| #ifndef PARLIB_ARCH_VCORE32_H |
| #define PARLIB_ARCH_VCORE32_H |
| |
| #ifndef PARLIB_ARCH_VCORE_H |
| #error "Do not include include vcore32.h directly" |
| #endif |
| |
| #include <ros/common.h> |
| #include <ros/trapframe.h> |
| #include <ros/procdata.h> |
| #include <ros/syscall.h> |
| #include <ros/arch/mmu.h> |
| #include <sys/vcore-tls.h> |
| |
| /* Here's how the HW popping works: It sets up the future stack pointer to |
| * have extra stuff after it, and then it pops the registers, then pops the new |
| * context's stack pointer. Then it uses the extra stuff (the new PC is on the |
| * stack, the location of notif_disabled, and a clobbered work register) to |
| * enable notifs, make sure notif IPIs weren't pending, restore the work reg, |
| * and then "ret". |
| * |
| * This is what the target uthread's stack will look like (growing down): |
| * |
| * Target ESP -> | u_thread's old stuff | the future %esp, tf->tf_esp |
| * | new eip | 0x04 below %esp (one slot is 0x04) |
| * | eflags space | 0x08 below |
| * | eax save space | 0x0c below |
| * | actual syscall | 0x10 below (0x30 space) |
| * | *sysc ptr to syscall | 0x40 below (0x10 + 0x30) |
| * | notif_pending_loc | 0x44 below (0x10 + 0x30) |
| * | notif_disabled_loc | 0x48 below (0x10 + 0x30) |
| * |
| * The important thing is that it can handle a notification after it enables |
| * notifications, and when it gets resumed it can ultimately run the new |
| * context. Enough state is saved in the running context and stack to continue |
| * running. |
| * |
| * Related to that is whether or not our stack pointer is sufficiently far down |
| * so that restarting *this* code won't clobber shit we need later. The way we |
| * do this is that we do any "stack jumping" before we enable interrupts/notifs. |
| * These jumps are when we directly modify esp, specifically in the down |
| * direction (subtracts). Adds would be okay. |
| * |
| * Another related concern is the storage for sysc. It used to be on the |
| * vcore's stack, but if an interrupt comes in before we use it, we trash the |
| * vcore's stack (and thus the storage for sysc!). Instead, we put it on the |
| * stack of the user tf. Moral: don't touch a vcore's stack with notifs |
| * enabled. */ |
| |
| /* Helper for writing the info we need later to the u_tf's stack. Note, this |
| * could get fucked if the struct syscall isn't a multiple of 4-bytes. Also, |
| * note this goes backwards, since memory reads up the stack. */ |
| struct restart_helper { |
| uint32_t notif_disab_loc; |
| uint32_t notif_pend_loc; |
| struct syscall *sysc; |
| struct syscall local_sysc; /* unused for now */ |
| uint32_t eax_save; |
| uint32_t eflags; |
| uint32_t eip; |
| }; |
| |
| /* Static syscall, used for self-notifying. We never wait on it, and we |
| * actually might submit it multiple times in parallel on different cores! |
| * While this may seem dangerous, the kernel needs to be able to handle this |
| * scenario. It's also important that we never wait on this, since for all but |
| * the first call, the DONE flag will be set. (Set once, then never reset) */ |
| extern struct syscall vc_entry; /* in x86/vcore.c */ |
| |
| static inline void pop_hw_tf(struct hw_trapframe *tf, uint32_t vcoreid) |
| { |
| struct restart_helper *rst; |
| struct preempt_data *vcpd = &__procdata.vcore_preempt_data[vcoreid]; |
| if (!tf->tf_cs) { /* sysenter TF. esp and eip are in other regs. */ |
| tf->tf_esp = tf->tf_regs.reg_ebp; |
| tf->tf_eip = tf->tf_regs.reg_edx; |
| } |
| /* The stuff we need to write will be below the current stack of the utf */ |
| rst = (struct restart_helper*)((void*)tf->tf_esp - |
| sizeof(struct restart_helper)); |
| /* Fill in the info we'll need later */ |
| rst->notif_disab_loc = (uint32_t)&vcpd->notif_disabled; |
| rst->notif_pend_loc = (uint32_t)&vcpd->notif_pending; |
| rst->sysc = &vc_entry; |
| rst->eax_save = 0; /* avoid bugs */ |
| rst->eflags = tf->tf_eflags; |
| rst->eip = tf->tf_eip; |
| |
| asm volatile ("movl %0,%%esp; " /* jump esp to the utf */ |
| "popal; " /* restore normal registers */ |
| "addl $0x24,%%esp; " /* move to the esp slot in the tf */ |
| "popl %%esp; " /* change to the utf's %esp */ |
| "subl $0x08,%%esp; " /* move esp to below eax's slot */ |
| "pushl %%eax; " /* save eax, will clobber soon */ |
| "movl %2,%%eax; " /* sizeof struct syscall */ |
| "addl $0x0c,%%eax; " /* more offset btw eax/notif_en_loc*/ |
| "subl %%eax,%%esp; " /* move to notif_en_loc slot */ |
| "popl %%eax; " /* load notif_disabled addr */ |
| "movb $0x00,(%%eax); " /* enable notifications */ |
| /* Need a wrmb() here so the write of enable_notif can't pass |
| * the read of notif_pending (racing with a potential |
| * cross-core call with proc_notify()). */ |
| "lock addl $0,(%%esp); " /* LOCK is a CPU mb() */ |
| /* From here down, we can get interrupted and restarted */ |
| "popl %%eax; " /* get notif_pending status */ |
| "testb $0x01,(%%eax); " /* test if a notif is pending */ |
| "jz 1f; " /* if not pending, skip syscall */ |
| /* Actual syscall. Note we don't wait on the async call */ |
| "popl %%eax; " /* &sysc, trap arg0 */ |
| "pushl %%edx; " /* save edx, will be trap arg1 */ |
| "movl $0x1,%%edx; " /* sending one async syscall: arg1 */ |
| "int %1; " /* fire the syscall */ |
| "popl %%edx; " /* restore regs after syscall */ |
| "jmp 2f; " /* skip 1:, already popped */ |
| "1: popl %%eax; " /* discard &sysc (on non-sc path) */ |
| "2: addl %2,%%esp; " /* jump over the sysc (both paths) */ |
| "popl %%eax; " /* restore tf's %eax */ |
| "popfl; " /* restore utf's eflags */ |
| "ret; " /* return to the new PC */ |
| : |
| : "g"(tf), "i"(T_SYSCALL), "i"(sizeof(struct syscall)) |
| : "memory"); |
| } |
| |
| static inline void pop_sw_tf(struct sw_trapframe *sw_tf, uint32_t vcoreid) |
| { |
| struct preempt_data *vcpd = &__procdata.vcore_preempt_data[vcoreid]; |
| |
| /* Restore callee-saved FPU state. We need to clear exceptions before |
| * reloading the FP CW, in case the new CW unmasks any. We also need to |
| * reset the tag word to clear out the stack. |
| * |
| * The main issue here is that while our context was saved in an |
| * ABI-complaint manner, we may be starting up on a somewhat random FPU |
| * state. Having gibberish in registers isn't a big deal, but some of the |
| * FP environment settings could cause trouble. If fnclex; emms isn't |
| * enough, we could also save/restore the entire FP env with fldenv, or do |
| * an fninit before fldcw. */ |
| asm volatile ("ldmxcsr %0" : : "m"(sw_tf->tf_mxcsr)); |
| asm volatile ("fnclex; emms; fldcw %0" : : "m"(sw_tf->tf_fpucw)); |
| /* Basic plan: restore all regs, off ecx as the sw_tf. Switch to the new |
| * stack, push the PC so we can pop it later. Use eax and edx for the |
| * locations of sysc and vcpd. Once on the new stack, we enable notifs, |
| * check if we missed one, and if so, self notify. */ |
| asm volatile ("movl 0x00(%0),%%ebp; " /* restore regs */ |
| "movl 0x04(%0),%%ebx; " |
| "movl 0x08(%0),%%esi; " |
| "movl 0x0c(%0),%%edi; " |
| "movl 0x10(%0),%%esp; " /* jump to future stack */ |
| "pushl 0x14(%0); " /* save PC for future ret */ |
| "movl %2,%%ecx; " /* vcpd loc into ecx */ |
| "addl %4,%%ecx; " /* notif_disabled loc into ecx */ |
| "movb $0x00,(%%ecx); " /* enable notifications */ |
| /* Need a wrmb() here so the write of enable_notif can't pass |
| * the read of notif_pending (racing with a potential |
| * cross-core call with proc_notify()). */ |
| "lock addl $0,(%%esp); " /* LOCK is a CPU mb() */ |
| /* From here down, we can get interrupted and restarted */ |
| "movl %2,%%ecx; " /* vcpd loc into ecx */ |
| "addl %5,%%ecx; " /* notif_pending loc into ecx */ |
| "testb $0x01,(%%ecx); " /* test if a notif is pending */ |
| "jz 1f; " /* if not pending, skip syscall */ |
| /* Actual syscall. Note we don't wait on the async call. |
| * &sysc is already in eax (trap arg0). */ |
| "movl $0x1,%%edx; " /* sending one async syscall: arg1 */ |
| "int %3; " /* fire the syscall */ |
| "1: ret; " /* retaddr was pushed earlier */ |
| : |
| : "c"(sw_tf), |
| "a"(&vc_entry), |
| "d"(vcpd), |
| "i"(T_SYSCALL), |
| "i"(offsetof(struct preempt_data, notif_disabled)), |
| "i"(offsetof(struct preempt_data, notif_pending)) |
| : "memory"); |
| } |
| |
| /* Pops a user context, reanabling notifications at the same time. A Userspace |
| * scheduler can call this when transitioning off the transition stack. |
| * |
| * At some point in vcore context before calling this, you need to clear |
| * notif_pending (do this by calling handle_events()). As a potential |
| * optimization, consider clearing the notif_pending flag / handle_events again |
| * (right before popping), right before calling this. If notif_pending is not |
| * clear, this will self_notify this core, since it should be because we missed |
| * a notification message while notifs were disabled. */ |
| static inline void pop_user_ctx(struct user_context *ctx, uint32_t vcoreid) |
| { |
| if (ctx->type == ROS_HW_CTX) |
| pop_hw_tf(&ctx->tf.hw_tf, vcoreid); |
| else |
| pop_sw_tf(&ctx->tf.sw_tf, vcoreid); |
| } |
| |
| /* Like the regular pop_user_ctx, but this one doesn't check or clear |
| * notif_pending. The only case where we use this is when an IRQ/notif |
| * interrupts a uthread that is in the process of disabling notifs. */ |
| static inline void pop_user_ctx_raw(struct user_context *ctx, uint32_t vcoreid) |
| { |
| struct hw_trapframe *tf = &ctx->tf.hw_tf; |
| assert(ctx->type == ROS_HW_CTX); |
| struct restart_helper *rst; |
| struct preempt_data *vcpd = &__procdata.vcore_preempt_data[vcoreid]; |
| if (!tf->tf_cs) { /* sysenter TF. esp and eip are in other regs. */ |
| tf->tf_esp = tf->tf_regs.reg_ebp; |
| tf->tf_eip = tf->tf_regs.reg_edx; |
| } |
| /* The stuff we need to write will be below the current stack of the utf */ |
| rst = (struct restart_helper*)((void*)tf->tf_esp - |
| sizeof(struct restart_helper)); |
| /* Fill in the info we'll need later */ |
| rst->notif_disab_loc = (uint32_t)&vcpd->notif_disabled; |
| rst->eax_save = 0; /* avoid bugs */ |
| rst->eflags = tf->tf_eflags; |
| rst->eip = tf->tf_eip; |
| |
| asm volatile ("movl %0,%%esp; " /* jump esp to the utf */ |
| "popal; " /* restore normal registers */ |
| "addl $0x24,%%esp; " /* move to the esp slot in the tf */ |
| "popl %%esp; " /* change to the utf's %esp */ |
| "subl $0x08,%%esp; " /* move esp to below eax's slot */ |
| "pushl %%eax; " /* save eax, will clobber soon */ |
| "movl %2,%%eax; " /* sizeof struct syscall */ |
| "addl $0x0c,%%eax; " /* more offset btw eax/notif_en_loc*/ |
| "subl %%eax,%%esp; " /* move to notif_en_loc slot */ |
| "popl %%eax; " /* load notif_disabled addr */ |
| "movb $0x00,(%%eax); " /* enable notifications */ |
| /* Here's where we differ from the regular pop_user_ctx(). |
| * We do the same pops/esp moves, just to keep things similar |
| * and simple, but don't do test, clear notif_pending, or |
| * call a syscall. */ |
| /* From here down, we can get interrupted and restarted */ |
| "popl %%eax; " /* get notif_pending status */ |
| "popl %%eax; " /* discard &sysc (on non-sc path) */ |
| "addl %2,%%esp; " /* jump over the sysc (both paths) */ |
| "popl %%eax; " /* restore tf's %eax */ |
| "popfl; " /* restore utf's eflags */ |
| "ret; " /* return to the new PC */ |
| : |
| : "g"(tf), "i"(T_SYSCALL), "i"(sizeof(struct syscall)) |
| : "memory"); |
| } |
| |
| /* Save's a SW context, setting the PC to the end of this function. We only |
| * save callee-saved registers (of the sysv abi). The compiler knows to save |
| * the others via the input/clobber lists. |
| * |
| * Callers of this function need to have at least one |
| * 'calling-convention-compliant' function call between this and any floating |
| * point, so that the compiler saves any caller-saved FP before getting to |
| * here. |
| * |
| * To some extent, TLS is 'callee-saved', in that no one ever expects it to |
| * change. We handle uthread TLS changes separately, since we often change to |
| * them early to set some variables. Arguably we should do this different. */ |
| static inline void save_user_ctx(struct user_context *ctx) |
| { |
| struct sw_trapframe *sw_tf = &ctx->tf.sw_tf; |
| long dummy; |
| ctx->type = ROS_SW_CTX; |
| asm volatile ("stmxcsr %0" : "=m"(sw_tf->tf_mxcsr)); |
| asm volatile ("fnstcw %0" : "=m"(sw_tf->tf_fpucw)); |
| /* Pretty simple: save all the regs, IAW the sys-v ABI */ |
| asm volatile ("movl %%ebp,0x00(%0); " |
| "movl %%ebx,0x04(%0); " |
| "movl %%esi,0x08(%0); " |
| "movl %%edi,0x0c(%0); " |
| "movl %%esp,0x10(%0); " |
| "leal 1f,%%eax; " /* get future eip */ |
| "movl %%eax,0x14(%0); " |
| "1: " /* where this tf will restart */ |
| : "=c"(dummy) /* force clobber for ecx */ |
| : "c"(sw_tf) |
| : "eax", "edx", "memory", "cc"); |
| } __attribute__((always_inline, returns_twice)) |
| |
| /* The old version, kept around for testing */ |
| static inline void save_user_ctx_hw(struct user_context *ctx) |
| { |
| struct hw_trapframe *tf = &ctx->tf.hw_tf; |
| ctx->type = ROS_HW_CTX; |
| memset(tf, 0, sizeof(struct hw_trapframe)); /* sanity */ |
| /* set CS and make sure eflags is okay */ |
| tf->tf_cs = GD_UT | 3; |
| tf->tf_eflags = 0x00000200; /* interrupts enabled. bare minimum eflags. */ |
| /* Save the regs and the future esp. */ |
| asm volatile("movl %%esp,(%0); " /* save esp in it's slot*/ |
| "pushl %%eax; " /* temp save eax */ |
| "leal 1f,%%eax; " /* get future eip */ |
| "movl %%eax,(%1); " /* store future eip */ |
| "popl %%eax; " /* restore eax */ |
| "movl %2,%%esp; " /* move to the beginning of the tf */ |
| "addl $0x20,%%esp; " /* move to after the push_regs */ |
| "pushal; " /* save regs */ |
| "addl $0x44,%%esp; " /* move to esp slot */ |
| "popl %%esp; " /* restore esp */ |
| "1: " /* where this tf will restart */ |
| : |
| : "g"(&tf->tf_esp), "g"(&tf->tf_eip), "g"(tf) |
| : "eax", "memory", "cc"); |
| } __attribute__((always_inline, returns_twice)) |
| |
| static inline void init_user_ctx(struct user_context *ctx, uint32_t entry_pt, |
| uint32_t stack_top) |
| { |
| struct sw_trapframe *sw_tf = &ctx->tf.sw_tf; |
| ctx->type = ROS_SW_CTX; |
| /* No need to bother with setting the other GP registers; the called |
| * function won't care about their contents. */ |
| sw_tf->tf_esp = stack_top; |
| sw_tf->tf_eip = entry_pt; |
| sw_tf->tf_mxcsr = 0x00001f80; /* x86 default mxcsr */ |
| sw_tf->tf_fpucw = 0x037f; /* x86 default FP CW */ |
| } |
| |
| // this is how we get our thread id on entry. |
| #define __vcore_id_on_entry \ |
| ({ \ |
| register int temp asm ("eax"); \ |
| temp; \ |
| }) |
| |
| /* For debugging. */ |
| #include <stdio.h> |
| static void print_hw_tf(struct hw_trapframe *tf) |
| { |
| printf("[user] HW TRAP frame %08p\n", tf); |
| printf(" edi 0x%08x\n", tf->tf_regs.reg_edi); |
| printf(" esi 0x%08x\n", tf->tf_regs.reg_esi); |
| printf(" ebp 0x%08x\n", tf->tf_regs.reg_ebp); |
| printf(" oesp 0x%08x\n", tf->tf_regs.reg_oesp); |
| printf(" ebx 0x%08x\n", tf->tf_regs.reg_ebx); |
| printf(" edx 0x%08x\n", tf->tf_regs.reg_edx); |
| printf(" ecx 0x%08x\n", tf->tf_regs.reg_ecx); |
| printf(" eax 0x%08x\n", tf->tf_regs.reg_eax); |
| printf(" gs 0x----%04x\n", tf->tf_gs); |
| printf(" fs 0x----%04x\n", tf->tf_fs); |
| printf(" es 0x----%04x\n", tf->tf_es); |
| printf(" ds 0x----%04x\n", tf->tf_ds); |
| printf(" trap 0x%08x\n", tf->tf_trapno); |
| printf(" err 0x%08x\n", tf->tf_err); |
| printf(" eip 0x%08x\n", tf->tf_eip); |
| printf(" cs 0x----%04x\n", tf->tf_cs); |
| printf(" flag 0x%08x\n", tf->tf_eflags); |
| printf(" esp 0x%08x\n", tf->tf_esp); |
| printf(" ss 0x----%04x\n", tf->tf_ss); |
| } |
| |
| static void print_sw_tf(struct sw_trapframe *sw_tf) |
| { |
| printf("[user] SW TRAP frame %08p\n", sw_tf); |
| printf(" ebp 0x%08x\n", sw_tf->tf_ebp); |
| printf(" ebx 0x%08x\n", sw_tf->tf_ebx); |
| printf(" esi 0x%08x\n", sw_tf->tf_esi); |
| printf(" edi 0x%08x\n", sw_tf->tf_edi); |
| printf(" esp 0x%08x\n", sw_tf->tf_esp); |
| printf(" eip 0x%08x\n", sw_tf->tf_eip); |
| printf(" mxcsr 0x%08x\n", sw_tf->tf_mxcsr); |
| printf(" fpucw 0x----%04x\n", sw_tf->tf_fpucw); |
| } |
| |
| static void print_user_context(struct user_context *ctx) |
| { |
| if (ctx->type == ROS_HW_CTX) |
| print_hw_tf(&ctx->tf.hw_tf); |
| else if (ctx->type == ROS_SW_CTX) |
| print_sw_tf(&ctx->tf.sw_tf); |
| else |
| printf("Unknown context type %d\n", ctx->type); |
| } |
| |
| #endif /* PARLIB_ARCH_VCORE32_H */ |
