/[gxemul]/trunk/src/devices/dev_ram.c
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Annotation of /trunk/src/devices/dev_ram.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 42 - (hide annotations)
Mon Oct 8 16:22:32 2007 UTC (16 years, 6 months ago) by dpavlin
File MIME type: text/plain
File size: 5609 byte(s)
++ trunk/HISTORY	(local)
$Id: HISTORY,v 1.1613 2007/06/15 20:11:26 debug Exp $
20070501	Continuing a little on m88k disassembly (control registers,
		more instructions).
		Adding a dummy mvme88k machine mode.
20070502	Re-adding MIPS load/store alignment exceptions.
20070503	Implementing more of the M88K disassembly code.
20070504	Adding disassembly of some more M88K load/store instructions.
		Implementing some relatively simple M88K instructions (br.n,
		xor[.u] imm, and[.u] imm).
20070505	Implementing M88K three-register and, or, xor, and jmp[.n],
		bsr[.n] including function call trace stuff.
		Applying a patch from Bruce M. Simpson which implements the
		SYSCON_BOARD_CPU_CLOCK_FREQ_ID object of the syscon call in
		the yamon PROM emulation.
20070506	Implementing M88K bb0[.n] and bb1[.n], and skeletons for
		ldcr and stcr (although no control regs are implemented yet).
20070509	Found and fixed the bug which caused Linux for QEMU_MIPS to
		stop working in 0.4.5.1: It was a faulty change to the MIPS
		'sc' and 'scd' instructions I made while going through gcc -W
		warnings on 20070428.
20070510	Updating the Linux/QEMU_MIPS section in guestoses.html to
		use mips-test-0.2.tar.gz instead of 0.1.
		A big thank you to Miod Vallat for sending me M88K manuals.
		Implementing more M88K instructions (addu, subu, div[u], mulu,
		ext[u], clr, set, cmp).
20070511	Fixing bugs in the M88K "and" and "and.u" instructions (found
		by comparing against the manual).
		Implementing more M88K instructions (mask[.u], mak, bcnd (auto-
		generated)) and some more control register details.
		Cleanup: Removing the experimental AVR emulation mode and
		corresponding devices; AVR emulation wasn't really meaningful.
		Implementing autogeneration of most M88K loads/stores. The
		rectangle drawing demo (with -O0) for M88K runs :-)
		Beginning on M88K exception handling.
		More M88K instructions: tb0, tb1, rte, sub, jsr[.n].
		Adding some skeleton MVME PROM ("BUG") emulation.
20070512	Fixing a bug in the M88K cmp instruction.
		Adding the M88K lda (scaled register) instruction.
		Fixing bugs in 64-bit (32-bit pairs) M88K loads/stores.
		Removing the unused tick_hz stuff from the machine struct.
		Implementing the M88K xmem instruction. OpenBSD/mvme88k gets
		far enough to display the Copyright banner :-)
		Implementing subu.co (guess), addu.co, addu.ci, ff0, and ff1.
		Adding a dev_mvme187, for MVME187-specific devices/registers.
		OpenBSD/mvme88k prints more boot messages. :)
20070515	Continuing on MVME187 emulation (adding more devices, beginning
		on the CMMUs, etc).
		Adding the M88K and.c, xor.c, and or.c instructions, and making
		sure that mul, div, etc cause exceptions if executed when SFD1
		is disabled.
20070517	Continuing on M88K and MVME187 emulation in general; moving
		the CMMU registers to the CPU struct, separating dev_pcc2 from
		dev_mvme187, and beginning on memory_m88k.c (BATC and PATC).
		Fixing a bug in 64-bit (32-bit pairs) M88K fast stores.
		Implementing the clock part of dev_mk48txx.
		Implementing the M88K fstcr and xcr instructions.
		Implementing m88k_cpu_tlbdump().
		Beginning on the implementation of a separate address space
		for M88K .usr loads/stores.
20070520	Removing the non-working (skeleton) Sandpoint, SonyNEWS, SHARK
		Dnard, and Zaurus machine modes.
		Experimenting with dyntrans to_be_translated read-ahead. It
		seems to give a very small performance increase for MIPS
		emulation, but a large performance degradation for SuperH. Hm.
20070522	Disabling correct SuperH ITLB emulation; it does not seem to be
		necessary in order to let SH4 guest OSes run, and it slows down
		userspace code.
		Implementing "samepage" branches for SuperH emulation, and some
		other minor speed hacks.
20070525	Continuing on M88K memory-related stuff: exceptions, memory
		transaction register contents, etc.
		Implementing the M88K subu.ci instruction.
		Removing the non-working (skeleton) Iyonix machine mode.
		OpenBSD/mvme88k reaches userland :-), starts executing
		/sbin/init's instructions, and issues a few syscalls, before
		crashing.
20070526	Fixing bugs in dev_mk48txx, so that OpenBSD/mvme88k detects
		the correct time-of-day.
		Implementing a generic IRQ controller for the test machines
		(dev_irqc), similar to a proposed patch from Petr Stepan.
		Experimenting some more with translation read-ahead.
		Adding an "expect" script for automated OpenBSD/landisk
		install regression/performance tests.
20070527	Adding a dummy mmEye (SH3) machine mode skeleton.
		FINALLY found the strange M88K bug I have been hunting: I had
		not emulated the SNIP value for exceptions occurring in
		branch delay slots correctly.
		Implementing correct exceptions for 64-bit M88K loads/stores.
		Address to symbol lookups are now disabled when M88K is
		running in usermode (because usermode addresses don't have
		anything to do with supervisor addresses).
20070531	Removing the mmEye machine mode skeleton.
20070604	Some minor code cleanup.
20070605	Moving src/useremul.c into a subdir (src/useremul/), and
		cleaning up some more legacy constructs.
		Adding -Wstrict-aliasing and -fstrict-aliasing detection to
		the configure script.
20070606	Adding a check for broken GCC on Solaris to the configure
		script. (GCC 3.4.3 on Solaris cannot handle static variables
		which are initialized to 0 or NULL. :-/)
		Removing the old (non-working) ARC emulation modes: NEC RD94,
		R94, R96, and R98, and the last traces of Olivetti M700 and
		Deskstation Tyne.
		Removing the non-working skeleton WDSC device (dev_wdsc).
20070607	Thinking about how to use the host's cc + ld at runtime to
		generate native code. (See experiments/native_cc_ld_test.i
		for an example.)
20070608	Adding a program counter sampling timer, which could be useful
		for native code generation experiments.
		The KN02_CSR_NRMMOD bit in the DECstation 5000/200 (KN02) CSR
		should always be set, to allow a 5000/200 PROM to boot.
20070609	Moving out breakpoint details from the machine struct into
		a helper struct, and removing the limit on max nr of
		breakpoints.
20070610	Moving out tick functions into a helper struct as well (which
		also gets rid of the max limit).
20070612	FINALLY figured out why Debian/DECstation stopped working when
		translation read-ahead was enabled: in src/memory_rw.c, the
		call to invalidate_code_translation was made also if the
		memory access was an instruction load (if the page was mapped
		as writable); it shouldn't be called in that case.
20070613	Implementing some more MIPS32/64 revision 2 instructions: di,
		ei, ext, dext, dextm, dextu, and ins.
20070614	Implementing an instruction combination for the NetBSD/arm
		idle loop (making the host not use any cpu if NetBSD/arm
		inside the emulator is not using any cpu).
		Increasing the nr of ARM VPH entries from 128 to 384.
20070615	Removing the ENABLE_arch stuff from the configure script, so
		that all included architectures are included in both release
		and development builds.
		Moving memory related helper functions from misc.c to memory.c.
		Adding preliminary instructions for netbooting NetBSD/pmppc to
		guestoses.html; it doesn't work yet, there are weird timeouts.
		Beginning a total rewrite of the userland emulation modes
		(removing all emulation modes, beginning from scratch with
		NetBSD/MIPS and FreeBSD/Alpha only).
20070616	After fixing a bug in the DEC21143 NIC (the TDSTAT_OWN bit was
		only cleared for the last segment when transmitting, not all
		segments), NetBSD/pmppc boots with root-on-nfs without the
		timeouts. Updating guestoses.html.
		Removing the skeleton PSP (Playstation Portable) mode.
		Moving X11-related stuff in the machine struct into a helper
		struct.
		Cleanup of out-of-memory checks, to use a new CHECK_ALLOCATION
		macro (which prints a meaningful error message).
		Adding a COMMENT to each machine and device (for automagic
		.index comment generation).
		Doing regression testing for the next release.

==============  RELEASE 0.4.6  ==============


1 dpavlin 4 /*
2 dpavlin 34 * Copyright (C) 2004-2007 Anders Gavare. All rights reserved.
3 dpavlin 4 *
4     * Redistribution and use in source and binary forms, with or without
5     * modification, are permitted provided that the following conditions are met:
6     *
7     * 1. Redistributions of source code must retain the above copyright
8     * notice, this list of conditions and the following disclaimer.
9     * 2. Redistributions in binary form must reproduce the above copyright
10     * notice, this list of conditions and the following disclaimer in the
11     * documentation and/or other materials provided with the distribution.
12     * 3. The name of the author may not be used to endorse or promote products
13     * derived from this software without specific prior written permission.
14     *
15     * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16     * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17     * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18     * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19     * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20     * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21     * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22     * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23     * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24     * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25     * SUCH DAMAGE.
26     *
27     *
28 dpavlin 42 * $Id: dev_ram.c,v 1.25 2007/06/15 19:57:33 debug Exp $
29 dpavlin 4 *
30 dpavlin 42 * COMMENT: A generic RAM (memory) device
31     *
32     * Note: This device can also be used to mirror/alias another part of RAM.
33 dpavlin 4 */
34    
35     #include <stdio.h>
36     #include <stdlib.h>
37     #include <string.h>
38     #include <sys/types.h>
39     #include <sys/mman.h>
40    
41     #include "cpu.h"
42     #include "devices.h"
43 dpavlin 18 #include "machine.h"
44 dpavlin 4 #include "memory.h"
45     #include "misc.h"
46    
47    
48     /* #define RAM_DEBUG */
49    
50     struct ram_data {
51 dpavlin 36 uint64_t baseaddress;
52    
53 dpavlin 4 int mode;
54     uint64_t otheraddress;
55    
56 dpavlin 18 /* If mode = DEV_RAM_MIRROR: */
57     uint64_t offset;
58    
59 dpavlin 4 /* If mode = DEV_RAM_RAM: */
60     unsigned char *data;
61     uint64_t length;
62     };
63    
64    
65 dpavlin 22 DEVICE_ACCESS(ram)
66 dpavlin 4 {
67     struct ram_data *d = extra;
68    
69     #ifdef RAM_DEBUG
70     if (writeflag==MEM_READ) {
71     debug("[ ram: read from 0x%x, len=%i ]\n",
72     (int)relative_addr, (int)len);
73     } else {
74     int i;
75     debug("[ ram: write to 0x%x:", (int)relative_addr);
76     for (i=0; i<len; i++)
77     debug(" %02x", data[i]);
78     debug(" (len=%i) ]\n", len);
79     }
80     #endif
81    
82     switch (d->mode) {
83 dpavlin 42
84 dpavlin 4 case DEV_RAM_MIRROR:
85     /* TODO: how about caches? */
86     return cpu->memory_rw(cpu, mem,
87     d->otheraddress + relative_addr, data, len,
88     writeflag, PHYSICAL);
89 dpavlin 42
90 dpavlin 4 case DEV_RAM_RAM:
91 dpavlin 36 if (writeflag == MEM_WRITE) {
92 dpavlin 4 memcpy(&d->data[relative_addr], data, len);
93 dpavlin 36
94     /* Invalidate any code translations on a write: */
95     if (cpu->invalidate_code_translation != NULL) {
96     cpu->invalidate_code_translation(
97     cpu, d->baseaddress + relative_addr,
98     INVALIDATE_PADDR);
99     }
100     } else {
101 dpavlin 4 memcpy(data, &d->data[relative_addr], len);
102 dpavlin 36 }
103 dpavlin 4 break;
104 dpavlin 42
105 dpavlin 4 default:
106     fatal("dev_ram_access(): unknown mode %i\n", d->mode);
107     exit(1);
108     }
109    
110     return 1;
111     }
112    
113    
114     /*
115     * dev_ram_init():
116 dpavlin 18 *
117     * Initializes a RAM or mirror device. Things get a bit complicated because
118     * of dyntrans (i.e. mirrored memory ranges should be entered into the
119     * translation arrays just as normal memory and other devices are).
120 dpavlin 4 */
121 dpavlin 18 void dev_ram_init(struct machine *machine, uint64_t baseaddr, uint64_t length,
122 dpavlin 4 int mode, uint64_t otheraddress)
123     {
124     struct ram_data *d;
125 dpavlin 20 int flags = DM_DEFAULT, points_to_ram = 1;
126 dpavlin 4
127 dpavlin 42 CHECK_ALLOCATION(d = malloc(sizeof(struct ram_data)));
128 dpavlin 4 memset(d, 0, sizeof(struct ram_data));
129    
130 dpavlin 18 if (mode & DEV_RAM_MIGHT_POINT_TO_DEVICES) {
131     mode &= ~DEV_RAM_MIGHT_POINT_TO_DEVICES;
132     points_to_ram = 0;
133     }
134    
135 dpavlin 4 d->mode = mode;
136 dpavlin 36 d->baseaddress = baseaddr;
137 dpavlin 4 d->otheraddress = otheraddress;
138    
139     switch (d->mode) {
140 dpavlin 18
141 dpavlin 4 case DEV_RAM_MIRROR:
142 dpavlin 18 /*
143     * Calculate the amount that the mirror memory is offset from
144     * the real (physical) memory. This is used in src/memory_rw.c
145 dpavlin 20 * with dyntrans accesses if DM_EMULATED_RAM is set.
146 dpavlin 18 */
147     d->offset = baseaddr - otheraddress;
148    
149     /* Aligned RAM? Then it works with dyntrans. */
150     if (points_to_ram &&
151     (baseaddr & (machine->arch_pagesize-1)) == 0 &&
152     (otheraddress & (machine->arch_pagesize - 1)) == 0 &&
153     (length & (machine->arch_pagesize - 1)) == 0)
154 dpavlin 20 flags |= DM_DYNTRANS_OK | DM_DYNTRANS_WRITE_OK
155     | DM_EMULATED_RAM;
156 dpavlin 18
157     memory_device_register(machine->memory, "ram [mirror]",
158     baseaddr, length, dev_ram_access, d, flags
159 dpavlin 20 | DM_READS_HAVE_NO_SIDE_EFFECTS, (void *) &d->offset);
160 dpavlin 4 break;
161 dpavlin 18
162 dpavlin 4 case DEV_RAM_RAM:
163     /*
164     * Allocate zero-filled RAM using mmap(). If mmap() failed,
165 dpavlin 42 * try malloc(), but then memset() must also be called, which
166 dpavlin 4 * can be slow for large chunks of memory.
167     */
168     d->length = length;
169     d->data = (unsigned char *) mmap(NULL, length,
170     PROT_READ | PROT_WRITE, MAP_ANON | MAP_PRIVATE, -1, 0);
171     if (d->data == NULL) {
172 dpavlin 42 CHECK_ALLOCATION(d->data = malloc(length));
173 dpavlin 4 memset(d->data, 0, length);
174     }
175 dpavlin 18
176     /* Aligned memory? Then it works with dyntrans. */
177     if ((baseaddr & (machine->arch_pagesize - 1)) == 0 &&
178     (length & (machine->arch_pagesize - 1)) == 0)
179 dpavlin 20 flags |= DM_DYNTRANS_OK | DM_DYNTRANS_WRITE_OK;
180 dpavlin 18
181     memory_device_register(machine->memory, "ram", baseaddr,
182     d->length, dev_ram_access, d, flags
183 dpavlin 20 | DM_READS_HAVE_NO_SIDE_EFFECTS, d->data);
184 dpavlin 4 break;
185 dpavlin 18
186 dpavlin 4 default:
187     fatal("dev_ram_access(): unknown mode %i\n", d->mode);
188     exit(1);
189     }
190     }
191    

  ViewVC Help
Powered by ViewVC 1.1.26