OpenBIOS August 2011

openbios@openbios.org

19 participants
88 discussions

Re: [OpenBIOS] [RFC] PowerPC Mac OS on Qemu
by William Hahne 10 Aug '11

10 Aug '11

Please disregard this post. I failed to properly split it up and I have reposted everything in it. William Hahne On Tue, Aug 9, 2011 at 2:46 PM, William Hahne <will07c5(a)gmail.com> wrote: > Hello, > > I am a Google Summer of Code student working on getting PowerPC Mac OS > running on Qemu. While I am not finished I need to upstream what I currently > have before the end of Summer of Code. My patch is for OpenBIOS but I am > cross-posting to both Qemu and OpenBIOS mailing lists since it is useful to > get feedback from both. > > One part of the patch I am particularly concerned about is the patch to > arch/ppc/qemu/init.c to provide a CPU and Timebase frequency. Qemu doesn't > report a CPU frequency and the reported Timebase frequency is too high and > causes the Mac OS scheduler to break. I hard-coded values that work but this > seems like an unacceptable long-term solution to me. A simple test for CPU > frequency might be the best solution here but seems a little redundant as > Mac OS later runs its own test, seemingly only relying on these values > during initialization (I am not 100% sure of this since it crashes before > reaching that point.) Feedback here would be especially appreciated. > > Another thing to note is in drivers/adb_kbd.c. "get-key-map" which returns > a map of the current pressed keys on the keyboard just returns a dummy > value. I felt it was a waste of time making a full implementation when all > it really gets you is the ability to use keyboard shortcuts for verbose or > single user mode. > > Other than those specific concerns I welcome general feedback on the patch, > since as I said I am hoping to get it in before the end of Summer of Code. > > Thanks, > William Hahne >

1 0

[RFC] Incorrect CPU timebase-frequency and clock-frequency in OpenBIOS
by William Hahne 10 Aug '11

10 Aug '11

Hello, I would like to apologize for not splitting up my last post. I am a Google Summer of Code student working on getting Mac OS X to work in Qemu. Currently the cpu device in OpenBIOS has incorrect timebase-frequency and clock-frequency properties. The Mac OS X scheduler relies on these values to calculate the length of a timeslice. Both values are taken from fw_cfg by OpenBIOS. The problem with timebase-frequency is that the value is far too high. Mac OS X depends on this value being sufficiently smaller than clock-frequency. The value for clock-frequency no longer even exists in fw_cfg and is therefore reported as 0. Mac OS X needs some value here even if it doesn't indicate an actual cpu frequency. The patch below for OpenBIOS provides some fake values that work, but since the values are hard-coded into OpenBIOS, it isn't a very nice solution. The best solutions I can think of are to either compute clock-frequency in OpenBIOS or have Qemu provide a value even if its purpose is only to appease Mac OS X or any other OS that depends on the value. If the timebase frequency is correct than it can be lowered otherwise it can be corrected. I am posting this RFC to get thoughts more on what to do with clock-frequency. For reference the timebase frequency is from the fw_cfg value FW_CFG_PPC_TBFREQ and clock frequency is from FW_CFG_PPC_CPUFREQ. Thanks, William Hahne Index: arch/ppc/qemu/init.c =================================================================== --- arch/ppc/qemu/init.c (revision 1041) +++ arch/ppc/qemu/init.c (working copy) @@ -250,12 +250,14 @@ push_str("icache-block-size"); fword("property"); - PUSH(fw_cfg_read_i32(FW_CFG_PPC_TBFREQ)); + /* XXX - Mac OS X needs valid values for timebase-frequency and clock-frequency + for scheduling to work. These are just temporary values and shouldn't be hardcoded here. */ + PUSH(16600000); fword("encode-int"); push_str("timebase-frequency"); fword("property"); - PUSH(fw_cfg_read_i32(FW_CFG_PPC_CPUFREQ)); + PUSH(266000000); fword("encode-int"); push_str("clock-frequency"); fword("property");

1 0

[PATCH 10/10] Adding filll Forth primitive
by William Hahne 09 Aug '11

09 Aug '11

This is a forth primitive that is required by BootX. It just fills some specified memory address and length with longs. Index: kernel/forth.c =================================================================== --- kernel/forth.c (revision 1041) +++ kernel/forth.c (working copy) @@ -1610,6 +1616,20 @@ memset(src, value, count); } +/* + * filll ( addr len byte -- ) + */ +static void filll(void) +{ + ucell value = POP(); + ucell count = POP(); + ucell *dest = (ucell *)cell2pointer(POP()); + + int i; + for (i = 0; i <= count / 4; i++) { + dest[i] = value; + } +} /* * unaligned-w@ ( addr -- w ) @@ -1963,4 +1983,5 @@ do_encode_file, /* $encode-file */ do_debug_xt, /* (debug */ do_debug_off, /* (debug-off) */ + filll /* filll */ }; Index: kernel/bootstrap.c =================================================================== --- kernel/bootstrap.c (revision 1041) +++ kernel/bootstrap.c (working copy) @@ -89,7 +89,7 @@ "here", "here!", "dobranch", "do?branch", "unaligned-w@", "unaligned-w!", "unaligned-l@", "unaligned-l!", "ioc@", "iow@", "iol@", "ioc!", "iow!", "iol!", "i", "j", "call", "sys-debug", - "$include", "$encode-file", "(debug", "(debug-off)" + "$include", "$encode-file", "(debug", "(debug-off)", "filll" }; /*

1 0

[PATCH 9/10] Divide by zero check
by William Hahne 09 Aug '11

09 Aug '11

It is possible for BootX to purposefully divide by zero expecting 0 as the result. Index: kernel/forth.c =================================================================== --- kernel/forth.c (revision 1041) +++ kernel/forth.c (working copy) @@ -1157,6 +1157,12 @@ { const ucell b = POP(); const ducell a = DPOP(); + + if (b == 0) { // can't divide by zero + PUSH(0); + DPUSH(0); + return; + } #ifdef NEED_FAKE_INT128_T if (a.hi != 0) { fprintf(stderr, "mudivmod called (0x%016llx %016llx / 0x%016llx)\n",

1 0

[PATCH 8/10] Check that stdout variable is not null
by William Hahne 09 Aug '11

09 Aug '11

BootX will sometimes set the stdout variable to null and then subsequently call io-emit which can eventually result in an infinite loop. Index: forth/admin/iocontrol.fs =================================================================== --- forth/admin/iocontrol.fs (revision 1041) +++ forth/admin/iocontrol.fs (working copy) @@ -80,6 +80,7 @@ ; : io-emit ( char -- ) + stdout @ 0= if drop exit then \ make sure stdout pointer is not null io-out-char c! io-out-char 1 " write" stdout @ $call-method drop ;

1 0

[PATCH 7/10] Apple local variables implementation in Forth
by William Hahne 09 Aug '11

09 Aug '11

This is an implementation of Apple local variables which is required to execute the Forth scripts in BootX Index: forth/bootstrap/bootstrap.fs =================================================================== --- forth/bootstrap/bootstrap.fs (revision 1041) +++ forth/bootstrap/bootstrap.fs (working copy) @@ -1413,10 +1413,135 @@ here last @ , latest ! \ write backlink and set latest ; +\ +\ Defer required by BootX +\ +defer spin + +\ +\ Local Variables (Apple specific) +\ + +\ +: has-locals ( -- true/false ) + skipws + ib >in @ + c@ 7b <> + if false exit then + ib >in @ + 1+ c@ 20 > + if false exit then + + true + ; + +: comp-str-char ( str len char -- true/false ) + swap 1 <> if 2drop false exit then + swap c@ <> if false exit then + true + ; + +variable locals_wordlist +variable locals_state \ 0 - reading args + \ 1 - ; + \ 2 - reading vars +: read-locals ( -- addr0 .. addrN addrCount ) + has-locals + not if 0 false exit then \ no locals + + 0 locals_state ! + 0 >r + + s" get-current" $find drop execute + s" wordlist" $find drop execute + dup locals_wordlist ! + s" set-current" $find drop execute + + parse-word 2drop \ ditch the { + + begin + parse-word + + 2dup 3b comp-str-char \ check for ; + if 1 locals_state ! then + + 2dup 7d comp-str-char \ check for } + not + while + locals_state @ 1 <> if \ the ; is not a local variable so ignore it + header + + locals_state @ 0= if \ only save the address if it is an arg + r> + here na1+ >r + 1+ >r + then + + 3 , 0 , + reveal + else \ if we hit a ; then move to next state + 2drop + 2 locals_state ! + then + repeat + + 2drop + + s" set-current" $find drop execute + + r> 0 + begin + 2dup + > + while + r> -rot + 1+ + repeat + + drop + + true + ; + +: begin-locals ( addr0 .. addrN count hasLocals -- ) + not if drop exit then + + dup 0> if + 0 do + ['] (lit) , , ['] ! , + loop + else drop then + + s" get-order" $find drop execute + locals_wordlist @ + swap 1+ + s" set-order" $find drop execute + ; + +: end-locals ( -- ) + locals_wordlist @ 0= if exit then + + 0 locals_wordlist ! + + s" get-order" $find drop execute + swap drop 1- + s" set-order" $find drop execute + ; + +: -> parse-word $find drop na1+ + ['] (lit) , , ['] ! , + ; immediate + +\ +\ 7.3.9.1 Defining words +\ + : : - parse-word header - 1 , ] + parse-word >r >r + read-locals + r> r> header + 1 , + begin-locals + ] ; : :noname @@ -1426,6 +1551,7 @@ ; : ; + end-locals ['] (semis) , reveal ['] [ execute ; immediate

1 0

[PATCH 6/10] Wrapped HFS+ incorrectly recognized
by William Hahne 09 Aug '11

09 Aug '11

The code for checking for a wrapped HFS+ volume looks at the wrong offset. Index: fs/hfsplus/hfsp_volume.c =================================================================== --- fs/hfsplus/hfsp_volume.c (revision 1041) +++ fs/hfsplus/hfsp_volume.c (working copy) @@ -302,7 +302,7 @@ os_read(fd, vol, 2, HFSP_BLOCKSZ_BITS); if (__be16_to_cpu(vol[0]) == HFS_VOLHEAD_SIG && - __be16_to_cpu(vol[0x7c]) == HFSP_VOLHEAD_SIG) { + __be16_to_cpu(vol[0x3e]) == HFSP_VOLHEAD_SIG) { ret = -1; } else if (__be16_to_cpu(vol[0]) == HFSP_VOLHEAD_SIG) { ret = -1;

1 0

[PATCH 5/10] Add draw-rectangle to the display package
by William Hahne 09 Aug '11

09 Aug '11

While not necessary for BootX to function this allows the Apple logo to be displayed during boot time. Index: packages/video.c =================================================================== --- packages/video.c (revision 1041) +++ packages/video.c (working copy) @@ -281,6 +281,42 @@ fill_rect( color_ind, x, y, w, h ); } +static void +video_draw_rect( void ) +{ + int h = POP(); + int w = POP(); + int y = POP(); + int x = POP(); + unsigned char* img = (unsigned char*)cell2pointer(POP()); + char *pp; + + if (!video.has_video || x < 0 || y < 0 || w <= 0 || h <= 0 || + x + w > video.fb.w || y + h > video.fb.h) + return; + + pp = (char*)video.fb.mphys + video.fb.rb * y; + for( ; h--; pp += video.fb.rb ) { + int ww = w; + + if( video.fb.depth == 24 || video.fb.depth == 32 ) { + unsigned long *p = (unsigned long*)pp + x; + while( ww-- ) + *p++ = get_color(*img++); + } else if( video.fb.depth == 16 || video.fb.depth == 15 ) { + unsigned short *p = (unsigned short*)pp + x; + while( ww-- ) + *p++ = get_color(*img++); + } else { + char *p = (char *)((unsigned short*)pp + x); + + while( ww-- ) + *p++ = get_color(*img++); + } + } + +} + /* ( addr len -- actual ) */ static void video_write(void) @@ -299,6 +335,7 @@ {"dimensions", video_dimensions }, {"set-colors", video_set_colors }, {"fill-rectangle", video_fill_rect }, + {"draw-rectangle", video_draw_rect }, {"color!", video_color_bang }, {"write", video_write }, };

1 0

[PATCH 4/10] Place framebuffer address in frame-buffer-adr
by William Hahne 09 Aug '11

09 Aug '11

BootX gets the framebuffer address to provide to Mac OS from the frame-buffer-adr value. This value was previously just set to null. Index: packages/video.c =================================================================== --- packages/video.c (revision 1041) +++ packages/video.c (working copy) @@ -332,6 +369,9 @@ video.has_video = 1; video.pal = malloc( 256 * sizeof(unsigned long) ); + PUSH(video.fb.mphys); + feval("to frame-buffer-adr"); + #ifdef CONFIG_PPC s = (video.fb.mphys & 0xfff); size = ((video.fb.h * video.fb.rb + s) + 0xfff) & ~0xfff;

1 0

[PATCH 3/10] Don't stop at APPLE_VOID partition
by William Hahne 09 Aug '11

09 Aug '11

The void partitions can be occasionally found on Mac OS discs (10.2 for example.) Previously mac-parts would just stop at the first void partition meaning it would never try to boot from the HFS partition. Index: packages/mac-parts.c =================================================================== --- packages/mac-parts.c (revision 1041) +++ packages/mac-parts.c (working copy) @@ -165,9 +165,12 @@ for (parnum = 1; parnum <= __be32_to_cpu(par.pmMapBlkCnt); parnum++) { SEEK( bs * parnum ); READ( &par, sizeof(par) ); - if( __be16_to_cpu(par.pmSig) != DESC_PART_SIGNATURE || - !__be32_to_cpu(par.pmPartBlkCnt) ) + if( __be16_to_cpu(par.pmSig) != DESC_PART_SIGNATURE) { break; + } + if ( !__be32_to_cpu(par.pmPartBlkCnt) ) { + continue; /* Just a void partition, ignore it. */ + } DPRINTF("found partition type: %s with status %x\n", par.pmPartType, __be32_to_cpu(par.pmPartStatus));

1 0

← Newer
1
2
3
4
5
6
7
8
9
Older →