X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=scandir.cc;h=c2a093feead2c15b535b368c7cb5ededed62a37b;hb=7680a5bda8a87e1f5d02ea07ef767878e3bb143c;hp=3e004fdfc29aee5af00282504fc94dcdc38f04aa;hpb=3cb5ea93e29ba5baf6a5d1f170bbc003d7e2fe35;p=cumulus.git

diff --git a/scandir.cc b/scandir.cc
index 3e004fd..c2a093f 100644
--- a/scandir.cc
+++ b/scandir.cc
@@ -15,12 +15,20 @@
 #include <vector>
 
 #include "store.h"
+#include "sha1.h"
 
 using std::string;
 using std::vector;
 
+static SegmentStore *segment_store;
 static OutputStream *info_dump = NULL;
 
+static SegmentPartitioner *index_segment, *data_segment;
+
+/* Buffer for holding a single block of data read from a file. */
+static const int LBS_BLOCK_SIZE = 1024 * 1024;
+static char *block_buf;
+
 void scandir(const string& path);
 
 /* Converts time to microseconds since the epoch. */
@@ -29,34 +37,75 @@ int64_t encode_time(time_t time)
     return (int64_t)time * 1000000;
 }
 
-void dumpfile(int fd)
+/* Read data from a file descriptor and return the amount of data read.  A
+ * short read (less than the requested size) will only occur if end-of-file is
+ * hit. */
+size_t file_read(int fd, char *buf, size_t maxlen)
+{
+    size_t bytes_read = 0;
+
+    while (true) {
+        ssize_t res = read(fd, buf, maxlen);
+        if (res < 0) {
+            if (errno == EINTR)
+                continue;
+            throw IOException("file_read: error reading");
+        } else if (res == 0) {
+            break;
+        } else {
+            bytes_read += res;
+            buf += res;
+            maxlen -= res;
+        }
+    }
+
+    return bytes_read;
+}
+
+/* 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)
 {
     struct stat stat_buf;
     fstat(fd, &stat_buf);
     int64_t size = 0;
 
-    char buf[4096];
-
     if ((stat_buf.st_mode & S_IFMT) != S_IFREG) {
         printf("file is no longer a regular file!\n");
         return;
     }
 
+    /* The index data consists of a sequence of pointers to the data blocks
+     * that actually comprise the file data.  This level of indirection is used
+     * so that the same data block can be used in multiple files, or multiple
+     * versions of the same file. */
+    struct uuid segment_uuid;
+    int object_id;
+    OutputStream *index_data = index_segment->new_object(&segment_uuid,
+                                                         &object_id);
+
+    SHA1Checksum hash;
     while (true) {
-        ssize_t res = read(fd, buf, sizeof(buf));
-        if (res < 0) {
-            if (errno == EINTR)
-                continue;
-            printf("Error while reading: %m\n");
-            return;
-        } else if (res == 0) {
+        struct uuid block_segment_uuid;
+        int block_object_id;
+
+        size_t bytes = file_read(fd, block_buf, LBS_BLOCK_SIZE);
+        if (bytes == 0)
             break;
-        } else {
-            size += res;
-        }
+
+        hash.process(block_buf, bytes);
+        OutputStream *block = data_segment->new_object(&block_segment_uuid,
+                                                       &block_object_id);
+        block->write(block_buf, bytes);
+        index_data->write_uuid(block_segment_uuid);
+        index_data->write_u32(block_object_id);
+
+        size += bytes;
     }
 
-    printf("    bytes=%Ld\n", size);
+    file_info["sha1"] = string((const char *)hash.checksum(),
+                               hash.checksum_size());
+    file_info["data"] = encode_objref(segment_uuid, object_id);
 }
 
 void scanfile(const string& path)
@@ -67,6 +116,9 @@ void scanfile(const string& path)
     char *buf;
     ssize_t len;
 
+    // Set to true if the item is a directory and we should recursively scan
+    bool recurse = false;
+
     dictionary file_info;
 
     lstat(path.c_str(), &stat_buf);
@@ -137,14 +189,14 @@ void scanfile(const string& path)
         flags = fcntl(fd, F_GETFL);
         fcntl(fd, F_SETFL, flags & ~O_NONBLOCK);
 
-        //dumpfile(fd);
         file_info["size"] = encode_u64(stat_buf.st_size);
+        dumpfile(fd, file_info);
         close(fd);
 
         break;
     case S_IFDIR:
         inode_type = 'd';
-        scandir(path);
+        recurse = true;
         break;
 
     default:
@@ -156,6 +208,11 @@ void scanfile(const string& path)
 
     info_dump->write_string(path);
     info_dump->write_dictionary(file_info);
+
+    // If we hit a directory, now that we've written the directory itself,
+    // recursively scan the directory.
+    if (recurse)
+        scandir(path);
 }
 
 void scandir(const string& path)
@@ -189,14 +246,17 @@ void scandir(const string& path)
 
 int main(int argc, char *argv[])
 {
-    FILE *dump = fopen("fileinfo", "w");
-    if (dump == NULL) {
-        fprintf(stderr, "Cannot open fileinfo: %m\n");
-        return 1;
-    }
+    block_buf = new char[LBS_BLOCK_SIZE];
+
+    segment_store = new SegmentStore(".");
+    SegmentWriter *sw = segment_store->new_segment();
+    info_dump = sw->new_object(NULL);
 
-    FileOutputStream os(dump);
-    info_dump = &os;
+    index_segment = new SegmentPartitioner(segment_store);
+    data_segment = new SegmentPartitioner(segment_store);
+
+    string uuid = SegmentWriter::format_uuid(sw->get_uuid());
+    printf("Backup UUID: %s\n", uuid.c_str());
 
     try {
         scanfile(".");
@@ -204,5 +264,9 @@ int main(int argc, char *argv[])
         fprintf(stderr, "IOException: %s\n", e.getError().c_str());
     }
 
+    delete index_segment;
+    delete data_segment;
+    delete sw;
+
     return 0;
 }