[Lguest] [PATCH 4/8] lguest: update commentry

Paul E. McKenney paulmck at linux.vnet.ibm.com
Fri Jul 24 09:00:11 EST 2009


On Thu, Jul 23, 2009 at 11:16:12PM +0930, Rusty Russell wrote:
> 
> Every so often, after code shuffles, I need to go through and unbitrot
> the Lguest Journey (see drivers/lguest/README).  Since we now use RCU in
> a simple form in one place I took the opportunity to expand that explanation.

Looks good!!!

Some comments inline.  Will you also be commenting the read side in
send_notify_to_eventfd()?

							Thanx, Paul

> Signed-off-by: Rusty Russell <rusty at rustcorp.com.au>
> Cc: Ingo Molnar <mingo at redhat.com>
> Cc: Paul McKenney <paulmck at linux.vnet.ibm.com>
> ---
>  Documentation/lguest/lguest.c       |  184 +++++++++++++++++++++++++++---------
>  arch/x86/include/asm/lguest_hcall.h |    8 -
>  arch/x86/lguest/boot.c              |   99 +++++++++++++++----
>  arch/x86/lguest/i386_head.S         |    2 
>  drivers/lguest/core.c               |    7 +
>  drivers/lguest/hypercalls.c         |    6 -
>  drivers/lguest/lguest_device.c      |   11 +-
>  drivers/lguest/lguest_user.c        |   47 ++++++++-
>  drivers/lguest/page_tables.c        |   84 ++++++++++++----
>  drivers/lguest/x86/core.c           |    2 
>  drivers/lguest/x86/switcher_32.S    |    6 -
>  11 files changed, 353 insertions(+), 103 deletions(-)
> 
> diff --git a/Documentation/lguest/lguest.c b/Documentation/lguest/lguest.c
> --- a/Documentation/lguest/lguest.c
> +++ b/Documentation/lguest/lguest.c
> @@ -49,7 +49,7 @@
>  #include "linux/virtio_ring.h"
>  #include "asm/bootparam.h"
>  /*L:110
> - * We can ignore the 39 include files we need for this program, but I do want
> + * We can ignore the 42 include files we need for this program, but I do want
>   * to draw attention to the use of kernel-style types.
>   *
>   * As Linus said, "C is a Spartan language, and so should your naming be."  I
> @@ -305,6 +305,11 @@ static void *map_zeroed_pages(unsigned i
>  		    PROT_READ|PROT_WRITE|PROT_EXEC, MAP_PRIVATE, fd, 0);
>  	if (addr == MAP_FAILED)
>  		err(1, "Mmaping %u pages of /dev/zero", num);
> +
> +	/*
> +	 * One neat mmap feature is that you can close the fd, and it
> +	 * stays mapped.
> +	 */
>  	close(fd);
> 
>  	return addr;
> @@ -557,7 +562,7 @@ static void tell_kernel(unsigned long st
>  }
>  /*:*/
> 
> -/*
> +/*L:200
>   * Device Handling.
>   *
>   * When the Guest gives us a buffer, it sends an array of addresses and sizes.
> @@ -608,7 +613,10 @@ static unsigned next_desc(struct vring_d
>  	return next;
>  }
> 
> -/* This actually sends the interrupt for this virtqueue */
> +/*
> + * This actually sends the interrupt for this virtqueue, if we've used a
> + * buffer.
> + */
>  static void trigger_irq(struct virtqueue *vq)
>  {
>  	unsigned long buf[] = { LHREQ_IRQ, vq->config.irq };
> @@ -629,12 +637,12 @@ static void trigger_irq(struct virtqueue
>  }
> 
>  /*
> - * This looks in the virtqueue and for the first available buffer, and converts
> + * This looks in the virtqueue for the first available buffer, and converts
>   * it to an iovec for convenient access.  Since descriptors consist of some
>   * number of output then some number of input descriptors, it's actually two
>   * iovecs, but we pack them into one and note how many of each there were.
>   *
> - * This function returns the descriptor number found.
> + * This function waits if necessary, and returns the descriptor number found.
>   */
>  static unsigned wait_for_vq_desc(struct virtqueue *vq,
>  				 struct iovec iov[],
> @@ -644,10 +652,14 @@ static unsigned wait_for_vq_desc(struct 
>  	struct vring_desc *desc;
>  	u16 last_avail = lg_last_avail(vq);
> 
> +	/* There's nothing available? */
>  	while (last_avail == vq->vring.avail->idx) {
>  		u64 event;
> 
> -		/* OK, tell Guest about progress up to now. */
> +		/*
> +		 * Since we're about to sleep, now is a good time to tell the
> +		 * Guest about what we've used up to now.
> +		 */
>  		trigger_irq(vq);
> 
>  		/* OK, now we need to know about added descriptors. */
> @@ -734,8 +746,9 @@ static unsigned wait_for_vq_desc(struct 
>  }
> 
>  /*
> - * After we've used one of their buffers, we tell them about it.  We'll then
> - * want to send them an interrupt, using trigger_irq().
> + * After we've used one of their buffers, we tell the Guest about it.  Sometime
> + * later we'll want to send them an interrupt using trigger_irq(); note that
> + * wait_for_vq_desc() does that for us if it has to wait.
>   */
>  static void add_used(struct virtqueue *vq, unsigned int head, int len)
>  {
> @@ -782,12 +795,12 @@ static void console_input(struct virtque
>  	struct console_abort *abort = vq->dev->priv;
>  	struct iovec iov[vq->vring.num];
> 
> -	/* Make sure there's a descriptor waiting. */
> +	/* Make sure there's a descriptor available. */
>  	head = wait_for_vq_desc(vq, iov, &out_num, &in_num);
>  	if (out_num)
>  		errx(1, "Output buffers in console in queue?");
> 
> -	/* Read it in. */
> +	/* Read into it.  This is where we usually wait. */
>  	len = readv(STDIN_FILENO, iov, in_num);
>  	if (len <= 0) {
>  		/* Ran out of input? */
> @@ -800,6 +813,7 @@ static void console_input(struct virtque
>  			pause();
>  	}
> 
> +	/* Tell the Guest we used a buffer. */
>  	add_used_and_trigger(vq, head, len);
> 
>  	/*
> @@ -834,15 +848,23 @@ static void console_output(struct virtqu
>  	unsigned int head, out, in;
>  	struct iovec iov[vq->vring.num];
> 
> +	/* We usually wait in here, for the Guest to give us something. */
>  	head = wait_for_vq_desc(vq, iov, &out, &in);
>  	if (in)
>  		errx(1, "Input buffers in console output queue?");
> +
> +	/* writev can return a partial write, so we loop here. */
>  	while (!iov_empty(iov, out)) {
>  		int len = writev(STDOUT_FILENO, iov, out);
>  		if (len <= 0)
>  			err(1, "Write to stdout gave %i", len);
>  		iov_consume(iov, out, len);
>  	}
> +
> +	/*
> +	 * We're finished with that buffer: if we're going to sleep,
> +	 * wait_for_vq_desc() will prod the Guest with an interrupt.
> +	 */
>  	add_used(vq, head, 0);
>  }
> 
> @@ -862,15 +884,30 @@ static void net_output(struct virtqueue 
>  	unsigned int head, out, in;
>  	struct iovec iov[vq->vring.num];
> 
> +	/* We usually wait in here for the Guest to give us a packet. */
>  	head = wait_for_vq_desc(vq, iov, &out, &in);
>  	if (in)
>  		errx(1, "Input buffers in net output queue?");
> +	/*
> +	 * Send the whole thing through to /dev/net/tun.  It expects the exact
> +	 * same format: what a coincidence!
> +	 */
>  	if (writev(net_info->tunfd, iov, out) < 0)
>  		errx(1, "Write to tun failed?");
> +
> +	/*
> +	 * Done with that one; wait_for_vq_desc() will send the interrupt if
> +	 * all packets are processed.
> +	 */
>  	add_used(vq, head, 0);
>  }
> 
> -/* Will reading from this file descriptor block? */
> +/*
> + * Handling network input is a bit trickier, because I've tried to optimize it.
> + *
> + * First we have a helper routine which tells is if from this file descriptor
> + * (ie. the /dev/net/tun device) will block:
> + */
>  static bool will_block(int fd)
>  {
>  	fd_set fdset;
> @@ -880,7 +917,11 @@ static bool will_block(int fd)
>  	return select(fd+1, &fdset, NULL, NULL, &zero) != 1;
>  }
> 
> -/* This handles packets coming in from the tun device to our Guest. */
> +/*
> + * This handles packets coming in from the tun device to our Guest.  Like all
> + * service routines, it gets called again as soon as it returns, so you don't
> + * see a while(1) loop here.
> + */
>  static void net_input(struct virtqueue *vq)
>  {
>  	int len;
> @@ -888,21 +929,38 @@ static void net_input(struct virtqueue *
>  	struct iovec iov[vq->vring.num];
>  	struct net_info *net_info = vq->dev->priv;
> 
> +	/*
> +	 * Get a descriptor to write an incoming packet into.  This will also
> +	 * send an interrupt if they're out of descriptors.
> +	 */
>  	head = wait_for_vq_desc(vq, iov, &out, &in);
>  	if (out)
>  		errx(1, "Output buffers in net input queue?");
> 
> -	/* Deliver interrupt now, since we're about to sleep. */
> +	/*
> +	 * If it looks like we'll block reading from the tun device, send them
> +	 * an interrupt.
> +	 */
>  	if (vq->pending_used && will_block(net_info->tunfd))
>  		trigger_irq(vq);
> 
> +	/*
> +	 * Read in the packet.  This is where we normally wait (when there's no
> +	 * incoming network traffic).
> +	 */
>  	len = readv(net_info->tunfd, iov, in);
>  	if (len <= 0)
>  		err(1, "Failed to read from tun.");
> +
> +	/*
> +	 * Mark that packet buffer as used, but don't interrupt here.  We want
> +	 * to wait until we've done as much work as we can.
> +	 */
>  	add_used(vq, head, len);
>  }
> +/*:*/
> 
> -/* This is the helper to create threads. */
> +/* This is the helper to create threads: run the service routine in a loop. */
>  static int do_thread(void *_vq)
>  {
>  	struct virtqueue *vq = _vq;
> @@ -950,11 +1008,14 @@ static void reset_device(struct device *
>  	signal(SIGCHLD, (void *)kill_launcher);
>  }
> 
> +/*L:216
> + * This actually creates the thread which services the virtqueue for a device.
> + */
>  static void create_thread(struct virtqueue *vq)
>  {
>  	/*
> -	 * Create stack for thread and run it.  Since the stack grows upwards,
> -	 * we point the stack pointer to the end of this region.
> +	 * Create stack for thread.  Since the stack grows upwards, we point
> +	 * the stack pointer to the end of this region.
>  	 */
>  	char *stack = malloc(32768);
>  	unsigned long args[] = { LHREQ_EVENTFD,
> @@ -966,17 +1027,22 @@ static void create_thread(struct virtque
>  		err(1, "Creating eventfd");
>  	args[2] = vq->eventfd;
> 
> -	/* Attach an eventfd to this virtqueue: it will go off
> -	 * when the Guest does an LHCALL_NOTIFY for this vq. */
> +	/*
> +	 * Attach an eventfd to this virtqueue: it will go off when the Guest
> +	 * does an LHCALL_NOTIFY for this vq.
> +	 */
>  	if (write(lguest_fd, &args, sizeof(args)) != 0)
>  		err(1, "Attaching eventfd");
> 
> -	/* CLONE_VM: because it has to access the Guest memory, and
> -	 * SIGCHLD so we get a signal if it dies. */
> +	/*
> +	 * CLONE_VM: because it has to access the Guest memory, and SIGCHLD so
> +	 * we get a signal if it dies.
> +	 */
>  	vq->thread = clone(do_thread, stack + 32768, CLONE_VM | SIGCHLD, vq);
>  	if (vq->thread == (pid_t)-1)
>  		err(1, "Creating clone");
> -	/* We close our local copy, now the child has it. */
> +
> +	/* We close our local copy now the child has it. */
>  	close(vq->eventfd);
>  }
> 
> @@ -1028,7 +1094,10 @@ static void update_device_status(struct 
>  	}
>  }
> 
> -/* This is the generic routine we call when the Guest uses LHCALL_NOTIFY. */
> +/*L:215
> + * This is the generic routine we call when the Guest uses LHCALL_NOTIFY.  In
> + * particular, it's used to notify us of device status changes during boot.
> + */
>  static void handle_output(unsigned long addr)
>  {
>  	struct device *i;
> @@ -1037,18 +1106,32 @@ static void handle_output(unsigned long 
>  	for (i = devices.dev; i; i = i->next) {
>  		struct virtqueue *vq;
> 
> -		/* Notifications to device descriptors update device status. */
> +		/*
> +		 * Notifications to device descriptors mean they updated the
> +		 * device status.
> +		 */
>  		if (from_guest_phys(addr) == i->desc) {
>  			update_device_status(i);
>  			return;
>  		}
> 
> -		/* Devices *can* be used before status is set to DRIVER_OK. */
> +		/*
> +		 * Devices *can* be used before status is set to DRIVER_OK.
> +		 * The original plan was that they would never do this: they
> +		 * would always finish setting up their status bits before
> +		 * actually touching the virtqueues.  In practice, we allowed
> +		 * them to, and they do (eg. the disk probes for partition
> +		 * tables as part of initialization).
> +		 *
> +		 * If we see this, we start the device: once it's running, we
> +		 * expect the device to catch all the notifications.
> +		 */
>  		for (vq = i->vq; vq; vq = vq->next) {
>  			if (addr != vq->config.pfn*getpagesize())
>  				continue;
>  			if (i->running)
>  				errx(1, "Notification on running %s", i->name);
> +			/* This just calls create_thread() for each virtqueue */
>  			start_device(i);
>  			return;
>  		}
> @@ -1132,6 +1215,11 @@ static void add_virtqueue(struct device 
>  	vq->next = NULL;
>  	vq->last_avail_idx = 0;
>  	vq->dev = dev;
> +
> +	/*
> +	 * This is the routine the service thread will run, and its Process ID
> +	 * once it's running.
> +	 */
>  	vq->service = service;
>  	vq->thread = (pid_t)-1;
> 
> @@ -1202,7 +1290,8 @@ static void set_config(struct device *de
> 
>  /*
>   * This routine does all the creation and setup of a new device, including
> - * calling new_dev_desc() to allocate the descriptor and device memory.
> + * calling new_dev_desc() to allocate the descriptor and device memory.  We
> + * don't actually start the service threads until later.
>   *
>   * See what I mean about userspace being boring?
>   */
> @@ -1478,19 +1567,7 @@ static void setup_tun_net(char *arg)
>  		verbose("device %u: tun %s: %s\n",
>  			devices.device_num, tapif, arg);
>  }
> -
> -/*
> - * Our block (disk) device should be really simple: the Guest asks for a block
> - * number and we read or write that position in the file.  Unfortunately, that
> - * was amazingly slow: the Guest waits until the read is finished before
> - * running anything else, even if it could have been doing useful work.
> - *
> - * We could use async I/O, except it's reputed to suck so hard that characters
> - * actually go missing from your code when you try to use it.
> - *
> - * So this was one reason why lguest now does all virtqueue servicing in
> - * separate threads: it's more efficient and more like a real device.
> - */
> +/*:*/
> 
>  /* This hangs off device->priv. */
>  struct vblk_info
> @@ -1512,8 +1589,16 @@ struct vblk_info
>  /*L:210
>   * The Disk
>   *
> - * Remember that the block device is handled by a separate I/O thread.  We head
> - * straight into the core of that thread here:
> + * The disk only has one virtqueue, so it only has one thread.  It is really
> + * simple: the Guest asks for a block number and we read or write that position
> + * in the file.
> + *
> + * Before we serviced each virtqueue in a separate thread, that was unacceptably
> + * slow: the Guest waits until the read is finished before running anything
> + * else, even if it could have been doing useful work.
> + *
> + * We could have used async I/O, except it's reputed to suck so hard that
> + * characters actually go missing from your code when you try to use it.
>   */
>  static void blk_request(struct virtqueue *vq)
>  {
> @@ -1525,7 +1610,10 @@ static void blk_request(struct virtqueue
>  	struct iovec iov[vq->vring.num];
>  	off64_t off;
> 
> -	/* Get the next request. */
> +	/*
> +	 * Get the next request, where we normally wait.  It triggers the
> +	 * interrupt to acknowledge previously serviced requests (if any).
> +	 */
>  	head = wait_for_vq_desc(vq, iov, &out_num, &in_num);
> 
>  	/*
> @@ -1539,6 +1627,10 @@ static void blk_request(struct virtqueue
> 
>  	out = convert(&iov[0], struct virtio_blk_outhdr);
>  	in = convert(&iov[out_num+in_num-1], u8);
> +	/*
> +	 * For historical reasons, block operations are expressed in 512 byte
> +	 * "sectors".
> +	 */
>  	off = out->sector * 512;
> 
>  	/*
> @@ -1614,6 +1706,7 @@ static void blk_request(struct virtqueue
>  	if (out->type & VIRTIO_BLK_T_BARRIER)
>  		fdatasync(vblk->fd);
> 
> +	/* Finished that request. */
>  	add_used(vq, head, wlen);
>  }
> 
> @@ -1682,9 +1775,8 @@ static void rng_input(struct virtqueue *
>  		errx(1, "Output buffers in rng?");
> 
>  	/*
> -	 * This is why we convert to iovecs: the readv() call uses them, and so
> -	 * it reads straight into the Guest's buffer.  We loop to make sure we
> -	 * fill it.
> +	 * Just like the console write, we loop to cover the whole iovec.
> +	 * In this case, short reads actually happen quite a bit.
>  	 */
>  	while (!iov_empty(iov, in_num)) {
>  		len = readv(rng_info->rfd, iov, in_num);
> @@ -1818,7 +1910,9 @@ int main(int argc, char *argv[])
>  	devices.lastdev = NULL;
>  	devices.next_irq = 1;
> 
> +	/* We're CPU 0.  In fact, that's the only CPU possible right now. */
>  	cpu_id = 0;
> +
>  	/*
>  	 * 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
> @@ -1926,7 +2020,7 @@ int main(int argc, char *argv[])
>  	 */
>  	tell_kernel(start);
> 
> -	/* Ensure that we terminate if a child dies. */
> +	/* Ensure that we terminate if a device-servicing child dies. */
>  	signal(SIGCHLD, kill_launcher);
> 
>  	/* If we exit via err(), this kills all the threads, restores tty. */
> diff --git a/arch/x86/include/asm/lguest_hcall.h b/arch/x86/include/asm/lguest_hcall.h
> --- a/arch/x86/include/asm/lguest_hcall.h
> +++ b/arch/x86/include/asm/lguest_hcall.h
> @@ -35,10 +35,10 @@
>   * operations?  There are two ways: the direct way is to make a "hypercall",
>   * to make requests of the Host Itself.
>   *
> - * We use the KVM hypercall mechanism. Seventeen hypercalls are
> - * available: the hypercall number is put in the %eax register, and the
> - * arguments (when required) are placed in %ebx, %ecx, %edx and %esi.
> - * If a return value makes sense, it's returned in %eax.
> + * We use the KVM hypercall mechanism, though completely different hypercall
> + * numbers. Seventeen hypercalls are available: the hypercall number is put in
> + * the %eax register, and the arguments (when required) are placed in %ebx,
> + * %ecx, %edx and %esi.  If a return value makes sense, it's returned in %eax.
>   *
>   * Grossly invalid calls result in Sudden Death at the hands of the vengeful
>   * Host, rather than returning failure.  This reflects Winston Churchill's
> diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c
> --- a/arch/x86/lguest/boot.c
> +++ b/arch/x86/lguest/boot.c
> @@ -154,6 +154,7 @@ static void lazy_hcall1(unsigned long ca
>  		async_hcall(call, arg1, 0, 0, 0);
>  }
> 
> +/* You can imagine what lazy_hcall2, 3 and 4 look like. :*/
>  static void lazy_hcall2(unsigned long call,
>  		       unsigned long arg1,
>  		       unsigned long arg2)
> @@ -189,8 +190,10 @@ static void lazy_hcall4(unsigned long ca
>  }
>  #endif
> 
> -/* When lazy mode is turned off reset the per-cpu lazy mode variable and then
> - * issue the do-nothing hypercall to flush any stored calls. */
> +/*G:036
> + * When lazy mode is turned off reset the per-cpu lazy mode variable and then
> + * issue the do-nothing hypercall to flush any stored calls.
> +:*/
>  static void lguest_leave_lazy_mmu_mode(void)
>  {
>  	kvm_hypercall0(LHCALL_FLUSH_ASYNC);
> @@ -250,13 +253,11 @@ extern void lg_irq_enable(void);
>  extern void lg_restore_fl(unsigned long flags);
> 
>  /*M:003
> - * Note that we don't check for outstanding interrupts when we re-enable them
> - * (or when we unmask an interrupt).  This seems to work for the moment, since
> - * interrupts are rare and we'll just get the interrupt on the next timer tick,
> - * but now we can run with CONFIG_NO_HZ, we should revisit this.  One way would
> - * be to put the "irq_enabled" field in a page by itself, and have the Host
> - * write-protect it when an interrupt comes in when irqs are disabled.  There
> - * will then be a page fault as soon as interrupts are re-enabled.
> + * We could be more efficient in our checking of outstanding interrupts, rather
> + * than using a branch.  One way would be to put the "irq_enabled" field in a
> + * page by itself, and have the Host write-protect it when an interrupt comes
> + * in when irqs are disabled.  There will then be a page fault as soon as
> + * interrupts are re-enabled.
>   *
>   * A better method is to implement soft interrupt disable generally for x86:
>   * instead of disabling interrupts, we set a flag.  If an interrupt does come
> @@ -568,7 +569,7 @@ static void lguest_write_cr4(unsigned lo
>   * cr3 ---> +---------+
>   *	    |  	   --------->+---------+
>   *	    |	      |	     | PADDR1  |
> - *	  Top-level   |	     | PADDR2  |
> + *	  Mid-level   |	     | PADDR2  |
>   *	  (PMD) page  |	     | 	       |
>   *	    |	      |	   Lower-level |
>   *	    |	      |	   (PTE) page  |
> @@ -588,23 +589,62 @@ static void lguest_write_cr4(unsigned lo
>   *    Index into top     Index into second      Offset within page
>   *  page directory page    pagetable page
>   *
> - * The kernel spends a lot of time changing both the top-level page directory
> - * and lower-level pagetable pages.  The Guest doesn't know physical addresses,
> - * so while it maintains these page tables exactly like normal, it also needs
> - * to keep the Host informed whenever it makes a change: the Host will create
> - * the real page tables based on the Guests'.
> + * Now, unfortunately, this isn't the whole story: Intel added Physical Address
> + * Extension (PAE) to allow 32 bit systems to use 64GB of memory (ie. 36 bits).
> + * These are held in 64-bit page table entries, so we can now only fit 512
> + * entries in a page, and the neat three-level tree breaks down.
> + *
> + * The result is a four level page table:
> + *
> + * cr3 --> [ 4 Upper  ]
> + *	   [   Level  ]
> + *	   [  Entries ]
> + *	   [(PUD Page)]---> +---------+
> + *	 		    |  	   --------->+---------+
> + *	 		    |	      |	     | PADDR1  |
> + *	 		  Mid-level   |	     | PADDR2  |
> + *	 		  (PMD) page  |	     | 	       |
> + *	 		    |	      |	   Lower-level |
> + *	 		    |	      |	   (PTE) page  |
> + *	 		    |	      |	     |	       |
> + *	 		      ....    	     	 ....
> + *
> + *
> + * And the virtual address is decoded as:
> + *
> + *         1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
> + *      |<-2->|<--- 9 bits ---->|<---- 9 bits --->|<------ 12 bits ------>|
> + * Index into    Index into mid    Index into lower    Offset within page
> + * top entries   directory page     pagetable page
> + *
> + * It's too hard to switch between these two formats at runtime, so Linux only
> + * supports one or the other depending on whether CONFIG_X86_PAE is set.  Many
> + * distributions turn it on, and not just for people with silly amounts of
> + * memory: the larger PTE entries allow room for the NX bit, which lets the
> + * kernel disable execution of pages and increase security.
> + *
> + * This was a problem for lguest, which couldn't run on these distributions;
> + * then Matias Zabaljauregui figured it all out and implemented it, and only a
> + * handful of puppies were crushed in the process!
> + *
> + * Back to our point: the kernel spends a lot of time changing both the
> + * top-level page directory and lower-level pagetable pages.  The Guest doesn't
> + * know physical addresses, so while it maintains these page tables exactly
> + * like normal, it also needs to keep the Host informed whenever it makes a
> + * change: the Host will create the real page tables based on the Guests'.
>   */
> 
>  /*
> - * The Guest calls this to set a second-level entry (pte), ie. to map a page
> - * into a process' address space.  We set the entry then tell the Host the
> - * toplevel and address this corresponds to.  The Guest uses one pagetable per
> - * process, so we need to tell the Host which one we're changing (mm->pgd).
> + * The Guest calls this after it has set a second-level entry (pte), ie. to map
> + * a page into a process' address space.  Wetell the Host the toplevel and
> + * address this corresponds to.  The Guest uses one pagetable per process, so
> + * we need to tell the Host which one we're changing (mm->pgd).
>   */
>  static void lguest_pte_update(struct mm_struct *mm, unsigned long addr,
>  			       pte_t *ptep)
>  {
>  #ifdef CONFIG_X86_PAE
> +	/* PAE needs to hand a 64 bit page table entry, so it uses two args. */
>  	lazy_hcall4(LHCALL_SET_PTE, __pa(mm->pgd), addr,
>  		    ptep->pte_low, ptep->pte_high);
>  #else
> @@ -612,6 +652,7 @@ static void lguest_pte_update(struct mm_
>  #endif
>  }
> 
> +/* This is the "set and update" combo-meal-deal version. */
>  static void lguest_set_pte_at(struct mm_struct *mm, unsigned long addr,
>  			      pte_t *ptep, pte_t pteval)
>  {
> @@ -672,6 +713,11 @@ static void lguest_set_pte(pte_t *ptep, 
>  }
> 
>  #ifdef CONFIG_X86_PAE
> +/*
> + * With 64-bit PTE values, we need to be careful setting them: if we set 32
> + * bits at a time, the hardware could see a weird half-set entry.  These
> + * versions ensure we update all 64 bits at once.
> + */
>  static void lguest_set_pte_atomic(pte_t *ptep, pte_t pte)
>  {
>  	native_set_pte_atomic(ptep, pte);
> @@ -679,13 +725,14 @@ static void lguest_set_pte_atomic(pte_t 
>  		lazy_hcall1(LHCALL_FLUSH_TLB, 1);
>  }
> 
> -void lguest_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
> +static void lguest_pte_clear(struct mm_struct *mm, unsigned long addr,
> +			     pte_t *ptep)
>  {
>  	native_pte_clear(mm, addr, ptep);
>  	lguest_pte_update(mm, addr, ptep);
>  }
> 
> -void lguest_pmd_clear(pmd_t *pmdp)
> +static void lguest_pmd_clear(pmd_t *pmdp)
>  {
>  	lguest_set_pmd(pmdp, __pmd(0));
>  }
> @@ -784,6 +831,14 @@ static void __init lguest_init_IRQ(void)
>  	irq_ctx_init(smp_processor_id());
>  }
> 
> +/*
> + * With CONFIG_SPARSE_IRQ, interrupt descriptors are allocated as-needed, so
> + * rather than set them in lguest_init_IRQ we are called here every time an
> + * lguest device needs an interrupt.
> + *
> + * FIXME: irq_to_desc_alloc_node() can fail due to lack of memory, we should
> + * pass that up!
> + */
>  void lguest_setup_irq(unsigned int irq)
>  {
>  	irq_to_desc_alloc_node(irq, 0);
> @@ -1298,7 +1353,7 @@ __init void lguest_init(void)
>  	 */
>  	switch_to_new_gdt(0);
> 
> -	/* As described in head_32.S, we map the first 128M of memory. */
> +	/* We actually boot with all memory mapped, but let's say 128MB. */
>  	max_pfn_mapped = (128*1024*1024) >> PAGE_SHIFT;
> 
>  	/*
> diff --git a/arch/x86/lguest/i386_head.S b/arch/x86/lguest/i386_head.S
> --- a/arch/x86/lguest/i386_head.S
> +++ b/arch/x86/lguest/i386_head.S
> @@ -102,6 +102,7 @@ send_interrupts:
>  	 * create one manually here.
>  	 */
>  	.byte 0x0f,0x01,0xc1 /* KVM_HYPERCALL */
> +	/* Put eax back the way we found it. */
>  	popl %eax
>  	ret
> 
> @@ -125,6 +126,7 @@ ENTRY(lg_restore_fl)
>  	jnz send_interrupts
>  	/* Again, the normal path has used no extra registers.  Clever, huh? */
>  	ret
> +/*:*/
> 
>  /* These demark the EIP range where host should never deliver interrupts. */
>  .global lguest_noirq_start
> diff --git a/drivers/lguest/core.c b/drivers/lguest/core.c
> --- a/drivers/lguest/core.c
> +++ b/drivers/lguest/core.c
> @@ -217,10 +217,15 @@ int run_guest(struct lg_cpu *cpu, unsign
> 
>  		/*
>  		 * It's possible the Guest did a NOTIFY hypercall to the
> -		 * Launcher, in which case we return from the read() now.
> +		 * Launcher.
>  		 */
>  		if (cpu->pending_notify) {
> +			/*
> +			 * Does it just needs to write to a registered
> +			 * eventfd (ie. the appropriate virtqueue thread)?
> +			 */
>  			if (!send_notify_to_eventfd(cpu)) {
> +				/* OK, we tell the main Laucher. */
>  				if (put_user(cpu->pending_notify, user))
>  					return -EFAULT;
>  				return sizeof(cpu->pending_notify);
> diff --git a/drivers/lguest/hypercalls.c b/drivers/lguest/hypercalls.c
> --- a/drivers/lguest/hypercalls.c
> +++ b/drivers/lguest/hypercalls.c
> @@ -59,7 +59,7 @@ static void do_hcall(struct lg_cpu *cpu,
>  	case LHCALL_SHUTDOWN: {
>  		char msg[128];
>  		/*
> -		 * Shutdown is such a trivial hypercall that we do it in four
> +		 * Shutdown is such a trivial hypercall that we do it in five
>  		 * lines right here.
>  		 *
>  		 * If the lgread fails, it will call kill_guest() itself; the
> @@ -245,6 +245,10 @@ static void initialize(struct lg_cpu *cp
>   * device), the Guest will still see the old page.  In practice, this never
>   * happens: why would the Guest read a page which it has never written to?  But
>   * a similar scenario might one day bite us, so it's worth mentioning.
> + *
> + * Note that if we used a shared anonymous mapping in the Launcher instead of
> + * mapping /dev/zero private, we wouldn't worry about cop-on-write.  And we
> + * need that to switch the Launcher to processes (away from threads) anyway.
>  :*/
> 
>  /*H:100
> diff --git a/drivers/lguest/lguest_device.c b/drivers/lguest/lguest_device.c
> --- a/drivers/lguest/lguest_device.c
> +++ b/drivers/lguest/lguest_device.c
> @@ -236,7 +236,7 @@ static void lg_notify(struct virtqueue *
>  extern void lguest_setup_irq(unsigned int irq);
> 
>  /*
> - * This routine finds the first virtqueue described in the configuration of
> + * This routine finds the Nth virtqueue described in the configuration of
>   * this device and sets it up.
>   *
>   * This is kind of an ugly duckling.  It'd be nicer to have a standard
> @@ -244,9 +244,6 @@ extern void lguest_setup_irq(unsigned in
>   * everyone wants to do it differently.  The KVM coders want the Guest to
>   * allocate its own pages and tell the Host where they are, but for lguest it's
>   * simpler for the Host to simply tell us where the pages are.
> - *
> - * So we provide drivers with a "find the Nth virtqueue and set it up"
> - * function.
>   */
>  static struct virtqueue *lg_find_vq(struct virtio_device *vdev,
>  				    unsigned index,
> @@ -422,7 +419,11 @@ static void add_lguest_device(struct lgu
> 
>  	/* This devices' parent is the lguest/ dir. */
>  	ldev->vdev.dev.parent = lguest_root;
> -	/* We have a unique device index thanks to the dev_index counter. */
> +	/*
> +	 * The device type comes straight from the descriptor.  There's also a
> +	 * device vendor field in the virtio_device struct, which we leave as
> +	 * 0.
> +	 */
>  	ldev->vdev.id.device = d->type;
>  	/*
>  	 * We have a simple set of routines for querying the device's
> diff --git a/drivers/lguest/lguest_user.c b/drivers/lguest/lguest_user.c
> --- a/drivers/lguest/lguest_user.c
> +++ b/drivers/lguest/lguest_user.c
> @@ -32,10 +32,36 @@ bool send_notify_to_eventfd(struct lg_cp
>  	return cpu->pending_notify == 0;
>  }
> 
> +/*L:055
> + * One of the more tricksy tricks in the Linux Kernel is a technique called
> + * Read Copy Update.  Since one point of lguest is to teach lguest journeyers
> + * about kernel coding, I use it here.  (In case you're curious, other purposes
> + * include learning about virtualization and instilling a deep appreciation for
> + * simplicity and puppies).
> + *
> + * We keep a simple array which maps LHCALL_NOTIFY values to eventfds, but we
> + * we add new eventfds without ever blocking readers from accessing the array.

Suggest s/we we/we/  ;-)

> + * The current Launcher only does this during boot, so that never happens.  But
> + * Read Copy Update is cool, and adding a lock risks damaging even more puppies
> + * than this code does.
> + *
> + * We allocate a brand new one-larger array, copy the old one and add our new
> + * element.  Then we make the lg eventfd pointer point to the new array.

Could you please also mention the role of rcu_assign_pointer()?

> + * That's the easy part: now we need to free the old one, but we need to make
> + * sure no slow CPU somewhere is still looking at it.  That's what
> + * synchronize_rcu does for us: waits until every CPU has indicated that it has
> + * moved on to know it's no longer using the old one.

This is protected on by lguest_lock, correct?

> + *
> + * If that's unclear, see http://en.wikipedia.org/wiki/Read-copy-update.
> + */
>  static int add_eventfd(struct lguest *lg, unsigned long addr, int fd)
>  {
>  	struct lg_eventfd_map *new, *old = lg->eventfds;
> 
> +	/*
> +	 * We don't allow notifications on address 0 anyway (pending_notify of
> +	 * 0 means "nothing pending).

s/"nothing pending/"nothing pending"/

> +	 */
>  	if (!addr)
>  		return -EINVAL;
> 
> @@ -75,6 +101,14 @@ static int add_eventfd(struct lguest *lg
>  	return 0;
>  }
> 
> +/*L:052
> + * Receiving notifications from the Guest is usually done by attaching a
> + * particular LHCALL_NOTIFY value to an event filedescriptor.  The eventfd will
> + * become readable when the Guest does an LHCALL_NOTIFY with that value.
> + *
> + * This is really convenient for processing each virtqueue in a separate
> + * thread.
> + */
>  static int attach_eventfd(struct lguest *lg, const unsigned long __user *input)
>  {
>  	unsigned long addr, fd;
> @@ -86,6 +120,11 @@ static int attach_eventfd(struct lguest 
>  	if (get_user(fd, input) != 0)
>  		return -EFAULT;
> 
> +	/*
> +	 * Just make sure two callers don't add eventfds at once.  We really
> +	 * only need to lock against callers adding to the same Guest, so using
> +	 * the Big Lguest Lock is overkill.  But this is setup, not a fast path.
> +	 */
>  	mutex_lock(&lguest_lock);
>  	err = add_eventfd(lg, addr, fd);
>  	mutex_unlock(&lguest_lock);
> @@ -106,6 +145,10 @@ static int user_send_irq(struct lg_cpu *
>  	if (irq >= LGUEST_IRQS)
>  		return -EINVAL;
> 
> +	/*
> +	 * Next time the Guest runs, the core code will see if it can deliver
> +	 * this interrupt.
> +	 */
>  	set_interrupt(cpu, irq);
>  	return 0;
>  }
> @@ -307,10 +350,10 @@ unlock:
>   * The first operation the Launcher does must be a write.  All writes
>   * start with an unsigned long number: for the first write this must be
>   * LHREQ_INITIALIZE to set up the Guest.  After that the Launcher can use
> - * writes of other values to send interrupts.
> + * writes of other values to send interrupts or set up receipt of notifications.
>   *
>   * Note that we overload the "offset" in the /dev/lguest file to indicate what
> - * CPU number we're dealing with.  Currently this is always 0, since we only
> + * CPU number we're dealing with.  Currently this is always 0 since we only
>   * support uniprocessor Guests, but you can see the beginnings of SMP support
>   * here.
>   */
> diff --git a/drivers/lguest/page_tables.c b/drivers/lguest/page_tables.c
> --- a/drivers/lguest/page_tables.c
> +++ b/drivers/lguest/page_tables.c
> @@ -29,10 +29,10 @@
>  /*H:300
>   * The Page Table Code
>   *
> - * We use two-level page tables for the Guest.  If you're not entirely
> - * comfortable with virtual addresses, physical addresses and page tables then
> - * I recommend you review arch/x86/lguest/boot.c's "Page Table Handling" (with
> - * diagrams!).
> + * We use two-level page tables for the Guest, or three-level with PAE.  If
> + * you're not entirely comfortable with virtual addresses, physical addresses
> + * and page tables then I recommend you review arch/x86/lguest/boot.c's "Page
> + * Table Handling" (with diagrams!).
>   *
>   * The Guest keeps page tables, but we maintain the actual ones here: these are
>   * called "shadow" page tables.  Which is a very Guest-centric name: these are
> @@ -52,9 +52,8 @@
>  :*/
> 
>  /*
> - * 1024 entries in a page table page maps 1024 pages: 4MB.  The Switcher is
> - * conveniently placed at the top 4MB, so it uses a separate, complete PTE
> - * page.
> + * The Switcher uses the complete top PTE page.  That's 1024 PTE entries (4MB)
> + * or 512 PTE entries with PAE (2MB).
>   */
>  #define SWITCHER_PGD_INDEX (PTRS_PER_PGD - 1)
> 
> @@ -81,7 +80,8 @@ static DEFINE_PER_CPU(pte_t *, switcher_
> 
>  /*H:320
>   * The page table code is curly enough to need helper functions to keep it
> - * clear and clean.
> + * clear and clean.  The kernel itself provides many of them; one advantage
> + * of insisting that the Guest and Host use the same CONFIG_PAE setting.
>   *
>   * There are two functions which return pointers to the shadow (aka "real")
>   * page tables.
> @@ -155,7 +155,7 @@ static pte_t *spte_addr(struct lg_cpu *c
>  }
> 
>  /*
> - * These two functions just like the above two, except they access the Guest
> + * These functions are just like the above two, except they access the Guest
>   * page tables.  Hence they return a Guest address.
>   */
>  static unsigned long gpgd_addr(struct lg_cpu *cpu, unsigned long vaddr)
> @@ -165,6 +165,7 @@ static unsigned long gpgd_addr(struct lg
>  }
> 
>  #ifdef CONFIG_X86_PAE
> +/* Follow the PGD to the PMD. */
>  static unsigned long gpmd_addr(pgd_t gpgd, unsigned long vaddr)
>  {
>  	unsigned long gpage = pgd_pfn(gpgd) << PAGE_SHIFT;
> @@ -172,6 +173,7 @@ static unsigned long gpmd_addr(pgd_t gpg
>  	return gpage + pmd_index(vaddr) * sizeof(pmd_t);
>  }
> 
> +/* Follow the PMD to the PTE. */
>  static unsigned long gpte_addr(struct lg_cpu *cpu,
>  			       pmd_t gpmd, unsigned long vaddr)
>  {
> @@ -181,6 +183,7 @@ static unsigned long gpte_addr(struct lg
>  	return gpage + pte_index(vaddr) * sizeof(pte_t);
>  }
>  #else
> +/* Follow the PGD to the PTE (no mid-level for !PAE). */
>  static unsigned long gpte_addr(struct lg_cpu *cpu,
>  				pgd_t gpgd, unsigned long vaddr)
>  {
> @@ -314,6 +317,7 @@ bool demand_page(struct lg_cpu *cpu, uns
>  	pte_t gpte;
>  	pte_t *spte;
> 
> +	/* Mid level for PAE. */
>  #ifdef CONFIG_X86_PAE
>  	pmd_t *spmd;
>  	pmd_t gpmd;
> @@ -391,6 +395,8 @@ bool demand_page(struct lg_cpu *cpu, uns
>  	 */
>  	gpte_ptr = gpte_addr(cpu, gpgd, vaddr);
>  #endif
> +
> +	/* Read the actual PTE value. */
>  	gpte = lgread(cpu, gpte_ptr, pte_t);
> 
>  	/* If this page isn't in the Guest page tables, we can't page it in. */
> @@ -507,6 +513,7 @@ void pin_page(struct lg_cpu *cpu, unsign
>  	if (!page_writable(cpu, vaddr) && !demand_page(cpu, vaddr, 2))
>  		kill_guest(cpu, "bad stack page %#lx", vaddr);
>  }
> +/*:*/
> 
>  #ifdef CONFIG_X86_PAE
>  static void release_pmd(pmd_t *spmd)
> @@ -543,7 +550,11 @@ static void release_pgd(pgd_t *spgd)
>  }
> 
>  #else /* !CONFIG_X86_PAE */
> -/*H:450 If we chase down the release_pgd() code, it looks like this: */
> +/*H:450
> + * If we chase down the release_pgd() code, the non-PAE version looks like
> + * this.  The PAE version is almost identical, but instead of calling
> + * release_pte it calls release_pmd(), which looks much like this.
> + */
>  static void release_pgd(pgd_t *spgd)
>  {
>  	/* If the entry's not present, there's nothing to release. */
> @@ -898,17 +909,21 @@ void guest_set_pgd(struct lguest *lg, un
>  		/* ... throw it away. */
>  		release_pgd(lg->pgdirs[pgdir].pgdir + idx);
>  }
> +
>  #ifdef CONFIG_X86_PAE
> +/* For setting a mid-level, we just throw everything away.  It's easy. */
>  void guest_set_pmd(struct lguest *lg, unsigned long pmdp, u32 idx)
>  {
>  	guest_pagetable_clear_all(&lg->cpus[0]);
>  }
>  #endif
> 
> -/*
> - * Once we know how much memory we have we can construct simple identity (which
> +/*H:505
> + * To get through boot, we construct simple identity page mappings (which
>   * set virtual == physical) and linear mappings which will get the Guest far
> - * enough into the boot to create its own.
> + * enough into the boot to create its own.  The linear mapping means we
> + * simplify the Guest boot, but it makes assumptions about their PAGE_OFFSET,
> + * as you'll see.
>   *
>   * We lay them out of the way, just below the initrd (which is why we need to
>   * know its size here).
> @@ -944,6 +959,10 @@ static unsigned long setup_pagetables(st
>  	linear = (void *)pgdir - linear_pages * PAGE_SIZE;
> 
>  #ifdef CONFIG_X86_PAE
> +	/*
> +	 * And the single mid page goes below that.  We only use one, but
> +	 * that's enough to map 1G, which definitely gets us through boot.
> +	 */
>  	pmds = (void *)linear - PAGE_SIZE;
>  #endif
>  	/*
> @@ -957,13 +976,14 @@ static unsigned long setup_pagetables(st
>  			return -EFAULT;
>  	}
> 
> +#ifdef CONFIG_X86_PAE
>  	/*
> -	 * The top level points to the linear page table pages above.
> -	 * We setup the identity and linear mappings here.
> +	 * Make the Guest PMD entries point to the corresponding place in the
> +	 * linear mapping (up to one page worth of PMD).
>  	 */
> -#ifdef CONFIG_X86_PAE
>  	for (i = j = 0; i < mapped_pages && j < PTRS_PER_PMD;
>  	     i += PTRS_PER_PTE, j++) {
> +		/* FIXME: native_set_pmd is overkill here. */
>  		native_set_pmd(&pmd, __pmd(((unsigned long)(linear + i)
>  		- mem_base) | _PAGE_PRESENT | _PAGE_RW | _PAGE_USER));
> 
> @@ -971,18 +991,36 @@ static unsigned long setup_pagetables(st
>  			return -EFAULT;
>  	}
> 
> +	/* One PGD entry, pointing to that PMD page. */
>  	set_pgd(&pgd, __pgd(((u32)pmds - mem_base) | _PAGE_PRESENT));
> +	/* Copy it in as the first PGD entry (ie. addresses 0-1G). */
>  	if (copy_to_user(&pgdir[0], &pgd, sizeof(pgd)) != 0)
>  		return -EFAULT;
> +	/*
> +	 * And the third PGD entry (ie. addresses 3G-4G).
> +	 *
> +	 * FIXME: This assumes that PAGE_OFFSET for the Guest is 0xC0000000.
> +	 */
>  	if (copy_to_user(&pgdir[3], &pgd, sizeof(pgd)) != 0)
>  		return -EFAULT;
>  #else
> +	/*
> +	 * The top level points to the linear page table pages above.
> +	 * We setup the identity and linear mappings here.
> +	 */
>  	phys_linear = (unsigned long)linear - mem_base;
>  	for (i = 0; i < mapped_pages; i += PTRS_PER_PTE) {
>  		pgd_t pgd;
> +		/*
> +		 * Create a PGD entry which points to the right part of the
> +		 * linear PTE pages.
> +		 */
>  		pgd = __pgd((phys_linear + i * sizeof(pte_t)) |
>  			    (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER));
> 
> +		/*
> +		 * Copy it into the PGD page at 0 and PAGE_OFFSET.
> +		 */
>  		if (copy_to_user(&pgdir[i / PTRS_PER_PTE], &pgd, sizeof(pgd))
>  		    || copy_to_user(&pgdir[pgd_index(PAGE_OFFSET)
>  					   + i / PTRS_PER_PTE],
> @@ -992,8 +1030,8 @@ static unsigned long setup_pagetables(st
>  #endif
> 
>  	/*
> -	 * We return the top level (guest-physical) address: remember where
> -	 * this is.
> +	 * We return the top level (guest-physical) address: we remember where
> +	 * this is to write it into lguest_data when the Guest initializes.
>  	 */
>  	return (unsigned long)pgdir - mem_base;
>  }
> @@ -1031,7 +1069,9 @@ int init_guest_pagetable(struct lguest *
>  	lg->pgdirs[0].pgdir = (pgd_t *)get_zeroed_page(GFP_KERNEL);
>  	if (!lg->pgdirs[0].pgdir)
>  		return -ENOMEM;
> +
>  #ifdef CONFIG_X86_PAE
> +	/* For PAE, we also create the initial mid-level. */
>  	pgd = lg->pgdirs[0].pgdir;
>  	pmd_table = (pmd_t *) get_zeroed_page(GFP_KERNEL);
>  	if (!pmd_table)
> @@ -1040,11 +1080,13 @@ int init_guest_pagetable(struct lguest *
>  	set_pgd(pgd + SWITCHER_PGD_INDEX,
>  		__pgd(__pa(pmd_table) | _PAGE_PRESENT));
>  #endif
> +
> +	/* This is the current page table. */
>  	lg->cpus[0].cpu_pgd = 0;
>  	return 0;
>  }
> 
> -/* When the Guest calls LHCALL_LGUEST_INIT we do more setup. */
> +/*H:508 When the Guest calls LHCALL_LGUEST_INIT we do more setup. */
>  void page_table_guest_data_init(struct lg_cpu *cpu)
>  {
>  	/* We get the kernel address: above this is all kernel memory. */
> @@ -1105,12 +1147,16 @@ void map_switcher_in_guest(struct lg_cpu
>  	pmd_t switcher_pmd;
>  	pmd_t *pmd_table;
> 
> +	/* FIXME: native_set_pmd is overkill here. */
>  	native_set_pmd(&switcher_pmd, pfn_pmd(__pa(switcher_pte_page) >>
>  		       PAGE_SHIFT, PAGE_KERNEL_EXEC));
> 
> +	/* Figure out where the pmd page is, by reading the PGD, and converting
> +	 * it to a virtual address. */
>  	pmd_table = __va(pgd_pfn(cpu->lg->
>  			pgdirs[cpu->cpu_pgd].pgdir[SWITCHER_PGD_INDEX])
>  								<< PAGE_SHIFT);
> +	/* Now write it into the shadow page table. */
>  	native_set_pmd(&pmd_table[SWITCHER_PMD_INDEX], switcher_pmd);
>  #else
>  	pgd_t switcher_pgd;
> diff --git a/drivers/lguest/x86/core.c b/drivers/lguest/x86/core.c
> --- a/drivers/lguest/x86/core.c
> +++ b/drivers/lguest/x86/core.c
> @@ -187,7 +187,7 @@ static void run_guest_once(struct lg_cpu
>   * also simplify copy_in_guest_info().  Note that we'd still need to restore
>   * things when we exit to Launcher userspace, but that's fairly easy.
>   *
> - * We could also try using this hooks for PGE, but that might be too expensive.
> + * We could also try using these hooks for PGE, but that might be too expensive.
>   *
>   * The hooks were designed for KVM, but we can also put them to good use.
>  :*/
> diff --git a/drivers/lguest/x86/switcher_32.S b/drivers/lguest/x86/switcher_32.S
> --- a/drivers/lguest/x86/switcher_32.S
> +++ b/drivers/lguest/x86/switcher_32.S
> @@ -1,7 +1,7 @@
>  /*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.
> + * This is the Switcher: code which sits at 0xFFC00000 (or 0xFFE00000) 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
> 


More information about the Lguest mailing list