public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: Luis Machado <luis.machado@arm.com>
To: Kuan-Ying Lee <Kuan-Ying.Lee@mediatek.com>,
	"gdb-patches@sourceware.org" <gdb-patches@sourceware.org>
Subject: Re: [PATCH, v4] [AArch64] MTE corefile support
Date: Mon, 6 Jun 2022 10:47:13 +0100	[thread overview]
Message-ID: <fd1bfc2a-7ba5-59df-867f-bb3960a90435@arm.com> (raw)
In-Reply-To: <04febcab961d9ae834ffdea960a7469a3a309b13.camel@mediatek.com>

On 6/6/22 10:42, Kuan-Ying Lee wrote:
> On Mon, 2022-06-06 at 17:28 +0800, Luis Machado via Gdb-patches wrote:
>> Ping?
>>
> 
> Hi Luis,
> 
> I don't see PT_AARCH64_MEMTAG_MTE in include/elf/common.h.
> 
> I think we need to add below in include/elf/common.h.

Based on binutils@ feedback, this constant was moved to include/elf/aarch64.h, and is
part of the binutils patch.

> 
> #define PT_AARCH64_MEMTAG_MTE (PT_LOPROC + 0x2)
> 
> Or do I miss something?
> 
> Best regards,
> Kuan-Ying Lee
> 
>> On 5/3/22 22:56, Luis Machado via Gdb-patches wrote:
>>> v4:
>>>
>>> - Updated documentation (added cross-references).
>>> - Updated the segment name from PT_ARM_MEMTAG_MTE to
>>>     PT_AARCH64_MEMTAG_MTE.
>>>
>>> v3:
>>>
>>> - Updated NEWS and documentation to be more thorough.
>>>
>>> v2:
>>>
>>> - Rework memory tag section handling to use generic section fields.
>>>
>>> --
>>>
>>> Teach GDB how to dump memory tags for AArch64 when using the gcore
>>> command
>>> and how to read memory tag data back from a core file generated by
>>> GDB
>>> (via gcore) or by the Linux kernel.
>>>
>>> The format is documented in the Linux Kernel documentation [1].
>>>
>>> Each tagged memory range (listed in /proc/<pid>/smaps) gets dumped
>>> to its
>>> own PT_AARCH64_MEMTAG_MTE segment. A section named ".memtag" is
>>> created for each
>>> of those segments when reading the core file back.
>>>
>>> To save a little bit of space, given MTE tags only take 4 bits, the
>>> memory tags
>>> are stored packed as 2 tags per byte.
>>>
>>> When reading the data back, the tags are unpacked.
>>>
>>> I've added a new testcase to exercise the feature.
>>>
>>> Build-tested with --enable-targets=all and regression tested on
>>> aarch64-linux
>>> Ubuntu 20.04.
>>>
>>> [1] Documentation/arm64/memory-tagging-extension.rst (Core Dump
>>> Support)
>>> ---
>>>    gdb/Makefile.in                              |   1 +
>>>    gdb/NEWS                                     |  10 ++
>>>    gdb/aarch64-linux-tdep.c                     | 167
>>> +++++++++++++++++++
>>>    gdb/arch/aarch64-mte-linux.c                 |  56 +++++++
>>>    gdb/arch/aarch64-mte-linux.h                 |  10 ++
>>>    gdb/corelow.c                                |  62 +++++++
>>>    gdb/defs.h                                   |   3 +-
>>>    gdb/doc/gdb.texinfo                          |  19 +++
>>>    gdb/gcore.c                                  |  83 ++++++++-
>>>    gdb/gdbarch-components.py                    |  35 ++++
>>>    gdb/gdbarch-gen.h                            |  26 +++
>>>    gdb/gdbarch.c                                |  96 +++++++++++
>>>    gdb/linux-tdep.c                             |  39 ++++-
>>>    gdb/memtag.c                                 |  61 +++++++
>>>    gdb/memtag.h                                 |  50 ++++++
>>>    gdb/testsuite/gdb.arch/aarch64-mte-gcore.c   |  93 +++++++++++
>>>    gdb/testsuite/gdb.arch/aarch64-mte-gcore.exp | 107 ++++++++++++
>>>    17 files changed, 910 insertions(+), 8 deletions(-)
>>>    create mode 100644 gdb/memtag.c
>>>    create mode 100644 gdb/memtag.h
>>>    create mode 100644 gdb/testsuite/gdb.arch/aarch64-mte-gcore.c
>>>    create mode 100644 gdb/testsuite/gdb.arch/aarch64-mte-gcore.exp
>>>
>>> diff --git a/gdb/Makefile.in b/gdb/Makefile.in
>>> index 418094775a5..fac9364bea4 100644
>>> --- a/gdb/Makefile.in
>>> +++ b/gdb/Makefile.in
>>> @@ -1120,6 +1120,7 @@ COMMON_SFILES = \
>>>    	memattr.c \
>>>    	memory-map.c \
>>>    	memrange.c \
>>> +	memtag.c \
>>>    	minidebug.c \
>>>    	minsyms.c \
>>>    	mipsread.c \
>>> diff --git a/gdb/NEWS b/gdb/NEWS
>>> index 982f4a1a18c..3d925dc3663 100644
>>> --- a/gdb/NEWS
>>> +++ b/gdb/NEWS
>>> @@ -3,6 +3,16 @@
>>>    
>>>    *** Changes since GDB 12
>>>    
>>> +* GDB now supports dumping memory tag data for AArch64 MTE.  It
>>> also supports
>>> +  reading memory tag data for AArch64 MTE from core files
>>> generated by
>>> +  the gcore command or the Linux kernel.
>>> +
>>> +  When a process uses memory-mapped pages protected by memory tags
>>> (for
>>> +  example, AArch64 MTE), this additional information will be
>>> recorded in
>>> +  the core file in the event of a crash or if GDB generates a core
>>> file
>>> +  from the current process state.  GDB will show this additional
>>> information
>>> +  automatically, or through one of the memory-tag subcommands.
>>> +
>>>    * GDB now supports hardware watchpoints on FreeBSD/Aarch64.
>>>    
>>>    * Remove support for building against Python 2, it is now only
>>> possible to
>>> diff --git a/gdb/aarch64-linux-tdep.c b/gdb/aarch64-linux-tdep.c
>>> index 55094b3d88b..12d98e71796 100644
>>> --- a/gdb/aarch64-linux-tdep.c
>>> +++ b/gdb/aarch64-linux-tdep.c
>>> @@ -53,6 +53,9 @@
>>>    
>>>    #include "gdbsupport/selftest.h"
>>>    
>>> +#include "elf/common.h"
>>> +#include "elf/aarch64.h"
>>> +
>>>    /* Signal frame handling.
>>>    
>>>          +------------+  ^
>>> @@ -1781,6 +1784,155 @@ aarch64_linux_report_signal_info (struct
>>> gdbarch *gdbarch,
>>>        }
>>>    }
>>>    
>>> +/* AArch64 Linux implementation of the
>>> gdbarch_create_memtag_section hook.  */
>>> +
>>> +static asection *
>>> +aarch64_linux_create_memtag_section (struct gdbarch *gdbarch, bfd
>>> *obfd,
>>> +				     CORE_ADDR address, size_t size)
>>> +{
>>> +  gdb_assert (obfd != nullptr);
>>> +  gdb_assert (size > 0);
>>> +
>>> +  /* Create the section and associated program header.  */
>>> +  asection *mte_section = bfd_make_section_anyway (obfd,
>>> "memtag");
>>> +
>>> +  if (mte_section == nullptr)
>>> +    return nullptr;
>>> +
>>> +  bfd_set_section_vma (mte_section, address);
>>> +  /* The size of the memory range covered by the memory tags.  We
>>> reuse the
>>> +     section's rawsize field for this purpose.  */
>>> +  mte_section->rawsize = size;
>>> +  /* Tags are stored packed as 2 tags per byte.  */
>>> +  bfd_set_section_size (mte_section, (size /
>>> AARCH64_MTE_GRANULE_SIZE) / 2);
>>> +  /* Make sure the section's flags has SEC_HAS_CONTENTS, otherwise
>>> BFD will
>>> +     refuse to write data to this section.  */
>>> +  bfd_set_section_flags (mte_section, SEC_HAS_CONTENTS);
>>> +
>>> +  /* Store program header information.  */
>>> +  bfd_record_phdr (obfd, PT_AARCH64_MEMTAG_MTE, 1, 0, 0, 0, 0, 0,
>>> 1,
>>> +		   &mte_section);
>>> +
>>> +  return mte_section;
>>> +}
>>> +
>>> +/* Maximum number of tags to request.  */
>>> +#define MAX_TAGS_TO_TRANSFER 1024
>>> +
>>> +/* AArch64 Linux implementation of the gdbarch_fill_memtag_section
>>> hook.  */
>>> +
>>> +static bool
>>> +aarch64_linux_fill_memtag_section (struct gdbarch *gdbarch,
>>> asection *osec)
>>> +{
>>> +  /* We only handle MTE tags for now.  */
>>> +
>>> +  size_t segment_size = osec->rawsize;
>>> +  CORE_ADDR start_address = bfd_section_vma (osec);
>>> +  CORE_ADDR end_address = start_address + segment_size;
>>> +
>>> +  /* Figure out how many tags we need to store in this memory
>>> range.  */
>>> +  size_t granules = aarch64_mte_get_tag_granules (start_address,
>>> segment_size,
>>> +						  AARCH64_MTE_GRANULE_S
>>> IZE);
>>> +
>>> +  /* If there are no tag granules to fetch, just return.  */
>>> +  if (granules == 0)
>>> +    return true;
>>> +
>>> +  CORE_ADDR address = start_address;
>>> +
>>> +  /* Vector of tags.  */
>>> +  gdb::byte_vector tags;
>>> +
>>> +  while (granules > 0)
>>> +    {
>>> +      /* Transfer tags in chunks.  */
>>> +      gdb::byte_vector tags_read;
>>> +      size_t xfer_len
>>> +	= (granules >= MAX_TAGS_TO_TRANSFER)?
>>> +	  MAX_TAGS_TO_TRANSFER * AARCH64_MTE_GRANULE_SIZE :
>>> +	  granules * AARCH64_MTE_GRANULE_SIZE;
>>> +
>>> +      if (!target_fetch_memtags (address, xfer_len, tags_read,
>>> +				 static_cast<int>
>>> (memtag_type::allocation)))
>>> +	{
>>> +	  warning (_("Failed to read MTE tags from memory range
>>> [%s,%s)."),
>>> +		     phex_nz (start_address, sizeof (start_address)),
>>> +		     phex_nz (end_address, sizeof (end_address)));
>>> +	  return false;
>>> +	}
>>> +
>>> +      /* Transfer over the tags that have been read.  */
>>> +      tags.insert (tags.end (), tags_read.begin (), tags_read.end
>>> ());
>>> +
>>> +      /* Adjust the remaining granules and starting address.  */
>>> +      granules -= tags_read.size ();
>>> +      address += tags_read.size () * AARCH64_MTE_GRANULE_SIZE;
>>> +    }
>>> +
>>> +  /* Pack the MTE tag bits.  */
>>> +  aarch64_mte_pack_tags (tags);
>>> +
>>> +  if (!bfd_set_section_contents (osec->owner, osec, tags.data (),
>>> +				 0, tags.size ()))
>>> +    {
>>> +      warning (_("Failed to write %s bytes of corefile memory "
>>> +		 "tag content (%s)."),
>>> +	       pulongest (tags.size ()),
>>> +	       bfd_errmsg (bfd_get_error ()));
>>> +    }
>>> +  return true;
>>> +}
>>> +
>>> +/* AArch64 Linux implementation of the
>>> gdbarch_decode_memtag_section
>>> +   hook.  Decode a memory tag section and return the requested
>>> tags.
>>> +
>>> +   The section is guaranteed to cover the [ADDRESS, ADDRESS +
>>> length)
>>> +   range.  */
>>> +
>>> +static gdb::byte_vector
>>> +aarch64_linux_decode_memtag_section (struct gdbarch *gdbarch,
>>> +				     bfd_section *section,
>>> +				     int type,
>>> +				     CORE_ADDR address, size_t length)
>>> +{
>>> +  gdb_assert (section != nullptr);
>>> +
>>> +  /* The requested address must not be less than section->vma.  */
>>> +  gdb_assert (section->vma <= address);
>>> +
>>> +  /* Figure out how many tags we need to fetch in this memory
>>> range.  */
>>> +  size_t granules = aarch64_mte_get_tag_granules (address, length,
>>> +						  AARCH64_MTE_GRANULE_S
>>> IZE);
>>> +  /* Sanity check.  */
>>> +  gdb_assert (granules > 0);
>>> +
>>> +  /* Fetch the total number of tags in the range [VMA, address +
>>> length).  */
>>> +  size_t granules_from_vma
>>> +    = aarch64_mte_get_tag_granules (section->vma,
>>> +				    address - section->vma + length,
>>> +				    AARCH64_MTE_GRANULE_SIZE);
>>> +
>>> +  /* Adjust the tags vector to contain the exact number of packed
>>> bytes.  */
>>> +  gdb::byte_vector tags (((granules - 1) >> 1) + 1);
>>> +
>>> +  /* Figure out the starting offset into the packed tags data.  */
>>> +  file_ptr offset = ((granules_from_vma - granules) >> 1);
>>> +
>>> +  if (!bfd_get_section_contents (section->owner, section,
>>> tags.data (),
>>> +				 offset, tags.size ()))
>>> +    error (_("Couldn't read contents from memtag section."));
>>> +
>>> +  /* At this point, the tags are packed 2 per byte.  Unpack them
>>> before
>>> +     returning.  */
>>> +  bool skip_first = ((granules_from_vma - granules) % 2) != 0;
>>> +  aarch64_mte_unpack_tags (tags, skip_first);
>>> +
>>> +  /* Resize to the exact number of tags that was requested.  */
>>> +  tags.resize (granules);
>>> +
>>> +  return tags;
>>> +}
>>> +
>>>    static void
>>>    aarch64_linux_init_abi (struct gdbarch_info info, struct gdbarch
>>> *gdbarch)
>>>    {
>>> @@ -1864,6 +2016,21 @@ aarch64_linux_init_abi (struct gdbarch_info
>>> info, struct gdbarch *gdbarch)
>>>    
>>>          set_gdbarch_report_signal_info (gdbarch,
>>>    				      aarch64_linux_report_signal_info)
>>> ;
>>> +
>>> +      /* Core file helpers.  */
>>> +
>>> +      /* Core file helper to create a memory tag section for a
>>> particular
>>> +	 PT_LOAD segment.  */
>>> +      set_gdbarch_create_memtag_section
>>> +	(gdbarch, aarch64_linux_create_memtag_section);
>>> +
>>> +      /* Core file helper to fill a memory tag section with tag
>>> data.  */
>>> +      set_gdbarch_fill_memtag_section
>>> +	(gdbarch, aarch64_linux_fill_memtag_section);
>>> +
>>> +      /* Core file helper to decode a memory tag section.  */
>>> +      set_gdbarch_decode_memtag_section (gdbarch,
>>> +					 aarch64_linux_decode_memtag_se
>>> ction);
>>>        }
>>>    
>>>      /* Initialize the aarch64_linux_record_tdep.  */
>>> diff --git a/gdb/arch/aarch64-mte-linux.c b/gdb/arch/aarch64-mte-
>>> linux.c
>>> index fc7a8cc00f7..3af6f364e91 100644
>>> --- a/gdb/arch/aarch64-mte-linux.c
>>> +++ b/gdb/arch/aarch64-mte-linux.c
>>> @@ -21,6 +21,62 @@
>>>    
>>>    /* See arch/aarch64-mte-linux.h */
>>>    
>>> +void
>>> +aarch64_mte_pack_tags (gdb::byte_vector &tags)
>>> +{
>>> +  /* Nothing to pack?  */
>>> +  if (tags.empty ())
>>> +    return;
>>> +
>>> +  /* If the tags vector has an odd number of elements, add another
>>> +     zeroed-out element to make it even.  This facilitates
>>> packing.  */
>>> +  if ((tags.size () % 2) != 0)
>>> +    tags.emplace_back (0);
>>> +
>>> +  for (int unpacked = 0, packed = 0; unpacked < tags.size ();
>>> +       unpacked += 2, packed++)
>>> +    tags[packed] = (tags[unpacked + 1] << 4) | tags[unpacked];
>>> +
>>> +  /* Now we have half the size.  */
>>> +  tags.resize (tags.size () / 2);
>>> +}
>>> +
>>> +/* See arch/aarch64-mte-linux.h */
>>> +
>>> +void
>>> +aarch64_mte_unpack_tags (gdb::byte_vector &tags, bool skip_first)
>>> +{
>>> +  /* Nothing to unpack?  */
>>> +  if (tags.empty ())
>>> +    return;
>>> +
>>> +  /* An unpacked MTE tags vector will have twice the number of
>>> elements
>>> +     compared to an unpacked one.  */
>>> +  gdb::byte_vector unpacked_tags (tags.size () * 2);
>>> +
>>> +  int unpacked = 0, packed = 0;
>>> +
>>> +  if (skip_first)
>>> +    {
>>> +      /* We are not interested in the first unpacked element, just
>>> discard
>>> +	 it.  */
>>> +      unpacked_tags[unpacked] = (tags[packed] >> 4) & 0xf;
>>> +      unpacked++;
>>> +      packed++;
>>> +    }
>>> +
>>> +  for (; packed < tags.size (); unpacked += 2, packed++)
>>> +    {
>>> +      unpacked_tags[unpacked] = tags[packed] & 0xf;
>>> +      unpacked_tags[unpacked + 1] = (tags[packed] >> 4) & 0xf;
>>> +    }
>>> +
>>> +  /* Update the original tags vector.  */
>>> +  tags = std::move (unpacked_tags);
>>> +}
>>> +
>>> +/* See arch/aarch64-mte-linux.h */
>>> +
>>>    size_t
>>>    aarch64_mte_get_tag_granules (CORE_ADDR addr, size_t len, size_t
>>> granule_size)
>>>    {
>>> diff --git a/gdb/arch/aarch64-mte-linux.h b/gdb/arch/aarch64-mte-
>>> linux.h
>>> index d158926feff..8a145b447aa 100644
>>> --- a/gdb/arch/aarch64-mte-linux.h
>>> +++ b/gdb/arch/aarch64-mte-linux.h
>>> @@ -32,6 +32,7 @@
>>>    
>>>    /* We have one tag per 16 bytes of memory.  */
>>>    #define AARCH64_MTE_GRANULE_SIZE 16
>>> +#define AARCH64_MTE_TAG_BIT_SIZE 4
>>>    #define AARCH64_MTE_LOGICAL_TAG_START_BIT 56
>>>    #define AARCH64_MTE_LOGICAL_MAX_VALUE 0xf
>>>    
>>> @@ -71,4 +72,13 @@ extern CORE_ADDR aarch64_mte_set_ltag (CORE_ADDR
>>> address, CORE_ADDR tag);
>>>       It is always possible to get the logical tag.  */
>>>    extern CORE_ADDR aarch64_mte_get_ltag (CORE_ADDR address);
>>>    
>>> +/* Given a TAGS vector containing 1 MTE tag per byte, pack the
>>> data as
>>> +   2 tags per byte and resize the vector.  */
>>> +void aarch64_mte_pack_tags (gdb::byte_vector &tags);
>>> +
>>> +/* Given a TAGS vector containing 2 MTE tags per byte, unpack the
>>> data as
>>> +   1 tag per byte and resize the vector.  If SKIP_FIRST is TRUE,
>>> skip the
>>> +   first unpacked element.  Otherwise leave it in the unpacked
>>> vector.  */
>>> +void aarch64_mte_unpack_tags (gdb::byte_vector &tags, bool
>>> skip_first);
>>> +
>>>    #endif /* ARCH_AARCH64_LINUX_H */
>>> diff --git a/gdb/corelow.c b/gdb/corelow.c
>>> index 8c33fb7ebb2..8b8994f80db 100644
>>> --- a/gdb/corelow.c
>>> +++ b/gdb/corelow.c
>>> @@ -52,6 +52,7 @@
>>>    #include <unordered_set>
>>>    #include "gdbcmd.h"
>>>    #include "xml-tdesc.h"
>>> +#include "memtag.h"
>>>    
>>>    #ifndef O_LARGEFILE
>>>    #define O_LARGEFILE 0
>>> @@ -101,6 +102,13 @@ class core_target final : public
>>> process_stratum_target
>>>    
>>>      bool info_proc (const char *, enum info_proc_what) override;
>>>    
>>> +  bool supports_memory_tagging () override;
>>> +
>>> +  /* Core file implementation of fetch_memtags.  Fetch the memory
>>> tags from
>>> +     core file notes.  */
>>> +  bool fetch_memtags (CORE_ADDR address, size_t len,
>>> +		      gdb::byte_vector &tags, int type) override;
>>> +
>>>      /* A few helpers.  */
>>>    
>>>      /* Getter, see variable definition.  */
>>> @@ -1162,6 +1170,60 @@ core_target::info_proc (const char *args,
>>> enum info_proc_what request)
>>>      return true;
>>>    }
>>>    
>>> +/* Implementation of the "supports_memory_tagging" target_ops
>>> method.  */
>>> +
>>> +bool
>>> +core_target::supports_memory_tagging ()
>>> +{
>>> +  /* Look for memory tag sections.  If they exist, that means this
>>> core file
>>> +     supports memory tagging.  */
>>> +
>>> +  return (bfd_get_section_by_name (core_bfd, "memtag") !=
>>> nullptr);
>>> +}
>>> +
>>> +/* Implementation of the "fetch_memtags" target_ops method.  */
>>> +
>>> +bool
>>> +core_target::fetch_memtags (CORE_ADDR address, size_t len,
>>> +			    gdb::byte_vector &tags, int type)
>>> +{
>>> +  struct gdbarch *gdbarch = target_gdbarch ();
>>> +
>>> +  /* Make sure we have a way to decode the memory tag notes.  */
>>> +  if (!gdbarch_decode_memtag_section_p (gdbarch))
>>> +    error (_("gdbarch_decode_memtag_section not implemented for
>>> this "
>>> +	     "architecture."));
>>> +
>>> +  memtag_section_info info;
>>> +  info.memtag_section = nullptr;
>>> +
>>> +  while (get_next_core_memtag_section (core_bfd,
>>> info.memtag_section,
>>> +				       address, info))
>>> +  {
>>> +    size_t adjusted_length
>>> +      = (address + len < info.end_address)? len :
>>> (info.end_address - address);
>>> +
>>> +    /* Decode the memory tag note and return the tags.  */
>>> +    gdb::byte_vector tags_read
>>> +      = gdbarch_decode_memtag_section (gdbarch,
>>> info.memtag_section, type,
>>> +				       address, adjusted_length);
>>> +
>>> +    /* Transfer over the tags that have been read.  */
>>> +    tags.insert (tags.end (), tags_read.begin (), tags_read.end
>>> ());
>>> +
>>> +    /* ADDRESS + LEN may cross the boundaries of a particular
>>> memory tag
>>> +       segment.  Check if we need to fetch tags from a different
>>> section.  */
>>> +    if (!tags_read.empty () && (address + len) < info.end_address)
>>> +      return true;
>>> +
>>> +    /* There are more tags to fetch.  Update ADDRESS and LEN.  */
>>> +    len -= (info.end_address - address);
>>> +    address = info.end_address;
>>> +  }
>>> +
>>> +  return false;
>>> +}
>>> +
>>>    /* Get a pointer to the current core target.  If not connected to
>>> a
>>>       core target, return NULL.  */
>>>    
>>> diff --git a/gdb/defs.h b/gdb/defs.h
>>> index 99bfdd526ff..51a7576a56a 100644
>>> --- a/gdb/defs.h
>>> +++ b/gdb/defs.h
>>> @@ -344,7 +344,8 @@ extern const char *pc_prefix (CORE_ADDR);
>>>    
>>>    typedef int (*find_memory_region_ftype) (CORE_ADDR addr, unsigned
>>> long size,
>>>    					 int read, int write, int exec,
>>> -					 int modified, void *data);
>>> +					 int modified, bool
>>> memory_tagged,
>>> +					 void *data);
>>>    
>>>    /* * Possible lvalue types.  Like enum language, this should be
>>> in
>>>       value.h, but needs to be here for the same reason.  */
>>> diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
>>> index 38ad2ac32b0..36f10f20cfb 100644
>>> --- a/gdb/doc/gdb.texinfo
>>> +++ b/gdb/doc/gdb.texinfo
>>> @@ -25555,6 +25555,25 @@ options that can be controlled at runtime
>>> and emulates the @code{prctl}
>>>    option @code{PR_SET_TAGGED_ADDR_CTRL}.  For further information,
>>> see the
>>>    documentation in the Linux kernel.
>>>    
>>> +@value{GDBN} supports dumping memory tag data to core files
>>> through the
>>> +@command{gcore} command and reading memory tag data from core
>>> files generated
>>> +by the @command{gcore} command or the Linux kernel.
>>> +
>>> +When a process uses memory-mapped pages protected by memory tags
>>> (for
>>> +example, AArch64 MTE), this additional information will be
>>> recorded in
>>> +the core file in the event of a crash or if @value{GDBN} generates
>>> a core file
>>> +from the current process state.
>>> +
>>> +The memory tag data will be used so developers can display the
>>> memory
>>> +tags from a particular memory region (using the @samp{m} modifier
>>> to the
>>> +@command{x} command, using the @command{print} command or using
>>> the various
>>> +@command{memory-tag} subcommands.
>>> +
>>> +In the case of a crash, @value{GDBN} will attempt to retrieve the
>>> memory tag
>>> +information automatically from the core file, and will show one of
>>> the above
>>> +messages depending on whether the synchronous or asynchronous mode
>>> is selected.
>>> +@xref{Memory Tagging}. @xref{Memory}.
>>> +
>>>    @node i386
>>>    @subsection x86 Architecture-specific Issues
>>>    
>>> diff --git a/gdb/gcore.c b/gdb/gcore.c
>>> index fdb22b72a07..b81ef81ab84 100644
>>> --- a/gdb/gcore.c
>>> +++ b/gdb/gcore.c
>>> @@ -349,6 +349,12 @@ make_output_phdrs (bfd *obfd, asection *osec)
>>>      int p_flags = 0;
>>>      int p_type = 0;
>>>    
>>> +  /* Memory tag segments have already been handled by the
>>> architecture, as
>>> +     those contain arch-specific information.  If we have one of
>>> those, just
>>> +     return.  */
>>> +  if (startswith (bfd_section_name (osec), "memtag"))
>>> +    return;
>>> +
>>>      /* FIXME: these constants may only be applicable for ELF.  */
>>>      if (startswith (bfd_section_name (osec), "load"))
>>>        p_type = PT_LOAD;
>>> @@ -371,7 +377,8 @@ make_output_phdrs (bfd *obfd, asection *osec)
>>>    
>>>    static int
>>>    gcore_create_callback (CORE_ADDR vaddr, unsigned long size, int
>>> read,
>>> -		       int write, int exec, int modified, void *data)
>>> +		       int write, int exec, int modified, bool
>>> memory_tagged,
>>> +		       void *data)
>>>    {
>>>      bfd *obfd = (bfd *) data;
>>>      asection *osec;
>>> @@ -454,6 +461,45 @@ gcore_create_callback (CORE_ADDR vaddr,
>>> unsigned long size, int read,
>>>      return 0;
>>>    }
>>>    
>>> +/* gdbarch_find_memory_region callback for creating a memory tag
>>> section.
>>> +   DATA is 'bfd *' for the core file GDB is creating.  */
>>> +
>>> +static int
>>> +gcore_create_memtag_section_callback (CORE_ADDR vaddr, unsigned
>>> long size,
>>> +				      int read, int write, int exec,
>>> +				      int modified, bool memory_tagged,
>>> +				      void *data)
>>> +{
>>> +  /* Are there memory tags in this particular memory map
>>> entry?  */
>>> +  if (!memory_tagged)
>>> +    return 0;
>>> +
>>> +  bfd *obfd = (bfd *) data;
>>> +
>>> +  /* Ask the architecture to create a memory tag section for this
>>> particular
>>> +     memory map entry.  It will be populated with contents later,
>>> as we can't
>>> +     start writing the contents before we have all the sections
>>> sorted out.  */
>>> +  asection *memtag_section
>>> +    = gdbarch_create_memtag_section (target_gdbarch (), obfd,
>>> vaddr, size);
>>> +
>>> +  if (memtag_section == nullptr)
>>> +    {
>>> +      warning (_("Couldn't make gcore memory tag segment: %s"),
>>> +	       bfd_errmsg (bfd_get_error ()));
>>> +      return 1;
>>> +    }
>>> +
>>> +  if (info_verbose)
>>> +    {
>>> +      gdb_printf (gdb_stdout, "Saved memory tag segment, %s bytes
>>> "
>>> +			      "at %s\n",
>>> +		  plongest (bfd_section_size (memtag_section)),
>>> +		  paddress (target_gdbarch (), vaddr));
>>> +    }
>>> +
>>> +  return 0;
>>> +}
>>> +
>>>    int
>>>    objfile_find_memory_regions (struct target_ops *self,
>>>    			     find_memory_region_ftype func, void *obfd)
>>> @@ -483,6 +529,7 @@ objfile_find_memory_regions (struct target_ops
>>> *self,
>>>    			   (flags & SEC_READONLY) == 0, /*
>>> Writable.  */
>>>    			   (flags & SEC_CODE) != 0, /* Executable.  */
>>>    			   1, /* MODIFIED is unknown, pass it as
>>> true.  */
>>> +			   false, /* No memory tags in the object
>>> file.  */
>>>    			   obfd);
>>>    	    if (ret != 0)
>>>    	      return ret;
>>> @@ -496,6 +543,7 @@ objfile_find_memory_regions (struct target_ops
>>> *self,
>>>    	     1, /* Stack section will be writable.  */
>>>    	     0, /* Stack section will not be executable.  */
>>>    	     1, /* Stack section will be modified.  */
>>> +	     false, /* No memory tags in the object file.  */
>>>    	     obfd);
>>>    
>>>      /* Make a heap segment.  */
>>> @@ -506,6 +554,7 @@ objfile_find_memory_regions (struct target_ops
>>> *self,
>>>    	     1, /* Heap section will be writable.  */
>>>    	     0, /* Heap section will not be executable.  */
>>>    	     1, /* Heap section will be modified.  */
>>> +	     false, /* No memory tags in the object file.  */
>>>    	     obfd);
>>>    
>>>      return 0;
>>> @@ -555,6 +604,20 @@ gcore_copy_callback (bfd *obfd, asection
>>> *osec)
>>>        }
>>>    }
>>>    
>>> +/* Callback to copy contents to a particular memory tag
>>> section.  */
>>> +
>>> +static void
>>> +gcore_copy_memtag_section_callback (bfd *obfd, asection *osec)
>>> +{
>>> +  /* We are only interested in "memtag" sections.  */
>>> +  if (!startswith (bfd_section_name (osec), "memtag"))
>>> +    return;
>>> +
>>> +  /* Fill the section with memory tag contents.  */
>>> +  if (!gdbarch_fill_memtag_section (target_gdbarch (), osec))
>>> +    error (_("Failed to fill memory tag section for core file."));
>>> +}
>>> +
>>>    static int
>>>    gcore_memory_sections (bfd *obfd)
>>>    {
>>> @@ -567,13 +630,27 @@ gcore_memory_sections (bfd *obfd)
>>>    	return 0;			/* FIXME: error return/msg?  */
>>>        }
>>>    
>>> +  /* Take care of dumping memory tags, if there are any.  */
>>> +  if (!gdbarch_find_memory_regions_p (target_gdbarch ())
>>> +      || gdbarch_find_memory_regions (target_gdbarch (),
>>> +				      gcore_create_memtag_section_callb
>>> ack,
>>> +				      obfd) != 0)
>>> +    {
>>> +      if (target_find_memory_regions
>>> (gcore_create_memtag_section_callback,
>>> +				      obfd) != 0)
>>> +	return 0;
>>> +    }
>>> +
>>>      /* Record phdrs for section-to-segment mapping.  */
>>>      for (asection *sect : gdb_bfd_sections (obfd))
>>>        make_output_phdrs (obfd, sect);
>>>    
>>> -  /* Copy memory region contents.  */
>>> +  /* Copy memory region and memory tag contents.  */
>>>      for (asection *sect : gdb_bfd_sections (obfd))
>>> -    gcore_copy_callback (obfd, sect);
>>> +    {
>>> +      gcore_copy_callback (obfd, sect);
>>> +      gcore_copy_memtag_section_callback (obfd, sect);
>>> +    }
>>>    
>>>      return 1;
>>>    }
>>> diff --git a/gdb/gdbarch-components.py b/gdb/gdbarch-components.py
>>> index e8f20c83ff0..6fa1b7591db 100644
>>> --- a/gdb/gdbarch-components.py
>>> +++ b/gdb/gdbarch-components.py
>>> @@ -1522,6 +1522,41 @@ Find core file memory regions
>>>        invalid=True,
>>>    )
>>>    
>>> +Method(
>>> +    comment="""
>>> +Given a bfd OBFD, segment ADDRESS and SIZE, create a memory tag
>>> section to be dumped to a core file
>>> +""",
>>> +    type="asection *",
>>> +    name="create_memtag_section",
>>> +    params=[("bfd *", "obfd"), ("CORE_ADDR", "address"),
>>> ("size_t", "size")],
>>> +    predicate=True,
>>> +    invalid=True,
>>> +)
>>> +
>>> +Method(
>>> +    comment="""
>>> +Given a memory tag section OSEC, fill OSEC's contents with the
>>> appropriate tag data
>>> +""",
>>> +    type="bool",
>>> +    name="fill_memtag_section",
>>> +    params=[("asection *", "osec")],
>>> +    predicate=True,
>>> +    invalid=True,
>>> +)
>>> +
>>> +Method(
>>> +    comment="""
>>> +Decode a memory tag SECTION and return the tags of type TYPE
>>> contained in
>>> +the memory range [ADDRESS, ADDRESS + LENGTH).
>>> +If no tags were found, return an empty vector.
>>> +""",
>>> +    type="gdb::byte_vector",
>>> +    name="decode_memtag_section",
>>> +    params=[("bfd_section *", "section"), ("int", "type"),
>>> ("CORE_ADDR", "address"), ("size_t", "length")],
>>> +    predicate=True,
>>> +    invalid=True,
>>> +)
>>> +
>>>    Method(
>>>        comment="""
>>>    Read offset OFFSET of TARGET_OBJECT_LIBRARIES formatted shared
>>> libraries list from
>>> diff --git a/gdb/gdbarch-gen.h b/gdb/gdbarch-gen.h
>>> index 882b9057b1a..1d19f51f21d 100644
>>> --- a/gdb/gdbarch-gen.h
>>> +++ b/gdb/gdbarch-gen.h
>>> @@ -874,6 +874,32 @@ typedef int
>>> (gdbarch_find_memory_regions_ftype) (struct gdbarch *gdbarch,
>>> find_m
>>>    extern int gdbarch_find_memory_regions (struct gdbarch *gdbarch,
>>> find_memory_region_ftype func, void *data);
>>>    extern void set_gdbarch_find_memory_regions (struct gdbarch
>>> *gdbarch, gdbarch_find_memory_regions_ftype *find_memory_regions);
>>>    
>>> +/* Given a bfd OBFD, segment ADDRESS and SIZE, create a memory tag
>>> section to be dumped to a core file */
>>> +
>>> +extern bool gdbarch_create_memtag_section_p (struct gdbarch
>>> *gdbarch);
>>> +
>>> +typedef asection * (gdbarch_create_memtag_section_ftype) (struct
>>> gdbarch *gdbarch, bfd *obfd, CORE_ADDR address, size_t size);
>>> +extern asection * gdbarch_create_memtag_section (struct gdbarch
>>> *gdbarch, bfd *obfd, CORE_ADDR address, size_t size);
>>> +extern void set_gdbarch_create_memtag_section (struct gdbarch
>>> *gdbarch, gdbarch_create_memtag_section_ftype
>>> *create_memtag_section);
>>> +
>>> +/* Given a memory tag section OSEC, fill OSEC's contents with the
>>> appropriate tag data */
>>> +
>>> +extern bool gdbarch_fill_memtag_section_p (struct gdbarch
>>> *gdbarch);
>>> +
>>> +typedef bool (gdbarch_fill_memtag_section_ftype) (struct gdbarch
>>> *gdbarch, asection *osec);
>>> +extern bool gdbarch_fill_memtag_section (struct gdbarch *gdbarch,
>>> asection *osec);
>>> +extern void set_gdbarch_fill_memtag_section (struct gdbarch
>>> *gdbarch, gdbarch_fill_memtag_section_ftype *fill_memtag_section);
>>> +
>>> +/* Decode a memory tag SECTION and return the tags of type TYPE
>>> contained in
>>> +   the memory range [ADDRESS, ADDRESS + LENGTH).
>>> +   If no tags were found, return an empty vector. */
>>> +
>>> +extern bool gdbarch_decode_memtag_section_p (struct gdbarch
>>> *gdbarch);
>>> +
>>> +typedef gdb::byte_vector (gdbarch_decode_memtag_section_ftype)
>>> (struct gdbarch *gdbarch, bfd_section *section, int type, CORE_ADDR
>>> address, size_t length);
>>> +extern gdb::byte_vector gdbarch_decode_memtag_section (struct
>>> gdbarch *gdbarch, bfd_section *section, int type, CORE_ADDR
>>> address, size_t length);
>>> +extern void set_gdbarch_decode_memtag_section (struct gdbarch
>>> *gdbarch, gdbarch_decode_memtag_section_ftype
>>> *decode_memtag_section);
>>> +
>>>    /* Read offset OFFSET of TARGET_OBJECT_LIBRARIES formatted shared
>>> libraries list from
>>>       core file into buffer READBUF with length LEN.  Return the
>>> number of bytes read
>>>       (zero indicates failure).
>>> diff --git a/gdb/gdbarch.c b/gdb/gdbarch.c
>>> index a588bdef61a..f5dbacb14e7 100644
>>> --- a/gdb/gdbarch.c
>>> +++ b/gdb/gdbarch.c
>>> @@ -171,6 +171,9 @@ struct gdbarch
>>>      gdbarch_iterate_over_regset_sections_ftype
>>> *iterate_over_regset_sections;
>>>      gdbarch_make_corefile_notes_ftype *make_corefile_notes;
>>>      gdbarch_find_memory_regions_ftype *find_memory_regions;
>>> +  gdbarch_create_memtag_section_ftype *create_memtag_section;
>>> +  gdbarch_fill_memtag_section_ftype *fill_memtag_section;
>>> +  gdbarch_decode_memtag_section_ftype *decode_memtag_section;
>>>      gdbarch_core_xfer_shared_libraries_ftype
>>> *core_xfer_shared_libraries;
>>>      gdbarch_core_xfer_shared_libraries_aix_ftype
>>> *core_xfer_shared_libraries_aix;
>>>      gdbarch_core_pid_to_str_ftype *core_pid_to_str;
>>> @@ -527,6 +530,9 @@ verify_gdbarch (struct gdbarch *gdbarch)
>>>      /* Skip verify of iterate_over_regset_sections, has
>>> predicate.  */
>>>      /* Skip verify of make_corefile_notes, has predicate.  */
>>>      /* Skip verify of find_memory_regions, has predicate.  */
>>> +  /* Skip verify of create_memtag_section, has predicate.  */
>>> +  /* Skip verify of fill_memtag_section, has predicate.  */
>>> +  /* Skip verify of decode_memtag_section, has predicate.  */
>>>      /* Skip verify of core_xfer_shared_libraries, has
>>> predicate.  */
>>>      /* Skip verify of core_xfer_shared_libraries_aix, has
>>> predicate.  */
>>>      /* Skip verify of core_pid_to_str, has predicate.  */
>>> @@ -1096,6 +1102,24 @@ gdbarch_dump (struct gdbarch *gdbarch,
>>> struct ui_file *file)
>>>      gdb_printf (file,
>>>                          "gdbarch_dump: find_memory_regions =
>>> <%s>\n",
>>>                          host_address_to_string (gdbarch-
>>>> find_memory_regions));
>>> +  gdb_printf (file,
>>> +                      "gdbarch_dump:
>>> gdbarch_create_memtag_section_p() = %d\n",
>>> +                      gdbarch_create_memtag_section_p (gdbarch));
>>> +  gdb_printf (file,
>>> +                      "gdbarch_dump: create_memtag_section =
>>> <%s>\n",
>>> +                      host_address_to_string (gdbarch-
>>>> create_memtag_section));
>>> +  gdb_printf (file,
>>> +                      "gdbarch_dump:
>>> gdbarch_fill_memtag_section_p() = %d\n",
>>> +                      gdbarch_fill_memtag_section_p (gdbarch));
>>> +  gdb_printf (file,
>>> +                      "gdbarch_dump: fill_memtag_section =
>>> <%s>\n",
>>> +                      host_address_to_string (gdbarch-
>>>> fill_memtag_section));
>>> +  gdb_printf (file,
>>> +                      "gdbarch_dump:
>>> gdbarch_decode_memtag_section_p() = %d\n",
>>> +                      gdbarch_decode_memtag_section_p (gdbarch));
>>> +  gdb_printf (file,
>>> +                      "gdbarch_dump: decode_memtag_section =
>>> <%s>\n",
>>> +                      host_address_to_string (gdbarch-
>>>> decode_memtag_section));
>>>      gdb_printf (file,
>>>                          "gdbarch_dump:
>>> gdbarch_core_xfer_shared_libraries_p() = %d\n",
>>>                          gdbarch_core_xfer_shared_libraries_p
>>> (gdbarch));
>>> @@ -3744,6 +3768,78 @@ set_gdbarch_find_memory_regions (struct
>>> gdbarch *gdbarch,
>>>      gdbarch->find_memory_regions = find_memory_regions;
>>>    }
>>>    
>>> +bool
>>> +gdbarch_create_memtag_section_p (struct gdbarch *gdbarch)
>>> +{
>>> +  gdb_assert (gdbarch != NULL);
>>> +  return gdbarch->create_memtag_section != NULL;
>>> +}
>>> +
>>> +asection *
>>> +gdbarch_create_memtag_section (struct gdbarch *gdbarch, bfd *obfd,
>>> CORE_ADDR address, size_t size)
>>> +{
>>> +  gdb_assert (gdbarch != NULL);
>>> +  gdb_assert (gdbarch->create_memtag_section != NULL);
>>> +  if (gdbarch_debug >= 2)
>>> +    gdb_printf (gdb_stdlog, "gdbarch_create_memtag_section
>>> called\n");
>>> +  return gdbarch->create_memtag_section (gdbarch, obfd, address,
>>> size);
>>> +}
>>> +
>>> +void
>>> +set_gdbarch_create_memtag_section (struct gdbarch *gdbarch,
>>> +                                   gdbarch_create_memtag_section_f
>>> type create_memtag_section)
>>> +{
>>> +  gdbarch->create_memtag_section = create_memtag_section;
>>> +}
>>> +
>>> +bool
>>> +gdbarch_fill_memtag_section_p (struct gdbarch *gdbarch)
>>> +{
>>> +  gdb_assert (gdbarch != NULL);
>>> +  return gdbarch->fill_memtag_section != NULL;
>>> +}
>>> +
>>> +bool
>>> +gdbarch_fill_memtag_section (struct gdbarch *gdbarch, asection
>>> *osec)
>>> +{
>>> +  gdb_assert (gdbarch != NULL);
>>> +  gdb_assert (gdbarch->fill_memtag_section != NULL);
>>> +  if (gdbarch_debug >= 2)
>>> +    gdb_printf (gdb_stdlog, "gdbarch_fill_memtag_section
>>> called\n");
>>> +  return gdbarch->fill_memtag_section (gdbarch, osec);
>>> +}
>>> +
>>> +void
>>> +set_gdbarch_fill_memtag_section (struct gdbarch *gdbarch,
>>> +                                 gdbarch_fill_memtag_section_ftype
>>> fill_memtag_section)
>>> +{
>>> +  gdbarch->fill_memtag_section = fill_memtag_section;
>>> +}
>>> +
>>> +bool
>>> +gdbarch_decode_memtag_section_p (struct gdbarch *gdbarch)
>>> +{
>>> +  gdb_assert (gdbarch != NULL);
>>> +  return gdbarch->decode_memtag_section != NULL;
>>> +}
>>> +
>>> +gdb::byte_vector
>>> +gdbarch_decode_memtag_section (struct gdbarch *gdbarch,
>>> bfd_section *section, int type, CORE_ADDR address, size_t length)
>>> +{
>>> +  gdb_assert (gdbarch != NULL);
>>> +  gdb_assert (gdbarch->decode_memtag_section != NULL);
>>> +  if (gdbarch_debug >= 2)
>>> +    gdb_printf (gdb_stdlog, "gdbarch_decode_memtag_section
>>> called\n");
>>> +  return gdbarch->decode_memtag_section (gdbarch, section, type,
>>> address, length);
>>> +}
>>> +
>>> +void
>>> +set_gdbarch_decode_memtag_section (struct gdbarch *gdbarch,
>>> +                                   gdbarch_decode_memtag_section_f
>>> type decode_memtag_section)
>>> +{
>>> +  gdbarch->decode_memtag_section = decode_memtag_section;
>>> +}
>>> +
>>>    bool
>>>    gdbarch_core_xfer_shared_libraries_p (struct gdbarch *gdbarch)
>>>    {
>>> diff --git a/gdb/linux-tdep.c b/gdb/linux-tdep.c
>>> index 4e728a06e7e..8a83ed320cf 100644
>>> --- a/gdb/linux-tdep.c
>>> +++ b/gdb/linux-tdep.c
>>> @@ -42,6 +42,7 @@
>>>    #include "gcore.h"
>>>    #include "gcore-elf.h"
>>>    #include "solib-svr4.h"
>>> +#include "memtag.h"
>>>    
>>>    #include <ctype.h>
>>>    #include <unordered_map>
>>> @@ -1320,6 +1321,7 @@ typedef int linux_find_memory_region_ftype
>>> (ULONGEST vaddr, ULONGEST size,
>>>    					    ULONGEST offset, ULONGEST
>>> inode,
>>>    					    int read, int write,
>>>    					    int exec, int modified,
>>> +					    bool memory_tagged,
>>>    					    const char *filename,
>>>    					    void *data);
>>>    
>>> @@ -1470,10 +1472,11 @@ parse_smaps_data (const char *data,
>>>      return smaps;
>>>    }
>>>    
>>> -/* See linux-tdep.h.  */
>>> +/* Helper that checks if an address is in a memory tag page for a
>>> live
>>> +   process.  */
>>>    
>>> -bool
>>> -linux_address_in_memtag_page (CORE_ADDR address)
>>> +static bool
>>> +linux_process_address_in_memtag_page (CORE_ADDR address)
>>>    {
>>>      if (current_inferior ()->fake_pid_p)
>>>        return false;
>>> @@ -1505,6 +1508,30 @@ linux_address_in_memtag_page (CORE_ADDR
>>> address)
>>>      return false;
>>>    }
>>>    
>>> +/* Helper that checks if an address is in a memory tag page for a
>>> core file
>>> +   process.  */
>>> +
>>> +static bool
>>> +linux_core_file_address_in_memtag_page (CORE_ADDR address)
>>> +{
>>> +  if (core_bfd == nullptr)
>>> +    return false;
>>> +
>>> +  memtag_section_info info;
>>> +  return get_next_core_memtag_section (core_bfd, nullptr, address,
>>> info);
>>> +}
>>> +
>>> +/* See linux-tdep.h.  */
>>> +
>>> +bool
>>> +linux_address_in_memtag_page (CORE_ADDR address)
>>> +{
>>> +  if (!target_has_execution ())
>>> +    return linux_core_file_address_in_memtag_page (address);
>>> +
>>> +  return linux_process_address_in_memtag_page (address);
>>> +}
>>> +
>>>    /* List memory regions in the inferior for a corefile.  */
>>>    
>>>    static int
>>> @@ -1593,6 +1620,7 @@ linux_find_memory_regions_full (struct
>>> gdbarch *gdbarch,
>>>    		map.offset, map.inode, map.read, map.write, map.exec,
>>>    		1, /* MODIFIED is true because we want to dump
>>>    		      the mapping.  */
>>> +		map.vmflags.memory_tagging != 0,
>>>    		map.filename.c_str (), obfd);
>>>    	}
>>>        }
>>> @@ -1621,12 +1649,14 @@ static int
>>>    linux_find_memory_regions_thunk (ULONGEST vaddr, ULONGEST size,
>>>    				 ULONGEST offset, ULONGEST inode,
>>>    				 int read, int write, int exec, int
>>> modified,
>>> +				 bool memory_tagged,
>>>    				 const char *filename, void *arg)
>>>    {
>>>      struct linux_find_memory_regions_data *data
>>>        = (struct linux_find_memory_regions_data *) arg;
>>>    
>>> -  return data->func (vaddr, size, read, write, exec, modified,
>>> data->obfd);
>>> +  return data->func (vaddr, size, read, write, exec, modified,
>>> memory_tagged,
>>> +		     data->obfd);
>>>    }
>>>    
>>>    /* A variant of linux_find_memory_regions_full that is suitable
>>> as the
>>> @@ -1675,6 +1705,7 @@ static int
>>>    linux_make_mappings_callback (ULONGEST vaddr, ULONGEST size,
>>>    			      ULONGEST offset, ULONGEST inode,
>>>    			      int read, int write, int exec, int
>>> modified,
>>> +			      bool memory_tagged,
>>>    			      const char *filename, void *data)
>>>    {
>>>      struct linux_make_mappings_data *map_data
>>> diff --git a/gdb/memtag.c b/gdb/memtag.c
>>> new file mode 100644
>>> index 00000000000..af86137c49d
>>> --- /dev/null
>>> +++ b/gdb/memtag.c
>>> @@ -0,0 +1,61 @@
>>> +/* GDB generic memory tagging functions.
>>> +
>>> +   Copyright (C) 2022 Free Software Foundation, Inc.
>>> +
>>> +   This file is part of GDB.
>>> +
>>> +   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 <
>>> https://urldefense.com/v3/__http://www.gnu.org/licenses/__;!!CTRNKA9wMg0ARbw!0jBVgsoE-RaTGT098YRLfC3gcQJlvfuMujbUqmtqoh3JjeD2BI4JNxmWCGaefncNfKBbmw$
>>>   >.  */
>>> +
>>> +#include "defs.h"
>>> +#include "memtag.h"
>>> +#include "bfd.h"
>>> +
>>> +/* See memtag.h */
>>> +
>>> +bool
>>> +get_next_core_memtag_section (bfd *abfd, asection *section,
>>> +			      CORE_ADDR address, memtag_section_info
>>> &info)
>>> +{
>>> +  /* If the caller provided no SECTION to start from, search from
>>> the
>>> +     beginning.  */
>>> +  if (section == nullptr)
>>> +    section = bfd_get_section_by_name (abfd, "memtag");
>>> +
>>> +  /* Go through all the memtag sections and figure out if ADDRESS
>>> +     falls within one of the memory ranges that contain tags.  */
>>> +  while (section != nullptr)
>>> +    {
>>> +      size_t memtag_range_size = section->rawsize;
>>> +      size_t tags_size = bfd_section_size (section);
>>> +
>>> +      /* Empty memory range and empty tag dump should not
>>> happen.  */
>>> +      gdb_assert (memtag_range_size != 0);
>>> +      gdb_assert (tags_size != 0);
>>> +
>>> +      CORE_ADDR start_address = bfd_section_vma (section);
>>> +      CORE_ADDR end_address = start_address + memtag_range_size;
>>> +
>>> +      /* Is the address within [start_address, end_address)?  */
>>> +      if (address >= start_address
>>> +	  && address < end_address)
>>> +	{
>>> +	  info.start_address = start_address;
>>> +	  info.end_address = end_address;
>>> +	  info.memtag_section = section;
>>> +	  return true;
>>> +	}
>>> +      section = bfd_get_next_section_by_name (abfd, section);
>>> +    }
>>> +  return false;
>>> +}
>>> diff --git a/gdb/memtag.h b/gdb/memtag.h
>>> new file mode 100644
>>> index 00000000000..fe908c1e5e3
>>> --- /dev/null
>>> +++ b/gdb/memtag.h
>>> @@ -0,0 +1,50 @@
>>> +/* GDB generic memory tagging definitions.
>>> +   Copyright (C) 2022 Free Software Foundation, Inc.
>>> +
>>> +   This file is part of GDB.
>>> +
>>> +   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 <
>>> https://urldefense.com/v3/__http://www.gnu.org/licenses/__;!!CTRNKA9wMg0ARbw!0jBVgsoE-RaTGT098YRLfC3gcQJlvfuMujbUqmtqoh3JjeD2BI4JNxmWCGaefncNfKBbmw$
>>>   >.  */
>>> +
>>> +#ifndef MEMTAG_H
>>> +#define MEMTAG_H
>>> +
>>> +#include "bfd.h"
>>> +
>>> +struct memtag_section_info
>>> +{
>>> +  /* The start address of the tagged memory range.  */
>>> +  CORE_ADDR start_address;
>>> +  /* The final address of the tagged memory range.  */
>>> +  CORE_ADDR end_address;
>>> +  /* The section containing tags for the memory range
>>> +     [start_address, end_address).  */
>>> +  asection *memtag_section;
>>> +};
>>> +
>>> +/* Helper function to walk through memory tag sections in a core
>>> file.
>>> +
>>> +   Return TRUE if there is a "memtag" section containing
>>> ADDRESS.  Return FALSE
>>> +   otherwise.
>>> +
>>> +   If SECTION is provided, search from that section onwards. If
>>> SECTION is
>>> +   nullptr, then start a new search.
>>> +
>>> +   If a "memtag" section containing ADDRESS is found, fill INFO
>>> with data
>>> +   about such section.  Otherwise leave it unchanged.  */
>>> +
>>> +bool get_next_core_memtag_section (bfd *abfd, asection *section,
>>> +				   CORE_ADDR address,
>>> +				   memtag_section_info &info);
>>> +
>>> +#endif /* MEMTAG_H */
>>> diff --git a/gdb/testsuite/gdb.arch/aarch64-mte-gcore.c
>>> b/gdb/testsuite/gdb.arch/aarch64-mte-gcore.c
>>> new file mode 100644
>>> index 00000000000..b20ebcff424
>>> --- /dev/null
>>> +++ b/gdb/testsuite/gdb.arch/aarch64-mte-gcore.c
>>> @@ -0,0 +1,93 @@
>>> +/* This test program is part of GDB, the GNU debugger.
>>> +
>>> +   Copyright 2021 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 <
>>> https://urldefense.com/v3/__http://www.gnu.org/licenses/__;!!CTRNKA9wMg0ARbw!0jBVgsoE-RaTGT098YRLfC3gcQJlvfuMujbUqmtqoh3JjeD2BI4JNxmWCGaefncNfKBbmw$
>>>   >.  */
>>> +
>>> +/* Exercise AArch64's Memory Tagging Extension with tagged
>>> pointers.  */
>>> +
>>> +/* This test was based on the documentation for the AArch64 Memory
>>> Tagging
>>> +   Extension from the Linux Kernel, found in the sources in
>>> +   Documentation/arm64/memory-tagging-extension.rst.  */
>>> +
>>> +#include <errno.h>
>>> +#include <stdio.h>
>>> +#include <stdlib.h>
>>> +#include <unistd.h>
>>> +#include <sys/auxv.h>
>>> +#include <sys/mman.h>
>>> +#include <sys/prctl.h>
>>> +
>>> +/* From arch/arm64/include/uapi/asm/hwcap.h */
>>> +#define HWCAP2_MTE              (1 << 18)
>>> +
>>> +/* From arch/arm64/include/uapi/asm/mman.h */
>>> +#define PROT_MTE  0x20
>>> +
>>> +/* From include/uapi/linux/prctl.h */
>>> +#define PR_SET_TAGGED_ADDR_CTRL 55
>>> +#define PR_GET_TAGGED_ADDR_CTRL 56
>>> +#define PR_TAGGED_ADDR_ENABLE	(1UL << 0)
>>> +#define PR_MTE_TCF_SHIFT	1
>>> +#define PR_MTE_TCF_SYNC		(1UL << PR_MTE_TCF_SHIFT)
>>> +#define PR_MTE_TAG_SHIFT	3
>>> +
>>> +void
>>> +access_memory (unsigned char *tagged_ptr)
>>> +{
>>> +  tagged_ptr[0] = 'a';
>>> +}
>>> +
>>> +int
>>> +main (int argc, char **argv)
>>> +{
>>> +  unsigned char *tagged_ptr;
>>> +  unsigned long page_sz = sysconf (_SC_PAGESIZE);
>>> +  unsigned long hwcap2 = getauxval(AT_HWCAP2);
>>> +
>>> +  /* Bail out if MTE is not supported.  */
>>> +  if (!(hwcap2 & HWCAP2_MTE))
>>> +    return 1;
>>> +
>>> +  /* Enable the tagged address ABI, synchronous MTE tag check
>>> faults and
>>> +     allow all non-zero tags in the randomly generated set.  */
>>> +  if (prctl (PR_SET_TAGGED_ADDR_CTRL,
>>> +	     PR_TAGGED_ADDR_ENABLE | PR_MTE_TCF_SYNC
>>> +	     | (0xfffe << PR_MTE_TAG_SHIFT),
>>> +	     0, 0, 0))
>>> +    {
>>> +      perror ("prctl () failed");
>>> +      return 1;
>>> +    }
>>> +
>>> +  /* Create a mapping that will have PROT_MTE set.  */
>>> +  tagged_ptr = mmap (0, page_sz, PROT_READ | PROT_WRITE,
>>> +		     MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
>>> +  if (tagged_ptr == MAP_FAILED)
>>> +    {
>>> +      perror ("mmap () failed");
>>> +      return 1;
>>> +    }
>>> +
>>> +  /* Enable MTE on the above anonymous mmap.  */
>>> +  if (mprotect (tagged_ptr, page_sz, PROT_READ | PROT_WRITE |
>>> PROT_MTE))
>>> +    {
>>> +      perror ("mprotect () failed");
>>> +      return 1;
>>> +    }
>>> +
>>> +  access_memory (tagged_ptr);
>>> +
>>> +  return 0;
>>> +}
>>> diff --git a/gdb/testsuite/gdb.arch/aarch64-mte-gcore.exp
>>> b/gdb/testsuite/gdb.arch/aarch64-mte-gcore.exp
>>> new file mode 100644
>>> index 00000000000..8a19c4b449e
>>> --- /dev/null
>>> +++ b/gdb/testsuite/gdb.arch/aarch64-mte-gcore.exp
>>> @@ -0,0 +1,107 @@
>>> +# Copyright (C) 2018-2021 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 <
>>> https://urldefense.com/v3/__http://www.gnu.org/licenses/__;!!CTRNKA9wMg0ARbw!0jBVgsoE-RaTGT098YRLfC3gcQJlvfuMujbUqmtqoh3JjeD2BI4JNxmWCGaefncNfKBbmw$
>>>   >.
>>> +
>>> +# This file is part of the gdb testsuite.
>>> +
>>> +# Test generating and reading a core file with MTE memory tags.
>>> +
>>> +if {![is_aarch64_target]} {
>>> +    verbose "Skipping ${gdb_test_file_name}."
>>> +    return
>>> +}
>>> +
>>> +standard_testfile
>>> +if { [prepare_for_testing "failed to prepare" ${testfile}
>>> ${srcfile}] } {
>>> +    return -1
>>> +}
>>> +
>>> +if ![runto_main] {
>>> +    untested "could not run to main"
>>> +    return -1
>>> +}
>>> +
>>> +# Targets that don't support memory tagging should not execute the
>>> +# runtime memory tagging tests.
>>> +if {![supports_memtag]} {
>>> +    unsupported "memory tagging unsupported"
>>> +    return -1
>>> +}
>>> +
>>> +gdb_breakpoint "access_memory"
>>> +
>>> +if [gdb_continue "access_memory"] {
>>> +    return -1
>>> +}
>>> +
>>> +# Set each tag granule to a different tag value, from 0x0 to 0xf.
>>> +set atag_msg "Allocation tag\\(s\\) updated successfully\."
>>> +for {set i 15} {$i >= 0} {incr i -1} {
>>> +    set index [expr [expr 15 - $i] * 16]
>>> +    set tag [format "%02x" $i]
>>> +    gdb_test "memory-tag set-allocation-tag &tagged_ptr\[$index\]
>>> 1 $tag" \
>>> +	     $atag_msg \
>>> +	     "set memory tag of &tagged_ptr\[$index\] to $tag"
>>> +}
>>> +
>>> +# Run until a crash and confirm GDB displays memory tag violation
>>> +# information.
>>> +gdb_test "continue" \
>>> +    [multi_line \
>>> +	"Program received signal SIGSEGV, Segmentation fault" \
>>> +	"Memory tag violation while accessing address $hex" \
>>> +	"Allocation tag $hex" \
>>> +	"Logical tag $hex\." \
>>> +	"$hex in access_memory \\(.*\\) at .*" \
>>> +	".*tagged_ptr\\\[0\\\] = 'a';"] \
>>> +	 "display tag violation information for live process"
>>> +
>>> +# Generate the core file.
>>> +set core_filename [standard_output_file "$testfile.core"]
>>> +set core_generated [gdb_gcore_cmd "$core_filename" "generate core
>>> file"]
>>> +
>>> +if { !$core_generated } {
>>> +    return -1
>>> +}
>>> +
>>> +clean_restart $binfile
>>> +
>>> +# Load the core file and make sure we see the tag violation fault
>>> +# information.
>>> +gdb_test "core $core_filename" \
>>> +    [multi_line \
>>> +	"Core was generated by.*\." \
>>> +	"Program terminated with signal SIGSEGV, Segmentation fault" \
>>> +	"Memory tag violation while accessing address $hex" \
>>> +	"Allocation tag 0xf" \
>>> +	"Logical tag 0x0\." \
>>> +	"#0.*$hex in access_memory \\(.*\\) at .*" \
>>> +	".*tagged_ptr\\\[0\\\] = 'a';"] \
>>> +	 "core file shows tag violation information"
>>> +
>>> +# Make sure we have the tag_ctl register.
>>> +gdb_test "info register tag_ctl" \
>>> +	 "tag_ctl.*$hex.*${::decimal}" \
>>> +	 "tag_ctl is available"
>>> +
>>> +# Check if the tag granules have the expected values.  If they do,
>>> that
>>> +# means the core file saved the tags properly and GDB has read
>>> them
>>> +# correctly.
>>> +for {set i 15} {$i >= 0} {incr i -1} {
>>> +    set index [expr [expr 15 - $i] * 16]
>>> +    set tag [format "%x" $i]
>>> +    gdb_test "memory-tag print-allocation-tag
>>> &tagged_ptr\[$index\]" \
>>> +	     "= 0x$tag" \
>>> +	     "memory tag of &tagged_ptr\[$index\] is correct"
>>> +}
> 


  reply	other threads:[~2022-06-06  9:47 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-31 14:03 Luis Machado
2022-04-21 15:20 ` [PATCH, v2] " Luis Machado
2022-04-21 15:52   ` Eli Zaretskii
2022-04-22  8:12     ` Luis Machado
2022-04-22  8:30       ` Eli Zaretskii
2022-04-22  8:37         ` Luis Machado
2022-04-22  8:43           ` Eli Zaretskii
2022-04-22  8:44             ` Luis Machado
2022-04-22 13:27 ` [PATCH 1/2] " Luis Machado
2022-04-22 13:33   ` Eli Zaretskii
2022-04-22 13:30 ` [PATCH, v3] " Luis Machado
2022-05-03 21:56 ` [PATCH, v4] " Luis Machado
2022-05-12 10:36   ` Luis Machado
2022-05-18 12:46   ` Luis Machado
2022-05-18 13:58     ` John Baldwin
2022-05-23  9:50       ` Luis Machado
2022-05-23  9:49     ` Luis Machado
2022-06-06  9:28   ` Luis Machado
2022-06-06  9:42     ` Kuan-Ying Lee
2022-06-06  9:47       ` Luis Machado [this message]
2022-06-06  9:54         ` Kuan-Ying Lee
2022-06-06 10:49     ` Eli Zaretskii
2022-06-22  9:04   ` Luis Machado
2022-06-27 14:51   ` Pedro Alves
2022-07-11 10:13     ` Luis Machado
2022-07-11 10:57   ` [PATCH] [AArch64,v5] " Luis Machado
2022-07-18 13:54     ` Pedro Alves
2022-07-19 14:25       ` Luis Machado
  -- strict thread matches above, loose matches on Subject: below --
2021-05-18 20:20 [PATCH] [AArch64] " Luis Machado
2021-05-31 16:44 ` [PATCH,v4][AArch64] " Luis Machado

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=fd1bfc2a-7ba5-59df-867f-bb3960a90435@arm.com \
    --to=luis.machado@arm.com \
    --cc=Kuan-Ying.Lee@mediatek.com \
    --cc=gdb-patches@sourceware.org \
    /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).