Fix #53467: Phar cannot compress large archives

When Phars are flushed, a new temporary file is created for each entry
which should be compressed, and the `compressed_filesize` is retrieved.
Afterwards, the Phar manifest is written, and only after that the files
are copied to the actual Phar.  So for each such entry there is an open
temp file, what easily exceeds the limit.

Therefore, we use a single temporary file for all entries, and store
the start offset in the otherwise unused `header_offset` member.  We
ensure that the `cfp` members are properly set to NULL even if flushing
fails, to avoid use after free scenarios.

This solution is based on a suggestion by @lserni[1].

Closes GH-6643.

[1] <https://github.com/box-project/box2/issues/80#issuecomment-77147371>

2 files changed, 42 insertions, 17 deletions

diff --git a/NEWS b/NEWS
index 9ca7f5624b..0ef6168f56 100644
--- a/NEWS
+++ b/NEWS
@@ -9,6 +9,7 @@ PHP                                                                        NEWS
   . Fixed bug #75850 (Unclear error message wrt. __halt_compiler() w/o
     semicolon) (cmb)
   . Fixed bug #70091 (Phar does not mark UTF-8 filenames in ZIP archives). (cmb)
+  . Fixed bug #53467 (Phar cannot compress large archives). (cmb, lserni)
 
 - Standard:
   . Fixed bug #80654 (file_get_contents() maxlen fails above (2**31)-1 bytes).
diff --git a/ext/phar/phar.c b/ext/phar/phar.c
index 2749283ff5..7cb1b06363 100644
--- a/ext/phar/phar.c
+++ b/ext/phar/phar.c
@@ -2508,6 +2508,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 	smart_str main_metadata_str = {0};
 	int free_user_stub, free_fp = 1, free_ufp = 1;
 	int manifest_hack = 0;
+	php_stream *shared_cfp = NULL;
 
 	if (phar->is_persistent) {
 		if (error) {
@@ -2788,10 +2789,13 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 			return EOF;
 		}
 
-		/* create new file that holds the compressed version */
+		/* create new file that holds the compressed versions */
 		/* work around inability to specify freedom in write and strictness
 		in read count */
-		entry->cfp = php_stream_fopen_tmpfile();
+		if (shared_cfp == NULL) {
+			shared_cfp = php_stream_fopen_tmpfile();
+		}
+		entry->cfp = shared_cfp;
 		if (!entry->cfp) {
 			if (error) {
 				spprintf(error, 0, "unable to create temporary file");
@@ -2800,8 +2804,11 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 				php_stream_close(oldfile);
 			}
 			php_stream_close(newfile);
-			return EOF;
+			goto cleanup;
 		}
+		/* for real phars, header_offset is unused; we misuse it here to store the offset in the temp file */
+		ZEND_ASSERT(entry->header_offset == 0);
+		entry->header_offset = php_stream_tell(entry->cfp);
 		php_stream_flush(file);
 		if (-1 == phar_seek_efp(entry, 0, SEEK_SET, 0, 0)) {
 			if (closeoldfile) {
@@ -2811,7 +2818,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 			if (error) {
 				spprintf(error, 0, "unable to seek to start of file \"%s\" while creating new phar \"%s\"", entry->filename, phar->fname);
 			}
-			return EOF;
+			goto cleanup;
 		}
 		php_stream_filter_append((&entry->cfp->writefilters), filter);
 		if (SUCCESS != php_stream_copy_to_stream_ex(file, entry->cfp, entry->uncompressed_filesize, NULL)) {
@@ -2822,15 +2829,14 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 			if (error) {
 				spprintf(error, 0, "unable to copy compressed file contents of file \"%s\" while creating new phar \"%s\"", entry->filename, phar->fname);
 			}
-			return EOF;
+			goto cleanup;
 		}
 		php_stream_filter_flush(filter, 1);
 		php_stream_flush(entry->cfp);
 		php_stream_filter_remove(filter, 1);
 		php_stream_seek(entry->cfp, 0, SEEK_END);
-		entry->compressed_filesize = (uint32_t) php_stream_tell(entry->cfp);
+		entry->compressed_filesize = ((uint32_t) php_stream_tell(entry->cfp)) - entry->header_offset;
 		/* generate crc on compressed file */
-		php_stream_rewind(entry->cfp);
 		entry->old_flags = entry->flags;
 		entry->is_modified = 1;
 		global_flags |= (entry->flags & PHAR_ENT_COMPRESSION_MASK);
@@ -2886,7 +2892,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 			spprintf(error, 0, "unable to write manifest header of new phar \"%s\"", phar->fname);
 		}
 
-		return EOF;
+		goto cleanup;
 	}
 
 	phar->alias_len = restore_alias_len;
@@ -2907,7 +2913,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 			spprintf(error, 0, "unable to write manifest meta-data of new phar \"%s\"", phar->fname);
 		}
 
-		return EOF;
+		goto cleanup;
 	}
 	smart_str_free(&main_metadata_str);
 
@@ -2942,7 +2948,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 					spprintf(error, 0, "unable to write filename of file \"%s\" to manifest of new phar \"%s\"", entry->filename, phar->fname);
 				}
 			}
-			return EOF;
+			goto cleanup;
 		}
 
 		/* set the manifest meta-data:
@@ -2975,7 +2981,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 				spprintf(error, 0, "unable to write temporary manifest of file \"%s\" to manifest of new phar \"%s\"", entry->filename, phar->fname);
 			}
 
-			return EOF;
+			goto cleanup;
 		}
 	} ZEND_HASH_FOREACH_END();
 	/* Hack - see bug #65028, add padding byte to the end of the manifest */
@@ -2991,7 +2997,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 				spprintf(error, 0, "unable to write manifest padding byte");
 			}
 
-			return EOF;
+			goto cleanup;
 		}
 	}
 
@@ -3004,7 +3010,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 
 		if (entry->cfp) {
 			file = entry->cfp;
-			php_stream_rewind(file);
+			php_stream_seek(file, entry->header_offset, SEEK_SET);
 		} else {
 			file = phar_get_efp(entry, 0);
 			if (-1 == phar_seek_efp(entry, 0, SEEK_SET, 0, 0)) {
@@ -3015,7 +3021,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 				if (error) {
 					spprintf(error, 0, "unable to seek to start of file \"%s\" while creating new phar \"%s\"", entry->filename, phar->fname);
 				}
-				return EOF;
+				goto cleanup;
 			}
 		}
 
@@ -3027,7 +3033,7 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 			if (error) {
 				spprintf(error, 0, "unable to seek to start of file \"%s\" while creating new phar \"%s\"", entry->filename, phar->fname);
 			}
-			return EOF;
+			goto cleanup;
 		}
 
 		/* this will have changed for all files that have either changed compression or been modified */
@@ -3044,14 +3050,14 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 				spprintf(error, 0, "unable to write contents of file \"%s\" to new phar \"%s\"", entry->filename, phar->fname);
 			}
 
-			return EOF;
+			goto cleanup;
 		}
 
 		entry->is_modified = 0;
 
 		if (entry->cfp) {
-			php_stream_close(entry->cfp);
 			entry->cfp = NULL;
+			entry->header_offset = 0;
 		}
 
 		if (entry->fp_type == PHAR_MOD) {
@@ -3067,6 +3073,11 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 		}
 	} ZEND_HASH_FOREACH_END();
 
+	if (shared_cfp != NULL) {
+		php_stream_close(shared_cfp);
+		shared_cfp = NULL;
+	}
+
 	/* append signature */
 	if (global_flags & PHAR_HDR_SIGNATURE) {
 		char sig_buf[4];
@@ -3197,6 +3208,19 @@ int phar_flush(phar_archive_data *phar, char *user_stub, zend_long len, int conv
 	}
 
 	return EOF;
+
+cleanup:
+	if (shared_cfp != NULL) {
+		php_stream_close(shared_cfp);
+	}
+	ZEND_HASH_FOREACH_PTR(&phar->manifest, entry) {
+		if (entry->cfp) {
+			entry->cfp = NULL;
+			entry->header_offset = 0;
+		}
+	} ZEND_HASH_FOREACH_END();
+
+	return EOF;
 }
 /* }}} */