X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=scandir.cc;h=0dfdbdba58518dd7e7f91ec9cd375a8c86110731;hb=0a76f30b578a2ace26090a4b9bbd74b04124a20e;hp=30b8d4572aa6eab67a9796018f8ba8c59310e6a4;hpb=738a672b1a4a95eb24b785c9238594de99f8e314;p=cumulus.git diff --git a/scandir.cc b/scandir.cc index 30b8d45..0dfdbdb 100644 --- a/scandir.cc +++ b/scandir.cc @@ -6,6 +6,8 @@ #include #include #include +#include +#include #include #include #include @@ -20,6 +22,7 @@ #include #include "format.h" +#include "localdb.h" #include "store.h" #include "sha1.h" @@ -36,6 +39,10 @@ static char *block_buf; 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. */ std::ostringstream metadata_root; @@ -98,8 +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); @@ -108,7 +116,7 @@ void dumpfile(int fd, dictionary &file_info) if ((stat_buf.st_mode & S_IFMT) != S_IFREG) { fprintf(stderr, "file is no longer a regular file!\n"); - return; + return -1; } /* The index data consists of a sequence of pointers to the data blocks @@ -123,15 +131,25 @@ 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()); size += bytes; } @@ -164,6 +182,8 @@ void dumpfile(int fd, dictionary &file_info) segment_list.insert(i->get_ref().get_segment()); delete i; } + + return size; } void scanfile(const string& path) @@ -173,6 +193,7 @@ 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 @@ -187,12 +208,20 @@ 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) { @@ -248,10 +277,18 @@ 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'; @@ -318,7 +355,20 @@ int main(int argc, char *argv[]) tss = new TarSegmentStore(backup_dest); - string desc_filename = backup_dest + "/snapshot.lbs"; + string database_path = backup_dest + "/localdb.sqlite"; + db = new LocalDb; + db->Open(database_path.c_str()); + + /* 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()); try { @@ -337,16 +387,20 @@ int main(int argc, char *argv[]) root->checksum(); segment_list.insert(root->get_ref().get_segment()); - descriptor << "root: " << root->get_ref().to_string() << "\n\n"; + 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; - descriptor << "segments:\n"; + descriptor << "Segments:\n"; for (std::set::iterator i = segment_list.begin(); i != segment_list.end(); ++i) { descriptor << " " << *i << "\n"; } + db->Close(); + tss->sync(); delete tss;