* [PATCH v2 1/2] Documentation: x86: Rename resctrl_ui.rst and add two errata to the file
2020-10-13 2:46 [PATCH v2 0/2] x86/resctrl: Workaround MBM errata Fenghua Yu
@ 2020-10-13 2:46 ` Fenghua Yu
2020-10-13 2:54 ` Randy Dunlap
2020-10-13 2:46 ` [PATCH v2 2/2] x86/resctrl: Correct MBM total and local values Fenghua Yu
1 sibling, 1 reply; 5+ messages in thread
From: Fenghua Yu @ 2020-10-13 2:46 UTC (permalink / raw)
To: Borislav Petkov, Thomas Gleixner, Ingo Molnar, Stephane Eranian,
Tony Luck, Reinette Chatre, Ravi V Shankar
Cc: linux-kernel, x86, Fenghua Yu
Intel Memory Bandwidth Monitoring (MBM) counters may report system
memory bandwidth incorrectly on some Intel processors. The errata are
reported in erratum SKX99 [1], erratum BDF102 [2] and RDT reference
manual [3].
To work around the errata, MBM total and local readings are corrected
using a correction factor table.
Since the correction factor table is not publicly documented anywhere,
the table and the errata are documented in Documentation/x86/resctrl.rst
for future reference. The resctrl.rst file is renamed from
Documentation/x86/resctrl_ui.rst because the file won't contain user
interface only anymore.
1. Erratum SKX99 in Intel Xeon Processor Scalable Family Specification
Update:
http://web.archive.org/web/20200716124958/https://www.intel.com/content/www/us/en/processors/xeon/scalable/xeon-scalable-spec-update.html
2. Erratum BDF102 in Intel Xeon E5-2600 v4 Processor Product Family
Specification Update:
http://web.archive.org/web/20191125200531/https://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/xeon-e5-v4-spec-update.pdf
3. The errata in Intel Resource Director Technology (Intel RDT) on 2nd
Generation Intel Xeon Scalable Processors Reference Manual:
https://software.intel.com/content/www/us/en/develop/articles/intel-resource-director-technology-rdt-reference-manual.html
Suggested-by: Borislav Petkov <bp@alien8.de>
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
---
Change Log:
v2:
- Document the correction factor table and errata in resctrl.rst (Boris).
- Change the documentation URLs to stable archive.org (Tony).
Documentation/conf.py | 2 +-
Documentation/x86/index.rst | 2 +-
.../x86/{resctrl_ui.rst => resctrl.rst} | 82 +++++++++++++++++++
3 files changed, 84 insertions(+), 2 deletions(-)
rename Documentation/x86/{resctrl_ui.rst => resctrl.rst} (92%)
diff --git a/Documentation/conf.py b/Documentation/conf.py
index c503188880d9..b5b2be8eec22 100644
--- a/Documentation/conf.py
+++ b/Documentation/conf.py
@@ -36,7 +36,7 @@ needs_sphinx = '1.3'
# Add any Sphinx extension module names here, as strings. They can be
# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
# ones.
-extensions = ['kerneldoc', 'rstFlatTable', 'kernel_include', 'cdomain',
+extensions = ['kerneldoc', 'rstFlatTable', 'kernel_include',
'kfigure', 'sphinx.ext.ifconfig', 'automarkup',
'maintainers_include', 'sphinx.ext.autosectionlabel' ]
diff --git a/Documentation/x86/index.rst b/Documentation/x86/index.rst
index 265d9e9a093b..49d2fd9f0e5b 100644
--- a/Documentation/x86/index.rst
+++ b/Documentation/x86/index.rst
@@ -25,7 +25,7 @@ x86-specific Documentation
pti
mds
microcode
- resctrl_ui
+ resctrl
tsx_async_abort
usb-legacy-support
i386/index
diff --git a/Documentation/x86/resctrl_ui.rst b/Documentation/x86/resctrl.rst
similarity index 92%
rename from Documentation/x86/resctrl_ui.rst
rename to Documentation/x86/resctrl.rst
index e59b7b93a9b4..8b8ca6de5e1f 100644
--- a/Documentation/x86/resctrl_ui.rst
+++ b/Documentation/x86/resctrl.rst
@@ -1209,3 +1209,85 @@ View the llc occupancy snapshot::
# cat /sys/fs/resctrl/p1/mon_data/mon_L3_00/llc_occupancy
11234000
+
+Intel RDT Errata
+================
+
+Intel MBM Counters May Report System Memory Bandwidth Incorrectly
+-----------------------------------------------------------------
+
+Errata SKX99 for Skylake server and BDF102 for Broadwell server.
+
+Problem: Intel Memory Bandwidth Monitoring (MBM) counters track metrics
+according to the assigned Resource Monitor ID (RMID) for that logical core.
+The IA32_QM_CTR register(MSR 0xC8E), used to report these metrics, may
+report incorrect system bandwidth for certain RMID values.
+
+Implication: Due to the errata, system memory bandwidth may not match
+what is reported.
+
+Workaround: The kernel works around the errata.
+
+MBM total and local readings are corrected by the following correction
+factor table for the errata:
+
++---------------+---------------+---------------+-----------------+
+|core count |rmid count |rmid threshold |correction factor|
++---------------+---------------+---------------+-----------------+
+|1 |8 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|2 |16 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|3 |24 |15 |0.969650 |
++---------------+---------------+---------------+-----------------+
+|4 |32 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|6 |48 |31 |0.969650 |
++---------------+---------------+---------------+-----------------+
+|7 |56 |47 |1.142857 |
++---------------+---------------+---------------+-----------------+
+|8 |64 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|9 |72 |63 |1.185115 |
++---------------+---------------+---------------+-----------------+
+|10 |80 |63 |1.066553 |
++---------------+---------------+---------------+-----------------+
+|11 |88 |79 |1.454545 |
++---------------+---------------+---------------+-----------------+
+|12 |96 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|13 |104 |95 |1.230769 |
++---------------+---------------+---------------+-----------------+
+|14 |112 |95 |1.142857 |
++---------------+---------------+---------------+-----------------+
+|15 |120 |95 |1.066667 |
++---------------+---------------+---------------+-----------------+
+|16 |128 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|17 |136 |127 |1.254863 |
++---------------+---------------+---------------+-----------------+
+|18 |144 |127 |1.185255 |
++---------------+---------------+---------------+-----------------+
+|19 |152 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|20 |160 |127 |1.066667 |
++---------------+---------------+---------------+-----------------+
+|21 |168 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|22 |176 |159 |1.454334 |
++---------------+---------------+---------------+-----------------+
+|23 |184 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|24 |192 |127 |0.969744 |
++---------------+---------------+---------------+-----------------+
+|25 |200 |191 |1.280246 |
++---------------+---------------+---------------+-----------------+
+|26 |208 |191 |1.230921 |
++---------------+---------------+---------------+-----------------+
+|27 |216 |0 |1.000000 |
++---------------+---------------+---------------+-----------------+
+|28 |224 |191 |1.143118 |
++---------------+---------------+---------------+-----------------+
+
+If rmid > rmid threshold, MBM total and local values should be multiplied
+by the correction factor.
--
2.28.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH v2 2/2] x86/resctrl: Correct MBM total and local values
2020-10-13 2:46 [PATCH v2 0/2] x86/resctrl: Workaround MBM errata Fenghua Yu
2020-10-13 2:46 ` [PATCH v2 1/2] Documentation: x86: Rename resctrl_ui.rst and add two errata to the file Fenghua Yu
@ 2020-10-13 2:46 ` Fenghua Yu
1 sibling, 0 replies; 5+ messages in thread
From: Fenghua Yu @ 2020-10-13 2:46 UTC (permalink / raw)
To: Borislav Petkov, Thomas Gleixner, Ingo Molnar, Stephane Eranian,
Tony Luck, Reinette Chatre, Ravi V Shankar
Cc: linux-kernel, x86, Fenghua Yu
Intel Memory Bandwidth Monitoring (MBM) counters may report system
memory bandwidth incorrectly on some Intel processors. The errata
SKX99 for Skylake server, BDF102 for Broadwell server, and the
correction factor table are documented in Documentation/x86/resctrl.rst.
Intel MBM counters track metrics according to the assigned Resource
Monitor ID (RMID) for that logical core. The IA32_QM_CTR register
(MSR 0xC8E), used to report these metrics, may report incorrect system
bandwidth for certain RMID values.
Due to the errata, system memory bandwidth may not match what is reported.
To work around the errata, MBM total and local readings are corrected
using the correction factor table. If rmid > rmid threshold, MBM total
and local values should be multiplied by the correction factor.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
---
Change Log:
v2:
- Add "const" for mbm_cf_table[] (Boris).
- Add verbs to corrected_mbm_counter() and intel_rdt_mbm_quirk() (Boris).
- Describe the errata in the commit message (Boris).
- Fix checkpatch.pl warning on "/* FALLTHROUGH */" (Boris).
arch/x86/kernel/cpu/resctrl/core.c | 4 ++
arch/x86/kernel/cpu/resctrl/internal.h | 1 +
arch/x86/kernel/cpu/resctrl/monitor.c | 83 +++++++++++++++++++++++++-
3 files changed, 86 insertions(+), 2 deletions(-)
diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
index 9e1712e8aef7..b3bdc477599e 100644
--- a/arch/x86/kernel/cpu/resctrl/core.c
+++ b/arch/x86/kernel/cpu/resctrl/core.c
@@ -893,6 +893,10 @@ static __init void __check_quirks_intel(void)
set_rdt_options("!cmt,!mbmtotal,!mbmlocal,!l3cat");
else
set_rdt_options("!l3cat");
+ fallthrough;
+ case INTEL_FAM6_BROADWELL_X:
+ intel_rdt_mbm_apply_quirk();
+ break;
}
}
diff --git a/arch/x86/kernel/cpu/resctrl/internal.h b/arch/x86/kernel/cpu/resctrl/internal.h
index 80fa997fae60..35ececc0b48b 100644
--- a/arch/x86/kernel/cpu/resctrl/internal.h
+++ b/arch/x86/kernel/cpu/resctrl/internal.h
@@ -616,6 +616,7 @@ void mon_event_read(struct rmid_read *rr, struct rdt_resource *r,
void mbm_setup_overflow_handler(struct rdt_domain *dom,
unsigned long delay_ms);
void mbm_handle_overflow(struct work_struct *work);
+void intel_rdt_mbm_apply_quirk(void);
bool is_mba_sc(struct rdt_resource *r);
void setup_default_ctrlval(struct rdt_resource *r, u32 *dc, u32 *dm);
u32 delay_bw_map(unsigned long bw, struct rdt_resource *r);
diff --git a/arch/x86/kernel/cpu/resctrl/monitor.c b/arch/x86/kernel/cpu/resctrl/monitor.c
index 54dffe574e67..88e9bf96a9db 100644
--- a/arch/x86/kernel/cpu/resctrl/monitor.c
+++ b/arch/x86/kernel/cpu/resctrl/monitor.c
@@ -18,6 +18,7 @@
#include <linux/module.h>
#include <linux/slab.h>
#include <asm/cpu_device_id.h>
+#include <asm/intel-family.h>
#include "internal.h"
struct rmid_entry {
@@ -64,6 +65,69 @@ unsigned int rdt_mon_features;
*/
unsigned int resctrl_cqm_threshold;
+#define CF(cf) ((unsigned long)(1048576 * (cf) + 0.5))
+
+/*
+ * The correction factor table is documented in Documentation/x86/resctrl.rst.
+ * If rmid > rmid threshold, MBM total and local values should be multiplied
+ * by the correction factor.
+ *
+ * The original table is modified for better code:
+ *
+ * 1. The threshold 0 is changed to rmid count - 1 so don't do correction
+ * for the case.
+ * 2. MBM total and local correction table indexed by core counter which is
+ * equal to (x86_cache_max_rmid + 1) / 8 - 1 and is from 0 up to 27.
+ * 3. The correction factor is normalized to 2^20 (1048576) so it's faster
+ * to calculate corrected value by shifting:
+ * corrected_value = (original_value * correction_factor) >> 20
+ */
+static const struct mbm_correction_factor_table {
+ u32 rmidthreshold;
+ u64 cf;
+} mbm_cf_table[] = {
+ {7, CF(1.000000)},
+ {15, CF(1.000000)},
+ {15, CF(0.969650)},
+ {31, CF(1.000000)},
+ {31, CF(1.066667)},
+ {31, CF(0.969650)},
+ {47, CF(1.142857)},
+ {63, CF(1.000000)},
+ {63, CF(1.185115)},
+ {63, CF(1.066553)},
+ {79, CF(1.454545)},
+ {95, CF(1.000000)},
+ {95, CF(1.230769)},
+ {95, CF(1.142857)},
+ {95, CF(1.066667)},
+ {127, CF(1.000000)},
+ {127, CF(1.254863)},
+ {127, CF(1.185255)},
+ {151, CF(1.000000)},
+ {127, CF(1.066667)},
+ {167, CF(1.000000)},
+ {159, CF(1.454334)},
+ {183, CF(1.000000)},
+ {127, CF(0.969744)},
+ {191, CF(1.280246)},
+ {191, CF(1.230921)},
+ {215, CF(1.000000)},
+ {191, CF(1.143118)},
+};
+
+static u32 mbm_cf_rmidthreshold = UINT_MAX;
+static u64 mbm_cf;
+
+static inline u64 get_corrected_mbm_count(u32 rmid, unsigned long val)
+{
+ /* Correct MBM value. */
+ if (rmid > mbm_cf_rmidthreshold)
+ val = (val * mbm_cf) >> 20;
+
+ return val;
+}
+
static inline struct rmid_entry *__rmid_entry(u32 rmid)
{
struct rmid_entry *entry;
@@ -260,7 +324,8 @@ static int __mon_event_count(u32 rmid, struct rmid_read *rr)
m->chunks += chunks;
m->prev_msr = tval;
- rr->val += m->chunks;
+ rr->val += get_corrected_mbm_count(rmid, m->chunks);
+
return 0;
}
@@ -280,7 +345,7 @@ static void mbm_bw_count(u32 rmid, struct rmid_read *rr)
chunks = mbm_overflow_count(m->prev_bw_msr, tval, rr->r->mbm_width);
m->chunks += chunks;
- cur_bw = (chunks * r->mon_scale) >> 20;
+ cur_bw = (get_corrected_mbm_count(rmid, chunks) * r->mon_scale) >> 20;
if (m->delta_comp)
m->delta_bw = abs(cur_bw - m->prev_bw);
@@ -644,3 +709,17 @@ int rdt_get_mon_l3_config(struct rdt_resource *r)
return 0;
}
+
+void intel_rdt_mbm_apply_quirk(void)
+{
+ int cf_index;
+
+ cf_index = (boot_cpu_data.x86_cache_max_rmid + 1) / 8 - 1;
+ if (cf_index >= ARRAY_SIZE(mbm_cf_table)) {
+ pr_info("No MBM correction factor available\n");
+ return;
+ }
+
+ mbm_cf_rmidthreshold = mbm_cf_table[cf_index].rmidthreshold;
+ mbm_cf = mbm_cf_table[cf_index].cf;
+}
--
2.28.0
^ permalink raw reply related [flat|nested] 5+ messages in thread