|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [Xen-devel] [PATCH v9 11/15] microcode: unify loading update during CPU resuming and AP wakeup
On Thu, Aug 22, 2019 at 04:10:46PM +0200, Roger Pau Monné wrote:
>On Mon, Aug 19, 2019 at 09:25:24AM +0800, Chao Gao wrote:
>> Both are loading the cached patch. Since APs call the unified function,
>> microcode_update_one(), during wakeup, the 'start_update' parameter
>> which originally used to distinguish BSP and APs is redundant. So remove
>> this parameter.
>>
>> Signed-off-by: Chao Gao <chao.gao@xxxxxxxxx>
>> ---
>> Note that here is a functional change: resuming a CPU would call
>> ->end_update() now while previously it wasn't. Not quite sure
>> whether it is correct.
>
>I guess that's required if it called start_update prior to calling
>end_update?
>
>>
>> Changes in v9:
>> - return -EOPNOTSUPP rather than 0 if microcode_ops is NULL in
>> microcode_update_one()
>> - rebase and fix conflicts.
>>
>> Changes in v8:
>> - split out from the previous patch
>> ---
>> xen/arch/x86/acpi/power.c | 2 +-
>> xen/arch/x86/microcode.c | 90
>> ++++++++++++++++++-----------------------
>> xen/arch/x86/smpboot.c | 5 +--
>> xen/include/asm-x86/processor.h | 4 +-
>> 4 files changed, 44 insertions(+), 57 deletions(-)
>>
>> diff --git a/xen/arch/x86/acpi/power.c b/xen/arch/x86/acpi/power.c
>> index 4f21903..24798d5 100644
>> --- a/xen/arch/x86/acpi/power.c
>> +++ b/xen/arch/x86/acpi/power.c
>> @@ -253,7 +253,7 @@ static int enter_state(u32 state)
>>
>> console_end_sync();
>>
>> - microcode_resume_cpu();
>> + microcode_update_one();
>>
>> if ( !recheck_cpu_features(0) )
>> panic("Missing previously available feature(s)\n");
>> diff --git a/xen/arch/x86/microcode.c b/xen/arch/x86/microcode.c
>> index a2febc7..bdd9c9f 100644
>> --- a/xen/arch/x86/microcode.c
>> +++ b/xen/arch/x86/microcode.c
>> @@ -203,24 +203,6 @@ static struct microcode_patch *parse_blob(const char
>> *buf, uint32_t len)
>> return NULL;
>> }
>>
>> -int microcode_resume_cpu(void)
>> -{
>> - int err;
>> - struct cpu_signature *sig = &this_cpu(cpu_sig);
>> -
>> - if ( !microcode_ops )
>> - return 0;
>> -
>> - spin_lock(µcode_mutex);
>> -
>> - err = microcode_ops->collect_cpu_info(sig);
>> - if ( likely(!err) )
>> - err = microcode_ops->apply_microcode(microcode_cache);
>> - spin_unlock(µcode_mutex);
>> -
>> - return err;
>> -}
>> -
>> void microcode_free_patch(struct microcode_patch *microcode_patch)
>> {
>> microcode_ops->free_patch(microcode_patch->mc);
>> @@ -384,11 +366,29 @@ static int __init microcode_init(void)
>> }
>> __initcall(microcode_init);
>>
>> -int __init early_microcode_update_cpu(bool start_update)
>> +/* Load a cached update to current cpu */
>> +int microcode_update_one(void)
>> +{
>> + int rc;
>> +
>> + if ( !microcode_ops )
>> + return -EOPNOTSUPP;
>> +
>> + rc = microcode_update_cpu(NULL);
>> +
>> + if ( microcode_ops->end_update )
>> + microcode_ops->end_update();
>
>Don't you need to call start_update before calling
>microcode_update_cpu?
No. On AMD side, osvw_status records the hardware erratum in the system.
As we don't assume all CPUs have the same erratum, each cpu calls
end_update to update osvw_status after ucode loading.
start_update just resets osvw_status to 0. And it is called once prior
to ucode loading on any CPU so that osvw_status can be recomputed.
Thanks
Chao
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel
|
![]() |
Lists.xenproject.org is hosted with RackSpace, monitoring our |