RE: [PATCH] Commandline Support for the H8300 Simulator.

["D.Venkatasubramanian, Noida" <dvenkat at noida dot hcltech dot com>]

Hi All,

Based on some suggestions from Andrew Cagney,
and others, made some changes to the patch.

GDB part is here:
http://sources.redhat.com/ml/gdb-patches/2003-03/msg00357.html
Newlib part is here:
http://sources.redhat.com/ml/newlib/2003/msg00178.html

A pseudo opcode needs to be added to h8300.h.

#define O_SYS_CMDLINE 120

The 120 value is purely arbitrary, just kept some
distance between previous last value, as we could,
theoretically add more File I/O operations there.

Could someone review and apply the patch. Or give
some comments.

Thanks and regards,

Venky

2003-03-17 D.Venkatasubramanian <dvenkat at noida dot hcltech dot com>

	* h8300.h: Added a pseudo opcode for Commandline
	processing.






>-----Original Message-----
>From: D.Venkatasubramanian, Noida 
>Sent: Saturday, February 22, 2003 9:11 PM
>To: Kazu Hirata; Nick Clifton; 'gdb-patches at sources dot redhat dot com';
>'newlib at sources dot redhat dot com'; binutils at sources dot redhat dot com;
>gnuh8 at gnuh8 dot org dot uk
>Cc: D.Venkatasubramanian, Noida
>Subject: [PATCH] Commandline Support for the H8300 Simulator.
>
>
>Hi All,
>
>I am submitting a patch for adding support for 
>the processing of commandline arguments on the 
>H8300 Simulator.
>
>A brief description of each of the changes 
>follows:
>
>(File : h8300_h_patch.txt)
>1) Added a pseudo opcode O_SYS_CMDLINE which 
>will act as a trap identifier in h8300.h.
>
>(File : crt0_S_patch.txt)
>2) An extra instruction enclosed between a 
>#ifdef __SIMULATOR_ARGV_SUPPORT__ / #endif 
>directives, which simulates a trap instruction.
>This instruction is a jsr to 0xcc, one of the 
>locations in the magic vector used by H8300.
>
>(File : configure_host_patch.txt)
>3) Added a macro -DSIMULATOR_ARGV_SUPPORT. The 
>configure.host file ensures that this macro 
>is activated when our target is a simulator.
>
>(File : inst_h_patch.txt)
>4) Added a new char ** variable which will 
>hold the address of the Commandline arguments 
>which the common files of the simulator would 
>provide.
>
>(File : compile_c_patch.txt)
>This requires a lengthy explanation
>5) This macro sets a variable addr_cmdline to 
>point to the location of the 8-bit high memory 
>area as defined in the linker scripts. If the 
>linker scripts change, then so would this 
>require a change.
>
> #define SET_CMDLINE_LOCATION \
>   if (h8300smode) \
>     addr_cmdline = 0xffff00L; \
>   else if (h8300hmode) \
>     addr_cmdline = 0x2ff00L; \
>   else \
>     addr_cmdline = 0xff00L;
>
>6) 0xcc is a magic vector location used to 
>simulate a trap instruction on H8300.
>Q : Why I chose 0xcc?
>A : A patch for File I/O implementation is 
>pending approval, and those require 0xc5 - 0xcb, 
>hence the 0xcc, once that patch is approved, the 
>Oxc4 would be rendered useless and we could then 
>use it instead of 0xcc. (Whether or not we want to 
>change to 0xc4 then is a matter of trivial opinion.)
>
> 		      switch (dst->src.literal)
>  			{
> 			case 0xc4:
>  			  dst->opcode = O (O_SYSCALL, SB);
> 			  break;
> 			case 0xcc:
> 			  dst->opcode = O (O_SYS_CMDLINE, SB);
> 			  break;
>  			}
> 		      /* End of Processing for system calls.  */
>
>7) Setting pointer to Commandline arguments 
>provided by the common simulator support.
> 
>   /* Command Line support.  */
>   if (argv != NULL)
>     {
>       ptr_CommandLine = argv;
>     }
>   
>8) The code starting with this in sim_resume 
>handles the actual simulation of the system call.
>
>case O (O_SYS_CMDLINE, SB):
>
>I have added a lot of comments to the code.
>
>Briefly, this saves the commandline arguments to 
>the 8-bit memory area, and sets up an array which 
>contains the addresses in the 8-bit area, where
>each argument starts. Once that is done, these 
>have to be saved in simulator memory consecutively.
>As I need memory which is non volatile, I steal 
>some memory from the stack and move the stack 
>pointer such that I have enough memory to store all
>these addresses. As teh stack pointer holds the new
>value at the end, this memory is rendered non-volatile.
>
>At this point, I store the addresses in the array 
>containing the address values for the commandline 
>arguments to this memory.
>
>Nick Clifton had pointed out that at the beginning, 
>there should be another value, a pointer to the 
>beginning of argv. But the H8300 compiler seems not 
>to require this. I set the no. of arguments to 
>Register 0, address of the first of the address on 
>the non-volatile memory to Register 1, and set the 
>new stack pointer to the top of this non-volatile 
>memory. (Slightly controversial stuff, that extra
>pointer to argv that Nick had specified, but adding
>that seems to break the support, the last argument
>gets left out.)
>
>Tested on H8300-hms and H8300-elf and does not seem
>to break anything. Also I am able to print the 
>commandline arguments correctly.
>
>I used a simple test case :
>int main(int argc, char *argv[])
>{
>        int i;
>        int j = 0;
>
>        printf ("Argc = %d\n", argc);
>
>        for (i = 0; i < argc; i++)
>                printf ("%s\n", argv[i]);
>        return 0;
>}
>
>Could someone review this patch and point out any issues. 
>Thought, this would be very useful feature combined with
>the File I/O feature for which I have already submitted
>a patch.
>
>I think I will end this long description here.
>
>Thank You Nick Clifton and Kazu Hirata and all those that 
>were helpful in the explanation of ABI and other issues.
>
>Thanks and Regards,
>
>Venky
>
>PS : There are two files complete_patch.txt and 
>complete_ChangeLog.txt which contain all the patches
>and ChangeLogs in single files respectively.
>
>Changelog Entries
>
>In include/opcode (File : binutils_ChangeLog.txt)
>
>
>2003-02-22 D.Venkatasubramanian <dvenkat at noida dot hcltech dot com>
>
>	* h8300.h: Added a pseudo opcodes for Commandline
>	processing.
>
>In newlib/ (File : newlib_ChangeLog.txt)
>
>2003-02-22  D.Venkatasubramanian  <dvenkat at noida dot hcltech dot com>
>
>	* libc/include/sys/h8300hms/crt0.S: Added an extra 
>	instruction enclosed by SIMULATOR_ARGV_SUPPORT macro 
>	for Commandline support when target is simulator.
>	* configure.host: Added -DSIMULATOR_ARGV_SUPPORT for
>	Commandline Support.
>
>In sim/h8300/ (File : sim_ChangeLog.txt)
>
>2003-02-22  D.Venkatasubramanian  <dvenkat at noida dot hcltech dot com>
>
>	* compile.c: Added #define SET_CMDLINE_LOCATION to 
>	set the location of 8-bit (256 locations) where the
>	Command Line arguments would be stored.
>	(decode): Added a TRAP to 0xcc for Commandline 
>	processing using pseudo opcode O_SYS_CMDLINE.
>	(sim_resume): Added handling of O_SYS_CMDLINE Trap.
>	(sim_create_inferior): Setting a pointer to 
>	Commandline Args array.
>	* inst.h: Added a new variable ptr_CommandLine for
>	storing pointer to Commandline array.
>

Attachment: h8300_h_patch.txt
Description: Text document

Attachment: binutils_ChangeLog.txt
Description: Text document