[linux-yocto] [yocto-4.12][PATCH 1/3] x86/intel_rdt: Move special case code for Haswell to a quirk function

Bruce Ashfield bruce.ashfield at windriver.com
Wed Jan 10 07:37:26 PST 2018


On 2018-01-10 10:26 AM, Bruce Ashfield wrote:
> On 2018-01-10 12:44 AM, qwang2 wrote:
>>
>>
>> On 2018年01月10日 11:11, Bruce Ashfield wrote:
>>> On 2018-01-09 9:07 PM, qwang2 wrote:
>>>> Hi Bruce,
>>>>
>>>> This commit is based on the patches at 
>>>> https://www.mail-archive.com/linux-yocto@yoctoproject.org/msg06131.html. 
>>>>
>>>>
>>>> It seems that those patches hasn't been merged, would you please 
>>>> help to check it?
>>>
>>> Send them all in a completes series.
>> Hi Bruce,
>>
>> I made all these patches in a completes series, and they are at the 
>> git-tree as below.
> 
> Something isn't adding up for me.
> 
> The series you sent was 3 patches, and the one you referenced was CQM
> based.
> 
> I did look at the original pull request you referenced on the list, but
> when I followed up that I had found the patches (in December), that was
> for a 4 patch series, not something as large as you have below.

And to be clear, that 4 patch series was for rdt changes, but
looking at my 4.12 standard/base .. they aren't staged either
(although I claim that I did stage them). I see the rdt patches
in that branch you sent, so I assume you are seeing them missing
as well.

I'm not sure what happened there, but let's reset and clarify
things.

Question 1: are there three functional blocks (aka features) that
you are looking to be merged ?

Question 2: can you resend that pull request with a clear division
between the three features (or two).

Cheers,

Bruce

> 
> i.e. when I look at that original email:
> 
> --------------
> Hi Bruce,
> 
> These patches enable CQM (Cache monitoring) enhancement. And they are
> alll picked from upstream.
> 
> Thanks,
> 
> Quanyang
> 
> 
> The following changes since commit 
> 4226b065fca4f630901d99b99d18c395ae3866fb:
> 
>     ipv4: net namespace does not inherit network configurations
> (2017-12-05 17:17:05 -0500)
> 
> are available in the git repository at:
> 
>     https://github.com/wangquanyang/linux-yocto-4.12.git standard/base
> 
> for you to fetch changes up to 2453602a06e94f4a26aee5ad9441bdef2a79d8d6:
> 
>     perf tests: Remove Intel CQM perf test (2017-12-19 17:56:01 +0800)
> ---------------
> 
> And that email has no summary of the changes from you, no overall
> diffstat and no information on how they were tested.
> 
> As a result, I can't easily tell your 3 patch series for rdt from
> all the CQM patches (how many are there in total?).
> 
> You can leave the patches in a git branch (I can fetch and pull them
> from there), but can you resend the summary information I need, with
> a breakdown between the two features, so I can tell which patches belong
> to which feature.
> 
> Bruce
> 
> 
>>
>> Thanks,
>> Quanyang
>>
>> The following changes since commit 
>> 4226b065fca4f630901d99b99d18c395ae3866fb:
>>
>>    ipv4: net namespace does not inherit network configurations 
>> (2017-12-05 17:17:05 -0500)
>>
>> are available in the git repository at:
>>
>>    https://github.com/wangquanyang/linux-yocto-4.12.git standard/base
>>
>> for you to fetch changes up to 1d07c0cde181b7746c6091dc276ff6a205d5ca23:
>>
>>    x86/intel_rdt: Turn off most RDT features on Skylake (2018-01-10 
>> 13:13:22 +0800)
>>
>> ----------------------------------------------------------------
>> Alexei Starovoitov (1):
>>        perf, bpf: Add BPF support to all perf_event types
>>
>> Arnaldo Carvalho de Melo (5):
>>        tools: Adopt __noreturn from kernel sources
>>        perf event-parse: Use pr_warning()
>>        tools: Adopt __printf from kernel sources
>>        perf tools: Remove warning()
>>        perf test: Add 'struct test *' to the test functions
>>
>> Colin Ian King (2):
>>        x86/intel_rdt: Remove redundant ternary operator on return
>>        x86/intel_rdt/cqm: Make integer rmid_limbo_count static
>>
>> Fenghua Yu (1):
>>        x86/intel_rdt: Show bitmask of shareable resource with other 
>> executing units
>>
>> Jithu Joseph (1):
>>        x86/intel_rdt: Remove redundant assignment
>>
>> Reinette Chatre (2):
>>        x86/intel_rdt: Mark rdt_root and closid_alloc as static
>>        x86/intel_rdt: Initialize bitmask of shareable resource if CDP 
>> enabled
>>
>> Sebastian Andrzej Siewior (1):
>>        perf/x86/intel/cqm: Use cpuhp_setup_state_cpuslocked()
>>
>> Tony Luck (10):
>>        x86/intel_rdt: Simplify info and base file lists
>>        x86/intel_rdt/mbm: Basic counting of MBM events (total and local)
>>        x86/intel_rdt: Add framework for better RDT UI diagnostics
>>        x86/intel_rdt: Add diagnostics when writing the schemata file
>>        x86/intel_rdt: Add diagnostics when writing the tasks file
>>        x86/intel_rdt: Add diagnostics when writing the cpus file
>>        x86/intel_rdt: Add diagnostics when making directories
>>        x86/intel_rdt: Move special case code for Haswell to a quirk 
>> function
>>        x86/intel_rdt: Add command line options for resource director 
>> technology
>>        x86/intel_rdt: Turn off most RDT features on Skylake
>>
>> Vikas Shivappa (29):
>>        x86/perf/cqm: Wipe out perf based cqm
>>        x86/intel_rdt/cqm: Documentation for resctrl based RDT Monitoring
>>        x86/intel_rdt: Introduce a common compile option for RDT
>>        x86/intel_rdt: Change file names to accommodate RDT monitor code
>>        x86/intel_rdt: Cleanup namespace to support RDT monitoring
>>        x86/intel_rdt: Make rdt_resources_all more readable
>>        x86/intel_rdt/cqm: Add RDT monitoring initialization
>>        x86/intel_rdt/cqm: Add RMID (Resource monitoring ID) management
>>        x86/intel_rdt/cqm: Add info files for RDT monitoring
>>        x86/intel_rdt: Prepare for RDT monitoring mkdir support
>>        x86/intel_rdt/cqm: Add mkdir support for RDT monitoring
>>        x86/intel_rdt: Change closid type from int to u32
>>        x86/intel_rdt/cqm: Add tasks file support
>>        x86/intel_rdt: Prepare to add RDT monitor cpus file support
>>        x86/intel_rdt/cqm: Add cpus file support
>>        x86/intel_rdt: Prepare for RDT monitor data support
>>        x86/intel_rdt/cqm: Add mon_data
>>        x86/intel_rdt: Separate the ctrl bits from rmdir
>>        x86/intel_rdt/cqm: Add rmdir support
>>        x86/intel_rdt/cqm: Add mount,umount support
>>        x86/intel_rdt: Introduce rdt_enable_key for scheduling
>>        x86/intel_rdt/cqm: Add sched_in support
>>        x86/intel_rdt/cqm: Add CPU hotplug support
>>        x86/intel_rdt/mbm: Add mbm counter initialization
>>        x86/intel_rdt/mbm: Handle counter overflow
>>        x86/intel_rdt/mbm: Fix MBM overflow handler during CPU hotplug
>>        x86/intel_rdt/cqm: Improve limbo list processing
>>        x86/intel_rdt/cqm: Clear the default RMID during hotcpu
>>        x86/intel_rdt: Modify the intel_pqr_state for better performance
>>
>> Xiaochen Shen (2):
>>        x86/intel_rdt: Fix a silent failure when writing zero value 
>> schemata
>>        perf tests: Remove Intel CQM perf test
>>
>>   Documentation/admin-guide/kernel-parameters.rst |    1 +
>>   Documentation/admin-guide/kernel-parameters.txt |    6 +
>>   Documentation/x86/intel_rdt_ui.txt |  323 ++++++++++++--
>>   MAINTAINERS |    2 +-
>>   arch/x86/Kconfig |   12 +-
>>   arch/x86/events/intel/Makefile |    2 +-
>>   arch/x86/events/intel/cqm.c | 1766 
>> -------------------------------------------------------------------------
>>   arch/x86/include/asm/intel_rdt.h |  286 ------------
>>   arch/x86/include/asm/intel_rdt_common.h |   27 --
>>   arch/x86/include/asm/intel_rdt_sched.h |   92 ++++
>>   arch/x86/kernel/cpu/Makefile |    2 +-
>>   arch/x86/kernel/cpu/intel_rdt.c |  376 +++++++++++++---
>>   arch/x86/kernel/cpu/intel_rdt.h |  447 +++++++++++++++++++
>>   arch/x86/kernel/cpu/{intel_rdt_schemata.c => 
>> intel_rdt_ctrlmondata.c} |  117 ++++-
>>   arch/x86/kernel/cpu/intel_rdt_monitor.c |  499 +++++++++++++++++++++
>>   arch/x86/kernel/cpu/intel_rdt_rdtgroup.c | 1248 
>> +++++++++++++++++++++++++++++++++++++++++----------
>>   arch/x86/kernel/process_32.c |    2 +-
>>   arch/x86/kernel/process_64.c |    2 +-
>>   include/linux/perf_event.h |   25 +-
>>   include/linux/sched.h |    5 +-
>>   kernel/bpf/arraymap.c |   28 +-
>>   kernel/events/core.c |   51 +--
>>   kernel/trace/bpf_trace.c |   22 +-
>>   tools/include/linux/compiler-gcc.h |    4 +
>>   tools/perf/arch/arm/util/cs-etm.c |    4 +-
>>   tools/perf/arch/x86/include/arch-tests.h |   10 +-
>>   tools/perf/arch/x86/tests/Build |    1 -
>>   tools/perf/arch/x86/tests/arch-tests.c |    4 -
>>   tools/perf/arch/x86/tests/insn-x86.c |    2 +-
>>   tools/perf/arch/x86/tests/intel-cqm.c |  127 ------
>>   tools/perf/arch/x86/tests/perf-time-to-tsc.c |    2 +-
>>   tools/perf/arch/x86/tests/rdpmc.c |    2 +-
>>   tools/perf/tests/attr.c |    2 +-
>>   tools/perf/tests/backward-ring-buffer.c |    2 +-
>>   tools/perf/tests/bitmap.c |    2 +-
>>   tools/perf/tests/bp_signal.c |    2 +-
>>   tools/perf/tests/bp_signal_overflow.c |    2 +-
>>   tools/perf/tests/bpf.c |    4 +-
>>   tools/perf/tests/builtin-test.c |    4 +-
>>   tools/perf/tests/clang.c |    4 +-
>>   tools/perf/tests/code-reading.c |    2 +-
>>   tools/perf/tests/cpumap.c |    4 +-
>>   tools/perf/tests/dso-data.c |    6 +-
>>   tools/perf/tests/dwarf-unwind.c |    2 +-
>>   tools/perf/tests/event-times.c |    2 +-
>>   tools/perf/tests/event_update.c |    2 +-
>>   tools/perf/tests/evsel-roundtrip-name.c |    2 +-
>>   tools/perf/tests/evsel-tp-sched.c |    2 +-
>>   tools/perf/tests/expr.c |    2 +-
>>   tools/perf/tests/fdarray.c |    4 +-
>>   tools/perf/tests/hists_cumulate.c |    2 +-
>>   tools/perf/tests/hists_filter.c |    2 +-
>>   tools/perf/tests/hists_link.c |    2 +-
>>   tools/perf/tests/hists_output.c |    2 +-
>>   tools/perf/tests/is_printable_array.c |    2 +-
>>   tools/perf/tests/keep-tracking.c |    2 +-
>>   tools/perf/tests/kmod-path.c |    2 +-
>>   tools/perf/tests/llvm.c |    2 +-
>>   tools/perf/tests/mmap-basic.c |    2 +-
>>   tools/perf/tests/mmap-thread-lookup.c |    2 +-
>>   tools/perf/tests/openat-syscall-all-cpus.c |    2 +-
>>   tools/perf/tests/openat-syscall-tp-fields.c |    2 +-
>>   tools/perf/tests/openat-syscall.c |    2 +-
>>   tools/perf/tests/parse-events.c |   15 +-
>>   tools/perf/tests/parse-no-sample-id-all.c |    2 +-
>>   tools/perf/tests/perf-hooks.c |    2 +-
>>   tools/perf/tests/perf-record.c |    2 +-
>>   tools/perf/tests/pmu.c |    2 +-
>>   tools/perf/tests/python-use.c |    2 +-
>>   tools/perf/tests/sample-parsing.c |    2 +-
>>   tools/perf/tests/sdt.c |    4 +-
>>   tools/perf/tests/stat.c |    6 +-
>>   tools/perf/tests/sw-clock.c |    2 +-
>>   tools/perf/tests/switch-tracking.c |    2 +-
>>   tools/perf/tests/task-exit.c |    2 +-
>>   tools/perf/tests/tests.h |  112 ++---
>>   tools/perf/tests/thread-map.c |    6 +-
>>   tools/perf/tests/thread-mg-share.c |    2 +-
>>   tools/perf/tests/topology.c |    2 +-
>>   tools/perf/tests/unit_number__scnprintf.c |    2 +-
>>   tools/perf/tests/vmlinux-kallsyms.c |    2 +-
>>   tools/perf/util/cache.h |    3 +-
>>   tools/perf/util/debug.h |   11 +-
>>   tools/perf/util/intel-pt-decoder/intel-pt-log.h |    4 +-
>>   tools/perf/util/probe-event.h |    4 +-
>>   tools/perf/util/scripting-engines/trace-event-python.c |    3 +-
>>   tools/perf/util/strbuf.h |    4 +-
>>   tools/perf/util/trace-event-parse.c |    4 +-
>>   tools/perf/util/usage.c |   26 +-
>>   tools/perf/util/util.h |   19 +-
>>   90 files changed, 2987 insertions(+), 2833 deletions(-)
>>   delete mode 100644 arch/x86/events/intel/cqm.c
>>   delete mode 100644 arch/x86/include/asm/intel_rdt.h
>>   delete mode 100644 arch/x86/include/asm/intel_rdt_common.h
>>   create mode 100644 arch/x86/include/asm/intel_rdt_sched.h
>>   create mode 100644 arch/x86/kernel/cpu/intel_rdt.h
>>   rename arch/x86/kernel/cpu/{intel_rdt_schemata.c => 
>> intel_rdt_ctrlmondata.c} (69%)
>>   create mode 100644 arch/x86/kernel/cpu/intel_rdt_monitor.c
>>   delete mode 100644 tools/perf/arch/x86/tests/intel-cqm.c
>>
>>>
>>> What I have in the tree, is what I have merged. If you have
>>> dependencies that aren't in my tree, send them as part of a
>>> combined series.
>>>
>>> Bruce
>>>
>>>>
>>>> Thanks,
>>>>
>>>> Quanyang
>>>>
>>>>
>>>> On 2018年01月10日 02:19, Bruce Ashfield wrote:
>>>>> This series doesn't apply to the current 4.12:
>>>>>
>>>>> -----------------
>>>>> Applying: x86/intel_rdt: Move special case code for Haswell to a 
>>>>> quirk function
>>>>> error: patch failed: arch/x86/kernel/cpu/intel_rdt.c:172
>>>>> error: arch/x86/kernel/cpu/intel_rdt.c: patch does not apply
>>>>> Patch failed at 0001 x86/intel_rdt: Move special case code for 
>>>>> Haswell to a quirk function
>>>>> The copy of the patch that failed is found in:
>>>>> /home/bruce/poky-kernel/linux-yocto-4.12.git/.git/rebase-apply/patch
>>>>> When you have resolved this problem, run "git am --continue".
>>>>> If you prefer to skip this patch, run "git am --skip" instead.
>>>>> To restore the original branch and stop patching, run "git am 
>>>>> --abort".
>>>>> ---------------
>>>>>
>>>>> Can you double check that the 4.12 you used to generate this series
>>>>> matches what I have staged on git.yoctoproject.org.
>>>>>
>>>>> Bruce
>>>>>
>>>>> On 01/08/2018 10:20 PM, quanyang.wang at windriver.com wrote:
>>>>>> From: Tony Luck <tony.luck at intel.com>
>>>>>>
>>>>>> commit 0576113a387e0c8a5d9e24b4cd62605d1c9c0db8 upstream
>>>>>>
>>>>>> No functional change, but lay the ground work for other per-model
>>>>>> quirks.
>>>>>>
>>>>>> Signed-off-by: Tony Luck <tony.luck at intel.com>
>>>>>> Signed-off-by: Thomas Gleixner <tglx at linutronix.de>
>>>>>> Cc: Fenghua" <fenghua.yu at intel.com>
>>>>>> Cc: Ravi V" <ravi.v.shankar at intel.com>
>>>>>> Cc: "Peter Zijlstra" <peterz at infradead.org>
>>>>>> Cc: "Stephane Eranian" <eranian at google.com>
>>>>>> Cc: "Andi Kleen" <ak at linux.intel.com>
>>>>>> Cc: "David Carrillo-Cisneros" <davidcc at google.com>
>>>>>> Cc: Vikas Shivappa <vikas.shivappa at linux.intel.com>
>>>>>> Link: 
>>>>>> http://lkml.kernel.org/r/f195a83751b5f8b1d8a78bd3c1914300c8fa3142.1503512900.git.tony.luck@intel.com 
>>>>>>
>>>>>> ---
>>>>>>   arch/x86/kernel/cpu/intel_rdt.c | 52 
>>>>>> ++++++++++++++++++++++-------------------
>>>>>>   1 file changed, 28 insertions(+), 24 deletions(-)
>>>>>>
>>>>>> diff --git a/arch/x86/kernel/cpu/intel_rdt.c 
>>>>>> b/arch/x86/kernel/cpu/intel_rdt.c
>>>>>> index dfc7c48..5af2f8f 100644
>>>>>> --- a/arch/x86/kernel/cpu/intel_rdt.c
>>>>>> +++ b/arch/x86/kernel/cpu/intel_rdt.c
>>>>>> @@ -172,34 +172,28 @@ static unsigned int cbm_idx(struct 
>>>>>> rdt_resource *r, unsigned int closid)
>>>>>>    * is always 20 on hsw server parts. The minimum cache bitmask 
>>>>>> length
>>>>>>    * allowed for HSW server is always 2 bits. Hardcode all of them.
>>>>>>    */
>>>>>> -static inline bool cache_alloc_hsw_probe(void)
>>>>>> +static inline void cache_alloc_hsw_probe(void)
>>>>>>   {
>>>>>> -    if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
>>>>>> -        boot_cpu_data.x86 == 6 &&
>>>>>> -        boot_cpu_data.x86_model == INTEL_FAM6_HASWELL_X) {
>>>>>> -        struct rdt_resource *r  = 
>>>>>> &rdt_resources_all[RDT_RESOURCE_L3];
>>>>>> -        u32 l, h, max_cbm = BIT_MASK(20) - 1;
>>>>>> +    struct rdt_resource *r  = &rdt_resources_all[RDT_RESOURCE_L3];
>>>>>> +    u32 l, h, max_cbm = BIT_MASK(20) - 1;
>>>>>>   -        if (wrmsr_safe(IA32_L3_CBM_BASE, max_cbm, 0))
>>>>>> -            return false;
>>>>>> -        rdmsr(IA32_L3_CBM_BASE, l, h);
>>>>>> -
>>>>>> -        /* If all the bits were set in MSR, return success */
>>>>>> -        if (l != max_cbm)
>>>>>> -            return false;
>>>>>> +    if (wrmsr_safe(IA32_L3_CBM_BASE, max_cbm, 0))
>>>>>> +        return;
>>>>>> +    rdmsr(IA32_L3_CBM_BASE, l, h);
>>>>>>   -        r->num_closid = 4;
>>>>>> -        r->default_ctrl = max_cbm;
>>>>>> -        r->cache.cbm_len = 20;
>>>>>> -        r->cache.shareable_bits = 0xc0000;
>>>>>> -        r->cache.min_cbm_bits = 2;
>>>>>> -        r->alloc_capable = true;
>>>>>> -        r->alloc_enabled = true;
>>>>>> +    /* If all the bits were set in MSR, return success */
>>>>>> +    if (l != max_cbm)
>>>>>> +        return;
>>>>>>   -        return true;
>>>>>> -    }
>>>>>> +    r->num_closid = 4;
>>>>>> +    r->default_ctrl = max_cbm;
>>>>>> +    r->cache.cbm_len = 20;
>>>>>> +    r->cache.shareable_bits = 0xc0000;
>>>>>> +    r->cache.min_cbm_bits = 2;
>>>>>> +    r->alloc_capable = true;
>>>>>> +    r->alloc_enabled = true;
>>>>>>   -    return false;
>>>>>> +    rdt_alloc_capable = true;
>>>>>>   }
>>>>>>     /*
>>>>>> @@ -648,7 +642,7 @@ static __init bool get_rdt_alloc_resources(void)
>>>>>>   {
>>>>>>       bool ret = false;
>>>>>>   -    if (cache_alloc_hsw_probe())
>>>>>> +    if (rdt_alloc_capable)
>>>>>>           return true;
>>>>>>         if (!boot_cpu_has(X86_FEATURE_RDT_A))
>>>>>> @@ -690,8 +684,18 @@ static __init bool get_rdt_mon_resources(void)
>>>>>>       return 
>>>>>> !rdt_get_mon_l3_config(&rdt_resources_all[RDT_RESOURCE_L3]);
>>>>>>   }
>>>>>>   +static __init void rdt_quirks(void)
>>>>>> +{
>>>>>> +    switch (boot_cpu_data.x86_model) {
>>>>>> +    case INTEL_FAM6_HASWELL_X:
>>>>>> +        cache_alloc_hsw_probe();
>>>>>> +        break;
>>>>>> +    }
>>>>>> +}
>>>>>> +
>>>>>>   static __init bool get_rdt_resources(void)
>>>>>>   {
>>>>>> +    rdt_quirks();
>>>>>>       rdt_alloc_capable = get_rdt_alloc_resources();
>>>>>>       rdt_mon_capable = get_rdt_mon_resources();
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>
> 



More information about the linux-yocto mailing list