Embedded systems and system-level programmable data sheet

7 embedded system design issues 7.1 system although embedded system applications face a very broad, the complexity of application vary, personalized requirements.

However, the overall design of embedded system requires consideration of key issues still can be summarized into the following areas: · in-depth analysis of the system, have a thorough understanding of the application depending on the application needs. ·, forming the overall scheme, select or to build a development platform, complete algorithms and architectures for mapping; formation of hardware and software co-design programme; · embedded processor selection; · real-time operating system (RTOS) selection; · development tools (such as compilers, emulators) selection or development; design · data paths; · Testability design; design · reusability; · to imitated card, verify that the establishment of the environment; · the target node of planning; 7.2 embedded processor embedded processor is based on application and user-oriented core embedded system, because the application-oriented, object-oriented features of the industry, for different applications and different semiconductor process, various kinds of embedded processors variants include more than 1000 can be divided into: · embedded microprocessor: arm, MIPS, PowerPC, 68000, 386EX; · embedded microcontroller: 8051, MC68HC05, MC68HC11, 68300, Am186, TMS320C2X; · embedded DSP processors: TMS320 series, DSP56000 series; · embedded SOC (System on chip): OMAP, M-Core; 7.3 embedded processor development with embedded applications of expanding, such as disk drives, mobile phone, PDA, network switches, printers, and so on, to define the concept of the "embedded" in itself is becoming increasingly difficult. But what is certain is that many embedded applications in areas such as performance, cost, power consumption, and other indicators are ever-changing requirements. In order to adapt to the different requirements of the application for a variety of processors are also rapidly. The following to apply a broad United Kingdom ARM company's embedded ARM processor, for example, for analysis. ARM processor microarchitecture of evolution--for the past 10 years, ARM processor developed from simple ARM7TDMI core development to the latest arm11EJ-S kernel, its major progress: · from simple three-line to 8-level advanced pipeline architecture. Single clock for comparison, the performance difference was not significant. · ARM11 is the ARM family of performance of a series. ARM7TM usually only in the following frequencies of 200MHz; in 0.13 Micron process, arm11TM run frequency up to 500-700MHz, brought more powerful ARM processors · performance. increasing line series, structure and design improvements. In arm1026EJ-STM kernel, with a return stack, reduce the subroutine returns the line refresh time, improves system performance arm11 increased ·. static and dynamic combinations of jump prejudging. · arm11 support single instruction multiple data (SIMD) instructions. 7.4 real-time operating system (RTOS) software is one of the core for embedded systems, embedded systems often have real-time requirements, and real-time operating systems with simple, compact, and other characteristics, as a result, embedded systems often use embedded real-time operating system. For embedded systems and real-time operating systems, however, not every embedded real-time operating system can support the embedded processor. Therefore, the embedded processor with embedded real-time operating system selection must also consider. Common embedded real-time operating system: VxWorks; · OS-9 ·; · PSOS; · QNX · VRTX;; · real time operating system based on Linux; · real time operating system based on Windows; · domestic independent development of real-time operating systems;

Linux operations command log

In a Linux Terminal, for the convenience of checking operations possible errors, as well as avoid screen scroll limit, we may record the operations log.

Common tool screen, script, and the tee. 1.screen — screenmanagerwithVT100/ANSIterminalemulatio > screen-L > here is what we do > exit after a similar file name: screenlog.0 > morescreenlog.0 so you can see what you have just done, find possible problems 2.script — maketypescriptofterminalsession > script > our operations > exit generates one such file: typescript > moretypescript here we do 3.tee – readfromstandardinputandwritetostandardoutputandfiles this command can read Terminal input and output to a terminal or file, sometimes it can be used to record make commands can produce output to the terminal by a large amount of content is output to the file. This way you can easily record a log of these commands. > Make2 > & 1 | teemake.log, of course, we can also redirect to a file > make > make.logPS: 2 > & 1 to log the error log ifyouwanttofilterthecontrolsymbols, trytousethe "col" commandliketis: $ catscreenlog.0 | col-b > screenlogor $ cattypescript | col-b > scriptlog

CreateProcessW implementation on full control of process creation

【 Preface 】 write this document when I met enough food, many problems, I would like to thank bkbll, SobeIt a1rsupply and pointing, TCH of hard work, only the birth of this document, this article there may be some errors, the error is due to my mistake, if you have any comments, welcome to http://www.itaq.org noted or E-mail: zf35@citiz.net 【 overview 】 implementation on the server to process created by the control has a great significance, by monitoring the process of creating, we can make the process of being allowed to run correctly, and it is not permitted by the program will create fails, this can prevent unknown Trojans, virus and worm threats on the server.

To achieve these purposes, you must create a process-related hookwindows API, under the insidethewindowsNT "and" NativeAPIReference records, plus the actual track, softIce windows create process API call procedures are as follows: 【 code 】 CreateProcessA-> CreateProcessW-> CreateProcessInternalW-> ... > end call-ZwCreateProcess this document we use CreateProcessW to achieve our goal, of course, you can also use other several API. This document demonstrates code is a little changes can be applied to arbitrary Ring3 function. For hook an API, you can use a lot of ways, this article use override function entry point approach to achieve hook CreateProcessW, more detailed information please refer to the windows SobeIt write down some way hookAPI. 【 Copy-on-write 】 initial test, I use the rewrite function aCreateProcessW softice is the entry point code, F5 to switch back to windows after the discovery of all things, but when I write a program to modify the entry point code CreateProcessW, your changes are only valid for this process, other processes for system did not produce any effect. Use softice track and found this process CreateProcessW virtual address is mapped to a new, and other process on a different physical address, if you read the Webcrazy copy-on-write mechanism ", it is not hard to see that this is the impact of copy-on-write mechanism. For the system DLLs, each DLL that is mapped to a different process on the same virtual address, these virtual address and point to the same physical address by which system the lowest resource consumption when a process tries to overwrite data in physical memory, in order not to affect other processes, the system automatically assigns a new physical memory, the original data in physical memory copy, overwrite, and then to overwrite memory of the process's virtual address remap it to a new physical memory, and other process or map in the original physical memory, this is the "copy-on-write technique" (copy-on-write), then the system is how to determine when you should use copy-on-write? this is a virtual address to determine, PTE when copy-on-write flag in the PTE is set, any of the virtual address of the write operation will cause a copy-on-writ 【 three viable 】 in order to achieve a global hook, we cannot be restricted by copy-on-write mechanism, now I thought of three approaches to achieving our aim. 1. through the driver to modify the page table entry (PTE) properties so that the corresponding virtual address CreateProcessW lost copy-on-write property, which in itself processes on CreateProcessW entry point code modifications will take effect for all processes in the system, enabling global hook. 2. through the windows themselves an object to the physical memory \\phymem for direct reading and writing, first positioning itself processes Eprocess (KTEB) (PS: how to navigate in Ring3 arbitrary process Eprocess please refer to the written before I get process Eprocess "), you can get after Eprocess process page catalog, and then use \\phymem read stored page catalog contents of physical memory, and then simulate operating system virtual address-> physical memory address conversion, the resulting CreateProcessW corresponding physical address, use \\phymem we avoid copy-on-write mechanism directly overwrite CreateProcessW. 3. through the most conventional means to achieve the objective, the first enumeration system all processes and then by VirtualAllocEx, VirtualProtectEx VirtualQueryEx, etc functions modify each process page properties, allocate new space. Finally our code using WriteProcessMemory wrote all process space, the use of override CreateProcessW entrance to Jmp * ** * ** * to go to our code, change the function of the flow of execution. The above three methods, method 1 is only a concept, not a reality, drop, I go back and try the page Welcome to master to achieve, and then mail a copy of the code give me: P method 2 I wrote a complete code to implement it, but in this documentNot for discussion, the document will become very long, I will be in a separate document specifically describes the approach of specific implementation. Method 3 so that the focus of this article, the following is a detailed description of the method 3. Query the base address and CreateProcessW properties here we use the function, its VirtualQueryEx prototype is as follows: SIZE_T VirtualQueryEx (HANDLEhProcess, LPCVOIDlpAddress, PMEMORY_BASIC_INFORMATIONlpBufer, SIZE_TdwLengt); Parameter description: HANDLEhProcess want to query the memory information on the process handle LPCVOIDlpAddress point to a memory area you want to query the pointer PMEMORY_BASIC_INFORMATIONlpBuffer MEMORY_BASIC_INFORMATION structure pointer points to SIZE_TdwLengthlpBuffer size after you call this function, related information is stored in the structure pointed to by lpBuffer is modified in the page properties for CreateProcessW a page that has the following properties: PAGE_EXECUTE PAGE_EXECUTE_READ PAGE_EXECUTE_READWRITE PAGE_EXECUTE_WRITECOPY PAGE_NOACCESS PAGE_READONLY PAGE_READWRITE PAGE_WRITECOPY we modify by VirtualProtectEx page properties: BOOL VirtualProtectEx (HANDLEhProcess, LPVOIDlpAddress, SIZE_TdwSize, DWORDflNewProtect, PDWORDlpflOldProtect); Parameter description: HANDLEhProce ss process handle LPVOIDlpAddress point to want to modify a pointer to the memory region SIZE_TdwSize modify the size of the memory area DWORDflNewProtect new page attributes PDWORDlpflOldProtect links to save the old page property of memory from the back of the code we can see that in order to overwrite the function entry point code, we must give it PAGE_EXECUTE_READWRITE property. In the process of allocating free space optical modify function entry point code is not enough. We must have written some code to take over the work, as CreateProcessW process space is isolated, in order to achieve the goal of a global hook, we must provide each process for a piece of space to store our code, this is going to use this function, VirtualAllocEx VirtualAllocEx prototype is as follows: LPVOID VirtualAllocEx (HANDLEhProcess, LPVOIDlpAddress, SIZE_TdwSize, DWORDflAllocationType, DWORDflProtect); Parameter description: HANDLEhProcess process handle LPVOIDlpAddress points to a pointer to the allocated memory area SIZE_TdwSize assigned area DWORDflAllocationType memory type DWORDflProtect new memory properties we use WriteProcessMemory function to write to the remote process's code and data, the prototype is as follows: BOOL WriteProcessMemory (HANDLEhProcess, LPVOIDlpBaseAddress, LPCVOIDlpBufer, SIZE_TnSize, SIZE_T * lpNumberOfBytesWritten); Parameter description: HANDLEhProcess process handle LPVOIDlpBaseAddress point to write address pointer LPCVOIDlpBuffer point to write the data pointer size of SIZE_T * SIZE_TnSizelpBuffer lpNumberOfBytesWritten actual number of bytes written for the magic in my compiler using WriteProcessMemory put myself to write a function JmpToAddress writes the contents of the remote process space, regardless of my JmpToAddress content is written to the space are E9 ** ** the few bytes, which made me very confused, from machine code, this is a relative jump instruction. It is come from it, in order to understand this problem, I used the VC Debug it, enter in the watch window, showing JmpToAddress JmpToAddress virtual address 0x00410XXX, and then open the memory window to view the memory storage of content, discovery is JmpToAddress code, this is strange, mysterious E9 * *** fromWhat to do, so I ask the a1rsupply, he told me that the debug version of the VC will generate a jump table, light, original play of magic is a compiler. In order to remote processes properly written code, we must calculate the real function address, the following I wrote a piece of code to calculate the real function address: __asm {pusad leaeax, JmpToAddress movecx, JmpToAddress shrecx, 8 addeax addeax, ecx, eax, 5 movJmpAfterCalc popad} solve trouble-positioning problem in writing the code, I encountered another big question is how to locate the address. I write JmpToAddress () function as follows: JmpToAddress void__declspec (naked) (void) {__a smjmp [HookedAddr]} in the local process this code no problem, but when it is written to the remote process problems will arise, we take a look at its assembly code, following jmp [00401Cxxx] we note this process in the virtual address is deposited in the HookedAddr address, but the remote process, this address points to is something else, jmp past will produce unexpected results, in order to achieve the correct behavior, we'll write to the remote process WriteProcessMemory HookedAddr content, and then use a relative address references it JmpAddress void__declspec (naked) (void) {__asmcalllag flag: __asmmovebx __asmpopeax __asmaddeax, 0x0e, [eax] __asmjmpebx} popeax, eax is deposited inside of this directive, the address of the virtual, plus a fixed value, [eax] is through WritePr

Firefox upgrade: System Administrator security considerations

Mozilla's Firefox is recognized as not only is an excellent browser, but its safety factor is also very high.

Many have used InternetExplorer or ActiveX technology in institutions because of their security and ease of operation and prepare for change to Firefox. High safety factor does not mean that Firefox can shield all viruses or no security vulnerabilities. However, in the most recent version, Mozilla 1.0 issued a series of upgrade package to make up for the Firefox security vulnerability in the latest version 1.0.6, the system has been up for a series of code vulnerabilities. However, the problem is that an organization's network administrator whether it is necessary to constantly track Firefox version updates and keep pace with it. In General, you can have two situations, to distinguish between the approach: one is the personal that is the system user end, desktop users, to be responsible for your own system's update, we called a poweruser. Another case is the software installed by system administrators control the end user only permission to run the program, you cannot remove the installation, we called JoeUser. 1.Powerusers Firefox's default settings, once a day to prompt for a new core program or plug-in, when Firefox found can be used to do the upgrade plugin, a green arrow will appear next to the throbber, this in the browser icon in the upper right corner of the download process will blink. If the user clicks on the green arrow, Firefox will start to download the latest version, before beginning the installation, Firefox will prompt the user to close the browser. In short, powerusers if necessary their own program of upgrading and installation. For an experienced user, these operations do not bother. But on the other hand, if your environment requires that the user won't have to install a program on the system, then this convenient and free will. 2. "JoeUsers" Firefox is as a cross-platform applications to write, so it on Windows, does not automatically through the central authentication system upgrade, or even automatically in the enterprise for new product configurations. This has always been considered to be the Firefox of the biggest flaws. At the same time, it is also the problem of Firefox users cannot change system settings, because many Firefox settings are not retained in the registry, but only as text exists. MozillaFirefox near new FrontMotion, is considered the most comprehensive solution to the problem of software, and therefore also has been a large number of users install MozillaFirefoxMSI, MSIs (MicrosoftInstallerpackages) was installed at the same time, it will issue a series of directives, so that the MSIs to each whole unit installed in the computer. MSIs has many advantages, such as its send mode for many administrators with a convenient, at the same time, it can first reconnaissance out during the upgrade process if there is a running Firefox process and prompts the user to close. More convenient is that FrontMotion also created a Firefox community to help administrators through Active Directory settings for Firefox to download, if you are just starting to set in the Organization and use Firefox, and you want to rely on existing systems rather than a specific solution, you can get help in the Firefox community. In addition, the FFDeploy script package BobTempleton is also a good choice, it is easier to configure, because use is the VBS script, programmers can easily be modified to configure a Firefox image. This is a relatively simple way, since it avoids a manageable environment deployment complexity of Firefox.

Linux system some SU command security recommendations

First of all we see su in man help page for an explanation, that is to say by su-runashellwithsubstituteuserandgroupIDs su we do not provide user and group name you can start a shell program.

Su is a binary executable file, command file path which is/bin/su, the following is from the command line query su file type and the path: case study 1: islab $ whic su/bin/su case study two: islab $ file/bin/su/bin/su: setuidELF32-bitLSBsharedobject, Intel80386, version1 (SYSV), forGNU/Linux2.6.9, stripped cases III: islab $ ll/bin/su-rwsr-xr-x1rootroot24060Jan102007/bin/su cases and cases II we can see the su is a setuid program (the setuid bit chmodu + s can be used, such as ls displays su file owner property hired setuid bit), in this case, you can get su than its owner higher privileges, i.e. run su, your permission will be promoted, with root permissions equivalent. Case study three we can see the file type is ELF32-bitLSBsharedobject (set the setuid bit), which means that the program requires libc functions library, also requires the use of the ELF interpreter, and comply with the LSB specification. Question: normal user can copy from other machines in the su command. A: they can copy from the other machine su command, but they will not be able to su to correct permissions settings such as chownroot and chmodu + s, and so on. So the copy over the su is not working correctly. Question 2: how to prevent normal users from executing the su command. A: 1.) You can set up a special group, only members of the group to perform the su command islab # groupaddweelislab # useraddweelislab # chownroot: mysql/bin/bas islab # chmod4750/bin/su2). Only the root user can perform the su command. islab#chmod4700/bin/su3）。 By pam library implementation only wheel group membership required to perform the su command, the following example adds the account to the wheel group zhaoke. Islab # groupaddweelislab # useraddweelislab # usermod-Gwheelzhaokeislab # ll/lib/security/pam_wheel.so-rwxr-xr-x1rootroot5692Feb222007/lib/security/pam_wheel.soislab # vi/etc/pam.d/su increase following line authrequired/lib/security/pam_wheel.souse_uid then save exit su profile. Question three: the average user though unable to perform the su command, but it is also possible through brute force attacks to obtain root password answer: normal user can shell or ssh on the root account for brute force attacks. We can consider a number of security tools such as pam_abl for ssh for protection. Pam_abl will be able to set the time on error log-in account to a temporary block. Of course, normal users can also elevate privileges through program vulnerabilities such as buffer overflows.

Embedded Linux in industrial control applications

4.2 software design for embedded operating system is the core of the entire embedded system.

As mentioned earlier, the embedded system memory and storage capacity is insufficient, you must design for reduction of linux. In the cutting process, the main technologies involved are the following. (1) the streamlining of the standard Linux kernel is oriented to the PC, it integrates many of the needs of the embedded PC system and unwanted features. Therefore, you can add or remove the independence of function block, you can compile the kernel, the only embedded system function modules and remove unnecessary function block. Then, recompile the kernel will be significantly reduced. (2) of the virtual memory system shield after analysis found that virtual memory is the leading Linux real-time is not strong. In the industrial control, certain tasks must meet certain requirements of real-time, shielding the kernel virtual memory management mechanism to enhance real-time Linux. When you want to change the kernel of a mechanism, generally do not have to write code on a large scale, you can use conditional compilation method. At the same time as the linux system on the application process is fair time allocation scheduling algorithm, this algorithm cannot guarantee that the system of real-time requirements, and requirements change. Change the way there are two: one by POSIX, the second is through the underlying programming. The author is linux real-time named pipe (FIFO) queue of special handling real-time task sequencing. In fact, real time named pipe like as never change real-time tasks, so the page can dramatically reduce the memory page without delay caused by uncertainty. (3) written device driver determines the basic functionality of the kernel, is written for a specific device driver, you can follow to write a Linux driver under the rules for writing. Writing a device driver should have the following features: • the initialize and release; • complete data from the kernel to the hardware device delivery and read data from hardware two functions; • read application passed to the device file data as well as the echo application requests for data; • the detection and treatment equipment error. (4) development of Flash-based JFS file system applications and critical data is typically a document to be stored in the Flash file system. JFFS2 file system is a log-structured, which means that it is essentially a long column node. Each node contains information about the parts of the file. JFFS2 is specifically designed for that kind of like flash chip embedded device creation, so it's entire design provides better Flash management with other file systems, incomparable advantages. Follows: ● JFFS2 at sector level Flash erase/write/read operations than Ext2 file system. ● JFFS2 provides better than Ext2fs crash/power-down protection. When you want to change small amounts of data, the Ext2 file system will copy the entire sectors into memory (DRAM), and in memory to be merged into a new data and then written back to the entire sector. But you can always change the JFFS2 need (not override) the entire sector, and also has a crash/power-down function security. Implementation of the above steps, a mini Linux operating system is constructed. Structure of Linux including process management, memory management and document management, etc. It supports multiple tasks in parallel, there is a complete TCP/IP Protocol, while Linux built-in on Ethernet controllers support, you can connect through Ethernet to Ethernet for remote configuration and monitoring. Will crop the good kernel porting to the target Board, first of all should be compiled into the kernel for the processor of the target code. Because of different hardware system of the transplantations boot code will differ, therefore, some kernel program may want to overwrite. Involves writing a Linux boot code and modify the relevant part of the architecture code mainly boot, memory management and interrupt handling. The company will be M-System DOC2000 as system boot device, the boot code can be placed in DOC. This system is powered on, boot code to perform basic hardware initialization, then the kernel image is loaded into memory and runs, in the end, then debug the good kernel and applications programming in Flash memory. Because of this crop of Linux was successfully ported to target platforms, so that you can run the start of the development system, you can apply under specific to developing applications. If the data acquisition module, data processing module, communications and data dissemination module and so on. 5 conclusion today, Internet applications are go to embedded device as the Center, therefore, labor control system combined with the Internet to achieve Internet is an inevitable trend. The embedded Linux kernel embedded microprocessor-based StrongarmSA1110 32-bit MCU system, and then by constructing a TCP/IP multiple network protocols and basic network communication protocols, and then use the embedded operating system on the underlying hardware and network protocols support, as well as on requirements of industrial control system real-time lin-ux kernel and virtual memory mechanisms to ensure the transformation, and control tasks to complete in real-time and reliability. It is foreseeable that this programme in the field of industrial control with good prospects, and a short development cycle, performance stable reliable, adaptable, and so on.

Linux + NT OS Loader mini-HOWTO translation Edition

1. the summary 2.NTOSLoader work?

3. how to install a WindowsNT and Linux on the same system on 3.1. install WindowsNT 3.2. partitioning 3.3. install Linux 3.4.Lilo 4.Bootpart 5.Linux part of some work 7 6.WindowsNT-do it again 8. exclude issues 9. other 9.1. reference 9.2. thanks 9.3. feedback 9.4. tainted NTLoader starts: FreeBSD 1. abstract this document describes using WindowsNTbootloader to start Linux method. this procedure has been in WindowsNT4.0Workstation Edition and tested Linux2.0.2.NTOSLoader work? NTOSloader hope from other operating system boot sectors into a file that Access reads the records. it and start the selected operating system, that is, or the operation of the different models of WindowsNT or any other operating system. (3) how to install a WindowsNT and Linux on the same system on 3.1. install WindowsNT first try installing WindowsNT. If you want to use NTFS as your file WindowsNT partition system, remember, the current version of Linux is unable to access NTFS partitions, one can read NTFS partitions of the Alpha version of the drivers can be found at http://www.informatik.hu-berlin.de/~loewis/ntfs. you can create another FAT partition as a data interchange, or you can use the DOS format of soft disc. 3.2. partitioning another mini-howto recommended to not use NT's "Disk Administrator" to build a Linux replacement as well as the root partition, it will leave some space, I also recommend using Linux fdisk. 3.3. install Linux boot from floppy disks, replacement and Linux root partition .Fdisk NTFS partitions will be considered as HPFS partitions, this is normal. start again from the magnet Linux and press your preferences Setup Linux. during the installation it is recommended you hang up the HPFS partition found: ignore when it vs. 3.4.Lilo when you do install Lilo's part, specify your Linux root partition as your boot device for your floppy disk's master boot record (MBR) is now a WindowsNT all. This means that your/etc/lilo.conf for root project and boot project will have the same value if you have an IDE hard disk and your Linux partition is in the second partition on your/etc/lilo.conf for boot project will look like: boot =/dev/da2 if you have two drives and your Linux lies on the second disc of the first partition on your/etc/lilo.conf for boot project will look like: boot =/dev/db1 to match your system's core implementation lilo. If you are not sure, then start from the magnet to examine core. If you are unable to start the WindowsNT, you are in trouble, I hope that you have established the most recent fix magnet. 4.Bootpart has a G.Vollant written by NT program called bootpasr can do for you the following two points made in this .Bootpart: http://ourworld.compuserve.com/homepages/gvollant/bootpart.htm. If you want to know how it operates, with the fifth and sixth section describes some of the steps. work your 5.Linux must start from the magnet Linux until some changing for the better until NT. now you need from your Linux root partition to start out of the magnetic zone if/dev/hda2 is your Linux partition, the dd command is: # ddif =/dev/hda2of =/bootsect.lnxbs = 512count = 1 if your bootsect.lnx exceeds 512 bit tuples so that something was amiss. now copy the files to a bootsect.lnx DOS formatted the disk, if this is your file transfer to the NTFS partition area method so you can do to copy # mcopy/bootsect.lnxa: or # mount-tmsdos/dev/fd0/mnt # copy/bootsect.lnx/mnt # umount/mnt 6.WindowsNT portion of work from copies of the files on disk to the c: ootsect.lnx. I will not tell you what to do. (translation: copya: bootsect.lnxc:, should be.) just like Linux lilo.conf, under WindowsNT is c: oot.ini. in compiling it before you use this command to remove its system and read-only properties: C:attrib-s-rc: oot.ini now use an editor to change the boot.ini file, for example notepad, like this:[bootloader]　　timeout=30　　default=multi(0)disk(0)rdisk(0)partition(1)WINNT　　[operatingsystems]　　multi(0)disk(0)rdisk(0)partition(1)WINNT="WindowsNTWorkstation...　　multi(0)disk(0)rdisk(0)partition(1)WINNT="WindowsNTWorkstation... C:BOOTSECT.LNX = "Linux" in this case only joined the last line in your store with this directive after the boot.ini back its properties: C:attrib + s + rc: oot.ini in turn off your WindowsNT and restarted after you should see something like this: OSLoaderV4.00 please select what you want to start the operating system: WindowsNTWorkstationVersion4.0 WindowsNTWorkstationVersion4.0 [VGAmode] Linux using the ↑ and ↓ keys to the options you want, select highlight theearth please press ENTER to confirm the selection of Linux and see LILOloadingzImage .... 7. do it again each time your Linux partition boot sectors are modified after the publication of a new copy of bootsect.lnx must transfer to C:BOOTSECT.LNX. one example is when you use lilo installation a new core when so you can see that such a system is not designed to test the core. 8. exclude problem if things are not working as expected, with a magnet to check if you can start a sexual/dev/hdb1 is your Linux partition, your/etc/lilo.conf has the following items: root =/dev/db1 boot =/dev/d0 insert a magnet and execute lilo. now try to start from magnet. If you're on Linux cannot/dev/hdb1 was activated, then NTOSloader also will not start it if you see a pile of 01010101, that your root disks do not have access to. check all your disks are detected by the BIOS. If the magnet you can boot your Linux partition, you can use the following instructions to check out your boot sectors to NTOSloader used: # ddif =/dev/fd0of =/bootsect.lnxbs = 512count = 1 you can retain this magnet in your WindowsNT Setup problems when you first aid magnet. 9. other 9.1. reference TheLinux + WindowsNTmini-OWTO TheFAQforFreeBSD2.X 9.2. thank thank XiaomingYi prompts these procedures in addition to the first floppy disk on a different disk for the operation. provide disk skills FrankDennler. 9.3. feedback comments are very welcome. 9.4. translation: to start FreeBSD because the translator NTLoader PC and Linux in addition to WindowsNT4.0, installed, then just follow the FreeBSD2.2 above to try NTLoader is there a way to start FreeBSD. my FreeBSD installed in/dev/sda2, FreeBSD 's/dev/sd0s2. follow the above practices I first put the bootsector stored/dev/sda2 into one archive: # ddif =/dev/sda2of =/bootsect.bsdbs = 512count = 1 and copy it to c: boot.ini again to add C:BOOTSECT.BSD = "FreeBSD", but this method could fail to boot FreeBSD! On the one hand, think that down, on the one hand, and did not give up suddenly thought, lilo, or you can take to start FreeBSD. Hence, do the following experiment, after entering the Linux will change to:/etc/lilo.conf boot =/dev/d0 install =/boot/boot.b vga = normal other =/dev/sda2 label = FreeBSD table =/dev/sda loader =/dev/os2_d.b into a format the disk, and then run lilo-v. then will the boot sectors disk into a file: # ddif =/dev/fd0of =/bootsect.bsdbs = 512count = 1 and into c:. Remove magnet after reboot, NTLoader menu under Select FreeBSD, then you see: LILOloadingFreeBSD > > FreeBSDBOOT  @ 0x10000: 639/65536kofmemory Usage: [[0:] [fd] (0, a)]/kernel] [-abcCdrsv] Use1: sd (0, a)/kerneltobootsd0ifitisBIOSdrive1 forfilelistorpressEnterforde Use?faults　　Boot: