Tom Hughes [Wed, 8 Jun 2011 09:10:40 +0000 (09:10 +0000)]
It seems that the perf_counter_open system call was renamed to
perf_event_open some time after we added it, so correct the name
wherever it appears to match the current kernel source.
Also fixup the PRE handler to do the check correctly, using the
size field of the structure to work out how much data there is.
Julian Seward [Tue, 7 Jun 2011 22:54:32 +0000 (22:54 +0000)]
Add a simple but (to me, at least) useful thing, if (0)'d by default,
to print a line of text approximately every 20 million SBs. This is
useful for monitoring the progress of long running programs.
Julian Seward [Tue, 7 Jun 2011 22:44:09 +0000 (22:44 +0000)]
Fix the computation of debuginfo bias values, in the case where the
debuginfo object has been not been prelinked but the main object has.
Fixes a segfault observed running any C++ application (eg,
drd/tests/annotate_smart_pointer) on Fedora 14 (32-bit), when the
debuginfo RPM for libstdc++ is installed.
Julian Seward [Tue, 7 Jun 2011 21:42:07 +0000 (21:42 +0000)]
find_nsegment_idx: double size of the lookup cache in an attempt
to mitigate the probable increase in miss rates resulting from
r11798 (which causes increased numbers of such queries).
Julian Seward [Tue, 7 Jun 2011 21:39:28 +0000 (21:39 +0000)]
Add a fourth --smc-check= variant, --smc-check=all-non-file. This
adds self-modifying-code checks to all guest code taken from mappings
which are not file backed, but omits checks in code from file backed
mappings. This has the effect of giving complete smc-coverage of JIT
generated code -- since that is invariably generated into anonymous
mapped areas -- without burdening non-JIT generated code with such
checks. Running Firefox 6, --smc-check=all-non-file reduces by a
factor of between 3 and 10 the number of translations requiring a self
check, compared to --smc-check=all. These changes depend on the vex
interface changes in r2158.
Bart Van Assche [Sun, 5 Jun 2011 08:51:47 +0000 (08:51 +0000)]
Make Valgrind build on Linux kernel 3.0 and beyond. Closes #274926. An official
statement that the Linux kernel 3.0 API and ABI is compatible with Linux kernel
2.6 can be found here: http://lkml.org/lkml/2011/5/29/204.
Julian Seward [Fri, 3 Jun 2011 23:27:39 +0000 (23:27 +0000)]
Followup fix to r11794 (generalise processing of PT_LOAD entries):
change some incorrect "rx->bias" (which due to lameness of C's type
system was valid) to "inrx->bias".
Julian Seward [Tue, 31 May 2011 07:17:54 +0000 (07:17 +0000)]
On arm-linux, give gcc "-mcpu=cortex-a8" so as to persuade it to
accept the armv6t2 (I think) insns (movt, movw) in
dispatch-arm-linux.S. Some older gccs apparently think they are
compiling for armv6 by default and reject them (gcc 4.4.1 on Ubuntu
9.10, for one.)
Julian Seward [Tue, 31 May 2011 07:09:06 +0000 (07:09 +0000)]
More gdbserver test fixes, from #214909 c 101:
fix mcsig(no)pass on arm Ubuntu10, arm thumb internal doc, improve
simulate_control_c
* new file docs/internals/arm_thumb_notes_gdbserver.txt
documentation about the subtilities of the thumb bit handling in gdbsrv.
* made the SIGFPE backtrace filtering less dependent on gdb/os/libc/...
* improved simulate_control_c : runs faster/less dependent on timeout value
Julian Seward [Sun, 29 May 2011 09:34:30 +0000 (09:34 +0000)]
Give the amd64-linux and x86-linux dispatchers two entry points, not one,
so as to avoid a GSP-changed check in the common case. See vex r2155.
(amd64-darwin and x86-darwin are now temporarily unbuildable.)
Julian Seward [Sat, 28 May 2011 17:07:53 +0000 (17:07 +0000)]
dispatch-x86-linux.S:
use test-based detection of GSP pointer changes.
Saves one load per SB.
dispatch-amd64-linux.S:
ditto
dispatch-amd64-linux.S:
use movabsq to get &VG_(tt_fast) into a register,
instead of an rsp-relative load from a constant pool.
Saves a second load per SB.
Julian Seward [Sat, 28 May 2011 11:05:44 +0000 (11:05 +0000)]
Remove another memory reference from the arm dispatcher loop, by using
the fact that all {VG,VEX}_TRC_VALUES have their lowest bit set. All
other targets can benefit from this trick too.
Julian Seward [Sat, 28 May 2011 10:16:58 +0000 (10:16 +0000)]
Get rid of a bunch of loads in the arm dispatcher inner loops, and
make some attempt to schedule for Cortex-A8. Improves overall IPC
for none running perf/bz2.c "-O" from 0.879 to 0.925.
Julian Seward [Tue, 17 May 2011 21:35:41 +0000 (21:35 +0000)]
gdbserver: (#214909 c 91)
add cleanup: line to none/tests/require-text-symbol-2.vgtest
In case Valgrind terminates abnormally, then no cleanup is done.
In this case, the abnormal termination is in the test which checks
--require-text-symbol=:*libc.so*:doesntexist
This patch adds a cleanup: line to the test.
Julian Seward [Tue, 17 May 2011 18:14:53 +0000 (18:14 +0000)]
gdbserver: (#214909 c 82)
ensure proper cleanup of gdbsrv FIFOs/shmem files with untraced fork/exec
* syswrap-{generic|darwin|aix5}.c : in PRE(sys_execve) : terminate gdbserver
* pub_core_gdbserver.h and m_gdbserver.c : add VG_(gdbserver_prerun_action),
factorising the actions to do by gdbserver at "startup" (i.e. a traced
fork or a traced exec).
* scheduler.c : implement startup action using VG_(gdbserver_prerun_action)
Julian Seward [Tue, 17 May 2011 17:15:07 +0000 (17:15 +0000)]
gdbserver: misc fixes (#214909 c 77)
Fix some tests on ppc-debian6,s390x + handled Nick Nethercote, Josef
Weidendorfer comments
* improved testing & related doc
- added option --vex-iropt-precise-memory-exns=yes to mcsig(no)pass.vgtest
+ updated manual-core.xml
- cleanup some comments in *.vgtest
- modified filter_gdb and filter_memcheck_monitor to
handle specific ppc/debian6.0 mcsig(no)pass output
handle specific s390x 'missing debug info'
- added more information in README_DEVELOPPERS on how to
investigate failing gdbserver tests.
* handled Nick Nethercote comment:
Replaced kludgy ms.snapshot detailed
by ms.detailed_snaphot
Updated documentation and test.
* handled Josef Weindendorfer comments:
- do not report an error if ptrace_scope file can't be read.
Instead, a debug trace is done if -d (debug) option given
- added an option -l to give the list of active Valgrind
gdbserver. Useful a.o. to support callgrind_control.
Updated documentation
- added ref. to vgdb help in the vgdb --help message
Julian Seward [Tue, 17 May 2011 16:35:11 +0000 (16:35 +0000)]
gdbserver: Fixes for ARM-Thumb (#214909 c 76)
fix arm thumb by transforming an address to its thumb form when needed
* added a function thumb_pc transforming a pc to its thumb form if needed
(using an heuristic to guess if this is a thumb address)
* when program counter is modified by gdb, use thumb_pc
* use thumb_pc in monitor command vg.translate
(I was able to check that this improves inferior call on a small
thumb compiled executable + mcinfcallRU test) but I could not compile
all tests with thumb).
Bart Van Assche [Sun, 15 May 2011 07:04:03 +0000 (07:04 +0000)]
Get rid of the remaining "set but not used" warnings reported by gcc 4.6 by
swapping the roles of the VALGRIND_DO_CLIENT_REQUEST() and
VALGRIND_DO_CLIENT_REQUEST_EXPR() macros. Also, many __attribute__((unused))
declarations on variables have been eliminated. Closes #269778.
Note: so far this patch has been tested on x86/Linux, amd64/Linux and
ppc64/Linux but not yet on any other supported CPU/OS combination.
Julian Seward [Wed, 11 May 2011 15:31:24 +0000 (15:31 +0000)]
When reading Dwarf3 variable type and location information, print a
line showing the number of variables read for each object. Currently
disabled -- is a sanity-check mechanism for exp-sgcheck.
Julian Seward [Tue, 10 May 2011 11:01:07 +0000 (11:01 +0000)]
Improvements for testing and compilation breakage for the GDB server
on various platforms:
* In all gdbserver_tests using gdb:
Made a more general way to remove the initial start message.
* tests using threads burning cpu modified to have only 1 thread.
This makes them independent of the scheduler fairness.
* filter_gdb and filter_vgdb enhanced to anonymise
some debian 6.0/ppc specific things
some s390x/gdb 7.0, gdb 7.1 specific things
* vgdb.c: added an #include <linux/ptrace.h> to fix compilation
on s390x fedora and suse. (Christian Boerntrager)
* fixed a bug in valgrind-low.c debug log :
when a register size is 0, its image cannot be output (and register
should not be transferred).
* added a parameter --keep-unfiltered to vg_regtest.in
This will make it easier to update filter_gdb:
in case gdbserver_tests are failing due to "artificial"
differences to be filtered, re-run the tests using:
perl tests/vg_regtest --keep-unfiltered gdbserver_tests
Then a tar file with all the *.out in gdbserver_tests
will allow me to better/faster update the filter_gdb.
* made a better detection of a working PTRACE_GETREGS at compile time
and/or at run-time.
This is the patch on bug 214909 comment 69.
(Philippe Waroquiers, philippe.waroquiers@skynet.be)
Julian Seward [Mon, 9 May 2011 22:51:14 +0000 (22:51 +0000)]
Back out r11705, which was an attempt to make reading of line number
info created by LLVM 2.9 work properly. As per long discussion in
#272189, this isn't actually possible -- LLVM 2.9 creates bogus line
number info, and the bogusness can't be worked around at the Valgrind
end.
Julian Seward [Mon, 9 May 2011 09:15:28 +0000 (09:15 +0000)]
ppc{32,64}-linux: mark VG_MINIMAL_LONGJMP as noreturn, since it is,
and not doing so leads to compiler warnings for functions that tail
call this one, which themselves are marked "attribute noreturn".
Julian Seward [Wed, 4 May 2011 09:07:38 +0000 (09:07 +0000)]
setup_client_stack: use have_exename to consistently guard uses
of VG_(args_the_exename), thereby avoiding a potential segfault.
Spotted by IBM's BEAM checker.
Julian Seward [Wed, 4 May 2011 09:06:17 +0000 (09:06 +0000)]
calling format_message: when passing frameNo == -1, also pass
tid == VG_INVALID_THREADID rather than an uninitialised ThreadId.
Also in format_message, improve precondition assertions for
frameNo and tid.
There's no error in the current code since if frameNo == -1 then
tid is unused, but it caused IBM's BEAM checker to complain.
Julian Seward [Wed, 4 May 2011 09:01:58 +0000 (09:01 +0000)]
VG_(env_unsetenv), VG_(env_clone): add assertions so as to cause
assertions instead of segfaults. Potential segfaults were detected by
IBM's BEAM checker.
Change the TT_FAST hash function for from "insn_address >> 2" to
"insn_address >> 1". The former is appropriate for ARM code, where
all insns are 4-sized and 4-aligned, but not for Thumb code, where the
minimum size and alignment is 2. The old scheme happened to work for
Thumb (indeed, any hash function would), but caused huge amounts of
conflict misses in the fast cache for some programs.
The change has been observed to reduce conflict misses by up to 100
times, and in some cases, improves performance significantly for Thumb
code. Performance of ARM code is unchanged or possibly a bit worse.
Change the default (minimum) client malloc alignment from 8 to 16
on ppc32-linux. This is needed to make Altivec-using code work
correctly. Noticed when running ./auxprogs/gsl16test with gcc-4.6
with args -mcpu=970 -g -O3 -ftree-vectorize on Memcheck, in
which case a few of the tests failed because malloc() returns
8 byte aligned memory when it should return 16-aligned memory.
Add alternative expected output cases for more recent glibcs (eg,
2.12), which print a minus sign for NaNs. Fixes #262989.
(Maynard Johnson, maynardj@us.ibm.com)
projects / thirdparty / valgrind.git / log
