1
0
mirror of https://git.FreeBSD.org/ports.git synced 2024-11-07 22:58:11 +00:00
freebsd-ports/lang/gnat/files/5fintman.adb
Steve Price 0b588ecd16 Upgrade to version 3.10p and do a little portlint cleaning.
PR:		5138
Submitted by:	Daniel M. Eischen <deischen@iworks.InterWorks.org>
1998-01-18 09:08:45 +00:00

236 lines
9.9 KiB
Ada

------------------------------------------------------------------------------
-- --
-- GNU ADA RUNTIME LIBRARY (GNARL) COMPONENTS --
-- --
-- S Y S T E M . I N T E R R U P T _ M A N A G E M E N T --
-- --
-- B o d y --
-- (Version for new GNARL) --
-- --
-- $Revision: 1.1 $ --
-- --
-- Copyright (C) 1991,1992,1993,1994,1995,1996 Florida State University --
-- --
-- GNARL is free software; you can redistribute it and/or modify it under --
-- terms of the GNU General Public License as published by the Free Soft- --
-- ware Foundation; either version 2, or (at your option) any later ver- --
-- sion. GNARL is distributed in the hope that it will be useful, but WITH- --
-- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY --
-- or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License --
-- for more details. You should have received a copy of the GNU General --
-- Public License distributed with GNARL; see file COPYING. If not, write --
-- to the Free Software Foundation, 59 Temple Place - Suite 330, Boston, --
-- MA 02111-1307, USA. --
-- --
-- As a special exception, if other files instantiate generics from this --
-- unit, or you link this unit with other files to produce an executable, --
-- this unit does not by itself cause the resulting executable to be --
-- covered by the GNU General Public License. This exception does not --
-- however invalidate any other reasons why the executable file might be --
-- covered by the GNU Public License. --
-- --
-- GNARL was developed by the GNARL team at Florida State University. It is --
-- now maintained by Ada Core Technologies Inc. in cooperation with Florida --
-- State University (http://www.gnat.com). --
-- --
------------------------------------------------------------------------------
-- This is the FreeBSD PTHREADS version of this package
-- This is only a first approximation.
-- It should be autogenerated by the m4 macro processor.
-- Contributed by Peter Burwood (gnat@arcangel.dircon.co.uk).
-- This file performs the system-dependent translation between machine
-- exceptions and the Ada exceptions, if any, that should be raised when
-- they occur. This version works for FreeBSD. Contributed by
-- Daniel M. Eischen (deischen@iworks.InterWorks.org).
-- PLEASE DO NOT add any dependences on other packages.
-- This package is designed to work with or without tasking support.
-- See the other warnings in the package specification before making
-- any modifications to this file.
-- Make a careful study of all signals available under the OS,
-- to see which need to be reserved, kept always unmasked,
-- or kept always unmasked.
-- Be on the lookout for special signals that
-- may be used by the thread library.
with Interfaces.C;
-- used for int and other types
with System.Error_Reporting;
-- used for Shutdown
with System.OS_Interface;
-- used for various Constants, Signal and types
package body System.Interrupt_Management is
use Interfaces.C;
use System.Error_Reporting;
use System.OS_Interface;
type Interrupt_List is array (Interrupt_ID range <>) of Interrupt_ID;
Exception_Interrupts : constant Interrupt_List :=
(SIGFPE, SIGILL, SIGSEGV, SIGBUS);
----------------------
-- Notify_Exception --
----------------------
-- This function identifies the Ada exception to be raised using
-- the information when the system received a synchronous signal.
-- Since this function is machine and OS dependent, different code
-- has to be provided for different target.
-- Language specs say signal handlers take exactly one arg, even
-- though FreeBSD actually supplies three. Ugh!
procedure Notify_Exception
(signo : Signal;
code : Interfaces.C.int;
context : access struct_sigcontext);
procedure Notify_Exception
(signo : Signal;
code : Interfaces.C.int;
context : access struct_sigcontext) is
begin
-- As long as we are using a longjmp to return control to the
-- exception handler on the runtime stack, we are safe. The original
-- signal mask (the one we had before coming into this signal catching
-- function) will be restored by the longjmp. Therefore, raising
-- an exception in this handler should be a safe operation.
-- Check that treatment of exception propagation here
-- is consistent with treatment of the abort signal in
-- System.Task_Primitives.Operations.
-- ?????
-- The code below is first approximation.
-- It would be nice to figure out more
-- precisely what exception has occurred.
-- One also should arrange to use an alternate stack for
-- recovery from stack overflow.
-- I don't understand the Linux kernel code well
-- enough to figure out how to do this yet.
-- I hope someone will look at this. --Ted Baker
-- How can SIGSEGV be split into constraint and storage errors ?
-- What should SIGILL really raise ? Some implemenations have
-- codes for different types of SIGILL and some raise Storage_Error.
-- What causes SIGBUS and should it be caught ?
-- Peter Burwood
case signo is
when SIGFPE =>
raise Constraint_Error;
when SIGILL =>
raise Constraint_Error;
when SIGSEGV =>
raise Storage_Error;
when SIGBUS =>
raise Storage_Error;
when others =>
pragma Assert (Shutdown ("Unexpected signal"));
null;
end case;
end Notify_Exception;
----------------
-- Initialize --
----------------
procedure Initialize is
act : aliased struct_sigaction;
old_act : aliased struct_sigaction;
mask : aliased sigset_t;
Result : Interfaces.C.int;
begin
Abort_Task_Interrupt := SIGABRT;
-- Change this if you want to use another signal for task abort.
-- SIGTERM might be a good one.
act.sa_handler := Notify_Exception'Address;
act.sa_flags := 16#010#;
-- Set sa_flags to SA_NODEFER so that during the handler execution
-- we do not change the Signal_Mask to be masked for the Signal.
-- This is a temporary fix to the problem that the Signal_Mask is
-- not restored after the exception (longjmp) from the handler.
-- The right fix should be made in sigsetjmp so that we save
-- the Signal_Set and restore it after a longjmp.
-- In that case, this field should be changed back to 0. ??? (Dong-Ik)
Result := sigemptyset (mask'Access);
pragma Assert (Result = 0
or else Shutdown ("GNULLI failure---sigemptyset"));
for I in Exception_Interrupts'Range loop
Result := sigaddset (mask'Access, Signal (Exception_Interrupts (I)));
pragma Assert (Result = 0
or else Shutdown ("GNULLI failure---sigaddset"));
end loop;
act.sa_mask := mask;
for I in Exception_Interrupts'Range loop
Keep_Unmasked (Exception_Interrupts (I)) := True;
Result :=
sigaction
(Signal (Exception_Interrupts (I)), act'Access,
old_act'Unchecked_Access);
pragma Assert (Result = 0
or else Shutdown ("GNULLI failure---sigaction"));
end loop;
Keep_Unmasked (Abort_Task_Interrupt) := true;
-- Keep_Unmasked (SIGBUS) := true;
Keep_Unmasked (SIGSTOP) := true;
Keep_Unmasked (SIGKILL) := true;
Keep_Unmasked (SIGINT) := true;
-- Keep_Unmasked (SIGEMT) := true;
-- Keep_Unmasked (SIGCHLD) := true;
-- Keep_Unmasked (SIGALRM) := true;
-- ???? The above signals have been found to need to be
-- kept unmasked on some systems, per Dong-Ik Oh.
-- I don't know whether the MIT/Provenzano threads
-- need these or any other signals unmasked at the thread level.
-- I hope somebody will take
-- the time to look it up. -- Ted Baker
-- FreeBSD uses SIGINFO to dump thread status to stdout. If
-- the user really wants to attach his own handler, let him.
-- FreeBSD pthreads uses setitimer/getitimer for thread scheduling.
-- It's not clear, but it looks as if it only needs SIGVTALRM
-- in order to handle the setitimer/getitimer operations. We
-- could probably allow SIGALARM, but we'll leave it as unmasked
-- for now. FreeBSD pthreads also needs SIGCHLD.
Keep_Unmasked (SIGCHLD) := true;
Keep_Unmasked (SIGALRM) := true;
Keep_Unmasked (SIGVTALRM) := true;
Reserve := Reserve or Keep_Unmasked or Keep_Masked;
Reserve (0) := true;
-- We do not have Signal 0 in reality. We just use this value
-- to identify non-existent signals (see s-intnam.ads). Therefore,
-- Signal 0 should not be used in all signal related operations hence
-- mark it as reserved.
end Initialize;
begin
Initialize;
end System.Interrupt_Management;