Skip to contentSkip to conversation

x86/HVM: drop stale check from hvm_load_cpu_msrs()

PATCH Superseded Jan Beulich 2 participants 3 messages Nov 29, 2022 – Nov 30, 2022

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86/HVM: drop stale check from hvm_load_cpu_msrs()

Up until f61685a66903 ("x86: remove defunct init/load/save_msr() hvm_funcs") the check of the _rsvd field served as an error check for the earlier hvm_funcs.save_msr() invocation. With that invocation gone the check makes no sense anymore. While dropping it also merge the two paths setting "err" to -ENXIO.

---
We could go further here, removing the local "err" variable altogether, by using "return -ENXIO". Thoughts.

xen/arch/x86/hvm/hvm.c +4 -5
@@ -1494,13 +1494,12 @@ static int cf_check hvm_load_cpu_msrs(st
         case MSR_AMD64_DR1_ADDRESS_MASK ... MSR_AMD64_DR3_ADDRESS_MASK:
             rc = guest_wrmsr(v, ctxt->msr[i].index, ctxt->msr[i].val);
 
-            if ( rc != X86EMUL_OKAY )
-                err = -ENXIO;
-            break;
+            if ( rc == X86EMUL_OKAY )
+                continue;
 
+            fallthrough;
         default:
-            if ( !ctxt->msr[i]._rsvd )
-                err = -ENXIO;
+            err = -ENXIO;
             break;
         }
     }
#1 lore raw
Quoted 10 lines · Jan Beulich

Up until f61685a66903 ("x86: remove defunct init/load/save_msr() hvm_funcs") the check of the _rsvd field served as an error check for the earlier hvm_funcs.save_msr() invocation. With that invocation gone the check makes no sense anymore. While dropping it also merge the two paths setting "err" to -ENXIO.

---
We could go further here, removing the local "err" variable altogether, by using "return -ENXIO". Thoughts.

'err' is a non-standard variable name, so yeah, why not.

That said, the current code has a split loop checking the incoming _rsvd fields in a first pass, and then calling guest_wrmsr() on the second pass.  This was also made pointless by the identified changeset, so the two loops ought to be merged.

~Andrew

#2 lore raw
Quoted 7 lines · Andrew Cooper
Quoted 6 lines · Jan Beulich

Up until f61685a66903 ("x86: remove defunct init/load/save_msr() hvm_funcs") the check of the _rsvd field served as an error check for the earlier hvm_funcs.save_msr() invocation. With that invocation gone the check makes no sense anymore. While dropping it also merge the two paths setting "err" to -ENXIO.

Quoted 6 lines
Quoted 4 lines

---
We could go further here, removing the local "err" variable altogether, by using "return -ENXIO". Thoughts.

'err' is a non-standard variable name, so yeah, why not.

Okay, I'll make a v2 for this.

Quoted 4 lines

That said, the current code has a split loop checking the incoming _rsvd fields in a first pass, and then calling guest_wrmsr() on the second pass.  This was also made pointless by the identified changeset, so the two loops ought to be merged.

Not really, no - it would violate the "Checking finished" comment (but of course we could also delete that one), but I'd also prefer to keep checking for all errors we can check for early _before_ starting to make any changes to the guest. Therefore if you really wanted that, I guess you'd need to make a follow-on change yourself, with a convincing justification (I wouldn't outright object to such a change, but I probably also wouldn't ack it, leaving that to someone else).

Jan