Re: Revamp sniffer; Was: [obish?sym;rfa:doc] Wire up vsyscall

[Andrew Cagney <cagney@gnu.org>]

[-- Attachment #1: Type: text/plain, Size: 1750 bytes --]

Ok,

I just figured out why amd64 GNU/Linux signal trampolines don't work 
correctly and the reason is scary!

After signal delivery (e.x., sigaltstack testcase) the stack looks like:

   sentinel: SENTINEL_FRAME
   catcher: NORMAL_FRAME, dwarf2 unwinder
   <sigtramp>: SIGTRAMP_FRAME, sigtramp unwinder
   thrower: NORMAL_FRAME, dwarf2 unwinder

while later after catcher has "finish"ed it looks like:

   sentinel: SENTINEL_FRAME
   __restore_rt: NORMAL_FRAME, dwarf2 unwinder
   <corrupt stack>

Notice how the first time round the trampoline was correctly identified 
but not the second.  What happened?

In the first case:

- linux jumps to catcher setting it's return address to <sigtramp>'s 
entry point
- gdb correctly identifies catcher as a callee or normal frame
- gdb goes to unwind catcher and starts sniffing <sigtramp>
- the dwarf2 sniffer, when sniffing <sigtramp>, asks for that frame's 
address-in-block, since the catcher is a normal frame, that translates 
to <catcher's-return-address>-1
- <catcher's-return-address>-1 translates into <sigtramp>-1
- since <sigtramp>-1 doesn't have any CFI (lucky eh?) dwarf2 dismisses 
the frame
- the sigtramp unwinder gets to take a wiff and identifies the frame

while in the second case (no catcher):

- the dwarf2 sniffer, when sniffing <sigtramp>, asks for that frame's 
address-in-block, since the next frame is a sentinel that translates 
into <catcher's-return-address>
- <catcher's-return-address> translates into <sigtramp>
- <sigtramp> does have CFI so dwarf2 claims the frame marking it 
"normal", outch!

So for the moment, I think the sigtramp frame should always be sniffed 
before the dwarf2 frame.  We can revisit the possability of modifing 
dwarf2 later, much later.

Andrew

[-- Attachment #2: Attached Message --]
[-- Type: message/rfc822, Size: 13206 bytes --]

[-- Attachment #2.1.1: Type: text/plain, Size: 5 bytes --]

ping

[-- Attachment #2.1.2: Attached Message --]
[-- Type: message/rfc822, Size: 9162 bytes --]

[-- Attachment #2.1.2.1.1: Type: text/plain, Size: 175 bytes --]

Mark,

Want do you want to do here?  Commit daniel's patch for the short term 
while I clean this up proper over the medium term (appending it to the 
vsyscall list)?

Andrew

[-- Attachment #2.1.2.1.2: Attached Message --]
[-- Type: message/rfc822, Size: 4991 bytes --]

From: Andrew Cagney <cagney@gnu.org>
To: Mark Kettenis <kettenis@chello.nl>
Cc: drow@false.org, gdb-patches@sources.redhat.com, roland@redhat.com
Subject: Revamp sniffer; Was: [obish?sym;rfa:doc] Wire up vsyscall
Date: Mon, 17 May 2004 13:14:48 -0400
Message-ID: <40A8F308.9040900@gnu.org>

> 2004-01-25  Daniel Jacobowitz  <drow@mvista.com>
> 
> 	* i386-tdep.c (i386_gdbarch_init): Check for signal frame first.
> 	* i386-linux-tdep.c (i386_linux_pc_in_sigtramp): Handle
> 	__kernel_sigreturn.
> 
[...]
> I'm somewhat reluctant to throw the CFI completely overboard.

Yes.

> How
> about extending the DWARF2 unwinder such that we can set the
> [SIGTRAMP_FRAME] attribute based on the function's name?

The frame attributes (tramp et.al.) in many cases are orthogonal to the 
unwind mechanism, but sometimes they are not).

At present the frame sniffer is focused solely on identifying and 
returning an unwinder.  The frame's type being identified as an 
unexpected sid effect.  I think this should be inverted vis:

frame_sniffer () ->
	- attributes
	- unwinder (or NULL, indicating sniff for unwinder)
	- base (or NULL, indicating sniff for base)
	- ...

so while the type is determined up front, other things are optional. 
That way here the tramp code would look like:

   if (name == _kernel_sigreturn)
     if (cfi at address)
       return { attribute=TRAMP; unwinder=cfi_unwinder; base=NULL };
     else
       return { attribute=TRAMP; unwinder=alt_unwinder; base=NULL };

this also opens the way for us doing:

   if (name == solib-trampoline)
     return { attribute=TRAMP|CALLEE; unwinder=NULL; base=NULL }

i.e., it's a trampoline (need to skip it) but let something more generic 
identify the unwinder.

Andrew