Recursive protection to _take_dump() (#412)

* added recursive check for DRIVE_DUMP error

* Test commit for multithreading with pthread

* Added join() to the pthread and change lock() for unlock()

* Removed the destroy() functions for the mutex and setted the mutex to be global

* Accepted new error messaages

* Added recusivity error message

* Fixed identation issues

* Removed the unnecessary thead safety

* added mutex for thread protection

* test commit for mutex

* test commit for mutex

* test commit for mutex

* test commit for mutex

* test commit for mutex

* test commit for mutex

* test commit for mutex

* test commit for mutex

* test commit for mutex

* mutex added with function to start it

* mutex added with function to start it

* Added recursive counter to sg_data structure and removed the mutex

* A sum operator is changed to a binary operator OR _parse_logPage

* Changed datatype from uint16_t to uint8_t from cast

* Wrong variable call fixed.

* Added null check to recursive_counter

* Implemented some optimizations to _parse_logPage and changed _take_dump to work with priv->recursive_counter

* Identation fixed.

* Changed the conditions flow to the original version

* changed datatype to recursive_counter

---------

Co-authored-by: moisesFigueroa <figueroa.moises@ibm.com>

Author: 2 people

messages/tape_linux_sg/root.txt

@@ -100,6 +100,7 @@ root:table {
 		30258W:string { "Cannot retrieve drive dump: failed to read buffer (%d)." }
 		30259W:string { "Cannot retrieve drive dump: failed to write to dump file (%d)." }
 		30260W:string { "Cannot retrieve drive dump: wrote %d bytes out, expected %d." }
+		30261W:string { "Cannot retrieve drive dump: failed to write to communicate with drive. Tried (%d) times." }
 		30261I:string { "Taking drive dump in buffer." }
 		30262I:string { "Forcing drive dump." }
 		30263I:string { "%s returns %s (%d) %s." }
@@ -136,6 +137,7 @@ root:table {
 		30294I:string { "Setting up timeout values from %s." }
 		30295I:string { "Have unstable TUR response, start over (Cur = %d, Prev = %d)." }
 		30296I:string { "Capturing a stable TUR at line %d." }
+		30297W:string { "Cannot retrieve drive dump: failed to communicate with drive. Tried (%d) times." }
 
 		30392D:string { "Backend %s %s." }
 		30393D:string { "Backend %s: %d %s." }

src/tape_drivers/linux/sg/sg_tape.c

@@ -103,6 +103,8 @@ struct sg_global_data global_data;
 #define TU_DEFAULT_TIMEOUT (60)
 #define MAX_RETRY          (100)
 
+#define MAX_TAKE_DUMP_ATTEMPTS (10)
+
 /* Forward references (For keep function order to struct tape_ops) */
 int sg_readpos(void *device, struct tc_position *pos);
 int sg_locate(void *device, struct tc_position dest, struct tc_position *pos);
@@ -118,19 +120,17 @@ static inline int _parse_logPage(const unsigned char *logdata,
 								 const uint16_t param, uint32_t *param_size,
 								 unsigned char *buf, const size_t bufsize)
 {
-	uint16_t page_len, param_code, param_len;
-	uint32_t i;
+	const uint16_t page_len = ((uint16_t)logdata[2] << 8) | (uint16_t)logdata[3];
+	uint16_t param_code, param_len;
+	uint32_t i = LOG_PAGE_HEADER_SIZE;
 	uint32_t ret = -EDEV_INTERNAL_ERROR;
 
-	page_len = ((uint16_t)logdata[2] << 8) + (uint16_t)logdata[3];
-	i = LOG_PAGE_HEADER_SIZE;
-
 	while(i < page_len)
 	{
-		param_code = ((uint16_t)logdata[i] << 8) + (uint16_t)logdata[i+1];
+		param_code = ((uint16_t)logdata[i] << 8) | (uint16_t)logdata[i+1];
 		param_len  = (uint16_t)logdata[i + LOG_PAGE_PARAMSIZE_OFFSET];
 
-		if(param_code == param)
+		if (param_code == param)
 		{
 			*param_size = param_len;
 			if(bufsize < param_len){
@@ -143,6 +143,7 @@ static inline int _parse_logPage(const unsigned char *logdata,
 				break;
 			}
 		}
+
 		i += param_len + LOG_PAGE_PARAM_OFFSET;
 	}
 
@@ -392,6 +393,13 @@ static int _take_dump(struct sg_data *priv, bool capture_unforced)
 	time_t    now;
 	struct tm *tm_now;
 
+	/* To check if the function became recursive */
+	if (priv->recursive_counter > MAX_TAKE_DUMP_ATTEMPTS) {
+		ltfsmsg(LTFS_WARN, 30297W, priv->recursive_counter);
+		return 0;
+	}
+	priv->recursive_counter++;
+
 	if (priv->vendor != VENDOR_IBM)
 		return 0;
 
@@ -426,6 +434,8 @@ static int _take_dump(struct sg_data *priv, bool capture_unforced)
 
 	ltfs_profiler_add_entry(priv->profiler, NULL, TAPEBEND_REQ_EXIT(REQ_TC_TAKEDUMPDRV));
 
+	priv->recursive_counter = 0;
+
 	return 0;
 }
 

src/tape_drivers/linux/sg/sg_tape.h

@@ -82,6 +82,7 @@ struct sg_data {
 	struct timeout_tape  *timeouts;            /**< Timeout table */
 	struct tc_drive_info info;                 /**< Drive information */
 	FILE*                profiler;             /**< The file pointer for profiler */
+	int                  recursive_counter;    /**< Recursive counter for take dump */
 };
 
 struct sg_global_data {