1375 lines · 37.6 KB · Assembly
RawDownload
/*  * Linux/PA-RISC Project (http://www.parisc-linux.org/) *  * System call entry code / Linux gateway page * Copyright (c) Matthew Wilcox 1999 <willy@infradead.org> * Licensed under the GNU GPL. * thanks to Philipp Rumpf, Mike Shaver and various others * sorry about the wall, puffin.. */ /*How does the Linux gateway page on PA-RISC work?------------------------------------------------The Linux gateway page on PA-RISC is "special".It actually has PAGE_GATEWAY bits set (this is linux terminology; in pariscterminology it's Execute, promote to PL0) in the page map.  So anythingexecuting on this page executes with kernel level privilege (there's more to itthan that: to have this happen, you also have to use a branch with a ,gatecompleter to activate the privilege promotion).  The upshot is that everythingthat runs on the gateway page runs at kernel privilege but with the currentuser process address space (although you have access to kernel space via %sr2).For the 0x100 syscall entry, we redo the space registers to point to the kerneladdress space (preserving the user address space in %sr3), move to wide mode ifrequired, save the user registers and branch into the kernel syscall entrypoint.  For all the other functions, we execute at kernel privilege but don'tflip address spaces. The basic upshot of this is that these code snippets areexecuted atomically (because the kernel can't be pre-empted) and they mayperform architecturally forbidden (to PL3) operations (like setting controlregisters).*/  #include <asm/asm-offsets.h>#include <asm/unistd.h>#include <asm/errno.h>#include <asm/page.h>#include <asm/psw.h>#include <asm/thread_info.h>#include <asm/assembly.h>#include <asm/processor.h>#include <asm/cache.h>#include <asm/spinlock_types.h> #include <linux/linkage.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 	.level          PA_ASM_LEVEL 	.macro	lws_pagefault_disable reg1,reg2	mfctl	%cr30, \reg2	ldo	TASK_PAGEFAULT_DISABLED(\reg2), \reg2	ldw	0(%sr2,\reg2), \reg1	ldo	1(\reg1), \reg1	stw	\reg1, 0(%sr2,\reg2)	.endm 	.macro	lws_pagefault_enable reg1,reg2	mfctl	%cr30, \reg2	ldo	TASK_PAGEFAULT_DISABLED(\reg2), \reg2	ldw	0(%sr2,\reg2), \reg1	ldo	-1(\reg1), \reg1	stw	\reg1, 0(%sr2,\reg2)	.endm 	/* raise exception if spinlock content is not zero or	 * __ARCH_SPIN_LOCK_UNLOCKED_VAL */	.macro	spinlock_check spin_val,tmpreg#ifdef CONFIG_LIGHTWEIGHT_SPINLOCK_CHECK	ldi	__ARCH_SPIN_LOCK_UNLOCKED_VAL, \tmpreg	andcm,=	\spin_val, \tmpreg, %r0	.word	SPINLOCK_BREAK_INSN#endif	.endm 	.text 	.import syscall_exit,code	.import syscall_exit_rfi,code 	/* 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 PAGE_SIZEENTRY(linux_gateway_page)         /* ADDRESS 0x00 to 0xb0 = 176 bytes / 4 bytes per insn = 44 insns */	.rept 44	KILL_INSN	.endr 	/* ADDRESS 0xb0 to 0xb8, lws uses two 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 (5) lws_entry:	gate	lws_start, %r0		/* increase privilege */	depi	PRIV_USER, 31, 2, %r31	/* Ensure we return into user mode. */ 	/* Fill from 0xb8 to 0xe0 */	.rept 10	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	PRIV_USER, 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 LINUX_GATEWAY_ADDRlinux_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 */ #ifdef CONFIG_64BIT	/* Store W bit on entry to the syscall in case it's a wide userland	 * process. */	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, %r211:	#endif 	/* We use a rsm/ssm pair to prevent sr3 from being clobbered	 * by external interrupts.	 */	mfsp    %sr7,%r1                        /* save user sr7 */	rsm	PSW_SM_I, %r0			/* disable interrupts */	mtsp    %r1,%sr3                        /* and store it in sr3 */ 	mfctl   %cr30,%r1	xor     %r1,%r30,%r30                   /* ye olde xor trick */	xor     %r1,%r30,%r1	xor     %r1,%r30,%r30 	LDREG	TASK_STACK(%r30),%r30		/* set up kernel stack */	ldo	FRAME_SIZE(%r30),%r30	/* 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 */	ssm	PSW_SM_I, %r0			/* enable interrupts */	STREGM	%r1,FRAME_SIZE(%r30)		/* save r1 (usp) here for now */	mfctl	%cr30,%r1			/* get task ptr in %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 CONFIG_64BIT	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)		/* Syscall number */	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   %r0, TASK_PT_ORIG_R28(%r1)      /* don't prohibit restarts */	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 CONFIG_64BIT	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	TASK_TI_FLAGS(%r1),%r1	ldi	_TIF_SYSCALL_TRACE_MASK, %r19	and,COND(=) %r1, %r19, %r0	b,n	.Ltracesys		/* Note!  We cannot use the syscall table that is mapped	nearby since the gateway page is mapped execute-only. */ #ifdef CONFIG_COMPAT	ldil	L%sys_call_table, %r1	or,ev	%r2,%r2,%r2	ldil	L%sys_call_table64, %r1	ldo	R%sys_call_table(%r1), %r19	or,ev	%r2,%r2,%r2	ldo	R%sys_call_table64(%r1), %r19#else	load32	sys_call_table, %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.	 */	mfctl	%cr30,%r1			/* get task ptr */	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 */ 	copy	%r2,%r26	ldil	L%do_syscall_trace_enter,%r1	ldil	L%tracesys_next,%r2	be	R%do_syscall_trace_enter(%sr7,%r1)	ldo	R%tracesys_next(%r2),%r2	tracesys_next:	/* do_syscall_trace_enter either returned the syscallno, or -1L,	 *  so we skip restoring the PT_GR20 below, since we pulled it from	 *  task->thread.regs.gr[20] above.	 */	copy	%ret0,%r20 	mfctl	%cr30,%r1			/* get task ptr */	LDREG   TASK_PT_GR28(%r1), %r28		/* Restore return value */	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	LDREG   TASK_PT_GR22(%r1), %r22	LDREG   TASK_PT_GR21(%r1), %r21#ifdef CONFIG_64BIT	ldo	-16(%r30),%r29			/* Reference param save area */#else	stw     %r22, -52(%r30)                 /* 5th argument */	stw     %r21, -56(%r30)                 /* 6th argument */#endif 	cmpib,COND(=),n -1,%r20,tracesys_exit /* seccomp may have returned -1 */	comiclr,>>	__NR_Linux_syscalls, %r20, %r0	b,n	.Ltracesys_nosys 	/* Note!  We cannot use the syscall table that is mapped	nearby since the gateway page is mapped execute-only. */ #ifdef CONFIG_COMPAT	LDREG	TASK_PT_GR30(%r1), %r19		/* get users sp back */	extrd,u	%r19,63,1,%r2			/* W hidden in bottom bit */ 	ldil	L%sys_call_table, %r1	or,ev	%r2,%r2,%r2	ldil	L%sys_call_table64, %r1	ldo	R%sys_call_table(%r1), %r19	or,ev	%r2,%r2,%r2	ldo	R%sys_call_table64(%r1), %r19#else	load32	sys_call_table, %r19#endif 	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 .Ltracesys_nosys:	ldo	-ENOSYS(%r0),%r28		/* set errno */ 	/* Do *not* call this function on the gateway page, because it	makes a direct call to syscall_trace. */	tracesys_exit:	mfctl	%cr30,%r1			/* get task ptr */#ifdef CONFIG_64BIT	ldo	-16(%r30),%r29			/* Reference param save area */#endif	ldo	TASK_REGS(%r1),%r26	BL	do_syscall_trace_exit,%r2	STREG   %r28,TASK_PT_GR28(%r1)          /* save return value now */	mfctl	%cr30,%r1			/* get task ptr */	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:	mfctl	%cr30,%r1			/* get task ptr */#ifdef CONFIG_64BIT	ldo	-16(%r30),%r29			/* Reference param save area */#endif	BL	do_syscall_trace_exit,%r2	ldo	TASK_REGS(%r1),%r26 	ldil	L%syscall_exit_rfi,%r1	be,n	R%syscall_exit_rfi(%sr7,%r1)  	/*********************************************************		32/64-bit Light-Weight-Syscall ABI 		* - Indicates a hint for userspace inline asm		implementations. 		Syscall number (caller-saves)	        - %r20	        * In asm clobber. 		Argument registers (caller-saves)	        - %r26, %r25, %r24, %r23, %r22	        * In asm input. 		Return registers (caller-saves)	        - %r28 (return), %r21 (errno)	        * In asm output. 		Caller-saves registers	        - %r1, %r27, %r29	        - %r2 (return pointer)	        - %r31 (ble link register)	        * In asm clobber. 		Callee-saves registers	        - %r3-%r18	        - %r30 (stack pointer)	        * Not in asm clobber. 		If userspace is 32-bit:		Callee-saves registers	        - %r19 (32-bit PIC register) 		Differences from 32-bit calling convention:		- Syscall number in %r20		- Additional argument register %r22 (arg4)		- Callee-saves %r19. 		If userspace is 64-bit:		Callee-saves registers		- %r27 (64-bit PIC register) 		Differences from 64-bit calling convention:		- Syscall number in %r20		- Additional argument register %r22 (arg4)		- Callee-saves %r27. 		Error codes returned by entry path: 		ENOSYS - r20 was an invalid LWS number. 	*********************************************************/lws_start: #ifdef CONFIG_64BIT	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

Lines 1–500 of 1375.