[PATCH] Only construct ucontext for SA_SIGINFO signal handlers

[PATCH] Only construct ucontext for SA_SIGINFO signal handlers

[Jon TURNEY]

	* exceptions.cc (call_signal_handler): Only bother to construct
	the ucontext for signal handlers with SA_SIGINFO set.

Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
---
 winsup/cygwin/ChangeLog     |  5 +++++
 winsup/cygwin/exceptions.cc | 46 ++++++++++++++++++++++++++-------------------
 2 files changed, 32 insertions(+), 19 deletions(-)

diff --git a/winsup/cygwin/ChangeLog b/winsup/cygwin/ChangeLog
index 3b0e111..0ddc795 100644
--- a/winsup/cygwin/ChangeLog
+++ b/winsup/cygwin/ChangeLog
@@ -1,5 +1,10 @@
 2015-04-02  Jon TURNEY  <jon.turney@dronecode.org.uk>
 
+	* exceptions.cc (call_signal_handler): Only bother to construct
+	the ucontext for signal handlers with SA_SIGINFO set.
+
+2015-04-02  Jon TURNEY  <jon.turney@dronecode.org.uk>
+
 	* include/cygwin/signal.h (struct __mcontext): 16-byte align.
 	* include/sys/ucontext.h (ucontext_t): Ditto.
 
diff --git a/winsup/cygwin/exceptions.cc b/winsup/cygwin/exceptions.cc
index 0d1f36d..bac550c 100644
--- a/winsup/cygwin/exceptions.cc
+++ b/winsup/cygwin/exceptions.cc
@@ -1490,33 +1490,41 @@ _cygtls::call_signal_handler ()
       siginfo_t thissi = infodata;
       void (*thisfunc) (int, siginfo_t *, void *) = func;
 
-      ucontext_t thiscontext;
-      thiscontext.uc_link = 0;
-      thiscontext.uc_flags = 0;
-      if (thissi.si_cyg)
-        memcpy (&thiscontext.uc_mcontext, ((cygwin_exception *)thissi.si_cyg)->context(), sizeof(CONTEXT));
-      else
-        RtlCaptureContext ((CONTEXT *)&thiscontext.uc_mcontext);
-        /* FIXME: Really this should be the context which the signal interrupted? */
-
-      /* FIXME: If/when sigaltstack is implemented, this will need to do
-         something more complicated */
-      thiscontext.uc_stack.ss_sp = NtCurrentTeb ()->Tib.StackBase;
-      thiscontext.uc_stack.ss_flags = 0;
-      if (!NtCurrentTeb ()->DeallocationStack)
-        thiscontext.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->Tib.StackLimit - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
-      else
-        thiscontext.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->DeallocationStack - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
+      ucontext_t context;
+      ucontext_t *thiscontext = NULL;
+
+      /* Only make a context for SA_SIGINFO handlers */
+      if (this_sa_flags & SA_SIGINFO)
+        {
+          context.uc_link = 0;
+          context.uc_flags = 0;
+          if (thissi.si_cyg)
+            memcpy (&context.uc_mcontext, ((cygwin_exception *)thissi.si_cyg)->context(), sizeof(CONTEXT));
+          else
+            RtlCaptureContext ((CONTEXT *)&context.uc_mcontext);
+            /* FIXME: Really this should be the context which the signal interrupted? */
+
+          /* FIXME: If/when sigaltstack is implemented, this will need to do
+             something more complicated */
+          context.uc_stack.ss_sp = NtCurrentTeb ()->Tib.StackBase;
+          context.uc_stack.ss_flags = 0;
+          if (!NtCurrentTeb ()->DeallocationStack)
+            context.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->Tib.StackLimit - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
+          else
+            context.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->DeallocationStack - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
+
+          thiscontext = &context;
+        }
 
       sigset_t this_oldmask = set_process_mask_delta ();
-      thiscontext.uc_sigmask = this_oldmask;
+      context.uc_sigmask = this_oldmask;
       int this_errno = saved_errno;
       reset_signal_arrived ();
       incyg = false;
       sig = 0;		/* Flag that we can accept another signal */
       unlock ();	/* unlock signal stack */
 
-      thisfunc (thissig, &thissi, &thiscontext);
+      thisfunc (thissig, &thissi, thiscontext);
       incyg = true;
 
       set_signal_mask (_my_tls.sigmask, this_oldmask);
-- 
2.1.4

Re: [PATCH] Only construct ucontext for SA_SIGINFO signal handlers

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 2133 bytes --]

On Apr  2 20:30, Jon TURNEY wrote:
> 	* exceptions.cc (call_signal_handler): Only bother to construct
> 	the ucontext for signal handlers with SA_SIGINFO set.

Looks good, except...

> +      ucontext_t context;
> +      ucontext_t *thiscontext = NULL;
> +
> +      /* Only make a context for SA_SIGINFO handlers */
> +      if (this_sa_flags & SA_SIGINFO)
> +        {
> +          context.uc_link = 0;
> +          context.uc_flags = 0;
> +          if (thissi.si_cyg)
> +            memcpy (&context.uc_mcontext, ((cygwin_exception *)thissi.si_cyg)->context(), sizeof(CONTEXT));
> +          else
> +            RtlCaptureContext ((CONTEXT *)&context.uc_mcontext);
> +            /* FIXME: Really this should be the context which the signal interrupted? */
> +
> +          /* FIXME: If/when sigaltstack is implemented, this will need to do
> +             something more complicated */
> +          context.uc_stack.ss_sp = NtCurrentTeb ()->Tib.StackBase;
> +          context.uc_stack.ss_flags = 0;
> +          if (!NtCurrentTeb ()->DeallocationStack)
> +            context.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->Tib.StackLimit - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
> +          else
> +            context.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->DeallocationStack - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
> +
> +          thiscontext = &context;
> +        }


>        sigset_t this_oldmask = set_process_mask_delta ();
> -      thiscontext.uc_sigmask = this_oldmask;
> +      context.uc_sigmask = this_oldmask;
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

This call to set_process_mask_delta() should occur before constructing
the context, so that filling in uc_sigmask can be moved into the above
`'if' branch.

On second thought, isn't this slightly wrong anyway?  Shouldn't that be

         context.uc_sigmask = _my_tls.sigmask;
	 context.uc_mcontext.oldmask = this_oldmask;

?


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH] Only construct ucontext for SA_SIGINFO signal handlers

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 2554 bytes --]

On Apr  3 13:18, Corinna Vinschen wrote:
> On Apr  2 20:30, Jon TURNEY wrote:
> > 	* exceptions.cc (call_signal_handler): Only bother to construct
> > 	the ucontext for signal handlers with SA_SIGINFO set.
> 
> Looks good, except...
> 
> > +      ucontext_t context;
> > +      ucontext_t *thiscontext = NULL;
> > +
> > +      /* Only make a context for SA_SIGINFO handlers */
> > +      if (this_sa_flags & SA_SIGINFO)
> > +        {
> > +          context.uc_link = 0;
> > +          context.uc_flags = 0;
> > +          if (thissi.si_cyg)
> > +            memcpy (&context.uc_mcontext, ((cygwin_exception *)thissi.si_cyg)->context(), sizeof(CONTEXT));
> > +          else
> > +            RtlCaptureContext ((CONTEXT *)&context.uc_mcontext);
> > +            /* FIXME: Really this should be the context which the signal interrupted? */
> > +
> > +          /* FIXME: If/when sigaltstack is implemented, this will need to do
> > +             something more complicated */
> > +          context.uc_stack.ss_sp = NtCurrentTeb ()->Tib.StackBase;
> > +          context.uc_stack.ss_flags = 0;
> > +          if (!NtCurrentTeb ()->DeallocationStack)
> > +            context.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->Tib.StackLimit - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
> > +          else
> > +            context.uc_stack.ss_size = (uintptr_t)NtCurrentTeb ()->DeallocationStack - (uintptr_t)NtCurrentTeb ()->Tib.StackBase;
> > +
> > +          thiscontext = &context;
> > +        }
> 
> 
> >        sigset_t this_oldmask = set_process_mask_delta ();
> > -      thiscontext.uc_sigmask = this_oldmask;
> > +      context.uc_sigmask = this_oldmask;
>          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> 
> This call to set_process_mask_delta() should occur before constructing
> the context, so that filling in uc_sigmask can be moved into the above
> `'if' branch.
> 
> On second thought, isn't this slightly wrong anyway?  Shouldn't that be
> 
>          context.uc_sigmask = _my_tls.sigmask;
> 	 context.uc_mcontext.oldmask = this_oldmask;

Oh, btw., what about cr2?  Right now, with the above code, it contains
a random value.  It should at least be zero'ed out.  Alternatively:

  context.uc_mcontext.cr2 = (thissi.si_signo == SIGSEGV
			     || thissi.si_signo == SIGBUS)
			    ? (uintptr_t) thissi.si_addr : 0;


Thanks,
Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH] Only construct ucontext for SA_SIGINFO signal handlers

[Jon TURNEY]

On 03/04/2015 13:17, Corinna Vinschen wrote:
> On Apr  3 13:18, Corinna Vinschen wrote:
>> On Apr  2 20:30, Jon TURNEY wrote:
>>
>>>         sigset_t this_oldmask = set_process_mask_delta ();
>>> -      thiscontext.uc_sigmask = this_oldmask;
>>> +      context.uc_sigmask = this_oldmask;
>>           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>
>> This call to set_process_mask_delta() should occur before constructing
>> the context, so that filling in uc_sigmask can be moved into the above
>> `'if' branch.

Ok, I will move it.

>> On second thought, isn't this slightly wrong anyway?  Shouldn't that be
>>
>>           context.uc_sigmask = _my_tls.sigmask;
>> 	 context.uc_mcontext.oldmask = this_oldmask;

As I wrote elsewhere:  You'll have to help me understand what the 
difference in meaning between ucontext_t.uc_sigmask and 
ucontext_t.uc_mcontext.oldmask is.

I don't see how the value of _my_tls.sigmask has any meaning at that 
point in the code.

> Oh, btw., what about cr2?  Right now, with the above code, it contains
> a random value.  It should at least be zero'ed out.  Alternatively:
>
>    context.uc_mcontext.cr2 = (thissi.si_signo == SIGSEGV
> 			     || thissi.si_signo == SIGBUS)
> 			    ? (uintptr_t) thissi.si_addr : 0;
>

Sure, but can we deal with that as a separate patch?

Re: [PATCH] Only construct ucontext for SA_SIGINFO signal handlers

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 2765 bytes --]

On Apr  3 13:51, Jon TURNEY wrote:
> On 03/04/2015 13:17, Corinna Vinschen wrote:
> >On Apr  3 13:18, Corinna Vinschen wrote:
> >>On Apr  2 20:30, Jon TURNEY wrote:
> >>
> >>>        sigset_t this_oldmask = set_process_mask_delta ();
> >>>-      thiscontext.uc_sigmask = this_oldmask;
> >>>+      context.uc_sigmask = this_oldmask;
> >>          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> >>
> >>This call to set_process_mask_delta() should occur before constructing
> >>the context, so that filling in uc_sigmask can be moved into the above
> >>`'if' branch.
> 
> Ok, I will move it.
> 
> >>On second thought, isn't this slightly wrong anyway?  Shouldn't that be
> >>
> >>          context.uc_sigmask = _my_tls.sigmask;
> >>	 context.uc_mcontext.oldmask = this_oldmask;
> 
> As I wrote elsewhere:  You'll have to help me understand what the difference
> in meaning between ucontext_t.uc_sigmask and ucontext_t.uc_mcontext.oldmask
> is.
> 
> I don't see how the value of _my_tls.sigmask has any meaning at that point
> in the code.

Ok, I had a look into the Linux source and searched the web, and here's
the problem.

One is that sigset_t on Linux is not just a 32 or 64 bit bitmask anymore,
but an array of ulong's used as a rather big sigmask.

OTOH, mcontext_t::oldmask is only the size of "unsigned long".  In fact,
as it turns out by inspecting the Linux kernel, oldmask is nothing else
than the first bits of uc_sigmask which fit into an unsigned long.  And
in the net I found that oldmask is just the old representation of
sigset_t, before the Linux kernel allowed more signals than fit into
a bitmask of unsigned long size.  In fact, it's only for backward compat,
but unused these days.

Given that, setting context.uc_sigmask to this_oldmask is apparently
the right thing to do.  For emulating backward compat (which we don't
need, but it also doesn't hurt), we could set oldmask to the same
value:

  context.uc_sigmask = context.uc_mcontext.oldmask = this_oldmask;

> >Oh, btw., what about cr2?  Right now, with the above code, it contains
> >a random value.  It should at least be zero'ed out.  Alternatively:
> >
> >   context.uc_mcontext.cr2 = (thissi.si_signo == SIGSEGV
> >			     || thissi.si_signo == SIGBUS)
> >			    ? (uintptr_t) thissi.si_addr : 0;
> >
> 
> Sure, but can we deal with that as a separate patch?

Yes, but you can just apply it as well.  cr2 is the address of a page
fault, so that's equivalent to the value in ExceptionInformation[1]
which, in turn, is stored in si_addr in exception::handle.


Thanks,
Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH] Only construct ucontext for SA_SIGINFO signal handlers

[Jon TURNEY]

On 03/04/2015 15:08, Corinna Vinschen wrote:
> On Apr  3 13:51, Jon TURNEY wrote:
>> On 03/04/2015 13:17, Corinna Vinschen wrote:
>>> On Apr  3 13:18, Corinna Vinschen wrote:
>>>> On Apr  2 20:30, Jon TURNEY wrote:
>>>>
>>>>>         sigset_t this_oldmask = set_process_mask_delta ();
>>>>> -      thiscontext.uc_sigmask = this_oldmask;
>>>>> +      context.uc_sigmask = this_oldmask;
>>>>           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>>>
>>>> This call to set_process_mask_delta() should occur before constructing
>>>> the context, so that filling in uc_sigmask can be moved into the above
>>>> `'if' branch.
>>
>> Ok, I will move it.
>>
>>>> On second thought, isn't this slightly wrong anyway?  Shouldn't that be
>>>>
>>>>           context.uc_sigmask = _my_tls.sigmask;
>>>> 	 context.uc_mcontext.oldmask = this_oldmask;
>>
>> As I wrote elsewhere:  You'll have to help me understand what the difference
>> in meaning between ucontext_t.uc_sigmask and ucontext_t.uc_mcontext.oldmask
>> is.
>>
>> I don't see how the value of _my_tls.sigmask has any meaning at that point
>> in the code.
>
> Ok, I had a look into the Linux source and searched the web, and here's
> the problem.
>
> One is that sigset_t on Linux is not just a 32 or 64 bit bitmask anymore,
> but an array of ulong's used as a rather big sigmask.
>
> OTOH, mcontext_t::oldmask is only the size of "unsigned long".  In fact,
> as it turns out by inspecting the Linux kernel, oldmask is nothing else
> than the first bits of uc_sigmask which fit into an unsigned long.  And
> in the net I found that oldmask is just the old representation of
> sigset_t, before the Linux kernel allowed more signals than fit into
> a bitmask of unsigned long size.  In fact, it's only for backward compat,
> but unused these days.
>
> Given that, setting context.uc_sigmask to this_oldmask is apparently
> the right thing to do.  For emulating backward compat (which we don't
> need, but it also doesn't hurt), we could set oldmask to the same
> value:
>
>    context.uc_sigmask = context.uc_mcontext.oldmask = this_oldmask;
>

Thank you very much for researching this.  I tried but wasn't able to 
discover anything much.

What you suggest seems right, so I'll make an updated patch including that.