perf: Sanitize perf directory structure, p1

From: Borislav Petkov <[email protected]>

Hi,

here's v1 ontop of rc3 + tip/master, i.e., hopefully I've picked up all relevant branches and rebased my pile ontop so that they don't break any patches in-flight.

Please apply,
thanks.

Changelog:

v0:

so what better time to do that fun than during the holidays when no one else is producing patches. :-)

Anyway, here's a part one conversion of the AMD perf events facilities. Directory structure looks like this:

arch/x86/events/

|-- amd
|   |-- core.c
|   |-- ibs.c
|   |-- iommu.c
|   |-- iommu.h
|   `-- uncore.c
|-- core.c

`-- Makefile

and I've kept the churn at a mininum. Holler if something's not kosher.

Borislav Petkov (5):
x86/events: Move perf_event.c
x86/events: Move perf_event_amd.c
x86/events: Move perf_event_amd_ibs.c
x86/events: Move perf_event_amd_iommu.*
x86/events: Move perf_event_amd_uncore.c

 arch/x86/Kbuild                                                   | 3 ++-
 arch/x86/events/Makefile                                          | 7 +++++++
 arch/x86/{kernel/cpu/perf_event_amd.c => events/amd/core.c}       | 2 +-
 arch/x86/{kernel/cpu/perf_event_amd_ibs.c => events/amd/ibs.c}    | 2 +-
 .../x86/{kernel/cpu/perf_event_amd_iommu.c => events/amd/iommu.c} | 4 ++--
 .../x86/{kernel/cpu/perf_event_amd_iommu.h => events/amd/iommu.h} | 0
 .../{kernel/cpu/perf_event_amd_uncore.c => events/amd/uncore.c}   | 0
 arch/x86/{kernel/cpu/perf_event.c => events/core.c}               | 2 +-
 arch/x86/kernel/cpu/Makefile                                      | 8 +-------
 9 files changed, 15 insertions(+), 13 deletions(-)
 create mode 100644 arch/x86/events/Makefile
 rename arch/x86/{kernel/cpu/perf_event_amd.c => events/amd/core.c} (99%)
 rename arch/x86/{kernel/cpu/perf_event_amd_ibs.c => events/amd/ibs.c} (99%)
 rename arch/x86/{kernel/cpu/perf_event_amd_iommu.c => events/amd/iommu.c} (99%)
 rename arch/x86/{kernel/cpu/perf_event_amd_iommu.h => events/amd/iommu.h} (100%)
 rename arch/x86/{kernel/cpu/perf_event_amd_uncore.c => events/amd/uncore.c} (100%)
 rename arch/x86/{kernel/cpu/perf_event.c => events/core.c} (99%)

2.3.5

From: Borislav Petkov <[email protected]>

Also, keep the churn at minimum by adjusting the include "perf_event.h" when each file gets moved.

---
 arch/x86/Kbuild                                     | 3 ++-
 arch/x86/events/Makefile                            | 1 +
 arch/x86/{kernel/cpu/perf_event.c => events/core.c} | 2 +-
 arch/x86/kernel/cpu/Makefile                        | 2 --
 4 files changed, 4 insertions(+), 4 deletions(-)
 create mode 100644 arch/x86/events/Makefile
 rename arch/x86/{kernel/cpu/perf_event.c => events/core.c} (99%)

index 1538562cc720..eb3abf8ac44e 100644
@@ -1,6 +1,7 @@
-
 obj-y += entry/
 
+obj-$(CONFIG_PERF_EVENTS) += events/
+
 obj-$(CONFIG_KVM) += kvm/
 
 # Xen paravirtualization support

new file mode 100644
index 000000000000..3fad3ce1bbab
@@ -0,0 +1 @@
+obj-y			+= core.o

similarity index 99%
rename from arch/x86/kernel/cpu/perf_event.c
rename to arch/x86/events/core.c
index 7402c8182813..90ca601b5d5c 100644
@@ -37,7 +37,7 @@
 #include <asm/desc.h>
 #include <asm/ldt.h>
 
-#include "perf_event.h"
+#include "../kernel/cpu/perf_event.h"
 
 struct x86_pmu x86_pmu __read_mostly;
 

index faa7b5204129..8cbc41b63447 100644
@@ -30,8 +30,6 @@ obj-$(CONFIG_CPU_SUP_CENTAUR)		+= centaur.o
 obj-$(CONFIG_CPU_SUP_TRANSMETA_32)	+= transmeta.o
 obj-$(CONFIG_CPU_SUP_UMC_32)		+= umc.o
 
-obj-$(CONFIG_PERF_EVENTS)		+= perf_event.o
-
 ifdef CONFIG_PERF_EVENTS
 obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd.o perf_event_amd_uncore.o
 ifdef CONFIG_AMD_IOMMU

2.3.5

I have to admit I've been falling behind on my lkml reading, but is there a good reason for moving all these files around?

I'm often using "git blame" to track down when bugs are introduced, and it's a big pain trying to do that across file moves like this. Although that's maybe just due to difficiencies in my git usage skills.

Vince

git blame still tells you which commit modified a particular line. That's not lost accross a file move.

git log stops per default when a file moved, but you can tell it not to do so via "--follow".

Thanks,

tglx

Yeah, as the PMU drivers grew organically the naming scheme became a bit messy:

- for example the uncore drivers are under kernel/cpu/, although they are not bound to CPUs.

- separation between various vendor drivers is not always chrystal clear.

- the 'kernel/' subdirectory is a needless anachronism, we try to move bits out of it when the opportunity presents itself.

- the new naming also better follows the core kernel/events/ perf code.

I presume after this round of patches is in, Boris will send another round to reorganize the Intel PMU drivers as well.

It seems to work reasonably well here:

241771ef016b5 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2008-12-03 10:39:53 +0100    1) /*
cdd6c482c9ff9 arch/x86/kernel/cpu/perf_event.c   (Ingo Molnar            2009-09-21 12:02:48 +0200    2)  * Performance events x86 architecture code
241771ef016b5 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2008-12-03 10:39:53 +0100    3)  *
98144511427c1 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2009-04-29 14:52:50 +0200    4)  *  Copyright (C) 2008 Thomas Gleixner <[email protected]>
98144511427c1 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2009-04-29 14:52:50 +0200    5)  *  Copyright (C) 2008-2009 Red Hat, Inc., Ingo Molnar
98144511427c1 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2009-04-29 14:52:50 +0200    6)  *  Copyright (C) 2009 Jaswinder Singh Rajput
98144511427c1 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2009-04-29 14:52:50 +0200    7)  *  Copyright (C) 2009 Advanced Micro Devices, Inc., Robert Richter
90eec103b96e3 arch/x86/kernel/cpu/perf_event.c   (Peter Zijlstra         2015-11-16 11:08:45 +0100    8)  *  Copyright (C) 2008-2009 Red Hat, Inc., Peter Zijlstra
30dd568c91260 arch/x86/kernel/cpu/perf_counter.c (Markus Metzger         2009-07-21 15:56:48 +0200    9)  *  Copyright (C) 2009 Intel Corporation, <[email protected]>
1da53e023029c arch/x86/kernel/cpu/perf_event.c   (Stephane Eranian       2010-01-18 10:58:01 +0200   10)  *  Copyright (C) 2009 Google, Inc., Stephane Eranian
241771ef016b5 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2008-12-03 10:39:53 +0100   11)  *
241771ef016b5 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2008-12-03 10:39:53 +0100   12)  *  For licencing details see kernel-base/COPYING
241771ef016b5 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2008-12-03 10:39:53 +0100   13)  */
241771ef016b5 arch/x86/kernel/cpu/perf_counter.c (Ingo Molnar            2008-12-03 10:39:53 +0100   14) 

[...]
c48b60538c3ba arch/x86/kernel/cpu/perf_event.c (Vince Weaver 2012-03-01 17:28:14 -0500 84) rdpmcl(hwc->event_base_rdpmc, new_raw_count);

see for example how Git still knows that we had a arch/x86/kernel/cpu/perf_counter.c initial file name.

Or how you did an optimization in c48b60538c3ba when the file was already named cpu/perf_event.c! :-)

Thanks,

Ingo

Yes, he will. He will base them ontop of newest tip/master too.

:-)

Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
-- 

yes, for simple renames it works, but once files start getting "reorganized" and split up it gets tricky. For example trying to track things across the single perf_event.h to the separate perf_event.h files after the UAPI split is always a pain.

But anyway, it sounds like the renames are being done for reasonably good reasons, it's just the tip commit message that came my way didn't have this extra context.

Thanks,

Vince

From: Borislav Petkov <[email protected]>

We distribute those in vendor subdirs, starting with .../events/amd/.

---
 arch/x86/events/Makefile                                    | 2 ++
 arch/x86/{kernel/cpu/perf_event_amd.c => events/amd/core.c} | 2 +-
 arch/x86/kernel/cpu/Makefile                                | 2 +-
 3 files changed, 4 insertions(+), 2 deletions(-)
 rename arch/x86/{kernel/cpu/perf_event_amd.c => events/amd/core.c} (99%)

index 3fad3ce1bbab..e0560b6dd77b 100644
@@ -1 +1,3 @@
 obj-y			+= core.o
+
+obj-$(CONFIG_CPU_SUP_AMD)               += amd/core.o

similarity index 99%
rename from arch/x86/kernel/cpu/perf_event_amd.c
rename to arch/x86/events/amd/core.c
index 58610539b048..51b16583679c 100644
@@ -5,7 +5,7 @@
 #include <linux/slab.h>
 #include <asm/apicdef.h>
 
-#include "perf_event.h"
+#include "../../kernel/cpu/perf_event.h"
 
 static __initconst const u64 amd_hw_cache_event_ids
 				[PERF_COUNT_HW_CACHE_MAX]

index 8cbc41b63447..2ba0e37acb2d 100644
@@ -31,7 +31,7 @@ obj-$(CONFIG_CPU_SUP_TRANSMETA_32)	+= transmeta.o
 obj-$(CONFIG_CPU_SUP_UMC_32)		+= umc.o
 
 ifdef CONFIG_PERF_EVENTS
-obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd.o perf_event_amd_uncore.o
+obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd_uncore.o
 ifdef CONFIG_AMD_IOMMU
 obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd_iommu.o
 endif

2.3.5

From: Borislav Petkov <[email protected]>

---
 arch/x86/events/Makefile                                       | 1 +
 arch/x86/{kernel/cpu/perf_event_amd_ibs.c => events/amd/ibs.c} | 2 +-
 arch/x86/kernel/cpu/Makefile                                   | 2 +-
 3 files changed, 3 insertions(+), 2 deletions(-)
 rename arch/x86/{kernel/cpu/perf_event_amd_ibs.c => events/amd/ibs.c} (99%)

index e0560b6dd77b..88f787350bf0 100644
@@ -1,3 +1,4 @@
 obj-y			+= core.o
 
 obj-$(CONFIG_CPU_SUP_AMD)               += amd/core.o
+obj-$(CONFIG_X86_LOCAL_APIC)            += amd/ibs.o

similarity index 99%
rename from arch/x86/kernel/cpu/perf_event_amd_ibs.c
rename to arch/x86/events/amd/ibs.c
index aa12f9509cfb..a8abd082d932 100644
@@ -14,7 +14,7 @@
 
 #include <asm/apic.h>
 
-#include "perf_event.h"
+#include "../../kernel/cpu/perf_event.h"
 
 static u32 ibs_caps;
 

index 2ba0e37acb2d..b1711586005b 100644
@@ -54,7 +54,7 @@ obj-$(CONFIG_X86_MCE)			+= mcheck/
 obj-$(CONFIG_MTRR)			+= mtrr/
 obj-$(CONFIG_MICROCODE)			+= microcode/
 
-obj-$(CONFIG_X86_LOCAL_APIC)		+= perfctr-watchdog.o perf_event_amd_ibs.o
+obj-$(CONFIG_X86_LOCAL_APIC)		+= perfctr-watchdog.o
 
 obj-$(CONFIG_HYPERVISOR_GUEST)		+= vmware.o hypervisor.o mshyperv.o
 

2.3.5

From: Borislav Petkov <[email protected]>

---
 arch/x86/events/Makefile                                           | 3 +++
 arch/x86/{kernel/cpu/perf_event_amd_iommu.c => events/amd/iommu.c} | 4 ++--
 arch/x86/{kernel/cpu/perf_event_amd_iommu.h => events/amd/iommu.h} | 0
 arch/x86/kernel/cpu/Makefile                                       | 3 ---
 4 files changed, 5 insertions(+), 5 deletions(-)
 rename arch/x86/{kernel/cpu/perf_event_amd_iommu.c => events/amd/iommu.c} (99%)
 rename arch/x86/{kernel/cpu/perf_event_amd_iommu.h => events/amd/iommu.h} (100%)

index 88f787350bf0..838195d90961 100644
@@ -2,3 +2,6 @@ obj-y			+= core.o
 
 obj-$(CONFIG_CPU_SUP_AMD)               += amd/core.o
 obj-$(CONFIG_X86_LOCAL_APIC)            += amd/ibs.o
+ifdef CONFIG_AMD_IOMMU
+obj-$(CONFIG_CPU_SUP_AMD)               += amd/iommu.o
+endif

similarity index 99%
rename from arch/x86/kernel/cpu/perf_event_amd_iommu.c
rename to arch/x86/events/amd/iommu.c
index 97242a9242bd..629bc700eb08 100644
@@ -16,8 +16,8 @@
 #include <linux/cpumask.h>
 #include <linux/slab.h>
 
-#include "perf_event.h"
-#include "perf_event_amd_iommu.h"
+#include "../../kernel/cpu/perf_event.h"
+#include "iommu.h"
 
 #define COUNTER_SHIFT		16
 

similarity index 100%
rename from arch/x86/kernel/cpu/perf_event_amd_iommu.h
rename to arch/x86/events/amd/iommu.h

index b1711586005b..37444ffd714d 100644
@@ -32,9 +32,6 @@ obj-$(CONFIG_CPU_SUP_UMC_32)		+= umc.o
 
 ifdef CONFIG_PERF_EVENTS
 obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd_uncore.o
-ifdef CONFIG_AMD_IOMMU
-obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd_iommu.o
-endif
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_p6.o perf_event_knc.o perf_event_p4.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_intel_lbr.o perf_event_intel_ds.o perf_event_intel.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_intel_rapl.o perf_event_intel_cqm.o

2.3.5

From: Borislav Petkov <[email protected]>

---
 arch/x86/events/Makefile                                             | 2 +-
 arch/x86/{kernel/cpu/perf_event_amd_uncore.c => events/amd/uncore.c} | 0
 arch/x86/kernel/cpu/Makefile                                         | 1 -
 3 files changed, 1 insertion(+), 2 deletions(-)
 rename arch/x86/{kernel/cpu/perf_event_amd_uncore.c => events/amd/uncore.c} (100%)

index 838195d90961..7d1ecff583b0 100644
@@ -1,6 +1,6 @@
 obj-y			+= core.o
 
-obj-$(CONFIG_CPU_SUP_AMD)               += amd/core.o
+obj-$(CONFIG_CPU_SUP_AMD)               += amd/core.o amd/uncore.o
 obj-$(CONFIG_X86_LOCAL_APIC)            += amd/ibs.o
 ifdef CONFIG_AMD_IOMMU
 obj-$(CONFIG_CPU_SUP_AMD)               += amd/iommu.o

similarity index 100%
rename from arch/x86/kernel/cpu/perf_event_amd_uncore.c
rename to arch/x86/events/amd/uncore.c

index 37444ffd714d..d8cf3338a035 100644
@@ -31,7 +31,6 @@ obj-$(CONFIG_CPU_SUP_TRANSMETA_32)	+= transmeta.o
 obj-$(CONFIG_CPU_SUP_UMC_32)		+= umc.o
 
 ifdef CONFIG_PERF_EVENTS
-obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd_uncore.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_p6.o perf_event_knc.o perf_event_p4.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_intel_lbr.o perf_event_intel_ds.o perf_event_intel.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_intel_rapl.o perf_event_intel_cqm.o

2.3.5