[PATCH 0/2] libsframe: fix some memory leaks

[PATCH 0/2] libsframe: fix some memory leaks

[Indu Bhagat]

Hello,

This patch set fixes some memory leaks in the libsframe and it's testsuite.

Testing notes:
- Reg tested natively on aarch64 and x86_64

Thanks,
Indu Bhagat (2):
  libsframe: fix a memory leak in sframe_decode
  libsframe: testsuite: fix memory leaks in testcases

 libsframe/sframe-impl.h                           | 15 +++++++++++----
 libsframe/sframe.c                                |  9 +++++++++
 .../testsuite/libsframe.decode/be-flipping.c      |  3 +++
 libsframe/testsuite/libsframe.decode/frecnt-1.c   |  3 +++
 libsframe/testsuite/libsframe.decode/frecnt-2.c   |  3 +++
 5 files changed, 29 insertions(+), 4 deletions(-)

-- 
2.37.2

[PATCH 1/2] libsframe: fix a memory leak in sframe_decode

[Indu Bhagat]

sframe_decode () needs to malloc a temporary buffer of the same size as
the input buffer (containing the SFrame section bytes) when endian
flipping is needed.  The decoder keeps the endian flipped contents in
this buffer for its usage.  This code is necessary when the target
endianneess is not the same as host endianness.

The malloc'd buffer needs to be kept track of, so that it can freed up in
sframe_decoder_free () later.

ChangeLog:

	* libsframe/sframe-impl.h (struct sframe_decoder_ctx): Add new
	member to keep track of the internally malloc'd buffer.
	* libsframe/sframe.c (sframe_decoder_free): Free it up.
	(sframe_decode): Update the reference to the buffer.
---
 libsframe/sframe-impl.h | 15 +++++++++++----
 libsframe/sframe.c      |  9 +++++++++
 2 files changed, 20 insertions(+), 4 deletions(-)

diff --git a/libsframe/sframe-impl.h b/libsframe/sframe-impl.h
index 0e61c977886..340d3b35042 100644
--- a/libsframe/sframe-impl.h
+++ b/libsframe/sframe-impl.h
@@ -32,10 +32,17 @@ extern "C"
 
 struct sframe_decoder_ctx
 {
-  sframe_header sfd_header;	      /* SFrame header.  */
-  uint32_t *sfd_funcdesc;	      /* SFrame function desc entries table.  */
-  void *sfd_fres;		      /* SFrame FRE table.  */
-  int sfd_fre_nbytes;		      /* Number of bytes needed for SFrame FREs.  */
+  /* SFrame header.  */
+  sframe_header sfd_header;
+  /* SFrame function desc entries table.  */
+  uint32_t *sfd_funcdesc;
+  /* SFrame FRE table.  */
+  void *sfd_fres;
+  /* Number of bytes needed for SFrame FREs.  */
+  int sfd_fre_nbytes;
+  /* Reference to the internally malloc'd buffer, if any, for endian flipping
+     the original input buffer before decoding.  */
+  void *sfd_buf;
 };
 
 struct sframe_encoder_ctx
diff --git a/libsframe/sframe.c b/libsframe/sframe.c
index b8fde2f04f8..e41c95b9944 100644
--- a/libsframe/sframe.c
+++ b/libsframe/sframe.c
@@ -548,6 +548,11 @@ sframe_decoder_free (sframe_decoder_ctx **decoder)
 	  free (dctx->sfd_fres);
 	  dctx->sfd_fres = NULL;
 	}
+      if (dctx->sfd_buf != NULL)
+	{
+	  free (dctx->sfd_buf);
+	  dctx->sfd_buf = NULL;
+	}
 
       free (*decoder);
       *decoder = NULL;
@@ -824,6 +829,10 @@ sframe_decode (const char *sf_buf, size_t sf_size, int *errp)
 	  return sframe_ret_set_errno (errp, SFRAME_ERR_BUF_INVAL);
 	}
       frame_buf = tempbuf;
+      /* This buffer is malloc'd when endian flipping the contents of the input
+	 buffer are needed.  Keep a reference to it so it can be free'd up
+	 later in sframe_decoder_free ().  */
+      dctx->sfd_buf = tempbuf;
     }
   else
     frame_buf = (char *)sf_buf;
-- 
2.37.2

[PATCH 2/2] libsframe: testsuite: fix memory leaks in testcases

[Indu Bhagat]

ChangeLog:

	* libsframe/testsuite/libsframe.decode/be-flipping.c: Free
	SFrame buffer.
	* libsframe/testsuite/libsframe.decode/frecnt-1.c: Likewise.
	* libsframe/testsuite/libsframe.decode/frecnt-2.c: Likewise.
---
 libsframe/testsuite/libsframe.decode/be-flipping.c | 3 +++
 libsframe/testsuite/libsframe.decode/frecnt-1.c    | 3 +++
 libsframe/testsuite/libsframe.decode/frecnt-2.c    | 3 +++
 3 files changed, 9 insertions(+)

diff --git a/libsframe/testsuite/libsframe.decode/be-flipping.c b/libsframe/testsuite/libsframe.decode/be-flipping.c
index d915f9bb278..378537e4e10 100644
--- a/libsframe/testsuite/libsframe.decode/be-flipping.c
+++ b/libsframe/testsuite/libsframe.decode/be-flipping.c
@@ -104,6 +104,9 @@ main (void)
   err = sframe_decoder_get_funcdesc (dctx, 0, &nfres, &fsize, &fstart, &finfo);
   TEST ("be-flipping: Decoder get FDE", err == 0);
   TEST ("be-flipping: Decoder FRE count", nfres == 5);
+
+  free (sf_buf);
+  sf_buf = NULL;
       
   sframe_decoder_free (&dctx);
   return 0;
diff --git a/libsframe/testsuite/libsframe.decode/frecnt-1.c b/libsframe/testsuite/libsframe.decode/frecnt-1.c
index 49861ed4f5c..bffa2ef37fb 100644
--- a/libsframe/testsuite/libsframe.decode/frecnt-1.c
+++ b/libsframe/testsuite/libsframe.decode/frecnt-1.c
@@ -89,6 +89,9 @@ main (void)
   TEST ("frecnt-1: Decoder get FDE", err == 0);
   TEST ("frecnt-1: Decoder FRE count", nfres == 4);
 
+  free (sf_buf);
+  sf_buf = NULL;
+
   sframe_decoder_free (&dctx);
   return 0;
 
diff --git a/libsframe/testsuite/libsframe.decode/frecnt-2.c b/libsframe/testsuite/libsframe.decode/frecnt-2.c
index 7c140d88c3a..cbb72fe5f08 100644
--- a/libsframe/testsuite/libsframe.decode/frecnt-2.c
+++ b/libsframe/testsuite/libsframe.decode/frecnt-2.c
@@ -94,6 +94,9 @@ main (void)
       TEST ("frecnt-2: Decode get FRE", nfres == 4);
     }
 
+  free (sf_buf);
+  sf_buf = NULL;
+
   sframe_decoder_free (&dctx);
   return 0;
 
-- 
2.37.2

Re: [PATCH 0/2] libsframe: fix some memory leaks

[Nick Clifton]

Hi Indu,

> This patch set fixes some memory leaks in the libsframe and it's testsuite.

Patch series approved.  Please apply.

Would you be interested in becoming an official maintainer for the libsframe
code ?  That way you can approve your own patches - as long as they are specific
to libsframe of course.

Cheers
   Nick

libsframe builder (Was: [PATCH 0/2] libsframe: fix some memory leaks)

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 517 bytes --]

Hi Indu,

On Thu, 2022-12-22 at 14:54 -0800, Indu Bhagat via Binutils wrote:
> This patch set fixes some memory leaks in the libsframe and it's
> testsuite.

I see you are an enthusiastic user of the binutils-try buildbot. But
changes under libsframe/ don't trigger new builds and the libsframe
testsuite isn't actually ran.

The following tweak to the buildbot master.cfg should fix that.

Please let buildbot@sourceware.org know if you need more tweaks to
better test the new code.

Cheers,

Mark

[-- Attachment #2: 0001-Add-libsframe-to-binutils_files.patch --]
[-- Type: text/x-patch, Size: 3031 bytes --]

From b58f0c5e82062f9c1160b72bba62a425c993fc79 Mon Sep 17 00:00:00 2001
From: Mark Wielaard <mark@klomp.org>
Date: Fri, 23 Dec 2022 14:44:53 +0100
Subject: [PATCH] Add libsframe to binutils_files

Because gdb and binutils share a git repo we only trigger builds
if they are part of gdb or binutils. libsframe is a new subdir
part of binutils.

Also add an explicit check-libsframe to binutils_step_checks.
---
 builder/master.cfg | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/builder/master.cfg b/builder/master.cfg
index 1e17aa3..8a66ae8 100644
--- a/builder/master.cfg
+++ b/builder/master.cfg
@@ -671,6 +671,7 @@ c['schedulers'].append(gccrust_bootstrap_scheduler)
 binutils_files = ["bfd/",
                   "binutils/", "gas/", "ld/",
                   "libctf/",
+                  "libsframe/",
                   "gold/", "elfcpp/",
                   "include/", "libiberty/", "opcodes/",
                   "configure", "Makefile.in"]
@@ -2106,14 +2107,17 @@ binutils_step_check = steps.Test(
         workdir='binutils-build',
         command=['make',
                  util.Interpolate('-j%(prop:ncpus)s'),
-                 'check-ld', 'check-gas', 'check-binutils'],
+                 'check-ld', 'check-gas', 'check-binutils',
+                 'check-libsframe'],
         name='make check',
         logfiles={ "ld.sum": "ld/ld.sum",
                    "ld.log": "ld/ld.log",
                    "gas.sum": "gas/testsuite/gas.sum",
                    "gas.log": "gas/testsuite/gas.log",
                    "binutils.sum": "binutils/binutils.sum",
-                   "binutils.log": "binutils/binutils.log" },
+                   "binutils.log": "binutils/binutils.log",
+                   "libsframe.sum": "libsframe/libsframe.sum",
+                   "libsframe.log": "libsframe/libsframe.log" },
         haltOnFailure=False, flunkOnFailure=True)
 binutils_steps_bunsen = bunsen_logfile_upload_cpio_steps(
         ["*.sum", "*.log"],
@@ -2123,7 +2127,8 @@ binutils_step_check_libctf = steps.Test(
         workdir='binutils-build',
         command=['make',
                  util.Interpolate('-j%(prop:ncpus)s'),
-                 'check-ld', 'check-gas', 'check-binutils', 'check-libctf'],
+                 'check-ld', 'check-gas', 'check-binutils', 'check-libctf',
+                 'check-libsframe'],
         name='make check',
         logfiles={ "ld.sum": "ld/ld.sum",
                    "ld.log": "ld/ld.log",
@@ -2131,6 +2136,8 @@ binutils_step_check_libctf = steps.Test(
                    "gas.log": "gas/testsuite/gas.log",
                    "binutils.sum": "binutils/binutils.sum",
                    "binutils.log": "binutils/binutils.log",
+                   "libsframe.sum": "libsframe/libsframe.sum",
+                   "libsframe.log": "libsframe/libsframe.log",
                    "libctf.sum": "libctf/libctf.sum",
                    "libctf.log": "libctf/libctf.log" },
         haltOnFailure=False, flunkOnFailure=True)
-- 
2.18.4

Re: libsframe builder (Was: [PATCH 0/2] libsframe: fix some memory leaks)

[Frank Ch. Eigler]

Hi -

>          command=['make',
>                   util.Interpolate('-j%(prop:ncpus)s'),
> -                 'check-ld', 'check-gas', 'check-binutils'],
> +                 'check-ld', 'check-gas', 'check-binutils',
> +                 'check-libsframe'],
>          name='make check',
>          logfiles={ "ld.sum": "ld/ld.sum",
>                     "ld.log": "ld/ld.log",
>                     "gas.sum": "gas/testsuite/gas.sum",
>                     "gas.log": "gas/testsuite/gas.log",
>                     "binutils.sum": "binutils/binutils.sum",
> -                   "binutils.log": "binutils/binutils.log" },
> +                   "binutils.log": "binutils/binutils.log",
> +                   "libsframe.sum": "libsframe/libsframe.sum",
> +                   "libsframe.log": "libsframe/libsframe.log" },
> [...]

(For the record, all of those .log/.sum files, and more, are pulled
into the bunsen database automagically, so listing them all here for
storage and presentation in the buildbot system is not really
necessary.)

- FChE

Re: libsframe builder (Was: [PATCH 0/2] libsframe: fix some memory leaks)

[Indu Bhagat]

On 12/23/22 06:22, Mark Wielaard wrote:
> Hi Indu,
> 
> On Thu, 2022-12-22 at 14:54 -0800, Indu Bhagat via Binutils wrote:
>> This patch set fixes some memory leaks in the libsframe and it's
>> testsuite.
> 
> I see you are an enthusiastic user of the binutils-try buildbot. But
> changes under libsframe/ don't trigger new builds and the libsframe
> testsuite isn't actually ran.
> 

heh, binutils-try buildbot's enthusiastic user I am, yes. I did wonder 
if there are enough resources for users to be doing this so frequently 
(with almost every patch set before commit), I hope its not a problem ?

Right, I noticed that libsframe/ testsuite is not being run, but 
binutils-try buildbot has still been useful to catch regressions to 
gas/ld by my code ...

Asking about adding libsframe/ testsuite was on my list, thanks for 
doing that.

Best Regards,
Indu

> The following tweak to the buildbot master.cfg should fix that.
> 
> Please let buildbot@sourceware.org know if you need more tweaks to
> better test the new code.
> 
> Cheers,
> 
> Mark

Re: [PATCH 0/2] libsframe: fix some memory leaks

[Indu Bhagat]

On 12/23/22 01:55, Nick Clifton wrote:
> Hi Indu,
> 
>> This patch set fixes some memory leaks in the libsframe and it's 
>> testsuite.
> 
> Patch series approved.  Please apply.
> 
> Would you be interested in becoming an official maintainer for the 
> libsframe
> code ?  That way you can approve your own patches - as long as they are 
> specific
> to libsframe of course.
> 

Yes, I am happy to do that.

Thanks for reviewing.

> Cheers
>    Nick
> 

Re: [PATCH 0/2] libsframe: fix some memory leaks

[Nick Clifton]

Hi Indu,

>> Would you be interested in becoming an official maintainer for the libsframe
>> code ?  That way you can approve your own patches - as long as they are specific
>> to libsframe of course.
> 
> Yes, I am happy to do that.

Brilliant - please create a patch to add yourself to the binutils/MAINTAINERS file
as the maintainer for libsframe.  And welcome to the wonderful world of maintainership!

Cheers
   Nick

Re: libsframe builder (Was: [PATCH 0/2] libsframe: fix some memory leaks)

[Mark Wielaard]

Hi Frank,

On Fri, 2022-12-23 at 09:41 -0500, Frank Ch. Eigler wrote:
> (For the record, all of those .log/.sum files, and more, are pulled
> into the bunsen database automagically, so listing them all here for
> storage and presentation in the buildbot system is not really
> necessary.)

You are of course right. But the bunsen link is not easily found. We
really need to tweak the web-interface and emails sent to prominently
feature the bunsen results link.

They can be found in the stdio log output of the "upload to bunsen"
step as https://builder.sourceware.org/testrun/<hashid>

Cheers,

Mark

Re: libsframe builder (Was: [PATCH 0/2] libsframe: fix some memory leaks)

[Mark Wielaard]

Hi Indu,

On Fri, 2022-12-23 at 07:24 -0800, Indu Bhagat wrote:
> heh, binutils-try buildbot's enthusiastic user I am, yes. I did
> wonder 
> if there are enough resources for users to be doing this so frequently 
> (with almost every patch set before commit), I hope its not a problem ?

It isn't a problem, a binutils build/check takes just a couple of
minutes and at least for the x86_64 and i386 builders there are more
than one worker. The most resource constrained one is the s390x worker,
but it is still able to catch up. Worst case you have to wait an hour
till your job is finally scheduled.

You can always check whether there are a lot of pending jobs at 
https://builder.sourceware.org/buildbot/#/pendingbuildrequests
(There are often a couple which take multiple hours but are only ran a
few times a day)

Cheers,

Mark