[Xen-devel] [PATCH RFC] x86/EFI: take care of fully eager FPU around runtime calls

["Jan Beulich" <jbeulich@xxxxxxxx>]

Just like in the HVM emulation case we must not set CR0.TS here in
fully eager mode. Note that idle vCPU-s never have
->arch.fully_eager_fpu set (for their initialization not going through
vcpu_init_fpu()), so we won't hit the respective ASSERT() in
vcpu_restore_fpu_eager().

Introduce a "curr" local variable and replace other uses of "current"
at the same time.

Reported-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
RFC: Not even compile tested, as I'm writing this from home. Also please
     excuse the formatting (hence the attachment) - our mail web frontend

     doesn't allow anything better.


--- a/xen/common/efi/runtime.c
+++ b/xen/common/efi/runtime.c
@@ -118,14 +118,16 @@ struct efi_rs_state efi_rs_enter(void)
 
 void efi_rs_leave(struct efi_rs_state *state)
 {
+    struct vcpu *curr = current;
+
     if ( !state->cr3 )
         return;
     write_cr3(state->cr3);
-    if ( is_pv_vcpu(current) && !is_idle_vcpu(current) )
+    if ( is_pv_vcpu(curr) && !is_idle_vcpu(curr) )
     {
         struct desc_ptr gdt_desc = {
             .limit = LAST_RESERVED_GDT_BYTE,
-            .base  = GDT_VIRT_START(current)
+            .base  = GDT_VIRT_START(curr)
         };
 
         asm volatile ( "lgdt %0" : : "m" (gdt_desc) );
@@ -133,7 +135,10 @@ void efi_rs_leave(struct efi_rs_state *state)
     irq_exit();
     efi_rs_on_cpu = NR_CPUS;
     spin_unlock(&efi_rs_lock);
-    stts();
+    if ( curr->arch.fully_eager_fpu )
+        vcpu_restore_fpu_eager(curr);
+    else
+        stts();
 }
 
 bool efi_rs_using_pgtables(void)

Attachment: x86-EFI-eager-FPU.patch
Description: Text document

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