- 08 Oct, 2009 4 commits
-
-
Arnaldo Carvalho de Melo authored
This removes the ovelapping of vmlinux addresses with modules, using the ELF section name when using --vmlinux and creating a unique DSO name when using /proc/kallsyms ([kernel].N). This is done by creating multiple 'struct map' instances for address ranges backed by DSOs that have just the symbols for that range and a name that is derived from the ELF section name.o Now it is possible to ask for just the symbols in some particular kernel section: $ perf report -m --vmlinux ../build/tip-recvmmsg/vmlinux \ --dsos [kernel].vsyscall_fn | head -15 52.73% Xorg [.] vread_hpet 18.61% firefox [.] vread_hpet 14.50% npviewer.bin [.] vread_hpet 6.83% compiz [.] vread_hpet 5.73% glxgears [.] vread_hpet 0.63% java [.] vread_hpet 0.30% gnome-terminal [.] vread_hpet 0.23% perf [.] vread_hpet 0.18% xchat [.] vread_hpet $ Now we don't have to first lookup the list of modules and then, if it fails, vmlinux symbols, its just a simple lookup for the map then the symbols, just like for threads. Reports generated using /proc/kallsyms and --vmlinux should provide the same results, modulo the DSO name for sections other than ".text". But they don't right now because things like: ffffffff81011c20-ffffffff81012068 system_call ffffffff81011c30-ffffffff81011c9b system_call_after_swapgs ffffffff81011c9c-ffffffff81011cb6 system_call_fastpath ffffffff81011cb7-ffffffff81011cbb ret_from_sys_call I.e. overlapping symbols, again some ASM special case that we have to fixup. Signed-off-by:Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1254934136-8503-1-git-send-email-acme@redhat.com> Signed-off-by:
Ingo Molnar <mingo@elte.hu>
-
Arnaldo Carvalho de Melo authored
Like printing every symbol created. Signed-off-by:
-
Frederic Weisbecker authored
When we get sched traces that involve a task that was already created before opening the event, we won't have the comm event for it. So if we can't find the comm event for a given thread, we look at the traces that may contain these informations. Before: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | :5124:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | :6244:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | :6245:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | After: ata/1:371 | 0.000 ms | 1 | avg: 3988.693 ms | max: 3988.693 ms | kondemand/1:421 | 0.096 ms | 3 | avg: 345.346 ms | max: 1035.989 ms | kondemand/0:420 | 0.025 ms | 3 | avg: 421.332 ms | max: 964.014 ms | firefox:5124 | 0.103 ms | 5 | avg: 74.082 ms | max: 277.194 ms | npviewer.bin:6244 | 0.691 ms | 9 | avg: 125.655 ms | max: 271.306 ms | firefox:5080 | 0.924 ms | 5 | avg: 53.833 ms | max: 257.828 ms | npviewer.bin:6225 | 21.871 ms | 53 | avg: 22.462 ms | max: 220.835 ms | npviewer.bin:6245 | 9.631 ms | 21 | avg: 41.864 ms | max: 213.349 ms | Signed-off-by:
Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1255012632-7882-1-git-send-email-fweisbec@gmail.com> Signed-off-by:
-
Frederic Weisbecker authored
This librarizes the perf.data file mapping and handling in various perf tools, roughly reducing the amount of code and fixing the places that mmap from beginning of the file whereas we want to mmap from the beginning of the data, leading to page fault because the mmap window is too small since the trace info are written in the file too. TODO: - convert perf timechart too Signed-off-by:
-
- 07 Oct, 2009 2 commits
-
-
Frederic Weisbecker authored
This drops the trace.info file and move its contents into the common perf.data file. This is done by creating a new trace_info section into this file. A user of perf headers needs to call perf_header__set_trace_info() to save the trace meta informations into the perf.data file. A file created by perf after his patch is unsupported by previous version because the size of the headers have increased. That said, it's two new fields that have been added in the end of the headers, and those could be ignored by previous versions if they just handled the dynamic header size and then ignore the unknow part. The offsets guarantee the compatibility. We'll do a -stable fix for that. But current previous versions handle the header size using its static size, not dynamic, then it's not backward compatible with trace records. Signed-off-by:
-
Frederic Weisbecker authored
Currently, we are mapping perf.data in the beginning of the file and use the data offset as a buffer offset. This may exceed the mapping area if the data offset is upper than page_size * mmap_window and result in a page fault (thing that happen if we merge trace.info in perf.data). Instead, let's start the mapping in the page that matches our data offset. v2: Drop a junk from another patch (trace_report() removal) Signed-off-by:
-
- 06 Oct, 2009 11 commits
-
-
Ingo Molnar authored
Use auto-freq events by default in perf record and perf top. This allows more consistent hardware event sampling, regardless of the intensity of the underlying event. It also keeps us from over-sampling on larger/busier systems. (also make surrounding initializations more consistent) Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: -
Tom Zanussi authored
Needed for distinguishing string fields in event stream processing. Signed-off-by:
Tom Zanussi <tzanussi@gmail.com> Acked-by:
-
Tom Zanussi authored
Needed to fully qualify event names for event stream processing. Signed-off-by:
-
Tom Zanussi authored
The sign info used for filters in the kernel is also useful to applications that process the trace stream. Add it to the format files and make it available to userspace. Signed-off-by:
-
Ingo Molnar authored
Merge reason: Upcoming patch is dependent on a fix in perf/urgent. Signed-off-by: -
Peter Zijlstra authored
Some architectures such as Sparc, ARM and MIPS (basically everything with flush_dcache_page()) need to deal with dcache aliases by carefully placing pages in both kernel and user maps. These architectures typically have to use vmalloc_user() for this. However, on other architectures, vmalloc() is not needed and has the downsides of being more restricted and slower than regular allocations. Signed-off-by:
Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by:
David Miller <davem@davemloft.net> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1254830228.21044.272.camel@laptop> Signed-off-by:
-
Chuck Ebbert authored
Signed-off-by:
Chuck Ebbert <cebbert@redhat.com> LKML-Reference: <200910061138.n96BcqkJ004709@int-mx03.intmail.prod.int.phx2.redhat.com> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Asm routines that end up having size equal to zero are not really zero sized, and as now we do kernel_maps__fixup_sym_end, at least for kernel routines this gets fixed. A similar fixup needs to be done for the userspace bits as well, but as this fixup started only because in /proc/kallsyms we don't have the end address nor the function size, it appeared here first. Signed-off-by:
-
Tom Zanussi authored
The syscall event definitions use long for the syscall exit ret value, but unsigned long for the same thing in the format and field definitions. Change them all to long. Signed-off-by:
-
Tom Zanussi authored
Add missing BLOCK_IOPOLL_SOFTIRQ entry. Signed-off-by:
-
Tom Zanussi authored
And some minor whitespace cleanup. Signed-off-by:
-
- 05 Oct, 2009 4 commits
-
-
Arnaldo Carvalho de Melo authored
In resolve_symbol, as we're moving to breaking the kernel symbols list per address ranges, i.e. kernel linking sections, so that we don't have a big kernel_map that in its range covers what is in the modules. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
$ cut -d' ' -f1 /proc/modules|grep _|wc -l 29 $ cut -d' ' -f1 /proc/modules|grep _|sed 's/$/.ko'/g|while read n;do find /lib/modules/`uname -r` -name $n;done|wc -l 12 For instance: $ grep ^aes_x86 /proc/modules aes_x86_64 9056 2 - Live 0xffffffffa0091000 $ l /lib/modules/2.6.31-tip/kernel/arch/x86/crypto/aes-x86_64.ko -rw-r--r-- 1 root root 136438 2009-09-22 19:05 /lib/modules/2.6.31-tip/kernel/arch/x86/crypto/aes-x86_64.ko Handle that by introducing a strxfrchar routine that replaces dashes with underscores when matching file names to loaded modules. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
So that we get kallsyms processing closer to vmlinux + modules symtabs processing. One change in behaviour is that since when one specifies --vmlinux -m should be used to ask for modules, so it is now for kallsyms as well. Also continue if one manages to load the vmlinux data but module processing fails, so that at least some analisys can be done with part of the needed symbols. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Signed-off-by:
-
- 04 Oct, 2009 2 commits
-
-
Chris Wilson authored
If we launch the child on behalf of the user, ensure that it dies along with ourselves when we are interrupted. Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Cc: Chris Wilson <chris@chris-wilson.co.uk> LKML-Reference: <1254616502-4728-1-git-send-email-chris@chris-wilson.co.uk> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
As it was not being exposed via any command line and with --dsos/--comms we can do this and even more, like asking for just kernel + some module: [root@doppio linux-2.6-tip]# perf report --dsos \[kernel\],\[drm\] --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules | head -15 # Samples: 619669 # # Overhead Command Shared Object Symbol # ........ ............... ............. ...... # 7.12% swapper [kernel] [k] read_hpet 6.86% init [kernel] [k] read_hpet 6.22% init [kernel] [k] mwait_idle_with_hints 5.34% swapper [kernel] [k] mwait_idle_with_hints 3.01% firefox [kernel] [.] vread_hpet 2.14% Xorg [drm] [k] drm_clflush_pages 2.09% pidgin [kernel] [.] vread_hpet 1.58% npviewer.bin [kernel] [.] vread_hpet 1.37% swapper [kernel] [k] hpet_next_event 1.23% Xorg [kernel] [k] read_hpet [root@doppio linux-2.6-tip]# Signed-off-by:
-
- 03 Oct, 2009 1 commit
-
-
Arnaldo Carvalho de Melo authored
Now perf report and annotate do the callgraph/hit processing in their specialized hist_entry__add functions. Signed-off-by:
-
- 02 Oct, 2009 1 commit
-
-
Arnaldo Carvalho de Melo authored
Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by:
-
- 01 Oct, 2009 5 commits
-
-
Mulyadi Santosa authored
Right now generate-cmdlist.sh is not executable, so we should call it as an argument ".". This fixes cases where due to different umask defaults the generate-cmdlist.sh script is not executable in a kernel tree checkout. Signed-off-by:
Mulyadi Santosa <mulyadi.santosa@gmail.com> Acked-by:
Sam Ravnborg <sam@ravnborg.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <f284c33d0909251201w422e9687x8cd3a784e85adf7d@mail.gmail.com> Signed-off-by:
-
Xiao Guangrong authored
While at it: we can traverse ctx->group_list to get all group leader, it should be safe since we hold ctx->mutex. Changlog v1->v2: - remove WARN_ON_ONCE() according to Peter Zijlstra's suggestion Signed-off-by:
Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Acked-by:
Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <4ABC5AF9.6060808@cn.fujitsu.com> Signed-off-by:
-
Xiao Guangrong authored
Paul Mackerras says: "Actually, looking at this more closely, it has to be a group leader anyway since it's at the top level of ctx->group_list. In fact I see four places where we do: list_for_each_entry(event, &ctx->group_list, group_entry) { if (event == event->group_leader) ... or the equivalent, three of which appear to have been introduced by afedadf2 ("perf_counter: Optimize sched in/out of counters") back in May by Peter Z. As far as I can see the if () is superfluous in each case (a singleton event will be a group of 1 and will have its group_leader pointing to itself)." [ See: http://marc.info/?l=linux-kernel&m=125361238901442&w=2 ] And Peter Zijlstra points out this is a bugfix: "The intent was to call event_sched_{in,out}() for single event groups because that's cheaper than group_sched_{in,out}(), however.. - as you noticed, I got the condition wrong, it should have read: list_empty(&event->sibling_list) - it failed to call group_can_go_on() which deals with ->exclusive. - it also doesn't call hw_perf_group_sched_in() which might break power." [ See: http://marc.info/?l=linux-kernel&m=125369523318583&w=2 ] Changelog v1->v2: - Fix the title name according to Peter Zijlstra's suggestion - Remove the comments and WARN_ON_ONCE() as Peter Zijlstra's suggestion Signed-off-by: -
Arjan van de Ven authored
For doing work on the Linux power management components, I need to make long (30+ seconds) traces. Currently, this then results in a HUGE svg file, with mostly process data that isn't interesting. This patch adds a --power-only mode to perf timechart that only outputs the CPU power section of the SVG; this significantly reduces the size of the SVG file, making even 30+ second traces viewable with inkscape. As a minor tweak for the same effect, the minimum text size is decreased; current inkscape cannot zoom in deep enough to show text this small, but it reduces inkscape compute time. Signed-off-by:
Arjan van de Ven <arjan@linux.intel.com> Cc: peterz@infradead.org LKML-Reference: <20090924154013.0675ab71@infradead.org> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Signed-off-by:
-
- 30 Sep, 2009 5 commits
-
-
Arnaldo Carvalho de Melo authored
Several variables are not used at all, cut'n'paste leftovers. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Several variables are not used at all, cut'n'paste leftovers. Also check if the sample_type is RAW earlier, to avoid needless searches. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: -
John Kacur authored
Move histogram related functions into their own files (hist.c and hist.h) and make use of them in builtin-annotate.c and builtin-report.c. Signed-off-by:
John Kacur <jkacur@redhat.com> Acked-by:
-
Arnaldo Carvalho de Melo authored
Signed-off-by:
Eric Dumazet <eric.dumazet@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <20090925220239.GA5488@ghostprotocols.net> Signed-off-by:
-
- 27 Sep, 2009 5 commits
-
-
Linus Torvalds authored
-
Linus Torvalds authored
.. duplicated by merging the same fix twice, for details see commit 0d9df251 ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild-fixes") Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org>
-
Dave Young authored
The following commit made console open fails while booting: commit b50989dc Author: Alan Cox <alan@linux.intel.com> Date: Sat Sep 19 13:13:22 2009 -0700 tty: make the kref destructor occur asynchronously Due to tty release routines run in a workqueue now, error like the following will be reported while booting: INIT open /dev/console Input/output error It also causes hibernation regression to appear as reported at http://bugzilla.kernel.org/show_bug.cgi?id=14229 The reason is that now there's latency issue with closing, but when we open a "closing not finished" tty, -EIO will be returned. Fix it as per the following Alan's suggestion: Fun but it's actually not a bug and the fix is wrong in itself as the port may be closing but not yet being destructed, in which case it seems to do the wrong thing. Opening a tty that is closing (and could be closing for long periods) is supposed to return -EIO. I suspect a better way to deal with this and keep the old console timing is to split tty->shutdown into two functions. tty->shutdown() - called synchronously just before we dump the tty onto the waitqueue for destruction tty->cleanup() - called when the destructor runs. We would then do the shutdown part which can occur in IRQ context fine, before queueing the rest of the release (from tty->magic = 0 ... the end) to occur asynchronously The USB update in -next would then need a call like if (tty->cleanup) tty->cleanup(tty); at the top of the async function and the USB shutdown to be split between shutdown and cleanup as the USB resource cleanup and final tidy cannot occur synchronously as it needs to sleep. In other words the logic becomes final kref put make object unfindable async clean it up Signed-off-by:
Dave Young <hidave.darkstar@gmail.com> [ rjw: Rebased on top of 2.6.31-git, reworked the changelog. ] Signed-off-by:
"Rafael J. Wysocki" <rjw@sisk.pl> [ Changed serial naming to match new rules, dropped tty_shutdown as per comments from Alan Stern - Linus ] Signed-off-by:
-
Linus Torvalds authored
Commit 3d5b6fb4 ("ACPI: Kill overly verbose "power state" log messages") removed the actual use of this variable, but didn't remove the variable itself, resulting in build warnings like drivers/acpi/processor_idle.c: In function ‘acpi_processor_power_init’: drivers/acpi/processor_idle.c:1169: warning: unused variable ‘i’ Just get rid of the now unused variable. Signed-off-by:
-
Alexey Dobriyan authored
* mark struct vm_area_struct::vm_ops as const * mark vm_ops in AGP code But leave TTM code alone, something is fishy there with global vm_ops being used. Signed-off-by:
Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by:
-
