This patch fixes a bug where target fails to resume after completing GDB FileIO.
We need to update target last run control information to decide resumption. This
was not being done for vcont packets.
Change-Id: I44bea31720f8b877dba97d77a202303d546ea5bd
Signed-off-by: Omair Javaid <omair.javaid@linaro.org>
Reviewed-on: http://openocd.zylin.com/4539
Tested-by: jenkins
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
In 2016, ARM released the second edition of the semihosting specs
("Semihosting for AArch32 and AArch64"), adding support for 64-bits.
To ease the reuse of the semihosting logic for other platforms
(like RISC-V), the semihosting code was isolated from the ARM
target and updated to the latest specs.
The new code is already in use since January (in GNU MCU Eclipse
OpenOCD) and no problems were reported, neither for ARM nor for
RISC-V targets, after more than 7K downloads.
The 2 new files were formatted with uncrustify.
Change-Id: Ie84dbd86a547323bb8a5d24eab68fc7dad013d96
Signed-off-by: Liviu Ionescu <ilg@livius.net>
Reviewed-on: http://openocd.zylin.com/4518
Tested-by: jenkins
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Use sector sizes instead of bank size.
Detect a gap between sectors and emit xml blocks accordingly.
Detect sector overflow over the bank size.
Change-Id: If0e0e44b0c3b93067b4d717c9c7b07c08582e57b
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: http://openocd.zylin.com/4436
Tested-by: jenkins
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
With this option enabled (it's disabled by default) errors accessing
registers are returned to gdb. Otherwise they are ignored and success is
reported to gdb. (This is the current behavior.)
We want this for RISC-V, but there's still some cleanup that needs to be
done before that can be upstreamed.
Signed-off-by: Tim Newsome <tim@sifive.com>
Change-Id: I7e56109ea52d18b780c14a07fb35f9e6e8979da4
Reviewed-on: http://openocd.zylin.com/4452
Reviewed-by: Steven Stallion <sstallion@gmail.com>
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Tested-by: jenkins
gdb assumes that a rtos can make any thread active at will in response
to a 'Hg' packet. It further assumes that it needs to step-over after
setting a breakpoint on frame #0 of any non-current thread. Both
assumptions are not valid for an actual rtos. We fake the step-over to
not trigger an internal error in gdb. See
https://sourceware.org/bugzilla/show_bug.cgi?id=22925 for details.
Change-Id: Ida60cd134033c1d58ada77b87fe664a58f61e2c0
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Reviewed-on: http://openocd.zylin.com/4448
Tested-by: jenkins
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Normally, when a ctrl-c is received from gdb, a SIGINT is reported back
unconditionally to tell gdb that the target has stopped in response.
However when a rtos support was configured, the rtos awareness overwrote
the signal with an actual thread state, which gdb then ignored and got
stuck without the user able to interrupt.
Change-Id: I40fd62333e020a8c4d9df0079270e84df9c77f88
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Reviewed-on: http://openocd.zylin.com/4445
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Although the leak is negligible, the clean heap on exit will ease
valgrind testing.
Change-Id: I3a7a9c8e8dc7557aa51d0b9caa244537e5e7007d
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: http://openocd.zylin.com/4410
Tested-by: jenkins
This patch adds support to generate multiple nested architecture defined
data types in gdb target xml generated by openOCD. Architecture defined
structs, unions, vectors nested in one or more architecture defined
types can be generated now.
Example:
<vector id="v2d" type="ieee_double" count="2"/>
<vector id="v2u" type="uint64" count="2"/>
<vector id="v2i" type="int64" count="2"/>
<union id="vnd">
<field name="f" type="v2d"/>
<field name="u" type="v2u"/>
<field name="s" type="v2i"/>
</union>
Change-Id: I0f3c5c6daf3d22cde7e4b7b4165d2e97e25872f7
Signed-off-by: Omair Javaid <omair.javaid@linaro.org>
Reviewed-on: http://openocd.zylin.com/4372
Tested-by: jenkins
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
this patch contains several changes to run control and state
handling together with gdb:
- graceful handling of target/gdb desync on resume, step and halt
- a default gdb-attach event executing the "halt" command, to meet gdb
expectation of target state when it attaches
- call target_poll() after Ctrl-C command from gdb
- call target_poll() after resume and step through a vCont packet
- fix log message forwarding on vCont stepping, also move an aarch64
log message from INFO to DEBUG level to prevent messing up the gdb
console during source-line stepping
- fix oversight in vCont support that messes up breakpoint handling
during stepping
Change-Id: Ic79db7c2b798a35283ff752e9b12475486a1f31a
Fixes: d301d8b42f
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Reviewed-on: http://openocd.zylin.com/4432
Tested-by: jenkins
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
make sure the RTOS thread database is updated early on a new
gdb connection.
Change-Id: I4da9ef30f8634263d697116cefc47976cd1970ad
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Reviewed-on: http://openocd.zylin.com/4000
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Tested-by: jenkins
The RTOS handlers present OS threads to gdb but the openocd
target layer only knows about CPU cores (hardware threads).
This patch allows closing this gap inside the RTOS handler.
The default implementation just returns the current core, but
a RTOS handler can provide its own function that associates a
an OS thread with a core.
Change-Id: I12cafe50b38a38b28057bc5d3a708aa20bf60515
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Reviewed-on: http://openocd.zylin.com/3997
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Tested-by: jenkins
Make gdb use target support for single-stepping if available.
Change-Id: Ie72345a1e749aefba7cd175ccbf5cf51d4f1a632
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Reviewed-on: http://openocd.zylin.com/3833
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
When listening on port 0, the system will assign a random open port. We
use this to run multiple OpenOCD instances against multiple simulators
as part of regression testing. This mechanism means the various test
instances don't have to coordinate to ensure they don't reuse any ports.
The required changes are minimal:
1. Don't increment the port number when it's 0.
2. Print out which port was assigned by the system.
Change-Id: I404c801fc405e9d8eb8420562c02e78d4db6242f
Signed-off-by: Tim Newsome <tim@sifive.com>
Reviewed-on: http://openocd.zylin.com/4316
Tested-by: jenkins
Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
When multiple targets are declared, it's not always obvious which
target the connection was made for, this can lead to very confusing
errors.
Reported by zjason on IRC.
Change-Id: I52906320394e89cb6cfe82054a3f94b27c999689
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Reviewed-on: http://openocd.zylin.com/4135
Tested-by: jenkins
Accept 64 bit addresses from GDB read memory packet.
Also allow breakpoint/stepping addresses to take 64bit values.
Change-Id: I9bf7b44affe24839cf30897c55ad17fdd29edf14
Signed-off-by: David Ung <david.ung.42@gmail.com>
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Define a target_addr_t type to support 32-bit and 64-bit addresses at
the same time. Also define matching TARGET_PRI*ADDR format macros as
well as a convenient TARGET_ADDR_FMT.
In targets that are 32-bit (avr32, nds32, arm7/9/11, fm4, xmc1000)
be least invasive by leaving the formatting unchanged apart from the
type;
for generic code adopt TARGET_ADDR_FMT as unified address format.
Don't silently change gdb formatting here, leave that to later.
Add COMMAND_PARSE_ADDRESS() macro to abstract the address type.
Implement it using its own parse_target_addr() function, in the hopes
of catching pointer type mismatches better.
Add '--disable-target64' configure option to revert to previous 32-bit
target address behavior.
Change-Id: I2e91d205862ceb14f94b3e72a7e99ee0373a85d5
Signed-off-by: Dongxue Zhang <elta.era@gmail.com>
Signed-off-by: David Ung <david.ung.42@gmail.com>
[AF: Default to enabling (Paul Fertser), rename macros, simplify]
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Matthias Welwarsky <matthias.welwarsky@sysgo.com>
Simplify hexify() and do not longer use 0 as special case for the
parameter 'count' to determine the string length of the binary input.
Instead, use strlen() outside of the function if needed.
Additionally, fix the return value and return the length of the
converted string. The old function always returned 2 * count.
Also, use more appropriate data types for the function parameters and
add a small documentation.
Change-Id: I133a8ab786b8f7c1296afcaf9c0a0b43881e5112
Signed-off-by: Marc Schink <openocd-dev@marcschink.de>
Reviewed-on: http://openocd.zylin.com/3793
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
This patch adds support for the qXfer:threads:read packet. In addition
to providing a more efficient method of updating thread state, recent
versions of GDB (7.11.1 and up) can also report remote thread names.
While thread names are not enabled in this patch due to its limited
applicability at the moment, it can be enabled at a later date with
little effort.
As a part of revamping how threads are presented to GDB, extra info
strings for each of the supported RTOSes were updated to match
conventions present in the GDB source code. For more information, see
remote_threads_extra_info() in remote.c. This results in a much smoother
experience when interacting with GDB.
It is also worth mentioning that use of qXfer:threads:read works around
a number of regressions in older versions of GDB regarding remote thread
display. Trust me, it's great.
Change-Id: I97dd6a93c342ceb9b9d0023b6359db0e5604c6e6
Signed-off-by: Steven Stallion <stallion@squareup.com>
Reviewed-on: http://openocd.zylin.com/3559
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-by: Paul Fertser <fercerpav@gmail.com>
To speed up downloads, OpenOCD sends gdb OK when a write is received,
even before the write has actually occurred. The failure is then
returned for the next write. That leads to the following confusing
behavior:
```
(gdb) p/x *((int*)0xdeadbeef)=8675309
$2 = 0x845fed
(gdb) p/x *((int*)0x80000000)=6874742
Cannot access memory at address 0x80000000
```
While it's actually the first write that failed.
This change hacks around this problem by not sending OK for small writes
(len<8) until the write has actually occurred. This does not impact
download speed, since during downloads (almost) all writes will have
much larger length.
Change-Id: I1f8b9bb19b0707487f840df6871e372e4ba228dd
Signed-off-by: Tim Newsome <tim@sifive.com>
Reviewed-on: http://openocd.zylin.com/3803
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
The current implementation is not suitable for user provided data
because it does not detect invalid inputs in many cases. For example,
the string "aa0xbb" is successfully converted to the 3 bytes: 0xaa,
0x00 and 0xbb. An other example is "aabi" which is successfully
converted to the 2 bytes: 0xaa and 0x0b. Both are obviously incorrect.
Make unhexify() robust on invalid data and use more appropriate data
types for its parameters. Also, add a small documentation for the
function.
Change-Id: Idb799beb86fc608b066c8a76365021ed44c7f890
Signed-off-by: Marc Schink <openocd-dev@marcschink.de>
Reviewed-on: http://openocd.zylin.com/3792
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
As per the documentation, used "disabled" as the value to disable, as this
is the same value to disable the telnet and tcl server.
Change-Id: Idc4a8580098ec1107dcc6e1f59e817ecdebc38ac
Signed-off-by: Sandeep Mistry <s.mistry@arduino.cc>
Reviewed-on: http://openocd.zylin.com/3175
Tested-by: jenkins
Reviewed-by: Cristian Maglie
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Although the documentation suggested this worked, and it is implemented
for tcl_port and telnet_port, the directive was not recognized for
gdb_port.
Change-Id: I38d95ee879ec3f6d551603b7313749a21e0e498e
Signed-off-by: Peter A. Bigot <pab@pabigot.com>
Reviewed-on: http://openocd.zylin.com/3637
Tested-by: jenkins
Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
I couldn't make OpenOCD to work with GDB. I was always getting this in GDB:
(gdb) target remote localhost:3333
Remote debugging using localhost:3333
Ignoring packet error, continuing...
Ignoring packet error, continuing...
Ignoring packet error, continuing...
Ignoring packet error, continuing...
Malformed response to offset query, timeout
(gdb)
While debugging gdb remote protocol, I have seen that gdb responds with:
w ++$?#3f
And those two '+' seems to confuse the OpenOCD parser, if it sees another
'+' sign it emits the DEBUG output and sets the noack_mode to 2. The
problem is that we weren't even IN noack mode, this was set to 0 and then
it explicitly sets it to 2 and thus turning the noack mode on.
Change-Id: If267c9226e57fa83121ded09cf69829f8f0b4b93
Signed-off-by: Matej Kupljen <matej.kupljen@gmail.com>
Reviewed-on: http://openocd.zylin.com/2545
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-by: Paul Fertser <fercerpav@gmail.com>
This allows GDB to automatically switch to the thread that has
been interrupted and show you where it has stopped.
Change-Id: Icb9500dc42a61eb977e9fac55ce9503c9926bf5d
Signed-off-by: Jon Burgess <jburgess777@gmail.com>
Reviewed-on: http://openocd.zylin.com/2303
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
We always have feature names defined by string literals and the
standard guarantees static storage duration for them. Hence, there's
no need duplicating and then freeing them.
Valgrind-tested.
Change-Id: I1b77f966c548e3694141c63bd8680735f0f47505
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Reviewed-on: http://openocd.zylin.com/2028
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Introduced by 537b06a81 (free non-malloced memory).
Rewrite to use standard C string routines and make returning annex
optional since it's not currently used.
Change-Id: Idf3698a482dfeff7fa5ea1660fd89122eb80b68d
Signed-off-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Reviewed-on: http://openocd.zylin.com/2023
Tested-by: jenkins
Reviewed-by: Paul Fertser <fercerpav@gmail.com>
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
commit da0d1e37 did not merge correctly, causing the build to fail.
Change-Id: I3f525054bb38b7ee29bf27309bb2e6a5bb8329c7
Signed-off-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-on: http://openocd.zylin.com/2020
Tested-by: jenkins
v4:
- changed first line of commit message
v3:
- added extra LOG_ERROR() message
v2:
- Added missing "goto error"
- free also the on extra element of features[]
In contrast to target_get_gdb_reg_list(), the list returned by
get_reg_features_list() consists of items which are itself
malloc'ed.
--> Free the list items prior freeing the list itself.
Additionally:
- gdb_generate_target_description():
o Do error handling similar as gdb_get_target_description_chunk() does.
- gdb_get_target_description_chunk()
o **features must be initialised prior an "goto error" can happen
Change-Id: Iad07824618c51084e0aa0499ee6fc96198b320f0
Signed-off-by: Christian Eggers <ceggers@gmx.de>
Reviewed-on: http://openocd.zylin.com/1917
Tested-by: jenkins
Reviewed-by: Trevor Woerner <trevor.woerner@linaro.org>
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
v2:
- Split work into separate patches
The received packet will not be altered in any of the processing functions.
Some it can be made "const".
Change-Id: I7bb410224cf6daa74a6c494624176ccb9ae638ac
Signed-off-by: Christian Eggers <ceggers@gmx.de>
Reviewed-on: http://openocd.zylin.com/1919
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Make a temporary copy of argument "buf" before modifying it.
This requires also returning annex as copy of "buf".
This change is necessary in order to make packet[] "const".
Change-Id: I41eddc2edba1a88384aa7f5591fe50f6ee6a135c
Signed-off-by: Christian Eggers <ceggers@gmx.de>
---
Changelog:
v4:
- Initialize annex to NULL
v3:
- Return "annex" as copy instead of removing it.
Reviewed-on: http://openocd.zylin.com/1924
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
v2:
removed curly braces around single statement if-block
packet[] has already been zero terminated in gdb_input_inner()
This change is necessary in order to make packet[] "const".
Change-Id: I978bbe52d151a63574db77fb747f596da256d377
Signed-off-by: Christian Eggers <ceggers@gmx.de>
Reviewed-on: http://openocd.zylin.com/1922
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Comment in gdb_server code mistakenly attributed constant warnings
about unexpected acknowledgement to the ACK packet sent by GDB when
establishing connection. However that is not the case, OpenOCD
gdb_server is handling this packet correctly without an warnings. This
warning instead was caused by the ACK packet which GDB sends right after
going into noack mode. Because OpenOCD gdb_server is already in noack
mode at this state, it emits warning about it. This is a documented GDB
behaviour
(https://sourceware.org/gdb/onlinedocs/gdb/Packet-Acknowledgment.html)
so there is no reason to scare users with warnings in this case.
This patch basically introduces two-level noack mode: after receiving
QStartNoAckMode gdb_server will set noack_mode to 1, then it will
receive this last ACK packet, but instead of printing a warning,
noack_mode will be increased to 2. Should there be any other ACK packets
after that, they will be properly reported by warning. All other code
that relies on noack_mode checks it for "!= 0", so there will be no
difference if it is 1 or 2.
Change-Id: I0e9c57fd93293bfe010390db2f3f161528b11d86
Signed-off-by: Anton Kolesov <anton.kolesov@synopsys.com>
Reviewed-on: http://openocd.zylin.com/1895
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
This prevents invalid free(NULL) (that crashes OpenOCD on desktop
distros).
With radare2 it's now a bit more useful, memory access works if I seek
to the correct address, can't test further as it apparently lacks thumb2
support.
Change-Id: I6ec32d09fd52dab53ba765d7f7519baa1f55d973
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Reviewed-on: http://openocd.zylin.com/1853
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Review and modify to conform to C99 integer types format specifiers.
Use arm-none-eabi toolchain to build successfully.
Change-Id: If855072a8f88886809309155ac6d031dcfcbc4b2
Signed-off-by: Hsiangkai Wang <hsiangkai@gmail.com>
Signed-off-by: Hsiangkai <hsiangkai@gmail.com>
Reviewed-on: http://openocd.zylin.com/1794
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
get_flash_bank_by_addr() iterates through all flash banks
trying to auto_probe() every bank, even if bank can belongs to
target other than requested, and this other target can be
in non-halted state, which leads to error message and
operation abort.
Same situation in gdb_new_connection() and gdb_memory_map():
get_flash_bank_by_num() tries to auto_probe() requested bank,
so first get bank by get_flash_bank_by_num_noprobe(), check
if it belongs to current connection's target and skip
get_flash_bank_by_num() (actually autoprobing) if not.
Change-Id: I48b3f93dddcd9283394f14ad2de3248397c75ab3
Signed-off-by: Sergey A. Borshch <sb-sf@users.sourceforge.net>
Reviewed-on: http://openocd.zylin.com/1813
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Check reg_list[i]->feature for NULL pointers before checking feature->name
in order to avoid NULL access.
Change-Id: If20401a6c5aa5e1e29f0fcf9bb95585b49658832
Signed-off-by: Luca BRUNO <lucab@debian.org>
Reviewed-on: http://openocd.zylin.com/1809
Tested-by: jenkins
Reviewed-by: Peter Stuge <peter@stuge.se>
Reviewed-by: Jens Bauer <jens@gpio.dk>
Don't use const on pointers that hold heap allocated data, because that
means functions that free them must cast away the const.
Do use const on pointer parameters or fields that needn't be modified.
Remove pointer casts that are no longer needed after fixing the constness.
Change-Id: I5d206f5019982fd1950bc6d6d07b6062dc24e886
Signed-off-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Reviewed-on: http://openocd.zylin.com/1668
Tested-by: jenkins
Reviewed-by: Mathias Küster <kesmtp@freenet.de>
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Fix use of uninitialized pointer passed to xml_printf, caught by valgrind.
Make sure a failed gdb_generate_target_description frees all allocated
memory and avoids touching its out argument.
Plug memory leak and check allocation in handle_gdb_save_tdesc_command.
Change-Id: I30e20f6760a6215b1b4496304acdf47347eed829
Signed-off-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Reviewed-on: http://openocd.zylin.com/1645
Tested-by: jenkins
Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-by: Franck Jullien <franck.jullien@gmail.com>
This enables gdb target description support by default, it will check that
this is supported by the active target.
Change-Id: Iea5a5b94c5e373af30dacb780020b79536caa1e3
Signed-off-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-on: http://openocd.zylin.com/1618
Tested-by: jenkins
Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Add support for the default gdb register description type "int" and "float".
When this is given to gdb it will use the bitsize to determine the reg size.
Change-Id: Iaeed594d1feece54044128eae1baff9858bdcae0
Signed-off-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-on: http://openocd.zylin.com/1622
Tested-by: jenkins
Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
