public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] ggc: do not wipe out unrelated data via gt_ggc_rtab
@ 2023-09-28 19:55 Sergei Trofimovich
  2023-09-29  8:32 ` Richard Biener
  0 siblings, 1 reply; 2+ messages in thread
From: Sergei Trofimovich @ 2023-09-28 19:55 UTC (permalink / raw)
  To: gcc-patches, Andrew Pinski, Richard Biener, David Malcolm, Martin Jambor
  Cc: Sergei Trofimovich, Sergei Trofimovich

From: Sergei Trofimovich <siarheit@google.com>

There are 3 GC root tables:

   gt_ggc_rtab
   gt_ggc_deletable_rtab
   gt_pch_scalar_rtab

`deletable` and `scalar` tables are both simple: each element always
contains a pointer to the beginning of the object and it's size is the
full object.

`rtab` is different: it's `base` is a pointer in the middle of the
struct and `stride` points to the next GC pointer in the array.

Before the change there were 2 problems:

1. We memset()ed not just pointers but data around them.
2. We wen out of bounds of the last object described by gt_ggc_rtab
   and triggered bootstrap failures in profile and asan bootstraps.

After the change we handle only pointers themselves like the rest of
ggc-common.cc code.

gcc/
	PR/111505
	* ggc-common.cc (ggc_zero_out_root_pointers): New helper.
	* ggc-common.cc (ggc_common_finalize): Use helper instead of
	memset() to wipe out pointers.
---
 gcc/ggc-common.cc | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/gcc/ggc-common.cc b/gcc/ggc-common.cc
index 95803fa95a1..39e2581affd 100644
--- a/gcc/ggc-common.cc
+++ b/gcc/ggc-common.cc
@@ -75,6 +75,18 @@ ggc_mark_root_tab (const_ggc_root_tab_t rt)
       (*rt->cb) (*(void **) ((char *)rt->base + rt->stride * i));
 }
 
+/* Zero out all the roots in the table RT.  */
+
+static void
+ggc_zero_rtab_roots (const_ggc_root_tab_t rt)
+{
+  size_t i;
+
+  for ( ; rt->base != NULL; rt++)
+    for (i = 0; i < rt->nelt; i++)
+      (*(void **) ((char *)rt->base + rt->stride * i)) = (void*)0;
+}
+
 /* Iterate through all registered roots and mark each element.  */
 
 void
@@ -1307,8 +1319,7 @@ ggc_common_finalize ()
       memset (rti->base, 0, rti->stride * rti->nelt);
 
   for (rt = gt_ggc_rtab; *rt; rt++)
-    for (rti = *rt; rti->base != NULL; rti++)
-      memset (rti->base, 0, rti->stride * rti->nelt);
+    ggc_zero_rtab_roots (*rt);
 
   for (rt = gt_pch_scalar_rtab; *rt; rt++)
     for (rti = *rt; rti->base != NULL; rti++)
-- 
2.42.0


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] ggc: do not wipe out unrelated data via gt_ggc_rtab
  2023-09-28 19:55 [PATCH] ggc: do not wipe out unrelated data via gt_ggc_rtab Sergei Trofimovich
@ 2023-09-29  8:32 ` Richard Biener
  0 siblings, 0 replies; 2+ messages in thread
From: Richard Biener @ 2023-09-29  8:32 UTC (permalink / raw)
  To: Sergei Trofimovich
  Cc: gcc-patches, Andrew Pinski, David Malcolm, Martin Jambor,
	Sergei Trofimovich

On Thu, 28 Sep 2023, Sergei Trofimovich wrote:

> From: Sergei Trofimovich <siarheit@google.com>
> 
> There are 3 GC root tables:
> 
>    gt_ggc_rtab
>    gt_ggc_deletable_rtab
>    gt_pch_scalar_rtab
> 
> `deletable` and `scalar` tables are both simple: each element always
> contains a pointer to the beginning of the object and it's size is the
> full object.
> 
> `rtab` is different: it's `base` is a pointer in the middle of the
> struct and `stride` points to the next GC pointer in the array.
> 
> Before the change there were 2 problems:
> 
> 1. We memset()ed not just pointers but data around them.
> 2. We wen out of bounds of the last object described by gt_ggc_rtab
>    and triggered bootstrap failures in profile and asan bootstraps.
> 
> After the change we handle only pointers themselves like the rest of
> ggc-common.cc code.

OK/

Thanks,
Richard.

> gcc/
> 	PR/111505
> 	* ggc-common.cc (ggc_zero_out_root_pointers): New helper.
> 	* ggc-common.cc (ggc_common_finalize): Use helper instead of
> 	memset() to wipe out pointers.
> ---
>  gcc/ggc-common.cc | 15 +++++++++++++--
>  1 file changed, 13 insertions(+), 2 deletions(-)
> 
> diff --git a/gcc/ggc-common.cc b/gcc/ggc-common.cc
> index 95803fa95a1..39e2581affd 100644
> --- a/gcc/ggc-common.cc
> +++ b/gcc/ggc-common.cc
> @@ -75,6 +75,18 @@ ggc_mark_root_tab (const_ggc_root_tab_t rt)
>        (*rt->cb) (*(void **) ((char *)rt->base + rt->stride * i));
>  }
>  
> +/* Zero out all the roots in the table RT.  */
> +
> +static void
> +ggc_zero_rtab_roots (const_ggc_root_tab_t rt)
> +{
> +  size_t i;
> +
> +  for ( ; rt->base != NULL; rt++)
> +    for (i = 0; i < rt->nelt; i++)
> +      (*(void **) ((char *)rt->base + rt->stride * i)) = (void*)0;
> +}
> +
>  /* Iterate through all registered roots and mark each element.  */
>  
>  void
> @@ -1307,8 +1319,7 @@ ggc_common_finalize ()
>        memset (rti->base, 0, rti->stride * rti->nelt);
>  
>    for (rt = gt_ggc_rtab; *rt; rt++)
> -    for (rti = *rt; rti->base != NULL; rti++)
> -      memset (rti->base, 0, rti->stride * rti->nelt);
> +    ggc_zero_rtab_roots (*rt);
>  
>    for (rt = gt_pch_scalar_rtab; *rt; rt++)
>      for (rti = *rt; rti->base != NULL; rti++)
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2023-09-29  8:32 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-09-28 19:55 [PATCH] ggc: do not wipe out unrelated data via gt_ggc_rtab Sergei Trofimovich
2023-09-29  8:32 ` Richard Biener

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).