6 Commits

Author SHA1 Message Date
Caleb J. Gardner 9a710b98cf Moved a lot of things to have explicit alloc arguments
Changed Table to be stateless
2026-02-15 10:31:42 -06:00
Caleb J. Gardner b7b99325da Slight improvement to how permissions are applied to folders
unsquashfs Verbose flag
2026-02-13 01:35:20 -06:00
Caleb J. Gardner b22e4d003d Update times 2026-02-12 08:57:36 -06:00
Caleb J. Gardner a41c37fef4 Updated unsquashfs version print out 2026-02-12 06:54:20 -06:00
Caleb J. Gardner 2d079d77f7 Exclusive file creation 2026-02-12 05:19:25 -06:00
Caleb J. Gardner 48f4235875 Fixed threads == 0 causing single threaded extraction.
Set exclusive file creation
2026-02-12 05:08:16 -06:00
11 changed files with 278 additions and 306 deletions
+9
View File
@@ -58,6 +58,7 @@ pub fn build(b: *std.Build) !void {
b.installArtifact(lib);
b.installArtifact(exe);
const run_step = b.step("run", "Run the app");
const run_cmd = b.addRunArtifact(exe);
run_step.dependOn(&run_cmd.step);
@@ -76,4 +77,12 @@ pub fn build(b: *std.Build) !void {
const test_step = b.step("test", "Run tests");
test_step.dependOn(&run_mod_tests.step);
test_step.dependOn(&run_exe_tests.step);
// Wanted by ZLS for better detection.
const exe_check = b.addExecutable(.{
.name = "unsquashfs",
.root_module = mod,
});
const check = b.step("check", "Check if unsquashfs compiles");
check.dependOn(&exe_check.step);
}
+31 -87
View File
@@ -31,122 +31,66 @@ pub const FragEntry = packed struct {
const Archive = @This();
// 4 Gigs
const DEFAULT_MEM_SIZE = 4 * 1024 * 1024 * 1024;
parent_alloc: std.mem.Allocator,
alloc: std.heap.ThreadSafeAllocator,
// alloc: std.heap.FixedBufferAllocator,
// fixed_buf: []u8,
thread_count: usize,
fil: OffsetFile,
super: Superblock,
setup: bool = false,
decomp: Decomp.DecompFn,
frag_table: Table(FragEntry) = undefined,
id_table: Table(u16) = undefined,
export_table: Table(InodeRef) = undefined,
frag_table: Table(FragEntry),
id_table: Table(u16),
export_table: Table(InodeRef),
/// Default settings using std.Thread.getCpuCount() threads and the minimum of 4gb or half of system memory for memory usage.
pub fn init(alloc: std.mem.Allocator, fil: File) !Archive {
return initAdvanced(
alloc,
fil,
0,
try std.Thread.getCpuCount(),
);
}
/// Create the Archive dictating the amount of threads & memory used.
/// If trying to extract a full archive, a large memory size & thread count could help.
/// If you're planning on only interacting with a small number of files, it should be fine to use few threads and a small memory size.
pub fn initAdvanced(alloc: std.mem.Allocator, fil: File, offset: u64, threads: usize) !Archive {
/// Begin reading a squashfs archive from the given File at the given offset.
pub fn init(fil: File, offset: u64) !Archive {
var super: Superblock = undefined;
const red = try fil.pread(@ptrCast(&super), offset);
std.debug.assert(red == @sizeOf(Superblock));
_ = try fil.pread(@ptrCast(&super), offset);
try super.validate();
// const fixed_buf = try alloc.alloc(u8, mem);
return .{
.parent_alloc = alloc,
.alloc = .{ .child_allocator = alloc },
// .fixed_buf = fixed_buf,
.thread_count = threads,
.fil = .init(fil, offset),
.decomp = switch (super.compression) {
const decomp: Decomp.DecompFn = switch (super.compression) {
.gzip => Decomp.gzipDecompress,
.lzma => Decomp.lzmaDecompress,
.xz => Decomp.xzDecompress,
.zstd => Decomp.zstdDecompress,
.lz4 => if (config.use_c_libs) Decomp.cLz4 else return error.Lz4Unsupported,
.lzo => if (config.use_c_libs and config.allow_lzo) Decomp.lzoDecompress else return error.LzoUnsupported,
},
};
const offset_fil: OffsetFile = .init(fil, offset);
return .{
.fil = offset_fil,
.super = super,
.decomp = decomp,
.frag_table = .init(offset_fil, decomp, super.frag_start, super.frag_count),
.id_table = .init(offset_fil, decomp, super.id_start, super.id_count),
.export_table = .init(offset_fil, decomp, super.export_start, super.inode_count),
};
}
pub fn deinit(self: *Archive) void {
// self.parent_alloc.free(self.fixed_buf);
if (self.setup) {
self.frag_table.deinit();
self.export_table.deinit();
self.id_table.deinit();
}
}
pub fn allocator(self: *Archive) std.mem.Allocator {
return self.alloc.allocator();
}
fn setupValues(self: *Archive) !void {
const alloc = self.allocator();
self.frag_table = try .init(alloc, self.fil, self.decomp, self.super.frag_start, self.super.frag_count);
self.id_table = try .init(alloc, self.fil, self.decomp, self.super.id_start, self.super.id_count);
self.export_table = try .init(alloc, self.fil, self.decomp, self.super.export_start, self.super.inode_count);
self.setup = true;
}
pub fn id(self: *Archive, idx: u32) !u16 {
if (!self.setup) try self.setupValues();
return self.id_table.get(idx);
}
pub fn frag(self: *Archive, idx: u32) !FragEntry {
if (!self.setup) try self.setupValues();
return self.frag_table.get(idx);
}
pub fn inode(self: *Archive, num: u32) !Inode {
pub fn inode(self: Archive, num: u32) !Inode {
if (!self.setup) try self.setupValues();
const ref = try self.export_table.get(num - 1);
var rdr = try self.fil.readerAt(ref.block_start + self.super.inode_start, &[0]u8{});
var meta: MetadataReader = .init(self.allocator(), &rdr.interface, &self.decomp);
var meta: MetadataReader = .init(self.alloc, &rdr.interface, &self.decomp);
try meta.interface.discardAll(ref.block_offset);
return try .read(self.allocator(), &meta.interface, self.super.block_size);
return try .read(self.alloc, &meta.interface, self.super.block_size);
}
pub fn root(self: *Archive) !SfsFile {
if (!self.setup) try self.setupValues();
pub fn root(self: Archive, alloc: std.mem.Allocator) !SfsFile {
var rdr = try self.fil.readerAt(self.super.root_ref.block_start + self.super.inode_start, &[0]u8{});
var meta: MetadataReader = .init(self.allocator(), &rdr.interface, self.decomp);
var meta: MetadataReader = .init(alloc, &rdr.interface, self.decomp);
try meta.interface.discardAll(self.super.root_ref.block_offset);
const in: Inode = try .read(self.allocator(), &meta.interface, self.super.block_size);
return .init(self, in, "");
const in: Inode = try .read(alloc, &meta.interface, self.super.block_size);
return .init(alloc, self, in, "");
}
pub fn open(self: *Archive, path: []const u8) !SfsFile {
if (!self.setup) try self.setupValues();
var root_fil = try self.root();
pub fn open(self: Archive, alloc: std.mem.Allocator, path: []const u8) !SfsFile {
var root_fil = try self.root(alloc);
defer if (!SfsFile.pathIsSelf(path)) root_fil.deinit();
return root_fil.open(path);
return root_fil.open(alloc, path);
}
pub fn extract(self: *Archive, path: []const u8, options: ExtractionOptions) !void {
if (!self.setup) try self.setupValues();
var alloc = self.allocator();
pub fn extract(self: Archive, alloc: std.mem.Allocator, path: []const u8, options: ExtractionOptions) !void {
var ext_path: []u8 = undefined;
if (std.fs.cwd().statFile(path)) |stat| {
if (stat.kind == .directory) {
@@ -162,8 +106,8 @@ pub fn extract(self: *Archive, path: []const u8, options: ExtractionOptions) !vo
}
defer if (ext_path.len > path.len) alloc.free(ext_path);
var rdr = try self.fil.readerAt(self.super.root_ref.block_start + self.super.inode_start, &[0]u8{});
var meta: MetadataReader = .init(self.allocator(), &rdr.interface, self.decomp);
var meta: MetadataReader = .init(alloc, &rdr.interface, self.decomp);
try meta.interface.discardAll(self.super.root_ref.block_offset);
const in: Inode = try .read(self.allocator(), &meta.interface, self.super.block_size);
try in.extractToThreaded(self, ext_path, options, self.thread_count);
const in: Inode = try .read(alloc, &meta.interface, self.super.block_size);
try in.extractTo(alloc, self, ext_path, options);
}
+10 -6
View File
@@ -15,7 +15,8 @@ const help_mgs =
\\
\\ -o <offset> Start reading the archive at the given offset.
\\
\\ -p <threads> Specify how many threads to use. If no present, the system's logical cores count is used.
\\ -p <threads> Specify how many threads to use. If no present or zero, the system's logical cores count is used.
\\ -v Verbose
\\
\\ --help Display this messages
\\ --version Display the version
@@ -28,6 +29,7 @@ var archive: []const u8 = "";
var extLoc: []const u8 = "squashfs-root";
var offset: u64 = 0;
var threads: u32 = 0;
var verbose: bool = false;
pub fn main() !void {
const alloc = std.heap.smp_allocator;
@@ -42,9 +44,8 @@ pub fn main() !void {
}
var fil: std.fs.File = try std.fs.cwd().openFile(archive, .{}); //TODO: Handle error gracefully.
defer fil.close();
var arc: squashfs.Archive = try .initAdvanced(alloc, fil, offset, threads); //TODO: Update when memory size matters. //TODO: Handle error gracefully.
defer arc.deinit();
try arc.extract(extLoc, .Default); //TODO: Handle error gracefully.
var arc: squashfs.Archive = try .init(fil, offset); //TODO: Update when memory size matters. //TODO: Handle error gracefully.
try arc.extract(alloc, extLoc, if (verbose) try .VerboseDefault(&out.interface) else try .Default()); //TODO: Handle error gracefully.
}
fn handleArgs(alloc: std.mem.Allocator, out: *Writer) !void {
@@ -82,10 +83,13 @@ fn handleArgs(alloc: std.mem.Allocator, out: *Writer) !void {
return errors.InvalidArguments;
};
continue;
} else if (std.mem.eql(u8, arg, "-v")) {
verbose = true;
continue;
} else if (std.mem.eql(u8, arg, "--version")) {
try out.print("zig-unsquashfs version ", .{});
try out.print("zig-unsquashfs v", .{});
try config.version.format(out);
try out.print("\nBuilt using Zig {s} with {} backend in {} mode.\n", .{ builtin.zig_version_string, builtin.zig_backend, builtin.mode });
try out.print("\nBuilt using Zig {s} in {} mode\n", .{ builtin.zig_version_string, builtin.mode });
std.process.exit(0);
return;
} else if (std.mem.eql(u8, arg, "--help")) {
+1
View File
@@ -29,6 +29,7 @@ pub const CompressionType = enum(u16) {
zstd,
};
/// A generic decompression function. alloc is only used for internal use and any allocations made will be freed.
pub const DecompFn = *const fn (alloc: std.mem.Allocator, in: []u8, out: []u8) anyerror!usize; // TODO: replace anyerror to definitive error types.
pub const gzipDecompress = if (config.use_c_libs) cGzip else zigGzip;
+28 -54
View File
@@ -24,39 +24,40 @@ const FileError = error{
const SfsFile = @This();
archive: *Archive,
alloc: std.mem.Allocator,
archive: Archive,
inode: Inode,
name: []const u8,
/// Initialize a new File.
/// name is copied to the File so can be safely freed afterwards.
pub fn init(archive: *Archive, inode: Inode, name: []const u8) !SfsFile {
const new_name = try archive.allocator().alloc(u8, name.len);
pub fn init(alloc: std.mem.Allocator, archive: Archive, inode: Inode, name: []const u8) !SfsFile {
const new_name = try alloc.alloc(u8, name.len);
@memcpy(new_name, name);
return .{
.alloc = alloc,
.archive = archive,
.inode = inode,
.name = new_name,
};
}
pub fn fromEntry(archive: *Archive, entry: DirEntry) !SfsFile {
pub fn fromEntry(alloc: std.mem.Allocator, archive: Archive, entry: DirEntry) !SfsFile {
var rdr = try archive.fil.readerAt(entry.block_start + archive.super.inode_start, &[0]u8{});
var meta: MetadataReader = .init(archive.allocator(), &rdr.interface, archive.decomp);
var meta: MetadataReader = .init(alloc, &rdr.interface, archive.decomp);
try meta.interface.discardAll(entry.block_offset);
const inode: Inode = try .read(archive.allocator(), &meta.interface, archive.super.block_size);
errdefer inode.deinit(archive.allocator());
return .init(archive, inode, entry.name);
const inode: Inode = try .read(alloc, &meta.interface, archive.super.block_size);
errdefer inode.deinit(alloc);
return .init(alloc, archive, inode, entry.name);
}
pub fn deinit(self: SfsFile) void {
var alloc = self.archive.allocator();
alloc.free(self.name);
self.inode.deinit(alloc);
self.alloc.free(self.name);
self.inode.deinit(self.alloc);
}
fn getEntries(self: SfsFile) ![]DirEntry {
return self.inode.dirEntries(self.archive.allocator(), self.archive.*);
fn getEntries(self: SfsFile, alloc: std.mem.Allocator) ![]DirEntry {
return self.inode.dirEntries(alloc, self.archive);
}
pub fn ownerUid(self: SfsFile) !u16 {
@@ -96,20 +97,19 @@ pub fn iterate(self: SfsFile) !Iterator {
}
/// Open a sub-file/folder within a directory at the given path.
/// If path is "", ".", "/", or "./", this File is returned.
pub fn open(self: SfsFile, path: []const u8) !SfsFile {
pub fn open(self: SfsFile, alloc: std.mem.Allocator, path: []const u8) !SfsFile {
if (!self.isDir()) return FileError.NotDirectory;
if (pathIsSelf(path)) return self;
// Recursively stip ending & leading path separators.
if (path[0] == '/') return self.open(path[1..]);
if (path[path.len - 1] == '/') return self.open(path[0 .. path.len - 1]);
if (path[0] == '/') return self.open(alloc, path[1..]);
if (path[path.len - 1] == '/') return self.open(alloc, path[0 .. path.len - 1]);
const idx = std.mem.indexOf(u8, path, "/") orelse path.len;
const first_element = path[0..idx];
if (std.mem.eql(u8, first_element, ".")) return self.open(path[idx + 1 ..]);
const entries = try self.getEntries();
if (std.mem.eql(u8, first_element, ".")) return self.open(alloc, path[idx + 1 ..]);
const entries = try self.getEntries(alloc);
defer {
var alloc = self.archive.allocator();
for (entries) |e| {
e.deinit(alloc);
}
@@ -122,12 +122,12 @@ pub fn open(self: SfsFile, path: []const u8) !SfsFile {
const comp = std.mem.order(u8, first_element, cur_slice[split].name);
switch (comp) {
.eq => {
var fil: SfsFile = try .fromEntry(self.archive, cur_slice[split]);
var fil: SfsFile = try .fromEntry(alloc, self.archive, cur_slice[split]);
if (idx == path.len) {
return fil;
}
defer fil.deinit();
return fil.open(path[idx + 1 ..]);
return fil.open(alloc, path[idx + 1 ..]);
},
.lt => cur_slice = cur_slice[0..split],
.gt => cur_slice = cur_slice[split + 1 ..],
@@ -170,35 +170,8 @@ pub fn devNum(self: SfsFile) !u32 {
/// Extract the given File to the path. If File is a regular file, the path must be a directory or not exist.
/// If the gievn path is a folder, the File's contents will be extracted within.
pub fn extract(self: *SfsFile, path: []const u8, options: ExtractionOptions) !void {
var alloc = self.archive.allocator();
var ext_path: []u8 = undefined;
if (std.fs.cwd().statFile(path)) |stat| {
if (stat.kind == .directory) {
if (!self.isDir()) {
const has_end_sep = path[path.len - 1] == '/';
const alloc_size = if (has_end_sep)
path.len + self.name.len
else
path.len + self.name.len + 1;
ext_path = try alloc.alloc(u8, alloc_size);
@memcpy(ext_path[0..path.len], path);
@memcpy(ext_path[ext_path.len - self.name.len ..], self.name);
if (!has_end_sep) ext_path[path.len] = '/';
} else {
ext_path = @constCast(path);
}
} else return FileError.ExtractionPathExists;
} else |err| {
if (err == error.FileNotFound) {
ext_path = @constCast(path);
} else {
std.log.err("Error stat-ing extraction path {s}: {}\n", .{ path, err });
return err;
}
}
defer if (ext_path.len > path.len) alloc.free(ext_path);
return self.inode.extractToThreaded(self.archive, path, options, self.archive.thread_count);
pub fn extract(self: SfsFile, alloc: std.mem.Allocator, path: []const u8, options: ExtractionOptions) !void {
return self.inode.extractTo(alloc, self.archive, path, options);
}
/// Utility function.
@@ -210,8 +183,10 @@ pub fn pathIsSelf(path: []const u8) bool {
}
pub const Iterator = struct {
alloc: std.mem.Allocator,
entries: []DirEntry,
archive: *Archive,
archive: Archive,
idx: u32 = 0,
@@ -221,10 +196,9 @@ pub const Iterator = struct {
return try SfsFile.fromEntry(self.archive, self.entries[self.idx]);
}
pub fn deinit(self: Iterator) void {
var alloc = self.archive.allocator();
for (self.entries) |e| {
e.deinit(alloc);
e.deinit(self.alloc);
}
alloc.free(self.entries);
self.alloc.free(self.entries);
}
};
+176 -111
View File
@@ -4,6 +4,7 @@ const std = @import("std");
const Reader = std.Io.Reader;
const WaitGroup = std.Thread.WaitGroup;
const Pool = std.Thread.Pool;
const Mutex = std.Thread.Mutex;
const Archive = @import("archive.zig");
const DirEntry = @import("dir_entry.zig");
@@ -92,7 +93,7 @@ pub fn read(alloc: std.mem.Allocator, rdr: *Reader, block_size: u32) !Inode {
},
};
}
pub fn readFromEntry(alloc: std.mem.Allocator, archive: *Archive, entry: DirEntry) !Inode {
pub fn readFromEntry(alloc: std.mem.Allocator, archive: Archive, entry: DirEntry) !Inode {
var rdr = try archive.fil.readerAt(archive.super.inode_start + entry.block_start, &[0]u8{});
var meta: MetadataReader = .init(alloc, &rdr.interface, archive.decomp);
try meta.interface.discardAll(entry.block_offset);
@@ -110,32 +111,42 @@ pub fn deinit(self: Inode, alloc: std.mem.Allocator) void {
}
/// Get the data reader for a file inode.
pub fn dataReader(self: Inode, alloc: std.mem.Allocator, archive: *Archive) !DataReader {
pub fn dataReader(self: Inode, alloc: std.mem.Allocator, archive: Archive) !DataReader {
return switch (self.hdr.inode_type) {
.file => readerFromData(alloc, archive, self.data.file),
.ext_file => readerFromData(alloc, archive, self.data.ext_file),
else => error.NotRegularFile,
};
}
fn readerFromData(alloc: std.mem.Allocator, archive: *Archive, data: anytype) !DataReader {
var out: DataReader = .init(alloc, archive.*, data.block_sizes, data.block_start, data.size);
if (data.frag_idx != 0xFFFFFFFF)
out.addFragment(try archive.frag(data.frag_idx), data.frag_block_offset);
return out;
fn readerFromData(alloc: std.mem.Allocator, archive: Archive, data: anytype) !DataReader {
return .init(
alloc,
archive,
data.block_sizes,
data.block_start,
data.size,
data.frag_block_offset,
if (data.frag_idx == 0xFFFFFFFF) null else try archive.frag_table.get(alloc, data.frag_idx),
);
}
/// Get a threaded data reader for a file inode.
pub fn threadedDataReader(self: Inode, alloc: std.mem.Allocator, archive: *Archive) !ThreadedDataReader {
pub fn threadedDataReader(self: Inode, alloc: std.mem.Allocator, archive: Archive) !ThreadedDataReader {
return switch (self.hdr.inode_type) {
.file => threadedReaderFromData(alloc, archive, self.data.file),
.ext_file => threadedReaderFromData(alloc, archive, self.data.ext_file),
else => error.NotRegularFile,
};
}
fn threadedReaderFromData(alloc: std.mem.Allocator, archive: *Archive, data: anytype) !ThreadedDataReader {
var out: ThreadedDataReader = .init(alloc, archive.*, data.block_sizes, data.block_start, data.size);
if (data.frag_idx != 0xFFFFFFFF)
out.addFragment(try archive.frag(data.frag_idx), data.frag_block_offset);
return out;
fn threadedReaderFromData(alloc: std.mem.Allocator, archive: Archive, data: anytype) !ThreadedDataReader {
return .init(
alloc,
archive,
data.block_sizes,
data.block_start,
data.size,
data.frag_block_offset,
if (data.frag_idx == 0xFFFFFFFF) null else try archive.frag_table.get(alloc, data.frag_idx),
);
}
/// Get the directory entries for a directory inode.
@@ -153,17 +164,17 @@ fn entriesFromData(alloc: std.mem.Allocator, archive: Archive, data: anytype) ![
return DirEntry.readDir(alloc, &meta.interface, data.size);
}
/// Extract the inode to the given path. Single threaded.
pub fn extractTo(self: Inode, archive: *Archive, path: []const u8, options: ExtractionOptions) !void {
/// Extract the inode to the given path.
pub fn extractTo(self: Inode, alloc: std.mem.Allocator, archive: Archive, path: []const u8, options: ExtractionOptions) !void {
if (options.threads > 1) return self.extractToThreaded(alloc, archive, path, options); // We go to a dedicated function for mutli-threaded
switch (self.hdr.inode_type) {
.dir, .ext_dir => {
// Removing any trailing separators since that's the easiest path forward.
if (path[path.len - 1] == '/') return self.extractTo(archive, path[0 .. path.len - 1], options);
if (path[path.len - 1] == '/') return self.extractTo(alloc, archive, path[0 .. path.len - 1], options);
std.fs.cwd().makeDir(path) catch |err| {
if (err != std.fs.Dir.MakeError.PathAlreadyExists) return err;
};
var alloc = archive.allocator();
const entries = try self.dirEntries(alloc, archive.*);
const entries = try self.dirEntries(alloc, archive);
defer {
for (entries) |entry| entry.deinit(alloc);
alloc.free(entries);
@@ -177,96 +188,137 @@ pub fn extractTo(self: Inode, archive: *Archive, path: []const u8, options: Extr
var inode: Inode = try readFromEntry(alloc, archive, entry);
defer inode.deinit(alloc);
try inode.extractTo(archive, new_path, options);
try inode.extractTo(alloc, archive, new_path, options);
}
},
.file, .ext_file => try self.extractRegFile(archive.allocator(), archive, path, options),
.file, .ext_file => try self.extractRegFile(alloc, archive, path, options),
.symlink, .ext_symlink => try self.extractSymlink(path),
else => try self.extractDevice(archive, path, options),
else => try self.extractDevice(alloc, archive, path, options),
}
}
const Perms = struct {
const Parent = struct {
alloc: std.mem.Allocator,
path: []const u8,
uid: u16,
gid: u16,
perm: u16,
mod_time: u32,
ignore_permissions: bool,
ignore_xattr: bool,
wg: WaitGroup,
mut: Mutex = .{},
fn create(alloc: std.mem.Allocator, hdr: Header, archive: Archive, path: []const u8, options: ExtractionOptions, dir_size: usize) !*Parent {
const out = try alloc.create(Parent);
errdefer alloc.destroy(out);
out.* = .{
.alloc = alloc,
.path = path,
.uid = try archive.id_table.get(alloc, hdr.uid_idx),
.gid = try archive.id_table.get(alloc, hdr.gid_idx),
.perm = hdr.permissions,
.mod_time = hdr.mod_time,
.ignore_permissions = options.ignore_permissions,
.ignore_xattr = options.ignore_xattr,
.wg = .{},
};
out.wg.startMany(dir_size);
return out;
}
/// Extract the inode to the given path. Multi-threaded.
/// Functions identically to extractTo on all but regular files and directories.
///
/// If threads <= 1, then this just calls extractTo.
pub fn extractToThreaded(self: Inode, archive: *Archive, path: []const u8, options: ExtractionOptions, threads: usize) !void {
if (threads <= 1) return self.extractTo(archive, path, options);
switch (self.hdr.inode_type) {
.dir, .ext_dir => {
// Removing any trailing separators since that's the easiest path forward.
if (path[path.len - 1] == '/') return self.extractToThreaded(archive, path[0 .. path.len - 1], options, threads);
// Fixed Allocator
// const mem_buf = archive.allocator().alloc(u8, 2 * 1024 * 1024 * 1024);
// defer archive.allocator().free(mem_buf);
// var fixed_alloc: std.heap.FixedBufferAllocator = .init(mem_buf);
// const alloc = fixed_alloc.threadSafeAllocator();
// Arena Allocator
var arena_alloc: std.heap.ArenaAllocator = .init(archive.allocator());
defer arena_alloc.deinit();
var thread_alloc: std.heap.ThreadSafeAllocator = .{ .child_allocator = arena_alloc.allocator() };
const alloc = thread_alloc.allocator();
var wg: WaitGroup = .{};
var perms: ?std.ArrayList(Perms) = if (options.ignore_permissions) null else try .initCapacity(alloc, 100);
// defer if(!options.ignore_permissions) perms.?.deinit(alloc); We don't need to do this due to ArenaAllocator
var pool: Pool = undefined;
try pool.init(.{ .allocator = alloc, .n_jobs = threads - 1 });
defer pool.deinit();
var out_err: ?anyerror = null;
wg.start();
self.extractThread(alloc, archive, path, options, &wg, &pool, &out_err, &perms);
pool.waitAndWork(&wg);
if (out_err != null) return out_err.?;
if (perms != null) {
for (perms.?.items) |p| {
fn finish(p: *Parent) !void {
p.mut.lock();
{
defer p.mut.unlock();
p.wg.finish();
if (!p.wg.isDone()) return;
}
defer p.alloc.destroy(p);
var fil = try std.fs.cwd().openFile(p.path, .{});
defer fil.close();
const time = @as(i128, p.mod_time) * 1000000000;
try fil.updateTimes(time, time);
if (p.ignore_permissions) {
try fil.chmod(p.perm);
try fil.chown(p.uid, p.gid);
}
}
};
/// Extract the inode to the given path. Multi-threaded.
/// Functions identically to extractTo on all but regular files and directories.
fn extractToThreaded(self: Inode, alloc: std.mem.Allocator, archive: Archive, path: []const u8, options: ExtractionOptions) !void {
switch (self.hdr.inode_type) {
.dir, .ext_dir => {
// Removing any trailing separators since that's the easiest path forward.
if (path[path.len - 1] == '/') return self.extractToThreaded(alloc, archive, path[0 .. path.len - 1], options);
// Arena Allocator
var arena_alloc: std.heap.ArenaAllocator = .init(alloc);
defer arena_alloc.deinit();
var thread_alloc: std.heap.ThreadSafeAllocator = .{ .child_allocator = arena_alloc.allocator() };
var wg: WaitGroup = .{};
// defer if(!options.ignore_permissions) perms.?.deinit(alloc); We don't need to do this due to ArenaAllocator
var pool: Pool = undefined;
try pool.init(.{ .allocator = alloc, .n_jobs = options.threads - 1 });
defer pool.deinit();
var out_err: ?anyerror = null;
wg.start();
self.extractThread(thread_alloc.allocator(), archive, path, options, &wg, &pool, &out_err, null);
pool.waitAndWork(&wg);
if (out_err != null) return out_err.?;
var fil = try std.fs.cwd().openFile(path, .{});
defer fil.close();
const time = @as(i128, self.hdr.mod_time) * 1000000000;
try fil.updateTimes(time, time);
if (options.ignore_permissions) {
try fil.chmod(self.hdr.permissions);
try fil.chown(try archive.id_table.get(alloc, self.hdr.uid_idx), try archive.id_table.get(alloc, self.hdr.gid_idx));
}
},
.file, .ext_file => {
const alloc = archive.allocator();
var pool: Pool = undefined;
try pool.init(.{ .allocator = alloc, .n_jobs = threads });
try pool.init(.{ .allocator = alloc, .n_jobs = options.threads - 1 });
defer pool.deinit();
try self.extractRegFileThreaded(alloc, archive, path, options, &pool);
var thread_alloc: std.heap.ThreadSafeAllocator = .{ .child_allocator = alloc };
try self.extractRegFileThreaded(thread_alloc.allocator(), archive, path, options, &pool);
if (!options.ignore_permissions) {
var fil = try std.fs.cwd().openFile(path, .{});
defer fil.close();
const time = @as(i128, self.hdr.mod_time) * 1000000000;
try fil.updateTimes(time, time);
if (!options.ignore_permissions) {
try fil.chmod(self.hdr.permissions);
try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
try fil.chown(try archive.id_table.get(alloc, self.hdr.uid_idx), try archive.id_table.get(alloc, self.hdr.gid_idx));
}
},
.symlink, .ext_symlink => try self.extractSymlink(path),
else => try self.extractDevice(archive, path, options),
else => try self.extractDevice(alloc, archive, path, options),
}
}
fn extractThreadEntry(
entry: DirEntry,
alloc: std.mem.Allocator,
archive: *Archive,
archive: Archive,
path: []const u8,
options: ExtractionOptions,
wg: *WaitGroup,
pool: *Pool,
out_err: *?anyerror,
perms: *?std.ArrayList(Perms),
parent: ?*Parent,
) void {
var new_path = alloc.alloc(u8, path.len + entry.name.len + 1) catch |err| {
wg.finish();
@@ -281,33 +333,48 @@ fn extractThreadEntry(
wg.finish();
return;
};
inode.extractThread(alloc, archive, new_path, options, wg, pool, out_err, perms);
inode.extractThread(alloc, archive, new_path, options, wg, pool, out_err, parent);
}
/// Extract threadedly the inode to the path.
fn extractThread(
self: Inode,
alloc: std.mem.Allocator,
archive: *Archive,
archive: Archive,
path: []const u8,
options: ExtractionOptions,
wg: *WaitGroup,
pool: *Pool,
out_err: *?anyerror,
perms: *?std.ArrayList(Perms),
parent: ?*Parent,
) void {
defer wg.finish();
if (options.verbose)
options.verbose_writer.?.print("Extracting inode #{} to {s}\n", .{ self.hdr.num, path }) catch {};
defer {
if (parent != null) parent.?.finish() catch |err| {
if (options.verbose)
options.verbose_writer.?.print("Error setting folder permission to {s}: {}\n", .{ parent.?.path, err }) catch {};
out_err.* = err;
};
wg.finish();
}
if (out_err.* != null) return;
switch (self.hdr.inode_type) {
.dir, .ext_dir => {
std.fs.cwd().makeDir(path) catch |err| {
if (err != std.fs.Dir.MakeError.PathAlreadyExists) {
_ = std.fs.cwd().makePathStatus(path) catch |err| {
if (options.verbose)
options.verbose_writer.?.print("Error creating {s}: {}\n", .{ path, err }) catch {};
out_err.* = err;
return;
}
};
const entries = self.dirEntries(alloc, archive.*) catch |err| {
const entries = self.dirEntries(alloc, archive) catch |err| {
if (options.verbose)
options.verbose_writer.?.print("Error getting directory entries for inode #{} (extracting to {s}): {}\n", .{ self.hdr.num, path, err }) catch {};
out_err.* = err;
return;
};
const p = Parent.create(alloc, self.hdr, archive, path, options, entries.len) catch |err| {
out_err.* = err;
return;
};
@@ -315,7 +382,7 @@ fn extractThread(
// defer files.deinit(alloc); We don't need to do this due to ArenaAllocator
for (entries) |entry| {
if (entry.inode_type == .dir) {
extractThreadEntry(entry, alloc, archive, path, options, wg, pool, out_err, perms);
extractThreadEntry(entry, alloc, archive, path, options, wg, pool, out_err, p);
continue;
}
pool.spawn(
@@ -329,45 +396,39 @@ fn extractThread(
wg,
pool,
out_err,
perms,
p,
},
) catch |err| {
wg.finish();
p.finish() catch |e| {
if (options.verbose)
options.verbose_writer.?.print("Error setting folder permission to {s}: {}\n", .{ p.path, e }) catch {};
};
if (options.verbose)
options.verbose_writer.?.print("Error starting extraction thread: {}\n", .{err}) catch {};
out_err.* = err;
continue;
};
}
if (!options.ignore_permissions) {
const new_val = perms.*.?.addOne(alloc) catch |err| {
out_err.* = err;
return;
};
new_val.* = .{
.path = path,
.uid = archive.id(self.hdr.uid_idx) catch |err| {
out_err.* = err;
return;
},
.gid = archive.id(self.hdr.gid_idx) catch |err| {
out_err.* = err;
return;
},
.perm = self.hdr.permissions,
};
}
},
.file, .ext_file => {
self.extractRegFileThreaded(alloc, archive, path, options, pool) catch |err| {
if (options.verbose)
options.verbose_writer.?.print("Error extracting file inode #{} to {s}: {}\n", .{ self.hdr.num, path, err }) catch {};
out_err.* = err;
};
},
.symlink, .ext_symlink => {
self.extractSymlink(path) catch |err| {
if (options.verbose)
options.verbose_writer.?.print("Error extracting symlink inode #{} to {s}: {}\n", .{ self.hdr.num, path, err }) catch {};
out_err.* = err;
};
},
else => {
self.extractDevice(archive, path, options) catch |err| {
self.extractDevice(alloc, archive, path, options) catch |err| {
if (options.verbose)
options.verbose_writer.?.print("Error extracting device/IPC inode #{} to {s}: {}\n", .{ self.hdr.num, path, err }) catch {};
out_err.* = err;
};
},
@@ -377,33 +438,37 @@ fn extractThread(
/// Optionally set owner & permissions.
///
/// Assumes the inode is a file or ext_file type.
fn extractRegFile(self: Inode, alloc: std.mem.Allocator, archive: *Archive, path: []const u8, options: ExtractionOptions) !void {
var fil = try std.fs.cwd().createFile(path, .{});
fn extractRegFile(self: Inode, alloc: std.mem.Allocator, archive: Archive, path: []const u8, options: ExtractionOptions) !void {
var fil = try std.fs.cwd().createFile(path, .{ .exclusive = true });
defer fil.close();
var wrt = fil.writer(&[0]u8{});
var dat_rdr = try self.dataReader(alloc, archive);
defer dat_rdr.deinit();
_ = try dat_rdr.interface.streamRemaining(&wrt.interface);
try wrt.interface.flush();
// updateTime is in nanoseconds (a billionth of a second). mod_time is in seconds.
// TODO: fix
// try fil.updateTimes(self.hdr.mod_time, self.hdr.mod_time);
const time = @as(i128, self.hdr.mod_time) * 1000000000;
try fil.updateTimes(time, time);
if (!options.ignore_permissions) {
try fil.chmod(self.hdr.permissions);
try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
try fil.chown(try archive.id_table.get(alloc, self.hdr.uid_idx), try archive.id_table.get(alloc, self.hdr.gid_idx));
}
}
/// Extract the inode file contents to the given path threadedly.
/// pool is used to spawn threads.
///
/// Assumes the inode is a file or ext_file type.
fn extractRegFileThreaded(self: Inode, alloc: std.mem.Allocator, archive: *Archive, path: []const u8, options: ExtractionOptions, pool: *Pool) !void {
fn extractRegFileThreaded(self: Inode, alloc: std.mem.Allocator, archive: Archive, path: []const u8, options: ExtractionOptions, pool: *Pool) !void {
var fil = try std.fs.cwd().createFile(path, .{});
defer fil.close();
var data = try self.threadedDataReader(alloc, archive);
try data.extractThreaded(fil, pool);
const time = @as(i128, self.hdr.mod_time) * 1000000000;
try fil.updateTimes(time, time);
if (!options.ignore_permissions) {
try fil.chmod(self.hdr.permissions);
try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
try fil.chown(try archive.id_table.get(alloc, self.hdr.uid_idx), try archive.id_table.get(alloc, self.hdr.gid_idx));
}
}
/// Creates the symlink described by the inode.
@@ -421,7 +486,7 @@ fn extractSymlink(self: Inode, path: []const u8) !void {
///
/// Optionally set owner & permissions.
/// Assumes the inode is a char_dev, block_dev, fifo, socket, or their extended counterparts.
fn extractDevice(self: Inode, archive: *Archive, path: []const u8, options: ExtractionOptions) !void {
fn extractDevice(self: Inode, alloc: std.mem.Allocator, archive: Archive, path: []const u8, options: ExtractionOptions) !void {
var mode: u32 = undefined;
var dev: u32 = 0;
switch (self.data) {
@@ -465,12 +530,12 @@ fn extractDevice(self: Inode, archive: *Archive, path: []const u8, options: Extr
},
}
var fil = try std.fs.cwd().openFile(path, .{});
// updateTime is in nanoseconds (a billionth of a second). mod_time is in seconds.
// TODO: fix
// try fil.updateTimes(self.hdr.mod_time, self.hdr.mod_time);
defer fil.close();
const time = @as(i128, self.hdr.mod_time) * 1000000000;
try fil.updateTimes(time, time);
if (!options.ignore_permissions) {
try fil.chmod(self.hdr.permissions);
try fil.chown(try archive.id(self.hdr.uid_idx), try archive.id(self.hdr.gid_idx));
try fil.chown(try archive.id_table.get(alloc, self.hdr.uid_idx), try archive.id_table.get(alloc, self.hdr.gid_idx));
}
if (!options.ignore_xattr) {
// TODO
+11 -2
View File
@@ -5,6 +5,8 @@ const Writer = std.Io.Writer;
const ExtractionOptions = @This();
/// The amount of threads to use for extraction.
threads: usize,
/// Don't set the file's owner & permissions after extraction
ignore_permissions: bool = false,
/// Don't set xattr values. Currently xattrs are never set anyway.
@@ -16,9 +18,16 @@ verbose: bool = false,
/// Where to print verbose log.
verbose_writer: ?*Writer = null,
pub const Default: ExtractionOptions = .{};
pub fn VerboseDefault(wrt: *Writer) ExtractionOptions {
pub const SingleThreaded: ExtractionOptions = .{ .threads = 1 };
/// Use std.Thread.getCpuCount threads for extraction.
pub fn Default() !ExtractionOptions {
return .{
.threads = try std.Thread.getCpuCount(),
};
}
pub fn VerboseDefault(wrt: *Writer) !ExtractionOptions {
return .{
.threads = try std.Thread.getCpuCount(),
.verbose = true,
.verbose_writer = wrt,
};
+7 -34
View File
@@ -16,61 +16,34 @@ pub fn Table(T: anytype) type {
const VALS_PER_BLOCK = 8192 / @sizeOf(T);
alloc: std.mem.Allocator,
fil: OffsetFile,
decomp: DecompFn,
tab_start: u64,
tab: std.AutoHashMap(u32, []T),
tab_start: u64,
values: u32,
mut: Mutex = .{},
pub fn init(alloc: std.mem.Allocator, fil: OffsetFile, decomp: DecompFn, tab_start: u64, values: u32) !Self {
pub fn init(fil: OffsetFile, decomp: DecompFn, tab_start: u64, values: u32) Self {
return .{
.alloc = alloc,
.fil = fil,
.decomp = decomp,
.tab_start = tab_start,
.tab = .init(alloc),
.tab_start = tab_start,
.values = values,
};
}
pub fn deinit(self: *Self) void {
var iter = self.tab.valueIterator();
while (iter.next()) |s| {
self.alloc.free(s.*);
}
self.tab.deinit();
}
pub fn get(self: *Self, idx: u32) !T {
if (idx >= self.values) return TableError.InvalidIndex;
pub fn get(self: Self, alloc: std.mem.Allocator, idx: u32) !T {
const block_num = idx / VALS_PER_BLOCK;
const idx_offset = idx - (block_num * VALS_PER_BLOCK);
if (self.tab.contains(block_num)) {
const block = self.tab.get(block_num).?;
return block[idx_offset];
}
self.mut.lock();
defer self.mut.unlock();
// Double check in case of race condition..
if (self.tab.contains(block_num)) {
const block = self.tab.get(block_num).?;
return block[idx_offset];
}
const is_last = (self.values - 1) / VALS_PER_BLOCK == block_num;
const slice_size = if (is_last) self.values - (block_num * VALS_PER_BLOCK) else VALS_PER_BLOCK;
const slice = try self.alloc.alloc(T, slice_size);
var slice: [VALS_PER_BLOCK]T = undefined;
var rdr = try self.fil.readerAt(self.tab_start + (8 * block_num), &[0]u8{});
var offset: u64 = 0;
try rdr.interface.readSliceEndian(u64, @ptrCast(&offset), .little);
rdr = try self.fil.readerAt(offset, &[0]u8{});
var meta: MetadataReader = .init(self.alloc, &rdr.interface, self.decomp);
try meta.interface.readSliceEndian(T, @ptrCast(slice), .little);
try self.tab.put(block_num, slice);
var meta: MetadataReader = .init(alloc, &rdr.interface, self.decomp);
try meta.interface.readSliceEndian(T, slice[0..slice_size], .little);
return slice[idx_offset];
}
};
+6 -9
View File
@@ -9,8 +9,7 @@ const TestArchive = "testing/LinuxPATest.sfs";
test "Basics" {
var fil = try std.fs.cwd().openFile(TestArchive, .{});
defer fil.close();
var sfs: Archive = try .init(std.testing.allocator, fil);
defer sfs.deinit();
const sfs: Archive = try .init(fil, 0);
if (sfs.super != LinuxPATestCorrectSuperblock) {
std.debug.print("Superblock wrong\nShould be: {}\n\nis: {}\n", .{ LinuxPATestCorrectSuperblock, sfs.super });
return error.BadSuperblock;
@@ -24,11 +23,10 @@ test "ExtractSingleFile" {
std.fs.cwd().deleteFile(TestFileExtractLocation) catch {};
var fil = try std.fs.cwd().openFile(TestArchive, .{});
defer fil.close();
var sfs: Archive = try .init(std.testing.allocator, fil);
defer sfs.deinit();
var test_fil = try sfs.open(TestFile);
var sfs: Archive = try .init(fil, 0);
var test_fil = try sfs.open(std.testing.allocator, TestFile);
defer test_fil.deinit();
try test_fil.extract(TestFileExtractLocation, .Default);
try test_fil.extract(std.testing.allocator, TestFileExtractLocation, try .Default());
//TODO: validate extracted file.
}
@@ -38,9 +36,8 @@ test "ExtractCompleteArchive" {
std.fs.cwd().deleteTree(TestFullExtractLocation) catch {};
var fil = try std.fs.cwd().openFile(TestArchive, .{});
defer fil.close();
var sfs: Archive = try .init(std.testing.allocator, fil);
defer sfs.deinit();
try sfs.extract(TestFullExtractLocation, .Default);
var sfs: Archive = try .init(fil, 0);
try sfs.extract(std.testing.allocator, TestFullExtractLocation, try .Default());
}
const LinuxPATestCorrectSuperblock: Superblock = .{
+4 -6
View File
@@ -29,13 +29,16 @@ interface: Reader,
cur_offset: u64,
block_idx: u32 = 0,
pub fn init(alloc: std.mem.Allocator, archive: Archive, blocks: []BlockSize, start: u64, size: u64) DataReader {
pub fn init(alloc: std.mem.Allocator, archive: Archive, blocks: []BlockSize, start: u64, size: u64, frag_offset: u32, frag_entry: ?FragEntry) DataReader {
return .{
.alloc = alloc,
.fil = archive.fil,
.decomp = archive.decomp,
.block_size = archive.super.block_size,
.blocks = blocks,
.frag = frag_entry,
.frag_offset = frag_offset,
.size = size,
.cur_offset = start,
.interface = .{
@@ -56,11 +59,6 @@ pub fn deinit(self: *DataReader) void {
self.interface.seek = 0;
}
pub fn addFragment(self: *DataReader, entry: FragEntry, frag_offset: u32) void {
self.frag = entry;
self.frag_offset = frag_offset;
}
fn numBlocks(self: DataReader) usize {
var res = self.blocks.len;
if (self.frag != null) res += 1;
+4 -6
View File
@@ -28,7 +28,7 @@ size: u64,
start_offset: u64,
pub fn init(alloc: std.mem.Allocator, archive: Archive, blocks: []BlockSize, start: u64, size: u64) ThreadedDataReader {
pub fn init(alloc: std.mem.Allocator, archive: Archive, blocks: []BlockSize, start: u64, size: u64, frag_offset: u32, frag_entry: ?FragEntry) ThreadedDataReader {
return .{
.alloc = alloc,
.fil = archive.fil,
@@ -37,12 +37,10 @@ pub fn init(alloc: std.mem.Allocator, archive: Archive, blocks: []BlockSize, sta
.blocks = blocks,
.size = size,
.start_offset = start,
};
}
pub fn addFragment(self: *ThreadedDataReader, entry: FragEntry, frag_offset: u32) void {
self.frag = entry;
self.frag_offset = frag_offset;
.frag_offset = frag_offset,
.frag = frag_entry,
};
}
fn numBlocks(self: ThreadedDataReader) usize {