| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| | |
| |
| |
| |
| |
| |
| | |
(probe_table::get_next_probe): Dissect using struct probe_table.
(sdt_var_expanding_visitor): Use lex_cast.
(dwarf_builder::build): Copy probe and location for TOK_MARK cases.
Call derive_probes for kprobe and utrace cases.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
query_module was already returning DW_CB_ABORT when a simple match was
found, but dwflpp::iterate_over_modules was ignoring that and instead
forcing the module loop to restart. The only way out of the loop was
with the pending_interrupts flag, which is only for signalled
interrupts.
Now iterate_over_modules will only attempt the dwfl_getmodules loop
once, since that loop will only abort if the CB returns DW_CB_ABORT.
Then query_module is also modified to return ABORT if pending_interrupts
is flagged.
My trusty test, stap -l syscall.*, is nearly 2x faster with this change.
Empirically, I found that the kernel object is always the first "module"
returned, so the syscall probepoints always gets to short-circuit the
loop right away.
|
| |\| |
|
| | |
| |
| |
| |
| |
| |
| | |
* tapsets.cxx (dwarf_builder::probe_table::probe_table): gcc 4.4
complains that shdr may be used uninitialized. I added returns to
ensure that it's ok, but gcc still complains. Set the thing to NULL
as well to silence the beast.
|
| | |
| |
| |
| |
| |
| | |
dwarf_builder::build to handle the .probes section
(sdt_var_expanding_visitor): New. Expand static probe $argN.
(dwarf_builder::build): Use probe_table. Add kprobe and utrace probe types.
|
| |\| |
|
| | |
| |
| |
| |
| |
| |
| | |
For a call like "stap -l 'syscall.*'", I found that ~10% of the time was
spent compiling the blacklist regexps over again for each probe point.
By moving this functionality into the kernel dwflpp instance, we can
reuse the regexps and get an easy speed boost.
|
| |\| |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Both kernel.function and kprobe.function were defining a global array
stap_unreg_kprobes to use in bulk kprobes unregistration. The compiler
allowed the duplicate definition as long as they were the same size, as
it was when exercised in buildok/thirtyone.
kprobe.function now uses a separate stap_unreg_kprobes2, and the
testcase is modified to produce an imbalanced number of probes.
|
| | |
| |
| |
| |
| |
| | |
When a kernel.function or kprobe.function fails in registration, we
usually print a WARNING and move on. With this patch, kprobes that have
the optional '?' flag will not print any WARNING.
|
| |\| |
|
| | | |
|
| | |\ |
|
| | | | |
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The method query_cu_containing_global_address was only called by
query_cu_containing_module_address, and the latter was just doing a
simple argument transform. They are now merged into a single method,
query_cu_containing_address. The function module_address_to_global is
also merged here at its only call site.
|
| | | |
| | |
| | |
| | |
| | | |
It's not a terribly clean split, but moving it helps reveals some of the
knots that need to be untangled.
|
| | | | |
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | |
| | | |
We know the full type of every tracepoint argument, so for those that
are pointers, print $$vars/$$parms using "%p". The integer-type
arguments continue to use the generic "%#x".
|
| | |/
| |
| |
| |
| |
| | |
* includes/sys/sdt.h (STAP_PROBE_DATA_): Mark .probes section SHF_ALLOC.
* tapsets.cxx (dwarf_builder::build): Search in either dwarf or main elf
file for .probes section.
|
| |\|
| |
| |
| |
| |
| | |
Conflicts:
tapsets.cxx
|
| | |
| |
| |
| |
| |
| | |
* tapsets.cxx (dwarf_builder::build): Add some comments, verbose log
messages and get Elf through dwarf_getelf if it exists before searching
for .probes section.
|
| | |
| |
| |
| |
| |
| | |
We have the saved_conversion_error field, but I wasn't using it. Now
@cast errors are saved in that field, so they're only seen if the
optimizer doesn't remove the @cast.
|
| | |
| |
| |
| |
| |
| |
| | |
Since translate_components is the one that knows the details of its
failures, it makes more sense to let it throw its own errors, instead of
relying on each caller to do it. The function now always either returns
successfully or throws an error.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The previous code recursed the entirety of translate_components, and
it seemed to be restarting the components list every time, so it would
only work if the anonymous portion was the first component. Even then,
examining the code output by semok/thirtythree revealed that it wasn't
fully translating the locations when multiple anonymous pieces were
involved.
Instead, it now recurses in a separate function, find_struct_member,
which does just enough to find the member die and return. It also
builds a vector of the locations passed through, so translate_components
can output code for the full chain of anonymity.
The generated code for semok/thirtythree's $page->mapping now appears to
match the offsets from my manual inspection of struct page. I also
added a test for $page->first_page->mapping, which works now but would
segfault the old code.
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| |
| | |
All of the timer.* tapsets are now built and handled in their own
tapset-timers.cxx, as the first step of many to pare down the current
monolithic tapsets.cxx.
|
| | | |
|
| |\|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
modsign.cxx
runtime/staprun/modverify.c
runtime/staprun/staprun_funcs.c
stap-authorize-server-cert
stap-authorize-signing-cert
stap-serverd
systemtap.spec
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
Newer gcc (4.4) places more DW_TAG_subprograms with DW_AT_declaration
at call sites. iterate_over_labels should only be concerned about
complete declarations.
* tapsets.cxx (dwflpp::iterate_over_labels): Skip DW_TAG_subprogram
that has attribute DW_AT_declaration.
|
| |\| |
|
| | |
| |
| |
| |
| |
| |
| | |
Instead of just structs and unions, we now also process typedefs, enums,
and base types in iterate_over_globals, so declaration_resolve can find
more variations. This especially useful to let @casts reference typedef
names.
|
| | |
| |
| |
| |
| | |
In translate_components, give a cleaner warning about trying to
dereference an enum type, rather than "unexpected type tag".
|
| |\|
| |
| |
| |
| |
| |
| | |
Conflicts:
configure
testsuite/configure
|
| | |
| |
| |
| |
| |
| |
| |
| | |
Older (2.6.18) based kernels defined struct kprobes symbol_name as char *,
but newer (2.6.27) based kernels defined the same field as const char *.
* tapsets.cxx (kprobe_derived_probe_group::emit_module_init):
Always cast to char * when assigning to kprobes.symbol_name field.
|
| | |
| |
| |
| |
| | |
* tapsets.cxx (kprobe_derived_probe): Use enter_k[ret]probe>>2<<_probe.
* testsuite/buildok/thirtyone.stp: Test this.
|
| | |
| |
| |
| |
| | |
The code to emit maxactive was almost there; it just needed to enable
the maxactive probe component and pass the value down.
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
All of our other kprobe types will let the script continue after a
registration failure, as per PR6749. The dwarfless kprobes should be
no exception.
Also, the way they were exiting was causing an unclean shutdown, so I
added tests to badkprobe.exp to make sure that cleanup is always allowed
to run.
|
| | | |
|
| |\|
| |
| |
| |
| |
| | |
Conflicts:
Makefile.in
|
| | |
| |
| |
| |
| |
| |
| | |
* tapsets.cxx: Add dwfl_report_offline_predicate to check pending
interrupts for an early abort.
* tapsets.h: Declare it.
* translate.cxx: Add callback to function call.
|
