On Saturday 12 July 2008 07:57:43 ron minnich wrote:
> and here I thought I knew sometning about the x86.
>
> anybody know why an INT 3 would generate a gpf? Verified under both
> Plan 9 and Linux. I'm flummoxed.

Hmm, good q.  Tracked it down here.

OK, because it's not a direct trap (it's an interrupt gate), we don't copy the
ring level across: the result is that we use the default entry, and that's RPL
1 only.

How's this?
Rusty.

lguest: Guest int3 fix

Ron Minnich noticed that guest userspace gets a GPF when it tries to int3:
we need to copy the privilege level from the guest-supplied IDT to the real
IDT.  int3 is the only common case where guest userspace expects to invoke
an interrupt, so that's the symptom of failing to do this.

Signed-off-by: Rusty Russell <[EMAIL PROTECTED]>

diff -r aacc459aba83 drivers/lguest/interrupts_and_traps.c
--- a/drivers/lguest/interrupts_and_traps.c     Sat Jul 12 16:22:02 2008 +1000
+++ b/drivers/lguest/interrupts_and_traps.c     Sat Jul 12 17:43:59 2008 +1000
@@ -406,7 +406,8 @@ void load_guest_idt_entry(struct lg_cpu 
  * deliver_trap() to bounce it back into the Guest. */
 static void default_idt_entry(struct desc_struct *idt,
                              int trap,
-                             const unsigned long handler)
+                             const unsigned long handler,
+                             const struct desc_struct *base)
 {
        /* A present interrupt gate. */
        u32 flags = 0x8e00;
@@ -415,6 +416,10 @@ static void default_idt_entry(struct des
         * the Guest to use the "int" instruction to trigger it. */
        if (trap == LGUEST_TRAP_ENTRY)
                flags |= (GUEST_PL << 13);
+       else if (base)
+               /* Copy priv. level from what Guest asked for.  This allows
+                * debug (int 3) traps from Guest userspace, for example. */
+               flags |= (base->b & 0x6000);
 
        /* Now pack it into the IDT entry in its weird format. */
        idt->a = (LGUEST_CS<<16) | (handler&0x0000FFFF);
@@ -428,7 +433,7 @@ void setup_default_idt_entries(struct lg
        unsigned int i;
 
        for (i = 0; i < ARRAY_SIZE(state->guest_idt); i++)
-               default_idt_entry(&state->guest_idt[i], i, def[i]);
+               default_idt_entry(&state->guest_idt[i], i, def[i], NULL);
 }
 
 /*H:240 We don't use the IDT entries in the "struct lguest" directly, instead
@@ -442,6 +447,8 @@ void copy_traps(const struct lg_cpu *cpu
        /* We can simply copy the direct traps, otherwise we use the default
         * ones in the Switcher: they will return to the Host. */
        for (i = 0; i < ARRAY_SIZE(cpu->arch.idt); i++) {
+               const struct desc_struct *gidt = &cpu->arch.idt[i];
+
                /* If no Guest can ever override this trap, leave it alone. */
                if (!direct_trap(i))
                        continue;
@@ -449,12 +456,15 @@ void copy_traps(const struct lg_cpu *cpu
                /* Only trap gates (type 15) can go direct to the Guest.
                 * Interrupt gates (type 14) disable interrupts as they are
                 * entered, which we never let the Guest do.  Not present
-                * entries (type 0x0) also can't go direct, of course. */
-               if (idt_type(cpu->arch.idt[i].a, cpu->arch.idt[i].b) == 0xF)
-                       idt[i] = cpu->arch.idt[i];
+                * entries (type 0x0) also can't go direct, of course.
+                *
+                * If it can't go direct, we still need to copy the priv. level:
+                * they might want to give userspace access to a software
+                * interrupt. */
+               if (idt_type(gidt->a, gidt->b) == 0xF)
+                       idt[i] = *gidt;
                else
-                       /* Reset it to the default. */
-                       default_idt_entry(&idt[i], i, def[i]);
+                       default_idt_entry(&idt[i], i, def[i], gidt);
        }
 }
 
_______________________________________________
Lguest mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/lguest

Reply via email to