libburnia/libburn
2
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Introduced alternative signal handling actions

Browse Source
This commit is contained in:
Thomas Schmitt
2010-03-05 09:08:16 +00:00
parent b4e5afd317
commit 3519b42c14
12 changed files with 395 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

93
libburn/libburn.h
View File

@ -736,17 +736,22 @@ void burn_finish(void);
/* ts A61002 */
/** Abort any running drive operation and finally call burn_finish().
You MUST calm down the busy drive if an aborting event occurs during a
/** Abort any running drive operation and eventually call burn_finish().
You MUST shut down the busy drives if an aborting event occurs during a
burn run. For that you may call this function either from your own signal
handling code or indirectly by activating the builtin signal handling:
handling code or indirectly by activating the built-in signal handling:
burn_set_signal_handling("my_app_name : ", NULL, 0);
Else you may eventually call burn_drive_cancel() on the active drive and
wait for it to assume state BURN_DRIVE_IDLE.
@param patience Maximum number of seconds to wait for drives to finish
Else you may eventually call burn_drive_cancel() on the active drives and
wait for them to assume state BURN_DRIVE_IDLE.
@param patience Maximum number of seconds to wait for drives to
finish.
@since 0.7.8 :
If this is -1, then only the cancel operations will
be performed and no burn_finish() will happen.
@param pacifier_func If not NULL: a function to produce appeasing messages.
See burn_abort_pacifier() for an example.
@param handle Opaque handle to be used with pacifier_func
@param handle Opaque handle to be used with pacifier_func
@return 1 ok, all went well
0 had to leave a drive in unclean state
<0 severe error, do no use libburn again
@ -2784,30 +2789,78 @@ int burn_set_messenger(void *messenger);
/* ts A61002 */
/* @since 0.2.6 */
/** The prototype of a handler function suitable for burn_set_abort_handling().
/** The prototype of a handler function suitable for burn_set_signal_handling()
Such a function has to return -2 if it does not want the process to
exit with value 1.
*/
typedef int (*burn_abort_handler_t)(void *handle, int signum, int flag);
/** Control builtin signal handling. See also burn_abort().
/** Control built-in signal handling. Either by setting an own handler or
by activating the built-in signal handler.
A function parameter handle of NULL activates the built-in abort handler.
Depending on mode it may cancel all drive operations, wait for all drives
to become idle, exit(1). It may also prepare function
burn_drive_get_status() for waiting and performing exit(1).
If text is not NULL then it is used as prefix for pacifier messages of
burn_abort_pacifier().
Before version 0.7.8 only action 0 was available. I.e. the built-in handler
waited for the drives to become idle and then performed exit(1) directly.
But FreeBSD 8.0 sometimes pauses the other threads until the signal handler
returns.
The new actions try to avoid this deadlock. It is advised to use action 3
burn_set_signal_handling(text, NULL, 48);
and call burn_is_aborting(0) frequently. If it replies 1, then call
burn_abort() and exit(1).
@param handle Opaque handle eventually pointing to an application
provided memory object
@param handler A function to be called on signals. It will get handle as
argument. It should finally call burn_abort(). See there.
@param mode : 0 call handler(handle, signum, 0) on nearly all signals
1 enable system default reaction on all signals
2 try to ignore nearly all signals
10 like mode 2 but handle SIGABRT like with mode 0
Arguments (text, NULL, 0) activate the builtin abort handler. It will
eventually call burn_abort() and then perform exit(1). If text is not NULL
then it is used as prefix for pacifier messages of burn_abort_pacifier().
argument. flag will be 0.
It should finally call burn_abort(). See there.
@param mode : bit0 - bit3:
Receiving signals:
0 Call handler(handle, signum, 0) on nearly all signals
1 Enable system default reaction on all signals
2 Try to ignore nearly all signals
10 like mode 2 but handle SIGABRT like with mode 0
bit4 - bit7: With handler == NULL :
Action of built-in handler. "control thread" is the one
which called burn_set_signal_handling().
All actions activate receive mode 2 to ignore further
signals.
0 Same as 1 (for pre-0.7.8 backward compatibility)
@since 0.7.8
1 Catch the control thread in abort handler, call
burn_abort(>0) and finally exit(1).
Does not always work with FreeBSD.
2 Call burn_abort(-1) and return from handler. When the
control thread calls burn_drive_get_status(), then do
burn_abort(>0) instead, and finally exit(1).
Does not always work with FreeBSD.
3 Call burn_abort(-1), return from handler. It is duty of
the application to detect a pending abort condition
by calling burn_is_aborting() and to wait for all
drives to become idle. E.g. by calling burn_abort(>0).
4 Like 3, but without calling burn_abort(-1). Only the
indicator of burn_is_aborting() gets set.
@since 0.2.6
*/
void burn_set_signal_handling(void *handle, burn_abort_handler_t handler,
int mode);
/* ts B00304 */
/* Inquire whether the built-in abort handler was triggered by a signal.
This has to be done to detect pending abort handling if signal handling
was set to the built-in handler and action was set to 2 or 3.
@param flag Bitfield for control purposes (unused yet, submit 0)
@return 0 = no abort was triggered
>0 = action that was triggered (action 0 is reported as 1)
@since 0.7.8
*/
int burn_is_aborting(int flag);
/* ts A70811 */
/** Write data in random access mode.
The drive must be grabbed successfully before calling this function which
@ -2955,5 +3008,11 @@ BURN_END_DECLS
#define Libburn_dummy_probe_write_modeS 1
/* ts B00225 */
/* Leave abort handler quickly and catch control thread in
burn_drive_get_status().
*/
#define Libburn_signal_handler_return_2 1
#endif /*LIBBURN_H*/