[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[RFC PATCH 1/1] xen: Use a global mapping for runstate


  • To: xen-devel@xxxxxxxxxxxxxxxxxxxx
  • From: Bertrand Marquis <bertrand.marquis@xxxxxxx>
  • Date: Thu, 28 May 2020 16:25:31 +0100
  • Arc-authentication-results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=arm.com; dmarc=pass action=none header.from=arm.com; dkim=pass header.d=arm.com; arc=none
  • Arc-message-signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=x49loOKd0eLq+Ci8SrugkRc00z80hcsQX17k9oTzHC4=; b=IKpw31S1XMY5MOdkvnBuhq6iOJunbGJ+ZFTdfZL/8bBFPxFk2CWNGFdlrkisV4qiVavUTM1LzzTNDl28r8pUxqiHQeT+5XcS7yW1EiCy3kDeME0QUapB25PInxPD+ZlWaYoWy309iNM270iOYj4umaO4qanSTvmqHtvYL5X8InUmYa5wMzsdzAl8NCgKoiHUP6w/kreCbPvBCAzWofTjYFjOuR8ytjOKQz/lxXuPGd7v79Bbr8UFOrhz5Q9YmjNxzem7O02RCJPwSoQCw3/W+uOjqt46cQMwWcixHbDAgVKie0tbF4afrMeAMcE8c60CnaCysrnQ9cPzBDJ9lHxanw==
  • Arc-seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=fGDDAeSlF0aiahxa37OAMA9PqojcavNfZKKGZp4gjwqMkKY1ugUov0s1sC70CA0C0uEtAXEW4WNDvIsQYhu8d1YbueN4RH9elvo7lcHLkZWi/f9ETEwtET+/+SkvX+wXJpbx5vvcsIXI4VZeh0coezBzqDkwi+4UtHCvvZOnGk3k3/xrj5kiw9GtOT/XyOGjLPSX3kfHRdw9Q9pn83a+H4VoARjSwYwUy04b2AVHlM019tMq9sXk2BOQkU0atCg/pTB3+sPGj9VtmdbRHdFePzFY9jDc7yezTj/SC7fY5841tO/uCRY2UieR7Xyz6kJ/sI1qupEjdx9hJrJForEQ9w==
  • Authentication-results-original: lists.xenproject.org; dkim=none (message not signed) header.d=none; lists.xenproject.org; dmarc=none action=none header.from=arm.com;
  • Cc: Stefano Stabellini <sstabellini@xxxxxxxxxx>, Julien Grall <julien@xxxxxxx>, Wei Liu <wl@xxxxxxx>, Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Ian Jackson <ian.jackson@xxxxxxxxxxxxx>, George Dunlap <george.dunlap@xxxxxxxxxx>, Jan Beulich <jbeulich@xxxxxxxx>, nd@xxxxxxx, Volodymyr Babchuk <Volodymyr_Babchuk@xxxxxxxx>, Roger Pau Monné <roger.pau@xxxxxxxxxx>
  • Delivery-date: Thu, 28 May 2020 15:27:10 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
  • Nodisclaimer: true
  • Original-authentication-results: lists.xenproject.org; dkim=none (message not signed) header.d=none;lists.xenproject.org; dmarc=none action=none header.from=arm.com;

At the moment on Arm, a Linux guest running with KTPI enabled will
cause the following error when a context switch happens in user mode:
(XEN) p2m.c:1890: d1v0: Failed to walk page-table va 0xffffff837ebe0cd0

This patch is modifying runstate handling to map the area given by the
guest inside Xen during the hypercall.
This is removing the guest virtual to physical conversion during context
switches which removes the bug and improve performance by preventing to
walk page tables during context switches.

--
In the current status, this patch is only working on Arm and needs to
be fixed on X86 (see #error on domain.c for missing get_page_from_gva).

Signed-off-by: Bertrand Marquis <bertrand.marquis@xxxxxxx>
---
 xen/arch/arm/domain.c   | 32 +++++++++-------
 xen/arch/x86/domain.c   | 51 ++++++++++++++-----------
 xen/common/domain.c     | 84 ++++++++++++++++++++++++++++++++++-------
 xen/include/xen/sched.h | 11 ++++--
 4 files changed, 124 insertions(+), 54 deletions(-)

diff --git a/xen/arch/arm/domain.c b/xen/arch/arm/domain.c
index 31169326b2..799b0e0103 100644
--- a/xen/arch/arm/domain.c
+++ b/xen/arch/arm/domain.c
@@ -278,33 +278,37 @@ static void ctxt_switch_to(struct vcpu *n)
 /* Update per-VCPU guest runstate shared memory area (if registered). */
 static void update_runstate_area(struct vcpu *v)
 {
-    void __user *guest_handle = NULL;
-    struct vcpu_runstate_info runstate;
+    struct vcpu_runstate_info *runstate;
 
-    if ( guest_handle_is_null(runstate_guest(v)) )
+    /* XXX why do we accept not to block here */
+    if ( !spin_trylock(&v->runstate_guest_lock) )
         return;
 
-    memcpy(&runstate, &v->runstate, sizeof(runstate));
+    runstate = runstate_guest(v);
+
+    if (runstate == NULL)
+    {
+        spin_unlock(&v->runstate_guest_lock);
+        return;
+    }
 
     if ( VM_ASSIST(v->domain, runstate_update_flag) )
     {
-        guest_handle = &v->runstate_guest.p->state_entry_time + 1;
-        guest_handle--;
-        runstate.state_entry_time |= XEN_RUNSTATE_UPDATE;
-        __raw_copy_to_guest(guest_handle,
-                            (void *)(&runstate.state_entry_time + 1) - 1, 1);
+        runstate->state_entry_time |= XEN_RUNSTATE_UPDATE;
         smp_wmb();
+        v->runstate.state_entry_time |= XEN_RUNSTATE_UPDATE;
     }
 
-    __copy_to_guest(runstate_guest(v), &runstate, 1);
+    memcpy(runstate, &v->runstate, sizeof(v->runstate));
 
-    if ( guest_handle )
+    if ( VM_ASSIST(v->domain, runstate_update_flag) )
     {
-        runstate.state_entry_time &= ~XEN_RUNSTATE_UPDATE;
+        runstate->state_entry_time &= ~XEN_RUNSTATE_UPDATE;
         smp_wmb();
-        __raw_copy_to_guest(guest_handle,
-                            (void *)(&runstate.state_entry_time + 1) - 1, 1);
+        v->runstate.state_entry_time &= ~XEN_RUNSTATE_UPDATE;
     }
+
+    spin_unlock(&v->runstate_guest_lock);
 }
 
 static void schedule_tail(struct vcpu *prev)
diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index 6327ba0790..10c6ceb561 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -1642,57 +1642,62 @@ void paravirt_ctxt_switch_to(struct vcpu *v)
 /* Update per-VCPU guest runstate shared memory area (if registered). */
 bool update_runstate_area(struct vcpu *v)
 {
-    bool rc;
     struct guest_memory_policy policy = { .nested_guest_mode = false };
-    void __user *guest_handle = NULL;
-    struct vcpu_runstate_info runstate;
+    struct vcpu_runstate_info *runstate;
 
-    if ( guest_handle_is_null(runstate_guest(v)) )
+    /* XXX: should we return false ? */
+    if ( !spin_trylock(&v->runstate_guest_lock) )
         return true;
 
-    update_guest_memory_policy(v, &policy);
+    runstate = runstate_guest(v);
 
-    memcpy(&runstate, &v->runstate, sizeof(runstate));
+    if (runstate == NULL)
+    {
+        spin_unlock(&v->runstate_guest_lock);
+        return true;
+    }
+
+    update_guest_memory_policy(v, &policy);
 
     if ( VM_ASSIST(v->domain, runstate_update_flag) )
     {
-        guest_handle = has_32bit_shinfo(v->domain)
-            ? &v->runstate_guest.compat.p->state_entry_time + 1
-            : &v->runstate_guest.native.p->state_entry_time + 1;
-        guest_handle--;
-        runstate.state_entry_time |= XEN_RUNSTATE_UPDATE;
-        __raw_copy_to_guest(guest_handle,
-                            (void *)(&runstate.state_entry_time + 1) - 1, 1);
+        runstate->state_entry_time |= XEN_RUNSTATE_UPDATE;
         smp_wmb();
+        if (has_32bit_shinfo(v->domain))
+            v->runstate_guest.compat->state_entry_time |= XEN_RUNSTATE_UPDATE;
+        else
+            v->runstate_guest.native->state_entry_time |= XEN_RUNSTATE_UPDATE;
     }
 
     if ( has_32bit_shinfo(v->domain) )
     {
         struct compat_vcpu_runstate_info info;
-
         XLAT_vcpu_runstate_info(&info, &runstate);
-        __copy_to_guest(v->runstate_guest.compat, &info, 1);
-        rc = true;
+        memcpy(v->runstate_guest.compat, &info, 1);
     }
     else
-        rc = __copy_to_guest(runstate_guest(v), &runstate, 1) !=
-             sizeof(runstate);
+        memcpy(runstate, &v->runstate, sizeof(v->runstate));
 
-    if ( guest_handle )
+    if ( VM_ASSIST(v->domain, runstate_update_flag) )
     {
-        runstate.state_entry_time &= ~XEN_RUNSTATE_UPDATE;
+        runstate->state_entry_time |= XEN_RUNSTATE_UPDATE;
         smp_wmb();
-        __raw_copy_to_guest(guest_handle,
-                            (void *)(&runstate.state_entry_time + 1) - 1, 1);
+        if (has_32bit_shinfo(v->domain))
+            v->runstate_guest.compat->state_entry_time |= XEN_RUNSTATE_UPDATE;
+        else
+            v->runstate_guest.native->state_entry_time |= XEN_RUNSTATE_UPDATE;
     }
 
+    spin_unlock(&v->runstate_guest_lock);
+
     update_guest_memory_policy(v, &policy);
 
-    return rc;
+    return true;
 }
 
 static void _update_runstate_area(struct vcpu *v)
 {
+    /* XXX: this should be removed */
     if ( !update_runstate_area(v) && is_pv_vcpu(v) &&
          !(v->arch.flags & TF_kernel_mode) )
         v->arch.pv.need_update_runstate_area = 1;
diff --git a/xen/common/domain.c b/xen/common/domain.c
index 7cc9526139..acc6f90ba3 100644
--- a/xen/common/domain.c
+++ b/xen/common/domain.c
@@ -161,6 +161,7 @@ struct vcpu *vcpu_create(struct domain *d, unsigned int 
vcpu_id)
     v->dirty_cpu = VCPU_CPU_CLEAN;
 
     spin_lock_init(&v->virq_lock);
+    spin_lock_init(&v->runstate_guest_lock);
 
     tasklet_init(&v->continue_hypercall_tasklet, NULL, NULL);
 
@@ -691,6 +692,66 @@ int rcu_lock_live_remote_domain_by_id(domid_t dom, struct 
domain **d)
     return 0;
 }
 
+static void  unmap_runstate_area(struct vcpu *v, unsigned int lock)
+{
+    mfn_t mfn;
+
+    if ( ! runstate_guest(v) )
+        return;
+
+    if (lock)
+        spin_lock(&v->runstate_guest_lock);
+
+    mfn = domain_page_map_to_mfn(runstate_guest(v));
+
+    unmap_domain_page_global((void *)
+                            ((unsigned long)v->runstate_guest &
+                             PAGE_MASK));
+
+    put_page_and_type(mfn_to_page(mfn));
+    runstate_guest(v) = NULL;
+
+    if (lock)
+        spin_unlock(&v->runstate_guest_lock);
+}
+
+static int map_runstate_area(struct vcpu *v,
+                    struct vcpu_register_runstate_memory_area *area)
+{
+    unsigned long offset = area->addr.p & ~PAGE_MASK;
+    void *mapping;
+    struct page_info *page;
+    size_t size = sizeof(struct vcpu_runstate_info);
+
+    ASSERT(runstate_guest(v) == NULL);
+
+    /* do not allow an area crossing 2 pages */
+    if ( offset > (PAGE_SIZE - size) )
+        return -EINVAL;
+
+#ifdef CONFIG_ARM
+    page = get_page_from_gva(v, area->addr.p, GV2M_WRITE);
+#else
+    /* XXX how to solve this one ? */
+#error get_page_from_gva is not available on other archs
+#endif
+    if ( !page )
+        return -EINVAL;
+
+    mapping = __map_domain_page_global(page);
+
+    if ( mapping == NULL )
+    {
+        put_page_and_type(page);
+        return -ENOMEM;
+    }
+
+    runstate_guest(v) = (struct vcpu_runstate_info *)
+        ((unsigned long)mapping + offset);
+
+    return 0;
+}
+
 int domain_kill(struct domain *d)
 {
     int rc = 0;
@@ -727,7 +788,10 @@ int domain_kill(struct domain *d)
         if ( cpupool_move_domain(d, cpupool0) )
             return -ERESTART;
         for_each_vcpu ( d, v )
+        {
+            unmap_runstate_area(v, 0);
             unmap_vcpu_info(v);
+        }
         d->is_dying = DOMDYING_dead;
         /* Mem event cleanup has to go here because the rings 
          * have to be put before we call put_domain. */
@@ -1167,7 +1231,7 @@ int domain_soft_reset(struct domain *d)
 
     for_each_vcpu ( d, v )
     {
-        set_xen_guest_handle(runstate_guest(v), NULL);
+        unmap_runstate_area(v, 1);
         unmap_vcpu_info(v);
     }
 
@@ -1484,7 +1548,6 @@ long do_vcpu_op(int cmd, unsigned int vcpuid, 
XEN_GUEST_HANDLE_PARAM(void) arg)
     case VCPUOP_register_runstate_memory_area:
     {
         struct vcpu_register_runstate_memory_area area;
-        struct vcpu_runstate_info runstate;
 
         rc = -EFAULT;
         if ( copy_from_guest(&area, arg, 1) )
@@ -1493,18 +1556,13 @@ long do_vcpu_op(int cmd, unsigned int vcpuid, 
XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( !guest_handle_okay(area.addr.h, 1) )
             break;
 
-        rc = 0;
-        runstate_guest(v) = area.addr.h;
+        spin_lock(&v->runstate_guest_lock);
 
-        if ( v == current )
-        {
-            __copy_to_guest(runstate_guest(v), &v->runstate, 1);
-        }
-        else
-        {
-            vcpu_runstate_get(v, &runstate);
-            __copy_to_guest(runstate_guest(v), &runstate, 1);
-        }
+        unmap_runstate_area(v, 0);
+
+        rc = map_runstate_area(v, &area);
+
+        spin_unlock(&v->runstate_guest_lock);
 
         break;
     }
diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
index ac53519d7f..ab87182e81 100644
--- a/xen/include/xen/sched.h
+++ b/xen/include/xen/sched.h
@@ -166,15 +166,18 @@ struct vcpu
     struct sched_unit *sched_unit;
 
     struct vcpu_runstate_info runstate;
+
+    spinlock_t      runstate_guest_lock;
+
 #ifndef CONFIG_COMPAT
 # define runstate_guest(v) ((v)->runstate_guest)
-    XEN_GUEST_HANDLE(vcpu_runstate_info_t) runstate_guest; /* guest address */
+    vcpu_runstate_info_t *runstate_guest; /* mapped address of guest copy */
 #else
 # define runstate_guest(v) ((v)->runstate_guest.native)
     union {
-        XEN_GUEST_HANDLE(vcpu_runstate_info_t) native;
-        XEN_GUEST_HANDLE(vcpu_runstate_info_compat_t) compat;
-    } runstate_guest; /* guest address */
+        vcpu_runstate_info_t *native;
+        vcpu_runstate_info_compat_t *compat;
+    } runstate_guest; /* mapped address of guest copy */
 #endif
     unsigned int     new_state;
 
-- 
2.17.1




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.