Remove include of blockstore_impl.h from disk_tool and osd

src/blockstore/blockstore.cpp

@@ -58,6 +58,17 @@ void blockstore_t::dump_diagnostics()
     return impl->dump_diagnostics();
 }
 
+std::string blockstore_t::get_op_diag(blockstore_op_t *op)
+{
+    char buf[256];
+    auto priv = PRIV(op);
+    if (priv->wait_for)
+        snprintf(buf, sizeof(buf), "state=%d wait=%d (detail=%ju)", priv->op_state, priv->wait_for, priv->wait_detail);
+    else
+        snprintf(buf, sizeof(buf), "state=%d", priv->op_state);
+    return std::string(buf);
+}
+
 uint32_t blockstore_t::get_block_size()
 {
     return impl->get_block_size();

src/blockstore/blockstore.h

@@ -17,11 +17,6 @@
 #include "ringloop.h"
 #include "timerfd_manager.h"
 
-// Memory alignment for direct I/O (usually 512 bytes)
-#ifndef DIRECT_IO_ALIGNMENT
-#define DIRECT_IO_ALIGNMENT 512
-#endif
-
 // Default block size is 128 KB, current allowed range is 4K - 128M
 #define DEFAULT_DATA_BLOCK_ORDER 17
 #define MIN_DATA_BLOCK_SIZE 4*1024
@@ -30,12 +25,6 @@
 
 #define MIN_JOURNAL_SIZE 1024*1024
 
-// "VITAstor"
-#define BLOCKSTORE_META_MAGIC_V1 0x726F747341544956l
-#define BLOCKSTORE_META_FORMAT_V1 1
-#define BLOCKSTORE_META_FORMAT_V2 2
-#define BLOCKSTORE_META_FORMAT_HEAP 3
-
 #define BS_OP_MIN 1
 #define BS_OP_READ 1
 #define BS_OP_WRITE 2
@@ -226,6 +215,9 @@ public:
     // Print diagnostics to stdout
     void dump_diagnostics();
 
+    // Get diagnostic string for an operation
+    std::string get_op_diag(blockstore_op_t *op);
+
     uint32_t get_block_size();
     uint64_t get_block_count();
     uint64_t get_free_block_count();

src/blockstore/blockstore_disk.h

@@ -8,6 +8,11 @@
 #include <string>
 #include <map>
 
+// Memory alignment for direct I/O (usually 512 bytes)
+#ifndef DIRECT_IO_ALIGNMENT
+#define DIRECT_IO_ALIGNMENT 512
+#endif
+
 #define BLOCKSTORE_CSUM_NONE 0
 // Lower byte of checksum type is its length
 #define BLOCKSTORE_CSUM_CRC32C 0x104

src/blockstore/blockstore_impl.h

@@ -5,6 +5,7 @@
 
 #include "blockstore.h"
 #include "blockstore_disk.h"
+#include "ondisk_formats.h"
 
 #include <sys/types.h>
 #include <sys/ioctl.h>
@@ -89,47 +90,6 @@
 
 #include "blockstore_journal.h"
 
-// "VITAstor"
-#define BLOCKSTORE_META_MAGIC_V1 0x726F747341544956l
-#define BLOCKSTORE_META_FORMAT_V1 1
-#define BLOCKSTORE_META_FORMAT_V2 2
-
-// metadata header (superblock)
-struct __attribute__((__packed__)) blockstore_meta_header_v1_t
-{
-    uint64_t zero;
-    uint64_t magic;
-    uint64_t version;
-    uint32_t meta_block_size;
-    uint32_t data_block_size;
-    uint32_t bitmap_granularity;
-};
-
-struct __attribute__((__packed__)) blockstore_meta_header_v2_t
-{
-    uint64_t zero;
-    uint64_t magic;
-    uint64_t version;
-    uint32_t meta_block_size;
-    uint32_t data_block_size;
-    uint32_t bitmap_granularity;
-    uint32_t data_csum_type;
-    uint32_t csum_block_size;
-    uint32_t header_csum;
-};
-
-// 32 bytes = 24 bytes + block bitmap (4 bytes by default) + external attributes (also bitmap, 4 bytes by default)
-// per "clean" entry on disk with fixed metadata tables
-struct __attribute__((__packed__)) clean_disk_entry
-{
-    object_id oid;
-    uint64_t version;
-    uint8_t bitmap[];
-    // Two more fields come after bitmap in metadata version 2:
-    // uint32_t data_csum[];
-    // uint32_t entry_csum;
-};
-
 // 32 = 16 + 16 bytes per "clean" entry in memory (object_id => clean_entry)
 struct __attribute__((__packed__)) clean_entry
 {

src/blockstore/blockstore_journal.h

@@ -3,147 +3,6 @@
 
 #pragma once
 
-#include "crc32c.h"
-#include <set>
-
-#define JOURNAL_MAGIC 0x4A33
-#define JOURNAL_VERSION_V1 1
-#define JOURNAL_VERSION_V2 2
-#define JOURNAL_BUFFER_SIZE 4*1024*1024
-#define JOURNAL_ENTRY_HEADER_SIZE 16
-
-// Journal entries
-// Journal entries are linked to each other by their crc32 value
-// The journal is almost a blockchain, because object versions constantly increase
-#define JE_MIN         0x01
-#define JE_START       0x01
-#define JE_SMALL_WRITE 0x02
-#define JE_BIG_WRITE   0x03
-#define JE_STABLE      0x04
-#define JE_DELETE      0x05
-#define JE_ROLLBACK    0x06
-#define JE_SMALL_WRITE_INSTANT 0x07
-#define JE_BIG_WRITE_INSTANT   0x08
-#define JE_MAX         0x08
-
-// crc32c comes first to ease calculation
-struct __attribute__((__packed__)) journal_entry_start
-{
-    uint32_t crc32;
-    uint16_t magic;
-    uint16_t type;
-    uint32_t size;
-    uint32_t reserved;
-    uint64_t journal_start;
-    uint64_t version;
-    uint32_t data_csum_type;
-    uint32_t csum_block_size;
-};
-#define JE_START_V0_SIZE 24
-#define JE_START_V1_SIZE 32
-#define JE_START_V2_SIZE 40
-
-struct __attribute__((__packed__)) journal_entry_small_write
-{
-    uint32_t crc32;
-    uint16_t magic;
-    uint16_t type;
-    uint32_t size;
-    uint32_t crc32_prev;
-    object_id oid;
-    uint64_t version;
-    uint32_t offset;
-    uint32_t len;
-    // small_write entries contain <len> bytes of data which is stored in next sectors
-    // data_offset is its offset within journal
-    uint64_t data_offset;
-    uint32_t crc32_data; // zero when data_csum_type != 0
-    // small_write and big_write entries are followed by the "external" bitmap
-    // its size is dynamic and included in journal entry's <size> field
-    uint8_t bitmap[];
-    // and then data checksums if data_csum_type != 0
-    // uint32_t data_crc32c[];
-};
-
-struct __attribute__((__packed__)) journal_entry_big_write
-{
-    uint32_t crc32;
-    uint16_t magic;
-    uint16_t type;
-    uint32_t size;
-    uint32_t crc32_prev;
-    object_id oid;
-    uint64_t version;
-    uint32_t offset;
-    uint32_t len;
-    uint64_t location;
-    // small_write and big_write entries are followed by the "external" bitmap
-    // its size is dynamic and included in journal entry's <size> field
-    uint8_t bitmap[];
-    // and then data checksums if data_csum_type != 0
-    // uint32_t data_crc32c[];
-};
-
-struct __attribute__((__packed__)) journal_entry_stable
-{
-    uint32_t crc32;
-    uint16_t magic;
-    uint16_t type;
-    uint32_t size;
-    uint32_t crc32_prev;
-    object_id oid;
-    uint64_t version;
-};
-
-struct __attribute__((__packed__)) journal_entry_rollback
-{
-    uint32_t crc32;
-    uint16_t magic;
-    uint16_t type;
-    uint32_t size;
-    uint32_t crc32_prev;
-    object_id oid;
-    uint64_t version;
-};
-
-struct __attribute__((__packed__)) journal_entry_del
-{
-    uint32_t crc32;
-    uint16_t magic;
-    uint16_t type;
-    uint32_t size;
-    uint32_t crc32_prev;
-    object_id oid;
-    uint64_t version;
-};
-
-struct __attribute__((__packed__)) journal_entry
-{
-    union
-    {
-        struct __attribute__((__packed__))
-        {
-            uint32_t crc32;
-            uint16_t magic;
-            uint16_t type;
-            uint32_t size;
-            uint32_t crc32_prev;
-        };
-        journal_entry_start start;
-        journal_entry_small_write small_write;
-        journal_entry_big_write big_write;
-        journal_entry_stable stable;
-        journal_entry_rollback rollback;
-        journal_entry_del del;
-    };
-};
-
-inline uint32_t je_crc32(journal_entry *je)
-{
-    // 0x48674bc7 = crc32(4 zero bytes)
-    return crc32c(0x48674bc7, ((uint8_t*)je)+4, je->size-4);
-}
-
 struct journal_sector_info_t
 {
     uint64_t offset;

src/blockstore/ondisk_formats.h

@@ -0,0 +1,186 @@
+// Metadata on-disk structures
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 (see README.md for details)
+
+#pragma once
+
+#include "crc32c.h"
+
+#define JOURNAL_MAGIC 0x4A33
+#define JOURNAL_VERSION_V1 1
+#define JOURNAL_VERSION_V2 2
+#define JOURNAL_BUFFER_SIZE 4*1024*1024
+#define JOURNAL_ENTRY_HEADER_SIZE 16
+
+// Journal entries
+// Journal entries are linked to each other by their crc32 value
+// The journal is almost a blockchain, because object versions constantly increase
+#define JE_MIN         0x01
+#define JE_START       0x01
+#define JE_SMALL_WRITE 0x02
+#define JE_BIG_WRITE   0x03
+#define JE_STABLE      0x04
+#define JE_DELETE      0x05
+#define JE_ROLLBACK    0x06
+#define JE_SMALL_WRITE_INSTANT 0x07
+#define JE_BIG_WRITE_INSTANT   0x08
+#define JE_MAX         0x08
+
+// crc32c comes first to ease calculation
+struct __attribute__((__packed__)) journal_entry_start
+{
+    uint32_t crc32;
+    uint16_t magic;
+    uint16_t type;
+    uint32_t size;
+    uint32_t reserved;
+    uint64_t journal_start;
+    uint64_t version;
+    uint32_t data_csum_type;
+    uint32_t csum_block_size;
+};
+#define JE_START_V0_SIZE 24
+#define JE_START_V1_SIZE 32
+#define JE_START_V2_SIZE 40
+
+struct __attribute__((__packed__)) journal_entry_small_write
+{
+    uint32_t crc32;
+    uint16_t magic;
+    uint16_t type;
+    uint32_t size;
+    uint32_t crc32_prev;
+    object_id oid;
+    uint64_t version;
+    uint32_t offset;
+    uint32_t len;
+    // small_write entries contain <len> bytes of data which is stored in next sectors
+    // data_offset is its offset within journal
+    uint64_t data_offset;
+    uint32_t crc32_data; // zero when data_csum_type != 0
+    // small_write and big_write entries are followed by the "external" bitmap
+    // its size is dynamic and included in journal entry's <size> field
+    uint8_t bitmap[];
+    // and then data checksums if data_csum_type != 0
+    // uint32_t data_crc32c[];
+};
+
+struct __attribute__((__packed__)) journal_entry_big_write
+{
+    uint32_t crc32;
+    uint16_t magic;
+    uint16_t type;
+    uint32_t size;
+    uint32_t crc32_prev;
+    object_id oid;
+    uint64_t version;
+    uint32_t offset;
+    uint32_t len;
+    uint64_t location;
+    // small_write and big_write entries are followed by the "external" bitmap
+    // its size is dynamic and included in journal entry's <size> field
+    uint8_t bitmap[];
+    // and then data checksums if data_csum_type != 0
+    // uint32_t data_crc32c[];
+};
+
+struct __attribute__((__packed__)) journal_entry_stable
+{
+    uint32_t crc32;
+    uint16_t magic;
+    uint16_t type;
+    uint32_t size;
+    uint32_t crc32_prev;
+    object_id oid;
+    uint64_t version;
+};
+
+struct __attribute__((__packed__)) journal_entry_rollback
+{
+    uint32_t crc32;
+    uint16_t magic;
+    uint16_t type;
+    uint32_t size;
+    uint32_t crc32_prev;
+    object_id oid;
+    uint64_t version;
+};
+
+struct __attribute__((__packed__)) journal_entry_del
+{
+    uint32_t crc32;
+    uint16_t magic;
+    uint16_t type;
+    uint32_t size;
+    uint32_t crc32_prev;
+    object_id oid;
+    uint64_t version;
+};
+
+struct __attribute__((__packed__)) journal_entry
+{
+    union
+    {
+        struct __attribute__((__packed__))
+        {
+            uint32_t crc32;
+            uint16_t magic;
+            uint16_t type;
+            uint32_t size;
+            uint32_t crc32_prev;
+        };
+        journal_entry_start start;
+        journal_entry_small_write small_write;
+        journal_entry_big_write big_write;
+        journal_entry_stable stable;
+        journal_entry_rollback rollback;
+        journal_entry_del del;
+    };
+};
+
+inline uint32_t je_crc32(journal_entry *je)
+{
+    // 0x48674bc7 = crc32(4 zero bytes)
+    return crc32c(0x48674bc7, ((uint8_t*)je)+4, je->size-4);
+}
+
+// "VITAstor"
+#define BLOCKSTORE_META_MAGIC_V1 0x726F747341544956l
+#define BLOCKSTORE_META_FORMAT_V1 1
+#define BLOCKSTORE_META_FORMAT_V2 2
+
+// metadata header (superblock)
+struct __attribute__((__packed__)) blockstore_meta_header_v1_t
+{
+    uint64_t zero;
+    uint64_t magic;
+    uint64_t version;
+    uint32_t meta_block_size;
+    uint32_t data_block_size;
+    uint32_t bitmap_granularity;
+};
+
+struct __attribute__((__packed__)) blockstore_meta_header_v2_t
+{
+    uint64_t zero;
+    uint64_t magic;
+    uint64_t version;
+    uint32_t meta_block_size;
+    uint32_t data_block_size;
+    uint32_t bitmap_granularity;
+    uint32_t data_csum_type;
+    uint32_t csum_block_size;
+    uint32_t header_csum;
+};
+
+// 32 bytes = 24 bytes + block bitmap (4 bytes by default) + external attributes (also bitmap, 4 bytes by default)
+// per "clean" entry on disk with fixed metadata tables
+struct __attribute__((__packed__)) clean_disk_entry
+{
+    object_id oid;
+    uint64_t version;
+    uint8_t bitmap[];
+    // Two more fields come after bitmap in metadata version 2:
+    // uint32_t data_csum[];
+    // uint32_t entry_csum;
+};

src/disk_tool/disk_tool.h

@@ -8,14 +8,17 @@
 #endif
 
 #include <map>
+#include <set>
 #include <vector>
 #include <string>
 #include <functional>
 
 #include "json11/json11.hpp"
 #include "blockstore_disk.h"
-#include "blockstore_impl.h"
+#include "blockstore.h"
+#include "ondisk_formats.h"
 #include "crc32c.h"
+#include "allocator.h"
 
 // vITADisk
 #define VITASTOR_DISK_MAGIC 0x6b73694441544976

src/disk_tool/disk_tool_journal.cpp

@@ -1,7 +1,10 @@
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 (see README.md for details)
 
+#include <unistd.h>
+
 #include "disk_tool.h"
+#include "malloc_or_die.h"
 
 int disk_tool_t::dump_journal()
 {

src/disk_tool/disk_tool_meta.cpp

@@ -5,6 +5,7 @@
 #include "rw_blocking.h"
 #include "osd_id.h"
 #include "json_util.h"
+#include "malloc_or_die.h"
 
 int disk_tool_t::process_meta(std::function<void(blockstore_meta_header_v2_t *)> hdr_fn,
     std::function<void(uint64_t, clean_disk_entry*, uint8_t*)> record_fn, bool do_open)

src/disk_tool/disk_tool_resize.cpp

@@ -4,6 +4,7 @@
 #include "disk_tool.h"
 #include "rw_blocking.h"
 #include "str_util.h"
+#include "malloc_or_die.h"
 
 #define DM_ST_EMPTY 0
 #define DM_ST_TO_READ 1

src/disk_tool/disk_tool_udev.cpp

@@ -7,6 +7,7 @@
 #include "rw_blocking.h"
 #include "str_util.h"
 #include "json_util.h"
+#include "malloc_or_die.h"
 
 struct __attribute__((__packed__)) vitastor_disk_superblock_t
 {

src/disk_tool/disk_tool_utils.cpp

@@ -2,11 +2,13 @@
 // License: VNPL-1.1 (see README.md for details)
 
 #include <sys/wait.h>
+#include <sys/ioctl.h>
 #include <dirent.h>
 
 #include "disk_tool.h"
 #include "rw_blocking.h"
 #include "str_util.h"
+#include "malloc_or_die.h"
 
 uint64_t sscanf_json(const char *fmt, const json11::Json & str)
 {

src/osd/osd.cpp

@@ -9,7 +9,6 @@
 #include <arpa/inet.h>
 
 #include "addr_util.h"
-#include "blockstore_impl.h"
 #include "osd_primary.h"
 #include "osd.h"
 #include "http_client.h"
@@ -650,11 +649,11 @@ void osd_t::print_slow()
                     op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
                 {
                     cur_slow_op_secondary++;
-                    bufprintf(" state=%d", op->bs_op ? PRIV(op->bs_op)->op_state : -1);
+                    if (op->bs_op)
-                    int wait_for = op->bs_op ? PRIV(op->bs_op)->wait_for : 0;
-                    if (wait_for)
                     {
-                        bufprintf(" wait=%d (detail=%ju)", wait_for, PRIV(op->bs_op)->wait_detail);
+                        auto diag = bs->get_op_diag(op->bs_op);
+                        if (diag != "")
+                            bufprintf(" %s", diag.c_str());
                     }
                 }
                 else if (op->req.hdr.opcode == OSD_OP_READ || op->req.hdr.opcode == OSD_OP_WRITE ||