Xen-Devel Archive on lore.kernel.org
 help / color / Atom feed
From: Sergey Dyasli <sergey.dyasli@citrix.com>
To: Chao Gao <chao.gao@intel.com>
Cc: "sergey.dyasli@citrix.com >> Sergey Dyasli"
	<sergey.dyasli@citrix.com>, "Ashok Raj" <ashok.raj@intel.com>,
	"Wei Liu" <wl@xen.org>,
	"Andrew Cooper" <andrew.cooper3@citrix.com>,
	"Jan Beulich" <jbeulich@suse.com>,
	xen-devel@lists.xenproject.org,
	"Roger Pau Monné" <roger.pau@citrix.com>
Subject: Re: [Xen-devel] [PATCH v9 15/15] microcode: block #NMI handling when loading an ucode
Date: Wed, 28 Aug 2019 09:52:00 +0100
Message-ID: <5b22b82b-e3cc-3ead-46f2-624ff9e5d2d7@citrix.com> (raw)
In-Reply-To: <20190827045159.GA28509@gao-cwp>

On 27/08/2019 05:52, Chao Gao wrote:
> On Mon, Aug 26, 2019 at 04:07:59PM +0800, Chao Gao wrote:
>> On Fri, Aug 23, 2019 at 09:46:37AM +0100, Sergey Dyasli wrote:
>>> On 19/08/2019 02:25, Chao Gao wrote:
>>>> register an nmi callback. And this callback does busy-loop on threads
>>>> which are waiting for loading completion. Control threads send NMI to
>>>> slave threads to prevent NMI acceptance during ucode loading.
>>>>
>>>> Signed-off-by: Chao Gao <chao.gao@intel.com>
>>>> ---
>>>> Changes in v9:
>>>>  - control threads send NMI to all other threads. Slave threads will
>>>>  stay in the NMI handling to prevent NMI acceptance during ucode
>>>>  loading. Note that self-nmi is invalid according to SDM.
>>>
>>> To me this looks like a half-measure: why keep only slave threads in
>>> the NMI handler, when master threads can update the microcode from
>>> inside the NMI handler as well?
>>
>> No special reason. Because the issue we want to address is that slave
>> threads might go to handle NMI and access MSRs when master thread is
>> loading ucode. So we only keep slave threads in the NMI handler.
>>
>>>
>>> You mention that self-nmi is invalid, but Xen has self_nmi() which is
>>> used for apply_alternatives() during boot, so can be trusted to work.
>>
>> Sorry, I meant using self shorthand to send self-nmi. I tried to use
>> self shorthand but got APIC error. And I agree that it is better to
>> make slave thread call self_nmi() itself.
>>
>>>
>>> I experimented a bit with the following approach: after loading_state
>>> becomes LOADING_CALLIN, each cpu issues a self_nmi() and rendezvous
>>> via cpu_callin_map into LOADING_ENTER to do a ucode update directly in
>>> the NMI handler. And it seems to work.
>>>
>>> Separate question is about the safety of this approach: can we be sure
>>> that a ucode update would not reset the status of the NMI latch? I.e.
>>> can it cause another NMI to be delivered while Xen already handles one?
>>
>> Ashok, what's your opinion on Sergey's approach and his concern?
> 
> Hi Sergey,
> 
> I talked with Ashok. We think your approach is better. I will follow
> your approach in v10. It would be much helpful if you post your patch
> so that I can just rebase it onto other patches.

Sure thing. The below code is my first attempt at improving the original
patch. It can benefit from some further refactoring.

---
 xen/arch/x86/microcode.c | 108 ++++++++++++++++++++++++++++-----------
 1 file changed, 79 insertions(+), 29 deletions(-)

diff --git a/xen/arch/x86/microcode.c b/xen/arch/x86/microcode.c
index 91f9e811f8..ba2363406f 100644
--- a/xen/arch/x86/microcode.c
+++ b/xen/arch/x86/microcode.c
@@ -36,8 +36,10 @@
 #include <xen/earlycpio.h>
 #include <xen/watchdog.h>

+#include <asm/apic.h>
 #include <asm/delay.h>
 #include <asm/msr.h>
+#include <asm/nmi.h>
 #include <asm/processor.h>
 #include <asm/setup.h>
 #include <asm/microcode.h>
@@ -232,6 +234,7 @@ DEFINE_PER_CPU(struct cpu_signature, cpu_sig);
  */
 static cpumask_t cpu_callin_map;
 static atomic_t cpu_out, cpu_updated;
+struct microcode_patch *nmi_patch;

 /*
  * Return a patch that covers current CPU. If there are multiple patches,
@@ -337,15 +340,25 @@ static int microcode_update_cpu(const struct microcode_patch *patch)
     return err;
 }

+static void slave_thread_work(void)
+{
+    /* Do nothing, just wait */
+    while ( loading_state != LOADING_EXIT )
+        cpu_relax();
+}
+
 static int slave_thread_fn(void)
 {
-    unsigned int cpu = smp_processor_id();
     unsigned int master = cpumask_first(this_cpu(cpu_sibling_mask));

     while ( loading_state != LOADING_CALLIN )
+    {
+        if ( loading_state == LOADING_EXIT )
+            return 0;
         cpu_relax();
+    }

-    cpumask_set_cpu(cpu, &cpu_callin_map);
+    self_nmi();

     while ( loading_state != LOADING_EXIT )
         cpu_relax();
@@ -356,30 +369,35 @@ static int slave_thread_fn(void)
     return 0;
 }

-static int master_thread_fn(const struct microcode_patch *patch)
+static void master_thread_work(void)
 {
-    unsigned int cpu = smp_processor_id();
-    int ret = 0;
-
-    while ( loading_state != LOADING_CALLIN )
-        cpu_relax();
-
-    cpumask_set_cpu(cpu, &cpu_callin_map);
+    int ret;

     while ( loading_state != LOADING_ENTER )
+    {
+        if ( loading_state == LOADING_EXIT )
+            return;
         cpu_relax();
+    }

-    /*
-     * If an error happened, control thread would set 'loading_state'
-     * to LOADING_EXIT. Don't perform ucode loading for this case
-     */
-    if ( loading_state == LOADING_EXIT )
-        return ret;
-
-    ret = microcode_ops->apply_microcode(patch);
+    ret = microcode_ops->apply_microcode(nmi_patch);
     if ( !ret )
         atomic_inc(&cpu_updated);
     atomic_inc(&cpu_out);
+}
+
+static int master_thread_fn(const struct microcode_patch *patch)
+{
+    int ret = 0;
+
+    while ( loading_state != LOADING_CALLIN )
+    {
+        if ( loading_state == LOADING_EXIT )
+            return ret;
+        cpu_relax();
+    }
+
+    self_nmi();

     while ( loading_state != LOADING_EXIT )
         cpu_relax();
@@ -387,35 +405,40 @@ static int master_thread_fn(const struct microcode_patch *patch)
     return ret;
 }

-static int control_thread_fn(const struct microcode_patch *patch)
+static void control_thread_work(void)
 {
-    unsigned int cpu = smp_processor_id(), done;
-    unsigned long tick;
     int ret;

-    /* Allow threads to call in */
-    loading_state = LOADING_CALLIN;
-    smp_mb();
-
-    cpumask_set_cpu(cpu, &cpu_callin_map);
-
     /* Waiting for all threads calling in */
     ret = wait_for_condition(wait_cpu_callin,
                              (void *)(unsigned long)num_online_cpus(),
                              MICROCODE_CALLIN_TIMEOUT_US);
     if ( ret ) {
         loading_state = LOADING_EXIT;
-        return ret;
+        return;
     }

     /* Let master threads load the given ucode update */
     loading_state = LOADING_ENTER;
     smp_mb();

-    ret = microcode_ops->apply_microcode(patch);
+    ret = microcode_ops->apply_microcode(nmi_patch);
     if ( !ret )
         atomic_inc(&cpu_updated);
     atomic_inc(&cpu_out);
+}
+
+static int control_thread_fn(const struct microcode_patch *patch)
+{
+    unsigned int done;
+    unsigned long tick;
+    int ret;
+
+    /* Allow threads to call in */
+    loading_state = LOADING_CALLIN;
+    smp_mb();
+
+    self_nmi();

     tick = rdtsc_ordered();
     /* Waiting for master threads finishing update */
@@ -481,12 +504,35 @@ static int do_microcode_update(void *patch)
     return ret;
 }

+static int microcode_nmi_callback(const struct cpu_user_regs *regs, int cpu)
+{
+    unsigned int master = cpumask_first(this_cpu(cpu_sibling_mask));
+    unsigned int controller = cpumask_first(&cpu_online_map);
+
+    /* System-generated NMI, will be ignored */
+    if ( loading_state == LOADING_PREPARE )
+        return 0;
+
+    ASSERT(loading_state == LOADING_CALLIN);
+    cpumask_set_cpu(cpu, &cpu_callin_map);
+
+    if ( cpu == controller )
+        control_thread_work();
+    else if ( cpu == master )
+        master_thread_work();
+    else
+        slave_thread_work();
+
+    return 0;
+}
+
 int microcode_update(XEN_GUEST_HANDLE_PARAM(const_void) buf, unsigned long len)
 {
     int ret;
     void *buffer;
     unsigned int cpu, updated;
     struct microcode_patch *patch;
+    nmi_callback_t *saved_nmi_callback;

     if ( len != (uint32_t)len )
         return -E2BIG;
@@ -551,6 +597,9 @@ int microcode_update(XEN_GUEST_HANDLE_PARAM(const_void) buf, unsigned long len)
      * watchdog timeout.
      */
     watchdog_disable();
+
+    nmi_patch = patch;
+    saved_nmi_callback = set_nmi_callback(microcode_nmi_callback);
     /*
      * Late loading dance. Why the heavy-handed stop_machine effort?
      *
@@ -563,6 +612,7 @@ int microcode_update(XEN_GUEST_HANDLE_PARAM(const_void) buf, unsigned long len)
      *   conservative and good.
      */
     ret = stop_machine_run(do_microcode_update, patch, NR_CPUS);
+    set_nmi_callback(saved_nmi_callback);
     watchdog_enable();

     updated = atomic_read(&cpu_updated);
-- 
2.17.1


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

  reply index

Thread overview: 57+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-19  1:25 [Xen-devel] [PATCH v9 00/15] improve late microcode loading Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 01/15] microcode/intel: extend microcode_update_match() Chao Gao
2019-08-28 15:12   ` Jan Beulich
2019-08-29  7:15     ` Chao Gao
2019-08-29  7:14       ` Jan Beulich
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 02/15] microcode/amd: fix memory leak Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 03/15] microcode/amd: distinguish old and mismatched ucode in microcode_fits() Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 04/15] microcode: introduce a global cache of ucode patch Chao Gao
2019-08-22 11:11   ` Roger Pau Monné
2019-08-28 15:21   ` Jan Beulich
2019-08-29 10:18   ` Jan Beulich
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 05/15] microcode: clean up microcode_resume_cpu Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 06/15] microcode: remove struct ucode_cpu_info Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 07/15] microcode: remove pointless 'cpu' parameter Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 08/15] microcode/amd: call svm_host_osvw_init() in common code Chao Gao
2019-08-22 13:08   ` Roger Pau Monné
2019-08-28 15:26   ` Jan Beulich
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 09/15] microcode: pass a patch pointer to apply_microcode() Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 10/15] microcode: split out apply_microcode() from cpu_request_microcode() Chao Gao
2019-08-22 13:59   ` Roger Pau Monné
2019-08-29 10:06     ` Jan Beulich
2019-08-30  3:22       ` Chao Gao
2019-08-30  7:25         ` Jan Beulich
2019-08-29 10:19   ` Jan Beulich
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 11/15] microcode: unify loading update during CPU resuming and AP wakeup Chao Gao
2019-08-22 14:10   ` Roger Pau Monné
2019-08-22 16:44     ` Chao Gao
2019-08-23  9:09       ` Roger Pau Monné
2019-08-29  7:37         ` Chao Gao
2019-08-29  8:16           ` Roger Pau Monné
2019-08-29 10:26           ` Jan Beulich
2019-08-29 10:29   ` Jan Beulich
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 12/15] microcode: reduce memory allocation and copy when creating a patch Chao Gao
2019-08-23  8:11   ` Roger Pau Monné
2019-08-26  7:03     ` Chao Gao
2019-08-26  8:11       ` Roger Pau Monné
2019-08-29 10:47   ` Jan Beulich
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 13/15] x86/microcode: Synchronize late microcode loading Chao Gao
2019-08-19 10:27   ` Sergey Dyasli
2019-08-19 14:49     ` Chao Gao
2019-08-29 12:06   ` Jan Beulich
2019-08-30  3:30     ` Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 14/15] microcode: remove microcode_update_lock Chao Gao
2019-08-19  1:25 ` [Xen-devel] [PATCH v9 15/15] microcode: block #NMI handling when loading an ucode Chao Gao
2019-08-23  8:46   ` Sergey Dyasli
2019-08-26  8:07     ` Chao Gao
2019-08-27  4:52       ` Chao Gao
2019-08-28  8:52         ` Sergey Dyasli [this message]
2019-08-29 12:11         ` Jan Beulich
2019-08-30  6:35           ` Chao Gao
2019-09-09  5:52             ` Chao Gao
2019-09-09  6:16               ` Jan Beulich
2019-08-29 12:22   ` Jan Beulich
2019-08-30  6:33     ` Chao Gao
2019-08-30  7:30       ` Jan Beulich
2019-08-22  7:51 ` [Xen-devel] [PATCH v9 00/15] improve late microcode loading Sergey Dyasli
2019-08-22 15:39   ` Chao Gao

Reply instructions:

You may reply publically 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=5b22b82b-e3cc-3ead-46f2-624ff9e5d2d7@citrix.com \
    --to=sergey.dyasli@citrix.com \
    --cc=andrew.cooper3@citrix.com \
    --cc=ashok.raj@intel.com \
    --cc=chao.gao@intel.com \
    --cc=jbeulich@suse.com \
    --cc=roger.pau@citrix.com \
    --cc=wl@xen.org \
    --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

Xen-Devel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/xen-devel/0 xen-devel/git/0.git
	git clone --mirror https://lore.kernel.org/xen-devel/1 xen-devel/git/1.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 xen-devel xen-devel/ https://lore.kernel.org/xen-devel \
		xen-devel@lists.xenproject.org xen-devel@lists.xen.org
	public-inbox-index xen-devel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.xenproject.lists.xen-devel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git