Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Andreas Arnez <arnez@linux.vnet.ibm.com>
To: gdb-patches@sourceware.org
Cc: Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
Subject: [PATCH 1/2] S390: Take value from sub-register if applicable
Date: Fri, 15 Apr 2016 10:43:00 -0000	[thread overview]
Message-ID: <1460716939-27850-2-git-send-email-arnez@linux.vnet.ibm.com> (raw)
In-Reply-To: <1460716939-27850-1-git-send-email-arnez@linux.vnet.ibm.com>

On an s390 target with vector registers store.exp emits various FAILs
while trying to access a variable located in an unwound floating-point
register.  In those cases the variable's DWARF location happens to be
DW_OP_reg24, which refers to FP register f8 or to vector register v8,
depending on whether the target has vector registers.  But while f8 is
call-saved and thus can be unwound by GDB, v8 is not.  Thus an uplevel
variable with such a location can neither be read nor written.

A similar situation occurs on an s390 (32-bit) target with high GPRs.
For that case we already have logic in s390_unwind_pseudo_register which
unwinds a full GPR by casting its lower half to the larger type.
However, that cast does not result in an lvalue, so an uplevel variable
based on such a location is readable, but not writable.  Again, this
results in various FAILs from store.exp.

Both issues are addressed with a change to the s390 implementation of
the gdbarch method 'value_from_register': If the given type is small
enough to fit in the appropriate sub-register, then take the value from
that sub-register instead of from the full register.

gdb/ChangeLog:

	* s390-linux-tdep.c (s390_value_from_register): Use sub-register
	if appropriate.
---
 gdb/s390-linux-tdep.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/gdb/s390-linux-tdep.c b/gdb/s390-linux-tdep.c
index fc57592..8b4efb1 100644
--- a/gdb/s390-linux-tdep.c
+++ b/gdb/s390-linux-tdep.c
@@ -518,12 +518,18 @@ s390_value_from_register (struct gdbarch *gdbarch, struct type *type,
 			  int regnum, struct frame_id frame_id)
 {
   struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);
-  struct value *value = default_value_from_register (gdbarch, type,
-						     regnum, frame_id);
-  check_typedef (type);
+  unsigned int len = TYPE_LENGTH (check_typedef (type));
+  struct value *value;
 
-  if ((regnum >= S390_F0_REGNUM && regnum <= S390_F15_REGNUM
-       && TYPE_LENGTH (type) < 8)
+  /* If the value fits in a sub-register, use that instead.  */
+  if (regnum_is_vxr_full (tdep, regnum) && len <= 8)
+    regnum = regnum - tdep->v0_full_regnum + S390_F0_REGNUM;
+  else if (regnum_is_gpr_full (tdep, regnum) && len <= 4)
+    regnum = regnum - tdep->gpr_full_regnum + S390_R0_REGNUM;
+
+  value = default_value_from_register (gdbarch, type, regnum, frame_id);
+
+  if ((regnum >= S390_F0_REGNUM && regnum <= S390_F15_REGNUM && len < 8)
       || regnum_is_vxr_full (tdep, regnum)
       || (regnum >= S390_V16_REGNUM && regnum <= S390_V31_REGNUM))
     set_value_offset (value, 0);
-- 
2.5.0


  reply	other threads:[~2016-04-15 10:43 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-04-15 10:42 [PATCH 0/2] Fixes for some s390 fails with store.exp Andreas Arnez
2016-04-15 10:43 ` Andreas Arnez [this message]
2016-04-15 18:08   ` [PATCH 1/2] S390: Take value from sub-register if applicable Ulrich Weigand
2016-04-15 10:43 ` [PATCH 2/2] Involve gdbarch in taking DWARF register pieces Andreas Arnez
2016-04-15 18:10   ` Ulrich Weigand
2016-04-15 18:37     ` Pedro Alves
2016-04-18 11:53       ` Andreas Arnez
2016-04-18 13:53         ` Pedro Alves
2016-04-18 15:02           ` Andreas Arnez
2016-04-18 15:55             ` Pedro Alves
2016-04-18 15:57               ` Doug Evans
2016-04-19 12:08               ` Andreas Arnez
2016-04-28 13:24                 ` [PING][PATCH " Andreas Arnez
2016-04-28 14:47                   ` Pedro Alves
2016-04-28 16:51                     ` Andreas Arnez
2016-04-28 18:16                       ` Andreas Arnez
2016-04-28 22:15                         ` Pedro Alves
2016-04-28 22:15                       ` Pedro Alves
2016-06-14 17:03                   ` Jan Kratochvil

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=1460716939-27850-2-git-send-email-arnez@linux.vnet.ibm.com \
    --to=arnez@linux.vnet.ibm.com \
    --cc=Ulrich.Weigand@de.ibm.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