This is in support of "Bug 375839 - Temporary storage exhausted, when long
sequence of vfmadd231ps instructions to be executed", and reduces code size by
around 3% in that case.
Julian Seward [Mon, 27 Mar 2017 18:32:10 +0000 (18:32 +0000)]
Rewrite dis_FMA so it generates not-quite-so-terrible code. It's still terrible
(breaks vectors into scalars) but this rewrite does it in a way which makes it
interact better with put-to-get forwarding. It also removes all the
Iop_Reinterp casting involved. For long sequences of FMA instructions this
reduces the amount of memcheck-generated code to about 75% of what it was
before. Improves the situation for
Bug 375839 - Temporary storage exhusted , when long sequence of vfmadd231ps instructions to be executed
but isn't a convincing fix.
Ivo Raisr [Fri, 24 Mar 2017 13:46:15 +0000 (13:46 +0000)]
Use consistently chase1() in MSVC specific transformation hacks.
This code is experimental and not used by default but should be self-consistent.
n-i-bz
Ivo Raisr [Thu, 23 Mar 2017 23:22:21 +0000 (23:22 +0000)]
Fix for 377698 - Missing memory check for futex() uaddr arg for FUTEX_WAKE,
and FUTEX_WAKE_BITSET, check only 4 args for FUTEX_WAKE_BITSET,
and 2 args for FUTEX_TRYLOCK_PI.
Fixes BZ#377698.
Patch by: diane.meirowitz@oracle.com
Julian Seward [Mon, 20 Mar 2017 21:34:02 +0000 (21:34 +0000)]
Bug 377717 - Fix massive space leak when reading compressed debuginfo sections.
This makes reading of compressed debuginfo usable for very large object files.
It also adds a bunch extra documentation about a tricky invariant in the
compressed debuginfo handling (a recursive cache refill path!) and adds a
whole bunch of assertions.
Follow up to fix 376956 syswrap of SNDDRV and DRM_IOCTL_VERSION causing some
addresses to be wrongly marked as addressable
As noted by Ivo, if the syscall fails, then we have a leak.
So, enable the flag SfPostOnFail if we allocate memory.
In the POST ioctl, check that FAILURE only happens for this drm ioctl,
and free the memory for both SUCCESS and FAILURE.
Do the POST_MEM_WRITE only if SUCCESS
Add missing break for the DRM ioctl operations that do not have any args
Due to this missing break, the code was falling through to
the case VKI_SNDRV_CTL_IOCTL_PVERSION:
and was then setting some bytes as defined at (whatever address is in) ARG3.
The lfdpx, stdpx, lfdp and stfdp instructions work on a register pair. The
register pair test must only be applied to these instructions in the
dis_fp_pair() function.
Updating NEWS file for the commit
VEX commit 3308 makes the fix in VEX/priv/guest_ppc_toIR.c
The lfdpx, stdpx, lfdp and stfdp instructions work on a register pair. The
register pair test must only be applied to these instructions in the
dis_fp_pair() function.
Mark Wielaard [Thu, 2 Mar 2017 11:14:36 +0000 (11:14 +0000)]
Fix two safe_to_deref/sizeof mismatches in sys_sigaction/sys_rt_sigaction.
Before dereferencing the sigaction pointer and reading the fields we
need to make sure the whole struct is safe_to_deref. We were using the
size of the pointer, but needed the size of the struct.
Ivo Raisr [Sun, 26 Feb 2017 21:37:04 +0000 (21:37 +0000)]
Introduce macro ASSUME_ALIGNED and use it in vki-linux.h.
Reduces the number of compiler warnings on MIPS platforms.
Partial fix for BZ#370028.
Patch by: Aleksandar Rikalo <aleksandar.rikalo@imgtec.com>
Carl Love [Mon, 20 Feb 2017 19:04:13 +0000 (19:04 +0000)]
Patch submitted by Will Schmidt for PPC64.
On ppc64*, the TOC ptr should not be considered a scratch
register, and should not be in the clobber list.
This is called out in newer GCC (i.e. gcc7) and triggers an error.
Thusly, remove R2 from the clobber list.
Signed-Off-By: Will Schmidt <will_schmidt@vnet.ibm.com>
Tested and committed by: Carl Love <cel@us.ibm.com>
modifty sleepers to have easier evaluation of interaction between cpu freq scaling
and scheduler lock (pipe based or futex based)
See http://www.valgrind.org/docs/manual/manual-core.html#manual-core.pthreads_perf_sched
for background info about cpu freq scaling and valgrind thread scheduler.
To reproduce the interaction, do:
for sched in --fair-sched=yes --fair-sched=no
do
for affinity in 0 1
do
echo $sched $affinity
time ./vg-in-place $sched -q ./gdbserver_tests/sleepers 1000000 0 1000 B-B-B-B- $affinity
done
done
which gives the below output (intel core i5-6402P, debian 8, kernel 3.16.0).
In summary: the fair scheduler is fair, the pipe based scheduler
can be really unfair (e.g. with --fair-sched=no and no affinity,
2 threads are finishing their work, while the 2 other threads are
starting their work only after the first 2 have fully finished).
The difference in timing is significant : 1m14s versus around 47 seconds.
Note: If the governor is set to performance, strangely, the time needed for
--fair-sched=no increases slighltly (to around 48 seconds).
The time for --fair-sched=yes with or without affinity is then also
to around 48 seconds.
Below is timing with on-demand governor:
--fair-sched=yes 0
loops/sleep_ms/burn/threads_spec/affinity: 1000000 0 1000 B-B-B-B- 0
Brussels ready to sleep and/or burn
London ready to sleep and/or burn
Petaouchnok ready to sleep and/or burn
main ready to sleep and/or burn
Brussels finished to sleep and/or burn
London finished to sleep and/or burn
Petaouchnok finished to sleep and/or burn
main finished to sleep and/or burn
real 1m14.582s
user 1m14.348s
sys 0m0.204s
--fair-sched=yes 1
loops/sleep_ms/burn/threads_spec/affinity: 1000000 0 1000 B-B-B-B- 1
Brussels ready to sleep and/or burn
London ready to sleep and/or burn
Petaouchnok ready to sleep and/or burn
main ready to sleep and/or burn
main finished to sleep and/or burn
Brussels finished to sleep and/or burn
Petaouchnok finished to sleep and/or burn
London finished to sleep and/or burn
real 0m46.785s
user 0m46.756s
sys 0m0.032s
--fair-sched=no 0
loops/sleep_ms/burn/threads_spec/affinity: 1000000 0 1000 B-B-B-B- 0
Brussels ready to sleep and/or burn
Brussels finished to sleep and/or burn
London ready to sleep and/or burn
London finished to sleep and/or burn
Petaouchnok ready to sleep and/or burn
main ready to sleep and/or burn
Petaouchnok finished to sleep and/or burn
main finished to sleep and/or burn
real 0m47.742s
user 0m48.224s
sys 0m0.084s
--fair-sched=no 1
loops/sleep_ms/burn/threads_spec/affinity: 1000000 0 1000 B-B-B-B- 1
Brussels ready to sleep and/or burn
London ready to sleep and/or burn
Petaouchnok ready to sleep and/or burn
main ready to sleep and/or burn
Brussels finished to sleep and/or burn
London finished to sleep and/or burn
main finished to sleep and/or burn
Petaouchnok finished to sleep and/or burn
Mark Wielaard [Sat, 18 Feb 2017 21:16:05 +0000 (21:16 +0000)]
Remove arm64 specific (unused) setup_child () and assign_guest_tls ().
valgrind svn r16186 (which fixed bug #342040 and #373192) introduced a
generic linux way of handling setup_child and assign_guest_tls. So there
is no need anymore for arch specific implementations. And gcc complained
they were unused.
Petar Jovanovic [Mon, 13 Feb 2017 16:15:24 +0000 (16:15 +0000)]
mips64: do correct 32-bit comparison for Iop_CmpNE32
Make sure that we take into account 32-bit size of values in comparison
on MIPS64-platforms. This is done either by sign extending these values
before comparison or sign extending xored values (depending on what
comparison we do). This should avoid false-positives like the one
reported in BZ #341481.
Patch based on code provided by Crestez Dan Leonard and Tamara Vlahovic.
Petar Jovanovic [Fri, 10 Feb 2017 17:58:40 +0000 (17:58 +0000)]
mips: rewrite mips_irgen_load_and_add32|64 and code around it
Make sure that mips_irgen_load_and_add32 gets both expected value and
new value, so the function code makes more sense and does load/store in
a atomic way.
Petar Jovanovic [Fri, 3 Feb 2017 01:19:55 +0000 (01:19 +0000)]
fix leak-segv-jmp test for platforms with 64K pagesize
Increase the size of allocated array, so mprotect call does not end up
protecting non-allocated areas. This enables the test to work on
platforms with pagesize=64K.
Issue discovered on MIPS XLP machine with 64K pagesize.
Petar Jovanovic [Fri, 3 Feb 2017 00:34:52 +0000 (00:34 +0000)]
add suppression for helgrind/tests/tc22_exit_w_lock
Function pthread_create indirectly calls function memcpy. Helgrind
considers that memcpy is not thread safe function. For error reported
from pthread_create there is the suppression helgrind---_dl_allocate_tls
in the file glibc-2.34567-NPTL-helgrind.supp.
Since glibc version 2.23, memcpy is implemented by __mempcpy_inline.
This causes that call to memcpy from pthread_create is no longer
recognized by the suppression.
In test helgrind/tests/tc22_exit_w_lock, pthread_create is called twice,
and second call reports error, which causes failing of the test.
This patch adds suppression for glibc 2.23 and greater.
Petar Jovanovic [Fri, 27 Jan 2017 18:20:50 +0000 (18:20 +0000)]
Adding new filter to memcheck/tests/filter_stderr
This should fix issue with sendmsg test and glibc 2.22.
Glibc 2.22 introduced sysdeps/unix/sysv/linux/sendmsg.c that has
__libc_sendmsg function implementation (in comparison to earlier
implementation in syscall-template.S).
So, test suite needs to filter out this case, otherwise we get test
diffs such as:
Syscall param sendmsg(msg) points to uninitialised byte(s)
- at 0x........: sendmsg (in /...libc...)
+ at 0x........: sendmsg (sendmsg.c:28)
which are false positives.
This fixes memcheck/tests/sendmsg (stderr) on platforms with 2.22+ glibc.
Petar Jovanovic [Fri, 27 Jan 2017 17:55:13 +0000 (17:55 +0000)]
mips: fix target_get_drv() function
Return correct Dtv location. Top of MIPS tcbhead structure is located
0x7000 bytes before the value of ULR. Dtv is the first of two pointers
in the tcbhead structure.
This fixes gdbserver_tests/hgtls on some MIPS platforms.
Allow memcheck to output the leak results as a callgrind xtree file.
* New command line options --xtree-leak=no|yes and --xtree-leak-file=<file>
to produce the end of execution leak report in a xtree callgrind format
file.
* New option 'xtleak' in the memcheck leak_check monitor command, to
produce the leak report in an xtree file.
* File name template arguments (such as --log-file, --xtree-memory-file, ...)
have a new %n format letter that is replaced by a sequence number.
Julian Seward [Mon, 16 Jan 2017 05:14:24 +0000 (05:14 +0000)]
dis_neon_data_3same: don't silently accept invalid instructions. Instead,
"return False" for any instruction not accepted by this function. Also,
add a few switch default backstops.
Petar Jovanovic [Fri, 13 Jan 2017 16:29:15 +0000 (16:29 +0000)]
mips64: update exp file for test_math
Leave the old exp file that covers cases in which __addtf3 and __subtf3
did not take into account rounding modes. New exp file is the same file
that already exists in mips32 folder, so we just create a symbolic link
to it.
projects / thirdparty / valgrind.git / log
