public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: Yvan Roux <yvan.roux@foss.st.com>
To: <gdb-patches@sourceware.org>, Luis Machado <luis.machado@arm.com>
Cc: Torbjorn SVENSSON <torbjorn.svensson@foss.st.com>
Subject: Re: [PATCH] gdb/arm: Sync sp with other *sp registers
Date: Thu, 7 Jul 2022 11:11:07 +0200	[thread overview]
Message-ID: <20220707091107.GA5239@gnbcxd0114.gnb.st.com> (raw)
In-Reply-To: <3a4bec66-7031-74ad-f7f4-d70af5dfe725@arm.com>

On Wed, Jul 06, 2022 at 01:34:59PM +0100, Luis Machado wrote:
> Hi,
> 
> On 6/30/22 08:13, Yvan Roux wrote:
> > Hi,
> > 
> > For Arm Cortex-M33 with security extensions, there are 4 different
> > stack pointers (msp_s, msp_ns, psp_s, psp_ns), without security
> > extensions and for other Cortex-M targets, there are 2 different
> > stack pointers (msp and psp).
> > 
> > This patch allows arch dependent unwinding override register values
> > for registers not covered by the DWARF data.  That means that sp
> > will always be in sync with one of the real stack pointers on
> > Arm targets that contains more than one stack pointer.
> > 
> > Signed-off-by: Torbjörn SVENSSON <torbjorn.svensson@foss.st.com>
> > Signed-off-by: Yvan Roux <yvan.roux@foss.st.com>
> > ---
> >   gdb/arm-tdep.c     | 94 ++++++++++++++++++++++++++++++++++++++++++++++
> >   gdb/dwarf2/frame.c | 58 ++++++++++++++++++++++++----
> >   gdb/dwarf2/frame.h |  9 +++++
> >   3 files changed, 153 insertions(+), 8 deletions(-)
> > 
> > diff --git a/gdb/arm-tdep.c b/gdb/arm-tdep.c
> > index 3a1b52c2380..48711cb2a65 100644
> > --- a/gdb/arm-tdep.c
> > +++ b/gdb/arm-tdep.c
> > @@ -4956,6 +4956,98 @@ arm_dwarf2_frame_init_reg (struct gdbarch *gdbarch, int regnum,
> >       }
> >   }
> > +/* Return the value of the unspecified register REGNUM for THIS_FRAME.  */
> > +
> > +static struct value *
> > +arm_dwarf2_frame_prev_unspecified_reg (struct frame_info *this_frame,
> > +				       int regnum)
> > +{
> > +  struct gdbarch *gdbarch = get_frame_arch (this_frame);
> > +  arm_gdbarch_tdep *tdep = (arm_gdbarch_tdep *) gdbarch_tdep (gdbarch);
> > +  bool override_with_sp_value = false;
> > +
> > +  if (tdep->have_sec_ext)
> > +    {
> > +      CORE_ADDR sp, msp_s, msp_ns, psp_s, psp_ns;
> > +      sp = get_frame_register_unsigned (this_frame, ARM_SP_REGNUM);
> > +      msp_s = get_frame_register_unsigned (this_frame,
> > +					   tdep->m_profile_msp_s_regnum);
> > +      msp_ns = get_frame_register_unsigned (this_frame,
> > +					    tdep->m_profile_msp_ns_regnum);
> > +      psp_s = get_frame_register_unsigned (this_frame,
> > +					   tdep->m_profile_psp_s_regnum);
> > +      psp_ns = get_frame_register_unsigned (this_frame,
> > +					    tdep->m_profile_psp_ns_regnum);
> > +
> > +      if (regnum == tdep->m_profile_msp_regnum && (msp_s == sp || msp_ns == sp))
> > +	{
> > +	  /* Use value of SP for MSP register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +      else if (regnum == tdep->m_profile_msp_s_regnum && msp_s == sp)
> > +	{
> > +	  /* Use value of SP for MSP_S register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +      else if (regnum == tdep->m_profile_msp_ns_regnum && msp_ns == sp)
> > +	{
> > +	  /* Use value of SP for MSP_NS register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +      else if (regnum == tdep->m_profile_psp_regnum && (psp_s == sp
> > +							|| psp_ns == sp))
> > +	{
> > +	  /* Use value of SP for PSP register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +      else if (regnum == tdep->m_profile_psp_s_regnum && psp_s == sp)
> > +	{
> > +	  /* Use value of SP for PSP_S register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +      else if (regnum == tdep->m_profile_psp_ns_regnum && psp_ns == sp)
> > +	{
> > +	  /* Use value of SP for PSP_NS register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +    }
> > +  else if (tdep->is_m)
> > +    {
> > +      CORE_ADDR sp, msp, psp;
> > +      sp = get_frame_register_unsigned (this_frame, ARM_SP_REGNUM);
> > +      msp = get_frame_register_unsigned (this_frame,
> > +					 tdep->m_profile_msp_regnum);
> > +      psp = get_frame_register_unsigned (this_frame,
> > +					 tdep->m_profile_psp_regnum);
> > +
> > +      if (regnum == tdep->m_profile_msp_regnum && sp == msp)
> > +	{
> > +	  /* Use value of SP for MSP register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +      else if (regnum == tdep->m_profile_psp_regnum && sp == psp)
> > +	{
> > +	  /* Use value of SP for PSP register.  */
> > +	  override_with_sp_value = true;
> > +	}
> > +    }
> > +
> > +  if (override_with_sp_value)
> > +    {
> > +      CORE_ADDR prev_sp;
> > +      struct frame_info *prev_frame = get_prev_frame (this_frame);
> > +      if (prev_frame)
> > +	prev_sp = get_frame_register_unsigned (prev_frame, ARM_SP_REGNUM);
> > +      else
> > +	prev_sp = get_frame_base (this_frame);
> > +
> > +      return frame_unwind_got_constant (this_frame, regnum, prev_sp);
> > +    }
> > +
> > +  /* Don't know how to fetch this value, use generic method.  */
> > +  return NULL;
> > +}
> > +
> >   /* Given BUF, which is OLD_LEN bytes ending at ENDADDR, expand
> >      the buffer to be NEW_LEN bytes ending at ENDADDR.  Return
> >      NULL if an error occurs.  BUF is freed.  */
> > @@ -10438,6 +10530,8 @@ arm_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
> >     gdbarch_init_osabi (info, gdbarch);
> >     dwarf2_frame_set_init_reg (gdbarch, arm_dwarf2_frame_init_reg);
> > +  dwarf2_frame_set_prev_unspecified_reg (gdbarch,
> > +					 arm_dwarf2_frame_prev_unspecified_reg);
> >     /* Add some default predicates.  */
> >     if (is_m)
> > diff --git a/gdb/dwarf2/frame.c b/gdb/dwarf2/frame.c
> > index 5878d72f922..c4b15b2d240 100644
> > --- a/gdb/dwarf2/frame.c
> > +++ b/gdb/dwarf2/frame.c
> > @@ -595,6 +595,10 @@ struct dwarf2_frame_ops
> >     /* Convert .eh_frame register number to DWARF register number, or
> >        adjust .debug_frame register number.  */
> >     int (*adjust_regnum) (struct gdbarch *, int, int);
> > +
> > +  /* Return the value of the unspecified register REGNUM for THIS_FRAME. */
> > +  struct value * (*prev_unspecified_reg) (struct frame_info *this_frame,
> > +					  int regnum);
> >   };
> >   /* Default architecture-specific register state initialization
> > @@ -677,6 +681,50 @@ dwarf2_frame_init_reg (struct gdbarch *gdbarch, int regnum,
> >     ops->init_reg (gdbarch, regnum, reg, this_frame);
> >   }
> > +/* Set the architecture-specific unspecified register unwinding
> > +   function for GDBARCH to PREV_UNSPECIFIED_REG.  */
> > +
> > +void
> > +dwarf2_frame_set_prev_unspecified_reg (struct gdbarch *gdbarch,
> > +				       struct value * (*prev_unspecified_reg)
> > +					      (struct frame_info *this_frame,
> > +					       int regnum))
> > +{
> > +  struct dwarf2_frame_ops *ops
> > +    = (struct dwarf2_frame_ops *) gdbarch_data (gdbarch, dwarf2_frame_data);
> > +
> > +  ops->prev_unspecified_reg = prev_unspecified_reg;
> > +}
> 
> I'm wondering if we really need to introduce a hook here. I understand we may need to sidestep the dwarf unwinder
> to do custom handling of the msp*/psp* registers, but wouldn't we be able to achieve the same with the existing
> dwarf hooks?
> 
> For example, we could register a hook with dwarf2_frame_set_init_reg (), like aarch64 (see gdb/aarch64-tdep.c:aarch64_dwarf2_frame_init_reg).
> 
> With such a hook, we could define our own rules for individual registers through DWARF2_FRAME_REG_FN.
> 
> Would that work?

Yes, that's indeed a better and cleaner solution, we've tested it and it works
fine.  I'll submit a new patchset.

-- 
Y.

      reply	other threads:[~2022-07-07  9:11 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-06-30  7:13 Yvan Roux
2022-07-06 12:34 ` Luis Machado
2022-07-07  9:11   ` Yvan Roux [this message]

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=20220707091107.GA5239@gnbcxd0114.gnb.st.com \
    --to=yvan.roux@foss.st.com \
    --cc=gdb-patches@sourceware.org \
    --cc=luis.machado@arm.com \
    --cc=torbjorn.svensson@foss.st.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).