-/*P:900 This is the Switcher: code which sits at 0xFFC00000 to do the low-level
- * Guest<->Host switch. It is as simple as it can be made, but it's naturally
- * very specific to x86.
+/*P:900 This is the Switcher: code which sits at 0xFFC00000 astride both the
+ * Host and Guest to do the low-level Guest<->Host switch. It is as simple as
+ * it can be made, but it's naturally very specific to x86.
*
* You have now completed Preparation. If this has whet your appetite; if you
* are feeling invigorated and refreshed then the next, more challenging stage
* can be found in "make Guest". :*/
+/*M:012 Lguest is meant to be simple: my rule of thumb is that 1% more LOC must
+ * gain at least 1% more performance. Since neither LOC nor performance can be
+ * measured beforehand, it generally means implementing a feature then deciding
+ * if it's worth it. And once it's implemented, who can say no?
+ *
+ * This is why I haven't implemented this idea myself. I want to, but I
+ * haven't. You could, though.
+ *
+ * The main place where lguest performance sucks is Guest page faulting. When
+ * a Guest userspace process hits an unmapped page we switch back to the Host,
+ * walk the page tables, find it's not mapped, switch back to the Guest page
+ * fault handler, which calls a hypercall to set the page table entry, then
+ * finally returns to userspace. That's two round-trips.
+ *
+ * If we had a small walker in the Switcher, we could quickly check the Guest
+ * page table and if the page isn't mapped, immediately reflect the fault back
+ * into the Guest. This means the Switcher would have to know the top of the
+ * Guest page table and the page fault handler address.
+ *
+ * For simplicity, the Guest should only handle the case where the privilege
+ * level of the fault is 3 and probably only not present or write faults. It
+ * should also detect recursive faults, and hand the original fault to the
+ * Host (which is actually really easy).
+ *
+ * Two questions remain. Would the performance gain outweigh the complexity?
+ * And who would write the verse documenting it? :*/
+
+/*M:011 Lguest64 handles NMI. This gave me NMI envy (until I looked at their
+ * code). It's worth doing though, since it would let us use oprofile in the
+ * Host when a Guest is running. :*/
+
/*S:100
* Welcome to the Switcher itself!
*
// All saved and there's now five steps before us:
// Stack, GDT, IDT, TSS
- // And last of all the page tables are flipped.
+ // Then last of all the page tables are flipped.
// Yet beware that our stack pointer must be
// Always valid lest an NMI hits
lgdt LGUEST_PAGES_guest_gdt_desc(%eax)
// The Guest's IDT we did partially
- // Move to the "struct lguest_pages" as well.
+ // Copy to "struct lguest_pages" as well.
lidt LGUEST_PAGES_guest_idt_desc(%eax)
// The TSS entry which controls traps
// Must be loaded up with "ltr" now:
+ // The GDT entry that TSS uses
+ // Changes type when we load it: damn Intel!
// For after we switch over our page tables
- // It (as the rest) will be writable no more.
- // (The GDT entry TSS needs
- // Changes type when we load it: damn Intel!)
+ // That entry will be read-only: we'd crash.
movl $(GDT_ENTRY_TSS*8), %edx
ltr %dx
// Look back now, before we take this last step!
// The Host's TSS entry was also marked used;
- // Let's clear it again, ere we return.
+ // Let's clear it again for our return.
// The GDT descriptor of the Host
// Points to the table after two "size" bytes
movl (LGUEST_PAGES_host_gdt_desc+2)(%eax), %edx
- // Clear the type field of "used" (byte 5, bit 2)
+ // Clear "used" from type field (byte 5, bit 2)
andb $0xFD, (GDT_ENTRY_TSS*8 + 5)(%edx)
// Once our page table's switched, the Guest is live!
// The page table change did one tricky thing:
// The Guest's register page has been mapped
- // Writable onto our %esp (stack) --
+ // Writable under our %esp (stack) --
// We can simply pop off all Guest regs.
popl %eax
popl %ebx
addl $8, %esp
// The last five stack slots hold return address
- // And everything needed to change privilege
- // Into the Guest privilege level of 1,
+ // And everything needed to switch privilege
+ // From Switcher's level 0 to Guest's 1,
// And the stack where the Guest had last left it.
// Interrupts are turned back on: we are Guest.
iret
-// There are two paths where we switch to the Host
+// We tread two paths to switch back to the Host
+// Yet both must save Guest state and restore Host
// So we put the routine in a macro.
-// We are on our way home, back to the Host
-// Interrupted out of the Guest, we come here.
#define SWITCH_TO_HOST \
/* We save the Guest state: all registers first \
* Laid out just as "struct lguest_regs" defines */ \
movl %esp, %eax; \
andl $(~(1 << PAGE_SHIFT - 1)), %eax; \
/* Save our trap number: the switch will obscure it \
- * (The Guest regs are not mapped here in the Host) \
+ * (In the Host the Guest regs are not mapped here) \
* %ebx holds it safe for deliver_to_host */ \
movl LGUEST_PAGES_regs_trapnum(%eax), %ebx; \
/* The Host GDT, IDT and stack! \
/* Switch to Host's GDT, IDT. */ \
lgdt LGUEST_PAGES_host_gdt_desc(%eax); \
lidt LGUEST_PAGES_host_idt_desc(%eax); \
- /* Restore the Host's stack where it's saved regs lie */ \
+ /* Restore the Host's stack where its saved regs lie */ \
movl LGUEST_PAGES_host_sp(%eax), %esp; \
- /* Last the TSS: our Host is complete */ \
+ /* Last the TSS: our Host is returned */ \
movl $(GDT_ENTRY_TSS*8), %edx; \
ltr %dx; \
/* Restore now the regs saved right at the first. */ \
popl %ds; \
popl %es
-// Here's where we come when the Guest has just trapped:
-// (Which trap we'll see has been pushed on the stack).
+// The first path is trod when the Guest has trapped:
+// (Which trap it was has been pushed on the stack).
// We need only switch back, and the Host will decode
// Why we came home, and what needs to be done.
return_to_host:
SWITCH_TO_HOST
iret
+// We are lead to the second path like so:
// An interrupt, with some cause external
// Has ajerked us rudely from the Guest's code
// Again we must return home to the Host
// But now we must go home via that place
// Where that interrupt was supposed to go
// Had we not been ensconced, running the Guest.
- // Here we see the cleverness of our stack:
+ // Here we see the trickness of run_guest_once():
// The Host stack is formed like an interrupt
// With EIP, CS and EFLAGS layered.
// Interrupt handlers end with "iret"
xorw %ax, %ax
orl %eax, %edx
// Now the address of the handler's in %edx
- // We call it now: its "iret" takes us home.
+ // We call it now: its "iret" drops us home.
jmp *%edx
// Every interrupt can come to us here
projects / linux-2.6 / blobdiff