All of lore.kernel.org
 help / color / mirror / Atom feed
From: Juergen Gross <jgross@suse.com>
To: xen-devel@lists.xenproject.org
Cc: Juergen Gross <jgross@suse.com>,
	andrew.cooper3@citrix.com, jbeulich@suse.com
Subject: [PATCH v7 5/9] xen/x86: disable global pages for domains with XPTI active
Date: Thu, 12 Apr 2018 20:09:40 +0200	[thread overview]
Message-ID: <20180412180944.31735-6-jgross@suse.com> (raw)
In-Reply-To: <20180412180944.31735-1-jgross@suse.com>

Instead of flushing the TLB from global pages when switching address
spaces with XPTI being active just disable global pages via %cr4
completely when a domain subject to XPTI is active. This avoids the
need for extra TLB flushes as loading %cr3 will remove all TLB
entries.

In order to avoid states with cr3/cr4 having inconsistent values
(e.g. global pages being activated while cr3 already specifies a XPTI
address space) move loading of the new cr4 value to write_ptbase()
(actually to switch_cr3_cr4() called by write_ptbase()).

This requires to use switch_cr3_cr4() instead of write_ptbase() when
building dom0 in order to avoid setting cr4 with cr4.smap set.

Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
V7:
- use switch_cr3_cr4() in dom0_build.c

V6:
- don't call read_cr4() multiple times in switch_cr3_cr4()
  (Andrew Cooper)

V4:
- don't use mmu_cr4_features for setting new cr4 value (Jan Beulich)
- use simpler scheme for setting X86_CR4_PGE in
  pv_guest_cr4_to_real_cr4() (Jan Beulich)

V3:
- move cr4 loading for all domains from *_ctxt_switch_to() to
  write_cr3_cr4() called by write_ptbase() (Jan Beulich)
- rebase
---
 xen/arch/x86/domain.c          |  5 -----
 xen/arch/x86/flushtlb.c        | 17 ++++++++++++-----
 xen/arch/x86/mm.c              | 14 +++++++++++---
 xen/arch/x86/pv/dom0_build.c   |  6 +++---
 xen/arch/x86/x86_64/entry.S    | 10 ----------
 xen/common/efi/runtime.c       |  4 ++--
 xen/include/asm-x86/domain.h   |  3 ++-
 xen/include/asm-x86/flushtlb.h |  2 +-
 8 files changed, 31 insertions(+), 30 deletions(-)

diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index 3d9c19d055..9b001a03ec 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -1523,17 +1523,12 @@ void paravirt_ctxt_switch_from(struct vcpu *v)
 void paravirt_ctxt_switch_to(struct vcpu *v)
 {
     root_pgentry_t *root_pgt = this_cpu(root_pgt);
-    unsigned long cr4;
 
     if ( root_pgt )
         root_pgt[root_table_offset(PERDOMAIN_VIRT_START)] =
             l4e_from_page(v->domain->arch.perdomain_l3_pg,
                           __PAGE_HYPERVISOR_RW);
 
-    cr4 = pv_guest_cr4_to_real_cr4(v);
-    if ( unlikely(cr4 != read_cr4()) )
-        write_cr4(cr4);
-
     if ( unlikely(v->arch.debugreg[7] & DR7_ACTIVE_MASK) )
         activate_debugregs(v);
 
diff --git a/xen/arch/x86/flushtlb.c b/xen/arch/x86/flushtlb.c
index fc3b0a3268..e28bf04a37 100644
--- a/xen/arch/x86/flushtlb.c
+++ b/xen/arch/x86/flushtlb.c
@@ -89,20 +89,27 @@ static void do_tlb_flush(void)
     post_flush(t);
 }
 
-void switch_cr3(unsigned long cr3)
+void switch_cr3_cr4(unsigned long cr3, unsigned long cr4)
 {
-    unsigned long flags, cr4;
+    unsigned long flags, old_cr4;
     u32 t;
 
     /* This non-reentrant function is sometimes called in interrupt context. */
     local_irq_save(flags);
 
     t = pre_flush();
-    cr4 = read_cr4();
 
-    write_cr4(cr4 & ~X86_CR4_PGE);
+    old_cr4 = read_cr4();
+    if ( old_cr4 & X86_CR4_PGE )
+    {
+        old_cr4 = cr4 & ~X86_CR4_PGE;
+        write_cr4(old_cr4);
+    }
+
     write_cr3(cr3);
-    write_cr4(cr4);
+
+    if ( old_cr4 != cr4 )
+        write_cr4(cr4);
 
     post_flush(t);
 
diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index 9c36614099..73a38e8715 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -503,20 +503,28 @@ void make_cr3(struct vcpu *v, mfn_t mfn)
 void write_ptbase(struct vcpu *v)
 {
     struct cpu_info *cpu_info = get_cpu_info();
+    unsigned long new_cr4;
+
+    new_cr4 = (is_pv_vcpu(v) && !is_idle_vcpu(v))
+              ? pv_guest_cr4_to_real_cr4(v)
+              : ((read_cr4() & ~X86_CR4_TSD) | X86_CR4_PGE);
 
     if ( is_pv_vcpu(v) && v->domain->arch.pv_domain.xpti )
     {
         cpu_info->root_pgt_changed = true;
         cpu_info->pv_cr3 = __pa(this_cpu(root_pgt));
-        switch_cr3(v->arch.cr3);
+        switch_cr3_cr4(v->arch.cr3, new_cr4);
     }
     else
     {
-        /* Make sure to clear xen_cr3 before pv_cr3; switch_cr3() serializes. */
+        /* Make sure to clear xen_cr3 before pv_cr3. */
         cpu_info->xen_cr3 = 0;
-        switch_cr3(v->arch.cr3);
+        /* switch_cr3_cr4() serializes. */
+        switch_cr3_cr4(v->arch.cr3, new_cr4);
         cpu_info->pv_cr3 = 0;
     }
+
+    ASSERT(is_pv_vcpu(v) || read_cr4() == mmu_cr4_features);
 }
 
 /*
diff --git a/xen/arch/x86/pv/dom0_build.c b/xen/arch/x86/pv/dom0_build.c
index d148395919..4465a059a8 100644
--- a/xen/arch/x86/pv/dom0_build.c
+++ b/xen/arch/x86/pv/dom0_build.c
@@ -717,7 +717,7 @@ int __init dom0_construct_pv(struct domain *d,
         update_cr3(v);
 
     /* We run on dom0's page tables for the final part of the build process. */
-    write_ptbase(v);
+    switch_cr3_cr4(v->arch.cr3, read_cr4());
     mapcache_override_current(v);
 
     /* Copy the OS image and free temporary buffer. */
@@ -738,7 +738,7 @@ int __init dom0_construct_pv(struct domain *d,
              (parms.virt_hypercall >= v_end) )
         {
             mapcache_override_current(NULL);
-            write_ptbase(current);
+            switch_cr3_cr4(current->arch.cr3, read_cr4());
             printk("Invalid HYPERCALL_PAGE field in ELF notes.\n");
             rc = -1;
             goto out;
@@ -871,7 +871,7 @@ int __init dom0_construct_pv(struct domain *d,
 
     /* Return to idle domain's page tables. */
     mapcache_override_current(NULL);
-    write_ptbase(current);
+    switch_cr3_cr4(current->arch.cr3, read_cr4());
 
     update_domain_wallclock_time(d);
 
diff --git a/xen/arch/x86/x86_64/entry.S b/xen/arch/x86/x86_64/entry.S
index dd42223b20..5f0758d64f 100644
--- a/xen/arch/x86/x86_64/entry.S
+++ b/xen/arch/x86/x86_64/entry.S
@@ -153,13 +153,8 @@ restore_all_guest:
                 ROOT_PAGETABLE_LAST_XEN_SLOT - 1) * 8, %rdi
         rep movsq
 .Lrag_copy_done:
-        mov   STACK_CPUINFO_FIELD(cr4)(%rdx), %rdi
         mov   %r9, STACK_CPUINFO_FIELD(xen_cr3)(%rdx)
-        mov   %rdi, %rsi
-        and   $~X86_CR4_PGE, %rdi
-        mov   %rdi, %cr4
         mov   %rax, %cr3
-        mov   %rsi, %cr4
 .Lrag_keep_cr3:
 
         /* Restore stashed SPEC_CTRL value. */
@@ -215,12 +210,7 @@ restore_all_xen:
          * so "g" will have to do.
          */
 UNLIKELY_START(g, exit_cr3)
-        mov   %cr4, %rdi
-        mov   %rdi, %rsi
-        and   $~X86_CR4_PGE, %rdi
-        mov   %rdi, %cr4
         mov   %rax, %cr3
-        mov   %rsi, %cr4
 UNLIKELY_END(exit_cr3)
 
         /* WARNING! `ret`, `call *`, `jmp *` not safe beyond this point. */
diff --git a/xen/common/efi/runtime.c b/xen/common/efi/runtime.c
index 4e5ddfef4f..070a70d784 100644
--- a/xen/common/efi/runtime.c
+++ b/xen/common/efi/runtime.c
@@ -111,7 +111,7 @@ struct efi_rs_state efi_rs_enter(void)
         lgdt(&gdt_desc);
     }
 
-    switch_cr3(virt_to_maddr(efi_l4_pgtable));
+    switch_cr3_cr4(virt_to_maddr(efi_l4_pgtable), read_cr4());
 
     return state;
 }
@@ -120,7 +120,7 @@ void efi_rs_leave(struct efi_rs_state *state)
 {
     if ( !state->cr3 )
         return;
-    switch_cr3(state->cr3);
+    switch_cr3_cr4(state->cr3, read_cr4());
     if ( is_pv_vcpu(current) && !is_idle_vcpu(current) )
     {
         struct desc_ptr gdt_desc = {
diff --git a/xen/include/asm-x86/domain.h b/xen/include/asm-x86/domain.h
index ed4199931a..b7894dc8c8 100644
--- a/xen/include/asm-x86/domain.h
+++ b/xen/include/asm-x86/domain.h
@@ -618,9 +618,10 @@ unsigned long pv_guest_cr4_fixup(const struct vcpu *, unsigned long guest_cr4);
 #define pv_guest_cr4_to_real_cr4(v)                         \
     (((v)->arch.pv_vcpu.ctrlreg[4]                          \
       | (mmu_cr4_features                                   \
-         & (X86_CR4_PGE | X86_CR4_PSE | X86_CR4_SMEP |      \
+         & (X86_CR4_PSE | X86_CR4_SMEP |                    \
             X86_CR4_SMAP | X86_CR4_OSXSAVE |                \
             X86_CR4_FSGSBASE))                              \
+      | ((v)->domain->arch.pv_domain.xpti ? 0 : X86_CR4_PGE) \
       | ((v)->domain->arch.vtsc ? X86_CR4_TSD : 0))         \
      & ~X86_CR4_DE)
 #define real_cr4_to_pv_guest_cr4(c)                         \
diff --git a/xen/include/asm-x86/flushtlb.h b/xen/include/asm-x86/flushtlb.h
index 5515f73b7f..73321f948a 100644
--- a/xen/include/asm-x86/flushtlb.h
+++ b/xen/include/asm-x86/flushtlb.h
@@ -84,7 +84,7 @@ static inline unsigned long read_cr3(void)
 }
 
 /* Write pagetable base and implicitly tick the tlbflush clock. */
-void switch_cr3(unsigned long cr3);
+void switch_cr3_cr4(unsigned long cr3, unsigned long cr4);
 
 /* flush_* flag fields: */
  /*
-- 
2.13.6


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

  parent reply	other threads:[~2018-04-12 18:11 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-04-12 18:09 [PATCH v7 0/9] xen/x86: various XPTI speedups Juergen Gross
2018-04-12 18:09 ` [PATCH v7 1/9] x86/xpti: avoid copying L4 page table contents when possible Juergen Gross
2018-04-13  8:43   ` Jan Beulich
2018-04-16 15:34     ` Tim Deegan
2018-04-16 18:22       ` Juergen Gross
2018-04-18  6:55         ` Juergen Gross
2018-04-12 18:09 ` [PATCH v7 2/9] xen/x86: add a function for modifying cr3 Juergen Gross
2018-04-12 18:09 ` [PATCH v7 3/9] xen/x86: support per-domain flag for xpti Juergen Gross
2018-04-12 18:09 ` [PATCH v7 4/9] xen/x86: use invpcid for flushing the TLB Juergen Gross
2018-04-12 18:09 ` Juergen Gross [this message]
2018-04-12 18:09 ` [PATCH v7 6/9] xen/x86: use flag byte for decision whether xen_cr3 is valid Juergen Gross
2018-04-12 18:09 ` [PATCH v7 7/9] xen/x86: convert pv_guest_cr4_to_real_cr4() to a function Juergen Gross
2018-04-12 18:09 ` [PATCH v7 8/9] xen/x86: add some cr3 helpers Juergen Gross
2018-04-12 18:09 ` [PATCH v7 9/9] xen/x86: use PCID feature Juergen Gross
2018-04-13  9:59 ` [PATCH v7 0/9] xen/x86: various XPTI speedups Andrew Cooper
2018-04-13 10:29   ` Juergen Gross

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20180412180944.31735-6-jgross@suse.com \
    --to=jgross@suse.com \
    --cc=andrew.cooper3@citrix.com \
    --cc=jbeulich@suse.com \
    --cc=xen-devel@lists.xenproject.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.