Mark Wielaard [Tue, 29 Dec 2009 20:05:55 +0000 (21:05 +0100)]
Fixup some memory tapset vm kernel function probe fallbacks.
Older kernels don't have all GFP constants defined, and the fallback
kernel function probe fallbacks don't have the same dwarf variable
names as the kernel trace point probes. So replace them with variables
that are available. bytes_alloc was sometimes a long and sometimes a
string, this caused scripts to fail depending on which alternative was
chosen for a particular kernel. So make it a long always.
This isn't a full solution since kmalloc is always inlined which makes
the kernel.function("kmalloc").return probe fail.
* tapset/memory.stp: Define __GFP_THISNODE, __GFP_RECLAIMABLE,
GFP_TEMPORARY, GFP_HIGHUSER_MOVABLE and GFP_THISNODE when not yet
defined.
(__vm.kmalloc.kp): Use $flags, not $gfp_flags. Set bytes_alloc equal
to bytes_req.
(__vm.kmem_cache_alloc.kp): Likewise. And use $cachep->buffer_size
for bytes_req.
(__vm.kmalloc_node.kp): Likewise.
(__vm.kmem_cache_alloc_node.kp): Likewise.
(__vm.kfree.kp): Use $ibjp for ptr, not $return.
(__vm.kmem_cache_free.kp): Likewise.
* testsuite/buildok/vm.tracepoints.stp: Move vm.kmalloc test to...
* testsuite/buildok/vm.tracepoints.kmalloc.stp: ... here.
Mark Wielaard [Tue, 29 Dec 2009 19:04:19 +0000 (20:04 +0100)]
Remove all caching from _stp_module_relocate.
This function now also does user space address relocation. The connection
with module loading/unloading was removed in commit 1b94bf which made all
symbol tables emitted at compile time. _stp_module_relocate is called from
the dwarf unwinder, and when doing some of the symbol lookup through
dwflpp::emit_adress now.
* runtime/sym.c (_stp_module_relocate): Remove last, last_sec and last_tsk
caching and invalidation code.
Tim Moore [Tue, 22 Dec 2009 22:30:56 +0000 (23:30 +0100)]
grapher: remove _zoomFactor from horizontal scaling calculation
This makes zooming more consistent and less buggy.
* grapher/Graph.cxx (Graph::draw): Set graph time spread using
_zoomFactor.
* grapher/Graph.hxx (Graph::getHorizontalScale): no _zoomFactor
* grapher/GraphWidget.cxx (on_motion_notify_event): Dragging increment
uses _zoomFactor.
(on_scroll_event): Set extents after changing the zoom factor with
the scroll wheel.
Tim Moore [Tue, 22 Dec 2009 18:48:20 +0000 (19:48 +0100)]
grapher: scale from right end of graph
The right side of the graph represents the most recent time. Since
there is never anything interesting to the right of that, it makes
sense to have the origin of the scaling be there.
* grapher/Graph.hxx (getHorizontalScale): new function
* grapher/GraphStyle.cxx (GraphStyleBar::draw, GraphStyleDot::draw,
GraphStyleEvent::draw): Use cairo transform functions to set up
scaling.
(GraphStyleBar::dataIndexAtPoint, GraphStyleEvent::dataIndexAtPoint):
Base calculations on scaling from right.
Tim Moore [Tue, 22 Dec 2009 10:35:38 +0000 (11:35 +0100)]
grapher: scroll continuously with time
Don't scale graph based on how much data will fit. This didn't work
very well and resulted in distracting, weird scale changes. We now
assume that scripts output their time (x axis) in milliseconds.
* grapher/Graph.hxx (setCurrentTime): New function.
* grapher/Graph.cxx (Graph::draw): Assume a fixed default scale of 1
pixel = 5 milliseconds and don't do any autoscaling.
* grapher/GraphWidget.cxx (GraphWidget constructor): Set global time
base on startup.
(on_expose_event): Don't search graphs for earliest time.
* grapher/GraphWidget.hxx (_timeBaseInitialized): delete
* grapher/Time.hxx: new file; interface to timeval.
Mark Wielaard [Mon, 21 Dec 2009 12:02:19 +0000 (13:02 +0100)]
PR11015 Support shared library reloading (in different processes)
* runtime/task_finder_vma.c (stap_remove_vma_map_info): Return negative
status on failure.
(stap_find_vma_map_info): Likewise.
(stap_find_vma_map_info_user): New function.
(stap_drop_vma_maps): New function.
* runtime/sym.h (addr): Renamed to static_addr, to store addresses for
sections which are always mapped at the same address.
(_stp_module_relocate): Add extra struct task_struct *tsk argument.
* runtime/sym.c (_stp_tf_exec_cb): New callback, calls stap_drop_vma_maps.
(_stp_tf_mmap_cb): Don't store address in module.section, but call
stap_add_vma_map_info() per tsk->group_leader for matched module.
Don't register empty/null modules.
(_stp_module_relocate): Take extra struct task_struct *tsk argument,
cache last tsk used. Only use section->static_addr for none dynamic
modules. Use stap_find_vma_map_info_user() to locate dynamic modules.
(_stp_mod_sec_lookup): Add extra argument unsigned long *rel_addr to
optionally store relative address when module/section found.
(_stp_kallsyms_lookup): Use _stp_mod_sec_lookup to find relative address.
(_stp_sym_init): Register _stp_tf_exec_cb in stap_task_finder_target.
Add error check to see if task finder could be initialized.
* dwflpp.cxx (emit_address): Pass NULL for kernel/modules and current for
user tasks to _stp_module_relocate.
* runtime/transport/symbols.c (_stp_do_relocation): Set new static_addr
_stp_section field.
* runtime/unwind.c (adjustStartLoc): Take new struct task_struct *tsk
argument and pass to stap_find_vma_map_info_user and _stp_module_relocate
to find adjusted addr.
(_stp_search_unwind_hdr): Pass through struct task_struct *tsk.
(unwind_frame): Likewise.
* tapset/context-symbols.stp (probemod): Add NULL to _stp_mod_sec_lookup
call to indicate we aren't interested in relative address.
* tapsets.cxx (dwarf_derived_probe_group::emit_module_init): Pass NULL to
_stp_module_relocate to indicate kernel/module address.
Josh Stone [Tue, 22 Dec 2009 02:26:47 +0000 (18:26 -0800)]
PR11112: Check the full %m/M buffer, and limit the length
We already had code in place to try a deref on the requested memory
buffer, but it was missing the static-precision case. Thus, it was
possible to craft an address that would pass the check on the first byte
but would pagefault at the end of the buffer.
While we're at it, we should also be limiting the number of bytes in
such a read, so even legitimately-huge buffers won't chew up kernel
time. I've arbitrarily chosen 1024 as the limit, but we can revisit
that later. (see also PR10490)
TODO: we need a reliable testcase where a starting address is valid but
the end address is bogus. In PR11112, the reproducer was using a huge
precision to run off the heap, but we need something that will
consistently work even with <1024 length.
Mark Wielaard [Mon, 21 Dec 2009 12:02:19 +0000 (13:02 +0100)]
PR11015 Support shared library reloading (in different processes)
* runtime/task_finder_vma.c (stap_remove_vma_map_info): Return negative
status on failure.
(stap_find_vma_map_info): Likewise.
(stap_find_vma_map_info_user): New function.
(stap_drop_vma_maps): New function.
* runtime/sym.h (addr): Renamed to static_addr, to store addresses for
sections which are always mapped at the same address.
(_stp_module_relocate): Add extra struct task_struct *tsk argument.
* runtime/sym.c (_stp_tf_exec_cb): New callback, calls stap_drop_vma_maps.
(_stp_tf_mmap_cb): Don't store address in module.section, but call
stap_add_vma_map_info() per tsk->group_leader for matched module.
Don't register empty/null modules.
(_stp_module_relocate): Take extra struct task_struct *tsk argument,
cache last tsk used. Only use section->static_addr for none dynamic
modules. Use stap_find_vma_map_info_user() to locate dynamic modules.
(_stp_mod_sec_lookup): Add extra argument unsigned long *rel_addr to
optionally store relative address when module/section found.
(_stp_kallsyms_lookup): Use _stp_mod_sec_lookup to find relative address.
(_stp_sym_init): Register _stp_tf_exec_cb in stap_task_finder_target.
Add error check to see if task finder could be initialized.
* dwflpp.cxx (emit_address): Pass NULL for kernel/modules and current for
user tasks to _stp_module_relocate.
* runtime/transport/symbols.c (_stp_do_relocation): Set new static_addr
_stp_section field.
* runtime/unwind.c (adjustStartLoc): Take new struct task_struct *tsk
argument and pass to stap_find_vma_map_info_user and _stp_module_relocate
to find adjusted addr.
(_stp_search_unwind_hdr): Pass through struct task_struct *tsk.
(unwind_frame): Likewise.
* tapset/context-symbols.stp (probemod): Add NULL to _stp_mod_sec_lookup
call to indicate we aren't interested in relative address.
* tapsets.cxx (dwarf_derived_probe_group::emit_module_init): Pass NULL to
_stp_module_relocate to indicate kernel/module address.
Frank Ch. Eigler [Fri, 18 Dec 2009 17:20:52 +0000 (12:20 -0500)]
PR10601 part 1: i386 and x86-64 regset for dwarf fetch/store_register()s
* runtime/loc2c-runtime.h (fetch_register, store_register):
forked into k_ (kernel) and u_ (user) varieties. Implement
i386 and x86_64 in terms of regset.h; fall back to k_* for
other architectures.
* tapsets.cxx: (*::visit_target_symbol): Emit macros to map
loc2c's fetch/store_register to loc2c-runtime's k_ or u_ as
appopriate.
Tim Moore [Fri, 18 Dec 2009 16:56:24 +0000 (17:56 +0100)]
Remove uprobes.h declaration from runtime.h
Turns out that it breaks on kernels that don't have utrace.
* runtime/runtime.h : Don't include uprobes.h
* runtime/stack.c: Include uprobes.h
* runtime/stack-i386.c: Check if uprobes is included at all.
* runtime/stack-x86_64.c: ditto
* tapsets.cxx (uprobe_derived_probe_group::emit_module_decls): put
uprobes.h include back in.
William Cohen [Thu, 17 Dec 2009 22:51:16 +0000 (17:51 -0500)]
Remove blank line in ucontext-unwind.stp for SystemTap Tapset Reference Manual
The automatic documentation extraction doesn't allow blank lines between the
the comment and the actual code. Removed the problem blank line to allow
the information to be extracted for the SystemTap Tapset Reference Manual.
Tim Moore [Thu, 17 Dec 2009 15:18:34 +0000 (16:18 +0100)]
support for a brief backtrace format
This only prints symbol+offset, or an address if the symbol isn't
known.
* runtime/runtime.h (SYM_VERBOSE_NO, SYM_VERBOSE_FULL,
SYM_VERBOSE_BRIEF): new constants
* runtime/stack.c (_stp_stack_print): support brief format
* runtime/sym.c (_stp_func_print): ditto
* tapset/ucontext-unwind.stp (print_ubacktrace_brief): new function
* testsuite/systemtap.context/fib.c: new test program
* testsuite/systemtap.context/fib.stp: new test
* testsuite/systemtap.context/fib.exp: new test
Tim Moore [Wed, 16 Dec 2009 11:00:55 +0000 (12:00 +0100)]
set the IP in return probes to the returned-to instruction
It's easily available in kretprobes and uretprobes and is consistent
with the rest of the program state.
* translate.cxx (emit_common_header) : add uretprobe_instance to context.
* tapsets.cxx (common_probe_entryfn_prologue): Initialize ri in
context to 0.
(dwarf_derived_probe_group::emit_module_decls): Change IP to return
address in kretprobes.
(uprobe_derived_probe_group::emit_module_decls): enter_uretprobe_probe:
set ri (uretprobe_instance) in context. Change IP to return
address in uretprobes. Don't emit uprobe include and #define
* runtime/runtime.h : Add includes and #define for uprobes.
* runtime/stack.c (_stp_stack_print, _stp_stack_snprint): Add extra
argument for uretprobe_instance.
* tapset/context-unwind.stp (print_backtrace, backtrace): Pass NULL
for uretprobe_instance to _stp_stack_print.
* tapset/ucontext-unwind.stp (print_ubacktrace, ubacktrace): pass
uretprobe_instance to _stp_stack_print
* testsuite/systemtap.context/uprobe_uaddr.exp : new test for uaddr in
function probes
* testsuite/systemtap.context/uprobe_uaddr.stp : new file
Dave Brolley [Mon, 14 Dec 2009 18:08:45 +0000 (13:08 -0500)]
PR 10905: stap-server initscript improvements
o Handle, -B, -I, -R options
o Allow specification of servers by pid
o Allow specification of servers by nickname.
o Advertise options used using avahi.
Use regs->ARM_cpsr instead, this should work with all versions
of linux found in git repo (versions >=2.6.12-rc2). Difference
between condition_codes() and regs->ARM_cpsr should not matter
for systemtap.
Stan Cox [Fri, 11 Dec 2009 21:59:55 +0000 (16:59 -0500)]
Use env(SYSTEMTAP_TESTAPPS) for all systemtap.apps
mysql.exp: Use env(SYSTEMTAP_TESTAPPS).
(mysqlrelease): Update and check for download failure.
postgres.exp: Use env(SYSTEMTAP_TESTAPPS).
stap-tcl.stp: Use .library("library").mark("mark").
tcl.exp: Likewise.
Tim Moore [Fri, 11 Dec 2009 13:03:47 +0000 (14:03 +0100)]
grapher: implement restarting a stap process
* grapher/StapParser.cxx (StapParser::disconnect): new function
* grapher/StapParser.hxx (StapProcess::StapProcess): initialize argv to 0
* grapher/grapher.cxx (StapLauncher::setArgs): Set argv to 0
(StapLauncher launch, launchUsingParser): Refactor launch(),
extracting function a that (re)launches a stap process using an
existing parser.
(StapLauncher::onChildDied): call disconnect() on dead parser.
(GrapherWindow::_graphicalLauncher, setGraphicalLauncher): delete
member, replacing with...
(graphicalLauncher): new variable
(ProcModelColumns): Store parser object in the list model instead of
just a StapProcess object.
(ProcWindow::onRestart): new function
(ProcWindow::refresh): Preserve the list selection when the process
list is refreshed.
(ProcWindow::onSelectionChanged): Manage the restart button's state.