void BX_CPU_C::task_switch(bx_selector_t *tss_selector,
bx_descriptor_t *tss_descriptor, unsigned source,
Bit32u dword1, Bit32u dword2)
{
Bit32u obase32; // base address of old TSS
Bit32u nbase32; // base address of new TSS
Bit32u temp32, newCR3;
Bit16u raw_cs_selector, raw_ss_selector, raw_ds_selector, raw_es_selector,
raw_fs_selector, raw_gs_selector, raw_ldt_selector;
Bit16u temp16, trap_word;
bx_selector_t cs_selector, ss_selector, ds_selector, es_selector,
fs_selector, gs_selector, ldt_selector;
bx_descriptor_t cs_descriptor, ss_descriptor, ldt_descriptor;
Bit32u old_TSS_max, new_TSS_max, old_TSS_limit, new_TSS_limit;
Bit32u newEAX, newECX, newEDX, newEBX;
Bit32u newESP, newEBP, newESI, newEDI;
Bit32u newEFLAGS, newEIP;
BX_DEBUG(("TASKING: ENTER"));
invalidate_prefetch_q();
// Discard any traps and inhibits for new context; traps will
// resume upon return.
BX_CPU_THIS_PTR debug_trap = 0;
BX_CPU_THIS_PTR inhibit_mask = 0;
// STEP 1: The following checks are made before calling task_switch(),
// for JMP & CALL only. These checks are NOT made for exceptions,
// interrupts & IRET.
//
// 1) TSS DPL must be >= CPL
// 2) TSS DPL must be >= TSS selector RPL
// 3) TSS descriptor is not busy.
// TSS must be present, else #NP(TSS selector)
if (tss_descriptor->p==0) {
BX_ERROR(("task_switch: TSS descriptor is not present !"));
exception(BX_NP_EXCEPTION, tss_selector->value & 0xfffc, 0);
}
// STEP 2: The processor performs limit-checking on the target TSS
// to verify that the TSS limit is greater than or equal
// to 67h (2Bh for 16-bit TSS).
// Gather info about old TSS
if (BX_CPU_THIS_PTR tr.cache.type <= 3) {
old_TSS_max = 0x29;
}
else {
old_TSS_max = 0x5F;
}
// Gather info about new TSS
if (tss_descriptor->type <= 3) { // {1,3}
new_TSS_max = 0x2B;
}
else { // tss_descriptor->type = {9,11}
new_TSS_max = 0x67;
}
obase32 = (Bit32u) BX_CPU_THIS_PTR tr.cache.u.system.base; // old TSS.base
old_TSS_limit = BX_CPU_THIS_PTR tr.cache.u.system.limit_scaled;
nbase32 = (Bit32u) tss_descriptor->u.system.base; // new TSS.base
new_TSS_limit = tss_descriptor->u.system.limit_scaled;
// TSS must have valid limit, else #TS(TSS selector)
if (tss_selector->ti || tss_descriptor->valid==0 ||
new_TSS_limit < new_TSS_max)
{
BX_ERROR(("task_switch(): new TSS limit < %d", new_TSS_max));
exception(BX_TS_EXCEPTION, tss_selector->value & 0xfffc, 0);
}
if (old_TSS_limit < old_TSS_max) {
BX_ERROR(("task_switch(): old TSS limit < %d", old_TSS_max));
exception(BX_TS_EXCEPTION, BX_CPU_THIS_PTR tr.selector.value & 0xfffc, 0);
}
if (obase32 == nbase32) {
BX_INFO(("TASK SWITCH: switching to the same TSS !"));
}
// Check that old TSS, new TSS, and all segment descriptors
// used in the task switch are paged in.
if (BX_CPU_THIS_PTR cr0.get_PG())
{
dtranslate_linear(obase32, 0, BX_WRITE); // new TSS
dtranslate_linear(obase32 + old_TSS_max, 0, BX_WRITE);
dtranslate_linear(nbase32, 0, BX_READ); // old TSS
dtranslate_linear(nbase32 + new_TSS_max, 0, BX_READ);
// ??? Humm, we check the new TSS region with READ above,
// but sometimes we need to write the link field in that
// region. We also sometimes update other fields, perhaps
// we need to WRITE check them here also, so that we keep
// the written state consistent (ie, we don't encounter a
// page fault in the middle).
if (source == BX_TASK_FROM_CALL_OR_INT)
{
dtranslate_linear(nbase32, 0, BX_WRITE);
dtranslate_linear(nbase32 + 2, 0, BX_WRITE);
}
}
// Privilege and busy checks done in CALL, JUMP, INT, IRET
// STEP 3: Save the current task state in the TSS. Up to this point,
// any exception that occurs aborts the task switch without
// changing the processor state.
/* save current machine state in old task's TSS */
Bit32u oldEFLAGS = read_eflags();
/* if moving to busy task, clear NT bit */
if (tss_descriptor->type == BX_SYS_SEGMENT_BUSY_286_TSS ||
tss_descriptor->type == BX_SYS_SEGMENT_BUSY_386_TSS)
{
oldEFLAGS &= ~EFlagsNTMask;
}
if (BX_CPU_THIS_PTR tr.cache.type <= 3) {
temp16 = IP; access_write_linear(Bit32u(obase32 + 14), 2, 0, &temp16);
temp16 = oldEFLAGS; access_write_linear(Bit32u(obase32 + 16), 2, 0, &temp16);
temp16 = AX; access_write_linear(Bit32u(obase32 + 18), 2, 0, &temp16);
temp16 = CX; access_write_linear(Bit32u(obase32 + 20), 2, 0, &temp16);
temp16 = DX; access_write_linear(Bit32u(obase32 + 22), 2, 0, &temp16);
temp16 = BX; access_write_linear(Bit32u(obase32 + 24), 2, 0, &temp16);
temp16 = SP; access_write_linear(Bit32u(obase32 + 26), 2, 0, &temp16);
temp16 = BP; access_write_linear(Bit32u(obase32 + 28), 2, 0, &temp16);
temp16 = SI; access_write_linear(Bit32u(obase32 + 30), 2, 0, &temp16);
temp16 = DI; access_write_linear(Bit32u(obase32 + 32), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_ES].selector.value;
access_write_linear(Bit32u(obase32 + 34), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_CS].selector.value;
access_write_linear(Bit32u(obase32 + 36), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_SS].selector.value;
access_write_linear(Bit32u(obase32 + 38), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_DS].selector.value;
access_write_linear(Bit32u(obase32 + 40), 2, 0, &temp16);
}
else {
temp32 = EIP; access_write_linear(Bit32u(obase32 + 0x20), 4, 0, &temp32);
temp32 = oldEFLAGS; access_write_linear(Bit32u(obase32 + 0x24), 4, 0, &temp32);
temp32 = EAX; access_write_linear(Bit32u(obase32 + 0x28), 4, 0, &temp32);
temp32 = ECX; access_write_linear(Bit32u(obase32 + 0x2c), 4, 0, &temp32);
temp32 = EDX; access_write_linear(Bit32u(obase32 + 0x30), 4, 0, &temp32);
temp32 = EBX; access_write_linear(Bit32u(obase32 + 0x34), 4, 0, &temp32);
temp32 = ESP; access_write_linear(Bit32u(obase32 + 0x38), 4, 0, &temp32);
temp32 = EBP; access_write_linear(Bit32u(obase32 + 0x3c), 4, 0, &temp32);
temp32 = ESI; access_write_linear(Bit32u(obase32 + 0x40), 4, 0, &temp32);
temp32 = EDI; access_write_linear(Bit32u(obase32 + 0x44), 4, 0, &temp32);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_ES].selector.value;
access_write_linear(Bit32u(obase32 + 0x48), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_CS].selector.value;
access_write_linear(Bit32u(obase32 + 0x4c), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_SS].selector.value;
access_write_linear(Bit32u(obase32 + 0x50), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_DS].selector.value;
access_write_linear(Bit32u(obase32 + 0x54), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_FS].selector.value;
access_write_linear(Bit32u(obase32 + 0x58), 2, 0, &temp16);
temp16 = BX_CPU_THIS_PTR sregs[BX_SEG_REG_GS].selector.value;
access_write_linear(Bit32u(obase32 + 0x5c), 2, 0, &temp16);
}
// effect on link field of new task
if (source == BX_TASK_FROM_CALL_OR_INT)
{
// set to selector of old task's TSS
temp16 = BX_CPU_THIS_PTR tr.selector.value;
access_write_linear(nbase32, 2, 0, &temp16);
}
// STEP 4: The new-task state is loaded from the TSS
if (tss_descriptor->type <= 3) {
access_read_linear(Bit32u(nbase32 + 14), 2, 0, BX_READ, &temp16);
newEIP = temp16; // zero out upper word
access_read_linear(Bit32u(nbase32 + 16), 2, 0, BX_READ, &temp16);
newEFLAGS = temp16;
// incoming TSS is 16bit:
// - upper word of general registers is set to 0xFFFF
// - upper word of eflags is zero'd
// - FS, GS are zero'd
// - upper word of eIP is zero'd
access_read_linear(Bit32u(nbase32 + 18), 2, 0, BX_READ, &temp16);
newEAX = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 20), 2, 0, BX_READ, &temp16);
newECX = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 22), 2, 0, BX_READ, &temp16);
newEDX = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 24), 2, 0, BX_READ, &temp16);
newEBX = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 26), 2, 0, BX_READ, &temp16);
newESP = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 28), 2, 0, BX_READ, &temp16);
newEBP = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 30), 2, 0, BX_READ, &temp16);
newESI = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 32), 2, 0, BX_READ, &temp16);
newEDI = 0xffff0000 | temp16;
access_read_linear(Bit32u(nbase32 + 34), 2, 0, BX_READ, &raw_es_selector);
access_read_linear(Bit32u(nbase32 + 36), 2, 0, BX_READ, &raw_cs_selector);
access_read_linear(Bit32u(nbase32 + 38), 2, 0, BX_READ, &raw_ss_selector);
access_read_linear(Bit32u(nbase32 + 40), 2, 0, BX_READ, &raw_ds_selector);
access_read_linear(Bit32u(nbase32 + 42), 2, 0, BX_READ, &raw_ldt_selector);
raw_fs_selector = 0; // use a NULL selector
raw_gs_selector = 0; // use a NULL selector
// No CR3 change for 286 task switch
newCR3 = 0; // keep compiler happy (not used)
trap_word = 0; // keep compiler happy (not used)
}
else {
if (BX_CPU_THIS_PTR cr0.get_PG())
access_read_linear(Bit32u(nbase32 + 0x1c), 4, 0, BX_READ, &newCR3);
else
newCR3 = 0; // keep compiler happy (not used)
access_read_linear(Bit32u(nbase32 + 0x20), 4, 0, BX_READ, &newEIP);
access_read_linear(Bit32u(nbase32 + 0x24), 4, 0, BX_READ, &newEFLAGS);
access_read_linear(Bit32u(nbase32 + 0x28), 4, 0, BX_READ, &newEAX);
access_read_linear(Bit32u(nbase32 + 0x2c), 4, 0, BX_READ, &newECX);
access_read_linear(Bit32u(nbase32 + 0x30), 4, 0, BX_READ, &newEDX);
access_read_linear(Bit32u(nbase32 + 0x34), 4, 0, BX_READ, &newEBX);
access_read_linear(Bit32u(nbase32 + 0x38), 4, 0, BX_READ, &newESP);
access_read_linear(Bit32u(nbase32 + 0x3c), 4, 0, BX_READ, &newEBP);
access_read_linear(Bit32u(nbase32 + 0x40), 4, 0, BX_READ, &newESI);
access_read_linear(Bit32u(nbase32 + 0x44), 4, 0, BX_READ, &newEDI);
access_read_linear(Bit32u(nbase32 + 0x48), 2, 0, BX_READ, &raw_es_selector);
access_read_linear(Bit32u(nbase32 + 0x4c), 2, 0, BX_READ, &raw_cs_selector);
access_read_linear(Bit32u(nbase32 + 0x50), 2, 0, BX_READ, &raw_ss_selector);
access_read_linear(Bit32u(nbase32 + 0x54), 2, 0, BX_READ, &raw_ds_selector);
access_read_linear(Bit32u(nbase32 + 0x58), 2, 0, BX_READ, &raw_fs_selector);
access_read_linear(Bit32u(nbase32 + 0x5c), 2, 0, BX_READ, &raw_gs_selector);
access_read_linear(Bit32u(nbase32 + 0x60), 2, 0, BX_READ, &raw_ldt_selector);
access_read_linear(Bit32u(nbase32 + 0x64), 2, 0, BX_READ, &trap_word);
}
// Step 5: If CALL, interrupt, or JMP, set busy flag in new task's
// TSS descriptor. If IRET, leave set.
if (source == BX_TASK_FROM_JUMP || source == BX_TASK_FROM_CALL_OR_INT)
{
// set the new task's busy bit
Bit32u laddr = (Bit32u)(BX_CPU_THIS_PTR gdtr.base) + (tss_selector->index<<3) + 4;
access_read_linear(laddr, 4, 0, BX_READ, &dword2);
dword2 |= 0x200;
access_write_linear(laddr, 4, 0, &dword2);
}
// Step 6: If JMP or IRET, clear busy bit in old task TSS descriptor,
// otherwise leave set.
// effect on Busy bit of old task
if (source == BX_TASK_FROM_JUMP || source == BX_TASK_FROM_IRET) {
// Bit is cleared
Bit32u laddr = (Bit32u) BX_CPU_THIS_PTR gdtr.base + (BX_CPU_THIS_PTR tr.selector.index<<3) + 4;
access_read_linear(laddr, 4, 0, BX_READ, &temp32);
temp32 &= ~0x200;
access_write_linear(laddr, 4, 0, &temp32);
}
//
// Commit point. At this point, we commit to the new
// context. If an unrecoverable error occurs in further
// processing, we complete the task switch without performing
// additional access and segment availablility checks and
// generate the appropriate exception prior to beginning
// execution of the new task.
//
// Step 7: Load the task register with the segment selector and
// descriptor for the new task TSS.
BX_CPU_THIS_PTR tr.selector = *tss_selector;
BX_CPU_THIS_PTR tr.cache = *tss_descriptor;
BX_CPU_THIS_PTR tr.cache.type |= 2; // mark TSS in TR as busy
// Step 8: Set TS flag in the CR0 image stored in the new task TSS.
BX_CPU_THIS_PTR cr0.set_TS(1);
// Task switch clears LE/L3/L2/L1/L0 in DR7
BX_CPU_THIS_PTR dr7 &= ~0x00000155;
// Step 9: If call or interrupt, set the NT flag in the eflags
// image stored in new task's TSS. If IRET or JMP,
// NT is restored from new TSS eflags image. (no change)
// effect on NT flag of new task
if (source == BX_TASK_FROM_CALL_OR_INT) {
newEFLAGS |= EFlagsNTMask; // NT flag is set
}
// Step 10: Load the new task (dynamic) state from new TSS.
// Any errors associated with loading and qualification of
// segment descriptors in this step occur in the new task's
// context. State loaded here includes LDTR, CR3,
// EFLAGS, EIP, general purpose registers, and segment
// descriptor parts of the segment registers.
if ((tss_descriptor->type >= 9) && BX_CPU_THIS_PTR cr0.get_PG()) {
// change CR3 only if it actually modified
if (newCR3 != BX_CPU_THIS_PTR cr3) {
SetCR3(newCR3); // Tell paging unit about new cr3 value
BX_DEBUG (("task_switch changing CR3 to 0x" FMT_PHY_ADDRX, newCR3));
BX_INSTR_TLB_CNTRL(BX_CPU_ID, BX_INSTR_TASKSWITCH, newCR3);
}
}
BX_CPU_THIS_PTR prev_rip = EIP = newEIP;
EAX = newEAX;
ECX = newECX;
EDX = newEDX;
EBX = newEBX;
ESP = newESP;
EBP = newEBP;
ESI = newESI;
EDI = newEDI;
writeEFlags(newEFLAGS, EFlagsValidMask);
// Fill in selectors for all segment registers. If errors
// occur later, the selectors will at least be loaded.
parse_selector(raw_cs_selector, &cs_selector);
BX_CPU_THIS_PTR sregs[BX_SEG_REG_CS].selector = cs_selector;
parse_selector(raw_ds_selector, &ds_selector);
BX_CPU_THIS_PTR sregs[BX_SEG_REG_DS].selector = ds_selector;
parse_selector(raw_es_selector, &es_selector);
BX_CPU_THIS_PTR sregs[BX_SEG_REG_ES].selector = es_selector;
parse_selector(raw_ss_selector, &ss_selector);
BX_CPU_THIS_PTR sregs[BX_SEG_REG_SS].selector = ss_selector;
parse_selector(raw_fs_selector, &fs_selector);
BX_CPU_THIS_PTR sregs[BX_SEG_REG_FS].selector = fs_selector;
parse_selector(raw_gs_selector, &gs_selector);
BX_CPU_THIS_PTR sregs[BX_SEG_REG_GS].selector = gs_selector;
parse_selector(raw_ldt_selector, &ldt_selector);
BX_CPU_THIS_PTR ldtr.selector = ldt_selector;
// Start out with invalid descriptor caches, fill in
// with values only as they are validated.
BX_CPU_THIS_PTR ldtr.cache.valid = 0;
BX_CPU_THIS_PTR ldtr.cache.u.system.limit = 0;
BX_CPU_THIS_PTR ldtr.cache.u.system.limit_scaled = 0;
BX_CPU_THIS_PTR sregs[BX_SEG_REG_ES].cache.valid = 0;
BX_CPU_THIS_PTR sregs[BX_SEG_REG_CS].cache.valid = 0;
BX_CPU_THIS_PTR sregs[BX_SEG_REG_SS].cache.valid = 0;
BX_CPU_THIS_PTR sregs[BX_SEG_REG_DS].cache.valid = 0;
BX_CPU_THIS_PTR sregs[BX_SEG_REG_FS].cache.valid = 0;
BX_CPU_THIS_PTR sregs[BX_SEG_REG_GS].cache.valid = 0;
// LDTR
if (ldt_selector.ti) {
// LDT selector must be in GDT
BX_INFO(("task_switch(exception after commit point): bad LDT selector TI=1"));
exception(BX_TS_EXCEPTION, raw_ldt_selector & 0xfffc, 0);
}
if ((raw_ldt_selector & 0xfffc) != 0) {
bx_bool good = fetch_raw_descriptor2(&ldt_selector, &dword1, &dword2);
if (!good) {
BX_ERROR(("task_switch(exception after commit point): bad LDT fetch"));
exception(BX_TS_EXCEPTION, raw_ldt_selector & 0xfffc, 0);
}
parse_descriptor(dword1, dword2, &ldt_descriptor);
// LDT selector of new task is valid, else #TS(new task's LDT)
if (ldt_descriptor.valid==0 ||
ldt_descriptor.type!=BX_SYS_SEGMENT_LDT ||
ldt_descriptor.segment)
{
BX_ERROR(("task_switch(exception after commit point): bad LDT segment"));
exception(BX_TS_EXCEPTION, raw_ldt_selector & 0xfffc, 0);
}
// LDT of new task is present in memory, else #TS(new tasks's LDT)
if (! IS_PRESENT(ldt_descriptor)) {
BX_ERROR(("task_switch(exception after commit point): LDT not present"));
exception(BX_TS_EXCEPTION, raw_ldt_selector & 0xfffc, 0);
}
// All checks pass, fill in LDTR shadow cache
BX_CPU_THIS_PTR ldtr.cache = ldt_descriptor;
}
else {
// NULL LDT selector is OK, leave cache invalid
}
if (v8086_mode()) {
// load seg regs as 8086 registers
load_seg_reg(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_CS], raw_cs_selector);
load_seg_reg(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_SS], raw_ss_selector);
load_seg_reg(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_DS], raw_ds_selector);
load_seg_reg(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_ES], raw_es_selector);
load_seg_reg(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_FS], raw_fs_selector);
load_seg_reg(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_GS], raw_gs_selector);
}
else {
// SS
if ((raw_ss_selector & 0xfffc) != 0)
{
bx_bool good = fetch_raw_descriptor2(&ss_selector, &dword1, &dword2);
if (!good) {
BX_ERROR(("task_switch(exception after commit point): bad SS fetch"));
exception(BX_TS_EXCEPTION, raw_ss_selector & 0xfffc, 0);
}
parse_descriptor(dword1, dword2, &ss_descriptor);
// SS selector must be within its descriptor table limits else #TS(SS)
// SS descriptor AR byte must must indicate writable data segment,
// else #TS(SS)
if (ss_descriptor.valid==0 || ss_descriptor.segment==0 ||
IS_CODE_SEGMENT(ss_descriptor.type) ||
!IS_DATA_SEGMENT_WRITEABLE(ss_descriptor.type))
{
BX_ERROR(("task_switch(exception after commit point): SS not valid or writeable segment"));
exception(BX_TS_EXCEPTION, raw_ss_selector & 0xfffc, 0);
}
//
// Stack segment is present in memory, else #SS(new stack segment)
//
if (! IS_PRESENT(ss_descriptor)) {
BX_ERROR(("task_switch(exception after commit point): SS not present"));
exception(BX_SS_EXCEPTION, raw_ss_selector & 0xfffc, 0);
}
// Stack segment DPL matches CS.RPL, else #TS(new stack segment)
if (ss_descriptor.dpl != cs_selector.rpl) {
BX_ERROR(("task_switch(exception after commit point): SS.rpl != CS.RPL"));
exception(BX_TS_EXCEPTION, raw_ss_selector & 0xfffc, 0);
}
// Stack segment DPL matches selector RPL, else #TS(new stack segment)
if (ss_descriptor.dpl != ss_selector.rpl) {
BX_ERROR(("task_switch(exception after commit point): SS.dpl != SS.rpl"));
exception(BX_TS_EXCEPTION, raw_ss_selector & 0xfffc, 0);
}
// All checks pass, fill in shadow cache
BX_CPU_THIS_PTR sregs[BX_SEG_REG_SS].cache = ss_descriptor;
}
else {
// SS selector is valid, else #TS(new stack segment)
BX_ERROR(("task_switch(exception after commit point): SS NULL"));
exception(BX_TS_EXCEPTION, raw_ss_selector & 0xfffc, 0);
}
// if new selector is not null then perform following checks:
// index must be within its descriptor table limits else #TS(selector)
// AR byte must indicate data or readable code else #TS(selector)
// if data or non-conforming code then:
// DPL must be >= CPL else #TS(selector)
// DPL must be >= RPL else #TS(selector)
// AR byte must indicate PRESENT else #NP(selector)
// load cache with new segment descriptor and set valid bit
// CS
if ((raw_cs_selector & 0xfffc) != 0) {
bx_bool good = fetch_raw_descriptor2(&cs_selector, &dword1, &dword2);
if (!good) {
BX_ERROR(("task_switch(exception after commit point): bad CS fetch"));
exception(BX_TS_EXCEPTION, raw_cs_selector & 0xfffc, 0);
}
parse_descriptor(dword1, dword2, &cs_descriptor);
// CS descriptor AR byte must indicate code segment else #TS(CS)
if (cs_descriptor.valid==0 || cs_descriptor.segment==0 ||
IS_DATA_SEGMENT(cs_descriptor.type))
{
BX_ERROR(("task_switch(exception after commit point): CS not valid executable seg"));
exception(BX_TS_EXCEPTION, raw_cs_selector & 0xfffc, 0);
}
// if non-conforming then DPL must equal selector RPL else #TS(CS)
if (IS_CODE_SEGMENT_NON_CONFORMING(cs_descriptor.type) &&
cs_descriptor.dpl != cs_selector.rpl)
{
BX_ERROR(("task_switch(exception after commit point): non-conforming: CS.dpl!=CS.RPL"));
exception(BX_TS_EXCEPTION, raw_cs_selector & 0xfffc, 0);
}
// if conforming then DPL must be <= selector RPL else #TS(CS)
if (IS_CODE_SEGMENT_CONFORMING(cs_descriptor.type) &&
cs_descriptor.dpl > cs_selector.rpl)
{
BX_ERROR(("task_switch(exception after commit point): conforming: CS.dpl>RPL"));
exception(BX_TS_EXCEPTION, raw_cs_selector & 0xfffc, 0);
}
// Code segment is present in memory, else #NP(new code segment)
if (! IS_PRESENT(cs_descriptor)) {
BX_ERROR(("task_switch(exception after commit point): CS.p==0"));
exception(BX_NP_EXCEPTION, raw_cs_selector & 0xfffc, 0);
}
// All checks pass, fill in shadow cache
BX_CPU_THIS_PTR sregs[BX_SEG_REG_CS].cache = cs_descriptor;
}
else {
// If new cs selector is null #TS(CS)
BX_ERROR(("task_switch(exception after commit point): CS NULL"));
exception(BX_TS_EXCEPTION, raw_cs_selector & 0xfffc, 0);
}
#if BX_SUPPORT_ICACHE
BX_CPU_THIS_PTR updateFetchModeMask();
#endif
#if BX_CPU_LEVEL >= 4 && BX_SUPPORT_ALIGNMENT_CHECK
handleAlignmentCheck(); // task switch, CPL was modified
#endif
task_switch_load_selector(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_DS],
&ds_selector, raw_ds_selector, cs_selector.rpl);
task_switch_load_selector(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_ES],
&es_selector, raw_es_selector, cs_selector.rpl);
task_switch_load_selector(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_FS],
&fs_selector, raw_fs_selector, cs_selector.rpl);
task_switch_load_selector(&BX_CPU_THIS_PTR sregs[BX_SEG_REG_GS],
&gs_selector, raw_gs_selector, cs_selector.rpl);
}
if ((tss_descriptor->type>=9) && (trap_word & 0x1)) {
BX_CPU_THIS_PTR debug_trap |= 0x00008000; // BT flag in DR6
BX_CPU_THIS_PTR async_event = 1; // so processor knows to check
BX_INFO(("task_switch: T bit set in new TSS"));
}
//
// Step 14: Begin execution of new task.
//
BX_DEBUG(("TASKING: LEAVE"));
}
dotraplinkage void __kprobes
do_general_protection(struct pt_regs *regs, long error_code)
{
struct task_struct *tsk;
conditional_sti(regs);
#ifdef CONFIG_X86_32
if (v8086_mode(regs))
goto gp_in_vm86;
#endif
tsk = current;
if (!user_mode_novm(regs))
goto gp_in_kernel;
#if defined(CONFIG_X86_32) && defined(CONFIG_PAX_PAGEEXEC)
if (!nx_enabled && tsk->mm && (tsk->mm->pax_flags & MF_PAX_PAGEEXEC)) {
struct mm_struct *mm = tsk->mm;
unsigned long limit;
down_write(&mm->mmap_sem);
limit = mm->context.user_cs_limit;
if (limit < TASK_SIZE) {
track_exec_limit(mm, limit, TASK_SIZE, VM_EXEC);
up_write(&mm->mmap_sem);
return;
}
up_write(&mm->mmap_sem);
}
#endif
tsk->thread.error_code = error_code;
tsk->thread.trap_no = 13;
if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
printk_ratelimit()) {
printk(KERN_INFO
"%s[%d] general protection ip:%lx sp:%lx error:%lx",
tsk->comm, task_pid_nr(tsk),
regs->ip, regs->sp, error_code);
print_vma_addr(" in ", regs->ip);
printk("\n");
}
force_sig(SIGSEGV, tsk);
return;
#ifdef CONFIG_X86_32
gp_in_vm86:
local_irq_enable();
handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code);
return;
#endif
gp_in_kernel:
if (fixup_exception(regs))
return;
tsk->thread.error_code = error_code;
tsk->thread.trap_no = 13;
if (notify_die(DIE_GPF, "general protection fault", regs,
error_code, 13, SIGSEGV) == NOTIFY_STOP)
return;
#if defined(CONFIG_X86_32) && defined(CONFIG_PAX_KERNEXEC)
if ((regs->cs & 0xFFFF) == __KERNEL_CS)
die("PAX: suspicious general protection fault", regs, error_code);
else
#endif
die("general protection fault", regs, error_code);
}
/*
* Our handling of the processor debug registers is non-trivial.
* We do not clear them on entry and exit from the kernel. Therefore
* it is possible to get a watchpoint trap here from inside the kernel.
* However, the code in ./ptrace.c has ensured that the user can
* only set watchpoints on userspace addresses. Therefore the in-kernel
* watchpoint trap can only occur in code which is reading/writing
* from user space. Such code must not hold kernel locks (since it
* can equally take a page fault), therefore it is safe to call
* force_sig_info even though that claims and releases locks.
*
* Code in ./signal.c ensures that the debug control register
* is restored before we deliver any signal, and therefore that
* user code runs with the correct debug control register even though
* we clear it here.
*
* Being careful here means that we don't have to be as careful in a
* lot of more complicated places (task switching can be a bit lazy
* about restoring all the debug state, and ptrace doesn't have to
* find every occurrence of the TF bit that could be saved away even
* by user code)
*
* May run on IST stack.
*/
dotraplinkage void __kprobes do_debug(struct pt_regs *regs, long error_code)
{
struct task_struct *tsk = current;
unsigned long condition;
int si_code;
get_debugreg(condition, 6);
/*
* The processor cleared BTF, so don't mark that we need it set.
*/
clear_tsk_thread_flag(tsk, TIF_DEBUGCTLMSR);
tsk->thread.debugctlmsr = 0;
if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code,
SIGTRAP) == NOTIFY_STOP)
return;
/* It's safe to allow irq's after DR6 has been saved */
preempt_conditional_sti(regs);
/* Mask out spurious debug traps due to lazy DR7 setting */
if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) {
if (!tsk->thread.debugreg7)
goto clear_dr7;
}
#ifdef CONFIG_X86_32
if (v8086_mode(regs))
goto debug_vm86;
#endif
/* Save debug status register where ptrace can see it */
tsk->thread.debugreg6 = condition;
/*
* Single-stepping through TF: make sure we ignore any events in
* kernel space (but re-enable TF when returning to user mode).
*/
if (condition & DR_STEP) {
if (!user_mode_novm(regs))
goto clear_TF_reenable;
}
si_code = get_si_code(condition);
/* Ok, finally something we can handle */
send_sigtrap(tsk, regs, error_code, si_code);
/*
* Disable additional traps. They'll be re-enabled when
* the signal is delivered.
*/
clear_dr7:
set_debugreg(0, 7);
preempt_conditional_cli(regs);
return;
#ifdef CONFIG_X86_32
debug_vm86:
/* reenable preemption: handle_vm86_trap() might sleep */
dec_preempt_count();
handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, 1);
conditional_cli(regs);
return;
#endif
clear_TF_reenable:
set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
regs->flags &= ~X86_EFLAGS_TF;
preempt_conditional_cli(regs);
return;
}
static void __kprobes
do_trap(int trapnr, int signr, char *str, struct pt_regs *regs,
long error_code, siginfo_t *info)
{
struct task_struct *tsk = current;
#ifdef CONFIG_X86_32
if (v8086_mode(regs)) {
/*
* traps 0, 1, 3, 4, and 5 should be forwarded to vm86.
* On nmi (interrupt 2), do_trap should not be called.
*/
if (trapnr < 6)
goto vm86_trap;
goto trap_signal;
}
#endif
if (!user_mode_novm(regs))
goto kernel_trap;
#ifdef CONFIG_X86_32
trap_signal:
#endif
/*
* We want error_code and trap_no set for userspace faults and
* kernelspace faults which result in die(), but not
* kernelspace faults which are fixed up. die() gives the
* process no chance to handle the signal and notice the
* kernel fault information, so that won't result in polluting
* the information about previously queued, but not yet
* delivered, faults. See also do_general_protection below.
*/
tsk->thread.error_code = error_code;
tsk->thread.trap_no = trapnr;
#ifdef CONFIG_X86_64
if (show_unhandled_signals && unhandled_signal(tsk, signr) &&
printk_ratelimit()) {
printk(KERN_INFO
"%s[%d] trap %s ip:%lx sp:%lx error:%lx",
tsk->comm, task_pid_nr(tsk), str,
regs->ip, regs->sp, error_code);
print_vma_addr(" in ", regs->ip);
printk("\n");
}
#endif
if (info)
force_sig_info(signr, info, tsk);
else
force_sig(signr, tsk);
return;
kernel_trap:
if (!fixup_exception(regs)) {
tsk->thread.error_code = error_code;
tsk->thread.trap_no = trapnr;
die(str, regs, error_code);
}
#ifdef CONFIG_PAX_REFCOUNT
if (trapnr == 4)
pax_report_refcount_overflow(regs);
#endif
return;
#ifdef CONFIG_X86_32
vm86_trap:
if (handle_vm86_trap((struct kernel_vm86_regs *) regs,
error_code, trapnr))
goto trap_signal;
return;
#endif
}