WO2009076405A1

WO2009076405A1 - Integration of secure data transfer applications for generic io devices

Info

Publication number: WO2009076405A1
Application number: PCT/US2008/086168
Authority: WO
Inventors: Xian Jun Liu; Robert C. Chang; Po Yuan; Junzhi Wang; Ron Barzilai; Bahman Qawami; Farshid Sabet-Sharghi
Original assignee: Sandisk Corporation
Priority date: 2007-12-13
Filing date: 2008-12-10
Publication date: 2009-06-18
Also published as: US9032154B2; US20090157971A1; EP2245527B1; KR101538619B1; EP2245527A1; TWI515558B; TW200935218A; KR20100113059A

Abstract

Techniques are presented for sending an application instruction from a hosting digital appliance to a portable medium, where the instruction is structured as one or more units whose size is a first size, or number of bytes. After flushing the contents of a cache, the instruction is written to the cache, where the cache is structured as logical blocks having a size that is a second size that is larger (in terms of number of bytes) than the first size. In writing the instruction (having a command part and, possibly, a data part), the start of the instruction is aligned with one of the logical block boundaries in the cache and the instruction is padded out with dummy data so that it fills an integral number of the cache blocks. When a response from a portable device to an instruction is received at a hosting digital appliance, the cache is similarly flushed prior to receiving the response. The response is then stored to align with a logical block boundary of the cache.

Description

INTEGRATION OF SECURE DATA TRANSFER APPLICATIONS FOR GENERIC IO DEVICES

FIELD OF THE INVENTION

[0001] This invention pertains generally to the field of using portable media, such as a secure memory card, with digital appliances, such as a telephone handset, and more particularly to techniques for transferring instructions between a digital appliance and a portable media in communication therewith.

BACKGROUND

[0002] Digital appliances, such as notebook or laptop computers, personal digital assistants (PDAs), or mobile phones, are commonly used with portable media, such as memory cards or USB drives. In some cases, secure data is stored on these portable media. The application responsible for securely transferring this data between the hosting appliance and the media often uses instructions that have a specific structure. As part of the transfer process, the host may cache the instruction; but as the host's operating system may unaware of the instruction's structure, and because the host's cache structure may not be compatible with the instruction's structure, the instruction, along with any associated data, may not be successfully transferred.

SUMMARY OF THE INVENTION

[0003] According to a one aspect, techniques are presented for sending an application instruction from a hosting digital appliance to a portable medium, where the instruction is structured as one or more units whose size is a first size, or number of bytes. After flushing the contents of a cache, the instruction is written to the cache, where the cache is structured as logical blocks having a size that is a second size that is larger (in terms of number of bytes) than the first size. In writing the instruction (having a command part and, possibly, a data part), the start of the instruction is aligned with one of the logical block boundaries in the cache and the instruction is padded out with dummy data so that it fills an integral number of the cache blocks. [0004] According to another aspect, when a response from a portable device to an instruction is received at a hosting digital appliance, the cache is similarly flushed prior to receiving the response. The response is then stored to align with a logical block boundary of the cache.

[0005] Various aspects, advantages, features and embodiments of the present invention are included in the following description of exemplary examples thereof, which description should be taken in conjunction with the accompanying drawings. All patents, patent applications, articles, other publications, documents and things referenced herein are hereby incorporated herein by this reference in their entirety for all purposes. To the extent of any inconsistency or conflict in the definition or use of terms between any of the incorporated publications, documents or things and the present application, those of the present application shall prevail.

BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The various aspects and features of the present invention may be better understood by examining the following figures, in which:

Figure 1 is a flowchart of a common sequence for a write process with caching;

Figure 2 is an application system diagram;

Figure 3 shows an example of a command structure;

Figure 4 is a table showing an example of a command structure;

Figure 5 illustrates the caching of an instruction, such as is shown in Figure 3, where the instruction does not like up with a cache boundary;

Figure 6 illustrates the caching of an instruction, such as is shown in Figure 3, where the instruction is aligned with a cache boundary and padded out;

Figure 7 illustrates a method of placing an instruction into cache; and

Figure 8 illustrates a method of placing a response into the host's cache.

DETAILED DESCRIPTION

Introduction

[0007] When transferring content between a portable medium, such as a memory card, and a hosting digital appliance, such as mobile telephone handset, a host's operation system will often use a caching mechanism for the content. This cache will typically have a structure where it is logically broken up into a number of blocks of a particular size. The content being transferred, whether data or an instruction (a command and maybe some associated data), may also be structured in a specific way as one or more sections of a certain size. When the content is structure in units of one size (for example, as 512 Byte sectors) and the cache is structured in units of another size (for example, in 2KB blocks), the caching mechanism may clash with the content's structural requirements.

[0008] This is particularly true with dealing with secure content and secure portable storage media, such as that marketed under the TrustedFlash name. In such cases, the more basic levels of a host's operating system that handle transfers between the host and the portable medium will see the instruction as just so much data, being unaware of any embedded commands in the instruction, and store it as such in cache. Examples of such instructions and their use are described in US patent number 7,237,046, US patent application number 10/899,260, and, particularly, US patent applications numbers 11/298,349 and 11/299,186. Under such arrangements, the host's operating system may treat the instruction as data to be written to a particular address on the medium; but, when received on the medium, this instruction will be interpreted as, for example, a write command to store data at different address on the medium. If the host then tries to issue a read command for this data, the address as seen by the host's operating system will not correspond to the actual address as interpreted on the medium; and as the host may then rely what is thinks is cached copy of the data, it will incorrectly use this cached data.

[0009] The techniques described here are widely applicable, both for portable media, such as memory cards, that are primarily used with a hosting digital appliance, and also digital appliances which themselves include non-volatile memory, such as embedded flash. Example of such digital appliances include desktop, notebook or laptop tablet computers, personal digital assistants (PDAs), mobile phone handsets, a variety of media players such as MP3 players, DVD players, gaming consoles, digital recording devices such as digital cameras, and others. Suitable portable media are available from SanDisk Corporation, the assignee hereof, which include those sold under trademarks TrustedFlash, CompactFlash (CF), Multi-Media Card (MMC), Secure Digital (SD), MicroSD and Memory Stick PRO. These memory cards are removably connectable with digital appliances through mating connectors that are different for most cards. Flash drives, sold as those sold under the Cruzer trademark, contain a plug according to the Universal Serial Bus (USB) standard, so can be plugged directly into any digital appliance having a USB receptacle. TrustedFlash cards and other secure cards (SIM, SMMC, SD, etc) allow secure storage of the both licenses and associated content. Also the following discussion will often refer to nonvolatile memory as flash, the techniques given here are not so limited and more generally apply to other non-volatile memory structures (such as described in US patent application 10/841,379, for example), for both portable devices and embedded memories. Although the teachings given here are not so limited, when reference is made to a particular embodiment for purposes of discussion, the following will often refer to the concrete example of a mobile telephone handset used with a memory card, such as that of the TrustedFlash variety, having instructions whose structure is based on a sector.

[0010] To circumvent such possible difficulties with the host caching, firmware associated with the secure transfer can be changed to be aware of the running time command size; the associated application toolkit needs to be integrated with the applications; instructions sent from the application to the platform operation system (OS), including write and read commands, start with a cache block size aligned sector; and that the write, read buffer size should be cache block size aligned.

[0011] Technically, to allow fast spread out of secure content media to users of digital appliances, integration for these platforms with the corresponding secure content toolkit is important. Due to the special requirements of most secure content technology and code size consideration, the corresponding application toolkit usually needs to be integrated into the operating system (OS) of a handset or other hosting device, so that the applications can call the application programming interfaces (APIs) provided by the toolkit to access the media. However, for certain standardized or open platforms, say Linux and WIPI, this solution can be complicated to implement since it is not easy to have the needed platform modifications to support secure content media. [0012] To overcome this limitation, every platform will provide a way for applications to access generic IO device; in case of a TrustedFlash card, this is the SD IO device. With this assumption, the applications can use the provided IO access to communicate with the secure content media card by integrating the corresponding toolkit into the applications. This can be used for any platforms, especially being good for those unmodified platforms, as long as the IO device is accessible for applications. The major difficulty of this solution is to deal with the caching mechanism, which is used most likely for every platform. The following section describes the nature of this caching mechanism, and the conflicts between its buffered IO and the earlier versions of secure content transfer firmware.

Problem of Buffered IO with Current Secure Transfer Firmware [0013] As noted, many command and data transfer processes rely upon a particular structure for instructions. For example, current TrustedFlash firmware is defined to communicate with the handset or other host through sector (512 Bytes) based instructions. The number of sectors for each instruction is predefined. This can cause a major IO problem for buffered IO only platforms. Due to the hosting platform's caching mechanism, the instructions sent to TrustedFlash firmware on the media may not be sector based, so that the TrustedFlash communication channel is broken. For example, a TrustedFlash CREATE ACR command is a one-sector write instruction. When an application sends a one-sector write to the platform OS, this write command may be changed to a write command with multiple sectors, and the starting write location may be changed as well, after this command reaches the secure transfer firmware on the media.

[0014] A common scenario for write command is the following:

. Application asks to write to the device n sectors on location m^th sector. . The OS will check the cache for related memory location; if it does not exist, the OS will read data from a location on the card, including the required location, and then stores it on the cache. . The OS will update the related sector.

. The OS may or may not flush the data back to the card right away. Figure 1 is a flow chart to describe this common behavior for write command as seen by the host. It should again be noted that although the application may see the instruction as a command, possibly with data attached, to the operating system, it is just interpreted as data.

[0015] Figure 1 starts at 101 with the application asking to write to the device. The cache is checked at 103 to determine whether a related location already exists in the cache; if so (the "Y" path), the data is written to the related location of the cache at 105. If a related cache location is not found (the "N" path out of 103), it is then determined whether the n (the number of sectors to write) is cache-block-size aligned, and m (the location for the write) is cache-block-size aligned at 107, and, if so, the process again goes to 105. The NO path out of 107 leads to 109, where the starting write location or buffer size buffer is adjusted as needed. For example, take the case of where the size of the cache blocks is 2KB and the hosting application sends the instruction data with a total of n=5 sectors and which starts on the seventh sector (m=7). The host will change the starting on the fifth and the total sectors of data as 8. 109 then goes on to 111, where the related data aligned with the cache block size is read from the device, before going to 105. Finally, the OS may flush the data back to the card right if forced. As a result of this caching mechanism, the TrustedFlash firmware may not be able to understand whatever command sent from the applications.

[0016] In practice, some platforms support direct IO, while others only provide buffered IO . In addition, the caching algorithms may be different from platform to platform. Therefore, a generic solution is quite preferable for integration with secure transfer application's toolkit. The following discussion present such an integrated solution, based on the example for TrustedFlash, that works whether the host platform uses direct IO or buffered IO . This is presented below as including four components: the structure of system integration; the modification of firmware; and the methods of solving caching for write and for read.

Integrating Secure Transfer Toolkit into Applications

[0017] Since a platform remains the same, for the applications to access the secure content card, the corresponding toolkit will need to integrate into the host's applications so that they can call the toolkit APIs to send, say, TrustedFlash commands to the TrustedFlash cards. Briefly, the toolkit provides full package of APIs for applications to access the secure media. This package offers many functions related with secure system, such as system operation, account operation, storage operation, and digital rights management (DRM) operation.

[0018] The applications will directly call the toolkit's APIs, and then the toolkit will talk to the corresponding secure medium by using the device operation methods provided by the platform OS. For example, in Linux, this methods are open(), read(), write(), and close(). According to each individual platform, this glue layer between toolkit and the platform is implemented in the Platform I/O Device Operation.

[0019] Figure 2 gives a block diagram of the structure. On the host side, the layer shown include the various customer applications 201 on the top, under which is the platform's operating system and file systems 203, and then the platform's block device driver 205 which is used to communicate with the portable medium 250. The toolkit 207 will then be between the costumer application layer 201 and the operation system and file systems layers. As shown in Figure 2, among the elements that can be included in the toolkit are digital rights management (DRM), data protection, customer APIs, storage, account, system, and utility blocks, as well the secure manager, secure file system, protocol driver, and platform I/O device operation sublayers.

The Modification of Secure Firmware

[0020] As discussed above, the instructions used in secure data protocols are often based on units of a particular number of bits. For example, the TrustedFlash instructions for communicating between a host and the TrustedFlash firmware on media are sector-based. The number of sectors for each instruction is predefined during the firmware loading. Figure 3 and table of Figure 4 show the TrustedFlash command structure as found in the TrustedFlash specification.

[0021] As shown in Figure 3, and as is developed further in US patent applications numbers 11/298,349 and 11/299,186, the exemplary instruction is composed of a first sector (512 Bytes) that holds the command. After this command block follows any associated sectors of data, again structured as sectors. For example, a write instruction for N sectors of data would consist of N+l sectors, a first for the write command followed by the data sectors. For commands without data, such as a status request, would just be the one sector of the command block. Any response, such as the data of a read command, would be treated by the read process of the last section below. The table of Figure 4 illustrates some details for the exemplary command block structure in the TrustedFlash protocol.

[0022] Figure 5 schematically illustrates the sort of misalignment that can occur under the prior art arrangement when an instruction based on units of a first size is written into a cache based on logical blocks of a different size. For this example, the instruction 501 is take to be made up of 5 sector-sized units, where the first sector 501a is the command portion and there are four sectors 501b of associated data. The cache 503 is structure in terms of 2KB (or, more accurately 2048Byte) blocks. (It will typically be the case that the size of the blocks used in the cache will be some multiple of a standard sector.) Without the sort of arrangements discussed below, the instruction could be cached at a arbitrary location in the cache as shown in Figure 5, with the resultant problems described above.

[0023] In order to make the media firmware understood the commands from platforms with buffered IO, the current firmware preferably knows the dynamical command size of each commands. To do so, one field can be introduced to add into the command block to specify the total size of current sending command. For example, in the table of Figure 4, a new field for the dummy sector count can be added into the second portion of the command sector beginning at byte 36. The predefined command lengths will be treated as the default values, and this field will tell the actual size of this command in run time. If this field is 0, then the default values will be used.

[0024] When a platform OS is using buffered IO, the behaviors for both read and write may be changed as described below. The following two sections describe the way to deal with the caching on write, when an instruction is sent to the medium, and read, when a response is returned. Briefly, this process is illustrated in Figure 6, which is similarly numbered to Figure 5, except as shown in Figure 6, the instruction is now aligned with a logical block boundary in the cache and any partially filled cache blocks are then padded out as shown by the hatched area 607. Method of Direct Write Command for Buffered IO

[0025] The secure transfer firmware is the agent inside the media (e.g., TrustedFlash) card to handle the commands. The secure transfer commands from the host are preferably sent to the card immediately. Therefore, if the platform OS does not send a write command to the card, but puts the data into the buffer instead, the medium will have no way to get the command and respond to it. In addition, for instructions structured with the command block required to be in a specific locations, as with the TrustedFlash command block that must be the first sector of the whole command, the sequence of the write command must follow the appropriate (e.g. TrustedFlash) protocol; any extra reads or writes in the command sequence before command completion, including get command status, will cause problems to firmware, so that the command will not succeed. Below is the method to solve these problems one by one.

[0026] The method of forcing the OS to flush write command is the following:

• Close the device

• Reopen the device

Some platforms may provide a function of flushing for write, in which case the corresponding command can be used. In others that lack such a command, or where flushing may not be working, the medium can be logically closed and reopened. To ensure the write command is flushed into the card, close-reopen always works.

[0027] Generally, when a card is inserted to the host, the host will initialize and mount the media/card. To access the contents of the media, it will open the media. It is considered as "logical" open. Related data, table, cache, and so on are established. After that, you can browse the directories, or open "file" for read/write operation. By "closing" the media, all cache contents, tables or data are cleared or invalidated. For example, the Linux operating system includes an open() command and a close() command. The close() command closes a file descriptor of the media so that it no longer refers to any file and may be used. Any record locks held on the file descriptor it was associated with, and owned with the process are removed. To logically reopen the device, a call to open() will create a new open file description. (These are described further in the Linux programmers manual.) [0028] The other problem is the potential change of write location, write buffer size, and command sequence. To avoid these, the toolkit will send to the medium a command in the following way:

• Be aware of the cache block size of particular platform

• Set the starting sector of write command aligned with the cache block size

• Make the buffer aligned with the cache block size, giving the correct size of the sending command.

This way, the OS does not need to adjust this write command since everything is cache-block-size aligned, and it does not need to read data form the card since the buffer size is block size aligned. As a result, the write command from the host will not be modified. If a platform supports direct IO, this method can still apply by simply setting the block size as 512 bytes.

[0029] The flow chart of Figure 7 describes this method of write command. The flow begins at 701 with the original status of the medium being open. The cache is then flushed at 703 by logically closing the device and then reopening it. Alternately, if the operating system has a specific command to flush the cache, this can be used instead.

[0030] The instruction is then placed in the flushed cache at block 705. As shown in Figure 6, the instruction 601 (including command block 601a and, in this example, four sectors of data) is stored so that it is aligned with a logical boundary in the cache 603. If this leaves any cache blocks only partially filled, these are then padded out to maintain the instruction as an integral number of logical blocks in the cache. In this example, as each logical block in the cache can hold 4 sectors, and the instruction has 5 sectors, when the instruction is contiguously cached, the last sector of the instruction only occupies part of a cache block. The hatching at 607 represents the added padded so that the host's operating system thinks the instruction is formed of complete cache blocks and treats it as such.

[0031] Returning to Figure 7, as shown in block 705, this process can include the command block being filled with the total size of the actual instruction or command and reading back the starting sector with the cache's logical block size aligned. The data, including any dummy sectors to fill out a complete cache block, are then filled. The starting write sector can then be checked to confirm the block alignment of the instruction. At block 707, the write command to send this instruction to the media can then be sent to the platform's operating system, where the write process will now start with the logical block address (LBA) that aligns with the cache's block boundaries and size.

Methods to Deal with Caching for IO Read

[0032] After receiving an instruction form the host, the medium may need to respond to the host, such as with a status or with data in response to a read request. Because of caching, whenever the application send a read command, the OS may just go to the cache to read it back if the related location data is in the cache. Therefore, the actual firmware response is not returned back to the application.

[0033] To deal with this problem, cache bypass can be used in the following way:

• The read relative staring logical block address (LBA) for each TrustedFalsh command changes for every read.

• The difference of starting read LBA between two read commands should be bigger that the cache block size.

• The starting read LBA should be aligned with cache block size.

• Close and reopen the device to make sure the OS flushing the read for application.

If a platform supports direct IO, this method can still apply by simply setting the cache block size as 512 bytes. The flow chart of Figure 8 describes this method of read command.

[0034] Figure 8 shows an exemplary embodiment for allowing the medium to respond to an instruction from the host. Typically, this read process will immediately follow the write process described in the last section, with no other commands being inserted between the write instruction and the read command. The flow again begins with the original device status being open, block 801.

[0035] At 803, the cache is flushed as at 703, with a close and then open or, if available, with a command to flush the cache. As the cache is now emptied, this prevents the operating system from just returning cached material instead of the actual response from the medium. At 805, the alignment is checked to see that when the received response is placed in the cache, it will be aligned with the logical block structure of the cache. Finally, the read command is sent to the platform's operating system.

[0036] Once these preparations have been made, the driver can then send the read instruction to the medium, which will then return the response. As noted, this may include actual secure user data (that, data as seen be the secure transfer system) or a non-data response, such as a status. The response is then cached aligned with the logical block boundaries of the cache. Padding, which secure transfer application would then ignore, may also be added to fill out an integral number of cache blocks.

[0037] Although the invention has been described with reference to particular embodiments, the description is only an example of the invention's application and should not be taken as a limitation. Consequently, various adaptations and combinations of features of the embodiments disclosed are within the scope of the invention as encompassed by the following claims.

Claims

IT IS CLAIMED:

1. A method of sending an application instruction from a hosting digital appliance to a portable medium in communication therewith, wherein the instruction is structured as one or more units whose size is a first number of bytes, the method comprising: flushing the contents of a cache on the digital appliance; and subsequently writing the instruction to the cache, where the cache is structured as a plurality of logical blocks whose size is a second number of bytes, the second number being greater than the first number, the writing including: aligning the start of the instruction with a boundary of the logical blocks; and padding the instruction with dummy data to have a size which is an integer multiple of the second number of bytes.

2. The method of claim 1, wherein the second number is a multiple of the first number.

3. The method of claim 1, further comprising: transferring the instruction from the cache to the portable medium.

4. The method of claim 1 , wherein the size of said units is a sector.

5. The method of claim 1, wherein the instruction includes a command portion and data portion.

6. The method of claim 5, wherein the portable medium includes a nonvolatile memory and the command is a write command for storing the data portion in the non-volatile memory.

7. The method of claim 5, wherein the command portion is one of said units in size and the data portion is one or more of said units in size.

8. The method of claim 1, wherein the instruction does not include a data portion.

9. The method of claim 8, wherein the instruction comprises a read data command.

10. The method of claim 9, further comprising: in response to the instruction, returning data from the a portable medium to the hosting digital appliance.

11. The method of claim 10, further comprising: storing the returned data in the cache, including aligning the start of the data with a boundary of the logical blocks.

12. The method of claim 8, further comprising: storing a response to the instruction in the cache, wherein the response is structured as one or more units whose size is the first number of bytes and the storing includes aligning the start of the response with a boundary of the logical blocks.

13. The method of claim 1, wherein flushing the contents of the cache comprises: closing the portable medium; and subsequently opening the portable medium.

14. A method of receiving from a portable medium an application response at a hosting digital appliance in communication therewith, wherein the response is structured as one or more units whose size is a first number of bytes, the method comprising: issuing an instruction from the digital appliance to the portable medium; flushing the contents of a cache on the digital appliance, where the cache is structured as a plurality of logical blocks whose size is a second number of bytes, the second number being greater than the first number; subsequently receiving a response to the instruction on the hosting digital appliance; and storing the response in the cache, the storing including: aligning the start of the response with a boundary of the logical blocks.

15. The method of claim 14, , wherein the second number is a multiple of the first number.

16. The method of claim 14, wherein the storing further includes: padding the response with dummy data to have a size which is an integer multiple of the second number of bytes.

17. The method of claim 14, wherein the instruction includes a read command and the response includes data.

18. The method of claim 14, wherein the response is a status.

19. The method of claim 14, wherein flushing the contents of the cache comprises: closing the portable medium; and subsequently opening the portable medium.

20. The method of claim 14, wherein the size of said units is a sector.