public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
* [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file
@ 2017-10-17 13:50 Djordje Todorovic
  2017-10-17 13:50 ` [PATCH 3/3] Add test for fetching TLS from " Djordje Todorovic
                   ` (2 more replies)
  0 siblings, 3 replies; 8+ messages in thread
From: Djordje Todorovic @ 2017-10-17 13:50 UTC (permalink / raw)
  To: macro
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica, Djordje Todorovic

On MIPS o32, n32 and n64 platforms information such as PID was not
correctly written into core file from GDB.

This fixes that behavior.

bfd/ChangeLog:

	* elf32-mips.c (elf32_mips_write_core_note): New function.
	(elf_backend_write_core_note): New macro.
	* elf64-mips.c (elf64_mips_write_core_note): New function.
	(elf_backend_write_core_note): New macro.
	* elfn32-mips.c (elf32_mips_write_core_note): New function.
	(elf_backend_write_core_note): New macro.
---
 bfd/elf32-mips.c  | 44 ++++++++++++++++++++++++++++++++++++++++++++
 bfd/elf64-mips.c  | 44 ++++++++++++++++++++++++++++++++++++++++++++
 bfd/elfn32-mips.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 132 insertions(+)

diff --git a/bfd/elf32-mips.c b/bfd/elf32-mips.c
index 17ca432..2ee15e9 100644
--- a/bfd/elf32-mips.c
+++ b/bfd/elf32-mips.c
@@ -2373,6 +2373,45 @@ elf32_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 
   return TRUE;
 }
+
+/* Write Linux core PRSTATUS note into core file.  */
+
+static char *
+elf32_mips_write_core_note (bfd *abfd, char *buf, int *bufsiz, int note_type,
+			      ...)
+{
+  switch (note_type)
+    {
+    default:
+      return NULL;
+
+    case NT_PRPSINFO:
+      BFD_FAIL ();
+      return NULL;
+
+    case NT_PRSTATUS:
+      {
+	char data[256];
+	va_list ap;
+	long pid;
+	int cursig;
+	const void *greg;
+
+	va_start (ap, note_type);
+	memset (data, 0, 72);
+	pid = va_arg (ap, long);
+	bfd_put_32 (abfd, pid, data + 24);
+	cursig = va_arg (ap, int);
+	bfd_put_16 (abfd, cursig, data + 12);
+	greg = va_arg (ap, const void *);
+	memcpy (data + 72, greg, 180);
+	memset (data + 252, 0, 4);
+	va_end (ap);
+	return elfcore_write_note (abfd, buf, bufsiz,
+				   "CORE", note_type, data, sizeof (data));
+      }
+    }
+}
 \f
 /* Depending on the target vector we generate some version of Irix
    executables or "normal" MIPS ELF ABI executables.  */
@@ -2554,6 +2593,9 @@ static const struct ecoff_debug_swap mips_elf32_ecoff_debug_swap = {
 #define ELF_COMMONPAGESIZE		0x1000
 #define elf32_bed			elf32_tradbed
 
+#undef elf_backend_write_core_note
+#define elf_backend_write_core_note elf32_mips_write_core_note
+
 /* Include the target file again for this target.  */
 #include "elf32-target.h"
 
@@ -2575,6 +2617,8 @@ static const struct ecoff_debug_swap mips_elf32_ecoff_debug_swap = {
 #undef	elf32_bed
 #define elf32_bed				elf32_fbsd_tradbed
 
+#undef elf_backend_write_core_note
+
 #include "elf32-target.h"
 /* Implement elf_backend_final_write_processing for VxWorks.  */
 
diff --git a/bfd/elf64-mips.c b/bfd/elf64-mips.c
index d51124b..f3997bd 100644
--- a/bfd/elf64-mips.c
+++ b/bfd/elf64-mips.c
@@ -4248,6 +4248,45 @@ elf64_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 
   return TRUE;
 }
+
+/* Write Linux core PRSTATUS note into core file.  */
+
+static char *
+elf64_mips_write_core_note (bfd *abfd, char *buf, int *bufsiz, int note_type,
+			      ...)
+{
+  switch (note_type)
+    {
+    default:
+      return NULL;
+
+    case NT_PRPSINFO:
+      BFD_FAIL ();
+      return NULL;
+
+    case NT_PRSTATUS:
+      {
+	char data[480];
+	va_list ap;
+	long pid;
+	int cursig;
+	const void *greg;
+
+	va_start (ap, note_type);
+	memset (data, 0, 112);
+	pid = va_arg (ap, long);
+	bfd_put_32 (abfd, pid, data + 32);
+	cursig = va_arg (ap, int);
+	bfd_put_16 (abfd, cursig, data + 12);
+	greg = va_arg (ap, const void *);
+	memcpy (data + 112, greg, 360);
+	memset (data + 472, 0, 8);
+	va_end (ap);
+	return elfcore_write_note (abfd, buf, bufsiz,
+				   "CORE", note_type, data, sizeof (data));
+      }
+    }
+}
 \f
 /* ECOFF swapping routines.  These are used when dealing with the
    .mdebug section, which is in the ECOFF debugging format.  */
@@ -4454,6 +4493,9 @@ const struct elf_size_info mips_elf64_size_info =
 #define ELF_COMMONPAGESIZE		0x1000
 #define elf64_bed			elf64_tradbed
 
+#undef elf_backend_write_core_note
+#define elf_backend_write_core_note elf64_mips_write_core_note
+
 /* Include the target file again for this target.  */
 #include "elf64-target.h"
 
@@ -4476,4 +4518,6 @@ const struct elf_size_info mips_elf64_size_info =
 #undef	elf64_bed
 #define elf64_bed				elf64_fbsd_tradbed
 
+#undef elf64_mips_write_core_note
+
 #include "elf64-target.h"
diff --git a/bfd/elfn32-mips.c b/bfd/elfn32-mips.c
index fdae183..d9d2020 100644
--- a/bfd/elfn32-mips.c
+++ b/bfd/elfn32-mips.c
@@ -3630,6 +3630,45 @@ elf_n32_mips_grok_freebsd_prstatus (bfd *abfd, Elf_Internal_Note *note)
   return _bfd_elfcore_make_pseudosection (abfd, ".reg",
 					  size, note->descpos + offset);
 }
+
+/* Write Linux core PRSTATUS note into core file.  */
+
+static char *
+elf32_mips_write_core_note (bfd *abfd, char *buf, int *bufsiz, int note_type,
+			      ...)
+{
+  switch (note_type)
+    {
+    default:
+      return NULL;
+
+    case NT_PRPSINFO:
+      BFD_FAIL ();
+      return NULL;
+
+    case NT_PRSTATUS:
+      {
+	char data[440];
+	va_list ap;
+	long pid;
+	int cursig;
+	const void *greg;
+
+	va_start (ap, note_type);
+	memset (data, 0, 72);
+	pid = va_arg (ap, long);
+	bfd_put_32 (abfd, pid, data + 24);
+	cursig = va_arg (ap, int);
+	bfd_put_16 (abfd, cursig, data + 12);
+	greg = va_arg (ap, const void *);
+	memcpy (data + 72, greg, 360);
+	memset (data + 432, 0, 8);
+	va_end (ap);
+	return elfcore_write_note (abfd, buf, bufsiz,
+				   "CORE", note_type, data, sizeof (data));
+      }
+    }
+}
 \f
 /* Depending on the target vector we generate some version of Irix
    executables or "normal" MIPS ELF ABI executables.  */
@@ -3807,6 +3846,9 @@ static const struct ecoff_debug_swap mips_elf32_ecoff_debug_swap = {
 #define ELF_COMMONPAGESIZE		0x1000
 #define elf32_bed			elf32_tradbed
 
+#undef elf_backend_write_core_note
+#define elf_backend_write_core_note elf32_mips_write_core_note
+
 /* Include the target file again for this target.  */
 #include "elf32-target.h"
 
@@ -3829,4 +3871,6 @@ static const struct ecoff_debug_swap mips_elf32_ecoff_debug_swap = {
 #undef	elf32_bed
 #define elf32_bed				elf32_fbsd_tradbed
 
+#undef elf_backend_write_core_note
+
 #include "elf32-target.h"
-- 
2.7.4

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH 2/3] BFD: Extract PID from MIPS core dump file
  2017-10-17 13:50 [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file Djordje Todorovic
  2017-10-17 13:50 ` [PATCH 3/3] Add test for fetching TLS from " Djordje Todorovic
@ 2017-10-17 13:50 ` Djordje Todorovic
  2017-10-18 13:36   ` Maciej W. Rozycki
  2017-10-18 13:36 ` [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file Maciej W. Rozycki
  2 siblings, 1 reply; 8+ messages in thread
From: Djordje Todorovic @ 2017-10-17 13:50 UTC (permalink / raw)
  To: macro
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica, Djordje Todorovic

On MIPS o32, n32 and n64 platforms, PID information was not
correctly propagated from core dump file to internal GDB
structures.  This patch fixes that behavior.

GDB needs correct PID in order to fetch value of TLS variable
from core file.

bfd/ChangeLog:

     * elf32-mips.c (elf32_mips_grok_psinfo): Extract core->pid.
     * elf64-mips.c (elf64_mips_grok_psinfo): Likewise.
     * elfn32-mips.c (elf32_mips_grok_psinfo): Likewise.
---
 bfd/elf32-mips.c  | 2 ++
 bfd/elf64-mips.c  | 2 ++
 bfd/elfn32-mips.c | 2 ++
 3 files changed, 6 insertions(+)

diff --git a/bfd/elf32-mips.c b/bfd/elf32-mips.c
index 2ee15e9..86c5e43 100644
--- a/bfd/elf32-mips.c
+++ b/bfd/elf32-mips.c
@@ -2353,6 +2353,8 @@ elf32_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 	return FALSE;
 
       case 128:		/* Linux/MIPS elf_prpsinfo */
+	elf_tdata (abfd)->core->pid
+	 = bfd_get_32 (abfd, note->descdata + 16);
 	elf_tdata (abfd)->core->program
 	 = _bfd_elfcore_strndup (abfd, note->descdata + 32, 16);
 	elf_tdata (abfd)->core->command
diff --git a/bfd/elf64-mips.c b/bfd/elf64-mips.c
index f3997bd..0227a0f 100644
--- a/bfd/elf64-mips.c
+++ b/bfd/elf64-mips.c
@@ -4228,6 +4228,8 @@ elf64_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 	return FALSE;
 
       case 136:		/* Linux/MIPS - N64 kernel elf_prpsinfo */
+	elf_tdata (abfd)->core->pid
+	 = bfd_get_32 (abfd, note->descdata + 24);
 	elf_tdata (abfd)->core->program
 	 = _bfd_elfcore_strndup (abfd, note->descdata + 40, 16);
 	elf_tdata (abfd)->core->command
diff --git a/bfd/elfn32-mips.c b/bfd/elfn32-mips.c
index d9d2020..cb7e26c 100644
--- a/bfd/elfn32-mips.c
+++ b/bfd/elfn32-mips.c
@@ -3560,6 +3560,8 @@ elf32_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 	return FALSE;
 
       case 128:		/* Linux/MIPS elf_prpsinfo */
+	elf_tdata (abfd)->core->pid
+	 = bfd_get_32 (abfd, note->descdata + 16);
 	elf_tdata (abfd)->core->program
 	 = _bfd_elfcore_strndup (abfd, note->descdata + 32, 16);
 	elf_tdata (abfd)->core->command
-- 
2.7.4

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH 3/3] Add test for fetching TLS from core file
  2017-10-17 13:50 [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file Djordje Todorovic
@ 2017-10-17 13:50 ` Djordje Todorovic
  2017-10-18 13:37   ` Maciej W. Rozycki
  2017-10-17 13:50 ` [PATCH 2/3] BFD: Extract PID from MIPS core dump file Djordje Todorovic
  2017-10-18 13:36 ` [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file Maciej W. Rozycki
  2 siblings, 1 reply; 8+ messages in thread
From: Djordje Todorovic @ 2017-10-17 13:50 UTC (permalink / raw)
  To: macro
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica, Djordje Todorovic

GDB needs correct PID in order to fetch value of TLS variable from
core file.  On MIPS platforms it was omitted and fetching value of
TLS variable was not available.

This adds a new test in order to be sure if GDB on native platforms
can successfully fetch value of TLS variable.

gdb/testsuite:

	* gdb.threads/tls-core.c: New file.
	* gdb.threads/tls-core.exp: Likewise.
---
 gdb/testsuite/gdb.threads/tls-core.c   | 44 +++++++++++++++++++++++++
 gdb/testsuite/gdb.threads/tls-core.exp | 60 ++++++++++++++++++++++++++++++++++
 2 files changed, 104 insertions(+)
 create mode 100644 gdb/testsuite/gdb.threads/tls-core.c
 create mode 100644 gdb/testsuite/gdb.threads/tls-core.exp

diff --git a/gdb/testsuite/gdb.threads/tls-core.c b/gdb/testsuite/gdb.threads/tls-core.c
new file mode 100644
index 0000000..e41c584
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/tls-core.c
@@ -0,0 +1,44 @@
+/* This test is part of GDB, the GNU debugger.
+
+   Copyright 2017 Free Software Foundation, Inc.
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <pthread.h>
+
+#define NUMBER_OF_THREADS 5
+
+int __thread foo = 0xdeadbeef;
+
+void *
+thread (void *in) /* BREAK */
+{
+  int *tmp = (int *) in;
+  int value = *tmp;
+  foo += *tmp;
+  while (1)
+    sleep (10);
+}
+
+int
+main (void)
+{
+  pthread_t threads[NUMBER_OF_THREADS];
+  int i;
+  for (i = 0; i < NUMBER_OF_THREADS; i++)
+    pthread_create (&threads[i], NULL, thread, &i);
+}
\ No newline at end of file
diff --git a/gdb/testsuite/gdb.threads/tls-core.exp b/gdb/testsuite/gdb.threads/tls-core.exp
new file mode 100644
index 0000000..dc291ab
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/tls-core.exp
@@ -0,0 +1,60 @@
+# Copyright 2017 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+standard_testfile
+
+if {[gdb_compile_pthreads "${srcdir}/${subdir}/${srcfile}" "${binfile}" \
+	 executable { debug }] != "" } {
+    return -1
+}
+
+
+clean_restart ${binfile}
+
+#
+# Set breakpoint at line 28.
+#
+gdb_breakpoint [gdb_get_line_number "BREAK"]
+gdb_test "run" "Starting program: .*"
+
+#
+# Generate corefile.
+#
+set corefile [standard_output_file gcore.test]
+set core_supported [gdb_gcore_cmd "$corefile" "save a corefile"]
+if {!$core_supported} {
+    return -1
+}
+
+gdb_exit
+
+#
+# Restart gdb and load generated corefile.
+#
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load ${binfile}
+
+set core_loaded [gdb_core_cmd "$corefile" "load generated corefile"]
+if { $core_loaded == -1 } {
+    # No use proceeding from here.
+    return
+}
+
+gdb_test "p/x foo" \
+                    "\\$\[0-9]+ = 0xdeadbeef" \
+                    "Printing thread-local storage variable."
+
+gdb_exit
-- 
2.7.4

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 2/3] BFD: Extract PID from MIPS core dump file
  2017-10-17 13:50 ` [PATCH 2/3] BFD: Extract PID from MIPS core dump file Djordje Todorovic
@ 2017-10-18 13:36   ` Maciej W. Rozycki
  0 siblings, 0 replies; 8+ messages in thread
From: Maciej W. Rozycki @ 2017-10-18 13:36 UTC (permalink / raw)
  To: Djordje Todorovic
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica

On Tue, 17 Oct 2017, Djordje Todorovic wrote:

> On MIPS o32, n32 and n64 platforms, PID information was not
> correctly propagated from core dump file to internal GDB
> structures.  This patch fixes that behavior.
> 
> GDB needs correct PID in order to fetch value of TLS variable
> from core file.

 This is OK.  Thanks.

  Maciej

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file
  2017-10-17 13:50 [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file Djordje Todorovic
  2017-10-17 13:50 ` [PATCH 3/3] Add test for fetching TLS from " Djordje Todorovic
  2017-10-17 13:50 ` [PATCH 2/3] BFD: Extract PID from MIPS core dump file Djordje Todorovic
@ 2017-10-18 13:36 ` Maciej W. Rozycki
  2 siblings, 0 replies; 8+ messages in thread
From: Maciej W. Rozycki @ 2017-10-18 13:36 UTC (permalink / raw)
  To: Djordje Todorovic
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica

On Tue, 17 Oct 2017, Djordje Todorovic wrote:

> diff --git a/bfd/elf32-mips.c b/bfd/elf32-mips.c
> index 17ca432..2ee15e9 100644
> --- a/bfd/elf32-mips.c
> +++ b/bfd/elf32-mips.c
> @@ -2373,6 +2373,45 @@ elf32_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
>  
>    return TRUE;
>  }
> +
> +/* Write Linux core PRSTATUS note into core file.  */
> +
> +static char *
> +elf32_mips_write_core_note (bfd *abfd, char *buf, int *bufsiz, int note_type,
> +			      ...)

 Formatting: please align the wrapped line to the inside of the opening 
parenthesis above, i.e.:

elf32_mips_write_core_note (bfd *abfd, char *buf, int *bufsiz, int note_type,
			    ...)

> @@ -2554,6 +2593,9 @@ static const struct ecoff_debug_swap mips_elf32_ecoff_debug_swap = {
>  #define ELF_COMMONPAGESIZE		0x1000
>  #define elf32_bed			elf32_tradbed
>  
> +#undef elf_backend_write_core_note
> +#define elf_backend_write_core_note elf32_mips_write_core_note

 Formatting: please use tabs to align the RHS to column #40 like all the 
other such definitions.

> diff --git a/bfd/elf64-mips.c b/bfd/elf64-mips.c
> index d51124b..f3997bd 100644
> --- a/bfd/elf64-mips.c
> +++ b/bfd/elf64-mips.c
> @@ -4248,6 +4248,45 @@ elf64_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
>  
>    return TRUE;
>  }
> +
> +/* Write Linux core PRSTATUS note into core file.  */
> +
> +static char *
> +elf64_mips_write_core_note (bfd *abfd, char *buf, int *bufsiz, int note_type,
> +			      ...)

 Formatting.

> @@ -4454,6 +4493,9 @@ const struct elf_size_info mips_elf64_size_info =
>  #define ELF_COMMONPAGESIZE		0x1000
>  #define elf64_bed			elf64_tradbed
>  
> +#undef elf_backend_write_core_note
> +#define elf_backend_write_core_note elf64_mips_write_core_note

 Formatting.

> diff --git a/bfd/elfn32-mips.c b/bfd/elfn32-mips.c
> index fdae183..d9d2020 100644
> --- a/bfd/elfn32-mips.c
> +++ b/bfd/elfn32-mips.c
> @@ -3630,6 +3630,45 @@ elf_n32_mips_grok_freebsd_prstatus (bfd *abfd, Elf_Internal_Note *note)
>    return _bfd_elfcore_make_pseudosection (abfd, ".reg",
>  					  size, note->descpos + offset);
>  }
> +
> +/* Write Linux core PRSTATUS note into core file.  */
> +
> +static char *
> +elf32_mips_write_core_note (bfd *abfd, char *buf, int *bufsiz, int note_type,
> +			      ...)

 Formatting.

> @@ -3807,6 +3846,9 @@ static const struct ecoff_debug_swap mips_elf32_ecoff_debug_swap = {
>  #define ELF_COMMONPAGESIZE		0x1000
>  #define elf32_bed			elf32_tradbed
>  
> +#undef elf_backend_write_core_note
> +#define elf_backend_write_core_note elf32_mips_write_core_note

 Formatting.

  Maciej

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 3/3] Add test for fetching TLS from core file
  2017-10-17 13:50 ` [PATCH 3/3] Add test for fetching TLS from " Djordje Todorovic
@ 2017-10-18 13:37   ` Maciej W. Rozycki
  2017-10-24 11:17     ` Pedro Alves
  0 siblings, 1 reply; 8+ messages in thread
From: Maciej W. Rozycki @ 2017-10-18 13:37 UTC (permalink / raw)
  To: Djordje Todorovic, Pedro Alves
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica

On Tue, 17 Oct 2017, Djordje Todorovic wrote:

> diff --git a/gdb/testsuite/gdb.threads/tls-core.c b/gdb/testsuite/gdb.threads/tls-core.c
> new file mode 100644
> index 0000000..e41c584
> --- /dev/null
> +++ b/gdb/testsuite/gdb.threads/tls-core.c
> @@ -0,0 +1,44 @@
> +/* This test is part of GDB, the GNU debugger.
> +
> +   Copyright 2017 Free Software Foundation, Inc.
> +
> +   This program is free software; you can redistribute it and/or modify
> +   it under the terms of the GNU General Public License as published by
> +   the Free Software Foundation; either version 3 of the License, or
> +   (at your option) any later version.
> +
> +   This program is distributed in the hope that it will be useful,
> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> +   GNU General Public License for more details.
> +
> +   You should have received a copy of the GNU General Public License
> +   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
> +
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +#include <pthread.h>
> +
> +#define NUMBER_OF_THREADS 5
> +
> +int __thread foo = 0xdeadbeef;
> +
> +void *
> +thread (void *in) /* BREAK */
> +{
> +  int *tmp = (int *) in;
> +  int value = *tmp;
> +  foo += *tmp;
> +  while (1)
> +    sleep (10);
> +}
> +
> +int
> +main (void)
> +{
> +  pthread_t threads[NUMBER_OF_THREADS];
> +  int i;
> +  for (i = 0; i < NUMBER_OF_THREADS; i++)
> +    pthread_create (&threads[i], NULL, thread, &i);

 Missing `return' statement here.

 Also `while (1) sleep (10);' perhaps, so that you don't cause a race with 
the whole process terminating (I think `pthread_join (&threads[0], NULL);' 
would do too; I'm sure there are other ways as well to prevent from 
running to exit(2)).  Overall please don't assume any particular thread 
scheduling characteristics of the underlying OS, and especially on SMP 
systems effects may be sometimes surprising.

> +}
> \ No newline at end of file

 Missing new line here.

> diff --git a/gdb/testsuite/gdb.threads/tls-core.exp b/gdb/testsuite/gdb.threads/tls-core.exp
> new file mode 100644
> index 0000000..dc291ab
> --- /dev/null
> +++ b/gdb/testsuite/gdb.threads/tls-core.exp
> @@ -0,0 +1,60 @@
> +# Copyright 2017 Free Software Foundation, Inc.
> +
> +# This program is free software; you can redistribute it and/or modify
> +# it under the terms of the GNU General Public License as published by
> +# the Free Software Foundation; either version 3 of the License, or
> +# (at your option) any later version.
> +#
> +# This program is distributed in the hope that it will be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> +# GNU General Public License for more details.
> +#
> +# You should have received a copy of the GNU General Public License
> +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
> +
> +standard_testfile
> +
> +if {[gdb_compile_pthreads "${srcdir}/${subdir}/${srcfile}" "${binfile}" \
> +	 executable { debug }] != "" } {

 Formatting: space after a tab here.

> +    return -1
> +}
> +
> +
> +clean_restart ${binfile}
> +
> +#
> +# Set breakpoint at line 28.
> +#
> +gdb_breakpoint [gdb_get_line_number "BREAK"]
> +gdb_test "run" "Starting program: .*"

 The `run' command isn't supported with all targets and we do not wan't to 
cause a test failure unrelated to the matter being verified.  Please use 
`runto_main' then, i.e.:

clean_restart ${binfile}

runto_main

Sorry to miss that before.

 Also since you set a breakpoint at a function (which wasn't evident 
before) you should be able to simplify processing and just use:

runto thread

instead of going through the dance to set and catch the breakpoint 
manually.

> +
> +#
> +# Generate corefile.
> +#
> +set corefile [standard_output_file gcore.test]
> +set core_supported [gdb_gcore_cmd "$corefile" "save a corefile"]
> +if {!$core_supported} {
> +    return -1
> +}
> +
> +gdb_exit
> +
> +#
> +# Restart gdb and load generated corefile.
> +#
> +gdb_start
> +gdb_reinitialize_dir $srcdir/$subdir
> +gdb_load ${binfile}
> +
> +set core_loaded [gdb_core_cmd "$corefile" "load generated corefile"]
> +if { $core_loaded == -1 } {

 I think this has to be `$core_loaded != 1' or suchlike as I doubt a core 
file with no registers recorded will be usable for the actual test being 
made here.  I missed this peculiarity before, sorry.

> +    # No use proceeding from here.
> +    return

 I missed this on the previous iteration, sorry: why no return result here 
(i.e. an empty string being returned) vs -1 elsewhere?

 I don't think we have a clear rule for return results defined and 
moreover the caller actually ignores it.  Regardless, we set the result, 
differently, across the test suite.

 The value of -1 seems to be used as an indicator of a fatal test 
processing failure, such as with the first `return' statement above.  And 
in the case of a failure `gdb_compile_pthreads' will have already called 
`unsupported', so there's nothing else for us to do here.

 The second `return' statement is neither a fatal test processing failure 
nor a test failure.  It merely marks the `gcore' feature not being 
supported and consequently the inability to continue with the test case.  
In this case `gdb_gcore_cmd' will have already called `unsupported' as 
expected.  So I think it should return 0 or nothing (i.e. an empty 
string).

 The third `return' statement actually marks a test failure, because GDB 
is supposed to be able to load a core file in all cases, but not a fatal 
test processing failure.  Again in this case `gdb_core_cmd' will have 
already called `fail' as expected.  So again I think it should return 0 or 
nothing.

 I'd like to hear from a general maintainer about this though.

 Pedro, would you be able to comment on my observations and, ultimately 
(once the details have been sorted out), approve this change (as 
technically it's outside my area of competence anyway)?

> +}
> +
> +gdb_test "p/x foo" \
> +                    "\\$\[0-9]+ = 0xdeadbeef" \
> +                    "Printing thread-local storage variable."

 Formatting: use tabs rather than spaces to indent; one per line will do.

 Please wait a couple of days to give Pedro an opportunity to chime in and 
then resubmit the whole series with the details above and 1/3 updated.

  Maciej

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 3/3] Add test for fetching TLS from core file
  2017-10-18 13:37   ` Maciej W. Rozycki
@ 2017-10-24 11:17     ` Pedro Alves
  0 siblings, 0 replies; 8+ messages in thread
From: Pedro Alves @ 2017-10-24 11:17 UTC (permalink / raw)
  To: Maciej W. Rozycki, Djordje Todorovic
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica

Hi Maciej,

Sorry, I hadn't noticed the question below directed at me.

On 10/18/2017 02:35 PM, Maciej W. Rozycki wrote:

>  Pedro, would you be able to comment on my observations and, 

They sound reasonable to me.

> ultimately 
> (once the details have been sorted out), approve this change (as 
> technically it's outside my area of competence anyway)?

Sure.

>> +}
>> +
>> +gdb_test "p/x foo" \
>> +                    "\\$\[0-9]+ = 0xdeadbeef" \
>> +                    "Printing thread-local storage variable."
> 
>  Formatting: use tabs rather than spaces to indent; one per line will do.
> 
>  Please wait a couple of days to give Pedro an opportunity to chime in and 
> then resubmit the whole series with the details above and 1/3 updated.
> 

There's code in the "thread" function that seemingly will never
be executed:

 +void *
 +thread (void *in) /* BREAK */
 +{
 +  int *tmp = (int *) in;
 +  int value = *tmp;
 +  foo += *tmp;
 +  while (1)
 +    sleep (10);
 +}

since there's a breakpoint set at "BREAK" and the testcase stops there
to generate a core with "gcore" as soon as the first thread
reaches the "BREAK" line.

But what's the point of that code in the "thread" function if
it's never going to be executed?  That only serves to obfuscate
as is.  I may well have missed something, of course.

 +gdb_test "p/x foo" \
 +                    "\\$\[0-9]+ = 0xdeadbeef" \
 +                    "Printing thread-local storage variable."

For test name/message, please write instead:

                 "print thread-local storage variable"

i.e., no uppercase, no period, and avoid gerund.

+gdb_exit
+
+#
+# Restart gdb and load generated corefile.
+#
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load ${binfile}

This above can all be replaced by a call to clean_restart.

Thanks,
Pedro Alves

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH 2/3] BFD: Extract PID from MIPS core dump file
  2017-10-25 14:15 Djordje Todorovic
@ 2017-10-25 14:15 ` Djordje Todorovic
  0 siblings, 0 replies; 8+ messages in thread
From: Djordje Todorovic @ 2017-10-25 14:15 UTC (permalink / raw)
  To: macro
  Cc: binutils, gdb-patches, asowda, petar.jovanovic, nemanja.popov,
	nikola.prica, palves, Djordje Todorovic

On MIPS o32, n32 and n64 platforms, PID information was not
correctly propagated from core dump file to internal GDB
structures.  This patch fixes that behavior.

GDB needs correct PID in order to fetch value of TLS variable
from core file.

bfd/ChangeLog:

     * elf32-mips.c (elf32_mips_grok_psinfo): Extract core->pid.
     * elf64-mips.c (elf64_mips_grok_psinfo): Likewise.
     * elfn32-mips.c (elf32_mips_grok_psinfo): Likewise.
---
 bfd/elf32-mips.c  | 2 ++
 bfd/elf64-mips.c  | 2 ++
 bfd/elfn32-mips.c | 2 ++
 3 files changed, 6 insertions(+)

diff --git a/bfd/elf32-mips.c b/bfd/elf32-mips.c
index 69c6061..de204bc 100644
--- a/bfd/elf32-mips.c
+++ b/bfd/elf32-mips.c
@@ -2353,6 +2353,8 @@ elf32_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 	return FALSE;
 
       case 128:		/* Linux/MIPS elf_prpsinfo */
+	elf_tdata (abfd)->core->pid
+	 = bfd_get_32 (abfd, note->descdata + 16);
 	elf_tdata (abfd)->core->program
 	 = _bfd_elfcore_strndup (abfd, note->descdata + 32, 16);
 	elf_tdata (abfd)->core->command
diff --git a/bfd/elf64-mips.c b/bfd/elf64-mips.c
index 69f365b..6a1b71f 100644
--- a/bfd/elf64-mips.c
+++ b/bfd/elf64-mips.c
@@ -4228,6 +4228,8 @@ elf64_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 	return FALSE;
 
       case 136:		/* Linux/MIPS - N64 kernel elf_prpsinfo */
+	elf_tdata (abfd)->core->pid
+	 = bfd_get_32 (abfd, note->descdata + 24);
 	elf_tdata (abfd)->core->program
 	 = _bfd_elfcore_strndup (abfd, note->descdata + 40, 16);
 	elf_tdata (abfd)->core->command
diff --git a/bfd/elfn32-mips.c b/bfd/elfn32-mips.c
index 790baee..3e34eef 100644
--- a/bfd/elfn32-mips.c
+++ b/bfd/elfn32-mips.c
@@ -3562,6 +3562,8 @@ elf32_mips_grok_psinfo (bfd *abfd, Elf_Internal_Note *note)
 	return FALSE;
 
       case 128:		/* Linux/MIPS elf_prpsinfo */
+	elf_tdata (abfd)->core->pid
+	 = bfd_get_32 (abfd, note->descdata + 16);
 	elf_tdata (abfd)->core->program
 	 = _bfd_elfcore_strndup (abfd, note->descdata + 32, 16);
 	elf_tdata (abfd)->core->command
-- 
2.7.4

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2017-10-25 14:15 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-10-17 13:50 [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file Djordje Todorovic
2017-10-17 13:50 ` [PATCH 3/3] Add test for fetching TLS from " Djordje Todorovic
2017-10-18 13:37   ` Maciej W. Rozycki
2017-10-24 11:17     ` Pedro Alves
2017-10-17 13:50 ` [PATCH 2/3] BFD: Extract PID from MIPS core dump file Djordje Todorovic
2017-10-18 13:36   ` Maciej W. Rozycki
2017-10-18 13:36 ` [PATCH 1/3] BFD: Write Linux core PRSTATUS note into MIPS core file Maciej W. Rozycki
2017-10-25 14:15 Djordje Todorovic
2017-10-25 14:15 ` [PATCH 2/3] BFD: Extract PID from MIPS core dump file Djordje Todorovic

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).