X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=scandir.cc;h=f8f6127cb62be40d9ba88b401f88f688cd1b0018;hb=57890fff9773a368f241a25c89931a5d67fccb4e;hp=2570be5cfd9a6153a00049e58e26b911331af9da;hpb=161345e7d1927c2e80f2ba69c344af41193113f3;p=cumulus.git diff --git a/scandir.cc b/scandir.cc index 2570be5..f8f6127 100644 --- a/scandir.cc +++ b/scandir.cc @@ -6,6 +6,8 @@ #include #include #include +#include +#include #include #include #include @@ -15,10 +17,12 @@ #include #include #include +#include #include #include #include "format.h" +#include "localdb.h" #include "store.h" #include "sha1.h" @@ -30,10 +34,14 @@ using std::ostream; static TarSegmentStore *tss = NULL; /* Buffer for holding a single block of data read from a file. */ -static const int LBS_BLOCK_SIZE = 1024 * 1024; +static const size_t LBS_BLOCK_SIZE = 1024 * 1024; static char *block_buf; -static const int LBS_METADATA_BLOCK_SIZE = 65536; +static const size_t LBS_METADATA_BLOCK_SIZE = 65536; + +/* Local database, which tracks objects written in this and previous + * invocations to help in creating incremental snapshots. */ +LocalDb *db; /* Contents of the root object. This will contain a set of indirect links to * the metadata objects. */ @@ -97,8 +105,9 @@ size_t file_read(int fd, char *buf, size_t maxlen) } /* Read the contents of a file (specified by an open file descriptor) and copy - * the data to the store. */ -void dumpfile(int fd, dictionary &file_info) + * the data to the store. Returns the size of the file (number of bytes + * dumped), or -1 on error. */ +int64_t dumpfile(int fd, dictionary &file_info) { struct stat stat_buf; fstat(fd, &stat_buf); @@ -106,8 +115,8 @@ void dumpfile(int fd, dictionary &file_info) list object_list; if ((stat_buf.st_mode & S_IFMT) != S_IFREG) { - printf("file is no longer a regular file!\n"); - return; + fprintf(stderr, "file is no longer a regular file!\n"); + return -1; } /* The index data consists of a sequence of pointers to the data blocks @@ -122,15 +131,27 @@ void dumpfile(int fd, dictionary &file_info) hash.process(block_buf, bytes); - // tarstore processing - LbsObject *o = new LbsObject; - o->set_group("data"); - o->set_data(block_buf, bytes); - o->write(tss); - object_list.push_back(o->get_name()); - segment_list.insert(o->get_ref().get_segment()); - delete o; + // Either find a copy of this block in an already-existing segment, or + // index it so it can be re-used in the future + SHA1Checksum block_hash; + block_hash.process(block_buf, bytes); + string block_csum = block_hash.checksum_str(); + ObjectReference ref = db->FindObject(block_csum, bytes); + + // Store a copy of the object if one does not yet exist + if (ref.get_segment().size() == 0) { + LbsObject *o = new LbsObject; + o->set_group("data"); + o->set_data(block_buf, bytes); + o->write(tss); + ref = o->get_ref(); + db->StoreObject(ref, block_csum, bytes); + delete o; + } + object_list.push_back(ref.to_string()); + segment_list.insert(ref.get_segment()); + db->UseObject(ref); size += bytes; } @@ -163,6 +184,8 @@ void dumpfile(int fd, dictionary &file_info) segment_list.insert(i->get_ref().get_segment()); delete i; } + + return size; } void scanfile(const string& path) @@ -172,6 +195,7 @@ void scanfile(const string& path) struct stat stat_buf; char *buf; ssize_t len; + int64_t file_size; list refs; // Set to true if the item is a directory and we should recursively scan @@ -186,12 +210,20 @@ void scanfile(const string& path) metadata << "name: " << uri_encode(path) << "\n"; file_info["mode"] = encode_int(stat_buf.st_mode & 07777); - file_info["atime"] = encode_int(stat_buf.st_atime); - file_info["ctime"] = encode_int(stat_buf.st_ctime); file_info["mtime"] = encode_int(stat_buf.st_mtime); file_info["user"] = encode_int(stat_buf.st_uid); file_info["group"] = encode_int(stat_buf.st_gid); + struct passwd *pwd = getpwuid(stat_buf.st_uid); + if (pwd != NULL) { + file_info["user"] += " (" + uri_encode(pwd->pw_name) + ")"; + } + + struct group *grp = getgrgid(stat_buf.st_gid); + if (pwd != NULL) { + file_info["group"] += " (" + uri_encode(grp->gr_name) + ")"; + } + char inode_type; switch (stat_buf.st_mode & S_IFMT) { @@ -216,16 +248,14 @@ void scanfile(const string& path) buf = new char[stat_buf.st_size + 2]; len = readlink(path.c_str(), buf, stat_buf.st_size + 1); if (len < 0) { - printf("error reading symlink: %m\n"); + fprintf(stderr, "error reading symlink: %m\n"); } else if (len <= stat_buf.st_size) { buf[len] = '\0'; - printf(" contents=%s\n", buf); + file_info["contents"] = uri_encode(buf); } else if (len > stat_buf.st_size) { - printf("error reading symlink: name truncated\n"); + fprintf(stderr, "error reading symlink: name truncated\n"); } - file_info["contents"] = uri_encode(buf); - delete[] buf; break; case S_IFREG: @@ -249,10 +279,18 @@ void scanfile(const string& path) flags = fcntl(fd, F_GETFL); fcntl(fd, F_SETFL, flags & ~O_NONBLOCK); - file_info["size"] = encode_int(stat_buf.st_size); - dumpfile(fd, file_info); + file_size = dumpfile(fd, file_info); + file_info["size"] = encode_int(file_size); close(fd); + if (file_size < 0) + return; // error occurred; do not dump file + + if (file_size != stat_buf.st_size) { + fprintf(stderr, "Warning: Size of %s changed during reading\n", + path.c_str()); + } + break; case S_IFDIR: inode_type = 'd'; @@ -284,7 +322,7 @@ void scandir(const string& path) DIR *dir = opendir(path.c_str()); if (dir == NULL) { - printf("Error: %m\n"); + fprintf(stderr, "Error: %m\n"); return; } @@ -312,11 +350,31 @@ int main(int argc, char *argv[]) { block_buf = new char[LBS_BLOCK_SIZE]; - if (argc > 1) { - tss = new TarSegmentStore(argv[1]); - } else { - tss = new TarSegmentStore("."); - } + string backup_dest = "."; + + if (argc > 1) + backup_dest = argv[1]; + + tss = new TarSegmentStore(backup_dest); + + /* Write a backup descriptor file, which says which segments are needed and + * where to start to restore this snapshot. The filename is based on the + * current time. */ + time_t now; + struct tm time_buf; + char desc_buf[256]; + time(&now); + localtime_r(&now, &time_buf); + strftime(desc_buf, sizeof(desc_buf), "%Y%m%dT%H%M%S", &time_buf); + string desc_filename = backup_dest + "/" + desc_buf + ".lbs"; + std::ofstream descriptor(desc_filename.c_str()); + + /* Open the local database which tracks all objects that are stored + * remotely, for efficient incrementals. Provide it with the name of this + * snapshot. */ + string database_path = backup_dest + "/localdb.sqlite"; + db = new LocalDb; + db->Open(database_path.c_str(), desc_buf); try { scanfile("."); @@ -334,16 +392,20 @@ int main(int argc, char *argv[]) root->checksum(); segment_list.insert(root->get_ref().get_segment()); - string r = root->get_ref().to_string(); - printf("\nroot: %s\n\n", r.c_str()); + descriptor << "Root: " << root->get_ref().to_string() << "\n"; + strftime(desc_buf, sizeof(desc_buf), "%Y-%m-%d %H:%M:%S %z", &time_buf); + descriptor << "Date: " << desc_buf << "\n"; + delete root; - printf("segments:\n"); + descriptor << "Segments:\n"; for (std::set::iterator i = segment_list.begin(); i != segment_list.end(); ++i) { - printf(" %s\n", i->c_str()); + descriptor << " " << *i << "\n"; } + db->Close(); + tss->sync(); delete tss;