proxmox-backup/src/pxar/encoder.rs

1147 lines
39 KiB
Rust
Raw Normal View History

//! *pxar* format encoder.
2018-12-30 12:47:27 +00:00
//!
//! This module contain the code to generate *pxar* archive files.
2018-12-30 12:47:27 +00:00
2018-12-27 12:15:10 +00:00
use failure::*;
use endian_trait::Endian;
use std::collections::{HashSet, HashMap};
2018-12-27 12:15:10 +00:00
use super::flags;
2018-12-27 12:15:10 +00:00
use super::format_definition::*;
use super::binary_search_tree::*;
use super::helper::*;
use super::match_pattern::*;
use crate::tools::fs;
use crate::tools::acl;
use crate::tools::xattr;
2018-12-27 12:15:10 +00:00
use std::io::Write;
use std::os::unix::io::AsRawFd;
use std::os::unix::ffi::OsStrExt;
use std::os::unix::io::RawFd;
use std::path::{Path, PathBuf};
use std::ffi::CStr;
2018-12-28 10:48:47 +00:00
2018-12-28 09:44:12 +00:00
use nix::NixPath;
2018-12-28 06:14:12 +00:00
use nix::fcntl::OFlag;
use nix::sys::stat::Mode;
use nix::errno::Errno;
use nix::sys::stat::FileStat;
2018-12-27 12:15:10 +00:00
use proxmox::tools::vec;
/// The format requires to build sorted directory lookup tables in
/// memory, so we restrict the number of allowed entries to limit
/// maximum memory usage.
pub const MAX_DIRECTORY_ENTRIES: usize = 256*1024;
2018-12-28 09:44:12 +00:00
2019-03-15 11:19:51 +00:00
#[derive(Eq, PartialEq, Hash)]
struct HardLinkInfo {
st_dev: u64,
st_ino: u64,
}
2019-03-15 07:24:32 +00:00
pub struct Encoder<'a, W: Write> {
2019-03-15 11:19:51 +00:00
base_path: PathBuf,
relative_path: PathBuf,
writer: &'a mut W,
2018-12-28 10:48:47 +00:00
writer_pos: usize,
2019-01-30 17:25:37 +00:00
_size: usize,
2018-12-28 10:48:47 +00:00
file_copy_buffer: Vec<u8>,
device_set: Option<HashSet<u64>>,
verbose: bool,
// Flags set by the user
feature_flags: u64,
// Flags signaling features supported by the filesystem
fs_feature_flags: u64,
hardlinks: HashMap<HardLinkInfo, (PathBuf, u64)>,
2018-12-27 12:15:10 +00:00
}
2019-03-15 07:24:32 +00:00
impl <'a, W: Write> Encoder<'a, W> {
2019-03-15 11:19:51 +00:00
// used for error reporting
fn full_path(&self) -> PathBuf {
self.base_path.join(&self.relative_path)
}
/// Create archive, write result data to ``writer``.
///
/// The ``device_set`` can be use used to limit included mount points.
///
/// - ``None``: include all mount points
/// - ``Some(set)``: only include devices listed in this set (the
/// root path device is automathically added to this list, so
/// you can pass an empty set if you want to archive a single
/// mount point.)
pub fn encode(
path: PathBuf,
dir: &mut nix::dir::Dir,
writer: &'a mut W,
device_set: Option<HashSet<u64>>,
verbose: bool,
skip_lost_and_found: bool, // fixme: should be a feature flag ??
feature_flags: u64,
) -> Result<(), Error> {
2018-12-28 10:48:47 +00:00
const FILE_COPY_BUFFER_SIZE: usize = 1024*1024;
2018-12-28 10:48:47 +00:00
let mut file_copy_buffer = Vec::with_capacity(FILE_COPY_BUFFER_SIZE);
unsafe { file_copy_buffer.set_len(FILE_COPY_BUFFER_SIZE); }
2018-12-27 12:15:10 +00:00
// todo: use scandirat??
let dir_fd = dir.as_raw_fd();
let stat = nix::sys::stat::fstat(dir_fd)
.map_err(|err| format_err!("fstat {:?} failed - {}", path, err))?;
if !is_directory(&stat) {
bail!("got unexpected file type {:?} (not a directory)", path);
}
let mut device_set = device_set.clone();
if let Some(ref mut set) = device_set {
set.insert(stat.st_dev);
}
let magic = detect_fs_type(dir_fd)?;
if is_virtual_file_system(magic) {
bail!("backup virtual file systems is disabled!");
}
let fs_feature_flags = flags::feature_flags_from_magic(magic);
let mut me = Self {
2019-03-15 11:19:51 +00:00
base_path: path,
relative_path: PathBuf::new(),
writer: writer,
writer_pos: 0,
_size: 0,
file_copy_buffer,
device_set,
verbose,
feature_flags,
fs_feature_flags,
2019-03-15 11:19:51 +00:00
hardlinks: HashMap::new(),
};
2019-03-15 11:19:51 +00:00
if verbose { println!("{:?}", me.full_path()); }
let mut excludes = Vec::new();
if skip_lost_and_found {
excludes.push(MatchPattern::from_line(b"**/lost+found").unwrap().unwrap());
}
me.encode_dir(dir, &stat, magic, excludes)?;
2018-12-27 12:15:10 +00:00
Ok(())
}
2018-12-28 10:48:47 +00:00
fn write(&mut self, buf: &[u8]) -> Result<(), Error> {
self.writer.write_all(buf)?;
2018-12-28 10:48:47 +00:00
self.writer_pos += buf.len();
Ok(())
}
2019-01-11 09:01:51 +00:00
fn write_item<T: Endian>(&mut self, item: T) -> Result<(), Error> {
2019-01-11 09:01:51 +00:00
let data = item.to_le();
let buffer = unsafe { std::slice::from_raw_parts(
&data as *const T as *const u8,
std::mem::size_of::<T>()
)};
self.write(buffer)?;
Ok(())
}
2018-12-28 10:48:47 +00:00
fn flush_copy_buffer(&mut self, size: usize) -> Result<(), Error> {
self.writer.write_all(&self.file_copy_buffer[..size])?;
2018-12-28 10:48:47 +00:00
self.writer_pos += size;
Ok(())
}
fn write_header(&mut self, htype: u64, size: u64) -> Result<(), Error> {
let size = size + (std::mem::size_of::<PxarHeader>() as u64);
self.write_item(PxarHeader { size, htype })?;
2018-12-28 10:48:47 +00:00
Ok(())
}
fn write_filename(&mut self, name: &CStr) -> Result<(), Error> {
let buffer = name.to_bytes_with_nul();
self.write_header(PXAR_FILENAME, buffer.len() as u64)?;
2018-12-28 10:48:47 +00:00
self.write(buffer)?;
Ok(())
2018-12-28 10:48:47 +00:00
}
fn create_entry(&self, stat: &FileStat) -> Result<PxarEntry, Error> {
let mode = if is_symlink(&stat) {
(libc::S_IFLNK | 0o777) as u64
} else {
(stat.st_mode & (libc::S_IFMT | 0o7777)) as u64
};
let mtime = stat.st_mtime * 1_000_000_000 + stat.st_mtime_nsec;
if mtime < 0 {
2019-03-15 11:19:51 +00:00
bail!("got strange mtime ({}) from fstat for {:?}.", mtime, self.full_path());
}
let entry = PxarEntry {
mode: mode,
flags: 0,
uid: stat.st_uid,
gid: stat.st_gid,
mtime: mtime as u64,
};
Ok(entry)
}
fn read_chattr(&self, fd: RawFd, entry: &mut PxarEntry) -> Result<(), Error> {
let mut attr: usize = 0;
let res = unsafe { fs::read_attr_fd(fd, &mut attr)};
if let Err(err) = res {
if let nix::Error::Sys(errno) = err {
if errno_is_unsupported(errno) { return Ok(()) };
}
2019-03-15 11:19:51 +00:00
bail!("read_attr_fd failed for {:?} - {}", self.full_path(), err);
}
let flags = flags::feature_flags_from_chattr(attr as u32);
entry.flags = entry.flags | flags;
Ok(())
}
fn read_fat_attr(&self, fd: RawFd, magic: i64, entry: &mut PxarEntry) -> Result<(), Error> {
use fs::magic::*;
if magic != MSDOS_SUPER_MAGIC && magic != FUSE_SUPER_MAGIC {
return Ok(());
}
let mut attr: u32 = 0;
let res = unsafe { fs::read_fat_attr_fd(fd, &mut attr)};
if let Err(err) = res {
if let nix::Error::Sys(errno) = err {
if errno_is_unsupported(errno) { return Ok(()) };
}
2019-03-15 11:19:51 +00:00
bail!("read_fat_attr_fd failed for {:?} - {}", self.full_path(), err);
}
let flags = flags::feature_flags_from_fat_attr(attr);
entry.flags = entry.flags | flags;
Ok(())
}
/// True if all of the given feature flags are set in the Encoder, false otherwise
fn has_features(&self, feature_flags: u64) -> bool {
(self.feature_flags & self.fs_feature_flags & feature_flags) == feature_flags
}
/// True if at least one of the given feature flags is set in the Encoder, false otherwise
fn has_some_features(&self, feature_flags: u64) -> bool {
(self.feature_flags & self.fs_feature_flags & feature_flags) != 0
}
fn read_xattrs(&self, fd: RawFd, stat: &FileStat) -> Result<(Vec<PxarXAttr>, Option<PxarFCaps>), Error> {
let mut xattrs = Vec::new();
let mut fcaps = None;
let flags = flags::WITH_XATTRS | flags::WITH_FCAPS;
if !self.has_some_features(flags) {
return Ok((xattrs, fcaps));
}
// Should never be called on symlinks, just in case check anyway
if is_symlink(&stat) {
return Ok((xattrs, fcaps));
}
let xattr_names = match xattr::flistxattr(fd) {
Ok(names) => names,
// Do not bail if the underlying endpoint does not supports xattrs
Err(Errno::EOPNOTSUPP) => return Ok((xattrs, fcaps)),
// Do not bail if the endpoint cannot carry xattrs (such as symlinks)
Err(Errno::EBADF) => return Ok((xattrs, fcaps)),
Err(err) => bail!("read_xattrs failed for {:?} - {}", self.full_path(), err),
};
for name in xattr_names.split(|c| *c == b'\0') {
// Only extract the relevant extended attributes
if !xattr::is_valid_xattr_name(&name) {
continue;
}
let value = match xattr::fgetxattr(fd, name) {
Ok(value) => value,
// Vanished between flistattr and getxattr, this is ok, silently ignore
Err(Errno::ENODATA) => continue,
Err(err) => bail!("read_xattrs failed for {:?} - {}", self.full_path(), err),
};
if xattr::is_security_capability(&name) {
if self.has_features(flags::WITH_FCAPS) {
// fcaps are stored in own format within the archive
fcaps = Some(PxarFCaps {
data: value,
});
}
} else if self.has_features(flags::WITH_XATTRS) {
xattrs.push(PxarXAttr {
name: name.to_vec(),
value: value,
});
}
}
xattrs.sort();
Ok((xattrs, fcaps))
}
fn read_acl(&self, fd: RawFd, stat: &FileStat, acl_type: acl::ACLType) -> Result<PxarACL, Error> {
let ret = PxarACL {
users: Vec::new(),
groups: Vec::new(),
group_obj: None,
default: None,
};
if !self.has_features(flags::WITH_ACL) {
return Ok(ret);
}
if is_symlink(&stat) {
return Ok(ret);
}
if acl_type == acl::ACL_TYPE_DEFAULT && !is_directory(&stat) {
bail!("ACL_TYPE_DEFAULT only defined for directories.");
}
// In order to be able to get ACLs with type ACL_TYPE_DEFAULT, we have
// to create a path for acl_get_file(). acl_get_fd() only allows to get
// ACL_TYPE_ACCESS attributes.
let proc_path = Path::new("/proc/self/fd/").join(fd.to_string());
let acl = match acl::ACL::get_file(&proc_path, acl_type) {
Ok(acl) => acl,
// Don't bail if underlying endpoint does not support acls
Err(Errno::EOPNOTSUPP) => return Ok(ret),
// Don't bail if the endpoint cannot carry acls
Err(Errno::EBADF) => return Ok(ret),
// Don't bail if there is no data
Err(Errno::ENODATA) => return Ok(ret),
Err(err) => bail!("error while reading ACL - {}", err),
};
self.process_acl(acl, acl_type)
}
fn process_acl(&self, acl: acl::ACL, acl_type: acl::ACLType) -> Result<PxarACL, Error> {
let mut acl_user = Vec::new();
let mut acl_group = Vec::new();
let mut acl_group_obj = None;
let mut acl_default = None;
let mut user_obj_permissions = None;
let mut group_obj_permissions = None;
let mut other_permissions = None;
let mut mask_permissions = None;
for entry in &mut acl.entries() {
let tag = entry.get_tag_type()?;
let permissions = entry.get_permissions()?;
match tag {
acl::ACL_USER_OBJ => user_obj_permissions = Some(permissions),
acl::ACL_GROUP_OBJ => group_obj_permissions = Some(permissions),
acl::ACL_OTHER => other_permissions = Some(permissions),
acl::ACL_MASK => mask_permissions = Some(permissions),
acl::ACL_USER => {
acl_user.push(PxarACLUser {
uid: entry.get_qualifier()?,
permissions: permissions,
});
},
acl::ACL_GROUP => {
acl_group.push(PxarACLGroup {
gid: entry.get_qualifier()?,
permissions: permissions,
});
},
_ => bail!("Unexpected ACL tag encountered!"),
}
}
acl_user.sort();
acl_group.sort();
match acl_type {
acl::ACL_TYPE_ACCESS => {
// The mask permissions are mapped to the stat group permissions
// in case that the ACL group permissions were set.
// Only in that case we need to store the group permissions,
// in the other cases they are identical to the stat group permissions.
if let (Some(gop), Some(_)) = (group_obj_permissions, mask_permissions) {
acl_group_obj = Some(PxarACLGroupObj {
permissions: gop,
});
}
},
acl::ACL_TYPE_DEFAULT => {
if user_obj_permissions != None ||
group_obj_permissions != None ||
other_permissions != None ||
mask_permissions != None
{
acl_default = Some(PxarACLDefault {
// The value is set to UINT64_MAX as placeholder if one
// of the permissions is not set
user_obj_permissions: user_obj_permissions.unwrap_or(std::u64::MAX),
group_obj_permissions: group_obj_permissions.unwrap_or(std::u64::MAX),
other_permissions: other_permissions.unwrap_or(std::u64::MAX),
mask_permissions: mask_permissions.unwrap_or(std::u64::MAX),
});
}
},
_ => bail!("Unexpected ACL type encountered"),
}
Ok(PxarACL {
users: acl_user,
groups: acl_group,
group_obj: acl_group_obj,
default: acl_default,
})
}
/// Read the quota project id for an inode, supported on ext4/XFS/FUSE/ZFS filesystems
fn read_quota_project_id(&self, fd: RawFd, magic: i64, stat: &FileStat) -> Result<Option<PxarQuotaProjID>, Error> {
if !(is_directory(&stat) || is_reg_file(&stat)) {
return Ok(None);
}
if !self.has_features(flags::WITH_QUOTA_PROJID) {
return Ok(None);
}
match magic {
fs::magic::EXT4_SUPER_MAGIC |
fs::magic::XFS_SUPER_MAGIC |
fs::magic::FUSE_SUPER_MAGIC |
fs::magic::ZFS_SUPER_MAGIC => {
let mut fsxattr = fs::FSXAttr::default();
let res = unsafe {
fs::fs_ioc_fsgetxattr(fd, &mut fsxattr)
};
// On some FUSE filesystems it can happen that ioctl is not supported.
// For these cases projid is set to 0 while the error is ignored.
if let Err(err) = res {
let errno = err.as_errno().ok_or_else(|| {
format_err!("error while reading quota project id for {:#?}", self.full_path())
})?;
if errno_is_unsupported(errno) {
return Ok(None);
} else {
bail!("error while reading quota project id for {:#?} - {}", self.full_path(), errno);
}
}
let projid = fsxattr.fsx_projid as u64;
if projid == 0 {
return Ok(None);
} else {
return Ok(Some(PxarQuotaProjID { projid }));
}
},
_ => return Ok(None),
}
}
fn write_entry(&mut self, entry: PxarEntry) -> Result<(), Error> {
self.write_header(PXAR_ENTRY, std::mem::size_of::<PxarEntry>() as u64)?;
self.write_item(entry)?;
Ok(())
}
fn write_xattr(&mut self, xattr: PxarXAttr) -> Result<(), Error> {
let size = xattr.name.len() + xattr.value.len() + 1; // +1 for '\0' separating name and value
self.write_header(PXAR_XATTR, size as u64)?;
self.write(xattr.name.as_slice())?;
self.write(&[0])?;
self.write(xattr.value.as_slice())?;
Ok(())
}
fn write_fcaps(&mut self, fcaps: Option<PxarFCaps>) -> Result<(), Error> {
if let Some(fcaps) = fcaps {
let size = fcaps.data.len();
self.write_header(PXAR_FCAPS, size as u64)?;
self.write(fcaps.data.as_slice())?;
}
Ok(())
}
fn write_acl_user(&mut self, acl_user: PxarACLUser) -> Result<(), Error> {
self.write_header(PXAR_ACL_USER, std::mem::size_of::<PxarACLUser>() as u64)?;
self.write_item(acl_user)?;
Ok(())
}
fn write_acl_group(&mut self, acl_group: PxarACLGroup) -> Result<(), Error> {
self.write_header(PXAR_ACL_GROUP, std::mem::size_of::<PxarACLGroup>() as u64)?;
self.write_item(acl_group)?;
Ok(())
}
fn write_acl_group_obj(&mut self, acl_group_obj: PxarACLGroupObj) -> Result<(), Error> {
self.write_header(PXAR_ACL_GROUP_OBJ, std::mem::size_of::<PxarACLGroupObj>() as u64)?;
self.write_item(acl_group_obj)?;
Ok(())
}
fn write_acl_default(&mut self, acl_default: PxarACLDefault) -> Result<(), Error> {
self.write_header(PXAR_ACL_DEFAULT, std::mem::size_of::<PxarACLDefault>() as u64)?;
self.write_item(acl_default)?;
Ok(())
}
fn write_acl_default_user(&mut self, acl_default_user: PxarACLUser) -> Result<(), Error> {
self.write_header(PXAR_ACL_DEFAULT_USER, std::mem::size_of::<PxarACLUser>() as u64)?;
self.write_item(acl_default_user)?;
Ok(())
}
fn write_acl_default_group(&mut self, acl_default_group: PxarACLGroup) -> Result<(), Error> {
self.write_header(PXAR_ACL_DEFAULT_GROUP, std::mem::size_of::<PxarACLGroup>() as u64)?;
self.write_item(acl_default_group)?;
Ok(())
}
fn write_quota_project_id(&mut self, projid: PxarQuotaProjID) -> Result<(), Error> {
self.write_header(PXAR_QUOTA_PROJID, std::mem::size_of::<PxarQuotaProjID>() as u64)?;
self.write_item(projid)?;
Ok(())
}
fn write_goodbye_table(&mut self, goodbye_offset: usize, goodbye_items: &mut [PxarGoodbyeItem]) -> Result<(), Error> {
goodbye_items.sort_unstable_by(|a, b| a.hash.cmp(&b.hash));
let item_count = goodbye_items.len();
let goodbye_table_size = (item_count + 1)*std::mem::size_of::<PxarGoodbyeItem>();
self.write_header(PXAR_GOODBYE, goodbye_table_size as u64)?;
if self.file_copy_buffer.len() < goodbye_table_size {
let need = goodbye_table_size - self.file_copy_buffer.len();
self.file_copy_buffer.reserve(need);
unsafe { self.file_copy_buffer.set_len(self.file_copy_buffer.capacity()); }
}
let buffer = &mut self.file_copy_buffer;
copy_binary_search_tree(item_count, |s, d| {
let item = &goodbye_items[s];
let offset = d*std::mem::size_of::<PxarGoodbyeItem>();
let dest = crate::tools::map_struct_mut::<PxarGoodbyeItem>(&mut buffer[offset..]).unwrap();
dest.offset = u64::to_le(item.offset);
dest.size = u64::to_le(item.size);
dest.hash = u64::to_le(item.hash);
});
// append PxarGoodbyeTail as last item
let offset = item_count*std::mem::size_of::<PxarGoodbyeItem>();
let dest = crate::tools::map_struct_mut::<PxarGoodbyeItem>(&mut buffer[offset..]).unwrap();
dest.offset = u64::to_le(goodbye_offset as u64);
dest.size = u64::to_le((goodbye_table_size + std::mem::size_of::<PxarHeader>()) as u64);
dest.hash = u64::to_le(PXAR_GOODBYE_TAIL_MARKER);
self.flush_copy_buffer(goodbye_table_size)?;
Ok(())
}
fn encode_dir(
&mut self,
dir: &mut nix::dir::Dir,
dir_stat: &FileStat,
magic: i64,
match_pattern: Vec<MatchPattern>,
) -> Result<(), Error> {
2018-12-27 12:15:10 +00:00
2019-03-15 11:19:51 +00:00
//println!("encode_dir: {:?} start {}", self.full_path(), self.writer_pos);
2018-12-27 12:15:10 +00:00
let mut name_list = vec![];
let rawfd = dir.as_raw_fd();
let dir_start_pos = self.writer_pos;
2018-12-27 12:15:10 +00:00
let mut dir_entry = self.create_entry(&dir_stat)?;
2018-12-28 09:44:12 +00:00
self.read_chattr(rawfd, &mut dir_entry)?;
self.read_fat_attr(rawfd, magic, &mut dir_entry)?;
// for each node in the directory tree, the filesystem features are
// checked based on the fs magic number.
self.fs_feature_flags = flags::feature_flags_from_magic(magic);
let (xattrs, fcaps) = self.read_xattrs(rawfd, &dir_stat)?;
let acl_access = self.read_acl(rawfd, &dir_stat, acl::ACL_TYPE_ACCESS)?;
let acl_default = self.read_acl(rawfd, &dir_stat, acl::ACL_TYPE_DEFAULT)?;
let projid = self.read_quota_project_id(rawfd, magic, &dir_stat)?;
2019-01-11 09:01:51 +00:00
self.write_entry(dir_entry)?;
for xattr in xattrs {
self.write_xattr(xattr)?;
}
self.write_fcaps(fcaps)?;
2018-12-28 10:48:47 +00:00
for user in acl_access.users {
self.write_acl_user(user)?;
}
for group in acl_access.groups {
self.write_acl_group(group)?;
}
if let Some(group_obj) = acl_access.group_obj {
self.write_acl_group_obj(group_obj)?;
}
for default_user in acl_default.users {
self.write_acl_default_user(default_user)?;
}
for default_group in acl_default.groups {
self.write_acl_default_group(default_group)?;
}
if let Some(default) = acl_default.default {
self.write_acl_default(default)?;
}
if let Some(projid) = projid {
self.write_quota_project_id(projid)?;
}
let include_children;
if is_virtual_file_system(magic) {
include_children = false;
} else {
if let Some(set) = &self.device_set {
include_children = set.contains(&dir_stat.st_dev);
} else {
include_children = true;
}
}
// Expand the exclude match pattern inherited from the parent by local entries, if present
let mut local_match_pattern = match_pattern.clone();
let pxar_exclude = match MatchPattern::from_file(rawfd, ".pxarexclude") {
Ok(Some((mut excludes, buffer, stat))) => {
local_match_pattern.append(&mut excludes);
Some((buffer, stat))
},
Ok(None) => None,
Err(err) => bail!("error while reading exclude file - {}", err),
};
if include_children {
for entry in dir.iter() {
let entry = entry.map_err(|err| {
format_err!("readir {:?} failed - {}", self.full_path(), err)
})?;
let filename = entry.file_name().to_owned();
2018-12-27 12:15:10 +00:00
let name = filename.to_bytes_with_nul();
if name == b".\0" || name == b"..\0" {
continue;
}
2018-12-27 12:15:10 +00:00
let stat = match nix::sys::stat::fstatat(rawfd, filename.as_ref(), nix::fcntl::AtFlags::AT_SYMLINK_NOFOLLOW) {
Ok(stat) => stat,
Err(nix::Error::Sys(Errno::ENOENT)) => {
let filename_osstr = std::ffi::OsStr::from_bytes(filename.to_bytes());
self.report_vanished_file(&self.full_path().join(filename_osstr))?;
continue;
},
Err(err) => bail!("fstat {:?} failed - {}", self.full_path(), err),
};
match match_filename(&filename, &stat, &local_match_pattern) {
(MatchType::Positive, _) => {
let filename_osstr = std::ffi::OsStr::from_bytes(filename.to_bytes());
eprintln!("matched by .pxarexclude entry - skipping: {:?}", self.full_path().join(filename_osstr));
},
(_, child_pattern) => name_list.push((filename, stat, child_pattern)),
}
if name_list.len() > MAX_DIRECTORY_ENTRIES {
bail!("too many directory items in {:?} (> {})", self.full_path(), MAX_DIRECTORY_ENTRIES);
}
}
} else {
2019-03-15 11:19:51 +00:00
eprintln!("skip mount point: {:?}", self.full_path());
2018-12-27 12:15:10 +00:00
}
name_list.sort_unstable_by(|a, b| a.0.cmp(&b.0));
2018-12-27 12:15:10 +00:00
let mut goodbye_items = vec![];
for (filename, stat, exclude_list) in name_list {
let start_pos = self.writer_pos;
if filename.as_bytes() == b".pxarexclude" {
if let Some((ref content, ref stat)) = pxar_exclude {
let filefd = match nix::fcntl::openat(rawfd, filename.as_ref(), OFlag::O_NOFOLLOW, Mode::empty()) {
Ok(filefd) => filefd,
Err(nix::Error::Sys(Errno::ENOENT)) => {
self.report_vanished_file(&self.full_path())?;
continue;
},
Err(err) => {
let filename_osstr = std::ffi::OsStr::from_bytes(filename.to_bytes());
bail!("open file {:?} failed - {}", self.full_path().join(filename_osstr), err);
},
};
2018-12-27 12:15:10 +00:00
let child_magic = if dir_stat.st_dev != stat.st_dev {
detect_fs_type(filefd)?
} else {
magic
};
self.write_filename(&filename)?;
self.encode_pxar_exclude(filefd, stat, child_magic, content)?;
continue;
}
}
self.relative_path.push(std::ffi::OsStr::from_bytes(filename.as_bytes()));
if self.verbose { println!("{:?}", self.full_path()); }
if is_directory(&stat) {
let mut dir = match nix::dir::Dir::openat(rawfd, filename.as_ref(), OFlag::O_DIRECTORY|OFlag::O_NOFOLLOW, Mode::empty()) {
Ok(dir) => dir,
Err(nix::Error::Sys(Errno::ENOENT)) => {
2019-03-15 11:19:51 +00:00
self.report_vanished_file(&self.full_path())?;
continue; // fixme!!
},
2019-03-15 11:19:51 +00:00
Err(err) => bail!("open dir {:?} failed - {}", self.full_path(), err),
};
let child_magic = if dir_stat.st_dev != stat.st_dev {
detect_fs_type(dir.as_raw_fd())?
} else {
magic
};
self.write_filename(&filename)?;
self.encode_dir(&mut dir, &stat, child_magic, exclude_list)?;
} else if is_reg_file(&stat) {
2019-03-15 11:19:51 +00:00
let mut hardlink_target = None;
2019-03-15 11:19:51 +00:00
if stat.st_nlink > 1 {
let link_info = HardLinkInfo { st_dev: stat.st_dev, st_ino: stat.st_ino };
hardlink_target = self.hardlinks.get(&link_info).map(|(v, offset)| {
let mut target = v.clone().into_os_string();
target.push("\0"); // add Nul byte
(target, (start_pos as u64) - offset)
});
if hardlink_target == None {
self.hardlinks.insert(link_info, (self.relative_path.clone(), start_pos as u64));
2019-03-15 11:19:51 +00:00
}
}
if let Some((target, offset)) = hardlink_target {
self.write_filename(&filename)?;
self.encode_hardlink(target.as_bytes(), offset)?;
} else {
let filefd = match nix::fcntl::openat(rawfd, filename.as_ref(), OFlag::O_NOFOLLOW, Mode::empty()) {
Ok(filefd) => filefd,
Err(nix::Error::Sys(Errno::ENOENT)) => {
self.report_vanished_file(&self.full_path())?;
continue;
},
Err(err) => bail!("open file {:?} failed - {}", self.full_path(), err),
};
let child_magic = if dir_stat.st_dev != stat.st_dev {
detect_fs_type(filefd)?
} else {
magic
};
self.write_filename(&filename)?;
let res = self.encode_file(filefd, &stat, child_magic);
let _ = nix::unistd::close(filefd); // ignore close errors
res?;
}
} else if is_symlink(&stat) {
let mut buffer = vec::undefined(libc::PATH_MAX as usize);
let res = filename.with_nix_path(|cstr| {
2018-12-28 13:51:43 +00:00
unsafe { libc::readlinkat(rawfd, cstr.as_ptr(), buffer.as_mut_ptr() as *mut libc::c_char, buffer.len()-1) }
})?;
match Errno::result(res) {
2018-12-28 13:51:43 +00:00
Ok(len) => {
buffer[len as usize] = 0u8; // add Nul byte
self.write_filename(&filename)?;
2018-12-28 13:51:43 +00:00
self.encode_symlink(&buffer[..((len+1) as usize)], &stat)?
}
Err(nix::Error::Sys(Errno::ENOENT)) => {
2019-03-15 11:19:51 +00:00
self.report_vanished_file(&self.full_path())?;
continue;
}
2019-03-15 11:19:51 +00:00
Err(err) => bail!("readlink {:?} failed - {}", self.full_path(), err),
}
} else if is_block_dev(&stat) || is_char_dev(&stat) {
if self.has_features(flags::WITH_DEVICE_NODES) {
self.write_filename(&filename)?;
self.encode_device(&stat)?;
} else {
eprintln!("skip device node: {:?}", self.full_path());
}
} else if is_fifo(&stat) {
if self.has_features(flags::WITH_FIFOS) {
self.write_filename(&filename)?;
self.encode_special(&stat)?;
} else {
eprintln!("skip fifo: {:?}", self.full_path());
}
} else if is_socket(&stat) {
if self.has_features(flags::WITH_SOCKETS) {
self.write_filename(&filename)?;
self.encode_special(&stat)?;
} else {
eprintln!("skip socket: {:?}", self.full_path());
}
} else {
2019-03-15 11:19:51 +00:00
bail!("unsupported file type (mode {:o} {:?})", stat.st_mode, self.full_path());
2018-12-27 12:15:10 +00:00
}
let end_pos = self.writer_pos;
goodbye_items.push(PxarGoodbyeItem {
offset: start_pos as u64,
size: (end_pos - start_pos) as u64,
2018-12-30 16:32:52 +00:00
hash: compute_goodbye_hash(filename.to_bytes()),
});
2019-03-15 11:19:51 +00:00
self.relative_path.pop();
2018-12-28 10:48:47 +00:00
}
2019-03-15 11:19:51 +00:00
//println!("encode_dir: {:?} end {}", self.full_path(), self.writer_pos);
// fixup goodby item offsets
let goodbye_start = self.writer_pos as u64;
for item in &mut goodbye_items {
item.offset = goodbye_start - item.offset;
}
let goodbye_offset = self.writer_pos - dir_start_pos;
self.write_goodbye_table(goodbye_offset, &mut goodbye_items)?;
2019-03-15 11:19:51 +00:00
//println!("encode_dir: {:?} end1 {}", self.full_path(), self.writer_pos);
Ok(())
}
fn encode_file(&mut self, filefd: RawFd, stat: &FileStat, magic: i64) -> Result<(), Error> {
2019-03-15 11:19:51 +00:00
//println!("encode_file: {:?}", self.full_path());
let mut entry = self.create_entry(&stat)?;
2018-12-28 09:44:12 +00:00
self.read_chattr(filefd, &mut entry)?;
self.read_fat_attr(filefd, magic, &mut entry)?;
let (xattrs, fcaps) = self.read_xattrs(filefd, &stat)?;
let acl_access = self.read_acl(filefd, &stat, acl::ACL_TYPE_ACCESS)?;
let projid = self.read_quota_project_id(filefd, magic, &stat)?;
2018-12-28 09:44:12 +00:00
2019-01-11 09:01:51 +00:00
self.write_entry(entry)?;
for xattr in xattrs {
self.write_xattr(xattr)?;
}
self.write_fcaps(fcaps)?;
for user in acl_access.users {
self.write_acl_user(user)?;
}
for group in acl_access.groups {
self.write_acl_group(group)?;
}
if let Some(group_obj) = acl_access.group_obj {
self.write_acl_group_obj(group_obj)?;
}
if let Some(projid) = projid {
self.write_quota_project_id(projid)?;
}
2018-12-28 09:44:12 +00:00
let include_payload;
if is_virtual_file_system(magic) {
include_payload = false;
} else {
if let Some(ref set) = &self.device_set {
include_payload = set.contains(&stat.st_dev);
} else {
include_payload = true;
}
}
if !include_payload {
2019-03-15 11:19:51 +00:00
eprintln!("skip content: {:?}", self.full_path());
self.write_header(PXAR_PAYLOAD, 0)?;
return Ok(());
}
2018-12-28 09:44:12 +00:00
let size = stat.st_size as u64;
self.write_header(PXAR_PAYLOAD, size)?;
2018-12-28 09:44:12 +00:00
let mut pos: u64 = 0;
loop {
let n = match nix::unistd::read(filefd, &mut self.file_copy_buffer) {
Ok(n) => n,
Err(nix::Error::Sys(Errno::EINTR)) => continue /* try again */,
2019-03-15 11:19:51 +00:00
Err(err) => bail!("read {:?} failed - {}", self.full_path(), err),
2018-12-28 09:44:12 +00:00
};
if n == 0 { // EOF
if pos != size {
// Note:: casync format cannot handle that
2019-03-15 11:19:51 +00:00
bail!("detected shrinked file {:?} ({} < {})", self.full_path(), pos, size);
2018-12-28 09:44:12 +00:00
}
break;
}
let mut next = pos + (n as u64);
if next > size { next = size; }
let count = (next - pos) as usize;
2018-12-28 10:48:47 +00:00
self.flush_copy_buffer(count)?;
2018-12-28 09:44:12 +00:00
pos = next;
2018-12-28 09:44:12 +00:00
if pos >= size { break; }
}
Ok(())
}
2019-01-11 11:22:00 +00:00
fn encode_device(&mut self, stat: &FileStat) -> Result<(), Error> {
let entry = self.create_entry(&stat)?;
2019-01-11 11:22:00 +00:00
self.write_entry(entry)?;
let major = unsafe { libc::major(stat.st_rdev) } as u64;
let minor = unsafe { libc::minor(stat.st_rdev) } as u64;
2019-03-15 11:19:51 +00:00
//println!("encode_device: {:?} {} {} {}", self.full_path(), stat.st_rdev, major, minor);
2019-01-11 11:22:00 +00:00
self.write_header(PXAR_DEVICE, std::mem::size_of::<PxarDevice>() as u64)?;
self.write_item(PxarDevice { major, minor })?;
2019-01-11 11:22:00 +00:00
Ok(())
}
// FIFO or Socket
fn encode_special(&mut self, stat: &FileStat) -> Result<(), Error> {
let entry = self.create_entry(&stat)?;
self.write_entry(entry)?;
Ok(())
}
fn encode_symlink(&mut self, target: &[u8], stat: &FileStat) -> Result<(), Error> {
2019-03-15 11:19:51 +00:00
//println!("encode_symlink: {:?} -> {:?}", self.full_path(), target);
2018-12-27 12:15:10 +00:00
let entry = self.create_entry(&stat)?;
2019-01-11 09:01:51 +00:00
self.write_entry(entry)?;
self.write_header(PXAR_SYMLINK, target.len() as u64)?;
2018-12-28 10:48:47 +00:00
self.write(target)?;
2018-12-27 12:15:10 +00:00
Ok(())
}
fn encode_hardlink(&mut self, target: &[u8], offset: u64) -> Result<(), Error> {
//println!("encode_hardlink: {:?} -> {:?}", self.full_path(), target);
// Note: HARDLINK replaces an ENTRY.
self.write_header(PXAR_FORMAT_HARDLINK, (target.len() as u64) + 8)?;
self.write_item(offset)?;
self.write(target)?;
Ok(())
}
fn encode_pxar_exclude(&mut self, filefd: RawFd, stat: &FileStat, magic: i64, content: &[u8]) -> Result<(), Error> {
let mut entry = self.create_entry(&stat)?;
self.read_chattr(filefd, &mut entry)?;
self.read_fat_attr(filefd, magic, &mut entry)?;
let (xattrs, fcaps) = self.read_xattrs(filefd, &stat)?;
let acl_access = self.read_acl(filefd, &stat, acl::ACL_TYPE_ACCESS)?;
let projid = self.read_quota_project_id(filefd, magic, &stat)?;
self.write_entry(entry)?;
for xattr in xattrs {
self.write_xattr(xattr)?;
}
self.write_fcaps(fcaps)?;
for user in acl_access.users {
self.write_acl_user(user)?;
}
for group in acl_access.groups {
self.write_acl_group(group)?;
}
if let Some(group_obj) = acl_access.group_obj {
self.write_acl_group_obj(group_obj)?;
}
if let Some(projid) = projid {
self.write_quota_project_id(projid)?;
}
let include_payload;
if is_virtual_file_system(magic) {
include_payload = false;
} else {
if let Some(set) = &self.device_set {
include_payload = set.contains(&stat.st_dev);
} else {
include_payload = true;
}
}
if !include_payload {
eprintln!("skip content: {:?}", self.full_path());
self.write_header(PXAR_PAYLOAD, 0)?;
return Ok(());
}
let size = content.len();
self.write_header(PXAR_PAYLOAD, size as u64)?;
self.writer.write_all(content)?;
self.writer_pos += size;
Ok(())
}
// the report_XXX method may raise and error - depending on encoder configuration
fn report_vanished_file(&self, path: &Path) -> Result<(), Error> {
eprintln!("WARNING: detected vanished file {:?}", path);
Ok(())
}
}
// If there is a match, an updated MatchPattern list to pass to the matched child is returned.
fn match_filename(
filename: &CStr,
stat: &FileStat,
match_pattern: &Vec<MatchPattern>
) -> (MatchType, Vec<MatchPattern>) {
let mut child_pattern = Vec::new();
let mut match_state = MatchType::None;
for pattern in match_pattern {
match pattern.matches_filename(filename, is_directory(&stat)) {
MatchType::None => {},
MatchType::Positive => match_state = MatchType::Positive,
MatchType::Negative => match_state = MatchType::Negative,
match_type => {
if match_state != MatchType::Positive && match_state != MatchType::Negative {
match_state = match_type;
}
child_pattern.push(pattern.get_rest_pattern());
},
}
}
(match_state, child_pattern)
}
fn errno_is_unsupported(errno: Errno) -> bool {
match errno {
Errno::ENOTTY | Errno::ENOSYS | Errno::EBADF | Errno::EOPNOTSUPP | Errno::EINVAL => {
true
}
_ => false,
}
}
fn detect_fs_type(fd: RawFd) -> Result<i64, Error> {
let mut fs_stat: libc::statfs = unsafe { std::mem::uninitialized() };
let res = unsafe { libc::fstatfs(fd, &mut fs_stat) };
Errno::result(res)?;
Ok(fs_stat.f_type)
}
#[inline(always)]
2019-01-30 17:25:37 +00:00
pub fn is_temporary_file_system(magic: i64) -> bool {
use fs::magic::*;
magic == RAMFS_MAGIC || magic == TMPFS_MAGIC
}
2019-01-30 17:25:37 +00:00
pub fn is_virtual_file_system(magic: i64) -> bool {
use fs::magic::*;
match magic {
BINFMTFS_MAGIC |
CGROUP2_SUPER_MAGIC |
CGROUP_SUPER_MAGIC |
CONFIGFS_MAGIC |
DEBUGFS_MAGIC |
DEVPTS_SUPER_MAGIC |
EFIVARFS_MAGIC |
FUSE_CTL_SUPER_MAGIC |
HUGETLBFS_MAGIC |
MQUEUE_MAGIC |
NFSD_MAGIC |
PROC_SUPER_MAGIC |
PSTOREFS_MAGIC |
RPCAUTH_GSSMAGIC |
SECURITYFS_MAGIC |
SELINUX_MAGIC |
SMACK_MAGIC |
SYSFS_MAGIC => true,
_ => false
}
}