Make 4K EEPROM default save type; support carts with no EEPROM #878

meeq · 2021-06-23T22:04:26Z

Proper fix for workaround in DragonMinded/libdragon#125 (review)

The Joybus response values have been confirmed on real hardware using the following test ROM which dumps the EEPROM status input and output data: eepromtest.zip

I was not able to run this test ROM on Mupen64Plus; all of my libdragon example ROMs get stuck at "Mupen64Plus Started...":

I'd appreciate some assistance on getting this to run. I was able to test this ROM successfully with cen64.

loganmc10 · 2021-06-24T00:15:55Z

I'll test this in a bit, I think it would be cleaner to just leave the type as a uint16_t and do

rx_buf[0] = (uint8_t)(cart->eeprom.type >> 0);
rx_buf[1] = (uint8_t)(cart->eeprom.type >> 8);
rx_buf[2] = (uint8_t)(cart->eeprom.type >> 0);

(Just repeat the first byte, which is 0 when present and 0xff when absent)

loganmc10 · 2021-06-24T00:31:44Z

The other thing I'll say: the way this is written, it defaults to "no eeprom". This means that any ROM not in the INI won't support eeprom saves. This might be a problem for ROM hacks/new ROMs (like a new version of Smash Remix or something).

I think it would be better to retain the default of 4k eeprom. I don't think we've run across a game that failed because it said it had eeprom when it shouldn't.

EDIT: Or maybe we just set it to "no eeprom" if the INI says it uses another save type like flash or sram

loganmc10 · 2021-06-24T01:01:09Z

Here is the result of the test ROM (after I applied this PR):

I think that looks right, but as I said, I think we should still default to 4k eeprom. It's a fairly safe assumption that if a commercial game is probing that address, it expects the eeprom to be there.

meeq · 2021-06-24T02:18:26Z

I have an updated EEPROM test ROM that exercises a variety of EEPROM probe command formats: eepromtest.zip

This branch has been updated to reflect a more-accurate understanding of how the real hardware handles EEPROM checking: If there is no EEPROM, the recv buffer will not be modified.

meeq · 2021-06-24T02:29:39Z

As for the "user-experience question" of whether it's actually a good idea to make this change: it's subjective.

According to this list of save types, 4K EEPROM is the second-most popular save format for N64 titles. There is a lot of sense in the argument that a user-friendly emulator should assume 4K EEPROM if no other save type is specified, since it (almost always) won't hurt and has a decent chance of helping.

My only argument against is that homebrew/mods that want other save types (such as Smash Remix which uses SRAM) will still be a broken experience that requires the user to select the right save type. So it will "just work" some of the time, but silently fail other times.

loganmc10 · 2021-06-24T02:42:31Z

-    JDT_EEPROM_4K        = 0x8000,  /* 4k EEPROM */
-    JDT_EEPROM_16K       = 0xc000,  /* 16k EEPROM */
+    JDT_EEPROM_4K        = 0x0080,  /* 4k EEPROM */
+   JDT_EEPROM_16K       = 0x00c0,  /* 16k EEPROM */

Did you mean to flip the order of these bytes? I don't think that's right.

Jj0YzL5nvJ

DK64 demos look better (less sped up), but now doesn't initialize the save file.

Based on EEPROM test ROM run on real hardware.

meeq · 2021-06-24T12:48:16Z

Sorry, I definitely messed up by flipping the bytes of the EEPROM identifiers. Let me try to figure out how to get my libdragon ROMs running on this branch so I can run the test ROM and verify these changes myself.

meeq · 2021-06-24T13:52:14Z

I have attached the libdragon test suite, which includes EEPROM Joybus command tests: testrom_emu.zip

Line numbers for test_eeprom failures are mapped to this file: libdragon/tests/test_eeprom.c

There are several other failures unrelated to EEPROM that are out of scope for this PR.

Without the changes in this PR, it is not possible to exercise the "EEPROM not detected" code-path since Mupen64Plus will always have a 4K EEPROM (unless 16K EEPROM is specified)

loganmc10 · 2021-06-24T14:11:06Z

Here are the results with the PR applied (default settings):

Core: IS64: **** Starting test: test_exception
Core: IS64: ASSERTION FAILED (test_exception.c:258):
Core: IS64: breakpoint_occured != 1 (0 != 1)
Core: IS64: Breakpoint exception was not triggered
Core: IS64: **** Starting test: test_timer_ticks
Core: IS64: ASSERTION FAILED (test_timer.c:195):
Core: IS64: t5-t0 < 1000
Core: IS64: invalid timer_ticks [start=ffffffe9]: 160000000b - 17000002fd
Core: IS64: **** Starting test: test_timer_oneshot
Core: IS64: **** Starting test: test_timer_slow_callback
Core: IS64: **** Starting test: test_timer_continuous
Core: IS64: **** Starting test: test_timer_mixed
Core: IS64: **** Starting test: test_timer_disabled_start
Core: IS64: **** Starting test: test_timer_disabled_restart
Core: IS64: **** Starting test: test_irq_reentrancy
Core: IS64: **** Starting test: test_eeprom
Core: IS64: 4K EEPROM detected.
Core: IS64: 4K EEPROM detected.
Core: IS64: 4K EEPROM detected.
Core: IS64: 4K EEPROM detected.
Core: IS64: **** Starting test: test_dfs_read
Core: IS64: **** Starting test: test_dfs_rom_addr
Core: IS64: **** Starting test: test_eepromfs
Core: IS64: **** Starting test: test_cache_invalidate
Core: IS64: ASSERTION FAILED (test_cache.c:34):
Core: IS64: [----deadbe] != [----aaaaaa]
Core: IS64:      ^^              ^^  idx: 0
Core: IS64: unexpected data in not-invalidated cached cacheline 0 (0/0)
Core: IS64: **** Starting test: test_debug_sdfs
Core: IS64: TEST SKIPPED:
Core: IS64: no SD support
Core: IS64: SKIP
Core: IS64: 
Core: IS64: Testsuite finished in 00:02
Core: IS64: Passed: 11 out of 16 (1 skipped)

When I add an entry for the ROM to the INI and set the SaveType to None:

Core: IS64: **** Starting test: test_exception
Core: IS64: ASSERTION FAILED (test_exception.c:258):
Core: IS64: breakpoint_occured != 1 (0 != 1)
Core: IS64: Breakpoint exception was not triggered
Core: IS64: **** Starting test: test_timer_ticks
Core: IS64: ASSERTION FAILED (test_timer.c:195):
Core: IS64: t5-t0 < 1000
Core: IS64: invalid timer_ticks [start=ffffffe9]: 160000000b - 17000002fd
Core: IS64: **** Starting test: test_timer_oneshot
Core: IS64: **** Starting test: test_timer_slow_callback
Core: IS64: **** Starting test: test_timer_continuous
Core: IS64: **** Starting test: test_timer_mixed
Core: IS64: **** Starting test: test_timer_disabled_start
Core: IS64: **** Starting test: test_timer_disabled_restart
Core: IS64: **** Starting test: test_irq_reentrancy
Core: IS64: **** Starting test: test_eeprom
Core: IS64: EEPROM not detected.
Core: IS64: EEPROM not detected.
Core: IS64: EEPROM not detected.
Core: IS64: EEPROM not detected.
Core: IS64: **** Starting test: test_dfs_read
Core: IS64: **** Starting test: test_dfs_rom_addr
Core: IS64: **** Starting test: test_eepromfs
Core: IS64: TEST SKIPPED:
Core: IS64: EEPROM not found; skipping eepfs tests
Core: IS64: SKIP
Core: IS64: **** Starting test: test_cache_invalidate
Core: IS64: ASSERTION FAILED (test_cache.c:34):
Core: IS64: [----deadbe] != [----aaaaaa]
Core: IS64:      ^^              ^^  idx: 0
Core: IS64: unexpected data in not-invalidated cached cacheline 0 (0/0)
Core: IS64: **** Starting test: test_debug_sdfs
Core: IS64: TEST SKIPPED:
Core: IS64: no SD support
Core: IS64: SKIP
Core: IS64: 
Core: IS64: Testsuite finished in 00:02
Core: IS64: Passed: 10 out of 16 (2 skipped)

The results (for eeprom) look good to me, PR looks good to me in the state it's in currently.

meeq · 2021-06-24T14:56:33Z

src/api/m64p_types.h

    SAVETYPE_SRAM,
    SAVETYPE_FLASH_RAM,
-    SAVETYPE_CONTROLLER_PACK,
+    SAVETYPE_CONTROLLER_PAK,


"Correct" spelling is "Pak": https://en.wikipedia.org/wiki/Nintendo_64_accessories#Controller_Pak

meeq · 2021-06-24T14:59:19Z

src/api/m64p_types.h

@@ -224,11 +224,11 @@ typedef enum

 typedef enum
 {
-    SAVETYPE_EEPROM_4KB = 0,
-    SAVETYPE_EEPROM_16KB,
+    SAVETYPE_EEPROM_4K = 0,


KB is not accurate; the EEPROM sizes are in kilobits: http://n64devkit.square7.ch/pro-man/pro26/26-06.htm

There are two types of EEPROM according to their capacities. One is the 4k-EEPROM. The capacity of this type is 4 kilobits (64 words * 64 bits = 512 bytes). The other is the 16k-EEPROM. The capacity of this type is 16 kilobits (256 words * 64 bits = 2048 bytes).

meeq · 2021-06-24T15:01:10Z

src/device/cart/cart.c

-        rx_buf[0] = (uint8_t)(cart->eeprom.type >> 0);
-        rx_buf[1] = (uint8_t)(cart->eeprom.type >> 8);
-        rx_buf[2] = 0x00;
+        if (cart->eeprom.type) {


This behavior has been tested on real hardware; if there is no EEPROM, the receive buffer is untouched.

src/api/m64p_types.h

loganmc10 · 2021-07-07T19:53:23Z

This LGTM, we've got a lot of pending PRs, if we don't start merging things we'll end up with conflicts soon enough

Jj0YzL5nvJ

I have been using this PR as a patch and have not experienced any noticeable issues.

Narann · 2021-07-07T20:56:52Z

Here we go.

meeq mentioned this pull request Jun 23, 2021

Implement EEPROM Filesystem and test ROM DragonMinded/libdragon#125

Merged

Jj0YzL5nvJ reviewed Jun 24, 2021

View reviewed changes

meeq added 3 commits June 24, 2021 08:46

Fix "No EEPROM present" Joybus status response

d234d2f

Improve accuracy of PIF EEPROM status command

bb1b14f

Based on EEPROM test ROM run on real hardware.

Default "Unknown" save type to EEPROM 4K

5d0fb8c

Revert byte swap on EEPROM Joybus identifiers

58e39dc

meeq changed the title ~~Fix EEPROM detection Joybus status response~~ Make 4K EEPROM default save type; support carts with no EEPROM Jun 24, 2021

meeq added 2 commits June 24, 2021 10:40

Rename save types for correctness

514fa1a

Revert m64p_rom_save_type reordering

700115d

meeq commented Jun 24, 2021

View reviewed changes

Revert loss of whitespace

d5c8dc1

meeq commented Jun 24, 2021

View reviewed changes

bsmiles32 reviewed Jun 24, 2021

View reviewed changes

src/api/m64p_types.h Show resolved Hide resolved

Preserve inaccurate m64p_rom_save_type symbols

e005ab7

Jj0YzL5nvJ approved these changes Jul 7, 2021

View reviewed changes

Narann merged commit 88b4301 into mupen64plus:master Jul 7, 2021

Jj0YzL5nvJ mentioned this pull request Nov 27, 2021

EEPROM16K save type needed for Donkey Kong 64 - Tag Anywhere (V5) (U) to save libretro/mupen64plus-libretro-nx#405

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make 4K EEPROM default save type; support carts with no EEPROM #878

Make 4K EEPROM default save type; support carts with no EEPROM #878

meeq commented Jun 23, 2021

loganmc10 commented Jun 24, 2021

loganmc10 commented Jun 24, 2021 •

edited

loganmc10 commented Jun 24, 2021 •

edited

meeq commented Jun 24, 2021

meeq commented Jun 24, 2021

loganmc10 commented Jun 24, 2021 •

edited

Jj0YzL5nvJ left a comment

meeq commented Jun 24, 2021 •

edited

meeq commented Jun 24, 2021 •

edited

loganmc10 commented Jun 24, 2021 •

edited

meeq Jun 24, 2021

meeq Jun 24, 2021

meeq Jun 24, 2021

loganmc10 commented Jul 7, 2021

Jj0YzL5nvJ left a comment

Narann commented Jul 7, 2021

Make 4K EEPROM default save type; support carts with no EEPROM #878

Make 4K EEPROM default save type; support carts with no EEPROM #878

Conversation

meeq commented Jun 23, 2021

loganmc10 commented Jun 24, 2021

loganmc10 commented Jun 24, 2021 • edited

loganmc10 commented Jun 24, 2021 • edited

meeq commented Jun 24, 2021

meeq commented Jun 24, 2021

loganmc10 commented Jun 24, 2021 • edited

Jj0YzL5nvJ left a comment

Choose a reason for hiding this comment

meeq commented Jun 24, 2021 • edited

meeq commented Jun 24, 2021 • edited

loganmc10 commented Jun 24, 2021 • edited

meeq Jun 24, 2021

Choose a reason for hiding this comment

meeq Jun 24, 2021

Choose a reason for hiding this comment

meeq Jun 24, 2021

Choose a reason for hiding this comment

loganmc10 commented Jul 7, 2021

Jj0YzL5nvJ left a comment

Choose a reason for hiding this comment

Narann commented Jul 7, 2021

loganmc10 commented Jun 24, 2021 •

edited

loganmc10 commented Jun 24, 2021 •

edited

loganmc10 commented Jun 24, 2021 •

edited

meeq commented Jun 24, 2021 •

edited

meeq commented Jun 24, 2021 •

edited

loganmc10 commented Jun 24, 2021 •

edited