9cdd694a797623b3d0f76ee69304dd85 linux-loop-hvr-2.4.16.0.patch

[packages/kernel.git] / linux-2.4.13-pre6-lvm-1.0.1rc4cvs.patch

--- linux/include/linux/lvm.h.orig	Tue Oct 23 13:55:20 2001
+++ linux/include/linux/lvm.h	Tue Oct 23 13:55:51 2001
@@ -9,7 +9,7 @@
  * May-July 1998
  * January-March,July,September,October,Dezember 1999
  * January,February,July,November 2000
- * January-March 2001
+ * January-March,June,July 2001
  *
  * lvm is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -67,6 +67,10 @@
  *    01/03/2001 - Revert to IOP10 and add VG_CREATE_OLD call for compatibility
  *    08/03/2001 - new lv_t (in core) version number 5: changed page member
  *                 to (struct kiobuf *) to use for COW exception table io
+ *    26/03/2001 - changed lv_v4 to lv_v5 in structure definition (HM)
+ *    21/06/2001 - changed BLOCK_SIZE back to 1024 for non S/390
+ *    22/06/2001 - added Andreas Dilger's PE on 4k boundary alignment enhancements
+ *    19/07/2001 - added rwsem compatibility macros for 2.2 kernels
  *
  */
 
@@ -74,8 +78,8 @@
 #ifndef _LVM_H_INCLUDE
 #define _LVM_H_INCLUDE
 
-#define LVM_RELEASE_NAME "0.9.1_beta6"
-#define LVM_RELEASE_DATE "12/03/2001"
+#define LVM_RELEASE_NAME "1.0.1-rc4"
+#define LVM_RELEASE_DATE "03/10/2001"
 
 #define	_LVM_KERNEL_H_VERSION	"LVM "LVM_RELEASE_NAME" ("LVM_RELEASE_DATE")"
 
@@ -102,16 +106,14 @@
    #define DEBUG_DEVICE
    #define DEBUG_KFREE
  */
-#endif				/* #ifdef __KERNEL__ */
 
-#ifndef __KERNEL__
-#define __KERNEL__
 #include <linux/kdev_t.h>
 #include <linux/list.h>
-#undef __KERNEL__
 #else
+#define __KERNEL__
 #include <linux/kdev_t.h>
 #include <linux/list.h>
+#undef __KERNEL__
 #endif				/* #ifndef __KERNEL__ */
 
 #include <asm/types.h>
@@ -123,6 +125,7 @@
 #include <asm/semaphore.h>
 #endif				/* #ifdef __KERNEL__ */
 
+
 #include <asm/page.h>
 
 #if !defined ( LVM_BLK_MAJOR) || !defined ( LVM_CHAR_MAJOR)
@@ -136,31 +139,18 @@
 #ifdef CONFIG_ARCH_S390
 #define BLOCK_SIZE	4096
 #else
-#define BLOCK_SIZE	512
+#define BLOCK_SIZE	1024
 #endif
 
 #ifndef	SECTOR_SIZE
 #define SECTOR_SIZE	512
 #endif
 
-#define LVM_STRUCT_VERSION	1	/* structure version */
+/* structure version */
+#define LVM_STRUCT_VERSION 1
 
 #define	LVM_DIR_PREFIX	"/dev/"
 
-/* set the default structure version */
-#if ( LVM_STRUCT_VERSION == 1)
-#define pv_t pv_v2_t
-#define lv_t lv_v5_t
-#define vg_t vg_v3_t
-#define pv_disk_t pv_disk_v2_t
-#define lv_disk_t lv_disk_v3_t
-#define vg_disk_t vg_disk_v2_t
-#define lv_block_exception_t lv_block_exception_v1_t
-#define lv_COW_table_disk_t lv_COW_table_disk_v1_t
-#endif
-
-
-
 /*
  * i/o protocol version
  *
@@ -234,40 +224,11 @@
  */
 
 /* DONT TOUCH THESE !!! */
-/* base of PV structure in disk partition */
-#define	LVM_PV_DISK_BASE  	0L
 
-/* size reserved for PV structure on disk */
-#define	LVM_PV_DISK_SIZE  	1024L
 
-/* base of VG structure in disk partition */
-#define	LVM_VG_DISK_BASE  	LVM_PV_DISK_SIZE
 
-/* size reserved for VG structure */
-#define	LVM_VG_DISK_SIZE  	( 9 * 512L)
-
-/* size reserved for timekeeping */
-#define	LVM_TIMESTAMP_DISK_BASE	( LVM_VG_DISK_BASE +  LVM_VG_DISK_SIZE)
-#define	LVM_TIMESTAMP_DISK_SIZE	512L	/* reserved for timekeeping */
-
-/* name list of physical volumes on disk */
-#define	LVM_PV_UUIDLIST_DISK_BASE ( LVM_TIMESTAMP_DISK_BASE + \
-                                    LVM_TIMESTAMP_DISK_SIZE)
-
-/* now for the dynamically calculated parts of the VGDA */
-#define	LVM_LV_DISK_OFFSET(a, b) ( (a)->lv_on_disk.base + \
-                                   sizeof ( lv_disk_t) * b)
-#define	LVM_DISK_SIZE(pv) 	 ( (pv)->pe_on_disk.base + \
-                                   (pv)->pe_on_disk.size)
-#define	LVM_PE_DISK_OFFSET(pe, pv)	( pe * pv->pe_size + \
-					  ( LVM_DISK_SIZE ( pv) / SECTOR_SIZE))
-#define	LVM_PE_ON_DISK_BASE(pv) \
-   { int rest; \
-     pv->pe_on_disk.base = pv->lv_on_disk.base + pv->lv_on_disk.size; \
-     if ( ( rest = pv->pe_on_disk.base % SECTOR_SIZE) != 0) \
-        pv->pe_on_disk.base += ( SECTOR_SIZE - rest); \
-   }
-/* END default disk spaces and offsets for PVs */
+
+
 
 
 /*
@@ -299,9 +260,8 @@
 #define	LVM_MAX_STRIPES		128	/* max # of stripes */
 #define	LVM_MAX_SIZE            ( 1024LU * 1024 / SECTOR_SIZE * 1024 * 1024)	/* 1TB[sectors] */
 #define	LVM_MAX_MIRRORS    	2	/* future use */
-#define	LVM_MIN_READ_AHEAD	0	/* minimum read ahead sectors */
-#define	LVM_DEFAULT_READ_AHEAD	1024	/* default read ahead sectors for 512k scsi segments */
-#define	LVM_MAX_READ_AHEAD	10000	/* maximum read ahead sectors */
+#define	LVM_MIN_READ_AHEAD	2	/* minimum read ahead sectors */
+#define	LVM_MAX_READ_AHEAD	120	/* maximum read ahead sectors */
 #define	LVM_MAX_LV_IO_TIMEOUT	60	/* seconds I/O timeout (future use) */
 #define	LVM_PARTITION           0xfe	/* LVM partition id */
 #define	LVM_NEW_PARTITION       0x8e	/* new LVM partition id (10/09/1999) */
@@ -312,25 +272,12 @@
 #define	LVM_SNAPSHOT_MIN_CHUNK	(PAGE_SIZE/1024)	/* 4 or 8 KB */
 
 #define	UNDEF	-1
-#define FALSE	0
-#define TRUE	1
-
-
-#define LVM_GET_COW_TABLE_CHUNKS_PER_PE(vg, lv) ( \
-	vg->pe_size / lv->lv_chunk_size)
-
-#define LVM_GET_COW_TABLE_ENTRIES_PER_PE(vg, lv) ( \
-{ \
-	int COW_table_entries_per_PE; \
-	int COW_table_chunks_per_PE; \
-\
-	COW_table_entries_per_PE = LVM_GET_COW_TABLE_CHUNKS_PER_PE(vg, lv); \
-	COW_table_chunks_per_PE = ( COW_table_entries_per_PE * sizeof(lv_COW_table_disk_t) / SECTOR_SIZE + lv->lv_chunk_size - 1) / lv->lv_chunk_size; \
-	COW_table_entries_per_PE - COW_table_chunks_per_PE;})
-
 
 /*
  * ioctls
+ * FIXME: the last parameter to _IO{W,R,WR} is a data type.  The macro will
+ *	  expand this using sizeof(), so putting "1" there is misleading
+ *	  because sizeof(1) = sizeof(int) = sizeof(2) = 4 on a 32-bit machine!
  */
 /* volume group */
 #define	VG_CREATE_OLD           _IOW ( 0xfe, 0x00, 1)
@@ -394,7 +341,12 @@
 #endif
 
 /* lock the logical volume manager */
+#if LVM_DRIVER_IOP_VERSION > 11
+#define	LVM_LOCK_LVM            _IO ( 0xfe, 0x9A)
+#else
+/* This is actually the same as _IO ( 0xff, 0x00), oops.  Remove for IOP 12+ */
 #define	LVM_LOCK_LVM            _IO ( 0xfe, 0x100)
+#endif
 /* END ioctls */
 
 
@@ -445,21 +397,21 @@
 #define	UUID_LEN		32	/* don't change!!! */
 
 /* copy on write tables in disk format */
-typedef struct {
+typedef struct lv_COW_table_disk_v1 {
 	uint64_t pv_org_number;
 	uint64_t pv_org_rsector;
 	uint64_t pv_snap_number;
 	uint64_t pv_snap_rsector;
-} lv_COW_table_disk_v1_t;
+} lv_COW_table_disk_t;
 
 /* remap physical sector/rdev pairs including hash */
-typedef struct {
+typedef struct lv_block_exception_v1 {
 	struct list_head hash;
 	uint32_t rsector_org;
 	kdev_t   rdev_org;
 	uint32_t rsector_new;
 	kdev_t   rdev_new;
-} lv_block_exception_v1_t;
+} lv_block_exception_t;
 
 /* disk stored pe information */
 typedef struct {
@@ -475,37 +427,11 @@
 
 
 /*
- * Structure Physical Volume (PV) Version 1
+ * physical volume structures
  */
 
 /* core */
-typedef struct {
-	char id[2];		/* Identifier */
-	unsigned short version;	/* HM lvm version */
-	lvm_disk_data_t pv_on_disk;
-	lvm_disk_data_t vg_on_disk;
-	lvm_disk_data_t pv_namelist_on_disk;
-	lvm_disk_data_t lv_on_disk;
-	lvm_disk_data_t pe_on_disk;
-	char pv_name[NAME_LEN];
-	char vg_name[NAME_LEN];
-	char system_id[NAME_LEN];	/* for vgexport/vgimport */
-	kdev_t pv_dev;
-	uint pv_number;
-	uint pv_status;
-	uint pv_allocatable;
-	uint pv_size;		/* HM */
-	uint lv_cur;
-	uint pe_size;
-	uint pe_total;
-	uint pe_allocated;
-	uint pe_stale;		/* for future use */
-	pe_disk_t *pe;		/* HM */
-	struct inode *inode;	/* HM */
-} pv_v1_t;
-
-/* core */
-typedef struct {
+typedef struct pv_v2 {
 	char id[2];		/* Identifier */
 	unsigned short version;	/* HM lvm version */
 	lvm_disk_data_t pv_on_disk;
@@ -527,36 +453,17 @@
 	uint pe_allocated;
 	uint pe_stale;		/* for future use */
 	pe_disk_t *pe;		/* HM */
-	struct inode *inode;	/* HM */
+	struct block_device *bd;
 	char pv_uuid[UUID_LEN+1];
-} pv_v2_t;
 
+#ifndef __KERNEL__
+	uint32_t pe_start;	/* in sectors */
+#endif
+} pv_t;
 
-/* disk */
-typedef struct {
-	uint8_t id[2];		/* Identifier */
-	uint16_t version;		/* HM lvm version */
-	lvm_disk_data_t pv_on_disk;
-	lvm_disk_data_t vg_on_disk;
-	lvm_disk_data_t pv_namelist_on_disk;
-	lvm_disk_data_t lv_on_disk;
-	lvm_disk_data_t pe_on_disk;
-	uint8_t pv_name[NAME_LEN];
-	uint8_t vg_name[NAME_LEN];
-	uint8_t system_id[NAME_LEN];	/* for vgexport/vgimport */
-	uint32_t pv_major;
-	uint32_t pv_number;
-	uint32_t pv_status;
-	uint32_t pv_allocatable;
-	uint32_t pv_size;		/* HM */
-	uint32_t lv_cur;
-	uint32_t pe_size;
-	uint32_t pe_total;
-	uint32_t pe_allocated;
-} pv_disk_v1_t;
 
 /* disk */
-typedef struct {
+typedef struct pv_disk_v2 {
 	uint8_t id[2];		/* Identifier */
 	uint16_t version;		/* HM lvm version */
 	lvm_disk_data_t pv_on_disk;
@@ -576,7 +483,11 @@
 	uint32_t pe_size;
 	uint32_t pe_total;
 	uint32_t pe_allocated;
-} pv_disk_v2_t;
+	
+	/* new in struct version 2 */
+	uint32_t pe_start;	        /* in sectors */
+
+} pv_disk_t;
 
 
 /*
@@ -600,7 +511,7 @@
 } le_remap_req_t;
 
 typedef struct lv_bmap {
-	ulong lv_block;
+	uint32_t lv_block;
 	dev_t lv_dev;
 } lv_bmap_t;
 
@@ -609,7 +520,7 @@
  */
 
 /* core */
-typedef struct lv_v4 {
+typedef struct lv_v5 {
 	char lv_name[NAME_LEN];
 	char vg_name[NAME_LEN];
 	uint lv_access;
@@ -632,9 +543,9 @@
 	uint lv_read_ahead;
 
 	/* delta to version 1 starts here */
-	struct lv_v4 *lv_snapshot_org;
-	struct lv_v4 *lv_snapshot_prev;
-	struct lv_v4 *lv_snapshot_next;
+	struct lv_v5 *lv_snapshot_org;
+	struct lv_v5 *lv_snapshot_prev;
+	struct lv_v5 *lv_snapshot_next;
 	lv_block_exception_t *lv_block_exception;
 	uint lv_remap_ptr;
 	uint lv_remap_end;
@@ -643,22 +554,22 @@
 #ifdef __KERNEL__
 	struct kiobuf *lv_iobuf;
 	struct kiobuf *lv_COW_table_iobuf;
-	struct semaphore lv_snapshot_sem;
+	struct rw_semaphore lv_lock;
 	struct list_head *lv_snapshot_hash_table;
 	uint32_t lv_snapshot_hash_table_size;
 	uint32_t lv_snapshot_hash_mask;
 	wait_queue_head_t lv_snapshot_wait;
 	int	lv_snapshot_use_rate;
-	void	*vg;
+	struct vg_v3	*vg;
 
 	uint lv_allocated_snapshot_le;
 #else
 	char dummy[200];
 #endif
-} lv_v5_t;
+} lv_t;
 
 /* disk */
-typedef struct {
+typedef struct lv_disk_v3 {
 	uint8_t lv_name[NAME_LEN];
 	uint8_t vg_name[NAME_LEN];
 	uint32_t lv_access;
@@ -680,36 +591,14 @@
 	uint32_t lv_allocation;
 	uint32_t lv_io_timeout;	/* for future use */
 	uint32_t lv_read_ahead;	/* HM */
-} lv_disk_v3_t;
+} lv_disk_t;
 
 /*
  * Structure Volume Group (VG) Version 1
  */
 
 /* core */
-typedef struct {
-	char vg_name[NAME_LEN];	/* volume group name */
-	uint vg_number;		/* volume group number */
-	uint vg_access;		/* read/write */
-	uint vg_status;		/* active or not */
-	uint lv_max;		/* maximum logical volumes */
-	uint lv_cur;		/* current logical volumes */
-	uint lv_open;		/* open    logical volumes */
-	uint pv_max;		/* maximum physical volumes */
-	uint pv_cur;		/* current physical volumes FU */
-	uint pv_act;		/* active physical volumes */
-	uint dummy;		/* was obsolete max_pe_per_pv */
-	uint vgda;		/* volume group descriptor arrays FU */
-	uint pe_size;		/* physical extent size in sectors */
-	uint pe_total;		/* total of physical extents */
-	uint pe_allocated;	/* allocated physical extents */
-	uint pvg_total;		/* physical volume groups FU */
-	struct proc_dir_entry *proc;
-	pv_t *pv[ABS_MAX_PV + 1];	/* physical volume struct pointers */
-	lv_t *lv[ABS_MAX_LV + 1];	/* logical  volume struct pointers */
-} vg_v1_t;
-
-typedef struct {
+typedef struct vg_v3 {
 	char vg_name[NAME_LEN];	/* volume group name */
 	uint vg_number;		/* volume group number */
 	uint vg_access;		/* read/write */
@@ -737,30 +626,11 @@
 #else
 	char dummy1[200];
 #endif
-} vg_v3_t;
+} vg_t;
 
 
 /* disk */
-typedef struct {
-	uint8_t vg_name[NAME_LEN];	/* volume group name */
-	uint32_t vg_number;	/* volume group number */
-	uint32_t vg_access;	/* read/write */
-	uint32_t vg_status;	/* active or not */
-	uint32_t lv_max;		/* maximum logical volumes */
-	uint32_t lv_cur;		/* current logical volumes */
-	uint32_t lv_open;		/* open    logical volumes */
-	uint32_t pv_max;		/* maximum physical volumes */
-	uint32_t pv_cur;		/* current physical volumes FU */
-	uint32_t pv_act;		/* active physical volumes */
-	uint32_t dummy;
-	uint32_t vgda;		/* volume group descriptor arrays FU */
-	uint32_t pe_size;		/* physical extent size in sectors */
-	uint32_t pe_total;		/* total of physical extents */
-	uint32_t pe_allocated;	/* allocated physical extents */
-	uint32_t pvg_total;	/* physical volume groups FU */
-} vg_disk_v1_t;
-
-typedef struct {
+typedef struct vg_disk_v2 {
 	uint8_t vg_uuid[UUID_LEN];	/* volume group UUID */
 	uint8_t vg_name_dummy[NAME_LEN-UUID_LEN];	/* rest of v1 VG name */
 	uint32_t vg_number;	/* volume group number */
@@ -778,7 +648,7 @@
 	uint32_t pe_total;		/* total of physical extents */
 	uint32_t pe_allocated;	/* allocated physical extents */
 	uint32_t pvg_total;	/* physical volume groups FU */
-} vg_disk_v2_t;
+} vg_disk_t;
 
 
 /*
@@ -838,4 +708,41 @@
 	int	rate;
 } lv_snapshot_use_rate_req_t;
 
+
+
+/* useful inlines */
+static inline ulong round_up(ulong n, ulong size) {
+	size--;
+	return (n + size) & ~size;
+}
+
+static inline ulong div_up(ulong n, ulong size) {
+	return round_up(n, size) / size;
+}
+
+/* FIXME: nasty capital letters */
+static int inline LVM_GET_COW_TABLE_CHUNKS_PER_PE(vg_t *vg, lv_t *lv) {
+	return vg->pe_size / lv->lv_chunk_size;
+}
+
+static int inline LVM_GET_COW_TABLE_ENTRIES_PER_PE(vg_t *vg, lv_t *lv) {
+	ulong chunks = vg->pe_size / lv->lv_chunk_size;
+	ulong entry_size = sizeof(lv_COW_table_disk_t);
+	ulong chunk_size = lv->lv_chunk_size * SECTOR_SIZE;
+	ulong entries = (vg->pe_size * SECTOR_SIZE) /
+		(entry_size + chunk_size);
+
+	if(chunks < 2)
+		return 0;
+
+	for(; entries; entries--)
+		if((div_up(entries * entry_size, chunk_size) + entries) <=
+		   chunks)
+			break;
+
+	return entries;
+}
+
+
 #endif				/* #ifndef _LVM_H_INCLUDE */
+
--- linux/drivers/md/lvm.c.orig	Tue Oct 23 13:55:17 2001
+++ linux/drivers/md/lvm.c	Tue Oct 23 13:55:51 2001
@@ -7,7 +7,7 @@
  * April-May,July-August,November 1998
  * January-March,May,July,September,October 1999
  * January,February,July,September-November 2000
- * January,February,March 2001
+ * January-April 2001
  *
  *
  * LVM driver is free software; you can redistribute it and/or modify
@@ -73,7 +73,7 @@
  *                 only other update ioctls are blocked now
  *               - fixed pv->pe to NULL for pv_status
  *               - using lv_req structure in lvm_chr_ioctl() now
- *               - fixed NULL ptr reference bug in lvm_do_lv_extendreduce()
+ *               - fixed NULL ptr reference bug in lvm_do_lv_extend_reduce()
  *                 caused by uncontiguous PV array in lvm_chr_ioctl(VG_REDUCE)
  *    09/02/1999 - changed BLKRASET and BLKRAGET in lvm_chr_ioctl() to
  *                 handle lgoical volume private read ahead sector
@@ -194,11 +194,25 @@
  *               - factored lvm_do_pv_flush out of lvm_chr_ioctl (HM)
  *    09/03/2001 - Added _lock_open_count to ensure we only drop the lock
  *                 when the locking process closes.
- *    05/04/2001 - lvm_map bugs: don't use b_blocknr/b_dev in lvm_map, it
- *                 destroys stacking devices. call b_end_io on failed maps.
- *                 (Jens Axboe)
- *    30/04/2001 - replace get_hardblock_size() with get_hardsect_size() for
- *                 2.4.4 kernel.
+ *    05/04/2001 - Defer writes to an extent that is being moved [JT]
+ *    05/04/2001 - use b_rdev and b_rsector rather than b_dev and b_blocknr in
+ *                 lvm_map() in order to make stacking devices more happy (HM)
+ *    11/04/2001 - cleaned up the pvmove queue code. I no longer retain the
+ *                 rw flag, instead WRITEA's are just dropped [JT]
+ *    30/04/2001 - added KERNEL_VERSION > 2.4.3 get_hardsect_size() rather
+ *                 than get_hardblocksize() call
+ *    03/05/2001 - Use copy_to/from_user to preserve pointers in
+ *                 lvm_do_status_by*
+ *    11/05/2001 - avoid accesses to inactive snapshot data in
+ *                 __update_hardsectsize() and lvm_do_lv_extend_reduce() (JW)
+ *    28/05/2001 - implemented missing BLKSSZGET ioctl
+ *    05/06/2001 - Move _pe_lock out of fast path for lvm_map when no PEs
+ *                 locked.  Make buffer queue flush not need locking.
+ *                 Fix lvm_user_bmap() to set b_rsector for new lvm_map(). [AED]
+ *    30/06/2001 - Speed up __update_hardsectsize() by checking if PVs have
+ *                 the same hardsectsize (very likely) before scanning all LEs
+ *                 in the LV each time.  [AED]
+ *    12/10/2001 - Use add/del_gendisk() routines in 2.4.10+
  *
  */
 
@@ -228,6 +242,7 @@
 #include <linux/locks.h>
 
 
+#include <linux/devfs_fs_kernel.h>
 #include <linux/smp_lock.h>
 #include <asm/ioctl.h>
 #include <asm/segment.h>
@@ -245,13 +260,9 @@
 
 #include "lvm-internal.h"
 
-#define	LVM_CORRECT_READ_AHEAD(a)		\
-do {						\
-	if ((a) < LVM_MIN_READ_AHEAD ||		\
-	    (a) > LVM_MAX_READ_AHEAD)		\
-		(a) = LVM_DEFAULT_READ_AHEAD;	\
-	read_ahead[MAJOR_NR] = (a);		\
-} while(0)
+#define	LVM_CORRECT_READ_AHEAD( a) \
+   if      ( a < LVM_MIN_READ_AHEAD || \
+             a > LVM_MAX_READ_AHEAD) a = LVM_MAX_READ_AHEAD;
 
 #ifndef WRITEA
 #  define WRITEA WRITE
@@ -316,6 +327,15 @@
 static void __update_hardsectsize(lv_t *lv);
 
 
+static void _queue_io(struct buffer_head *bh, int rw);
+static struct buffer_head *_dequeue_io(void);
+static void _flush_io(struct buffer_head *bh);
+
+static int _open_pv(pv_t *pv);
+static void _close_pv(pv_t *pv);
+
+static unsigned long _sectors_to_k(unsigned long sect);
+
 #ifdef LVM_HD_NAME
 void lvm_hd_name(char *, int);
 #endif
@@ -323,9 +343,7 @@
 
 
 /* variables */
-char *lvm_version = "LVM version "LVM_RELEASE_NAME"  by Heinz Mauelshagen  "
-                    "("LVM_RELEASE_DATE")\n";
-char *lvm_short_version = "version "LVM_RELEASE_NAME" ("LVM_RELEASE_DATE")";
+char *lvm_version = "LVM version "LVM_RELEASE_NAME"("LVM_RELEASE_DATE")";
 ushort lvm_iop_version = LVM_DRIVER_IOP_VERSION;
 int loadtime = 0;
 const char *const lvm_name = LVM_NAME;
@@ -334,11 +352,6 @@
 /* volume group descriptor area pointers */
 vg_t *vg[ABS_MAX_VG];
 
-static pv_t *pvp = NULL;
-static lv_t *lvp = NULL;
-static pe_t *pep = NULL;
-
-
 /* map from block minor number to VG and LV numbers */
 typedef struct {
 	int vg_number;
@@ -350,7 +363,7 @@
 /* Request structures (lvm_chr_ioctl()) */
 static pv_change_req_t pv_change_req;
 static pv_status_req_t pv_status_req;
-static pe_lock_req_t pe_lock_req;
+volatile static pe_lock_req_t pe_lock_req;
 static le_remap_req_t le_remap_req;
 static lv_req_t lv_req;
 
@@ -365,11 +378,14 @@
 static uint vg_count = 0;
 static long lvm_chr_open_count = 0;
 static DECLARE_WAIT_QUEUE_HEAD(lvm_wait);
-static DECLARE_WAIT_QUEUE_HEAD(lvm_map_wait);
 
 static spinlock_t lvm_lock = SPIN_LOCK_UNLOCKED;
 static spinlock_t lvm_snapshot_lock = SPIN_LOCK_UNLOCKED;
 
+static struct buffer_head *_pe_requests;
+static DECLARE_RWSEM(_pe_lock);
+
+
 struct file_operations lvm_chr_fops = {
 	open:		lvm_chr_open,
 	release:	lvm_chr_close,
@@ -379,7 +395,7 @@
 /* block device operations structure needed for 2.3.38? and above */
 struct block_device_operations lvm_blk_dops =
 {
-	open: 		lvm_blk_open,
+	open:		lvm_blk_open,
 	release:	lvm_blk_close,
 	ioctl:		lvm_blk_ioctl,
 };
@@ -393,29 +409,38 @@
 
 static struct gendisk lvm_gendisk =
 {
-	major:		MAJOR_NR,
-	major_name:	LVM_NAME,
-	minor_shift:	0,
-	max_p:		1,
-	part:		lvm_hd_struct,
-	sizes:		lvm_size,
-	nr_real:	MAX_LV,
+	MAJOR_NR,		/* major # */
+	LVM_NAME,		/* name of major */
+	0,			/* number of times minor is shifted
+				   to get real minor */
+	1,			/* maximum partitions per device */
+	lvm_hd_struct,		/* partition table */
+	lvm_size,		/* device size in blocks, copied
+				   to block_size[] */
+	MAX_LV,			/* number or real devices */
+	NULL,			/* internal */
+	NULL,			/* pointer to next gendisk struct (internal) */
 };
 
+
 /*
  * Driver initialization...
  */
 int lvm_init(void)
 {
-	if (register_chrdev(LVM_CHAR_MAJOR, lvm_name, &lvm_chr_fops) < 0) {
-		printk(KERN_ERR "%s -- register_chrdev failed\n", lvm_name);
+	if (devfs_register_chrdev(LVM_CHAR_MAJOR,
+				  lvm_name, &lvm_chr_fops) < 0) {
+		printk(KERN_ERR "%s -- devfs_register_chrdev failed\n",
+		       lvm_name);
 		return -EIO;
 	}
-	if (register_blkdev(MAJOR_NR, lvm_name, &lvm_blk_dops) < 0)
+	if (devfs_register_blkdev(MAJOR_NR, lvm_name, &lvm_blk_dops) < 0)
 	{
-		printk("%s -- register_blkdev failed\n", lvm_name);
-		if (unregister_chrdev(LVM_CHAR_MAJOR, lvm_name) < 0)
-			printk(KERN_ERR "%s -- unregister_chrdev failed\n", lvm_name);
+		printk("%s -- devfs_register_blkdev failed\n", lvm_name);
+		if (devfs_unregister_chrdev(LVM_CHAR_MAJOR, lvm_name) < 0)
+			printk(KERN_ERR
+			       "%s -- devfs_unregister_chrdev failed\n",
+			       lvm_name);
 		return -EIO;
 	}
 
@@ -423,6 +448,7 @@
 	lvm_init_vars();
 	lvm_geninit(&lvm_gendisk);
 
+	/* insert our gendisk at the corresponding major */
 	add_gendisk(&lvm_gendisk);
 
 #ifdef LVM_HD_NAME
@@ -433,20 +459,19 @@
 	blk_queue_make_request(BLK_DEFAULT_QUEUE(MAJOR_NR), lvm_make_request_fn);
 
 
+	/* initialise the pe lock */
+	pe_lock_req.lock = UNLOCK_PE;
+
 	/* optional read root VGDA */
 /*
    if ( *rootvg != 0) vg_read_with_pv_and_lv ( rootvg, &vg);
 */
 
-	printk(KERN_INFO
-	       "%s%s -- "
 #ifdef MODULE
-	       "Module"
+	printk(KERN_INFO "%s module loaded\n", lvm_version);
 #else
-	       "Driver"
+	printk(KERN_INFO "%s\n", lvm_version);
 #endif
-	       " successfully initialized\n",
-	       lvm_version, lvm_name);
 
 	return 0;
 } /* lvm_init() */
@@ -457,15 +482,16 @@
 
 static void lvm_cleanup(void)
 {
-	if (unregister_chrdev(LVM_CHAR_MAJOR, lvm_name) < 0) {
-		printk(KERN_ERR "%s -- unregister_chrdev failed\n", lvm_name);
-	}
-	if (unregister_blkdev(MAJOR_NR, lvm_name) < 0) {
-		printk(KERN_ERR "%s -- unregister_blkdev failed\n", lvm_name);
-	}
+	if (devfs_unregister_chrdev(LVM_CHAR_MAJOR, lvm_name) < 0)
+		printk(KERN_ERR "%s -- devfs_unregister_chrdev failed\n",
+		       lvm_name);
+	if (devfs_unregister_blkdev(MAJOR_NR, lvm_name) < 0)
+		printk(KERN_ERR "%s -- devfs_unregister_blkdev failed\n",
+		       lvm_name);
 
 
 
+	/* delete our gendisk from chain */
 	del_gendisk(&lvm_gendisk);
 
 	blk_size[MAJOR_NR] = NULL;
@@ -480,7 +506,9 @@
 	/* unregister with procfs and devfs */
 	lvm_fin_fs();
 
+#ifdef MODULE
 	printk(KERN_INFO "%s -- Module successfully deactivated\n", lvm_name);
+#endif
 
 	return;
 }	/* lvm_cleanup() */
@@ -497,8 +525,8 @@
 	lvm_lock = lvm_snapshot_lock = SPIN_LOCK_UNLOCKED;
 
 	pe_lock_req.lock = UNLOCK_PE;
-	pe_lock_req.data.lv_dev = \
-	pe_lock_req.data.pv_dev = \
+	pe_lock_req.data.lv_dev = 0;
+	pe_lock_req.data.pv_dev = 0;
 	pe_lock_req.data.pv_offset = 0;
 
 	/* Initialize VG pointers */
@@ -521,6 +549,9 @@
  *
  ********************************************************************/
 
+#define MODE_TO_STR(mode) (mode) & FMODE_READ ? "READ" : "", \
+			  (mode) & FMODE_WRITE ? "WRITE" : ""
+
 /*
  * character device open routine
  */
@@ -528,8 +559,8 @@
 {
 	int minor = MINOR(inode->i_rdev);
 
-	P_DEV("%s -- lvm_chr_open MINOR: %d  VG#: %d  mode: 0x%X  lock: %d\n",
-	      lvm_name, minor, VG_CHR(minor), file->f_mode, lock);
+	P_DEV("chr_open MINOR: %d  VG#: %d  mode: %s%s  lock: %d\n",
+	      minor, VG_CHR(minor), MODE_TO_STR(file->f_mode), lock);
 
 	/* super user validation */
 	if (!capable(CAP_SYS_ADMIN)) return -EACCES;
@@ -569,9 +600,8 @@
 	/* otherwise cc will complain about unused variables */
 	(void) lvm_lock;
 
-	P_IOCTL("%s -- lvm_chr_ioctl: command: 0x%X  MINOR: %d  "
-		"VG#: %d  mode: 0x%X\n",
-		lvm_name, command, minor, VG_CHR(minor), file->f_mode);
+	P_IOCTL("chr MINOR: %d  command: 0x%X  arg: %p  VG#: %d  mode: %s%s\n",
+		minor, command, arg, VG_CHR(minor), MODE_TO_STR(file->f_mode));
 
 #ifdef LVM_TOTAL_RESET
 	if (lvm_reset_spindown > 0) return -EACCES;
@@ -764,13 +794,10 @@
 /*
  * character device close routine
  */
-int lvm_chr_close(struct inode *inode, struct file *file)
+static int lvm_chr_close(struct inode *inode, struct file *file)
 {
-#ifdef DEBUG
-	int minor = MINOR(inode->i_rdev);
-	printk(KERN_DEBUG
-	     "%s -- lvm_chr_close   VG#: %d\n", lvm_name, VG_CHR(minor));
-#endif
+	P_DEV("chr_close MINOR: %d  VG#: %d\n",
+	      MINOR(inode->i_rdev), VG_CHR(MINOR(inode->i_rdev)));
 
 #ifdef LVM_TOTAL_RESET
 	if (lvm_reset_spindown > 0) {
@@ -784,6 +811,7 @@
 	spin_lock(&lvm_lock);
 	if(lock == current->pid) {
 		if(!_lock_open_count) {
+			P_DEV("chr_close: unlocking LVM for pid %d\n", lock);
 			lock = 0;
 			wake_up_interruptible(&lvm_wait);
 		} else
@@ -813,8 +841,8 @@
 	lv_t *lv_ptr;
 	vg_t *vg_ptr = vg[VG_BLK(minor)];
 
-	P_DEV("%s -- lvm_blk_open MINOR: %d  VG#: %d  LV#: %d  mode: 0x%X\n",
-	      lvm_name, minor, VG_BLK(minor), LV_BLK(minor), file->f_mode);
+	P_DEV("blk_open MINOR: %d  VG#: %d  LV#: %d  mode: %s%s\n",
+	      minor, VG_BLK(minor), LV_BLK(minor), MODE_TO_STR(file->f_mode));
 
 #ifdef LVM_TOTAL_RESET
 	if (lvm_reset_spindown > 0)
@@ -848,7 +876,7 @@
 
 		MOD_INC_USE_COUNT;
 
-		P_DEV("%s -- OPEN OK, LV size %d\n", lvm_name, lv_ptr->lv_size);
+		P_DEV("blk_open OK, LV size %d\n", lv_ptr->lv_size);
 
 		return 0;
 	}
@@ -868,22 +896,19 @@
 	void *arg = (void *) a;
 	struct hd_geometry *hd = (struct hd_geometry *) a;
 
-	P_IOCTL("%s -- lvm_blk_ioctl MINOR: %d  command: 0x%X  arg: %lX  "
-		"VG#: %dl  LV#: %d\n",
-		lvm_name, minor, command, (ulong) arg,
-		VG_BLK(minor), LV_BLK(minor));
+	P_IOCTL("blk MINOR: %d  command: 0x%X  arg: %p  VG#: %d  LV#: %d  "
+		"mode: %s%s\n", minor, command, arg, VG_BLK(minor),
+		LV_BLK(minor), MODE_TO_STR(file->f_mode));
 
 	switch (command) {
+	case BLKSSZGET:
+		/* get block device sector size as needed e.g. by fdisk */
+		return put_user(lvm_sectsize(inode->i_rdev), (int *) arg);
+
 	case BLKGETSIZE:
 		/* return device size */
-		P_IOCTL("%s -- lvm_blk_ioctl -- BLKGETSIZE: %u\n",
-			lvm_name, lv_ptr->lv_size);
-		if (put_user(lv_ptr->lv_size, (unsigned long *)arg))
-			return -EFAULT;
-		break;
-
-	case BLKGETSIZE64:
-		if (put_user((u64)lv_ptr->lv_size << 9, (u64 *)arg))
+		P_IOCTL("BLKGETSIZE: %u\n", lv_ptr->lv_size);
+		if (put_user(lv_ptr->lv_size, (long *)arg))
 			return -EFAULT;
 		break;
 
@@ -892,7 +917,7 @@
 		/* flush buffer cache */
 		if (!capable(CAP_SYS_ADMIN)) return -EACCES;
 
-		P_IOCTL("%s -- lvm_blk_ioctl -- BLKFLSBUF\n", lvm_name);
+		P_IOCTL("BLKFLSBUF\n");
 
 		fsync_dev(inode->i_rdev);
 		invalidate_buffers(inode->i_rdev);
@@ -903,20 +928,19 @@
 		/* set read ahead for block device */
 		if (!capable(CAP_SYS_ADMIN)) return -EACCES;
 
-		P_IOCTL("%s -- lvm_blk_ioctl -- BLKRASET: %ld sectors for %s\n",
-			lvm_name, (long) arg, kdevname(inode->i_rdev));
+		P_IOCTL("BLKRASET: %ld sectors for %s\n",
+			(long) arg, kdevname(inode->i_rdev));
 
 		if ((long) arg < LVM_MIN_READ_AHEAD ||
 		    (long) arg > LVM_MAX_READ_AHEAD)
 			return -EINVAL;
 		lv_ptr->lv_read_ahead = (long) arg;
-		read_ahead[MAJOR_NR] = lv_ptr->lv_read_ahead;
 		break;
 
 
 	case BLKRAGET:
 		/* get current read ahead setting */
-		P_IOCTL("%s -- lvm_blk_ioctl -- BLKRAGET\n", lvm_name);
+		P_IOCTL("BLKRAGET %d\n", lv_ptr->lv_read_ahead);
 		if (put_user(lv_ptr->lv_read_ahead, (long *)arg))
 			return -EFAULT;
 		break;
@@ -969,13 +993,12 @@
 		break;
 
 	case LV_BMAP:
-		/* turn logical block into (dev_t, block). non privileged. */
+		/* turn logical block into (dev_t, block).  non privileged. */
 		/* don't bmap a snapshot, since the mapping can change */
-		if (lv_ptr->lv_access & LV_SNAPSHOT)
+		if(lv_ptr->lv_access & LV_SNAPSHOT)
 			return -EPERM;
 
 		return lvm_user_bmap(inode, (struct lv_bmap *) arg);
-		break;
 
 	case LV_SET_ALLOCATION:
 		/* set allocation flags of a logical volume */
@@ -1006,11 +1029,8 @@
 	vg_t *vg_ptr = vg[VG_BLK(minor)];
 	lv_t *lv_ptr = vg_ptr->lv[LV_BLK(minor)];
 
-#ifdef DEBUG
-	printk(KERN_DEBUG
-	       "%s -- lvm_blk_close MINOR: %d  VG#: %d  LV#: %d\n",
-	       lvm_name, minor, VG_BLK(minor), LV_BLK(minor));
-#endif
+	P_DEV("blk_close MINOR: %d  VG#: %d  LV#: %d\n",
+	      minor, VG_BLK(minor), LV_BLK(minor));
 
 	if (lv_ptr->lv_open == 1) vg_ptr->lv_open--;
 	lv_ptr->lv_open--;
@@ -1063,17 +1083,17 @@
 		return -EFAULT;
 
 	memset(&bh,0,sizeof bh);
-	bh.b_rsector = block;
-	bh.b_dev = bh.b_rdev = inode->i_dev;
+	bh.b_blocknr = block;
+	bh.b_dev = bh.b_rdev = inode->i_rdev;
 	bh.b_size = lvm_get_blksize(bh.b_dev);
+	bh.b_rsector = block * (bh.b_size >> 9);
 	if ((err=lvm_map(&bh, READ)) < 0)  {
 		printk("lvm map failed: %d\n", err);
 		return -EINVAL;
 	}
 
-	return put_user(kdev_t_to_nr(bh.b_rdev), &user_result->lv_dev) ||
-	       put_user(bh.b_rsector/(bh.b_size>>9), &user_result->lv_block) ?
-	       -EFAULT : 0;
+	return (put_user(kdev_t_to_nr(bh.b_rdev), &user_result->lv_dev) ||
+		put_user(bh.b_rsector/(bh.b_size>>9), &user_result->lv_block));
 }
 
 
@@ -1081,16 +1101,68 @@
  * block device support function for /usr/src/linux/drivers/block/ll_rw_blk.c
  * (see init_module/lvm_init)
  */
-static inline void __remap_snapshot(kdev_t rdev, ulong rsector,
+static void __remap_snapshot(kdev_t rdev, ulong rsector,
 				    ulong pe_start, lv_t *lv, vg_t *vg) {
+
+	/* copy a chunk from the origin to a snapshot device */
+	down_write(&lv->lv_lock);
+
+	/* we must redo lvm_snapshot_remap_block in order to avoid a
+	   race condition in the gap where no lock was held */
 	if (!lvm_snapshot_remap_block(&rdev, &rsector, pe_start, lv) &&
 	    !lvm_snapshot_COW(rdev, rsector, pe_start, rsector, vg, lv))
 		lvm_write_COW_table_block(vg, lv);
+
+	up_write(&lv->lv_lock);
+}
+
+static inline void _remap_snapshot(kdev_t rdev, ulong rsector,
+				   ulong pe_start, lv_t *lv, vg_t *vg) {
+	int r;
+
+	/* check to see if this chunk is already in the snapshot */
+	down_read(&lv->lv_lock);
+	r = lvm_snapshot_remap_block(&rdev, &rsector, pe_start, lv);
+	up_read(&lv->lv_lock);
+
+	if (!r)
+		/* we haven't yet copied this block to the snapshot */
+		__remap_snapshot(rdev, rsector, pe_start, lv, vg);
+}
+
+
+/*
+ * extents destined for a pe that is on the move should be deferred
+ */
+static inline int _should_defer(kdev_t pv, ulong sector, uint32_t pe_size) {
+	return ((pe_lock_req.lock == LOCK_PE) &&
+		(pv == pe_lock_req.data.pv_dev) &&
+		(sector >= pe_lock_req.data.pv_offset) &&
+		(sector < (pe_lock_req.data.pv_offset + pe_size)));
 }
 
+static inline int _defer_extent(struct buffer_head *bh, int rw,
+				kdev_t pv, ulong sector, uint32_t pe_size)
+{
+	if (pe_lock_req.lock == LOCK_PE) {
+		down_read(&_pe_lock);
+		if (_should_defer(pv, sector, pe_size)) {
+			up_read(&_pe_lock);
+			down_write(&_pe_lock);
+			if (_should_defer(pv, sector, pe_size))
+				_queue_io(bh, rw);
+			up_write(&_pe_lock);
+			return 1;
+		}
+		up_read(&_pe_lock);
+	}
+	return 0;
+}
+
+
 static int lvm_map(struct buffer_head *bh, int rw)
 {
-	int minor = MINOR(bh->b_dev);
+	int minor = MINOR(bh->b_rdev);
 	ulong index;
 	ulong pe_start;
 	ulong size = bh->b_size >> 9;
@@ -1101,7 +1173,7 @@
 	lv_t *lv = vg_this->lv[LV_BLK(minor)];
 
 
-	down(&lv->lv_snapshot_sem);
+	down_read(&lv->lv_lock);
 	if (!(lv->lv_status & LV_ACTIVE)) {
 		printk(KERN_ALERT
 		       "%s - lvm_map: ll_rw_blk for inactive LV %s\n",
@@ -1119,7 +1191,7 @@
 
 	P_MAP("%s - lvm_map minor: %d  *rdev: %s  *rsector: %lu  size:%lu\n",
 	      lvm_name, minor,
-	      kdevname(bh->b_dev),
+	      kdevname(bh->b_rdev),
 	      rsector_org, size);
 
 	if (rsector_org + size > lv->lv_size) {
@@ -1130,7 +1202,7 @@
 		goto bad;
 	}
 
-remap:
+
 	if (lv->lv_stripes < 2) { /* linear mapping */
 		/* get the index */
 		index = rsector_org / vg_this->pe_size;
@@ -1167,36 +1239,33 @@
 		      rsector_map, stripe_length, stripe_index);
 	}
 
-	/* handle physical extents on the move */
-	if (pe_lock_req.lock == LOCK_PE) {
-		if (rdev_map == pe_lock_req.data.pv_dev &&
-		    rsector_map >= pe_lock_req.data.pv_offset &&
-		    rsector_map < (pe_lock_req.data.pv_offset +
-				   vg_this->pe_size)) {
-			sleep_on(&lvm_map_wait);
-			goto remap;
+	/*
+	 * Queue writes to physical extents on the move until move completes.
+	 * Don't get _pe_lock until there is a reasonable expectation that
+	 * we need to queue this request, because this is in the fast path.
+	 */
+	if (rw == WRITE || rw == WRITEA) {
+		if(_defer_extent(bh, rw, rdev_map,
+				 rsector_map, vg_this->pe_size)) {
+
+			up_read(&lv->lv_lock);
+			return 0;
 		}
-	}
 
-	/* statistic */
-	if (rw == WRITE || rw == WRITEA)
-		lv->lv_current_pe[index].writes++;
-	else
-		lv->lv_current_pe[index].reads++;
+		lv->lv_current_pe[index].writes++;	/* statistic */
+	} else
+		lv->lv_current_pe[index].reads++;	/* statistic */
 
-	/* snapshot volume exception handling on physical device
-           address base */
+	/* snapshot volume exception handling on physical device address base */
 	if (!(lv->lv_access & (LV_SNAPSHOT|LV_SNAPSHOT_ORG)))
 		goto out;
 
 	if (lv->lv_access & LV_SNAPSHOT) { /* remap snapshot */
-		if (lv->lv_block_exception)
-			lvm_snapshot_remap_block(&rdev_map, &rsector_map,
-						 pe_start, lv);
-		else
+		if (lvm_snapshot_remap_block(&rdev_map, &rsector_map,
+					     pe_start, lv) < 0)
 			goto bad;
 
-	} else if(rw == WRITE || rw == WRITEA) { /* snapshot origin */
+	} else if (rw == WRITE || rw == WRITEA) { /* snapshot origin */
 		lv_t *snap;
 
 		/* start with first snapshot and loop through all of
@@ -1209,22 +1278,20 @@
 
 			/* Serializes the COW with the accesses to the
 			   snapshot device */
-			down(&snap->lv_snapshot_sem);
-			__remap_snapshot(rdev_map, rsector_map,
+			_remap_snapshot(rdev_map, rsector_map,
 					 pe_start, snap, vg_this);
-			up(&snap->lv_snapshot_sem);
 		}
 	}
 
  out:
 	bh->b_rdev = rdev_map;
 	bh->b_rsector = rsector_map;
-	up(&lv->lv_snapshot_sem);
+	up_read(&lv->lv_lock);
 	return 1;
 
  bad:
 	buffer_IO_error(bh);
-	up(&lv->lv_snapshot_sem);
+	up_read(&lv->lv_lock);
 	return -1;
 } /* lvm_map() */
 
@@ -1258,15 +1325,10 @@
 /*
  * make request function
  */
-static int lvm_make_request_fn(request_queue_t *q, 
-			       int rw, 
-			       struct buffer_head *bh) 
-{
-	if (lvm_map(bh, rw) >= 0)
-		return 1;
-
-	buffer_IO_error(bh);
-	return 0;
+static int lvm_make_request_fn(request_queue_t *q,
+			       int rw,
+			       struct buffer_head *bh) {
+	return (lvm_map(bh, rw) <= 0) ? 0 : 1;
 }
 
 
@@ -1283,8 +1345,7 @@
 lock_try_again:
 	spin_lock(&lvm_lock);
 	if (lock != 0 && lock != current->pid) {
-		P_IOCTL("lvm_do_lock_lvm: %s is locked by pid %d ...\n",
-			lvm_name, lock);
+		P_DEV("lvm_do_lock_lvm: locked by pid %d ...\n", lock);
 		spin_unlock(&lvm_lock);
 		interruptible_sleep_on(&lvm_wait);
 		if (current->sigpending != 0)
@@ -1296,6 +1357,7 @@
 		goto lock_try_again;
 	}
 	lock = current->pid;
+	P_DEV("lvm_do_lock_lvm: locking LVM for pid %d\n", lock);
 	spin_unlock(&lvm_lock);
 	return 0;
 } /* lvm_do_lock_lvm */
@@ -1306,33 +1368,60 @@
  */
 static int lvm_do_pe_lock_unlock(vg_t *vg_ptr, void *arg)
 {
+	pe_lock_req_t new_lock;
+	struct buffer_head *bh;
 	uint p;
 
 	if (vg_ptr == NULL) return -ENXIO;
-	if (copy_from_user(&pe_lock_req, arg,
-			   sizeof(pe_lock_req_t)) != 0) return -EFAULT;
+	if (copy_from_user(&new_lock, arg, sizeof(new_lock)) != 0)
+		return -EFAULT;
 
-	switch (pe_lock_req.lock) {
+	switch (new_lock.lock) {
 	case LOCK_PE:
 		for (p = 0; p < vg_ptr->pv_max; p++) {
 			if (vg_ptr->pv[p] != NULL &&
-			    pe_lock_req.data.pv_dev ==
-			    vg_ptr->pv[p]->pv_dev)
+			    new_lock.data.pv_dev == vg_ptr->pv[p]->pv_dev)
 				break;
 		}
 		if (p == vg_ptr->pv_max) return -ENXIO;
 
-		pe_lock_req.lock = UNLOCK_PE;
+		/*
+		 * this sync releaves memory pressure to lessen the
+		 * likelyhood of pvmove being paged out - resulting in
+		 * deadlock.
+		 *
+		 * This method of doing a pvmove is broken
+		 */
 		fsync_dev(pe_lock_req.data.lv_dev);
+
+		down_write(&_pe_lock);
+		if (pe_lock_req.lock == LOCK_PE) {
+			up_write(&_pe_lock);
+			return -EBUSY;
+		}
+
+		/* Should we do to_kdev_t() on the pv_dev and lv_dev??? */
 		pe_lock_req.lock = LOCK_PE;
+		pe_lock_req.data.lv_dev = new_lock.data.lv_dev;
+		pe_lock_req.data.pv_dev = new_lock.data.pv_dev;
+		pe_lock_req.data.pv_offset = new_lock.data.pv_offset;
+		up_write(&_pe_lock);
+
+		/* some requests may have got through since the fsync */
+		fsync_dev(pe_lock_req.data.pv_dev);
 		break;
 
 	case UNLOCK_PE:
+		down_write(&_pe_lock);
 		pe_lock_req.lock = UNLOCK_PE;
-		pe_lock_req.data.lv_dev = \
-		pe_lock_req.data.pv_dev = \
+		pe_lock_req.data.lv_dev = 0;
+		pe_lock_req.data.pv_dev = 0;
 		pe_lock_req.data.pv_offset = 0;
-		wake_up(&lvm_map_wait);
+		bh = _dequeue_io();
+		up_write(&_pe_lock);
+
+		/* handle all deferred io for this PE */
+		_flush_io(bh);
 		break;
 
 	default:
@@ -1400,6 +1489,8 @@
 	}
 	/* get the volume group structure */
 	if (copy_from_user(vg_ptr, arg, sizeof(vg_t)) != 0) {
+		P_IOCTL("lvm_do_vg_create ERROR: copy VG ptr %p (%d bytes)\n",
+			arg, sizeof(vg_t));
 		kfree(vg_ptr);
 		return -EFAULT;
 	}
@@ -1409,8 +1500,9 @@
 
 	/* Validate it */
 	if (vg[VG_CHR(minor)] != NULL) {
-	    kfree(vg_ptr);
-	    return -EPERM;
+		P_IOCTL("lvm_do_vg_create ERROR: VG %d in use\n", minor);
+		kfree(vg_ptr);
+		return -EPERM;
 	}
 
 	/* we are not that active so far... */
@@ -1441,6 +1533,7 @@
 	/* get the physical volume structures */
 	vg_ptr->pv_act = vg_ptr->pv_cur = 0;
 	for (p = 0; p < vg_ptr->pv_max; p++) {
+		pv_t *pvp;
 		/* user space address */
 		if ((pvp = vg_ptr->pv[p]) != NULL) {
 			ret = lvm_do_pv_create(pvp, vg_ptr, p);
@@ -1464,9 +1557,12 @@
 	/* get the logical volume structures */
 	vg_ptr->lv_cur = 0;
 	for (l = 0; l < vg_ptr->lv_max; l++) {
+		lv_t *lvp;
 		/* user space address */
 		if ((lvp = vg_ptr->lv[l]) != NULL) {
 			if (copy_from_user(&lv, lvp, sizeof(lv_t)) != 0) {
+				P_IOCTL("ERROR: copying LV ptr %p (%d bytes)\n",
+					lvp, sizeof(lv_t));
 				lvm_do_vg_remove(minor);
 				return -EFAULT;
 			}
@@ -1488,7 +1584,7 @@
 	/* Second path to correct snapshot logical volumes which are not
 	   in place during first path above */
 	for (l = 0; l < ls; l++) {
-		lvp = snap_lv_ptr[l];
+		lv_t *lvp = snap_lv_ptr[l];
 		if (copy_from_user(&lv, lvp, sizeof(lv_t)) != 0) {
 			lvm_do_vg_remove(minor);
 			return -EFAULT;
@@ -1680,27 +1776,41 @@
  * character device support function physical volume create
  */
 static int lvm_do_pv_create(pv_t *pvp, vg_t *vg_ptr, ulong p) {
-	pv_t *pv_ptr = NULL;
+	pv_t *pv;
+	int err;
 
-	pv_ptr = vg_ptr->pv[p] = kmalloc(sizeof(pv_t),GFP_KERNEL);
-	if (pv_ptr == NULL) {
+	pv = kmalloc(sizeof(pv_t),GFP_KERNEL);
+	if (pv == NULL) {
 		printk(KERN_CRIT
 		       "%s -- PV_CREATE: kmalloc error PV at line %d\n",
 		       lvm_name, __LINE__);
 		return -ENOMEM;
 	}
-	if (copy_from_user(pv_ptr, pvp, sizeof(pv_t)) != 0) {
+
+	memset(pv, 0, sizeof(*pv));
+
+	if (copy_from_user(pv, pvp, sizeof(pv_t)) != 0) {
+		P_IOCTL("lvm_do_pv_create ERROR: copy PV ptr %p (%d bytes)\n",
+			pvp, sizeof(pv_t));
+		kfree(pv);
 		return -EFAULT;
 	}
+
+	if ((err = _open_pv(pv))) {
+		kfree(pv);
+		return err;
+	}
+
 	/* We don't need the PE list
 	   in kernel space as with LVs pe_t list (see below) */
-	pv_ptr->pe = NULL;
-	pv_ptr->pe_allocated = 0;
-	pv_ptr->pv_status = PV_ACTIVE;
+	pv->pe = NULL;
+	pv->pe_allocated = 0;
+	pv->pv_status = PV_ACTIVE;
 	vg_ptr->pv_act++;
 	vg_ptr->pv_cur++;
-	lvm_fs_create_pv(vg_ptr, pv_ptr);
+	lvm_fs_create_pv(vg_ptr, pv);
 
+	vg_ptr->pv[p] = pv;
 	return 0;
 } /* lvm_do_pv_create() */
 
@@ -1709,47 +1819,73 @@
  * character device support function physical volume remove
  */
 static int lvm_do_pv_remove(vg_t *vg_ptr, ulong p) {
-	pv_t *pv_ptr = vg_ptr->pv[p];
+	pv_t *pv = vg_ptr->pv[p];
 
-	lvm_fs_remove_pv(vg_ptr, pv_ptr);
+	lvm_fs_remove_pv(vg_ptr, pv);
 
-	vg_ptr->pe_total -= pv_ptr->pe_total;
+	vg_ptr->pe_total -= pv->pe_total;
 	vg_ptr->pv_cur--;
 	vg_ptr->pv_act--;
-#ifdef LVM_GET_INODE
-	lvm_clear_inode(pv_ptr->inode);
-#endif
-	kfree(pv_ptr);
+
+	_close_pv(pv);
+	kfree(pv);
+
 	vg_ptr->pv[p] = NULL;
 
 	return 0;
 }
 
 
-static void __update_hardsectsize(lv_t *lv) {
-	int le, e;
-	int max_hardsectsize = 0, hardsectsize;
-
-	for (le = 0; le < lv->lv_allocated_le; le++) {
-		hardsectsize = get_hardsect_size(lv->lv_current_pe[le].dev);
-		if (hardsectsize == 0)
-			hardsectsize = 512;
-		if (hardsectsize > max_hardsectsize)
-			max_hardsectsize = hardsectsize;
-	}
-
-	if (lv->lv_access & LV_SNAPSHOT) {
-		for (e = 0; e < lv->lv_remap_end; e++) {
-			hardsectsize =
-				get_hardsect_size(
-					lv->lv_block_exception[e].rdev_new);
-			if (hardsectsize == 0)
-				hardsectsize = 512;
-			if (hardsectsize > max_hardsectsize)
+static void __update_hardsectsize(lv_t *lv)
+{
+	int max_hardsectsize = 0, hardsectsize = 0;
+	int p;
+
+	/* Check PVs first to see if they all have same sector size */
+	for (p = 0; p < lv->vg->pv_cur; p++) {
+		pv_t *pv = lv->vg->pv[p];
+		if (pv && (hardsectsize = lvm_sectsize(pv->pv_dev))) {
+			if (max_hardsectsize == 0)
 				max_hardsectsize = hardsectsize;
+			else if (hardsectsize != max_hardsectsize) {
+				P_DEV("%s PV[%d] (%s) sector size %d, not %d\n",
+				      lv->lv_name, p, kdevname(pv->pv_dev),
+				      hardsectsize, max_hardsectsize);
+				break;
+			}
 		}
 	}
 
+	/* PVs have different block size, need to check each LE sector size */
+	if (hardsectsize != max_hardsectsize) {
+		int le;
+		for (le = 0; le < lv->lv_allocated_le; le++) {
+			hardsectsize = lvm_sectsize(lv->lv_current_pe[le].dev);
+			if (hardsectsize > max_hardsectsize) {
+				P_DEV("%s LE[%d] (%s) blocksize %d not %d\n",
+				      lv->lv_name, le,
+				      kdevname(lv->lv_current_pe[le].dev),
+				      hardsectsize, max_hardsectsize);
+				max_hardsectsize = hardsectsize;
+			}
+		}
+
+		/* only perform this operation on active snapshots */
+		if ((lv->lv_access & LV_SNAPSHOT) &&
+		    (lv->lv_status & LV_ACTIVE)) {
+			int e;
+			for (e = 0; e < lv->lv_remap_end; e++) {
+				hardsectsize = lvm_sectsize(lv->lv_block_exception[e].rdev_new);
+				if (hardsectsize > max_hardsectsize)
+					max_hardsectsize = hardsectsize;
+			}
+		}
+	}
+
+	if (max_hardsectsize == 0)
+		max_hardsectsize = SECTOR_SIZE;
+	P_DEV("hardblocksize for LV %s is %d\n",
+	      kdevname(lv->lv_dev), max_hardsectsize);
 	lvm_hardsectsizes[MINOR(lv->lv_dev)] = max_hardsectsize;
 }
 
@@ -1763,9 +1899,12 @@
 	lv_block_exception_t *lvbe = lv->lv_block_exception;
 	vg_t *vg_ptr = vg[VG_CHR(minor)];
 	lv_t *lv_ptr = NULL;
+	pe_t *pep;
 
-	if ((pep = lv->lv_current_pe) == NULL) return -EINVAL;
-	if (lv->lv_chunk_size > LVM_SNAPSHOT_MAX_CHUNK)
+	if (!(pep = lv->lv_current_pe))
+		return -EINVAL;
+
+	if (_sectors_to_k(lv->lv_chunk_size) > LVM_SNAPSHOT_MAX_CHUNK)
 		return -EINVAL;
 
 	for (l = 0; l < vg_ptr->lv_cur; l++) {
@@ -1797,8 +1936,8 @@
 
 	lv_status_save = lv_ptr->lv_status;
 	lv_ptr->lv_status &= ~LV_ACTIVE;
-	lv_ptr->lv_snapshot_org = \
-	lv_ptr->lv_snapshot_prev = \
+	lv_ptr->lv_snapshot_org = NULL;
+	lv_ptr->lv_snapshot_prev = NULL;
 	lv_ptr->lv_snapshot_next = NULL;
 	lv_ptr->lv_block_exception = NULL;
 	lv_ptr->lv_iobuf = NULL;
@@ -1806,7 +1945,8 @@
 	lv_ptr->lv_snapshot_hash_table = NULL;
 	lv_ptr->lv_snapshot_hash_table_size = 0;
 	lv_ptr->lv_snapshot_hash_mask = 0;
-	init_MUTEX(&lv_ptr->lv_snapshot_sem);
+	init_rwsem(&lv_ptr->lv_lock);
+
 	lv_ptr->lv_snapshot_use_rate = 0;
 
 	vg_ptr->lv[l] = lv_ptr;
@@ -1815,6 +1955,7 @@
 	   is not a snapshot logical volume */
 	if (!(lv_ptr->lv_access & LV_SNAPSHOT)) {
 		size = lv_ptr->lv_allocated_le * sizeof(pe_t);
+
 		if ((lv_ptr->lv_current_pe = vmalloc(size)) == NULL) {
 			printk(KERN_CRIT
 			       "%s -- LV_CREATE: vmalloc error LV_CURRENT_PE of %d Byte "
@@ -1826,6 +1967,8 @@
 			return -ENOMEM;
 		}
 		if (copy_from_user(lv_ptr->lv_current_pe, pep, size)) {
+			P_IOCTL("ERROR: copying PE ptr %p (%d bytes)\n",
+				pep, sizeof(size));
 			vfree(lv_ptr->lv_current_pe);
 			kfree(lv_ptr);
 			vg_ptr->lv[l] = NULL;
@@ -1847,6 +1990,15 @@
 			    vg_ptr->lv[LV_BLK(lv_ptr->lv_snapshot_minor)];
 			if (lv_ptr->lv_snapshot_org != NULL) {
 				size = lv_ptr->lv_remap_end * sizeof(lv_block_exception_t);
+
+				if(!size) {
+					printk(KERN_WARNING
+					       "%s -- zero length exception table requested\n",
+					       lvm_name);
+					kfree(lv_ptr);
+					return -EINVAL;
+				}
+
 				if ((lv_ptr->lv_block_exception = vmalloc(size)) == NULL) {
 					printk(KERN_CRIT
 					       "%s -- lvm_do_lv_create: vmalloc error LV_BLOCK_EXCEPTION "
@@ -1934,6 +2086,7 @@
 	LVM_CORRECT_READ_AHEAD(lv_ptr->lv_read_ahead);
 	vg_ptr->lv_cur++;
 	lv_ptr->lv_status = lv_status_save;
+	lv_ptr->vg = vg_ptr;
 
 	__update_hardsectsize(lv_ptr);
 
@@ -1948,7 +2101,7 @@
 		fsync_dev_lockfs(org->lv_dev);
 #endif
 
-		down(&org->lv_snapshot_sem);
+		down_write(&org->lv_lock);
 		org->lv_access |= LV_SNAPSHOT_ORG;
 		lv_ptr->lv_access &= ~LV_SNAPSHOT_ORG; /* this can only hide an userspace bug */
 
@@ -1957,7 +2110,7 @@
 		for (last = org; last->lv_snapshot_next; last = last->lv_snapshot_next);
 		lv_ptr->lv_snapshot_prev = last;
 		last->lv_snapshot_next = lv_ptr;
-		up(&org->lv_snapshot_sem);
+		up_write(&org->lv_lock);
 	}
 
 	/* activate the logical volume */
@@ -1973,14 +2126,12 @@
 
 #ifdef	LVM_VFS_ENHANCEMENT
 /* VFS function call to unlock the filesystem */
-	if (lv_ptr->lv_access & LV_SNAPSHOT) {
+	if (lv_ptr->lv_access & LV_SNAPSHOT)
 		unlockfs(lv_ptr->lv_snapshot_org->lv_dev);
-	}
 #endif
 
-	lv_ptr->vg = vg_ptr;
-
-	lvm_fs_create_lv(vg_ptr, lv_ptr);
+	lvm_gendisk.part[MINOR(lv_ptr->lv_dev)].de =
+	    lvm_fs_create_lv(vg_ptr, lv_ptr);
 	return 0;
 } /* lvm_do_lv_create() */
 
@@ -2026,7 +2177,7 @@
 		 * to the original lv before playing with it.
 		 */
 		lv_t * org = lv_ptr->lv_snapshot_org;
-		down(&org->lv_snapshot_sem);
+		down_write(&org->lv_lock);
 
 		/* remove this snapshot logical volume from the chain */
 		lv_ptr->lv_snapshot_prev->lv_snapshot_next = lv_ptr->lv_snapshot_next;
@@ -2039,7 +2190,7 @@
 		if (!org->lv_snapshot_next) {
 			org->lv_access &= ~LV_SNAPSHOT_ORG;
 		}
-		up(&org->lv_snapshot_sem);
+		up_write(&org->lv_lock);
 
 		lvm_snapshot_release(lv_ptr);
 
@@ -2060,6 +2211,7 @@
 	/* reset generic hd */
 	lvm_gendisk.part[MINOR(lv_ptr->lv_dev)].start_sect = -1;
 	lvm_gendisk.part[MINOR(lv_ptr->lv_dev)].nr_sects = 0;
+	lvm_gendisk.part[MINOR(lv_ptr->lv_dev)].de = 0;
 	lvm_size[MINOR(lv_ptr->lv_dev)] = 0;
 
 	/* reset VG/LV mapping */
@@ -2191,8 +2343,7 @@
 		new_stripe_size = new_lv->lv_allocated_le / new_lv->lv_stripes;
 		end = min(old_stripe_size, new_stripe_size);
 
-		for (i = source = dest = 0;
-		     i < new_lv->lv_stripes; i++) {
+		for (i = source = dest = 0; i < new_lv->lv_stripes; i++) {
 			for (j = 0; j < end; j++) {
 				new_lv->lv_current_pe[dest + j].reads +=
 				    old_lv->lv_current_pe[source + j].reads;
@@ -2227,23 +2378,27 @@
 
 	old_lv = vg_ptr->lv[l];
 
-	if (old_lv->lv_access & LV_SNAPSHOT)
-		r = __extend_reduce_snapshot(vg_ptr, old_lv, new_lv);
-	else
+	if (old_lv->lv_access & LV_SNAPSHOT) {
+		/* only perform this operation on active snapshots */
+		if (old_lv->lv_status & LV_ACTIVE)
+			r = __extend_reduce_snapshot(vg_ptr, old_lv, new_lv);
+		else
+			r = -EPERM;
+
+	} else
 		r = __extend_reduce(vg_ptr, old_lv, new_lv);
 
 	if(r)
 		return r;
 
 	/* copy relevent fields */
-	down(&old_lv->lv_snapshot_sem);
+	down_write(&old_lv->lv_lock);
 
 	if(new_lv->lv_access & LV_SNAPSHOT) {
-
 		size = (new_lv->lv_remap_end > old_lv->lv_remap_end) ?
 			old_lv->lv_remap_ptr : new_lv->lv_remap_end;
 		size *= sizeof(lv_block_exception_t);
-		memcpy(new_lv->lv_block_exception, 
+		memcpy(new_lv->lv_block_exception,
 		       old_lv->lv_block_exception, size);
 
 		old_lv->lv_remap_end = new_lv->lv_remap_end;
@@ -2258,7 +2413,7 @@
 		for (e = 0; e < new_lv->lv_remap_ptr; e++)
 			lvm_hash_link(new_lv->lv_block_exception + e,
 				      new_lv->lv_block_exception[e].rdev_org,
-				    new_lv->lv_block_exception[e].rsector_org,
+				      new_lv->lv_block_exception[e].rsector_org,
 				      new_lv);
 
 	} else {
@@ -2278,7 +2433,7 @@
 			lv_t *snap;
 			for(snap = old_lv->lv_snapshot_next; snap;
 			    snap = snap->lv_snapshot_next) {
-				down(&snap->lv_snapshot_sem);
+				down_write(&snap->lv_lock);
 				snap->lv_current_pe = old_lv->lv_current_pe;
 				snap->lv_allocated_le =
 					old_lv->lv_allocated_le;
@@ -2290,13 +2445,13 @@
 				lvm_size[MINOR(snap->lv_dev)] =
 					old_lv->lv_size >> 1;
 				__update_hardsectsize(snap);
-				up(&snap->lv_snapshot_sem);
+				up_write(&snap->lv_lock);
 			}
 		}
 	}
 
 	__update_hardsectsize(old_lv);
-	up(&old_lv->lv_snapshot_sem);
+	up_write(&old_lv->lv_lock);
 
 	return 0;
 } /* lvm_do_lv_extend_reduce() */
@@ -2325,8 +2480,10 @@
 		    strcmp(lv_ptr->lv_name,
 			   lv_status_byname_req.lv_name) == 0) {
 		        /* Save usermode pointers */
-		        saved_ptr1 = lv_status_byname_req.lv->lv_current_pe;
-			saved_ptr2 = lv_status_byname_req.lv->lv_block_exception;
+		        if (copy_from_user(&saved_ptr1, &lv_status_byname_req.lv->lv_current_pe, sizeof(void*)) != 0)
+				return -EFAULT;
+			if (copy_from_user(&saved_ptr2, &lv_status_byname_req.lv->lv_block_exception, sizeof(void*)) != 0)
+			        return -EFAULT;
 		        if (copy_to_user(lv_status_byname_req.lv,
 					 lv_ptr,
 					 sizeof(lv_t)) != 0)
@@ -2339,7 +2496,8 @@
 					return -EFAULT;
 			}
 			/* Restore usermode pointers */
-			lv_status_byname_req.lv->lv_current_pe = saved_ptr1;
+			if (copy_to_user(&lv_status_byname_req.lv->lv_current_pe, &saved_ptr1, sizeof(void*)) != 0)
+			        return -EFAULT;
 			return 0;
 		}
 	}
@@ -2368,8 +2526,11 @@
 		return -ENXIO;
 
 	/* Save usermode pointers */
-	saved_ptr1 = lv_status_byindex_req.lv->lv_current_pe;
-	saved_ptr2 = lv_status_byindex_req.lv->lv_block_exception;
+	if (copy_from_user(&saved_ptr1, &lv_status_byindex_req.lv->lv_current_pe, sizeof(void*)) != 0)
+	        return -EFAULT;
+	if (copy_from_user(&saved_ptr2, &lv_status_byindex_req.lv->lv_block_exception, sizeof(void*)) != 0)
+	        return -EFAULT;
+
 	if (copy_to_user(lv_status_byindex_req.lv, lv_ptr, sizeof(lv_t)) != 0)
 		return -EFAULT;
 	if (saved_ptr1 != NULL) {
@@ -2381,7 +2542,8 @@
 	}
 
 	/* Restore usermode pointers */
-	lv_status_byindex_req.lv->lv_current_pe = saved_ptr1;
+	if (copy_to_user(&lv_status_byindex_req.lv->lv_current_pe, &saved_ptr1, sizeof(void *)) != 0)
+	        return -EFAULT;
 
 	return 0;
 } /* lvm_do_lv_status_byindex() */
@@ -2411,8 +2573,10 @@
 	lv_ptr = vg_ptr->lv[l];
 
 	/* Save usermode pointers */
-	saved_ptr1 = lv_status_bydev_req.lv->lv_current_pe;
-	saved_ptr2 = lv_status_bydev_req.lv->lv_block_exception;
+	if (copy_from_user(&saved_ptr1, &lv_status_bydev_req.lv->lv_current_pe, sizeof(void*)) != 0)
+	        return -EFAULT;
+	if (copy_from_user(&saved_ptr2, &lv_status_bydev_req.lv->lv_block_exception, sizeof(void*)) != 0)
+	        return -EFAULT;
 
 	if (copy_to_user(lv_status_bydev_req.lv, lv_ptr, sizeof(lv_t)) != 0)
 		return -EFAULT;
@@ -2424,7 +2588,8 @@
 			return -EFAULT;
 	}
 	/* Restore usermode pointers */
-	lv_status_bydev_req.lv->lv_current_pe = saved_ptr1;
+	if (copy_to_user(&lv_status_bydev_req.lv->lv_current_pe, &saved_ptr1, sizeof(void *)) != 0)
+	        return -EFAULT;
 
 	return 0;
 } /* lvm_do_lv_status_bydev() */
@@ -2445,9 +2610,7 @@
 		if (lv_ptr->lv_dev == lv->lv_dev)
 		{
 			lvm_fs_remove_lv(vg_ptr, lv_ptr);
-			strncpy(lv_ptr->lv_name,
-				lv_req->lv_name,
-				NAME_LEN);
+			strncpy(lv_ptr->lv_name, lv_req->lv_name, NAME_LEN);
 			lvm_fs_create_lv(vg_ptr, lv_ptr);
 			break;
 		}
@@ -2465,9 +2628,7 @@
 {
 	uint p;
 	pv_t *pv_ptr;
-#ifdef LVM_GET_INODE
-	struct inode *inode_sav;
-#endif
+	struct block_device *bd;
 
 	if (vg_ptr == NULL) return -ENXIO;
 	if (copy_from_user(&pv_change_req, arg,
@@ -2479,20 +2640,17 @@
 		if (pv_ptr != NULL &&
 		    strcmp(pv_ptr->pv_name,
 			       pv_change_req.pv_name) == 0) {
-#ifdef LVM_GET_INODE
-			inode_sav = pv_ptr->inode;
-#endif
+
+			bd = pv_ptr->bd;
 			if (copy_from_user(pv_ptr,
 					   pv_change_req.pv,
 					   sizeof(pv_t)) != 0)
 				return -EFAULT;
+			pv_ptr->bd = bd;
 
 			/* We don't need the PE list
 			   in kernel space as with LVs pe_t list */
 			pv_ptr->pe = NULL;
-#ifdef LVM_GET_INODE
-			pv_ptr->inode = inode_sav;
-#endif
 			return 0;
 		}
 	}
@@ -2535,8 +2693,7 @@
 {
 	pv_flush_req_t pv_flush_req;
 
-	if (copy_from_user(&pv_flush_req, arg,
-			   sizeof(pv_flush_req)) != 0)
+	if (copy_from_user(&pv_flush_req, arg, sizeof(pv_flush_req)) != 0)
 		return -EFAULT;
 
 	fsync_dev(pv_flush_req.pv_dev);
@@ -2571,5 +2728,82 @@
 } /* lvm_gen_init() */
 
 
+
+/* Must have down_write(_pe_lock) when we enqueue buffers */
+static void _queue_io(struct buffer_head *bh, int rw) {
+	if (bh->b_reqnext) BUG();
+	bh->b_reqnext = _pe_requests;
+	_pe_requests = bh;
+}
+
+/* Must have down_write(_pe_lock) when we dequeue buffers */
+static struct buffer_head *_dequeue_io(void)
+{
+	struct buffer_head *bh = _pe_requests;
+	_pe_requests = NULL;
+	return bh;
+}
+
+/*
+ * We do not need to hold _pe_lock to flush buffers.  bh should be taken from
+ * _pe_requests under down_write(_pe_lock), and then _pe_requests can be set
+ * NULL and we drop _pe_lock.  Any new buffers defered at this time will be
+ * added to a new list, and the old buffers can have their I/O restarted
+ * asynchronously.
+ *
+ * If, for some reason, the same PE is locked again before all of these writes
+ * have finished, then these buffers will just be re-queued (i.e. no danger).
+ */
+static void _flush_io(struct buffer_head *bh)
+{
+	while (bh) {
+		struct buffer_head *next = bh->b_reqnext;
+		bh->b_reqnext = NULL;
+		/* resubmit this buffer head */
+		generic_make_request(WRITE, bh);
+		bh = next;
+	}
+}
+
+
+/*
+ * we must open the pv's before we use them
+ */
+static int _open_pv(pv_t *pv) {
+	int err;
+	struct block_device *bd;
+
+	if (!(bd = bdget(kdev_t_to_nr(pv->pv_dev))))
+		return -ENOMEM;
+
+	err = blkdev_get(bd, FMODE_READ|FMODE_WRITE, 0, BDEV_FILE);
+	if (err) {
+		bdput(bd);
+		return err;
+	}
+
+	pv->bd = bd;
+	return 0;
+}
+
+static void _close_pv(pv_t *pv) {
+	if(!pv || !pv->bd)
+		return;
+
+	blkdev_put(pv->bd, BDEV_FILE);
+	bdput(pv->bd);
+	pv->bd = 0;
+}
+
+
+static unsigned long _sectors_to_k(unsigned long sect)
+{
+	if(SECTOR_SIZE > 1024) {
+		return sect * (SECTOR_SIZE / 1024);
+	}
+
+	return sect / (1024 / SECTOR_SIZE);
+}
+
 module_init(lvm_init);
 module_exit(lvm_cleanup);
--- linux/drivers/md/lvm-internal.h.orig	Tue Oct 23 13:55:17 2001
+++ linux/drivers/md/lvm-internal.h	Tue Oct 23 13:55:51 2001
@@ -1,3 +1,4 @@
+
 /*
  * kernel/lvm_internal.h
  *
@@ -39,17 +40,20 @@
 
 /* global variables, defined in lvm.c */
 extern char *lvm_version;
-extern char *lvm_short_version;
 extern ushort lvm_iop_version;
 extern int loadtime;
 extern const char *const lvm_name;
 
 
+extern uint vg_count;
 extern vg_t *vg[];
 extern struct file_operations lvm_chr_fops;
 
 extern struct block_device_operations lvm_blk_dops;
 
+#define lvm_sectsize(dev) get_hardsect_size(dev)
+
+/* 2.4.8 had no global min/max macros, and 2.4.9's were flawed */
 
 /* debug macros */
 #ifdef DEBUG_IOCTL
@@ -96,7 +100,7 @@
 
 void lvm_fs_create_vg(vg_t *vg_ptr);
 void lvm_fs_remove_vg(vg_t *vg_ptr);
-void lvm_fs_create_lv(vg_t *vg_ptr, lv_t *lv);
+devfs_handle_t lvm_fs_create_lv(vg_t *vg_ptr, lv_t *lv);
 void lvm_fs_remove_lv(vg_t *vg_ptr, lv_t *lv);
 void lvm_fs_create_pv(vg_t *vg_ptr, pv_t *pv);
 void lvm_fs_remove_pv(vg_t *vg_ptr, pv_t *pv);
--- linux/drivers/md/lvm-snap.c.orig	Tue Oct 23 13:55:17 2001
+++ linux/drivers/md/lvm-snap.c	Tue Oct 23 13:55:51 2001
@@ -39,6 +39,8 @@
  *                 o pv number is returned in new uint * arg
  *                 o -1 returned on error
  *                 lvm_snapshot_fill_COW_table has a return value too.
+ *    15/10/2001 - fix snapshot alignment problem [CM]
+ *               - fix snapshot full oops (always check lv_block_exception) [CM]
  *
  */
 
@@ -49,6 +51,7 @@
 #include <linux/types.h>
 #include <linux/iobuf.h>
 #include <linux/lvm.h>
+#include <linux/devfs_fs_kernel.h>
 
 
 #include "lvm-internal.h"
@@ -140,10 +143,20 @@
 	unsigned long mask = lv->lv_snapshot_hash_mask;
 	int chunk_size = lv->lv_chunk_size;
 
+	if (!hash_table)
+		BUG();
 	hash_table = &hash_table[hashfn(org_dev, org_start, mask, chunk_size)];
 	list_add(&exception->hash, hash_table);
 }
 
+/*
+ * Determine if we already have a snapshot chunk for this block.
+ * Return: 1 if it the chunk already exists
+ *         0 if we need to COW this block and allocate a new chunk
+ *        -1 if the snapshot was disabled because it ran out of space
+ *
+ * We need to be holding at least a read lock on lv->lv_lock.
+ */
 int lvm_snapshot_remap_block(kdev_t * org_dev, unsigned long * org_sector,
 			     unsigned long pe_start, lv_t * lv)
 {
@@ -153,6 +166,9 @@
 	int chunk_size = lv->lv_chunk_size;
 	lv_block_exception_t * exception;
 
+	if (!lv->lv_block_exception)
+		return -1;
+
 	pe_off = pe_start % chunk_size;
 	pe_adjustment = (*org_sector-pe_off) % chunk_size;
 	__org_start = *org_sector - pe_adjustment;
@@ -196,19 +212,25 @@
 	       reason);
 }
 
-static inline void lvm_snapshot_prepare_blocks(unsigned long * blocks,
+static inline int lvm_snapshot_prepare_blocks(unsigned long *blocks,
 					       unsigned long start,
 					       int nr_sectors,
 					       int blocksize)
 {
 	int i, sectors_per_block, nr_blocks;
 
-	sectors_per_block = blocksize >> 9;
+	sectors_per_block = blocksize / SECTOR_SIZE;
+
+	if(start & (sectors_per_block - 1))
+		return 0;
+
 	nr_blocks = nr_sectors / sectors_per_block;
 	start /= sectors_per_block;
 
 	for (i = 0; i < nr_blocks; i++)
 		blocks[i] = start++;
+
+	return 1;
 }
 
 inline int lvm_get_blksize(kdev_t dev)
@@ -291,6 +313,8 @@
 
 /*
  * writes a COW exception table sector to disk (HM)
+ *
+ * We need to hold a write lock on lv_snap->lv_lock.
  */
 int lvm_write_COW_table_block(vg_t * vg, lv_t *lv_snap)
 {
@@ -309,6 +333,10 @@
  * if there is no exception storage space free any longer --> release snapshot.
  *
  * this routine gets called for each _first_ write to a physical chunk.
+ *
+ * We need to hold a write lock on lv_snap->lv_lock.  It is assumed that
+ * lv->lv_block_exception is non-NULL (checked by lvm_snapshot_remap_block())
+ * when this function is called.
  */
 int lvm_snapshot_COW(kdev_t org_phys_dev,
 		     unsigned long org_phys_sector,
@@ -318,6 +346,7 @@
 {
 	const char * reason;
 	unsigned long org_start, snap_start, snap_phys_dev, virt_start, pe_off;
+	unsigned long phys_start;
 	int idx = lv_snap->lv_remap_ptr, chunk_size = lv_snap->lv_chunk_size;
 	struct kiobuf * iobuf;
 	unsigned long blocks[KIO_MAX_SECTORS];
@@ -352,8 +381,8 @@
 
 	iobuf = lv_snap->lv_iobuf;
 
-	blksize_org = lvm_get_blksize(org_phys_dev);
-	blksize_snap = lvm_get_blksize(snap_phys_dev);
+	blksize_org = lvm_sectsize(org_phys_dev);
+	blksize_snap = lvm_sectsize(snap_phys_dev);
 	max_blksize = max(blksize_org, blksize_snap);
 	min_blksize = min(blksize_org, blksize_snap);
 	max_sectors = KIO_MAX_SECTORS * (min_blksize>>9);
@@ -361,6 +390,9 @@
 	if (chunk_size % (max_blksize>>9))
 		goto fail_blksize;
 
+	/* Don't change org_start, we need it to fill in the exception table */
+	phys_start = org_start;
+
 	while (chunk_size)
 	{
 		nr_sectors = min(chunk_size, max_sectors);
@@ -368,17 +400,24 @@
 
 		iobuf->length = nr_sectors << 9;
 
-		lvm_snapshot_prepare_blocks(blocks, org_start,
-					    nr_sectors, blksize_org);
+		if (!lvm_snapshot_prepare_blocks(blocks, phys_start,
+						 nr_sectors, blksize_org))
+			goto fail_prepare;
+
 		if (__brw_kiovec(READ, 1, &iobuf, org_phys_dev, blocks,
 				 blksize_org, lv_snap) != (nr_sectors<<9))
 			goto fail_raw_read;
 
-		lvm_snapshot_prepare_blocks(blocks, snap_start,
-					    nr_sectors, blksize_snap);
+		if (!lvm_snapshot_prepare_blocks(blocks, snap_start,
+						 nr_sectors, blksize_snap))
+			goto fail_prepare;
+
 		if (__brw_kiovec(WRITE, 1, &iobuf, snap_phys_dev, blocks,
 				 blksize_snap, lv_snap) != (nr_sectors<<9))
 			goto fail_raw_write;
+
+		phys_start += nr_sectors;
+		snap_start += nr_sectors;
 	}
 
 #ifdef DEBUG_SNAPSHOT
@@ -418,6 +457,11 @@
 fail_blksize:
 	reason = "blocksize error";
 	goto out;
+
+fail_prepare:
+	reason = "couldn't prepare kiovec blocks "
+		"(start probably isn't block aligned)";
+	goto out;
 }
 
 int lvm_snapshot_alloc_iobuf_pages(struct kiobuf * iobuf, int sectors)
@@ -588,7 +632,7 @@
 	snap_phys_dev = lv_snap->lv_block_exception[idx].rdev_new;
 	snap_pe_start = lv_snap->lv_block_exception[idx - (idx % COW_entries_per_pe)].rsector_new - lv_snap->lv_chunk_size;
 
-	blksize_snap = lvm_get_blksize(snap_phys_dev);
+	blksize_snap = lvm_sectsize(snap_phys_dev);
 
         COW_entries_per_block = blksize_snap / sizeof(lv_COW_table_disk_t);
         idx_COW_table = idx % COW_entries_per_pe % COW_entries_per_block;
@@ -637,7 +681,7 @@
 			idx++;
 			snap_phys_dev = lv_snap->lv_block_exception[idx].rdev_new;
 			snap_pe_start = lv_snap->lv_block_exception[idx - (idx % COW_entries_per_pe)].rsector_new - lv_snap->lv_chunk_size;
-			blksize_snap = lvm_get_blksize(snap_phys_dev);
+			blksize_snap = lvm_sectsize(snap_phys_dev);
 			blocks[0] = snap_pe_start >> (blksize_snap >> 10);
 		} else blocks[0]++;
 
--- linux/drivers/md/lvm-fs.c.orig	Tue Oct 23 13:55:17 2001
+++ linux/drivers/md/lvm-fs.c	Tue Oct 23 13:55:51 2001
@@ -3,7 +3,7 @@
  *
  * Copyright (C) 2001 Sistina Software
  *
- * January,February 2001
+ * January-April 2001
  *
  * LVM driver is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -26,6 +26,10 @@
  * Changelog
  *
  *    11/01/2001 - First version (Joe Thornber)
+ *    21/03/2001 - added display of stripes and stripe size (HM)
+ *    04/10/2001 - corrected devfs_register() call in lvm_init_fs()
+ *    11/04/2001 - don't devfs_register("lvm") as user-space always does it
+ *    10/05/2001 - show more of PV name in /proc/lvm/global
  *
  */
 
@@ -38,6 +42,7 @@
 
 #include <linux/devfs_fs_kernel.h>
 #include <linux/proc_fs.h>
+#include <linux/init.h>
 #include <linux/lvm.h>
 
 #include "lvm-internal.h"
@@ -58,7 +63,9 @@
 
 static void _show_uuid(const char *src, char *b, char *e);
 
+#if 0
 static devfs_handle_t lvm_devfs_handle;
+#endif
 static devfs_handle_t vg_devfs_handle[MAX_VG];
 static devfs_handle_t ch_devfs_handle[MAX_VG];
 static devfs_handle_t lv_devfs_handle[MAX_LV];
@@ -69,14 +76,16 @@
 /* inline functions */
 
 /* public interface */
-void lvm_init_fs() {
+void __init lvm_init_fs() {
 	struct proc_dir_entry *pde;
 
+/* User-space has already registered this */
+#if 0
 	lvm_devfs_handle = devfs_register(
-		0 , "lvm", 0, 0, LVM_CHAR_MAJOR,
+		0 , "lvm", 0, LVM_CHAR_MAJOR, 0,
 		S_IFCHR | S_IRUSR | S_IWUSR | S_IRGRP,
 		&lvm_chr_fops, NULL);
-
+#endif
 	lvm_proc_dir = create_proc_entry(LVM_DIR, S_IFDIR, &proc_root);
 	if (lvm_proc_dir) {
 		lvm_proc_vg_subdir = create_proc_entry(LVM_VG_SUBDIR, S_IFDIR,
@@ -87,8 +96,9 @@
 }
 
 void lvm_fin_fs() {
+#if 0
 	devfs_unregister (lvm_devfs_handle);
-
+#endif
 	remove_proc_entry(LVM_GLOBAL, lvm_proc_dir);
 	remove_proc_entry(LVM_VG_SUBDIR, lvm_proc_dir);
 	remove_proc_entry(LVM_DIR, &proc_root);
@@ -137,8 +147,14 @@
 
 	if(vg_ptr->vg_dir_pde) {
 		remove_proc_entry(LVM_LV_SUBDIR, vg_ptr->vg_dir_pde);
+		vg_ptr->lv_subdir_pde = NULL;
+
 		remove_proc_entry(LVM_PV_SUBDIR, vg_ptr->vg_dir_pde);
+		vg_ptr->pv_subdir_pde = NULL;
+
 		remove_proc_entry("group", vg_ptr->vg_dir_pde);
+		vg_ptr->vg_dir_pde = NULL;
+
 		remove_proc_entry(vg_ptr->vg_name, lvm_proc_vg_subdir);
 	}
 }
@@ -150,7 +166,7 @@
 	return name;
 }
 
-void lvm_fs_create_lv(vg_t *vg_ptr, lv_t *lv) {
+devfs_handle_t lvm_fs_create_lv(vg_t *vg_ptr, lv_t *lv) {
 	struct proc_dir_entry *pde;
 	const char *name = _basename(lv->lv_name);
 
@@ -165,6 +181,7 @@
 		pde->read_proc = _proc_read_lv;
 		pde->data = lv;
 	}
+	return lv_devfs_handle[MINOR(lv->lv_dev)];
 }
 
 void lvm_fs_remove_lv(vg_t *vg_ptr, lv_t *lv) {
@@ -256,6 +273,12 @@
 	sz += sprintf(page + sz, "number:       %u\n", lv->lv_number);
 	sz += sprintf(page + sz, "open:         %u\n", lv->lv_open);
 	sz += sprintf(page + sz, "allocation:   %u\n", lv->lv_allocation);
+	if(lv->lv_stripes > 1) {
+		sz += sprintf(page + sz, "stripes:      %u\n",
+			      lv->lv_stripes);
+		sz += sprintf(page + sz, "stripesize:   %u\n",
+			      lv->lv_stripesize);
+	}
 	sz += sprintf(page + sz, "device:       %02u:%02u\n",
 		      MAJOR(lv->lv_dev), MINOR(lv->lv_dev));
 
@@ -304,8 +327,8 @@
 
 #ifdef DEBUG_LVM_PROC_GET_INFO
 	printk(KERN_DEBUG
-	       "%s - lvm_proc_get_global_info CALLED  pos: %lu  count: %d  whence: %d\n",
-	       lvm_name, pos, count, whence);
+	       "%s - lvm_proc_get_global_info CALLED  pos: %lu  count: %d\n",
+	       lvm_name, pos, count);
 #endif
 
 	if(pos != 0 && buf != NULL)
@@ -361,7 +384,7 @@
 #endif
 			      " %s\n\n"
 			      "Total:  %d VG%s  %d PV%s  %d LV%s ",
-			      lvm_short_version,
+			      lvm_version,
 			      vg_counter, vg_counter == 1 ? "" : "s",
 			      pv_counter, pv_counter == 1 ? "" : "s",
 			      lv_counter, lv_counter == 1 ? "" : "s");
@@ -564,7 +587,7 @@
 	allocation_flag = 'A';
 	if (!(pv->pv_allocatable & PV_ALLOCATABLE))
 		allocation_flag = 'N';
-	pv_name = strrchr(pv->pv_name+1,'/');
+	pv_name = strchr(pv->pv_name+1,'/');
 	if ( pv_name == 0) pv_name = pv->pv_name;
 	else               pv_name++;
 	sz = sprintf(buf,
diff -ruN -X /home/joe/packages/dontdiff linux_2.4.1/drivers/md/lvm-snap.h linux/drivers/md/lvm-snap.h
--- linux_2.4.1/drivers/md/lvm-snap.h	Fri Feb 16 14:51:26 2001
+++ linux/drivers/md/lvm-snap.h	Thu Jan  1 01:00:00 1970
@@ -1,47 +0,0 @@
-/*
- * kernel/lvm-snap.h
- *
- * Copyright (C) 2001 Sistina Software
- *
- *
- * LVM driver is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2, or (at your option)
- * any later version.
- *
- * LVM driver is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with GNU CC; see the file COPYING.  If not, write to
- * the Free Software Foundation, 59 Temple Place - Suite 330,
- * Boston, MA 02111-1307, USA.
- *
- */
-
-/*
- * Changelog
- *
- *    05/01/2001:Joe Thornber - Factored this file out of lvm.c
- *
- */
-
-#ifndef LVM_SNAP_H
-#define LVM_SNAP_H
-
-/* external snapshot calls */
-extern inline int lvm_get_blksize(kdev_t);
-extern int lvm_snapshot_alloc(lv_t *);
-extern void lvm_snapshot_fill_COW_page(vg_t *, lv_t *);
-extern int lvm_snapshot_COW(kdev_t, ulong, ulong, ulong, lv_t *);
-extern int lvm_snapshot_remap_block(kdev_t *, ulong *, ulong, lv_t *);
-extern void lvm_snapshot_release(lv_t *); 
-extern int lvm_write_COW_table_block(vg_t *, lv_t *);
-extern inline void lvm_hash_link(lv_block_exception_t *, 
-				 kdev_t, ulong, lv_t *);
-extern int lvm_snapshot_alloc_hash_table(lv_t *);
-extern void lvm_drop_snapshot(lv_t *, const char *);
-
-#endif