Salvador Arroyo [Sat, 3 Nov 2012 11:39:27 +0000 (12:39 +0100)]
mips: optimize mips32_pracc_read_regs() code
Current code needs 101 pracc accesses for this function, this code needs 12 less.
There is a singularity in this code, is the only function that restore
a register from param out instead from pracc stack. Obviously the register
was previously stored at param out. This save 2 pracc accesses.
Change-Id: Ie95b6f983a3198dafc0eab2dd5acc11f871a8d83 Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/958 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Salvador Arroyo [Sat, 3 Nov 2012 10:29:46 +0000 (11:29 +0100)]
mips: optimize mips32_pracc_write_regs() code.
All the the loads are done with lui and ori instructions, there is
no need to save any register, they will be overwritten.
Like in the previous patch, for speed optimization in write code,
same instructions can be saved if the lower half word or the upper
half word is 0.
If the lower half word is 0, it can be loaded with only a lui instruction.
If the higher half word is 0 it can be done with an ori instruction with register 0.
This code saves 10 pracc accesses at a minimum, and 40 at a maximum,
obviously if register 2 to 31 are 0 or a half word is 0
Current code needs 91 pracc accesses.
Change-Id: I892c5b440191d0c7a474c96845d41c373b7fc637 Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/957 Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk> Tested-by: jenkins
Salvador Arroyo [Fri, 2 Nov 2012 15:08:31 +0000 (16:08 +0100)]
mips: optimize write code for speed
All the writes are done by the new function mips32_pracc_write_mem_generic().
The code is similar to the read generic code.
The reuse of register 15 as memory base address saves 3 pracc accesses.
The first write takes 13(12) pracc accesses and for additional writes 3(2).
Loading miniprograms should take 25% less time and loading fastdata transfer
handler code should be over 2x faster.
Change-Id: Ia3b24ba084af33be99da19f00a7fd4d1b291f350 Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/956 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Salvador Arroyo [Fri, 2 Nov 2012 09:14:58 +0000 (10:14 +0100)]
mips: optimize read code for speed
Really nothing new that not explained in previous patches.
The code is expanded as needed, there are no loops in pracc code.
For the first value pracc accesses are reduced from 39 to 16
and for aditional values from 10 to 3.
dump_image should work around 3x faster.
Change-Id: I37c9b13395c09eb52a91f10cdb6cbaedef8ab98b Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/955 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Salvador Arroyo [Thu, 1 Nov 2012 22:22:32 +0000 (23:22 +0100)]
mips: optimize mips32_pracc_read_u32() function
This function is highly optimized, there is not much to
improve.
Loading the base address for pracc access with the new
defined MIPS32_PRACC_BASE_ADDR saves one instruction.
The memory address is loaded in too steps. First the upper
address is loaded. The lower address is passed as an offset in
the memory load instruction.
The offset is signed, if the lower address is in the range of
0x8000 to 0xffff the offset is a negative value, and the upper
address must be incremented by 1.
Pracc accesses are now 12 instead of 14.
Change-Id: I286945b240ed5c5d5cc540780a41a8a5fa075da3 Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/952 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Salvador Arroyo [Thu, 1 Nov 2012 19:55:28 +0000 (20:55 +0100)]
mips: optimize CP0 read/write code
MIPS32_PRACC_BASE_ADDR is defined as 0xFF200000. Now is
possible to load the base address with a lui instruction and
only one pracc access.
Offsets to the pracc code addresses are defined to simplify the code
and probably make it a bit more readable or self-explained.
Change-Id: I853dd2d7fad52745931cc6e6be68c0ae156d897e Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/951 Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk> Tested-by: jenkins
Salvador Arroyo [Thu, 1 Nov 2012 14:45:12 +0000 (15:45 +0100)]
mips: code clean up in mips_m4k_debug_entry() function
The function mips_ejtag_read_debug() is defined in mips_ejtag.c
and is called only by mips_m4k_debug_entry() for reading the
CP0 debug register. The comment in this function is obviously wrong.
There is a generic function to read CP0 registers with similar code.
A call to mips32_cp0_read() should work in the same way.
The purpose of reading the debug register is to test if the DSS
bit is set and clear the SSt bit.
It is faster and easier if the SSt bit is cleared without any check.
Remark: DSS bit set only means that a debug single-step exception
ocurred, but it is not possible to step over a sdbbp instruction,
in this case DSS will not be set and the SSt bit not cleared by code.
Resume command at another address will step, so really the behavior
is not the same.
Change-Id: Ibd35f80e0f7669976d96f4ed813830cecf587971 Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/950 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Salvador Arroyo [Tue, 30 Oct 2012 22:15:35 +0000 (23:15 +0100)]
mips: optimize mips_ejtag_step_disable() code
The code is a bit large compared to mips_ejtag_step_enable().
With the mips32 xori instruction the code can be
reused.
The number of pracc accesses are reduced from 18 to 7.
Change-Id: If3974ebd64da4461c22b089796646990e68e1b72 Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/944 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Salvador Arroyo [Sat, 1 Sep 2012 11:50:39 +0000 (13:50 +0200)]
Pic32mx.cfg: Change system clock to 8Mhz after reset-init.
As for openocd 0.6.0-rc2 the function mips32_pracc_fastdata_xfer()
should now work at a scan frequency up to 1200Khz.
Mainly usefull to increase programming speed.
Also verify_image should be slightly faster.
Change-Id: I1e9b2be73690a4597e2f6ba069c1205026850f07 Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/805 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Freddie Chopin [Fri, 9 Nov 2012 22:18:23 +0000 (23:18 +0100)]
Revert "mpsse: Always perform a general reset of the MPSSE in mpsse_open()"
This reverts commit 452248af1d06cb1140b85f53ef4fdee1c746d807. This change
breaks all non-high speed adapters. The patch was not tested and did not get
any review.
Spencer Oliver [Mon, 5 Nov 2012 13:04:15 +0000 (13:04 +0000)]
target: add async algorithm timeout
An issue was observed when using an async algorithm with a target that had
not been previously reset beforehand. The target would enter a infinite
loop within target_run_flash_async_algorithm.
Add a timeout that will at least prevent this issue from happening. and also
suggest the user resets the target.
Change-Id: I5277e0d64e252d3d353e8d5bc9889a37fdc63060 Signed-off-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-on: http://openocd.zylin.com/949 Tested-by: jenkins Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Peter Horn [Mon, 8 Oct 2012 20:06:07 +0000 (22:06 +0200)]
cortex_m: Fix single stepping will not return to debug mode sometimes
This occurs when stepping past a breakpoint on a even address with
maskisr option set to auto
With -d3 the following log message appears in this case:
"Debug : Interrupt handlers didn't complete within time,
leaving target running"
Cause : Given a breakpoint is set on the lower half word and the PC is on
the upper half word. When another breakpoint is now set on the current PC
then resuming the core will not result in a break on the newly set
breakpoint. This has been observed on a STM32F1x, STM32F2x (CM3) but not
on a STM32F0x (CM0). It's not clear if this is a STM32F1/F2 only or a
general CM3 problem.
Change-Id: I384813f3bfdf935373b5e23cdb2d7f243c70cc00 Signed-off-by: Peter Horn <peter.horn@bluewin.ch>
Reviewed-on: http://openocd.zylin.com/864 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Gianluca Renzi [Thu, 25 Oct 2012 12:29:33 +0000 (14:29 +0200)]
Added support for NXP LPC1850 Microcontroller
Added a new configuration file for LPC18xx based boards, such as
HitexLPC1850RevA Evaluation Board, and all other based on the
same microcontroller by NXP.
Gianluca Renzi [Thu, 25 Oct 2012 12:26:42 +0000 (14:26 +0200)]
Generic LPC1850 board w/ SPIFI flash.
This config file is intended as an example of how to
use the lpcspifi flash driver, but it should be functional
for most LPC1850 boards utilizing SPIFI flash.
Spencer Oliver [Fri, 26 Oct 2012 11:59:29 +0000 (12:59 +0100)]
gdb: fix broken qCRC packet handling
The rtos layer was incorrectly handling a qCRC packet as a qC packet.
Make sure we check for the qCRC packet and return unhandled so the gdb
server gets a chance to handle it.
This packet is used in the gdb compare-sections cmd.
Change-Id: I21f8e5fa7225fccd13d65cf9e40186895065a7e3 Signed-off-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-on: http://openocd.zylin.com/933 Tested-by: jenkins Reviewed-by: Matthias Blaicher <matthias@blaicher.com> Reviewed-by: Peter Stuge <peter@stuge.se>
rtos: Fix wrong allocation in linux_get_symbol_list_to_lookup
linux_get_symbol_list_to_lookup allocates to few memory. On 64 bit
systems the error did not show due to char* being twice its size,
leaving accidentally enough space.
This patch makes linux_get_symbol_list_to_lookup behave identical
to all other RTOS.
Freddie Chopin [Sat, 6 Oct 2012 07:49:28 +0000 (09:49 +0200)]
Fix serious bug in LPC2xxx/LPC17xx flash algorithm.
Flash algorithm for LPC17xx/LPC2xxx was trying to "reuse" previously
allocated working area on next flashing which is not possible -
working areas are freed automatically on reset. This caused all but
first flashing attempts to fail. As there is no point in storing pointer
to working area, it was converted to local variable.
This patch adds ChibiOS/RT support. This patch requires at least
ChibiOS/RT development version starting from SVN revision 4734.
Note, that the Thread structures depend not only on the target
but also on the ChibiOS configuration at build time.
To correct this ChibiOS includes a new "memory signature" which
specifies the offsets.
Special thanks go to Peter Stuge and Spencer Oliver for their
continous input and feedback to this patch.
rtos: Don't crash on qSymbol GDB packet when no RTOS is configured
Commit 43902905bbc8fdc9d764f3843d697161d9b5fd92 fixed a bug but also
introduced a regression. The RTOS GDB packet handler is always called,
not only when an RTOS is actually configured, so it is important to
check if an RTOS has been configured or not before actually processing
the qSymbol packet.
Change-Id: I1aed54f6c2817e1ebf99ddcda051df4554ea5a3a Signed-off-by: Peter Stuge <peter@stuge.se>
Reviewed-on: http://openocd.zylin.com/907 Tested-by: jenkins
Peter Horn [Mon, 8 Oct 2012 18:55:01 +0000 (20:55 +0200)]
rlink: Fix DTC command timeout
With the current timeout setting i = 10 in drtc_run_timeout()
I get "Error: too many retries waiting for DTC status" when
loading a program into the FLASH of an STM32F1.
By experimentation a value of i = 22 was found to be the minimum
on my system. Therefore the value has been increased to i = 50.
Change-Id: Ib67fc648ccaad305871b81c2c39e49de53c330a0 Signed-off-by: Peter Horn <peter.horn@bluewin.ch>
Reviewed-on: http://openocd.zylin.com/863 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
The erase time for the option byte page is not directly specified but is
assumed to be the same as the other pages (or mass erase) which is 20 to
40 ms. The current timeout value is 10 which means 10 ms plus the time to
poll the status flag that many times.
With faster interfaces or drivers (such as when using the ftdi driver
instead of the ft2232 driver) the adapter delay is not enough in some
cases, unless the jtag freq is reduced as a workaround. The result is a
"timed out waiting for flash" error when trying to write the options.
Increase the timeout to a minimum of 100 ms, which is in line with the
other erase timeouts. Also make defines of both the erase and the program
timeouts.
Change-Id: Ia86e71505033c52b60ef30092000689fbb547a18 Signed-off-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Reviewed-on: http://openocd.zylin.com/902 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk>
Peter Stuge [Thu, 4 Oct 2012 13:17:53 +0000 (15:17 +0200)]
rtos: Rewrite rtos_qsymbol() and fix auto-detect false positive
Matthias Blaicher submitted a patch at http://openocd.zylin.com/#/c/891/
to fix the false positive; when no RTOS was detected OpenOCD used the
last RTOS in the list.
While reviewing the code affected by Matthias' patch a rewrite seemed
appropriate, to make the code readable.
Matthias has abandoned his change and this change also fixes the false
positive.
Peter Stuge [Thu, 4 Oct 2012 12:35:18 +0000 (14:35 +0200)]
rtos: Rewrite rtos_try_next() for readability
The new code is almost functionally equivalent to the old.
The function now returns 0 instead of -1 if target->rtos has not yet
been allocated. All call sites only test for success, and in practise
that is also the only thing that matters; if the function successfully
iterated to the next RTOS or not.
Other than that the only difference is that the code is now readable.
Many thanks to Matthias Blaicher for the fix to the iteration error!
In FLASHD_ErasePages AT91C_EFC_FCMD_EPA is used to erase sectors.
According to the datasheet FARG[15:2] defines the page from which
the erase will start.This page must be modulo 4, 8, 16 or 32
according to the number of pages to erase. FARG[1:0] defines the
number of pages to be erased. Previously (firstpage << 2) was used
to conform to this, seems it should not be shifted... Changed it
to (firstPage) | erasePages.
Spencer Oliver [Thu, 4 Oct 2012 13:01:14 +0000 (14:01 +0100)]
build: fix broken ftd2xx bus blaster
If configure is executed without --enable-ft2232_ftd2xx then the bus blaster
or presto will fail to build with unresolved external ftd2xx_status_string.
Make sure we run the ftd2xx build test if --enable-usb_blaster_ftd2xx is enabled.
Change-Id: I09d270d6fcd083d77f6785b8969d9acb3dfef11d Signed-off-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-on: http://openocd.zylin.com/892 Tested-by: jenkins Reviewed-by: Peter Stuge <peter@stuge.se>
Spencer Oliver [Mon, 24 Sep 2012 11:29:47 +0000 (12:29 +0100)]
ftdi: fix adapter_init rclk fallback
adapter_init expects jtag_get_speed (via ftdi_khz) to return a valid
fallback speed if the adapter does not support rclk. The call was failing
and so was the rest of the adapter init.
The makes the new ftdi driver emulate the old ftdi driver.
George Harris [Fri, 17 Aug 2012 00:53:32 +0000 (17:53 -0700)]
Added SPIFI flash driver, algorithms, and docs
Added a flash driver designed to allow program/erase of
memory-mapped SPI flash chips for LPC43xx/LPC18xx family
micros. This driver includes three algorithms - erase,
write, and SPIFI peripheral initialization (to allow
memory-mapped access after a reset). The driver has been
added to the flash driver table (drivers.c), and the
OpenOCD documentation has been updated to include the flash
driver configuration command.
George Harris [Fri, 17 Aug 2012 00:35:39 +0000 (17:35 -0700)]
SPI nor drivers refactor
Moved common SPI flash driver code (device table,
common commands) into flash/nor/spi.c and spi.h.
Updated flash/nor/stmsmi.c to reflect this refactor.
In at91sam3.c for Atmel SAM3 flash support, there are arithmetic errors in the functions sam3_page_read() and sam3_page_write().
Address locations are computed incorrectly due to an extra addition operation. This leads to memory locations being skipped during
flash writes and reads.
Smaller programs are written successfully into flash, with memory gaps, while larger programs of legitimate size fail because the
skipped memory is not utilized and therefore unavailable.
The changes address this condition, and have been tested with an Atmel SAM3X-EK evaluation board.
Spencer Oliver [Thu, 30 Aug 2012 11:53:34 +0000 (12:53 +0100)]
cortex_m: suggest using hardware srst if VECTRESET used
If the target does not support SYSRESETREQ we fall back to using VECTRESET.
This however does not reset the peripherals and we issue a warning to the user
to suggest using a reset-init script.
Also suggest that using hardware srst will give them the same functionality
as using SYSRESETREQ.
Change-Id: Ie1781c4b849fed66c52222e6539735537c879fb3 Signed-off-by: Spencer Oliver <spen@spen-soft.co.uk>
Reviewed-on: http://openocd.zylin.com/802 Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com> Tested-by: jenkins Reviewed-by: Freddie Chopin <freddie.chopin@gmail.com>
Salvador Arroyo [Sat, 25 Aug 2012 14:22:58 +0000 (16:22 +0200)]
Pic32mx: make row programming work with any offset
In function pic32mx_write_block() if the parameter
offset is not a multiple of row size the row offset
(offset % row_size) will be ignored by the flash
controller, shifting the code to the beginning of
the row.
Word programming gets it right.
Change-Id: I134913e3d533688f791bbcb0c6e8983524197f3c Signed-off-by: Salvador Arroyo <sarroyofdez@yahoo.es>
Reviewed-on: http://openocd.zylin.com/796 Tested-by: jenkins Reviewed-by: Spencer Oliver <spen@spen-soft.co.uk> Reviewed-by: Freddie Chopin <freddie.chopin@gmail.com>
projects / openocd / log