Loading arch/cris/arch-v10/kernel/entry.S +1 −244 Original line number Diff line number Diff line /* $Id: entry.S,v 1.28 2005/06/20 05:06:30 starvik Exp $ * /* * linux/arch/cris/entry.S * * Copyright (C) 2000, 2001, 2002 Axis Communications AB * * Authors: Bjorn Wesen (bjornw@axis.com) * * $Log: entry.S,v $ * Revision 1.28 2005/06/20 05:06:30 starvik * Remove unnecessary diff to kernel.org tree * * Revision 1.27 2005/03/04 08:16:16 starvik * Merge of Linux 2.6.11. * * Revision 1.26 2005/01/11 13:49:47 starvik * Added NMI handler. * * Revision 1.25 2004/12/27 11:18:32 starvik * Merge of Linux 2.6.10 (not functional yet). * * Revision 1.24 2004/12/22 10:41:23 starvik * Updates to make v10 compile with the latest SMP aware generic code (even * though v10 will never have SMP). * * Revision 1.23 2004/10/19 13:07:37 starvik * Merge of Linux 2.6.9 * * Revision 1.22 2004/06/21 10:29:55 starvik * Merge of Linux 2.6.7 * * Revision 1.21 2004/06/09 05:30:27 starvik * Clean up multiple interrupt handling. * Prevent interrupts from interrupting each other. * Handle all active interrupts. * * Revision 1.20 2004/06/08 08:55:32 starvik * Removed unused code * * Revision 1.19 2004/06/04 11:56:15 starvik * Implemented page table lookup for refills in assembler for improved performance. * * Revision 1.18 2004/05/11 12:28:25 starvik * Merge of Linux 2.6.6 * * Revision 1.17 2003/09/11 07:29:49 starvik * Merge of Linux 2.6.0-test5 * * Revision 1.16 2003/07/04 08:27:41 starvik * Merge of Linux 2.5.74 * * Revision 1.15 2003/04/09 07:32:55 starvik * resume should return task_struct, not thread_info * * Revision 1.14 2003/04/09 05:20:44 starvik * Merge of Linux 2.5.67 * * Revision 1.13 2002/12/11 15:42:02 starvik * Extracted v10 (ETRAX 100LX) specific stuff from arch/cris/kernel/*.c * * Revision 1.12 2002/12/10 09:00:10 starvik * Merge of Linux 2.5.51 * * Revision 1.11 2002/12/05 07:53:10 starvik * Corrected constants used with btstq * * Revision 1.10 2002/11/27 08:45:10 starvik * pid is in task_struct, not thread_info * * Revision 1.9 2002/11/26 09:52:05 starvik * Added preemptive kernel scheduling (if CONFIG_PREEMPT) * * Revision 1.8 2002/11/20 11:56:11 starvik * Merge of Linux 2.5.48 * * Revision 1.7 2002/11/18 13:02:42 starvik * Added fourth parameter to do_notify_resume * Minor cleanup * * Revision 1.6 2002/11/11 10:37:50 starvik * Use new asm-offset defines * Modified for new location of current->work etc * Removed SYMBOL_NAME from syscalls * Added some new syscalls * * Revision 1.5 2002/11/05 06:45:11 starvik * Merge of Linux 2.5.45 * * Revision 1.4 2002/02/05 15:41:31 bjornw * Rewritten to conform better to current 2.5 code (similar to arch/i386) * * Revision 1.3 2002/01/21 15:22:20 bjornw * NICE_DOGGY fix from 2.4 arch/cris * * Revision 1.37 2001/12/07 17:03:55 bjornw * Call a c-hook called watchdog_bite_hook instead of show_registers directly * * Revision 1.36 2001/11/22 13:36:36 bjornw * * In ret_from_intr, check regs->dccr for usermode reentrance instead of * DCCR explicitly (because the latter might not reflect current reality) * * In mmu_bus_fault, set $r9 _after_ calling the C-code instead of before * since $r9 is call-clobbered and is potentially needed afterwards * * Revision 1.35 2001/10/30 17:10:15 bjornw * Add some syscalls * * Revision 1.34 2001/10/01 14:45:03 bjornw * Removed underscores and added register prefixes * * Revision 1.33 2001/08/21 13:48:01 jonashg * Added fix by HP to avoid oops when doing a hard_reset_now. * * Revision 1.32 2001/08/14 04:32:02 hp * In _resume, add comment why R9 is saved; don't sound like it's call-saved. * * Revision 1.31 2001/07/25 16:07:42 bjornw * softirq_active/mask -> softirq_pending only * * Revision 1.30 2001/07/05 01:03:32 hp * - include asm/errno.h to get ENOSYS. * - Use ENOSYS, not local constant LENOSYS; tweak comments. * - Explain why .include, not #include is used. * - Make oops-register-dump if watchdog bits and it's not expected. * - Don't jsr, use jump _hard_reset_now, and skip spurious nop. * - Use correct section attribute for section .rodata. * - Adjust sys_ni_syscall fill number. * * Revision 1.29 2001/06/25 14:07:00 hp * Fix review comment. * * head.S: Use IO_STATE, IO_FIELD and IO_MASK constructs instead of * magic numbers. Add comment that -traditional must not be used. * * entry.S (SYMBOL_NAME): Change redefinition to use ## concatenation. * Correct and update comment. * * Makefile (.S.o): Don't use -traditional. Add comment why the * toplevel rule can't be used (now that there's a reason). * * Revision 1.28 2001/06/21 02:00:40 hp * * entry.S: Include asm/unistd.h. * (_sys_call_table): Use section .rodata, not .data. * (_kernel_thread): Move from... * * process.c: ... here. * * entryoffsets.c (VAL): Break out from... * (OF): Use VAL. * (LCLONE_VM): New asmified value from CLONE_VM. * * Revision 1.27 2001/05/29 11:25:27 markusl * In case of "spurious_interrupt", do hard_reset instead of hanging system in a loop... * * Revision 1.26 2001/05/15 15:46:03 bjornw * Include config.h now that we use some CONFIG_ options * * Revision 1.25 2001/05/15 05:38:47 hp * Tweaked code in _ret_from_sys_call * * Revision 1.24 2001/05/15 05:27:49 hp * Save r9 in r1 over function call rather than on stack. * * Revision 1.23 2001/05/15 05:10:00 hp * Generate entry.S structure offsets from C * * Revision 1.22 2001/04/17 13:58:39 orjanf * * Renamed CONFIG_KGDB to CONFIG_ETRAX_KGDB. * * Revision 1.21 2001/04/17 11:33:29 orjanf * Updated according to review: * * Included asm/sv_addr_ag.h to get macro for internal register. * * Corrected comment regarding system call argument passing. * * Removed comment about instruction being in a delay slot. * * Added comment about SYMBOL_NAME macro. * * Revision 1.20 2001/04/12 08:51:07 hp * - Add entry for sys_fcntl64. In fact copy last piece from i386 including ... * - .rept to fill table to safe state with sys_ni_syscall. * * Revision 1.19 2001/04/04 09:43:32 orjanf * * Moved do_sigtrap from traps.c to entry.S. * * LTASK_PID need not be global anymore. * * Revision 1.18 2001/03/26 09:25:02 markusl * Updated after review, should now handle USB interrupts correctly. * * Revision 1.17 2001/03/21 16:12:55 bjornw * * Always make room for the cpu status record in the frame, in order to * use the same framelength and layout for both mmu busfaults and normal * irqs. No need to check for the explicit CRIS_FRAME_FIXUP type anymore. * * Fixed bug with using addq for popping the stack in the epilogue - it * destroyed the flag register. Use instructions that don't affect the * flag register instead. * * Removed write to R_PORT_PA_DATA during spurious_interrupt * * Revision 1.16 2001/03/20 19:43:02 bjornw * * Get rid of esp0 setting * * Give a 7th argument to a systemcall - the stackframe * * Revision 1.15 2001/03/05 13:14:30 bjornw * Spelling fix * * Revision 1.14 2001/02/23 08:36:36 perf * New ABI; syscallnr=r9, arg5=mof, arg6=srp. * Corrected tracesys call check. * * Revision 1.13 2001/02/15 08:40:55 perf * H-P by way of perf; * - (_system_call): Don't read system call function address into r1. * - (RBFExit): There is no such thing as a null pop. Adjust sp by addq. * - (_system_call): Don't use r10 and don't save and restore it. * - (THREAD_ESP0): New constant. * - (_system_call): Inline set_esp0. * * Revision 1.12 2001/01/31 17:56:25 orjanf * Added definition of LTASK_PID and made it global. * * Revision 1.11 2001/01/10 21:13:29 bjornw * SYMBOL_NAME is defined incorrectly for the compiler options we currently use * * Revision 1.10 2000/12/18 23:47:56 bjornw * * Added syscall trace support (ptrace), completely untested of course * * Removed redundant check for NULL entries in syscall_table * * Revision 1.9 2000/11/21 16:40:51 bjornw * * New frame type used when an SBFS frame needs to be popped without * actually restarting the instruction * * Enable interrupts in signal_return (they did so in x86, I hope it's a good * idea) * * Revision 1.8 2000/11/17 16:53:35 bjornw * Added detection of frame-type in Rexit, so that mmu_bus_fault can * use ret_from_intr in the return-path to check for signals (like SEGV) * and other foul things that might have occurred during the fault. * * Revision 1.7 2000/10/06 15:04:28 bjornw * Include mof in register savings * * Revision 1.6 2000/09/12 16:02:44 bjornw * Linux-2.4.0-test7 derived updates * * Revision 1.5 2000/08/17 15:35:15 bjornw * 2.4.0-test6 changed local_irq_count and friends API * * Revision 1.4 2000/08/02 13:59:30 bjornw * Removed olduname and uname from the syscall list * * Revision 1.3 2000/07/31 13:32:58 bjornw * * Export ret_from_intr * * _resume updated (prev/last tjohejsan) * * timer_interrupt obsolete * * SIGSEGV detection in mmu_bus_fault temporarily disabled * * */ /* Loading Loading
arch/cris/arch-v10/kernel/entry.S +1 −244 Original line number Diff line number Diff line /* $Id: entry.S,v 1.28 2005/06/20 05:06:30 starvik Exp $ * /* * linux/arch/cris/entry.S * * Copyright (C) 2000, 2001, 2002 Axis Communications AB * * Authors: Bjorn Wesen (bjornw@axis.com) * * $Log: entry.S,v $ * Revision 1.28 2005/06/20 05:06:30 starvik * Remove unnecessary diff to kernel.org tree * * Revision 1.27 2005/03/04 08:16:16 starvik * Merge of Linux 2.6.11. * * Revision 1.26 2005/01/11 13:49:47 starvik * Added NMI handler. * * Revision 1.25 2004/12/27 11:18:32 starvik * Merge of Linux 2.6.10 (not functional yet). * * Revision 1.24 2004/12/22 10:41:23 starvik * Updates to make v10 compile with the latest SMP aware generic code (even * though v10 will never have SMP). * * Revision 1.23 2004/10/19 13:07:37 starvik * Merge of Linux 2.6.9 * * Revision 1.22 2004/06/21 10:29:55 starvik * Merge of Linux 2.6.7 * * Revision 1.21 2004/06/09 05:30:27 starvik * Clean up multiple interrupt handling. * Prevent interrupts from interrupting each other. * Handle all active interrupts. * * Revision 1.20 2004/06/08 08:55:32 starvik * Removed unused code * * Revision 1.19 2004/06/04 11:56:15 starvik * Implemented page table lookup for refills in assembler for improved performance. * * Revision 1.18 2004/05/11 12:28:25 starvik * Merge of Linux 2.6.6 * * Revision 1.17 2003/09/11 07:29:49 starvik * Merge of Linux 2.6.0-test5 * * Revision 1.16 2003/07/04 08:27:41 starvik * Merge of Linux 2.5.74 * * Revision 1.15 2003/04/09 07:32:55 starvik * resume should return task_struct, not thread_info * * Revision 1.14 2003/04/09 05:20:44 starvik * Merge of Linux 2.5.67 * * Revision 1.13 2002/12/11 15:42:02 starvik * Extracted v10 (ETRAX 100LX) specific stuff from arch/cris/kernel/*.c * * Revision 1.12 2002/12/10 09:00:10 starvik * Merge of Linux 2.5.51 * * Revision 1.11 2002/12/05 07:53:10 starvik * Corrected constants used with btstq * * Revision 1.10 2002/11/27 08:45:10 starvik * pid is in task_struct, not thread_info * * Revision 1.9 2002/11/26 09:52:05 starvik * Added preemptive kernel scheduling (if CONFIG_PREEMPT) * * Revision 1.8 2002/11/20 11:56:11 starvik * Merge of Linux 2.5.48 * * Revision 1.7 2002/11/18 13:02:42 starvik * Added fourth parameter to do_notify_resume * Minor cleanup * * Revision 1.6 2002/11/11 10:37:50 starvik * Use new asm-offset defines * Modified for new location of current->work etc * Removed SYMBOL_NAME from syscalls * Added some new syscalls * * Revision 1.5 2002/11/05 06:45:11 starvik * Merge of Linux 2.5.45 * * Revision 1.4 2002/02/05 15:41:31 bjornw * Rewritten to conform better to current 2.5 code (similar to arch/i386) * * Revision 1.3 2002/01/21 15:22:20 bjornw * NICE_DOGGY fix from 2.4 arch/cris * * Revision 1.37 2001/12/07 17:03:55 bjornw * Call a c-hook called watchdog_bite_hook instead of show_registers directly * * Revision 1.36 2001/11/22 13:36:36 bjornw * * In ret_from_intr, check regs->dccr for usermode reentrance instead of * DCCR explicitly (because the latter might not reflect current reality) * * In mmu_bus_fault, set $r9 _after_ calling the C-code instead of before * since $r9 is call-clobbered and is potentially needed afterwards * * Revision 1.35 2001/10/30 17:10:15 bjornw * Add some syscalls * * Revision 1.34 2001/10/01 14:45:03 bjornw * Removed underscores and added register prefixes * * Revision 1.33 2001/08/21 13:48:01 jonashg * Added fix by HP to avoid oops when doing a hard_reset_now. * * Revision 1.32 2001/08/14 04:32:02 hp * In _resume, add comment why R9 is saved; don't sound like it's call-saved. * * Revision 1.31 2001/07/25 16:07:42 bjornw * softirq_active/mask -> softirq_pending only * * Revision 1.30 2001/07/05 01:03:32 hp * - include asm/errno.h to get ENOSYS. * - Use ENOSYS, not local constant LENOSYS; tweak comments. * - Explain why .include, not #include is used. * - Make oops-register-dump if watchdog bits and it's not expected. * - Don't jsr, use jump _hard_reset_now, and skip spurious nop. * - Use correct section attribute for section .rodata. * - Adjust sys_ni_syscall fill number. * * Revision 1.29 2001/06/25 14:07:00 hp * Fix review comment. * * head.S: Use IO_STATE, IO_FIELD and IO_MASK constructs instead of * magic numbers. Add comment that -traditional must not be used. * * entry.S (SYMBOL_NAME): Change redefinition to use ## concatenation. * Correct and update comment. * * Makefile (.S.o): Don't use -traditional. Add comment why the * toplevel rule can't be used (now that there's a reason). * * Revision 1.28 2001/06/21 02:00:40 hp * * entry.S: Include asm/unistd.h. * (_sys_call_table): Use section .rodata, not .data. * (_kernel_thread): Move from... * * process.c: ... here. * * entryoffsets.c (VAL): Break out from... * (OF): Use VAL. * (LCLONE_VM): New asmified value from CLONE_VM. * * Revision 1.27 2001/05/29 11:25:27 markusl * In case of "spurious_interrupt", do hard_reset instead of hanging system in a loop... * * Revision 1.26 2001/05/15 15:46:03 bjornw * Include config.h now that we use some CONFIG_ options * * Revision 1.25 2001/05/15 05:38:47 hp * Tweaked code in _ret_from_sys_call * * Revision 1.24 2001/05/15 05:27:49 hp * Save r9 in r1 over function call rather than on stack. * * Revision 1.23 2001/05/15 05:10:00 hp * Generate entry.S structure offsets from C * * Revision 1.22 2001/04/17 13:58:39 orjanf * * Renamed CONFIG_KGDB to CONFIG_ETRAX_KGDB. * * Revision 1.21 2001/04/17 11:33:29 orjanf * Updated according to review: * * Included asm/sv_addr_ag.h to get macro for internal register. * * Corrected comment regarding system call argument passing. * * Removed comment about instruction being in a delay slot. * * Added comment about SYMBOL_NAME macro. * * Revision 1.20 2001/04/12 08:51:07 hp * - Add entry for sys_fcntl64. In fact copy last piece from i386 including ... * - .rept to fill table to safe state with sys_ni_syscall. * * Revision 1.19 2001/04/04 09:43:32 orjanf * * Moved do_sigtrap from traps.c to entry.S. * * LTASK_PID need not be global anymore. * * Revision 1.18 2001/03/26 09:25:02 markusl * Updated after review, should now handle USB interrupts correctly. * * Revision 1.17 2001/03/21 16:12:55 bjornw * * Always make room for the cpu status record in the frame, in order to * use the same framelength and layout for both mmu busfaults and normal * irqs. No need to check for the explicit CRIS_FRAME_FIXUP type anymore. * * Fixed bug with using addq for popping the stack in the epilogue - it * destroyed the flag register. Use instructions that don't affect the * flag register instead. * * Removed write to R_PORT_PA_DATA during spurious_interrupt * * Revision 1.16 2001/03/20 19:43:02 bjornw * * Get rid of esp0 setting * * Give a 7th argument to a systemcall - the stackframe * * Revision 1.15 2001/03/05 13:14:30 bjornw * Spelling fix * * Revision 1.14 2001/02/23 08:36:36 perf * New ABI; syscallnr=r9, arg5=mof, arg6=srp. * Corrected tracesys call check. * * Revision 1.13 2001/02/15 08:40:55 perf * H-P by way of perf; * - (_system_call): Don't read system call function address into r1. * - (RBFExit): There is no such thing as a null pop. Adjust sp by addq. * - (_system_call): Don't use r10 and don't save and restore it. * - (THREAD_ESP0): New constant. * - (_system_call): Inline set_esp0. * * Revision 1.12 2001/01/31 17:56:25 orjanf * Added definition of LTASK_PID and made it global. * * Revision 1.11 2001/01/10 21:13:29 bjornw * SYMBOL_NAME is defined incorrectly for the compiler options we currently use * * Revision 1.10 2000/12/18 23:47:56 bjornw * * Added syscall trace support (ptrace), completely untested of course * * Removed redundant check for NULL entries in syscall_table * * Revision 1.9 2000/11/21 16:40:51 bjornw * * New frame type used when an SBFS frame needs to be popped without * actually restarting the instruction * * Enable interrupts in signal_return (they did so in x86, I hope it's a good * idea) * * Revision 1.8 2000/11/17 16:53:35 bjornw * Added detection of frame-type in Rexit, so that mmu_bus_fault can * use ret_from_intr in the return-path to check for signals (like SEGV) * and other foul things that might have occurred during the fault. * * Revision 1.7 2000/10/06 15:04:28 bjornw * Include mof in register savings * * Revision 1.6 2000/09/12 16:02:44 bjornw * Linux-2.4.0-test7 derived updates * * Revision 1.5 2000/08/17 15:35:15 bjornw * 2.4.0-test6 changed local_irq_count and friends API * * Revision 1.4 2000/08/02 13:59:30 bjornw * Removed olduname and uname from the syscall list * * Revision 1.3 2000/07/31 13:32:58 bjornw * * Export ret_from_intr * * _resume updated (prev/last tjohejsan) * * timer_interrupt obsolete * * SIGSEGV detection in mmu_bus_fault temporarily disabled * * */ /* Loading
