diff options
Diffstat (limited to 'winsup/cygwin/cygtls.cc')
| -rw-r--r-- | winsup/cygwin/cygtls.cc | 69 |
1 files changed, 5 insertions, 64 deletions
diff --git a/winsup/cygwin/cygtls.cc b/winsup/cygwin/cygtls.cc index 979d96851..c59f0cbd0 100644 --- a/winsup/cygwin/cygtls.cc +++ b/winsup/cygwin/cygtls.cc @@ -1,6 +1,6 @@ /* cygtls.cc - Copyright 2003, 2004, 2005, 2006, 2007, 2008, 2009 Red Hat, Inc. + Copyright 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010 Red Hat, Inc. This software is a copyrighted work licensed under the terms of the Cygwin license. Please consult the file "CYGWIN_LICENSE" for @@ -15,6 +15,7 @@ details. */ #include "dtable.h" #include "cygheap.h" #include "sigproc.h" +#include "exception.h" class sentry { @@ -54,6 +55,9 @@ void _cygtls::call (DWORD (*func) (void *, void *), void *arg) { char buf[CYGTLS_PADSIZE]; + /* Initialize this thread's ability to respond to things like + SIGSEGV or SIGFPE. */ + exception protect; _my_tls.call2 (func, arg, buf); } @@ -89,9 +93,6 @@ _cygtls::init_thread (void *x, DWORD (*func) (void *, void *)) } local_clib._current_locale = "C"; locals.process_logmask = LOG_UPTO (LOG_DEBUG); - /* Initialize this thread's ability to respond to things like - SIGSEGV or SIGFPE. */ - init_exception_handler (); } thread_id = GetCurrentThreadId (); @@ -219,63 +220,3 @@ _cygtls::set_siginfo (sigpacket *pack) { infodata = pack->si; } - -/* Set up the exception handler for the current thread. The x86 uses segment - register fs, offset 0 to point to the current exception handler. */ - -extern exception_list *_except_list asm ("%fs:0"); - -void -_cygtls::init_exception_handler () -{ - /* Here in the distant past of 17-Jul-2009, we had an issue where Windows - 2008 became YA perplexed because the cygwin exception handler was added - at the start of the SEH while still being in the list further on. This - was because we added a loop by setting el.prev to _except_list here. - Since el is reused in this thread, and this function can be called - more than once when a dll is loaded, this is not a good thing. - - So, for now, until the next required tweak, we will just avoid adding the - cygwin exception handler if it is already on this list. This could present - a problem if some previous exception handler tries to do things that are - better left to Cygwin. I await the cygwin mailing list notification of - this event with bated breath. - (cgf 2009-07-17) - - A change in plans: In the not-so-distant past of 2010-02-23 it was - discovered that something was moving in ahead of cygwin's exception - handler so just detecting that the exception handler was loaded wasn't - good enough. I sort of anticipated this. So, the next step is to remove - the old exception handler from the list and add it to the beginning. - - The next step will probably be to call this function at various points - in cygwin (like from _cygtls::setup_fault maybe) to absolutely ensure that - we have control. For now, however, this seems good enough. - (cgf 2010-02-23) - */ - exception_list *e = _except_list; - if (e == &el) - return; - while (e && e != (exception_list *) -1) - if (e->prev != &el) - e = e->prev; - else - { - e->prev = el.prev; - break; - } - /* Apparently Windows stores some information about an exception and tries - to figure out if the SEH which returned 0 last time actually solved the - problem, or if the problem still persists (e.g. same exception at same - address). In this case Windows seems to decide that it can't trust - that SEH and calls the next handler in the chain instead. - - At one point this was a loop (el.prev = ⪙). This outsmarted the - above behaviour. Unfortunately this trick doesn't work anymore with - Windows 2008, which irremediably gets into an endless loop, taking 100% - CPU. That's why we reverted to a normal SEH chain and changed the way - the exception handler returns to the application. */ - el.handler = handle_exceptions; - el.prev = _except_list; - _except_list = ⪙ -} |