public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: Aleksandar Ristovski <aristovski@qnx.com>
To: Pedro Alves <palves@redhat.com>, gdb-patches@sourceware.org
Subject: Re: [PATCH 2/2] [nto] Improve ABI sniffing.
Date: Wed, 21 Oct 2015 14:42:00 -0000	[thread overview]
Message-ID: <5627944D.3050900@qnx.com> (raw)
In-Reply-To: <56275FB2.3050509@redhat.com>

[-- Attachment #1: Type: text/plain, Size: 2393 bytes --]

On 15-10-21 05:49 AM, Pedro Alves wrote:
...
>> +  char *note; // buffer holding the section contents
> 
> Please use /**/ format for comments, and write full sentences - start
> with uppercase, period at end.  Usually that leads to putting the comment
> above, e.g.:
> 
>  +  /* Buffer holding the section contents.  */
>  +  char *note;

Done.

> 
>> +  unsigned int namelen;
>> +  const char *name;
>> +
>> +  sectname = bfd_get_section_name (abfd, sect);
>> +  sectsize = bfd_section_size (abfd, sect);
>> +
>> +  /* TODO: limit the note size here, for now limit is 128 bytes
>> +     (enough to check the name and type).  */
> 
> This reads like limiting is left to do, but then it does
> implement a limit.  So this TODO comment is confusing.

Comment removed.

> 
> You should also make sure the section is the at least
> the minimum size you expect though.

Done.

...
>> +    {
>> +      note = alloca (sectsize);
> 
> For C++, write:
> 
>       note = (char *) alloca (sectsize);

Used XNEWVEC/XDELETEVEC instead.

> 
> 
>> +      bfd_get_section_contents (abfd, sect, note, 0, sectsize);
>> +      namelen = (unsigned int) bfd_h_get_32 (abfd, note);
>> +      name = note + 12;
>> +
>> +      if (namelen > 0
>> +	  && (0 == strcmp (name, QNX_NOTE_NAME)))
>> +        *(enum gdb_osabi *) obj = GDB_OSABI_QNXNTO;
>> +    }
>> +}
>> +
>>  enum gdb_osabi
>>  nto_elf_osabi_sniffer (bfd *abfd)
>>  {
>> -  if (nto_is_nto_target)
>> +  enum gdb_osabi osabi = GDB_OSABI_UNKNOWN;
>> +
>> +  bfd_map_over_sections (abfd,
>> +			 nto_sniff_abi_note_section,
>> +			 &osabi);
>> +
>> +  if (osabi == GDB_OSABI_UNKNOWN && nto_is_nto_target)
>>      return nto_is_nto_target (abfd);
> 
> ...
> nto_is_nto_target = procfs_is_nto_target;
> ...
> 
> static enum gdb_osabi
> procfs_is_nto_target (bfd *abfd)
> {
>   return GDB_OSABI_QNXNTO;
> }
> 
> So that basically could be rewritten as:
> 
>   if (osabi == GDB_OSABI_UNKNOWN && nto_is_nto_target)
>      return GDB_OSABI_QNXNTO;
> 
> 
> But, do we still need this nto_is_nto_target hack here?
> Now with proper sniffing, can't we just remove it altogether?

Removed. Rely on new sniffing only, no hard coded fallback.


> 
>> -  return GDB_OSABI_UNKNOWN;
>> +  return osabi;
>>  }
>>  
>>  static const char *nto_thread_state_str[] =
>>
> 
> 
> Thanks,
> Pedro Alves
> 
> 


New version of the patch attached.

Thank you,

Aleksandar Ristovski

[-- Attachment #2: 0002-nto-Improve-ABI-sniffing.patch --]
[-- Type: text/x-patch, Size: 2513 bytes --]

From 6df97e71696e47d85000413bf4a97cf5d43b11f1 Mon Sep 17 00:00:00 2001
From: Aleksandar Ristovski <aristovski@qnx.com>
Date: Wed, 21 Oct 2015 09:29:54 -0400
Subject: [PATCH 2/2] [nto] Improve ABI sniffing.

Use qnx specific notes to figure out the OS.

gdb/ChangeLog:
	* gdb/nto-tdep.c (QNX_NOTE_NAME, QNX_INFO_SECT_NAME): New defines.
	(nto_sniff_abi_note_section): New function.
	(nto_elf_osabi_sniffer): Use new function to recognize nto specific
	binary.
---
 gdb/nto-tdep.c | 55 ++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 52 insertions(+), 3 deletions(-)

diff --git a/gdb/nto-tdep.c b/gdb/nto-tdep.c
index e50d302..4e0cc84 100644
--- a/gdb/nto-tdep.c
+++ b/gdb/nto-tdep.c
@@ -32,6 +32,9 @@
 #include "gdbcore.h"
 #include "objfiles.h"
 
+#define QNX_NOTE_NAME	"QNX"
+#define QNX_INFO_SECT_NAME "QNX_info"
+
 #ifdef __CYGWIN__
 #include <sys/cygwin.h>
 #endif
@@ -332,12 +335,58 @@ nto_dummy_supply_regset (struct regcache *regcache, char *regs)
   /* Do nothing.  */
 }
 
+static void
+nto_sniff_abi_note_section (bfd *abfd, asection *sect, void *obj)
+{
+  const char *sectname;
+  unsigned int sectsize;
+  /* Buffer holding the section contents.  */
+  char *note;
+  unsigned int namelen;
+  const char *name;
+
+  sectname = bfd_get_section_name (abfd, sect);
+  sectsize = bfd_section_size (abfd, sect);
+
+  if (sectsize > 128)
+    sectsize = 128;
+
+  if (sectname != NULL && strstr (sectname, QNX_INFO_SECT_NAME) != NULL)
+    *(enum gdb_osabi *) obj = GDB_OSABI_QNXNTO;
+
+  if (sectname != NULL && strstr (sectname, "note") != NULL)
+    {
+      const unsigned sizeof_Elf_Nhdr = 12;
+
+      note = XNEWVEC (char, sectsize);
+      bfd_get_section_contents (abfd, sect, note, 0, sectsize);
+      namelen = (unsigned int) bfd_h_get_32 (abfd, note);
+      name = note + sizeof_Elf_Nhdr;
+      if (sectsize < namelen + sizeof_Elf_Nhdr
+	  || namelen > sizeof (QNX_NOTE_NAME) + 1)
+	{
+	  /* Can not be QNX note.  */
+	  XDELETEVEC (note);
+	  return;
+	}
+
+      if (namelen > 0 && 0 == strcmp (name, QNX_NOTE_NAME))
+        *(enum gdb_osabi *) obj = GDB_OSABI_QNXNTO;
+
+      XDELETEVEC (note);
+    }
+}
+
 enum gdb_osabi
 nto_elf_osabi_sniffer (bfd *abfd)
 {
-  if (nto_is_nto_target)
-    return nto_is_nto_target (abfd);
-  return GDB_OSABI_UNKNOWN;
+  enum gdb_osabi osabi = GDB_OSABI_UNKNOWN;
+
+  bfd_map_over_sections (abfd,
+			 nto_sniff_abi_note_section,
+			 &osabi);
+
+  return osabi;
 }
 
 static const char *nto_thread_state_str[] =
-- 
1.9.1


WARNING: multiple messages have this Message-ID
From: Aleksandar Ristovski <aristovski@qnx.com>
To: gdb-patches@sourceware.org
Subject: Re: [PATCH 2/2] [nto] Improve ABI sniffing.
Date: Wed, 21 Oct 2015 14:47:00 -0000	[thread overview]
Message-ID: <5627944D.3050900@qnx.com> (raw)
Message-ID: <20151021144700.2D8gGmbnLfZT_fYtvKOMOzbjs2B7wWaEWdjrGYCU5Ps@z> (raw)
In-Reply-To: <56275FB2.3050509@redhat.com>

[-- Attachment #1: Type: text/plain, Size: 2393 bytes --]

On 15-10-21 05:49 AM, Pedro Alves wrote:
...
>> +  char *note; // buffer holding the section contents
> 
> Please use /**/ format for comments, and write full sentences - start
> with uppercase, period at end.  Usually that leads to putting the comment
> above, e.g.:
> 
>  +  /* Buffer holding the section contents.  */
>  +  char *note;

Done.

> 
>> +  unsigned int namelen;
>> +  const char *name;
>> +
>> +  sectname = bfd_get_section_name (abfd, sect);
>> +  sectsize = bfd_section_size (abfd, sect);
>> +
>> +  /* TODO: limit the note size here, for now limit is 128 bytes
>> +     (enough to check the name and type).  */
> 
> This reads like limiting is left to do, but then it does
> implement a limit.  So this TODO comment is confusing.

Comment removed.

> 
> You should also make sure the section is the at least
> the minimum size you expect though.

Done.

...
>> +    {
>> +      note = alloca (sectsize);
> 
> For C++, write:
> 
>       note = (char *) alloca (sectsize);

Used XNEWVEC/XDELETEVEC instead.

> 
> 
>> +      bfd_get_section_contents (abfd, sect, note, 0, sectsize);
>> +      namelen = (unsigned int) bfd_h_get_32 (abfd, note);
>> +      name = note + 12;
>> +
>> +      if (namelen > 0
>> +	  && (0 == strcmp (name, QNX_NOTE_NAME)))
>> +        *(enum gdb_osabi *) obj = GDB_OSABI_QNXNTO;
>> +    }
>> +}
>> +
>>  enum gdb_osabi
>>  nto_elf_osabi_sniffer (bfd *abfd)
>>  {
>> -  if (nto_is_nto_target)
>> +  enum gdb_osabi osabi = GDB_OSABI_UNKNOWN;
>> +
>> +  bfd_map_over_sections (abfd,
>> +			 nto_sniff_abi_note_section,
>> +			 &osabi);
>> +
>> +  if (osabi == GDB_OSABI_UNKNOWN && nto_is_nto_target)
>>      return nto_is_nto_target (abfd);
> 
> ...
> nto_is_nto_target = procfs_is_nto_target;
> ...
> 
> static enum gdb_osabi
> procfs_is_nto_target (bfd *abfd)
> {
>   return GDB_OSABI_QNXNTO;
> }
> 
> So that basically could be rewritten as:
> 
>   if (osabi == GDB_OSABI_UNKNOWN && nto_is_nto_target)
>      return GDB_OSABI_QNXNTO;
> 
> 
> But, do we still need this nto_is_nto_target hack here?
> Now with proper sniffing, can't we just remove it altogether?

Removed. Rely on new sniffing only, no hard coded fallback.


> 
>> -  return GDB_OSABI_UNKNOWN;
>> +  return osabi;
>>  }
>>  
>>  static const char *nto_thread_state_str[] =
>>
> 
> 
> Thanks,
> Pedro Alves
> 
> 


New version of the patch attached.

Thank you,

Aleksandar Ristovski

[-- Attachment #2: 0002-nto-Improve-ABI-sniffing.patch --]
[-- Type: text/x-patch, Size: 2513 bytes --]

From 6df97e71696e47d85000413bf4a97cf5d43b11f1 Mon Sep 17 00:00:00 2001
From: Aleksandar Ristovski <aristovski@qnx.com>
Date: Wed, 21 Oct 2015 09:29:54 -0400
Subject: [PATCH 2/2] [nto] Improve ABI sniffing.

Use qnx specific notes to figure out the OS.

gdb/ChangeLog:
	* gdb/nto-tdep.c (QNX_NOTE_NAME, QNX_INFO_SECT_NAME): New defines.
	(nto_sniff_abi_note_section): New function.
	(nto_elf_osabi_sniffer): Use new function to recognize nto specific
	binary.
---
 gdb/nto-tdep.c | 55 ++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 52 insertions(+), 3 deletions(-)

diff --git a/gdb/nto-tdep.c b/gdb/nto-tdep.c
index e50d302..4e0cc84 100644
--- a/gdb/nto-tdep.c
+++ b/gdb/nto-tdep.c
@@ -32,6 +32,9 @@
 #include "gdbcore.h"
 #include "objfiles.h"
 
+#define QNX_NOTE_NAME	"QNX"
+#define QNX_INFO_SECT_NAME "QNX_info"
+
 #ifdef __CYGWIN__
 #include <sys/cygwin.h>
 #endif
@@ -332,12 +335,58 @@ nto_dummy_supply_regset (struct regcache *regcache, char *regs)
   /* Do nothing.  */
 }
 
+static void
+nto_sniff_abi_note_section (bfd *abfd, asection *sect, void *obj)
+{
+  const char *sectname;
+  unsigned int sectsize;
+  /* Buffer holding the section contents.  */
+  char *note;
+  unsigned int namelen;
+  const char *name;
+
+  sectname = bfd_get_section_name (abfd, sect);
+  sectsize = bfd_section_size (abfd, sect);
+
+  if (sectsize > 128)
+    sectsize = 128;
+
+  if (sectname != NULL && strstr (sectname, QNX_INFO_SECT_NAME) != NULL)
+    *(enum gdb_osabi *) obj = GDB_OSABI_QNXNTO;
+
+  if (sectname != NULL && strstr (sectname, "note") != NULL)
+    {
+      const unsigned sizeof_Elf_Nhdr = 12;
+
+      note = XNEWVEC (char, sectsize);
+      bfd_get_section_contents (abfd, sect, note, 0, sectsize);
+      namelen = (unsigned int) bfd_h_get_32 (abfd, note);
+      name = note + sizeof_Elf_Nhdr;
+      if (sectsize < namelen + sizeof_Elf_Nhdr
+	  || namelen > sizeof (QNX_NOTE_NAME) + 1)
+	{
+	  /* Can not be QNX note.  */
+	  XDELETEVEC (note);
+	  return;
+	}
+
+      if (namelen > 0 && 0 == strcmp (name, QNX_NOTE_NAME))
+        *(enum gdb_osabi *) obj = GDB_OSABI_QNXNTO;
+
+      XDELETEVEC (note);
+    }
+}
+
 enum gdb_osabi
 nto_elf_osabi_sniffer (bfd *abfd)
 {
-  if (nto_is_nto_target)
-    return nto_is_nto_target (abfd);
-  return GDB_OSABI_UNKNOWN;
+  enum gdb_osabi osabi = GDB_OSABI_UNKNOWN;
+
+  bfd_map_over_sections (abfd,
+			 nto_sniff_abi_note_section,
+			 &osabi);
+
+  return osabi;
 }
 
 static const char *nto_thread_state_str[] =
-- 
1.9.1


  reply	other threads:[~2015-10-21 13:34 UTC|newest]

Thread overview: 47+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-10-13 16:01 [PATCH 0/4] [nto] Nto fixes Aleksandar Ristovski
2015-10-13 16:01 ` [PATCH 4/4] [nto] Setup signals Aleksandar Ristovski
2015-10-16 16:16   ` Pedro Alves
2015-10-22 15:57     ` Aleksandar Ristovski
2015-10-22 15:58       ` Aleksandar Ristovski
2015-10-13 16:01 ` [PATCH 3/4] [nto] Fix nto target stopped by watchpoint Aleksandar Ristovski
2015-10-16 16:10   ` Pedro Alves
2015-10-20 18:42     ` [PATCH 0/2] (patch 3/4 v2) Broken down patch 3/4 Aleksandar Ristovski
2015-10-20 19:24       ` [PATCH 2/2] [nto] Improve ABI sniffing Aleksandar Ristovski
2015-10-21 10:39         ` Pedro Alves
2015-10-21 14:42           ` Aleksandar Ristovski [this message]
2015-10-21 14:47             ` Aleksandar Ristovski
2015-10-21 15:17             ` Pedro Alves
2015-10-21 15:37               ` Aleksandar Ristovski
2015-10-21 16:13                 ` Aleksandar Ristovski
2015-10-21 16:39                 ` Pedro Alves
2015-10-21 18:10                   ` Aleksandar Ristovski
2015-10-21 18:23                     ` Aleksandar Ristovski
2015-10-21  8:18       ` [PATCH 1/2] [nto] Fix nto target stopped by watchpoint Aleksandar Ristovski
2015-10-21 10:39         ` Pedro Alves
2015-10-21 17:51           ` Aleksandar Ristovski
2015-10-21 18:00             ` Aleksandar Ristovski
2015-10-13 16:01 ` [PATCH 1/4] [nto] Fix nto build Aleksandar Ristovski
2015-10-15 17:34   ` Pedro Alves
2015-10-13 16:01 ` [PATCH 2/4] [nto] Fixes for nto procfs Aleksandar Ristovski
2015-10-15 17:41   ` Pedro Alves
2015-10-20 12:43     ` Aleksandar Ristovski
2015-10-20 13:21       ` Aleksandar Ristovski
2015-10-20 14:28       ` Pedro Alves
2015-10-20 14:28         ` [PATCH 0/3] (patch 2/4, v2) Break patch 2/4 into 3 Aleksandar Ristovski
2015-10-20 14:28           ` [PATCH 2/3] (patch 2/4, v2) [nto] Implement TARGET_OBJECT_AUXV Aleksandar Ristovski
2015-10-20 15:24             ` Pedro Alves
2015-10-20 16:03               ` Aleksandar Ristovski
2015-10-20 16:48                 ` Pedro Alves
2015-10-20 17:08                   ` Aleksandar Ristovski
2015-10-20 17:13                     ` Aleksandar Ristovski
2015-10-20 18:11                     ` Aleksandar Ristovski
2015-10-20 18:39                       ` Aleksandar Ristovski
2015-10-20 18:11                     ` Pedro Alves
2015-10-20 14:29           ` [PATCH 1/3] (patch 2/4, v2) [nto] Fixes for nto procfs Aleksandar Ristovski
2015-10-20 15:20             ` Pedro Alves
2015-10-20 17:13               ` Aleksandar Ristovski
2015-10-20 17:14                 ` Aleksandar Ristovski
2015-10-20 15:03           ` [PATCH 3/3] (patch 2/4, v2) [nto] Implement procfs_pid_to_exec_file Aleksandar Ristovski
2015-10-20 15:25             ` Pedro Alves
2015-10-20 18:11               ` Aleksandar Ristovski
2015-10-20 18:19                 ` Aleksandar Ristovski

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=5627944D.3050900@qnx.com \
    --to=aristovski@qnx.com \
    --cc=gdb-patches@sourceware.org \
    --cc=palves@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).