From: David Carrillo-Cisneros <davidcc@google.com>
To: Fenghua Yu <fenghua.yu@intel.com>
Cc: Stephane Eranian <eranian@google.com>,
Paul Turner <pjt@google.com>,
linux-kernel@vger.kernel.org,
David Carrillo-Cisneros <davidcc@google.com>
Subject: [PATCH 2/2] x86/intel_rdt: Fix bug in initialization, locks and write cbm mask.
Date: Wed, 4 Nov 2015 18:19:29 -0800 [thread overview]
Message-ID: <1446689969-67703-2-git-send-email-davidcc@google.com> (raw)
In-Reply-To: <1446689969-67703-1-git-send-email-davidcc@google.com>
Fix bugs in patch series "x86:Intel Cache Allocation Technology Support"
patches by Fenghua Yu. Changes are:
1) Instruct task_css_check not to print a warning for
unnecesary lockdeps when calling from __rdt_intel_sched_in
since all callers are already synchronized by task_rq_lock().
2) Add missing mutex_locks surrounding accesses to clos_cbm_table.
3) Properly initialize online cpus in intel_rdt_late_init by using
intel_rdt_cpu_start() instead of rdt_cpumask_update().
4) Make cbm_validate_rdt_cgroup to actually use the children's mask
when validating children's masks (as it should).
Signed-off-by: David Carrillo-Cisneros <davidcc@google.com>
---
arch/x86/include/asm/intel_rdt.h | 12 +++++++++---
arch/x86/kernel/cpu/intel_rdt.c | 24 ++++++++++++++++++------
2 files changed, 27 insertions(+), 9 deletions(-)
diff --git a/arch/x86/include/asm/intel_rdt.h b/arch/x86/include/asm/intel_rdt.h
index fbe1e00..f487a93 100644
--- a/arch/x86/include/asm/intel_rdt.h
+++ b/arch/x86/include/asm/intel_rdt.h
@@ -37,11 +37,17 @@ static inline struct intel_rdt *parent_rdt(struct intel_rdt *ir)
}
/*
- * Return rdt group to which this task belongs.
+ * Return rdt group to which this task belongs without checking for lockdep.
*/
-static inline struct intel_rdt *task_rdt(struct task_struct *task)
+static inline struct intel_rdt *task_rdt_nocheck(struct task_struct *task)
{
- return css_rdt(task_css(task, intel_rdt_cgrp_id));
+ /*
+ * The checks for lockdep performed by task_subsys_state are not
+ * necessary when callers are properly synchronized by other locks.
+ * If the caller for this function is not properly synchronized
+ * use task_css instead.
+ */
+ return css_rdt(task_css_check(task, intel_rdt_cgrp_id, true));
}
/*
diff --git a/arch/x86/kernel/cpu/intel_rdt.c b/arch/x86/kernel/cpu/intel_rdt.c
index cb4d2ef..d5fa76f 100644
--- a/arch/x86/kernel/cpu/intel_rdt.c
+++ b/arch/x86/kernel/cpu/intel_rdt.c
@@ -115,7 +115,13 @@ static inline bool cache_alloc_supported(struct cpuinfo_x86 *c)
void __intel_rdt_sched_in(void *dummy)
{
struct intel_pqr_state *state = this_cpu_ptr(&pqr_state);
- struct intel_rdt *ir = task_rdt(current);
+
+ /*
+ * All callers are synchronized by task_rq_lock(); we do not use RCU
+ * which is pointless here. Thus, we call task_rdt_nocheck that avoids
+ * the lockdep checks.
+ */
+ struct intel_rdt *ir = task_rdt_nocheck(current);
if (ir->closid == state->closid)
return;
@@ -403,7 +409,9 @@ static int intel_cache_alloc_cbm_read(struct seq_file *m, void *v)
struct intel_rdt *ir = css_rdt(seq_css(m));
unsigned long l3_cbm = 0;
+ mutex_lock(&rdt_group_mutex);
clos_cbm_table_read(ir->closid, &l3_cbm);
+ mutex_unlock(&rdt_group_mutex);
seq_printf(m, "%08lx\n", l3_cbm);
return 0;
@@ -431,7 +439,7 @@ static int cbm_validate_rdt_cgroup(struct intel_rdt *ir, unsigned long cbmvalue)
rcu_read_lock();
rdt_for_each_child(css, ir) {
c = css_rdt(css);
- clos_cbm_table_read(par->closid, &cbm_tmp);
+ clos_cbm_table_read(c->closid, &cbm_tmp);
if (!bitmap_subset(&cbm_tmp, &cbmvalue, MAX_CBM_LENGTH)) {
rcu_read_unlock();
err = -EINVAL;
@@ -504,7 +512,6 @@ static int intel_cache_alloc_cbm_write(struct cgroup_subsys_state *css,
closcbm_map_dump();
out:
mutex_unlock(&rdt_group_mutex);
-
return err;
}
@@ -513,12 +520,16 @@ static void rdt_cgroup_init(void)
int max_cbm_len = boot_cpu_data.x86_cache_max_cbm_len;
u32 closid;
+ mutex_lock(&rdt_group_mutex);
+
closid_alloc(&closid);
WARN_ON(closid != 0);
rdt_root_group.closid = closid;
clos_cbm_table_update(closid, (1ULL << max_cbm_len) - 1);
+
+ mutex_unlock(&rdt_group_mutex);
}
static int __init intel_rdt_late_init(void)
@@ -552,15 +563,16 @@ static int __init intel_rdt_late_init(void)
cpu_notifier_register_begin();
for_each_online_cpu(i)
- rdt_cpumask_update(i);
-
+ intel_rdt_cpu_start(i);
__hotcpu_notifier(intel_rdt_cpu_notifier, 0);
cpu_notifier_register_done();
+
rdt_cgroup_init();
static_key_slow_inc(&rdt_enable_key);
- pr_info("Intel cache allocation enabled\n");
+ pr_info("Intel cache allocation enabled\n"
+ "max_closid:%u, max_cbm_len:%u\n", maxid, max_cbm_len);
out_err:
return err;
--
2.6.0.rc2.230.g3dd15c0
prev parent reply other threads:[~2015-11-05 2:20 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-10-02 6:09 [PATCH V15 00/11] x86: Intel Cache Allocation Technology Support Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 01/11] x86/intel_cqm: Modify hot cpu notification handling Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 02/11] x86/intel_rapl: " Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 03/11] x86/intel_rdt: Cache Allocation documentation Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 04/11] x86/intel_rdt: Add support for Cache Allocation detection Fenghua Yu
2015-11-04 14:51 ` Luiz Capitulino
2015-10-02 6:09 ` [PATCH V15 05/11] x86/intel_rdt: Add Class of service management Fenghua Yu
2015-11-04 14:55 ` Luiz Capitulino
2015-10-02 6:09 ` [PATCH V15 06/11] x86/intel_rdt: Add L3 cache capacity bitmask management Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 07/11] x86/intel_rdt: Implement scheduling support for Intel RDT Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 08/11] x86/intel_rdt: Hot cpu support for Cache Allocation Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 09/11] x86/intel_rdt: Intel haswell Cache Allocation enumeration Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 10/11] x86,cgroup/intel_rdt : Add intel_rdt cgroup documentation Fenghua Yu
2015-10-02 6:09 ` [PATCH V15 11/11] x86,cgroup/intel_rdt : Add a cgroup interface to manage Intel cache allocation Fenghua Yu
2015-11-18 20:58 ` Marcelo Tosatti
2015-11-18 21:27 ` Marcelo Tosatti
2015-12-16 22:00 ` Yu, Fenghua
2015-11-18 22:15 ` Marcelo Tosatti
2015-12-14 22:58 ` Yu, Fenghua
2015-10-11 19:50 ` [PATCH V15 00/11] x86: Intel Cache Allocation Technology Support Thomas Gleixner
2015-10-12 18:52 ` Yu, Fenghua
2015-10-12 19:58 ` Thomas Gleixner
2015-10-13 22:40 ` Marcelo Tosatti
2015-10-15 11:37 ` Peter Zijlstra
2015-10-16 0:17 ` Marcelo Tosatti
2015-10-16 9:44 ` Peter Zijlstra
2015-10-16 20:24 ` Marcelo Tosatti
2015-10-19 23:49 ` Marcelo Tosatti
2015-10-13 21:31 ` Marcelo Tosatti
2015-10-15 11:36 ` Peter Zijlstra
2015-10-16 2:28 ` Marcelo Tosatti
2015-10-16 9:50 ` Peter Zijlstra
2015-10-26 20:02 ` Marcelo Tosatti
2015-11-02 22:20 ` cat cgroup interface proposal (non hierarchical) was " Marcelo Tosatti
2015-11-04 14:42 ` Luiz Capitulino
2015-11-04 14:57 ` Thomas Gleixner
2015-11-04 15:12 ` Luiz Capitulino
2015-11-04 15:28 ` Thomas Gleixner
2015-11-04 15:35 ` Luiz Capitulino
2015-11-04 15:50 ` Thomas Gleixner
2015-11-05 2:19 ` [PATCH 1/2] x86/intel_rdt,intel_cqm: Remove build dependency of RDT code on CQM code David Carrillo-Cisneros
2015-11-05 2:19 ` David Carrillo-Cisneros [this message]
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=1446689969-67703-2-git-send-email-davidcc@google.com \
--to=davidcc@google.com \
--cc=eranian@google.com \
--cc=fenghua.yu@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=pjt@google.com \
/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).