# HG changeset patch
# User cl349@xxxxxxxxxxxxxxxxxxxx
# Node ID f27205ea60ef4b47734e4a8b8177ca1c2b25fa5d
# Parent 5aae0c8158b9782dd0110a97bbe1274da6e83f6d
# Parent ce018d2730c037b9d8e1a60c56328aebca5bfc48
merge?
diff -r 5aae0c8158b9 -r f27205ea60ef linux-2.6-xen-sparse/mm/memory.c
--- a/linux-2.6-xen-sparse/mm/memory.c Sat Sep 3 16:57:54 2005
+++ b/linux-2.6-xen-sparse/mm/memory.c Sat Sep 3 16:58:50 2005
@@ -1367,20 +1367,15 @@
struct page *old_page, *new_page;
unsigned long pfn = pte_pfn(pte);
pte_t entry;
+ struct page invalid_page;
if (unlikely(!pfn_valid(pfn))) {
- /*
- * This should really halt the system so it can be debugged or
- * at least the kernel stops what it's doing before it corrupts
- * data, but for the moment just pretend this is OOM.
- */
- pte_unmap(page_table);
- printk(KERN_ERR "do_wp_page: bogus page at address %08lx\n",
- address);
- spin_unlock(&mm->page_table_lock);
- return VM_FAULT_OOM;
- }
- old_page = pfn_to_page(pfn);
+ /* This can happen with /dev/mem (PROT_WRITE, MAP_PRIVATE). */
+ invalid_page.flags = (1<<PG_reserved) | (1<<PG_locked);
+ old_page = &invalid_page;
+ } else {
+ old_page = pfn_to_page(pfn);
+ }
if (!TestSetPageLocked(old_page)) {
int reuse = can_share_swap_page(old_page);
@@ -1416,7 +1411,13 @@
new_page = alloc_page_vma(GFP_HIGHUSER, vma, address);
if (!new_page)
goto no_new_page;
- copy_user_highpage(new_page, old_page, address);
+ if (old_page == &invalid_page) {
+ char *vto = kmap_atomic(new_page, KM_USER1);
+ copy_page(vto, (void *)(address & PAGE_MASK));
+ kunmap_atomic(vto, KM_USER1);
+ } else {
+ copy_user_highpage(new_page, old_page, address);
+ }
}
/*
* Re-check the pte - we dropped the lock
diff -r 5aae0c8158b9 -r f27205ea60ef tools/firmware/rombios/rombios.c
--- a/tools/firmware/rombios/rombios.c Sat Sep 3 16:57:54 2005
+++ b/tools/firmware/rombios/rombios.c Sat Sep 3 16:58:50 2005
@@ -31,7 +31,7 @@
// Xen full virtualization does not handle unaligned IO with page crossing.
// Disable 32-bit PIO as a workaround.
-#define NO_PIO32
+#undef NO_PIO32
// ROM BIOS compatability entry points:
diff -r 5aae0c8158b9 -r f27205ea60ef tools/firmware/vmxassist/Makefile
--- a/tools/firmware/vmxassist/Makefile Sat Sep 3 16:57:54 2005
+++ b/tools/firmware/vmxassist/Makefile Sat Sep 3 16:58:50 2005
@@ -24,7 +24,7 @@
# The emulator code lives in ROM space
TEXTADDR=0x000D0000
-DEFINES=-DDEBUG -DENABLE_VME -DTEXTADDR=${TEXTADDR}
+DEFINES=-DDEBUG -DTEXTADDR=${TEXTADDR}
XENINC=-I$(XEN_ROOT)/xen/include -I$(XEN_ROOT)/tools/libxc
#DEFINES=-DDEBUG -DTEST -DTEXTADDR=${TEXTADDR}
#XENINC=-I/home/leendert/xen/xeno-unstable.bk/xen/include
diff -r 5aae0c8158b9 -r f27205ea60ef tools/firmware/vmxassist/setup.c
--- a/tools/firmware/vmxassist/setup.c Sat Sep 3 16:57:54 2005
+++ b/tools/firmware/vmxassist/setup.c Sat Sep 3 16:58:50 2005
@@ -353,7 +353,7 @@
#endif
setup_gdt();
setup_idt();
-#ifdef ENABLE_VME
+#ifndef TEST
set_cr4(get_cr4() | CR4_VME);
#endif
setup_ctx();
diff -r 5aae0c8158b9 -r f27205ea60ef tools/firmware/vmxassist/vm86.c
--- a/tools/firmware/vmxassist/vm86.c Sat Sep 3 16:57:54 2005
+++ b/tools/firmware/vmxassist/vm86.c Sat Sep 3 16:58:50 2005
@@ -465,8 +465,7 @@
* Emulate a segment load in protected mode
*/
int
-load_seg(unsigned long sel, unsigned long *base, unsigned long *limit,
- union vmcs_arbytes *arbytes)
+load_seg(unsigned long sel, u32 *base, u32 *limit, union vmcs_arbytes *arbytes)
{
unsigned long long entry;
diff -r 5aae0c8158b9 -r f27205ea60ef tools/firmware/vmxassist/vmxloader.c
--- a/tools/firmware/vmxassist/vmxloader.c Sat Sep 3 16:57:54 2005
+++ b/tools/firmware/vmxassist/vmxloader.c Sat Sep 3 16:58:50 2005
@@ -110,8 +110,8 @@
}
#ifdef _ACPI_
puts("Loading ACPI ...\n");
- if (ACPI_PHYSICAL_ADDRESS+sizeof(acpi) <= 0xF0000 ){
- /* make sure acpi table does not overlap rombios
+ if (ACPI_PHYSICAL_ADDRESS+sizeof(acpi) <= 0xF0000 ){
+ /* make sure acpi table does not overlap rombios
* currently acpi less than 8K will be OK.
*/
memcpy((void *)ACPI_PHYSICAL_ADDRESS, acpi, sizeof(acpi));
@@ -122,5 +122,6 @@
memcpy((void *)TEXTADDR, vmxassist, sizeof(vmxassist));
puts("Go ...\n");
((void (*)())TEXTADDR)();
+ return 0;
}
diff -r 5aae0c8158b9 -r f27205ea60ef tools/ioemu/exec.c
--- a/tools/ioemu/exec.c Sat Sep 3 16:57:54 2005
+++ b/tools/ioemu/exec.c Sat Sep 3 16:58:50 2005
@@ -142,6 +142,10 @@
#else
setvbuf(logfile, NULL, _IOLBF, 0);
#endif
+/*
+ stdout = logfile;
+ stderr = logfile;
+*/
}
}
@@ -386,9 +390,6 @@
io_mem_write[io_index][1](io_mem_opaque[io_index], addr,
val);
l = 2;
} else {
- if (l!=1){
- fprintf(logfile, "ERROR 8 bit mmio\n");
- }
/* 8 bit access */
val = ldub_raw(buf);
io_mem_write[io_index][0](io_mem_opaque[io_index], addr,
val);
diff -r 5aae0c8158b9 -r f27205ea60ef tools/ioemu/hw/pcnet.c
--- a/tools/ioemu/hw/pcnet.c Sat Sep 3 16:57:54 2005
+++ b/tools/ioemu/hw/pcnet.c Sat Sep 3 16:58:50 2005
@@ -569,6 +569,10 @@
cpu_physical_memory_read(PHYSADDR(s, tmd.tmd0.tbadr),
s->buffer + s->xmit_pos, 4096 - tmd.tmd1.bcnt);
s->xmit_pos += 4096 - tmd.tmd1.bcnt;
+
+ tmd.tmd1.own = 0;
+ TMDSTORE(&tmd, PHYSADDR(s,CSR_CXDA(s)));
+
#ifdef PCNET_DEBUG
printf("pcnet_transmit size=%d\n", s->xmit_pos);
#endif
@@ -580,10 +584,10 @@
s->csr[0] &= ~0x0008; /* clear TDMD */
s->csr[4] |= 0x0004; /* set TXSTRT */
s->xmit_pos = -1;
- }
-
- tmd.tmd1.own = 0;
- TMDSTORE(&tmd, PHYSADDR(s,CSR_CXDA(s)));
+ } else {
+ tmd.tmd1.own = 0;
+ TMDSTORE(&tmd, PHYSADDR(s,CSR_CXDA(s)));
+ }
if (!CSR_TOKINTD(s) || (CSR_LTINTEN(s) && tmd.tmd1.ltint))
s->csr[0] |= 0x0200; /* set TINT */
diff -r 5aae0c8158b9 -r f27205ea60ef tools/ioemu/target-i386-dm/helper2.c
--- a/tools/ioemu/target-i386-dm/helper2.c Sat Sep 3 16:57:54 2005
+++ b/tools/ioemu/target-i386-dm/helper2.c Sat Sep 3 16:58:50 2005
@@ -169,133 +169,217 @@
unsigned long
do_inp(CPUState *env, unsigned long addr, unsigned long size)
{
- switch(size) {
- case 1:
- return cpu_inb(env, addr);
- case 2:
- return cpu_inw(env, addr);
- case 4:
- return cpu_inl(env, addr);
- default:
- fprintf(logfile, "inp: bad size: %lx %lx\n", addr, size);
- exit(-1);
- }
+ switch(size) {
+ case 1:
+ return cpu_inb(env, addr);
+ case 2:
+ return cpu_inw(env, addr);
+ case 4:
+ return cpu_inl(env, addr);
+ default:
+ fprintf(logfile, "inp: bad size: %lx %lx\n", addr, size);
+ exit(-1);
+ }
}
void
do_outp(CPUState *env, unsigned long addr, unsigned long size,
unsigned long val)
{
- switch(size) {
- case 1:
- return cpu_outb(env, addr, val);
- case 2:
- return cpu_outw(env, addr, val);
- case 4:
- return cpu_outl(env, addr, val);
- default:
- fprintf(logfile, "outp: bad size: %lx %lx\n", addr, size);
- exit(-1);
- }
+ switch(size) {
+ case 1:
+ return cpu_outb(env, addr, val);
+ case 2:
+ return cpu_outw(env, addr, val);
+ case 4:
+ return cpu_outl(env, addr, val);
+ default:
+ fprintf(logfile, "outp: bad size: %lx %lx\n", addr, size);
+ exit(-1);
+ }
}
extern void cpu_physical_memory_rw(target_phys_addr_t addr, uint8_t *buf,
int len, int is_write);
static inline void
-read_physical(target_phys_addr_t addr, unsigned long size, void *val)
-{
- return cpu_physical_memory_rw(addr, val, size, 0);
+read_physical(u64 addr, unsigned long size, void *val)
+{
+ return cpu_physical_memory_rw((target_phys_addr_t)addr, val, size, 0);
}
static inline void
-write_physical(target_phys_addr_t addr, unsigned long size, void *val)
-{
- return cpu_physical_memory_rw(addr, val, size, 1);
-}
-
-//send the ioreq to device model
-void cpu_dispatch_ioreq(CPUState *env, ioreq_t *req)
-{
- int i;
- int sign;
-
- sign = (req->df) ? -1 : 1;
-
- if ((!req->pdata_valid) && (req->dir == IOREQ_WRITE)) {
- if (req->size != 4) {
- // Bochs expects higher bits to be 0
- req->u.data &= (1UL << (8 * req->size))-1;
- }
- }
-
- if (req->port_mm == 0){//port io
- if(req->dir == IOREQ_READ){//read
- if (!req->pdata_valid) {
- req->u.data = do_inp(env, req->addr, req->size);
- } else {
- unsigned long tmp;
-
- for (i = 0; i < req->count; i++) {
- tmp = do_inp(env, req->addr, req->size);
-
write_physical((target_phys_addr_t)req->u.pdata + (sign * i * req->size),
- req->size, &tmp);
- }
- }
- } else if(req->dir == IOREQ_WRITE) {
- if (!req->pdata_valid) {
- do_outp(env, req->addr, req->size, req->u.data);
- } else {
- for (i = 0; i < req->count; i++) {
- unsigned long tmp;
-
-
read_physical((target_phys_addr_t)req->u.pdata + (sign * i * req->size),
req->size,
- &tmp);
- do_outp(env, req->addr, req->size, tmp);
- }
- }
-
- }
- } else if (req->port_mm == 1){//memory map io
+write_physical(u64 addr, unsigned long size, void *val)
+{
+ return cpu_physical_memory_rw((target_phys_addr_t)addr, val, size, 1);
+}
+
+void
+cpu_ioreq_pio(CPUState *env, ioreq_t *req)
+{
+ int i, sign;
+
+ sign = req->df ? -1 : 1;
+
+ if (req->dir == IOREQ_READ) {
if (!req->pdata_valid) {
- //handle stos
- if(req->dir == IOREQ_READ) { //read
- for (i = 0; i < req->count; i++) {
-
read_physical((target_phys_addr_t)req->addr + (sign * i * req->size),
req->size, &req->u.data);
- }
- } else if(req->dir == IOREQ_WRITE) { //write
- for (i = 0; i < req->count; i++) {
-
write_physical((target_phys_addr_t)req->addr + (sign * i * req->size),
req->size, &req->u.data);
- }
- }
+ req->u.data = do_inp(env, req->addr, req->size);
} else {
- //handle movs
- unsigned long tmp;
- if (req->dir == IOREQ_READ) {
- for (i = 0; i < req->count; i++) {
-
read_physical((target_phys_addr_t)req->addr + (sign * i * req->size),
req->size, &tmp);
-
write_physical((target_phys_addr_t)req->u.pdata + (sign * i * req->size),
req->size, &tmp);
- }
- } else if (req->dir == IOREQ_WRITE) {
- for (i = 0; i < req->count; i++) {
-
read_physical((target_phys_addr_t)req->u.pdata + (sign * i * req->size),
req->size, &tmp);
-
write_physical((target_phys_addr_t)req->addr + (sign * i * req->size),
req->size, &tmp);
- }
- }
- }
- }
- /* No state change if state = STATE_IORESP_HOOK */
- if (req->state == STATE_IOREQ_INPROCESS)
- req->state = STATE_IORESP_READY;
- env->send_event = 1;
+ unsigned long tmp;
+
+ for (i = 0; i < req->count; i++) {
+ tmp = do_inp(env, req->addr, req->size);
+ write_physical((target_phys_addr_t) req->u.pdata
+ + (sign * i * req->size),
+ req->size, &tmp);
+ }
+ }
+ } else if (req->dir == IOREQ_WRITE) {
+ if (!req->pdata_valid) {
+ do_outp(env, req->addr, req->size, req->u.data);
+ } else {
+ for (i = 0; i < req->count; i++) {
+ unsigned long tmp;
+
+ read_physical((target_phys_addr_t) req->u.pdata
+ + (sign * i * req->size),
+ req->size, &tmp);
+ do_outp(env, req->addr, req->size, tmp);
+ }
+ }
+ }
+}
+
+void
+cpu_ioreq_move(CPUState *env, ioreq_t *req)
+{
+ int i, sign;
+
+ sign = req->df ? -1 : 1;
+
+ if (!req->pdata_valid) {
+ if (req->dir == IOREQ_READ) {
+ for (i = 0; i < req->count; i++) {
+ read_physical(req->addr
+ + (sign * i * req->size),
+ req->size, &req->u.data);
+ }
+ } else if (req->dir == IOREQ_WRITE) {
+ for (i = 0; i < req->count; i++) {
+ write_physical(req->addr
+ + (sign * i * req->size),
+ req->size, &req->u.data);
+ }
+ }
+ } else {
+ unsigned long tmp;
+
+ if (req->dir == IOREQ_READ) {
+ for (i = 0; i < req->count; i++) {
+ read_physical(req->addr
+ + (sign * i * req->size),
+ req->size, &tmp);
+ write_physical((target_phys_addr_t )req->u.pdata
+ + (sign * i * req->size),
+ req->size, &tmp);
+ }
+ } else if (req->dir == IOREQ_WRITE) {
+ for (i = 0; i < req->count; i++) {
+ read_physical((target_phys_addr_t) req->u.pdata
+ + (sign * i * req->size),
+ req->size, &tmp);
+ write_physical(req->addr
+ + (sign * i * req->size),
+ req->size, &tmp);
+ }
+ }
+ }
+}
+
+void
+cpu_ioreq_and(CPUState *env, ioreq_t *req)
+{
+ unsigned long tmp1, tmp2;
+
+ if (req->pdata_valid != 0)
+ hw_error("expected scalar value");
+
+ read_physical(req->addr, req->size, &tmp1);
+ if (req->dir == IOREQ_WRITE) {
+ tmp2 = tmp1 & (unsigned long) req->u.data;
+ write_physical(req->addr, req->size, &tmp2);
+ }
+ req->u.data = tmp1;
+}
+
+void
+cpu_ioreq_or(CPUState *env, ioreq_t *req)
+{
+ unsigned long tmp1, tmp2;
+
+ if (req->pdata_valid != 0)
+ hw_error("expected scalar value");
+
+ read_physical(req->addr, req->size, &tmp1);
+ if (req->dir == IOREQ_WRITE) {
+ tmp2 = tmp1 | (unsigned long) req->u.data;
+ write_physical(req->addr, req->size, &tmp2);
+ }
+ req->u.data = tmp1;
+}
+
+void
+cpu_ioreq_xor(CPUState *env, ioreq_t *req)
+{
+ unsigned long tmp1, tmp2;
+
+ if (req->pdata_valid != 0)
+ hw_error("expected scalar value");
+
+ read_physical(req->addr, req->size, &tmp1);
+ if (req->dir == IOREQ_WRITE) {
+ tmp2 = tmp1 ^ (unsigned long) req->u.data;
+ write_physical(req->addr, req->size, &tmp2);
+ }
+ req->u.data = tmp1;
}
void
cpu_handle_ioreq(CPUState *env)
{
ioreq_t *req = cpu_get_ioreq();
- if (req)
- cpu_dispatch_ioreq(env, req);
+
+ if (req) {
+ if ((!req->pdata_valid) && (req->dir == IOREQ_WRITE)) {
+ if (req->size != 4)
+ req->u.data &= (1UL << (8 * req->size))-1;
+ }
+
+ switch (req->type) {
+ case IOREQ_TYPE_PIO:
+ cpu_ioreq_pio(env, req);
+ break;
+ case IOREQ_TYPE_COPY:
+ cpu_ioreq_move(env, req);
+ break;
+ case IOREQ_TYPE_AND:
+ cpu_ioreq_and(env, req);
+ break;
+ case IOREQ_TYPE_OR:
+ cpu_ioreq_or(env, req);
+ break;
+ case IOREQ_TYPE_XOR:
+ cpu_ioreq_xor(env, req);
+ break;
+ default:
+ hw_error("Invalid ioreq type 0x%x", req->type);
+ }
+
+ /* No state change if state = STATE_IORESP_HOOK */
+ if (req->state == STATE_IOREQ_INPROCESS)
+ req->state = STATE_IORESP_READY;
+ env->send_event = 1;
+ }
}
void
@@ -321,7 +405,7 @@
// Send a message on the event channel. Add the vector to the shared mem
// page.
- intr = &(shared_page->sp_global.pic_intr[0]);
+ intr = (unsigned long *) &(shared_page->sp_global.pic_intr[0]);
atomic_set_bit(vector, intr);
if (loglevel & CPU_LOG_INT)
fprintf(logfile, "injecting vector: %x\n", vector);
diff -r 5aae0c8158b9 -r f27205ea60ef tools/ioemu/vl.c
--- a/tools/ioemu/vl.c Sat Sep 3 16:57:54 2005
+++ b/tools/ioemu/vl.c Sat Sep 3 16:58:50 2005
@@ -413,6 +413,11 @@
fprintf(stderr, "qemu: hardware error: ");
vfprintf(stderr, fmt, ap);
fprintf(stderr, "\n");
+ if (logfile) {
+ fprintf(logfile, "qemu: hardware error: ");
+ vfprintf(logfile, fmt, ap);
+ fprintf(logfile, "\n");
+ }
va_end(ap);
abort();
}
diff -r 5aae0c8158b9 -r f27205ea60ef tools/libxc/xc_linux_save.c
--- a/tools/libxc/xc_linux_save.c Sat Sep 3 16:57:54 2005
+++ b/tools/libxc/xc_linux_save.c Sat Sep 3 16:58:50 2005
@@ -20,6 +20,24 @@
#define BATCH_SIZE 1024 /* 1024 pages (4MB) at a time */
#define MAX_MBIT_RATE 500
+
+
+/*
+** Default values for important tuning parameters. Can override by passing
+** non-zero replacement values to xc_linux_save().
+**
+** XXX SMH: should consider if want to be able to override MAX_MBIT_RATE too.
+**
+*/
+#define DEF_MAX_ITERS 29 /* limit us to 30 times round loop */
+#define DEF_MAX_FACTOR 3 /* never send more than 3x nr_pfns */
+
+
+
+/* Flags to control behaviour of xc_linux_save */
+#define XCFLAGS_LIVE 1
+#define XCFLAGS_DEBUG 2
+
#define DEBUG 0
@@ -320,18 +338,18 @@
xc_dominfo_t *info,
vcpu_guest_context_t *ctxt)
{
- int i=0;
+ int i = 0;
char ans[30];
printf("suspend\n");
fflush(stdout);
if (fgets(ans, sizeof(ans), stdin) == NULL) {
- ERR("failed reading suspend reply");
- return -1;
+ ERR("failed reading suspend reply");
+ return -1;
}
if (strncmp(ans, "done\n", 5)) {
- ERR("suspend reply incorrect: %s", ans);
- return -1;
+ ERR("suspend reply incorrect: %s", ans);
+ return -1;
}
retry:
@@ -377,19 +395,16 @@
return -1;
}
-int xc_linux_save(int xc_handle, int io_fd, u32 dom)
+int xc_linux_save(int xc_handle, int io_fd, u32 dom, u32 max_iters,
+ u32 max_factor, u32 flags)
{
xc_dominfo_t info;
int rc = 1, i, j, k, last_iter, iter = 0;
unsigned long mfn;
- int live = 0; // (ioctxt->flags & XCFLAGS_LIVE);
- int debug = 0; // (ioctxt->flags & XCFLAGS_DEBUG);
+ int live = (flags & XCFLAGS_LIVE);
+ int debug = (flags & XCFLAGS_DEBUG);
int sent_last_iter, skip_this_iter;
-
- /* Important tuning parameters */
- int max_iters = 29; /* limit us to 30 times round loop */
- int max_factor = 3; /* never send more than 3x nr_pfns */
/* The new domain's shared-info frame number. */
unsigned long shared_info_frame;
@@ -442,8 +457,16 @@
MBIT_RATE = START_MBIT_RATE;
- DPRINTF("xc_linux_save start %d\n", dom);
-
+
+ /* If no explicit control parameters given, use defaults */
+ if(!max_iters)
+ max_iters = DEF_MAX_ITERS;
+ if(!max_factor)
+ max_factor = DEF_MAX_FACTOR;
+
+
+ DPRINTF("xc_linux_save start DOM%u live=%s\n", dom, live?"true":"false");
+
if (mlock(&ctxt, sizeof(ctxt))) {
ERR("Unable to mlock ctxt");
return 1;
diff -r 5aae0c8158b9 -r f27205ea60ef tools/libxc/xenguest.h
--- a/tools/libxc/xenguest.h Sat Sep 3 16:57:54 2005
+++ b/tools/libxc/xenguest.h Sat Sep 3 16:58:50 2005
@@ -6,13 +6,12 @@
* Copyright (c) 2003-2004, K A Fraser.
*/
-#ifndef XENBUILD_H
-#define XENBUILD_H
+#ifndef XENGUEST_H
+#define XENGUEST_H
-#define XCFLAGS_VERBOSE 1
-#define XCFLAGS_LIVE 2
-#define XCFLAGS_DEBUG 4
-#define XCFLAGS_CONFIGURE 8
+#define XCFLAGS_LIVE 1
+#define XCFLAGS_DEBUG 2
+
/**
* This function will save a domain running Linux.
@@ -22,7 +21,8 @@
* @parm dom the id of the domain
* @return 0 on success, -1 on failure
*/
-int xc_linux_save(int xc_handle, int fd, uint32_t dom);
+int xc_linux_save(int xc_handle, int fd, uint32_t dom, uint32_t max_iters,
+ uint32_t max_factor, uint32_t flags /* XCFLAGS_xxx */);
/**
* This function will restore a saved domain running Linux.
@@ -35,8 +35,9 @@
* @parm store_mfn returned with the mfn of the store page
* @return 0 on success, -1 on failure
*/
-int xc_linux_restore(int xc_handle, int io_fd, uint32_t dom, unsigned long
nr_pfns,
- unsigned int store_evtchn, unsigned long *store_mfn);
+int xc_linux_restore(int xc_handle, int io_fd, uint32_t dom,
+ unsigned long nr_pfns, unsigned int store_evtchn,
+ unsigned long *store_mfn);
int xc_linux_build(int xc_handle,
uint32_t domid,
@@ -65,4 +66,4 @@
unsigned int store_evtchn,
unsigned long *store_mfn);
-#endif
+#endif // XENGUEST_H
diff -r 5aae0c8158b9 -r f27205ea60ef tools/python/xen/xend/XendCheckpoint.py
--- a/tools/python/xen/xend/XendCheckpoint.py Sat Sep 3 16:57:54 2005
+++ b/tools/python/xen/xend/XendCheckpoint.py Sat Sep 3 16:58:50 2005
@@ -34,7 +34,7 @@
raise XendError(errmsg)
return buf
-def save(xd, fd, dominfo):
+def save(xd, fd, dominfo, live):
write_exact(fd, SIGNATURE, "could not write guest state file: signature")
config = sxp.to_string(dominfo.sxpr())
@@ -42,8 +42,13 @@
"could not write guest state file: config len")
write_exact(fd, config, "could not write guest state file: config")
+ # xc_save takes three customization parameters: maxit, max_f, and flags
+ # the last controls whether or not save is 'live', while the first two
+ # further customize behaviour when 'live' save is enabled. Passing "0"
+ # simply uses the defaults compiled into libxenguest; see the comments
+ # and/or code in xc_linux_save() for more information.
cmd = [PATH_XC_SAVE, str(xc.handle()), str(fd),
- str(dominfo.id)]
+ str(dominfo.id), "0", "0", str(live) ]
log.info("[xc_save] " + join(cmd))
child = xPopen3(cmd, True, -1, [fd, xc.handle()])
diff -r 5aae0c8158b9 -r f27205ea60ef tools/python/xen/xend/XendDomain.py
--- a/tools/python/xen/xend/XendDomain.py Sat Sep 3 16:57:54 2005
+++ b/tools/python/xen/xend/XendDomain.py Sat Sep 3 16:58:50 2005
@@ -542,7 +542,7 @@
dominfo.name = "tmp-" + dominfo.name
try:
- XendCheckpoint.save(self, sock.fileno(), dominfo)
+ XendCheckpoint.save(self, sock.fileno(), dominfo, live)
except:
if dst == "localhost":
dominfo.name = string.replace(dominfo.name, "tmp-", "", 1)
@@ -563,7 +563,8 @@
fd = os.open(dst, os.O_WRONLY | os.O_CREAT | os.O_TRUNC)
- return XendCheckpoint.save(self, fd, dominfo)
+ # For now we don't support 'live checkpoint'
+ return XendCheckpoint.save(self, fd, dominfo, False)
except OSError, ex:
raise XendError("can't write guest state file %s: %s" %
diff -r 5aae0c8158b9 -r f27205ea60ef tools/xcutils/xc_save.c
--- a/tools/xcutils/xc_save.c Sat Sep 3 16:57:54 2005
+++ b/tools/xcutils/xc_save.c Sat Sep 3 16:58:50 2005
@@ -17,14 +17,17 @@
int
main(int argc, char **argv)
{
- unsigned int xc_fd, io_fd, domid;
+ unsigned int xc_fd, io_fd, domid, maxit, max_f, flags;
- if (argc != 4)
- errx(1, "usage: %s xcfd iofd domid", argv[0]);
+ if (argc != 7)
+ errx(1, "usage: %s xcfd iofd domid maxit maxf flags", argv[0]);
xc_fd = atoi(argv[1]);
io_fd = atoi(argv[2]);
domid = atoi(argv[3]);
+ maxit = atoi(argv[4]);
+ max_f = atoi(argv[5]);
+ flags = atoi(argv[6]);
- return xc_linux_save(xc_fd, io_fd, domid);
+ return xc_linux_save(xc_fd, io_fd, domid, maxit, max_f, flags);
}
diff -r 5aae0c8158b9 -r f27205ea60ef xen/arch/x86/shadow.c
--- a/xen/arch/x86/shadow.c Sat Sep 3 16:57:54 2005
+++ b/xen/arch/x86/shadow.c Sat Sep 3 16:58:50 2005
@@ -531,7 +531,7 @@
int i, init_table = 0;
__guest_get_l2e(v, va, &gl2e);
- ASSERT(l2e_get_flags(gl2e) & _PAGE_PRESENT);
+ ASSERT(guest_l2e_get_flags(gl2e) & _PAGE_PRESENT);
gl1pfn = l2e_get_pfn(gl2e);
if ( !(sl1mfn = __shadow_status(d, gl1pfn, PGT_l1_shadow)) )
@@ -1664,7 +1664,7 @@
return 0;
}
- ASSERT(l1e_get_flags(gpte) & _PAGE_RW);
+ ASSERT(guest_l1e_get_flags(gpte) & _PAGE_RW);
guest_l1e_add_flags(gpte, _PAGE_DIRTY | _PAGE_ACCESSED);
spte = l1e_from_pfn(gmfn, guest_l1e_get_flags(gpte) & ~_PAGE_GLOBAL);
diff -r 5aae0c8158b9 -r f27205ea60ef xen/arch/x86/vmx.c
--- a/xen/arch/x86/vmx.c Sat Sep 3 16:57:54 2005
+++ b/xen/arch/x86/vmx.c Sat Sep 3 16:58:50 2005
@@ -602,15 +602,66 @@
return 0;
}
+void send_pio_req(struct cpu_user_regs *regs, unsigned long port,
+ unsigned long count, int size, long value, int dir, int pvalid)
+{
+ struct vcpu *v = current;
+ vcpu_iodata_t *vio;
+ ioreq_t *p;
+
+ vio = get_vio(v->domain, v->vcpu_id);
+ if (vio == NULL) {
+ printk("bad shared page: %lx\n", (unsigned long) vio);
+ domain_crash_synchronous();
+ }
+
+ if (test_bit(ARCH_VMX_IO_WAIT, &v->arch.arch_vmx.flags)) {
+ printf("VMX I/O has not yet completed\n");
+ domain_crash_synchronous();
+ }
+ set_bit(ARCH_VMX_IO_WAIT, &v->arch.arch_vmx.flags);
+
+ p = &vio->vp_ioreq;
+ p->dir = dir;
+ p->pdata_valid = pvalid;
+
+ p->type = IOREQ_TYPE_PIO;
+ p->size = size;
+ p->addr = port;
+ p->count = count;
+ p->df = regs->eflags & EF_DF ? 1 : 0;
+
+ if (pvalid) {
+ if (vmx_paging_enabled(current))
+ p->u.pdata = (void *) gva_to_gpa(value);
+ else
+ p->u.pdata = (void *) value; /* guest VA == guest PA */
+ } else
+ p->u.data = value;
+
+ p->state = STATE_IOREQ_READY;
+
+ if (vmx_portio_intercept(p)) {
+ /* no blocking & no evtchn notification */
+ clear_bit(ARCH_VMX_IO_WAIT, &v->arch.arch_vmx.flags);
+ return;
+ }
+
+ evtchn_send(iopacket_port(v->domain));
+ vmx_wait_io();
+}
+
static void vmx_io_instruction(struct cpu_user_regs *regs,
unsigned long exit_qualification, unsigned long inst_len)
{
- struct vcpu *d = current;
- vcpu_iodata_t *vio;
- ioreq_t *p;
- unsigned long addr;
+ struct mi_per_cpu_info *mpcip;
unsigned long eip, cs, eflags;
+ unsigned long port, size, dir;
int vm86;
+
+ mpcip = ¤t->domain->arch.vmx_platform.mpci;
+ mpcip->instr = INSTR_PIO;
+ mpcip->flags = 0;
__vmread(GUEST_RIP, &eip);
__vmread(GUEST_CS_SELECTOR, &cs);
@@ -623,105 +674,87 @@
vm86, cs, eip, exit_qualification);
if (test_bit(6, &exit_qualification))
- addr = (exit_qualification >> 16) & (0xffff);
+ port = (exit_qualification >> 16) & 0xFFFF;
else
- addr = regs->edx & 0xffff;
- TRACE_VMEXIT (2,addr);
-
- vio = get_vio(d->domain, d->vcpu_id);
- if (vio == 0) {
- printk("bad shared page: %lx", (unsigned long) vio);
- domain_crash_synchronous();
- }
- p = &vio->vp_ioreq;
- p->dir = test_bit(3, &exit_qualification); /* direction */
-
- p->pdata_valid = 0;
- p->count = 1;
- p->size = (exit_qualification & 7) + 1;
+ port = regs->edx & 0xffff;
+ TRACE_VMEXIT(2, port);
+ size = (exit_qualification & 7) + 1;
+ dir = test_bit(3, &exit_qualification); /* direction */
if (test_bit(4, &exit_qualification)) { /* string instruction */
- unsigned long laddr;
-
- __vmread(GUEST_LINEAR_ADDRESS, &laddr);
+ unsigned long addr, count = 1;
+ int sign = regs->eflags & EF_DF ? -1 : 1;
+
+ __vmread(GUEST_LINEAR_ADDRESS, &addr);
+
/*
* In protected mode, guest linear address is invalid if the
* selector is null.
*/
- if (!vm86 && check_for_null_selector(eip)) {
- laddr = (p->dir == IOREQ_WRITE) ? regs->esi : regs->edi;
- }
- p->pdata_valid = 1;
-
- p->u.data = laddr;
- if (vmx_paging_enabled(d))
- p->u.pdata = (void *) gva_to_gpa(p->u.data);
- p->df = (eflags & X86_EFLAGS_DF) ? 1 : 0;
-
- if (test_bit(5, &exit_qualification)) /* "rep" prefix */
- p->count = vm86 ? regs->ecx & 0xFFFF : regs->ecx;
-
- /*
- * Split up string I/O operations that cross page boundaries. Don't
- * advance %eip so that "rep insb" will restart at the next page.
- */
- if ((p->u.data & PAGE_MASK) !=
- ((p->u.data + p->count * p->size - 1) & PAGE_MASK)) {
- VMX_DBG_LOG(DBG_LEVEL_2,
- "String I/O crosses page boundary (cs:eip=0x%lx:0x%lx)\n",
- cs, eip);
- if (p->u.data & (p->size - 1)) {
- printf("Unaligned string I/O operation (cs:eip=0x%lx:0x%lx)\n",
- cs, eip);
- domain_crash_synchronous();
- }
- p->count = (PAGE_SIZE - (p->u.data & ~PAGE_MASK)) / p->size;
- } else {
- __update_guest_eip(inst_len);
- }
- } else if (p->dir == IOREQ_WRITE) {
- p->u.data = regs->eax;
+ if (!vm86 && check_for_null_selector(eip))
+ addr = dir == IOREQ_WRITE ? regs->esi : regs->edi;
+
+ if (test_bit(5, &exit_qualification)) { /* "rep" prefix */
+ mpcip->flags |= REPZ;
+ count = vm86 ? regs->ecx & 0xFFFF : regs->ecx;
+ }
+
+ /*
+ * Handle string pio instructions that cross pages or that
+ * are unaligned. See the comments in vmx_platform.c/handle_mmio()
+ */
+ if ((addr & PAGE_MASK) != ((addr + size - 1) & PAGE_MASK)) {
+ unsigned long value = 0;
+
+ mpcip->flags |= OVERLAP;
+ if (dir == IOREQ_WRITE)
+ vmx_copy(&value, addr, size, VMX_COPY_IN);
+ send_pio_req(regs, port, 1, size, value, dir, 0);
+ } else {
+ if ((addr & PAGE_MASK) != ((addr + count * size - 1) & PAGE_MASK)) {
+ if (sign > 0)
+ count = (PAGE_SIZE - (addr & ~PAGE_MASK)) / size;
+ else
+ count = (addr & ~PAGE_MASK) / size;
+ } else
+ __update_guest_eip(inst_len);
+
+ send_pio_req(regs, port, count, size, addr, dir, 1);
+ }
+ } else {
__update_guest_eip(inst_len);
- } else
- __update_guest_eip(inst_len);
-
- p->addr = addr;
- p->port_mm = 0;
-
- /* Check if the packet needs to be intercepted */
- if (vmx_portio_intercept(p))
- /* no blocking & no evtchn notification */
- return;
-
- set_bit(ARCH_VMX_IO_WAIT, &d->arch.arch_vmx.flags);
- p->state = STATE_IOREQ_READY;
- evtchn_send(iopacket_port(d->domain));
- vmx_wait_io();
-}
-
-enum { COPY_IN = 0, COPY_OUT };
-
-static inline int
+ send_pio_req(regs, port, 1, size, regs->eax, dir, 0);
+ }
+}
+
+int
vmx_copy(void *buf, unsigned long laddr, int size, int dir)
{
+ unsigned long mfn;
char *addr;
- unsigned long mfn;
-
- if ( (size + (laddr & (PAGE_SIZE - 1))) >= PAGE_SIZE )
- {
- printf("vmx_copy exceeds page boundary\n");
- return 0;
- }
-
- mfn = get_mfn_from_pfn(laddr >> PAGE_SHIFT);
- addr = (char *)map_domain_page(mfn) + (laddr & ~PAGE_MASK);
-
- if (dir == COPY_IN)
- memcpy(buf, addr, size);
- else
- memcpy(addr, buf, size);
-
- unmap_domain_page(addr);
+ int count;
+
+ while (size > 0) {
+ count = PAGE_SIZE - (laddr & ~PAGE_MASK);
+ if (count > size)
+ count = size;
+
+ mfn = get_mfn_from_pfn(laddr >> PAGE_SHIFT);
+ /* XXX check whether laddr is valid */
+ addr = (char *)map_domain_page(mfn) + (laddr & ~PAGE_MASK);
+
+ if (dir == VMX_COPY_IN)
+ memcpy(buf, addr, count);
+ else
+ memcpy(addr, buf, count);
+
+ unmap_domain_page(addr);
+
+ laddr += count;
+ buf += count;
+ size -= count;
+ }
+
return 1;
}
@@ -908,7 +941,7 @@
u32 cp;
/* make sure vmxassist exists (this is not an error) */
- if (!vmx_copy(&magic, VMXASSIST_MAGIC_OFFSET, sizeof(magic), COPY_IN))
+ if (!vmx_copy(&magic, VMXASSIST_MAGIC_OFFSET, sizeof(magic), VMX_COPY_IN))
return 0;
if (magic != VMXASSIST_MAGIC)
return 0;
@@ -922,20 +955,20 @@
*/
case VMX_ASSIST_INVOKE:
/* save the old context */
- if (!vmx_copy(&cp, VMXASSIST_OLD_CONTEXT, sizeof(cp), COPY_IN))
+ if (!vmx_copy(&cp, VMXASSIST_OLD_CONTEXT, sizeof(cp), VMX_COPY_IN))
goto error;
if (cp != 0) {
if (!vmx_world_save(d, &c))
goto error;
- if (!vmx_copy(&c, cp, sizeof(c), COPY_OUT))
+ if (!vmx_copy(&c, cp, sizeof(c), VMX_COPY_OUT))
goto error;
}
/* restore the new context, this should activate vmxassist */
- if (!vmx_copy(&cp, VMXASSIST_NEW_CONTEXT, sizeof(cp), COPY_IN))
+ if (!vmx_copy(&cp, VMXASSIST_NEW_CONTEXT, sizeof(cp), VMX_COPY_IN))
goto error;
if (cp != 0) {
- if (!vmx_copy(&c, cp, sizeof(c), COPY_IN))
+ if (!vmx_copy(&c, cp, sizeof(c), VMX_COPY_IN))
goto error;
if (!vmx_world_restore(d, &c))
goto error;
@@ -949,10 +982,10 @@
*/
case VMX_ASSIST_RESTORE:
/* save the old context */
- if (!vmx_copy(&cp, VMXASSIST_OLD_CONTEXT, sizeof(cp), COPY_IN))
+ if (!vmx_copy(&cp, VMXASSIST_OLD_CONTEXT, sizeof(cp), VMX_COPY_IN))
goto error;
if (cp != 0) {
- if (!vmx_copy(&c, cp, sizeof(c), COPY_IN))
+ if (!vmx_copy(&c, cp, sizeof(c), VMX_COPY_IN))
goto error;
if (!vmx_world_restore(d, &c))
goto error;
@@ -1554,15 +1587,18 @@
__vmread(IDT_VECTORING_INFO_FIELD, &idtv_info_field);
if (idtv_info_field & INTR_INFO_VALID_MASK) {
- if ((idtv_info_field & 0x0700) != 0x400) { /* exclude soft ints */
- __vmwrite(VM_ENTRY_INTR_INFO_FIELD, idtv_info_field);
-
- if (idtv_info_field & 0x800) { /* valid error code */
- unsigned long error_code;
- __vmread(VM_EXIT_INTR_ERROR_CODE, &error_code);
- __vmwrite(VM_ENTRY_EXCEPTION_ERROR_CODE, error_code);
- }
- }
+ __vmwrite(VM_ENTRY_INTR_INFO_FIELD, idtv_info_field);
+
+ __vmread(VM_EXIT_INSTRUCTION_LEN, &inst_len);
+ if (inst_len >= 1 && inst_len <= 15)
+ __vmwrite(VM_ENTRY_INSTRUCTION_LEN, inst_len);
+
+ if (idtv_info_field & 0x800) { /* valid error code */
+ unsigned long error_code;
+ __vmread(IDT_VECTORING_ERROR_CODE, &error_code);
+ __vmwrite(VM_ENTRY_EXCEPTION_ERROR_CODE, error_code);
+ }
+
VMX_DBG_LOG(DBG_LEVEL_1, "idtv_info_field=%x", idtv_info_field);
}
diff -r 5aae0c8158b9 -r f27205ea60ef xen/arch/x86/vmx_intercept.c
--- a/xen/arch/x86/vmx_intercept.c Sat Sep 3 16:57:54 2005
+++ b/xen/arch/x86/vmx_intercept.c Sat Sep 3 16:58:50 2005
@@ -172,7 +172,7 @@
if (p->size != 1 ||
p->pdata_valid ||
- p->port_mm)
+ p->type != IOREQ_TYPE_PIO)
return 0;
if (p->addr == PIT_MODE &&
@@ -284,7 +284,5 @@
if (!reinit)
register_portio_handler(0x40, 4, intercept_pit_io);
}
-
-}
-
+}
#endif /* CONFIG_VMX */
diff -r 5aae0c8158b9 -r f27205ea60ef xen/arch/x86/vmx_io.c
--- a/xen/arch/x86/vmx_io.c Sat Sep 3 16:57:54 2005
+++ b/xen/arch/x86/vmx_io.c Sat Sep 3 16:58:50 2005
@@ -33,6 +33,7 @@
#include <asm/vmx_platform.h>
#include <asm/vmx_virpit.h>
#include <asm/apic.h>
+#include <asm/shadow.h>
#include <public/io/ioreq.h>
#include <public/io/vmx_vlapic.h>
@@ -123,7 +124,6 @@
regs->esp &= 0xFFFF0000;
regs->esp |= (value & 0xFFFF);
break;
-
case 5:
regs->ebp &= 0xFFFF0000;
regs->ebp |= (value & 0xFFFF);
@@ -207,7 +207,6 @@
*reg &= ~0xFFFF;
*reg |= (value & 0xFFFF);
break;
-
case LONG:
*reg &= ~0xFFFFFFFF;
*reg |= (value & 0xFFFFFFFF);
@@ -322,13 +321,319 @@
}
#endif
+extern long get_reg_value(int size, int index, int seg, struct cpu_user_regs
*regs);
+
+static inline void set_eflags_CF(int size, unsigned long v1,
+ unsigned long v2, struct cpu_user_regs *regs)
+{
+ unsigned long mask = (1 << (8 * size)) - 1;
+
+ if ((v1 & mask) > (v2 & mask))
+ regs->eflags |= X86_EFLAGS_CF;
+ else
+ regs->eflags &= ~X86_EFLAGS_CF;
+}
+
+static inline void set_eflags_OF(int size, unsigned long v1,
+ unsigned long v2, unsigned long v3, struct cpu_user_regs *regs)
+{
+ if ((v3 ^ v2) & (v3 ^ v1) & (1 << ((8 * size) - 1)))
+ regs->eflags |= X86_EFLAGS_OF;
+}
+
+static inline void set_eflags_AF(int size, unsigned long v1,
+ unsigned long v2, unsigned long v3, struct cpu_user_regs *regs)
+{
+ if ((v1 ^ v2 ^ v3) & 0x10)
+ regs->eflags |= X86_EFLAGS_AF;
+}
+
+static inline void set_eflags_ZF(int size, unsigned long v1,
+ struct cpu_user_regs *regs)
+{
+ unsigned long mask = (1 << (8 * size)) - 1;
+
+ if ((v1 & mask) == 0)
+ regs->eflags |= X86_EFLAGS_ZF;
+}
+
+static inline void set_eflags_SF(int size, unsigned long v1,
+ struct cpu_user_regs *regs)
+{
+ if (v1 & (1 << ((8 * size) - 1)))
+ regs->eflags |= X86_EFLAGS_SF;
+}
+
+static char parity_table[256] = {
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1
+};
+
+static inline void set_eflags_PF(int size, unsigned long v1,
+ struct cpu_user_regs *regs)
+{
+ if (parity_table[v1 & 0xFF])
+ regs->eflags |= X86_EFLAGS_PF;
+}
+
+static void vmx_pio_assist(struct cpu_user_regs *regs, ioreq_t *p,
+ struct mi_per_cpu_info *mpcip)
+{
+ unsigned long old_eax;
+ int sign = p->df ? -1 : 1;
+
+ if (p->dir == IOREQ_WRITE) {
+ if (p->pdata_valid) {
+ regs->esi += sign * p->count * p->size;
+ if (mpcip->flags & REPZ)
+ regs->ecx -= p->count;
+ }
+ } else {
+ if (mpcip->flags & OVERLAP) {
+ unsigned long addr;
+
+ regs->edi += sign * p->count * p->size;
+ if (mpcip->flags & REPZ)
+ regs->ecx -= p->count;
+
+ addr = regs->edi;
+ if (sign > 0)
+ addr -= p->size;
+ vmx_copy(&p->u.data, addr, p->size, VMX_COPY_OUT);
+ } else if (p->pdata_valid) {
+ regs->edi += sign * p->count * p->size;
+ if (mpcip->flags & REPZ)
+ regs->ecx -= p->count;
+ } else {
+ old_eax = regs->eax;
+ switch (p->size) {
+ case 1:
+ regs->eax = (old_eax & 0xffffff00) | (p->u.data & 0xff);
+ break;
+ case 2:
+ regs->eax = (old_eax & 0xffff0000) | (p->u.data & 0xffff);
+ break;
+ case 4:
+ regs->eax = (p->u.data & 0xffffffff);
+ break;
+ default:
+ printk("Error: %s unknown port size\n", __FUNCTION__);
+ domain_crash_synchronous();
+ }
+ }
+ }
+}
+
+static void vmx_mmio_assist(struct cpu_user_regs *regs, ioreq_t *p,
+ struct mi_per_cpu_info *mpcip)
+{
+ int sign = p->df ? -1 : 1;
+ int size = -1, index = -1;
+ unsigned long value = 0, diff = 0;
+ unsigned long src, dst;
+
+ src = mpcip->operand[0];
+ dst = mpcip->operand[1];
+ size = operand_size(src);
+
+ switch (mpcip->instr) {
+ case INSTR_MOV:
+ if (dst & REGISTER) {
+ index = operand_index(dst);
+ set_reg_value(size, index, 0, regs, p->u.data);
+ }
+ break;
+
+ case INSTR_MOVZ:
+ if (dst & REGISTER) {
+ index = operand_index(dst);
+ switch (size) {
+ case BYTE: p->u.data = p->u.data & 0xFFULL; break;
+ case WORD: p->u.data = p->u.data & 0xFFFFULL; break;
+ case LONG: p->u.data = p->u.data & 0xFFFFFFFFULL; break;
+ }
+ set_reg_value(operand_size(dst), index, 0, regs, p->u.data);
+ }
+ break;
+
+ case INSTR_MOVS:
+ sign = p->df ? -1 : 1;
+ regs->esi += sign * p->count * p->size;
+ regs->edi += sign * p->count * p->size;
+
+ if ((mpcip->flags & OVERLAP) && p->dir == IOREQ_READ) {
+ unsigned long addr = regs->edi;
+
+ if (sign > 0)
+ addr -= p->size;
+ vmx_copy(&p->u.data, addr, p->size, VMX_COPY_OUT);
+ }
+
+ if (mpcip->flags & REPZ)
+ regs->ecx -= p->count;
+ break;
+
+ case INSTR_STOS:
+ sign = p->df ? -1 : 1;
+ regs->edi += sign * p->count * p->size;
+ if (mpcip->flags & REPZ)
+ regs->ecx -= p->count;
+ break;
+
+ case INSTR_AND:
+ if (src & REGISTER) {
+ index = operand_index(src);
+ value = get_reg_value(size, index, 0, regs);
+ diff = (unsigned long) p->u.data & value;
+ } else if (src & IMMEDIATE) {
+ value = mpcip->immediate;
+ diff = (unsigned long) p->u.data & value;
+ } else if (src & MEMORY) {
+ index = operand_index(dst);
+ value = get_reg_value(size, index, 0, regs);
+ diff = (unsigned long) p->u.data & value;
+ set_reg_value(size, index, 0, regs, diff);
+ }
+
+ /*
+ * The OF and CF flags are cleared; the SF, ZF, and PF
+ * flags are set according to the result. The state of
+ * the AF flag is undefined.
+ */
+ regs->eflags &= ~(X86_EFLAGS_CF|X86_EFLAGS_PF|
+ X86_EFLAGS_ZF|X86_EFLAGS_SF|X86_EFLAGS_OF);
+ set_eflags_ZF(size, diff, regs);
+ set_eflags_SF(size, diff, regs);
+ set_eflags_PF(size, diff, regs);
+ break;
+
+ case INSTR_OR:
+ if (src & REGISTER) {
+ index = operand_index(src);
+ value = get_reg_value(size, index, 0, regs);
+ diff = (unsigned long) p->u.data | value;
+ } else if (src & IMMEDIATE) {
+ value = mpcip->immediate;
+ diff = (unsigned long) p->u.data | value;
+ } else if (src & MEMORY) {
+ index = operand_index(dst);
+ value = get_reg_value(size, index, 0, regs);
+ diff = (unsigned long) p->u.data | value;
+ set_reg_value(size, index, 0, regs, diff);
+ }
+
+ /*
+ * The OF and CF flags are cleared; the SF, ZF, and PF
+ * flags are set according to the result. The state of
+ * the AF flag is undefined.
+ */
+ regs->eflags &= ~(X86_EFLAGS_CF|X86_EFLAGS_PF|
+ X86_EFLAGS_ZF|X86_EFLAGS_SF|X86_EFLAGS_OF);
+ set_eflags_ZF(size, diff, regs);
+ set_eflags_SF(size, diff, regs);
+ set_eflags_PF(size, diff, regs);
+ break;
+
+ case INSTR_XOR:
+ if (src & REGISTER) {
+ index = operand_index(src);
+ value = get_reg_value(size, index, 0, regs);
+ diff = (unsigned long) p->u.data ^ value;
+ } else if (src & IMMEDIATE) {
+ value = mpcip->immediate;
+ diff = (unsigned long) p->u.data ^ value;
+ } else if (src & MEMORY) {
+ index = operand_index(dst);
+ value = get_reg_value(size, index, 0, regs);
+ diff = (unsigned long) p->u.data ^ value;
+ set_reg_value(size, index, 0, regs, diff);
+ }
+
+ /*
+ * The OF and CF flags are cleared; the SF, ZF, and PF
+ * flags are set according to the result. The state of
+ * the AF flag is undefined.
+ */
+ regs->eflags &= ~(X86_EFLAGS_CF|X86_EFLAGS_PF|
+ X86_EFLAGS_ZF|X86_EFLAGS_SF|X86_EFLAGS_OF);
+ set_eflags_ZF(size, diff, regs);
+ set_eflags_SF(size, diff, regs);
+ set_eflags_PF(size, diff, regs);
+ break;
+
+ case INSTR_CMP:
+ if (src & REGISTER) {
+ index = operand_index(src);
+ value = get_reg_value(size, index, 0, regs);
+ diff = (unsigned long) p->u.data - value;
+ } else if (src & IMMEDIATE) {
+ value = mpcip->immediate;
+ diff = (unsigned long) p->u.data - value;
+ } else if (src & MEMORY) {
+ index = operand_index(dst);
+ value = get_reg_value(size, index, 0, regs);
+ diff = value - (unsigned long) p->u.data;
+ }
+
+ /*
+ * The CF, OF, SF, ZF, AF, and PF flags are set according
+ * to the result
+ */
+ regs->eflags &= ~(X86_EFLAGS_CF|X86_EFLAGS_PF|X86_EFLAGS_AF|
+ X86_EFLAGS_ZF|X86_EFLAGS_SF|X86_EFLAGS_OF);
+ set_eflags_CF(size, value, (unsigned long) p->u.data, regs);
+ set_eflags_OF(size, diff, value, (unsigned long) p->u.data, regs);
+ set_eflags_AF(size, diff, value, (unsigned long) p->u.data, regs);
+ set_eflags_ZF(size, diff, regs);
+ set_eflags_SF(size, diff, regs);
+ set_eflags_PF(size, diff, regs);
+ break;
+
+ case INSTR_TEST:
+ if (src & REGISTER) {
+ index = operand_index(src);
+ value = get_reg_value(size, index, 0, regs);
+ } else if (src & IMMEDIATE) {
+ value = mpcip->immediate;
+ } else if (src & MEMORY) {
+ index = operand_index(dst);
+ value = get_reg_value(size, index, 0, regs);
+ }
+ diff = (unsigned long) p->u.data & value;
+
+ /*
+ * Sets the SF, ZF, and PF status flags. CF and OF are set to 0
+ */
+ regs->eflags &= ~(X86_EFLAGS_CF|X86_EFLAGS_PF|
+ X86_EFLAGS_ZF|X86_EFLAGS_SF|X86_EFLAGS_OF);
+ set_eflags_ZF(size, diff, regs);
+ set_eflags_SF(size, diff, regs);
+ set_eflags_PF(size, diff, regs);
+ break;
+ }
+
+ load_cpu_user_regs(regs);
+}
+
void vmx_io_assist(struct vcpu *v)
{
vcpu_iodata_t *vio;
ioreq_t *p;
struct cpu_user_regs *regs = guest_cpu_user_regs();
- unsigned long old_eax;
- int sign;
struct mi_per_cpu_info *mpci_p;
struct cpu_user_regs *inst_decoder_regs;
@@ -340,80 +645,26 @@
if (vio == 0) {
VMX_DBG_LOG(DBG_LEVEL_1,
"bad shared page: %lx", (unsigned long) vio);
+ printf("bad shared page: %lx\n", (unsigned long) vio);
domain_crash_synchronous();
}
+
p = &vio->vp_ioreq;
-
- if (p->state == STATE_IORESP_HOOK){
+ if (p->state == STATE_IORESP_HOOK)
vmx_hooks_assist(v);
- }
/* clear IO wait VMX flag */
if (test_bit(ARCH_VMX_IO_WAIT, &v->arch.arch_vmx.flags)) {
- if (p->state != STATE_IORESP_READY) {
- /* An interrupt send event raced us */
- return;
- } else {
- p->state = STATE_INVALID;
- }
- clear_bit(ARCH_VMX_IO_WAIT, &v->arch.arch_vmx.flags);
- } else {
- return;
- }
-
- sign = (p->df) ? -1 : 1;
- if (p->port_mm) {
- if (p->pdata_valid) {
- regs->esi += sign * p->count * p->size;
- regs->edi += sign * p->count * p->size;
- } else {
- if (p->dir == IOREQ_WRITE) {
- return;
- }
- int size = -1, index = -1;
-
- size = operand_size(v->domain->arch.vmx_platform.mpci.mmio_target);
- index =
operand_index(v->domain->arch.vmx_platform.mpci.mmio_target);
-
- if (v->domain->arch.vmx_platform.mpci.mmio_target & WZEROEXTEND) {
- p->u.data = p->u.data & 0xffff;
- }
- set_reg_value(size, index, 0, regs, p->u.data);
-
- }
- load_cpu_user_regs(regs);
- return;
- }
-
- if (p->dir == IOREQ_WRITE) {
- if (p->pdata_valid) {
- regs->esi += sign * p->count * p->size;
- regs->ecx -= p->count;
- }
- return;
- } else {
- if (p->pdata_valid) {
- regs->edi += sign * p->count * p->size;
- regs->ecx -= p->count;
- return;
- }
- }
-
- old_eax = regs->eax;
-
- switch(p->size) {
- case 1:
- regs->eax = (old_eax & 0xffffff00) | (p->u.data & 0xff);
- break;
- case 2:
- regs->eax = (old_eax & 0xffff0000) | (p->u.data & 0xffff);
- break;
- case 4:
- regs->eax = (p->u.data & 0xffffffff);
- break;
- default:
- printk("Error: %s unknwon port size\n", __FUNCTION__);
- domain_crash_synchronous();
+ if (p->state == STATE_IORESP_READY) {
+ p->state = STATE_INVALID;
+ clear_bit(ARCH_VMX_IO_WAIT, &v->arch.arch_vmx.flags);
+
+ if (p->type == IOREQ_TYPE_PIO)
+ vmx_pio_assist(regs, p, mpci_p);
+ else
+ vmx_mmio_assist(regs, p, mpci_p);
+ }
+ /* else an interrupt send event raced us */
}
}
@@ -456,8 +707,9 @@
int port = iopacket_port(current->domain);
do {
- if(!test_bit(port, ¤t->domain->shared_info->evtchn_pending[0]))
+ if (!test_bit(port, ¤t->domain->shared_info->evtchn_pending[0]))
do_block();
+
vmx_check_events(current);
if (!test_bit(ARCH_VMX_IO_WAIT, ¤t->arch.arch_vmx.flags))
break;
diff -r 5aae0c8158b9 -r f27205ea60ef xen/arch/x86/vmx_platform.c
--- a/xen/arch/x86/vmx_platform.c Sat Sep 3 16:57:54 2005
+++ b/xen/arch/x86/vmx_platform.c Sat Sep 3 16:58:50 2005
@@ -64,37 +64,37 @@
case QUAD:
return (long)(reg);
default:
- printk("Error: <__get_reg_value>Invalid reg size\n");
+ printf("Error: (__get_reg_value) Invalid reg size\n");
domain_crash_synchronous();
}
}
-static long get_reg_value(int size, int index, int seg, struct cpu_user_regs
*regs)
+long get_reg_value(int size, int index, int seg, struct cpu_user_regs *regs)
{
if (size == BYTE) {
switch (index) {
- case 0: //%al
+ case 0: /* %al */
return (char)(regs->rax & 0xFF);
- case 1: //%cl
+ case 1: /* %cl */
return (char)(regs->rcx & 0xFF);
- case 2: //%dl
+ case 2: /* %dl */
return (char)(regs->rdx & 0xFF);
- case 3: //%bl
+ case 3: /* %bl */
return (char)(regs->rbx & 0xFF);
- case 4: //%ah
+ case 4: /* %ah */
return (char)((regs->rax & 0xFF00) >> 8);
- case 5: //%ch
+ case 5: /* %ch */
return (char)((regs->rcx & 0xFF00) >> 8);
- case 6: //%dh
+ case 6: /* %dh */
return (char)((regs->rdx & 0xFF00) >> 8);
- case 7: //%bh
+ case 7: /* %bh */
return (char)((regs->rbx & 0xFF00) >> 8);
default:
- printk("Error: (get_reg_value)Invalid index value\n");
+ printf("Error: (get_reg_value) Invalid index value\n");
domain_crash_synchronous();
}
-
- }
+ }
+
switch (index) {
case 0: return __get_reg_value(regs->rax, size);
case 1: return __get_reg_value(regs->rcx, size);
@@ -113,7 +113,7 @@
case 14: return __get_reg_value(regs->r14, size);
case 15: return __get_reg_value(regs->r15, size);
default:
- printk("Error: (get_reg_value)Invalid index value\n");
+ printf("Error: (get_reg_value) Invalid index value\n");
domain_crash_synchronous();
}
}
@@ -129,117 +129,91 @@
__vmread(GUEST_RIP, ®s->eip);
}
-static long get_reg_value(int size, int index, int seg, struct cpu_user_regs
*regs)
+static inline long __get_reg_value(unsigned long reg, int size)
{
- /*
- * Reference the db_reg[] table
- */
- switch (size) {
- case BYTE:
+ switch(size) {
+ case WORD:
+ return (short)(reg & 0xFFFF);
+ case LONG:
+ return (int)(reg & 0xFFFFFFFF);
+ default:
+ printf("Error: (__get_reg_value) Invalid reg size\n");
+ domain_crash_synchronous();
+ }
+}
+
+long get_reg_value(int size, int index, int seg, struct cpu_user_regs *regs)
+{
+ if (size == BYTE) {
switch (index) {
- case 0: //%al
+ case 0: /* %al */
return (char)(regs->eax & 0xFF);
- case 1: //%cl
+ case 1: /* %cl */
return (char)(regs->ecx & 0xFF);
- case 2: //%dl
+ case 2: /* %dl */
return (char)(regs->edx & 0xFF);
- case 3: //%bl
+ case 3: /* %bl */
return (char)(regs->ebx & 0xFF);
- case 4: //%ah
+ case 4: /* %ah */
return (char)((regs->eax & 0xFF00) >> 8);
- case 5: //%ch
+ case 5: /* %ch */
return (char)((regs->ecx & 0xFF00) >> 8);
- case 6: //%dh
+ case 6: /* %dh */
return (char)((regs->edx & 0xFF00) >> 8);
- case 7: //%bh
+ case 7: /* %bh */
return (char)((regs->ebx & 0xFF00) >> 8);
default:
- printk("Error: (get_reg_value)size case 0 error\n");
+ printf("Error: (get_reg_value) Invalid index value\n");
domain_crash_synchronous();
}
- case WORD:
+ }
+
switch (index) {
- case 0: //%ax
- return (short)(regs->eax & 0xFFFF);
- case 1: //%cx
- return (short)(regs->ecx & 0xFFFF);
- case 2: //%dx
- return (short)(regs->edx & 0xFFFF);
- case 3: //%bx
- return (short)(regs->ebx & 0xFFFF);
- case 4: //%sp
- return (short)(regs->esp & 0xFFFF);
- break;
- case 5: //%bp
- return (short)(regs->ebp & 0xFFFF);
- case 6: //%si
- return (short)(regs->esi & 0xFFFF);
- case 7: //%di
- return (short)(regs->edi & 0xFFFF);
- default:
- printk("Error: (get_reg_value)size case 1 error\n");
- domain_crash_synchronous();
- }
- case LONG:
- switch (index) {
- case 0: //%eax
- return regs->eax;
- case 1: //%ecx
- return regs->ecx;
- case 2: //%edx
- return regs->edx;
-
- case 3: //%ebx
- return regs->ebx;
- case 4: //%esp
- return regs->esp;
- case 5: //%ebp
- return regs->ebp;
- case 6: //%esi
- return regs->esi;
- case 7: //%edi
- return regs->edi;
- default:
- printk("Error: (get_reg_value)size case 2 error\n");
- domain_crash_synchronous();
- }
+ case 0: return __get_reg_value(regs->eax, size);
+ case 1: return __get_reg_value(regs->ecx, size);
+ case 2: return __get_reg_value(regs->edx, size);
+ case 3: return __get_reg_value(regs->ebx, size);
+ case 4: return __get_reg_value(regs->esp, size);
+ case 5: return __get_reg_value(regs->ebp, size);
+ case 6: return __get_reg_value(regs->esi, size);
+ case 7: return __get_reg_value(regs->edi, size);
default:
- printk("Error: (get_reg_value)size case error\n");
+ printf("Error: (get_reg_value) Invalid index value\n");
domain_crash_synchronous();
}
}
#endif
-static inline const unsigned char *check_prefix(const unsigned char *inst,
struct instruction *thread_inst, unsigned char *rex_p)
+static inline unsigned char *check_prefix(unsigned char *inst,
+ struct instruction *thread_inst, unsigned char *rex_p)
{
while (1) {
switch (*inst) {
- /* rex prefix for em64t instructions*/
+ /* rex prefix for em64t instructions */
case 0x40 ... 0x4e:
*rex_p = *inst;
break;
-
- case 0xf3: //REPZ
+ case 0xf3: /* REPZ */
thread_inst->flags = REPZ;
- break;
- case 0xf2: //REPNZ
+ break;
+ case 0xf2: /* REPNZ */
thread_inst->flags = REPNZ;
- break;
- case 0xf0: //LOCK
+ break;
+ case 0xf0: /* LOCK */
break;
- case 0x2e: //CS
- case 0x36: //SS
- case 0x3e: //DS
- case 0x26: //ES
- case 0x64: //FS
- case 0x65: //GS
- thread_inst->seg_sel = *inst;
+ case 0x2e: /* CS */
+ case 0x36: /* SS */
+ case 0x3e: /* DS */
+ case 0x26: /* ES */
+ case 0x64: /* FS */
+ case 0x65: /* GS */
+ thread_inst->seg_sel = *inst;
break;
- case 0x66: //32bit->16bit
+ case 0x66: /* 32bit->16bit */
thread_inst->op_size = WORD;
break;
case 0x67:
- printf("Error: Not handling 0x67 (yet)\n");
+ printf("Error: Not handling 0x67 (yet)\n");
domain_crash_synchronous();
break;
default:
@@ -249,7 +223,7 @@
}
}
-static inline unsigned long get_immediate(int op16, const unsigned char *inst,
int op_size)
+static inline unsigned long get_immediate(int op16,const unsigned char *inst,
int op_size)
{
int mod, reg, rm;
unsigned long val = 0;
@@ -317,197 +291,299 @@
static void init_instruction(struct instruction *mmio_inst)
{
- memset(mmio_inst->i_name, '0', I_NAME_LEN);
- mmio_inst->op_size = 0;
- mmio_inst->offset = 0;
+ mmio_inst->instr = 0;
+ mmio_inst->op_size = 0;
mmio_inst->immediate = 0;
mmio_inst->seg_sel = 0;
- mmio_inst->op_num = 0;
mmio_inst->operand[0] = 0;
mmio_inst->operand[1] = 0;
- mmio_inst->operand[2] = 0;
mmio_inst->flags = 0;
}
#define GET_OP_SIZE_FOR_BYTE(op_size) \
- do {if (rex) op_size = BYTE_64;else op_size = BYTE;} while(0)
+ do { \
+ if (rex) \
+ op_size = BYTE_64; \
+ else \
+ op_size = BYTE; \
+ } while(0)
#define GET_OP_SIZE_FOR_NONEBYTE(op_size) \
- do {if (rex & 0x8) op_size = QUAD; else if (op_size != WORD) op_size =
LONG;} while(0)
-
-static int vmx_decode(const unsigned char *inst, struct instruction
*thread_inst)
+ do { \
+ if (rex & 0x8) \
+ op_size = QUAD; \
+ else if (op_size != WORD) \
+ op_size = LONG; \
+ } while(0)
+
+
+/*
+ * Decode mem,accumulator operands (as in <opcode> m8/m16/m32, al,ax,eax)
+ */
+static int mem_acc(unsigned char size, struct instruction *instr)
+{
+ instr->operand[0] = mk_operand(size, 0, 0, MEMORY);
+ instr->operand[1] = mk_operand(size, 0, 0, REGISTER);
+ return DECODE_success;
+}
+
+/*
+ * Decode accumulator,mem operands (as in <opcode> al,ax,eax, m8/m16/m32)
+ */
+static int acc_mem(unsigned char size, struct instruction *instr)
+{
+ instr->operand[0] = mk_operand(size, 0, 0, REGISTER);
+ instr->operand[1] = mk_operand(size, 0, 0, MEMORY);
+ return DECODE_success;
+}
+
+/*
+ * Decode mem,reg operands (as in <opcode> r32/16, m32/16)
+ */
+static int mem_reg(unsigned char size, unsigned char *opcode,
+ struct instruction *instr, unsigned char rex)
+{
+ int index = get_index(opcode + 1, rex);
+
+ instr->operand[0] = mk_operand(size, 0, 0, MEMORY);
+ instr->operand[1] = mk_operand(size, index, 0, REGISTER);
+ return DECODE_success;
+}
+
+/*
+ * Decode reg,mem operands (as in <opcode> m32/16, r32/16)
+ */
+static int reg_mem(unsigned char size, unsigned char *opcode,
+ struct instruction *instr, unsigned char rex)
+{
+ int index = get_index(opcode + 1, rex);
+
+ instr->operand[0] = mk_operand(size, index, 0, REGISTER);
+ instr->operand[1] = mk_operand(size, 0, 0, MEMORY);
+ return DECODE_success;
+}
+
+static int vmx_decode(unsigned char *opcode, struct instruction *instr)
{
unsigned long eflags;
int index, vm86 = 0;
unsigned char rex = 0;
unsigned char tmp_size = 0;
-
- init_instruction(thread_inst);
-
- inst = check_prefix(inst, thread_inst, &rex);
+ init_instruction(instr);
+
+ opcode = check_prefix(opcode, instr, &rex);
__vmread(GUEST_RFLAGS, &eflags);
if (eflags & X86_EFLAGS_VM)
vm86 = 1;
if (vm86) { /* meaning is reversed */
- if (thread_inst->op_size == WORD)
- thread_inst->op_size = LONG;
- else if (thread_inst->op_size == LONG)
- thread_inst->op_size = WORD;
- else if (thread_inst->op_size == 0)
- thread_inst->op_size = WORD;
- }
-
- switch(*inst) {
- case 0x81:
- /* This is only a workaround for cmpl instruction*/
- strcpy((char *)thread_inst->i_name, "cmp");
+ if (instr->op_size == WORD)
+ instr->op_size = LONG;
+ else if (instr->op_size == LONG)
+ instr->op_size = WORD;
+ else if (instr->op_size == 0)
+ instr->op_size = WORD;
+ }
+
+ switch (*opcode) {
+ case 0x0B: /* or m32/16, r32/16 */
+ instr->instr = INSTR_OR;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return mem_reg(instr->op_size, opcode, instr, rex);
+
+ case 0x20: /* and r8, m8 */
+ instr->instr = INSTR_AND;
+ GET_OP_SIZE_FOR_BYTE(instr->op_size);
+ return reg_mem(instr->op_size, opcode, instr, rex);
+
+ case 0x21: /* and r32/16, m32/16 */
+ instr->instr = INSTR_AND;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return reg_mem(instr->op_size, opcode, instr, rex);
+
+ case 0x23: /* and m32/16, r32/16 */
+ instr->instr = INSTR_AND;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return mem_reg(instr->op_size, opcode, instr, rex);
+
+ case 0x30: /* xor r8, m8 */
+ instr->instr = INSTR_XOR;
+ GET_OP_SIZE_FOR_BYTE(instr->op_size);
+ return reg_mem(instr->op_size, opcode, instr, rex);
+
+ case 0x31: /* xor r32/16, m32/16 */
+ instr->instr = INSTR_XOR;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return reg_mem(instr->op_size, opcode, instr, rex);
+
+ case 0x39: /* cmp r32/16, m32/16 */
+ instr->instr = INSTR_CMP;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return reg_mem(instr->op_size, opcode, instr, rex);
+
+ case 0x81:
+ if (((opcode[1] >> 3) & 7) == 7) { /* cmp $imm, m32/16 */
+ instr->instr = INSTR_CMP;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+
+ instr->operand[0] = mk_operand(instr->op_size, 0, 0, IMMEDIATE);
+ instr->immediate = get_immediate(vm86, opcode+1, BYTE);
+ instr->operand[1] = mk_operand(instr->op_size, 0, 0, MEMORY);
+
return DECODE_success;
-
- case 0x88:
- /* mov r8 to m8 */
- thread_inst->op_size = BYTE;
- index = get_index((inst + 1), rex);
- GET_OP_SIZE_FOR_BYTE(tmp_size);
- thread_inst->operand[0] = mk_operand(tmp_size, index, 0, REGISTER);
-
- break;
- case 0x89:
- /* mov r32/16 to m32/16 */
- index = get_index((inst + 1), rex);
- GET_OP_SIZE_FOR_NONEBYTE(thread_inst->op_size);
- thread_inst->operand[0] = mk_operand(thread_inst->op_size, index,
0, REGISTER);
-
- break;
- case 0x8a:
- /* mov m8 to r8 */
- thread_inst->op_size = BYTE;
- index = get_index((inst + 1), rex);
- GET_OP_SIZE_FOR_BYTE(tmp_size);
- thread_inst->operand[1] = mk_operand(tmp_size, index, 0, REGISTER);
- break;
- case 0x8b:
- /* mov r32/16 to m32/16 */
- index = get_index((inst + 1), rex);
- GET_OP_SIZE_FOR_NONEBYTE(thread_inst->op_size);
- thread_inst->operand[1] = mk_operand(thread_inst->op_size, index,
0, REGISTER);
- break;
- case 0x8c:
- case 0x8e:
- printk("%x, This opcode hasn't been handled yet!", *inst);
- return DECODE_failure;
- /* Not handle it yet. */
- case 0xa0:
- /* mov byte to al */
- thread_inst->op_size = BYTE;
- GET_OP_SIZE_FOR_BYTE(tmp_size);
- thread_inst->operand[1] = mk_operand(tmp_size, 0, 0, REGISTER);
- break;
- case 0xa1:
- /* mov word/doubleword to ax/eax */
- GET_OP_SIZE_FOR_NONEBYTE(thread_inst->op_size);
- thread_inst->operand[1] = mk_operand(thread_inst->op_size, 0, 0,
REGISTER);
-
- break;
- case 0xa2:
- /* mov al to (seg:offset) */
- thread_inst->op_size = BYTE;
- GET_OP_SIZE_FOR_BYTE(tmp_size);
- thread_inst->operand[0] = mk_operand(tmp_size, 0, 0, REGISTER);
- break;
- case 0xa3:
- /* mov ax/eax to (seg:offset) */
- GET_OP_SIZE_FOR_NONEBYTE(thread_inst->op_size);
- thread_inst->operand[0] = mk_operand(thread_inst->op_size, 0, 0,
REGISTER);
- break;
- case 0xa4:
- /* movsb */
- thread_inst->op_size = BYTE;
- strcpy((char *)thread_inst->i_name, "movs");
- return DECODE_success;
- case 0xa5:
- /* movsw/movsl */
- GET_OP_SIZE_FOR_NONEBYTE(thread_inst->op_size);
- strcpy((char *)thread_inst->i_name, "movs");
- return DECODE_success;
- case 0xaa:
- /* stosb */
- thread_inst->op_size = BYTE;
- strcpy((char *)thread_inst->i_name, "stosb");
- return DECODE_success;
- case 0xab:
- /* stosw/stosl */
- if (thread_inst->op_size == WORD) {
- strcpy((char *)thread_inst->i_name, "stosw");
- } else {
- thread_inst->op_size = LONG;
- strcpy((char *)thread_inst->i_name, "stosl");
- }
- return DECODE_success;
- case 0xc6:
- /* mov imm8 to m8 */
- thread_inst->op_size = BYTE;
- thread_inst->operand[0] = mk_operand(BYTE, 0, 0, IMMEDIATE);
- thread_inst->immediate = get_immediate(vm86,
- (inst+1), thread_inst->op_size);
- break;
- case 0xc7:
- /* mov imm16/32 to m16/32 */
- GET_OP_SIZE_FOR_NONEBYTE(thread_inst->op_size);
- thread_inst->operand[0] = mk_operand(thread_inst->op_size, 0, 0,
IMMEDIATE);
- thread_inst->immediate = get_immediate(vm86, (inst+1),
thread_inst->op_size);
+ } else
+ return DECODE_failure;
+
+ case 0x84: /* test m8, r8 */
+ instr->instr = INSTR_TEST;
+ instr->op_size = BYTE;
+ GET_OP_SIZE_FOR_BYTE(tmp_size);
+ return mem_reg(tmp_size, opcode, instr, rex);
+
+ case 0x88: /* mov r8, m8 */
+ instr->instr = INSTR_MOV;
+ instr->op_size = BYTE;
+ GET_OP_SIZE_FOR_BYTE(tmp_size);
+ return reg_mem(tmp_size, opcode, instr, rex);
+
+ case 0x89: /* mov r32/16, m32/16 */
+ instr->instr = INSTR_MOV;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return reg_mem(instr->op_size, opcode, instr, rex);
+
+ case 0x8A: /* mov m8, r8 */
+ instr->instr = INSTR_MOV;
+ instr->op_size = BYTE;
+ GET_OP_SIZE_FOR_BYTE(tmp_size);
+ return mem_reg(tmp_size, opcode, instr, rex);
+
+ case 0x8B: /* mov m32/16, r32/16 */
+ instr->instr = INSTR_MOV;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return mem_reg(instr->op_size, opcode, instr, rex);
+
+ case 0xA0: /* mov <addr>, al */
+ instr->instr = INSTR_MOV;
+ instr->op_size = BYTE;
+ GET_OP_SIZE_FOR_BYTE(tmp_size);
+ return mem_acc(tmp_size, instr);
+
+ case 0xA1: /* mov <addr>, ax/eax */
+ instr->instr = INSTR_MOV;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return mem_acc(instr->op_size, instr);
+
+ case 0xA2: /* mov al, <addr> */
+ instr->instr = INSTR_MOV;
+ instr->op_size = BYTE;
+ GET_OP_SIZE_FOR_BYTE(tmp_size);
+ return acc_mem(tmp_size, instr);
+
+ case 0xA3: /* mov ax/eax, <addr> */
+ instr->instr = INSTR_MOV;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return acc_mem(instr->op_size, instr);
+
+ case 0xA4: /* movsb */
+ instr->instr = INSTR_MOVS;
+ instr->op_size = BYTE;
+ return DECODE_success;
- break;
- case 0x0f:
- break;
- default:
- printk("%x, This opcode hasn't been handled yet!", *inst);
- return DECODE_failure;
- }
+ case 0xA5: /* movsw/movsl */
+ instr->instr = INSTR_MOVS;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return DECODE_success;
- strcpy((char *)thread_inst->i_name, "mov");
- if (*inst != 0x0f) {
+ case 0xAA: /* stosb */
+ instr->instr = INSTR_STOS;
+ instr->op_size = BYTE;
return DECODE_success;
- }
-
- inst++;
- switch (*inst) {
+
+ case 0xAB: /* stosw/stosl */
+ instr->instr = INSTR_STOS;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ return DECODE_success;
- /* movz */
- case 0xb6:
- index = get_index((inst + 1), rex);
- GET_OP_SIZE_FOR_NONEBYTE(thread_inst->op_size);
- thread_inst->operand[1] = mk_operand(thread_inst->op_size, index,
0, REGISTER);
- thread_inst->op_size = BYTE;
- strcpy((char *)thread_inst->i_name, "movzb");
+ case 0xC6:
+ if (((opcode[1] >> 3) & 7) == 0) { /* mov $imm8, m8 */
+ instr->instr = INSTR_MOV;
+ instr->op_size = BYTE;
+
+ instr->operand[0] = mk_operand(instr->op_size, 0, 0, IMMEDIATE);
+ instr->immediate = get_immediate(vm86, opcode+1, instr->op_size);
+ instr->operand[1] = mk_operand(instr->op_size, 0, 0, MEMORY);
return DECODE_success;
- case 0xb7:
- index = get_index((inst + 1), rex);
- if (rex & 0x8) {
- thread_inst->op_size = LONG;
- thread_inst->operand[1] = mk_operand(QUAD, index, 0,
REGISTER);
- } else {
- thread_inst->op_size = WORD;
- thread_inst->operand[1] = mk_operand(LONG, index, 0,
REGISTER);
- }
+ } else
+ return DECODE_failure;
- strcpy((char *)thread_inst->i_name, "movzw");
+ case 0xC7:
+ if (((opcode[1] >> 3) & 7) == 0) { /* mov $imm16/32, m16/32 */
+ instr->instr = INSTR_MOV;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+
+ instr->operand[0] = mk_operand(instr->op_size, 0, 0, IMMEDIATE);
+ instr->immediate = get_immediate(vm86, opcode+1, instr->op_size);
+ instr->operand[1] = mk_operand(instr->op_size, 0, 0, MEMORY);
return DECODE_success;
- default:
- printk("0f %x, This opcode hasn't been handled yet!", *inst);
- return DECODE_failure;
- }
-
- /* will never reach here */
- return DECODE_failure;
-}
-
+ } else
+ return DECODE_failure;
+
+ case 0xF6:
+ if (((opcode[1] >> 3) & 7) == 0) { /* testb $imm8, m8 */
+ instr->instr = INSTR_TEST;
+ instr->op_size = BYTE;
+
+ instr->operand[0] = mk_operand(instr->op_size, 0, 0, IMMEDIATE);
+ instr->immediate = get_immediate(vm86, opcode+1, instr->op_size);
+ instr->operand[1] = mk_operand(instr->op_size, 0, 0, MEMORY);
+
+ return DECODE_success;
+ } else
+ return DECODE_failure;
+
+ case 0x0F:
+ break;
+
+ default:
+ printf("%x, This opcode isn't handled yet!\n", *opcode);
+ return DECODE_failure;
+ }
+
+ switch (*++opcode) {
+ case 0xB6: /* movz m8, r16/r32 */
+ instr->instr = INSTR_MOVZ;
+ GET_OP_SIZE_FOR_NONEBYTE(instr->op_size);
+ index = get_index(opcode + 1, rex);
+ instr->operand[0] = mk_operand(BYTE, 0, 0, MEMORY);
+ instr->operand[1] = mk_operand(instr->op_size, index, 0, REGISTER);
+ return DECODE_success;
+
+ case 0xB7: /* movz m16, r32 */
+ instr->instr = INSTR_MOVZ;
+ index = get_index(opcode + 1, rex);
+ if (rex & 0x8) {
+ instr->op_size = LONG;
+ instr->operand[1] = mk_operand(QUAD, index, 0, REGISTER);
+ } else {
+ instr->op_size = WORD;
+ instr->operand[1] = mk_operand(LONG, index, 0, REGISTER);
+ }
+ instr->operand[0] = mk_operand(instr->op_size, 0, 0, MEMORY);
+ return DECODE_success;
+
+ default:
+ printf("0f %x, This opcode isn't handled yet\n", *opcode);
+ return DECODE_failure;
+ }
+}
+
+/* XXX use vmx_copy instead */
int inst_copy_from_guest(unsigned char *buf, unsigned long guest_eip, int
inst_len)
{
unsigned long gpa;
@@ -552,40 +628,27 @@
return inst_len+remaining;
}
-static int read_from_mmio(struct instruction *inst_p)
-{
- // Only for mov instruction now!!!
- if (inst_p->operand[1] & REGISTER)
- return 1;
-
- return 0;
-}
-
-// dir: 1 read from mmio
-// 0 write to mmio
-static void send_mmio_req(unsigned long gpa,
- struct instruction *inst_p, long value, int dir, int pvalid)
+void send_mmio_req(unsigned char type, unsigned long gpa,
+ unsigned long count, int size, long value, int dir, int pvalid)
{
struct vcpu *d = current;
vcpu_iodata_t *vio;
ioreq_t *p;
int vm86;
- struct mi_per_cpu_info *mpci_p;
- struct cpu_user_regs *inst_decoder_regs;
+ struct cpu_user_regs *regs;
extern long evtchn_send(int lport);
- mpci_p = ¤t->domain->arch.vmx_platform.mpci;
- inst_decoder_regs = mpci_p->inst_decoder_regs;
+ regs = current->domain->arch.vmx_platform.mpci.inst_decoder_regs;
vio = get_vio(d->domain, d->vcpu_id);
-
if (vio == NULL) {
- printk("bad shared page\n");
+ printf("bad shared page\n");
domain_crash_synchronous();
}
+
p = &vio->vp_ioreq;
- vm86 = inst_decoder_regs->eflags & X86_EFLAGS_VM;
+ vm86 = regs->eflags & X86_EFLAGS_VM;
if (test_bit(ARCH_VMX_IO_WAIT, &d->arch.arch_vmx.flags)) {
printf("VMX I/O has not yet completed\n");
@@ -596,24 +659,21 @@
p->dir = dir;
p->pdata_valid = pvalid;
- p->port_mm = 1;
- p->size = inst_p->op_size;
+ p->type = type;
+ p->size = size;
p->addr = gpa;
- p->u.data = value;
+ p->count = count;
+ p->df = regs->eflags & EF_DF ? 1 : 0;
+
+ if (pvalid) {
+ if (vmx_paging_enabled(current))
+ p->u.pdata = (void *) gva_to_gpa(value);
+ else
+ p->u.pdata = (void *) value; /* guest VA == guest PA */
+ } else
+ p->u.data = value;
p->state = STATE_IOREQ_READY;
-
- if (inst_p->flags & REPZ) {
- if (vm86)
- p->count = inst_decoder_regs->ecx & 0xFFFF;
- else
- p->count = inst_decoder_regs->ecx;
- p->df = (inst_decoder_regs->eflags & EF_DF) ? 1 : 0;
- } else
- p->count = 1;
-
- if ((pvalid) && vmx_paging_enabled(current))
- p->u.pdata = (void *) gva_to_gpa(p->u.data);
if (vmx_mmio_intercept(p)){
p->state = STATE_IORESP_READY;
@@ -625,18 +685,50 @@
vmx_wait_io();
}
+static void mmio_operands(int type, unsigned long gpa, struct instruction
*inst,
+ struct mi_per_cpu_info *mpcip, struct cpu_user_regs *regs)
+{
+ unsigned long value = 0;
+ int index, size;
+
+ size = operand_size(inst->operand[0]);
+
+ mpcip->flags = inst->flags;
+ mpcip->instr = inst->instr;
+ mpcip->operand[0] = inst->operand[0]; /* source */
+ mpcip->operand[1] = inst->operand[1]; /* destination */
+
+ if (inst->operand[0] & REGISTER) { /* dest is memory */
+ index = operand_index(inst->operand[0]);
+ value = get_reg_value(size, index, 0, regs);
+ send_mmio_req(type, gpa, 1, size, value, IOREQ_WRITE, 0);
+ } else if (inst->operand[0] & IMMEDIATE) { /* dest is memory */
+ value = inst->immediate;
+ send_mmio_req(type, gpa, 1, size, value, IOREQ_WRITE, 0);
+ } else if (inst->operand[0] & MEMORY) { /* dest is register */
+ /* send the request and wait for the value */
+ send_mmio_req(type, gpa, 1, size, 0, IOREQ_READ, 0);
+ } else {
+ printf("mmio_operands: invalid operand\n");
+ domain_crash_synchronous();
+ }
+}
+
+#define GET_REPEAT_COUNT() \
+ (mmio_inst.flags & REPZ ? (vm86 ? regs->ecx & 0xFFFF : regs->ecx) : 1)
+
void handle_mmio(unsigned long va, unsigned long gpa)
{
unsigned long eip, eflags, cs;
unsigned long inst_len, inst_addr;
- struct mi_per_cpu_info *mpci_p;
- struct cpu_user_regs *inst_decoder_regs;
+ struct mi_per_cpu_info *mpcip;
+ struct cpu_user_regs *regs;
struct instruction mmio_inst;
unsigned char inst[MAX_INST_LEN];
- int vm86, ret;
+ int i, vm86, ret;
- mpci_p = ¤t->domain->arch.vmx_platform.mpci;
- inst_decoder_regs = mpci_p->inst_decoder_regs;
+ mpcip = ¤t->domain->arch.vmx_platform.mpci;
+ regs = mpcip->inst_decoder_regs;
__vmread(GUEST_RIP, &eip);
__vmread(VM_EXIT_INSTRUCTION_LEN, &inst_len);
@@ -647,108 +739,142 @@
__vmread(GUEST_CS_SELECTOR, &cs);
inst_addr = (cs << 4) + eip;
} else
- inst_addr = eip; /* XXX should really look at GDT[cs].base too */
-
- memset(inst, '0', MAX_INST_LEN);
+ inst_addr = eip;
+
+ memset(inst, 0, MAX_INST_LEN);
ret = inst_copy_from_guest(inst, inst_addr, inst_len);
if (ret != inst_len) {
- printk("handle_mmio - EXIT: get guest instruction fault\n");
+ printf("handle_mmio - EXIT: get guest instruction fault\n");
domain_crash_synchronous();
}
-
init_instruction(&mmio_inst);
if (vmx_decode(inst, &mmio_inst) == DECODE_failure) {
- printk("vmx decode failure: eip=%lx, va=%lx\n %x %x %x %x\n", eip, va,
- inst[0], inst[1], inst[2], inst[3]);
+ printf("mmio opcode: va 0x%lx, gpa 0x%lx, len %ld:",
+ va, gpa, inst_len);
+ for (i = 0; i < inst_len; i++)
+ printf(" %02x", inst[i] & 0xFF);
+ printf("\n");
domain_crash_synchronous();
}
- __vmwrite(GUEST_RIP, eip + inst_len);
- store_cpu_user_regs(inst_decoder_regs);
-
- // Only handle "mov" and "movs" instructions!
- if (!strncmp((char *)mmio_inst.i_name, "movz", 4)) {
- if (read_from_mmio(&mmio_inst)) {
- // Send the request and waiting for return value.
- mpci_p->mmio_target = mmio_inst.operand[1] | WZEROEXTEND;
- send_mmio_req(gpa, &mmio_inst, 0, IOREQ_READ, 0);
- return ;
- } else {
- printk("handle_mmio - EXIT: movz error!\n");
- domain_crash_synchronous();
- }
- }
-
- if (!strncmp((char *)mmio_inst.i_name, "movs", 4)) {
+ store_cpu_user_regs(regs);
+ regs->eip += inst_len; /* advance %eip */
+
+ switch (mmio_inst.instr) {
+ case INSTR_MOV:
+ mmio_operands(IOREQ_TYPE_COPY, gpa, &mmio_inst, mpcip, regs);
+ break;
+
+ case INSTR_MOVS:
+ {
+ unsigned long count = GET_REPEAT_COUNT();
+ unsigned long size = mmio_inst.op_size;
+ int sign = regs->eflags & EF_DF ? -1 : 1;
unsigned long addr = 0;
int dir;
+ /* determine non-MMIO address */
if (vm86) {
unsigned long seg;
__vmread(GUEST_ES_SELECTOR, &seg);
- if (((seg << 4) + (inst_decoder_regs->edi & 0xFFFF)) == va) {
+ if (((seg << 4) + (regs->edi & 0xFFFF)) == va) {
dir = IOREQ_WRITE;
__vmread(GUEST_DS_SELECTOR, &seg);
- addr = (seg << 4) + (inst_decoder_regs->esi & 0xFFFF);
+ addr = (seg << 4) + (regs->esi & 0xFFFF);
} else {
dir = IOREQ_READ;
- addr = (seg << 4) + (inst_decoder_regs->edi & 0xFFFF);
+ addr = (seg << 4) + (regs->edi & 0xFFFF);
}
- } else { /* XXX should really look at GDT[ds/es].base too */
- if (va == inst_decoder_regs->edi) {
+ } else {
+ if (va == regs->edi) {
dir = IOREQ_WRITE;
- addr = inst_decoder_regs->esi;
+ addr = regs->esi;
} else {
dir = IOREQ_READ;
- addr = inst_decoder_regs->edi;
+ addr = regs->edi;
}
}
- send_mmio_req(gpa, &mmio_inst, addr, dir, 1);
- return;
- }
-
- if (!strncmp((char *)mmio_inst.i_name, "mov", 3)) {
- long value = 0;
- int size, index;
-
- if (read_from_mmio(&mmio_inst)) {
- // Send the request and waiting for return value.
- mpci_p->mmio_target = mmio_inst.operand[1];
- send_mmio_req(gpa, &mmio_inst, value, IOREQ_READ, 0);
- return;
- } else {
- // Write to MMIO
- if (mmio_inst.operand[0] & IMMEDIATE) {
- value = mmio_inst.immediate;
- } else if (mmio_inst.operand[0] & REGISTER) {
- size = operand_size(mmio_inst.operand[0]);
- index = operand_index(mmio_inst.operand[0]);
- value = get_reg_value(size, index, 0, inst_decoder_regs);
- } else {
- domain_crash_synchronous();
- }
- send_mmio_req(gpa, &mmio_inst, value, IOREQ_WRITE, 0);
- return;
- }
- }
-
- if (!strncmp((char *)mmio_inst.i_name, "stos", 4)) {
- send_mmio_req(gpa, &mmio_inst,
- inst_decoder_regs->eax, IOREQ_WRITE, 0);
- return;
- }
- /* Workaround for cmp instruction */
- if (!strncmp((char *)mmio_inst.i_name, "cmp", 3)) {
- inst_decoder_regs->eflags &= ~X86_EFLAGS_ZF;
- __vmwrite(GUEST_RFLAGS, inst_decoder_regs->eflags);
- return;
- }
-
- domain_crash_synchronous();
+ mpcip->flags = mmio_inst.flags;
+ mpcip->instr = mmio_inst.instr;
+
+ /*
+ * In case of a movs spanning multiple pages, we break the accesses
+ * up into multiple pages (the device model works with non-continguous
+ * physical guest pages). To copy just one page, we adjust %ecx and
+ * do not advance %eip so that the next "rep movs" copies the next page.
+ * Unaligned accesses, for example movsl starting at PGSZ-2, are
+ * turned into a single copy where we handle the overlapping memory
+ * copy ourself. After this copy succeeds, "rep movs" is executed
+ * again.
+ */
+ if ((addr & PAGE_MASK) != ((addr + size - 1) & PAGE_MASK)) {
+ unsigned long value = 0;
+
+ mpcip->flags |= OVERLAP;
+
+ regs->eip -= inst_len; /* do not advance %eip */
+
+ if (dir == IOREQ_WRITE)
+ vmx_copy(&value, addr, size, VMX_COPY_IN);
+ send_mmio_req(IOREQ_TYPE_COPY, gpa, 1, size, value, dir, 0);
+ } else {
+ if ((addr & PAGE_MASK) != ((addr + count * size - 1) & PAGE_MASK)) {
+ regs->eip -= inst_len; /* do not advance %eip */
+
+ if (sign > 0)
+ count = (PAGE_SIZE - (addr & ~PAGE_MASK)) / size;
+ else
+ count = (addr & ~PAGE_MASK) / size;
+ }
+
+ send_mmio_req(IOREQ_TYPE_COPY, gpa, count, size, addr, dir, 1);
+ }
+ break;
+ }
+
+ case INSTR_MOVZ:
+ mmio_operands(IOREQ_TYPE_COPY, gpa, &mmio_inst, mpcip, regs);
+ break;
+
+ case INSTR_STOS:
+ /*
+ * Since the destination is always in (contiguous) mmio space we don't
+ * need to break it up into pages.
+ */
+ mpcip->flags = mmio_inst.flags;
+ mpcip->instr = mmio_inst.instr;
+ send_mmio_req(IOREQ_TYPE_COPY, gpa,
+ GET_REPEAT_COUNT(), mmio_inst.op_size, regs->eax, IOREQ_WRITE, 0);
+ break;
+
+ case INSTR_OR:
+ mmio_operands(IOREQ_TYPE_OR, gpa, &mmio_inst, mpcip, regs);
+ break;
+
+ case INSTR_AND:
+ mmio_operands(IOREQ_TYPE_AND, gpa, &mmio_inst, mpcip, regs);
+ break;
+
+ case INSTR_XOR:
+ mmio_operands(IOREQ_TYPE_XOR, gpa, &mmio_inst, mpcip, regs);
+ break;
+
+ case INSTR_CMP:
+ mmio_operands(IOREQ_TYPE_COPY, gpa, &mmio_inst, mpcip, regs);
+ break;
+
+ case INSTR_TEST:
+ mmio_operands(IOREQ_TYPE_COPY, gpa, &mmio_inst, mpcip, regs);
+ break;
+
+ default:
+ printf("Unhandled MMIO instruction\n");
+ domain_crash_synchronous();
+ }
}
#endif /* CONFIG_VMX */
diff -r 5aae0c8158b9 -r f27205ea60ef xen/common/memory.c
--- a/xen/common/memory.c Sat Sep 3 16:57:54 2005
+++ b/xen/common/memory.c Sat Sep 3 16:58:50 2005
@@ -25,7 +25,8 @@
unsigned long *extent_list,
unsigned int nr_extents,
unsigned int extent_order,
- unsigned int flags)
+ unsigned int flags,
+ int *preempted)
{
struct pfn_info *page;
unsigned long i;
@@ -43,7 +44,10 @@
for ( i = 0; i < nr_extents; i++ )
{
if ( hypercall_preempt_check() )
- return i;
+ {
+ *preempted = 1;
+ return i;
+ }
if ( unlikely((page = alloc_domheap_pages(
d, extent_order, flags)) == NULL) )
@@ -67,7 +71,8 @@
unsigned long *extent_list,
unsigned int nr_extents,
unsigned int extent_order,
- unsigned int flags)
+ unsigned int flags,
+ int *preempted)
{
struct pfn_info *page;
unsigned long i, j, mpfn;
@@ -78,7 +83,10 @@
for ( i = 0; i < nr_extents; i++ )
{
if ( hypercall_preempt_check() )
- return i;
+ {
+ *preempted = 1;
+ return i;
+ }
if ( unlikely(__get_user(mpfn, &extent_list[i]) != 0) )
return i;
@@ -124,7 +132,7 @@
long do_memory_op(int cmd, void *arg)
{
struct domain *d;
- int rc, start_extent, op, flags = 0;
+ int rc, start_extent, op, flags = 0, preempted = 0;
struct xen_memory_reservation reservation;
op = cmd & ((1 << START_EXTENT_SHIFT) - 1);
@@ -165,19 +173,18 @@
reservation.extent_start,
reservation.nr_extents,
reservation.extent_order,
- flags);
+ flags,
+ &preempted);
if ( unlikely(reservation.domid != DOMID_SELF) )
put_domain(d);
rc += start_extent;
- if ( (rc != reservation.nr_extents) && hypercall_preempt_check() )
+ if ( preempted )
return hypercall2_create_continuation(
- __HYPERVISOR_memory_op,
- op | (rc << START_EXTENT_SHIFT),
- arg);
-
+ __HYPERVISOR_memory_op, op | (rc << START_EXTENT_SHIFT), arg);
+
break;
case XENMEM_maximum_ram_page:
diff -r 5aae0c8158b9 -r f27205ea60ef xen/include/asm-x86/vmx.h
--- a/xen/include/asm-x86/vmx.h Sat Sep 3 16:57:54 2005
+++ b/xen/include/asm-x86/vmx.h Sat Sep 3 16:58:50 2005
@@ -471,4 +471,7 @@
void load_cpu_user_regs(struct cpu_user_regs *regs);
void store_cpu_user_regs(struct cpu_user_regs *regs);
+enum { VMX_COPY_IN = 0, VMX_COPY_OUT };
+int vmx_copy(void *buf, unsigned long laddr, int size, int dir);
+
#endif /* __ASM_X86_VMX_H__ */
diff -r 5aae0c8158b9 -r f27205ea60ef xen/include/asm-x86/vmx_platform.h
--- a/xen/include/asm-x86/vmx_platform.h Sat Sep 3 16:57:54 2005
+++ b/xen/include/asm-x86/vmx_platform.h Sat Sep 3 16:58:50 2005
@@ -24,8 +24,7 @@
#include <asm/vmx_virpit.h>
#include <asm/vmx_intercept.h>
-#define MAX_OPERAND_NUM 3
-#define I_NAME_LEN 16
+#define MAX_OPERAND_NUM 2
#define mk_operand(size, index, seg, flag) \
(((size) << 24) | ((index) << 16) | ((seg) << 8) | (flag))
@@ -35,54 +34,60 @@
#define operand_index(operand) \
((operand >> 16) & 0xFF)
- //For instruction.operand[].size
+
+/* for instruction.operand[].size */
#define BYTE 1
#define WORD 2
#define LONG 4
#define QUAD 8
#define BYTE_64 16
- //For instruction.operand[].flag
+/* for instruction.operand[].flag */
#define REGISTER 0x1
#define MEMORY 0x2
#define IMMEDIATE 0x4
-#define WZEROEXTEND 0x8
- //For instruction.flags
+/* for instruction.flags */
#define REPZ 0x1
#define REPNZ 0x2
+#define OVERLAP 0x4
+
+#define INSTR_PIO 1
+#define INSTR_OR 2
+#define INSTR_AND 3
+#define INSTR_XOR 4
+#define INSTR_CMP 5
+#define INSTR_MOV 6
+#define INSTR_MOVS 7
+#define INSTR_MOVZ 8
+#define INSTR_STOS 9
+#define INSTR_TEST 10
struct instruction {
- __s8 i_name[I_NAME_LEN]; //Instruction's name
- __s16 op_size; //The operand's bit size, e.g. 16-bit or 32-bit.
-
- __u64 offset; //The effective address
- //offset = Base + (Index * Scale) + Displacement
-
+ __s8 instr; /* instruction type */
+ __s16 op_size; /* the operand's bit size, e.g. 16-bit or 32-bit */
__u64 immediate;
-
- __u16 seg_sel; //Segmentation selector
-
- __u32 operand[MAX_OPERAND_NUM]; //The order of operand is from AT&T
Assembly
- __s16 op_num; //The operand numbers
-
- __u32 flags; //
+ __u16 seg_sel; /* segmentation selector */
+ __u32 operand[MAX_OPERAND_NUM]; /* order is AT&T assembly */
+ __u32 flags;
};
#define MAX_INST_LEN 32
-struct mi_per_cpu_info
-{
- unsigned long mmio_target;
- struct cpu_user_regs *inst_decoder_regs;
+struct mi_per_cpu_info {
+ int flags;
+ int instr; /* instruction */
+ unsigned long operand[2]; /* operands */
+ unsigned long immediate; /* immediate portion */
+ struct cpu_user_regs *inst_decoder_regs; /* current context */
};
struct virtual_platform_def {
- unsigned long *real_mode_data; /* E820, etc. */
+ unsigned long *real_mode_data; /* E820, etc. */
unsigned long shared_page_va;
struct vmx_virpit_t vmx_pit;
struct vmx_handler_t vmx_handler;
- struct mi_per_cpu_info mpci; /* MMIO */
+ struct mi_per_cpu_info mpci; /* MMIO */
};
extern void handle_mmio(unsigned long, unsigned long);
diff -r 5aae0c8158b9 -r f27205ea60ef xen/include/public/io/ioreq.h
--- a/xen/include/public/io/ioreq.h Sat Sep 3 16:57:54 2005
+++ b/xen/include/public/io/ioreq.h Sat Sep 3 16:58:50 2005
@@ -29,9 +29,17 @@
#define STATE_IORESP_READY 3
#define STATE_IORESP_HOOK 4
-/* VMExit dispatcher should cooperate with instruction decoder to
- prepare this structure and notify service OS and DM by sending
- virq */
+#define IOREQ_TYPE_PIO 0 /* pio */
+#define IOREQ_TYPE_COPY 1 /* mmio ops */
+#define IOREQ_TYPE_AND 2
+#define IOREQ_TYPE_OR 3
+#define IOREQ_TYPE_XOR 4
+
+/*
+ * VMExit dispatcher should cooperate with instruction decoder to
+ * prepare this structure and notify service OS and DM by sending
+ * virq
+ */
typedef struct {
u64 addr; /* physical address */
u64 size; /* size in bytes */
@@ -43,8 +51,8 @@
u8 state:4;
u8 pdata_valid:1; /* if 1, use pdata above */
u8 dir:1; /* 1=read, 0=write */
- u8 port_mm:1; /* 0=portio, 1=mmio */
u8 df:1;
+ u8 type; /* I/O type */
} ioreq_t;
#define MAX_VECTOR 256
diff -r 5aae0c8158b9 -r f27205ea60ef tools/firmware/vmxassist/TODO
--- a/tools/firmware/vmxassist/TODO Sat Sep 3 16:57:54 2005
+++ /dev/null Sat Sep 3 16:58:50 2005
@@ -1,8 +0,0 @@
-
-- Use the VME extensions (interrupt handling)
-
-- Use E820 map in vmxassist instead of cmos hack
-
-- Add ACPI support (Nitin's patch)
-
-
_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog
|
Home