATTR_RI_MAX_CDB_CONTINUATION_LEN = 0xA,   /* 4        */
 	OSD_ATTR_RI_TOTAL_CAPACITY           = 0x80,  /* 8        */
 	OSD_ATTR_RI_USED_CAPACITY            = 0x81,  /* 8        */
 	OSD_ATTR_RI_NUMBER_OF_PARTITIONS     = 0xC0,  /* 8        */
 	OSD_ATTR_RI_CLOCK                    = 0x100, /* 6        */
+	OARI_DEFAULT_ISOLATION_METHOD        = 0X110, /* 1        */
+	OARI_SUPPORTED_ISOLATION_METHODS     = 0X111, /* 32       */
+
+	OARI_DATA_ATOMICITY_GUARANTEE                   = 0X120,   /* 8       */
+	OARI_DATA_ATOMICITY_ALIGNMENT                   = 0X121,   /* 8       */
+	OARI_ATTRIBUTES_ATOMICITY_GUARANTEE             = 0X122,   /* 8       */
+	OARI_DATA_ATTRIBUTES_ATOMICITY_MULTIPLIER       = 0X123,   /* 1       */
+
+	OARI_MAXIMUM_SNAPSHOTS_COUNT                    = 0X1C1,    /* 0 or 4 */
+	OARI_MAXIMUM_CLONES_COUNT                       = 0X1C2,    /* 0 or 4 */
+	OARI_MAXIMUM_BRANCH_DEPTH                       = 0X1CC,    /* 0 or 4 */
+	OARI_SUPPORTED_OBJECT_DUPLICATION_METHOD_FIRST  = 0X200,    /* 0 or 4 */
+	OARI_SUPPORTED_OBJECT_DUPLICATION_METHOD_LAST   = 0X2ff,    /* 0 or 4 */
+	OARI_SUPPORTED_TIME_OF_DUPLICATION_METHOD_FIRST = 0X300,    /* 0 or 4 */
+	OARI_SUPPORTED_TIME_OF_DUPLICATION_METHOD_LAST  = 0X30F,    /* 0 or 4 */
+	OARI_SUPPORT_FOR_DUPLICATED_OBJECT_FREEZING     = 0X310,    /* 0 or 4 */
+	OARI_SUPPORT_FOR_SNAPSHOT_REFRESHING            = 0X311,    /* 0 or 1 */
+	OARI_SUPPORTED_CDB_CONTINUATION_DESC_TYPE_FIRST = 0X7000001,/* 0 or 4 */
+	OARI_SUPPORTED_CDB_CONTINUATION_DESC_TYPE_LAST  = 0X700FFFF,/* 0 or 4 */
 };
 /* Root_Information_attributes_page does not have a get_page structure */
 
@@ -120,7 +143,15 @@ enum {
 	OSD_ATTR_PI_PARTITION_ID            = 0x1,     /* 8        */
 	OSD_ATTR_PI_USERNAME                = 0x9,     /* variable */
 	OSD_ATTR_PI_USED_CAPACITY           = 0x81,    /* 8        */
+	OSD_ATTR_PI_USED_CAPACITY_INCREMENT = 0x84,    /* 0 or 8   */
 	OSD_ATTR_PI_NUMBER_OF_OBJECTS       = 0xC1,    /* 8        */
+
+	OSD_ATTR_PI_ACTUAL_DATA_SPACE                      = 0xD1, /* 0 or 8 */
+	OSD_ATTR_PI_RESERVED_DATA_SPACE                    = 0xD2, /* 0 or 8 */
+	OSD_ATTR_PI_DEFAULT_SNAPSHOT_DUPLICATION_METHOD    = 0x200,/* 0 or 4 */
+	OSD_ATTR_PI_DEFAULT_CLONE_DUPLICATION_METHOD       = 0x201,/* 0 or 4 */
+	OSD_ATTR_PI_DEFAULT_SP_TIME_OF_DUPLICATION         = 0x300,/* 0 or 4 */
+	OSD_ATTR_PI_DEFAULT_CLONE_TIME_OF_DUPLICATION      = 0x301,/* 0 or 4 */
 };
 /* Partition Information attributes page does not have a get_page structure */
 
@@ -131,6 +162,7 @@ enum {
 	OSD_ATTR_CI_PARTITION_ID           = 0x1,       /* 8        */
 	OSD_ATTR_CI_COLLECTION_OBJECT_ID   = 0x2,       /* 8        */
 	OSD_ATTR_CI_USERNAME               = 0x9,       /* variable */
+	OSD_ATTR_CI_COLLECTION_TYPE        = 0xA,       /* 1        */
 	OSD_ATTR_CI_USED_CAPACITY          = 0x81,      /* 8        */
 };
 /* Collection Information attributes page does not have a get_page structure */
@@ -144,6 +176,8 @@ enum {
 	OSD_ATTR_OI_USERNAME             = 0x9,       /* variable */
 	OSD_ATTR_OI_USED_CAPACITY        = 0x81,      /* 8        */
 	OSD_ATTR_OI_LOGICAL_LENGTH       = 0x82,      /* 8        */
+	SD_ATTR_OI_ACTUAL_DATA_SPACE     = 0XD1,      /* 0 OR 8   */
+	SD_ATTR_OI_RESERVED_DATA_SPACE   = 0XD2,      /* 0 OR 8   */
 };
 /* Object Information attributes page does not have a get_page structure */
 
@@ -248,7 +282,18 @@ struct object_timestamps_attributes_page {
 	struct osd_timestamp data_modified_time;
 }  __packed;
 
-/* 7.1.2.19 Collections attributes page */
+/* OSD2r05: 7.1.3.19 Attributes Access attributes page
+ * (OSD_APAGE_PARTITION_ATTR_ACCESS)
+ *
+ * each attribute is of the form below. Total array length is deduced
+ * from the attribute's length
+ * (See allowed_attributes_access of the struct osd_cap_object_descriptor)
+ */
+struct attributes_access_attr {
+	struct osd_attributes_list_attrid attr_list[0];
+} __packed;
+
+/* OSD2r05: 7.1.2.21 Collections attributes page */
 /* TBD */
 
 /* 7.1.2.20 Root Policy/Security attributes page (OSD_APAGE_ROOT_SECURITY) */
@@ -324,4 +369,29 @@ struct object_security_attributes_page {
 	__be32 policy_access_tag;
 }  __packed;
 
+/* OSD2r05: 7.1.3.31 Current Command attributes page
+ * (OSD_APAGE_CURRENT_COMMAND)
+ */
+enum {
+	OSD_ATTR_CC_RESPONSE_INTEGRITY_CHECK_VALUE     = 0x1, /* 32  */
+	OSD_ATTR_CC_OBJECT_TYPE                        = 0x2, /* 1   */
+	OSD_ATTR_CC_PARTITION_ID                       = 0x3, /* 8   */
+	OSD_ATTR_CC_OBJECT_ID                          = 0x4, /* 8   */
+	OSD_ATTR_CC_STARTING_BYTE_ADDRESS_OF_APPEND    = 0x5, /* 8   */
+	OSD_ATTR_CC_CHANGE_IN_USED_CAPACITY            = 0x6, /* 8   */
+};
+
+/*TBD: osdv1_current_command_attributes_page */
+
+struct osdv2_current_command_attributes_page {
+	struct osd_attr_page_header hdr;  /* id=0xFFFFFFFE, size=0x44 */
+	u8 response_integrity_check_value[OSD_CRYPTO_KEYID_SIZE];
+	u8 object_type;
+	u8 reserved[3];
+	__be64 partition_id;
+	__be64 object_id;
+	__be64 starting_byte_address_of_append;
+	__be64 change_in_used_capacity;
+};
+
 #endif /*ndef __OSD_ATTRIBUTES_H__*/
-- 
1.6.2.1

[PATCH 03/10 resend] libosd: Better printout of OSD target system information

[Boaz Harrosh]

Shorten out the Attributes names.
Align all results on column 24.
Print system ID in a new line.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 drivers/scsi/osd/osd_initiator.c |   23 ++++++++++++-----------
 1 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index 1ce6b24..15f0bbc 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -118,39 +118,39 @@ static int _osd_print_system_info(struct osd_dev *od, void *caps)
 		_osd_ver_desc(or));
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_VENDOR_IDENTIFICATION [%s]\n",
+	OSD_INFO("VENDOR_IDENTIFICATION  [%s]\n",
 		(char *)pFirst);
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_PRODUCT_IDENTIFICATION [%s]\n",
+	OSD_INFO("PRODUCT_IDENTIFICATION [%s]\n",
 		(char *)pFirst);
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_PRODUCT_MODEL [%s]\n",
+	OSD_INFO("PRODUCT_MODEL          [%s]\n",
 		(char *)pFirst);
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_PRODUCT_REVISION_LEVEL [%u]\n",
+	OSD_INFO("PRODUCT_REVISION_LEVEL [%u]\n",
 		pFirst ? get_unaligned_be32(pFirst) : ~0U);
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_PRODUCT_SERIAL_NUMBER [%s]\n",
+	OSD_INFO("PRODUCT_SERIAL_NUMBER  [%s]\n",
 		(char *)pFirst);
 
 	pFirst = get_attrs[a].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_OSD_NAME [%s]\n", (char *)pFirst);
+	OSD_INFO("OSD_NAME               [%s]\n", (char *)pFirst);
 	a++;
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_TOTAL_CAPACITY [0x%llx]\n",
+	OSD_INFO("TOTAL_CAPACITY         [0x%llx]\n",
 		pFirst ? _LLU(get_unaligned_be64(pFirst)) : ~0ULL);
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_USED_CAPACITY [0x%llx]\n",
+	OSD_INFO("USED_CAPACITY          [0x%llx]\n",
 		pFirst ? _LLU(get_unaligned_be64(pFirst)) : ~0ULL);
 
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_NUMBER_OF_PARTITIONS [%llu]\n",
+	OSD_INFO("NUMBER_OF_PARTITIONS   [%llu]\n",
 		pFirst ? _LLU(get_unaligned_be64(pFirst)) : ~0ULL);
 
 	if (a >= nelem)
@@ -158,7 +158,7 @@ static int _osd_print_system_info(struct osd_dev *od, void *caps)
 
 	/* FIXME: Where are the time utilities */
 	pFirst = get_attrs[a++].val_ptr;
-	OSD_INFO("OSD_ATTR_RI_CLOCK [0x%02x%02x%02x%02x%02x%02x]\n",
+	OSD_INFO("CLOCK                  [0x%02x%02x%02x%02x%02x%02x]\n",
 		((char *)pFirst)[0], ((char *)pFirst)[1],
 		((char *)pFirst)[2], ((char *)pFirst)[3],
 		((char *)pFirst)[4], ((char *)pFirst)[5]);
@@ -169,7 +169,8 @@ static int _osd_print_system_info(struct osd_dev *od, void *caps)
 
 		hex_dump_to_buffer(get_attrs[a].val_ptr, len, 32, 1,
 				   sid_dump, sizeof(sid_dump), true);
-		OSD_INFO("OSD_ATTR_RI_OSD_SYSTEM_ID(%d) [%s]\n", len, sid_dump);
+		OSD_INFO("OSD_SYSTEM_ID(%d)\n"
+			 "        [%s]\n", len, sid_dump);
 		a++;
 	}
 out:
-- 
1.6.2.1

[PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API

[Boaz Harrosh]

By popular demand, define usefull wrappers for osd_req_read/write
that recieve kernel pointers. All users had their own.

Also remove these from exofs

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 drivers/scsi/osd/osd_initiator.c |   28 ++++++++++++++++++++++++++++
 fs/exofs/common.h                |    6 ------
 fs/exofs/osd.c                   |   26 --------------------------
 include/scsi/osd_initiator.h     |    4 ++++
 4 files changed, 32 insertions(+), 32 deletions(-)

diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index 15f0bbc..c98153b 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -789,6 +789,20 @@ void osd_req_write(struct osd_request *or,
 }
 EXPORT_SYMBOL(osd_req_write);
 
+int osd_req_write_kern(struct osd_request *or,
+	const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
+{
+	struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
+	struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
+
+	if (IS_ERR(bio))
+		return PTR_ERR(bio);
+
+	osd_req_write(or, obj, bio, offset);
+	return 0;
+}
+EXPORT_SYMBOL(osd_req_write_kern);
+
 /*TODO: void osd_req_append(struct osd_request *,
 	const struct osd_obj_id *, struct bio *data_out); */
 /*TODO: void osd_req_create_write(struct osd_request *,
@@ -824,6 +838,20 @@ void osd_req_read(struct osd_request *or,
 }
 EXPORT_SYMBOL(osd_req_read);
 
+int osd_req_read_kern(struct osd_request *or,
+	const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
+{
+	struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
+	struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
+
+	if (IS_ERR(bio))
+		return PTR_ERR(bio);
+
+	osd_req_read(or, obj, bio, offset);
+	return 0;
+}
+EXPORT_SYMBOL(osd_req_read_kern);
+
 void osd_req_get_attributes(struct osd_request *or,
 	const struct osd_obj_id *obj)
 {
diff --git a/fs/exofs/common.h b/fs/exofs/common.h
index b1512c4..24667ee 100644
--- a/fs/exofs/common.h
+++ b/fs/exofs/common.h
@@ -175,10 +175,4 @@ int exofs_async_op(struct osd_request *or,
 
 int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr);
 
-int osd_req_read_kern(struct osd_request *or,
-	const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
-
-int osd_req_write_kern(struct osd_request *or,
-	const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
-
 #endif /*ifndef __EXOFS_COM_H__*/
diff --git a/fs/exofs/osd.c b/fs/exofs/osd.c
index b249ae9..48cc4d1 100644
--- a/fs/exofs/osd.c
+++ b/fs/exofs/osd.c
@@ -125,29 +125,3 @@ int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr)
 
 	return -EIO;
 }
-
-int osd_req_read_kern(struct osd_request *or,
-	const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
-{
-	struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
-	struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
-
-	if (!bio)
-		return -ENOMEM;
-
-	osd_req_read(or, obj, bio, offset);
-	return 0;
-}
-
-int osd_req_write_kern(struct osd_request *or,
-	const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
-{
-	struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
-	struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
-
-	if (!bio)
-		return -ENOMEM;
-
-	osd_req_write(or, obj, bio, offset);
-	return 0;
-}
diff --git a/include/scsi/osd_initiator.h b/include/scsi/osd_initiator.h
index b24d961..6132790 100644
--- a/include/scsi/osd_initiator.h
+++ b/include/scsi/osd_initiator.h
@@ -364,6 +364,8 @@ void osd_req_remove_object(struct osd_request *or, struct osd_obj_id *);
 
 void osd_req_write(struct osd_request *or,
 	const struct osd_obj_id *, struct bio *data_out, u64 offset);
+int osd_req_write_kern(struct osd_request *or,
+	const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
 void osd_req_append(struct osd_request *or,
 	const struct osd_obj_id *, struct bio *data_out);/* NI */
 void osd_req_create_write(struct osd_request *or,
@@ -379,6 +381,8 @@ void osd_req_flush_object(struct osd_request *or,
 
 void osd_req_read(struct osd_request *or,
 	const struct osd_obj_id *, struct bio *data_in, u64 offset);
+int osd_req_read_kern(struct osd_request *or,
+	const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
 
 /*
  * Root/Partition/Collection/Object Attributes commands
-- 
1.6.2.1

[PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes

[Boaz Harrosh]

_osd_req_finalize_data_integrity was trying to deduce the number of
out_bytes from passed osd_request->out.bio. This is wrong when
the bio is chained. The caller of _osd_req_finalize_data_integrity
has more ready available information and should just pass it.

Also in the light of future support for CDB-continuation segment this is
a better solution.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 drivers/scsi/osd/osd_initiator.c |    9 +++++----
 1 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index c98153b..ba2ebae 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -1262,7 +1262,7 @@ static inline void osd_sec_parms_set_in_offset(bool is_v1,
 }
 
 static int _osd_req_finalize_data_integrity(struct osd_request *or,
-	bool has_in, bool has_out, const u8 *cap_key)
+	bool has_in, bool has_out, u64 out_data_bytes, const u8 *cap_key)
 {
 	struct osd_security_parameters *sec_parms = _osd_req_sec_params(or);
 	int ret;
@@ -1277,8 +1277,7 @@ static int _osd_req_finalize_data_integrity(struct osd_request *or,
 		};
 		unsigned pad;
 
-		or->out_data_integ.data_bytes = cpu_to_be64(
-			or->out.bio ? or->out.bio->bi_size : 0);
+		or->out_data_integ.data_bytes = cpu_to_be64(out_data_bytes);
 		or->out_data_integ.set_attributes_bytes = cpu_to_be64(
 			or->set_attr.total_bytes);
 		or->out_data_integ.get_attributes_bytes = cpu_to_be64(
@@ -1370,6 +1369,7 @@ int osd_finalize_request(struct osd_request *or,
 {
 	struct osd_cdb_head *cdbh = osd_cdb_head(&or->cdb);
 	bool has_in, has_out;
+	u64 out_data_bytes = or->out.total_bytes;
 	int ret;
 
 	if (options & OSD_REQ_FUA)
@@ -1439,7 +1439,8 @@ int osd_finalize_request(struct osd_request *or,
 		}
 	}
 
-	ret = _osd_req_finalize_data_integrity(or, has_in, has_out, cap_key);
+	ret = _osd_req_finalize_data_integrity(or, has_in, has_out,
+					       out_data_bytes, cap_key);
 	if (ret)
 		return ret;
 
-- 
1.6.2.1

[PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter

[Boaz Harrosh]

For supporting of chained-bios we can not inspect the first
bio only, as before. Caller shall pass the total length of the
request, ie. sum_bytes(bio-chain).

Also since the bio might be a chain we don't set it's direction
on behalf of it's callers. The bio direction should be properly
set prior to this call. So fix a couple of write users that now
need to set the bio direction properly

[In this patch I change both library code and user sites at
 exofs, to make it easy on integration. It should be submitted
 via James's scsi-misc tree.]

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
CC: Jeff Garzik <jeff@garzik.org>
---
 drivers/scsi/osd/osd_initiator.c |   23 +++++++++++++----------
 fs/exofs/inode.c                 |    5 +++--
 include/scsi/osd_initiator.h     |    4 ++--
 3 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index ba2ebae..3f5ec57 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -779,13 +779,14 @@ EXPORT_SYMBOL(osd_req_remove_object);
 */
 
 void osd_req_write(struct osd_request *or,
-	const struct osd_obj_id *obj, struct bio *bio, u64 offset)
+	const struct osd_obj_id *obj, u64 offset,
+	struct bio *bio, u64 len)
 {
-	_osd_req_encode_common(or, OSD_ACT_WRITE, obj, offset, bio->bi_size);
+	_osd_req_encode_common(or, OSD_ACT_WRITE, obj, offset, len);
 	WARN_ON(or->out.bio || or->out.total_bytes);
-	bio->bi_rw |= (1 << BIO_RW);
+	WARN_ON(0 ==  bio_rw_flagged(bio, BIO_RW));
 	or->out.bio = bio;
-	or->out.total_bytes = bio->bi_size;
+	or->out.total_bytes = len;
 }
 EXPORT_SYMBOL(osd_req_write);
 
@@ -798,7 +799,8 @@ int osd_req_write_kern(struct osd_request *or,
 	if (IS_ERR(bio))
 		return PTR_ERR(bio);
 
-	osd_req_write(or, obj, bio, offset);
+	bio->bi_rw |= (1 << BIO_RW); /* FIXME: bio_set_dir() */
+	osd_req_write(or, obj, offset, bio, len);
 	return 0;
 }
 EXPORT_SYMBOL(osd_req_write_kern);
@@ -828,13 +830,14 @@ void osd_req_flush_object(struct osd_request *or,
 EXPORT_SYMBOL(osd_req_flush_object);
 
 void osd_req_read(struct osd_request *or,
-	const struct osd_obj_id *obj, struct bio *bio, u64 offset)
+	const struct osd_obj_id *obj, u64 offset,
+	struct bio *bio, u64 len)
 {
-	_osd_req_encode_common(or, OSD_ACT_READ, obj, offset, bio->bi_size);
+	_osd_req_encode_common(or, OSD_ACT_READ, obj, offset, len);
 	WARN_ON(or->in.bio || or->in.total_bytes);
-	bio->bi_rw &= ~(1 << BIO_RW);
+	WARN_ON(1 == bio_rw_flagged(bio, BIO_RW));
 	or->in.bio = bio;
-	or->in.total_bytes = bio->bi_size;
+	or->in.total_bytes = len;
 }
 EXPORT_SYMBOL(osd_req_read);
 
@@ -847,7 +850,7 @@ int osd_req_read_kern(struct osd_request *or,
 	if (IS_ERR(bio))
 		return PTR_ERR(bio);
 
-	osd_req_read(or, obj, bio, offset);
+	osd_req_read(or, obj, offset, bio, len);
 	return 0;
 }
 EXPORT_SYMBOL(osd_req_read_kern);
diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c
index ba8d9fa..f79e8e5 100644
--- a/fs/exofs/inode.c
+++ b/fs/exofs/inode.c
@@ -266,7 +266,7 @@ static int read_exec(struct page_collect *pcol, bool is_sync)
 		goto err;
 	}
 
-	osd_req_read(or, &obj, pcol->bio, i_start);
+	osd_req_read(or, &obj, i_start, pcol->bio, pcol->length);
 
 	if (is_sync) {
 		exofs_sync_op(or, pcol->sbi->s_timeout, oi->i_cred);
@@ -522,7 +522,8 @@ static int write_exec(struct page_collect *pcol)
 
 	*pcol_copy = *pcol;
 
-	osd_req_write(or, &obj, pcol_copy->bio, i_start);
+	pcol_copy->bio->bi_rw |= (1 << BIO_RW); /* FIXME: bio_set_dir() */
+	osd_req_write(or, &obj, i_start, pcol_copy->bio, pcol_copy->length);
 	ret = exofs_async_op(or, writepages_done, pcol_copy, oi->i_cred);
 	if (unlikely(ret)) {
 		EXOFS_ERR("write_exec: exofs_async_op() Faild\n");
diff --git a/include/scsi/osd_initiator.h b/include/scsi/osd_initiator.h
index 6132790..8c1e3b8 100644
--- a/include/scsi/osd_initiator.h
+++ b/include/scsi/osd_initiator.h
@@ -363,7 +363,7 @@ void osd_req_create_object(struct osd_request *or, struct osd_obj_id *);
 void osd_req_remove_object(struct osd_request *or, struct osd_obj_id *);
 
 void osd_req_write(struct osd_request *or,
-	const struct osd_obj_id *, struct bio *data_out, u64 offset);
+	const struct osd_obj_id *obj, u64 offset, struct bio *bio, u64 len);
 int osd_req_write_kern(struct osd_request *or,
 	const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
 void osd_req_append(struct osd_request *or,
@@ -380,7 +380,7 @@ void osd_req_flush_object(struct osd_request *or,
 	/*V2*/ u64 offset, /*V2*/ u64 len);
 
 void osd_req_read(struct osd_request *or,
-	const struct osd_obj_id *, struct bio *data_in, u64 offset);
+	const struct osd_obj_id *obj, u64 offset, struct bio *bio, u64 len);
 int osd_req_read_kern(struct osd_request *or,
 	const struct osd_obj_id *obj, u64 offset, void *buff, u64 len);
 
-- 
1.6.2.1

[PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue

[Boaz Harrosh]

libosd users that need to work with bios, must sometime use
the request_queue associated with the osd_dev. Make a wrapper for
that, and convert all in-tree users.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 drivers/scsi/osd/osd_initiator.c |    6 +++---
 fs/exofs/inode.c                 |    3 +--
 include/scsi/osd_initiator.h     |    5 +++++
 3 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index 3f5ec57..3959797 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -670,7 +670,7 @@ static int _osd_req_list_objects(struct osd_request *or,
 	__be16 action, const struct osd_obj_id *obj, osd_id initial_id,
 	struct osd_obj_id_list *list, unsigned nelem)
 {
-	struct request_queue *q = or->osd_dev->scsi_device->request_queue;
+	struct request_queue *q = osd_request_queue(or->osd_dev);
 	u64 len = nelem * sizeof(osd_id) + sizeof(*list);
 	struct bio *bio;
 
@@ -793,7 +793,7 @@ EXPORT_SYMBOL(osd_req_write);
 int osd_req_write_kern(struct osd_request *or,
 	const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
 {
-	struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
+	struct request_queue *req_q = osd_request_queue(or->osd_dev);
 	struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
 
 	if (IS_ERR(bio))
@@ -844,7 +844,7 @@ EXPORT_SYMBOL(osd_req_read);
 int osd_req_read_kern(struct osd_request *or,
 	const struct osd_obj_id *obj, u64 offset, void* buff, u64 len)
 {
-	struct request_queue *req_q = or->osd_dev->scsi_device->request_queue;
+	struct request_queue *req_q = osd_request_queue(or->osd_dev);
 	struct bio *bio = bio_map_kern(req_q, buff, len, GFP_KERNEL);
 
 	if (IS_ERR(bio))
diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c
index f79e8e5..77d0a29 100644
--- a/fs/exofs/inode.c
+++ b/fs/exofs/inode.c
@@ -59,10 +59,9 @@ static void _pcol_init(struct page_collect *pcol, unsigned expected_pages,
 		struct inode *inode)
 {
 	struct exofs_sb_info *sbi = inode->i_sb->s_fs_info;
-	struct request_queue *req_q = sbi->s_dev->scsi_device->request_queue;
 
 	pcol->sbi = sbi;
-	pcol->req_q = req_q;
+	pcol->req_q = osd_request_queue(sbi->s_dev);
 	pcol->inode = inode;
 	pcol->expected_pages = expected_pages;
 
diff --git a/include/scsi/osd_initiator.h b/include/scsi/osd_initiator.h
index 8c1e3b8..b44dc53 100644
--- a/include/scsi/osd_initiator.h
+++ b/include/scsi/osd_initiator.h
@@ -18,6 +18,7 @@
 #include "osd_types.h"
 
 #include <linux/blkdev.h>
+#include <scsi/scsi_device.h>
 
 /* Note: "NI" in comments below means "Not Implemented yet" */
 
@@ -69,6 +70,10 @@ void osd_dev_fini(struct osd_dev *od);
 
 /* some hi level device operations */
 int osd_auto_detect_ver(struct osd_dev *od, void *caps);    /* GFP_KERNEL */
+static inline struct request_queue *osd_request_queue(struct osd_dev *od)
+{
+	return od->scsi_device->request_queue;
+}
 
 /* we might want to use function vector in the future */
 static inline void osd_dev_set_ver(struct osd_dev *od, enum osd_std_version v)
-- 
1.6.2.1

[PATCH 08/10] osduld: use filp_open() when looking up an osd-device

[Boaz Harrosh]

This patch was inspired by Al Viro, for simplifying and fixing the
retrieval of osd-devices by in-kernel users, eg: file systems.
In-Kernel users, now, go through the same path user-mode does by
opening a file on the osd char-device and though holding a reference
to both the device and the Module.

A file pointer was added to the osd_dev structure which is now
allocated for each user. The internal osd_dev is no longer exposed
outside of the uld. I wanted to do that for a long time so each
libosd user can have his own defaults on the device.

The API is left the same, so user code need not change.

It is no longer needed to open/close a file handle on the osd
char-device from user-mode, before mounting an exofs on it.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
CC: Al Viro <viro@ZenIV.linux.org.uk>
---
 drivers/scsi/osd/osd_uld.c   |   66 +++++++++++++++++++-----------------------
 include/scsi/osd_initiator.h |    1 +
 2 files changed, 31 insertions(+), 36 deletions(-)

diff --git a/drivers/scsi/osd/osd_uld.c b/drivers/scsi/osd/osd_uld.c
index 22b59e1..0bdef33 100644
--- a/drivers/scsi/osd/osd_uld.c
+++ b/drivers/scsi/osd/osd_uld.c
@@ -49,6 +49,7 @@
 #include <linux/device.h>
 #include <linux/idr.h>
 #include <linux/major.h>
+#include <linux/file.h>
 
 #include <scsi/scsi.h>
 #include <scsi/scsi_driver.h>
@@ -175,10 +176,9 @@ static const struct file_operations osd_fops = {
 
 struct osd_dev *osduld_path_lookup(const char *name)
 {
-	struct path path;
-	struct inode *inode;
-	struct cdev *cdev;
-	struct osd_uld_device *uninitialized_var(oud);
+	struct osd_uld_device *oud;
+	struct osd_dev *od;
+	struct file *file;
 	int error;
 
 	if (!name || !*name) {
@@ -186,52 +186,46 @@ struct osd_dev *osduld_path_lookup(const char *name)
 		return ERR_PTR(-EINVAL);
 	}
 
-	error = kern_path(name, LOOKUP_FOLLOW, &path);
-	if (error) {
-		OSD_ERR("path_lookup of %s failed=>%d\n", name, error);
-		return ERR_PTR(error);
-	}
+	od = kzalloc(sizeof(*od), GFP_KERNEL);
+	if (!od)
+		return ERR_PTR(-ENOMEM);
 
-	inode = path.dentry->d_inode;
-	error = -EINVAL; /* Not the right device e.g osd_uld_device */
-	if (!S_ISCHR(inode->i_mode)) {
-		OSD_DEBUG("!S_ISCHR()\n");
-		goto out;
+	file = filp_open(name, O_RDWR, 0);
+	if (IS_ERR(file)) {
+		error = PTR_ERR(file);
+		goto free_od;
 	}
 
-	cdev = inode->i_cdev;
-	if (!cdev) {
-		OSD_ERR("Before mounting an OSD Based filesystem\n");
-		OSD_ERR("  user-mode must open+close the %s device\n", name);
-		OSD_ERR("  Example: bash: echo < %s\n", name);
-		goto out;
+	if (file->f_op != &osd_fops){
+		error = -EINVAL;
+		goto close_file;
 	}
 
-	/* The Magic wand. Is it our char-dev */
-	/* TODO: Support sg devices */
-	if (cdev->owner != THIS_MODULE) {
-		OSD_ERR("Error mounting %s - is not an OSD device\n", name);
-		goto out;
-	}
+	oud = file->private_data;
 
-	oud = container_of(cdev, struct osd_uld_device, cdev);
+	*od = oud->od;
+	od->file = file;
 
-	__uld_get(oud);
-	error = 0;
+	return od;
 
-out:
-	path_put(&path);
-	return error ? ERR_PTR(error) : &oud->od;
+close_file:
+	fput(file);
+free_od:
+	kfree(od);
+	return ERR_PTR(error);
 }
 EXPORT_SYMBOL(osduld_path_lookup);
 
 void osduld_put_device(struct osd_dev *od)
 {
-	if (od) {
-		struct osd_uld_device *oud = container_of(od,
-						struct osd_uld_device, od);
 
-		__uld_put(oud);
+	if (od && !IS_ERR(od)) {
+		struct osd_uld_device *oud = od->file->private_data;
+
+		BUG_ON(od->scsi_device != oud->od.scsi_device);
+
+		fput(od->file);
+		kfree(od);
 	}
 }
 EXPORT_SYMBOL(osduld_put_device);
diff --git a/include/scsi/osd_initiator.h b/include/scsi/osd_initiator.h
index b44dc53..02bd9f7 100644
--- a/include/scsi/osd_initiator.h
+++ b/include/scsi/osd_initiator.h
@@ -48,6 +48,7 @@ enum osd_std_version {
  */
 struct osd_dev {
 	struct scsi_device *scsi_device;
+	struct file *file;
 	unsigned def_timeout;
 
 #ifdef OSD_VER1_SUPPORT
-- 
1.6.2.1

[PATCH 09/10] libosd: Use REQ_QUIET requests.

[Boaz Harrosh]

libosd has it's own sense decoding and printout. Don't
let scsi_lib duplicate that printout. (Which is done wrong
in regard to osd commands)

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 drivers/scsi/osd/osd_initiator.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index 3959797..71341ad 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -1340,6 +1340,8 @@ static int _init_blk_request(struct osd_request *or,
 
 	or->request = req;
 	req->cmd_type = REQ_TYPE_BLOCK_PC;
+	req->cmd_flags |= REQ_QUIET;
+
 	req->timeout = or->timeout;
 	req->retries = or->retries;
 	req->sense = or->sense;
-- 
1.6.2.1

[PATCH 10/10] osd: Remove out-of-tree left overs

[Boaz Harrosh]

* Delete Makefile. It is only used for out-of-tree compilation
  and was never needed. It slipped in by mistake.
* Remove from Kbuild all the out of tree stuff as promised.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 drivers/scsi/osd/Kbuild   |   25 -------------------------
 drivers/scsi/osd/Makefile |   37 -------------------------------------
 2 files changed, 0 insertions(+), 62 deletions(-)
 delete mode 100755 drivers/scsi/osd/Makefile

diff --git a/drivers/scsi/osd/Kbuild b/drivers/scsi/osd/Kbuild
index 0e207aa..5fd73d7 100644
--- a/drivers/scsi/osd/Kbuild
+++ b/drivers/scsi/osd/Kbuild
@@ -11,31 +11,6 @@
 # it under the terms of the GNU General Public License version 2
 #
 
-ifneq ($(OSD_INC),)
-# we are built out-of-tree Kconfigure everything as on
-
-CONFIG_SCSI_OSD_INITIATOR=m
-ccflags-y += -DCONFIG_SCSI_OSD_INITIATOR -DCONFIG_SCSI_OSD_INITIATOR_MODULE
-
-CONFIG_SCSI_OSD_ULD=m
-ccflags-y += -DCONFIG_SCSI_OSD_ULD -DCONFIG_SCSI_OSD_ULD_MODULE
-
-# CONFIG_SCSI_OSD_DPRINT_SENSE =
-#	0 - no print of errors
-#	1 - print errors
-#	2 - errors + warrnings
-ccflags-y += -DCONFIG_SCSI_OSD_DPRINT_SENSE=1
-
-# Uncomment to turn debug on
-# ccflags-y += -DCONFIG_SCSI_OSD_DEBUG
-
-# if we are built out-of-tree and the hosting kernel has OSD headers
-# then "ccflags-y +=" will not pick the out-off-tree headers. Only by doing
-# this it will work. This might break in future kernels
-LINUXINCLUDE := -I$(OSD_INC) $(LINUXINCLUDE)
-
-endif
-
 # libosd.ko - osd-initiator library
 libosd-y := osd_initiator.o
 obj-$(CONFIG_SCSI_OSD_INITIATOR) += libosd.o
diff --git a/drivers/scsi/osd/Makefile b/drivers/scsi/osd/Makefile
deleted file mode 100755
index d905344..0000000
--- a/drivers/scsi/osd/Makefile
+++ /dev/null
@@ -1,37 +0,0 @@
-#
-# Makefile for the OSD modules (out of tree)
-#
-# Copyright (C) 2008 Panasas Inc.  All rights reserved.
-#
-# Authors:
-#   Boaz Harrosh <bharrosh@panasas.com>
-#   Benny Halevy <bhalevy@panasas.com>
-#
-# This program is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License version 2
-#
-# This Makefile is used to call the kernel Makefile in case of an out-of-tree
-# build.
-# $KSRC should point to a Kernel source tree otherwise host's default is
-# used. (eg. /lib/modules/`uname -r`/build)
-
-# include path for out-of-tree Headers
-OSD_INC ?= `pwd`/../../../include
-
-# allow users to override these
-# e.g. to compile for a kernel that you aren't currently running
-KSRC ?= /lib/modules/$(shell uname -r)/build
-KBUILD_OUTPUT ?=
-ARCH ?=
-V ?= 0
-
-# this is the basic Kbuild out-of-tree invocation, with the M= option
-KBUILD_BASE = +$(MAKE) -C $(KSRC) M=`pwd` KBUILD_OUTPUT=$(KBUILD_OUTPUT) ARCH=$(ARCH) V=$(V)
-
-all: libosd
-
-libosd: ;
-	$(KBUILD_BASE) OSD_INC=$(OSD_INC) modules
-
-clean:
-	$(KBUILD_BASE) clean
-- 
1.6.2.1

Re: [PATCH 0/] OSD changes for 2.6.31

[Jeff Garzik]

Boaz Harrosh wrote:
> These are the accumulated osd patches for the next merge window,
> Linux 2.6.31. 
> 
> Most of these patches are a repost rebased on latest scsi-misc
> 
> List of patches:
> [PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
> [PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
> [PATCH 03/10 resend] libosd: Better printout of OSD target system information
> [PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
> [PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
> [PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter
> 
>    These are a resend. The last 2 patches are important for supporting Jeff's
>    osdblk driver. (The first 4 are old patches that thier time has come)
> 
> [PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
>    Simple wrapper
> 
> [PATCH 08/10] osduld: use filp_open() when looking up an osd-device
> 
>     This patch was inspired by Al Viro. It uses the in-kernel open-file
>     mechanism to look up the osd char-device for Kernel users like exofs
>     and later pNFS-Objlayout.
>     It was heavily tested and works well.
>     It fixes the problem we had before, of need to open+close a file handle
>     on the osd-char-device before mounting an exofs.
>     Please review
> 
> [PATCH 09/10] libosd: Use REQ_QUIET requests.
>     Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
>     because of the way we routinely receive a recovered scsi sense. Recently, for
>     reason I could not find, this condition started to print messages at:
>        scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
>     Use of a REQ_QUIET flag is needed to silence these prints.
>     OSD has it's own prints for scsi-sense returns, when needed.
> 
> [PATCH 10/10] osd: Remove out-of-tree left overs
> 
> log stat
>  drivers/scsi/osd/Kbuild          |   25 -----------
>  drivers/scsi/osd/Makefile        |   37 -----------------
>  drivers/scsi/osd/osd_initiator.c |   83 +++++++++++++++++++++++++++-----------
>  drivers/scsi/osd/osd_uld.c       |   66 ++++++++++++++----------------
>  fs/exofs/common.h                |    6 ---
>  fs/exofs/inode.c                 |    8 ++--
>  fs/exofs/osd.c                   |   26 ------------
>  include/scsi/osd_attributes.h    |   74 +++++++++++++++++++++++++++++++++-
>  include/scsi/osd_initiator.h     |   14 +++++-
>  include/scsi/osd_protocol.h      |    8 ++++
>  10 files changed, 185 insertions(+), 162 deletions(-)
>  delete mode 100755 drivers/scsi/osd/Makefile
> 
> These patches are available in the git repository at:
> 
>   git://git.open-osd.org/linux-open-osd.git osd-devel

Any chance osdblk could get added to this patchset sometime soon?

	Jeff

Re: [PATCH 0/] OSD changes for 2.6.31

[James Bottomley]

On Mon, 2009-05-25 at 10:39 -0400, Jeff Garzik wrote:
> Boaz Harrosh wrote:
> > These are the accumulated osd patches for the next merge window,
> > Linux 2.6.31. 
> > 
> > Most of these patches are a repost rebased on latest scsi-misc
> > 
> > List of patches:
> > [PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
> > [PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
> > [PATCH 03/10 resend] libosd: Better printout of OSD target system information
> > [PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
> > [PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
> > [PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter
> > 
> >    These are a resend. The last 2 patches are important for supporting Jeff's
> >    osdblk driver. (The first 4 are old patches that thier time has come)
> > 
> > [PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
> >    Simple wrapper
> > 
> > [PATCH 08/10] osduld: use filp_open() when looking up an osd-device
> > 
> >     This patch was inspired by Al Viro. It uses the in-kernel open-file
> >     mechanism to look up the osd char-device for Kernel users like exofs
> >     and later pNFS-Objlayout.
> >     It was heavily tested and works well.
> >     It fixes the problem we had before, of need to open+close a file handle
> >     on the osd-char-device before mounting an exofs.
> >     Please review
> > 
> > [PATCH 09/10] libosd: Use REQ_QUIET requests.
> >     Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
> >     because of the way we routinely receive a recovered scsi sense. Recently, for
> >     reason I could not find, this condition started to print messages at:
> >        scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
> >     Use of a REQ_QUIET flag is needed to silence these prints.
> >     OSD has it's own prints for scsi-sense returns, when needed.
> > 
> > [PATCH 10/10] osd: Remove out-of-tree left overs
> > 
> > log stat
> >  drivers/scsi/osd/Kbuild          |   25 -----------
> >  drivers/scsi/osd/Makefile        |   37 -----------------
> >  drivers/scsi/osd/osd_initiator.c |   83 +++++++++++++++++++++++++++-----------
> >  drivers/scsi/osd/osd_uld.c       |   66 ++++++++++++++----------------
> >  fs/exofs/common.h                |    6 ---
> >  fs/exofs/inode.c                 |    8 ++--
> >  fs/exofs/osd.c                   |   26 ------------
> >  include/scsi/osd_attributes.h    |   74 +++++++++++++++++++++++++++++++++-
> >  include/scsi/osd_initiator.h     |   14 +++++-
> >  include/scsi/osd_protocol.h      |    8 ++++
> >  10 files changed, 185 insertions(+), 162 deletions(-)
> >  delete mode 100755 drivers/scsi/osd/Makefile
> > 
> > These patches are available in the git repository at:
> > 
> >   git://git.open-osd.org/linux-open-osd.git osd-devel
> 
> Any chance osdblk could get added to this patchset sometime soon?

You mean for upstream submission?  You need to publish it in final form
and sign off on it first ...  Even the modified patch set is still
showing:

From: Jeff Garzik <jeff@garzik.org>

TODO: Commit log

NOT-Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
WILL-Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

James

Re: [PATCH 0/] OSD changes for 2.6.31

[Boaz Harrosh]

On 05/25/2009 05:54 PM, James Bottomley wrote:
> On Mon, 2009-05-25 at 10:39 -0400, Jeff Garzik wrote:
>> Boaz Harrosh wrote:
>>> These are the accumulated osd patches for the next merge window,
>>> Linux 2.6.31. 
>>>
>>> Most of these patches are a repost rebased on latest scsi-misc
>>>
>>> List of patches:
>>> [PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
>>> [PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
>>> [PATCH 03/10 resend] libosd: Better printout of OSD target system information
>>> [PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
>>> [PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
>>> [PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter
>>>
>>>    These are a resend. The last 2 patches are important for supporting Jeff's
>>>    osdblk driver. (The first 4 are old patches that thier time has come)
>>>
>>> [PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
>>>    Simple wrapper
>>>
>>> [PATCH 08/10] osduld: use filp_open() when looking up an osd-device
>>>
>>>     This patch was inspired by Al Viro. It uses the in-kernel open-file
>>>     mechanism to look up the osd char-device for Kernel users like exofs
>>>     and later pNFS-Objlayout.
>>>     It was heavily tested and works well.
>>>     It fixes the problem we had before, of need to open+close a file handle
>>>     on the osd-char-device before mounting an exofs.
>>>     Please review
>>>
>>> [PATCH 09/10] libosd: Use REQ_QUIET requests.
>>>     Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
>>>     because of the way we routinely receive a recovered scsi sense. Recently, for
>>>     reason I could not find, this condition started to print messages at:
>>>        scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
>>>     Use of a REQ_QUIET flag is needed to silence these prints.
>>>     OSD has it's own prints for scsi-sense returns, when needed.
>>>
>>> [PATCH 10/10] osd: Remove out-of-tree left overs
>>>
>>> log stat
>>>  drivers/scsi/osd/Kbuild          |   25 -----------
>>>  drivers/scsi/osd/Makefile        |   37 -----------------
>>>  drivers/scsi/osd/osd_initiator.c |   83 +++++++++++++++++++++++++++-----------
>>>  drivers/scsi/osd/osd_uld.c       |   66 ++++++++++++++----------------
>>>  fs/exofs/common.h                |    6 ---
>>>  fs/exofs/inode.c                 |    8 ++--
>>>  fs/exofs/osd.c                   |   26 ------------
>>>  include/scsi/osd_attributes.h    |   74 +++++++++++++++++++++++++++++++++-
>>>  include/scsi/osd_initiator.h     |   14 +++++-
>>>  include/scsi/osd_protocol.h      |    8 ++++
>>>  10 files changed, 185 insertions(+), 162 deletions(-)
>>>  delete mode 100755 drivers/scsi/osd/Makefile
>>>
>>> These patches are available in the git repository at:
>>>
>>>   git://git.open-osd.org/linux-open-osd.git osd-devel
>> Any chance osdblk could get added to this patchset sometime soon?
> 
> You mean for upstream submission?  You need to publish it in final form
> and sign off on it first ...  Even the modified patch set is still
> showing:
> 
> From: Jeff Garzik <jeff@garzik.org>
> 
> TODO: Commit log
> 
> NOT-Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
> WILL-Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
> 
> James
> 
> 

Right, and also it is for post-merge not scsi-misc, as it is dependent
on block/for-2.6.31.

James do you want to push this through your tree? or should I push it
through the exofs tree?

Thanks
Boaz

Re: [PATCH 0/] OSD changes for 2.6.31

[Jeff Garzik]

James Bottomley wrote:
> On Mon, 2009-05-25 at 10:39 -0400, Jeff Garzik wrote:
>> Boaz Harrosh wrote:
>>> These are the accumulated osd patches for the next merge window,
>>> Linux 2.6.31. 
>>>
>>> Most of these patches are a repost rebased on latest scsi-misc
>>>
>>> List of patches:
>>> [PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
>>> [PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
>>> [PATCH 03/10 resend] libosd: Better printout of OSD target system information
>>> [PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
>>> [PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
>>> [PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter
>>>
>>>    These are a resend. The last 2 patches are important for supporting Jeff's
>>>    osdblk driver. (The first 4 are old patches that thier time has come)
>>>
>>> [PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
>>>    Simple wrapper
>>>
>>> [PATCH 08/10] osduld: use filp_open() when looking up an osd-device
>>>
>>>     This patch was inspired by Al Viro. It uses the in-kernel open-file
>>>     mechanism to look up the osd char-device for Kernel users like exofs
>>>     and later pNFS-Objlayout.
>>>     It was heavily tested and works well.
>>>     It fixes the problem we had before, of need to open+close a file handle
>>>     on the osd-char-device before mounting an exofs.
>>>     Please review
>>>
>>> [PATCH 09/10] libosd: Use REQ_QUIET requests.
>>>     Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
>>>     because of the way we routinely receive a recovered scsi sense. Recently, for
>>>     reason I could not find, this condition started to print messages at:
>>>        scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
>>>     Use of a REQ_QUIET flag is needed to silence these prints.
>>>     OSD has it's own prints for scsi-sense returns, when needed.
>>>
>>> [PATCH 10/10] osd: Remove out-of-tree left overs
>>>
>>> log stat
>>>  drivers/scsi/osd/Kbuild          |   25 -----------
>>>  drivers/scsi/osd/Makefile        |   37 -----------------
>>>  drivers/scsi/osd/osd_initiator.c |   83 +++++++++++++++++++++++++++-----------
>>>  drivers/scsi/osd/osd_uld.c       |   66 ++++++++++++++----------------
>>>  fs/exofs/common.h                |    6 ---
>>>  fs/exofs/inode.c                 |    8 ++--
>>>  fs/exofs/osd.c                   |   26 ------------
>>>  include/scsi/osd_attributes.h    |   74 +++++++++++++++++++++++++++++++++-
>>>  include/scsi/osd_initiator.h     |   14 +++++-
>>>  include/scsi/osd_protocol.h      |    8 ++++
>>>  10 files changed, 185 insertions(+), 162 deletions(-)
>>>  delete mode 100755 drivers/scsi/osd/Makefile
>>>
>>> These patches are available in the git repository at:
>>>
>>>   git://git.open-osd.org/linux-open-osd.git osd-devel
>> Any chance osdblk could get added to this patchset sometime soon?
> 
> You mean for upstream submission?  You need to publish it in final form
> and sign off on it first ...  Even the modified patch set is still
> showing:
> 
> From: Jeff Garzik <jeff@garzik.org>
> 
> TODO: Commit log
> 
> NOT-Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
> WILL-Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>

Yes, that should be affected, hopefully, by the answer to 
<4A15D883.1060104@garzik.org> from May 21.

The main point is to get it queued somewhere.  I'm not particular about 
which queue or patchset.

	Jeff

Re: [PATCH 0/] OSD changes for 2.6.31

[Boaz Harrosh]

On 05/24/2009 07:58 PM, Boaz Harrosh wrote:
> These are the accumulated osd patches for the next merge window,
> Linux 2.6.31. 
> 
> Most of these patches are a repost rebased on latest scsi-misc
> 
> List of patches:
> [PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
> [PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
> [PATCH 03/10 resend] libosd: Better printout of OSD target system information
> [PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
> [PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
> [PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter
> 
>    These are a resend. The last 2 patches are important for supporting Jeff's
>    osdblk driver. (The first 4 are old patches that thier time has come)
> 
> [PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
>    Simple wrapper
> 
> [PATCH 08/10] osduld: use filp_open() when looking up an osd-device
> 
>     This patch was inspired by Al Viro. It uses the in-kernel open-file
>     mechanism to look up the osd char-device for Kernel users like exofs
>     and later pNFS-Objlayout.
>     It was heavily tested and works well.
>     It fixes the problem we had before, of need to open+close a file handle
>     on the osd-char-device before mounting an exofs.
>     Please review
> 
> [PATCH 09/10] libosd: Use REQ_QUIET requests.
>     Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
>     because of the way we routinely receive a recovered scsi sense. Recently, for
>     reason I could not find, this condition started to print messages at:
>        scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
>     Use of a REQ_QUIET flag is needed to silence these prints.
>     OSD has it's own prints for scsi-sense returns, when needed.
> 
> [PATCH 10/10] osd: Remove out-of-tree left overs
> 

The merge window has risen upon us and these patches are still not
in scsi-misc. (Luckily they where in linux-next through the osd tree)
If these will not go into 2.6.31 I'm very screwed, with the pNFS stuff,
Please reconsider?

Thanks in advance
Boaz

Re: [PATCH 0/] OSD changes for 2.6.31

[James Bottomley]

On Wed, 2009-06-10 at 12:10 +0300, Boaz Harrosh wrote:
> On 05/24/2009 07:58 PM, Boaz Harrosh wrote:
> > These are the accumulated osd patches for the next merge window,
> > Linux 2.6.31. 
> > 
> > Most of these patches are a repost rebased on latest scsi-misc
> > 
> > List of patches:
> > [PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
> > [PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
> > [PATCH 03/10 resend] libosd: Better printout of OSD target system information
> > [PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
> > [PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
> > [PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter
> > 
> >    These are a resend. The last 2 patches are important for supporting Jeff's
> >    osdblk driver. (The first 4 are old patches that thier time has come)
> > 
> > [PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
> >    Simple wrapper
> > 
> > [PATCH 08/10] osduld: use filp_open() when looking up an osd-device
> > 
> >     This patch was inspired by Al Viro. It uses the in-kernel open-file
> >     mechanism to look up the osd char-device for Kernel users like exofs
> >     and later pNFS-Objlayout.
> >     It was heavily tested and works well.
> >     It fixes the problem we had before, of need to open+close a file handle
> >     on the osd-char-device before mounting an exofs.
> >     Please review
> > 
> > [PATCH 09/10] libosd: Use REQ_QUIET requests.
> >     Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
> >     because of the way we routinely receive a recovered scsi sense. Recently, for
> >     reason I could not find, this condition started to print messages at:
> >        scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
> >     Use of a REQ_QUIET flag is needed to silence these prints.
> >     OSD has it's own prints for scsi-sense returns, when needed.
> > 
> > [PATCH 10/10] osd: Remove out-of-tree left overs
> > 
> 
> The merge window has risen upon us and these patches are still not
> in scsi-misc. (Luckily they where in linux-next through the osd tree)
> If these will not go into 2.6.31 I'm very screwed, with the pNFS stuff,
> Please reconsider?

I didn't reject them ... they just contain non scsi bits in fs/exofs.
However, since they're tightly entangled, I suppose the scsi-misc tree
will do ... just testing them out now.

James

Re: [PATCH 0/] OSD changes for 2.6.31

[Boaz Harrosh]

On 06/10/2009 05:03 PM, James Bottomley wrote:
> On Wed, 2009-06-10 at 12:10 +0300, Boaz Harrosh wrote:
>> On 05/24/2009 07:58 PM, Boaz Harrosh wrote:
>>> These are the accumulated osd patches for the next merge window,
>>> Linux 2.6.31. 
>>>
>>> Most of these patches are a repost rebased on latest scsi-misc
>>>
>>> List of patches:
>>> [PATCH 01/10 resend] libosd: OSD2r05: Additional command enums                                                        
>>> [PATCH 02/10 resend] libosd: OSD2r05: Attribute definitions
>>> [PATCH 03/10 resend] libosd: Better printout of OSD target system information
>>> [PATCH 04/10 resend] libosd: osd_req_{read,write}_kern new API
>>> [PATCH 05/10 resend] libosd: Let _osd_req_finalize_data_integrity receive number of out_bytes
>>> [PATCH 06/10 resend] libosd: osd_req_{read,write} takes a length parameter
>>>
>>>    These are a resend. The last 2 patches are important for supporting Jeff's
>>>    osdblk driver. (The first 4 are old patches that thier time has come)
>>>
>>> [PATCH 07/10] libosd: Define an osd_dev wrapper to retrieve the request_queue
>>>    Simple wrapper
>>>
>>> [PATCH 08/10] osduld: use filp_open() when looking up an osd-device
>>>
>>>     This patch was inspired by Al Viro. It uses the in-kernel open-file
>>>     mechanism to look up the osd char-device for Kernel users like exofs
>>>     and later pNFS-Objlayout.
>>>     It was heavily tested and works well.
>>>     It fixes the problem we had before, of need to open+close a file handle
>>>     on the osd-char-device before mounting an exofs.
>>>     Please review
>>>
>>> [PATCH 09/10] libosd: Use REQ_QUIET requests.
>>>     Recent scsi_lib changes, caused a dmesg span in exofs+osd regular use,
>>>     because of the way we routinely receive a recovered scsi sense. Recently, for
>>>     reason I could not find, this condition started to print messages at:
>>>        scsi_io_comlition:778 (look for "== RECOVERED_ERROR")
>>>     Use of a REQ_QUIET flag is needed to silence these prints.
>>>     OSD has it's own prints for scsi-sense returns, when needed.
>>>
>>> [PATCH 10/10] osd: Remove out-of-tree left overs
>>>
>> The merge window has risen upon us and these patches are still not
>> in scsi-misc. (Luckily they where in linux-next through the osd tree)
>> If these will not go into 2.6.31 I'm very screwed, with the pNFS stuff,
>> Please reconsider?
> 
> I didn't reject them ... they just contain non scsi bits in fs/exofs.
> However, since they're tightly entangled, I suppose the scsi-misc tree
> will do ... just testing them out now.
> 

Yes there is a comment about that in the patch commit log. The API
changed and I changed all in-tree users in one patch. (Since
I'm the maintainer of exofs, then naturally the Ack). I made
sure it will not conflict.

Some exofs patches are post-merge pending once scsi-misc will get
in.

> James
> 
> 

Thanks James
Boaz