| Age | Commit message (Collapse) | Author |
|---|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Windows version of this test utility was named perr.exe. Change
to match linux as ocsd-perr.exe.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Add more information on trc_pkt_lister, and add in list of
development only test utilities.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
The trc_pkt_lister man file was out of date.
Update to contain all the latest relevant information.
Adjust the makefile to install / clean man file separately from library.
Reported-by: Wookey <wookey@wookware.org>
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Fix missing ')'
When listing options, put dev only ones in separate section.
Reported-by: Wookey <wookey@wookware.org>
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Build of mem_acc_test app has no corresponding clean operation, which can cause
subsequent builds to fail. Add clean operation.
Reported-by: Wookey <wookey@wookware.org>
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Fixes a build warning on the decoder source.
Reported-by: Wookey <wookey@wookware.org>
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Add in new instructions from Arch 9.5 Feat_PAuth_LR, RETA<k>SPPC <label>
and RETA<k>SPPC <Rm> to iinstruction decoder to recognise these as
indirect branches.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Add snapshot to test multi-session decode
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Snapshots can declare multiple buffer in trace.ini:
[trace_buffers]
buffers=buffer1,buffer2
Update test program to process all listed buffers into single file. Assumes
config of first buffer valid for all.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
The memory space used when requesting memory that results in the _NACC packet
is now output as part of the packet printed string. string
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
cs-trace-disasm.py is pushed to linux kernel repo, and file
path is tools/perf/scripts/python/arm-cs-trace-disasm.py.
Fixes docs HOWTO.md and test script perf-test-script.bash
Reported as Github pull request #60
Signed-off-by: yuxiangyang <yuxiangyang326@gmail.com>
|
|
Master and channel IDs were not being correctly set when decoder was reset.
Additionally they were not reset when a VERSION packet is received (Github issue #61)
Reported-by: MommeSherif on github.com
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Setting GCCDIR=1 when using makefile.dev will cause the output directory to be
of the form builddir/<arch>/GCC_<ver>
Fix makefile to allow for version numbers longer than 3 characters.
e.g. 11.3
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Add docs for caching and instruction run limit on etmv4 / ete decoder.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Test updates to check the debug features enabled by the OPENCSD_INSTR_RANGE_LIMIT
and OPENCSD_ERR_ON_AA64_BAD_OPCODE env vars.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
The snapshot reader is updated to use the 'space' key in the [dump] section
of the snapshot files. This sets the memory space used when registering the
memory dump file with the decoder.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Test mem-acc mapper, accessors and caching
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Packet lister test program adds command line parameters to adjust memacc caching parameters
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Decode tree will read any OPENCSD_xx env vars and adjust caching accordingly
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Caching of memory data is now tunable either from the client by API, or via runtime
environment variables.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Add default constructor and initialise function to mem-acc-buf
and mem-acc-cb to allow declaration as local stack objects.
Fix memory issue in mapper - do not destroy accessors on clear, code that
adds then must delete them after removal from mapper.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Decode defines memory spaces used to access program image.
RME requires that new spaces for Realm and Root are required.
New types defined to allow clients to set memory blocks / handle callbacks
for these memory spaces.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Adds an optional limit on the number of instructions that will be allow in a single instruciton
range.
Limit not used unless an environment variable "OPENCSD_INSTR_RANGE_LIMIT" is set.
Use to debug possible decoder runaway if incorrect program image supplied
to decoder is suspected.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
M class exceptions can be tail chained with the address set to a fixed indicator
value and not being the preferred return address.
Fix this be handling that value and not trying to walk to the address.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
AArch64 instructions have an illegal opcode range when top 16 bits or 0x0000
Return error if one of these opcodes is seen.
Help debug issues when incorrect memory is provided to the decoder.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
VS2015 is deprecated as product no longer supported.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Remove residual VS2010 workarounds as support for VS2010 is no longer supported.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Add option to mute printed output to allow tests of library performance
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Possible overrun error when using iterators in mispredictAtoms.
Fix p0 stack to ensure iterator cannot run past end() element & correctly
take reference on element for later deleting after iterator is erased.
Reported-by yabinc <yabinc@google.com> (github issue #53, #58)
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Need local copy of Generic output object to correctly return via get func
and delete on C-API destroy decode tree.
Reported-by: 'holio0' (github issue #55)
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Adds in support for cores with FEAT_ITE (arch v9.4, ETE p1v3), which allows a
PE to inject instumentation data directly into the ETE trace stream.
This new packet is decoded and passed through to the decoder.
This generates a new generic packet type OCSD_GEN_TRC_ELEM_INSTRUMENTATION
containing the instrumentation payload.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Mispredict handler was removing unneeded addresses between the mispredict
packet and the atom packet, without deleting the item afterwards.
Erase now also deletes the unwanted item.
Fixes issue #52 from github
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
ocsdlib has ASAN enabled for this build - follow suit with test progs to
allow use of ASAN checks in debug.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
The handling for trace streams using HSYNC and FSYNC packets was not
correctly accounting for an odd number of HSYNC packets in a stream,
which could result in read past end of buffer.
Routine changed to ensure handling of 2 byte aligned end of buffer,
and FSYNC across a pair of input buffers.
Reported-by: yabinc <yabinc@google.com> (github issue #50)
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
The reset on 4xFSYNC for frame aligned input data blocks was
not correctly throwing an error on bad input, resulting in
subsequent block overruns
Additionally, a 4XFSYNC at the end of block could also result in
an read past end of block.
Reported-by: yabinc <yabinc@google.com> (github issue #49)
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Currently setting incorrect configuration flags on initialising the decode
tree for the demux formatter does not result in error messages being seen
Fix adjusts the demux module to split init and configure to allow attachement
of logger before configure.
Also reversed the priority of HSYNC and FSYNC alignment values.
Always 2 byte multiples on HSYNC or HSYNC + FSYNC, 4 byte if FSYNC only
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
|
Pathalogical case where 7 ID changes in a frame. all with data associated with previous ID,
may cause buffer overrun issue. Fix by adding in additional buffer for final data element.
Highly unlikely / impossible to occur in valid system generated frame data but could be
triggered by corrupt / invalid input data.
Reported-by: yabinc <yabinc@google.com> (github issue #51)
Signed-off-by: Mike Leach <mike.leach@linaro.org>
|
