// SPDX-License-Identifier: GPL-2.0-only/* * Copyright (C) 2001-2002 Sistina Software (UK) Limited. * Copyright (C) 2006-2008 Red Hat GmbH * * This file is released under the GPL. */ #include "dm-exception-store.h" #include <linux/ctype.h>#include <linux/mm.h>#include <linux/pagemap.h>#include <linux/vmalloc.h>#include <linux/export.h>#include <linux/slab.h>#include <linux/dm-io.h>#include <linux/dm-bufio.h> #define DM_MSG_PREFIX "persistent snapshot"#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32U /* 16KB */ #define DM_PREFETCH_CHUNKS 12 /* *--------------------------------------------------------------- * Persistent snapshots, by persistent we mean that the snapshot * will survive a reboot. *--------------------------------------------------------------- */ /* * We need to store a record of which parts of the origin have * been copied to the snapshot device. The snapshot code * requires that we copy exception chunks to chunk aligned areas * of the COW store. It makes sense therefore, to store the * metadata in chunk size blocks. * * There is no backward or forward compatibility implemented, * snapshots with different disk versions than the kernel will * not be usable. It is expected that "lvcreate" will blank out * the start of a fresh COW device before calling the snapshot * constructor. * * The first chunk of the COW device just contains the header. * After this there is a chunk filled with exception metadata, * followed by as many exception chunks as can fit in the * metadata areas. * * All on disk structures are in little-endian format. The end * of the exceptions info is indicated by an exception with a * new_chunk of 0, which is invalid since it would point to the * header chunk. */ /* * Magic for persistent snapshots: "SnAp" - Feeble isn't it. */#define SNAP_MAGIC 0x70416e53 /* * The on-disk version of the metadata. */#define SNAPSHOT_DISK_VERSION 1 #define NUM_SNAPSHOT_HDR_CHUNKS 1 struct disk_header { __le32 magic; /* * Is this snapshot valid. There is no way of recovering * an invalid snapshot. */ __le32 valid; /* * Simple, incrementing version. no backward * compatibility. */ __le32 version; /* In sectors */ __le32 chunk_size;} __packed; struct disk_exception { __le64 old_chunk; __le64 new_chunk;} __packed; struct core_exception { uint64_t old_chunk; uint64_t new_chunk;}; struct commit_callback { void (*callback)(void *ref, int success); void *context;}; /* * The top level structure for a persistent exception store. */struct pstore { struct dm_exception_store *store; int version; int valid; uint32_t exceptions_per_area; /* * Now that we have an asynchronous kcopyd there is no * need for large chunk sizes, so it wont hurt to have a * whole chunks worth of metadata in memory at once. */ void *area; /* * An area of zeros used to clear the next area. */ void *zero_area; /* * An area used for header. The header can be written * concurrently with metadata (when invalidating the snapshot), * so it needs a separate buffer. */ void *header_area; /* * Used to keep track of which metadata area the data in * 'chunk' refers to. */ chunk_t current_area; /* * The next free chunk for an exception. * * When creating exceptions, all the chunks here and above are * free. It holds the next chunk to be allocated. On rare * occasions (e.g. after a system crash) holes can be left in * the exception store because chunks can be committed out of * order. * * When merging exceptions, it does not necessarily mean all the * chunks here and above are free. It holds the value it would * have held if all chunks had been committed in order of * allocation. Consequently the value may occasionally be * slightly too low, but since it's only used for 'status' and * it can never reach its minimum value too early this doesn't * matter. */ chunk_t next_free; /* * The index of next free exception in the current * metadata area. */ uint32_t current_committed; atomic_t pending_count; uint32_t callback_count; struct commit_callback *callbacks; struct dm_io_client *io_client; struct workqueue_struct *metadata_wq;}; static int alloc_area(struct pstore *ps){ int r = -ENOMEM; size_t len; len = ps->store->chunk_size << SECTOR_SHIFT; /* * Allocate the chunk_size block of memory that will hold * a single metadata area. */ ps->area = vmalloc(len); if (!ps->area) goto err_area; ps->zero_area = vzalloc(len); if (!ps->zero_area) goto err_zero_area; ps->header_area = vmalloc(len); if (!ps->header_area) goto err_header_area; return 0; err_header_area: vfree(ps->zero_area); err_zero_area: vfree(ps->area); err_area: return r;} static void free_area(struct pstore *ps){ vfree(ps->area); ps->area = NULL; vfree(ps->zero_area); ps->zero_area = NULL; vfree(ps->header_area); ps->header_area = NULL;} struct mdata_req { struct dm_io_region *where; struct dm_io_request *io_req; struct work_struct work; int result;}; static void do_metadata(struct work_struct *work){ struct mdata_req *req = container_of(work, struct mdata_req, work); req->result = dm_io(req->io_req, 1, req->where, NULL, NULL, IOPRIO_DEFAULT);} /* * Read or write a chunk aligned and sized block of data from a device. */static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, blk_opf_t opf, int metadata){ struct dm_io_region where = { .bdev = dm_snap_cow(ps->store->snap)->bdev, .sector = ps->store->chunk_size * chunk, .count = ps->store->chunk_size, }; struct dm_io_request io_req = { .bi_opf = opf, .mem.type = DM_IO_VMA, .mem.ptr.vma = area, .client = ps->io_client, .notify.fn = NULL, }; struct mdata_req req; if (!metadata) return dm_io(&io_req, 1, &where, NULL, NULL, IOPRIO_DEFAULT); req.where = &where; req.io_req = &io_req; /* * Issue the synchronous I/O from a different thread * to avoid submit_bio_noacct recursion. */ INIT_WORK_ONSTACK(&req.work, do_metadata); queue_work(ps->metadata_wq, &req.work); flush_workqueue(ps->metadata_wq); destroy_work_on_stack(&req.work); return req.result;} /* * Convert a metadata area index to a chunk index. */static chunk_t area_location(struct pstore *ps, chunk_t area){ return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area);} static void skip_metadata(struct pstore *ps){ uint32_t stride = ps->exceptions_per_area + 1; chunk_t next_free = ps->next_free; if (sector_div(next_free, stride) == NUM_SNAPSHOT_HDR_CHUNKS) ps->next_free++;} /* * Read or write a metadata area. Remembering to skip the first * chunk which holds the header. */static int area_io(struct pstore *ps, blk_opf_t opf){ chunk_t chunk = area_location(ps, ps->current_area); return chunk_io(ps, ps->area, chunk, opf, 0);} static void zero_memory_area(struct pstore *ps){ memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);} static int zero_disk_area(struct pstore *ps, chunk_t area){ return chunk_io(ps, ps->zero_area, area_location(ps, area), REQ_OP_WRITE, 0);} static int read_header(struct pstore *ps, int *new_snapshot){ int r; struct disk_header *dh; unsigned int chunk_size; int chunk_size_supplied = 1; char *chunk_err; /* * Use default chunk size (or logical_block_size, if larger) * if none supplied */ if (!ps->store->chunk_size) { ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS, bdev_logical_block_size(dm_snap_cow(ps->store->snap)-> bdev) >> 9); ps->store->chunk_mask = ps->store->chunk_size - 1; ps->store->chunk_shift = __ffs(ps->store->chunk_size); chunk_size_supplied = 0; } ps->io_client = dm_io_client_create(); if (IS_ERR(ps->io_client)) return PTR_ERR(ps->io_client); r = alloc_area(ps); if (r) return r; r = chunk_io(ps, ps->header_area, 0, REQ_OP_READ, 1); if (r) goto bad; dh = ps->header_area; if (le32_to_cpu(dh->magic) == 0) { *new_snapshot = 1; return 0; } if (le32_to_cpu(dh->magic) != SNAP_MAGIC) { DMWARN("Invalid or corrupt snapshot"); r = -ENXIO; goto bad; } *new_snapshot = 0; ps->valid = le32_to_cpu(dh->valid); ps->version = le32_to_cpu(dh->version); chunk_size = le32_to_cpu(dh->chunk_size); if (ps->store->chunk_size == chunk_size) return 0; if (chunk_size_supplied) DMWARN("chunk size %u in device metadata overrides table chunk size of %u.", chunk_size, ps->store->chunk_size); /* We had a bogus chunk_size. Fix stuff up. */ free_area(ps); r = dm_exception_store_set_chunk_size(ps->store, chunk_size, &chunk_err); if (r) { DMERR("invalid on-disk chunk size %u: %s.", chunk_size, chunk_err); return r; } r = alloc_area(ps); return r; bad: free_area(ps); return r;} static int write_header(struct pstore *ps){ struct disk_header *dh; memset(ps->header_area, 0, ps->store->chunk_size << SECTOR_SHIFT); dh = ps->header_area; dh->magic = cpu_to_le32(SNAP_MAGIC); dh->valid = cpu_to_le32(ps->valid); dh->version = cpu_to_le32(ps->version); dh->chunk_size = cpu_to_le32(ps->store->chunk_size); return chunk_io(ps, ps->header_area, 0, REQ_OP_WRITE, 1);} /* * Access functions for the disk exceptions, these do the endian conversions. */static struct disk_exception *get_exception(struct pstore *ps, void *ps_area, uint32_t index){ BUG_ON(index >= ps->exceptions_per_area); return ((struct disk_exception *) ps_area) + index;} static void read_exception(struct pstore *ps, void *ps_area, uint32_t index, struct core_exception *result){ struct disk_exception *de = get_exception(ps, ps_area, index); /* copy it */ result->old_chunk = le64_to_cpu(de->old_chunk); result->new_chunk = le64_to_cpu(de->new_chunk);} static void write_exception(struct pstore *ps, uint32_t index, struct core_exception *e){ struct disk_exception *de = get_exception(ps, ps->area, index); /* copy it */ de->old_chunk = cpu_to_le64(e->old_chunk); de->new_chunk = cpu_to_le64(e->new_chunk);} static void clear_exception(struct pstore *ps, uint32_t index){ struct disk_exception *de = get_exception(ps, ps->area, index); /* clear it */ de->old_chunk = 0; de->new_chunk = 0;} /* * Registers the exceptions that are present in the current area. * 'full' is filled in to indicate if the area has been * filled. */static int insert_exceptions(struct pstore *ps, void *ps_area, int (*callback)(void *callback_context, chunk_t old, chunk_t new), void *callback_context, int *full){ int r; unsigned int i; struct core_exception e; /* presume the area is full */ *full = 1; for (i = 0; i < ps->exceptions_per_area; i++) { read_exception(ps, ps_area, i, &e); /* * If the new_chunk is pointing at the start of * the COW device, where the first metadata area * is we know that we've hit the end of the * exceptions. Therefore the area is not full. */ if (e.new_chunk == 0LL) { ps->current_committed = i; *full = 0; break; } /* * Keep track of the start of the free chunks. */ if (ps->next_free <= e.new_chunk) ps->next_free = e.new_chunk + 1; /* * Otherwise we add the exception to the snapshot. */ r = callback(callback_context, e.old_chunk, e.new_chunk); if (r) return r; } return 0;} static int read_exceptions(struct pstore *ps, int (*callback)(void *callback_context, chunk_t old, chunk_t new), void *callback_context){ int r, full = 1; struct dm_bufio_client *client; chunk_t prefetch_area = 0; client = dm_bufio_client_create(dm_snap_cow(ps->store->snap)->bdev, ps->store->chunk_size << SECTOR_SHIFT, 1, 0, NULL, NULL, 0); Lines 1–500 of 980. Load more
// SPDX-License-Identifier: GPL-2.0-only /* * Copyright (C) 2001-2002 Sistina Software (UK) Limited. * Copyright (C) 2006-2008 Red Hat GmbH * * This file is released under the GPL. */ #include "dm-exception-store.h" #include <linux/ctype.h> #include <linux/mm.h> #include <linux/pagemap.h> #include <linux/vmalloc.h> #include <linux/export.h> #include <linux/slab.h> #include <linux/dm-io.h> #include <linux/dm-bufio.h> #define DM_MSG_PREFIX "persistent snapshot" #define DM_CHUNK_SIZE_DEFAULT_SECTORS 32U /* 16KB */ #define DM_PREFETCH_CHUNKS 12 /* *--------------------------------------------------------------- * Persistent snapshots, by persistent we mean that the snapshot * will survive a reboot. *--------------------------------------------------------------- */ /* * We need to store a record of which parts of the origin have * been copied to the snapshot device. The snapshot code * requires that we copy exception chunks to chunk aligned areas * of the COW store. It makes sense therefore, to store the * metadata in chunk size blocks. * * There is no backward or forward compatibility implemented, * snapshots with different disk versions than the kernel will * not be usable. It is expected that "lvcreate" will blank out * the start of a fresh COW device before calling the snapshot * constructor. * * The first chunk of the COW device just contains the header. * After this there is a chunk filled with exception metadata, * followed by as many exception chunks as can fit in the * metadata areas. * * All on disk structures are in little-endian format. The end * of the exceptions info is indicated by an exception with a * new_chunk of 0, which is invalid since it would point to the * header chunk. */ /* * Magic for persistent snapshots: "SnAp" - Feeble isn't it. */ #define SNAP_MAGIC 0x70416e53 /* * The on-disk version of the metadata. */ #define SNAPSHOT_DISK_VERSION 1 #define NUM_SNAPSHOT_HDR_CHUNKS 1 struct disk_header { __le32 magic; /* * Is this snapshot valid. There is no way of recovering * an invalid snapshot. */ __le32 valid; /* * Simple, incrementing version. no backward * compatibility. */ __le32 version; /* In sectors */ __le32 chunk_size; } __packed; struct disk_exception { __le64 old_chunk; __le64 new_chunk; } __packed; struct core_exception { uint64_t old_chunk; uint64_t new_chunk; }; struct commit_callback { void (*callback)(void *ref, int success); void *context; }; /* * The top level structure for a persistent exception store. */ struct pstore { struct dm_exception_store *store; int version; int valid; uint32_t exceptions_per_area; /* * Now that we have an asynchronous kcopyd there is no * need for large chunk sizes, so it wont hurt to have a * whole chunks worth of metadata in memory at once. */ void *area; /* * An area of zeros used to clear the next area. */ void *zero_area; /* * An area used for header. The header can be written * concurrently with metadata (when invalidating the snapshot), * so it needs a separate buffer. */ void *header_area; /* * Used to keep track of which metadata area the data in * 'chunk' refers to. */ chunk_t current_area; /* * The next free chunk for an exception. * * When creating exceptions, all the chunks here and above are * free. It holds the next chunk to be allocated. On rare * occasions (e.g. after a system crash) holes can be left in * the exception store because chunks can be committed out of * order. * * When merging exceptions, it does not necessarily mean all the * chunks here and above are free. It holds the value it would * have held if all chunks had been committed in order of * allocation. Consequently the value may occasionally be * slightly too low, but since it's only used for 'status' and * it can never reach its minimum value too early this doesn't * matter. */ chunk_t next_free; /* * The index of next free exception in the current * metadata area. */ uint32_t current_committed; atomic_t pending_count; uint32_t callback_count; struct commit_callback *callbacks; struct dm_io_client *io_client; struct workqueue_struct *metadata_wq; }; static int alloc_area(struct pstore *ps) { int r = -ENOMEM; size_t len; len = ps->store->chunk_size << SECTOR_SHIFT; /* * Allocate the chunk_size block of memory that will hold * a single metadata area. */ ps->area = vmalloc(len); if (!ps->area) goto err_area; ps->zero_area = vzalloc(len); if (!ps->zero_area) goto err_zero_area; ps->header_area = vmalloc(len); if (!ps->header_area) goto err_header_area; return 0; err_header_area: vfree(ps->zero_area); err_zero_area: vfree(ps->area); err_area: return r; } static void free_area(struct pstore *ps) { vfree(ps->area); ps->area = NULL; vfree(ps->zero_area); ps->zero_area = NULL; vfree(ps->header_area); ps->header_area = NULL; } struct mdata_req { struct dm_io_region *where; struct dm_io_request *io_req; struct work_struct work; int result; }; static void do_metadata(struct work_struct *work) { struct mdata_req *req = container_of(work, struct mdata_req, work); req->result = dm_io(req->io_req, 1, req->where, NULL, NULL, IOPRIO_DEFAULT); } /* * Read or write a chunk aligned and sized block of data from a device. */ static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, blk_opf_t opf, int metadata) { struct dm_io_region where = { .bdev = dm_snap_cow(ps->store->snap)->bdev, .sector = ps->store->chunk_size * chunk, .count = ps->store->chunk_size, }; struct dm_io_request io_req = { .bi_opf = opf, .mem.type = DM_IO_VMA, .mem.ptr.vma = area, .client = ps->io_client, .notify.fn = NULL, }; struct mdata_req req; if (!metadata) return dm_io(&io_req, 1, &where, NULL, NULL, IOPRIO_DEFAULT); req.where = &where; req.io_req = &io_req; /* * Issue the synchronous I/O from a different thread * to avoid submit_bio_noacct recursion. */ INIT_WORK_ONSTACK(&req.work, do_metadata); queue_work(ps->metadata_wq, &req.work); flush_workqueue(ps->metadata_wq); destroy_work_on_stack(&req.work); return req.result; } /* * Convert a metadata area index to a chunk index. */ static chunk_t area_location(struct pstore *ps, chunk_t area) { return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area); } static void skip_metadata(struct pstore *ps) { uint32_t stride = ps->exceptions_per_area + 1; chunk_t next_free = ps->next_free; if (sector_div(next_free, stride) == NUM_SNAPSHOT_HDR_CHUNKS) ps->next_free++; } /* * Read or write a metadata area. Remembering to skip the first * chunk which holds the header. */ static int area_io(struct pstore *ps, blk_opf_t opf) { chunk_t chunk = area_location(ps, ps->current_area); return chunk_io(ps, ps->area, chunk, opf, 0); } static void zero_memory_area(struct pstore *ps) { memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT); } static int zero_disk_area(struct pstore *ps, chunk_t area) { return chunk_io(ps, ps->zero_area, area_location(ps, area), REQ_OP_WRITE, 0); } static int read_header(struct pstore *ps, int *new_snapshot) { int r; struct disk_header *dh; unsigned int chunk_size; int chunk_size_supplied = 1; char *chunk_err; /* * Use default chunk size (or logical_block_size, if larger) * if none supplied */ if (!ps->store->chunk_size) { ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS, bdev_logical_block_size(dm_snap_cow(ps->store->snap)-> bdev) >> 9); ps->store->chunk_mask = ps->store->chunk_size - 1; ps->store->chunk_shift = __ffs(ps->store->chunk_size); chunk_size_supplied = 0; } ps->io_client = dm_io_client_create(); if (IS_ERR(ps->io_client)) return PTR_ERR(ps->io_client); r = alloc_area(ps); if (r) return r; r = chunk_io(ps, ps->header_area, 0, REQ_OP_READ, 1); if (r) goto bad; dh = ps->header_area; if (le32_to_cpu(dh->magic) == 0) { *new_snapshot = 1; return 0; } if (le32_to_cpu(dh->magic) != SNAP_MAGIC) { DMWARN("Invalid or corrupt snapshot"); r = -ENXIO; goto bad; } *new_snapshot = 0; ps->valid = le32_to_cpu(dh->valid); ps->version = le32_to_cpu(dh->version); chunk_size = le32_to_cpu(dh->chunk_size); if (ps->store->chunk_size == chunk_size) return 0; if (chunk_size_supplied) DMWARN("chunk size %u in device metadata overrides table chunk size of %u.", chunk_size, ps->store->chunk_size); /* We had a bogus chunk_size. Fix stuff up. */ free_area(ps); r = dm_exception_store_set_chunk_size(ps->store, chunk_size, &chunk_err); if (r) { DMERR("invalid on-disk chunk size %u: %s.", chunk_size, chunk_err); return r; } r = alloc_area(ps); return r; bad: free_area(ps); return r; } static int write_header(struct pstore *ps) { struct disk_header *dh; memset(ps->header_area, 0, ps->store->chunk_size << SECTOR_SHIFT); dh = ps->header_area; dh->magic = cpu_to_le32(SNAP_MAGIC); dh->valid = cpu_to_le32(ps->valid); dh->version = cpu_to_le32(ps->version); dh->chunk_size = cpu_to_le32(ps->store->chunk_size); return chunk_io(ps, ps->header_area, 0, REQ_OP_WRITE, 1); } /* * Access functions for the disk exceptions, these do the endian conversions. */ static struct disk_exception *get_exception(struct pstore *ps, void *ps_area, uint32_t index) { BUG_ON(index >= ps->exceptions_per_area); return ((struct disk_exception *) ps_area) + index; } static void read_exception(struct pstore *ps, void *ps_area, uint32_t index, struct core_exception *result) { struct disk_exception *de = get_exception(ps, ps_area, index); /* copy it */ result->old_chunk = le64_to_cpu(de->old_chunk); result->new_chunk = le64_to_cpu(de->new_chunk); } static void write_exception(struct pstore *ps, uint32_t index, struct core_exception *e) { struct disk_exception *de = get_exception(ps, ps->area, index); /* copy it */ de->old_chunk = cpu_to_le64(e->old_chunk); de->new_chunk = cpu_to_le64(e->new_chunk); } static void clear_exception(struct pstore *ps, uint32_t index) { struct disk_exception *de = get_exception(ps, ps->area, index); /* clear it */ de->old_chunk = 0; de->new_chunk = 0; } /* * Registers the exceptions that are present in the current area. * 'full' is filled in to indicate if the area has been * filled. */ static int insert_exceptions(struct pstore *ps, void *ps_area, int (*callback)(void *callback_context, chunk_t old, chunk_t new), void *callback_context, int *full) { int r; unsigned int i; struct core_exception e; /* presume the area is full */ *full = 1; for (i = 0; i < ps->exceptions_per_area; i++) { read_exception(ps, ps_area, i, &e); /* * If the new_chunk is pointing at the start of * the COW device, where the first metadata area * is we know that we've hit the end of the * exceptions. Therefore the area is not full. */ if (e.new_chunk == 0LL) { ps->current_committed = i; *full = 0; break; } /* * Keep track of the start of the free chunks. */ if (ps->next_free <= e.new_chunk) ps->next_free = e.new_chunk + 1; /* * Otherwise we add the exception to the snapshot. */ r = callback(callback_context, e.old_chunk, e.new_chunk); if (r) return r; } return 0; } static int read_exceptions(struct pstore *ps, int (*callback)(void *callback_context, chunk_t old, chunk_t new), void *callback_context) { int r, full = 1; struct dm_bufio_client *client; chunk_t prefetch_area = 0; client = dm_bufio_client_create(dm_snap_cow(ps->store->snap)->bdev, ps->store->chunk_size << SECTOR_SHIFT, 1, 0, NULL, NULL, 0);