Re: RFA: ia64 portion of libunwind patch

["J. Johnston" <jjohnstn at redhat dot com>]

I have just added a new patch which accounts for Andrew's latest xfer_partial 
changes.

Ok to commit?

-- Jeff J.

2003-11-13  Jeff Johnston  <jjohnstn@redhat.com>

	* config/ia64/nm-linux.h (NATIVE_XFER_UNWIND_TABLE): New macro to
	set up the ia64 linux native target method for getting the kernel
	unwind table.
	* ia64-linux-nat.c (ia64_linux_xfer_unwind_table): New function.
	* inftarg.c (child_xfer_partial): Enable section of code to
	handle TARGET_OBJECT_UNWIND_TABLE.
	* target.h (target_object): Add new TARGET_OBJECT_UNWIND_TABLE macro.


J. Johnston wrote:
> I have chosen the first option to implement the syscall.  I have 
> included a patch here regarding this mechanism.
>
> Ok to commit?
>
> -- Jeff J.
>
> 2003-11-04  Jeff Johnston  <jjohnstn@redhat.com>
>
>     * config/ia64/nm-linux.h (CHILD_TO_XFER_PARTIAL): New macro to set 
> up           the ia64 linux native target to_xfer_partial method.
>     (ia64_linux_xfer_partial): New prototype.
>     * ia64-linux-nat.c (ia64_linux_xfer_partial): New function.
>     * inftarg.c (init_child_ops)[CHILD_TO_XFER_PARTIAL]: Set up xfer 
> partial method if one provided via macro.
>     * target.c (init_dummy_target): Initialize to_xfer_partial to default.
>     * target.h (target_object): Add new TARGET_OBJECT_UNWIND_TABLE enum.
>
>
> Andrew Cagney wrote:
>
> > > Nothing which involves a syscall is acceptable in a tdep file.  That's
> > > what the t means - target support.
> >
> >
> >
> > ("tdep" mysteriously means "architecture vector support".  That "t" 
> > really no longer makes sense :-()
> >
> > Yes, architecture vector shouldn't be directly making syscalls.  
> > Instead the architecture specific code should use the target vector to 
> > obtain this system information.  Here I think the best option is to 
> > add another TARGET_OBJECT_xxxx variant and then use the target_read() 
> > method to pull in the data.
> >
> > As for the native target side, ia64-linux-nat should export something 
> > like ia64_linux_child_read_partial that performs the syscall.  The 
> > tricky [er messy] part is wiring it into child_ops, I see there are 
> > lots of choices :-(
> >
> > - define a nm-*.c macro and have that enable the assignment (ex 
> > #define NATIVE_TO_READ_PARTIAL ia64_linux_child?) (ref KILL_INFERIOR)
> >
> > - have a callback set the to_read_partial method in child_ops 
> > (exec_set_find_memory_regions)
> >
> > - modify all the targets so that each implements the new method
> >
> > - others?
> >
> > Is this information available via /proc?  In a core file?
> >
> > Andrew
> >
> > PS: Note this pending patch.  Dependant on the timing you or I may 
> > need to tweak the name.
> > http://sources.redhat.com/ml/gdb-patches/2003-10/msg00795.html
Index: config/ia64/nm-linux.h
===================================================================
RCS file: /cvs/src/src/gdb/config/ia64/nm-linux.h,v
retrieving revision 1.10
diff -u -r1.10 nm-linux.h
--- config/ia64/nm-linux.h	20 Jun 2003 13:57:29 -0000	1.10
+++ config/ia64/nm-linux.h	14 Nov 2003 00:06:02 -0000
@@ -69,4 +69,15 @@
 extern int ia64_linux_remove_watchpoint (ptid_t ptid, CORE_ADDR addr,
                                          int len);
 
+#include "target.h"
+
+#define NATIVE_XFER_UNWIND_TABLE ia64_linux_xfer_unwind_table
+extern LONGEST ia64_linux_xfer_unwind_table (struct target_ops *ops, 
+					     enum target_object object,
+					     const char *annex, 
+					     void *readbuf,
+					     const void *writebuf,
+					     ULONGEST offset, 
+					     LONGEST len);
+
 #endif /* #ifndef NM_LINUX_H */
Index: target.h
===================================================================
RCS file: /cvs/src/src/gdb/target.h,v
retrieving revision 1.53
diff -u -r1.53 target.h
--- target.h	10 Nov 2003 21:20:44 -0000	1.53
+++ target.h	14 Nov 2003 00:06:03 -0000
@@ -223,8 +223,10 @@
   /* AVR target specific transfer.  See "avr-tdep.c" and "remote.c".  */
   TARGET_OBJECT_AVR,
   /* Transfer up-to LEN bytes of memory starting at OFFSET.  */
-  TARGET_OBJECT_MEMORY
-  /* Possible future ojbects: TARGET_OJBECT_FILE, TARGET_OBJECT_PROC,
+  TARGET_OBJECT_MEMORY,
+  /* Kernel Unwind Table.  See "ia64-tdep.c".  */
+  TARGET_OBJECT_UNWIND_TABLE,
+  /* Possible future objects: TARGET_OBJECT_FILE, TARGET_OBJECT_PROC,
      TARGET_OBJECT_AUXV, ...  */
 };
 
Index: inftarg.c
===================================================================
RCS file: /cvs/src/src/gdb/inftarg.c,v
retrieving revision 1.20
diff -u -r1.20 inftarg.c
--- inftarg.c	10 Nov 2003 21:20:44 -0000	1.20
+++ inftarg.c	14 Nov 2003 00:06:03 -0000
@@ -571,14 +571,12 @@
 				  NULL, ops);
       return -1;
 
-#if 0
     case TARGET_OBJECT_UNWIND_TABLE:
 #ifndef NATIVE_XFER_UNWIND_TABLE
 #define NATIVE_XFER_UNWIND_TABLE(OPS,OBJECT,ANNEX,WRITEBUF,READBUF,OFFSET,LEN) (-1)
 #endif
       return NATIVE_XFER_UNWIND_TABLE (ops, object, annex, readbuf, writebuf,
 				       offset, len);
-#endif
 
 #if 0
     case TARGET_OBJECT_AUXV:
Index: ia64-linux-nat.c
===================================================================
RCS file: /cvs/src/src/gdb/ia64-linux-nat.c,v
retrieving revision 1.20
diff -u -r1.20 ia64-linux-nat.c
--- ia64-linux-nat.c	2 Oct 2003 20:28:29 -0000	1.20
+++ ia64-linux-nat.c	14 Nov 2003 00:06:03 -0000
@@ -34,6 +34,7 @@
 #ifdef HAVE_SYS_REG_H
 #include <sys/reg.h>
 #endif
+#include <sys/syscall.h>
 #include <sys/user.h>
 
 #include <asm/ptrace_offsets.h>
@@ -644,4 +645,14 @@
   write_register_pid (IA64_PSR_REGNUM, psr, ptid);
 
   return (CORE_ADDR) siginfo.si_addr;
+}
+
+LONGEST 
+ia64_linux_xfer_unwind_table (struct target_ops *ops,
+			      enum target_object object,
+			      const char *annex,
+			      void *readbuf, const void *writebuf,
+			      ULONGEST offset, LONGEST len)
+{
+  return syscall (__NR_getunwind, readbuf, len);
 }