From b60dd35e33b622216c12d5804f949df087eb59fa Mon Sep 17 00:00:00 2001 From: David Brownell Date: Wed, 13 Jan 2010 23:33:53 -0800 Subject: User's Guide updates Capture various bits of useful information that have come up on the list but haven't yet gotten into the documentation: - Watchdog timers firing during JTAG debug need attention; - Some chips have special registers to help JTAG debug; - Cortex-M3 stepping example with IRQs and maskisr; - Clarifications re adaptive clocking: not all ARMs do it, and explain it a bit better. Signed-off-by: David Brownell --- doc/openocd.texi | 111 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 88 insertions(+), 23 deletions(-) (limited to 'doc') diff --git a/doc/openocd.texi b/doc/openocd.texi index 3e0b5db..4446046 100644 --- a/doc/openocd.texi +++ b/doc/openocd.texi @@ -942,6 +942,33 @@ handling issues like: @itemize @bullet +@item @b{Watchdog Timers}... +Watchog timers are typically used to automatically reset systems if +some application task doesn't periodically reset the timer. (The +assumption is that the system has locked up if the task can't run.) +When a JTAG debugger halts the system, that task won't be able to run +and reset the timer ... potentially causing resets in the middle of +your debug sessions. + +It's rarely a good idea to disable such watchdogs, since their usage +needs to be debugged just like all other parts of your firmware. +That might however be your only option. + +Look instead for chip-specific ways to stop the watchdog from counting +while the system is in a debug halt state. It may be simplest to set +that non-counting mode in your debugger startup scripts. You may however +need a different approach when, for example, a motor could be physically +damaged by firmware remaining inactive in a debug halt state. That might +involve a type of firmware mode where that "non-counting" mode is disabled +at the beginning then re-enabled at the end; a watchdog reset might fire +and complicate the debug session, but hardware (or people) would be +protected.@footnote{Note that many systems support a "monitor mode" debug +that is a somewhat cleaner way to address such issues. You can think of +it as only halting part of the system, maybe just one task, +instead of the whole thing. +At this writing, January 2010, OpenOCD based debugging does not support +monitor mode debug, only "halt mode" debug.} + @item @b{ARM Semihosting}... @cindex ARM semihosting When linked with a special runtime library provided with many @@ -964,7 +991,12 @@ via the @code{WFI} instruction (or its coprocessor equivalent, before ARMv7). You may want to @emph{disable that instruction} in source code, or otherwise prevent using that state, -to ensure you can get JTAG access at any time. +to ensure you can get JTAG access at any time.@footnote{As a more +polite alternative, some processors have special debug-oriented +registers which can be used to change various features including +how the low power states are clocked while debugging. +The STM32 DBGMCU_CR register is an example; at the cost of extra +power consumption, JTAG can be used during low power states.} For example, the OpenOCD @command{halt} command may not work for an idle processor otherwise. @@ -6699,8 +6731,10 @@ to debug remote targets. Setting up GDB to work with OpenOCD can involve several components: @itemize -@item OpenOCD itself may need to be configured. @xref{GDB Configuration}. -@item GDB itself may need configuration, as shown in this chapter. +@item The OpenOCD server support for GDB may need to be configured. +@xref{GDB Configuration}. +@item GDB's support for OpenOCD may need configuration, +as shown in this chapter. @item If you have a GUI environment like Eclipse, that also will probably need to be configured. @end itemize @@ -6803,6 +6837,24 @@ With that particular hardware (Cortex-M3) the hardware breakpoints only work for code running from flash memory. Most other ARM systems do not have such restrictions. +Another example of useful GDB configuration came from a user who +found that single stepping his Cortex-M3 didn't work well with IRQs +and an RTOS until he told GDB to disable the IRQs while stepping: + +@example +define hook-step +mon cortex_m3 maskisr on +end +define hookpost-step +mon cortex_m3 maskisr off +end +@end example + +Rather than typing such commands interactively, you may prefer to +save them in a file and have GDB execute them as it starts, perhaps +using a @file{.gdbinit} in your project directory or starting GDB +using @command{gdb -x filename}. + @section Programming using GDB @cindex Programming using GDB @@ -6947,36 +6999,48 @@ is jim, not real tcl). In digital circuit design it is often refered to as ``clock synchronisation'' the JTAG interface uses one clock (TCK or TCLK) -operating at some speed, your target is operating at another. The two -clocks are not synchronised, they are ``asynchronous'' +operating at some speed, your CPU target is operating at another. +The two clocks are not synchronised, they are ``asynchronous'' -In order for the two to work together they must be synchronised. Otherwise -the two systems will get out of sync with each other and nothing will -work. There are 2 basic options: +In order for the two to work together they must be synchronised +well enough to work; JTAG can't go ten times faster than the CPU, +for example. There are 2 basic options: @enumerate @item -Use a special circuit. +Use a special "adaptive clocking" circuit to change the JTAG +clock rate to match what the CPU currently supports. @item -One clock must be some multiple slower than the other. +The JTAG clock must be fixed at some speed that's enough slower than +the CPU clock that all TMS and TDI transitions can be detected. @end enumerate @b{Does this really matter?} For some chips and some situations, this -is a non-issue (i.e.: A 500MHz ARM926) but for others - for example some -Atmel SAM7 and SAM9 chips start operation from reset at 32kHz - -program/enable the oscillators and eventually the main clock. It is in -those critical times you must slow the JTAG clock to sometimes 1 to -4kHz. - -Imagine debugging a 500MHz ARM926 hand held battery powered device -that ``deep sleeps'' at 32kHz between every keystroke. It can be -painful. +is a non-issue, like a 500MHz ARM926 with a 5 MHz JTAG link; +the CPU has no difficulty keeping up with JTAG. +Startup sequences are often problematic though, as are other +situations where the CPU clock rate changes (perhaps to save +power). + +For example, Atmel AT91SAM chips start operation from reset with +a 32kHz system clock. Boot firmware may activate the main oscillator +and PLL before switching to a faster clock (perhaps that 500 MHz +ARM926 scenario). +If you're using JTAG to debug that startup sequence, you must slow +the JTAG clock to sometimes 1 to 4kHz. After startup completes, +JTAG can use a faster clock. + +Consider also debugging a 500MHz ARM926 hand held battery powered +device that enters a low power ``deep sleep'' mode, at 32kHz CPU +clock, between keystrokes unless it has work to do. When would +that 5 MHz JTAG clock be usable? @b{Solution #1 - A special circuit} -In order to make use of this, your JTAG dongle must support the RTCK +In order to make use of this, +both your CPU and your JTAG dongle must support the RTCK feature. Not all dongles support this - keep reading! -The RTCK signal often found in some ARM chips is used to help with +The RTCK ("Return TCK") signal in some ARM chips is used to help with this problem. ARM has a good description of the problem described at this link: @url{http://www.arm.com/support/faqdev/4170.html} [checked 28/nov/2008]. Link title: ``How does the JTAG synchronisation logic @@ -7013,8 +7077,9 @@ ARM11 cores use an 8:1 division. Note: Many FTDI2232C based JTAG dongles are limited to 6MHz. You can still debug the 'low power' situations - you just need to -manually adjust the clock speed at every step. While painful and -tedious, it is not always practical. +either use a fixed and very slow JTAG clock rate ... or else +manually adjust the clock speed at every step. (Adjusting is painful +and tedious, and is not always practical.) It is however easy to ``code your way around it'' - i.e.: Cheat a little, have a special debug mode in your application that does a ``high power -- cgit v1.1