What would be the simplest and most portable way (in the sense of only having to copy a few files to the target machine, like procdump is) to generate a kernel dump that has handle information?
procdump has the -mk option which generates a limited dump file pertaining to the specified process. It is reported in WinDbg as:
Mini Kernel Dump File: Only registers and stack trace are available. Most of the commands I try (!handle, !process 0 0) fail to read the data.
Seems that officially, windbg and kd would generate dumps (which would require kernel debugging).
A weird solution I found is using livekd with -ml: Generate live dump using native support (Windows 8.1 and above only).. livekd still looks for kd.exe, but does not use it :) so I can trick it with an empty file, and does not require kernel debugging. Any idea how that works?
LiveKD uses the undocumented NtSystemDebugControl API to capture the memory dump. While you can easily find information about that API online the easiest thing to do is just use LiveKD.
My dogfood machine hit a deadlock earlier this morning in a debug build. It's at a breakpoint in gdb now. What is my best next step before I kill it?
Turn off pagination: set pagination off
Turn on logging: set logging on
Get a stack trace: thread apply all bt
Get a core dump in case we want more info: generate-core-file
Confirm the core file and log file are reasonable (the core file should be binary, the log file should be text, both should be non-empty) before exiting from gdb.
Upload the log file to a Gist so we can all see it.
My question is how operating system loads
User space application to RAM. I know how
Bootloader works when we first turn computer on Bios simply reads 512 kb data till aa55 bootloader signature and loads bootloader to ram. Do regular userspace programms are handled in this way? If yes how? Because bootloader activated by bios and how user space program handled by operating system? More specifacally how execv() load program to RAM and start execution point for user space ?
Thanks in advance
Userspace programs are not handled like the bios, the Kernel will be involved in running a userspace program.
In general:
When a program is executed in shell, the shell will invoke system calls to create a new task in a new address space, read in the executable binary, and begin executing it.
To understand the details, you need to understand:
The elf format. Of course there are also other formats which can be used in Linux, elf is just the most common one, and a good starting point. Understanding elf will help you understand how the kernel loads the executable binary into memory precisely.
Linux process management; this will help you to understand how a program starts to run.
Reading the related codes in the kernel. fs/exec.c will be of great help.
The procedure varies among operating systems. Some systems have a background command interpreter that exists through the life of a process and within the process itself. When a program is run, the command interpreter stays in the background (in protected from user mode access). When the program completes, the command interpreter comes to the foreground and can run another program in the same process.
In the Eunuchs-world, the command interpreter is just a user-mode program. Whenever it runs a program it kicks off another process.
Both of these types of systems use a loader to configure the process address space for running a program. The executable file is a set of instructions that define how to lay out the address space,
This is significantly different from a bootloader. A bootloader blindly loads a block of stored data into memory. A program loader contains complex instructions for laying out a process address space that include handling shared libraries and doing address fixups.
I am looking for assistance with the proper GDB / OpenOCD initializion and running commands (external tools) to use within Eclipse for flash and RAM debugging, as well as the proper modifications or additions that need to be incorporated in a makefile for flash vs RAM building for this MCU, if this matters of course.
MCU: STM32F103VET6
I am using Eclipse Helios with Zylin Embedded CDT, Yagarto Tools and Bins, OpenOCD 0.4, and have an Olimex ARM-USB-OCD JTAG adapter.
I have already configured the ARM-USB-OCD and added it as an external tool in Eclipse. For initializing OpenOCD I used the following command in Eclipse. The board config file references the stm32 MCU:
openocd -f interface/olimex-arm-usb-ocd-h.cfg -f board/stm32f10x_128k_eval.cfg
When I run this within Eclipse everything appears to be working (GDB Interface, OpenOCD finds the MCU, etc). I can also telnet into OpenOCD and run commands.
So, I am stuck on the next part; initialization and commands for flash and RAM debugging, as well as erasing flash.
I read through several tutorials, and scoured the net, but have not been able to find anything particular to this processor. I am new to this, so I might not be recognizing an equivalent product for an example.
I'm working with the same tool chain to program and debug a STM32F107 board. Following are my observations to get an STM32Fxxx chip programmed and debugged under this toolchain.
Initial Starting Point
So at this point you've got a working OpenOCD to ARM-USB-OCD connection and so you should be all set on that end. Now the work is on getting Eclipse/Zylin/Yagarto GDB combination to properly talk to the STM32Fxxx through the OpenOCD/Olimex connection. One thing to keep in mind is that all the OpenOCD commands to issue are the run mode commands. The configuration scripts and command-line options to invoke the OpenOCD server are configuration mode commands. Once you issue the init command then the server enters run mode which opens up the set of commands you'll need next. You've probably done it somewhere else but I tack on a '-c "init"' option when I call the OpenOCD server like so:
openocd -f /path to scripts/olimex-arm-usb-ocd-h.cfg -f /path to targets/stm32f107.cfg -c "init"
The following commands I issue next are done by the Eclipse Debug Configurations dialogue. Under the Zylin Embedded debug (Native) section, I create a new configuration, give it a name, Project (optional), and absolute path to the binary that I want to program. Under the Debugger tab I set the debugger to Embedded GDB, point to the Yagarto GDB binary path, don't set a GDB command file, set GDB command set to Standard, and the protocol to mi.
The Commands Tab - Connect GDB to OpenOCD
So the next tab is the Commands tab and that's where the meat of the issue lies. You have two spaces Initialize and Run. Not sure exactly what the difference is except to guess that they occur pre- and post-invocation of GDB. Either way I haven't noticed a difference in how my commands are run.
But anyway, following the examples I found on the net, I filled the Initialize box with the following commands:
set remote hardware-breakpoint limit 6
set remote hardware-watchoint-limit 4
target remote localhost:3333
monitor halt
monitor poll
First two lines tell GDB how many breakpoints and watchpoints you have. Open OCD Manual Section 20.3 says GDB can't query for that information so I tell it myself. Next line commands GDB to connect to the remote target at the localhost over port 3333. The last line is a monitor command which tells GDB to pass the command on to the target without taking any action itself. In this case the target is OpenOCD and I'm giving it the command halt. After that I tell OpenOCD to switch to asynchronous mode of operation. As some of the following operations take a while, it's useful not to have OpenOCD block and wait for every operation.
Sidenote #1: If you're ever in doubt about the state of GDB or OpenOCD then you can use the Eclipse debug console to send commands to GDB or OpenOCD (via GDB monitor commands) after invoking this debug configuration.
The Commands Tab - Setting up the User Flash
Next are commands I give in the Run commands section:
monitor flash probe 0
monitor flash protect 0 0 127 off
monitor reset halt
monitor stm32x mass_erase 0
monitor flash write_image STM3210CTest/test_rom.elf
monitor flash protect 0 0 127 on
disconnect
target remote localhost:3333
monitor soft_reset_halt
to be explained in the following sections...
Setting up Access to User Flash Memory
First I issue an OpenOCD query to see if it can find the flash module and report the proper address. If it responds that it found the flash at address 0x08000000 then we're good. The 0 at the end specifies to get information about flash bank 0.
Sidenote #2: The STM32Fxxx part-specific data sheets have a memory map in section 4. Very useful to keep on hand as you work with the chip. Also as everything is accessed as a memory address, you'll come to know this layout like the back of your hand after a little programming time!
So after confirming that the flash has been properly configured we invoke the command to turn off write protection to the flash bank. PM0075 describes everything you need to know about programming the flash memory. What you need to know for this command is the flash bank, starting sector, ending sector, and whether to enable or disable write protection. The flash bank is defined in the configuration files you passed to OpenOCD and was confirmed by the previous command. Since I want to disable protection for the entire flash space I specify sectors 0 to 127. PM0075 explains how I got that number as it refers to how the flash memory is organized into 2KB pages for my (and your) device. My device has 256KB of flash so that means I have 128 pages. Your device has 512KB of flash so you'll have 256 pages. To confirm that your device's write-protection has been disabled properly, you can check the FLASH_WRPR register at address 0x40022020 using the OpenOCD command:
monitor mdw 0x40022020
The resulting word that it prints will be 0xffffffff which means all pages have their write protection disabled. 0x00000000 means all pages have write protection enabled.
Sidenote #3: On the subject of the memory commands, I bricked my chip twice as I was messing with the option bytes at the block starting at address 0x1ffff800. First time I set the read protection on the flash (kind of hard to figure out what your doing if you do that), second time I set the hardware watchdog which prevented me from doing anything afterwards since the watchdog kept firing off! Fixed it by using the OpenOCD memory access commands. Moral of the story is: With great power comes great responsibility.... Or another take is that if I shoot myself in the foot I can still fix things via JTAG.
Sidenote #4: One thing that'll happen if you try to write to protected flash memory is the FLASH_SR:WRPRTERR bit will be set. OpenOCD will report a more user-friendly error message.
Erasing the Flash
So after disabling the write protection, we need to erase the memory that you want to program. I do a mass erase which erases everything, you also have the option to erase by sector or address (I think). Either way you need to erase first before programming as the hardware checks for erasure first before allowing a write to occur. If the FLASH_SR:PGERR bit (0x4002200c) ever gets set during programming then you know you haven't erased that chunk of memory yet.
Sidenote #5: Erasing a bit in flash memory means setting it to 1.
Programming Your Binary
The next two lines after erasure writes the binary image to the flash and reenables the write protection. There isn't much more to say that isn't covered by PM0075. Basically any error that occurs when you issue flash write_image is probably related to the flash protection not being disabled. It's probably NOT OpenOCD though if you're curious you can take enable the debug output and follow what it does.
GDB Debugging
So finally after programming I disconnect GDB from the remote connection and then reconnect it to the target, do a soft-reset, and my GDB is now ready to debug. This last part I just figured out last night as I was trying to figure out why, after programming, GDB wouldn't properly stop at main() after reset. It kept going off into the weeds and blowing up.
My current thinking and from what I read in the OpenOCD and GDB manuals is that the remote connection is, first and foremost, meant to be used between GDB and a target that has already been configured and running. Well I'm using GDB to configure before I run so I think the symbol table or some other important info gets messed up during the programming. The OpenOCD manual says that the server automatically reports the memory and symbols when GDB connects but all that info probably becomes invalid when the chip gets programmed. Disconnecting and reconnecting I think refreshes the info GDB needs to debug properly. So that has led me to create another Debug Configuration, this one just connects and resets the target since I don't necessarily need to program the chip every time I want to use GDB.
Whew! Done! Kind of long but this took me 3 weekends to figure out so isn't too terribly bad I think...
Final sidenote: During my time debugging I found that OpenOCD debug output to be invaluable to me understanding what OpenOCD was doing under the covers. To program a STM32x chip you need to unlock the flash registers, flip the right bits, and can only write a half-word at a time. For a while I was questioning whether OpenOCD was doing this properly but after looking through the OpenOCD debug output and comparing it against what the PM0075 instructions were, I was able to confirm that it did indeed follow the proper steps to do each operation. I also found I was duplicating steps that OpenOCD was already doing so I was able to cut out instructions that weren't helping! So moral of the story: Debug output is your friend!
I struggled getting JLink to work with a STM3240XX and found a statement in the JLink GDB server documentation saying that after loading flash you must issue a "target reset":
"When debugging in flash the stack pointer and the PC are set automatically when the target is reset after the flash download. Without reset after download, the stack pointer and the PC need to be initialized correctly, typically in the .gdbinit file."
When I added a "target reset" in the Run box of the debugger Setup of Eclipse, suddenly everything worked. I did not have this problem with a Kinetis K60.
The document also explains how to manually set the stack pointer and pc directly if you don't want to issue a reset. It may not be the disconnect/connect that solves the problem but the reset.
What i use after the last sentence in the Comannd Tab - 'Run' Commands, is:
symbol-file STM3210CTest/test_rom.elf
thbreak main
continue
The thbreak main sentence is what makes gdb stop at main.
Are those dump the full dump? What's the best dump which can be analyzed later using Windbg?
Task manager, process explorer and AdPlus are all able to create full user mode dumps, which will contain all the details needed for debugging using WinDbg.
Special attention is required when creating dumps for 32 bit processes running on 64 bit Windows. Please see my answer to this question for additional detail.