Merge branch '2020-06-23-thin-remap'

This commit is contained in:
Joe Thornber 2020-07-22 16:08:19 +01:00
commit 67622f3c76
14 changed files with 2155 additions and 44 deletions

54
Cargo.lock generated
View File

@ -21,6 +21,11 @@ dependencies = [
"winapi 0.3.8 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "anyhow"
version = "1.0.31"
source = "registry+https://github.com/rust-lang/crates.io-index"
[[package]]
name = "arrayvec"
version = "0.4.12"
@ -100,6 +105,11 @@ dependencies = [
"regex 1.3.9 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "fixedbitset"
version = "0.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
[[package]]
name = "flate2"
version = "1.0.14"
@ -240,6 +250,14 @@ dependencies = [
"unicode-xid 0.2.0 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "quick-xml"
version = "0.18.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
dependencies = [
"memchr 2.3.3 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "quickcheck"
version = "0.9.2"
@ -306,6 +324,11 @@ dependencies = [
"rand_core 0.5.1 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "redox_syscall"
version = "0.1.56"
source = "registry+https://github.com/rust-lang/crates.io-index"
[[package]]
name = "regex"
version = "1.3.9"
@ -322,6 +345,14 @@ name = "regex-syntax"
version = "0.6.18"
source = "registry+https://github.com/rust-lang/crates.io-index"
[[package]]
name = "remove_dir_all"
version = "0.5.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
dependencies = [
"winapi 0.3.8 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "rustc_version"
version = "0.2.3"
@ -368,6 +399,19 @@ dependencies = [
"unicode-xid 0.2.0 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "tempfile"
version = "3.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
dependencies = [
"cfg-if 0.1.9 (registry+https://github.com/rust-lang/crates.io-index)",
"libc 0.2.71 (registry+https://github.com/rust-lang/crates.io-index)",
"rand 0.7.3 (registry+https://github.com/rust-lang/crates.io-index)",
"redox_syscall 0.1.56 (registry+https://github.com/rust-lang/crates.io-index)",
"remove_dir_all 0.5.3 (registry+https://github.com/rust-lang/crates.io-index)",
"winapi 0.3.8 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
name = "textwrap"
version = "0.11.0"
@ -380,9 +424,11 @@ dependencies = [
name = "thinp"
version = "0.1.0"
dependencies = [
"anyhow 1.0.31 (registry+https://github.com/rust-lang/crates.io-index)",
"byteorder 1.3.4 (registry+https://github.com/rust-lang/crates.io-index)",
"clap 2.33.1 (registry+https://github.com/rust-lang/crates.io-index)",
"crc32c 0.4.0 (registry+https://github.com/rust-lang/crates.io-index)",
"fixedbitset 0.3.0 (registry+https://github.com/rust-lang/crates.io-index)",
"flate2 1.0.14 (registry+https://github.com/rust-lang/crates.io-index)",
"libc 0.2.71 (registry+https://github.com/rust-lang/crates.io-index)",
"nix 0.17.0 (registry+https://github.com/rust-lang/crates.io-index)",
@ -390,9 +436,11 @@ dependencies = [
"num-derive 0.3.0 (registry+https://github.com/rust-lang/crates.io-index)",
"num-traits 0.2.11 (registry+https://github.com/rust-lang/crates.io-index)",
"num_cpus 1.13.0 (registry+https://github.com/rust-lang/crates.io-index)",
"quick-xml 0.18.1 (registry+https://github.com/rust-lang/crates.io-index)",
"quickcheck 0.9.2 (registry+https://github.com/rust-lang/crates.io-index)",
"quickcheck_macros 0.9.1 (registry+https://github.com/rust-lang/crates.io-index)",
"rand 0.7.3 (registry+https://github.com/rust-lang/crates.io-index)",
"tempfile 3.1.0 (registry+https://github.com/rust-lang/crates.io-index)",
]
[[package]]
@ -456,6 +504,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
"checksum adler32 1.0.4 (registry+https://github.com/rust-lang/crates.io-index)" = "5d2e7343e7fc9de883d1b0341e0b13970f764c14101234857d2ddafa1cb1cac2"
"checksum aho-corasick 0.7.10 (registry+https://github.com/rust-lang/crates.io-index)" = "8716408b8bc624ed7f65d223ddb9ac2d044c0547b6fa4b0d554f3a9540496ada"
"checksum ansi_term 0.11.0 (registry+https://github.com/rust-lang/crates.io-index)" = "ee49baf6cb617b853aa8d93bf420db2383fab46d314482ca2803b40d5fde979b"
"checksum anyhow 1.0.31 (registry+https://github.com/rust-lang/crates.io-index)" = "85bb70cc08ec97ca5450e6eba421deeea5f172c0fc61f78b5357b2a8e8be195f"
"checksum arrayvec 0.4.12 (registry+https://github.com/rust-lang/crates.io-index)" = "cd9fd44efafa8690358b7408d253adf110036b88f55672a933f01d616ad9b1b9"
"checksum atty 0.2.14 (registry+https://github.com/rust-lang/crates.io-index)" = "d9b39be18770d11421cdb1b9947a45dd3f37e93092cbf377614828a319d5fee8"
"checksum autocfg 1.0.0 (registry+https://github.com/rust-lang/crates.io-index)" = "f8aac770f1885fd7e387acedd76065302551364496e46b3dd00860b2f8359b9d"
@ -467,6 +516,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
"checksum crc32c 0.4.0 (registry+https://github.com/rust-lang/crates.io-index)" = "77ba37ef26c12988c1cee882d522d65e1d5d2ad8c3864665b88ee92767ed84c5"
"checksum crc32fast 1.2.0 (registry+https://github.com/rust-lang/crates.io-index)" = "ba125de2af0df55319f41944744ad91c71113bf74a4646efff39afe1f6842db1"
"checksum env_logger 0.7.1 (registry+https://github.com/rust-lang/crates.io-index)" = "44533bbbb3bb3c1fa17d9f2e4e38bbbaf8396ba82193c4cb1b6445d711445d36"
"checksum fixedbitset 0.3.0 (registry+https://github.com/rust-lang/crates.io-index)" = "2fc4fcacf5cd3681968f6524ea159383132937739c6c40dabab9e37ed515911b"
"checksum flate2 1.0.14 (registry+https://github.com/rust-lang/crates.io-index)" = "2cfff41391129e0a856d6d822600b8d71179d46879e310417eb9c762eb178b42"
"checksum getrandom 0.1.14 (registry+https://github.com/rust-lang/crates.io-index)" = "7abc8dd8451921606d809ba32e95b6111925cd2906060d2dcc29c070220503eb"
"checksum hermit-abi 0.1.13 (registry+https://github.com/rust-lang/crates.io-index)" = "91780f809e750b0a89f5544be56617ff6b1227ee485bcb06ebe10cdf89bd3b71"
@ -484,6 +534,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
"checksum num_cpus 1.13.0 (registry+https://github.com/rust-lang/crates.io-index)" = "05499f3756671c15885fee9034446956fff3f243d6077b91e5767df161f766b3"
"checksum ppv-lite86 0.2.8 (registry+https://github.com/rust-lang/crates.io-index)" = "237a5ed80e274dbc66f86bd59c1e25edc039660be53194b5fe0a482e0f2612ea"
"checksum proc-macro2 1.0.18 (registry+https://github.com/rust-lang/crates.io-index)" = "beae6331a816b1f65d04c45b078fd8e6c93e8071771f41b8163255bbd8d7c8fa"
"checksum quick-xml 0.18.1 (registry+https://github.com/rust-lang/crates.io-index)" = "3cc440ee4802a86e357165021e3e255a9143724da31db1e2ea540214c96a0f82"
"checksum quickcheck 0.9.2 (registry+https://github.com/rust-lang/crates.io-index)" = "a44883e74aa97ad63db83c4bf8ca490f02b2fc02f92575e720c8551e843c945f"
"checksum quickcheck_macros 0.9.1 (registry+https://github.com/rust-lang/crates.io-index)" = "608c156fd8e97febc07dc9c2e2c80bf74cfc6ef26893eae3daf8bc2bc94a4b7f"
"checksum quote 1.0.6 (registry+https://github.com/rust-lang/crates.io-index)" = "54a21852a652ad6f610c9510194f398ff6f8692e334fd1145fed931f7fbe44ea"
@ -491,8 +542,10 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
"checksum rand_chacha 0.2.2 (registry+https://github.com/rust-lang/crates.io-index)" = "f4c8ed856279c9737206bf725bf36935d8666ead7aa69b52be55af369d193402"
"checksum rand_core 0.5.1 (registry+https://github.com/rust-lang/crates.io-index)" = "90bde5296fc891b0cef12a6d03ddccc162ce7b2aff54160af9338f8d40df6d19"
"checksum rand_hc 0.2.0 (registry+https://github.com/rust-lang/crates.io-index)" = "ca3129af7b92a17112d59ad498c6f81eaf463253766b90396d39ea7a39d6613c"
"checksum redox_syscall 0.1.56 (registry+https://github.com/rust-lang/crates.io-index)" = "2439c63f3f6139d1b57529d16bc3b8bb855230c8efcc5d3a896c8bea7c3b1e84"
"checksum regex 1.3.9 (registry+https://github.com/rust-lang/crates.io-index)" = "9c3780fcf44b193bc4d09f36d2a3c87b251da4a046c87795a0d35f4f927ad8e6"
"checksum regex-syntax 0.6.18 (registry+https://github.com/rust-lang/crates.io-index)" = "26412eb97c6b088a6997e05f69403a802a92d520de2f8e63c2b65f9e0f47c4e8"
"checksum remove_dir_all 0.5.3 (registry+https://github.com/rust-lang/crates.io-index)" = "3acd125665422973a33ac9d3dd2df85edad0f4ae9b00dafb1a05e43a9f5ef8e7"
"checksum rustc_version 0.2.3 (registry+https://github.com/rust-lang/crates.io-index)" = "138e3e0acb6c9fb258b19b67cb8abd63c00679d2851805ea151465464fe9030a"
"checksum ryu 1.0.5 (registry+https://github.com/rust-lang/crates.io-index)" = "71d301d4193d031abdd79ff7e3dd721168a9572ef3fe51a1517aba235bd8f86e"
"checksum semver 0.9.0 (registry+https://github.com/rust-lang/crates.io-index)" = "1d7eb9ef2c18661902cc47e535f9bc51b78acd254da71d375c2f6720d9a40403"
@ -500,6 +553,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
"checksum static_assertions 0.3.4 (registry+https://github.com/rust-lang/crates.io-index)" = "7f3eb36b47e512f8f1c9e3d10c2c1965bc992bd9cdb024fa581e2194501c83d3"
"checksum strsim 0.8.0 (registry+https://github.com/rust-lang/crates.io-index)" = "8ea5119cdb4c55b55d432abb513a0429384878c15dde60cc77b1c99de1a95a6a"
"checksum syn 1.0.30 (registry+https://github.com/rust-lang/crates.io-index)" = "93a56fabc59dce20fe48b6c832cc249c713e7ed88fa28b0ee0a3bfcaae5fe4e2"
"checksum tempfile 3.1.0 (registry+https://github.com/rust-lang/crates.io-index)" = "7a6e24d9338a0a5be79593e2fa15a648add6138caa803e2d5bc782c371732ca9"
"checksum textwrap 0.11.0 (registry+https://github.com/rust-lang/crates.io-index)" = "d326610f408c7a4eb6f51c37c330e496b08506c9457c9d34287ecc38809fb060"
"checksum thread_local 1.0.1 (registry+https://github.com/rust-lang/crates.io-index)" = "d40c6d1b69745a6ec6fb1ca717914848da4b44ae29d9b3080cbee91d72a69b14"
"checksum unicode-width 0.1.7 (registry+https://github.com/rust-lang/crates.io-index)" = "caaa9d531767d1ff2150b9332433f32a24622147e5ebb1f26409d5da67afd479"

View File

@ -6,15 +6,19 @@ edition = "2018"
license = "GPL3"
[dependencies]
anyhow = "1.0"
byteorder = "1.3"
clap = "2.33"
crc32c = "0.4"
flate2 = "1.0"
fixedbitset = "0.3"
libc = "0.2.71"
quick-xml = "0.18"
nix = "0.17"
nom = "5.1"
num_cpus = "1.13"
rand = "0.7"
tempfile = "3.1"
num-traits = "0.2"
num-derive = "0.3"

View File

@ -2,6 +2,7 @@ extern crate clap;
extern crate thinp;
use clap::{App, Arg};
use std::path::Path;
use std::process::exit;
use thinp::file_utils;
@ -23,14 +24,14 @@ fn main() {
.takes_value(true));
let matches = parser.get_matches();
let input_file = matches.value_of("INPUT").unwrap();
let output_file = matches.value_of("OUTPUT").unwrap();
let input_file = Path::new(matches.value_of("INPUT").unwrap());
let output_file = Path::new(matches.value_of("OUTPUT").unwrap());
if !file_utils::file_exists(input_file) {
eprintln!("Couldn't find input file '{}'.", &input_file);
if !file_utils::file_exists(&input_file) {
eprintln!("Couldn't find input file '{}'.", &input_file.display());
exit(1);
}
if let Err(reason) = thinp::pack::toplevel::pack(&input_file, &output_file) {
println!("Application error: {}\n", reason);
exit(1);

View File

@ -2,6 +2,7 @@ extern crate clap;
extern crate thinp;
use clap::{App, Arg};
use std::path::Path;
use std::process;
use thinp::file_utils;
@ -9,30 +10,34 @@ use std::process::exit;
fn main() {
let parser = App::new("thin_metadata_unpack")
.version(thinp::version::TOOLS_VERSION)
.version(thinp::version::TOOLS_VERSION)
.about("Unpack a compressed file of thin metadata.")
.arg(Arg::with_name("INPUT")
.help("Specify thinp metadata binary device/file")
.required(true)
.short("i")
.value_name("DEV")
.takes_value(true))
.arg(Arg::with_name("OUTPUT")
.help("Specify packed output file")
.required(true)
.short("o")
.value_name("FILE")
.takes_value(true));
.arg(
Arg::with_name("INPUT")
.help("Specify thinp metadata binary device/file")
.required(true)
.short("i")
.value_name("DEV")
.takes_value(true),
)
.arg(
Arg::with_name("OUTPUT")
.help("Specify packed output file")
.required(true)
.short("o")
.value_name("FILE")
.takes_value(true),
);
let matches = parser.get_matches();
let input_file = matches.value_of("INPUT").unwrap();
let output_file = matches.value_of("OUTPUT").unwrap();
let input_file = Path::new(matches.value_of("INPUT").unwrap());
let output_file = Path::new(matches.value_of("OUTPUT").unwrap());
if !file_utils::file_exists(input_file) {
eprintln!("Couldn't find input file '{}'.", &input_file);
eprintln!("Couldn't find input file '{}'.", &input_file.display());
exit(1);
}
if let Err(reason) = thinp::pack::toplevel::unpack(&input_file, &output_file) {
println!("Application error: {}", reason);
process::exit(1);

74
src/bin/thin_shrink.rs Normal file
View File

@ -0,0 +1,74 @@
extern crate clap;
extern crate thinp;
use clap::{App, Arg};
use std::path::Path;
use std::process::exit;
use thinp::file_utils;
fn main() {
let parser = App::new("thin_shrink")
.version(thinp::version::TOOLS_VERSION)
.about("Rewrite xml metadata and move data in an inactive pool.")
.arg(
Arg::with_name("INPUT")
.help("Specify thinp metadata xml file")
.required(true)
.long("input")
.value_name("INPUT")
.takes_value(true),
)
.arg(
Arg::with_name("OUTPUT")
.help("Specify output xml file")
.required(true)
.long("output")
.value_name("OUTPUT")
.takes_value(true),
)
.arg(
Arg::with_name("DATA")
.help("Specify pool data device where data will be moved")
.required(true)
.long("data")
.value_name("DATA")
.takes_value(true),
)
.arg(
Arg::with_name("NOCOPY")
.help("Skip the copying of data, useful for benchmarking")
.required(false)
.long("no-copy")
.value_name("NOCOPY")
.takes_value(false),
)
.arg(
Arg::with_name("SIZE")
.help("Specify new size for the pool (in data blocks)")
.required(true)
.long("nr-blocks")
.value_name("SIZE")
.takes_value(true),
);
let matches = parser.get_matches();
// FIXME: check these look like xml
let input_file = Path::new(matches.value_of("INPUT").unwrap());
let output_file = Path::new(matches.value_of("OUTPUT").unwrap());
let size = matches.value_of("SIZE").unwrap().parse::<u64>().unwrap();
let data_file = Path::new(matches.value_of("DATA").unwrap());
let do_copy = !matches.is_present("NOCOPY");
if !file_utils::file_exists(input_file) {
eprintln!("Couldn't find input file '{}'.", input_file.display());
exit(1);
}
if let Err(reason) =
thinp::shrink::toplevel::shrink(&input_file, &output_file, &data_file, size, do_copy)
{
println!("Application error: {}\n", reason);
exit(1);
}
}

View File

@ -1,8 +1,11 @@
use nix::sys::stat;
use nix::sys::stat::{FileStat, SFlag};
use std::fs::{File, OpenOptions};
use std::io;
use std::fs::File;
use std::io::{Seek, Write};
use std::os::unix::io::AsRawFd;
use std::path::Path;
use tempfile::tempfile;
//---------------------------------------
@ -11,15 +14,13 @@ fn check_bits(mode: u32, flag: &SFlag) -> bool {
}
pub fn is_file_or_blk(info: FileStat) -> bool {
check_bits(info.st_mode, &stat::SFlag::S_IFBLK) ||
check_bits(info.st_mode, &stat::SFlag::S_IFREG)
check_bits(info.st_mode, &stat::SFlag::S_IFBLK)
|| check_bits(info.st_mode, &stat::SFlag::S_IFREG)
}
pub fn file_exists(path: &str) -> bool {
pub fn file_exists(path: &Path) -> bool {
match stat::stat(path) {
Ok(info) => {
is_file_or_blk(info)
}
Ok(info) => is_file_or_blk(info),
_ => {
// FIXME: assuming all errors indicate the file doesn't
// exist.
@ -39,19 +40,19 @@ pub fn fail<T>(msg: &str) -> io::Result<T> {
Err(e)
}
fn get_device_size(path: &str) -> io::Result<u64> {
let file = File::open(path)?;
fn get_device_size(path: &Path) -> io::Result<u64> {
let file = File::open(path)?;
let fd = file.as_raw_fd();
let mut cap = 0u64;
unsafe {
match ioctl_blkgetsize64(fd, &mut cap) {
Ok(_) => {Ok(cap)}
_ => {fail("BLKGETSIZE64 ioctl failed")}
}
match ioctl_blkgetsize64(fd, &mut cap) {
Ok(_) => Ok(cap),
_ => fail("BLKGETSIZE64 ioctl failed"),
}
}
}
pub fn file_size(path: &str) -> io::Result<u64> {
pub fn file_size(path: &Path) -> io::Result<u64> {
match stat::stat(path) {
Ok(info) => {
if check_bits(info.st_mode, &SFlag::S_IFREG) {
@ -60,12 +61,40 @@ pub fn file_size(path: &str) -> io::Result<u64> {
get_device_size(path)
} else {
fail("not a regular file or block device")
}
}
}
_ => {
fail("stat failed")
}
}
_ => fail("stat failed"),
}
}
//---------------------------------------
fn set_size<W: Write + Seek>(w: &mut W, nr_bytes: u64) -> io::Result<()> {
let zeroes: Vec<u8> = vec![0; 1];
if nr_bytes > 0 {
w.seek(io::SeekFrom::Start(nr_bytes - 1))?;
w.write_all(&zeroes)?;
}
Ok(())
}
pub fn temp_file_sized(nr_bytes: u64) -> io::Result<std::fs::File> {
let mut file = tempfile()?;
set_size(&mut file, nr_bytes)?;
Ok(file)
}
pub fn create_sized_file(path: &Path, nr_bytes: u64) -> io::Result<std::fs::File> {
let mut file = OpenOptions::new()
.read(false)
.write(true)
.create(true)
.truncate(true)
.open(path)?;
set_size(&mut file, nr_bytes)?;
Ok(file)
}
//---------------------------------------

View File

@ -1,3 +1,4 @@
extern crate anyhow;
extern crate byteorder;
extern crate crc32c;
extern crate flate2;
@ -18,4 +19,6 @@ pub mod block_manager;
pub mod check;
pub mod file_utils;
pub mod pack;
pub mod shrink;
pub mod thin;
pub mod version;

View File

@ -6,6 +6,7 @@ use std::os::unix::fs::OpenOptionsExt;
use std::{
error::Error,
fs::OpenOptions,
path::Path,
io,
io::prelude::*,
io::Cursor,
@ -67,7 +68,7 @@ fn mk_chunk_vecs(nr_blocks: u64, nr_jobs: u64) -> Vec<Vec<(u64, u64)>> {
vs
}
pub fn pack(input_file: &str, output_file: &str) -> Result<(), Box<dyn Error>> {
pub fn pack(input_file: &Path, output_file: &Path) -> Result<(), Box<dyn Error>> {
let nr_blocks = get_nr_blocks(&input_file)?;
let nr_jobs = std::cmp::max(1, std::cmp::min(num_cpus::get() as u64, nr_blocks / 128));
let chunk_vecs = mk_chunk_vecs(nr_blocks, nr_jobs);
@ -192,7 +193,7 @@ where
r.read_u64::<LittleEndian>()
}
fn get_nr_blocks(path: &str) -> io::Result<u64> {
fn get_nr_blocks(path: &Path) -> io::Result<u64> {
let len = file_utils::file_size(path)?;
Ok(len / (BLOCK_SIZE as u64))
}
@ -306,7 +307,7 @@ where
Ok(())
}
pub fn unpack(input_file: &str, output_file: &str) -> Result<(), Box<dyn Error>> {
pub fn unpack(input_file: &Path, output_file: &Path) -> Result<(), Box<dyn Error>> {
let mut input = OpenOptions::new()
.read(true)
.write(false)

61
src/shrink/copier.rs Normal file
View File

@ -0,0 +1,61 @@
use anyhow::Result;
use std::fs::OpenOptions;
use std::path::Path;
use std::io::{Seek, SeekFrom, Write, Read};
//use std::os::unix::fs::OpenOptionsExt;
pub type Sector = u64;
#[derive(Debug)]
pub struct Region {
pub src: Sector,
pub dest: Sector,
pub len: Sector,
}
fn copy_step<W>(file: &mut W, src_byte: u64, dest_byte: u64, len: usize) -> Result<()>
where
W: Write + Seek + Read,
{
let mut buf = vec![0; len];
file.seek(SeekFrom::Start(src_byte))?;
file.read_exact(&mut buf)?;
file.seek(SeekFrom::Start(dest_byte))?;
file.write_all(&buf)?;
Ok(())
}
fn copy_region<W>(file: &mut W, r: &Region) -> Result<()>
where
W: Write + Seek + Read,
{
const MAX_BYTES: Sector = 1024 * 1024 * 64;
let src_bytes = r.src * 512;
let dest_bytes = r.dest * 512;
let len_bytes = r.len * 512;
let mut written = 0;
while written != len_bytes {
let step = u64::min(len_bytes - written, MAX_BYTES);
copy_step(file, src_bytes + written, dest_bytes + written, step as usize)?;
written += step;
}
Ok(())
}
pub fn copy(path: &Path, regions: &[Region]) -> Result<()> {
let mut input = OpenOptions::new()
.read(true)
.write(true)
//.custom_flags(libc::O_DIRECT)
.open(path)?;
for r in regions {
eprintln!("copying {:?}", r);
copy_region(&mut input, r)?;
}
input.flush()?;
Ok(())
}

3
src/shrink/mod.rs Normal file
View File

@ -0,0 +1,3 @@
pub mod toplevel;
mod copier;

528
src/shrink/toplevel.rs Normal file
View File

@ -0,0 +1,528 @@
use anyhow::{anyhow, Result};
use fixedbitset::FixedBitSet;
use std::fs::OpenOptions;
use std::io::Write;
use std::os::unix::fs::OpenOptionsExt;
use std::path::Path;
use crate::shrink::copier::{self, Region};
use crate::thin::xml::{self, Visit};
//---------------------------------------
#[derive(Debug)]
struct Pass1 {
// FIXME: Inefficient, use a range_set of some description
allocated_blocks: FixedBitSet,
nr_blocks: u64,
/// High blocks are beyond the new, reduced end of the pool. These
/// will need to be moved.
nr_high_blocks: u64,
block_size: Option<u64>,
}
impl Pass1 {
fn new(nr_blocks: u64) -> Pass1 {
Pass1 {
allocated_blocks: FixedBitSet::with_capacity(0),
nr_blocks,
nr_high_blocks: 0,
block_size: None,
}
}
}
impl xml::MetadataVisitor for Pass1 {
fn superblock_b(&mut self, sb: &xml::Superblock) -> Result<Visit> {
self.allocated_blocks.grow(sb.nr_data_blocks as usize);
self.block_size = Some(sb.data_block_size as u64);
Ok(Visit::Continue)
}
fn superblock_e(&mut self) -> Result<Visit> {
Ok(Visit::Continue)
}
fn device_b(&mut self, _d: &xml::Device) -> Result<Visit> {
Ok(Visit::Continue)
}
fn device_e(&mut self) -> Result<Visit> {
Ok(Visit::Continue)
}
fn map(&mut self, m: &xml::Map) -> Result<Visit> {
for i in m.data_begin..(m.data_begin + m.len) {
if i > self.nr_blocks {
self.nr_high_blocks += 1;
}
self.allocated_blocks.insert(i as usize);
}
Ok(Visit::Continue)
}
fn eof(&mut self) -> Result<Visit> {
Ok(Visit::Continue)
}
}
//---------------------------------------
// Writes remapped xml
struct Pass2<W: Write> {
writer: xml::XmlWriter<W>,
nr_blocks: u64,
remaps: Vec<(BlockRange, BlockRange)>,
}
impl<W: Write> Pass2<W> {
fn new(w: W, nr_blocks: u64, remaps: Vec<(BlockRange, BlockRange)>) -> Pass2<W> {
Pass2 {
writer: xml::XmlWriter::new(w),
nr_blocks,
remaps,
}
}
}
impl<W: Write> xml::MetadataVisitor for Pass2<W> {
fn superblock_b(&mut self, sb: &xml::Superblock) -> Result<Visit> {
self.writer.superblock_b(sb)
}
fn superblock_e(&mut self) -> Result<Visit> {
self.writer.superblock_e()
}
fn device_b(&mut self, d: &xml::Device) -> Result<Visit> {
self.writer.device_b(d)
}
fn device_e(&mut self) -> Result<Visit> {
self.writer.device_e()
}
fn map(&mut self, m: &xml::Map) -> Result<Visit> {
if m.data_begin + m.len < self.nr_blocks {
// no remapping needed.
self.writer.map(m)?;
} else {
let r = m.data_begin..(m.data_begin + m.len);
let remaps = remap(&r, &self.remaps);
let mut written = 0;
for r in remaps {
self.writer.map(&xml::Map {
thin_begin: m.thin_begin + written,
data_begin: r.start,
time: m.time,
len: range_len(&r),
})?;
written += range_len(&r);
}
}
Ok(Visit::Continue)
}
fn eof(&mut self) -> Result<Visit> {
self.writer.eof()
}
}
//---------------------------------------
type BlockRange = std::ops::Range<u64>;
fn bits_to_ranges(bits: &FixedBitSet) -> Vec<BlockRange> {
let mut ranges = Vec::new();
let mut start = None;
for i in 0..bits.len() {
match (bits[i], start) {
(false, None) => {}
(true, None) => {
start = Some((i as u64, 1));
}
(false, Some((b, len))) => {
ranges.push(b..(b + len));
start = None;
}
(true, Some((b, len))) => {
start = Some((b, len + 1));
}
}
}
if let Some((b, len)) = start {
ranges.push(b..(b + len));
}
ranges
}
// Splits the ranges into those below threshold, and those equal or
// above threshold below threshold, and those equal or above threshold
fn ranges_split(ranges: &[BlockRange], threshold: u64) -> (Vec<BlockRange>, Vec<BlockRange>) {
use std::ops::Range;
let mut below = Vec::new();
let mut above = Vec::new();
for r in ranges {
match r {
Range { start, end } if *end <= threshold => below.push(*start..*end),
Range { start, end } if *start < threshold => {
below.push(*start..threshold);
above.push(threshold..*end);
}
Range { start, end } => above.push(*start..*end),
}
}
(below, above)
}
fn negate_ranges(ranges: &[BlockRange], upper_limit: u64) -> Vec<BlockRange> {
use std::ops::Range;
let mut result = Vec::new();
let mut cursor = 0;
for r in ranges {
match r {
Range { start, end } if cursor < *start => {
result.push(cursor..*start);
cursor = *end;
}
Range { start: _, end } => {
cursor = *end;
}
}
}
if cursor < upper_limit {
result.push(cursor..upper_limit);
}
result
}
fn range_len(r: &BlockRange) -> u64 {
r.end - r.start
}
fn ranges_total(rs: &[BlockRange]) -> u64 {
rs.iter().fold(0, |sum, r| sum + range_len(r))
}
// Assumes there is enough space to remap.
fn build_remaps(ranges: Vec<BlockRange>, free: Vec<BlockRange>) -> Vec<(BlockRange, BlockRange)> {
use std::cmp::Ordering;
let mut remap = Vec::new();
let mut range_iter = ranges.into_iter();
let mut free_iter = free.into_iter();
let mut r_ = range_iter.next();
let mut f_ = free_iter.next();
while let (Some(r), Some(f)) = (r_, f_) {
let rlen = range_len(&r);
let flen = range_len(&f);
match rlen.cmp(&flen) {
Ordering::Less => {
// range fits into the free chunk
remap.push((r, f.start..(f.start + rlen)));
f_ = Some((f.start + rlen)..f.end);
r_ = range_iter.next();
}
Ordering::Equal => {
remap.push((r, f));
r_ = range_iter.next();
f_ = free_iter.next();
}
Ordering::Greater => {
remap.push((r.start..(r.start + flen), f));
r_ = Some((r.start + flen)..r.end);
f_ = free_iter.next();
}
}
}
remap
}
#[test]
fn test_build_remaps() {
struct Test {
ranges: Vec<BlockRange>,
free: Vec<BlockRange>,
result: Vec<(BlockRange, BlockRange)>,
}
let tests = vec![
Test {
ranges: vec![],
free: vec![],
result: vec![],
},
Test {
ranges: vec![],
free: vec![0..100],
result: vec![],
},
Test {
ranges: vec![1000..1002],
free: vec![0..100],
result: vec![(1000..1002, 0..2)],
},
Test {
ranges: vec![1000..1002, 1100..1110],
free: vec![0..100],
result: vec![(1000..1002, 0..2), (1100..1110, 2..12)],
},
Test {
ranges: vec![100..120],
free: vec![0..5, 20..23, 30..50],
result: vec![(100..105, 0..5), (105..108, 20..23), (108..120, 30..42)],
},
];
for t in tests {
assert_eq!(build_remaps(t.ranges, t.free), t.result);
}
}
fn overlaps(r1: &BlockRange, r2: &BlockRange, index: usize) -> Option<usize> {
if r1.start >= r2.end {
return None;
}
if r2.start >= r1.end {
return None;
}
Some(index)
}
// Finds the index of the first entry that overlaps r.
fn find_first(r: &BlockRange, remaps: &[(BlockRange, BlockRange)]) -> Option<usize> {
if remaps.is_empty() {
return None;
}
match remaps.binary_search_by_key(&r.start, |(from, _)| from.start) {
Ok(n) => Some(n),
Err(n) => {
if n == 0 {
let (from, _) = &remaps[n];
overlaps(&r, &from, n)
} else if n == remaps.len() {
let (from, _) = &remaps[n - 1];
overlaps(&r, from, n - 1)
} else {
// Need to check the previous entry
let (from, _) = &remaps[n - 1];
overlaps(&r, &from, n - 1).or_else(|| {
let (from, _) = &remaps[n];
overlaps(&r, &from, n)
})
}
}
}
}
fn is_empty(r: &BlockRange) -> bool {
r.start == r.end
}
// remaps must be in sorted order by from.start.
fn remap(r: &BlockRange, remaps: &[(BlockRange, BlockRange)]) -> Vec<BlockRange> {
let mut remap = Vec::new();
let mut r = r.start..r.end;
if let Some(index) = find_first(&r, &remaps) {
let mut index = index;
loop {
let (from, to) = &remaps[index];
// There may be a prefix that doesn't overlap with 'from'
if r.start < from.start {
let len = u64::min(range_len(&r), from.start - r.start);
remap.push(r.start..(r.start + len));
r = (r.start + len)..r.end;
if is_empty(&r) {
break;
}
}
let to = (to.start + (r.start - from.start))..to.end;
let from = r.start..from.end;
let rlen = range_len(&r);
let flen = range_len(&from);
let len = u64::min(rlen, flen);
remap.push(to.start..(to.start + len));
r = (r.start + len)..r.end;
if is_empty(&r) {
break;
}
if len == flen {
index += 1;
}
if index == remaps.len() {
remap.push(r.start..r.end);
break;
}
}
} else {
remap.push(r.start..r.end);
}
remap
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn remap_test() {
struct Test {
remaps: Vec<(BlockRange, BlockRange)>,
input: BlockRange,
output: Vec<BlockRange>,
}
let tests = [
Test {
remaps: vec![],
input: 0..1,
output: vec![0..1],
},
Test {
remaps: vec![],
input: 100..1000,
output: vec![100..1000],
},
Test {
remaps: vec![(10..20, 110..120)],
input: 0..5,
output: vec![0..5],
},
Test {
remaps: vec![(10..20, 110..120)],
input: 10..20,
output: vec![110..120],
},
Test {
remaps: vec![(10..20, 110..120)],
input: 5..15,
output: vec![5..10, 110..115],
},
Test {
remaps: vec![(10..20, 110..120)],
input: 5..25,
output: vec![5..10, 110..120, 20..25],
},
Test {
remaps: vec![(10..20, 110..120)],
input: 15..25,
output: vec![115..120, 20..25],
},
Test {
remaps: vec![(10..20, 110..120)],
input: 25..35,
output: vec![25..35],
},
Test {
remaps: vec![(10..20, 110..120), (30..40, 230..240)],
input: 0..50,
output: vec![0..10, 110..120, 20..30, 230..240, 40..50],
},
];
for t in &tests {
let rs = remap(&t.input, &t.remaps);
assert_eq!(rs, t.output);
}
}
}
fn build_copy_regions(remaps: &[(BlockRange, BlockRange)], block_size: u64) -> Vec<Region> {
let mut rs = Vec::new();
for (from, to) in remaps {
rs.push(Region {
src: from.start * block_size,
dest: to.start * block_size,
len: range_len(&from) * block_size,
});
}
rs
}
fn process_xml<MV: xml::MetadataVisitor>(input_path: &Path, pass: &mut MV) -> Result<()> {
let input = OpenOptions::new()
.read(true)
.write(false)
.custom_flags(libc::O_EXCL)
.open(input_path)?;
xml::read(input, pass)?;
Ok(())
}
pub fn shrink(
input_path: &Path,
output_path: &Path,
data_path: &Path,
nr_blocks: u64,
do_copy: bool,
) -> Result<()> {
let mut pass1 = Pass1::new(nr_blocks);
eprint!("Reading xml...");
process_xml(input_path, &mut pass1)?;
eprintln!("done");
eprintln!("{} blocks need moving", pass1.nr_high_blocks);
let ranges = bits_to_ranges(&pass1.allocated_blocks);
let (below, above) = ranges_split(&ranges, nr_blocks);
let free = negate_ranges(&below, nr_blocks);
let free_blocks = ranges_total(&free);
eprintln!("{} free blocks.", free_blocks);
if free_blocks < pass1.nr_high_blocks {
return Err(anyhow!("Insufficient space"));
}
let remaps = build_remaps(above, free);
if do_copy {
let regions = build_copy_regions(&remaps, pass1.block_size.unwrap() as u64);
copier::copy(data_path, &regions)?;
} else {
eprintln!("skipping copy");
}
let output = OpenOptions::new()
.read(false)
.write(true)
.create(true)
.open(output_path)?;
let mut pass2 = Pass2::new(output, nr_blocks, remaps);
eprint!("writing new xml...");
process_xml(input_path, &mut pass2)?;
eprintln!("done.");
Ok(())
}
//---------------------------------------

1
src/thin/mod.rs Normal file
View File

@ -0,0 +1 @@
pub mod xml;

413
src/thin/xml.rs Normal file
View File

@ -0,0 +1,413 @@
use anyhow::Result;
use std::{borrow::Cow, fmt::Display, io::prelude::*, io::BufReader, io::Write};
use quick_xml::events::attributes::Attribute;
use quick_xml::events::{BytesEnd, BytesStart, Event};
use quick_xml::{Reader, Writer};
//---------------------------------------
#[derive(Clone)]
pub struct Superblock {
pub uuid: String,
pub time: u64,
pub transaction: u64,
pub flags: Option<u32>,
pub version: Option<u32>,
pub data_block_size: u32,
pub nr_data_blocks: u64,
pub metadata_snap: Option<u64>,
}
#[derive(Clone)]
pub struct Device {
pub dev_id: u32,
pub mapped_blocks: u64,
pub transaction: u64,
pub creation_time: u64,
pub snap_time: u64,
}
#[derive(Clone)]
pub struct Map {
pub thin_begin: u64,
pub data_begin: u64,
pub time: u32,
pub len: u64,
}
#[derive(Clone)]
pub enum Visit {
Continue,
Stop,
}
pub trait MetadataVisitor {
fn superblock_b(&mut self, sb: &Superblock) -> Result<Visit>;
fn superblock_e(&mut self) -> Result<Visit>;
fn device_b(&mut self, d: &Device) -> Result<Visit>;
fn device_e(&mut self) -> Result<Visit>;
fn map(&mut self, m: &Map) -> Result<Visit>;
fn eof(&mut self) -> Result<Visit>;
}
pub struct XmlWriter<W: Write> {
w: Writer<W>,
}
impl<W: Write> XmlWriter<W> {
pub fn new(w: W) -> XmlWriter<W> {
XmlWriter {
w: Writer::new_with_indent(w, 0x20, 2),
}
}
}
fn mk_attr_<'a, T: Display>(n: T) -> Cow<'a, [u8]> {
let str = format!("{}", n);
Cow::Owned(str.into_bytes())
}
fn mk_attr<T: Display>(key: &[u8], value: T) -> Attribute {
Attribute {
key,
value: mk_attr_(value),
}
}
const XML_VERSION: u32 = 2;
impl<W: Write> MetadataVisitor for XmlWriter<W> {
fn superblock_b(&mut self, sb: &Superblock) -> Result<Visit> {
let tag = b"superblock";
let mut elem = BytesStart::owned(tag.to_vec(), tag.len());
elem.push_attribute(mk_attr(b"uuid", sb.uuid.clone()));
elem.push_attribute(mk_attr(b"time", sb.time));
elem.push_attribute(mk_attr(b"transaction", sb.transaction));
if let Some(flags) = sb.flags {
// FIXME: is this really a nr?
elem.push_attribute(mk_attr(b"flags", flags));
}
elem.push_attribute(mk_attr(b"version", XML_VERSION));
elem.push_attribute(mk_attr(b"data_block_size", sb.data_block_size));
elem.push_attribute(mk_attr(b"nr_data_blocks", sb.nr_data_blocks));
if let Some(snap) = sb.metadata_snap {
elem.push_attribute(mk_attr(b"metadata_snap", snap));
}
self.w.write_event(Event::Start(elem))?;
Ok(Visit::Continue)
}
fn superblock_e(&mut self) -> Result<Visit> {
self.w
.write_event(Event::End(BytesEnd::borrowed(b"superblock")))?;
Ok(Visit::Continue)
}
fn device_b(&mut self, d: &Device) -> Result<Visit> {
let tag = b"device";
let mut elem = BytesStart::owned(tag.to_vec(), tag.len());
elem.push_attribute(mk_attr(b"dev_id", d.dev_id));
elem.push_attribute(mk_attr(b"mapped_blocks", d.mapped_blocks));
elem.push_attribute(mk_attr(b"transaction", d.transaction));
elem.push_attribute(mk_attr(b"creation_time", d.creation_time));
elem.push_attribute(mk_attr(b"snap_time", d.snap_time));
self.w.write_event(Event::Start(elem))?;
Ok(Visit::Continue)
}
fn device_e(&mut self) -> Result<Visit> {
self.w
.write_event(Event::End(BytesEnd::borrowed(b"device")))?;
Ok(Visit::Continue)
}
fn map(&mut self, m: &Map) -> Result<Visit> {
match m.len {
1 => {
let tag = b"single_mapping";
let mut elem = BytesStart::owned(tag.to_vec(), tag.len());
elem.push_attribute(mk_attr(b"origin_block", m.thin_begin));
elem.push_attribute(mk_attr(b"data_block", m.data_begin));
elem.push_attribute(mk_attr(b"time", m.time));
self.w.write_event(Event::Empty(elem))?;
}
_ => {
let tag = b"range_mapping";
let mut elem = BytesStart::owned(tag.to_vec(), tag.len());
elem.push_attribute(mk_attr(b"origin_begin", m.thin_begin));
elem.push_attribute(mk_attr(b"data_begin", m.data_begin));
elem.push_attribute(mk_attr(b"length", m.len));
elem.push_attribute(mk_attr(b"time", m.time));
self.w.write_event(Event::Empty(elem))?;
}
}
Ok(Visit::Continue)
}
fn eof(&mut self) -> Result<Visit> {
let w = self.w.inner();
w.flush()?;
Ok(Visit::Continue)
}
}
//---------------------------------------
// FIXME: nasty unwraps
fn string_val(kv: &Attribute) -> String {
let v = kv.unescaped_value().unwrap();
let bytes = v.to_vec();
String::from_utf8(bytes).unwrap()
}
// FIXME: there's got to be a way of doing this without copying the string
fn u64_val(kv: &Attribute) -> Result<u64> {
let n = string_val(kv).parse::<u64>()?;
Ok(n)
}
fn u32_val(kv: &Attribute) -> Result<u32> {
let n = string_val(kv).parse::<u32>()?;
Ok(n)
}
fn bad_attr<T>(_tag: &str, _attr: &[u8]) -> Result<T> {
todo!();
}
fn missing_attr<T>(_tag: &str, _attr: &str) -> Result<T> {
todo!();
}
fn check_attr<T>(tag: &str, name: &str, maybe_v: Option<T>) -> Result<T> {
match maybe_v {
None => missing_attr(tag, name),
Some(v) => Ok(v),
}
}
fn parse_superblock(e: &BytesStart) -> Result<Superblock> {
let mut uuid: Option<String> = None;
let mut time: Option<u64> = None;
let mut transaction: Option<u64> = None;
let mut flags: Option<u32> = None;
let mut version: Option<u32> = None;
let mut data_block_size: Option<u32> = None;
let mut nr_data_blocks: Option<u64> = None;
let mut metadata_snap: Option<u64> = None;
for a in e.attributes() {
let kv = a.unwrap();
match kv.key {
b"uuid" => uuid = Some(string_val(&kv)),
b"time" => time = Some(u64_val(&kv)?),
b"transaction" => transaction = Some(u64_val(&kv)?),
b"flags" => flags = Some(u32_val(&kv)?),
b"version" => version = Some(u32_val(&kv)?),
b"data_block_size" => data_block_size = Some(u32_val(&kv)?),
b"nr_data_blocks" => nr_data_blocks = Some(u64_val(&kv)?),
b"metadata_snap" => metadata_snap = Some(u64_val(&kv)?),
_ => return bad_attr("superblock", kv.key),
}
}
let tag = "superblock";
Ok(Superblock {
uuid: check_attr(tag, "uuid", uuid)?,
time: check_attr(tag, "time", time)?,
transaction: check_attr(tag, "transaction", transaction)?,
flags,
version,
data_block_size: check_attr(tag, "data_block_size", data_block_size)?,
nr_data_blocks: check_attr(tag, "nr_data_blocks", nr_data_blocks)?,
metadata_snap,
})
}
fn parse_device(e: &BytesStart) -> Result<Device> {
let mut dev_id: Option<u32> = None;
let mut mapped_blocks: Option<u64> = None;
let mut transaction: Option<u64> = None;
let mut creation_time: Option<u64> = None;
let mut snap_time: Option<u64> = None;
for a in e.attributes() {
let kv = a.unwrap();
match kv.key {
b"dev_id" => dev_id = Some(u32_val(&kv)?),
b"mapped_blocks" => mapped_blocks = Some(u64_val(&kv)?),
b"transaction" => transaction = Some(u64_val(&kv)?),
b"creation_time" => creation_time = Some(u64_val(&kv)?),
b"snap_time" => snap_time = Some(u64_val(&kv)?),
_ => return bad_attr("device", kv.key),
}
}
let tag = "device";
Ok(Device {
dev_id: check_attr(tag, "dev_id", dev_id)?,
mapped_blocks: check_attr(tag, "mapped_blocks", mapped_blocks)?,
transaction: check_attr(tag, "transaction", transaction)?,
creation_time: check_attr(tag, "creation_time", creation_time)?,
snap_time: check_attr(tag, "snap_time", snap_time)?,
})
}
fn parse_single_map(e: &BytesStart) -> Result<Map> {
let mut thin_begin: Option<u64> = None;
let mut data_begin: Option<u64> = None;
let mut time: Option<u32> = None;
for a in e.attributes() {
let kv = a.unwrap();
match kv.key {
b"origin_block" => thin_begin = Some(u64_val(&kv)?),
b"data_block" => data_begin = Some(u64_val(&kv)?),
b"time" => time = Some(u32_val(&kv)?),
_ => return bad_attr("single_mapping", kv.key),
}
}
let tag = "single_mapping";
Ok(Map {
thin_begin: check_attr(tag, "origin_block", thin_begin)?,
data_begin: check_attr(tag, "data_block", data_begin)?,
time: check_attr(tag, "time", time)?,
len: 1,
})
}
fn parse_range_map(e: &BytesStart) -> Result<Map> {
let mut thin_begin: Option<u64> = None;
let mut data_begin: Option<u64> = None;
let mut time: Option<u32> = None;
let mut length: Option<u64> = None;
for a in e.attributes() {
let kv = a.unwrap();
match kv.key {
b"origin_begin" => thin_begin = Some(u64_val(&kv)?),
b"data_begin" => data_begin = Some(u64_val(&kv)?),
b"time" => time = Some(u32_val(&kv)?),
b"length" => length = Some(u64_val(&kv)?),
_ => return bad_attr("range_mapping", kv.key),
}
}
let tag = "range_mapping";
Ok(Map {
thin_begin: check_attr(tag, "origin_begin", thin_begin)?,
data_begin: check_attr(tag, "data_begin", data_begin)?,
time: check_attr(tag, "time", time)?,
len: check_attr(tag, "length", length)?,
})
}
fn handle_event<R, M>(reader: &mut Reader<R>, buf: &mut Vec<u8>, visitor: &mut M) -> Result<Visit>
where
R: Read + BufRead,
M: MetadataVisitor,
{
match reader.read_event(buf) {
Ok(Event::Start(ref e)) => match e.name() {
b"superblock" => visitor.superblock_b(&parse_superblock(e)?),
b"device" => visitor.device_b(&parse_device(e)?),
_ => todo!(),
},
Ok(Event::End(ref e)) => match e.name() {
b"superblock" => visitor.superblock_e(),
b"device" => visitor.device_e(),
_ => todo!(),
},
Ok(Event::Empty(ref e)) => match e.name() {
b"single_mapping" => visitor.map(&parse_single_map(e)?),
b"range_mapping" => visitor.map(&parse_range_map(e)?),
_ => todo!(),
},
Ok(Event::Text(_)) => Ok(Visit::Continue),
Ok(Event::Comment(_)) => Ok(Visit::Continue),
Ok(Event::Eof) => {
visitor.eof()?;
Ok(Visit::Stop)
}
Ok(_) => todo!(),
// FIXME: don't panic!
Err(e) => panic!("error parsing xml {:?}", e),
}
}
pub fn read<R, M>(input: R, visitor: &mut M) -> Result<()>
where
R: Read,
M: MetadataVisitor,
{
let input = BufReader::new(input);
let mut reader = Reader::from_reader(input);
reader.trim_text(true);
let mut buf = Vec::new();
loop {
match handle_event(&mut reader, &mut buf, visitor)? {
Visit::Continue => {}
Visit::Stop => break,
}
}
Ok(())
}
//---------------------------------------
struct SBVisitor {
superblock: Option<Superblock>,
}
impl MetadataVisitor for SBVisitor {
fn superblock_b(&mut self, sb: &Superblock) -> Result<Visit> {
self.superblock = Some(sb.clone());
Ok(Visit::Stop)
}
fn superblock_e(&mut self) -> Result<Visit> {
Ok(Visit::Continue)
}
fn device_b(&mut self, _d: &Device) -> Result<Visit> {
Ok(Visit::Continue)
}
fn device_e(&mut self) -> Result<Visit> {
Ok(Visit::Continue)
}
fn map(&mut self, _m: &Map) -> Result<Visit> {
Ok(Visit::Continue)
}
fn eof(&mut self) -> Result<Visit> {
Ok(Visit::Stop)
}
}
pub fn read_superblock<R>(input: R) -> Result<Superblock>
where
R: Read,
{
let mut v = SBVisitor {superblock: None};
read(input, &mut v)?;
Ok(v.superblock.unwrap())
}
//---------------------------------------

934
tests/thin_shrink.rs Normal file
View File

@ -0,0 +1,934 @@
use anyhow::{anyhow, Result};
use byteorder::{LittleEndian, ReadBytesExt, WriteBytesExt};
use rand::prelude::*;
use std::collections::VecDeque;
use std::fs::OpenOptions;
use std::io::{Cursor, Read, Seek, SeekFrom, Write};
use std::ops::Range;
use std::path::{Path, PathBuf};
use tempfile::tempdir;
use thinp::file_utils;
use thinp::thin::xml::{self, Visit};
//------------------------------------
#[derive(Debug)]
struct ThinBlock {
thin_id: u32,
thin_block: u64,
data_block: u64,
block_size: usize,
}
struct ThinReadRef {
pub data: Vec<u8>,
}
struct ThinWriteRef<'a, W: Write + Seek> {
file: &'a mut W,
block_byte: u64,
pub data: Vec<u8>,
}
impl ThinBlock {
fn read_ref<R: Read + Seek>(&self, r: &mut R) -> Result<ThinReadRef> {
let mut rr = ThinReadRef {
data: vec![0; self.block_size * 512],
};
let byte = self.data_block * (self.block_size as u64) * 512;
r.seek(SeekFrom::Start(byte))?;
r.read_exact(&mut rr.data)?;
Ok(rr)
}
fn zero_ref<'a, W: Write + Seek>(&self, w: &'a mut W) -> ThinWriteRef<'a, W> {
ThinWriteRef {
file: w,
block_byte: self.data_block * (self.block_size as u64) * 512,
data: vec![0; self.block_size * 512],
}
}
//fn write_ref<'a, W>(&self, w: &'a mut W) -> Result<ThinWriteRef<'a, W>>
//where
//W: Read + Write + Seek,
//{
//let mut data = vec![0; self.block_size];
//w.seek(SeekFrom::Start(self.data_block * (self.block_size as u64)))?;
//w.read_exact(&mut data[0..])?;
//
//let wr = ThinWriteRef {
//file: w,
//block_byte: self.data_block * (self.block_size as u64),
//data: vec![0; self.block_size],
//};
//
//Ok(wr)
//}
}
impl<'a, W: Write + Seek> Drop for ThinWriteRef<'a, W> {
fn drop(&mut self) {
// FIXME: We shouldn't panic in a drop function, so any IO
// errors will have to make their way back to the user
// another way (eg, via a flush() method).
self.file.seek(SeekFrom::Start(self.block_byte)).unwrap();
self.file.write_all(&self.data).unwrap();
}
}
//------------------------------------
trait ThinVisitor {
fn thin_block(&mut self, tb: &ThinBlock) -> Result<()>;
}
struct ThinXmlVisitor<'a, V: ThinVisitor> {
inner: &'a mut V,
block_size: Option<u32>,
thin_id: Option<u32>,
}
impl<'a, V: ThinVisitor> xml::MetadataVisitor for ThinXmlVisitor<'a, V> {
fn superblock_b(&mut self, sb: &xml::Superblock) -> Result<Visit> {
self.block_size = Some(sb.data_block_size);
Ok(Visit::Continue)
}
fn superblock_e(&mut self) -> Result<Visit> {
Ok(Visit::Continue)
}
fn device_b(&mut self, d: &xml::Device) -> Result<Visit> {
self.thin_id = Some(d.dev_id);
Ok(Visit::Continue)
}
fn device_e(&mut self) -> Result<Visit> {
Ok(Visit::Continue)
}
fn map(&mut self, m: &xml::Map) -> Result<Visit> {
for i in 0..m.len {
let block = ThinBlock {
thin_id: self.thin_id.unwrap(),
thin_block: m.thin_begin + i,
data_block: m.data_begin + i,
block_size: self.block_size.unwrap() as usize,
};
self.inner.thin_block(&block)?;
}
Ok(Visit::Continue)
}
fn eof(&mut self) -> Result<Visit> {
Ok(Visit::Stop)
}
}
fn thin_visit<R, M>(input: R, visitor: &mut M) -> Result<()>
where
R: Read,
M: ThinVisitor,
{
let mut xml_visitor = ThinXmlVisitor {
inner: visitor,
block_size: None,
thin_id: None,
};
xml::read(input, &mut xml_visitor)
}
//------------------------------------
// To test thin_shrink we'd like to stamp a known pattern across the
// provisioned areas of the thins in the pool, do the shrink, verify
// the patterns.
// A simple linear congruence generator used to create the data to
// go into the thin blocks.
struct Generator {
x: u64,
a: u64,
c: u64,
}
impl Generator {
fn new() -> Generator {
Generator {
x: 0,
a: 6364136223846793005,
c: 1442695040888963407,
}
}
fn step(&mut self) {
self.x = self.a.wrapping_mul(self.x).wrapping_add(self.c)
}
fn fill_buffer(&mut self, seed: u64, bytes: &mut [u8]) -> Result<()> {
self.x = seed;
assert!(bytes.len() % 8 == 0);
let nr_words = bytes.len() / 8;
let mut out = Cursor::new(bytes);
for _ in 0..nr_words {
out.write_u64::<LittleEndian>(self.x)?;
self.step();
}
Ok(())
}
fn verify_buffer(&mut self, seed: u64, bytes: &[u8]) -> Result<bool> {
self.x = seed;
assert!(bytes.len() % 8 == 0);
let nr_words = bytes.len() / 8;
let mut input = Cursor::new(bytes);
for _ in 0..nr_words {
let w = input.read_u64::<LittleEndian>()?;
if w != self.x {
eprintln!("{} != {}", w, self.x);
return Ok(false);
}
self.step();
}
Ok(true)
}
}
//------------------------------------
struct Stamper<'a, W: Write + Seek> {
data_file: &'a mut W,
seed: u64,
}
impl<'a, W: Write + Seek> Stamper<'a, W> {
fn new(w: &'a mut W, seed: u64) -> Stamper<'a, W> {
Stamper { data_file: w, seed }
}
}
impl<'a, W: Write + Seek> ThinVisitor for Stamper<'a, W> {
fn thin_block(&mut self, b: &ThinBlock) -> Result<()> {
let mut wr = b.zero_ref(self.data_file);
let mut gen = Generator::new();
gen.fill_buffer(self.seed ^ (b.thin_id as u64) ^ b.thin_block, &mut wr.data)?;
Ok(())
}
}
//------------------------------------
struct Verifier<'a, R: Read + Seek> {
data_file: &'a mut R,
seed: u64,
}
impl<'a, R: Read + Seek> Verifier<'a, R> {
fn new(r: &'a mut R, seed: u64) -> Verifier<'a, R> {
Verifier { data_file: r, seed }
}
}
impl<'a, R: Read + Seek> ThinVisitor for Verifier<'a, R> {
fn thin_block(&mut self, b: &ThinBlock) -> Result<()> {
let rr = b.read_ref(self.data_file)?;
let mut gen = Generator::new();
if !gen.verify_buffer(self.seed ^ (b.thin_id as u64) ^ b.thin_block, &rr.data)? {
return Err(anyhow!("data verify failed for {:?}", b));
}
Ok(())
}
}
//------------------------------------
fn mk_path(dir: &Path, file: &str) -> PathBuf {
let mut p = PathBuf::new();
p.push(dir);
p.push(PathBuf::from(file));
p
}
fn generate_xml(path: &Path, g: &mut dyn Scenario) -> Result<()> {
let xml_out = OpenOptions::new()
.read(false)
.write(true)
.create(true)
.truncate(true)
.open(path)?;
let mut w = xml::XmlWriter::new(xml_out);
g.generate_xml(&mut w)
}
fn create_data_file(data_path: &Path, xml_path: &Path) -> Result<()> {
let input = OpenOptions::new().read(true).write(false).open(xml_path)?;
let sb = xml::read_superblock(input)?;
let nr_blocks = sb.nr_data_blocks as u64;
let block_size = sb.data_block_size as u64 * 512;
let _file = file_utils::create_sized_file(data_path, nr_blocks * block_size)?;
Ok(())
}
fn stamp(xml_path: &Path, data_path: &Path, seed: u64) -> Result<()> {
let mut data = OpenOptions::new()
.read(false)
.write(true)
.open(&data_path)?;
let xml = OpenOptions::new().read(true).write(false).open(&xml_path)?;
let mut stamper = Stamper::new(&mut data, seed);
thin_visit(xml, &mut stamper)
}
fn verify(xml_path: &Path, data_path: &Path, seed: u64) -> Result<()> {
let mut data = OpenOptions::new()
.read(true)
.write(false)
.open(&data_path)?;
let xml = OpenOptions::new().read(true).write(false).open(&xml_path)?;
let mut verifier = Verifier::new(&mut data, seed);
thin_visit(xml, &mut verifier)
}
trait Scenario {
fn generate_xml(&mut self, v: &mut dyn xml::MetadataVisitor) -> Result<()>;
fn get_new_nr_blocks(&self) -> u64;
}
fn test_shrink(scenario: &mut dyn Scenario) -> Result<()> {
let dir = tempdir()?;
let xml_before = mk_path(dir.path(), "before.xml");
let xml_after = mk_path(dir.path(), "after.xml");
let data_path = mk_path(dir.path(), "metadata.bin");
generate_xml(&xml_before, scenario)?;
create_data_file(&data_path, &xml_before)?;
let mut rng = rand::thread_rng();
let seed = rng.gen::<u64>();
stamp(&xml_before, &data_path, seed)?;
verify(&xml_before, &data_path, seed)?;
let new_nr_blocks = scenario.get_new_nr_blocks();
thinp::shrink::toplevel::shrink(&xml_before, &xml_after, &data_path, new_nr_blocks, true)?;
verify(&xml_after, &data_path, seed)?;
Ok(())
}
//------------------------------------
fn common_sb(nr_blocks: u64) -> xml::Superblock {
xml::Superblock {
uuid: "".to_string(),
time: 0,
transaction: 0,
flags: None,
version: None,
data_block_size: 32,
nr_data_blocks: nr_blocks,
metadata_snap: None,
}
}
struct EmptyPoolS {}
impl Scenario for EmptyPoolS {
fn generate_xml(&mut self, v: &mut dyn xml::MetadataVisitor) -> Result<()> {
v.superblock_b(&common_sb(1024))?;
v.superblock_e()?;
Ok(())
}
fn get_new_nr_blocks(&self) -> u64 {
512
}
}
#[test]
fn shrink_empty_pool() -> Result<()> {
let mut s = EmptyPoolS {};
test_shrink(&mut s)
}
//------------------------------------
struct SingleThinS {
offset: u64,
len: u64,
old_nr_data_blocks: u64,
new_nr_data_blocks: u64,
}
impl SingleThinS {
fn new(offset: u64, len: u64, old_nr_data_blocks: u64, new_nr_data_blocks: u64) -> Self {
SingleThinS {
offset,
len,
old_nr_data_blocks,
new_nr_data_blocks,
}
}
}
impl Scenario for SingleThinS {
fn generate_xml(&mut self, v: &mut dyn xml::MetadataVisitor) -> Result<()> {
v.superblock_b(&common_sb(self.old_nr_data_blocks))?;
v.device_b(&xml::Device {
dev_id: 0,
mapped_blocks: self.len,
transaction: 0,
creation_time: 0,
snap_time: 0,
})?;
v.map(&xml::Map {
thin_begin: 0,
data_begin: self.offset,
time: 0,
len: self.len,
})?;
v.device_e()?;
v.superblock_e()?;
Ok(())
}
fn get_new_nr_blocks(&self) -> u64 {
self.new_nr_data_blocks
}
}
#[test]
fn shrink_single_no_move_1() -> Result<()> {
let mut s = SingleThinS::new(0, 1024, 2048, 1280);
test_shrink(&mut s)
}
#[test]
fn shrink_single_no_move_2() -> Result<()> {
let mut s = SingleThinS::new(100, 1024, 2048, 1280);
test_shrink(&mut s)
}
#[test]
fn shrink_single_no_move_3() -> Result<()> {
let mut s = SingleThinS::new(1024, 1024, 2048, 2048);
test_shrink(&mut s)
}
#[test]
fn shrink_single_partial_move() -> Result<()> {
let mut s = SingleThinS::new(1024, 1024, 2048, 1280);
test_shrink(&mut s)
}
#[test]
fn shrink_single_total_move() -> Result<()> {
let mut s = SingleThinS::new(2048, 1024, 1024 + 2048, 1280);
test_shrink(&mut s)
}
#[test]
fn shrink_insufficient_space() -> Result<()> {
let mut s = SingleThinS::new(0, 2048, 3000, 1280);
match test_shrink(&mut s) {
Ok(_) => Err(anyhow!("Shrink unexpectedly succeeded")),
Err(_) => Ok(()),
}
}
//------------------------------------
struct FragmentedS {
nr_thins: u32,
thin_size: u64,
old_nr_data_blocks: u64,
new_nr_data_blocks: u64,
}
impl FragmentedS {
fn new(nr_thins: u32, thin_size: u64) -> Self {
let old_size = (nr_thins as u64) * thin_size;
FragmentedS {
nr_thins,
thin_size,
old_nr_data_blocks: (nr_thins as u64) * thin_size,
new_nr_data_blocks: old_size * 3 / 4,
}
}
}
#[derive(Clone)]
struct ThinRun {
thin_id: u32,
thin_begin: u64,
len: u64,
}
#[derive(Clone, Debug, Copy)]
struct MappedRun {
thin_id: u32,
thin_begin: u64,
data_begin: u64,
len: u64,
}
fn mk_runs(thin_id: u32, total_len: u64, run_len: std::ops::Range<u64>) -> Vec<ThinRun> {
let mut runs = Vec::new();
let mut b = 0u64;
while b < total_len {
let len = u64::min(
total_len - b,
thread_rng().gen_range(run_len.start, run_len.end),
);
runs.push(ThinRun {
thin_id: thin_id,
thin_begin: b,
len,
});
b += len;
}
runs
}
impl Scenario for FragmentedS {
fn generate_xml(&mut self, v: &mut dyn xml::MetadataVisitor) -> Result<()> {
// Allocate each thin fully, in runs between 1 and 16.
let mut runs = Vec::new();
for thin in 0..self.nr_thins {
runs.append(&mut mk_runs(thin, self.thin_size, 1..17));
}
// Shuffle
runs.shuffle(&mut rand::thread_rng());
// map across the data
let mut maps = Vec::new();
let mut b = 0;
for r in &runs {
maps.push(MappedRun {
thin_id: r.thin_id,
thin_begin: r.thin_begin,
data_begin: b,
len: r.len,
});
b += r.len;
}
// drop half the mappings, which leaves us free runs
let mut dropped = Vec::new();
for i in 0..maps.len() {
if i % 2 == 0 {
dropped.push(maps[i].clone());
}
}
// Unshuffle. This isn't strictly necc. but makes the xml
// more readable.
use std::cmp::Ordering;
maps.sort_by(|&l, &r| match l.thin_id.cmp(&r.thin_id) {
Ordering::Equal => l.thin_begin.cmp(&r.thin_begin),
o => o,
});
// write the xml
v.superblock_b(&common_sb(self.old_nr_data_blocks))?;
for thin in 0..self.nr_thins {
v.device_b(&xml::Device {
dev_id: thin,
mapped_blocks: self.thin_size,
transaction: 0,
creation_time: 0,
snap_time: 0,
})?;
for m in &dropped {
if m.thin_id != thin {
continue;
}
v.map(&xml::Map {
thin_begin: m.thin_begin,
data_begin: m.data_begin,
time: 0,
len: m.len,
})?;
}
v.device_e()?;
}
v.superblock_e()?;
Ok(())
}
fn get_new_nr_blocks(&self) -> u64 {
self.new_nr_data_blocks
}
}
#[test]
fn shrink_fragmented_thin_1() -> Result<()> {
let mut s = FragmentedS::new(1, 2048);
test_shrink(&mut s)
}
#[test]
fn shrink_fragmented_thin_2() -> Result<()> {
let mut s = FragmentedS::new(2, 2048);
test_shrink(&mut s)
}
#[test]
fn shrink_fragmented_thin_8() -> Result<()> {
let mut s = FragmentedS::new(2, 2048);
test_shrink(&mut s)
}
#[test]
fn shrink_fragmented_thin_64() -> Result<()> {
let mut s = FragmentedS::new(2, 2048);
test_shrink(&mut s)
}
//------------------------------------
struct Allocator {
runs: VecDeque<Range<u64>>,
}
impl Allocator {
fn new_shuffled(total_len: u64, run_len: Range<u64>) -> Allocator {
let mut runs = Vec::new();
let mut b = 0u64;
while b < total_len {
let len = u64::min(
total_len - b,
thread_rng().gen_range(run_len.start, run_len.end),
);
runs.push(b..(b + len));
b += len;
}
runs.shuffle(&mut thread_rng());
let runs: VecDeque<Range<u64>> = runs.iter().map(|r| r.clone()).collect();
Allocator { runs }
}
fn is_empty(&self) -> bool {
self.runs.is_empty()
}
fn alloc(&mut self, len: u64) -> Result<Vec<Range<u64>>> {
let mut len = len;
let mut runs = Vec::new();
while len > 0 {
let r = self.runs.pop_front();
if r.is_none() {
return Err(anyhow!("could not allocate; out of space"));
}
let mut r = r.unwrap();
let rlen = r.end - r.start;
if len < rlen {
runs.push(r.start..(r.start + len));
// We need to push something back.
self.runs.push_front((r.start + len)..r.end);
len = 0;
} else {
runs.push(r.start..r.end);
len -= rlen;
}
}
Ok(runs)
}
}
// Having explicitly unmapped regions makes it easier to
// apply snapshots.
#[derive(Clone)]
enum Run {
Mapped { data_begin: u64, len: u64 },
UnMapped { len: u64 },
}
impl Run {
fn len(&self) -> u64 {
match self {
Run::Mapped {
data_begin: _data_begin,
len,
} => *len,
Run::UnMapped { len } => *len,
}
}
fn split(&self, n: u64) -> (Option<Run>, Option<Run>) {
if n == 0 {
return (None, Some(self.clone()));
} else {
if self.len() <= n {
return (Some(self.clone()), None);
} else {
match self {
Run::Mapped { data_begin, len } => (
Some(Run::Mapped {
data_begin: *data_begin,
len: n,
}),
Some(Run::Mapped {
data_begin: data_begin + n,
len: len - n,
}),
),
Run::UnMapped { len } => (
Some(Run::UnMapped { len: n }),
Some(Run::UnMapped { len: len - n }),
),
}
}
}
}
}
#[derive(Clone)]
struct ThinDev {
thin_id: u32,
dev_size: u64,
runs: Vec<Run>,
}
impl ThinDev {
fn emit(&self, v: &mut dyn xml::MetadataVisitor) -> Result<()> {
v.device_b(&xml::Device {
dev_id: self.thin_id,
mapped_blocks: self.dev_size,
transaction: 0,
creation_time: 0,
snap_time: 0,
})?;
let mut b = 0;
for r in &self.runs {
match r {
Run::Mapped { data_begin, len } => {
v.map(&xml::Map {
thin_begin: b,
data_begin: *data_begin,
time: 0,
len: *len,
})?;
b += len;
}
Run::UnMapped { len } => {
b += len;
}
}
}
v.device_e()?;
Ok(())
}
}
#[derive(Clone)]
enum SnapRunType {
Same,
Diff,
Hole,
}
#[derive(Clone)]
struct SnapRun(SnapRunType, u64);
fn mk_origin(thin_id: u32, total_len: u64, allocator: &mut Allocator) -> Result<ThinDev> {
let mut runs = Vec::new();
let mut b = 0;
while b < total_len {
let len = u64::min(thread_rng().gen_range(16, 64), total_len - b);
match thread_rng().gen_range(0, 2) {
0 => {
for data in allocator.alloc(len)? {
assert!(data.end >= data.start);
runs.push(Run::Mapped {
data_begin: data.start,
len: data.end - data.start,
});
}
}
1 => {
runs.push(Run::UnMapped { len });
}
_ => {
return Err(anyhow!("bad value returned from rng"));
}
};
b += len;
}
Ok(ThinDev {
thin_id,
dev_size: total_len,
runs,
})
}
fn mk_snap_mapping(
total_len: u64,
run_len: Range<u64>,
same_percent: usize,
diff_percent: usize,
) -> Vec<SnapRun> {
let mut runs = Vec::new();
let mut b = 0u64;
while b < total_len {
let len = u64::min(
total_len - b,
thread_rng().gen_range(run_len.start, run_len.end),
);
let n = thread_rng().gen_range(0, 100);
if n < same_percent {
runs.push(SnapRun(SnapRunType::Same, len));
} else if n < diff_percent {
runs.push(SnapRun(SnapRunType::Diff, len));
} else {
runs.push(SnapRun(SnapRunType::Hole, len));
}
b += len;
}
runs
}
fn split_runs(mut n: u64, runs: &Vec<Run>) -> (Vec<Run>, Vec<Run>) {
let mut before = Vec::new();
let mut after = Vec::new();
for r in runs {
match r.split(n) {
(Some(lhs), None) => {
before.push(lhs);
}
(Some(lhs), Some(rhs)) => {
before.push(lhs);
after.push(rhs);
}
(None, Some(rhs)) => {
after.push(rhs);
}
(None, None) => {}
}
n -= r.len();
}
(before, after)
}
fn apply_snap_runs(
origin: &Vec<Run>,
snap: &Vec<SnapRun>,
allocator: &mut Allocator,
) -> Result<Vec<Run>> {
let mut origin = origin.clone();
let mut runs = Vec::new();
for SnapRun(st, slen) in snap {
let (os, rest) = split_runs(*slen, &origin);
match st {
SnapRunType::Same => {
for o in os {
runs.push(o);
}
}
SnapRunType::Diff => {
for data in allocator.alloc(*slen)? {
runs.push(Run::Mapped {
data_begin: data.start,
len: data.end - data.start,
});
}
}
SnapRunType::Hole => {
runs.push(Run::UnMapped { len: *slen });
}
}
origin = rest;
}
Ok(runs)
}
// Snapshots share mappings, not neccessarily the entire ranges.
struct SnapS {
len: u64,
nr_snaps: u32,
// Snaps will differ from the origin by this percentage
percent_change: usize,
old_nr_data_blocks: u64,
new_nr_data_blocks: u64,
}
impl SnapS {
fn new(len: u64, nr_snaps: u32, percent_change: usize) -> Self {
let delta = len * (nr_snaps as u64) * (percent_change as u64) / 100;
let old_nr_data_blocks = len + 3 * delta;
let new_nr_data_blocks = len + 2 * delta;
SnapS {
len,
nr_snaps,
percent_change,
old_nr_data_blocks,
new_nr_data_blocks,
}
}
}
impl Scenario for SnapS {
fn generate_xml(&mut self, v: &mut dyn xml::MetadataVisitor) -> Result<()> {
let mut allocator = Allocator::new_shuffled(self.old_nr_data_blocks, 64..512);
let origin = mk_origin(0, self.len, &mut allocator)?;
v.superblock_b(&common_sb(self.old_nr_data_blocks))?;
origin.emit(v)?;
v.superblock_e()?;
Ok(())
}
fn get_new_nr_blocks(&self) -> u64 {
self.new_nr_data_blocks
}
}
#[test]
fn shrink_identical_snap() -> Result<()> {
let mut s = SnapS::new(1024, 1, 0);
test_shrink(&mut s)
}
//------------------------------------