Description
-----------
Change ppc64 oprofile kernel driver to use the SLOT bits (MMCRA[37:39]only on
older processors where those bits are defined.
Background
----------
The performance monitor unit of the 64-bit POWER processor family has the
ability to collect accurate instruction-level samples when profiling on marked
events (i.e., "PM_MRK_<event-name>"). In processors prior to POWER6, the MMCRA
register contained "slot information" that the oprofile kernel driver used to
adjust the value latched in the SIAR at the time of a PMU interrupt. But as of
POWER6, these slot bits in MMCRA are no longer necessary for oprofile to use,
since the SIAR itself holds the accurate sampled instruction address. With
POWER6, these MMCRA slot bits were zero'ed out by hardware so oprofile's use of
these slot bits was, in effect, a NOP. But with POWER7, these bits are no
longer zero'ed out; however, they serve some other purpose rather than slot
information. Thus, using these bits on POWER7 to adjust the SIAR value results
in samples being attributed to the wrong instructions. The attached patch
changes the oprofile kernel driver to ignore these slot bits on all newer
processors starting with POWER6.
Signed-off-by: Maynard Johnson <maynardj@us.ibm.com>
Signed-off-by: Michael Wolf <mjw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
For ppc750 processors which use 4 performance counters instead of the
6 G4 uses but otherwise is compatible with G4.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/oprofile/cell/spu_profiler.c is missing a asm/time.h
include which is required for ppc_proc_freq. This can cause compile
failures for some config combinations.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Convert arch/powerpc/ over to long long based u64:
-#ifdef __powerpc64__
-# include <asm-generic/int-l64.h>
-#else
-# include <asm-generic/int-ll64.h>
-#endif
+#include <asm-generic/int-ll64.h>
This will avoid reoccuring spurious warnings in core kernel code that
comes when people test on their own hardware. (i.e. x86 in ~98% of the
cases) This is what x86 uses and it generally helps keep 64-bit code
32-bit clean too.
[Adjusted to not impact user mode (from paulus) - sfr]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile: (31 commits)
powerpc/oprofile: fix whitespaces in op_model_cell.c
powerpc/oprofile: IBM CELL: add SPU event profiling support
powerpc/oprofile: fix cell/pr_util.h
powerpc/oprofile: IBM CELL: cleanup and restructuring
oprofile: make new cpu buffer functions part of the api
oprofile: remove #ifdef CONFIG_OPROFILE_IBS in non-ibs code
ring_buffer: fix ring_buffer_event_length()
oprofile: use new data sample format for ibs
oprofile: add op_cpu_buffer_get_data()
oprofile: add op_cpu_buffer_add_data()
oprofile: rework implementation of cpu buffer events
oprofile: modify op_cpu_buffer_read_entry()
oprofile: add op_cpu_buffer_write_reserve()
oprofile: rename variables in add_ibs_begin()
oprofile: rename add_sample() in cpu_buffer.c
oprofile: rename variable ibs_allowed to has_ibs in op_model_amd.c
oprofile: making add_sample_entry() inline
oprofile: remove backtrace code for ibs
oprofile: remove unused ibs macro
oprofile: remove unused components in struct oprofile_cpu_buffer
...
This patch adds the SPU event based profiling funcitonality for the
IBM Cell processor. Previously, the CELL OProfile kernel code supported
PPU event, PPU cycle profiling and SPU cycle profiling. The addition of
SPU event profiling allows the users to identify where in their SPU code
various SPU evnets are occuring. This should help users further identify
issues with their code. Note, SPU profiling has some limitations due to HW
constraints. Only one event at a time can be used for profiling and SPU event
profiling must be time sliced across all of the SPUs in a node.
The patch adds a new arch specific file to the OProfile file system. The
file has bit 0 set to indicate that the kernel supports SPU event profiling.
The user tool must check this file/bit to make sure the kernel supports
SPU event profiling before trying to do SPU event profiling. The user tool
check is part of the user tool patch for SPU event profiling.
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Robert Richter <robert.richter@amd.com>
This patch restructures and cleans up the code a bit to make it
easier to add new functionality later. The patch makes no
functional changes to the existing code.
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Robert Richter <robert.richter@amd.com>
struct dentry is one of the most critical structures in the kernel. So it's
sad to see it going neglected.
With CONFIG_PROFILING turned on (which is probably the common case at least
for distros and kernel developers), sizeof(struct dcache) == 208 here
(64-bit). This gives 19 objects per slab.
I packed d_mounted into a hole, and took another 4 bytes off the inline
name length to take the padding out from the end of the structure. This
shinks it to 200 bytes. I could have gone the other way and increased the
length to 40, but I'm aiming for a magic number, read on...
I then got rid of the d_cookie pointer. This shrinks it to 192 bytes. Rant:
why was this ever a good idea? The cookie system should increase its hash
size or use a tree or something if lookups are a problem. Also the "fast
dcookie lookups" in oprofile should be moved into the dcookie code -- how
can oprofile possibly care about the dcookie_mutex? It gets dropped after
get_dcookie() returns so it can't be providing any sort of protection.
At 192 bytes, 21 objects fit into a 4K page, saving about 3MB on my system
with ~140 000 entries allocated. 192 is also a multiple of 64, so we get
nice cacheline alignment on 64 and 32 byte line systems -- any given dentry
will now require 3 cachelines to touch all fields wheras previously it
would require 4.
I know the inline name size was chosen quite carefully, however with the
reduction in cacheline footprint, it should actually be just about as fast
to do a name lookup for a 36 character name as it was before the patch (and
faster for other sizes). The memory footprint savings for names which are
<= 32 or > 36 bytes long should more than make up for the memory cost for
33-36 byte names.
Performance is a feature...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The size of the pm_signal_local array should be equal to the
number of SPUs being configured in the array. Currently, the
array is of size 4 (NR_PHYS_CTRS) but being indexed by a for
loop from 0 to 7 (NUM_SPUS_PER_NODE). This could potentially
cause an oops or random memory corruption since the pm_signal_local
array is on the stack. This fixes it.
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
* 'v28-range-hrtimers-for-linus-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (37 commits)
hrtimers: add missing docbook comments to struct hrtimer
hrtimers: simplify hrtimer_peek_ahead_timers()
hrtimers: fix docbook comments
DECLARE_PER_CPU needs linux/percpu.h
hrtimers: fix typo
rangetimers: fix the bug reported by Ingo for real
rangetimer: fix BUG_ON reported by Ingo
rangetimer: fix x86 build failure for the !HRTIMERS case
select: fix alpha OSF wrapper
select: fix alpha OSF wrapper
hrtimer: peek at the timer queue just before going idle
hrtimer: make the futex() system call use the per process slack value
hrtimer: make the nanosleep() syscall use the per process slack
hrtimer: fix signed/unsigned bug in slack estimator
hrtimer: show the timer ranges in /proc/timer_list
hrtimer: incorporate feedback from Peter Zijlstra
hrtimer: add a hrtimer_start_range() function
hrtimer: another build fix
hrtimer: fix build bug found by Ingo
hrtimer: make select() and poll() use the hrtimer range feature
...
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile: (21 commits)
OProfile: Fix buffer synchronization for IBS
oprofile: hotplug cpu fix
oprofile: fixing whitespaces in arch/x86/oprofile/*
oprofile: fixing whitespaces in arch/x86/oprofile/*
oprofile: fixing whitespaces in drivers/oprofile/*
x86/oprofile: add the logic for enabling additional IBS bits
x86/oprofile: reordering functions in nmi_int.c
x86/oprofile: removing unused function parameter in add_ibs_begin()
oprofile: more whitespace fixes
oprofile: whitespace fixes
OProfile: Rename IBS sysfs dir into "ibs_op"
OProfile: Rework string handling in setup_ibs_files()
OProfile: Rework oprofile_add_ibs_sample() function
oprofile: discover counters for op ppro too
oprofile: Implement Intel architectural perfmon support
oprofile: Don't report Nehalem as core_2
oprofile: drop const in num counters field
Revert "Oprofile Multiplexing Patch"
x86, oprofile: BUG: using smp_processor_id() in preemptible code
x86/oprofile: fix on_each_cpu build error
...
Manually fixed trivial conflicts in
drivers/oprofile/{cpu_buffer.c,event_buffer.h}
The issue is the SPU code is not holding the kernel mutex lock while
adding samples to the kernel buffer.
This patch creates per SPU buffers to hold the data. Data
is added to the buffers from in interrupt context. The data
is periodically pushed to the kernel buffer via a new Oprofile
function oprofile_put_buff(). The oprofile_put_buff() function
is called via a work queue enabling the funtion to acquire the
mutex lock.
The existing user controls for adjusting the per CPU buffer
size is used to control the size of the per SPU buffers.
Similarly, overflows of the SPU buffers are reported by
incrementing the per CPU buffer stats. This eliminates the
need to have architecture specific controls for the per SPU
buffers which is not acceptable to the OProfile user tool
maintainer.
The export of the oprofile add_event_entry() is removed as it
is no longer needed given this patch.
Note, this patch has not addressed the issue of indexing arrays
by the spu number. This still needs to be fixed as the spu
numbering is not guarenteed to be 0 to max_num_spus-1.
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Maynard Johnson <maynardj@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Acked-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Offset is unsigned and when an address isn't found in the vma map
vma_map_lookup() returns the vma physical address + 0x10000000.
vma_map_lookup used to return 0xffffffff on a failed lookup, but
a change was made to return the vma physical address + 0x10000000
There are two callers of vam_map_lookup: one of them correctly
deals with this new return value, but the other (below) did not.
Signed-off-by: Roel Kluin <12o3l@tiscali.nl>
Acked-by: Maynard Johnson <maynardj@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In order to be able to do range hrtimers we need to use accessor functions
to the "expire" member of the hrtimer struct.
This patch converts powerpc/oprofile to these accessors.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
The file arch/powerpc/kernel/sysfs.c is currently only compiled for
64-bit kernels. It contain code to register CPU sysdevs in sysfs and
add various properties such as cache topology and raw access by root
to performance monitor counters (PMCs). A lot of that can be re-used
as is on 32-bits.
This makes the file be built for both, with appropriate ifdef'ing
for the few bits that are really 64-bit specific, and adds some
support for the raw PMCs for 75x and 74xx processors.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
It's not even passed on to smp_call_function() anymore, since that
was removed. So kill it.
Acked-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
__FUNCTION__ is gcc-specific, use __func__
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch enables OProfile callgraph support for the Cell processor. The
original code was just calling a function to add the PC value, now it will
call a function that first checks the callgraph depth. Callgraph is already
enabled on the other Power platforms.
Signed-off-by: Bob Nelson <rrnelson@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
get_dcookie() is always called with a dentry and a vfsmount from a struct
path. Make get_dcookie() take it directly as an argument.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Jan Blunck <jblunck@suse.de>
Acked-by: Christoph Hellwig <hch@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Some of the more recent e300 cores have the same performance monitor
implementation as the e500. e300 isn't book-e, so the name isn't
really appropriate. In preparation for e300 support, rename a bunch
of fsl_booke things to say fsl_emb (Freescale Embedded Performance Monitors).
Signed-off-by: Andy Fleming <afleming@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Remove the deprecated __attribute_used__.
[Introduce __section in a few places to silence checkpatch /sam]
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Fix debug_bus_control and group_control PMU register values set up in
set_pm_event(). Initialize variables before calling set_pm_event().
Delete unused static array and code that initialized it.
Rename constant to better reflect usage.
Signed-off-by: Bob Nelson <rrnelson@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Quoting Randy:
"It seems sad that this patch sources Kconfig.marker, a 7-line file,
20-something times. Yes, you (we) don't want to put those 7 lines into
20-something different files, so sourcing is the right thing.
However, what you did for avr32 seems more on the right track to me: make
_one_ Instrumentation support menu that includes PROFILING, OPROFILE, KPROBES,
and MARKERS and then use (source) that in all of the arches."
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The new Cell EDAC driver needs that file, oprofile also does ugly
path tricks to get to it, it's time to move it to asm-powerpc. While
at it, rename it to be consistent with cell-pmu.h (and dashes look
nicer than underscores anyway).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
From: Maynard Johnson <mpjohn@us.ibm.com>
This patch updates the existing arch/powerpc/oprofile/op_model_cell.c
to add in the SPU profiling capabilities. In addition, a 'cell' subdirectory
was added to arch/powerpc/oprofile to hold Cell-specific SPU profiling code.
Exports spu_set_profile_private_kref and spu_get_profile_private_kref which
are used by OProfile to store private profile information in spufs data
structures.
Also incorporated several fixes from other patches (rrn). Check pointer
returned from kzalloc. Eliminated unnecessary cast. Better error
handling and cleanup in the related area. 64-bit unsigned long parameter
was being demoted to 32-bit unsigned int and eventually promoted back to
unsigned long.
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Maynard Johnson <mpjohn@us.ibm.com>
Signed-off-by: Bob Nelson <rrnelson@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Oprofile enhanced instruction sampling support.
When performing instruction sampling, the mmcra[SLOT] field can be used to
more accurately identify the address of the sampled instruction.
Tested on power4, js20, power5 and power5+.
Signed-off-by: Will Schmidt <will_schmidt@vnet.ibm.com>
cc: Maynard Johnson <maynardj@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Fix the misspellings of "propogate", "writting" and (oh, the shame
:-) "kenrel" in the source tree.
Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Use DEFINE_SPINLOCK instead of initializing spinlocks to
SPIN_LOCK_UNLOCKED, since DEFINE_SPINLOCK is better for lockdep.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Oprofile support for PA6T, kernel side.
Also rename the PA6T_SPRN.* defines to SPRN_PA6T.*.
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch adds code to deal with conversion of
logical cpu to cbe nodes. It removes code that
assummed there were two logical CPUs per CBE.
Signed-off-by: Christian Krafft <krafft@de.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
The code was setting up the debug bus for group 21 when profiling on the
event PPU CYCLES. The debug bus is not actually used by the hardware
performance counters when counting PPU CYCLES. Setting up the debug bus
for PPU CYCLES causes signal routing conflicts on the debug bus when
profiling PPU cycles and another PPU event. This patch fixes the code to
only setup the debug bus to route the performance signals for the non
PPU CYCLE events.
Signed-off-by: Maynard Johnson <mpjohn@us.ibm.com>
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
This is a clean up patch that includes the following changes:
-Some comments were added to clarify the code based on feedback
from the community.
-The write_pm_cntrl() and set_count_mode() were passed a
structure element from a global variable. The argument was
removed so the functions now just operate on the global directly.
-The set_pm_event() function call in the cell_virtual_cntr()
routine was moved to a for-loop before the for_each_cpu loop
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Maynard Johnson <mpjohn@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Clean up the ctr_read/write a bit. It's currently defined in the
include but only used in one C file each. The only exception is the
classic version, so keep that in the include and define in the C file
as appropriate.
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
A few code paths need to check whether or not they are running
on the PS3's LV1 hypervisor before making hcalls. This introduces
a new firmware feature bit for this, FW_FEATURE_PS3_LV1.
Now when both PS3 and IBM_CELL_BLADE are enabled, but not PSERIES,
FW_FEATURE_PS3_LV1 and FW_FEATURE_LPAR get enabled at compile time,
which is a bug. The same problem can also happen for (PPC_ISERIES &&
!PPC_PSERIES && PPC_SOMETHING_ELSE). In order to solve this, I
introduce a new CONFIG_PPC_NATIVE option that is set when at least
one platform is selected that can run without a hypervisor and then
turns the firmware feature check into a run-time option.
The new cell oprofile support that was recently merged does not
work on hypervisor based platforms like the PS3, therefore make
it depend on PPC_CELL_NATIVE instead of PPC_CELL. This may change
if we get oprofile support for PS3.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Add PPU event-based and cycle-based profiling support to Oprofile for Cell.
Oprofile is expected to collect data on all CPUs simultaneously.
However, there is one set of performance counters per node. There are
two hardware threads or virtual CPUs on each node. Hence, OProfile must
multiplex in time the performance counter collection on the two virtual
CPUs.
The multiplexing of the performance counters is done by a virtual
counter routine. Initially, the counters are configured to collect data
on the even CPUs in the system, one CPU per node. In order to capture
the PC for the virtual CPU when the performance counter interrupt occurs
(the specified number of events between samples has occurred), the even
processors are configured to handle the performance counter interrupts
for their node. The virtual counter routine is called via a kernel
timer after the virtual sample time. The routine stops the counters,
saves the current counts, loads the last counts for the other virtual
CPU on the node, sets interrupts to be handled by the other virtual CPU
and restarts the counters, the virtual timer routine is scheduled to run
again. The virtual sample time is kept relatively small to make sure
sampling occurs on both CPUs on the node with a relatively small
granularity. Whenever the counters overflow, the performance counter
interrupt is called to collect the PC for the CPU where data is being
collected.
The oprofile driver relies on a firmware RTAS call to setup the debug bus
to route the desired signals to the performance counter hardware to be
counted. The RTAS call must set the routing registers appropriately in
each of the islands to pass the signals down the debug bus as well as
routing the signals from a particular island onto the bus. There is a
second firmware RTAS call to reset the debug bus to the non pass thru
state when the counters are not in use.
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Maynard Johnson <mpjohn@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Fixed a compile error in building the 85xx support with oprofile, and in
the process cleaned up some issues with the fsl_booke performance monitor
code.
* Reorganized FSL Book-E performance monitoring code so that the 7450
wouldn't be built if the e500 was, and cleaned it up so it was more
self-contained.
* Added a cpu_setup function for FSL Book-E. The original
cpu_setup function prototype had no arguments, assuming that
the reg_setup function would copy the required information into
variables which represented the registers. This was silly for
e500, since it has 1 register per counter (rather than 3 for
all counters), so the code has been restructured to have
cpu_setup take the current counter config array as an argument,
with op_powerpc_setup() invoking op_powerpc_cpu_setup() through
on_each_cpu(), and op_powerpc_cpu_setup() invoking the
model-specific cpu_setup function with an argument. The
argument is ignored on all other platforms at present.
* Fixed a confusing line where a trinary operator only had two
arguments
Signed-off-by: Andrew Fleming <afleming@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Many files include the filename at the beginning, serveral used a wrong one.
Signed-off-by: Uwe Zeisberger <Uwe_Zeisberger@digi.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
