June 2007 - coreboot - mail.coreboot.org

[LinuxBIOS] [RFC] v3: dts fixes and questions
by Uwe Hermann July 3, 2007

July 3, 2007

Here's a patch which streamlines the dts contents a bit: - Every "thing" should have an "enable;" line (or am I missing some reason why they shouldn't?) - Every "thing" should have the "enable;" as first entry, if it has one. - Constructors should come before other stuff, but after the "enable;". Comments? Does the order in which variables are defined (e.g. ide0_enable) and in which the constructors are called matter? I.e. is there a semantic difference between { ide0_enable = "0"; ide1_enable = "0"; constructor = "i82371eb_constructors"; }; and { constructor = "i82371eb_constructors"; ide0_enable = "0"; ide1_enable = "0"; }; ? See patch for my proposal, but please note that it may be incorrent, that's why I'm asking... Uwe. -- http://www.hermann-uwe.de | http://www.holsham-traders.de http://www.crazy-hacks.org | http://www.unmaintained-free-software.org

5 12

Re: [LinuxBIOS] r387 - in LinuxBIOSv3/arch/x86: . geodelx
by Peter Stuge July 3, 2007

July 3, 2007

On Wed, Jun 27, 2007 at 11:07:10PM +0200, svn(a)openbios.org wrote: > +++ LinuxBIOSv3/arch/x86/Makefile 2007-06-27 21:07:10 UTC (rev 387) .. > +$(obj)/arch/x86/geodelx/stage0.o: $(src)/arch/x86/geodelx/stage0.S > + $(Q)mkdir -p $(obj)/arch/x86/geodelx Uh? The rule depends on a file in a directory created in the rule? > + $(Q)printf " CC $(subst $(shell pwd)/,,$(@))\n" > + $(Q)$(CC) -E $(LINUXBIOSINCLUDE) $< \ > + -o $(obj)/arch/x86/stage0_asm.s -DBOOTBLK=0x1f00 \ > + -DRESRVED=0xf0 -DDATE=\"`date +%Y/%m/%d`\" Maybe even the svn rev date? > + /* Setup access to the cache for under 640K. Note MC not setup yet. */ > + msr.hi = 0x20000000; > + msr.lo = 0xfff80; > + wrmsr(MSR_GLIU0 + 0x20, msr); > + > + msr.hi = 0x20000000; > + msr.lo = 0x80fffe0; > + wrmsr(MSR_GLIU0 + 0x21, msr); > + > + msr.hi = 0x20000000; > + msr.lo = 0xfff80; > + wrmsr(MSR_GLIU1 + 0x20, msr); > + > + msr.hi = 0x20000000; > + msr.lo = 0x80fffe0; > + wrmsr(MSR_GLIU1 + 0x21, msr); No nice #defines around for these? > +/** > + * start_time1 Starts Timer 1 for port 61 use. FIXME try to figure > + * out what these values mean. > + */ > +void start_timer1(void) > +{ > + outb(0x56, 0x43); > + outb(0x12, 0x41); > +} 0x43 is PIT command/control. 0x41 is PIT counter 1. 0x56 means write counter 1 lower 8 bits in next IO, set the counter mode to square wave generator (count down to 0 from programmed value twice in a row, alternating the output signal) counting in 16-bit binary mode. 0x12 is written to the counter. Used for RAM refresh on XT/AT but the port 61 reference indicates it has to do with the (emulated?) keyboard controller. > + msr_t msrGlcpSysRstpll; This guy could use a better name. At least no caps. > + msrGlcpSysRstpll.lo &= > + ~(0xFF << RSTPPL_LOWER_HOLD_COUNT_SHIFT); > + msrGlcpSysRstpll.lo |= > + (0xDE << RSTPPL_LOWER_HOLD_COUNT_SHIFT); Why 0xDE ? > + /* You should never get here..... The chip has reset. */ > + printk(BIOS_EMERG,"CONFIGURING PLL FAILURE -- HALT\n"); > + post_code(POST_PLL_RESET_FAIL); > + __asm__ __volatile__("hlt\n"); How about that hlt() function? Remember we want to put panic room code there when we have too much spare time. > +/** > + * Return the CPU clock rate. Rates in this system are always returned > + * as multkiples of 33 Mhz. > + * > + */ > +u32 cpu_speed(void) Is there a doxygen syntax for the return value? Also please say which unit the value is in. MHz? > +/** > + * Return the Geode Link clock rate. Rates in this system are always > + * returned as multkiples of 33 Mhz. > + * > + */ > +u32 geode_link_speed(void) Ditto. > + speed = ((((msr.hi >> RSTPLL_UPPER_GLMULT_SHIFT) & 0x1F) + 1) * 333) / 10; > + if ((((((msr.hi >> RSTPLL_UPPER_GLMULT_SHIFT) & 0x1F) + 1) * 333) % 10) > 5) { Maybe that 0x1F is a #define? > +/** > + * Return the PCI bus clock rate. Rates in this system are always > + * returned as multkiples of 33 Mhz. > + * > + */ > +u32 pci_speed(void) Doxygen return value comment here too? > + msrnum = GLCP_DELAY_CONTROLS; > + msr = rdmsr(msrnum); > + if (msr.lo & ~(0x7C0)) { > + return; > + } The juju must stay but 0x7c0 seems like a #define? > + if (spdbyte0 == 0 || spdbyte1 == 0) { > + /* one dimm solution */ > + if (spdbyte1 == 0) { > + msr.hi |= 0x000800000; > + } > + spdbyte0 += spdbyte1; > + if (spdbyte0 > 8) { > + /* large dimm */ > + if (glspeed < 334) { > + msr.hi |= 0x0837100AA; > + msr.lo |= 0x056960004; > + } else { > + msr.hi |= 0x082710055; > + msr.lo |= 0x056960004; > + } > + } else if (spdbyte0 > 4) { > + /* medium dimm */ > + if (glspeed < 334) { > + msr.hi |= 0x0837100AA; > + msr.lo |= 0x056960004; > + } else { > + msr.hi |= 0x0827100AA; > + msr.lo |= 0x056960004; > + } > + } else { > + /* small dimm */ > + if (glspeed < 334) { > + msr.hi |= 0x0837100FF; > + msr.lo |= 0x056960004; > + } else { > + msr.hi |= 0x0827100FF; > + msr.lo |= 0x056960004; > + } > + } > + } else { > + /* two dimm solution */ > + spdbyte0 += spdbyte1; > + if (spdbyte0 > 24) { > + /* huge dimms */ > + if (glspeed < 334) { > + msr.hi |= 0x0B37100A5; > + msr.lo |= 0x056960004; > + } else { > + msr.hi |= 0x0B2710000; > + msr.lo |= 0x056960004; > + } > + } else if (spdbyte0 > 16) { > + /* large dimms */ > + if (glspeed < 334) { > + msr.hi |= 0x0B37100A5; > + msr.lo |= 0x056960004; > + } else { > + msr.hi |= 0x0B27100A5; > + msr.lo |= 0x056960004; > + } > + } else if (spdbyte0 >= 8) { > + /* medium dimms */ > + if (glspeed < 334) { > + msr.hi |= 0x0937100A5; > + msr.lo |= 0x056960004; > + } else { > + msr.hi |= 0x0C27100A5; > + msr.lo |= 0x056960004; > + } > + } else { > + /* small dimms */ > + if (glspeed < 334) { > + msr.hi |= 0x0837100A5; > + msr.lo |= 0x056960004; > + } else { > + msr.hi |= 0x082710000; > + msr.lo |= 0x056960004; > + } > + } Each if statements just changes a few bits. Couldn't this be made more readable? > +DCacheSetupBad: > + hlt /* Issues */ > + jmp DCacheSetupBad Hehe, yes, we have issues. Should we aim for panic room code in assembly? > +lout: > + /* Restore the BIST result. */ > + movl %ebp, %eax > + /* We need to set ebp? No need. */ > + movl %esp, %ebp > + pushl %eax /* BIST */ > + call stage1_main > + /* We will not go back. */ I remember commenting on this before; please handle a return or change it to a jmp. (I like the latter.) > +fixed_mtrr_msr: > + .long 0x250, 0x258, 0x259 > + .long 0x268, 0x269, 0x26A > + .long 0x26B, 0x26C, 0x26D > + .long 0x26E, 0x26F > +var_mtrr_msr: > + .long 0x200, 0x201, 0x202, 0x203 > + .long 0x204, 0x205, 0x206, 0x207 ..or this would be executed on return.. //Peter

2 1

[LinuxBIOS] [PATCH][v3] small fix for nrv2b
by Patrick Georgi July 3, 2007

July 3, 2007

Hi, this is actually a brown paper bag patch, as my last patch managed to break the build of nrv2b. It's basically some reordering and an additional preprocessor define on the command line. I'm sorry that this issue slipped through (I didn't notice that my build didn't cover nrv2b anymore) Regards, Patrick Georgi Signed-off-by: Patrick Georgi <patrick(a)georgi-clan.de> Index: util/nrv2b/nrv2b.c =================================================================== --- util/nrv2b/nrv2b.c (revision 367) +++ util/nrv2b/nrv2b.c (working copy) @@ -1309,9 +1309,47 @@ out = malloc(*out_len); ucl_nrv2b_99_compress(in, in_len, out, out_len, 0 ); } +#endif +#ifdef DECODE + +#define GETBIT_8(bb, src, ilen) \ + (((bb = bb & 0x7f ? bb*2 : ((unsigned)src[ilen++]*2+1)) >> 8) & 1) + +#define GETBIT_LE16(bb, src, ilen) \ + (bb*=2,bb&0xffff ? (bb>>16)&1 : (ilen+=2,((bb=(src[ilen-2]+src[ilen-1]*256u)*2+1)>>16)&1)) + +#define GETBIT_LE32(bb, src, ilen) \ + (bc > 0 ? ((bb>>--bc)&1) : (bc=31,\ + bb=*(const uint32_t *)((src)+ilen),ilen+=4,(bb>>31)&1)) + +#if ENDIAN == 0 && BITSIZE == 8 +#define GETBIT(bb, src, ilen) GETBIT_8(bb, src, ilen) +#endif +#if ENDIAN == 0 && BITSIZE == 16 +#define GETBIT(bb, src, ilen) GETBIT_LE16(bb, src, ilen) +#endif +#if ENDIAN == 0 && BITSIZE == 32 +#define GETBIT(bb, src, ilen) GETBIT_LE32(bb, src, ilen) +#endif + +#ifndef GETBIT +#error "Bad Combination of ENDIAN and BITSIZE values specified" +#endif + +#undef SAFE + +#ifdef SAFE +#define FAIL(x,r) if (x) { Error(r); } +#else +#define FAIL(x,r) +#endif + +#ifdef COMPACT void do_nrv2b_uncompress(char* dst, char* src, unsigned long len) { unsigned long ilen = 0, olen = 0, last_m_off = 1; + uint32_t bb = 0; + unsigned bc = 0; for (;;) { unsigned int m_off, m_len; while (GETBIT(bb, src, ilen)) { @@ -1359,42 +1397,7 @@ } FAIL(ilen < src_len, "input not consumed"); } -#endif - -#ifdef DECODE - -#define GETBIT_8(bb, src, ilen) \ - (((bb = bb & 0x7f ? bb*2 : ((unsigned)src[ilen++]*2+1)) >> 8) & 1) - -#define GETBIT_LE16(bb, src, ilen) \ - (bb*=2,bb&0xffff ? (bb>>16)&1 : (ilen+=2,((bb=(src[ilen-2]+src[ilen-1]*256u)*2+1)>>16)&1)) - -#define GETBIT_LE32(bb, src, ilen) \ - (bc > 0 ? ((bb>>--bc)&1) : (bc=31,\ - bb=*(const uint32_t *)((src)+ilen),ilen+=4,(bb>>31)&1)) - -#if ENDIAN == 0 && BITSIZE == 8 -#define GETBIT(bb, src, ilen) GETBIT_8(bb, src, ilen) -#endif -#if ENDIAN == 0 && BITSIZE == 16 -#define GETBIT(bb, src, ilen) GETBIT_LE16(bb, src, ilen) -#endif -#if ENDIAN == 0 && BITSIZE == 32 -#define GETBIT(bb, src, ilen) GETBIT_LE32(bb, src, ilen) -#endif - -#ifndef GETBIT -#error "Bad Combination of ENDIAN and BITSIZE values specified" -#endif - -#undef SAFE - -#ifdef SAFE -#define FAIL(x,r) if (x) { Error(r); } #else -#define FAIL(x,r) -#endif - void Decode(void) /* recover */ { uint32_t tw; @@ -1494,6 +1497,7 @@ free(dst); } #endif +#endif #ifdef MAIN int main(int argc, char *argv[]) Index: util/nrv2b/Makefile =================================================================== --- util/nrv2b/Makefile (revision 367) +++ util/nrv2b/Makefile (working copy) @@ -25,7 +25,7 @@ $(obj)/util/nrv2b/nrv2b-compress.o: nrv2bdir $(src)/util/nrv2b/nrv2b.c $(Q)printf " HOSTCC $(subst $(shell pwd)/,,$(@))\n" - $(Q)$(HOSTCC) $(HOSTCFLAGS) -DENCODE -DVERBOSE -DCOMPACT -DNDEBUG -DBITSIZE=32 -DENDIAN=0 -c -o $@ $(src)/util/nrv2b/nrv2b.c + $(Q)$(HOSTCC) $(HOSTCFLAGS) -DENCODE -DDECODE -DVERBOSE -DCOMPACT -DNDEBUG -DBITSIZE=32 -DENDIAN=0 -c -o $@ $(src)/util/nrv2b/nrv2b.c nrv2bdir: $(Q)printf " BUILD NRV2B\n"

2 1

Re: [LinuxBIOS] r413 - in LinuxBIOSv3: arch/x86 include mainboard/artecgroup/dbe61 southbridge/amd/cs5536 superio/winbond/w83627hf
by Peter Stuge July 3, 2007

July 3, 2007

On Fri, Jun 29, 2007 at 04:36:03PM +0200, svn(a)openbios.org wrote: > +#define ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0])) Need extra () around a or no? sizeof((a)) / sizeof(((a))[0])) ? //Peter

2 1

[LinuxBIOS] [PATCH] v3 - spinlock cleanup
by Stefan Reinauer July 3, 2007

July 3, 2007

* fix copyright headers to v2 only as suggested by Uwe * drop silly code when not using SMP * fix device/device.c * drop spinlock_t, use struct spinlock instead. Signed-off-by: Stefan Reinauer <stepan(a)coresystems.de> Index: include/spinlock.h =================================================================== --- include/spinlock.h (revision 427) +++ include/spinlock.h (working copy) @@ -2,11 +2,11 @@ * This file is part of the LinuxBIOS project. * * Copyright (C) 2001 Linux Networx + * Copyright (C) 2007 coresystems GmbH * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * the Free Software Foundation; version 2 of the License. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of @@ -22,23 +22,17 @@ #define SPINLOCK_H #if defined(CONFIG_SMP) && CONFIG_SMP == 1 + #include <arch/spinlock.h> + #else /* !CONFIG_SMP */ -/* Most GCC versions have a nasty bug with empty initializers */ -#if (__GNUC__ > 2) -typedef struct { } spinlock_t; -#define SPIN_LOCK_UNLOCKED (spinlock_t) { } -#else -typedef struct { int gcc_is_buggy; } spinlock_t; -#define SPIN_LOCK_UNLOCKED (spinlock_t) { 0 } -#endif - #define barrier() do {} while(0) #define spin_is_locked(lock) 0 #define spin_unlock_wait(lock) do {} while(0) #define spin_lock(lock) do {} while(0) #define spin_unlock(lock) do {} while(0) +#define spin_define(lock) /* empty */ #endif #endif /* SPINLOCK_H */ Index: include/arch/x86/arch/spinlock.h =================================================================== --- include/arch/x86/arch/spinlock.h (revision 427) +++ include/arch/x86/arch/spinlock.h (working copy) @@ -2,11 +2,11 @@ * This file is part of the LinuxBIOS project. * * Copyright (C) 2001 Linux Networx + * Copyright (C) 2007 coresystems GmbH * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * the Free Software Foundation; version 2 of the License. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of @@ -25,12 +25,12 @@ * Your basic SMP spinlocks, allowing only a single CPU anywhere */ -typedef struct { +struct spinlock { volatile unsigned int lock; -} spinlock_t; +}; -#define SPIN_LOCK_UNLOCKED (spinlock_t) { 1 } +#define SPIN_LOCK_UNLOCKED (struct spinlock) { 1 } /* * Simple spin lock operations. There are two variants, one clears IRQ's @@ -60,18 +60,21 @@ #define spin_unlock_string \ "movb $1,%0" -static inline __attribute__((always_inline)) void spin_lock(spinlock_t *lock) +static inline __attribute__((always_inline)) void spin_lock(struct spinlock *lock) { __asm__ __volatile__( spin_lock_string :"=m" (lock->lock) : : "memory"); } -static inline __attribute__((always_inline)) void spin_unlock(spinlock_t *lock) +static inline __attribute__((always_inline)) void spin_unlock(struct spinlock *lock) { __asm__ __volatile__( spin_unlock_string :"=m" (lock->lock) : : "memory"); } +#define spin_define(spin) static struct spinlock spin = SPIN_LOCK_UNLOCKED + + #endif /* ARCH_SPINLOCK_H */ Index: device/device.c =================================================================== --- device/device.c (revision 427) +++ device/device.c (working copy) @@ -183,7 +183,9 @@ * @return Pointer to the newly created device structure. * @see device_path */ -// static spinlock_t dev_lock = SPIN_LOCK_UNLOCKED; + +spin_define(dev_lock); + struct device *alloc_dev(struct bus *parent, struct device_path *path, struct device_id *devid) { -- coresystems GmbH • Brahmsstr. 16 • D-79104 Freiburg i. Br. Tel.: +49 761 7668825 • Fax: +49 761 7664613 Email: info(a)coresystems.de • http://www.coresystems.de/ Registergericht: Amtsgericht Freiburg • HRB 7656 Geschäftsführer: Stefan Reinauer • Ust-IdNr.: DE245674866

3 7

[LinuxBIOS] [PATCH] v3 - put lar size in the bootblock
by Stefan Reinauer July 3, 2007

July 3, 2007

Hi, this patch puts the lar size in the bootblock and reads it from there. Why? This way we don't need to recompile the image when the size of the LinuxBIOS image changes. This alows building images for 50 motherboards and equipping each with 10 payloads, resulting in 500 images while you only have to build each payload once and each motherboard, too. There's also a small "fix" allowing the compression type to be case insensitive. Not really relevant I guess. Signed-off-by: Stefan Reinauer <stepan(a)coresystems.de> Index: util/lar/bootblock.c =================================================================== --- util/lar/bootblock.c (revision 426) +++ util/lar/bootblock.c (working copy) @@ -74,16 +74,16 @@ int fixup_bootblock(void) { - /* Per definition the bootblock starts with 256 empty bytes. - * These are utilized to make the bootblock part of a lar file, - * and store the image size. - * - * We will also calculate a checksum here. - */ + int i; + uint32_t *size_pos; - /* first try. Clear out ugly left-over from ld hack */ - bootblock_code[bootblock_len - 13] = '\0'; - bootblock_code[bootblock_len - 12] = '\0'; + /* This cleans out the area after the reset vector */ + for(i=13; i>0; i--) + bootblock_code[bootblock_len - i] = '\0'; + + /* add lar size to image */ + size_pos=(uint32_t *)(bootblock_code+bootblock_len-12); + size_pos[0] = get_larsize(); return 0; } Index: util/lar/lar.c =================================================================== --- util/lar/lar.c (revision 426) +++ util/lar/lar.c (working copy) @@ -96,10 +96,10 @@ larmode = CREATE; break; case 'C': - if (strcmp("lzma", optarg) == 0) { + if (strcasecmp("lzma", optarg) == 0) { algo = lzma; } - if (strcmp("nrv2b", optarg) == 0) { + if (strcasecmp("nrv2b", optarg) == 0) { algo = nrv2b; } break; Index: arch/x86/stage1.c =================================================================== --- arch/x86/stage1.c (revision 426) +++ arch/x86/stage1.c (working copy) @@ -101,8 +101,12 @@ // FIXME this should be defined in the VPD area // but NOT IN THE CODE. - archive.len=(CONFIG_LINUXBIOS_ROMSIZE_KB)*1024; - archive.start=(void *)(0UL-(CONFIG_LINUXBIOS_ROMSIZE_KB*1024)); + /* The len field starts behind the reset vector on x86. + * The start is not correct for all platforms. sc520 will + * need some hands on here. + */ + archive.len = *(u32 *)0xfffffff4; + archive.start =(void *)(0UL-archive.len); // FIXME check integrity -- coresystems GmbH • Brahmsstr. 16 • D-79104 Freiburg i. Br. Tel.: +49 761 7668825 • Fax: +49 761 7664613 Email: info(a)coresystems.de • http://www.coresystems.de/

5 6

[LinuxBIOS] ELF loader in v3
by Patrick Georgi July 2, 2007

July 2, 2007

Hi, I got quite far with adding decompression to lar (lzma'd data runs fine, nrv2b decompression seems to hang), but ran into issues that need discussion. Currently, two files in the lar archive are parsed directly out of rom (using the find_file function which returns a pointer to there). One of them is "option_table", a 900 byte file (which compresses to 350 or so), where I could live with it staying uncompressed in image. The other one is the payload. My current work around, is to just define a place in memory where to decode it to and then load the ELF image from there. This complicates ELF loading as now there is a risk that the raw ELF file and the ELF image overlap. (Right now, I just use an address that I know is beyond the ELF's range) I propose a new API for the ELF loader that hides this issue and ensures that decompression works fine. It would be used like this: copy_elf(&archive, "normal/payload", &entrypoint, &last_address) This function looks up "normal/payload" in the archive starting at archive, and returns success in the return value. entrypoint could then be used to jump into the image. last_address contains the last byte used by the ELF image. The rationale for returning the entrypoint and last_address is to allow further loads, so multipart payloads can be used, which provides a cleaner and simpler way to allow modular payloads. Specifically, I'd like to see this for the grub2 support I work on, as it's very trivial to just push the modules as "payload.data" into the lar archive, load it right behind the ELF image and have grub2 find its modules at _end. Regards, Patrick Georgi

8 24

Re: [LinuxBIOS] r418 - in LinuxBIOSv3: arch/x86 device include include/arch/x86 include/arch/x86/arch include/device lib util/dtc
by Uwe Hermann July 1, 2007

July 1, 2007

On Fri, Jun 29, 2007 at 06:57:23PM +0200, svn(a)openbios.org wrote: > * start using arch/foo.h again instead of archfoo.h (trivial) > * make constructor an initializer. I don't understand. What's an initializer in this context? > + if(c && c->ops && c->ops->constructor) > + c->ops->constructor(dev, c); > + else > + printk(BIOS_INFO, "No constructor called for %s.\n", > + dev_id_string(&c->id)); BIOS_DEBUG? > Copied: LinuxBIOSv3/include/arch/x86/arch/elf.h (from rev 417, LinuxBIOSv3/include/arch/x86/archelf.h) > =================================================================== > --- LinuxBIOSv3/include/arch/x86/arch/elf.h (rev 0) > +++ LinuxBIOSv3/include/arch/x86/arch/elf.h 2007-06-29 16:57:23 UTC (rev 418) Missing license header. Trivial file, so feel free to add yourself as copyright owner. > @@ -0,0 +1,8 @@ > +#ifndef ARCH_ELF_H > +#define ARCH_ELF_H Following our current guidelines this should be ARCH_X86_ARCH_ELF_H, which looks a bit stupid, but I think it's necessary. ARCH_ELF_H doesn't work as we'd have another ARCH_ELF_H if we add another architecture... > + > +#define ELF_CLASS ELFCLASS32 > +#define ELF_DATA ELFDATA2LSB > +#define ELF_ARCH EM_386 > + > +#endif /* ARCH_ELF_H */ > > Added: LinuxBIOSv3/include/arch/x86/arch/spinlock.h > =================================================================== > --- LinuxBIOSv3/include/arch/x86/arch/spinlock.h (rev 0) > +++ LinuxBIOSv3/include/arch/x86/arch/spinlock.h 2007-06-29 16:57:23 UTC (rev 418) > @@ -0,0 +1,83 @@ > +/* > + * This file is part of the LinuxBIOS project. > + * > + * Copyright (C) 2001 Linux Networx > + * > + * This program is free software; you can redistribute it and/or modify > + * it under the terms of the GNU General Public License as published by > + * the Free Software Foundation; either version 2 of the License, or > + * (at your option) any later version. This should be GPLv2, I think. We know from email conversation with Eric Biederman / Linux Networx that their code is GPLv2 only. But this file is based on Linux code anyway, thus definately GPLv2, I'll update the header and HACKING accordingly (later). > + * > + * This program is distributed in the hope that it will be useful, > + * but WITHOUT ANY WARRANTY; without even the implied warranty of > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the > + * GNU General Public License for more details. > + * > + * You should have received a copy of the GNU General Public License > + * along with this program; if not, write to the Free Software > + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA > + */ > + > +#ifndef ARCH_SPINLOCK_H > +#define ARCH_SPINLOCK_H See above. > + > +/* > + * Your basic SMP spinlocks, allowing only a single CPU anywhere > + */ > + > +typedef struct { > + volatile unsigned int lock; Why volatile? I think the Linux file doesn't have that. > +} spinlock_t; Drop spinlock_t and use 'struct spinlock' as we do for other structs? > + > + > +#define SPIN_LOCK_UNLOCKED (spinlock_t) { 1 } > + > +/* > + * Simple spin lock operations. There are two variants, one clears IRQ's > + * on the local processor, one does not. > + * > + * We make no fairness assumptions. They have a cost. > + */ > +#define barrier() __asm__ __volatile__("": : :"memory") #define barrier() (__asm__ __volatile__("": : :"memory")) (better safe than sorry) > +#define spin_is_locked(x) (*(volatile char *)(&(x)->lock) <= 0) > +#define spin_unlock_wait(x) do { barrier(); } while(spin_is_locked(x)) #define spin_unlock_wait(x) (do { barrier(); } while(spin_is_locked(x))) > + > +#define spin_lock_string \ > + "\n1:\t" \ > + "lock ; decb %0\n\t" \ > + "js 2f\n" \ > + ".section .text.lock,\"ax\"\n" \ > + "2:\t" \ > + "cmpb $0,%0\n\t" \ > + "rep;nop\n\t" \ > + "jle 2b\n\t" \ > + "jmp 1b\n" \ > + ".previous" > + > +/* > + * This works. Despite all the confusion. > + */ > +#define spin_unlock_string \ > + "movb $1,%0" Shall we drop these #defines? They're only used in one place and a small comment which says what the asm does should be enough. > +static inline __attribute__((always_inline)) void spin_lock(spinlock_t *lock) > +{ > + __asm__ __volatile__( > + spin_lock_string > + :"=m" (lock->lock) : : "memory"); > +} > + > +static inline __attribute__((always_inline)) void spin_unlock(spinlock_t *lock) > +{ > + __asm__ __volatile__( > + spin_unlock_string > + :"=m" (lock->lock) : : "memory"); > +} > Modified: LinuxBIOSv3/util/dtc/dtc-parser.y > =================================================================== > --- LinuxBIOSv3/util/dtc/dtc-parser.y 2007-06-29 15:32:19 UTC (rev 417) > +++ LinuxBIOSv3/util/dtc/dtc-parser.y 2007-06-29 16:57:23 UTC (rev 418) > @@ -142,7 +142,7 @@ > } > ')' ';' { > /* convention: first property is labeled with path */ > - $6->label = strdup($3.val); > + $6->label = strdup((char *)$3.val); Is this the correct fix for the warning? I didn't notice this warning before so it must have been introduced by some recent change? Uwe. -- http://www.hermann-uwe.de | http://www.holsham-traders.de http://www.crazy-hacks.org | http://www.unmaintained-free-software.org

4 5

[LinuxBIOS] r429 - in LinuxBIOSv3: device include include/arch/x86/arch
by svn＠openbios.org July 1, 2007

July 1, 2007

Author: stepan Date: 2007-07-01 00:55:10 +0200 (Sun, 01 Jul 2007) New Revision: 429 Modified: LinuxBIOSv3/device/device.c LinuxBIOSv3/include/arch/x86/arch/spinlock.h LinuxBIOSv3/include/spinlock.h Log: * fix copyright headers to v2 only as suggested by Uwe * drop silly code when not using SMP * fix device/device.c * drop spinlock_t, use struct spinlock instead. Signed-off-by: Stefan Reinauer <stepan(a)coresystems.de> Acked-by: Ronald G. Minnich <rminnich(a)gmail.com> one kitten died already! ;-) Modified: LinuxBIOSv3/device/device.c =================================================================== --- LinuxBIOSv3/device/device.c 2007-06-30 22:09:57 UTC (rev 428) +++ LinuxBIOSv3/device/device.c 2007-06-30 22:55:10 UTC (rev 429) @@ -183,7 +183,9 @@ * @return Pointer to the newly created device structure. * @see device_path */ -// static spinlock_t dev_lock = SPIN_LOCK_UNLOCKED; + +spin_define(dev_lock); + struct device *alloc_dev(struct bus *parent, struct device_path *path, struct device_id *devid) { Modified: LinuxBIOSv3/include/arch/x86/arch/spinlock.h =================================================================== --- LinuxBIOSv3/include/arch/x86/arch/spinlock.h 2007-06-30 22:09:57 UTC (rev 428) +++ LinuxBIOSv3/include/arch/x86/arch/spinlock.h 2007-06-30 22:55:10 UTC (rev 429) @@ -2,11 +2,11 @@ * This file is part of the LinuxBIOS project. * * Copyright (C) 2001 Linux Networx + * Copyright (C) 2007 coresystems GmbH * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * the Free Software Foundation; version 2 of the License. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of @@ -25,12 +25,12 @@ * Your basic SMP spinlocks, allowing only a single CPU anywhere */ -typedef struct { +struct spinlock { volatile unsigned int lock; -} spinlock_t; +}; -#define SPIN_LOCK_UNLOCKED (spinlock_t) { 1 } +#define SPIN_LOCK_UNLOCKED (struct spinlock) { 1 } /* * Simple spin lock operations. There are two variants, one clears IRQ's @@ -60,18 +60,21 @@ #define spin_unlock_string \ "movb $1,%0" -static inline __attribute__((always_inline)) void spin_lock(spinlock_t *lock) +static inline __attribute__((always_inline)) void spin_lock(struct spinlock *lock) { __asm__ __volatile__( spin_lock_string :"=m" (lock->lock) : : "memory"); } -static inline __attribute__((always_inline)) void spin_unlock(spinlock_t *lock) +static inline __attribute__((always_inline)) void spin_unlock(struct spinlock *lock) { __asm__ __volatile__( spin_unlock_string :"=m" (lock->lock) : : "memory"); } +#define spin_define(spin) static struct spinlock spin = SPIN_LOCK_UNLOCKED + + #endif /* ARCH_SPINLOCK_H */ Modified: LinuxBIOSv3/include/spinlock.h =================================================================== --- LinuxBIOSv3/include/spinlock.h 2007-06-30 22:09:57 UTC (rev 428) +++ LinuxBIOSv3/include/spinlock.h 2007-06-30 22:55:10 UTC (rev 429) @@ -2,11 +2,11 @@ * This file is part of the LinuxBIOS project. * * Copyright (C) 2001 Linux Networx + * Copyright (C) 2007 coresystems GmbH * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * the Free Software Foundation; version 2 of the License. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of @@ -22,23 +22,17 @@ #define SPINLOCK_H #if defined(CONFIG_SMP) && CONFIG_SMP == 1 + #include <arch/spinlock.h> + #else /* !CONFIG_SMP */ -/* Most GCC versions have a nasty bug with empty initializers */ -#if (__GNUC__ > 2) -typedef struct { } spinlock_t; -#define SPIN_LOCK_UNLOCKED (spinlock_t) { } -#else -typedef struct { int gcc_is_buggy; } spinlock_t; -#define SPIN_LOCK_UNLOCKED (spinlock_t) { 0 } -#endif - #define barrier() do {} while(0) #define spin_is_locked(lock) 0 #define spin_unlock_wait(lock) do {} while(0) #define spin_lock(lock) do {} while(0) #define spin_unlock(lock) do {} while(0) +#define spin_define(lock) /* empty */ #endif #endif /* SPINLOCK_H */

1 0

Re: [LinuxBIOS] r427 - in LinuxBIOSv3: arch/x86 util/lar
by Uwe Hermann July 1, 2007

July 1, 2007

On Sun, Jul 01, 2007 at 12:03:39AM +0200, svn(a)openbios.org wrote: > Author: stepan > Date: 2007-07-01 00:03:39 +0200 (Sun, 01 Jul 2007) > New Revision: 427 Patch sent: Sat, 30 Jun 2007 23:58:15 +0200 ACK: Sat, 30 Jun 2007 15:00:18 -0700 Commit: Sun, 1 Jul 2007 00:03:39 +0200 2:03 minutes til ack, 3:21 til commit. *cough* My theory: you guys are stuck in a room with lots of kittens and a terrorist. Every time a patch doesn't get committed in less than 5 minutes the terrorist kills a kitten. You must love kittens quite a lot :) Uwe. -- http://www.hermann-uwe.de | http://www.holsham-traders.de http://www.crazy-hacks.org | http://www.unmaintained-free-software.org

2 1