]> git.proxmox.com Git - proxmox-backup.git/blobdiff - src/pxar/encoder.rs
src/pxar/encoder.rs: allow to pass list of devices
[proxmox-backup.git] / src / pxar / encoder.rs
index c77acb6eadd1ff17edd8cf9e98686dffd19038f3..4ffe906420b86d3648b5f73163f2cfc0756d6380 100644 (file)
@@ -4,11 +4,12 @@
 
 use failure::*;
 use endian_trait::Endian;
-use std::collections::HashMap;
+use std::collections::{HashSet, HashMap};
 
 use super::format_definition::*;
 use super::binary_search_tree::*;
 use super::helper::*;
+use super::exclude_pattern::*;
 use crate::tools::fs;
 use crate::tools::acl;
 use crate::tools::xattr;
@@ -27,7 +28,7 @@ use nix::sys::stat::Mode;
 use nix::errno::Errno;
 use nix::sys::stat::FileStat;
 
-use crate::tools::vec;
+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
@@ -47,10 +48,12 @@ pub struct Encoder<'a, W: Write> {
     writer_pos: usize,
     _size: usize,
     file_copy_buffer: Vec<u8>,
-    all_file_systems: bool,
-    root_st_dev: u64,
+    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)>,
 }
 
@@ -61,11 +64,20 @@ impl <'a, W: Write> Encoder<'a, W> {
         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,
-        all_file_systems: bool,
+        device_set: Option<HashSet<u64>>,
         verbose: bool,
         feature_flags: u64,
     ) -> Result<(), Error> {
@@ -79,21 +91,26 @@ impl <'a, W: Write> Encoder<'a, W> {
         // todo: use scandirat??
 
         let dir_fd = dir.as_raw_fd();
-        let stat = match nix::sys::stat::fstat(dir_fd) {
-            Ok(stat) => stat,
-            Err(err) => bail!("fstat {:?} failed - {}", path, err),
-        };
+        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 = feature_flags_from_magic(magic);
+
         let mut me = Self {
             base_path: path,
             relative_path: PathBuf::new(),
@@ -101,16 +118,16 @@ impl <'a, W: Write> Encoder<'a, W> {
             writer_pos: 0,
             _size: 0,
             file_copy_buffer,
-            all_file_systems,
-            root_st_dev: stat.st_dev,
+            device_set,
             verbose,
             feature_flags,
+            fs_feature_flags,
             hardlinks: HashMap::new(),
         };
 
         if verbose { println!("{:?}", me.full_path()); }
 
-        me.encode_dir(dir, &stat, magic)?;
+        me.encode_dir(dir, &stat, magic, Vec::new())?;
 
         Ok(())
     }
@@ -223,12 +240,12 @@ impl <'a, W: Write> Encoder<'a, W> {
 
     /// 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 & feature_flags) == feature_flags
+        (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 & feature_flags) != 0
+        (self.feature_flags & self.fs_feature_flags & feature_flags) != 0
     }
 
     fn read_xattrs(&self, fd: RawFd, stat: &FileStat) -> Result<(Vec<CaFormatXAttr>, Option<CaFormatFCaps>), Error> {
@@ -554,7 +571,7 @@ impl <'a, W: Write> Encoder<'a, W> {
         Ok(())
     }
 
-    fn encode_dir(&mut self, dir: &mut nix::dir::Dir, dir_stat: &FileStat, magic: i64)  -> Result<(), Error> {
+    fn encode_dir(&mut self, dir: &mut nix::dir::Dir, dir_stat: &FileStat, magic: i64, match_pattern: Vec<PxarExcludePattern>)  -> Result<(), Error> {
 
         //println!("encode_dir: {:?} start {}", self.full_path(), self.writer_pos);
 
@@ -568,6 +585,11 @@ impl <'a, W: Write> Encoder<'a, W> {
 
         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 = 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)?;
@@ -602,23 +624,30 @@ impl <'a, W: Write> Encoder<'a, W> {
             self.write_quota_project_id(projid)?;
         }
 
-        let mut dir_count = 0;
-
         let include_children;
         if is_virtual_file_system(magic) {
             include_children = false;
         } else {
-            include_children = (self.root_st_dev == dir_stat.st_dev) || self.all_file_systems;
+            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 PxarExcludePattern::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() {
-                dir_count += 1;
-                if dir_count > MAX_DIRECTORY_ENTRIES {
-                    bail!("too many directory items in {:?} (> {})",
-                          self.full_path(), MAX_DIRECTORY_ENTRIES);
-                }
-
                 let entry =  entry.map_err(|err| {
                     format_err!("readir {:?} failed - {}", self.full_path(), err)
                 })?;
@@ -629,31 +658,68 @@ impl <'a, W: Write> Encoder<'a, W> {
                     continue;
                 }
 
-                name_list.push(filename);
+                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_exclude_pattern(&filename, &stat, &local_match_pattern) {
+                    (MatchType::Exclude, _) => {
+                        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 {
             eprintln!("skip mount point: {:?}", self.full_path());
         }
 
-        name_list.sort_unstable_by(|a, b| a.cmp(&b));
+        name_list.sort_unstable_by(|a, b| a.0.cmp(&b.0));
 
         let mut goodbye_items = vec![];
 
-        for filename in &name_list {
-            self.relative_path.push(std::ffi::OsStr::from_bytes(filename.as_bytes()));
+        for (filename, stat, exclude_list) in name_list {
+            let start_pos = self.writer_pos;
 
-            if self.verbose { println!("{:?}", self.full_path()); }
+            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);
+                        },
+                    };
+
+                    let child_magic = if dir_stat.st_dev != stat.st_dev {
+                        detect_fs_type(filefd)?
+                    } else {
+                        magic
+                    };
 
-            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)) => {
-                    self.report_vanished_file(&self.full_path())?;
+                    self.write_filename(&filename)?;
+                    self.encode_pxar_exclude(filefd, stat, child_magic, content)?;
                     continue;
                 }
-                Err(err) => bail!("fstat {:?} failed - {}", self.full_path(), err),
-            };
+            }
 
-            let start_pos = self.writer_pos;
+            self.relative_path.push(std::ffi::OsStr::from_bytes(filename.as_bytes()));
+
+            if self.verbose { println!("{:?}", self.full_path()); }
 
             if is_directory(&stat) {
 
@@ -673,7 +739,7 @@ impl <'a, W: Write> Encoder<'a, W> {
                 };
 
                 self.write_filename(&filename)?;
-                self.encode_dir(&mut dir, &stat, child_magic)?;
+                self.encode_dir(&mut dir, &stat, child_magic, exclude_list)?;
 
             } else if is_reg_file(&stat) {
 
@@ -809,7 +875,11 @@ impl <'a, W: Write> Encoder<'a, W> {
         if is_virtual_file_system(magic) {
             include_payload = false;
         } else {
-            include_payload = (stat.st_dev == self.root_st_dev) || self.all_file_systems;
+            if let Some(ref set) = &self.device_set {
+                include_payload = set.contains(&stat.st_dev);
+            } else {
+                include_payload = true;
+            }
         }
 
         if !include_payload {
@@ -905,6 +975,58 @@ impl <'a, W: Write> Encoder<'a, W> {
         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(CA_FORMAT_PAYLOAD, 0)?;
+            return Ok(());
+        }
+
+        let size = content.len();
+        self.write_header(CA_FORMAT_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> {
@@ -915,6 +1037,38 @@ impl <'a, W: Write> Encoder<'a, W> {
     }
 }
 
+// If there is a match, an updated PxarExcludePattern list to pass to the matched child is returned.
+fn match_exclude_pattern(
+    filename: &CStr,
+    stat: &FileStat,
+    match_pattern: &Vec<PxarExcludePattern>
+) ->  (MatchType, Vec<PxarExcludePattern>) {
+    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::Exclude =>  match_state = MatchType::Exclude,
+            MatchType::Include =>  match_state = MatchType::Include,
+            MatchType::PartialExclude =>  {
+                if match_state != MatchType::Exclude && match_state != MatchType::Include {
+                    match_state = MatchType::PartialExclude;
+                }
+                child_pattern.push(pattern.get_rest_pattern());
+            },
+            MatchType::PartialInclude =>  {
+                if match_state != MatchType::Exclude && match_state != MatchType::Include {
+                    match_state = MatchType::PartialInclude;
+                }
+                child_pattern.push(pattern.get_rest_pattern());
+            },
+        }
+    }
+
+    (match_state, child_pattern)
+}
+
 fn errno_is_unsupported(errno: Errno) -> bool {
 
     match errno {
@@ -933,35 +1087,6 @@ fn detect_fs_type(fd: RawFd) -> Result<i64, Error> {
     Ok(fs_stat.f_type)
 }
 
-
-// from /usr/include/linux/magic.h
-// and from casync util.h
-pub const BINFMTFS_MAGIC: i64 =        0x42494e4d;
-pub const CGROUP2_SUPER_MAGIC: i64 =   0x63677270;
-pub const CGROUP_SUPER_MAGIC: i64 =    0x0027e0eb;
-pub const CONFIGFS_MAGIC: i64 =        0x62656570;
-pub const DEBUGFS_MAGIC: i64 =         0x64626720;
-pub const DEVPTS_SUPER_MAGIC: i64 =    0x00001cd1;
-pub const EFIVARFS_MAGIC: i64 =        0xde5e81e4;
-pub const FUSE_CTL_SUPER_MAGIC: i64 =  0x65735543;
-pub const HUGETLBFS_MAGIC: i64 =       0x958458f6;
-pub const MQUEUE_MAGIC: i64 =          0x19800202;
-pub const NFSD_MAGIC: i64 =            0x6e667364;
-pub const PROC_SUPER_MAGIC: i64 =      0x00009fa0;
-pub const PSTOREFS_MAGIC: i64 =        0x6165676C;
-pub const RPCAUTH_GSSMAGIC: i64 =      0x67596969;
-pub const SECURITYFS_MAGIC: i64 =      0x73636673;
-pub const SELINUX_MAGIC: i64 =         0xf97cff8c;
-pub const SMACK_MAGIC: i64 =           0x43415d53;
-pub const RAMFS_MAGIC: i64 =           0x858458f6;
-pub const TMPFS_MAGIC: i64 =           0x01021994;
-pub const SYSFS_MAGIC: i64 =           0x62656572;
-pub const MSDOS_SUPER_MAGIC: i64 =     0x00004d44;
-pub const FUSE_SUPER_MAGIC: i64 =      0x65735546;
-pub const EXT4_SUPER_MAGIC: i64 =      0x0000EF53;
-pub const XFS_SUPER_MAGIC: i64 =       0x58465342;
-
-
 #[inline(always)]
 pub fn is_temporary_file_system(magic: i64) -> bool {
     magic == RAMFS_MAGIC || magic == TMPFS_MAGIC