From: Chao Gao <chao.gao@intel.com>
To: xen-devel@lists.xenproject.org
Cc: "Sergey Dyasli" <sergey.dyasli@citrix.com>,
"Wei Liu" <wei.liu2@citrix.com>,
"Ashok Raj" <ashok.raj@intel.com>,
"Andrew Cooper" <andrew.cooper3@citrix.com>,
"Jan Beulich" <jbeulich@suse.com>,
"Chao Gao" <chao.gao@intel.com>,
"Roger Pau Monné" <roger.pau@citrix.com>
Subject: [PATCH v6 12/12] microcode: update microcode on cores in parallel
Date: Mon, 11 Mar 2019 15:57:36 +0800 [thread overview]
Message-ID: <1552291056-20286-13-git-send-email-chao.gao@intel.com> (raw)
In-Reply-To: <1552291056-20286-1-git-send-email-chao.gao@intel.com>
microcode_update_lock and microcode_mutex prevent cores from updating
microcode in parallel. Below changes are made to support parallel
microcode update on cores.
microcode_update_lock is removed. This lock was to prevent logic threads
of a same core from updating microcode at the same time. But due to
using a global lock, it also prevented parallel microcode updating on
different cores. We ensure the siblings of a same core won't update
microcode at the caller level of apply_microcode() instead of inside
it:
1. For late microcode update, do_microcode_update() ensures that only
one sibiling thread of a core can update microcode.
2. For microcode update during system startup or CPU-hotplug, the new
early_ucode_update_lock() can guarantee update serialization of logical
threads. We use this sub-optimal method because it is hard to implement a
per-core lock at that early stage where sibling info isn't initialized yet.
3. get/put_cpu_bitmaps() prevents the concurrency of CPU-hotplug and
late microcode update.
microcode_mutex is replaced by a rwlock. microcode_mutex was used to
prevent concurrent accesses to 'uci' (already removed in previous patch)
and microcode_cache. Now the only shared resource which needs to be
protected is the microcode_cache. A rwlock allows multiple readers (one
thread of each core) to access the global cache and update microcode
simultaneously. Because the rwlock may be held in stop_machine context,
where interrupt is disabled, irq{save, restore} variants are used to
get/release the rwlock.
Note that printk in apply_microcode() and svm_host_osvm_init() (for AMD
only) are still processed sequentially.
Signed-off-by: Chao Gao <chao.gao@intel.com>
---
Changes in v6:
- introduce early_ucode_update_lock to serialize early ucode update.
Changes in v5:
- newly add
---
xen/arch/x86/microcode.c | 65 ++++++++++++++++++++++++++----------------
xen/arch/x86/microcode_amd.c | 8 +-----
xen/arch/x86/microcode_intel.c | 9 +-----
3 files changed, 43 insertions(+), 39 deletions(-)
diff --git a/xen/arch/x86/microcode.c b/xen/arch/x86/microcode.c
index 96bcef6..e7df70f 100644
--- a/xen/arch/x86/microcode.c
+++ b/xen/arch/x86/microcode.c
@@ -204,7 +204,9 @@ scan:
const struct microcode_ops *microcode_ops;
-static DEFINE_SPINLOCK(microcode_mutex);
+static DEFINE_RWLOCK(cache_rwlock);
+/* Hold this lock during early microcode update */
+static DEFINE_SPINLOCK(early_ucode_update_lock);
DEFINE_PER_CPU(struct cpu_signature, cpu_sig);
@@ -224,40 +226,47 @@ static atomic_t cpu_in, cpu_out;
*/
bool microcode_save_patch(struct microcode_patch *new)
{
- struct microcode_patch *old;
+ struct microcode_patch *old, *free = NULL;
+ enum microcode_match_result result = MIS_UCODE;
+ unsigned long flags;
- ASSERT(spin_is_locked(µcode_mutex));
+ write_lock_irqsave(&cache_rwlock, flags);
list_for_each_entry(old, µcode_cache, list)
{
- enum microcode_match_result result =
- microcode_ops->compare_patch(new, old);
+ result = microcode_ops->compare_patch(new, old);
if ( result == OLD_UCODE )
{
- microcode_ops->free_patch(new);
- return false;
+ free = new;
+ break;
}
else if ( result == NEW_UCODE )
{
list_replace(&old->list, &new->list);
- microcode_ops->free_patch(old);
- return true;
+ free = old;
+ break;
}
else /* result == MIS_UCODE */
continue;
}
- if ( microcode_ops->is_patch_compatible &&
- !microcode_ops->is_patch_compatible(new) )
+ if ( result == MIS_UCODE )
{
- xfree(new);
- return false;
+ if ( microcode_ops->is_patch_compatible &&
+ !microcode_ops->is_patch_compatible(new) )
+ free = new;
+ else
+ list_add_tail(&new->list, µcode_cache);
}
- list_add_tail(&new->list, µcode_cache);
+ write_unlock_irqrestore(&cache_rwlock, flags);
+
+ /* free useless patches after interrupt enabled */
+ if ( free )
+ microcode_ops->free_patch(free);
- return true;
+ return free != new;
}
/*
@@ -269,8 +278,6 @@ const struct microcode_patch *microcode_find_patch(void)
{
const struct microcode_patch *microcode_patch;
- ASSERT(spin_is_locked(µcode_mutex));
-
list_for_each_entry(microcode_patch, µcode_cache, list)
{
if ( microcode_ops->match_cpu(microcode_patch) )
@@ -288,11 +295,9 @@ static int microcode_parse_blob(char *buf, uint32_t len)
{
int ret;
- spin_lock(µcode_mutex);
ret = microcode_ops->collect_cpu_info(&this_cpu(cpu_sig));
if ( likely(!ret) )
ret = microcode_ops->cpu_request_microcode(buf, len);
- spin_unlock(µcode_mutex);
return ret;
}
@@ -300,12 +305,15 @@ static int microcode_parse_blob(char *buf, uint32_t len)
static int microcode_update_cpu(void)
{
int ret;
+ unsigned long flags;
- spin_lock(µcode_mutex);
ret = microcode_ops->collect_cpu_info(&this_cpu(cpu_sig));
- if ( likely(!ret) )
- ret = microcode_ops->apply_microcode();
- spin_unlock(µcode_mutex);
+ if ( unlikely(ret) )
+ return ret;
+
+ read_lock_irqsave(&cache_rwlock, flags);
+ ret = microcode_ops->apply_microcode();
+ read_unlock_irqrestore(&cache_rwlock, flags);
return ret;
}
@@ -493,7 +501,16 @@ __initcall(microcode_init);
int early_microcode_update_cpu(void)
{
- return microcode_ops ? microcode_update_cpu() : 0;
+ int rc;
+
+ if ( !microcode_ops )
+ return 0;
+
+ spin_lock(&early_ucode_update_lock);
+ rc = microcode_update_cpu();
+ spin_unlock(&early_ucode_update_lock);
+
+ return rc;
}
/*
diff --git a/xen/arch/x86/microcode_amd.c b/xen/arch/x86/microcode_amd.c
index 5c25ff2..4aa8fdc 100644
--- a/xen/arch/x86/microcode_amd.c
+++ b/xen/arch/x86/microcode_amd.c
@@ -74,9 +74,6 @@ struct mpbhdr {
uint8_t data[];
};
-/* serialize access to the physical write */
-static DEFINE_SPINLOCK(microcode_update_lock);
-
/* See comment in start_update() for cases when this routine fails */
static int collect_cpu_info(struct cpu_signature *csig)
{
@@ -251,7 +248,6 @@ static enum microcode_match_result compare_patch(
static int apply_microcode(void)
{
- unsigned long flags;
uint32_t rev;
const struct microcode_header_amd *hdr;
const struct microcode_patch *patch;
@@ -265,15 +261,13 @@ static int apply_microcode(void)
hdr = patch->data;
- spin_lock_irqsave(µcode_update_lock, flags);
+ BUG_ON(local_irq_is_enabled());
hw_err = wrmsr_safe(MSR_AMD_PATCHLOADER, (unsigned long)hdr);
/* get patch id after patching */
rdmsrl(MSR_AMD_PATCHLEVEL, rev);
- spin_unlock_irqrestore(µcode_update_lock, flags);
-
/*
* Some processors leave the ucode blob mapping as UC after the update.
* Flush the mapping to regain normal cacheability.
diff --git a/xen/arch/x86/microcode_intel.c b/xen/arch/x86/microcode_intel.c
index d5ef145..56f3956 100644
--- a/xen/arch/x86/microcode_intel.c
+++ b/xen/arch/x86/microcode_intel.c
@@ -95,9 +95,6 @@ struct extended_sigtable {
#define STEPPING_MASK ~0xfU
-/* serialize access to the physical write to MSR 0x79 */
-static DEFINE_SPINLOCK(microcode_update_lock);
-
static int collect_cpu_info(struct cpu_signature *csig)
{
unsigned int cpu_num = smp_processor_id();
@@ -334,7 +331,6 @@ static int get_matching_microcode(const void *mc, unsigned int cpu)
static int apply_microcode(void)
{
- unsigned long flags;
uint64_t msr_content;
unsigned int val[2];
unsigned int cpu_num = raw_smp_processor_id();
@@ -347,9 +343,7 @@ static int apply_microcode(void)
return -EINVAL;
mc_intel = patch->data;
-
- /* serialize access to the physical write to MSR 0x79 */
- spin_lock_irqsave(µcode_update_lock, flags);
+ BUG_ON(local_irq_is_enabled());
/*
* Writeback and invalidate caches before updating microcode to avoid
@@ -368,7 +362,6 @@ static int apply_microcode(void)
rdmsrl(MSR_IA32_UCODE_REV, msr_content);
val[1] = (uint32_t)(msr_content >> 32);
- spin_unlock_irqrestore(µcode_update_lock, flags);
if ( val[1] != mc_intel->hdr.rev )
{
printk(KERN_ERR "microcode: CPU%d update from revision "
--
1.8.3.1
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
next prev parent reply other threads:[~2019-03-11 7:54 UTC|newest]
Thread overview: 50+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-03-11 7:57 [PATCH v6 00/12] improve late microcode loading Chao Gao
2019-03-11 7:57 ` [PATCH v6 01/12] misc/xenmicrocode: Upload a microcode blob to the hypervisor Chao Gao
2019-03-12 15:27 ` Roger Pau Monné
2019-03-13 5:05 ` Chao Gao
2019-03-13 9:24 ` Wei Liu
2019-03-25 9:38 ` Sergey Dyasli
2019-04-02 2:26 ` Chao Gao
2019-03-11 7:57 ` [PATCH v6 02/12] microcode/intel: use union to get fields without shifting and masking Chao Gao
2019-03-12 15:33 ` Roger Pau Monné
2019-03-12 16:43 ` Jan Beulich
2019-03-12 18:23 ` Wei Liu
2019-03-11 7:57 ` [PATCH v6 03/12] microcode/intel: extend microcode_update_match() Chao Gao
2019-03-11 7:57 ` [PATCH v6 04/12] microcode: introduce a global cache of ucode patch Chao Gao
2019-03-12 16:53 ` Roger Pau Monné
2019-03-12 23:31 ` Raj, Ashok
2019-03-13 5:28 ` Chao Gao
2019-03-13 7:39 ` Jan Beulich
2019-03-13 10:30 ` Andrew Cooper
2019-03-13 17:04 ` Andrew Cooper
2019-03-14 7:42 ` Jan Beulich
2019-03-13 16:36 ` Sergey Dyasli
2019-03-14 1:39 ` Chao Gao
2019-03-11 7:57 ` [PATCH v6 05/12] microcode: only save compatible ucode patches Chao Gao
2019-03-12 17:03 ` Roger Pau Monné
2019-03-13 7:45 ` Jan Beulich
2019-03-11 7:57 ` [PATCH v6 06/12] microcode: remove struct ucode_cpu_info Chao Gao
2019-03-11 7:57 ` [PATCH v6 07/12] microcode: remove pointless 'cpu' parameter Chao Gao
2019-03-11 7:57 ` [PATCH v6 08/12] microcode: split out apply_microcode() from cpu_request_microcode() Chao Gao
2019-03-11 7:57 ` [PATCH v6 09/12] microcode: remove struct microcode_info Chao Gao
2019-03-11 7:57 ` [PATCH v6 10/12] microcode/intel: Writeback and invalidate caches before updating microcode Chao Gao
2019-03-21 11:08 ` Sergey Dyasli
2019-03-11 7:57 ` [PATCH v6 11/12] x86/microcode: Synchronize late microcode loading Chao Gao
2019-03-13 0:07 ` Raj, Ashok
2019-03-13 5:02 ` Chao Gao
2019-03-13 7:54 ` Jan Beulich
2019-03-13 8:02 ` Jan Beulich
2019-03-14 12:39 ` Andrew Cooper
2019-03-14 18:57 ` Raj, Ashok
2019-03-14 20:25 ` Thomas Gleixner
2019-03-15 9:40 ` Andrew Cooper
2019-03-15 10:44 ` Thomas Gleixner
2019-03-14 13:01 ` Chao Gao
2019-03-14 13:08 ` Jan Beulich
2019-03-11 7:57 ` Chao Gao [this message]
2019-03-21 12:24 ` [RFC PATCH v6 13/12] microcode: add sequential application policy Sergey Dyasli
2019-03-21 14:25 ` Chao Gao
2019-03-26 16:23 ` Jan Beulich
2019-03-19 20:22 ` [PATCH v6 00/12] improve late microcode loading Woods, Brian
2019-03-19 21:39 ` Woods, Brian
2019-03-20 8:58 ` Chao Gao
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=1552291056-20286-13-git-send-email-chao.gao@intel.com \
--to=chao.gao@intel.com \
--cc=andrew.cooper3@citrix.com \
--cc=ashok.raj@intel.com \
--cc=jbeulich@suse.com \
--cc=roger.pau@citrix.com \
--cc=sergey.dyasli@citrix.com \
--cc=wei.liu2@citrix.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).