Method and system for failsafe recovery and upgrade of an embedded operating system

[0002] The present invention relates to embedded operating systems and, in particular, to a method and system, for upgrade and recovery of an embedded operating system that do not rely on read-only memory.

[0003] Computer operating systems represent a well-known and mature technology of fundamental importance to the computer hardware and software industries. Computer operating systems have evolved into relatively large collections of software programs and routines that provide an interface between the processing hardware of a computer system and the users of that processing capability, including human users, application programs, peripheral devices, appliances, and remote devices and computers interconnected with the computer system via one or more communications networks. For example, a computer operating system is responsible for managing the internal memory resources of a computer system and allocation of those resources to active processes, as well as for managing loading and execution of programs associated with active processes and multiplexing of concurrently active processes. Computer operating systems include various device drivers that control peripheral devices interconnected with the computer system and that manage data transfer between the computer system and the peripheral devices, memory management facilities, command interpreters, file system facilities, and many additional functions. Modern computer operating systems are also responsible for providing an interactive environment with which human users interact with the computer system.

[0004] Initially, computer systems did not include operating systems. Operating systems evolved to provide efficient and robust servicing of concurrent human users of computer systems. The evolution of operating systems has been tightly coupled to the evolution of computer system hardware capabilities. As the hardware capabilities of peripheral devices and the functions and requirements for peripheral devices have increased, the concept of embedded operating systems that run within peripheral devices, such as printers, has emerged. Embedded operating systems internally manage peripheral and consumer electronic devices just as computer operating systems manage computer systems. In fact, certain operating systems initially developed as computer operating systems are currently being transformed into embedded operating systems for managing peripheral and consumer electronic devices, such as printers, home entertainment devices, and other such electronic devices.

[0005] FIG. 1 is a block diagram of a typical computing environment managed by an embedded operating system within a peripheral or consumer electronic device. A typical, high-end device includes a processor 102 that executes an embedded operating system and various software routines that control operation of the device, a random access memory (“RAM”) 104, a read-only memory (“ROM”) 106, a hard disk or other mass storage device 108, a communications controller, such as an Ethernet controller, fibre channel controller, or modem 110, and various controllers 112 and 114 that control electrical and electromechanical components of the device. In a printer, for example, various controllers, such as controllers 112 and 114, may control electrical motors that feed sheets of paper and that control mechanical and electromechanical components such as print heads, collation mechanisms, and LCD front panel displays and pushbuttons that display information to a user and through which a user may input data to the routines that control operation of the printer. Typically, the processor 102, RAM 104, and ROM 106 may be interconnected via a high speed bus 116, that is, in turn, interconnected with a lower-speed bus 118 via a bus bridge 120. The lower-speed bus 118 typically interconnects the mass storage device 108 and controllers 110, 112, and 114 with the processor 102 and the RAM 104.

[0006] In current devices that employ embedded operating systems, instructions that represent the embedded operating system or critical components of the embedded operating system are stored in ROM 106 so that, when the device is powered on, the embedded operating system or critical components of the embedded operating system can be reliably retrieved from ROM for execution by the processor 102. The failure rate of ROM is much lower than the failure rate of mass storage devices so that, by using ROM to store the embedded operating system or critical components of the embedded operating system, the device can be powered up to a functional state even when the mass storage device becomes unreliable or fails completely. ROM data is maintained through power-on and power-off cycles, as is data stored on disk drives, but RAM data is not, and therefore RAM cannot be used for storing non-volatile copies of the embedded operating system. In addition to the critical components of the operating system, called the kernel, an operating system typically requires a data environment, called a file system, in which less frequently invoked routines and data files are stored. This data environment is typically stored in current peripheral and consumer electronic devices in file systems on the mass storage device 108 that allow data to be both read from and written to the file system. However, file systems that allow both reads and writes are potentially error prone, and sufficient robustness of the data can be obtained only by complex and expensive software technologies or combinations of specialized hardware and software technologies.

[0007] The kernel and associated data environment that together compose an embedded operating system are referred to as the “primary image.” Normally, when errors occur that corrupt the embedded operating system, user intervention is required to rebuild the primary image from stored backup copies of the primary image, routines stored in ROM, or a combination of backup copies and ROM-based information. ROM memory is an expensive component, small in capacity, and is relatively expensive and time-consuming to update. Designers, users, and manufacturers of peripheral and consumer electronic devices have thus recognized the need for robustly storing an embedded operating system within an electronic device without using ROM memory and without using complex software or software and hardware solutions for robustly storing archival copies of primary images. In addition, designers, manufacturers, and users of peripheral and consumer electronic devices have recognized the need for embedded operating systems that can automatically detect and correct file system corruption, including corruption of the primary image, without user intervention, and that can be easily updated to newer or alternate versions from remote locations without extensive user intervention.

[0008] In one embodiment of the present invention within a peripheral or consumer electronic device, a read-only disk partition /boot is employed as a disk-based ROM that stores a failsafe copy of an initial operating system (“OS”) kernel as well as a backup copy of a primary image comprising a compressed archive file containing the data environment and operating system routines that are sufficient to boot up and install an embedded operating system. A second generally read-only disk partition /dist stores a backup copy of the primary image as well. The initial OS kernel, upon power-on or other boot event, is instantiated in a RAM-disk partition and runs several script routines that verify the integrity of the primary image and automatically rebuild the primary image in case of corruption. These script routines also perform integrity checks on the /boot partition and can rebuild the /boot partition in the case of a “destructive” upgrade operation invoked as a result of a new primary image that has been downloaded to the /dist partition. By using the /boot partition as a disk-based ROM, the expensive ROM component typically employed in currently available peripheral and consumer electronic devices can be eliminated. Detection and correction of file system corruption is automatic, as is installation and upgrade to a new embedded OS version. Because the /boot partition that serves as disk-based ROM is read-only, corruption of the primary image and OS kernel stored within the /boot partition is extremely improbable, providing the robustness and reliability currently provided by storing critical routines and data in a ROM component.

[0009] FIG. 1 is a block diagram of a typical computing environment managed by an embedded operating system within a peripheral or consumer electronic device.

[0010] FIG. 2 illustrates partitioning of the hard disk drive according to the described embodiment of the present invention.

[0011] FIG. 3 illustrates the power-on boot process.

[0012] FIG. 4 illustrates two important restoration operations that may be carried out by script routines executed by the initial OS kernel.

[0013] FIG. 5 illustrates the process for downloading a new version of the primary image and/or initial OS kernel and installing the new version.

[0014] One embodiment of the present invention employs a read-only /boot partition on a 4.1 gigabyte Integrated Device Electronics (“IDE”) disk drive as a disk-based ROM for reliably storing a backup primary image and initial OS kernel for an embedded Linux operating system that controls a printer peripheral device. In this embodiment, the printer peripheral device does not include a traditional ROM for storing the OS kernel, as in currently available and previous printer peripheral devices. Duplication and management of backup archives of the primary image provide, in this embodiment of the present invention, the reliability and robustness provided by ROM in currently available and previous peripheral devices. It is important to note that many different embedded operating systems, in addition to the embedded Linux operating system, can be recovered and upgraded with slight changes to the described embodiment.

[0015] FIG. 2 illustrates partitioning of the hard disk drive according to the described embodiment of the present invention. The hard disk drive is partitioned to include a root partition 202, the /boot partition 204, the /dist partition 206, a swap partition 208, a /var partition 210, and a /var/spool partition 212. The embedded operating system supports partitioning of the hard disk drive via the Linux sfdisk command according to the configuration table, stored in the file hda.conf, shown below as Table 1. It should be noted that partitions for an IDE disk drive are normally labeled by strings having the form “/dev/hda” concatenated with an integer, as can be seen in the first column of Table 1, below. Normally, file systems are mounted within partitions. Thus, a reference to the /boot partition is a reference, in the described embodiment, to the partition “/dev/hda1” which serves as the mount point for the file system “/boot.” Generally, /boot may be referred to, in the following, as a partition and may be alternately referred to as a file system, depending on the context of the reference. The same alternative referencing may be applied to the other partitions/file systems to be described below.

[0016] The first column of Table 1 lists the partition labels for each partition, the values in the second column indicate the size, in megabytes (“MB”), of the corresponding partitions, and the values in the third column indicate the partition types of the corresponding partitions, where “83” indicates a Linux native partition, “5” indicates an extended partition, and “82” indicates a Linux swap partition. The ‘r’ prefix on values in the second column indicate that the partition should expand to the “remaining” space on the disk but should at least be the indicated size in megabytes (“MB”).

[0017] The standard Linux file fstab contains a description of the partitioning and mount points for file systems for a hard disk drive. The contents of the file fstab, following partitioning of a hard disk drive by sfdisk according to the hda.config file shown in Table 1 and mounting of the various file systems employed in the described embodiment using the Linux mount command, is shown below as Table 2:

[0018] Note that the first row in Table 2, describing the partition “/proc,” describes a virtual file system that is used by various operating system routines to report information about the OS kernel but that, being a virtual file system, does not correspond to a physical hard disk partition. The first column of Table 2 references a block special device in which the file system indicated in the second column has been mounted. The third column describes the type of file system, where “proc” designates virtual proc file system, “ext2” designates a local file system, “swap” designates a file system used for virtual memory swapping by the Linux operating system. The fourth column of Table 2 lists attributes of the various file systems, including the following attributes: (1) “defaults,” an attribute standing for the attributes list “rw,suid,dev,exec,auto,nouser,async;” (2) “rw,” an attribute indicating read-write access; (3) “exec,” an attribute that indicates that execution of binary files from the file system is allowed; (4) “sync,” an attribute that specifies that only synchronous writes to files of the file system are permitted; (5) “noauto,” an attribute that indicates that the file system should not be automatically or implicitly mounted; (6) “auto,” an attribute that indicates that the file system may be mounted implicitly using the mount option “−a;” (7) “suid,” an attribute that indicates enablement set-user-identifier or set-group identifier bits associated with the file system; (8) “async,” an attribute that specifies that writes to files of the file system may be carried out asynchronously or, in other words, with intermediate RAM caching; (9) “grpquota,” an attribute that specifies the file system is regulated by size quotas as they relate to users or groups; and (10) “ro,” an attribute inidcating read-only access.

[0019] Referring again to FIG. 2, it can be seen that each of the file systems 202, 204, 206, 208, 210, and 212 corresponds to file systems described by entries in Table 2. In the Linux operating system, directories are specified in hierarchical fashion, starting with the root directory “/.” Thus, the top-level directories “/boot,” “/dist,” “/var,” and “/var/spool” along with the root directory “/,” each occupies a separate partition. The special file system “swap” and the virtual file system “proc” are not accessible through the file system hierarchy. The /dist partition 206 contains the files “primary.tar.gz” 216 and “primary.tar” 218. The file “primary.tar.gz” is a compressed archive file containing the files that represent the primary image for the operating system. The file “primary.tar” is an uncompressed primary image archive file. The /boot partition 204 contains a number of files including another copy of the file “primary.tar.gz” 220, the file “failsafe.gz” 222, and the file “vmlinuz” 224. The copy of “primary.tar.gz” in the /boot partition is the compressed factory-installed primary image archive file that can be used to restore the primary image in the event of data corruption. The file “failsafe.gz” is a compressed file containing an initial RAM-disk root file system that is loaded upon power up or reboot and that executes a number of script routines, to be described in detail below, that implement that robustness and reliability characteristics of the disk-based ROM of the present invention. The file “vmlinuz” 224 is a compressed Linux embedded operating system kernel.

[0020] FIG. 3 illustrates the power-on boot process. When the appliance that includes the described embodiment of the present invention (“print spooler”) is powered on, the contents of the file “failsafe.gz” 222 within the file system “/boot” is uncompressed into an initial RAM-disk file system 304 contained within RAM (104 in FIG. 1). This boot information is specified in the file lilo.conf shown below in Table 3 and processed by the Linux Loader (“LILO”).

[0021] This initial OS kernel executes a number of script routines, including the script routines “linuxrc” and “install,” to be described below in detail, that check the hard disk for corruption and, if corruption is present, restores the contents of the various file systems. Then, following completion of execution of the script routines “linuxrc” and “install,” the primary Linux embedded file system, a copy of which is contained in the file primary.tar.gz 220, is mounted, as directed by the lilo.conf statement “root=/dev/hda5,” after the RAM-disk file system is automatically unmounted. Following mounting of the primary Linux embedded file system, the routine ‘/sbin/init’ is executed, which then brings up the rest of the embedded Linux operating system. Note that the same kernel continues to execute for the embedded Linux operating system as for the initial OS, but that additional routines and processes that support additional functionality are included in the embedded Linux operating system.

[0022] FIG. 4 illustrates two important restoration operations that may be carried out by script routines executed by the initial OS kernel. If the initial OS kernel determines that the file “primary.tar” 218 in the file system “/dist” is damaged or corrupt, then the initial OS kernel can recreate this file by uncompressing the file “primary.tar.gz,” as indicated by arrow 402. However, if the file “primary.tar.gz” 216 has been corrupted, then the initial OS kernel can replace the corrupted file “primary.tar.gz” with a copy of the file “primary.tar.gz” on the file system “/boot,” as indicated in FIG. 4 by arrow 404.

[0023] FIG. 5 illustrates the process for downloading a new version of the primary image and installing the new version. While the print spooler is running, a new version of the primary image may be downloaded via a communications network and stored in the file system “/dist,” replacing the current copy of the file “primary.tar.gz,” as indicated by arrow 502 in FIG. 5. Following a reboot of the printer, and during execution of a script routine by the initial OS kernel, the presence of the new “primary.tar.gz” file is detected and the new “primary.tar.gz” file is uncompressed, as indicated by arrow 503, to create a new “primary.tar” archive file, which is then provided to the Linux command “tar” to selectively regenerate the “/boot”, “/”, “/var” “/var/spool” partitions as indicated by arrows 504, 505, 506, and 507 in FIG. 5.

[0024] By maintaining the file system “/boot” as read-only, except for very short periods during a destructive installation of a new version of the primary image and/or initial OS kernel, as shown in FIG. 5, the files maintained in the “/boot” partition are essentially incorruptible and serve under the present invention in the same capacity as data instructions stored in ROM in currently-available and prior embedded systems. By automatically detecting file system corruption and newly downloaded OS versions during the booting process, by script routines to be described below, and repairing detected corruption and/or installing newly downloaded OS versions, the described embodiment provides reliability and robustness, previously effected using ROM-based data and instruction storage and complex software mechanisms as well as automatic upgrades, both without requiring user intervention in most cases.

[0025] Table 4, below, contains an alternative description of the partitions and file systems described above with reference to Tables 1 and 2 and FIGS. 2-5.

[0026] The /boot partition is the failsafe boot partition that includes a compressed archive of the factory-installed primary image. The /dist partition is the point to which upgrades are downloaded in order to install new or alternate versions of the primary image. The swap partition is disk space used for virtual memory swapping. The root partition “/” is the primary root partition for the system. The /var partition includes files that contain configuration data and other variable data, and the partition “/var/spool” comprises the bulk of high disk space and is used for printing tasks and applications, including spooling, logging, and other such information.

[0027] The initial OS kernel invokes three script routines, described below, to check the above-described file systems for corruption, repair any detected corruptions, and to install any new or alternate versions of the primary image and/or initial OS kernel. These script routines include the script routines “linuxrc,” “install,” and “rc.sysinit,” which are discussed in order below. Each script routine, including linuxrc, begins with the number of default definitions comprising string assignments to local variables. The defaults for the routine “linuxrc” are provided below:

[0028] Next, script routines may include a number of functions used by the main function of the script routines. The functions included in the script routine “linuxrc” follow below:

[0029] The function “debug,” comprising lines 44-51, prints a message to a front panel LCD device of the printer and starts a new instance of the Linux shell command interpreter. The function “debug” allows a developer to continue to interact with the printer. The function “debug” may be disabled, not included in, or changed in production versions of linuxrc as well as in the routine “install,” to be discussed below. The function “error,” comprising lines 54-71, prints an error message to an LCD device on the front panel. The functions “progress,” comprising lines 74-89, “message,” comprising lines 92-109, and “question,” comprising lines 112-126, print an activity message to an LCD device on the front panel. The function “powerCheck,” comprising lines 129-140, checks to see if the power button of the printer has been pressed and, if so, shuts down the print spooler after printing out a shutting down message, explicitly unmounting the boot and disk partitions, and unmounting all other partitions using the Linux command “umount.” The function “version_string,” comprising lines 143-188, parses a boot file to extract information from the boot file that is concatenated together to form a boot version label file that includes major and minor version numbers and a build label.

[0030] The main body of the script routine “linuxrc” is provided below:

[0031] On lines 193-194, linuxrc sets the values of several local variables to 0 (the Boolean value FALSE, by script routine convention). On lines 197-198, linuxrc sets up an environment for execution of the script routine by exporting the path variable, set above in the default section on line 2, via the Linux command export, and setting the file creation mask to the octal value “000” via the Linux command “umask.” On lines 201-209, linuxrc mounts all the file systems described above with reference to FIGS. 2-5 and Tables 1-4. On lines 212-214, linuxrc installs a driver module that controls the LCD front panel of the print spooler. On lines 217-219, linuxrc installs a soft power driver module and then, on line 222, clears the power button by reading it once via the soft-power-driver-module command “checkpwr.” On line 225, linuxrc calls the function “powerCheck,” described above, and then sets the name of the host system on lines 226-227. Note that linuxrc calls the function “powerCheck” periodically throughout the remaining lines in order to detect cycling of the power button by a user in order to shut down the print spooler. On lines 231-241, linuxrc queries a user, via the front panel, whether or not the user wishes to restore factory default settings and, if so, sets the variable “installFactory,” on line 238, to 1 (Boolean value TRUE), after which the function “progress” is called to report installation of factory defaults. Note that the function “progress” is called at various points by linuxrc to inform the user, via the front panel LCD, of the activities currently undertaken by linuxrc. On lines 245-250, linuxrc calls the Linux command “e2fsck” to check for integrity of the file system “/boot,” and, if possible, repair any defects. If the system command “e2fsck” returns a return value greater than 1, then an unrecoverable error has been detected by e2fsck and linuxrc then calls the functions “error” and “debug” to report the corruption and to provide a new shell command environment to allow a developer to continue to interact with the print spooler via the root console. If, on the other hand, the file system “/boot” is uncorrupted, then, on line 251, linuxrc mounts the file system “/boot” in read-only mode via the Linux command “mount.” Similarly, on line 254, linuxrc checks the file system “/dist” for corruption. If the system call “e2fsck” detects an unrecoverable error, then, on line 258, linuxrc calls the Linux routine “mke2fs” to reformat the disk partition /dist and sets the local variable “installFactory” to 1 so that the backup factory image stored in the /boot partition will be copied to the /dist partition. On line 261, linuxrc mounts the /dist partition and, on line 263, linuxrc displays the /boot partition primary image version via the function “version_string.” On line 267, linuxrc tests to see whether the primary image file is present in the /dist partition and, if not, sets the local variable “installFactory” to 1 on line 268. Otherwise, on line 270, linuxrc attempts to decompress the primary image file in the /dist partition and, if decompression fails, sets the local variable “installFactory” to 1 on line 272. If the local variable “installFactory” has been set to 1 in the preceding lines, then, on lines 279-283, linuxrc remounts the /dist partition to allow for writing, copies the compressed primary image from the /boot partition to the /dist partition, removes the uncompressed primary image archive from the /dist partition, and then remounts the /dist partition for read-only access. On lines 287-295, linuxrc checks to see if the uncompressed primary image archive file is present in the /dist partition and if the uncompressed primary image archive file can be successfully used by the system routine “tar” to reconstruct a file hierarchy. If the uncompressed primary image archive file is either not present or cannot be manipulated by the routine “tar,” then the local variable “uncompress” is set to 1 (Boolean value TRUE). If the local variable “uncompress” has been set to 1, then, on lines 300-303, linuxrc remounts the /dist partition to allow for writing, uncompresses the compressed primary image file to create the uncompressed primary image archive file, and then remounts the /dist partition for read-only access. On lines 308-314, linuxrc extracts the install script, to be described below, from the uncompressed primary image archive file in the /dist partition into the directory “/sbin” via the Linux command “tar” and changes the access mode of the extracted install script to allow it to be executed via the command “chmod.” On line 315, linuxrc invokes the script “install”, passing to the script “install” the local variables “installFactory” and “uncompress.” If the script “install” successfully runs, then the file systems are checked and verified, and reconstructed if necessary. Then, linuxrc removes the LCD driver and soft power drivers on lines 321-322 and 325-326, and then explicitly unmounts the /boot and /dist file systems on lines 330 and 331 and unmounts the remaining file systems on line 333. When linuxrc finishes execution, as discussed above, the primary Linux embedded file system is mounted and the full embedded Linux operating system is brought up, starting with execution of “/sbin/init.” However, if the script “install” fails, then, in certain cases, control returns to linuxrc on line 316, in which case linuxrc invokes the function “debug” to start a new shell environment after printing out an error message.

[0032] The script routine “install,” called by the previously-described script routine “linuxrc,” is provided below. First, as with linuxrc, a list of default definitions is provided:

[0033] Next, install includes a set of functions that are called from the main function of install. These functions are provided below:

[0034] The function “debug,” comprising lines 47-54, echoes a message from the front panel of the printer and then invokes a new shell environment to allow continued interaction of the developer with the print spooler in the face of a unrecoverable error. The function “compatible_version,” comprising lines 59-85, compares string representations of a boot version label, to determine whether the version string, supplied as the second argument, is less than or greater than the version strings supplied as the first argument, depending on a comparator operator supplied as the third argument. The function “validate_version,” comprising lines 88-178, compares the current major/minor/build version label of the current boot version to a new major/minor/build boot label and to a comma-separated validity expression to determine whether or not the new boot version is greater than the current boot version and the current boot version is less than a boot label contained in the validity expression or whether the current boot version falls within a range of boot versions in the validity expression, in which case the function validate_version returns the value “0.” Otherwise, the function “validate_version” returns the value “−1.” On lines 93-96, the function “validate_version” calls a shell script in the file “/boot/product” to set the local variable “bootVer” to the major/minor/build boot version label of the current OS. Then, in the while-loop, comprising lines 98-124, the function “validate_version” selects each of the comma-separated terms from the validity expression supplied as a fourth argument. When the selected term is a hyphen-separated version range, the function “validate_version” compares the current boot version to the range, and when the term is a single boot version label, the function “validate_version” compares the current boot version with the new boot version, supplied as arguments 1-3, and to the boot version term selected from the validity expression. The function “error,” comprising lines 131-148, prints an error message to the LCD of the front panel of the printer. The function “message,” comprising lines 151-168, prints an activity message to the LCD of the front panel of the printer. The function “need_destruct,” comprising lines 171-181, extracts the shell script “/boot/product” from the uncompressed primary image archive file in the /dist partition, calls the extracted shell script to determine the boot version label of the primary image, and then calls the function “validate_version” to determine whether the boot version in the uncompressed primary image archive file indicates that a destructive install is required or, in other words, whether the /boot partition needs to be rebuilt due to reception by the print spooler device of a new primary image representing a more recent version. The function “need_partition,” comprising lines 184-195, extracts the file hda.conf from the uncompressed primary image archive file in the /dist partition and then compares the extracted file hda.conf to the current hda.conf for the hard disk to determine whether the two files are different, indicating that the hard disk needs to be repartitioned for installing a new primary image. If so, any newly specified partitions are created via a call to the function “parttool” on line 190. The function “progress,” comprising lines 198-211, prints an activity message to the front panel LCD. The function “powerCheck,” comprising lines 218-232, is similar to the function “powerCheck” included in the previously-described script routine “linuxrc.”

[0035] The main body of the install script routine is provided below:

[0036] Note that, as with the previously-described script routine “linuxrc,” the script routine “install” periodically calls the function “powerCheck” to detect power shutoffs. Also, the script routine “install” periodically prints out progress messages to the front panel LCD. On line 244, install calls the Linux command “mkswap” to set up the swapping partition for virtual memory paging. On lines 249-271, install determines whether a destructive upgrade has been indicated explicitly, via arguments passed into instal, or whether a newly downloaded primary image has a more recent version number than the current primary image, indicating the need for a destructive upgrade and, in either case, proceeds with a destructive upgrade on lines 254-269. First, the /boot partition is remounted to allow for writes on line 255. Then a directory change is made to the root directory on line 256 and the uncompressed primary image archive file is processed using the Linux command “tar,” on line 257, to extract the primary image from the uncompressed primary image archive and place the appropriate files and directories within the file systems of the hard disk. Then, on line 258, a compressed primary image file is copied from the /dist partition to the /boot partition and, on line 259, a new boot loader is installed into the system via the Linux command lilo. On line 260, the /boot partition is remounted for read-only access. Thus, lines 256-259 represent the only time that the /boot partition is exposed to write access. As discussed above, at all other times the /boot partition has read-only access and therefore is protected from corruption. On line 261, the /dist partition is remounted for write access, and the uncompressed primary image archive file in the /dist partition is removed on line 262, followed by remounting of the /dist partition for read-only access. Finally, the system is rebooted using the command “reboot,” on line 269, which will end processing of install and eventually invoke the script routine “linuxrc,” previously described above. On lines 275-278, install determines whether the disk drive needs to be repartitioned and, if so, effects the repartitioning. On lines 282-288, install calls the Linux command e2fsck to verify integrity of the / partition and, if an unrecoverable corruption is detected, reformats the / partition via the Linux command mke2fs, on line 287, additionally setting the local variable “installRoot” to 1 on line 288, to indicate that the files for the root partition need to be reinstalled. Similarly, install checks the /var and /var/spool partitions for corruption and reformats those partitions upon detecting unrecoverable corruption on lines 295-303. On lines 316-329, install additionally checks to make sure that “install” files are located within each of the verified partitions and, if not, sets appropriate local variables to indicate that the files of the partitions must be reinstalled. Finally, in three blocks of code comprising lines 333-342, 346-360, and 364-389, install reinstalls the files of the /, /var, and /var/spool partitions by extracting the appropriate files from the uncompressed primary image archive file on the /dist partition.

[0037] The script “rc.sysinit” is called upon booting of the primary image or, in other words, during the transition from the initial OS kernel to the full Linux embedded operating system that controls the print spooler except during the initial booting and configuration of the print spooler. The script “rc.sysinit” is provided below:

[0038] The above script “rc.sysinit” is provide for completeness of description of the described embodiment of the present invention, but will not be described to the same level of detail as the scripts “linuxrc” and “install” are described above, since the script “rc.sysinit” does not manipulate the primary image files in the /boot and /dist partitions that, along with the script routines “linuxrc” and “install,” embody a large portion of the disk-based ROM feature of the present invention. The script routine “rc.sysinit” sets the path variable on line 15 and mounts all the file systems on lines 20-34. Then, rc.sysinit installs the LCD driver module on lines 40-43, starts up virtual memory swapping on line 51, loads a networking module on line 54, creates a host name, if necessary, on lines 58-59, and, on lines 63-68, reads in a configuration data file or sets local variables “NETWORKING” and “HOSTNAME” indicating that the configuration file could not be found and that networking cannot therefore be enabled. Then, in the remaining code, rc.sysinit enables a disk quota mechanism for the local file systems, performs maintenance on certain local files, sets the system clock, and sets up networking and initializes serial ports.

[0039] Although the present invention has been described in terms of a particular embodiment, it is not intended that the invention be limited to this embodiment. Modifications within the spirit of the invention will be apparent to those skilled in the art. For example, the present invention may be implemented in many different computer languages with any number of different modular, functional, and data structure organizations. The particular partitionings of hard disk drives included in electronic devices may vary, the types of hard disk drives included in the electronic devices may vary, and the present invention may be practiced in a wide variety of peripheral and consumer electronic devices that use embedded operating systems for control. The primary images may be stored in a single archival file, multiple archival files, single compressed archival files, or multiple compressed archival files. Various types of access privilege assignment strategies may be applied to guarantee that the ROM-disk feature of the present invention provides robust and reliable storage of the backup primary image and other critical files. In the described embodiment, the Linux operating system served as the embedded operating system, but the techniques employed in the described embodiment may also be used to support failsafe recovery and upgrade of other embedded operating systems.

[0040] The foregoing description, for purposes of explanation, used specific nomenclature to provide a thorough understanding of the invention. However, it will be apparent to one skilled in the art that the specific details are not required in order to practice the invention. The foregoing descriptions of specific embodiments of the present invention are presented for purpose of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed. Obviously, many modifications and variations are possible in view of the above teachings. The embodiments are shown and described in order to best explain the principles of the invention and its practical applications, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the following claims and their equivalents: