mirror of
https://github.com/Fishwaldo/Star64_linux.git
synced 2025-04-30 12:23:57 +00:00
616 commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
Giovanni Gherdovich
|
1567c3e346 |
x86, sched: Add support for frequency invariance
Implement arch_scale_freq_capacity() for 'modern' x86. This function
is used by the scheduler to correctly account usage in the face of
DVFS.
The present patch addresses Intel processors specifically and has positive
performance and performance-per-watt implications for the schedutil cpufreq
governor, bringing it closer to, if not on-par with, the powersave governor
from the intel_pstate driver/framework.
Large performance gains are obtained when the machine is lightly loaded and
no regression are observed at saturation. The benchmarks with the largest
gains are kernel compilation, tbench (the networking version of dbench) and
shell-intensive workloads.
1. FREQUENCY INVARIANCE: MOTIVATION
* Without it, a task looks larger if the CPU runs slower
2. PECULIARITIES OF X86
* freq invariance accounting requires knowing the ratio freq_curr/freq_max
2.1 CURRENT FREQUENCY
* Use delta_APERF / delta_MPERF * freq_base (a.k.a "BusyMHz")
2.2 MAX FREQUENCY
* It varies with time (turbo). As an approximation, we set it to a
constant, i.e. 4-cores turbo frequency.
3. EFFECTS ON THE SCHEDUTIL FREQUENCY GOVERNOR
* The invariant schedutil's formula has no feedback loop and reacts faster
to utilization changes
4. KNOWN LIMITATIONS
* In some cases tasks can't reach max util despite how hard they try
5. PERFORMANCE TESTING
5.1 MACHINES
* Skylake, Broadwell, Haswell
5.2 SETUP
* baseline Linux v5.2 w/ non-invariant schedutil. Tested freq_max = 1-2-3-4-8-12
active cores turbo w/ invariant schedutil, and intel_pstate/powersave
5.3 BENCHMARK RESULTS
5.3.1 NEUTRAL BENCHMARKS
* NAS Parallel Benchmark (HPC), hackbench
5.3.2 NON-NEUTRAL BENCHMARKS
* tbench (10-30% better), kernbench (10-15% better),
shell-intensive-scripts (30-50% better)
* no regressions
5.3.3 SELECTION OF DETAILED RESULTS
5.3.4 POWER CONSUMPTION, PERFORMANCE-PER-WATT
* dbench (5% worse on one machine), kernbench (3% worse),
tbench (5-10% better), shell-intensive-scripts (10-40% better)
6. MICROARCH'ES ADDRESSED HERE
* Xeon Core before Scalable Performance processors line (Xeon Gold/Platinum
etc have different MSRs semantic for querying turbo levels)
7. REFERENCES
* MMTests performance testing framework, github.com/gormanm/mmtests
+-------------------------------------------------------------------------+
| 1. FREQUENCY INVARIANCE: MOTIVATION
+-------------------------------------------------------------------------+
For example; suppose a CPU has two frequencies: 500 and 1000 Mhz. When
running a task that would consume 1/3rd of a CPU at 1000 MHz, it would
appear to consume 2/3rd (or 66.6%) when running at 500 MHz, giving the
false impression this CPU is almost at capacity, even though it can go
faster [*]. In a nutshell, without frequency scale-invariance tasks look
larger just because the CPU is running slower.
[*] (footnote: this assumes a linear frequency/performance relation; which
everybody knows to be false, but given realities its the best approximation
we can make.)
+-------------------------------------------------------------------------+
| 2. PECULIARITIES OF X86
+-------------------------------------------------------------------------+
Accounting for frequency changes in PELT signals requires the computation of
the ratio freq_curr / freq_max. On x86 neither of those terms is readily
available.
2.1 CURRENT FREQUENCY
====================
Since modern x86 has hardware control over the actual frequency we run
at (because amongst other things, Turbo-Mode), we cannot simply use
the frequency as requested through cpufreq.
Instead we use the APERF/MPERF MSRs to compute the effective frequency
over the recent past. Also, because reading MSRs is expensive, don't
do so every time we need the value, but amortize the cost by doing it
every tick.
2.2 MAX FREQUENCY
=================
Obtaining freq_max is also non-trivial because at any time the hardware can
provide a frequency boost to a selected subset of cores if the package has
enough power to spare (eg: Turbo Boost). This means that the maximum frequency
available to a given core changes with time.
The approach taken in this change is to arbitrarily set freq_max to a constant
value at boot. The value chosen is the "4-cores (4C) turbo frequency" on most
microarchitectures, after evaluating the following candidates:
* 1-core (1C) turbo frequency (the fastest turbo state available)
* around base frequency (a.k.a. max P-state)
* something in between, such as 4C turbo
To interpret these options, consider that this is the denominator in
freq_curr/freq_max, and that ratio will be used to scale PELT signals such as
util_avg and load_avg. A large denominator will undershoot (util_avg looks a
bit smaller than it really is), viceversa with a smaller denominator PELT
signals will tend to overshoot. Given that PELT drives frequency selection
in the schedutil governor, we will have:
freq_max set to | effect on DVFS
--------------------+------------------
1C turbo | power efficiency (lower freq choices)
base freq | performance (higher util_avg, higher freq requests)
4C turbo | a bit of both
4C turbo proves to be a good compromise in a number of benchmarks (see below).
+-------------------------------------------------------------------------+
| 3. EFFECTS ON THE SCHEDUTIL FREQUENCY GOVERNOR
+-------------------------------------------------------------------------+
Once an architecture implements a frequency scale-invariant utilization (the
PELT signal util_avg), schedutil switches its frequency selection formula from
freq_next = 1.25 * freq_curr * util [non-invariant util signal]
to
freq_next = 1.25 * freq_max * util [invariant util signal]
where, in the second formula, freq_max is set to the 1C turbo frequency (max
turbo). The advantage of the second formula, whose usage we unlock with this
patch, is that freq_next doesn't depend on the current frequency in an
iterative fashion, but can jump to any frequency in a single update. This
absence of feedback in the formula makes it quicker to react to utilization
changes and more robust against pathological instabilities.
Compare it to the update formula of intel_pstate/powersave:
freq_next = 1.25 * freq_max * Busy%
where again freq_max is 1C turbo and Busy% is the percentage of time not spent
idling (calculated with delta_MPERF / delta_TSC); essentially the same as
invariant schedutil, and largely responsible for intel_pstate/powersave good
reputation. The non-invariant schedutil formula is derived from the invariant
one by approximating util_inv with util_raw * freq_curr / freq_max, but this
has limitations.
Testing shows improved performances due to better frequency selections when
the machine is lightly loaded, and essentially no change in behaviour at
saturation / overutilization.
+-------------------------------------------------------------------------+
| 4. KNOWN LIMITATIONS
+-------------------------------------------------------------------------+
It's been shown that it is possible to create pathological scenarios where a
CPU-bound task cannot reach max utilization, if the normalizing factor
freq_max is fixed to a constant value (see [Lelli-2018]).
If freq_max is set to 4C turbo as we do here, one needs to peg at least 5
cores in a package doing some busywork, and observe that none of those task
will ever reach max util (1024) because they're all running at less than the
4C turbo frequency.
While this concern still applies, we believe the performance benefit of
frequency scale-invariant PELT signals outweights the cost of this limitation.
[Lelli-2018]
https://lore.kernel.org/lkml/20180517150418.GF22493@localhost.localdomain/
+-------------------------------------------------------------------------+
| 5. PERFORMANCE TESTING
+-------------------------------------------------------------------------+
5.1 MACHINES
============
We tested the patch on three machines, with Skylake, Broadwell and Haswell
CPUs. The details are below, together with the available turbo ratios as
reported by the appropriate MSRs.
* 8x-SKYLAKE-UMA:
Single socket E3-1240 v5, Skylake 4 cores/8 threads
Max EFFiciency, BASE frequency and available turbo levels (MHz):
EFFIC 800 |********
BASE 3500 |***********************************
4C 3700 |*************************************
3C 3800 |**************************************
2C 3900 |***************************************
1C 3900 |***************************************
* 80x-BROADWELL-NUMA:
Two sockets E5-2698 v4, 2x Broadwell 20 cores/40 threads
Max EFFiciency, BASE frequency and available turbo levels (MHz):
EFFIC 1200 |************
BASE 2200 |**********************
8C 2900 |*****************************
7C 3000 |******************************
6C 3100 |*******************************
5C 3200 |********************************
4C 3300 |*********************************
3C 3400 |**********************************
2C 3600 |************************************
1C 3600 |************************************
* 48x-HASWELL-NUMA
Two sockets E5-2670 v3, 2x Haswell 12 cores/24 threads
Max EFFiciency, BASE frequency and available turbo levels (MHz):
EFFIC 1200 |************
BASE 2300 |***********************
12C 2600 |**************************
11C 2600 |**************************
10C 2600 |**************************
9C 2600 |**************************
8C 2600 |**************************
7C 2600 |**************************
6C 2600 |**************************
5C 2700 |***************************
4C 2800 |****************************
3C 2900 |*****************************
2C 3100 |*******************************
1C 3100 |*******************************
5.2 SETUP
=========
* The baseline is Linux v5.2 with schedutil (non-invariant) and the intel_pstate
driver in passive mode.
* The rationale for choosing the various freq_max values to test have been to
try all the 1-2-3-4C turbo levels (note that 1C and 2C turbo are identical
on all machines), plus one more value closer to base_freq but still in the
turbo range (8C turbo for both 80x-BROADWELL-NUMA and 48x-HASWELL-NUMA).
* In addition we've run all tests with intel_pstate/powersave for comparison.
* The filesystem is always XFS, the userspace is openSUSE Leap 15.1.
* 8x-SKYLAKE-UMA is capable of HWP (Hardware-Managed P-States), so the runs
with active intel_pstate on this machine use that.
This gives, in terms of combinations tested on each machine:
* 8x-SKYLAKE-UMA
* Baseline: Linux v5.2, non-invariant schedutil, intel_pstate passive
* intel_pstate active + powersave + HWP
* invariant schedutil, freq_max = 1C turbo
* invariant schedutil, freq_max = 3C turbo
* invariant schedutil, freq_max = 4C turbo
* both 80x-BROADWELL-NUMA and 48x-HASWELL-NUMA
* [same as 8x-SKYLAKE-UMA, but no HWP capable]
* invariant schedutil, freq_max = 8C turbo
(which on 48x-HASWELL-NUMA is the same as 12C turbo, or "all cores turbo")
5.3 BENCHMARK RESULTS
=====================
5.3.1 NEUTRAL BENCHMARKS
------------------------
Tests that didn't show any measurable difference in performance on any of the
test machines between non-invariant schedutil and our patch are:
* NAS Parallel Benchmarks (NPB) using either MPI or openMP for IPC, any
computational kernel
* flexible I/O (FIO)
* hackbench (using threads or processes, and using pipes or sockets)
5.3.2 NON-NEUTRAL BENCHMARKS
----------------------------
What follow are summary tables where each benchmark result is given a score.
* A tilde (~) means a neutral result, i.e. no difference from baseline.
* Scores are computed with the ratio result_new / result_baseline, so a tilde
means a score of 1.00.
* The results in the score ratio are the geometric means of results running
the benchmark with different parameters (eg: for kernbench: using 1, 2, 4,
... number of processes; for pgbench: varying the number of clients, and so
on).
* The first three tables show higher-is-better kind of tests (i.e. measured in
operations/second), the subsequent three show lower-is-better kind of tests
(i.e. the workload is fixed and we measure elapsed time, think kernbench).
* "gitsource" is a name we made up for the test consisting in running the
entire unit tests suite of the Git SCM and measuring how long it takes. We
take it as a typical example of shell-intensive serialized workload.
* In the "I_PSTATE" column we have the results for intel_pstate/powersave. Other
columns show invariant schedutil for different values of freq_max. 4C turbo
is circled as it's the value we've chosen for the final implementation.
80x-BROADWELL-NUMA (comparison ratio; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
pgbench-ro 1.14 ~ ~ | 1.11 | 1.14
pgbench-rw ~ ~ ~ | ~ | ~
netperf-udp 1.06 ~ 1.06 | 1.05 | 1.07
netperf-tcp ~ 1.03 ~ | 1.01 | 1.02
tbench4 1.57 1.18 1.22 | 1.30 | 1.56
+------+
8x-SKYLAKE-UMA (comparison ratio; higher is better)
+------+
I_PSTATE/HWP 1C 3C | 4C |
pgbench-ro ~ ~ ~ | ~ |
pgbench-rw ~ ~ ~ | ~ |
netperf-udp ~ ~ ~ | ~ |
netperf-tcp ~ ~ ~ | ~ |
tbench4 1.30 1.14 1.14 | 1.16 |
+------+
48x-HASWELL-NUMA (comparison ratio; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 12C
pgbench-ro 1.15 ~ ~ | 1.06 | 1.16
pgbench-rw ~ ~ ~ | ~ | ~
netperf-udp 1.05 0.97 1.04 | 1.04 | 1.02
netperf-tcp 0.96 1.01 1.01 | 1.01 | 1.01
tbench4 1.50 1.05 1.13 | 1.13 | 1.25
+------+
In the table above we see that active intel_pstate is slightly better than our
4C-turbo patch (both in reference to the baseline non-invariant schedutil) on
read-only pgbench and much better on tbench. Both cases are notable in which
it shows that lowering our freq_max (to 8C-turbo and 12C-turbo on
80x-BROADWELL-NUMA and 48x-HASWELL-NUMA respectively) helps invariant
schedutil to get closer.
If we ignore active intel_pstate and focus on the comparison with baseline
alone, there are several instances of double-digit performance improvement.
80x-BROADWELL-NUMA (comparison ratio; lower is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
dbench4 1.23 0.95 0.95 | 0.95 | 0.95
kernbench 0.93 0.83 0.83 | 0.83 | 0.82
gitsource 0.98 0.49 0.49 | 0.49 | 0.48
+------+
8x-SKYLAKE-UMA (comparison ratio; lower is better)
+------+
I_PSTATE/HWP 1C 3C | 4C |
dbench4 ~ ~ ~ | ~ |
kernbench ~ ~ ~ | ~ |
gitsource 0.92 0.55 0.55 | 0.55 |
+------+
48x-HASWELL-NUMA (comparison ratio; lower is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
dbench4 ~ ~ ~ | ~ | ~
kernbench 0.94 0.90 0.89 | 0.90 | 0.90
gitsource 0.97 0.69 0.69 | 0.69 | 0.69
+------+
dbench is not very remarkable here, unless we notice how poorly active
intel_pstate is performing on 80x-BROADWELL-NUMA: 23% regression versus
non-invariant schedutil. We repeated that run getting consistent results. Out
of scope for the patch at hand, but deserving future investigation. Other than
that, we previously ran this campaign with Linux v5.0 and saw the patch doing
better on dbench a the time. We haven't checked closely and can only speculate
at this point.
On the NUMA boxes kernbench gets 10-15% improvements on average; we'll see in
the detailed tables that the gains concentrate on low process counts (lightly
loaded machines).
The test we call "gitsource" (running the git unit test suite, a long-running
single-threaded shell script) appears rather spectacular in this table (gains
of 30-50% depending on the machine). It is to be noted, however, that
gitsource has no adjustable parameters (such as the number of jobs in
kernbench, which we average over in order to get a single-number summary
score) and is exactly the kind of low-parallelism workload that benefits the
most from this patch. When looking at the detailed tables of kernbench or
tbench4, at low process or client counts one can see similar numbers.
5.3.3 SELECTION OF DETAILED RESULTS
-----------------------------------
Machine : 48x-HASWELL-NUMA
Benchmark : tbench4 (i.e. dbench4 over the network, actually loopback)
Varying parameter : number of clients
Unit : MB/sec (higher is better)
5.2.0 vanilla (BASELINE) 5.2.0 intel_pstate 5.2.0 1C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Hmean 1 126.73 +- 0.31% ( ) 315.91 +- 0.66% ( 149.28%) 125.03 +- 0.76% ( -1.34%)
Hmean 2 258.04 +- 0.62% ( ) 614.16 +- 0.51% ( 138.01%) 269.58 +- 1.45% ( 4.47%)
Hmean 4 514.30 +- 0.67% ( ) 1146.58 +- 0.54% ( 122.94%) 533.84 +- 1.99% ( 3.80%)
Hmean 8 1111.38 +- 2.52% ( ) 2159.78 +- 0.38% ( 94.33%) 1359.92 +- 1.56% ( 22.36%)
Hmean 16 2286.47 +- 1.36% ( ) 3338.29 +- 0.21% ( 46.00%) 2720.20 +- 0.52% ( 18.97%)
Hmean 32 4704.84 +- 0.35% ( ) 4759.03 +- 0.43% ( 1.15%) 4774.48 +- 0.30% ( 1.48%)
Hmean 64 7578.04 +- 0.27% ( ) 7533.70 +- 0.43% ( -0.59%) 7462.17 +- 0.65% ( -1.53%)
Hmean 128 6998.52 +- 0.16% ( ) 6987.59 +- 0.12% ( -0.16%) 6909.17 +- 0.14% ( -1.28%)
Hmean 192 6901.35 +- 0.25% ( ) 6913.16 +- 0.10% ( 0.17%) 6855.47 +- 0.21% ( -0.66%)
5.2.0 3C-turbo 5.2.0 4C-turbo 5.2.0 12C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Hmean 1 128.43 +- 0.28% ( 1.34%) 130.64 +- 3.81% ( 3.09%) 153.71 +- 5.89% ( 21.30%)
Hmean 2 311.70 +- 6.15% ( 20.79%) 281.66 +- 3.40% ( 9.15%) 305.08 +- 5.70% ( 18.23%)
Hmean 4 641.98 +- 2.32% ( 24.83%) 623.88 +- 5.28% ( 21.31%) 906.84 +- 4.65% ( 76.32%)
Hmean 8 1633.31 +- 1.56% ( 46.96%) 1714.16 +- 0.93% ( 54.24%) 2095.74 +- 0.47% ( 88.57%)
Hmean 16 3047.24 +- 0.42% ( 33.27%) 3155.02 +- 0.30% ( 37.99%) 3634.58 +- 0.15% ( 58.96%)
Hmean 32 4734.31 +- 0.60% ( 0.63%) 4804.38 +- 0.23% ( 2.12%) 4674.62 +- 0.27% ( -0.64%)
Hmean 64 7699.74 +- 0.35% ( 1.61%) 7499.72 +- 0.34% ( -1.03%) 7659.03 +- 0.25% ( 1.07%)
Hmean 128 6935.18 +- 0.15% ( -0.91%) 6942.54 +- 0.10% ( -0.80%) 7004.85 +- 0.12% ( 0.09%)
Hmean 192 6901.62 +- 0.12% ( 0.00%) 6856.93 +- 0.10% ( -0.64%) 6978.74 +- 0.10% ( 1.12%)
This is one of the cases where the patch still can't surpass active
intel_pstate, not even when freq_max is as low as 12C-turbo. Otherwise, gains are
visible up to 16 clients and the saturated scenario is the same as baseline.
The scores in the summary table from the previous sections are ratios of
geometric means of the results over different clients, as seen in this table.
Machine : 80x-BROADWELL-NUMA
Benchmark : kernbench (kernel compilation)
Varying parameter : number of jobs
Unit : seconds (lower is better)
5.2.0 vanilla (BASELINE) 5.2.0 intel_pstate 5.2.0 1C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 2 379.68 +- 0.06% ( ) 330.20 +- 0.43% ( 13.03%) 285.93 +- 0.07% ( 24.69%)
Amean 4 200.15 +- 0.24% ( ) 175.89 +- 0.22% ( 12.12%) 153.78 +- 0.25% ( 23.17%)
Amean 8 106.20 +- 0.31% ( ) 95.54 +- 0.23% ( 10.03%) 86.74 +- 0.10% ( 18.32%)
Amean 16 56.96 +- 1.31% ( ) 53.25 +- 1.22% ( 6.50%) 48.34 +- 1.73% ( 15.13%)
Amean 32 34.80 +- 2.46% ( ) 33.81 +- 0.77% ( 2.83%) 30.28 +- 1.59% ( 12.99%)
Amean 64 26.11 +- 1.63% ( ) 25.04 +- 1.07% ( 4.10%) 22.41 +- 2.37% ( 14.16%)
Amean 128 24.80 +- 1.36% ( ) 23.57 +- 1.23% ( 4.93%) 21.44 +- 1.37% ( 13.55%)
Amean 160 24.85 +- 0.56% ( ) 23.85 +- 1.17% ( 4.06%) 21.25 +- 1.12% ( 14.49%)
5.2.0 3C-turbo 5.2.0 4C-turbo 5.2.0 8C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 2 284.08 +- 0.13% ( 25.18%) 283.96 +- 0.51% ( 25.21%) 285.05 +- 0.21% ( 24.92%)
Amean 4 153.18 +- 0.22% ( 23.47%) 154.70 +- 1.64% ( 22.71%) 153.64 +- 0.30% ( 23.24%)
Amean 8 87.06 +- 0.28% ( 18.02%) 86.77 +- 0.46% ( 18.29%) 86.78 +- 0.22% ( 18.28%)
Amean 16 48.03 +- 0.93% ( 15.68%) 47.75 +- 1.99% ( 16.17%) 47.52 +- 1.61% ( 16.57%)
Amean 32 30.23 +- 1.20% ( 13.14%) 30.08 +- 1.67% ( 13.57%) 30.07 +- 1.67% ( 13.60%)
Amean 64 22.59 +- 2.02% ( 13.50%) 22.63 +- 0.81% ( 13.32%) 22.42 +- 0.76% ( 14.12%)
Amean 128 21.37 +- 0.67% ( 13.82%) 21.31 +- 1.15% ( 14.07%) 21.17 +- 1.93% ( 14.63%)
Amean 160 21.68 +- 0.57% ( 12.76%) 21.18 +- 1.74% ( 14.77%) 21.22 +- 1.00% ( 14.61%)
The patch outperform active intel_pstate (and baseline) by a considerable
margin; the summary table from the previous section says 4C turbo and active
intel_pstate are 0.83 and 0.93 against baseline respectively, so 4C turbo is
0.83/0.93=0.89 against intel_pstate (~10% better on average). There is no
noticeable difference with regard to the value of freq_max.
Machine : 8x-SKYLAKE-UMA
Benchmark : gitsource (time to run the git unit test suite)
Varying parameter : none
Unit : seconds (lower is better)
5.2.0 vanilla 5.2.0 intel_pstate/hwp 5.2.0 1C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 858.85 +- 1.16% ( ) 791.94 +- 0.21% ( 7.79%) 474.95 ( 44.70%)
5.2.0 3C-turbo 5.2.0 4C-turbo
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Amean 475.26 +- 0.20% ( 44.66%) 474.34 +- 0.13% ( 44.77%)
In this test, which is of interest as representing shell-intensive
(i.e. fork-intensive) serialized workloads, invariant schedutil outperforms
intel_pstate/powersave by a whopping 40% margin.
5.3.4 POWER CONSUMPTION, PERFORMANCE-PER-WATT
---------------------------------------------
The following table shows average power consumption in watt for each
benchmark. Data comes from turbostat (package average), which in turn is read
from the RAPL interface on CPUs. We know the patch affects CPU frequencies so
it's reasonable to ignore other power consumers (such as memory or I/O). Also,
we don't have a power meter available in the lab so RAPL is the best we have.
turbostat sampled average power every 10 seconds for the entire duration of
each benchmark. We took all those values and averaged them (i.e. with don't
have detail on a per-parameter granularity, only on whole benchmarks).
80x-BROADWELL-NUMA (power consumption, watts)
+--------+
BASELINE I_PSTATE 1C 3C | 4C | 8C
pgbench-ro 130.01 142.77 131.11 132.45 | 134.65 | 136.84
pgbench-rw 68.30 60.83 71.45 71.70 | 71.65 | 72.54
dbench4 90.25 59.06 101.43 99.89 | 101.10 | 102.94
netperf-udp 65.70 69.81 66.02 68.03 | 68.27 | 68.95
netperf-tcp 88.08 87.96 88.97 88.89 | 88.85 | 88.20
tbench4 142.32 176.73 153.02 163.91 | 165.58 | 176.07
kernbench 92.94 101.95 114.91 115.47 | 115.52 | 115.10
gitsource 40.92 41.87 75.14 75.20 | 75.40 | 75.70
+--------+
8x-SKYLAKE-UMA (power consumption, watts)
+--------+
BASELINE I_PSTATE/HWP 1C 3C | 4C |
pgbench-ro 46.49 46.68 46.56 46.59 | 46.52 |
pgbench-rw 29.34 31.38 30.98 31.00 | 31.00 |
dbench4 27.28 27.37 27.49 27.41 | 27.38 |
netperf-udp 22.33 22.41 22.36 22.35 | 22.36 |
netperf-tcp 27.29 27.29 27.30 27.31 | 27.33 |
tbench4 41.13 45.61 43.10 43.33 | 43.56 |
kernbench 42.56 42.63 43.01 43.01 | 43.01 |
gitsource 13.32 13.69 17.33 17.30 | 17.35 |
+--------+
48x-HASWELL-NUMA (power consumption, watts)
+--------+
BASELINE I_PSTATE 1C 3C | 4C | 12C
pgbench-ro 128.84 136.04 129.87 132.43 | 132.30 | 134.86
pgbench-rw 37.68 37.92 37.17 37.74 | 37.73 | 37.31
dbench4 28.56 28.73 28.60 28.73 | 28.70 | 28.79
netperf-udp 56.70 60.44 56.79 57.42 | 57.54 | 57.52
netperf-tcp 75.49 75.27 75.87 76.02 | 76.01 | 75.95
tbench4 115.44 139.51 119.53 123.07 | 123.97 | 130.22
kernbench 83.23 91.55 95.58 95.69 | 95.72 | 96.04
gitsource 36.79 36.99 39.99 40.34 | 40.35 | 40.23
+--------+
A lower power consumption isn't necessarily better, it depends on what is done
with that energy. Here are tables with the ratio of performance-per-watt on
each machine and benchmark. Higher is always better; a tilde (~) means a
neutral ratio (i.e. 1.00).
80x-BROADWELL-NUMA (performance-per-watt ratios; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 8C
pgbench-ro 1.04 1.06 0.94 | 1.07 | 1.08
pgbench-rw 1.10 0.97 0.96 | 0.96 | 0.97
dbench4 1.24 0.94 0.95 | 0.94 | 0.92
netperf-udp ~ 1.02 1.02 | ~ | 1.02
netperf-tcp ~ 1.02 ~ | ~ | 1.02
tbench4 1.26 1.10 1.06 | 1.12 | 1.26
kernbench 0.98 0.97 0.97 | 0.97 | 0.98
gitsource ~ 1.11 1.11 | 1.11 | 1.13
+------+
8x-SKYLAKE-UMA (performance-per-watt ratios; higher is better)
+------+
I_PSTATE/HWP 1C 3C | 4C |
pgbench-ro ~ ~ ~ | ~ |
pgbench-rw 0.95 0.97 0.96 | 0.96 |
dbench4 ~ ~ ~ | ~ |
netperf-udp ~ ~ ~ | ~ |
netperf-tcp ~ ~ ~ | ~ |
tbench4 1.17 1.09 1.08 | 1.10 |
kernbench ~ ~ ~ | ~ |
gitsource 1.06 1.40 1.40 | 1.40 |
+------+
48x-HASWELL-NUMA (performance-per-watt ratios; higher is better)
+------+
I_PSTATE 1C 3C | 4C | 12C
pgbench-ro 1.09 ~ 1.09 | 1.03 | 1.11
pgbench-rw ~ 0.86 ~ | ~ | 0.86
dbench4 ~ 1.02 1.02 | 1.02 | ~
netperf-udp ~ 0.97 1.03 | 1.02 | ~
netperf-tcp 0.96 ~ ~ | ~ | ~
tbench4 1.24 ~ 1.06 | 1.05 | 1.11
kernbench 0.97 0.97 0.98 | 0.97 | 0.96
gitsource 1.03 1.33 1.32 | 1.32 | 1.33
+------+
These results are overall pleasing: in plenty of cases we observe
performance-per-watt improvements. The few regressions (read/write pgbench and
dbench on the Broadwell machine) are of small magnitude. kernbench loses a few
percentage points (it has a 10-15% performance improvement, but apparently the
increase in power consumption is larger than that). tbench4 and gitsource, which
benefit the most from the patch, keep a positive score in this table which is
a welcome surprise; that suggests that in those particular workloads the
non-invariant schedutil (and active intel_pstate, too) makes some rather
suboptimal frequency selections.
+-------------------------------------------------------------------------+
| 6. MICROARCH'ES ADDRESSED HERE
+-------------------------------------------------------------------------+
The patch addresses Xeon Core processors that use MSR_PLATFORM_INFO and
MSR_TURBO_RATIO_LIMIT to advertise their base frequency and turbo frequencies
respectively. This excludes the recent Xeon Scalable Performance processors
line (Xeon Gold, Platinum etc) whose MSRs have to be parsed differently.
Subsequent patches will address:
* Xeon Scalable Performance processors and Atom Goldmont/Goldmont Plus
* Xeon Phi (Knights Landing, Knights Mill)
* Atom Silvermont
+-------------------------------------------------------------------------+
| 7. REFERENCES
+-------------------------------------------------------------------------+
Tests have been run with the help of the MMTests performance testing
framework, see github.com/gormanm/mmtests. The configuration file names for
the benchmark used are:
db-pgbench-timed-ro-small-xfs
db-pgbench-timed-rw-small-xfs
io-dbench4-async-xfs
network-netperf-unbound
network-tbench
scheduler-unbound
workload-kerndevel-xfs
workload-shellscripts-xfs
hpc-nas-c-class-mpi-full-xfs
hpc-nas-c-class-omp-full
All those benchmarks are generally available on the web:
pgbench: https://www.postgresql.org/docs/10/pgbench.html
netperf: https://hewlettpackard.github.io/netperf/
dbench/tbench: https://dbench.samba.org/
gitsource: git unit test suite, github.com/git/git
NAS Parallel Benchmarks: https://www.nas.nasa.gov/publications/npb.html
hackbench: https://people.redhat.com/mingo/cfs-scheduler/tools/hackbench.c
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Doug Smythies <dsmythies@telus.net>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lkml.kernel.org/r/20200122151617.531-2-ggherdovich@suse.cz
|
||
Linus Torvalds
|
c5f12fdb8b |
Merge branch 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 apic updates from Thomas Gleixner: - Cleanup the apic IPI implementation by removing duplicated code and consolidating the functions into the APIC core. - Implement a safe variant of the IPI broadcast mode. Contrary to earlier attempts this uses the core tracking of which CPUs have been brought online at least once so that a broadcast does not end up in some dead end in BIOS/SMM code when the CPU is still waiting for init. Once all CPUs have been brought up once, IPI broadcasting is enabled. Before that regular one by one IPIs are issued. - Drop the paravirt CR8 related functions as they have no user anymore - Initialize the APIC TPR to block interrupt 16-31 as they are reserved for CPU exceptions and should never be raised by any well behaving device. - Emit a warning when vector space exhaustion breaks the admin set affinity of an interrupt. - Make sure to use the NMI fallback when shutdown via reboot vector IPI fails. The original code had conditions which prevent the code path to be reached. - Annotate various APIC config variables as RO after init. [ The ipi broadcase change came in earlier through the cpu hotplug branch, but I left the explanation in the commit message since it was shared between the two different branches - Linus ] * 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (28 commits) x86/apic/vector: Warn when vector space exhaustion breaks affinity x86/apic: Annotate global config variables as "read-only after init" x86/apic/x2apic: Implement IPI shorthands support x86/apic/flat64: Remove the IPI shorthand decision logic x86/apic: Share common IPI helpers x86/apic: Remove the shorthand decision logic x86/smp: Enhance native_send_call_func_ipi() x86/smp: Move smp_function_call implementations into IPI code x86/apic: Provide and use helper for send_IPI_allbutself() x86/apic: Add static key to Control IPI shorthands x86/apic: Move no_ipi_broadcast() out of 32bit x86/apic: Add NMI_VECTOR wait to IPI shorthand x86/apic: Remove dest argument from __default_send_IPI_shortcut() x86/hotplug: Silence APIC and NMI when CPU is dead x86/cpu: Move arch_smt_update() to a neutral place x86/apic/uv: Make x2apic_extra_bits static x86/apic: Consolidate the apic local headers x86/apic: Move apic_flat_64 header into apic directory x86/apic: Move ipi header into apic directory x86/apic: Cleanup the include maze ... |
||
Thomas Gleixner
|
60dcaad573 |
x86/hotplug: Silence APIC and NMI when CPU is dead
In order to support IPI/NMI broadcasting via the shorthand mechanism side effects of shorthands need to be mitigated: Shorthand IPIs and NMIs hit all CPUs including unplugged CPUs Neither of those can be handled on unplugged CPUs for obvious reasons. It would be trivial to just fully disable the APIC via the enable bit in MSR_APICBASE. But that's not possible because clearing that bit on systems based on the 3 wire APIC bus would require a hardware reset to bring it back as the APIC would lose track of bus arbitration. On systems with FSB delivery APICBASE could be disabled, but it has to be guaranteed that no interrupt is sent to the APIC while in that state and it's not clear from the SDM whether it still responds to INIT/SIPI messages. Therefore stay on the safe side and switch the APIC into soft disabled mode so it won't deliver any regular vector to the CPU. NMIs are still propagated to the 'dead' CPUs. To mitigate that add a check for the CPU being offline on early nmi entry and if so bail. Note, this cannot use the stop/restart_nmi() magic which is used in the alternatives code. A dead CPU cannot invoke nmi_enter() or anything else due to RCU and other reasons. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/alpine.DEB.2.21.1907241723290.1791@nanos.tec.linutronix.de |
||
Pingfan Liu
|
6973210242 |
x86/realmode: Remove trampoline_status
There is no reader of trampoline_status, it's only written.
It turns out that after commit
|
||
Zhenzhong Duan
|
090d54bcbc |
Revert "x86/paravirt: Set up the virt_spin_lock_key after static keys get initialized"
This reverts commit |
||
|
Thomas Gleixner
|
7652ac9201 |
x86/asm: Move native_write_cr0/4() out of line
The pinning of sensitive CR0 and CR4 bits caused a boot crash when loading the kvm_intel module on a kernel compiled with CONFIG_PARAVIRT=n. The reason is that the static key which controls the pinning is marked RO after init. The kvm_intel module contains a CR4 write which requires to update the static key entry list. That obviously does not work when the key is in a RO section. With CONFIG_PARAVIRT enabled this does not happen because the CR4 write uses the paravirt indirection and the actual write function is built in. As the key is intended to be immutable after init, move native_write_cr0/4() out of line. While at it consolidate the update of the cr4 shadow variable and store the value right away when the pinning is initialized on a booting CPU. No point in reading it back 20 instructions later. This allows to confine the static key and the pinning variable to cpu/common and allows to mark them static. Fixes: |
||
|
Linus Torvalds
|
222a21d295 |
Merge branch 'x86-topology-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 topology updates from Ingo Molnar: "Implement multi-die topology support on Intel CPUs and expose the die topology to user-space tooling, by Len Brown, Kan Liang and Zhang Rui. These changes should have no effect on the kernel's existing understanding of topologies, i.e. there should be no behavioral impact on cache, NUMA, scheduler, perf and other topologies and overall system performance" * 'x86-topology-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: perf/x86/intel/rapl: Cosmetic rename internal variables in response to multi-die/pkg support perf/x86/intel/uncore: Cosmetic renames in response to multi-die/pkg support hwmon/coretemp: Cosmetic: Rename internal variables to zones from packages thermal/x86_pkg_temp_thermal: Cosmetic: Rename internal variables to zones from packages perf/x86/intel/cstate: Support multi-die/package perf/x86/intel/rapl: Support multi-die/package perf/x86/intel/uncore: Support multi-die/package topology: Create core_cpus and die_cpus sysfs attributes topology: Create package_cpus sysfs attribute hwmon/coretemp: Support multi-die/package powercap/intel_rapl: Update RAPL domain name and debug messages thermal/x86_pkg_temp_thermal: Support multi-die/package powercap/intel_rapl: Support multi-die/package powercap/intel_rapl: Simplify rapl_find_package() x86/topology: Define topology_logical_die_id() x86/topology: Define topology_die_id() cpu/topology: Export die_id x86/topology: Create topology_max_die_per_package() x86/topology: Add CPUID.1F multi-die/package support |
||
Kees Cook
|
873d50d58f |
x86/asm: Pin sensitive CR4 bits
Several recent exploits have used direct calls to the native_write_cr4() function to disable SMEP and SMAP before then continuing their exploits using userspace memory access. Direct calls of this form can be mitigate by pinning bits of CR4 so that they cannot be changed through a common function. This is not intended to be a general ROP protection (which would require CFI to defend against properly), but rather a way to avoid trivial direct function calling (or CFI bypasses via a matching function prototype) as seen in: https://googleprojectzero.blogspot.com/2017/05/exploiting-linux-kernel-via-packet.html (https://github.com/xairy/kernel-exploits/tree/master/CVE-2017-7308) The goals of this change: - Pin specific bits (SMEP, SMAP, and UMIP) when writing CR4. - Avoid setting the bits too early (they must become pinned only after CPU feature detection and selection has finished). - Pinning mask needs to be read-only during normal runtime. - Pinning needs to be checked after write to validate the cr4 state Using __ro_after_init on the mask is done so it can't be first disabled with a malicious write. Since these bits are global state (once established by the boot CPU and kernel boot parameters), they are safe to write to secondary CPUs before those CPUs have finished feature detection. As such, the bits are set at the first cr4 write, so that cr4 write bugs can be detected (instead of silently papered over). This uses a few bytes less storage of a location we don't have: read-only per-CPU data. A check is performed after the register write because an attack could just skip directly to the register write. Such a direct jump is possible because of how this function may be built by the compiler (especially due to the removal of frame pointers) where it doesn't add a stack frame (function exit may only be a retq without pops) which is sufficient for trivial exploitation like in the timer overwrites mentioned above). The asm argument constraints gain the "+" modifier to convince the compiler that it shouldn't make ordering assumptions about the arguments or memory, and treat them as changed. Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Dave Hansen <dave.hansen@intel.com> Cc: kernel-hardening@lists.openwall.com Link: https://lkml.kernel.org/r/20190618045503.39105-3-keescook@chromium.org |
||
|
Thomas Gleixner
|
9ff554e9be |
treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 82
Based on 1 normalized pattern(s): this code is released under the gnu general public license version 2 or later extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 3 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Richard Fontana <rfontana@redhat.com> Reviewed-by: Armijn Hemel <armijn@tjaldur.nl> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190520075211.232210963@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> |
||
Len Brown
|
2e4c54dac7 |
topology: Create core_cpus and die_cpus sysfs attributes
Create CPU topology sysfs attributes: "core_cpus" and "core_cpus_list" These attributes represent all of the logical CPUs that share the same core. These attriutes is synonymous with the existing "thread_siblings" and "thread_siblings_list" attribute, which will be deprecated. Create CPU topology sysfs attributes: "die_cpus" and "die_cpus_list". These attributes represent all of the logical CPUs that share the same die. Suggested-by: Brice Goglin <Brice.Goglin@inria.fr> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Ingo Molnar <mingo@kernel.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/071c23a298cd27ede6ed0b6460cae190d193364f.1557769318.git.len.brown@intel.com |
||
|
Len Brown
|
212bf4fdb7 |
x86/topology: Define topology_logical_die_id()
Define topology_logical_die_id() ala existing topology_logical_package_id() Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Ingo Molnar <mingo@kernel.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/2f3526e25ae14fbeff26fb26e877d159df8946d9.1557769318.git.len.brown@intel.com |
||
|
Len Brown
|
7745f03eb3 |
x86/topology: Add CPUID.1F multi-die/package support
Some new systems have multiple software-visible die within each package. Update Linux parsing of the Intel CPUID "Extended Topology Leaf" to handle either CPUID.B, or the new CPUID.1F. Add cpuinfo_x86.die_id and cpuinfo_x86.max_dies to store the result. die_id will be non-zero only for multi-die/package systems. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Ingo Molnar <mingo@kernel.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: linux-doc@vger.kernel.org Link: https://lkml.kernel.org/r/7b23d2d26d717b8e14ba137c94b70943f1ae4b5c.1557769318.git.len.brown@intel.com |
||
|
Linus Torvalds
|
948a64995a |
Merge branch 'x86-topology-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 topology updates from Ingo Molnar: "Two main changes: preparatory changes for Intel multi-die topology support, plus a syslog message tweak" * 'x86-topology-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86/topology: Make DEBUG_HOTPLUG_CPU0 pr_info() more descriptive x86/smpboot: Rename match_die() to match_pkg() topology: Simplify cputopology.txt formatting and wording x86/topology: Fix documentation typo |
||
|
Len Brown
|
169d086996 |
x86/smpboot: Rename match_die() to match_pkg()
Syntax only, no functional or semantic change. This routine matches packages, not die, so name it thus. Signed-off-by: Len Brown <len.brown@intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will.deacon@arm.com> Link: http://lkml.kernel.org/r/7ca18c4ae7816a1f9eda37414725df676e63589d.1551160674.git.len.brown@intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
|
Thomas Gleixner
|
66c7ceb47f |
x86/irq/32: Handle irq stack allocation failure proper
irq_ctx_init() crashes hard on page allocation failures. While that's ok during early boot, it's just wrong in the CPU hotplug bringup code. Check the page allocation failure and return -ENOMEM and handle it at the call sites. On early boot the only way out is to BUG(), but on CPU hotplug there is no reason to crash, so just abort the operation. Rename the function to something more sensible while at it. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Alison Schofield <alison.schofield@intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Anshuman Khandual <anshuman.khandual@arm.com> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Nicolai Stange <nstange@suse.de> Cc: Pu Wen <puwen@hygon.cn> Cc: Sean Christopherson <sean.j.christopherson@intel.com> Cc: Shaokun Zhang <zhangshaokun@hisilicon.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Cc: x86-ml <x86@kernel.org> Cc: xen-devel@lists.xenproject.org Cc: Yazen Ghannam <yazen.ghannam@amd.com> Cc: Yi Wang <wang.yi59@zte.com.cn> Cc: Zhenzhong Duan <zhenzhong.duan@oracle.com> Link: https://lkml.kernel.org/r/20190414160146.089060584@linutronix.de |
||
|
Linus Torvalds
|
bcd49c3dd1 |
Merge branch 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 cleanups from Ingo Molnar: "Various cleanups and simplifications, none of them really stands out, they are all over the place" * 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86/uaccess: Remove unused __addr_ok() macro x86/smpboot: Remove unused phys_id variable x86/mm/dump_pagetables: Remove the unused prev_pud variable x86/fpu: Move init_xstate_size() to __init section x86/cpu_entry_area: Move percpu_setup_debug_store() to __init section x86/mtrr: Remove unused variable x86/boot/compressed/64: Explain paging_prepare()'s return value x86/resctrl: Remove duplicate MSR_MISC_FEATURE_CONTROL definition x86/asm/suspend: Drop ENTRY from local data x86/hw_breakpoints, kprobes: Remove kprobes ifdeffery x86/boot: Save several bytes in decompressor x86/trap: Remove useless declaration x86/mm/tlb: Remove unused cpu variable x86/events: Mark expected switch-case fall-throughs x86/asm-prototypes: Remove duplicate include <asm/page.h> x86/kernel: Mark expected switch-case fall-throughs x86/insn-eval: Mark expected switch-case fall-through x86/platform/UV: Replace kmalloc() and memset() with k[cz]alloc() calls x86/e820: Replace kmalloc() + memcpy() with kmemdup() |
||
Anshuman Khandual
|
98fa15f34c |
mm: replace all open encodings for NUMA_NO_NODE
Patch series "Replace all open encodings for NUMA_NO_NODE", v3. All these places for replacement were found by running the following grep patterns on the entire kernel code. Please let me know if this might have missed some instances. This might also have replaced some false positives. I will appreciate suggestions, inputs and review. 1. git grep "nid == -1" 2. git grep "node == -1" 3. git grep "nid = -1" 4. git grep "node = -1" This patch (of 2): At present there are multiple places where invalid node number is encoded as -1. Even though implicitly understood it is always better to have macros in there. Replace these open encodings for an invalid node number with the global macro NUMA_NO_NODE. This helps remove NUMA related assumptions like 'invalid node' from various places redirecting them to a common definition. Link: http://lkml.kernel.org/r/1545127933-10711-2-git-send-email-anshuman.khandual@arm.com Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com> Reviewed-by: David Hildenbrand <david@redhat.com> Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> [ixgbe] Acked-by: Jens Axboe <axboe@kernel.dk> [mtip32xx] Acked-by: Vinod Koul <vkoul@kernel.org> [dmaengine.c] Acked-by: Michael Ellerman <mpe@ellerman.id.au> [powerpc] Acked-by: Doug Ledford <dledford@redhat.com> [drivers/infiniband] Cc: Joseph Qi <jiangqi903@gmail.com> Cc: Hans Verkuil <hverkuil@xs4all.nl> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> |
||
Shaokun Zhang
|
f91fecc09e |
x86/smpboot: Remove unused phys_id variable
The 'phys_id' local variable became unused after commit
|
||
Hui Wang
|
aa02ef099c |
x86/topology: Use total_cpus for max logical packages calculation
nr_cpu_ids can be limited on the command line via nr_cpus=. This can break the logical package management because it results in a smaller number of packages while in kdump kernel. Check below case: There is a two sockets system, each socket has 8 cores, which has 16 logical cpus while HT was turn on. 0 1 2 3 4 5 6 7 | 16 17 18 19 20 21 22 23 cores on socket 0 threads on socket 0 8 9 10 11 12 13 14 15 | 24 25 26 27 28 29 30 31 cores on socket 1 threads on socket 1 While starting the kdump kernel with command line option nr_cpus=16 panic was triggered on one of the cpus 24-31 eg. 26, then online cpu will be 1-15, 26(cpu 0 was disabled in kdump), ncpus will be 16 and __max_logical_packages will be 1, but actually two packages were booted on. This issue can reproduced by set kdump option nr_cpus=<real physical core numbers>, and then trigger panic on last socket's thread, for example: taskset -c 26 echo c > /proc/sysrq-trigger Use total_cpus which will not be limited by nr_cpus command line to calculate the value of __max_logical_packages. Signed-off-by: Hui Wang <john.wanghui@huawei.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: <guijianfeng@huawei.com> Cc: <wencongyang2@huawei.com> Cc: <douliyang1@huawei.com> Cc: <qiaonuohan@huawei.com> Link: https://lkml.kernel.org/r/20181107023643.22174-1-john.wanghui@huawei.com |
||
Mike Rapoport
|
57c8a661d9 |
mm: remove include/linux/bootmem.h
Move remaining definitions and declarations from include/linux/bootmem.h into include/linux/memblock.h and remove the redundant header. The includes were replaced with the semantic patch below and then semi-automated removal of duplicated '#include <linux/memblock.h> @@ @@ - #include <linux/bootmem.h> + #include <linux/memblock.h> [sfr@canb.auug.org.au: dma-direct: fix up for the removal of linux/bootmem.h] Link: http://lkml.kernel.org/r/20181002185342.133d1680@canb.auug.org.au [sfr@canb.auug.org.au: powerpc: fix up for removal of linux/bootmem.h] Link: http://lkml.kernel.org/r/20181005161406.73ef8727@canb.auug.org.au [sfr@canb.auug.org.au: x86/kaslr, ACPI/NUMA: fix for linux/bootmem.h removal] Link: http://lkml.kernel.org/r/20181008190341.5e396491@canb.auug.org.au Link: http://lkml.kernel.org/r/1536927045-23536-30-git-send-email-rppt@linux.vnet.ibm.com Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au> Acked-by: Michal Hocko <mhocko@suse.com> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Chris Zankel <chris@zankel.net> Cc: "David S. Miller" <davem@davemloft.net> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Greentime Hu <green.hu@gmail.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Guan Xuetao <gxt@pku.edu.cn> Cc: Ingo Molnar <mingo@redhat.com> Cc: "James E.J. Bottomley" <jejb@parisc-linux.org> Cc: Jonas Bonn <jonas@southpole.se> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Ley Foon Tan <lftan@altera.com> Cc: Mark Salter <msalter@redhat.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Matt Turner <mattst88@gmail.com> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Michal Simek <monstr@monstr.eu> Cc: Palmer Dabbelt <palmer@sifive.com> Cc: Paul Burton <paul.burton@mips.com> Cc: Richard Kuo <rkuo@codeaurora.org> Cc: Richard Weinberger <richard@nod.at> Cc: Rich Felker <dalias@libc.org> Cc: Russell King <linux@armlinux.org.uk> Cc: Serge Semin <fancer.lancer@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tony Luck <tony.luck@intel.com> Cc: Vineet Gupta <vgupta@synopsys.com> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> |
||
Pu Wen
|
0b13bec787 |
x86/smpboot: Do not use BSP INIT delay and MWAIT to idle on Dhyana
The Hygon Dhyana CPU uses no delay in smp_quirk_init_udelay(), and does HLT on idle just like AMD does. Signed-off-by: Pu Wen <puwen@hygon.cn> Signed-off-by: Borislav Petkov <bp@suse.de> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: bp@alien8.de Cc: tglx@linutronix.de Cc: mingo@redhat.com Cc: hpa@zytor.com Cc: x86@kernel.org Cc: thomas.lendacky@amd.com Link: https://lkml.kernel.org/r/87000fa82e273f5967c908448414228faf61e077.1537533369.git.puwen@hygon.cn |
||
|
Thomas Gleixner
|
f2701b77bb |
Merge 4.18-rc7 into master to pick up the KVM dependcy
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> |
||
Nicolai Stange
|
447ae31667 |
x86: Don't include linux/irq.h from asm/hardirq.h
The next patch in this series will have to make the definition of
irq_cpustat_t available to entering_irq().
Inclusion of asm/hardirq.h into asm/apic.h would cause circular header
dependencies like
asm/smp.h
asm/apic.h
asm/hardirq.h
linux/irq.h
linux/topology.h
linux/smp.h
asm/smp.h
or
linux/gfp.h
linux/mmzone.h
asm/mmzone.h
asm/mmzone_64.h
asm/smp.h
asm/apic.h
asm/hardirq.h
linux/irq.h
linux/irqdesc.h
linux/kobject.h
linux/sysfs.h
linux/kernfs.h
linux/idr.h
linux/gfp.h
and others.
This causes compilation errors because of the header guards becoming
effective in the second inclusion: symbols/macros that had been defined
before wouldn't be available to intermediate headers in the #include chain
anymore.
A possible workaround would be to move the definition of irq_cpustat_t
into its own header and include that from both, asm/hardirq.h and
asm/apic.h.
However, this wouldn't solve the real problem, namely asm/harirq.h
unnecessarily pulling in all the linux/irq.h cruft: nothing in
asm/hardirq.h itself requires it. Also, note that there are some other
archs, like e.g. arm64, which don't have that #include in their
asm/hardirq.h.
Remove the linux/irq.h #include from x86' asm/hardirq.h.
Fix resulting compilation errors by adding appropriate #includes to *.c
files as needed.
Note that some of these *.c files could be cleaned up a bit wrt. to their
set of #includes, but that should better be done from separate patches, if
at all.
Signed-off-by: Nicolai Stange <nstange@suse.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
||
|
Zhenzhong Duan
|
4fb5f58e8d |
x86/mm/32: Initialize the CR4 shadow before __flush_tlb_all()
On 32-bit kernels, __flush_tlb_all() may have read the CR4 shadow before the initialization of CR4 shadow in cpu_init(). Fix it by adding an explicit cr4_init_shadow() call into start_secondary() which is the first function called on non-boot SMP CPUs - ahead of the __flush_tlb_all() call. ( This is somewhat of a layering violation, but start_secondary() does CR4 bootstrap in the PCID case anyway. ) Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Link: http://lkml.kernel.org/r/b07b6ae9-4b57-4b40-b9bc-50c2c67f1d91@default Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
|
Thomas Gleixner
|
f048c399e0 |
x86/topology: Provide topology_smt_supported()
Provide information whether SMT is supoorted by the CPUs. Preparatory patch for SMT control mechanism. Suggested-by: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@kernel.org> |
||
|
Thomas Gleixner
|
6a4d2657e0 |
x86/smp: Provide topology_is_primary_thread()
If the CPU is supporting SMT then the primary thread can be found by checking the lower APIC ID bits for zero. smp_num_siblings is used to build the mask for the APIC ID bits which need to be taken into account. This uses the MPTABLE or ACPI/MADT supplied APIC ID, which can be different than the initial APIC ID in CPUID. But according to AMD the lower bits have to be consistent. Intel gave a tentative confirmation as well. Preparatory patch to support disabling SMT at boot/runtime. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Acked-by: Ingo Molnar <mingo@kernel.org> |
||
|
Linus Torvalds
|
5cef8c2a22 |
Merge branch 'x86-boot-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 boot updates from Ingo Molnar: - Centaur CPU updates (David Wang) - AMD and other CPU topology enumeration improvements and fixes (Borislav Petkov, Thomas Gleixner, Suravee Suthikulpanit) - Continued 5-level paging work (Kirill A. Shutemov) * 'x86-boot-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86/mm: Mark __pgtable_l5_enabled __initdata x86/mm: Mark p4d_offset() __always_inline x86/mm: Introduce the 'no5lvl' kernel parameter x86/mm: Stop pretending pgtable_l5_enabled is a variable x86/mm: Unify pgtable_l5_enabled usage in early boot code x86/boot/compressed/64: Fix trampoline page table address calculation x86/CPU: Move x86_cpuinfo::x86_max_cores assignment to detect_num_cpu_cores() x86/Centaur: Report correct CPU/cache topology x86/CPU: Move cpu_detect_cache_sizes() into init_intel_cacheinfo() x86/CPU: Make intel_num_cpu_cores() generic x86/CPU: Move cpu local function declarations to local header x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available x86/CPU: Modify detect_extended_topology() to return result x86/CPU/AMD: Calculate last level cache ID from number of sharing threads x86/CPU: Rename intel_cacheinfo.c to cacheinfo.c perf/events/amd/uncore: Fix amd_uncore_llc ID to use pre-defined cpu_llc_id x86/CPU/AMD: Have smp_num_siblings and cpu_llc_id always be present x86/Centaur: Initialize supported CPU features properly |
||
Ingo Molnar
|
177bfd725b |
Merge branches 'x86/urgent' and 'core/urgent' into x86/boot, to pick up fixes and avoid conflicts
Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
|
Thomas Gleixner
|
1f50ddb4f4 |
x86/speculation: Handle HT correctly on AMD
The AMD64_LS_CFG MSR is a per core MSR on Family 17H CPUs. That means when hyperthreading is enabled the SSBD bit toggle needs to take both cores into account. Otherwise the following situation can happen: CPU0 CPU1 disable SSB disable SSB enable SSB <- Enables it for the Core, i.e. for CPU0 as well So after the SSB enable on CPU1 the task on CPU0 runs with SSB enabled again. On Intel the SSBD control is per core as well, but the synchronization logic is implemented behind the per thread SPEC_CTRL MSR. It works like this: CORE_SPEC_CTRL = THREAD0_SPEC_CTRL | THREAD1_SPEC_CTRL i.e. if one of the threads enables a mitigation then this affects both and the mitigation is only disabled in the core when both threads disabled it. Add the necessary synchronization logic for AMD family 17H. Unfortunately that requires a spinlock to serialize the access to the MSR, but the locks are only shared between siblings. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Borislav Petkov <bp@suse.de> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> |
||
Borislav Petkov
|
f8b64d08dd |
x86/CPU/AMD: Have smp_num_siblings and cpu_llc_id always be present
Move smp_num_siblings and cpu_llc_id to cpu/common.c so that they're always present as symbols and not only in the CONFIG_SMP case. Then, other code using them doesn't need ugly ifdeffery anymore. Get rid of some ifdeffery. Signed-off-by: Borislav Petkov <bpetkov@suse.de> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1524864877-111962-2-git-send-email-suravee.suthikulpanit@amd.com |
||
Yazen Ghannam
|
da6fa7ef67 |
x86/smpboot: Don't use mwait_play_dead() on AMD systems
Recent AMD systems support using MWAIT for C1 state. However, MWAIT will not allow deeper cstates than C1 on current systems. play_dead() expects to use the deepest state available. The deepest state available on AMD systems is reached through SystemIO or HALT. If MWAIT is available, it is preferred over the other methods, so the CPU never reaches the deepest possible state. Don't try to use MWAIT to play_dead() on AMD systems. Instead, use CPUIDLE to enter the deepest state advertised by firmware. If CPUIDLE is not available then fallback to HALT. Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: stable@vger.kernel.org Cc: Yazen Ghannam <Yazen.Ghannam@amd.com> Link: https://lkml.kernel.org/r/20180403140228.58540-1-Yazen.Ghannam@amd.com |
||
Alison Schofield
|
1340ccfa9a |
x86,sched: Allow topologies where NUMA nodes share an LLC
Intel's Skylake Server CPUs have a different LLC topology than previous
generations. When in Sub-NUMA-Clustering (SNC) mode, the package is divided
into two "slices", each containing half the cores, half the LLC, and one
memory controller and each slice is enumerated to Linux as a NUMA
node. This is similar to how the cores and LLC were arranged for the
Cluster-On-Die (CoD) feature.
CoD allowed the same cache line to be present in each half of the LLC.
But, with SNC, each line is only ever present in *one* slice. This means
that the portion of the LLC *available* to a CPU depends on the data being
accessed:
Remote socket: entire package LLC is shared
Local socket->local slice: data goes into local slice LLC
Local socket->remote slice: data goes into remote-slice LLC. Slightly
higher latency than local slice LLC.
The biggest implication from this is that a process accessing all
NUMA-local memory only sees half the LLC capacity.
The CPU describes its cache hierarchy with the CPUID instruction. One of
the CPUID leaves enumerates the "logical processors sharing this
cache". This information is used for scheduling decisions so that tasks
move more freely between CPUs sharing the cache.
But, the CPUID for the SNC configuration discussed above enumerates the LLC
as being shared by the entire package. This is not 100% precise because the
entire cache is not usable by all accesses. But, it *is* the way the
hardware enumerates itself, and this is not likely to change.
The userspace visible impact of all the above is that the sysfs info
reports the entire LLC as being available to the entire package. As noted
above, this is not true for local socket accesses. This patch does not
correct the sysfs info. It is the same, pre and post patch.
The current code emits the following warning:
sched: CPU #3's llc-sibling CPU #0 is not on the same node! [node: 1 != 0]. Ignoring dependency.
The warning is coming from the topology_sane() check in smpboot.c because
the topology is not matching the expectations of the model for obvious
reasons.
To fix this, add a vendor and model specific check to never call
topology_sane() for these systems. Also, just like "Cluster-on-Die" disable
the "coregroup" sched_domain_topology_level and use NUMA information from
the SRAT alone.
This is OK at least on the hardware we are immediately concerned about
because the LLC sharing happens at both the slice and at the package level,
which are also NUMA boundaries.
Signed-off-by: Alison Schofield <alison.schofield@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Borislav Petkov <bp@suse.de>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: brice.goglin@gmail.com
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: "H. Peter Anvin" <hpa@linux.intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: https://lkml.kernel.org/r/20180407002130.GA18984@alison-desk.jf.intel.com
|
||
Samuel Neves
|
4596749339 |
x86/topology: Update the 'cpu cores' field in /proc/cpuinfo correctly across CPU hotplug operations
Without this fix, /proc/cpuinfo will display an incorrect amount of CPU cores, after bringing them offline and online again, as exemplified below: $ cat /proc/cpuinfo | grep cores cpu cores : 4 cpu cores : 8 cpu cores : 8 cpu cores : 20 cpu cores : 4 cpu cores : 3 cpu cores : 2 cpu cores : 2 This patch fixes this by always zeroing the booted_cores variable upon turning off a logical CPU. Tested-by: Dou Liyang <douly.fnst@cn.fujitsu.com> Signed-off-by: Samuel Neves <sneves@dei.uc.pt> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: jgross@suse.com Cc: luto@kernel.org Cc: prarit@redhat.com Cc: vkuznets@redhat.com Link: http://lkml.kernel.org/r/20180221205036.5244-1-sneves@dei.uc.pt Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
Prarit Bhargava
|
63e708f826 |
x86/xen: Calculate __max_logical_packages on PV domains
The kernel panics on PV domains because native_smp_cpus_done() is
only called for HVM domains.
Calculate __max_logical_packages for PV domains.
Fixes:
|
||
Masayoshi Mizuma
|
295cc7eb31 |
x86/smpboot: Fix uncore_pci_remove() indexing bug when hot-removing a physical CPU
When a physical CPU is hot-removed, the following warning messages
are shown while the uncore device is removed in uncore_pci_remove():
WARNING: CPU: 120 PID: 5 at arch/x86/events/intel/uncore.c:988
uncore_pci_remove+0xf1/0x110
...
CPU: 120 PID: 5 Comm: kworker/u1024:0 Not tainted 4.15.0-rc8 #1
Workqueue: kacpi_hotplug acpi_hotplug_work_fn
...
Call Trace:
pci_device_remove+0x36/0xb0
device_release_driver_internal+0x145/0x210
pci_stop_bus_device+0x76/0xa0
pci_stop_root_bus+0x44/0x60
acpi_pci_root_remove+0x1f/0x80
acpi_bus_trim+0x54/0x90
acpi_bus_trim+0x2e/0x90
acpi_device_hotplug+0x2bc/0x4b0
acpi_hotplug_work_fn+0x1a/0x30
process_one_work+0x141/0x340
worker_thread+0x47/0x3e0
kthread+0xf5/0x130
When uncore_pci_remove() runs, it tries to get the package ID to
clear the value of uncore_extra_pci_dev[].dev[] by using
topology_phys_to_logical_pkg(). The warning messesages are
shown because topology_phys_to_logical_pkg() returns -1.
arch/x86/events/intel/uncore.c:
static void uncore_pci_remove(struct pci_dev *pdev)
{
...
phys_id = uncore_pcibus_to_physid(pdev->bus);
...
pkg = topology_phys_to_logical_pkg(phys_id); // returns -1
for (i = 0; i < UNCORE_EXTRA_PCI_DEV_MAX; i++) {
if (uncore_extra_pci_dev[pkg].dev[i] == pdev) {
uncore_extra_pci_dev[pkg].dev[i] = NULL;
break;
}
}
WARN_ON_ONCE(i >= UNCORE_EXTRA_PCI_DEV_MAX); // <=========== HERE!!
topology_phys_to_logical_pkg() tries to find
cpuinfo_x86->phys_proc_id that matches the phys_pkg argument.
arch/x86/kernel/smpboot.c:
int topology_phys_to_logical_pkg(unsigned int phys_pkg)
{
int cpu;
for_each_possible_cpu(cpu) {
struct cpuinfo_x86 *c = &cpu_data(cpu);
if (c->initialized && c->phys_proc_id == phys_pkg)
return c->logical_proc_id;
}
return -1;
}
However, the phys_proc_id was already set to 0 by remove_siblinginfo()
when the CPU was offlined.
So, topology_phys_to_logical_pkg() cannot find the correct
logical_proc_id and always returns -1.
As the result, uncore_pci_remove() calls WARN_ON_ONCE() and the warning
messages are shown.
What is worse is that the bogus 'pkg' index results in two bugs:
- We dereference uncore_extra_pci_dev[] with a negative index
- We fail to clean up a stale pointer in uncore_extra_pci_dev[][]
To fix these bugs, remove the clearing of ->phys_proc_id from remove_siblinginfo().
This should not cause any problems, because ->phys_proc_id is not
used after it is hot-removed and it is re-set while hot-adding.
Signed-off-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: yasu.isimatu@gmail.com
Cc: <stable@vger.kernel.org>
Fixes:
|
||
|
Linus Torvalds
|
3ccabd6d9d |
Merge branch 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 cleanups from Ingo Molnar: "Misc cleanups" * 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86: Remove unused IOMMU_STRESS Kconfig x86/extable: Mark exception handler functions visible x86/timer: Don't inline __const_udelay x86/headers: Remove duplicate #includes |
||
Jan Kiszka
|
e348caef8b |
x86/platform: Control warm reset setup via legacy feature flag
Allow to turn off the setup of BIOS-managed warm reset via a new flag in x86_legacy_features. Besides the UV1, the upcoming jailhose guest support needs this switched off. Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: jailhouse-dev@googlegroups.com Link: https://lkml.kernel.org/r/44376558129d70a2c1527959811371ef4b82e829.1511770314.git.jan.kiszka@siemens.com |
||
|
Linus Torvalds
|
52c90f2d32 |
Merge branch 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 page table isolation fixes from Thomas Gleixner:
"Four patches addressing the PTI fallout as discussed and debugged
yesterday:
- Remove stale and pointless TLB flush invocations from the hotplug
code
- Remove stale preempt_disable/enable from __native_flush_tlb()
- Plug the memory leak in the write_ldt() error path"
* 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/ldt: Make LDT pgtable free conditional
x86/ldt: Plug memory leak in error path
x86/mm: Remove preempt_disable/enable() from __native_flush_tlb()
x86/smpboot: Remove stale TLB flush invocations
|
||
|
Thomas Gleixner
|
322f8b8b34 |
x86/smpboot: Remove stale TLB flush invocations
smpboot_setup_warm_reset_vector() and smpboot_restore_warm_reset_vector() invoke local_flush_tlb() for no obvious reason. Digging in history revealed that the original code in the 2.1 era added those because the code manipulated a swapper_pg_dir pagetable entry. The pagetable manipulation was removed long ago in the 2.3 timeframe, but the TLB flush invocations stayed around forever. Remove them along with the pointless pr_debug()s which come from the same 2.1 change. Reported-by: Dominik Brodowski <linux@dominikbrodowski.net> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: <stable@vger.kernel.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Linus Torvalds <torvalds@linuxfoundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20171230211829.586548655@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
|
Linus Torvalds
|
caf9a82657 |
Merge branch 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 PTI preparatory patches from Thomas Gleixner:
"Todays Advent calendar window contains twentyfour easy to digest
patches. The original plan was to have twenty three matching the date,
but a late fixup made that moot.
- Move the cpu_entry_area mapping out of the fixmap into a separate
address space. That's necessary because the fixmap becomes too big
with NRCPUS=8192 and this caused already subtle and hard to
diagnose failures.
The top most patch is fresh from today and cures a brain slip of
that tall grumpy german greybeard, who ignored the intricacies of
32bit wraparounds.
- Limit the number of CPUs on 32bit to 64. That's insane big already,
but at least it's small enough to prevent address space issues with
the cpu_entry_area map, which have been observed and debugged with
the fixmap code
- A few TLB flush fixes in various places plus documentation which of
the TLB functions should be used for what.
- Rename the SYSENTER stack to CPU_ENTRY_AREA stack as it is used for
more than sysenter now and keeping the name makes backtraces
confusing.
- Prevent LDT inheritance on exec() by moving it to arch_dup_mmap(),
which is only invoked on fork().
- Make vysycall more robust.
- A few fixes and cleanups of the debug_pagetables code. Check
PAGE_PRESENT instead of checking the PTE for 0 and a cleanup of the
C89 initialization of the address hint array which already was out
of sync with the index enums.
- Move the ESPFIX init to a different place to prepare for PTI.
- Several code moves with no functional change to make PTI
integration simpler and header files less convoluted.
- Documentation fixes and clarifications"
* 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (24 commits)
x86/cpu_entry_area: Prevent wraparound in setup_cpu_entry_area_ptes() on 32bit
init: Invoke init_espfix_bsp() from mm_init()
x86/cpu_entry_area: Move it out of the fixmap
x86/cpu_entry_area: Move it to a separate unit
x86/mm: Create asm/invpcid.h
x86/mm: Put MMU to hardware ASID translation in one place
x86/mm: Remove hard-coded ASID limit checks
x86/mm: Move the CR3 construction functions to tlbflush.h
x86/mm: Add comments to clarify which TLB-flush functions are supposed to flush what
x86/mm: Remove superfluous barriers
x86/mm: Use __flush_tlb_one() for kernel memory
x86/microcode: Dont abuse the TLB-flush interface
x86/uv: Use the right TLB-flush API
x86/entry: Rename SYSENTER_stack to CPU_ENTRY_AREA_entry_stack
x86/doc: Remove obvious weirdnesses from the x86 MM layout documentation
x86/mm/64: Improve the memory map documentation
x86/ldt: Prevent LDT inheritance on exec
x86/ldt: Rework locking
arch, mm: Allow arch_dup_mmap() to fail
x86/vsyscall/64: Warn and fail vsyscall emulation in NATIVE mode
...
|
||
|
Thomas Gleixner
|
613e396bc0 |
init: Invoke init_espfix_bsp() from mm_init()
init_espfix_bsp() needs to be invoked before the page table isolation initialization. Move it into mm_init() which is the place where pti_init() will be added. While at it get rid of the #ifdeffery and provide proper stub functions. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
|
Ingo Molnar
|
0fd2e9c53d |
Merge commit 'upstream-x86-entry' into WIP.x86/mm
Pull in a minimal set of v4.15 entry code changes, for a base for the MM isolation patches. Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
Pravin Shedge
|
81bf665d00 |
x86/headers: Remove duplicate #includes
These duplicate includes have been found with scripts/checkincludes.pl but they have been removed manually to avoid removing false positives. Signed-off-by: Pravin Shedge <pravin.shedge4linux@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: ard.biesheuvel@linaro.org Cc: boris.ostrovsky@oracle.com Cc: geert@linux-m68k.org Cc: jgross@suse.com Cc: linux-efi@vger.kernel.org Cc: luto@kernel.org Cc: matt@codeblueprint.co.uk Cc: thomas.lendacky@amd.com Cc: tim.c.chen@linux.intel.com Cc: xen-devel@lists.xenproject.org Link: http://lkml.kernel.org/r/1513024951-9221-1-git-send-email-pravin.shedge4linux@gmail.com Signed-off-by: Ingo Molnar <mingo@kernel.org> |
||
|
Prarit Bhargava
|
947134d9b0 |
x86/smpboot: Do not use smp_num_siblings in __max_logical_packages calculation
Documentation/x86/topology.txt defines smp_num_siblings as "The number of threads in a core". Since commit |
||
Chunyu Hu
|
55d2d0ad2f |
x86/idt: Load idt early in start_secondary
On a secondary, idt is first loaded in cpu_init() with load_current_idt(),
i.e. no exceptions can be handled before that point.
The conversion of WARN() to use UD requires the IDT being loaded earlier as
any warning between start_secondary() and load_curren_idt() in cpu_init()
will result in an unhandled @UD exception and therefore fail the bringup of
the CPU.
Install the IDT handlers right in start_secondary() before calling cpu_init().
[ tglx: Massaged changelog ]
Fixes:
|
||
|
Prarit Bhargava
|
b4c0a7326f |
x86/smpboot: Fix __max_logical_packages estimate
A system booted with a small number of cores enabled per package panics because the estimate of __max_logical_packages is too low. This occurs when the total number of active cores across all packages is less than the maximum core count for a single package. e.g.: On a 4 package system with 20 cores/package where only 4 cores are enabled on each package, the value of __max_logical_packages is calculated as DIV_ROUND_UP(16 / 20) = 1 and not 4. Calculate __max_logical_packages after the cpu enumeration has completed. Use the boot cpu's data to extrapolate the number of packages. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Tom Lendacky <thomas.lendacky@amd.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Kan Liang <kan.liang@intel.com> Cc: He Chen <he.chen@linux.intel.com> Cc: Stephane Eranian <eranian@google.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Piotr Luc <piotr.luc@intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arvind Yadav <arvind.yadav.cs@gmail.com> Cc: Vitaly Kuznetsov <vkuznets@redhat.com> Cc: Borislav Petkov <bp@suse.de> Cc: Tim Chen <tim.c.chen@linux.intel.com> Cc: Mathias Krause <minipli@googlemail.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Link: https://lkml.kernel.org/r/20171114124257.22013-4-prarit@redhat.com |
||
Andi Kleen
|
30bb981185 |
x86/topology: Avoid wasting 128k for package id array
Analyzing large early boot allocations unveiled the logical package id
storage as a prominent memory waste. Since commit
|
||
|
Linus Torvalds
|
b18d62891a |
Merge branch 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 APIC updates from Thomas Gleixner:
"This update provides a major overhaul of the APIC initialization and
vector allocation code:
- Unification of the APIC and interrupt mode setup which was
scattered all over the place and was hard to follow. This also
distangles the timer setup from the APIC initialization which
brings a clear separation of functionality.
Great detective work from Dou Lyiang!
- Refactoring of the x86 vector allocation mechanism. The existing
code was based on nested loops and rather convoluted APIC callbacks
which had a horrible worst case behaviour and tried to serve all
different use cases in one go. This led to quite odd hacks when
supporting the new managed interupt facility for multiqueue devices
and made it more or less impossible to deal with the vector space
exhaustion which was a major roadblock for server hibernation.
Aside of that the code dealing with cpu hotplug and the system
vectors was disconnected from the actual vector management and
allocation code, which made it hard to follow and maintain.
Utilizing the new bitmap matrix allocator core mechanism, the new
allocator and management code consolidates the handling of system
vectors, legacy vectors, cpu hotplug mechanisms and the actual
allocation which needs to be aware of system and legacy vectors and
hotplug constraints into a single consistent entity.
This has one visible change: The support for multi CPU targets of
interrupts, which is only available on a certain subset of
CPUs/APIC variants has been removed in favour of single interrupt
targets. A proper analysis of the multi CPU target feature revealed
that there is no real advantage as the vast majority of interrupts
end up on the CPU with the lowest APIC id in the set of target CPUs
anyway. That change was agreed on by the relevant folks and allowed
to simplify the implementation significantly and to replace rather
fragile constructs like the vector cleanup IPI with straight
forward and solid code.
Furthermore this allowed to cleanly separate the allocation details
for legacy, normal and managed interrupts:
* Legacy interrupts are not longer wasting 16 vectors
unconditionally
* Managed interrupts have now a guaranteed vector reservation, but
the actual vector assignment happens when the interrupt is
requested. It's guaranteed not to fail.
* Normal interrupts no longer allocate vectors unconditionally
when the interrupt is set up (IO/APIC init or MSI(X) enable).
The mechanism has been switched to a best effort reservation
mode. The actual allocation happens when the interrupt is
requested. Contrary to managed interrupts the request can fail
due to vector space exhaustion, but drivers must handle a fail
of request_irq() anyway. When the interrupt is freed, the vector
is handed back as well.
This solves a long standing problem with large unconditional
vector allocations for a certain class of enterprise devices
which prevented server hibernation due to vector space
exhaustion when the unused allocated vectors had to be migrated
to CPU0 while unplugging all non boot CPUs.
The code has been equipped with trace points and detailed debugfs
information to aid analysis of the vector space"
* 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (60 commits)
x86/vector/msi: Select CONFIG_GENERIC_IRQ_RESERVATION_MODE
PCI/MSI: Set MSI_FLAG_MUST_REACTIVATE in core code
genirq: Add config option for reservation mode
x86/vector: Use correct per cpu variable in free_moved_vector()
x86/apic/vector: Ignore set_affinity call for inactive interrupts
x86/apic: Fix spelling mistake: "symmectic" -> "symmetric"
x86/apic: Use dead_cpu instead of current CPU when cleaning up
ACPI/init: Invoke early ACPI initialization earlier
x86/vector: Respect affinity mask in irq descriptor
x86/irq: Simplify hotplug vector accounting
x86/vector: Switch IOAPIC to global reservation mode
x86/vector/msi: Switch to global reservation mode
x86/vector: Handle managed interrupts proper
x86/io_apic: Reevaluate vector configuration on activate()
iommu/amd: Reevaluate vector configuration on activate()
iommu/vt-d: Reevaluate vector configuration on activate()
x86/apic/msi: Force reactivation of interrupts at startup time
x86/vector: Untangle internal state from irq_cfg
x86/vector: Compile SMP only code conditionally
x86/apic: Remove unused callbacks
...
|
||
|
Linus Torvalds
|
6a9f70b0a5 |
Merge branch 'x86-boot-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 boot updates from Ingo Molnar: "Three smaller changes: - clang fix - boot message beautification - unnecessary header inclusion removal" * 'x86-boot-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86/boot: Disable Clang warnings about GNU extensions x86/boot: Remove unnecessary #include <generated/utsrelease.h> x86/boot: Spell out "boot CPU" for BP |
||
|
Linus Torvalds
|
d6ec9d9a4d |
Merge branch 'x86-asm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 core updates from Ingo Molnar:
"Note that in this cycle most of the x86 topics interacted at a level
that caused them to be merged into tip:x86/asm - but this should be a
temporary phenomenon, hopefully we'll back to the usual patterns in
the next merge window.
The main changes in this cycle were:
Hardware enablement:
- Add support for the Intel UMIP (User Mode Instruction Prevention)
CPU feature. This is a security feature that disables certain
instructions such as SGDT, SLDT, SIDT, SMSW and STR. (Ricardo Neri)
[ Note that this is disabled by default for now, there are some
smaller enhancements in the pipeline that I'll follow up with in
the next 1-2 days, which allows this to be enabled by default.]
- Add support for the AMD SEV (Secure Encrypted Virtualization) CPU
feature, on top of SME (Secure Memory Encryption) support that was
added in v4.14. (Tom Lendacky, Brijesh Singh)
- Enable new SSE/AVX/AVX512 CPU features: AVX512_VBMI2, GFNI, VAES,
VPCLMULQDQ, AVX512_VNNI, AVX512_BITALG. (Gayatri Kammela)
Other changes:
- A big series of entry code simplifications and enhancements (Andy
Lutomirski)
- Make the ORC unwinder default on x86 and various objtool
enhancements. (Josh Poimboeuf)
- 5-level paging enhancements (Kirill A. Shutemov)
- Micro-optimize the entry code a bit (Borislav Petkov)
- Improve the handling of interdependent CPU features in the early
FPU init code (Andi Kleen)
- Build system enhancements (Changbin Du, Masahiro Yamada)
- ... plus misc enhancements, fixes and cleanups"
* 'x86-asm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (118 commits)
x86/build: Make the boot image generation less verbose
selftests/x86: Add tests for the STR and SLDT instructions
selftests/x86: Add tests for User-Mode Instruction Prevention
x86/traps: Fix up general protection faults caused by UMIP
x86/umip: Enable User-Mode Instruction Prevention at runtime
x86/umip: Force a page fault when unable to copy emulated result to user
x86/umip: Add emulation code for UMIP instructions
x86/cpufeature: Add User-Mode Instruction Prevention definitions
x86/insn-eval: Add support to resolve 16-bit address encodings
x86/insn-eval: Handle 32-bit address encodings in virtual-8086 mode
x86/insn-eval: Add wrapper function for 32 and 64-bit addresses
x86/insn-eval: Add support to resolve 32-bit address encodings
x86/insn-eval: Compute linear address in several utility functions
resource: Fix resource_size.cocci warnings
X86/KVM: Clear encryption attribute when SEV is active
X86/KVM: Decrypt shared per-cpu variables when SEV is active
percpu: Introduce DEFINE_PER_CPU_DECRYPTED
x86: Add support for changing memory encryption attribute in early boot
x86/io: Unroll string I/O when SEV is active
x86/boot: Add early boot support when running with SEV active
...
|