lk/linux-kernel-test
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus

Browse Source 
* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus: (27 commits)
  lguest: use __PAGE_KERNEL instead of _PAGE_KERNEL
  lguest: Use explicit includes rateher than indirect
  lguest: get rid of lg variable assignments
  lguest: change gpte_addr header
  lguest: move changed bitmap to lg_cpu
  lguest: move last_pages to lg_cpu
  lguest: change last_guest to last_cpu
  lguest: change spte_addr header
  lguest: per-vcpu lguest pgdir management
  lguest: make pending notifications per-vcpu
  lguest: makes special fields be per-vcpu
  lguest: per-vcpu lguest task management
  lguest: replace lguest_arch with lg_cpu_arch.
  lguest: make registers per-vcpu
  lguest: make emulate_insn receive a vcpu struct.
  lguest: map_switcher_in_guest() per-vcpu
  lguest: per-vcpu interrupt processing.
  lguest: per-vcpu lguest timers
  lguest: make hypercalls use the vcpu struct
  lguest: make write() operation smp aware
  ...

Manual conflict resolved (maybe even correctly, who knows) in
drivers/lguest/x86/core.c
This commit is contained in:
Linus Torvalds
2008-01-31 09:35:32 +11:00
parent 44c3b59102 84f12e39c8
commit d145c7253c
13 changed files with 576 additions and 460 deletions
Download Patch File Download Diff File Expand all files Collapse all files

49
Documentation/lguest/lguest.c
View File

@@ -79,6 +79,9 @@ static void *guest_base;
/* The maximum guest physical address allowed, and maximum possible. */ /* The maximum guest physical address allowed, and maximum possible. */
static unsigned long guest_limit, guest_max; static unsigned long guest_limit, guest_max;
/* a per-cpu variable indicating whose vcpu is currently running */
static unsigned int __thread cpu_id;
/* This is our list of devices. */ /* This is our list of devices. */
struct device_list struct device_list
{ {
@@ -153,6 +156,9 @@ struct virtqueue
void (*handle_output)(int fd, struct virtqueue *me); void (*handle_output)(int fd, struct virtqueue *me);
}; };
/* Remember the arguments to the program so we can "reboot" */
static char **main_args;
/* Since guest is UP and we don't run at the same time, we don't need barriers. /* Since guest is UP and we don't run at the same time, we don't need barriers.
* But I include them in the code in case others copy it. */ * But I include them in the code in case others copy it. */
#define wmb() #define wmb()
@@ -554,7 +560,7 @@ static void wake_parent(int pipefd, int lguest_fd)
else else
FD_CLR(-fd - 1, &devices.infds); FD_CLR(-fd - 1, &devices.infds);
} else /* Send LHREQ_BREAK command. */ } else /* Send LHREQ_BREAK command. */
write(lguest_fd, args, sizeof(args)); pwrite(lguest_fd, args, sizeof(args), cpu_id);
} }
} }
@@ -1489,7 +1495,9 @@ static void setup_block_file(const char *filename)
/* Create stack for thread and run it */ /* Create stack for thread and run it */
stack = malloc(32768); stack = malloc(32768);
if (clone(io_thread, stack + 32768, CLONE_VM, dev) == -1) /* SIGCHLD - We dont "wait" for our cloned thread, so prevent it from
* becoming a zombie. */
if (clone(io_thread, stack + 32768, CLONE_VM | SIGCHLD, dev) == -1)
err(1, "Creating clone"); err(1, "Creating clone");
/* We don't need to keep the I/O thread's end of the pipes open. */ /* We don't need to keep the I/O thread's end of the pipes open. */
@@ -1499,7 +1507,21 @@ static void setup_block_file(const char *filename)
verbose("device %u: virtblock %llu sectors\n", verbose("device %u: virtblock %llu sectors\n",
devices.device_num, cap); devices.device_num, cap);
} }
/* That's the end of device setup. */ /* That's the end of device setup. :*/
/* Reboot */
static void __attribute__((noreturn)) restart_guest(void)
{
unsigned int i;
/* Closing pipes causes the waker thread and io_threads to die, and
* closing /dev/lguest cleans up the Guest. Since we don't track all
* open fds, we simply close everything beyond stderr. */
for (i = 3; i < FD_SETSIZE; i++)
close(i);
execv(main_args[0], main_args);
err(1, "Could not exec %s", main_args[0]);
}
/*L:220 Finally we reach the core of the Launcher, which runs the Guest, serves /*L:220 Finally we reach the core of the Launcher, which runs the Guest, serves
* its input and output, and finally, lays it to rest. */ * its input and output, and finally, lays it to rest. */
@@ -1511,7 +1533,8 @@ static void __attribute__((noreturn)) run_guest(int lguest_fd)
int readval; int readval;
/* We read from the /dev/lguest device to run the Guest. */ /* We read from the /dev/lguest device to run the Guest. */
readval = read(lguest_fd, &notify_addr, sizeof(notify_addr)); readval = pread(lguest_fd, &notify_addr,
sizeof(notify_addr), cpu_id);
/* One unsigned long means the Guest did HCALL_NOTIFY */ /* One unsigned long means the Guest did HCALL_NOTIFY */
if (readval == sizeof(notify_addr)) { if (readval == sizeof(notify_addr)) {
@@ -1521,16 +1544,23 @@ static void __attribute__((noreturn)) run_guest(int lguest_fd)
/* ENOENT means the Guest died. Reading tells us why. */ /* ENOENT means the Guest died. Reading tells us why. */
} else if (errno == ENOENT) { } else if (errno == ENOENT) {
char reason[1024] = { 0 }; char reason[1024] = { 0 };
read(lguest_fd, reason, sizeof(reason)-1); pread(lguest_fd, reason, sizeof(reason)-1, cpu_id);
errx(1, "%s", reason); errx(1, "%s", reason);
/* ERESTART means that we need to reboot the guest */
} else if (errno == ERESTART) {
restart_guest();
/* EAGAIN means the Waker wanted us to look at some input. /* EAGAIN means the Waker wanted us to look at some input.
* Anything else means a bug or incompatible change. */ * Anything else means a bug or incompatible change. */
} else if (errno != EAGAIN) } else if (errno != EAGAIN)
err(1, "Running guest failed"); err(1, "Running guest failed");
/* Only service input on thread for CPU 0. */
if (cpu_id != 0)
continue;
/* Service input, then unset the BREAK to release the Waker. */ /* Service input, then unset the BREAK to release the Waker. */
handle_input(lguest_fd); handle_input(lguest_fd);
if (write(lguest_fd, args, sizeof(args)) < 0) if (pwrite(lguest_fd, args, sizeof(args), cpu_id) < 0)
err(1, "Resetting break"); err(1, "Resetting break");
} }
} }
@@ -1571,6 +1601,12 @@ int main(int argc, char *argv[])
/* If they specify an initrd file to load. */ /* If they specify an initrd file to load. */
const char *initrd_name = NULL; const char *initrd_name = NULL;
/* Save the args: we "reboot" by execing ourselves again. */
main_args = argv;
/* We don't "wait" for the children, so prevent them from becoming
* zombies. */
signal(SIGCHLD, SIG_IGN);
/* First we initialize the device list. Since console and network /* First we initialize the device list. Since console and network
* device receive input from a file descriptor, we keep an fdset * device receive input from a file descriptor, we keep an fdset
* (infds) and the maximum fd number (max_infd) with the head of the * (infds) and the maximum fd number (max_infd) with the head of the
@@ -1582,6 +1618,7 @@ int main(int argc, char *argv[])
devices.lastdev = &devices.dev; devices.lastdev = &devices.dev;
devices.next_irq = 1; devices.next_irq = 1;
cpu_id = 0;
/* We need to know how much memory so we can set up the device /* We need to know how much memory so we can set up the device
* descriptor and memory pages for the devices as we parse the command * descriptor and memory pages for the devices as we parse the command
* line. So we quickly look through the arguments to find the amount * line. So we quickly look through the arguments to find the amount

11
arch/x86/lguest/boot.c
View File

@@ -67,6 +67,7 @@
#include <asm/mce.h> #include <asm/mce.h>
#include <asm/io.h> #include <asm/io.h>
#include <asm/i387.h> #include <asm/i387.h>
#include <asm/reboot.h> /* for struct machine_ops */
/*G:010 Welcome to the Guest! /*G:010 Welcome to the Guest!
* *
@@ -813,7 +814,7 @@ static void lguest_safe_halt(void)
* rather than virtual addresses, so we use __pa() here. */ * rather than virtual addresses, so we use __pa() here. */
static void lguest_power_off(void) static void lguest_power_off(void)
{ {
hcall(LHCALL_CRASH, __pa("Power down"), 0, 0); hcall(LHCALL_SHUTDOWN, __pa("Power down"), LGUEST_SHUTDOWN_POWEROFF, 0);
} }
/* /*
@@ -823,7 +824,7 @@ static void lguest_power_off(void)
*/ */
static int lguest_panic(struct notifier_block *nb, unsigned long l, void *p) static int lguest_panic(struct notifier_block *nb, unsigned long l, void *p)
{ {
hcall(LHCALL_CRASH, __pa(p), 0, 0); hcall(LHCALL_SHUTDOWN, __pa(p), LGUEST_SHUTDOWN_POWEROFF, 0);
/* The hcall won't return, but to keep gcc happy, we're "done". */ /* The hcall won't return, but to keep gcc happy, we're "done". */
return NOTIFY_DONE; return NOTIFY_DONE;
} }
@@ -927,6 +928,11 @@ static unsigned lguest_patch(u8 type, u16 clobber, void *ibuf,
return insn_len; return insn_len;
} }
static void lguest_restart(char *reason)
{
hcall(LHCALL_SHUTDOWN, __pa(reason), LGUEST_SHUTDOWN_RESTART, 0);
}
/*G:030 Once we get to lguest_init(), we know we're a Guest. The pv_ops /*G:030 Once we get to lguest_init(), we know we're a Guest. The pv_ops
* structures in the kernel provide points for (almost) every routine we have * structures in the kernel provide points for (almost) every routine we have
* to override to avoid privileged instructions. */ * to override to avoid privileged instructions. */
@@ -1060,6 +1066,7 @@ __init void lguest_init(void)
* the Guest routine to power off. */ * the Guest routine to power off. */
pm_power_off = lguest_power_off; pm_power_off = lguest_power_off;
machine_ops.restart = lguest_restart;
/* Now we're set up, call start_kernel() in init/main.c and we proceed /* Now we're set up, call start_kernel() in init/main.c and we proceed
* to boot as normal. It never returns. */ * to boot as normal. It never returns. */
start_kernel(); start_kernel();

2
drivers/Makefile
View File

@@ -72,7 +72,7 @@ obj-$(CONFIG_ISDN) += isdn/
obj-$(CONFIG_EDAC) += edac/ obj-$(CONFIG_EDAC) += edac/
obj-$(CONFIG_MCA) += mca/ obj-$(CONFIG_MCA) += mca/
obj-$(CONFIG_EISA) += eisa/ obj-$(CONFIG_EISA) += eisa/
obj-$(CONFIG_LGUEST_GUEST) += lguest/ obj-y += lguest/
obj-$(CONFIG_CPU_FREQ) += cpufreq/ obj-$(CONFIG_CPU_FREQ) += cpufreq/
obj-$(CONFIG_CPU_IDLE) += cpuidle/ obj-$(CONFIG_CPU_IDLE) += cpuidle/
obj-$(CONFIG_MMC) += mmc/ obj-$(CONFIG_MMC) += mmc/

46
drivers/lguest/core.c
View File

@@ -151,43 +151,43 @@ int lguest_address_ok(const struct lguest *lg,
/* This routine copies memory from the Guest. Here we can see how useful the /* This routine copies memory from the Guest. Here we can see how useful the
* kill_lguest() routine we met in the Launcher can be: we return a random * kill_lguest() routine we met in the Launcher can be: we return a random
* value (all zeroes) instead of needing to return an error. */ * value (all zeroes) instead of needing to return an error. */
void __lgread(struct lguest *lg, void *b, unsigned long addr, unsigned bytes) void __lgread(struct lg_cpu *cpu, void *b, unsigned long addr, unsigned bytes)
{ {
if (!lguest_address_ok(lg, addr, bytes) if (!lguest_address_ok(cpu->lg, addr, bytes)
|| copy_from_user(b, lg->mem_base + addr, bytes) != 0) { || copy_from_user(b, cpu->lg->mem_base + addr, bytes) != 0) {
/* copy_from_user should do this, but as we rely on it... */ /* copy_from_user should do this, but as we rely on it... */
memset(b, 0, bytes); memset(b, 0, bytes);
kill_guest(lg, "bad read address %#lx len %u", addr, bytes); kill_guest(cpu, "bad read address %#lx len %u", addr, bytes);
} }
} }
/* This is the write (copy into guest) version. */ /* This is the write (copy into guest) version. */
void __lgwrite(struct lguest *lg, unsigned long addr, const void *b, void __lgwrite(struct lg_cpu *cpu, unsigned long addr, const void *b,
unsigned bytes) unsigned bytes)
{ {
if (!lguest_address_ok(lg, addr, bytes) if (!lguest_address_ok(cpu->lg, addr, bytes)
|| copy_to_user(lg->mem_base + addr, b, bytes) != 0) || copy_to_user(cpu->lg->mem_base + addr, b, bytes) != 0)
kill_guest(lg, "bad write address %#lx len %u", addr, bytes); kill_guest(cpu, "bad write address %#lx len %u", addr, bytes);
} }
/*:*/ /*:*/
/*H:030 Let's jump straight to the the main loop which runs the Guest. /*H:030 Let's jump straight to the the main loop which runs the Guest.
* Remember, this is called by the Launcher reading /dev/lguest, and we keep * Remember, this is called by the Launcher reading /dev/lguest, and we keep
* going around and around until something interesting happens. */ * going around and around until something interesting happens. */
int run_guest(struct lguest *lg, unsigned long __user *user) int run_guest(struct lg_cpu *cpu, unsigned long __user *user)
{ {
/* We stop running once the Guest is dead. */ /* We stop running once the Guest is dead. */
while (!lg->dead) { while (!cpu->lg->dead) {
/* First we run any hypercalls the Guest wants done. */ /* First we run any hypercalls the Guest wants done. */
if (lg->hcall) if (cpu->hcall)
do_hypercalls(lg); do_hypercalls(cpu);
/* It's possible the Guest did a NOTIFY hypercall to the /* It's possible the Guest did a NOTIFY hypercall to the
* Launcher, in which case we return from the read() now. */ * Launcher, in which case we return from the read() now. */
if (lg->pending_notify) { if (cpu->pending_notify) {
if (put_user(lg->pending_notify, user)) if (put_user(cpu->pending_notify, user))
return -EFAULT; return -EFAULT;
return sizeof(lg->pending_notify); return sizeof(cpu->pending_notify);
} }
/* Check for signals */ /* Check for signals */
@@ -195,13 +195,13 @@ int run_guest(struct lguest *lg, unsigned long __user *user)
return -ERESTARTSYS; return -ERESTARTSYS;
/* If Waker set break_out, return to Launcher. */ /* If Waker set break_out, return to Launcher. */
if (lg->break_out) if (cpu->break_out)
return -EAGAIN; return -EAGAIN;
/* Check if there are any interrupts which can be delivered /* Check if there are any interrupts which can be delivered
* now: if so, this sets up the hander to be executed when we * now: if so, this sets up the hander to be executed when we
* next run the Guest. */ * next run the Guest. */
maybe_do_interrupt(lg); maybe_do_interrupt(cpu);
/* All long-lived kernel loops need to check with this horrible /* All long-lived kernel loops need to check with this horrible
* thing called the freezer. If the Host is trying to suspend, * thing called the freezer. If the Host is trying to suspend,
@@ -210,12 +210,12 @@ int run_guest(struct lguest *lg, unsigned long __user *user)
/* Just make absolutely sure the Guest is still alive. One of /* Just make absolutely sure the Guest is still alive. One of
* those hypercalls could have been fatal, for example. */ * those hypercalls could have been fatal, for example. */
if (lg->dead) if (cpu->lg->dead)
break; break;
/* If the Guest asked to be stopped, we sleep. The Guest's /* If the Guest asked to be stopped, we sleep. The Guest's
* clock timer or LHCALL_BREAK from the Waker will wake us. */ * clock timer or LHCALL_BREAK from the Waker will wake us. */
if (lg->halted) { if (cpu->halted) {
set_current_state(TASK_INTERRUPTIBLE); set_current_state(TASK_INTERRUPTIBLE);
schedule(); schedule();
continue; continue;
@@ -226,15 +226,17 @@ int run_guest(struct lguest *lg, unsigned long __user *user)
local_irq_disable(); local_irq_disable();
/* Actually run the Guest until something happens. */ /* Actually run the Guest until something happens. */
lguest_arch_run_guest(lg); lguest_arch_run_guest(cpu);
/* Now we're ready to be interrupted or moved to other CPUs */ /* Now we're ready to be interrupted or moved to other CPUs */
local_irq_enable(); local_irq_enable();
/* Now we deal with whatever happened to the Guest. */ /* Now we deal with whatever happened to the Guest. */
lguest_arch_handle_trap(lg); lguest_arch_handle_trap(cpu);
} }
if (cpu->lg->dead == ERR_PTR(-ERESTART))
return -ERESTART;
/* The Guest is dead => "No such file or directory" */ /* The Guest is dead => "No such file or directory" */
return -ENOENT; return -ENOENT;
} }
@@ -253,7 +255,7 @@ static int __init init(void)
/* Lguest can't run under Xen, VMI or itself. It does Tricky Stuff. */ /* Lguest can't run under Xen, VMI or itself. It does Tricky Stuff. */
if (paravirt_enabled()) { if (paravirt_enabled()) {
printk("lguest is afraid of %s\n", pv_info.name); printk("lguest is afraid of being a guest\n");
return -EPERM; return -EPERM;
} }

106
drivers/lguest/hypercalls.c
View File

@@ -23,13 +23,14 @@
#include <linux/uaccess.h> #include <linux/uaccess.h>
#include <linux/syscalls.h> #include <linux/syscalls.h>
#include <linux/mm.h> #include <linux/mm.h>
#include <linux/ktime.h>
#include <asm/page.h> #include <asm/page.h>
#include <asm/pgtable.h> #include <asm/pgtable.h>
#include "lg.h" #include "lg.h"
/*H:120 This is the core hypercall routine: where the Guest gets what it wants. /*H:120 This is the core hypercall routine: where the Guest gets what it wants.
* Or gets killed. Or, in the case of LHCALL_CRASH, both. */ * Or gets killed. Or, in the case of LHCALL_CRASH, both. */
static void do_hcall(struct lguest *lg, struct hcall_args *args) static void do_hcall(struct lg_cpu *cpu, struct hcall_args *args)
{ {
switch (args->arg0) { switch (args->arg0) {
case LHCALL_FLUSH_ASYNC: case LHCALL_FLUSH_ASYNC:
@@ -39,60 +40,62 @@ static void do_hcall(struct lguest *lg, struct hcall_args *args)
case LHCALL_LGUEST_INIT: case LHCALL_LGUEST_INIT:
/* You can't get here unless you're already initialized. Don't /* You can't get here unless you're already initialized. Don't
* do that. */ * do that. */
kill_guest(lg, "already have lguest_data"); kill_guest(cpu, "already have lguest_data");
break; break;
case LHCALL_CRASH: { case LHCALL_SHUTDOWN: {
/* Crash is such a trivial hypercall that we do it in four /* Shutdown is such a trivial hypercall that we do it in four
* lines right here. */ * lines right here. */
char msg[128]; char msg[128];
/* If the lgread fails, it will call kill_guest() itself; the /* If the lgread fails, it will call kill_guest() itself; the
* kill_guest() with the message will be ignored. */ * kill_guest() with the message will be ignored. */
__lgread(lg, msg, args->arg1, sizeof(msg)); __lgread(cpu, msg, args->arg1, sizeof(msg));
msg[sizeof(msg)-1] = '\0'; msg[sizeof(msg)-1] = '\0';
kill_guest(lg, "CRASH: %s", msg); kill_guest(cpu, "CRASH: %s", msg);
if (args->arg2 == LGUEST_SHUTDOWN_RESTART)
cpu->lg->dead = ERR_PTR(-ERESTART);
break; break;
} }
case LHCALL_FLUSH_TLB: case LHCALL_FLUSH_TLB:
/* FLUSH_TLB comes in two flavors, depending on the /* FLUSH_TLB comes in two flavors, depending on the
* argument: */ * argument: */
if (args->arg1) if (args->arg1)
guest_pagetable_clear_all(lg); guest_pagetable_clear_all(cpu);
else else
guest_pagetable_flush_user(lg); guest_pagetable_flush_user(cpu);
break; break;
/* All these calls simply pass the arguments through to the right /* All these calls simply pass the arguments through to the right
* routines. */ * routines. */
case LHCALL_NEW_PGTABLE: case LHCALL_NEW_PGTABLE:
guest_new_pagetable(lg, args->arg1); guest_new_pagetable(cpu, args->arg1);
break; break;
case LHCALL_SET_STACK: case LHCALL_SET_STACK:
guest_set_stack(lg, args->arg1, args->arg2, args->arg3); guest_set_stack(cpu, args->arg1, args->arg2, args->arg3);
break; break;
case LHCALL_SET_PTE: case LHCALL_SET_PTE:
guest_set_pte(lg, args->arg1, args->arg2, __pte(args->arg3)); guest_set_pte(cpu, args->arg1, args->arg2, __pte(args->arg3));
break; break;
case LHCALL_SET_PMD: case LHCALL_SET_PMD:
guest_set_pmd(lg, args->arg1, args->arg2); guest_set_pmd(cpu->lg, args->arg1, args->arg2);
break; break;
case LHCALL_SET_CLOCKEVENT: case LHCALL_SET_CLOCKEVENT:
guest_set_clockevent(lg, args->arg1); guest_set_clockevent(cpu, args->arg1);
break; break;
case LHCALL_TS: case LHCALL_TS:
/* This sets the TS flag, as we saw used in run_guest(). */ /* This sets the TS flag, as we saw used in run_guest(). */
lg->ts = args->arg1; cpu->ts = args->arg1;
break; break;
case LHCALL_HALT: case LHCALL_HALT:
/* Similarly, this sets the halted flag for run_guest(). */ /* Similarly, this sets the halted flag for run_guest(). */
lg->halted = 1; cpu->halted = 1;
break; break;
case LHCALL_NOTIFY: case LHCALL_NOTIFY:
lg->pending_notify = args->arg1; cpu->pending_notify = args->arg1;
break; break;
default: default:
/* It should be an architecture-specific hypercall. */ /* It should be an architecture-specific hypercall. */
if (lguest_arch_do_hcall(lg, args)) if (lguest_arch_do_hcall(cpu, args))
kill_guest(lg, "Bad hypercall %li\n", args->arg0); kill_guest(cpu, "Bad hypercall %li\n", args->arg0);
} }
} }
/*:*/ /*:*/
@@ -104,13 +107,13 @@ static void do_hcall(struct lguest *lg, struct hcall_args *args)
* Guest put them in the ring, but we also promise the Guest that they will * Guest put them in the ring, but we also promise the Guest that they will
* happen before any normal hypercall (which is why we check this before * happen before any normal hypercall (which is why we check this before
* checking for a normal hcall). */ * checking for a normal hcall). */
static void do_async_hcalls(struct lguest *lg) static void do_async_hcalls(struct lg_cpu *cpu)
{ {
unsigned int i; unsigned int i;
u8 st[LHCALL_RING_SIZE]; u8 st[LHCALL_RING_SIZE];
/* For simplicity, we copy the entire call status array in at once. */ /* For simplicity, we copy the entire call status array in at once. */
if (copy_from_user(&st, &lg->lguest_data->hcall_status, sizeof(st))) if (copy_from_user(&st, &cpu->lg->lguest_data->hcall_status, sizeof(st)))
return; return;
/* We process "struct lguest_data"s hcalls[] ring once. */ /* We process "struct lguest_data"s hcalls[] ring once. */
@@ -119,7 +122,7 @@ static void do_async_hcalls(struct lguest *lg)
/* We remember where we were up to from last time. This makes /* We remember where we were up to from last time. This makes
* sure that the hypercalls are done in the order the Guest * sure that the hypercalls are done in the order the Guest
* places them in the ring. */ * places them in the ring. */
unsigned int n = lg->next_hcall; unsigned int n = cpu->next_hcall;
/* 0xFF means there's no call here (yet). */ /* 0xFF means there's no call here (yet). */
if (st[n] == 0xFF) if (st[n] == 0xFF)
@@ -127,65 +130,65 @@ static void do_async_hcalls(struct lguest *lg)
/* OK, we have hypercall. Increment the "next_hcall" cursor, /* OK, we have hypercall. Increment the "next_hcall" cursor,
* and wrap back to 0 if we reach the end. */ * and wrap back to 0 if we reach the end. */
if (++lg->next_hcall == LHCALL_RING_SIZE) if (++cpu->next_hcall == LHCALL_RING_SIZE)
lg->next_hcall = 0; cpu->next_hcall = 0;
/* Copy the hypercall arguments into a local copy of /* Copy the hypercall arguments into a local copy of
* the hcall_args struct. */ * the hcall_args struct. */
if (copy_from_user(&args, &lg->lguest_data->hcalls[n], if (copy_from_user(&args, &cpu->lg->lguest_data->hcalls[n],
sizeof(struct hcall_args))) { sizeof(struct hcall_args))) {
kill_guest(lg, "Fetching async hypercalls"); kill_guest(cpu, "Fetching async hypercalls");
break; break;
} }
/* Do the hypercall, same as a normal one. */ /* Do the hypercall, same as a normal one. */
do_hcall(lg, &args); do_hcall(cpu, &args);
/* Mark the hypercall done. */ /* Mark the hypercall done. */
if (put_user(0xFF, &lg->lguest_data->hcall_status[n])) { if (put_user(0xFF, &cpu->lg->lguest_data->hcall_status[n])) {
kill_guest(lg, "Writing result for async hypercall"); kill_guest(cpu, "Writing result for async hypercall");
break; break;
} }
/* Stop doing hypercalls if they want to notify the Launcher: /* Stop doing hypercalls if they want to notify the Launcher:
* it needs to service this first. */ * it needs to service this first. */
if (lg->pending_notify) if (cpu->pending_notify)
break; break;
} }
} }
/* Last of all, we look at what happens first of all. The very first time the /* Last of all, we look at what happens first of all. The very first time the
* Guest makes a hypercall, we end up here to set things up: */ * Guest makes a hypercall, we end up here to set things up: */
static void initialize(struct lguest *lg) static void initialize(struct lg_cpu *cpu)
{ {
/* You can't do anything until you're initialized. The Guest knows the /* You can't do anything until you're initialized. The Guest knows the
* rules, so we're unforgiving here. */ * rules, so we're unforgiving here. */
if (lg->hcall->arg0 != LHCALL_LGUEST_INIT) { if (cpu->hcall->arg0 != LHCALL_LGUEST_INIT) {
kill_guest(lg, "hypercall %li before INIT", lg->hcall->arg0); kill_guest(cpu, "hypercall %li before INIT", cpu->hcall->arg0);
return; return;
} }
if (lguest_arch_init_hypercalls(lg)) if (lguest_arch_init_hypercalls(cpu))
kill_guest(lg, "bad guest page %p", lg->lguest_data); kill_guest(cpu, "bad guest page %p", cpu->lg->lguest_data);
/* The Guest tells us where we're not to deliver interrupts by putting /* The Guest tells us where we're not to deliver interrupts by putting
* the range of addresses into "struct lguest_data". */ * the range of addresses into "struct lguest_data". */
if (get_user(lg->noirq_start, &lg->lguest_data->noirq_start) if (get_user(cpu->lg->noirq_start, &cpu->lg->lguest_data->noirq_start)
|| get_user(lg->noirq_end, &lg->lguest_data->noirq_end)) || get_user(cpu->lg->noirq_end, &cpu->lg->lguest_data->noirq_end))
kill_guest(lg, "bad guest page %p", lg->lguest_data); kill_guest(cpu, "bad guest page %p", cpu->lg->lguest_data);
/* We write the current time into the Guest's data page once so it can /* We write the current time into the Guest's data page once so it can
* set its clock. */ * set its clock. */
write_timestamp(lg); write_timestamp(cpu);
/* page_tables.c will also do some setup. */ /* page_tables.c will also do some setup. */
page_table_guest_data_init(lg); page_table_guest_data_init(cpu);
/* This is the one case where the above accesses might have been the /* This is the one case where the above accesses might have been the
* first write to a Guest page. This may have caused a copy-on-write * first write to a Guest page. This may have caused a copy-on-write
* fault, but the old page might be (read-only) in the Guest * fault, but the old page might be (read-only) in the Guest
* pagetable. */ * pagetable. */
guest_pagetable_clear_all(lg); guest_pagetable_clear_all(cpu);
} }
/*H:100 /*H:100
@@ -194,27 +197,27 @@ static void initialize(struct lguest *lg)
* Remember from the Guest, hypercalls come in two flavors: normal and * Remember from the Guest, hypercalls come in two flavors: normal and
* asynchronous. This file handles both of types. * asynchronous. This file handles both of types.
*/ */
void do_hypercalls(struct lguest *lg) void do_hypercalls(struct lg_cpu *cpu)
{ {
/* Not initialized yet? This hypercall must do it. */ /* Not initialized yet? This hypercall must do it. */
if (unlikely(!lg->lguest_data)) { if (unlikely(!cpu->lg->lguest_data)) {
/* Set up the "struct lguest_data" */ /* Set up the "struct lguest_data" */
initialize(lg); initialize(cpu);
/* Hcall is done. */ /* Hcall is done. */
lg->hcall = NULL; cpu->hcall = NULL;
return; return;
} }
/* The Guest has initialized. /* The Guest has initialized.
* *
* Look in the hypercall ring for the async hypercalls: */ * Look in the hypercall ring for the async hypercalls: */
do_async_hcalls(lg); do_async_hcalls(cpu);
/* If we stopped reading the hypercall ring because the Guest did a /* If we stopped reading the hypercall ring because the Guest did a
* NOTIFY to the Launcher, we want to return now. Otherwise we do * NOTIFY to the Launcher, we want to return now. Otherwise we do
* the hypercall. */ * the hypercall. */
if (!lg->pending_notify) { if (!cpu->pending_notify) {
do_hcall(lg, lg->hcall); do_hcall(cpu, cpu->hcall);
/* Tricky point: we reset the hcall pointer to mark the /* Tricky point: we reset the hcall pointer to mark the
* hypercall as "done". We use the hcall pointer rather than * hypercall as "done". We use the hcall pointer rather than
* the trap number to indicate a hypercall is pending. * the trap number to indicate a hypercall is pending.
@@ -225,16 +228,17 @@ void do_hypercalls(struct lguest *lg)
* Launcher, the run_guest() loop will exit without running the * Launcher, the run_guest() loop will exit without running the
* Guest. When it comes back it would try to re-run the * Guest. When it comes back it would try to re-run the
* hypercall. */ * hypercall. */
lg->hcall = NULL; cpu->hcall = NULL;
} }
} }
/* This routine supplies the Guest with time: it's used for wallclock time at /* This routine supplies the Guest with time: it's used for wallclock time at
* initial boot and as a rough time source if the TSC isn't available. */ * initial boot and as a rough time source if the TSC isn't available. */
void write_timestamp(struct lguest *lg) void write_timestamp(struct lg_cpu *cpu)
{ {
struct timespec now; struct timespec now;
ktime_get_real_ts(&now); ktime_get_real_ts(&now);
if (copy_to_user(&lg->lguest_data->time, &now, sizeof(struct timespec))) if (copy_to_user(&cpu->lg->lguest_data->time,
kill_guest(lg, "Writing timestamp"); &now, sizeof(struct timespec)))
kill_guest(cpu, "Writing timestamp");
} }

149
drivers/lguest/interrupts_and_traps.c
View File

@@ -41,11 +41,11 @@ static int idt_present(u32 lo, u32 hi)
/* We need a helper to "push" a value onto the Guest's stack, since that's a /* We need a helper to "push" a value onto the Guest's stack, since that's a
* big part of what delivering an interrupt does. */ * big part of what delivering an interrupt does. */
static void push_guest_stack(struct lguest *lg, unsigned long *gstack, u32 val) static void push_guest_stack(struct lg_cpu *cpu, unsigned long *gstack, u32 val)
{ {
/* Stack grows upwards: move stack then write value. */ /* Stack grows upwards: move stack then write value. */
*gstack -= 4; *gstack -= 4;
lgwrite(lg, *gstack, u32, val); lgwrite(cpu, *gstack, u32, val);
} }
/*H:210 The set_guest_interrupt() routine actually delivers the interrupt or /*H:210 The set_guest_interrupt() routine actually delivers the interrupt or
@@ -60,7 +60,7 @@ static void push_guest_stack(struct lguest *lg, unsigned long *gstack, u32 val)
* We set up the stack just like the CPU does for a real interrupt, so it's * We set up the stack just like the CPU does for a real interrupt, so it's
* identical for the Guest (and the standard "iret" instruction will undo * identical for the Guest (and the standard "iret" instruction will undo
* it). */ * it). */
static void set_guest_interrupt(struct lguest *lg, u32 lo, u32 hi, int has_err) static void set_guest_interrupt(struct lg_cpu *cpu, u32 lo, u32 hi, int has_err)
{ {
unsigned long gstack, origstack; unsigned long gstack, origstack;
u32 eflags, ss, irq_enable; u32 eflags, ss, irq_enable;
@@ -69,59 +69,59 @@ static void set_guest_interrupt(struct lguest *lg, u32 lo, u32 hi, int has_err)
/* There are two cases for interrupts: one where the Guest is already /* There are two cases for interrupts: one where the Guest is already
* in the kernel, and a more complex one where the Guest is in * in the kernel, and a more complex one where the Guest is in
* userspace. We check the privilege level to find out. */ * userspace. We check the privilege level to find out. */
if ((lg->regs->ss&0x3) != GUEST_PL) { if ((cpu->regs->ss&0x3) != GUEST_PL) {
/* The Guest told us their kernel stack with the SET_STACK /* The Guest told us their kernel stack with the SET_STACK
* hypercall: both the virtual address and the segment */ * hypercall: both the virtual address and the segment */
virtstack = lg->esp1; virtstack = cpu->esp1;
ss = lg->ss1; ss = cpu->ss1;
origstack = gstack = guest_pa(lg, virtstack); origstack = gstack = guest_pa(cpu, virtstack);
/* We push the old stack segment and pointer onto the new /* We push the old stack segment and pointer onto the new
* stack: when the Guest does an "iret" back from the interrupt * stack: when the Guest does an "iret" back from the interrupt
* handler the CPU will notice they're dropping privilege * handler the CPU will notice they're dropping privilege
* levels and expect these here. */ * levels and expect these here. */
push_guest_stack(lg, &gstack, lg->regs->ss); push_guest_stack(cpu, &gstack, cpu->regs->ss);
push_guest_stack(lg, &gstack, lg->regs->esp); push_guest_stack(cpu, &gstack, cpu->regs->esp);
} else { } else {
/* We're staying on the same Guest (kernel) stack. */ /* We're staying on the same Guest (kernel) stack. */
virtstack = lg->regs->esp; virtstack = cpu->regs->esp;
ss = lg->regs->ss; ss = cpu->regs->ss;
origstack = gstack = guest_pa(lg, virtstack); origstack = gstack = guest_pa(cpu, virtstack);
} }
/* Remember that we never let the Guest actually disable interrupts, so /* Remember that we never let the Guest actually disable interrupts, so
* the "Interrupt Flag" bit is always set. We copy that bit from the * the "Interrupt Flag" bit is always set. We copy that bit from the
* Guest's "irq_enabled" field into the eflags word: we saw the Guest * Guest's "irq_enabled" field into the eflags word: we saw the Guest
* copy it back in "lguest_iret". */ * copy it back in "lguest_iret". */
eflags = lg->regs->eflags; eflags = cpu->regs->eflags;
if (get_user(irq_enable, &lg->lguest_data->irq_enabled) == 0 if (get_user(irq_enable, &cpu->lg->lguest_data->irq_enabled) == 0
&& !(irq_enable & X86_EFLAGS_IF)) && !(irq_enable & X86_EFLAGS_IF))
eflags &= ~X86_EFLAGS_IF; eflags &= ~X86_EFLAGS_IF;
/* An interrupt is expected to push three things on the stack: the old /* An interrupt is expected to push three things on the stack: the old
* "eflags" word, the old code segment, and the old instruction * "eflags" word, the old code segment, and the old instruction
* pointer. */ * pointer. */
push_guest_stack(lg, &gstack, eflags); push_guest_stack(cpu, &gstack, eflags);
push_guest_stack(lg, &gstack, lg->regs->cs); push_guest_stack(cpu, &gstack, cpu->regs->cs);
push_guest_stack(lg, &gstack, lg->regs->eip); push_guest_stack(cpu, &gstack, cpu->regs->eip);
/* For the six traps which supply an error code, we push that, too. */ /* For the six traps which supply an error code, we push that, too. */
if (has_err) if (has_err)
push_guest_stack(lg, &gstack, lg->regs->errcode); push_guest_stack(cpu, &gstack, cpu->regs->errcode);
/* Now we've pushed all the old state, we change the stack, the code /* Now we've pushed all the old state, we change the stack, the code
* segment and the address to execute. */ * segment and the address to execute. */
lg->regs->ss = ss; cpu->regs->ss = ss;
lg->regs->esp = virtstack + (gstack - origstack); cpu->regs->esp = virtstack + (gstack - origstack);
lg->regs->cs = (__KERNEL_CS|GUEST_PL); cpu->regs->cs = (__KERNEL_CS|GUEST_PL);
lg->regs->eip = idt_address(lo, hi); cpu->regs->eip = idt_address(lo, hi);
/* There are two kinds of interrupt handlers: 0xE is an "interrupt /* There are two kinds of interrupt handlers: 0xE is an "interrupt
* gate" which expects interrupts to be disabled on entry. */ * gate" which expects interrupts to be disabled on entry. */
if (idt_type(lo, hi) == 0xE) if (idt_type(lo, hi) == 0xE)
if (put_user(0, &lg->lguest_data->irq_enabled)) if (put_user(0, &cpu->lg->lguest_data->irq_enabled))
kill_guest(lg, "Disabling interrupts"); kill_guest(cpu, "Disabling interrupts");
} }
/*H:205 /*H:205
@@ -129,23 +129,23 @@ static void set_guest_interrupt(struct lguest *lg, u32 lo, u32 hi, int has_err)
* *
* maybe_do_interrupt() gets called before every entry to the Guest, to see if * maybe_do_interrupt() gets called before every entry to the Guest, to see if
* we should divert the Guest to running an interrupt handler. */ * we should divert the Guest to running an interrupt handler. */
void maybe_do_interrupt(struct lguest *lg) void maybe_do_interrupt(struct lg_cpu *cpu)
{ {
unsigned int irq; unsigned int irq;
DECLARE_BITMAP(blk, LGUEST_IRQS); DECLARE_BITMAP(blk, LGUEST_IRQS);
struct desc_struct *idt; struct desc_struct *idt;
/* If the Guest hasn't even initialized yet, we can do nothing. */ /* If the Guest hasn't even initialized yet, we can do nothing. */
if (!lg->lguest_data) if (!cpu->lg->lguest_data)
return; return;
/* Take our "irqs_pending" array and remove any interrupts the Guest /* Take our "irqs_pending" array and remove any interrupts the Guest
* wants blocked: the result ends up in "blk". */ * wants blocked: the result ends up in "blk". */
if (copy_from_user(&blk, lg->lguest_data->blocked_interrupts, if (copy_from_user(&blk, cpu->lg->lguest_data->blocked_interrupts,
sizeof(blk))) sizeof(blk)))
return; return;
bitmap_andnot(blk, lg->irqs_pending, blk, LGUEST_IRQS); bitmap_andnot(blk, cpu->irqs_pending, blk, LGUEST_IRQS);
/* Find the first interrupt. */ /* Find the first interrupt. */
irq = find_first_bit(blk, LGUEST_IRQS); irq = find_first_bit(blk, LGUEST_IRQS);
@@ -155,19 +155,20 @@ void maybe_do_interrupt(struct lguest *lg)
/* They may be in the middle of an iret, where they asked us never to /* They may be in the middle of an iret, where they asked us never to
* deliver interrupts. */ * deliver interrupts. */
if (lg->regs->eip >= lg->noirq_start && lg->regs->eip < lg->noirq_end) if (cpu->regs->eip >= cpu->lg->noirq_start &&
(cpu->regs->eip < cpu->lg->noirq_end))
return; return;
/* If they're halted, interrupts restart them. */ /* If they're halted, interrupts restart them. */
if (lg->halted) { if (cpu->halted) {
/* Re-enable interrupts. */ /* Re-enable interrupts. */
if (put_user(X86_EFLAGS_IF, &lg->lguest_data->irq_enabled)) if (put_user(X86_EFLAGS_IF, &cpu->lg->lguest_data->irq_enabled))
kill_guest(lg, "Re-enabling interrupts"); kill_guest(cpu, "Re-enabling interrupts");
lg->halted = 0; cpu->halted = 0;
} else { } else {
/* Otherwise we check if they have interrupts disabled. */ /* Otherwise we check if they have interrupts disabled. */
u32 irq_enabled; u32 irq_enabled;
if (get_user(irq_enabled, &lg->lguest_data->irq_enabled)) if (get_user(irq_enabled, &cpu->lg->lguest_data->irq_enabled))
irq_enabled = 0; irq_enabled = 0;
if (!irq_enabled) if (!irq_enabled)
return; return;
@@ -176,15 +177,15 @@ void maybe_do_interrupt(struct lguest *lg)
/* Look at the IDT entry the Guest gave us for this interrupt. The /* Look at the IDT entry the Guest gave us for this interrupt. The
* first 32 (FIRST_EXTERNAL_VECTOR) entries are for traps, so we skip * first 32 (FIRST_EXTERNAL_VECTOR) entries are for traps, so we skip
* over them. */ * over them. */
idt = &lg->arch.idt[FIRST_EXTERNAL_VECTOR+irq]; idt = &cpu->arch.idt[FIRST_EXTERNAL_VECTOR+irq];
/* If they don't have a handler (yet?), we just ignore it */ /* If they don't have a handler (yet?), we just ignore it */
if (idt_present(idt->a, idt->b)) { if (idt_present(idt->a, idt->b)) {
/* OK, mark it no longer pending and deliver it. */ /* OK, mark it no longer pending and deliver it. */
clear_bit(irq, lg->irqs_pending); clear_bit(irq, cpu->irqs_pending);
/* set_guest_interrupt() takes the interrupt descriptor and a /* set_guest_interrupt() takes the interrupt descriptor and a
* flag to say whether this interrupt pushes an error code onto * flag to say whether this interrupt pushes an error code onto
* the stack as well: virtual interrupts never do. */ * the stack as well: virtual interrupts never do. */
set_guest_interrupt(lg, idt->a, idt->b, 0); set_guest_interrupt(cpu, idt->a, idt->b, 0);
} }
/* Every time we deliver an interrupt, we update the timestamp in the /* Every time we deliver an interrupt, we update the timestamp in the
@@ -192,7 +193,7 @@ void maybe_do_interrupt(struct lguest *lg)
* did this more often, but it can actually be quite slow: doing it * did this more often, but it can actually be quite slow: doing it
* here is a compromise which means at least it gets updated every * here is a compromise which means at least it gets updated every
* timer interrupt. */ * timer interrupt. */
write_timestamp(lg); write_timestamp(cpu);
} }
/*:*/ /*:*/
@@ -245,19 +246,19 @@ static int has_err(unsigned int trap)
} }
/* deliver_trap() returns true if it could deliver the trap. */ /* deliver_trap() returns true if it could deliver the trap. */
int deliver_trap(struct lguest *lg, unsigned int num) int deliver_trap(struct lg_cpu *cpu, unsigned int num)
{ {
/* Trap numbers are always 8 bit, but we set an impossible trap number /* Trap numbers are always 8 bit, but we set an impossible trap number
* for traps inside the Switcher, so check that here. */ * for traps inside the Switcher, so check that here. */
if (num >= ARRAY_SIZE(lg->arch.idt)) if (num >= ARRAY_SIZE(cpu->arch.idt))
return 0; return 0;
/* Early on the Guest hasn't set the IDT entries (or maybe it put a /* Early on the Guest hasn't set the IDT entries (or maybe it put a
* bogus one in): if we fail here, the Guest will be killed. */ * bogus one in): if we fail here, the Guest will be killed. */
if (!idt_present(lg->arch.idt[num].a, lg->arch.idt[num].b)) if (!idt_present(cpu->arch.idt[num].a, cpu->arch.idt[num].b))
return 0; return 0;
set_guest_interrupt(lg, lg->arch.idt[num].a, lg->arch.idt[num].b, set_guest_interrupt(cpu, cpu->arch.idt[num].a,
has_err(num)); cpu->arch.idt[num].b, has_err(num));
return 1; return 1;
} }
@@ -309,18 +310,18 @@ static int direct_trap(unsigned int num)
* the Guest. * the Guest.
* *
* Which is deeply unfair, because (literally!) it wasn't the Guests' fault. */ * Which is deeply unfair, because (literally!) it wasn't the Guests' fault. */
void pin_stack_pages(struct lguest *lg) void pin_stack_pages(struct lg_cpu *cpu)
{ {
unsigned int i; unsigned int i;
/* Depending on the CONFIG_4KSTACKS option, the Guest can have one or /* Depending on the CONFIG_4KSTACKS option, the Guest can have one or
* two pages of stack space. */ * two pages of stack space. */
for (i = 0; i < lg->stack_pages; i++) for (i = 0; i < cpu->lg->stack_pages; i++)
/* The stack grows *upwards*, so the address we're given is the /* The stack grows *upwards*, so the address we're given is the
* start of the page after the kernel stack. Subtract one to * start of the page after the kernel stack. Subtract one to
* get back onto the first stack page, and keep subtracting to * get back onto the first stack page, and keep subtracting to
* get to the rest of the stack pages. */ * get to the rest of the stack pages. */
pin_page(lg, lg->esp1 - 1 - i * PAGE_SIZE); pin_page(cpu, cpu->esp1 - 1 - i * PAGE_SIZE);
} }
/* Direct traps also mean that we need to know whenever the Guest wants to use /* Direct traps also mean that we need to know whenever the Guest wants to use
@@ -331,21 +332,21 @@ void pin_stack_pages(struct lguest *lg)
* *
* In Linux each process has its own kernel stack, so this happens a lot: we * In Linux each process has its own kernel stack, so this happens a lot: we
* change stacks on each context switch. */ * change stacks on each context switch. */
void guest_set_stack(struct lguest *lg, u32 seg, u32 esp, unsigned int pages) void guest_set_stack(struct lg_cpu *cpu, u32 seg, u32 esp, unsigned int pages)
{ {
/* You are not allowed have a stack segment with privilege level 0: bad /* You are not allowed have a stack segment with privilege level 0: bad
* Guest! */ * Guest! */
if ((seg & 0x3) != GUEST_PL) if ((seg & 0x3) != GUEST_PL)
kill_guest(lg, "bad stack segment %i", seg); kill_guest(cpu, "bad stack segment %i", seg);
/* We only expect one or two stack pages. */ /* We only expect one or two stack pages. */
if (pages > 2) if (pages > 2)
kill_guest(lg, "bad stack pages %u", pages); kill_guest(cpu, "bad stack pages %u", pages);
/* Save where the stack is, and how many pages */ /* Save where the stack is, and how many pages */
lg->ss1 = seg; cpu->ss1 = seg;
lg->esp1 = esp; cpu->esp1 = esp;
lg->stack_pages = pages; cpu->lg->stack_pages = pages;
/* Make sure the new stack pages are mapped */ /* Make sure the new stack pages are mapped */
pin_stack_pages(lg); pin_stack_pages(cpu);
} }
/* All this reference to mapping stacks leads us neatly into the other complex /* All this reference to mapping stacks leads us neatly into the other complex
@@ -353,7 +354,7 @@ void guest_set_stack(struct lguest *lg, u32 seg, u32 esp, unsigned int pages)
/*H:235 This is the routine which actually checks the Guest's IDT entry and /*H:235 This is the routine which actually checks the Guest's IDT entry and
* transfers it into the entry in "struct lguest": */ * transfers it into the entry in "struct lguest": */
static void set_trap(struct lguest *lg, struct desc_struct *trap, static void set_trap(struct lg_cpu *cpu, struct desc_struct *trap,
unsigned int num, u32 lo, u32 hi) unsigned int num, u32 lo, u32 hi)
{ {
u8 type = idt_type(lo, hi); u8 type = idt_type(lo, hi);
@@ -366,7 +367,7 @@ static void set_trap(struct lguest *lg, struct desc_struct *trap,
/* We only support interrupt and trap gates. */ /* We only support interrupt and trap gates. */
if (type != 0xE && type != 0xF) if (type != 0xE && type != 0xF)
kill_guest(lg, "bad IDT type %i", type); kill_guest(cpu, "bad IDT type %i", type);
/* We only copy the handler address, present bit, privilege level and /* We only copy the handler address, present bit, privilege level and
* type. The privilege level controls where the trap can be triggered * type. The privilege level controls where the trap can be triggered
@@ -383,7 +384,7 @@ static void set_trap(struct lguest *lg, struct desc_struct *trap,
* *
* We saw the Guest setting Interrupt Descriptor Table (IDT) entries with the * We saw the Guest setting Interrupt Descriptor Table (IDT) entries with the
* LHCALL_LOAD_IDT_ENTRY hypercall before: that comes here. */ * LHCALL_LOAD_IDT_ENTRY hypercall before: that comes here. */
void load_guest_idt_entry(struct lguest *lg, unsigned int num, u32 lo, u32 hi) void load_guest_idt_entry(struct lg_cpu *cpu, unsigned int num, u32 lo, u32 hi)
{ {
/* Guest never handles: NMI, doublefault, spurious interrupt or /* Guest never handles: NMI, doublefault, spurious interrupt or
* hypercall. We ignore when it tries to set them. */ * hypercall. We ignore when it tries to set them. */
@@ -392,13 +393,13 @@ void load_guest_idt_entry(struct lguest *lg, unsigned int num, u32 lo, u32 hi)
/* Mark the IDT as changed: next time the Guest runs we'll know we have /* Mark the IDT as changed: next time the Guest runs we'll know we have
* to copy this again. */ * to copy this again. */
lg->changed |= CHANGED_IDT; cpu->changed |= CHANGED_IDT;
/* Check that the Guest doesn't try to step outside the bounds. */ /* Check that the Guest doesn't try to step outside the bounds. */
if (num >= ARRAY_SIZE(lg->arch.idt)) if (num >= ARRAY_SIZE(cpu->arch.idt))
kill_guest(lg, "Setting idt entry %u", num); kill_guest(cpu, "Setting idt entry %u", num);
else else
set_trap(lg, &lg->arch.idt[num], num, lo, hi); set_trap(cpu, &cpu->arch.idt[num], num, lo, hi);
} }
/* The default entry for each interrupt points into the Switcher routines which /* The default entry for each interrupt points into the Switcher routines which
@@ -434,14 +435,14 @@ void setup_default_idt_entries(struct lguest_ro_state *state,
/*H:240 We don't use the IDT entries in the "struct lguest" directly, instead /*H:240 We don't use the IDT entries in the "struct lguest" directly, instead
* we copy them into the IDT which we've set up for Guests on this CPU, just * we copy them into the IDT which we've set up for Guests on this CPU, just
* before we run the Guest. This routine does that copy. */ * before we run the Guest. This routine does that copy. */
void copy_traps(const struct lguest *lg, struct desc_struct *idt, void copy_traps(const struct lg_cpu *cpu, struct desc_struct *idt,
const unsigned long *def) const unsigned long *def)
{ {
unsigned int i; unsigned int i;
/* We can simply copy the direct traps, otherwise we use the default /* We can simply copy the direct traps, otherwise we use the default
* ones in the Switcher: they will return to the Host. */ * ones in the Switcher: they will return to the Host. */
for (i = 0; i < ARRAY_SIZE(lg->arch.idt); i++) { for (i = 0; i < ARRAY_SIZE(cpu->arch.idt); i++) {
/* If no Guest can ever override this trap, leave it alone. */ /* If no Guest can ever override this trap, leave it alone. */
if (!direct_trap(i)) if (!direct_trap(i))
continue; continue;
@@ -450,8 +451,8 @@ void copy_traps(const struct lguest *lg, struct desc_struct *idt,
* Interrupt gates (type 14) disable interrupts as they are * Interrupt gates (type 14) disable interrupts as they are
* entered, which we never let the Guest do. Not present * entered, which we never let the Guest do. Not present
* entries (type 0x0) also can't go direct, of course. */ * entries (type 0x0) also can't go direct, of course. */
if (idt_type(lg->arch.idt[i].a, lg->arch.idt[i].b) == 0xF) if (idt_type(cpu->arch.idt[i].a, cpu->arch.idt[i].b) == 0xF)
idt[i] = lg->arch.idt[i]; idt[i] = cpu->arch.idt[i];
else else
/* Reset it to the default. */ /* Reset it to the default. */
default_idt_entry(&idt[i], i, def[i]); default_idt_entry(&idt[i], i, def[i]);
@@ -470,13 +471,13 @@ void copy_traps(const struct lguest *lg, struct desc_struct *idt,
* infrastructure to set a callback at that time. * infrastructure to set a callback at that time.
* *
* 0 means "turn off the clock". */ * 0 means "turn off the clock". */
void guest_set_clockevent(struct lguest *lg, unsigned long delta) void guest_set_clockevent(struct lg_cpu *cpu, unsigned long delta)
{ {
ktime_t expires; ktime_t expires;
if (unlikely(delta == 0)) { if (unlikely(delta == 0)) {
/* Clock event device is shutting down. */ /* Clock event device is shutting down. */
hrtimer_cancel(&lg->hrt); hrtimer_cancel(&cpu->hrt);
return; return;
} }
@@ -484,25 +485,25 @@ void guest_set_clockevent(struct lguest *lg, unsigned long delta)
* all the time between now and the timer interrupt it asked for. This * all the time between now and the timer interrupt it asked for. This
* is almost always the right thing to do. */ * is almost always the right thing to do. */
expires = ktime_add_ns(ktime_get_real(), delta); expires = ktime_add_ns(ktime_get_real(), delta);
hrtimer_start(&lg->hrt, expires, HRTIMER_MODE_ABS); hrtimer_start(&cpu->hrt, expires, HRTIMER_MODE_ABS);
} }
/* This is the function called when the Guest's timer expires. */ /* This is the function called when the Guest's timer expires. */
static enum hrtimer_restart clockdev_fn(struct hrtimer *timer) static enum hrtimer_restart clockdev_fn(struct hrtimer *timer)
{ {
struct lguest *lg = container_of(timer, struct lguest, hrt); struct lg_cpu *cpu = container_of(timer, struct lg_cpu, hrt);
/* Remember the first interrupt is the timer interrupt. */ /* Remember the first interrupt is the timer interrupt. */
set_bit(0, lg->irqs_pending); set_bit(0, cpu->irqs_pending);
/* If the Guest is actually stopped, we need to wake it up. */ /* If the Guest is actually stopped, we need to wake it up. */
if (lg->halted) if (cpu->halted)
wake_up_process(lg->tsk); wake_up_process(cpu->tsk);
return HRTIMER_NORESTART; return HRTIMER_NORESTART;
} }
/* This sets up the timer for this Guest. */ /* This sets up the timer for this Guest. */
void init_clockdev(struct lguest *lg) void init_clockdev(struct lg_cpu *cpu)
{ {
hrtimer_init(&lg->hrt, CLOCK_REALTIME, HRTIMER_MODE_ABS); hrtimer_init(&cpu->hrt, CLOCK_REALTIME, HRTIMER_MODE_ABS);
lg->hrt.function = clockdev_fn; cpu->hrt.function = clockdev_fn;
} }

164
drivers/lguest/lg.h
View File

@@ -8,6 +8,7 @@
#include <linux/lguest.h> #include <linux/lguest.h>
#include <linux/lguest_launcher.h> #include <linux/lguest_launcher.h>
#include <linux/wait.h> #include <linux/wait.h>
#include <linux/hrtimer.h>
#include <linux/err.h> #include <linux/err.h>
#include <asm/semaphore.h> #include <asm/semaphore.h>
@@ -38,58 +39,72 @@ struct lguest_pages
#define CHANGED_GDT_TLS 4 /* Actually a subset of CHANGED_GDT */ #define CHANGED_GDT_TLS 4 /* Actually a subset of CHANGED_GDT */
#define CHANGED_ALL 3 #define CHANGED_ALL 3
/* The private info the thread maintains about the guest. */ struct lguest;
struct lguest
{ struct lg_cpu {
unsigned int id;
struct lguest *lg;
struct task_struct *tsk;
struct mm_struct *mm; /* == tsk->mm, but that becomes NULL on exit */
u32 cr2;
int ts;
u32 esp1;
u8 ss1;
/* Bitmap of what has changed: see CHANGED_* above. */
int changed;
unsigned long pending_notify; /* pfn from LHCALL_NOTIFY */
/* At end of a page shared mapped over lguest_pages in guest. */ /* At end of a page shared mapped over lguest_pages in guest. */
unsigned long regs_page; unsigned long regs_page;
struct lguest_regs *regs; struct lguest_regs *regs;
struct lguest_pages *last_pages;
int cpu_pgd; /* which pgd this cpu is currently using */
/* If a hypercall was asked for, this points to the arguments. */
struct hcall_args *hcall;
u32 next_hcall;
/* Virtual clock device */
struct hrtimer hrt;
/* Do we need to stop what we're doing and return to userspace? */
int break_out;
wait_queue_head_t break_wq;
int halted;
/* Pending virtual interrupts */
DECLARE_BITMAP(irqs_pending, LGUEST_IRQS);
struct lg_cpu_arch arch;
};
/* The private info the thread maintains about the guest. */
struct lguest
{
struct lguest_data __user *lguest_data; struct lguest_data __user *lguest_data;
struct task_struct *tsk; struct lg_cpu cpus[NR_CPUS];
struct mm_struct *mm; /* == tsk->mm, but that becomes NULL on exit */ unsigned int nr_cpus;
u32 pfn_limit; u32 pfn_limit;
/* This provides the offset to the base of guest-physical /* This provides the offset to the base of guest-physical
* memory in the Launcher. */ * memory in the Launcher. */
void __user *mem_base; void __user *mem_base;
unsigned long kernel_address; unsigned long kernel_address;
u32 cr2;
int halted;
int ts;
u32 next_hcall;
u32 esp1;
u8 ss1;
/* If a hypercall was asked for, this points to the arguments. */
struct hcall_args *hcall;
/* Do we need to stop what we're doing and return to userspace? */
int break_out;
wait_queue_head_t break_wq;
/* Bitmap of what has changed: see CHANGED_* above. */
int changed;
struct lguest_pages *last_pages;
/* We keep a small number of these. */
u32 pgdidx;
struct pgdir pgdirs[4]; struct pgdir pgdirs[4];
unsigned long noirq_start, noirq_end; unsigned long noirq_start, noirq_end;
unsigned long pending_notify; /* pfn from LHCALL_NOTIFY */
unsigned int stack_pages; unsigned int stack_pages;
u32 tsc_khz; u32 tsc_khz;
/* Dead? */ /* Dead? */
const char *dead; const char *dead;
struct lguest_arch arch;
/* Virtual clock device */
struct hrtimer hrt;
/* Pending virtual interrupts */
DECLARE_BITMAP(irqs_pending, LGUEST_IRQS);
}; };
extern struct mutex lguest_lock; extern struct mutex lguest_lock;
@@ -97,26 +112,26 @@ extern struct mutex lguest_lock;
/* core.c: */ /* core.c: */
int lguest_address_ok(const struct lguest *lg, int lguest_address_ok(const struct lguest *lg,
unsigned long addr, unsigned long len); unsigned long addr, unsigned long len);
void __lgread(struct lguest *, void *, unsigned long, unsigned); void __lgread(struct lg_cpu *, void *, unsigned long, unsigned);
void __lgwrite(struct lguest *, unsigned long, const void *, unsigned); void __lgwrite(struct lg_cpu *, unsigned long, const void *, unsigned);
/*H:035 Using memory-copy operations like that is usually inconvient, so we /*H:035 Using memory-copy operations like that is usually inconvient, so we
* have the following helper macros which read and write a specific type (often * have the following helper macros which read and write a specific type (often
* an unsigned long). * an unsigned long).
* *
* This reads into a variable of the given type then returns that. */ * This reads into a variable of the given type then returns that. */
#define lgread(lg, addr, type) \ #define lgread(cpu, addr, type) \
({ type _v; __lgread((lg), &_v, (addr), sizeof(_v)); _v; }) ({ type _v; __lgread((cpu), &_v, (addr), sizeof(_v)); _v; })
/* This checks that the variable is of the given type, then writes it out. */ /* This checks that the variable is of the given type, then writes it out. */
#define lgwrite(lg, addr, type, val) \ #define lgwrite(cpu, addr, type, val) \
do { \ do { \
typecheck(type, val); \ typecheck(type, val); \
__lgwrite((lg), (addr), &(val), sizeof(val)); \ __lgwrite((cpu), (addr), &(val), sizeof(val)); \
} while(0) } while(0)
/* (end of memory access helper routines) :*/ /* (end of memory access helper routines) :*/
int run_guest(struct lguest *lg, unsigned long __user *user); int run_guest(struct lg_cpu *cpu, unsigned long __user *user);
/* Helper macros to obtain the first 12 or the last 20 bits, this is only the /* Helper macros to obtain the first 12 or the last 20 bits, this is only the
* first step in the migration to the kernel types. pte_pfn is already defined * first step in the migration to the kernel types. pte_pfn is already defined
@@ -126,52 +141,53 @@ int run_guest(struct lguest *lg, unsigned long __user *user);
#define pgd_pfn(x) (pgd_val(x) >> PAGE_SHIFT) #define pgd_pfn(x) (pgd_val(x) >> PAGE_SHIFT)
/* interrupts_and_traps.c: */ /* interrupts_and_traps.c: */
void maybe_do_interrupt(struct lguest *lg); void maybe_do_interrupt(struct lg_cpu *cpu);
int deliver_trap(struct lguest *lg, unsigned int num); int deliver_trap(struct lg_cpu *cpu, unsigned int num);
void load_guest_idt_entry(struct lguest *lg, unsigned int i, u32 low, u32 hi); void load_guest_idt_entry(struct lg_cpu *cpu, unsigned int i,
void guest_set_stack(struct lguest *lg, u32 seg, u32 esp, unsigned int pages); u32 low, u32 hi);
void pin_stack_pages(struct lguest *lg); void guest_set_stack(struct lg_cpu *cpu, u32 seg, u32 esp, unsigned int pages);
void pin_stack_pages(struct lg_cpu *cpu);
void setup_default_idt_entries(struct lguest_ro_state *state, void setup_default_idt_entries(struct lguest_ro_state *state,
const unsigned long *def); const unsigned long *def);
void copy_traps(const struct lguest *lg, struct desc_struct *idt, void copy_traps(const struct lg_cpu *cpu, struct desc_struct *idt,
const unsigned long *def); const unsigned long *def);
void guest_set_clockevent(struct lguest *lg, unsigned long delta); void guest_set_clockevent(struct lg_cpu *cpu, unsigned long delta);
void init_clockdev(struct lguest *lg); void init_clockdev(struct lg_cpu *cpu);
bool check_syscall_vector(struct lguest *lg); bool check_syscall_vector(struct lguest *lg);
int init_interrupts(void); int init_interrupts(void);
void free_interrupts(void); void free_interrupts(void);
/* segments.c: */ /* segments.c: */
void setup_default_gdt_entries(struct lguest_ro_state *state); void setup_default_gdt_entries(struct lguest_ro_state *state);
void setup_guest_gdt(struct lguest *lg); void setup_guest_gdt(struct lg_cpu *cpu);
void load_guest_gdt(struct lguest *lg, unsigned long table, u32 num); void load_guest_gdt(struct lg_cpu *cpu, unsigned long table, u32 num);
void guest_load_tls(struct lguest *lg, unsigned long tls_array); void guest_load_tls(struct lg_cpu *cpu, unsigned long tls_array);
void copy_gdt(const struct lguest *lg, struct desc_struct *gdt); void copy_gdt(const struct lg_cpu *cpu, struct desc_struct *gdt);
void copy_gdt_tls(const struct lguest *lg, struct desc_struct *gdt); void copy_gdt_tls(const struct lg_cpu *cpu, struct desc_struct *gdt);
/* page_tables.c: */ /* page_tables.c: */
int init_guest_pagetable(struct lguest *lg, unsigned long pgtable); int init_guest_pagetable(struct lguest *lg, unsigned long pgtable);
void free_guest_pagetable(struct lguest *lg); void free_guest_pagetable(struct lguest *lg);
void guest_new_pagetable(struct lguest *lg, unsigned long pgtable); void guest_new_pagetable(struct lg_cpu *cpu, unsigned long pgtable);
void guest_set_pmd(struct lguest *lg, unsigned long gpgdir, u32 i); void guest_set_pmd(struct lguest *lg, unsigned long gpgdir, u32 i);
void guest_pagetable_clear_all(struct lguest *lg); void guest_pagetable_clear_all(struct lg_cpu *cpu);
void guest_pagetable_flush_user(struct lguest *lg); void guest_pagetable_flush_user(struct lg_cpu *cpu);
void guest_set_pte(struct lguest *lg, unsigned long gpgdir, void guest_set_pte(struct lg_cpu *cpu, unsigned long gpgdir,
unsigned long vaddr, pte_t val); unsigned long vaddr, pte_t val);
void map_switcher_in_guest(struct lguest *lg, struct lguest_pages *pages); void map_switcher_in_guest(struct lg_cpu *cpu, struct lguest_pages *pages);
int demand_page(struct lguest *info, unsigned long cr2, int errcode); int demand_page(struct lg_cpu *cpu, unsigned long cr2, int errcode);
void pin_page(struct lguest *lg, unsigned long vaddr); void pin_page(struct lg_cpu *cpu, unsigned long vaddr);
unsigned long guest_pa(struct lguest *lg, unsigned long vaddr); unsigned long guest_pa(struct lg_cpu *cpu, unsigned long vaddr);
void page_table_guest_data_init(struct lguest *lg); void page_table_guest_data_init(struct lg_cpu *cpu);
/* <arch>/core.c: */ /* <arch>/core.c: */
void lguest_arch_host_init(void); void lguest_arch_host_init(void);
void lguest_arch_host_fini(void); void lguest_arch_host_fini(void);
void lguest_arch_run_guest(struct lguest *lg); void lguest_arch_run_guest(struct lg_cpu *cpu);
void lguest_arch_handle_trap(struct lguest *lg); void lguest_arch_handle_trap(struct lg_cpu *cpu);
int lguest_arch_init_hypercalls(struct lguest *lg); int lguest_arch_init_hypercalls(struct lg_cpu *cpu);
int lguest_arch_do_hcall(struct lguest *lg, struct hcall_args *args); int lguest_arch_do_hcall(struct lg_cpu *cpu, struct hcall_args *args);
void lguest_arch_setup_regs(struct lguest *lg, unsigned long start); void lguest_arch_setup_regs(struct lg_cpu *cpu, unsigned long start);
/* <arch>/switcher.S: */ /* <arch>/switcher.S: */
extern char start_switcher_text[], end_switcher_text[], switch_to_guest[]; extern char start_switcher_text[], end_switcher_text[], switch_to_guest[];
@@ -181,8 +197,8 @@ int lguest_device_init(void);
void lguest_device_remove(void); void lguest_device_remove(void);
/* hypercalls.c: */ /* hypercalls.c: */
void do_hypercalls(struct lguest *lg); void do_hypercalls(struct lg_cpu *cpu);
void write_timestamp(struct lguest *lg); void write_timestamp(struct lg_cpu *cpu);
/*L:035 /*L:035
* Let's step aside for the moment, to study one important routine that's used * Let's step aside for the moment, to study one important routine that's used
@@ -208,12 +224,12 @@ void write_timestamp(struct lguest *lg);
* Like any macro which uses an "if", it is safely wrapped in a run-once "do { * Like any macro which uses an "if", it is safely wrapped in a run-once "do {
* } while(0)". * } while(0)".
*/ */
#define kill_guest(lg, fmt...) \ #define kill_guest(cpu, fmt...) \
do { \ do { \
if (!(lg)->dead) { \ if (!(cpu)->lg->dead) { \
(lg)->dead = kasprintf(GFP_ATOMIC, fmt); \ (cpu)->lg->dead = kasprintf(GFP_ATOMIC, fmt); \
if (!(lg)->dead) \ if (!(cpu)->lg->dead) \
(lg)->dead = ERR_PTR(-ENOMEM); \ (cpu)->lg->dead = ERR_PTR(-ENOMEM); \
} \ } \
} while(0) } while(0)
/* (End of aside) :*/ /* (End of aside) :*/

147
drivers/lguest/lguest_user.c
View File

@@ -6,6 +6,7 @@
#include <linux/uaccess.h> #include <linux/uaccess.h>
#include <linux/miscdevice.h> #include <linux/miscdevice.h>
#include <linux/fs.h> #include <linux/fs.h>
#include <linux/sched.h>
#include "lg.h" #include "lg.h"
/*L:055 When something happens, the Waker process needs a way to stop the /*L:055 When something happens, the Waker process needs a way to stop the
@@ -13,7 +14,7 @@
* LHREQ_BREAK and the value "1" to /dev/lguest to do this. Once the Launcher * LHREQ_BREAK and the value "1" to /dev/lguest to do this. Once the Launcher
* has done whatever needs attention, it writes LHREQ_BREAK and "0" to release * has done whatever needs attention, it writes LHREQ_BREAK and "0" to release
* the Waker. */ * the Waker. */
static int break_guest_out(struct lguest *lg, const unsigned long __user *input) static int break_guest_out(struct lg_cpu *cpu, const unsigned long __user*input)
{ {
unsigned long on; unsigned long on;
@@ -22,21 +23,21 @@ static int break_guest_out(struct lguest *lg, const unsigned long __user *input)
return -EFAULT; return -EFAULT;
if (on) { if (on) {
lg->break_out = 1; cpu->break_out = 1;
/* Pop it out of the Guest (may be running on different CPU) */ /* Pop it out of the Guest (may be running on different CPU) */
wake_up_process(lg->tsk); wake_up_process(cpu->tsk);
/* Wait for them to reset it */ /* Wait for them to reset it */
return wait_event_interruptible(lg->break_wq, !lg->break_out); return wait_event_interruptible(cpu->break_wq, !cpu->break_out);
} else { } else {
lg->break_out = 0; cpu->break_out = 0;
wake_up(&lg->break_wq); wake_up(&cpu->break_wq);
return 0; return 0;
} }
} }
/*L:050 Sending an interrupt is done by writing LHREQ_IRQ and an interrupt /*L:050 Sending an interrupt is done by writing LHREQ_IRQ and an interrupt
* number to /dev/lguest. */ * number to /dev/lguest. */
static int user_send_irq(struct lguest *lg, const unsigned long __user *input) static int user_send_irq(struct lg_cpu *cpu, const unsigned long __user *input)
{ {
unsigned long irq; unsigned long irq;
@@ -46,7 +47,7 @@ static int user_send_irq(struct lguest *lg, const unsigned long __user *input)
return -EINVAL; return -EINVAL;
/* Next time the Guest runs, the core code will see if it can deliver /* Next time the Guest runs, the core code will see if it can deliver
* this interrupt. */ * this interrupt. */
set_bit(irq, lg->irqs_pending); set_bit(irq, cpu->irqs_pending);
return 0; return 0;
} }
@@ -55,13 +56,21 @@ static int user_send_irq(struct lguest *lg, const unsigned long __user *input)
static ssize_t read(struct file *file, char __user *user, size_t size,loff_t*o) static ssize_t read(struct file *file, char __user *user, size_t size,loff_t*o)
{ {
struct lguest *lg = file->private_data; struct lguest *lg = file->private_data;
struct lg_cpu *cpu;
unsigned int cpu_id = *o;
/* You must write LHREQ_INITIALIZE first! */ /* You must write LHREQ_INITIALIZE first! */
if (!lg) if (!lg)
return -EINVAL; return -EINVAL;
/* Watch out for arbitrary vcpu indexes! */
if (cpu_id >= lg->nr_cpus)
return -EINVAL;
cpu = &lg->cpus[cpu_id];
/* If you're not the task which owns the Guest, go away. */ /* If you're not the task which owns the Guest, go away. */
if (current != lg->tsk) if (current != cpu->tsk)
return -EPERM; return -EPERM;
/* If the guest is already dead, we indicate why */ /* If the guest is already dead, we indicate why */
@@ -81,11 +90,53 @@ static ssize_t read(struct file *file, char __user *user, size_t size,loff_t*o)
/* If we returned from read() last time because the Guest notified, /* If we returned from read() last time because the Guest notified,
* clear the flag. */ * clear the flag. */
if (lg->pending_notify) if (cpu->pending_notify)
lg->pending_notify = 0; cpu->pending_notify = 0;
/* Run the Guest until something interesting happens. */ /* Run the Guest until something interesting happens. */
return run_guest(lg, (unsigned long __user *)user); return run_guest(cpu, (unsigned long __user *)user);
}
static int lg_cpu_start(struct lg_cpu *cpu, unsigned id, unsigned long start_ip)
{
if (id >= NR_CPUS)
return -EINVAL;
cpu->id = id;
cpu->lg = container_of((cpu - id), struct lguest, cpus[0]);
cpu->lg->nr_cpus++;
init_clockdev(cpu);
/* We need a complete page for the Guest registers: they are accessible
* to the Guest and we can only grant it access to whole pages. */
cpu->regs_page = get_zeroed_page(GFP_KERNEL);
if (!cpu->regs_page)
return -ENOMEM;
/* We actually put the registers at the bottom of the page. */
cpu->regs = (void *)cpu->regs_page + PAGE_SIZE - sizeof(*cpu->regs);
/* Now we initialize the Guest's registers, handing it the start
* address. */
lguest_arch_setup_regs(cpu, start_ip);
/* Initialize the queue for the waker to wait on */
init_waitqueue_head(&cpu->break_wq);
/* We keep a pointer to the Launcher task (ie. current task) for when
* other Guests want to wake this one (inter-Guest I/O). */
cpu->tsk = current;
/* We need to keep a pointer to the Launcher's memory map, because if
* the Launcher dies we need to clean it up. If we don't keep a
* reference, it is destroyed before close() is called. */
cpu->mm = get_task_mm(cpu->tsk);
/* We remember which CPU's pages this Guest used last, for optimization
* when the same Guest runs on the same CPU twice. */
cpu->last_pages = NULL;
return 0;
} }
/*L:020 The initialization write supplies 4 pointer sized (32 or 64 bit) /*L:020 The initialization write supplies 4 pointer sized (32 or 64 bit)
@@ -134,15 +185,10 @@ static int initialize(struct file *file, const unsigned long __user *input)
lg->mem_base = (void __user *)(long)args[0]; lg->mem_base = (void __user *)(long)args[0];
lg->pfn_limit = args[1]; lg->pfn_limit = args[1];
/* We need a complete page for the Guest registers: they are accessible /* This is the first cpu */
* to the Guest and we can only grant it access to whole pages. */ err = lg_cpu_start(&lg->cpus[0], 0, args[3]);
lg->regs_page = get_zeroed_page(GFP_KERNEL); if (err)
if (!lg->regs_page) {
err = -ENOMEM;
goto release_guest; goto release_guest;
}
/* We actually put the registers at the bottom of the page. */
lg->regs = (void *)lg->regs_page + PAGE_SIZE - sizeof(*lg->regs);
/* Initialize the Guest's shadow page tables, using the toplevel /* Initialize the Guest's shadow page tables, using the toplevel
* address the Launcher gave us. This allocates memory, so can * address the Launcher gave us. This allocates memory, so can
@@ -151,28 +197,6 @@ static int initialize(struct file *file, const unsigned long __user *input)
if (err) if (err)
goto free_regs; goto free_regs;
/* Now we initialize the Guest's registers, handing it the start
* address. */
lguest_arch_setup_regs(lg, args[3]);
/* The timer for lguest's clock needs initialization. */
init_clockdev(lg);
/* We keep a pointer to the Launcher task (ie. current task) for when
* other Guests want to wake this one (inter-Guest I/O). */
lg->tsk = current;
/* We need to keep a pointer to the Launcher's memory map, because if
* the Launcher dies we need to clean it up. If we don't keep a
* reference, it is destroyed before close() is called. */
lg->mm = get_task_mm(lg->tsk);
/* Initialize the queue for the waker to wait on */
init_waitqueue_head(&lg->break_wq);
/* We remember which CPU's pages this Guest used last, for optimization
* when the same Guest runs on the same CPU twice. */
lg->last_pages = NULL;
/* We keep our "struct lguest" in the file's private_data. */ /* We keep our "struct lguest" in the file's private_data. */
file->private_data = lg; file->private_data = lg;
@@ -182,7 +206,8 @@ static int initialize(struct file *file, const unsigned long __user *input)
return sizeof(args); return sizeof(args);
free_regs: free_regs:
free_page(lg->regs_page); /* FIXME: This should be in free_vcpu */
free_page(lg->cpus[0].regs_page);
release_guest: release_guest:
kfree(lg); kfree(lg);
unlock: unlock:
@@ -202,30 +227,37 @@ static ssize_t write(struct file *file, const char __user *in,
struct lguest *lg = file->private_data; struct lguest *lg = file->private_data;
const unsigned long __user *input = (const unsigned long __user *)in; const unsigned long __user *input = (const unsigned long __user *)in;
unsigned long req; unsigned long req;
struct lg_cpu *uninitialized_var(cpu);
unsigned int cpu_id = *off;
if (get_user(req, input) != 0) if (get_user(req, input) != 0)
return -EFAULT; return -EFAULT;
input++; input++;
/* If you haven't initialized, you must do that first. */ /* If you haven't initialized, you must do that first. */
if (req != LHREQ_INITIALIZE && !lg) if (req != LHREQ_INITIALIZE) {
return -EINVAL; if (!lg || (cpu_id >= lg->nr_cpus))
return -EINVAL;
cpu = &lg->cpus[cpu_id];
if (!cpu)
return -EINVAL;
}
/* Once the Guest is dead, all you can do is read() why it died. */ /* Once the Guest is dead, all you can do is read() why it died. */
if (lg && lg->dead) if (lg && lg->dead)
return -ENOENT; return -ENOENT;
/* If you're not the task which owns the Guest, you can only break */ /* If you're not the task which owns the Guest, you can only break */
if (lg && current != lg->tsk && req != LHREQ_BREAK) if (lg && current != cpu->tsk && req != LHREQ_BREAK)
return -EPERM; return -EPERM;
switch (req) { switch (req) {
case LHREQ_INITIALIZE: case LHREQ_INITIALIZE:
return initialize(file, input); return initialize(file, input);
case LHREQ_IRQ: case LHREQ_IRQ:
return user_send_irq(lg, input); return user_send_irq(cpu, input);
case LHREQ_BREAK: case LHREQ_BREAK:
return break_guest_out(lg, input); return break_guest_out(cpu, input);
default: default:
return -EINVAL; return -EINVAL;
} }
@@ -241,6 +273,7 @@ static ssize_t write(struct file *file, const char __user *in,
static int close(struct inode *inode, struct file *file) static int close(struct inode *inode, struct file *file)
{ {
struct lguest *lg = file->private_data; struct lguest *lg = file->private_data;
unsigned int i;
/* If we never successfully initialized, there's nothing to clean up */ /* If we never successfully initialized, there's nothing to clean up */
if (!lg) if (!lg)
@@ -249,19 +282,23 @@ static int close(struct inode *inode, struct file *file)
/* We need the big lock, to protect from inter-guest I/O and other /* We need the big lock, to protect from inter-guest I/O and other
* Launchers initializing guests. */ * Launchers initializing guests. */
mutex_lock(&lguest_lock); mutex_lock(&lguest_lock);
/* Cancels the hrtimer set via LHCALL_SET_CLOCKEVENT. */
hrtimer_cancel(&lg->hrt);
/* Free up the shadow page tables for the Guest. */ /* Free up the shadow page tables for the Guest. */
free_guest_pagetable(lg); free_guest_pagetable(lg);
/* Now all the memory cleanups are done, it's safe to release the
* Launcher's memory management structure. */ for (i = 0; i < lg->nr_cpus; i++) {
mmput(lg->mm); /* Cancels the hrtimer set via LHCALL_SET_CLOCKEVENT. */
hrtimer_cancel(&lg->cpus[i].hrt);
/* We can free up the register page we allocated. */
free_page(lg->cpus[i].regs_page);
/* Now all the memory cleanups are done, it's safe to release
* the Launcher's memory management structure. */
mmput(lg->cpus[i].mm);
}
/* If lg->dead doesn't contain an error code it will be NULL or a /* If lg->dead doesn't contain an error code it will be NULL or a
* kmalloc()ed string, either of which is ok to hand to kfree(). */ * kmalloc()ed string, either of which is ok to hand to kfree(). */
if (!IS_ERR(lg->dead)) if (!IS_ERR(lg->dead))
kfree(lg->dead); kfree(lg->dead);
/* We can free up the register page we allocated. */
free_page(lg->regs_page);
/* We clear the entire structure, which also marks it as free for the /* We clear the entire structure, which also marks it as free for the
* next user. */ * next user. */
memset(lg, 0, sizeof(*lg)); memset(lg, 0, sizeof(*lg));

179
drivers/lguest/page_tables.c
View File

@@ -68,23 +68,23 @@ static DEFINE_PER_CPU(pte_t *, switcher_pte_pages);
* page directory entry (PGD) for that address. Since we keep track of several * page directory entry (PGD) for that address. Since we keep track of several
* page tables, the "i" argument tells us which one we're interested in (it's * page tables, the "i" argument tells us which one we're interested in (it's
* usually the current one). */ * usually the current one). */
static pgd_t *spgd_addr(struct lguest *lg, u32 i, unsigned long vaddr) static pgd_t *spgd_addr(struct lg_cpu *cpu, u32 i, unsigned long vaddr)
{ {
unsigned int index = pgd_index(vaddr); unsigned int index = pgd_index(vaddr);
/* We kill any Guest trying to touch the Switcher addresses. */ /* We kill any Guest trying to touch the Switcher addresses. */
if (index >= SWITCHER_PGD_INDEX) { if (index >= SWITCHER_PGD_INDEX) {
kill_guest(lg, "attempt to access switcher pages"); kill_guest(cpu, "attempt to access switcher pages");
index = 0; index = 0;
} }
/* Return a pointer index'th pgd entry for the i'th page table. */ /* Return a pointer index'th pgd entry for the i'th page table. */
return &lg->pgdirs[i].pgdir[index]; return &cpu->lg->pgdirs[i].pgdir[index];
} }
/* This routine then takes the page directory entry returned above, which /* This routine then takes the page directory entry returned above, which
* contains the address of the page table entry (PTE) page. It then returns a * contains the address of the page table entry (PTE) page. It then returns a
* pointer to the PTE entry for the given address. */ * pointer to the PTE entry for the given address. */
static pte_t *spte_addr(struct lguest *lg, pgd_t spgd, unsigned long vaddr) static pte_t *spte_addr(pgd_t spgd, unsigned long vaddr)
{ {
pte_t *page = __va(pgd_pfn(spgd) << PAGE_SHIFT); pte_t *page = __va(pgd_pfn(spgd) << PAGE_SHIFT);
/* You should never call this if the PGD entry wasn't valid */ /* You should never call this if the PGD entry wasn't valid */
@@ -94,14 +94,13 @@ static pte_t *spte_addr(struct lguest *lg, pgd_t spgd, unsigned long vaddr)
/* These two functions just like the above two, except they access the Guest /* These two functions just like the above two, except they access the Guest
* page tables. Hence they return a Guest address. */ * page tables. Hence they return a Guest address. */
static unsigned long gpgd_addr(struct lguest *lg, unsigned long vaddr) static unsigned long gpgd_addr(struct lg_cpu *cpu, unsigned long vaddr)
{ {
unsigned int index = vaddr >> (PGDIR_SHIFT); unsigned int index = vaddr >> (PGDIR_SHIFT);
return lg->pgdirs[lg->pgdidx].gpgdir + index * sizeof(pgd_t); return cpu->lg->pgdirs[cpu->cpu_pgd].gpgdir + index * sizeof(pgd_t);
} }
static unsigned long gpte_addr(struct lguest *lg, static unsigned long gpte_addr(pgd_t gpgd, unsigned long vaddr)
pgd_t gpgd, unsigned long vaddr)
{ {
unsigned long gpage = pgd_pfn(gpgd) << PAGE_SHIFT; unsigned long gpage = pgd_pfn(gpgd) << PAGE_SHIFT;
BUG_ON(!(pgd_flags(gpgd) & _PAGE_PRESENT)); BUG_ON(!(pgd_flags(gpgd) & _PAGE_PRESENT));
@@ -138,7 +137,7 @@ static unsigned long get_pfn(unsigned long virtpfn, int write)
* entry can be a little tricky. The flags are (almost) the same, but the * entry can be a little tricky. The flags are (almost) the same, but the
* Guest PTE contains a virtual page number: the CPU needs the real page * Guest PTE contains a virtual page number: the CPU needs the real page
* number. */ * number. */
static pte_t gpte_to_spte(struct lguest *lg, pte_t gpte, int write) static pte_t gpte_to_spte(struct lg_cpu *cpu, pte_t gpte, int write)
{ {
unsigned long pfn, base, flags; unsigned long pfn, base, flags;
@@ -149,7 +148,7 @@ static pte_t gpte_to_spte(struct lguest *lg, pte_t gpte, int write)
flags = (pte_flags(gpte) & ~_PAGE_GLOBAL); flags = (pte_flags(gpte) & ~_PAGE_GLOBAL);
/* The Guest's pages are offset inside the Launcher. */ /* The Guest's pages are offset inside the Launcher. */
base = (unsigned long)lg->mem_base / PAGE_SIZE; base = (unsigned long)cpu->lg->mem_base / PAGE_SIZE;
/* We need a temporary "unsigned long" variable to hold the answer from /* We need a temporary "unsigned long" variable to hold the answer from
* get_pfn(), because it returns 0xFFFFFFFF on failure, which wouldn't * get_pfn(), because it returns 0xFFFFFFFF on failure, which wouldn't
@@ -157,7 +156,7 @@ static pte_t gpte_to_spte(struct lguest *lg, pte_t gpte, int write)
* page, given the virtual number. */ * page, given the virtual number. */
pfn = get_pfn(base + pte_pfn(gpte), write); pfn = get_pfn(base + pte_pfn(gpte), write);
if (pfn == -1UL) { if (pfn == -1UL) {
kill_guest(lg, "failed to get page %lu", pte_pfn(gpte)); kill_guest(cpu, "failed to get page %lu", pte_pfn(gpte));
/* When we destroy the Guest, we'll go through the shadow page /* When we destroy the Guest, we'll go through the shadow page
* tables and release_pte() them. Make sure we don't think * tables and release_pte() them. Make sure we don't think
* this one is valid! */ * this one is valid! */
@@ -177,17 +176,18 @@ static void release_pte(pte_t pte)
} }
/*:*/ /*:*/
static void check_gpte(struct lguest *lg, pte_t gpte) static void check_gpte(struct lg_cpu *cpu, pte_t gpte)
{ {
if ((pte_flags(gpte) & (_PAGE_PWT|_PAGE_PSE)) if ((pte_flags(gpte) & (_PAGE_PWT|_PAGE_PSE))
|| pte_pfn(gpte) >= lg->pfn_limit) || pte_pfn(gpte) >= cpu->lg->pfn_limit)
kill_guest(lg, "bad page table entry"); kill_guest(cpu, "bad page table entry");
} }
static void check_gpgd(struct lguest *lg, pgd_t gpgd) static void check_gpgd(struct lg_cpu *cpu, pgd_t gpgd)
{ {
if ((pgd_flags(gpgd) & ~_PAGE_TABLE) || pgd_pfn(gpgd) >= lg->pfn_limit) if ((pgd_flags(gpgd) & ~_PAGE_TABLE) ||
kill_guest(lg, "bad page directory entry"); (pgd_pfn(gpgd) >= cpu->lg->pfn_limit))
kill_guest(cpu, "bad page directory entry");
} }
/*H:330 /*H:330
@@ -200,7 +200,7 @@ static void check_gpgd(struct lguest *lg, pgd_t gpgd)
* *
* If we fixed up the fault (ie. we mapped the address), this routine returns * If we fixed up the fault (ie. we mapped the address), this routine returns
* true. Otherwise, it was a real fault and we need to tell the Guest. */ * true. Otherwise, it was a real fault and we need to tell the Guest. */
int demand_page(struct lguest *lg, unsigned long vaddr, int errcode) int demand_page(struct lg_cpu *cpu, unsigned long vaddr, int errcode)
{ {
pgd_t gpgd; pgd_t gpgd;
pgd_t *spgd; pgd_t *spgd;
@@ -209,24 +209,24 @@ int demand_page(struct lguest *lg, unsigned long vaddr, int errcode)
pte_t *spte; pte_t *spte;
/* First step: get the top-level Guest page table entry. */ /* First step: get the top-level Guest page table entry. */
gpgd = lgread(lg, gpgd_addr(lg, vaddr), pgd_t); gpgd = lgread(cpu, gpgd_addr(cpu, vaddr), pgd_t);
/* Toplevel not present? We can't map it in. */ /* Toplevel not present? We can't map it in. */
if (!(pgd_flags(gpgd) & _PAGE_PRESENT)) if (!(pgd_flags(gpgd) & _PAGE_PRESENT))
return 0; return 0;
/* Now look at the matching shadow entry. */ /* Now look at the matching shadow entry. */
spgd = spgd_addr(lg, lg->pgdidx, vaddr); spgd = spgd_addr(cpu, cpu->cpu_pgd, vaddr);
if (!(pgd_flags(*spgd) & _PAGE_PRESENT)) { if (!(pgd_flags(*spgd) & _PAGE_PRESENT)) {
/* No shadow entry: allocate a new shadow PTE page. */ /* No shadow entry: allocate a new shadow PTE page. */
unsigned long ptepage = get_zeroed_page(GFP_KERNEL); unsigned long ptepage = get_zeroed_page(GFP_KERNEL);
/* This is not really the Guest's fault, but killing it is /* This is not really the Guest's fault, but killing it is
* simple for this corner case. */ * simple for this corner case. */
if (!ptepage) { if (!ptepage) {
kill_guest(lg, "out of memory allocating pte page"); kill_guest(cpu, "out of memory allocating pte page");
return 0; return 0;
} }
/* We check that the Guest pgd is OK. */ /* We check that the Guest pgd is OK. */
check_gpgd(lg, gpgd); check_gpgd(cpu, gpgd);
/* And we copy the flags to the shadow PGD entry. The page /* And we copy the flags to the shadow PGD entry. The page
* number in the shadow PGD is the page we just allocated. */ * number in the shadow PGD is the page we just allocated. */
*spgd = __pgd(__pa(ptepage) | pgd_flags(gpgd)); *spgd = __pgd(__pa(ptepage) | pgd_flags(gpgd));
@@ -234,8 +234,8 @@ int demand_page(struct lguest *lg, unsigned long vaddr, int errcode)
/* OK, now we look at the lower level in the Guest page table: keep its /* OK, now we look at the lower level in the Guest page table: keep its
* address, because we might update it later. */ * address, because we might update it later. */
gpte_ptr = gpte_addr(lg, gpgd, vaddr); gpte_ptr = gpte_addr(gpgd, vaddr);
gpte = lgread(lg, gpte_ptr, pte_t); gpte = lgread(cpu, gpte_ptr, pte_t);
/* If this page isn't in the Guest page tables, we can't page it in. */ /* If this page isn't in the Guest page tables, we can't page it in. */
if (!(pte_flags(gpte) & _PAGE_PRESENT)) if (!(pte_flags(gpte) & _PAGE_PRESENT))
@@ -252,7 +252,7 @@ int demand_page(struct lguest *lg, unsigned long vaddr, int errcode)
/* Check that the Guest PTE flags are OK, and the page number is below /* Check that the Guest PTE flags are OK, and the page number is below
* the pfn_limit (ie. not mapping the Launcher binary). */ * the pfn_limit (ie. not mapping the Launcher binary). */
check_gpte(lg, gpte); check_gpte(cpu, gpte);
/* Add the _PAGE_ACCESSED and (for a write) _PAGE_DIRTY flag */ /* Add the _PAGE_ACCESSED and (for a write) _PAGE_DIRTY flag */
gpte = pte_mkyoung(gpte); gpte = pte_mkyoung(gpte);
@@ -260,7 +260,7 @@ int demand_page(struct lguest *lg, unsigned long vaddr, int errcode)
gpte = pte_mkdirty(gpte); gpte = pte_mkdirty(gpte);
/* Get the pointer to the shadow PTE entry we're going to set. */ /* Get the pointer to the shadow PTE entry we're going to set. */
spte = spte_addr(lg, *spgd, vaddr); spte = spte_addr(*spgd, vaddr);
/* If there was a valid shadow PTE entry here before, we release it. /* If there was a valid shadow PTE entry here before, we release it.
* This can happen with a write to a previously read-only entry. */ * This can happen with a write to a previously read-only entry. */
release_pte(*spte); release_pte(*spte);
@@ -268,17 +268,17 @@ int demand_page(struct lguest *lg, unsigned long vaddr, int errcode)
/* If this is a write, we insist that the Guest page is writable (the /* If this is a write, we insist that the Guest page is writable (the
* final arg to gpte_to_spte()). */ * final arg to gpte_to_spte()). */
if (pte_dirty(gpte)) if (pte_dirty(gpte))
*spte = gpte_to_spte(lg, gpte, 1); *spte = gpte_to_spte(cpu, gpte, 1);
else else
/* If this is a read, don't set the "writable" bit in the page /* If this is a read, don't set the "writable" bit in the page
* table entry, even if the Guest says it's writable. That way * table entry, even if the Guest says it's writable. That way
* we will come back here when a write does actually occur, so * we will come back here when a write does actually occur, so
* we can update the Guest's _PAGE_DIRTY flag. */ * we can update the Guest's _PAGE_DIRTY flag. */
*spte = gpte_to_spte(lg, pte_wrprotect(gpte), 0); *spte = gpte_to_spte(cpu, pte_wrprotect(gpte), 0);
/* Finally, we write the Guest PTE entry back: we've set the /* Finally, we write the Guest PTE entry back: we've set the
* _PAGE_ACCESSED and maybe the _PAGE_DIRTY flags. */ * _PAGE_ACCESSED and maybe the _PAGE_DIRTY flags. */
lgwrite(lg, gpte_ptr, pte_t, gpte); lgwrite(cpu, gpte_ptr, pte_t, gpte);
/* The fault is fixed, the page table is populated, the mapping /* The fault is fixed, the page table is populated, the mapping
* manipulated, the result returned and the code complete. A small * manipulated, the result returned and the code complete. A small
@@ -297,19 +297,19 @@ int demand_page(struct lguest *lg, unsigned long vaddr, int errcode)
* *
* This is a quick version which answers the question: is this virtual address * This is a quick version which answers the question: is this virtual address
* mapped by the shadow page tables, and is it writable? */ * mapped by the shadow page tables, and is it writable? */
static int page_writable(struct lguest *lg, unsigned long vaddr) static int page_writable(struct lg_cpu *cpu, unsigned long vaddr)
{ {
pgd_t *spgd; pgd_t *spgd;
unsigned long flags; unsigned long flags;
/* Look at the current top level entry: is it present? */ /* Look at the current top level entry: is it present? */
spgd = spgd_addr(lg, lg->pgdidx, vaddr); spgd = spgd_addr(cpu, cpu->cpu_pgd, vaddr);
if (!(pgd_flags(*spgd) & _PAGE_PRESENT)) if (!(pgd_flags(*spgd) & _PAGE_PRESENT))
return 0; return 0;
/* Check the flags on the pte entry itself: it must be present and /* Check the flags on the pte entry itself: it must be present and
* writable. */ * writable. */
flags = pte_flags(*(spte_addr(lg, *spgd, vaddr))); flags = pte_flags(*(spte_addr(*spgd, vaddr)));
return (flags & (_PAGE_PRESENT|_PAGE_RW)) == (_PAGE_PRESENT|_PAGE_RW); return (flags & (_PAGE_PRESENT|_PAGE_RW)) == (_PAGE_PRESENT|_PAGE_RW);
} }
@@ -317,10 +317,10 @@ static int page_writable(struct lguest *lg, unsigned long vaddr)
/* So, when pin_stack_pages() asks us to pin a page, we check if it's already /* So, when pin_stack_pages() asks us to pin a page, we check if it's already
* in the page tables, and if not, we call demand_page() with error code 2 * in the page tables, and if not, we call demand_page() with error code 2
* (meaning "write"). */ * (meaning "write"). */
void pin_page(struct lguest *lg, unsigned long vaddr) void pin_page(struct lg_cpu *cpu, unsigned long vaddr)
{ {
if (!page_writable(lg, vaddr) && !demand_page(lg, vaddr, 2)) if (!page_writable(cpu, vaddr) && !demand_page(cpu, vaddr, 2))
kill_guest(lg, "bad stack page %#lx", vaddr); kill_guest(cpu, "bad stack page %#lx", vaddr);
} }
/*H:450 If we chase down the release_pgd() code, it looks like this: */ /*H:450 If we chase down the release_pgd() code, it looks like this: */
@@ -358,28 +358,28 @@ static void flush_user_mappings(struct lguest *lg, int idx)
* *
* The Guest has a hypercall to throw away the page tables: it's used when a * The Guest has a hypercall to throw away the page tables: it's used when a
* large number of mappings have been changed. */ * large number of mappings have been changed. */
void guest_pagetable_flush_user(struct lguest *lg) void guest_pagetable_flush_user(struct lg_cpu *cpu)
{ {
/* Drop the userspace part of the current page table. */ /* Drop the userspace part of the current page table. */
flush_user_mappings(lg, lg->pgdidx); flush_user_mappings(cpu->lg, cpu->cpu_pgd);
} }
/*:*/ /*:*/
/* We walk down the guest page tables to get a guest-physical address */ /* We walk down the guest page tables to get a guest-physical address */
unsigned long guest_pa(struct lguest *lg, unsigned long vaddr) unsigned long guest_pa(struct lg_cpu *cpu, unsigned long vaddr)
{ {
pgd_t gpgd; pgd_t gpgd;
pte_t gpte; pte_t gpte;
/* First step: get the top-level Guest page table entry. */ /* First step: get the top-level Guest page table entry. */
gpgd = lgread(lg, gpgd_addr(lg, vaddr), pgd_t); gpgd = lgread(cpu, gpgd_addr(cpu, vaddr), pgd_t);
/* Toplevel not present? We can't map it in. */ /* Toplevel not present? We can't map it in. */
if (!(pgd_flags(gpgd) & _PAGE_PRESENT)) if (!(pgd_flags(gpgd) & _PAGE_PRESENT))
kill_guest(lg, "Bad address %#lx", vaddr); kill_guest(cpu, "Bad address %#lx", vaddr);
gpte = lgread(lg, gpte_addr(lg, gpgd, vaddr), pte_t); gpte = lgread(cpu, gpte_addr(gpgd, vaddr), pte_t);
if (!(pte_flags(gpte) & _PAGE_PRESENT)) if (!(pte_flags(gpte) & _PAGE_PRESENT))
kill_guest(lg, "Bad address %#lx", vaddr); kill_guest(cpu, "Bad address %#lx", vaddr);
return pte_pfn(gpte) * PAGE_SIZE | (vaddr & ~PAGE_MASK); return pte_pfn(gpte) * PAGE_SIZE | (vaddr & ~PAGE_MASK);
} }
@@ -399,7 +399,7 @@ static unsigned int find_pgdir(struct lguest *lg, unsigned long pgtable)
/*H:435 And this is us, creating the new page directory. If we really do /*H:435 And this is us, creating the new page directory. If we really do
* allocate a new one (and so the kernel parts are not there), we set * allocate a new one (and so the kernel parts are not there), we set
* blank_pgdir. */ * blank_pgdir. */
static unsigned int new_pgdir(struct lguest *lg, static unsigned int new_pgdir(struct lg_cpu *cpu,
unsigned long gpgdir, unsigned long gpgdir,
int *blank_pgdir) int *blank_pgdir)
{ {
@@ -407,22 +407,23 @@ static unsigned int new_pgdir(struct lguest *lg,
/* We pick one entry at random to throw out. Choosing the Least /* We pick one entry at random to throw out. Choosing the Least
* Recently Used might be better, but this is easy. */ * Recently Used might be better, but this is easy. */
next = random32() % ARRAY_SIZE(lg->pgdirs); next = random32() % ARRAY_SIZE(cpu->lg->pgdirs);
/* If it's never been allocated at all before, try now. */ /* If it's never been allocated at all before, try now. */
if (!lg->pgdirs[next].pgdir) { if (!cpu->lg->pgdirs[next].pgdir) {
lg->pgdirs[next].pgdir = (pgd_t *)get_zeroed_page(GFP_KERNEL); cpu->lg->pgdirs[next].pgdir =
(pgd_t *)get_zeroed_page(GFP_KERNEL);
/* If the allocation fails, just keep using the one we have */ /* If the allocation fails, just keep using the one we have */
if (!lg->pgdirs[next].pgdir) if (!cpu->lg->pgdirs[next].pgdir)
next = lg->pgdidx; next = cpu->cpu_pgd;
else else
/* This is a blank page, so there are no kernel /* This is a blank page, so there are no kernel
* mappings: caller must map the stack! */ * mappings: caller must map the stack! */
*blank_pgdir = 1; *blank_pgdir = 1;
} }
/* Record which Guest toplevel this shadows. */ /* Record which Guest toplevel this shadows. */
lg->pgdirs[next].gpgdir = gpgdir; cpu->lg->pgdirs[next].gpgdir = gpgdir;
/* Release all the non-kernel mappings. */ /* Release all the non-kernel mappings. */
flush_user_mappings(lg, next); flush_user_mappings(cpu->lg, next);
return next; return next;
} }
@@ -432,21 +433,21 @@ static unsigned int new_pgdir(struct lguest *lg,
* Now we've seen all the page table setting and manipulation, let's see what * Now we've seen all the page table setting and manipulation, let's see what
* what happens when the Guest changes page tables (ie. changes the top-level * what happens when the Guest changes page tables (ie. changes the top-level
* pgdir). This occurs on almost every context switch. */ * pgdir). This occurs on almost every context switch. */
void guest_new_pagetable(struct lguest *lg, unsigned long pgtable) void guest_new_pagetable(struct lg_cpu *cpu, unsigned long pgtable)
{ {
int newpgdir, repin = 0; int newpgdir, repin = 0;
/* Look to see if we have this one already. */ /* Look to see if we have this one already. */
newpgdir = find_pgdir(lg, pgtable); newpgdir = find_pgdir(cpu->lg, pgtable);
/* If not, we allocate or mug an existing one: if it's a fresh one, /* If not, we allocate or mug an existing one: if it's a fresh one,
* repin gets set to 1. */ * repin gets set to 1. */
if (newpgdir == ARRAY_SIZE(lg->pgdirs)) if (newpgdir == ARRAY_SIZE(cpu->lg->pgdirs))
newpgdir = new_pgdir(lg, pgtable, &repin); newpgdir = new_pgdir(cpu, pgtable, &repin);
/* Change the current pgd index to the new one. */ /* Change the current pgd index to the new one. */
lg->pgdidx = newpgdir; cpu->cpu_pgd = newpgdir;
/* If it was completely blank, we map in the Guest kernel stack */ /* If it was completely blank, we map in the Guest kernel stack */
if (repin) if (repin)
pin_stack_pages(lg); pin_stack_pages(cpu);
} }
/*H:470 Finally, a routine which throws away everything: all PGD entries in all /*H:470 Finally, a routine which throws away everything: all PGD entries in all
@@ -468,11 +469,11 @@ static void release_all_pagetables(struct lguest *lg)
* mapping. Since kernel mappings are in every page table, it's easiest to * mapping. Since kernel mappings are in every page table, it's easiest to
* throw them all away. This traps the Guest in amber for a while as * throw them all away. This traps the Guest in amber for a while as
* everything faults back in, but it's rare. */ * everything faults back in, but it's rare. */
void guest_pagetable_clear_all(struct lguest *lg) void guest_pagetable_clear_all(struct lg_cpu *cpu)
{ {
release_all_pagetables(lg); release_all_pagetables(cpu->lg);
/* We need the Guest kernel stack mapped again. */ /* We need the Guest kernel stack mapped again. */
pin_stack_pages(lg); pin_stack_pages(cpu);
} }
/*:*/ /*:*/
/*M:009 Since we throw away all mappings when a kernel mapping changes, our /*M:009 Since we throw away all mappings when a kernel mapping changes, our
@@ -497,24 +498,24 @@ void guest_pagetable_clear_all(struct lguest *lg)
* _PAGE_ACCESSED then we can put a read-only PTE entry in immediately, and if * _PAGE_ACCESSED then we can put a read-only PTE entry in immediately, and if
* they set _PAGE_DIRTY then we can put a writable PTE entry in immediately. * they set _PAGE_DIRTY then we can put a writable PTE entry in immediately.
*/ */
static void do_set_pte(struct lguest *lg, int idx, static void do_set_pte(struct lg_cpu *cpu, int idx,
unsigned long vaddr, pte_t gpte) unsigned long vaddr, pte_t gpte)
{ {
/* Look up the matching shadow page directory entry. */ /* Look up the matching shadow page directory entry. */
pgd_t *spgd = spgd_addr(lg, idx, vaddr); pgd_t *spgd = spgd_addr(cpu, idx, vaddr);
/* If the top level isn't present, there's no entry to update. */ /* If the top level isn't present, there's no entry to update. */
if (pgd_flags(*spgd) & _PAGE_PRESENT) { if (pgd_flags(*spgd) & _PAGE_PRESENT) {
/* Otherwise, we start by releasing the existing entry. */ /* Otherwise, we start by releasing the existing entry. */
pte_t *spte = spte_addr(lg, *spgd, vaddr); pte_t *spte = spte_addr(*spgd, vaddr);
release_pte(*spte); release_pte(*spte);
/* If they're setting this entry as dirty or accessed, we might /* If they're setting this entry as dirty or accessed, we might
* as well put that entry they've given us in now. This shaves * as well put that entry they've given us in now. This shaves
* 10% off a copy-on-write micro-benchmark. */ * 10% off a copy-on-write micro-benchmark. */
if (pte_flags(gpte) & (_PAGE_DIRTY | _PAGE_ACCESSED)) { if (pte_flags(gpte) & (_PAGE_DIRTY | _PAGE_ACCESSED)) {
check_gpte(lg, gpte); check_gpte(cpu, gpte);
*spte = gpte_to_spte(lg, gpte, *spte = gpte_to_spte(cpu, gpte,
pte_flags(gpte) & _PAGE_DIRTY); pte_flags(gpte) & _PAGE_DIRTY);
} else } else
/* Otherwise kill it and we can demand_page() it in /* Otherwise kill it and we can demand_page() it in
@@ -533,22 +534,22 @@ static void do_set_pte(struct lguest *lg, int idx,
* *
* The benefit is that when we have to track a new page table, we can copy keep * The benefit is that when we have to track a new page table, we can copy keep
* all the kernel mappings. This speeds up context switch immensely. */ * all the kernel mappings. This speeds up context switch immensely. */
void guest_set_pte(struct lguest *lg, void guest_set_pte(struct lg_cpu *cpu,
unsigned long gpgdir, unsigned long vaddr, pte_t gpte) unsigned long gpgdir, unsigned long vaddr, pte_t gpte)
{ {
/* Kernel mappings must be changed on all top levels. Slow, but /* Kernel mappings must be changed on all top levels. Slow, but
* doesn't happen often. */ * doesn't happen often. */
if (vaddr >= lg->kernel_address) { if (vaddr >= cpu->lg->kernel_address) {
unsigned int i; unsigned int i;
for (i = 0; i < ARRAY_SIZE(lg->pgdirs); i++) for (i = 0; i < ARRAY_SIZE(cpu->lg->pgdirs); i++)
if (lg->pgdirs[i].pgdir) if (cpu->lg->pgdirs[i].pgdir)
do_set_pte(lg, i, vaddr, gpte); do_set_pte(cpu, i, vaddr, gpte);
} else { } else {
/* Is this page table one we have a shadow for? */ /* Is this page table one we have a shadow for? */
int pgdir = find_pgdir(lg, gpgdir); int pgdir = find_pgdir(cpu->lg, gpgdir);
if (pgdir != ARRAY_SIZE(lg->pgdirs)) if (pgdir != ARRAY_SIZE(cpu->lg->pgdirs))
/* If so, do the update. */ /* If so, do the update. */
do_set_pte(lg, pgdir, vaddr, gpte); do_set_pte(cpu, pgdir, vaddr, gpte);
} }
} }
@@ -590,30 +591,32 @@ int init_guest_pagetable(struct lguest *lg, unsigned long pgtable)
{ {
/* We start on the first shadow page table, and give it a blank PGD /* We start on the first shadow page table, and give it a blank PGD
* page. */ * page. */
lg->pgdidx = 0; lg->pgdirs[0].gpgdir = pgtable;
lg->pgdirs[lg->pgdidx].gpgdir = pgtable; lg->pgdirs[0].pgdir = (pgd_t *)get_zeroed_page(GFP_KERNEL);
lg->pgdirs[lg->pgdidx].pgdir = (pgd_t*)get_zeroed_page(GFP_KERNEL); if (!lg->pgdirs[0].pgdir)
if (!lg->pgdirs[lg->pgdidx].pgdir)
return -ENOMEM; return -ENOMEM;
lg->cpus[0].cpu_pgd = 0;
return 0; return 0;
} }
/* When the Guest calls LHCALL_LGUEST_INIT we do more setup. */ /* When the Guest calls LHCALL_LGUEST_INIT we do more setup. */
void page_table_guest_data_init(struct lguest *lg) void page_table_guest_data_init(struct lg_cpu *cpu)
{ {
/* We get the kernel address: above this is all kernel memory. */ /* We get the kernel address: above this is all kernel memory. */
if (get_user(lg->kernel_address, &lg->lguest_data->kernel_address) if (get_user(cpu->lg->kernel_address,
&cpu->lg->lguest_data->kernel_address)
/* We tell the Guest that it can't use the top 4MB of virtual /* We tell the Guest that it can't use the top 4MB of virtual
* addresses used by the Switcher. */ * addresses used by the Switcher. */
|| put_user(4U*1024*1024, &lg->lguest_data->reserve_mem) || put_user(4U*1024*1024, &cpu->lg->lguest_data->reserve_mem)
|| put_user(lg->pgdirs[lg->pgdidx].gpgdir,&lg->lguest_data->pgdir)) || put_user(cpu->lg->pgdirs[0].gpgdir, &cpu->lg->lguest_data->pgdir))
kill_guest(lg, "bad guest page %p", lg->lguest_data); kill_guest(cpu, "bad guest page %p", cpu->lg->lguest_data);
/* In flush_user_mappings() we loop from 0 to /* In flush_user_mappings() we loop from 0 to
* "pgd_index(lg->kernel_address)". This assumes it won't hit the * "pgd_index(lg->kernel_address)". This assumes it won't hit the
* Switcher mappings, so check that now. */ * Switcher mappings, so check that now. */
if (pgd_index(lg->kernel_address) >= SWITCHER_PGD_INDEX) if (pgd_index(cpu->lg->kernel_address) >= SWITCHER_PGD_INDEX)
kill_guest(lg, "bad kernel address %#lx", lg->kernel_address); kill_guest(cpu, "bad kernel address %#lx",
cpu->lg->kernel_address);
} }
/* When a Guest dies, our cleanup is fairly simple. */ /* When a Guest dies, our cleanup is fairly simple. */
@@ -634,17 +637,18 @@ void free_guest_pagetable(struct lguest *lg)
* Guest (and not the pages for other CPUs). We have the appropriate PTE pages * Guest (and not the pages for other CPUs). We have the appropriate PTE pages
* for each CPU already set up, we just need to hook them in now we know which * for each CPU already set up, we just need to hook them in now we know which
* Guest is about to run on this CPU. */ * Guest is about to run on this CPU. */
void map_switcher_in_guest(struct lguest *lg, struct lguest_pages *pages) void map_switcher_in_guest(struct lg_cpu *cpu, struct lguest_pages *pages)
{ {
pte_t *switcher_pte_page = __get_cpu_var(switcher_pte_pages); pte_t *switcher_pte_page = __get_cpu_var(switcher_pte_pages);
pgd_t switcher_pgd; pgd_t switcher_pgd;
pte_t regs_pte; pte_t regs_pte;
unsigned long pfn;
/* Make the last PGD entry for this Guest point to the Switcher's PTE /* Make the last PGD entry for this Guest point to the Switcher's PTE
* page for this CPU (with appropriate flags). */ * page for this CPU (with appropriate flags). */
switcher_pgd = __pgd(__pa(switcher_pte_page) | _PAGE_KERNEL); switcher_pgd = __pgd(__pa(switcher_pte_page) | __PAGE_KERNEL);
lg->pgdirs[lg->pgdidx].pgdir[SWITCHER_PGD_INDEX] = switcher_pgd; cpu->lg->pgdirs[cpu->cpu_pgd].pgdir[SWITCHER_PGD_INDEX] = switcher_pgd;
/* We also change the Switcher PTE page. When we're running the Guest, /* We also change the Switcher PTE page. When we're running the Guest,
* we want the Guest's "regs" page to appear where the first Switcher * we want the Guest's "regs" page to appear where the first Switcher
@@ -653,7 +657,8 @@ void map_switcher_in_guest(struct lguest *lg, struct lguest_pages *pages)
* CPU's "struct lguest_pages": if we make sure the Guest's register * CPU's "struct lguest_pages": if we make sure the Guest's register
* page is already mapped there, we don't have to copy them out * page is already mapped there, we don't have to copy them out
* again. */ * again. */
regs_pte = pfn_pte (__pa(lg->regs_page) >> PAGE_SHIFT, __pgprot(_PAGE_KERNEL)); pfn = __pa(cpu->regs_page) >> PAGE_SHIFT;
regs_pte = pfn_pte(pfn, __pgprot(__PAGE_KERNEL));
switcher_pte_page[(unsigned long)pages/PAGE_SIZE%PTRS_PER_PTE] = regs_pte; switcher_pte_page[(unsigned long)pages/PAGE_SIZE%PTRS_PER_PTE] = regs_pte;
} }
/*:*/ /*:*/

48
drivers/lguest/segments.c
View File

@@ -58,7 +58,7 @@ static int ignored_gdt(unsigned int num)
* Protection Fault in the Switcher when it restores a Guest segment register * Protection Fault in the Switcher when it restores a Guest segment register
* which tries to use that entry. Then we kill the Guest for causing such a * which tries to use that entry. Then we kill the Guest for causing such a
* mess: the message will be "unhandled trap 256". */ * mess: the message will be "unhandled trap 256". */
static void fixup_gdt_table(struct lguest *lg, unsigned start, unsigned end) static void fixup_gdt_table(struct lg_cpu *cpu, unsigned start, unsigned end)
{ {
unsigned int i; unsigned int i;
@@ -71,14 +71,14 @@ static void fixup_gdt_table(struct lguest *lg, unsigned start, unsigned end)
/* Segment descriptors contain a privilege level: the Guest is /* Segment descriptors contain a privilege level: the Guest is
* sometimes careless and leaves this as 0, even though it's * sometimes careless and leaves this as 0, even though it's
* running at privilege level 1. If so, we fix it here. */ * running at privilege level 1. If so, we fix it here. */
if ((lg->arch.gdt[i].b & 0x00006000) == 0) if ((cpu->arch.gdt[i].b & 0x00006000) == 0)
lg->arch.gdt[i].b |= (GUEST_PL << 13); cpu->arch.gdt[i].b |= (GUEST_PL << 13);
/* Each descriptor has an "accessed" bit. If we don't set it /* Each descriptor has an "accessed" bit. If we don't set it
* now, the CPU will try to set it when the Guest first loads * now, the CPU will try to set it when the Guest first loads
* that entry into a segment register. But the GDT isn't * that entry into a segment register. But the GDT isn't
* writable by the Guest, so bad things can happen. */ * writable by the Guest, so bad things can happen. */
lg->arch.gdt[i].b |= 0x00000100; cpu->arch.gdt[i].b |= 0x00000100;
} }
} }
@@ -109,31 +109,31 @@ void setup_default_gdt_entries(struct lguest_ro_state *state)
/* This routine sets up the initial Guest GDT for booting. All entries start /* This routine sets up the initial Guest GDT for booting. All entries start
* as 0 (unusable). */ * as 0 (unusable). */
void setup_guest_gdt(struct lguest *lg) void setup_guest_gdt(struct lg_cpu *cpu)
{ {
/* Start with full 0-4G segments... */ /* Start with full 0-4G segments... */
lg->arch.gdt[GDT_ENTRY_KERNEL_CS] = FULL_EXEC_SEGMENT; cpu->arch.gdt[GDT_ENTRY_KERNEL_CS] = FULL_EXEC_SEGMENT;
lg->arch.gdt[GDT_ENTRY_KERNEL_DS] = FULL_SEGMENT; cpu->arch.gdt[GDT_ENTRY_KERNEL_DS] = FULL_SEGMENT;
/* ...except the Guest is allowed to use them, so set the privilege /* ...except the Guest is allowed to use them, so set the privilege
* level appropriately in the flags. */ * level appropriately in the flags. */
lg->arch.gdt[GDT_ENTRY_KERNEL_CS].b |= (GUEST_PL << 13); cpu->arch.gdt[GDT_ENTRY_KERNEL_CS].b |= (GUEST_PL << 13);
lg->arch.gdt[GDT_ENTRY_KERNEL_DS].b |= (GUEST_PL << 13); cpu->arch.gdt[GDT_ENTRY_KERNEL_DS].b |= (GUEST_PL << 13);
} }
/*H:650 An optimization of copy_gdt(), for just the three "thead-local storage" /*H:650 An optimization of copy_gdt(), for just the three "thead-local storage"
* entries. */ * entries. */
void copy_gdt_tls(const struct lguest *lg, struct desc_struct *gdt) void copy_gdt_tls(const struct lg_cpu *cpu, struct desc_struct *gdt)
{ {
unsigned int i; unsigned int i;
for (i = GDT_ENTRY_TLS_MIN; i <= GDT_ENTRY_TLS_MAX; i++) for (i = GDT_ENTRY_TLS_MIN; i <= GDT_ENTRY_TLS_MAX; i++)
gdt[i] = lg->arch.gdt[i]; gdt[i] = cpu->arch.gdt[i];
} }
/*H:640 When the Guest is run on a different CPU, or the GDT entries have /*H:640 When the Guest is run on a different CPU, or the GDT entries have
* changed, copy_gdt() is called to copy the Guest's GDT entries across to this * changed, copy_gdt() is called to copy the Guest's GDT entries across to this
* CPU's GDT. */ * CPU's GDT. */
void copy_gdt(const struct lguest *lg, struct desc_struct *gdt) void copy_gdt(const struct lg_cpu *cpu, struct desc_struct *gdt)
{ {
unsigned int i; unsigned int i;
@@ -141,38 +141,38 @@ void copy_gdt(const struct lguest *lg, struct desc_struct *gdt)
* replaced. See ignored_gdt() above. */ * replaced. See ignored_gdt() above. */
for (i = 0; i < GDT_ENTRIES; i++) for (i = 0; i < GDT_ENTRIES; i++)
if (!ignored_gdt(i)) if (!ignored_gdt(i))
gdt[i] = lg->arch.gdt[i]; gdt[i] = cpu->arch.gdt[i];
} }
/*H:620 This is where the Guest asks us to load a new GDT (LHCALL_LOAD_GDT). /*H:620 This is where the Guest asks us to load a new GDT (LHCALL_LOAD_GDT).
* We copy it from the Guest and tweak the entries. */ * We copy it from the Guest and tweak the entries. */
void load_guest_gdt(struct lguest *lg, unsigned long table, u32 num) void load_guest_gdt(struct lg_cpu *cpu, unsigned long table, u32 num)
{ {
/* We assume the Guest has the same number of GDT entries as the /* We assume the Guest has the same number of GDT entries as the
* Host, otherwise we'd have to dynamically allocate the Guest GDT. */ * Host, otherwise we'd have to dynamically allocate the Guest GDT. */
if (num > ARRAY_SIZE(lg->arch.gdt)) if (num > ARRAY_SIZE(cpu->arch.gdt))
kill_guest(lg, "too many gdt entries %i", num); kill_guest(cpu, "too many gdt entries %i", num);
/* We read the whole thing in, then fix it up. */ /* We read the whole thing in, then fix it up. */
__lgread(lg, lg->arch.gdt, table, num * sizeof(lg->arch.gdt[0])); __lgread(cpu, cpu->arch.gdt, table, num * sizeof(cpu->arch.gdt[0]));
fixup_gdt_table(lg, 0, ARRAY_SIZE(lg->arch.gdt)); fixup_gdt_table(cpu, 0, ARRAY_SIZE(cpu->arch.gdt));
/* Mark that the GDT changed so the core knows it has to copy it again, /* Mark that the GDT changed so the core knows it has to copy it again,
* even if the Guest is run on the same CPU. */ * even if the Guest is run on the same CPU. */
lg->changed |= CHANGED_GDT; cpu->changed |= CHANGED_GDT;
} }
/* This is the fast-track version for just changing the three TLS entries. /* This is the fast-track version for just changing the three TLS entries.
* Remember that this happens on every context switch, so it's worth * Remember that this happens on every context switch, so it's worth
* optimizing. But wouldn't it be neater to have a single hypercall to cover * optimizing. But wouldn't it be neater to have a single hypercall to cover
* both cases? */ * both cases? */
void guest_load_tls(struct lguest *lg, unsigned long gtls) void guest_load_tls(struct lg_cpu *cpu, unsigned long gtls)
{ {
struct desc_struct *tls = &lg->arch.gdt[GDT_ENTRY_TLS_MIN]; struct desc_struct *tls = &cpu->arch.gdt[GDT_ENTRY_TLS_MIN];
__lgread(lg, tls, gtls, sizeof(*tls)*GDT_ENTRY_TLS_ENTRIES); __lgread(cpu, tls, gtls, sizeof(*tls)*GDT_ENTRY_TLS_ENTRIES);
fixup_gdt_table(lg, GDT_ENTRY_TLS_MIN, GDT_ENTRY_TLS_MAX+1); fixup_gdt_table(cpu, GDT_ENTRY_TLS_MIN, GDT_ENTRY_TLS_MAX+1);
/* Note that just the TLS entries have changed. */ /* Note that just the TLS entries have changed. */
lg->changed |= CHANGED_GDT_TLS; cpu->changed |= CHANGED_GDT_TLS;
} }
/*:*/ /*:*/

127
drivers/lguest/x86/core.c
View File

@@ -60,7 +60,7 @@ static struct lguest_pages *lguest_pages(unsigned int cpu)
(SWITCHER_ADDR + SHARED_SWITCHER_PAGES*PAGE_SIZE))[cpu]); (SWITCHER_ADDR + SHARED_SWITCHER_PAGES*PAGE_SIZE))[cpu]);
} }
static DEFINE_PER_CPU(struct lguest *, last_guest); static DEFINE_PER_CPU(struct lg_cpu *, last_cpu);
/*S:010 /*S:010
* We approach the Switcher. * We approach the Switcher.
@@ -73,16 +73,16 @@ static DEFINE_PER_CPU(struct lguest *, last_guest);
* since it last ran. We saw this set in interrupts_and_traps.c and * since it last ran. We saw this set in interrupts_and_traps.c and
* segments.c. * segments.c.
*/ */
static void copy_in_guest_info(struct lguest *lg, struct lguest_pages *pages) static void copy_in_guest_info(struct lg_cpu *cpu, struct lguest_pages *pages)
{ {
/* Copying all this data can be quite expensive. We usually run the /* Copying all this data can be quite expensive. We usually run the
* same Guest we ran last time (and that Guest hasn't run anywhere else * same Guest we ran last time (and that Guest hasn't run anywhere else
* meanwhile). If that's not the case, we pretend everything in the * meanwhile). If that's not the case, we pretend everything in the
* Guest has changed. */ * Guest has changed. */
if (__get_cpu_var(last_guest) != lg || lg->last_pages != pages) { if (__get_cpu_var(last_cpu) != cpu || cpu->last_pages != pages) {
__get_cpu_var(last_guest) = lg; __get_cpu_var(last_cpu) = cpu;
lg->last_pages = pages; cpu->last_pages = pages;
lg->changed = CHANGED_ALL; cpu->changed = CHANGED_ALL;
} }
/* These copies are pretty cheap, so we do them unconditionally: */ /* These copies are pretty cheap, so we do them unconditionally: */
@@ -90,42 +90,42 @@ static void copy_in_guest_info(struct lguest *lg, struct lguest_pages *pages)
pages->state.host_cr3 = __pa(current->mm->pgd); pages->state.host_cr3 = __pa(current->mm->pgd);
/* Set up the Guest's page tables to see this CPU's pages (and no /* Set up the Guest's page tables to see this CPU's pages (and no
* other CPU's pages). */ * other CPU's pages). */
map_switcher_in_guest(lg, pages); map_switcher_in_guest(cpu, pages);
/* Set up the two "TSS" members which tell the CPU what stack to use /* Set up the two "TSS" members which tell the CPU what stack to use
* for traps which do directly into the Guest (ie. traps at privilege * for traps which do directly into the Guest (ie. traps at privilege
* level 1). */ * level 1). */
pages->state.guest_tss.sp1 = lg->esp1; pages->state.guest_tss.esp1 = cpu->esp1;
pages->state.guest_tss.ss1 = lg->ss1; pages->state.guest_tss.ss1 = cpu->ss1;
/* Copy direct-to-Guest trap entries. */ /* Copy direct-to-Guest trap entries. */
if (lg->changed & CHANGED_IDT) if (cpu->changed & CHANGED_IDT)
copy_traps(lg, pages->state.guest_idt, default_idt_entries); copy_traps(cpu, pages->state.guest_idt, default_idt_entries);
/* Copy all GDT entries which the Guest can change. */ /* Copy all GDT entries which the Guest can change. */
if (lg->changed & CHANGED_GDT) if (cpu->changed & CHANGED_GDT)
copy_gdt(lg, pages->state.guest_gdt); copy_gdt(cpu, pages->state.guest_gdt);
/* If only the TLS entries have changed, copy them. */ /* If only the TLS entries have changed, copy them. */
else if (lg->changed & CHANGED_GDT_TLS) else if (cpu->changed & CHANGED_GDT_TLS)
copy_gdt_tls(lg, pages->state.guest_gdt); copy_gdt_tls(cpu, pages->state.guest_gdt);
/* Mark the Guest as unchanged for next time. */ /* Mark the Guest as unchanged for next time. */
lg->changed = 0; cpu->changed = 0;
} }
/* Finally: the code to actually call into the Switcher to run the Guest. */ /* Finally: the code to actually call into the Switcher to run the Guest. */
static void run_guest_once(struct lguest *lg, struct lguest_pages *pages) static void run_guest_once(struct lg_cpu *cpu, struct lguest_pages *pages)
{ {
/* This is a dummy value we need for GCC's sake. */ /* This is a dummy value we need for GCC's sake. */
unsigned int clobber; unsigned int clobber;
/* Copy the guest-specific information into this CPU's "struct /* Copy the guest-specific information into this CPU's "struct
* lguest_pages". */ * lguest_pages". */
copy_in_guest_info(lg, pages); copy_in_guest_info(cpu, pages);
/* Set the trap number to 256 (impossible value). If we fault while /* Set the trap number to 256 (impossible value). If we fault while
* switching to the Guest (bad segment registers or bug), this will * switching to the Guest (bad segment registers or bug), this will
* cause us to abort the Guest. */ * cause us to abort the Guest. */
lg->regs->trapnum = 256; cpu->regs->trapnum = 256;
/* Now: we push the "eflags" register on the stack, then do an "lcall". /* Now: we push the "eflags" register on the stack, then do an "lcall".
* This is how we change from using the kernel code segment to using * This is how we change from using the kernel code segment to using
@@ -143,7 +143,7 @@ static void run_guest_once(struct lguest *lg, struct lguest_pages *pages)
* 0-th argument above, ie "a"). %ebx contains the * 0-th argument above, ie "a"). %ebx contains the
* physical address of the Guest's top-level page * physical address of the Guest's top-level page
* directory. */ * directory. */
: "0"(pages), "1"(__pa(lg->pgdirs[lg->pgdidx].pgdir)) : "0"(pages), "1"(__pa(cpu->lg->pgdirs[cpu->cpu_pgd].pgdir))
/* We tell gcc that all these registers could change, /* We tell gcc that all these registers could change,
* which means we don't have to save and restore them in * which means we don't have to save and restore them in
* the Switcher. */ * the Switcher. */
@@ -161,12 +161,12 @@ static void run_guest_once(struct lguest *lg, struct lguest_pages *pages)
/*H:040 This is the i386-specific code to setup and run the Guest. Interrupts /*H:040 This is the i386-specific code to setup and run the Guest. Interrupts
* are disabled: we own the CPU. */ * are disabled: we own the CPU. */
void lguest_arch_run_guest(struct lguest *lg) void lguest_arch_run_guest(struct lg_cpu *cpu)
{ {
/* Remember the awfully-named TS bit? If the Guest has asked to set it /* Remember the awfully-named TS bit? If the Guest has asked to set it
* we set it now, so we can trap and pass that trap to the Guest if it * we set it now, so we can trap and pass that trap to the Guest if it
* uses the FPU. */ * uses the FPU. */
if (lg->ts) if (cpu->ts)
lguest_set_ts(); lguest_set_ts();
/* SYSENTER is an optimized way of doing system calls. We can't allow /* SYSENTER is an optimized way of doing system calls. We can't allow
@@ -180,7 +180,7 @@ void lguest_arch_run_guest(struct lguest *lg)
/* Now we actually run the Guest. It will return when something /* Now we actually run the Guest. It will return when something
* interesting happens, and we can examine its registers to see what it * interesting happens, and we can examine its registers to see what it
* was doing. */ * was doing. */
run_guest_once(lg, lguest_pages(raw_smp_processor_id())); run_guest_once(cpu, lguest_pages(raw_smp_processor_id()));
/* Note that the "regs" pointer contains two extra entries which are /* Note that the "regs" pointer contains two extra entries which are
* not really registers: a trap number which says what interrupt or * not really registers: a trap number which says what interrupt or
@@ -191,11 +191,11 @@ void lguest_arch_run_guest(struct lguest *lg)
* bad virtual address. We have to grab this now, because once we * bad virtual address. We have to grab this now, because once we
* re-enable interrupts an interrupt could fault and thus overwrite * re-enable interrupts an interrupt could fault and thus overwrite
* cr2, or we could even move off to a different CPU. */ * cr2, or we could even move off to a different CPU. */
if (lg->regs->trapnum == 14) if (cpu->regs->trapnum == 14)
lg->arch.last_pagefault = read_cr2(); cpu->arch.last_pagefault = read_cr2();
/* Similarly, if we took a trap because the Guest used the FPU, /* Similarly, if we took a trap because the Guest used the FPU,
* we have to restore the FPU it expects to see. */ * we have to restore the FPU it expects to see. */
else if (lg->regs->trapnum == 7) else if (cpu->regs->trapnum == 7)
math_state_restore(); math_state_restore();
/* Restore SYSENTER if it's supposed to be on. */ /* Restore SYSENTER if it's supposed to be on. */
@@ -214,22 +214,22 @@ void lguest_arch_run_guest(struct lguest *lg)
* When the Guest uses one of these instructions, we get a trap (General * When the Guest uses one of these instructions, we get a trap (General
* Protection Fault) and come here. We see if it's one of those troublesome * Protection Fault) and come here. We see if it's one of those troublesome
* instructions and skip over it. We return true if we did. */ * instructions and skip over it. We return true if we did. */
static int emulate_insn(struct lguest *lg) static int emulate_insn(struct lg_cpu *cpu)
{ {
u8 insn; u8 insn;
unsigned int insnlen = 0, in = 0, shift = 0; unsigned int insnlen = 0, in = 0, shift = 0;
/* The eip contains the *virtual* address of the Guest's instruction: /* The eip contains the *virtual* address of the Guest's instruction:
* guest_pa just subtracts the Guest's page_offset. */ * guest_pa just subtracts the Guest's page_offset. */
unsigned long physaddr = guest_pa(lg, lg->regs->eip); unsigned long physaddr = guest_pa(cpu, cpu->regs->eip);
/* This must be the Guest kernel trying to do something, not userspace! /* This must be the Guest kernel trying to do something, not userspace!
* The bottom two bits of the CS segment register are the privilege * The bottom two bits of the CS segment register are the privilege
* level. */ * level. */
if ((lg->regs->cs & 3) != GUEST_PL) if ((cpu->regs->cs & 3) != GUEST_PL)
return 0; return 0;
/* Decoding x86 instructions is icky. */ /* Decoding x86 instructions is icky. */
insn = lgread(lg, physaddr, u8); insn = lgread(cpu, physaddr, u8);
/* 0x66 is an "operand prefix". It means it's using the upper 16 bits /* 0x66 is an "operand prefix". It means it's using the upper 16 bits
of the eax register. */ of the eax register. */
@@ -237,7 +237,7 @@ static int emulate_insn(struct lguest *lg)
shift = 16; shift = 16;
/* The instruction is 1 byte so far, read the next byte. */ /* The instruction is 1 byte so far, read the next byte. */
insnlen = 1; insnlen = 1;
insn = lgread(lg, physaddr + insnlen, u8); insn = lgread(cpu, physaddr + insnlen, u8);
} }
/* We can ignore the lower bit for the moment and decode the 4 opcodes /* We can ignore the lower bit for the moment and decode the 4 opcodes
@@ -268,26 +268,26 @@ static int emulate_insn(struct lguest *lg)
if (in) { if (in) {
/* Lower bit tells is whether it's a 16 or 32 bit access */ /* Lower bit tells is whether it's a 16 or 32 bit access */
if (insn & 0x1) if (insn & 0x1)
lg->regs->eax = 0xFFFFFFFF; cpu->regs->eax = 0xFFFFFFFF;
else else
lg->regs->eax |= (0xFFFF << shift); cpu->regs->eax |= (0xFFFF << shift);
} }
/* Finally, we've "done" the instruction, so move past it. */ /* Finally, we've "done" the instruction, so move past it. */
lg->regs->eip += insnlen; cpu->regs->eip += insnlen;
/* Success! */ /* Success! */
return 1; return 1;
} }
/*H:050 Once we've re-enabled interrupts, we look at why the Guest exited. */ /*H:050 Once we've re-enabled interrupts, we look at why the Guest exited. */
void lguest_arch_handle_trap(struct lguest *lg) void lguest_arch_handle_trap(struct lg_cpu *cpu)
{ {
switch (lg->regs->trapnum) { switch (cpu->regs->trapnum) {
case 13: /* We've intercepted a General Protection Fault. */ case 13: /* We've intercepted a General Protection Fault. */
/* Check if this was one of those annoying IN or OUT /* Check if this was one of those annoying IN or OUT
* instructions which we need to emulate. If so, we just go * instructions which we need to emulate. If so, we just go
* back into the Guest after we've done it. */ * back into the Guest after we've done it. */
if (lg->regs->errcode == 0) { if (cpu->regs->errcode == 0) {
if (emulate_insn(lg)) if (emulate_insn(cpu))
return; return;
} }
break; break;
@@ -301,7 +301,8 @@ void lguest_arch_handle_trap(struct lguest *lg)
* *
* The errcode tells whether this was a read or a write, and * The errcode tells whether this was a read or a write, and
* whether kernel or userspace code. */ * whether kernel or userspace code. */
if (demand_page(lg, lg->arch.last_pagefault, lg->regs->errcode)) if (demand_page(cpu, cpu->arch.last_pagefault,
cpu->regs->errcode))
return; return;
/* OK, it's really not there (or not OK): the Guest needs to /* OK, it's really not there (or not OK): the Guest needs to
@@ -311,15 +312,16 @@ void lguest_arch_handle_trap(struct lguest *lg)
* Note that if the Guest were really messed up, this could * Note that if the Guest were really messed up, this could
* happen before it's done the LHCALL_LGUEST_INIT hypercall, so * happen before it's done the LHCALL_LGUEST_INIT hypercall, so
* lg->lguest_data could be NULL */ * lg->lguest_data could be NULL */
if (lg->lguest_data && if (cpu->lg->lguest_data &&
put_user(lg->arch.last_pagefault, &lg->lguest_data->cr2)) put_user(cpu->arch.last_pagefault,
kill_guest(lg, "Writing cr2"); &cpu->lg->lguest_data->cr2))
kill_guest(cpu, "Writing cr2");
break; break;
case 7: /* We've intercepted a Device Not Available fault. */ case 7: /* We've intercepted a Device Not Available fault. */
/* If the Guest doesn't want to know, we already restored the /* If the Guest doesn't want to know, we already restored the
* Floating Point Unit, so we just continue without telling * Floating Point Unit, so we just continue without telling
* it. */ * it. */
if (!lg->ts) if (!cpu->ts)
return; return;
break; break;
case 32 ... 255: case 32 ... 255:
@@ -332,19 +334,19 @@ void lguest_arch_handle_trap(struct lguest *lg)
case LGUEST_TRAP_ENTRY: case LGUEST_TRAP_ENTRY:
/* Our 'struct hcall_args' maps directly over our regs: we set /* Our 'struct hcall_args' maps directly over our regs: we set
* up the pointer now to indicate a hypercall is pending. */ * up the pointer now to indicate a hypercall is pending. */
lg->hcall = (struct hcall_args *)lg->regs; cpu->hcall = (struct hcall_args *)cpu->regs;
return; return;
} }
/* We didn't handle the trap, so it needs to go to the Guest. */ /* We didn't handle the trap, so it needs to go to the Guest. */
if (!deliver_trap(lg, lg->regs->trapnum)) if (!deliver_trap(cpu, cpu->regs->trapnum))
/* If the Guest doesn't have a handler (either it hasn't /* If the Guest doesn't have a handler (either it hasn't
* registered any yet, or it's one of the faults we don't let * registered any yet, or it's one of the faults we don't let
* it handle), it dies with a cryptic error message. */ * it handle), it dies with a cryptic error message. */
kill_guest(lg, "unhandled trap %li at %#lx (%#lx)", kill_guest(cpu, "unhandled trap %li at %#lx (%#lx)",
lg->regs->trapnum, lg->regs->eip, cpu->regs->trapnum, cpu->regs->eip,
lg->regs->trapnum == 14 ? lg->arch.last_pagefault cpu->regs->trapnum == 14 ? cpu->arch.last_pagefault
: lg->regs->errcode); : cpu->regs->errcode);
} }
/* Now we can look at each of the routines this calls, in increasing order of /* Now we can look at each of the routines this calls, in increasing order of
@@ -487,17 +489,17 @@ void __exit lguest_arch_host_fini(void)
/*H:122 The i386-specific hypercalls simply farm out to the right functions. */ /*H:122 The i386-specific hypercalls simply farm out to the right functions. */
int lguest_arch_do_hcall(struct lguest *lg, struct hcall_args *args) int lguest_arch_do_hcall(struct lg_cpu *cpu, struct hcall_args *args)
{ {
switch (args->arg0) { switch (args->arg0) {
case LHCALL_LOAD_GDT: case LHCALL_LOAD_GDT:
load_guest_gdt(lg, args->arg1, args->arg2); load_guest_gdt(cpu, args->arg1, args->arg2);
break; break;
case LHCALL_LOAD_IDT_ENTRY: case LHCALL_LOAD_IDT_ENTRY:
load_guest_idt_entry(lg, args->arg1, args->arg2, args->arg3); load_guest_idt_entry(cpu, args->arg1, args->arg2, args->arg3);
break; break;
case LHCALL_LOAD_TLS: case LHCALL_LOAD_TLS:
guest_load_tls(lg, args->arg1); guest_load_tls(cpu, args->arg1);
break; break;
default: default:
/* Bad Guest. Bad! */ /* Bad Guest. Bad! */
@@ -507,13 +509,14 @@ int lguest_arch_do_hcall(struct lguest *lg, struct hcall_args *args)
} }
/*H:126 i386-specific hypercall initialization: */ /*H:126 i386-specific hypercall initialization: */
int lguest_arch_init_hypercalls(struct lguest *lg) int lguest_arch_init_hypercalls(struct lg_cpu *cpu)
{ {
u32 tsc_speed; u32 tsc_speed;
/* The pointer to the Guest's "struct lguest_data" is the only /* The pointer to the Guest's "struct lguest_data" is the only
* argument. We check that address now. */ * argument. We check that address now. */
if (!lguest_address_ok(lg, lg->hcall->arg1, sizeof(*lg->lguest_data))) if (!lguest_address_ok(cpu->lg, cpu->hcall->arg1,
sizeof(*cpu->lg->lguest_data)))
return -EFAULT; return -EFAULT;
/* Having checked it, we simply set lg->lguest_data to point straight /* Having checked it, we simply set lg->lguest_data to point straight
@@ -521,7 +524,7 @@ int lguest_arch_init_hypercalls(struct lguest *lg)
* copy_to_user/from_user from now on, instead of lgread/write. I put * copy_to_user/from_user from now on, instead of lgread/write. I put
* this in to show that I'm not immune to writing stupid * this in to show that I'm not immune to writing stupid
* optimizations. */ * optimizations. */
lg->lguest_data = lg->mem_base + lg->hcall->arg1; cpu->lg->lguest_data = cpu->lg->mem_base + cpu->hcall->arg1;
/* We insist that the Time Stamp Counter exist and doesn't change with /* We insist that the Time Stamp Counter exist and doesn't change with
* cpu frequency. Some devious chip manufacturers decided that TSC * cpu frequency. Some devious chip manufacturers decided that TSC
@@ -534,12 +537,12 @@ int lguest_arch_init_hypercalls(struct lguest *lg)
tsc_speed = tsc_khz; tsc_speed = tsc_khz;
else else
tsc_speed = 0; tsc_speed = 0;
if (put_user(tsc_speed, &lg->lguest_data->tsc_khz)) if (put_user(tsc_speed, &cpu->lg->lguest_data->tsc_khz))
return -EFAULT; return -EFAULT;
/* The interrupt code might not like the system call vector. */ /* The interrupt code might not like the system call vector. */
if (!check_syscall_vector(lg)) if (!check_syscall_vector(cpu->lg))
kill_guest(lg, "bad syscall vector"); kill_guest(cpu, "bad syscall vector");
return 0; return 0;
} }
@@ -548,9 +551,9 @@ int lguest_arch_init_hypercalls(struct lguest *lg)
* *
* Most of the Guest's registers are left alone: we used get_zeroed_page() to * Most of the Guest's registers are left alone: we used get_zeroed_page() to
* allocate the structure, so they will be 0. */ * allocate the structure, so they will be 0. */
void lguest_arch_setup_regs(struct lguest *lg, unsigned long start) void lguest_arch_setup_regs(struct lg_cpu *cpu, unsigned long start)
{ {
struct lguest_regs *regs = lg->regs; struct lguest_regs *regs = cpu->regs;
/* There are four "segment" registers which the Guest needs to boot: /* There are four "segment" registers which the Guest needs to boot:
* The "code segment" register (cs) refers to the kernel code segment * The "code segment" register (cs) refers to the kernel code segment
@@ -577,5 +580,5 @@ void lguest_arch_setup_regs(struct lguest *lg, unsigned long start)
/* There are a couple of GDT entries the Guest expects when first /* There are a couple of GDT entries the Guest expects when first
* booting. */ * booting. */
setup_guest_gdt(lg); setup_guest_gdt(cpu);
} }

2
include/asm-x86/lguest.h
View File

@@ -56,7 +56,7 @@ struct lguest_ro_state
struct desc_struct guest_gdt[GDT_ENTRIES]; struct desc_struct guest_gdt[GDT_ENTRIES];
}; };
struct lguest_arch struct lg_cpu_arch
{ {
/* The GDT entries copied into lguest_ro_state when running. */ /* The GDT entries copied into lguest_ro_state when running. */
struct desc_struct gdt[GDT_ENTRIES]; struct desc_struct gdt[GDT_ENTRIES];

6
include/asm-x86/lguest_hcall.h
View File

@@ -4,7 +4,7 @@
#define LHCALL_FLUSH_ASYNC 0 #define LHCALL_FLUSH_ASYNC 0
#define LHCALL_LGUEST_INIT 1 #define LHCALL_LGUEST_INIT 1
#define LHCALL_CRASH 2 #define LHCALL_SHUTDOWN 2
#define LHCALL_LOAD_GDT 3 #define LHCALL_LOAD_GDT 3
#define LHCALL_NEW_PGTABLE 4 #define LHCALL_NEW_PGTABLE 4
#define LHCALL_FLUSH_TLB 5 #define LHCALL_FLUSH_TLB 5
@@ -20,6 +20,10 @@
#define LGUEST_TRAP_ENTRY 0x1F #define LGUEST_TRAP_ENTRY 0x1F
/* Argument number 3 to LHCALL_LGUEST_SHUTDOWN */
#define LGUEST_SHUTDOWN_POWEROFF 1
#define LGUEST_SHUTDOWN_RESTART 2
#ifndef __ASSEMBLY__ #ifndef __ASSEMBLY__
#include <asm/hw_irq.h> #include <asm/hw_irq.h>