void show_regs(struct pt_regs * regs)
{
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
printk("\n");
printk("Pid: %d, comm: %20s\n", current->pid, current->comm);
printk("EIP: %04x:[<%08lx>] CPU: %d\n",0xffff & regs->xcs,regs->eip, smp_processor_id());
print_symbol("EIP is at %s\n", regs->eip);
if (user_mode_vm(regs))
printk(" ESP: %04x:%08lx",0xffff & regs->xss,regs->esp);
printk(" EFLAGS: %08lx %s (%s %.*s)\n",
regs->eflags, print_tainted(), system_utsname.release,
(int)strcspn(system_utsname.version, " "),
system_utsname.version);
printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
regs->eax,regs->ebx,regs->ecx,regs->edx);
printk("ESI: %08lx EDI: %08lx EBP: %08lx",
regs->esi, regs->edi, regs->ebp);
printk(" DS: %04x ES: %04x\n",
0xffff & regs->xds,0xffff & regs->xes);
cr0 = read_cr0();
cr2 = read_cr2();
cr3 = read_cr3();
cr4 = read_cr4_safe();
printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n", cr0, cr2, cr3, cr4);
show_trace(NULL, regs, ®s->esp);
}
/*H:040
* This is the i386-specific code to setup and run the Guest. Interrupts
* are disabled: we own the CPU.
*/
void lguest_arch_run_guest(struct lg_cpu *cpu)
{
/*
* Remember the awfully-named TS bit? If the Guest has asked to set it
* we set it now, so we can trap and pass that trap to the Guest if it
* uses the FPU.
*/
if (cpu->ts && user_has_fpu())
stts();
/*
* SYSENTER is an optimized way of doing system calls. We can't allow
* it because it always jumps to privilege level 0. A normal Guest
* won't try it because we don't advertise it in CPUID, but a malicious
* Guest (or malicious Guest userspace program) could, so we tell the
* CPU to disable it before running the Guest.
*/
if (boot_cpu_has(X86_FEATURE_SEP))
wrmsr(MSR_IA32_SYSENTER_CS, 0, 0);
/*
* Now we actually run the Guest. It will return when something
* interesting happens, and we can examine its registers to see what it
* was doing.
*/
run_guest_once(cpu, lguest_pages(raw_smp_processor_id()));
/*
* Note that the "regs" structure contains two extra entries which are
* not really registers: a trap number which says what interrupt or
* trap made the switcher code come back, and an error code which some
* traps set.
*/
/* Restore SYSENTER if it's supposed to be on. */
if (boot_cpu_has(X86_FEATURE_SEP))
wrmsr(MSR_IA32_SYSENTER_CS, __KERNEL_CS, 0);
/* Clear the host TS bit if it was set above. */
if (cpu->ts && user_has_fpu())
clts();
/*
* If the Guest page faulted, then the cr2 register will tell us the
* bad virtual address. We have to grab this now, because once we
* re-enable interrupts an interrupt could fault and thus overwrite
* cr2, or we could even move off to a different CPU.
*/
if (cpu->regs->trapnum == 14)
cpu->arch.last_pagefault = read_cr2();
/*
* Similarly, if we took a trap because the Guest used the FPU,
* we have to restore the FPU it expects to see.
* math_state_restore() may sleep and we may even move off to
* a different CPU. So all the critical stuff should be done
* before this.
*/
else if (cpu->regs->trapnum == 7 && !user_has_fpu())
math_state_restore();
}
static void __save_processor_state(struct saved_context *ctxt)
{
mtrr_save_fixed_ranges(NULL);
kernel_fpu_begin();
/*
* descriptor tables
*/
store_gdt(&ctxt->gdt);
store_idt(&ctxt->idt);
store_tr(ctxt->tr);
/*
* segment registers
*/
savesegment(es, ctxt->es);
savesegment(fs, ctxt->fs);
savesegment(gs, ctxt->gs);
savesegment(ss, ctxt->ss);
/*
* control registers
*/
ctxt->cr0 = read_cr0();
ctxt->cr2 = read_cr2();
ctxt->cr3 = read_cr3();
ctxt->cr4 = read_cr4_safe();
}
void
do_page_fault(struct pt_regs *regs, unsigned int vector)
{
/* Kernel space exception fixup check */
if (fixup_exception(regs))
return;
unsigned long cr2 = read_cr2();
#ifdef CONFIG_KGDB
kgdb_breakpoint();
#endif
struct siginfo s;
memset(&s, 0, sizeof(struct siginfo));
s.si_signo = SIGSEGV;
s.si_errno = 0;
s.si_addr = (void *)cr2;
s.si_code = SEGV_ACCERR; // SEGV_MAPERR;
int i = sigsend(current->aspace->id, ANY_ID, SIGSEGV, &s);
printk(">> PAGE FAULT! Sent signal %d. CR2 is %p, cpu %d\n", i, (void *)cr2, this_cpu);
show_registers(regs);
/* If the fault occurred in kernel space, or the init_task, go down */
if ( (regs->rip >= PAGE_OFFSET) ||
(current->aspace->id == INIT_ASPACE_ID))
{
local_irq_disable();
halt();
}
}
void PageTable::handle_fault(REGS * _r)
{
/* Page fault handler */
//Console::puts("Page Fault Handler Called\n");
unsigned long addr = 0, page_dir_index = 0, page_table_index = 0, page_offset = 0;
unsigned long page_fault_dir = read_cr2();
//Console::puts("Virtual Address = ");
Console::putui(page_fault_dir);
/* Page directory index is stored in the first 10 bits of the virtual address */
page_dir_index = SHIFT_RIGHT (page_fault_dir, PAGE_DIR_SHIFT) & 0x3FF;
/* Page table index is stored in the second 10 bits of the virtual address */
page_table_index = SHIFT_RIGHT (page_fault_dir, PAGE_TABLE_SHIFT) & 0x3FF;
/* Page offset is stored in the first 12 bits of the virtual address */
page_offset = page_fault_dir & 0xFFF;
if ((current_page_table->page_directory[page_dir_index] & 0x1) == 0)
/* The page table is not present in the page directory.
* So allocate a frame from the kernel frame pool and mark the page as present.
*/
{
current_page_table->page_directory[page_dir_index] = ((PageTable::kernel_mem_pool->get_frame()) * PageTable::PAGE_SIZE ) | 3;
}
((unsigned long*)((current_page_table->page_directory[page_dir_index]) & 0xFFFFF000) )[page_table_index] = ( (PageTable::process_mem_pool->get_frame()) * (PageTable::PAGE_SIZE)) | 3;
}
void PageTable::handle_fault(REGS * _r)
{
if(!(_r->err_code & 0x2))
Console::puts("It's read only.\n");
if(!(_r->err_code & 0x1)) //if the requested page is not in memory
{
unsigned long virtualaddr=read_cr2();
unsigned long pdindex=virtualaddr>>22; //we need first 10bit
unsigned long * pd=(unsigned long *)read_cr3(); //get physical memory address which is also vm address of page directory
Console::putui((unsigned int)pd[pdindex]);
if((pd[pdindex] & 0x1)==0) //pagetable is not present
{
pd[pdindex]=(kernel_mem_pool->get_frame()*4096) | 3; //request frame address for storing page table,
//by calling kernel_mem_pool, we assume pagetable for a process reside in kernel memory
/*pt=(unsigned long *)(pd[pdindex] & 0xFFFFF000);
for(unsigned long i;i<=1023;i++)
pt[i]=2;*/
}
unsigned long * pt=(unsigned long *)(pd[pdindex] & 0xFFFFF000);
//Console::putui((unsigned int)pd[pdindex]);
unsigned long ptindex=(virtualaddr>>12) & 0x03FF; //we need next 10bit in the middle
pt[ptindex]=(process_mem_pool->get_frame()*4096) | 3; //request frame address for storing the page
//Console::putui((unsigned int)pt[ptindex]);
//for(;;);
}
/*-------------------------------------------------------------------------
* pfint - paging fault ISR
*-------------------------------------------------------------------------
*/
SYSCALL pfint()
{
kprintf("---------ISR----------\n");
unsigned int pfaddr = read_cr2();// faulted address
//Check that pfaddr is a legal address ????
unsigned int pdbaddr = read_cr3()&(~0xfff); //page directory base address
unsigned int p = pfaddr >> 22; // upper ten bits of faulted address
unsigned int q = (pfaddr >> 12)&0x3ff; // middle ten bits of faulted address
unsigned int offset = pfaddr & 0xfff; // last twelve bits
pd_t * pt = (pd_t *)pdbaddr + p ; //pth page table
//kprintf("%x\n",*pt);
if(pt->pd_pres == 0) // if page table is not present
{
int frm_num = get_frm(1,FR_TBL,pfaddr>>12);
//kprintf("%d\n",frm_num);
if(frm_num == SYSERR)
return SYSERR;
pt->pd_base = (0x00400000 + frm_num*4096) >> 12;
pt->pd_pres = 1;
write_cr3(pdbaddr);
kprintf("faulted addr: %x xth page table: %x, content: %x\n",pfaddr,pt,p);
return OK;
}
/**
@brief Page fault handler
*/
asmregparm void do_page_fault(struct pt_regs * regs, long error_code)
{
unsigned long error_address = read_cr2();
bool in_user = address_in_user(error_address);
struct ko_thread *current;
//printk("PAGE fault: eip %x addr %x code %x\n", regs->ip, error_address, error_code);
/*
Access violation.
读、写和u/s异常
如果是用户模式，那么访问核心层空间时崩溃，其他地方尝试修复
如果是特权模式，都尝试修复
*/
if (error_code & PAGE_FAULT_U)
{
if (in_user == false)
goto die_cur;
}
current = kt_current();
if (unlikely(false == ks_exception(current, error_address, error_code)))
goto die_cur;
return;
die_cur:
printk("TODO: Thread deing :eip %x addr %x code %x\n", regs->ip, error_address, error_code);
//TODO kill thread
//fak_arch_x86_dump_register(regs);
kt_delete_current();
}
void do_page_fault(struct pt_regs *regs, unsigned long error_code)
{
unsigned long addr = read_cr2();
struct sched_shutdown sched_shutdown = { .reason = SHUTDOWN_crash };
if ((error_code & TRAP_PF_WRITE) && handle_cow(addr))
return;
/* If we are already handling a page fault, and got another one
that means we faulted in pagetable walk. Continuing here would cause
a recursive fault */
if(handling_pg_fault == 1)
{
printk("Page fault in pagetable walk (access to invalid memory?).\n");
HYPERVISOR_sched_op(SCHEDOP_shutdown, &sched_shutdown);
}
handling_pg_fault++;
barrier();
printk("Page fault at linear address %p, rip %p, regs %p, sp %p, our_sp %p, code %lx\n",
addr, regs->rip, regs, regs->rsp, &addr, error_code);
dump_regs(regs);
//do_stack_walk(regs->rbp);
dump_mem(regs->rsp);
dump_mem(regs->rbp);
dump_mem(regs->rip);
page_walk(addr);
HYPERVISOR_sched_op(SCHEDOP_shutdown, &sched_shutdown);
/* We should never get here ... but still */
handling_pg_fault--;
}
void
do_page_fault(struct pt_regs *regs, unsigned int vector)
{
/* Kernel space exception fixup check */
if (fixup_exception(regs))
return;
unsigned long cr2 = read_cr2();
#ifdef CONFIG_KGDB
kgdb_breakpoint();
#endif
struct siginfo s;
memset(&s, 0, sizeof(struct siginfo));
s.si_signo = SIGSEGV;
s.si_errno = 0;
s.si_addr = (void *)cr2;
s.si_code = SEGV_ACCERR; // SEGV_MAPERR;
int i = sigsend(current->aspace->id, ANY_ID, SIGSEGV, &s);
printk(">> PAGE FAULT! Sent signal %d: CR2 is %p\n", i, s.si_addr);
show_registers(regs);
while (1) {}
}
void __show_registers(struct pt_regs *regs, int all)
{
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
unsigned long d0, d1, d2, d3, d6, d7;
unsigned long sp;
unsigned short ss, gs;
if (user_mode_vm(regs)) {
sp = regs->sp;
ss = regs->ss & 0xffff;
savesegment(gs, gs);
} else {
sp = (unsigned long) (®s->sp);
savesegment(ss, ss);
savesegment(gs, gs);
}
printk("\n");
printk("Pid: %d, comm: %s %s (%s %.*s)\n",
task_pid_nr(current), current->comm,
print_tainted(), init_utsname()->release,
(int)strcspn(init_utsname()->version, " "),
init_utsname()->version);
printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
(u16)regs->cs, regs->ip, regs->flags,
smp_processor_id());
print_symbol("EIP is at %s\n", regs->ip);
printk("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
regs->ax, regs->bx, regs->cx, regs->dx);
printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
regs->si, regs->di, regs->bp, sp);
printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
(u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
if (!all)
return;
cr0 = read_cr0();
cr2 = read_cr2();
cr3 = read_cr3();
cr4 = read_cr4_safe();
printk("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
cr0, cr2, cr3, cr4);
get_debugreg(d0, 0);
get_debugreg(d1, 1);
get_debugreg(d2, 2);
get_debugreg(d3, 3);
printk("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
d0, d1, d2, d3);
get_debugreg(d6, 6);
get_debugreg(d7, 7);
printk("DR6: %08lx DR7: %08lx\n",
d6, d7);
}
void __show_regs(struct pt_regs *regs, int all)
{
#ifdef NOT_FOR_L4
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
unsigned long d0, d1, d2, d3, d6, d7;
#endif
unsigned long sp;
unsigned short ss, gs;
if (user_mode_vm(regs)) {
sp = regs->sp;
ss = regs->ss & 0xffff;
gs = get_user_gs(regs);
} else {
sp = kernel_stack_pointer(regs);
savesegment(ss, ss);
savesegment(gs, gs);
}
show_regs_common();
printk(KERN_DEFAULT "EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
(u16)regs->cs, regs->ip, regs->flags,
smp_processor_id());
print_symbol("EIP is at %s\n", regs->ip);
printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
regs->ax, regs->bx, regs->cx, regs->dx);
printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
regs->si, regs->di, regs->bp, sp);
printk(KERN_DEFAULT " DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
(u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
if (!all)
return;
#ifdef NOT_FOR_L4
cr0 = read_cr0();
cr2 = read_cr2();
cr3 = read_cr3();
cr4 = read_cr4_safe();
printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
cr0, cr2, cr3, cr4);
get_debugreg(d0, 0);
get_debugreg(d1, 1);
get_debugreg(d2, 2);
get_debugreg(d3, 3);
printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
d0, d1, d2, d3);
get_debugreg(d6, 6);
get_debugreg(d7, 7);
printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n",
d6, d7);
#endif
}
void fault(struct tcb *tcb, struct ccb *ccb) {
if (tcb->ivect == 32) {
ccb->lapic->eoi = 0;
scheduler_schedule();
return;
}
if (tcb->ivect == 0x100) {
if (tcb->rdi == 0 && tcb->state != TCB_STATE_RUNNING) {
ccb_unload_tcb();
}
scheduler_schedule();
return;
}
if (tcb->ivect == 14) {
extern uint64_t read_cr2(void);
log(ERROR, "page fault at addr %p (ip %p, sp %p, id %d)", read_cr2(), tcb->rip, tcb->rsp, tcb->id);
log(ERROR, "translation: %p %p", pcx_get_trans(NULL, read_cr2()), pcx_get_flags(NULL, read_cr2()));
queue_pagefault(ccb_unload_tcb());
interrupt_vector_fire(__PINION_INTERRUPT_PAGEFAULT);
scheduler_schedule();
return;
}
if (tcb->ivect == 8) {
log(ERROR, "PANIC: double fault");
for(;;);
}
if (tcb->ivect < 32) {
queue_miscfault(ccb_unload_tcb());
interrupt_vector_fire(__PINION_INTERRUPT_MISCFAULT);
scheduler_schedule();
return;
}
}
void __show_regs(struct pt_regs *regs, int all)
{
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
unsigned long d0, d1, d2, d3, d6, d7;
unsigned long sp;
unsigned short ss, gs;
if (user_mode(regs)) {
sp = regs->sp;
ss = regs->ss & 0xffff;
gs = get_user_gs(regs);
} else {
sp = kernel_stack_pointer(regs);
savesegment(ss, ss);
savesegment(gs, gs);
}
printk(KERN_DEFAULT "EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
(u16)regs->cs, regs->ip, regs->flags,
smp_processor_id());
print_symbol("EIP is at %s\n", regs->ip);
printk(KERN_DEFAULT "EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
regs->ax, regs->bx, regs->cx, regs->dx);
printk(KERN_DEFAULT "ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
regs->si, regs->di, regs->bp, sp);
printk(KERN_DEFAULT " DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
(u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
if (!all)
return;
cr0 = read_cr0();
cr2 = read_cr2();
cr3 = read_cr3();
cr4 = __read_cr4_safe();
printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
cr0, cr2, cr3, cr4);
get_debugreg(d0, 0);
get_debugreg(d1, 1);
get_debugreg(d2, 2);
get_debugreg(d3, 3);
get_debugreg(d6, 6);
get_debugreg(d7, 7);
/* Only print out debug registers if they are in their non-default state. */
if ((d0 == 0) && (d1 == 0) && (d2 == 0) && (d3 == 0) &&
(d6 == DR6_RESERVED) && (d7 == 0x400))
return;
printk(KERN_DEFAULT "DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
d0, d1, d2, d3);
printk(KERN_DEFAULT "DR6: %08lx DR7: %08lx\n",
d6, d7);
}
void FH_page_fault(struct fault_ctx *fctx)
{
uintptr_t fault_addr, fixup_addr;
struct intr_stack_frame *stack_frame = fctx->istack_frame;
fault_addr = read_cr2();
fixup_addr = fixup_fault_address(stack_frame->rip);
if (IS_KERNEL_FAULT(fctx) && !fixup_addr) {
display_unhandled_pf_info(fctx, fault_addr);
goto stop_cpu;
}
else {
vmm_t *vmm = current_task()->task_mm;
uint32_t errmask = 0;
int ret = -EFAULT;
if (!PFAULT_READ(fctx->errcode))
errmask |= PFLT_WRITE;
if (PFAULT_NXE(fctx->errcode))
errmask |= PFLT_NOEXEC;
if (PFAULT_PROTECT(fctx->errcode))
errmask |= PFLT_PROTECT;
else
errmask |= PFLT_NOT_PRESENT;
ret = vmm_handle_page_fault(vmm, fault_addr, errmask);
if (!ret) {
return;
}
if (fixup_addr) {
goto handle_fixup;
}
display_unhandled_pf_info(fctx, fault_addr);
#ifdef CONFIG_DUMP_VMM_ON_FAULT
kprintf_fault("========= [Vrages Tree] =========");
vmranges_print_tree_dbg(vmm);
#endif /* CONFIG_DUMP_VMM_ON_FAULT */
#ifdef CONFIG_SEND_SIGSEGV_ON_FAULTS
send_sigsegv(fault_addr, ret);
return;
#endif /* CONFIG_SEND_SIGSEGV_ON_FAULTS */
}
stop_cpu:
__stop_cpu();
handle_fixup:
kprintf_fault("FIXUP = %p\n", fixup_addr);
stack_frame->rip = fixup_addr;
return;
}
dotraplinkage void do_page_fault(struct pt_regs *regs, long error_code)
{
unsigned long address = read_cr2();
__do_page_fault(regs, error_code, address);
printk("BUG: unable to handle kernel paging request at %#lx\n", address);
printk("CPU: %d PID: %u Comm: %s\n", smp_processor_id(), current->pid, current->comm);
show_regs(regs);
panic("panic at #PF");
}
void dump_regs(struct irq_regs *regs)
{
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
unsigned long d0, d1, d2, d3, d6, d7;
unsigned long sp;
printf("EIP: %04x:[<%08lx>] EFLAGS: %08lx\n",
(u16)regs->xcs, regs->eip, regs->eflags);
printf("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
regs->eax, regs->ebx, regs->ecx, regs->edx);
printf("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
regs->esi, regs->edi, regs->ebp, regs->esp);
printf(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
(u16)regs->xds, (u16)regs->xes, (u16)regs->xfs, (u16)regs->xgs, (u16)regs->xss);
cr0 = read_cr0();
cr2 = read_cr2();
cr3 = read_cr3();
cr4 = read_cr4();
printf("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
cr0, cr2, cr3, cr4);
d0 = get_debugreg(0);
d1 = get_debugreg(1);
d2 = get_debugreg(2);
d3 = get_debugreg(3);
printf("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
d0, d1, d2, d3);
d6 = get_debugreg(6);
d7 = get_debugreg(7);
printf("DR6: %08lx DR7: %08lx\n",
d6, d7);
printf("Stack:\n");
sp = regs->esp;
sp += 64;
while (sp > (regs->esp - 16)) {
if (sp == regs->esp)
printf("--->");
else
printf(" ");
printf("0x%8.8lx : 0x%8.8lx\n", sp, (ulong)readl(sp));
sp -= 4;
}
}
void page_fault(unsigned int n, istate_t *istate)
{
uintptr_t badvaddr = read_cr2();
if (istate->error_word & PFERR_CODE_RSVD)
panic("Reserved bit set in page table entry.");
pf_access_t access;
if (istate->error_word & PFERR_CODE_RW)
access = PF_ACCESS_WRITE;
else if (istate->error_word & PFERR_CODE_ID)
access = PF_ACCESS_EXEC;
else
access = PF_ACCESS_READ;
(void) as_page_fault(badvaddr, access, istate);
}
void PageTable::handle_fault(REGS* _r)
{
/* check what exception? */
unsigned long faulting_addr = read_cr2();
/* compute the index of the page dir and page table */
int page_dir_index = faulting_addr >> 22;
int page_table_index = (faulting_addr >> 12) & MASK;
/* Perform the recursive page dir trick */
unsigned long* virt_page_directory = (unsigned long*)0xFFFFF000;
unsigned long* virt_page_table = 0;
/* check the present bit of the dir */
if(virt_page_directory[page_dir_index] == 2)
{
/* create page table, assign new frame for the faulting address */
unsigned long frame = process_mem_pool->get_frame();
// mark as present
virt_page_directory[page_dir_index] = frame | 3;
frame = process_mem_pool->get_frame();
virt_page_table = (unsigned long*)(0xFFC00000 | (page_dir_index << 12));
for(int i = 0; i < 1024; i++)
// mark as absent
virt_page_table[i] = 0 | 2;
// mark as present the page containing the faulting addr
virt_page_table[page_table_index] = frame | 3;
}
else
{
// page table present, assign page
unsigned long new_frame = process_mem_pool->get_frame_address(process_mem_pool->get_frame());
// unsigned long new_frame = process_mem_pool->get_frame();
virt_page_table = (unsigned long*)(0xFFC00000 | (page_dir_index << 12));
// page containing the faulting address to be marked present
virt_page_table[page_table_index] = new_frame | 3;
}
}
void lguest_arch_run_guest(struct lg_cpu *cpu)
{
if (cpu->ts)
unlazy_fpu(current);
if (boot_cpu_has(X86_FEATURE_SEP))
wrmsr(MSR_IA32_SYSENTER_CS, 0, 0);
run_guest_once(cpu, lguest_pages(raw_smp_processor_id()));
if (boot_cpu_has(X86_FEATURE_SEP))
wrmsr(MSR_IA32_SYSENTER_CS, __KERNEL_CS, 0);
if (cpu->regs->trapnum == 14)
cpu->arch.last_pagefault = read_cr2();
else if (cpu->regs->trapnum == 7)
math_state_restore();
}
/* All of our Exception handling Interrupt Service Routines will
* point to this function. This will tell us what exception has
* happened! Right now, we simply halt the system by hitting an
* endless loop. All ISRs disable interrupts while they are being
* serviced as a 'locking' mechanism to prevent an IRQ from
* happening and messing up kernel data structures */
extern "C" void fault_handler(struct cpu_state *r)
{
if (r->int_no < 32)
{
bool handled = false;
switch( r->int_no )
{
case 14: // Page fault
handled = paging_handle_fault( r, read_cr2() );
break;
default:
debug_bochs_printf( "EXCEPTION!" );
debug_bochs_printf( " (%x) (err_code = %x)\n", r->int_no, r->err_code );
debug_bochs_printf( "%s Exception\n", exception_messages[r->int_no] );
break;
}
if( !handled )
{
debug_bochs_printf( "\nRegister dump:\n" );
/*
unsigned int gs, fs, es, ds;
unsigned int edi, esi, ebp, esp, ebx, edx, ecx, eax;
unsigned int int_no, err_code;
unsigned int eip, cs, eflags, useresp, ss;
*/
debug_bochs_printf( "EAX = %x EBX = %x ECX = %x EDX = %x\n", r->eax, r->ebx, r->ecx, r->edx );
debug_bochs_printf( "ESP = %x EBP = %x ESI = %x EDI = %x\n", r->esp, r->ebp, r->esi, r->edi );
debug_bochs_printf( "GS = %x FS = %x ES = %x DS = %x\n", r->gs, r->fs, r->es, r->ds );
debug_bochs_printf( "EIP = %x CS = %x EFLAGS = %x SS = %x\n", r->eip, r->cs, r->eflags, r->ss );
debug_bochs_printf( "System Halted!\n" );
for (;;);
}
}
}
/*-------------------------------------------------------------------------
* pfint - paging fault ISR
*-------------------------------------------------------------------------
*/
SYSCALL pfint() {
STATWORD ps;
disable(ps);
int vp, p, q;
int i;
unsigned long fault_addr;
int store, pageth;
int new_frame, m_frame = -1;
pt_t *ptr_pt;
fault_addr = read_cr2();
vp = (fault_addr & ~(NBPG - 1)) >> 12;
//If process tries to access bs that it has not got using getbs.
if (bsm_lookup(currpid, fault_addr, &store, &pageth) == SYSERR) {
kill(currpid);
}
pd_t *ptr_pd = (pd_t *) (proctab[currpid].pdbr);
p = (fault_addr & 0xFFC00000) >> 22;
q = (fault_addr & 0x003FF000) >> 12;
ptr_pd = ptr_pd + p;
if (ptr_pd->pd_pres == 0) {
get_frm(&new_frame);
frm_tab[new_frame].fr_pid = currpid;
frm_tab[new_frame].fr_type = FR_TBL;
new_frame = new_frame + FRAME0;
add_page_dir(ptr_pd, new_frame, ((proctab[currpid].pdbr & ~(NBPG - 1)) >> 12) - FRAME0);
} else
void dump_regs(struct irq_regs *regs)
{
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
unsigned long d0, d1, d2, d3, d6, d7;
printf("EIP: %04x:[<%08lx>] EFLAGS: %08lx\n",
(u16)regs->xcs, regs->eip, regs->eflags);
printf("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
regs->eax, regs->ebx, regs->ecx, regs->edx);
printf("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
regs->esi, regs->edi, regs->ebp, regs->esp);
printf(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
(u16)regs->xds, (u16)regs->xes, (u16)regs->xfs, (u16)regs->xgs, (u16)regs->xss);
cr0 = read_cr0();
cr2 = read_cr2();
cr3 = read_cr3();
cr4 = read_cr4();
printf("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
cr0, cr2, cr3, cr4);
d0 = get_debugreg(0);
d1 = get_debugreg(1);
d2 = get_debugreg(2);
d3 = get_debugreg(3);
printf("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
d0, d1, d2, d3);
d6 = get_debugreg(6);
d7 = get_debugreg(7);
printf("DR6: %08lx DR7: %08lx\n",
d6, d7);
}
SYSCALL pfint()
{
unsigned long cr2,physical_addr;
virt_addr_t * vaddr;
int vp,s,o,avail,*store,*pageth;
unsigned int p,q,pt;
pd_t *pd;
pt_t *new_pt;
STATWORD ps;
// Disable interrupts
disable(ps);
if(GDB)
kprintf("\n*************pfint is running!************\n");
// Get the faulted address. The processor loads the CR2 register
// with the 32-bit address that generated the exception.
/* 1. Get the faulted address. */
cr2 = read_cr2();
vaddr = (virt_addr_t *)(&cr2);
if(GDB)
kprintf("&cr2=%x, cr2=%x, &vaddr=%x, vaddr=%x\n",&cr2,cr2,&vaddr,vaddr);
/* 2. Let 'vp' be the virtual page number of the page containing of the faulted address */
vp = a2pno(cr2);
if(GDB)
kprintf("vp=%d,\n",vp);
/* 3. Let pd point to the current page directory. */
pd = proctab[currpid].pdbr;
if(GDB)
kprintf("pd=%x,\n",pd);
/* 4. Check that a is a legal address (i.e., it has been mapped).
If it is not, print an error message and kill the process. */
pageth = getmem( sizeof(int *) );
store = getmem( sizeof(int *) );
if( SYSERR == bsm_lookup(currpid, vp, store, pageth)){
kprintf("ERROR: This virtual address hasn't been mapped!\n");
kill(currpid);
}
/* 5. Let p be the upper ten bits of a. [p represents page dirctory offset] */
/* 6. Let q be the bits [21:12] of a. [p represents page table offset.]
/* 7.1 Let pt point to the pth page table.*/
p = vaddr->pd_offset;
q = vaddr->pt_offset;
pt = vaddr->pg_offset;
if(GDB)
kprintf("p=%d,q=%d,pt=%d\n",p,q,pt);
/* 7.2 If the pth page table does not exist obtain a frame for it and initialize it. */
if(pd[p].pd_pres != 1){
if(GDB)
kprintf("**obtain a frame for the new page table. \n");
avail = get_frm(); //get the id of a new frame from frm_tab[];
if (avail == -1) {
if(GDB)
kprintf("Could not create page table!\n");
restore(ps);
return SYSERR;
}
//initialize frame[avail], update the process_id and frame_type of this frame.
init_frm(avail, currpid, FR_TBL);
frm_tab[avail].fr_upper_t = pa2frid((unsigned long) pd);
if(GDB)
kprintf("upper page table @frame[%d] pd=%x, a2pno(pd)=%d\n",frm_tab[avail].fr_upper_t, pd, a2pno((unsigned long) pd));
new_pt = frid2pa(avail);
init_pt(new_pt);
//update this page_table_entry in the page_directory.
pd[p].pd_pres = 1;
pd[p].pd_write = 1;
pd[p].pd_user = 0; // not sure about the usage;
pd[p].pd_pwt = 0;
pd[p].pd_pcd = 0;
pd[p].pd_acc = 0;
pd[p].pd_mbz = 0;
pd[p].pd_fmb = 0;
pd[p].pd_global = 0;
pd[p].pd_avail = 0; // not in use right now.
pd[p].pd_base = a2pno((unsigned long) new_pt); /* location of page table */
if(GDB)
kprintf("New page_table(%x)@frame[%d] updated in page_directory[%d]@(frame[%d])\n",
new_pt, avail, p, frm_tab[avail].fr_upper_t);
if(GDB)
kprintf("q=%d, new_pt[q]=%x, new_pt=%x, pd[p].pd_base=%d\n",
q, new_pt[q], new_pt, pd[p].pd_base);
}
//if the page table has already existed, just need to refcnt++;
else
{
int avail = pd[p].pd_base -1024;
frm_tab[avail].fr_refcnt++;
if(GDB)
kprintf("frm_tab[%d].fr_refcnt = %d, frame_type: %d\n",avail, frm_tab[avail].fr_refcnt, frm_tab[avail].fr_type);
}
/* 8.1 Using the backing store map, find the store s and page offset o which correspond to vp. */
//already saved in 'store' and 'pageth'
s = *store;
o = *pageth;
/* 8.2 In the inverted page table increment the reference count of the frame which holds pt.
This indicates that one more of pt's entries is marked "present." */
avail = find_frm(currpid,vp);
if (avail == -1)
{
if(GDB)
kprintf("allocating a page for the page fault\n");
avail = get_frm();
if(avail == -1)
{
if(GDB)
kprintf("ATTENTION! Frames full. ###Replacement NEEDED!###\n");
int frame_number = proctab[currpid].nframes-1;
int frame_id = proc_frames[currpid][0];
//update_proc_frames(pid,frame_number);
int i;
for (i = 0; i+1 < frame_number; ++i)
{
proc_frames[currpid][i] = proc_frames[currpid][i+1];
}
proctab[currpid].nframes = frame_number;
int pid = frm_tab[frame_id].fr_pid;
int upper_id = frm_tab[frame_id].fr_upper_t;
vp = frm_tab[frame_id].fr_vpno;
if(GDB)
kprintf("currpid=%d, frame[%d].pid=%d .vpno=%d, upper_frame[%d].ref=%d\n",currpid,frame_id,pid,vp,upper_id,frm_tab[upper_id].fr_refcnt);
p = vp>>10;
q = vp &0x003ff;
new_pt = vp2pa(pd[p].pd_base);
new_pt[q].pt_pres = 0;
new_pt[q].pt_write = 1;
new_pt[q].pt_base = 0;
if(GDB)
kprintf("pd_offset=%d, pt_offset=%d, pt_dirty=%d\n",p,q,new_pt[q].pt_dirty);
if(new_pt[q].pt_dirty == 1)
{
//write back and
pageth = getmem( sizeof(int *) );
store = getmem( sizeof(int *) );
if( SYSERR == bsm_lookup(currpid, vp, store, pageth)){
kprintf("ERROR: This virtual address hasn't been mapped!\n");
kill(currpid);
}
if(GDB)
kprintf("maping found: {pid: %d, vpno: %d, store: %d, pageth: %d}\n",currpid,vp,*store,*pageth);
write_bs((char *)new_pt, *store, *pageth);
}
init_pt(new_pt);
reset_frm(frame_id);
frm_tab[upper_id].fr_refcnt -= 2; //it is 2, not 1.
if(frm_tab[upper_id].fr_refcnt <= 0){
//mark the appropriate entry in pd as being not present, and free pt.
}
//invalidate the TLB entry for the page vp using the invlpg instruction
if(pid == currpid) {
set_PDBR(currpid);
}
}
else
{
init_frm(avail, currpid, FR_PAGE);
frm_tab[avail].fr_upper_t = pd[p].pd_base-FRAME0;
if(GDB)
kprintf("upper page table @frame[%d]\n",frm_tab[avail].fr_upper_t);
frm_tab[avail].fr_vpno = vp;
int counter = proctab[currpid].nframes;
proc_frames[currpid][counter] = frm_tab[avail].fr_id;
proctab[currpid].nframes++;
if(GDB)
kprintf("proc_frames[%d][%d] = frame[%d]\n",currpid,counter,avail);
// Add this frame to head of the frame list within the bs of this process
//(frm_tab[avail].bs_next)->fr_vpno
//, proctab[currpid].bsmap[s].frames->bs_next
if(GDB)
kprintf("&frm_tab[avail].bs_next = %x\n",frm_tab[avail].bs_next, &frm_tab[avail].bs_next);
if(GDB)
kprintf("proctab[%d].bsmap[%d].frames = %x, ->vpno=%d, ->bs_next=%x\n",currpid, s, proctab[currpid].bsmap[s].frames, proctab[currpid].bsmap[s].frames->fr_vpno, proctab[currpid].bsmap[s].frames->bs_next);
frm_tab[avail].bs_next = getmem(sizeof(fr_map_t *));
frm_tab[avail].bs_next = proctab[currpid].bsmap[s].frames;
proctab[currpid].bsmap[s].frames = &frm_tab[avail];
fr_map_t *frame = proctab[currpid].bsmap[s].frames;
int i = frame->fr_vpno;
if(GDB)
kprintf("i = %d\n",i);
if(GDB)
kprintf("~~~frame[%d] linked to frame[%d]\n", avail, frame->bs_next==NULL?-1:frame->bs_next->fr_id);
if(GDB)
kprintf("frame[%d].bs_next = %x, &**=%x\n",avail,frm_tab[avail].bs_next, &frm_tab[avail].bs_next);
if(GDB)
kprintf("proctab[%d].bsmap[%d].frames = %x, ->vpno=%d, ->bs_next=%x\n",currpid, s, proctab[currpid].bsmap[s].frames, proctab[currpid].bsmap[s].frames->fr_vpno, proctab[currpid].bsmap[s].frames->bs_next);
if(GDB)
kprintf("Mapping frame[%d](ppno[%d]) to {pid[%d], vpno[%d]} -> {bs[%d],offset:%d}\n",
avail,frid2vpno(avail),currpid,vp,s,o);
physical_addr = frid2pa(avail);
read_bs(physical_addr,s,o);
if(GDB)
kprintf("copied from bs[%d]:offset[%d] to vp[%d]@(%x)\n",s,o,vp,vp2pa(vp));
}
}
/* The page fault handler. */
void PageTable::handle_fault(REGS * _r) {
unsigned int err_code = _r->err_code & 1;
unsigned long fault_address = read_cr2();
switch(err_code) {
case KERNEL_READ_PAGE_NOT_PRESENT:
case KERNEL_WRITE_PAGE_NOT_PRESENT:
unsigned long page_directory_index = fault_address >> 22; //constify this
unsigned long page_table_index = (fault_address & 0x003FFFFF) >> 12; //constify this
unsigned long *page_directory = current_page_table->get_page_directory();
if(page_directory[page_directory_index] & 1) { //constify this
//page table entry is present in the directory
unsigned long *page_table = (unsigned long *)(page_directory[page_directory_index] & (0xFFFFF000));
if(page_table[page_table_index] & 1) {
//page entry is present in the page table
//why did the page fault occur? :P
}
else {
//constify the 3 - PT entry control bits
unsigned long *requested_page_from_framepool;
if(fault_address >= shared_size) {
requested_page_from_framepool =
(unsigned long *)(process_mem_pool->get_frame() * PAGE_SIZE);
} else {
requested_page_from_framepool =
(unsigned long *)(kernel_mem_pool->get_frame() * PAGE_SIZE);
}
create_page_table_entry(requested_page_from_framepool,
page_table,
page_table_index,
3);
//handle the fault by reading/writing to that location or do what is required
}
} else {
//create page table
unsigned long *requested_page_from_framepool =
(unsigned long *)(kernel_mem_pool->get_frame() * PAGE_SIZE);
create_page_table_entry(requested_page_from_framepool,
page_directory,
page_directory_index,
3);
unsigned long *page_table = (unsigned long *)(page_directory[page_directory_index] & (0xFFFFF000));
if(fault_address >= shared_size) {
requested_page_from_framepool =
(unsigned long *)(process_mem_pool->get_frame() * PAGE_SIZE);
} else {
requested_page_from_framepool =
(unsigned long *)(kernel_mem_pool->get_frame() * PAGE_SIZE);
}
create_page_table_entry(requested_page_from_framepool,
page_table,
page_table_index,
3);
//handle the fault by reading/writing to that location or do what is required
}
break;
}
/*
switch(err_code) {
case KERNEL_READ_PAGE_NOT_PRESENT:
break;
case KERNEL_READ_PROTECTION_FAULT:
break;
case KERNEL_WRITE_PAGE_NOT_PRESENT:
break;
case KERNEL_WRITE_PROTECTION_FAULT:
break;
case USER_READ_PAGE_NOT_PRESENT:
break;
case USER_READ_PROTECTION_FAULT:
break;
case USER_WRITE_PAGE_NOT_PRESENT:
break;
case USER_WRITE_PROTECTION_FAULT:
break;
}*/
}
/*-------------------------------------------------------------------------
* pfint - paging fault ISR
*-------------------------------------------------------------------------
*/
SYSCALL pfint()
{
STATWORD ps;
disable(ps);
unsigned long int eip = read_cr2();
unsigned long int pd_offset = eip >> 22;
unsigned long int pt_offset = eip >>12;
pt_offset = pt_offset & 1023;
unsigned long offset = eip;
offset = offset & 4095;
eip = eip >> 12; //vpno
int i = 0, flag = 0;
int backing_store_page, bs_id;
for(i = 0 ; i < NBS; i++)
{
if(proctab[currpid].map[i].bs_pid == currpid)
{
if(proctab[currpid].map[i].next == NULL)
{
if(proctab[currpid].map[i].bs_vpno == eip) //we found the exact page
{
backing_store_page = proctab[currpid].map[i].starting_page;
flag = 1;
}
else if(proctab[currpid].map[i].bs_vpno < eip && (proctab[currpid].map[i].bs_vpno+proctab[currpid].map[i].bs_npages) >= eip)
{
backing_store_page = proctab[currpid].map[i].starting_page + eip - proctab[currpid].map[i].bs_vpno;
flag = 1;
}
}
else
{
bs_map_t *move = &(proctab[currpid].map[i]);
while(move != NULL)
{
if(move->bs_vpno == eip) //we found the exact page
{
backing_store_page = move->starting_page;
flag = 1;
break;
}
else if(move->bs_vpno < eip && (move->bs_npages+move->bs_vpno) >= eip)
{
backing_store_page = move->starting_page + eip - move->bs_vpno;
flag = 1;
break;
}
move = move->next;
}
}
}
if(flag==1)
{
bs_id = i;
break;
}
}
unsigned long *bs_addr = (unsigned long *)(backing_store_page*NBPG);
//populate page table
//checking if page dir is empty
pd_t *ptr1 = (pd_t *)(proctab[currpid].pdbr);
ptr1 += pd_offset;
pt_t *ptr;
if(ptr1->pd_pres == 1) //page table exists hence add entry to that
{
ptr = (pt_t *)((ptr1->pd_base)*NBPG);
frm_tab[(ptr1->pd_base)-FRAME0].refcnt++;
frm_map[(ptr1->pd_base)-FRAME0].fr_refcnt++;
}
else //we need to create a page table, add our free_frame entry to it and add the page table entry to the directory
{
//kprintf("\nin else %d\n",ptr1);
int pt_frame = get_frm();
frm_tab[pt_frame-FRAME0].status = FRM_PGT;
frm_tab[pt_frame-FRAME0].refcnt++;
frm_map[pt_frame-FRAME0].fr_status = FRM_MAPPED;
frm_map[pt_frame-FRAME0].fr_pid = currpid;
frm_map[pt_frame-FRAME0].fr_refcnt++;
frm_map[pt_frame-FRAME0].fr_type = FR_TBL;
ptr = (pt_t*)(pt_frame*NBPG);
//add the above table to page directory
ptr1->pd_pres = 1;
ptr1->pd_write = 1;
ptr1->pd_user = 0;
ptr1->pd_pwt = 0;
ptr1->pd_pcd = 0;
ptr1->pd_acc = 0;
ptr1->pd_mbz = 0;
ptr1->pd_fmb = 0;
ptr1->pd_global = 0;
ptr1->pd_avail = 0;
ptr1->pd_base = pt_frame;
}
//add entry to page table
ptr += pt_offset;
ptr->pt_pres = 1;
ptr->pt_write = 1;
ptr->pt_user = 0;
ptr->pt_pwt = 0;
ptr->pt_pcd = 0;
ptr->pt_acc = 0;
ptr->pt_dirty = 0;
ptr->pt_mbz = 0;
ptr->pt_global = 0;
ptr->pt_avail = 0;
//getting a free frame an setting the frame mappings
int free_frame;
if(page_table[backing_store_page-total_bs_available] != -1)
{
free_frame = page_table[backing_store_page-total_bs_available];
frm_map[free_frame-FRAME0].fr_refcnt++;
frm_tab[free_frame-FRAME0].refcnt++;
fr_map_t *map = (fr_map_t *)getmem(sizeof(fr_map_t));
map->fr_pid = currpid;
map->fr_vpno = eip;
map->shared = NULL;
fr_map_t *next = (fr_map_t *)getmem(sizeof(fr_map_t));
next = &(frm_map[free_frame-FRAME0]);
while(next->shared != NULL)
next = next->shared;
next->shared = map;
}
else
{
//obtain free frame
free_frame = get_frm();
frm_tab[free_frame-FRAME0].status = FRM_BS;
frm_tab[free_frame-FRAME0].refcnt++;
frm_tab[free_frame-FRAME0].bs = bs_id;
frm_tab[free_frame-FRAME0].bs_page = backing_store_page;
frm_tab[free_frame-FRAME0].bs_next = NULL;
frm_tab[free_frame-FRAME0].fifo = NULL;
frm_tab[free_frame-FRAME0].age = 128;
frm_map[free_frame-FRAME0].fr_status = FRM_MAPPED;
frm_map[free_frame-FRAME0].fr_pid = currpid;
frm_map[free_frame-FRAME0].fr_vpno = eip;
frm_map[free_frame-FRAME0].fr_refcnt++;
frm_map[free_frame-FRAME0].fr_type = FR_PAGE;
frm_map[free_frame-FRAME0].bs_page_num = backing_store_page;
page_table[backing_store_page-total_bs_available] = free_frame;
//set bs mappings
if(bs_tab[bs_id].frm == NULL)
bs_tab[bs_id].frm = &frm_tab[free_frame-FRAME0];
else
{
frame_t *move = (frame_t *)getmem(sizeof(frame_t));
move = bs_tab[bs_id].frm;
while(move->bs_next != NULL)
{
move = move->bs_next;
}
move->bs_next = &frm_tab[free_frame-FRAME0];
}
//adding this frame to the fifo queue
if(fifo_head == NULL)
{
//queue is empty
fifo_head = &frm_tab[free_frame-FRAME0];
fifo_tail = fifo_head;
}
else
{
fifo_tail->fifo = &frm_tab[free_frame-FRAME0];
fifo_tail = &frm_tab[free_frame-FRAME0];
}
}
unsigned long *dest_addr = (unsigned long *)(free_frame*NBPG);
//copy page from bs to phy
for(i = 0; i < NBPG/sizeof(unsigned long); i++)
{
*dest_addr = *bs_addr;
dest_addr++;
bs_addr++;
}
ptr->pt_base = free_frame;
restore(ps);
return OK;
}
static void dump_regs(struct irq_regs *regs)
{
unsigned long cs, eip, eflags;
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
unsigned long d0, d1, d2, d3, d6, d7;
unsigned long sp;
/*
* Some exceptions cause an error code to be saved on the current stack
* after the EIP value. We should extract CS/EIP/EFLAGS from different
* position on the stack based on the exception number.
*/
switch (regs->irq_id) {
case EXC_DF:
case EXC_TS:
case EXC_NP:
case EXC_SS:
case EXC_GP:
case EXC_PF:
case EXC_AC:
cs = regs->context.ctx2.xcs;
eip = regs->context.ctx2.eip;
eflags = regs->context.ctx2.eflags;
/* We should fix up the ESP due to error code */
regs->esp += 4;
break;
default:
cs = regs->context.ctx1.xcs;
eip = regs->context.ctx1.eip;
eflags = regs->context.ctx1.eflags;
break;
}
printf("EIP: %04x:[<%08lx>] EFLAGS: %08lx\n",
(u16)cs, eip, eflags);
if (gd->flags & GD_FLG_RELOC)
printf("Original EIP :[<%08lx>]\n", eip - gd->reloc_off);
printf("EAX: %08lx EBX: %08lx ECX: %08lx EDX: %08lx\n",
regs->eax, regs->ebx, regs->ecx, regs->edx);
printf("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
regs->esi, regs->edi, regs->ebp, regs->esp);
printf(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
(u16)regs->xds, (u16)regs->xes, (u16)regs->xfs,
(u16)regs->xgs, (u16)regs->xss);
cr0 = read_cr0();
cr2 = read_cr2();
cr3 = read_cr3();
cr4 = read_cr4();
printf("CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
cr0, cr2, cr3, cr4);
d0 = get_debugreg(0);
d1 = get_debugreg(1);
d2 = get_debugreg(2);
d3 = get_debugreg(3);
printf("DR0: %08lx DR1: %08lx DR2: %08lx DR3: %08lx\n",
d0, d1, d2, d3);
d6 = get_debugreg(6);
d7 = get_debugreg(7);
printf("DR6: %08lx DR7: %08lx\n",
d6, d7);
printf("Stack:\n");
sp = regs->esp;
sp += 64;
while (sp > (regs->esp - 16)) {
if (sp == regs->esp)
printf("--->");
else
printf(" ");
printf("0x%8.8lx : 0x%8.8lx\n", sp, (ulong)readl(sp));
sp -= 4;
}
}
/*******************************************************************************
* Name: pf_handler
*
* Desc: High level page fault handling code. The low level page fault
* interrupt handler is written in assembly. It'll setup the required
* registers (faulted address in CR2) and the stack frame with the
* error code. This routine is responsible for the following:
* 1. Read the faulted address and lookup BS to find if this address is
* mapped. If so, get the BS id and the offset within the BS. If
* not, this is an illegal access - kill the process and move on.
* 2. Actual paging starts here. Lookup the proctab to find the pid's
* pdir base. From the faulted address, we can get the pdir offset.
* Using these two, check if a page table exist for the faulted
* address. If not create one.
* 3. If the page table is presnet, get the ptbl offset from the
* faulted vaddr. This points to the location of the page table
* entry. Now, check if the frame assoicated with this page is
* already present in the memory (shared pages). If so, update the
* page table's 'pres' bit to reflect this and increment the
* frame's refcount. If not, allocate a new frame and update the
* page table entry to reflect that the frame is present.
* 4. Processor caches the paging entries (TLB) maintained by software
* and uses them whenever possible. When a pgt entry's 'pres' bit
* is cleared, we need to flush the entry from the processor
* cache so that the proceessor would use the updated software
* data. This is described in detail in section 4.8 of Intel IA32
* software developer manual (vol 3). There are many ways to force
* the processor to use the software tables, than hardware cache.
* One such way is to reload teh CR0 register. So, if any of the
* 'pres' bits are modified, we reload the CR0 register.
*
* Params: None.
*
* Returns: SYSCALL
* OK - on success
* SYSERR - on error
******************************************************************************/
SYSCALL
pf_handler(void)
{
int bs_id = EMPTY;
int bs_offset = EMPTY;
int fr_id = EMPTY;
uint32_t pdir_offset = 0;
uint32_t ptbl_offset = 0;
uint32_t page_offset = 0;
uint32_t fault_addr = 0;
pd_t *pdir = NULL;
pt_t *ptbl = NULL;
frm_map_t *frptr = NULL;
frm_map_t *pt_frptr = NULL;
virt_addr_t *fault_vaddr = NULL;
STATWORD ps;
disable(ps);
DTRACE_START;
DTRACE("DBG$ %d %s> inside high-level page fault handler\n", \
currpid, __func__);
/* vaddr : pdir_offset : ptbl_offset : page_offset
* 32 bits : 10 bits : 10 bits : 12 bits
*/
fault_addr = read_cr2();
/* The virtual address is 32-bits. So, we directly read the required set of
* bits by assigining it to a strcutre with appropriate bit-fields.
*/
fault_vaddr = (virt_addr_t *) (&fault_addr);
pdir_offset = (uint32_t) fault_vaddr->pd_offset;
ptbl_offset = (uint32_t) fault_vaddr->pt_offset;
page_offset = (uint32_t) fault_vaddr->pg_offset;
DTRACE("DBG$ %d %s> faulted vaddr 0x%08x, vpage %d\n", \
currpid, __func__, fault_addr, VADDR_TO_VPAGE(fault_addr));
DTRACE("DBG$ %d %s> pd %d, pt %d, pg %d\n", \
currpid, __func__, pdir_offset, ptbl_offset, page_offset);
/* Check the BS for a mapping for the faulted vaddr and the pid. If present,
* record the BS id and the offset within the BS. If not present, it's
* illeagal memory access. Kill the process and return.
*/
if (SYSERR == bsm_lookup(currpid, fault_addr, &bs_id, &bs_offset,
NULL, NULL)) {
DTRACE("DBG$ %d %s> bsm_lookup() failed\n", currpid, __func__);
DTRACE("DBG$ %d %s> pid %d will be killed\n", \
currpid, __func__, currpid);
kprintf("\n\n");
kprintf("FATAL ERROR: Process '%s' with pid '%d' is trying to access " \
"virtual memory out of its range! \nThe process will be " \
"terminated.\n", P_GET_PNAME(currpid), currpid);
kprintf("\n\n");
sleep(9);
DTRACE_END;
restore(ps);
kill(currpid);
goto RESTORE_AND_RETURN_ERROR;
}
/* Get the currpid's page directory and index to the appropriate pgt. If pgt
* isn't present, create one.
*/
pdir = P_GET_PDIR(currpid);
if (FALSE == PD_GET_PRES(pdir, pdir_offset)) {
DTRACE("DBG$ %d %s> pgt not present for pid %d, pd offset %d, " \
"pt offset %d, pg offset %d, vaddr 0x%08x\n", currpid, \
__func__, currpid, pdir_offset, ptbl_offset, page_offset, \
fault_addr);
ptbl = new_pgt();
if (NULL == ptbl) {
DTRACE("DBG$ %d %s> new_pgt() failed\n", currpid, __func__);
goto RESTORE_AND_RETURN_ERROR;
}
/* Fill-in few meta-data for the pgt frame just created. */
pt_frptr = FR_GET_FPTR(FR_PA_TO_ID(ptbl));
pt_frptr->fr_pid = currpid;
/* Set the 'pres' and 'write' bits alone. Rest would've been zeroed
* out by new_pgt(). Also, set the base of the new page table.
*/
pdir[pdir_offset].pd_pres = 1;
pdir[pdir_offset].pd_write = 1;
pdir[pdir_offset].pd_base = VADDR_TO_VPAGE((unsigned) ptbl);
} else {
DTRACE("DBG$ %d %s> ptbl already present at 0x%08x, fr id %d\n", \
currpid, __func__, VPAGE_TO_VADDR(PD_GET_BASE(pdir, pdir_offset)),\
FR_PA_TO_ID(VPAGE_TO_VADDR(PD_GET_BASE(pdir, pdir_offset))));
}
ptbl = (pt_t *) VPAGE_TO_VADDR(PD_GET_BASE(pdir, pdir_offset));
DTRACE("DBG$ %d %s> ptbl present at 0x%08x, fr id %d\n", \
currpid, __func__, ptbl, FR_PA_TO_ID(ptbl));
/* Find if a frame representing the same BS id and offset is present in the
* memory (shared pages). If so, just update the page table entry and
* increment teh refcount.
*/
if (EMPTY == (fr_id = is_frm_present(bs_id, bs_offset))) {
DTRACE("DBG$ %d %s> frame not present.. creating a new frame\n", \
currpid, __func__);
frptr = get_frm(FR_PAGE);
if (NULL == frptr) {
DTRACE("DBG$ %d %s> get_frm() failed\n", currpid, __func__);
goto RESTORE_AND_RETURN_ERROR;
}
fr_id = frptr->fr_id;
frm_pidmap_oper(fr_id, getpid(), FR_OP_PMAP_SET);
frm_record_details(fr_id, getpid(), VADDR_TO_VPAGE(fault_addr));
/* Read the appropriate page from BS onto the new frame. */
if (SYSERR == read_bs((char *) FR_ID_TO_PA(fr_id), bs_id, bs_offset)) {
DTRACE("DBG$ %d %s> read_bs() failed for fr id %d, bs %d, " \
"offset %d\n", currpid, __func__, fr_id, bs_id, bs_offset);
goto RESTORE_AND_RETURN_ERROR;
} else {
DTRACE("DBG$ %d %s> reading for fr id %d, bs %d, offset %d\n", \
currpid, __func__, fr_id, bs_id, bs_offset);
}
/* Fill-in the new BS details in the frame. */
frptr->fr_type = FR_PAGE;
frptr->fr_bs = bs_id;
frptr->fr_bsoffset = bs_offset;
inc_frm_refcount(fr_id);
#ifdef DBG_ON
print_frm_id(fr_id);
#endif
} else {
/* A frame representing the same BS and offset is already present in the
* memory. So, just increment the refcount of the frame.
*/
frm_pidmap_oper(fr_id, getpid(), FR_OP_PMAP_SET);
frm_record_details(fr_id, getpid(), VADDR_TO_VPAGE(fault_addr));
inc_frm_refcount(fr_id);
}
/* In both cases (frame present and frame not present), we need to update
* the page table entry as at this point, the frame is loaded onto the
* memory. Do the following w.r.t. the pgt frame:
* 1. Set the 'pres' bit in the page table entry corresponding to the
* newly loaded page to reflect that the page is present in the
* memory.
* 2. Set the 'write' bit (as given in PA3 description).
* 3. Update the 'base' of the page entry corresponding to the newly
* created page to point to the frame.
* 4. Increment the refcount of the pgt frame. Unlike data frames, where
* refocunt denotes the # of processes that map to the actual
* physical frame, pgt frame's refcount reflects the # of pages
* (that are part of this pgt) that are present in the memory.
* This will be decremented when a frame is paged out and the
* pgt frame will be freed when the refcount reaches zero.
*/
ptbl[ptbl_offset].pt_pres = 1;
ptbl[ptbl_offset].pt_write = 1;
ptbl[ptbl_offset].pt_base = FR_ID_TO_VPAGE(fr_id);
inc_frm_refcount(FR_PA_TO_ID(ptbl));
/* Reload the CR0 register would force the processor to flush the tables
* that the processor maintains in hardware cache and to use the updated
* software tables.
*/
enable_paging();
DTRACE("DBG$ %d %s> returning OK\n", currpid, __func__);
DTRACE_END;
restore(ps);
return OK;
RESTORE_AND_RETURN_ERROR:
DTRACE("DBG$ %d %s> returning SYSERR\n", currpid, __func__);
DTRACE_END;
restore(ps);
return SYSERR;
}
/*-------------------------------------------------------------------------
* pfint - paging fault ISR
*-------------------------------------------------------------------------
*/
SYSCALL pfint()
{
STATWORD ps;
disable(ps);
unsigned long int eip = read_cr2();
//kprintf("\n\n#PF in %s, cr2:%x",proctab[currpid].pname, eip);
unsigned long int pd_offset = eip >> 22;
unsigned long int pt_offset = eip >>12;
pt_offset = pt_offset & 1023;
unsigned long offset = eip;
offset = offset & 4095;
//kprintf("\nEIP read is %lu\n",eip);
eip = eip >> 12; //vpno
//kprintf("\nEIP shifted is %lu\n",eip);
int i = 0, flag = 0;
int backing_store_page, bs_id;
for(i = 0 ; i < NBS; i++)
{
if(proctab[currpid].map[i].pid == currpid)
{
//kprintf("\nif\n");
if(proctab[currpid].map[i].next == NULL)
{
//kprintf("\nif if\n");
if(proctab[currpid].map[i].vpno == eip) //we found the exact page
{
backing_store_page = proctab[currpid].map[i].base_page;
flag = 1;
//kprintf("\nif if if\n");
}
else if(proctab[currpid].map[i].vpno < eip && (proctab[currpid].map[i].vpno+proctab[currpid].map[i].npages) >= eip) //we found the page in the range of base_page - npages
{
backing_store_page = proctab[currpid].map[i].base_page + eip - proctab[currpid].map[i].vpno;
flag = 1;
//kprintf("\nif if else if\n");
}
}
else
{
//kprintf("\nelse\n");
bs_map_t *jump = &(proctab[currpid].map[i]);
while(jump != NULL)
{
//kprintf("\nwhile\n");
if(jump->vpno == eip) //we found the exact page
{
backing_store_page = jump->base_page;
flag = 1;
//kprintf("\nif else if\n");
break;
}
else if(jump->vpno < eip && (jump->npages+jump->vpno) >= eip) //we found the page in the range of base_page - npages
{
backing_store_page = jump->base_page + eip - jump->vpno;
flag = 1;
//kprintf("\nif else else if\n");
break;
}
jump = jump->next;
}
}
}
if(flag)
{
bs_id = i;
break;
}
}
//kprintf("\nin pfint bs %d, bs_page %d",bs_id,backing_store_page);
unsigned long *bs_addr = (unsigned long *)(backing_store_page*NBPG);
//populate page table
//checking if page dir is empty
pd_t *ptr1 = (pd_t *)(proctab[currpid].pdbr);
//kprintf("\nbefore %d, %lu\n",ptr1,pd_offset);
ptr1 += pd_offset;
pt_t *ptr;
if(ptr1->pd_pres == 1) //page table exists hence add entry to that
{
ptr = (pt_t *)((ptr1->pd_base)*NBPG);
frm_tab[(ptr1->pd_base)-FRAME0].refcnt++;
frm_map[(ptr1->pd_base)-FRAME0].fr_refcnt++;
}
else //we need to create a page table, add our free_frm entry to it and add the page table entry to the directory
{
//kprintf("\nin else %d\n",ptr1);
int pt_frame = get_frm();
frm_tab[pt_frame-FRAME0].status = FRM_PGT;
frm_tab[pt_frame-FRAME0].refcnt++;
frm_map[pt_frame-FRAME0].fr_status = FRM_MAPPED;
frm_map[pt_frame-FRAME0].fr_pid = currpid;
frm_map[pt_frame-FRAME0].fr_refcnt++;
frm_map[pt_frame-FRAME0].fr_type = FR_TBL;
ptr = (pt_t*)(pt_frame*NBPG);
//add the above table to page directory
ptr1->pd_pres = 1;
ptr1->pd_write = 1;
ptr1->pd_user = 0;
ptr1->pd_pwt = 0;
ptr1->pd_pcd = 0;
ptr1->pd_acc = 0;
ptr1->pd_mbz = 0;
ptr1->pd_fmb = 0;
ptr1->pd_global = 0;
ptr1->pd_avail = 0;
ptr1->pd_base = pt_frame;
//kprintf("\nget_frm return frame %d. To be page table for process %s",pt_frame, proctab[currpid].pname);
}
//add entry to page table
ptr += pt_offset;
ptr->pt_pres = 1;
ptr->pt_write = 1;
ptr->pt_user = 0;
ptr->pt_pwt = 0;
ptr->pt_pcd = 0;
ptr->pt_acc = 0;
ptr->pt_dirty = 0;
ptr->pt_mbz = 0;
ptr->pt_global = 0;
ptr->pt_avail = 0;
//getting a free frame an setting the frame mappings
int free_frm;
if(ni_page_table[backing_store_page-total_bs_left] != -1) // if backing store is already mapped then we share the frame
{
free_frm = ni_page_table[backing_store_page-total_bs_left];
//kprintf("\nIN NI frm = %d, ni = %d\n",free_frm, ni_page_table[backing_store_page-total_bs_left]);
frm_map[free_frm-FRAME0].fr_refcnt++;
frm_tab[free_frm-FRAME0].refcnt++;
//adding this vpno and pid to the shared frame map list
fr_map_t *map = (fr_map_t *)getmem(sizeof(fr_map_t));
map->fr_pid = currpid;
map->fr_vpno = eip;
map->shared = NULL;
fr_map_t *next = (fr_map_t *)getmem(sizeof(fr_map_t));
next = &(frm_map[free_frm-FRAME0]);
while(next->shared != NULL)
next = next->shared;
next->shared = map;
//printing the list
/*kprintf("\nSHARED VPNO: ");
next = &(frm_map[free_frm-FRAME0]);
while(next != NULL)
{
kprintf("%d -> ",next->fr_vpno);
next = next->shared;
}*/
}
else
{
free_frm = get_frm();
frm_tab[free_frm-FRAME0].status = FRM_BS;
frm_tab[free_frm-FRAME0].refcnt++;
frm_tab[free_frm-FRAME0].bs = bs_id;
frm_tab[free_frm-FRAME0].bs_page = backing_store_page;
frm_tab[free_frm-FRAME0].bs_next = NULL;
frm_tab[free_frm-FRAME0].fifo = NULL;
frm_tab[free_frm-FRAME0].age = 128;
frm_map[free_frm-FRAME0].fr_status = FRM_MAPPED;
frm_map[free_frm-FRAME0].fr_pid = currpid;
frm_map[free_frm-FRAME0].fr_vpno = eip;
frm_map[free_frm-FRAME0].fr_refcnt++;
frm_map[free_frm-FRAME0].fr_type = FR_PAGE;
frm_map[free_frm-FRAME0].bs_page_num = backing_store_page;
ni_page_table[backing_store_page-total_bs_left] = free_frm;
//set bs mappings
if(bs_tab[bs_id].frm == NULL)
bs_tab[bs_id].frm = &frm_tab[free_frm-FRAME0];
else
{
//frame_t *jump = &frm_tab[free_frm-FRAME0];
frame_t *jump = (frame_t *)getmem(sizeof(frame_t));
jump = bs_tab[bs_id].frm;
while(jump->bs_next != NULL)
{
jump = jump->bs_next;
//kprintf("\njumping\n");
}
jump->bs_next = &frm_tab[free_frm-FRAME0];
}
//adding this frame to the fifo queue
if(fifo_head == NULL)
{
//queue is empty
fifo_head = &frm_tab[free_frm-FRAME0];
fifo_tail = fifo_head;
}
else
{
fifo_tail->fifo = &frm_tab[free_frm-FRAME0];
fifo_tail = &frm_tab[free_frm-FRAME0];
}
}
//kprintf("\nget_frm return frame %d.", free_frm);
unsigned long *dst_addr = (unsigned long *)(free_frm*NBPG);
//kprintf("\n\n Virtual page %d mapped to bs page %d, bs id %d, mapped to frame %d, %lu",eip, backing_store_page, bs_id, free_frm,dst_addr);
/*frame_t *next = fifo_head;
kprintf("\nFIFO : ");
while(next != NULL)
{
kprintf("%d-> ",next->frame_num);
next = next->fifo;
}
kprintf("\n");*/
//copy page from bs to phy
for(i = 0; i < NBPG/sizeof(unsigned long); i++)
{
*dst_addr = *bs_addr;
dst_addr++;
bs_addr++;
}
ptr->pt_base = free_frm;
restore(ps);
//kprintf("\nmap bs%d/page: %d to frame %d",bs_id, (backing_store_page-proctab[currpid].map[bs_id].base_page), free_frm);
return OK;
}
/*-------------------------------------------------------------------------
* pfint - paging fault ISR
*-------------------------------------------------------------------------
*/
SYSCALL pfint() {
STATWORD ps;
disable(ps);
int store, pageth;
unsigned long vaddr = read_cr2();
unsigned int p_offset = (vaddr & 0xFFC00000) >> 22;
unsigned int q_offset = (vaddr & 0x3FF000) >> 12;
bsm_lookup(currpid, vaddr, &store, &pageth);
frame_t *frame_backing_store = obtain_frame();
bs_tab[store].pg_to_frm_map[pageth] = frame_backing_store->frm_num;
frame_backing_store->bs = store;
frame_backing_store->bs_page = pageth;
frame_backing_store->status = FRM_BS;
frame_backing_store->fr_type = FR_PAGE;
read_bs((char *)(frame_backing_store->frm_num * 4096), store, pageth );
frame_t * obtained_free_frame = frame_backing_store;
// unsigned long pdbr = add_pg_dir_entry_for_pg_fault(currpid, p_offset, q_offset, obtained_free_frame);
unsigned long page_to_frame;
int avail = 0;
struct pentry *pptr = &proctab[currpid];
frame_t *directory = pptr->pd;
pd_t *table_descriptor = (pd_t *) ((4096 * directory->frm_num) + p_offset * sizeof(pd_t));
if (table_descriptor->pd_pres == 1)
{
frame_t *page_frame;
int i = 0;
for (i = 0; i < NFRAMES; i++) {
if (frm_tab[i].frm_num == table_descriptor->pd_base)
{
page_frame= &frm_tab[i];
}
}
pt_t *offset_value = (pt_t *) (4096 * page_frame->frm_num );
offset_value =offset_value+q_offset;
pt_t ptr;
ptr.pt_pres = 1;
ptr.pt_write = 1;
ptr.pt_user = 0;
ptr.pt_pwt = 0;
ptr.pt_pcd = 0;
ptr.pt_acc = 0;
ptr.pt_dirty = 0;
ptr.pt_mbz = 0;
ptr.pt_global = 0;
ptr.pt_avail = 0;
ptr.pt_base = obtained_free_frame->frm_num;
*offset_value = ptr;
}
else
{
frame_t *page_frame = obtain_frame();
page_frame->status = FRM_PGT;
page_frame->fr_type = FRM_PGT;
page_frame->fr_pid = currpid;
pd_t *table_descriptor = (pd_t *) (4096 * directory->frm_num);
table_descriptor =table_descriptor+p_offset;
pd_t ptr1;
ptr1.pd_pres = 1;
ptr1.pd_write = 1;
ptr1.pd_user = 0;
ptr1.pd_pwt = 0;
ptr1.pd_pcd = 0;
ptr1.pd_acc = 0;
ptr1.pd_mbz = 0;
ptr1.pd_fmb = 0;
ptr1.pd_global = 0;
ptr1.pd_avail = 0;
ptr1.pd_base = page_frame->frm_num;
*table_descriptor = ptr1;
pt_t *offset_value = (pt_t *) (4096 * page_frame->frm_num );
offset_value =offset_value+q_offset;
pt_t ptr;
ptr.pt_pres = 1;
ptr.pt_write = 1;
ptr.pt_user = 0;
ptr.pt_pwt = 0;
ptr.pt_pcd = 0;
ptr.pt_acc = 0;
ptr.pt_dirty = 0;
ptr.pt_mbz = 0;
ptr.pt_global = 0;
ptr.pt_avail = 0;
ptr.pt_base = obtained_free_frame->frm_num;
*offset_value = ptr;
page_to_frame = page_frame->frm_num ;
}
unsigned long pdbr= pptr->pdbr;
struct pentry *pptr1 = &proctab[currpid];
bs_map_t *map = &(pptr1->map[store]);
if(map->frm == NULL)
map->frm = obtained_free_frame;
else{
frame_t *tmp = map->frm;
while(tmp->bs_next != NULL)
tmp = tmp->bs_next;
tmp->bs_next = obtained_free_frame;
}
obtained_free_frame->fr_vpno = map->vpno + obtained_free_frame->bs_page;
obtained_free_frame->fr_pid = currpid;
write_cr3(pdbr * NBPG);
restore(ps);
return OK;
}
