Hi -
On 2016-09-06 at 14:10 Ronald G. Minnich wrote:
> + if (! p512) {
> + werrstr("page table allocation failed: %r\n");
Just curious - did this work without clobbering the original errstr?
If it did, then I guess printf internally was buffering the input
before clobbering errstr, which is nice.
> + ret = posix_memalign((void **)&stack, 4096, DEFAULT_STACK_SIZE);
> + if (ret)
> + return ret;
> + /* touch the top word on the stack so we don't page fault
> + * on that in the VM. */
> + stack[DEFAULT_STACK_SIZE/sizeof(uint64_t) - 1] = 0;
> + vm_tf->tf_rsp = (uint64_t)stack;
I think that sets rsp to the bottom of the stack, instead of the top.
Maybe something like this:
vm_tf->tf_rsp = &stack[DEFAULT_STACK_SIZE/sizeof(uint64_t) - 1];
/* touch the top word on the stack so we don't page fault
* on that in the VM. */
*(uintptr_t*)vm_tf->tf_rsp = 0;
(barring compiler warnings, etc).
Also, checkpatch had a few hits. I fixed those and attached the
edited patch.
Barret
--
You received this message because you are subscribed to the Google Groups
"Akaros" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
For more options, visit https://groups.google.com/d/optout.
From: "Ronald G. Minnich" <[email protected]>
Author: "Ronald G. Minnich"
Email: [email protected]
Subject: VMM: add two vthread functions.
Date: Tue, 6 Sep 2016 14:10:12 -0700
The intent of these functions it to give users a familiar
API (pthreads). I just had the experience of running the
naming by a non-Akaros person and the reaction was really
positive to the pthreads similarity.
Here's a usage example:
struct virtual_machine vm[1];
static volatile int count;
static void vmcall(void *a)
{
while (count < 1000000) {
__asm__ __volatile__("vmcall\n\t");
count++;
}
count++;
while (1);
}
int main(int argc, char **argv)
{
vthread_attr_init(vm, 0);
vthread_create(vm, 0, (uint64_t) vmcall, NULL);
while(count < 2) {
}
return 0;
}
Change-Id: I3e72d8a0efd89482d2f9856daccb7945f7ae1c92
Signed-off-by: Ronald G. Minnich <[email protected]>
---
user/vmm/include/vmm/sched.h | 4 +
user/vmm/include/vmm/vmm.h | 7 ++
user/vmm/vthread.c | 175 +++++++++++++++++++++++++++++++++++++++++++
3 files changed, 186 insertions(+)
create mode 100644 user/vmm/vthread.c
diff --git a/user/vmm/include/vmm/sched.h b/user/vmm/include/vmm/sched.h
index c9a7136..403ca3f 100644
--- a/user/vmm/include/vmm/sched.h
+++ b/user/vmm/include/vmm/sched.h
@@ -80,4 +80,8 @@ void start_guest_thread(struct guest_thread *gth);
struct task_thread *vmm_run_task(struct virtual_machine *vm,
void (*func)(void *), void *arg);
+int vthread_attr_init(struct virtual_machine *vm, int vmmflags);
+int vthread_attr_kernel_init(struct virtual_machine *vm, int vmmflags);
+int vthread_create(struct virtual_machine *vm, int guest, void *rip, void *arg);
+
__END_DECLS
diff --git a/user/vmm/include/vmm/vmm.h b/user/vmm/include/vmm/vmm.h
index 5139b2a..6577651 100644
--- a/user/vmm/include/vmm/vmm.h
+++ b/user/vmm/include/vmm/vmm.h
@@ -27,12 +27,19 @@ struct virtual_machine {
struct guest_thread **gths;
unsigned int nr_gpcs;
struct vmm_gpcore_init *gpcis;
+ bool vminit;
/* TODO: put these in appropriate structures. e.g., virtio things in
* something related to virtio. low4k in something related to the guest's
* memory. */
uint8_t *low4k;
struct virtio_mmio_dev *virtio_mmio_devices[VIRTIO_MMIO_MAX_NUM_DEV];
+
+ /* Default root pointer to use if one is not set in a
+ * guest thread. We expect this to be the common case,
+ * where all guests share a page table. It's not required
+ * however. */
+ void *root;
};
char *regname(uint8_t reg);
diff --git a/user/vmm/vthread.c b/user/vmm/vthread.c
new file mode 100644
index 0000000..9b7a397
--- /dev/null
+++ b/user/vmm/vthread.c
@@ -0,0 +1,175 @@
+/* Copyright (c) 2016 Google Inc.
+ *
+ * See LICENSE for details.
+ *
+ * Helper functions for virtual machines */
+
+#include <errno.h>
+#include <stdlib.h>
+#include <parlib/bitmask.h>
+#include <parlib/uthread.h>
+#include <sys/mman.h>
+#include <sys/syscall.h>
+#include <vmm/vmm.h>
+
+
+static void *page(void *addr, int count)
+{
+ void *v;
+ unsigned long flags = MAP_POPULATE | MAP_ANONYMOUS;
+
+ if (addr)
+ flags |= MAP_FIXED;
+ return mmap(addr, count * 4096, PROT_READ | PROT_WRITE, flags, -1, 0);
+}
+
+/* vmsetup is a basic helper function used by vthread_attr_init
+ * and vthread_attr_kernel_init. */
+static int vmsetup(struct virtual_machine *vm, int flags)
+{
+ unsigned long long *p512, *p1;
+ struct vm_trapframe *vm_tf;
+ int i, ret;
+ uint8_t *p;
+
+ if (vm->vminit)
+ return -EBUSY;
+
+ if (vm->nr_gpcs == 0)
+ vm->nr_gpcs = 1;
+
+ vm->gpcis = calloc(vm->nr_gpcs, sizeof(*vm->gpcis));
+
+ /* Set up default page mappings. The common case,
+ * for user VM threads and kernel VM threads, is that
+ * they need some kind of initial page tables. The kernels
+ * will almost always throw them away; the user VM threads
+ * will almost always continue to use them. Using two
+ * pages and setting up an initial page table is
+ * cheap and makes users lives easier. This initial
+ * page table can grow to 512 GiB, which should be enough
+ * for now.
+ *
+ * At the same time, we allow users to select other
+ * arrangements if they wish. Here's a simple example: is it
+ * possible someone will want a different guest page table for
+ * every guest? Yes.
+ *
+ * We lock the page table to 0x1000000 for now. We can't just
+ * let it pick anything as it may pick something the guest
+ * can't address (i.e. outside EPT range). */
+
+ /* Allocate 2 pages for page table pages: a page of
+ * 512 GiB PTEs with only one entry filled to point to
+ * a page of 1 GiB PTEs; a page of 1 GiB PTEs with
+ * only one entry filled. */
+
+ p512 = page((void *)0x1000000, 2);
+ if (!p512) {
+ werrstr("page table allocation failed: %r\n");
+ return -1;
+ }
+ p1 = &p512[512];
+ vm->root = p512;
+
+ /* Set up a 1:1 ("identity") page mapping from host
+ * virtual to guest physical for 1 GiB. This mapping
+ * is used unless the guest (e.g. Linux) sets up its
+ * own page tables. Be aware that the values stored in
+ * the table are physical addresses. This is subtle
+ * and mistakes are easily disguised due to the
+ * identity mapping, so take care when manipulating
+ * these mappings. Note: we don't yet have symbols for
+ * "start of virtual address common to host and guest"
+ * so we just use the first GiB for now. */
+ p512[PML4(0x400000)] = (uint64_t)p1 | PTE_KERN_RW;
+ p1[PML3(0x400000)] = PTE_PS | PTE_KERN_RW;
+
+ /* technically, we don't need these pages for the
+ * all guests. Currently, the kernel requires them. */
+ for (i = 0; i < vm->nr_gpcs; i++) {
+ p = page(NULL, 3);
+ if (!p) {
+ werrstr("Can't allocate 3 pages for guest %d: %r", i);
+ return -1;
+ }
+ vm->gpcis[i].posted_irq_desc = &p[0];
+ vm->gpcis[i].vapic_addr = &p[4096];
+ vm->gpcis[i].apic_addr = &p[8192];
+ }
+
+ ret = vmm_init(vm, flags);
+ if (ret)
+ return ret;
+
+ for (i = 0; i < vm->nr_gpcs; i++) {
+ vm_tf = gth_to_vmtf(vm->gths[i]);
+ vm_tf->tf_cr3 = (uint64_t)p512;
+ }
+ vm->vminit = 1;
+
+ return 0;
+}
+
+/* vthread_addr sets up a virtual_machine struct such that functions
+ * can start up VM guests. It is like pthread_attr in that it sets up
+ * default attributes and can be used in vthread_create calls. If
+ * vm->nrgpcs is not set then the vm will be set up for 1 guest. */
+int vthread_attr_init(struct virtual_machine *vm, int vmmflags)
+{
+ return vmsetup(vm, vmmflags);
+}
+
+/* vthread_attr_kernel_init sets up minimum basic attributes for
+ * running a kernel, as opposed to just user mode. This setup
+ * includes an APIC page at 0xfee00000, to be shared by all cores. */
+int vthread_attr_kernel_init(struct virtual_machine *vm, int vmmflags)
+{
+ int ret;
+ int i;
+ uint32_t *apic;
+
+ ret = vmsetup(vm, vmmflags);
+ if (ret)
+ return ret;
+
+ for (i = 0; i < vm->nr_gpcs; i++) {
+ apic = vm->gpcis[i].apic_addr;
+ apic[0x30 / 4] = 0x01060015;
+ }
+ return 0;
+}
+
+#define DEFAULT_STACK_SIZE 65536
+/* vthread_create creates and starts a VM guest. The interface is intended
+ * to be as much like pthread_create as possible. */
+int vthread_create(struct virtual_machine *vm, int guest, void *rip, void *arg)
+{
+ struct vm_trapframe *vm_tf;
+ int ret;
+ uint64_t *stack;
+
+ if (!vm->vminit)
+ return -EAGAIN;
+
+ if (guest > vm->nr_gpcs)
+ return -ENOENT;
+
+ vm_tf = gth_to_vmtf(vm->gths[guest]);
+
+ /* For now we make the default VM stack pretty small.
+ * We can grow it as needed. */
+ if (!vm_tf->tf_rsp) {
+ ret = posix_memalign((void **)&stack, 4096, DEFAULT_STACK_SIZE);
+ if (ret)
+ return ret;
+ /* touch the top word on the stack so we don't page fault
+ * on that in the VM. */
+ stack[DEFAULT_STACK_SIZE/sizeof(uint64_t) - 1] = 0;
+ vm_tf->tf_rsp = (uint64_t)stack;
+ }
+ vm_tf->tf_rip = (uint64_t)rip;
+ vm_tf->tf_rdi = (uint64_t)arg;
+ start_guest_thread(vm->gths[guest]);
+ return 0;
+}
--
2.8.0.rc3.226.g39d4020
--
You received this message because you are subscribed to the Google Groups "Akaros" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
To post to this group, send email to [email protected].
For more options, visit https://groups.google.com/d/optout.
