Bernhard Kaindl | f212ec4 | 2008-01-30 13:34:11 +0100 | [diff] [blame] | 1 | |
| 2 | Using physical DMA provided by OHCI-1394 FireWire controllers for debugging |
| 3 | --------------------------------------------------------------------------- |
| 4 | |
| 5 | Introduction |
| 6 | ------------ |
| 7 | |
| 8 | Basically all FireWire controllers which are in use today are compliant |
| 9 | to the OHCI-1394 specification which defines the controller to be a PCI |
| 10 | bus master which uses DMA to offload data transfers from the CPU and has |
| 11 | a "Physical Response Unit" which executes specific requests by employing |
| 12 | PCI-Bus master DMA after applying filters defined by the OHCI-1394 driver. |
| 13 | |
| 14 | Once properly configured, remote machines can send these requests to |
| 15 | ask the OHCI-1394 controller to perform read and write requests on |
| 16 | physical system memory and, for read requests, send the result of |
| 17 | the physical memory read back to the requester. |
| 18 | |
| 19 | With that, it is possible to debug issues by reading interesting memory |
| 20 | locations such as buffers like the printk buffer or the process table. |
| 21 | |
| 22 | Retrieving a full system memory dump is also possible over the FireWire, |
| 23 | using data transfer rates in the order of 10MB/s or more. |
| 24 | |
| 25 | Memory access is currently limited to the low 4G of physical address |
| 26 | space which can be a problem on IA64 machines where memory is located |
| 27 | mostly above that limit, but it is rarely a problem on more common |
| 28 | hardware such as hardware based on x86, x86-64 and PowerPC. |
| 29 | |
| 30 | Together with a early initialization of the OHCI-1394 controller for debugging, |
| 31 | this facility proved most useful for examining long debugs logs in the printk |
| 32 | buffer on to debug early boot problems in areas like ACPI where the system |
| 33 | fails to boot and other means for debugging (serial port) are either not |
| 34 | available (notebooks) or too slow for extensive debug information (like ACPI). |
| 35 | |
| 36 | Drivers |
| 37 | ------- |
| 38 | |
Stefan Richter | 09d7328 | 2008-02-18 21:38:35 +0100 | [diff] [blame] | 39 | The ohci1394 driver in drivers/ieee1394 initializes the OHCI-1394 controllers |
| 40 | to a working state and enables physical DMA by default for all remote nodes. |
| 41 | This can be turned off by ohci1394's module parameter phys_dma=0. |
Bernhard Kaindl | f212ec4 | 2008-01-30 13:34:11 +0100 | [diff] [blame] | 42 | |
Stefan Richter | 09d7328 | 2008-02-18 21:38:35 +0100 | [diff] [blame] | 43 | The alternative firewire-ohci driver in drivers/firewire uses filtered physical |
Stefan Richter | 080de8c | 2008-02-28 20:54:43 +0100 | [diff] [blame] | 44 | DMA by default, which is more secure but not suitable for remote debugging. |
| 45 | Compile the driver with CONFIG_FIREWIRE_OHCI_REMOTE_DMA (Kernel hacking menu: |
| 46 | Remote debugging over FireWire with firewire-ohci) to get unfiltered physical |
| 47 | DMA. |
Stefan Richter | 09d7328 | 2008-02-18 21:38:35 +0100 | [diff] [blame] | 48 | |
Stefan Richter | 080de8c | 2008-02-28 20:54:43 +0100 | [diff] [blame] | 49 | Because ohci1394 and firewire-ohci depend on the PCI enumeration to be |
| 50 | completed, an initialization routine which runs pretty early has been |
| 51 | implemented for x86. This routine runs long before console_init() can be |
| 52 | called, i.e. before the printk buffer appears on the console. |
Bernhard Kaindl | f212ec4 | 2008-01-30 13:34:11 +0100 | [diff] [blame] | 53 | |
| 54 | To activate it, enable CONFIG_PROVIDE_OHCI1394_DMA_INIT (Kernel hacking menu: |
Stefan Richter | 080de8c | 2008-02-28 20:54:43 +0100 | [diff] [blame] | 55 | Remote debugging over FireWire early on boot) and pass the parameter |
| 56 | "ohci1394_dma=early" to the recompiled kernel on boot. |
Bernhard Kaindl | f212ec4 | 2008-01-30 13:34:11 +0100 | [diff] [blame] | 57 | |
| 58 | Tools |
| 59 | ----- |
| 60 | |
| 61 | firescope - Originally developed by Benjamin Herrenschmidt, Andi Kleen ported |
| 62 | it from PowerPC to x86 and x86_64 and added functionality, firescope can now |
| 63 | be used to view the printk buffer of a remote machine, even with live update. |
| 64 | |
| 65 | Bernhard Kaindl enhanced firescope to support accessing 64-bit machines |
| 66 | from 32-bit firescope and vice versa: |
| 67 | - ftp://ftp.suse.de/private/bk/firewire/tools/firescope-0.2.2.tar.bz2 |
| 68 | |
| 69 | and he implemented fast system dump (alpha version - read README.txt): |
| 70 | - ftp://ftp.suse.de/private/bk/firewire/tools/firedump-0.1.tar.bz2 |
| 71 | |
| 72 | There is also a gdb proxy for firewire which allows to use gdb to access |
| 73 | data which can be referenced from symbols found by gdb in vmlinux: |
| 74 | - ftp://ftp.suse.de/private/bk/firewire/tools/fireproxy-0.33.tar.bz2 |
| 75 | |
| 76 | The latest version of this gdb proxy (fireproxy-0.34) can communicate (not |
| 77 | yet stable) with kgdb over an memory-based communication module (kgdbom). |
| 78 | |
| 79 | Getting Started |
| 80 | --------------- |
| 81 | |
| 82 | The OHCI-1394 specification regulates that the OHCI-1394 controller must |
| 83 | disable all physical DMA on each bus reset. |
| 84 | |
| 85 | This means that if you want to debug an issue in a system state where |
| 86 | interrupts are disabled and where no polling of the OHCI-1394 controller |
| 87 | for bus resets takes place, you have to establish any FireWire cable |
| 88 | connections and fully initialize all FireWire hardware __before__ the |
| 89 | system enters such state. |
| 90 | |
| 91 | Step-by-step instructions for using firescope with early OHCI initialization: |
| 92 | |
| 93 | 1) Verify that your hardware is supported: |
| 94 | |
| 95 | Load the ohci1394 or the fw-ohci module and check your kernel logs. |
| 96 | You should see a line similar to |
| 97 | |
| 98 | ohci1394: fw-host0: OHCI-1394 1.1 (PCI): IRQ=[18] MMIO=[fe9ff800-fe9fffff] |
| 99 | ... Max Packet=[2048] IR/IT contexts=[4/8] |
| 100 | |
| 101 | when loading the driver. If you have no supported controller, many PCI, |
| 102 | CardBus and even some Express cards which are fully compliant to OHCI-1394 |
| 103 | specification are available. If it requires no driver for Windows operating |
| 104 | systems, it most likely is. Only specialized shops have cards which are not |
| 105 | compliant, they are based on TI PCILynx chips and require drivers for Win- |
| 106 | dows operating systems. |
| 107 | |
| 108 | 2) Establish a working FireWire cable connection: |
| 109 | |
| 110 | Any FireWire cable, as long at it provides electrically and mechanically |
| 111 | stable connection and has matching connectors (there are small 4-pin and |
| 112 | large 6-pin FireWire ports) will do. |
| 113 | |
| 114 | If an driver is running on both machines you should see a line like |
| 115 | |
| 116 | ieee1394: Node added: ID:BUS[0-01:1023] GUID[0090270001b84bba] |
| 117 | |
| 118 | on both machines in the kernel log when the cable is plugged in |
| 119 | and connects the two machines. |
| 120 | |
| 121 | 3) Test physical DMA using firescope: |
| 122 | |
| 123 | On the debug host, |
| 124 | - load the raw1394 module, |
| 125 | - make sure that /dev/raw1394 is accessible, |
| 126 | then start firescope: |
| 127 | |
| 128 | $ firescope |
| 129 | Port 0 (ohci1394) opened, 2 nodes detected |
| 130 | |
| 131 | FireScope |
| 132 | --------- |
| 133 | Target : <unspecified> |
| 134 | Gen : 1 |
| 135 | [Ctrl-T] choose target |
| 136 | [Ctrl-H] this menu |
| 137 | [Ctrl-Q] quit |
| 138 | |
| 139 | ------> Press Ctrl-T now, the output should be similar to: |
| 140 | |
| 141 | 2 nodes available, local node is: 0 |
| 142 | 0: ffc0, uuid: 00000000 00000000 [LOCAL] |
| 143 | 1: ffc1, uuid: 00279000 ba4bb801 |
| 144 | |
| 145 | Besides the [LOCAL] node, it must show another node without error message. |
| 146 | |
| 147 | 4) Prepare for debugging with early OHCI-1394 initialization: |
| 148 | |
| 149 | 4.1) Kernel compilation and installation on debug target |
| 150 | |
| 151 | Compile the kernel to be debugged with CONFIG_PROVIDE_OHCI1394_DMA_INIT |
| 152 | (Kernel hacking: Provide code for enabling DMA over FireWire early on boot) |
| 153 | enabled and install it on the machine to be debugged (debug target). |
| 154 | |
| 155 | 4.2) Transfer the System.map of the debugged kernel to the debug host |
| 156 | |
| 157 | Copy the System.map of the kernel be debugged to the debug host (the host |
| 158 | which is connected to the debugged machine over the FireWire cable). |
| 159 | |
| 160 | 5) Retrieving the printk buffer contents: |
| 161 | |
| 162 | With the FireWire cable connected, the OHCI-1394 driver on the debugging |
| 163 | host loaded, reboot the debugged machine, booting the kernel which has |
| 164 | CONFIG_PROVIDE_OHCI1394_DMA_INIT enabled, with the option ohci1394_dma=early. |
| 165 | |
| 166 | Then, on the debugging host, run firescope, for example by using -A: |
| 167 | |
| 168 | firescope -A System.map-of-debug-target-kernel |
| 169 | |
| 170 | Note: -A automatically attaches to the first non-local node. It only works |
| 171 | reliably if only connected two machines are connected using FireWire. |
| 172 | |
| 173 | After having attached to the debug target, press Ctrl-D to view the |
| 174 | complete printk buffer or Ctrl-U to enter auto update mode and get an |
| 175 | updated live view of recent kernel messages logged on the debug target. |
| 176 | |
| 177 | Call "firescope -h" to get more information on firescope's options. |
| 178 | |
| 179 | Notes |
| 180 | ----- |
| 181 | Documentation and specifications: ftp://ftp.suse.de/private/bk/firewire/docs |
| 182 | |
| 183 | FireWire is a trademark of Apple Inc. - for more information please refer to: |
| 184 | http://en.wikipedia.org/wiki/FireWire |