diff options
Diffstat (limited to 'arch/parisc/kernel/syscall.S')
| -rw-r--r-- | arch/parisc/kernel/syscall.S | 703 |
1 files changed, 703 insertions, 0 deletions
diff --git a/arch/parisc/kernel/syscall.S b/arch/parisc/kernel/syscall.S new file mode 100644 index 000000000000..32ea701f4d20 --- /dev/null +++ b/arch/parisc/kernel/syscall.S @@ -0,0 +1,703 @@ +/* + * Linux/PA-RISC Project (http://www.parisc-linux.org/) + * + * System call entry code Copyright (c) Matthew Wilcox 1999 <willy@bofh.ai> + * Licensed under the GNU GPL. + * thanks to Philipp Rumpf, Mike Shaver and various others + * sorry about the wall, puffin.. + */ + +#include <asm/offsets.h> +#include <asm/unistd.h> +#include <asm/errno.h> +#include <asm/psw.h> +#include <asm/thread_info.h> + +#include <asm/assembly.h> +#include <asm/processor.h> + + /* We fill the empty parts of the gateway page with + * something that will kill the kernel or a + * userspace application. + */ +#define KILL_INSN break 0,0 + +#include <linux/config.h> /* for CONFIG_SMP */ + +#ifdef __LP64__ + .level 2.0w +#else + .level 1.1 +#endif + +#ifndef __LP64__ + .macro fixup_branch,lbl + b \lbl + .endm +#else + .macro fixup_branch,lbl + ldil L%\lbl, %r1 + ldo R%\lbl(%r1), %r1 + bv,n %r0(%r1) + .endm +#endif + + .text + + .import syscall_exit,code + .import syscall_exit_rfi,code + .export linux_gateway_page + + /* Linux gateway page is aliased to virtual page 0 in the kernel + * address space. Since it is a gateway page it cannot be + * dereferenced, so null pointers will still fault. We start + * the actual entry point at 0x100. We put break instructions + * at the beginning of the page to trap null indirect function + * pointers. + */ + + .align 4096 +linux_gateway_page: + + /* ADDRESS 0x00 to 0xb0 = 176 bytes / 4 bytes per insn = 44 insns */ + .rept 44 + KILL_INSN + .endr + + /* ADDRESS 0xb0 to 0xb4, lws uses 1 insns for entry */ + /* Light-weight-syscall entry must always be located at 0xb0 */ + /* WARNING: Keep this number updated with table size changes */ +#define __NR_lws_entries (2) + +lws_entry: + /* Unconditional branch to lws_start, located on the + same gateway page */ + b,n lws_start + + /* Fill from 0xb4 to 0xe0 */ + .rept 11 + KILL_INSN + .endr + + /* This function MUST be located at 0xe0 for glibc's threading + mechanism to work. DO NOT MOVE THIS CODE EVER! */ +set_thread_pointer: + gate .+8, %r0 /* increase privilege */ + depi 3, 31, 2, %r31 /* Ensure we return into user mode. */ + be 0(%sr7,%r31) /* return to user space */ + mtctl %r26, %cr27 /* move arg0 to the control register */ + + /* Increase the chance of trapping if random jumps occur to this + address, fill from 0xf0 to 0x100 */ + .rept 4 + KILL_INSN + .endr + +/* This address must remain fixed at 0x100 for glibc's syscalls to work */ + .align 256 +linux_gateway_entry: + gate .+8, %r0 /* become privileged */ + mtsp %r0,%sr4 /* get kernel space into sr4 */ + mtsp %r0,%sr5 /* get kernel space into sr5 */ + mtsp %r0,%sr6 /* get kernel space into sr6 */ + mfsp %sr7,%r1 /* save user sr7 */ + mtsp %r1,%sr3 /* and store it in sr3 */ + +#ifdef __LP64__ + /* for now we can *always* set the W bit on entry to the syscall + * since we don't support wide userland processes. We could + * also save the current SM other than in r0 and restore it on + * exit from the syscall, and also use that value to know + * whether to do narrow or wide syscalls. -PB + */ + ssm PSW_SM_W, %r1 + extrd,u %r1,PSW_W_BIT,1,%r1 + /* sp must be aligned on 4, so deposit the W bit setting into + * the bottom of sp temporarily */ + or,ev %r1,%r30,%r30 + b,n 1f + /* The top halves of argument registers must be cleared on syscall + * entry from narrow executable. + */ + depdi 0, 31, 32, %r26 + depdi 0, 31, 32, %r25 + depdi 0, 31, 32, %r24 + depdi 0, 31, 32, %r23 + depdi 0, 31, 32, %r22 + depdi 0, 31, 32, %r21 +1: +#endif + mfctl %cr30,%r1 + xor %r1,%r30,%r30 /* ye olde xor trick */ + xor %r1,%r30,%r1 + xor %r1,%r30,%r30 + + ldo THREAD_SZ_ALGN+FRAME_SIZE(%r30),%r30 /* set up kernel stack */ + + /* N.B.: It is critical that we don't set sr7 to 0 until r30 + * contains a valid kernel stack pointer. It is also + * critical that we don't start using the kernel stack + * until after sr7 has been set to 0. + */ + + mtsp %r0,%sr7 /* get kernel space into sr7 */ + STREGM %r1,FRAME_SIZE(%r30) /* save r1 (usp) here for now */ + mfctl %cr30,%r1 /* get task ptr in %r1 */ + LDREG TI_TASK(%r1),%r1 + + /* Save some registers for sigcontext and potential task + switch (see entry.S for the details of which ones are + saved/restored). TASK_PT_PSW is zeroed so we can see whether + a process is on a syscall or not. For an interrupt the real + PSW value is stored. This is needed for gdb and sys_ptrace. */ + STREG %r0, TASK_PT_PSW(%r1) + STREG %r2, TASK_PT_GR2(%r1) /* preserve rp */ + STREG %r19, TASK_PT_GR19(%r1) + + LDREGM -FRAME_SIZE(%r30), %r2 /* get users sp back */ +#ifdef __LP64__ + extrd,u %r2,63,1,%r19 /* W hidden in bottom bit */ +#if 0 + xor %r19,%r2,%r2 /* clear bottom bit */ + depd,z %r19,1,1,%r19 + std %r19,TASK_PT_PSW(%r1) +#endif +#endif + STREG %r2, TASK_PT_GR30(%r1) /* ... and save it */ + + STREG %r20, TASK_PT_GR20(%r1) + STREG %r21, TASK_PT_GR21(%r1) + STREG %r22, TASK_PT_GR22(%r1) + STREG %r23, TASK_PT_GR23(%r1) /* 4th argument */ + STREG %r24, TASK_PT_GR24(%r1) /* 3rd argument */ + STREG %r25, TASK_PT_GR25(%r1) /* 2nd argument */ + STREG %r26, TASK_PT_GR26(%r1) /* 1st argument */ + STREG %r27, TASK_PT_GR27(%r1) /* user dp */ + STREG %r28, TASK_PT_GR28(%r1) /* return value 0 */ + STREG %r28, TASK_PT_ORIG_R28(%r1) /* return value 0 (saved for signals) */ + STREG %r29, TASK_PT_GR29(%r1) /* return value 1 */ + STREG %r31, TASK_PT_GR31(%r1) /* preserve syscall return ptr */ + + ldo TASK_PT_FR0(%r1), %r27 /* save fpregs from the kernel */ + save_fp %r27 /* or potential task switch */ + + mfctl %cr11, %r27 /* i.e. SAR */ + STREG %r27, TASK_PT_SAR(%r1) + + loadgp + +#ifdef __LP64__ + ldo -16(%r30),%r29 /* Reference param save area */ + copy %r19,%r2 /* W bit back to r2 */ +#else + /* no need to save these on stack in wide mode because the first 8 + * args are passed in registers */ + stw %r22, -52(%r30) /* 5th argument */ + stw %r21, -56(%r30) /* 6th argument */ +#endif + + /* Are we being ptraced? */ + mfctl %cr30, %r1 + LDREG TI_TASK(%r1),%r1 + LDREG TASK_PTRACE(%r1), %r1 + bb,<,n %r1,31,.Ltracesys + + /* Note! We cannot use the syscall table that is mapped + nearby since the gateway page is mapped execute-only. */ + +#ifdef __LP64__ + ldil L%sys_call_table, %r1 + or,= %r2,%r2,%r2 + addil L%(sys_call_table64-sys_call_table), %r1 + ldo R%sys_call_table(%r1), %r19 + or,= %r2,%r2,%r2 + ldo R%sys_call_table64(%r1), %r19 +#else + ldil L%sys_call_table, %r1 + ldo R%sys_call_table(%r1), %r19 +#endif + comiclr,>>= __NR_Linux_syscalls, %r20, %r0 + b,n .Lsyscall_nosys + + LDREGX %r20(%r19), %r19 + + /* If this is a sys_rt_sigreturn call, and the signal was received + * when not in_syscall, then we want to return via syscall_exit_rfi, + * not syscall_exit. Signal no. in r20, in_syscall in r25 (see + * trampoline code in signal.c). + */ + ldi __NR_rt_sigreturn,%r2 + comb,= %r2,%r20,.Lrt_sigreturn +.Lin_syscall: + ldil L%syscall_exit,%r2 + be 0(%sr7,%r19) + ldo R%syscall_exit(%r2),%r2 +.Lrt_sigreturn: + comib,<> 0,%r25,.Lin_syscall + ldil L%syscall_exit_rfi,%r2 + be 0(%sr7,%r19) + ldo R%syscall_exit_rfi(%r2),%r2 + + /* Note! Because we are not running where we were linked, any + calls to functions external to this file must be indirect. To + be safe, we apply the opposite rule to functions within this + file, with local labels given to them to ensure correctness. */ + +.Lsyscall_nosys: +syscall_nosys: + ldil L%syscall_exit,%r1 + be R%syscall_exit(%sr7,%r1) + ldo -ENOSYS(%r0),%r28 /* set errno */ + + +/* Warning! This trace code is a virtual duplicate of the code above so be + * sure to maintain both! */ +.Ltracesys: +tracesys: + /* Need to save more registers so the debugger can see where we + * are. This saves only the lower 8 bits of PSW, so that the C + * bit is still clear on syscalls, and the D bit is set if this + * full register save path has been executed. We check the D + * bit on syscall_return_rfi to determine which registers to + * restore. An interrupt results in a full PSW saved with the + * C bit set, a non-straced syscall entry results in C and D clear + * in the saved PSW. + */ + ldo -THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r1 /* get task ptr */ + LDREG TI_TASK(%r1), %r1 + ssm 0,%r2 + STREG %r2,TASK_PT_PSW(%r1) /* Lower 8 bits only!! */ + mfsp %sr0,%r2 + STREG %r2,TASK_PT_SR0(%r1) + mfsp %sr1,%r2 + STREG %r2,TASK_PT_SR1(%r1) + mfsp %sr2,%r2 + STREG %r2,TASK_PT_SR2(%r1) + mfsp %sr3,%r2 + STREG %r2,TASK_PT_SR3(%r1) + STREG %r2,TASK_PT_SR4(%r1) + STREG %r2,TASK_PT_SR5(%r1) + STREG %r2,TASK_PT_SR6(%r1) + STREG %r2,TASK_PT_SR7(%r1) + STREG %r2,TASK_PT_IASQ0(%r1) + STREG %r2,TASK_PT_IASQ1(%r1) + LDREG TASK_PT_GR31(%r1),%r2 + STREG %r2,TASK_PT_IAOQ0(%r1) + ldo 4(%r2),%r2 + STREG %r2,TASK_PT_IAOQ1(%r1) + ldo TASK_REGS(%r1),%r2 + /* reg_save %r2 */ + STREG %r3,PT_GR3(%r2) + STREG %r4,PT_GR4(%r2) + STREG %r5,PT_GR5(%r2) + STREG %r6,PT_GR6(%r2) + STREG %r7,PT_GR7(%r2) + STREG %r8,PT_GR8(%r2) + STREG %r9,PT_GR9(%r2) + STREG %r10,PT_GR10(%r2) + STREG %r11,PT_GR11(%r2) + STREG %r12,PT_GR12(%r2) + STREG %r13,PT_GR13(%r2) + STREG %r14,PT_GR14(%r2) + STREG %r15,PT_GR15(%r2) + STREG %r16,PT_GR16(%r2) + STREG %r17,PT_GR17(%r2) + STREG %r18,PT_GR18(%r2) + /* Finished saving things for the debugger */ + + ldil L%syscall_trace,%r1 + ldil L%tracesys_next,%r2 + be R%syscall_trace(%sr7,%r1) + ldo R%tracesys_next(%r2),%r2 + +tracesys_next: + ldil L%sys_call_table,%r1 + ldo R%sys_call_table(%r1), %r19 + + ldo -THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r1 /* get task ptr */ + LDREG TI_TASK(%r1), %r1 + LDREG TASK_PT_GR20(%r1), %r20 + LDREG TASK_PT_GR26(%r1), %r26 /* Restore the users args */ + LDREG TASK_PT_GR25(%r1), %r25 + LDREG TASK_PT_GR24(%r1), %r24 + LDREG TASK_PT_GR23(%r1), %r23 +#ifdef __LP64__ + LDREG TASK_PT_GR22(%r1), %r22 + LDREG TASK_PT_GR21(%r1), %r21 + ldo -16(%r30),%r29 /* Reference param save area */ +#endif + + comiclr,>>= __NR_Linux_syscalls, %r20, %r0 + b,n .Lsyscall_nosys + + LDREGX %r20(%r19), %r19 + + /* If this is a sys_rt_sigreturn call, and the signal was received + * when not in_syscall, then we want to return via syscall_exit_rfi, + * not syscall_exit. Signal no. in r20, in_syscall in r25 (see + * trampoline code in signal.c). + */ + ldi __NR_rt_sigreturn,%r2 + comb,= %r2,%r20,.Ltrace_rt_sigreturn +.Ltrace_in_syscall: + ldil L%tracesys_exit,%r2 + be 0(%sr7,%r19) + ldo R%tracesys_exit(%r2),%r2 + + /* Do *not* call this function on the gateway page, because it + makes a direct call to syscall_trace. */ + +tracesys_exit: + ldo -THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r1 /* get task ptr */ + LDREG TI_TASK(%r1), %r1 +#ifdef __LP64__ + ldo -16(%r30),%r29 /* Reference param save area */ +#endif + bl syscall_trace, %r2 + STREG %r28,TASK_PT_GR28(%r1) /* save return value now */ + ldo -THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r1 /* get task ptr */ + LDREG TI_TASK(%r1), %r1 + LDREG TASK_PT_GR28(%r1), %r28 /* Restore return val. */ + + ldil L%syscall_exit,%r1 + be,n R%syscall_exit(%sr7,%r1) + +.Ltrace_rt_sigreturn: + comib,<> 0,%r25,.Ltrace_in_syscall + ldil L%tracesys_sigexit,%r2 + be 0(%sr7,%r19) + ldo R%tracesys_sigexit(%r2),%r2 + +tracesys_sigexit: + ldo -THREAD_SZ_ALGN-FRAME_SIZE(%r30),%r1 /* get task ptr */ + LDREG 0(%r1), %r1 +#ifdef __LP64__ + ldo -16(%r30),%r29 /* Reference param save area */ +#endif + bl syscall_trace, %r2 + nop + + ldil L%syscall_exit_rfi,%r1 + be,n R%syscall_exit_rfi(%sr7,%r1) + + + /********************************************************* + Light-weight-syscall code + + r20 - lws number + r26,r25,r24,r23,r22 - Input registers + r28 - Function return register + r21 - Error code. + + Scracth: Any of the above that aren't being + currently used, including r1. + + Return pointer: r31 (Not usable) + + Error codes returned by entry path: + + ENOSYS - r20 was an invalid LWS number. + + *********************************************************/ +lws_start: + /* Gate and ensure we return to userspace */ + gate .+8, %r0 + depi 3, 31, 2, %r31 /* Ensure we return to userspace */ + +#ifdef __LP64__ + /* FIXME: If we are a 64-bit kernel just + * turn this on unconditionally. + */ + ssm PSW_SM_W, %r1 + extrd,u %r1,PSW_W_BIT,1,%r1 + /* sp must be aligned on 4, so deposit the W bit setting into + * the bottom of sp temporarily */ + or,ev %r1,%r30,%r30 + + /* Clip LWS number to a 32-bit value always */ + depdi 0, 31, 32, %r20 +#endif + + /* Is the lws entry number valid? */ + comiclr,>>= __NR_lws_entries, %r20, %r0 + b,n lws_exit_nosys + + /* WARNING: Trashing sr2 and sr3 */ + mfsp %sr7,%r1 /* get userspace into sr3 */ + mtsp %r1,%sr3 + mtsp %r0,%sr2 /* get kernel space into sr2 */ + + /* Load table start */ + ldil L%lws_table, %r1 + ldo R%lws_table(%r1), %r28 /* Scratch use of r28 */ + LDREGX %r20(%sr2,r28), %r21 /* Scratch use of r21 */ + + /* Jump to lws, lws table pointers already relocated */ + be,n 0(%sr2,%r21) + +lws_exit_nosys: + ldo -ENOSYS(%r0),%r21 /* set errno */ + /* Fall through: Return to userspace */ + +lws_exit: +#ifdef __LP64__ + /* decide whether to reset the wide mode bit + * + * For a syscall, the W bit is stored in the lowest bit + * of sp. Extract it and reset W if it is zero */ + extrd,u,*<> %r30,63,1,%r1 + rsm PSW_SM_W, %r0 + /* now reset the lowest bit of sp if it was set */ + xor %r30,%r1,%r30 +#endif + be,n 0(%sr3, %r31) + + + + /*************************************************** + Implementing CAS as an atomic operation: + + %r26 - Address to examine + %r25 - Old value to check (old) + %r24 - New value to set (new) + %r28 - Return prev through this register. + %r21 - Kernel error code + + If debugging is DISabled: + + %r21 has the following meanings: + + EAGAIN - CAS is busy, ldcw failed, try again. + EFAULT - Read or write failed. + + If debugging is enabled: + + EDEADLOCK - CAS called recursively. + EAGAIN && r28 == 1 - CAS is busy. Lock contended. + EAGAIN && r28 == 2 - CAS is busy. ldcw failed. + EFAULT - Read or write failed. + + Scratch: r20, r28, r1 + + ****************************************************/ + + /* Do not enable LWS debugging */ +#define ENABLE_LWS_DEBUG 0 + + /* ELF64 Process entry path */ +lws_compare_and_swap64: +#ifdef __LP64__ + b,n lws_compare_and_swap +#else + /* If we are not a 64-bit kernel, then we don't + * implement having 64-bit input registers + */ + b,n lws_exit_nosys +#endif + + /* ELF32 Process entry path */ +lws_compare_and_swap32: +#ifdef __LP64__ + /* Clip all the input registers */ + depdi 0, 31, 32, %r26 + depdi 0, 31, 32, %r25 + depdi 0, 31, 32, %r24 +#endif + +lws_compare_and_swap: +#ifdef CONFIG_SMP + /* Load start of lock table */ + ldil L%lws_lock_start, %r20 + ldo R%lws_lock_start(%r20), %r28 + + /* Extract four bits from r26 and hash lock (Bits 4-7) */ + extru %r26, 27, 4, %r20 + + /* Find lock to use, the hash is either one of 0 to + 15, multiplied by 16 (keep it 16-byte aligned) + and add to the lock table offset. */ + shlw %r20, 4, %r20 + add %r20, %r28, %r20 + +# ifdef ENABLE_LWS_DEBUG + /* + DEBUG, check for deadlock! + If the thread register values are the same + then we were the one that locked it last and + this is a recurisve call that will deadlock. + We *must* giveup this call and fail. + */ + ldw 4(%sr2,%r20), %r28 /* Load thread register */ + mfctl %cr27, %r21 /* Get current thread register */ + cmpb,<>,n %r21, %r28, cas_lock /* Called recursive? */ + b lws_exit /* Return error! */ + ldo -EDEADLOCK(%r0), %r21 +cas_lock: + cmpb,=,n %r0, %r28, cas_nocontend /* Is nobody using it? */ + ldo 1(%r0), %r28 /* 1st case */ + b lws_exit /* Contended... */ + ldo -EAGAIN(%r0), %r21 /* Spin in userspace */ +cas_nocontend: +# endif +/* ENABLE_LWS_DEBUG */ + + ldcw 0(%sr2,%r20), %r28 /* Try to acquire the lock */ + cmpb,<>,n %r0, %r28, cas_action /* Did we get it? */ +cas_wouldblock: + ldo 2(%r0), %r28 /* 2nd case */ + b lws_exit /* Contended... */ + ldo -EAGAIN(%r0), %r21 /* Spin in userspace */ +#endif +/* CONFIG_SMP */ + + /* + prev = *addr; + if ( prev == old ) + *addr = new; + return prev; + */ + + /* NOTES: + This all works becuse intr_do_signal + and schedule both check the return iasq + and see that we are on the kernel page + so this process is never scheduled off + or is ever sent any signal of any sort, + thus it is wholly atomic from usrspaces + perspective + */ +cas_action: +#if defined CONFIG_SMP && defined ENABLE_LWS_DEBUG + /* DEBUG */ + mfctl %cr27, %r1 + stw %r1, 4(%sr2,%r20) +#endif + /* The load and store could fail */ +1: ldw 0(%sr3,%r26), %r28 + sub,<> %r28, %r25, %r0 +2: stw %r24, 0(%sr3,%r26) +#ifdef CONFIG_SMP + /* Free lock */ + stw %r20, 0(%sr2,%r20) +# ifdef ENABLE_LWS_DEBUG + /* Clear thread register indicator */ + stw %r0, 4(%sr2,%r20) +# endif +#endif + /* Return to userspace, set no error */ + b lws_exit + copy %r0, %r21 + +3: + /* Error occured on load or store */ +#ifdef CONFIG_SMP + /* Free lock */ + stw %r20, 0(%sr2,%r20) +# ifdef ENABLE_LWS_DEBUG + stw %r0, 4(%sr2,%r20) +# endif +#endif + b lws_exit + ldo -EFAULT(%r0),%r21 /* set errno */ + nop + nop + nop + nop + + /* Two exception table entries, one for the load, + the other for the store. Either return -EFAULT. + Each of the entries must be relocated. */ + .section __ex_table,"aw" +#ifdef __LP64__ + /* Pad the address calculation */ + .word 0,(2b - linux_gateway_page) + .word 0,(3b - linux_gateway_page) +#else + .word (2b - linux_gateway_page) + .word (3b - linux_gateway_page) +#endif + .previous + + .section __ex_table,"aw" +#ifdef __LP64__ + /* Pad the address calculation */ + .word 0,(1b - linux_gateway_page) + .word 0,(3b - linux_gateway_page) +#else + .word (1b - linux_gateway_page) + .word (3b - linux_gateway_page) +#endif + .previous + +end_compare_and_swap: + + /* Make sure nothing else is placed on this page */ + .align 4096 + .export end_linux_gateway_page +end_linux_gateway_page: + + /* Relocate symbols assuming linux_gateway_page is mapped + to virtual address 0x0 */ +#ifdef __LP64__ + /* FIXME: The code will always be on the gateay page + and thus it will be on the first 4k, the + assembler seems to think that the final + subtraction result is only a word in + length, so we pad the value. + */ +#define LWS_ENTRY(_name_) .word 0,(lws_##_name_ - linux_gateway_page) +#else +#define LWS_ENTRY(_name_) .word (lws_##_name_ - linux_gateway_page) +#endif + + .align 4096 + /* Light-weight-syscall table */ + /* Start of lws table. */ + .export lws_table +.Llws_table: +lws_table: + LWS_ENTRY(compare_and_swap32) /* 0 - ELF32 Atomic compare and swap */ + LWS_ENTRY(compare_and_swap64) /* 1 - ELF64 Atomic compare and swap */ + /* End of lws table */ + + .align 4096 + .export sys_call_table +.Lsys_call_table: +sys_call_table: +#include "syscall_table.S" + +#ifdef __LP64__ + .align 4096 + .export sys_call_table64 +.Lsys_call_table64: +sys_call_table64: +#define SYSCALL_TABLE_64BIT +#include "syscall_table.S" +#endif + +#ifdef CONFIG_SMP + /* + All light-weight-syscall atomic operations + will use this set of locks + */ + .section .data + .align 4096 + .export lws_lock_start +.Llws_lock_start: +lws_lock_start: + /* lws locks */ + .align 16 + .rept 16 + /* Keep locks aligned at 16-bytes */ + .word 1 + .word 0 + .word 0 + .word 0 + .endr + .previous +#endif +/* CONFIG_SMP for lws_lock_start */ + +.end + + |