[Bug optimization/15248] New: Reload may generate stores to read-only memory

[Bug optimization/15248] New: Reload may generate stores to read-only memory

[ebotcazou at gcc dot gnu dot org]

Under certain circumstances, the reload pass may generate stores to read-only
memory, thus producing executables that will segfault at run time.

The pattern is the following: a REG_EQUAL note in a set of a pseudo containing a
MEM/u is promoted by local-alloc to a REG_EQUIV note; during reload, the pseudo
doesn't get a hard reg so is replaced by its equivalent memory location as the
destination of the set.

This had affected the 3.3 branch (PR opt/13634, opt/13472, opt/14529) and the
3.4 branch (PR opt/15112), and was fixed by not recording MEM/u as equivalent
memory locations for pseudos altogether.

More sophisticated solutions can probably be devised so as to avoid the
unnecessary spills to the frame that are currently produced.

-- 
           Summary: Reload may generate stores to read-only memory
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P2
         Component: optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ebotcazou at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: *-*-*
  GCC host triplet: *-*-*
GCC target triplet: *-*-*


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug optimization/15248] Reload may generate stores to read-only memory

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-05-02 14:04 -------
Confirmed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-05-02 14:04:36
               date|                            |
   Target Milestone|---                         |3.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] Reload may generate stores to read-only memory

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-09-17 10:38 -------
Does this bug still exist, now that RTX_UNCHANGING is gone?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] Reload may generate stores to read-only memory

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-09-19 13:13 -------
> Does this bug still exist, now that RTX_UNCHANGING is gone?

I think so.  From the point of view of read-only memory, Richard's patch was
only a renaming, in particular in local-alloc.c, so I think the scenario I
described can still happen.  Note that we don't have testcases for mainline though.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] Reload may generate stores to read-only memory

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.0.0                       |---


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] Reload may generate stores to read-only memory

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-04-19 06:54 -------
The problem has resurfaced on mainline:
http://gcc.gnu.org/ml/gcc/2005-04/msg01052.html


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |law at redhat dot com
   Last reconfirmed|2004-08-01 00:24:37         |2005-04-19 06:54:12
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0/4.1 Regression] Reload may generate stores to read-only memory

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2005-04-19 10:03 -------
Eric, if you confirm that this bug is fixed in 3.3 and 3.4, then this is a 
(latent) regression on mainline and 4.0.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |3.4.4 3.3.5
            Summary|Reload may generate stores  |[4.0/4.1 Regression] Reload
                   |to read-only memory         |may generate stores to read-
                   |                            |only memory
   Target Milestone|---                         |4.0.1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0/4.1 Regression] Reload may generate stores to read-only memory

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-04-20 17:04 -------
> Eric, if you confirm that this bug is fixed in 3.3 and 3.4, then this is a 
> (latent) regression on mainline and 4.0.

It is indeed.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-04-30 22:10 -------
The same fix was applied to mainline.  Jeff, please consider backporting your
patch to 4.0 branch so that we be definitely done with this nasty problem.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.0/4.1 Regression] Reload |[4.0 Regression] Reload may
                   |may generate stores to read-|generate stores to read-only
                   |only memory                 |memory


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[law at redhat dot com]

------- Additional Comments From law at redhat dot com  2005-05-03 19:25 -------
Subject: Re:  [4.0 Regression] Reload may
	generate stores to read-only memory

On Sat, 2005-04-30 at 22:10 +0000, ebotcazou at gcc dot gnu dot org
wrote:
> ------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-04-30 22:10 -------
> The same fix was applied to mainline.  Jeff, please consider backporting your
> patch to 4.0 branch so that we be definitely done with this nasty problem.
No strong opinions here -- it's certainly a latent bug, but triggering
it is relatively difficult with 4.0.

Basically we need to evaluate the risk/reward.  We can fix this with the
exact same patch that we're using in mainline right now which hasn't
caused us any problems yet.  That patch is also very similar to what was
done for one or more of the 3.x release branches.  I consider the patch
relatively safe.

What's the reward -- a latent code generation bug is fixed.

So is the reward worth the small risk the patch poses?  

[ Note that if we had a bug report with 4.0 which was traceable to
  this problem, then I'd clearly land on the side of installing the
  patch. ]


Jeff



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[mark at codesourcery dot com]

------- Additional Comments From mark at codesourcery dot com  2005-05-03 19:35 -------
Subject: Re:  [4.0 Regression] Reload may	generate
 stores to read-only memory

Jeffrey A Law wrote:
> On Sat, 2005-04-30 at 22:10 +0000, ebotcazou at gcc dot gnu dot org
> wrote:
> 
>>------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-04-30 22:10 -------
>>The same fix was applied to mainline.  Jeff, please consider backporting your
>>patch to 4.0 branch so that we be definitely done with this nasty problem.
> 
> No strong opinions here -- it's certainly a latent bug, but triggering
> it is relatively difficult with 4.0.
> 
> Basically we need to evaluate the risk/reward.  We can fix this with the
> exact same patch that we're using in mainline right now which hasn't
> caused us any problems yet.  That patch is also very similar to what was
> done for one or more of the 3.x release branches.  I consider the patch
> relatively safe.
> 
> What's the reward -- a latent code generation bug is fixed.
> 
> So is the reward worth the small risk the patch poses?  
> 
> [ Note that if we had a bug report with 4.0 which was traceable to
>   this problem, then I'd clearly land on the side of installing the
>   patch. ]

I don't know that I'm in a position to make a good decision here, but I 
do agree with Jeff -- it would be a lot more compelling to make the 
change if there was actually a demonstrable problem -- especially coming 
from non-synthetic code.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-05-07 16:52 -------
> I don't know that I'm in a position to make a good decision here, but I 
> do agree with Jeff -- it would be a lot more compelling to make the 
> change if there was actually a demonstrable problem -- especially coming 
> from non-synthetic code.

I presume you'd like to have a testcase that fails with 4.0.x?  Because I can
hardly think of a more demonstrable problem than this one: Jeff's analysis for
4.1.x was exactly the same as mine for 3.3.x and nothing has changed in the code.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mark at codesourcery dot com
      Known to work|3.4.4 3.3.5                 |3.4.4 3.3.5 4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[mark at codesourcery dot com]

------- Additional Comments From mark at codesourcery dot com  2005-05-09 00:20 -------
Subject: Re:  [4.0 Regression] Reload may generate
 stores to read-only memory

ebotcazou at gcc dot gnu dot org wrote:
> ------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-05-07 16:52 -------
> 
>>I don't know that I'm in a position to make a good decision here, but I 
>>do agree with Jeff -- it would be a lot more compelling to make the 
>>change if there was actually a demonstrable problem -- especially coming 
>>from non-synthetic code.
> 
> 
> I presume you'd like to have a testcase that fails with 4.0.x?  Because I can
> hardly think of a more demonstrable problem than this one: Jeff's analysis for
> 4.1.x was exactly the same as mine for 3.3.x and nothing has changed in the code.

Yes, a 4.0 testcase was what I had in mind.  If you don't think it's 
easy to come up with one, but you are sure that people are likely to run 
into this bug, let me know.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[law at redhat dot com]

------- Additional Comments From law at redhat dot com  2005-05-17 17:03 -------
Subject: Re:  [4.0 Regression] Reload may
	generate stores to read-only memory

On Sat, 2005-05-07 at 16:52 +0000, ebotcazou at gcc dot gnu dot org
wrote:
> ------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-05-07 16:52 -------
> > I don't know that I'm in a position to make a good decision here, but I 
> > do agree with Jeff -- it would be a lot more compelling to make the 
> > change if there was actually a demonstrable problem -- especially coming 
> > from non-synthetic code.
> 
> I presume you'd like to have a testcase that fails with 4.0.x?  Because I can
> hardly think of a more demonstrable problem than this one: Jeff's analysis for
> 4.1.x was exactly the same as mine for 3.3.x and nothing has changed in the code.
It's still a relatively difficult bug to tickle.  It only appeared in
the mainline as a side effect of improvements to the jump threading code

Block duplication followed by redundancy elimination resulted in having
a pseudo-pseudo copy where the source pseudo was equivalent to a 
readonly mem and the RTL optimizers did not eliminate the pseudo-pseudo
copy.

Bernd has done some additional work on this problem and we may be
converging on a better solution than simply ignoring these problematical
equivalences.

Jeff



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-07-06 16:41 -------
Postponed until 4.0.2.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.0.1                       |4.0.2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[law at redhat dot com]

------- Additional Comments From law at redhat dot com  2005-07-06 16:47 -------
Subject: Re:  [4.0 Regression] Reload may
	generate stores to read-only memory

On Wed, 2005-07-06 at 16:41 +0000, mmitchel at gcc dot gnu dot org
wrote:
> ------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-07-06 16:41 -------
> Postponed until 4.0.2.
Probably wise.  I know Bernd had to go through a couple iterations on
the fixes.  Hopefully they've settled down enough that we can consider
the final fix for 4.0.2.

jeff




-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-07-07 10:19 -------
> Probably wise.  I know Bernd had to go through a couple iterations on
> the fixes.  Hopefully they've settled down enough that we can consider
> the final fix for 4.0.2.

Or we could simply put the "quick fix" on the 4.0 branch, like on 3.3 and 3.4.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[bonzini at gcc dot gnu dot org]

------- Additional Comments From bonzini at gcc dot gnu dot org  2005-08-18 09:45 -------
The "quick fix", for the record, is
http://gcc.gnu.org/ml/gcc-patches/2005-04/txt00083.txt

It makes sense to apply this to 4.0.2, and be done with this bug, since Jeff and
others decided on mainline that this problem is rare enough to care about the
pessimization.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-
                   |                            |patches/2005-04/txt00083.txt
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[mmitchel at gcc dot gnu dot org]

------- Additional Comments From mmitchel at gcc dot gnu dot org  2005-08-22 16:31 -------
Let's go with the quick fix, for 4.0.2, for now.  If we get a better fix, fine;
but let's not have this bug in 4.0.2.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[bonzini at gcc dot gnu dot org]

------- Additional Comments From bonzini at gcc dot gnu dot org  2005-08-29 14:41 -------
All branches now have the "quick fix".

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248

[Bug rtl-optimization/15248] [4.0 Regression] Reload may generate stores to read-only memory

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-08-29 14:58 -------
Subject: Bug 15248

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-4_0-branch
Changes by:	bonzini@gcc.gnu.org	2005-08-29 14:41:20

Modified files:
	gcc            : reload1.c ChangeLog 

Log message:
	2005-08-29  Paolo Bonzini  <bonzini@gnu.org>
	
	PR rtl-optimization/15248
	* reload1.c (reload): Ignore equivalences between pseudos and
	read only memory.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/reload1.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.461.6.3&r2=1.461.6.4
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=2.7592.2.400&r2=2.7592.2.401



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15248