Changeset 434 for trunk/libs

Timestamp:

Feb 14, 2018, 3:41:31 PM (6 years ago)

Author:

alain

Message:

blap

File:

• trunk/libs/stdio.h (modified) (3 diffs)

trunk/libs/stdio.h

@@ -264 +264 @@
 
 /***************************************************************************************** 
- * This function implements the POSIX "kill" system call.
+ * This function implements the "kill" system call on the user side.
  * It register the signal defined by the <sig_id> argument in all thread descriptors 
  * of a target process identified by the <pid> argument. This is done in all clusters
  * containing threads for the target process.
+ * It can be executed by any thread running in any cluster, as this function uses
+ * remote access to traverse the list of process copies stored in the owner cluster,
+ * and the RPC_SIGNAL_RISE to signal the remote threads. 
+ * This function does nothing for (sig_id == 0). This can be used to check process pid.
+ * TODO : This first implementation supports only SIGKILL / SIGSTOP / SIGCONT values.
  ***************************************************************************************** 
  * @ pid      : target process identifier.
- * @ sig_id   : index defining the signal type (from 1 to 31). 
+ * @ sig_id   : index defining the signal type.
  * @ return 0 if success / returns -1 if failure.
  ****************************************************************************************/
@@ -277 +282 @@
 
 /***************************************************************************************** 
- * This function implements the POSIX "getpid" system call.
+ * This function implements the "getpid" system call on the user side.
  ***************************************************************************************** 
  * @ returns the process PID for the calling thread process.
@@ -284 +289 @@
 
 /***************************************************************************************** 
- * This function implement the POSIX "fork" system call.
+ * This function implement the "fork" system call on the user side.
  * The calling process descriptor (parent process), and the associated thread descriptor
- * are replicated in the same cluster as the calling thread, but the new process (child 
- * process) is registered in another target cluster, that is the new process owner. 
- * The child process and the associated main thread will be migrated to the target cluster
- * later, when the child process makes an "exec" or any other system call. 
+ * are replicated in a - likely - remote cluster, that becomes the new process owner. 
+ * The child process get a new PID is linked to the parent PID. The child process inherit
+ * from the parent process the memory image, and all open files (including the TXT).
+ * The child process becomes the TXT terminal owner.
  * The target cluster depends on the "fork_user" flag and "fork_cxy" variable that can be
  * stored in the calling thread descriptor by the specific fork_place() system call.
  * If not, the kernel function makes a query to the DQDT to select the target cluster. 
  ***************************************************************************************** 
- * @ returns child process PID if success / returns -1 if failure
+ * @ if success, returns child process PID to parent, and return O to child.
+ * @ if failure, returns -1 to parent / no child process is created.
  ****************************************************************************************/
 int fork();
 
 /***************************************************************************************** 
- * This function implement the "exec" system call.
- * It is executed in the client cluster, but the new process descriptor and main thread
- * must be created in a server cluster, that is generally another cluster. 
- * - if the server_cluster is the client cluster, call directly the process_make_exec()
- *   function to create a new process, and launch a new thread in local cluster.
- * - if the target_cluster is remote, call rpc_process_exec_client() to execute the
- *   process_make_exec() on the remote cluster.
- * In both case this function build an exec_info_t structure containing all informations
- * required to build the new process descriptor and the associated thread.
- * Finally, the calling process and thread are deleted.
+ * This function implement the "exec" system call on the user side.
+ * It creates, in the same cluster as the calling thread, a new process descriptor,
+ * and a new associated main thread descriptor, executing a new memory image defined 
+ * by the <filename> argument. This new process inherit from the old process the PID
+ * and the PPID, as well as all open files (including the TXT). 
+ * The old process descriptor, and all its threads are blocked, and marked for deletion.
+ * Therefore the exec syscall does not return to the calling thread in case of success.
+ * This function build an exec_info_t structure containing the new process arguments,
+ * as defined by the <arv> argument, and the new process environment variables, 
+ * as defined by the <envp>  argument.
+ * TODO : the <argv> and <envp> arguments are not supported yet (both must be NULL).
  ***************************************************************************************** 
  * @ filename : string pointer on .elf filename (virtual pointer in user space)
  * @ argv     : array of strings on process arguments (virtual pointers in user space)
  * @ envp     : array of strings on environment variables (virtual pointers in user space)
- * @ returns O if success / returns -1 if failure.
+ * @ does not return if success / returns -1 if failure.
  ****************************************************************************************/
 int exec( char  * filename,