com php-src: Fix off by 1 problem.: NEWS ext/mysqlnd/mysqlnd_ ps_codec.c ext/mysqlnd/mysqlnd_wireprotocol.c

From: Date: Mon, 12 Dec 2016 19:11:02 +0000
Subject: com php-src: Fix off by 1 problem.: NEWS ext/mysqlnd/mysqlnd_ ps_codec.c ext/mysqlnd/mysqlnd_wireprotocol.c
Groups: php.cvs 
Request: Send a blank email to php-cvs+get-96308@lists.php.net to get a copy of this message
Commit: e15c418c4c765d3a0c65a5ff15dcf37b1e556a66 Author: Andrey Hristov <andrey@php.net> Mon, 12 Dec 2016 21:11:02 +0200 Parents: 7a8774ade4383cd1c98d46e7e64f73f869a110e2 Branches: PHP-7.0 Link: http://git.php.net/?p=php-src.git;a=commitdiff;h=e15c418c4c765d3a0c65a5ff15dcf37b1e556a66 Log: Fix off by 1 problem. The problem was manifestated only with BIT columns and only when more than one row was fetched. The problem was coming from the fact that in pre-7.0 times mysqlnd was using a no-copy optimization. This optimization kept the strings (and also the BIT mask equivalents as strings) in the packet and the zval referred to them. 7.0+ zvals cannot use no-copy and always copy. Because of this the allocated memory for the packet was reduced by 1 by the person who ported the driver, but the starting address of the bit area wasn't reduced. Because of this the bit_area started at wrong address and the length decoded wrong. Changed paths: M NEWS M ext/mysqlnd/mysqlnd_ps_codec.c M ext/mysqlnd/mysqlnd_wireprotocol.c Diff: diff --git a/NEWS b/NEWS index fe09e4f..2285d5d 100644 --- a/NEWS +++ b/NEWS @@ -12,6 +12,10 @@ PHP NEWS . Fixed bug #73679 (DOTNET read access violation using invalid codepage). (Anatol) +- Mysqlnd: + . Fixed issue with decoding BIT columns when having more than one rows in the + result set. 7.0+ problem. (Andrey) + - PCRE: . Fixed bug #73612 (preg_*() may leak memory). (cmb) diff --git a/ext/mysqlnd/mysqlnd_ps_codec.c b/ext/mysqlnd/mysqlnd_ps_codec.c index e0b6c56..da24363 100644 --- a/ext/mysqlnd/mysqlnd_ps_codec.c +++ b/ext/mysqlnd/mysqlnd_ps_codec.c @@ -88,6 +88,7 @@ ps_fetch_from_1_to_8_bytes(zval * zv, const MYSQLND_FIELD * const field, unsigne } else { DBG_INF("stringify"); tmp_len = sprintf((char *)&tmp, MYSQLND_LLU_SPEC, uval); + DBG_INF_FMT("value=%s", tmp); } } } else { diff --git a/ext/mysqlnd/mysqlnd_wireprotocol.c b/ext/mysqlnd/mysqlnd_wireprotocol.c index 5871c3c..9f2aafab 100644 --- a/ext/mysqlnd/mysqlnd_wireprotocol.c +++ b/ext/mysqlnd/mysqlnd_wireprotocol.c @@ -1607,7 +1607,8 @@ php_mysqlnd_rowp_read_text_protocol_aux(MYSQLND_MEMORY_POOL_CHUNK * row_buffer, zval *current_field, *end_field, *start_field; zend_uchar * p = row_buffer->ptr; size_t data_size = row_buffer->app; - zend_uchar * bit_area = (zend_uchar*) row_buffer->ptr + data_size + 1; /* we allocate from here */ + /* we allocate from here. In pre-7.0 it was +1, as there was an additional \0 for the last string in the packet - because of the zval optimizations - using no-copy */ + zend_uchar * bit_area = (zend_uchar*) row_buffer->ptr + data_size; const zend_uchar * const packet_end = (zend_uchar*) row_buffer->ptr + data_size; DBG_ENTER("php_mysqlnd_rowp_read_text_protocol_aux"); @@ -1734,9 +1735,25 @@ php_mysqlnd_rowp_read_text_protocol_aux(MYSQLND_MEMORY_POOL_CHUNK * row_buffer, */ p -= len; if (Z_TYPE_P(current_field) == IS_LONG) { + /* + Andrey : See below. No need of bit_area, as we can use on stack for this. + The bit area should be removed - the prealloc_more_bytes in php_mysqlnd_read_row_ex() + + char tmp[22]; + const size_t tmp_len = sprintf((char *)&tmp, MYSQLND_LLU_SPEC, Z_LVAL_P(current_field)); + ZVAL_STRINGL(current_field, tmp, tmp_len); + */ bit_area += 1 + sprintf((char *)start, ZEND_LONG_FMT, Z_LVAL_P(current_field)); ZVAL_STRINGL(current_field, (char *) start, bit_area - start - 1); - } else if (Z_TYPE_P(current_field) == IS_STRING){ + } else if (Z_TYPE_P(current_field) == IS_STRING) { + /* + Andrey : This is totally sensless, but I am not gonna remove it in a production version. + This copies the data from the zval to the bit area. The destroys the original value + and creates the same one from the bit area. No need. It was making sense in pre-7.0 + when we used zval IS_STRING with no-copy that referred to the bit area. + The bit area has no sense in both the case of IS_LONG and IS_STRING as 7.0 zval + IS_STRING always copies. + */ memcpy(bit_area, Z_STRVAL_P(current_field), Z_STRLEN_P(current_field)); bit_area += Z_STRLEN_P(current_field); *bit_area++ = '\0'; @@ -1815,7 +1832,15 @@ php_mysqlnd_rowp_read(void * _packet, MYSQLND_CONN_DATA * conn) packet_type_to_statistic_packet_count[PROT_ROW_PACKET], 1); - /* packet->row_buffer->ptr is of size 'data_size + 1' */ + /* + packet->row_buffer->ptr is of size 'data_size' + in pre-7.0 it was really 'data_size + 1' although it was counted as 'data_size' + The +1 was for the additional byte needed to \0 terminate the last string in the row. + This was needed as the zvals of pre-7.0 could use external memory (no copy param to ZVAL_STRINGL). + However, in 7.0+ the strings always copy. Thus this +1 byte was removed. Also the optimization or \0 + terminating every string, which did overwrite the lengths from the packet. For this reason we needed + to keep (and copy) the lengths externally. + */ packet->header.size = data_size; packet->row_buffer->app = data_size;

« previous php.cvs (#96308) next »