Index: qemu/configure =================================================================== --- qemu.orig/configure 2006-08-26 16:31:53.000000000 +0100 +++ qemu/configure 2006-08-26 16:31:53.000000000 +0100 @@ -97,6 +97,7 @@ build_docs="no" build_acpi_tables="no" uname_release="" +nptl="yes" # OS specific targetos=`uname -s` @@ -243,6 +244,8 @@ ;; --enable-iasl) build_acpi_tables="yes" ;; + --disable-nptl) nptl="no" + ;; esac done @@ -441,6 +444,23 @@ fi fi +# check NPTL support +cat > $TMPC < +void foo() +{ +#ifndef CLONE_SETTLS +#error bork +#endif +} +EOF + +if $cc -c -o $TMPO $TMPC 2> /dev/null ; then + : +else + nptl="no" +fi + ########################################## # SDL probe @@ -559,6 +579,7 @@ fi echo "FMOD support $fmod $fmod_support" echo "kqemu support $kqemu" +echo "NPTL support $nptl" echo "Documentation $build_docs" [ ! -z "$uname_release" ] && \ echo "uname -r $uname_release" @@ -880,6 +901,14 @@ echo "SDL_CFLAGS=`$sdl_config --cflags`" >> $config_mak fi fi +else + if test "$nptl" = "yes" ; then + case "$target_cpu" in + arm | armeb) + echo "#define USE_NPTL 1" >> $config_h + ;; + esac + fi fi if test "$cocoa" = "yes" ; then Index: qemu/exec-all.h =================================================================== --- qemu.orig/exec-all.h 2006-08-26 16:28:32.000000000 +0100 +++ qemu/exec-all.h 2006-08-26 16:31:53.000000000 +0100 @@ -347,163 +347,7 @@ extern CPUReadMemoryFunc *io_mem_read[IO_MEM_NB_ENTRIES][4]; extern void *io_mem_opaque[IO_MEM_NB_ENTRIES]; -#ifdef __powerpc__ -static inline int testandset (int *p) -{ - int ret; - __asm__ __volatile__ ( - "0: lwarx %0,0,%1\n" - " xor. %0,%3,%0\n" - " bne 1f\n" - " stwcx. %2,0,%1\n" - " bne- 0b\n" - "1: " - : "=&r" (ret) - : "r" (p), "r" (1), "r" (0) - : "cr0", "memory"); - return ret; -} -#endif - -#ifdef __i386__ -static inline int testandset (int *p) -{ - long int readval = 0; - - __asm__ __volatile__ ("lock; cmpxchgl %2, %0" - : "+m" (*p), "+a" (readval) - : "r" (1) - : "cc"); - return readval; -} -#endif - -#ifdef __x86_64__ -static inline int testandset (int *p) -{ - long int readval = 0; - - __asm__ __volatile__ ("lock; cmpxchgl %2, %0" - : "+m" (*p), "+a" (readval) - : "r" (1) - : "cc"); - return readval; -} -#endif - -#ifdef __s390__ -static inline int testandset (int *p) -{ - int ret; - - __asm__ __volatile__ ("0: cs %0,%1,0(%2)\n" - " jl 0b" - : "=&d" (ret) - : "r" (1), "a" (p), "0" (*p) - : "cc", "memory" ); - return ret; -} -#endif - -#ifdef __alpha__ -static inline int testandset (int *p) -{ - int ret; - unsigned long one; - - __asm__ __volatile__ ("0: mov 1,%2\n" - " ldl_l %0,%1\n" - " stl_c %2,%1\n" - " beq %2,1f\n" - ".subsection 2\n" - "1: br 0b\n" - ".previous" - : "=r" (ret), "=m" (*p), "=r" (one) - : "m" (*p)); - return ret; -} -#endif - -#ifdef __sparc__ -static inline int testandset (int *p) -{ - int ret; - - __asm__ __volatile__("ldstub [%1], %0" - : "=r" (ret) - : "r" (p) - : "memory"); - - return (ret ? 1 : 0); -} -#endif - -#ifdef __arm__ -static inline int testandset (int *spinlock) -{ - register unsigned int ret; - __asm__ __volatile__("swp %0, %1, [%2]" - : "=r"(ret) - : "0"(1), "r"(spinlock)); - - return ret; -} -#endif - -#ifdef __mc68000 -static inline int testandset (int *p) -{ - char ret; - __asm__ __volatile__("tas %1; sne %0" - : "=r" (ret) - : "m" (p) - : "cc","memory"); - return ret; -} -#endif - -#ifdef __ia64 -#include - -static inline int testandset (int *p) -{ - return __sync_lock_test_and_set (p, 1); -} -#endif - -typedef int spinlock_t; - -#define SPIN_LOCK_UNLOCKED 0 - -#if defined(CONFIG_USER_ONLY) -static inline void spin_lock(spinlock_t *lock) -{ - while (testandset(lock)); -} - -static inline void spin_unlock(spinlock_t *lock) -{ - *lock = 0; -} - -static inline int spin_trylock(spinlock_t *lock) -{ - return !testandset(lock); -} -#else -static inline void spin_lock(spinlock_t *lock) -{ -} - -static inline void spin_unlock(spinlock_t *lock) -{ -} - -static inline int spin_trylock(spinlock_t *lock) -{ - return 1; -} -#endif +#include "qemu_spinlock.h" extern spinlock_t tb_lock; Index: qemu/linux-user/arm/syscall.h =================================================================== --- qemu.orig/linux-user/arm/syscall.h 2006-03-09 19:18:11.000000000 +0000 +++ qemu/linux-user/arm/syscall.h 2006-08-26 16:31:53.000000000 +0100 @@ -28,7 +28,9 @@ #define ARM_SYSCALL_BASE 0x900000 #define ARM_THUMB_SYSCALL 0 -#define ARM_NR_cacheflush (ARM_SYSCALL_BASE + 0xf0000 + 2) +#define ARM_NR_BASE 0xf0000 +#define ARM_NR_cacheflush (ARM_NR_BASE + 2) +#define ARM_NR_set_tls (ARM_NR_BASE + 5) #define ARM_NR_semihosting 0x123456 #define ARM_NR_thumb_semihosting 0xAB Index: qemu/linux-user/main.c =================================================================== --- qemu.orig/linux-user/main.c 2006-08-26 16:28:40.000000000 +0100 +++ qemu/linux-user/main.c 2006-08-26 16:31:53.000000000 +0100 @@ -309,6 +309,50 @@ } } +/* Handle a jump to the kernel code page. */ +static int +do_kernel_trap(CPUARMState *env) +{ + uint32_t addr; + uint32_t *ptr; + uint32_t cpsr; + + switch (env->regs[15]) { + case 0xffff0fc0: /* __kernel_cmpxchg */ + /* XXX: This only works between threads, not between processes. + Use native atomic operations. */ + /* ??? This probably breaks horribly if the access segfaults. */ + cpu_lock(); + ptr = (uint32_t *)env->regs[2]; + cpsr = cpsr_read(env); + if (*ptr == env->regs[0]) { + *ptr = env->regs[1]; + env->regs[0] = 0; + cpsr |= CPSR_C; + } else { + env->regs[0] = -1; + cpsr &= ~CPSR_C; + } + cpsr_write(env, cpsr, CPSR_C); + cpu_unlock(); + break; + case 0xffff0fe0: /* __kernel_get_tls */ + env->regs[0] = env->cp15.c13_tls; + break; + default: + return 1; + } + /* Jump back to the caller. */ + addr = env->regs[14]; + if (addr & 1) { + env->thumb = 1; + addr &= ~1; + } + env->regs[15] = addr; + + return 0; +} + void cpu_loop(CPUARMState *env) { int trapnr; @@ -365,10 +409,8 @@ } } - if (n == ARM_NR_cacheflush) { - arm_cache_flush(env->regs[0], env->regs[1]); - } else if (n == ARM_NR_semihosting - || n == ARM_NR_thumb_semihosting) { + if (n == ARM_NR_semihosting + || n == ARM_NR_thumb_semihosting) { env->regs[0] = do_arm_semihosting (env); } else if (n == 0 || n >= ARM_SYSCALL_BASE || (env->thumb && n == ARM_THUMB_SYSCALL)) { @@ -379,14 +421,34 @@ n -= ARM_SYSCALL_BASE; env->eabi = 0; } - env->regs[0] = do_syscall(env, - n, - env->regs[0], - env->regs[1], - env->regs[2], - env->regs[3], - env->regs[4], - env->regs[5]); + if ( n > ARM_NR_BASE) { + switch (n) + { + case ARM_NR_cacheflush: + arm_cache_flush(env->regs[0], env->regs[1]); + break; +#ifdef USE_NPTL + case ARM_NR_set_tls: + cpu_set_tls(env, env->regs[0]); + env->regs[0] = 0; + break; +#endif + default: + printf ("Error: Bad syscall: %x\n", n); + goto error; + } + } + else + { + env->regs[0] = do_syscall(env, + n, + env->regs[0], + env->regs[1], + env->regs[2], + env->regs[3], + env->regs[4], + env->regs[5]); + } } else { goto error; } @@ -425,6 +487,10 @@ } } break; + case EXCP_KERNEL_TRAP: + if (do_kernel_trap(env)) + goto error; + break; default: error: fprintf(stderr, "qemu: unhandled CPU exception 0x%x - aborting\n", @@ -1639,6 +1705,10 @@ ts->heap_base = info->brk; /* This will be filled in on the first SYS_HEAPINFO call. */ ts->heap_limit = 0; + /* Register the magic kernel code page. The cpu will generate a + special exception when it tries to execute code here. We can't + put real code here because it may be in use by the host kernel. */ + page_set_flags(0xffff0000, 0xffff0fff, 0); } #elif defined(TARGET_SPARC) { Index: qemu/linux-user/qemu.h =================================================================== --- qemu.orig/linux-user/qemu.h 2006-08-26 16:28:40.000000000 +0100 +++ qemu/linux-user/qemu.h 2006-08-26 16:33:50.000000000 +0100 @@ -75,6 +75,9 @@ uint32_t v86mask; #endif int used; /* non zero if used */ +#ifdef USE_NPTL + uint32_t *child_tidptr; +#endif struct image_info *info; uint8_t stack[0]; } __attribute__((aligned(16))) TaskState; Index: qemu/linux-user/syscall.c =================================================================== --- qemu.orig/linux-user/syscall.c 2006-08-26 16:28:40.000000000 +0100 +++ qemu/linux-user/syscall.c 2006-08-26 16:31:53.000000000 +0100 @@ -66,9 +66,18 @@ #include #include "qemu.h" +#include "qemu_spinlock.h" //#define DEBUG +#ifdef USE_NPTL +#define CLONE_NPTL_FLAGS2 (CLONE_SETTLS | \ + CLONE_PARENT_SETTID | CLONE_CHILD_SETTID | CLONE_CHILD_CLEARTID) +#else +/* XXX: Hardcode the above values. */ +#define CLONE_NPTL_FLAGS2 0 +#endif + #if defined(TARGET_I386) || defined(TARGET_ARM) || defined(TARGET_SPARC) /* 16 bit uid wrappers emulation */ #define USE_UID16 @@ -1602,20 +1611,38 @@ thread/process */ #define NEW_STACK_SIZE 8192 +#ifdef USE_NPTL +static spinlock_t nptl_lock = SPIN_LOCK_UNLOCKED; +#endif + static int clone_func(void *arg) { CPUState *env = arg; +#ifdef HAVE_NPTL + /* Wait until the parent has finshed initializing the tls state. */ + while (!spin_trylock(&nptl_lock)) + usleep(1); + spin_unlock(&nptl_lock); +#endif cpu_loop(env); /* never exits */ return 0; } -int do_fork(CPUState *env, unsigned int flags, unsigned long newsp) +int do_fork(CPUState *env, unsigned int flags, unsigned long newsp, + uint32_t *parent_tidptr, void *newtls, + uint32_t *child_tidptr) { int ret; TaskState *ts; uint8_t *new_stack; CPUState *new_env; +#ifdef USE_NPTL + unsigned int nptl_flags; + + if (flags & CLONE_PARENT_SETTID) + *parent_tidptr = gettid(); +#endif if (flags & CLONE_VM) { ts = malloc(sizeof(TaskState) + NEW_STACK_SIZE); @@ -1665,16 +1692,60 @@ #error unsupported target CPU #endif new_env->opaque = ts; +#ifdef USE_NPTL + nptl_flags = flags; + flags &= ~CLONE_NPTL_FLAGS2; + if (nptl_flags & CLONE_CHILD_CLEARTID) { + ts->child_tidptr = child_tidptr; + } + if (nptl_flags & CLONE_SETTLS) + cpu_set_tls (new_env, newtls); + /* Grab the global cpu lock so that the thread setup appears + atomic. */ + if (nptl_flags & CLONE_CHILD_SETTID) + spin_lock(&nptl_lock); +#else + if (flags & CLONE_NPTL_FLAGS2) + return -EINVAL; +#endif + #ifdef __ia64__ ret = __clone2(clone_func, new_stack + NEW_STACK_SIZE, flags, new_env); #else ret = clone(clone_func, new_stack + NEW_STACK_SIZE, flags, new_env); #endif +#ifdef USE_NPTL + if (ret != -1) { + if (nptl_flags & CLONE_CHILD_SETTID) + *child_tidptr = ret; + } + /* Allow the child to continue. */ + if (nptl_flags & CLONE_CHILD_SETTID) + spin_unlock(&nptl_lock); +#endif } else { - /* if no CLONE_VM, we consider it is a fork */ - if ((flags & ~CSIGNAL) != 0) - return -EINVAL; - ret = fork(); + /* if no CLONE_VM, we consider it is a fork */ + if ((flags & ~(CSIGNAL | CLONE_NPTL_FLAGS2)) != 0) + return -EINVAL; + ret = fork(); +#ifdef USE_NPTL + /* There is a race condition here. The parent process could + theoretically read the TID in the child process before the child + tid is set. This would require using either ptrace + (not implemented) or having *_tidptr to point at a shared memory + mapping. We can't repeat the spinlock hack used above because + the child process gets its own copy of the lock. */ + if (ret == 0) { + /* Child Process. */ + if (flags & CLONE_CHILD_SETTID) + *child_tidptr = gettid(); + ts = (TaskState *)env->opaque; + if (flags & CLONE_CHILD_CLEARTID) + ts->child_tidptr = child_tidptr; + if (flags & CLONE_SETTLS) + cpu_set_tls (env, newtls); + } +#endif } return ret; } @@ -1918,7 +1989,7 @@ ret = do_brk(arg1); break; case TARGET_NR_fork: - ret = get_errno(do_fork(cpu_env, SIGCHLD, 0)); + ret = get_errno(do_fork(cpu_env, SIGCHLD, 0, NULL, NULL, NULL)); break; case TARGET_NR_waitpid: { @@ -2989,7 +3060,8 @@ ret = get_errno(fsync(arg1)); break; case TARGET_NR_clone: - ret = get_errno(do_fork(cpu_env, arg1, arg2)); + ret = get_errno(do_fork(cpu_env, arg1, arg2, (uint32_t *)arg3, + (void *)arg4, (uint32_t *)arg5)); break; #ifdef __NR_exit_group /* new thread calls */ @@ -3339,7 +3411,8 @@ #endif #ifdef TARGET_NR_vfork case TARGET_NR_vfork: - ret = get_errno(do_fork(cpu_env, CLONE_VFORK | CLONE_VM | SIGCHLD, 0)); + ret = get_errno(do_fork(cpu_env, CLONE_VFORK | CLONE_VM | SIGCHLD, 0, + NULL, NULL, NULL)); break; #endif #ifdef TARGET_NR_ugetrlimit @@ -3838,4 +3911,3 @@ #endif return ret; } - Index: qemu/target-arm/cpu.h =================================================================== --- qemu.orig/target-arm/cpu.h 2006-03-09 19:18:27.000000000 +0000 +++ qemu/target-arm/cpu.h 2006-08-26 16:31:53.000000000 +0100 @@ -35,6 +35,9 @@ #define EXCP_IRQ 5 #define EXCP_FIQ 6 #define EXCP_BKPT 7 +#define EXCP_KERNEL_TRAP 8 /* Jumped to kernel code page. */ + + /* We currently assume float and double are IEEE single and double precision respectively. @@ -85,6 +88,7 @@ uint32_t c9_data; uint32_t c13_fcse; /* FCSE PID. */ uint32_t c13_context; /* Context ID. */ + uint32_t c13_tls; /* Paul Brook told me to just add this ;) */ } cp15; /* Internal CPU feature flags. */ @@ -135,6 +139,15 @@ int cpu_arm_signal_handler(int host_signum, struct siginfo *info, void *puc); +void cpu_lock(void); +void cpu_unlock(void); +#if defined(USE_NPTL) +static inline void cpu_set_tls(CPUARMState *env, void *newtls) +{ + env->cp15.c13_tls = (uint32_t)newtls; +} +#endif + #define CPSR_M (0x1f) #define CPSR_T (1 << 5) #define CPSR_F (1 << 6) @@ -146,7 +159,11 @@ #define CPSR_J (1 << 24) #define CPSR_IT_0_1 (3 << 25) #define CPSR_Q (1 << 27) -#define CPSR_NZCV (0xf << 28) +#define CPSR_V (1 << 28) +#define CPSR_C (1 << 29) +#define CPSR_Z (1 << 30) +#define CPSR_N (1 << 31) +#define CPSR_NZCV (CPSR_N | CPSR_Z | CPSR_C | CPSR_V) #define CACHED_CPSR_BITS (CPSR_T | CPSR_Q | CPSR_NZCV) /* Return the current CPSR value. */ Index: qemu/target-arm/exec.h =================================================================== --- qemu.orig/target-arm/exec.h 2006-03-09 19:18:27.000000000 +0000 +++ qemu/target-arm/exec.h 2006-08-26 16:31:53.000000000 +0100 @@ -51,8 +51,6 @@ /* In op_helper.c */ -void cpu_lock(void); -void cpu_unlock(void); void helper_set_cp15(CPUState *, uint32_t, uint32_t); uint32_t helper_get_cp15(CPUState *, uint32_t); Index: qemu/target-arm/op.c =================================================================== --- qemu.orig/target-arm/op.c 2006-08-26 16:28:48.000000000 +0100 +++ qemu/target-arm/op.c 2006-08-26 16:31:53.000000000 +0100 @@ -891,6 +891,12 @@ cpu_loop_exit(); } +void OPPROTO op_kernel_trap(void) +{ + env->exception_index = EXCP_KERNEL_TRAP; + cpu_loop_exit(); +} + /* VFP support. We follow the convention used for VFP instrunctions: Single precition routines have a "s" suffix, double precision a "d" suffix. */ Index: qemu/target-arm/translate.c =================================================================== --- qemu.orig/target-arm/translate.c 2006-08-26 16:28:48.000000000 +0100 +++ qemu/target-arm/translate.c 2006-08-26 16:31:53.000000000 +0100 @@ -2382,6 +2382,7 @@ s->is_jmp = DISAS_JUMP; } + /* generate intermediate code in gen_opc_buf and gen_opparam_buf for basic block 'tb'. If search_pc is TRUE, also generate PC information for each intermediate instruction. */ @@ -2416,6 +2417,15 @@ nb_gen_labels = 0; lj = -1; do { +#ifdef CONFIG_USER_ONLY + /* Intercept jump to the magic kernel page. */ + if (dc->pc > 0xffff0000) { + gen_op_kernel_trap(); + dc->is_jmp = DISAS_UPDATE; + break; + } +#endif + if (env->nb_breakpoints > 0) { for(j = 0; j < env->nb_breakpoints; j++) { if (env->breakpoints[j] == dc->pc) { Index: qemu/qemu_spinlock.h =================================================================== --- /dev/null 1970-01-01 00:00:00.000000000 +0000 +++ qemu/qemu_spinlock.h 2006-08-26 16:31:53.000000000 +0100 @@ -0,0 +1,182 @@ +/* + * internal execution defines for qemu + * + * Copyright (c) 2003 Fabrice Bellard + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library 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 + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#ifndef _QEMU_SPINLOCK_H +#define _QEMU_SPINLOCK_H + +#ifdef __powerpc__ +static inline int testandset (int *p) +{ + int ret; + __asm__ __volatile__ ( + "0: lwarx %0,0,%1\n" + " xor. %0,%3,%0\n" + " bne 1f\n" + " stwcx. %2,0,%1\n" + " bne- 0b\n" + "1: " + : "=&r" (ret) + : "r" (p), "r" (1), "r" (0) + : "cr0", "memory"); + return ret; +} +#endif + +#ifdef __i386__ +static inline int testandset (int *p) +{ + long int readval = 0; + + __asm__ __volatile__ ("lock; cmpxchgl %2, %0" + : "+m" (*p), "+a" (readval) + : "r" (1) + : "cc"); + return readval; +} +#endif + +#ifdef __x86_64__ +static inline int testandset (int *p) +{ + long int readval = 0; + + __asm__ __volatile__ ("lock; cmpxchgl %2, %0" + : "+m" (*p), "+a" (readval) + : "r" (1) + : "cc"); + return readval; +} +#endif + +#ifdef __s390__ +static inline int testandset (int *p) +{ + int ret; + + __asm__ __volatile__ ("0: cs %0,%1,0(%2)\n" + " jl 0b" + : "=&d" (ret) + : "r" (1), "a" (p), "0" (*p) + : "cc", "memory" ); + return ret; +} +#endif + +#ifdef __alpha__ +static inline int testandset (int *p) +{ + int ret; + unsigned long one; + + __asm__ __volatile__ ("0: mov 1,%2\n" + " ldl_l %0,%1\n" + " stl_c %2,%1\n" + " beq %2,1f\n" + ".subsection 2\n" + "1: br 0b\n" + ".previous" + : "=r" (ret), "=m" (*p), "=r" (one) + : "m" (*p)); + return ret; +} +#endif + +#ifdef __sparc__ +static inline int testandset (int *p) +{ + int ret; + + __asm__ __volatile__("ldstub [%1], %0" + : "=r" (ret) + : "r" (p) + : "memory"); + + return (ret ? 1 : 0); +} +#endif + +#ifdef __arm__ +static inline int testandset (int *spinlock) +{ + register unsigned int ret; + __asm__ __volatile__("swp %0, %1, [%2]" + : "=r"(ret) + : "0"(1), "r"(spinlock)); + + return ret; +} +#endif + +#ifdef __mc68000 +static inline int testandset (int *p) +{ + char ret; + __asm__ __volatile__("tas %1; sne %0" + : "=r" (ret) + : "m" (p) + : "cc","memory"); + return ret; +} +#endif + +#ifdef __ia64 +#include + +static inline int testandset (int *p) +{ + return __sync_lock_test_and_set (p, 1); +} +#endif + +typedef int spinlock_t; + +#define SPIN_LOCK_UNLOCKED 0 + +#if defined(CONFIG_USER_ONLY) +static inline void spin_lock(spinlock_t *lock) +{ + while (testandset(lock)); +} + +static inline void spin_unlock(spinlock_t *lock) +{ + *lock = 0; +} + +static inline int spin_trylock(spinlock_t *lock) +{ + return !testandset(lock); +} +#else +static inline void spin_lock(spinlock_t *lock) +{ +} + +static inline void spin_unlock(spinlock_t *lock) +{ +} + +static inline int spin_trylock(spinlock_t *lock) +{ + return 1; +} +#endif + +#endif /* ! _QEMU_SPINLOCK_H */