]> sourceware.org Git - systemtap.git/log
systemtap.git
12 years agoFix PR13641 by not calling a sleeping function from an invalid context.
David Smith [Thu, 2 Feb 2012 16:36:44 +0000 (10:36 -0600)]
Fix PR13641 by not calling a sleeping function from an invalid context.

* testsuite/systemtap.examples/process/pfiles.stp: To avoid the
  possibility of sleeping while holding a lock, change GFP_KERNEL
  allocation to GFP_ATOMIC allocation.  Also convert i_mode2str() from an
  embedded-C function.

12 years agoMake stopwatches.stp example executable.
Mark Wielaard [Thu, 2 Feb 2012 08:39:41 +0000 (09:39 +0100)]
Make stopwatches.stp example executable.

12 years agopost-release version bump to 1.8
Frank Ch. Eigler [Thu, 2 Feb 2012 04:12:17 +0000 (23:12 -0500)]
post-release version bump to 1.8

12 years agorelease 1.7 prep release-1.7
Frank Ch. Eigler [Thu, 2 Feb 2012 02:25:30 +0000 (21:25 -0500)]
release 1.7 prep

12 years agoi18n: update-po
Frank Ch. Eigler [Wed, 1 Feb 2012 18:29:45 +0000 (13:29 -0500)]
i18n: update-po

12 years agoMove warnings for embedded privilege credentials in modules to -v.
Dave Brolley [Wed, 1 Feb 2012 17:48:45 +0000 (12:48 -0500)]
Move warnings for embedded privilege credentials in modules to -v.

12 years agoGive information about online servers and which servers were specified when no suitab...
Dave Brolley [Wed, 1 Feb 2012 16:58:34 +0000 (11:58 -0500)]
Give information about online servers and which servers were specified when no suitable servers are found.

12 years agoruntime: document reason runtime/staprun* are compiled with -Wno-format-nonliteral
Frank Ch. Eigler [Wed, 1 Feb 2012 17:19:03 +0000 (12:19 -0500)]
runtime: document reason runtime/staprun* are compiled with -Wno-format-nonliteral

12 years agoModiified eventcount.stp example to wrap the map
Chris Meek [Wed, 1 Feb 2012 17:05:58 +0000 (12:05 -0500)]
Modiified eventcount.stp example to wrap the map

general/eventcount.stp example now has a sample of map wrapping.

12 years agobump copyright date on startup message
Frank Ch. Eigler [Wed, 1 Feb 2012 16:05:44 +0000 (11:05 -0500)]
bump copyright date on startup message

12 years agoMerge branch 'mjw/create_atomic'
Mark Wielaard [Wed, 1 Feb 2012 13:08:14 +0000 (14:08 +0100)]
Merge branch 'mjw/create_atomic'

12 years agoPR13640: don't start stap-server service by default
Frank Ch. Eigler [Wed, 1 Feb 2012 12:42:52 +0000 (07:42 -0500)]
PR13640: don't start stap-server service by default

12 years agotestsuite: make #! .stp files executable to make rpmlint happy
Frank Ch. Eigler [Wed, 1 Feb 2012 12:40:42 +0000 (07:40 -0500)]
testsuite: make #! .stp files executable to make rpmlint happy

12 years agoPR13461: systemtap.spec: reorganize subpackages, new -client, -devel
Frank Ch. Eigler [Tue, 31 Jan 2012 21:12:43 +0000 (16:12 -0500)]
PR13461: systemtap.spec: reorganize subpackages, new -client, -devel

12 years agoutrace_compatibility.h: Fix typo/thinko in UTRACE_ATTACH_ATOMIC comment.
Mark Wielaard [Wed, 1 Feb 2012 12:25:13 +0000 (13:25 +0100)]
utrace_compatibility.h: Fix typo/thinko in UTRACE_ATTACH_ATOMIC comment.

s/non-interrupt/interrupt/ context.

12 years agotapset/task.stp (pid2task) Extend scope of rcu_read_lock to include pid2task.
Mark Wielaard [Wed, 1 Feb 2012 11:58:39 +0000 (12:58 +0100)]
tapset/task.stp (pid2task) Extend scope of rcu_read_lock to include pid2task.

12 years agoitrace.c: rcu_dereference needs to be called under rcu_read_lock.
Mark Wielaard [Wed, 1 Feb 2012 09:53:25 +0000 (10:53 +0100)]
itrace.c: rcu_dereference needs to be called under rcu_read_lock.

12 years agoUse UTRACE_ATTACH_ATOMIC when calling utrace create under lock.
Mark Wielaard [Tue, 31 Jan 2012 22:06:30 +0000 (23:06 +0100)]
Use UTRACE_ATTACH_ATOMIC when calling utrace create under lock.

UTRACE_ATTACH_ATOMIC is very new, but we need it when calling
utrace create while holding a lock or in interrupt context.
If it doesn't exist we do as before (define it as zero).

See http://sourceware.org/ml/systemtap/2011-q4/msg00340.html

12 years agoUpdates the example indices.
William Cohen [Tue, 31 Jan 2012 16:53:51 +0000 (11:53 -0500)]
Updates the example indices.

12 years agoAdd example and buildok test for stopwatch.stp tapset
William Cohen [Tue, 31 Jan 2012 16:51:36 +0000 (11:51 -0500)]
Add example and buildok test for stopwatch.stp tapset

12 years agoAdd the stopwatch.stp tapset
William Cohen [Tue, 31 Jan 2012 15:50:53 +0000 (10:50 -0500)]
Add the stopwatch.stp tapset

The stopwatch.stp tapset provides multiple, independent timers to user
scripts.  Stopwatches can be created by the user script at
anytime. The created stopwatches can be stopped and started by the
user script. The times from the stopwatches can be read in seconds,
milliseconds, microsecons, and nanoseconds.

Signed-off-by: William Cohen <wcohen@redhat.com>
12 years agoPR12773 Update: Fixed abrt install path
Chris Meek [Tue, 31 Jan 2012 16:28:49 +0000 (11:28 -0500)]
PR12773 Update: Fixed abrt install path

abrt-action-install-debuginfo-to-abrt-cache is now
installed in /usr/libexec/ rather than in /usr/bin/.

It now checks both paths, so it should work with
older versions of abrt as well as the newest ones.

12 years ago13631 - Detection of Reqested Privilege Level in stap-serverd
Dave Brolley [Mon, 30 Jan 2012 20:42:54 +0000 (15:42 -0500)]
13631 - Detection of Reqested Privilege Level in stap-serverd

Test for the privilege levels which require signing explicitely.

12 years agoPR 13631 and other rework of multi-privilege level support
Dave Brolley [Mon, 30 Jan 2012 18:40:16 +0000 (13:40 -0500)]
PR 13631 and other rework of multi-privilege level support

  - PR 13631 - Detection of Reqested Privilege Level in stap-server
    - Parsing of the client's command now uses getopt_long and shares
      the getopt input with session.cxx
      - cmdline.cxx, cmdline.h
  - Do not allow multiple privilege-setting options to specify
    different privilege levels.
  - No longer a need for stap-serverd to generate a --privilege
    option for the spawned stap.
  - stap-serverd cannot (and no longer does) rely on knowledge of
    the highest signed privilege level when evaluating the user's
    credentials.
  - Simplified user credentials test against those required by the module
    in stap-serverd.
  - Some kernel.* probes were being registered for use by stapdev only.
    Corrected to be stapdev | stapsys.
    - Corrected testsuite problem which was hiding this bug.

12 years agotask_finder2: update an unprivileged check for stapsys
Josh Stone [Sat, 28 Jan 2012 00:16:56 +0000 (16:16 -0800)]
task_finder2: update an unprivileged check for stapsys

* runtime/task_finder2.c (stap_start_task_finder): Update the macro
  around the "only probe their own threads" check to exempt both stapdev
  and stapsys modules (as __stp_utrace_attach_match_filename does).

12 years agoAUTHORS bump
Frank Ch. Eigler [Fri, 27 Jan 2012 16:38:31 +0000 (11:38 -0500)]
AUTHORS bump

12 years agoPR13628: override LANG for stap-prep operation in non-english locales
Otavio Pontes [Fri, 27 Jan 2012 14:20:24 +0000 (09:20 -0500)]
PR13628: override LANG for stap-prep operation in non-english locales

12 years agoPR13626 Add extra whitespace in sdt.h to deal with C11/C++11 UDF.
Mark Wielaard [Thu, 26 Jan 2012 22:58:38 +0000 (23:58 +0100)]
PR13626 Add extra whitespace in sdt.h to deal with C11/C++11 UDF.

gcc 4.7 supports C11/C++11 user defined literals (UDF):
https://en.wikipedia.org/wiki/C%2B%2B11#User-defined_literals
In std=c++0x mode one now needs to add whitespace between string literals.
See http://gcc.gnu.org/gcc-4.7/porting_to.html

12 years agoPR13546 Cont'd: Added stress tests
Chris Meek [Thu, 26 Jan 2012 21:17:50 +0000 (16:17 -0500)]
PR13546 Cont'd: Added stress tests

Tests both normal and statistics arrays with integer,
string and two string keys with 2048 enties each.

12 years agoPR13546 Cont'd: Updated NEWS
Chris Meek [Thu, 26 Jan 2012 19:01:26 +0000 (14:01 -0500)]
PR13546 Cont'd: Updated NEWS

12 years agoPR13546 Cont'd: Fixed failing test cases
Chris Meek [Thu, 26 Jan 2012 18:56:07 +0000 (13:56 -0500)]
PR13546 Cont'd: Fixed failing test cases

Since we introduced a random seed to the hash function,
certain tests that printed results in an unordered
fashion could now fail, since the order is not necessarily
the same from run to run. These tests have been fixed and now
print in a certain order as to avoid those errors.

12 years agoPR13546 - Improve Map Hashing
Chris Meek [Thu, 26 Jan 2012 18:48:46 +0000 (13:48 -0500)]
PR13546 - Improve Map Hashing

- Made str_hash() in map.c more hashy.
- Introduced a runtime random seed to the hash calculations
  to reduce the chance of deliberate hash collision inducing
  attacks.
- Made the HASH_TABLE_SIZE depend on MAXMAPENTRIES, rather than
  just a static 256. This effectivly increases the size of the
  hash table.

12 years agoDon't squash to pass 4 when using --remote
Josh Stone [Thu, 26 Jan 2012 18:11:23 +0000 (10:11 -0800)]
Don't squash to pass 4 when using --remote

Our native_build check is a little too zealous, squashing runs to pass 4
when the local kernel doesn't match the command-line arguments.  When
we're running on a remote host, the local kernel is irrelevant.

This issue only cropped up if the user provided explicit -r/-a arguments
to match the remote -- otherwise the implicit searching was doing the
right thing with matching each kernel.

* session.cxx (systemtap_session::check_options): Remote targets don't
  care a whit whether the local kernel matches the build.

12 years agocredit for systemtap.spec mktemp
Peter Robinson [Thu, 26 Jan 2012 16:46:55 +0000 (11:46 -0500)]
credit for systemtap.spec mktemp

recognizing original version of commit 87600139

12 years agoPR13539
Jim Keniston [Thu, 26 Jan 2012 15:22:52 +0000 (10:22 -0500)]
PR13539

(Add credit for Jim Keniston's earlier version of these uprobes patches.)

12 years agoPR13539: fight uprobes unregister-vs-exit races
Srikar Dronamraju [Thu, 26 Jan 2012 14:56:31 +0000 (09:56 -0500)]
PR13539: fight uprobes unregister-vs-exit races

While I still cannot see a reason how uprobe_{free,put}_process can
race uprobe_report_{exit,exec}, I certainly think somebit of cleanup
can be done.  However I am dont think we need to do a utask or uproc lookup
from the table. Especially in case of callbacks.

Mostly similar to what Jim proposed.
I haven't tested this patch myself and I couldn't reproduce the problem.
[fche corrected some typos and removed meddlesome printk's]

12 years agoRegenerate systemtap.examples index to include new pstrace_exec.stp.
Mark Wielaard [Wed, 25 Jan 2012 09:13:08 +0000 (10:13 +0100)]
Regenerate systemtap.examples index to include new pstrace_exec.stp.

12 years agoAdd the pstrace function to tapset with documentation and example test
William Cohen [Tue, 24 Jan 2012 16:58:39 +0000 (11:58 -0500)]
Add the pstrace function to tapset with documentation and example test

The pstrace function generates a string that traces the processes from
the task_struct passed in back to the ancestor spawned by init(1).
This allows you to determine what processes are ancestors for a particular
process.

12 years agodump_symbol_tables: kernel entry functions are NOTYPE.
Mark Wielaard [Tue, 24 Jan 2012 13:55:35 +0000 (14:55 +0100)]
dump_symbol_tables: kernel entry functions are NOTYPE.

Commit a09fc5 restricted symbol inclusion of STT_NOTYPE to ET_REL files
(kernel modules) to get fn-desc in .odp which don't occur anywhere else.
But kernel entry functions are also not STT_FUNC but are marked STT_NOTYPE.

12 years agoimprove error message if script requires -g but lacks it, part 3
Frank Ch. Eigler [Tue, 24 Jan 2012 03:42:52 +0000 (22:42 -0500)]
improve error message if script requires -g but lacks it, part 3

* tapsets.cxx (dwarf_var_expanding_visitor::visit_target_symbol): Suggest -g
  if missing.
  (dwarf_cast_expanding_visitor::visit_cast_op): Ditto.
  (tracepoint_var_expanding_visitor::visit_target_symbol_arg): Ditto too.

12 years agoDocument --log, --port and --ssl in the stap-server(8) man page.
Dave Brolley [Fri, 20 Jan 2012 20:15:47 +0000 (15:15 -0500)]
Document --log, --port and --ssl in the stap-server(8) man page.

12 years agoOnly run .plt tests on x86.
Stan Cox [Thu, 19 Jan 2012 21:45:31 +0000 (16:45 -0500)]
Only run .plt tests on x86.

* unprivileged_myproc.exp:  .plt is only supported on x86.
* unprivileged_probes.exp:  Likewise.

12 years agoMerge branch 'master' of ssh://sources.redhat.com/git/systemtap
William Cohen [Thu, 19 Jan 2012 15:22:52 +0000 (10:22 -0500)]
Merge branch 'master' of ssh://sources.redhat.com/git/systemtap

12 years agoImprove extraction of kernel version for unprivileged_embedded_C.exp
William Cohen [Thu, 19 Jan 2012 14:50:08 +0000 (09:50 -0500)]
Improve extraction of kernel version for unprivileged_embedded_C.exp

The tcl code to extract the kernel version from the "uname -r" would
include extraneous characters for some cases, for example "3.2.0+"
would end up with "+" character in something that should only be a number.
This change eliminates the extraneous characters.

12 years agoImprove extraction of kernel version for unprivileged_embedded_C.exp
William Cohen [Thu, 19 Jan 2012 14:50:08 +0000 (09:50 -0500)]
Improve extraction of kernel version for unprivileged_embedded_C.exp

The tcl code to extract the kernel version from the "uname -r" would
include extraneous characters for some cases, for example "3.2.0+"
would end up with "+" chaacter in something that should only be a number.
This change eliminates the extraneous characters.

12 years agoimprove error message if script requires -g but lacks it, part 2
Frank Ch. Eigler [Thu, 19 Jan 2012 00:55:24 +0000 (19:55 -0500)]
improve error message if script requires -g but lacks it, part 2

* tapsets.cxx (*probe_builder::build): Mention stap -g for statement.absolute
  probes.

12 years agoimprove error message if script requires -g but lacks it
Frank Ch. Eigler [Wed, 18 Jan 2012 21:01:31 +0000 (16:01 -0500)]
improve error message if script requires -g but lacks it

* parse.cxx (parse_value, parse::parse_embeddedcode): Wordify.

12 years agoImprove extraction of kernel version for bz6503.exp
William Cohen [Tue, 17 Jan 2012 21:28:20 +0000 (16:28 -0500)]
Improve extraction of kernel version for bz6503.exp

The tcl code to extract the kernel version from the "uname -r" would
include extraneous characters for some cases, for example "3.2.0+"
would end up with "+" chaacter in something that should only be a number.
This change eliminates the extraneous characters.

12 years agoRemove bogus executable bit from process_by_cmd.stp.
Mark Wielaard [Sat, 14 Jan 2012 18:22:38 +0000 (19:22 +0100)]
Remove bogus executable bit from process_by_cmd.stp.

12 years agoFixed /bin/mktemp require.
David Smith [Fri, 13 Jan 2012 15:44:08 +0000 (09:44 -0600)]
Fixed /bin/mktemp require.

12 years agoOnly run .plt tests on x86.
Stan Cox [Thu, 12 Jan 2012 15:54:19 +0000 (10:54 -0500)]
Only run .plt tests on x86.

* semok.exp: .plt is only supported on x86.

12 years agoUpdate to mysql 5.5.19.
Stan Cox [Wed, 11 Jan 2012 03:36:02 +0000 (22:36 -0500)]
Update to mysql 5.5.19.

* mysql.exp (run_tests):  Update mysql_install_db and mysqld invocations.
(main):  Use cmake.  xfail if mysql testrun fails as that is in mysql's realm.

12 years agogcc 4.7 need to #include <unistd.h> for sleep, sysconf, etc.
Mark Wielaard [Mon, 9 Jan 2012 12:01:23 +0000 (13:01 +0100)]
gcc 4.7 need to #include <unistd.h> for sleep, sysconf, etc.

12 years agoSmall unprivileged_myproc.exp testcase fix.
David Smith [Fri, 6 Jan 2012 17:26:56 +0000 (11:26 -0600)]
Small unprivileged_myproc.exp testcase fix.

* testsuite/systemtap.unprivileged/unprivileged_myproc.exp: Small fix to
  catch kill errors (so the test won't error out).

12 years agoMention task_finder2 work in NEWS.
David Smith [Thu, 5 Jan 2012 18:58:29 +0000 (12:58 -0600)]
Mention task_finder2 work in NEWS.

12 years agoFix gcc 4.7 compile issues (most related to gcc PR49745).
David Smith [Thu, 5 Jan 2012 17:16:27 +0000 (11:16 -0600)]
Fix gcc 4.7 compile issues (most related to gcc PR49745).

* cache.cxx: Include <unistd.h>
* session.cxx: Ditto.
* setupdwfl.cxx: Ditto.
* translate.cxx (emit_symbol_data): Set correct type on structure
  initialization to avoid gcc 4.7 error.

12 years agoDo not exit sdt_misc.exp if test #1 fails.
Stan Cox [Tue, 3 Jan 2012 23:13:22 +0000 (18:13 -0500)]
Do not exit sdt_misc.exp if test #1 fails.

* sdt_misc.exp:  Continue if test #1 fails as V1 on s390 may fault getting $arg1

12 years agoruntime: autoconf for fs_supers hlist change
Frank Ch. Eigler [Tue, 3 Jan 2012 21:46:42 +0000 (16:46 -0500)]
runtime: autoconf for fs_supers hlist change

* runtime/autoconf-fs_supers-hlist.c: New test.
* buildrun.cxx (compile_pass): Run it.
* runtime/transport/transport.c (_stp_get_root_dir): Use
  hlist_entry etc. if needed.

12 years agoruntime: autoconf for relay umode_t change
Frank Ch. Eigler [Tue, 3 Jan 2012 20:02:25 +0000 (15:02 -0500)]
runtime: autoconf for relay umode_t change

* runtime/autoconf-relay-umode_t.c: New test.
* buildrun.cxx (compile_pass): Run it.
* runtime/transport/relay_v2.c (__stp_relay_create_buf_file_callback):
  Change int -> umode_t if needed.

12 years agotestsuite: unload uprobes at start of run
Frank Ch. Eigler [Fri, 30 Dec 2011 16:53:52 +0000 (11:53 -0500)]
testsuite: unload uprobes at start of run

* testsuite/lib/systemtap.exp (setup_systemtap_environment):
  In case we have an old one still sitting around, zap it, to
  force its rebuild.

12 years agowhen building uprobes.ko, don't suppress make stderr
Frank Ch. Eigler [Fri, 30 Dec 2011 16:52:45 +0000 (11:52 -0500)]
when building uprobes.ko, don't suppress make stderr

* buildrun.cxx (make_uprobes): Build with same quietness policy
  as base systemtap module, rather than -vvvv policy for tracepoint/query
  test modules.

12 years agorefresh AUTHORS
Frank Ch. Eigler [Sat, 31 Dec 2011 03:41:09 +0000 (22:41 -0500)]
refresh AUTHORS

12 years agorebuild example indexes after keyword tweaks
Frank Ch. Eigler [Sat, 31 Dec 2011 03:40:32 +0000 (22:40 -0500)]
rebuild example indexes after keyword tweaks

12 years agoCorrect and consistent .meta files for example scripts: README
Fredrik Östman [Fri, 23 Dec 2011 08:35:04 +0000 (08:35 +0000)]
Correct and consistent .meta files for example scripts: README

Signed-off-by: Fredrik Östman <fredrik_oestman@mentor.com>
12 years agoCorrect and consistent .meta files for examples
Fredrik Östman [Fri, 23 Dec 2011 08:21:30 +0000 (08:21 +0000)]
Correct and consistent .meta files for examples

Signed-off-by: Fredrik Östman <fredrik_oestman@mentor.com>
12 years agodiagnostics: Let --vp 003 explain why char[] vs. char* is chosen for probe metadata
Frank Ch. Eigler [Thu, 29 Dec 2011 15:18:39 +0000 (10:18 -0500)]
diagnostics: Let --vp 003 explain why char[] vs. char* is chosen for probe metadata

New:
adapt location:14max - 4045132/288938tot =>[]
adapt derivation:39max - 11268582/288938tot =>[]
adapt pp:161max - 24877446/288938tot =>*
adapt pn:161max - 24877446/288938tot =>*

* translate.cxx (translate_pass): Change verbosity>2 diagnostics.

12 years agostap-report: also include lsmod results
Frank Ch. Eigler [Wed, 28 Dec 2011 16:54:23 +0000 (11:54 -0500)]
stap-report: also include lsmod results

12 years agoUse function.call and function.return in all thread_indent() examples.
Mark Wielaard [Sun, 25 Dec 2011 19:56:53 +0000 (20:56 +0100)]
Use function.call and function.return in all thread_indent() examples.

If the .call modifier isn't used the function probes will match both .call
and .inline, but .return doesn't match any inlined functions leading to
unbalanced thread_indent output.

12 years ago$SYSTEMTAP_SYNC: new environment variable for staprun
Frank Ch. Eigler [Fri, 23 Dec 2011 19:49:56 +0000 (14:49 -0500)]
$SYSTEMTAP_SYNC: new environment variable for staprun

If set, it runs a sync(2) before the stap init_module(2)s.  This is
only for helping bring up stap on unstable environments, so is not
formally documented.

12 years agostap-report: include CONFIG_MODULE_* items too
Frank Ch. Eigler [Fri, 23 Dec 2011 19:38:15 +0000 (14:38 -0500)]
stap-report: include CONFIG_MODULE_* items too

12 years agotapset: separate env_var() from context.stp
Frank Ch. Eigler [Thu, 22 Dec 2011 02:06:24 +0000 (21:06 -0500)]
tapset: separate env_var() from context.stp

If it's left in context.stp, it will drag in tokenize.stp, which in
turns drags in an unoptimizable  %{ #define STAP_NEED_CONTEXT_TOKENIZE 1 %}
which in turns enlarges the context.

* tapset/context-envvar.stp: New file. Steal env_var() from ...
* tapset/context.stp: ... from here.

12 years agotestsuite: disable unprivileged_* tests on non-uprobes platforms
Frank Ch. Eigler [Wed, 21 Dec 2011 23:16:07 +0000 (18:16 -0500)]
testsuite: disable unprivileged_* tests on non-uprobes platforms

12 years agoi18n: update po/* stuff from transifex and source changes
Frank Ch. Eigler [Wed, 21 Dec 2011 19:57:22 +0000 (14:57 -0500)]
i18n: update po/* stuff from transifex and source changes

12 years agotestsuite: typo fix for tcl uprobe_p'ing
Frank Ch. Eigler [Wed, 21 Dec 2011 19:05:25 +0000 (14:05 -0500)]
testsuite: typo fix for tcl uprobe_p'ing

commit 55e2452 introduced a typo.  It was introduced by a random
quantum fluctuation, honest.

12 years agotestsuite: more use [uprobes_p] as applicability test
Frank Ch. Eigler [Wed, 21 Dec 2011 18:48:12 +0000 (13:48 -0500)]
testsuite: more use [uprobes_p] as applicability test

12 years agoPR13536: eliminate staprun -u uprobes check/load race condition.
Frank Ch. Eigler [Wed, 21 Dec 2011 18:34:05 +0000 (13:34 -0500)]
PR13536: eliminate staprun -u uprobes check/load race condition.

(Note that there is no security impact to this change.  The worst that
could happen before were unnecessary staprun startup errors.)

* runtime/staprun/staprun_funcs.c (insert_module): Return a nontrivial
  error return code in some cases.  (More needed!)  Don't kvetch about
  a failing init_module(2).
* runtime/staprun/staprun.c (enable_uprobes): Don't grep kallsyms for
  an optimistic early abort.  Just call insert_module.  Don't freak upon
  an EEXIST.
  (insert_stap_module): Kvetch about a failing insert_module().

* testsuite/systemtap.base/buildid.exp: Stop using that profane artefact
  of an obsolete past, "staprun -u".

12 years agotestsuite: use [uprobes_p] as applicability test
Frank Ch. Eigler [Wed, 21 Dec 2011 18:30:28 +0000 (13:30 -0500)]
testsuite: use [uprobes_p] as applicability test

This is necessary as [utrace_p] is becoming trueish for
tracepoint/taskfinder2 kernels.

12 years agomodinfo: Apply to the uprobes hash name too
Josh Stone [Tue, 20 Dec 2011 22:36:53 +0000 (14:36 -0800)]
modinfo: Apply to the uprobes hash name too

Since commit 76d01f36 added the modinfo values to uprobes.ko, then we
need to account for that in the hashed name used for caching.

12 years agoPR13274 related: Added class to save and restore variables
Chris Meek [Tue, 20 Dec 2011 22:14:50 +0000 (17:14 -0500)]
PR13274 related: Added class to save and restore variables

Added a class save_and_restore to automatically save the value of
a variable, and restore it and the end of the calling function.
Can optionally take a variable to initialize the value to something
after saving it.

12 years agoPR13274: Fixed testsuite for standard warning processing (elaborate.cxx)
Chris Meek [Tue, 20 Dec 2011 22:11:00 +0000 (17:11 -0500)]
PR13274: Fixed testsuite for standard warning processing (elaborate.cxx)

12 years agoPR13274: Standardized Warning Processing (elaborate.cxx)
Chris Meek [Tue, 20 Dec 2011 22:04:31 +0000 (17:04 -0500)]
PR13274: Standardized Warning Processing (elaborate.cxx)

Changed warning printing to use session.print_warning(),
and cleaned up warnings being printed.

Also changed the logic in semantic_pass_optimize1 and 2
to either print all warnings if the verbosity is high enough
or suppress all warnings after the 1st itteration.

12 years agoPR13274: Fixed testsuite for standard warning processing (without elaborate.cxx)
Chris Meek [Thu, 15 Dec 2011 19:03:47 +0000 (14:03 -0500)]
PR13274: Fixed testsuite for standard warning processing (without elaborate.cxx)

Since we are now using "WARNING" rather than "Warning", some of the
expect statements in some test cases needed to be tweaked.

12 years agoPR13274: Standardized Warning Processing (without elaborate.cxx)
Chris Meek [Thu, 15 Dec 2011 18:54:43 +0000 (13:54 -0500)]
PR13274: Standardized Warning Processing (without elaborate.cxx)

Changed it so we always use session.print_warning() rather than
an explicit call to clog or cerr. (There are a few cases where
the session is not available, so they were left as is, but
marked up with an // XXX needs-session comment.)

Also changed print_warning() to check for suppress_warnings, so
we no longer have to explicitly check for it when calling
print_warning(). There are a few cases where I left in the
explicit check, specifically when there was more code than
just a simple call to print_warning() within it.

Also standardized the use of "WARNING:". All warnings should now
use "WARNING:" rather than "Warning:" or "warning:".

12 years agoTest with postgres-9.1.2
Stan Cox [Tue, 20 Dec 2011 21:45:06 +0000 (16:45 -0500)]
Test with postgres-9.1.2

* postgres.exp (postgresrelease):  Update.

12 years agomodinfo: also apply to uprobes.ko
Frank Ch. Eigler [Tue, 20 Dec 2011 20:52:27 +0000 (15:52 -0500)]
modinfo: also apply to uprobes.ko

Continue commit 633e5ca70 with respect to a self-built uprobes.ko,
in that --modinfo k=V values should be propagated there too.

12 years agoPR13489: don't confuse staprun's "is uprobes in the kernel?" logic
Frank Ch. Eigler [Tue, 20 Dec 2011 20:49:35 +0000 (15:49 -0500)]
PR13489: don't confuse staprun's "is uprobes in the kernel?" logic

After recent commits associated with PR13489, a stap module may
include a few symbols named kallsyms_*, specifically
*unregister_uprobe.  This confused staprun's simpleminded grep on
/proc/kallsyms to guess that uprobes per se was in the kernel, even
though it was just some peer stap module.  So now grep for a more
picky regexp, including all-important whitespace.

* runtime/staprun/staprun.c (enable_uprobes): Look for
  " unregister_uprobe\t" instead of "unregister_uprobe".

12 years agoUpdate DEPRECATION section of stap.1 with list of deprecated probe types.
David Smith [Tue, 20 Dec 2011 20:11:57 +0000 (14:11 -0600)]
Update DEPRECATION section of stap.1 with list of deprecated probe types.

12 years agoPR13520: force-create $SYSTEMTAP_DIR
Frank Ch. Eigler [Tue, 20 Dec 2011 16:41:15 +0000 (11:41 -0500)]
PR13520: force-create $SYSTEMTAP_DIR

wcohen reports dejagnu/tcl errors otherwise, due to commit f1d4f431.

12 years agoPR13520: perspire to keep lockdep enabled, despite TAINT_OOT_MODULE logic
Frank Ch. Eigler [Tue, 20 Dec 2011 04:43:16 +0000 (23:43 -0500)]
PR13520: perspire to keep lockdep enabled, despite TAINT_OOT_MODULE logic

lockdep is nice to have while running the systemtap testsuite on
recent kernels.

* testsuite/lib/systemtap.exp (setup_systemtap_environment): Make it so.

12 years agoPR13520: add support for loading default options from $SYSTEMTAP_DIR/rc
Frank Ch. Eigler [Tue, 20 Dec 2011 04:39:37 +0000 (23:39 -0500)]
PR13520: add support for loading default options from $SYSTEMTAP_DIR/rc

* main.cxx (main): Load some wordexp()d lines from $SYSTEMTAP_DIR/rc,
  prefixing incoming argc/argv.
* NEWS, stap.1, session.cxx (usage): Document it.

12 years agotask-finder2: fix thinko in uprobes_p test
Frank Ch. Eigler [Tue, 20 Dec 2011 04:36:18 +0000 (23:36 -0500)]
task-finder2: fix thinko in uprobes_p test

commit 771b667 missed a part.

* testsuite/lib/systemtap.exp (uprobes_p): Set path.

12 years agoMake a simple hint for the ``cancelable'' logging exercise.
William Cohen [Mon, 19 Dec 2011 21:30:42 +0000 (16:30 -0500)]
Make a simple hint for the ``cancelable'' logging exercise.

12 years agoRevert "Add in hint for ``cancelable'' logging tapset exercise."
William Cohen [Mon, 19 Dec 2011 21:20:59 +0000 (16:20 -0500)]
Revert "Add in hint for ``cancelable'' logging tapset exercise."

This reverts commit c19343f71a8809cd46080e0ad968ad3006e9e980.

12 years agoAdd in hint for ``cancelable'' logging tapset exercise.
William Cohen [Mon, 19 Dec 2011 21:19:32 +0000 (16:19 -0500)]
Add in hint for ``cancelable'' logging tapset exercise.

12 years agoRevert "Remove "cancelable logging" exercise, the speculative.stp tapset implements it"
William Cohen [Mon, 19 Dec 2011 21:00:01 +0000 (16:00 -0500)]
Revert "Remove "cancelable logging" exercise, the speculative.stp tapset implements it"

This reverts commit 7dd177f6e6ce6488e65b6bbcdd7d98bed15e5689.

12 years agoRemove "cancelable logging" exercise, the speculative.stp tapset implements it
William Cohen [Mon, 19 Dec 2011 20:45:10 +0000 (15:45 -0500)]
Remove "cancelable logging" exercise, the speculative.stp tapset implements it

12 years agoAdd a test for the speculative.stp tapset
William Cohen [Fri, 16 Dec 2011 19:47:00 +0000 (14:47 -0500)]
Add a test for the speculative.stp tapset

A simple test to make sure that the speculative.stp tapset can be compiled
and used.

12 years agoAdd the speculative.stp tapset
William Cohen [Fri, 16 Dec 2011 18:40:01 +0000 (13:40 -0500)]
Add the speculative.stp tapset

The speculative.stp tapset allow one to speculative add things to
output buffers and then later commit or discard the information in the
buffers.  Four functions in the tapset:

speculation() - function to give an id for speculative buffer
speculate() - add output to a speculative buffer
discard() - remove output for a speculative buffer
commit() - output data for a speculative buffer

12 years agoSync up task_finder2.c with task_finder.c
David Smith [Mon, 19 Dec 2011 20:25:36 +0000 (14:25 -0600)]
Sync up task_finder2.c with task_finder.c

* runtime/task_finder2.c: Sync up with runtime/task_finder.c.  Remove some
  unused code and update a privilege check.

12 years agoPR13489, the neverending kludge story. Resolve RHEL5's old utrace conflicts.
Frank Ch. Eigler [Mon, 19 Dec 2011 14:03:58 +0000 (09:03 -0500)]
PR13489, the neverending kludge story.  Resolve RHEL5's old utrace conflicts.

RHEL5's old utrace uses utrace_native_view() instead of
task_user_regset_view(), so we shouldn't be so stuck on trying to find
it.  To add insult to injury, any attempt to use kallsyms* to look up
the latter results in an unloadable module since kallsyms* is not
exported in these old kernels.

* transport/transport.c (_stp_transport_init): Only look for
  task_user_regset_view for other than old rhel5 utrace.

This page took 0.288273 seconds and 5 git commands to generate.